DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices
@ 2016-06-06 22:08 Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 02/39] bnxt: add HWRM init code Stephen Hurd
                   ` (40 more replies)
  0 siblings, 41 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the initial skeleton for bnxt driver along with the
nic guide to tie into the build system.
At this point, the driver simply fails init.

v4:
Fix a warning that the document isn't included in any toctree
Also remove a PCI ID added erroneously.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 MAINTAINERS                                     |   5 ++
 config/common_base                              |   5 ++
 doc/guides/nics/bnxt.rst                        |  49 +++++++++++
 doc/guides/nics/index.rst                       |   1 +
 drivers/net/Makefile                            |   1 +
 drivers/net/bnxt/Makefile                       |  63 ++++++++++++++
 drivers/net/bnxt/bnxt_ethdev.c                  | 104 ++++++++++++++++++++++++
 drivers/net/bnxt/rte_pmd_bnxt_version.map       |   4 +
 lib/librte_eal/common/include/rte_pci_dev_ids.h |  38 +++++++--
 mk/rte.app.mk                                   |   1 +
 10 files changed, 266 insertions(+), 5 deletions(-)
 create mode 100644 doc/guides/nics/bnxt.rst
 create mode 100644 drivers/net/bnxt/Makefile
 create mode 100644 drivers/net/bnxt/bnxt_ethdev.c
 create mode 100644 drivers/net/bnxt/rte_pmd_bnxt_version.map

diff --git a/MAINTAINERS b/MAINTAINERS
index 3e8558f..8892086 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -400,6 +400,11 @@ M: Declan Doherty <declan.doherty@intel.com>
 F: drivers/crypto/null/
 F: doc/guides/cryptodevs/null.rst
 
+Broadcom BNXT PMD
+M: Stephen Hurd <stephen.hurd@broadcom.com>
+F: drivers/net/bnxt/
+F: doc/guides/nics/bnxt.rst
+
 
 Packet processing
 -----------------
diff --git a/config/common_base b/config/common_base
index 47c26f6..dc298e9 100644
--- a/config/common_base
+++ b/config/common_base
@@ -245,6 +245,11 @@ CONFIG_RTE_LIBRTE_NFP_PMD=n
 CONFIG_RTE_LIBRTE_NFP_DEBUG=n
 
 #
+# Compile burst-oriented Broadcom BNXT PMD driver
+#
+CONFIG_RTE_LIBRTE_BNXT_PMD=y
+
+#
 # Compile software PMD backed by SZEDATA2 device
 #
 CONFIG_RTE_LIBRTE_PMD_SZEDATA2=n
diff --git a/doc/guides/nics/bnxt.rst b/doc/guides/nics/bnxt.rst
new file mode 100644
index 0000000..2669e98
--- /dev/null
+++ b/doc/guides/nics/bnxt.rst
@@ -0,0 +1,49 @@
+..  BSD LICENSE
+    Copyright 2016 Broadcom Limited
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions
+    are met:
+
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in
+    the documentation and/or other materials provided with the
+    distribution.
+    * Neither the name of Broadcom Limited nor the names of its
+    contributors may be used to endorse or promote products derived
+    from this software without specific prior written permission.
+
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+bnxt poll mode driver library
+=============================
+
+The bnxt poll mode library (**librte_pmd_bnxt**) implements support for
+**Broadcom NetXtreme® C-Series**.  These adapters support Standards-
+compliant 10/25/50Gbps 30MPPS full-duplex throughput.
+
+Information about this family of adapters can be found in the
+`NetXtreme® Brand section <https://www.broadcom.com/products/ethernet-communication-and-switching?technology%5B%5D=88>`_
+of the `Broadcom web site <http://www.broadcom.com/>`_.
+
+Limitations
+-----------
+
+With the current driver, allocated mbufs must be large enough to hold
+the entire received frame.  If the mbufs are not large enough, the
+packets will be dropped.  This is most limiting when jumbo frames are
+used.
+
+SR-IOV is not supported.
diff --git a/doc/guides/nics/index.rst b/doc/guides/nics/index.rst
index 0b13698..ffe011e 100644
--- a/doc/guides/nics/index.rst
+++ b/doc/guides/nics/index.rst
@@ -36,6 +36,7 @@ Network Interface Controller Drivers
     :numbered:
 
     overview
+    bnxt
     bnx2x
     cxgbe
     e1000em
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index 6ba7658..3832706 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -45,6 +45,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_MLX4_PMD) += mlx4
 DIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5
 DIRS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD) += mpipe
 DIRS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp
+DIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_NULL) += null
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_PCAP) += pcap
 DIRS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede
diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
new file mode 100644
index 0000000..f6333fd
--- /dev/null
+++ b/drivers/net/bnxt/Makefile
@@ -0,0 +1,63 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
+#   Copyright(c) 2014 6WIND S.A.
+#   Copyright(c) Broadcom Limited.
+#   All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_bnxt.a
+
+LIBABIVER := 1
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_bnxt_version.map
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+
+#
+# Export include files
+#
+SYMLINK-y-include +=
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_eal
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
new file mode 100644
index 0000000..d6c0d51
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -0,0 +1,104 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include <rte_dev.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+
+#define DRV_MODULE_NAME		"bnxt"
+static const char bnxt_version[] =
+	"Broadcom Cumulus driver " DRV_MODULE_NAME "\n";
+
+static struct rte_pci_id bnxt_pci_id_map[] = {
+#define RTE_PCI_DEV_ID_DECL_BNXT(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
+#include "rte_pci_dev_ids.h"
+	{.device_id = 0},
+};
+
+/*
+ * Initialization
+ */
+
+static int
+bnxt_dev_init(struct rte_eth_dev *eth_dev)
+{
+	static int version_printed;
+
+	if (version_printed++ == 0)
+		RTE_LOG(INFO, PMD, "%s", bnxt_version);
+
+	if (eth_dev->pci_dev->addr.function >= 2 &&
+			eth_dev->pci_dev->addr.function < 4) {
+		RTE_LOG(ERR, PMD, "Function not enabled %x:\n",
+			eth_dev->pci_dev->addr.function);
+		return -ENOMEM;
+	}
+
+	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
+	return -EPERM;
+}
+
+static int
+bnxt_dev_uninit(struct rte_eth_dev *eth_dev __rte_unused) {
+	return 0;
+}
+
+static struct eth_driver bnxt_rte_pmd = {
+	.pci_drv = {
+		    .name = "rte_" DRV_MODULE_NAME "_pmd",
+		    .id_table = bnxt_pci_id_map,
+		    .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
+		    },
+	.eth_dev_init = bnxt_dev_init,
+	.eth_dev_uninit = bnxt_dev_uninit,
+	.dev_private_size = 32 /* this must be non-zero apparently */,
+};
+
+static int bnxt_rte_pmd_init(const char *name, const char *params __rte_unused)
+{
+	RTE_LOG(INFO, PMD, "bnxt_rte_pmd_init() called for %s\n", name);
+	rte_eth_driver_register(&bnxt_rte_pmd);
+	return 0;
+}
+
+static struct rte_driver bnxt_pmd_drv = {
+	.name = "eth_bnxt",
+	.type = PMD_PDEV,
+	.init = bnxt_rte_pmd_init,
+};
+
+PMD_REGISTER_DRIVER(bnxt_pmd_drv);
diff --git a/drivers/net/bnxt/rte_pmd_bnxt_version.map b/drivers/net/bnxt/rte_pmd_bnxt_version.map
new file mode 100644
index 0000000..349c6e1
--- /dev/null
+++ b/drivers/net/bnxt/rte_pmd_bnxt_version.map
@@ -0,0 +1,4 @@
+DPDK_16.04 {
+
+	local: *;
+};
diff --git a/lib/librte_eal/common/include/rte_pci_dev_ids.h b/lib/librte_eal/common/include/rte_pci_dev_ids.h
index cf7b548..f63cdc0 100644
--- a/lib/librte_eal/common/include/rte_pci_dev_ids.h
+++ b/lib/librte_eal/common/include/rte_pci_dev_ids.h
@@ -63,11 +63,12 @@
  * This file contains a list of the PCI device IDs recognised by DPDK, which
  * can be used to fill out an array of structures describing the devices.
  *
- * Currently four families of devices are recognised: those supported by the
- * IGB driver, by EM driver, those supported by the IXGBE driver, and by virtio
- * driver which is a para virtualization driver running in guest virtual machine.
- * The inclusion of these in an array built using this file depends on the
- * definition of
+ * Currently five families of devices are recognised: those supported by the
+ * IGB driver, by EM driver, those supported by the IXGBE driver, those
+ * supported by the BNXT driver, and by virtio driver which is a para
+ * virtualization driver running in guest virtual machine. The inclusion of
+ * these in an array built using this file depends on the definition of
+ * RTE_PCI_DEV_ID_DECL_BNXT
  * RTE_PCI_DEV_ID_DECL_EM
  * RTE_PCI_DEV_ID_DECL_IGB
  * RTE_PCI_DEV_ID_DECL_IGBVF
@@ -152,6 +153,10 @@
 #define RTE_PCI_DEV_ID_DECL_BNX2XVF(vend, dev)
 #endif
 
+#ifndef RTE_PCI_DEV_ID_DECL_BNXT
+#define RTE_PCI_DEV_ID_DECL_BNXT(vend, dev)
+#endif
+
 #ifndef PCI_VENDOR_ID_INTEL
 /** Vendor ID used by Intel devices */
 #define PCI_VENDOR_ID_INTEL 0x8086
@@ -686,6 +691,28 @@ RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57811_MF)
 RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57840_MF)
 #endif
 
+/****************** Broadcom bnxt devices ******************/
+
+#define BROADCOM_DEV_ID_57301                  0x16c8
+#define BROADCOM_DEV_ID_57302                  0x16c9
+#define BROADCOM_DEV_ID_57304_PF               0x16ca
+#define BROADCOM_DEV_ID_57304_VF               0x16cb
+#define BROADCOM_DEV_ID_57402                  0x16d0
+#define BROADCOM_DEV_ID_57404                  0x16d1
+#define BROADCOM_DEV_ID_57406_PF               0x16d2
+#define BROADCOM_DEV_ID_57406_VF               0x16d3
+#define BROADCOM_DEV_ID_57406_MF               0x16d4
+
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57301)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF)
+
 /*
  * Undef all RTE_PCI_DEV_ID_DECL_* here.
  */
@@ -702,3 +729,4 @@ RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57840_MF)
 #undef RTE_PCI_DEV_ID_DECL_VMXNET3
 #undef RTE_PCI_DEV_ID_DECL_FM10K
 #undef RTE_PCI_DEV_ID_DECL_FM10KVF
+#undef RTE_PCI_DEV_ID_DECL_BNXT
diff --git a/mk/rte.app.mk b/mk/rte.app.mk
index b84b56d..effab79 100644
--- a/mk/rte.app.mk
+++ b/mk/rte.app.mk
@@ -143,6 +143,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_ENA_PMD)        += -lrte_pmd_ena
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MLX4_PMD)       += -lrte_pmd_mlx4
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MLX5_PMD)       += -lrte_pmd_mlx5
 _LDLIBS-$(CONFIG_RTE_LIBRTE_NFP_PMD)        += -lrte_pmd_nfp
+_LDLIBS-$(CONFIG_RTE_LIBRTE_BNXT_PMD)       += -lrte_pmd_bnxt
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2)   += -lrte_pmd_szedata2
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD)      += -lrte_pmd_mpipe
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_RING)       += -lrte_pmd_ring
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 02/39] bnxt: add HWRM init code
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 03/39] bnxt: add driver register/unregister support Stephen Hurd
                   ` (39 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Start adding support to use the HWRM API.
Hardware Resource Manager or HWRM in short, is a set of API provided
by the firmware running in the ASIC to manage the various resources.

Initial commit just performs necessary HWRM queries for init, then
fails as before.

The used HWRM calls so far:
bnxt_hwrm_func_qcaps:
	Queries device capabilities.

bnxt_hwrm_ver_get:
	Gets the firmware version and interface specifications.
	Returns an error if the firmware on the device is not
	supported by the driver and ensures the response space
	is large enough for the largest possible response.

bnxt_hwrm_queue_qportcfg:
	Required to get the default queue ID.

v4:
Fix few issues highlighted by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                | 114 ++++
 drivers/net/bnxt/bnxt_ethdev.c         | 111 ++++
 drivers/net/bnxt/bnxt_hwrm.c           | 324 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  53 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 954 +++++++++++++++++++++++++++++++++
 6 files changed, 1557 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt.h
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.c
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.h
 create mode 100644 drivers/net/bnxt/hsi_struct_def_dpdk.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index f6333fd..9965597 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -49,6 +49,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 # all source are stored in SRCS-y
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 
 #
 # Export include files
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
new file mode 100644
index 0000000..8cb7f5b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt.h
@@ -0,0 +1,114 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_H_
+#define _BNXT_H_
+
+#include <inttypes.h>
+#include <sys/queue.h>
+
+#include <rte_ethdev.h>
+#include <rte_memory.h>
+#include <rte_lcore.h>
+#include <rte_spinlock.h>
+
+struct bnxt_vf_info {
+	uint16_t		fw_fid;
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+	uint16_t		max_rsscos_ctx;
+	uint16_t		max_cp_rings;
+	uint16_t		max_tx_rings;
+	uint16_t		max_rx_rings;
+	uint16_t		max_l2_ctx;
+	uint16_t		max_vnics;
+	struct bnxt_pf_info	*pf;
+};
+
+struct bnxt_pf_info {
+#define BNXT_FIRST_PF_FID	1
+#define BNXT_MAX_VFS(bp)	(bp->pf.max_vfs)
+#define BNXT_FIRST_VF_FID	128
+#define BNXT_PF_RINGS_USED(bp)	bnxt_get_num_queues(bp)
+#define BNXT_PF_RINGS_AVAIL(bp)	(bp->pf.max_cp_rings - BNXT_PF_RINGS_USED(bp))
+	uint32_t		fw_fid;
+	uint8_t			port_id;
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+	uint16_t		max_rsscos_ctx;
+	uint16_t		max_cp_rings;
+	uint16_t		max_tx_rings;
+	uint16_t		max_rx_rings;
+	uint16_t		max_l2_ctx;
+	uint16_t		max_vnics;
+	uint16_t		first_vf_id;
+	uint16_t		active_vfs;
+	uint16_t		max_vfs;
+	void			*vf_req_buf;
+	phys_addr_t		vf_req_buf_dma_addr;
+	uint32_t		vf_req_fwd[8];
+	struct bnxt_vf_info	*vf;
+};
+
+#define BNXT_COS_QUEUE_COUNT	8
+struct bnxt_cos_queue_info {
+	uint8_t	id;
+	uint8_t	profile;
+};
+
+struct bnxt {
+	void				*bar0;
+
+	struct rte_eth_dev		*eth_dev;
+	struct rte_pci_device		*pdev;
+
+	uint32_t		flags;
+#define BNXT_FLAG_VF		(1 << 1)
+#define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
+#define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
+
+#define MAX_NUM_MAC_ADDR	32
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+
+	uint16_t			hwrm_cmd_seq;
+	void				*hwrm_cmd_resp_addr;
+	phys_addr_t			hwrm_cmd_resp_dma_addr;
+	rte_spinlock_t			hwrm_lock;
+	uint16_t			max_req_len;
+	uint16_t			max_resp_len;
+
+	struct bnxt_cos_queue_info	cos_queue[BNXT_COS_QUEUE_COUNT];
+
+	struct bnxt_pf_info		pf;
+	struct bnxt_vf_info		vf;
+};
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index d6c0d51..8ebd742 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -39,6 +39,9 @@
 #include <rte_malloc.h>
 #include <rte_cycles.h>
 
+#include "bnxt.h"
+#include "bnxt_hwrm.h"
+
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
 	"Broadcom Cumulus driver " DRV_MODULE_NAME "\n";
@@ -49,14 +52,69 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
+static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	rte_free(eth_dev->data->mac_addrs);
+	bnxt_free_hwrm_resources(bp);
+}
+
 /*
  * Initialization
  */
 
+static struct eth_dev_ops bnxt_dev_ops = {
+	.dev_close = bnxt_dev_close_op,
+};
+
+static bool bnxt_vf_pciid(uint16_t id)
+{
+	if (id == BROADCOM_DEV_ID_57304_VF ||
+	    id == BROADCOM_DEV_ID_57406_VF)
+		return true;
+	return false;
+}
+
+static int bnxt_init_board(struct rte_eth_dev *eth_dev)
+{
+	int rc;
+	struct bnxt *bp = eth_dev->data->dev_private;
+
+	/* enable device (incl. PCI PM wakeup), and bus-mastering */
+	if (!eth_dev->pci_dev->mem_resource[0].addr) {
+		RTE_LOG(ERR, PMD,
+			"Cannot find PCI device base address, aborting\n");
+		rc = -ENODEV;
+		goto init_err_disable;
+	}
+
+	bp->eth_dev = eth_dev;
+	bp->pdev = eth_dev->pci_dev;
+
+	bp->bar0 = (void *)eth_dev->pci_dev->mem_resource[0].addr;
+	if (!bp->bar0) {
+		RTE_LOG(ERR, PMD, "Cannot map device registers, aborting\n");
+		rc = -ENOMEM;
+		goto init_err_release;
+	}
+	return 0;
+
+init_err_release:
+	if (bp->bar0)
+		bp->bar0 = NULL;
+
+init_err_disable:
+
+	return rc;
+}
+
 static int
 bnxt_dev_init(struct rte_eth_dev *eth_dev)
 {
 	static int version_printed;
+	struct bnxt *bp;
+	int rc;
 
 	if (version_printed++ == 0)
 		RTE_LOG(INFO, PMD, "%s", bnxt_version);
@@ -69,7 +127,60 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	}
 
 	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
+	bp = eth_dev->data->dev_private;
+
+	if (bnxt_vf_pciid(eth_dev->pci_dev->id.device_id))
+		bp->flags |= BNXT_FLAG_VF;
+
+	rc = bnxt_init_board(eth_dev);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Board initialization failed rc: %x\n", rc);
+		goto error;
+	}
+	eth_dev->dev_ops = &bnxt_dev_ops;
+	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
+	/* eth_dev->tx_pkt_burst = &bnxt_xmit_pkts; */
+
+	rc = bnxt_alloc_hwrm_resources(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"hwrm resource allocation failure rc: %x\n", rc);
+		goto error;
+	}
+	rc = bnxt_hwrm_ver_get(bp);
+	if (rc)
+		goto error;
+	bnxt_hwrm_queue_qportcfg(bp);
+
+	/* Get the MAX capabilities for this function */
+	rc = bnxt_hwrm_func_qcaps(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "hwrm query capability failure rc: %x\n", rc);
+		goto error_free;
+	}
+	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
+					ETHER_ADDR_LEN * MAX_NUM_MAC_ADDR, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		RTE_LOG(ERR, PMD,
+			"Failed to alloc %u bytes needed to store MAC addr tbl",
+			ETHER_ADDR_LEN * MAX_NUM_MAC_ADDR);
+		rc = -ENOMEM;
+		goto error_free;
+	}
+	/* Copy the permanent MAC from the qcap response address now. */
+	if (BNXT_PF(bp))
+		memcpy(bp->mac_addr, bp->pf.mac_addr, sizeof(bp->mac_addr));
+	else
+		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
+	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
+
 	return -EPERM;
+
+error_free:
+	bnxt_dev_close_op(eth_dev);
+error:
+	return rc;
 }
 
 static int
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
new file mode 100644
index 0000000..e187121
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -0,0 +1,324 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_byteorder.h>
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+
+#include "bnxt.h"
+#include "bnxt_hwrm.h"
+#include "hsi_struct_def_dpdk.h"
+
+#define HWRM_CMD_TIMEOUT		2000
+
+/*
+ * HWRM Functions (sent to HWRM)
+ * These are named bnxt_hwrm_*() and return -1 if bnxt_hwrm_send_message()
+ * fails (ie: a timeout), and a positive non-zero HWRM error code if the HWRM
+ * command was failed by the ChiMP.
+ */
+
+static int bnxt_hwrm_send_message_locked(struct bnxt *bp, void *msg,
+					uint32_t msg_len)
+{
+	unsigned int i;
+	struct input *req = msg;
+	struct output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t *data = msg;
+	uint8_t *bar;
+	uint8_t *valid;
+
+	/* Write request msg to hwrm channel */
+	for (i = 0; i < msg_len; i += 4) {
+		bar = (uint8_t *)bp->bar0 + i;
+		*(volatile uint32_t *)bar = *data;
+		data++;
+	}
+
+	/* Zero the rest of the request space */
+	for (; i < bp->max_req_len; i += 4) {
+		bar = (uint8_t *)bp->bar0 + i;
+		*(volatile uint32_t *)bar = 0;
+	}
+
+	/* Ring channel doorbell */
+	bar = (uint8_t *)bp->bar0 + 0x100;
+	*(volatile uint32_t *)bar = 1;
+
+	/* Poll for the valid bit */
+	for (i = 0; i < HWRM_CMD_TIMEOUT; i++) {
+		/* Sanity check on the resp->resp_len */
+		rte_rmb();
+		if (resp->resp_len && resp->resp_len <=
+		    bp->max_resp_len) {
+			/* Last byte of resp contains the valid key */
+			valid = (uint8_t *)resp + resp->resp_len - 1;
+			if (*valid == HWRM_RESP_VALID_KEY)
+				break;
+		}
+		rte_delay_us(600);
+	}
+
+	if (i >= HWRM_CMD_TIMEOUT) {
+		RTE_LOG(ERR, PMD, "Error sending msg %x\n",
+			req->req_type);
+		goto err_ret;
+	}
+	return 0;
+
+err_ret:
+	return -1;
+}
+
+static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
+{
+	int rc;
+
+	rte_spinlock_lock(&bp->hwrm_lock);
+	rc = bnxt_hwrm_send_message_locked(bp, msg, msg_len);
+	rte_spinlock_unlock(&bp->hwrm_lock);
+	return rc;
+}
+
+#define HWRM_PREP(req, type, cr, resp) \
+	memset(bp->hwrm_cmd_resp_addr, 0, bp->max_resp_len); \
+	req.req_type = rte_cpu_to_le_16(HWRM_##type); \
+	req.cmpl_ring = rte_cpu_to_le_16(cr); \
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++); \
+	req.target_id = rte_cpu_to_le_16(0xffff); \
+	req.resp_addr = rte_cpu_to_le_64(bp->hwrm_cmd_resp_dma_addr)
+
+#define HWRM_CHECK_RESULT \
+	{ \
+		if (rc) { \
+			RTE_LOG(ERR, PMD, "%s failed rc:%d\n", \
+				__func__, rc); \
+			return rc; \
+		} \
+		if (resp->error_code) { \
+			rc = rte_le_to_cpu_16(resp->error_code); \
+			RTE_LOG(ERR, PMD, "%s error %d\n", __func__, rc); \
+			return rc; \
+		} \
+	}
+
+int bnxt_hwrm_func_qcaps(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_func_qcaps_input req = {.req_type = 0 };
+	struct hwrm_func_qcaps_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, FUNC_QCAPS, -1, resp);
+
+	req.fid = rte_cpu_to_le_16(0xffff);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		pf->fw_fid = rte_le_to_cpu_32(resp->fid);
+		pf->port_id = resp->port_id;
+		memcpy(pf->mac_addr, resp->perm_mac_address, ETHER_ADDR_LEN);
+		pf->max_rsscos_ctx = rte_le_to_cpu_16(resp->max_rsscos_ctx);
+		pf->max_cp_rings = rte_le_to_cpu_16(resp->max_cmpl_rings);
+		pf->max_tx_rings = rte_le_to_cpu_16(resp->max_tx_rings);
+		pf->max_rx_rings = rte_le_to_cpu_16(resp->max_rx_rings);
+		pf->max_l2_ctx = rte_le_to_cpu_16(resp->max_l2_ctxs);
+		pf->max_vnics = rte_le_to_cpu_16(resp->max_vnics);
+		pf->first_vf_id = rte_le_to_cpu_16(resp->first_vf_id);
+		pf->max_vfs = rte_le_to_cpu_16(resp->max_vfs);
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		vf->fw_fid = rte_le_to_cpu_32(resp->fid);
+		memcpy(vf->mac_addr, &resp->perm_mac_address, ETHER_ADDR_LEN);
+		vf->max_rsscos_ctx = rte_le_to_cpu_16(resp->max_rsscos_ctx);
+		vf->max_cp_rings = rte_le_to_cpu_16(resp->max_cmpl_rings);
+		vf->max_tx_rings = rte_le_to_cpu_16(resp->max_tx_rings);
+		vf->max_rx_rings = rte_le_to_cpu_16(resp->max_rx_rings);
+		vf->max_l2_ctx = rte_le_to_cpu_16(resp->max_l2_ctxs);
+		vf->max_vnics = rte_le_to_cpu_16(resp->max_vnics);
+	}
+
+	return rc;
+}
+
+int bnxt_hwrm_ver_get(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_ver_get_input req = {.req_type = 0 };
+	struct hwrm_ver_get_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t my_version;
+	uint32_t fw_version;
+	uint16_t max_resp_len;
+	char type[RTE_MEMZONE_NAMESIZE];
+
+	HWRM_PREP(req, VER_GET, -1, resp);
+
+	req.hwrm_intf_maj = HWRM_VERSION_MAJOR;
+	req.hwrm_intf_min = HWRM_VERSION_MINOR;
+	req.hwrm_intf_upd = HWRM_VERSION_UPDATE;
+
+	/*
+	 * Hold the lock since we may be adjusting the response pointers.
+	 */
+	rte_spinlock_lock(&bp->hwrm_lock);
+	rc = bnxt_hwrm_send_message_locked(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	RTE_LOG(INFO, PMD, "%d.%d.%d:%d.%d.%d\n",
+		resp->hwrm_intf_maj, resp->hwrm_intf_min,
+		resp->hwrm_intf_upd,
+		resp->hwrm_fw_maj, resp->hwrm_fw_min, resp->hwrm_fw_bld);
+
+	my_version = HWRM_VERSION_MAJOR << 16;
+	my_version |= HWRM_VERSION_MINOR << 8;
+	my_version |= HWRM_VERSION_UPDATE;
+
+	fw_version = resp->hwrm_intf_maj << 16;
+	fw_version |= resp->hwrm_intf_min << 8;
+	fw_version |= resp->hwrm_intf_upd;
+
+	if (resp->hwrm_intf_maj != HWRM_VERSION_MAJOR) {
+		RTE_LOG(ERR, PMD, "Unsupported firmware API version\n");
+		rc = -EINVAL;
+		goto error;
+	}
+
+	if (my_version != fw_version) {
+		RTE_LOG(INFO, PMD, "BNXT Driver/HWRM API mismatch.\n");
+		if (my_version < fw_version) {
+			RTE_LOG(INFO, PMD,
+				"Firmware API version is newer than driver.\n");
+			RTE_LOG(INFO, PMD,
+				"The driver may be missing features.\n");
+		} else {
+			RTE_LOG(INFO, PMD,
+				"Firmware API version is older than driver.\n");
+			RTE_LOG(INFO, PMD,
+				"Not all driver features may be functional.\n");
+		}
+	}
+
+	if (bp->max_req_len > resp->max_req_win_len) {
+		RTE_LOG(ERR, PMD, "Unsupported request length\n");
+		rc = -EINVAL;
+	}
+	bp->max_req_len = resp->max_req_win_len;
+	max_resp_len = resp->max_resp_len;
+	if (bp->max_resp_len != max_resp_len) {
+		sprintf(type, "bnxt_hwrm_%04x:%02x:%02x:%02x",
+			bp->pdev->addr.domain, bp->pdev->addr.bus,
+			bp->pdev->addr.devid, bp->pdev->addr.function);
+
+		rte_free(bp->hwrm_cmd_resp_addr);
+
+		bp->hwrm_cmd_resp_addr = rte_malloc(type, max_resp_len, 0);
+		if (bp->hwrm_cmd_resp_addr == NULL) {
+			rc = -ENOMEM;
+			goto error;
+		}
+		bp->hwrm_cmd_resp_dma_addr =
+			rte_malloc_virt2phy(bp->hwrm_cmd_resp_addr);
+		bp->max_resp_len = max_resp_len;
+	}
+
+error:
+	rte_spinlock_unlock(&bp->hwrm_lock);
+	return rc;
+}
+
+int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_queue_qportcfg_input req = {.req_type = 0 };
+	struct hwrm_queue_qportcfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, QUEUE_QPORTCFG, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+#define GET_QUEUE_INFO(x) \
+	bp->cos_queue[x].id = resp->queue_id##x; \
+	bp->cos_queue[x].profile = resp->queue_id##x##_service_profile
+
+	GET_QUEUE_INFO(0);
+	GET_QUEUE_INFO(1);
+	GET_QUEUE_INFO(2);
+	GET_QUEUE_INFO(3);
+	GET_QUEUE_INFO(4);
+	GET_QUEUE_INFO(5);
+	GET_QUEUE_INFO(6);
+	GET_QUEUE_INFO(7);
+
+	return rc;
+}
+
+/*
+ * HWRM utility functions
+ */
+
+void bnxt_free_hwrm_resources(struct bnxt *bp)
+{
+	/* Release memzone */
+	rte_free(bp->hwrm_cmd_resp_addr);
+	bp->hwrm_cmd_resp_addr = NULL;
+	bp->hwrm_cmd_resp_dma_addr = 0;
+}
+
+int bnxt_alloc_hwrm_resources(struct bnxt *bp)
+{
+	struct rte_pci_device *pdev = bp->pdev;
+	char type[RTE_MEMZONE_NAMESIZE];
+
+	sprintf(type, "bnxt_hwrm_%04x:%02x:%02x:%02x", pdev->addr.domain,
+		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+	bp->max_req_len = HWRM_MAX_REQ_LEN;
+	bp->max_resp_len = HWRM_MAX_RESP_LEN;
+	bp->hwrm_cmd_resp_addr = rte_malloc(type, bp->max_resp_len, 0);
+	if (bp->hwrm_cmd_resp_addr == NULL)
+		return -ENOMEM;
+	bp->hwrm_cmd_resp_dma_addr =
+		rte_malloc_virt2phy(bp->hwrm_cmd_resp_addr);
+	rte_spinlock_init(&bp->hwrm_lock);
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
new file mode 100644
index 0000000..e35e8c0
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -0,0 +1,53 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_HWRM_H_
+#define _BNXT_HWRM_H_
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include "bnxt.h"
+
+#define HWRM_SEQ_ID_INVALID -1U
+
+int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+
+int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
+
+int bnxt_hwrm_ver_get(struct bnxt *bp);
+
+void bnxt_free_hwrm_resources(struct bnxt *bp);
+int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
new file mode 100644
index 0000000..1667927
--- /dev/null
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -0,0 +1,954 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _HSI_STRUCT_DEF_EXTERNAL_H_
+#define _HSI_STRUCT_DEF_EXTERNAL_H_
+
+/* HW Resource Manager Specification 1.2.0 */
+#define HWRM_VERSION_MAJOR	1
+#define HWRM_VERSION_MINOR	2
+#define HWRM_VERSION_UPDATE	0
+
+/*
+ * Following is the signature for HWRM message field that indicates not
+ * applicable (All F's). Need to cast it the size of the field if needed.
+ */
+#define HWRM_MAX_REQ_LEN	(128)  /* hwrm_func_buf_rgtr */
+#define HWRM_MAX_RESP_LEN	(176)  /* hwrm_func_qstats */
+#define HWRM_RESP_VALID_KEY	1 /* valid key for HWRM response */
+
+/*
+ * Request types
+ */
+#define HWRM_VER_GET                                      (UINT32_C(0x0))
+#define HWRM_FUNC_QCAPS                                   (UINT32_C(0x15))
+#define HWRM_QUEUE_QPORTCFG                               (UINT32_C(0x30))
+
+/*
+ * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
+ * inside the chip. The HWRM is implemented in firmware, and runs on embedded
+ * processors inside the chip. This firmware is vital part of the chip's
+ * hardware. The chip can not be used by driver without it.
+ */
+
+/* Input (16 bytes) */
+struct input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+} __attribute__((packed));
+
+/* Output (8 bytes) */
+struct output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+} __attribute__((packed));
+
+/* hwrm_func_qcaps */
+/*
+ * Description: This command returns capabilities of a function. The input FID
+ * value is used to indicate what function is being queried. This allows a
+ * physical function driver to query virtual functions that are children of the
+ * physical function. The output FID value is needed to configure Rings and
+ * MSI-X vectors so their DMA operations appear correctly on the PCI bus.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_func_qcaps_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Function ID of the function that is being queried. 0xFF... (All Fs)
+	 * if the query is for the requesting function.
+	 */
+	uint16_t fid;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (80 bytes) */
+struct hwrm_func_qcaps_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * FID value. This value is used to identify operations on the PCI bus
+	 * as belonging to a particular PCI function.
+	 */
+	uint16_t fid;
+
+	/*
+	 * Port ID of port that this function is associated with. Valid only for
+	 * the PF. 0xFF... (All Fs) if this function is not associated with any
+	 * port. 0xFF... (All Fs) if this function is called from a VF.
+	 */
+	uint16_t port_id;
+
+	/* If 1, then Push mode is supported on this function. */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_PUSH_MODE_SUPPORTED   UINT32_C(0x1)
+	/*
+	 * If 1, then the global MSI-X auto-masking is enabled for the device.
+	 */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_GLOBAL_MSIX_AUTOMASKING \
+								UINT32_C(0x2)
+	/*
+	 * If 1, then the Precision Time Protocol (PTP) processing is supported
+	 * on this function. The HWRM should enable PTP on only a single
+	 * Physical Function (PF) per port.
+	 */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_PTP_SUPPORTED         UINT32_C(0x4)
+	uint32_t flags;
+
+	/*
+	 * This value is current MAC address configured for this function. A
+	 * value of 00-00-00-00-00-00 indicates no MAC address is currently
+	 * configured.
+	 */
+	uint8_t perm_mac_address[6];
+
+	/*
+	 * The maximum number of RSS/COS contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_rsscos_ctx;
+
+	/*
+	 * The maximum number of completion rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_cmpl_rings;
+
+	/*
+	 * The maximum number of transmit rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_tx_rings;
+
+	/*
+	 * The maximum number of receive rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_rx_rings;
+
+	/*
+	 * The maximum number of L2 contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_l2_ctxs;
+
+	/* The maximum number of VNICs that can be allocated to the function. */
+	uint16_t max_vnics;
+
+	/*
+	 * The identifier for the first VF enabled on a PF. This is valid only
+	 * on the PF with SR-IOV enabled. 0xFF... (All Fs) if this command is
+	 * called on a PF with SR-IOV disabled or on a VF.
+	 */
+	uint16_t first_vf_id;
+
+	/*
+	 * The maximum number of VFs that can be allocated to the function. This
+	 * is valid only on the PF with SR-IOV enabled. 0xFF... (All Fs) if this
+	 * command is called on a PF with SR-IOV disabled or on a VF.
+	 */
+	uint16_t max_vfs;
+
+	/*
+	 * The maximum number of statistic contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_stat_ctx;
+
+	/*
+	 * The maximum number of Encapsulation records that can be offloaded by
+	 * this function.
+	 */
+	uint32_t max_encap_records;
+
+	/*
+	 * The maximum number of decapsulation records that can be offloaded by
+	 * this function.
+	 */
+	uint32_t max_decap_records;
+
+	/*
+	 * The maximum number of Exact Match (EM) flows that can be offloaded by
+	 * this function on the TX side.
+	 */
+	uint32_t max_tx_em_flows;
+
+	/*
+	 * The maximum number of Wildcard Match (WM) flows that can be offloaded
+	 * by this function on the TX side.
+	 */
+	uint32_t max_tx_wm_flows;
+
+	/*
+	 * The maximum number of Exact Match (EM) flows that can be offloaded by
+	 * this function on the RX side.
+	 */
+	uint32_t max_rx_em_flows;
+
+	/*
+	 * The maximum number of Wildcard Match (WM) flows that can be offloaded
+	 * by this function on the RX side.
+	 */
+	uint32_t max_rx_wm_flows;
+
+	/*
+	 * The maximum number of multicast filters that can be supported by this
+	 * function on the RX side.
+	 */
+	uint32_t max_mcast_filters;
+
+	/*
+	 * The maximum value of flow_id that can be supported in completion
+	 * records.
+	 */
+	uint32_t max_flow_id;
+
+	/*
+	 * The maximum number of HW ring groups that can be supported on this
+	 * function.
+	 */
+	uint32_t max_hw_ring_grps;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ver_get */
+/*
+ * Description: This function is called by a driver to determine the HWRM
+ * interface version supported by the HWRM firmware, the version of HWRM
+ * firmware implementation, the name of HWRM firmware, the versions of other
+ * embedded firmwares, and the names of other embedded firmwares, etc. Any
+ * interface or firmware version with major = 0, minor = 0, and update = 0 shall
+ * be considered an invalid version.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ver_get_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This field represents the major version of HWRM interface
+	 * specification supported by the driver HWRM implementation. The
+	 * interface major version is intended to change only when non backward
+	 * compatible changes are made to the HWRM interface specification.
+	 */
+	uint8_t hwrm_intf_maj;
+
+	/*
+	 * This field represents the minor version of HWRM interface
+	 * specification supported by the driver HWRM implementation. A change
+	 * in interface minor version is used to reflect significant backward
+	 * compatible modification to HWRM interface specification. This can be
+	 * due to addition or removal of functionality. HWRM interface
+	 * specifications with the same major version but different minor
+	 * versions are compatible.
+	 */
+	uint8_t hwrm_intf_min;
+
+	/*
+	 * This field represents the update version of HWRM interface
+	 * specification supported by the driver HWRM implementation. The
+	 * interface update version is used to reflect minor changes or bug
+	 * fixes to a released HWRM interface specification.
+	 */
+	uint8_t hwrm_intf_upd;
+
+	uint8_t unused_0[5];
+} __attribute__((packed));
+
+/* Output (128 bytes) */
+struct hwrm_ver_get_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This field represents the major version of HWRM interface
+	 * specification supported by the HWRM implementation. The interface
+	 * major version is intended to change only when non backward compatible
+	 * changes are made to the HWRM interface specification. A HWRM
+	 * implementation that is compliant with this specification shall
+	 * provide value of 1 in this field.
+	 */
+	uint8_t hwrm_intf_maj;
+
+	/*
+	 * This field represents the minor version of HWRM interface
+	 * specification supported by the HWRM implementation. A change in
+	 * interface minor version is used to reflect significant backward
+	 * compatible modification to HWRM interface specification. This can be
+	 * due to addition or removal of functionality. HWRM interface
+	 * specifications with the same major version but different minor
+	 * versions are compatible. A HWRM implementation that is compliant with
+	 * this specification shall provide value of 0 in this field.
+	 */
+	uint8_t hwrm_intf_min;
+
+	/*
+	 * This field represents the update version of HWRM interface
+	 * specification supported by the HWRM implementation. The interface
+	 * update version is used to reflect minor changes or bug fixes to a
+	 * released HWRM interface specification. A HWRM implementation that is
+	 * compliant with this specification shall provide value of 1 in this
+	 * field.
+	 */
+	uint8_t hwrm_intf_upd;
+
+	uint8_t hwrm_intf_rsvd;
+
+	/*
+	 * This field represents the major version of HWRM firmware. A change in
+	 * firmware major version represents a major firmware release.
+	 */
+	uint8_t hwrm_fw_maj;
+
+	/*
+	 * This field represents the minor version of HWRM firmware. A change in
+	 * firmware minor version represents significant firmware functionality
+	 * changes.
+	 */
+	uint8_t hwrm_fw_min;
+
+	/*
+	 * This field represents the build version of HWRM firmware. A change in
+	 * firmware build version represents bug fixes to a released firmware.
+	 */
+	uint8_t hwrm_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version of the HWRM firmware.
+	 */
+	uint8_t hwrm_fw_rsvd;
+
+	/*
+	 * This field represents the major version of mgmt firmware. A change in
+	 * major version represents a major release.
+	 */
+	uint8_t mgmt_fw_maj;
+
+	/*
+	 * This field represents the minor version of mgmt firmware. A change in
+	 * minor version represents significant functionality changes.
+	 */
+	uint8_t mgmt_fw_min;
+
+	/*
+	 * This field represents the build version of mgmt firmware. A change in
+	 * update version represents bug fixes.
+	 */
+	uint8_t mgmt_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t mgmt_fw_rsvd;
+
+	/*
+	 * This field represents the major version of network control firmware.
+	 * A change in major version represents a major release.
+	 */
+	uint8_t netctrl_fw_maj;
+
+	/*
+	 * This field represents the minor version of network control firmware.
+	 * A change in minor version represents significant functionality
+	 * changes.
+	 */
+	uint8_t netctrl_fw_min;
+
+	/*
+	 * This field represents the build version of network control firmware.
+	 * A change in update version represents bug fixes.
+	 */
+	uint8_t netctrl_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t netctrl_fw_rsvd;
+
+	/*
+	 * This field is reserved for future use. The responder should set it to
+	 * 0. The requester should ignore this field.
+	 */
+	uint32_t reserved1;
+
+	/*
+	 * This field represents the major version of RoCE firmware. A change in
+	 * major version represents a major release.
+	 */
+	uint8_t roce_fw_maj;
+
+	/*
+	 * This field represents the minor version of RoCE firmware. A change in
+	 * minor version represents significant functionality changes.
+	 */
+	uint8_t roce_fw_min;
+
+	/*
+	 * This field represents the build version of RoCE firmware. A change in
+	 * update version represents bug fixes.
+	 */
+	uint8_t roce_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t roce_fw_rsvd;
+
+	/*
+	 * This field represents the name of HWRM FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char hwrm_fw_name[16];
+
+	/*
+	 * This field represents the name of mgmt FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char mgmt_fw_name[16];
+
+	/*
+	 * This field represents the name of network control firmware (ASCII
+	 * chars without NULL at the end).
+	 */
+	char netctrl_fw_name[16];
+
+	/*
+	 * This field is reserved for future use. The responder should set it to
+	 * 0. The requester should ignore this field.
+	 */
+	uint32_t reserved2[4];
+
+	/*
+	 * This field represents the name of RoCE FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char roce_fw_name[16];
+
+	/* This field returns the chip number. */
+	uint16_t chip_num;
+
+	/* This field returns the revision of chip. */
+	uint8_t chip_rev;
+
+	/* This field returns the chip metal number. */
+	uint8_t chip_metal;
+
+	/* This field returns the bond id of the chip. */
+	uint8_t chip_bond_id;
+
+	/*
+	 * This value indicates the type of platform used for chip
+	 * implementation.
+	 */
+	/* ASIC */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_ASIC \
+							(UINT32_C(0x0) << 0)
+	/* FPGA platform of the chip. */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_FPGA \
+							(UINT32_C(0x1) << 0)
+	/* Palladium platform of the chip. */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_PALLADIUM \
+							(UINT32_C(0x2) << 0)
+	uint8_t chip_platform_type;
+
+	/*
+	 * This field returns the maximum value of request window that is
+	 * supported by the HWRM. The request window is mapped into device
+	 * address space using MMIO.
+	 */
+	uint16_t max_req_win_len;
+
+	/*
+	 * This field returns the maximum value of response buffer in bytes. If
+	 * a request specifies the response buffer length that is greater than
+	 * this value, then the HWRM should fail it. The value of this field
+	 * shall be 4KB or more.
+	 */
+	uint16_t max_resp_len;
+
+	/*
+	 * This field returns the default request timeout value in milliseconds.
+	 */
+	uint16_t def_req_timeout;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_queue_qportcfg */
+/*
+ * Description: This function is called by a driver to query queue configuration
+ * of a port. # The HWRM shall at least advertise one queue with lossy service
+ * profile. # The driver shall use this command to query queue ids before
+ * configuring or using any queues. # If a service profile is not set for a
+ * queue, then the driver shall not use that queue without configuring a service
+ * profile for it. # If the driver is not allowed to configure service profiles,
+ * then the driver shall only use queues for which service profiles are pre-
+ * configured.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_queue_qportcfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Enumeration denoting the RX, TX type of the resource. This
+	 * enumeration is used for resources that are similar for both TX and RX
+	 * paths of the chip.
+	 */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH \
+							UINT32_C(0x1)
+		/* tx path */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_TX \
+							(UINT32_C(0x0) << 0)
+		/* rx path */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_RX \
+							(UINT32_C(0x1) << 0)
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_LAST \
+					HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_RX
+	uint32_t flags;
+
+	/*
+	 * Port ID of port for which the queue configuration is being queried.
+	 * This field is only required when sent by IPC.
+	 */
+	uint16_t port_id;
+
+	uint16_t unused_0;
+} __attribute__((packed));
+
+/* Output (32 bytes) */
+struct hwrm_queue_qportcfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* The maximum number of queues that can be configured. */
+	uint8_t max_configurable_queues;
+
+	/* The maximum number of lossless queues that can be configured. */
+	uint8_t max_configurable_lossless_queues;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed. If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure queues using
+	 * hwrm_queue_cfg.
+	 */
+	uint8_t queue_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure queue buffers
+	 * using hwrm_queue_buffers_cfg.
+	 */
+	uint8_t queue_buffers_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure PFC using
+	 * hwrm_queue_pfcenable_cfg.
+	 */
+	uint8_t queue_pfcenable_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure Priority to CoS
+	 * mapping using hwrm_queue_pri2cos_cfg.
+	 */
+	uint8_t queue_pri2cos_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure CoS Bandwidth
+	 * configuration using hwrm_queue_cos2bw_cfg.
+	 */
+	uint8_t queue_cos2bw_cfg_allowed;
+
+	/* ID of CoS Queue 0. FF - Invalid id */
+	uint8_t queue_id0;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id0_service_profile;
+
+	/* ID of CoS Queue 1. FF - Invalid id */
+	uint8_t queue_id1;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id1_service_profile;
+
+	/* ID of CoS Queue 2. FF - Invalid id */
+	uint8_t queue_id2;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id2_service_profile;
+
+	/* ID of CoS Queue 3. FF - Invalid id */
+	uint8_t queue_id3;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id3_service_profile;
+
+	/* ID of CoS Queue 4. FF - Invalid id */
+	uint8_t queue_id4;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id4_service_profile;
+
+	/* ID of CoS Queue 5. FF - Invalid id */
+	uint8_t queue_id5;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id5_service_profile;
+
+	/* ID of CoS Queue 6. FF - Invalid id */
+	uint8_t queue_id6_service_profile;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id6;
+
+	/* ID of CoS Queue 7. FF - Invalid id */
+	uint8_t queue_id7;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id7_service_profile;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 03/39] bnxt: add driver register/unregister support
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 02/39] bnxt: add HWRM init code Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 04/39] bnxt: add dev infos get operation Stephen Hurd
                   ` (38 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Move init() cleanup into uninit() function
Fix .dev_private_size
Add require hwrm calls:
	bnxt_hwrm_func_driver_register()
	bnxt_hwrm_func_driver_unregister()

v4:
Address review comment regarding removal of bnxt_dev_close_op

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt.h                |   1 +
 drivers/net/bnxt/bnxt_ethdev.c         |  38 ++++-
 drivers/net/bnxt/bnxt_hwrm.c           |  50 ++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 277 ++++++++++++++++++++++++++++++++-
 5 files changed, 358 insertions(+), 11 deletions(-)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index 8cb7f5b..ed057ef 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -91,6 +91,7 @@ struct bnxt {
 	struct rte_pci_device		*pdev;
 
 	uint32_t		flags;
+#define BNXT_FLAG_REGISTERED	(1 << 0)
 #define BNXT_FLAG_VF		(1 << 1)
 #define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
 #define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 8ebd742..26e6447 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -123,7 +123,8 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 			eth_dev->pci_dev->addr.function < 4) {
 		RTE_LOG(ERR, PMD, "Function not enabled %x:\n",
 			eth_dev->pci_dev->addr.function);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto error;
 	}
 
 	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
@@ -146,11 +147,11 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	if (rc) {
 		RTE_LOG(ERR, PMD,
 			"hwrm resource allocation failure rc: %x\n", rc);
-		goto error;
+		goto error_free;
 	}
 	rc = bnxt_hwrm_ver_get(bp);
 	if (rc)
-		goto error;
+		goto error_free;
 	bnxt_hwrm_queue_qportcfg(bp);
 
 	/* Get the MAX capabilities for this function */
@@ -175,17 +176,38 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
 
-	return -EPERM;
+	rc = bnxt_hwrm_func_driver_register(bp, 0,
+					    bp->pf.vf_req_fwd);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Failed to register driver");
+		rc = -EBUSY;
+		goto error_free;
+	}
+
+	RTE_LOG(INFO, PMD,
+		DRV_MODULE_NAME " found at mem %" PRIx64 ", node addr %pM\n",
+		eth_dev->pci_dev->mem_resource[0].phys_addr,
+		eth_dev->pci_dev->mem_resource[0].addr);
+
+	return 0;
 
 error_free:
-	bnxt_dev_close_op(eth_dev);
+	eth_dev->driver->eth_dev_uninit(eth_dev);
 error:
 	return rc;
 }
 
 static int
-bnxt_dev_uninit(struct rte_eth_dev *eth_dev __rte_unused) {
-	return 0;
+bnxt_dev_uninit(struct rte_eth_dev *eth_dev) {
+	struct bnxt *bp = eth_dev->data->dev_private;
+	int rc;
+
+	if (eth_dev->data->mac_addrs)
+		rte_free(eth_dev->data->mac_addrs);
+	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
+	bnxt_free_hwrm_resources(bp);
+	return rc;
 }
 
 static struct eth_driver bnxt_rte_pmd = {
@@ -196,7 +218,7 @@ static struct eth_driver bnxt_rte_pmd = {
 		    },
 	.eth_dev_init = bnxt_dev_init,
 	.eth_dev_uninit = bnxt_dev_uninit,
-	.dev_private_size = 32 /* this must be non-zero apparently */,
+	.dev_private_size = sizeof(struct bnxt),
 };
 
 static int bnxt_rte_pmd_init(const char *name, const char *params __rte_unused)
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index e187121..8aba8cd 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -36,6 +36,7 @@
 #include <rte_cycles.h>
 #include <rte_malloc.h>
 #include <rte_memzone.h>
+#include <rte_version.h>
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
@@ -178,6 +179,34 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
+				   uint32_t *vf_req_fwd)
+{
+	int rc;
+	struct hwrm_func_drv_rgtr_input req = {.req_type = 0 };
+	struct hwrm_func_drv_rgtr_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (bp->flags & BNXT_FLAG_REGISTERED)
+		return 0;
+
+	HWRM_PREP(req, FUNC_DRV_RGTR, -1, resp);
+	req.flags = flags;
+	req.enables = HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VER;
+	req.ver_maj = RTE_VER_YEAR;
+	req.ver_min = RTE_VER_MONTH;
+	req.ver_upd = RTE_VER_MINOR;
+
+	memcpy(req.vf_req_fwd, vf_req_fwd, sizeof(req.vf_req_fwd));
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->flags |= BNXT_FLAG_REGISTERED;
+
+	return rc;
+}
+
 int bnxt_hwrm_ver_get(struct bnxt *bp)
 {
 	int rc = 0;
@@ -264,6 +293,27 @@ error:
 	return rc;
 }
 
+int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags)
+{
+	int rc;
+	struct hwrm_func_drv_unrgtr_input req = {.req_type = 0 };
+	struct hwrm_func_drv_unrgtr_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (!(bp->flags & BNXT_FLAG_REGISTERED))
+		return 0;
+
+	HWRM_PREP(req, FUNC_DRV_UNRGTR, -1, resp);
+	req.flags = flags;
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->flags &= ~BNXT_FLAG_REGISTERED;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index e35e8c0..6f2e445 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,7 +41,10 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
+				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 1667927..be3ed0f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -50,9 +50,11 @@
 /*
  * Request types
  */
-#define HWRM_VER_GET                                      (UINT32_C(0x0))
-#define HWRM_FUNC_QCAPS                                   (UINT32_C(0x15))
-#define HWRM_QUEUE_QPORTCFG                               (UINT32_C(0x30))
+#define HWRM_VER_GET		(UINT32_C(0x0))
+#define HWRM_FUNC_QCAPS		(UINT32_C(0x15))
+#define HWRM_FUNC_DRV_UNRGTR	(UINT32_C(0x1a))
+#define HWRM_FUNC_DRV_RGTR	(UINT32_C(0x1d))
+#define HWRM_QUEUE_QPORTCFG	(UINT32_C(0x30))
 
 /*
  * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
@@ -951,4 +953,273 @@ struct hwrm_queue_qportcfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_func_drv_rgtr */
+/*
+ * Description: This command is used by the function driver to register its
+ * information with the HWRM. A function driver shall implement this command. A
+ * function driver shall use this command during the driver initialization right
+ * after the HWRM version discovery and default ring resources allocation.
+ */
+
+/* Input (80 bytes) */
+struct hwrm_func_drv_rgtr_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the function driver is requesting all requests
+	 * from its children VF drivers to be forwarded to itself. This flag can
+	 * only be set by the PF driver. If a VF driver sets this flag, it
+	 * should be ignored by the HWRM.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_FLAGS_FWD_ALL_MODE        UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the function is requesting none of the requests
+	 * from its children VF drivers to be forwarded to itself. This flag can
+	 * only be set by the PF driver. If a VF driver sets this flag, it
+	 * should be ignored by the HWRM.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_FLAGS_FWD_NONE_MODE       UINT32_C(0x2)
+	uint32_t flags;
+
+	/* This bit must be '1' for the os_type field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_OS_TYPE           UINT32_C(0x1)
+	/* This bit must be '1' for the ver field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VER               UINT32_C(0x2)
+	/* This bit must be '1' for the timestamp field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_TIMESTAMP         UINT32_C(0x4)
+	/* This bit must be '1' for the vf_req_fwd field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VF_REQ_FWD        UINT32_C(0x8)
+	/*
+	 * This bit must be '1' for the async_event_fwd field to be configured.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_ASYNC_EVENT_FWD \
+								UINT32_C(0x10)
+	uint32_t enables;
+
+	/* This value indicates the type of OS. */
+		/* Unknown */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* Other OS not listed below. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_OTHER \
+							(UINT32_C(0x1) << 0)
+		/* MSDOS OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_MSDOS \
+							(UINT32_C(0xe) << 0)
+		/* Windows OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WINDOWS \
+							(UINT32_C(0x12) << 0)
+		/* Solaris OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_SOLARIS \
+							(UINT32_C(0x1d) << 0)
+		/* Linux OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_LINUX \
+							(UINT32_C(0x24) << 0)
+		/* FreeBSD OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_FREEBSD \
+							(UINT32_C(0x2a) << 0)
+		/* VMware ESXi OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_ESXI \
+							(UINT32_C(0x68) << 0)
+		/* Microsoft Windows 8 64-bit OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WIN864 \
+							(UINT32_C(0x73) << 0)
+		/* Microsoft Windows Server 2012 R2 OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WIN2012R2 \
+							(UINT32_C(0x74) << 0)
+	uint16_t os_type;
+
+	/* This is the major version of the driver. */
+	uint8_t ver_maj;
+
+	/* This is the minor version of the driver. */
+	uint8_t ver_min;
+
+	/* This is the update version of the driver. */
+	uint8_t ver_upd;
+
+	uint8_t unused_0;
+	uint16_t unused_1;
+
+	/*
+	 * This is a 32-bit timestamp provided by the driver for keep alive. The
+	 * timestamp is in multiples of 1ms.
+	 */
+	uint32_t timestamp;
+
+	uint32_t unused_2;
+
+	/*
+	 * This is a 256-bit bit mask provided by the PF driver for letting the
+	 * HWRM know what commands issued by the VF driver to the HWRM should be
+	 * forwarded to the PF driver. Nth bit refers to the Nth req_type.
+	 * Setting Nth bit to 1 indicates that requests from the VF driver with
+	 * req_type equal to N shall be forwarded to the parent PF driver. This
+	 * field is not valid for the VF driver.
+	 */
+	uint32_t vf_req_fwd[8];
+
+	/*
+	 * This is a 256-bit bit mask provided by the function driver (PF or VF
+	 * driver) to indicate the list of asynchronous event completions to be
+	 * forwarded. Nth bit refers to the Nth event_id. Setting Nth bit to 1
+	 * by the function driver shall result in the HWRM forwarding
+	 * asynchronous event completion with event_id equal to N. If all bits
+	 * are set to 0 (value of 0), then the HWRM shall not forward any
+	 * asynchronous event completion to this function driver.
+	 */
+	uint32_t async_event_fwd[8];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_func_drv_rgtr_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_func_drv_unrgtr */
+/*
+ * Description: This command is used by the function driver to un register with
+ * the HWRM. A function driver shall implement this command. A function driver
+ * shall use this command during the driver unloading.
+ */
+/* Input (24 bytes) */
+
+struct hwrm_func_drv_unrgtr_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the function driver is notifying the HWRM to
+	 * prepare for the shutdown.
+	 */
+	#define HWRM_FUNC_DRV_UNRGTR_INPUT_FLAGS_PREPARE_FOR_SHUTDOWN \
+							UINT32_C(0x1)
+	uint32_t flags;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_func_drv_unrgtr_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 #endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 04/39] bnxt: add dev infos get operation
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 02/39] bnxt: add HWRM init code Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 03/39] bnxt: add driver register/unregister support Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 05/39] bnxt: add dev configure operation Stephen Hurd
                   ` (37 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Gets device info from the bp structure filled in the init() function.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt.h        |  3 ++
 drivers/net/bnxt/bnxt_ethdev.c | 95 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 98 insertions(+)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index ed057ef..f8707b2 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -42,6 +42,9 @@
 #include <rte_lcore.h>
 #include <rte_spinlock.h>
 
+#define BNXT_MAX_MTU		9000
+#define VLAN_TAG_SIZE		4
+
 struct bnxt_vf_info {
 	uint16_t		fw_fid;
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 26e6447..a8a9912 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -61,10 +61,105 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 }
 
 /*
+ * Device configuration and status function
+ */
+
+static void bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
+				  struct rte_eth_dev_info *dev_info)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	uint16_t max_vnics, i, j, vpool, vrxq;
+
+	/* MAC Specifics */
+	dev_info->max_mac_addrs = MAX_NUM_MAC_ADDR;
+	dev_info->max_hash_mac_addrs = 0;
+
+	/* PF/VF specifics */
+	if (BNXT_PF(bp)) {
+		dev_info->max_rx_queues = bp->pf.max_rx_rings;
+		dev_info->max_tx_queues = bp->pf.max_tx_rings;
+		dev_info->max_vfs = bp->pf.active_vfs;
+		dev_info->reta_size = bp->pf.max_rsscos_ctx;
+		max_vnics = bp->pf.max_vnics;
+	} else {
+		dev_info->max_rx_queues = bp->vf.max_rx_rings;
+		dev_info->max_tx_queues = bp->vf.max_tx_rings;
+		dev_info->reta_size = bp->vf.max_rsscos_ctx;
+		max_vnics = bp->vf.max_vnics;
+	}
+
+	/* Fast path specifics */
+	dev_info->min_rx_bufsize = 1;
+	dev_info->max_rx_pktlen = BNXT_MAX_MTU + ETHER_HDR_LEN + ETHER_CRC_LEN
+				  + VLAN_TAG_SIZE;
+	dev_info->rx_offload_capa = 0;
+	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_IPV4_CKSUM |
+					DEV_TX_OFFLOAD_TCP_CKSUM |
+					DEV_TX_OFFLOAD_UDP_CKSUM |
+					DEV_TX_OFFLOAD_TCP_TSO;
+
+	/* *INDENT-OFF* */
+	dev_info->default_rxconf = (struct rte_eth_rxconf) {
+		.rx_thresh = {
+			.pthresh = 8,
+			.hthresh = 8,
+			.wthresh = 0,
+		},
+		.rx_free_thresh = 32,
+		.rx_drop_en = 0,
+	};
+
+	dev_info->default_txconf = (struct rte_eth_txconf) {
+		.tx_thresh = {
+			.pthresh = 32,
+			.hthresh = 0,
+			.wthresh = 0,
+		},
+		.tx_free_thresh = 32,
+		.tx_rs_thresh = 32,
+		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
+			     ETH_TXQ_FLAGS_NOOFFLOADS,
+	};
+	/* *INDENT-ON* */
+
+	/*
+	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
+	 *       need further investigation.
+	 */
+
+	/* VMDq resources */
+	vpool = 64; /* ETH_64_POOLS */
+	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
+	for (i = 0; i < 4; vpool >>= 1, i++) {
+		if (max_vnics > vpool) {
+			for (j = 0; j < 5; vrxq >>= 1, j++) {
+				if (dev_info->max_rx_queues > vrxq) {
+					if (vpool > vrxq)
+						vpool = vrxq;
+					goto found;
+				}
+			}
+			/* Not enough resources to support VMDq */
+			break;
+		}
+	}
+	/* Not enough resources to support VMDq */
+	vpool = 0;
+	vrxq = 0;
+found:
+	dev_info->max_vmdq_pools = vpool;
+	dev_info->vmdq_queue_num = vrxq;
+
+	dev_info->vmdq_pool_base = 0;
+	dev_info->vmdq_queue_base = 0;
+}
+
+/*
  * Initialization
  */
 
 static struct eth_dev_ops bnxt_dev_ops = {
+	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 };
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 05/39] bnxt: add dev configure operation
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (2 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 04/39] bnxt: add dev infos get operation Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 06/39] bnxt: add vnic functions and structs Stephen Hurd
                   ` (36 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the bnxt_hwrm_port_phy_cfg() HWRM call,
and copies required information into the new struct bnxt_link_info.

v4:
Fixed few issues identified by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt.h                |  32 +++
 drivers/net/bnxt/bnxt_ethdev.c         |  24 ++
 drivers/net/bnxt/bnxt_hwrm.c           | 232 +++++++++++++++-
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 470 +++++++++++++++++++++++++++++++++
 5 files changed, 758 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index f8707b2..bfce91e 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -81,6 +81,29 @@ struct bnxt_pf_info {
 	struct bnxt_vf_info	*vf;
 };
 
+/* Max wait time is 10 * 100ms = 1s */
+#define BNXT_LINK_WAIT_CNT	10
+#define BNXT_LINK_WAIT_INTERVAL	100
+struct bnxt_link_info {
+	uint8_t			phy_flags;
+	uint8_t			mac_type;
+	uint8_t			phy_link_status;
+	uint8_t			loop_back;
+	uint8_t			link_up;
+	uint8_t			duplex;
+	uint8_t			pause;
+	uint8_t			force_pause;
+	uint8_t			auto_pause;
+	uint8_t			auto_mode;
+#define PHY_VER_LEN		3
+	uint8_t			phy_ver[PHY_VER_LEN];
+	uint16_t		link_speed;
+	uint16_t		support_speeds;
+	uint16_t		auto_link_speed;
+	uint16_t		auto_link_speed_mask;
+	uint32_t		preemphasis;
+};
+
 #define BNXT_COS_QUEUE_COUNT	8
 struct bnxt_cos_queue_info {
 	uint8_t	id;
@@ -99,6 +122,14 @@ struct bnxt {
 #define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
 #define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
 
+	unsigned int		rx_nr_rings;
+	unsigned int		rx_cp_nr_rings;
+	struct bnxt_rx_queue **rx_queues;
+
+	unsigned int		tx_nr_rings;
+	unsigned int		tx_cp_nr_rings;
+	struct bnxt_tx_queue **tx_queues;
+
 #define MAX_NUM_MAC_ADDR	32
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
 
@@ -109,6 +140,7 @@ struct bnxt {
 	uint16_t			max_req_len;
 	uint16_t			max_resp_len;
 
+	struct bnxt_link_info	link_info;
 	struct bnxt_cos_queue_info	cos_queue[BNXT_COS_QUEUE_COUNT];
 
 	struct bnxt_pf_info		pf;
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index a8a9912..b46d2ce 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -154,6 +154,29 @@ found:
 	dev_info->vmdq_queue_base = 0;
 }
 
+/* Configure the device based on the configuration provided */
+static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	int rc;
+
+	bp->rx_queues = (void *)eth_dev->data->rx_queues;
+	bp->tx_queues = (void *)eth_dev->data->tx_queues;
+
+	/* Inherit new configurations */
+	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
+	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
+	bp->rx_cp_nr_rings = bp->rx_nr_rings;
+	bp->tx_cp_nr_rings = bp->tx_nr_rings;
+
+	if (eth_dev->data->dev_conf.rxmode.jumbo_frame)
+		eth_dev->data->mtu =
+				eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
+				ETHER_HDR_LEN - ETHER_CRC_LEN - VLAN_TAG_SIZE;
+	rc = bnxt_set_hwrm_link_config(bp, true);
+	return rc;
+}
+
 /*
  * Initialization
  */
@@ -161,6 +184,7 @@ found:
 static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
+	.dev_configure = bnxt_dev_configure_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 8aba8cd..a2d7815 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -83,7 +83,7 @@ static int bnxt_hwrm_send_message_locked(struct bnxt *bp, void *msg,
 		/* Sanity check on the resp->resp_len */
 		rte_rmb();
 		if (resp->resp_len && resp->resp_len <=
-		    bp->max_resp_len) {
+				bp->max_resp_len) {
 			/* Last byte of resp contains the valid key */
 			valid = (uint8_t *)resp + resp->resp_len - 1;
 			if (*valid == HWRM_RESP_VALID_KEY)
@@ -314,6 +314,61 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags)
 	return rc;
 }
 
+static int bnxt_hwrm_port_phy_cfg(struct bnxt *bp, struct bnxt_link_info *conf)
+{
+	int rc = 0;
+	struct hwrm_port_phy_cfg_input req = {.req_type = 0};
+	struct hwrm_port_phy_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, PORT_PHY_CFG, -1, resp);
+
+	req.flags = conf->phy_flags;
+	if (conf->link_up) {
+		req.force_link_speed = conf->link_speed;
+		/*
+		 * Note, ChiMP FW 20.2.1 and 20.2.2 return an error when we set
+		 * any auto mode, even "none".
+		 */
+		if (req.auto_mode == HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE) {
+			req.flags |= HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE;
+		} else {
+			req.auto_mode = conf->auto_mode;
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_MODE;
+			req.auto_link_speed_mask = conf->auto_link_speed_mask;
+			req.enables |=
+			   HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED_MASK;
+			req.auto_link_speed = conf->auto_link_speed;
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED;
+		}
+		req.auto_duplex = conf->duplex;
+		req.enables |= HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_DUPLEX;
+		req.auto_pause = conf->auto_pause;
+		/* Set force_pause if there is no auto or if there is a force */
+		if (req.auto_pause)
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_PAUSE;
+		else
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE;
+		req.force_pause = conf->force_pause;
+		if (req.force_pause)
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE;
+	} else {
+		req.flags &= ~HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG;
+		req.flags |= HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE_LINK_DOWN;
+		req.force_link_speed = 0;
+	}
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
@@ -372,3 +427,178 @@ int bnxt_alloc_hwrm_resources(struct bnxt *bp)
 
 	return 0;
 }
+
+static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
+{
+	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
+
+	if ((conf_link_speed & ETH_LINK_SPEED_FIXED) == ETH_LINK_SPEED_AUTONEG)
+		return HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
+
+	switch (conf_link_speed) {
+	case ETH_LINK_SPEED_10M_HD:
+	case ETH_LINK_SPEED_100M_HD:
+		return HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF;
+	}
+	return hw_link_duplex;
+}
+
+static uint16_t bnxt_parse_eth_link_speed(uint32_t conf_link_speed)
+{
+	uint16_t eth_link_speed = 0;
+
+	if ((conf_link_speed & ETH_LINK_SPEED_FIXED) == ETH_LINK_SPEED_AUTONEG)
+		return ETH_LINK_SPEED_AUTONEG;
+
+	switch (conf_link_speed & ~ETH_LINK_SPEED_FIXED) {
+	case ETH_LINK_SPEED_100M:
+	case ETH_LINK_SPEED_100M_HD:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MB;
+		break;
+	case ETH_LINK_SPEED_1G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB;
+		break;
+	case ETH_LINK_SPEED_2_5G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB;
+		break;
+	case ETH_LINK_SPEED_10G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB;
+		break;
+	case ETH_LINK_SPEED_20G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB;
+		break;
+	case ETH_LINK_SPEED_25G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB;
+		break;
+	case ETH_LINK_SPEED_40G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB;
+		break;
+	case ETH_LINK_SPEED_50G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB;
+		break;
+	default:
+		RTE_LOG(ERR, PMD,
+			"Unsupported link speed %d; default to AUTO\n",
+			conf_link_speed);
+		break;
+	}
+	return eth_link_speed;
+}
+
+#define BNXT_SUPPORTED_SPEEDS (ETH_LINK_SPEED_100M | ETH_LINK_SPEED_100M_HD | \
+		ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G | \
+		ETH_LINK_SPEED_10G | ETH_LINK_SPEED_20G | ETH_LINK_SPEED_25G | \
+		ETH_LINK_SPEED_40G | ETH_LINK_SPEED_50G)
+
+static int bnxt_valid_link_speed(uint32_t link_speed, uint8_t port_id)
+{
+	uint32_t one_speed;
+
+	if (link_speed == ETH_LINK_SPEED_AUTONEG)
+		return 0;
+
+	if (link_speed & ETH_LINK_SPEED_FIXED) {
+		one_speed = link_speed & ~ETH_LINK_SPEED_FIXED;
+
+		if (one_speed & (one_speed - 1)) {
+			RTE_LOG(ERR, PMD,
+				"Invalid advertised speeds (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+		if ((one_speed & BNXT_SUPPORTED_SPEEDS) != one_speed) {
+			RTE_LOG(ERR, PMD,
+				"Unsupported advertised speed (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+	} else {
+		if (!(link_speed & BNXT_SUPPORTED_SPEEDS)) {
+			RTE_LOG(ERR, PMD,
+				"Unsupported advertised speeds (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
+static uint16_t bnxt_parse_eth_link_speed_mask(uint32_t link_speed)
+{
+	uint16_t ret = 0;
+
+	if (link_speed == ETH_LINK_SPEED_AUTONEG)
+		link_speed = BNXT_SUPPORTED_SPEEDS;
+
+	if (link_speed & ETH_LINK_SPEED_100M)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB;
+	if (link_speed & ETH_LINK_SPEED_100M_HD)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB;
+	if (link_speed & ETH_LINK_SPEED_1G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB;
+	if (link_speed & ETH_LINK_SPEED_2_5G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB;
+	if (link_speed & ETH_LINK_SPEED_10G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB;
+	if (link_speed & ETH_LINK_SPEED_20G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB;
+	if (link_speed & ETH_LINK_SPEED_25G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB;
+	if (link_speed & ETH_LINK_SPEED_40G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB;
+	if (link_speed & ETH_LINK_SPEED_50G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB;
+	return ret;
+}
+
+int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up)
+{
+	int rc = 0;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_link_info link_req;
+	uint16_t speed;
+
+	rc = bnxt_valid_link_speed(dev_conf->link_speeds,
+			bp->eth_dev->data->port_id);
+	if (rc)
+		goto error;
+
+	memset(&link_req, 0, sizeof(link_req));
+	speed = bnxt_parse_eth_link_speed(dev_conf->link_speeds);
+	link_req.link_up = link_up;
+	if (speed == 0) {
+		link_req.phy_flags =
+				HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG;
+		link_req.auto_mode =
+				HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_OR_BELOW;
+		link_req.auto_link_speed_mask =
+			bnxt_parse_eth_link_speed_mask(dev_conf->link_speeds);
+		link_req.auto_link_speed =
+				HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_50GB;
+	} else {
+		link_req.auto_mode = HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE;
+		link_req.phy_flags = HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE |
+			HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESET_PHY;
+		link_req.link_speed = speed;
+	}
+	link_req.duplex = bnxt_parse_eth_link_duplex(dev_conf->link_speeds);
+	link_req.auto_pause = bp->link_info.auto_pause;
+	link_req.force_pause = bp->link_info.force_pause;
+
+	rc = bnxt_hwrm_port_phy_cfg(bp, &link_req);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Set link config failed with rc %d\n", rc);
+	}
+
+error:
+	return rc;
+}
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 6f2e445..eef3be6 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -52,5 +52,6 @@ int bnxt_hwrm_ver_get(struct bnxt *bp);
 
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
 
 #endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index be3ed0f..e36774e 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -54,6 +54,7 @@
 #define HWRM_FUNC_QCAPS		(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR	(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR	(UINT32_C(0x1d))
+#define HWRM_PORT_PHY_CFG                                 (UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG	(UINT32_C(0x30))
 
 /*
@@ -345,6 +346,475 @@ struct hwrm_func_qcaps_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_port_phy_cfg */
+/*
+ * Description: This command configures the PHY device for the port. It allows
+ * setting of the most generic settings for the PHY. The HWRM shall complete
+ * this command as soon as PHY settings are configured. They may not be applied
+ * when the command response is provided. A VF driver shall not be allowed to
+ * configure PHY using this command. In a network partition mode, a PF driver
+ * shall not be allowed to configure PHY using this command.
+ */
+
+/* Input (56 bytes) */
+struct hwrm_port_phy_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is set to '1', the PHY for the port shall be reset. #
+	 * If this bit is set to 1, then the HWRM shall reset the PHY after
+	 * applying PHY configuration changes specified in this command. # In
+	 * order to guarantee that PHY configuration changes specified in this
+	 * command take effect, the HWRM client should set this flag to 1. # If
+	 * this bit is not set to 1, then the HWRM may reset the PHY depending
+	 * on the current PHY configuration and settings specified in this
+	 * command.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESET_PHY            UINT32_C(0x1)
+	/*
+	 * When this bit is set to '1', the link shall be forced to be taken
+	 * down. # When this bit is set to '1", all other command input settings
+	 * related to the link speed shall be ignored. Once the link state is
+	 * forced down, it can be explicitly cleared from that state by setting
+	 * this flag to '0'. # If this flag is set to '0', then the link shall
+	 * be cleared from forced down state if the link is in forced down
+	 * state. There may be conditions (e.g. out-of-band or sideband
+	 * configuration changes for the link) outside the scope of the HWRM
+	 * implementation that may clear forced down link state.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE_LINK_DOWN      UINT32_C(0x2)
+	/*
+	 * When this bit is set to '1', the link shall be forced to the
+	 * force_link_speed value. When this bit is set to '1', the HWRM client
+	 * should not enable any of the auto negotiation related fields
+	 * represented by auto_XXX fields in this command. When this bit is set
+	 * to '1' and the HWRM client has enabled a auto_XXX field in this
+	 * command, then the HWRM shall ignore the enabled auto_XXX field. When
+	 * this bit is set to zero, the link shall be allowed to autoneg.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE                UINT32_C(0x4)
+	/*
+	 * When this bit is set to '1', the auto-negotiation process shall be
+	 * restarted on the link.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG      UINT32_C(0x8)
+	/*
+	 * When this bit is set to '1', Energy Efficient Ethernet (EEE) is
+	 * requested to be enabled on this link. If EEE is not supported on this
+	 * port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_ENABLE	UINT32_C(0x10)
+	/*
+	 * When this bit is set to '1', Energy Efficient Ethernet (EEE) is
+	 * requested to be disabled on this link. If EEE is not supported on
+	 * this port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_DISABLE	UINT32_C(0x20)
+	/*
+	 * When this bit is set to '1' and EEE is enabled on this link, then TX
+	 * LPI is requested to be enabled on the link. If EEE is not supported
+	 * on this port, then this flag shall be ignored by the HWRM. If EEE is
+	 * disabled on this port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_TX_LPI	UINT32_C(0x40)
+	uint32_t flags;
+
+	/* This bit must be '1' for the auto_mode field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_MODE          UINT32_C(0x1)
+	/* This bit must be '1' for the auto_duplex field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_DUPLEX        UINT32_C(0x2)
+	/* This bit must be '1' for the auto_pause field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_PAUSE         UINT32_C(0x4)
+	/*
+	 * This bit must be '1' for the auto_link_speed field to be configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED    UINT32_C(0x8)
+	/*
+	 * This bit must be '1' for the auto_link_speed_mask field to be
+	 * configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED_MASK \
+								UINT32_C(0x10)
+	/* This bit must be '1' for the wirespeed field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_WIRESPEED	UINT32_C(0x20)
+	/* This bit must be '1' for the lpbk field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_LPBK		UINT32_C(0x40)
+	/* This bit must be '1' for the preemphasis field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_PREEMPHASIS	UINT32_C(0x80)
+	/* This bit must be '1' for the force_pause field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE	UINT32_C(0x100)
+	/*
+	 * This bit must be '1' for the eee_link_speed_mask field to be
+	 * configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_EEE_LINK_SPEED_MASK \
+								UINT32_C(0x200)
+	/* This bit must be '1' for the tx_lpi_timer field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_TX_LPI_TIMER	UINT32_C(0x400)
+	uint32_t enables;
+
+	/* Port ID of port that is to be configured. */
+	uint16_t port_id;
+
+	/*
+	 * This is the speed that will be used if the force bit is '1'. If
+	 * unsupported speed is selected, an error will be generated.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t force_link_speed;
+
+	/*
+	 * This value is used to identify what autoneg mode is used when the
+	 * link speed is not being forced.
+	 */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE	(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ALL_SPEEDS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_SPEED_MASK \
+							(UINT32_C(0x4) << 0)
+	uint8_t auto_mode;
+
+	/*
+	 * This is the duplex setting that will be used if the autoneg_mode is
+	 * "one_speed" or "one_or_below".
+	 */
+		/* Half Duplex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF \
+							(UINT32_C(0x0) << 0)
+		/* Full duplex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_FULL \
+							(UINT32_C(0x1) << 0)
+		/* Both Half and Full dupex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH \
+							(UINT32_C(0x2) << 0)
+	uint8_t auto_duplex;
+
+	/*
+	 * This value is used to configure the pause that will be used for
+	 * autonegotiation. Add text on the usage of auto_pause and force_pause.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX              UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX              UINT32_C(0x2)
+	/*
+	 * When set to 1, the advertisement of pause is enabled. # When the
+	 * auto_mode is not set to none and this flag is set to 1, then the
+	 * auto_pause bits on this port are being advertised and autoneg pause
+	 * results are being interpreted. # When the auto_mode is not set to
+	 * none and this flag is set to 0, the pause is forced as indicated in
+	 * force_pause, and also advertised as auto_pause bits, but the autoneg
+	 * results are not interpreted since the pause configuration is being
+	 * forced. # When the auto_mode is set to none and this flag is set to
+	 * 1, auto_pause bits should be ignored and should be set to 0.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_AUTONEG_PAUSE   UINT32_C(0x4)
+	uint8_t auto_pause;
+
+	uint8_t unused_0;
+
+	/*
+	 * This is the speed that will be used if the autoneg_mode is
+	 * "one_speed" or "one_or_below". If an unsupported speed is selected,
+	 * an error will be generated.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t auto_link_speed;
+
+	/*
+	 * This is a mask of link speeds that will be used if autoneg_mode is
+	 * "mask". If unsupported speed is enabled an error will be generated.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MB \
+							UINT32_C(0x2000)
+	uint16_t auto_link_speed_mask;
+
+	/* This value controls the wirespeed feature. */
+		/* Wirespeed feature is disabled. */
+	#define HWRM_PORT_PHY_CFG_INPUT_WIRESPEED_OFF	(UINT32_C(0x0) << 0)
+		/* Wirespeed feature is enabled. */
+	#define HWRM_PORT_PHY_CFG_INPUT_WIRESPEED_ON	(UINT32_C(0x1) << 0)
+	uint8_t wirespeed;
+
+	/* This value controls the loopback setting for the PHY. */
+		/* No loopback is selected. Normal operation. */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_NONE	(UINT32_C(0x0) << 0)
+		/*
+		 * The HW will be configured with local loopback such that host
+		 * data is sent back to the host without modification.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_LOCAL	(UINT32_C(0x1) << 0)
+		/*
+		 * The HW will be configured with remote loopback such that port
+		 * logic will send packets back out the transmitter that are
+		 * received.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_REMOTE	(UINT32_C(0x2) << 0)
+	uint8_t lpbk;
+
+	/*
+	 * This value is used to configure the pause that will be used for force
+	 * mode.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX             UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX             UINT32_C(0x2)
+	uint8_t force_pause;
+
+	uint8_t unused_1;
+
+	/*
+	 * This value controls the pre-emphasis to be used for the link. Driver
+	 * should not set this value (use enable.preemphasis = 0) unless driver
+	 * is sure of setting. Normally HWRM FW will determine proper pre-
+	 * emphasis.
+	 */
+	uint32_t preemphasis;
+
+	/*
+	 * Setting for link speed mask that is used to advertise speeds during
+	 * autonegotiation when EEE is enabled. This field is valid only when
+	 * EEE is enabled. The speeds specified in this field shall be a subset
+	 * of speeds specified in auto_link_speed_mask. If EEE is enabled,then
+	 * at least one speed shall be provided in this mask.
+	 */
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD1  UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_100MB  UINT32_C(0x2)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD2  UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_1GB    UINT32_C(0x8)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD3 \
+								UINT32_C(0x10)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD4 \
+								UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_10GB \
+								UINT32_C(0x40)
+	uint16_t eee_link_speed_mask;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * Reuested setting of TX LPI timer in microseconds. This field is valid
+	 * only when EEE is enabled and TX LPI is enabled.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_TX_LPI_TIMER_MASK \
+							UINT32_C(0xffffff)
+	#define HWRM_PORT_PHY_CFG_INPUT_TX_LPI_TIMER_SFT           0
+	uint32_t tx_lpi_timer;
+
+	uint32_t unused_4;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_port_phy_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_ver_get */
 /*
  * Description: This function is called by a driver to determine the HWRM
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 06/39] bnxt: add vnic functions and structs
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (3 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 05/39] bnxt: add dev configure operation Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 07/39] bnxt: declare ring structs and free() func Stephen Hurd
                   ` (35 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add functions to allocate, initialize, and free vnics.

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |  14 ++
 drivers/net/bnxt/bnxt_vnic.c           | 277 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_vnic.h           |  80 ++++++++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h |   5 +-
 5 files changed, 376 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/bnxt/bnxt_vnic.c
 create mode 100644 drivers/net/bnxt/bnxt_vnic.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 9965597..c57afaa 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
 # Export include files
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index bfce91e..d0f84f4 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -45,6 +45,13 @@
 #define BNXT_MAX_MTU		9000
 #define VLAN_TAG_SIZE		4
 
+enum bnxt_hw_context {
+	HW_CONTEXT_NONE     = 0,
+	HW_CONTEXT_IS_RSS   = 1,
+	HW_CONTEXT_IS_COS   = 2,
+	HW_CONTEXT_IS_LB    = 3,
+};
+
 struct bnxt_vf_info {
 	uint16_t		fw_fid;
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
@@ -130,6 +137,13 @@ struct bnxt {
 	unsigned int		tx_cp_nr_rings;
 	struct bnxt_tx_queue **tx_queues;
 
+	struct bnxt_vnic_info	*vnic_info;
+	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
+
+	/* VNIC pointer for flow filter (VMDq) pools */
+#define MAX_FF_POOLS	ETH_64_POOLS
+	STAILQ_HEAD(, bnxt_vnic_info)	ff_pool[MAX_FF_POOLS];
+
 #define MAX_NUM_MAC_ADDR	32
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
 
diff --git a/drivers/net/bnxt/bnxt_vnic.c b/drivers/net/bnxt/bnxt_vnic.c
new file mode 100644
index 0000000..c04c4c7
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_vnic.c
@@ -0,0 +1,277 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_memzone.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * VNIC Functions
+ */
+
+static void prandom_bytes(void *dest_ptr, size_t len)
+{
+	char *dest = (char *)dest_ptr;
+	uint64_t rb;
+
+	while (len) {
+		rb = rte_rand();
+		if (len >= 8) {
+			memcpy(dest, &rb, 8);
+			len -= 8;
+			dest += 8;
+		} else {
+			memcpy(dest, &rb, len);
+			dest += len;
+			len = 0;
+		}
+	}
+}
+
+void bnxt_init_vnics(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	uint16_t max_vnics;
+	int i, j;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	STAILQ_INIT(&bp->free_vnic_list);
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+		vnic->fw_vnic_id = (uint16_t)HWRM_NA_SIGNATURE;
+		vnic->fw_rss_cos_lb_ctx = (uint16_t)HWRM_NA_SIGNATURE;
+		vnic->ctx_is_rss_cos_lb = HW_CONTEXT_NONE;
+
+		for (j = 0; j < MAX_QUEUES_PER_VNIC; j++)
+			vnic->fw_grp_ids[j] = (uint16_t)HWRM_NA_SIGNATURE;
+
+		prandom_bytes(vnic->rss_hash_key, HW_HASH_KEY_SIZE);
+		STAILQ_INIT(&vnic->filter);
+		STAILQ_INSERT_TAIL(&bp->free_vnic_list, vnic, next);
+	}
+	for (i = 0; i < MAX_FF_POOLS; i++)
+		STAILQ_INIT(&bp->ff_pool[i]);
+}
+
+int bnxt_free_vnic(struct bnxt *bp, struct bnxt_vnic_info *vnic,
+			  int pool)
+{
+	struct bnxt_vnic_info *temp;
+
+	temp = STAILQ_FIRST(&bp->ff_pool[pool]);
+	while (temp) {
+		if (temp == vnic) {
+			STAILQ_REMOVE(&bp->ff_pool[pool], vnic,
+				      bnxt_vnic_info, next);
+			vnic->fw_vnic_id = (uint16_t)HWRM_NA_SIGNATURE;
+			STAILQ_INSERT_TAIL(&bp->free_vnic_list, vnic,
+					   next);
+			return 0;
+		}
+		temp = STAILQ_NEXT(temp, next);
+	}
+	RTE_LOG(ERR, PMD, "VNIC %p is not found in pool[%d]\n", vnic, pool);
+	return -EINVAL;
+}
+
+struct bnxt_vnic_info *bnxt_alloc_vnic(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+
+	/* Find the 1st unused vnic from the free_vnic_list pool*/
+	vnic = STAILQ_FIRST(&bp->free_vnic_list);
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "No more free VNIC resources\n");
+		return NULL;
+	}
+	STAILQ_REMOVE_HEAD(&bp->free_vnic_list, next);
+	return vnic;
+}
+
+void bnxt_free_all_vnics(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *temp, *next;
+	int i;
+
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		temp = STAILQ_FIRST(&bp->ff_pool[i]);
+		while (temp) {
+			next = STAILQ_NEXT(temp, next);
+			STAILQ_REMOVE(&bp->ff_pool[i], temp, bnxt_vnic_info,
+				      next);
+			STAILQ_INSERT_TAIL(&bp->free_vnic_list, temp, next);
+			temp = next;
+		}
+	}
+}
+
+void bnxt_free_vnic_attributes(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+
+	STAILQ_FOREACH(vnic, &bp->free_vnic_list, next) {
+		if (vnic->rss_table) {
+			/* 'Unreserve' the rss_table */
+			/* N/A */
+
+			vnic->rss_table = NULL;
+		}
+
+		if (vnic->rss_hash_key) {
+			/* 'Unreserve' the rss_hash_key */
+			/* N/A */
+
+			vnic->rss_hash_key = NULL;
+		}
+	}
+}
+
+int bnxt_alloc_vnic_attributes(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	struct rte_pci_device *pdev = bp->pdev;
+	const struct rte_memzone *mz;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+	int entry_length = RTE_CACHE_LINE_ROUNDUP(
+				HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table) +
+				HW_HASH_KEY_SIZE);
+	uint16_t max_vnics;
+	int i;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
+		 "bnxt_%04x:%02x:%02x:%02x_vnicattr", pdev->addr.domain,
+		 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
+	mz = rte_memzone_lookup(mz_name);
+	if (!mz) {
+		mz = rte_memzone_reserve(mz_name,
+					 entry_length * max_vnics,
+					 SOCKET_ID_ANY,
+					 RTE_MEMZONE_2MB |
+					 RTE_MEMZONE_SIZE_HINT_ONLY);
+		if (!mz)
+			return -ENOMEM;
+	}
+
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+
+		/* Allocate rss table and hash key */
+		vnic->rss_table =
+			(void *)((char *)mz->addr + (entry_length * i));
+		memset(vnic->rss_table, -1, entry_length);
+
+		vnic->rss_table_dma_addr = mz->phys_addr + (entry_length * i);
+		vnic->rss_hash_key = (void *)((char *)vnic->rss_table +
+			     HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table));
+
+		vnic->rss_hash_key_dma_addr = vnic->rss_table_dma_addr +
+			     HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table);
+	}
+
+	return 0;
+}
+
+void bnxt_free_vnic_mem(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	uint16_t max_vnics, i;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+		if (vnic->fw_vnic_id != (uint16_t)HWRM_NA_SIGNATURE) {
+			RTE_LOG(ERR, PMD, "VNIC is not freed yet!\n");
+			/* TODO Call HWRM to free VNIC */
+		}
+	}
+
+	rte_free(bp->vnic_info);
+	bp->vnic_info = NULL;
+}
+
+int bnxt_alloc_vnic_mem(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic_mem;
+	uint16_t max_vnics;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	/* Allocate memory for VNIC pool and filter pool */
+	vnic_mem = rte_zmalloc("bnxt_vnic_info",
+			       max_vnics * sizeof(struct bnxt_vnic_info), 0);
+	if (vnic_mem == NULL) {
+		RTE_LOG(ERR, PMD, "Failed to alloc memory for %d VNICs",
+			max_vnics);
+		return -ENOMEM;
+	}
+	bp->vnic_info = vnic_mem;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_vnic.h b/drivers/net/bnxt/bnxt_vnic.h
new file mode 100644
index 0000000..9671ba4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_vnic.h
@@ -0,0 +1,80 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_VNIC_H_
+#define _BNXT_VNIC_H_
+
+#include <sys/queue.h>
+#include <stdbool.h>
+
+struct bnxt_vnic_info {
+	STAILQ_ENTRY(bnxt_vnic_info)	next;
+	uint8_t		ff_pool_idx;
+
+	uint16_t	fw_vnic_id; /* returned by Chimp during alloc */
+	uint16_t	fw_rss_cos_lb_ctx;
+	uint16_t	ctx_is_rss_cos_lb;
+#define MAX_NUM_TRAFFIC_CLASSES		8
+#define MAX_NUM_RSS_QUEUES_PER_VNIC	16
+#define MAX_QUEUES_PER_VNIC	(MAX_NUM_RSS_QUEUES_PER_VNIC + \
+				 MAX_NUM_TRAFFIC_CLASSES)
+	uint16_t	start_grp_id;
+	uint16_t	end_grp_id;
+	uint16_t	fw_grp_ids[MAX_QUEUES_PER_VNIC];
+	uint16_t	hash_type;
+	phys_addr_t	rss_table_dma_addr;
+	uint16_t	*rss_table;
+	phys_addr_t	rss_hash_key_dma_addr;
+	void		*rss_hash_key;
+	uint32_t	flags;
+#define BNXT_VNIC_INFO_PROMISC			(1 << 0)
+#define BNXT_VNIC_INFO_ALLMULTI			(1 << 1)
+
+	bool		vlan_strip;
+	bool		func_default;
+
+	STAILQ_HEAD(, bnxt_filter_info)	filter;
+};
+
+struct bnxt;
+void bnxt_init_vnics(struct bnxt *bp);
+int bnxt_free_vnic(struct bnxt *bp, struct bnxt_vnic_info *vnic,
+			  int pool);
+struct bnxt_vnic_info *bnxt_alloc_vnic(struct bnxt *bp);
+void bnxt_free_all_vnics(struct bnxt *bp);
+void bnxt_free_vnic_attributes(struct bnxt *bp);
+int bnxt_alloc_vnic_attributes(struct bnxt *bp);
+void bnxt_free_vnic_mem(struct bnxt *bp);
+int bnxt_alloc_vnic_mem(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index e36774e..3c032e0 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -43,8 +43,11 @@
  * Following is the signature for HWRM message field that indicates not
  * applicable (All F's). Need to cast it the size of the field if needed.
  */
+#define HWRM_NA_SIGNATURE        ((uint32_t)(-1))
 #define HWRM_MAX_REQ_LEN	(128)  /* hwrm_func_buf_rgtr */
 #define HWRM_MAX_RESP_LEN	(176)  /* hwrm_func_qstats */
+#define HW_HASH_INDEX_SIZE      0x80    /* 7 bit indirection table index. */
+#define HW_HASH_KEY_SIZE        40
 #define HWRM_RESP_VALID_KEY	1 /* valid key for HWRM response */
 
 /*
@@ -54,7 +57,7 @@
 #define HWRM_FUNC_QCAPS		(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR	(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR	(UINT32_C(0x1d))
-#define HWRM_PORT_PHY_CFG                                 (UINT32_C(0x20))
+#define HWRM_PORT_PHY_CFG	(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG	(UINT32_C(0x30))
 
 /*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 07/39] bnxt: declare ring structs and free() func
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (4 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 06/39] bnxt: add vnic functions and structs Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 08/39] bnxt: add completion ring support Stephen Hurd
                   ` (34 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Declare ring structures and a ring free() function.
These are generic ring mamagement functions which will be used to create
Tx, Rx and Completion rings in the subsequent patches.

v4:
Address checkpatch warnings.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile    |  1 +
 drivers/net/bnxt/bnxt_ring.c | 47 ++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ring.h | 92 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 140 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_ring.c
 create mode 100644 drivers/net/bnxt/bnxt_ring.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index c57afaa..757ea62 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
new file mode 100644
index 0000000..d3b70cc
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -0,0 +1,47 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "bnxt.h"
+#include "bnxt_ring.h"
+
+/*
+ * Generic ring handling
+ */
+
+void bnxt_free_ring(struct bnxt_ring_struct *ring)
+{
+	if (ring->vmem_size && *ring->vmem) {
+		memset((char *)*ring->vmem, 0, ring->vmem_size);
+		*ring->vmem = NULL;
+	}
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
new file mode 100644
index 0000000..ebbd759
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -0,0 +1,92 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RING_H_
+#define _BNXT_RING_H_
+
+#include <inttypes.h>
+
+#include <rte_memory.h>
+
+#define RING_NEXT(ring, idx)		(((idx) + 1) & (ring)->ring_mask)
+
+#define RTE_MBUF_DATA_DMA_ADDR(mb) \
+	((uint64_t)((mb)->buf_physaddr + (mb)->data_off))
+
+#define DB_IDX_MASK						0xffffff
+#define DB_IDX_VALID						(0x1 << 26)
+#define DB_IRQ_DIS						(0x1 << 27)
+#define DB_KEY_TX						(0x0 << 28)
+#define DB_KEY_RX						(0x1 << 28)
+#define DB_KEY_CP						(0x2 << 28)
+#define DB_KEY_ST						(0x3 << 28)
+#define DB_KEY_TX_PUSH						(0x4 << 28)
+#define DB_LONG_TX_PUSH						(0x2 << 24)
+
+#define DEFAULT_CP_RING_SIZE	256
+#define DEFAULT_RX_RING_SIZE	256
+#define DEFAULT_TX_RING_SIZE	256
+
+#define MAX_TPA		128
+
+/* These assume 4k pages */
+#define MAX_RX_DESC_CNT (8 * 1024)
+#define MAX_TX_DESC_CNT (4 * 1024)
+#define MAX_CP_DESC_CNT (16 * 1024)
+
+#define INVALID_HW_RING_ID      ((uint16_t)-1)
+
+struct bnxt_ring_struct {
+	void			*bd;
+	phys_addr_t		bd_dma;
+	uint32_t		ring_size;
+	uint32_t		ring_mask;
+
+	int			vmem_size;
+	void			**vmem;
+
+	uint16_t		fw_ring_id; /* Ring id filled by Chimp FW */
+};
+
+struct bnxt_ring_grp_info {
+	uint16_t	fw_stats_ctx;
+	uint16_t	fw_grp_id;
+	uint16_t	rx_fw_ring_id;
+	uint16_t	cp_fw_ring_id;
+	uint16_t	ag_fw_ring_id;
+};
+
+struct bnxt;
+void bnxt_free_ring(struct bnxt_ring_struct *ring);
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 08/39] bnxt: add completion ring support
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (5 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 07/39] bnxt: declare ring structs and free() func Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 09/39] bnxt: add L2 filter alloc/init/free Stephen Hurd
                   ` (33 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Structures, macros, and functions for working with completion rings
in the driver.

Completion Ring is used by the Ethernet controller to provide the
status of transmitted & received packets, report errors, status changes
to the host software.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   5 +
 drivers/net/bnxt/bnxt_cpr.c            | 140 +++++++++++++++++++
 drivers/net/bnxt/bnxt_cpr.h            |  88 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.c           |  18 +++
 drivers/net/bnxt/bnxt_hwrm.h           |   2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 239 ++++++++++++++++++++++++++++++++-
 7 files changed, 487 insertions(+), 6 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_cpr.c
 create mode 100644 drivers/net/bnxt/bnxt_cpr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 757ea62..afd1690 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -48,6 +48,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 # all source are stored in SRCS-y
 #
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index d0f84f4..bdd355f 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -42,6 +42,8 @@
 #include <rte_lcore.h>
 #include <rte_spinlock.h>
 
+#include "bnxt_cpr.h"
+
 #define BNXT_MAX_MTU		9000
 #define VLAN_TAG_SIZE		4
 
@@ -137,6 +139,9 @@ struct bnxt {
 	unsigned int		tx_cp_nr_rings;
 	struct bnxt_tx_queue **tx_queues;
 
+	/* Default completion ring */
+	struct bnxt_cp_ring_info	*def_cp_ring;
+
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
new file mode 100644
index 0000000..ba8c0d4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -0,0 +1,140 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
+
+/*
+ * Async event handling
+ */
+void bnxt_handle_async_event(struct bnxt *bp __rte_unused,
+			     struct cmpl_base *cmp)
+{
+	struct hwrm_async_event_cmpl *async_cmp =
+				(struct hwrm_async_event_cmpl *)cmp;
+
+	/* TODO: HWRM async events are not defined yet */
+	/* Needs to handle: link events, error events, etc. */
+	switch (async_cmp->event_id) {
+	case 0:
+		/* Assume LINK_CHANGE == 0 */
+		RTE_LOG(INFO, PMD, "Link change event\n");
+
+		/* Can just prompt the update_op routine to do a qcfg
+		 * instead of doing the actual qcfg
+		 */
+		break;
+	case 1:
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "handle_async_event id = 0x%x\n",
+			async_cmp->event_id);
+		break;
+	}
+}
+
+void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmpl)
+{
+	struct hwrm_fwd_req_cmpl *fwd_cmpl = (struct hwrm_fwd_req_cmpl *)cmpl;
+	struct input *fwd_cmd;
+	uint16_t logical_vf_id, error_code;
+
+	/* Qualify the fwd request */
+	if (fwd_cmpl->source_id < bp->pf.first_vf_id) {
+		RTE_LOG(ERR, PMD,
+			"FWD req's source_id 0x%x > first_vf_id 0x%x\n",
+			fwd_cmpl->source_id, bp->pf.first_vf_id);
+		error_code = HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED;
+		goto reject;
+	} else if (fwd_cmpl->req_len_type >> HWRM_FWD_REQ_CMPL_REQ_LEN_SFT >
+		   128 - sizeof(struct input)) {
+		RTE_LOG(ERR, PMD,
+		    "FWD req's cmd len 0x%x > 108 bytes allowed\n",
+		    fwd_cmpl->req_len_type >> HWRM_FWD_REQ_CMPL_REQ_LEN_SFT);
+		error_code = HWRM_ERR_CODE_INVALID_PARAMS;
+		goto reject;
+	}
+
+	/* Locate VF's forwarded command */
+	logical_vf_id = fwd_cmpl->source_id - bp->pf.first_vf_id;
+	fwd_cmd = (struct input *)((uint8_t *)bp->pf.vf_req_buf +
+		   (logical_vf_id * 128));
+
+	/* Provision the request */
+	switch (fwd_cmd->req_type) {
+	case HWRM_CFA_L2_FILTER_ALLOC:
+	case HWRM_CFA_L2_FILTER_FREE:
+	case HWRM_CFA_L2_FILTER_CFG:
+	case HWRM_CFA_L2_SET_RX_MASK:
+		break;
+	default:
+		error_code = HWRM_ERR_CODE_INVALID_PARAMS;
+		goto reject;
+	}
+
+	/* Forward */
+	fwd_cmd->target_id = fwd_cmpl->source_id;
+	bnxt_hwrm_exec_fwd_resp(bp, fwd_cmd);
+	return;
+
+reject:
+	/* TODO: Encap the reject error resp into the hwrm_err_iput? */
+	/* Use the error_code for the reject cmd */
+	RTE_LOG(ERR, PMD,
+		"Error 0x%x found in the forward request\n", error_code);
+}
+
+/* For the default completion ring only */
+void bnxt_free_def_cp_ring(struct bnxt *bp)
+{
+	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+	struct bnxt_ring_struct *ring = cpr->cp_ring_struct;
+
+	bnxt_free_ring(ring);
+}
+
+/* For the default completion ring only */
+void bnxt_init_def_ring_struct(struct bnxt *bp)
+{
+	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+	struct bnxt_ring_struct *ring = cpr->cp_ring_struct;
+
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->ring_size = rte_align32pow2(DEFAULT_CP_RING_SIZE);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
new file mode 100644
index 0000000..878c7c9
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -0,0 +1,88 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_CPR_H_
+#define _BNXT_CPR_H_
+
+#include "hsi_struct_def_dpdk.h"
+
+#define CMP_VALID(cmp, raw_cons, ring)					\
+	(!!(((struct cmpl_base *)(cmp))->info3_v & CMPL_BASE_V) ==	\
+	 !((raw_cons) & ((ring)->ring_size)))
+
+#define CMP_TYPE(cmp)						\
+	(((struct cmpl_base *)cmp)->type & CMPL_BASE_TYPE_MASK)
+
+#define ADV_RAW_CMP(idx, n)	((idx) + (n))
+#define NEXT_RAW_CMP(idx)	ADV_RAW_CMP(idx, 1)
+#define RING_CMP(ring, idx)	((idx) & (ring)->ring_mask)
+#define NEXT_CMP(idx)		RING_CMP(ADV_RAW_CMP(idx, 1))
+
+#define DB_CP_REARM_FLAGS	(DB_KEY_CP | DB_IDX_VALID)
+#define DB_CP_FLAGS		(DB_KEY_CP | DB_IDX_VALID | DB_IRQ_DIS)
+
+#define B_CP_DB_REARM(cpr, raw_cons)					\
+		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_REARM_FLAGS | \
+				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+
+#define B_CP_DIS_DB(cpr, raw_cons)					\
+		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
+				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+
+struct bnxt_ring_struct;
+struct bnxt_cp_ring_info {
+	uint32_t		cp_raw_cons;
+	void			*cp_doorbell;
+
+	struct cmpl_base	*cp_desc_ring;
+
+	phys_addr_t		cp_desc_mapping;
+
+	struct ctx_hw_stats	*hw_stats;
+	phys_addr_t		hw_stats_map;
+	uint32_t		hw_stats_ctx_id;
+
+	struct bnxt_ring_struct	*cp_ring_struct;
+};
+
+#define RX_CMP_L2_ERRORS						\
+	(RX_PKT_CMPL_ERRORS_BUFFER_ERROR_MASK | RX_PKT_CMPL_ERRORS_CRC_ERROR)
+
+
+struct bnxt;
+void bnxt_free_def_cp_ring(struct bnxt *bp);
+void bnxt_init_def_ring_struct(struct bnxt *bp);
+void bnxt_handle_async_event(struct bnxt *bp, struct cmpl_base *cmp);
+void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmp);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index a2d7815..f591ead 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -135,6 +135,24 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
+{
+	int rc;
+	struct hwrm_exec_fwd_resp_input req = {.req_type = 0 };
+	struct hwrm_exec_fwd_resp_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, EXEC_FWD_RESP, -1, resp);
+
+	memcpy(req.encap_request, fwd_cmd,
+	       sizeof(req.encap_request));
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index eef3be6..b792313 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,6 +41,8 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
+
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 3c032e0..9efb68b 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -53,12 +53,143 @@
 /*
  * Request types
  */
-#define HWRM_VER_GET		(UINT32_C(0x0))
-#define HWRM_FUNC_QCAPS		(UINT32_C(0x15))
-#define HWRM_FUNC_DRV_UNRGTR	(UINT32_C(0x1a))
-#define HWRM_FUNC_DRV_RGTR	(UINT32_C(0x1d))
-#define HWRM_PORT_PHY_CFG	(UINT32_C(0x20))
-#define HWRM_QUEUE_QPORTCFG	(UINT32_C(0x30))
+#define HWRM_VER_GET			(UINT32_C(0x0))
+#define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
+#define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
+#define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
+#define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
+#define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+#define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
+#define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
+#define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
+#define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
+
+/* Return Codes */
+#define HWRM_ERR_CODE_INVALID_PARAMS                      (UINT32_C(0x2))
+#define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED              (UINT32_C(0x3))
+
+/* HWRM Forwarded Request (16 bytes) */
+struct hwrm_fwd_req_cmpl {
+	/* Length of forwarded request in bytes. */
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define HWRM_FWD_REQ_CMPL_TYPE_MASK		UINT32_C(0x3f)
+	#define HWRM_FWD_REQ_CMPL_TYPE_SFT		0
+		/* Forwarded HWRM Request */
+	#define HWRM_FWD_REQ_CMPL_TYPE_HWRM_FWD_REQ	(UINT32_C(0x22) << 0)
+	/* Length of forwarded request in bytes. */
+	#define HWRM_FWD_REQ_CMPL_REQ_LEN_MASK		UINT32_C(0xffc0)
+	#define HWRM_FWD_REQ_CMPL_REQ_LEN_SFT		6
+	uint16_t req_len_type;
+
+	/*
+	 * Source ID of this request. Typically used in forwarding requests and
+	 * responses. 0x0 - 0xFFF8 - Used for function ids 0xFFF8 - 0xFFFE -
+	 * Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t source_id;
+
+	uint32_t unused_0;
+
+	/* Address of forwarded request. */
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define HWRM_FWD_REQ_CMPL_V			UINT32_C(0x1)
+	/* Address of forwarded request. */
+	#define HWRM_FWD_REQ_CMPL_REQ_BUF_ADDR_MASK	UINT32_C(0xfffffffe)
+	#define HWRM_FWD_REQ_CMPL_REQ_BUF_ADDR_SFT	1
+	uint64_t req_buf_addr_v;
+} __attribute__((packed));
+
+/* HWRM Asynchronous Event Completion Record (16 bytes) */
+struct hwrm_async_event_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_MASK		UINT32_C(0x3f)
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_SFT		0
+		/* HWRM Asynchronous Event Information */
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_HWRM_ASYNC_EVENT \
+							(UINT32_C(0x2e) << 0)
+	uint16_t type;
+
+	/* Identifiers of events. */
+		/* Link status changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_STATUS_CHANGE \
+							(UINT32_C(0x0) << 0)
+		/* Link MTU changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_MTU_CHANGE \
+							(UINT32_C(0x1) << 0)
+		/* Link speed changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_SPEED_CHANGE \
+							(UINT32_C(0x2) << 0)
+		/* DCB Configuration changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_DCB_CONFIG_CHANGE \
+							(UINT32_C(0x3) << 0)
+		/* Port connection not allowed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PORT_CONN_NOT_ALLOWED \
+							(UINT32_C(0x4) << 0)
+		/* Link speed configuration was not allowed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_SPEED_CFG_NOT_ALLOWED \
+							(UINT32_C(0x5) << 0)
+		/* Function driver unloaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_FUNC_DRVR_UNLOAD \
+							(UINT32_C(0x10) << 0)
+		/* Function driver loaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_FUNC_DRVR_LOAD \
+							(UINT32_C(0x11) << 0)
+		/* PF driver unloaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_DRVR_UNLOAD \
+							(UINT32_C(0x20) << 0)
+		/* PF driver loaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_DRVR_LOAD \
+							(UINT32_C(0x21) << 0)
+		/* VF Function Level Reset (FLR) */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_VF_FLR	(UINT32_C(0x30) << 0)
+		/* VF MAC Address Change */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_VF_MAC_ADDR_CHANGE \
+							(UINT32_C(0x31) << 0)
+		/* PF-VF communication channel status change. */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_VF_COMM_STATUS_CHANGE \
+							(UINT32_C(0x32) << 0)
+		/* HWRM Error */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_HWRM_ERROR \
+							(UINT32_C(0xff) << 0)
+	uint16_t event_id;
+
+	/* Event specific data */
+	uint32_t event_data2;
+
+	/* opaque is 7 b */
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define HWRM_ASYNC_EVENT_CMPL_V				UINT32_C(0x1)
+	/* opaque is 7 b */
+	#define HWRM_ASYNC_EVENT_CMPL_OPAQUE_MASK		UINT32_C(0xfe)
+	#define HWRM_ASYNC_EVENT_CMPL_OPAQUE_SFT		1
+	uint8_t opaque_v;
+
+	/* 8-lsb timestamp from POR (100-msec resolution) */
+	uint8_t timestamp_lo;
+
+	/* 16-lsb timestamp from POR (100-msec resolution) */
+	uint16_t timestamp_hi;
+
+	/* Event specific data */
+	uint32_t event_data1;
+} __attribute__((packed));
 
 /*
  * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
@@ -122,6 +253,102 @@ struct output {
 	uint16_t resp_len;
 } __attribute__((packed));
 
+/* hwrm_exec_fwd_resp */
+/*
+ * Description: This command is used to send an encapsulated request to the
+ * HWRM. This command instructs the HWRM to execute the request and forward the
+ * response of the encapsulated request to the location specified in the
+ * original request that is encapsulated. The target id of this command shall be
+ * set to 0xFFFF (HWRM). The response location in this command shall be used to
+ * acknowledge the receipt of the encapsulated request and forwarding of the
+ * response.
+ */
+
+/* Input (128 bytes) */
+struct hwrm_exec_fwd_resp_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This is an encapsulated request. This request should be executed by
+	 * the HWRM and the response should be provided in the response buffer
+	 * inside the encapsulated request.
+	 */
+	uint32_t encap_request[26];
+
+	/*
+	 * This value indicates the target id of the response to the
+	 * encapsulated request. 0x0 - 0xFFF8 - Used for function ids 0xFFF8 -
+	 * 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t encap_resp_target_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_exec_fwd_resp_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_func_qcaps */
 /*
  * Description: This command returns capabilities of a function. The input FID
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 09/39] bnxt: add L2 filter alloc/init/free
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (6 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 08/39] bnxt: add completion ring support Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 10/39] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
                   ` (32 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add the L2 filter structure and the alloc/init/free functions for
dealing with them.

A filter is used to identify traffic that contains a matching set of
parameters like unicast or broadcast MAC address or a VLAN tag amongst
other things which then allows the ASIC to direct the  incoming traffic
to an appropriate VNIC or Rx ring.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   3 +
 drivers/net/bnxt/bnxt_filter.c         | 175 +++++++++++++
 drivers/net/bnxt/bnxt_filter.h         |  74 ++++++
 drivers/net/bnxt/bnxt_hwrm.c           |  21 ++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 456 +++++++++++++++++++++++++++++++++
 7 files changed, 733 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_filter.c
 create mode 100644 drivers/net/bnxt/bnxt_filter.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index afd1690..b7834b1 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index bdd355f..49aa38b 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -145,6 +145,9 @@ struct bnxt {
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
+	struct bnxt_filter_info	*filter_info;
+	STAILQ_HEAD(, bnxt_filter_info)	free_filter_list;
+
 	/* VNIC pointer for flow filter (VMDq) pools */
 #define MAX_FF_POOLS	ETH_64_POOLS
 	STAILQ_HEAD(, bnxt_vnic_info)	ff_pool[MAX_FF_POOLS];
diff --git a/drivers/net/bnxt/bnxt_filter.c b/drivers/net/bnxt/bnxt_filter.c
new file mode 100644
index 0000000..f03a1dc
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_filter.c
@@ -0,0 +1,175 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <sys/queue.h>
+
+#include <rte_log.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * Filter Functions
+ */
+
+struct bnxt_filter_info *bnxt_alloc_filter(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+
+	/* Find the 1st unused filter from the free_filter_list pool*/
+	filter = STAILQ_FIRST(&bp->free_filter_list);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "No more free filter resources\n");
+		return NULL;
+	}
+	STAILQ_REMOVE_HEAD(&bp->free_filter_list, next);
+
+	/* Default to L2 MAC Addr filter */
+	filter->flags = HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX;
+	filter->enables = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
+			HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK;
+	memcpy(filter->l2_addr, bp->eth_dev->data->mac_addrs->addr_bytes,
+	       ETHER_ADDR_LEN);
+	memset(filter->l2_addr_mask, 0xff, ETHER_ADDR_LEN);
+	return filter;
+}
+
+void bnxt_init_filters(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+	int i, max_filters;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	STAILQ_INIT(&bp->free_filter_list);
+	for (i = 0; i < max_filters; i++) {
+		filter = &bp->filter_info[i];
+		filter->fw_l2_filter_id = -1;
+		STAILQ_INSERT_TAIL(&bp->free_filter_list, filter, next);
+	}
+}
+
+void bnxt_free_all_filters(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter, *temp_filter;
+	int i;
+
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			filter = STAILQ_FIRST(&vnic->filter);
+			while (filter) {
+				temp_filter = STAILQ_NEXT(filter, next);
+				STAILQ_REMOVE(&vnic->filter, filter,
+					      bnxt_filter_info, next);
+				STAILQ_INSERT_TAIL(&bp->free_filter_list,
+						   filter, next);
+				filter = temp_filter;
+			}
+			STAILQ_INIT(&vnic->filter);
+		}
+	}
+}
+
+void bnxt_free_filter_mem(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+	uint16_t max_filters, i;
+	int rc = 0;
+
+	/* Ensure that all filters are freed */
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	for (i = 0; i < max_filters; i++) {
+		filter = &bp->filter_info[i];
+		if (filter->fw_l2_filter_id != ((uint64_t)-1)) {
+			RTE_LOG(ERR, PMD, "HWRM filter is not freed??\n");
+			/* Call HWRM to try to free filter again */
+			rc = bnxt_hwrm_clear_filter(bp, filter);
+			if (rc)
+				RTE_LOG(ERR, PMD,
+				       "HWRM filter cannot be freed rc = %d\n",
+					rc);
+		}
+		filter->fw_l2_filter_id = -1;
+	}
+	STAILQ_INIT(&bp->free_filter_list);
+
+	rte_free(bp->filter_info);
+	bp->filter_info = NULL;
+}
+
+int bnxt_alloc_filter_mem(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter_mem;
+	uint16_t max_filters;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	/* Allocate memory for VNIC pool and filter pool */
+	filter_mem = rte_zmalloc("bnxt_filter_info",
+				 max_filters * sizeof(struct bnxt_filter_info),
+				 0);
+	if (filter_mem == NULL) {
+		RTE_LOG(ERR, PMD, "Failed to alloc memory for %d filters",
+			max_filters);
+		return -ENOMEM;
+	}
+	bp->filter_info = filter_mem;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_filter.h b/drivers/net/bnxt/bnxt_filter.h
new file mode 100644
index 0000000..06fe134
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_filter.h
@@ -0,0 +1,74 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_FILTER_H_
+#define _BNXT_FILTER_H_
+
+#include <rte_ether.h>
+
+struct bnxt;
+struct bnxt_filter_info {
+	STAILQ_ENTRY(bnxt_filter_info)	next;
+	uint64_t		fw_l2_filter_id;
+#define INVALID_MAC_INDEX	((uint16_t)-1)
+	uint16_t		mac_index;
+
+	/* Filter Characteristics */
+	uint32_t		flags;
+	uint32_t		enables;
+	uint8_t			l2_addr[ETHER_ADDR_LEN];
+	uint8_t			l2_addr_mask[ETHER_ADDR_LEN];
+	uint16_t		l2_ovlan;
+	uint16_t		l2_ovlan_mask;
+	uint16_t		l2_ivlan;
+	uint16_t		l2_ivlan_mask;
+	uint8_t			t_l2_addr[ETHER_ADDR_LEN];
+	uint8_t			t_l2_addr_mask[ETHER_ADDR_LEN];
+	uint16_t		t_l2_ovlan;
+	uint16_t		t_l2_ovlan_mask;
+	uint16_t		t_l2_ivlan;
+	uint16_t		t_l2_ivlan_mask;
+	uint8_t			tunnel_type;
+	uint16_t		mirror_vnic_id;
+	uint32_t		vni;
+	uint8_t			pri_hint;
+	uint64_t		l2_filter_id_hint;
+};
+
+struct bnxt_filter_info *bnxt_alloc_filter(struct bnxt *bp);
+void bnxt_init_filters(struct bnxt *bp);
+void bnxt_free_all_filters(struct bnxt *bp);
+void bnxt_free_filter_mem(struct bnxt *bp);
+int bnxt_alloc_filter_mem(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index f591ead..5d9a991 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -39,6 +39,7 @@
 #include <rte_version.h>
 
 #include "bnxt.h"
+#include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -135,6 +136,26 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_clear_filter(struct bnxt *bp,
+			   struct bnxt_filter_info *filter)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_filter_free_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_filter_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, CFA_L2_FILTER_FREE, -1, resp);
+
+	req.l2_filter_id = rte_cpu_to_le_64(filter->fw_l2_filter_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	filter->fw_l2_filter_id = -1;
+
+	return 0;
+}
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
 {
 	int rc;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index b792313..c48ba3f 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,6 +41,9 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_clear_filter(struct bnxt *bp,
+			   struct bnxt_filter_info *filter);
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 9efb68b..ca42a87 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -253,6 +253,462 @@ struct output {
 	uint16_t resp_len;
 } __attribute__((packed));
 
+/* hwrm_cfa_l2_filter_alloc */
+/*
+ * A filter is used to identify traffic that contains a matching set of
+ * parameters like unicast or broadcast MAC address or a VLAN tag amongst
+ * other things which then allows the ASIC to direct the  incoming traffic
+ * to an appropriate VNIC or Rx ring.
+ */
+
+/* Input (96 bytes) */
+struct hwrm_cfa_l2_filter_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Enumeration denoting the RX, TX type of the resource. This
+	 * enumeration is used for resources that are similar for both TX and RX
+	 * paths of the chip.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH \
+							UINT32_C(0x1)
+		/* tx path */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_TX \
+							(UINT32_C(0x0) << 0)
+		/* rx path */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX \
+							(UINT32_C(0x1) << 0)
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_LAST \
+				HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX
+	/*
+	 * Setting of this flag indicates the applicability to the loopback
+	 * path.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_LOOPBACK \
+							UINT32_C(0x2)
+	/*
+	 * Setting of this flag indicates drop action. If this flag is not set,
+	 * then it should be considered accept action.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_DROP \
+							UINT32_C(0x4)
+	/*
+	 * If this flag is set, all t_l2_* fields are invalid and they should
+	 * not be specified. If this flag is set, then l2_* fields refer to
+	 * fields of outermost L2 header.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST \
+							UINT32_C(0x8)
+	uint32_t flags;
+
+	/* This bit must be '1' for the l2_addr field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR \
+							UINT32_C(0x1)
+	/* This bit must be '1' for the l2_addr_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK \
+							UINT32_C(0x2)
+	/* This bit must be '1' for the l2_ovlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN \
+							UINT32_C(0x4)
+	/* This bit must be '1' for the l2_ovlan_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN_MASK \
+							UINT32_C(0x8)
+	/* This bit must be '1' for the l2_ivlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN \
+							UINT32_C(0x10)
+	/* This bit must be '1' for the l2_ivlan_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK \
+							UINT32_C(0x20)
+	/* This bit must be '1' for the t_l2_addr field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_ADDR \
+							UINT32_C(0x40)
+	/*
+	 * This bit must be '1' for the t_l2_addr_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_ADDR_MASK \
+							UINT32_C(0x80)
+	/* This bit must be '1' for the t_l2_ovlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_OVLAN \
+							UINT32_C(0x100)
+	/*
+	 * This bit must be '1' for the t_l2_ovlan_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_OVLAN_MASK \
+							UINT32_C(0x200)
+	/* This bit must be '1' for the t_l2_ivlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_IVLAN \
+							UINT32_C(0x400)
+	/*
+	 * This bit must be '1' for the t_l2_ivlan_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_IVLAN_MASK \
+							UINT32_C(0x800)
+	/* This bit must be '1' for the src_type field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_SRC_TYPE \
+							UINT32_C(0x1000)
+	/* This bit must be '1' for the src_id field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_SRC_ID \
+							UINT32_C(0x2000)
+	/* This bit must be '1' for the tunnel_type field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE \
+							UINT32_C(0x4000)
+	/* This bit must be '1' for the dst_id field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_DST_ID \
+							UINT32_C(0x8000)
+	/*
+	 * This bit must be '1' for the mirror_vnic_id field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_MIRROR_VNIC_ID \
+							UINT32_C(0x10000)
+	uint32_t enables;
+
+	/*
+	 * This value sets the match value for the L2 MAC address. Destination
+	 * MAC address for RX path. Source MAC address for TX path.
+	 */
+	uint8_t l2_addr[6];
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+
+	/*
+	 * This value sets the mask value for the L2 address. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint8_t l2_addr_mask[6];
+
+	/* This value sets VLAN ID value for outer VLAN. */
+	uint16_t l2_ovlan;
+
+	/*
+	 * This value sets the mask value for the ovlan id. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint16_t l2_ovlan_mask;
+
+	/* This value sets VLAN ID value for inner VLAN. */
+	uint16_t l2_ivlan;
+
+	/*
+	 * This value sets the mask value for the ivlan id. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint16_t l2_ivlan_mask;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This value sets the match value for the tunnel L2 MAC address.
+	 * Destination MAC address for RX path. Source MAC address for TX path.
+	 */
+	uint8_t t_l2_addr[6];
+
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/*
+	 * This value sets the mask value for the tunnel L2 address. A value of
+	 * 0 will mask the corresponding bit from compare.
+	 */
+	uint8_t t_l2_addr_mask[6];
+
+	/* This value sets VLAN ID value for tunnel outer VLAN. */
+	uint16_t t_l2_ovlan;
+
+	/*
+	 * This value sets the mask value for the tunnel ovlan id. A value of 0
+	 * will mask the corresponding bit from compare.
+	 */
+	uint16_t t_l2_ovlan_mask;
+
+	/* This value sets VLAN ID value for tunnel inner VLAN. */
+	uint16_t t_l2_ivlan;
+
+	/*
+	 * This value sets the mask value for the tunnel ivlan id. A value of 0
+	 * will mask the corresponding bit from compare.
+	 */
+	uint16_t t_l2_ivlan_mask;
+
+	/* This value identifies the type of source of the packet. */
+		/* Network port */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_NPORT \
+							(UINT32_C(0x0) << 0)
+		/* Physical function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_PF \
+							(UINT32_C(0x1) << 0)
+		/* Virtual function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_VF \
+							(UINT32_C(0x2) << 0)
+		/* Virtual NIC of a function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_VNIC \
+							(UINT32_C(0x3) << 0)
+		/* Embedded processor for CFA management */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_KONG \
+							(UINT32_C(0x4) << 0)
+		/* Embedded processor for OOB management */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_APE \
+							(UINT32_C(0x5) << 0)
+		/* Embedded processor for RoCE */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_BONO \
+							(UINT32_C(0x6) << 0)
+		/* Embedded processor for network proxy functions */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_TANG \
+							(UINT32_C(0x7) << 0)
+	uint8_t src_type;
+
+	uint8_t unused_6;
+	/*
+	 * This value is the id of the source. For a network port, it represents
+	 * port_id. For a physical function, it represents fid. For a virtual
+	 * function, it represents vf_id. For a vnic, it represents vnic_id. For
+	 * embedded processors, this id is not valid. Notes: 1. The function ID
+	 * is implied if it src_id is not provided for a src_type that is either
+	 */
+	uint32_t src_id;
+
+	/* Tunnel Type. */
+		/* Non-tunnel */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_NONTUNNEL \
+							(UINT32_C(0x0) << 0)
+		/* Virtual eXtensible Local Area Network (VXLAN) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_VXLAN \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Network Virtualization Generic Routing Encapsulation (NVGRE)
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_NVGRE \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Generic Routing Encapsulation (GRE) inside Ethernet payload
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_L2GRE \
+							(UINT32_C(0x3) << 0)
+		/* IP in IP */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_IPIP \
+							(UINT32_C(0x4) << 0)
+		/* Generic Network Virtualization Encapsulation (Geneve) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_GENEVE \
+							(UINT32_C(0x5) << 0)
+		/* Multi-Protocol Lable Switching (MPLS) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_MPLS \
+							(UINT32_C(0x6) << 0)
+		/* Stateless Transport Tunnel (STT) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_STT \
+							(UINT32_C(0x7) << 0)
+		/*
+		 * Generic Routing Encapsulation (GRE) inside IP datagram
+		 * payload
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_IPGRE \
+							(UINT32_C(0x8) << 0)
+		/* Any tunneled traffic */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_ANYTUNNEL \
+							(UINT32_C(0xff) << 0)
+	uint8_t tunnel_type;
+
+	uint8_t unused_7;
+
+	/*
+	 * If set, this value shall represent the Logical VNIC ID of the
+	 * destination VNIC for the RX path and network port id of the
+	 * destination port for the TX path.
+	 */
+	uint16_t dst_id;
+
+	/* Logical VNIC ID of the VNIC where traffic is mirrored. */
+	uint16_t mirror_vnic_id;
+
+	/*
+	 * This hint is provided to help in placing the filter in the filter
+	 * table.
+	 */
+		/* No preference */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_NO_PREFER \
+							(UINT32_C(0x0) << 0)
+		/* Above the given filter */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_ABOVE_FILTER \
+							(UINT32_C(0x1) << 0)
+		/* Below the given filter */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_BELOW_FILTER \
+							(UINT32_C(0x2) << 0)
+		/* As high as possible */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_MAX \
+							(UINT32_C(0x3) << 0)
+		/* As low as possible */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_MIN \
+							(UINT32_C(0x4) << 0)
+	uint8_t pri_hint;
+
+	uint8_t unused_8;
+	uint32_t unused_9;
+
+	/*
+	 * This is the ID of the filter that goes along with the pri_hint. This
+	 * field is valid only for the following values. 1 - Above the given
+	 * filter 2 - Below the given filter
+	 */
+	uint64_t l2_filter_id_hint;
+} __attribute__((packed));
+
+/* Output (24 bytes) */
+struct hwrm_cfa_l2_filter_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This value identifies a set of CFA data structures used for an L2
+	 * context.
+	 */
+	uint64_t l2_filter_id;
+
+	/*
+	 * This is the ID of the flow associated with this filter. This value
+	 * shall be used to match and associate the flow identifier returned in
+	 * completion records. A value of 0xFFFFFFFF shall indicate no flow id.
+	 */
+	uint32_t flow_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_cfa_l2_filter_free */
+/*
+ * Description: Free a L2 filter. The HWRM shall free all associated filter
+ * resources with the L2 filter.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_cfa_l2_filter_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This value identifies a set of CFA data structures used for an L2
+	 * context.
+	 */
+	uint64_t l2_filter_id;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_cfa_l2_filter_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_exec_fwd_resp */
 /*
  * Description: This command is used to send an encapsulated request to the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 10/39] bnxt: add Tx queue operations (nonfunctional)
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (7 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 09/39] bnxt: add L2 filter alloc/init/free Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 11/39] bnxt: add Rx queue create/destroy operations Stephen Hurd
                   ` (31 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add code to create/destroy TX queues. This still requires support to
create a TX ring in the ASIC which will be completed in a future commit.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile      |   1 +
 drivers/net/bnxt/bnxt_ethdev.c |   3 +
 drivers/net/bnxt/bnxt_txq.c    | 125 +++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_txq.h    |  75 +++++++++++++++++++++++++
 4 files changed, 204 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_txq.c
 create mode 100644 drivers/net/bnxt/bnxt_txq.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index b7834b1..13a90b9 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -53,6 +53,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b46d2ce..77a6d92 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -41,6 +41,7 @@
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -185,6 +186,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.tx_queue_setup = bnxt_tx_queue_setup_op,
+	.tx_queue_release = bnxt_tx_queue_release_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
new file mode 100644
index 0000000..cb3dd8e
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -0,0 +1,125 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_ring.h"
+#include "bnxt_txq.h"
+
+/*
+ * TX Queues
+ */
+
+void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+
+	/* 'Unreserve' rte_memzone */
+	/* N/A */
+
+	if (cpr->hw_stats)
+		cpr->hw_stats = NULL;
+}
+
+static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq __rte_unused)
+{
+	/* TODO: Requires interaction with TX ring */
+}
+
+void bnxt_free_tx_mbufs(struct bnxt *bp)
+{
+	struct bnxt_tx_queue *txq;
+	int i;
+
+	for (i = 0; i < (int)bp->tx_nr_rings; i++) {
+		txq = bp->tx_queues[i];
+		bnxt_tx_queue_release_mbufs(txq);
+	}
+}
+
+void bnxt_tx_queue_release_op(void *tx_queue)
+{
+	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
+
+	if (txq) {
+		/* TODO: Free ring and stats here */
+		rte_free(txq);
+	}
+}
+
+int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_txconf *tx_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_tx_queue *txq;
+
+	if (!nb_desc || nb_desc > MAX_TX_DESC_CNT) {
+		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
+		return -EINVAL;
+	}
+
+	if (eth_dev->data->tx_queues) {
+		txq = eth_dev->data->tx_queues[queue_idx];
+		if (txq) {
+			bnxt_tx_queue_release_op(txq);
+			txq = NULL;
+		}
+	}
+	txq = rte_zmalloc_socket("bnxt_tx_queue", sizeof(struct bnxt_tx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq == NULL) {
+		RTE_LOG(ERR, PMD, "bnxt_tx_queue allocation failed!");
+		return -ENOMEM;
+	}
+	txq->bp = bp;
+	txq->nb_tx_desc = nb_desc;
+	txq->tx_free_thresh = tx_conf->tx_free_thresh;
+
+	/* TODO: Initialize ring structure */
+
+	txq->queue_id = queue_idx;
+	txq->port_id = eth_dev->data->port_id;
+
+	/* TODO: Allocate TX ring hardware descriptors */
+
+	/* TODO: Initialize the ring */
+
+	eth_dev->data->tx_queues[queue_idx] = txq;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_txq.h b/drivers/net/bnxt/bnxt_txq.h
new file mode 100644
index 0000000..16f3a0b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txq.h
@@ -0,0 +1,75 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_TXQ_H_
+#define _BNXT_TXQ_H_
+
+struct bnxt_tx_ring_info;
+struct bnxt_cp_ring_info;
+struct bnxt_tx_queue {
+	uint16_t		nb_tx_desc;    /* number of TX descriptors */
+	uint16_t		tx_free_thresh;/* minimum TX before freeing */
+	/** Index to last TX descriptor to have been cleaned. */
+	uint16_t		last_desc_cleaned;
+	/** Total number of TX descriptors ready to be allocated. */
+	uint16_t		tx_next_dd; /* next desc to scan for DD bit */
+	uint16_t		tx_next_rs; /* next desc to set RS bit */
+	uint16_t		queue_id; /* TX queue index */
+	uint16_t		reg_idx; /* TX queue register index */
+	uint8_t			port_id; /* Device port identifier */
+	uint8_t			pthresh; /* Prefetch threshold register */
+	uint8_t			hthresh; /* Host threshold register */
+	uint8_t			wthresh; /* Write-back threshold reg */
+	uint32_t		txq_flags; /* Holds flags for this TXq */
+	uint32_t		ctx_curr; /* Hardware context states */
+	uint8_t			tx_deferred_start; /* not in global dev start */
+
+	struct bnxt		*bp;
+	int			index;
+	int			tx_wake_thresh;
+	struct bnxt_tx_ring_info	*tx_ring;
+
+	unsigned int		cp_nr_rings;
+	struct bnxt_cp_ring_info	*cp_ring;
+};
+
+void bnxt_free_txq_stats(struct bnxt_tx_queue *txq);
+void bnxt_free_tx_mbufs(struct bnxt *bp);
+void bnxt_tx_queue_release_op(void *tx_queue);
+int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_txconf *tx_conf);
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 11/39] bnxt: add Rx queue create/destroy operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (8 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 10/39] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 12/39] bnxt: Add statistics operations Stephen Hurd
                   ` (30 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Adds initial create/destroy queue code. Still requires RX ring support
which will be brought in subsequent patches to be functional.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   2 +
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +
 drivers/net/bnxt/bnxt_rxq.c            | 288 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_rxq.h            |  74 +++++++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 121 ++++++++++++++
 6 files changed, 489 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_rxq.c
 create mode 100644 drivers/net/bnxt/bnxt_rxq.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 13a90b9..21ed71c 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -53,6 +53,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index 49aa38b..f7cf9d1 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -142,6 +142,8 @@ struct bnxt {
 	/* Default completion ring */
 	struct bnxt_cp_ring_info	*def_cp_ring;
 
