DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS
@ 2018-03-21  7:51 Rosen Xu
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 1/5] Add Intel FPGA BUS Command Parse Code Rosen Xu
                   ` (5 more replies)
  0 siblings, 6 replies; 19+ messages in thread
From: Rosen Xu @ 2018-03-21  7:51 UTC (permalink / raw)
  To: dev
  Cc: declan.doherty, bruce.richardson, shreyansh.jain, tianfei.zhang,
	hao.wu, gaetan.rivet

Intel FPGA BUS in DPDK
-------------------------

This patch set introduces Intel FPGA BUS support in DPDK.

Motivation
==========

FPGA is used more and more widely in Cloud and NFV, one primary reason is
that FPGA not only provide ASIC performance but also it's more flexible 
than ASIC. FPGA use Partial Reconfigure(PR) Parts of Bitstream to achieve 
its flexibility. Another reason is that one FPGA can be shared 
by different Users, and each User can use some of AFUs of One FPGA.

That means One FPGA Device Bitstream is divided into many Parts of 
Bitstream(each Part of Bitstream is defined as AFU-Accelerated 
Function Unit), and each AFU is a Hardware Acceleration Unit and 
it can dynamically Reload respectively.

Scope
=====

The Intel FPGA BUS implementation is target towards various FPGA Devices 
use PR to provide many Acceleration Function. Specific PMDs may also 
bind to its AFU. And Applications don't care they are using ASIC 
Acceleration or FPGA AFU Acceleration.

Proposed Solution
=================
 - Involve Rawdev to take FPGA Partial Configuration(Download/PR)
 - Defined FPGA-BUS for Acceleration Drivers of AFUs
   - FPGA PCI Scan(1st Scan) follows DPDK UIO/VFIO PCI Scan Process, 
     probe Intel FPGA Rawdev Driver
   - AFU Scan(2nd Scan) bind DPDK Driver to FPGA Partial-Bitstream

Status
=====
With integrating Intel PSG FPGA Software Stack OPAE(Open Programmable 
Acceleration Engine) Share Code, Intel FPGA BUS runs well in 
Intel PSG FPGA Cards.

Patch set Information
====================

This patch set includes 6 patches:
* 0 : Introduce the Intel FPGA BUS library and enable its compilation.
* 1 : Adds command parse code, for start-up application 
      with Intel FPGA BUS.
* 2 : Adds Driver Probe Code, for AFU Drivers should probed 
      after PCI Drivers.
* 3 : Adds Intel FPGA BUS library code, for AFU Device scan
      and AFU Drivers probe.
* 4 : Adds a Intel FPGA rawdevice driver, for FPGA Device Management 
      such as PR.
* 5 : Adds Intel OPAE(Open Programmable Acceleration Engine) Share Code, 
      it's Intel FPGA Software Stack.

Rosen Xu (5):
  Add Intel FPGA BUS Command Parse Code
  Add Intel FPGA BUS Probe Code
  Add Intel FPGA BUS Lib Code
  Add Intel FPGA BUS Rawdev Code
  Add Intel OPAE Share Code

 drivers/bus/ifpga/Makefile                         |   64 +
 drivers/bus/ifpga/ifpga_bus.c                      |  573 +++++++
 drivers/bus/ifpga/ifpga_common.c                   |  154 ++
 drivers/bus/ifpga/ifpga_common.h                   |   25 +
 drivers/bus/ifpga/ifpga_logs.h                     |   32 +
 drivers/bus/ifpga/rte_bus_ifpga.h                  |  141 ++
 drivers/bus/ifpga/rte_bus_ifpga_version.map        |    8 +
 drivers/raw/Makefile                               |    1 +
 drivers/raw/ifpga_rawdev/Makefile                  |   63 +
 drivers/raw/ifpga_rawdev/base/Makefile             |   54 +
 drivers/raw/ifpga_rawdev/base/ifpga_api.c          |  543 +++++++
 drivers/raw/ifpga_rawdev/base/ifpga_api.h          |   77 +
 drivers/raw/ifpga_rawdev/base/ifpga_compat.h       |   84 +
 drivers/raw/ifpga_rawdev/base/ifpga_defines.h      | 1696 ++++++++++++++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c    |  861 ++++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h    |   38 +
 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c  |  340 ++++
 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h  |  197 +++
 drivers/raw/ifpga_rawdev/base/ifpga_fme.c          |  763 +++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c    |  328 ++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c    |  430 +++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c    |  742 +++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c       |  395 +++++
 drivers/raw/ifpga_rawdev/base/ifpga_hw.h           |  152 ++
 drivers/raw/ifpga_rawdev/base/ifpga_port.c         |  730 +++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_port_error.c   |  236 +++
 drivers/raw/ifpga_rawdev/base/opae_debug.c         |  126 ++
 drivers/raw/ifpga_rawdev/base/opae_debug.h         |   46 +
 drivers/raw/ifpga_rawdev/base/opae_hw_api.c        |  389 +++++
 drivers/raw/ifpga_rawdev/base/opae_hw_api.h        |  276 ++++
 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c  |  151 ++
 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h  |  293 ++++
 drivers/raw/ifpga_rawdev/base/opae_osdep.h         |  111 ++
 .../ifpga_rawdev/base/osdep_raw/osdep_generic.h    |  104 ++
 .../ifpga_rawdev/base/osdep_rte/osdep_generic.h    |   72 +
 drivers/raw/ifpga_rawdev/ifpga_rawdev.c            |  486 ++++++
 drivers/raw/ifpga_rawdev/ifpga_rawdev.h            |   38 +
 drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c    |   99 ++
 .../ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map  |    4 +
 lib/librte_eal/common/eal_common_bus.c             |   14 +-
 lib/librte_eal/common/eal_common_options.c         |    8 +-
 lib/librte_eal/common/eal_options.h                |    2 +
 42 files changed, 10944 insertions(+), 2 deletions(-)
 create mode 100644 drivers/bus/ifpga/Makefile
 create mode 100644 drivers/bus/ifpga/ifpga_bus.c
 create mode 100644 drivers/bus/ifpga/ifpga_common.c
 create mode 100644 drivers/bus/ifpga/ifpga_common.h
 create mode 100644 drivers/bus/ifpga/ifpga_logs.h
 create mode 100644 drivers/bus/ifpga/rte_bus_ifpga.h
 create mode 100644 drivers/bus/ifpga/rte_bus_ifpga_version.map
 create mode 100644 drivers/raw/ifpga_rawdev/Makefile
 create mode 100644 drivers/raw/ifpga_rawdev/base/Makefile
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_api.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_api.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_compat.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_defines.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_hw.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_port.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_port_error.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_debug.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_debug.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_hw_api.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_hw_api.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_osdep.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/osdep_raw/osdep_generic.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/osdep_rte/osdep_generic.h
 create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev.c
 create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev.h
 create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c
 create mode 100644 drivers/raw/ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map

-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [dpdk-dev] [PATCH V2 1/5] Add Intel FPGA BUS Command Parse Code
  2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
@ 2018-03-21  7:51 ` Rosen Xu
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code Rosen Xu
                   ` (4 subsequent siblings)
  5 siblings, 0 replies; 19+ messages in thread
From: Rosen Xu @ 2018-03-21  7:51 UTC (permalink / raw)
  To: dev
  Cc: declan.doherty, bruce.richardson, shreyansh.jain, tianfei.zhang,
	hao.wu, gaetan.rivet

Signed-off-by: Rosen Xu <rosen.xu@intel.com>
---
 lib/librte_eal/common/eal_common_options.c | 8 +++++++-
 lib/librte_eal/common/eal_options.h        | 2 ++
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/lib/librte_eal/common/eal_common_options.c b/lib/librte_eal/common/eal_common_options.c
index 9f2f8d2..4fe0875 100644
--- a/lib/librte_eal/common/eal_common_options.c
+++ b/lib/librte_eal/common/eal_common_options.c
@@ -73,6 +73,7 @@
 	{OPT_VDEV,              1, NULL, OPT_VDEV_NUM             },
 	{OPT_VFIO_INTR,         1, NULL, OPT_VFIO_INTR_NUM        },
 	{OPT_VMWARE_TSC_MAP,    0, NULL, OPT_VMWARE_TSC_MAP_NUM   },
+	{OPT_IFPGA,             1, NULL, OPT_IFPGA_NUM            },
 	{0,                     0, NULL, 0                        }
 };
 
@@ -1160,7 +1161,12 @@ static int xdigit2val(unsigned char c)
 
 		core_parsed = LCORE_OPT_MAP;
 		break;
-
+	case OPT_IFPGA_NUM:
+		if (eal_option_device_add(RTE_DEVTYPE_VIRTUAL,
+				optarg) < 0) {
+			return -1;
+		}
+		break;
 	/* don't know what to do, leave this to caller */
 	default:
 		return 1;
diff --git a/lib/librte_eal/common/eal_options.h b/lib/librte_eal/common/eal_options.h
index e86c711..bdbb2c4 100644
--- a/lib/librte_eal/common/eal_options.h
+++ b/lib/librte_eal/common/eal_options.h
@@ -55,6 +55,8 @@ enum {
 	OPT_VFIO_INTR_NUM,
 #define OPT_VMWARE_TSC_MAP    "vmware-tsc-map"
 	OPT_VMWARE_TSC_MAP_NUM,
+#define OPT_IFPGA              "ifpga"
+	OPT_IFPGA_NUM,
 	OPT_LONG_MAX_NUM
 };
 
-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code
  2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 1/5] Add Intel FPGA BUS Command Parse Code Rosen Xu
@ 2018-03-21  7:51 ` Rosen Xu
  2018-03-21  9:07   ` Shreyansh Jain
  2018-03-21 10:05   ` Gaëtan Rivet
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code Rosen Xu
                   ` (3 subsequent siblings)
  5 siblings, 2 replies; 19+ messages in thread
From: Rosen Xu @ 2018-03-21  7:51 UTC (permalink / raw)
  To: dev
  Cc: declan.doherty, bruce.richardson, shreyansh.jain, tianfei.zhang,
	hao.wu, gaetan.rivet

Signed-off-by: Rosen Xu <rosen.xu@intel.com>
---
 lib/librte_eal/common/eal_common_bus.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/lib/librte_eal/common/eal_common_bus.c b/lib/librte_eal/common/eal_common_bus.c
index 3e022d5..e3bcebe 100644
--- a/lib/librte_eal/common/eal_common_bus.c
+++ b/lib/librte_eal/common/eal_common_bus.c
@@ -87,7 +87,7 @@ struct rte_bus_list rte_bus_list =
 rte_bus_probe(void)
 {
 	int ret;
-	struct rte_bus *bus, *vbus = NULL;
+	struct rte_bus *bus, *vbus = NULL, *ifpga_bus = NULL;
 
 	TAILQ_FOREACH(bus, &rte_bus_list, next) {
 		if (!strcmp(bus->name, "vdev")) {
@@ -95,6 +95,11 @@ struct rte_bus_list rte_bus_list =
 			continue;
 		}
 
+		if (!strcmp(bus->name, "ifpga")) {
+			ifpga_bus = bus;
+			continue;
+		}
+
 		ret = bus->probe();
 		if (ret)
 			RTE_LOG(ERR, EAL, "Bus (%s) probe failed.\n",
@@ -108,6 +113,13 @@ struct rte_bus_list rte_bus_list =
 				vbus->name);
 	}
 
+	if (ifpga_bus) {
+		ret = ifpga_bus->probe();
+		if (ret)
+			RTE_LOG(ERR, EAL, "Scan for (%s) bus failed.\n",
+				ifpga_bus->name);
+	}
+
 	return 0;
 }
 
-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 1/5] Add Intel FPGA BUS Command Parse Code Rosen Xu
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code Rosen Xu
@ 2018-03-21  7:51 ` Rosen Xu
  2018-03-21  9:28   ` Shreyansh Jain
  2018-03-21 10:20   ` Gaëtan Rivet
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 4/5] Add Intel FPGA BUS Rawdev Code Rosen Xu
                   ` (2 subsequent siblings)
  5 siblings, 2 replies; 19+ messages in thread
From: Rosen Xu @ 2018-03-21  7:51 UTC (permalink / raw)
  To: dev
  Cc: declan.doherty, bruce.richardson, shreyansh.jain, tianfei.zhang,
	hao.wu, gaetan.rivet

Signed-off-by: Rosen Xu <rosen.xu@intel.com>
---
 drivers/bus/ifpga/Makefile                  |  64 ++++
 drivers/bus/ifpga/ifpga_bus.c               | 573 ++++++++++++++++++++++++++++
 drivers/bus/ifpga/ifpga_common.c            | 154 ++++++++
 drivers/bus/ifpga/ifpga_common.h            |  25 ++
 drivers/bus/ifpga/ifpga_logs.h              |  32 ++
 drivers/bus/ifpga/rte_bus_ifpga.h           | 141 +++++++
 drivers/bus/ifpga/rte_bus_ifpga_version.map |   8 +
 7 files changed, 997 insertions(+)
 create mode 100644 drivers/bus/ifpga/Makefile
 create mode 100644 drivers/bus/ifpga/ifpga_bus.c
 create mode 100644 drivers/bus/ifpga/ifpga_common.c
 create mode 100644 drivers/bus/ifpga/ifpga_common.h
 create mode 100644 drivers/bus/ifpga/ifpga_logs.h
 create mode 100644 drivers/bus/ifpga/rte_bus_ifpga.h
 create mode 100644 drivers/bus/ifpga/rte_bus_ifpga_version.map

diff --git a/drivers/bus/ifpga/Makefile b/drivers/bus/ifpga/Makefile
new file mode 100644
index 0000000..c71f186
--- /dev/null
+++ b/drivers/bus/ifpga/Makefile
@@ -0,0 +1,64 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
+#   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 $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_bus_ifpga.a
+LIBABIVER := 1
+EXPORT_MAP := rte_bus_ifpga_version.map
+
+ifeq ($(CONFIG_RTE_LIBRTE_DPAA2_DEBUG_INIT),y)
+CFLAGS += -O0 -g
+CFLAGS += "-Wno-error"
+else
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+endif
+
+CFLAGS += -I$(RTE_SDK)/drivers/bus/ifpga
+CFLAGS += -I$(RTE_SDK)/drivers/bus/pci
+CFLAGS += -I$(RTE_SDK)/lib/librte_eal/linuxapp/eal
+CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
+#CFLAGS += -I$(RTE_SDK)/lib/librte_rawdev
+#LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring -lrte_rawdev
+LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
+#LDLIBS += -lrte_ethdev
+
+VPATH += $(SRCDIR)/base
+
+SRCS-y += \
+        ifpga_bus.c \
+        ifpga_common.c
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/bus/ifpga/ifpga_bus.c b/drivers/bus/ifpga/ifpga_bus.c
new file mode 100644
index 0000000..ff72b74
--- /dev/null
+++ b/drivers/bus/ifpga/ifpga_bus.c
@@ -0,0 +1,573 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#include <string.h>
+#include <inttypes.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <sys/queue.h>
+#include <sys/mman.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include <rte_errno.h>
+#include <rte_bus.h>
+#include <rte_per_lcore.h>
+#include <rte_memory.h>
+#include <rte_memzone.h>
+#include <rte_eal.h>
+#include <rte_common.h>
+
+#include <rte_devargs.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
+#include <rte_kvargs.h>
+#include <rte_alarm.h>
+
+#include "rte_rawdev.h"
+#include "rte_rawdev_pmd.h"
+#include "rte_bus_ifpga.h"
+#include "ifpga_logs.h"
+#include "ifpga_common.h"
+
+int ifpga_bus_logtype;
+
+/*register a ifpga bus based driver */
+void rte_ifpga_driver_register(struct rte_afu_driver *driver)
+{
+	RTE_VERIFY(driver);
+
+	TAILQ_INSERT_TAIL(&rte_ifpga_bus.driver_list, driver, next);
+}
+
+/*un-register a fpga bus based driver */
+void rte_ifpga_driver_unregister(struct rte_afu_driver *driver)
+{
+	TAILQ_REMOVE(&rte_ifpga_bus.driver_list, driver, next);
+}
+
+static struct rte_ifpga_device *
+ifpga_find_ifpga_dev(const struct rte_pci_addr *pci_addr)
+{
+	struct rte_ifpga_device *ifpga_dev = NULL;
+
+	TAILQ_FOREACH(ifpga_dev, &rte_ifpga_bus.ifpga_list, next) {
+		if (!ifpga_pci_addr_cmp_1(&ifpga_dev->pci_addr, pci_addr))
+			return ifpga_dev;
+	}
+	return NULL;
+}
+
+static struct rte_afu_device *
+ifpga_find_afu_dev(const struct rte_afu_id *afu_id)
+{
+	struct rte_afu_device *afu_dev = NULL;
+
+	TAILQ_FOREACH(afu_dev, &rte_ifpga_bus.afu_list, next) {
+		if (!ifpga_afu_id_cmp(&afu_dev->id, afu_id))
+			return afu_dev;
+	}
+	return NULL;
+}
+
+static const char *valid_args[] = {
+#define IFPGA_ARG_BDF          "bdf"
+	IFPGA_ARG_BDF,
+#define IFPGA_ARG_PORT         "port"
+	IFPGA_ARG_PORT,
+#define IFPGA_ARG_PATH         "path"
+	IFPGA_ARG_PATH,
+#define IFPGA_ARG_UUID_HIGH    "uuid_high"
+	IFPGA_ARG_UUID_HIGH,
+#define IFPGA_ARG_UUID_LOW     "uuid_low"
+	IFPGA_ARG_UUID_LOW,
+#define IFPGA_ARG_PR_ENABLE     "pr_enable"
+	IFPGA_ARG_PR_ENABLE,
+#define IFPGA_ARG_DEBUG         "debug"
+	IFPGA_ARG_DEBUG,
+	NULL
+};
+
+/*
+ * Scan the content of the FPGA bus, and the devices in the devices
+ * list
+ */
+static struct rte_afu_device *
+rte_ifpga_scan_one(struct rte_devargs *devargs,
+	struct rte_rawdev *rawdev)
+{
+	struct rte_kvargs *kvlist = NULL;
+	struct rte_bus *pci_bus = NULL;
+	struct rte_device *dev = NULL;
+	struct rte_afu_device *afu_dev = NULL;
+	struct rte_afu_pr_conf afu_pr_conf;
+	int ret = 0;
+	char *path = NULL;
+	int pr_enable = 1;
+	int debug = 0;
+
+	memset((char *)(&afu_pr_conf), 0, sizeof(struct rte_afu_pr_conf));
+
+	kvlist = rte_kvargs_parse(devargs->args, valid_args);
+	if (!kvlist) {
+		IFPGA_BUS_ERR("error when parsing param");
+		goto end;
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_BDF) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_BDF,
+			&ifpga_get_bdf_arg, &afu_pr_conf.afu_id.pci_addr) < 0) {
+			IFPGA_BUS_ERR("error to parse %s", IFPGA_ARG_BDF);
+			goto end;
+		}
+	} else {
+		IFPGA_BUS_ERR("arg %s is mandatory for ifpga bus",
+			IFPGA_ARG_PATH);
+		goto end;
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_PORT) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_PORT,
+		&ifpga_get_integer32_arg, &afu_pr_conf.afu_id.port) < 0) {
+			IFPGA_BUS_ERR("error to parse %s",
+				     IFPGA_ARG_PORT);
+			goto end;
+		}
+	} else {
+		IFPGA_BUS_ERR("arg %s is mandatory for ifpga bus",
+			  IFPGA_ARG_PATH);
+		goto end;
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_PATH) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_PATH,
+				       &ifpga_get_string_arg, &path) < 0) {
+			IFPGA_BUS_ERR("error to parse %s",
+				     IFPGA_ARG_PATH);
+			goto end;
+		}
+	} else {
+		IFPGA_BUS_ERR("arg %s is mandatory for ifpga bus",
+			  IFPGA_ARG_PATH);
+		goto end;
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_UUID_HIGH) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_UUID_HIGH,
+		&ifpga_get_integer64_arg, &afu_pr_conf.afu_id.uuid_high) < 0) {
+			IFPGA_BUS_ERR("error to parse %s",
+				     IFPGA_ARG_UUID_HIGH);
+			goto end;
+		}
+	} else {
+		IFPGA_BUS_ERR("arg %s is mandatory for ifpga bus",
+			  IFPGA_ARG_PATH);
+		goto end;
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_UUID_LOW) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_UUID_LOW,
+		&ifpga_get_integer64_arg, &afu_pr_conf.afu_id.uuid_low) < 0) {
+			IFPGA_BUS_ERR("error to parse %s",
+				     IFPGA_ARG_UUID_LOW);
+			goto end;
+		}
+	} else {
+		IFPGA_BUS_ERR("arg %s is mandatory for ifpga bus",
+			  IFPGA_ARG_PATH);
+		goto end;
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_PR_ENABLE) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_PR_ENABLE,
+			&ifpga_get_integer32_arg, &pr_enable) < 0) {
+			IFPGA_BUS_ERR("error to parse %s",
+				     IFPGA_ARG_UUID_HIGH);
+			goto end;
+		}
+	}
+
+	if (rte_kvargs_count(kvlist, IFPGA_ARG_DEBUG) == 1) {
+		if (rte_kvargs_process(kvlist, IFPGA_ARG_DEBUG,
+				       &ifpga_get_integer32_arg, &debug) < 0) {
+			IFPGA_BUS_ERR("error to parse %s",
+				     IFPGA_ARG_UUID_HIGH);
+			goto end;
+		}
+	}
+
+	if (!debug) {
+		pci_bus = rte_bus_find_by_name("pci");
+		if (pci_bus == NULL) {
+			IFPGA_BUS_ERR("unable to find PCI bus\n");
+			goto end;
+		}
+
+		dev = pci_bus->find_device(NULL,
+				ifpga_pci_addr_cmp,
+				&afu_pr_conf.afu_id.pci_addr);
+		if (dev == NULL) {
+			IFPGA_BUS_ERR("unable to find PCI device\n");
+			goto end;
+		}
+	} else {
+		IFPGA_BUS_DEBUG("pci_addr domain   : %x\n",
+				afu_pr_conf.afu_id.pci_addr.domain);
+		IFPGA_BUS_DEBUG("pci_addr bus      : %x\n",
+				afu_pr_conf.afu_id.pci_addr.bus);
+		IFPGA_BUS_DEBUG("pci_addr devid    : %x\n",
+				afu_pr_conf.afu_id.pci_addr.devid);
+		IFPGA_BUS_DEBUG("pci_addr function : %x\n",
+				afu_pr_conf.afu_id.pci_addr.function);
+		IFPGA_BUS_DEBUG("uuid_low          : %lx\n",
+				afu_pr_conf.afu_id.uuid_low);
+		IFPGA_BUS_DEBUG("uuid_high         : %lx\n",
+				afu_pr_conf.afu_id.uuid_high);
+		IFPGA_BUS_DEBUG("afu port          : %x\n",
+				afu_pr_conf.afu_id.port);
+	}
+
+	if (ifpga_find_afu_dev(&afu_pr_conf.afu_id))
+		goto end;
+
+
+	afu_dev = calloc(1, sizeof(*afu_dev));
+	if (!afu_dev)
+		goto end;
+
+	afu_dev->device.devargs = devargs;
+	afu_dev->device.numa_node = SOCKET_ID_ANY;
+	afu_dev->device.name = devargs->name;
+	afu_dev->rawdev = rawdev;
+	afu_dev->id.pci_addr.domain = afu_pr_conf.afu_id.pci_addr.domain;
+	afu_dev->id.pci_addr.bus = afu_pr_conf.afu_id.pci_addr.bus;
+	afu_dev->id.pci_addr.devid = afu_pr_conf.afu_id.pci_addr.devid;
+	afu_dev->id.pci_addr.function = afu_pr_conf.afu_id.pci_addr.function;
+	afu_dev->id.uuid_low  = afu_pr_conf.afu_id.uuid_low;
+	afu_dev->id.uuid_high = afu_pr_conf.afu_id.uuid_high;
+	afu_dev->id.port      = afu_pr_conf.afu_id.port;
+
+	if (rawdev->dev_ops && rawdev->dev_ops->dev_info_get)
+		rawdev->dev_ops->dev_info_get(rawdev, afu_dev);
+
+	if (rawdev->dev_ops &&
+		rawdev->dev_ops->dev_start &&
+		rawdev->dev_ops->dev_start(rawdev))
+		goto free_dev;
+	if (pr_enable) {
+		strncpy(afu_pr_conf.bs_path, path, strlen(path));
+		if (rawdev->dev_ops->firmware_load &&
+			rawdev->dev_ops->firmware_load(rawdev,
+					&afu_pr_conf)){
+			printf("firmware load error %d\n", ret);
+			goto free_dev;
+		}
+	}
+
+	return afu_dev;
+
+free_dev:
+	free(afu_dev);
+end:
+	if (kvlist)
+		rte_kvargs_free(kvlist);
+	if (path)
+		free(path);
+
+	return NULL;
+}
+
+/*
+ * Scan the content of the FPGA bus, and the devices in the devices
+ * list
+ */
+static int
+rte_ifpga_scan(void)
+{
+	struct rte_ifpga_device *ifpga_dev;
+	struct rte_devargs *devargs;
+	struct rte_kvargs *kvlist = NULL;
+	struct rte_pci_addr pci_addr;
+
+	/* for FPGA devices we scan the devargs_list populated via cmdline */
+	TAILQ_FOREACH(devargs, &devargs_list, next) {
+		if (devargs->bus != &rte_ifpga_bus.bus)
+			continue;
+
+		kvlist = rte_kvargs_parse(devargs->args, valid_args);
+		if (!kvlist) {
+			IFPGA_BUS_ERR("error when parsing param");
+			goto end;
+		}
+
+		if (rte_kvargs_count(kvlist, IFPGA_ARG_BDF) == 1) {
+			if (rte_kvargs_process(kvlist, IFPGA_ARG_BDF,
+				&ifpga_get_bdf_arg, &pci_addr) < 0) {
+				IFPGA_BUS_ERR("error to parse %s",
+					IFPGA_ARG_BDF);
+				goto end;
+		}
+	} else {
+		IFPGA_BUS_ERR("arg %s is mandatory for ifpga bus",
+			IFPGA_ARG_PATH);
+		goto end;
+	}
+
+	if (ifpga_find_ifpga_dev(&pci_addr))
+		goto end;
+
+	ifpga_dev = calloc(1, sizeof(*ifpga_dev));
+	if (!ifpga_dev)
+		goto end;
+
+		ifpga_dev->pci_addr.bus      = pci_addr.bus;
+		ifpga_dev->pci_addr.devid    = pci_addr.devid;
+		ifpga_dev->pci_addr.function = pci_addr.function;
+
+		TAILQ_INSERT_TAIL(&rte_ifpga_bus.ifpga_list, ifpga_dev, next);
+	}
+
+	return 0;
+
+end:
+	if (kvlist)
+		rte_kvargs_free(kvlist);
+
+	return 0;
+}
+
+/*
+ * Scan the content of the PCI bus, and call the probe() function for
+ * all registered drivers that have a matching entry in its id_table
+ * for discovered devices.
+ */
+int
+rte_ifpga_scan_2nd(void *conf, void *id)
+{
+	struct rte_ifpga_device *ifpga_dev;
+	struct rte_rawdev *rdev;
+	struct rte_pci_addr *pci_addr;
+	struct rte_devargs *devargs;
+	struct rte_afu_device *afu_dev = NULL;
+
+	rdev     = conf;
+	pci_addr = id;
+
+	ifpga_dev = ifpga_find_ifpga_dev(pci_addr);
+	if (!ifpga_dev)
+		return 0;
+
+	ifpga_dev->rdev = rdev;
+
+    /* for FPGA devices we scan the devargs_list populated via cmdline */
+	TAILQ_FOREACH(devargs, &devargs_list, next) {
+	    afu_dev = rte_ifpga_scan_one(devargs, rdev);
+		if (afu_dev != NULL)
+		    TAILQ_INSERT_TAIL(&rte_ifpga_bus.afu_list, afu_dev, next);
+		else
+			return 0;
+	}
+
+	return 0;
+}
+
+static int
+ifpga_probe_one_driver(struct rte_afu_driver *drv,
+			struct rte_afu_device *afu_dev)
+{
+	int ret;
+
+	if ((drv->id.pci_addr.bus == afu_dev->id.pci_addr.bus) &&
+		(drv->id.pci_addr.devid == afu_dev->id.pci_addr.devid) &&
+		(drv->id.pci_addr.function == afu_dev->id.pci_addr.function) &&
+		(drv->id.uuid_low == afu_dev->id.uuid_low) &&
+		(drv->id.uuid_high == afu_dev->id.uuid_high) &&
+		(drv->id.port == afu_dev->id.port)) {
+
+		afu_dev->driver = drv;
+
+		/* call the driver probe() function */
+		ret = drv->probe(afu_dev);
+		if (ret)
+			afu_dev->driver = NULL;
+		return ret;
+	}
+
+	/* return positive value if driver doesn't support this device */
+	return 1;
+}
+
+static int
+ifpga_probe_all_drivers(struct rte_afu_device *afu_dev)
+{
+	const char *name;
+	struct rte_afu_driver *drv = NULL;
+	int rc;
+
+	if (afu_dev == NULL)
+		return -1;
+
+	/* Check if a driver is already loaded */
+	if (afu_dev->driver != NULL)
+		return 0;
+
+	name = rte_ifpga_device_name(afu_dev);
+	IFPGA_BUS_DEBUG("Search driver %s to probe device %s\n", name,
+		rte_ifpga_device_name(afu_dev));
+
+	TAILQ_FOREACH(drv, &rte_ifpga_bus.driver_list, next) {
+		rc = ifpga_probe_one_driver(drv, afu_dev);
+		if (rc < 0)
+			/* negative value is an error */
+			return -1;
+		if (rc > 0)
+			/* positive value means driver doesn't support it */
+			continue;
+		return 0;
+	}
+	return 1;
+}
+
+/*
+ * Scan the content of the PCI bus, and call the probe() function for
+ * all registered drivers that have a matching entry in its id_table
+ * for discovered devices.
+ */
+static int
+rte_ifpga_probe(void)
+{
+	struct rte_afu_device *afu_dev = NULL;
+	int ret = 0;
+
+	TAILQ_FOREACH(afu_dev, &rte_ifpga_bus.afu_list, next) {
+
+		if (afu_dev->device.driver)
+			continue;
+
+		ret = ifpga_probe_all_drivers(afu_dev);
+		if (ret < 0)
+		    IFPGA_BUS_ERR("failed to initialize %s device\n",
+				rte_ifpga_device_name(afu_dev));
+	}
+
+	return 0;
+}
+
+static int
+rte_ifpga_plug(struct rte_device *dev)
+{
+	return ifpga_probe_all_drivers(RTE_DEV_TO_AFU(dev));
+}
+
+static int ifpga_remove_driver(struct rte_afu_device *afu_dev)
+{
+	const char *name;
+	const struct rte_afu_driver *driver;
+
+	name = rte_ifpga_device_name(afu_dev);
+	if (!afu_dev->device.driver) {
+		IFPGA_BUS_DEBUG("no driver attach to device %s\n", name);
+		return 1;
+	}
+
+	driver = container_of(afu_dev->device.driver,
+		const struct rte_afu_driver,
+		driver);
+	return driver->remove(afu_dev);
+}
+
+static int
+rte_ifpga_unplug(struct rte_device *dev)
+{
+	struct rte_afu_device *afu_dev;
+	struct rte_devargs *devargs;
+	int ret;
+
+	if (dev == NULL)
+		return -EINVAL;
+
+	afu_dev = RTE_DEV_TO_AFU(dev);
+	if (!dev)
+		return -ENOENT;
+
+	devargs = dev->devargs;
+
+	ret = ifpga_remove_driver(afu_dev);
+	if (ret)
+		return ret;
+
+	TAILQ_REMOVE(&rte_ifpga_bus.afu_list, afu_dev, next);
+
+	TAILQ_REMOVE(&devargs_list, devargs, next);
+
+	free(devargs->args);
+	free(devargs);
+	free(afu_dev);
+	return 0;
+
+}
+
+static struct rte_device *
+rte_ifpga_find_device(const struct rte_device *start,
+	rte_dev_cmp_t cmp, const void *data)
+{
+	struct rte_afu_device *afu_dev;
+
+	TAILQ_FOREACH(afu_dev, &rte_ifpga_bus.afu_list, next) {
+		if (start && &afu_dev->device == start) {
+			start = NULL;
+			continue;
+		}
+		if (cmp(&afu_dev->device, data) == 0)
+			return &afu_dev->device;
+	}
+	return NULL;
+}
+static int
+rte_ifpga_parse(const char *name, void *addr)
+{
+	struct rte_afu_driver **out = addr;
+	struct rte_afu_driver *driver = NULL;
+
+	TAILQ_FOREACH(driver, &rte_ifpga_bus.driver_list, next) {
+		if (strncmp(driver->driver.name, name,
+			    strlen(driver->driver.name)) == 0)
+			break;
+		if (driver->driver.alias &&
+		    strncmp(driver->driver.alias, name,
+			    strlen(driver->driver.alias)) == 0)
+			break;
+	}
+	if (driver != NULL &&
+	    addr != NULL)
+		*out = driver;
+	return driver == NULL;
+}
+
+struct rte_ifpga_bus rte_ifpga_bus = {
+	 .bus = {
+		 .scan        = rte_ifpga_scan,
+		 .probe       = rte_ifpga_probe,
+		 .find_device = rte_ifpga_find_device,
+	     .plug        = rte_ifpga_plug,
+	     .unplug      = rte_ifpga_unplug,
+	     .parse       = rte_ifpga_parse,
+	 },
+	.ifpga_list  = TAILQ_HEAD_INITIALIZER(rte_ifpga_bus.ifpga_list),
+	.afu_list    = TAILQ_HEAD_INITIALIZER(rte_ifpga_bus.afu_list),
+	.driver_list = TAILQ_HEAD_INITIALIZER(rte_ifpga_bus.driver_list),
+};
+
+RTE_REGISTER_BUS(IFPGA_BUS_NAME, rte_ifpga_bus.bus);
+
+RTE_INIT(ifpga_init_log)
+{
+	ifpga_bus_logtype = rte_log_register("bus.ifpga");
+	if (ifpga_bus_logtype >= 0)
+		rte_log_set_level(ifpga_bus_logtype, RTE_LOG_NOTICE);
+}
+
diff --git a/drivers/bus/ifpga/ifpga_common.c b/drivers/bus/ifpga/ifpga_common.c
new file mode 100644
index 0000000..8d98854
--- /dev/null
+++ b/drivers/bus/ifpga/ifpga_common.c
@@ -0,0 +1,154 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#include <string.h>
+#include <inttypes.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <sys/queue.h>
+#include <sys/mman.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include <rte_errno.h>
+#include <rte_bus.h>
+#include <rte_per_lcore.h>
+#include <rte_memory.h>
+#include <rte_memzone.h>
+#include <rte_eal.h>
+#include <rte_common.h>
+
+#include <rte_devargs.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
+#include <rte_kvargs.h>
+#include <rte_alarm.h>
+
+#include "rte_bus_ifpga.h"
+#include "ifpga_logs.h"
+#include "ifpga_common.h"
+
+int ifpga_get_string_arg(const char *key __rte_unused,
+	const char *value, void *extra_args)
+{
+	if (!value || !extra_args)
+		return -EINVAL;
+
+	*(char **)extra_args = strdup(value);
+
+	if (!*(char **)extra_args)
+		return -ENOMEM;
+
+	return 0;
+}
+int ifpga_get_integer32_arg(const char *key __rte_unused,
+	const char *value, void *extra_args)
+{
+	if (!value || !extra_args)
+		return -EINVAL;
+
+	*(int *)extra_args = strtoull(value, NULL, 0);
+
+	return 0;
+}
+int ifpga_get_integer64_arg(const char *key __rte_unused,
+	const char *value, void *extra_args)
+{
+	if (!value || !extra_args)
+		return -EINVAL;
+
+	*(uint64_t *)extra_args = strtoull(value, NULL, 0);
+
+	return 0;
+}
+int ifpga_get_unsigned_long(const char *str, int base)
+{
+	unsigned long num;
+	char *end = NULL;
+
+	errno = 0;
+
+	num = strtoul(str, &end, base);
+	if ((str[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
+		return -1;
+
+	return num;
+
+}
+int ifpga_get_bdf_arg(const char *key __rte_unused,
+	const char *value, void *extra_args)
+{
+#define MAX_PATH_LEN 1024
+	struct rte_pci_addr *addr;
+	int num[4];
+	char str[MAX_PATH_LEN];
+	int i, j;
+
+	if (!value || !extra_args)
+		return -EINVAL;
+
+	addr = (struct rte_pci_addr *)extra_args;
+	strcpy(str, value);
+	memset(num, 0, 4 * sizeof(num[0]));
+	i = strlen(str) - 1;
+	j = 3;
+	while (i > 0 && j >= 0) {
+		while ((str[i - 1] != ':' && str[i - 1] != '.') && i > 0)
+			i--;
+		num[j--] = ifpga_get_unsigned_long(&str[i], 16);
+		i--;
+		if (i >= 0)
+			str[i] = '\0';
+	}
+	addr->domain = num[0];
+	addr->bus = num[1];
+	addr->devid = num[2];
+	addr->function = num[3];
+	printf("[%s]: bdf %04d:%02d:%02d.%02d\n",
+			__func__,
+			addr->domain,
+			addr->bus,
+			addr->devid,
+			addr->function);
+
+	return 0;
+}
+
+int ifpga_pci_addr_cmp_1(const struct rte_pci_addr *pci_addr0,
+	const struct rte_pci_addr *pci_addr1)
+{
+	if ((pci_addr0->bus == pci_addr1->bus) &&
+		(pci_addr0->devid == pci_addr1->devid) &&
+		(pci_addr0->function == pci_addr1->function)) {
+		return 0;
+	}
+
+	return 1;
+}
+
+int ifpga_afu_id_cmp(const struct rte_afu_id *afu_id0,
+	const struct rte_afu_id *afu_id1)
+{
+	if ((afu_id0->pci_addr.bus       == afu_id1->pci_addr.bus) &&
+		(afu_id0->pci_addr.devid     == afu_id1->pci_addr.devid) &&
+		(afu_id0->pci_addr.function  == afu_id1->pci_addr.function) &&
+		(afu_id0->uuid_low           == afu_id1->uuid_low) &&
+		(afu_id0->uuid_high          == afu_id1->uuid_high) &&
+		(afu_id0->port               == afu_id1->port)) {
+		return 0;
+	} else
+		return 1;
+}
+int ifpga_pci_addr_cmp(const struct rte_device *dev,
+	const void *_pci_addr)
+{
+	struct rte_pci_device *pdev;
+	const struct rte_pci_addr *paddr = _pci_addr;
+
+	pdev = RTE_DEV_TO_PCI(*(struct rte_device **)(void *)&dev);
+	return rte_eal_compare_pci_addr(&pdev->addr, paddr);
+}
diff --git a/drivers/bus/ifpga/ifpga_common.h b/drivers/bus/ifpga/ifpga_common.h
new file mode 100644
index 0000000..23b68d4
--- /dev/null
+++ b/drivers/bus/ifpga/ifpga_common.h
@@ -0,0 +1,25 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2015 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#ifndef _IFPGA_COMMON_H_
+#define _IFPGA_COMMON_H_
+
+int ifpga_get_string_arg(const char *key __rte_unused,
+	const char *value, void *extra_args);
+int ifpga_get_integer32_arg(const char *key __rte_unused,
+	const char *value, void *extra_args);
+int ifpga_get_integer64_arg(const char *key __rte_unused,
+	const char *value, void *extra_args);
+int ifpga_get_unsigned_long(const char *str, int base);
+int ifpga_get_bdf_arg(const char *key __rte_unused,
+	const char *value, void *extra_args);
+int ifpga_pci_addr_cmp_1(const struct rte_pci_addr *pci_addr0,
+	const struct rte_pci_addr *pci_addr1);
+int ifpga_afu_id_cmp(const struct rte_afu_id *afu_id0,
+	const struct rte_afu_id *afu_id1);
+int ifpga_pci_addr_cmp(const struct rte_device *dev,
+	const void *_pci_addr);
+
+#endif /* _IFPGA_COMMON_H_ */
diff --git a/drivers/bus/ifpga/ifpga_logs.h b/drivers/bus/ifpga/ifpga_logs.h
new file mode 100644
index 0000000..36b9b3f
--- /dev/null
+++ b/drivers/bus/ifpga/ifpga_logs.h
@@ -0,0 +1,32 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2015 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#ifndef _IFPGA_BUS_LOGS_H_
+#define _IFPGA_BUS_LOGS_H_
+
+#include <rte_log.h>
+
+extern int ifpga_bus_logtype;
+
+#define IFPGA_LOG(level, fmt, args...) \
+	rte_log(RTE_LOG_ ## level, ifpga_bus_logtype, "%s(): " fmt "\n", \
+		__func__, ##args)
+
+#define IFPGA_BUS_LOG(level, fmt, args...) \
+	rte_log(RTE_LOG_ ## level, ifpga_bus_logtype, "%s(): " fmt "\n", \
+		__func__, ##args)
+
+#define IFPGA_BUS_FUNC_TRACE() IFPGA_BUS_LOG(DEBUG, ">>")
+
+#define IFPGA_BUS_DEBUG(fmt, args...) \
+	IFPGA_BUS_LOG(DEBUG, fmt, ## args)
+#define IFPGA_BUS_INFO(fmt, args...) \
+	IFPGA_BUS_LOG(INFO, fmt, ## args)
+#define IFPGA_BUS_ERR(fmt, args...) \
+	IFPGA_BUS_LOG(ERR, fmt, ## args)
+#define IFPGA_BUS_WARN(fmt, args...) \
+	IFPGA_BUS_LOG(WARNING, fmt, ## args)
+
+#endif /* _IFPGA_BUS_LOGS_H_ */
diff --git a/drivers/bus/ifpga/rte_bus_ifpga.h b/drivers/bus/ifpga/rte_bus_ifpga.h
new file mode 100644
index 0000000..3ffa30a
--- /dev/null
+++ b/drivers/bus/ifpga/rte_bus_ifpga.h
@@ -0,0 +1,141 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2015 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#ifndef _RTE_BUS_IFPGA_H_
+#define _RTE_BUS_IFPGA_H_
+
+/**
+ * @file
+ *
+ * RTE PCI Bus Interface
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <rte_bus.h>
+#include <rte_pci.h>
+
+/** Name of Intel FPGA Bus */
+#define IFPGA_BUS_NAME ifpga
+
+/* Forward declarations */
+struct rte_ifpga_device;
+struct rte_afu_device;
+struct rte_afu_driver;
+
+/** List of Intel FPGA devices */
+TAILQ_HEAD(rte_ifpga_device_list, rte_ifpga_device);
+/** List of Intel AFU devices */
+TAILQ_HEAD(rte_afu_device_list, rte_afu_device);
+/** List of AFU drivers */
+TAILQ_HEAD(rte_afu_driver_list, rte_afu_driver);
+
+#define IFPGA_BUS_BITSTREAM_PATH_MAX_LEN 256
+
+/**
+ * A structure describing an ID for a AFU driver. Each driver provides a
+ * table of these IDs for each device that it supports.
+ */
+struct rte_afu_id {
+	struct rte_pci_addr pci_addr;
+	uint64_t uuid_low;
+	uint64_t uuid_high;
+	int      port;
+} __attribute__ ((packed));
+
+/**
+ * A structure pr configuration AFU driver.
+ */
+
+struct rte_afu_pr_conf {
+	struct rte_afu_id afu_id;
+	int pr_enable;
+	char		bs_path[IFPGA_BUS_BITSTREAM_PATH_MAX_LEN];
+};
+
+#define AFU_PRI_STR_SIZE (PCI_PRI_STR_SIZE + 8)
+
+/**
+ * A structure describing a fpga device.
+ */
+struct rte_ifpga_device {
+	TAILQ_ENTRY(rte_ifpga_device) next;       /**< Next in device list. */
+	struct rte_pci_addr pci_addr;
+	struct rte_rawdev *rdev;
+};
+
+/**
+ * A structure describing a AFU device.
+ */
+struct rte_afu_device {
+	TAILQ_ENTRY(rte_afu_device) next;       /**< Next in device list. */
+	struct rte_device device;               /**< Inherit core device */
+	struct rte_rawdev *rawdev;
+	struct rte_afu_id id;                   /**< AFU id within FPGA. */
+	uint32_t num_region;   /**< number of regions found */
+	struct rte_mem_resource mem_resource[PCI_MAX_RESOURCE];
+						/**< PCI Memory Resource */
+	struct rte_intr_handle intr_handle;     /**< Interrupt handle */
+	struct rte_afu_driver *driver;          /**< Associated driver */
+	char path[IFPGA_BUS_BITSTREAM_PATH_MAX_LEN];
+} __attribute__ ((packed));
+
+/**
+ * @internal
+ * Helper macro for drivers that need to convert to struct rte_afu_device.
+ */
+#define RTE_DEV_TO_AFU(ptr) \
+	container_of(ptr, struct rte_afu_device, device)
+
+/**
+ * Initialisation function for the driver called during PCI probing.
+ */
+typedef int (afu_probe_t)(struct rte_afu_device *);
+
+/**
+ * Uninitialisation function for the driver called during hotplugging.
+ */
+typedef int (afu_remove_t)(struct rte_afu_device *);
+
+/**
+ * A structure describing a PCI device.
+ */
+struct rte_afu_driver {
+	TAILQ_ENTRY(rte_afu_driver) next;       /**< Next afu driver. */
+	struct rte_driver driver;               /**< Inherit core driver. */
+	afu_probe_t *probe;                     /**< Device Probe function. */
+	afu_remove_t *remove;                   /**< Device Remove function. */
+	struct rte_afu_id id;	                /**< AFU id within FPGA. */
+	uint32_t drv_flags;         /**< Flags contolling handling of device. */
+};
+
+/**
+ * Structure describing the Intel FPGA bus
+ */
+struct rte_ifpga_bus {
+	struct rte_bus bus;               /**< Inherit the generic class */
+	struct rte_ifpga_device_list ifpga_list;  /**< List of FPGA devices */
+	struct rte_afu_device_list afu_list;  /**< List of AFU devices */
+	struct rte_afu_driver_list driver_list;  /**< List of FPGA drivers */
+};
+
+static inline const char *
+rte_ifpga_device_name(const struct rte_afu_device *afu)
+{
+	if (afu && afu->device.name)
+		return afu->device.name;
+	return NULL;
+}
+
+extern struct rte_ifpga_bus rte_ifpga_bus;
+
+void rte_ifpga_driver_register(struct rte_afu_driver *driver);
+void rte_ifpga_driver_unregister(struct rte_afu_driver *driver);
+int rte_ifpga_scan_2nd(void *conf, void *id);
+
+
+#endif /* _RTE_BUS_IFPGA_H_ */
diff --git a/drivers/bus/ifpga/rte_bus_ifpga_version.map b/drivers/bus/ifpga/rte_bus_ifpga_version.map
new file mode 100644
index 0000000..e2aa7da
--- /dev/null
+++ b/drivers/bus/ifpga/rte_bus_ifpga_version.map
@@ -0,0 +1,8 @@
+DPDK_17.11 {
+	global:
+
+    rte_ifpga_driver_register;
+    rte_ifpga_driver_unregister;
+
+	local: *;
+};
-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [dpdk-dev] [PATCH V2 4/5] Add Intel FPGA BUS Rawdev Code
  2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
                   ` (2 preceding siblings ...)
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code Rosen Xu
@ 2018-03-21  7:51 ` Rosen Xu
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 5/5] Add Intel OPAE Share Code Rosen Xu
  2018-03-21 10:00 ` [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Gaëtan Rivet
  5 siblings, 0 replies; 19+ messages in thread
From: Rosen Xu @ 2018-03-21  7:51 UTC (permalink / raw)
  To: dev
  Cc: declan.doherty, bruce.richardson, shreyansh.jain, tianfei.zhang,
	hao.wu, gaetan.rivet, Yanglong Wu