+	unsigned int		nr_vnics;
+
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 77a6d92..7e7d1ab 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -41,6 +41,7 @@
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
 #include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
@@ -186,6 +187,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.rx_queue_setup = bnxt_rx_queue_setup_op,
+	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
 };
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
new file mode 100644
index 0000000..11e9466
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -0,0 +1,288 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
+#include "bnxt_rxq.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * RX Queues
+ */
+
+void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+
+	/* 'Unreserve' rte_memzone */
+	/* N/A */
+
+	if (cpr->hw_stats)
+		cpr->hw_stats = NULL;
+}
+
+int bnxt_mq_rx_configure(struct bnxt *bp)
+{
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	unsigned int i, j, nb_q_per_grp, ring_idx;
+	int start_grp_id, end_grp_id, rc = 0;
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter;
+	struct bnxt_rx_queue *rxq;
+
+	bp->nr_vnics = 0;
+
+	/* Single queue mode */
+	if (bp->rx_cp_nr_rings < 2) {
+		vnic = bnxt_alloc_vnic(bp);
+		if (!vnic) {
+			RTE_LOG(ERR, PMD, "VNIC alloc failed\n");
+			rc = -ENOMEM;
+			goto err_out;
+		}
+		STAILQ_INSERT_TAIL(&bp->ff_pool[0], vnic, next);
+		bp->nr_vnics++;
+
+		rxq = bp->eth_dev->data->rx_queues[0];
+		rxq->vnic = vnic;
+
+		vnic->func_default = true;
+		vnic->ff_pool_idx = 0;
+		vnic->start_grp_id = 1;
+		vnic->end_grp_id = vnic->start_grp_id +
+				   bp->rx_cp_nr_rings - 1;
+		filter = bnxt_alloc_filter(bp);
+		if (!filter) {
+			RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+			rc = -ENOMEM;
+			goto err_out;
+		}
+		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+		goto out;
+	}
+
+	/* Multi-queue mode */
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG) {
+		/* VMDq ONLY, VMDq+RSS, VMDq+DCB, VMDq+DCB+RSS */
+		enum rte_eth_nb_pools pools;
+
+		switch (dev_conf->rxmode.mq_mode) {
+		case ETH_MQ_RX_VMDQ_RSS:
+		case ETH_MQ_RX_VMDQ_ONLY:
+			{
+				const struct rte_eth_vmdq_rx_conf *conf =
+				    &dev_conf->rx_adv_conf.vmdq_rx_conf;
+
+				/* ETH_8/64_POOLs */
+				pools = conf->nb_queue_pools;
+				break;
+			}
+		default:
+			RTE_LOG(ERR, PMD, "Unsupported mq_mod %d\n",
+				dev_conf->rxmode.mq_mode);
+			rc = -EINVAL;
+			goto err_out;
+		}
+		/* For each pool, allocate MACVLAN CFA rule & VNIC */
+		if (!pools) {
+			RTE_LOG(ERR, PMD,
+				"VMDq pool not set, defaulted to 64\n");
+			pools = ETH_64_POOLS;
+		}
+		nb_q_per_grp = bp->rx_cp_nr_rings / pools;
+		start_grp_id = 1;
+		end_grp_id = start_grp_id + nb_q_per_grp - 1;
+
+		ring_idx = 0;
+		for (i = 0; i < pools; i++) {
+			vnic = bnxt_alloc_vnic(bp);
+			if (!vnic) {
+				RTE_LOG(ERR, PMD,
+					"VNIC alloc failed\n");
+				rc = -ENOMEM;
+				goto err_out;
+			}
+			STAILQ_INSERT_TAIL(&bp->ff_pool[i], vnic, next);
+			bp->nr_vnics++;
+
+			for (j = 0; j < nb_q_per_grp; j++, ring_idx++) {
+				rxq = bp->eth_dev->data->rx_queues[ring_idx];
+				rxq->vnic = vnic;
+			}
+			if (i == 0)
+				vnic->func_default = true;
+			vnic->ff_pool_idx = i;
+			vnic->start_grp_id = start_grp_id;
+			vnic->end_grp_id = end_grp_id;
+
+			filter = bnxt_alloc_filter(bp);
+			if (!filter) {
+				RTE_LOG(ERR, PMD,
+					"L2 filter alloc failed\n");
+				rc = -ENOMEM;
+				goto err_out;
+			}
+			/*
+			 * TODO: Configure & associate CFA rule for
+			 * each VNIC for each VMDq with MACVLAN, MACVLAN+TC
+			 */
+			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+
+			start_grp_id = end_grp_id + 1;
+			end_grp_id += nb_q_per_grp;
+		}
+		goto out;
+	}
+
+	/* Non-VMDq mode - RSS, DCB, RSS+DCB */
+	/* Init default VNIC for RSS or DCB only */
+	vnic = bnxt_alloc_vnic(bp);
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "VNIC alloc failed\n");
+		rc = -ENOMEM;
+		goto err_out;
+	}
+	/* Partition the rx queues for the single pool */
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		rxq = bp->eth_dev->data->rx_queues[i];
+		rxq->vnic = vnic;
+	}
+	STAILQ_INSERT_TAIL(&bp->ff_pool[0], vnic, next);
+	bp->nr_vnics++;
+
+	vnic->func_default = true;
+	vnic->ff_pool_idx = 0;
+	vnic->start_grp_id = 1;
+	vnic->end_grp_id = vnic->start_grp_id +
+			   bp->rx_cp_nr_rings - 1;
+	filter = bnxt_alloc_filter(bp);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+		rc = -ENOMEM;
+		goto err_out;
+	}
+	STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
+		vnic->hash_type =
+			HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4 |
+			HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+
+out:
+	return rc;
+
+err_out:
+	/* Free allocated vnic/filters */
+
+	return rc;
+}
+
+static void bnxt_rx_queue_release_mbufs(struct bnxt_rx_queue *rxq __rte_unused)
+{
+	/* TODO: Requires interaction with TX ring */
+}
+
+void bnxt_free_rx_mbufs(struct bnxt *bp)
+{
+	struct bnxt_rx_queue *rxq;
+	int i;
+
+	for (i = 0; i < (int)bp->rx_nr_rings; i++) {
+		rxq = bp->rx_queues[i];
+		bnxt_rx_queue_release_mbufs(rxq);
+	}
+}
+
+void bnxt_rx_queue_release_op(void *rx_queue)
+{
+	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
+
+	if (rxq) {
+		bnxt_rx_queue_release_mbufs(rxq);
+
+		/* TODO: Free ring and stats here */
+
+		rte_free(rxq);
+	}
+}
+
+int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_rxconf *rx_conf,
+			       struct rte_mempool *mp)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_rx_queue *rxq;
+
+	if (!nb_desc || nb_desc > MAX_RX_DESC_CNT) {
+		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
+		return -EINVAL;
+	}
+
+	if (eth_dev->data->rx_queues) {
+		rxq = eth_dev->data->rx_queues[queue_idx];
+		if (rxq)
+			bnxt_rx_queue_release_op(rxq);
+	}
+	rxq = rte_zmalloc_socket("bnxt_rx_queue", sizeof(struct bnxt_rx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (!rxq) {
+		RTE_LOG(ERR, PMD, "bnxt_rx_queue allocation failed!");
+		return -ENOMEM;
+	}
+	rxq->bp = bp;
+	rxq->mb_pool = mp;
+	rxq->nb_rx_desc = nb_desc;
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+
+	/* TODO: Initialize ring structure */
+
+	rxq->queue_id = queue_idx;
+	rxq->port_id = eth_dev->data->port_id;
+	rxq->crc_len = (uint8_t)((eth_dev->data->dev_conf.rxmode.hw_strip_crc) ?
+				0 : ETHER_CRC_LEN);
+
+	eth_dev->data->rx_queues[queue_idx] = rxq;
+	/* TODO: Allocate RX ring hardware descriptors */
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_rxq.h b/drivers/net/bnxt/bnxt_rxq.h
new file mode 100644
index 0000000..9554329
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxq.h
@@ -0,0 +1,74 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RQX_H_
+#define _BNXT_RQX_H_
+
+struct bnxt;
+struct bnxt_rx_ring_info;
+struct bnxt_cp_ring_info;
+struct bnxt_rx_queue {
+	struct rte_mempool	*mb_pool; /* mbuf pool for RX ring */
+	struct rte_mbuf		*pkt_first_seg; /* 1st seg of pkt */
+	struct rte_mbuf		*pkt_last_seg; /* Last seg of pkt */
+	uint64_t		mbuf_initializer; /* val to init mbuf */
+	uint16_t		nb_rx_desc; /* num of RX desc */
+	uint16_t		rx_tail; /* cur val of RDT register */
+	uint16_t		nb_rx_hold; /* num held free RX desc */
+	uint16_t		rx_free_thresh; /* max free RX desc to hold */
+	uint16_t		queue_id; /* RX queue index */
+	uint16_t		reg_idx; /* RX queue register index */
+	uint8_t			port_id; /* Device port identifier */
+	uint8_t			crc_len; /* 0 if CRC stripped, 4 otherwise */
+
+	struct bnxt		*bp;
+	struct bnxt_vnic_info	*vnic;
+
+	uint32_t			rx_buf_size;
+	uint32_t			rx_buf_use_size;  /* useable size */
+	struct bnxt_rx_ring_info	*rx_ring;
+	struct bnxt_cp_ring_info	*cp_ring;
+};
+
+void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq);
+int bnxt_mq_rx_configure(struct bnxt *bp);
+void bnxt_rx_queue_release_op(void *rx_queue);
+int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_rxconf *rx_conf,
+			       struct rte_mempool *mp);
+void bnxt_free_rx_mbufs(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ca42a87..bbec5bb 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1897,6 +1897,127 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_vnic_rss_cfg */
+/* Description: This function is used to enable RSS configuration. */
+
+/* Input (48 bytes) */
+struct hwrm_vnic_rss_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over source and
+	 * destination IPv4 addresses of IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4		UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv4 addresses and source/destination ports of
+	 * TCP/IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv4 addresses and source/destination ports of
+	 * UDP/IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4	UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over source and
+	 * destination IPv4 addresses of IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6		UINT32_C(0x8)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv6 addresses and source/destination ports of
+	 * TCP/IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6	UINT32_C(0x10)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv6 addresses and source/destination ports of
+	 * UDP/IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6	UINT32_C(0x20)
+	uint32_t hash_type;
+
+	uint32_t unused_0;
+
+	/* This is the address for rss ring group table */
+	uint64_t ring_grp_tbl_addr;
+
+	/* This is the address for rss hash key table */
+	uint64_t hash_key_tbl_addr;
+
+	/* Index to the rss indirection table. */
+	uint16_t rss_ctx_idx;
+
+	uint16_t unused_1[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_rss_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* Output (32 bytes) */
 struct hwrm_queue_qportcfg_output {
 	/*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 12/39] bnxt: Add statistics operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (9 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 11/39] bnxt: add Rx queue create/destroy operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 13/39] bnxt: initial Tx code implementation Stephen Hurd
                   ` (29 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add get and clear staitstics operations and the asociated HWRM calls.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_cpr.c            |   1 +
 drivers/net/bnxt/bnxt_cpr.h            |   2 -
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +
 drivers/net/bnxt/bnxt_hwrm.c           |  49 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   8 +-
 drivers/net/bnxt/bnxt_rxq.c            |   1 +
 drivers/net/bnxt/bnxt_stats.c          | 142 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_stats.h          |  44 ++++++++++
 drivers/net/bnxt/bnxt_txq.c            |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 107 +++++++++++++++++++++++++
 11 files changed, 355 insertions(+), 4 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_stats.c
 create mode 100644 drivers/net/bnxt/bnxt_stats.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 21ed71c..f6a04f8 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -54,6 +54,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index ba8c0d4..06a618a 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -35,6 +35,7 @@
 #include "bnxt_cpr.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
+#include "hsi_struct_def_dpdk.h"
 
 /*
  * Async event handling
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index 878c7c9..e6333fc 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -34,8 +34,6 @@
 #ifndef _BNXT_CPR_H_
 #define _BNXT_CPR_H_
 
-#include "hsi_struct_def_dpdk.h"
-
 #define CMP_VALID(cmp, raw_cons, ring)					\
 	(!!(((struct cmpl_base *)(cmp))->info3_v & CMPL_BASE_V) ==	\
 	 !((raw_cons) & ((ring)->ring_size)))
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 7e7d1ab..3453509 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -42,6 +42,7 @@
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_stats.h"
 #include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
@@ -187,6 +188,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.stats_get = bnxt_stats_get_op,
+	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 5d9a991..2574bd0 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -39,8 +39,11 @@
 #include <rte_version.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
+#include "bnxt_txq.h"
 #include "hsi_struct_def_dpdk.h"
 
 #define HWRM_CMD_TIMEOUT		2000
@@ -436,10 +439,56 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
+{
+	int rc = 0;
+	struct hwrm_stat_ctx_clr_stats_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_clr_stats_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_CLR_STATS, -1, resp);
+
+	if (cpr->hw_stats_ctx_id == (uint32_t)HWRM_NA_SIGNATURE)
+		return rc;
+
+	req.stat_ctx_id = rte_cpu_to_le_16(cpr->hw_stats_ctx_id);
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
 
+int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq;
+		struct bnxt_rx_queue *rxq;
+		struct bnxt_cp_ring_info *cpr;
+
+		if (i >= bp->rx_cp_nr_rings) {
+			txq = bp->tx_queues[i - bp->rx_cp_nr_rings];
+			cpr = txq->cp_ring;
+		} else {
+			rxq = bp->rx_queues[i];
+			cpr = rxq->cp_ring;
+		}
+
+		rc = bnxt_hwrm_stat_clear(bp, cpr);
+		if (rc)
+			return rc;
+	}
+	return 0;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index c48ba3f..0861417 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -37,10 +37,11 @@
 #include <inttypes.h>
 #include <stdbool.h>
 
-#include "bnxt.h"
-
 #define HWRM_SEQ_ID_INVALID -1U
 
+struct bnxt;
+struct bnxt_filter_info;
+struct bnxt_cp_ring_info;
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter);
 
@@ -53,8 +54,11 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
+int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
+
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index 11e9466..1ae6f66 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -36,6 +36,7 @@
 #include <rte_malloc.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
diff --git a/drivers/net/bnxt/bnxt_stats.c b/drivers/net/bnxt/bnxt_stats.c
new file mode 100644
index 0000000..6f1c760
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_stats.c
@@ -0,0 +1,142 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_byteorder.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
+#include "bnxt_stats.h"
+#include "bnxt_txq.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * Statistics functions
+ */
+
+void bnxt_free_stats(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+
+		bnxt_free_txq_stats(txq);
+	}
+	for (i = 0; i < (int)bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+
+		bnxt_free_rxq_stats(rxq);
+	}
+}
+
+void bnxt_stats_get_op(struct rte_eth_dev *eth_dev,
+			   struct rte_eth_stats *bnxt_stats)
+{
+	unsigned int i;
+	struct bnxt *bp = eth_dev->data->dev_private;
+
+	memset(bnxt_stats, 0, sizeof(*bnxt_stats));
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		struct ctx_hw_stats64 *hw_stats =
+		    (struct ctx_hw_stats64 *)cpr->hw_stats;
+
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_ucast_pkts);
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_mcast_pkts);
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_bcast_pkts);
+
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_ucast_bytes);
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_mcast_bytes);
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_bcast_bytes);
+
+		/*
+		 * TBD: No clear mapping to this... we don't seem
+		 * to have a stat specifically for dropped due to
+		 * insufficient mbufs.
+		 */
+		bnxt_stats->q_errors[i] = 0;
+
+		/* These get replaced once the *_QSTATS commands work */
+		bnxt_stats->ipackets += bnxt_stats->q_ipackets[i];
+		bnxt_stats->ibytes += bnxt_stats->q_ibytes[i];
+		bnxt_stats->imissed += bnxt_stats->q_errors[i];
+		bnxt_stats->ierrors +=
+				rte_le_to_cpu_64(hw_stats->rx_err_pkts);
+	}
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		struct ctx_hw_stats64 *hw_stats =
+		    (struct ctx_hw_stats64 *)cpr->hw_stats;
+
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_ucast_pkts);
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_mcast_pkts);
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_bcast_pkts);
+
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_ucast_bytes);
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_mcast_bytes);
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_bcast_bytes);
+
+		/* These get replaced once the *_QSTATS commands work */
+		bnxt_stats->opackets += bnxt_stats->q_opackets[i];
+		bnxt_stats->obytes +=  bnxt_stats->q_obytes[i];
+		bnxt_stats->oerrors += rte_le_to_cpu_64(hw_stats->tx_drop_pkts);
+		bnxt_stats->oerrors += rte_le_to_cpu_64(hw_stats->tx_err_pkts);
+	}
+}
+
+void bnxt_stats_reset_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	bnxt_clear_all_hwrm_stat_ctxs(bp);
+}
diff --git a/drivers/net/bnxt/bnxt_stats.h b/drivers/net/bnxt/bnxt_stats.h
new file mode 100644
index 0000000..65408a4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_stats.h
@@ -0,0 +1,44 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_STATS_H_
+#define _BNXT_STATS_H_
+
+#include <rte_ethdev.h>
+
+void bnxt_free_stats(struct bnxt *bp);
+void bnxt_stats_get_op(struct rte_eth_dev *eth_dev,
+			   struct rte_eth_stats *bnxt_stats);
+void bnxt_stats_reset_op(struct rte_eth_dev *eth_dev);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index cb3dd8e..a3648c2 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -36,6 +36,7 @@
 #include <rte_malloc.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index bbec5bb..91a83d1 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -34,6 +34,35 @@
 #ifndef _HSI_STRUCT_DEF_EXTERNAL_H_
 #define _HSI_STRUCT_DEF_EXTERNAL_H_
 
+/*
+ * per-context HW statistics -- chip view
+ */
+
+struct ctx_hw_stats64 {
+	uint64_t rx_ucast_pkts;
+	uint64_t rx_mcast_pkts;
+	uint64_t rx_bcast_pkts;
+	uint64_t rx_drop_pkts;
+	uint64_t rx_err_pkts;
+	uint64_t rx_ucast_bytes;
+	uint64_t rx_mcast_bytes;
+	uint64_t rx_bcast_bytes;
+
+	uint64_t tx_ucast_pkts;
+	uint64_t tx_mcast_pkts;
+	uint64_t tx_bcast_pkts;
+	uint64_t tx_drop_pkts;
+	uint64_t tx_err_pkts;
+	uint64_t tx_ucast_bytes;
+	uint64_t tx_mcast_bytes;
+	uint64_t tx_bcast_bytes;
+
+	uint64_t tpa_pkts;
+	uint64_t tpa_bytes;
+	uint64_t tpa_events;
+	uint64_t tpa_aborts;
+} ctx_hw_stats64_t;
+
 /* HW Resource Manager Specification 1.2.0 */
 #define HWRM_VERSION_MAJOR	1
 #define HWRM_VERSION_MINOR	2
@@ -63,6 +92,7 @@
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
 /* Return Codes */
@@ -1897,6 +1927,83 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_clr_stats */
+/* Description: This command clears statistics of a context. */
+
+/* Input (24 bytes) */
+struct hwrm_stat_ctx_clr_stats_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* ID of the statistics context that is being queried. */
+	uint32_t stat_ctx_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_stat_ctx_clr_stats_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 13/39] bnxt: initial Tx code implementation
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (10 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 12/39] bnxt: Add statistics operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 14/39] bnxt: initial Rx " Stephen Hurd
                   ` (28 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Initial implementation of tx_pkt_burst for transmit.
Add code to allocate rings to bnxt_ring.c
This allows creation of rings in ASIC, which is used by the Tx function.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_cpr.h            |   4 +-
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +-
 drivers/net/bnxt/bnxt_ring.c           | 145 ++++++++++
 drivers/net/bnxt/bnxt_ring.h           |   8 +
 drivers/net/bnxt/bnxt_txq.c            |  42 ++-
 drivers/net/bnxt/bnxt_txr.c            | 314 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_txr.h            |  71 +++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 512 +++++++++++++++++++++++++++++++++
 9 files changed, 1091 insertions(+), 9 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_txr.c
 create mode 100644 drivers/net/bnxt/bnxt_txr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index f6a04f8..0785681 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -56,6 +56,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index e6333fc..f104281 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -51,11 +51,11 @@
 
 #define B_CP_DB_REARM(cpr, raw_cons)					\
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_REARM_FLAGS | \
-				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
 #define B_CP_DIS_DB(cpr, raw_cons)					\
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
-				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
 struct bnxt_ring_struct;
 struct bnxt_cp_ring_info {
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3453509..4ace543 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -44,6 +44,7 @@
 #include "bnxt_rxq.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -269,7 +270,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	}
 	eth_dev->dev_ops = &bnxt_dev_ops;
 	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
-	/* eth_dev->tx_pkt_burst = &bnxt_xmit_pkts; */
+	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
 
 	rc = bnxt_alloc_hwrm_resources(bp);
 	if (rc) {
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index d3b70cc..be77bbe 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -31,8 +31,14 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_memzone.h>
+
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_ring.h"
+#include "bnxt_txr.h"
+
+#include "hsi_struct_def_dpdk.h"
 
 /*
  * Generic ring handling
@@ -45,3 +51,142 @@ void bnxt_free_ring(struct bnxt_ring_struct *ring)
 		*ring->vmem = NULL;
 	}
 }
+
+/*
+ * Allocates a completion ring with vmem and stats optionally also allocating
+ * a TX and/or RX ring.  Passing NULL as tx_ring_info and/or rx_ring_info
+ * to not allocate them.
+ *
+ * Order in the allocation is:
+ * stats - Always non-zero length
+ * cp vmem - Always zero-length, supported for the bnxt_ring_struct abstraction
+ * tx vmem - Only non-zero length if tx_ring_info is not NULL
+ * rx vmem - Only non-zero length if rx_ring_info is not NULL
+ * cp bd ring - Always non-zero length
+ * tx bd ring - Only non-zero length if tx_ring_info is not NULL
+ * rx bd ring - Only non-zero length if rx_ring_info is not NULL
+ */
+int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
+			    struct bnxt_tx_ring_info *tx_ring_info,
+			    struct bnxt_rx_ring_info *rx_ring_info,
+			    struct bnxt_cp_ring_info *cp_ring_info,
+			    const char *suffix)
+{
+	struct bnxt_ring_struct *cp_ring = cp_ring_info->cp_ring_struct;
+	struct bnxt_ring_struct *tx_ring;
+	/* TODO: RX ring */
+	/* struct bnxt_ring_struct *rx_ring; */
+	struct rte_pci_device *pdev = bp->pdev;
+	const struct rte_memzone *mz = NULL;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+
+	int stats_len = (tx_ring_info || rx_ring_info) ?
+	    RTE_CACHE_LINE_ROUNDUP(sizeof(struct ctx_hw_stats64)) : 0;
+
+	int cp_vmem_start = stats_len;
+	int cp_vmem_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->vmem_size);
+
+	int tx_vmem_start = cp_vmem_start + cp_vmem_len;
+	int tx_vmem_len =
+	    tx_ring_info ? RTE_CACHE_LINE_ROUNDUP(tx_ring_info->
+						tx_ring_struct->vmem_size) : 0;
+
+	int rx_vmem_start = tx_vmem_start + tx_vmem_len;
+	/* TODO: RX ring */
+	int rx_vmem_len = 0;
+	/*
+	 * rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
+	 * rx_ring_struct->vmem_size) : 0;
+	 */
+
+	int cp_ring_start = rx_vmem_start + rx_vmem_len;
+	int cp_ring_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->ring_size *
+						 sizeof(struct cmpl_base));
+
+	int tx_ring_start = cp_ring_start + cp_ring_len;
+	int tx_ring_len = tx_ring_info ?
+	    RTE_CACHE_LINE_ROUNDUP(tx_ring_info->tx_ring_struct->ring_size *
+				   sizeof(struct tx_bd_long)) : 0;
+
+	int rx_ring_start = tx_ring_start + tx_ring_len;
+	/* TODO: RX ring */
+	int rx_ring_len = 0;
+	/*
+	 * rx_ring_info ?
+	 * RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
+	 * sizeof(struct rx_prod_pkt_bd)) : 0;
+	 */
+
+	int total_alloc_len = rx_ring_start + rx_ring_len;
+
+	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
+		 "bnxt_%04x:%02x:%02x:%02x-%04x_%s", pdev->addr.domain,
+		 pdev->addr.bus, pdev->addr.devid, pdev->addr.function, qidx,
+		 suffix);
+	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
+	mz = rte_memzone_lookup(mz_name);
+	if (!mz) {
+		mz = rte_memzone_reserve(mz_name, total_alloc_len,
+					 SOCKET_ID_ANY,
+					 RTE_MEMZONE_2MB |
+					 RTE_MEMZONE_SIZE_HINT_ONLY);
+		if (mz == NULL)
+			return -ENOMEM;
+	}
+	memset(mz->addr, 0, mz->len);
+
+	if (tx_ring_info) {
+		tx_ring = tx_ring_info->tx_ring_struct;
+
+		tx_ring->bd = ((char *)mz->addr + tx_ring_start);
+		tx_ring_info->tx_desc_ring = (struct tx_bd_long *)tx_ring->bd;
+		tx_ring->bd_dma = mz->phys_addr + tx_ring_start;
+		tx_ring_info->tx_desc_mapping = tx_ring->bd_dma;
+
+		if (!tx_ring->bd)
+			return -ENOMEM;
+		if (tx_ring->vmem_size) {
+			tx_ring->vmem =
+			    (void **)((char *)mz->addr + tx_vmem_start);
+			tx_ring_info->tx_buf_ring =
+			    (struct bnxt_sw_tx_bd *)tx_ring->vmem;
+		}
+	}
+
+/*
+ *	if (rx_ring_info) {
+ *		rx_ring = &rx_ring_info->rx_ring_struct;
+ *
+ *		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
+ *		rx_ring_info->rx_desc_ring =
+ *		    (struct rx_prod_pkt_bd *)rx_ring->bd;
+ *		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
+ *		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+ *
+ *		if (!rx_ring->bd)
+ *			return -ENOMEM;
+ *		if (rx_ring->vmem_size) {
+ *			rx_ring->vmem =
+ *			    (void **)((char *)mz->addr + rx_vmem_start);
+ *			rx_ring_info->rx_buf_ring =
+ *			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
+ *		}
+ *	}
+ */
+
+	cp_ring->bd = ((char *)mz->addr + cp_ring_start);
+	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
+	cp_ring_info->cp_desc_ring = cp_ring->bd;
+	cp_ring_info->cp_desc_mapping = cp_ring->bd_dma;
+
+	if (!cp_ring->bd)
+		return -ENOMEM;
+	if (cp_ring->vmem_size)
+		*cp_ring->vmem = ((char *)mz->addr + stats_len);
+	if (stats_len) {
+		cp_ring_info->hw_stats = mz->addr;
+		cp_ring_info->hw_stats_map = mz->phys_addr;
+	}
+	cp_ring_info->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index ebbd759..281587c 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -87,6 +87,14 @@ struct bnxt_ring_grp_info {
 };
 
 struct bnxt;
+struct bnxt_tx_ring_info;
+struct bnxt_rx_ring_info;
+struct bnxt_cp_ring_info;
 void bnxt_free_ring(struct bnxt_ring_struct *ring);
+int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
+			    struct bnxt_tx_ring_info *tx_ring_info,
+			    struct bnxt_rx_ring_info *rx_ring_info,
+			    struct bnxt_cp_ring_info *cp_ring_info,
+			    const char *suffix);
 
 #endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index a3648c2..7aba199 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -39,6 +39,7 @@
 #include "bnxt_cpr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 
 /*
  * TX Queues
@@ -55,9 +56,20 @@ void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
 		cpr->hw_stats = NULL;
 }
 
-static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq __rte_unused)
+static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq)
 {
-	/* TODO: Requires interaction with TX ring */
+	struct bnxt_sw_tx_bd *sw_ring;
+	uint16_t i;
+
+	sw_ring = txq->tx_ring->tx_buf_ring;
+	if (sw_ring) {
+		for (i = 0; i < txq->tx_ring->tx_ring_struct->ring_size; i++) {
+			if (sw_ring[i].mbuf) {
+				rte_pktmbuf_free(sw_ring[i].mbuf);
+				sw_ring[i].mbuf = NULL;
+			}
+		}
+	}
 }
 
 void bnxt_free_tx_mbufs(struct bnxt *bp)
@@ -76,7 +88,15 @@ void bnxt_tx_queue_release_op(void *tx_queue)
 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
 
 	if (txq) {
-		/* TODO: Free ring and stats here */
+		/* Free TX ring hardware descriptors */
+		bnxt_tx_queue_release_mbufs(txq);
+		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
+
+		/* Free TX completion ring hardware descriptors */
+		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
+
+		bnxt_free_txq_stats(txq);
+
 		rte_free(txq);
 	}
 }
@@ -112,14 +132,24 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	txq->nb_tx_desc = nb_desc;
 	txq->tx_free_thresh = tx_conf->tx_free_thresh;
 
-	/* TODO: Initialize ring structure */
+	bnxt_init_tx_ring_struct(txq);
 
 	txq->queue_id = queue_idx;
 	txq->port_id = eth_dev->data->port_id;
 
-	/* TODO: Allocate TX ring hardware descriptors */
+	/* Allocate TX ring hardware descriptors */
+	if (bnxt_alloc_rings(bp, queue_idx, txq->tx_ring, NULL, txq->cp_ring,
+			"bnxt_tx_ring")) {
+		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for tx_ring failed!");
+		bnxt_tx_queue_release_op(txq);
+		return -ENOMEM;
+	}
 
-	/* TODO: Initialize the ring */
+	if (bnxt_init_one_tx_ring(txq)) {
+		RTE_LOG(ERR, PMD, "bnxt_init_one_tx_ring failed!");
+		bnxt_tx_queue_release_op(txq);
+		return -ENOMEM;
+	}
 
 	eth_dev->data->tx_queues[queue_idx] = txq;
 	return 0;
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
new file mode 100644
index 0000000..2314410
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -0,0 +1,314 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_byteorder.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_ring.h"
+#include "bnxt_txq.h"
+#include "bnxt_txr.h"
+#include "hsi_struct_def_dpdk.h"
+#include <stdbool.h>
+
+/*
+ * TX Ring handling
+ */
+
+void bnxt_free_tx_rings(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->tx_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+
+		if (!txq)
+			continue;
+
+		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
+		/* TODO: free() txq->tx_ring and txq->tx_ring->tx_ring_struct */
+		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
+		/* TODO: free() txq->cp_ring and txq->cp_ring->cp_ring_struct */
+
+		rte_free(txq);
+		bp->tx_queues[i] = NULL;
+	}
+}
+
+int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	struct bnxt_ring_struct *ring = txr->tx_ring_struct;
+
+	txq->tx_wake_thresh = ring->ring_size / 2;
+	ring->fw_ring_id = INVALID_HW_RING_ID;
+
+	return 0;
+}
+
+void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_tx_ring_info *txr;
+	struct bnxt_ring_struct *ring;
+
+	/* TODO: These need to be allocated */
+	txr = txq->tx_ring;
+	ring = txr->tx_ring_struct;
+	ring->ring_size = rte_align32pow2(txq->nb_tx_desc + 1);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)txr->tx_desc_ring;
+	ring->bd_dma = txr->tx_desc_mapping;
+	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_tx_bd);
+	ring->vmem = (void **)&txr->tx_buf_ring;
+
+	/* TODO: These need to be allocated */
+	cpr = txq->cp_ring;
+	ring = cpr->cp_ring_struct;
+	ring->ring_size = txr->tx_ring_struct->ring_size;
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
+
+static inline uint32_t bnxt_tx_avail(struct bnxt_tx_ring_info *txr)
+{
+	/* Tell compiler to fetch tx indices from memory. */
+	rte_compiler_barrier();
+
+	return txr->tx_ring_struct->ring_size -
+		((txr->tx_prod - txr->tx_cons) &
+			txr->tx_ring_struct->ring_mask) - 1;
+}
+
+static uint16_t bnxt_start_xmit(struct rte_mbuf *tx_pkt,
+				struct bnxt_tx_queue *txq)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	struct tx_bd_long *txbd;
+	struct tx_bd_long_hi *txbd1;
+	uint32_t vlan_tag_flags, cfa_action;
+	bool long_bd = false;
+	uint16_t last_prod = 0;
+	struct rte_mbuf *m_seg;
+	struct bnxt_sw_tx_bd *tx_buf;
+	static const uint32_t lhint_arr[4] = {
+		TX_BD_LONG_FLAGS_LHINT_LT512,
+		TX_BD_LONG_FLAGS_LHINT_LT1K,
+		TX_BD_LONG_FLAGS_LHINT_LT2K,
+		TX_BD_LONG_FLAGS_LHINT_LT2K
+	};
+
+	if (tx_pkt->ol_flags & (PKT_TX_TCP_SEG | PKT_TX_TCP_CKSUM |
+				PKT_TX_UDP_CKSUM | PKT_TX_IP_CKSUM |
+				PKT_TX_VLAN_PKT))
+		long_bd = true;
+
+	tx_buf = &txr->tx_buf_ring[txr->tx_prod];
+	tx_buf->mbuf = tx_pkt;
+	tx_buf->nr_bds = long_bd + tx_pkt->nb_segs;
+	last_prod = (txr->tx_prod + tx_buf->nr_bds - 1) &
+				txr->tx_ring_struct->ring_mask;
+
+	if (unlikely(bnxt_tx_avail(txr) < tx_buf->nr_bds))
+		return -ENOMEM;
+
+	txbd = &txr->tx_desc_ring[txr->tx_prod];
+	txbd->opaque = txr->tx_prod;
+	txbd->flags_type = tx_buf->nr_bds << TX_BD_LONG_FLAGS_BD_CNT_SFT;
+	txbd->len = tx_pkt->data_len;
+	if (txbd->len >= 2014)
+		txbd->flags_type |= TX_BD_LONG_FLAGS_LHINT_GTE2K;
+	else
+		txbd->flags_type |= lhint_arr[txbd->len >> 9];
+	txbd->addr = rte_cpu_to_le_32(RTE_MBUF_DATA_DMA_ADDR(tx_buf->mbuf));
+
+	if (long_bd) {
+		txbd->flags_type |= TX_BD_LONG_TYPE_TX_BD_LONG;
+		vlan_tag_flags = 0;
+		cfa_action = 0;
+		if (tx_buf->mbuf->ol_flags & PKT_TX_VLAN_PKT) {
+			/* shurd: Should this mask at
+			 * TX_BD_LONG_CFA_META_VLAN_VID_MASK?
+			 */
+			vlan_tag_flags = TX_BD_LONG_CFA_META_KEY_VLAN_TAG |
+				tx_buf->mbuf->vlan_tci;
+			/* Currently supports 8021Q, 8021AD vlan offloads
+			 * QINQ1, QINQ2, QINQ3 vlan headers are deprecated
+			 */
+			/* DPDK only supports 802.11q VLAN packets */
+			vlan_tag_flags |=
+					TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
+		}
+
+		txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+
+		txbd1 = (struct tx_bd_long_hi *)
+					&txr->tx_desc_ring[txr->tx_prod];
+		txbd1->lflags = 0;
+		txbd1->cfa_meta = vlan_tag_flags;
+		txbd1->cfa_action = cfa_action;
+
+		if (tx_pkt->ol_flags & PKT_TX_TCP_SEG) {
+			/* TSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_LSO;
+			txbd1->hdr_size = tx_pkt->l2_len + tx_pkt->l3_len +
+					tx_pkt->l4_len;
+			txbd1->mss = tx_pkt->tso_segsz;
+
+		} else if (tx_pkt->ol_flags & (PKT_TX_TCP_CKSUM |
+					PKT_TX_UDP_CKSUM)) {
+			/* TCP/UDP CSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_TCP_UDP_CHKSUM;
+			txbd1->mss = 0;
+
+		} else if (tx_pkt->ol_flags & PKT_TX_IP_CKSUM) {
+			/* IP CSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_IP_CHKSUM;
+			txbd1->mss = 0;
+		}
+	} else {
+		txbd->flags_type |= TX_BD_SHORT_TYPE_TX_BD_SHORT;
+	}
+
+	m_seg = tx_pkt->next;
+	/* i is set at the end of the if(long_bd) block */
+	while (txr->tx_prod != last_prod) {
+		txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+		tx_buf = &txr->tx_buf_ring[txr->tx_prod];
+
+		txbd = &txr->tx_desc_ring[txr->tx_prod];
+		txbd->addr = rte_cpu_to_le_32(RTE_MBUF_DATA_DMA_ADDR(m_seg));
+		txbd->flags_type = TX_BD_SHORT_TYPE_TX_BD_SHORT;
+		txbd->len = m_seg->data_len;
+
+		m_seg = m_seg->next;
+	}
+
+	txbd->flags_type |= TX_BD_LONG_FLAGS_PACKET_END;
+
+	txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+
+	return 0;
+}
+
+static void bnxt_tx_cmp(struct bnxt_tx_queue *txq, int nr_pkts)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	uint16_t cons = txr->tx_cons;
+	int i, j;
+
+	for (i = 0; i < nr_pkts; i++) {
+		struct bnxt_sw_tx_bd *tx_buf;
+		struct rte_mbuf *mbuf;
+
+		tx_buf = &txr->tx_buf_ring[cons];
+		cons = RING_NEXT(txr->tx_ring_struct, cons);
+		mbuf = tx_buf->mbuf;
+		tx_buf->mbuf = NULL;
+
+		/* EW - no need to unmap DMA memory? */
+
+		for (j = 1; j < tx_buf->nr_bds; j++)
+			cons = RING_NEXT(txr->tx_ring_struct, cons);
+		rte_pktmbuf_free(mbuf);
+	}
+
+	txr->tx_cons = cons;
+}
+
+static int bnxt_handle_tx_cp(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+	uint32_t raw_cons = cpr->cp_raw_cons;
+	uint32_t cons;
+	int nb_tx_pkts = 0;
+	struct tx_cmpl *txcmp;
+
+	if ((txq->tx_ring->tx_ring_struct->ring_size -
+			(bnxt_tx_avail(txq->tx_ring))) >
+			txq->tx_free_thresh) {
+		while (1) {
+			cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
+			txcmp = (struct tx_cmpl *)&cpr->cp_desc_ring[cons];
+
+			if (!CMP_VALID(txcmp, raw_cons, cpr->cp_ring_struct))
+				break;
+
+			if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
+				nb_tx_pkts++;
+			else
+				RTE_LOG(DEBUG, PMD,
+						"Unhandled CMP type %02x\n",
+						CMP_TYPE(txcmp));
+			raw_cons = NEXT_RAW_CMP(raw_cons);
+		}
+		if (nb_tx_pkts)
+			bnxt_tx_cmp(txq, nb_tx_pkts);
+		cpr->cp_raw_cons = raw_cons;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+	}
+	return nb_tx_pkts;
+}
+
+uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct bnxt_tx_queue *txq = tx_queue;
+	uint16_t nb_tx_pkts = 0;
+	uint16_t db_mask = txq->tx_ring->tx_ring_struct->ring_size >> 2;
+	uint16_t last_db_mask = 0;
+
+	/* Handle TX completions */
+	bnxt_handle_tx_cp(txq);
+
+	/* Handle TX burst request */
+	for (nb_tx_pkts = 0; nb_tx_pkts < nb_pkts; nb_tx_pkts++) {
+		if (bnxt_start_xmit(tx_pkts[nb_tx_pkts], txq)) {
+			break;
+		} else if ((nb_tx_pkts & db_mask) != last_db_mask) {
+			B_TX_DB(txq->tx_ring->tx_doorbell,
+					txq->tx_ring->tx_prod);
+			last_db_mask = nb_tx_pkts & db_mask;
+		}
+	}
+	if (nb_tx_pkts)
+		B_TX_DB(txq->tx_ring->tx_doorbell, txq->tx_ring->tx_prod);
+
+	return nb_tx_pkts;
+}
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
new file mode 100644
index 0000000..1797a3d
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -0,0 +1,71 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_TXR_H_
+#define _BNXT_TXR_H_
+
+#define MAX_TX_RINGS	16
+#define BNXT_TX_PUSH_THRESH 92
+
+#define B_TX_DB(db, prod)						\
+		(*(uint32_t *)db = (DB_KEY_TX | prod))
+
+struct bnxt_tx_ring_info {
+	uint16_t		tx_prod;
+	uint16_t		tx_cons;
+	void			*tx_doorbell;
+
+	struct tx_bd_long	*tx_desc_ring;
+	struct bnxt_sw_tx_bd	*tx_buf_ring;
+
+	phys_addr_t		tx_desc_mapping;
+
+#define BNXT_DEV_STATE_CLOSING	0x1
+	uint32_t		dev_state;
+
+	struct bnxt_ring_struct	*tx_ring_struct;
+};
+
+struct bnxt_sw_tx_bd {
+	struct rte_mbuf		*mbuf; /* mbuf associated with TX descriptor */
+	uint8_t			is_gso;
+	unsigned short		nr_bds;
+};
+
+void bnxt_free_tx_rings(struct bnxt *bp);
+int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
+void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq);
+uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 91a83d1..1543f20 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -99,6 +99,518 @@ struct ctx_hw_stats64 {
 #define HWRM_ERR_CODE_INVALID_PARAMS                      (UINT32_C(0x2))
 #define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED              (UINT32_C(0x3))
 
+/* Short TX BD (16 bytes) */
+struct tx_bd_short {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	/* This value identifies the type of buffer descriptor. */
+	#define TX_BD_SHORT_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_BD_SHORT_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 16B long and is used for normal L2
+		 * packet transmission.
+		 */
+	#define TX_BD_SHORT_TYPE_TX_BD_SHORT		(UINT32_C(0x0) << 0)
+	/*
+	 * If set to 1, the packet ends with the data in the buffer pointed to
+	 * by this descriptor. This flag must be valid on every BD.
+	 */
+	#define TX_BD_SHORT_FLAGS_PACKET_END		UINT32_C(0x40)
+	/*
+	 * If set to 1, the device will not generate a completion for this
+	 * transmit packet unless there is an error in it's processing. If this
+	 * bit is set to 0, then the packet will be completed normally. This bit
+	 * must be valid only on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_NO_CMPL		UINT32_C(0x80)
+	/*
+	 * This value indicates how many 16B BD locations are consumed in the
+	 * ring by this packet. A value of 1 indicates that this BD is the only
+	 * BD (and that the it is a short BD). A value of 3 indicates either 3
+	 * short BDs or 1 long BD and one short BD in the packet. A value of 0
+	 * indicates that there are 32 BD locations in the packet (the maximum).
+	 * This field is valid only on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_BD_CNT_MASK		UINT32_C(0x1f00)
+	#define TX_BD_SHORT_FLAGS_BD_CNT_SFT		8
+	/*
+	 * This value is a hint for the length of the entire packet. It is used
+	 * by the chip to optimize internal processing. The packet will be
+	 * dropped if the hint is too short. This field is valid only on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_LHINT_MASK		UINT32_C(0x6000)
+	#define TX_BD_SHORT_FLAGS_LHINT_SFT		13
+		/* indicates packet length < 512B */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT512		(UINT32_C(0x0) << 13)
+		/* indicates 512 <= packet length < 1KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT1K		(UINT32_C(0x1) << 13)
+		/* indicates 1KB <= packet length < 2KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT2K		(UINT32_C(0x2) << 13)
+		/* indicates packet length >= 2KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_GTE2K		(UINT32_C(0x3) << 13)
+	#define TX_BD_SHORT_FLAGS_LHINT_LAST	TX_BD_SHORT_FLAGS_LHINT_GTE2K
+	/*
+	 * If set to 1, the device immediately updates the Send Consumer Index
+	 * after the buffer associated with this descriptor has been transferred
+	 * via DMA to NIC memory from host memory. An interrupt may or may not
+	 * be generated according to the state of the interrupt avoidance
+	 * mechanisms. If this bit is set to 0, then the Consumer Index is only
+	 * updated as soon as one of the host interrupt coalescing conditions
+	 * has been met. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_COAL_NOW		UINT32_C(0x8000)
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_BD_SHORT_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the host physical buffer this BD describes in
+	 * bytes. This field must be valid on all BDs of a packet.
+	 */
+	uint16_t len;
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with the
+	 * transmit BD. This field must be valid on the first BD of a packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address for the portion of the packet
+	 * described by this TX BD. This value must be valid on all BDs of a
+	 * packet.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
+/* Long TX BD (32 bytes split to 2 16-byte struct) */
+struct tx_bd_long {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	/* This value identifies the type of buffer descriptor. */
+	#define TX_BD_LONG_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_BD_LONG_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 32B long and is used for normal L2
+		 * packet transmission.
+		 */
+	#define TX_BD_LONG_TYPE_TX_BD_LONG		(UINT32_C(0x10) << 0)
+	/*
+	 * If set to 1, the packet ends with the data in the buffer pointed to
+	 * by this descriptor. This flag must be valid on every BD.
+	 */
+	#define TX_BD_LONG_FLAGS_PACKET_END		UINT32_C(0x40)
+	/*
+	 * If set to 1, the device will not generate a completion for this
+	 * transmit packet unless there is an error in it's processing. If this
+	 * bit is set to 0, then the packet will be completed normally. This bit
+	 * must be valid only on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_NO_CMPL		UINT32_C(0x80)
+	/*
+	 * This value indicates how many 16B BD locations are consumed in the
+	 * ring by this packet. A value of 1 indicates that this BD is the only
+	 * BD (and that the it is a short BD). A value of 3 indicates either 3
+	 * short BDs or 1 long BD and one short BD in the packet. A value of 0
+	 * indicates that there are 32 BD locations in the packet (the maximum).
+	 * This field is valid only on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_BD_CNT_MASK		UINT32_C(0x1f00)
+	#define TX_BD_LONG_FLAGS_BD_CNT_SFT		8
+	/*
+	 * This value is a hint for the length of the entire packet. It is used
+	 * by the chip to optimize internal processing. The packet will be
+	 * dropped if the hint is too short. This field is valid only on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_LHINT_MASK		UINT32_C(0x6000)
+	#define TX_BD_LONG_FLAGS_LHINT_SFT		13
+		/* indicates packet length < 512B */
+	#define TX_BD_LONG_FLAGS_LHINT_LT512		(UINT32_C(0x0) << 13)
+		/* indicates 512 <= packet length < 1KB */
+	#define TX_BD_LONG_FLAGS_LHINT_LT1K		(UINT32_C(0x1) << 13)
+		/* indicates 1KB <= packet length < 2KB */
+	#define TX_BD_LONG_FLAGS_LHINT_LT2K		(UINT32_C(0x2) << 13)
+		/* indicates packet length >= 2KB */
+	#define TX_BD_LONG_FLAGS_LHINT_GTE2K		(UINT32_C(0x3) << 13)
+	#define TX_BD_LONG_FLAGS_LHINT_LAST	TX_BD_LONG_FLAGS_LHINT_GTE2K
+	/*
+	 * If set to 1, the device immediately updates the Send Consumer Index
+	 * after the buffer associated with this descriptor has been transferred
+	 * via DMA to NIC memory from host memory. An interrupt may or may not
+	 * be generated according to the state of the interrupt avoidance
+	 * mechanisms. If this bit is set to 0, then the Consumer Index is only
+	 * updated as soon as one of the host interrupt coalescing conditions
+	 * has been met. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_COAL_NOW		UINT32_C(0x8000)
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	#define TX_BD_LONG_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_BD_LONG_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the host physical buffer this BD describes in
+	 * bytes. This field must be valid on all BDs of a packet.
+	 */
+	uint16_t len;
+
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with the
+	 * transmit BD. This field must be valid on the first BD of a packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address for the portion of the packet
+	 * described by this TX BD. This value must be valid on all BDs of a
+	 * packet.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
+/* last 16 bytes of Long TX BD */
+
+struct tx_bd_long_hi {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Their value on other BDs of the packet will be ignored.
+	 */
+	/*
+	 * If set to 1, the controller replaces the TCP/UPD checksum fields of
+	 * normal TCP/UPD checksum, or the inner TCP/UDP checksum field of the
+	 * encapsulated TCP/UDP packets with the hardware calculated TCP/UDP
+	 * checksum for the packet associated with this descriptor. This bit
+	 * must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_TCP_UDP_CHKSUM	UINT32_C(0x1)
+	/*
+	 * If set to 1, the controller replaces the IP checksum of the normal
+	 * packets, or the inner IP checksum of the encapsulated packets with
+	 * the hardware calculated IP checksum for the packet associated with
+	 * this descriptor. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_IP_CHKSUM		UINT32_C(0x2)
+	/*
+	 * If set to 1, the controller will not append an Ethernet CRC to the
+	 * end of the frame. This bit must be valid on the first BD of a packet.
+	 * Packet must be 64B or longer when this flag is set. It is not useful
+	 * to use this bit with any form of TX offload such as CSO or LSO. The
+	 * intent is that the packet from the host already has a valid Ethernet
+	 * CRC on the packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_NOCRC			UINT32_C(0x4)
+	/*
+	 * If set to 1, the device will record the time at which the packet was
+	 * actually transmitted at the TX MAC. This bit must be valid on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_STAMP			UINT32_C(0x8)
+	/*
+	 * If set to 1, The controller replaces the tunnel IP checksum field
+	 * with hardware calculated IP checksum for the IP header of the packet
+	 * associated with this descriptor. In case of VXLAN, the controller
+	 * also replaces the outer header UDP checksum with hardware calculated
+	 * UDP checksum for the packet associated with this descriptor.
+	 */
+	#define TX_BD_LONG_LFLAGS_T_IP_CHKSUM		UINT32_C(0x10)
+	/*
+	 * If set to 1, the device will treat this packet with LSO(Large Send
+	 * Offload) processing for both normal or encapsulated packets, which is
+	 * a form of TCP segmentation. When this bit is 1, the hdr_size and mss
+	 * fields must be valid. The driver doesn't need to set t_ip_chksum,
+	 * ip_chksum, and tcp_udp_chksum flags since the controller will replace
+	 * the appropriate checksum fields for segmented packets. When this bit
+	 * is 1, the hdr_size and mss fields must be valid.
+	 */
+	#define TX_BD_LONG_LFLAGS_LSO			UINT32_C(0x20)
+	/*
+	 * If set to zero when LSO is '1', then the IPID will be treated as a
+	 * 16b number and will be wrapped if it exceeds a value of 0xffff. If
+	 * set to one when LSO is '1', then the IPID will be treated as a 15b
+	 * number and will be wrapped if it exceeds a value 0f 0x7fff.
+	 */
+	#define TX_BD_LONG_LFLAGS_IPID_FMT		UINT32_C(0x40)
+	/*
+	 * If set to zero when LSO is '1', then the IPID of the tunnel IP header
+	 * will not be modified during LSO operations. If set to one when LSO is
+	 * '1', then the IPID of the tunnel IP header will be incremented for
+	 * each subsequent segment of an LSO operation.
+	 */
+	#define TX_BD_LONG_LFLAGS_T_IPID		UINT32_C(0x80)
+	/*
+	 * If set to '1', then the RoCE ICRC will be appended to the packet.
+	 * Packet must be a valid RoCE format packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_ROCE_CRC		UINT32_C(0x100)
+	/*
+	 * If set to '1', then the FCoE CRC will be appended to the packet.
+	 * Packet must be a valid FCoE format packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_FCOE_CRC		UINT32_C(0x200)
+	uint16_t lflags;
+
+	/*
+	 * When LSO is '1', this field must contain the offset of the TCP
+	 * payload from the beginning of the packet in as 16b words. In case of
+	 * encapsulated/tunneling packet, this field contains the offset of the
+	 * inner TCP payload from beginning of the packet as 16-bit words. This
+	 * value must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_HDR_SIZE_MASK		UINT32_C(0x1ff)
+	#define TX_BD_LONG_HDR_SIZE_SFT			0
+	uint16_t hdr_size;
+
+	/*
+	 * This is the MSS value that will be used to do the LSO processing. The
+	 * value is the length in bytes of the TCP payload for each segment
+	 * generated by the LSO operation. This value must be valid on the first
+	 * BD of a packet.
+	 */
+	#define TX_BD_LONG_MSS_MASK			UINT32_C(0x7fff)
+	#define TX_BD_LONG_MSS_SFT			0
+	uint32_t mss;
+
+	uint16_t unused_2;
+
+	/*
+	 * This value selects a CFA action to perform on the packet. Set this
+	 * value to zero if no CFA action is desired. This value must be valid
+	 * on the first BD of a packet.
+	 */
+	uint16_t cfa_action;
+
+	/*
+	 * This value is action meta-data that defines CFA edit operations that
+	 * are done in addition to any action editing.
+	 */
+	/* When key=1, This is the VLAN tag VID value. */
+	#define TX_BD_LONG_CFA_META_VLAN_VID_MASK	UINT32_C(0xfff)
+	#define TX_BD_LONG_CFA_META_VLAN_VID_SFT	0
+	/* When key=1, This is the VLAN tag DE value. */
+	#define TX_BD_LONG_CFA_META_VLAN_DE		UINT32_C(0x1000)
+	/* When key=1, This is the VLAN tag PRI value. */
+	#define TX_BD_LONG_CFA_META_VLAN_PRI_MASK	UINT32_C(0xe000)
+	#define TX_BD_LONG_CFA_META_VLAN_PRI_SFT	13
+	/* When key=1, This is the VLAN tag TPID select value. */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_MASK	UINT32_C(0x70000)
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_SFT	16
+		/* 0x88a8 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8	(UINT32_C(0x0) << 16)
+		/* 0x8100 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100	(UINT32_C(0x1) << 16)
+		/* 0x9100 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100	(UINT32_C(0x2) << 16)
+		/* 0x9200 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200	(UINT32_C(0x3) << 16)
+		/* 0x9300 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300	(UINT32_C(0x4) << 16)
+		/* Value programmed in CFA VLANTPID register. */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPIDCFG	(UINT32_C(0x5) << 16)
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_LAST \
+					TX_BD_LONG_CFA_META_VLAN_TPID_TPIDCFG
+	/* When key=1, This is the VLAN tag TPID select value. */
+	#define TX_BD_LONG_CFA_META_VLAN_RESERVED_MASK	UINT32_C(0xff80000)
+	#define TX_BD_LONG_CFA_META_VLAN_RESERVED_SFT	19
+	/*
+	 * This field identifies the type of edit to be performed on the packet.
+	 * This value must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_CFA_META_KEY_MASK		UINT32_C(0xf0000000)
+	#define TX_BD_LONG_CFA_META_KEY_SFT		28
+		/* No editing */
+	#define TX_BD_LONG_CFA_META_KEY_NONE		(UINT32_C(0x0) << 28)
+		/*
+		 * - meta[17:16] - TPID select value (0 = 0x8100). - meta[15:12]
+		 * - PRI/DE value. - meta[11:0] - VID value.
+		 */
+	#define TX_BD_LONG_CFA_META_KEY_VLAN_TAG	(UINT32_C(0x1) << 28)
+	#define TX_BD_LONG_CFA_META_KEY_LAST	TX_BD_LONG_CFA_META_KEY_VLAN_TAG
+	uint32_t cfa_meta;
+} __attribute__((packed));
+
+/* Completion Ring Structures */
+/* Note: This structure is used by the HWRM to communicate HWRM Error. */
+/* Base Completion Record (16 bytes) */
+struct cmpl_base {
+	/* unused is 10 b */
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define CMPL_BASE_TYPE_MASK			UINT32_C(0x3f)
+	#define CMPL_BASE_TYPE_SFT			0
+		/* TX L2 completion: Completion of TX packet. Length = 16B */
+	#define CMPL_BASE_TYPE_TX_L2			(UINT32_C(0x0) << 0)
+		/*
+		 * RX L2 completion: Completion of and L2 RX packet.
+		 * Length = 32B
+		*/
+	#define CMPL_BASE_TYPE_RX_L2			(UINT32_C(0x11) << 0)
+		/*
+		 * RX Aggregation Buffer completion : Completion of an L2
+		 * aggregation buffer in support of TPA, HDS, or Jumbo packet
+		 * completion. Length = 16B
+		 */
+	#define CMPL_BASE_TYPE_RX_AGG			(UINT32_C(0x12) << 0)
+		/*
+		 * RX L2 TPA Start Completion: Completion at the beginning of a
+		 * TPA operation. Length = 32B
+		 */
+	#define CMPL_BASE_TYPE_RX_TPA_START		(UINT32_C(0x13) << 0)
+		/*
+		 * RX L2 TPA End Completion: Completion at the end of a TPA
+		 * operation. Length = 32B
+		 */
+	#define CMPL_BASE_TYPE_RX_TPA_END		(UINT32_C(0x15) << 0)
+		/*
+		 * Statistics Ejection Completion: Completion of statistics data
+		 * ejection buffer. Length = 16B
+		 */
+	#define CMPL_BASE_TYPE_STAT_EJECT		(UINT32_C(0x1a) << 0)
+		/* HWRM Command Completion: Completion of an HWRM command. */
+	#define CMPL_BASE_TYPE_HWRM_DONE		(UINT32_C(0x20) << 0)
+		/* Forwarded HWRM Request */
+	#define CMPL_BASE_TYPE_HWRM_FWD_REQ		(UINT32_C(0x22) << 0)
+		/* Forwarded HWRM Response */
+	#define CMPL_BASE_TYPE_HWRM_FWD_RESP		(UINT32_C(0x24) << 0)
+		/* HWRM Asynchronous Event Information */
+	#define CMPL_BASE_TYPE_HWRM_ASYNC_EVENT		(UINT32_C(0x2e) << 0)
+		/* CQ Notification */
+	#define CMPL_BASE_TYPE_CQ_NOTIFICATION		(UINT32_C(0x30) << 0)
+		/* SRQ Threshold Event */
+	#define CMPL_BASE_TYPE_SRQ_EVENT		(UINT32_C(0x32) << 0)
+		/* DBQ Threshold Event */
+	#define CMPL_BASE_TYPE_DBQ_EVENT		(UINT32_C(0x34) << 0)
+		/* QP Async Notification */
+	#define CMPL_BASE_TYPE_QP_EVENT			(UINT32_C(0x38) << 0)
+		/* Function Async Notification */
+	#define CMPL_BASE_TYPE_FUNC_EVENT		(UINT32_C(0x3a) << 0)
+	uint16_t type;
+
+	uint16_t info1;
+	uint32_t info2;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define CMPL_BASE_V				UINT32_C(0x1)
+	/* info3 is 31 b */
+	#define CMPL_BASE_INFO3_MASK			UINT32_C(0xfffffffe)
+	#define CMPL_BASE_INFO3_SFT			1
+	uint32_t info3_v;
+
+	uint32_t info4;
+} __attribute__((packed));
+
+/* TX Completion Record (16 bytes) */
+struct tx_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define TX_CMPL_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_CMPL_TYPE_SFT			0
+		/* TX L2 completion: Completion of TX packet. Length = 16B */
+	#define TX_CMPL_TYPE_TX_L2			(UINT32_C(0x0) << 0)
+	/*
+	 * When this bit is '1', it indicates a packet that has an error of some
+	 * type. Type of error is indicated in error_flags.
+	 */
+	#define TX_CMPL_FLAGS_ERROR			UINT32_C(0x40)
+	/*
+	 * When this bit is '1', it indicates that the packet completed was
+	 * transmitted using the push acceleration data provided by the driver.
+	 * When this bit is '0', it indicates that the packet had not push
+	 * acceleration data written or was executed as a normal packet even
+	 * though push data was provided.
+	 */
+	#define TX_CMPL_FLAGS_PUSH			UINT32_C(0x80)
+	#define TX_CMPL_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_CMPL_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	uint16_t unused_0;
+
+	/*
+	 * This is a copy of the opaque field from the first TX BD of this
+	 * transmitted packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define TX_CMPL_V				UINT32_C(0x1)
+	/*
+	 * This error indicates that there was some sort of problem with the BDs
+	 * for the packet.
+	 */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_MASK	UINT32_C(0xe)
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_SFT		1
+		/* No error */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_NO_ERROR	(UINT32_C(0x0) << 1)
+		/* Bad Format: BDs were not formatted correctly. */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_BAD_FMT	(UINT32_C(0x2) << 1)
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_LAST \
+					TX_CMPL_ERRORS_BUFFER_ERROR_BAD_FMT
+	/*
+	 * When this bit is '1', it indicates that the length of the packet was
+	 * zero. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_ZERO_LENGTH_PKT		UINT32_C(0x10)
+	/*
+	 * When this bit is '1', it indicates that the packet was longer than
+	 * the programmed limit in TDI. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_EXCESSIVE_BD_LENGTH	UINT32_C(0x20)
+	/*
+	 * When this bit is '1', it indicates that one or more of the BDs
+	 * associated with this packet generated a PCI error. This probably
+	 * means the address was not valid.
+	 */
+	#define TX_CMPL_ERRORS_DMA_ERROR		UINT32_C(0x40)
+	/*
+	 * When this bit is '1', it indicates that the packet was longer than
+	 * indicated by the hint. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_HINT_TOO_SHORT		UINT32_C(0x80)
+	/*
+	 * When this bit is '1', it indicates that the packet was dropped due to
+	 * Poison TLP error on one or more of the TLPs in the PXP completion.
+	 */
+	#define TX_CMPL_ERRORS_POISON_TLP_ERROR		UINT32_C(0x100)
+	#define TX_CMPL_ERRORS_MASK			UINT32_C(0xfffe)
+	#define TX_CMPL_ERRORS_SFT			1
+	uint16_t errors_v;
+
+	uint16_t unused_1;
+	uint32_t unused_2;
+} __attribute__((packed)) tx_cmpl_t, *ptx_cmpl_t;
+
 /* HWRM Forwarded Request (16 bytes) */
 struct hwrm_fwd_req_cmpl {
 	/* Length of forwarded request in bytes. */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 14/39] bnxt: initial Rx code implementation
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (11 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 13/39] bnxt: initial Tx code implementation Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 15/39] bnxt: Code to alloc/free ring Stephen Hurd
                   ` (27 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Initial implementation of rx_pkt_burst
Add code to allocate rings to bnxt_ring.c

v4:
Use rte_mbuf_raw_alloc instead of the now depricated
__rte_mbuf_raw_alloc and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +-
 drivers/net/bnxt/bnxt_ring.c           |  60 ++---
 drivers/net/bnxt/bnxt_rxq.c            |  34 ++-
 drivers/net/bnxt/bnxt_rxr.c            | 341 ++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_rxr.h            |  62 +++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 474 +++++++++++++++++++++++++++++++++
 7 files changed, 935 insertions(+), 40 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_rxr.c
 create mode 100644 drivers/net/bnxt/bnxt_rxr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 0785681..4d35412 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -54,6 +54,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txr.c
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 4ace543..6888363 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -42,6 +42,7 @@
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
 #include "bnxt_txr.h"
@@ -269,7 +270,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 		goto error;
 	}
 	eth_dev->dev_ops = &bnxt_dev_ops;
-	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
+	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
 
 	rc = bnxt_alloc_hwrm_resources(bp);
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index be77bbe..fab0e27 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -36,6 +36,7 @@
 #include "bnxt.h"
 #include "bnxt_cpr.h"
 #include "bnxt_ring.h"
+#include "bnxt_rxr.h"
 #include "bnxt_txr.h"
 
 #include "hsi_struct_def_dpdk.h"
@@ -73,9 +74,8 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 			    const char *suffix)
 {
 	struct bnxt_ring_struct *cp_ring = cp_ring_info->cp_ring_struct;
+	struct bnxt_ring_struct *rx_ring;
 	struct bnxt_ring_struct *tx_ring;
-	/* TODO: RX ring */
-	/* struct bnxt_ring_struct *rx_ring; */
 	struct rte_pci_device *pdev = bp->pdev;
 	const struct rte_memzone *mz = NULL;
 	char mz_name[RTE_MEMZONE_NAMESIZE];
@@ -92,12 +92,8 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 						tx_ring_struct->vmem_size) : 0;
 
 	int rx_vmem_start = tx_vmem_start + tx_vmem_len;
-	/* TODO: RX ring */
-	int rx_vmem_len = 0;
-	/*
-	 * rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
-	 * rx_ring_struct->vmem_size) : 0;
-	 */
+	int rx_vmem_len = rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
+						rx_ring_struct->vmem_size) : 0;
 
 	int cp_ring_start = rx_vmem_start + rx_vmem_len;
 	int cp_ring_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->ring_size *
@@ -109,13 +105,9 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 				   sizeof(struct tx_bd_long)) : 0;
 
 	int rx_ring_start = tx_ring_start + tx_ring_len;
-	/* TODO: RX ring */
-	int rx_ring_len = 0;
-	/*
-	 * rx_ring_info ?
-	 * RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
-	 * sizeof(struct rx_prod_pkt_bd)) : 0;
-	 */
+	int rx_ring_len =  rx_ring_info ?
+		RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
+		sizeof(struct rx_prod_pkt_bd)) : 0;
 
 	int total_alloc_len = rx_ring_start + rx_ring_len;
 
@@ -153,26 +145,24 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		}
 	}
 
-/*
- *	if (rx_ring_info) {
- *		rx_ring = &rx_ring_info->rx_ring_struct;
- *
- *		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
- *		rx_ring_info->rx_desc_ring =
- *		    (struct rx_prod_pkt_bd *)rx_ring->bd;
- *		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
- *		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
- *
- *		if (!rx_ring->bd)
- *			return -ENOMEM;
- *		if (rx_ring->vmem_size) {
- *			rx_ring->vmem =
- *			    (void **)((char *)mz->addr + rx_vmem_start);
- *			rx_ring_info->rx_buf_ring =
- *			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
- *		}
- *	}
- */
+	if (rx_ring_info) {
+		rx_ring = rx_ring_info->rx_ring_struct;
+
+		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
+		rx_ring_info->rx_desc_ring =
+		    (struct rx_prod_pkt_bd *)rx_ring->bd;
+		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
+		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+
+		if (!rx_ring->bd)
+			return -ENOMEM;
+		if (rx_ring->vmem_size) {
+			rx_ring->vmem =
+			    (void **)((char *)mz->addr + rx_vmem_start);
+			rx_ring_info->rx_buf_ring =
+			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
+		}
+	}
 
 	cp_ring->bd = ((char *)mz->addr + cp_ring_start);
 	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index 1ae6f66..13fd08c 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -41,6 +41,7 @@
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -217,7 +218,20 @@ err_out:
 
 static void bnxt_rx_queue_release_mbufs(struct bnxt_rx_queue *rxq __rte_unused)
 {
-	/* TODO: Requires interaction with TX ring */
+	struct bnxt_sw_rx_bd *sw_ring;
+	uint16_t i;
+
+	if (rxq) {
+		sw_ring = rxq->rx_ring->rx_buf_ring;
+		if (sw_ring) {
+			for (i = 0; i < rxq->nb_rx_desc; i++) {
+				if (sw_ring[i].mbuf) {
+					rte_pktmbuf_free_seg(sw_ring[i].mbuf);
+					sw_ring[i].mbuf = NULL;
+				}
+			}
+		}
+	}
 }
 
 void bnxt_free_rx_mbufs(struct bnxt *bp)
@@ -238,7 +252,13 @@ void bnxt_rx_queue_release_op(void *rx_queue)
 	if (rxq) {
 		bnxt_rx_queue_release_mbufs(rxq);
 
-		/* TODO: Free ring and stats here */
+		/* Free RX ring hardware descriptors */
+		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
+
+		/* Free RX completion ring hardware descriptors */
+		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+
+		bnxt_free_rxq_stats(rxq);
 
 		rte_free(rxq);
 	}
@@ -275,7 +295,7 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	rxq->nb_rx_desc = nb_desc;
 	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
 
-	/* TODO: Initialize ring structure */
+	bnxt_init_rx_ring_struct(rxq);
 
 	rxq->queue_id = queue_idx;
 	rxq->port_id = eth_dev->data->port_id;
@@ -283,7 +303,13 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 				0 : ETHER_CRC_LEN);
 
 	eth_dev->data->rx_queues[queue_idx] = rxq;
-	/* TODO: Allocate RX ring hardware descriptors */
+	/* Allocate RX ring hardware descriptors */
+	if (bnxt_alloc_rings(bp, queue_idx, NULL, rxq->rx_ring, rxq->cp_ring,
+			"bnxt_rx_ring")) {
+		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for rx_ring failed!");
+		bnxt_rx_queue_release_op(rxq);
+		return -ENOMEM;
+	}
 
 	return 0;
 }
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
new file mode 100644
index 0000000..43d6979
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -0,0 +1,341 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include <rte_byteorder.h>
+#include <rte_malloc.h>
+#include <rte_memory.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_ring.h"
+#include "bnxt_rxr.h"
+#include "bnxt_rxq.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * RX Ring handling
+ */
+
+static inline struct rte_mbuf *__bnxt_alloc_rx_data(struct rte_mempool *mb)
+{
+	struct rte_mbuf *data;
+
+	data = rte_mbuf_raw_alloc(mb);
+
+	return data;
+}
+
+static inline int bnxt_alloc_rx_data(struct bnxt_rx_queue *rxq,
+				     struct bnxt_rx_ring_info *rxr,
+				     uint16_t prod)
+{
+	struct rx_prod_pkt_bd *rxbd = &rxr->rx_desc_ring[prod];
+	struct bnxt_sw_rx_bd *rx_buf = &rxr->rx_buf_ring[prod];
+	struct rte_mbuf *data;
+
+	data = __bnxt_alloc_rx_data(rxq->mb_pool);
+	if (!data)
+		return -ENOMEM;
+
+	rx_buf->mbuf = data;
+
+	rxbd->addr = rte_cpu_to_le_64(RTE_MBUF_DATA_DMA_ADDR(rx_buf->mbuf));
+
+	return 0;
+}
+
+static void bnxt_reuse_rx_mbuf(struct bnxt_rx_ring_info *rxr, uint16_t cons,
+			       struct rte_mbuf *mbuf)
+{
+	uint16_t prod = rxr->rx_prod;
+	struct bnxt_sw_rx_bd *prod_rx_buf;
+	struct rx_prod_pkt_bd *prod_bd, *cons_bd;
+
+	prod_rx_buf = &rxr->rx_buf_ring[prod];
+
+	prod_rx_buf->mbuf = mbuf;
+
+	prod_bd = &rxr->rx_desc_ring[prod];
+	cons_bd = &rxr->rx_desc_ring[cons];
+
+	prod_bd->addr = cons_bd->addr;
+}
+
+static uint16_t bnxt_rx_pkt(struct rte_mbuf **rx_pkt,
+			    struct bnxt_rx_queue *rxq, uint32_t *raw_cons)
+{
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+	struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+	struct rx_pkt_cmpl *rxcmp;
+	struct rx_pkt_cmpl_hi *rxcmp1;
+	uint32_t tmp_raw_cons = *raw_cons;
+	uint16_t cons, prod, cp_cons =
+	    RING_CMP(cpr->cp_ring_struct, tmp_raw_cons);
+	struct bnxt_sw_rx_bd *rx_buf;
+	struct rte_mbuf *mbuf;
+	int rc = 0;
+
+	rxcmp = (struct rx_pkt_cmpl *)
+	    &cpr->cp_desc_ring[cp_cons];
+
+	tmp_raw_cons = NEXT_RAW_CMP(tmp_raw_cons);
+	cp_cons = RING_CMP(cpr->cp_ring_struct, tmp_raw_cons);
+	rxcmp1 = (struct rx_pkt_cmpl_hi *)&cpr->cp_desc_ring[cp_cons];
+
+	if (!CMP_VALID(rxcmp1, tmp_raw_cons, cpr->cp_ring_struct))
+		return -EBUSY;
+
+	prod = rxr->rx_prod;
+
+	/* EW - GRO deferred to phase 3 */
+	cons = rxcmp->opaque;
+	rx_buf = &rxr->rx_buf_ring[cons];
+	mbuf = rx_buf->mbuf;
+	rte_prefetch0(mbuf);
+
+	mbuf->nb_segs = 1;
+	mbuf->next = NULL;
+	mbuf->pkt_len = rxcmp->len;
+	mbuf->data_len = mbuf->pkt_len;
+	mbuf->port = rxq->port_id;
+	mbuf->ol_flags = 0;
+	if (rxcmp->flags_type & RX_PKT_CMPL_FLAGS_RSS_VALID) {
+		mbuf->hash.rss = rxcmp->rss_hash;
+		mbuf->ol_flags |= PKT_RX_RSS_HASH;
+	} else {
+		mbuf->hash.fdir.id = rxcmp1->cfa_code;
+		mbuf->ol_flags |= PKT_RX_FDIR | PKT_RX_FDIR_ID;
+	}
+	if (rxcmp1->flags2 & RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN) {
+		mbuf->vlan_tci = rxcmp1->metadata &
+			(RX_PKT_CMPL_METADATA_VID_MASK |
+			RX_PKT_CMPL_METADATA_DE |
+			RX_PKT_CMPL_METADATA_PRI_MASK);
+		mbuf->ol_flags |= PKT_RX_VLAN_PKT;
+	}
+
+	rx_buf->mbuf = NULL;
+	if (rxcmp1->errors_v2 & RX_CMP_L2_ERRORS) {
+		/* Re-install the mbuf back to the rx ring */
+		bnxt_reuse_rx_mbuf(rxr, cons, mbuf);
+
+		rc = -EIO;
+		goto next_rx;
+	}
+	/*
+	 * TODO: Redesign this....
+	 * If the allocation fails, the packet does not get received.
+	 * Simply returning this will result in slowly falling behind
+	 * on the producer ring buffers.
+	 * Instead, "filling up" the producer just before ringing the
+	 * doorbell could be a better solution since it will let the
+	 * producer ring starve until memory is available again pushing
+	 * the drops into hardware and getting them out of the driver
+	 * allowing recovery to a full producer ring.
+	 *
+	 * This could also help with cache usage by preventing per-packet
+	 * calls in favour of a tight loop with the same function being called
+	 * in it.
+	 */
+	if (bnxt_alloc_rx_data(rxq, rxr, prod)) {
+		RTE_LOG(ERR, PMD, "mbuf alloc failed with prod=0x%x\n", prod);
+		rc = -ENOMEM;
+		goto next_rx;
+	}
+
+	/*
+	 * All MBUFs are allocated with the same size under DPDK,
+	 * no optimization for rx_copy_thresh
+	 */
+
+	/* AGG buf operation is deferred */
+
+	/* EW - VLAN reception.  Must compare against the ol_flags */
+
+	*rx_pkt = mbuf;
+next_rx:
+	rxr->rx_prod = RING_NEXT(rxr->rx_ring_struct, prod);
+
+	*raw_cons = tmp_raw_cons;
+
+	return rc;
+}
+
+uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct bnxt_rx_queue *rxq = rx_queue;
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+	struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+	uint32_t raw_cons = cpr->cp_raw_cons;
+	uint32_t cons;
+	int nb_rx_pkts = 0;
+	bool rx_event = false;
+	struct rx_pkt_cmpl *rxcmp;
+
+	/* Handle RX burst request */
+	while (1) {
+		int rc;
+
+		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
+		rte_prefetch0(&cpr->cp_desc_ring[cons]);
+		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
+
+		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
+			break;
+
+		/* TODO: Avoid magic numbers... */
+		if ((CMP_TYPE(rxcmp) & 0x30) == 0x10) {
+			rc = bnxt_rx_pkt(&rx_pkts[nb_rx_pkts], rxq, &raw_cons);
+			if (likely(!rc))
+				nb_rx_pkts++;
+			else if (rc == -EBUSY)	/* partial completion */
+				break;
+			rx_event = true;
+		}
+		raw_cons = NEXT_RAW_CMP(raw_cons);
+		if (nb_rx_pkts == nb_pkts)
+			break;
+	}
+	if (raw_cons == cpr->cp_raw_cons) {
+		/*
+		 * For PMD, there is no need to keep on pushing to REARM
+		 * the doorbell if there are no new completions
+		 */
+		return nb_rx_pkts;
+	}
+	cpr->cp_raw_cons = raw_cons;
+
+	B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+	if (rx_event)
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+	return nb_rx_pkts;
+}
+
+void bnxt_free_rx_rings(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->rx_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+
+		if (!rxq)
+			continue;
+
+		/* TODO: free() rxq->rx_ring and rxq->rx_ring->rx_ring_struct */
+		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
+		/* TODO: free() rxq->cp_ring and rxq->cp_ring->cp_ring_struct */
+		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+
+		rte_free(rxq);
+		bp->rx_queues[i] = NULL;
+	}
+}
+
+void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt *bp = rxq->bp;
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_rx_ring_info *rxr;
+	struct bnxt_ring_struct *ring;
+
+	rxq->rx_buf_use_size = bp->eth_dev->data->mtu +
+			       ETHER_HDR_LEN + ETHER_CRC_LEN +
+			       (2 * VLAN_TAG_SIZE);
+	rxq->rx_buf_size = rxq->rx_buf_use_size + sizeof(struct rte_mbuf);
+
+	rxr = rxq->rx_ring;
+	ring = rxr->rx_ring_struct;
+	ring->ring_size = rte_align32pow2(rxq->nb_rx_desc);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)rxr->rx_desc_ring;
+	ring->bd_dma = rxr->rx_desc_mapping;
+	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_rx_bd);
+	ring->vmem = (void **)&rxr->rx_buf_ring;
+
+	cpr = rxq->cp_ring;
+	ring = cpr->cp_ring_struct;
+	ring->ring_size = rxr->rx_ring_struct->ring_size * 2;
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
+
+static void bnxt_init_rxbds(struct bnxt_ring_struct *ring, uint32_t type,
+			    uint16_t len)
+{
+	uint32_t j;
+	struct rx_prod_pkt_bd *rx_bd_ring = (struct rx_prod_pkt_bd *)ring->bd;
+
+	if (!rx_bd_ring)
+		return;
+	for (j = 0; j < ring->ring_size; j++) {
+		rx_bd_ring[j].flags_type = type;
+		rx_bd_ring[j].len = len;
+		rx_bd_ring[j].opaque = j;
+	}
+}
+
+int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt_rx_ring_info *rxr;
+	struct bnxt_ring_struct *ring;
+	uint32_t prod, type;
+	unsigned int i;
+
+	type = RX_PROD_PKT_BD_TYPE_RX_PROD_PKT | RX_PROD_PKT_BD_FLAGS_EOP_PAD;
+
+	/* TODO: These need to be allocated */
+	rxr = rxq->rx_ring;
+	ring = rxr->rx_ring_struct;
+	bnxt_init_rxbds(ring, type, rxq->rx_buf_use_size);
+
+	prod = rxr->rx_prod;
+	for (i = 0; i < ring->ring_size; i++) {
+		if (bnxt_alloc_rx_data(rxq, rxr, prod) != 0) {
+			RTE_LOG(WARNING, PMD,
+				"init'ed rx ring %d with %d/%d mbufs only\n",
+				rxq->queue_id, i, ring->ring_size);
+			break;
+		}
+		rxr->rx_prod = prod;
+		prod = RING_NEXT(rxr->rx_ring_struct, prod);
+	}
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
new file mode 100644
index 0000000..e9bae3f
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -0,0 +1,62 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RXR_H_
+#define _BNXT_RXR_H_
+
+#define B_RX_DB(db, prod)						\
+		(*(uint32_t *)db = (DB_KEY_RX | prod))
+
+struct bnxt_sw_rx_bd {
+	struct rte_mbuf		*mbuf; /* data associated with RX descriptor */
+};
+
+struct bnxt_rx_ring_info {
+	uint16_t		rx_prod;
+	void			*rx_doorbell;
+
+	struct rx_prod_pkt_bd	*rx_desc_ring;
+	struct bnxt_sw_rx_bd	*rx_buf_ring; /* sw ring */
+
+	phys_addr_t		rx_desc_mapping;
+
+	struct bnxt_ring_struct	*rx_ring_struct;
+};
+
+uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+			       uint16_t nb_pkts);
+void bnxt_free_rx_rings(struct bnxt *bp);
+void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq);
+int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 1543f20..8b30787 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -446,6 +446,79 @@ struct tx_bd_long_hi {
 	uint32_t cfa_meta;
 } __attribute__((packed));
 
+/* RX Producer Packet BD (16 bytes) */
+struct rx_prod_pkt_bd {
+	/* This value identifies the type of buffer descriptor. */
+	#define RX_PROD_PKT_BD_TYPE_MASK		UINT32_C(0x3f)
+	#define RX_PROD_PKT_BD_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 16B long and is an RX Producer (ie.
+		 * empty) buffer descriptor.
+		 */
+	#define RX_PROD_PKT_BD_TYPE_RX_PROD_PKT		(UINT32_C(0x4) << 0)
+	/*
+	 * If set to 1, the packet will be placed at the address plus 2B. The 2
+	 * Bytes of padding will be written as zero.
+	 */
+	/*
+	 * This is intended to be used when the host buffer is cache-line
+	 * aligned to produce packets that are easy to parse in host memory
+	 * while still allowing writes to be cache line aligned.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_SOP_PAD		UINT32_C(0x40)
+	/*
+	 * If set to 1, the packet write will be padded out to the nearest
+	 * cache-line with zero value padding.
+	 */
+	/*
+	 * If receive buffers start/end on cache-line boundaries, this feature
+	 * will ensure that all data writes on the PCI bus start/end on cache
+	 * line boundaries.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_EOP_PAD		UINT32_C(0x80)
+	/*
+	 * This value is the number of additional buffers in the ring that
+	 * describe the buffer space to be consumed for the this packet. If the
+	 * value is zero, then the packet must fit within the space described by
+	 * this BD. If this value is 1 or more, it indicates how many additional
+	 * "buffer" BDs are in the ring immediately following this BD to be used
+	 * for the same network packet. Even if the packet to be placed does not
+	 * need all the additional buffers, they will be consumed anyway.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_BUFFERS_MASK	UINT32_C(0x300)
+	#define RX_PROD_PKT_BD_FLAGS_BUFFERS_SFT	8
+	#define RX_PROD_PKT_BD_FLAGS_MASK		UINT32_C(0xffc0)
+	#define RX_PROD_PKT_BD_FLAGS_SFT		6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length in Bytes of the host physical buffer where data
+	 * for the packet may be placed in host memory.
+	 */
+	/*
+	 * While this is a Byte resolution value, it is often advantageous to
+	 * ensure that the buffers provided end on a host cache line.
+	 */
+	uint16_t len;
+
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with this
+	 * receive buffer set.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address where data for the packet may by
+	 * placed in host memory.
+	 */
+	/*
+	 * While this is a Byte resolution value, it is often advantageous to
+	 * ensure that the buffers provide start on a host cache line.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
 /* Completion Ring Structures */
 /* Note: This structure is used by the HWRM to communicate HWRM Error. */
 /* Base Completion Record (16 bytes) */
@@ -611,6 +684,407 @@ struct tx_cmpl {
 	uint32_t unused_2;
 } __attribute__((packed)) tx_cmpl_t, *ptx_cmpl_t;
 
+/* RX Packet Completion Record (32 bytes split to 2 16-byte struct) */
+struct rx_pkt_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define RX_PKT_CMPL_TYPE_MASK			UINT32_C(0x3f)
+	#define RX_PKT_CMPL_TYPE_SFT			0
+		/*
+		 * RX L2 completion: Completion of and L2 RX packet.
+		 * Length = 32B
+		 */
+	#define RX_PKT_CMPL_TYPE_RX_L2			(UINT32_C(0x11) << 0)
+	/*
+	 * When this bit is '1', it indicates a packet that has an error of some
+	 * type. Type of error is indicated in error_flags.
+	 */
+	#define RX_PKT_CMPL_FLAGS_ERROR			UINT32_C(0x40)
+	/* This field indicates how the packet was placed in the buffer. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_MASK	UINT32_C(0x380)
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_SFT		7
+		/* Normal: Packet was placed using normal algorithm. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_NORMAL	(UINT32_C(0x0) << 7)
+		/* Jumbo: Packet was placed using jumbo algorithm. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_JUMBO	(UINT32_C(0x1) << 7)
+		/*
+		 * Header/Data Separation: Packet was placed using Header/Data
+		 * separation algorithm. The separation location is indicated by
+		 * the itype field.
+		 */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_HDS		(UINT32_C(0x2) << 7)
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_LAST \
+						RX_PKT_CMPL_FLAGS_PLACEMENT_HDS
+	/* This bit is '1' if the RSS field in this completion is valid. */
+	#define RX_PKT_CMPL_FLAGS_RSS_VALID		UINT32_C(0x400)
+	/*
+	 * This value indicates what the inner packet determined for the packet
+	 * was.
+	 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_MASK		UINT32_C(0xf000)
+	#define RX_PKT_CMPL_FLAGS_ITYPE_SFT		12
+		/* Not Known: Indicates that the packet type was not known. */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_NOT_KNOWN	(UINT32_C(0x0) << 12)
+		/*
+		 * IP Packet: Indicates that the packet was an IP packet, but
+		 * further classification was not possible.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_IP		(UINT32_C(0x1) << 12)
+		/*
+		 * TCP Packet: Indicates that the packet was IP and TCP. This
+		 * indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_TCP		(UINT32_C(0x2) << 12)
+		/*
+		 * UDP Packet: Indicates that the packet was IP and UDP. This
+		 * indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_UDP		(UINT32_C(0x3) << 12)
+		/*
+		 * FCoE Packet: Indicates that the packet was recognized as a
+		 * FCoE. This also indicates that the payload_offset field is
+		 * valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_FCOE		(UINT32_C(0x4) << 12)
+		/*
+		 * RoCE Packet: Indicates that the packet was recognized as a
+		 * RoCE. This also indicates that the payload_offset field is
+		 * valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_ROCE		(UINT32_C(0x5) << 12)
+		/*
+		 * ICMP Packet: Indicates that the packet was recognized as
+		 * ICMP. This indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_ICMP		(UINT32_C(0x7) << 12)
+		/*
+		 * PtP packet wo/timestamp: Indicates that the packet was
+		 * recognized as a PtP packet.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_PTP_WO_TIMESTAMP \
+							(UINT32_C(0x8) << 12)
+		/*
+		 * PtP packet w/timestamp: Indicates that the packet was
+		 * recognized as a PtP packet and that a timestamp was taken for
+		 * the packet.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_PTP_W_TIMESTAMP	(UINT32_C(0x9) << 12)
+	#define RX_PKT_CMPL_FLAGS_ITYPE_LAST \
+					RX_PKT_CMPL_FLAGS_ITYPE_PTP_W_TIMESTAMP
+	#define RX_PKT_CMPL_FLAGS_MASK			UINT32_C(0xffc0)
+	#define RX_PKT_CMPL_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the data for the packet stored in the buffer(s)
+	 * identified by the opaque value. This includes the packet BD and any
+	 * associated buffer BDs. This does not include the the length of any
+	 * data places in aggregation BDs.
+	 */
+	uint16_t len;
+
+	/*
+	 * This is a copy of the opaque field from the RX BD this completion
+	 * corresponds to.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define RX_PKT_CMPL_V1				UINT32_C(0x1)
+	/*
+	 * This value is the number of aggregation buffers that follow this
+	 * entry in the completion ring that are a part of this packet. If the
+	 * value is zero, then the packet is completely contained in the buffer
+	 * space provided for the packet in the RX ring.
+	 */
+	#define RX_PKT_CMPL_AGG_BUFS_MASK		UINT32_C(0x3e)
+	#define RX_PKT_CMPL_AGG_BUFS_SFT		1
+	uint8_t agg_bufs_v1;
+
+	/*
+	 * This is the RSS hash type for the packet. The value is packed
+	 * {tuple_extrac_op[1:0],rss_profile_id[4:0],tuple_extrac_op[2]}.
+	 */
+	uint8_t rss_hash_type;
+
+	/*
+	 * This value indicates the offset from the beginning of the packet
+	 * where the inner payload starts. This value is valid for TCP, UDP,
+	 * FCoE, and RoCE packets.
+	 */
+	uint8_t payload_offset;
+
+	uint8_t unused_1;
+
+	/*
+	 * This value is the RSS hash value calculated for the packet based on
+	 * the mode bits and key value in the VNIC.
+	 */
+	uint32_t rss_hash;
+} __attribute__((packed));
+
+/* last 16 bytes of RX Packet Completion Record */
+struct rx_pkt_cmpl_hi {
+	/*
+	 * This indicates that the ip checksum was calculated for the inner
+	 * packet and that the ip_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_IP_CS_CALC		UINT32_C(0x1)
+	/*
+	 * This indicates that the TCP, UDP or ICMP checksum was calculated for
+	 * the inner packet and that the l4_cs_error field indicates if there
+	 * was an error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_L4_CS_CALC		UINT32_C(0x2)
+	/*
+	 * This indicates that the ip checksum was calculated for the tunnel
+	 * header and that the t_ip_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_T_IP_CS_CALC		UINT32_C(0x4)
+	/*
+	 * This indicates that the UDP checksum was calculated for the tunnel
+	 * packet and that the t_l4_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_T_L4_CS_CALC		UINT32_C(0x8)
+	/* This value indicates what format the metadata field is. */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_MASK	UINT32_C(0xf0)
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_SFT	4
+		/* No metadata informtaion. Value is zero. */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_NONE	(UINT32_C(0x0) << 4)
+		/*
+		 * The metadata field contains the VLAN tag and TPID value. -
+		 * metadata[11:0] contains the vlan VID value. - metadata[12]
+		 * contains the vlan DE value. - metadata[15:13] contains the
+		 * vlan PRI value. - metadata[31:16] contains the vlan TPID
+		 * value.
+		 */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN	(UINT32_C(0x1) << 4)
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_LAST \
+					RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN
+	/*
+	 * This field indicates the IP type for the inner-most IP header. A
+	 * value of '0' indicates IPv4. A value of '1' indicates IPv6. This
+	 * value is only valid if itype indicates a packet with an IP header.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_IP_TYPE		UINT32_C(0x100)
+	uint32_t flags2;
+
+	/*
+	 * This is data from the CFA block as indicated by the meta_format
+	 * field.
+	 */
+	/* When meta_format=1, this value is the VLAN VID. */
+	#define RX_PKT_CMPL_METADATA_VID_MASK		UINT32_C(0xfff)
+	#define RX_PKT_CMPL_METADATA_VID_SFT		0
+	/* When meta_format=1, this value is the VLAN DE. */
+	#define RX_PKT_CMPL_METADATA_DE			UINT32_C(0x1000)
+	/* When meta_format=1, this value is the VLAN PRI. */
+	#define RX_PKT_CMPL_METADATA_PRI_MASK		UINT32_C(0xe000)
+	#define RX_PKT_CMPL_METADATA_PRI_SFT		13
+	/* When meta_format=1, this value is the VLAN TPID. */
+	#define RX_PKT_CMPL_METADATA_TPID_MASK		UINT32_C(0xffff0000)
+	#define RX_PKT_CMPL_METADATA_TPID_SFT		16
+	uint32_t metadata;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define RX_PKT_CMPL_V2				UINT32_C(0x1)
+	/*
+	 * This error indicates that there was some sort of problem with the BDs
+	 * for the packet that was found after part of the packet was already
+	 * placed. The packet should be treated as invalid.
+	 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_MASK	UINT32_C(0xe)
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_SFT	1
+		/* No buffer error */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_NO_BUFFER \
+							(UINT32_C(0x0) << 1)
+		/*
+		 * Did Not Fit: Packet did not fit into packet buffer provided.
+		 * For regular placement, this means the packet did not fit in
+		 * the buffer provided. For HDS and jumbo placement, this means
+		 * that the packet could not be placed into 7 physical buffers
+		 * or less.
+		 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_DID_NOT_FIT \
+							(UINT32_C(0x1) << 1)
+		/*
+		 * Not On Chip: All BDs needed for the packet were not on-chip
+		 * when the packet arrived.
+		 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_NOT_ON_CHIP \
+							(UINT32_C(0x2) << 1)
+		/* Bad Format: BDs were not formatted correctly. */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_BAD_FORMAT \
+							(UINT32_C(0x3) << 1)
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_BUFFER_ERROR_BAD_FORMAT
+	/* This indicates that there was an error in the IP header checksum. */
+	#define RX_PKT_CMPL_ERRORS_IP_CS_ERROR		UINT32_C(0x10)
+	/*
+	 * This indicates that there was an error in the TCP, UDP or ICMP
+	 * checksum.
+	 */
+	#define RX_PKT_CMPL_ERRORS_L4_CS_ERROR		UINT32_C(0x20)
+	/*
+	 * This indicates that there was an error in the tunnel IP header
+	 * checksum.
+	 */
+	#define RX_PKT_CMPL_ERRORS_T_IP_CS_ERROR	UINT32_C(0x40)
+	/* This indicates that there was an error in the tunnel UDP checksum. */
+	#define RX_PKT_CMPL_ERRORS_T_L4_CS_ERROR	UINT32_C(0x80)
+	/*
+	 * This indicates that there was a CRC error on either an FCoE or RoCE
+	 * packet. The itype indicates the packet type.
+	 */
+	#define RX_PKT_CMPL_ERRORS_CRC_ERROR		UINT32_C(0x100)
+	/*
+	 * This indicates that there was an error in the tunnel portion of the
+	 * packet when this field is non-zero.
+	 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_MASK	UINT32_C(0xe00)
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_SFT	9
+		/*
+		 * No additional error occurred on the tunnel portion of the
+		 * packet of the packet does not have a tunnel.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_NO_ERROR	(UINT32_C(0x0) << 9)
+		/*
+		 * Indicates that IP header version does not match expectation
+		 * from L2 Ethertype for IPv4 and IPv6 in the tunnel header.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_VERSION \
+							(UINT32_C(0x1) << 9)
+		/*
+		 * Indicates that header length is out of range in the tunnel
+		 * header. Valid for IPv4.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_HDR_LEN \
+							(UINT32_C(0x2) << 9)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the PPPoE header length for a tunnel PPPoE packet.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_TUNNEL_TOTAL_ERROR \
+							(UINT32_C(0x3) << 9)
+		/*
+		 * Indicates that physical packet is shorter than that claimed
+		 * by the tunnel l3 header length. Valid for IPv4, or IPv6
+		 * tunnel packet packets.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_IP_TOTAL_ERROR \
+							(UINT32_C(0x4) << 9)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the tunnel UDP header length for a tunnel UDP
+		 * packet that is not fragmented.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_UDP_TOTAL_ERROR \
+							(UINT32_C(0x5) << 9)
+		/*
+		 * indicates that the IPv4 TTL or IPv6 hop limit check have
+		 * failed (e.g. TTL = 0) in the tunnel header. Valid for IPv4,
+		 * and IPv6.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_TTL \
+							(UINT32_C(0x6) << 9)
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_TTL
+	/*
+	 * This indicates that there was an error in the inner portion of the
+	 * packet when this field is non-zero.
+	 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_MASK	UINT32_C(0xf000)
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_SFT	12
+		/*
+		 * No additional error occurred on the tunnel portion of the
+		 * packet of the packet does not have a tunnel.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_NO_ERROR	(UINT32_C(0x0) << 12)
+		/*
+		 * Indicates that IP header version does not match expectation
+		 * from L2 Ethertype for IPv4 and IPv6 or that option other than
+		 * VFT was parsed on FCoE packet.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_VERSION \
+							(UINT32_C(0x1) << 12)
+		/*
+		 * indicates that header length is out of range. Valid for IPv4
+		 * and RoCE
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_HDR_LEN \
+							(UINT32_C(0x2) << 12)
+		/*
+		 * indicates that the IPv4 TTL or IPv6 hop limit check have
+		 * failed (e.g. TTL = 0). Valid for IPv4, and IPv6
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_TTL	(UINT32_C(0x3) << 12)
+		/*
+		 * Indicates that physical packet is shorter than that claimed
+		 * by the l3 header length. Valid for IPv4, IPv6 packet or RoCE
+		 * packets.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_IP_TOTAL_ERROR \
+							(UINT32_C(0x4) << 12)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the UDP header length for a UDP packet that is not
+		 * fragmented.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_UDP_TOTAL_ERROR \
+							(UINT32_C(0x5) << 12)
+		/*
+		 * Indicates that TCP header length > IP payload. Valid for TCP
+		 * packets only.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_HDR_LEN \
+							(UINT32_C(0x6) << 12)
+		/* Indicates that TCP header length < 5. Valid for TCP. */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_HDR_LEN_TOO_SMALL \
+							(UINT32_C(0x7) << 12)
+		/*
+		 * Indicates that TCP option headers result in a TCP header size
+		 * that does not match data offset in TCP header. Valid for TCP.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_OPT_LEN \
+							(UINT32_C(0x8) << 12)
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_OPT_LEN
+	#define RX_PKT_CMPL_ERRORS_MASK			UINT32_C(0xfffe)
+	#define RX_PKT_CMPL_ERRORS_SFT			1
+	uint16_t errors_v2;
+
+	/*
+	 * This field identifies the CFA action rule that was used for this
+	 * packet.
+	 */
+	uint16_t cfa_code;
+
+	/*
+	 * This value holds the reordering sequence number for the packet. If
+	 * the reordering sequence is not valid, then this value is zero. The
+	 * reordering domain for the packet is in the bottom 8 to 10b of the
+	 * rss_hash value. The bottom 20b of this value contain the ordering
+	 * domain value for the packet.
+	 */
+	#define RX_PKT_CMPL_REORDER_MASK		UINT32_C(0xffffff)
+	#define RX_PKT_CMPL_REORDER_SFT			0
+	uint32_t reorder;
+} __attribute__((packed));
+
 /* HWRM Forwarded Request (16 bytes) */
 struct hwrm_fwd_req_cmpl {
 	/* Length of forwarded request in bytes. */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 15/39] bnxt: Code to alloc/free ring
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (12 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 14/39] bnxt: initial Rx " Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 16/39] bnxt: add HWRM function reset command Stephen Hurd
                   ` (26 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Perform allocation and free()ing of ring and information structures
for the TX, RX, and completion rings. The previous patches had
so far provided top level stubs, while this patch does the real
allocation and freeing of the memory.

v4:
- Address review comments and fix issues pointed out by checkpatch.
- Change the argument passed to bnxt_alloc_rings.
 Instead of passing bnxt_tx_ring and bnxt_rx_ring,
 shorten them to txr and rxr respectively.
- Add code to free the reserved memzone

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_cpr.c  | 28 +++++++++++++++++++++++-----
 drivers/net/bnxt/bnxt_cpr.h  |  2 +-
 drivers/net/bnxt/bnxt_ring.c |  4 ++++
 drivers/net/bnxt/bnxt_ring.h |  1 +
 drivers/net/bnxt/bnxt_rxq.c  | 22 +++++++++++++---------
 drivers/net/bnxt/bnxt_rxr.c  | 42 ++++++++++++++++++++++++++++++++++--------
 drivers/net/bnxt/bnxt_rxr.h  |  2 +-
 drivers/net/bnxt/bnxt_txq.c  | 28 +++++++++++++++++-----------
 drivers/net/bnxt/bnxt_txr.c  | 43 ++++++++++++++++++++++++++++++++++---------
 drivers/net/bnxt/bnxt_txr.h  |  2 +-
 10 files changed, 129 insertions(+), 45 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index 06a618a..98f3ca2 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -31,6 +31,8 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_malloc.h>
+
 #include "bnxt.h"
 #include "bnxt_cpr.h"
 #include "bnxt_hwrm.h"
@@ -121,21 +123,37 @@ reject:
 void bnxt_free_def_cp_ring(struct bnxt *bp)
 {
 	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-	struct bnxt_ring_struct *ring = cpr->cp_ring_struct;
 
-	bnxt_free_ring(ring);
+	bnxt_free_ring(cpr->cp_ring_struct);
+	rte_free(cpr->cp_ring_struct);
+	rte_free(cpr);
 }
 
 /* For the default completion ring only */
-void bnxt_init_def_ring_struct(struct bnxt *bp)
+int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id)
 {
-	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-	struct bnxt_ring_struct *ring = cpr->cp_ring_struct;
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_ring_struct *ring;
 
+	cpr = rte_zmalloc_socket("bnxt_cp_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	bp->def_cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_cp_ring_struct",
+				  sizeof(struct bnxt_ring_struct),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->ring_size = rte_align32pow2(DEFAULT_CP_RING_SIZE);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index f104281..3e25a75 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -79,7 +79,7 @@ struct bnxt_cp_ring_info {
 
 struct bnxt;
 void bnxt_free_def_cp_ring(struct bnxt *bp);
-void bnxt_init_def_ring_struct(struct bnxt *bp);
+int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id);
 void bnxt_handle_async_event(struct bnxt *bp, struct cmpl_base *cmp);
 void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmp);
 
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index fab0e27..675c5b8 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -51,6 +51,7 @@ void bnxt_free_ring(struct bnxt_ring_struct *ring)
 		memset((char *)*ring->vmem, 0, ring->vmem_size);
 		*ring->vmem = NULL;
 	}
+	rte_memzone_free((const struct rte_memzone *)ring->mem_zone);
 }
 
 /*
@@ -134,6 +135,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		tx_ring_info->tx_desc_ring = (struct tx_bd_long *)tx_ring->bd;
 		tx_ring->bd_dma = mz->phys_addr + tx_ring_start;
 		tx_ring_info->tx_desc_mapping = tx_ring->bd_dma;
+		tx_ring->mem_zone = (const void *)mz;
 
 		if (!tx_ring->bd)
 			return -ENOMEM;
@@ -153,6 +155,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		    (struct rx_prod_pkt_bd *)rx_ring->bd;
 		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
 		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+		rx_ring->mem_zone = (const void *)mz;
 
 		if (!rx_ring->bd)
 			return -ENOMEM;
@@ -168,6 +171,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
 	cp_ring_info->cp_desc_ring = cp_ring->bd;
 	cp_ring_info->cp_desc_mapping = cp_ring->bd_dma;
+	cp_ring->mem_zone = (const void *)mz;
 
 	if (!cp_ring->bd)
 		return -ENOMEM;
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index 281587c..c7c3b33 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -76,6 +76,7 @@ struct bnxt_ring_struct {
 	void			**vmem;
 
 	uint16_t		fw_ring_id; /* Ring id filled by Chimp FW */
+	const void		*mem_zone;
 };
 
 struct bnxt_ring_grp_info {
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index 13fd08c..cddf17d 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -53,9 +53,6 @@ void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq)
 {
 	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
 
-	/* 'Unreserve' rte_memzone */
-	/* N/A */
-
 	if (cpr->hw_stats)
 		cpr->hw_stats = NULL;
 }
@@ -273,10 +270,12 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
 	struct bnxt_rx_queue *rxq;
+	int rc = 0;
 
 	if (!nb_desc || nb_desc > MAX_RX_DESC_CNT) {
 		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	if (eth_dev->data->rx_queues) {
@@ -288,14 +287,17 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 				 RTE_CACHE_LINE_SIZE, socket_id);
 	if (!rxq) {
 		RTE_LOG(ERR, PMD, "bnxt_rx_queue allocation failed!");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 	rxq->bp = bp;
 	rxq->mb_pool = mp;
 	rxq->nb_rx_desc = nb_desc;
 	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
 
-	bnxt_init_rx_ring_struct(rxq);
+	rc = bnxt_init_rx_ring_struct(rxq, socket_id);
+	if (rc)
+		goto out;
 
 	rxq->queue_id = queue_idx;
 	rxq->port_id = eth_dev->data->port_id;
@@ -305,11 +307,13 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	eth_dev->data->rx_queues[queue_idx] = rxq;
 	/* Allocate RX ring hardware descriptors */
 	if (bnxt_alloc_rings(bp, queue_idx, NULL, rxq->rx_ring, rxq->cp_ring,
-			"bnxt_rx_ring")) {
+			"rxr")) {
 		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for rx_ring failed!");
 		bnxt_rx_queue_release_op(rxq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
-	return 0;
+out:
+	return rc;
 }
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
index 43d6979..0ac6f21 100644
--- a/drivers/net/bnxt/bnxt_rxr.c
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -255,17 +255,20 @@ void bnxt_free_rx_rings(struct bnxt *bp)
 		if (!rxq)
 			continue;
 
-		/* TODO: free() rxq->rx_ring and rxq->rx_ring->rx_ring_struct */
 		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
-		/* TODO: free() rxq->cp_ring and rxq->cp_ring->cp_ring_struct */
+		rte_free(rxq->rx_ring->rx_ring_struct);
+		rte_free(rxq->rx_ring);
+
 		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+		rte_free(rxq->cp_ring->cp_ring_struct);
+		rte_free(rxq->cp_ring);
 
 		rte_free(rxq);
 		bp->rx_queues[i] = NULL;
 	}
 }
 
-void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
+int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 {
 	struct bnxt *bp = rxq->bp;
 	struct bnxt_cp_ring_info *cpr;
@@ -277,8 +280,19 @@ void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
 			       (2 * VLAN_TAG_SIZE);
 	rxq->rx_buf_size = rxq->rx_buf_use_size + sizeof(struct rte_mbuf);
 
-	rxr = rxq->rx_ring;
-	ring = rxr->rx_ring_struct;
+	rxr = rte_zmalloc_socket("bnxt_rx_ring",
+				 sizeof(struct bnxt_rx_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxr == NULL)
+		return -ENOMEM;
+	rxq->rx_ring = rxr;
+
+	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
+				   sizeof(struct bnxt_ring_struct),
+				   RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	rxr->rx_ring_struct = ring;
 	ring->ring_size = rte_align32pow2(rxq->nb_rx_desc);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)rxr->rx_desc_ring;
@@ -286,14 +300,27 @@ void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
 	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_rx_bd);
 	ring->vmem = (void **)&rxr->rx_buf_ring;
 
-	cpr = rxq->cp_ring;
-	ring = cpr->cp_ring_struct;
+	cpr = rte_zmalloc_socket("bnxt_rx_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	rxq->cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
+				   sizeof(struct bnxt_ring_struct),
+				   RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->ring_size = rxr->rx_ring_struct->ring_size * 2;
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
 
 static void bnxt_init_rxbds(struct bnxt_ring_struct *ring, uint32_t type,
@@ -320,7 +347,6 @@ int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
 
 	type = RX_PROD_PKT_BD_TYPE_RX_PROD_PKT | RX_PROD_PKT_BD_FLAGS_EOP_PAD;
 
-	/* TODO: These need to be allocated */
 	rxr = rxq->rx_ring;
 	ring = rxr->rx_ring_struct;
 	bnxt_init_rxbds(ring, type, rxq->rx_buf_use_size);
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
index e9bae3f..7ba8f7b 100644
--- a/drivers/net/bnxt/bnxt_rxr.h
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -56,7 +56,7 @@ struct bnxt_rx_ring_info {
 uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 			       uint16_t nb_pkts);
 void bnxt_free_rx_rings(struct bnxt *bp);
-void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq);
+int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id);
 int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq);
 
 #endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 7aba199..99ddddd 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -49,9 +49,6 @@ void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
 {
 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
 
-	/* 'Unreserve' rte_memzone */
-	/* N/A */
-
 	if (cpr->hw_stats)
 		cpr->hw_stats = NULL;
 }
@@ -109,10 +106,12 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
 	struct bnxt_tx_queue *txq;
+	int rc = 0;
 
 	if (!nb_desc || nb_desc > MAX_TX_DESC_CNT) {
 		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	if (eth_dev->data->tx_queues) {
@@ -124,33 +123,40 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	}
 	txq = rte_zmalloc_socket("bnxt_tx_queue", sizeof(struct bnxt_tx_queue),
 				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq == NULL) {
+	if (!txq) {
 		RTE_LOG(ERR, PMD, "bnxt_tx_queue allocation failed!");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 	txq->bp = bp;
 	txq->nb_tx_desc = nb_desc;
 	txq->tx_free_thresh = tx_conf->tx_free_thresh;
 
-	bnxt_init_tx_ring_struct(txq);
+	rc = bnxt_init_tx_ring_struct(txq, socket_id);
+	if (rc)
+		goto out;
 
 	txq->queue_id = queue_idx;
 	txq->port_id = eth_dev->data->port_id;
 
 	/* Allocate TX ring hardware descriptors */
 	if (bnxt_alloc_rings(bp, queue_idx, txq->tx_ring, NULL, txq->cp_ring,
-			"bnxt_tx_ring")) {
+			"txr")) {
 		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for tx_ring failed!");
 		bnxt_tx_queue_release_op(txq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
 	if (bnxt_init_one_tx_ring(txq)) {
 		RTE_LOG(ERR, PMD, "bnxt_init_one_tx_ring failed!");
 		bnxt_tx_queue_release_op(txq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
 	eth_dev->data->tx_queues[queue_idx] = txq;
-	return 0;
+
+out:
+	return rc;
 }
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
index 2314410..eef01c7 100644
--- a/drivers/net/bnxt/bnxt_txr.c
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -59,9 +59,12 @@ void bnxt_free_tx_rings(struct bnxt *bp)
 			continue;
 
 		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
-		/* TODO: free() txq->tx_ring and txq->tx_ring->tx_ring_struct */
+		rte_free(txq->tx_ring->tx_ring_struct);
+		rte_free(txq->tx_ring);
+
 		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
-		/* TODO: free() txq->cp_ring and txq->cp_ring->cp_ring_struct */
+		rte_free(txq->cp_ring->cp_ring_struct);
+		rte_free(txq->cp_ring);
 
 		rte_free(txq);
 		bp->tx_queues[i] = NULL;
@@ -79,15 +82,25 @@ int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
 	return 0;
 }
 
-void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
+int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id)
 {
 	struct bnxt_cp_ring_info *cpr;
 	struct bnxt_tx_ring_info *txr;
 	struct bnxt_ring_struct *ring;
 
-	/* TODO: These need to be allocated */
-	txr = txq->tx_ring;
-	ring = txr->tx_ring_struct;
+	txr = rte_zmalloc_socket("bnxt_tx_ring",
+				 sizeof(struct bnxt_tx_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txr == NULL)
+		return -ENOMEM;
+	txq->tx_ring = txr;
+
+	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
+				  sizeof(struct bnxt_ring_struct),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	txr->tx_ring_struct = ring;
 	ring->ring_size = rte_align32pow2(txq->nb_tx_desc + 1);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)txr->tx_desc_ring;
@@ -95,15 +108,27 @@ void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
 	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_tx_bd);
 	ring->vmem = (void **)&txr->tx_buf_ring;
 
-	/* TODO: These need to be allocated */
-	cpr = txq->cp_ring;
-	ring = cpr->cp_ring_struct;
+	cpr = rte_zmalloc_socket("bnxt_tx_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	txq->cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
+				  sizeof(struct bnxt_ring_struct),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->ring_size = txr->tx_ring_struct->ring_size;
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
 
 static inline uint32_t bnxt_tx_avail(struct bnxt_tx_ring_info *txr)
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
index 1797a3d..56d0b08 100644
--- a/drivers/net/bnxt/bnxt_txr.h
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -64,7 +64,7 @@ struct bnxt_sw_tx_bd {
 
 void bnxt_free_tx_rings(struct bnxt *bp);
 int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
-void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq);
+int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id);
 uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 			       uint16_t nb_pkts);
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 16/39] bnxt: add HWRM function reset command
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (13 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 15/39] bnxt: Code to alloc/free ring Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 17/39] bnxt: add HWRM vnic alloc function Stephen Hurd
                   ` (25 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add bnxt_hwrm_func_reset() function and supporting structs and macros.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           |  17 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 129 +++++++++++++++++++++++++++++++++
 3 files changed, 147 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 2574bd0..3400083 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -221,6 +221,23 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_func_reset(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_func_reset_input req = {.req_type = 0 };
+	struct hwrm_func_reset_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, FUNC_RESET, -1, resp);
+
+	req.enables = rte_cpu_to_le_32(0);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd)
 {
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 0861417..4fa94aa 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -50,6 +50,7 @@ int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+int bnxt_hwrm_func_reset(struct bnxt *bp);
 int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 8b30787..6209368 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -83,6 +83,7 @@ struct ctx_hw_stats64 {
  * Request types
  */
 #define HWRM_VER_GET			(UINT32_C(0x0))
+#define HWRM_FUNC_RESET			(UINT32_C(0x11))
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
@@ -2048,6 +2049,134 @@ struct hwrm_func_qcaps_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_func_reset */
+/*
+ * Description: This command resets a hardware function (PCIe function) and
+ * frees any resources used by the function. This command shall be initiated by
+ * the driver after an FLR has occurred to prepare the function for re-use. This
+ * command may also be initiated by a driver prior to doing it's own
+ * configuration. This command puts the function into the reset state. In the
+ * reset state, global and port related features of the chip are not available.
+ */
+/*
+ * Note: This command will reset a function that has already been disabled or
+ * idled. The command returns all the resources owned by the function so a new
+ * driver may allocate and configure resources normally.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_func_reset_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This bit must be '1' for the vf_id_valid field to be configured. */
+	#define HWRM_FUNC_RESET_INPUT_ENABLES_VF_ID_VALID \
+							UINT32_C(0x1)
+	uint32_t enables;
+
+	/*
+	 * The ID of the VF that this PF is trying to reset. Only the parent PF
+	 * shall be allowed to reset a child VF. A parent PF driver shall use
+	 * this field only when a specific child VF is requested to be reset.
+	 */
+	uint16_t vf_id;
+
+	/* This value indicates the level of a function reset. */
+		/*
+		 * Reset the caller function and its children VFs (if any). If
+		 * no children functions exist, then reset the caller function
+		 * only.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETALL \
+							(UINT32_C(0x0) << 0)
+		/* Reset the caller function only */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETME \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Reset all children VFs of the caller function driver if the
+		 * caller is a PF driver. It is an error to specify this level
+		 * by a VF driver. It is an error to specify this level by a PF
+		 * driver with no children VFs.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETCHILDREN \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Reset a specific VF of the caller function driver if the
+		 * caller is the parent PF driver. It is an error to specify
+		 * this level by a VF driver. It is an error to specify this
+		 * level by a PF driver that is not the parent of the VF that is
+		 * being requested to reset.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETVF \
+							(UINT32_C(0x3) << 0)
+	uint8_t func_reset_level;
+
+	uint8_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_func_reset_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_port_phy_cfg */
 /*
  * Description: This command configures the PHY device for the port. It allows
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 17/39] bnxt: add HWRM vnic alloc function
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (14 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 16/39] bnxt: add HWRM function reset command Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 18/39] bnxt: add HWRM vnic free function Stephen Hurd
                   ` (24 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This requires a group info array in struct bnxt, so add that, save
the max size from the func_qcap response, and alloc/free in init/uninit

As mentioned in the previous patch, A VNIC represents a virtual interface.
It is a resource in the RX path of the chip and is used to setup various
target actions such as RSS, MAC filtering etc.. for the physical function
in use.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt.h                |  2 +
 drivers/net/bnxt/bnxt_hwrm.c           | 33 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 99 ++++++++++++++++++++++++++++++++++
 4 files changed, 136 insertions(+)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index f7cf9d1..df1f771 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -141,6 +141,8 @@ struct bnxt {
 
 	/* Default completion ring */
 	struct bnxt_cp_ring_info	*def_cp_ring;
+	uint32_t		max_ring_grps;
+	struct bnxt_ring_grp_info	*grp_info;
 
 	unsigned int		nr_vnics;
 
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 3400083..77afb81 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -43,7 +43,9 @@
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
 #define HWRM_CMD_TIMEOUT		2000
@@ -191,6 +193,7 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 
 	HWRM_CHECK_RESULT;
 
+	bp->max_ring_grps = rte_le_to_cpu_32(resp->max_hw_ring_grps);
 	if (BNXT_PF(bp)) {
 		struct bnxt_pf_info *pf = &bp->pf;
 
@@ -477,6 +480,36 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0, i, j;
+	struct hwrm_vnic_alloc_input req = {.req_type = 0 };
+	struct hwrm_vnic_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	/* map ring groups to this vnic */
+	for (i = vnic->start_grp_id, j = 0; i <= vnic->end_grp_id; i++, j++) {
+		if (bp->grp_info[i].fw_grp_id == (uint16_t)HWRM_NA_SIGNATURE) {
+			RTE_LOG(ERR, PMD,
+				"Not enough ring groups avail:%x req:%x\n", j,
+				(vnic->end_grp_id - vnic->start_grp_id) + 1);
+			break;
+		}
+		vnic->fw_grp_ids[j] = bp->grp_info[i].fw_grp_id;
+	}
+
+	vnic->fw_rss_cos_lb_ctx = (uint16_t)HWRM_NA_SIGNATURE;
+	vnic->ctx_is_rss_cos_lb = HW_CONTEXT_NONE;
+
+	HWRM_PREP(req, VNIC_ALLOC, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_vnic_id = rte_le_to_cpu_16(resp->vnic_id);
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 4fa94aa..62dc801 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -59,6 +59,8 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 6209368..eedd368 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -89,6 +89,7 @@ struct ctx_hw_stats64 {
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+#define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3119,6 +3120,104 @@ struct hwrm_stat_ctx_clr_stats_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_alloc */
+/*
+ * Description: This VNIC is a resource in the RX side of the chip that is used
+ * to represent a virtual host "interface". # At the time of VNIC allocation or
+ * configuration, the function can specify whether it wants the requested VNIC
+ * to be the default VNIC for the function or not. # If a function requests
+ * allocation of a VNIC for the first time and a VNIC is successfully allocated
+ * by the HWRM, then the HWRM shall make the allocated VNIC as the default VNIC
+ * for that function. # The default VNIC shall be used for the default action
+ * for a partition or function. # For each VNIC allocated on a function, a
+ * mapping on the RX side to map the allocated VNIC to source virtual interface
+ * shall be performed by the HWRM. This should be hidden to the function driver
+ * requesting the VNIC allocation. This enables broadcast/multicast replication
+ * with source knockout. # If multicast replication with source knockout is
+ * enabled, then the internal VNIC to SVIF mapping data structures shall be
+ * programmed at the time of VNIC allocation.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_vnic_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', this VNIC is requested to be the default VNIC
+	 * for this function.
+	 */
+	#define HWRM_VNIC_ALLOC_INPUT_FLAGS_DEFAULT                UINT32_C(0x1)
+	uint32_t flags;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* Logical vnic ID */
+	uint32_t vnic_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 18/39] bnxt: add HWRM vnic free function
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (15 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 17/39] bnxt: add HWRM vnic alloc function Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 19/39] bnxt: add HWRM vnic configure function Stephen Hurd
                   ` (23 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Frees a vnic allocated by vnic_alloc in the previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           | 21 +++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 82 ++++++++++++++++++++++++++++++++++
 3 files changed, 104 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 77afb81..02fb0c4 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -510,6 +510,27 @@ int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_free_input req = {.req_type = 0 };
+	struct hwrm_vnic_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (vnic->fw_vnic_id == INVALID_HW_RING_ID)
+		return rc;
+
+	HWRM_PREP(req, VNIC_FREE, -1, resp);
+
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_vnic_id = INVALID_HW_RING_ID;
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 62dc801..887ad2d 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -59,6 +59,7 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index eedd368..0771897 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -90,6 +90,7 @@ struct ctx_hw_stats64 {
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
+#define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3218,6 +3219,87 @@ struct hwrm_vnic_alloc_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_free */
+/*
+ * Description: Free a VNIC resource. Idle any resources associated with the
+ * VNIC as well as the VNIC. Reset and release all resources associated with the
+ * VNIC.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_vnic_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Logical vnic ID */
+	uint32_t vnic_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 19/39] bnxt: add HWRM vnic configure function
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (16 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 18/39] bnxt: add HWRM vnic free function Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 20/39] bnxt: add API to allow configuration of vnic Stephen Hurd
                   ` (22 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

This patch configures the properties and actions of the vnic
allocated by vnic_alloc function from the previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           |  34 ++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +-
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 155 +++++++++++++++++++++++++++++++++
 3 files changed, 191 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 02fb0c4..be4020f 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -510,6 +510,40 @@ int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_cfg_input req = {.req_type = 0 };
+	struct hwrm_vnic_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_CFG, -1, resp);
+
+	/* Only RSS support for now TBD: COS & LB */
+	req.enables =
+	    rte_cpu_to_le_32(HWRM_VNIC_CFG_INPUT_ENABLES_DFLT_RING_GRP |
+			     HWRM_VNIC_CFG_INPUT_ENABLES_RSS_RULE |
+			     HWRM_VNIC_CFG_INPUT_ENABLES_MRU);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+	req.dflt_ring_grp =
+		rte_cpu_to_le_16(bp->grp_info[vnic->start_grp_id].fw_grp_id);
+	req.rss_rule = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+	req.cos_rule = rte_cpu_to_le_16(0xffff);
+	req.lb_rule = rte_cpu_to_le_16(0xffff);
+	req.mru = rte_cpu_to_le_16(bp->eth_dev->data->mtu + ETHER_HDR_LEN +
+				   ETHER_CRC_LEN + VLAN_TAG_SIZE);
+	if (vnic->func_default)
+		req.flags = 1;
+	if (vnic->vlan_strip)
+		req.flags |=
+		    rte_cpu_to_le_32(HWRM_VNIC_CFG_INPUT_FLAGS_VLAN_STRIP_MODE);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 887ad2d..b5cf090 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -59,8 +59,9 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
-int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 0771897..ef0b37a 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -91,6 +91,7 @@ struct ctx_hw_stats64 {
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
+#define HWRM_VNIC_CFG			(UINT32_C(0x42))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3219,6 +3220,160 @@ struct hwrm_vnic_alloc_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_cfg */
+/* Description: Configure the RX VNIC structure. */
+
+/* Input (40 bytes) */
+struct hwrm_vnic_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the VNIC is requested to be the default VNIC
+	 * for the function.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_DEFAULT		UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the VNIC is being configured to strip VLAN in
+	 * the RX path. If set to '0', then VLAN stripping is disabled on this
+	 * VNIC.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_VLAN_STRIP_MODE	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the VNIC is being configured to buffer receive
+	 * packets in the hardware until the host posts new receive buffers. If
+	 * set to '0', then bd_stall is being configured to be disabled on this
+	 * VNIC.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_BD_STALL_MODE		UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the VNIC is being configured to receive both
+	 * RoCE and non-RoCE traffic. If set to '0', then this VNIC is not
+	 * configured to be operating in dual VNIC mode.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_ROCE_DUAL_VNIC_MODE	UINT32_C(0x8)
+	/*
+	 * When this flag is set to '1', the VNIC is requested to be configured
+	 * to receive only RoCE traffic. If this flag is set to '0', then this
+	 * flag shall be ignored by the HWRM. If roce_dual_vnic_mode flag is set
+	 * to '1', then the HWRM client shall not set this flag to '1'.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_ROCE_ONLY_VNIC_MODE	UINT32_C(0x10)
+	uint32_t flags;
+
+	/* This bit must be '1' for the dflt_ring_grp field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_DFLT_RING_GRP	UINT32_C(0x1)
+	/* This bit must be '1' for the rss_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_RSS_RULE		UINT32_C(0x2)
+	/* This bit must be '1' for the cos_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_COS_RULE		UINT32_C(0x4)
+	/* This bit must be '1' for the lb_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_LB_RULE		UINT32_C(0x8)
+	/* This bit must be '1' for the mru field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_MRU			UINT32_C(0x10)
+	uint32_t enables;
+
+	/* Logical vnic ID */
+	uint16_t vnic_id;
+
+	/*
+	 * Default Completion ring for the VNIC. This ring will be chosen if
+	 * packet does not match any RSS rules and if there is no COS rule.
+	 */
+	uint16_t dflt_ring_grp;
+
+	/*
+	 * RSS ID for RSS rule/table structure. 0xFF... (All Fs) if there is no
+	 * RSS rule.
+	 */
+	uint16_t rss_rule;
+
+	/*
+	 * RSS ID for COS rule/table structure. 0xFF... (All Fs) if there is no
+	 * COS rule.
+	 */
+	uint16_t cos_rule;
+
+	/*
+	 * RSS ID for load balancing rule/table structure. 0xFF... (All Fs) if
+	 * there is no LB rule.
+	 */
+	uint16_t lb_rule;
+
+	/*
+	 * The maximum receive unit of the vnic. Each vnic is associated with a
+	 * function. The vnic mru value overwrites the mru setting of the
+	 * associated function. The HWRM shall make sure that vnic mru does not
+	 * exceed the mru of the port the function is associated with.
+	 */
+	uint16_t mru;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_free */
 /*
  * Description: Free a VNIC resource. Idle any resources associated with the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 20/39] bnxt: add API to allow configuration of vnic
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (17 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 19/39] bnxt: add HWRM vnic configure function Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 21/39] bnxt: add HWRM API to configure RSS Stephen Hurd
                   ` (21 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds APIs to allow configuration of a VNIC.
The functions alloc and free the COS and Load Balance context
corresponding to the VNIC in the chip.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           |  38 ++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 153 +++++++++++++++++++++++++++++++++
 3 files changed, 193 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index be4020f..87853f6 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -544,6 +544,44 @@ int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cos_lb_ctx_alloc_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cos_lb_ctx_alloc_output *resp =
+						bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_COS_LB_CTX_ALLOC, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_rss_cos_lb_ctx = rte_le_to_cpu_16(resp->rss_cos_lb_ctx_id);
+
+	return rc;
+}
+
+int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cos_lb_ctx_free_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cos_lb_ctx_free_output *resp =
+						bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_COS_LB_CTX_FREE, -1, resp);
+
+	req.rss_cos_lb_ctx_id = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_rss_cos_lb_ctx = INVALID_HW_RING_ID;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index b5cf090..b7f6b20 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -61,6 +61,8 @@ int bnxt_hwrm_ver_get(struct bnxt *bp);
 
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ef0b37a..6412df2 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -92,6 +92,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
+#define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
+#define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3576,6 +3578,157 @@ struct hwrm_vnic_rss_cfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* Input (16 bytes) */
+struct hwrm_vnic_rss_cos_lb_ctx_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_vnic_rss_cos_lb_ctx_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* rss_cos_lb_ctx_id is 16 b */
+	uint16_t rss_cos_lb_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+	uint8_t unused_4;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_vnic_rss_cos_lb_ctx_free */
+/* Description: This function can be used to free COS/Load Balance context. */
+/* Input (24 bytes) */
+
+struct hwrm_vnic_rss_cos_lb_ctx_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* rss_cos_lb_ctx_id is 16 b */
+	uint16_t rss_cos_lb_ctx_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_rss_cos_lb_ctx_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* Output (32 bytes) */
 struct hwrm_queue_qportcfg_output {
 	/*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 21/39] bnxt: add HWRM API to configure RSS
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (18 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 20/39] bnxt: add API to allow configuration of vnic Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 22/39] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
                   ` (20 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

As mentioned earlier:
A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

The HWRM API defined in this patch will be used to enable RSS
configuration of the VNIC.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           | 24 ++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  2 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h |  1 +
 3 files changed, 27 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 87853f6..c43c2da 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -603,6 +603,30 @@ int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
+			   struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cfg_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_CFG, -1, resp);
+
+	req.hash_type = rte_cpu_to_le_32(vnic->hash_type);
+
+	req.ring_grp_tbl_addr =
+	    rte_cpu_to_le_64(vnic->rss_table_dma_addr);
+	req.hash_key_tbl_addr =
+	    rte_cpu_to_le_64(vnic->rss_hash_key_dma_addr);
+	req.rss_ctx_idx = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index b7f6b20..7c12c6d 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -64,6 +64,8 @@ int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
+			   struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 6412df2..72d4984 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -92,6 +92,7 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
+#define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 22/39] bnxt: add API for L2 Rx mask set/clear functions
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (19 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 21/39] bnxt: add HWRM API to configure RSS Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 23/39] bnxt: add HWRM stats context allocation Stephen Hurd
                   ` (19 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

These HWRM APIs allow setting and clearing of Rx masks in L2 context
per VNIC.

v4:
Address review comments.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           |  45 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 135 +++++++++++++++++++++++++++++++++
 3 files changed, 183 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index c43c2da..7f39db0 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -141,6 +141,51 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_set_rx_mask_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_set_rx_mask_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, CFA_L2_SET_RX_MASK, -1, resp);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+	req.mask = 0;
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
+int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_set_rx_mask_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_set_rx_mask_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t mask = 0;
+
+	HWRM_PREP(req, CFA_L2_SET_RX_MASK, -1, resp);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	/* FIXME add multicast flag, when multicast adding options is supported
+	 * by ethtool.
+	 */
+	if (vnic->flags & BNXT_VNIC_INFO_PROMISC)
+		mask = HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_PROMISCUOUS;
+	if (vnic->flags & BNXT_VNIC_INFO_ALLMULTI)
+		mask = HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_ALL_MCAST;
+	req.mask = rte_cpu_to_le_32(HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_MCAST |
+				    HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_BCAST |
+				    mask);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter)
 {
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 7c12c6d..915cf2a 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -42,6 +42,9 @@
 struct bnxt;
 struct bnxt_filter_info;
 struct bnxt_cp_ring_info;
+int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp,
+				   struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter);
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 72d4984..f8f6a3f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1732,6 +1732,141 @@ struct hwrm_cfa_l2_filter_free_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_cfa_l2_set_rx_mask */
+/* Description: This command will set rx mask of the function. */
+
+/* Input (40 bytes) */
+struct hwrm_cfa_l2_set_rx_mask_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* VNIC ID */
+	uint32_t vnic_id;
+
+	/* Reserved for future use. */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_RESERVED	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the function is requested to accept multi-cast
+	 * packets specified by the multicast addr table.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_MCAST	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the function is requested to accept all multi-
+	 * cast packets.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_ALL_MCAST	UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the function is requested to accept broadcast
+	 * packets.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_BCAST	UINT32_C(0x8)
+	/*
+	 * When this bit is '1', the function is requested to be put in the
+	 * promiscuous mode. The HWRM should accept any function to set up
+	 * promiscuous mode. The HWRM shall follow the semantics below for the
+	 * promiscuous mode support. # When partitioning is not enabled on a
+	 * port (i.e. single PF on the port), then the PF shall be allowed to be
+	 * in the promiscuous mode. When the PF is in the promiscuous mode, then
+	 * it shall receive all host bound traffic on that port. # When
+	 * partitioning is enabled on a port (i.e. multiple PFs per port) and a
+	 * PF on that port is in the promiscuous mode, then the PF receives all
+	 * traffic within that partition as identified by a unique identifier
+	 * for the PF (e.g. S-Tag). If a unique outer VLAN for the PF is
+	 * specified, then the setting of promiscuous mode on that PF shall
+	 * result in the PF receiving all host bound traffic with matching outer
+	 * VLAN. # A VF shall can be set in the promiscuous mode. In the
+	 * promiscuous mode, the VF does not receive any traffic unless a unique
+	 * outer VLAN for the VF is specified. If a unique outer VLAN for the VF
+	 * is specified, then the setting of promiscuous mode on that VF shall
+	 * result in the VF receiving all host bound traffic with the matching
+	 * outer VLAN. # The HWRM shall allow the setting of promiscuous mode on
+	 * a function independently from the promiscuous mode settings on other
+	 * functions.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_PROMISCUOUS	UINT32_C(0x10)
+	/*
+	 * If this flag is set, the corresponding RX filters shall be set up to
+	 * cover multicast/broadcast filters for the outermost Layer 2
+	 * destination MAC address field.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_OUTERMOST	UINT32_C(0x20)
+	uint32_t mask;
+
+	/* This is the address for mcast address tbl. */
+	uint64_t mc_tbl_addr;
+
+	/*
+	 * This value indicates how many entries in mc_tbl are valid. Each entry
+	 * is 6 bytes.
+	 */
+	uint32_t num_mc_entries;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_cfa_l2_set_rx_mask_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_exec_fwd_resp */
 /*
  * Description: This command is used to send an encapsulated request to the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 23/39] bnxt: add HWRM stats context allocation
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (20 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 22/39] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 24/39] bnxt: add HWRM ring alloc/free functions Stephen Hurd
                   ` (18 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API code to allocate a statistics context in the ASIC.
This API will be called by the previously submitted "add statistics
operations patch".

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           | 52 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  3 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 89 ++++++++++++++++++++++++++++++++++
 3 files changed, 144 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 7f39db0..5d0fbf1 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -525,6 +525,31 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 	return rc;
 }
 
+int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
+			     struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	int rc;
+	struct hwrm_stat_ctx_alloc_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_ALLOC, -1, resp);
+
+	req.update_period_ms = rte_cpu_to_le_32(1000);
+
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+	req.stats_dma_addr =
+	    rte_cpu_to_le_64(cpr->hw_stats_map);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	cpr->hw_stats_ctx_id = rte_le_to_cpu_16(resp->stat_ctx_id);
+	bp->grp_info[idx].fw_stats_ctx = cpr->hw_stats_ctx_id;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0, i, j;
@@ -701,6 +726,33 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq;
+		struct bnxt_rx_queue *rxq;
+		struct bnxt_cp_ring_info *cpr;
+		unsigned int idx = i + 1;
+
+		if (i >= bp->rx_cp_nr_rings) {
+			txq = bp->tx_queues[i - bp->rx_cp_nr_rings];
+			cpr = txq->cp_ring;
+		} else {
+			rxq = bp->rx_queues[i];
+			cpr = rxq->cp_ring;
+		}
+
+		rc = bnxt_hwrm_stat_ctx_alloc(bp, cpr, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 915cf2a..f41361e 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -59,6 +59,8 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
+int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
+			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
@@ -70,6 +72,7 @@ int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 			   struct bnxt_vnic_info *vnic);
 
+int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index f8f6a3f..28362c9 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -99,6 +99,7 @@ struct ctx_hw_stats64 {
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_STAT_CTX_ALLOC		(UINT32_C(0xb0))
 #define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
@@ -3183,6 +3184,94 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_alloc */
+/*
+ * Description: This command allocates and does basic preparation for a stat
+ * context.
+ */
+
+/* Input (32 bytes) */
+struct hwrm_stat_ctx_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This is the address for statistic block. */
+	uint64_t stats_dma_addr;
+
+	/*
+	 * The statistic block update period in ms. e.g. 250ms, 500ms, 750ms,
+	 * 1000ms.
+	 */
+	uint32_t update_period_ms;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_stat_ctx_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This is the statistics context ID value. */
+	uint32_t stat_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_clr_stats */
 /* Description: This command clears statistics of a context. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 24/39] bnxt: add HWRM ring alloc/free functions
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (21 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 23/39] bnxt: add HWRM stats context allocation Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 25/39] bnxt: add ring group " Stephen Hurd
                   ` (17 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API calls to allocate and free TX, RX and Completion rings
in the hardware along with the associated structs and definitions.

As mentioned earlier, a completion ring is used by the Ethernet
controller to provide the status of transmitted & received packets,
report errors, status changes to the host software.

v4:
Address review comments.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           | 108 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   7 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 305 +++++++++++++++++++++++++++++++++
 3 files changed, 420 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 5d0fbf1..6152856 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -504,6 +504,114 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_ring_alloc(struct bnxt *bp,
+			 struct bnxt_ring_struct *ring,
+			 uint32_t ring_type, uint32_t map_index,
+			 uint32_t stats_ctx_id)
+{
+	int rc = 0;
+	struct hwrm_ring_alloc_input req = {.req_type = 0 };
+	struct hwrm_ring_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_ALLOC, -1, resp);
+
+	req.enables = rte_cpu_to_le_32(0);
+
+	req.page_tbl_addr = rte_cpu_to_le_64(ring->bd_dma);
+	req.fbo = rte_cpu_to_le_32(0);
+	/* Association of ring index with doorbell index */
+	req.logical_id = rte_cpu_to_le_16(map_index);
+
+	switch (ring_type) {
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_TX:
+		req.queue_id = bp->cos_queue[0].id;
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_RX:
+		req.ring_type = ring_type;
+		req.cmpl_ring_id =
+		    rte_cpu_to_le_16(bp->grp_info[map_index].cp_fw_ring_id);
+		req.length = rte_cpu_to_le_32(ring->ring_size);
+		req.stat_ctx_id = rte_cpu_to_le_16(stats_ctx_id);
+		req.enables = rte_cpu_to_le_32(rte_le_to_cpu_32(req.enables) |
+			HWRM_RING_ALLOC_INPUT_ENABLES_STAT_CTX_ID_VALID);
+		break;
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL:
+		req.ring_type = ring_type;
+		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_POLL;
+		req.length = rte_cpu_to_le_32(ring->ring_size);
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "hwrm alloc invalid ring type %d\n",
+			ring_type);
+		return -1;
+	}
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	if (rc || resp->error_code) {
+		if (rc == 0 && resp->error_code)
+			rc = rte_le_to_cpu_16(resp->error_code);
+		switch (ring_type) {
+		case HWRM_RING_FREE_INPUT_RING_TYPE_CMPL:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc cp failed. rc:%d\n", rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_RX:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc rx failed. rc:%d\n", rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_TX:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc tx failed. rc:%d\n", rc);
+			return rc;
+		default:
+			RTE_LOG(ERR, PMD, "Invalid ring. rc:%d\n", rc);
+			return rc;
+		}
+	}
+
+	ring->fw_ring_id = rte_le_to_cpu_16(resp->ring_id);
+	return rc;
+}
+
+int bnxt_hwrm_ring_free(struct bnxt *bp,
+			struct bnxt_ring_struct *ring, uint32_t ring_type)
+{
+	int rc;
+	struct hwrm_ring_free_input req = {.req_type = 0 };
+	struct hwrm_ring_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_FREE, -1, resp);
+
+	req.ring_type = ring_type;
+	req.ring_id = rte_cpu_to_le_16(ring->fw_ring_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	if (rc || resp->error_code) {
+		if (rc == 0 && resp->error_code)
+			rc = rte_le_to_cpu_16(resp->error_code);
+
+		switch (ring_type) {
+		case HWRM_RING_FREE_INPUT_RING_TYPE_CMPL:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free cp failed. rc:%d\n",
+				rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_RX:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free rx failed. rc:%d\n",
+				rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_TX:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free tx failed. rc:%d\n",
+				rc);
+			return rc;
+		default:
+			RTE_LOG(ERR, PMD, "Invalid ring, rc:%d\n", rc);
+			return rc;
+		}
+	}
+	return 0;
+}
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index f41361e..e4fc243 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -58,6 +58,13 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
+int bnxt_hwrm_ring_alloc(struct bnxt *bp,
+			 struct bnxt_ring_struct *ring,
+			 uint32_t ring_type, uint32_t map_index,
+			 uint32_t stats_ctx_id);
+int bnxt_hwrm_ring_free(struct bnxt *bp,
+			struct bnxt_ring_struct *ring, uint32_t ring_type);
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 28362c9..e6280b6 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -93,6 +93,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
 #define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
+#define HWRM_RING_ALLOC			(UINT32_C(0x50))
+#define HWRM_RING_FREE			(UINT32_C(0x51))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
@@ -3184,6 +3186,309 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_ring_alloc */
+/*
+ * Description: This command allocates and does basic preparation for a ring.
+ */
+
+/* Input (80 bytes) */
+struct hwrm_ring_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This bit must be '1' for the Reserved1 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED1		UINT32_C(0x1)
+	/* This bit must be '1' for the Reserved2 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED2		UINT32_C(0x2)
+	/* This bit must be '1' for the Reserved3 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED3		UINT32_C(0x4)
+	/*
+	 * This bit must be '1' for the stat_ctx_id_valid field to be
+	 * configured.
+	 */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_STAT_CTX_ID_VALID	UINT32_C(0x8)
+	/* This bit must be '1' for the Reserved4 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED4		UINT32_C(0x10)
+	/* This bit must be '1' for the max_bw_valid field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_MAX_BW_VALID	UINT32_C(0x20)
+	uint32_t enables;
+
+	/* Ring Type. */
+		/* Completion Ring (CR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL	(UINT32_C(0x0) << 0)
+		/* TX Ring (TR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_TX	(UINT32_C(0x1) << 0)
+		/* RX Ring (RR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_RX	(UINT32_C(0x2) << 0)
+	uint8_t ring_type;
+
+	uint8_t unused_0;
+	uint16_t unused_1;
+
+	/* This value is a pointer to the page table for the Ring. */
+	uint64_t page_tbl_addr;
+
+	/* First Byte Offset of the first entry in the first page. */
+	uint32_t fbo;
+
+	/*
+	 * Actual page size in 2^page_size. The supported range is increments in
+	 * powers of 2 from 16 bytes to 1GB. - 4 = 16 B Page size is 16 B. - 12
+	 * = 4 KB Page size is 4 KB. - 13 = 8 KB Page size is 8 KB. - 16 = 64 KB
+	 * Page size is 64 KB. - 22 = 2 MB Page size is 2 MB. - 23 = 4 MB Page
+	 * size is 4 MB. - 31 = 1 GB Page size is 1 GB.
+	 */
+	uint8_t page_size;
+
+	/*
+	 * This value indicates the depth of page table. For this version of the
+	 * specification, value other than 0 or 1 shall be considered as an
+	 * invalid value. When the page_tbl_depth = 0, then it is treated as a
+	 * special case with the following. 1. FBO and page size fields are not
+	 * valid. 2. page_tbl_addr is the physical address of the first element
+	 * of the ring.
+	 */
+	uint8_t page_tbl_depth;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * Number of 16B units in the ring. Minimum size for a ring is 16 16B
+	 * entries.
+	 */
+	uint32_t length;
+
+	/*
+	 * Logical ring number for the ring to be allocated. This value
+	 * determines the position in the doorbell area where the update to the
+	 * ring will be made. For completion rings, this value is also the MSI-X
+	 * vector number for the function the completion ring is associated
+	 * with.
+	 */
+	uint16_t logical_id;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This value
+	 * indicates what completion ring the TX ring is associated with.
+	 */
+	uint16_t cmpl_ring_id;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This value
+	 * indicates what CoS queue the TX ring is associated with.
+	 */
+	uint16_t queue_id;
+
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved1;
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint16_t reserved2;
+
+	uint8_t unused_6;
+	uint8_t unused_7;
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved3;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This input
+	 * indicates what statistics context this ring should be associated
+	 * with.
+	 */
+	uint32_t stat_ctx_id;
+
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved4;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. Maximum BW
+	 * allocated to this TX ring in Mbps. The HWRM will translate this value
+	 * into byte counter and time interval used for this ring inside the
+	 * device.
+	 */
+	uint32_t max_bw;
+
+	/*
+	 * This field is used only when ring_type is a Completion ring. This
+	 * value indicates what interrupt mode should be used on this completion
+	 * ring. Note: In the legacy interrupt mode, no more than 16 completion
+	 * rings are allowed.
+	 */
+		/* Legacy INTA */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_LEGACY	(UINT32_C(0x0) << 0)
+		/* Reserved */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_RSVD	(UINT32_C(0x1) << 0)
+		/* MSI-X */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_MSIX	(UINT32_C(0x2) << 0)
+		/* No Interrupt - Polled mode */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_POLL	(UINT32_C(0x3) << 0)
+	uint8_t int_mode;
+
+	uint8_t unused_8[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_ring_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* Physical number of ring allocated. */
+	uint16_t ring_id;
+
+	/* Logical number of ring allocated. */
+	uint16_t logical_ring_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ring_free */
+/*
+ * Description: This command is used to free a ring and associated resources.
+ */
+/* Input (24 bytes) */
+
+struct hwrm_ring_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Ring Type. */
+		/* Completion Ring (CR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_CMPL	(UINT32_C(0x0) << 0)
+		/* TX Ring (TR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_TX	(UINT32_C(0x1) << 0)
+		/* RX Ring (RR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_RX	(UINT32_C(0x2) << 0)
+	uint8_t ring_type;
+
+	uint8_t unused_0;
+
+	/* Physical number of ring allocated. */
+	uint16_t ring_id;
+
+	uint32_t unused_1;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_alloc */
 /*
  * Description: This command allocates and does basic preparation for a stat
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 25/39] bnxt: add ring group alloc/free functions
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (22 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 24/39] bnxt: add HWRM ring alloc/free functions Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 26/39] bnxt: add HWRM stat context free function Stephen Hurd
                   ` (16 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API for ring group alloc/free functions, associated structs and
definitions.
This API allocates and does basic preparation for a ring group in ASIC.
A ring group is identified by an index. It consists of Rx ring id,
completion ring id and a statistics context.

v4:
Address issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           |  84 +++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   4 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 185 +++++++++++++++++++++++++++++++++
 3 files changed, 273 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 6152856..4c4f707 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -612,6 +612,47 @@ int bnxt_hwrm_ring_free(struct bnxt *bp,
 	return 0;
 }
 
+int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx)
+{
+	int rc = 0;
+	struct hwrm_ring_grp_alloc_input req = {.req_type = 0 };
+	struct hwrm_ring_grp_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_GRP_ALLOC, -1, resp);
+
+	req.cr = rte_cpu_to_le_16(bp->grp_info[idx].cp_fw_ring_id);
+	req.rr = rte_cpu_to_le_16(bp->grp_info[idx].rx_fw_ring_id);
+	req.ar = rte_cpu_to_le_16(bp->grp_info[idx].ag_fw_ring_id);
+	req.sc = rte_cpu_to_le_16(bp->grp_info[idx].fw_stats_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->grp_info[idx].fw_grp_id =
+	    rte_le_to_cpu_16(resp->ring_group_id);
+
+	return rc;
+}
+
+int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx)
+{
+	int rc;
+	struct hwrm_ring_grp_free_input req = {.req_type = 0 };
+	struct hwrm_ring_grp_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_GRP_FREE, -1, resp);
+
+	req.ring_group_id = rte_cpu_to_le_16(bp->grp_info[idx].fw_grp_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->grp_info[idx].fw_grp_id = INVALID_HW_RING_ID;
+	return rc;
+}
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 {
 	int rc = 0;
@@ -861,6 +902,49 @@ int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
+{
+	uint16_t i;
+	uint32_t rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (bp->grp_info[idx].fw_grp_id == INVALID_HW_RING_ID) {
+			RTE_LOG(ERR, PMD,
+				"Attempt to free invalid ring group %d\n",
+				idx);
+			continue;
+		}
+
+		rc = bnxt_hwrm_ring_grp_free(bp, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
+int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp)
+{
+	uint16_t i;
+	uint32_t rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (bp->grp_info[idx].cp_fw_ring_id == INVALID_HW_RING_ID ||
+		    bp->grp_info[idx].rx_fw_ring_id == INVALID_HW_RING_ID)
+			continue;
+
+		rc = bnxt_hwrm_ring_grp_alloc(bp, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index e4fc243..fb088cf 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -64,6 +64,8 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 			 uint32_t stats_ctx_id);
 int bnxt_hwrm_ring_free(struct bnxt *bp,
 			struct bnxt_ring_struct *ring, uint32_t ring_type);
+int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx);
+int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
@@ -81,6 +83,8 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
+int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index e6280b6..4e2eb9f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -95,6 +95,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
 #define HWRM_RING_ALLOC			(UINT32_C(0x50))
 #define HWRM_RING_FREE			(UINT32_C(0x51))
+#define HWRM_RING_GRP_ALLOC		(UINT32_C(0x60))
+#define HWRM_RING_GRP_FREE		(UINT32_C(0x61))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
@@ -3489,6 +3491,189 @@ struct hwrm_ring_free_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_ring_grp_alloc */
+/*
+ * Description: This API allocates and does basic preparation for a ring group.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ring_grp_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This value identifies the CR associated with the ring group. */
+	uint16_t cr;
+
+	/* This value identifies the main RR associated with the ring group. */
+	uint16_t rr;
+
+	/*
+	 * This value identifies the aggregation RR associated with the ring
+	 * group. If this value is 0xFF... (All Fs), then no Aggregation ring
+	 * will be set.
+	 */
+	uint16_t ar;
+
+	/*
+	 * This value identifies the statistics context associated with the ring
+	 * group.
+	 */
+	uint16_t sc;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_grp_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This is the ring group ID value. Use this value to program the
+	 * default ring group for the VNIC or as table entries in an RSS/COS
+	 * context.
+	 */
+	uint32_t ring_group_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ring_grp_free */
+/*
+ * Description: This API frees a ring group and associated resources. # If a
+ * ring in the ring group is reset or free, then the associated rings in the
+ * ring group shall also be reset/free using hwrm_ring_free. # A function driver
+ * shall always use hwrm_ring_grp_free after freeing all rings in a group. # As
+ * a part of executing this command, the HWRM shall reset all associated ring
+ * group resources.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ring_grp_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This is the ring group ID value. */
+	uint32_t ring_group_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_grp_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_alloc */
 /*
  * Description: This command allocates and does basic preparation for a stat
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 26/39] bnxt: add HWRM stat context free function
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (23 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 25/39] bnxt: add ring group " Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 27/39] bnxt: Add HWRM API to set and clear filters Stephen Hurd
                   ` (15 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add function and associated structures and definitions to free
statistics context from the ASIC.

v4:
Address review comments and fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           | 44 ++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  3 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 81 ++++++++++++++++++++++++++++++++++
 3 files changed, 128 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 4c4f707..d3e77d5 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -699,6 +699,28 @@ int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 	return rc;
 }
 
+int bnxt_hwrm_stat_ctx_free(struct bnxt *bp,
+			    struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	int rc;
+	struct hwrm_stat_ctx_free_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_FREE, -1, resp);
+
+	req.stat_ctx_id = rte_cpu_to_le_16(cpr->hw_stats_ctx_id);
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	cpr->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
+	bp->grp_info[idx].fw_stats_ctx = cpr->hw_stats_ctx_id;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0, i, j;
@@ -875,6 +897,28 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	int rc;
+	unsigned int i;
+	struct bnxt_cp_ring_info *cpr;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (i >= bp->rx_cp_nr_rings)
+			cpr = bp->tx_queues[i - bp->rx_cp_nr_rings]->cp_ring;
+		else
+			cpr = bp->rx_queues[i]->cp_ring;
+		if (cpr->hw_stats_ctx_id != HWRM_NA_SIGNATURE) {
+			rc = bnxt_hwrm_stat_ctx_free(bp, cpr, idx);
+			if (rc)
+				return rc;
+		}
+	}
+	return 0;
+}
+
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
 {
 	unsigned int i;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index fb088cf..5665762 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -70,6 +70,8 @@ int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
+int bnxt_hwrm_stat_ctx_free(struct bnxt *bp,
+			    struct bnxt_cp_ring_info *cpr, unsigned int idx);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
@@ -83,6 +85,7 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 4e2eb9f..d58295a 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -104,6 +104,7 @@ struct ctx_hw_stats64 {
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
 #define HWRM_STAT_CTX_ALLOC		(UINT32_C(0xb0))
+#define HWRM_STAT_CTX_FREE		(UINT32_C(0xb1))
 #define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
@@ -3839,6 +3840,86 @@ struct hwrm_stat_ctx_clr_stats_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_free */
+/* Description: This command is used to free a stat context. */
+/* Input (24 bytes) */
+
+struct hwrm_stat_ctx_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* ID of the statistics context that is being queried. */
+	uint32_t stat_ctx_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_stat_ctx_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This is the statistics context ID value. */
+	uint32_t stat_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_alloc */
 /*
  * Description: This VNIC is a resource in the RX side of the chip that is used
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 27/39] bnxt: Add HWRM API to set and clear filters
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (24 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 26/39] bnxt: add HWRM stat context free function Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 28/39] bnxt: add ring alloc, free and group init Stephen Hurd
                   ` (14 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to set and clear L2 filters from the corresponding
VNIC. These filters will determine the characteristics of Rx traffic.

v4:
Separated this code from the previous patch as it had nothing to
do with freeing of statistics context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c | 69 ++++++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h |  6 ++++
 2 files changed, 75 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index d3e77d5..f8e9d20 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -206,6 +206,49 @@ int bnxt_hwrm_clear_filter(struct bnxt *bp,
 	return 0;
 }
 
+int bnxt_hwrm_set_filter(struct bnxt *bp,
+			 struct bnxt_vnic_info *vnic,
+			 struct bnxt_filter_info *filter)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_filter_alloc_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_filter_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t enables = 0;
+
+	HWRM_PREP(req, CFA_L2_FILTER_ALLOC, -1, resp);
+
+	req.flags = rte_cpu_to_le_32(filter->flags);
+
+	enables = filter->enables |
+	      HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_DST_ID;
+	req.dst_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR)
+		memcpy(req.l2_addr, filter->l2_addr,
+		       ETHER_ADDR_LEN);
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK)
+		memcpy(req.l2_addr_mask, filter->l2_addr_mask,
+		       ETHER_ADDR_LEN);
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN)
+		req.l2_ovlan = filter->l2_ovlan;
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN_MASK)
+		req.l2_ovlan_mask = filter->l2_ovlan_mask;
+
+	req.enables = rte_cpu_to_le_32(enables);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	filter->fw_l2_filter_id = rte_le_to_cpu_64(resp->l2_filter_id);
+
+	return rc;
+}
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
 {
 	int rc;
@@ -1016,6 +1059,32 @@ int bnxt_alloc_hwrm_resources(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	struct bnxt_filter_info *filter;
+	int rc = 0;
+
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		rc = bnxt_hwrm_clear_filter(bp, filter);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
+int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	struct bnxt_filter_info *filter;
+	int rc = 0;
+
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		rc = bnxt_hwrm_set_filter(bp, vnic, filter);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
 static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
 {
 	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 5665762..55728df 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -47,6 +47,9 @@ int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp,
 int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter);
+int bnxt_hwrm_set_filter(struct bnxt *bp,
+			 struct bnxt_vnic_info *vnic,
+			 struct bnxt_filter_info *filter);
 
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 
@@ -88,6 +91,9 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
+int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+void bnxt_free_all_hwrm_resources(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 28/39] bnxt: add ring alloc, free and group init
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (25 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 27/39] bnxt: Add HWRM API to set and clear filters Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 29/39] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
                   ` (13 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add a function to initialize ring groups, and a function to
allocate and free the rings via HWRM.