Signed-off-by: Rosen Xu <rosen.xu@intel.com>
Signed-off-by: Yanglong Wu  <yanglong.wu@intel.com>
---
 drivers/raw/Makefile                               |   1 +
 drivers/raw/ifpga_rawdev/Makefile                  |  63 +++
 drivers/raw/ifpga_rawdev/ifpga_rawdev.c            | 486 +++++++++++++++++++++
 drivers/raw/ifpga_rawdev/ifpga_rawdev.h            |  38 ++
 drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c    |  99 +++++
 .../ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map  |   4 +
 6 files changed, 691 insertions(+)
 create mode 100644 drivers/raw/ifpga_rawdev/Makefile
 create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev.c
 create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev.h
 create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c
 create mode 100644 drivers/raw/ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map

diff --git a/drivers/raw/Makefile b/drivers/raw/Makefile
index da7c8b4..47cc567 100644
--- a/drivers/raw/Makefile
+++ b/drivers/raw/Makefile
@@ -5,5 +5,6 @@ include $(RTE_SDK)/mk/rte.vars.mk
 
 # DIRS-$(<configuration>) += <directory>
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_SKELETON_RAWDEV) += skeleton_rawdev
+DIRS-y += ifpga_rawdev
 
 include $(RTE_SDK)/mk/rte.subdir.mk
diff --git a/drivers/raw/ifpga_rawdev/Makefile b/drivers/raw/ifpga_rawdev/Makefile
new file mode 100644
index 0000000..7efc2f5
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/Makefile
@@ -0,0 +1,63 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
+#   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_ifpga_rawdev.a
+
+CFLAGS += -DALLOW_EXPERIMENTAL_API
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -I$(RTE_SDK)/drivers/bus/ifpga
+CFLAGS += -I$(RTE_SDK)/drivers/raw/ifpga_rawdev
+LDLIBS += -lrte_eal
+LDLIBS += -lrte_rawdev
+LDLIBS += -lrte_bus_vdev
+LDLIBS += -lrte_kvargs
+
+EXPORT_MAP := rte_pmd_ifpga_rawdev_version.map
+
+LIBABIVER := 1
+
+VPATH += $(SRCDIR)/base
+
+include $(RTE_SDK)/drivers/raw/ifpga_rawdev/base/Makefile
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_SKELETON_RAWDEV) += ifpga_rawdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_SKELETON_RAWDEV) += ifpga_rawdev_example.c
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/raw/ifpga_rawdev/ifpga_rawdev.c b/drivers/raw/ifpga_rawdev/ifpga_rawdev.c
new file mode 100644
index 0000000..a2a41d2
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/ifpga_rawdev.c
@@ -0,0 +1,486 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#include <string.h>
+#include <dirent.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <fcntl.h>
+#include <rte_log.h>
+#include <rte_bus.h>
+#include <rte_eal_memconfig.h>
+#include <rte_malloc.h>
+#include <rte_devargs.h>
+#include <rte_memcpy.h>
+#include <rte_ethdev.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
+#include <rte_kvargs.h>
+#include <rte_alarm.h>
+
+#include <rte_errno.h>
+#include <rte_per_lcore.h>
+#include <rte_memory.h>
+#include <rte_memzone.h>
+#include <rte_eal.h>
+#include <rte_common.h>
+
+#include "base/opae_hw_api.h"
+#include "rte_rawdev.h"
+#include "rte_rawdev_pmd.h"
+#include "ifpga_logs.h"
+#include "ifpga_common.h"
+#include "rte_bus_ifpga.h"
+#include "ifpga_rawdev.h"
+
+int ifpga_rawdev_logtype;
+
+#define PCI_VENDOR_ID_INTEL          0x8086
+/* PCI Device ID */
+#define PCIe_DEVICE_ID_RCiEP0_MCP    0xBCBD
+#define PCIe_DEVICE_ID_RCiEP0_SKX_P  0xBCC0
+#define PCIe_DEVICE_ID_RCiEP0_DCP    0x09C4
+/* VF Device */
+#define PCIe_DEVICE_ID_VF_MCP        0xBCBF
+#define PCIe_DEVICE_ID_VF_SKX_P      0xBCC1
+#define PCIe_DEVICE_ID_VF_DCP        0x09C5
+#define RTE_MAX_RAW_DEVICE           10
+
+static const struct rte_pci_id pci_ifpga_map[] = {
+	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCIe_DEVICE_ID_RCiEP0_MCP) },
+	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCIe_DEVICE_ID_VF_MCP) },
+	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCIe_DEVICE_ID_RCiEP0_SKX_P) },
+	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCIe_DEVICE_ID_VF_SKX_P) },
+	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCIe_DEVICE_ID_RCiEP0_DCP) },
+	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCIe_DEVICE_ID_VF_DCP) },
+	{ .vendor_id = 0, /* sentinel */ },
+};
+
+static int ifpga_fill_afu_dev(struct opae_accelerator *acc,
+		struct rte_afu_device *afu_dev)
+{
+	struct rte_mem_resource *res = afu_dev->mem_resource;
+	struct opae_acc_region_info region_info;
+	struct opae_acc_info info;
+	unsigned long i;
+	int ret;
+
+	ret = opae_acc_get_info(acc, &info);
+	if (ret)
+		return ret;
+
+	if (info.num_regions > PCI_MAX_RESOURCE)
+		return -EFAULT;
+
+	afu_dev->num_region = info.num_regions;
+
+	for (i = 0; i < info.num_regions; i++) {
+		region_info.index = i;
+		ret = opae_acc_get_region_info(acc, &region_info);
+		if (ret)
+			return ret;
+
+		if ((region_info.flags & ACC_REGION_MMIO) &&
+		    (region_info.flags & ACC_REGION_READ) &&
+		    (region_info.flags & ACC_REGION_WRITE)) {
+			res[i].phys_addr = region_info.phys_addr;
+			res[i].len = region_info.len;
+			res[i].addr = region_info.addr;
+		} else
+			return -EFAULT;
+	}
+
+	return 0;
+}
+
+static void ifpga_rawdev_info_get(struct rte_rawdev *dev,
+				     rte_rawdev_obj_t dev_info)
+{
+	struct opae_adapter *adapter;
+	struct opae_accelerator *acc;
+	struct rte_afu_device *afu_dev;
+
+	IFPGA_RAWDEV_PMD_FUNC_TRACE();
+
+	if (!dev_info) {
+		IFPGA_RAWDEV_PMD_ERR("Invalid request");
+		return;
+	}
+
+	adapter = ifpga_rawdev_get_priv(dev);
+	if (!adapter)
+		return;
+
+	afu_dev = dev_info;
+	afu_dev->rawdev = dev;
+
+	/* find opae_accelerator and fill info into afu_device */
+	opae_adapter_for_each_acc(adapter, acc) {
+		if (acc->index != afu_dev->id.port)
+			continue;
+
+		if (ifpga_fill_afu_dev(acc, afu_dev)) {
+			IFPGA_RAWDEV_PMD_ERR("cannot get info\n");
+			return;
+		}
+	}
+}
+
+static int ifpga_rawdev_start(struct rte_rawdev *dev)
+{
+	int ret = 0;
+	struct opae_adapter *adapter;
+
+	IFPGA_RAWDEV_PMD_FUNC_TRACE();
+
+	RTE_FUNC_PTR_OR_ERR_RET(dev, -EINVAL);
+
+	adapter = ifpga_rawdev_get_priv(dev);
+	if (!adapter)
+		return -ENODEV;
+
+	return ret;
+}
+
+static void ifpga_rawdev_stop(struct rte_rawdev *dev)
+{
+	dev->started = 0;
+}
+
+static int ifpga_rawdev_close(struct rte_rawdev *dev)
+{
+	if (!dev)
+		return 1;
+	else
+		return 0;
+}
+
+static int ifpga_rawdev_reset(struct rte_rawdev *dev)
+{
+	if (!dev)
+		return 1;
+	else
+		return 0;
+}
+
+static int
+fpga_pr(struct rte_rawdev *raw_dev, u32 port_id, u64 *buffer, u32 size,
+			u64 *status){
+
+	struct opae_adapter *adapter;
+	struct opae_manager *mgr;
+	struct opae_accelerator *acc;
+	struct opae_bridge *br;
+	int ret;
+
+	adapter = ifpga_rawdev_get_priv(raw_dev);
+	if (!adapter)
+		return -ENODEV;
+
+	mgr = opae_adapter_get_mgr(adapter);
+	if (!mgr)
+		return -ENODEV;
+
+	acc = opae_adapter_get_acc(adapter, port_id);
+	if (!acc)
+		return -ENODEV;
+
+	br = opae_acc_get_br(acc);
+	if (!br)
+		return -ENODEV;
+
+	ret = opae_manager_flash(mgr, port_id, buffer, size, status);
+	if (ret) {
+		printf("%s pr error %d\n", __func__, ret);
+		return ret;
+	}
+
+	usleep(100);
+	ret = opae_bridge_reset(br);
+	if (ret) {
+		printf("%s reset port:%d error %d\n", __func__, port_id, ret);
+		return ret;
+	}
+
+	return ret;
+}
+
+static int rte_fpga_do_pr(struct rte_rawdev *rawdev, int port_id,
+		const char *file_name)
+{
+	struct stat file_stat;
+	int file_fd;
+	int ret = 0;
+	u32 buffer_size;
+	void *buffer;
+	u64 pr_error;
+
+	if (!file_name)
+		return -EINVAL;
+
+	file_fd = open(file_name, O_RDONLY);
+	if (file_fd < 0) {
+		printf("%s: open file error: %s\n", __func__, file_name);
+		printf("Message : %s\n", strerror(errno));
+		return -EINVAL;
+	}
+	ret = stat(file_name, &file_stat);
+	if (ret) {
+		printf("stat on bitstream file failed: %s\n", file_name);
+		return -EINVAL;
+	}
+	buffer_size = file_stat.st_size;
+	printf("bitstream file size: %u\n", buffer_size);
+	buffer = rte_malloc(NULL, buffer_size, 0);
+	if (!buffer) {
+		ret = -ENOMEM;
+		goto close_fd;
+	}
+
+	/*read the raw data*/
+	if (buffer_size != read(file_fd, (void *)buffer, buffer_size)) {
+		ret = -EINVAL;
+		goto free_buffer;
+	}
+
+	/*do PR now*/
+	ret = fpga_pr(rawdev, port_id, buffer, buffer_size, &pr_error);
+	printf("downloading to device port %d....%s.\n", port_id,
+		ret ? "failed" : "success");
+	if (ret) {
+		ret = -EINVAL;
+		//raw_dev->ops->show_pr_error(pr_error);
+		goto free_buffer;
+	}
+
+free_buffer:
+	if (buffer)
+		rte_free(buffer);
+close_fd:
+	close(file_fd);
+	file_fd = 0;
+	return ret;
+}
+
+static int ifpga_rawdev_pr(struct rte_rawdev *dev,
+	rte_rawdev_obj_t pr_conf)
+{
+	struct opae_adapter *adapter;
+	struct rte_afu_pr_conf *afu_pr_conf;
+	int ret;
+
+	IFPGA_RAWDEV_PMD_FUNC_TRACE();
+
+	adapter = ifpga_rawdev_get_priv(dev);
+	if (!adapter)
+		return -ENODEV;
+
+	if (!pr_conf)
+		return -EINVAL;
+
+	afu_pr_conf = pr_conf;
+
+	ret = rte_fpga_do_pr(dev, afu_pr_conf->afu_id.port,
+			afu_pr_conf->bs_path);
+	if (ret) {
+		printf("===do pr error %d\n", ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+static const struct rte_rawdev_ops ifpga_rawdev_ops = {
+	.dev_info_get = ifpga_rawdev_info_get,
+	.dev_configure = NULL,
+	.dev_start = ifpga_rawdev_start,
+	.dev_stop = ifpga_rawdev_stop,
+	.dev_close = ifpga_rawdev_close,
+	.dev_reset = ifpga_rawdev_reset,
+
+	.queue_def_conf = NULL,
+	.queue_setup = NULL,
+	.queue_release = NULL,
+
+	.attr_get = NULL,
+	.attr_set = NULL,
+
+	.enqueue_bufs = NULL,
+	.dequeue_bufs = NULL,
+
+	.dump = NULL,
+
+	.xstats_get = NULL,
+	.xstats_get_names = NULL,
+	.xstats_get_by_name = NULL,
+	.xstats_reset = NULL,
+
+	.firmware_status_get = NULL,
+	.firmware_version_get = NULL,
+	.firmware_load = ifpga_rawdev_pr,
+	.firmware_unload = NULL,
+
+	.dev_selftest = NULL,
+};
+
+static int
+ifpga_rawdev_create(struct rte_pci_device *pci_dev,
+			int socket_id)
+{
+	int ret = 0;
+	struct rte_rawdev *rawdev = NULL;
+	struct opae_adapter *adapter;
+	struct opae_manager *mgr;
+	struct opae_adapter_data_pci *data;
+	char name[RTE_RAWDEV_NAME_MAX_LEN];
+	int i;
+
+	if (!pci_dev) {
+		IFPGA_RAWDEV_PMD_ERR("Invalid pci_dev of the device!");
+		ret = -EINVAL;
+		goto cleanup;
+	}
+
+	memset(name, 0, sizeof(name));
+	snprintf(name, RTE_RAWDEV_NAME_MAX_LEN, "IFPGA:%x:%x:%x",
+	pci_dev->addr.bus, pci_dev->addr.devid, pci_dev->addr.function);
+
+	IFPGA_RAWDEV_PMD_INFO("Init %s on NUMA node %d", name, rte_socket_id());
+
+	/* Allocate device structure */
+	rawdev = rte_rawdev_pmd_allocate(name, sizeof(struct opae_adapter),
+					 socket_id);
+	if (rawdev == NULL) {
+		IFPGA_RAWDEV_PMD_ERR("Unable to allocate rawdevice");
+		ret = -EINVAL;
+		goto cleanup;
+	}
+
+	/* alloc OPAE_FPGA_PCI data to register to OPAE hardware level API */
+	data = opae_adapter_data_alloc(OPAE_FPGA_PCI);
+	if (!data)
+		return -ENOMEM;
+
+	/* init opae_adapter_data_pci for device specific information */
+	for (i = 0; i < PCI_MAX_RESOURCE; i++) {
+		data->region[i].phys_addr = pci_dev->mem_resource[i].phys_addr;
+		data->region[i].len = pci_dev->mem_resource[i].len;
+		data->region[i].addr = pci_dev->mem_resource[i].addr;
+	}
+	data->device_id = pci_dev->id.device_id;
+	data->vendor_id = pci_dev->id.vendor_id;
+
+	/* create a opae_adapter based on above device data */
+	adapter = opae_adapter_alloc(pci_dev->device.name, data);
+	if (!adapter) {
+		opae_adapter_data_free(data);
+		return -ENOMEM;
+	}
+
+	rawdev->dev_ops = &ifpga_rawdev_ops;
+	rawdev->device = &pci_dev->device;
+	rawdev->driver_name = pci_dev->device.driver->name;
+
+	rawdev->dev_private = adapter;
+
+	/* must enumerate the adapter before use it */
+	ret = opae_adapter_enumerate(adapter);
+	if (ret)
+		return ret;
+
+	/* set opae_manager to rawdev */
+	mgr = opae_adapter_get_mgr(adapter);
+	if (mgr) {
+		/*PF function*/
+		IFPGA_RAWDEV_PMD_INFO("this is a PF function");
+	}
+
+	rte_ifpga_scan_2nd(rawdev, &pci_dev->addr);
+
+	return ret;
+
+cleanup:
+	if (rawdev)
+		rte_rawdev_pmd_release(rawdev);
+
+	return ret;
+}
+
+static int
+ifpga_rawdev_destroy(struct rte_pci_device *pci_dev)
+{
+	int ret;
+	struct rte_rawdev *rawdev;
+	char name[RTE_RAWDEV_NAME_MAX_LEN];
+	struct opae_adapter *adapter;
+
+	if (!pci_dev) {
+		IFPGA_RAWDEV_PMD_ERR("Invalid pci_dev of the device!");
+		ret = -EINVAL;
+		return ret;
+	}
+
+	memset(name, 0, sizeof(name));
+	snprintf(name, RTE_RAWDEV_NAME_MAX_LEN, "IFPGA:%x:%x:%x",
+		pci_dev->addr.bus, pci_dev->addr.devid, pci_dev->addr.function);
+
+	IFPGA_RAWDEV_PMD_INFO("Closing %s on NUMA node %d",
+		name, rte_socket_id());
+
+	rawdev = rte_rawdev_pmd_get_named_dev(name);
+	if (!rawdev) {
+		IFPGA_RAWDEV_PMD_ERR("Invalid device name (%s)", name);
+		return -EINVAL;
+	}
+
+	adapter = ifpga_rawdev_get_priv(rawdev);
+	if (!adapter)
+		return -ENODEV;
+
+	opae_adapter_data_free(adapter->data);
+	opae_adapter_free(adapter);
+
+	/* rte_rawdev_close is called by pmd_release */
+	ret = rte_rawdev_pmd_release(rawdev);
+	if (ret)
+		IFPGA_RAWDEV_PMD_DEBUG("Device cleanup failed");
+
+	return 0;
+}
+
+static int
+ifpga_rawdev_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev)
+{
+
+	printf("## %s\n", __func__);
+	return ifpga_rawdev_create(pci_dev, rte_socket_id());
+}
+
+static int ifpga_rawdev_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return ifpga_rawdev_destroy(pci_dev);
+}
+
+static struct rte_pci_driver rte_ifpga_rawdev_pmd = {
+	.id_table  = pci_ifpga_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe     = ifpga_rawdev_pci_probe,
+	.remove    = ifpga_rawdev_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(ifpga_rawdev_pci_driver, rte_ifpga_rawdev_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(ifpga_rawdev_pci_driver, rte_ifpga_rawdev_pmd);
+RTE_PMD_REGISTER_KMOD_DEP(ifpga_rawdev_pci_driver, "* igb_uio | uio_pci_generic | vfio-pci");
+
+RTE_INIT(ifpga_rawdev_init_log);
+static void
+ifpga_rawdev_init_log(void)
+{
+	ifpga_rawdev_logtype = rte_log_register("driver.raw.init");
+	if (ifpga_rawdev_logtype >= 0)
+		rte_log_set_level(ifpga_rawdev_logtype, RTE_LOG_NOTICE);
+}
diff --git a/drivers/raw/ifpga_rawdev/ifpga_rawdev.h b/drivers/raw/ifpga_rawdev/ifpga_rawdev.h
new file mode 100644
index 0000000..f346d1a
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/ifpga_rawdev.h
@@ -0,0 +1,38 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2015 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#ifndef _IFPGA_RAWDEV_H_
+#define _IFPGA_RAWDEV_H_
+
+extern int ifpga_rawdev_logtype;
+
+#define IFPGA_RAWDEV_PMD_LOG(level, fmt, args...) \
+	rte_log(RTE_LOG_ ## level, ifpga_rawdev_logtype, "%s(): " fmt "\n", \
+		__func__, ##args)
+
+#define IFPGA_RAWDEV_PMD_FUNC_TRACE() IFPGA_RAWDEV_PMD_LOG(DEBUG, ">>")
+
+#define IFPGA_RAWDEV_PMD_DEBUG(fmt, args...) \
+	IFPGA_RAWDEV_PMD_LOG(DEBUG, fmt, ## args)
+#define IFPGA_RAWDEV_PMD_INFO(fmt, args...) \
+	IFPGA_RAWDEV_PMD_LOG(INFO, fmt, ## args)
+#define IFPGA_RAWDEV_PMD_ERR(fmt, args...) \
+	IFPGA_RAWDEV_PMD_LOG(ERR, fmt, ## args)
+#define IFPGA_RAWDEV_PMD_WARN(fmt, args...) \
+	IFPGA_RAWDEV_PMD_LOG(WARNING, fmt, ## args)
+
+enum ifpga_rawdev_device_state {
+	IFPGA_IDLE,
+	IFPGA_READY,
+	IFPGA_ERROR
+};
+
+static inline struct opae_adapter *
+ifpga_rawdev_get_priv(const struct rte_rawdev *rawdev)
+{
+	return rawdev->dev_private;
+}
+
+#endif /* _IFPGA_RAWDEV_H_ */
diff --git a/drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c b/drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c
new file mode 100644
index 0000000..e073db2
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c
@@ -0,0 +1,99 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation.
+ * Copyright 2013-2014 6WIND S.A.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <sys/types.h>
+#include <sys/queue.h>
+#include <netinet/in.h>
+#include <setjmp.h>
+#include <stdarg.h>
+#include <ctype.h>
+#include <errno.h>
+#include <getopt.h>
+#include <signal.h>
+#include <stdbool.h>
+
+#include <rte_common.h>
+#include <rte_log.h>
+#include <rte_malloc.h>
+#include <rte_memory.h>
+#include <rte_memcpy.h>
+#include <rte_eal.h>
+#include <rte_launch.h>
+#include <rte_atomic.h>
+#include <rte_cycles.h>
+#include <rte_prefetch.h>
+#include <rte_lcore.h>
+#include <rte_per_lcore.h>
+#include <rte_branch_prediction.h>
+#include <rte_interrupts.h>
+#include <rte_random.h>
+#include <rte_debug.h>
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_io.h>
+
+#include <rte_errno.h>
+#include <rte_bus.h>
+#include <rte_memzone.h>
+
+#include <rte_devargs.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
+#include <rte_alarm.h>
+
+#include "rte_bus_ifpga.h"
+#include "ifpga_logs.h"
+
+#define RTE_PMD_REGISTER_AFU(nm, afudrv)\
+RTE_INIT(afudrvinitfn_ ##afudrv);\
+static const char *afudrvinit_ ## nm ## _alias;\
+static void afudrvinitfn_ ##afudrv(void)\
+{\
+	(afudrv).driver.name = RTE_STR(nm);\
+	(afudrv).driver.alias = afudrvinit_ ## nm ## _alias;\
+	rte_ifpga_driver_register(&afudrv);\
+} \
+RTE_PMD_EXPORT_NAME(nm, __COUNTER__)
+
+#define RTE_PMD_REGISTER_AFU_ALIAS(nm, alias)\
+static const char *afudrvinit_ ## nm ## _alias = RTE_STR(alias)
+
+static int afu_dev_probe(struct rte_afu_device *afu_dev)
+{
+	if (!afu_dev)
+		return 1;
+	else
+		return 0;
+}
+static int afu_dev_remove(struct rte_afu_device *afu_dev)
+{
+	if (!afu_dev)
+		return 1;
+	else
+		return 0;
+}
+
+static struct rte_afu_driver afu_dev_driver = {
+	.probe = afu_dev_probe,
+	.remove = afu_dev_remove,
+};
+
+RTE_PMD_REGISTER_AFU(net_afu_drv_example, afu_dev_driver);
+RTE_PMD_REGISTER_AFU_ALIAS(net_afu_drv_example, afu_dev);
+RTE_PMD_REGISTER_PARAM_STRING(net_afu_drv_example,
+	"bdf=<string> "
+	"port=<int> "
+	"uudi_high=<int64> "
+	"uuid_low=<int64> "
+	"path=<string> "
+	"pr_enable=<int>"
+	"debug=<int>");
diff --git a/drivers/raw/ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map b/drivers/raw/ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map
new file mode 100644
index 0000000..179140f
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map
@@ -0,0 +1,4 @@
+DPDK_18.02 {
+
+	local: *;
+};
-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [dpdk-dev] [PATCH V2 5/5] Add Intel OPAE Share Code
  2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
                   ` (3 preceding siblings ...)
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 4/5] Add Intel FPGA BUS Rawdev Code Rosen Xu
@ 2018-03-21  7:51 ` Rosen Xu
  2018-03-21 10:00 ` [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Gaëtan Rivet
  5 siblings, 0 replies; 19+ messages in thread
From: Rosen Xu @ 2018-03-21  7:51 UTC (permalink / raw)
  To: dev
  Cc: declan.doherty, bruce.richardson, shreyansh.jain, tianfei.zhang,
	hao.wu, gaetan.rivet

Signed-off-by: Tianfei Zhang <tianfei.zhang@intel.com>
Signed-off-by: Wu, Hao <hao.wu@intel.com>
---
 drivers/raw/ifpga_rawdev/base/Makefile             |   54 +
 drivers/raw/ifpga_rawdev/base/ifpga_api.c          |  543 +++++++
 drivers/raw/ifpga_rawdev/base/ifpga_api.h          |   77 +
 drivers/raw/ifpga_rawdev/base/ifpga_compat.h       |   84 +
 drivers/raw/ifpga_rawdev/base/ifpga_defines.h      | 1696 ++++++++++++++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c    |  861 ++++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h    |   38 +
 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c  |  340 ++++
 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h  |  197 +++
 drivers/raw/ifpga_rawdev/base/ifpga_fme.c          |  763 +++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c    |  328 ++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c    |  430 +++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c    |  742 +++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c       |  395 +++++
 drivers/raw/ifpga_rawdev/base/ifpga_hw.h           |  152 ++
 drivers/raw/ifpga_rawdev/base/ifpga_port.c         |  730 +++++++++
 drivers/raw/ifpga_rawdev/base/ifpga_port_error.c   |  236 +++
 drivers/raw/ifpga_rawdev/base/opae_debug.c         |  126 ++
 drivers/raw/ifpga_rawdev/base/opae_debug.h         |   46 +
 drivers/raw/ifpga_rawdev/base/opae_hw_api.c        |  389 +++++
 drivers/raw/ifpga_rawdev/base/opae_hw_api.h        |  276 ++++
 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c  |  151 ++
 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h  |  293 ++++
 drivers/raw/ifpga_rawdev/base/opae_osdep.h         |  111 ++
 .../ifpga_rawdev/base/osdep_raw/osdep_generic.h    |  104 ++
 .../ifpga_rawdev/base/osdep_rte/osdep_generic.h    |   72 +
 26 files changed, 9234 insertions(+)
 create mode 100644 drivers/raw/ifpga_rawdev/base/Makefile
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_api.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_api.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_compat.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_defines.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_hw.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_port.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_port_error.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_debug.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_debug.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_hw_api.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_hw_api.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/opae_osdep.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/osdep_raw/osdep_generic.h
 create mode 100644 drivers/raw/ifpga_rawdev/base/osdep_rte/osdep_generic.h