This should be the last functionality needed to add start/stop
device operations.

v4:
Address review comment to merge another patch into this to avoid
a compilation issue. Fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c | 112 +++++++++++++++++++++++++++++++++++++++-
 drivers/net/bnxt/bnxt_hwrm.h |   1 +
 drivers/net/bnxt/bnxt_ring.c | 119 +++++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ring.h |   2 +
 4 files changed, 233 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index f8e9d20..dc1bce7 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -43,8 +43,10 @@
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -579,7 +581,11 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 		break;
 	case HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL:
 		req.ring_type = ring_type;
-		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_POLL;
+		/*
+		 * TODO: Some HWRM versions crash with
+		 * HWRM_RING_ALLOC_INPUT_INT_MODE_POLL
+		 */
+		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_MSIX;
 		req.length = rte_cpu_to_le_32(ring->ring_size);
 		break;
 	default:
@@ -1012,6 +1018,84 @@ int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
 	return rc;
 }
 
+static void bnxt_free_cp_ring(struct bnxt *bp,
+			      struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+
+	bnxt_hwrm_ring_free(bp, cp_ring,
+			HWRM_RING_FREE_INPUT_RING_TYPE_CMPL);
+	cp_ring->fw_ring_id = INVALID_HW_RING_ID;
+	bp->grp_info[idx].cp_fw_ring_id = INVALID_HW_RING_ID;
+	memset(cpr->cp_desc_ring, 0, cpr->cp_ring_struct->ring_size *
+			sizeof(*cpr->cp_desc_ring));
+	cpr->cp_raw_cons = 0;
+}
+
+int bnxt_free_all_hwrm_rings(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_tx_ring_info *txr = txq->tx_ring;
+		struct bnxt_ring_struct *ring = txr->tx_ring_struct;
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		unsigned int idx = bp->rx_cp_nr_rings + i + 1;
+
+		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
+			bnxt_hwrm_ring_free(bp, ring,
+					HWRM_RING_FREE_INPUT_RING_TYPE_TX);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
+			memset(txr->tx_desc_ring, 0,
+					txr->tx_ring_struct->ring_size *
+					sizeof(*txr->tx_desc_ring));
+			memset(txr->tx_buf_ring, 0,
+					txr->tx_ring_struct->ring_size *
+					sizeof(*txr->tx_buf_ring));
+			txr->tx_prod = 0;
+			txr->tx_cons = 0;
+		}
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, idx);
+	}
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+		struct bnxt_ring_struct *ring = rxr->rx_ring_struct;
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		unsigned int idx = i + 1;
+
+		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
+			bnxt_hwrm_ring_free(bp, ring,
+					HWRM_RING_FREE_INPUT_RING_TYPE_RX);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
+			bp->grp_info[idx].rx_fw_ring_id = INVALID_HW_RING_ID;
+			memset(rxr->rx_desc_ring, 0,
+					rxr->rx_ring_struct->ring_size *
+					sizeof(*rxr->rx_desc_ring));
+			memset(rxr->rx_buf_ring, 0,
+					rxr->rx_ring_struct->ring_size *
+					sizeof(*rxr->rx_buf_ring));
+			rxr->rx_prod = 0;
+		}
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, idx);
+	}
+
+	/* Default completion ring */
+	{
+		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, 0);
+	}
+
+	return rc;
+}
+
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp)
 {
 	uint16_t i;
@@ -1085,6 +1169,32 @@ int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+void bnxt_free_all_hwrm_resources(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	unsigned int i;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+	bnxt_hwrm_cfa_l2_clear_rx_mask(bp, vnic);
+
+	/* VNIC resources */
+	for (i = 0; i < bp->nr_vnics; i++) {
+		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+		bnxt_clear_hwrm_vnic_filters(bp, vnic);
+
+		bnxt_hwrm_vnic_ctx_free(bp, vnic);
+		bnxt_hwrm_vnic_free(bp, vnic);
+	}
+	/* Ring resources */
+	bnxt_free_all_hwrm_rings(bp);
+	bnxt_free_all_hwrm_ring_grps(bp);
+	bnxt_free_all_hwrm_stat_ctxs(bp);
+}
+
 static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
 {
 	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 55728df..abf81fb 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -89,6 +89,7 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_rings(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index 675c5b8..49cfab9 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -35,8 +35,11 @@
 
 #include "bnxt.h"
 #include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
+#include "bnxt_rxq.h"
 #include "bnxt_rxr.h"
+#include "bnxt_txq.h"
 #include "bnxt_txr.h"
 
 #include "hsi_struct_def_dpdk.h"
@@ -55,6 +58,19 @@ void bnxt_free_ring(struct bnxt_ring_struct *ring)
 }
 
 /*
+ * Ring groups
+ */
+
+void bnxt_init_ring_grps(struct bnxt *bp)
+{
+	unsigned int i;
+
+	for (i = 0; i < bp->max_ring_grps; i++)
+		memset(&bp->grp_info[i], (uint8_t)HWRM_NA_SIGNATURE,
+		       sizeof(struct bnxt_ring_grp_info));
+}
+
+/*
  * Allocates a completion ring with vmem and stats optionally also allocating
  * a TX and/or RX ring.  Passing NULL as tx_ring_info and/or rx_ring_info
  * to not allocate them.
@@ -184,3 +200,106 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 	cp_ring_info->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
 	return 0;
 }
+
+/* ring_grp usage:
+ * [0] = default completion ring
+ * [1 -> +rx_cp_nr_rings] = rx_cp, rx rings
+ * [1+rx_cp_nr_rings + 1 -> +tx_cp_nr_rings] = tx_cp, tx rings
+ */
+int bnxt_alloc_hwrm_rings(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	/* Default completion ring */
+	{
+		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+		struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					  HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					  0, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+		bp->grp_info[0].cp_fw_ring_id = cp_ring->fw_ring_id;
+	}
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+		struct bnxt_ring_struct *ring = rxr->rx_ring_struct;
+		unsigned int idx = i + 1;
+
+		/* Rx cmpl */
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					idx, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].cp_fw_ring_id = cp_ring->fw_ring_id;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+
+		/* Rx ring */
+		rc = bnxt_hwrm_ring_alloc(bp, ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_RX,
+					idx, cpr->hw_stats_ctx_id);
+		if (rc)
+			goto err_out;
+		rxr->rx_prod = 0;
+		rxr->rx_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].rx_fw_ring_id = ring->fw_ring_id;
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+		if (bnxt_init_one_rx_ring(rxq)) {
+			RTE_LOG(ERR, PMD, "bnxt_init_one_rx_ring failed!");
+			bnxt_rx_queue_release_op(rxq);
+			return -ENOMEM;
+		}
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+	}
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_tx_ring_info *txr = txq->tx_ring;
+		struct bnxt_ring_struct *ring = txr->tx_ring_struct;
+		unsigned int idx = 1 + bp->rx_cp_nr_rings + i;
+
+		/* Tx cmpl */
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					idx, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].cp_fw_ring_id = cp_ring->fw_ring_id;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+
+		/* Tx ring */
+		rc = bnxt_hwrm_ring_alloc(bp, ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_TX,
+					idx, cpr->hw_stats_ctx_id);
+		if (rc)
+			goto err_out;
+
+		txr->tx_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+	}
+
+err_out:
+	return rc;
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index c7c3b33..e548586 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -92,10 +92,12 @@ struct bnxt_tx_ring_info;
 struct bnxt_rx_ring_info;
 struct bnxt_cp_ring_info;
 void bnxt_free_ring(struct bnxt_ring_struct *ring);
+void bnxt_init_ring_grps(struct bnxt *bp);
 int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 			    struct bnxt_tx_ring_info *tx_ring_info,
 			    struct bnxt_rx_ring_info *rx_ring_info,
 			    struct bnxt_cp_ring_info *cp_ring_info,
 			    const char *suffix);
+int bnxt_alloc_hwrm_rings(struct bnxt *bp);
 
 #endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 29/39] bnxt: add HWRM port PHY config call and helpers
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (26 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 28/39] bnxt: add ring alloc, free and group init Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations Stephen Hurd
                   ` (12 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM calls to query the port's PHY and link configuration.
This HWRM command and helper function like bnxt_get_hwrm_link_config()
and bnxt_parse_hw_link_speed() parse the link state.

v4:
Fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_hwrm.c           | 120 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 790 +++++++++++++++++++++++++++++++++
 3 files changed, 911 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index dc1bce7..978e379 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -521,6 +521,43 @@ static int bnxt_hwrm_port_phy_cfg(struct bnxt *bp, struct bnxt_link_info *conf)
 	return rc;
 }
 
+static int bnxt_hwrm_port_phy_qcfg(struct bnxt *bp,
+				   struct bnxt_link_info *link_info)
+{
+	int rc = 0;
+	struct hwrm_port_phy_qcfg_input req = {.req_type = 0};
+	struct hwrm_port_phy_qcfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, PORT_PHY_QCFG, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	link_info->phy_link_status = resp->link;
+	if (link_info->phy_link_status == HWRM_PORT_PHY_QCFG_OUTPUT_LINK_LINK) {
+		link_info->link_up = 1;
+		link_info->link_speed = rte_le_to_cpu_16(resp->link_speed);
+	} else {
+		link_info->link_up = 0;
+		link_info->link_speed = 0;
+	}
+	link_info->duplex = resp->duplex;
+	link_info->pause = resp->pause;
+	link_info->auto_pause = resp->auto_pause;
+	link_info->force_pause = resp->force_pause;
+	link_info->auto_mode = resp->auto_mode;
+
+	link_info->support_speeds = rte_le_to_cpu_16(resp->support_speeds);
+	link_info->auto_link_speed = rte_le_to_cpu_16(resp->auto_link_speed);
+	link_info->preemphasis = rte_le_to_cpu_32(resp->preemphasis);
+	link_info->phy_ver[0] = resp->phy_maj;
+	link_info->phy_ver[1] = resp->phy_min;
+	link_info->phy_ver[2] = resp->phy_bld;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
@@ -1326,6 +1363,89 @@ static uint16_t bnxt_parse_eth_link_speed_mask(uint32_t link_speed)
 	return ret;
 }
 
+static uint32_t bnxt_parse_hw_link_speed(uint16_t hw_link_speed)
+{
+	uint32_t eth_link_speed = ETH_SPEED_NUM_NONE;
+
+	switch (hw_link_speed) {
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB:
+		eth_link_speed = ETH_SPEED_NUM_100M;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_1GB:
+		eth_link_speed = ETH_SPEED_NUM_1G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2_5GB:
+		eth_link_speed = ETH_SPEED_NUM_2_5G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10GB:
+		eth_link_speed = ETH_SPEED_NUM_10G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_20GB:
+		eth_link_speed = ETH_SPEED_NUM_20G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_25GB:
+		eth_link_speed = ETH_SPEED_NUM_25G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_40GB:
+		eth_link_speed = ETH_SPEED_NUM_40G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_50GB:
+		eth_link_speed = ETH_SPEED_NUM_50G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2GB:
+	default:
+		RTE_LOG(ERR, PMD, "HWRM link speed %d not defined\n",
+			hw_link_speed);
+		break;
+	}
+	return eth_link_speed;
+}
+
+static uint16_t bnxt_parse_hw_link_duplex(uint16_t hw_link_duplex)
+{
+	uint16_t eth_link_duplex = ETH_LINK_FULL_DUPLEX;
+
+	switch (hw_link_duplex) {
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH:
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_FULL:
+		eth_link_duplex = ETH_LINK_FULL_DUPLEX;
+		break;
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF:
+		eth_link_duplex = ETH_LINK_HALF_DUPLEX;
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "HWRM link duplex %d not defined\n",
+			hw_link_duplex);
+		break;
+	}
+	return eth_link_duplex;
+}
+
+int bnxt_get_hwrm_link_config(struct bnxt *bp, struct rte_eth_link *link)
+{
+	int rc = 0;
+	struct bnxt_link_info *link_info = &bp->link_info;
+
+	rc = bnxt_hwrm_port_phy_qcfg(bp, link_info);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Get link config failed with rc %d\n", rc);
+		goto exit;
+	}
+	if (link_info->link_up)
+		link->link_speed =
+			bnxt_parse_hw_link_speed(link_info->link_speed);
+	else
+		link->link_speed = ETH_LINK_SPEED_10M;
+	link->link_duplex = bnxt_parse_hw_link_duplex(link_info->duplex);
+	link->link_status = link_info->link_up;
+	link->link_autoneg = link_info->auto_mode ==
+		HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE ?
+		ETH_LINK_SPEED_FIXED : ETH_LINK_SPEED_AUTONEG;
+exit:
+	return rc;
+}
+
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index abf81fb..d1aee1c 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -97,6 +97,7 @@ int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 void bnxt_free_all_hwrm_resources(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+int bnxt_get_hwrm_link_config(struct bnxt *bp, struct rte_eth_link *link);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
 
 #endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index d58295a..f2db3ea 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -88,6 +88,7 @@ struct ctx_hw_stats64 {
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
+#define HWRM_PORT_PHY_QCFG		(UINT32_C(0x27))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
@@ -2793,6 +2794,795 @@ struct hwrm_port_phy_cfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_port_phy_qcfg */
+/* Description: This command queries the PHY configuration for the port. */
+/* Input (24 bytes) */
+
+struct hwrm_port_phy_qcfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Port ID of port that is to be queried. */
+	uint16_t port_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (96 bytes) */
+struct hwrm_port_phy_qcfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This value indicates the current link status. */
+		/* There is no link or cable detected. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_NO_LINK	(UINT32_C(0x0) << 0)
+		/* There is no link, but a cable has been detected. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SIGNAL	(UINT32_C(0x1) << 0)
+		/* There is a link. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_LINK	(UINT32_C(0x2) << 0)
+	uint8_t link;
+
+	uint8_t unused_0;
+
+	/* This value indicates the current link speed of the connection. */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t link_speed;
+
+	/* This value is indicates the duplex of the current connection. */
+		/* Half Duplex connection. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_DUPLEX_HALF	(UINT32_C(0x0) << 0)
+		/* Full duplex connection. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_DUPLEX_FULL	(UINT32_C(0x1) << 0)
+	uint8_t duplex;
+
+	/*
+	 * This value is used to indicate the current pause configuration. When
+	 * autoneg is enabled, this value represents the autoneg results of
+	 * pause configuration.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX	UINT32_C(0x2)
+	uint8_t pause;
+
+	/*
+	 * The supported speeds for the port. This is a bit mask. For each speed
+	 * that is supported, the corrresponding bit will be set to '1'.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10MB \
+							UINT32_C(0x2000)
+	uint16_t support_speeds;
+
+	/*
+	 * Current setting of forced link speed. When the link speed is not
+	 * being forced, this value shall be set to 0.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t force_link_speed;
+
+	/* Current setting of auto negotiation mode. */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ALL_SPEEDS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_SPEED_MASK \
+							(UINT32_C(0x4) << 0)
+	uint8_t auto_mode;
+
+	/*
+	 * Current setting of pause autonegotiation. Move autoneg_pause flag
+	 * here.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_TX	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_RX	UINT32_C(0x2)
+	/*
+	 * When set to 1, the advertisement of pause is enabled. # When the
+	 * auto_mode is not set to none and this flag is set to 1, then the
+	 * auto_pause bits on this port are being advertised and autoneg pause
+	 * results are being interpreted. # When the auto_mode is not set to
+	 * none and this flag is set to 0, the pause is forced as indicated in
+	 * force_pause, and also advertised as auto_pause bits, but the autoneg
+	 * results are not interpreted since the pause configuration is being
+	 * forced. # When the auto_mode is set to none and this flag is set to
+	 * 1, auto_pause bits should be ignored and should be set to 0.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_AUTONEG_PAUSE \
+							UINT32_C(0x4)
+	uint8_t auto_pause;
+
+	/*
+	 * Current setting for auto_link_speed. This field is only valid when
+	 * auto_mode is set to "one_speed" or "one_or_below".
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t auto_link_speed;
+
+	/*
+	 * Current setting for auto_link_speed_mask that is used to advertise
+	 * speeds during autonegotiation. This field is only valid when
+	 * auto_mode is set to "mask". The speeds specified in this field shall
+	 * be a subset of supported speeds on this port.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10MB \
+							UINT32_C(0x2000)
+	uint16_t auto_link_speed_mask;
+
+	/* Current setting for wirespeed. */
+		/* Wirespeed feature is disabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_WIRESPEED_OFF	(UINT32_C(0x0) << 0)
+		/* Wirespeed feature is enabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_WIRESPEED_ON	(UINT32_C(0x1) << 0)
+	uint8_t wirespeed;
+
+	/* Current setting for loopback. */
+		/* No loopback is selected. Normal operation. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_NONE	(UINT32_C(0x0) << 0)
+		/*
+		 * The HW will be configured with local loopback such that host
+		 * data is sent back to the host without modification.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_LOCAL	(UINT32_C(0x1) << 0)
+		/*
+		 * The HW will be configured with remote loopback such that port
+		 * logic will send packets back out the transmitter that are
+		 * received.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_REMOTE	(UINT32_C(0x2) << 0)
+	uint8_t lpbk;
+
+	/*
+	 * Current setting of forced pause. When the pause configuration is not
+	 * being forced, then this value shall be set to 0.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_PAUSE_TX \
+							UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_PAUSE_RX \
+							UINT32_C(0x2)
+	uint8_t force_pause;
+
+	/*
+	 * This value indicates the current status of the optics module on this
+	 * port.
+	 */
+		/* Module is inserted and accepted */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Module is rejected and transmit side Laser is disabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_DISABLETX \
+							(UINT32_C(0x1) << 0)
+		/* Module mismatch warning. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG \
+							(UINT32_C(0x2) << 0)
+		/* Module is rejected and powered down. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_PWRDOWN \
+							(UINT32_C(0x3) << 0)
+		/* Module is not inserted. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NOTINSERTED \
+							(UINT32_C(0x4) << 0)
+		/* Module status is not applicable. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NOTAPPLICABLE \
+							(UINT32_C(0xff) << 0)
+	uint8_t module_status;
+
+	/* Current setting for preemphasis. */
+	uint32_t preemphasis;
+
+	/* This field represents the major version of the PHY. */
+	uint8_t phy_maj;
+
+	/* This field represents the minor version of the PHY. */
+	uint8_t phy_min;
+
+	/* This field represents the build version of the PHY. */
+	uint8_t phy_bld;
+
+	/* This value represents a PHY type. */
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* BASE-CR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASECR \
+							(UINT32_C(0x1) << 0)
+		/* BASE-KR4 (Deprecated) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR4 \
+							(UINT32_C(0x2) << 0)
+		/* BASE-LR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASELR \
+							(UINT32_C(0x3) << 0)
+		/* BASE-SR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASESR \
+							(UINT32_C(0x4) << 0)
+		/* BASE-KR2 (Deprecated) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR2 \
+							(UINT32_C(0x5) << 0)
+		/* BASE-KX */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKX \
+							(UINT32_C(0x6) << 0)
+		/* BASE-KR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR \
+							(UINT32_C(0x7) << 0)
+		/* BASE-T */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASET \
+							(UINT32_C(0x8) << 0)
+		/* EEE capable BASE-T */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASETE \
+							(UINT32_C(0x9) << 0)
+		/* SGMII connected external PHY */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_SGMIIEXTPHY \
+							(UINT32_C(0xa) << 0)
+	uint8_t phy_type;
+
+	/* This value represents a media type. */
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* Twisted Pair */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_TP	(UINT32_C(0x1) << 0)
+		/* Direct Attached Copper */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_DAC \
+							(UINT32_C(0x2) << 0)
+		/* Fiber */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_FIBRE \
+							(UINT32_C(0x3) << 0)
+	uint8_t media_type;
+
+	/* This value represents a transceiver type. */
+		/* PHY and MAC are in the same package */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_PKG_TYPE_XCVR_INTERNAL \
+							(UINT32_C(0x1) << 0)
+		/* PHY and MAC are in different packages */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_PKG_TYPE_XCVR_EXTERNAL \
+							(UINT32_C(0x2) << 0)
+	uint8_t xcvr_pkg_type;
+
+	/*
+	 * This field represents flags related to EEE configuration. These EEE
+	 * configuration flags are valid only when the auto_mode is not set to
+	 * none (in other words autonegotiation is enabled).
+	 */
+	/* This field represents PHY address. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_ADDR_MASK	UINT32_C(0x1f)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_ADDR_SFT	0
+	/*
+	 * When set to 1, Energy Efficient Ethernet (EEE) mode is enabled.
+	 * Speeds for autoneg with EEE mode enabled are based on
+	 * eee_link_speed_mask.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_ENABLED \
+							UINT32_C(0x20)
+	/*
+	 * This flag is valid only when eee_enabled is set to 1. # If
+	 * eee_enabled is set to 0, then EEE mode is disabled and this flag
+	 * shall be ignored. # If eee_enabled is set to 1 and this flag is set
+	 * to 1, then Energy Efficient Ethernet (EEE) mode is enabled and in
+	 * use. # If eee_enabled is set to 1 and this flag is set to 0, then
+	 * Energy Efficient Ethernet (EEE) mode is enabled but is currently not
+	 * in use.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_ACTIVE \
+							UINT32_C(0x40)
+	/*
+	 * This flag is valid only when eee_enabled is set to 1. # If
+	 * eee_enabled is set to 0, then EEE mode is disabled and this flag
+	 * shall be ignored. # If eee_enabled is set to 1 and this flag is set
+	 * to 1, then Energy Efficient Ethernet (EEE) mode is enabled and TX LPI
+	 * is enabled. # If eee_enabled is set to 1 and this flag is set to 0,
+	 * then Energy Efficient Ethernet (EEE) mode is enabled but TX LPI is
+	 * disabled.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_TX_LPI \
+							UINT32_C(0x80)
+	/*
+	 * This field represents flags related to EEE configuration. These EEE
+	 * configuration flags are valid only when the auto_mode is not set to
+	 * none (in other words autonegotiation is enabled).
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_MASK \
+							UINT32_C(0xe0)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_SFT	5
+	uint8_t eee_config_phy_addr;
+
+	/* Reserved field, set to 0 */
+	/*
+	 * When set to 1, the parallel detection is used to determine the speed
+	 * of the link partner. Parallel detection is used when a
+	 * autonegotiation capable device is connected to a link parter that is
+	 * not capable of autonegotiation.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PARALLEL_DETECT \
+							UINT32_C(0x1)
+	/* Reserved field, set to 0 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_RESERVED_MASK	UINT32_C(0xfe)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_RESERVED_SFT	1
+	uint8_t parallel_detect;
+
+	/*
+	 * The advertised speeds for the port by the link partner. Each
+	 * advertised speed will be set to '1'.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10MB \
+							UINT32_C(0x2000)
+	uint16_t link_partner_adv_speeds;
+
+	/*
+	 * The advertised autoneg for the port by the link partner. This field
+	 * is deprecated and should be set to 0.
+	 */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ALL_SPEEDS\
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_SPEED_MASK\
+							(UINT32_C(0x4) << 0)
+	uint8_t link_partner_adv_auto_mode;
+
+	/* The advertised pause settings on the port by the link partner. */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_PAUSE_TX \
+							UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_PAUSE_RX \
+							UINT32_C(0x2)
+	uint8_t link_partner_adv_pause;
+
+	/*
+	 * Current setting for link speed mask that is used to advertise speeds
+	 * during autonegotiation when EEE is enabled. This field is valid only
+	 * when eee_enabled flags is set to 1. The speeds specified in this
+	 * field shall be a subset of speeds specified in auto_link_speed_mask.
+	 */
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD1 \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD2 \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD3 \
+							UINT32_C(0x10)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD4 \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	uint16_t adv_eee_link_speed_mask;
+
+	/*
+	 * Current setting for link speed mask that is advertised by the link
+	 * partner when EEE is enabled. This field is valid only when
+	 * eee_enabled flags is set to 1.
+	 */
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD1 \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD2 \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD3 \
+							UINT32_C(0x10)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD4 \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	uint16_t link_partner_adv_eee_link_speed_mask;
+
+	/* This value represents transceiver identifier type. */
+	/*
+	 * Current setting of TX LPI timer in microseconds. This field is valid
+	 * only when_eee_enabled flag is set to 1 and tx_lpi_enabled is set to
+	 * 1.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_TX_LPI_TIMER_MASK \
+							UINT32_C(0xffffff)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_TX_LPI_TIMER_SFT         0
+	/* This value represents transceiver identifier type. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_MASK \
+							UINT32_C(0xff000000)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_SFT \
+							24
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 24)
+		/* SFP/SFP+/SFP28 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_SFP \
+							(UINT32_C(0x3) << 24)
+		/* QSFP */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFP \
+							(UINT32_C(0xc) << 24)
+		/* QSFP+ */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFPPLUS \
+							(UINT32_C(0xd) << 24)
+		/* QSFP28 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFP28 \
+							(UINT32_C(0x11) << 24)
+	uint32_t xcvr_identifier_type_tx_lpi_timer;
+
+	uint32_t unused_1;
+
+	/*
+	 * Up to 16 bytes of null padded ASCII string representing PHY vendor.
+	 * If the string is set to null, then the vendor name is not available.
+	 */
+	char phy_vendor_name[16];
+
+	/*
+	 * Up to 16 bytes of null padded ASCII string that identifies vendor
+	 * specific part number of the PHY. If the string is set to null, then
+	 * the vendor specific part number is not available.
+	 */
+	char phy_vendor_partnumber[16];
+
+	uint32_t unused_2;
+	uint8_t unused_3;
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_ver_get */
 /*
  * Description: This function is called by a driver to determine the HWRM
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (27 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 29/39] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-08 10:02   ` Bruce Richardson
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 31/39] bnxt: add promiscuous enable/disable operations Stephen Hurd
                   ` (11 subsequent siblings)
  40 siblings, 1 reply; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to add the start, stop and link update dev_ops.
The BNXT driver will now minimally pass traffic with testpmd.

v4:
- Fix issues pointed out by checkpatch.
- Shorten the string passed for reserving memzone
when default completion ring is created.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 269 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 269 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 6888363..ac82876 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -40,12 +40,17 @@
 #include <rte_cycles.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
 #include "bnxt_rxq.h"
 #include "bnxt_rxr.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
 #include "bnxt_txr.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -65,6 +70,177 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 	bnxt_free_hwrm_resources(bp);
 }
 
+/***********************/
+
+/*
+ * High level utility functions
+ */
+
+static void bnxt_free_mem(struct bnxt *bp)
+{
+	bnxt_free_filter_mem(bp);
+	bnxt_free_vnic_attributes(bp);
+	bnxt_free_vnic_mem(bp);
+
+	bnxt_free_stats(bp);
+	bnxt_free_tx_rings(bp);
+	bnxt_free_rx_rings(bp);
+	bnxt_free_def_cp_ring(bp);
+}
+
+static int bnxt_alloc_mem(struct bnxt *bp)
+{
+	int rc;
+
+	/* Default completion ring */
+	rc = bnxt_init_def_ring_struct(bp, SOCKET_ID_ANY);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_rings(bp, 0, NULL, NULL,
+			      bp->def_cp_ring, "def_cp");
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_vnic_mem(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_vnic_attributes(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_filter_mem(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	return 0;
+
+alloc_mem_err:
+	bnxt_free_mem(bp);
+	return rc;
+}
+
+static int bnxt_init_chip(struct bnxt *bp)
+{
+	unsigned int i, rss_idx, fw_idx;
+	int rc;
+
+	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM stat ctx alloc failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_alloc_hwrm_rings(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM ring alloc failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM ring grp alloc failure: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_mq_rx_configure(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "MQ mode configure failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	/* VNIC configuration */
+	for (i = 0; i < bp->nr_vnics; i++) {
+		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+		rc = bnxt_hwrm_vnic_alloc(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic alloc failure rc: %x\n",
+				rc);
+			goto err_out;
+		}
+
+		rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD,
+				"HWRM vnic ctx alloc failure rc: %x\n", rc);
+			goto err_out;
+		}
+
+		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic cfg failure rc: %x\n", rc);
+			goto err_out;
+		}
+
+		rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic filter failure rc: %x\n",
+				rc);
+			goto err_out;
+		}
+		if (vnic->rss_table && vnic->hash_type) {
+			/*
+			 * Fill the RSS hash & redirection table with
+			 * ring group ids for all VNICs
+			 */
+			for (rss_idx = 0, fw_idx = 0;
+			     rss_idx < HW_HASH_INDEX_SIZE;
+			     rss_idx++, fw_idx++) {
+				if (vnic->fw_grp_ids[fw_idx] ==
+				    INVALID_HW_RING_ID)
+					fw_idx = 0;
+				vnic->rss_table[rss_idx] =
+						vnic->fw_grp_ids[fw_idx];
+			}
+			rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+			if (rc) {
+				RTE_LOG(ERR, PMD,
+					"HWRM vnic set RSS failure rc: %x\n",
+					rc);
+				goto err_out;
+			}
+		}
+	}
+	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0]);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	return 0;
+
+err_out:
+	bnxt_free_all_hwrm_resources(bp);
+
+	return rc;
+}
+
+static int bnxt_shutdown_nic(struct bnxt *bp)
+{
+	bnxt_free_all_hwrm_resources(bp);
+	bnxt_free_all_filters(bp);
+	bnxt_free_all_vnics(bp);
+	return 0;
+}
+
+static int bnxt_init_nic(struct bnxt *bp)
+{
+	int rc;
+
+	bnxt_init_ring_grps(bp);
+	bnxt_init_vnics(bp);
+	bnxt_init_filters(bp);
+
+	rc = bnxt_init_chip(bp);
+	if (rc)
+		return rc;
+
+	return 0;
+}
+
 /*
  * Device configuration and status function
  */
@@ -182,6 +358,85 @@ static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
 	return rc;
 }
 
+static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	int rc;
+
+	rc = bnxt_hwrm_func_reset(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "hwrm chip reset failure rc: %x\n", rc);
+		rc = -1;
+		goto error;
+	}
+
+	rc = bnxt_alloc_mem(bp);
+	if (rc)
+		goto error;
+
+	rc = bnxt_init_nic(bp);
+	if (rc)
+		goto error;
+
+	return 0;
+
+error:
+	bnxt_shutdown_nic(bp);
+	bnxt_free_tx_mbufs(bp);
+	bnxt_free_rx_mbufs(bp);
+	bnxt_free_mem(bp);
+	return rc;
+}
+
+/* Unload the driver, release resources */
+static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	if (bp->eth_dev->data->dev_started) {
+		/* TBD: STOP HW queues DMA */
+		eth_dev->data->dev_link.link_status = 0;
+	}
+	bnxt_shutdown_nic(bp);
+}
+
+static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
+			       int wait_to_complete)
+{
+	int rc = 0;
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_link new;
+	unsigned int cnt = BNXT_LINK_WAIT_CNT;
+
+	memset(&new, 0, sizeof(new));
+	do {
+		/* Retrieve link info from hardware */
+		rc = bnxt_get_hwrm_link_config(bp, &new);
+		if (rc) {
+			new.link_speed = ETH_LINK_SPEED_100M;
+			new.link_duplex = ETH_LINK_FULL_DUPLEX;
+			RTE_LOG(ERR, PMD,
+				"Failed to retrieve link rc = 0x%x!", rc);
+			goto out;
+		}
+		if (!wait_to_complete)
+			break;
+
+		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
+
+	} while (!new.link_status && cnt--);
+
+	/* Timed out or success */
+	if (new.link_status) {
+		/* Update only if success */
+		eth_dev->data->dev_link.link_duplex = new.link_duplex;
+		eth_dev->data->dev_link.link_speed = new.link_speed;
+	}
+	eth_dev->data->dev_link.link_status = new.link_status;
+out:
+	return rc;
+}
+
 /*
  * Initialization
  */
@@ -190,12 +445,15 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.dev_start = bnxt_dev_start_op,
+	.dev_stop = bnxt_dev_stop_op,
 	.stats_get = bnxt_stats_get_op,
 	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
+	.link_update = bnxt_link_update_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
@@ -305,6 +563,15 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	else
 		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
+	bp->grp_info = rte_zmalloc("bnxt_grp_info",
+				sizeof(*bp->grp_info) * bp->max_ring_grps, 0);
+	if (!bp->grp_info) {
+		RTE_LOG(ERR, PMD,
+			"Failed to alloc %zu bytes needed to store group info table\n",
+			sizeof(*bp->grp_info) * bp->max_ring_grps);
+		rc = -ENOMEM;
+		goto error_free;
+	}
 
 	rc = bnxt_hwrm_func_driver_register(bp, 0,
 					    bp->pf.vf_req_fwd);
@@ -335,6 +602,8 @@ bnxt_dev_uninit(struct rte_eth_dev *eth_dev) {
 
 	if (eth_dev->data->mac_addrs)
 		rte_free(eth_dev->data->mac_addrs);
+	if (bp->grp_info)
+		rte_free(bp->grp_info);
 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
 	bnxt_free_hwrm_resources(bp);
 	return rc;
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 31/39] bnxt: add promiscuous enable/disable operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (28 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 32/39] bnxt: add all multicast " Stephen Hurd
                   ` (10 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the promiscuous mode enable and disable dev_ops.

v4:
Fix couple of typos in the commit message.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index ac82876..3fce540 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -437,6 +437,34 @@ out:
 	return rc;
 }
 
+static void bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
+static void bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
 /*
  * Initialization
  */
@@ -454,6 +482,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
 	.link_update = bnxt_link_update_op,
+	.promiscuous_enable = bnxt_promiscuous_enable_op,
+	.promiscuous_disable = bnxt_promiscuous_disable_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 32/39] bnxt: add all multicast enable/disable operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (29 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 31/39] bnxt: add promiscuous enable/disable operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 33/39] bnxt: free memory in close operation Stephen Hurd
                   ` (9 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds dev_ops to enable/disable multicast traffic.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3fce540..d3a624f 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -465,6 +465,34 @@ static void bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
 	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
 }
 
+static void bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
+static void bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
 /*
  * Initialization
  */
@@ -484,6 +512,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
+	.allmulticast_enable = bnxt_allmulticast_enable_op,
+	.allmulticast_disable = bnxt_allmulticast_disable_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 33/39] bnxt: free memory in close operation
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (30 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 32/39] bnxt: add all multicast " Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 34/39] bnxt: add MAC address add/remove dev_ops Stephen Hurd
                   ` (8 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to free all resources except the one corresponding
to HWRM, which are required to notify the HWRM that the driver is unloaded
(these are freed in uninit()).

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index d3a624f..4254531 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -62,14 +62,6 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
-static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
-{
-	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
-
-	rte_free(eth_dev->data->mac_addrs);
-	bnxt_free_hwrm_resources(bp);
-}
-
 /***********************/
 
 /*
@@ -388,6 +380,16 @@ error:
 	return rc;
 }
 
+static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	bnxt_free_tx_mbufs(bp);
+	bnxt_free_rx_mbufs(bp);
+	bnxt_free_mem(bp);
+	rte_free(eth_dev->data->mac_addrs);
+}
+
 /* Unload the driver, release resources */
 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 {
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 34/39] bnxt: add MAC address add/remove dev_ops
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (31 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 33/39] bnxt: free memory in close operation Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 35/39] bnxt: add set link up/down operations Stephen Hurd
                   ` (7 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds dev_ops to Add/Remove MAC addresses.

v4:
Fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 71 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 71 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 4254531..34a5873 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -402,6 +402,75 @@ static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 	bnxt_shutdown_nic(bp);
 }
 
+static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
+				    uint32_t index)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter, *temp_filter;
+	int i;
+
+	/*
+	 * Loop through all VNICs from the specified filter flow pools to
+	 * remove the corresponding MAC addr filter
+	 */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		if (!(pool_mask & (1 << i)))
+			continue;
+
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			filter = STAILQ_FIRST(&vnic->filter);
+			while (filter) {
+				temp_filter = STAILQ_NEXT(filter, next);
+				if (filter->mac_index == index) {
+					STAILQ_REMOVE(&vnic->filter, filter,
+						      bnxt_filter_info, next);
+					bnxt_hwrm_clear_filter(bp, filter);
+					filter->mac_index = INVALID_MAC_INDEX;
+					memset(&filter->l2_addr, 0,
+					       ETHER_ADDR_LEN);
+					STAILQ_INSERT_TAIL(
+							&bp->free_filter_list,
+							filter, next);
+				}
+				filter = temp_filter;
+			}
+		}
+	}
+}
+
+static void bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
+				 struct ether_addr *mac_addr,
+				 uint32_t index, uint32_t pool)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = STAILQ_FIRST(&bp->ff_pool[pool]);
+	struct bnxt_filter_info *filter;
+
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "VNIC not found for pool %d!\n", pool);
+		return;
+	}
+	/* Attach requested MAC address to the new l2_filter */
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		if (filter->mac_index == index) {
+			RTE_LOG(ERR, PMD,
+				"MAC addr already existed for pool %d\n", pool);
+			return;
+		}
+	}
+	filter = bnxt_alloc_filter(bp);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+		return;
+	}
+	STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+	filter->mac_index = index;
+	memcpy(filter->l2_addr, mac_addr, ETHER_ADDR_LEN);
+	bnxt_hwrm_set_filter(bp, vnic, filter);
+}
+
 static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
 			       int wait_to_complete)
 {
@@ -516,6 +585,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
 	.allmulticast_enable = bnxt_allmulticast_enable_op,
 	.allmulticast_disable = bnxt_allmulticast_disable_op,
+	.mac_addr_add = bnxt_mac_addr_add_op,
+	.mac_addr_remove = bnxt_mac_addr_remove_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 35/39] bnxt: add set link up/down operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (32 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 34/39] bnxt: add MAC address add/remove dev_ops Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 36/39] bnxt: add reta update/query operations Stephen Hurd
                   ` (6 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Adds dev_ops to set link UP or DOWN as appropriate.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 34a5873..b04010c 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -380,6 +380,24 @@ error:
 	return rc;
 }
 
+static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	eth_dev->data->dev_link.link_status = 1;
+	bnxt_set_hwrm_link_config(bp, true);
+	return 0;
+}
+
+static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	eth_dev->data->dev_link.link_status = 0;
+	bnxt_set_hwrm_link_config(bp, false);
+	return 0;
+}
+
 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
@@ -574,6 +592,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_configure = bnxt_dev_configure_op,
 	.dev_start = bnxt_dev_start_op,
 	.dev_stop = bnxt_dev_stop_op,
+	.dev_set_link_up = bnxt_dev_set_link_up_op,
+	.dev_set_link_down = bnxt_dev_set_link_down_op,
 	.stats_get = bnxt_stats_get_op,
 	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 36/39] bnxt: add reta update/query operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (33 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 35/39] bnxt: add set link up/down operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 37/39] bnxt: add RSS device operations Stephen Hurd
                   ` (5 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add code to Update/query reta dev_ops

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 56 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 56 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b04010c..b3b76f1 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -582,6 +582,60 @@ static void bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
 	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
 }
 
+static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
+			    struct rte_eth_rss_reta_entry64 *reta_conf,
+			    uint16_t reta_size)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_vnic_info *vnic;
+	int i;
+
+	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
+		return -EINVAL;
+
+	if (reta_size != HW_HASH_INDEX_SIZE) {
+		RTE_LOG(ERR, PMD, "The configured hash table lookup size "
+			"(%d) must equal the size supported by the hardware "
+			"(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+		return -EINVAL;
+	}
+	/* Update the RSS VNIC(s) */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			memcpy(vnic->rss_table, reta_conf, reta_size);
+
+			bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+		}
+	}
+	return 0;
+}
+
+static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
+			      struct rte_eth_rss_reta_entry64 *reta_conf,
+			      uint16_t reta_size)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+
+	/* Retrieve from the default VNIC */
+	if (!vnic)
+		return -EINVAL;
+	if (!vnic->rss_table)
+		return -EINVAL;
+
+	if (reta_size != HW_HASH_INDEX_SIZE) {
+		RTE_LOG(ERR, PMD, "The configured hash table lookup size "
+			"(%d) must equal the size supported by the hardware "
+			"(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+		return -EINVAL;
+	}
+	/* EW - need to revisit here copying from u64 to u16 */
+	memcpy(reta_conf, vnic->rss_table, reta_size);
+
+	return 0;
+}
+
 /*
  * Initialization
  */
@@ -600,6 +654,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
+	.reta_update = bnxt_reta_update_op,
+	.reta_query = bnxt_reta_query_op,
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 37/39] bnxt: add RSS device operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (34 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 36/39] bnxt: add reta update/query operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 38/39] bnxt: add flow control operations Stephen Hurd
                   ` (4 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add rss_hash_update and rss_hash_conf_get dev_ops

v4:
Fix issues pointed out by checkpatch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 121 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 121 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b3b76f1..3c7f868 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -62,6 +62,14 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
+#define BNXT_ETH_RSS_SUPPORT (	\
+	ETH_RSS_IPV4 |		\
+	ETH_RSS_NONFRAG_IPV4_TCP |	\
+	ETH_RSS_NONFRAG_IPV4_UDP |	\
+	ETH_RSS_IPV6 |		\
+	ETH_RSS_NONFRAG_IPV6_TCP |	\
+	ETH_RSS_NONFRAG_IPV6_UDP)
+
 /***********************/
 
 /*
@@ -636,6 +644,117 @@ static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
 	return 0;
 }
 
+static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
+				   struct rte_eth_rss_conf *rss_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_vnic_info *vnic;
+	uint16_t hash_type = 0;
+	int i;
+
+	/*
+	 * If RSS enablement were different than dev_configure,
+	 * then return -EINVAL
+	 */
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
+		if (!rss_conf->rss_hf)
+			return -EINVAL;
+	} else {
+		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
+			return -EINVAL;
+	}
+	if (rss_conf->rss_hf & ETH_RSS_IPV4)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV4_TCP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV4_UDP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_IPV6)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV6_TCP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV6_UDP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
+
+	/* Update the RSS VNIC(s) */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			vnic->hash_type = hash_type;
+
+			/*
+			 * Use the supplied key if the key length is
+			 * acceptable and the rss_key is not NULL
+			 */
+			if (rss_conf->rss_key &&
+			    rss_conf->rss_key_len <= HW_HASH_KEY_SIZE)
+				memcpy(vnic->rss_hash_key, rss_conf->rss_key,
+				       rss_conf->rss_key_len);
+
+			bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+		}
+	}
+	return 0;
+}
+
+static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
+				     struct rte_eth_rss_conf *rss_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+	int len;
+	uint32_t hash_types;
+
+	/* RSS configuration is the same for all VNICs */
+	if (vnic && vnic->rss_hash_key) {
+		if (rss_conf->rss_key) {
+			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
+			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
+			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
+		}
+
+		hash_types = vnic->hash_type;
+		rss_conf->rss_hf = 0;
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_IPV4;
+			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_IPV6;
+			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
+		}
+		if (hash_types) {
+			RTE_LOG(ERR, PMD,
+				"Unknwon RSS config from firmware (%08x), RSS disabled",
+				vnic->hash_type);
+			return -ENOTSUP;
+		}
+	} else {
+		rss_conf->rss_hf = 0;
+	}
+	return 0;
+}
+
 /*
  * Initialization
  */
@@ -656,6 +775,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.tx_queue_release = bnxt_tx_queue_release_op,
 	.reta_update = bnxt_reta_update_op,
 	.reta_query = bnxt_reta_query_op,
+	.rss_hash_update = bnxt_rss_hash_update_op,
+	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 38/39] bnxt: add flow control operations
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (35 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 37/39] bnxt: add RSS device operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 39/39] bnxt: Replace bnxt_ring_struct with bnxt_ring Stephen Hurd
                   ` (3 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add flow_ctrl_get and flow_ctrl_set dev_ops.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_ethdev.c | 83 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 83 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3c7f868..406e38a 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -755,6 +755,87 @@ static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
 	return 0;
 }
 
+static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
+			       struct rte_eth_fc_conf *fc_conf __rte_unused)
+{
+	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+	struct rte_eth_link link_info;
+	int rc;
+
+	rc = bnxt_get_hwrm_link_config(bp, &link_info);
+	if (rc)
+		return rc;
+
+	memset(fc_conf, 0, sizeof(*fc_conf));
+	if (bp->link_info.auto_pause)
+		fc_conf->autoneg = 1;
+	switch (bp->link_info.pause) {
+	case 0:
+		fc_conf->mode = RTE_FC_NONE;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
+		fc_conf->mode = RTE_FC_TX_PAUSE;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
+		fc_conf->mode = RTE_FC_RX_PAUSE;
+		break;
+	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
+			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
+		fc_conf->mode = RTE_FC_FULL;
+		break;
+	}
+	return 0;
+}
+
+static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
+			       struct rte_eth_fc_conf *fc_conf)
+{
+	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+
+	switch (fc_conf->mode) {
+	case RTE_FC_NONE:
+		bp->link_info.auto_pause = 0;
+		bp->link_info.force_pause = 0;
+		break;
+	case RTE_FC_RX_PAUSE:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
+		}
+		break;
+	case RTE_FC_TX_PAUSE:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
+		}
+		break;
+	case RTE_FC_FULL:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
+		}
+		break;
+	}
+	return bnxt_set_hwrm_link_config(bp, true);
+}
+
 /*
  * Initialization
  */
@@ -784,6 +865,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.allmulticast_disable = bnxt_allmulticast_disable_op,
 	.mac_addr_add = bnxt_mac_addr_add_op,
 	.mac_addr_remove = bnxt_mac_addr_remove_op,
+	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
+	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v4 39/39] bnxt: Replace bnxt_ring_struct with bnxt_ring
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (36 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 38/39] bnxt: add flow control operations Stephen Hurd
@ 2016-06-06 22:08 ` Stephen Hurd
  2016-06-07  6:25 ` [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Thomas Monjalon
                   ` (2 subsequent siblings)
  40 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-06 22:08 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

As pointed in the previous round of review,
Having struct at the end of the structure bnxt_ring_struct is a redundant.
Replace it with bnxt_ring.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
---
 drivers/net/bnxt/bnxt_cpr.c  |  4 ++--
 drivers/net/bnxt/bnxt_cpr.h  |  4 ++--
 drivers/net/bnxt/bnxt_hwrm.c | 10 +++++-----
 drivers/net/bnxt/bnxt_hwrm.h |  4 ++--
 drivers/net/bnxt/bnxt_ring.c | 20 ++++++++++----------
 drivers/net/bnxt/bnxt_ring.h |  4 ++--
 drivers/net/bnxt/bnxt_rxr.c  | 10 +++++-----
 drivers/net/bnxt/bnxt_rxr.h  |  2 +-
 drivers/net/bnxt/bnxt_txr.c  |  8 ++++----
 drivers/net/bnxt/bnxt_txr.h  |  2 +-
 10 files changed, 34 insertions(+), 34 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index 98f3ca2..f0bfa1f 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -133,7 +133,7 @@ void bnxt_free_def_cp_ring(struct bnxt *bp)
 int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id)
 {
 	struct bnxt_cp_ring_info *cpr;
-	struct bnxt_ring_struct *ring;
+	struct bnxt_ring *ring;
 
 	cpr = rte_zmalloc_socket("bnxt_cp_ring",
 				 sizeof(struct bnxt_cp_ring_info),
@@ -143,7 +143,7 @@ int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id)
 	bp->def_cp_ring = cpr;
 
 	ring = rte_zmalloc_socket("bnxt_cp_ring_struct",
-				  sizeof(struct bnxt_ring_struct),
+				  sizeof(struct bnxt_ring),
 				  RTE_CACHE_LINE_SIZE, socket_id);
 	if (ring == NULL)
 		return -ENOMEM;
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index 3e25a75..c176f8c 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -57,7 +57,7 @@
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
 				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
-struct bnxt_ring_struct;
+struct bnxt_ring;
 struct bnxt_cp_ring_info {
 	uint32_t		cp_raw_cons;
 	void			*cp_doorbell;
@@ -70,7 +70,7 @@ struct bnxt_cp_ring_info {
 	phys_addr_t		hw_stats_map;
 	uint32_t		hw_stats_ctx_id;
 
-	struct bnxt_ring_struct	*cp_ring_struct;
+	struct bnxt_ring	*cp_ring_struct;
 };
 
 #define RX_CMP_L2_ERRORS						\
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 978e379..5d81a60 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -587,7 +587,7 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 }
 
 int bnxt_hwrm_ring_alloc(struct bnxt *bp,
-			 struct bnxt_ring_struct *ring,
+			 struct bnxt_ring *ring,
 			 uint32_t ring_type, uint32_t map_index,
 			 uint32_t stats_ctx_id)
 {
@@ -660,7 +660,7 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 }
 
 int bnxt_hwrm_ring_free(struct bnxt *bp,
-			struct bnxt_ring_struct *ring, uint32_t ring_type)
+			struct bnxt_ring *ring, uint32_t ring_type)
 {
 	int rc;
 	struct hwrm_ring_free_input req = {.req_type = 0 };
@@ -1058,7 +1058,7 @@ int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
 static void bnxt_free_cp_ring(struct bnxt *bp,
 			      struct bnxt_cp_ring_info *cpr, unsigned int idx)
 {
-	struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+	struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
 
 	bnxt_hwrm_ring_free(bp, cp_ring,
 			HWRM_RING_FREE_INPUT_RING_TYPE_CMPL);
@@ -1077,7 +1077,7 @@ int bnxt_free_all_hwrm_rings(struct bnxt *bp)
 	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
 		struct bnxt_tx_queue *txq = bp->tx_queues[i];
 		struct bnxt_tx_ring_info *txr = txq->tx_ring;
-		struct bnxt_ring_struct *ring = txr->tx_ring_struct;
+		struct bnxt_ring *ring = txr->tx_ring_struct;
 		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
 		unsigned int idx = bp->rx_cp_nr_rings + i + 1;
 
@@ -1101,7 +1101,7 @@ int bnxt_free_all_hwrm_rings(struct bnxt *bp)
 	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
 		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
 		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
-		struct bnxt_ring_struct *ring = rxr->rx_ring_struct;
+		struct bnxt_ring *ring = rxr->rx_ring_struct;
 		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
 		unsigned int idx = i + 1;
 
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index d1aee1c..7ad5f51 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -62,11 +62,11 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
 int bnxt_hwrm_ring_alloc(struct bnxt *bp,
-			 struct bnxt_ring_struct *ring,
+			 struct bnxt_ring *ring,
 			 uint32_t ring_type, uint32_t map_index,
 			 uint32_t stats_ctx_id);
 int bnxt_hwrm_ring_free(struct bnxt *bp,
-			struct bnxt_ring_struct *ring, uint32_t ring_type);
+			struct bnxt_ring *ring, uint32_t ring_type);
 int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx);
 int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index 49cfab9..4e0bd35 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -48,7 +48,7 @@
  * Generic ring handling
  */
 
-void bnxt_free_ring(struct bnxt_ring_struct *ring)
+void bnxt_free_ring(struct bnxt_ring *ring)
 {
 	if (ring->vmem_size && *ring->vmem) {
 		memset((char *)*ring->vmem, 0, ring->vmem_size);
@@ -77,7 +77,7 @@ void bnxt_init_ring_grps(struct bnxt *bp)
  *
  * Order in the allocation is:
  * stats - Always non-zero length
- * cp vmem - Always zero-length, supported for the bnxt_ring_struct abstraction
+ * cp vmem - Always zero-length, supported for the bnxt_ring abstraction
  * tx vmem - Only non-zero length if tx_ring_info is not NULL
  * rx vmem - Only non-zero length if rx_ring_info is not NULL
  * cp bd ring - Always non-zero length
@@ -90,9 +90,9 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 			    struct bnxt_cp_ring_info *cp_ring_info,
 			    const char *suffix)
 {
-	struct bnxt_ring_struct *cp_ring = cp_ring_info->cp_ring_struct;
-	struct bnxt_ring_struct *rx_ring;
-	struct bnxt_ring_struct *tx_ring;
+	struct bnxt_ring *cp_ring = cp_ring_info->cp_ring_struct;
+	struct bnxt_ring *rx_ring;
+	struct bnxt_ring *tx_ring;
 	struct rte_pci_device *pdev = bp->pdev;
 	const struct rte_memzone *mz = NULL;
 	char mz_name[RTE_MEMZONE_NAMESIZE];
@@ -214,7 +214,7 @@ int bnxt_alloc_hwrm_rings(struct bnxt *bp)
 	/* Default completion ring */
 	{
 		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-		struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
 
 		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
 					  HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
@@ -230,9 +230,9 @@ int bnxt_alloc_hwrm_rings(struct bnxt *bp)
 	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
 		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
 		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
-		struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
 		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
-		struct bnxt_ring_struct *ring = rxr->rx_ring_struct;
+		struct bnxt_ring *ring = rxr->rx_ring_struct;
 		unsigned int idx = i + 1;
 
 		/* Rx cmpl */
@@ -270,9 +270,9 @@ int bnxt_alloc_hwrm_rings(struct bnxt *bp)
 	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
 		struct bnxt_tx_queue *txq = bp->tx_queues[i];
 		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
-		struct bnxt_ring_struct *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
 		struct bnxt_tx_ring_info *txr = txq->tx_ring;
-		struct bnxt_ring_struct *ring = txr->tx_ring_struct;
+		struct bnxt_ring *ring = txr->tx_ring_struct;
 		unsigned int idx = 1 + bp->rx_cp_nr_rings + i;
 
 		/* Tx cmpl */
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index e548586..8656549 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -66,7 +66,7 @@
 
 #define INVALID_HW_RING_ID      ((uint16_t)-1)
 
-struct bnxt_ring_struct {
+struct bnxt_ring {
 	void			*bd;
 	phys_addr_t		bd_dma;
 	uint32_t		ring_size;
@@ -91,7 +91,7 @@ struct bnxt;
 struct bnxt_tx_ring_info;
 struct bnxt_rx_ring_info;
 struct bnxt_cp_ring_info;
-void bnxt_free_ring(struct bnxt_ring_struct *ring);
+void bnxt_free_ring(struct bnxt_ring *ring);
 void bnxt_init_ring_grps(struct bnxt *bp);
 int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 			    struct bnxt_tx_ring_info *tx_ring_info,
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
index 0ac6f21..5d93de2 100644
--- a/drivers/net/bnxt/bnxt_rxr.c
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -273,7 +273,7 @@ int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 	struct bnxt *bp = rxq->bp;
 	struct bnxt_cp_ring_info *cpr;
 	struct bnxt_rx_ring_info *rxr;
-	struct bnxt_ring_struct *ring;
+	struct bnxt_ring *ring;
 
 	rxq->rx_buf_use_size = bp->eth_dev->data->mtu +
 			       ETHER_HDR_LEN + ETHER_CRC_LEN +
@@ -288,7 +288,7 @@ int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 	rxq->rx_ring = rxr;
 
 	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
-				   sizeof(struct bnxt_ring_struct),
+				   sizeof(struct bnxt_ring),
 				   RTE_CACHE_LINE_SIZE, socket_id);
 	if (ring == NULL)
 		return -ENOMEM;
@@ -308,7 +308,7 @@ int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 	rxq->cp_ring = cpr;
 
 	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
-				   sizeof(struct bnxt_ring_struct),
+				   sizeof(struct bnxt_ring),
 				   RTE_CACHE_LINE_SIZE, socket_id);
 	if (ring == NULL)
 		return -ENOMEM;
@@ -323,7 +323,7 @@ int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 	return 0;
 }
 
-static void bnxt_init_rxbds(struct bnxt_ring_struct *ring, uint32_t type,
+static void bnxt_init_rxbds(struct bnxt_ring *ring, uint32_t type,
 			    uint16_t len)
 {
 	uint32_t j;
@@ -341,7 +341,7 @@ static void bnxt_init_rxbds(struct bnxt_ring_struct *ring, uint32_t type,
 int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
 {
 	struct bnxt_rx_ring_info *rxr;
-	struct bnxt_ring_struct *ring;
+	struct bnxt_ring *ring;
 	uint32_t prod, type;
 	unsigned int i;
 
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
index 7ba8f7b..f766b26 100644
--- a/drivers/net/bnxt/bnxt_rxr.h
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -50,7 +50,7 @@ struct bnxt_rx_ring_info {
 
 	phys_addr_t		rx_desc_mapping;
 
-	struct bnxt_ring_struct	*rx_ring_struct;
+	struct bnxt_ring	*rx_ring_struct;
 };
 
 uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
index eef01c7..8bf8fee 100644
--- a/drivers/net/bnxt/bnxt_txr.c
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -74,7 +74,7 @@ void bnxt_free_tx_rings(struct bnxt *bp)
 int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
 {
 	struct bnxt_tx_ring_info *txr = txq->tx_ring;
-	struct bnxt_ring_struct *ring = txr->tx_ring_struct;
+	struct bnxt_ring *ring = txr->tx_ring_struct;
 
 	txq->tx_wake_thresh = ring->ring_size / 2;
 	ring->fw_ring_id = INVALID_HW_RING_ID;
@@ -86,7 +86,7 @@ int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id)
 {
 	struct bnxt_cp_ring_info *cpr;
 	struct bnxt_tx_ring_info *txr;
-	struct bnxt_ring_struct *ring;
+	struct bnxt_ring *ring;
 
 	txr = rte_zmalloc_socket("bnxt_tx_ring",
 				 sizeof(struct bnxt_tx_ring_info),
@@ -96,7 +96,7 @@ int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id)
 	txq->tx_ring = txr;
 
 	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
-				  sizeof(struct bnxt_ring_struct),
+				  sizeof(struct bnxt_ring),
 				  RTE_CACHE_LINE_SIZE, socket_id);
 	if (ring == NULL)
 		return -ENOMEM;
@@ -116,7 +116,7 @@ int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id)
 	txq->cp_ring = cpr;
 
 	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
-				  sizeof(struct bnxt_ring_struct),
+				  sizeof(struct bnxt_ring),
 				  RTE_CACHE_LINE_SIZE, socket_id);
 	if (ring == NULL)
 		return -ENOMEM;
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
index 56d0b08..2176aca 100644
--- a/drivers/net/bnxt/bnxt_txr.h
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -53,7 +53,7 @@ struct bnxt_tx_ring_info {
 #define BNXT_DEV_STATE_CLOSING	0x1
 	uint32_t		dev_state;
 
-	struct bnxt_ring_struct	*tx_ring_struct;
+	struct bnxt_ring	*tx_ring_struct;
 };
 
 struct bnxt_sw_tx_bd {
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (37 preceding siblings ...)
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 39/39] bnxt: Replace bnxt_ring_struct with bnxt_ring Stephen Hurd
@ 2016-06-07  6:25 ` Thomas Monjalon
  2016-06-08 10:31   ` Bruce Richardson
  2016-06-08 10:21 ` Bruce Richardson
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
  40 siblings, 1 reply; 132+ messages in thread
From: Thomas Monjalon @ 2016-06-07  6:25 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde, bruce.richardson

Hi Stephen,
Reminder from http://dpdk.org/dev#send:

git send-email -39 -v4 --cover-letter --annotate
	--in-reply-to <Message-ID of the previous patch>

Please do not forget --in-reply-to. Thanks

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations
  2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations Stephen Hurd
@ 2016-06-08 10:02   ` Bruce Richardson
  2016-06-08 10:41     ` Bruce Richardson
  0 siblings, 1 reply; 132+ messages in thread
From: Bruce Richardson @ 2016-06-08 10:02 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Mon, Jun 06, 2016 at 03:08:34PM -0700, Stephen Hurd wrote:
> From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> 
> This patch adds code to add the start, stop and link update dev_ops.
> The BNXT driver will now minimally pass traffic with testpmd.
> 
> v4:
> - Fix issues pointed out by checkpatch.
> - Shorten the string passed for reserving memzone
> when default completion ring is created.
> 
> Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> Reviewed-by: David Christensen <david.christensen@broadcom.com>
> Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> ---
>  drivers/net/bnxt/bnxt_ethdev.c | 269 +++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 269 insertions(+)
> 
I get compilation errors after applying this patch:

== Build drivers/net/bnxt
  CC bnxt_ethdev.o
/home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c: In function ‘bnxt_init_chip’:
/home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:135:7: error: implicit declaration of function ‘bnxt_alloc_hwrm_rings’ [-Werror=implicit-function-declaration]
  rc = bnxt_alloc_hwrm_rings(bp);
       ^
/home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:135:2: error: nested extern declaration of ‘bnxt_alloc_hwrm_rings’ [-Werror=nested-externs]
  rc = bnxt_alloc_hwrm_rings(bp);
  ^
/home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c: In function ‘bnxt_init_nic’:
/home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:233:2: error: implicit declaration of function ‘bnxt_init_ring_grps’ [-Werror=implicit-function-declaration]
  bnxt_init_ring_grps(bp);
  ^
/home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:233:2: error: nested extern declaration of ‘bnxt_init_ring_grps’ [-Werror=nested-externs]
cc1: all warnings being treated as errors
/home/bruce/next-net/dpdk-next-net/mk/internal/rte.compile-pre.mk:126: recipe for target 'bnxt_ethdev.o' failed
make[5]: *** [bnxt_ethdev.o] Error 1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (38 preceding siblings ...)
  2016-06-07  6:25 ` [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Thomas Monjalon
@ 2016-06-08 10:21 ` Bruce Richardson
  2016-06-08 10:41   ` Bruce Richardson
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
  40 siblings, 1 reply; 132+ messages in thread
From: Bruce Richardson @ 2016-06-08 10:21 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Mon, Jun 06, 2016 at 03:08:05PM -0700, Stephen Hurd wrote:
> From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> 
> This patch adds the initial skeleton for bnxt driver along with the
> nic guide to tie into the build system.
> At this point, the driver simply fails init.
> 
> v4:
> Fix a warning that the document isn't included in any toctree
> Also remove a PCI ID added erroneously.
> 
> Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> Reviewed-by: David Christensen <david.christensen@broadcom.com>
> Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> ---
Hi Stephen, Ajit,

in the absense of a cover letter, I'll post my overall comments on this set here.

Thanks for the updated v4, I'm not seeing any checkpatch issues with the patches
that have applied and compiled up cleanly. However,

* the build is broken by patch 30, and none of the later patches 31-38 seem
to fix it for me. Is there a header file include missing in that patch or 
something? [I'm using gcc 5.3.1 on Fedora 23]
* patch 39 fails to apply for me with rejects on other files in the driver,
which is very strange. [drivers/net/bnxt/bnxt_hwrm.c, 
drivers/net/bnxt/bnxt_ring.c and drivers/net/bnxt/bnxt_ring.h]

Apart from this, the other concern I still have is with the explanation
accompaning some of the patches, especially for those to with rings. There are
many patches throughout the set which seem to be doing the same thing, adding
allocate and free functions for rings. 

For example:
Patch 28 is titled "add ring alloc, free and group init". For a start it's
unclear from the title, whether the alloc and free refers to individual rings
or to the groups. If it's referring to the rings themselves, then how is this
different functionality from:
Patch 7: add ring structs and free() func
Patch 10/11: add TX/RX queue create/destroy operations
Patch 15: code to alloc/free ring
Patch 24: add HWRM ring alloc/free functions

Or if it's to do with allocating and freeing the groups, it would seem to be
the same functionality as patch 25: "add ring group alloc/free functions".

In some cases, the commit message does add some detail, e.g. patches 7 and 10
point out what they don't cover, but the rest is still very unclear, as to what
each of the 5/6 patches for ring create/free are really doing and how they
work together. I'm not sure exactly how best to do this without understanding
the details of these patches, but one way might be to list out the different
part of the ring allocation/free in each patch and then explain what part of
that process this patch is doing and how it fits in the sequence. Otherwise,
maybe some of the patches may need to be merged if they are very closely related.

Can you please look to improve the commit messages when you do rework to fix
the compilation and patch application errors.

Thanks,
/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-07  6:25 ` [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Thomas Monjalon
@ 2016-06-08 10:31   ` Bruce Richardson
  0 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-08 10:31 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: Stephen Hurd, dev, ajit.khaparde

On Tue, Jun 07, 2016 at 08:25:44AM +0200, Thomas Monjalon wrote:
> Hi Stephen,
> Reminder from http://dpdk.org/dev#send:
> 
> git send-email -39 -v4 --cover-letter --annotate
> 	--in-reply-to <Message-ID of the previous patch>
> 
> Please do not forget --in-reply-to. Thanks

Three other minor style updates as well that would make my life a little easier:

* The sign-off type tags at the bottom of the email are kept in logical
chronological order (irrespective of actual chronological order). This means
that all signed-off-by tags go together followed by a reviewed-by tag afterwards.

* Watch out for initial caps at the start of a commit title. [This will be caught
by the check-git-log.sh script if you want to run that]

* The history updates from vN to vN+1 are best put after the signoffs and prefixed
with a cutline, so that they get stripped automatically when the patch is applied.
Otherwise I have to delete them manually from the commit message. The format of
the messages is therefore best done as:

	bnxt: <title>

	<commit message>

	Signed-off-by: <author1>
	Signed-off-by: <author2>
	Reviewed-by: <reviewer1>

	---
	vN changes
	<description of changes>

Thanks,
/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-08 10:21 ` Bruce Richardson
@ 2016-06-08 10:41   ` Bruce Richardson
  2016-06-14 13:14     ` Bruce Richardson
  0 siblings, 1 reply; 132+ messages in thread
From: Bruce Richardson @ 2016-06-08 10:41 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Wed, Jun 08, 2016 at 11:21:23AM +0100, Bruce Richardson wrote:
> On Mon, Jun 06, 2016 at 03:08:05PM -0700, Stephen Hurd wrote:
> > From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> > 
> > This patch adds the initial skeleton for bnxt driver along with the
> > nic guide to tie into the build system.
> > At this point, the driver simply fails init.
> > 
> > v4:
> > Fix a warning that the document isn't included in any toctree
> > Also remove a PCI ID added erroneously.
> > 
> > Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> > Reviewed-by: David Christensen <david.christensen@broadcom.com>
> > Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> > ---
> Hi Stephen, Ajit,
> 
> in the absense of a cover letter, I'll post my overall comments on this set here.
> 
> Thanks for the updated v4, I'm not seeing any checkpatch issues with the patches
> that have applied and compiled up cleanly. However,
> 
> * the build is broken by patch 30, and none of the later patches 31-38 seem
> to fix it for me. Is there a header file include missing in that patch or 
> something? [I'm using gcc 5.3.1 on Fedora 23]
> * patch 39 fails to apply for me with rejects on other files in the driver,
> which is very strange. [drivers/net/bnxt/bnxt_hwrm.c, 
> drivers/net/bnxt/bnxt_ring.c and drivers/net/bnxt/bnxt_ring.h]

Sorry, my bad here, the build is not broken, the patch 30 now applies fine
and compiles ok. I had somehow missed applying an earlier patch (patch 28)
after reviewing it, and that causes the failures.

Please ignore the above comments.

> 
> Apart from this, the other concern I still have is with the explanation
> accompaning some of the patches, especially for those to with rings. There are
> many patches throughout the set which seem to be doing the same thing, adding
> allocate and free functions for rings. 
> 
> For example:
> Patch 28 is titled "add ring alloc, free and group init". For a start it's
> unclear from the title, whether the alloc and free refers to individual rings
> or to the groups. If it's referring to the rings themselves, then how is this
> different functionality from:
> Patch 7: add ring structs and free() func
> Patch 10/11: add TX/RX queue create/destroy operations
> Patch 15: code to alloc/free ring
> Patch 24: add HWRM ring alloc/free functions
> 
> Or if it's to do with allocating and freeing the groups, it would seem to be
> the same functionality as patch 25: "add ring group alloc/free functions".
> 
> In some cases, the commit message does add some detail, e.g. patches 7 and 10
> point out what they don't cover, but the rest is still very unclear, as to what
> each of the 5/6 patches for ring create/free are really doing and how they
> work together. I'm not sure exactly how best to do this without understanding
> the details of these patches, but one way might be to list out the different
> part of the ring allocation/free in each patch and then explain what part of
> that process this patch is doing and how it fits in the sequence. Otherwise,
> maybe some of the patches may need to be merged if they are very closely related.
> 
> Can you please look to improve the commit messages when you do rework to fix
> the compilation and patch application errors.

Since the build break was my mistake, a new rev of the patches may not be
absolutely necessary. If it's more convenient and is not too complicated, you
can perhaps just post updated comments for the above-mentioned patches to the
list and I can fix up the commit messages on patch apply. However, the patches
and commit messages are quite confusing to read as they are right now.

Thanks,
/Bruce
> 
> Thanks,
> /Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations
  2016-06-08 10:02   ` Bruce Richardson
@ 2016-06-08 10:41     ` Bruce Richardson
  0 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-08 10:41 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Wed, Jun 08, 2016 at 11:02:08AM +0100, Bruce Richardson wrote:
> On Mon, Jun 06, 2016 at 03:08:34PM -0700, Stephen Hurd wrote:
> > From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> > 
> > This patch adds code to add the start, stop and link update dev_ops.
> > The BNXT driver will now minimally pass traffic with testpmd.
> > 
> > v4:
> > - Fix issues pointed out by checkpatch.
> > - Shorten the string passed for reserving memzone
> > when default completion ring is created.
> > 
> > Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> > Reviewed-by: David Christensen <david.christensen@broadcom.com>
> > Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> > ---
> >  drivers/net/bnxt/bnxt_ethdev.c | 269 +++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 269 insertions(+)
> > 
> I get compilation errors after applying this patch:
> 
> == Build drivers/net/bnxt
>   CC bnxt_ethdev.o
> /home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c: In function ‘bnxt_init_chip’:
> /home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:135:7: error: implicit declaration of function ‘bnxt_alloc_hwrm_rings’ [-Werror=implicit-function-declaration]
>   rc = bnxt_alloc_hwrm_rings(bp);
>        ^
> /home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:135:2: error: nested extern declaration of ‘bnxt_alloc_hwrm_rings’ [-Werror=nested-externs]
>   rc = bnxt_alloc_hwrm_rings(bp);
>   ^
> /home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c: In function ‘bnxt_init_nic’:
> /home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:233:2: error: implicit declaration of function ‘bnxt_init_ring_grps’ [-Werror=implicit-function-declaration]
>   bnxt_init_ring_grps(bp);
>   ^
> /home/bruce/next-net/dpdk-next-net/drivers/net/bnxt/bnxt_ethdev.c:233:2: error: nested extern declaration of ‘bnxt_init_ring_grps’ [-Werror=nested-externs]
> cc1: all warnings being treated as errors
> /home/bruce/next-net/dpdk-next-net/mk/internal/rte.compile-pre.mk:126: recipe for target 'bnxt_ethdev.o' failed
> make[5]: *** [bnxt_ethdev.o] Error 1
> 
Please ignore, my mistake, as I was missing patch 28.

/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-08 10:41   ` Bruce Richardson
@ 2016-06-14 13:14     ` Bruce Richardson
  0 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-14 13:14 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Wed, Jun 08, 2016 at 11:41:23AM +0100, Bruce Richardson wrote:
> On Wed, Jun 08, 2016 at 11:21:23AM +0100, Bruce Richardson wrote:
> > On Mon, Jun 06, 2016 at 03:08:05PM -0700, Stephen Hurd wrote:
> > > From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> > > 
<snip> 
> Since the build break was my mistake, a new rev of the patches may not be
> absolutely necessary. If it's more convenient and is not too complicated, you
> can perhaps just post updated comments for the above-mentioned patches to the
> list and I can fix up the commit messages on patch apply. However, the patches
> and commit messages are quite confusing to read as they are right now.
> 
Ping on this.

To apply the set, I'm still looking for, at minimum, commit message
clarification for the patches which set up the NIC queues. The deadline by which
all patches are meant to be ready for merge is at the end of this week, so
please prioritize this so we can get this driver into 16.07 release.

Thanks,
/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 01/38] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                   ` (39 preceding siblings ...)
  2016-06-08 10:21 ` Bruce Richardson
@ 2016-06-14 22:55 ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code Stephen Hurd
                     ` (39 more replies)
  40 siblings, 40 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the initial skeleton for bnxt driver along with the
nic guide to tie into the build system.
At this point, the driver simply fails init.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix a warning that the document isn't included in any toctree
Also remove a PCI ID added erroneously.

v5:
Re-format footer
---
 MAINTAINERS                                     |   5 ++
 config/common_base                              |   5 ++
 doc/guides/nics/bnxt.rst                        |  49 +++++++++++
 doc/guides/nics/index.rst                       |   1 +
 drivers/net/Makefile                            |   1 +
 drivers/net/bnxt/Makefile                       |  63 ++++++++++++++
 drivers/net/bnxt/bnxt_ethdev.c                  | 109 ++++++++++++++++++++++++
 drivers/net/bnxt/rte_pmd_bnxt_version.map       |   4 +
 lib/librte_eal/common/include/rte_pci_dev_ids.h |  38 +++++++--
 mk/rte.app.mk                                   |   1 +
 10 files changed, 271 insertions(+), 5 deletions(-)
 create mode 100644 doc/guides/nics/bnxt.rst
 create mode 100644 drivers/net/bnxt/Makefile
 create mode 100644 drivers/net/bnxt/bnxt_ethdev.c
 create mode 100644 drivers/net/bnxt/rte_pmd_bnxt_version.map

diff --git a/MAINTAINERS b/MAINTAINERS
index 3e6b70c..78d8530 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -401,6 +401,11 @@ M: Declan Doherty <declan.doherty@intel.com>
 F: drivers/crypto/null/
 F: doc/guides/cryptodevs/null.rst
 
+Broadcom BNXT PMD
+M: Stephen Hurd <stephen.hurd@broadcom.com>
+F: drivers/net/bnxt/
+F: doc/guides/nics/bnxt.rst
+
 
 Packet processing
 -----------------
diff --git a/config/common_base b/config/common_base
index b9ba405..71be7cd 100644
--- a/config/common_base
+++ b/config/common_base
@@ -245,6 +245,11 @@ CONFIG_RTE_LIBRTE_NFP_PMD=n
 CONFIG_RTE_LIBRTE_NFP_DEBUG=n
 
 #
+# Compile burst-oriented Broadcom BNXT PMD driver
+#
+CONFIG_RTE_LIBRTE_BNXT_PMD=y
+
+#
 # Compile software PMD backed by SZEDATA2 device
 #
 CONFIG_RTE_LIBRTE_PMD_SZEDATA2=n
diff --git a/doc/guides/nics/bnxt.rst b/doc/guides/nics/bnxt.rst
new file mode 100644
index 0000000..2669e98
--- /dev/null
+++ b/doc/guides/nics/bnxt.rst
@@ -0,0 +1,49 @@
+..  BSD LICENSE
+    Copyright 2016 Broadcom Limited
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions
+    are met:
+
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in
+    the documentation and/or other materials provided with the
+    distribution.
+    * Neither the name of Broadcom Limited nor the names of its
+    contributors may be used to endorse or promote products derived
+    from this software without specific prior written permission.
+
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+bnxt poll mode driver library
+=============================
+
+The bnxt poll mode library (**librte_pmd_bnxt**) implements support for
+**Broadcom NetXtreme® C-Series**.  These adapters support Standards-
+compliant 10/25/50Gbps 30MPPS full-duplex throughput.
+
+Information about this family of adapters can be found in the
+`NetXtreme® Brand section <https://www.broadcom.com/products/ethernet-communication-and-switching?technology%5B%5D=88>`_
+of the `Broadcom web site <http://www.broadcom.com/>`_.
+
+Limitations
+-----------
+
+With the current driver, allocated mbufs must be large enough to hold
+the entire received frame.  If the mbufs are not large enough, the
+packets will be dropped.  This is most limiting when jumbo frames are
+used.
+
+SR-IOV is not supported.
diff --git a/doc/guides/nics/index.rst b/doc/guides/nics/index.rst
index 0b13698..ffe011e 100644
--- a/doc/guides/nics/index.rst
+++ b/doc/guides/nics/index.rst
@@ -36,6 +36,7 @@ Network Interface Controller Drivers
     :numbered:
 
     overview
+    bnxt
     bnx2x
     cxgbe
     e1000em
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index 6ba7658..3832706 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -45,6 +45,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_MLX4_PMD) += mlx4
 DIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5
 DIRS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD) += mpipe
 DIRS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp
+DIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_NULL) += null
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_PCAP) += pcap
 DIRS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede
diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
new file mode 100644
index 0000000..f6333fd
--- /dev/null
+++ b/drivers/net/bnxt/Makefile
@@ -0,0 +1,63 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
+#   Copyright(c) 2014 6WIND S.A.
+#   Copyright(c) Broadcom Limited.
+#   All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_bnxt.a
+
+LIBABIVER := 1
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_bnxt_version.map
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+
+#
+# Export include files
+#
+SYMLINK-y-include +=
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_eal
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
new file mode 100644
index 0000000..af8a65f
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -0,0 +1,109 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include <rte_dev.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+
+#define DRV_MODULE_NAME		"bnxt"
+static const char bnxt_version[] =
+	"Broadcom Cumulus driver " DRV_MODULE_NAME "\n";
+
+static struct rte_pci_id bnxt_pci_id_map[] = {
+#define RTE_PCI_DEV_ID_DECL_BNXT(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
+#include "rte_pci_dev_ids.h"
+	{.device_id = 0},
+};
+
+/*
+ * Initialization
+ */
+
+static int
+bnxt_dev_init(struct rte_eth_dev *eth_dev)
+{
+	static int version_printed;
+	int rc;
+
+	if (version_printed++ == 0)
+		RTE_LOG(INFO, PMD, "%s", bnxt_version);
+
+	if (eth_dev->pci_dev->addr.function >= 2 &&
+			eth_dev->pci_dev->addr.function < 4) {
+		RTE_LOG(ERR, PMD, "Function not enabled %x:\n",
+			eth_dev->pci_dev->addr.function);
+		rc = -ENOMEM;
+		goto error;
+	}
+
+	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
+	rc = -EPERM;
+
+error:
+	return rc;
+}
+
+static int
+bnxt_dev_uninit(struct rte_eth_dev *eth_dev __rte_unused) {
+	return 0;
+}
+
+static struct eth_driver bnxt_rte_pmd = {
+	.pci_drv = {
+		    .name = "rte_" DRV_MODULE_NAME "_pmd",
+		    .id_table = bnxt_pci_id_map,
+		    .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
+		    },
+	.eth_dev_init = bnxt_dev_init,
+	.eth_dev_uninit = bnxt_dev_uninit,
+	.dev_private_size = 32 /* this must be non-zero apparently */,
+};
+
+static int bnxt_rte_pmd_init(const char *name, const char *params __rte_unused)
+{
+	RTE_LOG(INFO, PMD, "bnxt_rte_pmd_init() called for %s\n", name);
+	rte_eth_driver_register(&bnxt_rte_pmd);
+	return 0;
+}
+
+static struct rte_driver bnxt_pmd_drv = {
+	.name = "eth_bnxt",
+	.type = PMD_PDEV,
+	.init = bnxt_rte_pmd_init,
+};
+
+PMD_REGISTER_DRIVER(bnxt_pmd_drv);
diff --git a/drivers/net/bnxt/rte_pmd_bnxt_version.map b/drivers/net/bnxt/rte_pmd_bnxt_version.map
new file mode 100644
index 0000000..349c6e1
--- /dev/null
+++ b/drivers/net/bnxt/rte_pmd_bnxt_version.map
@@ -0,0 +1,4 @@
+DPDK_16.04 {
+
+	local: *;
+};
diff --git a/lib/librte_eal/common/include/rte_pci_dev_ids.h b/lib/librte_eal/common/include/rte_pci_dev_ids.h
index cf7b548..f63cdc0 100644
--- a/lib/librte_eal/common/include/rte_pci_dev_ids.h
+++ b/lib/librte_eal/common/include/rte_pci_dev_ids.h
@@ -63,11 +63,12 @@
  * This file contains a list of the PCI device IDs recognised by DPDK, which
  * can be used to fill out an array of structures describing the devices.
  *
- * Currently four families of devices are recognised: those supported by the
- * IGB driver, by EM driver, those supported by the IXGBE driver, and by virtio
- * driver which is a para virtualization driver running in guest virtual machine.
- * The inclusion of these in an array built using this file depends on the
- * definition of
+ * Currently five families of devices are recognised: those supported by the
+ * IGB driver, by EM driver, those supported by the IXGBE driver, those
+ * supported by the BNXT driver, and by virtio driver which is a para
+ * virtualization driver running in guest virtual machine. The inclusion of
+ * these in an array built using this file depends on the definition of
+ * RTE_PCI_DEV_ID_DECL_BNXT
  * RTE_PCI_DEV_ID_DECL_EM
  * RTE_PCI_DEV_ID_DECL_IGB
  * RTE_PCI_DEV_ID_DECL_IGBVF
@@ -152,6 +153,10 @@
 #define RTE_PCI_DEV_ID_DECL_BNX2XVF(vend, dev)
 #endif
 
+#ifndef RTE_PCI_DEV_ID_DECL_BNXT
+#define RTE_PCI_DEV_ID_DECL_BNXT(vend, dev)
+#endif
+
 #ifndef PCI_VENDOR_ID_INTEL
 /** Vendor ID used by Intel devices */
 #define PCI_VENDOR_ID_INTEL 0x8086
@@ -686,6 +691,28 @@ RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57811_MF)
 RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57840_MF)
 #endif
 
+/****************** Broadcom bnxt devices ******************/
+
+#define BROADCOM_DEV_ID_57301                  0x16c8
+#define BROADCOM_DEV_ID_57302                  0x16c9
+#define BROADCOM_DEV_ID_57304_PF               0x16ca
+#define BROADCOM_DEV_ID_57304_VF               0x16cb
+#define BROADCOM_DEV_ID_57402                  0x16d0
+#define BROADCOM_DEV_ID_57404                  0x16d1
+#define BROADCOM_DEV_ID_57406_PF               0x16d2
+#define BROADCOM_DEV_ID_57406_VF               0x16d3
+#define BROADCOM_DEV_ID_57406_MF               0x16d4
+
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57301)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF)
+
 /*
  * Undef all RTE_PCI_DEV_ID_DECL_* here.
  */
@@ -702,3 +729,4 @@ RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57840_MF)
 #undef RTE_PCI_DEV_ID_DECL_VMXNET3
 #undef RTE_PCI_DEV_ID_DECL_FM10K
 #undef RTE_PCI_DEV_ID_DECL_FM10KVF
+#undef RTE_PCI_DEV_ID_DECL_BNXT
diff --git a/mk/rte.app.mk b/mk/rte.app.mk
index e9969fc..c62ba64 100644
--- a/mk/rte.app.mk
+++ b/mk/rte.app.mk
@@ -103,6 +103,7 @@ ifeq ($(CONFIG_RTE_BUILD_SHARED_LIB),n)
 
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_AF_PACKET)  += -lrte_pmd_af_packet
 _LDLIBS-$(CONFIG_RTE_LIBRTE_BNX2X_PMD)      += -lrte_pmd_bnx2x -lz
+_LDLIBS-$(CONFIG_RTE_LIBRTE_BNXT_PMD)       += -lrte_pmd_bnxt
 _LDLIBS-$(CONFIG_RTE_LIBRTE_CXGBE_PMD)      += -lrte_pmd_cxgbe
 _LDLIBS-$(CONFIG_RTE_LIBRTE_E1000_PMD)      += -lrte_pmd_e1000
 _LDLIBS-$(CONFIG_RTE_LIBRTE_ENA_PMD)        += -lrte_pmd_ena
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-15 16:25     ` Ferruh Yigit
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 03/38] bnxt: add driver register/unregister support Stephen Hurd
                     ` (38 subsequent siblings)
  39 siblings, 1 reply; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Start adding support to use the HWRM API.
Hardware Resource Manager or HWRM in short, is a set of API provided
by the firmware running in the ASIC to manage the various resources.

Initial commit just performs necessary HWRM queries for init, then
fails as before.

Now that struct bnxt is non-zero size, we can set dev_private_size
correctly.

The used HWRM calls so far:
bnxt_hwrm_func_qcaps:
	This command returns capabilities of a function.

bnxt_hwrm_ver_get:
	This function is called by a driver to determine the HWRM
	interface version supported by the HWRM firmware, the
	version of HWRM firmware implementation, the name of HWRM
	firmware, the versions of other embedded firmwares, and
	the names of other embedded firmwares, etc.  Gets the
	firmware version and interface specifications.  Returns
	an error if the firmware on the device is not supported
	by the driver and ensures the response space is large
	enough for the largest possible response.

bnxt_hwrm_queue_qportcfg:
	This function is called by a driver to query queue
	configuration of a port.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix few issues highlighted by checkpatch.

v5:
Fix footer formatting
Switch to using the HWRM function descriptions from the source
documentation
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                | 114 ++++
 drivers/net/bnxt/bnxt_ethdev.c         | 111 +++-
 drivers/net/bnxt/bnxt_hwrm.c           | 324 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  53 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 954 +++++++++++++++++++++++++++++++++
 6 files changed, 1555 insertions(+), 2 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt.h
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.c
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.h
 create mode 100644 drivers/net/bnxt/hsi_struct_def_dpdk.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index f6333fd..9965597 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -49,6 +49,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 # all source are stored in SRCS-y
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 
 #
 # Export include files
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
new file mode 100644
index 0000000..8cb7f5b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt.h
@@ -0,0 +1,114 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_H_
+#define _BNXT_H_
+
+#include <inttypes.h>
+#include <sys/queue.h>
+
+#include <rte_ethdev.h>
+#include <rte_memory.h>
+#include <rte_lcore.h>
+#include <rte_spinlock.h>
+
+struct bnxt_vf_info {
+	uint16_t		fw_fid;
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+	uint16_t		max_rsscos_ctx;
+	uint16_t		max_cp_rings;
+	uint16_t		max_tx_rings;
+	uint16_t		max_rx_rings;
+	uint16_t		max_l2_ctx;
+	uint16_t		max_vnics;
+	struct bnxt_pf_info	*pf;
+};
+
+struct bnxt_pf_info {
+#define BNXT_FIRST_PF_FID	1
+#define BNXT_MAX_VFS(bp)	(bp->pf.max_vfs)
+#define BNXT_FIRST_VF_FID	128
+#define BNXT_PF_RINGS_USED(bp)	bnxt_get_num_queues(bp)
+#define BNXT_PF_RINGS_AVAIL(bp)	(bp->pf.max_cp_rings - BNXT_PF_RINGS_USED(bp))
+	uint32_t		fw_fid;
+	uint8_t			port_id;
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+	uint16_t		max_rsscos_ctx;
+	uint16_t		max_cp_rings;
+	uint16_t		max_tx_rings;
+	uint16_t		max_rx_rings;
+	uint16_t		max_l2_ctx;
+	uint16_t		max_vnics;
+	uint16_t		first_vf_id;
+	uint16_t		active_vfs;
+	uint16_t		max_vfs;
+	void			*vf_req_buf;
+	phys_addr_t		vf_req_buf_dma_addr;
+	uint32_t		vf_req_fwd[8];
+	struct bnxt_vf_info	*vf;
+};
+
+#define BNXT_COS_QUEUE_COUNT	8
+struct bnxt_cos_queue_info {
+	uint8_t	id;
+	uint8_t	profile;
+};
+
+struct bnxt {
+	void				*bar0;
+
+	struct rte_eth_dev		*eth_dev;
+	struct rte_pci_device		*pdev;
+
+	uint32_t		flags;
+#define BNXT_FLAG_VF		(1 << 1)
+#define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
+#define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
+
+#define MAX_NUM_MAC_ADDR	32
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+
+	uint16_t			hwrm_cmd_seq;
+	void				*hwrm_cmd_resp_addr;
+	phys_addr_t			hwrm_cmd_resp_dma_addr;
+	rte_spinlock_t			hwrm_lock;
+	uint16_t			max_req_len;
+	uint16_t			max_resp_len;
+
+	struct bnxt_cos_queue_info	cos_queue[BNXT_COS_QUEUE_COUNT];
+
+	struct bnxt_pf_info		pf;
+	struct bnxt_vf_info		vf;
+};
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index af8a65f..32ed081 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -39,6 +39,9 @@
 #include <rte_malloc.h>
 #include <rte_cycles.h>
 
+#include "bnxt.h"
+#include "bnxt_hwrm.h"
+
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
 	"Broadcom Cumulus driver " DRV_MODULE_NAME "\n";
@@ -49,14 +52,68 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
+static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	rte_free(eth_dev->data->mac_addrs);
+	bnxt_free_hwrm_resources(bp);
+}
+
 /*
  * Initialization
  */
 
+static struct eth_dev_ops bnxt_dev_ops = {
+	.dev_close = bnxt_dev_close_op,
+};
+
+static bool bnxt_vf_pciid(uint16_t id)
+{
+	if (id == BROADCOM_DEV_ID_57304_VF ||
+	    id == BROADCOM_DEV_ID_57406_VF)
+		return true;
+	return false;
+}
+
+static int bnxt_init_board(struct rte_eth_dev *eth_dev)
+{
+	int rc;
+	struct bnxt *bp = eth_dev->data->dev_private;
+
+	/* enable device (incl. PCI PM wakeup), and bus-mastering */
+	if (!eth_dev->pci_dev->mem_resource[0].addr) {
+		RTE_LOG(ERR, PMD,
+			"Cannot find PCI device base address, aborting\n");
+		rc = -ENODEV;
+		goto init_err_disable;
+	}
+
+	bp->eth_dev = eth_dev;
+	bp->pdev = eth_dev->pci_dev;
+
+	bp->bar0 = (void *)eth_dev->pci_dev->mem_resource[0].addr;
+	if (!bp->bar0) {
+		RTE_LOG(ERR, PMD, "Cannot map device registers, aborting\n");
+		rc = -ENOMEM;
+		goto init_err_release;
+	}
+	return 0;
+
+init_err_release:
+	if (bp->bar0)
+		bp->bar0 = NULL;
+
+init_err_disable:
+
+	return rc;
+}
+
 static int
 bnxt_dev_init(struct rte_eth_dev *eth_dev)
 {
 	static int version_printed;
+	struct bnxt *bp;
 	int rc;
 
 	if (version_printed++ == 0)
@@ -71,8 +128,58 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	}
 
 	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
-	rc = -EPERM;
+	bp = eth_dev->data->dev_private;
+
+	if (bnxt_vf_pciid(eth_dev->pci_dev->id.device_id))
+		bp->flags |= BNXT_FLAG_VF;
+
+	rc = bnxt_init_board(eth_dev);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Board initialization failed rc: %x\n", rc);
+		goto error;
+	}
+	eth_dev->dev_ops = &bnxt_dev_ops;
+	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
+	/* eth_dev->tx_pkt_burst = &bnxt_xmit_pkts; */
+
+	rc = bnxt_alloc_hwrm_resources(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"hwrm resource allocation failure rc: %x\n", rc);
+		goto error_free;
+	}
+	rc = bnxt_hwrm_ver_get(bp);
+	if (rc)
+		goto error_free;
+	bnxt_hwrm_queue_qportcfg(bp);
+
+	/* Get the MAX capabilities for this function */
+	rc = bnxt_hwrm_func_qcaps(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "hwrm query capability failure rc: %x\n", rc);
+		goto error_free;
+	}
+	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
+					ETHER_ADDR_LEN * MAX_NUM_MAC_ADDR, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		RTE_LOG(ERR, PMD,
+			"Failed to alloc %u bytes needed to store MAC addr tbl",
+			ETHER_ADDR_LEN * MAX_NUM_MAC_ADDR);
+		rc = -ENOMEM;
+		goto error_free;
+	}
+	/* Copy the permanent MAC from the qcap response address now. */
+	if (BNXT_PF(bp))
+		memcpy(bp->mac_addr, bp->pf.mac_addr, sizeof(bp->mac_addr));
+	else
+		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
+	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
+
+	return -EPERM;
 
+error_free:
+	bnxt_dev_close_op(eth_dev);
 error:
 	return rc;
 }
@@ -90,7 +197,7 @@ static struct eth_driver bnxt_rte_pmd = {
 		    },
 	.eth_dev_init = bnxt_dev_init,
 	.eth_dev_uninit = bnxt_dev_uninit,
-	.dev_private_size = 32 /* this must be non-zero apparently */,
+	.dev_private_size = sizeof(struct bnxt),
 };
 
 static int bnxt_rte_pmd_init(const char *name, const char *params __rte_unused)
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
new file mode 100644
index 0000000..e187121
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -0,0 +1,324 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_byteorder.h>
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+
+#include "bnxt.h"
+#include "bnxt_hwrm.h"
+#include "hsi_struct_def_dpdk.h"
+
+#define HWRM_CMD_TIMEOUT		2000
+
+/*
+ * HWRM Functions (sent to HWRM)
+ * These are named bnxt_hwrm_*() and return -1 if bnxt_hwrm_send_message()
+ * fails (ie: a timeout), and a positive non-zero HWRM error code if the HWRM
+ * command was failed by the ChiMP.
+ */
+
+static int bnxt_hwrm_send_message_locked(struct bnxt *bp, void *msg,
+					uint32_t msg_len)
+{
+	unsigned int i;
+	struct input *req = msg;
+	struct output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t *data = msg;
+	uint8_t *bar;
+	uint8_t *valid;
+
+	/* Write request msg to hwrm channel */
+	for (i = 0; i < msg_len; i += 4) {
+		bar = (uint8_t *)bp->bar0 + i;
+		*(volatile uint32_t *)bar = *data;
+		data++;
+	}
+
+	/* Zero the rest of the request space */
+	for (; i < bp->max_req_len; i += 4) {
+		bar = (uint8_t *)bp->bar0 + i;
+		*(volatile uint32_t *)bar = 0;
+	}
+
+	/* Ring channel doorbell */
+	bar = (uint8_t *)bp->bar0 + 0x100;
+	*(volatile uint32_t *)bar = 1;
+
+	/* Poll for the valid bit */
+	for (i = 0; i < HWRM_CMD_TIMEOUT; i++) {
+		/* Sanity check on the resp->resp_len */
+		rte_rmb();
+		if (resp->resp_len && resp->resp_len <=
+		    bp->max_resp_len) {
+			/* Last byte of resp contains the valid key */
+			valid = (uint8_t *)resp + resp->resp_len - 1;
+			if (*valid == HWRM_RESP_VALID_KEY)
+				break;
+		}
+		rte_delay_us(600);
+	}
+
+	if (i >= HWRM_CMD_TIMEOUT) {
+		RTE_LOG(ERR, PMD, "Error sending msg %x\n",
+			req->req_type);
+		goto err_ret;
+	}
+	return 0;
+
+err_ret:
+	return -1;
+}
+
+static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
+{
+	int rc;
+
+	rte_spinlock_lock(&bp->hwrm_lock);
+	rc = bnxt_hwrm_send_message_locked(bp, msg, msg_len);
+	rte_spinlock_unlock(&bp->hwrm_lock);
+	return rc;
+}
+
+#define HWRM_PREP(req, type, cr, resp) \
+	memset(bp->hwrm_cmd_resp_addr, 0, bp->max_resp_len); \
+	req.req_type = rte_cpu_to_le_16(HWRM_##type); \
+	req.cmpl_ring = rte_cpu_to_le_16(cr); \
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++); \
+	req.target_id = rte_cpu_to_le_16(0xffff); \
+	req.resp_addr = rte_cpu_to_le_64(bp->hwrm_cmd_resp_dma_addr)
+
+#define HWRM_CHECK_RESULT \
+	{ \
+		if (rc) { \
+			RTE_LOG(ERR, PMD, "%s failed rc:%d\n", \
+				__func__, rc); \
+			return rc; \
+		} \
+		if (resp->error_code) { \
+			rc = rte_le_to_cpu_16(resp->error_code); \
+			RTE_LOG(ERR, PMD, "%s error %d\n", __func__, rc); \
+			return rc; \
+		} \
+	}
+
+int bnxt_hwrm_func_qcaps(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_func_qcaps_input req = {.req_type = 0 };
+	struct hwrm_func_qcaps_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, FUNC_QCAPS, -1, resp);
+
+	req.fid = rte_cpu_to_le_16(0xffff);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		pf->fw_fid = rte_le_to_cpu_32(resp->fid);
+		pf->port_id = resp->port_id;
+		memcpy(pf->mac_addr, resp->perm_mac_address, ETHER_ADDR_LEN);
+		pf->max_rsscos_ctx = rte_le_to_cpu_16(resp->max_rsscos_ctx);
+		pf->max_cp_rings = rte_le_to_cpu_16(resp->max_cmpl_rings);
+		pf->max_tx_rings = rte_le_to_cpu_16(resp->max_tx_rings);
+		pf->max_rx_rings = rte_le_to_cpu_16(resp->max_rx_rings);
+		pf->max_l2_ctx = rte_le_to_cpu_16(resp->max_l2_ctxs);
+		pf->max_vnics = rte_le_to_cpu_16(resp->max_vnics);
+		pf->first_vf_id = rte_le_to_cpu_16(resp->first_vf_id);
+		pf->max_vfs = rte_le_to_cpu_16(resp->max_vfs);
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		vf->fw_fid = rte_le_to_cpu_32(resp->fid);
+		memcpy(vf->mac_addr, &resp->perm_mac_address, ETHER_ADDR_LEN);
+		vf->max_rsscos_ctx = rte_le_to_cpu_16(resp->max_rsscos_ctx);
+		vf->max_cp_rings = rte_le_to_cpu_16(resp->max_cmpl_rings);
+		vf->max_tx_rings = rte_le_to_cpu_16(resp->max_tx_rings);
+		vf->max_rx_rings = rte_le_to_cpu_16(resp->max_rx_rings);
+		vf->max_l2_ctx = rte_le_to_cpu_16(resp->max_l2_ctxs);
+		vf->max_vnics = rte_le_to_cpu_16(resp->max_vnics);
+	}
+
+	return rc;
+}
+
+int bnxt_hwrm_ver_get(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_ver_get_input req = {.req_type = 0 };
+	struct hwrm_ver_get_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t my_version;
+	uint32_t fw_version;
+	uint16_t max_resp_len;
+	char type[RTE_MEMZONE_NAMESIZE];
+
+	HWRM_PREP(req, VER_GET, -1, resp);
+
+	req.hwrm_intf_maj = HWRM_VERSION_MAJOR;
+	req.hwrm_intf_min = HWRM_VERSION_MINOR;
+	req.hwrm_intf_upd = HWRM_VERSION_UPDATE;
+
+	/*
+	 * Hold the lock since we may be adjusting the response pointers.
+	 */
+	rte_spinlock_lock(&bp->hwrm_lock);
+	rc = bnxt_hwrm_send_message_locked(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	RTE_LOG(INFO, PMD, "%d.%d.%d:%d.%d.%d\n",
+		resp->hwrm_intf_maj, resp->hwrm_intf_min,
+		resp->hwrm_intf_upd,
+		resp->hwrm_fw_maj, resp->hwrm_fw_min, resp->hwrm_fw_bld);
+
+	my_version = HWRM_VERSION_MAJOR << 16;
+	my_version |= HWRM_VERSION_MINOR << 8;
+	my_version |= HWRM_VERSION_UPDATE;
+
+	fw_version = resp->hwrm_intf_maj << 16;
+	fw_version |= resp->hwrm_intf_min << 8;
+	fw_version |= resp->hwrm_intf_upd;
+
+	if (resp->hwrm_intf_maj != HWRM_VERSION_MAJOR) {
+		RTE_LOG(ERR, PMD, "Unsupported firmware API version\n");
+		rc = -EINVAL;
+		goto error;
+	}
+
+	if (my_version != fw_version) {
+		RTE_LOG(INFO, PMD, "BNXT Driver/HWRM API mismatch.\n");
+		if (my_version < fw_version) {
+			RTE_LOG(INFO, PMD,
+				"Firmware API version is newer than driver.\n");
+			RTE_LOG(INFO, PMD,
+				"The driver may be missing features.\n");
+		} else {
+			RTE_LOG(INFO, PMD,
+				"Firmware API version is older than driver.\n");
+			RTE_LOG(INFO, PMD,
+				"Not all driver features may be functional.\n");
+		}
+	}
+
+	if (bp->max_req_len > resp->max_req_win_len) {
+		RTE_LOG(ERR, PMD, "Unsupported request length\n");
+		rc = -EINVAL;
+	}
+	bp->max_req_len = resp->max_req_win_len;
+	max_resp_len = resp->max_resp_len;
+	if (bp->max_resp_len != max_resp_len) {
+		sprintf(type, "bnxt_hwrm_%04x:%02x:%02x:%02x",
+			bp->pdev->addr.domain, bp->pdev->addr.bus,
+			bp->pdev->addr.devid, bp->pdev->addr.function);
+
+		rte_free(bp->hwrm_cmd_resp_addr);
+
+		bp->hwrm_cmd_resp_addr = rte_malloc(type, max_resp_len, 0);
+		if (bp->hwrm_cmd_resp_addr == NULL) {
+			rc = -ENOMEM;
+			goto error;
+		}
+		bp->hwrm_cmd_resp_dma_addr =
+			rte_malloc_virt2phy(bp->hwrm_cmd_resp_addr);
+		bp->max_resp_len = max_resp_len;
+	}
+
+error:
+	rte_spinlock_unlock(&bp->hwrm_lock);
+	return rc;
+}
+
+int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_queue_qportcfg_input req = {.req_type = 0 };
+	struct hwrm_queue_qportcfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, QUEUE_QPORTCFG, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+#define GET_QUEUE_INFO(x) \
+	bp->cos_queue[x].id = resp->queue_id##x; \
+	bp->cos_queue[x].profile = resp->queue_id##x##_service_profile
+
+	GET_QUEUE_INFO(0);
+	GET_QUEUE_INFO(1);
+	GET_QUEUE_INFO(2);
+	GET_QUEUE_INFO(3);
+	GET_QUEUE_INFO(4);
+	GET_QUEUE_INFO(5);
+	GET_QUEUE_INFO(6);
+	GET_QUEUE_INFO(7);
+
+	return rc;
+}
+
+/*
+ * HWRM utility functions
+ */
+
+void bnxt_free_hwrm_resources(struct bnxt *bp)
+{
+	/* Release memzone */
+	rte_free(bp->hwrm_cmd_resp_addr);
+	bp->hwrm_cmd_resp_addr = NULL;
+	bp->hwrm_cmd_resp_dma_addr = 0;
+}
+
+int bnxt_alloc_hwrm_resources(struct bnxt *bp)
+{
+	struct rte_pci_device *pdev = bp->pdev;
+	char type[RTE_MEMZONE_NAMESIZE];
+
+	sprintf(type, "bnxt_hwrm_%04x:%02x:%02x:%02x", pdev->addr.domain,
+		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+	bp->max_req_len = HWRM_MAX_REQ_LEN;
+	bp->max_resp_len = HWRM_MAX_RESP_LEN;
+	bp->hwrm_cmd_resp_addr = rte_malloc(type, bp->max_resp_len, 0);
+	if (bp->hwrm_cmd_resp_addr == NULL)
+		return -ENOMEM;
+	bp->hwrm_cmd_resp_dma_addr =
+		rte_malloc_virt2phy(bp->hwrm_cmd_resp_addr);
+	rte_spinlock_init(&bp->hwrm_lock);
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
new file mode 100644
index 0000000..e35e8c0
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -0,0 +1,53 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_HWRM_H_
+#define _BNXT_HWRM_H_
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include "bnxt.h"
+
+#define HWRM_SEQ_ID_INVALID -1U
+
+int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+
+int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
+
+int bnxt_hwrm_ver_get(struct bnxt *bp);
+
+void bnxt_free_hwrm_resources(struct bnxt *bp);
+int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
new file mode 100644
index 0000000..c0bdbda
--- /dev/null
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -0,0 +1,954 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _HSI_STRUCT_DEF_EXTERNAL_H_
+#define _HSI_STRUCT_DEF_EXTERNAL_H_
+
+/* HW Resource Manager Specification 1.2.0 */
+#define HWRM_VERSION_MAJOR	1
+#define HWRM_VERSION_MINOR	2
+#define HWRM_VERSION_UPDATE	0
+
+/*
+ * Following is the signature for HWRM message field that indicates not
+ * applicable (All F's). Need to cast it the size of the field if needed.
+ */
+#define HWRM_MAX_REQ_LEN	(128)  /* hwrm_func_buf_rgtr */
+#define HWRM_MAX_RESP_LEN	(176)  /* hwrm_func_qstats */
+#define HWRM_RESP_VALID_KEY	1 /* valid key for HWRM response */
+
+/*
+ * Request types
+ */
+#define HWRM_VER_GET			(UINT32_C(0x0))
+#define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
+#define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+
+/*
+ * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
+ * inside the chip. The HWRM is implemented in firmware, and runs on embedded
+ * processors inside the chip. This firmware is vital part of the chip's
+ * hardware. The chip can not be used by driver without it.
+ */
+
+/* Input (16 bytes) */
+struct input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+} __attribute__((packed));
+
+/* Output (8 bytes) */
+struct output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+} __attribute__((packed));
+
+/* hwrm_func_qcaps */
+/*
+ * Description: This command returns capabilities of a function. The input FID
+ * value is used to indicate what function is being queried. This allows a
+ * physical function driver to query virtual functions that are children of the
+ * physical function. The output FID value is needed to configure Rings and
+ * MSI-X vectors so their DMA operations appear correctly on the PCI bus.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_func_qcaps_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Function ID of the function that is being queried. 0xFF... (All Fs)
+	 * if the query is for the requesting function.
+	 */
+	uint16_t fid;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (80 bytes) */
+struct hwrm_func_qcaps_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * FID value. This value is used to identify operations on the PCI bus
+	 * as belonging to a particular PCI function.
+	 */
+	uint16_t fid;
+
+	/*
+	 * Port ID of port that this function is associated with. Valid only for
+	 * the PF. 0xFF... (All Fs) if this function is not associated with any
+	 * port. 0xFF... (All Fs) if this function is called from a VF.
+	 */
+	uint16_t port_id;
+
+	/* If 1, then Push mode is supported on this function. */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_PUSH_MODE_SUPPORTED   UINT32_C(0x1)
+	/*
+	 * If 1, then the global MSI-X auto-masking is enabled for the device.
+	 */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_GLOBAL_MSIX_AUTOMASKING \
+								UINT32_C(0x2)
+	/*
+	 * If 1, then the Precision Time Protocol (PTP) processing is supported
+	 * on this function. The HWRM should enable PTP on only a single
+	 * Physical Function (PF) per port.
+	 */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_PTP_SUPPORTED         UINT32_C(0x4)
+	uint32_t flags;
+
+	/*
+	 * This value is current MAC address configured for this function. A
+	 * value of 00-00-00-00-00-00 indicates no MAC address is currently
+	 * configured.
+	 */
+	uint8_t perm_mac_address[6];
+
+	/*
+	 * The maximum number of RSS/COS contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_rsscos_ctx;
+
+	/*
+	 * The maximum number of completion rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_cmpl_rings;
+
+	/*
+	 * The maximum number of transmit rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_tx_rings;
+
+	/*
+	 * The maximum number of receive rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_rx_rings;
+
+	/*
+	 * The maximum number of L2 contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_l2_ctxs;
+
+	/* The maximum number of VNICs that can be allocated to the function. */
+	uint16_t max_vnics;
+
+	/*
+	 * The identifier for the first VF enabled on a PF. This is valid only
+	 * on the PF with SR-IOV enabled. 0xFF... (All Fs) if this command is
+	 * called on a PF with SR-IOV disabled or on a VF.
+	 */
+	uint16_t first_vf_id;
+
+	/*
+	 * The maximum number of VFs that can be allocated to the function. This
+	 * is valid only on the PF with SR-IOV enabled. 0xFF... (All Fs) if this
+	 * command is called on a PF with SR-IOV disabled or on a VF.
+	 */
+	uint16_t max_vfs;
+
+	/*
+	 * The maximum number of statistic contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_stat_ctx;
+
+	/*
+	 * The maximum number of Encapsulation records that can be offloaded by
+	 * this function.
+	 */
+	uint32_t max_encap_records;
+
+	/*
+	 * The maximum number of decapsulation records that can be offloaded by
+	 * this function.
+	 */
+	uint32_t max_decap_records;
+
+	/*
+	 * The maximum number of Exact Match (EM) flows that can be offloaded by
+	 * this function on the TX side.
+	 */
+	uint32_t max_tx_em_flows;
+
+	/*
+	 * The maximum number of Wildcard Match (WM) flows that can be offloaded
+	 * by this function on the TX side.
+	 */
+	uint32_t max_tx_wm_flows;
+
+	/*
+	 * The maximum number of Exact Match (EM) flows that can be offloaded by
+	 * this function on the RX side.
+	 */
+	uint32_t max_rx_em_flows;
+
+	/*
+	 * The maximum number of Wildcard Match (WM) flows that can be offloaded
+	 * by this function on the RX side.
+	 */
+	uint32_t max_rx_wm_flows;
+
+	/*
+	 * The maximum number of multicast filters that can be supported by this
+	 * function on the RX side.
+	 */
+	uint32_t max_mcast_filters;
+
+	/*
+	 * The maximum value of flow_id that can be supported in completion
+	 * records.
+	 */
+	uint32_t max_flow_id;
+
+	/*
+	 * The maximum number of HW ring groups that can be supported on this
+	 * function.
+	 */
+	uint32_t max_hw_ring_grps;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ver_get */
+/*
+ * Description: This function is called by a driver to determine the HWRM
+ * interface version supported by the HWRM firmware, the version of HWRM
+ * firmware implementation, the name of HWRM firmware, the versions of other
+ * embedded firmwares, and the names of other embedded firmwares, etc. Any
+ * interface or firmware version with major = 0, minor = 0, and update = 0 shall
+ * be considered an invalid version.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ver_get_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This field represents the major version of HWRM interface
+	 * specification supported by the driver HWRM implementation. The
+	 * interface major version is intended to change only when non backward
+	 * compatible changes are made to the HWRM interface specification.
+	 */
+	uint8_t hwrm_intf_maj;
+
+	/*
+	 * This field represents the minor version of HWRM interface
+	 * specification supported by the driver HWRM implementation. A change
+	 * in interface minor version is used to reflect significant backward
+	 * compatible modification to HWRM interface specification. This can be
+	 * due to addition or removal of functionality. HWRM interface
+	 * specifications with the same major version but different minor
+	 * versions are compatible.
+	 */
+	uint8_t hwrm_intf_min;
+
+	/*
+	 * This field represents the update version of HWRM interface
+	 * specification supported by the driver HWRM implementation. The
+	 * interface update version is used to reflect minor changes or bug
+	 * fixes to a released HWRM interface specification.
+	 */
+	uint8_t hwrm_intf_upd;
+
+	uint8_t unused_0[5];
+} __attribute__((packed));
+
+/* Output (128 bytes) */
+struct hwrm_ver_get_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This field represents the major version of HWRM interface
+	 * specification supported by the HWRM implementation. The interface
+	 * major version is intended to change only when non backward compatible
+	 * changes are made to the HWRM interface specification. A HWRM
+	 * implementation that is compliant with this specification shall
+	 * provide value of 1 in this field.
+	 */
+	uint8_t hwrm_intf_maj;
+
+	/*
+	 * This field represents the minor version of HWRM interface
+	 * specification supported by the HWRM implementation. A change in
+	 * interface minor version is used to reflect significant backward
+	 * compatible modification to HWRM interface specification. This can be
+	 * due to addition or removal of functionality. HWRM interface
+	 * specifications with the same major version but different minor
+	 * versions are compatible. A HWRM implementation that is compliant with
+	 * this specification shall provide value of 0 in this field.
+	 */
+	uint8_t hwrm_intf_min;
+
+	/*
+	 * This field represents the update version of HWRM interface
+	 * specification supported by the HWRM implementation. The interface
+	 * update version is used to reflect minor changes or bug fixes to a
+	 * released HWRM interface specification. A HWRM implementation that is
+	 * compliant with this specification shall provide value of 1 in this
+	 * field.
+	 */
+	uint8_t hwrm_intf_upd;
+
+	uint8_t hwrm_intf_rsvd;
+
+	/*
+	 * This field represents the major version of HWRM firmware. A change in
+	 * firmware major version represents a major firmware release.
+	 */
+	uint8_t hwrm_fw_maj;
+
+	/*
+	 * This field represents the minor version of HWRM firmware. A change in
+	 * firmware minor version represents significant firmware functionality
+	 * changes.
+	 */
+	uint8_t hwrm_fw_min;
+
+	/*
+	 * This field represents the build version of HWRM firmware. A change in
+	 * firmware build version represents bug fixes to a released firmware.
+	 */
+	uint8_t hwrm_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version of the HWRM firmware.
+	 */
+	uint8_t hwrm_fw_rsvd;
+
+	/*
+	 * This field represents the major version of mgmt firmware. A change in
+	 * major version represents a major release.
+	 */
+	uint8_t mgmt_fw_maj;
+
+	/*
+	 * This field represents the minor version of mgmt firmware. A change in
+	 * minor version represents significant functionality changes.
+	 */
+	uint8_t mgmt_fw_min;
+
+	/*
+	 * This field represents the build version of mgmt firmware. A change in
+	 * update version represents bug fixes.
+	 */
+	uint8_t mgmt_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t mgmt_fw_rsvd;
+
+	/*
+	 * This field represents the major version of network control firmware.
+	 * A change in major version represents a major release.
+	 */
+	uint8_t netctrl_fw_maj;
+
+	/*
+	 * This field represents the minor version of network control firmware.
+	 * A change in minor version represents significant functionality
+	 * changes.
+	 */
+	uint8_t netctrl_fw_min;
+
+	/*
+	 * This field represents the build version of network control firmware.
+	 * A change in update version represents bug fixes.
+	 */
+	uint8_t netctrl_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t netctrl_fw_rsvd;
+
+	/*
+	 * This field is reserved for future use. The responder should set it to
+	 * 0. The requester should ignore this field.
+	 */
+	uint32_t reserved1;
+
+	/*
+	 * This field represents the major version of RoCE firmware. A change in
+	 * major version represents a major release.
+	 */
+	uint8_t roce_fw_maj;
+
+	/*
+	 * This field represents the minor version of RoCE firmware. A change in
+	 * minor version represents significant functionality changes.
+	 */
+	uint8_t roce_fw_min;
+
+	/*
+	 * This field represents the build version of RoCE firmware. A change in
+	 * update version represents bug fixes.
+	 */
+	uint8_t roce_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t roce_fw_rsvd;
+
+	/*
+	 * This field represents the name of HWRM FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char hwrm_fw_name[16];
+
+	/*
+	 * This field represents the name of mgmt FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char mgmt_fw_name[16];
+
+	/*
+	 * This field represents the name of network control firmware (ASCII
+	 * chars without NULL at the end).
+	 */
+	char netctrl_fw_name[16];
+
+	/*
+	 * This field is reserved for future use. The responder should set it to
+	 * 0. The requester should ignore this field.
+	 */
+	uint32_t reserved2[4];
+
+	/*
+	 * This field represents the name of RoCE FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char roce_fw_name[16];
+
+	/* This field returns the chip number. */
+	uint16_t chip_num;
+
+	/* This field returns the revision of chip. */
+	uint8_t chip_rev;
+
+	/* This field returns the chip metal number. */
+	uint8_t chip_metal;
+
+	/* This field returns the bond id of the chip. */
+	uint8_t chip_bond_id;
+
+	/*
+	 * This value indicates the type of platform used for chip
+	 * implementation.
+	 */
+	/* ASIC */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_ASIC \
+							(UINT32_C(0x0) << 0)
+	/* FPGA platform of the chip. */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_FPGA \
+							(UINT32_C(0x1) << 0)
+	/* Palladium platform of the chip. */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_PALLADIUM \
+							(UINT32_C(0x2) << 0)
+	uint8_t chip_platform_type;
+
+	/*
+	 * This field returns the maximum value of request window that is
+	 * supported by the HWRM. The request window is mapped into device
+	 * address space using MMIO.
+	 */
+	uint16_t max_req_win_len;
+
+	/*
+	 * This field returns the maximum value of response buffer in bytes. If
+	 * a request specifies the response buffer length that is greater than
+	 * this value, then the HWRM should fail it. The value of this field
+	 * shall be 4KB or more.
+	 */
+	uint16_t max_resp_len;
+
+	/*
+	 * This field returns the default request timeout value in milliseconds.
+	 */
+	uint16_t def_req_timeout;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_queue_qportcfg */
+/*
+ * Description: This function is called by a driver to query queue configuration
+ * of a port. # The HWRM shall at least advertise one queue with lossy service
+ * profile. # The driver shall use this command to query queue ids before
+ * configuring or using any queues. # If a service profile is not set for a
+ * queue, then the driver shall not use that queue without configuring a service
+ * profile for it. # If the driver is not allowed to configure service profiles,
+ * then the driver shall only use queues for which service profiles are pre-
+ * configured.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_queue_qportcfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Enumeration denoting the RX, TX type of the resource. This
+	 * enumeration is used for resources that are similar for both TX and RX
+	 * paths of the chip.
+	 */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH \
+							UINT32_C(0x1)
+		/* tx path */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_TX \
+							(UINT32_C(0x0) << 0)
+		/* rx path */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_RX \
+							(UINT32_C(0x1) << 0)
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_LAST \
+					HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_RX
+	uint32_t flags;
+
+	/*
+	 * Port ID of port for which the queue configuration is being queried.
+	 * This field is only required when sent by IPC.
+	 */
+	uint16_t port_id;
+
+	uint16_t unused_0;
+} __attribute__((packed));
+
+/* Output (32 bytes) */
+struct hwrm_queue_qportcfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* The maximum number of queues that can be configured. */
+	uint8_t max_configurable_queues;
+
+	/* The maximum number of lossless queues that can be configured. */
+	uint8_t max_configurable_lossless_queues;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed. If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure queues using
+	 * hwrm_queue_cfg.
+	 */
+	uint8_t queue_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure queue buffers
+	 * using hwrm_queue_buffers_cfg.
+	 */
+	uint8_t queue_buffers_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure PFC using
+	 * hwrm_queue_pfcenable_cfg.
+	 */
+	uint8_t queue_pfcenable_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure Priority to CoS
+	 * mapping using hwrm_queue_pri2cos_cfg.
+	 */
+	uint8_t queue_pri2cos_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure CoS Bandwidth
+	 * configuration using hwrm_queue_cos2bw_cfg.
+	 */
+	uint8_t queue_cos2bw_cfg_allowed;
+
+	/* ID of CoS Queue 0. FF - Invalid id */
+	uint8_t queue_id0;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id0_service_profile;
+
+	/* ID of CoS Queue 1. FF - Invalid id */
+	uint8_t queue_id1;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id1_service_profile;
+
+	/* ID of CoS Queue 2. FF - Invalid id */
+	uint8_t queue_id2;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id2_service_profile;
+
+	/* ID of CoS Queue 3. FF - Invalid id */
+	uint8_t queue_id3;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id3_service_profile;
+
+	/* ID of CoS Queue 4. FF - Invalid id */
+	uint8_t queue_id4;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id4_service_profile;
+
+	/* ID of CoS Queue 5. FF - Invalid id */
+	uint8_t queue_id5;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id5_service_profile;
+
+	/* ID of CoS Queue 6. FF - Invalid id */
+	uint8_t queue_id6_service_profile;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id6;
+
+	/* ID of CoS Queue 7. FF - Invalid id */
+	uint8_t queue_id7;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id7_service_profile;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 03/38] bnxt: add driver register/unregister support
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 04/38] bnxt: add dev infos get operation Stephen Hurd
                     ` (37 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Move init() cleanup into uninit() function
Fix .dev_private_size

New HWRM calls:
bnxt_hwrm_func_driver_register:
	This command is used by the function driver to register
	its information with the HWRM.

bnxt_hwrm_func_driver_unregister:
	This command is used by the function driver to unregister
	with the HWRM.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comment regarding removal of bnxt_dev_close_op

v5:
Reformat footer
Add HWRM function descriptions
---
 drivers/net/bnxt/bnxt.h                |   1 +
 drivers/net/bnxt/bnxt_ethdev.c         |  29 +++-
 drivers/net/bnxt/bnxt_hwrm.c           |  50 ++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 271 +++++++++++++++++++++++++++++++++
 5 files changed, 350 insertions(+), 4 deletions(-)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index 8cb7f5b..ed057ef 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -91,6 +91,7 @@ struct bnxt {
 	struct rte_pci_device		*pdev;
 
 	uint32_t		flags;
+#define BNXT_FLAG_REGISTERED	(1 << 0)
 #define BNXT_FLAG_VF		(1 << 1)
 #define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
 #define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 32ed081..26e6447 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -176,17 +176,38 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
 
-	return -EPERM;
+	rc = bnxt_hwrm_func_driver_register(bp, 0,
+					    bp->pf.vf_req_fwd);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Failed to register driver");
+		rc = -EBUSY;
+		goto error_free;
+	}
+
+	RTE_LOG(INFO, PMD,
+		DRV_MODULE_NAME " found at mem %" PRIx64 ", node addr %pM\n",
+		eth_dev->pci_dev->mem_resource[0].phys_addr,
+		eth_dev->pci_dev->mem_resource[0].addr);
+
+	return 0;
 
 error_free:
-	bnxt_dev_close_op(eth_dev);
+	eth_dev->driver->eth_dev_uninit(eth_dev);
 error:
 	return rc;
 }
 
 static int
-bnxt_dev_uninit(struct rte_eth_dev *eth_dev __rte_unused) {
-	return 0;
+bnxt_dev_uninit(struct rte_eth_dev *eth_dev) {
+	struct bnxt *bp = eth_dev->data->dev_private;
+	int rc;
+
+	if (eth_dev->data->mac_addrs)
+		rte_free(eth_dev->data->mac_addrs);
+	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
+	bnxt_free_hwrm_resources(bp);
+	return rc;
 }
 
 static struct eth_driver bnxt_rte_pmd = {
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index e187121..8aba8cd 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -36,6 +36,7 @@
 #include <rte_cycles.h>
 #include <rte_malloc.h>
 #include <rte_memzone.h>
+#include <rte_version.h>
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
@@ -178,6 +179,34 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
+				   uint32_t *vf_req_fwd)
+{
+	int rc;
+	struct hwrm_func_drv_rgtr_input req = {.req_type = 0 };
+	struct hwrm_func_drv_rgtr_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (bp->flags & BNXT_FLAG_REGISTERED)
+		return 0;
+
+	HWRM_PREP(req, FUNC_DRV_RGTR, -1, resp);
+	req.flags = flags;
+	req.enables = HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VER;
+	req.ver_maj = RTE_VER_YEAR;
+	req.ver_min = RTE_VER_MONTH;
+	req.ver_upd = RTE_VER_MINOR;
+
+	memcpy(req.vf_req_fwd, vf_req_fwd, sizeof(req.vf_req_fwd));
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->flags |= BNXT_FLAG_REGISTERED;
+
+	return rc;
+}
+
 int bnxt_hwrm_ver_get(struct bnxt *bp)
 {
 	int rc = 0;
@@ -264,6 +293,27 @@ error:
 	return rc;
 }
 
+int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags)
+{
+	int rc;
+	struct hwrm_func_drv_unrgtr_input req = {.req_type = 0 };
+	struct hwrm_func_drv_unrgtr_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (!(bp->flags & BNXT_FLAG_REGISTERED))
+		return 0;
+
+	HWRM_PREP(req, FUNC_DRV_UNRGTR, -1, resp);
+	req.flags = flags;
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->flags &= ~BNXT_FLAG_REGISTERED;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index e35e8c0..6f2e445 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,7 +41,10 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
+				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index c0bdbda..8d816f4 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -52,6 +52,8 @@
  */
 #define HWRM_VER_GET			(UINT32_C(0x0))
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
+#define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
+#define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 
 /*
@@ -951,4 +953,273 @@ struct hwrm_queue_qportcfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_func_drv_rgtr */
+/*
+ * Description: This command is used by the function driver to register its
+ * information with the HWRM. A function driver shall implement this command. A
+ * function driver shall use this command during the driver initialization right
+ * after the HWRM version discovery and default ring resources allocation.
+ */
+
+/* Input (80 bytes) */
+struct hwrm_func_drv_rgtr_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the function driver is requesting all requests
+	 * from its children VF drivers to be forwarded to itself. This flag can
+	 * only be set by the PF driver. If a VF driver sets this flag, it
+	 * should be ignored by the HWRM.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_FLAGS_FWD_ALL_MODE        UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the function is requesting none of the requests
+	 * from its children VF drivers to be forwarded to itself. This flag can
+	 * only be set by the PF driver. If a VF driver sets this flag, it
+	 * should be ignored by the HWRM.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_FLAGS_FWD_NONE_MODE       UINT32_C(0x2)
+	uint32_t flags;
+
+	/* This bit must be '1' for the os_type field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_OS_TYPE           UINT32_C(0x1)
+	/* This bit must be '1' for the ver field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VER               UINT32_C(0x2)
+	/* This bit must be '1' for the timestamp field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_TIMESTAMP         UINT32_C(0x4)
+	/* This bit must be '1' for the vf_req_fwd field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VF_REQ_FWD        UINT32_C(0x8)
+	/*
+	 * This bit must be '1' for the async_event_fwd field to be configured.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_ASYNC_EVENT_FWD \
+								UINT32_C(0x10)
+	uint32_t enables;
+
+	/* This value indicates the type of OS. */
+		/* Unknown */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* Other OS not listed below. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_OTHER \
+							(UINT32_C(0x1) << 0)
+		/* MSDOS OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_MSDOS \
+							(UINT32_C(0xe) << 0)
+		/* Windows OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WINDOWS \
+							(UINT32_C(0x12) << 0)
+		/* Solaris OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_SOLARIS \
+							(UINT32_C(0x1d) << 0)
+		/* Linux OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_LINUX \
+							(UINT32_C(0x24) << 0)
+		/* FreeBSD OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_FREEBSD \
+							(UINT32_C(0x2a) << 0)
+		/* VMware ESXi OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_ESXI \
+							(UINT32_C(0x68) << 0)
+		/* Microsoft Windows 8 64-bit OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WIN864 \
+							(UINT32_C(0x73) << 0)
+		/* Microsoft Windows Server 2012 R2 OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WIN2012R2 \
+							(UINT32_C(0x74) << 0)
+	uint16_t os_type;
+
+	/* This is the major version of the driver. */
+	uint8_t ver_maj;
+
+	/* This is the minor version of the driver. */
+	uint8_t ver_min;
+
+	/* This is the update version of the driver. */
+	uint8_t ver_upd;
+
+	uint8_t unused_0;
+	uint16_t unused_1;
+
+	/*
+	 * This is a 32-bit timestamp provided by the driver for keep alive. The
+	 * timestamp is in multiples of 1ms.
+	 */
+	uint32_t timestamp;
+
+	uint32_t unused_2;
+
+	/*
+	 * This is a 256-bit bit mask provided by the PF driver for letting the
+	 * HWRM know what commands issued by the VF driver to the HWRM should be
+	 * forwarded to the PF driver. Nth bit refers to the Nth req_type.
+	 * Setting Nth bit to 1 indicates that requests from the VF driver with
+	 * req_type equal to N shall be forwarded to the parent PF driver. This
+	 * field is not valid for the VF driver.
+	 */
+	uint32_t vf_req_fwd[8];
+
+	/*
+	 * This is a 256-bit bit mask provided by the function driver (PF or VF
+	 * driver) to indicate the list of asynchronous event completions to be
+	 * forwarded. Nth bit refers to the Nth event_id. Setting Nth bit to 1
+	 * by the function driver shall result in the HWRM forwarding
+	 * asynchronous event completion with event_id equal to N. If all bits
+	 * are set to 0 (value of 0), then the HWRM shall not forward any
+	 * asynchronous event completion to this function driver.
+	 */
+	uint32_t async_event_fwd[8];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_func_drv_rgtr_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_func_drv_unrgtr */
+/*
+ * Description: This command is used by the function driver to un register with
+ * the HWRM. A function driver shall implement this command. A function driver
+ * shall use this command during the driver unloading.
+ */
+/* Input (24 bytes) */
+
+struct hwrm_func_drv_unrgtr_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the function driver is notifying the HWRM to
+	 * prepare for the shutdown.
+	 */
+	#define HWRM_FUNC_DRV_UNRGTR_INPUT_FLAGS_PREPARE_FOR_SHUTDOWN \
+							UINT32_C(0x1)
+	uint32_t flags;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_func_drv_unrgtr_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 #endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 04/38] bnxt: add dev infos get operation
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 03/38] bnxt: add driver register/unregister support Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 05/38] bnxt: add dev configure operation Stephen Hurd
                     ` (36 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Gets device info from the bp structure filled in the init() function.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder footer
---
 drivers/net/bnxt/bnxt.h        |  3 ++
 drivers/net/bnxt/bnxt_ethdev.c | 95 ++++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.c   |  2 +-
 3 files changed, 99 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index ed057ef..f8707b2 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -42,6 +42,9 @@
 #include <rte_lcore.h>
 #include <rte_spinlock.h>
 
+#define BNXT_MAX_MTU		9000
+#define VLAN_TAG_SIZE		4
+
 struct bnxt_vf_info {
 	uint16_t		fw_fid;
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 26e6447..a8a9912 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -61,10 +61,105 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 }
 
 /*
+ * Device configuration and status function
+ */
+
+static void bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
+				  struct rte_eth_dev_info *dev_info)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	uint16_t max_vnics, i, j, vpool, vrxq;
+
+	/* MAC Specifics */
+	dev_info->max_mac_addrs = MAX_NUM_MAC_ADDR;
+	dev_info->max_hash_mac_addrs = 0;
+
+	/* PF/VF specifics */
+	if (BNXT_PF(bp)) {
+		dev_info->max_rx_queues = bp->pf.max_rx_rings;
+		dev_info->max_tx_queues = bp->pf.max_tx_rings;
+		dev_info->max_vfs = bp->pf.active_vfs;
+		dev_info->reta_size = bp->pf.max_rsscos_ctx;
+		max_vnics = bp->pf.max_vnics;
+	} else {
+		dev_info->max_rx_queues = bp->vf.max_rx_rings;
+		dev_info->max_tx_queues = bp->vf.max_tx_rings;
+		dev_info->reta_size = bp->vf.max_rsscos_ctx;
+		max_vnics = bp->vf.max_vnics;
+	}
+
+	/* Fast path specifics */
+	dev_info->min_rx_bufsize = 1;
+	dev_info->max_rx_pktlen = BNXT_MAX_MTU + ETHER_HDR_LEN + ETHER_CRC_LEN
+				  + VLAN_TAG_SIZE;
+	dev_info->rx_offload_capa = 0;
+	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_IPV4_CKSUM |
+					DEV_TX_OFFLOAD_TCP_CKSUM |
+					DEV_TX_OFFLOAD_UDP_CKSUM |
+					DEV_TX_OFFLOAD_TCP_TSO;
+
+	/* *INDENT-OFF* */
+	dev_info->default_rxconf = (struct rte_eth_rxconf) {
+		.rx_thresh = {
+			.pthresh = 8,
+			.hthresh = 8,
+			.wthresh = 0,
+		},
+		.rx_free_thresh = 32,
+		.rx_drop_en = 0,
+	};
+
+	dev_info->default_txconf = (struct rte_eth_txconf) {
+		.tx_thresh = {
+			.pthresh = 32,
+			.hthresh = 0,
+			.wthresh = 0,
+		},
+		.tx_free_thresh = 32,
+		.tx_rs_thresh = 32,
+		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
+			     ETH_TXQ_FLAGS_NOOFFLOADS,
+	};
+	/* *INDENT-ON* */
+
+	/*
+	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
+	 *       need further investigation.
+	 */
+
+	/* VMDq resources */
+	vpool = 64; /* ETH_64_POOLS */
+	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
+	for (i = 0; i < 4; vpool >>= 1, i++) {
+		if (max_vnics > vpool) {
+			for (j = 0; j < 5; vrxq >>= 1, j++) {
+				if (dev_info->max_rx_queues > vrxq) {
+					if (vpool > vrxq)
+						vpool = vrxq;
+					goto found;
+				}
+			}
+			/* Not enough resources to support VMDq */
+			break;
+		}
+	}
+	/* Not enough resources to support VMDq */
+	vpool = 0;
+	vrxq = 0;
+found:
+	dev_info->max_vmdq_pools = vpool;
+	dev_info->vmdq_queue_num = vrxq;
+
+	dev_info->vmdq_pool_base = 0;
+	dev_info->vmdq_queue_base = 0;
+}
+
+/*
  * Initialization
  */
 
 static struct eth_dev_ops bnxt_dev_ops = {
+	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 };
 
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 8aba8cd..3eabf0c 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -83,7 +83,7 @@ static int bnxt_hwrm_send_message_locked(struct bnxt *bp, void *msg,
 		/* Sanity check on the resp->resp_len */
 		rte_rmb();
 		if (resp->resp_len && resp->resp_len <=
-		    bp->max_resp_len) {
+				bp->max_resp_len) {
 			/* Last byte of resp contains the valid key */
 			valid = (uint8_t *)resp + resp->resp_len - 1;
 			if (*valid == HWRM_RESP_VALID_KEY)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 05/38] bnxt: add dev configure operation
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (2 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 04/38] bnxt: add dev infos get operation Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 06/38] bnxt: add vnic functions and structs Stephen Hurd
                     ` (35 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

The dev_configure_op function calls bnxt_set_hwrm_link_config() to
setup the PHY.  This calls the new  bnxt_parse_eth_link_*() functions
to translate from the DPDK macro values to those used by HWRM calls,
then calls bnxt_hwrm_port_phy_cfg() to issue the HWRM call.