diff --git a/drivers/raw/ifpga_rawdev/base/Makefile b/drivers/raw/ifpga_rawdev/base/Makefile
new file mode 100644
index 0000000..9c0d64b
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/Makefile
@@ -0,0 +1,54 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2017-2018 Intel Corporation. All rights reserved.
+#   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.
+
+ifneq ($(CONFIG_RTE_LIBRTE_EAL),)
+OSDEP := osdep_rte
+else
+OSDEP := osdep_raw
+endif
+
+CFLAGS += -I$(RTE_SDK)/drivers/raw/ifpga_rawdev/base/$(OSDEP)
+
+SRCS-y += ifpga_api.c
+SRCS-y += ifpga_enumerate.c
+SRCS-y += ifpga_feature_dev.c
+SRCS-y += ifpga_fme.c
+SRCS-y += ifpga_fme_iperf.c
+SRCS-y += ifpga_fme_dperf.c
+SRCS-y += ifpga_fme_error.c
+SRCS-y += ifpga_port.c
+SRCS-y += ifpga_port_error.c
+SRCS-y += ifpga_fme_pr.c
+SRCS-y += opae_hw_api.c
+SRCS-y += opae_ifpga_hw_api.c
+SRCS-y += opae_debug.c
+
+SRCS-y += $(wildcard $(SRCDIR)/base/$(OSDEP)/*.c)
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_api.c b/drivers/raw/ifpga_rawdev/base/ifpga_api.c
new file mode 100644
index 0000000..54a2e42
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_api.c
@@ -0,0 +1,543 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_api.h"
+#include "ifpga_enumerate.h"
+#include "ifpga_feature_dev.h"
+
+#include "opae_hw_api.h"
+
+/* Accelerator APIs */
+static int ifpga_acc_set_irq(struct opae_accelerator *acc,
+			     u32 start, u32 count, s32 evtfds[])
+{
+	struct ifpga_afu_info *afu_info = acc->data;
+	struct opae_bridge *br = acc->br;
+	struct ifpga_port_hw *port;
+	struct fpga_uafu_irq_set irq_set;
+
+	if (!br || !br->data)
+		return -EINVAL;
+
+	if (start >= afu_info->num_irqs || start + count > afu_info->num_irqs)
+		return -EINVAL;
+
+	port = br->data;
+
+	irq_set.start = start;
+	irq_set.count = count;
+	irq_set.evtfds = evtfds;
+
+	return ifpga_set_irq(port->parent, FEATURE_FIU_ID_PORT, port->port_id,
+			     PORT_FEATURE_ID_UINT, &irq_set);
+}
+
+static int ifpga_acc_get_info(struct opae_accelerator *acc,
+			      struct opae_acc_info *info)
+{
+	struct ifpga_afu_info *afu_info = acc->data;
+
+	if (!afu_info)
+		return -ENODEV;
+
+	info->num_regions = afu_info->num_regions;
+	info->num_irqs = afu_info->num_irqs;
+
+	return 0;
+}
+
+static int ifpga_acc_get_region_info(struct opae_accelerator *acc,
+				     struct opae_acc_region_info *info)
+{
+	struct ifpga_afu_info *afu_info = acc->data;
+
+	if (!afu_info)
+		return -EINVAL;
+
+	if (afu_info->num_regions >= info->index)
+		return -EINVAL;
+
+	/* always one RW region only for AFU now */
+	info->flags = ACC_REGION_READ | ACC_REGION_WRITE | ACC_REGION_MMIO;
+	info->len = afu_info->region[info->index].len;
+
+	return 0;
+}
+
+static int ifpga_acc_read(struct opae_accelerator *acc, unsigned int region_idx,
+			  u64 offset, unsigned int byte, void *data)
+{
+	struct ifpga_afu_info *afu_info = acc->data;
+	struct opae_reg_region *region;
+
+	if (!afu_info)
+		return -EINVAL;
+
+	if (offset + byte <= offset)
+		return -EINVAL;
+
+	if (region_idx >= afu_info->num_regions)
+		return -EINVAL;
+
+	region = &afu_info->region[region_idx];
+	if (offset + byte > region->len)
+		return -EINVAL;
+
+	switch (byte) {
+	case 8:
+		*(u64  *)data = opae_readq(region->addr + offset);
+		break;
+	case 4:
+		*(u32 *)data = opae_readl(region->addr + offset);
+		break;
+	case 2:
+		*(u16 *)data = opae_readw(region->addr + offset);
+		break;
+	case 1:
+		*(u8 *)data = opae_readb(region->addr + offset);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int ifpga_acc_write(struct opae_accelerator *acc,
+			   unsigned int region_idx, u64 offset,
+			   unsigned int byte, void *data)
+{
+	struct ifpga_afu_info *afu_info = acc->data;
+	struct opae_reg_region *region;
+
+	if (!afu_info)
+		return -EINVAL;
+
+	if (offset + byte <= offset)
+		return -EINVAL;
+
+	if (region_idx >= afu_info->num_regions)
+		return -EINVAL;
+
+	region = &afu_info->region[region_idx];
+	if (offset + byte > region->len)
+		return -EINVAL;
+
+	/* normal mmio case */
+	switch (byte) {
+	case 8:
+		opae_writeq(*(u64 *)data, region->addr + offset);
+		break;
+	case 4:
+		opae_writel(*(u32 *)data, region->addr + offset);
+		break;
+	case 2:
+		opae_writew(*(u16 *)data, region->addr + offset);
+		break;
+	case 1:
+		opae_writeb(*(u8 *)data, region->addr + offset);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+struct opae_accelerator_ops ifpga_acc_ops = {
+	.read = ifpga_acc_read,
+	.write = ifpga_acc_write,
+	.set_irq = ifpga_acc_set_irq,
+	.get_info = ifpga_acc_get_info,
+	.get_region_info = ifpga_acc_get_region_info,
+};
+
+/* Bridge APIs */
+
+static int ifpga_br_reset(struct opae_bridge *br)
+{
+	struct ifpga_port_hw *port = br->data;
+
+	return fpga_port_reset(port);
+}
+
+struct opae_bridge_ops ifpga_br_ops = {
+	.reset = ifpga_br_reset,
+};
+
+/* Manager APIs */
+static int ifpga_mgr_flash(struct opae_manager *mgr, int id, void *buf,
+			   u32 size, u64 *status)
+{
+	struct ifpga_fme_hw *fme = mgr->data;
+	struct ifpga_hw *hw = fme->parent;
+
+	return ifpga_pr(hw, id, buf, size, status);
+}
+
+struct opae_manager_ops ifpga_mgr_ops = {
+	.flash = ifpga_mgr_flash,
+};
+
+/* Adapter APIs */
+static int ifpga_adapter_enumerate(struct opae_adapter *adapter)
+{
+	struct ifpga_hw *hw = malloc(sizeof(*hw));
+
+	if (hw) {
+		memset(hw, 0, sizeof(*hw));
+		hw->pci_data = adapter->data;
+		hw->adapter = adapter;
+		if (ifpga_bus_enumerate(hw))
+			goto error;
+		return ifpga_bus_init(hw);
+	}
+
+error:
+	return -ENOMEM;
+}
+
+struct opae_adapter_ops ifpga_adapter_ops = {
+	.enumerate = ifpga_adapter_enumerate,
+};
+
+/**
+ *  ifpga_enumerate - enumerate the Device Feature List
+ *  @hw: pointer to the HW structure
+ *
+ *  This function enumerate the Device Feature List to discover
+ *  the FME and Port devices. The FME device and Port devices
+ *  will fill into HW structure when enumerate done.
+ *
+ *  @return
+ *   - 0: Success, device enumated.
+ *   - <0: Error code returned in enumeration.
+ **/
+int ifpga_enumerate(struct ifpga_hw *hw)
+{
+	return ifpga_bus_enumerate(hw);
+}
+
+/**
+ *  ifpga_reset_port - reset a port device
+ *  @hw: pointer to the HW structure
+ *  @port_id: port device id
+ *
+ *  @return
+ *   - 0: Success
+ *   - <0: Failure to reset a port device.
+ **/
+int ifpga_reset_port(struct ifpga_hw *hw, u32 port_id)
+{
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	return fpga_port_reset(&hw->port[port_id]);
+}
+
+/**
+ *  ifpga_get_afu_mmio_addr - get UAFU MMIO address
+ *  @hw: pointer to the HW structure
+ *  @port_id: port device id
+ *  @mem_resource: afu mmio resource
+ *  @num_resource: number of resource
+ *
+ *  return 0 on success or error code
+ **/
+int ifpga_get_afu_mmio_info(struct ifpga_hw *hw, u32 port_id,
+		struct rte_mem_resource *mem_resource,
+		u32 *num_resource)
+{
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	mem_resource->addr =
+		hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].addr;
+	mem_resource->len =
+		hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].size;
+	mem_resource->phys_addr =
+		hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].phys_addr;
+
+	*num_resource = 1;
+
+	return 0;
+}
+
+/**
+ *  ifpga_get_afu_uuid - get AFU uuid
+ *  @hw: pointer to the HW structure
+ *  @port_id: port device id
+ *  @uuid: the AFU's uuid
+ *
+ *  @return
+ *   - 0: Success
+ *   - <0: Failure to get uuid.
+ **/
+int ifpga_get_afu_uuid(struct ifpga_hw *hw, u32 port_id, struct uuid *uuid)
+{
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	if (!hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].size)
+		return -ENODEV;
+
+	return fpga_get_afu_uuid(&hw->port[port_id], uuid);
+}
+
+/**
+ *  ifpga_pr - do the partial reconfiguration for a given port device
+ *  @hw: pointer to the HW structure
+ *  @port_id: the port device id
+ *  @buffer: the buffer of the bitstream
+ *  @size: the size of the bitstream
+ *  @status: hardware status including PR error code if return -EIO.
+ *
+ *  @return
+ *   - 0: Success, partial reconfiguration finished.
+ *   - <0: Error code returned in partial reconfiguration.
+ **/
+int ifpga_pr(struct ifpga_hw *hw, u32 port_id, void *buffer, u32 size,
+	     u64 *status)
+{
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	return do_pr(hw, port_id, buffer, size, status);
+}
+
+int ifpga_get_prop(struct ifpga_hw *hw, u32 fiu_id, u32 port_id,
+		   struct feature_prop *prop)
+{
+	if (!hw || !prop)
+		return -EINVAL;
+
+	switch (fiu_id) {
+	case FEATURE_FIU_ID_FME:
+		return fme_get_prop(&hw->fme, prop);
+	case FEATURE_FIU_ID_PORT:
+		if (!is_valid_port_id(hw, port_id))
+			return -ENODEV;
+		return port_get_prop(&hw->port[port_id], prop);
+	}
+
+	return -ENOENT;
+}
+
+int ifpga_set_prop(struct ifpga_hw *hw, u32 fiu_id, u32 port_id,
+		   struct feature_prop *prop)
+{
+	if (!hw || !prop)
+		return -EINVAL;
+
+	switch (fiu_id) {
+	case FEATURE_FIU_ID_FME:
+		return fme_set_prop(&hw->fme, prop);
+	case FEATURE_FIU_ID_PORT:
+		if (!is_valid_port_id(hw, port_id))
+			return -ENODEV;
+		return port_set_prop(&hw->port[port_id], prop);
+	}
+
+	return -ENOENT;
+}
+
+int ifpga_set_irq(struct ifpga_hw *hw, u32 fiu_id, u32 port_id,
+		  u32 feature_id, void *irq_set)
+{
+	if (!hw || !irq_set)
+		return -EINVAL;
+
+	switch (fiu_id) {
+	case FEATURE_FIU_ID_FME:
+		return fme_set_irq(&hw->fme, feature_id, irq_set);
+	case FEATURE_FIU_ID_PORT:
+		if (!is_valid_port_id(hw, port_id))
+			return -ENODEV;
+		return port_set_irq(&hw->port[port_id], feature_id, irq_set);
+	}
+
+	return -ENOENT;
+}
+
+int ifpga_port_umsg_enable(struct ifpga_hw *hw, u32 port_id, bool enable)
+{
+	struct ifpga_port_hw *port = get_port(hw, port_id);
+
+	if (!port)
+		return -ENODEV;
+
+	spinlock_lock(&port->lock);
+	if (afu_port_umsg_enable(port, enable))
+		return -ENODEV;
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+int ifpga_port_umsg_set_mode(struct ifpga_hw *hw, u32 port_id, u32 mode)
+{
+	struct ifpga_port_hw *port = get_port(hw, port_id);
+
+	if (!port)
+		return -ENODEV;
+
+	spinlock_lock(&port->lock);
+	if (afu_port_umsg_set_mode(port, mode))
+		return -ENODEV;
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+void ifpga_show_pr_error(u64 pr_error)
+{
+#if 0
+	int i = 0;
+
+	for_each_set_bit(i, &pr_error, PR_MAX_ERR_NUM)
+		printf("PR Error: %s\n", pr_err_msg[i]);
+#endif
+	printf("PR Error: 0x%lx\n", pr_error);
+}
+
+int ifpga_mmio64_read(struct ifpga_hw *hw, u32 port_id, u64 offset, u64 *value)
+{
+	u8 *addr;
+	u32 size;
+
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	addr = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].addr;
+	size = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].size;
+
+	if (offset > size)
+		return -EINVAL;
+
+	*value = readq(addr + offset);
+
+	return 0;
+}
+
+int ifpga_mmio64_write(struct ifpga_hw *hw, u32 port_id, u64 offset, u64 value)
+{
+	u8 *addr;
+	u32 size;
+
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	addr = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].addr;
+	size = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].size;
+
+	if (offset > size)
+		return -EINVAL;
+
+	writeq(value, addr + offset);
+
+	return 0;
+}
+
+int ifpga_mmio32_read(struct ifpga_hw *hw, u32 port_id, u64 offset, u32 *value)
+{
+	u8 *addr;
+	u32 size;
+
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	addr = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].addr;
+	size = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].size;
+
+	if (offset > size)
+		return -EINVAL;
+
+	*value = readl(addr + offset);
+
+	return 0;
+}
+
+int ifpga_mmio32_write(struct ifpga_hw *hw, u32 port_id, u64 offset, u32 value)
+{
+	u8 *addr;
+	u32 size;
+
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	addr = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].addr;
+	size = hw->port[port_id].sub_feature[PORT_FEATURE_ID_UAFU].size;
+
+	if (offset > size)
+		return -EINVAL;
+
+	writel(value, addr + offset);
+
+	return 0;
+}
+
+int ifpga_fme_hw_init(struct ifpga_hw *hw)
+{
+	return fme_hw_init(&hw->fme);
+}
+
+void ifpga_fme_hw_uinit(struct ifpga_hw *hw)
+{
+	fme_hw_uinit(&hw->fme);
+}
+
+void ifpga_port_hw_uinit(struct ifpga_hw *hw, int port_id)
+{
+	struct ifpga_port_hw *port;
+
+	if (!is_valid_port_id(hw, port_id))
+		return;
+
+	port = &hw->port[port_id];
+
+	if (port->state == IFPGA_PORT_ATTACHED)
+		port_hw_uinit(port);
+}
+
+int ifpga_port_hw_init(struct ifpga_hw *hw, int port_id)
+{
+	struct ifpga_port_hw *port;
+
+	if (!is_valid_port_id(hw, port_id))
+		return -ENODEV;
+
+	port = &hw->port[port_id];
+
+	if (port->state != IFPGA_PORT_ATTACHED)
+		return -EINVAL;
+
+	return port_hw_init(port);
+}
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_api.h b/drivers/raw/ifpga_rawdev/base/ifpga_api.h
new file mode 100644
index 0000000..f8bca80
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_api.h
@@ -0,0 +1,77 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _IFPGA_API_H_
+#define _IFPGA_API_H_
+
+#include "opae_hw_api.h"
+#include "ifpga_hw.h"
+
+extern struct opae_adapter_ops ifpga_adapter_ops;
+extern struct opae_manager_ops ifpga_mgr_ops;
+extern struct opae_bridge_ops ifpga_br_ops;
+extern struct opae_accelerator_ops ifpga_acc_ops;
+
+/* common APIs */
+int ifpga_enumerate(struct ifpga_hw *hw);
+int ifpga_fme_hw_init(struct ifpga_hw *hw);
+void ifpga_fme_hw_uinit(struct ifpga_hw *hw);
+int ifpga_port_hw_init(struct ifpga_hw *hw, int port_id);
+void ifpga_port_hw_uinit(struct ifpga_hw *hw, int port_id);
+int ifpga_get_prop(struct ifpga_hw *hw, u32 fiu_id, u32 port_id,
+		   struct feature_prop *prop);
+int ifpga_set_prop(struct ifpga_hw *hw, u32 fiu_id, u32 port_id,
+		   struct feature_prop *prop);
+int ifpga_set_irq(struct ifpga_hw *hw, u32 fiu_id, u32 port_id,
+		  u32 feature_id, void *irq_set);
+
+/* Port/AFU APIs */
+int ifpga_reset_port(struct ifpga_hw *hw, u32 port_id);
+int ifpga_get_afu_mmio_info(struct ifpga_hw *hw, u32 port_id,
+			struct rte_mem_resource *mem_resource,
+			u32 *num_resource);
+int ifpga_get_afu_uuid(struct ifpga_hw *hw, u32 port_id, struct uuid *uuid);
+
+int ifpga_mmio64_read(struct ifpga_hw *hw, u32 port_id, u64 offset, u64 *value);
+int ifpga_mmio64_write(struct ifpga_hw *hw, u32 port_id, u64 offset, u64 value);
+int ifpga_mmio32_read(struct ifpga_hw *hw, u32 port_id, u64 offset, u32 *value);
+int ifpga_mmio32_write(struct ifpga_hw *hw, u32 port_id, u64 offset, u32 value);
+
+/* FME APIs */
+int ifpga_pr(struct ifpga_hw *hw, u32 port_id, void *buffer, u32 size,
+	     u64 *status);
+void ifpga_show_pr_error(u64 pr_error);
+
+/* Port APIs */
+int ifpga_port_umsg_enable(struct ifpga_hw *hw, u32 port_id, bool enable);
+int ifpga_port_umsg_set_mode(struct ifpga_hw *hw, u32 port_id, u32 mode);
+
+#endif /* _IFPGA_API_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_compat.h b/drivers/raw/ifpga_rawdev/base/ifpga_compat.h
new file mode 100644
index 0000000..6670630
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_compat.h
@@ -0,0 +1,84 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _IFPGA_COMPAT_H_
+#define _IFPGA_COMPAT_H_
+
+#include "opae_osdep.h"
+
+#undef container_of
+#define container_of(ptr, type, member) ({ \
+		typeof(((type *)0)->member)(*__mptr) = (ptr); \
+		(type *)((char *)__mptr - offsetof(type, member)); })
+
+#define PAGE_SHIFT       12
+#define PAGE_SIZE        (1 << PAGE_SHIFT)
+#define PAGE_MASK        (~(PAGE_SIZE - 1))
+#define PAGE_ALIGN(addr) (((addr) + PAGE_SIZE - 1) & PAGE_MASK)
+#define ALIGN(x, a)  (((x) + (a) - 1) & ~((a) - 1))
+
+#define IS_ALIGNED(x, a)		(((x) & ((typeof(x))(a) - 1)) == 0)
+#define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
+
+#define readl(addr) opae_readl(addr)
+#define readq(addr) opae_readq(addr)
+#define writel(value, addr) opae_writel(value, addr)
+#define writeq(value, addr) opae_writeq(value, addr)
+
+#define malloc(size) opae_malloc(size)
+#define zmalloc(size) opae_zmalloc(size)
+#define free(size) opae_free(size)
+
+/*
+ * Wait register's _field to be changed to the given value (_expect's _field)
+ * by polling with given interval and timeout.
+ */
+#define fpga_wait_register_field(_field, _expect, _reg_addr, _timeout, _invl)\
+({									     \
+	int wait = 0;							     \
+	int ret = -ETIMEDOUT;						     \
+	typeof(_expect) value;						     \
+	for (; wait <= _timeout; wait += _invl) {			     \
+		value.csr = readq(_reg_addr);				     \
+		if (_expect._field == value._field) {			     \
+			ret = 0;					     \
+			break;						     \
+		}							     \
+		udelay(_invl);						     \
+	}								     \
+	ret;								     \
+})
+
+#define __maybe_unused __attribute__((__unused__))
+
+#define UNUSED(x)	(void)(x)
+
+#endif /* _IFPGA_COMPAT_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_defines.h b/drivers/raw/ifpga_rawdev/base/ifpga_defines.h
new file mode 100644
index 0000000..7fc99c2
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_defines.h
@@ -0,0 +1,1696 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _IFPGA_DEFINES_H_
+#define _IFPGA_DEFINES_H_
+
+#include "ifpga_compat.h"
+
+#define MAX_FPGA_PORT_NUM  4
+
+#define FME_FEATURE_HEADER          "fme_hdr"
+#define FME_FEATURE_THERMAL_MGMT    "fme_thermal"
+#define FME_FEATURE_POWER_MGMT      "fme_power"
+#define FME_FEATURE_GLOBAL_IPERF    "fme_iperf"
+#define FME_FEATURE_GLOBAL_ERR      "fme_error"
+#define FME_FEATURE_PR_MGMT         "fme_pr"
+#define FME_FEATURE_HSSI_ETH        "fme_hssi"
+#define FME_FEATURE_GLOBAL_DPERF    "fme_dperf"
+#define FME_FEATURE_QSPI_FLASH	    "fme_qspi_flash"
+
+#define PORT_FEATURE_HEADER         "port_hdr"
+#define PORT_FEATURE_UAFU           "port_uafu"
+#define PORT_FEATURE_ERR            "port_err"
+#define PORT_FEATURE_UMSG           "port_umsg"
+#define PORT_FEATURE_PR             "port_pr"
+#define PORT_FEATURE_UINT           "port_uint"
+#define PORT_FEATURE_STP            "port_stp"
+
+/*
+ * do not check the revision id as id may be dynamic under
+ * some cases, e.g, UAFU.
+ */
+#define SKIP_REVISION_CHECK		0xff
+
+#define FME_HEADER_REVISION		1
+#define FME_THERMAL_MGMT_REVISION	0
+#define FME_POWER_MGMT_REVISION		1
+#define FME_GLOBAL_IPERF_REVISION	1
+#define FME_GLOBAL_ERR_REVISION		1
+#define FME_PR_MGMT_REVISION		2
+#define FME_HSSI_ETH_REVISION		0
+#define FME_GLOBAL_DPERF_REVISION	0
+#define FME_QSPI_REVISION		0
+
+#define PORT_HEADER_REVISION		0
+/* UAFU's header info depends on the downloaded GBS */
+#define PORT_UAFU_REVISION		SKIP_REVISION_CHECK
+#define PORT_ERR_REVISION		1
+#define PORT_UMSG_REVISION		0
+#define PORT_UINT_REVISION		0
+#define PORT_STP_REVISION		1
+
+#define FEATURE_TYPE_AFU	0x1
+#define FEATURE_TYPE_BBB        0x2
+#define FEATURE_TYPE_PRIVATE	0x3
+#define FEATURE_TYPE_FIU	0x4
+
+#define FEATURE_FIU_ID_FME	0x0
+#define FEATURE_FIU_ID_PORT	0x1
+
+#define FEATURE_ID_HEADER	0x0
+#define FEATURE_ID_AFU		0xff
+
+enum fpga_id_type {
+	FME_ID,
+	PORT_ID,
+	FPGA_ID_MAX,
+};
+
+enum fme_feature_id {
+	FME_FEATURE_ID_HEADER = 0x0,
+
+	FME_FEATURE_ID_THERMAL_MGMT	= 0x1,
+	FME_FEATURE_ID_POWER_MGMT = 0x2,
+	FME_FEATURE_ID_GLOBAL_IPERF = 0x3,
+	FME_FEATURE_ID_GLOBAL_ERR = 0x4,
+	FME_FEATURE_ID_PR_MGMT = 0x5,
+	FME_FEATURE_ID_HSSI_ETH = 0x6,
+	FME_FEATURE_ID_GLOBAL_DPERF = 0x7,
+	FME_FEATURE_ID_QSPI_FLASH = 0x8,
+
+	/* one for fme header. */
+	FME_FEATURE_ID_MAX = 0x9,
+};
+
+enum port_feature_id {
+	PORT_FEATURE_ID_HEADER = 0x0,
+	PORT_FEATURE_ID_ERROR = 0x1,
+	PORT_FEATURE_ID_UMSG = 0x2,
+	PORT_FEATURE_ID_UINT = 0x3,
+	PORT_FEATURE_ID_STP = 0x4,
+	PORT_FEATURE_ID_UAFU = 0x5,
+	PORT_FEATURE_ID_MAX = 0x6,
+};
+
+struct uuid {
+	u8 b[16];
+};
+
+/*
+ * All headers and structures must be byte-packed to match the
+ * SAS spec.
+ */
+#pragma pack(1)
+
+struct feature_header {
+	union {
+		u64 csr;
+		struct {
+			u16 id:12;
+			u8  revision:4;
+			u32 next_header_offset:24;
+			u8  end_of_list:1;
+			u32 reserved:19;
+			u8  type:4;
+		};
+	};
+};
+
+struct feature_bbb_header {
+	struct uuid guid;
+};
+
+struct feature_afu_header {
+	struct uuid guid;
+	union {
+		u64 csr;
+		struct {
+			u64 next_afu:24;
+			u64 reserved:40;
+		};
+	};
+};
+
+struct feature_fiu_header {
+	struct uuid guid;
+	union {
+		u64 csr;
+		struct {
+			u64 next_afu:24;
+			u64 reserved:40;
+		};
+	};
+};
+
+struct feature_fme_capability {
+	union {
+		u64 csr;
+		struct {
+			u8  fabric_verid;	/* Fabric version ID */
+			u8  socket_id:1;	/* Socket id */
+			u8  rsvd1:3;		/* Reserved */
+			/* pci0 link available yes /no */
+			u8  pci0_link_avile:1;
+			/* pci1 link available yes /no */
+			u8  pci1_link_avile:1;
+			/* Coherent (QPI/UPI) link available yes /no */
+			u8  qpi_link_avile:1;
+			u8  rsvd2:1;		/* Reserved */
+			/* IOMMU or VT-d supported  yes/no */
+			u8  iommu_support:1;
+			u8  num_ports:3;	/* Number of ports */
+			u8  sf_fab_ctl:1;	/* Internal validation bit */
+			u8  rsvd3:3;		/* Reserved */
+			/*
+			 * Address width supported in bits
+			 * BXT -0x26 , SKX -0x30
+			 */
+			u8  address_width_bits:6;
+			u8  rsvd4:2;		/* Reserved */
+			/* Size of cache supported in kb */
+			u16 cache_size:12;
+			u8  cache_assoc:4;	/* Cache Associativity */
+			u16 rsvd5:15;		/* Reserved */
+			u8  lock_bit:1;		/* Lock bit */
+		};
+	};
+};
+
+#define FME_AFU_ACCESS_PF		0
+#define FME_AFU_ACCESS_VF		1
+
+struct feature_fme_port {
+	union {
+		u64 csr;
+		struct {
+			u32 port_offset:24;
+			u8  reserved1;
+			u8  port_bar:3;
+			u32 reserved2:20;
+			u8  afu_access_control:1;
+			u8  reserved3:4;
+			u8  port_implemented:1;
+			u8  reserved4:3;
+		};
+	};
+};
+
+struct feature_fme_fab_status {
+	union {
+		u64 csr;
+		struct {
+			u8  upilink_status:4;   /* UPI Link Status */
+			u8  rsvd1:4;		/* Reserved */
+			u8  pci0link_status:1;  /* pci0 link status */
+			u8  rsvd2:3;            /* Reserved */
+			u8  pci1link_status:1;  /* pci1 link status */
+			u64 rsvd3:51;           /* Reserved */
+		};
+	};
+};
+
+struct feature_fme_genprotrange2_base {
+	union {
+		u64 csr;
+		struct {
+			u16 rsvd1;           /* Reserved */
+			/* Base Address of memory range */
+			u8  protected_base_addrss:4;
+			u64 rsvd2:44;           /* Reserved */
+		};
+	};
+};
+
+struct feature_fme_genprotrange2_limit {
+	union {
+		u64 csr;
+		struct {
+			u16 rsvd1;           /* Reserved */
+			/* Limit Address of memory range */
+			u8  protected_limit_addrss:4;
+			u16 rsvd2:11;           /* Reserved */
+			u8  enable:1;        /* Enable GENPROTRANGE check */
+			u32 rsvd3;           /* Reserved */
+		};
+	};
+};
+
+struct feature_fme_dxe_lock {
+	union {
+		u64 csr;
+		struct {
+			/*
+			 * Determines write access to the DXE region CSRs
+			 * 1 - CSR region is locked;
+			 * 0 - it is open for write access.
+			 */
+			u8  dxe_early_lock:1;
+			/*
+			 * Determines write access to the HSSI CSR
+			 * 1 - CSR region is locked;
+			 * 0 - it is open for write access.
+			 */
+			u8  dxe_late_lock:1;
+			u64 rsvd:62;
+		};
+	};
+};
+
+#define HSSI_ID_NO_HASSI	0
+#define HSSI_ID_PCIE_RP		1
+#define HSSI_ID_ETHERNET	2
+
+struct feature_fme_bitstream_id {
+	union {
+		u64 csr;
+		struct {
+			u32 gitrepo_hash:32;	/* GIT repository hash */
+			/*
+			 * HSSI configuration identifier:
+			 * 0 - No HSSI
+			 * 1 - PCIe-RP
+			 * 2 - Ethernet
+			 */
+			u8  hssi_id:4;
+			u16 rsvd1:12;		/* Reserved */
+			/* Bitstream version patch number */
+			u8  bs_verpatch:4;
+			/* Bitstream version minor number */
+			u8  bs_verminor:4;
+			/* Bitstream version major number */
+			u8  bs_vermajor:4;
+			/* Bitstream version debug number */
+			u8  bs_verdebug:4;
+		};
+	};
+};
+
+struct feature_fme_bitstream_md {
+	union {
+		u64 csr;
+		struct {
+			/* Seed number userd for synthesis flow */
+			u8  synth_seed:4;
+			/* Synthesis date(day number - 2 digits) */
+			u8  synth_day:8;
+			/* Synthesis date(month number - 2 digits) */
+			u8  synth_month:8;
+			/* Synthesis date(year number - 2 digits) */
+			u8  synth_year:8;
+			u64 rsvd:36;		/* Reserved */
+		};
+	};
+};
+
+struct feature_fme_iommu_ctrl {
+	union {
+		u64 csr;
+		struct {
+			/* Disables IOMMU prefetcher for C0 channel */
+			u8 prefetch_disableC0:1;
+			/* Disables IOMMU prefetcher for C1 channel */
+			u8 prefetch_disableC1:1;
+			/* Disables IOMMU partial cache line writes */
+			u8 prefetch_wrdisable:1;
+			u8 rsvd1:1;		/* Reserved */
+			/*
+			 * Select counter and read value from register
+			 * iommu_stat.dbg_counters
+			 * 0 - Number of 4K page translation response
+			 * 1 - Number of 2M page translation response
+			 * 2 - Number of 1G page translation response
+			 */
+			u8 counter_sel:2;
+			u32 rsvd2:26;		/* Reserved */
+			/* Connected to IOMMU SIP Capabilities */
+			u32 capecap_defeature;
+		};
+	};
+};
+
+struct feature_fme_iommu_stat {
+	union {
+		u64 csr;
+		struct {
+			/* Translation Enable bit from IOMMU SIP */
+			u8 translation_enable:1;
+			/* Drain request in progress */
+			u8 drain_req_inprog:1;
+			/* Invalidation current state */
+			u8 inv_state:3;
+			/* C0 Response Buffer current state */
+			u8 respbuffer_stateC0:3;
+			/* C1 Response Buffer current state */
+			u8 respbuffer_stateC1:3;
+			/* Last request ID to IOMMU SIP */
+			u8 last_reqID:4;
+			/* Last IOMMU SIP response ID value */
+			u8 last_respID:4;
+			/* Last IOMMU SIP response status value */
+			u8 last_respstatus:3;
+			/* C0 Transaction Buffer is not empty */
+			u8 transbuf_notEmptyC0:1;
+			/* C1 Transaction Buffer is not empty */
+			u8 transbuf_notEmptyC1:1;
+			/* C0 Request FIFO is not empty */
+			u8 reqFIFO_notemptyC0:1;
+			/* C1 Request FIFO is not empty */
+			u8 reqFIFO_notemptyC1:1;
+			/* C0 Response FIFO is not empty */
+			u8 respFIFO_notemptyC0:1;
+			/* C1 Response FIFO is not empty */
+			u8 respFIFO_notemptyC1:1;
+			/* C0 Response FIFO overflow detected */
+			u8 respFIFO_overflowC0:1;
+			/* C1 Response FIFO overflow detected */
+			u8 respFIFO_overflowC1:1;
+			/* C0 Transaction Buffer overflow detected */
+			u8 tranbuf_overflowC0:1;
+			/* C1 Transaction Buffer overflow detected */
+			u8 tranbuf_overflowC1:1;
+			/* Request FIFO overflow detected */
+			u8 reqFIFO_overflow:1;
+			/* IOMMU memory read in progress */
+			u8 memrd_inprog:1;
+			/* IOMMU memory write in progress */
+			u8 memwr_inprog:1;
+			u8 rsvd1:1;	/* Reserved */
+			/* Value of counter selected by iommu_ctl.counter_sel */
+			u16 dbg_counters:16;
+			u16 rsvd2:12;	/* Reserved */
+		};
+	};
+};
+
+struct feature_fme_pcie0_ctrl {
+	union {
+		u64 csr;
+		struct {
+			u64 vtd_bar_lock:1;	/* Lock VT-D BAR register */
+			u64 rsvd1:3;
+			u64 rciep:1;		/* Configure PCIE0 as RCiEP */
+			u64 rsvd2:59;
+		};
+	};
+};
+
+struct feature_fme_llpr_smrr_base {
+	union {
+		u64 csr;
+		struct {
+			u64 rsvd1:12;
+			u64 base:20;	/* SMRR2 memory range base address */
+			u64 rsvd2:32;
+		};
+	};
+};
+
+struct feature_fme_llpr_smrr_mask {
+	union {
+		u64 csr;
+		struct {
+			u64 rsvd1:11;
+			u64 valid:1;	/* LLPR_SMRR rule is valid or not */
+			/*
+			 * SMRR memory range mask which determines the range
+			 * of region being mapped
+			 */
+			u64 phys_mask:20;
+			u64 rsvd2:32;
+		};
+	};
+};
+
+struct feature_fme_llpr_smrr2_base {
+	union {
+		u64 csr;
+		struct {
+			u64 rsvd1:12;
+			u64 base:20;	/* SMRR2 memory range base address */
+			u64 rsvd2:32;
+		};
+	};
+};
+
+struct feature_fme_llpr_smrr2_mask {
+	union {
+		u64 csr;
+		struct {
+			u64 rsvd1:11;
+			u64 valid:1;	/* LLPR_SMRR2 rule is valid or not */
+			/*
+			 * SMRR2 memory range mask which determines the range
+			 * of region being mapped
+			 */
+			u64 phys_mask:20;
+			u64 rsvd2:32;
+		};
+	};
+};
+
+struct feature_fme_llpr_meseg_base {
+	union {
+		u64 csr;
+		struct {
+			/* A[45:19] of base address memory range */
+			u64 me_base:27;
+			u64 rsvd:37;
+		};
+	};
+};
+
+struct feature_fme_llpr_meseg_limit {
+	union {
+		u64 csr;
+		struct {
+			/* A[45:19] of limit address memory range */
+			u64 me_limit:27;
+			u64 rsvd1:4;
+			u64 enable:1;	/* Enable LLPR MESEG rule */
+			u64 rsvd2:32;
+		};
+	};
+};
+
+struct feature_fme_header {
+	struct feature_header header;
+	struct feature_afu_header afu_header;
+	u64 reserved;
+	u64 scratchpad;
+	struct feature_fme_capability capability;
+	struct feature_fme_port port[MAX_FPGA_PORT_NUM];
+	struct feature_fme_fab_status fab_status;
+	struct feature_fme_bitstream_id bitstream_id;
+	struct feature_fme_bitstream_md bitstream_md;
+	struct feature_fme_genprotrange2_base genprotrange2_base;
+	struct feature_fme_genprotrange2_limit genprotrange2_limit;
+	struct feature_fme_dxe_lock dxe_lock;
+	struct feature_fme_iommu_ctrl iommu_ctrl;
+	struct feature_fme_iommu_stat iommu_stat;
+	struct feature_fme_pcie0_ctrl pcie0_control;
+	struct feature_fme_llpr_smrr_base smrr_base;
+	struct feature_fme_llpr_smrr_mask smrr_mask;
+	struct feature_fme_llpr_smrr2_base smrr2_base;
+	struct feature_fme_llpr_smrr2_mask smrr2_mask;
+	struct feature_fme_llpr_meseg_base meseg_base;
+	struct feature_fme_llpr_meseg_limit meseg_limit;
+};
+
+struct feature_port_capability {
+	union {
+		u64 csr;
+		struct {
+			u8 port_number:2;	/* Port Number 0-3 */
+			u8 rsvd1:6;		/* Reserved */
+			u16 mmio_size;		/* User MMIO size in KB */
+			u8 rsvd2;		/* Reserved */
+			u8 sp_intr_num:4;	/* Supported interrupts num */
+			u32 rsvd3:28;		/* Reserved */
+		};
+	};
+};
+
+struct feature_port_control {
+	union {
+		u64 csr;
+		struct {
+			u8 port_sftrst:1;	/* Port Soft Reset */
+			u8 rsvd1:1;		/* Reserved */
+			u8 latency_tolerance:1;/* '1' >= 40us, '0' < 40us */
+			u8 rsvd2:1;		/* Reserved */
+			u8 port_sftrst_ack:1;	/* HW ACK for Soft Reset */
+			u64 rsvd3:59;		/* Reserved */
+		};
+	};
+};
+
+#define PORT_POWER_STATE_NORMAL		0
+#define PORT_POWER_STATE_AP1		1
+#define PORT_POWER_STATE_AP2		2
+#define PORT_POWER_STATE_AP6		6
+
+struct feature_port_status {
+	union {
+		u64 csr;
+		struct {
+			u8 port_freeze:1;	/* '1' - freezed '0' - normal */
+			u8 rsvd1:7;		/* Reserved */
+			u8 power_state:4;	/* Power State */
+			u8 ap1_event:1;		/* AP1 event was detected  */
+			u8 ap2_event:1;		/* AP2 event was detected  */
+			u64 rsvd2:50;		/* Reserved */
+		};
+	};
+};
+
+/* Port Header Register Set */
+struct feature_port_header {
+	struct feature_header header;
+	struct feature_afu_header afu_header;
+	u64 port_mailbox;
+	u64 scratchpad;
+	struct feature_port_capability capability;
+	struct feature_port_control control;
+	struct feature_port_status status;
+	u64 rsvd2;
+	u64 user_clk_freq_cmd0;
+	u64 user_clk_freq_cmd1;
+	u64 user_clk_freq_sts0;
+	u64 user_clk_freq_sts1;
+};
+
+struct feature_fme_tmp_threshold {
+	union {
+		u64 csr;
+		struct {
+			u8  tmp_thshold1:7;	  /* temperature Threshold 1 */
+			/* temperature Threshold 1 enable/disable */
+			u8  tmp_thshold1_enable:1;
+			u8  tmp_thshold2:7;       /* temperature Threshold 2 */
+			/* temperature Threshold 2 enable /disable */
+			u8  tmp_thshold2_enable:1;
+			u8  pro_hot_setpoint:7;   /* Proc Hot set point */
+			u8  rsvd4:1;              /* Reserved */
+			u8  therm_trip_thshold:7; /* Thermeal Trip Threshold */
+			u8  rsvd3:1;              /* Reserved */
+			u8  thshold1_status:1;	  /* Threshold 1 Status */
+			u8  thshold2_status:1;    /* Threshold 2 Status */
+			u8  rsvd5:1;              /* Reserved */
+			/* Thermeal Trip Threshold status */
+			u8  therm_trip_thshold_status:1;
+			u8  rsvd6:4;		  /* Reserved */
+			/* Validation mode- Force Proc Hot */
+			u8  valmodeforce:1;
+			/* Validation mode - Therm trip Hot */
+			u8  valmodetherm:1;
+			u8  rsvd2:2;              /* Reserved */
+			u8  thshold_policy:1;     /* threshold policy */
+			u32 rsvd:19;              /* Reserved */
+		};
+	};
+};
+
+/* Temperature Sensor Read values format 1 */
+struct feature_fme_temp_rdsensor_fmt1 {
+	union {
+		u64 csr;
+		struct {
+			/* Reads out FPGA temperature in celsius */
+			u8  fpga_temp:7;
+			u8  rsvd0:1;			/* Reserved */
+			/* Temperature reading sequence number */
+			u16 tmp_reading_seq_num;
+			/* Temperature reading is valid */
+			u8  tmp_reading_valid:1;
+			u8  rsvd1:7;			/* Reserved */
+			u16 dbg_mode:10;		/* Debug mode */
+			u32 rsvd2:22;			/* Reserved */
+		};
+	};
+};
+
+/* Temperature sensor read values format 2 */
+struct feature_fme_temp_rdsensor_fmt2 {
+	u64 rsvd;	/* Reserved */
+};
+
+/* Temperature Threshold Capability Register */
+struct feature_fme_tmp_threshold_cap {
+	union {
+		u64 csr;
+		struct {
+			/* Temperature Threshold Unsupported */
+			u8  tmp_thshold_disabled:1;
+			u64 rsvd:63;			/* Reserved */
+		};
+	};
+};
+
+/* FME THERNAL FEATURE */
+struct feature_fme_thermal {
+	struct feature_header header;
+	struct feature_fme_tmp_threshold threshold;
+	struct feature_fme_temp_rdsensor_fmt1 rdsensor_fm1;
+	struct feature_fme_temp_rdsensor_fmt2 rdsensor_fm2;
+	struct feature_fme_tmp_threshold_cap threshold_cap;
+};
+
+/* Power Status register */
+struct feature_fme_pm_status {
+	union {
+		u64 csr;
+		struct {
+			/* FPGA Power consumed, The format is to be defined */
+			u32 pwr_consumed:18;
+			/* FPGA Latency Tolerance Reporting */
+			u8  fpga_latency_report:1;
+			u64 rsvd:45;			/* Reserved */
+		};
+	};
+};
+
+/* AP Thresholds */
+struct feature_fme_pm_ap_threshold {
+	union {
+		u64 csr;
+		struct {
+			/*
+			 * Number of clocks (5ns period) for assertion
+			 * of FME_data
+			 */
+			u8  threshold1:7;
+			u8  rsvd1:1;
+			u8  threshold2:7;
+			u8  rsvd2:1;
+			u8  threshold1_status:1;
+			u8  threshold2_status:1;
+			u64 rsvd3:46;		/* Reserved */
+		};
+	};
+};
+
+/* Xeon Power Limit */
+struct feature_fme_pm_xeon_limit {
+	union {
+		u64 csr;
+		struct {
+			/* Power limit in Watts in 12.3 format */
+			u16 pwr_limit:15;
+			/* Indicates that power limit has been written */
+			u8  enable:1;
+			/* 0 - Turbe range, 1 - Entire range */
+			u8  clamping:1;
+			/* Time constant in XXYYY format */
+			u8  time:7;
+			u64 rsvd:40;		/* Reserved */
+		};
+	};
+};
+
+/* MCP Power Limit */
+struct feature_fme_pm_fpga_limit {
+	union {
+		u64 csr;
+		struct {
+			/* Power limit in Watts in 12.3 format */
+			u16 pwr_limit:15;
+			/* Indicates that power limit has been written */
+			u8  enable:1;
+			/* 0 - Turbe range, 1 - Entire range */
+			u8  clamping:1;
+			/* Time constant in XXYYY format */
+			u8  time:7;
+			u64 rsvd:40;		/* Reserved */
+		};
+	};
+};
+
+/* FME POWER FEATURE */
+struct feature_fme_power {
+	struct feature_header header;
+	struct feature_fme_pm_status status;
+	struct feature_fme_pm_ap_threshold threshold;
+	struct feature_fme_pm_xeon_limit xeon_limit;
+	struct feature_fme_pm_fpga_limit fpga_limit;
+};
+
+#define CACHE_CHANNEL_RD	0
+#define CACHE_CHANNEL_WR	1
+
+enum iperf_cache_events {
+	IPERF_CACHE_RD_HIT,
+	IPERF_CACHE_WR_HIT,
+	IPERF_CACHE_RD_MISS,
+	IPERF_CACHE_WR_MISS,
+	IPERF_CACHE_RSVD, /* reserved */
+	IPERF_CACHE_HOLD_REQ,
+	IPERF_CACHE_DATA_WR_PORT_CONTEN,
+	IPERF_CACHE_TAG_WR_PORT_CONTEN,
+	IPERF_CACHE_TX_REQ_STALL,
+	IPERF_CACHE_RX_REQ_STALL,
+	IPERF_CACHE_EVICTIONS,
+};
+
+/* FPMON Cache Control */
+struct feature_fme_ifpmon_ch_ctl {
+	union {
+		u64 csr;
+		struct {
+			u8  reset_counters:1;	/* Reset Counters */
+			u8  rsvd1:7;		/* Reserved */
+			u8  freeze:1;		/* Freeze if set to 1 */
+			u8  rsvd2:7;		/* Reserved */
+			u8  cache_event:4;	/* Select the cache event */
+			u8  cci_chsel:1;	/* Select the channel */
+			u64 rsvd3:43;		/* Reserved */
+		};
+	};
+};
+
+/* FPMON Cache Counter */
+struct feature_fme_ifpmon_ch_ctr {
+	union {
+		u64 csr;
+		struct {
+			/* Cache Counter for even addresse */
+			u64 cache_counter:48;
+			u16 rsvd:12;		/* Reserved */
+			/* Cache Event being reported */
+			u8  event_code:4;
+		};
+	};
+};
+
+enum iperf_fab_events {
+	IPERF_FAB_PCIE0_RD,
+	IPERF_FAB_PCIE0_WR,
+	IPERF_FAB_PCIE1_RD,
+	IPERF_FAB_PCIE1_WR,
+	IPERF_FAB_UPI_RD,
+	IPERF_FAB_UPI_WR,
+	IPERF_FAB_MMIO_RD,
+	IPERF_FAB_MMIO_WR,
+};
+
+#define FAB_DISABLE_FILTER     0
+#define FAB_ENABLE_FILTER      1
+
+/* FPMON FAB Control */
+struct feature_fme_ifpmon_fab_ctl {
+	union {
+		u64 csr;
+		struct {
+			u8  reset_counters:1;	/* Reset Counters */
+			u8  rsvd:7;		/* Reserved */
+			u8  freeze:1;		/* Set to 1 frozen counter */
+			u8  rsvd1:7;		/* Reserved */
+			u8  fab_evtcode:4;	/* Fabric Event Code */
+			u8  port_id:2;		/* Port ID */
+			u8  rsvd2:1;		/* Reserved */
+			u8  port_filter:1;	/* Port Filter */
+			u64 rsvd3:40;		/* Reserved */
+		};
+	};
+};
+
+/* FPMON Event Counter */
+struct feature_fme_ifpmon_fab_ctr {
+	union {
+		u64 csr;
+		struct {
+			u64 fab_cnt:60;	/* Fabric event counter */
+			/* Fabric event code being reported */
+			u8  event_code:4;
+		};
+	};
+};
+
+/* FPMON Clock Counter */
+struct feature_fme_ifpmon_clk_ctr {
+	u64 afu_interf_clock;		/* Clk_16UI (AFU clock) counter. */
+};
+
+enum iperf_vtd_events {
+	IPERF_VTD_AFU_MEM_RD_TRANS,
+	IPERF_VTD_AFU_MEM_WR_TRANS,
+	IPERF_VTD_AFU_DEVTLB_RD_HIT,
+	IPERF_VTD_AFU_DEVTLB_WR_HIT,
+	IPERF_VTD_DEVTLB_4K_FILL,
+	IPERF_VTD_DEVTLB_2M_FILL,
+	IPERF_VTD_DEVTLB_1G_FILL,
+};
+
+/* VT-d control register */
+struct feature_fme_ifpmon_vtd_ctl {
+	union {
+		u64 csr;
+		struct {
+			u8  reset_counters:1;	/* Reset Counters */
+			u8  rsvd:7;		/* Reserved */
+			u8  freeze:1;		/* Set to 1 frozen counter */
+			u8  rsvd1:7;		/* Reserved */
+			u8  vtd_evtcode:4;	/* VTd and TLB event code */
+			u64 rsvd2:44;		/* Reserved */
+		};
+	};
+};
+
+/* VT-d event counter */
+struct feature_fme_ifpmon_vtd_ctr {
+	union {
+		u64 csr;
+		struct {
+			u64 vtd_counter:48;	/* VTd event counter */
+			u16 rsvd:12;		/* Reserved */
+			u8  event_code:4;	/* VTd event code */
+		};
+	};
+};
+
+enum iperf_vtd_sip_events {
+	IPERF_VTD_SIP_IOTLB_4K_HIT,
+	IPERF_VTD_SIP_IOTLB_2M_HIT,
+	IPERF_VTD_SIP_IOTLB_1G_HIT,
+	IPERF_VTD_SIP_SLPWC_L3_HIT,
+	IPERF_VTD_SIP_SLPWC_L4_HIT,
+	IPERF_VTD_SIP_RCC_HIT,
+	IPERF_VTD_SIP_IOTLB_4K_MISS,
+	IPERF_VTD_SIP_IOTLB_2M_MISS,
+	IPERF_VTD_SIP_IOTLB_1G_MISS,
+	IPERF_VTD_SIP_SLPWC_L3_MISS,
+	IPERF_VTD_SIP_SLPWC_L4_MISS,
+	IPERF_VTD_SIP_RCC_MISS,
+};
+
+/* VT-d SIP control register */
+struct feature_fme_ifpmon_vtd_sip_ctl {
+	union {
+		u64 csr;
+		struct {
+			u8  reset_counters:1;	/* Reset Counters */
+			u8  rsvd:7;		/* Reserved */
+			u8  freeze:1;		/* Set to 1 frozen counter */
+			u8  rsvd1:7;		/* Reserved */
+			u8  vtd_evtcode:4;	/* VTd and TLB event code */
+			u64 rsvd2:44;		/* Reserved */
+		};
+	};
+};
+
+/* VT-d SIP event counter */
+struct feature_fme_ifpmon_vtd_sip_ctr {
+	union {
+		u64 csr;
+		struct {
+			u64 vtd_counter:48;	/* VTd event counter */
+			u16 rsvd:12;		/* Reserved */
+			u8 event_code:4;	/* VTd event code */
+		};
+	};
+};
+
+/* FME IPERF FEATURE */
+struct feature_fme_iperf {
+	struct feature_header header;
+	struct feature_fme_ifpmon_ch_ctl ch_ctl;
+	struct feature_fme_ifpmon_ch_ctr ch_ctr0;
+	struct feature_fme_ifpmon_ch_ctr ch_ctr1;
+	struct feature_fme_ifpmon_fab_ctl fab_ctl;
+	struct feature_fme_ifpmon_fab_ctr fab_ctr;
+	struct feature_fme_ifpmon_clk_ctr clk;
+	struct feature_fme_ifpmon_vtd_ctl vtd_ctl;
+	struct feature_fme_ifpmon_vtd_ctr vtd_ctr;
+	struct feature_fme_ifpmon_vtd_sip_ctl vtd_sip_ctl;
+	struct feature_fme_ifpmon_vtd_sip_ctr vtd_sip_ctr;
+};
+
+enum dperf_fab_events {
+	DPERF_FAB_PCIE0_RD,
+	DPERF_FAB_PCIE0_WR,
+	DPERF_FAB_MMIO_RD = 6,
+	DPERF_FAB_MMIO_WR,
+};
+
+#define DCP_FAB_DISABLE_FILTER     0
+#define DCP_FAB_ENABLE_FILTER      1
+
+/* FPMON FAB Control */
+struct feature_fme_dfpmon_fab_ctl {
+	union {
+		u64 csr;
+		struct {
+			u8  reset_counters:1;	/* Reset Counters */
+			u8  rsvd:7;		/* Reserved */
+			u8  freeze:1;		/* Set to 1 frozen counter */
+			u8  rsvd1:7;		/* Reserved */
+			u8  fab_evtcode:4;	/* Fabric Event Code */
+			u8  port_id:2;		/* Port ID */
+			u8  rsvd2:1;		/* Reserved */
+			u8  port_filter:1;	/* Port Filter */
+			u64 rsvd3:40;		/* Reserved */
+		};
+	};
+};
+
+/* FPMON Event Counter */
+struct feature_fme_dfpmon_fab_ctr {
+	union {
+		u64 csr;
+		struct {
+			u64 fab_cnt:60;	/* Fabric event counter */
+			/* Fabric event code being reported */
+			u8  event_code:4;
+		};
+	};
+};
+
+/* FPMON Clock Counter */
+struct feature_fme_dfpmon_clk_ctr {
+	u64 afu_interf_clock;		/* Clk_16UI (AFU clock) counter. */
+};
+
+/* FME DPERF FEATURE */
+struct feature_fme_dperf {
+	struct feature_header header;
+	u64 rsvd[3];
+	struct feature_fme_dfpmon_fab_ctl fab_ctl;
+	struct feature_fme_dfpmon_fab_ctr fab_ctr;
+	struct feature_fme_dfpmon_clk_ctr clk;
+};
+
+struct feature_fme_error0 {
+#define FME_ERROR0_MASK        0xFFUL
+#define FME_ERROR0_MASK_DEFAULT 0x40UL  /* pcode workaround */
+	union {
+		u64 csr;
+		struct {
+			u8  fabric_err:1;	/* Fabric error */
+			u8  fabfifo_overflow:1;	/* Fabric fifo overflow */
+			u8  kticdc_parity_err:2;/* KTI CDC Parity Error */
+			u8  iommu_parity_err:1;	/* IOMMU Parity error */
+			/* AFU PF/VF access mismatch detected */
+			u8  afu_acc_mode_err:1;
+			u8  mbp_err:1;		/* Indicates an MBP event */
+			/* PCIE0 CDC Parity Error */
+			u8  pcie0cdc_parity_err:5;
+			/* PCIE1 CDC Parity Error */
+			u8  pcie1cdc_parity_err:5;
+			/* CVL CDC Parity Error */
+			u8  cvlcdc_parity_err:3;
+			u64 rsvd:44;		/* Reserved */
+		};
+	};
+};
+
+/* PCIe0 Error Status register */
+struct feature_fme_pcie0_error {
+#define FME_PCIE0_ERROR_MASK   0xFFUL
+	union {
+		u64 csr;
+		struct {
+			u8  formattype_err:1;	/* TLP format/type error */
+			u8  MWAddr_err:1;	/* TLP MW address error */
+			u8  MWAddrLength_err:1;	/* TLP MW length error */
+			u8  MRAddr_err:1;	/* TLP MR address error */
+			u8  MRAddrLength_err:1;	/* TLP MR length error */
+			u8  cpl_tag_err:1;	/* TLP CPL tag error */
+			u8  cpl_status_err:1;	/* TLP CPL status error */
+			u8  cpl_timeout_err:1;	/* TLP CPL timeout */
+			u8  cci_parity_err:1;	/* CCI bridge parity error */
+			u8  rxpoison_tlp_err:1;	/* Received a TLP with EP set */
+			u64 rsvd:52;		/* Reserved */
+			u8  vfnumb_err:1;	/* Number of error VF */
+			u8  funct_type_err:1;	/* Virtual (1) or Physical */
+		};
+	};
+};
+
+/* PCIe1 Error Status register */
+struct feature_fme_pcie1_error {
+#define FME_PCIE1_ERROR_MASK   0xFFUL
+	union {
+		u64 csr;
+		struct {
+			u8  formattype_err:1;	/* TLP format/type error */
+			u8  MWAddr_err:1;	/* TLP MW address error */
+			u8  MWAddrLength_err:1;	/* TLP MW length error */
+			u8  MRAddr_err:1;	/* TLP MR address error */
+			u8  MRAddrLength_err:1;	/* TLP MR length error */
+			u8  cpl_tag_err:1;	/* TLP CPL tag error */
+			u8  cpl_status_err:1;	/* TLP CPL status error */
+			u8  cpl_timeout_err:1;	/* TLP CPL timeout */
+			u8  cci_parity_err:1;	/* CCI bridge parity error */
+			u8  rxpoison_tlp_err:1;	/* Received a TLP with EP set */
+			u64 rsvd:54;		/* Reserved */
+		};
+	};
+};
+
+/* FME First Error register */
+struct feature_fme_first_error {
+#define FME_FIRST_ERROR_MASK   ((1UL << 60) - 1)
+	union {
+		u64 csr;
+		struct {
+			/*
+			 * Indicates the Error Register that was
+			 * triggered first
+			 */
+			u64 err_reg_status:60;
+			/*
+			 * Holds 60 LSBs from the Error register that was
+			 * triggered first
+			 */
+			u8 errReg_id:4;
+		};
+	};
+};
+
+/* FME Next Error register */
+struct feature_fme_next_error {
+#define FME_NEXT_ERROR_MASK    ((1UL << 60) - 1)
+	union {
+		u64 csr;
+		struct {
+			/*
+			 * Indicates the Error Register that was
+			 * triggered second
+			 */
+			u64 err_reg_status:60;
+			/*
+			 * Holds 60 LSBs from the Error register that was
+			 * triggered second
+			 */
+			u8  errReg_id:4;
+		};
+	};
+};
+
+/* RAS Non Fatal Error Status register */
+struct feature_fme_ras_nonfaterror {
+	union {
+		u64 csr;
+		struct {
+			/* thremal threshold AP1 */
+			u8  temp_thresh_ap1:1;
+			/* thremal threshold AP2 */
+			u8  temp_thresh_ap2:1;
+			u8  pcie_error:1;	/* pcie Error */
+			u8  portfatal_error:1;	/* port fatal error */
+			u8  proc_hot:1;		/* Indicates a ProcHot event */
+			/* Indicates an AFU PF/VF access mismatch */
+			u8  afu_acc_mode_err:1;
+			/* Injected nonfata Error */
+			u8  injected_nonfata_err:1;
+			u8  rsvd1:2;
+			/* Temperature threshold triggered AP6*/
+			u8  temp_thresh_AP6:1;
+			/* Power threshold triggered AP1 */
+			u8  power_thresh_AP1:1;
+			/* Power threshold triggered AP2 */
+			u8  power_thresh_AP2:1;
+			/* Indicates a MBP event */
+			u8  mbp_err:1;
+			u64 rsvd2:51;		/* Reserved */
+		};
+	};
+};
+
+/* RAS Catastrophic Fatal Error Status register */
+struct feature_fme_ras_catfaterror {
+	union {
+		u64 csr;
+		struct {
+			/* KTI Link layer error detected */
+			u8  ktilink_fatal_err:1;
+			/* tag-n-cache error detected */
+			u8  tagcch_fatal_err:1;
+			/* CCI error detected */
+			u8  cci_fatal_err:1;
+			/* KTI Protocol error detected */
+			u8  ktiprpto_fatal_err:1;
+			/* Fatal DRAM error detected */
+			u8  dram_fatal_err:1;
+			/* IOMMU detected */
+			u8  iommu_fatal_err:1;
+			/* Fabric Fatal Error */
+			u8  fabric_fatal_err:1;
+			/* PCIe possion Error */
+			u8  pcie_poison_err:1;
+			/* Injected fatal Error */
+			u8  inject_fata_err:1;
+			/* Catastrophic CRC Error */
+			u8  crc_catast_err:1;
+			/* Catastrophic Thermal Error */
+			u8  therm_catast_err:1;
+			/* Injected Catastrophic Error */
+			u8  injected_catast_err:1;
+			u64 rsvd:52;
+		};
+	};
+};
+
+/* RAS Error injection register */
+struct feature_fme_ras_error_inj {
+#define FME_RAS_ERROR_INJ_MASK      0x7UL
+	union {
+		u64 csr;
+		struct {
+			u8  catast_error:1;	/* Catastrophic error flag */
+			u8  fatal_error:1;	/* Fatal error flag */
+			u8  nonfatal_error:1;	/* NonFatal error flag */
+			u64 rsvd:61;		/* Reserved */
+		};
+	};
+};
+
+/* FME error capabilities */
+struct feature_fme_error_capability {
+	union {
+	u64 csr;
+		struct {
+			u8 support_intr:1;
+			/* MSI-X vector table entry number */
+			u16 intr_vector_num:12;
+			u64 rsvd:51;	/* Reserved */
+		};
+	};
+};
+
+/* FME ERR FEATURE */
+struct feature_fme_err {
+	struct feature_header header;
+	struct feature_fme_error0 fme_err_mask;
+	struct feature_fme_error0 fme_err;
+	struct feature_fme_pcie0_error pcie0_err_mask;
+	struct feature_fme_pcie0_error pcie0_err;
+	struct feature_fme_pcie1_error pcie1_err_mask;
+	struct feature_fme_pcie1_error pcie1_err;
+	struct feature_fme_first_error fme_first_err;
+	struct feature_fme_next_error fme_next_err;
+	struct feature_fme_ras_nonfaterror ras_nonfat_mask;
+	struct feature_fme_ras_nonfaterror ras_nonfaterr;
+	struct feature_fme_ras_catfaterror ras_catfat_mask;
+	struct feature_fme_ras_catfaterror ras_catfaterr;
+	struct feature_fme_ras_error_inj ras_error_inj;
+	struct feature_fme_error_capability fme_err_capability;
+};
+
+/* FME Partial Reconfiguration Control */
+struct feature_fme_pr_ctl {
+	union {
+		u64 csr;
+		struct {
+			u8  pr_reset:1;		/* Reset PR Engine */
+			u8  rsvd3:3;		/* Reserved */
+			u8  pr_reset_ack:1;	/* Reset PR Engine Ack */
+			u8  rsvd4:3;		/* Reserved */
+			u8  pr_regionid:2;	/* PR Region ID */
+			u8  rsvd1:2;		/* Reserved */
+			u8  pr_start_req:1;	/* PR Start Request */
+			u8  pr_push_complete:1;	/* PR Data push complete */
+			u8  pr_kind:1;		/* PR Data push complete */
+			u32 rsvd:17;		/* Reserved */
+			u32 config_data;	/* Config data TBD */
+		};
+	};
+};
+
+/* FME Partial Reconfiguration Status */
+struct feature_fme_pr_status {
+	union {
+		u64 csr;
+		struct {
+			u16 pr_credit:9;	/* PR Credits */
+			u8  rsvd2:7;		/* Reserved */
+			u8  pr_status:1;	/* PR status */
+			u8  rsvd:3;		/* Reserved */
+			/* Altra PR Controller Block status */
+			u8  pr_controller_status:3;
+			u8  rsvd1:1;            /* Reserved */
+			u8  pr_host_status:4;   /* PR Host status */
+			u8  rsvd3:4;		/* Reserved */
+			/* Security Block Status fields (TBD) */
+			u32 security_bstatus;
+		};
+	};
+};
+
+/* FME Partial Reconfiguration Data */
+struct feature_fme_pr_data {
+	union {
+		u64 csr;	/* PR data from the raw-binary file */
+		struct {
+			/* PR data from the raw-binary file */
+			u32 pr_data_raw;
+			u32 rsvd;
+		};
+	};
+};
+
+/* FME PR Public Key */
+struct feature_fme_pr_key {
+	u64 key;		/* FME PR Public Hash */
+};
+
+/* FME PR FEATURE */
+struct feature_fme_pr {
+	struct feature_header header;
+	/*Partial Reconfiguration control */
+	struct feature_fme_pr_ctl	ccip_fme_pr_control;
+
+	/* Partial Reconfiguration Status */
+	struct feature_fme_pr_status	ccip_fme_pr_status;
+
+	/* Partial Reconfiguration data */
+	struct feature_fme_pr_data	ccip_fme_pr_data;
+
+	/* Partial Reconfiguration data */
+	u64				ccip_fme_pr_err;
+
+	u64 rsvd1[3];
+
+	/* Partial Reconfiguration data registers */
+	u64 fme_pr_data1;
+	u64 fme_pr_data2;
+	u64 fme_pr_data3;
+	u64 fme_pr_data4;
+	u64 fme_pr_data5;
+	u64 fme_pr_data6;
+	u64 fme_pr_data7;
+	u64 fme_pr_data8;
+
+	u64 rsvd2[5];
+
+	/* PR Interface ID */
+	u64 fme_pr_intfc_id_l;
+	u64 fme_pr_intfc_id_h;
+
+	/* MSIX filed to be Added */
+};
+
+/* FME HSSI Control */
+struct feature_fme_hssi_eth_ctrl {
+	union {
+		u64 csr;
+		struct {
+			u32 data:32;		/* HSSI data */
+			u16 address:16;		/* HSSI address */
+			/*
+			 * HSSI comamnd
+			 * 0x0 - No request
+			 * 0x08 - SW register RD request
+			 * 0x10 - SW register WR request
+			 * 0x40 - Auxiliar bus RD request
+			 * 0x80 - Auxiliar bus WR request
+			 */
+			u16 cmd:16;
+		};
+	};
+};
+
+/* FME HSSI Status */
+struct feature_fme_hssi_eth_stat {
+	union {
+		u64 csr;
+		struct {
+			u32 data:32;		/* HSSI data */
+			u8  acknowledge:1;	/* HSSI acknowledge */
+			u8  spare:1;		/* HSSI spare */
+			u32 rsvd:30;		/* Reserved */
+		};
+	};
+};
+
+/* FME HSSI FEATURE */
+struct feature_fme_hssi {
+	struct feature_header header;
+	struct feature_fme_hssi_eth_ctrl	hssi_control;
+	struct feature_fme_hssi_eth_stat	hssi_status;
+};
+
+#define PORT_ERR_MASK		0xfff0703ff001f
+struct feature_port_err_key {
+	union {
+		u64 csr;
+		struct {
+			/* Tx Channel0: Overflow */
+			u8 tx_ch0_overflow:1;
+			/* Tx Channel0: Invalid request encoding */
+			u8 tx_ch0_invaldreq :1;
+			/* Tx Channel0: Request with cl_len=3 not supported */
+			u8 tx_ch0_cl_len3:1;
+			/* Tx Channel0: Request with cl_len=2 not aligned 2CL */
+			u8 tx_ch0_cl_len2:1;
+			/* Tx Channel0: Request with cl_len=4 not aligned 4CL */
+			u8 tx_ch0_cl_len4:1;
+
+			u16 rsvd1:4;			/* Reserved */
+
+			/* AFU MMIO RD received while PORT is in reset */
+			u8 mmio_rd_whilerst:1;
+			/* AFU MMIO WR received while PORT is in reset */
+			u8 mmio_wr_whilerst:1;
+
+			u16 rsvd2:5;			/* Reserved */
+
+			/* Tx Channel1: Overflow */
+			u8 tx_ch1_overflow:1;
+			/* Tx Channel1: Invalid request encoding */
+			u8 tx_ch1_invaldreq:1;
+			/* Tx Channel1: Request with cl_len=3 not supported */
+			u8 tx_ch1_cl_len3:1;
+			/* Tx Channel1: Request with cl_len=2 not aligned 2CL */
+			u8 tx_ch1_cl_len2:1;
+			/* Tx Channel1: Request with cl_len=4 not aligned 4CL */
+			u8 tx_ch1_cl_len4:1;
+
+			/* Tx Channel1: Insufficient data payload */
+			u8 tx_ch1_insuff_data:1;
+			/* Tx Channel1: Data payload overrun */
+			u8 tx_ch1_data_overrun:1;
+			/* Tx Channel1 : Incorrect address */
+			u8 tx_ch1_incorr_addr:1;
+			/* Tx Channel1 : NON-Zero SOP Detected */
+			u8 tx_ch1_nzsop:1;
+			/* Tx Channel1 : Illegal VC_SEL, atomic request VLO */
+			u8 tx_ch1_illegal_vcsel:1;
+
+			u8 rsvd3:6;			/* Reserved */
+
+			/* MMIO Read Timeout in AFU */
+			u8 mmioread_timeout:1;
+
+			/* Tx Channel2: FIFO Overflow */
+			u8 tx_ch2_fifo_overflow:1;
+
+			/* MMIO read is not matching pending request */
+			u8 unexp_mmio_resp:1;
+
+			u8 rsvd4:5;			/* Reserved */
+
+			/* Number of pending Requests: counter overflow */
+			u8 tx_req_counter_overflow:1;
+			/* Req with Address violating SMM Range */
+			u8 llpr_smrr_err:1;
+			/* Req with Address violating second SMM Range */
+			u8 llpr_smrr2_err:1;
+			/* Req with Address violating ME Stolen message */
+			u8 llpr_mesg_err:1;
+			/* Req with Address violating Generic Protected Range */
+			u8 genprot_range_err:1;
+			/* Req with Address violating Legacy Range low */
+			u8 legrange_low_err:1;
+			/* Req with Address violating Legacy Range High */
+			u8 legrange_high_err:1;
+			/* Req with Address violating VGA memory range */
+			u8 vgmem_range_err:1;
+			u8 page_fault_err:1;		/* Page fault */
+			u8 pmr_err:1;			/* PMR Error */
+			u8 ap6_event:1;			/* AP6 event */
+			/* VF FLR detected on Port with PF access control */
+			u8 vfflr_access_err:1;
+			u16 rsvd5:12;			/* Reserved */
+		};
+	};
+};
+
+/* Port first error register, not contain all error bits in error register. */
+struct feature_port_first_err_key {
+	union {
+		u64 csr;
+		struct {
+			u8 tx_ch0_overflow:1;
+			u8 tx_ch0_invaldreq :1;
+			u8 tx_ch0_cl_len3:1;
+			u8 tx_ch0_cl_len2:1;
+			u8 tx_ch0_cl_len4:1;
+			u8 rsvd1:4;			/* Reserved */
+			u8 mmio_rd_whilerst:1;
+			u8 mmio_wr_whilerst:1;
+			u8 rsvd2:5;			/* Reserved */
+			u8 tx_ch1_overflow:1;
+			u8 tx_ch1_invaldreq:1;
+			u8 tx_ch1_cl_len3:1;
+			u8 tx_ch1_cl_len2:1;
+			u8 tx_ch1_cl_len4:1;
+			u8 tx_ch1_insuff_data:1;
+			u8 tx_ch1_data_overrun:1;
+			u8 tx_ch1_incorr_addr:1;
+			u8 tx_ch1_nzsop:1;
+			u8 tx_ch1_illegal_vcsel:1;
+			u8 rsvd3:6;			/* Reserved */
+			u8 mmioread_timeout:1;
+			u8 tx_ch2_fifo_overflow:1;
+			u8 rsvd4:6;			/* Reserved */
+			u8 tx_req_counter_overflow:1;
+			u32 rsvd5:23;			/* Reserved */
+		};
+	};
+};
+
+/* Port malformed Req0 */
+struct feature_port_malformed_req0 {
+	u64 header_lsb;
+};
+
+/* Port malformed Req1 */
+struct feature_port_malformed_req1 {
+	u64 header_msb;
+};
+
+/* Port debug register */
+struct feature_port_debug {
+	u64 port_debug;
+};
+
+/* Port error capabilities */
+struct feature_port_err_capability {
+	union {
+		u64 csr;
+		struct {
+			u8  support_intr:1;
+			/* MSI-X vector table entry number */
+			u16 intr_vector_num:12;
+			u64 rsvd:51;            /* Reserved */
+		};
+	};
+};
+
+/* PORT FEATURE ERROR */
+struct feature_port_error {
+	struct feature_header header;
+	struct feature_port_err_key error_mask;
+	struct feature_port_err_key port_error;
+	struct feature_port_first_err_key port_first_error;
+	struct feature_port_malformed_req0 malreq0;
+	struct feature_port_malformed_req1 malreq1;
+	struct feature_port_debug port_debug;
+	struct feature_port_err_capability error_capability;
+};
+
+/* Port UMSG Capability */
+struct feature_port_umsg_cap {
+	union {
+		u64 csr;
+		struct {
+			/* Number of umsg allocated to this port */
+			u8 umsg_allocated;
+			/* Enable / Disable UMsg engine for this port */
+			u8 umsg_enable:1;
+			/* Usmg initialization status */
+			u8 umsg_init_complete:1;
+			/* IOMMU can not translate the umsg base address */
+			u8 umsg_trans_error:1;
+			u64 rsvd:53;		/* Reserved */
+		};
+	};
+};
+
+/* Port UMSG base address */
+struct feature_port_umsg_baseaddr {
+	union {
+		u64 csr;
+		struct {
+			u64 base_addr:48;	/* 48 bit physical address */
+			u16 rsvd;		/* Reserved */
+		};
+	};
+};
+
+struct feature_port_umsg_mode {
+	union {
+		u64 csr;
+		struct {
+			u32 umsg_hint_enable;	/* UMSG hint enable/disable */
+			u32 rsvd;		/* Reserved */
+		};
+	};
+};
+
+/* PORT FEATURE UMSG */
+struct feature_port_umsg {
+	struct feature_header header;
+	struct feature_port_umsg_cap capability;
+	struct feature_port_umsg_baseaddr baseaddr;
+	struct feature_port_umsg_mode mode;
+};
+
+#define UMSG_EN_POLL_INVL 10 /* us */
+#define UMSG_EN_POLL_TIMEOUT 1000 /* us */
+
+/* Port UINT Capability */
+struct feature_port_uint_cap {
+	union {
+		u64 csr;
+		struct {
+			u16 intr_num:12;	/* Supported interrupts num */
+			/* First MSI-X vector table entry number */
+			u16 first_vec_num:12;
+			u64 rsvd:40;
+		};
+	};
+};
+
+/* PORT FEATURE UINT */
+struct feature_port_uint {
+	struct feature_header header;
+	struct feature_port_uint_cap capability;
+};
+
+/* STP region supports mmap operation, so use page aligned size. */
+#define PORT_FEATURE_STP_REGION_SIZE PAGE_ALIGN(sizeof(struct feature_port_stp))
+
+/* Port STP status register (for debug only)*/
+struct feature_port_stp_status {
+	union {
+		u64 csr;
+		struct {
+			/* SLD Hub end-point read/write timeout */
+			u8 sld_ep_timeout:1;
+			/* Remote STP in reset/disable */
+			u8 rstp_disabled:1;
+			u8 unsupported_read:1;
+			/* MMIO timeout detected and faked with a response */
+			u8 mmio_timeout:1;
+			u8 txfifo_count:4;
+			u8 rxfifo_count:4;
+			u8 txfifo_overflow:1;
+			u8 txfifo_underflow:1;
+			u8 rxfifo_overflow:1;
+			u8 rxfifo_underflow:1;
+			/* Number of MMIO write requests */
+			u16 write_requests;
+			/* Number of MMIO read requests */
+			u16 read_requests;
+			/* Number of MMIO read responses */
+			u16 read_responses;
+		};
+	};
+};
+
+/*
+ * PORT FEATURE STP
+ * Most registers in STP region are not touched by driver, but mmapped to user
+ * space. So they are not defined in below data structure, as its actual size
+ * is 0x18c per spec.
+ */
+struct feature_port_stp {
+	struct feature_header header;
+	struct feature_port_stp_status stp_status;
+};
+
+/**
+ * enum fpga_pr_states - fpga PR states
+ * @FPGA_PR_STATE_UNKNOWN: can't determine state
+ * @FPGA_PR_STATE_WRITE_INIT: preparing FPGA for programming
+ * @FPGA_PR_STATE_WRITE_INIT_ERR: Error during WRITE_INIT stage
+ * @FPGA_PR_STATE_WRITE: writing image to FPGA
+ * @FPGA_PR_STATE_WRITE_ERR: Error while writing FPGA
+ * @FPGA_PR_STATE_WRITE_COMPLETE: Doing post programming steps
+ * @FPGA_PR_STATE_WRITE_COMPLETE_ERR: Error during WRITE_COMPLETE
+ * @FPGA_PR_STATE_OPERATING: FPGA PR done
+ */
+enum fpga_pr_states {
+	/* canot determine state states */
+	FPGA_PR_STATE_UNKNOWN,
+
+	/* write sequence: init, write, complete */
+	FPGA_PR_STATE_WRITE_INIT,
+	FPGA_PR_STATE_WRITE_INIT_ERR,
+	FPGA_PR_STATE_WRITE,
+	FPGA_PR_STATE_WRITE_ERR,
+	FPGA_PR_STATE_WRITE_COMPLETE,
+	FPGA_PR_STATE_WRITE_COMPLETE_ERR,
+
+	/* FPGA PR done */
+	FPGA_PR_STATE_DONE,
+};
+
+/*
+ * FPGA Manager flags
+ * FPGA_MGR_PARTIAL_RECONFIG: do partial reconfiguration if supported
+ */
+#define FPGA_MGR_PARTIAL_RECONFIG	BIT(0)
+
+/**
+ * struct fpga_pr_info - specific information to a FPGA PR
+ * @flags: boolean flags as defined above
+ * @pr_err: PR error code
+ * @state: fpga manager state
+ * @port_id: port id
+ */
+struct fpga_pr_info {
+	u32 flags;
+	u64 pr_err;
+	enum fpga_pr_states state;
+	int port_id;
+};
+
+#define DEFINE_FPGA_PR_ERR_MSG(_name_)			\
+static const char * const _name_[] = {			\
+	"PR operation error detected",			\
+	"PR CRC error detected",			\
+	"PR incompatiable bitstream error detected",	\
+	"PR IP protocol error detected",		\
+	"PR FIFO overflow error detected",		\
+	"PR timeout error detected",			\
+	"PR secure load error detected",		\
+}
+
+#define RST_POLL_INVL 10 /* us */
+#define RST_POLL_TIMEOUT 1000 /* us */
+
+#define PR_WAIT_TIMEOUT   15000000
+
+#define PR_HOST_STATUS_IDLE	0
+#define PR_MAX_ERR_NUM	7
+
+DEFINE_FPGA_PR_ERR_MSG(pr_err_msg);
+
+/*
+ * green bitstream header must be byte-packed to match the
+ * real file format.
+ */
+struct bts_header {
+	u64 guid_h;
+	u64 guid_l;
+	u32 metadata_len;
+};
+
+#define GBS_GUID_H		0x414750466e6f6558
+#define GBS_GUID_L		0x31303076534247b7
+#define is_valid_bts(bts_hdr)				\
+	(((bts_hdr)->guid_h == GBS_GUID_H) &&		\
+	((bts_hdr)->guid_l == GBS_GUID_L))
+
+#endif /* _BASE_IFPGA_DEFINES_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c b/drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c
new file mode 100644
index 0000000..c6c081a
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c
@@ -0,0 +1,861 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "opae_hw_api.h"
+#include "ifpga_api.h"
+
+#include "ifpga_hw.h"
+#include "ifpga_enumerate.h"
+#include "ifpga_feature_dev.h"
+
+struct build_feature_devs_info {
+	struct opae_adapter_data_pci *pci_data;
+
+	struct ifpga_afu_info *acc_info;
+
+	void *fiu;
+	enum fpga_id_type current_type;
+	int current_port_id;
+
+	void *ioaddr;
+	void *ioend;
+	uint64_t phys_addr;
+	int current_bar;
+
+	void *pfme_hdr;
+
+	struct ifpga_hw *hw;
+};
+
+struct feature_info {
+	const char *name;
+	uint32_t resource_size;
+	int feature_index;
+	int revision_id;
+	unsigned int vec_start;
+	unsigned int vec_cnt;
+
+	struct feature_ops *ops;
+};
+
+/* indexed by fme feature IDs which are defined in 'enum fme_feature_id'. */
+static struct feature_info fme_features[] = {
+	{
+		.name = FME_FEATURE_HEADER,
+		.resource_size = sizeof(struct feature_fme_header),
+		.feature_index = FME_FEATURE_ID_HEADER,
+		.revision_id = FME_HEADER_REVISION,
+		.ops = &fme_hdr_ops,
+	},
+	{
+		.name = FME_FEATURE_THERMAL_MGMT,
+		.resource_size = sizeof(struct feature_fme_thermal),
+		.feature_index = FME_FEATURE_ID_THERMAL_MGMT,
+		.revision_id = FME_THERMAL_MGMT_REVISION,
+		.ops = &fme_thermal_mgmt_ops,
+	},
+	{
+		.name = FME_FEATURE_POWER_MGMT,
+		.resource_size = sizeof(struct feature_fme_power),
+		.feature_index = FME_FEATURE_ID_POWER_MGMT,
+		.revision_id = FME_POWER_MGMT_REVISION,
+		.ops = &fme_power_mgmt_ops,
+	},
+	{
+		.name = FME_FEATURE_GLOBAL_IPERF,
+		.resource_size = sizeof(struct feature_fme_iperf),
+		.feature_index = FME_FEATURE_ID_GLOBAL_IPERF,
+		.revision_id = FME_GLOBAL_IPERF_REVISION,
+		.ops = &fme_global_iperf_ops,
+	},
+	{
+		.name = FME_FEATURE_GLOBAL_ERR,
+		.resource_size = sizeof(struct feature_fme_err),
+		.feature_index = FME_FEATURE_ID_GLOBAL_ERR,
+		.revision_id = FME_GLOBAL_ERR_REVISION,
+		.ops = &fme_global_err_ops,
+	},
+	{
+		.name = FME_FEATURE_PR_MGMT,
+		.resource_size = sizeof(struct feature_fme_pr),
+		.feature_index = FME_FEATURE_ID_PR_MGMT,
+		.revision_id = FME_PR_MGMT_REVISION,
+		.ops = &fme_pr_mgmt_ops,
+	},
+	{
+		.name = FME_FEATURE_HSSI_ETH,
+		.resource_size = sizeof(struct feature_fme_hssi),
+		.feature_index = FME_FEATURE_ID_HSSI_ETH,
+		.revision_id = FME_HSSI_ETH_REVISION
+	},
+	{
+		.name = FME_FEATURE_GLOBAL_DPERF,
+		.resource_size = sizeof(struct feature_fme_dperf),
+		.feature_index = FME_FEATURE_ID_GLOBAL_DPERF,
+		.revision_id = FME_GLOBAL_DPERF_REVISION,
+		.ops = &fme_global_dperf_ops,
+	}
+};
+
+static struct feature_info port_features[] = {
+	{
+		.name = PORT_FEATURE_HEADER,
+		.resource_size = sizeof(struct feature_port_header),
+		.feature_index = PORT_FEATURE_ID_HEADER,
+		.revision_id = PORT_HEADER_REVISION,
+		.ops = &port_hdr_ops,
+	},
+	{
+		.name = PORT_FEATURE_ERR,
+		.resource_size = sizeof(struct feature_port_error),
+		.feature_index = PORT_FEATURE_ID_ERROR,
+		.revision_id = PORT_ERR_REVISION,
+		.ops = &port_error_ops,
+	},
+	{
+		.name = PORT_FEATURE_UMSG,
+		.resource_size = sizeof(struct feature_port_umsg),
+		.feature_index = PORT_FEATURE_ID_UMSG,
+		.revision_id = PORT_UMSG_REVISION,
+		.ops = &port_umsg_ops,
+	},
+	{
+		.name = PORT_FEATURE_UINT,
+		.resource_size = sizeof(struct feature_port_uint),
+		.feature_index = PORT_FEATURE_ID_UINT,
+		.revision_id = PORT_UINT_REVISION,
+		.ops = &port_uint_ops,
+	},
+	{
+		.name = PORT_FEATURE_STP,
+		.resource_size = PORT_FEATURE_STP_REGION_SIZE,
+		.feature_index = PORT_FEATURE_ID_STP,
+		.revision_id = PORT_STP_REVISION,
+		.ops = &port_stp_ops,
+	},
+	{
+		.name = PORT_FEATURE_UAFU,
+		/* UAFU feature size should be read from PORT_CAP.MMIOSIZE.
+		 * Will set uafu feature size while parse port device.
+		 */
+		.resource_size = 0,
+		.feature_index = PORT_FEATURE_ID_UAFU,
+		.revision_id = PORT_UAFU_REVISION
+	},
+};
+
+static u64 feature_id(void __iomem *start)
+{
+	struct feature_header header;
+
+	header.csr = readq(start);
+
+	switch (header.type) {
+	case FEATURE_TYPE_FIU:
+		return FEATURE_ID_HEADER;
+	case FEATURE_TYPE_PRIVATE:
+		return header.id;
+	case FEATURE_TYPE_AFU:
+		return FEATURE_ID_AFU;
+	}
+
+	WARN_ON(1);
+	return 0;
+}
+
+static int
+build_info_add_sub_feature(struct build_feature_devs_info *binfo,
+			   struct feature_info *finfo, void __iomem *start)
+{
+	struct ifpga_hw *hw = binfo->hw;
+	struct feature *feature = NULL;
+	int feature_idx = finfo->feature_index;
+	unsigned int vec_start = finfo->vec_start;
+	unsigned int vec_cnt = finfo->vec_cnt;
+	struct feature_irq_ctx *ctx = NULL;
+	int port_id, ret = 0;
+	unsigned int i;
+
+	if (binfo->current_type == FME_ID) {
+		feature = &hw->fme.sub_feature[feature_idx];
+		feature->parent = &hw->fme;
+	} else if (binfo->current_type == PORT_ID) {
+		port_id = binfo->current_port_id;
+		feature = &hw->port[port_id].sub_feature[feature_idx];
+		feature->parent = &hw->port[port_id];
+	} else {
+		return -EFAULT;
+	}
+
+	feature->state = IFPGA_FEATURE_ATTACHED;
+	feature->addr = start;
+	feature->id = feature_id(start);
+	feature->size = finfo->resource_size;
+	feature->name = finfo->name;
+	feature->revision = finfo->revision_id;
+	feature->ops = finfo->ops;
+	feature->phys_addr = binfo->phys_addr +
+				((u8 *)start - (u8 *)binfo->ioaddr);
+
+	if (vec_cnt) {
+		if (vec_start + vec_cnt <= vec_start)
+			return -EINVAL;
+
+		ctx = zmalloc(sizeof(*ctx) * vec_cnt);
+		if (!ctx)
+			return -ENOMEM;
+
+		for (i = 0; i < vec_cnt; i++) {
+			ctx[i].eventfd = -1;
+			ctx[i].idx = vec_start + i;
+		}
+	}
+
+	feature->ctx = ctx;
+	feature->ctx_num = vec_cnt;
+	feature->vfio_dev_fd = binfo->pci_data->vfio_dev_fd;
+
+	return ret;
+}
+
+static int
+create_feature_instance(struct build_feature_devs_info *binfo,
+			void __iomem *start, struct feature_info *finfo)
+{
+	struct feature_header *hdr = start;
+
+	if (((u8 *)binfo->ioend - (u8 *)start) < finfo->resource_size)
+		return -EINVAL;
+
+	if (finfo->revision_id != SKIP_REVISION_CHECK &&
+	    hdr->revision > finfo->revision_id) {
+		dev_err(binfo, "feature %s revision :default:%x, now at:%x, mis-match.\n",
+			finfo->name, finfo->revision_id, hdr->revision);
+	}
+
+	return build_info_add_sub_feature(binfo, finfo, start);
+}
+
+/*
+ * UAFU GUID is dynamic as it can be changed after FME downloads different
+ * Green Bitstream to the port, so we treat the unknown GUIDs which are
+ * attached on port's feature list as UAFU.
+ */
+static bool feature_is_UAFU(struct build_feature_devs_info *binfo)
+{
+	if (binfo->current_type != PORT_ID)
+		return false;
+
+	return true;
+}
+
+static int parse_feature_port_uafu(struct build_feature_devs_info *binfo,
+				   struct feature_header *hdr)
+{
+	enum port_feature_id id = PORT_FEATURE_ID_UAFU;
+	struct ifpga_afu_info *info;
+	void *start = (void *)hdr;
+	int ret;
+
+	if (port_features[id].resource_size) {
+		ret = create_feature_instance(binfo, hdr, &port_features[id]);