New HWRM calls:
bnxt_hwrm_port_phy_cfg:
	This command configures the PHY device for the port.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fixed few issues identified by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt.h                |  32 +++
 drivers/net/bnxt/bnxt_ethdev.c         |  24 ++
 drivers/net/bnxt/bnxt_hwrm.c           | 230 ++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 470 +++++++++++++++++++++++++++++++++
 5 files changed, 757 insertions(+)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index f8707b2..bfce91e 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -81,6 +81,29 @@ struct bnxt_pf_info {
 	struct bnxt_vf_info	*vf;
 };
 
+/* Max wait time is 10 * 100ms = 1s */
+#define BNXT_LINK_WAIT_CNT	10
+#define BNXT_LINK_WAIT_INTERVAL	100
+struct bnxt_link_info {
+	uint8_t			phy_flags;
+	uint8_t			mac_type;
+	uint8_t			phy_link_status;
+	uint8_t			loop_back;
+	uint8_t			link_up;
+	uint8_t			duplex;
+	uint8_t			pause;
+	uint8_t			force_pause;
+	uint8_t			auto_pause;
+	uint8_t			auto_mode;
+#define PHY_VER_LEN		3
+	uint8_t			phy_ver[PHY_VER_LEN];
+	uint16_t		link_speed;
+	uint16_t		support_speeds;
+	uint16_t		auto_link_speed;
+	uint16_t		auto_link_speed_mask;
+	uint32_t		preemphasis;
+};
+
 #define BNXT_COS_QUEUE_COUNT	8
 struct bnxt_cos_queue_info {
 	uint8_t	id;
@@ -99,6 +122,14 @@ struct bnxt {
 #define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
 #define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
 
+	unsigned int		rx_nr_rings;
+	unsigned int		rx_cp_nr_rings;
+	struct bnxt_rx_queue **rx_queues;
+
+	unsigned int		tx_nr_rings;
+	unsigned int		tx_cp_nr_rings;
+	struct bnxt_tx_queue **tx_queues;
+
 #define MAX_NUM_MAC_ADDR	32
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
 
@@ -109,6 +140,7 @@ struct bnxt {
 	uint16_t			max_req_len;
 	uint16_t			max_resp_len;
 
+	struct bnxt_link_info	link_info;
 	struct bnxt_cos_queue_info	cos_queue[BNXT_COS_QUEUE_COUNT];
 
 	struct bnxt_pf_info		pf;
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index a8a9912..b46d2ce 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -154,6 +154,29 @@ found:
 	dev_info->vmdq_queue_base = 0;
 }
 
+/* Configure the device based on the configuration provided */
+static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	int rc;
+
+	bp->rx_queues = (void *)eth_dev->data->rx_queues;
+	bp->tx_queues = (void *)eth_dev->data->tx_queues;
+
+	/* Inherit new configurations */
+	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
+	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
+	bp->rx_cp_nr_rings = bp->rx_nr_rings;
+	bp->tx_cp_nr_rings = bp->tx_nr_rings;
+
+	if (eth_dev->data->dev_conf.rxmode.jumbo_frame)
+		eth_dev->data->mtu =
+				eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
+				ETHER_HDR_LEN - ETHER_CRC_LEN - VLAN_TAG_SIZE;
+	rc = bnxt_set_hwrm_link_config(bp, true);
+	return rc;
+}
+
 /*
  * Initialization
  */
@@ -161,6 +184,7 @@ found:
 static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
+	.dev_configure = bnxt_dev_configure_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 3eabf0c..a2d7815 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -314,6 +314,61 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags)
 	return rc;
 }
 
+static int bnxt_hwrm_port_phy_cfg(struct bnxt *bp, struct bnxt_link_info *conf)
+{
+	int rc = 0;
+	struct hwrm_port_phy_cfg_input req = {.req_type = 0};
+	struct hwrm_port_phy_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, PORT_PHY_CFG, -1, resp);
+
+	req.flags = conf->phy_flags;
+	if (conf->link_up) {
+		req.force_link_speed = conf->link_speed;
+		/*
+		 * Note, ChiMP FW 20.2.1 and 20.2.2 return an error when we set
+		 * any auto mode, even "none".
+		 */
+		if (req.auto_mode == HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE) {
+			req.flags |= HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE;
+		} else {
+			req.auto_mode = conf->auto_mode;
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_MODE;
+			req.auto_link_speed_mask = conf->auto_link_speed_mask;
+			req.enables |=
+			   HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED_MASK;
+			req.auto_link_speed = conf->auto_link_speed;
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED;
+		}
+		req.auto_duplex = conf->duplex;
+		req.enables |= HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_DUPLEX;
+		req.auto_pause = conf->auto_pause;
+		/* Set force_pause if there is no auto or if there is a force */
+		if (req.auto_pause)
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_PAUSE;
+		else
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE;
+		req.force_pause = conf->force_pause;
+		if (req.force_pause)
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE;
+	} else {
+		req.flags &= ~HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG;
+		req.flags |= HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE_LINK_DOWN;
+		req.force_link_speed = 0;
+	}
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
@@ -372,3 +427,178 @@ int bnxt_alloc_hwrm_resources(struct bnxt *bp)
 
 	return 0;
 }
+
+static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
+{
+	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
+
+	if ((conf_link_speed & ETH_LINK_SPEED_FIXED) == ETH_LINK_SPEED_AUTONEG)
+		return HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
+
+	switch (conf_link_speed) {
+	case ETH_LINK_SPEED_10M_HD:
+	case ETH_LINK_SPEED_100M_HD:
+		return HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF;
+	}
+	return hw_link_duplex;
+}
+
+static uint16_t bnxt_parse_eth_link_speed(uint32_t conf_link_speed)
+{
+	uint16_t eth_link_speed = 0;
+
+	if ((conf_link_speed & ETH_LINK_SPEED_FIXED) == ETH_LINK_SPEED_AUTONEG)
+		return ETH_LINK_SPEED_AUTONEG;
+
+	switch (conf_link_speed & ~ETH_LINK_SPEED_FIXED) {
+	case ETH_LINK_SPEED_100M:
+	case ETH_LINK_SPEED_100M_HD:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MB;
+		break;
+	case ETH_LINK_SPEED_1G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB;
+		break;
+	case ETH_LINK_SPEED_2_5G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB;
+		break;
+	case ETH_LINK_SPEED_10G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB;
+		break;
+	case ETH_LINK_SPEED_20G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB;
+		break;
+	case ETH_LINK_SPEED_25G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB;
+		break;
+	case ETH_LINK_SPEED_40G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB;
+		break;
+	case ETH_LINK_SPEED_50G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB;
+		break;
+	default:
+		RTE_LOG(ERR, PMD,
+			"Unsupported link speed %d; default to AUTO\n",
+			conf_link_speed);
+		break;
+	}
+	return eth_link_speed;
+}
+
+#define BNXT_SUPPORTED_SPEEDS (ETH_LINK_SPEED_100M | ETH_LINK_SPEED_100M_HD | \
+		ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G | \
+		ETH_LINK_SPEED_10G | ETH_LINK_SPEED_20G | ETH_LINK_SPEED_25G | \
+		ETH_LINK_SPEED_40G | ETH_LINK_SPEED_50G)
+
+static int bnxt_valid_link_speed(uint32_t link_speed, uint8_t port_id)
+{
+	uint32_t one_speed;
+
+	if (link_speed == ETH_LINK_SPEED_AUTONEG)
+		return 0;
+
+	if (link_speed & ETH_LINK_SPEED_FIXED) {
+		one_speed = link_speed & ~ETH_LINK_SPEED_FIXED;
+
+		if (one_speed & (one_speed - 1)) {
+			RTE_LOG(ERR, PMD,
+				"Invalid advertised speeds (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+		if ((one_speed & BNXT_SUPPORTED_SPEEDS) != one_speed) {
+			RTE_LOG(ERR, PMD,
+				"Unsupported advertised speed (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+	} else {
+		if (!(link_speed & BNXT_SUPPORTED_SPEEDS)) {
+			RTE_LOG(ERR, PMD,
+				"Unsupported advertised speeds (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
+static uint16_t bnxt_parse_eth_link_speed_mask(uint32_t link_speed)
+{
+	uint16_t ret = 0;
+
+	if (link_speed == ETH_LINK_SPEED_AUTONEG)
+		link_speed = BNXT_SUPPORTED_SPEEDS;
+
+	if (link_speed & ETH_LINK_SPEED_100M)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB;
+	if (link_speed & ETH_LINK_SPEED_100M_HD)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB;
+	if (link_speed & ETH_LINK_SPEED_1G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB;
+	if (link_speed & ETH_LINK_SPEED_2_5G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB;
+	if (link_speed & ETH_LINK_SPEED_10G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB;
+	if (link_speed & ETH_LINK_SPEED_20G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB;
+	if (link_speed & ETH_LINK_SPEED_25G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB;
+	if (link_speed & ETH_LINK_SPEED_40G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB;
+	if (link_speed & ETH_LINK_SPEED_50G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB;
+	return ret;
+}
+
+int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up)
+{
+	int rc = 0;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_link_info link_req;
+	uint16_t speed;
+
+	rc = bnxt_valid_link_speed(dev_conf->link_speeds,
+			bp->eth_dev->data->port_id);
+	if (rc)
+		goto error;
+
+	memset(&link_req, 0, sizeof(link_req));
+	speed = bnxt_parse_eth_link_speed(dev_conf->link_speeds);
+	link_req.link_up = link_up;
+	if (speed == 0) {
+		link_req.phy_flags =
+				HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG;
+		link_req.auto_mode =
+				HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_OR_BELOW;
+		link_req.auto_link_speed_mask =
+			bnxt_parse_eth_link_speed_mask(dev_conf->link_speeds);
+		link_req.auto_link_speed =
+				HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_50GB;
+	} else {
+		link_req.auto_mode = HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE;
+		link_req.phy_flags = HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE |
+			HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESET_PHY;
+		link_req.link_speed = speed;
+	}
+	link_req.duplex = bnxt_parse_eth_link_duplex(dev_conf->link_speeds);
+	link_req.auto_pause = bp->link_info.auto_pause;
+	link_req.force_pause = bp->link_info.force_pause;
+
+	rc = bnxt_hwrm_port_phy_cfg(bp, &link_req);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Set link config failed with rc %d\n", rc);
+	}
+
+error:
+	return rc;
+}
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 6f2e445..eef3be6 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -52,5 +52,6 @@ int bnxt_hwrm_ver_get(struct bnxt *bp);
 
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
 
 #endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 8d816f4..cd2b14e 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -54,6 +54,7 @@
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
+#define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 
 /*
@@ -345,6 +346,475 @@ struct hwrm_func_qcaps_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_port_phy_cfg */
+/*
+ * Description: This command configures the PHY device for the port. It allows
+ * setting of the most generic settings for the PHY. The HWRM shall complete
+ * this command as soon as PHY settings are configured. They may not be applied
+ * when the command response is provided. A VF driver shall not be allowed to
+ * configure PHY using this command. In a network partition mode, a PF driver
+ * shall not be allowed to configure PHY using this command.
+ */
+
+/* Input (56 bytes) */
+struct hwrm_port_phy_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is set to '1', the PHY for the port shall be reset. #
+	 * If this bit is set to 1, then the HWRM shall reset the PHY after
+	 * applying PHY configuration changes specified in this command. # In
+	 * order to guarantee that PHY configuration changes specified in this
+	 * command take effect, the HWRM client should set this flag to 1. # If
+	 * this bit is not set to 1, then the HWRM may reset the PHY depending
+	 * on the current PHY configuration and settings specified in this
+	 * command.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESET_PHY            UINT32_C(0x1)
+	/*
+	 * When this bit is set to '1', the link shall be forced to be taken
+	 * down. # When this bit is set to '1", all other command input settings
+	 * related to the link speed shall be ignored. Once the link state is
+	 * forced down, it can be explicitly cleared from that state by setting
+	 * this flag to '0'. # If this flag is set to '0', then the link shall
+	 * be cleared from forced down state if the link is in forced down
+	 * state. There may be conditions (e.g. out-of-band or sideband
+	 * configuration changes for the link) outside the scope of the HWRM
+	 * implementation that may clear forced down link state.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE_LINK_DOWN      UINT32_C(0x2)
+	/*
+	 * When this bit is set to '1', the link shall be forced to the
+	 * force_link_speed value. When this bit is set to '1', the HWRM client
+	 * should not enable any of the auto negotiation related fields
+	 * represented by auto_XXX fields in this command. When this bit is set
+	 * to '1' and the HWRM client has enabled a auto_XXX field in this
+	 * command, then the HWRM shall ignore the enabled auto_XXX field. When
+	 * this bit is set to zero, the link shall be allowed to autoneg.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE                UINT32_C(0x4)
+	/*
+	 * When this bit is set to '1', the auto-negotiation process shall be
+	 * restarted on the link.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG      UINT32_C(0x8)
+	/*
+	 * When this bit is set to '1', Energy Efficient Ethernet (EEE) is
+	 * requested to be enabled on this link. If EEE is not supported on this
+	 * port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_ENABLE	UINT32_C(0x10)
+	/*
+	 * When this bit is set to '1', Energy Efficient Ethernet (EEE) is
+	 * requested to be disabled on this link. If EEE is not supported on
+	 * this port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_DISABLE	UINT32_C(0x20)
+	/*
+	 * When this bit is set to '1' and EEE is enabled on this link, then TX
+	 * LPI is requested to be enabled on the link. If EEE is not supported
+	 * on this port, then this flag shall be ignored by the HWRM. If EEE is
+	 * disabled on this port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_TX_LPI	UINT32_C(0x40)
+	uint32_t flags;
+
+	/* This bit must be '1' for the auto_mode field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_MODE          UINT32_C(0x1)
+	/* This bit must be '1' for the auto_duplex field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_DUPLEX        UINT32_C(0x2)
+	/* This bit must be '1' for the auto_pause field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_PAUSE         UINT32_C(0x4)
+	/*
+	 * This bit must be '1' for the auto_link_speed field to be configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED    UINT32_C(0x8)
+	/*
+	 * This bit must be '1' for the auto_link_speed_mask field to be
+	 * configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED_MASK \
+								UINT32_C(0x10)
+	/* This bit must be '1' for the wirespeed field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_WIRESPEED	UINT32_C(0x20)
+	/* This bit must be '1' for the lpbk field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_LPBK		UINT32_C(0x40)
+	/* This bit must be '1' for the preemphasis field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_PREEMPHASIS	UINT32_C(0x80)
+	/* This bit must be '1' for the force_pause field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE	UINT32_C(0x100)
+	/*
+	 * This bit must be '1' for the eee_link_speed_mask field to be
+	 * configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_EEE_LINK_SPEED_MASK \
+								UINT32_C(0x200)
+	/* This bit must be '1' for the tx_lpi_timer field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_TX_LPI_TIMER	UINT32_C(0x400)
+	uint32_t enables;
+
+	/* Port ID of port that is to be configured. */
+	uint16_t port_id;
+
+	/*
+	 * This is the speed that will be used if the force bit is '1'. If
+	 * unsupported speed is selected, an error will be generated.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t force_link_speed;
+
+	/*
+	 * This value is used to identify what autoneg mode is used when the
+	 * link speed is not being forced.
+	 */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE	(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ALL_SPEEDS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_SPEED_MASK \
+							(UINT32_C(0x4) << 0)
+	uint8_t auto_mode;
+
+	/*
+	 * This is the duplex setting that will be used if the autoneg_mode is
+	 * "one_speed" or "one_or_below".
+	 */
+		/* Half Duplex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF \
+							(UINT32_C(0x0) << 0)
+		/* Full duplex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_FULL \
+							(UINT32_C(0x1) << 0)
+		/* Both Half and Full dupex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH \
+							(UINT32_C(0x2) << 0)
+	uint8_t auto_duplex;
+
+	/*
+	 * This value is used to configure the pause that will be used for
+	 * autonegotiation. Add text on the usage of auto_pause and force_pause.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX              UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX              UINT32_C(0x2)
+	/*
+	 * When set to 1, the advertisement of pause is enabled. # When the
+	 * auto_mode is not set to none and this flag is set to 1, then the
+	 * auto_pause bits on this port are being advertised and autoneg pause
+	 * results are being interpreted. # When the auto_mode is not set to
+	 * none and this flag is set to 0, the pause is forced as indicated in
+	 * force_pause, and also advertised as auto_pause bits, but the autoneg
+	 * results are not interpreted since the pause configuration is being
+	 * forced. # When the auto_mode is set to none and this flag is set to
+	 * 1, auto_pause bits should be ignored and should be set to 0.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_AUTONEG_PAUSE   UINT32_C(0x4)
+	uint8_t auto_pause;
+
+	uint8_t unused_0;
+
+	/*
+	 * This is the speed that will be used if the autoneg_mode is
+	 * "one_speed" or "one_or_below". If an unsupported speed is selected,
+	 * an error will be generated.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t auto_link_speed;
+
+	/*
+	 * This is a mask of link speeds that will be used if autoneg_mode is
+	 * "mask". If unsupported speed is enabled an error will be generated.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MB \
+							UINT32_C(0x2000)
+	uint16_t auto_link_speed_mask;
+
+	/* This value controls the wirespeed feature. */
+		/* Wirespeed feature is disabled. */
+	#define HWRM_PORT_PHY_CFG_INPUT_WIRESPEED_OFF	(UINT32_C(0x0) << 0)
+		/* Wirespeed feature is enabled. */
+	#define HWRM_PORT_PHY_CFG_INPUT_WIRESPEED_ON	(UINT32_C(0x1) << 0)
+	uint8_t wirespeed;
+
+	/* This value controls the loopback setting for the PHY. */
+		/* No loopback is selected. Normal operation. */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_NONE	(UINT32_C(0x0) << 0)
+		/*
+		 * The HW will be configured with local loopback such that host
+		 * data is sent back to the host without modification.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_LOCAL	(UINT32_C(0x1) << 0)
+		/*
+		 * The HW will be configured with remote loopback such that port
+		 * logic will send packets back out the transmitter that are
+		 * received.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_REMOTE	(UINT32_C(0x2) << 0)
+	uint8_t lpbk;
+
+	/*
+	 * This value is used to configure the pause that will be used for force
+	 * mode.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX             UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX             UINT32_C(0x2)
+	uint8_t force_pause;
+
+	uint8_t unused_1;
+
+	/*
+	 * This value controls the pre-emphasis to be used for the link. Driver
+	 * should not set this value (use enable.preemphasis = 0) unless driver
+	 * is sure of setting. Normally HWRM FW will determine proper pre-
+	 * emphasis.
+	 */
+	uint32_t preemphasis;
+
+	/*
+	 * Setting for link speed mask that is used to advertise speeds during
+	 * autonegotiation when EEE is enabled. This field is valid only when
+	 * EEE is enabled. The speeds specified in this field shall be a subset
+	 * of speeds specified in auto_link_speed_mask. If EEE is enabled,then
+	 * at least one speed shall be provided in this mask.
+	 */
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD1  UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_100MB  UINT32_C(0x2)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD2  UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_1GB    UINT32_C(0x8)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD3 \
+								UINT32_C(0x10)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD4 \
+								UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_10GB \
+								UINT32_C(0x40)
+	uint16_t eee_link_speed_mask;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * Reuested setting of TX LPI timer in microseconds. This field is valid
+	 * only when EEE is enabled and TX LPI is enabled.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_TX_LPI_TIMER_MASK \
+							UINT32_C(0xffffff)
+	#define HWRM_PORT_PHY_CFG_INPUT_TX_LPI_TIMER_SFT           0
+	uint32_t tx_lpi_timer;
+
+	uint32_t unused_4;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_port_phy_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_ver_get */
 /*
  * Description: This function is called by a driver to determine the HWRM
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 06/38] bnxt: add vnic functions and structs
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (3 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 05/38] bnxt: add dev configure operation Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 07/38] bnxt: declare generic ring structs and free() func Stephen Hurd
                     ` (34 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add functions to allocate, initialize, and free vnics.

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |  14 ++
 drivers/net/bnxt/bnxt_vnic.c           | 277 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_vnic.h           |  80 ++++++++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h |   3 +
 5 files changed, 375 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_vnic.c
 create mode 100644 drivers/net/bnxt/bnxt_vnic.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 9965597..c57afaa 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
 # Export include files
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index bfce91e..d0f84f4 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -45,6 +45,13 @@
 #define BNXT_MAX_MTU		9000
 #define VLAN_TAG_SIZE		4
 
+enum bnxt_hw_context {
+	HW_CONTEXT_NONE     = 0,
+	HW_CONTEXT_IS_RSS   = 1,
+	HW_CONTEXT_IS_COS   = 2,
+	HW_CONTEXT_IS_LB    = 3,
+};
+
 struct bnxt_vf_info {
 	uint16_t		fw_fid;
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
@@ -130,6 +137,13 @@ struct bnxt {
 	unsigned int		tx_cp_nr_rings;
 	struct bnxt_tx_queue **tx_queues;
 
+	struct bnxt_vnic_info	*vnic_info;
+	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
+
+	/* VNIC pointer for flow filter (VMDq) pools */
+#define MAX_FF_POOLS	ETH_64_POOLS
+	STAILQ_HEAD(, bnxt_vnic_info)	ff_pool[MAX_FF_POOLS];
+
 #define MAX_NUM_MAC_ADDR	32
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
 
diff --git a/drivers/net/bnxt/bnxt_vnic.c b/drivers/net/bnxt/bnxt_vnic.c
new file mode 100644
index 0000000..c04c4c7
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_vnic.c
@@ -0,0 +1,277 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_memzone.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * VNIC Functions
+ */
+
+static void prandom_bytes(void *dest_ptr, size_t len)
+{
+	char *dest = (char *)dest_ptr;
+	uint64_t rb;
+
+	while (len) {
+		rb = rte_rand();
+		if (len >= 8) {
+			memcpy(dest, &rb, 8);
+			len -= 8;
+			dest += 8;
+		} else {
+			memcpy(dest, &rb, len);
+			dest += len;
+			len = 0;
+		}
+	}
+}
+
+void bnxt_init_vnics(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	uint16_t max_vnics;
+	int i, j;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	STAILQ_INIT(&bp->free_vnic_list);
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+		vnic->fw_vnic_id = (uint16_t)HWRM_NA_SIGNATURE;
+		vnic->fw_rss_cos_lb_ctx = (uint16_t)HWRM_NA_SIGNATURE;
+		vnic->ctx_is_rss_cos_lb = HW_CONTEXT_NONE;
+
+		for (j = 0; j < MAX_QUEUES_PER_VNIC; j++)
+			vnic->fw_grp_ids[j] = (uint16_t)HWRM_NA_SIGNATURE;
+
+		prandom_bytes(vnic->rss_hash_key, HW_HASH_KEY_SIZE);
+		STAILQ_INIT(&vnic->filter);
+		STAILQ_INSERT_TAIL(&bp->free_vnic_list, vnic, next);
+	}
+	for (i = 0; i < MAX_FF_POOLS; i++)
+		STAILQ_INIT(&bp->ff_pool[i]);
+}
+
+int bnxt_free_vnic(struct bnxt *bp, struct bnxt_vnic_info *vnic,
+			  int pool)
+{
+	struct bnxt_vnic_info *temp;
+
+	temp = STAILQ_FIRST(&bp->ff_pool[pool]);
+	while (temp) {
+		if (temp == vnic) {
+			STAILQ_REMOVE(&bp->ff_pool[pool], vnic,
+				      bnxt_vnic_info, next);
+			vnic->fw_vnic_id = (uint16_t)HWRM_NA_SIGNATURE;
+			STAILQ_INSERT_TAIL(&bp->free_vnic_list, vnic,
+					   next);
+			return 0;
+		}
+		temp = STAILQ_NEXT(temp, next);
+	}
+	RTE_LOG(ERR, PMD, "VNIC %p is not found in pool[%d]\n", vnic, pool);
+	return -EINVAL;
+}
+
+struct bnxt_vnic_info *bnxt_alloc_vnic(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+
+	/* Find the 1st unused vnic from the free_vnic_list pool*/
+	vnic = STAILQ_FIRST(&bp->free_vnic_list);
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "No more free VNIC resources\n");
+		return NULL;
+	}
+	STAILQ_REMOVE_HEAD(&bp->free_vnic_list, next);
+	return vnic;
+}
+
+void bnxt_free_all_vnics(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *temp, *next;
+	int i;
+
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		temp = STAILQ_FIRST(&bp->ff_pool[i]);
+		while (temp) {
+			next = STAILQ_NEXT(temp, next);
+			STAILQ_REMOVE(&bp->ff_pool[i], temp, bnxt_vnic_info,
+				      next);
+			STAILQ_INSERT_TAIL(&bp->free_vnic_list, temp, next);
+			temp = next;
+		}
+	}
+}
+
+void bnxt_free_vnic_attributes(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+
+	STAILQ_FOREACH(vnic, &bp->free_vnic_list, next) {
+		if (vnic->rss_table) {
+			/* 'Unreserve' the rss_table */
+			/* N/A */
+
+			vnic->rss_table = NULL;
+		}
+
+		if (vnic->rss_hash_key) {
+			/* 'Unreserve' the rss_hash_key */
+			/* N/A */
+
+			vnic->rss_hash_key = NULL;
+		}
+	}
+}
+
+int bnxt_alloc_vnic_attributes(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	struct rte_pci_device *pdev = bp->pdev;
+	const struct rte_memzone *mz;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+	int entry_length = RTE_CACHE_LINE_ROUNDUP(
+				HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table) +
+				HW_HASH_KEY_SIZE);
+	uint16_t max_vnics;
+	int i;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
+		 "bnxt_%04x:%02x:%02x:%02x_vnicattr", pdev->addr.domain,
+		 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
+	mz = rte_memzone_lookup(mz_name);
+	if (!mz) {
+		mz = rte_memzone_reserve(mz_name,
+					 entry_length * max_vnics,
+					 SOCKET_ID_ANY,
+					 RTE_MEMZONE_2MB |
+					 RTE_MEMZONE_SIZE_HINT_ONLY);
+		if (!mz)
+			return -ENOMEM;
+	}
+
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+
+		/* Allocate rss table and hash key */
+		vnic->rss_table =
+			(void *)((char *)mz->addr + (entry_length * i));
+		memset(vnic->rss_table, -1, entry_length);
+
+		vnic->rss_table_dma_addr = mz->phys_addr + (entry_length * i);
+		vnic->rss_hash_key = (void *)((char *)vnic->rss_table +
+			     HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table));
+
+		vnic->rss_hash_key_dma_addr = vnic->rss_table_dma_addr +
+			     HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table);
+	}
+
+	return 0;
+}
+
+void bnxt_free_vnic_mem(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	uint16_t max_vnics, i;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+		if (vnic->fw_vnic_id != (uint16_t)HWRM_NA_SIGNATURE) {
+			RTE_LOG(ERR, PMD, "VNIC is not freed yet!\n");
+			/* TODO Call HWRM to free VNIC */
+		}
+	}
+
+	rte_free(bp->vnic_info);
+	bp->vnic_info = NULL;
+}
+
+int bnxt_alloc_vnic_mem(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic_mem;
+	uint16_t max_vnics;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	/* Allocate memory for VNIC pool and filter pool */
+	vnic_mem = rte_zmalloc("bnxt_vnic_info",
+			       max_vnics * sizeof(struct bnxt_vnic_info), 0);
+	if (vnic_mem == NULL) {
+		RTE_LOG(ERR, PMD, "Failed to alloc memory for %d VNICs",
+			max_vnics);
+		return -ENOMEM;
+	}
+	bp->vnic_info = vnic_mem;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_vnic.h b/drivers/net/bnxt/bnxt_vnic.h
new file mode 100644
index 0000000..9671ba4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_vnic.h
@@ -0,0 +1,80 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_VNIC_H_
+#define _BNXT_VNIC_H_
+
+#include <sys/queue.h>
+#include <stdbool.h>
+
+struct bnxt_vnic_info {
+	STAILQ_ENTRY(bnxt_vnic_info)	next;
+	uint8_t		ff_pool_idx;
+
+	uint16_t	fw_vnic_id; /* returned by Chimp during alloc */
+	uint16_t	fw_rss_cos_lb_ctx;
+	uint16_t	ctx_is_rss_cos_lb;
+#define MAX_NUM_TRAFFIC_CLASSES		8
+#define MAX_NUM_RSS_QUEUES_PER_VNIC	16
+#define MAX_QUEUES_PER_VNIC	(MAX_NUM_RSS_QUEUES_PER_VNIC + \
+				 MAX_NUM_TRAFFIC_CLASSES)
+	uint16_t	start_grp_id;
+	uint16_t	end_grp_id;
+	uint16_t	fw_grp_ids[MAX_QUEUES_PER_VNIC];
+	uint16_t	hash_type;
+	phys_addr_t	rss_table_dma_addr;
+	uint16_t	*rss_table;
+	phys_addr_t	rss_hash_key_dma_addr;
+	void		*rss_hash_key;
+	uint32_t	flags;
+#define BNXT_VNIC_INFO_PROMISC			(1 << 0)
+#define BNXT_VNIC_INFO_ALLMULTI			(1 << 1)
+
+	bool		vlan_strip;
+	bool		func_default;
+
+	STAILQ_HEAD(, bnxt_filter_info)	filter;
+};
+
+struct bnxt;
+void bnxt_init_vnics(struct bnxt *bp);
+int bnxt_free_vnic(struct bnxt *bp, struct bnxt_vnic_info *vnic,
+			  int pool);
+struct bnxt_vnic_info *bnxt_alloc_vnic(struct bnxt *bp);
+void bnxt_free_all_vnics(struct bnxt *bp);
+void bnxt_free_vnic_attributes(struct bnxt *bp);
+int bnxt_alloc_vnic_attributes(struct bnxt *bp);
+void bnxt_free_vnic_mem(struct bnxt *bp);
+int bnxt_alloc_vnic_mem(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index cd2b14e..92e39ca 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -43,8 +43,11 @@
  * Following is the signature for HWRM message field that indicates not
  * applicable (All F's). Need to cast it the size of the field if needed.
  */
+#define HWRM_NA_SIGNATURE        ((uint32_t)(-1))
 #define HWRM_MAX_REQ_LEN	(128)  /* hwrm_func_buf_rgtr */
 #define HWRM_MAX_RESP_LEN	(176)  /* hwrm_func_qstats */
+#define HW_HASH_INDEX_SIZE      0x80    /* 7 bit indirection table index. */
+#define HW_HASH_KEY_SIZE        40
 #define HWRM_RESP_VALID_KEY	1 /* valid key for HWRM response */
 
 /*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 07/38] bnxt: declare generic ring structs and free() func
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (4 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 06/38] bnxt: add vnic functions and structs Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 08/38] bnxt: add completion ring support Stephen Hurd
                     ` (33 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Declare generic ring structures and a free() function. These are
generic ring mamagement functions which will be used to create Tx,
Rx and Completion rings in the subsequent patches, and tie them to
the HWRM managed ring resources.

This generic ring structure is shared all the ring types and tracks
the the host Buffer Descriptors (BDs) and the HWRM assigned ID.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address checkpatch warnings.

v5:
Reorder footer
---
 drivers/net/bnxt/Makefile    |  1 +
 drivers/net/bnxt/bnxt_ring.c | 47 ++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ring.h | 92 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 140 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_ring.c
 create mode 100644 drivers/net/bnxt/bnxt_ring.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index c57afaa..757ea62 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
new file mode 100644
index 0000000..b57d3ba
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -0,0 +1,47 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "bnxt.h"
+#include "bnxt_ring.h"
+
+/*
+ * Generic ring handling
+ */
+
+void bnxt_free_ring(struct bnxt_ring *ring)
+{
+	if (ring->vmem_size && *ring->vmem) {
+		memset((char *)*ring->vmem, 0, ring->vmem_size);
+		*ring->vmem = NULL;
+	}
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
new file mode 100644
index 0000000..3abfb04
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -0,0 +1,92 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RING_H_
+#define _BNXT_RING_H_
+
+#include <inttypes.h>
+
+#include <rte_memory.h>
+
+#define RING_NEXT(ring, idx)		(((idx) + 1) & (ring)->ring_mask)
+
+#define RTE_MBUF_DATA_DMA_ADDR(mb) \
+	((uint64_t)((mb)->buf_physaddr + (mb)->data_off))
+
+#define DB_IDX_MASK						0xffffff
+#define DB_IDX_VALID						(0x1 << 26)
+#define DB_IRQ_DIS						(0x1 << 27)
+#define DB_KEY_TX						(0x0 << 28)
+#define DB_KEY_RX						(0x1 << 28)
+#define DB_KEY_CP						(0x2 << 28)
+#define DB_KEY_ST						(0x3 << 28)
+#define DB_KEY_TX_PUSH						(0x4 << 28)
+#define DB_LONG_TX_PUSH						(0x2 << 24)
+
+#define DEFAULT_CP_RING_SIZE	256
+#define DEFAULT_RX_RING_SIZE	256
+#define DEFAULT_TX_RING_SIZE	256
+
+#define MAX_TPA		128
+
+/* These assume 4k pages */
+#define MAX_RX_DESC_CNT (8 * 1024)
+#define MAX_TX_DESC_CNT (4 * 1024)
+#define MAX_CP_DESC_CNT (16 * 1024)
+
+#define INVALID_HW_RING_ID      ((uint16_t)-1)
+
+struct bnxt_ring {
+	void			*bd;
+	phys_addr_t		bd_dma;
+	uint32_t		ring_size;
+	uint32_t		ring_mask;
+
+	int			vmem_size;
+	void			**vmem;
+
+	uint16_t		fw_ring_id; /* Ring id filled by Chimp FW */
+};
+
+struct bnxt_ring_grp_info {
+	uint16_t	fw_stats_ctx;
+	uint16_t	fw_grp_id;
+	uint16_t	rx_fw_ring_id;
+	uint16_t	cp_fw_ring_id;
+	uint16_t	ag_fw_ring_id;
+};
+
+struct bnxt;
+void bnxt_free_ring(struct bnxt_ring *ring);
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 08/38] bnxt: add completion ring support
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (5 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 07/38] bnxt: declare generic ring structs and free() func Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 09/38] bnxt: add L2 filter alloc/init/free Stephen Hurd
                     ` (32 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Structures, macros, and functions for working with completion rings
in the driver.

Completion Ring is used by the Ethernet controller to provide the
status of transmitted & received packets, report errors, report
status changes to the host software, and inter-function forwarding
requests.  In addition to the generic ring features, a completion ring
can have a statistics context that has statistics periodically DMAed
to host memory, along with a consumer index.

bnxt_handle_async_event() handles completions not related to a specific
transmit or receive ring such as link status changes which arrive on
the default completion ring.

Other physical or virtual functions on the same device may send an HWRM
command forward request.  In this case, we will pass it through
unvalidated. In the future, we will be able to have the PF monitor and
control VF access to the HWRM interface if needed.

New HWRM Calls:
bnxt_hwrm_exec_fwd_resp:
	Execute an encapsulated command and forward the response.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Reorder footer
Expand patch description
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   5 +
 drivers/net/bnxt/bnxt_cpr.c            | 140 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_cpr.h            |  88 +++++++++++++
 drivers/net/bnxt/bnxt_hwrm.c           |  18 +++
 drivers/net/bnxt/bnxt_hwrm.h           |   2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 227 +++++++++++++++++++++++++++++++++
 7 files changed, 481 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_cpr.c
 create mode 100644 drivers/net/bnxt/bnxt_cpr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 757ea62..afd1690 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -48,6 +48,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 # all source are stored in SRCS-y
 #
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index d0f84f4..bdd355f 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -42,6 +42,8 @@
 #include <rte_lcore.h>
 #include <rte_spinlock.h>
 
+#include "bnxt_cpr.h"
+
 #define BNXT_MAX_MTU		9000
 #define VLAN_TAG_SIZE		4
 
@@ -137,6 +139,9 @@ struct bnxt {
 	unsigned int		tx_cp_nr_rings;
 	struct bnxt_tx_queue **tx_queues;
 
+	/* Default completion ring */
+	struct bnxt_cp_ring_info	*def_cp_ring;
+
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
new file mode 100644
index 0000000..7de2b9c
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -0,0 +1,140 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
+
+/*
+ * Async event handling
+ */
+void bnxt_handle_async_event(struct bnxt *bp __rte_unused,
+			     struct cmpl_base *cmp)
+{
+	struct hwrm_async_event_cmpl *async_cmp =
+				(struct hwrm_async_event_cmpl *)cmp;
+
+	/* TODO: HWRM async events are not defined yet */
+	/* Needs to handle: link events, error events, etc. */
+	switch (async_cmp->event_id) {
+	case 0:
+		/* Assume LINK_CHANGE == 0 */
+		RTE_LOG(INFO, PMD, "Link change event\n");
+
+		/* Can just prompt the update_op routine to do a qcfg
+		 * instead of doing the actual qcfg
+		 */
+		break;
+	case 1:
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "handle_async_event id = 0x%x\n",
+			async_cmp->event_id);
+		break;
+	}
+}
+
+void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmpl)
+{
+	struct hwrm_fwd_req_cmpl *fwd_cmpl = (struct hwrm_fwd_req_cmpl *)cmpl;
+	struct input *fwd_cmd;
+	uint16_t logical_vf_id, error_code;
+
+	/* Qualify the fwd request */
+	if (fwd_cmpl->source_id < bp->pf.first_vf_id) {
+		RTE_LOG(ERR, PMD,
+			"FWD req's source_id 0x%x > first_vf_id 0x%x\n",
+			fwd_cmpl->source_id, bp->pf.first_vf_id);
+		error_code = HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED;
+		goto reject;
+	} else if (fwd_cmpl->req_len_type >> HWRM_FWD_REQ_CMPL_REQ_LEN_SFT >
+		   128 - sizeof(struct input)) {
+		RTE_LOG(ERR, PMD,
+		    "FWD req's cmd len 0x%x > 108 bytes allowed\n",
+		    fwd_cmpl->req_len_type >> HWRM_FWD_REQ_CMPL_REQ_LEN_SFT);
+		error_code = HWRM_ERR_CODE_INVALID_PARAMS;
+		goto reject;
+	}
+
+	/* Locate VF's forwarded command */
+	logical_vf_id = fwd_cmpl->source_id - bp->pf.first_vf_id;
+	fwd_cmd = (struct input *)((uint8_t *)bp->pf.vf_req_buf +
+		   (logical_vf_id * 128));
+
+	/* Provision the request */
+	switch (fwd_cmd->req_type) {
+	case HWRM_CFA_L2_FILTER_ALLOC:
+	case HWRM_CFA_L2_FILTER_FREE:
+	case HWRM_CFA_L2_FILTER_CFG:
+	case HWRM_CFA_L2_SET_RX_MASK:
+		break;
+	default:
+		error_code = HWRM_ERR_CODE_INVALID_PARAMS;
+		goto reject;
+	}
+
+	/* Forward */
+	fwd_cmd->target_id = fwd_cmpl->source_id;
+	bnxt_hwrm_exec_fwd_resp(bp, fwd_cmd);
+	return;
+
+reject:
+	/* TODO: Encap the reject error resp into the hwrm_err_iput? */
+	/* Use the error_code for the reject cmd */
+	RTE_LOG(ERR, PMD,
+		"Error 0x%x found in the forward request\n", error_code);
+}
+
+/* For the default completion ring only */
+void bnxt_free_def_cp_ring(struct bnxt *bp)
+{
+	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+	struct bnxt_ring *ring = cpr->cp_ring_struct;
+
+	bnxt_free_ring(ring);
+}
+
+/* For the default completion ring only */
+void bnxt_init_def_ring_struct(struct bnxt *bp)
+{
+	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+	struct bnxt_ring *ring = cpr->cp_ring_struct;
+
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->ring_size = rte_align32pow2(DEFAULT_CP_RING_SIZE);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
new file mode 100644
index 0000000..ecabc5f
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -0,0 +1,88 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_CPR_H_
+#define _BNXT_CPR_H_
+
+#include "hsi_struct_def_dpdk.h"
+
+#define CMP_VALID(cmp, raw_cons, ring)					\
+	(!!(((struct cmpl_base *)(cmp))->info3_v & CMPL_BASE_V) ==	\
+	 !((raw_cons) & ((ring)->ring_size)))
+
+#define CMP_TYPE(cmp)						\
+	(((struct cmpl_base *)cmp)->type & CMPL_BASE_TYPE_MASK)
+
+#define ADV_RAW_CMP(idx, n)	((idx) + (n))
+#define NEXT_RAW_CMP(idx)	ADV_RAW_CMP(idx, 1)
+#define RING_CMP(ring, idx)	((idx) & (ring)->ring_mask)
+#define NEXT_CMP(idx)		RING_CMP(ADV_RAW_CMP(idx, 1))
+
+#define DB_CP_REARM_FLAGS	(DB_KEY_CP | DB_IDX_VALID)
+#define DB_CP_FLAGS		(DB_KEY_CP | DB_IDX_VALID | DB_IRQ_DIS)
+
+#define B_CP_DB_REARM(cpr, raw_cons)					\
+		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_REARM_FLAGS | \
+				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+
+#define B_CP_DIS_DB(cpr, raw_cons)					\
+		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
+				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+
+struct bnxt_ring;
+struct bnxt_cp_ring_info {
+	uint32_t		cp_raw_cons;
+	void			*cp_doorbell;
+
+	struct cmpl_base	*cp_desc_ring;
+
+	phys_addr_t		cp_desc_mapping;
+
+	struct ctx_hw_stats	*hw_stats;
+	phys_addr_t		hw_stats_map;
+	uint32_t		hw_stats_ctx_id;
+
+	struct bnxt_ring	*cp_ring_struct;
+};
+
+#define RX_CMP_L2_ERRORS						\
+	(RX_PKT_CMPL_ERRORS_BUFFER_ERROR_MASK | RX_PKT_CMPL_ERRORS_CRC_ERROR)
+
+
+struct bnxt;
+void bnxt_free_def_cp_ring(struct bnxt *bp);
+void bnxt_init_def_ring_struct(struct bnxt *bp);
+void bnxt_handle_async_event(struct bnxt *bp, struct cmpl_base *cmp);
+void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmp);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index a2d7815..f591ead 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -135,6 +135,24 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
+{
+	int rc;
+	struct hwrm_exec_fwd_resp_input req = {.req_type = 0 };
+	struct hwrm_exec_fwd_resp_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, EXEC_FWD_RESP, -1, resp);
+
+	memcpy(req.encap_request, fwd_cmd,
+	       sizeof(req.encap_request));
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index eef3be6..b792313 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,6 +41,8 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
+
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 92e39ca..9efb68b 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -59,6 +59,137 @@
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+#define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
+#define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
+#define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
+#define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
+
+/* Return Codes */
+#define HWRM_ERR_CODE_INVALID_PARAMS                      (UINT32_C(0x2))
+#define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED              (UINT32_C(0x3))
+
+/* HWRM Forwarded Request (16 bytes) */
+struct hwrm_fwd_req_cmpl {
+	/* Length of forwarded request in bytes. */
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define HWRM_FWD_REQ_CMPL_TYPE_MASK		UINT32_C(0x3f)
+	#define HWRM_FWD_REQ_CMPL_TYPE_SFT		0
+		/* Forwarded HWRM Request */
+	#define HWRM_FWD_REQ_CMPL_TYPE_HWRM_FWD_REQ	(UINT32_C(0x22) << 0)
+	/* Length of forwarded request in bytes. */
+	#define HWRM_FWD_REQ_CMPL_REQ_LEN_MASK		UINT32_C(0xffc0)
+	#define HWRM_FWD_REQ_CMPL_REQ_LEN_SFT		6
+	uint16_t req_len_type;
+
+	/*
+	 * Source ID of this request. Typically used in forwarding requests and
+	 * responses. 0x0 - 0xFFF8 - Used for function ids 0xFFF8 - 0xFFFE -
+	 * Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t source_id;
+
+	uint32_t unused_0;
+
+	/* Address of forwarded request. */
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define HWRM_FWD_REQ_CMPL_V			UINT32_C(0x1)
+	/* Address of forwarded request. */
+	#define HWRM_FWD_REQ_CMPL_REQ_BUF_ADDR_MASK	UINT32_C(0xfffffffe)
+	#define HWRM_FWD_REQ_CMPL_REQ_BUF_ADDR_SFT	1
+	uint64_t req_buf_addr_v;
+} __attribute__((packed));
+
+/* HWRM Asynchronous Event Completion Record (16 bytes) */
+struct hwrm_async_event_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_MASK		UINT32_C(0x3f)
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_SFT		0
+		/* HWRM Asynchronous Event Information */
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_HWRM_ASYNC_EVENT \
+							(UINT32_C(0x2e) << 0)
+	uint16_t type;
+
+	/* Identifiers of events. */
+		/* Link status changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_STATUS_CHANGE \
+							(UINT32_C(0x0) << 0)
+		/* Link MTU changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_MTU_CHANGE \
+							(UINT32_C(0x1) << 0)
+		/* Link speed changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_SPEED_CHANGE \
+							(UINT32_C(0x2) << 0)
+		/* DCB Configuration changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_DCB_CONFIG_CHANGE \
+							(UINT32_C(0x3) << 0)
+		/* Port connection not allowed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PORT_CONN_NOT_ALLOWED \
+							(UINT32_C(0x4) << 0)
+		/* Link speed configuration was not allowed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_SPEED_CFG_NOT_ALLOWED \
+							(UINT32_C(0x5) << 0)
+		/* Function driver unloaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_FUNC_DRVR_UNLOAD \
+							(UINT32_C(0x10) << 0)
+		/* Function driver loaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_FUNC_DRVR_LOAD \
+							(UINT32_C(0x11) << 0)
+		/* PF driver unloaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_DRVR_UNLOAD \
+							(UINT32_C(0x20) << 0)
+		/* PF driver loaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_DRVR_LOAD \
+							(UINT32_C(0x21) << 0)
+		/* VF Function Level Reset (FLR) */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_VF_FLR	(UINT32_C(0x30) << 0)
+		/* VF MAC Address Change */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_VF_MAC_ADDR_CHANGE \
+							(UINT32_C(0x31) << 0)
+		/* PF-VF communication channel status change. */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_VF_COMM_STATUS_CHANGE \
+							(UINT32_C(0x32) << 0)
+		/* HWRM Error */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_HWRM_ERROR \
+							(UINT32_C(0xff) << 0)
+	uint16_t event_id;
+
+	/* Event specific data */
+	uint32_t event_data2;
+
+	/* opaque is 7 b */
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define HWRM_ASYNC_EVENT_CMPL_V				UINT32_C(0x1)
+	/* opaque is 7 b */
+	#define HWRM_ASYNC_EVENT_CMPL_OPAQUE_MASK		UINT32_C(0xfe)
+	#define HWRM_ASYNC_EVENT_CMPL_OPAQUE_SFT		1
+	uint8_t opaque_v;
+
+	/* 8-lsb timestamp from POR (100-msec resolution) */
+	uint8_t timestamp_lo;
+
+	/* 16-lsb timestamp from POR (100-msec resolution) */
+	uint16_t timestamp_hi;
+
+	/* Event specific data */
+	uint32_t event_data1;
+} __attribute__((packed));
 
 /*
  * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
@@ -122,6 +253,102 @@ struct output {
 	uint16_t resp_len;
 } __attribute__((packed));
 
+/* hwrm_exec_fwd_resp */
+/*
+ * Description: This command is used to send an encapsulated request to the
+ * HWRM. This command instructs the HWRM to execute the request and forward the
+ * response of the encapsulated request to the location specified in the
+ * original request that is encapsulated. The target id of this command shall be
+ * set to 0xFFFF (HWRM). The response location in this command shall be used to
+ * acknowledge the receipt of the encapsulated request and forwarding of the
+ * response.
+ */
+
+/* Input (128 bytes) */
+struct hwrm_exec_fwd_resp_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This is an encapsulated request. This request should be executed by
+	 * the HWRM and the response should be provided in the response buffer
+	 * inside the encapsulated request.
+	 */
+	uint32_t encap_request[26];
+
+	/*
+	 * This value indicates the target id of the response to the
+	 * encapsulated request. 0x0 - 0xFFF8 - Used for function ids 0xFFF8 -
+	 * 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t encap_resp_target_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_exec_fwd_resp_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_func_qcaps */
 /*
  * Description: This command returns capabilities of a function. The input FID
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 09/38] bnxt: add L2 filter alloc/init/free
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (6 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 08/38] bnxt: add completion ring support Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 10/38] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
                     ` (31 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add the L2 filter structure and the alloc/init/free functions for
dealing with them.

A filter is used to identify traffic that contains a matching set of
parameters like unicast or broadcast MAC address or a VLAN tag amongst
other things which then allows the ASIC to direct the  incoming traffic
to an appropriate VNIC or Rx ring.

New HWRM calls:
bnxt_hwrm_clear_filter:
	Free a L2 filter.

bnxt_hwrm_set_filter
	Allocate an An L2 filter or a L2 context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Move bnxt_hwrm_set_filter() into this patch
Add short descriptions of new HWRM calls
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   3 +
 drivers/net/bnxt/bnxt_filter.c         | 175 +++++++++++++
 drivers/net/bnxt/bnxt_filter.h         |  74 ++++++
 drivers/net/bnxt/bnxt_hwrm.c           |  65 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   6 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 456 +++++++++++++++++++++++++++++++++
 7 files changed, 780 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_filter.c
 create mode 100644 drivers/net/bnxt/bnxt_filter.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index afd1690..b7834b1 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index bdd355f..49aa38b 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -145,6 +145,9 @@ struct bnxt {
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
+	struct bnxt_filter_info	*filter_info;
+	STAILQ_HEAD(, bnxt_filter_info)	free_filter_list;
+
 	/* VNIC pointer for flow filter (VMDq) pools */
 #define MAX_FF_POOLS	ETH_64_POOLS
 	STAILQ_HEAD(, bnxt_vnic_info)	ff_pool[MAX_FF_POOLS];
diff --git a/drivers/net/bnxt/bnxt_filter.c b/drivers/net/bnxt/bnxt_filter.c
new file mode 100644
index 0000000..f03a1dc
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_filter.c
@@ -0,0 +1,175 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <sys/queue.h>
+
+#include <rte_log.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * Filter Functions
+ */
+
+struct bnxt_filter_info *bnxt_alloc_filter(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+
+	/* Find the 1st unused filter from the free_filter_list pool*/
+	filter = STAILQ_FIRST(&bp->free_filter_list);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "No more free filter resources\n");
+		return NULL;
+	}
+	STAILQ_REMOVE_HEAD(&bp->free_filter_list, next);
+
+	/* Default to L2 MAC Addr filter */
+	filter->flags = HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX;
+	filter->enables = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
+			HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK;
+	memcpy(filter->l2_addr, bp->eth_dev->data->mac_addrs->addr_bytes,
+	       ETHER_ADDR_LEN);
+	memset(filter->l2_addr_mask, 0xff, ETHER_ADDR_LEN);
+	return filter;
+}
+
+void bnxt_init_filters(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+	int i, max_filters;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	STAILQ_INIT(&bp->free_filter_list);
+	for (i = 0; i < max_filters; i++) {
+		filter = &bp->filter_info[i];
+		filter->fw_l2_filter_id = -1;
+		STAILQ_INSERT_TAIL(&bp->free_filter_list, filter, next);
+	}
+}
+
+void bnxt_free_all_filters(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter, *temp_filter;
+	int i;
+
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			filter = STAILQ_FIRST(&vnic->filter);
+			while (filter) {
+				temp_filter = STAILQ_NEXT(filter, next);
+				STAILQ_REMOVE(&vnic->filter, filter,
+					      bnxt_filter_info, next);
+				STAILQ_INSERT_TAIL(&bp->free_filter_list,
+						   filter, next);
+				filter = temp_filter;
+			}
+			STAILQ_INIT(&vnic->filter);
+		}
+	}
+}
+
+void bnxt_free_filter_mem(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+	uint16_t max_filters, i;
+	int rc = 0;
+
+	/* Ensure that all filters are freed */
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	for (i = 0; i < max_filters; i++) {
+		filter = &bp->filter_info[i];
+		if (filter->fw_l2_filter_id != ((uint64_t)-1)) {
+			RTE_LOG(ERR, PMD, "HWRM filter is not freed??\n");
+			/* Call HWRM to try to free filter again */
+			rc = bnxt_hwrm_clear_filter(bp, filter);
+			if (rc)
+				RTE_LOG(ERR, PMD,
+				       "HWRM filter cannot be freed rc = %d\n",
+					rc);
+		}
+		filter->fw_l2_filter_id = -1;
+	}
+	STAILQ_INIT(&bp->free_filter_list);
+
+	rte_free(bp->filter_info);
+	bp->filter_info = NULL;
+}
+
+int bnxt_alloc_filter_mem(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter_mem;
+	uint16_t max_filters;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	/* Allocate memory for VNIC pool and filter pool */
+	filter_mem = rte_zmalloc("bnxt_filter_info",
+				 max_filters * sizeof(struct bnxt_filter_info),
+				 0);
+	if (filter_mem == NULL) {
+		RTE_LOG(ERR, PMD, "Failed to alloc memory for %d filters",
+			max_filters);
+		return -ENOMEM;
+	}
+	bp->filter_info = filter_mem;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_filter.h b/drivers/net/bnxt/bnxt_filter.h
new file mode 100644
index 0000000..06fe134
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_filter.h
@@ -0,0 +1,74 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_FILTER_H_
+#define _BNXT_FILTER_H_
+
+#include <rte_ether.h>
+
+struct bnxt;
+struct bnxt_filter_info {
+	STAILQ_ENTRY(bnxt_filter_info)	next;
+	uint64_t		fw_l2_filter_id;
+#define INVALID_MAC_INDEX	((uint16_t)-1)
+	uint16_t		mac_index;
+
+	/* Filter Characteristics */
+	uint32_t		flags;
+	uint32_t		enables;
+	uint8_t			l2_addr[ETHER_ADDR_LEN];
+	uint8_t			l2_addr_mask[ETHER_ADDR_LEN];
+	uint16_t		l2_ovlan;
+	uint16_t		l2_ovlan_mask;
+	uint16_t		l2_ivlan;
+	uint16_t		l2_ivlan_mask;
+	uint8_t			t_l2_addr[ETHER_ADDR_LEN];
+	uint8_t			t_l2_addr_mask[ETHER_ADDR_LEN];
+	uint16_t		t_l2_ovlan;
+	uint16_t		t_l2_ovlan_mask;
+	uint16_t		t_l2_ivlan;
+	uint16_t		t_l2_ivlan_mask;
+	uint8_t			tunnel_type;
+	uint16_t		mirror_vnic_id;
+	uint32_t		vni;
+	uint8_t			pri_hint;
+	uint64_t		l2_filter_id_hint;
+};
+
+struct bnxt_filter_info *bnxt_alloc_filter(struct bnxt *bp);
+void bnxt_init_filters(struct bnxt *bp);
+void bnxt_free_all_filters(struct bnxt *bp);
+void bnxt_free_filter_mem(struct bnxt *bp);
+int bnxt_alloc_filter_mem(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index f591ead..65cd4a7 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -39,7 +39,9 @@
 #include <rte_version.h>
 
 #include "bnxt.h"
+#include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
 #define HWRM_CMD_TIMEOUT		2000
@@ -135,6 +137,69 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_clear_filter(struct bnxt *bp,
+			   struct bnxt_filter_info *filter)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_filter_free_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_filter_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, CFA_L2_FILTER_FREE, -1, resp);
+
+	req.l2_filter_id = rte_cpu_to_le_64(filter->fw_l2_filter_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	filter->fw_l2_filter_id = -1;
+
+	return 0;
+}
+
+int bnxt_hwrm_set_filter(struct bnxt *bp,
+			 struct bnxt_vnic_info *vnic,
+			 struct bnxt_filter_info *filter)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_filter_alloc_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_filter_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t enables = 0;
+
+	HWRM_PREP(req, CFA_L2_FILTER_ALLOC, -1, resp);
+
+	req.flags = rte_cpu_to_le_32(filter->flags);
+
+	enables = filter->enables |
+	      HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_DST_ID;
+	req.dst_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR)
+		memcpy(req.l2_addr, filter->l2_addr,
+		       ETHER_ADDR_LEN);
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK)
+		memcpy(req.l2_addr_mask, filter->l2_addr_mask,
+		       ETHER_ADDR_LEN);
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN)
+		req.l2_ovlan = filter->l2_ovlan;
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN_MASK)
+		req.l2_ovlan_mask = filter->l2_ovlan_mask;
+
+	req.enables = rte_cpu_to_le_32(enables);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	filter->fw_l2_filter_id = rte_le_to_cpu_64(resp->l2_filter_id);
+
+	return rc;
+}
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
 {
 	int rc;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index b792313..bc6fb08 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,6 +41,12 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_clear_filter(struct bnxt *bp,
+			   struct bnxt_filter_info *filter);
+int bnxt_hwrm_set_filter(struct bnxt *bp,
+			 struct bnxt_vnic_info *vnic,
+			 struct bnxt_filter_info *filter);
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 9efb68b..ca42a87 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -253,6 +253,462 @@ struct output {
 	uint16_t resp_len;
 } __attribute__((packed));
 
+/* hwrm_cfa_l2_filter_alloc */
+/*
+ * A filter is used to identify traffic that contains a matching set of
+ * parameters like unicast or broadcast MAC address or a VLAN tag amongst
+ * other things which then allows the ASIC to direct the  incoming traffic
+ * to an appropriate VNIC or Rx ring.
+ */
+
+/* Input (96 bytes) */
+struct hwrm_cfa_l2_filter_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Enumeration denoting the RX, TX type of the resource. This
+	 * enumeration is used for resources that are similar for both TX and RX
+	 * paths of the chip.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH \
+							UINT32_C(0x1)
+		/* tx path */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_TX \
+							(UINT32_C(0x0) << 0)
+		/* rx path */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX \
+							(UINT32_C(0x1) << 0)
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_LAST \
+				HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX
+	/*
+	 * Setting of this flag indicates the applicability to the loopback
+	 * path.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_LOOPBACK \
+							UINT32_C(0x2)
+	/*
+	 * Setting of this flag indicates drop action. If this flag is not set,
+	 * then it should be considered accept action.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_DROP \
+							UINT32_C(0x4)
+	/*
+	 * If this flag is set, all t_l2_* fields are invalid and they should
+	 * not be specified. If this flag is set, then l2_* fields refer to
+	 * fields of outermost L2 header.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST \
+							UINT32_C(0x8)
+	uint32_t flags;
+
+	/* This bit must be '1' for the l2_addr field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR \
+							UINT32_C(0x1)
+	/* This bit must be '1' for the l2_addr_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK \
+							UINT32_C(0x2)
+	/* This bit must be '1' for the l2_ovlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN \
+							UINT32_C(0x4)
+	/* This bit must be '1' for the l2_ovlan_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN_MASK \
+							UINT32_C(0x8)
+	/* This bit must be '1' for the l2_ivlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN \
+							UINT32_C(0x10)
+	/* This bit must be '1' for the l2_ivlan_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK \
+							UINT32_C(0x20)
+	/* This bit must be '1' for the t_l2_addr field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_ADDR \
+							UINT32_C(0x40)
+	/*
+	 * This bit must be '1' for the t_l2_addr_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_ADDR_MASK \
+							UINT32_C(0x80)
+	/* This bit must be '1' for the t_l2_ovlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_OVLAN \
+							UINT32_C(0x100)
+	/*
+	 * This bit must be '1' for the t_l2_ovlan_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_OVLAN_MASK \
+							UINT32_C(0x200)
+	/* This bit must be '1' for the t_l2_ivlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_IVLAN \
+							UINT32_C(0x400)
+	/*
+	 * This bit must be '1' for the t_l2_ivlan_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_IVLAN_MASK \
+							UINT32_C(0x800)
+	/* This bit must be '1' for the src_type field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_SRC_TYPE \
+							UINT32_C(0x1000)
+	/* This bit must be '1' for the src_id field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_SRC_ID \
+							UINT32_C(0x2000)
+	/* This bit must be '1' for the tunnel_type field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE \
+							UINT32_C(0x4000)
+	/* This bit must be '1' for the dst_id field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_DST_ID \
+							UINT32_C(0x8000)
+	/*
+	 * This bit must be '1' for the mirror_vnic_id field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_MIRROR_VNIC_ID \
+							UINT32_C(0x10000)
+	uint32_t enables;
+
+	/*
+	 * This value sets the match value for the L2 MAC address. Destination
+	 * MAC address for RX path. Source MAC address for TX path.
+	 */
+	uint8_t l2_addr[6];
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+
+	/*
+	 * This value sets the mask value for the L2 address. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint8_t l2_addr_mask[6];
+
+	/* This value sets VLAN ID value for outer VLAN. */
+	uint16_t l2_ovlan;
+
+	/*
+	 * This value sets the mask value for the ovlan id. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint16_t l2_ovlan_mask;
+
+	/* This value sets VLAN ID value for inner VLAN. */
+	uint16_t l2_ivlan;
+
+	/*
+	 * This value sets the mask value for the ivlan id. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint16_t l2_ivlan_mask;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This value sets the match value for the tunnel L2 MAC address.
+	 * Destination MAC address for RX path. Source MAC address for TX path.
+	 */
+	uint8_t t_l2_addr[6];
+
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/*
+	 * This value sets the mask value for the tunnel L2 address. A value of
+	 * 0 will mask the corresponding bit from compare.
+	 */
+	uint8_t t_l2_addr_mask[6];
+
+	/* This value sets VLAN ID value for tunnel outer VLAN. */
+	uint16_t t_l2_ovlan;
+
+	/*
+	 * This value sets the mask value for the tunnel ovlan id. A value of 0
+	 * will mask the corresponding bit from compare.
+	 */
+	uint16_t t_l2_ovlan_mask;
+
+	/* This value sets VLAN ID value for tunnel inner VLAN. */
+	uint16_t t_l2_ivlan;
+
+	/*
+	 * This value sets the mask value for the tunnel ivlan id. A value of 0
+	 * will mask the corresponding bit from compare.
+	 */
+	uint16_t t_l2_ivlan_mask;
+
+	/* This value identifies the type of source of the packet. */
+		/* Network port */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_NPORT \
+							(UINT32_C(0x0) << 0)
+		/* Physical function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_PF \
+							(UINT32_C(0x1) << 0)
+		/* Virtual function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_VF \
+							(UINT32_C(0x2) << 0)
+		/* Virtual NIC of a function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_VNIC \
+							(UINT32_C(0x3) << 0)
+		/* Embedded processor for CFA management */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_KONG \
+							(UINT32_C(0x4) << 0)
+		/* Embedded processor for OOB management */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_APE \
+							(UINT32_C(0x5) << 0)
+		/* Embedded processor for RoCE */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_BONO \
+							(UINT32_C(0x6) << 0)
+		/* Embedded processor for network proxy functions */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_TANG \
+							(UINT32_C(0x7) << 0)
+	uint8_t src_type;
+
+	uint8_t unused_6;
+	/*
+	 * This value is the id of the source. For a network port, it represents
+	 * port_id. For a physical function, it represents fid. For a virtual
+	 * function, it represents vf_id. For a vnic, it represents vnic_id. For
+	 * embedded processors, this id is not valid. Notes: 1. The function ID
+	 * is implied if it src_id is not provided for a src_type that is either
+	 */
+	uint32_t src_id;
+
+	/* Tunnel Type. */
+		/* Non-tunnel */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_NONTUNNEL \
+							(UINT32_C(0x0) << 0)
+		/* Virtual eXtensible Local Area Network (VXLAN) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_VXLAN \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Network Virtualization Generic Routing Encapsulation (NVGRE)
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_NVGRE \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Generic Routing Encapsulation (GRE) inside Ethernet payload
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_L2GRE \
+							(UINT32_C(0x3) << 0)
+		/* IP in IP */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_IPIP \
+							(UINT32_C(0x4) << 0)
+		/* Generic Network Virtualization Encapsulation (Geneve) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_GENEVE \
+							(UINT32_C(0x5) << 0)
+		/* Multi-Protocol Lable Switching (MPLS) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_MPLS \
+							(UINT32_C(0x6) << 0)
+		/* Stateless Transport Tunnel (STT) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_STT \
+							(UINT32_C(0x7) << 0)
+		/*
+		 * Generic Routing Encapsulation (GRE) inside IP datagram
+		 * payload
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_IPGRE \
+							(UINT32_C(0x8) << 0)
+		/* Any tunneled traffic */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_ANYTUNNEL \
+							(UINT32_C(0xff) << 0)
+	uint8_t tunnel_type;
+
+	uint8_t unused_7;
+
+	/*
+	 * If set, this value shall represent the Logical VNIC ID of the
+	 * destination VNIC for the RX path and network port id of the
+	 * destination port for the TX path.
+	 */
+	uint16_t dst_id;
+
+	/* Logical VNIC ID of the VNIC where traffic is mirrored. */
+	uint16_t mirror_vnic_id;
+
+	/*
+	 * This hint is provided to help in placing the filter in the filter
+	 * table.
+	 */
+		/* No preference */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_NO_PREFER \
+							(UINT32_C(0x0) << 0)
+		/* Above the given filter */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_ABOVE_FILTER \
+							(UINT32_C(0x1) << 0)
+		/* Below the given filter */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_BELOW_FILTER \
+							(UINT32_C(0x2) << 0)
+		/* As high as possible */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_MAX \
+							(UINT32_C(0x3) << 0)
+		/* As low as possible */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_MIN \
+							(UINT32_C(0x4) << 0)
+	uint8_t pri_hint;
+
+	uint8_t unused_8;
+	uint32_t unused_9;
+
+	/*
+	 * This is the ID of the filter that goes along with the pri_hint. This
+	 * field is valid only for the following values. 1 - Above the given
+	 * filter 2 - Below the given filter
+	 */
+	uint64_t l2_filter_id_hint;
+} __attribute__((packed));
+
+/* Output (24 bytes) */
+struct hwrm_cfa_l2_filter_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This value identifies a set of CFA data structures used for an L2
+	 * context.
+	 */
+	uint64_t l2_filter_id;
+
+	/*
+	 * This is the ID of the flow associated with this filter. This value
+	 * shall be used to match and associate the flow identifier returned in
+	 * completion records. A value of 0xFFFFFFFF shall indicate no flow id.
+	 */
+	uint32_t flow_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_cfa_l2_filter_free */
+/*
+ * Description: Free a L2 filter. The HWRM shall free all associated filter
+ * resources with the L2 filter.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_cfa_l2_filter_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This value identifies a set of CFA data structures used for an L2
+	 * context.
+	 */
+	uint64_t l2_filter_id;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_cfa_l2_filter_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_exec_fwd_resp */
 /*
  * Description: This command is used to send an encapsulated request to the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 10/38] bnxt: add Tx queue operations (nonfunctional)
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (7 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 09/38] bnxt: add L2 filter alloc/init/free Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 11/38] bnxt: add Rx queue create/destroy operations Stephen Hurd
                     ` (30 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

In this patch we are adding the bnxt_tx_queue_setup_op() and
bnxt_tx_queue_release_op() functions. These will be tied to the
tx_queue_setup and tx_queue_release dev_ops in a subsequent patch.
In these functions we allocate/free memory for the TX queues.

This still requires support to create a TX ring in the ASIC which
will be completed in a future commit. Each Tx queue created via the
tx_queue_setup dev_op will have an associated Tx ring in the hardware.

A Tx request coming on the Tx queue gets sent to the corresponding
Tx ring in the ASIC for subsequent transmission.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/Makefile      |   1 +
 drivers/net/bnxt/bnxt_ethdev.c |   3 +
 drivers/net/bnxt/bnxt_txq.c    | 124 +++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_txq.h    |  75 +++++++++++++++++++++++++
 4 files changed, 203 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_txq.c
 create mode 100644 drivers/net/bnxt/bnxt_txq.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index b7834b1..13a90b9 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -53,6 +53,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b46d2ce..77a6d92 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -41,6 +41,7 @@
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -185,6 +186,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.tx_queue_setup = bnxt_tx_queue_setup_op,
+	.tx_queue_release = bnxt_tx_queue_release_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
new file mode 100644
index 0000000..8928984
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -0,0 +1,124 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_ring.h"
+#include "bnxt_txq.h"
+
+/*
+ * TX Queues
+ */
+
+void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+
+	/* 'Unreserve' rte_memzone */
+
+	if (cpr->hw_stats)
+		cpr->hw_stats = NULL;
+}
+
+static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq __rte_unused)
+{
+	/* TODO: Requires interaction with TX ring */
+}
+
+void bnxt_free_tx_mbufs(struct bnxt *bp)
+{
+	struct bnxt_tx_queue *txq;
+	int i;
+
+	for (i = 0; i < (int)bp->tx_nr_rings; i++) {
+		txq = bp->tx_queues[i];
+		bnxt_tx_queue_release_mbufs(txq);
+	}
+}
+
+void bnxt_tx_queue_release_op(void *tx_queue)
+{
+	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
+
+	if (txq) {
+		/* TODO: Free ring and stats here */
+		rte_free(txq);
+	}
+}
+
+int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_txconf *tx_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_tx_queue *txq;
+
+	if (!nb_desc || nb_desc > MAX_TX_DESC_CNT) {
+		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
+		return -EINVAL;
+	}
+
+	if (eth_dev->data->tx_queues) {
+		txq = eth_dev->data->tx_queues[queue_idx];
+		if (txq) {
+			bnxt_tx_queue_release_op(txq);
+			txq = NULL;
+		}
+	}
+	txq = rte_zmalloc_socket("bnxt_tx_queue", sizeof(struct bnxt_tx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq == NULL) {
+		RTE_LOG(ERR, PMD, "bnxt_tx_queue allocation failed!");
+		return -ENOMEM;
+	}
+	txq->bp = bp;
+	txq->nb_tx_desc = nb_desc;
+	txq->tx_free_thresh = tx_conf->tx_free_thresh;
+
+	/* TODO: Initialize ring structure */
+
+	txq->queue_id = queue_idx;
+	txq->port_id = eth_dev->data->port_id;
+
+	/* TODO: Allocate TX ring hardware descriptors */
+
+	/* TODO: Initialize the ring */
+
+	eth_dev->data->tx_queues[queue_idx] = txq;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_txq.h b/drivers/net/bnxt/bnxt_txq.h
new file mode 100644
index 0000000..16f3a0b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txq.h
@@ -0,0 +1,75 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_TXQ_H_
+#define _BNXT_TXQ_H_
+
+struct bnxt_tx_ring_info;
+struct bnxt_cp_ring_info;
+struct bnxt_tx_queue {
+	uint16_t		nb_tx_desc;    /* number of TX descriptors */
+	uint16_t		tx_free_thresh;/* minimum TX before freeing */
+	/** Index to last TX descriptor to have been cleaned. */
+	uint16_t		last_desc_cleaned;
+	/** Total number of TX descriptors ready to be allocated. */
+	uint16_t		tx_next_dd; /* next desc to scan for DD bit */
+	uint16_t		tx_next_rs; /* next desc to set RS bit */
+	uint16_t		queue_id; /* TX queue index */
+	uint16_t		reg_idx; /* TX queue register index */
+	uint8_t			port_id; /* Device port identifier */
+	uint8_t			pthresh; /* Prefetch threshold register */
+	uint8_t			hthresh; /* Host threshold register */
+	uint8_t			wthresh; /* Write-back threshold reg */
+	uint32_t		txq_flags; /* Holds flags for this TXq */
+	uint32_t		ctx_curr; /* Hardware context states */
+	uint8_t			tx_deferred_start; /* not in global dev start */
+
+	struct bnxt		*bp;
+	int			index;
+	int			tx_wake_thresh;
+	struct bnxt_tx_ring_info	*tx_ring;
+
+	unsigned int		cp_nr_rings;
+	struct bnxt_cp_ring_info	*cp_ring;
+};
+
+void bnxt_free_txq_stats(struct bnxt_tx_queue *txq);
+void bnxt_free_tx_mbufs(struct bnxt *bp);
+void bnxt_tx_queue_release_op(void *tx_queue);
+int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_txconf *tx_conf);
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 11/38] bnxt: add Rx queue create/destroy operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (8 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 10/38] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 12/38] bnxt: add statistics operations Stephen Hurd
                     ` (29 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

In this patch we are adding the bnxt_rx_queue_setup_op() and
bnxt_rx_queue_release_op() functions. These will be tied to the
rx_queue_setup and rx_queue_release dev_ops in a subsequent patch.
In these functions we allocate/free memory for the RX queues.

This still requires support to create a RX ring in the ASIC which
will be completed in a future commit. Each Rx queue created via the
rx_queue_setup dev_op will have an associated Rx ring in the hardware.

The Rx logic in the hardware picks a Rx ring for each Rx frame received
by the hardware depending on the properties like RSS, MAC and VLAN
settings configured in the hardware. These packets in the end arrive
on the Rx queue corresponding to the Rx ring in the hardware.

We are also adding some functions like bnxt_mq_rx_configure()
bnxt_free_rx_mbufs() and bnxt_free_rxq_stats() which will be used in
subsequent patches.

We are also adding hwrm_vnic_rss_cfg_* structures, which will be used
in subsequent patches to enable RSS configuration.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   2 +
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +
 drivers/net/bnxt/bnxt_rxq.c            | 287 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_rxq.h            |  74 +++++++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 121 ++++++++++++++
 6 files changed, 488 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_rxq.c
 create mode 100644 drivers/net/bnxt/bnxt_rxq.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 13a90b9..21ed71c 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -53,6 +53,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index 49aa38b..f7cf9d1 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -142,6 +142,8 @@ struct bnxt {
 	/* Default completion ring */
 	struct bnxt_cp_ring_info	*def_cp_ring;
 
+	unsigned int		nr_vnics;
+
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 77a6d92..7e7d1ab 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -41,6 +41,7 @@
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
 #include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
@@ -186,6 +187,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.rx_queue_setup = bnxt_rx_queue_setup_op,
+	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
 };
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
new file mode 100644
index 0000000..d5c3173
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -0,0 +1,287 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
+#include "bnxt_rxq.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * RX Queues
+ */
+
+void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+
+	/* 'Unreserve' rte_memzone */
+
+	if (cpr->hw_stats)
+		cpr->hw_stats = NULL;
+}
+
+int bnxt_mq_rx_configure(struct bnxt *bp)
+{
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	unsigned int i, j, nb_q_per_grp, ring_idx;
+	int start_grp_id, end_grp_id, rc = 0;
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter;
+	struct bnxt_rx_queue *rxq;
+
+	bp->nr_vnics = 0;
+
+	/* Single queue mode */
+	if (bp->rx_cp_nr_rings < 2) {
+		vnic = bnxt_alloc_vnic(bp);
+		if (!vnic) {
+			RTE_LOG(ERR, PMD, "VNIC alloc failed\n");
+			rc = -ENOMEM;
+			goto err_out;
+		}
+		STAILQ_INSERT_TAIL(&bp->ff_pool[0], vnic, next);
+		bp->nr_vnics++;
+
+		rxq = bp->eth_dev->data->rx_queues[0];
+		rxq->vnic = vnic;
+
+		vnic->func_default = true;
+		vnic->ff_pool_idx = 0;
+		vnic->start_grp_id = 1;
+		vnic->end_grp_id = vnic->start_grp_id +
+				   bp->rx_cp_nr_rings - 1;
+		filter = bnxt_alloc_filter(bp);
+		if (!filter) {
+			RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+			rc = -ENOMEM;
+			goto err_out;
+		}
+		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+		goto out;
+	}
+
+	/* Multi-queue mode */
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG) {
+		/* VMDq ONLY, VMDq+RSS, VMDq+DCB, VMDq+DCB+RSS */
+		enum rte_eth_nb_pools pools;
+
+		switch (dev_conf->rxmode.mq_mode) {
+		case ETH_MQ_RX_VMDQ_RSS:
+		case ETH_MQ_RX_VMDQ_ONLY:
+			{
+				const struct rte_eth_vmdq_rx_conf *conf =
+				    &dev_conf->rx_adv_conf.vmdq_rx_conf;
+
+				/* ETH_8/64_POOLs */
+				pools = conf->nb_queue_pools;
+				break;
+			}
+		default:
+			RTE_LOG(ERR, PMD, "Unsupported mq_mod %d\n",
+				dev_conf->rxmode.mq_mode);
+			rc = -EINVAL;
+			goto err_out;
+		}
+		/* For each pool, allocate MACVLAN CFA rule & VNIC */
+		if (!pools) {
+			RTE_LOG(ERR, PMD,
+				"VMDq pool not set, defaulted to 64\n");
+			pools = ETH_64_POOLS;
+		}
+		nb_q_per_grp = bp->rx_cp_nr_rings / pools;
+		start_grp_id = 1;
+		end_grp_id = start_grp_id + nb_q_per_grp - 1;
+
+		ring_idx = 0;
+		for (i = 0; i < pools; i++) {
+			vnic = bnxt_alloc_vnic(bp);
+			if (!vnic) {
+				RTE_LOG(ERR, PMD,
+					"VNIC alloc failed\n");
+				rc = -ENOMEM;
+				goto err_out;
+			}
+			STAILQ_INSERT_TAIL(&bp->ff_pool[i], vnic, next);
+			bp->nr_vnics++;
+
+			for (j = 0; j < nb_q_per_grp; j++, ring_idx++) {
+				rxq = bp->eth_dev->data->rx_queues[ring_idx];
+				rxq->vnic = vnic;
+			}
+			if (i == 0)
+				vnic->func_default = true;
+			vnic->ff_pool_idx = i;
+			vnic->start_grp_id = start_grp_id;
+			vnic->end_grp_id = end_grp_id;
+
+			filter = bnxt_alloc_filter(bp);
+			if (!filter) {
+				RTE_LOG(ERR, PMD,
+					"L2 filter alloc failed\n");
+				rc = -ENOMEM;
+				goto err_out;
+			}
+			/*
+			 * TODO: Configure & associate CFA rule for
+			 * each VNIC for each VMDq with MACVLAN, MACVLAN+TC
+			 */
+			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+
+			start_grp_id = end_grp_id + 1;
+			end_grp_id += nb_q_per_grp;
+		}
+		goto out;
+	}
+
+	/* Non-VMDq mode - RSS, DCB, RSS+DCB */
+	/* Init default VNIC for RSS or DCB only */
+	vnic = bnxt_alloc_vnic(bp);
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "VNIC alloc failed\n");
+		rc = -ENOMEM;
+		goto err_out;
+	}
+	/* Partition the rx queues for the single pool */
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		rxq = bp->eth_dev->data->rx_queues[i];
+		rxq->vnic = vnic;
+	}
+	STAILQ_INSERT_TAIL(&bp->ff_pool[0], vnic, next);
+	bp->nr_vnics++;
+
+	vnic->func_default = true;
+	vnic->ff_pool_idx = 0;
+	vnic->start_grp_id = 1;
+	vnic->end_grp_id = vnic->start_grp_id +
+			   bp->rx_cp_nr_rings - 1;
+	filter = bnxt_alloc_filter(bp);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+		rc = -ENOMEM;
+		goto err_out;
+	}
+	STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
+		vnic->hash_type =
+			HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4 |
+			HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+
+out:
+	return rc;
+
+err_out:
+	/* Free allocated vnic/filters */
+
+	return rc;
+}
+
+static void bnxt_rx_queue_release_mbufs(struct bnxt_rx_queue *rxq __rte_unused)
+{
+	/* TODO: Requires interaction with TX ring */
+}
+
+void bnxt_free_rx_mbufs(struct bnxt *bp)
+{
+	struct bnxt_rx_queue *rxq;
+	int i;
+
+	for (i = 0; i < (int)bp->rx_nr_rings; i++) {
+		rxq = bp->rx_queues[i];
+		bnxt_rx_queue_release_mbufs(rxq);
+	}
+}
+
+void bnxt_rx_queue_release_op(void *rx_queue)
+{
+	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
+
+	if (rxq) {
+		bnxt_rx_queue_release_mbufs(rxq);
+
+		/* TODO: Free ring and stats here */
+
+		rte_free(rxq);
+	}
+}
+
+int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_rxconf *rx_conf,
+			       struct rte_mempool *mp)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_rx_queue *rxq;
+
+	if (!nb_desc || nb_desc > MAX_RX_DESC_CNT) {
+		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
+		return -EINVAL;
+	}
+
+	if (eth_dev->data->rx_queues) {
+		rxq = eth_dev->data->rx_queues[queue_idx];
+		if (rxq)
+			bnxt_rx_queue_release_op(rxq);
+	}
+	rxq = rte_zmalloc_socket("bnxt_rx_queue", sizeof(struct bnxt_rx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (!rxq) {
+		RTE_LOG(ERR, PMD, "bnxt_rx_queue allocation failed!");
+		return -ENOMEM;
+	}
+	rxq->bp = bp;
+	rxq->mb_pool = mp;
+	rxq->nb_rx_desc = nb_desc;
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+
+	/* TODO: Initialize ring structure */
+
+	rxq->queue_id = queue_idx;
+	rxq->port_id = eth_dev->data->port_id;
+	rxq->crc_len = (uint8_t)((eth_dev->data->dev_conf.rxmode.hw_strip_crc) ?
+				0 : ETHER_CRC_LEN);
+
+	eth_dev->data->rx_queues[queue_idx] = rxq;
+	/* TODO: Allocate RX ring hardware descriptors */
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_rxq.h b/drivers/net/bnxt/bnxt_rxq.h
new file mode 100644
index 0000000..9554329
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxq.h
@@ -0,0 +1,74 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RQX_H_
+#define _BNXT_RQX_H_
+
+struct bnxt;
+struct bnxt_rx_ring_info;
+struct bnxt_cp_ring_info;
+struct bnxt_rx_queue {
+	struct rte_mempool	*mb_pool; /* mbuf pool for RX ring */
+	struct rte_mbuf		*pkt_first_seg; /* 1st seg of pkt */
+	struct rte_mbuf		*pkt_last_seg; /* Last seg of pkt */
+	uint64_t		mbuf_initializer; /* val to init mbuf */
+	uint16_t		nb_rx_desc; /* num of RX desc */
+	uint16_t		rx_tail; /* cur val of RDT register */
+	uint16_t		nb_rx_hold; /* num held free RX desc */
+	uint16_t		rx_free_thresh; /* max free RX desc to hold */
+	uint16_t		queue_id; /* RX queue index */
+	uint16_t		reg_idx; /* RX queue register index */
+	uint8_t			port_id; /* Device port identifier */
+	uint8_t			crc_len; /* 0 if CRC stripped, 4 otherwise */
+
+	struct bnxt		*bp;
+	struct bnxt_vnic_info	*vnic;
+
+	uint32_t			rx_buf_size;
+	uint32_t			rx_buf_use_size;  /* useable size */
+	struct bnxt_rx_ring_info	*rx_ring;
+	struct bnxt_cp_ring_info	*cp_ring;
+};
+
+void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq);
+int bnxt_mq_rx_configure(struct bnxt *bp);
+void bnxt_rx_queue_release_op(void *rx_queue);
+int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_rxconf *rx_conf,
+			       struct rte_mempool *mp);
+void bnxt_free_rx_mbufs(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ca42a87..bbec5bb 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1897,6 +1897,127 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_vnic_rss_cfg */
+/* Description: This function is used to enable RSS configuration. */
+
+/* Input (48 bytes) */
+struct hwrm_vnic_rss_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over source and
+	 * destination IPv4 addresses of IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4		UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv4 addresses and source/destination ports of
+	 * TCP/IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv4 addresses and source/destination ports of
+	 * UDP/IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4	UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over source and
+	 * destination IPv4 addresses of IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6		UINT32_C(0x8)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv6 addresses and source/destination ports of
+	 * TCP/IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6	UINT32_C(0x10)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv6 addresses and source/destination ports of
+	 * UDP/IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6	UINT32_C(0x20)
+	uint32_t hash_type;
+
+	uint32_t unused_0;
+
+	/* This is the address for rss ring group table */
+	uint64_t ring_grp_tbl_addr;
+
+	/* This is the address for rss hash key table */
+	uint64_t hash_key_tbl_addr;
+
+	/* Index to the rss indirection table. */
+	uint16_t rss_ctx_idx;
+
+	uint16_t unused_1[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_rss_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* Output (32 bytes) */
 struct hwrm_queue_qportcfg_output {
 	/*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 12/38] bnxt: add statistics operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (9 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 11/38] bnxt: add Rx queue create/destroy operations Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 13/38] bnxt: add initial Tx code implementation Stephen Hurd
                     ` (28 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add the bnxt_stats_get_op() and bnxt_stats_reset_op() dev_ops to
get and reset staitstics. It also brings in the asociated HWRM calls
to handle the requests appropriately.

We also have the bnxt_free_stats() function which will be used in the
follow on patches to free the memory allocated by the driver for
statistics.

New HWRM calls:
bnxt_hwrm_stat_clear:
	This command clears statistics of a context

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Add more description for the patch
Add HWRM function descriptions
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_cpr.c            |   1 +
 drivers/net/bnxt/bnxt_cpr.h            |   2 -
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +
 drivers/net/bnxt/bnxt_hwrm.c           |  49 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   7 +-
 drivers/net/bnxt/bnxt_rxq.c            |   1 +
 drivers/net/bnxt/bnxt_stats.c          | 142 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_stats.h          |  44 ++++++++++
 drivers/net/bnxt/bnxt_txq.c            |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 107 +++++++++++++++++++++++++
 11 files changed, 355 insertions(+), 3 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_stats.c
 create mode 100644 drivers/net/bnxt/bnxt_stats.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 21ed71c..f6a04f8 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -54,6 +54,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index 7de2b9c..5a4b13d 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -35,6 +35,7 @@
 #include "bnxt_cpr.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
+#include "hsi_struct_def_dpdk.h"
 
 /*
  * Async event handling
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index ecabc5f..2be22de 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -34,8 +34,6 @@
 #ifndef _BNXT_CPR_H_
 #define _BNXT_CPR_H_
 
-#include "hsi_struct_def_dpdk.h"
-
 #define CMP_VALID(cmp, raw_cons, ring)					\
 	(!!(((struct cmpl_base *)(cmp))->info3_v & CMPL_BASE_V) ==	\
 	 !((raw_cons) & ((ring)->ring_size)))
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 7e7d1ab..3453509 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -42,6 +42,7 @@
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_stats.h"
 #include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
@@ -187,6 +188,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.stats_get = bnxt_stats_get_op,
+	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 65cd4a7..633683b 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -39,8 +39,11 @@
 #include <rte_version.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
+#include "bnxt_txq.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -480,10 +483,56 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
+{
+	int rc = 0;
+	struct hwrm_stat_ctx_clr_stats_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_clr_stats_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_CLR_STATS, -1, resp);
+
+	if (cpr->hw_stats_ctx_id == (uint32_t)HWRM_NA_SIGNATURE)
+		return rc;
+
+	req.stat_ctx_id = rte_cpu_to_le_16(cpr->hw_stats_ctx_id);
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
 
+int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq;
+		struct bnxt_rx_queue *rxq;
+		struct bnxt_cp_ring_info *cpr;
+
+		if (i >= bp->rx_cp_nr_rings) {
+			txq = bp->tx_queues[i - bp->rx_cp_nr_rings];
+			cpr = txq->cp_ring;
+		} else {
+			rxq = bp->rx_queues[i];
+			cpr = rxq->cp_ring;
+		}
+
+		rc = bnxt_hwrm_stat_clear(bp, cpr);
+		if (rc)
+			return rc;
+	}
+	return 0;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index bc6fb08..75c3442 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -37,7 +37,9 @@
 #include <inttypes.h>
 #include <stdbool.h>
 
-#include "bnxt.h"
+struct bnxt;
+struct bnxt_filter_info;
+struct bnxt_cp_ring_info;
 
 #define HWRM_SEQ_ID_INVALID -1U
 
@@ -56,8 +58,11 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
+int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
+
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index d5c3173..19c0e3f 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -36,6 +36,7 @@
 #include <rte_malloc.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
diff --git a/drivers/net/bnxt/bnxt_stats.c b/drivers/net/bnxt/bnxt_stats.c
new file mode 100644
index 0000000..6f1c760
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_stats.c
@@ -0,0 +1,142 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_byteorder.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
+#include "bnxt_stats.h"
+#include "bnxt_txq.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * Statistics functions
+ */
+
+void bnxt_free_stats(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+
+		bnxt_free_txq_stats(txq);
+	}
+	for (i = 0; i < (int)bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+
+		bnxt_free_rxq_stats(rxq);
+	}
+}
+
+void bnxt_stats_get_op(struct rte_eth_dev *eth_dev,
+			   struct rte_eth_stats *bnxt_stats)
+{
+	unsigned int i;
+	struct bnxt *bp = eth_dev->data->dev_private;
+
+	memset(bnxt_stats, 0, sizeof(*bnxt_stats));
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		struct ctx_hw_stats64 *hw_stats =
+		    (struct ctx_hw_stats64 *)cpr->hw_stats;
+
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_ucast_pkts);
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_mcast_pkts);
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_bcast_pkts);
+
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_ucast_bytes);
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_mcast_bytes);
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_bcast_bytes);
+
+		/*
+		 * TBD: No clear mapping to this... we don't seem
+		 * to have a stat specifically for dropped due to
+		 * insufficient mbufs.
+		 */
+		bnxt_stats->q_errors[i] = 0;
+
+		/* These get replaced once the *_QSTATS commands work */
+		bnxt_stats->ipackets += bnxt_stats->q_ipackets[i];
+		bnxt_stats->ibytes += bnxt_stats->q_ibytes[i];
+		bnxt_stats->imissed += bnxt_stats->q_errors[i];
+		bnxt_stats->ierrors +=
+				rte_le_to_cpu_64(hw_stats->rx_err_pkts);
+	}
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		struct ctx_hw_stats64 *hw_stats =
+		    (struct ctx_hw_stats64 *)cpr->hw_stats;
+
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_ucast_pkts);
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_mcast_pkts);
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_bcast_pkts);
+
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_ucast_bytes);
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_mcast_bytes);
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_bcast_bytes);
+
+		/* These get replaced once the *_QSTATS commands work */
+		bnxt_stats->opackets += bnxt_stats->q_opackets[i];
+		bnxt_stats->obytes +=  bnxt_stats->q_obytes[i];
+		bnxt_stats->oerrors += rte_le_to_cpu_64(hw_stats->tx_drop_pkts);
+		bnxt_stats->oerrors += rte_le_to_cpu_64(hw_stats->tx_err_pkts);
+	}
+}
+
+void bnxt_stats_reset_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	bnxt_clear_all_hwrm_stat_ctxs(bp);
+}
diff --git a/drivers/net/bnxt/bnxt_stats.h b/drivers/net/bnxt/bnxt_stats.h
new file mode 100644
index 0000000..65408a4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_stats.h
@@ -0,0 +1,44 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_STATS_H_
+#define _BNXT_STATS_H_
+
+#include <rte_ethdev.h>
+
+void bnxt_free_stats(struct bnxt *bp);
+void bnxt_stats_get_op(struct rte_eth_dev *eth_dev,
+			   struct rte_eth_stats *bnxt_stats);
+void bnxt_stats_reset_op(struct rte_eth_dev *eth_dev);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 8928984..8793c9d 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -36,6 +36,7 @@
 #include <rte_malloc.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index bbec5bb..91a83d1 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -34,6 +34,35 @@
 #ifndef _HSI_STRUCT_DEF_EXTERNAL_H_
 #define _HSI_STRUCT_DEF_EXTERNAL_H_
 
+/*
+ * per-context HW statistics -- chip view
+ */
+
+struct ctx_hw_stats64 {
+	uint64_t rx_ucast_pkts;
+	uint64_t rx_mcast_pkts;
+	uint64_t rx_bcast_pkts;
+	uint64_t rx_drop_pkts;
+	uint64_t rx_err_pkts;
+	uint64_t rx_ucast_bytes;
+	uint64_t rx_mcast_bytes;
+	uint64_t rx_bcast_bytes;
+
+	uint64_t tx_ucast_pkts;
+	uint64_t tx_mcast_pkts;
+	uint64_t tx_bcast_pkts;
+	uint64_t tx_drop_pkts;
+	uint64_t tx_err_pkts;
+	uint64_t tx_ucast_bytes;
+	uint64_t tx_mcast_bytes;
+	uint64_t tx_bcast_bytes;
+
+	uint64_t tpa_pkts;
+	uint64_t tpa_bytes;
+	uint64_t tpa_events;
+	uint64_t tpa_aborts;
+} ctx_hw_stats64_t;
+
 /* HW Resource Manager Specification 1.2.0 */
 #define HWRM_VERSION_MAJOR	1
 #define HWRM_VERSION_MINOR	2
@@ -63,6 +92,7 @@
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
 /* Return Codes */
@@ -1897,6 +1927,83 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_clr_stats */
+/* Description: This command clears statistics of a context. */
+
+/* Input (24 bytes) */
+struct hwrm_stat_ctx_clr_stats_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* ID of the statistics context that is being queried. */
+	uint32_t stat_ctx_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_stat_ctx_clr_stats_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 13/38] bnxt: add initial Tx code implementation
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (10 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 12/38] bnxt: add statistics operations Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 14/38] bnxt: add initial Rx " Stephen Hurd
                     ` (27 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Initial implementation of tx_pkt_burst for transmit.
bnxt_xmit_pkts() is the top level function that is called during Tx.

bnxt_handle_tx_cp() is used to check and process the Tx completions
generated for the Tx Buffer Descriptors sent by the hardware.

This patch also adds code to allocate rings in the hardware.
For each Tx queue allocated in the PMD driver, a corresponding ring
in hardware will be created. Every time a Tx request is initiated
via the bnxt_xmit_pkts() call, a Buffer Descriptor is created and
is sent to the hardware via the associated Tx ring.

On completing the Tx operation, the hardware will generates the status
in the form of a completion. This completion is processed by the
bnxt_handle_tx_cp() function.

Functions like bnxt_init_tx_ring_struct() and bnxt_init_one_tx_ring()
are used to initialize various members of the structure before
starting Tx operations.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Expand the patch description
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_cpr.h            |   4 +-
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +-
 drivers/net/bnxt/bnxt_ring.c           | 145 ++++++++++
 drivers/net/bnxt/bnxt_ring.h           |   8 +
 drivers/net/bnxt/bnxt_txq.c            |  42 ++-
 drivers/net/bnxt/bnxt_txr.c            | 314 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_txr.h            |  71 +++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 512 +++++++++++++++++++++++++++++++++
 9 files changed, 1091 insertions(+), 9 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_txr.c
 create mode 100644 drivers/net/bnxt/bnxt_txr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index f6a04f8..0785681 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -56,6 +56,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index 2be22de..4bd5606 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -51,11 +51,11 @@
 
 #define B_CP_DB_REARM(cpr, raw_cons)					\
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_REARM_FLAGS | \
-				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
 #define B_CP_DIS_DB(cpr, raw_cons)					\
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
-				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
 struct bnxt_ring;
 struct bnxt_cp_ring_info {
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3453509..4ace543 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -44,6 +44,7 @@
 #include "bnxt_rxq.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -269,7 +270,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	}
 	eth_dev->dev_ops = &bnxt_dev_ops;
 	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
-	/* eth_dev->tx_pkt_burst = &bnxt_xmit_pkts; */
+	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
 
 	rc = bnxt_alloc_hwrm_resources(bp);
 	if (rc) {
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index b57d3ba..2645dda 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -31,8 +31,14 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_memzone.h>
+
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_ring.h"
+#include "bnxt_txr.h"
+
+#include "hsi_struct_def_dpdk.h"
 
 /*
  * Generic ring handling
@@ -45,3 +51,142 @@ void bnxt_free_ring(struct bnxt_ring *ring)
 		*ring->vmem = NULL;
 	}
 }
+
+/*
+ * Allocates a completion ring with vmem and stats optionally also allocating
+ * a TX and/or RX ring.  Passing NULL as tx_ring_info and/or rx_ring_info
+ * to not allocate them.
+ *
+ * Order in the allocation is:
+ * stats - Always non-zero length
+ * cp vmem - Always zero-length, supported for the bnxt_ring abstraction
+ * tx vmem - Only non-zero length if tx_ring_info is not NULL
+ * rx vmem - Only non-zero length if rx_ring_info is not NULL
+ * cp bd ring - Always non-zero length
+ * tx bd ring - Only non-zero length if tx_ring_info is not NULL
+ * rx bd ring - Only non-zero length if rx_ring_info is not NULL
+ */
+int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
+			    struct bnxt_tx_ring_info *tx_ring_info,
+			    struct bnxt_rx_ring_info *rx_ring_info,
+			    struct bnxt_cp_ring_info *cp_ring_info,
+			    const char *suffix)
+{
+	struct bnxt_ring *cp_ring = cp_ring_info->cp_ring_struct;
+	struct bnxt_ring *tx_ring;
+	/* TODO: RX ring */
+	/* struct bnxt_ring *rx_ring; */
+	struct rte_pci_device *pdev = bp->pdev;
+	const struct rte_memzone *mz = NULL;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+
+	int stats_len = (tx_ring_info || rx_ring_info) ?
+	    RTE_CACHE_LINE_ROUNDUP(sizeof(struct ctx_hw_stats64)) : 0;
+
+	int cp_vmem_start = stats_len;
+	int cp_vmem_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->vmem_size);
+
+	int tx_vmem_start = cp_vmem_start + cp_vmem_len;
+	int tx_vmem_len =
+	    tx_ring_info ? RTE_CACHE_LINE_ROUNDUP(tx_ring_info->
+						tx_ring_struct->vmem_size) : 0;
+
+	int rx_vmem_start = tx_vmem_start + tx_vmem_len;
+	/* TODO: RX ring */
+	int rx_vmem_len = 0;
+	/*
+	 * rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
+	 * rx_ring_struct->vmem_size) : 0;
+	 */
+
+	int cp_ring_start = rx_vmem_start + rx_vmem_len;
+	int cp_ring_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->ring_size *
+						 sizeof(struct cmpl_base));
+
+	int tx_ring_start = cp_ring_start + cp_ring_len;
+	int tx_ring_len = tx_ring_info ?
+	    RTE_CACHE_LINE_ROUNDUP(tx_ring_info->tx_ring_struct->ring_size *
+				   sizeof(struct tx_bd_long)) : 0;
+
+	int rx_ring_start = tx_ring_start + tx_ring_len;
+	/* TODO: RX ring */
+	int rx_ring_len = 0;
+	/*
+	 * rx_ring_info ?
+	 * RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
+	 * sizeof(struct rx_prod_pkt_bd)) : 0;
+	 */
+
+	int total_alloc_len = rx_ring_start + rx_ring_len;
+
+	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
+		 "bnxt_%04x:%02x:%02x:%02x-%04x_%s", pdev->addr.domain,
+		 pdev->addr.bus, pdev->addr.devid, pdev->addr.function, qidx,
+		 suffix);
+	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
+	mz = rte_memzone_lookup(mz_name);
+	if (!mz) {
+		mz = rte_memzone_reserve(mz_name, total_alloc_len,
+					 SOCKET_ID_ANY,
+					 RTE_MEMZONE_2MB |
+					 RTE_MEMZONE_SIZE_HINT_ONLY);
+		if (mz == NULL)
+			return -ENOMEM;
+	}
+	memset(mz->addr, 0, mz->len);
+
+	if (tx_ring_info) {
+		tx_ring = tx_ring_info->tx_ring_struct;
+
+		tx_ring->bd = ((char *)mz->addr + tx_ring_start);
+		tx_ring_info->tx_desc_ring = (struct tx_bd_long *)tx_ring->bd;
+		tx_ring->bd_dma = mz->phys_addr + tx_ring_start;
+		tx_ring_info->tx_desc_mapping = tx_ring->bd_dma;
+
+		if (!tx_ring->bd)
+			return -ENOMEM;
+		if (tx_ring->vmem_size) {
+			tx_ring->vmem =
+			    (void **)((char *)mz->addr + tx_vmem_start);
+			tx_ring_info->tx_buf_ring =
+			    (struct bnxt_sw_tx_bd *)tx_ring->vmem;
+		}
+	}
+
+/*
+ *	if (rx_ring_info) {
+ *		rx_ring = &rx_ring_info->rx_ring_struct;
+ *
+ *		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
+ *		rx_ring_info->rx_desc_ring =
+ *		    (struct rx_prod_pkt_bd *)rx_ring->bd;
+ *		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
+ *		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+ *
+ *		if (!rx_ring->bd)
+ *			return -ENOMEM;
+ *		if (rx_ring->vmem_size) {
+ *			rx_ring->vmem =
+ *			    (void **)((char *)mz->addr + rx_vmem_start);
+ *			rx_ring_info->rx_buf_ring =
+ *			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
+ *		}
+ *	}
+ */
+
+	cp_ring->bd = ((char *)mz->addr + cp_ring_start);
+	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
+	cp_ring_info->cp_desc_ring = cp_ring->bd;
+	cp_ring_info->cp_desc_mapping = cp_ring->bd_dma;
+
+	if (!cp_ring->bd)
+		return -ENOMEM;
+	if (cp_ring->vmem_size)
+		*cp_ring->vmem = ((char *)mz->addr + stats_len);
+	if (stats_len) {
+		cp_ring_info->hw_stats = mz->addr;
+		cp_ring_info->hw_stats_map = mz->phys_addr;
+	}
+	cp_ring_info->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index 3abfb04..1e9d3be 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -87,6 +87,14 @@ struct bnxt_ring_grp_info {
 };
 
 struct bnxt;
+struct bnxt_tx_ring_info;
+struct bnxt_rx_ring_info;
+struct bnxt_cp_ring_info;
 void bnxt_free_ring(struct bnxt_ring *ring);
+int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
+			    struct bnxt_tx_ring_info *tx_ring_info,
+			    struct bnxt_rx_ring_info *rx_ring_info,
+			    struct bnxt_cp_ring_info *cp_ring_info,
+			    const char *suffix);
 
 #endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 8793c9d..460f71a 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -39,6 +39,7 @@
 #include "bnxt_cpr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 
 /*
  * TX Queues
@@ -54,9 +55,20 @@ void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
 		cpr->hw_stats = NULL;
 }
 
-static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq __rte_unused)
+static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq)
 {
-	/* TODO: Requires interaction with TX ring */
+	struct bnxt_sw_tx_bd *sw_ring;
+	uint16_t i;
+
+	sw_ring = txq->tx_ring->tx_buf_ring;
+	if (sw_ring) {
+		for (i = 0; i < txq->tx_ring->tx_ring_struct->ring_size; i++) {
+			if (sw_ring[i].mbuf) {
+				rte_pktmbuf_free(sw_ring[i].mbuf);
+				sw_ring[i].mbuf = NULL;
+			}
+		}
+	}
 }
 
 void bnxt_free_tx_mbufs(struct bnxt *bp)
@@ -75,7 +87,15 @@ void bnxt_tx_queue_release_op(void *tx_queue)
 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
 
 	if (txq) {
-		/* TODO: Free ring and stats here */
+		/* Free TX ring hardware descriptors */
+		bnxt_tx_queue_release_mbufs(txq);
+		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
+
+		/* Free TX completion ring hardware descriptors */
+		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
+
+		bnxt_free_txq_stats(txq);
+
 		rte_free(txq);
 	}
 }
@@ -111,14 +131,24 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	txq->nb_tx_desc = nb_desc;
 	txq->tx_free_thresh = tx_conf->tx_free_thresh;
 
-	/* TODO: Initialize ring structure */
+	bnxt_init_tx_ring_struct(txq);
 
 	txq->queue_id = queue_idx;
 	txq->port_id = eth_dev->data->port_id;
 
-	/* TODO: Allocate TX ring hardware descriptors */
+	/* Allocate TX ring hardware descriptors */
+	if (bnxt_alloc_rings(bp, queue_idx, txq->tx_ring, NULL, txq->cp_ring,
+			"txr")) {
+		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for tx_ring failed!");
+		bnxt_tx_queue_release_op(txq);
+		return -ENOMEM;
+	}
 
-	/* TODO: Initialize the ring */
+	if (bnxt_init_one_tx_ring(txq)) {
+		RTE_LOG(ERR, PMD, "bnxt_init_one_tx_ring failed!");
+		bnxt_tx_queue_release_op(txq);
+		return -ENOMEM;
+	}
 
 	eth_dev->data->tx_queues[queue_idx] = txq;
 	return 0;
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
new file mode 100644
index 0000000..9064e8d
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -0,0 +1,314 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_byteorder.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_ring.h"
+#include "bnxt_txq.h"
+#include "bnxt_txr.h"
+#include "hsi_struct_def_dpdk.h"
+#include <stdbool.h>
+
+/*
+ * TX Ring handling
+ */
+
+void bnxt_free_tx_rings(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->tx_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+
+		if (!txq)
+			continue;
+
+		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
+		/* TODO: free() txq->tx_ring and txq->tx_ring->tx_ring_struct */
+		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
+		/* TODO: free() txq->cp_ring and txq->cp_ring->cp_ring_struct */
+
+		rte_free(txq);
+		bp->tx_queues[i] = NULL;
+	}
+}
+
+int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	struct bnxt_ring *ring = txr->tx_ring_struct;
+
+	txq->tx_wake_thresh = ring->ring_size / 2;
+	ring->fw_ring_id = INVALID_HW_RING_ID;
+
+	return 0;
+}
+
+void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_tx_ring_info *txr;
+	struct bnxt_ring *ring;
+
+	/* TODO: These need to be allocated */
+	txr = txq->tx_ring;
+	ring = txr->tx_ring_struct;
+	ring->ring_size = rte_align32pow2(txq->nb_tx_desc + 1);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)txr->tx_desc_ring;
+	ring->bd_dma = txr->tx_desc_mapping;
+	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_tx_bd);
+	ring->vmem = (void **)&txr->tx_buf_ring;
+
+	/* TODO: These need to be allocated */
+	cpr = txq->cp_ring;
+	ring = cpr->cp_ring_struct;
+	ring->ring_size = txr->tx_ring_struct->ring_size;
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
+
+static inline uint32_t bnxt_tx_avail(struct bnxt_tx_ring_info *txr)
+{
+	/* Tell compiler to fetch tx indices from memory. */
+	rte_compiler_barrier();
+
+	return txr->tx_ring_struct->ring_size -
+		((txr->tx_prod - txr->tx_cons) &
+			txr->tx_ring_struct->ring_mask) - 1;
+}
+
+static uint16_t bnxt_start_xmit(struct rte_mbuf *tx_pkt,
+				struct bnxt_tx_queue *txq)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	struct tx_bd_long *txbd;
+	struct tx_bd_long_hi *txbd1;
+	uint32_t vlan_tag_flags, cfa_action;
+	bool long_bd = false;
+	uint16_t last_prod = 0;
+	struct rte_mbuf *m_seg;
+	struct bnxt_sw_tx_bd *tx_buf;
+	static const uint32_t lhint_arr[4] = {
+		TX_BD_LONG_FLAGS_LHINT_LT512,
+		TX_BD_LONG_FLAGS_LHINT_LT1K,
+		TX_BD_LONG_FLAGS_LHINT_LT2K,
+		TX_BD_LONG_FLAGS_LHINT_LT2K
+	};
+
+	if (tx_pkt->ol_flags & (PKT_TX_TCP_SEG | PKT_TX_TCP_CKSUM |
+				PKT_TX_UDP_CKSUM | PKT_TX_IP_CKSUM |
+				PKT_TX_VLAN_PKT))
+		long_bd = true;
+
+	tx_buf = &txr->tx_buf_ring[txr->tx_prod];
+	tx_buf->mbuf = tx_pkt;
+	tx_buf->nr_bds = long_bd + tx_pkt->nb_segs;
+	last_prod = (txr->tx_prod + tx_buf->nr_bds - 1) &
+				txr->tx_ring_struct->ring_mask;
+
+	if (unlikely(bnxt_tx_avail(txr) < tx_buf->nr_bds))
+		return -ENOMEM;
+
+	txbd = &txr->tx_desc_ring[txr->tx_prod];
+	txbd->opaque = txr->tx_prod;
+	txbd->flags_type = tx_buf->nr_bds << TX_BD_LONG_FLAGS_BD_CNT_SFT;
+	txbd->len = tx_pkt->data_len;
+	if (txbd->len >= 2014)
+		txbd->flags_type |= TX_BD_LONG_FLAGS_LHINT_GTE2K;
+	else
+		txbd->flags_type |= lhint_arr[txbd->len >> 9];
+	txbd->addr = rte_cpu_to_le_32(RTE_MBUF_DATA_DMA_ADDR(tx_buf->mbuf));
+
+	if (long_bd) {
+		txbd->flags_type |= TX_BD_LONG_TYPE_TX_BD_LONG;
+		vlan_tag_flags = 0;
+		cfa_action = 0;
+		if (tx_buf->mbuf->ol_flags & PKT_TX_VLAN_PKT) {
+			/* shurd: Should this mask at
+			 * TX_BD_LONG_CFA_META_VLAN_VID_MASK?
+			 */
+			vlan_tag_flags = TX_BD_LONG_CFA_META_KEY_VLAN_TAG |
+				tx_buf->mbuf->vlan_tci;
+			/* Currently supports 8021Q, 8021AD vlan offloads
+			 * QINQ1, QINQ2, QINQ3 vlan headers are deprecated
+			 */
+			/* DPDK only supports 802.11q VLAN packets */
+			vlan_tag_flags |=
+					TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
+		}
+
+		txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+
+		txbd1 = (struct tx_bd_long_hi *)
+					&txr->tx_desc_ring[txr->tx_prod];
+		txbd1->lflags = 0;
+		txbd1->cfa_meta = vlan_tag_flags;
+		txbd1->cfa_action = cfa_action;
+
+		if (tx_pkt->ol_flags & PKT_TX_TCP_SEG) {
+			/* TSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_LSO;
+			txbd1->hdr_size = tx_pkt->l2_len + tx_pkt->l3_len +
+					tx_pkt->l4_len;
+			txbd1->mss = tx_pkt->tso_segsz;
+
+		} else if (tx_pkt->ol_flags & (PKT_TX_TCP_CKSUM |
+					PKT_TX_UDP_CKSUM)) {
+			/* TCP/UDP CSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_TCP_UDP_CHKSUM;
+			txbd1->mss = 0;
+
+		} else if (tx_pkt->ol_flags & PKT_TX_IP_CKSUM) {
+			/* IP CSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_IP_CHKSUM;
+			txbd1->mss = 0;
+		}
+	} else {
+		txbd->flags_type |= TX_BD_SHORT_TYPE_TX_BD_SHORT;
+	}
+
+	m_seg = tx_pkt->next;
+	/* i is set at the end of the if(long_bd) block */
+	while (txr->tx_prod != last_prod) {
+		txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+		tx_buf = &txr->tx_buf_ring[txr->tx_prod];
+
+		txbd = &txr->tx_desc_ring[txr->tx_prod];
+		txbd->addr = rte_cpu_to_le_32(RTE_MBUF_DATA_DMA_ADDR(m_seg));
+		txbd->flags_type = TX_BD_SHORT_TYPE_TX_BD_SHORT;
+		txbd->len = m_seg->data_len;
+
+		m_seg = m_seg->next;
+	}
+
+	txbd->flags_type |= TX_BD_LONG_FLAGS_PACKET_END;
+
+	txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+
+	return 0;
+}
+
+static void bnxt_tx_cmp(struct bnxt_tx_queue *txq, int nr_pkts)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	uint16_t cons = txr->tx_cons;
+	int i, j;
+
+	for (i = 0; i < nr_pkts; i++) {
+		struct bnxt_sw_tx_bd *tx_buf;
+		struct rte_mbuf *mbuf;
+
+		tx_buf = &txr->tx_buf_ring[cons];
+		cons = RING_NEXT(txr->tx_ring_struct, cons);
+		mbuf = tx_buf->mbuf;
+		tx_buf->mbuf = NULL;
+
+		/* EW - no need to unmap DMA memory? */
+
+		for (j = 1; j < tx_buf->nr_bds; j++)
+			cons = RING_NEXT(txr->tx_ring_struct, cons);
+		rte_pktmbuf_free(mbuf);
+	}
+
+	txr->tx_cons = cons;
+}
+
+static int bnxt_handle_tx_cp(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+	uint32_t raw_cons = cpr->cp_raw_cons;
+	uint32_t cons;
+	int nb_tx_pkts = 0;
+	struct tx_cmpl *txcmp;
+
+	if ((txq->tx_ring->tx_ring_struct->ring_size -
+			(bnxt_tx_avail(txq->tx_ring))) >
+			txq->tx_free_thresh) {
+		while (1) {
+			cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
+			txcmp = (struct tx_cmpl *)&cpr->cp_desc_ring[cons];
+
+			if (!CMP_VALID(txcmp, raw_cons, cpr->cp_ring_struct))
+				break;
+
+			if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
+				nb_tx_pkts++;
+			else
+				RTE_LOG(DEBUG, PMD,
+						"Unhandled CMP type %02x\n",
+						CMP_TYPE(txcmp));
+			raw_cons = NEXT_RAW_CMP(raw_cons);
+		}
+		if (nb_tx_pkts)
+			bnxt_tx_cmp(txq, nb_tx_pkts);
+		cpr->cp_raw_cons = raw_cons;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+	}
+	return nb_tx_pkts;
+}
+
+uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct bnxt_tx_queue *txq = tx_queue;
+	uint16_t nb_tx_pkts = 0;
+	uint16_t db_mask = txq->tx_ring->tx_ring_struct->ring_size >> 2;
+	uint16_t last_db_mask = 0;
+
+	/* Handle TX completions */
+	bnxt_handle_tx_cp(txq);
+
+	/* Handle TX burst request */
+	for (nb_tx_pkts = 0; nb_tx_pkts < nb_pkts; nb_tx_pkts++) {
+		if (bnxt_start_xmit(tx_pkts[nb_tx_pkts], txq)) {
+			break;
+		} else if ((nb_tx_pkts & db_mask) != last_db_mask) {
+			B_TX_DB(txq->tx_ring->tx_doorbell,
+					txq->tx_ring->tx_prod);
+			last_db_mask = nb_tx_pkts & db_mask;
+		}
+	}
+	if (nb_tx_pkts)
+		B_TX_DB(txq->tx_ring->tx_doorbell, txq->tx_ring->tx_prod);
+
+	return nb_tx_pkts;
+}
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
new file mode 100644
index 0000000..712ce65
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -0,0 +1,71 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_TXR_H_
+#define _BNXT_TXR_H_
+
+#define MAX_TX_RINGS	16
+#define BNXT_TX_PUSH_THRESH 92
+
+#define B_TX_DB(db, prod)						\
+		(*(uint32_t *)db = (DB_KEY_TX | prod))
+
+struct bnxt_tx_ring_info {
+	uint16_t		tx_prod;
+	uint16_t		tx_cons;
+	void			*tx_doorbell;
+
+	struct tx_bd_long	*tx_desc_ring;
+	struct bnxt_sw_tx_bd	*tx_buf_ring;
+
+	phys_addr_t		tx_desc_mapping;
+
+#define BNXT_DEV_STATE_CLOSING	0x1
+	uint32_t		dev_state;
+
+	struct bnxt_ring	*tx_ring_struct;
+};
+
+struct bnxt_sw_tx_bd {
+	struct rte_mbuf		*mbuf; /* mbuf associated with TX descriptor */
+	uint8_t			is_gso;
+	unsigned short		nr_bds;
+};
+
+void bnxt_free_tx_rings(struct bnxt *bp);
+int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
+void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq);
+uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 91a83d1..1543f20 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -99,6 +99,518 @@ struct ctx_hw_stats64 {
 #define HWRM_ERR_CODE_INVALID_PARAMS                      (UINT32_C(0x2))
 #define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED              (UINT32_C(0x3))
 
+/* Short TX BD (16 bytes) */
+struct tx_bd_short {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	/* This value identifies the type of buffer descriptor. */
+	#define TX_BD_SHORT_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_BD_SHORT_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 16B long and is used for normal L2
+		 * packet transmission.
+		 */
+	#define TX_BD_SHORT_TYPE_TX_BD_SHORT		(UINT32_C(0x0) << 0)
+	/*
+	 * If set to 1, the packet ends with the data in the buffer pointed to
+	 * by this descriptor. This flag must be valid on every BD.
+	 */
+	#define TX_BD_SHORT_FLAGS_PACKET_END		UINT32_C(0x40)
+	/*
+	 * If set to 1, the device will not generate a completion for this
+	 * transmit packet unless there is an error in it's processing. If this
+	 * bit is set to 0, then the packet will be completed normally. This bit
+	 * must be valid only on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_NO_CMPL		UINT32_C(0x80)
+	/*
+	 * This value indicates how many 16B BD locations are consumed in the
+	 * ring by this packet. A value of 1 indicates that this BD is the only
+	 * BD (and that the it is a short BD). A value of 3 indicates either 3
+	 * short BDs or 1 long BD and one short BD in the packet. A value of 0
+	 * indicates that there are 32 BD locations in the packet (the maximum).
+	 * This field is valid only on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_BD_CNT_MASK		UINT32_C(0x1f00)
+	#define TX_BD_SHORT_FLAGS_BD_CNT_SFT		8
+	/*
+	 * This value is a hint for the length of the entire packet. It is used
+	 * by the chip to optimize internal processing. The packet will be
+	 * dropped if the hint is too short. This field is valid only on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_LHINT_MASK		UINT32_C(0x6000)
+	#define TX_BD_SHORT_FLAGS_LHINT_SFT		13
+		/* indicates packet length < 512B */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT512		(UINT32_C(0x0) << 13)
+		/* indicates 512 <= packet length < 1KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT1K		(UINT32_C(0x1) << 13)
+		/* indicates 1KB <= packet length < 2KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT2K		(UINT32_C(0x2) << 13)
+		/* indicates packet length >= 2KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_GTE2K		(UINT32_C(0x3) << 13)
+	#define TX_BD_SHORT_FLAGS_LHINT_LAST	TX_BD_SHORT_FLAGS_LHINT_GTE2K
+	/*
+	 * If set to 1, the device immediately updates the Send Consumer Index
+	 * after the buffer associated with this descriptor has been transferred
+	 * via DMA to NIC memory from host memory. An interrupt may or may not
+	 * be generated according to the state of the interrupt avoidance
+	 * mechanisms. If this bit is set to 0, then the Consumer Index is only
+	 * updated as soon as one of the host interrupt coalescing conditions
+	 * has been met. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_COAL_NOW		UINT32_C(0x8000)
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_BD_SHORT_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the host physical buffer this BD describes in
+	 * bytes. This field must be valid on all BDs of a packet.
+	 */
+	uint16_t len;
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with the
+	 * transmit BD. This field must be valid on the first BD of a packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address for the portion of the packet
+	 * described by this TX BD. This value must be valid on all BDs of a
+	 * packet.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
+/* Long TX BD (32 bytes split to 2 16-byte struct) */
+struct tx_bd_long {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	/* This value identifies the type of buffer descriptor. */
+	#define TX_BD_LONG_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_BD_LONG_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 32B long and is used for normal L2
+		 * packet transmission.
+		 */
+	#define TX_BD_LONG_TYPE_TX_BD_LONG		(UINT32_C(0x10) << 0)
+	/*
+	 * If set to 1, the packet ends with the data in the buffer pointed to
+	 * by this descriptor. This flag must be valid on every BD.
+	 */
+	#define TX_BD_LONG_FLAGS_PACKET_END		UINT32_C(0x40)
+	/*
+	 * If set to 1, the device will not generate a completion for this
+	 * transmit packet unless there is an error in it's processing. If this
+	 * bit is set to 0, then the packet will be completed normally. This bit
+	 * must be valid only on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_NO_CMPL		UINT32_C(0x80)
+	/*
+	 * This value indicates how many 16B BD locations are consumed in the
+	 * ring by this packet. A value of 1 indicates that this BD is the only
+	 * BD (and that the it is a short BD). A value of 3 indicates either 3
+	 * short BDs or 1 long BD and one short BD in the packet. A value of 0
+	 * indicates that there are 32 BD locations in the packet (the maximum).
+	 * This field is valid only on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_BD_CNT_MASK		UINT32_C(0x1f00)
+	#define TX_BD_LONG_FLAGS_BD_CNT_SFT		8
+	/*
+	 * This value is a hint for the length of the entire packet. It is used
+	 * by the chip to optimize internal processing. The packet will be
+	 * dropped if the hint is too short. This field is valid only on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_LHINT_MASK		UINT32_C(0x6000)
+	#define TX_BD_LONG_FLAGS_LHINT_SFT		13
+		/* indicates packet length < 512B */
+	#define TX_BD_LONG_FLAGS_LHINT_LT512		(UINT32_C(0x0) << 13)
+		/* indicates 512 <= packet length < 1KB */
+	#define TX_BD_LONG_FLAGS_LHINT_LT1K		(UINT32_C(0x1) << 13)
+		/* indicates 1KB <= packet length < 2KB */
+	#define TX_BD_LONG_FLAGS_LHINT_LT2K		(UINT32_C(0x2) << 13)
+		/* indicates packet length >= 2KB */
+	#define TX_BD_LONG_FLAGS_LHINT_GTE2K		(UINT32_C(0x3) << 13)
+	#define TX_BD_LONG_FLAGS_LHINT_LAST	TX_BD_LONG_FLAGS_LHINT_GTE2K
+	/*
+	 * If set to 1, the device immediately updates the Send Consumer Index
+	 * after the buffer associated with this descriptor has been transferred
+	 * via DMA to NIC memory from host memory. An interrupt may or may not
+	 * be generated according to the state of the interrupt avoidance
+	 * mechanisms. If this bit is set to 0, then the Consumer Index is only
+	 * updated as soon as one of the host interrupt coalescing conditions
+	 * has been met. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_COAL_NOW		UINT32_C(0x8000)
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	#define TX_BD_LONG_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_BD_LONG_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the host physical buffer this BD describes in
+	 * bytes. This field must be valid on all BDs of a packet.
+	 */
+	uint16_t len;
+
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with the
+	 * transmit BD. This field must be valid on the first BD of a packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address for the portion of the packet
+	 * described by this TX BD. This value must be valid on all BDs of a
+	 * packet.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
+/* last 16 bytes of Long TX BD */
+
+struct tx_bd_long_hi {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Their value on other BDs of the packet will be ignored.
+	 */
+	/*
+	 * If set to 1, the controller replaces the TCP/UPD checksum fields of
+	 * normal TCP/UPD checksum, or the inner TCP/UDP checksum field of the
+	 * encapsulated TCP/UDP packets with the hardware calculated TCP/UDP
+	 * checksum for the packet associated with this descriptor. This bit
+	 * must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_TCP_UDP_CHKSUM	UINT32_C(0x1)
+	/*
+	 * If set to 1, the controller replaces the IP checksum of the normal
+	 * packets, or the inner IP checksum of the encapsulated packets with
+	 * the hardware calculated IP checksum for the packet associated with
+	 * this descriptor. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_IP_CHKSUM		UINT32_C(0x2)
+	/*
+	 * If set to 1, the controller will not append an Ethernet CRC to the
+	 * end of the frame. This bit must be valid on the first BD of a packet.
+	 * Packet must be 64B or longer when this flag is set. It is not useful
+	 * to use this bit with any form of TX offload such as CSO or LSO. The
+	 * intent is that the packet from the host already has a valid Ethernet
+	 * CRC on the packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_NOCRC			UINT32_C(0x4)
+	/*
+	 * If set to 1, the device will record the time at which the packet was
+	 * actually transmitted at the TX MAC. This bit must be valid on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_STAMP			UINT32_C(0x8)
+	/*
+	 * If set to 1, The controller replaces the tunnel IP checksum field
+	 * with hardware calculated IP checksum for the IP header of the packet
+	 * associated with this descriptor. In case of VXLAN, the controller
+	 * also replaces the outer header UDP checksum with hardware calculated
+	 * UDP checksum for the packet associated with this descriptor.
+	 */
+	#define TX_BD_LONG_LFLAGS_T_IP_CHKSUM		UINT32_C(0x10)
+	/*
+	 * If set to 1, the device will treat this packet with LSO(Large Send
+	 * Offload) processing for both normal or encapsulated packets, which is
+	 * a form of TCP segmentation. When this bit is 1, the hdr_size and mss
+	 * fields must be valid. The driver doesn't need to set t_ip_chksum,
+	 * ip_chksum, and tcp_udp_chksum flags since the controller will replace
+	 * the appropriate checksum fields for segmented packets. When this bit
+	 * is 1, the hdr_size and mss fields must be valid.
+	 */
+	#define TX_BD_LONG_LFLAGS_LSO			UINT32_C(0x20)
+	/*
+	 * If set to zero when LSO is '1', then the IPID will be treated as a
+	 * 16b number and will be wrapped if it exceeds a value of 0xffff. If
+	 * set to one when LSO is '1', then the IPID will be treated as a 15b
+	 * number and will be wrapped if it exceeds a value 0f 0x7fff.
+	 */
+	#define TX_BD_LONG_LFLAGS_IPID_FMT		UINT32_C(0x40)
+	/*
+	 * If set to zero when LSO is '1', then the IPID of the tunnel IP header
+	 * will not be modified during LSO operations. If set to one when LSO is
+	 * '1', then the IPID of the tunnel IP header will be incremented for
+	 * each subsequent segment of an LSO operation.
+	 */
+	#define TX_BD_LONG_LFLAGS_T_IPID		UINT32_C(0x80)
+	/*
+	 * If set to '1', then the RoCE ICRC will be appended to the packet.
+	 * Packet must be a valid RoCE format packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_ROCE_CRC		UINT32_C(0x100)
+	/*
+	 * If set to '1', then the FCoE CRC will be appended to the packet.
+	 * Packet must be a valid FCoE format packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_FCOE_CRC		UINT32_C(0x200)
+	uint16_t lflags;
+
+	/*
+	 * When LSO is '1', this field must contain the offset of the TCP
+	 * payload from the beginning of the packet in as 16b words. In case of
+	 * encapsulated/tunneling packet, this field contains the offset of the
+	 * inner TCP payload from beginning of the packet as 16-bit words. This
+	 * value must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_HDR_SIZE_MASK		UINT32_C(0x1ff)
+	#define TX_BD_LONG_HDR_SIZE_SFT			0
+	uint16_t hdr_size;
+
+	/*
+	 * This is the MSS value that will be used to do the LSO processing. The
+	 * value is the length in bytes of the TCP payload for each segment
+	 * generated by the LSO operation. This value must be valid on the first
+	 * BD of a packet.
+	 */
+	#define TX_BD_LONG_MSS_MASK			UINT32_C(0x7fff)
+	#define TX_BD_LONG_MSS_SFT			0
+	uint32_t mss;
+
+	uint16_t unused_2;
+
+	/*
+	 * This value selects a CFA action to perform on the packet. Set this
+	 * value to zero if no CFA action is desired. This value must be valid
+	 * on the first BD of a packet.
+	 */
+	uint16_t cfa_action;
+
+	/*
+	 * This value is action meta-data that defines CFA edit operations that
+	 * are done in addition to any action editing.
+	 */
+	/* When key=1, This is the VLAN tag VID value. */
+	#define TX_BD_LONG_CFA_META_VLAN_VID_MASK	UINT32_C(0xfff)
+	#define TX_BD_LONG_CFA_META_VLAN_VID_SFT	0
+	/* When key=1, This is the VLAN tag DE value. */
+	#define TX_BD_LONG_CFA_META_VLAN_DE		UINT32_C(0x1000)
+	/* When key=1, This is the VLAN tag PRI value. */
+	#define TX_BD_LONG_CFA_META_VLAN_PRI_MASK	UINT32_C(0xe000)
+	#define TX_BD_LONG_CFA_META_VLAN_PRI_SFT	13
+	/* When key=1, This is the VLAN tag TPID select value. */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_MASK	UINT32_C(0x70000)
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_SFT	16
+		/* 0x88a8 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8	(UINT32_C(0x0) << 16)
+		/* 0x8100 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100	(UINT32_C(0x1) << 16)
+		/* 0x9100 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100	(UINT32_C(0x2) << 16)
+		/* 0x9200 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200	(UINT32_C(0x3) << 16)
+		/* 0x9300 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300	(UINT32_C(0x4) << 16)
+		/* Value programmed in CFA VLANTPID register. */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPIDCFG	(UINT32_C(0x5) << 16)
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_LAST \
+					TX_BD_LONG_CFA_META_VLAN_TPID_TPIDCFG
+	/* When key=1, This is the VLAN tag TPID select value. */
+	#define TX_BD_LONG_CFA_META_VLAN_RESERVED_MASK	UINT32_C(0xff80000)
+	#define TX_BD_LONG_CFA_META_VLAN_RESERVED_SFT	19
+	/*
+	 * This field identifies the type of edit to be performed on the packet.
+	 * This value must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_CFA_META_KEY_MASK		UINT32_C(0xf0000000)
+	#define TX_BD_LONG_CFA_META_KEY_SFT		28
+		/* No editing */
+	#define TX_BD_LONG_CFA_META_KEY_NONE		(UINT32_C(0x0) << 28)
+		/*
+		 * - meta[17:16] - TPID select value (0 = 0x8100). - meta[15:12]
+		 * - PRI/DE value. - meta[11:0] - VID value.
+		 */
+	#define TX_BD_LONG_CFA_META_KEY_VLAN_TAG	(UINT32_C(0x1) << 28)
+	#define TX_BD_LONG_CFA_META_KEY_LAST	TX_BD_LONG_CFA_META_KEY_VLAN_TAG
+	uint32_t cfa_meta;
+} __attribute__((packed));
+
+/* Completion Ring Structures */
+/* Note: This structure is used by the HWRM to communicate HWRM Error. */
+/* Base Completion Record (16 bytes) */
+struct cmpl_base {
+	/* unused is 10 b */
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define CMPL_BASE_TYPE_MASK			UINT32_C(0x3f)
+	#define CMPL_BASE_TYPE_SFT			0
+		/* TX L2 completion: Completion of TX packet. Length = 16B */
+	#define CMPL_BASE_TYPE_TX_L2			(UINT32_C(0x0) << 0)
+		/*
+		 * RX L2 completion: Completion of and L2 RX packet.
+		 * Length = 32B
+		*/
+	#define CMPL_BASE_TYPE_RX_L2			(UINT32_C(0x11) << 0)
+		/*
+		 * RX Aggregation Buffer completion : Completion of an L2
+		 * aggregation buffer in support of TPA, HDS, or Jumbo packet
+		 * completion. Length = 16B
+		 */
+	#define CMPL_BASE_TYPE_RX_AGG			(UINT32_C(0x12) << 0)
+		/*
+		 * RX L2 TPA Start Completion: Completion at the beginning of a
+		 * TPA operation. Length = 32B
+		 */
+	#define CMPL_BASE_TYPE_RX_TPA_START		(UINT32_C(0x13) << 0)
+		/*
+		 * RX L2 TPA End Completion: Completion at the end of a TPA
+		 * operation. Length = 32B
+		 */
+	#define CMPL_BASE_TYPE_RX_TPA_END		(UINT32_C(0x15) << 0)
+		/*
+		 * Statistics Ejection Completion: Completion of statistics data
+		 * ejection buffer. Length = 16B
+		 */
+	#define CMPL_BASE_TYPE_STAT_EJECT		(UINT32_C(0x1a) << 0)
+		/* HWRM Command Completion: Completion of an HWRM command. */
+	#define CMPL_BASE_TYPE_HWRM_DONE		(UINT32_C(0x20) << 0)
+		/* Forwarded HWRM Request */
+	#define CMPL_BASE_TYPE_HWRM_FWD_REQ		(UINT32_C(0x22) << 0)
+		/* Forwarded HWRM Response */
+	#define CMPL_BASE_TYPE_HWRM_FWD_RESP		(UINT32_C(0x24) << 0)
+		/* HWRM Asynchronous Event Information */
+	#define CMPL_BASE_TYPE_HWRM_ASYNC_EVENT		(UINT32_C(0x2e) << 0)
+		/* CQ Notification */
+	#define CMPL_BASE_TYPE_CQ_NOTIFICATION		(UINT32_C(0x30) << 0)
+		/* SRQ Threshold Event */
+	#define CMPL_BASE_TYPE_SRQ_EVENT		(UINT32_C(0x32) << 0)
+		/* DBQ Threshold Event */
+	#define CMPL_BASE_TYPE_DBQ_EVENT		(UINT32_C(0x34) << 0)
+		/* QP Async Notification */
+	#define CMPL_BASE_TYPE_QP_EVENT			(UINT32_C(0x38) << 0)
+		/* Function Async Notification */
+	#define CMPL_BASE_TYPE_FUNC_EVENT		(UINT32_C(0x3a) << 0)
+	uint16_t type;
+
+	uint16_t info1;
+	uint32_t info2;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define CMPL_BASE_V				UINT32_C(0x1)
+	/* info3 is 31 b */
+	#define CMPL_BASE_INFO3_MASK			UINT32_C(0xfffffffe)
+	#define CMPL_BASE_INFO3_SFT			1
+	uint32_t info3_v;
+
+	uint32_t info4;
+} __attribute__((packed));
+
+/* TX Completion Record (16 bytes) */
+struct tx_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define TX_CMPL_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_CMPL_TYPE_SFT			0
+		/* TX L2 completion: Completion of TX packet. Length = 16B */
+	#define TX_CMPL_TYPE_TX_L2			(UINT32_C(0x0) << 0)
+	/*
+	 * When this bit is '1', it indicates a packet that has an error of some
+	 * type. Type of error is indicated in error_flags.
+	 */
+	#define TX_CMPL_FLAGS_ERROR			UINT32_C(0x40)
+	/*
+	 * When this bit is '1', it indicates that the packet completed was
+	 * transmitted using the push acceleration data provided by the driver.
+	 * When this bit is '0', it indicates that the packet had not push
+	 * acceleration data written or was executed as a normal packet even
+	 * though push data was provided.
+	 */
+	#define TX_CMPL_FLAGS_PUSH			UINT32_C(0x80)
+	#define TX_CMPL_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_CMPL_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	uint16_t unused_0;
+
+	/*
+	 * This is a copy of the opaque field from the first TX BD of this
+	 * transmitted packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define TX_CMPL_V				UINT32_C(0x1)
+	/*
+	 * This error indicates that there was some sort of problem with the BDs
+	 * for the packet.
+	 */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_MASK	UINT32_C(0xe)
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_SFT		1
+		/* No error */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_NO_ERROR	(UINT32_C(0x0) << 1)
+		/* Bad Format: BDs were not formatted correctly. */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_BAD_FMT	(UINT32_C(0x2) << 1)
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_LAST \
+					TX_CMPL_ERRORS_BUFFER_ERROR_BAD_FMT
+	/*
+	 * When this bit is '1', it indicates that the length of the packet was
+	 * zero. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_ZERO_LENGTH_PKT		UINT32_C(0x10)
+	/*
+	 * When this bit is '1', it indicates that the packet was longer than
+	 * the programmed limit in TDI. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_EXCESSIVE_BD_LENGTH	UINT32_C(0x20)
+	/*
+	 * When this bit is '1', it indicates that one or more of the BDs
+	 * associated with this packet generated a PCI error. This probably
+	 * means the address was not valid.
+	 */
+	#define TX_CMPL_ERRORS_DMA_ERROR		UINT32_C(0x40)
+	/*
+	 * When this bit is '1', it indicates that the packet was longer than
+	 * indicated by the hint. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_HINT_TOO_SHORT		UINT32_C(0x80)
+	/*
+	 * When this bit is '1', it indicates that the packet was dropped due to
+	 * Poison TLP error on one or more of the TLPs in the PXP completion.
+	 */
+	#define TX_CMPL_ERRORS_POISON_TLP_ERROR		UINT32_C(0x100)
+	#define TX_CMPL_ERRORS_MASK			UINT32_C(0xfffe)
+	#define TX_CMPL_ERRORS_SFT			1
+	uint16_t errors_v;
+
+	uint16_t unused_1;
+	uint32_t unused_2;
+} __attribute__((packed)) tx_cmpl_t, *ptx_cmpl_t;
+
 /* HWRM Forwarded Request (16 bytes) */
 struct hwrm_fwd_req_cmpl {
 	/* Length of forwarded request in bytes. */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 14/38] bnxt: add initial Rx code implementation
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (11 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 13/38] bnxt: add initial Tx code implementation Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings Stephen Hurd
                     ` (26 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds initial implementation of rx_pkt_burst() function fo Rx.
bnxt_recv_pkts() is the top level function for doing Rx.