+		//port_features[id].resource_size = 0;
+	} else {
+		dev_err(binfo, "the uafu feature header is mis-configured.\n");
+		ret = -EINVAL;
+	}
+
+	if (ret)
+		return ret;
+
+	/* FIXME: need to figure out a better name */
+	info = malloc(sizeof(*info));
+	if (!info)
+		return -ENOMEM;
+
+	info->region[0].addr = start;
+	info->region[0].phys_addr = binfo->phys_addr +
+			(uint8_t *)start - (uint8_t *)binfo->ioaddr;
+	info->region[0].len = port_features[id].resource_size;
+	port_features[id].resource_size = 0;
+	info->num_regions = 1;
+
+	binfo->acc_info = info;
+
+	return ret;
+}
+
+static int parse_feature_afus(struct build_feature_devs_info *binfo,
+			      struct feature_header *hdr)
+{
+	int ret;
+	struct feature_afu_header *afu_hdr, header;
+	u8 __iomem *start;
+	u8 __iomem *end = binfo->ioend;
+
+	start = (u8 __iomem *)hdr;
+	for (; start < end; start += header.next_afu) {
+		if (end - start <
+			(unsigned int)(sizeof(*afu_hdr) + sizeof(*hdr)))
+			return -EINVAL;
+
+		hdr = (struct feature_header *)start;
+		afu_hdr = (struct feature_afu_header *)(hdr + 1);
+		header.csr = readq(&afu_hdr->csr);
+
+		if (feature_is_UAFU(binfo))
+			ret = parse_feature_port_uafu(binfo, hdr);
+			if (ret)
+				return ret;
+
+		if (!header.next_afu)
+			break;
+	}
+
+	return 0;
+}
+
+/* create and register proper private data */
+static int build_info_commit_dev(struct build_feature_devs_info *binfo)
+{
+	struct ifpga_afu_info *info = binfo->acc_info;
+	struct ifpga_hw *hw = binfo->hw;
+	struct opae_manager *mgr;
+	struct opae_bridge *br;
+	struct opae_accelerator *acc;
+
+	if (!binfo->fiu)
+		return 0;
+
+	if (binfo->current_type == PORT_ID) {
+		/* return error if no valid acc info data structure */
+		if (!info)
+			return -EFAULT;
+
+		br = opae_bridge_alloc(hw->adapter->name, &ifpga_br_ops,
+				       binfo->fiu);
+		if (!br)
+			return -ENOMEM;
+
+		br->id = binfo->current_port_id;
+
+		/* update irq info */
+		info->num_irqs = port_features[PORT_FEATURE_ID_UINT].vec_cnt;
+
+		acc = opae_accelerator_alloc(hw->adapter->name,
+					     &ifpga_acc_ops, info);
+		if (!acc) {
+			opae_bridge_free(br);
+			return -ENOMEM;
+		}
+
+		acc->br = br;
+		acc->index = br->id;
+
+		opae_adapter_add_acc(hw->adapter, acc);
+
+	} else if (binfo->current_type == FME_ID) {
+		mgr = opae_manager_alloc(hw->adapter->name, &ifpga_mgr_ops,
+					 binfo->fiu);
+		if (!mgr)
+			return -ENOMEM;
+
+		mgr->adapter = hw->adapter;
+		hw->adapter->mgr = mgr;
+	}
+
+	binfo->fiu = NULL;
+
+	return 0;
+}
+
+static int
+build_info_create_dev(struct build_feature_devs_info *binfo,
+		      enum fpga_id_type type, unsigned int index)
+{
+	int ret;
+
+	ret = build_info_commit_dev(binfo);
+	if (ret)
+		return ret;
+
+	binfo->current_type = type;
+
+	if (type == FME_ID) {
+		binfo->fiu = &binfo->hw->fme;
+	} else if (type == PORT_ID) {
+		binfo->fiu = &binfo->hw->port[index];
+		binfo->current_port_id = index;
+	}
+
+	return 0;
+}
+
+static int parse_feature_fme(struct build_feature_devs_info *binfo,
+			     struct feature_header *start)
+{
+	struct ifpga_hw *hw = binfo->hw;
+	struct ifpga_fme_hw *fme = &hw->fme;
+	int ret;
+
+	ret = build_info_create_dev(binfo, FME_ID, 0);
+	if (ret)
+		return ret;
+
+	/* Update FME states */
+	fme->state = IFPGA_FME_IMPLEMENTED;
+	fme->parent = hw;
+	spinlock_init(&fme->lock);
+
+	return create_feature_instance(binfo, start,
+				       &fme_features[FME_FEATURE_ID_HEADER]);
+}
+
+static int parse_feature_port(struct build_feature_devs_info *binfo,
+			      void __iomem *start)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_capability capability;
+	struct ifpga_hw *hw = binfo->hw;
+	struct ifpga_port_hw *port;
+	unsigned int port_id;
+	int ret;
+
+	/* Get current port's id */
+	port_hdr = (struct feature_port_header *)start;
+	capability.csr = readq(&port_hdr->capability);
+	port_id = capability.port_number;
+
+	ret = build_info_create_dev(binfo, PORT_ID, port_id);
+	if (ret)
+		return ret;
+
+	/*found a Port device*/
+	port = &hw->port[port_id];
+	port->port_id = binfo->current_port_id;
+	port->parent = hw;
+	port->state = IFPGA_PORT_ATTACHED;
+	spinlock_init(&port->lock);
+
+	return create_feature_instance(binfo, start,
+				      &port_features[PORT_FEATURE_ID_HEADER]);
+}
+
+static void enable_port_uafu(struct build_feature_devs_info *binfo,
+			     void __iomem *start)
+{
+	enum port_feature_id id = PORT_FEATURE_ID_UAFU;
+	struct feature_port_header *port_hdr;
+	struct feature_port_capability capability;
+	struct ifpga_port_hw *port = &binfo->hw->port[binfo->current_port_id];
+
+	port_hdr = (struct feature_port_header *)start;
+	capability.csr = readq(&port_hdr->capability);
+	port_features[id].resource_size = (capability.mmio_size << 10);
+
+	/*
+	 * From SAS spec, to Enable UAFU, we should reset related port,
+	 * or the whole mmio space in this UAFU will be invalid
+	 */
+	if (port_features[id].resource_size)
+		fpga_port_reset(port);
+}
+
+static int parse_feature_fiu(struct build_feature_devs_info *binfo,
+			     struct feature_header *hdr)
+{
+	struct feature_header header;
+	struct feature_fiu_header *fiu_hdr, fiu_header;
+	u8 __iomem *start = (u8 __iomem *)hdr;
+	int ret;
+
+	header.csr = readq(hdr);
+
+	switch (header.id) {
+	case FEATURE_FIU_ID_FME:
+		ret = parse_feature_fme(binfo, hdr);
+		//check_features_header(binfo->pdev, hdr, FPGA_DEVT_FME, 0);
+		binfo->pfme_hdr = hdr;
+		if (ret)
+			return ret;
+		break;
+	case FEATURE_FIU_ID_PORT:
+		ret = parse_feature_port(binfo, hdr);
+		//check_features_header(binfo->pdev, hdr, FPGA_DEVT_PORT, 0);
+		enable_port_uafu(binfo, hdr);
+		if (ret)
+			return ret;
+
+		/* Check Port FIU's next_afu pointer to User AFU DFH */
+		fiu_hdr = (struct feature_fiu_header *)(hdr + 1);
+		fiu_header.csr = readq(&fiu_hdr->csr);
+
+		if (fiu_header.next_afu) {
+			start += fiu_header.next_afu;
+			ret = parse_feature_afus(binfo,
+					(struct feature_header *)start);
+			if (ret)
+				return ret;
+		} else {
+			dev_info(binfo, "No AFUs detected on Port\n");
+		}
+
+		break;
+	default:
+		dev_info(binfo, "FIU TYPE %d is not supported yet.\n",
+			 header.id);
+	}
+
+	return 0;
+}
+
+static void parse_feature_irqs(struct build_feature_devs_info *binfo,
+				void __iomem *start, struct feature_info *finfo)
+{
+	finfo->vec_start = 0;
+	finfo->vec_cnt = 0;
+
+	UNUSED(binfo);
+
+	if (!strcmp(finfo->name, PORT_FEATURE_UINT)) {
+		struct feature_port_uint *port_uint = start;
+		struct feature_port_uint_cap uint_cap;
+
+		uint_cap.csr = readq(&port_uint->capability);
+		if (uint_cap.intr_num) {
+			finfo->vec_start = uint_cap.first_vec_num;
+			finfo->vec_cnt = uint_cap.intr_num;
+		} else {
+			dev_debug(binfo, "UAFU doesn't support interrupt\n");
+		}
+
+	} else if (!strcmp(finfo->name, PORT_FEATURE_ERR)) {
+		struct feature_port_error *port_err = start;
+		struct feature_port_err_capability port_err_cap;
+
+		port_err_cap.csr = readq(&port_err->error_capability);
+		if (port_err_cap.support_intr) {
+			finfo->vec_start = port_err_cap.intr_vector_num;
+			finfo->vec_cnt = 1;
+		} else
+			dev_debug(&binfo, "Port error doesn't support interrupt\n");
+
+	} else if (!strcmp(finfo->name, FME_FEATURE_GLOBAL_ERR)) {
+		struct feature_fme_err *fme_err = start;
+		struct feature_fme_error_capability fme_err_cap;
+
+		fme_err_cap.csr = readq(&fme_err->fme_err_capability);
+		if (fme_err_cap.support_intr) {
+			finfo->vec_start = fme_err_cap.intr_vector_num;
+			finfo->vec_cnt = 1;
+		} else
+			dev_debug(&binfo, "FME error doesn't support interrupt\n");
+	}
+}
+
+static int parse_feature_fme_private(struct build_feature_devs_info *binfo,
+				     struct feature_header *hdr)
+{
+	struct feature_header header;
+
+	header.csr = readq(hdr);
+
+	if (header.id >= ARRAY_SIZE(fme_features)) {
+		dev_err(binfo, "FME feature id %x is not supported yet.\n",
+			header.id);
+		return 0;
+	}
+
+	parse_feature_irqs(binfo, hdr, &fme_features[header.id]);
+
+	//check_features_header(binfo->pdev, hdr, FPGA_DEVT_FME, header.id);
+
+	return create_feature_instance(binfo, hdr, &fme_features[header.id]);
+}
+
+static int parse_feature_port_private(struct build_feature_devs_info *binfo,
+				      struct feature_header *hdr)
+{
+	struct feature_header header;
+	enum port_feature_id id;
+
+	header.csr = readq(hdr);
+	/*
+	 * the region of port feature id is [0x10, 0x13], + 1 to reserve 0
+	 * which is dedicated for port-hdr.
+	 */
+	id = (header.id & 0x000f) + 1;
+
+	if (id >= ARRAY_SIZE(port_features)) {
+		dev_err(binfo, "Port feature id %x is not supported yet.\n",
+			header.id);
+		return 0;
+	}
+
+	parse_feature_irqs(binfo, hdr, &port_features[id]);
+
+	//check_features_header(binfo->pdev, hdr, FPGA_DEVT_PORT, id);
+
+	return create_feature_instance(binfo, hdr, &port_features[id]);
+}
+
+static int parse_feature_private(struct build_feature_devs_info *binfo,
+				 struct feature_header *hdr)
+{
+	struct feature_header header;
+
+	header.csr = readq(hdr);
+
+	switch (binfo->current_type) {
+	case FME_ID:
+		return parse_feature_fme_private(binfo, hdr);
+	case PORT_ID:
+		return parse_feature_port_private(binfo, hdr);
+	default:
+		dev_err(binfo, "private feature %x belonging to AFU %d (unknown_type) is not supported yet.\n",
+			header.id, binfo->current_type);
+	}
+	return 0;
+}
+
+static int parse_feature(struct build_feature_devs_info *binfo,
+			 struct feature_header *hdr)
+{
+	struct feature_header header;
+	int ret = 0;
+
+	header.csr = readq(hdr);
+
+	switch (header.type) {
+	case FEATURE_TYPE_AFU:
+		ret = parse_feature_afus(binfo, hdr);
+		break;
+	case FEATURE_TYPE_PRIVATE:
+		ret = parse_feature_private(binfo, hdr);
+		break;
+	case FEATURE_TYPE_FIU:
+		ret = parse_feature_fiu(binfo, hdr);
+		break;
+	default:
+		dev_err(binfo, "Feature Type %x is not supported.\n",
+			hdr->type);
+	};
+
+	return ret;
+}
+
+static int
+parse_feature_list(struct build_feature_devs_info *binfo, u8 __iomem *start)
+{
+	struct feature_header *hdr, header;
+	u8 __iomem *end = (u8 __iomem *)binfo->ioend;
+	int ret = 0;
+
+	for (; start < end; start += header.next_header_offset) {
+		if (end - start < (unsigned int)sizeof(*hdr)) {
+			dev_err(binfo, "The region is too small to contain a feature.\n");
+			ret =  -EINVAL;
+			break;
+		}
+
+		hdr = (struct feature_header *)start;
+		ret = parse_feature(binfo, hdr);
+		if (ret)
+			return ret;
+
+		header.csr = readq(hdr);
+		if (!header.next_header_offset)
+			break;
+	}
+
+	return build_info_commit_dev(binfo);
+}
+
+/* switch the memory mapping to BAR# @bar */
+static int parse_switch_to(struct build_feature_devs_info *binfo, int bar)
+{
+	struct opae_adapter_data_pci *pci_data = binfo->pci_data;
+
+	if (!pci_data->region[bar].addr)
+		return -ENOMEM;
+
+	binfo->ioaddr = pci_data->region[bar].addr;
+	binfo->ioend = (u8 __iomem *)binfo->ioaddr + pci_data->region[bar].len;
+	binfo->phys_addr = pci_data->region[bar].phys_addr;
+	binfo->current_bar = bar;
+
+	return 0;
+}
+
+static int parse_ports_from_fme(struct build_feature_devs_info *binfo)
+{
+	struct feature_fme_header *fme_hdr;
+	struct feature_fme_port port;
+	int i = 0, ret = 0;
+
+	if (!binfo->pfme_hdr) {
+		dev_info(binfo,  "VF is detected.\n");
+		return ret;
+	}
+
+	fme_hdr = binfo->pfme_hdr;
+
+	do {
+		port.csr = readq(&fme_hdr->port[i]);
+		if (!port.port_implemented)
+			break;
+
+		/* skip port which only could be accessed via VF */
+		if (port.afu_access_control == FME_AFU_ACCESS_VF)
+			continue;
+
+		ret = parse_switch_to(binfo, port.port_bar);
+		if (ret)
+			break;
+
+		ret = parse_feature_list(binfo,
+			(u8 __iomem *)binfo->ioaddr + port.port_offset);
+		if (ret)
+			break;
+	} while (++i < MAX_FPGA_PORT_NUM);
+
+	return ret;
+}
+
+static struct build_feature_devs_info *
+build_info_alloc_and_init(struct ifpga_hw *hw)
+{
+	struct build_feature_devs_info *binfo;
+
+	binfo = zmalloc(sizeof(*binfo));
+	if (!binfo)
+		return binfo;
+
+	binfo->hw = hw;
+	binfo->pci_data = hw->pci_data;
+
+	/* fpga feature list starts from BAR 0 */
+	if (parse_switch_to(binfo, 0)) {
+		free(binfo);
+		return NULL;
+	}
+
+	return binfo;
+}
+
+static void build_info_free(struct build_feature_devs_info *binfo)
+{
+	free(binfo);
+}
+
+static void ifpga_print_device_feature_list(struct ifpga_hw *hw)
+{
+	struct ifpga_fme_hw *fme = &hw->fme;
+	struct ifpga_port_hw *port;
+	struct feature *feature;
+	int i, j;
+
+#if 0
+	dev_info(hw, "found a fpga device : %x:%x.%x\n", pci->addr.bus,
+			pci->addr.devid, pci->addr.function);
+#endif
+
+	dev_info(hw, "found fme_device, is in PF: %s\n",
+			is_ifpga_hw_pf(hw) ? "yes":"no");
+
+	for (i = 0; i < FME_FEATURE_ID_MAX; i++) {
+		feature = &fme->sub_feature[i];
+		if (feature->state != IFPGA_FEATURE_ATTACHED)
+			continue;
+
+		dev_info(hw, "%12s:	0x%llx - 0x%llx  - paddr: 0x%lx\n",
+			 feature->name, (unsigned long long)feature->addr,
+			 (unsigned long long)feature->addr + feature->size - 1,
+			 feature->phys_addr);
+	}
+
+	for (i = 0; i < MAX_FPGA_PORT_NUM; i++) {
+		port = &hw->port[i];
+
+		if (port->state != IFPGA_PORT_ATTACHED)
+			continue;
+
+		dev_info(hw, "port device: %d\n", port->port_id);
+
+		for (j = 0; j < PORT_FEATURE_ID_MAX; j++) {
+			feature = &port->sub_feature[j];
+			if (feature->state != IFPGA_FEATURE_ATTACHED)
+				continue;
+
+			dev_info(hw, "%12s:	0x%llx - 0x%llx  - paddr:0x%lx\n",
+				 feature->name,
+				 (unsigned long long)feature->addr,
+				 (unsigned long long)feature->addr +
+				 feature->size - 1,
+				 feature->phys_addr);
+		}
+	}
+}
+
+int ifpga_bus_enumerate(struct ifpga_hw *hw)
+{
+	struct build_feature_devs_info *binfo;
+	int ret;
+
+	binfo = build_info_alloc_and_init(hw);
+	if (!binfo)
+		return -ENOMEM;
+
+	ret = parse_feature_list(binfo, binfo->ioaddr);
+	if (ret)
+		goto exit;
+
+	ret = parse_ports_from_fme(binfo);
+	if (ret)
+		goto exit;
+
+	ifpga_print_device_feature_list(hw);
+
+exit:
+	build_info_free(binfo);
+	return ret;
+}
+
+int ifpga_bus_init(struct ifpga_hw *hw)
+{
+	int i;
+
+	fme_hw_init(&hw->fme);
+	for (i = 0; i < MAX_FPGA_PORT_NUM; i++)
+		port_hw_init(&hw->port[i]);
+
+	return 0;
+}
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h b/drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h
new file mode 100644
index 0000000..0df7b2f
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h
@@ -0,0 +1,38 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _IFPGA_ENUMERATE_H_
+#define _IFPGA_ENUMERATE_H_
+
+int ifpga_bus_init(struct ifpga_hw *hw);
+int ifpga_bus_enumerate(struct ifpga_hw *hw);
+
+#endif /* _IFPGA_ENUMERATE_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c b/drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c
new file mode 100644
index 0000000..6bc8b90
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c
@@ -0,0 +1,340 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 <linux/vfio.h>
+#include <sys/ioctl.h>
+
+#include "ifpga_feature_dev.h"
+
+/*
+ * Enable Port by clear the port soft reset bit, which is set by default.
+ * The AFU is unable to respond to any MMIO access while in reset.
+ * __fpga_port_enable function should only be used after __fpga_port_disable
+ * function.
+ */
+void __fpga_port_enable(struct ifpga_port_hw *port)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_control control;
+
+	WARN_ON(!port->disable_count);
+
+	if (--port->disable_count != 0)
+		return;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+	WARN_ON(!port_hdr);
+
+	control.csr = readq(&port_hdr->control);
+	control.port_sftrst = 0x0;
+	writeq(control.csr, &port_hdr->control);
+}
+
+int __fpga_port_disable(struct ifpga_port_hw *port)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_control control;
+
+	if (port->disable_count++ != 0)
+		return 0;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+	WARN_ON(!port_hdr);
+
+	/* Set port soft reset */
+	control.csr = readq(&port_hdr->control);
+	control.port_sftrst = 0x1;
+	writeq(control.csr, &port_hdr->control);
+
+	/*
+	 * HW sets ack bit to 1 when all outstanding requests have been drained
+	 * on this port and minimum soft reset pulse width has elapsed.
+	 * Driver polls port_soft_reset_ack to determine if reset done by HW.
+	 */
+	control.port_sftrst_ack = 1;
+
+	if (fpga_wait_register_field(port_sftrst_ack, control,
+				     &port_hdr->control, RST_POLL_TIMEOUT,
+				     RST_POLL_INVL)) {
+		dev_err(port, "timeout, fail to reset device\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+int fpga_get_afu_uuid(struct ifpga_port_hw *port, struct uuid *uuid)
+{
+	struct feature_port_header *port_hdr;
+	u64 guidl, guidh;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port, PORT_FEATURE_ID_UAFU);
+
+	spinlock_lock(&port->lock);
+	guidl = readq(&port_hdr->afu_header.guid.b[0]);
+	guidh = readq(&port_hdr->afu_header.guid.b[8]);
+	spinlock_unlock(&port->lock);
+
+	printf("%s: guidl=0x%lx, guidh=0x%lx\n", __func__, guidl, guidh);
+
+	memcpy(uuid->b, &guidl, sizeof(u64));
+	memcpy(uuid->b + 8, &guidh, sizeof(u64));
+
+	return 0;
+}
+
+/* Mask / Unmask Port Errors by the Error Mask register. */
+void port_err_mask(struct ifpga_port_hw *port, bool mask)
+{
+	struct feature_port_error *port_err;
+	struct feature_port_err_key err_mask;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+
+	if (mask)
+		err_mask.csr = PORT_ERR_MASK;
+	else
+		err_mask.csr = 0;
+
+	writeq(err_mask.csr, &port_err->error_mask);
+}
+
+/* Clear All Port Errors. */
+int port_err_clear(struct ifpga_port_hw *port, u64 err)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_error *port_err;
+	struct feature_port_err_key mask;
+	struct feature_port_first_err_key first;
+	struct feature_port_status status;
+	int ret = 0;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	/*
+	 * Clear All Port Errors
+	 *
+	 * - Check for AP6 State
+	 * - Halt Port by keeping Port in reset
+	 * - Set PORT Error mask to all 1 to mask errors
+	 * - Clear all errors
+	 * - Set Port mask to all 0 to enable errors
+	 * - All errors start capturing new errors
+	 * - Enable Port by pulling the port out of reset
+	 */
+
+	/* If device is still in AP6 state, can not clear any error.*/
+	status.csr = readq(&port_hdr->status);
+	if (status.power_state == PORT_POWER_STATE_AP6) {
+		dev_err(dev, "Could not clear errors, device in AP6 state.\n");
+		return -EBUSY;
+	}
+
+	/* Halt Port by keeping Port in reset */
+	ret = __fpga_port_disable(port);
+	if (ret)
+		return ret;
+
+	/* Mask all errors */
+	port_err_mask(port, true);
+
+	/* Clear errors if err input matches with current port errors.*/
+	mask.csr = readq(&port_err->port_error);
+
+	if (mask.csr == err) {
+		writeq(mask.csr, &port_err->port_error);
+
+		first.csr = readq(&port_err->port_first_error);
+		writeq(first.csr, &port_err->port_first_error);
+	} else
+		ret = -EBUSY;
+
+	/* Clear mask */
+	port_err_mask(port, false);
+
+	/* Enable the Port by clear the reset */
+	__fpga_port_enable(port);
+
+	return ret;
+}
+
+int port_clear_error(struct ifpga_port_hw *port)
+{
+	struct feature_port_error *port_err;
+	struct feature_port_err_key error;
+	u64 port_error;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+	port_error = error.csr = readq(&port_err->port_error);
+
+	dev_info(port, "read port error: 0x%lx\n", port_error);
+
+	return port_err_clear(port, port_error);
+}
+
+void fme_hw_uinit(struct ifpga_fme_hw *fme)
+{
+	struct feature *feature;
+	int i;
+
+	if (fme->state != IFPGA_FME_IMPLEMENTED)
+		return;
+
+	for (i = 0; i < FME_FEATURE_ID_MAX; i++) {
+		feature = &fme->sub_feature[i];
+		if (feature->state == IFPGA_FEATURE_ATTACHED &&
+			feature->ops && feature->ops->uinit)
+			feature->ops->uinit(feature);
+	}
+}
+
+int fme_hw_init(struct ifpga_fme_hw *fme)
+{
+	struct feature *feature;
+	int i, ret;
+
+	if (fme->state != IFPGA_FME_IMPLEMENTED)
+		return -EINVAL;
+
+	for (i = 0; i < FME_FEATURE_ID_MAX; i++) {
+		feature = &fme->sub_feature[i];
+		if (feature->state == IFPGA_FEATURE_ATTACHED &&
+			feature->ops && feature->ops->init) {
+			ret = feature->ops->init(feature);
+			if (ret) {
+				fme_hw_uinit(fme);
+				return ret;
+			}
+		}
+	}
+
+	return 0;
+}
+
+void port_hw_uinit(struct ifpga_port_hw *port)
+{
+	struct feature *feature;
+	int i;
+
+	for (i = 0; i < PORT_FEATURE_ID_MAX; i++) {
+		feature = &port->sub_feature[i];
+		if (feature->state == IFPGA_FEATURE_ATTACHED &&
+			feature->ops && feature->ops->uinit)
+			feature->ops->uinit(feature);
+	}
+}
+
+int port_hw_init(struct ifpga_port_hw *port)
+{
+	struct feature *feature;
+	int i, ret;
+
+	if (port->state == IFPGA_PORT_UNUSED)
+		return 0;
+
+	for (i = 0; i < PORT_FEATURE_ID_MAX; i++) {
+		feature = &port->sub_feature[i];
+		if (feature->ops && feature->ops->init)
+			ret = feature->ops->init(feature);
+			if (ret) {
+				port_hw_uinit(port);
+				return ret;
+			}
+	}
+
+	return 0;
+}
+
+/*
+ * FIXME: we should get msix vec count during pci enumeration instead of
+ * below hardcode value.
+ */
+#define FPGA_MSIX_VEC_COUNT	20
+/* irq set buffer length for interrupt */
+#define MSIX_IRQ_SET_BUF_LEN (sizeof(struct vfio_irq_set) + \
+				sizeof(int) * FPGA_MSIX_VEC_COUNT)
+
+/* only support msix for now*/
+static int vfio_msix_enable_block(s32 vfio_dev_fd, unsigned int vec_start,
+				  unsigned int count, s32 *fds)
+{
+	char irq_set_buf[MSIX_IRQ_SET_BUF_LEN];
+	struct vfio_irq_set *irq_set;
+	int len, ret;
+	int *fd_ptr;
+
+	len = sizeof(irq_set_buf);
+
+	irq_set = (struct vfio_irq_set *)irq_set_buf;
+	irq_set->argsz = len;
+	irq_set->count = count;
+	irq_set->flags = VFIO_IRQ_SET_DATA_EVENTFD |
+				VFIO_IRQ_SET_ACTION_TRIGGER;
+	irq_set->index = VFIO_PCI_MSIX_IRQ_INDEX;
+	irq_set->start = vec_start;
+
+	fd_ptr = (int *)&irq_set->data;
+	memcpy(fd_ptr, fds, sizeof(int) * count);
+
+	ret = ioctl(vfio_dev_fd, VFIO_DEVICE_SET_IRQS, irq_set);
+	if (ret)
+		printf("Error enabling MSI-X interrupts\n");
+
+	return ret;
+}
+
+int fpga_msix_set_block(struct feature *feature, unsigned int start,
+			unsigned int count, s32 *fds)
+{
+	struct feature_irq_ctx *ctx = feature->ctx;
+	unsigned int i;
+	int ret;
+
+	if (start >= feature->ctx_num || start + count > feature->ctx_num)
+		return -EINVAL;
+
+	/* assume that each feature has continuous vector space in msix*/
+	ret = vfio_msix_enable_block(feature->vfio_dev_fd,
+				     ctx[start].idx, count, fds);
+	if (!ret) {
+		for (i = 0; i < count; i++)
+			ctx[i].eventfd = fds[i];
+	}
+
+	return ret;
+}
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h b/drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h
new file mode 100644
index 0000000..1dc6eff
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h
@@ -0,0 +1,197 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _IFPGA_FEATURE_DEV_H_
+#define _IFPGA_FEATURE_DEV_H_
+
+#include "ifpga_hw.h"
+
+static inline struct ifpga_port_hw *
+get_port(struct ifpga_hw *hw, u32 port_id)
+{
+	if (!is_valid_port_id(hw, port_id))
+		return NULL;
+
+	return &hw->port[port_id];
+}
+
+#define ifpga_for_each_feature(hw, feature)		\
+	for ((feature) = (hw)->sub_feature;			\
+	   (feature) < (hw)->sub_feature + (FME_FEATURE_ID_MAX); (feature)++)
+
+static inline struct feature *
+get_fme_feature_by_id(struct ifpga_fme_hw *fme, u64 id)
+{
+	struct feature *feature;
+
+	ifpga_for_each_feature(fme, feature) {
+		if (feature->id == id)
+			return feature;
+	}
+
+	return NULL;
+}
+
+static inline struct feature *
+get_port_feature_by_id(struct ifpga_port_hw *port, u64 id)
+{
+	struct feature *feature;
+
+	ifpga_for_each_feature(port, feature) {
+		if (feature->id == id)
+			return feature;
+	}
+
+	return NULL;
+}
+
+static inline void  *
+get_fme_feature_ioaddr_by_index(struct ifpga_fme_hw *fme, int index)
+{
+	return fme->sub_feature[index].addr;
+}
+
+static inline void  *
+get_port_feature_ioaddr_by_index(struct ifpga_port_hw *port, int index)
+{
+	return port->sub_feature[index].addr;
+}
+
+static inline bool
+is_fme_feature_present(struct ifpga_fme_hw *fme, int index)
+{
+	return !!get_fme_feature_ioaddr_by_index(fme, index);
+}
+
+static inline bool
+is_port_feature_present(struct ifpga_port_hw *port, int index)
+{
+	return !!get_port_feature_ioaddr_by_index(port, index);
+}
+
+int fpga_get_afu_uuid(struct ifpga_port_hw *port, struct uuid *uuid);
+
+int __fpga_port_disable(struct ifpga_port_hw *port);
+void __fpga_port_enable(struct ifpga_port_hw *port);
+
+static inline int fpga_port_disable(struct ifpga_port_hw *port)
+{
+	int ret;
+
+	spinlock_lock(&port->lock);
+	ret = __fpga_port_disable(port);
+	spinlock_unlock(&port->lock);
+	return ret;
+}
+
+static inline int fpga_port_enable(struct ifpga_port_hw *port)
+{
+	spinlock_lock(&port->lock);
+	__fpga_port_enable(port);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static inline int __fpga_port_reset(struct ifpga_port_hw *port)
+{
+	int ret;
+
+	ret = __fpga_port_disable(port);
+	if (ret)
+		return ret;
+
+	__fpga_port_enable(port);
+
+	return 0;
+}
+
+static inline int fpga_port_reset(struct ifpga_port_hw *port)
+{
+	int ret;
+
+	spinlock_lock(&port->lock);
+	ret = __fpga_port_reset(port);
+	spinlock_unlock(&port->lock);
+	return ret;
+}
+
+int do_pr(struct ifpga_hw *hw, u32 port_id, void *buffer, u32 size,
+	   u64 *status);
+
+int fme_get_prop(struct ifpga_fme_hw *fme, struct feature_prop *prop);
+int fme_set_prop(struct ifpga_fme_hw *fme, struct feature_prop *prop);
+int fme_set_irq(struct ifpga_fme_hw *fme, u32 feature_id, void *irq_set);
+
+int fme_hw_init(struct ifpga_fme_hw *fme);
+void fme_hw_uinit(struct ifpga_fme_hw *fme);
+void port_hw_uinit(struct ifpga_port_hw *port);
+int port_hw_init(struct ifpga_port_hw *port);
+int port_clear_error(struct ifpga_port_hw *port);
+void port_err_mask(struct ifpga_port_hw *port, bool mask);
+int port_err_clear(struct ifpga_port_hw *port, u64 err);
+
+extern struct feature_ops fme_hdr_ops;
+extern struct feature_ops fme_thermal_mgmt_ops;
+extern struct feature_ops fme_power_mgmt_ops;
+extern struct feature_ops fme_global_err_ops;
+extern struct feature_ops fme_pr_mgmt_ops;
+extern struct feature_ops fme_global_iperf_ops;
+extern struct feature_ops fme_global_dperf_ops;
+
+int port_get_prop(struct ifpga_port_hw *port, struct feature_prop *prop);
+int port_set_prop(struct ifpga_port_hw *port, struct feature_prop *prop);
+
+/* This struct is used when parsing uafu irq_set */
+struct fpga_uafu_irq_set {
+	u32 start;
+	u32 count;
+	s32 *evtfds;
+};
+
+int port_set_irq(struct ifpga_port_hw *port, u32 feature_id, void *irq_set);
+
+extern struct feature_ops port_hdr_ops;
+extern struct feature_ops port_error_ops;
+extern struct feature_ops port_stp_ops;
+extern struct feature_ops port_umsg_ops;
+extern struct feature_ops port_uint_ops;
+
+int afu_port_umsg_enable(struct ifpga_port_hw *port, bool enable);
+int afu_port_umsg_set_mode(struct ifpga_port_hw *port, u32 mode);
+
+/* help functions for feature ops */
+int fpga_msix_set_block(struct feature *feature, unsigned int start,
+			unsigned int count, s32 *fds);
+void port_check_reg(void __iomem *addr, const char *reg_name, u64 dflt);
+
+
+#endif /* _IFPGA_FEATURE_DEV_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_fme.c b/drivers/raw/ifpga_rawdev/base/ifpga_fme.c
new file mode 100644
index 0000000..ecfddfc
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_fme.c
@@ -0,0 +1,763 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+#define PWR_THRESHOLD_MAX       0x7F
+
+int fme_get_prop(struct ifpga_fme_hw *fme, struct feature_prop *prop)
+{
+	struct feature *feature;
+
+	if (!fme)
+		return -ENOENT;
+
+	feature = get_fme_feature_by_id(fme, prop->feature_id);
+
+	if (feature && feature->ops && feature->ops->get_prop)
+		return feature->ops->get_prop(feature, prop);
+
+	return -ENOENT;
+}
+
+int fme_set_prop(struct ifpga_fme_hw *fme, struct feature_prop *prop)
+{
+	struct feature *feature;
+
+	if (!fme)
+		return -ENOENT;
+
+	feature = get_fme_feature_by_id(fme, prop->feature_id);
+
+	if (feature && feature->ops && feature->ops->set_prop)
+		return feature->ops->set_prop(feature, prop);
+
+	return -ENOENT;
+}
+
+int fme_set_irq(struct ifpga_fme_hw *fme, u32 feature_id, void *irq_set)
+{
+	struct feature *feature;
+
+	if (!fme)
+		return -ENOENT;
+
+	feature = get_fme_feature_by_id(fme, feature_id);
+
+	if (feature && feature->ops && feature->ops->set_irq)
+		return feature->ops->set_irq(feature, irq_set);
+
+	return -ENOENT;
+}
+
+/* fme private feature head */
+static int fme_hdr_init(struct feature *feature)
+{
+	struct feature_fme_header *fme_hdr;
+
+	fme_hdr = (struct feature_fme_header *)feature->addr;
+
+	dev_info(NULL, "FME HDR Init.\n");
+	dev_info(NULL, "FME cap %llx.\n",
+		 (unsigned long long)fme_hdr->capability.csr);
+
+	return 0;
+}
+
+static void fme_hdr_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME HDR UInit.\n");
+}
+
+static int fme_hdr_get_revision(struct ifpga_fme_hw *fme, u64 *revision)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+	struct feature_header header;
+
+	header.csr = readq(&fme_hdr->header);
+	*revision = header.revision;
+
+	return 0;
+}
+
+static int fme_hdr_get_ports_num(struct ifpga_fme_hw *fme, u64 *ports_num)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+	struct feature_fme_capability fme_capability;
+
+	fme_capability.csr = readq(&fme_hdr->capability);
+	*ports_num = fme_capability.num_ports;
+
+	return 0;
+}
+
+static int fme_hdr_get_cache_size(struct ifpga_fme_hw *fme, u64 *cache_size)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+	struct feature_fme_capability fme_capability;
+
+	fme_capability.csr = readq(&fme_hdr->capability);
+	*cache_size = fme_capability.cache_size;
+
+	return 0;
+}
+
+static int fme_hdr_get_version(struct ifpga_fme_hw *fme, u64 *version)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+	struct feature_fme_capability fme_capability;
+
+	fme_capability.csr = readq(&fme_hdr->capability);
+	*version = fme_capability.fabric_verid;
+
+	return 0;
+}
+
+static int fme_hdr_get_socket_id(struct ifpga_fme_hw *fme, u64 *socket_id)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+	struct feature_fme_capability fme_capability;
+
+	fme_capability.csr = readq(&fme_hdr->capability);
+	*socket_id = fme_capability.socket_id;
+
+	return 0;
+}
+
+static int fme_hdr_get_bitstream_id(struct ifpga_fme_hw *fme,
+				    u64 *bitstream_id)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+
+	*bitstream_id = readq(&fme_hdr->bitstream_id);
+
+	return 0;
+}
+
+static int fme_hdr_get_bitstream_metadata(struct ifpga_fme_hw *fme,
+					  u64 *bitstream_metadata)
+{
+	struct feature_fme_header *fme_hdr
+		= get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+
+	*bitstream_metadata = readq(&fme_hdr->bitstream_md);
+
+	return 0;
+}
+
+static int
+fme_hdr_get_prop(struct feature *feature, struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+
+	switch (prop->prop_id) {
+	case FME_HDR_PROP_REVISION:
+		return fme_hdr_get_revision(fme, &prop->data);
+	case FME_HDR_PROP_PORTS_NUM:
+		return fme_hdr_get_ports_num(fme, &prop->data);
+	case FME_HDR_PROP_CACHE_SIZE:
+		return fme_hdr_get_cache_size(fme, &prop->data);
+	case FME_HDR_PROP_VERSION:
+		return fme_hdr_get_version(fme, &prop->data);
+	case FME_HDR_PROP_SOCKET_ID:
+		return fme_hdr_get_socket_id(fme, &prop->data);
+	case FME_HDR_PROP_BITSTREAM_ID:
+		return fme_hdr_get_bitstream_id(fme, &prop->data);
+	case FME_HDR_PROP_BITSTREAM_METADATA:
+		return fme_hdr_get_bitstream_metadata(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+struct feature_ops fme_hdr_ops = {
+	.init = fme_hdr_init,
+	.uinit = fme_hdr_uinit,
+	.get_prop = fme_hdr_get_prop,
+};
+
+/* thermal management */
+static int fme_thermal_get_threshold1(struct ifpga_fme_hw *fme, u64 *thres1)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold temp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_threshold.csr = readq(&fme_thermal->threshold);
+	*thres1 = temp_threshold.tmp_thshold1;
+
+	return 0;
+}
+
+static int fme_thermal_set_threshold1(struct ifpga_fme_hw *fme, u64 thres1)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_header *fme_hdr;
+	struct feature_fme_tmp_threshold tmp_threshold;
+	struct feature_fme_capability fme_capability;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_THERMAL_MGMT);
+	fme_hdr = get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+
+	spinlock_lock(&fme->lock);
+	tmp_threshold.csr = readq(&fme_thermal->threshold);
+	fme_capability.csr = readq(&fme_hdr->capability);
+
+	if (fme_capability.lock_bit == 1) {
+		spinlock_unlock(&fme->lock);
+		return -EBUSY;
+	} else if (thres1 > 100) {
+		spinlock_unlock(&fme->lock);
+		return -EINVAL;
+	} else if (thres1 == 0) {
+		tmp_threshold.tmp_thshold1_enable = 0;
+		tmp_threshold.tmp_thshold1 = thres1;
+	} else {
+		tmp_threshold.tmp_thshold1_enable = 1;
+		tmp_threshold.tmp_thshold1 = thres1;
+	}
+
+	writeq(tmp_threshold.csr, &fme_thermal->threshold);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_thermal_get_threshold2(struct ifpga_fme_hw *fme, u64 *thres2)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold temp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_threshold.csr = readq(&fme_thermal->threshold);
+	*thres2 = temp_threshold.tmp_thshold2;
+
+	return 0;
+}
+
+static int fme_thermal_set_threshold2(struct ifpga_fme_hw *fme, u64 thres2)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_header *fme_hdr;
+	struct feature_fme_tmp_threshold tmp_threshold;
+	struct feature_fme_capability fme_capability;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_THERMAL_MGMT);
+	fme_hdr = get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+
+	spinlock_lock(&fme->lock);
+	tmp_threshold.csr = readq(&fme_thermal->threshold);
+	fme_capability.csr = readq(&fme_hdr->capability);
+
+	if (fme_capability.lock_bit == 1) {
+		spinlock_unlock(&fme->lock);
+		return -EBUSY;
+	} else if (thres2 > 100) {
+		spinlock_unlock(&fme->lock);
+		return -EINVAL;
+	} else if (thres2 == 0) {
+		tmp_threshold.tmp_thshold2_enable = 0;
+		tmp_threshold.tmp_thshold2 = thres2;
+	} else {
+		tmp_threshold.tmp_thshold2_enable = 1;
+		tmp_threshold.tmp_thshold2 = thres2;
+	}
+
+	writeq(tmp_threshold.csr, &fme_thermal->threshold);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_thermal_get_threshold_trip(struct ifpga_fme_hw *fme,
+					  u64 *thres_trip)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold temp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_threshold.csr = readq(&fme_thermal->threshold);
+	*thres_trip = temp_threshold.therm_trip_thshold;
+
+	return 0;
+}
+
+static int fme_thermal_get_threshold1_reached(struct ifpga_fme_hw *fme,
+					      u64 *thres1_reached)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold temp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_threshold.csr = readq(&fme_thermal->threshold);
+	*thres1_reached = temp_threshold.thshold1_status;
+
+	return 0;
+}
+
+static int fme_thermal_get_threshold2_reached(struct ifpga_fme_hw *fme,
+					      u64 *thres1_reached)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold temp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_threshold.csr = readq(&fme_thermal->threshold);
+	*thres1_reached = temp_threshold.thshold2_status;
+
+	return 0;
+}
+
+static int fme_thermal_get_threshold1_policy(struct ifpga_fme_hw *fme,
+					     u64 *thres1_policy)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold temp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_threshold.csr = readq(&fme_thermal->threshold);
+	*thres1_policy = temp_threshold.thshold_policy;
+
+	return 0;
+}
+
+static int fme_thermal_set_threshold1_policy(struct ifpga_fme_hw *fme,
+					     u64 thres1_policy)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold tmp_threshold;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_THERMAL_MGMT);
+
+	spinlock_lock(&fme->lock);
+	tmp_threshold.csr = readq(&fme_thermal->threshold);
+
+	if (thres1_policy == 0)
+		tmp_threshold.thshold_policy = 0;
+	else if (thres1_policy == 1)
+		tmp_threshold.thshold_policy = 1;
+	else {
+		spinlock_unlock(&fme->lock);
+		return -EINVAL;
+	}
+
+	writeq(tmp_threshold.csr, &fme_thermal->threshold);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_thermal_get_temperature(struct ifpga_fme_hw *fme, u64 *temp)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_temp_rdsensor_fmt1 temp_rdsensor_fmt1;
+
+	fme_thermal = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_THERMAL_MGMT);
+
+	temp_rdsensor_fmt1.csr = readq(&fme_thermal->rdsensor_fm1);
+	*temp = temp_rdsensor_fmt1.fpga_temp;
+
+	return 0;
+}
+
+static int fme_thermal_get_revision(struct ifpga_fme_hw *fme, u64 *revision)
+{
+	struct feature_fme_thermal *fme_thermal
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_THERMAL_MGMT);
+	struct feature_header header;
+
+	header.csr = readq(&fme_thermal->header);
+	*revision = header.revision;
+
+	return 0;
+}
+
+#define FME_THERMAL_CAP_NO_TMP_THRESHOLD	0x1
+
+static int fme_thermal_mgmt_init(struct feature *feature)
+{
+	struct feature_fme_thermal *fme_thermal;
+	struct feature_fme_tmp_threshold_cap thermal_cap;
+
+	UNUSED(feature);
+
+	dev_info(NULL, "FME thermal mgmt Init.\n");
+
+	fme_thermal = (struct feature_fme_thermal *)feature->addr;
+	thermal_cap.csr = readq(&fme_thermal->threshold_cap);
+
+	dev_info(NULL, "FME thermal cap %llx.\n",
+		 (unsigned long long)fme_thermal->threshold_cap.csr);
+
+	if (thermal_cap.tmp_thshold_disabled)
+		feature->cap |= FME_THERMAL_CAP_NO_TMP_THRESHOLD;
+
+	return 0;
+}
+
+static void fme_thermal_mgmt_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME thermal mgmt UInit.\n");
+}
+
+static int
+fme_thermal_set_prop(struct feature *feature, struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+
+	if (feature->cap & FME_THERMAL_CAP_NO_TMP_THRESHOLD)
+		return -ENOENT;
+
+	switch (prop->prop_id) {
+	case FME_THERMAL_PROP_THRESHOLD1:
+		return fme_thermal_set_threshold1(fme, prop->data);
+	case FME_THERMAL_PROP_THRESHOLD2:
+		return fme_thermal_set_threshold2(fme, prop->data);
+	case FME_THERMAL_PROP_THRESHOLD1_POLICY:
+		return fme_thermal_set_threshold1_policy(fme, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int
+fme_thermal_get_prop(struct feature *feature, struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+
+	if (feature->cap & FME_THERMAL_CAP_NO_TMP_THRESHOLD &&
+	    prop->prop_id != FME_THERMAL_PROP_TEMPERATURE &&
+	    prop->prop_id != FME_THERMAL_PROP_REVISION)
+		return -ENOENT;
+
+	switch (prop->prop_id) {
+	case FME_THERMAL_PROP_THRESHOLD1:
+		return fme_thermal_get_threshold1(fme, &prop->data);
+	case FME_THERMAL_PROP_THRESHOLD2:
+		return fme_thermal_get_threshold2(fme, &prop->data);
+	case FME_THERMAL_PROP_THRESHOLD_TRIP:
+		return fme_thermal_get_threshold_trip(fme, &prop->data);
+	case FME_THERMAL_PROP_THRESHOLD1_REACHED:
+		return fme_thermal_get_threshold1_reached(fme, &prop->data);
+	case FME_THERMAL_PROP_THRESHOLD2_REACHED:
+		return fme_thermal_get_threshold2_reached(fme, &prop->data);
+	case FME_THERMAL_PROP_THRESHOLD1_POLICY:
+		return fme_thermal_get_threshold1_policy(fme, &prop->data);
+	case FME_THERMAL_PROP_TEMPERATURE:
+		return fme_thermal_get_temperature(fme, &prop->data);
+	case FME_THERMAL_PROP_REVISION:
+		return fme_thermal_get_revision(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+struct feature_ops fme_thermal_mgmt_ops = {
+	.init = fme_thermal_mgmt_init,
+	.uinit = fme_thermal_mgmt_uinit,
+	.get_prop = fme_thermal_get_prop,
+	.set_prop = fme_thermal_set_prop,
+};
+
+
+static int fme_pwr_get_consumed(struct ifpga_fme_hw *fme, u64 *consumed)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_status pm_status;
+
+	pm_status.csr = readq(&fme_power->status);
+
+	*consumed = pm_status.pwr_consumed;
+
+	return 0;
+}
+
+static int fme_pwr_get_threshold1(struct ifpga_fme_hw *fme, u64 *threshold)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_ap_threshold pm_ap_threshold;
+
+	pm_ap_threshold.csr = readq(&fme_power->threshold);
+
+	*threshold = pm_ap_threshold.threshold1;
+
+	return 0;
+}
+
+static int fme_pwr_set_threshold1(struct ifpga_fme_hw *fme, u64 threshold)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_ap_threshold pm_ap_threshold;
+
+	spinlock_lock(&fme->lock);
+	pm_ap_threshold.csr = readq(&fme_power->threshold);
+
+	if (threshold <= PWR_THRESHOLD_MAX)
+		pm_ap_threshold.threshold1 = threshold;
+	else {
+		spinlock_unlock(&fme->lock);
+		return -EINVAL;
+	}
+
+	writeq(pm_ap_threshold.csr, &fme_power->threshold);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_pwr_get_threshold2(struct ifpga_fme_hw *fme, u64 *threshold)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_ap_threshold pm_ap_threshold;
+
+	pm_ap_threshold.csr = readq(&fme_power->threshold);
+
+	*threshold = pm_ap_threshold.threshold2;
+
+	return 0;
+}
+
+static int fme_pwr_set_threshold2(struct ifpga_fme_hw *fme, u64 threshold)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_ap_threshold pm_ap_threshold;
+
+	spinlock_lock(&fme->lock);
+	pm_ap_threshold.csr = readq(&fme_power->threshold);
+
+	if (threshold <= PWR_THRESHOLD_MAX)
+		pm_ap_threshold.threshold2 = threshold;
+	else {
+		spinlock_unlock(&fme->lock);
+		return -EINVAL;
+	}
+
+	writeq(pm_ap_threshold.csr, &fme_power->threshold);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_pwr_get_threshold1_status(struct ifpga_fme_hw *fme,
+					 u64 *threshold_status)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_ap_threshold pm_ap_threshold;
+
+	pm_ap_threshold.csr = readq(&fme_power->threshold);
+
+	*threshold_status = pm_ap_threshold.threshold1_status;
+
+	return 0;
+}
+
+static int fme_pwr_get_threshold2_status(struct ifpga_fme_hw *fme,
+					 u64 *threshold_status)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_ap_threshold pm_ap_threshold;
+
+	pm_ap_threshold.csr = readq(&fme_power->threshold);
+
+	*threshold_status = pm_ap_threshold.threshold2_status;
+
+	return 0;
+}
+
+static int fme_pwr_get_rtl(struct ifpga_fme_hw *fme, u64 *rtl)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_status pm_status;
+
+	pm_status.csr = readq(&fme_power->status);
+
+	*rtl = pm_status.fpga_latency_report;
+
+	return 0;
+}
+
+static int fme_pwr_get_xeon_limit(struct ifpga_fme_hw *fme, u64 *limit)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_xeon_limit xeon_limit;
+
+	xeon_limit.csr = readq(&fme_power->xeon_limit);
+
+	if (!xeon_limit.enable)
+		xeon_limit.pwr_limit = 0;
+
+	*limit = xeon_limit.pwr_limit;
+
+	return 0;
+
+}
+
+static int fme_pwr_get_fpga_limit(struct ifpga_fme_hw *fme, u64 *limit)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_POWER_MGMT);
+	struct feature_fme_pm_fpga_limit fpga_limit;
+
+	fpga_limit.csr = readq(&fme_power->fpga_limit);
+
+	if (!fpga_limit.enable)
+		fpga_limit.pwr_limit = 0;
+
+	*limit = fpga_limit.pwr_limit;
+
+	return 0;
+}
+
+static int fme_pwr_get_revision(struct ifpga_fme_hw *fme, u64 *revision)
+{
+	struct feature_fme_power *fme_power
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_POWER_MGMT);
+	struct feature_header header;
+
+	header.csr = readq(&fme_power->header);
+	*revision = header.revision;
+
+	return 0;
+}
+
+static int fme_power_mgmt_init(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME power mgmt Init.\n");
+
+	return 0;
+}
+
+static void fme_power_mgmt_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME power mgmt UInit.\n");
+}
+
+static int fme_power_mgmt_get_prop(struct feature *feature,
+				   struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+
+	switch (prop->prop_id) {
+	case FME_PWR_PROP_CONSUMED:
+		return fme_pwr_get_consumed(fme, &prop->data);
+	case FME_PWR_PROP_THRESHOLD1:
+		return fme_pwr_get_threshold1(fme, &prop->data);
+	case FME_PWR_PROP_THRESHOLD2:
+		return fme_pwr_get_threshold2(fme, &prop->data);
+	case FME_PWR_PROP_THRESHOLD1_STATUS:
+		return fme_pwr_get_threshold1_status(fme, &prop->data);
+	case FME_PWR_PROP_THRESHOLD2_STATUS:
+		return fme_pwr_get_threshold2_status(fme, &prop->data);
+	case FME_PWR_PROP_RTL:
+		return fme_pwr_get_rtl(fme, &prop->data);
+	case FME_PWR_PROP_XEON_LIMIT:
+		return fme_pwr_get_xeon_limit(fme, &prop->data);
+	case FME_PWR_PROP_FPGA_LIMIT:
+		return fme_pwr_get_fpga_limit(fme, &prop->data);
+	case FME_PWR_PROP_REVISION:
+		return fme_pwr_get_revision(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_power_mgmt_set_prop(struct feature *feature,
+				   struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+
+	switch (prop->prop_id) {
+	case FME_PWR_PROP_THRESHOLD1:
+		return fme_pwr_set_threshold1(fme, prop->data);
+	case FME_PWR_PROP_THRESHOLD2:
+		return fme_pwr_set_threshold2(fme, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+struct feature_ops fme_power_mgmt_ops = {
+	.init = fme_power_mgmt_init,
+	.uinit = fme_power_mgmt_uinit,
+	.get_prop = fme_power_mgmt_get_prop,
+	.set_prop = fme_power_mgmt_set_prop,
+};
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c b/drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c
new file mode 100644
index 0000000..dbbd564
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c
@@ -0,0 +1,328 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+#define PERF_OBJ_ROOT_ID	0xff
+
+static int fme_dperf_get_clock(struct ifpga_fme_hw *fme, u64 *clock)
+{
+	struct feature_fme_dperf *dperf;
+	struct feature_fme_dfpmon_clk_ctr clk;
+
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+	clk.afu_interf_clock = readq(&dperf->clk);
+
+	*clock = clk.afu_interf_clock;
+	return 0;
+}
+
+static int fme_dperf_get_revision(struct ifpga_fme_hw *fme, u64 *revision)
+{
+	struct feature_fme_dperf *dperf;
+	struct feature_header header;
+
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+	header.csr = readq(&dperf->header);
+	*revision = header.revision;
+
+	return 0;
+}
+
+#define DPERF_TIMEOUT	30
+
+static bool fabric_pobj_is_enabled(int port_id,
+				   struct feature_fme_dperf *dperf)
+{
+	struct feature_fme_dfpmon_fab_ctl ctl;
+
+	ctl.csr = readq(&dperf->fab_ctl);
+
+	if (ctl.port_filter == FAB_DISABLE_FILTER)
+		return port_id == PERF_OBJ_ROOT_ID;
+
+	return port_id == ctl.port_id;
+}
+
+static u64 read_fabric_counter(struct ifpga_fme_hw *fme, u8 port_id,
+			       enum dperf_fab_events fab_event)
+{
+	struct feature_fme_dfpmon_fab_ctl ctl;
+	struct feature_fme_dfpmon_fab_ctr ctr;
+	struct feature_fme_dperf *dperf;
+	u64 counter = 0;
+
+	spinlock_lock(&fme->lock);
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+
+	/* if it is disabled, force the counter to return zero. */
+	if (!fabric_pobj_is_enabled(port_id, dperf))
+		goto exit;
+
+	ctl.csr = readq(&dperf->fab_ctl);
+	ctl.fab_evtcode = fab_event;
+	writeq(ctl.csr, &dperf->fab_ctl);
+
+	ctr.event_code = fab_event;
+
+	if (fpga_wait_register_field(event_code, ctr,
+				     &dperf->fab_ctr, DPERF_TIMEOUT, 1)) {
+		dev_err(fme, "timeout, unmatched VTd event type in counter registers.\n");
+		spinlock_unlock(&fme->lock);
+		return -ETIMEDOUT;
+	}
+
+	ctr.csr = readq(&dperf->fab_ctr);
+	counter = ctr.fab_cnt;
+exit:
+	spinlock_unlock(&fme->lock);
+	return counter;
+}
+
+#define FAB_PORT_SHOW(name, event)					\
+static int fme_dperf_get_fab_port_##name(struct ifpga_fme_hw *fme,	\
+					 u8 port_id, u64 *counter)	\
+{									\
+	*counter = read_fabric_counter(fme, port_id, event);		\
+	return 0;							\
+}
+
+FAB_PORT_SHOW(pcie0_read, DPERF_FAB_PCIE0_RD);
+FAB_PORT_SHOW(pcie0_write, DPERF_FAB_PCIE0_WR);
+FAB_PORT_SHOW(mmio_read, DPERF_FAB_MMIO_RD);
+FAB_PORT_SHOW(mmio_write, DPERF_FAB_MMIO_WR);
+
+static int fme_dperf_get_fab_port_enable(struct ifpga_fme_hw *fme,
+					 u8 port_id, u64 *enable)
+{
+	struct feature_fme_dperf *dperf;
+	int status;
+
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+
+	status = fabric_pobj_is_enabled(port_id, dperf);
+	*enable = (u64)status;
+
+	return 0;
+}
+
+/*
+ * If enable one port or all port event counter in fabric, other
+ * fabric event counter originally enabled will be disable automatically.
+ */
+static int fme_dperf_set_fab_port_enable(struct ifpga_fme_hw *fme,
+					 u8 port_id, u64 enable)
+{
+	struct feature_fme_dfpmon_fab_ctl ctl;
+	struct feature_fme_dperf *dperf;
+	bool state;
+
+	state = !!enable;
+
+	if (!state)
+		return -EINVAL;
+
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+
+	/* if it is already enabled. */
+	if (fabric_pobj_is_enabled(port_id, dperf))
+		return 0;
+
+	spinlock_lock(&fme->lock);
+	ctl.csr = readq(&dperf->fab_ctl);
+	if (port_id == PERF_OBJ_ROOT_ID)
+		ctl.port_filter = FAB_DISABLE_FILTER;
+	else {
+		ctl.port_filter = FAB_ENABLE_FILTER;
+		ctl.port_id = port_id;
+	}
+
+	writeq(ctl.csr, &dperf->fab_ctl);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_dperf_get_fab_freeze(struct ifpga_fme_hw *fme, u64 *freeze)
+{
+	struct feature_fme_dperf *dperf;
+	struct feature_fme_dfpmon_fab_ctl ctl;
+
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+	ctl.csr = readq(&dperf->fab_ctl);
+	*freeze = (u64)ctl.freeze;
+
+	return 0;
+}
+
+static int fme_dperf_set_fab_freeze(struct ifpga_fme_hw *fme, u64 freeze)
+{
+	struct feature_fme_dperf *dperf;
+	struct feature_fme_dfpmon_fab_ctl ctl;
+	bool state;
+
+	state = !!freeze;
+
+	spinlock_lock(&fme->lock);
+	dperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_DPERF);
+	ctl.csr = readq(&dperf->fab_ctl);
+	ctl.freeze = state;
+	writeq(ctl.csr, &dperf->fab_ctl);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+#define PERF_MAX_PORT_NUM	1
+
+static int fme_global_dperf_init(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME global_dperf Init.\n");
+
+	return 0;
+}
+
+static void fme_global_dperf_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME global_dperf UInit.\n");
+}
+
+static int fme_dperf_fab_get_prop(struct feature *feature,
+				  struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x1: /* FREEZE */
+		return fme_dperf_get_fab_freeze(fme, &prop->data);
+	case 0x2: /* PCIE0_READ */
+		return fme_dperf_get_fab_port_pcie0_read(fme, sub, &prop->data);
+	case 0x3: /* PCIE0_WRITE */
+		return fme_dperf_get_fab_port_pcie0_write(fme, sub,
+							  &prop->data);
+	case 0x4: /* MMIO_READ */
+		return fme_dperf_get_fab_port_mmio_read(fme, sub, &prop->data);
+	case 0x5: /* MMIO_WRITE */
+		return fme_dperf_get_fab_port_mmio_write(fme, sub, &prop->data);
+	case 0x6: /* ENABLE */
+		return fme_dperf_get_fab_port_enable(fme, sub, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_dperf_root_get_prop(struct feature *feature,
+				   struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	if (sub != PERF_PROP_SUB_UNUSED)
+		return -ENOENT;
+
+	switch (id) {
+	case 0x1: /* CLOCK */
+		return fme_dperf_get_clock(fme, &prop->data);
+	case 0x2: /* REVISION */
+		return fme_dperf_get_revision(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_dperf_get_prop(struct feature *feature,
+				     struct feature_prop *prop)
+{
+	u8 top = GET_FIELD(PROP_TOP, prop->prop_id);
+
+	switch (top) {
+	case PERF_PROP_TOP_FAB:
+		return fme_dperf_fab_get_prop(feature, prop);
+	case PERF_PROP_TOP_UNUSED:
+		return fme_dperf_root_get_prop(feature, prop);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_dperf_fab_set_prop(struct feature *feature,
+				  struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x1: /* FREEZE - fab root only prop */
+		if (sub != PERF_PROP_SUB_UNUSED)
+			return -ENOENT;
+		return fme_dperf_set_fab_freeze(fme, prop->data);
+	case 0x6: /* ENABLE - fab both root and sub */
+		return fme_dperf_set_fab_port_enable(fme, sub, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_dperf_set_prop(struct feature *feature,
+				     struct feature_prop *prop)
+{
+	u8 top = GET_FIELD(PROP_TOP, prop->prop_id);
+
+	switch (top) {
+	case PERF_PROP_TOP_FAB:
+		return fme_dperf_fab_set_prop(feature, prop);
+	}
+
+	return -ENOENT;
+}
+
+struct feature_ops fme_global_dperf_ops = {
+	.init = fme_global_dperf_init,
+	.uinit = fme_global_dperf_uinit,
+	.get_prop = fme_global_dperf_get_prop,
+	.set_prop = fme_global_dperf_set_prop,
+
+};
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c b/drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c
new file mode 100644
index 0000000..a8a6a3c
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c
@@ -0,0 +1,430 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+static int fme_err_get_errors(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_error0 fme_error0;
+
+	fme_error0.csr = readq(&fme_err->fme_err);
+	*val = fme_error0.csr;
+
+	return 0;
+}
+
+static int fme_err_get_first_error(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_first_error fme_first_err;
+
+	fme_first_err.csr = readq(&fme_err->fme_first_err);
+	*val = fme_first_err.err_reg_status;
+
+	return 0;
+}
+
+static int fme_err_get_next_error(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_next_error fme_next_err;
+
+	fme_next_err.csr = readq(&fme_err->fme_next_err);
+	*val = fme_next_err.err_reg_status;
+
+	return 0;
+}
+
+static int fme_err_set_clear(struct ifpga_fme_hw *fme, u64 val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_error0 fme_error0;
+	struct feature_fme_first_error fme_first_err;
+	struct feature_fme_next_error fme_next_err;
+	int ret = 0;
+
+	spinlock_lock(&fme->lock);
+	writeq(FME_ERROR0_MASK, &fme_err->fme_err_mask);
+
+	fme_error0.csr = readq(&fme_err->fme_err);
+	if (val != fme_error0.csr) {
+		ret = -EBUSY;
+		goto exit;
+	}
+
+	fme_first_err.csr = readq(&fme_err->fme_first_err);
+	fme_next_err.csr = readq(&fme_err->fme_next_err);
+
+	writeq(fme_error0.csr & FME_ERROR0_MASK, &fme_err->fme_err);
+	writeq(fme_first_err.csr & FME_FIRST_ERROR_MASK,
+	       &fme_err->fme_first_err);
+	writeq(fme_next_err.csr & FME_NEXT_ERROR_MASK,
+	       &fme_err->fme_next_err);
+
+exit:
+	writeq(FME_ERROR0_MASK_DEFAULT, &fme_err->fme_err_mask);
+	spinlock_unlock(&fme->lock);
+
+	return ret;
+}
+
+static int fme_err_get_revision(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_header header;
+
+	header.csr = readq(&fme_err->header);
+	*val = header.revision;
+
+	return 0;
+}
+
+static int fme_err_get_pcie0_errors(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_pcie0_error pcie0_err;
+
+	pcie0_err.csr = readq(&fme_err->pcie0_err);
+	*val = pcie0_err.csr;
+
+	return 0;
+}
+
+static int fme_err_set_pcie0_errors(struct ifpga_fme_hw *fme, u64 val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_pcie0_error pcie0_err;
+	int ret = 0;
+
+	spinlock_lock(&fme->lock);
+	writeq(FME_PCIE0_ERROR_MASK, &fme_err->pcie0_err_mask);
+
+	pcie0_err.csr = readq(&fme_err->pcie0_err);
+	if (val != pcie0_err.csr)
+		ret = -EBUSY;
+	else
+		writeq(pcie0_err.csr & FME_PCIE0_ERROR_MASK,
+		       &fme_err->pcie0_err);
+
+	writeq(0UL, &fme_err->pcie0_err_mask);
+	spinlock_unlock(&fme->lock);
+
+	return ret;
+}
+
+static int fme_err_get_pcie1_errors(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_pcie1_error pcie1_err;
+
+	pcie1_err.csr = readq(&fme_err->pcie1_err);
+	*val = pcie1_err.csr;
+
+	return 0;
+}
+
+static int fme_err_set_pcie1_errors(struct ifpga_fme_hw *fme, u64 val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_pcie1_error pcie1_err;
+	int ret = 0;
+
+	spinlock_lock(&fme->lock);
+	writeq(FME_PCIE1_ERROR_MASK, &fme_err->pcie1_err_mask);
+
+	pcie1_err.csr = readq(&fme_err->pcie1_err);
+	if (val != pcie1_err.csr)
+		ret = -EBUSY;
+	else
+		writeq(pcie1_err.csr & FME_PCIE1_ERROR_MASK,
+		       &fme_err->pcie1_err);
+
+	writeq(0UL, &fme_err->pcie1_err_mask);
+	spinlock_unlock(&fme->lock);
+
+	return ret;
+}
+
+static int fme_err_get_nonfatal_errors(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_ras_nonfaterror ras_nonfaterr;
+
+	ras_nonfaterr.csr = readq(&fme_err->ras_nonfaterr);
+	*val = ras_nonfaterr.csr;
+
+	return 0;
+}
+
+static int fme_err_get_catfatal_errors(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_ras_catfaterror ras_catfaterr;
+
+	ras_catfaterr.csr = readq(&fme_err->ras_catfaterr);
+	*val = ras_catfaterr.csr;
+
+	return 0;
+}
+
+static int fme_err_get_inject_errors(struct ifpga_fme_hw *fme, u64 *val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_ras_error_inj ras_error_inj;
+
+	ras_error_inj.csr = readq(&fme_err->ras_error_inj);
+	*val = ras_error_inj.csr & FME_RAS_ERROR_INJ_MASK;
+
+	return 0;
+}
+
+static int fme_err_set_inject_errors(struct ifpga_fme_hw *fme, u64 val)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+					      FME_FEATURE_ID_GLOBAL_ERR);
+	struct feature_fme_ras_error_inj ras_error_inj;
+
+	spinlock_lock(&fme->lock);
+	ras_error_inj.csr = readq(&fme_err->ras_error_inj);
+
+	if (val <= FME_RAS_ERROR_INJ_MASK) {
+		ras_error_inj.csr = val;
+	} else {
+		spinlock_unlock(&fme->lock);
+		return -EINVAL;
+	}
+
+	writeq(ras_error_inj.csr, &fme_err->ras_error_inj);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static void fme_error_enable(struct ifpga_fme_hw *fme)
+{
+	struct feature_fme_err *fme_err
+		= get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_GLOBAL_ERR);
+
+	writeq(FME_ERROR0_MASK_DEFAULT, &fme_err->fme_err_mask);
+	writeq(0UL, &fme_err->pcie0_err_mask);
+	writeq(0UL, &fme_err->pcie1_err_mask);
+	writeq(0UL, &fme_err->ras_nonfat_mask);
+	writeq(0UL, &fme_err->ras_catfat_mask);
+}
+
+static int fme_global_error_init(struct feature *feature)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+
+	fme_error_enable(fme);
+
+	if (feature->ctx_num)
+		fme->capability |= FPGA_FME_CAP_ERR_IRQ;
+
+	return 0;
+}
+
+static void fme_global_error_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+}
+
+static int fme_err_fme_err_get_prop(struct feature *feature,
+				    struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x1: /* ERRORS */
+		return fme_err_get_errors(fme, &prop->data);
+	case 0x2: /* FIRST_ERROR */
+		return fme_err_get_first_error(fme, &prop->data);
+	case 0x3: /* NEXT_ERROR */
+		return fme_err_get_next_error(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_err_root_get_prop(struct feature *feature,
+				 struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x5: /* REVISION */
+		return fme_err_get_revision(fme, &prop->data);
+	case 0x6: /* PCIE0_ERRORS */
+		return fme_err_get_pcie0_errors(fme, &prop->data);
+	case 0x7: /* PCIE1_ERRORS */
+		return fme_err_get_pcie1_errors(fme, &prop->data);
+	case 0x8: /* NONFATAL_ERRORS */
+		return fme_err_get_nonfatal_errors(fme, &prop->data);
+	case 0x9: /* CATFATAL_ERRORS */
+		return fme_err_get_catfatal_errors(fme, &prop->data);
+	case 0xa: /* INJECT_ERRORS */
+		return fme_err_get_inject_errors(fme, &prop->data);
+	case 0xb: /* REVISION*/
+		return fme_err_get_revision(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_error_get_prop(struct feature *feature,
+				     struct feature_prop *prop)
+{
+	u8 top = GET_FIELD(PROP_TOP, prop->prop_id);
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+
+	/* PROP_SUB is never used */
+	if (sub != PROP_SUB_UNUSED)
+		return -ENOENT;
+
+	switch (top) {
+	case ERR_PROP_TOP_FME_ERR:
+		return fme_err_fme_err_get_prop(feature, prop);
+	case ERR_PROP_TOP_UNUSED:
+		return fme_err_root_get_prop(feature, prop);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_err_fme_err_set_prop(struct feature *feature,
+				    struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x4: /* CLEAR */
+		return fme_err_set_clear(fme, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_err_root_set_prop(struct feature *feature,
+				 struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x6: /* PCIE0_ERRORS */
+		return fme_err_set_pcie0_errors(fme, prop->data);
+	case 0x7: /* PCIE1_ERRORS */
+		return fme_err_set_pcie1_errors(fme, prop->data);
+	case 0xa: /* INJECT_ERRORS */
+		return fme_err_set_inject_errors(fme, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_error_set_prop(struct feature *feature,
+				     struct feature_prop *prop)
+{
+	u8 top = GET_FIELD(PROP_TOP, prop->prop_id);
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+
+	/* PROP_SUB is never used */
+	if (sub != PROP_SUB_UNUSED)
+		return -ENOENT;
+
+	switch (top) {
+	case ERR_PROP_TOP_FME_ERR:
+		return fme_err_fme_err_set_prop(feature, prop);
+	case ERR_PROP_TOP_UNUSED:
+		return fme_err_root_set_prop(feature, prop);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_err_set_irq(struct feature *feature, void *irq_set)
+{
+	struct fpga_fme_err_irq_set *err_irq_set =
+			(struct fpga_fme_err_irq_set *)irq_set;
+	struct ifpga_fme_hw *fme;
+	int ret;
+
+	fme = (struct ifpga_fme_hw *)feature->parent;
+
+	spinlock_lock(&fme->lock);
+	if (!(fme->capability & FPGA_FME_CAP_ERR_IRQ)) {
+		spinlock_unlock(&fme->lock);
+		return -ENODEV;
+	}
+
+	ret = fpga_msix_set_block(feature, 0, 1, &err_irq_set->evtfd);
+	spinlock_unlock(&fme->lock);
+
+	return ret;
+}
+
+struct feature_ops fme_global_err_ops = {
+	.init = fme_global_error_init,
+	.uinit = fme_global_error_uinit,
+	.get_prop = fme_global_error_get_prop,
+	.set_prop = fme_global_error_set_prop,
+	.set_irq = fme_global_err_set_irq,
+};
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c b/drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c
new file mode 100644
index 0000000..33c3598
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c
@@ -0,0 +1,742 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+#define PERF_OBJ_ROOT_ID	0xff
+
+static int fme_iperf_get_clock(struct ifpga_fme_hw *fme, u64 *clock)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_fme_ifpmon_clk_ctr clk;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+				FME_FEATURE_ID_GLOBAL_IPERF);
+	clk.afu_interf_clock = readq(&iperf->clk);
+
+	*clock = clk.afu_interf_clock;
+	return 0;
+}
+
+static int fme_iperf_get_revision(struct ifpga_fme_hw *fme, u64 *revision)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_header header;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+						FME_FEATURE_ID_GLOBAL_IPERF);
+	header.csr = readq(&iperf->header);
+	*revision = header.revision;
+
+	return 0;
+}
+
+static int fme_iperf_get_cache_freeze(struct ifpga_fme_hw *fme, u64 *freeze)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_fme_ifpmon_ch_ctl ctl;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->ch_ctl);
+	*freeze = (u64)ctl.freeze;
+	return 0;
+}
+
+static int fme_iperf_set_cache_freeze(struct ifpga_fme_hw *fme, u64 freeze)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_fme_ifpmon_ch_ctl ctl;
+	bool state;
+
+	state = !!freeze;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->ch_ctl);
+	ctl.freeze = state;
+	writeq(ctl.csr, &iperf->ch_ctl);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+#define IPERF_TIMEOUT	30
+
+static u64 read_cache_counter(struct ifpga_fme_hw *fme,
+				  u8 channel, enum iperf_cache_events event)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_fme_ifpmon_ch_ctl ctl;
+	struct feature_fme_ifpmon_ch_ctr ctr0, ctr1;
+	u64 counter;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_GLOBAL_IPERF);
+
+	/* set channel access type and cache event code. */
+	ctl.csr = readq(&iperf->ch_ctl);
+	ctl.cci_chsel = channel;
+	ctl.cache_event = event;
+	writeq(ctl.csr, &iperf->ch_ctl);
+
+	/* check the event type in the counter registers */
+	ctr0.event_code = event;
+
+	if (fpga_wait_register_field(event_code, ctr0,
+				     &iperf->ch_ctr0, IPERF_TIMEOUT, 1)) {
+		dev_err(fme, "timeout, unmatched cache event type in counter registers.\n");
+		spinlock_unlock(&fme->lock);
+		return -ETIMEDOUT;
+	}
+
+	ctr0.csr = readq(&iperf->ch_ctr0);
+	ctr1.csr = readq(&iperf->ch_ctr1);
+	counter = ctr0.cache_counter + ctr1.cache_counter;
+	spinlock_unlock(&fme->lock);
+
+	return counter;
+}
+
+#define CACHE_SHOW(name, type, event)					\
+static int fme_iperf_get_cache_##name(struct ifpga_fme_hw *fme,		\
+					u64 *counter)			\
+{									\
+	*counter = read_cache_counter(fme, type, event);		\
+	return 0;							\
+}
+
+CACHE_SHOW(read_hit, CACHE_CHANNEL_RD, IPERF_CACHE_RD_HIT);
+CACHE_SHOW(read_miss, CACHE_CHANNEL_RD, IPERF_CACHE_RD_MISS);
+CACHE_SHOW(write_hit, CACHE_CHANNEL_WR, IPERF_CACHE_WR_HIT);
+CACHE_SHOW(write_miss, CACHE_CHANNEL_WR, IPERF_CACHE_WR_MISS);
+CACHE_SHOW(hold_request, CACHE_CHANNEL_RD, IPERF_CACHE_HOLD_REQ);
+CACHE_SHOW(tx_req_stall, CACHE_CHANNEL_RD, IPERF_CACHE_TX_REQ_STALL);
+CACHE_SHOW(rx_req_stall, CACHE_CHANNEL_RD, IPERF_CACHE_RX_REQ_STALL);
+CACHE_SHOW(rx_eviction, CACHE_CHANNEL_RD, IPERF_CACHE_EVICTIONS);
+CACHE_SHOW(data_write_port_contention, CACHE_CHANNEL_WR,
+	   IPERF_CACHE_DATA_WR_PORT_CONTEN);
+CACHE_SHOW(tag_write_port_contention, CACHE_CHANNEL_WR,
+	   IPERF_CACHE_TAG_WR_PORT_CONTEN);
+
+static int fme_iperf_get_vtd_freeze(struct ifpga_fme_hw *fme, u64 *freeze)
+{
+	struct feature_fme_ifpmon_vtd_ctl ctl;
+	struct feature_fme_iperf *iperf;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->vtd_ctl);
+	*freeze = (u64)ctl.freeze;
+
+	return 0;
+}
+
+static int fme_iperf_set_vtd_freeze(struct ifpga_fme_hw *fme, u64 freeze)
+{
+	struct feature_fme_ifpmon_vtd_ctl ctl;
+	struct feature_fme_iperf *iperf;
+	bool state;
+
+	state = !!freeze;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->vtd_ctl);
+	ctl.freeze = state;
+	writeq(ctl.csr, &iperf->vtd_ctl);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static u64 read_iommu_sip_counter(struct ifpga_fme_hw *fme,
+				      enum iperf_vtd_sip_events event)
+{
+	struct feature_fme_ifpmon_vtd_sip_ctl sip_ctl;
+	struct feature_fme_ifpmon_vtd_sip_ctr sip_ctr;
+	struct feature_fme_iperf *iperf;
+	u64 counter;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+	sip_ctl.csr = readq(&iperf->vtd_sip_ctl);
+	sip_ctl.vtd_evtcode = event;
+	writeq(sip_ctl.csr, &iperf->vtd_sip_ctl);
+
+	sip_ctr.event_code = event;
+
+	if (fpga_wait_register_field(event_code, sip_ctr,
+				     &iperf->vtd_sip_ctr, IPERF_TIMEOUT, 1)) {
+		dev_err(fme, "timeout, unmatched VTd SIP event type in counter registers\n");
+		spinlock_unlock(&fme->lock);
+		return -ETIMEDOUT;
+	}
+
+	sip_ctr.csr = readq(&iperf->vtd_sip_ctr);
+	counter = sip_ctr.vtd_counter;
+	spinlock_unlock(&fme->lock);
+
+	return counter;
+}
+
+#define VTD_SIP_SHOW(name, event)					\
+static int fme_iperf_get_vtd_sip_##name(struct ifpga_fme_hw *fme,	\
+						u64 *counter)		\
+{									\
+	*counter = read_iommu_sip_counter(fme, event);			\
+	return 0;							\
+}
+
+VTD_SIP_SHOW(iotlb_4k_hit, IPERF_VTD_SIP_IOTLB_4K_HIT);
+VTD_SIP_SHOW(iotlb_2m_hit, IPERF_VTD_SIP_IOTLB_2M_HIT);
+VTD_SIP_SHOW(iotlb_1g_hit, IPERF_VTD_SIP_IOTLB_1G_HIT);
+VTD_SIP_SHOW(slpwc_l3_hit, IPERF_VTD_SIP_SLPWC_L3_HIT);