This patch also adds code to allocate rings in the ASIC.

For each Rx queue allocated in the PMD driver, a corresponding ring
in hardware will be created. Every time a frame is received a Rx ring
is selected based on the hardware configuration like RSS, MAC or VLAN,
COS and such. The hardware uses a completion ring to indicate the
availability of a packet.

This patch also brings in functions like bnxt_init_one_rx_ring()
bnxt_init_rx_ring_struct() which initializes various structures before
a Rx can begin.

bnxt_init_rxbds() initializes the Rx Buffer Descriptors while
bnxt_alloc_rx_data() allocates a buffer in the host to receive the
incoming packet.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Use rte_mbuf_raw_alloc instead of the now depricated
__rte_mbuf_raw_alloc and fix issues pointed out by checkpatch.

v5:
Expand the patch description
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +-
 drivers/net/bnxt/bnxt_ring.c           |  61 ++---
 drivers/net/bnxt/bnxt_rxq.c            |  34 ++-
 drivers/net/bnxt/bnxt_rxr.c            | 341 ++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_rxr.h            |  62 +++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 474 +++++++++++++++++++++++++++++++++
 7 files changed, 936 insertions(+), 40 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_rxr.c
 create mode 100644 drivers/net/bnxt/bnxt_rxr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 0785681..4d35412 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -54,6 +54,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txr.c
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 4ace543..6888363 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -42,6 +42,7 @@
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
 #include "bnxt_txr.h"
@@ -269,7 +270,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 		goto error;
 	}
 	eth_dev->dev_ops = &bnxt_dev_ops;
-	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
+	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
 
 	rc = bnxt_alloc_hwrm_resources(bp);
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index 2645dda..f547a9e 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -36,6 +36,7 @@
 #include "bnxt.h"
 #include "bnxt_cpr.h"
 #include "bnxt_ring.h"
+#include "bnxt_rxr.h"
 #include "bnxt_txr.h"
 
 #include "hsi_struct_def_dpdk.h"
@@ -74,8 +75,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 {
 	struct bnxt_ring *cp_ring = cp_ring_info->cp_ring_struct;
 	struct bnxt_ring *tx_ring;
-	/* TODO: RX ring */
-	/* struct bnxt_ring *rx_ring; */
+	struct bnxt_ring *rx_ring;
 	struct rte_pci_device *pdev = bp->pdev;
 	const struct rte_memzone *mz = NULL;
 	char mz_name[RTE_MEMZONE_NAMESIZE];
@@ -92,12 +92,9 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 						tx_ring_struct->vmem_size) : 0;
 
 	int rx_vmem_start = tx_vmem_start + tx_vmem_len;
-	/* TODO: RX ring */
-	int rx_vmem_len = 0;
-	/*
-	 * rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
-	 * rx_ring_struct->vmem_size) : 0;
-	 */
+	int rx_vmem_len = rx_ring_info ?
+		RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
+						rx_ring_struct->vmem_size) : 0;
 
 	int cp_ring_start = rx_vmem_start + rx_vmem_len;
 	int cp_ring_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->ring_size *
@@ -109,13 +106,9 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 				   sizeof(struct tx_bd_long)) : 0;
 
 	int rx_ring_start = tx_ring_start + tx_ring_len;
-	/* TODO: RX ring */
-	int rx_ring_len = 0;
-	/*
-	 * rx_ring_info ?
-	 * RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
-	 * sizeof(struct rx_prod_pkt_bd)) : 0;
-	 */
+	int rx_ring_len =  rx_ring_info ?
+		RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
+		sizeof(struct rx_prod_pkt_bd)) : 0;
 
 	int total_alloc_len = rx_ring_start + rx_ring_len;
 
@@ -153,26 +146,24 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		}
 	}
 
-/*
- *	if (rx_ring_info) {
- *		rx_ring = &rx_ring_info->rx_ring_struct;
- *
- *		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
- *		rx_ring_info->rx_desc_ring =
- *		    (struct rx_prod_pkt_bd *)rx_ring->bd;
- *		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
- *		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
- *
- *		if (!rx_ring->bd)
- *			return -ENOMEM;
- *		if (rx_ring->vmem_size) {
- *			rx_ring->vmem =
- *			    (void **)((char *)mz->addr + rx_vmem_start);
- *			rx_ring_info->rx_buf_ring =
- *			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
- *		}
- *	}
- */
+	if (rx_ring_info) {
+		rx_ring = rx_ring_info->rx_ring_struct;
+
+		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
+		rx_ring_info->rx_desc_ring =
+		    (struct rx_prod_pkt_bd *)rx_ring->bd;
+		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
+		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+
+		if (!rx_ring->bd)
+			return -ENOMEM;
+		if (rx_ring->vmem_size) {
+			rx_ring->vmem =
+			    (void **)((char *)mz->addr + rx_vmem_start);
+			rx_ring_info->rx_buf_ring =
+			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
+		}
+	}
 
 	cp_ring->bd = ((char *)mz->addr + cp_ring_start);
 	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index 19c0e3f..eeda631 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -41,6 +41,7 @@
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -216,7 +217,20 @@ err_out:
 
 static void bnxt_rx_queue_release_mbufs(struct bnxt_rx_queue *rxq __rte_unused)
 {
-	/* TODO: Requires interaction with TX ring */
+	struct bnxt_sw_rx_bd *sw_ring;
+	uint16_t i;
+
+	if (rxq) {
+		sw_ring = rxq->rx_ring->rx_buf_ring;
+		if (sw_ring) {
+			for (i = 0; i < rxq->nb_rx_desc; i++) {
+				if (sw_ring[i].mbuf) {
+					rte_pktmbuf_free_seg(sw_ring[i].mbuf);
+					sw_ring[i].mbuf = NULL;
+				}
+			}
+		}
+	}
 }
 
 void bnxt_free_rx_mbufs(struct bnxt *bp)
@@ -237,7 +251,13 @@ void bnxt_rx_queue_release_op(void *rx_queue)
 	if (rxq) {
 		bnxt_rx_queue_release_mbufs(rxq);
 
-		/* TODO: Free ring and stats here */
+		/* Free RX ring hardware descriptors */
+		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
+
+		/* Free RX completion ring hardware descriptors */
+		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+
+		bnxt_free_rxq_stats(rxq);
 
 		rte_free(rxq);
 	}
@@ -274,7 +294,7 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	rxq->nb_rx_desc = nb_desc;
 	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
 
-	/* TODO: Initialize ring structure */
+	bnxt_init_rx_ring_struct(rxq);
 
 	rxq->queue_id = queue_idx;
 	rxq->port_id = eth_dev->data->port_id;
@@ -282,7 +302,13 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 				0 : ETHER_CRC_LEN);
 
 	eth_dev->data->rx_queues[queue_idx] = rxq;
-	/* TODO: Allocate RX ring hardware descriptors */
+	/* Allocate RX ring hardware descriptors */
+	if (bnxt_alloc_rings(bp, queue_idx, NULL, rxq->rx_ring, rxq->cp_ring,
+			"rxr")) {
+		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for rx_ring failed!");
+		bnxt_rx_queue_release_op(rxq);
+		return -ENOMEM;
+	}
 
 	return 0;
 }
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
new file mode 100644
index 0000000..b02395b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -0,0 +1,341 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include <rte_byteorder.h>
+#include <rte_malloc.h>
+#include <rte_memory.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_ring.h"
+#include "bnxt_rxr.h"
+#include "bnxt_rxq.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * RX Ring handling
+ */
+
+static inline struct rte_mbuf *__bnxt_alloc_rx_data(struct rte_mempool *mb)
+{
+	struct rte_mbuf *data;
+
+	data = rte_mbuf_raw_alloc(mb);
+
+	return data;
+}
+
+static inline int bnxt_alloc_rx_data(struct bnxt_rx_queue *rxq,
+				     struct bnxt_rx_ring_info *rxr,
+				     uint16_t prod)
+{
+	struct rx_prod_pkt_bd *rxbd = &rxr->rx_desc_ring[prod];
+	struct bnxt_sw_rx_bd *rx_buf = &rxr->rx_buf_ring[prod];
+	struct rte_mbuf *data;
+
+	data = __bnxt_alloc_rx_data(rxq->mb_pool);
+	if (!data)
+		return -ENOMEM;
+
+	rx_buf->mbuf = data;
+
+	rxbd->addr = rte_cpu_to_le_64(RTE_MBUF_DATA_DMA_ADDR(rx_buf->mbuf));
+
+	return 0;
+}
+
+static void bnxt_reuse_rx_mbuf(struct bnxt_rx_ring_info *rxr, uint16_t cons,
+			       struct rte_mbuf *mbuf)
+{
+	uint16_t prod = rxr->rx_prod;
+	struct bnxt_sw_rx_bd *prod_rx_buf;
+	struct rx_prod_pkt_bd *prod_bd, *cons_bd;
+
+	prod_rx_buf = &rxr->rx_buf_ring[prod];
+
+	prod_rx_buf->mbuf = mbuf;
+
+	prod_bd = &rxr->rx_desc_ring[prod];
+	cons_bd = &rxr->rx_desc_ring[cons];
+
+	prod_bd->addr = cons_bd->addr;
+}
+
+static uint16_t bnxt_rx_pkt(struct rte_mbuf **rx_pkt,
+			    struct bnxt_rx_queue *rxq, uint32_t *raw_cons)
+{
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+	struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+	struct rx_pkt_cmpl *rxcmp;
+	struct rx_pkt_cmpl_hi *rxcmp1;
+	uint32_t tmp_raw_cons = *raw_cons;
+	uint16_t cons, prod, cp_cons =
+	    RING_CMP(cpr->cp_ring_struct, tmp_raw_cons);
+	struct bnxt_sw_rx_bd *rx_buf;
+	struct rte_mbuf *mbuf;
+	int rc = 0;
+
+	rxcmp = (struct rx_pkt_cmpl *)
+	    &cpr->cp_desc_ring[cp_cons];
+
+	tmp_raw_cons = NEXT_RAW_CMP(tmp_raw_cons);
+	cp_cons = RING_CMP(cpr->cp_ring_struct, tmp_raw_cons);
+	rxcmp1 = (struct rx_pkt_cmpl_hi *)&cpr->cp_desc_ring[cp_cons];
+
+	if (!CMP_VALID(rxcmp1, tmp_raw_cons, cpr->cp_ring_struct))
+		return -EBUSY;
+
+	prod = rxr->rx_prod;
+
+	/* EW - GRO deferred to phase 3 */
+	cons = rxcmp->opaque;
+	rx_buf = &rxr->rx_buf_ring[cons];
+	mbuf = rx_buf->mbuf;
+	rte_prefetch0(mbuf);
+
+	mbuf->nb_segs = 1;
+	mbuf->next = NULL;
+	mbuf->pkt_len = rxcmp->len;
+	mbuf->data_len = mbuf->pkt_len;
+	mbuf->port = rxq->port_id;
+	mbuf->ol_flags = 0;
+	if (rxcmp->flags_type & RX_PKT_CMPL_FLAGS_RSS_VALID) {
+		mbuf->hash.rss = rxcmp->rss_hash;
+		mbuf->ol_flags |= PKT_RX_RSS_HASH;
+	} else {
+		mbuf->hash.fdir.id = rxcmp1->cfa_code;
+		mbuf->ol_flags |= PKT_RX_FDIR | PKT_RX_FDIR_ID;
+	}
+	if (rxcmp1->flags2 & RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN) {
+		mbuf->vlan_tci = rxcmp1->metadata &
+			(RX_PKT_CMPL_METADATA_VID_MASK |
+			RX_PKT_CMPL_METADATA_DE |
+			RX_PKT_CMPL_METADATA_PRI_MASK);
+		mbuf->ol_flags |= PKT_RX_VLAN_PKT;
+	}
+
+	rx_buf->mbuf = NULL;
+	if (rxcmp1->errors_v2 & RX_CMP_L2_ERRORS) {
+		/* Re-install the mbuf back to the rx ring */
+		bnxt_reuse_rx_mbuf(rxr, cons, mbuf);
+
+		rc = -EIO;
+		goto next_rx;
+	}
+	/*
+	 * TODO: Redesign this....
+	 * If the allocation fails, the packet does not get received.
+	 * Simply returning this will result in slowly falling behind
+	 * on the producer ring buffers.
+	 * Instead, "filling up" the producer just before ringing the
+	 * doorbell could be a better solution since it will let the
+	 * producer ring starve until memory is available again pushing
+	 * the drops into hardware and getting them out of the driver
+	 * allowing recovery to a full producer ring.
+	 *
+	 * This could also help with cache usage by preventing per-packet
+	 * calls in favour of a tight loop with the same function being called
+	 * in it.
+	 */
+	if (bnxt_alloc_rx_data(rxq, rxr, prod)) {
+		RTE_LOG(ERR, PMD, "mbuf alloc failed with prod=0x%x\n", prod);
+		rc = -ENOMEM;
+		goto next_rx;
+	}
+
+	/*
+	 * All MBUFs are allocated with the same size under DPDK,
+	 * no optimization for rx_copy_thresh
+	 */
+
+	/* AGG buf operation is deferred */
+
+	/* EW - VLAN reception.  Must compare against the ol_flags */
+
+	*rx_pkt = mbuf;
+next_rx:
+	rxr->rx_prod = RING_NEXT(rxr->rx_ring_struct, prod);
+
+	*raw_cons = tmp_raw_cons;
+
+	return rc;
+}
+
+uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct bnxt_rx_queue *rxq = rx_queue;
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+	struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+	uint32_t raw_cons = cpr->cp_raw_cons;
+	uint32_t cons;
+	int nb_rx_pkts = 0;
+	bool rx_event = false;
+	struct rx_pkt_cmpl *rxcmp;
+
+	/* Handle RX burst request */
+	while (1) {
+		int rc;
+
+		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
+		rte_prefetch0(&cpr->cp_desc_ring[cons]);
+		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
+
+		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
+			break;
+
+		/* TODO: Avoid magic numbers... */
+		if ((CMP_TYPE(rxcmp) & 0x30) == 0x10) {
+			rc = bnxt_rx_pkt(&rx_pkts[nb_rx_pkts], rxq, &raw_cons);
+			if (likely(!rc))
+				nb_rx_pkts++;
+			else if (rc == -EBUSY)	/* partial completion */
+				break;
+			rx_event = true;
+		}
+		raw_cons = NEXT_RAW_CMP(raw_cons);
+		if (nb_rx_pkts == nb_pkts)
+			break;
+	}
+	if (raw_cons == cpr->cp_raw_cons) {
+		/*
+		 * For PMD, there is no need to keep on pushing to REARM
+		 * the doorbell if there are no new completions
+		 */
+		return nb_rx_pkts;
+	}
+	cpr->cp_raw_cons = raw_cons;
+
+	B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+	if (rx_event)
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+	return nb_rx_pkts;
+}
+
+void bnxt_free_rx_rings(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->rx_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+
+		if (!rxq)
+			continue;
+
+		/* TODO: free() rxq->rx_ring and rxq->rx_ring->rx_ring_struct */
+		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
+		/* TODO: free() rxq->cp_ring and rxq->cp_ring->cp_ring_struct */
+		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+
+		rte_free(rxq);
+		bp->rx_queues[i] = NULL;
+	}
+}
+
+void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt *bp = rxq->bp;
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_rx_ring_info *rxr;
+	struct bnxt_ring *ring;
+
+	rxq->rx_buf_use_size = bp->eth_dev->data->mtu +
+			       ETHER_HDR_LEN + ETHER_CRC_LEN +
+			       (2 * VLAN_TAG_SIZE);
+	rxq->rx_buf_size = rxq->rx_buf_use_size + sizeof(struct rte_mbuf);
+
+	rxr = rxq->rx_ring;
+	ring = rxr->rx_ring_struct;
+	ring->ring_size = rte_align32pow2(rxq->nb_rx_desc);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)rxr->rx_desc_ring;
+	ring->bd_dma = rxr->rx_desc_mapping;
+	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_rx_bd);
+	ring->vmem = (void **)&rxr->rx_buf_ring;
+
+	cpr = rxq->cp_ring;
+	ring = cpr->cp_ring_struct;
+	ring->ring_size = rxr->rx_ring_struct->ring_size * 2;
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
+
+static void bnxt_init_rxbds(struct bnxt_ring *ring, uint32_t type,
+			    uint16_t len)
+{
+	uint32_t j;
+	struct rx_prod_pkt_bd *rx_bd_ring = (struct rx_prod_pkt_bd *)ring->bd;
+
+	if (!rx_bd_ring)
+		return;
+	for (j = 0; j < ring->ring_size; j++) {
+		rx_bd_ring[j].flags_type = type;
+		rx_bd_ring[j].len = len;
+		rx_bd_ring[j].opaque = j;
+	}
+}
+
+int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt_rx_ring_info *rxr;
+	struct bnxt_ring *ring;
+	uint32_t prod, type;
+	unsigned int i;
+
+	type = RX_PROD_PKT_BD_TYPE_RX_PROD_PKT | RX_PROD_PKT_BD_FLAGS_EOP_PAD;
+
+	/* TODO: These need to be allocated */
+	rxr = rxq->rx_ring;
+	ring = rxr->rx_ring_struct;
+	bnxt_init_rxbds(ring, type, rxq->rx_buf_use_size);
+
+	prod = rxr->rx_prod;
+	for (i = 0; i < ring->ring_size; i++) {
+		if (bnxt_alloc_rx_data(rxq, rxr, prod) != 0) {
+			RTE_LOG(WARNING, PMD,
+				"init'ed rx ring %d with %d/%d mbufs only\n",
+				rxq->queue_id, i, ring->ring_size);
+			break;
+		}
+		rxr->rx_prod = prod;
+		prod = RING_NEXT(rxr->rx_ring_struct, prod);
+	}
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
new file mode 100644
index 0000000..95c61a0
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -0,0 +1,62 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RXR_H_
+#define _BNXT_RXR_H_
+
+#define B_RX_DB(db, prod)						\
+		(*(uint32_t *)db = (DB_KEY_RX | prod))
+
+struct bnxt_sw_rx_bd {
+	struct rte_mbuf		*mbuf; /* data associated with RX descriptor */
+};
+
+struct bnxt_rx_ring_info {
+	uint16_t		rx_prod;
+	void			*rx_doorbell;
+
+	struct rx_prod_pkt_bd	*rx_desc_ring;
+	struct bnxt_sw_rx_bd	*rx_buf_ring; /* sw ring */
+
+	phys_addr_t		rx_desc_mapping;
+
+	struct bnxt_ring	*rx_ring_struct;
+};
+
+uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+			       uint16_t nb_pkts);
+void bnxt_free_rx_rings(struct bnxt *bp);
+void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq);
+int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 1543f20..8b30787 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -446,6 +446,79 @@ struct tx_bd_long_hi {
 	uint32_t cfa_meta;
 } __attribute__((packed));
 
+/* RX Producer Packet BD (16 bytes) */
+struct rx_prod_pkt_bd {
+	/* This value identifies the type of buffer descriptor. */
+	#define RX_PROD_PKT_BD_TYPE_MASK		UINT32_C(0x3f)
+	#define RX_PROD_PKT_BD_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 16B long and is an RX Producer (ie.
+		 * empty) buffer descriptor.
+		 */
+	#define RX_PROD_PKT_BD_TYPE_RX_PROD_PKT		(UINT32_C(0x4) << 0)
+	/*
+	 * If set to 1, the packet will be placed at the address plus 2B. The 2
+	 * Bytes of padding will be written as zero.
+	 */
+	/*
+	 * This is intended to be used when the host buffer is cache-line
+	 * aligned to produce packets that are easy to parse in host memory
+	 * while still allowing writes to be cache line aligned.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_SOP_PAD		UINT32_C(0x40)
+	/*
+	 * If set to 1, the packet write will be padded out to the nearest
+	 * cache-line with zero value padding.
+	 */
+	/*
+	 * If receive buffers start/end on cache-line boundaries, this feature
+	 * will ensure that all data writes on the PCI bus start/end on cache
+	 * line boundaries.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_EOP_PAD		UINT32_C(0x80)
+	/*
+	 * This value is the number of additional buffers in the ring that
+	 * describe the buffer space to be consumed for the this packet. If the
+	 * value is zero, then the packet must fit within the space described by
+	 * this BD. If this value is 1 or more, it indicates how many additional
+	 * "buffer" BDs are in the ring immediately following this BD to be used
+	 * for the same network packet. Even if the packet to be placed does not
+	 * need all the additional buffers, they will be consumed anyway.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_BUFFERS_MASK	UINT32_C(0x300)
+	#define RX_PROD_PKT_BD_FLAGS_BUFFERS_SFT	8
+	#define RX_PROD_PKT_BD_FLAGS_MASK		UINT32_C(0xffc0)
+	#define RX_PROD_PKT_BD_FLAGS_SFT		6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length in Bytes of the host physical buffer where data
+	 * for the packet may be placed in host memory.
+	 */
+	/*
+	 * While this is a Byte resolution value, it is often advantageous to
+	 * ensure that the buffers provided end on a host cache line.
+	 */
+	uint16_t len;
+
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with this
+	 * receive buffer set.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address where data for the packet may by
+	 * placed in host memory.
+	 */
+	/*
+	 * While this is a Byte resolution value, it is often advantageous to
+	 * ensure that the buffers provide start on a host cache line.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
 /* Completion Ring Structures */
 /* Note: This structure is used by the HWRM to communicate HWRM Error. */
 /* Base Completion Record (16 bytes) */
@@ -611,6 +684,407 @@ struct tx_cmpl {
 	uint32_t unused_2;
 } __attribute__((packed)) tx_cmpl_t, *ptx_cmpl_t;
 
+/* RX Packet Completion Record (32 bytes split to 2 16-byte struct) */
+struct rx_pkt_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define RX_PKT_CMPL_TYPE_MASK			UINT32_C(0x3f)
+	#define RX_PKT_CMPL_TYPE_SFT			0
+		/*
+		 * RX L2 completion: Completion of and L2 RX packet.
+		 * Length = 32B
+		 */
+	#define RX_PKT_CMPL_TYPE_RX_L2			(UINT32_C(0x11) << 0)
+	/*
+	 * When this bit is '1', it indicates a packet that has an error of some
+	 * type. Type of error is indicated in error_flags.
+	 */
+	#define RX_PKT_CMPL_FLAGS_ERROR			UINT32_C(0x40)
+	/* This field indicates how the packet was placed in the buffer. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_MASK	UINT32_C(0x380)
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_SFT		7
+		/* Normal: Packet was placed using normal algorithm. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_NORMAL	(UINT32_C(0x0) << 7)
+		/* Jumbo: Packet was placed using jumbo algorithm. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_JUMBO	(UINT32_C(0x1) << 7)
+		/*
+		 * Header/Data Separation: Packet was placed using Header/Data
+		 * separation algorithm. The separation location is indicated by
+		 * the itype field.
+		 */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_HDS		(UINT32_C(0x2) << 7)
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_LAST \
+						RX_PKT_CMPL_FLAGS_PLACEMENT_HDS
+	/* This bit is '1' if the RSS field in this completion is valid. */
+	#define RX_PKT_CMPL_FLAGS_RSS_VALID		UINT32_C(0x400)
+	/*
+	 * This value indicates what the inner packet determined for the packet
+	 * was.
+	 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_MASK		UINT32_C(0xf000)
+	#define RX_PKT_CMPL_FLAGS_ITYPE_SFT		12
+		/* Not Known: Indicates that the packet type was not known. */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_NOT_KNOWN	(UINT32_C(0x0) << 12)
+		/*
+		 * IP Packet: Indicates that the packet was an IP packet, but
+		 * further classification was not possible.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_IP		(UINT32_C(0x1) << 12)
+		/*
+		 * TCP Packet: Indicates that the packet was IP and TCP. This
+		 * indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_TCP		(UINT32_C(0x2) << 12)
+		/*
+		 * UDP Packet: Indicates that the packet was IP and UDP. This
+		 * indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_UDP		(UINT32_C(0x3) << 12)
+		/*
+		 * FCoE Packet: Indicates that the packet was recognized as a
+		 * FCoE. This also indicates that the payload_offset field is
+		 * valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_FCOE		(UINT32_C(0x4) << 12)
+		/*
+		 * RoCE Packet: Indicates that the packet was recognized as a
+		 * RoCE. This also indicates that the payload_offset field is
+		 * valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_ROCE		(UINT32_C(0x5) << 12)
+		/*
+		 * ICMP Packet: Indicates that the packet was recognized as
+		 * ICMP. This indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_ICMP		(UINT32_C(0x7) << 12)
+		/*
+		 * PtP packet wo/timestamp: Indicates that the packet was
+		 * recognized as a PtP packet.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_PTP_WO_TIMESTAMP \
+							(UINT32_C(0x8) << 12)
+		/*
+		 * PtP packet w/timestamp: Indicates that the packet was
+		 * recognized as a PtP packet and that a timestamp was taken for
+		 * the packet.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_PTP_W_TIMESTAMP	(UINT32_C(0x9) << 12)
+	#define RX_PKT_CMPL_FLAGS_ITYPE_LAST \
+					RX_PKT_CMPL_FLAGS_ITYPE_PTP_W_TIMESTAMP
+	#define RX_PKT_CMPL_FLAGS_MASK			UINT32_C(0xffc0)
+	#define RX_PKT_CMPL_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the data for the packet stored in the buffer(s)
+	 * identified by the opaque value. This includes the packet BD and any
+	 * associated buffer BDs. This does not include the the length of any
+	 * data places in aggregation BDs.
+	 */
+	uint16_t len;
+
+	/*
+	 * This is a copy of the opaque field from the RX BD this completion
+	 * corresponds to.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define RX_PKT_CMPL_V1				UINT32_C(0x1)
+	/*
+	 * This value is the number of aggregation buffers that follow this
+	 * entry in the completion ring that are a part of this packet. If the
+	 * value is zero, then the packet is completely contained in the buffer
+	 * space provided for the packet in the RX ring.
+	 */
+	#define RX_PKT_CMPL_AGG_BUFS_MASK		UINT32_C(0x3e)
+	#define RX_PKT_CMPL_AGG_BUFS_SFT		1
+	uint8_t agg_bufs_v1;
+
+	/*
+	 * This is the RSS hash type for the packet. The value is packed
+	 * {tuple_extrac_op[1:0],rss_profile_id[4:0],tuple_extrac_op[2]}.
+	 */
+	uint8_t rss_hash_type;
+
+	/*
+	 * This value indicates the offset from the beginning of the packet
+	 * where the inner payload starts. This value is valid for TCP, UDP,
+	 * FCoE, and RoCE packets.
+	 */
+	uint8_t payload_offset;
+
+	uint8_t unused_1;
+
+	/*
+	 * This value is the RSS hash value calculated for the packet based on
+	 * the mode bits and key value in the VNIC.
+	 */
+	uint32_t rss_hash;
+} __attribute__((packed));
+
+/* last 16 bytes of RX Packet Completion Record */
+struct rx_pkt_cmpl_hi {
+	/*
+	 * This indicates that the ip checksum was calculated for the inner
+	 * packet and that the ip_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_IP_CS_CALC		UINT32_C(0x1)
+	/*
+	 * This indicates that the TCP, UDP or ICMP checksum was calculated for
+	 * the inner packet and that the l4_cs_error field indicates if there
+	 * was an error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_L4_CS_CALC		UINT32_C(0x2)
+	/*
+	 * This indicates that the ip checksum was calculated for the tunnel
+	 * header and that the t_ip_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_T_IP_CS_CALC		UINT32_C(0x4)
+	/*
+	 * This indicates that the UDP checksum was calculated for the tunnel
+	 * packet and that the t_l4_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_T_L4_CS_CALC		UINT32_C(0x8)
+	/* This value indicates what format the metadata field is. */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_MASK	UINT32_C(0xf0)
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_SFT	4
+		/* No metadata informtaion. Value is zero. */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_NONE	(UINT32_C(0x0) << 4)
+		/*
+		 * The metadata field contains the VLAN tag and TPID value. -
+		 * metadata[11:0] contains the vlan VID value. - metadata[12]
+		 * contains the vlan DE value. - metadata[15:13] contains the
+		 * vlan PRI value. - metadata[31:16] contains the vlan TPID
+		 * value.
+		 */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN	(UINT32_C(0x1) << 4)
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_LAST \
+					RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN
+	/*
+	 * This field indicates the IP type for the inner-most IP header. A
+	 * value of '0' indicates IPv4. A value of '1' indicates IPv6. This
+	 * value is only valid if itype indicates a packet with an IP header.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_IP_TYPE		UINT32_C(0x100)
+	uint32_t flags2;
+
+	/*
+	 * This is data from the CFA block as indicated by the meta_format
+	 * field.
+	 */
+	/* When meta_format=1, this value is the VLAN VID. */
+	#define RX_PKT_CMPL_METADATA_VID_MASK		UINT32_C(0xfff)
+	#define RX_PKT_CMPL_METADATA_VID_SFT		0
+	/* When meta_format=1, this value is the VLAN DE. */
+	#define RX_PKT_CMPL_METADATA_DE			UINT32_C(0x1000)
+	/* When meta_format=1, this value is the VLAN PRI. */
+	#define RX_PKT_CMPL_METADATA_PRI_MASK		UINT32_C(0xe000)
+	#define RX_PKT_CMPL_METADATA_PRI_SFT		13
+	/* When meta_format=1, this value is the VLAN TPID. */
+	#define RX_PKT_CMPL_METADATA_TPID_MASK		UINT32_C(0xffff0000)
+	#define RX_PKT_CMPL_METADATA_TPID_SFT		16
+	uint32_t metadata;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define RX_PKT_CMPL_V2				UINT32_C(0x1)
+	/*
+	 * This error indicates that there was some sort of problem with the BDs
+	 * for the packet that was found after part of the packet was already
+	 * placed. The packet should be treated as invalid.
+	 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_MASK	UINT32_C(0xe)
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_SFT	1
+		/* No buffer error */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_NO_BUFFER \
+							(UINT32_C(0x0) << 1)
+		/*
+		 * Did Not Fit: Packet did not fit into packet buffer provided.
+		 * For regular placement, this means the packet did not fit in
+		 * the buffer provided. For HDS and jumbo placement, this means
+		 * that the packet could not be placed into 7 physical buffers
+		 * or less.
+		 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_DID_NOT_FIT \
+							(UINT32_C(0x1) << 1)
+		/*
+		 * Not On Chip: All BDs needed for the packet were not on-chip
+		 * when the packet arrived.
+		 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_NOT_ON_CHIP \
+							(UINT32_C(0x2) << 1)
+		/* Bad Format: BDs were not formatted correctly. */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_BAD_FORMAT \
+							(UINT32_C(0x3) << 1)
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_BUFFER_ERROR_BAD_FORMAT
+	/* This indicates that there was an error in the IP header checksum. */
+	#define RX_PKT_CMPL_ERRORS_IP_CS_ERROR		UINT32_C(0x10)
+	/*
+	 * This indicates that there was an error in the TCP, UDP or ICMP
+	 * checksum.
+	 */
+	#define RX_PKT_CMPL_ERRORS_L4_CS_ERROR		UINT32_C(0x20)
+	/*
+	 * This indicates that there was an error in the tunnel IP header
+	 * checksum.
+	 */
+	#define RX_PKT_CMPL_ERRORS_T_IP_CS_ERROR	UINT32_C(0x40)
+	/* This indicates that there was an error in the tunnel UDP checksum. */
+	#define RX_PKT_CMPL_ERRORS_T_L4_CS_ERROR	UINT32_C(0x80)
+	/*
+	 * This indicates that there was a CRC error on either an FCoE or RoCE
+	 * packet. The itype indicates the packet type.
+	 */
+	#define RX_PKT_CMPL_ERRORS_CRC_ERROR		UINT32_C(0x100)
+	/*
+	 * This indicates that there was an error in the tunnel portion of the
+	 * packet when this field is non-zero.
+	 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_MASK	UINT32_C(0xe00)
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_SFT	9
+		/*
+		 * No additional error occurred on the tunnel portion of the
+		 * packet of the packet does not have a tunnel.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_NO_ERROR	(UINT32_C(0x0) << 9)
+		/*
+		 * Indicates that IP header version does not match expectation
+		 * from L2 Ethertype for IPv4 and IPv6 in the tunnel header.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_VERSION \
+							(UINT32_C(0x1) << 9)
+		/*
+		 * Indicates that header length is out of range in the tunnel
+		 * header. Valid for IPv4.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_HDR_LEN \
+							(UINT32_C(0x2) << 9)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the PPPoE header length for a tunnel PPPoE packet.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_TUNNEL_TOTAL_ERROR \
+							(UINT32_C(0x3) << 9)
+		/*
+		 * Indicates that physical packet is shorter than that claimed
+		 * by the tunnel l3 header length. Valid for IPv4, or IPv6
+		 * tunnel packet packets.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_IP_TOTAL_ERROR \
+							(UINT32_C(0x4) << 9)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the tunnel UDP header length for a tunnel UDP
+		 * packet that is not fragmented.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_UDP_TOTAL_ERROR \
+							(UINT32_C(0x5) << 9)
+		/*
+		 * indicates that the IPv4 TTL or IPv6 hop limit check have
+		 * failed (e.g. TTL = 0) in the tunnel header. Valid for IPv4,
+		 * and IPv6.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_TTL \
+							(UINT32_C(0x6) << 9)
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_TTL
+	/*
+	 * This indicates that there was an error in the inner portion of the
+	 * packet when this field is non-zero.
+	 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_MASK	UINT32_C(0xf000)
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_SFT	12
+		/*
+		 * No additional error occurred on the tunnel portion of the
+		 * packet of the packet does not have a tunnel.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_NO_ERROR	(UINT32_C(0x0) << 12)
+		/*
+		 * Indicates that IP header version does not match expectation
+		 * from L2 Ethertype for IPv4 and IPv6 or that option other than
+		 * VFT was parsed on FCoE packet.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_VERSION \
+							(UINT32_C(0x1) << 12)
+		/*
+		 * indicates that header length is out of range. Valid for IPv4
+		 * and RoCE
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_HDR_LEN \
+							(UINT32_C(0x2) << 12)
+		/*
+		 * indicates that the IPv4 TTL or IPv6 hop limit check have
+		 * failed (e.g. TTL = 0). Valid for IPv4, and IPv6
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_TTL	(UINT32_C(0x3) << 12)
+		/*
+		 * Indicates that physical packet is shorter than that claimed
+		 * by the l3 header length. Valid for IPv4, IPv6 packet or RoCE
+		 * packets.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_IP_TOTAL_ERROR \
+							(UINT32_C(0x4) << 12)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the UDP header length for a UDP packet that is not
+		 * fragmented.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_UDP_TOTAL_ERROR \
+							(UINT32_C(0x5) << 12)
+		/*
+		 * Indicates that TCP header length > IP payload. Valid for TCP
+		 * packets only.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_HDR_LEN \
+							(UINT32_C(0x6) << 12)
+		/* Indicates that TCP header length < 5. Valid for TCP. */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_HDR_LEN_TOO_SMALL \
+							(UINT32_C(0x7) << 12)
+		/*
+		 * Indicates that TCP option headers result in a TCP header size
+		 * that does not match data offset in TCP header. Valid for TCP.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_OPT_LEN \
+							(UINT32_C(0x8) << 12)
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_OPT_LEN
+	#define RX_PKT_CMPL_ERRORS_MASK			UINT32_C(0xfffe)
+	#define RX_PKT_CMPL_ERRORS_SFT			1
+	uint16_t errors_v2;
+
+	/*
+	 * This field identifies the CFA action rule that was used for this
+	 * packet.
+	 */
+	uint16_t cfa_code;
+
+	/*
+	 * This value holds the reordering sequence number for the packet. If
+	 * the reordering sequence is not valid, then this value is zero. The
+	 * reordering domain for the packet is in the bottom 8 to 10b of the
+	 * rss_hash value. The bottom 20b of this value contain the ordering
+	 * domain value for the packet.
+	 */
+	#define RX_PKT_CMPL_REORDER_MASK		UINT32_C(0xffffff)
+	#define RX_PKT_CMPL_REORDER_SFT			0
+	uint32_t reorder;
+} __attribute__((packed));
+
 /* HWRM Forwarded Request (16 bytes) */
 struct hwrm_fwd_req_cmpl {
 	/* Length of forwarded request in bytes. */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (12 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 14/38] bnxt: add initial Rx " Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 16/38] bnxt: add HWRM function reset command Stephen Hurd
                     ` (25 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Perform allocation and free()ing of ring and information structures
for the TX, RX, and completion rings. The previous patches had
so far provided top level stubs and generic ring support, while this
patch does the real allocation and freeing of the memory specific to
each different type of generic ring.

For example bnxt_init_tx_ring_struct() or bnxt_init_rx_ring_struct() is
now allocating memory based on the socked_id being provided.

bnxt_tx_queue_setup_op() or bnxt_rx_queue_setup_op() have gone through
some reformatting to perform a graceful cleanup in case memory
allocation fails.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.
Change the argument passed to bnxt_alloc_rings.
 Instead of passing bnxt_tx_ring and bnxt_rx_ring,
 shorten them to txr and rxr respectively.
Add code to free the reserved memzone

v5:
More descriptive commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_cpr.c  | 28 +++++++++++++++++++++++-----
 drivers/net/bnxt/bnxt_cpr.h  |  2 +-
 drivers/net/bnxt/bnxt_ring.c |  4 ++++
 drivers/net/bnxt/bnxt_ring.h |  1 +
 drivers/net/bnxt/bnxt_rxq.c  | 19 ++++++++++++-------
 drivers/net/bnxt/bnxt_rxr.c  | 42 ++++++++++++++++++++++++++++++++++--------
 drivers/net/bnxt/bnxt_rxr.h  |  2 +-
 drivers/net/bnxt/bnxt_txq.c  | 25 ++++++++++++++++---------
 drivers/net/bnxt/bnxt_txr.c  | 43 ++++++++++++++++++++++++++++++++++---------
 drivers/net/bnxt/bnxt_txr.h  |  2 +-
 10 files changed, 127 insertions(+), 41 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index 5a4b13d..60c277a 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -31,6 +31,8 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_malloc.h>
+
 #include "bnxt.h"
 #include "bnxt_cpr.h"
 #include "bnxt_hwrm.h"
@@ -121,21 +123,37 @@ reject:
 void bnxt_free_def_cp_ring(struct bnxt *bp)
 {
 	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-	struct bnxt_ring *ring = cpr->cp_ring_struct;
 
-	bnxt_free_ring(ring);
+	bnxt_free_ring(cpr->cp_ring_struct);
+	rte_free(cpr->cp_ring_struct);
+	rte_free(cpr);
 }
 
 /* For the default completion ring only */
-void bnxt_init_def_ring_struct(struct bnxt *bp)
+int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id)
 {
-	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-	struct bnxt_ring *ring = cpr->cp_ring_struct;
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_ring *ring;
 
+	cpr = rte_zmalloc_socket("cpr",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	bp->def_cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_cp_ring_struct",
+				  sizeof(struct bnxt_ring),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->ring_size = rte_align32pow2(DEFAULT_CP_RING_SIZE);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index 4bd5606..c176f8c 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -79,7 +79,7 @@ struct bnxt_cp_ring_info {
 
 struct bnxt;
 void bnxt_free_def_cp_ring(struct bnxt *bp);
-void bnxt_init_def_ring_struct(struct bnxt *bp);
+int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id);
 void bnxt_handle_async_event(struct bnxt *bp, struct cmpl_base *cmp);
 void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmp);
 
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index f547a9e..67b009a 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -51,6 +51,7 @@ void bnxt_free_ring(struct bnxt_ring *ring)
 		memset((char *)*ring->vmem, 0, ring->vmem_size);
 		*ring->vmem = NULL;
 	}
+	rte_memzone_free((const struct rte_memzone *)ring->mem_zone);
 }
 
 /*
@@ -135,6 +136,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		tx_ring_info->tx_desc_ring = (struct tx_bd_long *)tx_ring->bd;
 		tx_ring->bd_dma = mz->phys_addr + tx_ring_start;
 		tx_ring_info->tx_desc_mapping = tx_ring->bd_dma;
+		tx_ring->mem_zone = (const void *)mz;
 
 		if (!tx_ring->bd)
 			return -ENOMEM;
@@ -154,6 +156,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		    (struct rx_prod_pkt_bd *)rx_ring->bd;
 		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
 		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+		rx_ring->mem_zone = (const void *)mz;
 
 		if (!rx_ring->bd)
 			return -ENOMEM;
@@ -169,6 +172,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
 	cp_ring_info->cp_desc_ring = cp_ring->bd;
 	cp_ring_info->cp_desc_mapping = cp_ring->bd_dma;
+	cp_ring->mem_zone = (const void *)mz;
 
 	if (!cp_ring->bd)
 		return -ENOMEM;
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index 1e9d3be..f6ab214 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -76,6 +76,7 @@ struct bnxt_ring {
 	void			**vmem;
 
 	uint16_t		fw_ring_id; /* Ring id filled by Chimp FW */
+	const void		*mem_zone;
 };
 
 struct bnxt_ring_grp_info {
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index eeda631..cddf17d 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -53,8 +53,6 @@ void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq)
 {
 	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
 
-	/* 'Unreserve' rte_memzone */
-
 	if (cpr->hw_stats)
 		cpr->hw_stats = NULL;
 }
@@ -272,10 +270,12 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
 	struct bnxt_rx_queue *rxq;
+	int rc = 0;
 
 	if (!nb_desc || nb_desc > MAX_RX_DESC_CNT) {
 		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	if (eth_dev->data->rx_queues) {
@@ -287,14 +287,17 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 				 RTE_CACHE_LINE_SIZE, socket_id);
 	if (!rxq) {
 		RTE_LOG(ERR, PMD, "bnxt_rx_queue allocation failed!");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 	rxq->bp = bp;
 	rxq->mb_pool = mp;
 	rxq->nb_rx_desc = nb_desc;
 	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
 
-	bnxt_init_rx_ring_struct(rxq);
+	rc = bnxt_init_rx_ring_struct(rxq, socket_id);
+	if (rc)
+		goto out;
 
 	rxq->queue_id = queue_idx;
 	rxq->port_id = eth_dev->data->port_id;
@@ -307,8 +310,10 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 			"rxr")) {
 		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for rx_ring failed!");
 		bnxt_rx_queue_release_op(rxq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
-	return 0;
+out:
+	return rc;
 }
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
index b02395b..5d93de2 100644
--- a/drivers/net/bnxt/bnxt_rxr.c
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -255,17 +255,20 @@ void bnxt_free_rx_rings(struct bnxt *bp)
 		if (!rxq)
 			continue;
 
-		/* TODO: free() rxq->rx_ring and rxq->rx_ring->rx_ring_struct */
 		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
-		/* TODO: free() rxq->cp_ring and rxq->cp_ring->cp_ring_struct */
+		rte_free(rxq->rx_ring->rx_ring_struct);
+		rte_free(rxq->rx_ring);
+
 		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+		rte_free(rxq->cp_ring->cp_ring_struct);
+		rte_free(rxq->cp_ring);
 
 		rte_free(rxq);
 		bp->rx_queues[i] = NULL;
 	}
 }
 
-void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
+int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 {
 	struct bnxt *bp = rxq->bp;
 	struct bnxt_cp_ring_info *cpr;
@@ -277,8 +280,19 @@ void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
 			       (2 * VLAN_TAG_SIZE);
 	rxq->rx_buf_size = rxq->rx_buf_use_size + sizeof(struct rte_mbuf);
 
-	rxr = rxq->rx_ring;
-	ring = rxr->rx_ring_struct;
+	rxr = rte_zmalloc_socket("bnxt_rx_ring",
+				 sizeof(struct bnxt_rx_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxr == NULL)
+		return -ENOMEM;
+	rxq->rx_ring = rxr;
+
+	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
+				   sizeof(struct bnxt_ring),
+				   RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	rxr->rx_ring_struct = ring;
 	ring->ring_size = rte_align32pow2(rxq->nb_rx_desc);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)rxr->rx_desc_ring;
@@ -286,14 +300,27 @@ void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
 	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_rx_bd);
 	ring->vmem = (void **)&rxr->rx_buf_ring;
 
-	cpr = rxq->cp_ring;
-	ring = cpr->cp_ring_struct;
+	cpr = rte_zmalloc_socket("bnxt_rx_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	rxq->cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
+				   sizeof(struct bnxt_ring),
+				   RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->ring_size = rxr->rx_ring_struct->ring_size * 2;
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
 
 static void bnxt_init_rxbds(struct bnxt_ring *ring, uint32_t type,
@@ -320,7 +347,6 @@ int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
 
 	type = RX_PROD_PKT_BD_TYPE_RX_PROD_PKT | RX_PROD_PKT_BD_FLAGS_EOP_PAD;
 
-	/* TODO: These need to be allocated */
 	rxr = rxq->rx_ring;
 	ring = rxr->rx_ring_struct;
 	bnxt_init_rxbds(ring, type, rxq->rx_buf_use_size);
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
index 95c61a0..f766b26 100644
--- a/drivers/net/bnxt/bnxt_rxr.h
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -56,7 +56,7 @@ struct bnxt_rx_ring_info {
 uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 			       uint16_t nb_pkts);
 void bnxt_free_rx_rings(struct bnxt *bp);
-void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq);
+int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id);
 int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq);
 
 #endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 460f71a..99ddddd 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -49,8 +49,6 @@ void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
 {
 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
 
-	/* 'Unreserve' rte_memzone */
-
 	if (cpr->hw_stats)
 		cpr->hw_stats = NULL;
 }
@@ -108,10 +106,12 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
 	struct bnxt_tx_queue *txq;
+	int rc = 0;
 
 	if (!nb_desc || nb_desc > MAX_TX_DESC_CNT) {
 		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	if (eth_dev->data->tx_queues) {
@@ -123,15 +123,18 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	}
 	txq = rte_zmalloc_socket("bnxt_tx_queue", sizeof(struct bnxt_tx_queue),
 				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq == NULL) {
+	if (!txq) {
 		RTE_LOG(ERR, PMD, "bnxt_tx_queue allocation failed!");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 	txq->bp = bp;
 	txq->nb_tx_desc = nb_desc;
 	txq->tx_free_thresh = tx_conf->tx_free_thresh;
 
-	bnxt_init_tx_ring_struct(txq);
+	rc = bnxt_init_tx_ring_struct(txq, socket_id);
+	if (rc)
+		goto out;
 
 	txq->queue_id = queue_idx;
 	txq->port_id = eth_dev->data->port_id;
@@ -141,15 +144,19 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 			"txr")) {
 		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for tx_ring failed!");
 		bnxt_tx_queue_release_op(txq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
 	if (bnxt_init_one_tx_ring(txq)) {
 		RTE_LOG(ERR, PMD, "bnxt_init_one_tx_ring failed!");
 		bnxt_tx_queue_release_op(txq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
 	eth_dev->data->tx_queues[queue_idx] = txq;
-	return 0;
+
+out:
+	return rc;
 }
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
index 9064e8d..8bf8fee 100644
--- a/drivers/net/bnxt/bnxt_txr.c
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -59,9 +59,12 @@ void bnxt_free_tx_rings(struct bnxt *bp)
 			continue;
 
 		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
-		/* TODO: free() txq->tx_ring and txq->tx_ring->tx_ring_struct */
+		rte_free(txq->tx_ring->tx_ring_struct);
+		rte_free(txq->tx_ring);
+
 		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
-		/* TODO: free() txq->cp_ring and txq->cp_ring->cp_ring_struct */
+		rte_free(txq->cp_ring->cp_ring_struct);
+		rte_free(txq->cp_ring);
 
 		rte_free(txq);
 		bp->tx_queues[i] = NULL;
@@ -79,15 +82,25 @@ int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
 	return 0;
 }
 
-void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
+int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id)
 {
 	struct bnxt_cp_ring_info *cpr;
 	struct bnxt_tx_ring_info *txr;
 	struct bnxt_ring *ring;
 
-	/* TODO: These need to be allocated */
-	txr = txq->tx_ring;
-	ring = txr->tx_ring_struct;
+	txr = rte_zmalloc_socket("bnxt_tx_ring",
+				 sizeof(struct bnxt_tx_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txr == NULL)
+		return -ENOMEM;
+	txq->tx_ring = txr;
+
+	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
+				  sizeof(struct bnxt_ring),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	txr->tx_ring_struct = ring;
 	ring->ring_size = rte_align32pow2(txq->nb_tx_desc + 1);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)txr->tx_desc_ring;
@@ -95,15 +108,27 @@ void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
 	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_tx_bd);
 	ring->vmem = (void **)&txr->tx_buf_ring;
 
-	/* TODO: These need to be allocated */
-	cpr = txq->cp_ring;
-	ring = cpr->cp_ring_struct;
+	cpr = rte_zmalloc_socket("bnxt_tx_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	txq->cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
+				  sizeof(struct bnxt_ring),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->ring_size = txr->tx_ring_struct->ring_size;
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
 
 static inline uint32_t bnxt_tx_avail(struct bnxt_tx_ring_info *txr)
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
index 712ce65..2176aca 100644
--- a/drivers/net/bnxt/bnxt_txr.h
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -64,7 +64,7 @@ struct bnxt_sw_tx_bd {
 
 void bnxt_free_tx_rings(struct bnxt *bp);
 int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
-void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq);
+int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id);
 uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 			       uint16_t nb_pkts);
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 16/38] bnxt: add HWRM function reset command
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (13 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 17/38] bnxt: add HWRM vnic alloc function Stephen Hurd
                     ` (24 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add bnxt_hwrm_func_reset() function and supporting structs and macros.

New HWRM calls:
bnxt_hwrm_func_reset:
	This command puts the function into the reset state.
	In the reset state, global and port related features of the
	chip are not available.

This command resets a hardware function (PCIe function) and
frees any resources used by the function. This command initiated by
the driver prepare the function for re-use. This command may also be
initiated by a driver prior to doing it's own configuration.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
More descriptive commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           |  17 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 129 +++++++++++++++++++++++++++++++++
 3 files changed, 147 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 633683b..d1ca60a 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -265,6 +265,23 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_func_reset(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_func_reset_input req = {.req_type = 0 };
+	struct hwrm_func_reset_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, FUNC_RESET, -1, resp);
+
+	req.enables = rte_cpu_to_le_32(0);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd)
 {
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 75c3442..2679d09 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -54,6 +54,7 @@ int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+int bnxt_hwrm_func_reset(struct bnxt *bp);
 int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 8b30787..6209368 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -83,6 +83,7 @@ struct ctx_hw_stats64 {
  * Request types
  */
 #define HWRM_VER_GET			(UINT32_C(0x0))
+#define HWRM_FUNC_RESET			(UINT32_C(0x11))
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
@@ -2048,6 +2049,134 @@ struct hwrm_func_qcaps_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_func_reset */
+/*
+ * Description: This command resets a hardware function (PCIe function) and
+ * frees any resources used by the function. This command shall be initiated by
+ * the driver after an FLR has occurred to prepare the function for re-use. This
+ * command may also be initiated by a driver prior to doing it's own
+ * configuration. This command puts the function into the reset state. In the
+ * reset state, global and port related features of the chip are not available.
+ */
+/*
+ * Note: This command will reset a function that has already been disabled or
+ * idled. The command returns all the resources owned by the function so a new
+ * driver may allocate and configure resources normally.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_func_reset_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This bit must be '1' for the vf_id_valid field to be configured. */
+	#define HWRM_FUNC_RESET_INPUT_ENABLES_VF_ID_VALID \
+							UINT32_C(0x1)
+	uint32_t enables;
+
+	/*
+	 * The ID of the VF that this PF is trying to reset. Only the parent PF
+	 * shall be allowed to reset a child VF. A parent PF driver shall use
+	 * this field only when a specific child VF is requested to be reset.
+	 */
+	uint16_t vf_id;
+
+	/* This value indicates the level of a function reset. */
+		/*
+		 * Reset the caller function and its children VFs (if any). If
+		 * no children functions exist, then reset the caller function
+		 * only.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETALL \
+							(UINT32_C(0x0) << 0)
+		/* Reset the caller function only */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETME \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Reset all children VFs of the caller function driver if the
+		 * caller is a PF driver. It is an error to specify this level
+		 * by a VF driver. It is an error to specify this level by a PF
+		 * driver with no children VFs.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETCHILDREN \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Reset a specific VF of the caller function driver if the
+		 * caller is the parent PF driver. It is an error to specify
+		 * this level by a VF driver. It is an error to specify this
+		 * level by a PF driver that is not the parent of the VF that is
+		 * being requested to reset.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETVF \
+							(UINT32_C(0x3) << 0)
+	uint8_t func_reset_level;
+
+	uint8_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_func_reset_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_port_phy_cfg */
 /*
  * Description: This command configures the PHY device for the port. It allows
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 17/38] bnxt: add HWRM vnic alloc function
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (14 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 16/38] bnxt: add HWRM function reset command Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 18/38] bnxt: add HWRM vnic free function Stephen Hurd
                     ` (23 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This requires a group info array in struct bnxt, so add that, we can save
the max size from the func_qcap response, and alloc/free in init/uninit

A VNIC represents a virtual interface.
It is a resource in the RX path of the chip and is used to setup various
target actions such as RSS, MAC filtering etc.. for the physical function
in use.

New HWRM call:
bnxt_hwrm_vnic_alloc:
	Allocates a VNIC resource in the hardware.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt.h                |  2 +
 drivers/net/bnxt/bnxt_hwrm.c           | 32 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 99 ++++++++++++++++++++++++++++++++++
 4 files changed, 135 insertions(+)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index f7cf9d1..df1f771 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -141,6 +141,8 @@ struct bnxt {
 
 	/* Default completion ring */
 	struct bnxt_cp_ring_info	*def_cp_ring;
+	uint32_t		max_ring_grps;
+	struct bnxt_ring_grp_info	*grp_info;
 
 	unsigned int		nr_vnics;
 
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index d1ca60a..fb0b7bf 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -43,6 +43,7 @@
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_ring.h"
 #include "bnxt_txq.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
@@ -235,6 +236,7 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 
 	HWRM_CHECK_RESULT;
 
+	bp->max_ring_grps = rte_le_to_cpu_32(resp->max_hw_ring_grps);
 	if (BNXT_PF(bp)) {
 		struct bnxt_pf_info *pf = &bp->pf;
 
@@ -521,6 +523,36 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0, i, j;
+	struct hwrm_vnic_alloc_input req = {.req_type = 0 };
+	struct hwrm_vnic_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	/* map ring groups to this vnic */
+	for (i = vnic->start_grp_id, j = 0; i <= vnic->end_grp_id; i++, j++) {
+		if (bp->grp_info[i].fw_grp_id == (uint16_t)HWRM_NA_SIGNATURE) {
+			RTE_LOG(ERR, PMD,
+				"Not enough ring groups avail:%x req:%x\n", j,
+				(vnic->end_grp_id - vnic->start_grp_id) + 1);
+			break;
+		}
+		vnic->fw_grp_ids[j] = bp->grp_info[i].fw_grp_id;
+	}
+
+	vnic->fw_rss_cos_lb_ctx = (uint16_t)HWRM_NA_SIGNATURE;
+	vnic->ctx_is_rss_cos_lb = HW_CONTEXT_NONE;
+
+	HWRM_PREP(req, VNIC_ALLOC, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_vnic_id = rte_le_to_cpu_16(resp->vnic_id);
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 2679d09..7d09773 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,6 +63,8 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 6209368..eedd368 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -89,6 +89,7 @@ struct ctx_hw_stats64 {
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+#define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3119,6 +3120,104 @@ struct hwrm_stat_ctx_clr_stats_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_alloc */
+/*
+ * Description: This VNIC is a resource in the RX side of the chip that is used
+ * to represent a virtual host "interface". # At the time of VNIC allocation or
+ * configuration, the function can specify whether it wants the requested VNIC
+ * to be the default VNIC for the function or not. # If a function requests
+ * allocation of a VNIC for the first time and a VNIC is successfully allocated
+ * by the HWRM, then the HWRM shall make the allocated VNIC as the default VNIC
+ * for that function. # The default VNIC shall be used for the default action
+ * for a partition or function. # For each VNIC allocated on a function, a
+ * mapping on the RX side to map the allocated VNIC to source virtual interface
+ * shall be performed by the HWRM. This should be hidden to the function driver
+ * requesting the VNIC allocation. This enables broadcast/multicast replication
+ * with source knockout. # If multicast replication with source knockout is
+ * enabled, then the internal VNIC to SVIF mapping data structures shall be
+ * programmed at the time of VNIC allocation.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_vnic_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', this VNIC is requested to be the default VNIC
+	 * for this function.
+	 */
+	#define HWRM_VNIC_ALLOC_INPUT_FLAGS_DEFAULT                UINT32_C(0x1)
+	uint32_t flags;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* Logical vnic ID */
+	uint32_t vnic_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 18/38] bnxt: add HWRM vnic free function
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (15 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 17/38] bnxt: add HWRM vnic alloc function Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 19/38] bnxt: add HWRM vnic configure function Stephen Hurd
                     ` (22 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

In this patch we add a new HWRM API to free a VNIC.

A VNIC represents a virtual interface.
It is a resource in the RX path of the chip and is used to setup
various target actions such as RSS, MAC filtering etc.. for the physical
function in use.

New HWRM call:
bnxt_hwrm_vnic_free:
	Frees a vnic allocated by the bnxt_hwrm_vnic_alloc() function.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           | 21 +++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 82 ++++++++++++++++++++++++++++++++++
 3 files changed, 104 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index fb0b7bf..fbbb81d 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -553,6 +553,27 @@ int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_free_input req = {.req_type = 0 };
+	struct hwrm_vnic_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (vnic->fw_vnic_id == INVALID_HW_RING_ID)
+		return rc;
+
+	HWRM_PREP(req, VNIC_FREE, -1, resp);
+
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_vnic_id = INVALID_HW_RING_ID;
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 7d09773..93fb759 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,6 +63,7 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index eedd368..0771897 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -90,6 +90,7 @@ struct ctx_hw_stats64 {
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
+#define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3218,6 +3219,87 @@ struct hwrm_vnic_alloc_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_free */
+/*
+ * Description: Free a VNIC resource. Idle any resources associated with the
+ * VNIC as well as the VNIC. Reset and release all resources associated with the
+ * VNIC.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_vnic_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Logical vnic ID */
+	uint32_t vnic_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 19/38] bnxt: add HWRM vnic configure function
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (16 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 18/38] bnxt: add HWRM vnic free function Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 20/38] bnxt: add API to allow configuration of vnic Stephen Hurd
                     ` (21 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

This patch configures the properties and actions of the vnic
allocated by vnic_alloc function from the previous patch.

bnxt_hwrm_vnic_cfg:
	Configure the VNIC structure in hardware.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           |  34 ++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +-
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 155 +++++++++++++++++++++++++++++++++
 3 files changed, 191 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index fbbb81d..0c54129 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -553,6 +553,40 @@ int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_cfg_input req = {.req_type = 0 };
+	struct hwrm_vnic_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_CFG, -1, resp);
+
+	/* Only RSS support for now TBD: COS & LB */
+	req.enables =
+	    rte_cpu_to_le_32(HWRM_VNIC_CFG_INPUT_ENABLES_DFLT_RING_GRP |
+			     HWRM_VNIC_CFG_INPUT_ENABLES_RSS_RULE |
+			     HWRM_VNIC_CFG_INPUT_ENABLES_MRU);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+	req.dflt_ring_grp =
+		rte_cpu_to_le_16(bp->grp_info[vnic->start_grp_id].fw_grp_id);
+	req.rss_rule = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+	req.cos_rule = rte_cpu_to_le_16(0xffff);
+	req.lb_rule = rte_cpu_to_le_16(0xffff);
+	req.mru = rte_cpu_to_le_16(bp->eth_dev->data->mtu + ETHER_HDR_LEN +
+				   ETHER_CRC_LEN + VLAN_TAG_SIZE);
+	if (vnic->func_default)
+		req.flags = 1;
+	if (vnic->vlan_strip)
+		req.flags |=
+		    rte_cpu_to_le_32(HWRM_VNIC_CFG_INPUT_FLAGS_VLAN_STRIP_MODE);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 93fb759..cdaa295 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,8 +63,9 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
-int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 0771897..ef0b37a 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -91,6 +91,7 @@ struct ctx_hw_stats64 {
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
+#define HWRM_VNIC_CFG			(UINT32_C(0x42))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3219,6 +3220,160 @@ struct hwrm_vnic_alloc_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_cfg */
+/* Description: Configure the RX VNIC structure. */
+
+/* Input (40 bytes) */
+struct hwrm_vnic_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the VNIC is requested to be the default VNIC
+	 * for the function.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_DEFAULT		UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the VNIC is being configured to strip VLAN in
+	 * the RX path. If set to '0', then VLAN stripping is disabled on this
+	 * VNIC.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_VLAN_STRIP_MODE	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the VNIC is being configured to buffer receive
+	 * packets in the hardware until the host posts new receive buffers. If
+	 * set to '0', then bd_stall is being configured to be disabled on this
+	 * VNIC.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_BD_STALL_MODE		UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the VNIC is being configured to receive both
+	 * RoCE and non-RoCE traffic. If set to '0', then this VNIC is not
+	 * configured to be operating in dual VNIC mode.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_ROCE_DUAL_VNIC_MODE	UINT32_C(0x8)
+	/*
+	 * When this flag is set to '1', the VNIC is requested to be configured
+	 * to receive only RoCE traffic. If this flag is set to '0', then this
+	 * flag shall be ignored by the HWRM. If roce_dual_vnic_mode flag is set
+	 * to '1', then the HWRM client shall not set this flag to '1'.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_ROCE_ONLY_VNIC_MODE	UINT32_C(0x10)
+	uint32_t flags;
+
+	/* This bit must be '1' for the dflt_ring_grp field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_DFLT_RING_GRP	UINT32_C(0x1)
+	/* This bit must be '1' for the rss_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_RSS_RULE		UINT32_C(0x2)
+	/* This bit must be '1' for the cos_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_COS_RULE		UINT32_C(0x4)
+	/* This bit must be '1' for the lb_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_LB_RULE		UINT32_C(0x8)
+	/* This bit must be '1' for the mru field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_MRU			UINT32_C(0x10)
+	uint32_t enables;
+
+	/* Logical vnic ID */
+	uint16_t vnic_id;
+
+	/*
+	 * Default Completion ring for the VNIC. This ring will be chosen if
+	 * packet does not match any RSS rules and if there is no COS rule.
+	 */
+	uint16_t dflt_ring_grp;
+
+	/*
+	 * RSS ID for RSS rule/table structure. 0xFF... (All Fs) if there is no
+	 * RSS rule.
+	 */
+	uint16_t rss_rule;
+
+	/*
+	 * RSS ID for COS rule/table structure. 0xFF... (All Fs) if there is no
+	 * COS rule.
+	 */
+	uint16_t cos_rule;
+
+	/*
+	 * RSS ID for load balancing rule/table structure. 0xFF... (All Fs) if
+	 * there is no LB rule.
+	 */
+	uint16_t lb_rule;
+
+	/*
+	 * The maximum receive unit of the vnic. Each vnic is associated with a
+	 * function. The vnic mru value overwrites the mru setting of the
+	 * associated function. The HWRM shall make sure that vnic mru does not
+	 * exceed the mru of the port the function is associated with.
+	 */
+	uint16_t mru;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_free */
 /*
  * Description: Free a VNIC resource. Idle any resources associated with the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 20/38] bnxt: add API to allow configuration of vnic
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (17 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 19/38] bnxt: add HWRM vnic configure function Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 21/38] bnxt: add HWRM API to configure RSS Stephen Hurd
                     ` (20 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds APIs to allow configuration of a VNIC.
The functions alloc and free the Class of Service or COS and
Load Balance context corresponding to the VNIC in the chip.

New HWRM calls:
bnxt_hwrm_vnic_ctx_alloc:
	Used to allocate COS/Load Balance context of VNIC
bnxt_hwrm_vnic_ctx_free:
	Used to free COS/Load Balance context of VNIC

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
More description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           |  38 ++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 153 +++++++++++++++++++++++++++++++++
 3 files changed, 193 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 0c54129..5a5063f 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -587,6 +587,44 @@ int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cos_lb_ctx_alloc_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cos_lb_ctx_alloc_output *resp =
+						bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_COS_LB_CTX_ALLOC, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_rss_cos_lb_ctx = rte_le_to_cpu_16(resp->rss_cos_lb_ctx_id);
+
+	return rc;
+}
+
+int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cos_lb_ctx_free_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cos_lb_ctx_free_output *resp =
+						bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_COS_LB_CTX_FREE, -1, resp);
+
+	req.rss_cos_lb_ctx_id = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_rss_cos_lb_ctx = INVALID_HW_RING_ID;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index cdaa295..fbb8aba 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -65,6 +65,8 @@ int bnxt_hwrm_ver_get(struct bnxt *bp);
 
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ef0b37a..6412df2 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -92,6 +92,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
+#define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
+#define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3576,6 +3578,157 @@ struct hwrm_vnic_rss_cfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* Input (16 bytes) */
+struct hwrm_vnic_rss_cos_lb_ctx_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_vnic_rss_cos_lb_ctx_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* rss_cos_lb_ctx_id is 16 b */
+	uint16_t rss_cos_lb_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+	uint8_t unused_4;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_vnic_rss_cos_lb_ctx_free */
+/* Description: This function can be used to free COS/Load Balance context. */
+/* Input (24 bytes) */
+
+struct hwrm_vnic_rss_cos_lb_ctx_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* rss_cos_lb_ctx_id is 16 b */
+	uint16_t rss_cos_lb_ctx_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_rss_cos_lb_ctx_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* Output (32 bytes) */
 struct hwrm_queue_qportcfg_output {
 	/*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 21/38] bnxt: add HWRM API to configure RSS
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (18 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 20/38] bnxt: add API to allow configuration of vnic Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 22/38] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
                     ` (19 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

New HWRM call:
bnxt_hwrm_vnic_rss_cfg:
	Used to enable RSS configuration of the VNIC.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Expand the commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           | 24 ++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  2 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h |  1 +
 3 files changed, 27 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 5a5063f..29b0729 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -646,6 +646,30 @@ int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
+			   struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cfg_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_CFG, -1, resp);
+
+	req.hash_type = rte_cpu_to_le_32(vnic->hash_type);
+
+	req.ring_grp_tbl_addr =
+	    rte_cpu_to_le_64(vnic->rss_table_dma_addr);
+	req.hash_key_tbl_addr =
+	    rte_cpu_to_le_64(vnic->rss_hash_key_dma_addr);
+	req.rss_ctx_idx = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index fbb8aba..8c7b939 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -68,6 +68,8 @@ int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
+			   struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 6412df2..72d4984 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -92,6 +92,7 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
+#define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 22/38] bnxt: add API for L2 Rx mask set/clear functions
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (19 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 21/38] bnxt: add HWRM API to configure RSS Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 23/38] bnxt: add HWRM API for stats context allocation Stephen Hurd
                     ` (18 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add API to set/clear L2 Rx mask.

New HWRM calls:
bnxt_hwrm_cfa_l2_clear_rx_mask
bnxt_hwrm_cfa_l2_set_rx_mask

These HWRM APIs allow setting and clearing of Rx masks in L2 context
per VNIC.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           |  45 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 135 +++++++++++++++++++++++++++++++++
 3 files changed, 183 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 29b0729..2c43159 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -141,6 +141,51 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_set_rx_mask_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_set_rx_mask_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, CFA_L2_SET_RX_MASK, -1, resp);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+	req.mask = 0;
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
+int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_set_rx_mask_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_set_rx_mask_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t mask = 0;
+
+	HWRM_PREP(req, CFA_L2_SET_RX_MASK, -1, resp);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	/* FIXME add multicast flag, when multicast adding options is supported
+	 * by ethtool.
+	 */
+	if (vnic->flags & BNXT_VNIC_INFO_PROMISC)
+		mask = HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_PROMISCUOUS;
+	if (vnic->flags & BNXT_VNIC_INFO_ALLMULTI)
+		mask = HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_ALL_MCAST;
+	req.mask = rte_cpu_to_le_32(HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_MCAST |
+				    HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_BCAST |
+				    mask);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter)
 {
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 8c7b939..8037242 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -43,6 +43,9 @@ struct bnxt_cp_ring_info;
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp,
+				   struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter);
 int bnxt_hwrm_set_filter(struct bnxt *bp,
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 72d4984..f8f6a3f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1732,6 +1732,141 @@ struct hwrm_cfa_l2_filter_free_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_cfa_l2_set_rx_mask */
+/* Description: This command will set rx mask of the function. */
+
+/* Input (40 bytes) */
+struct hwrm_cfa_l2_set_rx_mask_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* VNIC ID */
+	uint32_t vnic_id;
+
+	/* Reserved for future use. */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_RESERVED	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the function is requested to accept multi-cast
+	 * packets specified by the multicast addr table.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_MCAST	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the function is requested to accept all multi-
+	 * cast packets.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_ALL_MCAST	UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the function is requested to accept broadcast
+	 * packets.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_BCAST	UINT32_C(0x8)
+	/*
+	 * When this bit is '1', the function is requested to be put in the
+	 * promiscuous mode. The HWRM should accept any function to set up
+	 * promiscuous mode. The HWRM shall follow the semantics below for the
+	 * promiscuous mode support. # When partitioning is not enabled on a
+	 * port (i.e. single PF on the port), then the PF shall be allowed to be
+	 * in the promiscuous mode. When the PF is in the promiscuous mode, then
+	 * it shall receive all host bound traffic on that port. # When
+	 * partitioning is enabled on a port (i.e. multiple PFs per port) and a
+	 * PF on that port is in the promiscuous mode, then the PF receives all
+	 * traffic within that partition as identified by a unique identifier
+	 * for the PF (e.g. S-Tag). If a unique outer VLAN for the PF is
+	 * specified, then the setting of promiscuous mode on that PF shall
+	 * result in the PF receiving all host bound traffic with matching outer
+	 * VLAN. # A VF shall can be set in the promiscuous mode. In the
+	 * promiscuous mode, the VF does not receive any traffic unless a unique
+	 * outer VLAN for the VF is specified. If a unique outer VLAN for the VF
+	 * is specified, then the setting of promiscuous mode on that VF shall
+	 * result in the VF receiving all host bound traffic with the matching
+	 * outer VLAN. # The HWRM shall allow the setting of promiscuous mode on
+	 * a function independently from the promiscuous mode settings on other
+	 * functions.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_PROMISCUOUS	UINT32_C(0x10)
+	/*
+	 * If this flag is set, the corresponding RX filters shall be set up to
+	 * cover multicast/broadcast filters for the outermost Layer 2
+	 * destination MAC address field.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_OUTERMOST	UINT32_C(0x20)
+	uint32_t mask;
+
+	/* This is the address for mcast address tbl. */
+	uint64_t mc_tbl_addr;
+
+	/*
+	 * This value indicates how many entries in mc_tbl are valid. Each entry
+	 * is 6 bytes.
+	 */
+	uint32_t num_mc_entries;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_cfa_l2_set_rx_mask_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_exec_fwd_resp */
 /*
  * Description: This command is used to send an encapsulated request to the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 23/38] bnxt: add HWRM API for stats context allocation
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (20 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 22/38] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 24/38] bnxt: add HWRM ring alloc/free functions Stephen Hurd
                     ` (17 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API code to allocate a statistics context in the ASIC.
This API will be called by the previously submitted "add statistics
operations patch".

New HWRM call:
bnxt_hwrm_stat_ctx_alloc:
	This command allocates and does basic preparation for a stat
context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 52 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  3 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 89 ++++++++++++++++++++++++++++++++++
 3 files changed, 144 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 2c43159..420a1cc 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -568,6 +568,31 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 	return rc;
 }
 
+int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
+			     struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	int rc;
+	struct hwrm_stat_ctx_alloc_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_ALLOC, -1, resp);
+
+	req.update_period_ms = rte_cpu_to_le_32(1000);
+
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+	req.stats_dma_addr =
+	    rte_cpu_to_le_64(cpr->hw_stats_map);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	cpr->hw_stats_ctx_id = rte_le_to_cpu_16(resp->stat_ctx_id);
+	bp->grp_info[idx].fw_stats_ctx = cpr->hw_stats_ctx_id;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0, i, j;
@@ -744,6 +769,33 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq;
+		struct bnxt_rx_queue *rxq;
+		struct bnxt_cp_ring_info *cpr;
+		unsigned int idx = i + 1;
+
+		if (i >= bp->rx_cp_nr_rings) {
+			txq = bp->tx_queues[i - bp->rx_cp_nr_rings];
+			cpr = txq->cp_ring;
+		} else {
+			rxq = bp->rx_queues[i];
+			cpr = rxq->cp_ring;
+		}
+
+		rc = bnxt_hwrm_stat_ctx_alloc(bp, cpr, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 8037242..49966f5 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,6 +63,8 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
+int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
+			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
@@ -74,6 +76,7 @@ int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 			   struct bnxt_vnic_info *vnic);
 
+int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index f8f6a3f..28362c9 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -99,6 +99,7 @@ struct ctx_hw_stats64 {
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_STAT_CTX_ALLOC		(UINT32_C(0xb0))
 #define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
@@ -3183,6 +3184,94 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_alloc */
+/*
+ * Description: This command allocates and does basic preparation for a stat
+ * context.
+ */
+
+/* Input (32 bytes) */
+struct hwrm_stat_ctx_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This is the address for statistic block. */
+	uint64_t stats_dma_addr;
+
+	/*
+	 * The statistic block update period in ms. e.g. 250ms, 500ms, 750ms,
+	 * 1000ms.
+	 */
+	uint32_t update_period_ms;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_stat_ctx_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This is the statistics context ID value. */
+	uint32_t stat_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_clr_stats */
 /* Description: This command clears statistics of a context. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 24/38] bnxt: add HWRM ring alloc/free functions
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (21 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 23/38] bnxt: add HWRM API for stats context allocation Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 25/38] bnxt: add ring group " Stephen Hurd
                     ` (16 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API calls to allocate and free TX, RX and Completion rings
in the hardware along with the associated structs and definitions.

This informs the hardware of how the specific rings were set up in the
host and allocates them in the HWRM, setting up the doorbell registers
etc. as needed, returning an ID for the ring.

Basic ring alloc/free calls:
bnxt_hwrm_ring_alloc
	This command allocates and does basic preparation for a ring.
bnxt_hwrm_ring_free
	This command is used to free a ring and associated resources.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 108 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   7 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 305 +++++++++++++++++++++++++++++++++
 3 files changed, 420 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 420a1cc..565b76f 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -547,6 +547,114 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_ring_alloc(struct bnxt *bp,
+			 struct bnxt_ring *ring,
+			 uint32_t ring_type, uint32_t map_index,
+			 uint32_t stats_ctx_id)
+{
+	int rc = 0;
+	struct hwrm_ring_alloc_input req = {.req_type = 0 };
+	struct hwrm_ring_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_ALLOC, -1, resp);
+
+	req.enables = rte_cpu_to_le_32(0);
+
+	req.page_tbl_addr = rte_cpu_to_le_64(ring->bd_dma);
+	req.fbo = rte_cpu_to_le_32(0);
+	/* Association of ring index with doorbell index */
+	req.logical_id = rte_cpu_to_le_16(map_index);
+
+	switch (ring_type) {
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_TX:
+		req.queue_id = bp->cos_queue[0].id;
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_RX:
+		req.ring_type = ring_type;
+		req.cmpl_ring_id =
+		    rte_cpu_to_le_16(bp->grp_info[map_index].cp_fw_ring_id);
+		req.length = rte_cpu_to_le_32(ring->ring_size);
+		req.stat_ctx_id = rte_cpu_to_le_16(stats_ctx_id);
+		req.enables = rte_cpu_to_le_32(rte_le_to_cpu_32(req.enables) |
+			HWRM_RING_ALLOC_INPUT_ENABLES_STAT_CTX_ID_VALID);
+		break;
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL:
+		req.ring_type = ring_type;
+		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_POLL;
+		req.length = rte_cpu_to_le_32(ring->ring_size);
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "hwrm alloc invalid ring type %d\n",
+			ring_type);
+		return -1;
+	}
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	if (rc || resp->error_code) {
+		if (rc == 0 && resp->error_code)
+			rc = rte_le_to_cpu_16(resp->error_code);
+		switch (ring_type) {
+		case HWRM_RING_FREE_INPUT_RING_TYPE_CMPL:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc cp failed. rc:%d\n", rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_RX:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc rx failed. rc:%d\n", rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_TX:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc tx failed. rc:%d\n", rc);
+			return rc;
+		default:
+			RTE_LOG(ERR, PMD, "Invalid ring. rc:%d\n", rc);
+			return rc;
+		}
+	}
+
+	ring->fw_ring_id = rte_le_to_cpu_16(resp->ring_id);
+	return rc;
+}
+
+int bnxt_hwrm_ring_free(struct bnxt *bp,
+			struct bnxt_ring *ring, uint32_t ring_type)
+{
+	int rc;
+	struct hwrm_ring_free_input req = {.req_type = 0 };
+	struct hwrm_ring_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_FREE, -1, resp);
+
+	req.ring_type = ring_type;
+	req.ring_id = rte_cpu_to_le_16(ring->fw_ring_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	if (rc || resp->error_code) {
+		if (rc == 0 && resp->error_code)
+			rc = rte_le_to_cpu_16(resp->error_code);
+
+		switch (ring_type) {
+		case HWRM_RING_FREE_INPUT_RING_TYPE_CMPL:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free cp failed. rc:%d\n",
+				rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_RX:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free rx failed. rc:%d\n",
+				rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_TX:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free tx failed. rc:%d\n",
+				rc);
+			return rc;
+		default:
+			RTE_LOG(ERR, PMD, "Invalid ring, rc:%d\n", rc);
+			return rc;
+		}
+	}
+	return 0;
+}
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 49966f5..ebde5fb 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -62,6 +62,13 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
+int bnxt_hwrm_ring_alloc(struct bnxt *bp,
+			 struct bnxt_ring *ring,
+			 uint32_t ring_type, uint32_t map_index,
+			 uint32_t stats_ctx_id);
+int bnxt_hwrm_ring_free(struct bnxt *bp,
+			struct bnxt_ring *ring, uint32_t ring_type);
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 28362c9..e6280b6 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -93,6 +93,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
 #define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
+#define HWRM_RING_ALLOC			(UINT32_C(0x50))
+#define HWRM_RING_FREE			(UINT32_C(0x51))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
@@ -3184,6 +3186,309 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_ring_alloc */
+/*
+ * Description: This command allocates and does basic preparation for a ring.
+ */
+
+/* Input (80 bytes) */
+struct hwrm_ring_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This bit must be '1' for the Reserved1 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED1		UINT32_C(0x1)
+	/* This bit must be '1' for the Reserved2 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED2		UINT32_C(0x2)
+	/* This bit must be '1' for the Reserved3 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED3		UINT32_C(0x4)
+	/*
+	 * This bit must be '1' for the stat_ctx_id_valid field to be
+	 * configured.
+	 */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_STAT_CTX_ID_VALID	UINT32_C(0x8)
+	/* This bit must be '1' for the Reserved4 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED4		UINT32_C(0x10)
+	/* This bit must be '1' for the max_bw_valid field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_MAX_BW_VALID	UINT32_C(0x20)
+	uint32_t enables;
+
+	/* Ring Type. */
+		/* Completion Ring (CR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL	(UINT32_C(0x0) << 0)
+		/* TX Ring (TR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_TX	(UINT32_C(0x1) << 0)
+		/* RX Ring (RR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_RX	(UINT32_C(0x2) << 0)
+	uint8_t ring_type;
+
+	uint8_t unused_0;
+	uint16_t unused_1;
+
+	/* This value is a pointer to the page table for the Ring. */
+	uint64_t page_tbl_addr;
+
+	/* First Byte Offset of the first entry in the first page. */
+	uint32_t fbo;
+
+	/*
+	 * Actual page size in 2^page_size. The supported range is increments in
+	 * powers of 2 from 16 bytes to 1GB. - 4 = 16 B Page size is 16 B. - 12
+	 * = 4 KB Page size is 4 KB. - 13 = 8 KB Page size is 8 KB. - 16 = 64 KB
+	 * Page size is 64 KB. - 22 = 2 MB Page size is 2 MB. - 23 = 4 MB Page
+	 * size is 4 MB. - 31 = 1 GB Page size is 1 GB.
+	 */
+	uint8_t page_size;
+
+	/*
+	 * This value indicates the depth of page table. For this version of the
+	 * specification, value other than 0 or 1 shall be considered as an
+	 * invalid value. When the page_tbl_depth = 0, then it is treated as a
+	 * special case with the following. 1. FBO and page size fields are not
+	 * valid. 2. page_tbl_addr is the physical address of the first element
+	 * of the ring.
+	 */
+	uint8_t page_tbl_depth;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * Number of 16B units in the ring. Minimum size for a ring is 16 16B
+	 * entries.
+	 */
+	uint32_t length;
+
+	/*
+	 * Logical ring number for the ring to be allocated. This value
+	 * determines the position in the doorbell area where the update to the
+	 * ring will be made. For completion rings, this value is also the MSI-X
+	 * vector number for the function the completion ring is associated
+	 * with.
+	 */
+	uint16_t logical_id;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This value
+	 * indicates what completion ring the TX ring is associated with.
+	 */
+	uint16_t cmpl_ring_id;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This value
+	 * indicates what CoS queue the TX ring is associated with.
+	 */
+	uint16_t queue_id;
+
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved1;
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint16_t reserved2;
+
+	uint8_t unused_6;
+	uint8_t unused_7;
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved3;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This input
+	 * indicates what statistics context this ring should be associated
+	 * with.
+	 */
+	uint32_t stat_ctx_id;
+
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved4;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. Maximum BW
+	 * allocated to this TX ring in Mbps. The HWRM will translate this value
+	 * into byte counter and time interval used for this ring inside the
+	 * device.
+	 */
+	uint32_t max_bw;
+
+	/*
+	 * This field is used only when ring_type is a Completion ring. This
+	 * value indicates what interrupt mode should be used on this completion
+	 * ring. Note: In the legacy interrupt mode, no more than 16 completion
+	 * rings are allowed.
+	 */
+		/* Legacy INTA */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_LEGACY	(UINT32_C(0x0) << 0)
+		/* Reserved */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_RSVD	(UINT32_C(0x1) << 0)
+		/* MSI-X */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_MSIX	(UINT32_C(0x2) << 0)
+		/* No Interrupt - Polled mode */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_POLL	(UINT32_C(0x3) << 0)
+	uint8_t int_mode;
+
+	uint8_t unused_8[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_ring_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* Physical number of ring allocated. */
+	uint16_t ring_id;
+
+	/* Logical number of ring allocated. */
+	uint16_t logical_ring_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ring_free */
+/*
+ * Description: This command is used to free a ring and associated resources.
+ */
+/* Input (24 bytes) */
+
+struct hwrm_ring_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Ring Type. */
+		/* Completion Ring (CR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_CMPL	(UINT32_C(0x0) << 0)
+		/* TX Ring (TR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_TX	(UINT32_C(0x1) << 0)
+		/* RX Ring (RR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_RX	(UINT32_C(0x2) << 0)
+	uint8_t ring_type;
+
+	uint8_t unused_0;
+
+	/* Physical number of ring allocated. */
+	uint16_t ring_id;
+
+	uint32_t unused_1;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_alloc */
 /*
  * Description: This command allocates and does basic preparation for a stat
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 25/38] bnxt: add ring group alloc/free functions
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (22 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 24/38] bnxt: add HWRM ring alloc/free functions Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 26/38] bnxt: add HWRM stat context free function Stephen Hurd
                     ` (15 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API for ring group alloc/free functions, associated structs and
definitions.
This API allocates and does basic preparation for a ring group in ASIC.
A ring group is identified by an index. It consists of Rx ring id,
completion ring id and a statistics context.

New HWRM calls:
bnxt_hwrm_ring_grp_alloc
	Allocates and does basic preparation for a ring group
bnxt_hwrm_ring_grp_free
	Frees and does cleansup resources of a ring group

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address issues pointed out by checkpatch.

v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           |  84 +++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   4 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 185 +++++++++++++++++++++++++++++++++
 3 files changed, 273 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 565b76f..f335401 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -655,6 +655,47 @@ int bnxt_hwrm_ring_free(struct bnxt *bp,
 	return 0;
 }
 
+int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx)
+{
+	int rc = 0;
+	struct hwrm_ring_grp_alloc_input req = {.req_type = 0 };
+	struct hwrm_ring_grp_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_GRP_ALLOC, -1, resp);
+
+	req.cr = rte_cpu_to_le_16(bp->grp_info[idx].cp_fw_ring_id);
+	req.rr = rte_cpu_to_le_16(bp->grp_info[idx].rx_fw_ring_id);
+	req.ar = rte_cpu_to_le_16(bp->grp_info[idx].ag_fw_ring_id);
+	req.sc = rte_cpu_to_le_16(bp->grp_info[idx].fw_stats_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->grp_info[idx].fw_grp_id =
+	    rte_le_to_cpu_16(resp->ring_group_id);
+
+	return rc;
+}
+
+int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx)
+{
+	int rc;
+	struct hwrm_ring_grp_free_input req = {.req_type = 0 };
+	struct hwrm_ring_grp_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_GRP_FREE, -1, resp);
+
+	req.ring_group_id = rte_cpu_to_le_16(bp->grp_info[idx].fw_grp_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->grp_info[idx].fw_grp_id = INVALID_HW_RING_ID;
+	return rc;
+}
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 {
 	int rc = 0;
@@ -904,6 +945,49 @@ int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
+{
+	uint16_t i;
+	uint32_t rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (bp->grp_info[idx].fw_grp_id == INVALID_HW_RING_ID) {
+			RTE_LOG(ERR, PMD,
+				"Attempt to free invalid ring group %d\n",
+				idx);
+			continue;
+		}
+
+		rc = bnxt_hwrm_ring_grp_free(bp, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
+int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp)
+{
+	uint16_t i;
+	uint32_t rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (bp->grp_info[idx].cp_fw_ring_id == INVALID_HW_RING_ID ||
+		    bp->grp_info[idx].rx_fw_ring_id == INVALID_HW_RING_ID)
+			continue;
+
+		rc = bnxt_hwrm_ring_grp_alloc(bp, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index ebde5fb..a9a8dc6 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -68,6 +68,8 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 			 uint32_t stats_ctx_id);
 int bnxt_hwrm_ring_free(struct bnxt *bp,
 			struct bnxt_ring *ring, uint32_t ring_type);
+int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx);
+int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
@@ -85,6 +87,8 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
+int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index e6280b6..4e2eb9f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -95,6 +95,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
 #define HWRM_RING_ALLOC			(UINT32_C(0x50))
 #define HWRM_RING_FREE			(UINT32_C(0x51))
+#define HWRM_RING_GRP_ALLOC		(UINT32_C(0x60))
+#define HWRM_RING_GRP_FREE		(UINT32_C(0x61))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
@@ -3489,6 +3491,189 @@ struct hwrm_ring_free_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_ring_grp_alloc */
+/*
+ * Description: This API allocates and does basic preparation for a ring group.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ring_grp_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This value identifies the CR associated with the ring group. */
+	uint16_t cr;
+
+	/* This value identifies the main RR associated with the ring group. */
+	uint16_t rr;
+
+	/*
+	 * This value identifies the aggregation RR associated with the ring
+	 * group. If this value is 0xFF... (All Fs), then no Aggregation ring
+	 * will be set.
+	 */
+	uint16_t ar;
+
+	/*
+	 * This value identifies the statistics context associated with the ring
+	 * group.
+	 */
+	uint16_t sc;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_grp_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This is the ring group ID value. Use this value to program the
+	 * default ring group for the VNIC or as table entries in an RSS/COS
+	 * context.
+	 */
+	uint32_t ring_group_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ring_grp_free */
+/*
+ * Description: This API frees a ring group and associated resources. # If a
+ * ring in the ring group is reset or free, then the associated rings in the
+ * ring group shall also be reset/free using hwrm_ring_free. # A function driver
+ * shall always use hwrm_ring_grp_free after freeing all rings in a group. # As
+ * a part of executing this command, the HWRM shall reset all associated ring
+ * group resources.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ring_grp_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This is the ring group ID value. */
+	uint32_t ring_group_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_grp_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_alloc */
 /*
  * Description: This command allocates and does basic preparation for a stat
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 26/38] bnxt: add HWRM stat context free function
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (23 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 25/38] bnxt: add ring group " Stephen Hurd
@ 2016-06-14 22:55   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 27/38] bnxt: add HWRM API to set and clear filters Stephen Hurd
                     ` (14 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:55 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add function and associated structures and definitions to free
statistics context from the ASIC.

New HWRM call:
bnxt_hwrm_stat_ctx_free
	This command is used to free a stat context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 44 ++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  3 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 81 ++++++++++++++++++++++++++++++++++
 3 files changed, 128 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index f335401..cb30109 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -742,6 +742,28 @@ int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 	return rc;
 }
 
+int bnxt_hwrm_stat_ctx_free(struct bnxt *bp,
+			    struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	int rc;
+	struct hwrm_stat_ctx_free_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_FREE, -1, resp);
+
+	req.stat_ctx_id = rte_cpu_to_le_16(cpr->hw_stats_ctx_id);
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	cpr->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
+	bp->grp_info[idx].fw_stats_ctx = cpr->hw_stats_ctx_id;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0, i, j;
@@ -918,6 +940,28 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	int rc;
+	unsigned int i;
+	struct bnxt_cp_ring_info *cpr;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (i >= bp->rx_cp_nr_rings)
+			cpr = bp->tx_queues[i - bp->rx_cp_nr_rings]->cp_ring;
+		else
+			cpr = bp->rx_queues[i]->cp_ring;
+		if (cpr->hw_stats_ctx_id != HWRM_NA_SIGNATURE) {
+			rc = bnxt_hwrm_stat_ctx_free(bp, cpr, idx);
+			if (rc)
+				return rc;
+		}
+	}
+	return 0;
+}
+
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
 {
 	unsigned int i;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index a9a8dc6..ffb45b1 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -74,6 +74,8 @@ int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
+int bnxt_hwrm_stat_ctx_free(struct bnxt *bp,
+			    struct bnxt_cp_ring_info *cpr, unsigned int idx);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
@@ -87,6 +89,7 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 4e2eb9f..d58295a 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -104,6 +104,7 @@ struct ctx_hw_stats64 {
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
 #define HWRM_STAT_CTX_ALLOC		(UINT32_C(0xb0))
+#define HWRM_STAT_CTX_FREE		(UINT32_C(0xb1))
 #define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
@@ -3839,6 +3840,86 @@ struct hwrm_stat_ctx_clr_stats_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_free */
+/* Description: This command is used to free a stat context. */
+/* Input (24 bytes) */
+
+struct hwrm_stat_ctx_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* ID of the statistics context that is being queried. */
+	uint32_t stat_ctx_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_stat_ctx_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This is the statistics context ID value. */
+	uint32_t stat_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_alloc */
 /*
  * Description: This VNIC is a resource in the RX side of the chip that is used
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 27/38] bnxt: add HWRM API to set and clear filters
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (24 preceding siblings ...)
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 26/38] bnxt: add HWRM stat context free function Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 28/38] bnxt: allocate and free all HWRM rings and groups Stephen Hurd
                     ` (13 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

New HWRM call:
bnxt_clear_hwrm_vnic_filters
	This patch adds code to set and clear L2 filters from the
	corresponding VNIC. These filters will determine the Rx flows

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Separated this code from the previous patch as it had nothing to
do with freeing of statistics context.

v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c | 26 ++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h |  3 +++
 2 files changed, 29 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index cb30109..13f9e1c 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -1059,6 +1059,32 @@ int bnxt_alloc_hwrm_resources(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	struct bnxt_filter_info *filter;
+	int rc = 0;
+
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		rc = bnxt_hwrm_clear_filter(bp, filter);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
+int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	struct bnxt_filter_info *filter;
+	int rc = 0;
+
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		rc = bnxt_hwrm_set_filter(bp, vnic, filter);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
 static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
 {
 	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index ffb45b1..9c26118 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -92,6 +92,9 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
+int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+void bnxt_free_all_hwrm_resources(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 28/38] bnxt: allocate and free all HWRM rings and groups
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (25 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 27/38] bnxt: add HWRM API to set and clear filters Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 29/38] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
                     ` (12 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add a top level functions to initialize ring groups, and functions
to allocate and free all the rings via HWRM.

A ring group is identified by an index. It consists of Rx or Tx ring id,
completion ring id and a statistics context. Once a ring group is
initialized, use this group index while creating the rings in the ASIC
using the appropriate HWRM API added via earlier patches.

Functions added:
bnxt_free_cp_ring
	Calls the HWRM function generic ring free with arugments specific
	to a completion ring and sanitizes the host completion structure
bnxt_free_all_hwrm_rings
	Frees all the HWRM allocated hardware rings
bnxt_free_all_hwrm_resources
	Frees all the resources allocated via the HRM in the hardware
bnxt_alloc_hwrm_rings
	Allocates all the HWRM rings needed in the current configuration

This should be the last functionality needed to add start/stop
device operations.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comment to merge another patch into this to avoid
a compilation issue. Fix issues pointed out by checkpatch.

v5:
Reorder the footer.
Remove instances of bnxt_ring_struct
---
 drivers/net/bnxt/bnxt_hwrm.c | 112 +++++++++++++++++++++++++++++++++++++++-
 drivers/net/bnxt/bnxt_hwrm.h |   1 +
 drivers/net/bnxt/bnxt_ring.c | 119 +++++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ring.h |   2 +
 4 files changed, 233 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 13f9e1c..fcc4811 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -43,8 +43,10 @@
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -579,7 +581,11 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 		break;
 	case HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL:
 		req.ring_type = ring_type;
-		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_POLL;
+		/*
+		 * TODO: Some HWRM versions crash with
+		 * HWRM_RING_ALLOC_INPUT_INT_MODE_POLL
+		 */
+		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_MSIX;
 		req.length = rte_cpu_to_le_32(ring->ring_size);
 		break;
 	default:
@@ -1012,6 +1018,84 @@ int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
 	return rc;
 }
 
+static void bnxt_free_cp_ring(struct bnxt *bp,
+			      struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+
+	bnxt_hwrm_ring_free(bp, cp_ring,
+			HWRM_RING_FREE_INPUT_RING_TYPE_CMPL);
+	cp_ring->fw_ring_id = INVALID_HW_RING_ID;
+	bp->grp_info[idx].cp_fw_ring_id = INVALID_HW_RING_ID;
+	memset(cpr->cp_desc_ring, 0, cpr->cp_ring_struct->ring_size *
+			sizeof(*cpr->cp_desc_ring));
+	cpr->cp_raw_cons = 0;
+}
+
+int bnxt_free_all_hwrm_rings(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_tx_ring_info *txr = txq->tx_ring;
+		struct bnxt_ring *ring = txr->tx_ring_struct;
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		unsigned int idx = bp->rx_cp_nr_rings + i + 1;
+
+		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
+			bnxt_hwrm_ring_free(bp, ring,
+					HWRM_RING_FREE_INPUT_RING_TYPE_TX);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
+			memset(txr->tx_desc_ring, 0,
+					txr->tx_ring_struct->ring_size *
+					sizeof(*txr->tx_desc_ring));
+			memset(txr->tx_buf_ring, 0,
+					txr->tx_ring_struct->ring_size *
+					sizeof(*txr->tx_buf_ring));
+			txr->tx_prod = 0;
+			txr->tx_cons = 0;
+		}
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, idx);
+	}
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+		struct bnxt_ring *ring = rxr->rx_ring_struct;
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		unsigned int idx = i + 1;
+
+		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
+			bnxt_hwrm_ring_free(bp, ring,
+					HWRM_RING_FREE_INPUT_RING_TYPE_RX);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
+			bp->grp_info[idx].rx_fw_ring_id = INVALID_HW_RING_ID;
+			memset(rxr->rx_desc_ring, 0,
+					rxr->rx_ring_struct->ring_size *
+					sizeof(*rxr->rx_desc_ring));
+			memset(rxr->rx_buf_ring, 0,
+					rxr->rx_ring_struct->ring_size *
+					sizeof(*rxr->rx_buf_ring));
+			rxr->rx_prod = 0;
+		}
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, idx);
+	}
+
+	/* Default completion ring */
+	{
+		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, 0);
+	}
+
+	return rc;
+}
+
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp)
 {
 	uint16_t i;
@@ -1085,6 +1169,32 @@ int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+void bnxt_free_all_hwrm_resources(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	unsigned int i;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+	bnxt_hwrm_cfa_l2_clear_rx_mask(bp, vnic);
+
+	/* VNIC resources */
+	for (i = 0; i < bp->nr_vnics; i++) {
+		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+		bnxt_clear_hwrm_vnic_filters(bp, vnic);
+
+		bnxt_hwrm_vnic_ctx_free(bp, vnic);
+		bnxt_hwrm_vnic_free(bp, vnic);
+	}
+	/* Ring resources */
+	bnxt_free_all_hwrm_rings(bp);
+	bnxt_free_all_hwrm_ring_grps(bp);
+	bnxt_free_all_hwrm_stat_ctxs(bp);
+}
+
 static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
 {
 	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 9c26118..c28e8d8 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -90,6 +90,7 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_rings(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index 67b009a..3f81ffc 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -35,8 +35,11 @@
 
 #include "bnxt.h"
 #include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
+#include "bnxt_rxq.h"
 #include "bnxt_rxr.h"
+#include "bnxt_txq.h"
 #include "bnxt_txr.h"
 
 #include "hsi_struct_def_dpdk.h"
@@ -55,6 +58,19 @@ void bnxt_free_ring(struct bnxt_ring *ring)
 }
 
 /*
+ * Ring groups
+ */
+
+void bnxt_init_ring_grps(struct bnxt *bp)
+{
+	unsigned int i;
+
+	for (i = 0; i < bp->max_ring_grps; i++)
+		memset(&bp->grp_info[i], (uint8_t)HWRM_NA_SIGNATURE,
+		       sizeof(struct bnxt_ring_grp_info));
+}
+
+/*
  * Allocates a completion ring with vmem and stats optionally also allocating
  * a TX and/or RX ring.  Passing NULL as tx_ring_info and/or rx_ring_info
  * to not allocate them.
@@ -185,3 +201,106 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 	cp_ring_info->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
 	return 0;
 }
+
+/* ring_grp usage:
+ * [0] = default completion ring
+ * [1 -> +rx_cp_nr_rings] = rx_cp, rx rings
+ * [1+rx_cp_nr_rings + 1 -> +tx_cp_nr_rings] = tx_cp, tx rings
+ */
+int bnxt_alloc_hwrm_rings(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	/* Default completion ring */
+	{
+		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					  HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					  0, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+		bp->grp_info[0].cp_fw_ring_id = cp_ring->fw_ring_id;
+	}
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+		struct bnxt_ring *ring = rxr->rx_ring_struct;
+		unsigned int idx = i + 1;
+
+		/* Rx cmpl */
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					idx, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].cp_fw_ring_id = cp_ring->fw_ring_id;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+
+		/* Rx ring */
+		rc = bnxt_hwrm_ring_alloc(bp, ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_RX,
+					idx, cpr->hw_stats_ctx_id);
+		if (rc)
+			goto err_out;
+		rxr->rx_prod = 0;
+		rxr->rx_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].rx_fw_ring_id = ring->fw_ring_id;
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+		if (bnxt_init_one_rx_ring(rxq)) {
+			RTE_LOG(ERR, PMD, "bnxt_init_one_rx_ring failed!");
+			bnxt_rx_queue_release_op(rxq);
+			return -ENOMEM;
+		}
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+	}
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_tx_ring_info *txr = txq->tx_ring;
+		struct bnxt_ring *ring = txr->tx_ring_struct;
+		unsigned int idx = 1 + bp->rx_cp_nr_rings + i;
+
+		/* Tx cmpl */
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					idx, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].cp_fw_ring_id = cp_ring->fw_ring_id;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+
+		/* Tx ring */
+		rc = bnxt_hwrm_ring_alloc(bp, ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_TX,
+					idx, cpr->hw_stats_ctx_id);
+		if (rc)
+			goto err_out;
+
+		txr->tx_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+	}
+
+err_out:
+	return rc;
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index f6ab214..8656549 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -92,10 +92,12 @@ struct bnxt_tx_ring_info;
 struct bnxt_rx_ring_info;
 struct bnxt_cp_ring_info;
 void bnxt_free_ring(struct bnxt_ring *ring);
+void bnxt_init_ring_grps(struct bnxt *bp);
 int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 			    struct bnxt_tx_ring_info *tx_ring_info,
 			    struct bnxt_rx_ring_info *rx_ring_info,
 			    struct bnxt_cp_ring_info *cp_ring_info,
 			    const char *suffix);
+int bnxt_alloc_hwrm_rings(struct bnxt *bp);
 
 #endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 29/38] bnxt: add HWRM port PHY config call and helpers
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (26 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 28/38] bnxt: allocate and free all HWRM rings and groups Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 30/38] bnxt: add start/stop/link update operations Stephen Hurd
                     ` (11 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM calls to query the port's PHY and link configuration.

New HWRM call:
bnxt_hwrm_port_phy_qcfg
	This command queries the PHY configuration for the port

Also adding helper function like bnxt_get_hwrm_link_config()
and bnxt_parse_hw_link_speed() parse the link state.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.

v5:
Reorder footer
Rewrite the patch description.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 120 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 790 +++++++++++++++++++++++++++++++++
 3 files changed, 911 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index fcc4811..5d81a60 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -521,6 +521,43 @@ static int bnxt_hwrm_port_phy_cfg(struct bnxt *bp, struct bnxt_link_info *conf)
 	return rc;
 }
 
+static int bnxt_hwrm_port_phy_qcfg(struct bnxt *bp,
+				   struct bnxt_link_info *link_info)
+{
+	int rc = 0;
+	struct hwrm_port_phy_qcfg_input req = {.req_type = 0};
+	struct hwrm_port_phy_qcfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, PORT_PHY_QCFG, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	link_info->phy_link_status = resp->link;
+	if (link_info->phy_link_status == HWRM_PORT_PHY_QCFG_OUTPUT_LINK_LINK) {
+		link_info->link_up = 1;
+		link_info->link_speed = rte_le_to_cpu_16(resp->link_speed);
+	} else {
+		link_info->link_up = 0;
+		link_info->link_speed = 0;
+	}
+	link_info->duplex = resp->duplex;
+	link_info->pause = resp->pause;
+	link_info->auto_pause = resp->auto_pause;
+	link_info->force_pause = resp->force_pause;
+	link_info->auto_mode = resp->auto_mode;
+
+	link_info->support_speeds = rte_le_to_cpu_16(resp->support_speeds);
+	link_info->auto_link_speed = rte_le_to_cpu_16(resp->auto_link_speed);
+	link_info->preemphasis = rte_le_to_cpu_32(resp->preemphasis);
+	link_info->phy_ver[0] = resp->phy_maj;
+	link_info->phy_ver[1] = resp->phy_min;
+	link_info->phy_ver[2] = resp->phy_bld;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
@@ -1326,6 +1363,89 @@ static uint16_t bnxt_parse_eth_link_speed_mask(uint32_t link_speed)
 	return ret;
 }
 
+static uint32_t bnxt_parse_hw_link_speed(uint16_t hw_link_speed)
+{
+	uint32_t eth_link_speed = ETH_SPEED_NUM_NONE;
+
+	switch (hw_link_speed) {
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB:
+		eth_link_speed = ETH_SPEED_NUM_100M;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_1GB:
+		eth_link_speed = ETH_SPEED_NUM_1G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2_5GB:
+		eth_link_speed = ETH_SPEED_NUM_2_5G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10GB:
+		eth_link_speed = ETH_SPEED_NUM_10G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_20GB:
+		eth_link_speed = ETH_SPEED_NUM_20G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_25GB:
+		eth_link_speed = ETH_SPEED_NUM_25G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_40GB:
+		eth_link_speed = ETH_SPEED_NUM_40G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_50GB:
+		eth_link_speed = ETH_SPEED_NUM_50G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2GB:
+	default:
+		RTE_LOG(ERR, PMD, "HWRM link speed %d not defined\n",
+			hw_link_speed);
+		break;
+	}
+	return eth_link_speed;
+}
+
+static uint16_t bnxt_parse_hw_link_duplex(uint16_t hw_link_duplex)
+{
+	uint16_t eth_link_duplex = ETH_LINK_FULL_DUPLEX;
+
+	switch (hw_link_duplex) {
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH:
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_FULL:
+		eth_link_duplex = ETH_LINK_FULL_DUPLEX;
+		break;
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF:
+		eth_link_duplex = ETH_LINK_HALF_DUPLEX;
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "HWRM link duplex %d not defined\n",
+			hw_link_duplex);
+		break;
+	}
+	return eth_link_duplex;
+}
+
+int bnxt_get_hwrm_link_config(struct bnxt *bp, struct rte_eth_link *link)
+{
+	int rc = 0;
+	struct bnxt_link_info *link_info = &bp->link_info;
+
+	rc = bnxt_hwrm_port_phy_qcfg(bp, link_info);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Get link config failed with rc %d\n", rc);
+		goto exit;
+	}
+	if (link_info->link_up)
+		link->link_speed =
+			bnxt_parse_hw_link_speed(link_info->link_speed);
+	else
+		link->link_speed = ETH_LINK_SPEED_10M;
+	link->link_duplex = bnxt_parse_hw_link_duplex(link_info->duplex);
+	link->link_status = link_info->link_up;
+	link->link_autoneg = link_info->auto_mode ==
+		HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE ?
+		ETH_LINK_SPEED_FIXED : ETH_LINK_SPEED_AUTONEG;
+exit:
+	return rc;
+}
+
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index c28e8d8..a508024 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -98,6 +98,7 @@ int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 void bnxt_free_all_hwrm_resources(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+int bnxt_get_hwrm_link_config(struct bnxt *bp, struct rte_eth_link *link);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
 
 #endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index d58295a..f2db3ea 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -88,6 +88,7 @@ struct ctx_hw_stats64 {
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
+#define HWRM_PORT_PHY_QCFG		(UINT32_C(0x27))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
@@ -2793,6 +2794,795 @@ struct hwrm_port_phy_cfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_port_phy_qcfg */
+/* Description: This command queries the PHY configuration for the port. */
+/* Input (24 bytes) */
+
+struct hwrm_port_phy_qcfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Port ID of port that is to be queried. */
+	uint16_t port_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (96 bytes) */
+struct hwrm_port_phy_qcfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This value indicates the current link status. */
+		/* There is no link or cable detected. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_NO_LINK	(UINT32_C(0x0) << 0)
+		/* There is no link, but a cable has been detected. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SIGNAL	(UINT32_C(0x1) << 0)
+		/* There is a link. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_LINK	(UINT32_C(0x2) << 0)
+	uint8_t link;
+
+	uint8_t unused_0;
+
+	/* This value indicates the current link speed of the connection. */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t link_speed;
+
+	/* This value is indicates the duplex of the current connection. */
+		/* Half Duplex connection. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_DUPLEX_HALF	(UINT32_C(0x0) << 0)
+		/* Full duplex connection. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_DUPLEX_FULL	(UINT32_C(0x1) << 0)
+	uint8_t duplex;
+
+	/*
+	 * This value is used to indicate the current pause configuration. When
+	 * autoneg is enabled, this value represents the autoneg results of
+	 * pause configuration.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX	UINT32_C(0x2)
+	uint8_t pause;
+
+	/*
+	 * The supported speeds for the port. This is a bit mask. For each speed
+	 * that is supported, the corrresponding bit will be set to '1'.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10MB \
+							UINT32_C(0x2000)
+	uint16_t support_speeds;
+
+	/*
+	 * Current setting of forced link speed. When the link speed is not
+	 * being forced, this value shall be set to 0.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t force_link_speed;
+
+	/* Current setting of auto negotiation mode. */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ALL_SPEEDS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_SPEED_MASK \
+							(UINT32_C(0x4) << 0)
+	uint8_t auto_mode;
+
+	/*
+	 * Current setting of pause autonegotiation. Move autoneg_pause flag
+	 * here.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_TX	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_RX	UINT32_C(0x2)
+	/*
+	 * When set to 1, the advertisement of pause is enabled. # When the
+	 * auto_mode is not set to none and this flag is set to 1, then the
+	 * auto_pause bits on this port are being advertised and autoneg pause
+	 * results are being interpreted. # When the auto_mode is not set to
+	 * none and this flag is set to 0, the pause is forced as indicated in
+	 * force_pause, and also advertised as auto_pause bits, but the autoneg
+	 * results are not interpreted since the pause configuration is being
+	 * forced. # When the auto_mode is set to none and this flag is set to
+	 * 1, auto_pause bits should be ignored and should be set to 0.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_AUTONEG_PAUSE \
+							UINT32_C(0x4)
+	uint8_t auto_pause;
+
+	/*
+	 * Current setting for auto_link_speed. This field is only valid when
+	 * auto_mode is set to "one_speed" or "one_or_below".
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t auto_link_speed;
+
+	/*
+	 * Current setting for auto_link_speed_mask that is used to advertise
+	 * speeds during autonegotiation. This field is only valid when
+	 * auto_mode is set to "mask". The speeds specified in this field shall
+	 * be a subset of supported speeds on this port.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10MB \
+							UINT32_C(0x2000)
+	uint16_t auto_link_speed_mask;
+
+	/* Current setting for wirespeed. */
+		/* Wirespeed feature is disabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_WIRESPEED_OFF	(UINT32_C(0x0) << 0)
+		/* Wirespeed feature is enabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_WIRESPEED_ON	(UINT32_C(0x1) << 0)
+	uint8_t wirespeed;
+
+	/* Current setting for loopback. */
+		/* No loopback is selected. Normal operation. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_NONE	(UINT32_C(0x0) << 0)
+		/*
+		 * The HW will be configured with local loopback such that host
+		 * data is sent back to the host without modification.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_LOCAL	(UINT32_C(0x1) << 0)
+		/*
+		 * The HW will be configured with remote loopback such that port
+		 * logic will send packets back out the transmitter that are
+		 * received.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_REMOTE	(UINT32_C(0x2) << 0)
+	uint8_t lpbk;
+
+	/*
+	 * Current setting of forced pause. When the pause configuration is not
+	 * being forced, then this value shall be set to 0.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_PAUSE_TX \
+							UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_PAUSE_RX \
+							UINT32_C(0x2)
+	uint8_t force_pause;
+
+	/*
+	 * This value indicates the current status of the optics module on this
+	 * port.
+	 */
+		/* Module is inserted and accepted */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Module is rejected and transmit side Laser is disabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_DISABLETX \
+							(UINT32_C(0x1) << 0)
+		/* Module mismatch warning. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG \
+							(UINT32_C(0x2) << 0)
+		/* Module is rejected and powered down. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_PWRDOWN \
+							(UINT32_C(0x3) << 0)
+		/* Module is not inserted. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NOTINSERTED \
+							(UINT32_C(0x4) << 0)
+		/* Module status is not applicable. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NOTAPPLICABLE \
+							(UINT32_C(0xff) << 0)
+	uint8_t module_status;
+
+	/* Current setting for preemphasis. */
+	uint32_t preemphasis;
+
+	/* This field represents the major version of the PHY. */
+	uint8_t phy_maj;
+
+	/* This field represents the minor version of the PHY. */
+	uint8_t phy_min;
+
+	/* This field represents the build version of the PHY. */
+	uint8_t phy_bld;
+
+	/* This value represents a PHY type. */
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* BASE-CR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASECR \
+							(UINT32_C(0x1) << 0)
+		/* BASE-KR4 (Deprecated) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR4 \
+							(UINT32_C(0x2) << 0)
+		/* BASE-LR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASELR \
+							(UINT32_C(0x3) << 0)
+		/* BASE-SR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASESR \
+							(UINT32_C(0x4) << 0)
+		/* BASE-KR2 (Deprecated) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR2 \
+							(UINT32_C(0x5) << 0)
+		/* BASE-KX */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKX \
+							(UINT32_C(0x6) << 0)
+		/* BASE-KR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR \
+							(UINT32_C(0x7) << 0)
+		/* BASE-T */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASET \
+							(UINT32_C(0x8) << 0)
+		/* EEE capable BASE-T */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASETE \
+							(UINT32_C(0x9) << 0)
+		/* SGMII connected external PHY */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_SGMIIEXTPHY \
+							(UINT32_C(0xa) << 0)
+	uint8_t phy_type;
+
+	/* This value represents a media type. */
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* Twisted Pair */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_TP	(UINT32_C(0x1) << 0)
+		/* Direct Attached Copper */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_DAC \
+							(UINT32_C(0x2) << 0)
+		/* Fiber */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_FIBRE \
+							(UINT32_C(0x3) << 0)
+	uint8_t media_type;
+
+	/* This value represents a transceiver type. */
+		/* PHY and MAC are in the same package */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_PKG_TYPE_XCVR_INTERNAL \
+							(UINT32_C(0x1) << 0)
+		/* PHY and MAC are in different packages */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_PKG_TYPE_XCVR_EXTERNAL \
+							(UINT32_C(0x2) << 0)
+	uint8_t xcvr_pkg_type;
+
+	/*
+	 * This field represents flags related to EEE configuration. These EEE
+	 * configuration flags are valid only when the auto_mode is not set to
+	 * none (in other words autonegotiation is enabled).
+	 */
+	/* This field represents PHY address. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_ADDR_MASK	UINT32_C(0x1f)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_ADDR_SFT	0
+	/*
+	 * When set to 1, Energy Efficient Ethernet (EEE) mode is enabled.
+	 * Speeds for autoneg with EEE mode enabled are based on
+	 * eee_link_speed_mask.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_ENABLED \
+							UINT32_C(0x20)
+	/*
+	 * This flag is valid only when eee_enabled is set to 1. # If
+	 * eee_enabled is set to 0, then EEE mode is disabled and this flag
+	 * shall be ignored. # If eee_enabled is set to 1 and this flag is set
+	 * to 1, then Energy Efficient Ethernet (EEE) mode is enabled and in
+	 * use. # If eee_enabled is set to 1 and this flag is set to 0, then
+	 * Energy Efficient Ethernet (EEE) mode is enabled but is currently not
+	 * in use.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_ACTIVE \
+							UINT32_C(0x40)
+	/*
+	 * This flag is valid only when eee_enabled is set to 1. # If
+	 * eee_enabled is set to 0, then EEE mode is disabled and this flag
+	 * shall be ignored. # If eee_enabled is set to 1 and this flag is set
+	 * to 1, then Energy Efficient Ethernet (EEE) mode is enabled and TX LPI
+	 * is enabled. # If eee_enabled is set to 1 and this flag is set to 0,
+	 * then Energy Efficient Ethernet (EEE) mode is enabled but TX LPI is
+	 * disabled.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_TX_LPI \
+							UINT32_C(0x80)
+	/*
+	 * This field represents flags related to EEE configuration. These EEE
+	 * configuration flags are valid only when the auto_mode is not set to
+	 * none (in other words autonegotiation is enabled).
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_MASK \
+							UINT32_C(0xe0)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_SFT	5
+	uint8_t eee_config_phy_addr;
+
+	/* Reserved field, set to 0 */
+	/*
+	 * When set to 1, the parallel detection is used to determine the speed
+	 * of the link partner. Parallel detection is used when a
+	 * autonegotiation capable device is connected to a link parter that is
+	 * not capable of autonegotiation.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PARALLEL_DETECT \
+							UINT32_C(0x1)
+	/* Reserved field, set to 0 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_RESERVED_MASK	UINT32_C(0xfe)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_RESERVED_SFT	1
+	uint8_t parallel_detect;
+
+	/*
+	 * The advertised speeds for the port by the link partner. Each
+	 * advertised speed will be set to '1'.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10MB \
+							UINT32_C(0x2000)
+	uint16_t link_partner_adv_speeds;
+
+	/*
+	 * The advertised autoneg for the port by the link partner. This field
+	 * is deprecated and should be set to 0.
+	 */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ALL_SPEEDS\
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_SPEED_MASK\
+							(UINT32_C(0x4) << 0)
+	uint8_t link_partner_adv_auto_mode;
+
+	/* The advertised pause settings on the port by the link partner. */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_PAUSE_TX \
+							UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_PAUSE_RX \
+							UINT32_C(0x2)
+	uint8_t link_partner_adv_pause;
+
+	/*
+	 * Current setting for link speed mask that is used to advertise speeds
+	 * during autonegotiation when EEE is enabled. This field is valid only
+	 * when eee_enabled flags is set to 1. The speeds specified in this
+	 * field shall be a subset of speeds specified in auto_link_speed_mask.
+	 */
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD1 \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD2 \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD3 \
+							UINT32_C(0x10)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD4 \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	uint16_t adv_eee_link_speed_mask;
+
+	/*
+	 * Current setting for link speed mask that is advertised by the link
+	 * partner when EEE is enabled. This field is valid only when
+	 * eee_enabled flags is set to 1.
+	 */
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD1 \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD2 \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD3 \
+							UINT32_C(0x10)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD4 \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	uint16_t link_partner_adv_eee_link_speed_mask;
+
+	/* This value represents transceiver identifier type. */
+	/*
+	 * Current setting of TX LPI timer in microseconds. This field is valid
+	 * only when_eee_enabled flag is set to 1 and tx_lpi_enabled is set to
+	 * 1.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_TX_LPI_TIMER_MASK \
+							UINT32_C(0xffffff)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_TX_LPI_TIMER_SFT         0
+	/* This value represents transceiver identifier type. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_MASK \
+							UINT32_C(0xff000000)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_SFT \
+							24
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 24)
+		/* SFP/SFP+/SFP28 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_SFP \
+							(UINT32_C(0x3) << 24)
+		/* QSFP */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFP \
+							(UINT32_C(0xc) << 24)
+		/* QSFP+ */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFPPLUS \
+							(UINT32_C(0xd) << 24)
+		/* QSFP28 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFP28 \
+							(UINT32_C(0x11) << 24)
+	uint32_t xcvr_identifier_type_tx_lpi_timer;
+
+	uint32_t unused_1;
+
+	/*
+	 * Up to 16 bytes of null padded ASCII string representing PHY vendor.
+	 * If the string is set to null, then the vendor name is not available.
+	 */
+	char phy_vendor_name[16];
+
+	/*
+	 * Up to 16 bytes of null padded ASCII string that identifies vendor
+	 * specific part number of the PHY. If the string is set to null, then
+	 * the vendor specific part number is not available.
+	 */
+	char phy_vendor_partnumber[16];
+
+	uint32_t unused_2;
+	uint8_t unused_3;
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_ver_get */
 /*
  * Description: This function is called by a driver to determine the HWRM
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 30/38] bnxt: add start/stop/link update operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (27 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 29/38] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 31/38] bnxt: add promiscuous enable/disable operations Stephen Hurd
                     ` (10 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to add the start, stop and link update dev_ops.

Also adds wraper functions like bnxt_init_chip(), bnxt_init_nic(),
bnxt_alloc_mem(), bnxt_free_mem()

The BNXT driver will now minimally pass traffic with testpmd.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.
Shorten the string passed for reserving memzone when default completion
ring is created.

v5:
Reorder footer
---
 drivers/net/bnxt/bnxt_ethdev.c | 269 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 269 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 6888363..ac82876 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -40,12 +40,17 @@
 #include <rte_cycles.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
 #include "bnxt_rxq.h"
 #include "bnxt_rxr.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
 #include "bnxt_txr.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -65,6 +70,177 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 	bnxt_free_hwrm_resources(bp);
 }
 
+/***********************/
+
+/*
+ * High level utility functions
+ */
+
+static void bnxt_free_mem(struct bnxt *bp)
+{
+	bnxt_free_filter_mem(bp);
+	bnxt_free_vnic_attributes(bp);
+	bnxt_free_vnic_mem(bp);
+
+	bnxt_free_stats(bp);
+	bnxt_free_tx_rings(bp);
+	bnxt_free_rx_rings(bp);
+	bnxt_free_def_cp_ring(bp);
+}
+
+static int bnxt_alloc_mem(struct bnxt *bp)
+{
+	int rc;
+
+	/* Default completion ring */
+	rc = bnxt_init_def_ring_struct(bp, SOCKET_ID_ANY);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_rings(bp, 0, NULL, NULL,
+			      bp->def_cp_ring, "def_cp");
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_vnic_mem(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_vnic_attributes(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_filter_mem(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	return 0;
+
+alloc_mem_err:
+	bnxt_free_mem(bp);
+	return rc;
+}
+
+static int bnxt_init_chip(struct bnxt *bp)
+{
+	unsigned int i, rss_idx, fw_idx;
+	int rc;
+
+	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM stat ctx alloc failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_alloc_hwrm_rings(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM ring alloc failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM ring grp alloc failure: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_mq_rx_configure(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "MQ mode configure failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	/* VNIC configuration */
+	for (i = 0; i < bp->nr_vnics; i++) {
+		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+		rc = bnxt_hwrm_vnic_alloc(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic alloc failure rc: %x\n",
+				rc);
+			goto err_out;
+		}
+
+		rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD,
+				"HWRM vnic ctx alloc failure rc: %x\n", rc);
+			goto err_out;
+		}
+
+		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic cfg failure rc: %x\n", rc);
+			goto err_out;
+		}
+
+		rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic filter failure rc: %x\n",
+				rc);
+			goto err_out;
+		}
+		if (vnic->rss_table && vnic->hash_type) {
+			/*
+			 * Fill the RSS hash & redirection table with
+			 * ring group ids for all VNICs
+			 */
+			for (rss_idx = 0, fw_idx = 0;
+			     rss_idx < HW_HASH_INDEX_SIZE;
+			     rss_idx++, fw_idx++) {
+				if (vnic->fw_grp_ids[fw_idx] ==
+				    INVALID_HW_RING_ID)
+					fw_idx = 0;
+				vnic->rss_table[rss_idx] =
+						vnic->fw_grp_ids[fw_idx];
+			}
+			rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+			if (rc) {
+				RTE_LOG(ERR, PMD,
+					"HWRM vnic set RSS failure rc: %x\n",
+					rc);
+				goto err_out;
+			}
+		}
+	}
+	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0]);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	return 0;
+
+err_out:
+	bnxt_free_all_hwrm_resources(bp);
+
+	return rc;
+}
+
+static int bnxt_shutdown_nic(struct bnxt *bp)
+{
+	bnxt_free_all_hwrm_resources(bp);
+	bnxt_free_all_filters(bp);
+	bnxt_free_all_vnics(bp);
+	return 0;
+}
+
+static int bnxt_init_nic(struct bnxt *bp)
+{
+	int rc;
+
+	bnxt_init_ring_grps(bp);
+	bnxt_init_vnics(bp);
+	bnxt_init_filters(bp);
+
+	rc = bnxt_init_chip(bp);
+	if (rc)
+		return rc;
+
+	return 0;
+}
+
 /*
  * Device configuration and status function
  */
@@ -182,6 +358,85 @@ static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
 	return rc;
 }
 
+static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	int rc;
+
+	rc = bnxt_hwrm_func_reset(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "hwrm chip reset failure rc: %x\n", rc);
+		rc = -1;
+		goto error;
+	}
+
+	rc = bnxt_alloc_mem(bp);
+	if (rc)
+		goto error;
+
+	rc = bnxt_init_nic(bp);
+	if (rc)
+		goto error;
+
+	return 0;
+
+error:
+	bnxt_shutdown_nic(bp);
+	bnxt_free_tx_mbufs(bp);
+	bnxt_free_rx_mbufs(bp);
+	bnxt_free_mem(bp);
+	return rc;
+}
+
+/* Unload the driver, release resources */
+static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	if (bp->eth_dev->data->dev_started) {
+		/* TBD: STOP HW queues DMA */
+		eth_dev->data->dev_link.link_status = 0;
+	}
+	bnxt_shutdown_nic(bp);
+}
+
+static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
+			       int wait_to_complete)
+{
+	int rc = 0;
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_link new;
+	unsigned int cnt = BNXT_LINK_WAIT_CNT;
+
+	memset(&new, 0, sizeof(new));
+	do {
+		/* Retrieve link info from hardware */
+		rc = bnxt_get_hwrm_link_config(bp, &new);
+		if (rc) {
+			new.link_speed = ETH_LINK_SPEED_100M;
+			new.link_duplex = ETH_LINK_FULL_DUPLEX;
+			RTE_LOG(ERR, PMD,
+				"Failed to retrieve link rc = 0x%x!", rc);
+			goto out;
+		}
+		if (!wait_to_complete)
+			break;
+
+		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
+
+	} while (!new.link_status && cnt--);
+
+	/* Timed out or success */
+	if (new.link_status) {
+		/* Update only if success */
+		eth_dev->data->dev_link.link_duplex = new.link_duplex;
+		eth_dev->data->dev_link.link_speed = new.link_speed;
+	}
+	eth_dev->data->dev_link.link_status = new.link_status;
+out:
+	return rc;
+}
+
 /*
  * Initialization
  */
@@ -190,12 +445,15 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.dev_start = bnxt_dev_start_op,
+	.dev_stop = bnxt_dev_stop_op,
 	.stats_get = bnxt_stats_get_op,
 	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
+	.link_update = bnxt_link_update_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
@@ -305,6 +563,15 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	else
 		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
+	bp->grp_info = rte_zmalloc("bnxt_grp_info",
+				sizeof(*bp->grp_info) * bp->max_ring_grps, 0);
+	if (!bp->grp_info) {
+		RTE_LOG(ERR, PMD,
+			"Failed to alloc %zu bytes needed to store group info table\n",
+			sizeof(*bp->grp_info) * bp->max_ring_grps);
+		rc = -ENOMEM;
+		goto error_free;
+	}
 
 	rc = bnxt_hwrm_func_driver_register(bp, 0,
 					    bp->pf.vf_req_fwd);
@@ -335,6 +602,8 @@ bnxt_dev_uninit(struct rte_eth_dev *eth_dev) {
 
 	if (eth_dev->data->mac_addrs)
 		rte_free(eth_dev->data->mac_addrs);
+	if (bp->grp_info)
+		rte_free(bp->grp_info);
 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
 	bnxt_free_hwrm_resources(bp);
 	return rc;
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 31/38] bnxt: add promiscuous enable/disable operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (28 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 30/38] bnxt: add start/stop/link update operations Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 32/38] bnxt: add all multicast " Stephen Hurd
                     ` (9 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the promiscuous mode enable and disable dev_ops.
Uses the bnxt_hwrm_cfa_l2_set_rx_mask() API added in the earlier patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix couple of typos in the commit message.

v5:
Reorder footer
Rewrite the patch description.
---
 drivers/net/bnxt/bnxt_ethdev.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index ac82876..3fce540 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -437,6 +437,34 @@ out:
 	return rc;
 }
 
+static void bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
+static void bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
 /*
  * Initialization
  */
@@ -454,6 +482,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
 	.link_update = bnxt_link_update_op,
+	.promiscuous_enable = bnxt_promiscuous_enable_op,
+	.promiscuous_disable = bnxt_promiscuous_disable_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 32/38] bnxt: add all multicast enable/disable operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (29 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 31/38] bnxt: add promiscuous enable/disable operations Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 33/38] bnxt: free memory in close operation Stephen Hurd
                     ` (8 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds dev_ops to enable/disable multicast traffic.
Uses the bnxt_hwrm_cfa_l2_set_rx_mask() API added in the earlier patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder footer
Rewrite the patch description.
---
 drivers/net/bnxt/bnxt_ethdev.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3fce540..d3a624f 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -465,6 +465,34 @@ static void bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
 	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
 }
 
+static void bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
+static void bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
 /*
  * Initialization
  */
@@ -484,6 +512,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
+	.allmulticast_enable = bnxt_allmulticast_enable_op,
+	.allmulticast_disable = bnxt_allmulticast_disable_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 33/38] bnxt: free memory in close operation
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (30 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 32/38] bnxt: add all multicast " Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 34/38] bnxt: add MAC address add/remove dev ops Stephen Hurd
                     ` (7 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to free all resources except the one corresponding
to HWRM, which are required to notify the HWRM that the driver is unloaded
(these are freed in uninit()).

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index d3a624f..4254531 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -62,14 +62,6 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
-static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
-{
-	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
-
-	rte_free(eth_dev->data->mac_addrs);
-	bnxt_free_hwrm_resources(bp);
-}
-
 /***********************/
 
 /*
@@ -388,6 +380,16 @@ error:
 	return rc;
 }
 
+static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	bnxt_free_tx_mbufs(bp);
+	bnxt_free_rx_mbufs(bp);
+	bnxt_free_mem(bp);
+	rte_free(eth_dev->data->mac_addrs);
+}
+
 /* Unload the driver, release resources */
 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 {
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 34/38] bnxt: add MAC address add/remove dev ops
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (31 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 33/38] bnxt: free memory in close operation Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 35/38] bnxt: add set link up/down operations Stephen Hurd
                     ` (6 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds dev_ops to Add/Remove MAC addresses.
These use the bnxt_hwrm_set_filter() defined in the previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 71 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 71 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 4254531..34a5873 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -402,6 +402,75 @@ static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 	bnxt_shutdown_nic(bp);
 }
 
+static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
+				    uint32_t index)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter, *temp_filter;
+	int i;
+
+	/*
+	 * Loop through all VNICs from the specified filter flow pools to
+	 * remove the corresponding MAC addr filter
+	 */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		if (!(pool_mask & (1 << i)))
+			continue;
+
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			filter = STAILQ_FIRST(&vnic->filter);
+			while (filter) {
+				temp_filter = STAILQ_NEXT(filter, next);
+				if (filter->mac_index == index) {
+					STAILQ_REMOVE(&vnic->filter, filter,
+						      bnxt_filter_info, next);
+					bnxt_hwrm_clear_filter(bp, filter);
+					filter->mac_index = INVALID_MAC_INDEX;
+					memset(&filter->l2_addr, 0,
+					       ETHER_ADDR_LEN);
+					STAILQ_INSERT_TAIL(
+							&bp->free_filter_list,
+							filter, next);
+				}
+				filter = temp_filter;
+			}
+		}
+	}
+}
+
+static void bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
+				 struct ether_addr *mac_addr,
+				 uint32_t index, uint32_t pool)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = STAILQ_FIRST(&bp->ff_pool[pool]);
+	struct bnxt_filter_info *filter;
+
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "VNIC not found for pool %d!\n", pool);
+		return;
+	}
+	/* Attach requested MAC address to the new l2_filter */
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		if (filter->mac_index == index) {
+			RTE_LOG(ERR, PMD,
+				"MAC addr already existed for pool %d\n", pool);
+			return;
+		}
+	}
+	filter = bnxt_alloc_filter(bp);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+		return;
+	}
+	STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+	filter->mac_index = index;
+	memcpy(filter->l2_addr, mac_addr, ETHER_ADDR_LEN);
+	bnxt_hwrm_set_filter(bp, vnic, filter);
+}
+
 static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
 			       int wait_to_complete)
 {
@@ -516,6 +585,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
 	.allmulticast_enable = bnxt_allmulticast_enable_op,
 	.allmulticast_disable = bnxt_allmulticast_disable_op,
+	.mac_addr_add = bnxt_mac_addr_add_op,
+	.mac_addr_remove = bnxt_mac_addr_remove_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 35/38] bnxt: add set link up/down operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (32 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 34/38] bnxt: add MAC address add/remove dev ops Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 36/38] bnxt: add reta update/query operations Stephen Hurd
                     ` (5 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Adds dev_ops to set link UP or DOWN as appropriate.
Uses the bnxt_set_hwrm_link_config() API added in previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 34a5873..b04010c 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -380,6 +380,24 @@ error:
 	return rc;
 }
 
+static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	eth_dev->data->dev_link.link_status = 1;
+	bnxt_set_hwrm_link_config(bp, true);
+	return 0;
+}
+
+static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	eth_dev->data->dev_link.link_status = 0;
+	bnxt_set_hwrm_link_config(bp, false);
+	return 0;
+}
+
 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
@@ -574,6 +592,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_configure = bnxt_dev_configure_op,
 	.dev_start = bnxt_dev_start_op,
 	.dev_stop = bnxt_dev_stop_op,
+	.dev_set_link_up = bnxt_dev_set_link_up_op,
+	.dev_set_link_down = bnxt_dev_set_link_down_op,
 	.stats_get = bnxt_stats_get_op,
 	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 36/38] bnxt: add reta update/query operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (33 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 35/38] bnxt: add set link up/down operations Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 37/38] bnxt: add RSS device operations Stephen Hurd
                     ` (4 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add code to Update/query reta dev_ops
Uses the bnxt_hwrm_vnic_rss_cfg() HWRM API added earlier.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 56 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 56 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b04010c..b3b76f1 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -582,6 +582,60 @@ static void bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
 	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
 }
 
+static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
+			    struct rte_eth_rss_reta_entry64 *reta_conf,
+			    uint16_t reta_size)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_vnic_info *vnic;
+	int i;
+
+	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
+		return -EINVAL;
+
+	if (reta_size != HW_HASH_INDEX_SIZE) {
+		RTE_LOG(ERR, PMD, "The configured hash table lookup size "
+			"(%d) must equal the size supported by the hardware "
+			"(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+		return -EINVAL;
+	}
+	/* Update the RSS VNIC(s) */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			memcpy(vnic->rss_table, reta_conf, reta_size);
+
+			bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+		}
+	}
+	return 0;
+}
+
+static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
+			      struct rte_eth_rss_reta_entry64 *reta_conf,
+			      uint16_t reta_size)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+
+	/* Retrieve from the default VNIC */
+	if (!vnic)
+		return -EINVAL;
+	if (!vnic->rss_table)
+		return -EINVAL;
+
+	if (reta_size != HW_HASH_INDEX_SIZE) {
+		RTE_LOG(ERR, PMD, "The configured hash table lookup size "
+			"(%d) must equal the size supported by the hardware "
+			"(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+		return -EINVAL;
+	}
+	/* EW - need to revisit here copying from u64 to u16 */
+	memcpy(reta_conf, vnic->rss_table, reta_size);
+
+	return 0;
+}
+
 /*
  * Initialization
  */
@@ -600,6 +654,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
+	.reta_update = bnxt_reta_update_op,
+	.reta_query = bnxt_reta_query_op,
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 37/38] bnxt: add RSS device operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (34 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 36/38] bnxt: add reta update/query operations Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 38/38] bnxt: add flow control operations Stephen Hurd
                     ` (3 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add rss_hash_update and rss_hash_conf_get dev_ops
Uses the bnxt_hwrm_vnic_rss_cfg() HWRM API added in the previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 121 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 121 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b3b76f1..3c7f868 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -62,6 +62,14 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
+#define BNXT_ETH_RSS_SUPPORT (	\
+	ETH_RSS_IPV4 |		\
+	ETH_RSS_NONFRAG_IPV4_TCP |	\
+	ETH_RSS_NONFRAG_IPV4_UDP |	\
+	ETH_RSS_IPV6 |		\
+	ETH_RSS_NONFRAG_IPV6_TCP |	\
+	ETH_RSS_NONFRAG_IPV6_UDP)
+
 /***********************/
 
 /*
@@ -636,6 +644,117 @@ static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
 	return 0;
 }
 
+static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
+				   struct rte_eth_rss_conf *rss_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_vnic_info *vnic;
+	uint16_t hash_type = 0;
+	int i;
+
+	/*
+	 * If RSS enablement were different than dev_configure,
+	 * then return -EINVAL
+	 */
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
+		if (!rss_conf->rss_hf)
+			return -EINVAL;
+	} else {
+		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
+			return -EINVAL;
+	}
+	if (rss_conf->rss_hf & ETH_RSS_IPV4)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV4_TCP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV4_UDP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_IPV6)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV6_TCP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV6_UDP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
+
+	/* Update the RSS VNIC(s) */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			vnic->hash_type = hash_type;
+
+			/*
+			 * Use the supplied key if the key length is
+			 * acceptable and the rss_key is not NULL
+			 */
+			if (rss_conf->rss_key &&
+			    rss_conf->rss_key_len <= HW_HASH_KEY_SIZE)
+				memcpy(vnic->rss_hash_key, rss_conf->rss_key,
+				       rss_conf->rss_key_len);
+
+			bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+		}
+	}
+	return 0;
+}
+
+static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
+				     struct rte_eth_rss_conf *rss_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+	int len;
+	uint32_t hash_types;
+
+	/* RSS configuration is the same for all VNICs */
+	if (vnic && vnic->rss_hash_key) {
+		if (rss_conf->rss_key) {
+			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
+			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
+			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
+		}
+
+		hash_types = vnic->hash_type;
+		rss_conf->rss_hf = 0;
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_IPV4;
+			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_IPV6;
+			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
+		}
+		if (hash_types) {
+			RTE_LOG(ERR, PMD,
+				"Unknwon RSS config from firmware (%08x), RSS disabled",
+				vnic->hash_type);
+			return -ENOTSUP;
+		}
+	} else {
+		rss_conf->rss_hf = 0;
+	}
+	return 0;
+}
+
 /*
  * Initialization
  */
@@ -656,6 +775,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.tx_queue_release = bnxt_tx_queue_release_op,
 	.reta_update = bnxt_reta_update_op,
 	.reta_query = bnxt_reta_query_op,
+	.rss_hash_update = bnxt_rss_hash_update_op,
+	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v5 38/38] bnxt: add flow control operations
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (35 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 37/38] bnxt: add RSS device operations Stephen Hurd
@ 2016-06-14 22:56   ` Stephen Hurd
  2016-06-15 15:28   ` [dpdk-dev] [PATCH v5 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Bruce Richardson
                     ` (2 subsequent siblings)
  39 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-14 22:56 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add flow_ctrl_get and flow_ctrl_set dev_ops.
Uses the bnxt_set_hwrm_link_config() HWRM API added in earlier patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 83 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 83 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3c7f868..406e38a 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -755,6 +755,87 @@ static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
 	return 0;
 }
 
+static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
+			       struct rte_eth_fc_conf *fc_conf __rte_unused)
+{
+	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+	struct rte_eth_link link_info;
+	int rc;
+
+	rc = bnxt_get_hwrm_link_config(bp, &link_info);
+	if (rc)
+		return rc;
+
+	memset(fc_conf, 0, sizeof(*fc_conf));
+	if (bp->link_info.auto_pause)
+		fc_conf->autoneg = 1;
+	switch (bp->link_info.pause) {
+	case 0:
+		fc_conf->mode = RTE_FC_NONE;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
+		fc_conf->mode = RTE_FC_TX_PAUSE;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
+		fc_conf->mode = RTE_FC_RX_PAUSE;
+		break;
+	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
+			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
+		fc_conf->mode = RTE_FC_FULL;
+		break;
+	}
+	return 0;
+}
+
+static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
+			       struct rte_eth_fc_conf *fc_conf)
+{
+	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+
+	switch (fc_conf->mode) {
+	case RTE_FC_NONE:
+		bp->link_info.auto_pause = 0;
+		bp->link_info.force_pause = 0;
+		break;
+	case RTE_FC_RX_PAUSE:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
+		}
+		break;
+	case RTE_FC_TX_PAUSE:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
+		}
+		break;
+	case RTE_FC_FULL:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
+		}
+		break;
+	}
+	return bnxt_set_hwrm_link_config(bp, true);
+}
+
 /*
  * Initialization
  */
@@ -784,6 +865,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.allmulticast_disable = bnxt_allmulticast_disable_op,
 	.mac_addr_add = bnxt_mac_addr_add_op,
 	.mac_addr_remove = bnxt_mac_addr_remove_op,
+	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
+	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v5 01/38] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (36 preceding siblings ...)
  2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 38/38] bnxt: add flow control operations Stephen Hurd
@ 2016-06-15 15:28   ` Bruce Richardson
  2016-06-15 15:30   ` Bruce Richardson
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
  39 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-15 15:28 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

One reminder for when sending future patchset - or other revisions of this
set, please include a cover letter. Any patchset over 4/5 patches generally
needs one, and it allows us to see the combined diffstats for the whole set.

/Bruce

On Tue, Jun 14, 2016 at 03:55:34PM -0700, Stephen Hurd wrote:
> From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> 
> This patch adds the initial skeleton for bnxt driver along with the
> nic guide to tie into the build system.
> At this point, the driver simply fails init.
> 
> Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> Reviewed-by: David Christensen <david.christensen@broadcom.com>
> 
> --
> v4:
> Fix a warning that the document isn't included in any toctree
> Also remove a PCI ID added erroneously.
> 
> v5:
> Re-format footer

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v5 01/38] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (37 preceding siblings ...)
  2016-06-15 15:28   ` [dpdk-dev] [PATCH v5 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Bruce Richardson
@ 2016-06-15 15:30   ` Bruce Richardson
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
  39 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-15 15:30 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Tue, Jun 14, 2016 at 03:55:34PM -0700, Stephen Hurd wrote:
> From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> 
> This patch adds the initial skeleton for bnxt driver along with the
> nic guide to tie into the build system.
> At this point, the driver simply fails init.
> 
> Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> Reviewed-by: David Christensen <david.christensen@broadcom.com>
> 
> --
> v4:
> Fix a warning that the document isn't included in any toctree
> Also remove a PCI ID added erroneously.
> 
> v5:
> Re-format footer
> ---
>  MAINTAINERS                                     |   5 ++
>  config/common_base                              |   5 ++
>  doc/guides/nics/bnxt.rst                        |  49 +++++++++++
>  doc/guides/nics/index.rst                       |   1 +

Sorry I missed this in earlier reviews, but an update to the overview.rst doc
is missing here. It needs a new column added for the new driver.

Regards,
/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code
  2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code Stephen Hurd
@ 2016-06-15 16:25     ` Ferruh Yigit
  2016-06-15 20:28       ` Stephen Hurd
  0 siblings, 1 reply; 132+ messages in thread
From: Ferruh Yigit @ 2016-06-15 16:25 UTC (permalink / raw)
  To: Stephen Hurd, dev, ajit.khaparde, bruce.richardson

On 6/14/2016 11:55 PM, Stephen Hurd wrote:
> From: Ajit Khaparde <ajit.khaparde@broadcom.com>
> 
> Start adding support to use the HWRM API.
> Hardware Resource Manager or HWRM in short, is a set of API provided
> by the firmware running in the ASIC to manage the various resources.
> 
> Initial commit just performs necessary HWRM queries for init, then
> fails as before.
> 
> Now that struct bnxt is non-zero size, we can set dev_private_size
> correctly.
> 
> The used HWRM calls so far:
> bnxt_hwrm_func_qcaps:
> 	This command returns capabilities of a function.
> 
> bnxt_hwrm_ver_get:
> 	This function is called by a driver to determine the HWRM
> 	interface version supported by the HWRM firmware, the
> 	version of HWRM firmware implementation, the name of HWRM
> 	firmware, the versions of other embedded firmwares, and
> 	the names of other embedded firmwares, etc.  Gets the
> 	firmware version and interface specifications.  Returns
> 	an error if the firmware on the device is not supported
> 	by the driver and ensures the response space is large
> 	enough for the largest possible response.
> 
> bnxt_hwrm_queue_qportcfg:
> 	This function is called by a driver to query queue
> 	configuration of a port.
> 
> Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
> Reviewed-by: David Christensen <david.christensen@broadcom.com>
> 
> --
> v4:
> Fix few issues highlighted by checkpatch.
> 
> v5:
> Fix footer formatting
> Switch to using the HWRM function descriptions from the source
> documentation
> ---

Hi Stephen,

I run basic checks on the patch set, mainly it compiles and applies
fine, overall patch set looks good.

Only get following checkpatch warning:
Only v5,02/38 has one WARNING:MACRO_WITH_FLOW_CONTROL

And my script caught following typos on commit messages:
v5-07-38: mamagement
v5-12-38: staitstics, asociated
v5-14-38: fo
v5-25-38: cleansup
v5-28-38: arugments
v5-30-38: wraper

If there won't be a new version of patch set, perhaps we can ask Bruce
to fix typos when applying J

Thanks,
ferruh

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code
  2016-06-15 16:25     ` Ferruh Yigit
@ 2016-06-15 20:28       ` Stephen Hurd
  2016-06-16  9:46         ` Bruce Richardson
  0 siblings, 1 reply; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 20:28 UTC (permalink / raw)
  To: Ferruh Yigit; +Cc: dev, Ajit Kumar Khaparde, Bruce Richardson

On Wed, Jun 15, 2016 at 9:25 AM, Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

>
> Hi Stephen,
>
> I run basic checks on the patch set, mainly it compiles and applies
> fine, overall patch set looks good.
>
> Only get following checkpatch warning:
> Only v5,02/38 has one WARNING:MACRO_WITH_FLOW_CONTROL
>

This is intentional, is this style of copy/paste avoidance not allowed?


-- 
Stephen Hurd
Principal Engineer - Software Development
Broadcom Corporation
949-926-8039
stephen.hurd@broadcom.com

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library
  2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
                     ` (38 preceding siblings ...)
  2016-06-15 15:30   ` Bruce Richardson
@ 2016-06-15 21:23   ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
                       ` (38 more replies)
  39 siblings, 39 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

The bnxt poll mode library (librte_pmd_bnxt) implements support for
Broadcom NetXtreme C-Series.  These adapters support Standards-
compliant 10/25/50Gbps 30MPPS full-duplex throughput.

Information about this family of adapters can be found in the
NetXtreme Brand section https://goo.gl/4H7q63 of the Broadcom web
site http://www.broadcom.com/

With the current driver, allocated mbufs must be large enough to hold
the entire received frame.  If the mbufs are not large enough, the
packets will be dropped.  This is most limiting when jumbo frames are
used.

Ajit Khaparde (38):
  bnxt: new driver for Broadcom NetXtreme-C devices
  bnxt: add HWRM init code
  bnxt: add driver register/unregister support
  bnxt: add dev infos get operation
  bnxt: add dev configure operation
  bnxt: add vnic functions and structs
  bnxt: declare generic ring structs and free() func
  bnxt: add completion ring support
  bnxt: add L2 filter alloc/init/free
  bnxt: add Tx queue operations (nonfunctional)
  bnxt: add Rx queue create/destroy operations
  bnxt: add statistics operations
  bnxt: add initial Tx code implementation
  bnxt: add initial Rx code implementation
  bnxt: add code to alloc/free Tx Rx and cmpl rings
  bnxt: add HWRM function reset command
  bnxt: add HWRM vnic alloc function
  bnxt: add HWRM vnic free function
  bnxt: add HWRM vnic configure function
  bnxt: add API to allow configuration of vnic
  bnxt: add HWRM API to configure RSS
  bnxt: add API for L2 Rx mask set/clear functions
  bnxt: add HWRM API for stats context allocation
  bnxt: add HWRM ring alloc/free functions
  bnxt: add ring group alloc/free functions
  bnxt: add HWRM stat context free function
  bnxt: add HWRM API to set and clear filters
  bnxt: allocate and free all HWRM rings and groups
  bnxt: add HWRM port PHY config call and helpers
  bnxt: add start/stop/link update operations
  bnxt: add promiscuous enable/disable operations
  bnxt: add all multicast enable/disable operations
  bnxt: free memory in close operation
  bnxt: add MAC address add/remove dev ops
  bnxt: add set link up/down operations
  bnxt: add reta update/query operations
  bnxt: add RSS device operations
  bnxt: add flow control operations

 MAINTAINERS                                     |    5 +
 config/common_base                              |    5 +
 doc/guides/nics/bnxt.rst                        |   49 +
 doc/guides/nics/index.rst                       |    1 +
 doc/guides/nics/overview.rst                    |  138 +-
 drivers/net/Makefile                            |    1 +
 drivers/net/bnxt/Makefile                       |   73 +
 drivers/net/bnxt/bnxt.h                         |  176 +
 drivers/net/bnxt/bnxt_cpr.c                     |  159 +
 drivers/net/bnxt/bnxt_cpr.h                     |   86 +
 drivers/net/bnxt/bnxt_ethdev.c                  | 1049 ++++
 drivers/net/bnxt/bnxt_filter.c                  |  175 +
 drivers/net/bnxt/bnxt_filter.h                  |   74 +
 drivers/net/bnxt/bnxt_hwrm.c                    | 1491 ++++++
 drivers/net/bnxt/bnxt_hwrm.h                    |  104 +
 drivers/net/bnxt/bnxt_ring.c                    |  306 ++
 drivers/net/bnxt/bnxt_ring.h                    |  103 +
 drivers/net/bnxt/bnxt_rxq.c                     |  319 ++
 drivers/net/bnxt/bnxt_rxq.h                     |   74 +
 drivers/net/bnxt/bnxt_rxr.c                     |  367 ++
 drivers/net/bnxt/bnxt_rxr.h                     |   62 +
 drivers/net/bnxt/bnxt_stats.c                   |  142 +
 drivers/net/bnxt/bnxt_stats.h                   |   44 +
 drivers/net/bnxt/bnxt_txq.c                     |  162 +
 drivers/net/bnxt/bnxt_txq.h                     |   75 +
 drivers/net/bnxt/bnxt_txr.c                     |  339 ++
 drivers/net/bnxt/bnxt_txr.h                     |   71 +
 drivers/net/bnxt/bnxt_vnic.c                    |  277 ++
 drivers/net/bnxt/bnxt_vnic.h                    |   80 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h          | 5799 +++++++++++++++++++++++
 drivers/net/bnxt/rte_pmd_bnxt_version.map       |    4 +
 lib/librte_eal/common/include/rte_pci_dev_ids.h |   38 +-
 mk/rte.app.mk                                   |    1 +
 33 files changed, 11775 insertions(+), 74 deletions(-)
 create mode 100644 doc/guides/nics/bnxt.rst
 create mode 100644 drivers/net/bnxt/Makefile
 create mode 100644 drivers/net/bnxt/bnxt.h
 create mode 100644 drivers/net/bnxt/bnxt_cpr.c
 create mode 100644 drivers/net/bnxt/bnxt_cpr.h
 create mode 100644 drivers/net/bnxt/bnxt_ethdev.c
 create mode 100644 drivers/net/bnxt/bnxt_filter.c
 create mode 100644 drivers/net/bnxt/bnxt_filter.h
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.c
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.h
 create mode 100644 drivers/net/bnxt/bnxt_ring.c
 create mode 100644 drivers/net/bnxt/bnxt_ring.h
 create mode 100644 drivers/net/bnxt/bnxt_rxq.c
 create mode 100644 drivers/net/bnxt/bnxt_rxq.h
 create mode 100644 drivers/net/bnxt/bnxt_rxr.c
 create mode 100644 drivers/net/bnxt/bnxt_rxr.h
 create mode 100644 drivers/net/bnxt/bnxt_stats.c
 create mode 100644 drivers/net/bnxt/bnxt_stats.h
 create mode 100644 drivers/net/bnxt/bnxt_txq.c
 create mode 100644 drivers/net/bnxt/bnxt_txq.h
 create mode 100644 drivers/net/bnxt/bnxt_txr.c
 create mode 100644 drivers/net/bnxt/bnxt_txr.h
 create mode 100644 drivers/net/bnxt/bnxt_vnic.c
 create mode 100644 drivers/net/bnxt/bnxt_vnic.h
 create mode 100644 drivers/net/bnxt/hsi_struct_def_dpdk.h
 create mode 100644 drivers/net/bnxt/rte_pmd_bnxt_version.map

-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 01/38] bnxt: new driver for Broadcom NetXtreme-C devices
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 02/38] bnxt: add HWRM init code Stephen Hurd
                       ` (37 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the initial skeleton for bnxt driver along with the
nic guide to tie into the build system.
At this point, the driver simply fails init.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix a warning that the document isn't included in any toctree
Also remove a PCI ID added erroneously.

v5:
Re-format footer

v6:
Add a column for the bnxt driver in doc/guides/nics/overview.rst
---
 MAINTAINERS                                     |   5 +
 config/common_base                              |   5 +
 doc/guides/nics/bnxt.rst                        |  49 +++++++++
 doc/guides/nics/index.rst                       |   1 +
 doc/guides/nics/overview.rst                    | 138 ++++++++++++------------
 drivers/net/Makefile                            |   1 +
 drivers/net/bnxt/Makefile                       |  63 +++++++++++
 drivers/net/bnxt/bnxt_ethdev.c                  | 109 +++++++++++++++++++
 drivers/net/bnxt/rte_pmd_bnxt_version.map       |   4 +
 lib/librte_eal/common/include/rte_pci_dev_ids.h |  38 ++++++-
 mk/rte.app.mk                                   |   1 +
 11 files changed, 340 insertions(+), 74 deletions(-)
 create mode 100644 doc/guides/nics/bnxt.rst
 create mode 100644 drivers/net/bnxt/Makefile
 create mode 100644 drivers/net/bnxt/bnxt_ethdev.c
 create mode 100644 drivers/net/bnxt/rte_pmd_bnxt_version.map

diff --git a/MAINTAINERS b/MAINTAINERS
index 3e6b70c..78d8530 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -401,6 +401,11 @@ M: Declan Doherty <declan.doherty@intel.com>
 F: drivers/crypto/null/
 F: doc/guides/cryptodevs/null.rst
 
+Broadcom BNXT PMD
+M: Stephen Hurd <stephen.hurd@broadcom.com>
+F: drivers/net/bnxt/
+F: doc/guides/nics/bnxt.rst
+
 
 Packet processing
 -----------------
diff --git a/config/common_base b/config/common_base
index b9ba405..71be7cd 100644
--- a/config/common_base
+++ b/config/common_base
@@ -245,6 +245,11 @@ CONFIG_RTE_LIBRTE_NFP_PMD=n
 CONFIG_RTE_LIBRTE_NFP_DEBUG=n
 
 #
+# Compile burst-oriented Broadcom BNXT PMD driver
+#
+CONFIG_RTE_LIBRTE_BNXT_PMD=y
+
+#
 # Compile software PMD backed by SZEDATA2 device
 #
 CONFIG_RTE_LIBRTE_PMD_SZEDATA2=n
diff --git a/doc/guides/nics/bnxt.rst b/doc/guides/nics/bnxt.rst
new file mode 100644
index 0000000..2669e98
--- /dev/null
+++ b/doc/guides/nics/bnxt.rst
@@ -0,0 +1,49 @@
+..  BSD LICENSE
+    Copyright 2016 Broadcom Limited
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions
+    are met:
+
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in
+    the documentation and/or other materials provided with the
+    distribution.
+    * Neither the name of Broadcom Limited nor the names of its
+    contributors may be used to endorse or promote products derived
+    from this software without specific prior written permission.
+
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+bnxt poll mode driver library
+=============================
+
+The bnxt poll mode library (**librte_pmd_bnxt**) implements support for
+**Broadcom NetXtreme® C-Series**.  These adapters support Standards-
+compliant 10/25/50Gbps 30MPPS full-duplex throughput.
+
+Information about this family of adapters can be found in the
+`NetXtreme® Brand section <https://www.broadcom.com/products/ethernet-communication-and-switching?technology%5B%5D=88>`_
+of the `Broadcom web site <http://www.broadcom.com/>`_.
+
+Limitations
+-----------
+
+With the current driver, allocated mbufs must be large enough to hold
+the entire received frame.  If the mbufs are not large enough, the
+packets will be dropped.  This is most limiting when jumbo frames are
+used.
+
+SR-IOV is not supported.
diff --git a/doc/guides/nics/index.rst b/doc/guides/nics/index.rst
index 0b13698..ffe011e 100644
--- a/doc/guides/nics/index.rst
+++ b/doc/guides/nics/index.rst
@@ -36,6 +36,7 @@ Network Interface Controller Drivers
     :numbered:
 
     overview
+    bnxt
     bnx2x
     cxgbe
     e1000em
diff --git a/doc/guides/nics/overview.rst b/doc/guides/nics/overview.rst
index 0bd8fae..01d043a 100644
--- a/doc/guides/nics/overview.rst
+++ b/doc/guides/nics/overview.rst
@@ -74,79 +74,79 @@ Most of these differences are summarized below.
 
 .. table:: Features availability in networking drivers
 
-   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
-   Feature              a b b b c e e e i i i i i i i i i i f f f f m m m n n p q q r s v v v v x
-                        f n n o x 1 n n 4 4 4 4 g g x x x x m m m m l l p f u c e e i z h i i m e
-                        p x x n g 0 a i 0 0 0 0 b b g g g g 1 1 1 1 x x i p l a d d n e o r r x n
-                        a 2 2 d b 0   c e e e e   v b b b b 0 0 0 0 4 5 p   l p e e g d s t t n v
-                        c x x i e 0       . v v   f e e e e k k k k     e         v   a t i i e i
-                        k   v n           . f f       . v v   . v v               f   t   o o t r
-                        e   f g           .   .       . f f   . f f                   a     . 3 t
-                        t                 v   v       v   v   v   v                   2     v
-                                          e   e       e   e   e   e                         e
-                                          c   c       c   c   c   c                         c
-   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
+   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
+   Feature              a b b b b c e e e i i i i i i i i i i f f f f m m m n n p q q r s v v v v x
+                        f n n n o x 1 n n 4 4 4 4 g g x x x x m m m m l l p f u c e e i z h i i m e
+                        p x x x n g 0 a i 0 0 0 0 b b g g g g 1 1 1 1 x x i p l a d d n e o r r x n
+                        a 2 2 t d b 0   c e e e e   v b b b b 0 0 0 0 4 5 p   l p e e g d s t t n v
+                        c x x   i e 0       . v v   f e e e e k k k k     e         v   a t i i e i
+                        k   v   n           . f f       . v v   . v v               f   t   o o t r
+                        e   f   g           .   .       . f f   . f f                   a     . 3 t
+                        t                   v   v       v   v   v   v                   2     v
+                                            e   e       e   e   e   e                         e
+                                            c   c       c   c   c   c                         c
+   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
    Speed capabilities
-   Link status            Y Y   Y Y   Y Y Y     Y   Y Y Y Y         Y Y         Y Y   Y Y Y Y
-   Link status event      Y Y     Y     Y Y     Y   Y Y             Y Y         Y Y     Y
+   Link status            Y Y Y   Y Y   Y Y Y     Y   Y Y Y Y         Y Y         Y Y   Y Y Y Y
+   Link status event      Y Y       Y     Y Y     Y   Y Y             Y Y         Y Y     Y
    Queue status event                                                                   Y
-   Rx interrupt                   Y     Y Y Y Y Y Y Y Y Y Y Y Y Y Y
-   Queue start/stop             Y   Y Y Y Y Y Y     Y Y     Y Y Y Y Y Y               Y   Y Y
-   MTU update                   Y Y Y           Y   Y Y Y Y         Y Y
-   Jumbo frame                  Y Y Y Y Y Y Y Y Y   Y Y Y Y Y Y Y Y Y Y       Y Y Y
-   Scattered Rx                 Y Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y               Y   Y
-   LRO                                              Y Y Y Y
-   TSO                          Y   Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y
-   Promiscuous mode       Y Y   Y Y   Y Y Y Y Y Y Y Y Y     Y Y     Y Y         Y Y   Y   Y Y
-   Allmulticast mode            Y Y     Y Y Y Y Y Y Y Y Y Y Y Y     Y Y         Y Y   Y   Y Y
-   Unicast MAC filter     Y Y     Y   Y Y Y Y Y Y Y Y Y Y Y Y Y     Y Y         Y Y       Y Y
-   Multicast MAC filter   Y Y         Y Y Y Y Y             Y Y     Y Y         Y Y       Y Y
-   RSS hash                     Y   Y Y Y Y Y Y Y   Y Y Y Y Y Y Y Y Y Y         Y Y
-   RSS key update                   Y   Y Y Y Y Y   Y Y Y Y Y Y Y Y   Y
-   RSS reta update                  Y   Y Y Y Y Y   Y Y Y Y Y Y Y Y   Y
-   VMDq                                 Y Y     Y   Y Y     Y Y
-   SR-IOV                   Y       Y   Y Y     Y   Y Y             Y Y           Y
-   DCB                                  Y Y     Y   Y Y
-   VLAN filter                    Y   Y Y Y Y Y Y Y Y Y Y Y Y Y     Y Y         Y Y       Y Y
-   Ethertype filter                     Y Y     Y   Y Y
-   N-tuple filter                               Y   Y Y
-   SYN filter                                   Y   Y Y
-   Tunnel filter                        Y Y         Y Y
-   Flexible filter                              Y
-   Hash filter                          Y Y Y Y
-   Flow director                        Y Y         Y Y               Y
-   Flow control                 Y Y     Y Y     Y   Y Y                         Y Y
-   Rate limitation                                  Y Y
-   Traffic mirroring                    Y Y         Y Y
-   CRC offload                  Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y   Y         Y Y
-   VLAN offload                 Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y   Y         Y Y
-   QinQ offload                   Y     Y   Y   Y Y Y   Y
-   L3 checksum offload          Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y Y Y
-   L4 checksum offload          Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y Y Y
-   Inner L3 checksum                Y   Y   Y       Y   Y           Y
-   Inner L4 checksum                Y   Y   Y       Y   Y           Y
-   Packet type parsing          Y     Y Y   Y   Y Y Y   Y   Y Y Y Y Y Y         Y Y
-   Timesync                             Y Y     Y   Y Y
-   Basic stats            Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y       Y Y Y   Y Y Y Y
-   Extended stats                   Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y                   Y Y
-   Stats per queue              Y                   Y Y     Y Y Y Y Y Y         Y Y   Y   Y Y
-   EEPROM dump                                  Y   Y Y
-   Registers dump                               Y Y Y Y Y Y
-   Multiprocess aware                   Y Y Y Y     Y Y Y Y Y Y Y Y Y Y       Y Y Y
-   BSD nic_uio                  Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y                       Y Y
-   Linux UIO              Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y             Y Y       Y Y
-   Linux VFIO                   Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y                       Y Y
-   Other kdrv                                                       Y Y               Y
-   ARMv7                                                                      Y           Y Y
-   ARMv8                                                                      Y           Y Y
-   Power8                                                           Y Y       Y
-   TILE-Gx                                                                    Y
-   x86-32                       Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y       Y         Y Y Y
-   x86-64                 Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y       Y Y Y   Y Y Y Y
-   Usage doc              Y Y   Y     Y                             Y Y       Y Y Y   Y   Y
+   Rx interrupt                     Y     Y Y Y Y Y Y Y Y Y Y Y Y Y Y
+   Queue start/stop           Y   Y   Y Y Y Y Y Y     Y Y     Y Y Y Y Y Y               Y   Y Y
+   MTU update                     Y Y Y           Y   Y Y Y Y         Y Y
+   Jumbo frame                    Y Y Y Y Y Y Y Y Y   Y Y Y Y Y Y Y Y Y Y       Y Y Y
+   Scattered Rx                   Y Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y               Y   Y
+   LRO                                                Y Y Y Y
+   TSO                            Y   Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y
+   Promiscuous mode       Y Y Y   Y Y   Y Y Y Y Y Y Y Y Y     Y Y     Y Y         Y Y   Y   Y Y
+   Allmulticast mode              Y Y     Y Y Y Y Y Y Y Y Y Y Y Y     Y Y         Y Y   Y   Y Y
+   Unicast MAC filter     Y Y Y     Y   Y Y Y Y Y Y Y Y Y Y Y Y Y     Y Y         Y Y       Y Y
+   Multicast MAC filter   Y Y Y         Y Y Y Y Y             Y Y     Y Y         Y Y       Y Y
+   RSS hash                       Y   Y Y Y Y Y Y Y   Y Y Y Y Y Y Y Y Y Y         Y Y
+   RSS key update                     Y   Y Y Y Y Y   Y Y Y Y Y Y Y Y   Y
+   RSS reta update            Y       Y   Y Y Y Y Y   Y Y Y Y Y Y Y Y   Y
+   VMDq                                   Y Y     Y   Y Y     Y Y
+   SR-IOV                   Y         Y   Y Y     Y   Y Y             Y Y           Y
+   DCB                                    Y Y     Y   Y Y
+   VLAN filter                      Y   Y Y Y Y Y Y Y Y Y Y Y Y Y     Y Y         Y Y       Y Y
+   Ethertype filter                       Y Y     Y   Y Y
+   N-tuple filter                                 Y   Y Y
+   SYN filter                                     Y   Y Y
+   Tunnel filter                          Y Y         Y Y
+   Flexible filter                                Y
+   Hash filter                            Y Y Y Y
+   Flow director                          Y Y         Y Y               Y
+   Flow control                   Y Y     Y Y     Y   Y Y                         Y Y
+   Rate limitation                                    Y Y
+   Traffic mirroring                      Y Y         Y Y
+   CRC offload                    Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y   Y         Y Y
+   VLAN offload                   Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y   Y         Y Y
+   QinQ offload                     Y     Y   Y   Y Y Y   Y
+   L3 checksum offload            Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y Y Y
+   L4 checksum offload            Y Y Y Y Y   Y   Y Y Y   Y   Y Y Y Y Y Y
+   Inner L3 checksum                  Y   Y   Y       Y   Y           Y
+   Inner L4 checksum                  Y   Y   Y       Y   Y           Y
+   Packet type parsing            Y     Y Y   Y   Y Y Y   Y   Y Y Y Y Y Y         Y Y
+   Timesync                               Y Y     Y   Y Y
+   Basic stats            Y Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y       Y Y Y   Y Y Y Y
+   Extended stats             Y       Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y                   Y Y
+   Stats per queue                Y                   Y Y     Y Y Y Y Y Y         Y Y   Y   Y Y
+   EEPROM dump                                    Y   Y Y
+   Registers dump                                 Y Y Y Y Y Y
+   Multiprocess aware                     Y Y Y Y     Y Y Y Y Y Y Y Y Y Y       Y Y Y
+   BSD nic_uio                    Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y                       Y Y
+   Linux UIO              Y Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y             Y Y       Y Y
+   Linux VFIO                     Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y                       Y Y
+   Other kdrv                                                         Y Y               Y
+   ARMv7                                                                        Y           Y Y
+   ARMv8                                                                        Y           Y Y
+   Power8                                                             Y Y       Y
+   TILE-Gx                                                                      Y
+   x86-32                         Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y       Y         Y Y Y
+   x86-64                 Y Y Y   Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y       Y Y Y   Y Y Y Y
+   Usage doc              Y Y     Y     Y                             Y Y       Y Y Y   Y   Y
    Design doc
    Perf doc
-   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
+   ==================== = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
 
 .. Note::
 
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index 6ba7658..3832706 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -45,6 +45,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_MLX4_PMD) += mlx4
 DIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5
 DIRS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD) += mpipe
 DIRS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp
+DIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_NULL) += null
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_PCAP) += pcap
 DIRS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede
diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
new file mode 100644
index 0000000..f6333fd
--- /dev/null
+++ b/drivers/net/bnxt/Makefile
@@ -0,0 +1,63 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
+#   Copyright(c) 2014 6WIND S.A.
+#   Copyright(c) Broadcom Limited.
+#   All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_bnxt.a
+
+LIBABIVER := 1
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_bnxt_version.map
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+
+#
+# Export include files
+#
+SYMLINK-y-include +=
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += lib/librte_eal
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
new file mode 100644
index 0000000..af8a65f
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -0,0 +1,109 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include <rte_dev.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+
+#define DRV_MODULE_NAME		"bnxt"
+static const char bnxt_version[] =
+	"Broadcom Cumulus driver " DRV_MODULE_NAME "\n";
+
+static struct rte_pci_id bnxt_pci_id_map[] = {
+#define RTE_PCI_DEV_ID_DECL_BNXT(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
+#include "rte_pci_dev_ids.h"
+	{.device_id = 0},
+};
+
+/*
+ * Initialization
+ */
+
+static int
+bnxt_dev_init(struct rte_eth_dev *eth_dev)
+{
+	static int version_printed;
+	int rc;
+
+	if (version_printed++ == 0)
+		RTE_LOG(INFO, PMD, "%s", bnxt_version);
+
+	if (eth_dev->pci_dev->addr.function >= 2 &&
+			eth_dev->pci_dev->addr.function < 4) {
+		RTE_LOG(ERR, PMD, "Function not enabled %x:\n",
+			eth_dev->pci_dev->addr.function);
+		rc = -ENOMEM;
+		goto error;
+	}
+
+	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
+	rc = -EPERM;
+
+error:
+	return rc;
+}
+
+static int
+bnxt_dev_uninit(struct rte_eth_dev *eth_dev __rte_unused) {
+	return 0;
+}
+
+static struct eth_driver bnxt_rte_pmd = {
+	.pci_drv = {
+		    .name = "rte_" DRV_MODULE_NAME "_pmd",
+		    .id_table = bnxt_pci_id_map,
+		    .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
+		    },
+	.eth_dev_init = bnxt_dev_init,
+	.eth_dev_uninit = bnxt_dev_uninit,
+	.dev_private_size = 32 /* this must be non-zero apparently */,
+};
+
+static int bnxt_rte_pmd_init(const char *name, const char *params __rte_unused)
+{
+	RTE_LOG(INFO, PMD, "bnxt_rte_pmd_init() called for %s\n", name);
+	rte_eth_driver_register(&bnxt_rte_pmd);
+	return 0;
+}
+
+static struct rte_driver bnxt_pmd_drv = {
+	.name = "eth_bnxt",
+	.type = PMD_PDEV,
+	.init = bnxt_rte_pmd_init,
+};
+
+PMD_REGISTER_DRIVER(bnxt_pmd_drv);
diff --git a/drivers/net/bnxt/rte_pmd_bnxt_version.map b/drivers/net/bnxt/rte_pmd_bnxt_version.map
new file mode 100644
index 0000000..349c6e1
--- /dev/null
+++ b/drivers/net/bnxt/rte_pmd_bnxt_version.map
@@ -0,0 +1,4 @@
+DPDK_16.04 {
+
+	local: *;
+};
diff --git a/lib/librte_eal/common/include/rte_pci_dev_ids.h b/lib/librte_eal/common/include/rte_pci_dev_ids.h
index cf7b548..f63cdc0 100644
--- a/lib/librte_eal/common/include/rte_pci_dev_ids.h
+++ b/lib/librte_eal/common/include/rte_pci_dev_ids.h
@@ -63,11 +63,12 @@
  * This file contains a list of the PCI device IDs recognised by DPDK, which
  * can be used to fill out an array of structures describing the devices.
  *
- * Currently four families of devices are recognised: those supported by the
- * IGB driver, by EM driver, those supported by the IXGBE driver, and by virtio
- * driver which is a para virtualization driver running in guest virtual machine.
- * The inclusion of these in an array built using this file depends on the
- * definition of
+ * Currently five families of devices are recognised: those supported by the
+ * IGB driver, by EM driver, those supported by the IXGBE driver, those
+ * supported by the BNXT driver, and by virtio driver which is a para
+ * virtualization driver running in guest virtual machine. The inclusion of
+ * these in an array built using this file depends on the definition of
+ * RTE_PCI_DEV_ID_DECL_BNXT
  * RTE_PCI_DEV_ID_DECL_EM
  * RTE_PCI_DEV_ID_DECL_IGB
  * RTE_PCI_DEV_ID_DECL_IGBVF
@@ -152,6 +153,10 @@
 #define RTE_PCI_DEV_ID_DECL_BNX2XVF(vend, dev)
 #endif
 
+#ifndef RTE_PCI_DEV_ID_DECL_BNXT
+#define RTE_PCI_DEV_ID_DECL_BNXT(vend, dev)
+#endif
+
 #ifndef PCI_VENDOR_ID_INTEL
 /** Vendor ID used by Intel devices */
 #define PCI_VENDOR_ID_INTEL 0x8086
@@ -686,6 +691,28 @@ RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57811_MF)
 RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57840_MF)
 #endif
 
+/****************** Broadcom bnxt devices ******************/
+
+#define BROADCOM_DEV_ID_57301                  0x16c8
+#define BROADCOM_DEV_ID_57302                  0x16c9
+#define BROADCOM_DEV_ID_57304_PF               0x16ca
+#define BROADCOM_DEV_ID_57304_VF               0x16cb
+#define BROADCOM_DEV_ID_57402                  0x16d0
+#define BROADCOM_DEV_ID_57404                  0x16d1
+#define BROADCOM_DEV_ID_57406_PF               0x16d2
+#define BROADCOM_DEV_ID_57406_VF               0x16d3
+#define BROADCOM_DEV_ID_57406_MF               0x16d4
+
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57301)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF)
+RTE_PCI_DEV_ID_DECL_BNXT(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF)
+
 /*
  * Undef all RTE_PCI_DEV_ID_DECL_* here.
  */
@@ -702,3 +729,4 @@ RTE_PCI_DEV_ID_DECL_BNX2X(PCI_VENDOR_ID_BROADCOM, BNX2X_DEV_ID_57840_MF)
 #undef RTE_PCI_DEV_ID_DECL_VMXNET3
 #undef RTE_PCI_DEV_ID_DECL_FM10K
 #undef RTE_PCI_DEV_ID_DECL_FM10KVF
+#undef RTE_PCI_DEV_ID_DECL_BNXT
diff --git a/mk/rte.app.mk b/mk/rte.app.mk
index e9969fc..c62ba64 100644
--- a/mk/rte.app.mk
+++ b/mk/rte.app.mk
@@ -103,6 +103,7 @@ ifeq ($(CONFIG_RTE_BUILD_SHARED_LIB),n)
 
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_AF_PACKET)  += -lrte_pmd_af_packet
 _LDLIBS-$(CONFIG_RTE_LIBRTE_BNX2X_PMD)      += -lrte_pmd_bnx2x -lz
+_LDLIBS-$(CONFIG_RTE_LIBRTE_BNXT_PMD)       += -lrte_pmd_bnxt
 _LDLIBS-$(CONFIG_RTE_LIBRTE_CXGBE_PMD)      += -lrte_pmd_cxgbe
 _LDLIBS-$(CONFIG_RTE_LIBRTE_E1000_PMD)      += -lrte_pmd_e1000
 _LDLIBS-$(CONFIG_RTE_LIBRTE_ENA_PMD)        += -lrte_pmd_ena
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 02/38] bnxt: add HWRM init code
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 03/38] bnxt: add driver register/unregister support Stephen Hurd
                       ` (36 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Start adding support to use the HWRM API.
Hardware Resource Manager or HWRM in short, is a set of API provided
by the firmware running in the ASIC to manage the various resources.

Initial commit just performs necessary HWRM queries for init, then
fails as before.

Now that struct bnxt is non-zero size, we can set dev_private_size
correctly.

The used HWRM calls so far:
bnxt_hwrm_func_qcaps:
	This command returns capabilities of a function.

bnxt_hwrm_ver_get:
	This function is called by a driver to determine the HWRM
	interface version supported by the HWRM firmware, the
	version of HWRM firmware implementation, the name of HWRM
	firmware, the versions of other embedded firmwares, and
	the names of other embedded firmwares, etc.  Gets the
	firmware version and interface specifications.  Returns
	an error if the firmware on the device is not supported
	by the driver and ensures the response space is large
	enough for the largest possible response.

bnxt_hwrm_queue_qportcfg:
	This function is called by a driver to query queue
	configuration of a port.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix few issues highlighted by checkpatch.

v5:
Fix footer formatting
Switch to using the HWRM function descriptions from the source
documentation
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                | 114 ++++
 drivers/net/bnxt/bnxt_ethdev.c         | 111 +++-
 drivers/net/bnxt/bnxt_hwrm.c           | 324 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  53 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 954 +++++++++++++++++++++++++++++++++
 6 files changed, 1555 insertions(+), 2 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt.h
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.c
 create mode 100644 drivers/net/bnxt/bnxt_hwrm.h
 create mode 100644 drivers/net/bnxt/hsi_struct_def_dpdk.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index f6333fd..9965597 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -49,6 +49,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 # all source are stored in SRCS-y
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 
 #
 # Export include files
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
new file mode 100644
index 0000000..8cb7f5b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt.h
@@ -0,0 +1,114 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_H_
+#define _BNXT_H_
+
+#include <inttypes.h>
+#include <sys/queue.h>
+
+#include <rte_ethdev.h>
+#include <rte_memory.h>
+#include <rte_lcore.h>
+#include <rte_spinlock.h>
+
+struct bnxt_vf_info {
+	uint16_t		fw_fid;
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+	uint16_t		max_rsscos_ctx;
+	uint16_t		max_cp_rings;
+	uint16_t		max_tx_rings;
+	uint16_t		max_rx_rings;
+	uint16_t		max_l2_ctx;
+	uint16_t		max_vnics;
+	struct bnxt_pf_info	*pf;
+};
+
+struct bnxt_pf_info {
+#define BNXT_FIRST_PF_FID	1
+#define BNXT_MAX_VFS(bp)	(bp->pf.max_vfs)
+#define BNXT_FIRST_VF_FID	128
+#define BNXT_PF_RINGS_USED(bp)	bnxt_get_num_queues(bp)
+#define BNXT_PF_RINGS_AVAIL(bp)	(bp->pf.max_cp_rings - BNXT_PF_RINGS_USED(bp))
+	uint32_t		fw_fid;
+	uint8_t			port_id;
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+	uint16_t		max_rsscos_ctx;
+	uint16_t		max_cp_rings;
+	uint16_t		max_tx_rings;
+	uint16_t		max_rx_rings;
+	uint16_t		max_l2_ctx;
+	uint16_t		max_vnics;
+	uint16_t		first_vf_id;
+	uint16_t		active_vfs;
+	uint16_t		max_vfs;
+	void			*vf_req_buf;
+	phys_addr_t		vf_req_buf_dma_addr;
+	uint32_t		vf_req_fwd[8];
+	struct bnxt_vf_info	*vf;
+};
+
+#define BNXT_COS_QUEUE_COUNT	8
+struct bnxt_cos_queue_info {
+	uint8_t	id;
+	uint8_t	profile;
+};
+
+struct bnxt {
+	void				*bar0;
+
+	struct rte_eth_dev		*eth_dev;
+	struct rte_pci_device		*pdev;
+
+	uint32_t		flags;
+#define BNXT_FLAG_VF		(1 << 1)
+#define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
+#define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
+
+#define MAX_NUM_MAC_ADDR	32
+	uint8_t			mac_addr[ETHER_ADDR_LEN];
+
+	uint16_t			hwrm_cmd_seq;
+	void				*hwrm_cmd_resp_addr;
+	phys_addr_t			hwrm_cmd_resp_dma_addr;
+	rte_spinlock_t			hwrm_lock;
+	uint16_t			max_req_len;
+	uint16_t			max_resp_len;
+
+	struct bnxt_cos_queue_info	cos_queue[BNXT_COS_QUEUE_COUNT];
+
+	struct bnxt_pf_info		pf;
+	struct bnxt_vf_info		vf;
+};
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index af8a65f..32ed081 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -39,6 +39,9 @@
 #include <rte_malloc.h>
 #include <rte_cycles.h>
 
+#include "bnxt.h"
+#include "bnxt_hwrm.h"
+
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
 	"Broadcom Cumulus driver " DRV_MODULE_NAME "\n";
@@ -49,14 +52,68 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
+static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	rte_free(eth_dev->data->mac_addrs);
+	bnxt_free_hwrm_resources(bp);
+}
+
 /*
  * Initialization
  */
 
+static struct eth_dev_ops bnxt_dev_ops = {
+	.dev_close = bnxt_dev_close_op,
+};
+
+static bool bnxt_vf_pciid(uint16_t id)
+{
+	if (id == BROADCOM_DEV_ID_57304_VF ||
+	    id == BROADCOM_DEV_ID_57406_VF)
+		return true;
+	return false;
+}
+
+static int bnxt_init_board(struct rte_eth_dev *eth_dev)
+{
+	int rc;
+	struct bnxt *bp = eth_dev->data->dev_private;
+
+	/* enable device (incl. PCI PM wakeup), and bus-mastering */
+	if (!eth_dev->pci_dev->mem_resource[0].addr) {
+		RTE_LOG(ERR, PMD,
+			"Cannot find PCI device base address, aborting\n");
+		rc = -ENODEV;
+		goto init_err_disable;
+	}
+
+	bp->eth_dev = eth_dev;
+	bp->pdev = eth_dev->pci_dev;
+
+	bp->bar0 = (void *)eth_dev->pci_dev->mem_resource[0].addr;
+	if (!bp->bar0) {
+		RTE_LOG(ERR, PMD, "Cannot map device registers, aborting\n");
+		rc = -ENOMEM;
+		goto init_err_release;
+	}
+	return 0;
+
+init_err_release:
+	if (bp->bar0)
+		bp->bar0 = NULL;
+
+init_err_disable:
+
+	return rc;
+}
+
 static int
 bnxt_dev_init(struct rte_eth_dev *eth_dev)
 {
 	static int version_printed;
+	struct bnxt *bp;
 	int rc;
 
 	if (version_printed++ == 0)
@@ -71,8 +128,58 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	}
 
 	rte_eth_copy_pci_info(eth_dev, eth_dev->pci_dev);
-	rc = -EPERM;
+	bp = eth_dev->data->dev_private;
+
+	if (bnxt_vf_pciid(eth_dev->pci_dev->id.device_id))
+		bp->flags |= BNXT_FLAG_VF;
+
+	rc = bnxt_init_board(eth_dev);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Board initialization failed rc: %x\n", rc);
+		goto error;
+	}
+	eth_dev->dev_ops = &bnxt_dev_ops;
+	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
+	/* eth_dev->tx_pkt_burst = &bnxt_xmit_pkts; */
+
+	rc = bnxt_alloc_hwrm_resources(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"hwrm resource allocation failure rc: %x\n", rc);
+		goto error_free;
+	}
+	rc = bnxt_hwrm_ver_get(bp);
+	if (rc)
+		goto error_free;
+	bnxt_hwrm_queue_qportcfg(bp);
+
+	/* Get the MAX capabilities for this function */
+	rc = bnxt_hwrm_func_qcaps(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "hwrm query capability failure rc: %x\n", rc);
+		goto error_free;
+	}
+	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
+					ETHER_ADDR_LEN * MAX_NUM_MAC_ADDR, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		RTE_LOG(ERR, PMD,
+			"Failed to alloc %u bytes needed to store MAC addr tbl",
+			ETHER_ADDR_LEN * MAX_NUM_MAC_ADDR);
+		rc = -ENOMEM;
+		goto error_free;
+	}
+	/* Copy the permanent MAC from the qcap response address now. */
+	if (BNXT_PF(bp))
+		memcpy(bp->mac_addr, bp->pf.mac_addr, sizeof(bp->mac_addr));
+	else
+		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
+	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
+
+	return -EPERM;
 
+error_free:
+	bnxt_dev_close_op(eth_dev);
 error:
 	return rc;
 }
@@ -90,7 +197,7 @@ static struct eth_driver bnxt_rte_pmd = {
 		    },
 	.eth_dev_init = bnxt_dev_init,
 	.eth_dev_uninit = bnxt_dev_uninit,
-	.dev_private_size = 32 /* this must be non-zero apparently */,
+	.dev_private_size = sizeof(struct bnxt),
 };
 
 static int bnxt_rte_pmd_init(const char *name, const char *params __rte_unused)
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
new file mode 100644
index 0000000..e187121
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -0,0 +1,324 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_byteorder.h>
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+
+#include "bnxt.h"
+#include "bnxt_hwrm.h"
+#include "hsi_struct_def_dpdk.h"
+
+#define HWRM_CMD_TIMEOUT		2000
+
+/*
+ * HWRM Functions (sent to HWRM)
+ * These are named bnxt_hwrm_*() and return -1 if bnxt_hwrm_send_message()
+ * fails (ie: a timeout), and a positive non-zero HWRM error code if the HWRM
+ * command was failed by the ChiMP.
+ */
+
+static int bnxt_hwrm_send_message_locked(struct bnxt *bp, void *msg,
+					uint32_t msg_len)
+{
+	unsigned int i;
+	struct input *req = msg;
+	struct output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t *data = msg;
+	uint8_t *bar;
+	uint8_t *valid;
+
+	/* Write request msg to hwrm channel */
+	for (i = 0; i < msg_len; i += 4) {
+		bar = (uint8_t *)bp->bar0 + i;
+		*(volatile uint32_t *)bar = *data;
+		data++;
+	}
+
+	/* Zero the rest of the request space */
+	for (; i < bp->max_req_len; i += 4) {
+		bar = (uint8_t *)bp->bar0 + i;
+		*(volatile uint32_t *)bar = 0;
+	}
+
+	/* Ring channel doorbell */
+	bar = (uint8_t *)bp->bar0 + 0x100;
+	*(volatile uint32_t *)bar = 1;
+
+	/* Poll for the valid bit */
+	for (i = 0; i < HWRM_CMD_TIMEOUT; i++) {
+		/* Sanity check on the resp->resp_len */
+		rte_rmb();
+		if (resp->resp_len && resp->resp_len <=
+		    bp->max_resp_len) {
+			/* Last byte of resp contains the valid key */
+			valid = (uint8_t *)resp + resp->resp_len - 1;
+			if (*valid == HWRM_RESP_VALID_KEY)
+				break;
+		}
+		rte_delay_us(600);
+	}
+
+	if (i >= HWRM_CMD_TIMEOUT) {
+		RTE_LOG(ERR, PMD, "Error sending msg %x\n",
+			req->req_type);
+		goto err_ret;
+	}
+	return 0;
+
+err_ret:
+	return -1;
+}
+
+static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
+{
+	int rc;
+
+	rte_spinlock_lock(&bp->hwrm_lock);
+	rc = bnxt_hwrm_send_message_locked(bp, msg, msg_len);
+	rte_spinlock_unlock(&bp->hwrm_lock);
+	return rc;
+}
+
+#define HWRM_PREP(req, type, cr, resp) \
+	memset(bp->hwrm_cmd_resp_addr, 0, bp->max_resp_len); \
+	req.req_type = rte_cpu_to_le_16(HWRM_##type); \
+	req.cmpl_ring = rte_cpu_to_le_16(cr); \
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++); \
+	req.target_id = rte_cpu_to_le_16(0xffff); \
+	req.resp_addr = rte_cpu_to_le_64(bp->hwrm_cmd_resp_dma_addr)
+
+#define HWRM_CHECK_RESULT \
+	{ \
+		if (rc) { \
+			RTE_LOG(ERR, PMD, "%s failed rc:%d\n", \
+				__func__, rc); \
+			return rc; \
+		} \
+		if (resp->error_code) { \
+			rc = rte_le_to_cpu_16(resp->error_code); \
+			RTE_LOG(ERR, PMD, "%s error %d\n", __func__, rc); \
+			return rc; \
+		} \
+	}
+
+int bnxt_hwrm_func_qcaps(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_func_qcaps_input req = {.req_type = 0 };
+	struct hwrm_func_qcaps_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, FUNC_QCAPS, -1, resp);
+
+	req.fid = rte_cpu_to_le_16(0xffff);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		pf->fw_fid = rte_le_to_cpu_32(resp->fid);
+		pf->port_id = resp->port_id;
+		memcpy(pf->mac_addr, resp->perm_mac_address, ETHER_ADDR_LEN);
+		pf->max_rsscos_ctx = rte_le_to_cpu_16(resp->max_rsscos_ctx);
+		pf->max_cp_rings = rte_le_to_cpu_16(resp->max_cmpl_rings);
+		pf->max_tx_rings = rte_le_to_cpu_16(resp->max_tx_rings);
+		pf->max_rx_rings = rte_le_to_cpu_16(resp->max_rx_rings);
+		pf->max_l2_ctx = rte_le_to_cpu_16(resp->max_l2_ctxs);
+		pf->max_vnics = rte_le_to_cpu_16(resp->max_vnics);
+		pf->first_vf_id = rte_le_to_cpu_16(resp->first_vf_id);
+		pf->max_vfs = rte_le_to_cpu_16(resp->max_vfs);
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		vf->fw_fid = rte_le_to_cpu_32(resp->fid);
+		memcpy(vf->mac_addr, &resp->perm_mac_address, ETHER_ADDR_LEN);
+		vf->max_rsscos_ctx = rte_le_to_cpu_16(resp->max_rsscos_ctx);
+		vf->max_cp_rings = rte_le_to_cpu_16(resp->max_cmpl_rings);
+		vf->max_tx_rings = rte_le_to_cpu_16(resp->max_tx_rings);
+		vf->max_rx_rings = rte_le_to_cpu_16(resp->max_rx_rings);
+		vf->max_l2_ctx = rte_le_to_cpu_16(resp->max_l2_ctxs);
+		vf->max_vnics = rte_le_to_cpu_16(resp->max_vnics);
+	}
+
+	return rc;
+}
+
+int bnxt_hwrm_ver_get(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_ver_get_input req = {.req_type = 0 };
+	struct hwrm_ver_get_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t my_version;
+	uint32_t fw_version;
+	uint16_t max_resp_len;
+	char type[RTE_MEMZONE_NAMESIZE];
+
+	HWRM_PREP(req, VER_GET, -1, resp);
+
+	req.hwrm_intf_maj = HWRM_VERSION_MAJOR;
+	req.hwrm_intf_min = HWRM_VERSION_MINOR;
+	req.hwrm_intf_upd = HWRM_VERSION_UPDATE;
+
+	/*
+	 * Hold the lock since we may be adjusting the response pointers.
+	 */
+	rte_spinlock_lock(&bp->hwrm_lock);
+	rc = bnxt_hwrm_send_message_locked(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	RTE_LOG(INFO, PMD, "%d.%d.%d:%d.%d.%d\n",
+		resp->hwrm_intf_maj, resp->hwrm_intf_min,
+		resp->hwrm_intf_upd,
+		resp->hwrm_fw_maj, resp->hwrm_fw_min, resp->hwrm_fw_bld);
+
+	my_version = HWRM_VERSION_MAJOR << 16;
+	my_version |= HWRM_VERSION_MINOR << 8;
+	my_version |= HWRM_VERSION_UPDATE;
+
+	fw_version = resp->hwrm_intf_maj << 16;
+	fw_version |= resp->hwrm_intf_min << 8;
+	fw_version |= resp->hwrm_intf_upd;
+
+	if (resp->hwrm_intf_maj != HWRM_VERSION_MAJOR) {
+		RTE_LOG(ERR, PMD, "Unsupported firmware API version\n");
+		rc = -EINVAL;
+		goto error;
+	}
+
+	if (my_version != fw_version) {
+		RTE_LOG(INFO, PMD, "BNXT Driver/HWRM API mismatch.\n");
+		if (my_version < fw_version) {
+			RTE_LOG(INFO, PMD,
+				"Firmware API version is newer than driver.\n");
+			RTE_LOG(INFO, PMD,
+				"The driver may be missing features.\n");
+		} else {
+			RTE_LOG(INFO, PMD,
+				"Firmware API version is older than driver.\n");
+			RTE_LOG(INFO, PMD,
+				"Not all driver features may be functional.\n");
+		}
+	}
+
+	if (bp->max_req_len > resp->max_req_win_len) {
+		RTE_LOG(ERR, PMD, "Unsupported request length\n");
+		rc = -EINVAL;
+	}
+	bp->max_req_len = resp->max_req_win_len;
+	max_resp_len = resp->max_resp_len;
+	if (bp->max_resp_len != max_resp_len) {
+		sprintf(type, "bnxt_hwrm_%04x:%02x:%02x:%02x",
+			bp->pdev->addr.domain, bp->pdev->addr.bus,
+			bp->pdev->addr.devid, bp->pdev->addr.function);
+
+		rte_free(bp->hwrm_cmd_resp_addr);
+
+		bp->hwrm_cmd_resp_addr = rte_malloc(type, max_resp_len, 0);
+		if (bp->hwrm_cmd_resp_addr == NULL) {
+			rc = -ENOMEM;
+			goto error;
+		}
+		bp->hwrm_cmd_resp_dma_addr =
+			rte_malloc_virt2phy(bp->hwrm_cmd_resp_addr);
+		bp->max_resp_len = max_resp_len;
+	}
+
+error:
+	rte_spinlock_unlock(&bp->hwrm_lock);
+	return rc;
+}
+
+int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_queue_qportcfg_input req = {.req_type = 0 };
+	struct hwrm_queue_qportcfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, QUEUE_QPORTCFG, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+#define GET_QUEUE_INFO(x) \
+	bp->cos_queue[x].id = resp->queue_id##x; \
+	bp->cos_queue[x].profile = resp->queue_id##x##_service_profile
+
+	GET_QUEUE_INFO(0);
+	GET_QUEUE_INFO(1);
+	GET_QUEUE_INFO(2);
+	GET_QUEUE_INFO(3);
+	GET_QUEUE_INFO(4);
+	GET_QUEUE_INFO(5);
+	GET_QUEUE_INFO(6);
+	GET_QUEUE_INFO(7);
+
+	return rc;
+}
+
+/*
+ * HWRM utility functions
+ */
+
+void bnxt_free_hwrm_resources(struct bnxt *bp)
+{
+	/* Release memzone */
+	rte_free(bp->hwrm_cmd_resp_addr);
+	bp->hwrm_cmd_resp_addr = NULL;
+	bp->hwrm_cmd_resp_dma_addr = 0;
+}
+
+int bnxt_alloc_hwrm_resources(struct bnxt *bp)
+{
+	struct rte_pci_device *pdev = bp->pdev;
+	char type[RTE_MEMZONE_NAMESIZE];
+
+	sprintf(type, "bnxt_hwrm_%04x:%02x:%02x:%02x", pdev->addr.domain,
+		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+	bp->max_req_len = HWRM_MAX_REQ_LEN;
+	bp->max_resp_len = HWRM_MAX_RESP_LEN;
+	bp->hwrm_cmd_resp_addr = rte_malloc(type, bp->max_resp_len, 0);
+	if (bp->hwrm_cmd_resp_addr == NULL)
+		return -ENOMEM;
+	bp->hwrm_cmd_resp_dma_addr =
+		rte_malloc_virt2phy(bp->hwrm_cmd_resp_addr);
+	rte_spinlock_init(&bp->hwrm_lock);
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
new file mode 100644
index 0000000..e35e8c0
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -0,0 +1,53 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_HWRM_H_
+#define _BNXT_HWRM_H_
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include "bnxt.h"
+
+#define HWRM_SEQ_ID_INVALID -1U
+
+int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+
+int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
+
+int bnxt_hwrm_ver_get(struct bnxt *bp);
+
+void bnxt_free_hwrm_resources(struct bnxt *bp);
+int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
new file mode 100644
index 0000000..c0bdbda
--- /dev/null
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -0,0 +1,954 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _HSI_STRUCT_DEF_EXTERNAL_H_
+#define _HSI_STRUCT_DEF_EXTERNAL_H_
+
+/* HW Resource Manager Specification 1.2.0 */
+#define HWRM_VERSION_MAJOR	1
+#define HWRM_VERSION_MINOR	2
+#define HWRM_VERSION_UPDATE	0
+
+/*
+ * Following is the signature for HWRM message field that indicates not
+ * applicable (All F's). Need to cast it the size of the field if needed.
+ */
+#define HWRM_MAX_REQ_LEN	(128)  /* hwrm_func_buf_rgtr */
+#define HWRM_MAX_RESP_LEN	(176)  /* hwrm_func_qstats */
+#define HWRM_RESP_VALID_KEY	1 /* valid key for HWRM response */
+
+/*
+ * Request types
+ */
+#define HWRM_VER_GET			(UINT32_C(0x0))
+#define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
+#define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+
+/*
+ * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
+ * inside the chip. The HWRM is implemented in firmware, and runs on embedded
+ * processors inside the chip. This firmware is vital part of the chip's
+ * hardware. The chip can not be used by driver without it.
+ */
+
+/* Input (16 bytes) */
+struct input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+} __attribute__((packed));
+
+/* Output (8 bytes) */
+struct output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+} __attribute__((packed));
+
+/* hwrm_func_qcaps */
+/*
+ * Description: This command returns capabilities of a function. The input FID
+ * value is used to indicate what function is being queried. This allows a
+ * physical function driver to query virtual functions that are children of the
+ * physical function. The output FID value is needed to configure Rings and
+ * MSI-X vectors so their DMA operations appear correctly on the PCI bus.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_func_qcaps_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Function ID of the function that is being queried. 0xFF... (All Fs)
+	 * if the query is for the requesting function.
+	 */
+	uint16_t fid;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (80 bytes) */
+struct hwrm_func_qcaps_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * FID value. This value is used to identify operations on the PCI bus
+	 * as belonging to a particular PCI function.
+	 */
+	uint16_t fid;
+
+	/*
+	 * Port ID of port that this function is associated with. Valid only for
+	 * the PF. 0xFF... (All Fs) if this function is not associated with any
+	 * port. 0xFF... (All Fs) if this function is called from a VF.
+	 */
+	uint16_t port_id;
+
+	/* If 1, then Push mode is supported on this function. */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_PUSH_MODE_SUPPORTED   UINT32_C(0x1)
+	/*
+	 * If 1, then the global MSI-X auto-masking is enabled for the device.
+	 */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_GLOBAL_MSIX_AUTOMASKING \
+								UINT32_C(0x2)
+	/*
+	 * If 1, then the Precision Time Protocol (PTP) processing is supported
+	 * on this function. The HWRM should enable PTP on only a single
+	 * Physical Function (PF) per port.
+	 */
+	#define HWRM_FUNC_QCAPS_OUTPUT_FLAGS_PTP_SUPPORTED         UINT32_C(0x4)
+	uint32_t flags;
+
+	/*
+	 * This value is current MAC address configured for this function. A
+	 * value of 00-00-00-00-00-00 indicates no MAC address is currently
+	 * configured.
+	 */
+	uint8_t perm_mac_address[6];
+
+	/*
+	 * The maximum number of RSS/COS contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_rsscos_ctx;
+
+	/*
+	 * The maximum number of completion rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_cmpl_rings;
+
+	/*
+	 * The maximum number of transmit rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_tx_rings;
+
+	/*
+	 * The maximum number of receive rings that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_rx_rings;
+
+	/*
+	 * The maximum number of L2 contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_l2_ctxs;
+
+	/* The maximum number of VNICs that can be allocated to the function. */
+	uint16_t max_vnics;
+
+	/*
+	 * The identifier for the first VF enabled on a PF. This is valid only
+	 * on the PF with SR-IOV enabled. 0xFF... (All Fs) if this command is
+	 * called on a PF with SR-IOV disabled or on a VF.
+	 */
+	uint16_t first_vf_id;
+
+	/*
+	 * The maximum number of VFs that can be allocated to the function. This
+	 * is valid only on the PF with SR-IOV enabled. 0xFF... (All Fs) if this
+	 * command is called on a PF with SR-IOV disabled or on a VF.
+	 */
+	uint16_t max_vfs;
+
+	/*
+	 * The maximum number of statistic contexts that can be allocated to the
+	 * function.
+	 */
+	uint16_t max_stat_ctx;
+
+	/*
+	 * The maximum number of Encapsulation records that can be offloaded by
+	 * this function.
+	 */
+	uint32_t max_encap_records;
+
+	/*
+	 * The maximum number of decapsulation records that can be offloaded by
+	 * this function.
+	 */
+	uint32_t max_decap_records;
+
+	/*
+	 * The maximum number of Exact Match (EM) flows that can be offloaded by
+	 * this function on the TX side.
+	 */
+	uint32_t max_tx_em_flows;
+
+	/*
+	 * The maximum number of Wildcard Match (WM) flows that can be offloaded
+	 * by this function on the TX side.
+	 */
+	uint32_t max_tx_wm_flows;
+
+	/*
+	 * The maximum number of Exact Match (EM) flows that can be offloaded by
+	 * this function on the RX side.
+	 */
+	uint32_t max_rx_em_flows;
+
+	/*
+	 * The maximum number of Wildcard Match (WM) flows that can be offloaded
+	 * by this function on the RX side.
+	 */
+	uint32_t max_rx_wm_flows;
+
+	/*
+	 * The maximum number of multicast filters that can be supported by this
+	 * function on the RX side.
+	 */
+	uint32_t max_mcast_filters;
+
+	/*
+	 * The maximum value of flow_id that can be supported in completion
+	 * records.
+	 */
+	uint32_t max_flow_id;
+
+	/*
+	 * The maximum number of HW ring groups that can be supported on this
+	 * function.
+	 */
+	uint32_t max_hw_ring_grps;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ver_get */
+/*
+ * Description: This function is called by a driver to determine the HWRM
+ * interface version supported by the HWRM firmware, the version of HWRM
+ * firmware implementation, the name of HWRM firmware, the versions of other
+ * embedded firmwares, and the names of other embedded firmwares, etc. Any
+ * interface or firmware version with major = 0, minor = 0, and update = 0 shall
+ * be considered an invalid version.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ver_get_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This field represents the major version of HWRM interface
+	 * specification supported by the driver HWRM implementation. The
+	 * interface major version is intended to change only when non backward
+	 * compatible changes are made to the HWRM interface specification.
+	 */
+	uint8_t hwrm_intf_maj;
+
+	/*
+	 * This field represents the minor version of HWRM interface
+	 * specification supported by the driver HWRM implementation. A change
+	 * in interface minor version is used to reflect significant backward
+	 * compatible modification to HWRM interface specification. This can be
+	 * due to addition or removal of functionality. HWRM interface
+	 * specifications with the same major version but different minor
+	 * versions are compatible.
+	 */
+	uint8_t hwrm_intf_min;
+
+	/*
+	 * This field represents the update version of HWRM interface
+	 * specification supported by the driver HWRM implementation. The
+	 * interface update version is used to reflect minor changes or bug
+	 * fixes to a released HWRM interface specification.
+	 */
+	uint8_t hwrm_intf_upd;
+
+	uint8_t unused_0[5];
+} __attribute__((packed));
+
+/* Output (128 bytes) */
+struct hwrm_ver_get_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This field represents the major version of HWRM interface
+	 * specification supported by the HWRM implementation. The interface
+	 * major version is intended to change only when non backward compatible
+	 * changes are made to the HWRM interface specification. A HWRM
+	 * implementation that is compliant with this specification shall
+	 * provide value of 1 in this field.
+	 */
+	uint8_t hwrm_intf_maj;
+
+	/*
+	 * This field represents the minor version of HWRM interface
+	 * specification supported by the HWRM implementation. A change in
+	 * interface minor version is used to reflect significant backward
+	 * compatible modification to HWRM interface specification. This can be
+	 * due to addition or removal of functionality. HWRM interface
+	 * specifications with the same major version but different minor
+	 * versions are compatible. A HWRM implementation that is compliant with
+	 * this specification shall provide value of 0 in this field.
+	 */
+	uint8_t hwrm_intf_min;
+
+	/*
+	 * This field represents the update version of HWRM interface
+	 * specification supported by the HWRM implementation. The interface
+	 * update version is used to reflect minor changes or bug fixes to a
+	 * released HWRM interface specification. A HWRM implementation that is
+	 * compliant with this specification shall provide value of 1 in this
+	 * field.
+	 */
+	uint8_t hwrm_intf_upd;
+
+	uint8_t hwrm_intf_rsvd;
+
+	/*
+	 * This field represents the major version of HWRM firmware. A change in
+	 * firmware major version represents a major firmware release.
+	 */
+	uint8_t hwrm_fw_maj;
+
+	/*
+	 * This field represents the minor version of HWRM firmware. A change in
+	 * firmware minor version represents significant firmware functionality
+	 * changes.
+	 */
+	uint8_t hwrm_fw_min;
+
+	/*
+	 * This field represents the build version of HWRM firmware. A change in
+	 * firmware build version represents bug fixes to a released firmware.
+	 */
+	uint8_t hwrm_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version of the HWRM firmware.
+	 */
+	uint8_t hwrm_fw_rsvd;
+
+	/*
+	 * This field represents the major version of mgmt firmware. A change in
+	 * major version represents a major release.
+	 */
+	uint8_t mgmt_fw_maj;
+
+	/*
+	 * This field represents the minor version of mgmt firmware. A change in
+	 * minor version represents significant functionality changes.
+	 */
+	uint8_t mgmt_fw_min;
+
+	/*
+	 * This field represents the build version of mgmt firmware. A change in
+	 * update version represents bug fixes.
+	 */
+	uint8_t mgmt_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t mgmt_fw_rsvd;
+
+	/*
+	 * This field represents the major version of network control firmware.
+	 * A change in major version represents a major release.
+	 */
+	uint8_t netctrl_fw_maj;
+
+	/*
+	 * This field represents the minor version of network control firmware.
+	 * A change in minor version represents significant functionality
+	 * changes.
+	 */
+	uint8_t netctrl_fw_min;
+
+	/*
+	 * This field represents the build version of network control firmware.
+	 * A change in update version represents bug fixes.
+	 */
+	uint8_t netctrl_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t netctrl_fw_rsvd;
+
+	/*
+	 * This field is reserved for future use. The responder should set it to
+	 * 0. The requester should ignore this field.
+	 */
+	uint32_t reserved1;
+
+	/*
+	 * This field represents the major version of RoCE firmware. A change in
+	 * major version represents a major release.
+	 */
+	uint8_t roce_fw_maj;
+
+	/*
+	 * This field represents the minor version of RoCE firmware. A change in
+	 * minor version represents significant functionality changes.
+	 */
+	uint8_t roce_fw_min;
+
+	/*
+	 * This field represents the build version of RoCE firmware. A change in
+	 * update version represents bug fixes.
+	 */
+	uint8_t roce_fw_bld;
+
+	/*
+	 * This field is a reserved field. This field can be used to represent
+	 * firmware branches or customer specific releases tied to a specific
+	 * (major,minor,update) version
+	 */
+	uint8_t roce_fw_rsvd;
+
+	/*
+	 * This field represents the name of HWRM FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char hwrm_fw_name[16];
+
+	/*
+	 * This field represents the name of mgmt FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char mgmt_fw_name[16];
+
+	/*
+	 * This field represents the name of network control firmware (ASCII
+	 * chars without NULL at the end).
+	 */
+	char netctrl_fw_name[16];
+
+	/*
+	 * This field is reserved for future use. The responder should set it to
+	 * 0. The requester should ignore this field.
+	 */
+	uint32_t reserved2[4];
+
+	/*
+	 * This field represents the name of RoCE FW (ASCII chars without NULL
+	 * at the end).
+	 */
+	char roce_fw_name[16];
+
+	/* This field returns the chip number. */
+	uint16_t chip_num;
+
+	/* This field returns the revision of chip. */
+	uint8_t chip_rev;
+
+	/* This field returns the chip metal number. */
+	uint8_t chip_metal;
+
+	/* This field returns the bond id of the chip. */
+	uint8_t chip_bond_id;
+
+	/*
+	 * This value indicates the type of platform used for chip
+	 * implementation.
+	 */
+	/* ASIC */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_ASIC \
+							(UINT32_C(0x0) << 0)
+	/* FPGA platform of the chip. */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_FPGA \
+							(UINT32_C(0x1) << 0)
+	/* Palladium platform of the chip. */
+	#define HWRM_VER_GET_OUTPUT_CHIP_PLATFORM_TYPE_PALLADIUM \
+							(UINT32_C(0x2) << 0)
+	uint8_t chip_platform_type;
+
+	/*
+	 * This field returns the maximum value of request window that is
+	 * supported by the HWRM. The request window is mapped into device
+	 * address space using MMIO.
+	 */
+	uint16_t max_req_win_len;
+
+	/*
+	 * This field returns the maximum value of response buffer in bytes. If
+	 * a request specifies the response buffer length that is greater than
+	 * this value, then the HWRM should fail it. The value of this field
+	 * shall be 4KB or more.
+	 */
+	uint16_t max_resp_len;
+
+	/*
+	 * This field returns the default request timeout value in milliseconds.
+	 */
+	uint16_t def_req_timeout;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_queue_qportcfg */
+/*
+ * Description: This function is called by a driver to query queue configuration
+ * of a port. # The HWRM shall at least advertise one queue with lossy service
+ * profile. # The driver shall use this command to query queue ids before
+ * configuring or using any queues. # If a service profile is not set for a
+ * queue, then the driver shall not use that queue without configuring a service
+ * profile for it. # If the driver is not allowed to configure service profiles,
+ * then the driver shall only use queues for which service profiles are pre-
+ * configured.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_queue_qportcfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Enumeration denoting the RX, TX type of the resource. This
+	 * enumeration is used for resources that are similar for both TX and RX
+	 * paths of the chip.
+	 */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH \
+							UINT32_C(0x1)
+		/* tx path */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_TX \
+							(UINT32_C(0x0) << 0)
+		/* rx path */
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_RX \
+							(UINT32_C(0x1) << 0)
+	#define HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_LAST \
+					HWRM_QUEUE_QPORTCFG_INPUT_FLAGS_PATH_RX
+	uint32_t flags;
+
+	/*
+	 * Port ID of port for which the queue configuration is being queried.
+	 * This field is only required when sent by IPC.
+	 */
+	uint16_t port_id;
+
+	uint16_t unused_0;
+} __attribute__((packed));
+
+/* Output (32 bytes) */
+struct hwrm_queue_qportcfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* The maximum number of queues that can be configured. */
+	uint8_t max_configurable_queues;
+
+	/* The maximum number of lossless queues that can be configured. */
+	uint8_t max_configurable_lossless_queues;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed. If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure queues using
+	 * hwrm_queue_cfg.
+	 */
+	uint8_t queue_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure queue buffers
+	 * using hwrm_queue_buffers_cfg.
+	 */
+	uint8_t queue_buffers_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure PFC using
+	 * hwrm_queue_pfcenable_cfg.
+	 */
+	uint8_t queue_pfcenable_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure Priority to CoS
+	 * mapping using hwrm_queue_pri2cos_cfg.
+	 */
+	uint8_t queue_pri2cos_cfg_allowed;
+
+	/*
+	 * 0 - Not allowed. Non-zero - Allowed If this value is non-zero, then
+	 * the HWRM shall allow the host SW driver to configure CoS Bandwidth
+	 * configuration using hwrm_queue_cos2bw_cfg.
+	 */
+	uint8_t queue_cos2bw_cfg_allowed;
+
+	/* ID of CoS Queue 0. FF - Invalid id */
+	uint8_t queue_id0;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID0_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id0_service_profile;
+
+	/* ID of CoS Queue 1. FF - Invalid id */
+	uint8_t queue_id1;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID1_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id1_service_profile;
+
+	/* ID of CoS Queue 2. FF - Invalid id */
+	uint8_t queue_id2;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID2_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id2_service_profile;
+
+	/* ID of CoS Queue 3. FF - Invalid id */
+	uint8_t queue_id3;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID3_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id3_service_profile;
+
+	/* ID of CoS Queue 4. FF - Invalid id */
+	uint8_t queue_id4;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID4_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id4_service_profile;
+
+	/* ID of CoS Queue 5. FF - Invalid id */
+	uint8_t queue_id5;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID5_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id5_service_profile;
+
+	/* ID of CoS Queue 6. FF - Invalid id */
+	uint8_t queue_id6_service_profile;
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID6_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id6;
+
+	/* ID of CoS Queue 7. FF - Invalid id */
+	uint8_t queue_id7;
+
+	/* This value is applicable to CoS queues only. */
+		/* Lossy (best-effort) */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_LOSSY \
+							(UINT32_C(0x0) << 0)
+		/* Lossless */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_LOSSLESS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Set to 0xFF... (All Fs) if there is no service profile
+		 * specified
+		 */
+	#define HWRM_QUEUE_QPORTCFG_OUTPUT_QUEUE_ID7_SERVICE_PROFILE_UNKNOWN \
+							(UINT32_C(0xff) << 0)
+	uint8_t queue_id7_service_profile;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 03/38] bnxt: add driver register/unregister support
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 02/38] bnxt: add HWRM init code Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 04/38] bnxt: add dev infos get operation Stephen Hurd
                       ` (35 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Move init() cleanup into uninit() function
Fix .dev_private_size

New HWRM calls:
bnxt_hwrm_func_driver_register:
	This command is used by the function driver to register
	its information with the HWRM.

bnxt_hwrm_func_driver_unregister:
	This command is used by the function driver to unregister
	with the HWRM.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comment regarding removal of bnxt_dev_close_op

v5:
Reformat footer
Add HWRM function descriptions
---
 drivers/net/bnxt/bnxt.h                |   1 +
 drivers/net/bnxt/bnxt_ethdev.c         |  29 +++-
 drivers/net/bnxt/bnxt_hwrm.c           |  50 ++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 271 +++++++++++++++++++++++++++++++++
 5 files changed, 350 insertions(+), 4 deletions(-)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index 8cb7f5b..ed057ef 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -91,6 +91,7 @@ struct bnxt {
 	struct rte_pci_device		*pdev;
 
 	uint32_t		flags;
+#define BNXT_FLAG_REGISTERED	(1 << 0)
 #define BNXT_FLAG_VF		(1 << 1)
 #define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
 #define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 32ed081..26e6447 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -176,17 +176,38 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
 
-	return -EPERM;
+	rc = bnxt_hwrm_func_driver_register(bp, 0,
+					    bp->pf.vf_req_fwd);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Failed to register driver");
+		rc = -EBUSY;
+		goto error_free;
+	}
+
+	RTE_LOG(INFO, PMD,
+		DRV_MODULE_NAME " found at mem %" PRIx64 ", node addr %pM\n",
+		eth_dev->pci_dev->mem_resource[0].phys_addr,
+		eth_dev->pci_dev->mem_resource[0].addr);
+
+	return 0;
 
 error_free:
-	bnxt_dev_close_op(eth_dev);
+	eth_dev->driver->eth_dev_uninit(eth_dev);
 error:
 	return rc;
 }
 
 static int
-bnxt_dev_uninit(struct rte_eth_dev *eth_dev __rte_unused) {
-	return 0;
+bnxt_dev_uninit(struct rte_eth_dev *eth_dev) {
+	struct bnxt *bp = eth_dev->data->dev_private;
+	int rc;
+
+	if (eth_dev->data->mac_addrs)
+		rte_free(eth_dev->data->mac_addrs);
+	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
+	bnxt_free_hwrm_resources(bp);
+	return rc;
 }
 
 static struct eth_driver bnxt_rte_pmd = {
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index e187121..8aba8cd 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -36,6 +36,7 @@
 #include <rte_cycles.h>
 #include <rte_malloc.h>
 #include <rte_memzone.h>
+#include <rte_version.h>
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
@@ -178,6 +179,34 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
+				   uint32_t *vf_req_fwd)
+{
+	int rc;
+	struct hwrm_func_drv_rgtr_input req = {.req_type = 0 };
+	struct hwrm_func_drv_rgtr_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (bp->flags & BNXT_FLAG_REGISTERED)
+		return 0;
+
+	HWRM_PREP(req, FUNC_DRV_RGTR, -1, resp);
+	req.flags = flags;
+	req.enables = HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VER;
+	req.ver_maj = RTE_VER_YEAR;
+	req.ver_min = RTE_VER_MONTH;
+	req.ver_upd = RTE_VER_MINOR;
+
+	memcpy(req.vf_req_fwd, vf_req_fwd, sizeof(req.vf_req_fwd));
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->flags |= BNXT_FLAG_REGISTERED;
+
+	return rc;
+}
+
 int bnxt_hwrm_ver_get(struct bnxt *bp)
 {
 	int rc = 0;
@@ -264,6 +293,27 @@ error:
 	return rc;
 }
 
+int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags)
+{
+	int rc;
+	struct hwrm_func_drv_unrgtr_input req = {.req_type = 0 };
+	struct hwrm_func_drv_unrgtr_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (!(bp->flags & BNXT_FLAG_REGISTERED))
+		return 0;
+
+	HWRM_PREP(req, FUNC_DRV_UNRGTR, -1, resp);
+	req.flags = flags;
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->flags &= ~BNXT_FLAG_REGISTERED;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index e35e8c0..6f2e445 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,7 +41,10 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
+				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index c0bdbda..8d816f4 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -52,6 +52,8 @@
  */
 #define HWRM_VER_GET			(UINT32_C(0x0))
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
+#define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
+#define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 
 /*
@@ -951,4 +953,273 @@ struct hwrm_queue_qportcfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_func_drv_rgtr */
+/*
+ * Description: This command is used by the function driver to register its
+ * information with the HWRM. A function driver shall implement this command. A
+ * function driver shall use this command during the driver initialization right
+ * after the HWRM version discovery and default ring resources allocation.
+ */
+
+/* Input (80 bytes) */
+struct hwrm_func_drv_rgtr_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the function driver is requesting all requests
+	 * from its children VF drivers to be forwarded to itself. This flag can
+	 * only be set by the PF driver. If a VF driver sets this flag, it
+	 * should be ignored by the HWRM.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_FLAGS_FWD_ALL_MODE        UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the function is requesting none of the requests
+	 * from its children VF drivers to be forwarded to itself. This flag can
+	 * only be set by the PF driver. If a VF driver sets this flag, it
+	 * should be ignored by the HWRM.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_FLAGS_FWD_NONE_MODE       UINT32_C(0x2)
+	uint32_t flags;
+
+	/* This bit must be '1' for the os_type field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_OS_TYPE           UINT32_C(0x1)
+	/* This bit must be '1' for the ver field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VER               UINT32_C(0x2)
+	/* This bit must be '1' for the timestamp field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_TIMESTAMP         UINT32_C(0x4)
+	/* This bit must be '1' for the vf_req_fwd field to be configured. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_VF_REQ_FWD        UINT32_C(0x8)
+	/*
+	 * This bit must be '1' for the async_event_fwd field to be configured.
+	 */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_ENABLES_ASYNC_EVENT_FWD \
+								UINT32_C(0x10)
+	uint32_t enables;
+
+	/* This value indicates the type of OS. */
+		/* Unknown */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* Other OS not listed below. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_OTHER \
+							(UINT32_C(0x1) << 0)
+		/* MSDOS OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_MSDOS \
+							(UINT32_C(0xe) << 0)
+		/* Windows OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WINDOWS \
+							(UINT32_C(0x12) << 0)
+		/* Solaris OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_SOLARIS \
+							(UINT32_C(0x1d) << 0)
+		/* Linux OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_LINUX \
+							(UINT32_C(0x24) << 0)
+		/* FreeBSD OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_FREEBSD \
+							(UINT32_C(0x2a) << 0)
+		/* VMware ESXi OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_ESXI \
+							(UINT32_C(0x68) << 0)
+		/* Microsoft Windows 8 64-bit OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WIN864 \
+							(UINT32_C(0x73) << 0)
+		/* Microsoft Windows Server 2012 R2 OS. */
+	#define HWRM_FUNC_DRV_RGTR_INPUT_OS_TYPE_WIN2012R2 \
+							(UINT32_C(0x74) << 0)
+	uint16_t os_type;
+
+	/* This is the major version of the driver. */
+	uint8_t ver_maj;
+
+	/* This is the minor version of the driver. */
+	uint8_t ver_min;
+
+	/* This is the update version of the driver. */
+	uint8_t ver_upd;
+
+	uint8_t unused_0;
+	uint16_t unused_1;
+
+	/*
+	 * This is a 32-bit timestamp provided by the driver for keep alive. The
+	 * timestamp is in multiples of 1ms.
+	 */
+	uint32_t timestamp;
+
+	uint32_t unused_2;
+
+	/*
+	 * This is a 256-bit bit mask provided by the PF driver for letting the
+	 * HWRM know what commands issued by the VF driver to the HWRM should be
+	 * forwarded to the PF driver. Nth bit refers to the Nth req_type.
+	 * Setting Nth bit to 1 indicates that requests from the VF driver with
+	 * req_type equal to N shall be forwarded to the parent PF driver. This
+	 * field is not valid for the VF driver.
+	 */
+	uint32_t vf_req_fwd[8];
+
+	/*
+	 * This is a 256-bit bit mask provided by the function driver (PF or VF
+	 * driver) to indicate the list of asynchronous event completions to be
+	 * forwarded. Nth bit refers to the Nth event_id. Setting Nth bit to 1
+	 * by the function driver shall result in the HWRM forwarding
+	 * asynchronous event completion with event_id equal to N. If all bits
+	 * are set to 0 (value of 0), then the HWRM shall not forward any
+	 * asynchronous event completion to this function driver.
+	 */
+	uint32_t async_event_fwd[8];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_func_drv_rgtr_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_func_drv_unrgtr */
+/*
+ * Description: This command is used by the function driver to un register with
+ * the HWRM. A function driver shall implement this command. A function driver
+ * shall use this command during the driver unloading.
+ */
+/* Input (24 bytes) */
+
+struct hwrm_func_drv_unrgtr_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the function driver is notifying the HWRM to
+	 * prepare for the shutdown.
+	 */
+	#define HWRM_FUNC_DRV_UNRGTR_INPUT_FLAGS_PREPARE_FOR_SHUTDOWN \
+							UINT32_C(0x1)
+	uint32_t flags;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_func_drv_unrgtr_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 #endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 04/38] bnxt: add dev infos get operation
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (2 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 03/38] bnxt: add driver register/unregister support Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 05/38] bnxt: add dev configure operation Stephen Hurd
                       ` (34 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Gets device info from the bp structure filled in the init() function.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder footer
---
 drivers/net/bnxt/bnxt.h        |  3 ++
 drivers/net/bnxt/bnxt_ethdev.c | 95 ++++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.c   |  2 +-
 3 files changed, 99 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index ed057ef..f8707b2 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -42,6 +42,9 @@
 #include <rte_lcore.h>
 #include <rte_spinlock.h>
 
+#define BNXT_MAX_MTU		9000
+#define VLAN_TAG_SIZE		4
+
 struct bnxt_vf_info {
 	uint16_t		fw_fid;
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 26e6447..a8a9912 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -61,10 +61,105 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 }
 
 /*
+ * Device configuration and status function
+ */
+
+static void bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
+				  struct rte_eth_dev_info *dev_info)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	uint16_t max_vnics, i, j, vpool, vrxq;
+
+	/* MAC Specifics */
+	dev_info->max_mac_addrs = MAX_NUM_MAC_ADDR;
+	dev_info->max_hash_mac_addrs = 0;
+
+	/* PF/VF specifics */
+	if (BNXT_PF(bp)) {
+		dev_info->max_rx_queues = bp->pf.max_rx_rings;
+		dev_info->max_tx_queues = bp->pf.max_tx_rings;
+		dev_info->max_vfs = bp->pf.active_vfs;
+		dev_info->reta_size = bp->pf.max_rsscos_ctx;
+		max_vnics = bp->pf.max_vnics;
+	} else {
+		dev_info->max_rx_queues = bp->vf.max_rx_rings;
+		dev_info->max_tx_queues = bp->vf.max_tx_rings;
+		dev_info->reta_size = bp->vf.max_rsscos_ctx;
+		max_vnics = bp->vf.max_vnics;
+	}
+
+	/* Fast path specifics */
+	dev_info->min_rx_bufsize = 1;
+	dev_info->max_rx_pktlen = BNXT_MAX_MTU + ETHER_HDR_LEN + ETHER_CRC_LEN
+				  + VLAN_TAG_SIZE;
+	dev_info->rx_offload_capa = 0;
+	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_IPV4_CKSUM |
+					DEV_TX_OFFLOAD_TCP_CKSUM |
+					DEV_TX_OFFLOAD_UDP_CKSUM |
+					DEV_TX_OFFLOAD_TCP_TSO;
+
+	/* *INDENT-OFF* */
+	dev_info->default_rxconf = (struct rte_eth_rxconf) {
+		.rx_thresh = {
+			.pthresh = 8,
+			.hthresh = 8,
+			.wthresh = 0,
+		},
+		.rx_free_thresh = 32,
+		.rx_drop_en = 0,
+	};
+
+	dev_info->default_txconf = (struct rte_eth_txconf) {
+		.tx_thresh = {
+			.pthresh = 32,
+			.hthresh = 0,
+			.wthresh = 0,
+		},
+		.tx_free_thresh = 32,
+		.tx_rs_thresh = 32,
+		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
+			     ETH_TXQ_FLAGS_NOOFFLOADS,
+	};
+	/* *INDENT-ON* */
+
+	/*
+	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
+	 *       need further investigation.
+	 */
+
+	/* VMDq resources */
+	vpool = 64; /* ETH_64_POOLS */
+	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
+	for (i = 0; i < 4; vpool >>= 1, i++) {
+		if (max_vnics > vpool) {
+			for (j = 0; j < 5; vrxq >>= 1, j++) {
+				if (dev_info->max_rx_queues > vrxq) {
+					if (vpool > vrxq)
+						vpool = vrxq;
+					goto found;
+				}
+			}
+			/* Not enough resources to support VMDq */
+			break;
+		}
+	}
+	/* Not enough resources to support VMDq */
+	vpool = 0;
+	vrxq = 0;
+found:
+	dev_info->max_vmdq_pools = vpool;
+	dev_info->vmdq_queue_num = vrxq;
+
+	dev_info->vmdq_pool_base = 0;
+	dev_info->vmdq_queue_base = 0;
+}
+
+/*
  * Initialization
  */
 
 static struct eth_dev_ops bnxt_dev_ops = {
+	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 };
 
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 8aba8cd..3eabf0c 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -83,7 +83,7 @@ static int bnxt_hwrm_send_message_locked(struct bnxt *bp, void *msg,
 		/* Sanity check on the resp->resp_len */
 		rte_rmb();
 		if (resp->resp_len && resp->resp_len <=
-		    bp->max_resp_len) {
+				bp->max_resp_len) {
 			/* Last byte of resp contains the valid key */
 			valid = (uint8_t *)resp + resp->resp_len - 1;
 			if (*valid == HWRM_RESP_VALID_KEY)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 05/38] bnxt: add dev configure operation
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (3 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 04/38] bnxt: add dev infos get operation Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 06/38] bnxt: add vnic functions and structs Stephen Hurd
                       ` (33 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

The dev_configure_op function calls bnxt_set_hwrm_link_config() to
setup the PHY.  This calls the new  bnxt_parse_eth_link_*() functions
to translate from the DPDK macro values to those used by HWRM calls,
then calls bnxt_hwrm_port_phy_cfg() to issue the HWRM call.

New HWRM calls:
bnxt_hwrm_port_phy_cfg:
	This command configures the PHY device for the port.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fixed few issues identified by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt.h                |  32 +++
 drivers/net/bnxt/bnxt_ethdev.c         |  24 ++
 drivers/net/bnxt/bnxt_hwrm.c           | 230 ++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 470 +++++++++++++++++++++++++++++++++
 5 files changed, 757 insertions(+)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index f8707b2..bfce91e 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -81,6 +81,29 @@ struct bnxt_pf_info {
 	struct bnxt_vf_info	*vf;
 };
 
+/* Max wait time is 10 * 100ms = 1s */
+#define BNXT_LINK_WAIT_CNT	10
+#define BNXT_LINK_WAIT_INTERVAL	100
+struct bnxt_link_info {
+	uint8_t			phy_flags;
+	uint8_t			mac_type;
+	uint8_t			phy_link_status;
+	uint8_t			loop_back;
+	uint8_t			link_up;
+	uint8_t			duplex;
+	uint8_t			pause;
+	uint8_t			force_pause;
+	uint8_t			auto_pause;
+	uint8_t			auto_mode;
+#define PHY_VER_LEN		3
+	uint8_t			phy_ver[PHY_VER_LEN];
+	uint16_t		link_speed;
+	uint16_t		support_speeds;
+	uint16_t		auto_link_speed;
+	uint16_t		auto_link_speed_mask;
+	uint32_t		preemphasis;
+};
+
 #define BNXT_COS_QUEUE_COUNT	8
 struct bnxt_cos_queue_info {
 	uint8_t	id;
@@ -99,6 +122,14 @@ struct bnxt {
 #define BNXT_PF(bp)		(!((bp)->flags & BNXT_FLAG_VF))
 #define BNXT_VF(bp)		((bp)->flags & BNXT_FLAG_VF)
 
+	unsigned int		rx_nr_rings;
+	unsigned int		rx_cp_nr_rings;
+	struct bnxt_rx_queue **rx_queues;
+
+	unsigned int		tx_nr_rings;
+	unsigned int		tx_cp_nr_rings;
+	struct bnxt_tx_queue **tx_queues;
+
 #define MAX_NUM_MAC_ADDR	32
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
 
@@ -109,6 +140,7 @@ struct bnxt {
 	uint16_t			max_req_len;
 	uint16_t			max_resp_len;
 
+	struct bnxt_link_info	link_info;
 	struct bnxt_cos_queue_info	cos_queue[BNXT_COS_QUEUE_COUNT];
 
 	struct bnxt_pf_info		pf;
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index a8a9912..b46d2ce 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -154,6 +154,29 @@ found:
 	dev_info->vmdq_queue_base = 0;
 }
 
+/* Configure the device based on the configuration provided */
+static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	int rc;
+
+	bp->rx_queues = (void *)eth_dev->data->rx_queues;
+	bp->tx_queues = (void *)eth_dev->data->tx_queues;
+
+	/* Inherit new configurations */
+	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
+	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
+	bp->rx_cp_nr_rings = bp->rx_nr_rings;
+	bp->tx_cp_nr_rings = bp->tx_nr_rings;
+
+	if (eth_dev->data->dev_conf.rxmode.jumbo_frame)
+		eth_dev->data->mtu =
+				eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
+				ETHER_HDR_LEN - ETHER_CRC_LEN - VLAN_TAG_SIZE;
+	rc = bnxt_set_hwrm_link_config(bp, true);
+	return rc;
+}
+
 /*
  * Initialization
  */
@@ -161,6 +184,7 @@ found:
 static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
+	.dev_configure = bnxt_dev_configure_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 3eabf0c..a2d7815 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -314,6 +314,61 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags)
 	return rc;
 }
 
+static int bnxt_hwrm_port_phy_cfg(struct bnxt *bp, struct bnxt_link_info *conf)
+{
+	int rc = 0;
+	struct hwrm_port_phy_cfg_input req = {.req_type = 0};
+	struct hwrm_port_phy_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, PORT_PHY_CFG, -1, resp);
+
+	req.flags = conf->phy_flags;
+	if (conf->link_up) {
+		req.force_link_speed = conf->link_speed;
+		/*
+		 * Note, ChiMP FW 20.2.1 and 20.2.2 return an error when we set
+		 * any auto mode, even "none".
+		 */
+		if (req.auto_mode == HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE) {
+			req.flags |= HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE;
+		} else {
+			req.auto_mode = conf->auto_mode;
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_MODE;
+			req.auto_link_speed_mask = conf->auto_link_speed_mask;
+			req.enables |=
+			   HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED_MASK;
+			req.auto_link_speed = conf->auto_link_speed;
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED;
+		}
+		req.auto_duplex = conf->duplex;
+		req.enables |= HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_DUPLEX;
+		req.auto_pause = conf->auto_pause;
+		/* Set force_pause if there is no auto or if there is a force */
+		if (req.auto_pause)
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_PAUSE;
+		else
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE;
+		req.force_pause = conf->force_pause;
+		if (req.force_pause)
+			req.enables |=
+				HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE;
+	} else {
+		req.flags &= ~HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG;
+		req.flags |= HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE_LINK_DOWN;
+		req.force_link_speed = 0;
+	}
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
@@ -372,3 +427,178 @@ int bnxt_alloc_hwrm_resources(struct bnxt *bp)
 
 	return 0;
 }
+
+static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
+{
+	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
+
+	if ((conf_link_speed & ETH_LINK_SPEED_FIXED) == ETH_LINK_SPEED_AUTONEG)
+		return HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
+
+	switch (conf_link_speed) {
+	case ETH_LINK_SPEED_10M_HD:
+	case ETH_LINK_SPEED_100M_HD:
+		return HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF;
+	}
+	return hw_link_duplex;
+}
+
+static uint16_t bnxt_parse_eth_link_speed(uint32_t conf_link_speed)
+{
+	uint16_t eth_link_speed = 0;
+
+	if ((conf_link_speed & ETH_LINK_SPEED_FIXED) == ETH_LINK_SPEED_AUTONEG)
+		return ETH_LINK_SPEED_AUTONEG;
+
+	switch (conf_link_speed & ~ETH_LINK_SPEED_FIXED) {
+	case ETH_LINK_SPEED_100M:
+	case ETH_LINK_SPEED_100M_HD:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MB;
+		break;
+	case ETH_LINK_SPEED_1G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB;
+		break;
+	case ETH_LINK_SPEED_2_5G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB;
+		break;
+	case ETH_LINK_SPEED_10G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB;
+		break;
+	case ETH_LINK_SPEED_20G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB;
+		break;
+	case ETH_LINK_SPEED_25G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB;
+		break;
+	case ETH_LINK_SPEED_40G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB;
+		break;
+	case ETH_LINK_SPEED_50G:
+		eth_link_speed =
+			HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB;
+		break;
+	default:
+		RTE_LOG(ERR, PMD,
+			"Unsupported link speed %d; default to AUTO\n",
+			conf_link_speed);
+		break;
+	}
+	return eth_link_speed;
+}
+
+#define BNXT_SUPPORTED_SPEEDS (ETH_LINK_SPEED_100M | ETH_LINK_SPEED_100M_HD | \
+		ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G | \
+		ETH_LINK_SPEED_10G | ETH_LINK_SPEED_20G | ETH_LINK_SPEED_25G | \
+		ETH_LINK_SPEED_40G | ETH_LINK_SPEED_50G)
+
+static int bnxt_valid_link_speed(uint32_t link_speed, uint8_t port_id)
+{
+	uint32_t one_speed;
+
+	if (link_speed == ETH_LINK_SPEED_AUTONEG)
+		return 0;
+
+	if (link_speed & ETH_LINK_SPEED_FIXED) {
+		one_speed = link_speed & ~ETH_LINK_SPEED_FIXED;
+
+		if (one_speed & (one_speed - 1)) {
+			RTE_LOG(ERR, PMD,
+				"Invalid advertised speeds (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+		if ((one_speed & BNXT_SUPPORTED_SPEEDS) != one_speed) {
+			RTE_LOG(ERR, PMD,
+				"Unsupported advertised speed (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+	} else {
+		if (!(link_speed & BNXT_SUPPORTED_SPEEDS)) {
+			RTE_LOG(ERR, PMD,
+				"Unsupported advertised speeds (%u) for port %u\n",
+				link_speed, port_id);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
+static uint16_t bnxt_parse_eth_link_speed_mask(uint32_t link_speed)
+{
+	uint16_t ret = 0;
+
+	if (link_speed == ETH_LINK_SPEED_AUTONEG)
+		link_speed = BNXT_SUPPORTED_SPEEDS;
+
+	if (link_speed & ETH_LINK_SPEED_100M)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB;
+	if (link_speed & ETH_LINK_SPEED_100M_HD)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB;
+	if (link_speed & ETH_LINK_SPEED_1G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB;
+	if (link_speed & ETH_LINK_SPEED_2_5G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB;
+	if (link_speed & ETH_LINK_SPEED_10G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB;
+	if (link_speed & ETH_LINK_SPEED_20G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB;
+	if (link_speed & ETH_LINK_SPEED_25G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB;
+	if (link_speed & ETH_LINK_SPEED_40G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB;
+	if (link_speed & ETH_LINK_SPEED_50G)
+		ret |= HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB;
+	return ret;
+}
+
+int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up)
+{
+	int rc = 0;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_link_info link_req;
+	uint16_t speed;
+
+	rc = bnxt_valid_link_speed(dev_conf->link_speeds,
+			bp->eth_dev->data->port_id);
+	if (rc)
+		goto error;
+
+	memset(&link_req, 0, sizeof(link_req));
+	speed = bnxt_parse_eth_link_speed(dev_conf->link_speeds);
+	link_req.link_up = link_up;
+	if (speed == 0) {
+		link_req.phy_flags =
+				HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG;
+		link_req.auto_mode =
+				HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_OR_BELOW;
+		link_req.auto_link_speed_mask =
+			bnxt_parse_eth_link_speed_mask(dev_conf->link_speeds);
+		link_req.auto_link_speed =
+				HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_50GB;
+	} else {
+		link_req.auto_mode = HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE;
+		link_req.phy_flags = HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE |
+			HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESET_PHY;
+		link_req.link_speed = speed;
+	}
+	link_req.duplex = bnxt_parse_eth_link_duplex(dev_conf->link_speeds);
+	link_req.auto_pause = bp->link_info.auto_pause;
+	link_req.force_pause = bp->link_info.force_pause;
+
+	rc = bnxt_hwrm_port_phy_cfg(bp, &link_req);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Set link config failed with rc %d\n", rc);
+	}
+
+error:
+	return rc;
+}
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 6f2e445..eef3be6 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -52,5 +52,6 @@ int bnxt_hwrm_ver_get(struct bnxt *bp);
 
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
 
 #endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 8d816f4..cd2b14e 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -54,6 +54,7 @@
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
+#define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 
 /*
@@ -345,6 +346,475 @@ struct hwrm_func_qcaps_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_port_phy_cfg */
+/*
+ * Description: This command configures the PHY device for the port. It allows
+ * setting of the most generic settings for the PHY. The HWRM shall complete
+ * this command as soon as PHY settings are configured. They may not be applied
+ * when the command response is provided. A VF driver shall not be allowed to
+ * configure PHY using this command. In a network partition mode, a PF driver
+ * shall not be allowed to configure PHY using this command.
+ */
+
+/* Input (56 bytes) */
+struct hwrm_port_phy_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is set to '1', the PHY for the port shall be reset. #
+	 * If this bit is set to 1, then the HWRM shall reset the PHY after
+	 * applying PHY configuration changes specified in this command. # In
+	 * order to guarantee that PHY configuration changes specified in this
+	 * command take effect, the HWRM client should set this flag to 1. # If
+	 * this bit is not set to 1, then the HWRM may reset the PHY depending
+	 * on the current PHY configuration and settings specified in this
+	 * command.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESET_PHY            UINT32_C(0x1)
+	/*
+	 * When this bit is set to '1', the link shall be forced to be taken
+	 * down. # When this bit is set to '1", all other command input settings
+	 * related to the link speed shall be ignored. Once the link state is
+	 * forced down, it can be explicitly cleared from that state by setting
+	 * this flag to '0'. # If this flag is set to '0', then the link shall
+	 * be cleared from forced down state if the link is in forced down
+	 * state. There may be conditions (e.g. out-of-band or sideband
+	 * configuration changes for the link) outside the scope of the HWRM
+	 * implementation that may clear forced down link state.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE_LINK_DOWN      UINT32_C(0x2)
+	/*
+	 * When this bit is set to '1', the link shall be forced to the
+	 * force_link_speed value. When this bit is set to '1', the HWRM client
+	 * should not enable any of the auto negotiation related fields
+	 * represented by auto_XXX fields in this command. When this bit is set
+	 * to '1' and the HWRM client has enabled a auto_XXX field in this
+	 * command, then the HWRM shall ignore the enabled auto_XXX field. When
+	 * this bit is set to zero, the link shall be allowed to autoneg.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_FORCE                UINT32_C(0x4)
+	/*
+	 * When this bit is set to '1', the auto-negotiation process shall be
+	 * restarted on the link.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_RESTART_AUTONEG      UINT32_C(0x8)
+	/*
+	 * When this bit is set to '1', Energy Efficient Ethernet (EEE) is
+	 * requested to be enabled on this link. If EEE is not supported on this
+	 * port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_ENABLE	UINT32_C(0x10)
+	/*
+	 * When this bit is set to '1', Energy Efficient Ethernet (EEE) is
+	 * requested to be disabled on this link. If EEE is not supported on
+	 * this port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_DISABLE	UINT32_C(0x20)
+	/*
+	 * When this bit is set to '1' and EEE is enabled on this link, then TX
+	 * LPI is requested to be enabled on the link. If EEE is not supported
+	 * on this port, then this flag shall be ignored by the HWRM. If EEE is
+	 * disabled on this port, then this flag shall be ignored by the HWRM.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FLAGS_EEE_TX_LPI	UINT32_C(0x40)
+	uint32_t flags;
+
+	/* This bit must be '1' for the auto_mode field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_MODE          UINT32_C(0x1)
+	/* This bit must be '1' for the auto_duplex field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_DUPLEX        UINT32_C(0x2)
+	/* This bit must be '1' for the auto_pause field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_PAUSE         UINT32_C(0x4)
+	/*
+	 * This bit must be '1' for the auto_link_speed field to be configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED    UINT32_C(0x8)
+	/*
+	 * This bit must be '1' for the auto_link_speed_mask field to be
+	 * configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_AUTO_LINK_SPEED_MASK \
+								UINT32_C(0x10)
+	/* This bit must be '1' for the wirespeed field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_WIRESPEED	UINT32_C(0x20)
+	/* This bit must be '1' for the lpbk field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_LPBK		UINT32_C(0x40)
+	/* This bit must be '1' for the preemphasis field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_PREEMPHASIS	UINT32_C(0x80)
+	/* This bit must be '1' for the force_pause field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_FORCE_PAUSE	UINT32_C(0x100)
+	/*
+	 * This bit must be '1' for the eee_link_speed_mask field to be
+	 * configured.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_EEE_LINK_SPEED_MASK \
+								UINT32_C(0x200)
+	/* This bit must be '1' for the tx_lpi_timer field to be configured. */
+	#define HWRM_PORT_PHY_CFG_INPUT_ENABLES_TX_LPI_TIMER	UINT32_C(0x400)
+	uint32_t enables;
+
+	/* Port ID of port that is to be configured. */
+	uint16_t port_id;
+
+	/*
+	 * This is the speed that will be used if the force bit is '1'. If
+	 * unsupported speed is selected, an error will be generated.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t force_link_speed;
+
+	/*
+	 * This value is used to identify what autoneg mode is used when the
+	 * link speed is not being forced.
+	 */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_NONE	(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ALL_SPEEDS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_MODE_SPEED_MASK \
+							(UINT32_C(0x4) << 0)
+	uint8_t auto_mode;
+
+	/*
+	 * This is the duplex setting that will be used if the autoneg_mode is
+	 * "one_speed" or "one_or_below".
+	 */
+		/* Half Duplex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF \
+							(UINT32_C(0x0) << 0)
+		/* Full duplex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_FULL \
+							(UINT32_C(0x1) << 0)
+		/* Both Half and Full dupex will be requested. */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH \
+							(UINT32_C(0x2) << 0)
+	uint8_t auto_duplex;
+
+	/*
+	 * This value is used to configure the pause that will be used for
+	 * autonegotiation. Add text on the usage of auto_pause and force_pause.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX              UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX              UINT32_C(0x2)
+	/*
+	 * When set to 1, the advertisement of pause is enabled. # When the
+	 * auto_mode is not set to none and this flag is set to 1, then the
+	 * auto_pause bits on this port are being advertised and autoneg pause
+	 * results are being interpreted. # When the auto_mode is not set to
+	 * none and this flag is set to 0, the pause is forced as indicated in
+	 * force_pause, and also advertised as auto_pause bits, but the autoneg
+	 * results are not interpreted since the pause configuration is being
+	 * forced. # When the auto_mode is set to none and this flag is set to
+	 * 1, auto_pause bits should be ignored and should be set to 0.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_AUTONEG_PAUSE   UINT32_C(0x4)
+	uint8_t auto_pause;
+
+	uint8_t unused_0;
+
+	/*
+	 * This is the speed that will be used if the autoneg_mode is
+	 * "one_speed" or "one_or_below". If an unsupported speed is selected,
+	 * an error will be generated.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t auto_link_speed;
+
+	/*
+	 * This is a mask of link speeds that will be used if autoneg_mode is
+	 * "mask". If unsupported speed is enabled an error will be generated.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_AUTO_LINK_SPEED_MASK_10MB \
+							UINT32_C(0x2000)
+	uint16_t auto_link_speed_mask;
+
+	/* This value controls the wirespeed feature. */
+		/* Wirespeed feature is disabled. */
+	#define HWRM_PORT_PHY_CFG_INPUT_WIRESPEED_OFF	(UINT32_C(0x0) << 0)
+		/* Wirespeed feature is enabled. */
+	#define HWRM_PORT_PHY_CFG_INPUT_WIRESPEED_ON	(UINT32_C(0x1) << 0)
+	uint8_t wirespeed;
+
+	/* This value controls the loopback setting for the PHY. */
+		/* No loopback is selected. Normal operation. */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_NONE	(UINT32_C(0x0) << 0)
+		/*
+		 * The HW will be configured with local loopback such that host
+		 * data is sent back to the host without modification.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_LOCAL	(UINT32_C(0x1) << 0)
+		/*
+		 * The HW will be configured with remote loopback such that port
+		 * logic will send packets back out the transmitter that are
+		 * received.
+		 */
+	#define HWRM_PORT_PHY_CFG_INPUT_LPBK_REMOTE	(UINT32_C(0x2) << 0)
+	uint8_t lpbk;
+
+	/*
+	 * This value is used to configure the pause that will be used for force
+	 * mode.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX             UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX             UINT32_C(0x2)
+	uint8_t force_pause;
+
+	uint8_t unused_1;
+
+	/*
+	 * This value controls the pre-emphasis to be used for the link. Driver
+	 * should not set this value (use enable.preemphasis = 0) unless driver
+	 * is sure of setting. Normally HWRM FW will determine proper pre-
+	 * emphasis.
+	 */
+	uint32_t preemphasis;
+
+	/*
+	 * Setting for link speed mask that is used to advertise speeds during
+	 * autonegotiation when EEE is enabled. This field is valid only when
+	 * EEE is enabled. The speeds specified in this field shall be a subset
+	 * of speeds specified in auto_link_speed_mask. If EEE is enabled,then
+	 * at least one speed shall be provided in this mask.
+	 */
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD1  UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_100MB  UINT32_C(0x2)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD2  UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_1GB    UINT32_C(0x8)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD3 \
+								UINT32_C(0x10)
+	/* Reserved */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_RSVD4 \
+								UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_CFG_INPUT_EEE_LINK_SPEED_MASK_10GB \
+								UINT32_C(0x40)
+	uint16_t eee_link_speed_mask;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * Reuested setting of TX LPI timer in microseconds. This field is valid
+	 * only when EEE is enabled and TX LPI is enabled.
+	 */
+	#define HWRM_PORT_PHY_CFG_INPUT_TX_LPI_TIMER_MASK \
+							UINT32_C(0xffffff)
+	#define HWRM_PORT_PHY_CFG_INPUT_TX_LPI_TIMER_SFT           0
+	uint32_t tx_lpi_timer;
+
+	uint32_t unused_4;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_port_phy_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_ver_get */
 /*
  * Description: This function is called by a driver to determine the HWRM
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 06/38] bnxt: add vnic functions and structs
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (4 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 05/38] bnxt: add dev configure operation Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 07/38] bnxt: declare generic ring structs and free() func Stephen Hurd
                       ` (32 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add functions to allocate, initialize, and free vnics.

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |  14 ++
 drivers/net/bnxt/bnxt_vnic.c           | 277 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_vnic.h           |  80 ++++++++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h |   3 +
 5 files changed, 375 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_vnic.c
 create mode 100644 drivers/net/bnxt/bnxt_vnic.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 9965597..c57afaa 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
 # Export include files
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index bfce91e..d0f84f4 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -45,6 +45,13 @@
 #define BNXT_MAX_MTU		9000
 #define VLAN_TAG_SIZE		4
 
+enum bnxt_hw_context {
+	HW_CONTEXT_NONE     = 0,
+	HW_CONTEXT_IS_RSS   = 1,
+	HW_CONTEXT_IS_COS   = 2,
+	HW_CONTEXT_IS_LB    = 3,
+};
+
 struct bnxt_vf_info {
 	uint16_t		fw_fid;
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
@@ -130,6 +137,13 @@ struct bnxt {
 	unsigned int		tx_cp_nr_rings;
 	struct bnxt_tx_queue **tx_queues;
 
+	struct bnxt_vnic_info	*vnic_info;
+	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
+
+	/* VNIC pointer for flow filter (VMDq) pools */
+#define MAX_FF_POOLS	ETH_64_POOLS
+	STAILQ_HEAD(, bnxt_vnic_info)	ff_pool[MAX_FF_POOLS];
+
 #define MAX_NUM_MAC_ADDR	32
 	uint8_t			mac_addr[ETHER_ADDR_LEN];
 
diff --git a/drivers/net/bnxt/bnxt_vnic.c b/drivers/net/bnxt/bnxt_vnic.c
new file mode 100644
index 0000000..c04c4c7
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_vnic.c
@@ -0,0 +1,277 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_memzone.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * VNIC Functions
+ */
+
+static void prandom_bytes(void *dest_ptr, size_t len)
+{
+	char *dest = (char *)dest_ptr;
+	uint64_t rb;
+
+	while (len) {
+		rb = rte_rand();
+		if (len >= 8) {
+			memcpy(dest, &rb, 8);
+			len -= 8;
+			dest += 8;
+		} else {
+			memcpy(dest, &rb, len);
+			dest += len;
+			len = 0;
+		}
+	}
+}
+
+void bnxt_init_vnics(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	uint16_t max_vnics;
+	int i, j;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	STAILQ_INIT(&bp->free_vnic_list);
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+		vnic->fw_vnic_id = (uint16_t)HWRM_NA_SIGNATURE;
+		vnic->fw_rss_cos_lb_ctx = (uint16_t)HWRM_NA_SIGNATURE;
+		vnic->ctx_is_rss_cos_lb = HW_CONTEXT_NONE;
+
+		for (j = 0; j < MAX_QUEUES_PER_VNIC; j++)
+			vnic->fw_grp_ids[j] = (uint16_t)HWRM_NA_SIGNATURE;
+
+		prandom_bytes(vnic->rss_hash_key, HW_HASH_KEY_SIZE);
+		STAILQ_INIT(&vnic->filter);
+		STAILQ_INSERT_TAIL(&bp->free_vnic_list, vnic, next);
+	}
+	for (i = 0; i < MAX_FF_POOLS; i++)
+		STAILQ_INIT(&bp->ff_pool[i]);
+}
+
+int bnxt_free_vnic(struct bnxt *bp, struct bnxt_vnic_info *vnic,
+			  int pool)
+{
+	struct bnxt_vnic_info *temp;
+
+	temp = STAILQ_FIRST(&bp->ff_pool[pool]);
+	while (temp) {
+		if (temp == vnic) {
+			STAILQ_REMOVE(&bp->ff_pool[pool], vnic,
+				      bnxt_vnic_info, next);
+			vnic->fw_vnic_id = (uint16_t)HWRM_NA_SIGNATURE;
+			STAILQ_INSERT_TAIL(&bp->free_vnic_list, vnic,
+					   next);
+			return 0;
+		}
+		temp = STAILQ_NEXT(temp, next);
+	}
+	RTE_LOG(ERR, PMD, "VNIC %p is not found in pool[%d]\n", vnic, pool);
+	return -EINVAL;
+}
+
+struct bnxt_vnic_info *bnxt_alloc_vnic(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+
+	/* Find the 1st unused vnic from the free_vnic_list pool*/
+	vnic = STAILQ_FIRST(&bp->free_vnic_list);
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "No more free VNIC resources\n");
+		return NULL;
+	}
+	STAILQ_REMOVE_HEAD(&bp->free_vnic_list, next);
+	return vnic;
+}
+
+void bnxt_free_all_vnics(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *temp, *next;
+	int i;
+
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		temp = STAILQ_FIRST(&bp->ff_pool[i]);
+		while (temp) {
+			next = STAILQ_NEXT(temp, next);
+			STAILQ_REMOVE(&bp->ff_pool[i], temp, bnxt_vnic_info,
+				      next);
+			STAILQ_INSERT_TAIL(&bp->free_vnic_list, temp, next);
+			temp = next;
+		}
+	}
+}
+
+void bnxt_free_vnic_attributes(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+
+	STAILQ_FOREACH(vnic, &bp->free_vnic_list, next) {
+		if (vnic->rss_table) {
+			/* 'Unreserve' the rss_table */
+			/* N/A */
+
+			vnic->rss_table = NULL;
+		}
+
+		if (vnic->rss_hash_key) {
+			/* 'Unreserve' the rss_hash_key */
+			/* N/A */
+
+			vnic->rss_hash_key = NULL;
+		}
+	}
+}
+
+int bnxt_alloc_vnic_attributes(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	struct rte_pci_device *pdev = bp->pdev;
+	const struct rte_memzone *mz;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+	int entry_length = RTE_CACHE_LINE_ROUNDUP(
+				HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table) +
+				HW_HASH_KEY_SIZE);
+	uint16_t max_vnics;
+	int i;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
+		 "bnxt_%04x:%02x:%02x:%02x_vnicattr", pdev->addr.domain,
+		 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
+	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
+	mz = rte_memzone_lookup(mz_name);
+	if (!mz) {
+		mz = rte_memzone_reserve(mz_name,
+					 entry_length * max_vnics,
+					 SOCKET_ID_ANY,
+					 RTE_MEMZONE_2MB |
+					 RTE_MEMZONE_SIZE_HINT_ONLY);
+		if (!mz)
+			return -ENOMEM;
+	}
+
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+
+		/* Allocate rss table and hash key */
+		vnic->rss_table =
+			(void *)((char *)mz->addr + (entry_length * i));
+		memset(vnic->rss_table, -1, entry_length);
+
+		vnic->rss_table_dma_addr = mz->phys_addr + (entry_length * i);
+		vnic->rss_hash_key = (void *)((char *)vnic->rss_table +
+			     HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table));
+
+		vnic->rss_hash_key_dma_addr = vnic->rss_table_dma_addr +
+			     HW_HASH_INDEX_SIZE * sizeof(*vnic->rss_table);
+	}
+
+	return 0;
+}
+
+void bnxt_free_vnic_mem(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	uint16_t max_vnics, i;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	for (i = 0; i < max_vnics; i++) {
+		vnic = &bp->vnic_info[i];
+		if (vnic->fw_vnic_id != (uint16_t)HWRM_NA_SIGNATURE) {
+			RTE_LOG(ERR, PMD, "VNIC is not freed yet!\n");
+			/* TODO Call HWRM to free VNIC */
+		}
+	}
+
+	rte_free(bp->vnic_info);
+	bp->vnic_info = NULL;
+}
+
+int bnxt_alloc_vnic_mem(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic_mem;
+	uint16_t max_vnics;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_vnics = pf->max_vnics;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_vnics = vf->max_vnics;
+	}
+	/* Allocate memory for VNIC pool and filter pool */
+	vnic_mem = rte_zmalloc("bnxt_vnic_info",
+			       max_vnics * sizeof(struct bnxt_vnic_info), 0);
+	if (vnic_mem == NULL) {
+		RTE_LOG(ERR, PMD, "Failed to alloc memory for %d VNICs",
+			max_vnics);
+		return -ENOMEM;
+	}
+	bp->vnic_info = vnic_mem;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_vnic.h b/drivers/net/bnxt/bnxt_vnic.h
new file mode 100644
index 0000000..9671ba4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_vnic.h
@@ -0,0 +1,80 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_VNIC_H_
+#define _BNXT_VNIC_H_
+
+#include <sys/queue.h>
+#include <stdbool.h>
+
+struct bnxt_vnic_info {
+	STAILQ_ENTRY(bnxt_vnic_info)	next;
+	uint8_t		ff_pool_idx;
+
+	uint16_t	fw_vnic_id; /* returned by Chimp during alloc */
+	uint16_t	fw_rss_cos_lb_ctx;
+	uint16_t	ctx_is_rss_cos_lb;
+#define MAX_NUM_TRAFFIC_CLASSES		8
+#define MAX_NUM_RSS_QUEUES_PER_VNIC	16
+#define MAX_QUEUES_PER_VNIC	(MAX_NUM_RSS_QUEUES_PER_VNIC + \
+				 MAX_NUM_TRAFFIC_CLASSES)
+	uint16_t	start_grp_id;
+	uint16_t	end_grp_id;
+	uint16_t	fw_grp_ids[MAX_QUEUES_PER_VNIC];
+	uint16_t	hash_type;
+	phys_addr_t	rss_table_dma_addr;
+	uint16_t	*rss_table;
+	phys_addr_t	rss_hash_key_dma_addr;
+	void		*rss_hash_key;
+	uint32_t	flags;
+#define BNXT_VNIC_INFO_PROMISC			(1 << 0)
+#define BNXT_VNIC_INFO_ALLMULTI			(1 << 1)
+
+	bool		vlan_strip;
+	bool		func_default;
+
+	STAILQ_HEAD(, bnxt_filter_info)	filter;
+};
+
+struct bnxt;
+void bnxt_init_vnics(struct bnxt *bp);
+int bnxt_free_vnic(struct bnxt *bp, struct bnxt_vnic_info *vnic,
+			  int pool);
+struct bnxt_vnic_info *bnxt_alloc_vnic(struct bnxt *bp);
+void bnxt_free_all_vnics(struct bnxt *bp);
+void bnxt_free_vnic_attributes(struct bnxt *bp);
+int bnxt_alloc_vnic_attributes(struct bnxt *bp);
+void bnxt_free_vnic_mem(struct bnxt *bp);
+int bnxt_alloc_vnic_mem(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index cd2b14e..92e39ca 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -43,8 +43,11 @@
  * Following is the signature for HWRM message field that indicates not
  * applicable (All F's). Need to cast it the size of the field if needed.
  */
+#define HWRM_NA_SIGNATURE        ((uint32_t)(-1))
 #define HWRM_MAX_REQ_LEN	(128)  /* hwrm_func_buf_rgtr */
 #define HWRM_MAX_RESP_LEN	(176)  /* hwrm_func_qstats */
+#define HW_HASH_INDEX_SIZE      0x80    /* 7 bit indirection table index. */
+#define HW_HASH_KEY_SIZE        40
 #define HWRM_RESP_VALID_KEY	1 /* valid key for HWRM response */
 
 /*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 07/38] bnxt: declare generic ring structs and free() func
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (5 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 06/38] bnxt: add vnic functions and structs Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 08/38] bnxt: add completion ring support Stephen Hurd
                       ` (31 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Declare generic ring structures and a free() function. These are
generic ring management functions which will be used to create Tx,
Rx and Completion rings in the subsequent patches, and tie them to
the HWRM managed ring resources.

This generic ring structure is shared all the ring types and tracks
the the host Buffer Descriptors (BDs) and the HWRM assigned ID.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address checkpatch warnings.

v5:
Reorder footer

v6:
Correct typo
---
 drivers/net/bnxt/Makefile    |  1 +
 drivers/net/bnxt/bnxt_ring.c | 47 ++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ring.h | 92 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 140 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_ring.c
 create mode 100644 drivers/net/bnxt/bnxt_ring.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index c57afaa..757ea62 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
new file mode 100644
index 0000000..b57d3ba
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -0,0 +1,47 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "bnxt.h"
+#include "bnxt_ring.h"
+
+/*
+ * Generic ring handling
+ */
+
+void bnxt_free_ring(struct bnxt_ring *ring)
+{
+	if (ring->vmem_size && *ring->vmem) {
+		memset((char *)*ring->vmem, 0, ring->vmem_size);
+		*ring->vmem = NULL;
+	}
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
new file mode 100644
index 0000000..3abfb04
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -0,0 +1,92 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RING_H_
+#define _BNXT_RING_H_
+
+#include <inttypes.h>
+
+#include <rte_memory.h>
+
+#define RING_NEXT(ring, idx)		(((idx) + 1) & (ring)->ring_mask)
+
+#define RTE_MBUF_DATA_DMA_ADDR(mb) \
+	((uint64_t)((mb)->buf_physaddr + (mb)->data_off))
+
+#define DB_IDX_MASK						0xffffff
+#define DB_IDX_VALID						(0x1 << 26)
+#define DB_IRQ_DIS						(0x1 << 27)
+#define DB_KEY_TX						(0x0 << 28)
+#define DB_KEY_RX						(0x1 << 28)
+#define DB_KEY_CP						(0x2 << 28)
+#define DB_KEY_ST						(0x3 << 28)
+#define DB_KEY_TX_PUSH						(0x4 << 28)
+#define DB_LONG_TX_PUSH						(0x2 << 24)
+
+#define DEFAULT_CP_RING_SIZE	256
+#define DEFAULT_RX_RING_SIZE	256
+#define DEFAULT_TX_RING_SIZE	256
+
+#define MAX_TPA		128
+
+/* These assume 4k pages */
+#define MAX_RX_DESC_CNT (8 * 1024)
+#define MAX_TX_DESC_CNT (4 * 1024)
+#define MAX_CP_DESC_CNT (16 * 1024)
+
+#define INVALID_HW_RING_ID      ((uint16_t)-1)
+
+struct bnxt_ring {
+	void			*bd;
+	phys_addr_t		bd_dma;
+	uint32_t		ring_size;
+	uint32_t		ring_mask;
+
+	int			vmem_size;
+	void			**vmem;
+
+	uint16_t		fw_ring_id; /* Ring id filled by Chimp FW */
+};
+
+struct bnxt_ring_grp_info {
+	uint16_t	fw_stats_ctx;
+	uint16_t	fw_grp_id;
+	uint16_t	rx_fw_ring_id;
+	uint16_t	cp_fw_ring_id;
+	uint16_t	ag_fw_ring_id;
+};
+
+struct bnxt;
+void bnxt_free_ring(struct bnxt_ring *ring);
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 08/38] bnxt: add completion ring support
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (6 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 07/38] bnxt: declare generic ring structs and free() func Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 09/38] bnxt: add L2 filter alloc/init/free Stephen Hurd
                       ` (30 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Structures, macros, and functions for working with completion rings
in the driver.

Completion Ring is used by the Ethernet controller to provide the
status of transmitted & received packets, report errors, report
status changes to the host software, and inter-function forwarding
requests.  In addition to the generic ring features, a completion ring
can have a statistics context that has statistics periodically DMAed
to host memory, along with a consumer index.

bnxt_handle_async_event() handles completions not related to a specific
transmit or receive ring such as link status changes which arrive on
the default completion ring.

Other physical or virtual functions on the same device may send an HWRM
command forward request.  In this case, we will pass it through
unvalidated. In the future, we will be able to have the PF monitor and
control VF access to the HWRM interface if needed.

New HWRM Calls:
bnxt_hwrm_exec_fwd_resp:
	Execute an encapsulated command and forward the response.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Reorder footer
Expand patch description
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   5 +
 drivers/net/bnxt/bnxt_cpr.c            | 140 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_cpr.h            |  88 +++++++++++++
 drivers/net/bnxt/bnxt_hwrm.c           |  18 +++
 drivers/net/bnxt/bnxt_hwrm.h           |   2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 227 +++++++++++++++++++++++++++++++++
 7 files changed, 481 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_cpr.c
 create mode 100644 drivers/net/bnxt/bnxt_cpr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 757ea62..afd1690 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -48,6 +48,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 # all source are stored in SRCS-y
 #
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index d0f84f4..bdd355f 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -42,6 +42,8 @@
 #include <rte_lcore.h>
 #include <rte_spinlock.h>
 
+#include "bnxt_cpr.h"
+
 #define BNXT_MAX_MTU		9000
 #define VLAN_TAG_SIZE		4
 
@@ -137,6 +139,9 @@ struct bnxt {
 	unsigned int		tx_cp_nr_rings;
 	struct bnxt_tx_queue **tx_queues;
 
+	/* Default completion ring */
+	struct bnxt_cp_ring_info	*def_cp_ring;
+
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
new file mode 100644
index 0000000..7de2b9c
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -0,0 +1,140 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
+
+/*
+ * Async event handling
+ */
+void bnxt_handle_async_event(struct bnxt *bp __rte_unused,
+			     struct cmpl_base *cmp)
+{
+	struct hwrm_async_event_cmpl *async_cmp =
+				(struct hwrm_async_event_cmpl *)cmp;
+
+	/* TODO: HWRM async events are not defined yet */
+	/* Needs to handle: link events, error events, etc. */
+	switch (async_cmp->event_id) {
+	case 0:
+		/* Assume LINK_CHANGE == 0 */
+		RTE_LOG(INFO, PMD, "Link change event\n");
+
+		/* Can just prompt the update_op routine to do a qcfg
+		 * instead of doing the actual qcfg
+		 */
+		break;
+	case 1:
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "handle_async_event id = 0x%x\n",
+			async_cmp->event_id);
+		break;
+	}
+}
+
+void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmpl)
+{
+	struct hwrm_fwd_req_cmpl *fwd_cmpl = (struct hwrm_fwd_req_cmpl *)cmpl;
+	struct input *fwd_cmd;
+	uint16_t logical_vf_id, error_code;
+
+	/* Qualify the fwd request */
+	if (fwd_cmpl->source_id < bp->pf.first_vf_id) {
+		RTE_LOG(ERR, PMD,
+			"FWD req's source_id 0x%x > first_vf_id 0x%x\n",
+			fwd_cmpl->source_id, bp->pf.first_vf_id);
+		error_code = HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED;
+		goto reject;
+	} else if (fwd_cmpl->req_len_type >> HWRM_FWD_REQ_CMPL_REQ_LEN_SFT >
+		   128 - sizeof(struct input)) {
+		RTE_LOG(ERR, PMD,
+		    "FWD req's cmd len 0x%x > 108 bytes allowed\n",
+		    fwd_cmpl->req_len_type >> HWRM_FWD_REQ_CMPL_REQ_LEN_SFT);
+		error_code = HWRM_ERR_CODE_INVALID_PARAMS;
+		goto reject;
+	}
+
+	/* Locate VF's forwarded command */
+	logical_vf_id = fwd_cmpl->source_id - bp->pf.first_vf_id;
+	fwd_cmd = (struct input *)((uint8_t *)bp->pf.vf_req_buf +
+		   (logical_vf_id * 128));
+
+	/* Provision the request */
+	switch (fwd_cmd->req_type) {
+	case HWRM_CFA_L2_FILTER_ALLOC:
+	case HWRM_CFA_L2_FILTER_FREE:
+	case HWRM_CFA_L2_FILTER_CFG:
+	case HWRM_CFA_L2_SET_RX_MASK:
+		break;
+	default:
+		error_code = HWRM_ERR_CODE_INVALID_PARAMS;
+		goto reject;
+	}
+
+	/* Forward */
+	fwd_cmd->target_id = fwd_cmpl->source_id;
+	bnxt_hwrm_exec_fwd_resp(bp, fwd_cmd);
+	return;
+
+reject:
+	/* TODO: Encap the reject error resp into the hwrm_err_iput? */
+	/* Use the error_code for the reject cmd */
+	RTE_LOG(ERR, PMD,
+		"Error 0x%x found in the forward request\n", error_code);
+}
+
+/* For the default completion ring only */
+void bnxt_free_def_cp_ring(struct bnxt *bp)
+{
+	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+	struct bnxt_ring *ring = cpr->cp_ring_struct;
+
+	bnxt_free_ring(ring);
+}
+
+/* For the default completion ring only */
+void bnxt_init_def_ring_struct(struct bnxt *bp)
+{
+	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+	struct bnxt_ring *ring = cpr->cp_ring_struct;
+
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->ring_size = rte_align32pow2(DEFAULT_CP_RING_SIZE);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
new file mode 100644
index 0000000..ecabc5f
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -0,0 +1,88 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_CPR_H_
+#define _BNXT_CPR_H_
+
+#include "hsi_struct_def_dpdk.h"
+
+#define CMP_VALID(cmp, raw_cons, ring)					\
+	(!!(((struct cmpl_base *)(cmp))->info3_v & CMPL_BASE_V) ==	\
+	 !((raw_cons) & ((ring)->ring_size)))
+
+#define CMP_TYPE(cmp)						\
+	(((struct cmpl_base *)cmp)->type & CMPL_BASE_TYPE_MASK)
+
+#define ADV_RAW_CMP(idx, n)	((idx) + (n))
+#define NEXT_RAW_CMP(idx)	ADV_RAW_CMP(idx, 1)
+#define RING_CMP(ring, idx)	((idx) & (ring)->ring_mask)
+#define NEXT_CMP(idx)		RING_CMP(ADV_RAW_CMP(idx, 1))
+
+#define DB_CP_REARM_FLAGS	(DB_KEY_CP | DB_IDX_VALID)
+#define DB_CP_FLAGS		(DB_KEY_CP | DB_IDX_VALID | DB_IRQ_DIS)
+
+#define B_CP_DB_REARM(cpr, raw_cons)					\
+		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_REARM_FLAGS | \
+				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+
+#define B_CP_DIS_DB(cpr, raw_cons)					\
+		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
+				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+
+struct bnxt_ring;
+struct bnxt_cp_ring_info {
+	uint32_t		cp_raw_cons;
+	void			*cp_doorbell;
+
+	struct cmpl_base	*cp_desc_ring;
+
+	phys_addr_t		cp_desc_mapping;
+
+	struct ctx_hw_stats	*hw_stats;
+	phys_addr_t		hw_stats_map;
+	uint32_t		hw_stats_ctx_id;
+
+	struct bnxt_ring	*cp_ring_struct;
+};
+
+#define RX_CMP_L2_ERRORS						\
+	(RX_PKT_CMPL_ERRORS_BUFFER_ERROR_MASK | RX_PKT_CMPL_ERRORS_CRC_ERROR)
+
+
+struct bnxt;
+void bnxt_free_def_cp_ring(struct bnxt *bp);
+void bnxt_init_def_ring_struct(struct bnxt *bp);
+void bnxt_handle_async_event(struct bnxt *bp, struct cmpl_base *cmp);
+void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmp);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index a2d7815..f591ead 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -135,6 +135,24 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
+{
+	int rc;
+	struct hwrm_exec_fwd_resp_input req = {.req_type = 0 };
+	struct hwrm_exec_fwd_resp_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, EXEC_FWD_RESP, -1, resp);
+
+	memcpy(req.encap_request, fwd_cmd,
+	       sizeof(req.encap_request));
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index eef3be6..b792313 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,6 +41,8 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
+
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 92e39ca..9efb68b 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -59,6 +59,137 @@
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+#define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
+#define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
+#define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
+#define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
+
+/* Return Codes */
+#define HWRM_ERR_CODE_INVALID_PARAMS                      (UINT32_C(0x2))
+#define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED              (UINT32_C(0x3))
+
+/* HWRM Forwarded Request (16 bytes) */
+struct hwrm_fwd_req_cmpl {
+	/* Length of forwarded request in bytes. */
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define HWRM_FWD_REQ_CMPL_TYPE_MASK		UINT32_C(0x3f)
+	#define HWRM_FWD_REQ_CMPL_TYPE_SFT		0
+		/* Forwarded HWRM Request */
+	#define HWRM_FWD_REQ_CMPL_TYPE_HWRM_FWD_REQ	(UINT32_C(0x22) << 0)
+	/* Length of forwarded request in bytes. */
+	#define HWRM_FWD_REQ_CMPL_REQ_LEN_MASK		UINT32_C(0xffc0)
+	#define HWRM_FWD_REQ_CMPL_REQ_LEN_SFT		6
+	uint16_t req_len_type;
+
+	/*
+	 * Source ID of this request. Typically used in forwarding requests and
+	 * responses. 0x0 - 0xFFF8 - Used for function ids 0xFFF8 - 0xFFFE -
+	 * Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t source_id;
+
+	uint32_t unused_0;
+
+	/* Address of forwarded request. */
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define HWRM_FWD_REQ_CMPL_V			UINT32_C(0x1)
+	/* Address of forwarded request. */
+	#define HWRM_FWD_REQ_CMPL_REQ_BUF_ADDR_MASK	UINT32_C(0xfffffffe)
+	#define HWRM_FWD_REQ_CMPL_REQ_BUF_ADDR_SFT	1
+	uint64_t req_buf_addr_v;
+} __attribute__((packed));
+
+/* HWRM Asynchronous Event Completion Record (16 bytes) */
+struct hwrm_async_event_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_MASK		UINT32_C(0x3f)
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_SFT		0
+		/* HWRM Asynchronous Event Information */
+	#define HWRM_ASYNC_EVENT_CMPL_TYPE_HWRM_ASYNC_EVENT \
+							(UINT32_C(0x2e) << 0)
+	uint16_t type;
+
+	/* Identifiers of events. */
+		/* Link status changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_STATUS_CHANGE \
+							(UINT32_C(0x0) << 0)
+		/* Link MTU changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_MTU_CHANGE \
+							(UINT32_C(0x1) << 0)
+		/* Link speed changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_SPEED_CHANGE \
+							(UINT32_C(0x2) << 0)
+		/* DCB Configuration changed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_DCB_CONFIG_CHANGE \
+							(UINT32_C(0x3) << 0)
+		/* Port connection not allowed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PORT_CONN_NOT_ALLOWED \
+							(UINT32_C(0x4) << 0)
+		/* Link speed configuration was not allowed */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_LINK_SPEED_CFG_NOT_ALLOWED \
+							(UINT32_C(0x5) << 0)
+		/* Function driver unloaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_FUNC_DRVR_UNLOAD \
+							(UINT32_C(0x10) << 0)
+		/* Function driver loaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_FUNC_DRVR_LOAD \
+							(UINT32_C(0x11) << 0)
+		/* PF driver unloaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_DRVR_UNLOAD \
+							(UINT32_C(0x20) << 0)
+		/* PF driver loaded */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_DRVR_LOAD \
+							(UINT32_C(0x21) << 0)
+		/* VF Function Level Reset (FLR) */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_VF_FLR	(UINT32_C(0x30) << 0)
+		/* VF MAC Address Change */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_VF_MAC_ADDR_CHANGE \
+							(UINT32_C(0x31) << 0)
+		/* PF-VF communication channel status change. */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_PF_VF_COMM_STATUS_CHANGE \
+							(UINT32_C(0x32) << 0)
+		/* HWRM Error */
+	#define HWRM_ASYNC_EVENT_CMPL_EVENT_ID_HWRM_ERROR \
+							(UINT32_C(0xff) << 0)
+	uint16_t event_id;
+
+	/* Event specific data */
+	uint32_t event_data2;
+
+	/* opaque is 7 b */
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define HWRM_ASYNC_EVENT_CMPL_V				UINT32_C(0x1)
+	/* opaque is 7 b */
+	#define HWRM_ASYNC_EVENT_CMPL_OPAQUE_MASK		UINT32_C(0xfe)
+	#define HWRM_ASYNC_EVENT_CMPL_OPAQUE_SFT		1
+	uint8_t opaque_v;
+
+	/* 8-lsb timestamp from POR (100-msec resolution) */
+	uint8_t timestamp_lo;
+
+	/* 16-lsb timestamp from POR (100-msec resolution) */
+	uint16_t timestamp_hi;
+
+	/* Event specific data */
+	uint32_t event_data1;
+} __attribute__((packed));
 
 /*
  * Note: The Hardware Resource Manager (HWRM) manages various hardware resources
@@ -122,6 +253,102 @@ struct output {
 	uint16_t resp_len;
 } __attribute__((packed));
 
+/* hwrm_exec_fwd_resp */
+/*
+ * Description: This command is used to send an encapsulated request to the
+ * HWRM. This command instructs the HWRM to execute the request and forward the
+ * response of the encapsulated request to the location specified in the
+ * original request that is encapsulated. The target id of this command shall be
+ * set to 0xFFFF (HWRM). The response location in this command shall be used to
+ * acknowledge the receipt of the encapsulated request and forwarding of the
+ * response.
+ */
+
+/* Input (128 bytes) */
+struct hwrm_exec_fwd_resp_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This is an encapsulated request. This request should be executed by
+	 * the HWRM and the response should be provided in the response buffer
+	 * inside the encapsulated request.
+	 */
+	uint32_t encap_request[26];
+
+	/*
+	 * This value indicates the target id of the response to the
+	 * encapsulated request. 0x0 - 0xFFF8 - Used for function ids 0xFFF8 -
+	 * 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t encap_resp_target_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_exec_fwd_resp_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_func_qcaps */
 /*
  * Description: This command returns capabilities of a function. The input FID
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 09/38] bnxt: add L2 filter alloc/init/free
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (7 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 08/38] bnxt: add completion ring support Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 10/38] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
                       ` (29 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add the L2 filter structure and the alloc/init/free functions for
dealing with them.

A filter is used to identify traffic that contains a matching set of
parameters like unicast or broadcast MAC address or a VLAN tag amongst
other things which then allows the ASIC to direct the  incoming traffic
to an appropriate VNIC or Rx ring.

New HWRM calls:
bnxt_hwrm_clear_filter:
	Free a L2 filter.

bnxt_hwrm_set_filter
	Allocate an An L2 filter or a L2 context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Move bnxt_hwrm_set_filter() into this patch
Add short descriptions of new HWRM calls
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   3 +
 drivers/net/bnxt/bnxt_filter.c         | 175 +++++++++++++
 drivers/net/bnxt/bnxt_filter.h         |  74 ++++++
 drivers/net/bnxt/bnxt_hwrm.c           |  65 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   6 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 456 +++++++++++++++++++++++++++++++++
 7 files changed, 780 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_filter.c
 create mode 100644 drivers/net/bnxt/bnxt_filter.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index afd1690..b7834b1 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -50,6 +50,7 @@ EXPORT_MAP := rte_pmd_bnxt_version.map
 #
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_cpr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index bdd355f..49aa38b 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -145,6 +145,9 @@ struct bnxt {
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
+	struct bnxt_filter_info	*filter_info;
+	STAILQ_HEAD(, bnxt_filter_info)	free_filter_list;
+
 	/* VNIC pointer for flow filter (VMDq) pools */
 #define MAX_FF_POOLS	ETH_64_POOLS
 	STAILQ_HEAD(, bnxt_vnic_info)	ff_pool[MAX_FF_POOLS];
diff --git a/drivers/net/bnxt/bnxt_filter.c b/drivers/net/bnxt/bnxt_filter.c
new file mode 100644
index 0000000..f03a1dc
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_filter.c
@@ -0,0 +1,175 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <sys/queue.h>
+
+#include <rte_log.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * Filter Functions
+ */
+
+struct bnxt_filter_info *bnxt_alloc_filter(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+
+	/* Find the 1st unused filter from the free_filter_list pool*/
+	filter = STAILQ_FIRST(&bp->free_filter_list);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "No more free filter resources\n");
+		return NULL;
+	}
+	STAILQ_REMOVE_HEAD(&bp->free_filter_list, next);
+
+	/* Default to L2 MAC Addr filter */
+	filter->flags = HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX;
+	filter->enables = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
+			HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK;
+	memcpy(filter->l2_addr, bp->eth_dev->data->mac_addrs->addr_bytes,
+	       ETHER_ADDR_LEN);
+	memset(filter->l2_addr_mask, 0xff, ETHER_ADDR_LEN);
+	return filter;
+}
+
+void bnxt_init_filters(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+	int i, max_filters;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	STAILQ_INIT(&bp->free_filter_list);
+	for (i = 0; i < max_filters; i++) {
+		filter = &bp->filter_info[i];
+		filter->fw_l2_filter_id = -1;
+		STAILQ_INSERT_TAIL(&bp->free_filter_list, filter, next);
+	}
+}
+
+void bnxt_free_all_filters(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter, *temp_filter;
+	int i;
+
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			filter = STAILQ_FIRST(&vnic->filter);
+			while (filter) {
+				temp_filter = STAILQ_NEXT(filter, next);
+				STAILQ_REMOVE(&vnic->filter, filter,
+					      bnxt_filter_info, next);
+				STAILQ_INSERT_TAIL(&bp->free_filter_list,
+						   filter, next);
+				filter = temp_filter;
+			}
+			STAILQ_INIT(&vnic->filter);
+		}
+	}
+}
+
+void bnxt_free_filter_mem(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter;
+	uint16_t max_filters, i;
+	int rc = 0;
+
+	/* Ensure that all filters are freed */
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	for (i = 0; i < max_filters; i++) {
+		filter = &bp->filter_info[i];
+		if (filter->fw_l2_filter_id != ((uint64_t)-1)) {
+			RTE_LOG(ERR, PMD, "HWRM filter is not freed??\n");
+			/* Call HWRM to try to free filter again */
+			rc = bnxt_hwrm_clear_filter(bp, filter);
+			if (rc)
+				RTE_LOG(ERR, PMD,
+				       "HWRM filter cannot be freed rc = %d\n",
+					rc);
+		}
+		filter->fw_l2_filter_id = -1;
+	}
+	STAILQ_INIT(&bp->free_filter_list);
+
+	rte_free(bp->filter_info);
+	bp->filter_info = NULL;
+}
+
+int bnxt_alloc_filter_mem(struct bnxt *bp)
+{
+	struct bnxt_filter_info *filter_mem;
+	uint16_t max_filters;
+
+	if (BNXT_PF(bp)) {
+		struct bnxt_pf_info *pf = &bp->pf;
+
+		max_filters = pf->max_l2_ctx;
+	} else {
+		struct bnxt_vf_info *vf = &bp->vf;
+
+		max_filters = vf->max_l2_ctx;
+	}
+	/* Allocate memory for VNIC pool and filter pool */
+	filter_mem = rte_zmalloc("bnxt_filter_info",
+				 max_filters * sizeof(struct bnxt_filter_info),
+				 0);
+	if (filter_mem == NULL) {
+		RTE_LOG(ERR, PMD, "Failed to alloc memory for %d filters",
+			max_filters);
+		return -ENOMEM;
+	}
+	bp->filter_info = filter_mem;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_filter.h b/drivers/net/bnxt/bnxt_filter.h
new file mode 100644
index 0000000..06fe134
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_filter.h
@@ -0,0 +1,74 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_FILTER_H_
+#define _BNXT_FILTER_H_
+
+#include <rte_ether.h>
+
+struct bnxt;
+struct bnxt_filter_info {
+	STAILQ_ENTRY(bnxt_filter_info)	next;
+	uint64_t		fw_l2_filter_id;
+#define INVALID_MAC_INDEX	((uint16_t)-1)
+	uint16_t		mac_index;
+
+	/* Filter Characteristics */
+	uint32_t		flags;
+	uint32_t		enables;
+	uint8_t			l2_addr[ETHER_ADDR_LEN];
+	uint8_t			l2_addr_mask[ETHER_ADDR_LEN];
+	uint16_t		l2_ovlan;
+	uint16_t		l2_ovlan_mask;
+	uint16_t		l2_ivlan;
+	uint16_t		l2_ivlan_mask;
+	uint8_t			t_l2_addr[ETHER_ADDR_LEN];
+	uint8_t			t_l2_addr_mask[ETHER_ADDR_LEN];
+	uint16_t		t_l2_ovlan;
+	uint16_t		t_l2_ovlan_mask;
+	uint16_t		t_l2_ivlan;
+	uint16_t		t_l2_ivlan_mask;
+	uint8_t			tunnel_type;
+	uint16_t		mirror_vnic_id;
+	uint32_t		vni;
+	uint8_t			pri_hint;
+	uint64_t		l2_filter_id_hint;
+};
+
+struct bnxt_filter_info *bnxt_alloc_filter(struct bnxt *bp);
+void bnxt_init_filters(struct bnxt *bp);
+void bnxt_free_all_filters(struct bnxt *bp);
+void bnxt_free_filter_mem(struct bnxt *bp);
+int bnxt_alloc_filter_mem(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index f591ead..65cd4a7 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -39,7 +39,9 @@
 #include <rte_version.h>
 
 #include "bnxt.h"
+#include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
 #define HWRM_CMD_TIMEOUT		2000
@@ -135,6 +137,69 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_clear_filter(struct bnxt *bp,
+			   struct bnxt_filter_info *filter)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_filter_free_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_filter_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, CFA_L2_FILTER_FREE, -1, resp);
+
+	req.l2_filter_id = rte_cpu_to_le_64(filter->fw_l2_filter_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	filter->fw_l2_filter_id = -1;
+
+	return 0;
+}
+
+int bnxt_hwrm_set_filter(struct bnxt *bp,
+			 struct bnxt_vnic_info *vnic,
+			 struct bnxt_filter_info *filter)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_filter_alloc_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_filter_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t enables = 0;
+
+	HWRM_PREP(req, CFA_L2_FILTER_ALLOC, -1, resp);
+
+	req.flags = rte_cpu_to_le_32(filter->flags);
+
+	enables = filter->enables |
+	      HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_DST_ID;
+	req.dst_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR)
+		memcpy(req.l2_addr, filter->l2_addr,
+		       ETHER_ADDR_LEN);
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK)
+		memcpy(req.l2_addr_mask, filter->l2_addr_mask,
+		       ETHER_ADDR_LEN);
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN)
+		req.l2_ovlan = filter->l2_ovlan;
+	if (enables &
+	    HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN_MASK)
+		req.l2_ovlan_mask = filter->l2_ovlan_mask;
+
+	req.enables = rte_cpu_to_le_32(enables);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	filter->fw_l2_filter_id = rte_le_to_cpu_64(resp->l2_filter_id);
+
+	return rc;
+}
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd)
 {
 	int rc;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index b792313..bc6fb08 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -41,6 +41,12 @@
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_clear_filter(struct bnxt *bp,
+			   struct bnxt_filter_info *filter);
+int bnxt_hwrm_set_filter(struct bnxt *bp,
+			 struct bnxt_vnic_info *vnic,
+			 struct bnxt_filter_info *filter);
+
 int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 9efb68b..ca42a87 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -253,6 +253,462 @@ struct output {
 	uint16_t resp_len;
 } __attribute__((packed));
 
+/* hwrm_cfa_l2_filter_alloc */
+/*
+ * A filter is used to identify traffic that contains a matching set of
+ * parameters like unicast or broadcast MAC address or a VLAN tag amongst
+ * other things which then allows the ASIC to direct the  incoming traffic
+ * to an appropriate VNIC or Rx ring.
+ */
+
+/* Input (96 bytes) */
+struct hwrm_cfa_l2_filter_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * Enumeration denoting the RX, TX type of the resource. This
+	 * enumeration is used for resources that are similar for both TX and RX
+	 * paths of the chip.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH \
+							UINT32_C(0x1)
+		/* tx path */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_TX \
+							(UINT32_C(0x0) << 0)
+		/* rx path */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX \
+							(UINT32_C(0x1) << 0)
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_LAST \
+				HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX
+	/*
+	 * Setting of this flag indicates the applicability to the loopback
+	 * path.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_LOOPBACK \
+							UINT32_C(0x2)
+	/*
+	 * Setting of this flag indicates drop action. If this flag is not set,
+	 * then it should be considered accept action.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_DROP \
+							UINT32_C(0x4)
+	/*
+	 * If this flag is set, all t_l2_* fields are invalid and they should
+	 * not be specified. If this flag is set, then l2_* fields refer to
+	 * fields of outermost L2 header.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST \
+							UINT32_C(0x8)
+	uint32_t flags;
+
+	/* This bit must be '1' for the l2_addr field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR \
+							UINT32_C(0x1)
+	/* This bit must be '1' for the l2_addr_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK \
+							UINT32_C(0x2)
+	/* This bit must be '1' for the l2_ovlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN \
+							UINT32_C(0x4)
+	/* This bit must be '1' for the l2_ovlan_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_OVLAN_MASK \
+							UINT32_C(0x8)
+	/* This bit must be '1' for the l2_ivlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN \
+							UINT32_C(0x10)
+	/* This bit must be '1' for the l2_ivlan_mask field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK \
+							UINT32_C(0x20)
+	/* This bit must be '1' for the t_l2_addr field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_ADDR \
+							UINT32_C(0x40)
+	/*
+	 * This bit must be '1' for the t_l2_addr_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_ADDR_MASK \
+							UINT32_C(0x80)
+	/* This bit must be '1' for the t_l2_ovlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_OVLAN \
+							UINT32_C(0x100)
+	/*
+	 * This bit must be '1' for the t_l2_ovlan_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_OVLAN_MASK \
+							UINT32_C(0x200)
+	/* This bit must be '1' for the t_l2_ivlan field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_IVLAN \
+							UINT32_C(0x400)
+	/*
+	 * This bit must be '1' for the t_l2_ivlan_mask field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_T_L2_IVLAN_MASK \
+							UINT32_C(0x800)
+	/* This bit must be '1' for the src_type field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_SRC_TYPE \
+							UINT32_C(0x1000)
+	/* This bit must be '1' for the src_id field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_SRC_ID \
+							UINT32_C(0x2000)
+	/* This bit must be '1' for the tunnel_type field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE \
+							UINT32_C(0x4000)
+	/* This bit must be '1' for the dst_id field to be configured. */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_DST_ID \
+							UINT32_C(0x8000)
+	/*
+	 * This bit must be '1' for the mirror_vnic_id field to be configured.
+	 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_MIRROR_VNIC_ID \
+							UINT32_C(0x10000)
+	uint32_t enables;
+
+	/*
+	 * This value sets the match value for the L2 MAC address. Destination
+	 * MAC address for RX path. Source MAC address for TX path.
+	 */
+	uint8_t l2_addr[6];
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+
+	/*
+	 * This value sets the mask value for the L2 address. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint8_t l2_addr_mask[6];
+
+	/* This value sets VLAN ID value for outer VLAN. */
+	uint16_t l2_ovlan;
+
+	/*
+	 * This value sets the mask value for the ovlan id. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint16_t l2_ovlan_mask;
+
+	/* This value sets VLAN ID value for inner VLAN. */
+	uint16_t l2_ivlan;
+
+	/*
+	 * This value sets the mask value for the ivlan id. A value of 0 will
+	 * mask the corresponding bit from compare.
+	 */
+	uint16_t l2_ivlan_mask;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This value sets the match value for the tunnel L2 MAC address.
+	 * Destination MAC address for RX path. Source MAC address for TX path.
+	 */
+	uint8_t t_l2_addr[6];
+
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/*
+	 * This value sets the mask value for the tunnel L2 address. A value of
+	 * 0 will mask the corresponding bit from compare.
+	 */
+	uint8_t t_l2_addr_mask[6];
+
+	/* This value sets VLAN ID value for tunnel outer VLAN. */
+	uint16_t t_l2_ovlan;
+
+	/*
+	 * This value sets the mask value for the tunnel ovlan id. A value of 0
+	 * will mask the corresponding bit from compare.
+	 */
+	uint16_t t_l2_ovlan_mask;
+
+	/* This value sets VLAN ID value for tunnel inner VLAN. */
+	uint16_t t_l2_ivlan;
+
+	/*
+	 * This value sets the mask value for the tunnel ivlan id. A value of 0
+	 * will mask the corresponding bit from compare.
+	 */
+	uint16_t t_l2_ivlan_mask;
+
+	/* This value identifies the type of source of the packet. */
+		/* Network port */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_NPORT \
+							(UINT32_C(0x0) << 0)
+		/* Physical function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_PF \
+							(UINT32_C(0x1) << 0)
+		/* Virtual function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_VF \
+							(UINT32_C(0x2) << 0)
+		/* Virtual NIC of a function */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_VNIC \
+							(UINT32_C(0x3) << 0)
+		/* Embedded processor for CFA management */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_KONG \
+							(UINT32_C(0x4) << 0)
+		/* Embedded processor for OOB management */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_APE \
+							(UINT32_C(0x5) << 0)
+		/* Embedded processor for RoCE */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_BONO \
+							(UINT32_C(0x6) << 0)
+		/* Embedded processor for network proxy functions */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_SRC_TYPE_TANG \
+							(UINT32_C(0x7) << 0)
+	uint8_t src_type;
+
+	uint8_t unused_6;
+	/*
+	 * This value is the id of the source. For a network port, it represents
+	 * port_id. For a physical function, it represents fid. For a virtual
+	 * function, it represents vf_id. For a vnic, it represents vnic_id. For
+	 * embedded processors, this id is not valid. Notes: 1. The function ID
+	 * is implied if it src_id is not provided for a src_type that is either
+	 */
+	uint32_t src_id;
+
+	/* Tunnel Type. */
+		/* Non-tunnel */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_NONTUNNEL \
+							(UINT32_C(0x0) << 0)
+		/* Virtual eXtensible Local Area Network (VXLAN) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_VXLAN \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Network Virtualization Generic Routing Encapsulation (NVGRE)
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_NVGRE \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Generic Routing Encapsulation (GRE) inside Ethernet payload
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_L2GRE \
+							(UINT32_C(0x3) << 0)
+		/* IP in IP */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_IPIP \
+							(UINT32_C(0x4) << 0)
+		/* Generic Network Virtualization Encapsulation (Geneve) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_GENEVE \
+							(UINT32_C(0x5) << 0)
+		/* Multi-Protocol Lable Switching (MPLS) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_MPLS \
+							(UINT32_C(0x6) << 0)
+		/* Stateless Transport Tunnel (STT) */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_STT \
+							(UINT32_C(0x7) << 0)
+		/*
+		 * Generic Routing Encapsulation (GRE) inside IP datagram
+		 * payload
+		 */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_IPGRE \
+							(UINT32_C(0x8) << 0)
+		/* Any tunneled traffic */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_TUNNEL_TYPE_ANYTUNNEL \
+							(UINT32_C(0xff) << 0)
+	uint8_t tunnel_type;
+
+	uint8_t unused_7;
+
+	/*
+	 * If set, this value shall represent the Logical VNIC ID of the
+	 * destination VNIC for the RX path and network port id of the
+	 * destination port for the TX path.
+	 */
+	uint16_t dst_id;
+
+	/* Logical VNIC ID of the VNIC where traffic is mirrored. */
+	uint16_t mirror_vnic_id;
+
+	/*
+	 * This hint is provided to help in placing the filter in the filter
+	 * table.
+	 */
+		/* No preference */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_NO_PREFER \
+							(UINT32_C(0x0) << 0)
+		/* Above the given filter */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_ABOVE_FILTER \
+							(UINT32_C(0x1) << 0)
+		/* Below the given filter */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_BELOW_FILTER \
+							(UINT32_C(0x2) << 0)
+		/* As high as possible */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_MAX \
+							(UINT32_C(0x3) << 0)
+		/* As low as possible */
+	#define HWRM_CFA_L2_FILTER_ALLOC_INPUT_PRI_HINT_MIN \
+							(UINT32_C(0x4) << 0)
+	uint8_t pri_hint;
+
+	uint8_t unused_8;
+	uint32_t unused_9;
+
+	/*
+	 * This is the ID of the filter that goes along with the pri_hint. This
+	 * field is valid only for the following values. 1 - Above the given
+	 * filter 2 - Below the given filter
+	 */
+	uint64_t l2_filter_id_hint;
+} __attribute__((packed));
+
+/* Output (24 bytes) */
+struct hwrm_cfa_l2_filter_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This value identifies a set of CFA data structures used for an L2
+	 * context.
+	 */
+	uint64_t l2_filter_id;
+
+	/*
+	 * This is the ID of the flow associated with this filter. This value
+	 * shall be used to match and associate the flow identifier returned in
+	 * completion records. A value of 0xFFFFFFFF shall indicate no flow id.
+	 */
+	uint32_t flow_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_cfa_l2_filter_free */
+/*
+ * Description: Free a L2 filter. The HWRM shall free all associated filter
+ * resources with the L2 filter.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_cfa_l2_filter_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * This value identifies a set of CFA data structures used for an L2
+	 * context.
+	 */
+	uint64_t l2_filter_id;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_cfa_l2_filter_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_exec_fwd_resp */
 /*
  * Description: This command is used to send an encapsulated request to the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 10/38] bnxt: add Tx queue operations (nonfunctional)
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (8 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 09/38] bnxt: add L2 filter alloc/init/free Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 11/38] bnxt: add Rx queue create/destroy operations Stephen Hurd
                       ` (28 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

In this patch we are adding the bnxt_tx_queue_setup_op() and
bnxt_tx_queue_release_op() functions. These will be tied to the
tx_queue_setup and tx_queue_release dev_ops in a subsequent patch.
In these functions we allocate/free memory for the TX queues.

This still requires support to create a TX ring in the ASIC which
will be completed in a future commit. Each Tx queue created via the
tx_queue_setup dev_op will have an associated Tx ring in the hardware.

A Tx request coming on the Tx queue gets sent to the corresponding
Tx ring in the ASIC for subsequent transmission.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/Makefile      |   1 +
 drivers/net/bnxt/bnxt_ethdev.c |   3 +
 drivers/net/bnxt/bnxt_txq.c    | 124 +++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_txq.h    |  75 +++++++++++++++++++++++++
 4 files changed, 203 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_txq.c
 create mode 100644 drivers/net/bnxt/bnxt_txq.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index b7834b1..13a90b9 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -53,6 +53,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b46d2ce..77a6d92 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -41,6 +41,7 @@
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -185,6 +186,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.tx_queue_setup = bnxt_tx_queue_setup_op,
+	.tx_queue_release = bnxt_tx_queue_release_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
new file mode 100644
index 0000000..8928984
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -0,0 +1,124 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_ring.h"
+#include "bnxt_txq.h"
+
+/*
+ * TX Queues
+ */
+
+void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+
+	/* 'Unreserve' rte_memzone */
+
+	if (cpr->hw_stats)
+		cpr->hw_stats = NULL;
+}
+
+static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq __rte_unused)
+{
+	/* TODO: Requires interaction with TX ring */
+}
+
+void bnxt_free_tx_mbufs(struct bnxt *bp)
+{
+	struct bnxt_tx_queue *txq;
+	int i;
+
+	for (i = 0; i < (int)bp->tx_nr_rings; i++) {
+		txq = bp->tx_queues[i];
+		bnxt_tx_queue_release_mbufs(txq);
+	}
+}
+
+void bnxt_tx_queue_release_op(void *tx_queue)
+{
+	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
+
+	if (txq) {
+		/* TODO: Free ring and stats here */
+		rte_free(txq);
+	}
+}
+
+int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_txconf *tx_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_tx_queue *txq;
+
+	if (!nb_desc || nb_desc > MAX_TX_DESC_CNT) {
+		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
+		return -EINVAL;
+	}
+
+	if (eth_dev->data->tx_queues) {
+		txq = eth_dev->data->tx_queues[queue_idx];
+		if (txq) {
+			bnxt_tx_queue_release_op(txq);
+			txq = NULL;
+		}
+	}
+	txq = rte_zmalloc_socket("bnxt_tx_queue", sizeof(struct bnxt_tx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq == NULL) {
+		RTE_LOG(ERR, PMD, "bnxt_tx_queue allocation failed!");
+		return -ENOMEM;
+	}
+	txq->bp = bp;
+	txq->nb_tx_desc = nb_desc;
+	txq->tx_free_thresh = tx_conf->tx_free_thresh;
+
+	/* TODO: Initialize ring structure */
+
+	txq->queue_id = queue_idx;
+	txq->port_id = eth_dev->data->port_id;
+
+	/* TODO: Allocate TX ring hardware descriptors */
+
+	/* TODO: Initialize the ring */
+
+	eth_dev->data->tx_queues[queue_idx] = txq;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_txq.h b/drivers/net/bnxt/bnxt_txq.h
new file mode 100644
index 0000000..16f3a0b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txq.h
@@ -0,0 +1,75 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_TXQ_H_
+#define _BNXT_TXQ_H_
+
+struct bnxt_tx_ring_info;
+struct bnxt_cp_ring_info;
+struct bnxt_tx_queue {
+	uint16_t		nb_tx_desc;    /* number of TX descriptors */
+	uint16_t		tx_free_thresh;/* minimum TX before freeing */
+	/** Index to last TX descriptor to have been cleaned. */
+	uint16_t		last_desc_cleaned;
+	/** Total number of TX descriptors ready to be allocated. */
+	uint16_t		tx_next_dd; /* next desc to scan for DD bit */
+	uint16_t		tx_next_rs; /* next desc to set RS bit */
+	uint16_t		queue_id; /* TX queue index */
+	uint16_t		reg_idx; /* TX queue register index */
+	uint8_t			port_id; /* Device port identifier */
+	uint8_t			pthresh; /* Prefetch threshold register */
+	uint8_t			hthresh; /* Host threshold register */
+	uint8_t			wthresh; /* Write-back threshold reg */
+	uint32_t		txq_flags; /* Holds flags for this TXq */
+	uint32_t		ctx_curr; /* Hardware context states */
+	uint8_t			tx_deferred_start; /* not in global dev start */
+
+	struct bnxt		*bp;
+	int			index;
+	int			tx_wake_thresh;
+	struct bnxt_tx_ring_info	*tx_ring;
+
+	unsigned int		cp_nr_rings;
+	struct bnxt_cp_ring_info	*cp_ring;
+};
+
+void bnxt_free_txq_stats(struct bnxt_tx_queue *txq);
+void bnxt_free_tx_mbufs(struct bnxt *bp);
+void bnxt_tx_queue_release_op(void *tx_queue);
+int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_txconf *tx_conf);
+
+#endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 11/38] bnxt: add Rx queue create/destroy operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (9 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 10/38] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 12/38] bnxt: add statistics operations Stephen Hurd
                       ` (27 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

In this patch we are adding the bnxt_rx_queue_setup_op() and
bnxt_rx_queue_release_op() functions. These will be tied to the
rx_queue_setup and rx_queue_release dev_ops in a subsequent patch.
In these functions we allocate/free memory for the RX queues.

This still requires support to create a RX ring in the ASIC which
will be completed in a future commit. Each Rx queue created via the
rx_queue_setup dev_op will have an associated Rx ring in the hardware.

The Rx logic in the hardware picks a Rx ring for each Rx frame received
by the hardware depending on the properties like RSS, MAC and VLAN
settings configured in the hardware. These packets in the end arrive
on the Rx queue corresponding to the Rx ring in the hardware.

We are also adding some functions like bnxt_mq_rx_configure()
bnxt_free_rx_mbufs() and bnxt_free_rxq_stats() which will be used in
subsequent patches.

We are also adding hwrm_vnic_rss_cfg_* structures, which will be used
in subsequent patches to enable RSS configuration.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt.h                |   2 +
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +
 drivers/net/bnxt/bnxt_rxq.c            | 287 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_rxq.h            |  74 +++++++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 121 ++++++++++++++
 6 files changed, 488 insertions(+)
 create mode 100644 drivers/net/bnxt/bnxt_rxq.c
 create mode 100644 drivers/net/bnxt/bnxt_rxq.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 13a90b9..21ed71c 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -53,6 +53,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index 49aa38b..f7cf9d1 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -142,6 +142,8 @@ struct bnxt {
 	/* Default completion ring */
 	struct bnxt_cp_ring_info	*def_cp_ring;
 
+	unsigned int		nr_vnics;
+
 	struct bnxt_vnic_info	*vnic_info;
 	STAILQ_HEAD(, bnxt_vnic_info)	free_vnic_list;
 
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 77a6d92..7e7d1ab 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -41,6 +41,7 @@
 
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
 #include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
@@ -186,6 +187,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.rx_queue_setup = bnxt_rx_queue_setup_op,
+	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
 };
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
new file mode 100644
index 0000000..d5c3173
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -0,0 +1,287 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_filter.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
+#include "bnxt_rxq.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * RX Queues
+ */
+
+void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+
+	/* 'Unreserve' rte_memzone */
+
+	if (cpr->hw_stats)
+		cpr->hw_stats = NULL;
+}
+
+int bnxt_mq_rx_configure(struct bnxt *bp)
+{
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	unsigned int i, j, nb_q_per_grp, ring_idx;
+	int start_grp_id, end_grp_id, rc = 0;
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter;
+	struct bnxt_rx_queue *rxq;
+
+	bp->nr_vnics = 0;
+
+	/* Single queue mode */
+	if (bp->rx_cp_nr_rings < 2) {
+		vnic = bnxt_alloc_vnic(bp);
+		if (!vnic) {
+			RTE_LOG(ERR, PMD, "VNIC alloc failed\n");
+			rc = -ENOMEM;
+			goto err_out;
+		}
+		STAILQ_INSERT_TAIL(&bp->ff_pool[0], vnic, next);
+		bp->nr_vnics++;
+
+		rxq = bp->eth_dev->data->rx_queues[0];
+		rxq->vnic = vnic;
+
+		vnic->func_default = true;
+		vnic->ff_pool_idx = 0;
+		vnic->start_grp_id = 1;
+		vnic->end_grp_id = vnic->start_grp_id +
+				   bp->rx_cp_nr_rings - 1;
+		filter = bnxt_alloc_filter(bp);
+		if (!filter) {
+			RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+			rc = -ENOMEM;
+			goto err_out;
+		}
+		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+		goto out;
+	}
+
+	/* Multi-queue mode */
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG) {
+		/* VMDq ONLY, VMDq+RSS, VMDq+DCB, VMDq+DCB+RSS */
+		enum rte_eth_nb_pools pools;
+
+		switch (dev_conf->rxmode.mq_mode) {
+		case ETH_MQ_RX_VMDQ_RSS:
+		case ETH_MQ_RX_VMDQ_ONLY:
+			{
+				const struct rte_eth_vmdq_rx_conf *conf =
+				    &dev_conf->rx_adv_conf.vmdq_rx_conf;
+
+				/* ETH_8/64_POOLs */
+				pools = conf->nb_queue_pools;
+				break;
+			}
+		default:
+			RTE_LOG(ERR, PMD, "Unsupported mq_mod %d\n",
+				dev_conf->rxmode.mq_mode);
+			rc = -EINVAL;
+			goto err_out;
+		}
+		/* For each pool, allocate MACVLAN CFA rule & VNIC */
+		if (!pools) {
+			RTE_LOG(ERR, PMD,
+				"VMDq pool not set, defaulted to 64\n");
+			pools = ETH_64_POOLS;
+		}
+		nb_q_per_grp = bp->rx_cp_nr_rings / pools;
+		start_grp_id = 1;
+		end_grp_id = start_grp_id + nb_q_per_grp - 1;
+
+		ring_idx = 0;
+		for (i = 0; i < pools; i++) {
+			vnic = bnxt_alloc_vnic(bp);
+			if (!vnic) {
+				RTE_LOG(ERR, PMD,
+					"VNIC alloc failed\n");
+				rc = -ENOMEM;
+				goto err_out;
+			}
+			STAILQ_INSERT_TAIL(&bp->ff_pool[i], vnic, next);
+			bp->nr_vnics++;
+
+			for (j = 0; j < nb_q_per_grp; j++, ring_idx++) {
+				rxq = bp->eth_dev->data->rx_queues[ring_idx];
+				rxq->vnic = vnic;
+			}
+			if (i == 0)
+				vnic->func_default = true;
+			vnic->ff_pool_idx = i;
+			vnic->start_grp_id = start_grp_id;
+			vnic->end_grp_id = end_grp_id;
+
+			filter = bnxt_alloc_filter(bp);
+			if (!filter) {
+				RTE_LOG(ERR, PMD,
+					"L2 filter alloc failed\n");
+				rc = -ENOMEM;
+				goto err_out;
+			}
+			/*
+			 * TODO: Configure & associate CFA rule for
+			 * each VNIC for each VMDq with MACVLAN, MACVLAN+TC
+			 */
+			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+
+			start_grp_id = end_grp_id + 1;
+			end_grp_id += nb_q_per_grp;
+		}
+		goto out;
+	}
+
+	/* Non-VMDq mode - RSS, DCB, RSS+DCB */
+	/* Init default VNIC for RSS or DCB only */
+	vnic = bnxt_alloc_vnic(bp);
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "VNIC alloc failed\n");
+		rc = -ENOMEM;
+		goto err_out;
+	}
+	/* Partition the rx queues for the single pool */
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		rxq = bp->eth_dev->data->rx_queues[i];
+		rxq->vnic = vnic;
+	}
+	STAILQ_INSERT_TAIL(&bp->ff_pool[0], vnic, next);
+	bp->nr_vnics++;
+
+	vnic->func_default = true;
+	vnic->ff_pool_idx = 0;
+	vnic->start_grp_id = 1;
+	vnic->end_grp_id = vnic->start_grp_id +
+			   bp->rx_cp_nr_rings - 1;
+	filter = bnxt_alloc_filter(bp);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+		rc = -ENOMEM;
+		goto err_out;
+	}
+	STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
+		vnic->hash_type =
+			HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4 |
+			HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+
+out:
+	return rc;
+
+err_out:
+	/* Free allocated vnic/filters */
+
+	return rc;
+}
+
+static void bnxt_rx_queue_release_mbufs(struct bnxt_rx_queue *rxq __rte_unused)
+{
+	/* TODO: Requires interaction with TX ring */
+}
+
+void bnxt_free_rx_mbufs(struct bnxt *bp)
+{
+	struct bnxt_rx_queue *rxq;
+	int i;
+
+	for (i = 0; i < (int)bp->rx_nr_rings; i++) {
+		rxq = bp->rx_queues[i];
+		bnxt_rx_queue_release_mbufs(rxq);
+	}
+}
+
+void bnxt_rx_queue_release_op(void *rx_queue)
+{
+	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
+
+	if (rxq) {
+		bnxt_rx_queue_release_mbufs(rxq);
+
+		/* TODO: Free ring and stats here */
+
+		rte_free(rxq);
+	}
+}
+
+int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_rxconf *rx_conf,
+			       struct rte_mempool *mp)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_rx_queue *rxq;
+
+	if (!nb_desc || nb_desc > MAX_RX_DESC_CNT) {
+		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
+		return -EINVAL;
+	}
+
+	if (eth_dev->data->rx_queues) {
+		rxq = eth_dev->data->rx_queues[queue_idx];
+		if (rxq)
+			bnxt_rx_queue_release_op(rxq);
+	}
+	rxq = rte_zmalloc_socket("bnxt_rx_queue", sizeof(struct bnxt_rx_queue),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (!rxq) {
+		RTE_LOG(ERR, PMD, "bnxt_rx_queue allocation failed!");
+		return -ENOMEM;
+	}
+	rxq->bp = bp;
+	rxq->mb_pool = mp;
+	rxq->nb_rx_desc = nb_desc;
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+
+	/* TODO: Initialize ring structure */
+
+	rxq->queue_id = queue_idx;
+	rxq->port_id = eth_dev->data->port_id;
+	rxq->crc_len = (uint8_t)((eth_dev->data->dev_conf.rxmode.hw_strip_crc) ?
+				0 : ETHER_CRC_LEN);
+
+	eth_dev->data->rx_queues[queue_idx] = rxq;
+	/* TODO: Allocate RX ring hardware descriptors */
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_rxq.h b/drivers/net/bnxt/bnxt_rxq.h
new file mode 100644
index 0000000..9554329
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxq.h
@@ -0,0 +1,74 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RQX_H_
+#define _BNXT_RQX_H_
+
+struct bnxt;
+struct bnxt_rx_ring_info;
+struct bnxt_cp_ring_info;
+struct bnxt_rx_queue {
+	struct rte_mempool	*mb_pool; /* mbuf pool for RX ring */
+	struct rte_mbuf		*pkt_first_seg; /* 1st seg of pkt */
+	struct rte_mbuf		*pkt_last_seg; /* Last seg of pkt */
+	uint64_t		mbuf_initializer; /* val to init mbuf */
+	uint16_t		nb_rx_desc; /* num of RX desc */
+	uint16_t		rx_tail; /* cur val of RDT register */
+	uint16_t		nb_rx_hold; /* num held free RX desc */
+	uint16_t		rx_free_thresh; /* max free RX desc to hold */
+	uint16_t		queue_id; /* RX queue index */
+	uint16_t		reg_idx; /* RX queue register index */
+	uint8_t			port_id; /* Device port identifier */
+	uint8_t			crc_len; /* 0 if CRC stripped, 4 otherwise */
+
+	struct bnxt		*bp;
+	struct bnxt_vnic_info	*vnic;
+
+	uint32_t			rx_buf_size;
+	uint32_t			rx_buf_use_size;  /* useable size */
+	struct bnxt_rx_ring_info	*rx_ring;
+	struct bnxt_cp_ring_info	*cp_ring;
+};
+
+void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq);
+int bnxt_mq_rx_configure(struct bnxt *bp);
+void bnxt_rx_queue_release_op(void *rx_queue);
+int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
+			       uint16_t queue_idx,
+			       uint16_t nb_desc,
+			       unsigned int socket_id,
+			       const struct rte_eth_rxconf *rx_conf,
+			       struct rte_mempool *mp);
+void bnxt_free_rx_mbufs(struct bnxt *bp);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ca42a87..bbec5bb 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1897,6 +1897,127 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_vnic_rss_cfg */
+/* Description: This function is used to enable RSS configuration. */
+
+/* Input (48 bytes) */
+struct hwrm_vnic_rss_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over source and
+	 * destination IPv4 addresses of IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4		UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv4 addresses and source/destination ports of
+	 * TCP/IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv4 addresses and source/destination ports of
+	 * UDP/IPv4 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4	UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over source and
+	 * destination IPv4 addresses of IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6		UINT32_C(0x8)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv6 addresses and source/destination ports of
+	 * TCP/IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6	UINT32_C(0x10)
+	/*
+	 * When this bit is '1', the RSS hash shall be computed over
+	 * source/destination IPv6 addresses and source/destination ports of
+	 * UDP/IPv6 packets.
+	 */
+	#define HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6	UINT32_C(0x20)
+	uint32_t hash_type;
+
+	uint32_t unused_0;
+
+	/* This is the address for rss ring group table */
+	uint64_t ring_grp_tbl_addr;
+
+	/* This is the address for rss hash key table */
+	uint64_t hash_key_tbl_addr;
+
+	/* Index to the rss indirection table. */
+	uint16_t rss_ctx_idx;
+
+	uint16_t unused_1[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_rss_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* Output (32 bytes) */
 struct hwrm_queue_qportcfg_output {
 	/*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 12/38] bnxt: add statistics operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (10 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 11/38] bnxt: add Rx queue create/destroy operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 13/38] bnxt: add initial Tx code implementation Stephen Hurd
                       ` (26 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add the bnxt_stats_get_op() and bnxt_stats_reset_op() dev_ops to
get and reset statistics. It also brings in the associated HWRM calls
to handle the requests appropriately.

We also have the bnxt_free_stats() function which will be used in the
follow on patches to free the memory allocated by the driver for
statistics.

New HWRM calls:
bnxt_hwrm_stat_clear:
	This command clears statistics of a context

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Add more description for the patch
Add HWRM function descriptions
Reorder footer

v6:
Correct typos
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_cpr.c            |   1 +
 drivers/net/bnxt/bnxt_cpr.h            |   2 -
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +
 drivers/net/bnxt/bnxt_hwrm.c           |  49 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   7 +-
 drivers/net/bnxt/bnxt_rxq.c            |   1 +
 drivers/net/bnxt/bnxt_stats.c          | 142 +++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_stats.h          |  44 ++++++++++
 drivers/net/bnxt/bnxt_txq.c            |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 107 +++++++++++++++++++++++++
 11 files changed, 355 insertions(+), 3 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_stats.c
 create mode 100644 drivers/net/bnxt/bnxt_stats.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 21ed71c..f6a04f8 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -54,6 +54,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index 7de2b9c..5a4b13d 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -35,6 +35,7 @@
 #include "bnxt_cpr.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
+#include "hsi_struct_def_dpdk.h"
 
 /*
  * Async event handling
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index ecabc5f..2be22de 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -34,8 +34,6 @@
 #ifndef _BNXT_CPR_H_
 #define _BNXT_CPR_H_
 
-#include "hsi_struct_def_dpdk.h"
-
 #define CMP_VALID(cmp, raw_cons, ring)					\
 	(!!(((struct cmpl_base *)(cmp))->info3_v & CMPL_BASE_V) ==	\
 	 !((raw_cons) & ((ring)->ring_size)))
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 7e7d1ab..3453509 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -42,6 +42,7 @@
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_stats.h"
 #include "bnxt_txq.h"
 
 #define DRV_MODULE_NAME		"bnxt"
@@ -187,6 +188,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.stats_get = bnxt_stats_get_op,
+	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 65cd4a7..633683b 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -39,8 +39,11 @@
 #include <rte_version.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
+#include "bnxt_txq.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -480,10 +483,56 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
+{
+	int rc = 0;
+	struct hwrm_stat_ctx_clr_stats_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_clr_stats_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_CLR_STATS, -1, resp);
+
+	if (cpr->hw_stats_ctx_id == (uint32_t)HWRM_NA_SIGNATURE)
+		return rc;
+
+	req.stat_ctx_id = rte_cpu_to_le_16(cpr->hw_stats_ctx_id);
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
 
+int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq;
+		struct bnxt_rx_queue *rxq;
+		struct bnxt_cp_ring_info *cpr;
+
+		if (i >= bp->rx_cp_nr_rings) {
+			txq = bp->tx_queues[i - bp->rx_cp_nr_rings];
+			cpr = txq->cp_ring;
+		} else {
+			rxq = bp->rx_queues[i];
+			cpr = rxq->cp_ring;
+		}
+
+		rc = bnxt_hwrm_stat_clear(bp, cpr);
+		if (rc)
+			return rc;
+	}
+	return 0;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index bc6fb08..75c3442 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -37,7 +37,9 @@
 #include <inttypes.h>
 #include <stdbool.h>
 
-#include "bnxt.h"
+struct bnxt;
+struct bnxt_filter_info;
+struct bnxt_cp_ring_info;
 
 #define HWRM_SEQ_ID_INVALID -1U
 
@@ -56,8 +58,11 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
+int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
+
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index d5c3173..19c0e3f 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -36,6 +36,7 @@
 #include <rte_malloc.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
diff --git a/drivers/net/bnxt/bnxt_stats.c b/drivers/net/bnxt/bnxt_stats.c
new file mode 100644
index 0000000..6f1c760
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_stats.c
@@ -0,0 +1,142 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_byteorder.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
+#include "bnxt_rxq.h"
+#include "bnxt_stats.h"
+#include "bnxt_txq.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * Statistics functions
+ */
+
+void bnxt_free_stats(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+
+		bnxt_free_txq_stats(txq);
+	}
+	for (i = 0; i < (int)bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+
+		bnxt_free_rxq_stats(rxq);
+	}
+}
+
+void bnxt_stats_get_op(struct rte_eth_dev *eth_dev,
+			   struct rte_eth_stats *bnxt_stats)
+{
+	unsigned int i;
+	struct bnxt *bp = eth_dev->data->dev_private;
+
+	memset(bnxt_stats, 0, sizeof(*bnxt_stats));
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		struct ctx_hw_stats64 *hw_stats =
+		    (struct ctx_hw_stats64 *)cpr->hw_stats;
+
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_ucast_pkts);
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_mcast_pkts);
+		bnxt_stats->q_ipackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_bcast_pkts);
+
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_ucast_bytes);
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_mcast_bytes);
+		bnxt_stats->q_ibytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->rx_bcast_bytes);
+
+		/*
+		 * TBD: No clear mapping to this... we don't seem
+		 * to have a stat specifically for dropped due to
+		 * insufficient mbufs.
+		 */
+		bnxt_stats->q_errors[i] = 0;
+
+		/* These get replaced once the *_QSTATS commands work */
+		bnxt_stats->ipackets += bnxt_stats->q_ipackets[i];
+		bnxt_stats->ibytes += bnxt_stats->q_ibytes[i];
+		bnxt_stats->imissed += bnxt_stats->q_errors[i];
+		bnxt_stats->ierrors +=
+				rte_le_to_cpu_64(hw_stats->rx_err_pkts);
+	}
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		struct ctx_hw_stats64 *hw_stats =
+		    (struct ctx_hw_stats64 *)cpr->hw_stats;
+
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_ucast_pkts);
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_mcast_pkts);
+		bnxt_stats->q_opackets[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_bcast_pkts);
+
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_ucast_bytes);
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_mcast_bytes);
+		bnxt_stats->q_obytes[i] +=
+		    rte_le_to_cpu_64(hw_stats->tx_bcast_bytes);
+
+		/* These get replaced once the *_QSTATS commands work */
+		bnxt_stats->opackets += bnxt_stats->q_opackets[i];
+		bnxt_stats->obytes +=  bnxt_stats->q_obytes[i];
+		bnxt_stats->oerrors += rte_le_to_cpu_64(hw_stats->tx_drop_pkts);
+		bnxt_stats->oerrors += rte_le_to_cpu_64(hw_stats->tx_err_pkts);
+	}
+}
+
+void bnxt_stats_reset_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	bnxt_clear_all_hwrm_stat_ctxs(bp);
+}
diff --git a/drivers/net/bnxt/bnxt_stats.h b/drivers/net/bnxt/bnxt_stats.h
new file mode 100644
index 0000000..65408a4
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_stats.h
@@ -0,0 +1,44 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2014-2015 Broadcom Corporation.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_STATS_H_
+#define _BNXT_STATS_H_
+
+#include <rte_ethdev.h>
+
+void bnxt_free_stats(struct bnxt *bp);
+void bnxt_stats_get_op(struct rte_eth_dev *eth_dev,
+			   struct rte_eth_stats *bnxt_stats);
+void bnxt_stats_reset_op(struct rte_eth_dev *eth_dev);
+
+#endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 8928984..8793c9d 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -36,6 +36,7 @@
 #include <rte_malloc.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
 
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index bbec5bb..91a83d1 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -34,6 +34,35 @@
 #ifndef _HSI_STRUCT_DEF_EXTERNAL_H_
 #define _HSI_STRUCT_DEF_EXTERNAL_H_
 
+/*
+ * per-context HW statistics -- chip view
+ */
+
+struct ctx_hw_stats64 {
+	uint64_t rx_ucast_pkts;
+	uint64_t rx_mcast_pkts;
+	uint64_t rx_bcast_pkts;
+	uint64_t rx_drop_pkts;
+	uint64_t rx_err_pkts;
+	uint64_t rx_ucast_bytes;
+	uint64_t rx_mcast_bytes;
+	uint64_t rx_bcast_bytes;
+
+	uint64_t tx_ucast_pkts;
+	uint64_t tx_mcast_pkts;
+	uint64_t tx_bcast_pkts;
+	uint64_t tx_drop_pkts;
+	uint64_t tx_err_pkts;
+	uint64_t tx_ucast_bytes;
+	uint64_t tx_mcast_bytes;
+	uint64_t tx_bcast_bytes;
+
+	uint64_t tpa_pkts;
+	uint64_t tpa_bytes;
+	uint64_t tpa_events;
+	uint64_t tpa_aborts;
+} ctx_hw_stats64_t;
+
 /* HW Resource Manager Specification 1.2.0 */
 #define HWRM_VERSION_MAJOR	1
 #define HWRM_VERSION_MINOR	2
@@ -63,6 +92,7 @@
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
 /* Return Codes */
@@ -1897,6 +1927,83 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_clr_stats */
+/* Description: This command clears statistics of a context. */
+
+/* Input (24 bytes) */
+struct hwrm_stat_ctx_clr_stats_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* ID of the statistics context that is being queried. */
+	uint32_t stat_ctx_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_stat_ctx_clr_stats_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 13/38] bnxt: add initial Tx code implementation
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (11 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 12/38] bnxt: add statistics operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 14/38] bnxt: add initial Rx " Stephen Hurd
                       ` (25 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Initial implementation of tx_pkt_burst for transmit.
bnxt_xmit_pkts() is the top level function that is called during Tx.

bnxt_handle_tx_cp() is used to check and process the Tx completions
generated for the Tx Buffer Descriptors sent by the hardware.

This patch also adds code to allocate rings in the hardware.
For each Tx queue allocated in the PMD driver, a corresponding ring
in hardware will be created. Every time a Tx request is initiated
via the bnxt_xmit_pkts() call, a Buffer Descriptor is created and
is sent to the hardware via the associated Tx ring.

On completing the Tx operation, the hardware will generates the status
in the form of a completion. This completion is processed by the
bnxt_handle_tx_cp() function.

Functions like bnxt_init_tx_ring_struct() and bnxt_init_one_tx_ring()
are used to initialize various members of the structure before
starting Tx operations.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Expand the patch description
Reorder footer
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_cpr.h            |   4 +-
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +-
 drivers/net/bnxt/bnxt_ring.c           | 145 ++++++++++
 drivers/net/bnxt/bnxt_ring.h           |   8 +
 drivers/net/bnxt/bnxt_txq.c            |  42 ++-
 drivers/net/bnxt/bnxt_txr.c            | 314 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_txr.h            |  71 +++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 512 +++++++++++++++++++++++++++++++++
 9 files changed, 1091 insertions(+), 9 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_txr.c
 create mode 100644 drivers/net/bnxt/bnxt_txr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index f6a04f8..0785681 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -56,6 +56,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_vnic.c
 
 #
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index 2be22de..4bd5606 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -51,11 +51,11 @@
 
 #define B_CP_DB_REARM(cpr, raw_cons)					\
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_REARM_FLAGS | \
-				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
 #define B_CP_DIS_DB(cpr, raw_cons)					\
 		(*(uint32_t *)((cpr)->cp_doorbell) = (DB_CP_FLAGS |	\
-				RING_CMP(&cpr->cp_ring_struct, raw_cons)))
+				RING_CMP(cpr->cp_ring_struct, raw_cons)))
 
 struct bnxt_ring;
 struct bnxt_cp_ring_info {
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3453509..4ace543 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -44,6 +44,7 @@
 #include "bnxt_rxq.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -269,7 +270,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	}
 	eth_dev->dev_ops = &bnxt_dev_ops;
 	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
-	/* eth_dev->tx_pkt_burst = &bnxt_xmit_pkts; */
+	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
 
 	rc = bnxt_alloc_hwrm_resources(bp);
 	if (rc) {
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index b57d3ba..2645dda 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -31,8 +31,14 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_memzone.h>
+
 #include "bnxt.h"
+#include "bnxt_cpr.h"
 #include "bnxt_ring.h"
+#include "bnxt_txr.h"
+
+#include "hsi_struct_def_dpdk.h"
 
 /*
  * Generic ring handling
@@ -45,3 +51,142 @@ void bnxt_free_ring(struct bnxt_ring *ring)
 		*ring->vmem = NULL;
 	}
 }
+
+/*
+ * Allocates a completion ring with vmem and stats optionally also allocating
+ * a TX and/or RX ring.  Passing NULL as tx_ring_info and/or rx_ring_info
+ * to not allocate them.
+ *
+ * Order in the allocation is:
+ * stats - Always non-zero length
+ * cp vmem - Always zero-length, supported for the bnxt_ring abstraction
+ * tx vmem - Only non-zero length if tx_ring_info is not NULL
+ * rx vmem - Only non-zero length if rx_ring_info is not NULL
+ * cp bd ring - Always non-zero length
+ * tx bd ring - Only non-zero length if tx_ring_info is not NULL
+ * rx bd ring - Only non-zero length if rx_ring_info is not NULL
+ */
+int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
+			    struct bnxt_tx_ring_info *tx_ring_info,
+			    struct bnxt_rx_ring_info *rx_ring_info,
+			    struct bnxt_cp_ring_info *cp_ring_info,
+			    const char *suffix)
+{
+	struct bnxt_ring *cp_ring = cp_ring_info->cp_ring_struct;
+	struct bnxt_ring *tx_ring;
+	/* TODO: RX ring */
+	/* struct bnxt_ring *rx_ring; */
+	struct rte_pci_device *pdev = bp->pdev;
+	const struct rte_memzone *mz = NULL;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+
+	int stats_len = (tx_ring_info || rx_ring_info) ?
+	    RTE_CACHE_LINE_ROUNDUP(sizeof(struct ctx_hw_stats64)) : 0;
+
+	int cp_vmem_start = stats_len;
+	int cp_vmem_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->vmem_size);
+
+	int tx_vmem_start = cp_vmem_start + cp_vmem_len;
+	int tx_vmem_len =
+	    tx_ring_info ? RTE_CACHE_LINE_ROUNDUP(tx_ring_info->
+						tx_ring_struct->vmem_size) : 0;
+
+	int rx_vmem_start = tx_vmem_start + tx_vmem_len;
+	/* TODO: RX ring */
+	int rx_vmem_len = 0;
+	/*
+	 * rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
+	 * rx_ring_struct->vmem_size) : 0;
+	 */
+
+	int cp_ring_start = rx_vmem_start + rx_vmem_len;
+	int cp_ring_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->ring_size *
+						 sizeof(struct cmpl_base));
+
+	int tx_ring_start = cp_ring_start + cp_ring_len;
+	int tx_ring_len = tx_ring_info ?
+	    RTE_CACHE_LINE_ROUNDUP(tx_ring_info->tx_ring_struct->ring_size *
+				   sizeof(struct tx_bd_long)) : 0;
+
+	int rx_ring_start = tx_ring_start + tx_ring_len;
+	/* TODO: RX ring */
+	int rx_ring_len = 0;
+	/*
+	 * rx_ring_info ?
+	 * RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
+	 * sizeof(struct rx_prod_pkt_bd)) : 0;
+	 */
+
+	int total_alloc_len = rx_ring_start + rx_ring_len;
+
+	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
+		 "bnxt_%04x:%02x:%02x:%02x-%04x_%s", pdev->addr.domain,
+		 pdev->addr.bus, pdev->addr.devid, pdev->addr.function, qidx,
+		 suffix);
+	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
+	mz = rte_memzone_lookup(mz_name);
+	if (!mz) {
+		mz = rte_memzone_reserve(mz_name, total_alloc_len,
+					 SOCKET_ID_ANY,
+					 RTE_MEMZONE_2MB |
+					 RTE_MEMZONE_SIZE_HINT_ONLY);
+		if (mz == NULL)
+			return -ENOMEM;
+	}
+	memset(mz->addr, 0, mz->len);
+
+	if (tx_ring_info) {
+		tx_ring = tx_ring_info->tx_ring_struct;
+
+		tx_ring->bd = ((char *)mz->addr + tx_ring_start);
+		tx_ring_info->tx_desc_ring = (struct tx_bd_long *)tx_ring->bd;
+		tx_ring->bd_dma = mz->phys_addr + tx_ring_start;
+		tx_ring_info->tx_desc_mapping = tx_ring->bd_dma;
+
+		if (!tx_ring->bd)
+			return -ENOMEM;
+		if (tx_ring->vmem_size) {
+			tx_ring->vmem =
+			    (void **)((char *)mz->addr + tx_vmem_start);
+			tx_ring_info->tx_buf_ring =
+			    (struct bnxt_sw_tx_bd *)tx_ring->vmem;
+		}
+	}
+
+/*
+ *	if (rx_ring_info) {
+ *		rx_ring = &rx_ring_info->rx_ring_struct;
+ *
+ *		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
+ *		rx_ring_info->rx_desc_ring =
+ *		    (struct rx_prod_pkt_bd *)rx_ring->bd;
+ *		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
+ *		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+ *
+ *		if (!rx_ring->bd)
+ *			return -ENOMEM;
+ *		if (rx_ring->vmem_size) {
+ *			rx_ring->vmem =
+ *			    (void **)((char *)mz->addr + rx_vmem_start);
+ *			rx_ring_info->rx_buf_ring =
+ *			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
+ *		}
+ *	}
+ */
+
+	cp_ring->bd = ((char *)mz->addr + cp_ring_start);
+	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
+	cp_ring_info->cp_desc_ring = cp_ring->bd;
+	cp_ring_info->cp_desc_mapping = cp_ring->bd_dma;
+
+	if (!cp_ring->bd)
+		return -ENOMEM;
+	if (cp_ring->vmem_size)
+		*cp_ring->vmem = ((char *)mz->addr + stats_len);
+	if (stats_len) {
+		cp_ring_info->hw_stats = mz->addr;
+		cp_ring_info->hw_stats_map = mz->phys_addr;
+	}
+	cp_ring_info->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index 3abfb04..1e9d3be 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -87,6 +87,14 @@ struct bnxt_ring_grp_info {
 };
 
 struct bnxt;
+struct bnxt_tx_ring_info;
+struct bnxt_rx_ring_info;
+struct bnxt_cp_ring_info;
 void bnxt_free_ring(struct bnxt_ring *ring);
+int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
+			    struct bnxt_tx_ring_info *tx_ring_info,
+			    struct bnxt_rx_ring_info *rx_ring_info,
+			    struct bnxt_cp_ring_info *cp_ring_info,
+			    const char *suffix);
 
 #endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 8793c9d..460f71a 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -39,6 +39,7 @@
 #include "bnxt_cpr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 
 /*
  * TX Queues
@@ -54,9 +55,20 @@ void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
 		cpr->hw_stats = NULL;
 }
 
-static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq __rte_unused)
+static void bnxt_tx_queue_release_mbufs(struct bnxt_tx_queue *txq)
 {
-	/* TODO: Requires interaction with TX ring */
+	struct bnxt_sw_tx_bd *sw_ring;
+	uint16_t i;
+
+	sw_ring = txq->tx_ring->tx_buf_ring;
+	if (sw_ring) {
+		for (i = 0; i < txq->tx_ring->tx_ring_struct->ring_size; i++) {
+			if (sw_ring[i].mbuf) {
+				rte_pktmbuf_free(sw_ring[i].mbuf);
+				sw_ring[i].mbuf = NULL;
+			}
+		}
+	}
 }
 
 void bnxt_free_tx_mbufs(struct bnxt *bp)
@@ -75,7 +87,15 @@ void bnxt_tx_queue_release_op(void *tx_queue)
 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
 
 	if (txq) {
-		/* TODO: Free ring and stats here */
+		/* Free TX ring hardware descriptors */
+		bnxt_tx_queue_release_mbufs(txq);
+		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
+
+		/* Free TX completion ring hardware descriptors */
+		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
+
+		bnxt_free_txq_stats(txq);
+
 		rte_free(txq);
 	}
 }
@@ -111,14 +131,24 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	txq->nb_tx_desc = nb_desc;
 	txq->tx_free_thresh = tx_conf->tx_free_thresh;
 
-	/* TODO: Initialize ring structure */
+	bnxt_init_tx_ring_struct(txq);
 
 	txq->queue_id = queue_idx;
 	txq->port_id = eth_dev->data->port_id;
 
-	/* TODO: Allocate TX ring hardware descriptors */
+	/* Allocate TX ring hardware descriptors */
+	if (bnxt_alloc_rings(bp, queue_idx, txq->tx_ring, NULL, txq->cp_ring,
+			"txr")) {
+		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for tx_ring failed!");
+		bnxt_tx_queue_release_op(txq);
+		return -ENOMEM;
+	}
 
-	/* TODO: Initialize the ring */
+	if (bnxt_init_one_tx_ring(txq)) {
+		RTE_LOG(ERR, PMD, "bnxt_init_one_tx_ring failed!");
+		bnxt_tx_queue_release_op(txq);
+		return -ENOMEM;
+	}
 
 	eth_dev->data->tx_queues[queue_idx] = txq;
 	return 0;
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
new file mode 100644
index 0000000..9064e8d
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -0,0 +1,314 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+
+#include <rte_byteorder.h>
+#include <rte_malloc.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_ring.h"
+#include "bnxt_txq.h"
+#include "bnxt_txr.h"
+#include "hsi_struct_def_dpdk.h"
+#include <stdbool.h>
+
+/*
+ * TX Ring handling
+ */
+
+void bnxt_free_tx_rings(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->tx_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+
+		if (!txq)
+			continue;
+
+		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
+		/* TODO: free() txq->tx_ring and txq->tx_ring->tx_ring_struct */
+		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
+		/* TODO: free() txq->cp_ring and txq->cp_ring->cp_ring_struct */
+
+		rte_free(txq);
+		bp->tx_queues[i] = NULL;
+	}
+}
+
+int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	struct bnxt_ring *ring = txr->tx_ring_struct;
+
+	txq->tx_wake_thresh = ring->ring_size / 2;
+	ring->fw_ring_id = INVALID_HW_RING_ID;
+
+	return 0;
+}
+
+void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_tx_ring_info *txr;
+	struct bnxt_ring *ring;
+
+	/* TODO: These need to be allocated */
+	txr = txq->tx_ring;
+	ring = txr->tx_ring_struct;
+	ring->ring_size = rte_align32pow2(txq->nb_tx_desc + 1);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)txr->tx_desc_ring;
+	ring->bd_dma = txr->tx_desc_mapping;
+	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_tx_bd);
+	ring->vmem = (void **)&txr->tx_buf_ring;
+
+	/* TODO: These need to be allocated */
+	cpr = txq->cp_ring;
+	ring = cpr->cp_ring_struct;
+	ring->ring_size = txr->tx_ring_struct->ring_size;
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
+
+static inline uint32_t bnxt_tx_avail(struct bnxt_tx_ring_info *txr)
+{
+	/* Tell compiler to fetch tx indices from memory. */
+	rte_compiler_barrier();
+
+	return txr->tx_ring_struct->ring_size -
+		((txr->tx_prod - txr->tx_cons) &
+			txr->tx_ring_struct->ring_mask) - 1;
+}
+
+static uint16_t bnxt_start_xmit(struct rte_mbuf *tx_pkt,
+				struct bnxt_tx_queue *txq)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	struct tx_bd_long *txbd;
+	struct tx_bd_long_hi *txbd1;
+	uint32_t vlan_tag_flags, cfa_action;
+	bool long_bd = false;
+	uint16_t last_prod = 0;
+	struct rte_mbuf *m_seg;
+	struct bnxt_sw_tx_bd *tx_buf;
+	static const uint32_t lhint_arr[4] = {
+		TX_BD_LONG_FLAGS_LHINT_LT512,
+		TX_BD_LONG_FLAGS_LHINT_LT1K,
+		TX_BD_LONG_FLAGS_LHINT_LT2K,
+		TX_BD_LONG_FLAGS_LHINT_LT2K
+	};
+
+	if (tx_pkt->ol_flags & (PKT_TX_TCP_SEG | PKT_TX_TCP_CKSUM |
+				PKT_TX_UDP_CKSUM | PKT_TX_IP_CKSUM |
+				PKT_TX_VLAN_PKT))
+		long_bd = true;
+
+	tx_buf = &txr->tx_buf_ring[txr->tx_prod];
+	tx_buf->mbuf = tx_pkt;
+	tx_buf->nr_bds = long_bd + tx_pkt->nb_segs;
+	last_prod = (txr->tx_prod + tx_buf->nr_bds - 1) &
+				txr->tx_ring_struct->ring_mask;
+
+	if (unlikely(bnxt_tx_avail(txr) < tx_buf->nr_bds))
+		return -ENOMEM;
+
+	txbd = &txr->tx_desc_ring[txr->tx_prod];
+	txbd->opaque = txr->tx_prod;
+	txbd->flags_type = tx_buf->nr_bds << TX_BD_LONG_FLAGS_BD_CNT_SFT;
+	txbd->len = tx_pkt->data_len;
+	if (txbd->len >= 2014)
+		txbd->flags_type |= TX_BD_LONG_FLAGS_LHINT_GTE2K;
+	else
+		txbd->flags_type |= lhint_arr[txbd->len >> 9];
+	txbd->addr = rte_cpu_to_le_32(RTE_MBUF_DATA_DMA_ADDR(tx_buf->mbuf));
+
+	if (long_bd) {
+		txbd->flags_type |= TX_BD_LONG_TYPE_TX_BD_LONG;
+		vlan_tag_flags = 0;
+		cfa_action = 0;
+		if (tx_buf->mbuf->ol_flags & PKT_TX_VLAN_PKT) {
+			/* shurd: Should this mask at
+			 * TX_BD_LONG_CFA_META_VLAN_VID_MASK?
+			 */
+			vlan_tag_flags = TX_BD_LONG_CFA_META_KEY_VLAN_TAG |
+				tx_buf->mbuf->vlan_tci;
+			/* Currently supports 8021Q, 8021AD vlan offloads
+			 * QINQ1, QINQ2, QINQ3 vlan headers are deprecated
+			 */
+			/* DPDK only supports 802.11q VLAN packets */
+			vlan_tag_flags |=
+					TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
+		}
+
+		txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+
+		txbd1 = (struct tx_bd_long_hi *)
+					&txr->tx_desc_ring[txr->tx_prod];
+		txbd1->lflags = 0;
+		txbd1->cfa_meta = vlan_tag_flags;
+		txbd1->cfa_action = cfa_action;
+
+		if (tx_pkt->ol_flags & PKT_TX_TCP_SEG) {
+			/* TSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_LSO;
+			txbd1->hdr_size = tx_pkt->l2_len + tx_pkt->l3_len +
+					tx_pkt->l4_len;
+			txbd1->mss = tx_pkt->tso_segsz;
+
+		} else if (tx_pkt->ol_flags & (PKT_TX_TCP_CKSUM |
+					PKT_TX_UDP_CKSUM)) {
+			/* TCP/UDP CSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_TCP_UDP_CHKSUM;
+			txbd1->mss = 0;
+
+		} else if (tx_pkt->ol_flags & PKT_TX_IP_CKSUM) {
+			/* IP CSO */
+			txbd1->lflags = TX_BD_LONG_LFLAGS_IP_CHKSUM;
+			txbd1->mss = 0;
+		}
+	} else {
+		txbd->flags_type |= TX_BD_SHORT_TYPE_TX_BD_SHORT;
+	}
+
+	m_seg = tx_pkt->next;
+	/* i is set at the end of the if(long_bd) block */
+	while (txr->tx_prod != last_prod) {
+		txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+		tx_buf = &txr->tx_buf_ring[txr->tx_prod];
+
+		txbd = &txr->tx_desc_ring[txr->tx_prod];
+		txbd->addr = rte_cpu_to_le_32(RTE_MBUF_DATA_DMA_ADDR(m_seg));
+		txbd->flags_type = TX_BD_SHORT_TYPE_TX_BD_SHORT;
+		txbd->len = m_seg->data_len;
+
+		m_seg = m_seg->next;
+	}
+
+	txbd->flags_type |= TX_BD_LONG_FLAGS_PACKET_END;
+
+	txr->tx_prod = RING_NEXT(txr->tx_ring_struct, txr->tx_prod);
+
+	return 0;
+}
+
+static void bnxt_tx_cmp(struct bnxt_tx_queue *txq, int nr_pkts)
+{
+	struct bnxt_tx_ring_info *txr = txq->tx_ring;
+	uint16_t cons = txr->tx_cons;
+	int i, j;
+
+	for (i = 0; i < nr_pkts; i++) {
+		struct bnxt_sw_tx_bd *tx_buf;
+		struct rte_mbuf *mbuf;
+
+		tx_buf = &txr->tx_buf_ring[cons];
+		cons = RING_NEXT(txr->tx_ring_struct, cons);
+		mbuf = tx_buf->mbuf;
+		tx_buf->mbuf = NULL;
+
+		/* EW - no need to unmap DMA memory? */
+
+		for (j = 1; j < tx_buf->nr_bds; j++)
+			cons = RING_NEXT(txr->tx_ring_struct, cons);
+		rte_pktmbuf_free(mbuf);
+	}
+
+	txr->tx_cons = cons;
+}
+
+static int bnxt_handle_tx_cp(struct bnxt_tx_queue *txq)
+{
+	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+	uint32_t raw_cons = cpr->cp_raw_cons;
+	uint32_t cons;
+	int nb_tx_pkts = 0;
+	struct tx_cmpl *txcmp;
+
+	if ((txq->tx_ring->tx_ring_struct->ring_size -
+			(bnxt_tx_avail(txq->tx_ring))) >
+			txq->tx_free_thresh) {
+		while (1) {
+			cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
+			txcmp = (struct tx_cmpl *)&cpr->cp_desc_ring[cons];
+
+			if (!CMP_VALID(txcmp, raw_cons, cpr->cp_ring_struct))
+				break;
+
+			if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
+				nb_tx_pkts++;
+			else
+				RTE_LOG(DEBUG, PMD,
+						"Unhandled CMP type %02x\n",
+						CMP_TYPE(txcmp));
+			raw_cons = NEXT_RAW_CMP(raw_cons);
+		}
+		if (nb_tx_pkts)
+			bnxt_tx_cmp(txq, nb_tx_pkts);
+		cpr->cp_raw_cons = raw_cons;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+	}
+	return nb_tx_pkts;
+}
+
+uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct bnxt_tx_queue *txq = tx_queue;
+	uint16_t nb_tx_pkts = 0;
+	uint16_t db_mask = txq->tx_ring->tx_ring_struct->ring_size >> 2;
+	uint16_t last_db_mask = 0;
+
+	/* Handle TX completions */
+	bnxt_handle_tx_cp(txq);
+
+	/* Handle TX burst request */
+	for (nb_tx_pkts = 0; nb_tx_pkts < nb_pkts; nb_tx_pkts++) {
+		if (bnxt_start_xmit(tx_pkts[nb_tx_pkts], txq)) {
+			break;
+		} else if ((nb_tx_pkts & db_mask) != last_db_mask) {
+			B_TX_DB(txq->tx_ring->tx_doorbell,
+					txq->tx_ring->tx_prod);
+			last_db_mask = nb_tx_pkts & db_mask;
+		}
+	}
+	if (nb_tx_pkts)
+		B_TX_DB(txq->tx_ring->tx_doorbell, txq->tx_ring->tx_prod);
+
+	return nb_tx_pkts;
+}
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
new file mode 100644
index 0000000..712ce65
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -0,0 +1,71 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_TXR_H_
+#define _BNXT_TXR_H_
+
+#define MAX_TX_RINGS	16
+#define BNXT_TX_PUSH_THRESH 92
+
+#define B_TX_DB(db, prod)						\
+		(*(uint32_t *)db = (DB_KEY_TX | prod))
+
+struct bnxt_tx_ring_info {
+	uint16_t		tx_prod;
+	uint16_t		tx_cons;
+	void			*tx_doorbell;
+
+	struct tx_bd_long	*tx_desc_ring;
+	struct bnxt_sw_tx_bd	*tx_buf_ring;
+
+	phys_addr_t		tx_desc_mapping;
+
+#define BNXT_DEV_STATE_CLOSING	0x1
+	uint32_t		dev_state;
+
+	struct bnxt_ring	*tx_ring_struct;
+};
+
+struct bnxt_sw_tx_bd {
+	struct rte_mbuf		*mbuf; /* mbuf associated with TX descriptor */
+	uint8_t			is_gso;
+	unsigned short		nr_bds;
+};
+
+void bnxt_free_tx_rings(struct bnxt *bp);
+int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
+void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq);
+uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 91a83d1..1543f20 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -99,6 +99,518 @@ struct ctx_hw_stats64 {
 #define HWRM_ERR_CODE_INVALID_PARAMS                      (UINT32_C(0x2))
 #define HWRM_ERR_CODE_RESOURCE_ACCESS_DENIED              (UINT32_C(0x3))
 
+/* Short TX BD (16 bytes) */
+struct tx_bd_short {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	/* This value identifies the type of buffer descriptor. */
+	#define TX_BD_SHORT_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_BD_SHORT_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 16B long and is used for normal L2
+		 * packet transmission.
+		 */
+	#define TX_BD_SHORT_TYPE_TX_BD_SHORT		(UINT32_C(0x0) << 0)
+	/*
+	 * If set to 1, the packet ends with the data in the buffer pointed to
+	 * by this descriptor. This flag must be valid on every BD.
+	 */
+	#define TX_BD_SHORT_FLAGS_PACKET_END		UINT32_C(0x40)
+	/*
+	 * If set to 1, the device will not generate a completion for this
+	 * transmit packet unless there is an error in it's processing. If this
+	 * bit is set to 0, then the packet will be completed normally. This bit
+	 * must be valid only on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_NO_CMPL		UINT32_C(0x80)
+	/*
+	 * This value indicates how many 16B BD locations are consumed in the
+	 * ring by this packet. A value of 1 indicates that this BD is the only
+	 * BD (and that the it is a short BD). A value of 3 indicates either 3
+	 * short BDs or 1 long BD and one short BD in the packet. A value of 0
+	 * indicates that there are 32 BD locations in the packet (the maximum).
+	 * This field is valid only on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_BD_CNT_MASK		UINT32_C(0x1f00)
+	#define TX_BD_SHORT_FLAGS_BD_CNT_SFT		8
+	/*
+	 * This value is a hint for the length of the entire packet. It is used
+	 * by the chip to optimize internal processing. The packet will be
+	 * dropped if the hint is too short. This field is valid only on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_LHINT_MASK		UINT32_C(0x6000)
+	#define TX_BD_SHORT_FLAGS_LHINT_SFT		13
+		/* indicates packet length < 512B */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT512		(UINT32_C(0x0) << 13)
+		/* indicates 512 <= packet length < 1KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT1K		(UINT32_C(0x1) << 13)
+		/* indicates 1KB <= packet length < 2KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_LT2K		(UINT32_C(0x2) << 13)
+		/* indicates packet length >= 2KB */
+	#define TX_BD_SHORT_FLAGS_LHINT_GTE2K		(UINT32_C(0x3) << 13)
+	#define TX_BD_SHORT_FLAGS_LHINT_LAST	TX_BD_SHORT_FLAGS_LHINT_GTE2K
+	/*
+	 * If set to 1, the device immediately updates the Send Consumer Index
+	 * after the buffer associated with this descriptor has been transferred
+	 * via DMA to NIC memory from host memory. An interrupt may or may not
+	 * be generated according to the state of the interrupt avoidance
+	 * mechanisms. If this bit is set to 0, then the Consumer Index is only
+	 * updated as soon as one of the host interrupt coalescing conditions
+	 * has been met. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_COAL_NOW		UINT32_C(0x8000)
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	#define TX_BD_SHORT_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_BD_SHORT_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the host physical buffer this BD describes in
+	 * bytes. This field must be valid on all BDs of a packet.
+	 */
+	uint16_t len;
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with the
+	 * transmit BD. This field must be valid on the first BD of a packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address for the portion of the packet
+	 * described by this TX BD. This value must be valid on all BDs of a
+	 * packet.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
+/* Long TX BD (32 bytes split to 2 16-byte struct) */
+struct tx_bd_long {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	/* This value identifies the type of buffer descriptor. */
+	#define TX_BD_LONG_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_BD_LONG_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 32B long and is used for normal L2
+		 * packet transmission.
+		 */
+	#define TX_BD_LONG_TYPE_TX_BD_LONG		(UINT32_C(0x10) << 0)
+	/*
+	 * If set to 1, the packet ends with the data in the buffer pointed to
+	 * by this descriptor. This flag must be valid on every BD.
+	 */
+	#define TX_BD_LONG_FLAGS_PACKET_END		UINT32_C(0x40)
+	/*
+	 * If set to 1, the device will not generate a completion for this
+	 * transmit packet unless there is an error in it's processing. If this
+	 * bit is set to 0, then the packet will be completed normally. This bit
+	 * must be valid only on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_NO_CMPL		UINT32_C(0x80)
+	/*
+	 * This value indicates how many 16B BD locations are consumed in the
+	 * ring by this packet. A value of 1 indicates that this BD is the only
+	 * BD (and that the it is a short BD). A value of 3 indicates either 3
+	 * short BDs or 1 long BD and one short BD in the packet. A value of 0
+	 * indicates that there are 32 BD locations in the packet (the maximum).
+	 * This field is valid only on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_BD_CNT_MASK		UINT32_C(0x1f00)
+	#define TX_BD_LONG_FLAGS_BD_CNT_SFT		8
+	/*
+	 * This value is a hint for the length of the entire packet. It is used
+	 * by the chip to optimize internal processing. The packet will be
+	 * dropped if the hint is too short. This field is valid only on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_LHINT_MASK		UINT32_C(0x6000)
+	#define TX_BD_LONG_FLAGS_LHINT_SFT		13
+		/* indicates packet length < 512B */
+	#define TX_BD_LONG_FLAGS_LHINT_LT512		(UINT32_C(0x0) << 13)
+		/* indicates 512 <= packet length < 1KB */
+	#define TX_BD_LONG_FLAGS_LHINT_LT1K		(UINT32_C(0x1) << 13)
+		/* indicates 1KB <= packet length < 2KB */
+	#define TX_BD_LONG_FLAGS_LHINT_LT2K		(UINT32_C(0x2) << 13)
+		/* indicates packet length >= 2KB */
+	#define TX_BD_LONG_FLAGS_LHINT_GTE2K		(UINT32_C(0x3) << 13)
+	#define TX_BD_LONG_FLAGS_LHINT_LAST	TX_BD_LONG_FLAGS_LHINT_GTE2K
+	/*
+	 * If set to 1, the device immediately updates the Send Consumer Index
+	 * after the buffer associated with this descriptor has been transferred
+	 * via DMA to NIC memory from host memory. An interrupt may or may not
+	 * be generated according to the state of the interrupt avoidance
+	 * mechanisms. If this bit is set to 0, then the Consumer Index is only
+	 * updated as soon as one of the host interrupt coalescing conditions
+	 * has been met. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_FLAGS_COAL_NOW		UINT32_C(0x8000)
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Only the packet_end bit must be valid for the remaining BDs of a
+	 * packet.
+	 */
+	#define TX_BD_LONG_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_BD_LONG_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the host physical buffer this BD describes in
+	 * bytes. This field must be valid on all BDs of a packet.
+	 */
+	uint16_t len;
+
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with the
+	 * transmit BD. This field must be valid on the first BD of a packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address for the portion of the packet
+	 * described by this TX BD. This value must be valid on all BDs of a
+	 * packet.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
+/* last 16 bytes of Long TX BD */
+
+struct tx_bd_long_hi {
+	/*
+	 * All bits in this field must be valid on the first BD of a packet.
+	 * Their value on other BDs of the packet will be ignored.
+	 */
+	/*
+	 * If set to 1, the controller replaces the TCP/UPD checksum fields of
+	 * normal TCP/UPD checksum, or the inner TCP/UDP checksum field of the
+	 * encapsulated TCP/UDP packets with the hardware calculated TCP/UDP
+	 * checksum for the packet associated with this descriptor. This bit
+	 * must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_TCP_UDP_CHKSUM	UINT32_C(0x1)
+	/*
+	 * If set to 1, the controller replaces the IP checksum of the normal
+	 * packets, or the inner IP checksum of the encapsulated packets with
+	 * the hardware calculated IP checksum for the packet associated with
+	 * this descriptor. This bit must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_IP_CHKSUM		UINT32_C(0x2)
+	/*
+	 * If set to 1, the controller will not append an Ethernet CRC to the
+	 * end of the frame. This bit must be valid on the first BD of a packet.
+	 * Packet must be 64B or longer when this flag is set. It is not useful
+	 * to use this bit with any form of TX offload such as CSO or LSO. The
+	 * intent is that the packet from the host already has a valid Ethernet
+	 * CRC on the packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_NOCRC			UINT32_C(0x4)
+	/*
+	 * If set to 1, the device will record the time at which the packet was
+	 * actually transmitted at the TX MAC. This bit must be valid on the
+	 * first BD of a packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_STAMP			UINT32_C(0x8)
+	/*
+	 * If set to 1, The controller replaces the tunnel IP checksum field
+	 * with hardware calculated IP checksum for the IP header of the packet
+	 * associated with this descriptor. In case of VXLAN, the controller
+	 * also replaces the outer header UDP checksum with hardware calculated
+	 * UDP checksum for the packet associated with this descriptor.
+	 */
+	#define TX_BD_LONG_LFLAGS_T_IP_CHKSUM		UINT32_C(0x10)
+	/*
+	 * If set to 1, the device will treat this packet with LSO(Large Send
+	 * Offload) processing for both normal or encapsulated packets, which is
+	 * a form of TCP segmentation. When this bit is 1, the hdr_size and mss
+	 * fields must be valid. The driver doesn't need to set t_ip_chksum,
+	 * ip_chksum, and tcp_udp_chksum flags since the controller will replace
+	 * the appropriate checksum fields for segmented packets. When this bit
+	 * is 1, the hdr_size and mss fields must be valid.
+	 */
+	#define TX_BD_LONG_LFLAGS_LSO			UINT32_C(0x20)
+	/*
+	 * If set to zero when LSO is '1', then the IPID will be treated as a
+	 * 16b number and will be wrapped if it exceeds a value of 0xffff. If
+	 * set to one when LSO is '1', then the IPID will be treated as a 15b
+	 * number and will be wrapped if it exceeds a value 0f 0x7fff.
+	 */
+	#define TX_BD_LONG_LFLAGS_IPID_FMT		UINT32_C(0x40)
+	/*
+	 * If set to zero when LSO is '1', then the IPID of the tunnel IP header
+	 * will not be modified during LSO operations. If set to one when LSO is
+	 * '1', then the IPID of the tunnel IP header will be incremented for
+	 * each subsequent segment of an LSO operation.
+	 */
+	#define TX_BD_LONG_LFLAGS_T_IPID		UINT32_C(0x80)
+	/*
+	 * If set to '1', then the RoCE ICRC will be appended to the packet.
+	 * Packet must be a valid RoCE format packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_ROCE_CRC		UINT32_C(0x100)
+	/*
+	 * If set to '1', then the FCoE CRC will be appended to the packet.
+	 * Packet must be a valid FCoE format packet.
+	 */
+	#define TX_BD_LONG_LFLAGS_FCOE_CRC		UINT32_C(0x200)
+	uint16_t lflags;
+
+	/*
+	 * When LSO is '1', this field must contain the offset of the TCP
+	 * payload from the beginning of the packet in as 16b words. In case of
+	 * encapsulated/tunneling packet, this field contains the offset of the
+	 * inner TCP payload from beginning of the packet as 16-bit words. This
+	 * value must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_HDR_SIZE_MASK		UINT32_C(0x1ff)
+	#define TX_BD_LONG_HDR_SIZE_SFT			0
+	uint16_t hdr_size;
+
+	/*
+	 * This is the MSS value that will be used to do the LSO processing. The
+	 * value is the length in bytes of the TCP payload for each segment
+	 * generated by the LSO operation. This value must be valid on the first
+	 * BD of a packet.
+	 */
+	#define TX_BD_LONG_MSS_MASK			UINT32_C(0x7fff)
+	#define TX_BD_LONG_MSS_SFT			0
+	uint32_t mss;
+
+	uint16_t unused_2;
+
+	/*
+	 * This value selects a CFA action to perform on the packet. Set this
+	 * value to zero if no CFA action is desired. This value must be valid
+	 * on the first BD of a packet.
+	 */
+	uint16_t cfa_action;
+
+	/*
+	 * This value is action meta-data that defines CFA edit operations that
+	 * are done in addition to any action editing.
+	 */
+	/* When key=1, This is the VLAN tag VID value. */
+	#define TX_BD_LONG_CFA_META_VLAN_VID_MASK	UINT32_C(0xfff)
+	#define TX_BD_LONG_CFA_META_VLAN_VID_SFT	0
+	/* When key=1, This is the VLAN tag DE value. */
+	#define TX_BD_LONG_CFA_META_VLAN_DE		UINT32_C(0x1000)
+	/* When key=1, This is the VLAN tag PRI value. */
+	#define TX_BD_LONG_CFA_META_VLAN_PRI_MASK	UINT32_C(0xe000)
+	#define TX_BD_LONG_CFA_META_VLAN_PRI_SFT	13
+	/* When key=1, This is the VLAN tag TPID select value. */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_MASK	UINT32_C(0x70000)
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_SFT	16
+		/* 0x88a8 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8	(UINT32_C(0x0) << 16)
+		/* 0x8100 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100	(UINT32_C(0x1) << 16)
+		/* 0x9100 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100	(UINT32_C(0x2) << 16)
+		/* 0x9200 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200	(UINT32_C(0x3) << 16)
+		/* 0x9300 */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300	(UINT32_C(0x4) << 16)
+		/* Value programmed in CFA VLANTPID register. */
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_TPIDCFG	(UINT32_C(0x5) << 16)
+	#define TX_BD_LONG_CFA_META_VLAN_TPID_LAST \
+					TX_BD_LONG_CFA_META_VLAN_TPID_TPIDCFG
+	/* When key=1, This is the VLAN tag TPID select value. */
+	#define TX_BD_LONG_CFA_META_VLAN_RESERVED_MASK	UINT32_C(0xff80000)
+	#define TX_BD_LONG_CFA_META_VLAN_RESERVED_SFT	19
+	/*
+	 * This field identifies the type of edit to be performed on the packet.
+	 * This value must be valid on the first BD of a packet.
+	 */
+	#define TX_BD_LONG_CFA_META_KEY_MASK		UINT32_C(0xf0000000)
+	#define TX_BD_LONG_CFA_META_KEY_SFT		28
+		/* No editing */
+	#define TX_BD_LONG_CFA_META_KEY_NONE		(UINT32_C(0x0) << 28)
+		/*
+		 * - meta[17:16] - TPID select value (0 = 0x8100). - meta[15:12]
+		 * - PRI/DE value. - meta[11:0] - VID value.
+		 */
+	#define TX_BD_LONG_CFA_META_KEY_VLAN_TAG	(UINT32_C(0x1) << 28)
+	#define TX_BD_LONG_CFA_META_KEY_LAST	TX_BD_LONG_CFA_META_KEY_VLAN_TAG
+	uint32_t cfa_meta;
+} __attribute__((packed));
+
+/* Completion Ring Structures */
+/* Note: This structure is used by the HWRM to communicate HWRM Error. */
+/* Base Completion Record (16 bytes) */
+struct cmpl_base {
+	/* unused is 10 b */
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define CMPL_BASE_TYPE_MASK			UINT32_C(0x3f)
+	#define CMPL_BASE_TYPE_SFT			0
+		/* TX L2 completion: Completion of TX packet. Length = 16B */
+	#define CMPL_BASE_TYPE_TX_L2			(UINT32_C(0x0) << 0)
+		/*
+		 * RX L2 completion: Completion of and L2 RX packet.
+		 * Length = 32B
+		*/
+	#define CMPL_BASE_TYPE_RX_L2			(UINT32_C(0x11) << 0)
+		/*
+		 * RX Aggregation Buffer completion : Completion of an L2
+		 * aggregation buffer in support of TPA, HDS, or Jumbo packet
+		 * completion. Length = 16B
+		 */
+	#define CMPL_BASE_TYPE_RX_AGG			(UINT32_C(0x12) << 0)
+		/*
+		 * RX L2 TPA Start Completion: Completion at the beginning of a
+		 * TPA operation. Length = 32B
+		 */
+	#define CMPL_BASE_TYPE_RX_TPA_START		(UINT32_C(0x13) << 0)
+		/*
+		 * RX L2 TPA End Completion: Completion at the end of a TPA
+		 * operation. Length = 32B
+		 */
+	#define CMPL_BASE_TYPE_RX_TPA_END		(UINT32_C(0x15) << 0)
+		/*
+		 * Statistics Ejection Completion: Completion of statistics data
+		 * ejection buffer. Length = 16B
+		 */
+	#define CMPL_BASE_TYPE_STAT_EJECT		(UINT32_C(0x1a) << 0)
+		/* HWRM Command Completion: Completion of an HWRM command. */
+	#define CMPL_BASE_TYPE_HWRM_DONE		(UINT32_C(0x20) << 0)
+		/* Forwarded HWRM Request */
+	#define CMPL_BASE_TYPE_HWRM_FWD_REQ		(UINT32_C(0x22) << 0)
+		/* Forwarded HWRM Response */
+	#define CMPL_BASE_TYPE_HWRM_FWD_RESP		(UINT32_C(0x24) << 0)
+		/* HWRM Asynchronous Event Information */
+	#define CMPL_BASE_TYPE_HWRM_ASYNC_EVENT		(UINT32_C(0x2e) << 0)
+		/* CQ Notification */
+	#define CMPL_BASE_TYPE_CQ_NOTIFICATION		(UINT32_C(0x30) << 0)
+		/* SRQ Threshold Event */
+	#define CMPL_BASE_TYPE_SRQ_EVENT		(UINT32_C(0x32) << 0)
+		/* DBQ Threshold Event */
+	#define CMPL_BASE_TYPE_DBQ_EVENT		(UINT32_C(0x34) << 0)
+		/* QP Async Notification */
+	#define CMPL_BASE_TYPE_QP_EVENT			(UINT32_C(0x38) << 0)
+		/* Function Async Notification */
+	#define CMPL_BASE_TYPE_FUNC_EVENT		(UINT32_C(0x3a) << 0)
+	uint16_t type;
+
+	uint16_t info1;
+	uint32_t info2;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define CMPL_BASE_V				UINT32_C(0x1)
+	/* info3 is 31 b */
+	#define CMPL_BASE_INFO3_MASK			UINT32_C(0xfffffffe)
+	#define CMPL_BASE_INFO3_SFT			1
+	uint32_t info3_v;
+
+	uint32_t info4;
+} __attribute__((packed));
+
+/* TX Completion Record (16 bytes) */
+struct tx_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define TX_CMPL_TYPE_MASK			UINT32_C(0x3f)
+	#define TX_CMPL_TYPE_SFT			0
+		/* TX L2 completion: Completion of TX packet. Length = 16B */
+	#define TX_CMPL_TYPE_TX_L2			(UINT32_C(0x0) << 0)
+	/*
+	 * When this bit is '1', it indicates a packet that has an error of some
+	 * type. Type of error is indicated in error_flags.
+	 */
+	#define TX_CMPL_FLAGS_ERROR			UINT32_C(0x40)
+	/*
+	 * When this bit is '1', it indicates that the packet completed was
+	 * transmitted using the push acceleration data provided by the driver.
+	 * When this bit is '0', it indicates that the packet had not push
+	 * acceleration data written or was executed as a normal packet even
+	 * though push data was provided.
+	 */
+	#define TX_CMPL_FLAGS_PUSH			UINT32_C(0x80)
+	#define TX_CMPL_FLAGS_MASK			UINT32_C(0xffc0)
+	#define TX_CMPL_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	uint16_t unused_0;
+
+	/*
+	 * This is a copy of the opaque field from the first TX BD of this
+	 * transmitted packet.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define TX_CMPL_V				UINT32_C(0x1)
+	/*
+	 * This error indicates that there was some sort of problem with the BDs
+	 * for the packet.
+	 */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_MASK	UINT32_C(0xe)
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_SFT		1
+		/* No error */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_NO_ERROR	(UINT32_C(0x0) << 1)
+		/* Bad Format: BDs were not formatted correctly. */
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_BAD_FMT	(UINT32_C(0x2) << 1)
+	#define TX_CMPL_ERRORS_BUFFER_ERROR_LAST \
+					TX_CMPL_ERRORS_BUFFER_ERROR_BAD_FMT
+	/*
+	 * When this bit is '1', it indicates that the length of the packet was
+	 * zero. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_ZERO_LENGTH_PKT		UINT32_C(0x10)
+	/*
+	 * When this bit is '1', it indicates that the packet was longer than
+	 * the programmed limit in TDI. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_EXCESSIVE_BD_LENGTH	UINT32_C(0x20)
+	/*
+	 * When this bit is '1', it indicates that one or more of the BDs
+	 * associated with this packet generated a PCI error. This probably
+	 * means the address was not valid.
+	 */
+	#define TX_CMPL_ERRORS_DMA_ERROR		UINT32_C(0x40)
+	/*
+	 * When this bit is '1', it indicates that the packet was longer than
+	 * indicated by the hint. No packet was transmitted.
+	 */
+	#define TX_CMPL_ERRORS_HINT_TOO_SHORT		UINT32_C(0x80)
+	/*
+	 * When this bit is '1', it indicates that the packet was dropped due to
+	 * Poison TLP error on one or more of the TLPs in the PXP completion.
+	 */
+	#define TX_CMPL_ERRORS_POISON_TLP_ERROR		UINT32_C(0x100)
+	#define TX_CMPL_ERRORS_MASK			UINT32_C(0xfffe)
+	#define TX_CMPL_ERRORS_SFT			1
+	uint16_t errors_v;
+
+	uint16_t unused_1;
+	uint32_t unused_2;
+} __attribute__((packed)) tx_cmpl_t, *ptx_cmpl_t;
+
 /* HWRM Forwarded Request (16 bytes) */
 struct hwrm_fwd_req_cmpl {
 	/* Length of forwarded request in bytes. */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 14/38] bnxt: add initial Rx code implementation
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (12 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 13/38] bnxt: add initial Tx code implementation Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings Stephen Hurd
                       ` (24 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds initial implementation of rx_pkt_burst() function for Rx.
bnxt_recv_pkts() is the top level function for doing Rx.

This patch also adds code to allocate rings in the ASIC.

For each Rx queue allocated in the PMD driver, a corresponding ring
in hardware will be created. Every time a frame is received a Rx ring
is selected based on the hardware configuration like RSS, MAC or VLAN,
COS and such. The hardware uses a completion ring to indicate the
availability of a packet.

This patch also brings in functions like bnxt_init_one_rx_ring()
bnxt_init_rx_ring_struct() which initializes various structures before
a Rx can begin.

bnxt_init_rxbds() initializes the Rx Buffer Descriptors while
bnxt_alloc_rx_data() allocates a buffer in the host to receive the
incoming packet.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Use rte_mbuf_raw_alloc instead of the now depricated
__rte_mbuf_raw_alloc and fix issues pointed out by checkpatch.

v5:
Expand the patch description
Reorder footer

v6:
Correct typos
---
 drivers/net/bnxt/Makefile              |   1 +
 drivers/net/bnxt/bnxt_ethdev.c         |   3 +-
 drivers/net/bnxt/bnxt_ring.c           |  61 ++---
 drivers/net/bnxt/bnxt_rxq.c            |  34 ++-
 drivers/net/bnxt/bnxt_rxr.c            | 341 ++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_rxr.h            |  62 +++++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 474 +++++++++++++++++++++++++++++++++
 7 files changed, 936 insertions(+), 40 deletions(-)
 create mode 100644 drivers/net/bnxt/bnxt_rxr.c
 create mode 100644 drivers/net/bnxt/bnxt_rxr.h

diff --git a/drivers/net/bnxt/Makefile b/drivers/net/bnxt/Makefile
index 0785681..4d35412 100644
--- a/drivers/net/bnxt/Makefile
+++ b/drivers/net/bnxt/Makefile
@@ -54,6 +54,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_filter.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_hwrm.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_rxr.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_stats.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txq.c
 SRCS-$(CONFIG_RTE_LIBRTE_BNXT_PMD) += bnxt_txr.c
diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 4ace543..6888363 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -42,6 +42,7 @@
 #include "bnxt.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
 #include "bnxt_txr.h"
@@ -269,7 +270,7 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 		goto error;
 	}
 	eth_dev->dev_ops = &bnxt_dev_ops;
-	/* eth_dev->rx_pkt_burst = &bnxt_recv_pkts; */
+	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
 
 	rc = bnxt_alloc_hwrm_resources(bp);
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index 2645dda..f547a9e 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -36,6 +36,7 @@
 #include "bnxt.h"
 #include "bnxt_cpr.h"
 #include "bnxt_ring.h"
+#include "bnxt_rxr.h"
 #include "bnxt_txr.h"
 
 #include "hsi_struct_def_dpdk.h"
@@ -74,8 +75,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 {
 	struct bnxt_ring *cp_ring = cp_ring_info->cp_ring_struct;
 	struct bnxt_ring *tx_ring;
-	/* TODO: RX ring */
-	/* struct bnxt_ring *rx_ring; */
+	struct bnxt_ring *rx_ring;
 	struct rte_pci_device *pdev = bp->pdev;
 	const struct rte_memzone *mz = NULL;
 	char mz_name[RTE_MEMZONE_NAMESIZE];
@@ -92,12 +92,9 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 						tx_ring_struct->vmem_size) : 0;
 
 	int rx_vmem_start = tx_vmem_start + tx_vmem_len;
-	/* TODO: RX ring */
-	int rx_vmem_len = 0;
-	/*
-	 * rx_ring_info ? RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
-	 * rx_ring_struct->vmem_size) : 0;
-	 */
+	int rx_vmem_len = rx_ring_info ?
+		RTE_CACHE_LINE_ROUNDUP(rx_ring_info->
+						rx_ring_struct->vmem_size) : 0;
 
 	int cp_ring_start = rx_vmem_start + rx_vmem_len;
 	int cp_ring_len = RTE_CACHE_LINE_ROUNDUP(cp_ring->ring_size *
@@ -109,13 +106,9 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 				   sizeof(struct tx_bd_long)) : 0;
 
 	int rx_ring_start = tx_ring_start + tx_ring_len;
-	/* TODO: RX ring */
-	int rx_ring_len = 0;
-	/*
-	 * rx_ring_info ?
-	 * RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
-	 * sizeof(struct rx_prod_pkt_bd)) : 0;
-	 */
+	int rx_ring_len =  rx_ring_info ?
+		RTE_CACHE_LINE_ROUNDUP(rx_ring_info->rx_ring_struct->ring_size *
+		sizeof(struct rx_prod_pkt_bd)) : 0;
 
 	int total_alloc_len = rx_ring_start + rx_ring_len;
 
@@ -153,26 +146,24 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		}
 	}
 
-/*
- *	if (rx_ring_info) {
- *		rx_ring = &rx_ring_info->rx_ring_struct;
- *
- *		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
- *		rx_ring_info->rx_desc_ring =
- *		    (struct rx_prod_pkt_bd *)rx_ring->bd;
- *		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
- *		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
- *
- *		if (!rx_ring->bd)
- *			return -ENOMEM;
- *		if (rx_ring->vmem_size) {
- *			rx_ring->vmem =
- *			    (void **)((char *)mz->addr + rx_vmem_start);
- *			rx_ring_info->rx_buf_ring =
- *			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
- *		}
- *	}
- */
+	if (rx_ring_info) {
+		rx_ring = rx_ring_info->rx_ring_struct;
+
+		rx_ring->bd = ((char *)mz->addr + rx_ring_start);
+		rx_ring_info->rx_desc_ring =
+		    (struct rx_prod_pkt_bd *)rx_ring->bd;
+		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
+		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+
+		if (!rx_ring->bd)
+			return -ENOMEM;
+		if (rx_ring->vmem_size) {
+			rx_ring->vmem =
+			    (void **)((char *)mz->addr + rx_vmem_start);
+			rx_ring_info->rx_buf_ring =
+			    (struct bnxt_sw_rx_bd *)rx_ring->vmem;
+		}
+	}
 
 	cp_ring->bd = ((char *)mz->addr + cp_ring_start);
 	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index 19c0e3f..eeda631 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -41,6 +41,7 @@
 #include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -216,7 +217,20 @@ err_out:
 
 static void bnxt_rx_queue_release_mbufs(struct bnxt_rx_queue *rxq __rte_unused)
 {
-	/* TODO: Requires interaction with TX ring */
+	struct bnxt_sw_rx_bd *sw_ring;
+	uint16_t i;
+
+	if (rxq) {
+		sw_ring = rxq->rx_ring->rx_buf_ring;
+		if (sw_ring) {
+			for (i = 0; i < rxq->nb_rx_desc; i++) {
+				if (sw_ring[i].mbuf) {
+					rte_pktmbuf_free_seg(sw_ring[i].mbuf);
+					sw_ring[i].mbuf = NULL;
+				}
+			}
+		}
+	}
 }
 
 void bnxt_free_rx_mbufs(struct bnxt *bp)
@@ -237,7 +251,13 @@ void bnxt_rx_queue_release_op(void *rx_queue)
 	if (rxq) {
 		bnxt_rx_queue_release_mbufs(rxq);
 
-		/* TODO: Free ring and stats here */
+		/* Free RX ring hardware descriptors */
+		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
+
+		/* Free RX completion ring hardware descriptors */
+		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+
+		bnxt_free_rxq_stats(rxq);
 
 		rte_free(rxq);
 	}
@@ -274,7 +294,7 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	rxq->nb_rx_desc = nb_desc;
 	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
 
-	/* TODO: Initialize ring structure */
+	bnxt_init_rx_ring_struct(rxq);
 
 	rxq->queue_id = queue_idx;
 	rxq->port_id = eth_dev->data->port_id;
@@ -282,7 +302,13 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 				0 : ETHER_CRC_LEN);
 
 	eth_dev->data->rx_queues[queue_idx] = rxq;
-	/* TODO: Allocate RX ring hardware descriptors */
+	/* Allocate RX ring hardware descriptors */
+	if (bnxt_alloc_rings(bp, queue_idx, NULL, rxq->rx_ring, rxq->cp_ring,
+			"rxr")) {
+		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for rx_ring failed!");
+		bnxt_rx_queue_release_op(rxq);
+		return -ENOMEM;
+	}
 
 	return 0;
 }
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
new file mode 100644
index 0000000..b02395b
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -0,0 +1,341 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <inttypes.h>
+#include <stdbool.h>
+
+#include <rte_byteorder.h>
+#include <rte_malloc.h>
+#include <rte_memory.h>
+
+#include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_ring.h"
+#include "bnxt_rxr.h"
+#include "bnxt_rxq.h"
+#include "hsi_struct_def_dpdk.h"
+
+/*
+ * RX Ring handling
+ */
+
+static inline struct rte_mbuf *__bnxt_alloc_rx_data(struct rte_mempool *mb)
+{
+	struct rte_mbuf *data;
+
+	data = rte_mbuf_raw_alloc(mb);
+
+	return data;
+}
+
+static inline int bnxt_alloc_rx_data(struct bnxt_rx_queue *rxq,
+				     struct bnxt_rx_ring_info *rxr,
+				     uint16_t prod)
+{
+	struct rx_prod_pkt_bd *rxbd = &rxr->rx_desc_ring[prod];
+	struct bnxt_sw_rx_bd *rx_buf = &rxr->rx_buf_ring[prod];
+	struct rte_mbuf *data;
+
+	data = __bnxt_alloc_rx_data(rxq->mb_pool);
+	if (!data)
+		return -ENOMEM;
+
+	rx_buf->mbuf = data;
+
+	rxbd->addr = rte_cpu_to_le_64(RTE_MBUF_DATA_DMA_ADDR(rx_buf->mbuf));
+
+	return 0;
+}
+
+static void bnxt_reuse_rx_mbuf(struct bnxt_rx_ring_info *rxr, uint16_t cons,
+			       struct rte_mbuf *mbuf)
+{
+	uint16_t prod = rxr->rx_prod;
+	struct bnxt_sw_rx_bd *prod_rx_buf;
+	struct rx_prod_pkt_bd *prod_bd, *cons_bd;
+
+	prod_rx_buf = &rxr->rx_buf_ring[prod];
+
+	prod_rx_buf->mbuf = mbuf;
+
+	prod_bd = &rxr->rx_desc_ring[prod];
+	cons_bd = &rxr->rx_desc_ring[cons];
+
+	prod_bd->addr = cons_bd->addr;
+}
+
+static uint16_t bnxt_rx_pkt(struct rte_mbuf **rx_pkt,
+			    struct bnxt_rx_queue *rxq, uint32_t *raw_cons)
+{
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+	struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+	struct rx_pkt_cmpl *rxcmp;
+	struct rx_pkt_cmpl_hi *rxcmp1;
+	uint32_t tmp_raw_cons = *raw_cons;
+	uint16_t cons, prod, cp_cons =
+	    RING_CMP(cpr->cp_ring_struct, tmp_raw_cons);
+	struct bnxt_sw_rx_bd *rx_buf;
+	struct rte_mbuf *mbuf;
+	int rc = 0;
+
+	rxcmp = (struct rx_pkt_cmpl *)
+	    &cpr->cp_desc_ring[cp_cons];
+
+	tmp_raw_cons = NEXT_RAW_CMP(tmp_raw_cons);
+	cp_cons = RING_CMP(cpr->cp_ring_struct, tmp_raw_cons);
+	rxcmp1 = (struct rx_pkt_cmpl_hi *)&cpr->cp_desc_ring[cp_cons];
+
+	if (!CMP_VALID(rxcmp1, tmp_raw_cons, cpr->cp_ring_struct))
+		return -EBUSY;
+
+	prod = rxr->rx_prod;
+
+	/* EW - GRO deferred to phase 3 */
+	cons = rxcmp->opaque;
+	rx_buf = &rxr->rx_buf_ring[cons];
+	mbuf = rx_buf->mbuf;
+	rte_prefetch0(mbuf);
+
+	mbuf->nb_segs = 1;
+	mbuf->next = NULL;
+	mbuf->pkt_len = rxcmp->len;
+	mbuf->data_len = mbuf->pkt_len;
+	mbuf->port = rxq->port_id;
+	mbuf->ol_flags = 0;
+	if (rxcmp->flags_type & RX_PKT_CMPL_FLAGS_RSS_VALID) {
+		mbuf->hash.rss = rxcmp->rss_hash;
+		mbuf->ol_flags |= PKT_RX_RSS_HASH;
+	} else {
+		mbuf->hash.fdir.id = rxcmp1->cfa_code;
+		mbuf->ol_flags |= PKT_RX_FDIR | PKT_RX_FDIR_ID;
+	}
+	if (rxcmp1->flags2 & RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN) {
+		mbuf->vlan_tci = rxcmp1->metadata &
+			(RX_PKT_CMPL_METADATA_VID_MASK |
+			RX_PKT_CMPL_METADATA_DE |
+			RX_PKT_CMPL_METADATA_PRI_MASK);
+		mbuf->ol_flags |= PKT_RX_VLAN_PKT;
+	}
+
+	rx_buf->mbuf = NULL;
+	if (rxcmp1->errors_v2 & RX_CMP_L2_ERRORS) {
+		/* Re-install the mbuf back to the rx ring */
+		bnxt_reuse_rx_mbuf(rxr, cons, mbuf);
+
+		rc = -EIO;
+		goto next_rx;
+	}
+	/*
+	 * TODO: Redesign this....
+	 * If the allocation fails, the packet does not get received.
+	 * Simply returning this will result in slowly falling behind
+	 * on the producer ring buffers.
+	 * Instead, "filling up" the producer just before ringing the
+	 * doorbell could be a better solution since it will let the
+	 * producer ring starve until memory is available again pushing
+	 * the drops into hardware and getting them out of the driver
+	 * allowing recovery to a full producer ring.
+	 *
+	 * This could also help with cache usage by preventing per-packet
+	 * calls in favour of a tight loop with the same function being called
+	 * in it.
+	 */
+	if (bnxt_alloc_rx_data(rxq, rxr, prod)) {
+		RTE_LOG(ERR, PMD, "mbuf alloc failed with prod=0x%x\n", prod);
+		rc = -ENOMEM;
+		goto next_rx;
+	}
+
+	/*
+	 * All MBUFs are allocated with the same size under DPDK,
+	 * no optimization for rx_copy_thresh
+	 */
+
+	/* AGG buf operation is deferred */
+
+	/* EW - VLAN reception.  Must compare against the ol_flags */
+
+	*rx_pkt = mbuf;
+next_rx:
+	rxr->rx_prod = RING_NEXT(rxr->rx_ring_struct, prod);
+
+	*raw_cons = tmp_raw_cons;
+
+	return rc;
+}
+
+uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct bnxt_rx_queue *rxq = rx_queue;
+	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+	struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+	uint32_t raw_cons = cpr->cp_raw_cons;
+	uint32_t cons;
+	int nb_rx_pkts = 0;
+	bool rx_event = false;
+	struct rx_pkt_cmpl *rxcmp;
+
+	/* Handle RX burst request */
+	while (1) {
+		int rc;
+
+		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
+		rte_prefetch0(&cpr->cp_desc_ring[cons]);
+		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
+
+		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
+			break;
+
+		/* TODO: Avoid magic numbers... */
+		if ((CMP_TYPE(rxcmp) & 0x30) == 0x10) {
+			rc = bnxt_rx_pkt(&rx_pkts[nb_rx_pkts], rxq, &raw_cons);
+			if (likely(!rc))
+				nb_rx_pkts++;
+			else if (rc == -EBUSY)	/* partial completion */
+				break;
+			rx_event = true;
+		}
+		raw_cons = NEXT_RAW_CMP(raw_cons);
+		if (nb_rx_pkts == nb_pkts)
+			break;
+	}
+	if (raw_cons == cpr->cp_raw_cons) {
+		/*
+		 * For PMD, there is no need to keep on pushing to REARM
+		 * the doorbell if there are no new completions
+		 */
+		return nb_rx_pkts;
+	}
+	cpr->cp_raw_cons = raw_cons;
+
+	B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+	if (rx_event)
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+	return nb_rx_pkts;
+}
+
+void bnxt_free_rx_rings(struct bnxt *bp)
+{
+	int i;
+
+	for (i = 0; i < (int)bp->rx_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+
+		if (!rxq)
+			continue;
+
+		/* TODO: free() rxq->rx_ring and rxq->rx_ring->rx_ring_struct */
+		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
+		/* TODO: free() rxq->cp_ring and rxq->cp_ring->cp_ring_struct */
+		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+
+		rte_free(rxq);
+		bp->rx_queues[i] = NULL;
+	}
+}
+
+void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt *bp = rxq->bp;
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_rx_ring_info *rxr;
+	struct bnxt_ring *ring;
+
+	rxq->rx_buf_use_size = bp->eth_dev->data->mtu +
+			       ETHER_HDR_LEN + ETHER_CRC_LEN +
+			       (2 * VLAN_TAG_SIZE);
+	rxq->rx_buf_size = rxq->rx_buf_use_size + sizeof(struct rte_mbuf);
+
+	rxr = rxq->rx_ring;
+	ring = rxr->rx_ring_struct;
+	ring->ring_size = rte_align32pow2(rxq->nb_rx_desc);
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)rxr->rx_desc_ring;
+	ring->bd_dma = rxr->rx_desc_mapping;
+	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_rx_bd);
+	ring->vmem = (void **)&rxr->rx_buf_ring;
+
+	cpr = rxq->cp_ring;
+	ring = cpr->cp_ring_struct;
+	ring->ring_size = rxr->rx_ring_struct->ring_size * 2;
+	ring->ring_mask = ring->ring_size - 1;
+	ring->bd = (void *)cpr->cp_desc_ring;
+	ring->bd_dma = cpr->cp_desc_mapping;
+	ring->vmem_size = 0;
+	ring->vmem = NULL;
+}
+
+static void bnxt_init_rxbds(struct bnxt_ring *ring, uint32_t type,
+			    uint16_t len)
+{
+	uint32_t j;
+	struct rx_prod_pkt_bd *rx_bd_ring = (struct rx_prod_pkt_bd *)ring->bd;
+
+	if (!rx_bd_ring)
+		return;
+	for (j = 0; j < ring->ring_size; j++) {
+		rx_bd_ring[j].flags_type = type;
+		rx_bd_ring[j].len = len;
+		rx_bd_ring[j].opaque = j;
+	}
+}
+
+int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
+{
+	struct bnxt_rx_ring_info *rxr;
+	struct bnxt_ring *ring;
+	uint32_t prod, type;
+	unsigned int i;
+
+	type = RX_PROD_PKT_BD_TYPE_RX_PROD_PKT | RX_PROD_PKT_BD_FLAGS_EOP_PAD;
+
+	/* TODO: These need to be allocated */
+	rxr = rxq->rx_ring;
+	ring = rxr->rx_ring_struct;
+	bnxt_init_rxbds(ring, type, rxq->rx_buf_use_size);
+
+	prod = rxr->rx_prod;
+	for (i = 0; i < ring->ring_size; i++) {
+		if (bnxt_alloc_rx_data(rxq, rxr, prod) != 0) {
+			RTE_LOG(WARNING, PMD,
+				"init'ed rx ring %d with %d/%d mbufs only\n",
+				rxq->queue_id, i, ring->ring_size);
+			break;
+		}
+		rxr->rx_prod = prod;
+		prod = RING_NEXT(rxr->rx_ring_struct, prod);
+	}
+
+	return 0;
+}
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
new file mode 100644
index 0000000..95c61a0
--- /dev/null
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -0,0 +1,62 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) Broadcom Limited.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Broadcom Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _BNXT_RXR_H_
+#define _BNXT_RXR_H_
+
+#define B_RX_DB(db, prod)						\
+		(*(uint32_t *)db = (DB_KEY_RX | prod))
+
+struct bnxt_sw_rx_bd {
+	struct rte_mbuf		*mbuf; /* data associated with RX descriptor */
+};
+
+struct bnxt_rx_ring_info {
+	uint16_t		rx_prod;
+	void			*rx_doorbell;
+
+	struct rx_prod_pkt_bd	*rx_desc_ring;
+	struct bnxt_sw_rx_bd	*rx_buf_ring; /* sw ring */
+
+	phys_addr_t		rx_desc_mapping;
+
+	struct bnxt_ring	*rx_ring_struct;
+};
+
+uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+			       uint16_t nb_pkts);
+void bnxt_free_rx_rings(struct bnxt *bp);
+void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq);
+int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq);
+
+#endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 1543f20..8b30787 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -446,6 +446,79 @@ struct tx_bd_long_hi {
 	uint32_t cfa_meta;
 } __attribute__((packed));
 
+/* RX Producer Packet BD (16 bytes) */
+struct rx_prod_pkt_bd {
+	/* This value identifies the type of buffer descriptor. */
+	#define RX_PROD_PKT_BD_TYPE_MASK		UINT32_C(0x3f)
+	#define RX_PROD_PKT_BD_TYPE_SFT			0
+		/*
+		 * Indicates that this BD is 16B long and is an RX Producer (ie.
+		 * empty) buffer descriptor.
+		 */
+	#define RX_PROD_PKT_BD_TYPE_RX_PROD_PKT		(UINT32_C(0x4) << 0)
+	/*
+	 * If set to 1, the packet will be placed at the address plus 2B. The 2
+	 * Bytes of padding will be written as zero.
+	 */
+	/*
+	 * This is intended to be used when the host buffer is cache-line
+	 * aligned to produce packets that are easy to parse in host memory
+	 * while still allowing writes to be cache line aligned.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_SOP_PAD		UINT32_C(0x40)
+	/*
+	 * If set to 1, the packet write will be padded out to the nearest
+	 * cache-line with zero value padding.
+	 */
+	/*
+	 * If receive buffers start/end on cache-line boundaries, this feature
+	 * will ensure that all data writes on the PCI bus start/end on cache
+	 * line boundaries.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_EOP_PAD		UINT32_C(0x80)
+	/*
+	 * This value is the number of additional buffers in the ring that
+	 * describe the buffer space to be consumed for the this packet. If the
+	 * value is zero, then the packet must fit within the space described by
+	 * this BD. If this value is 1 or more, it indicates how many additional
+	 * "buffer" BDs are in the ring immediately following this BD to be used
+	 * for the same network packet. Even if the packet to be placed does not
+	 * need all the additional buffers, they will be consumed anyway.
+	 */
+	#define RX_PROD_PKT_BD_FLAGS_BUFFERS_MASK	UINT32_C(0x300)
+	#define RX_PROD_PKT_BD_FLAGS_BUFFERS_SFT	8
+	#define RX_PROD_PKT_BD_FLAGS_MASK		UINT32_C(0xffc0)
+	#define RX_PROD_PKT_BD_FLAGS_SFT		6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length in Bytes of the host physical buffer where data
+	 * for the packet may be placed in host memory.
+	 */
+	/*
+	 * While this is a Byte resolution value, it is often advantageous to
+	 * ensure that the buffers provided end on a host cache line.
+	 */
+	uint16_t len;
+
+	/*
+	 * The opaque data field is pass through to the completion and can be
+	 * used for any data that the driver wants to associate with this
+	 * receive buffer set.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This is the host physical address where data for the packet may by
+	 * placed in host memory.
+	 */
+	/*
+	 * While this is a Byte resolution value, it is often advantageous to
+	 * ensure that the buffers provide start on a host cache line.
+	 */
+	uint64_t addr;
+} __attribute__((packed));
+
 /* Completion Ring Structures */
 /* Note: This structure is used by the HWRM to communicate HWRM Error. */
 /* Base Completion Record (16 bytes) */
@@ -611,6 +684,407 @@ struct tx_cmpl {
 	uint32_t unused_2;
 } __attribute__((packed)) tx_cmpl_t, *ptx_cmpl_t;
 
+/* RX Packet Completion Record (32 bytes split to 2 16-byte struct) */
+struct rx_pkt_cmpl {
+	/*
+	 * This field indicates the exact type of the completion. By convention,
+	 * the LSB identifies the length of the record in 16B units. Even values
+	 * indicate 16B records. Odd values indicate 32B records.
+	 */
+	#define RX_PKT_CMPL_TYPE_MASK			UINT32_C(0x3f)
+	#define RX_PKT_CMPL_TYPE_SFT			0
+		/*
+		 * RX L2 completion: Completion of and L2 RX packet.
+		 * Length = 32B
+		 */
+	#define RX_PKT_CMPL_TYPE_RX_L2			(UINT32_C(0x11) << 0)
+	/*
+	 * When this bit is '1', it indicates a packet that has an error of some
+	 * type. Type of error is indicated in error_flags.
+	 */
+	#define RX_PKT_CMPL_FLAGS_ERROR			UINT32_C(0x40)
+	/* This field indicates how the packet was placed in the buffer. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_MASK	UINT32_C(0x380)
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_SFT		7
+		/* Normal: Packet was placed using normal algorithm. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_NORMAL	(UINT32_C(0x0) << 7)
+		/* Jumbo: Packet was placed using jumbo algorithm. */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_JUMBO	(UINT32_C(0x1) << 7)
+		/*
+		 * Header/Data Separation: Packet was placed using Header/Data
+		 * separation algorithm. The separation location is indicated by
+		 * the itype field.
+		 */
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_HDS		(UINT32_C(0x2) << 7)
+	#define RX_PKT_CMPL_FLAGS_PLACEMENT_LAST \
+						RX_PKT_CMPL_FLAGS_PLACEMENT_HDS
+	/* This bit is '1' if the RSS field in this completion is valid. */
+	#define RX_PKT_CMPL_FLAGS_RSS_VALID		UINT32_C(0x400)
+	/*
+	 * This value indicates what the inner packet determined for the packet
+	 * was.
+	 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_MASK		UINT32_C(0xf000)
+	#define RX_PKT_CMPL_FLAGS_ITYPE_SFT		12
+		/* Not Known: Indicates that the packet type was not known. */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_NOT_KNOWN	(UINT32_C(0x0) << 12)
+		/*
+		 * IP Packet: Indicates that the packet was an IP packet, but
+		 * further classification was not possible.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_IP		(UINT32_C(0x1) << 12)
+		/*
+		 * TCP Packet: Indicates that the packet was IP and TCP. This
+		 * indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_TCP		(UINT32_C(0x2) << 12)
+		/*
+		 * UDP Packet: Indicates that the packet was IP and UDP. This
+		 * indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_UDP		(UINT32_C(0x3) << 12)
+		/*
+		 * FCoE Packet: Indicates that the packet was recognized as a
+		 * FCoE. This also indicates that the payload_offset field is
+		 * valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_FCOE		(UINT32_C(0x4) << 12)
+		/*
+		 * RoCE Packet: Indicates that the packet was recognized as a
+		 * RoCE. This also indicates that the payload_offset field is
+		 * valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_ROCE		(UINT32_C(0x5) << 12)
+		/*
+		 * ICMP Packet: Indicates that the packet was recognized as
+		 * ICMP. This indicates that the payload_offset field is valid.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_ICMP		(UINT32_C(0x7) << 12)
+		/*
+		 * PtP packet wo/timestamp: Indicates that the packet was
+		 * recognized as a PtP packet.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_PTP_WO_TIMESTAMP \
+							(UINT32_C(0x8) << 12)
+		/*
+		 * PtP packet w/timestamp: Indicates that the packet was
+		 * recognized as a PtP packet and that a timestamp was taken for
+		 * the packet.
+		 */
+	#define RX_PKT_CMPL_FLAGS_ITYPE_PTP_W_TIMESTAMP	(UINT32_C(0x9) << 12)
+	#define RX_PKT_CMPL_FLAGS_ITYPE_LAST \
+					RX_PKT_CMPL_FLAGS_ITYPE_PTP_W_TIMESTAMP
+	#define RX_PKT_CMPL_FLAGS_MASK			UINT32_C(0xffc0)
+	#define RX_PKT_CMPL_FLAGS_SFT			6
+	uint16_t flags_type;
+
+	/*
+	 * This is the length of the data for the packet stored in the buffer(s)
+	 * identified by the opaque value. This includes the packet BD and any
+	 * associated buffer BDs. This does not include the the length of any
+	 * data places in aggregation BDs.
+	 */
+	uint16_t len;
+
+	/*
+	 * This is a copy of the opaque field from the RX BD this completion
+	 * corresponds to.
+	 */
+	uint32_t opaque;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define RX_PKT_CMPL_V1				UINT32_C(0x1)
+	/*
+	 * This value is the number of aggregation buffers that follow this
+	 * entry in the completion ring that are a part of this packet. If the
+	 * value is zero, then the packet is completely contained in the buffer
+	 * space provided for the packet in the RX ring.
+	 */
+	#define RX_PKT_CMPL_AGG_BUFS_MASK		UINT32_C(0x3e)
+	#define RX_PKT_CMPL_AGG_BUFS_SFT		1
+	uint8_t agg_bufs_v1;
+
+	/*
+	 * This is the RSS hash type for the packet. The value is packed
+	 * {tuple_extrac_op[1:0],rss_profile_id[4:0],tuple_extrac_op[2]}.
+	 */
+	uint8_t rss_hash_type;
+
+	/*
+	 * This value indicates the offset from the beginning of the packet
+	 * where the inner payload starts. This value is valid for TCP, UDP,
+	 * FCoE, and RoCE packets.
+	 */
+	uint8_t payload_offset;
+
+	uint8_t unused_1;
+
+	/*
+	 * This value is the RSS hash value calculated for the packet based on
+	 * the mode bits and key value in the VNIC.
+	 */
+	uint32_t rss_hash;
+} __attribute__((packed));
+
+/* last 16 bytes of RX Packet Completion Record */
+struct rx_pkt_cmpl_hi {
+	/*
+	 * This indicates that the ip checksum was calculated for the inner
+	 * packet and that the ip_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_IP_CS_CALC		UINT32_C(0x1)
+	/*
+	 * This indicates that the TCP, UDP or ICMP checksum was calculated for
+	 * the inner packet and that the l4_cs_error field indicates if there
+	 * was an error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_L4_CS_CALC		UINT32_C(0x2)
+	/*
+	 * This indicates that the ip checksum was calculated for the tunnel
+	 * header and that the t_ip_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_T_IP_CS_CALC		UINT32_C(0x4)
+	/*
+	 * This indicates that the UDP checksum was calculated for the tunnel
+	 * packet and that the t_l4_cs_error field indicates if there was an
+	 * error.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_T_L4_CS_CALC		UINT32_C(0x8)
+	/* This value indicates what format the metadata field is. */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_MASK	UINT32_C(0xf0)
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_SFT	4
+		/* No metadata informtaion. Value is zero. */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_NONE	(UINT32_C(0x0) << 4)
+		/*
+		 * The metadata field contains the VLAN tag and TPID value. -
+		 * metadata[11:0] contains the vlan VID value. - metadata[12]
+		 * contains the vlan DE value. - metadata[15:13] contains the
+		 * vlan PRI value. - metadata[31:16] contains the vlan TPID
+		 * value.
+		 */
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN	(UINT32_C(0x1) << 4)
+	#define RX_PKT_CMPL_FLAGS2_META_FORMAT_LAST \
+					RX_PKT_CMPL_FLAGS2_META_FORMAT_VLAN
+	/*
+	 * This field indicates the IP type for the inner-most IP header. A
+	 * value of '0' indicates IPv4. A value of '1' indicates IPv6. This
+	 * value is only valid if itype indicates a packet with an IP header.
+	 */
+	#define RX_PKT_CMPL_FLAGS2_IP_TYPE		UINT32_C(0x100)
+	uint32_t flags2;
+
+	/*
+	 * This is data from the CFA block as indicated by the meta_format
+	 * field.
+	 */
+	/* When meta_format=1, this value is the VLAN VID. */
+	#define RX_PKT_CMPL_METADATA_VID_MASK		UINT32_C(0xfff)
+	#define RX_PKT_CMPL_METADATA_VID_SFT		0
+	/* When meta_format=1, this value is the VLAN DE. */
+	#define RX_PKT_CMPL_METADATA_DE			UINT32_C(0x1000)
+	/* When meta_format=1, this value is the VLAN PRI. */
+	#define RX_PKT_CMPL_METADATA_PRI_MASK		UINT32_C(0xe000)
+	#define RX_PKT_CMPL_METADATA_PRI_SFT		13
+	/* When meta_format=1, this value is the VLAN TPID. */
+	#define RX_PKT_CMPL_METADATA_TPID_MASK		UINT32_C(0xffff0000)
+	#define RX_PKT_CMPL_METADATA_TPID_SFT		16
+	uint32_t metadata;
+
+	/*
+	 * This value is written by the NIC such that it will be different for
+	 * each pass through the completion queue. The even passes will write 1.
+	 * The odd passes will write 0.
+	 */
+	#define RX_PKT_CMPL_V2				UINT32_C(0x1)
+	/*
+	 * This error indicates that there was some sort of problem with the BDs
+	 * for the packet that was found after part of the packet was already
+	 * placed. The packet should be treated as invalid.
+	 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_MASK	UINT32_C(0xe)
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_SFT	1
+		/* No buffer error */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_NO_BUFFER \
+							(UINT32_C(0x0) << 1)
+		/*
+		 * Did Not Fit: Packet did not fit into packet buffer provided.
+		 * For regular placement, this means the packet did not fit in
+		 * the buffer provided. For HDS and jumbo placement, this means
+		 * that the packet could not be placed into 7 physical buffers
+		 * or less.
+		 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_DID_NOT_FIT \
+							(UINT32_C(0x1) << 1)
+		/*
+		 * Not On Chip: All BDs needed for the packet were not on-chip
+		 * when the packet arrived.
+		 */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_NOT_ON_CHIP \
+							(UINT32_C(0x2) << 1)
+		/* Bad Format: BDs were not formatted correctly. */
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_BAD_FORMAT \
+							(UINT32_C(0x3) << 1)
+	#define RX_PKT_CMPL_ERRORS_BUFFER_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_BUFFER_ERROR_BAD_FORMAT
+	/* This indicates that there was an error in the IP header checksum. */
+	#define RX_PKT_CMPL_ERRORS_IP_CS_ERROR		UINT32_C(0x10)
+	/*
+	 * This indicates that there was an error in the TCP, UDP or ICMP
+	 * checksum.
+	 */
+	#define RX_PKT_CMPL_ERRORS_L4_CS_ERROR		UINT32_C(0x20)
+	/*
+	 * This indicates that there was an error in the tunnel IP header
+	 * checksum.
+	 */
+	#define RX_PKT_CMPL_ERRORS_T_IP_CS_ERROR	UINT32_C(0x40)
+	/* This indicates that there was an error in the tunnel UDP checksum. */
+	#define RX_PKT_CMPL_ERRORS_T_L4_CS_ERROR	UINT32_C(0x80)
+	/*
+	 * This indicates that there was a CRC error on either an FCoE or RoCE
+	 * packet. The itype indicates the packet type.
+	 */
+	#define RX_PKT_CMPL_ERRORS_CRC_ERROR		UINT32_C(0x100)
+	/*
+	 * This indicates that there was an error in the tunnel portion of the
+	 * packet when this field is non-zero.
+	 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_MASK	UINT32_C(0xe00)
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_SFT	9
+		/*
+		 * No additional error occurred on the tunnel portion of the
+		 * packet of the packet does not have a tunnel.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_NO_ERROR	(UINT32_C(0x0) << 9)
+		/*
+		 * Indicates that IP header version does not match expectation
+		 * from L2 Ethertype for IPv4 and IPv6 in the tunnel header.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_VERSION \
+							(UINT32_C(0x1) << 9)
+		/*
+		 * Indicates that header length is out of range in the tunnel
+		 * header. Valid for IPv4.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_HDR_LEN \
+							(UINT32_C(0x2) << 9)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the PPPoE header length for a tunnel PPPoE packet.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_TUNNEL_TOTAL_ERROR \
+							(UINT32_C(0x3) << 9)
+		/*
+		 * Indicates that physical packet is shorter than that claimed
+		 * by the tunnel l3 header length. Valid for IPv4, or IPv6
+		 * tunnel packet packets.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_IP_TOTAL_ERROR \
+							(UINT32_C(0x4) << 9)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the tunnel UDP header length for a tunnel UDP
+		 * packet that is not fragmented.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_UDP_TOTAL_ERROR \
+							(UINT32_C(0x5) << 9)
+		/*
+		 * indicates that the IPv4 TTL or IPv6 hop limit check have
+		 * failed (e.g. TTL = 0) in the tunnel header. Valid for IPv4,
+		 * and IPv6.
+		 */
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_TTL \
+							(UINT32_C(0x6) << 9)
+	#define RX_PKT_CMPL_ERRORS_T_PKT_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_T_PKT_ERROR_T_L3_BAD_TTL
+	/*
+	 * This indicates that there was an error in the inner portion of the
+	 * packet when this field is non-zero.
+	 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_MASK	UINT32_C(0xf000)
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_SFT	12
+		/*
+		 * No additional error occurred on the tunnel portion of the
+		 * packet of the packet does not have a tunnel.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_NO_ERROR	(UINT32_C(0x0) << 12)
+		/*
+		 * Indicates that IP header version does not match expectation
+		 * from L2 Ethertype for IPv4 and IPv6 or that option other than
+		 * VFT was parsed on FCoE packet.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_VERSION \
+							(UINT32_C(0x1) << 12)
+		/*
+		 * indicates that header length is out of range. Valid for IPv4
+		 * and RoCE
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_HDR_LEN \
+							(UINT32_C(0x2) << 12)
+		/*
+		 * indicates that the IPv4 TTL or IPv6 hop limit check have
+		 * failed (e.g. TTL = 0). Valid for IPv4, and IPv6
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L3_BAD_TTL	(UINT32_C(0x3) << 12)
+		/*
+		 * Indicates that physical packet is shorter than that claimed
+		 * by the l3 header length. Valid for IPv4, IPv6 packet or RoCE
+		 * packets.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_IP_TOTAL_ERROR \
+							(UINT32_C(0x4) << 12)
+		/*
+		 * Indicates that the physical packet is shorter than that
+		 * claimed by the UDP header length for a UDP packet that is not
+		 * fragmented.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_UDP_TOTAL_ERROR \
+							(UINT32_C(0x5) << 12)
+		/*
+		 * Indicates that TCP header length > IP payload. Valid for TCP
+		 * packets only.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_HDR_LEN \
+							(UINT32_C(0x6) << 12)
+		/* Indicates that TCP header length < 5. Valid for TCP. */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_HDR_LEN_TOO_SMALL \
+							(UINT32_C(0x7) << 12)
+		/*
+		 * Indicates that TCP option headers result in a TCP header size
+		 * that does not match data offset in TCP header. Valid for TCP.
+		 */
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_OPT_LEN \
+							(UINT32_C(0x8) << 12)
+	#define RX_PKT_CMPL_ERRORS_PKT_ERROR_LAST \
+				RX_PKT_CMPL_ERRORS_PKT_ERROR_L4_BAD_OPT_LEN
+	#define RX_PKT_CMPL_ERRORS_MASK			UINT32_C(0xfffe)
+	#define RX_PKT_CMPL_ERRORS_SFT			1
+	uint16_t errors_v2;
+
+	/*
+	 * This field identifies the CFA action rule that was used for this
+	 * packet.
+	 */
+	uint16_t cfa_code;
+
+	/*
+	 * This value holds the reordering sequence number for the packet. If
+	 * the reordering sequence is not valid, then this value is zero. The
+	 * reordering domain for the packet is in the bottom 8 to 10b of the
+	 * rss_hash value. The bottom 20b of this value contain the ordering
+	 * domain value for the packet.
+	 */
+	#define RX_PKT_CMPL_REORDER_MASK		UINT32_C(0xffffff)
+	#define RX_PKT_CMPL_REORDER_SFT			0
+	uint32_t reorder;
+} __attribute__((packed));
+
 /* HWRM Forwarded Request (16 bytes) */
 struct hwrm_fwd_req_cmpl {
 	/* Length of forwarded request in bytes. */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (13 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 14/38] bnxt: add initial Rx " Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 16/38] bnxt: add HWRM function reset command Stephen Hurd
                       ` (23 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Perform allocation and free()ing of ring and information structures
for the TX, RX, and completion rings. The previous patches had
so far provided top level stubs and generic ring support, while this
patch does the real allocation and freeing of the memory specific to
each different type of generic ring.

For example bnxt_init_tx_ring_struct() or bnxt_init_rx_ring_struct() is
now allocating memory based on the socked_id being provided.

bnxt_tx_queue_setup_op() or bnxt_rx_queue_setup_op() have gone through
some reformatting to perform a graceful cleanup in case memory
allocation fails.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.
Change the argument passed to bnxt_alloc_rings.
 Instead of passing bnxt_tx_ring and bnxt_rx_ring,
 shorten them to txr and rxr respectively.
Add code to free the reserved memzone

v5:
More descriptive commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_cpr.c  | 28 +++++++++++++++++++++++-----
 drivers/net/bnxt/bnxt_cpr.h  |  2 +-
 drivers/net/bnxt/bnxt_ring.c |  4 ++++
 drivers/net/bnxt/bnxt_ring.h |  1 +
 drivers/net/bnxt/bnxt_rxq.c  | 19 ++++++++++++-------
 drivers/net/bnxt/bnxt_rxr.c  | 42 ++++++++++++++++++++++++++++++++++--------
 drivers/net/bnxt/bnxt_rxr.h  |  2 +-
 drivers/net/bnxt/bnxt_txq.c  | 25 ++++++++++++++++---------
 drivers/net/bnxt/bnxt_txr.c  | 43 ++++++++++++++++++++++++++++++++++---------
 drivers/net/bnxt/bnxt_txr.h  |  2 +-
 10 files changed, 127 insertions(+), 41 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_cpr.c b/drivers/net/bnxt/bnxt_cpr.c
index 5a4b13d..60c277a 100644
--- a/drivers/net/bnxt/bnxt_cpr.c
+++ b/drivers/net/bnxt/bnxt_cpr.c
@@ -31,6 +31,8 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_malloc.h>
+
 #include "bnxt.h"
 #include "bnxt_cpr.h"
 #include "bnxt_hwrm.h"
@@ -121,21 +123,37 @@ reject:
 void bnxt_free_def_cp_ring(struct bnxt *bp)
 {
 	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-	struct bnxt_ring *ring = cpr->cp_ring_struct;
 
-	bnxt_free_ring(ring);
+	bnxt_free_ring(cpr->cp_ring_struct);
+	rte_free(cpr->cp_ring_struct);
+	rte_free(cpr);
 }
 
 /* For the default completion ring only */
-void bnxt_init_def_ring_struct(struct bnxt *bp)
+int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id)
 {
-	struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
-	struct bnxt_ring *ring = cpr->cp_ring_struct;
+	struct bnxt_cp_ring_info *cpr;
+	struct bnxt_ring *ring;
 
+	cpr = rte_zmalloc_socket("cpr",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	bp->def_cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_cp_ring_struct",
+				  sizeof(struct bnxt_ring),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->ring_size = rte_align32pow2(DEFAULT_CP_RING_SIZE);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
diff --git a/drivers/net/bnxt/bnxt_cpr.h b/drivers/net/bnxt/bnxt_cpr.h
index 4bd5606..c176f8c 100644
--- a/drivers/net/bnxt/bnxt_cpr.h
+++ b/drivers/net/bnxt/bnxt_cpr.h
@@ -79,7 +79,7 @@ struct bnxt_cp_ring_info {
 
 struct bnxt;
 void bnxt_free_def_cp_ring(struct bnxt *bp);
-void bnxt_init_def_ring_struct(struct bnxt *bp);
+int bnxt_init_def_ring_struct(struct bnxt *bp, unsigned int socket_id);
 void bnxt_handle_async_event(struct bnxt *bp, struct cmpl_base *cmp);
 void bnxt_handle_fwd_req(struct bnxt *bp, struct cmpl_base *cmp);
 
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index f547a9e..67b009a 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -51,6 +51,7 @@ void bnxt_free_ring(struct bnxt_ring *ring)
 		memset((char *)*ring->vmem, 0, ring->vmem_size);
 		*ring->vmem = NULL;
 	}
+	rte_memzone_free((const struct rte_memzone *)ring->mem_zone);
 }
 
 /*
@@ -135,6 +136,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		tx_ring_info->tx_desc_ring = (struct tx_bd_long *)tx_ring->bd;
 		tx_ring->bd_dma = mz->phys_addr + tx_ring_start;
 		tx_ring_info->tx_desc_mapping = tx_ring->bd_dma;
+		tx_ring->mem_zone = (const void *)mz;
 
 		if (!tx_ring->bd)
 			return -ENOMEM;
@@ -154,6 +156,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 		    (struct rx_prod_pkt_bd *)rx_ring->bd;
 		rx_ring->bd_dma = mz->phys_addr + rx_ring_start;
 		rx_ring_info->rx_desc_mapping = rx_ring->bd_dma;
+		rx_ring->mem_zone = (const void *)mz;
 
 		if (!rx_ring->bd)
 			return -ENOMEM;
@@ -169,6 +172,7 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 	cp_ring->bd_dma = mz->phys_addr + cp_ring_start;
 	cp_ring_info->cp_desc_ring = cp_ring->bd;
 	cp_ring_info->cp_desc_mapping = cp_ring->bd_dma;
+	cp_ring->mem_zone = (const void *)mz;
 
 	if (!cp_ring->bd)
 		return -ENOMEM;
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index 1e9d3be..f6ab214 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -76,6 +76,7 @@ struct bnxt_ring {
 	void			**vmem;
 
 	uint16_t		fw_ring_id; /* Ring id filled by Chimp FW */
+	const void		*mem_zone;
 };
 
 struct bnxt_ring_grp_info {
diff --git a/drivers/net/bnxt/bnxt_rxq.c b/drivers/net/bnxt/bnxt_rxq.c
index eeda631..cddf17d 100644
--- a/drivers/net/bnxt/bnxt_rxq.c
+++ b/drivers/net/bnxt/bnxt_rxq.c
@@ -53,8 +53,6 @@ void bnxt_free_rxq_stats(struct bnxt_rx_queue *rxq)
 {
 	struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
 
-	/* 'Unreserve' rte_memzone */
-
 	if (cpr->hw_stats)
 		cpr->hw_stats = NULL;
 }
@@ -272,10 +270,12 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
 	struct bnxt_rx_queue *rxq;
+	int rc = 0;
 
 	if (!nb_desc || nb_desc > MAX_RX_DESC_CNT) {
 		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	if (eth_dev->data->rx_queues) {
@@ -287,14 +287,17 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 				 RTE_CACHE_LINE_SIZE, socket_id);
 	if (!rxq) {
 		RTE_LOG(ERR, PMD, "bnxt_rx_queue allocation failed!");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 	rxq->bp = bp;
 	rxq->mb_pool = mp;
 	rxq->nb_rx_desc = nb_desc;
 	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
 
-	bnxt_init_rx_ring_struct(rxq);
+	rc = bnxt_init_rx_ring_struct(rxq, socket_id);
+	if (rc)
+		goto out;
 
 	rxq->queue_id = queue_idx;
 	rxq->port_id = eth_dev->data->port_id;
@@ -307,8 +310,10 @@ int bnxt_rx_queue_setup_op(struct rte_eth_dev *eth_dev,
 			"rxr")) {
 		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for rx_ring failed!");
 		bnxt_rx_queue_release_op(rxq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
-	return 0;
+out:
+	return rc;
 }
diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
index b02395b..5d93de2 100644
--- a/drivers/net/bnxt/bnxt_rxr.c
+++ b/drivers/net/bnxt/bnxt_rxr.c
@@ -255,17 +255,20 @@ void bnxt_free_rx_rings(struct bnxt *bp)
 		if (!rxq)
 			continue;
 
-		/* TODO: free() rxq->rx_ring and rxq->rx_ring->rx_ring_struct */
 		bnxt_free_ring(rxq->rx_ring->rx_ring_struct);
-		/* TODO: free() rxq->cp_ring and rxq->cp_ring->cp_ring_struct */
+		rte_free(rxq->rx_ring->rx_ring_struct);
+		rte_free(rxq->rx_ring);
+
 		bnxt_free_ring(rxq->cp_ring->cp_ring_struct);
+		rte_free(rxq->cp_ring->cp_ring_struct);
+		rte_free(rxq->cp_ring);
 
 		rte_free(rxq);
 		bp->rx_queues[i] = NULL;
 	}
 }
 
-void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
+int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id)
 {
 	struct bnxt *bp = rxq->bp;
 	struct bnxt_cp_ring_info *cpr;
@@ -277,8 +280,19 @@ void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
 			       (2 * VLAN_TAG_SIZE);
 	rxq->rx_buf_size = rxq->rx_buf_use_size + sizeof(struct rte_mbuf);
 
-	rxr = rxq->rx_ring;
-	ring = rxr->rx_ring_struct;
+	rxr = rte_zmalloc_socket("bnxt_rx_ring",
+				 sizeof(struct bnxt_rx_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxr == NULL)
+		return -ENOMEM;
+	rxq->rx_ring = rxr;
+
+	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
+				   sizeof(struct bnxt_ring),
+				   RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	rxr->rx_ring_struct = ring;
 	ring->ring_size = rte_align32pow2(rxq->nb_rx_desc);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)rxr->rx_desc_ring;
@@ -286,14 +300,27 @@ void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq)
 	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_rx_bd);
 	ring->vmem = (void **)&rxr->rx_buf_ring;
 
-	cpr = rxq->cp_ring;
-	ring = cpr->cp_ring_struct;
+	cpr = rte_zmalloc_socket("bnxt_rx_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	rxq->cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_rx_ring_struct",
+				   sizeof(struct bnxt_ring),
+				   RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->ring_size = rxr->rx_ring_struct->ring_size * 2;
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
 
 static void bnxt_init_rxbds(struct bnxt_ring *ring, uint32_t type,
@@ -320,7 +347,6 @@ int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
 
 	type = RX_PROD_PKT_BD_TYPE_RX_PROD_PKT | RX_PROD_PKT_BD_FLAGS_EOP_PAD;
 
-	/* TODO: These need to be allocated */
 	rxr = rxq->rx_ring;
 	ring = rxr->rx_ring_struct;
 	bnxt_init_rxbds(ring, type, rxq->rx_buf_use_size);
diff --git a/drivers/net/bnxt/bnxt_rxr.h b/drivers/net/bnxt/bnxt_rxr.h
index 95c61a0..f766b26 100644
--- a/drivers/net/bnxt/bnxt_rxr.h
+++ b/drivers/net/bnxt/bnxt_rxr.h
@@ -56,7 +56,7 @@ struct bnxt_rx_ring_info {
 uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 			       uint16_t nb_pkts);
 void bnxt_free_rx_rings(struct bnxt *bp);
-void bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq);
+int bnxt_init_rx_ring_struct(struct bnxt_rx_queue *rxq, unsigned int socket_id);
 int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq);
 
 #endif
diff --git a/drivers/net/bnxt/bnxt_txq.c b/drivers/net/bnxt/bnxt_txq.c
index 460f71a..99ddddd 100644
--- a/drivers/net/bnxt/bnxt_txq.c
+++ b/drivers/net/bnxt/bnxt_txq.c
@@ -49,8 +49,6 @@ void bnxt_free_txq_stats(struct bnxt_tx_queue *txq)
 {
 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
 
-	/* 'Unreserve' rte_memzone */
-
 	if (cpr->hw_stats)
 		cpr->hw_stats = NULL;
 }
@@ -108,10 +106,12 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
 	struct bnxt_tx_queue *txq;
+	int rc = 0;
 
 	if (!nb_desc || nb_desc > MAX_TX_DESC_CNT) {
 		RTE_LOG(ERR, PMD, "nb_desc %d is invalid", nb_desc);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	if (eth_dev->data->tx_queues) {
@@ -123,15 +123,18 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 	}
 	txq = rte_zmalloc_socket("bnxt_tx_queue", sizeof(struct bnxt_tx_queue),
 				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq == NULL) {
+	if (!txq) {
 		RTE_LOG(ERR, PMD, "bnxt_tx_queue allocation failed!");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 	txq->bp = bp;
 	txq->nb_tx_desc = nb_desc;
 	txq->tx_free_thresh = tx_conf->tx_free_thresh;
 
-	bnxt_init_tx_ring_struct(txq);
+	rc = bnxt_init_tx_ring_struct(txq, socket_id);
+	if (rc)
+		goto out;
 
 	txq->queue_id = queue_idx;
 	txq->port_id = eth_dev->data->port_id;
@@ -141,15 +144,19 @@ int bnxt_tx_queue_setup_op(struct rte_eth_dev *eth_dev,
 			"txr")) {
 		RTE_LOG(ERR, PMD, "ring_dma_zone_reserve for tx_ring failed!");
 		bnxt_tx_queue_release_op(txq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
 	if (bnxt_init_one_tx_ring(txq)) {
 		RTE_LOG(ERR, PMD, "bnxt_init_one_tx_ring failed!");
 		bnxt_tx_queue_release_op(txq);
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out;
 	}
 
 	eth_dev->data->tx_queues[queue_idx] = txq;
-	return 0;
+
+out:
+	return rc;
 }
diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
index 9064e8d..8bf8fee 100644
--- a/drivers/net/bnxt/bnxt_txr.c
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -59,9 +59,12 @@ void bnxt_free_tx_rings(struct bnxt *bp)
 			continue;
 
 		bnxt_free_ring(txq->tx_ring->tx_ring_struct);
-		/* TODO: free() txq->tx_ring and txq->tx_ring->tx_ring_struct */
+		rte_free(txq->tx_ring->tx_ring_struct);
+		rte_free(txq->tx_ring);
+
 		bnxt_free_ring(txq->cp_ring->cp_ring_struct);
-		/* TODO: free() txq->cp_ring and txq->cp_ring->cp_ring_struct */
+		rte_free(txq->cp_ring->cp_ring_struct);
+		rte_free(txq->cp_ring);
 
 		rte_free(txq);
 		bp->tx_queues[i] = NULL;
@@ -79,15 +82,25 @@ int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq)
 	return 0;
 }
 
-void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
+int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id)
 {
 	struct bnxt_cp_ring_info *cpr;
 	struct bnxt_tx_ring_info *txr;
 	struct bnxt_ring *ring;
 
-	/* TODO: These need to be allocated */
-	txr = txq->tx_ring;
-	ring = txr->tx_ring_struct;
+	txr = rte_zmalloc_socket("bnxt_tx_ring",
+				 sizeof(struct bnxt_tx_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txr == NULL)
+		return -ENOMEM;
+	txq->tx_ring = txr;
+
+	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
+				  sizeof(struct bnxt_ring),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	txr->tx_ring_struct = ring;
 	ring->ring_size = rte_align32pow2(txq->nb_tx_desc + 1);
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)txr->tx_desc_ring;
@@ -95,15 +108,27 @@ void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq)
 	ring->vmem_size = ring->ring_size * sizeof(struct bnxt_sw_tx_bd);
 	ring->vmem = (void **)&txr->tx_buf_ring;
 
-	/* TODO: These need to be allocated */
-	cpr = txq->cp_ring;
-	ring = cpr->cp_ring_struct;
+	cpr = rte_zmalloc_socket("bnxt_tx_ring",
+				 sizeof(struct bnxt_cp_ring_info),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (cpr == NULL)
+		return -ENOMEM;
+	txq->cp_ring = cpr;
+
+	ring = rte_zmalloc_socket("bnxt_tx_ring_struct",
+				  sizeof(struct bnxt_ring),
+				  RTE_CACHE_LINE_SIZE, socket_id);
+	if (ring == NULL)
+		return -ENOMEM;
+	cpr->cp_ring_struct = ring;
 	ring->ring_size = txr->tx_ring_struct->ring_size;
 	ring->ring_mask = ring->ring_size - 1;
 	ring->bd = (void *)cpr->cp_desc_ring;
 	ring->bd_dma = cpr->cp_desc_mapping;
 	ring->vmem_size = 0;
 	ring->vmem = NULL;
+
+	return 0;
 }
 
 static inline uint32_t bnxt_tx_avail(struct bnxt_tx_ring_info *txr)
diff --git a/drivers/net/bnxt/bnxt_txr.h b/drivers/net/bnxt/bnxt_txr.h
index 712ce65..2176aca 100644
--- a/drivers/net/bnxt/bnxt_txr.h
+++ b/drivers/net/bnxt/bnxt_txr.h
@@ -64,7 +64,7 @@ struct bnxt_sw_tx_bd {
 
 void bnxt_free_tx_rings(struct bnxt *bp);
 int bnxt_init_one_tx_ring(struct bnxt_tx_queue *txq);
-void bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq);
+int bnxt_init_tx_ring_struct(struct bnxt_tx_queue *txq, unsigned int socket_id);
 uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 			       uint16_t nb_pkts);
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 16/38] bnxt: add HWRM function reset command
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (14 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 17/38] bnxt: add HWRM vnic alloc function Stephen Hurd
                       ` (22 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add bnxt_hwrm_func_reset() function and supporting structs and macros.

New HWRM calls:
bnxt_hwrm_func_reset:
	This command puts the function into the reset state.
	In the reset state, global and port related features of the
	chip are not available.

This command resets a hardware function (PCIe function) and
frees any resources used by the function. This command initiated by
the driver prepare the function for re-use. This command may also be
initiated by a driver prior to doing it's own configuration.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
More descriptive commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           |  17 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 129 +++++++++++++++++++++++++++++++++
 3 files changed, 147 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 633683b..d1ca60a 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -265,6 +265,23 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_func_reset(struct bnxt *bp)
+{
+	int rc = 0;
+	struct hwrm_func_reset_input req = {.req_type = 0 };
+	struct hwrm_func_reset_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, FUNC_RESET, -1, resp);
+
+	req.enables = rte_cpu_to_le_32(0);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd)
 {
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 75c3442..2679d09 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -54,6 +54,7 @@ int bnxt_hwrm_exec_fwd_resp(struct bnxt *bp, void *fwd_cmd);
 int bnxt_hwrm_func_driver_register(struct bnxt *bp, uint32_t flags,
 				   uint32_t *vf_req_fwd);
 int bnxt_hwrm_func_qcaps(struct bnxt *bp);
+int bnxt_hwrm_func_reset(struct bnxt *bp);
 int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 8b30787..6209368 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -83,6 +83,7 @@ struct ctx_hw_stats64 {
  * Request types
  */
 #define HWRM_VER_GET			(UINT32_C(0x0))
+#define HWRM_FUNC_RESET			(UINT32_C(0x11))
 #define HWRM_FUNC_QCAPS			(UINT32_C(0x15))
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
@@ -2048,6 +2049,134 @@ struct hwrm_func_qcaps_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_func_reset */
+/*
+ * Description: This command resets a hardware function (PCIe function) and
+ * frees any resources used by the function. This command shall be initiated by
+ * the driver after an FLR has occurred to prepare the function for re-use. This
+ * command may also be initiated by a driver prior to doing it's own
+ * configuration. This command puts the function into the reset state. In the
+ * reset state, global and port related features of the chip are not available.
+ */
+/*
+ * Note: This command will reset a function that has already been disabled or
+ * idled. The command returns all the resources owned by the function so a new
+ * driver may allocate and configure resources normally.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_func_reset_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This bit must be '1' for the vf_id_valid field to be configured. */
+	#define HWRM_FUNC_RESET_INPUT_ENABLES_VF_ID_VALID \
+							UINT32_C(0x1)
+	uint32_t enables;
+
+	/*
+	 * The ID of the VF that this PF is trying to reset. Only the parent PF
+	 * shall be allowed to reset a child VF. A parent PF driver shall use
+	 * this field only when a specific child VF is requested to be reset.
+	 */
+	uint16_t vf_id;
+
+	/* This value indicates the level of a function reset. */
+		/*
+		 * Reset the caller function and its children VFs (if any). If
+		 * no children functions exist, then reset the caller function
+		 * only.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETALL \
+							(UINT32_C(0x0) << 0)
+		/* Reset the caller function only */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETME \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Reset all children VFs of the caller function driver if the
+		 * caller is a PF driver. It is an error to specify this level
+		 * by a VF driver. It is an error to specify this level by a PF
+		 * driver with no children VFs.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETCHILDREN \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Reset a specific VF of the caller function driver if the
+		 * caller is the parent PF driver. It is an error to specify
+		 * this level by a VF driver. It is an error to specify this
+		 * level by a PF driver that is not the parent of the VF that is
+		 * being requested to reset.
+		 */
+	#define HWRM_FUNC_RESET_INPUT_FUNC_RESET_LEVEL_RESETVF \
+							(UINT32_C(0x3) << 0)
+	uint8_t func_reset_level;
+
+	uint8_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_func_reset_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_port_phy_cfg */
 /*
  * Description: This command configures the PHY device for the port. It allows
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 17/38] bnxt: add HWRM vnic alloc function
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (15 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 16/38] bnxt: add HWRM function reset command Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 18/38] bnxt: add HWRM vnic free function Stephen Hurd
                       ` (21 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This requires a group info array in struct bnxt, so add that, we can save
the max size from the func_qcap response, and alloc/free in init/uninit

A VNIC represents a virtual interface.
It is a resource in the RX path of the chip and is used to setup various
target actions such as RSS, MAC filtering etc.. for the physical function
in use.

New HWRM call:
bnxt_hwrm_vnic_alloc:
	Allocates a VNIC resource in the hardware.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt.h                |  2 +
 drivers/net/bnxt/bnxt_hwrm.c           | 32 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 99 ++++++++++++++++++++++++++++++++++
 4 files changed, 135 insertions(+)

diff --git a/drivers/net/bnxt/bnxt.h b/drivers/net/bnxt/bnxt.h
index f7cf9d1..df1f771 100644
--- a/drivers/net/bnxt/bnxt.h
+++ b/drivers/net/bnxt/bnxt.h
@@ -141,6 +141,8 @@ struct bnxt {
 
 	/* Default completion ring */
 	struct bnxt_cp_ring_info	*def_cp_ring;
+	uint32_t		max_ring_grps;
+	struct bnxt_ring_grp_info	*grp_info;
 
 	unsigned int		nr_vnics;
 
diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index d1ca60a..fb0b7bf 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -43,6 +43,7 @@
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_ring.h"
 #include "bnxt_txq.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
@@ -235,6 +236,7 @@ int bnxt_hwrm_func_qcaps(struct bnxt *bp)
 
 	HWRM_CHECK_RESULT;
 
+	bp->max_ring_grps = rte_le_to_cpu_32(resp->max_hw_ring_grps);
 	if (BNXT_PF(bp)) {
 		struct bnxt_pf_info *pf = &bp->pf;
 
@@ -521,6 +523,36 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0, i, j;
+	struct hwrm_vnic_alloc_input req = {.req_type = 0 };
+	struct hwrm_vnic_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	/* map ring groups to this vnic */
+	for (i = vnic->start_grp_id, j = 0; i <= vnic->end_grp_id; i++, j++) {
+		if (bp->grp_info[i].fw_grp_id == (uint16_t)HWRM_NA_SIGNATURE) {
+			RTE_LOG(ERR, PMD,
+				"Not enough ring groups avail:%x req:%x\n", j,
+				(vnic->end_grp_id - vnic->start_grp_id) + 1);
+			break;
+		}
+		vnic->fw_grp_ids[j] = bp->grp_info[i].fw_grp_id;
+	}
+
+	vnic->fw_rss_cos_lb_ctx = (uint16_t)HWRM_NA_SIGNATURE;
+	vnic->ctx_is_rss_cos_lb = HW_CONTEXT_NONE;
+
+	HWRM_PREP(req, VNIC_ALLOC, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_vnic_id = rte_le_to_cpu_16(resp->vnic_id);
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 2679d09..7d09773 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,6 +63,8 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 6209368..eedd368 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -89,6 +89,7 @@ struct ctx_hw_stats64 {
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
+#define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3119,6 +3120,104 @@ struct hwrm_stat_ctx_clr_stats_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_alloc */
+/*
+ * Description: This VNIC is a resource in the RX side of the chip that is used
+ * to represent a virtual host "interface". # At the time of VNIC allocation or
+ * configuration, the function can specify whether it wants the requested VNIC
+ * to be the default VNIC for the function or not. # If a function requests
+ * allocation of a VNIC for the first time and a VNIC is successfully allocated
+ * by the HWRM, then the HWRM shall make the allocated VNIC as the default VNIC
+ * for that function. # The default VNIC shall be used for the default action
+ * for a partition or function. # For each VNIC allocated on a function, a
+ * mapping on the RX side to map the allocated VNIC to source virtual interface
+ * shall be performed by the HWRM. This should be hidden to the function driver
+ * requesting the VNIC allocation. This enables broadcast/multicast replication
+ * with source knockout. # If multicast replication with source knockout is
+ * enabled, then the internal VNIC to SVIF mapping data structures shall be
+ * programmed at the time of VNIC allocation.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_vnic_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', this VNIC is requested to be the default VNIC
+	 * for this function.
+	 */
+	#define HWRM_VNIC_ALLOC_INPUT_FLAGS_DEFAULT                UINT32_C(0x1)
+	uint32_t flags;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* Logical vnic ID */
+	uint32_t vnic_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 18/38] bnxt: add HWRM vnic free function
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (16 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 17/38] bnxt: add HWRM vnic alloc function Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 19/38] bnxt: add HWRM vnic configure function Stephen Hurd
                       ` (20 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

In this patch we add a new HWRM API to free a VNIC.

A VNIC represents a virtual interface.
It is a resource in the RX path of the chip and is used to setup
various target actions such as RSS, MAC filtering etc.. for the physical
function in use.

New HWRM call:
bnxt_hwrm_vnic_free:
	Frees a vnic allocated by the bnxt_hwrm_vnic_alloc() function.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           | 21 +++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 82 ++++++++++++++++++++++++++++++++++
 3 files changed, 104 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index fb0b7bf..fbbb81d 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -553,6 +553,27 @@ int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_free_input req = {.req_type = 0 };
+	struct hwrm_vnic_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	if (vnic->fw_vnic_id == INVALID_HW_RING_ID)
+		return rc;
+
+	HWRM_PREP(req, VNIC_FREE, -1, resp);
+
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_vnic_id = INVALID_HW_RING_ID;
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 7d09773..93fb759 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,6 +63,7 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index eedd368..0771897 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -90,6 +90,7 @@ struct ctx_hw_stats64 {
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
+#define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3218,6 +3219,87 @@ struct hwrm_vnic_alloc_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_free */
+/*
+ * Description: Free a VNIC resource. Idle any resources associated with the
+ * VNIC as well as the VNIC. Reset and release all resources associated with the
+ * VNIC.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_vnic_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Logical vnic ID */
+	uint32_t vnic_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_rss_cfg */
 /* Description: This function is used to enable RSS configuration. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 19/38] bnxt: add HWRM vnic configure function
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (17 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 18/38] bnxt: add HWRM vnic free function Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 20/38] bnxt: add API to allow configuration of vnic Stephen Hurd
                       ` (19 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

This patch configures the properties and actions of the vnic
allocated by vnic_alloc function from the previous patch.

bnxt_hwrm_vnic_cfg:
	Configure the VNIC structure in hardware.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Add more description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           |  34 ++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +-
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 155 +++++++++++++++++++++++++++++++++
 3 files changed, 191 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index fbbb81d..0c54129 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -553,6 +553,40 @@ int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_cfg_input req = {.req_type = 0 };
+	struct hwrm_vnic_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_CFG, -1, resp);
+
+	/* Only RSS support for now TBD: COS & LB */
+	req.enables =
+	    rte_cpu_to_le_32(HWRM_VNIC_CFG_INPUT_ENABLES_DFLT_RING_GRP |
+			     HWRM_VNIC_CFG_INPUT_ENABLES_RSS_RULE |
+			     HWRM_VNIC_CFG_INPUT_ENABLES_MRU);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+	req.dflt_ring_grp =
+		rte_cpu_to_le_16(bp->grp_info[vnic->start_grp_id].fw_grp_id);
+	req.rss_rule = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+	req.cos_rule = rte_cpu_to_le_16(0xffff);
+	req.lb_rule = rte_cpu_to_le_16(0xffff);
+	req.mru = rte_cpu_to_le_16(bp->eth_dev->data->mtu + ETHER_HDR_LEN +
+				   ETHER_CRC_LEN + VLAN_TAG_SIZE);
+	if (vnic->func_default)
+		req.flags = 1;
+	if (vnic->vlan_strip)
+		req.flags |=
+		    rte_cpu_to_le_32(HWRM_VNIC_CFG_INPUT_FLAGS_VLAN_STRIP_MODE);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 93fb759..cdaa295 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,8 +63,9 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
-int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 0771897..ef0b37a 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -91,6 +91,7 @@ struct ctx_hw_stats64 {
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
+#define HWRM_VNIC_CFG			(UINT32_C(0x42))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3219,6 +3220,160 @@ struct hwrm_vnic_alloc_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_vnic_cfg */
+/* Description: Configure the RX VNIC structure. */
+
+/* Input (40 bytes) */
+struct hwrm_vnic_cfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/*
+	 * When this bit is '1', the VNIC is requested to be the default VNIC
+	 * for the function.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_DEFAULT		UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the VNIC is being configured to strip VLAN in
+	 * the RX path. If set to '0', then VLAN stripping is disabled on this
+	 * VNIC.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_VLAN_STRIP_MODE	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the VNIC is being configured to buffer receive
+	 * packets in the hardware until the host posts new receive buffers. If
+	 * set to '0', then bd_stall is being configured to be disabled on this
+	 * VNIC.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_BD_STALL_MODE		UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the VNIC is being configured to receive both
+	 * RoCE and non-RoCE traffic. If set to '0', then this VNIC is not
+	 * configured to be operating in dual VNIC mode.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_ROCE_DUAL_VNIC_MODE	UINT32_C(0x8)
+	/*
+	 * When this flag is set to '1', the VNIC is requested to be configured
+	 * to receive only RoCE traffic. If this flag is set to '0', then this
+	 * flag shall be ignored by the HWRM. If roce_dual_vnic_mode flag is set
+	 * to '1', then the HWRM client shall not set this flag to '1'.
+	 */
+	#define HWRM_VNIC_CFG_INPUT_FLAGS_ROCE_ONLY_VNIC_MODE	UINT32_C(0x10)
+	uint32_t flags;
+
+	/* This bit must be '1' for the dflt_ring_grp field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_DFLT_RING_GRP	UINT32_C(0x1)
+	/* This bit must be '1' for the rss_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_RSS_RULE		UINT32_C(0x2)
+	/* This bit must be '1' for the cos_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_COS_RULE		UINT32_C(0x4)
+	/* This bit must be '1' for the lb_rule field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_LB_RULE		UINT32_C(0x8)
+	/* This bit must be '1' for the mru field to be configured. */
+	#define HWRM_VNIC_CFG_INPUT_ENABLES_MRU			UINT32_C(0x10)
+	uint32_t enables;
+
+	/* Logical vnic ID */
+	uint16_t vnic_id;
+
+	/*
+	 * Default Completion ring for the VNIC. This ring will be chosen if
+	 * packet does not match any RSS rules and if there is no COS rule.
+	 */
+	uint16_t dflt_ring_grp;
+
+	/*
+	 * RSS ID for RSS rule/table structure. 0xFF... (All Fs) if there is no
+	 * RSS rule.
+	 */
+	uint16_t rss_rule;
+
+	/*
+	 * RSS ID for COS rule/table structure. 0xFF... (All Fs) if there is no
+	 * COS rule.
+	 */
+	uint16_t cos_rule;
+
+	/*
+	 * RSS ID for load balancing rule/table structure. 0xFF... (All Fs) if
+	 * there is no LB rule.
+	 */
+	uint16_t lb_rule;
+
+	/*
+	 * The maximum receive unit of the vnic. Each vnic is associated with a
+	 * function. The vnic mru value overwrites the mru setting of the
+	 * associated function. The HWRM shall make sure that vnic mru does not
+	 * exceed the mru of the port the function is associated with.
+	 */
+	uint16_t mru;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_cfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_free */
 /*
  * Description: Free a VNIC resource. Idle any resources associated with the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 20/38] bnxt: add API to allow configuration of vnic
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (18 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 19/38] bnxt: add HWRM vnic configure function Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 21/38] bnxt: add HWRM API to configure RSS Stephen Hurd
                       ` (18 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds APIs to allow configuration of a VNIC.
The functions alloc and free the Class of Service or COS and
Load Balance context corresponding to the VNIC in the chip.

New HWRM calls:
bnxt_hwrm_vnic_ctx_alloc:
	Used to allocate COS/Load Balance context of VNIC
bnxt_hwrm_vnic_ctx_free:
	Used to free COS/Load Balance context of VNIC

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
More description in commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           |  38 ++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   2 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 153 +++++++++++++++++++++++++++++++++
 3 files changed, 193 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 0c54129..5a5063f 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -587,6 +587,44 @@ int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cos_lb_ctx_alloc_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cos_lb_ctx_alloc_output *resp =
+						bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_COS_LB_CTX_ALLOC, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_rss_cos_lb_ctx = rte_le_to_cpu_16(resp->rss_cos_lb_ctx_id);
+
+	return rc;
+}
+
+int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cos_lb_ctx_free_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cos_lb_ctx_free_output *resp =
+						bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_COS_LB_CTX_FREE, -1, resp);
+
+	req.rss_cos_lb_ctx_id = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	vnic->fw_rss_cos_lb_ctx = INVALID_HW_RING_ID;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index cdaa295..fbb8aba 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -65,6 +65,8 @@ int bnxt_hwrm_ver_get(struct bnxt *bp);
 
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index ef0b37a..6412df2 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -92,6 +92,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
+#define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
+#define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
@@ -3576,6 +3578,157 @@ struct hwrm_vnic_rss_cfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* Input (16 bytes) */
+struct hwrm_vnic_rss_cos_lb_ctx_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_vnic_rss_cos_lb_ctx_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* rss_cos_lb_ctx_id is 16 b */
+	uint16_t rss_cos_lb_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+	uint8_t unused_4;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_vnic_rss_cos_lb_ctx_free */
+/* Description: This function can be used to free COS/Load Balance context. */
+/* Input (24 bytes) */
+
+struct hwrm_vnic_rss_cos_lb_ctx_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* rss_cos_lb_ctx_id is 16 b */
+	uint16_t rss_cos_lb_ctx_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_vnic_rss_cos_lb_ctx_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* Output (32 bytes) */
 struct hwrm_queue_qportcfg_output {
 	/*
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 21/38] bnxt: add HWRM API to configure RSS
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (19 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 20/38] bnxt: add API to allow configuration of vnic Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 22/38] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
                       ` (17 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

A VNIC represents a virtual interface. It is a resource in the RX path
of the chip and is used to setup various target actions such as RSS,
MAC filtering etc.. for the physical function in use.

New HWRM call:
bnxt_hwrm_vnic_rss_cfg:
	Used to enable RSS configuration of the VNIC.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Expand the commit log
Reorder footer
---
 drivers/net/bnxt/bnxt_hwrm.c           | 24 ++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  2 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h |  1 +
 3 files changed, 27 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 5a5063f..29b0729 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -646,6 +646,30 @@ int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
+			   struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_vnic_rss_cfg_input req = {.req_type = 0 };
+	struct hwrm_vnic_rss_cfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, VNIC_RSS_CFG, -1, resp);
+
+	req.hash_type = rte_cpu_to_le_32(vnic->hash_type);
+
+	req.ring_grp_tbl_addr =
+	    rte_cpu_to_le_64(vnic->rss_table_dma_addr);
+	req.hash_key_tbl_addr =
+	    rte_cpu_to_le_64(vnic->rss_hash_key_dma_addr);
+	req.rss_ctx_idx = rte_cpu_to_le_16(vnic->fw_rss_cos_lb_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 /*
  * HWRM utility functions
  */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index fbb8aba..8c7b939 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -68,6 +68,8 @@ int bnxt_hwrm_vnic_cfg(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_ctx_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_ctx_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
+			   struct bnxt_vnic_info *vnic);
 
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 6412df2..72d4984 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -92,6 +92,7 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
+#define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 22/38] bnxt: add API for L2 Rx mask set/clear functions
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (20 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 21/38] bnxt: add HWRM API to configure RSS Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 23/38] bnxt: add HWRM API for stats context allocation Stephen Hurd
                       ` (16 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add API to set/clear L2 Rx mask.

New HWRM calls:
bnxt_hwrm_cfa_l2_clear_rx_mask
bnxt_hwrm_cfa_l2_set_rx_mask

These HWRM APIs allow setting and clearing of Rx masks in L2 context
per VNIC.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           |  45 +++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   3 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 135 +++++++++++++++++++++++++++++++++
 3 files changed, 183 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 29b0729..2c43159 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -141,6 +141,51 @@ static int bnxt_hwrm_send_message(struct bnxt *bp, void *msg, uint32_t msg_len)
 		} \
 	}
 
+int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_set_rx_mask_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_set_rx_mask_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, CFA_L2_SET_RX_MASK, -1, resp);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+	req.mask = 0;
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
+int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	int rc = 0;
+	struct hwrm_cfa_l2_set_rx_mask_input req = {.req_type = 0 };
+	struct hwrm_cfa_l2_set_rx_mask_output *resp = bp->hwrm_cmd_resp_addr;
+	uint32_t mask = 0;
+
+	HWRM_PREP(req, CFA_L2_SET_RX_MASK, -1, resp);
+	req.vnic_id = rte_cpu_to_le_16(vnic->fw_vnic_id);
+
+	/* FIXME add multicast flag, when multicast adding options is supported
+	 * by ethtool.
+	 */
+	if (vnic->flags & BNXT_VNIC_INFO_PROMISC)
+		mask = HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_PROMISCUOUS;
+	if (vnic->flags & BNXT_VNIC_INFO_ALLMULTI)
+		mask = HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_ALL_MCAST;
+	req.mask = rte_cpu_to_le_32(HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_MCAST |
+				    HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_BCAST |
+				    mask);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	return rc;
+}
+
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter)
 {
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 8c7b939..8037242 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -43,6 +43,9 @@ struct bnxt_cp_ring_info;
 
 #define HWRM_SEQ_ID_INVALID -1U
 
+int bnxt_hwrm_cfa_l2_clear_rx_mask(struct bnxt *bp,
+				   struct bnxt_vnic_info *vnic);
+int bnxt_hwrm_cfa_l2_set_rx_mask(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_clear_filter(struct bnxt *bp,
 			   struct bnxt_filter_info *filter);
 int bnxt_hwrm_set_filter(struct bnxt *bp,
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 72d4984..f8f6a3f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -1732,6 +1732,141 @@ struct hwrm_cfa_l2_filter_free_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_cfa_l2_set_rx_mask */
+/* Description: This command will set rx mask of the function. */
+
+/* Input (40 bytes) */
+struct hwrm_cfa_l2_set_rx_mask_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* VNIC ID */
+	uint32_t vnic_id;
+
+	/* Reserved for future use. */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_RESERVED	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', the function is requested to accept multi-cast
+	 * packets specified by the multicast addr table.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_MCAST	UINT32_C(0x2)
+	/*
+	 * When this bit is '1', the function is requested to accept all multi-
+	 * cast packets.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_ALL_MCAST	UINT32_C(0x4)
+	/*
+	 * When this bit is '1', the function is requested to accept broadcast
+	 * packets.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_BCAST	UINT32_C(0x8)
+	/*
+	 * When this bit is '1', the function is requested to be put in the
+	 * promiscuous mode. The HWRM should accept any function to set up
+	 * promiscuous mode. The HWRM shall follow the semantics below for the
+	 * promiscuous mode support. # When partitioning is not enabled on a
+	 * port (i.e. single PF on the port), then the PF shall be allowed to be
+	 * in the promiscuous mode. When the PF is in the promiscuous mode, then
+	 * it shall receive all host bound traffic on that port. # When
+	 * partitioning is enabled on a port (i.e. multiple PFs per port) and a
+	 * PF on that port is in the promiscuous mode, then the PF receives all
+	 * traffic within that partition as identified by a unique identifier
+	 * for the PF (e.g. S-Tag). If a unique outer VLAN for the PF is
+	 * specified, then the setting of promiscuous mode on that PF shall
+	 * result in the PF receiving all host bound traffic with matching outer
+	 * VLAN. # A VF shall can be set in the promiscuous mode. In the
+	 * promiscuous mode, the VF does not receive any traffic unless a unique
+	 * outer VLAN for the VF is specified. If a unique outer VLAN for the VF
+	 * is specified, then the setting of promiscuous mode on that VF shall
+	 * result in the VF receiving all host bound traffic with the matching
+	 * outer VLAN. # The HWRM shall allow the setting of promiscuous mode on
+	 * a function independently from the promiscuous mode settings on other
+	 * functions.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_PROMISCUOUS	UINT32_C(0x10)
+	/*
+	 * If this flag is set, the corresponding RX filters shall be set up to
+	 * cover multicast/broadcast filters for the outermost Layer 2
+	 * destination MAC address field.
+	 */
+	#define HWRM_CFA_L2_SET_RX_MASK_INPUT_MASK_OUTERMOST	UINT32_C(0x20)
+	uint32_t mask;
+
+	/* This is the address for mcast address tbl. */
+	uint64_t mc_tbl_addr;
+
+	/*
+	 * This value indicates how many entries in mc_tbl are valid. Each entry
+	 * is 6 bytes.
+	 */
+	uint32_t num_mc_entries;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_cfa_l2_set_rx_mask_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_exec_fwd_resp */
 /*
  * Description: This command is used to send an encapsulated request to the
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 23/38] bnxt: add HWRM API for stats context allocation
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (21 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 22/38] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 24/38] bnxt: add HWRM ring alloc/free functions Stephen Hurd
                       ` (15 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API code to allocate a statistics context in the ASIC.
This API will be called by the previously submitted "add statistics
operations patch".

New HWRM call:
bnxt_hwrm_stat_ctx_alloc:
	This command allocates and does basic preparation for a stat
context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 52 ++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  3 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 89 ++++++++++++++++++++++++++++++++++
 3 files changed, 144 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 2c43159..420a1cc 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -568,6 +568,31 @@ int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 	return rc;
 }
 
+int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
+			     struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	int rc;
+	struct hwrm_stat_ctx_alloc_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_ALLOC, -1, resp);
+
+	req.update_period_ms = rte_cpu_to_le_32(1000);
+
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+	req.stats_dma_addr =
+	    rte_cpu_to_le_64(cpr->hw_stats_map);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	cpr->hw_stats_ctx_id = rte_le_to_cpu_16(resp->stat_ctx_id);
+	bp->grp_info[idx].fw_stats_ctx = cpr->hw_stats_ctx_id;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0, i, j;
@@ -744,6 +769,33 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq;
+		struct bnxt_rx_queue *rxq;
+		struct bnxt_cp_ring_info *cpr;
+		unsigned int idx = i + 1;
+
+		if (i >= bp->rx_cp_nr_rings) {
+			txq = bp->tx_queues[i - bp->rx_cp_nr_rings];
+			cpr = txq->cp_ring;
+		} else {
+			rxq = bp->rx_queues[i];
+			cpr = rxq->cp_ring;
+		}
+
+		rc = bnxt_hwrm_stat_ctx_alloc(bp, cpr, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 8037242..49966f5 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -63,6 +63,8 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
+int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
+			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
@@ -74,6 +76,7 @@ int bnxt_hwrm_vnic_free(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 			   struct bnxt_vnic_info *vnic);
 
+int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index f8f6a3f..28362c9 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -99,6 +99,7 @@ struct ctx_hw_stats64 {
 #define HWRM_CFA_L2_FILTER_FREE		(UINT32_C(0x91))
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
+#define HWRM_STAT_CTX_ALLOC		(UINT32_C(0xb0))
 #define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
@@ -3183,6 +3184,94 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_alloc */
+/*
+ * Description: This command allocates and does basic preparation for a stat
+ * context.
+ */
+
+/* Input (32 bytes) */
+struct hwrm_stat_ctx_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This is the address for statistic block. */
+	uint64_t stats_dma_addr;
+
+	/*
+	 * The statistic block update period in ms. e.g. 250ms, 500ms, 750ms,
+	 * 1000ms.
+	 */
+	uint32_t update_period_ms;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_stat_ctx_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This is the statistics context ID value. */
+	uint32_t stat_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_clr_stats */
 /* Description: This command clears statistics of a context. */
 
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 24/38] bnxt: add HWRM ring alloc/free functions
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (22 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 23/38] bnxt: add HWRM API for stats context allocation Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 25/38] bnxt: add ring group " Stephen Hurd
                       ` (14 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API calls to allocate and free TX, RX and Completion rings
in the hardware along with the associated structs and definitions.

This informs the hardware of how the specific rings were set up in the
host and allocates them in the HWRM, setting up the doorbell registers
etc. as needed, returning an ID for the ring.

Basic ring alloc/free calls:
bnxt_hwrm_ring_alloc
	This command allocates and does basic preparation for a ring.
bnxt_hwrm_ring_free
	This command is used to free a ring and associated resources.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 108 ++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   7 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 305 +++++++++++++++++++++++++++++++++
 3 files changed, 420 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 420a1cc..565b76f 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -547,6 +547,114 @@ int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_hwrm_ring_alloc(struct bnxt *bp,
+			 struct bnxt_ring *ring,
+			 uint32_t ring_type, uint32_t map_index,
+			 uint32_t stats_ctx_id)
+{
+	int rc = 0;
+	struct hwrm_ring_alloc_input req = {.req_type = 0 };
+	struct hwrm_ring_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_ALLOC, -1, resp);
+
+	req.enables = rte_cpu_to_le_32(0);
+
+	req.page_tbl_addr = rte_cpu_to_le_64(ring->bd_dma);
+	req.fbo = rte_cpu_to_le_32(0);
+	/* Association of ring index with doorbell index */
+	req.logical_id = rte_cpu_to_le_16(map_index);
+
+	switch (ring_type) {
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_TX:
+		req.queue_id = bp->cos_queue[0].id;
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_RX:
+		req.ring_type = ring_type;
+		req.cmpl_ring_id =
+		    rte_cpu_to_le_16(bp->grp_info[map_index].cp_fw_ring_id);
+		req.length = rte_cpu_to_le_32(ring->ring_size);
+		req.stat_ctx_id = rte_cpu_to_le_16(stats_ctx_id);
+		req.enables = rte_cpu_to_le_32(rte_le_to_cpu_32(req.enables) |
+			HWRM_RING_ALLOC_INPUT_ENABLES_STAT_CTX_ID_VALID);
+		break;
+	case HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL:
+		req.ring_type = ring_type;
+		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_POLL;
+		req.length = rte_cpu_to_le_32(ring->ring_size);
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "hwrm alloc invalid ring type %d\n",
+			ring_type);
+		return -1;
+	}
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	if (rc || resp->error_code) {
+		if (rc == 0 && resp->error_code)
+			rc = rte_le_to_cpu_16(resp->error_code);
+		switch (ring_type) {
+		case HWRM_RING_FREE_INPUT_RING_TYPE_CMPL:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc cp failed. rc:%d\n", rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_RX:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc rx failed. rc:%d\n", rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_TX:
+			RTE_LOG(ERR, PMD,
+				"hwrm_ring_alloc tx failed. rc:%d\n", rc);
+			return rc;
+		default:
+			RTE_LOG(ERR, PMD, "Invalid ring. rc:%d\n", rc);
+			return rc;
+		}
+	}
+
+	ring->fw_ring_id = rte_le_to_cpu_16(resp->ring_id);
+	return rc;
+}
+
+int bnxt_hwrm_ring_free(struct bnxt *bp,
+			struct bnxt_ring *ring, uint32_t ring_type)
+{
+	int rc;
+	struct hwrm_ring_free_input req = {.req_type = 0 };
+	struct hwrm_ring_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_FREE, -1, resp);
+
+	req.ring_type = ring_type;
+	req.ring_id = rte_cpu_to_le_16(ring->fw_ring_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	if (rc || resp->error_code) {
+		if (rc == 0 && resp->error_code)
+			rc = rte_le_to_cpu_16(resp->error_code);
+
+		switch (ring_type) {
+		case HWRM_RING_FREE_INPUT_RING_TYPE_CMPL:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free cp failed. rc:%d\n",
+				rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_RX:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free rx failed. rc:%d\n",
+				rc);
+			return rc;
+		case HWRM_RING_FREE_INPUT_RING_TYPE_TX:
+			RTE_LOG(ERR, PMD, "hwrm_ring_free tx failed. rc:%d\n",
+				rc);
+			return rc;
+		default:
+			RTE_LOG(ERR, PMD, "Invalid ring, rc:%d\n", rc);
+			return rc;
+		}
+	}
+	return 0;
+}
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 49966f5..ebde5fb 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -62,6 +62,13 @@ int bnxt_hwrm_func_driver_unregister(struct bnxt *bp, uint32_t flags);
 
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp);
 
+int bnxt_hwrm_ring_alloc(struct bnxt *bp,
+			 struct bnxt_ring *ring,
+			 uint32_t ring_type, uint32_t map_index,
+			 uint32_t stats_ctx_id);
+int bnxt_hwrm_ring_free(struct bnxt *bp,
+			struct bnxt_ring *ring, uint32_t ring_type);
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 28362c9..e6280b6 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -93,6 +93,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
 #define HWRM_VNIC_CFG			(UINT32_C(0x42))
 #define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
+#define HWRM_RING_ALLOC			(UINT32_C(0x50))
+#define HWRM_RING_FREE			(UINT32_C(0x51))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
@@ -3184,6 +3186,309 @@ struct hwrm_queue_qportcfg_input {
 	uint16_t unused_0;
 } __attribute__((packed));
 
+/* hwrm_ring_alloc */
+/*
+ * Description: This command allocates and does basic preparation for a ring.
+ */
+
+/* Input (80 bytes) */
+struct hwrm_ring_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This bit must be '1' for the Reserved1 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED1		UINT32_C(0x1)
+	/* This bit must be '1' for the Reserved2 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED2		UINT32_C(0x2)
+	/* This bit must be '1' for the Reserved3 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED3		UINT32_C(0x4)
+	/*
+	 * This bit must be '1' for the stat_ctx_id_valid field to be
+	 * configured.
+	 */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_STAT_CTX_ID_VALID	UINT32_C(0x8)
+	/* This bit must be '1' for the Reserved4 field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_RESERVED4		UINT32_C(0x10)
+	/* This bit must be '1' for the max_bw_valid field to be configured. */
+	#define HWRM_RING_ALLOC_INPUT_ENABLES_MAX_BW_VALID	UINT32_C(0x20)
+	uint32_t enables;
+
+	/* Ring Type. */
+		/* Completion Ring (CR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL	(UINT32_C(0x0) << 0)
+		/* TX Ring (TR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_TX	(UINT32_C(0x1) << 0)
+		/* RX Ring (RR) */
+	#define HWRM_RING_ALLOC_INPUT_RING_TYPE_RX	(UINT32_C(0x2) << 0)
+	uint8_t ring_type;
+
+	uint8_t unused_0;
+	uint16_t unused_1;
+
+	/* This value is a pointer to the page table for the Ring. */
+	uint64_t page_tbl_addr;
+
+	/* First Byte Offset of the first entry in the first page. */
+	uint32_t fbo;
+
+	/*
+	 * Actual page size in 2^page_size. The supported range is increments in
+	 * powers of 2 from 16 bytes to 1GB. - 4 = 16 B Page size is 16 B. - 12
+	 * = 4 KB Page size is 4 KB. - 13 = 8 KB Page size is 8 KB. - 16 = 64 KB
+	 * Page size is 64 KB. - 22 = 2 MB Page size is 2 MB. - 23 = 4 MB Page
+	 * size is 4 MB. - 31 = 1 GB Page size is 1 GB.
+	 */
+	uint8_t page_size;
+
+	/*
+	 * This value indicates the depth of page table. For this version of the
+	 * specification, value other than 0 or 1 shall be considered as an
+	 * invalid value. When the page_tbl_depth = 0, then it is treated as a
+	 * special case with the following. 1. FBO and page size fields are not
+	 * valid. 2. page_tbl_addr is the physical address of the first element
+	 * of the ring.
+	 */
+	uint8_t page_tbl_depth;
+
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * Number of 16B units in the ring. Minimum size for a ring is 16 16B
+	 * entries.
+	 */
+	uint32_t length;
+
+	/*
+	 * Logical ring number for the ring to be allocated. This value
+	 * determines the position in the doorbell area where the update to the
+	 * ring will be made. For completion rings, this value is also the MSI-X
+	 * vector number for the function the completion ring is associated
+	 * with.
+	 */
+	uint16_t logical_id;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This value
+	 * indicates what completion ring the TX ring is associated with.
+	 */
+	uint16_t cmpl_ring_id;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This value
+	 * indicates what CoS queue the TX ring is associated with.
+	 */
+	uint16_t queue_id;
+
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved1;
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint16_t reserved2;
+
+	uint8_t unused_6;
+	uint8_t unused_7;
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved3;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. This input
+	 * indicates what statistics context this ring should be associated
+	 * with.
+	 */
+	uint32_t stat_ctx_id;
+
+	/* This field is reserved for the future use. It shall be set to 0. */
+	uint32_t reserved4;
+
+	/*
+	 * This field is used only when ring_type is a TX ring. Maximum BW
+	 * allocated to this TX ring in Mbps. The HWRM will translate this value
+	 * into byte counter and time interval used for this ring inside the
+	 * device.
+	 */
+	uint32_t max_bw;
+
+	/*
+	 * This field is used only when ring_type is a Completion ring. This
+	 * value indicates what interrupt mode should be used on this completion
+	 * ring. Note: In the legacy interrupt mode, no more than 16 completion
+	 * rings are allowed.
+	 */
+		/* Legacy INTA */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_LEGACY	(UINT32_C(0x0) << 0)
+		/* Reserved */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_RSVD	(UINT32_C(0x1) << 0)
+		/* MSI-X */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_MSIX	(UINT32_C(0x2) << 0)
+		/* No Interrupt - Polled mode */
+	#define HWRM_RING_ALLOC_INPUT_INT_MODE_POLL	(UINT32_C(0x3) << 0)
+	uint8_t int_mode;
+
+	uint8_t unused_8[3];
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_ring_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* Physical number of ring allocated. */
+	uint16_t ring_id;
+
+	/* Logical number of ring allocated. */
+	uint16_t logical_ring_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ring_free */
+/*
+ * Description: This command is used to free a ring and associated resources.
+ */
+/* Input (24 bytes) */
+
+struct hwrm_ring_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Ring Type. */
+		/* Completion Ring (CR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_CMPL	(UINT32_C(0x0) << 0)
+		/* TX Ring (TR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_TX	(UINT32_C(0x1) << 0)
+		/* RX Ring (RR) */
+	#define HWRM_RING_FREE_INPUT_RING_TYPE_RX	(UINT32_C(0x2) << 0)
+	uint8_t ring_type;
+
+	uint8_t unused_0;
+
+	/* Physical number of ring allocated. */
+	uint16_t ring_id;
+
+	uint32_t unused_1;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_alloc */
 /*
  * Description: This command allocates and does basic preparation for a stat
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 25/38] bnxt: add ring group alloc/free functions
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (23 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 24/38] bnxt: add HWRM ring alloc/free functions Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 26/38] bnxt: add HWRM stat context free function Stephen Hurd
                       ` (13 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM API for ring group alloc/free functions, associated structs and
definitions.
This API allocates and does basic preparation for a ring group in ASIC.
A ring group is identified by an index. It consists of Rx ring id,
completion ring id and a statistics context.

New HWRM calls:
bnxt_hwrm_ring_grp_alloc
	Allocates and does basic preparation for a ring group
bnxt_hwrm_ring_grp_free
	Frees and does cleanup resources of a ring group

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address issues pointed out by checkpatch.

v5:
Reorder the footer.

v6:
Correct typos
---
 drivers/net/bnxt/bnxt_hwrm.c           |  84 +++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |   4 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 185 +++++++++++++++++++++++++++++++++
 3 files changed, 273 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 565b76f..f335401 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -655,6 +655,47 @@ int bnxt_hwrm_ring_free(struct bnxt *bp,
 	return 0;
 }
 
+int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx)
+{
+	int rc = 0;
+	struct hwrm_ring_grp_alloc_input req = {.req_type = 0 };
+	struct hwrm_ring_grp_alloc_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_GRP_ALLOC, -1, resp);
+
+	req.cr = rte_cpu_to_le_16(bp->grp_info[idx].cp_fw_ring_id);
+	req.rr = rte_cpu_to_le_16(bp->grp_info[idx].rx_fw_ring_id);
+	req.ar = rte_cpu_to_le_16(bp->grp_info[idx].ag_fw_ring_id);
+	req.sc = rte_cpu_to_le_16(bp->grp_info[idx].fw_stats_ctx);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->grp_info[idx].fw_grp_id =
+	    rte_le_to_cpu_16(resp->ring_group_id);
+
+	return rc;
+}
+
+int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx)
+{
+	int rc;
+	struct hwrm_ring_grp_free_input req = {.req_type = 0 };
+	struct hwrm_ring_grp_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, RING_GRP_FREE, -1, resp);
+
+	req.ring_group_id = rte_cpu_to_le_16(bp->grp_info[idx].fw_grp_id);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	bp->grp_info[idx].fw_grp_id = INVALID_HW_RING_ID;
+	return rc;
+}
+
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr)
 {
 	int rc = 0;
@@ -904,6 +945,49 @@ int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return rc;
 }
 
+int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
+{
+	uint16_t i;
+	uint32_t rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (bp->grp_info[idx].fw_grp_id == INVALID_HW_RING_ID) {
+			RTE_LOG(ERR, PMD,
+				"Attempt to free invalid ring group %d\n",
+				idx);
+			continue;
+		}
+
+		rc = bnxt_hwrm_ring_grp_free(bp, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
+int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp)
+{
+	uint16_t i;
+	uint32_t rc = 0;
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (bp->grp_info[idx].cp_fw_ring_id == INVALID_HW_RING_ID ||
+		    bp->grp_info[idx].rx_fw_ring_id == INVALID_HW_RING_ID)
+			continue;
+
+		rc = bnxt_hwrm_ring_grp_alloc(bp, idx);
+
+		if (rc)
+			return rc;
+	}
+	return rc;
+}
+
 void bnxt_free_hwrm_resources(struct bnxt *bp)
 {
 	/* Release memzone */
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index ebde5fb..a9a8dc6 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -68,6 +68,8 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 			 uint32_t stats_ctx_id);
 int bnxt_hwrm_ring_free(struct bnxt *bp,
 			struct bnxt_ring *ring, uint32_t ring_type);
+int bnxt_hwrm_ring_grp_alloc(struct bnxt *bp, unsigned int idx);
+int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
@@ -85,6 +87,8 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
+int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index e6280b6..4e2eb9f 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -95,6 +95,8 @@ struct ctx_hw_stats64 {
 #define HWRM_VNIC_RSS_CFG		(UINT32_C(0x46))
 #define HWRM_RING_ALLOC			(UINT32_C(0x50))
 #define HWRM_RING_FREE			(UINT32_C(0x51))
+#define HWRM_RING_GRP_ALLOC		(UINT32_C(0x60))
+#define HWRM_RING_GRP_FREE		(UINT32_C(0x61))
 #define HWRM_VNIC_RSS_COS_LB_CTX_ALLOC	(UINT32_C(0x70))
 #define HWRM_VNIC_RSS_COS_LB_CTX_FREE	(UINT32_C(0x71))
 #define HWRM_CFA_L2_FILTER_ALLOC	(UINT32_C(0x90))
@@ -3489,6 +3491,189 @@ struct hwrm_ring_free_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_ring_grp_alloc */
+/*
+ * Description: This API allocates and does basic preparation for a ring group.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ring_grp_alloc_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This value identifies the CR associated with the ring group. */
+	uint16_t cr;
+
+	/* This value identifies the main RR associated with the ring group. */
+	uint16_t rr;
+
+	/*
+	 * This value identifies the aggregation RR associated with the ring
+	 * group. If this value is 0xFF... (All Fs), then no Aggregation ring
+	 * will be set.
+	 */
+	uint16_t ar;
+
+	/*
+	 * This value identifies the statistics context associated with the ring
+	 * group.
+	 */
+	uint16_t sc;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_grp_alloc_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/*
+	 * This is the ring group ID value. Use this value to program the
+	 * default ring group for the VNIC or as table entries in an RSS/COS
+	 * context.
+	 */
+	uint32_t ring_group_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
+/* hwrm_ring_grp_free */
+/*
+ * Description: This API frees a ring group and associated resources. # If a
+ * ring in the ring group is reset or free, then the associated rings in the
+ * ring group shall also be reset/free using hwrm_ring_free. # A function driver
+ * shall always use hwrm_ring_grp_free after freeing all rings in a group. # As
+ * a part of executing this command, the HWRM shall reset all associated ring
+ * group resources.
+ */
+
+/* Input (24 bytes) */
+struct hwrm_ring_grp_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* This is the ring group ID value. */
+	uint32_t ring_group_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+struct hwrm_ring_grp_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	uint32_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+	uint8_t unused_3;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_stat_ctx_alloc */
 /*
  * Description: This command allocates and does basic preparation for a stat
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 26/38] bnxt: add HWRM stat context free function
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (24 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 25/38] bnxt: add ring group " Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 27/38] bnxt: add HWRM API to set and clear filters Stephen Hurd
                       ` (12 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add function and associated structures and definitions to free
statistics context from the ASIC.

New HWRM call:
bnxt_hwrm_stat_ctx_free
	This command is used to free a stat context.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comments and fix issues pointed out by checkpatch.

v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 44 ++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h           |  3 ++
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 81 ++++++++++++++++++++++++++++++++++
 3 files changed, 128 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index f335401..cb30109 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -742,6 +742,28 @@ int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 	return rc;
 }
 
+int bnxt_hwrm_stat_ctx_free(struct bnxt *bp,
+			    struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	int rc;
+	struct hwrm_stat_ctx_free_input req = {.req_type = 0 };
+	struct hwrm_stat_ctx_free_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, STAT_CTX_FREE, -1, resp);
+
+	req.stat_ctx_id = rte_cpu_to_le_16(cpr->hw_stats_ctx_id);
+	req.seq_id = rte_cpu_to_le_16(bp->hwrm_cmd_seq++);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	cpr->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
+	bp->grp_info[idx].fw_stats_ctx = cpr->hw_stats_ctx_id;
+
+	return rc;
+}
+
 int bnxt_hwrm_vnic_alloc(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 {
 	int rc = 0, i, j;
@@ -918,6 +940,28 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp)
+{
+	int rc;
+	unsigned int i;
+	struct bnxt_cp_ring_info *cpr;
+
+	for (i = 0; i < bp->rx_cp_nr_rings + bp->tx_cp_nr_rings; i++) {
+		unsigned int idx = i + 1;
+
+		if (i >= bp->rx_cp_nr_rings)
+			cpr = bp->tx_queues[i - bp->rx_cp_nr_rings]->cp_ring;
+		else
+			cpr = bp->rx_queues[i]->cp_ring;
+		if (cpr->hw_stats_ctx_id != HWRM_NA_SIGNATURE) {
+			rc = bnxt_hwrm_stat_ctx_free(bp, cpr, idx);
+			if (rc)
+				return rc;
+		}
+	}
+	return 0;
+}
+
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp)
 {
 	unsigned int i;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index a9a8dc6..ffb45b1 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -74,6 +74,8 @@ int bnxt_hwrm_ring_grp_free(struct bnxt *bp, unsigned int idx);
 int bnxt_hwrm_stat_clear(struct bnxt *bp, struct bnxt_cp_ring_info *cpr);
 int bnxt_hwrm_stat_ctx_alloc(struct bnxt *bp,
 			     struct bnxt_cp_ring_info *cpr, unsigned int idx);
+int bnxt_hwrm_stat_ctx_free(struct bnxt *bp,
+			    struct bnxt_cp_ring_info *cpr, unsigned int idx);
 
 int bnxt_hwrm_ver_get(struct bnxt *bp);
 
@@ -87,6 +89,7 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index 4e2eb9f..d58295a 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -104,6 +104,7 @@ struct ctx_hw_stats64 {
 #define HWRM_CFA_L2_FILTER_CFG		(UINT32_C(0x92))
 #define HWRM_CFA_L2_SET_RX_MASK		(UINT32_C(0x93))
 #define HWRM_STAT_CTX_ALLOC		(UINT32_C(0xb0))
+#define HWRM_STAT_CTX_FREE		(UINT32_C(0xb1))
 #define HWRM_STAT_CTX_CLR_STATS		(UINT32_C(0xb3))
 #define HWRM_EXEC_FWD_RESP		(UINT32_C(0xd0))
 
@@ -3839,6 +3840,86 @@ struct hwrm_stat_ctx_clr_stats_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_stat_ctx_free */
+/* Description: This command is used to free a stat context. */
+/* Input (24 bytes) */
+
+struct hwrm_stat_ctx_free_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* ID of the statistics context that is being queried. */
+	uint32_t stat_ctx_id;
+
+	uint32_t unused_0;
+} __attribute__((packed));
+
+/* Output (16 bytes) */
+
+struct hwrm_stat_ctx_free_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This is the statistics context ID value. */
+	uint32_t stat_ctx_id;
+
+	uint8_t unused_0;
+	uint8_t unused_1;
+	uint8_t unused_2;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_vnic_alloc */
 /*
  * Description: This VNIC is a resource in the RX side of the chip that is used
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 27/38] bnxt: add HWRM API to set and clear filters
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (25 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 26/38] bnxt: add HWRM stat context free function Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 28/38] bnxt: allocate and free all HWRM rings and groups Stephen Hurd
                       ` (11 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

New HWRM call:
bnxt_clear_hwrm_vnic_filters
	This patch adds code to set and clear L2 filters from the
	corresponding VNIC. These filters will determine the Rx flows

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Separated this code from the previous patch as it had nothing to
do with freeing of statistics context.

v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_hwrm.c | 26 ++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_hwrm.h |  3 +++
 2 files changed, 29 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index cb30109..13f9e1c 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -1059,6 +1059,32 @@ int bnxt_alloc_hwrm_resources(struct bnxt *bp)
 	return 0;
 }
 
+int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	struct bnxt_filter_info *filter;
+	int rc = 0;
+
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		rc = bnxt_hwrm_clear_filter(bp, filter);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
+int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
+{
+	struct bnxt_filter_info *filter;
+	int rc = 0;
+
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		rc = bnxt_hwrm_set_filter(bp, vnic, filter);
+		if (rc)
+			break;
+	}
+	return rc;
+}
+
 static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
 {
 	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index ffb45b1..9c26118 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -92,6 +92,9 @@ int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
+int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
+void bnxt_free_all_hwrm_resources(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 28/38] bnxt: allocate and free all HWRM rings and groups
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (26 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 27/38] bnxt: add HWRM API to set and clear filters Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 29/38] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
                       ` (10 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add a top level functions to initialize ring groups, and functions
to allocate and free all the rings via HWRM.

A ring group is identified by an index. It consists of Rx or Tx ring id,
completion ring id and a statistics context. Once a ring group is
initialized, use this group index while creating the rings in the ASIC
using the appropriate HWRM API added via earlier patches.

Functions added:
bnxt_free_cp_ring
	Calls the HWRM function generic ring free with arguments specific
	to a completion ring and sanitizes the host completion structure
bnxt_free_all_hwrm_rings
	Frees all the HWRM allocated hardware rings
bnxt_free_all_hwrm_resources
	Frees all the resources allocated via the HRM in the hardware
bnxt_alloc_hwrm_rings
	Allocates all the HWRM rings needed in the current configuration

This should be the last functionality needed to add start/stop
device operations.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Address review comment to merge another patch into this to avoid
a compilation issue. Fix issues pointed out by checkpatch.

v5:
Reorder the footer.
Remove instances of bnxt_ring_struct

v6:
Correct typos
---
 drivers/net/bnxt/bnxt_hwrm.c | 112 +++++++++++++++++++++++++++++++++++++++-
 drivers/net/bnxt/bnxt_hwrm.h |   1 +
 drivers/net/bnxt/bnxt_ring.c | 119 +++++++++++++++++++++++++++++++++++++++++++
 drivers/net/bnxt/bnxt_ring.h |   2 +
 4 files changed, 233 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index 13f9e1c..fcc4811 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -43,8 +43,10 @@
 #include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
 #include "bnxt_rxq.h"
+#include "bnxt_rxr.h"
 #include "bnxt_ring.h"
 #include "bnxt_txq.h"
+#include "bnxt_txr.h"
 #include "bnxt_vnic.h"
 #include "hsi_struct_def_dpdk.h"
 
@@ -579,7 +581,11 @@ int bnxt_hwrm_ring_alloc(struct bnxt *bp,
 		break;
 	case HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL:
 		req.ring_type = ring_type;
-		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_POLL;
+		/*
+		 * TODO: Some HWRM versions crash with
+		 * HWRM_RING_ALLOC_INPUT_INT_MODE_POLL
+		 */
+		req.int_mode = HWRM_RING_ALLOC_INPUT_INT_MODE_MSIX;
 		req.length = rte_cpu_to_le_32(ring->ring_size);
 		break;
 	default:
@@ -1012,6 +1018,84 @@ int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp)
 	return rc;
 }
 
+static void bnxt_free_cp_ring(struct bnxt *bp,
+			      struct bnxt_cp_ring_info *cpr, unsigned int idx)
+{
+	struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+
+	bnxt_hwrm_ring_free(bp, cp_ring,
+			HWRM_RING_FREE_INPUT_RING_TYPE_CMPL);
+	cp_ring->fw_ring_id = INVALID_HW_RING_ID;
+	bp->grp_info[idx].cp_fw_ring_id = INVALID_HW_RING_ID;
+	memset(cpr->cp_desc_ring, 0, cpr->cp_ring_struct->ring_size *
+			sizeof(*cpr->cp_desc_ring));
+	cpr->cp_raw_cons = 0;
+}
+
+int bnxt_free_all_hwrm_rings(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_tx_ring_info *txr = txq->tx_ring;
+		struct bnxt_ring *ring = txr->tx_ring_struct;
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		unsigned int idx = bp->rx_cp_nr_rings + i + 1;
+
+		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
+			bnxt_hwrm_ring_free(bp, ring,
+					HWRM_RING_FREE_INPUT_RING_TYPE_TX);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
+			memset(txr->tx_desc_ring, 0,
+					txr->tx_ring_struct->ring_size *
+					sizeof(*txr->tx_desc_ring));
+			memset(txr->tx_buf_ring, 0,
+					txr->tx_ring_struct->ring_size *
+					sizeof(*txr->tx_buf_ring));
+			txr->tx_prod = 0;
+			txr->tx_cons = 0;
+		}
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, idx);
+	}
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+		struct bnxt_ring *ring = rxr->rx_ring_struct;
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		unsigned int idx = i + 1;
+
+		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
+			bnxt_hwrm_ring_free(bp, ring,
+					HWRM_RING_FREE_INPUT_RING_TYPE_RX);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
+			bp->grp_info[idx].rx_fw_ring_id = INVALID_HW_RING_ID;
+			memset(rxr->rx_desc_ring, 0,
+					rxr->rx_ring_struct->ring_size *
+					sizeof(*rxr->rx_desc_ring));
+			memset(rxr->rx_buf_ring, 0,
+					rxr->rx_ring_struct->ring_size *
+					sizeof(*rxr->rx_buf_ring));
+			rxr->rx_prod = 0;
+		}
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, idx);
+	}
+
+	/* Default completion ring */
+	{
+		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+
+		if (cpr->cp_ring_struct->fw_ring_id != INVALID_HW_RING_ID)
+			bnxt_free_cp_ring(bp, cpr, 0);
+	}
+
+	return rc;
+}
+
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp)
 {
 	uint16_t i;
@@ -1085,6 +1169,32 @@ int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic)
 	return rc;
 }
 
+void bnxt_free_all_hwrm_resources(struct bnxt *bp)
+{
+	struct bnxt_vnic_info *vnic;
+	unsigned int i;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+	bnxt_hwrm_cfa_l2_clear_rx_mask(bp, vnic);
+
+	/* VNIC resources */
+	for (i = 0; i < bp->nr_vnics; i++) {
+		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+		bnxt_clear_hwrm_vnic_filters(bp, vnic);
+
+		bnxt_hwrm_vnic_ctx_free(bp, vnic);
+		bnxt_hwrm_vnic_free(bp, vnic);
+	}
+	/* Ring resources */
+	bnxt_free_all_hwrm_rings(bp);
+	bnxt_free_all_hwrm_ring_grps(bp);
+	bnxt_free_all_hwrm_stat_ctxs(bp);
+}
+
 static uint16_t bnxt_parse_eth_link_duplex(uint32_t conf_link_speed)
 {
 	uint8_t hw_link_duplex = HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index 9c26118..c28e8d8 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -90,6 +90,7 @@ int bnxt_hwrm_vnic_rss_cfg(struct bnxt *bp,
 int bnxt_alloc_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_clear_all_hwrm_stat_ctxs(struct bnxt *bp);
 int bnxt_free_all_hwrm_stat_ctxs(struct bnxt *bp);
+int bnxt_free_all_hwrm_rings(struct bnxt *bp);
 int bnxt_free_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_alloc_all_hwrm_ring_grps(struct bnxt *bp);
 int bnxt_set_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
diff --git a/drivers/net/bnxt/bnxt_ring.c b/drivers/net/bnxt/bnxt_ring.c
index 67b009a..3f81ffc 100644
--- a/drivers/net/bnxt/bnxt_ring.c
+++ b/drivers/net/bnxt/bnxt_ring.c
@@ -35,8 +35,11 @@
 
 #include "bnxt.h"
 #include "bnxt_cpr.h"
+#include "bnxt_hwrm.h"
 #include "bnxt_ring.h"
+#include "bnxt_rxq.h"
 #include "bnxt_rxr.h"
+#include "bnxt_txq.h"
 #include "bnxt_txr.h"
 
 #include "hsi_struct_def_dpdk.h"
@@ -55,6 +58,19 @@ void bnxt_free_ring(struct bnxt_ring *ring)
 }
 
 /*
+ * Ring groups
+ */
+
+void bnxt_init_ring_grps(struct bnxt *bp)
+{
+	unsigned int i;
+
+	for (i = 0; i < bp->max_ring_grps; i++)
+		memset(&bp->grp_info[i], (uint8_t)HWRM_NA_SIGNATURE,
+		       sizeof(struct bnxt_ring_grp_info));
+}
+
+/*
  * Allocates a completion ring with vmem and stats optionally also allocating
  * a TX and/or RX ring.  Passing NULL as tx_ring_info and/or rx_ring_info
  * to not allocate them.
@@ -185,3 +201,106 @@ int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 	cp_ring_info->hw_stats_ctx_id = HWRM_NA_SIGNATURE;
 	return 0;
 }
+
+/* ring_grp usage:
+ * [0] = default completion ring
+ * [1 -> +rx_cp_nr_rings] = rx_cp, rx rings
+ * [1+rx_cp_nr_rings + 1 -> +tx_cp_nr_rings] = tx_cp, tx rings
+ */
+int bnxt_alloc_hwrm_rings(struct bnxt *bp)
+{
+	unsigned int i;
+	int rc = 0;
+
+	/* Default completion ring */
+	{
+		struct bnxt_cp_ring_info *cpr = bp->def_cp_ring;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					  HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					  0, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+		bp->grp_info[0].cp_fw_ring_id = cp_ring->fw_ring_id;
+	}
+
+	for (i = 0; i < bp->rx_cp_nr_rings; i++) {
+		struct bnxt_rx_queue *rxq = bp->rx_queues[i];
+		struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_rx_ring_info *rxr = rxq->rx_ring;
+		struct bnxt_ring *ring = rxr->rx_ring_struct;
+		unsigned int idx = i + 1;
+
+		/* Rx cmpl */
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					idx, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].cp_fw_ring_id = cp_ring->fw_ring_id;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+
+		/* Rx ring */
+		rc = bnxt_hwrm_ring_alloc(bp, ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_RX,
+					idx, cpr->hw_stats_ctx_id);
+		if (rc)
+			goto err_out;
+		rxr->rx_prod = 0;
+		rxr->rx_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].rx_fw_ring_id = ring->fw_ring_id;
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+		if (bnxt_init_one_rx_ring(rxq)) {
+			RTE_LOG(ERR, PMD, "bnxt_init_one_rx_ring failed!");
+			bnxt_rx_queue_release_op(rxq);
+			return -ENOMEM;
+		}
+		B_RX_DB(rxr->rx_doorbell, rxr->rx_prod);
+	}
+
+	for (i = 0; i < bp->tx_cp_nr_rings; i++) {
+		struct bnxt_tx_queue *txq = bp->tx_queues[i];
+		struct bnxt_cp_ring_info *cpr = txq->cp_ring;
+		struct bnxt_ring *cp_ring = cpr->cp_ring_struct;
+		struct bnxt_tx_ring_info *txr = txq->tx_ring;
+		struct bnxt_ring *ring = txr->tx_ring_struct;
+		unsigned int idx = 1 + bp->rx_cp_nr_rings + i;
+
+		/* Tx cmpl */
+		rc = bnxt_hwrm_ring_alloc(bp, cp_ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_CMPL,
+					idx, HWRM_NA_SIGNATURE);
+		if (rc)
+			goto err_out;
+
+		cpr->cp_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+		bp->grp_info[idx].cp_fw_ring_id = cp_ring->fw_ring_id;
+		B_CP_DIS_DB(cpr, cpr->cp_raw_cons);
+
+		/* Tx ring */
+		rc = bnxt_hwrm_ring_alloc(bp, ring,
+					HWRM_RING_ALLOC_INPUT_RING_TYPE_TX,
+					idx, cpr->hw_stats_ctx_id);
+		if (rc)
+			goto err_out;
+
+		txr->tx_doorbell =
+		    (char *)bp->eth_dev->pci_dev->mem_resource[2].addr +
+		    idx * 0x80;
+	}
+
+err_out:
+	return rc;
+}
diff --git a/drivers/net/bnxt/bnxt_ring.h b/drivers/net/bnxt/bnxt_ring.h
index f6ab214..8656549 100644
--- a/drivers/net/bnxt/bnxt_ring.h
+++ b/drivers/net/bnxt/bnxt_ring.h
@@ -92,10 +92,12 @@ struct bnxt_tx_ring_info;
 struct bnxt_rx_ring_info;
 struct bnxt_cp_ring_info;
 void bnxt_free_ring(struct bnxt_ring *ring);
+void bnxt_init_ring_grps(struct bnxt *bp);
 int bnxt_alloc_rings(struct bnxt *bp, uint16_t qidx,
 			    struct bnxt_tx_ring_info *tx_ring_info,
 			    struct bnxt_rx_ring_info *rx_ring_info,
 			    struct bnxt_cp_ring_info *cp_ring_info,
 			    const char *suffix);
+int bnxt_alloc_hwrm_rings(struct bnxt *bp);
 
 #endif
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 29/38] bnxt: add HWRM port PHY config call and helpers
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (27 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 28/38] bnxt: allocate and free all HWRM rings and groups Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 30/38] bnxt: add start/stop/link update operations Stephen Hurd
                       ` (9 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add HWRM calls to query the port's PHY and link configuration.

New HWRM call:
bnxt_hwrm_port_phy_qcfg
	This command queries the PHY configuration for the port

Also adding helper function like bnxt_get_hwrm_link_config()
and bnxt_parse_hw_link_speed() parse the link state.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.

v5:
Reorder footer
Rewrite the patch description.
---
 drivers/net/bnxt/bnxt_hwrm.c           | 120 +++++
 drivers/net/bnxt/bnxt_hwrm.h           |   1 +
 drivers/net/bnxt/hsi_struct_def_dpdk.h | 790 +++++++++++++++++++++++++++++++++
 3 files changed, 911 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_hwrm.c b/drivers/net/bnxt/bnxt_hwrm.c
index fcc4811..5d81a60 100644
--- a/drivers/net/bnxt/bnxt_hwrm.c
+++ b/drivers/net/bnxt/bnxt_hwrm.c
@@ -521,6 +521,43 @@ static int bnxt_hwrm_port_phy_cfg(struct bnxt *bp, struct bnxt_link_info *conf)
 	return rc;
 }
 
+static int bnxt_hwrm_port_phy_qcfg(struct bnxt *bp,
+				   struct bnxt_link_info *link_info)
+{
+	int rc = 0;
+	struct hwrm_port_phy_qcfg_input req = {.req_type = 0};
+	struct hwrm_port_phy_qcfg_output *resp = bp->hwrm_cmd_resp_addr;
+
+	HWRM_PREP(req, PORT_PHY_QCFG, -1, resp);
+
+	rc = bnxt_hwrm_send_message(bp, &req, sizeof(req));
+
+	HWRM_CHECK_RESULT;
+
+	link_info->phy_link_status = resp->link;
+	if (link_info->phy_link_status == HWRM_PORT_PHY_QCFG_OUTPUT_LINK_LINK) {
+		link_info->link_up = 1;
+		link_info->link_speed = rte_le_to_cpu_16(resp->link_speed);
+	} else {
+		link_info->link_up = 0;
+		link_info->link_speed = 0;
+	}
+	link_info->duplex = resp->duplex;
+	link_info->pause = resp->pause;
+	link_info->auto_pause = resp->auto_pause;
+	link_info->force_pause = resp->force_pause;
+	link_info->auto_mode = resp->auto_mode;
+
+	link_info->support_speeds = rte_le_to_cpu_16(resp->support_speeds);
+	link_info->auto_link_speed = rte_le_to_cpu_16(resp->auto_link_speed);
+	link_info->preemphasis = rte_le_to_cpu_32(resp->preemphasis);
+	link_info->phy_ver[0] = resp->phy_maj;
+	link_info->phy_ver[1] = resp->phy_min;
+	link_info->phy_ver[2] = resp->phy_bld;
+
+	return rc;
+}
+
 int bnxt_hwrm_queue_qportcfg(struct bnxt *bp)
 {
 	int rc = 0;
@@ -1326,6 +1363,89 @@ static uint16_t bnxt_parse_eth_link_speed_mask(uint32_t link_speed)
 	return ret;
 }
 
+static uint32_t bnxt_parse_hw_link_speed(uint16_t hw_link_speed)
+{
+	uint32_t eth_link_speed = ETH_SPEED_NUM_NONE;
+
+	switch (hw_link_speed) {
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB:
+		eth_link_speed = ETH_SPEED_NUM_100M;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_1GB:
+		eth_link_speed = ETH_SPEED_NUM_1G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2_5GB:
+		eth_link_speed = ETH_SPEED_NUM_2_5G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10GB:
+		eth_link_speed = ETH_SPEED_NUM_10G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_20GB:
+		eth_link_speed = ETH_SPEED_NUM_20G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_25GB:
+		eth_link_speed = ETH_SPEED_NUM_25G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_40GB:
+		eth_link_speed = ETH_SPEED_NUM_40G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_50GB:
+		eth_link_speed = ETH_SPEED_NUM_50G;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2GB:
+	default:
+		RTE_LOG(ERR, PMD, "HWRM link speed %d not defined\n",
+			hw_link_speed);
+		break;
+	}
+	return eth_link_speed;
+}
+
+static uint16_t bnxt_parse_hw_link_duplex(uint16_t hw_link_duplex)
+{
+	uint16_t eth_link_duplex = ETH_LINK_FULL_DUPLEX;
+
+	switch (hw_link_duplex) {
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_BOTH:
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_FULL:
+		eth_link_duplex = ETH_LINK_FULL_DUPLEX;
+		break;
+	case HWRM_PORT_PHY_CFG_INPUT_AUTO_DUPLEX_HALF:
+		eth_link_duplex = ETH_LINK_HALF_DUPLEX;
+		break;
+	default:
+		RTE_LOG(ERR, PMD, "HWRM link duplex %d not defined\n",
+			hw_link_duplex);
+		break;
+	}
+	return eth_link_duplex;
+}
+
+int bnxt_get_hwrm_link_config(struct bnxt *bp, struct rte_eth_link *link)
+{
+	int rc = 0;
+	struct bnxt_link_info *link_info = &bp->link_info;
+
+	rc = bnxt_hwrm_port_phy_qcfg(bp, link_info);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"Get link config failed with rc %d\n", rc);
+		goto exit;
+	}
+	if (link_info->link_up)
+		link->link_speed =
+			bnxt_parse_hw_link_speed(link_info->link_speed);
+	else
+		link->link_speed = ETH_LINK_SPEED_10M;
+	link->link_duplex = bnxt_parse_hw_link_duplex(link_info->duplex);
+	link->link_status = link_info->link_up;
+	link->link_autoneg = link_info->auto_mode ==
+		HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE ?
+		ETH_LINK_SPEED_FIXED : ETH_LINK_SPEED_AUTONEG;
+exit:
+	return rc;
+}
+
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up)
 {
 	int rc = 0;
diff --git a/drivers/net/bnxt/bnxt_hwrm.h b/drivers/net/bnxt/bnxt_hwrm.h
index c28e8d8..a508024 100644
--- a/drivers/net/bnxt/bnxt_hwrm.h
+++ b/drivers/net/bnxt/bnxt_hwrm.h
@@ -98,6 +98,7 @@ int bnxt_clear_hwrm_vnic_filters(struct bnxt *bp, struct bnxt_vnic_info *vnic);
 void bnxt_free_all_hwrm_resources(struct bnxt *bp);
 void bnxt_free_hwrm_resources(struct bnxt *bp);
 int bnxt_alloc_hwrm_resources(struct bnxt *bp);
+int bnxt_get_hwrm_link_config(struct bnxt *bp, struct rte_eth_link *link);
 int bnxt_set_hwrm_link_config(struct bnxt *bp, bool link_up);
 
 #endif
diff --git a/drivers/net/bnxt/hsi_struct_def_dpdk.h b/drivers/net/bnxt/hsi_struct_def_dpdk.h
index d58295a..f2db3ea 100644
--- a/drivers/net/bnxt/hsi_struct_def_dpdk.h
+++ b/drivers/net/bnxt/hsi_struct_def_dpdk.h
@@ -88,6 +88,7 @@ struct ctx_hw_stats64 {
 #define HWRM_FUNC_DRV_UNRGTR		(UINT32_C(0x1a))
 #define HWRM_FUNC_DRV_RGTR		(UINT32_C(0x1d))
 #define HWRM_PORT_PHY_CFG		(UINT32_C(0x20))
+#define HWRM_PORT_PHY_QCFG		(UINT32_C(0x27))
 #define HWRM_QUEUE_QPORTCFG		(UINT32_C(0x30))
 #define HWRM_VNIC_ALLOC			(UINT32_C(0x40))
 #define HWRM_VNIC_FREE			(UINT32_C(0x41))
@@ -2793,6 +2794,795 @@ struct hwrm_port_phy_cfg_output {
 	uint8_t valid;
 } __attribute__((packed));
 
+/* hwrm_port_phy_qcfg */
+/* Description: This command queries the PHY configuration for the port. */
+/* Input (24 bytes) */
+
+struct hwrm_port_phy_qcfg_input {
+	/*
+	 * This value indicates what type of request this is. The format for the
+	 * rest of the command is determined by this field.
+	 */
+	uint16_t req_type;
+
+	/*
+	 * This value indicates the what completion ring the request will be
+	 * optionally completed on. If the value is -1, then no CR completion
+	 * will be generated. Any other value must be a valid CR ring_id value
+	 * for this function.
+	 */
+	uint16_t cmpl_ring;
+
+	/* This value indicates the command sequence number. */
+	uint16_t seq_id;
+
+	/*
+	 * Target ID of this command. 0x0 - 0xFFF8 - Used for function ids
+	 * 0xFFF8 - 0xFFFE - Reserved for internal processors 0xFFFF - HWRM
+	 */
+	uint16_t target_id;
+
+	/*
+	 * This is the host address where the response will be written when the
+	 * request is complete. This area must be 16B aligned and must be
+	 * cleared to zero before the request is made.
+	 */
+	uint64_t resp_addr;
+
+	/* Port ID of port that is to be queried. */
+	uint16_t port_id;
+
+	uint16_t unused_0[3];
+} __attribute__((packed));
+
+/* Output (96 bytes) */
+struct hwrm_port_phy_qcfg_output {
+	/*
+	 * Pass/Fail or error type Note: receiver to verify the in parameters,
+	 * and fail the call with an error when appropriate
+	 */
+	uint16_t error_code;
+
+	/* This field returns the type of original request. */
+	uint16_t req_type;
+
+	/* This field provides original sequence number of the command. */
+	uint16_t seq_id;
+
+	/*
+	 * This field is the length of the response in bytes. The last byte of
+	 * the response is a valid flag that will read as '1' when the command
+	 * has been completely written to memory.
+	 */
+	uint16_t resp_len;
+
+	/* This value indicates the current link status. */
+		/* There is no link or cable detected. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_NO_LINK	(UINT32_C(0x0) << 0)
+		/* There is no link, but a cable has been detected. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SIGNAL	(UINT32_C(0x1) << 0)
+		/* There is a link. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_LINK	(UINT32_C(0x2) << 0)
+	uint8_t link;
+
+	uint8_t unused_0;
+
+	/* This value indicates the current link speed of the connection. */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t link_speed;
+
+	/* This value is indicates the duplex of the current connection. */
+		/* Half Duplex connection. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_DUPLEX_HALF	(UINT32_C(0x0) << 0)
+		/* Full duplex connection. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_DUPLEX_FULL	(UINT32_C(0x1) << 0)
+	uint8_t duplex;
+
+	/*
+	 * This value is used to indicate the current pause configuration. When
+	 * autoneg is enabled, this value represents the autoneg results of
+	 * pause configuration.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX	UINT32_C(0x2)
+	uint8_t pause;
+
+	/*
+	 * The supported speeds for the port. This is a bit mask. For each speed
+	 * that is supported, the corrresponding bit will be set to '1'.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10MB \
+							UINT32_C(0x2000)
+	uint16_t support_speeds;
+
+	/*
+	 * Current setting of forced link speed. When the link speed is not
+	 * being forced, this value shall be set to 0.
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t force_link_speed;
+
+	/* Current setting of auto negotiation mode. */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ALL_SPEEDS \
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_SPEED_MASK \
+							(UINT32_C(0x4) << 0)
+	uint8_t auto_mode;
+
+	/*
+	 * Current setting of pause autonegotiation. Move autoneg_pause flag
+	 * here.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_TX	UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages has been
+	 * requested. Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_RX	UINT32_C(0x2)
+	/*
+	 * When set to 1, the advertisement of pause is enabled. # When the
+	 * auto_mode is not set to none and this flag is set to 1, then the
+	 * auto_pause bits on this port are being advertised and autoneg pause
+	 * results are being interpreted. # When the auto_mode is not set to
+	 * none and this flag is set to 0, the pause is forced as indicated in
+	 * force_pause, and also advertised as auto_pause bits, but the autoneg
+	 * results are not interpreted since the pause configuration is being
+	 * forced. # When the auto_mode is set to none and this flag is set to
+	 * 1, auto_pause bits should be ignored and should be set to 0.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_PAUSE_AUTONEG_PAUSE \
+							UINT32_C(0x4)
+	uint8_t auto_pause;
+
+	/*
+	 * Current setting for auto_link_speed. This field is only valid when
+	 * auto_mode is set to "one_speed" or "one_or_below".
+	 */
+		/* 100Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_100MB \
+							(UINT32_C(0x1) << 0)
+		/* 1Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_1GB \
+							(UINT32_C(0xa) << 0)
+		/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_2GB \
+							(UINT32_C(0x14) << 0)
+		/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_2_5GB \
+							(UINT32_C(0x19) << 0)
+		/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_10GB \
+							(UINT32_C(0x64) << 0)
+		/* 20Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_20GB \
+							(UINT32_C(0xc8) << 0)
+		/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_25GB \
+							(UINT32_C(0xfa) << 0)
+		/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_40GB \
+							(UINT32_C(0x190) << 0)
+		/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_50GB \
+							(UINT32_C(0x1f4) << 0)
+		/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_100GB \
+							(UINT32_C(0x3e8) << 0)
+		/* 10Mb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_10MB \
+							(UINT32_C(0xffff) << 0)
+	uint16_t auto_link_speed;
+
+	/*
+	 * Current setting for auto_link_speed_mask that is used to advertise
+	 * speeds during autonegotiation. This field is only valid when
+	 * auto_mode is set to "mask". The speeds specified in this field shall
+	 * be a subset of supported speeds on this port.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_LINK_SPEED_MASK_10MB \
+							UINT32_C(0x2000)
+	uint16_t auto_link_speed_mask;
+
+	/* Current setting for wirespeed. */
+		/* Wirespeed feature is disabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_WIRESPEED_OFF	(UINT32_C(0x0) << 0)
+		/* Wirespeed feature is enabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_WIRESPEED_ON	(UINT32_C(0x1) << 0)
+	uint8_t wirespeed;
+
+	/* Current setting for loopback. */
+		/* No loopback is selected. Normal operation. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_NONE	(UINT32_C(0x0) << 0)
+		/*
+		 * The HW will be configured with local loopback such that host
+		 * data is sent back to the host without modification.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_LOCAL	(UINT32_C(0x1) << 0)
+		/*
+		 * The HW will be configured with remote loopback such that port
+		 * logic will send packets back out the transmitter that are
+		 * received.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LPBK_REMOTE	(UINT32_C(0x2) << 0)
+	uint8_t lpbk;
+
+	/*
+	 * Current setting of forced pause. When the pause configuration is not
+	 * being forced, then this value shall be set to 0.
+	 */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_PAUSE_TX \
+							UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_FORCE_PAUSE_RX \
+							UINT32_C(0x2)
+	uint8_t force_pause;
+
+	/*
+	 * This value indicates the current status of the optics module on this
+	 * port.
+	 */
+		/* Module is inserted and accepted */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Module is rejected and transmit side Laser is disabled. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_DISABLETX \
+							(UINT32_C(0x1) << 0)
+		/* Module mismatch warning. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG \
+							(UINT32_C(0x2) << 0)
+		/* Module is rejected and powered down. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_PWRDOWN \
+							(UINT32_C(0x3) << 0)
+		/* Module is not inserted. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NOTINSERTED \
+							(UINT32_C(0x4) << 0)
+		/* Module status is not applicable. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_NOTAPPLICABLE \
+							(UINT32_C(0xff) << 0)
+	uint8_t module_status;
+
+	/* Current setting for preemphasis. */
+	uint32_t preemphasis;
+
+	/* This field represents the major version of the PHY. */
+	uint8_t phy_maj;
+
+	/* This field represents the minor version of the PHY. */
+	uint8_t phy_min;
+
+	/* This field represents the build version of the PHY. */
+	uint8_t phy_bld;
+
+	/* This value represents a PHY type. */
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* BASE-CR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASECR \
+							(UINT32_C(0x1) << 0)
+		/* BASE-KR4 (Deprecated) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR4 \
+							(UINT32_C(0x2) << 0)
+		/* BASE-LR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASELR \
+							(UINT32_C(0x3) << 0)
+		/* BASE-SR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASESR \
+							(UINT32_C(0x4) << 0)
+		/* BASE-KR2 (Deprecated) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR2 \
+							(UINT32_C(0x5) << 0)
+		/* BASE-KX */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKX \
+							(UINT32_C(0x6) << 0)
+		/* BASE-KR */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASEKR \
+							(UINT32_C(0x7) << 0)
+		/* BASE-T */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASET \
+							(UINT32_C(0x8) << 0)
+		/* EEE capable BASE-T */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_BASETE \
+							(UINT32_C(0x9) << 0)
+		/* SGMII connected external PHY */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_TYPE_SGMIIEXTPHY \
+							(UINT32_C(0xa) << 0)
+	uint8_t phy_type;
+
+	/* This value represents a media type. */
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 0)
+		/* Twisted Pair */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_TP	(UINT32_C(0x1) << 0)
+		/* Direct Attached Copper */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_DAC \
+							(UINT32_C(0x2) << 0)
+		/* Fiber */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_MEDIA_TYPE_FIBRE \
+							(UINT32_C(0x3) << 0)
+	uint8_t media_type;
+
+	/* This value represents a transceiver type. */
+		/* PHY and MAC are in the same package */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_PKG_TYPE_XCVR_INTERNAL \
+							(UINT32_C(0x1) << 0)
+		/* PHY and MAC are in different packages */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_PKG_TYPE_XCVR_EXTERNAL \
+							(UINT32_C(0x2) << 0)
+	uint8_t xcvr_pkg_type;
+
+	/*
+	 * This field represents flags related to EEE configuration. These EEE
+	 * configuration flags are valid only when the auto_mode is not set to
+	 * none (in other words autonegotiation is enabled).
+	 */
+	/* This field represents PHY address. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_ADDR_MASK	UINT32_C(0x1f)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PHY_ADDR_SFT	0
+	/*
+	 * When set to 1, Energy Efficient Ethernet (EEE) mode is enabled.
+	 * Speeds for autoneg with EEE mode enabled are based on
+	 * eee_link_speed_mask.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_ENABLED \
+							UINT32_C(0x20)
+	/*
+	 * This flag is valid only when eee_enabled is set to 1. # If
+	 * eee_enabled is set to 0, then EEE mode is disabled and this flag
+	 * shall be ignored. # If eee_enabled is set to 1 and this flag is set
+	 * to 1, then Energy Efficient Ethernet (EEE) mode is enabled and in
+	 * use. # If eee_enabled is set to 1 and this flag is set to 0, then
+	 * Energy Efficient Ethernet (EEE) mode is enabled but is currently not
+	 * in use.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_ACTIVE \
+							UINT32_C(0x40)
+	/*
+	 * This flag is valid only when eee_enabled is set to 1. # If
+	 * eee_enabled is set to 0, then EEE mode is disabled and this flag
+	 * shall be ignored. # If eee_enabled is set to 1 and this flag is set
+	 * to 1, then Energy Efficient Ethernet (EEE) mode is enabled and TX LPI
+	 * is enabled. # If eee_enabled is set to 1 and this flag is set to 0,
+	 * then Energy Efficient Ethernet (EEE) mode is enabled but TX LPI is
+	 * disabled.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_EEE_TX_LPI \
+							UINT32_C(0x80)
+	/*
+	 * This field represents flags related to EEE configuration. These EEE
+	 * configuration flags are valid only when the auto_mode is not set to
+	 * none (in other words autonegotiation is enabled).
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_MASK \
+							UINT32_C(0xe0)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_EEE_CONFIG_SFT	5
+	uint8_t eee_config_phy_addr;
+
+	/* Reserved field, set to 0 */
+	/*
+	 * When set to 1, the parallel detection is used to determine the speed
+	 * of the link partner. Parallel detection is used when a
+	 * autonegotiation capable device is connected to a link parter that is
+	 * not capable of autonegotiation.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_PARALLEL_DETECT \
+							UINT32_C(0x1)
+	/* Reserved field, set to 0 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_RESERVED_MASK	UINT32_C(0xfe)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_RESERVED_SFT	1
+	uint8_t parallel_detect;
+
+	/*
+	 * The advertised speeds for the port by the link partner. Each
+	 * advertised speed will be set to '1'.
+	 */
+	/* 100Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100MBHD \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100MB \
+							UINT32_C(0x2)
+	/* 1Gb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_1GBHD \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_1GB \
+							UINT32_C(0x8)
+	/* 2Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_2GB \
+							UINT32_C(0x10)
+	/* 2.5Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_2_5GB \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10GB \
+							UINT32_C(0x40)
+	/* 20Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_20GB \
+							UINT32_C(0x80)
+	/* 25Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_25GB \
+							UINT32_C(0x100)
+	/* 40Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_40GB \
+							UINT32_C(0x200)
+	/* 50Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_50GB \
+							UINT32_C(0x400)
+	/* 100Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_100GB \
+							UINT32_C(0x800)
+	/* 10Mb link speed (Half-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10MBHD \
+							UINT32_C(0x1000)
+	/* 10Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_SPEEDS_10MB \
+							UINT32_C(0x2000)
+	uint16_t link_partner_adv_speeds;
+
+	/*
+	 * The advertised autoneg for the port by the link partner. This field
+	 * is deprecated and should be set to 0.
+	 */
+		/*
+		 * Disable autoneg or autoneg disabled. No speeds are selected.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_NONE \
+							(UINT32_C(0x0) << 0)
+		/* Select all possible speeds for autoneg mode. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ALL_SPEEDS\
+							(UINT32_C(0x1) << 0)
+		/*
+		 * Select only the auto_link_speed speed for autoneg mode. This
+		 * mode has been DEPRECATED. An HWRM client should not use this
+		 * mode.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ONE_SPEED \
+							(UINT32_C(0x2) << 0)
+		/*
+		 * Select the auto_link_speed or any speed below that speed for
+		 * autoneg. This mode has been DEPRECATED. An HWRM client should
+		 * not use this mode.
+		 */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_ONE_OR_BELOW \
+							(UINT32_C(0x3) << 0)
+		/*
+		 * Select the speeds based on the corresponding link speed mask
+		 * value that is provided.
+		 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_AUTO_MODE_SPEED_MASK\
+							(UINT32_C(0x4) << 0)
+	uint8_t link_partner_adv_auto_mode;
+
+	/* The advertised pause settings on the port by the link partner. */
+	/*
+	 * When this bit is '1', Generation of tx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_PAUSE_TX \
+							UINT32_C(0x1)
+	/*
+	 * When this bit is '1', Reception of rx pause messages is supported.
+	 * Disabled otherwise.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_PAUSE_RX \
+							UINT32_C(0x2)
+	uint8_t link_partner_adv_pause;
+
+	/*
+	 * Current setting for link speed mask that is used to advertise speeds
+	 * during autonegotiation when EEE is enabled. This field is valid only
+	 * when eee_enabled flags is set to 1. The speeds specified in this
+	 * field shall be a subset of speeds specified in auto_link_speed_mask.
+	 */
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD1 \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD2 \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD3 \
+							UINT32_C(0x10)
+	/* Reserved */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_RSVD4 \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_ADV_EEE_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	uint16_t adv_eee_link_speed_mask;
+
+	/*
+	 * Current setting for link speed mask that is advertised by the link
+	 * partner when EEE is enabled. This field is valid only when
+	 * eee_enabled flags is set to 1.
+	 */
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD1 \
+							UINT32_C(0x1)
+	/* 100Mb link speed (Full-duplex) */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_100MB \
+							UINT32_C(0x2)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD2 \
+							UINT32_C(0x4)
+	/* 1Gb link speed (Full-duplex) */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_1GB \
+							UINT32_C(0x8)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD3 \
+							UINT32_C(0x10)
+	/* Reserved */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_RSVD4 \
+							UINT32_C(0x20)
+	/* 10Gb link speed */
+	#define \
+	HWRM_PORT_PHY_QCFG_OUTPUT_LINK_PARTNER_ADV_EEE_LINK_SPEED_MASK_10GB \
+							UINT32_C(0x40)
+	uint16_t link_partner_adv_eee_link_speed_mask;
+
+	/* This value represents transceiver identifier type. */
+	/*
+	 * Current setting of TX LPI timer in microseconds. This field is valid
+	 * only when_eee_enabled flag is set to 1 and tx_lpi_enabled is set to
+	 * 1.
+	 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_TX_LPI_TIMER_MASK \
+							UINT32_C(0xffffff)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_TX_LPI_TIMER_SFT         0
+	/* This value represents transceiver identifier type. */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_MASK \
+							UINT32_C(0xff000000)
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_SFT \
+							24
+		/* Unknown */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_UNKNOWN \
+							(UINT32_C(0x0) << 24)
+		/* SFP/SFP+/SFP28 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_SFP \
+							(UINT32_C(0x3) << 24)
+		/* QSFP */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFP \
+							(UINT32_C(0xc) << 24)
+		/* QSFP+ */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFPPLUS \
+							(UINT32_C(0xd) << 24)
+		/* QSFP28 */
+	#define HWRM_PORT_PHY_QCFG_OUTPUT_XCVR_IDENTIFIER_TYPE_QSFP28 \
+							(UINT32_C(0x11) << 24)
+	uint32_t xcvr_identifier_type_tx_lpi_timer;
+
+	uint32_t unused_1;
+
+	/*
+	 * Up to 16 bytes of null padded ASCII string representing PHY vendor.
+	 * If the string is set to null, then the vendor name is not available.
+	 */
+	char phy_vendor_name[16];
+
+	/*
+	 * Up to 16 bytes of null padded ASCII string that identifies vendor
+	 * specific part number of the PHY. If the string is set to null, then
+	 * the vendor specific part number is not available.
+	 */
+	char phy_vendor_partnumber[16];
+
+	uint32_t unused_2;
+	uint8_t unused_3;
+	uint8_t unused_4;
+	uint8_t unused_5;
+
+	/*
+	 * This field is used in Output records to indicate that the output is
+	 * completely written to RAM. This field should be read as '1' to
+	 * indicate that the output has been completely written. When writing a
+	 * command completion or response to an internal processor, the order of
+	 * writes has to be such that this field is written last.
+	 */
+	uint8_t valid;
+} __attribute__((packed));
+
 /* hwrm_ver_get */
 /*
  * Description: This function is called by a driver to determine the HWRM
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 30/38] bnxt: add start/stop/link update operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (28 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 29/38] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 31/38] bnxt: add promiscuous enable/disable operations Stephen Hurd
                       ` (8 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to add the start, stop and link update dev_ops.

Also adds wrapper functions like bnxt_init_chip(), bnxt_init_nic(),
bnxt_alloc_mem(), bnxt_free_mem()

The BNXT driver will now minimally pass traffic with testpmd.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.
Shorten the string passed for reserving memzone when default completion
ring is created.

v5:
Reorder footer

v6:
Correct typos
---
 drivers/net/bnxt/bnxt_ethdev.c | 269 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 269 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 6888363..ac82876 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -40,12 +40,17 @@
 #include <rte_cycles.h>
 
 #include "bnxt.h"
+#include "bnxt_cpr.h"
+#include "bnxt_filter.h"
 #include "bnxt_hwrm.h"
+#include "bnxt_ring.h"
 #include "bnxt_rxq.h"
 #include "bnxt_rxr.h"
 #include "bnxt_stats.h"
 #include "bnxt_txq.h"
 #include "bnxt_txr.h"
+#include "bnxt_vnic.h"
+#include "hsi_struct_def_dpdk.h"
 
 #define DRV_MODULE_NAME		"bnxt"
 static const char bnxt_version[] =
@@ -65,6 +70,177 @@ static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 	bnxt_free_hwrm_resources(bp);
 }
 
+/***********************/
+
+/*
+ * High level utility functions
+ */
+
+static void bnxt_free_mem(struct bnxt *bp)
+{
+	bnxt_free_filter_mem(bp);
+	bnxt_free_vnic_attributes(bp);
+	bnxt_free_vnic_mem(bp);
+
+	bnxt_free_stats(bp);
+	bnxt_free_tx_rings(bp);
+	bnxt_free_rx_rings(bp);
+	bnxt_free_def_cp_ring(bp);
+}
+
+static int bnxt_alloc_mem(struct bnxt *bp)
+{
+	int rc;
+
+	/* Default completion ring */
+	rc = bnxt_init_def_ring_struct(bp, SOCKET_ID_ANY);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_rings(bp, 0, NULL, NULL,
+			      bp->def_cp_ring, "def_cp");
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_vnic_mem(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_vnic_attributes(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	rc = bnxt_alloc_filter_mem(bp);
+	if (rc)
+		goto alloc_mem_err;
+
+	return 0;
+
+alloc_mem_err:
+	bnxt_free_mem(bp);
+	return rc;
+}
+
+static int bnxt_init_chip(struct bnxt *bp)
+{
+	unsigned int i, rss_idx, fw_idx;
+	int rc;
+
+	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM stat ctx alloc failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_alloc_hwrm_rings(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM ring alloc failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "HWRM ring grp alloc failure: %x\n", rc);
+		goto err_out;
+	}
+
+	rc = bnxt_mq_rx_configure(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "MQ mode configure failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	/* VNIC configuration */
+	for (i = 0; i < bp->nr_vnics; i++) {
+		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
+
+		rc = bnxt_hwrm_vnic_alloc(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic alloc failure rc: %x\n",
+				rc);
+			goto err_out;
+		}
+
+		rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD,
+				"HWRM vnic ctx alloc failure rc: %x\n", rc);
+			goto err_out;
+		}
+
+		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic cfg failure rc: %x\n", rc);
+			goto err_out;
+		}
+
+		rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
+		if (rc) {
+			RTE_LOG(ERR, PMD, "HWRM vnic filter failure rc: %x\n",
+				rc);
+			goto err_out;
+		}
+		if (vnic->rss_table && vnic->hash_type) {
+			/*
+			 * Fill the RSS hash & redirection table with
+			 * ring group ids for all VNICs
+			 */
+			for (rss_idx = 0, fw_idx = 0;
+			     rss_idx < HW_HASH_INDEX_SIZE;
+			     rss_idx++, fw_idx++) {
+				if (vnic->fw_grp_ids[fw_idx] ==
+				    INVALID_HW_RING_ID)
+					fw_idx = 0;
+				vnic->rss_table[rss_idx] =
+						vnic->fw_grp_ids[fw_idx];
+			}
+			rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+			if (rc) {
+				RTE_LOG(ERR, PMD,
+					"HWRM vnic set RSS failure rc: %x\n",
+					rc);
+				goto err_out;
+			}
+		}
+	}
+	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0]);
+	if (rc) {
+		RTE_LOG(ERR, PMD,
+			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
+		goto err_out;
+	}
+
+	return 0;
+
+err_out:
+	bnxt_free_all_hwrm_resources(bp);
+
+	return rc;
+}
+
+static int bnxt_shutdown_nic(struct bnxt *bp)
+{
+	bnxt_free_all_hwrm_resources(bp);
+	bnxt_free_all_filters(bp);
+	bnxt_free_all_vnics(bp);
+	return 0;
+}
+
+static int bnxt_init_nic(struct bnxt *bp)
+{
+	int rc;
+
+	bnxt_init_ring_grps(bp);
+	bnxt_init_vnics(bp);
+	bnxt_init_filters(bp);
+
+	rc = bnxt_init_chip(bp);
+	if (rc)
+		return rc;
+
+	return 0;
+}
+
 /*
  * Device configuration and status function
  */
@@ -182,6 +358,85 @@ static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
 	return rc;
 }
 
+static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	int rc;
+
+	rc = bnxt_hwrm_func_reset(bp);
+	if (rc) {
+		RTE_LOG(ERR, PMD, "hwrm chip reset failure rc: %x\n", rc);
+		rc = -1;
+		goto error;
+	}
+
+	rc = bnxt_alloc_mem(bp);
+	if (rc)
+		goto error;
+
+	rc = bnxt_init_nic(bp);
+	if (rc)
+		goto error;
+
+	return 0;
+
+error:
+	bnxt_shutdown_nic(bp);
+	bnxt_free_tx_mbufs(bp);
+	bnxt_free_rx_mbufs(bp);
+	bnxt_free_mem(bp);
+	return rc;
+}
+
+/* Unload the driver, release resources */
+static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	if (bp->eth_dev->data->dev_started) {
+		/* TBD: STOP HW queues DMA */
+		eth_dev->data->dev_link.link_status = 0;
+	}
+	bnxt_shutdown_nic(bp);
+}
+
+static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
+			       int wait_to_complete)
+{
+	int rc = 0;
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_link new;
+	unsigned int cnt = BNXT_LINK_WAIT_CNT;
+
+	memset(&new, 0, sizeof(new));
+	do {
+		/* Retrieve link info from hardware */
+		rc = bnxt_get_hwrm_link_config(bp, &new);
+		if (rc) {
+			new.link_speed = ETH_LINK_SPEED_100M;
+			new.link_duplex = ETH_LINK_FULL_DUPLEX;
+			RTE_LOG(ERR, PMD,
+				"Failed to retrieve link rc = 0x%x!", rc);
+			goto out;
+		}
+		if (!wait_to_complete)
+			break;
+
+		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
+
+	} while (!new.link_status && cnt--);
+
+	/* Timed out or success */
+	if (new.link_status) {
+		/* Update only if success */
+		eth_dev->data->dev_link.link_duplex = new.link_duplex;
+		eth_dev->data->dev_link.link_speed = new.link_speed;
+	}
+	eth_dev->data->dev_link.link_status = new.link_status;
+out:
+	return rc;
+}
+
 /*
  * Initialization
  */
@@ -190,12 +445,15 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_infos_get = bnxt_dev_info_get_op,
 	.dev_close = bnxt_dev_close_op,
 	.dev_configure = bnxt_dev_configure_op,
+	.dev_start = bnxt_dev_start_op,
+	.dev_stop = bnxt_dev_stop_op,
 	.stats_get = bnxt_stats_get_op,
 	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
+	.link_update = bnxt_link_update_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
@@ -305,6 +563,15 @@ bnxt_dev_init(struct rte_eth_dev *eth_dev)
 	else
 		memcpy(bp->mac_addr, bp->vf.mac_addr, sizeof(bp->mac_addr));
 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, ETHER_ADDR_LEN);
+	bp->grp_info = rte_zmalloc("bnxt_grp_info",
+				sizeof(*bp->grp_info) * bp->max_ring_grps, 0);
+	if (!bp->grp_info) {
+		RTE_LOG(ERR, PMD,
+			"Failed to alloc %zu bytes needed to store group info table\n",
+			sizeof(*bp->grp_info) * bp->max_ring_grps);
+		rc = -ENOMEM;
+		goto error_free;
+	}
 
 	rc = bnxt_hwrm_func_driver_register(bp, 0,
 					    bp->pf.vf_req_fwd);
@@ -335,6 +602,8 @@ bnxt_dev_uninit(struct rte_eth_dev *eth_dev) {
 
 	if (eth_dev->data->mac_addrs)
 		rte_free(eth_dev->data->mac_addrs);
+	if (bp->grp_info)
+		rte_free(bp->grp_info);
 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
 	bnxt_free_hwrm_resources(bp);
 	return rc;
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 31/38] bnxt: add promiscuous enable/disable operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (29 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 30/38] bnxt: add start/stop/link update operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 32/38] bnxt: add all multicast " Stephen Hurd
                       ` (7 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds the promiscuous mode enable and disable dev_ops.
Uses the bnxt_hwrm_cfa_l2_set_rx_mask() API added in the earlier patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix couple of typos in the commit message.

v5:
Reorder footer
Rewrite the patch description.
---
 drivers/net/bnxt/bnxt_ethdev.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index ac82876..3fce540 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -437,6 +437,34 @@ out:
 	return rc;
 }
 
+static void bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
+static void bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
 /*
  * Initialization
  */
@@ -454,6 +482,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
 	.link_update = bnxt_link_update_op,
+	.promiscuous_enable = bnxt_promiscuous_enable_op,
+	.promiscuous_disable = bnxt_promiscuous_disable_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 32/38] bnxt: add all multicast enable/disable operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (30 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 31/38] bnxt: add promiscuous enable/disable operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 33/38] bnxt: free memory in close operation Stephen Hurd
                       ` (6 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds dev_ops to enable/disable multicast traffic.
Uses the bnxt_hwrm_cfa_l2_set_rx_mask() API added in the earlier patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder footer
Rewrite the patch description.
---
 drivers/net/bnxt/bnxt_ethdev.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3fce540..d3a624f 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -465,6 +465,34 @@ static void bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
 	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
 }
 
+static void bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
+static void bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic;
+
+	if (bp->vnic_info == NULL)
+		return;
+
+	vnic = &bp->vnic_info[0];
+
+	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
+	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
+}
+
 /*
  * Initialization
  */
@@ -484,6 +512,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
+	.allmulticast_enable = bnxt_allmulticast_enable_op,
+	.allmulticast_disable = bnxt_allmulticast_disable_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 33/38] bnxt: free memory in close operation
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (31 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 32/38] bnxt: add all multicast " Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 34/38] bnxt: add MAC address add/remove dev ops Stephen Hurd
                       ` (5 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds code to free all resources except the one corresponding
to HWRM, which are required to notify the HWRM that the driver is unloaded
(these are freed in uninit()).

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index d3a624f..4254531 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -62,14 +62,6 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
-static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
-{
-	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
-
-	rte_free(eth_dev->data->mac_addrs);
-	bnxt_free_hwrm_resources(bp);
-}
-
 /***********************/
 
 /*
@@ -388,6 +380,16 @@ error:
 	return rc;
 }
 
+static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	bnxt_free_tx_mbufs(bp);
+	bnxt_free_rx_mbufs(bp);
+	bnxt_free_mem(bp);
+	rte_free(eth_dev->data->mac_addrs);
+}
+
 /* Unload the driver, release resources */
 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 {
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 34/38] bnxt: add MAC address add/remove dev ops
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (32 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 33/38] bnxt: free memory in close operation Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 35/38] bnxt: add set link up/down operations Stephen Hurd
                       ` (4 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

This patch adds dev_ops to Add/Remove MAC addresses.
These use the bnxt_hwrm_set_filter() defined in the previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 71 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 71 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 4254531..34a5873 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -402,6 +402,75 @@ static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 	bnxt_shutdown_nic(bp);
 }
 
+static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
+				    uint32_t index)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
+	struct bnxt_vnic_info *vnic;
+	struct bnxt_filter_info *filter, *temp_filter;
+	int i;
+
+	/*
+	 * Loop through all VNICs from the specified filter flow pools to
+	 * remove the corresponding MAC addr filter
+	 */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		if (!(pool_mask & (1 << i)))
+			continue;
+
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			filter = STAILQ_FIRST(&vnic->filter);
+			while (filter) {
+				temp_filter = STAILQ_NEXT(filter, next);
+				if (filter->mac_index == index) {
+					STAILQ_REMOVE(&vnic->filter, filter,
+						      bnxt_filter_info, next);
+					bnxt_hwrm_clear_filter(bp, filter);
+					filter->mac_index = INVALID_MAC_INDEX;
+					memset(&filter->l2_addr, 0,
+					       ETHER_ADDR_LEN);
+					STAILQ_INSERT_TAIL(
+							&bp->free_filter_list,
+							filter, next);
+				}
+				filter = temp_filter;
+			}
+		}
+	}
+}
+
+static void bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
+				 struct ether_addr *mac_addr,
+				 uint32_t index, uint32_t pool)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = STAILQ_FIRST(&bp->ff_pool[pool]);
+	struct bnxt_filter_info *filter;
+
+	if (!vnic) {
+		RTE_LOG(ERR, PMD, "VNIC not found for pool %d!\n", pool);
+		return;
+	}
+	/* Attach requested MAC address to the new l2_filter */
+	STAILQ_FOREACH(filter, &vnic->filter, next) {
+		if (filter->mac_index == index) {
+			RTE_LOG(ERR, PMD,
+				"MAC addr already existed for pool %d\n", pool);
+			return;
+		}
+	}
+	filter = bnxt_alloc_filter(bp);
+	if (!filter) {
+		RTE_LOG(ERR, PMD, "L2 filter alloc failed\n");
+		return;
+	}
+	STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
+	filter->mac_index = index;
+	memcpy(filter->l2_addr, mac_addr, ETHER_ADDR_LEN);
+	bnxt_hwrm_set_filter(bp, vnic, filter);
+}
+
 static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
 			       int wait_to_complete)
 {
@@ -516,6 +585,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
 	.allmulticast_enable = bnxt_allmulticast_enable_op,
 	.allmulticast_disable = bnxt_allmulticast_disable_op,
+	.mac_addr_add = bnxt_mac_addr_add_op,
+	.mac_addr_remove = bnxt_mac_addr_remove_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 35/38] bnxt: add set link up/down operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (33 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 34/38] bnxt: add MAC address add/remove dev ops Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 36/38] bnxt: add reta update/query operations Stephen Hurd
                       ` (3 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Adds dev_ops to set link UP or DOWN as appropriate.
Uses the bnxt_set_hwrm_link_config() API added in previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 34a5873..b04010c 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -380,6 +380,24 @@ error:
 	return rc;
 }
 
+static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	eth_dev->data->dev_link.link_status = 1;
+	bnxt_set_hwrm_link_config(bp, true);
+	return 0;
+}
+
+static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+
+	eth_dev->data->dev_link.link_status = 0;
+	bnxt_set_hwrm_link_config(bp, false);
+	return 0;
+}
+
 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 {
 	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
@@ -574,6 +592,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.dev_configure = bnxt_dev_configure_op,
 	.dev_start = bnxt_dev_start_op,
 	.dev_stop = bnxt_dev_stop_op,
+	.dev_set_link_up = bnxt_dev_set_link_up_op,
+	.dev_set_link_down = bnxt_dev_set_link_down_op,
 	.stats_get = bnxt_stats_get_op,
 	.stats_reset = bnxt_stats_reset_op,
 	.rx_queue_setup = bnxt_rx_queue_setup_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 36/38] bnxt: add reta update/query operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (34 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 35/38] bnxt: add set link up/down operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 37/38] bnxt: add RSS device operations Stephen Hurd
                       ` (2 subsequent siblings)
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add code to Update/query reta dev_ops
Uses the bnxt_hwrm_vnic_rss_cfg() HWRM API added earlier.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 56 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 56 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b04010c..b3b76f1 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -582,6 +582,60 @@ static void bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
 	bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic);
 }
 
+static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
+			    struct rte_eth_rss_reta_entry64 *reta_conf,
+			    uint16_t reta_size)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_vnic_info *vnic;
+	int i;
+
+	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
+		return -EINVAL;
+
+	if (reta_size != HW_HASH_INDEX_SIZE) {
+		RTE_LOG(ERR, PMD, "The configured hash table lookup size "
+			"(%d) must equal the size supported by the hardware "
+			"(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+		return -EINVAL;
+	}
+	/* Update the RSS VNIC(s) */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			memcpy(vnic->rss_table, reta_conf, reta_size);
+
+			bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+		}
+	}
+	return 0;
+}
+
+static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
+			      struct rte_eth_rss_reta_entry64 *reta_conf,
+			      uint16_t reta_size)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+
+	/* Retrieve from the default VNIC */
+	if (!vnic)
+		return -EINVAL;
+	if (!vnic->rss_table)
+		return -EINVAL;
+
+	if (reta_size != HW_HASH_INDEX_SIZE) {
+		RTE_LOG(ERR, PMD, "The configured hash table lookup size "
+			"(%d) must equal the size supported by the hardware "
+			"(%d)\n", reta_size, HW_HASH_INDEX_SIZE);
+		return -EINVAL;
+	}
+	/* EW - need to revisit here copying from u64 to u16 */
+	memcpy(reta_conf, vnic->rss_table, reta_size);
+
+	return 0;
+}
+
 /*
  * Initialization
  */
@@ -600,6 +654,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.rx_queue_release = bnxt_rx_queue_release_op,
 	.tx_queue_setup = bnxt_tx_queue_setup_op,
 	.tx_queue_release = bnxt_tx_queue_release_op,
+	.reta_update = bnxt_reta_update_op,
+	.reta_query = bnxt_reta_query_op,
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 37/38] bnxt: add RSS device operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (35 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 36/38] bnxt: add reta update/query operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 38/38] bnxt: add flow control operations Stephen Hurd
  2016-06-16 14:24     ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Bruce Richardson
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add rss_hash_update and rss_hash_conf_get dev_ops
Uses the bnxt_hwrm_vnic_rss_cfg() HWRM API added in the previous patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v4:
Fix issues pointed out by checkpatch.

v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 121 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 121 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index b3b76f1..3c7f868 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -62,6 +62,14 @@ static struct rte_pci_id bnxt_pci_id_map[] = {
 	{.device_id = 0},
 };
 
+#define BNXT_ETH_RSS_SUPPORT (	\
+	ETH_RSS_IPV4 |		\
+	ETH_RSS_NONFRAG_IPV4_TCP |	\
+	ETH_RSS_NONFRAG_IPV4_UDP |	\
+	ETH_RSS_IPV6 |		\
+	ETH_RSS_NONFRAG_IPV6_TCP |	\
+	ETH_RSS_NONFRAG_IPV6_UDP)
+
 /***********************/
 
 /*
@@ -636,6 +644,117 @@ static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
 	return 0;
 }
 
+static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
+				   struct rte_eth_rss_conf *rss_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
+	struct bnxt_vnic_info *vnic;
+	uint16_t hash_type = 0;
+	int i;
+
+	/*
+	 * If RSS enablement were different than dev_configure,
+	 * then return -EINVAL
+	 */
+	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
+		if (!rss_conf->rss_hf)
+			return -EINVAL;
+	} else {
+		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
+			return -EINVAL;
+	}
+	if (rss_conf->rss_hf & ETH_RSS_IPV4)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV4_TCP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV4_UDP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
+	if (rss_conf->rss_hf & ETH_RSS_IPV6)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV6_TCP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
+	if (rss_conf->rss_hf & ETH_RSS_NONFRAG_IPV6_UDP)
+		hash_type |= HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
+
+	/* Update the RSS VNIC(s) */
+	for (i = 0; i < MAX_FF_POOLS; i++) {
+		STAILQ_FOREACH(vnic, &bp->ff_pool[i], next) {
+			vnic->hash_type = hash_type;
+
+			/*
+			 * Use the supplied key if the key length is
+			 * acceptable and the rss_key is not NULL
+			 */
+			if (rss_conf->rss_key &&
+			    rss_conf->rss_key_len <= HW_HASH_KEY_SIZE)
+				memcpy(vnic->rss_hash_key, rss_conf->rss_key,
+				       rss_conf->rss_key_len);
+
+			bnxt_hwrm_vnic_rss_cfg(bp, vnic);
+		}
+	}
+	return 0;
+}
+
+static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
+				     struct rte_eth_rss_conf *rss_conf)
+{
+	struct bnxt *bp = (struct bnxt *)eth_dev->data->dev_private;
+	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
+	int len;
+	uint32_t hash_types;
+
+	/* RSS configuration is the same for all VNICs */
+	if (vnic && vnic->rss_hash_key) {
+		if (rss_conf->rss_key) {
+			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
+			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
+			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
+		}
+
+		hash_types = vnic->hash_type;
+		rss_conf->rss_hf = 0;
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_IPV4;
+			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_IPV6;
+			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
+		}
+		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
+			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
+			hash_types &=
+				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
+		}
+		if (hash_types) {
+			RTE_LOG(ERR, PMD,
+				"Unknwon RSS config from firmware (%08x), RSS disabled",
+				vnic->hash_type);
+			return -ENOTSUP;
+		}
+	} else {
+		rss_conf->rss_hf = 0;
+	}
+	return 0;
+}
+
 /*
  * Initialization
  */
@@ -656,6 +775,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.tx_queue_release = bnxt_tx_queue_release_op,
 	.reta_update = bnxt_reta_update_op,
 	.reta_query = bnxt_reta_query_op,
+	.rss_hash_update = bnxt_rss_hash_update_op,
+	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
 	.link_update = bnxt_link_update_op,
 	.promiscuous_enable = bnxt_promiscuous_enable_op,
 	.promiscuous_disable = bnxt_promiscuous_disable_op,
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* [dpdk-dev] [PATCH v6 38/38] bnxt: add flow control operations
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (36 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 37/38] bnxt: add RSS device operations Stephen Hurd
@ 2016-06-15 21:23     ` Stephen Hurd
  2016-06-16 14:24     ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Bruce Richardson
  38 siblings, 0 replies; 132+ messages in thread
From: Stephen Hurd @ 2016-06-15 21:23 UTC (permalink / raw)
  To: dev, ajit.khaparde, bruce.richardson

From: Ajit Khaparde <ajit.khaparde@broadcom.com>

Add flow_ctrl_get and flow_ctrl_set dev_ops.
Uses the bnxt_set_hwrm_link_config() HWRM API added in earlier patch.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Stephen Hurd <stephen.hurd@broadcom.com>
Reviewed-by: David Christensen <david.christensen@broadcom.com>

--
v5:
Rewrite the patch description.
Reorder the footer.
---
 drivers/net/bnxt/bnxt_ethdev.c | 83 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 83 insertions(+)

diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
index 3c7f868..406e38a 100644
--- a/drivers/net/bnxt/bnxt_ethdev.c
+++ b/drivers/net/bnxt/bnxt_ethdev.c
@@ -755,6 +755,87 @@ static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
 	return 0;
 }
 
+static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
+			       struct rte_eth_fc_conf *fc_conf __rte_unused)
+{
+	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+	struct rte_eth_link link_info;
+	int rc;
+
+	rc = bnxt_get_hwrm_link_config(bp, &link_info);
+	if (rc)
+		return rc;
+
+	memset(fc_conf, 0, sizeof(*fc_conf));
+	if (bp->link_info.auto_pause)
+		fc_conf->autoneg = 1;
+	switch (bp->link_info.pause) {
+	case 0:
+		fc_conf->mode = RTE_FC_NONE;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
+		fc_conf->mode = RTE_FC_TX_PAUSE;
+		break;
+	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
+		fc_conf->mode = RTE_FC_RX_PAUSE;
+		break;
+	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
+			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
+		fc_conf->mode = RTE_FC_FULL;
+		break;
+	}
+	return 0;
+}
+
+static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
+			       struct rte_eth_fc_conf *fc_conf)
+{
+	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
+
+	switch (fc_conf->mode) {
+	case RTE_FC_NONE:
+		bp->link_info.auto_pause = 0;
+		bp->link_info.force_pause = 0;
+		break;
+	case RTE_FC_RX_PAUSE:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
+		}
+		break;
+	case RTE_FC_TX_PAUSE:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
+		}
+		break;
+	case RTE_FC_FULL:
+		if (fc_conf->autoneg) {
+			bp->link_info.auto_pause =
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
+					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
+			bp->link_info.force_pause = 0;
+		} else {
+			bp->link_info.auto_pause = 0;
+			bp->link_info.force_pause =
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
+					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
+		}
+		break;
+	}
+	return bnxt_set_hwrm_link_config(bp, true);
+}
+
 /*
  * Initialization
  */
@@ -784,6 +865,8 @@ static struct eth_dev_ops bnxt_dev_ops = {
 	.allmulticast_disable = bnxt_allmulticast_disable_op,
 	.mac_addr_add = bnxt_mac_addr_add_op,
 	.mac_addr_remove = bnxt_mac_addr_remove_op,
+	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
+	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
 };
 
 static bool bnxt_vf_pciid(uint16_t id)
-- 
1.9.1

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code
  2016-06-15 20:28       ` Stephen Hurd
@ 2016-06-16  9:46         ` Bruce Richardson
  0 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-16  9:46 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: Ferruh Yigit, dev, Ajit Kumar Khaparde

On Wed, Jun 15, 2016 at 01:28:13PM -0700, Stephen Hurd wrote:
> On Wed, Jun 15, 2016 at 9:25 AM, Ferruh Yigit <ferruh.yigit@intel.com>
> wrote:
> 
> >
> > Hi Stephen,
> >
> > I run basic checks on the patch set, mainly it compiles and applies
> > fine, overall patch set looks good.
> >
> > Only get following checkpatch warning:
> > Only v5,02/38 has one WARNING:MACRO_WITH_FLOW_CONTROL
> >
> 
> This is intentional, is this style of copy/paste avoidance not allowed?
> 
It's fine. Some checkpatch exceptions like this are allowed when they make sense.

Regards,
/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library
  2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
                       ` (37 preceding siblings ...)
  2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 38/38] bnxt: add flow control operations Stephen Hurd
@ 2016-06-16 14:24     ` Bruce Richardson
  2016-06-16 18:51       ` Ajit Khaparde
  38 siblings, 1 reply; 132+ messages in thread
From: Bruce Richardson @ 2016-06-16 14:24 UTC (permalink / raw)
  To: Stephen Hurd; +Cc: dev, ajit.khaparde

On Wed, Jun 15, 2016 at 02:23:00PM -0700, Stephen Hurd wrote:
> The bnxt poll mode library (librte_pmd_bnxt) implements support for
> Broadcom NetXtreme C-Series.  These adapters support Standards-
> compliant 10/25/50Gbps 30MPPS full-duplex throughput.
> 
> Information about this family of adapters can be found in the
> NetXtreme Brand section https://goo.gl/4H7q63 of the Broadcom web
> site http://www.broadcom.com/
> 
> With the current driver, allocated mbufs must be large enough to hold
> the entire received frame.  If the mbufs are not large enough, the
> packets will be dropped.  This is most limiting when jumbo frames are
> used.
> 

Applied to dpdk-next-net/rel_16_07

On apply I got conflicts with the nic overview document, so please check the
resulting information in that document is correct in the next-net tree.
I also added a very short entry to the release notes for this new driver as
part of patch 1, since that was missing. Please also check that for correctness
and send on any additional comments/corrections you want on that.

Thanks for all the work on this driver.

Regards,
/Bruce

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library
  2016-06-16 14:24     ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Bruce Richardson
@ 2016-06-16 18:51       ` Ajit Khaparde
  2016-06-21 14:25         ` Ajit Khaparde
  0 siblings, 1 reply; 132+ messages in thread
From: Ajit Khaparde @ 2016-06-16 18:51 UTC (permalink / raw)
  To: Bruce Richardson; +Cc: Stephen Hurd, dev

On Thu, Jun 16, 2016 at 9:24 AM, Bruce Richardson <
bruce.richardson@intel.com> wrote:

> On Wed, Jun 15, 2016 at 02:23:00PM -0700, Stephen Hurd wrote:
> > The bnxt poll mode library (librte_pmd_bnxt) implements support for
> > Broadcom NetXtreme C-Series.  These adapters support Standards-
> > compliant 10/25/50Gbps 30MPPS full-duplex throughput.
> >
> > Information about this family of adapters can be found in the
> > NetXtreme Brand section https://goo.gl/4H7q63 of the Broadcom web
> > site http://www.broadcom.com/
> >
> > With the current driver, allocated mbufs must be large enough to hold
> > the entire received frame.  If the mbufs are not large enough, the
> > packets will be dropped.  This is most limiting when jumbo frames are
> > used.
> >
>
> Applied to dpdk-next-net/rel_16_07
>
> On apply I got conflicts with the nic overview document, so please check
> the
> resulting information in that document is correct in the next-net tree.
> I also added a very short entry to the release notes for this new driver as
> part of patch 1, since that was missing. Please also check that for
> correctness
> and send on any additional comments/corrections you want on that.
>
Thanks Bruce. I had a cursory glance ​and it looked good.
We will update the
​m​ further
if necessary.
​


> Thanks for all the work on this driver.
>
> Regards,
> /Bruce
>

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library
  2016-06-16 18:51       ` Ajit Khaparde
@ 2016-06-21 14:25         ` Ajit Khaparde
  2016-06-21 14:46           ` Bruce Richardson
  0 siblings, 1 reply; 132+ messages in thread
From: Ajit Khaparde @ 2016-06-21 14:25 UTC (permalink / raw)
  To: Bruce Richardson; +Cc: Stephen Hurd, dev

Bruce,
We have a set of patches which add support to a couple of new Broadcom PCI
devices.
Should I use the dpdk-next-net/rel_16_07 to push them out?
Let me know if there is any other way to do this.

Thanks

On Thu, Jun 16, 2016 at 1:51 PM, Ajit Khaparde <ajit.khaparde@broadcom.com>
wrote:

>
> On Thu, Jun 16, 2016 at 9:24 AM, Bruce Richardson <
> bruce.richardson@intel.com> wrote:
>
>> On Wed, Jun 15, 2016 at 02:23:00PM -0700, Stephen Hurd wrote:
>> > The bnxt poll mode library (librte_pmd_bnxt) implements support for
>> > Broadcom NetXtreme C-Series.  These adapters support Standards-
>> > compliant 10/25/50Gbps 30MPPS full-duplex throughput.
>> >
>> > Information about this family of adapters can be found in the
>> > NetXtreme Brand section https://goo.gl/4H7q63 of the Broadcom web
>> > site http://www.broadcom.com/
>> >
>> > With the current driver, allocated mbufs must be large enough to hold
>> > the entire received frame.  If the mbufs are not large enough, the
>> > packets will be dropped.  This is most limiting when jumbo frames are
>> > used.
>> >
>>
>> Applied to dpdk-next-net/rel_16_07
>>
>> On apply I got conflicts with the nic overview document, so please check
>> the
>> resulting information in that document is correct in the next-net tree.
>> I also added a very short entry to the release notes for this new driver
>> as
>> part of patch 1, since that was missing. Please also check that for
>> correctness
>> and send on any additional comments/corrections you want on that.
>>
> Thanks Bruce. I had a cursory glance ​and it looked good.
> We will update the
> ​m​ further
> if necessary.
> ​
>
>
>> Thanks for all the work on this driver.
>>
>> Regards,
>> /Bruce
>>
>
>

^ permalink raw reply	[flat|nested] 132+ messages in thread

* Re: [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library
  2016-06-21 14:25         ` Ajit Khaparde
@ 2016-06-21 14:46           ` Bruce Richardson
  0 siblings, 0 replies; 132+ messages in thread
From: Bruce Richardson @ 2016-06-21 14:46 UTC (permalink / raw)
  To: Ajit Khaparde; +Cc: Stephen Hurd, dev

On Tue, Jun 21, 2016 at 09:25:18AM -0500, Ajit Khaparde wrote:
> Bruce,
> We have a set of patches which add support to a couple of new Broadcom PCI
> devices.
> Should I use the dpdk-next-net/rel_16_07 to push them out?
> Let me know if there is any other way to do this.
> 
> Thanks
> 

They should be based on dpdk-next-net/rel_16_07.

However, be aware that it is very late in the DPDK 16.07 release cycle for us
to take new content in, so please minimise the patchset to only the changes that
are absolutely necessary for you. At this stage we can only take contributions
on a best-effort basis, so having a smaller patchset will increase your chances
that the patch may be merged.

Regards,
/Bruce

> On Thu, Jun 16, 2016 at 1:51 PM, Ajit Khaparde <ajit.khaparde@broadcom.com>
> wrote:
> 
> >
> > On Thu, Jun 16, 2016 at 9:24 AM, Bruce Richardson <
> > bruce.richardson@intel.com> wrote:
> >
> >> On Wed, Jun 15, 2016 at 02:23:00PM -0700, Stephen Hurd wrote:
> >> > The bnxt poll mode library (librte_pmd_bnxt) implements support for
> >> > Broadcom NetXtreme C-Series.  These adapters support Standards-
> >> > compliant 10/25/50Gbps 30MPPS full-duplex throughput.
> >> >
> >> > Information about this family of adapters can be found in the
> >> > NetXtreme Brand section https://goo.gl/4H7q63 of the Broadcom web
> >> > site http://www.broadcom.com/
> >> >
> >> > With the current driver, allocated mbufs must be large enough to hold
> >> > the entire received frame.  If the mbufs are not large enough, the
> >> > packets will be dropped.  This is most limiting when jumbo frames are
> >> > used.
> >> >
> >>
> >> Applied to dpdk-next-net/rel_16_07
> >>
> >> On apply I got conflicts with the nic overview document, so please check
> >> the
> >> resulting information in that document is correct in the next-net tree.
> >> I also added a very short entry to the release notes for this new driver
> >> as
> >> part of patch 1, since that was missing. Please also check that for
> >> correctness
> >> and send on any additional comments/corrections you want on that.
> >>
> > Thanks Bruce. I had a cursory glance ​and it looked good.
> > We will update the
> > ​m​ further
> > if necessary.
> > ​
> >
> >
> >> Thanks for all the work on this driver.
> >>
> >> Regards,
> >> /Bruce
> >>
> >
> >

^ permalink raw reply	[flat|nested] 132+ messages in thread

end of thread, other threads:[~2016-06-21 14:46 UTC | newest]

Thread overview: 132+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-06-06 22:08 [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 02/39] bnxt: add HWRM init code Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 03/39] bnxt: add driver register/unregister support Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 04/39] bnxt: add dev infos get operation Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 05/39] bnxt: add dev configure operation Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 06/39] bnxt: add vnic functions and structs Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 07/39] bnxt: declare ring structs and free() func Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 08/39] bnxt: add completion ring support Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 09/39] bnxt: add L2 filter alloc/init/free Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 10/39] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 11/39] bnxt: add Rx queue create/destroy operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 12/39] bnxt: Add statistics operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 13/39] bnxt: initial Tx code implementation Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 14/39] bnxt: initial Rx " Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 15/39] bnxt: Code to alloc/free ring Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 16/39] bnxt: add HWRM function reset command Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 17/39] bnxt: add HWRM vnic alloc function Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 18/39] bnxt: add HWRM vnic free function Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 19/39] bnxt: add HWRM vnic configure function Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 20/39] bnxt: add API to allow configuration of vnic Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 21/39] bnxt: add HWRM API to configure RSS Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 22/39] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 23/39] bnxt: add HWRM stats context allocation Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 24/39] bnxt: add HWRM ring alloc/free functions Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 25/39] bnxt: add ring group " Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 26/39] bnxt: add HWRM stat context free function Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 27/39] bnxt: Add HWRM API to set and clear filters Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 28/39] bnxt: add ring alloc, free and group init Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 29/39] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 30/39] bnxt: add start/stop/link update operations Stephen Hurd
2016-06-08 10:02   ` Bruce Richardson
2016-06-08 10:41     ` Bruce Richardson
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 31/39] bnxt: add promiscuous enable/disable operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 32/39] bnxt: add all multicast " Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 33/39] bnxt: free memory in close operation Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 34/39] bnxt: add MAC address add/remove dev_ops Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 35/39] bnxt: add set link up/down operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 36/39] bnxt: add reta update/query operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 37/39] bnxt: add RSS device operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 38/39] bnxt: add flow control operations Stephen Hurd
2016-06-06 22:08 ` [dpdk-dev] [PATCH v4 39/39] bnxt: Replace bnxt_ring_struct with bnxt_ring Stephen Hurd
2016-06-07  6:25 ` [dpdk-dev] [PATCH v4 01/39] bnxt: new driver for Broadcom NetXtreme-C devices Thomas Monjalon
2016-06-08 10:31   ` Bruce Richardson
2016-06-08 10:21 ` Bruce Richardson
2016-06-08 10:41   ` Bruce Richardson
2016-06-14 13:14     ` Bruce Richardson
2016-06-14 22:55 ` [dpdk-dev] [PATCH v5 01/38] " Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 02/38] bnxt: add HWRM init code Stephen Hurd
2016-06-15 16:25     ` Ferruh Yigit
2016-06-15 20:28       ` Stephen Hurd
2016-06-16  9:46         ` Bruce Richardson
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 03/38] bnxt: add driver register/unregister support Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 04/38] bnxt: add dev infos get operation Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 05/38] bnxt: add dev configure operation Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 06/38] bnxt: add vnic functions and structs Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 07/38] bnxt: declare generic ring structs and free() func Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 08/38] bnxt: add completion ring support Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 09/38] bnxt: add L2 filter alloc/init/free Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 10/38] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 11/38] bnxt: add Rx queue create/destroy operations Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 12/38] bnxt: add statistics operations Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 13/38] bnxt: add initial Tx code implementation Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 14/38] bnxt: add initial Rx " Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 16/38] bnxt: add HWRM function reset command Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 17/38] bnxt: add HWRM vnic alloc function Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 18/38] bnxt: add HWRM vnic free function Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 19/38] bnxt: add HWRM vnic configure function Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 20/38] bnxt: add API to allow configuration of vnic Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 21/38] bnxt: add HWRM API to configure RSS Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 22/38] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 23/38] bnxt: add HWRM API for stats context allocation Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 24/38] bnxt: add HWRM ring alloc/free functions Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 25/38] bnxt: add ring group " Stephen Hurd
2016-06-14 22:55   ` [dpdk-dev] [PATCH v5 26/38] bnxt: add HWRM stat context free function Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 27/38] bnxt: add HWRM API to set and clear filters Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 28/38] bnxt: allocate and free all HWRM rings and groups Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 29/38] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 30/38] bnxt: add start/stop/link update operations Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 31/38] bnxt: add promiscuous enable/disable operations Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 32/38] bnxt: add all multicast " Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 33/38] bnxt: free memory in close operation Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 34/38] bnxt: add MAC address add/remove dev ops Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 35/38] bnxt: add set link up/down operations Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 36/38] bnxt: add reta update/query operations Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 37/38] bnxt: add RSS device operations Stephen Hurd
2016-06-14 22:56   ` [dpdk-dev] [PATCH v5 38/38] bnxt: add flow control operations Stephen Hurd
2016-06-15 15:28   ` [dpdk-dev] [PATCH v5 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Bruce Richardson
2016-06-15 15:30   ` Bruce Richardson
2016-06-15 21:23   ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 01/38] bnxt: new driver for Broadcom NetXtreme-C devices Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 02/38] bnxt: add HWRM init code Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 03/38] bnxt: add driver register/unregister support Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 04/38] bnxt: add dev infos get operation Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 05/38] bnxt: add dev configure operation Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 06/38] bnxt: add vnic functions and structs Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 07/38] bnxt: declare generic ring structs and free() func Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 08/38] bnxt: add completion ring support Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 09/38] bnxt: add L2 filter alloc/init/free Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 10/38] bnxt: add Tx queue operations (nonfunctional) Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 11/38] bnxt: add Rx queue create/destroy operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 12/38] bnxt: add statistics operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 13/38] bnxt: add initial Tx code implementation Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 14/38] bnxt: add initial Rx " Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 15/38] bnxt: add code to alloc/free Tx Rx and cmpl rings Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 16/38] bnxt: add HWRM function reset command Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 17/38] bnxt: add HWRM vnic alloc function Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 18/38] bnxt: add HWRM vnic free function Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 19/38] bnxt: add HWRM vnic configure function Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 20/38] bnxt: add API to allow configuration of vnic Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 21/38] bnxt: add HWRM API to configure RSS Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 22/38] bnxt: add API for L2 Rx mask set/clear functions Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 23/38] bnxt: add HWRM API for stats context allocation Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 24/38] bnxt: add HWRM ring alloc/free functions Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 25/38] bnxt: add ring group " Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 26/38] bnxt: add HWRM stat context free function Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 27/38] bnxt: add HWRM API to set and clear filters Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 28/38] bnxt: allocate and free all HWRM rings and groups Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 29/38] bnxt: add HWRM port PHY config call and helpers Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 30/38] bnxt: add start/stop/link update operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 31/38] bnxt: add promiscuous enable/disable operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 32/38] bnxt: add all multicast " Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 33/38] bnxt: free memory in close operation Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 34/38] bnxt: add MAC address add/remove dev ops Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 35/38] bnxt: add set link up/down operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 36/38] bnxt: add reta update/query operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 37/38] bnxt: add RSS device operations Stephen Hurd
2016-06-15 21:23     ` [dpdk-dev] [PATCH v6 38/38] bnxt: add flow control operations Stephen Hurd
2016-06-16 14:24     ` [dpdk-dev] [PATCH v6 00/38] new bnxt poll mode driver library Bruce Richardson
2016-06-16 18:51       ` Ajit Khaparde
2016-06-21 14:25         ` Ajit Khaparde
2016-06-21 14:46           ` Bruce Richardson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).