+VTD_SIP_SHOW(slpwc_l4_hit, IPERF_VTD_SIP_SLPWC_L4_HIT);
+VTD_SIP_SHOW(rcc_hit, IPERF_VTD_SIP_RCC_HIT);
+VTD_SIP_SHOW(iotlb_4k_miss, IPERF_VTD_SIP_IOTLB_4K_MISS);
+VTD_SIP_SHOW(iotlb_2m_miss, IPERF_VTD_SIP_IOTLB_2M_MISS);
+VTD_SIP_SHOW(iotlb_1g_miss, IPERF_VTD_SIP_IOTLB_1G_MISS);
+VTD_SIP_SHOW(slpwc_l3_miss, IPERF_VTD_SIP_SLPWC_L3_MISS);
+VTD_SIP_SHOW(slpwc_l4_miss, IPERF_VTD_SIP_SLPWC_L4_MISS);
+VTD_SIP_SHOW(rcc_miss, IPERF_VTD_SIP_RCC_MISS);
+
+static u64 read_iommu_counter(struct ifpga_fme_hw *fme, u8 port_id,
+			      enum iperf_vtd_events base_event)
+{
+	struct feature_fme_ifpmon_vtd_ctl ctl;
+	struct feature_fme_ifpmon_vtd_ctr ctr;
+	struct feature_fme_iperf *iperf;
+	enum iperf_vtd_events event = base_event + port_id;
+	u64 counter;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->vtd_ctl);
+	ctl.vtd_evtcode = event;
+	writeq(ctl.csr, &iperf->vtd_ctl);
+
+	ctr.event_code = event;
+
+	if (fpga_wait_register_field(event_code, ctr,
+				     &iperf->vtd_ctr, IPERF_TIMEOUT, 1)) {
+		dev_err(fme, "timeout, unmatched VTd event type in counter registers.\n");
+		spinlock_unlock(&fme->lock);
+		return -ETIMEDOUT;
+	}
+
+	ctr.csr = readq(&iperf->vtd_ctr);
+	counter = ctr.vtd_counter;
+	spinlock_unlock(&fme->lock);
+
+	return counter;
+}
+
+#define VTD_PORT_SHOW(name, base_event)					\
+static int fme_iperf_get_vtd_port_##name(struct ifpga_fme_hw *fme,	\
+				u8 port_id, u64 *counter)		\
+{									\
+	*counter = read_iommu_counter(fme, port_id, base_event);	\
+	return 0;							\
+}
+
+VTD_PORT_SHOW(read_transaction, IPERF_VTD_AFU_MEM_RD_TRANS);
+VTD_PORT_SHOW(write_transaction, IPERF_VTD_AFU_MEM_WR_TRANS);
+VTD_PORT_SHOW(devtlb_read_hit, IPERF_VTD_AFU_DEVTLB_RD_HIT);
+VTD_PORT_SHOW(devtlb_write_hit, IPERF_VTD_AFU_DEVTLB_WR_HIT);
+VTD_PORT_SHOW(devtlb_4k_fill, IPERF_VTD_DEVTLB_4K_FILL);
+VTD_PORT_SHOW(devtlb_2m_fill, IPERF_VTD_DEVTLB_2M_FILL);
+VTD_PORT_SHOW(devtlb_1g_fill, IPERF_VTD_DEVTLB_1G_FILL);
+
+static bool fabric_pobj_is_enabled(u8 port_id, struct feature_fme_iperf *iperf)
+{
+	struct feature_fme_ifpmon_fab_ctl ctl;
+
+	ctl.csr = readq(&iperf->fab_ctl);
+
+	if (ctl.port_filter == FAB_DISABLE_FILTER)
+		return port_id == PERF_OBJ_ROOT_ID;
+
+	return port_id == ctl.port_id;
+}
+
+static u64 read_fabric_counter(struct ifpga_fme_hw *fme, u8 port_id,
+			       enum iperf_fab_events fab_event)
+{
+	struct feature_fme_ifpmon_fab_ctl ctl;
+	struct feature_fme_ifpmon_fab_ctr ctr;
+	struct feature_fme_iperf *iperf;
+	u64 counter = 0;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+
+	/* if it is disabled, force the counter to return zero. */
+	if (!fabric_pobj_is_enabled(port_id, iperf))
+		goto exit;
+
+	ctl.csr = readq(&iperf->fab_ctl);
+	ctl.fab_evtcode = fab_event;
+	writeq(ctl.csr, &iperf->fab_ctl);
+
+	ctr.event_code = fab_event;
+
+	if (fpga_wait_register_field(event_code, ctr,
+				     &iperf->fab_ctr, IPERF_TIMEOUT, 1)) {
+		dev_err(fme, "timeout, unmatched VTd event type in counter registers.\n");
+		spinlock_unlock(&fme->lock);
+		return -ETIMEDOUT;
+	}
+
+	ctr.csr = readq(&iperf->fab_ctr);
+	counter = ctr.fab_cnt;
+exit:
+	spinlock_unlock(&fme->lock);
+	return counter;
+}
+
+#define FAB_PORT_SHOW(name, event)					\
+static int fme_iperf_get_fab_port_##name(struct ifpga_fme_hw *fme,	\
+				u8 port_id, u64 *counter)		\
+{									\
+	*counter = read_fabric_counter(fme, port_id, event);		\
+	return 0;							\
+}
+
+FAB_PORT_SHOW(pcie0_read, IPERF_FAB_PCIE0_RD);
+FAB_PORT_SHOW(pcie0_write, IPERF_FAB_PCIE0_WR);
+FAB_PORT_SHOW(pcie1_read, IPERF_FAB_PCIE1_RD);
+FAB_PORT_SHOW(pcie1_write, IPERF_FAB_PCIE1_WR);
+FAB_PORT_SHOW(upi_read, IPERF_FAB_UPI_RD);
+FAB_PORT_SHOW(upi_write, IPERF_FAB_UPI_WR);
+FAB_PORT_SHOW(mmio_read, IPERF_FAB_MMIO_RD);
+FAB_PORT_SHOW(mmio_write, IPERF_FAB_MMIO_WR);
+
+static int fme_iperf_get_fab_port_enable(struct ifpga_fme_hw *fme,
+					 u8 port_id, u64 *enable)
+{
+	struct feature_fme_iperf *iperf;
+	int status;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+
+	status = fabric_pobj_is_enabled(port_id, iperf);
+	*enable = (u64)status;
+
+	return 0;
+}
+
+/*
+ * If enable one port or all port event counter in fabric, other
+ * fabric event counter originally enabled will be disable automatically.
+ */
+static int fme_iperf_set_fab_port_enable(struct ifpga_fme_hw *fme,
+					 u8 port_id, u64 enable)
+{
+	struct feature_fme_ifpmon_fab_ctl ctl;
+	struct feature_fme_iperf *iperf;
+	bool state;
+
+	state = !!enable;
+
+	if (!state)
+		return -EINVAL;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+
+	/* if it is already enabled. */
+	if (fabric_pobj_is_enabled(port_id, iperf))
+		return 0;
+
+	spinlock_lock(&fme->lock);
+	ctl.csr = readq(&iperf->fab_ctl);
+	if (port_id == PERF_OBJ_ROOT_ID)
+		ctl.port_filter = FAB_DISABLE_FILTER;
+	else {
+		ctl.port_filter = FAB_ENABLE_FILTER;
+		ctl.port_id = port_id;
+	}
+
+	writeq(ctl.csr, &iperf->fab_ctl);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+static int fme_iperf_get_fab_freeze(struct ifpga_fme_hw *fme, u64 *freeze)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_fme_ifpmon_fab_ctl ctl;
+
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->fab_ctl);
+	*freeze = (u64)ctl.freeze;
+
+	return 0;
+}
+
+static int fme_iperf_set_fab_freeze(struct ifpga_fme_hw *fme, u64 freeze)
+{
+	struct feature_fme_iperf *iperf;
+	struct feature_fme_ifpmon_fab_ctl ctl;
+	bool state;
+
+	state = !!freeze;
+
+	spinlock_lock(&fme->lock);
+	iperf = get_fme_feature_ioaddr_by_index(fme,
+					    FME_FEATURE_ID_GLOBAL_IPERF);
+	ctl.csr = readq(&iperf->fab_ctl);
+	ctl.freeze = state;
+	writeq(ctl.csr, &iperf->fab_ctl);
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+#define PERF_MAX_PORT_NUM	1
+#define FME_IPERF_CAP_IOMMU	0x1
+
+static int fme_global_iperf_init(struct feature *feature)
+{
+	struct ifpga_fme_hw *fme;
+	struct feature_fme_header *fme_hdr;
+	struct feature_fme_capability fme_capability;
+
+	dev_info(NULL, "FME global_iperf Init.\n");
+
+	fme = (struct ifpga_fme_hw *)feature->parent;
+	fme_hdr = get_fme_feature_ioaddr_by_index(fme, FME_FEATURE_ID_HEADER);
+
+	/* check if iommu is not supported on this device. */
+	fme_capability.csr = readq(&fme_hdr->capability);
+	dev_info(NULL, "FME HEAD fme_capability %llx.\n",
+		 (unsigned long long)fme_hdr->capability.csr);
+
+	if (fme_capability.iommu_support)
+		feature->cap |= FME_IPERF_CAP_IOMMU;
+
+	return 0;
+}
+
+static void fme_global_iperf_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME global_iperf UInit.\n");
+}
+
+static int fme_iperf_root_get_prop(struct feature *feature,
+				   struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	if (sub != PERF_PROP_SUB_UNUSED)
+		return -ENOENT;
+
+	switch (id) {
+	case 0x1: /* CLOCK */
+		return fme_iperf_get_clock(fme, &prop->data);
+	case 0x2: /* REVISION */
+		return fme_iperf_get_revision(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_iperf_cache_get_prop(struct feature *feature,
+				    struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	if (sub != PERF_PROP_SUB_UNUSED)
+		return -ENOENT;
+
+	switch (id) {
+	case 0x1: /* FREEZE */
+		return fme_iperf_get_cache_freeze(fme, &prop->data);
+	case 0x2: /* READ_HIT */
+		return fme_iperf_get_cache_read_hit(fme, &prop->data);
+	case 0x3: /* READ_MISS */
+		return fme_iperf_get_cache_read_miss(fme, &prop->data);
+	case 0x4: /* WRITE_HIT */
+		return fme_iperf_get_cache_write_hit(fme, &prop->data);
+	case 0x5: /* WRITE_MISS */
+		return fme_iperf_get_cache_write_miss(fme, &prop->data);
+	case 0x6: /* HOLD_REQUEST */
+		return fme_iperf_get_cache_hold_request(fme, &prop->data);
+	case 0x7: /* TX_REQ_STALL */
+		return fme_iperf_get_cache_tx_req_stall(fme, &prop->data);
+	case 0x8: /* RX_REQ_STALL */
+		return fme_iperf_get_cache_rx_req_stall(fme, &prop->data);
+	case 0x9: /* RX_EVICTION */
+		return fme_iperf_get_cache_rx_eviction(fme, &prop->data);
+	case 0xa: /* DATA_WRITE_PORT_CONTENTION */
+		return fme_iperf_get_cache_data_write_port_contention(fme,
+							&prop->data);
+	case 0xb: /* TAG_WRITE_PORT_CONTENTION */
+		return fme_iperf_get_cache_tag_write_port_contention(fme,
+							&prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_iperf_vtd_root_get_prop(struct feature *feature,
+				       struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x1: /* FREEZE */
+		return fme_iperf_get_vtd_freeze(fme, &prop->data);
+	case 0x2: /* IOTLB_4K_HIT */
+		return fme_iperf_get_vtd_sip_iotlb_4k_hit(fme, &prop->data);
+	case 0x3: /* IOTLB_2M_HIT */
+		return fme_iperf_get_vtd_sip_iotlb_2m_hit(fme, &prop->data);
+	case 0x4: /* IOTLB_1G_HIT */
+		return fme_iperf_get_vtd_sip_iotlb_1g_hit(fme, &prop->data);
+	case 0x5: /* SLPWC_L3_HIT */
+		return fme_iperf_get_vtd_sip_slpwc_l3_hit(fme, &prop->data);
+	case 0x6: /* SLPWC_L4_HIT */
+		return fme_iperf_get_vtd_sip_slpwc_l4_hit(fme, &prop->data);
+	case 0x7: /* RCC_HIT */
+		return fme_iperf_get_vtd_sip_rcc_hit(fme, &prop->data);
+	case 0x8: /* IOTLB_4K_MISS */
+		return fme_iperf_get_vtd_sip_iotlb_4k_miss(fme, &prop->data);
+	case 0x9: /* IOTLB_2M_MISS */
+		return fme_iperf_get_vtd_sip_iotlb_2m_miss(fme, &prop->data);
+	case 0xa: /* IOTLB_1G_MISS */
+		return fme_iperf_get_vtd_sip_iotlb_1g_miss(fme, &prop->data);
+	case 0xb: /* SLPWC_L3_MISS */
+		return fme_iperf_get_vtd_sip_slpwc_l3_miss(fme, &prop->data);
+	case 0xc: /* SLPWC_L4_MISS */
+		return fme_iperf_get_vtd_sip_slpwc_l4_miss(fme, &prop->data);
+	case 0xd: /* RCC_MISS */
+		return fme_iperf_get_vtd_sip_rcc_miss(fme, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_iperf_vtd_sub_get_prop(struct feature *feature,
+				      struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+
+	if (sub > PERF_MAX_PORT_NUM)
+		return -ENOENT;
+
+	switch (id) {
+	case 0xe: /* READ_TRANSACTION */
+		return fme_iperf_get_vtd_port_read_transaction(fme, sub,
+							       &prop->data);
+	case 0xf: /* WRITE_TRANSACTION */
+		return fme_iperf_get_vtd_port_write_transaction(fme, sub,
+								&prop->data);
+	case 0x10: /* DEVTLB_READ_HIT */
+		return fme_iperf_get_vtd_port_devtlb_read_hit(fme, sub,
+							      &prop->data);
+	case 0x11: /* DEVTLB_WRITE_HIT */
+		return fme_iperf_get_vtd_port_devtlb_write_hit(fme, sub,
+							       &prop->data);
+	case 0x12: /* DEVTLB_4K_FILL */
+		return fme_iperf_get_vtd_port_devtlb_4k_fill(fme, sub,
+							     &prop->data);
+	case 0x13: /* DEVTLB_2M_FILL */
+		return fme_iperf_get_vtd_port_devtlb_2m_fill(fme, sub,
+							     &prop->data);
+	case 0x14: /* DEVTLB_1G_FILL */
+		return fme_iperf_get_vtd_port_devtlb_1g_fill(fme, sub,
+							     &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_iperf_vtd_get_prop(struct feature *feature,
+				  struct feature_prop *prop)
+{
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+
+	if (sub == PERF_PROP_SUB_UNUSED)
+		return fme_iperf_vtd_root_get_prop(feature, prop);
+
+	return fme_iperf_vtd_sub_get_prop(feature, prop);
+}
+
+static int fme_iperf_fab_get_prop(struct feature *feature,
+				  struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	/* Other properties are present for both top and sub levels */
+	switch (id) {
+	case 0x1: /* FREEZE */
+		if (sub != PERF_PROP_SUB_UNUSED)
+			return -ENOENT;
+		return fme_iperf_get_fab_freeze(fme, &prop->data);
+	case 0x2: /* PCIE0_READ */
+		return fme_iperf_get_fab_port_pcie0_read(fme, sub,
+							 &prop->data);
+	case 0x3: /* PCIE0_WRITE */
+		return fme_iperf_get_fab_port_pcie0_write(fme, sub,
+							  &prop->data);
+	case 0x4: /* PCIE1_READ */
+		return fme_iperf_get_fab_port_pcie1_read(fme, sub,
+							 &prop->data);
+	case 0x5: /* PCIE1_WRITE */
+		return fme_iperf_get_fab_port_pcie1_write(fme, sub,
+							  &prop->data);
+	case 0x6: /* UPI_READ */
+		return fme_iperf_get_fab_port_upi_read(fme, sub,
+						       &prop->data);
+	case 0x7: /* UPI_WRITE */
+		return fme_iperf_get_fab_port_upi_write(fme, sub,
+							&prop->data);
+	case 0x8: /* MMIO_READ */
+		return fme_iperf_get_fab_port_mmio_read(fme, sub,
+							&prop->data);
+	case 0x9: /* MMIO_WRITE */
+		return fme_iperf_get_fab_port_mmio_write(fme, sub,
+							 &prop->data);
+	case 0xa: /* ENABLE */
+		return fme_iperf_get_fab_port_enable(fme, sub, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_iperf_get_prop(struct feature *feature,
+				     struct feature_prop *prop)
+{
+	u8 top = GET_FIELD(PROP_TOP, prop->prop_id);
+
+	switch (top) {
+	case PERF_PROP_TOP_CACHE:
+		return fme_iperf_cache_get_prop(feature, prop);
+	case PERF_PROP_TOP_VTD:
+		return fme_iperf_vtd_get_prop(feature, prop);
+	case PERF_PROP_TOP_FAB:
+		return fme_iperf_fab_get_prop(feature, prop);
+	case PERF_PROP_TOP_UNUSED:
+		return fme_iperf_root_get_prop(feature, prop);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_iperf_cache_set_prop(struct feature *feature,
+				    struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	if (sub == PERF_PROP_SUB_UNUSED && id == 0x1) /* FREEZE */
+		return fme_iperf_set_cache_freeze(fme, prop->data);
+
+	return -ENOENT;
+}
+
+static int fme_iperf_vtd_set_prop(struct feature *feature,
+				  struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	if (sub == PERF_PROP_SUB_UNUSED && id == 0x1) /* FREEZE */
+		return fme_iperf_set_vtd_freeze(fme, prop->data);
+
+	return -ENOENT;
+}
+
+static int fme_iperf_fab_set_prop(struct feature *feature,
+				  struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme = feature->parent;
+	u8 sub = GET_FIELD(PROP_SUB, prop->prop_id);
+	u16 id = GET_FIELD(PROP_ID, prop->prop_id);
+
+	switch (id) {
+	case 0x1: /* FREEZE */
+		if (sub != PERF_PROP_SUB_UNUSED)
+			return -ENOENT;
+		return fme_iperf_set_fab_freeze(fme, prop->data);
+	case 0xa: /* ENABLE */
+		return fme_iperf_set_fab_port_enable(fme, sub, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int fme_global_iperf_set_prop(struct feature *feature,
+				     struct feature_prop *prop)
+{
+	u8 top = GET_FIELD(PROP_TOP, prop->prop_id);
+
+	switch (top) {
+	case PERF_PROP_TOP_CACHE:
+		return fme_iperf_cache_set_prop(feature, prop);
+	case PERF_PROP_TOP_VTD:
+		return fme_iperf_vtd_set_prop(feature, prop);
+	case PERF_PROP_TOP_FAB:
+		return fme_iperf_fab_set_prop(feature, prop);
+	}
+
+	return -ENOENT;
+}
+
+struct feature_ops fme_global_iperf_ops = {
+	.init = fme_global_iperf_init,
+	.uinit = fme_global_iperf_uinit,
+	.get_prop = fme_global_iperf_get_prop,
+	.set_prop = fme_global_iperf_set_prop,
+};
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c b/drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c
new file mode 100644
index 0000000..fec6013
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c
@@ -0,0 +1,395 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+//#if defined(RTE_ARCH_X86) && defined(RTE_MACHINE_CPUFLAG_AVX512F)
+#if 1
+static inline void copy512(const void *src, void *dst)
+{
+	asm volatile("vmovdqu64 (%0), %%zmm0;"
+		     "vmovntdq %%zmm0, (%1);"
+		     :
+		     : "r"(src), "r"(dst));
+}
+#else
+static inline void copy512(void *src, void *dst)
+{
+	WARN_ON(1);
+}
+#endif
+
+static u64
+pr_err_handle(struct feature_fme_pr *fme_pr)
+{
+	struct feature_fme_pr_status fme_pr_status;
+	unsigned long err_code;
+	u64 fme_pr_error;
+#if 0
+	int i = 0;
+#endif
+
+	fme_pr_status.csr = readq(&fme_pr->ccip_fme_pr_status);
+	if (!fme_pr_status.pr_status)
+		return 0;
+
+	err_code = readq(&fme_pr->ccip_fme_pr_err);
+	fme_pr_error = err_code;
+#if 0
+	for_each_set_bit(i, &err_code, PR_MAX_ERR_NUM)
+		dev_info(NULL, "%s\n", pr_err_msg[i]);
+#endif
+	writeq(fme_pr_error, &fme_pr->ccip_fme_pr_err);
+	return fme_pr_error;
+}
+
+static int fme_pr_write_init(struct ifpga_fme_hw *fme_dev,
+			     struct fpga_pr_info *info)
+{
+	struct feature_fme_pr *fme_pr;
+	struct feature_fme_pr_ctl fme_pr_ctl;
+	struct feature_fme_pr_status fme_pr_status;
+
+	fme_pr = get_fme_feature_ioaddr_by_index(fme_dev,
+						 FME_FEATURE_ID_PR_MGMT);
+	if (!fme_pr)
+		return -EINVAL;
+
+	if (info->flags != FPGA_MGR_PARTIAL_RECONFIG)
+		return -EINVAL;
+
+	dev_info(fme_dev, "resetting PR before initiated PR\n");
+
+	fme_pr_ctl.csr = readq(&fme_pr->ccip_fme_pr_control);
+	fme_pr_ctl.pr_reset = 1;
+	writeq(fme_pr_ctl.csr, &fme_pr->ccip_fme_pr_control);
+
+	fme_pr_ctl.pr_reset_ack = 1;
+
+	if (fpga_wait_register_field(pr_reset_ack, fme_pr_ctl,
+		&fme_pr->ccip_fme_pr_control, PR_WAIT_TIMEOUT, 1)) {
+		dev_err(fme_dev, "maximum PR timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	fme_pr_ctl.csr = readq(&fme_pr->ccip_fme_pr_control);
+	fme_pr_ctl.pr_reset = 0;
+	writeq(fme_pr_ctl.csr, &fme_pr->ccip_fme_pr_control);
+
+	dev_info(fme_dev, "waiting for PR resource in HW to be initialized and ready\n");
+
+	fme_pr_status.pr_host_status = PR_HOST_STATUS_IDLE;
+
+	if (fpga_wait_register_field(pr_host_status, fme_pr_status,
+		&fme_pr->ccip_fme_pr_status, PR_WAIT_TIMEOUT, 1)) {
+		dev_err(fme_dev, "maximum PR timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	dev_info(fme_dev, "check if have any previous PR error\n");
+	pr_err_handle(fme_pr);
+	return 0;
+}
+
+static int fme_pr_write(struct ifpga_fme_hw *fme_dev,
+			int port_id, const char *buf, size_t count,
+			struct fpga_pr_info *info)
+{
+	struct feature_fme_pr *fme_pr;
+	struct feature_fme_pr_ctl fme_pr_ctl;
+	struct feature_fme_pr_status fme_pr_status;
+	struct feature_fme_pr_data fme_pr_data;
+	int delay, pr_credit;
+	int ret = 0;
+
+	fme_pr = get_fme_feature_ioaddr_by_index(fme_dev,
+						 FME_FEATURE_ID_PR_MGMT);
+	if (!fme_pr)
+		return -EINVAL;
+
+	dev_info(fme_dev, "set PR port ID and start request\n");
+
+	fme_pr_ctl.csr = readq(&fme_pr->ccip_fme_pr_control);
+	fme_pr_ctl.pr_regionid = port_id;
+	fme_pr_ctl.pr_start_req = 1;
+	writeq(fme_pr_ctl.csr, &fme_pr->ccip_fme_pr_control);
+
+	dev_info(fme_dev, "pushing data from bitstream to HW\n");
+
+	fme_pr_status.csr = readq(&fme_pr->ccip_fme_pr_status);
+	pr_credit = fme_pr_status.pr_credit;
+
+	while (count > 0) {
+		delay = 0;
+		while (pr_credit <= 1) {
+			if (delay++ > PR_WAIT_TIMEOUT) {
+				dev_err(fme_dev, "maximum try\n");
+
+				info->pr_err = pr_err_handle(fme_pr);
+				return info->pr_err ? -EIO : -ETIMEDOUT;
+			}
+			udelay(1);
+
+			fme_pr_status.csr = readq(&fme_pr->ccip_fme_pr_status);
+			pr_credit = fme_pr_status.pr_credit;
+		};
+
+		if (count >= fme_dev->pr_bandwidth) {
+			switch (fme_dev->pr_bandwidth) {
+			case 4:
+				fme_pr_data.rsvd = 0;
+				fme_pr_data.pr_data_raw = *((const u32 *)buf);
+				writeq(fme_pr_data.csr,
+				       &fme_pr->ccip_fme_pr_data);
+				break;
+			case 64:
+				copy512(buf, &fme_pr->fme_pr_data1);
+				break;
+			default:
+				ret = -EFAULT;
+				goto done;
+			}
+
+			buf += fme_dev->pr_bandwidth;
+			count -= fme_dev->pr_bandwidth;
+			pr_credit--;
+		} else {
+			WARN_ON(1);
+			return -EINVAL;
+			goto done;
+		}
+	}
+
+done:
+	return ret;
+}
+
+static int fme_pr_write_complete(struct ifpga_fme_hw *fme_dev,
+				 struct fpga_pr_info *info)
+{
+	struct feature_fme_pr *fme_pr;
+	struct feature_fme_pr_ctl fme_pr_ctl;
+
+	fme_pr = get_fme_feature_ioaddr_by_index(fme_dev,
+						 FME_FEATURE_ID_PR_MGMT);
+
+	fme_pr_ctl.csr = readq(&fme_pr->ccip_fme_pr_control);
+	fme_pr_ctl.pr_push_complete = 1;
+	writeq(fme_pr_ctl.csr, &fme_pr->ccip_fme_pr_control);
+
+	dev_info(fme_dev, "green bitstream push complete\n");
+	dev_info(fme_dev, "waiting for HW to release PR resource\n");
+
+	fme_pr_ctl.pr_start_req = 0;
+
+	if (fpga_wait_register_field(pr_start_req, fme_pr_ctl,
+		&fme_pr->ccip_fme_pr_control, PR_WAIT_TIMEOUT, 1)) {
+		printf("maximum try.\n");
+		return -ETIMEDOUT;
+	}
+
+	dev_info(fme_dev, "PR operation complete, checking status\n");
+	info->pr_err = pr_err_handle(fme_pr);
+	if (info->pr_err)
+		return -EIO;
+
+	dev_info(fme_dev, "PR done successfully\n");
+	return 0;
+}
+
+static int fpga_pr_buf_load(struct ifpga_fme_hw *fme_dev,
+			    struct fpga_pr_info *info, const char *buf,
+			    size_t count)
+{
+	int ret;
+
+	info->state = FPGA_PR_STATE_WRITE_INIT;
+	ret = fme_pr_write_init(fme_dev, info);
+	if (ret) {
+		dev_err(fme_dev, "Error preparing FPGA for writing\n");
+		info->state = FPGA_PR_STATE_WRITE_INIT_ERR;
+		return ret;
+	}
+
+	/*
+	 * Write the FPGA image to the FPGA.
+	 */
+	info->state = FPGA_PR_STATE_WRITE;
+	ret = fme_pr_write(fme_dev, info->port_id, buf, count, info);
+	if (ret) {
+		dev_err(fme_dev, "Error while writing image data to FPGA\n");
+		info->state = FPGA_PR_STATE_WRITE_ERR;
+		return ret;
+	}
+
+	/*
+	 * After all the FPGA image has been written, do the device specific
+	 * steps to finish and set the FPGA into operating mode.
+	 */
+	info->state = FPGA_PR_STATE_WRITE_COMPLETE;
+	ret = fme_pr_write_complete(fme_dev, info);
+	if (ret) {
+		dev_err(fme_dev, "Error after writing image data to FPGA\n");
+		info->state = FPGA_PR_STATE_WRITE_COMPLETE_ERR;
+		return ret;
+	}
+	info->state = FPGA_PR_STATE_DONE;
+
+	return 0;
+}
+
+static int fme_pr(struct ifpga_hw *hw, u32 port_id, void *buffer, u32 size,
+		  u64 *status)
+{
+	struct feature_fme_header *fme_hdr;
+	struct feature_fme_capability fme_capability;
+	struct ifpga_fme_hw *fme = &hw->fme;
+	struct fpga_pr_info info;
+	struct ifpga_port_hw *port;
+	int ret = 0;
+
+	if (!buffer || size == 0)
+		return -EINVAL;
+	if (fme->state != IFPGA_FME_IMPLEMENTED)
+		return -EINVAL;
+
+	/*
+	 * Padding extra zeros to align PR buffer with PR bandwidth, HW will
+	 * ignore these zeros automatically.
+	 */
+	size = ALIGN(size, fme->pr_bandwidth);
+
+	/* get fme header region */
+	fme_hdr = get_fme_feature_ioaddr_by_index(fme,
+						  FME_FEATURE_ID_HEADER);
+	if (!fme_hdr)
+		return -EINVAL;
+
+	/* check port id */
+	fme_capability.csr = readq(&fme_hdr->capability);
+	if (port_id >= fme_capability.num_ports) {
+		dev_err(fme,  "port number more than maximum\n");
+		return -EINVAL;
+	}
+
+	memset(&info, 0, sizeof(struct fpga_pr_info));
+	info.flags = FPGA_MGR_PARTIAL_RECONFIG;
+	info.port_id = port_id;
+
+	spinlock_lock(&fme->lock);
+
+	/* get port device by port_id */
+	port = &hw->port[port_id];
+
+	/* Disable Port before PR */
+	fpga_port_disable(port);
+
+	ret = fpga_pr_buf_load(fme, &info, (void *)buffer, size);
+
+	*status = info.pr_err;
+
+	/* Re-enable Port after PR finished */
+	fpga_port_enable(port);
+	spinlock_unlock(&fme->lock);
+
+	return ret;
+}
+
+int do_pr(struct ifpga_hw *hw, u32 port_id, void *buffer, u32 size, u64 *status)
+{
+	struct bts_header *bts_hdr;
+	void *buf;
+	struct ifpga_port_hw *port;
+	int ret;
+
+	if (!buffer || size == 0) {
+		dev_err(hw, "invalid parameter\n");
+		return -EINVAL;
+	}
+
+	bts_hdr = (struct bts_header *)buffer;
+
+	if (is_valid_bts(bts_hdr)) {
+		dev_info(hw, "this is a valid bitsteam..\n");
+		size -= (sizeof(struct bts_header) +
+				     bts_hdr->metadata_len);
+		buf = (u8 *)buffer + sizeof(struct bts_header) +
+			       bts_hdr->metadata_len;
+	} else {
+		return -EINVAL;
+	}
+
+	/* clean port error before do PR */
+	port = &hw->port[port_id];
+	ret = port_clear_error(port);
+	if (ret) {
+		dev_err(hw, "port cannot clear error\n");
+		return -EINVAL;
+	}
+
+	return fme_pr(hw, port_id, buf, size, status);
+}
+
+static int fme_pr_mgmt_init(struct feature *feature)
+{
+	struct feature_fme_pr *fme_pr;
+	struct feature_header fme_pr_header;
+	struct ifpga_fme_hw *fme;
+
+	dev_info(NULL, "FME PR MGMT Init.\n");
+
+	fme = (struct ifpga_fme_hw *)feature->parent;
+
+	fme_pr = (struct feature_fme_pr *)feature->addr;
+
+	fme_pr_header.csr = readq(&fme_pr->header);
+	if (fme_pr_header.revision == 2) {
+		dev_info(NULL, "using 512-bit PR\n");
+		fme->pr_bandwidth = 64;
+	} else {
+		dev_info(NULL, "using 32-bit PR\n");
+		fme->pr_bandwidth = 4;
+	}
+
+	return 0;
+}
+
+static void fme_pr_mgmt_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "FME PR MGMT UInit.\n");
+}
+
+struct feature_ops fme_pr_mgmt_ops = {
+	.init = fme_pr_mgmt_init,
+	.uinit = fme_pr_mgmt_uinit,
+};
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_hw.h b/drivers/raw/ifpga_rawdev/base/ifpga_hw.h
new file mode 100644
index 0000000..5be615c
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_hw.h
@@ -0,0 +1,152 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _IFPGA_HW_H_
+#define _IFPGA_HW_H_
+
+#include "ifpga_defines.h"
+#include "opae_ifpga_hw_api.h"
+
+enum ifpga_feature_state {
+	IFPGA_FEATURE_UNUSED = 0,
+	IFPGA_FEATURE_ATTACHED,
+};
+
+struct feature_irq_ctx {
+	int eventfd;
+	int idx;
+};
+
+struct feature {
+	enum ifpga_feature_state state;
+	const char *name;
+	u64 id;
+	u8 *addr;
+	uint64_t phys_addr;
+	u32 size;
+	int revision;
+	u64 cap;
+	int vfio_dev_fd;
+	struct feature_irq_ctx *ctx;
+	unsigned int ctx_num;
+
+	void *parent;		/* to parent hw data structure */
+
+	struct feature_ops *ops;/* callback to this private feature */
+};
+
+struct feature_ops {
+	int (*init)(struct feature *feature);
+	void (*uinit)(struct feature *feature);
+	int (*test)(struct feature *feature);
+	int (*get_prop)(struct feature *feature, struct feature_prop *prop);
+	int (*set_prop)(struct feature *feature, struct feature_prop *prop);
+	int (*set_irq)(struct feature *feature, void *irq_set);
+};
+
+enum ifpga_fme_state {
+	IFPGA_FME_UNUSED = 0,
+	IFPGA_FME_IMPLEMENTED,
+};
+
+struct ifpga_fme_hw {
+	enum ifpga_fme_state state;
+
+	struct feature sub_feature[FME_FEATURE_ID_MAX];
+	spinlock_t lock;	/* protect hardware access */
+
+	void *parent;		/* pointer to ifpga_hw */
+
+	/* provied by HEADER feature */
+	u32 port_num;
+	struct uuid bitstream_id;
+	u64 bitstream_md;
+	size_t pr_bandwidth;
+	u32 socket_id;
+	u32 fabric_version_id;
+	u32 cache_size;
+
+	u32 capability;
+};
+
+enum ifpga_port_state {
+	IFPGA_PORT_UNUSED = 0,
+	IFPGA_PORT_ATTACHED,
+	IFPGA_PORT_DETACHED,
+};
+
+struct ifpga_port_hw {
+	enum ifpga_port_state state;
+
+	struct feature sub_feature[PORT_FEATURE_ID_MAX];
+	spinlock_t lock;	/* protect access to hw */
+
+	void *parent;		/* pointer to ifpga_hw */
+
+	int port_id;		/* provied by HEADER feature */
+	struct uuid afu_id;	/* provied by User AFU feature */
+
+	unsigned int disable_count;
+
+	u32 capability;
+	u32 num_umsgs;	/* The number of allocated umsgs */
+	u32 num_uafu_irqs;	/* The number of uafu interrupts */
+};
+
+#define AFU_MAX_REGION 1
+
+struct ifpga_afu_info {
+	struct opae_reg_region region[AFU_MAX_REGION];
+	unsigned int num_regions;
+	unsigned int num_irqs;
+};
+
+struct ifpga_hw {
+	struct opae_adapter *adapter;
+	struct opae_adapter_data_pci *pci_data;
+
+	struct ifpga_fme_hw fme;
+	struct ifpga_port_hw port[MAX_FPGA_PORT_NUM];
+};
+
+static inline bool is_ifpga_hw_pf(struct ifpga_hw *hw)
+{
+	return hw->fme.state != IFPGA_FME_UNUSED;
+}
+
+static inline bool is_valid_port_id(struct ifpga_hw *hw, u32 port_id)
+{
+	if (port_id >= MAX_FPGA_PORT_NUM ||
+		(hw->port[port_id].state != IFPGA_PORT_ATTACHED))
+		return false;
+	return true;
+}
+#endif /* _IFPGA_HW_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_port.c b/drivers/raw/ifpga_rawdev/base/ifpga_port.c
new file mode 100644
index 0000000..bf2e248
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_port.c
@@ -0,0 +1,730 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+int port_get_prop(struct ifpga_port_hw *port, struct feature_prop *prop)
+{
+	struct feature *feature;
+
+	if (!port)
+		return -ENOENT;
+
+	feature = get_port_feature_by_id(port, prop->feature_id);
+
+	if (feature && feature->ops && feature->ops->get_prop)
+		return feature->ops->get_prop(feature, prop);
+
+	return -ENOENT;
+}
+
+int port_set_prop(struct ifpga_port_hw *port, struct feature_prop *prop)
+{
+	struct feature *feature;
+
+	if (!port)
+		return -ENOENT;
+
+	feature = get_port_feature_by_id(port, prop->feature_id);
+
+	if (feature && feature->ops && feature->ops->set_prop)
+		return feature->ops->set_prop(feature, prop);
+
+	return -ENOENT;
+}
+
+int port_set_irq(struct ifpga_port_hw *port, u32 feature_id, void *irq_set)
+{
+	struct feature *feature;
+
+	if (!port)
+		return -ENOENT;
+
+	feature = get_port_feature_by_id(port, feature_id);
+
+	if (feature && feature->ops && feature->ops->set_irq)
+		return feature->ops->set_irq(feature, irq_set);
+
+	return -ENOENT;
+}
+
+
+void port_check_reg(void __iomem *addr,
+		    const char *reg_name, u64 dflt)
+{
+	u64 value = readq(addr);
+
+	UNUSED(reg_name);
+
+	if (value != dflt)
+		dev_debug(NULL, "%s: incorrect value 0x%llx vs defautl 0x%llx\n",
+			  reg_name, (unsigned long long)value,
+			  (unsigned long long)dflt);
+}
+
+static int port_get_revision(struct ifpga_port_hw *port, u64 *revision)
+{
+	struct feature_port_header *port_hdr
+		= get_port_feature_ioaddr_by_index(port,
+						   PORT_FEATURE_ID_HEADER);
+	struct feature_header header;
+
+	header.csr = readq(&port_hdr->header);
+
+	*revision = header.revision;
+
+	return 0;
+}
+
+static int port_get_portidx(struct ifpga_port_hw *port, u64 *idx)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_capability capability;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	capability.csr = readq(&port_hdr->capability);
+	*idx = capability.port_number;
+
+	return 0;
+}
+
+static int port_get_latency_tolerance(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_control control;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	control.csr = readq(&port_hdr->control);
+	*val = control.latency_tolerance;
+
+	return 0;
+}
+
+static int port_get_ap1_event(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_status status;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	status.csr = readq(&port_hdr->status);
+	spinlock_unlock(&port->lock);
+
+	*val = status.ap1_event;
+
+	return 0;
+}
+
+static int port_set_ap1_event(struct ifpga_port_hw *port, u64 val)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_status status;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	status.csr = readq(&port_hdr->status);
+	status.ap1_event = val;
+	writeq(status.csr, &port_hdr->status);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_get_ap2_event(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_status status;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	status.csr = readq(&port_hdr->status);
+	spinlock_unlock(&port->lock);
+
+	*val = status.ap2_event;
+
+	return 0;
+}
+
+static int port_set_ap2_event(struct ifpga_port_hw *port, u64 val)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_status status;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	status.csr = readq(&port_hdr->status);
+	status.ap2_event = val;
+	writeq(status.csr, &port_hdr->status);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_get_power_state(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+	struct feature_port_status status;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	status.csr = readq(&port_hdr->status);
+	spinlock_unlock(&port->lock);
+
+	*val = status.power_state;
+
+	return 0;
+}
+
+static int port_get_userclk_freqcmd(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	*val = readq(&port_hdr->user_clk_freq_cmd0);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_set_userclk_freqcmd(struct ifpga_port_hw *port, u64 val)
+{
+	struct feature_port_header *port_hdr;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	writeq(val, &port_hdr->user_clk_freq_cmd0);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_get_userclk_freqcntrcmd(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	*val = readq(&port_hdr->user_clk_freq_cmd1);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_set_userclk_freqcntrcmd(struct ifpga_port_hw *port, u64 val)
+{
+	struct feature_port_header *port_hdr;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	writeq(val, &port_hdr->user_clk_freq_cmd1);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_get_userclk_freqsts(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	*val = readq(&port_hdr->user_clk_freq_sts0);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static int port_get_userclk_freqcntrsts(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_header *port_hdr;
+
+	port_hdr = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_HEADER);
+
+	spinlock_lock(&port->lock);
+	*val = readq(&port_hdr->user_clk_freq_sts1);
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static struct feature_port_header hdr_dflt = {
+	.port_mailbox		= 0x0000000000000000,
+	.scratchpad		= 0x0000000000000000,
+	.capability = {
+		.csr		= 0x0000000100010000,
+	},
+	.control = {
+		/* Port Reset Bit is cleared in PCIe driver */
+		.csr		= 0x0000000000000004,
+	},
+	.status = {
+		.csr		= 0x0000000000000000,
+	},
+	.rsvd2			= 0x0000000000000000,
+	.user_clk_freq_cmd0	= 0x0000000000000000,
+	.user_clk_freq_cmd1	= 0x0000000000000000,
+	.user_clk_freq_sts0	= 0x0000000000000000,
+	.user_clk_freq_sts1	= 0x0000000000000000,
+};
+
+static int port_hdr_test(struct feature *feature)
+{
+	struct feature_port_header *port_hdr =
+		(struct feature_port_header *)feature->addr;
+
+	/* Check if default value of hardware registers matches with spec */
+	port_check_reg(&port_hdr->port_mailbox,
+			"hdr:port_mailbox", hdr_dflt.port_mailbox);
+	port_check_reg(&port_hdr->scratchpad,
+			"hdr:scratchpad", hdr_dflt.scratchpad);
+	port_check_reg(&port_hdr->capability,
+			"hdr:capability", hdr_dflt.capability.csr);
+	port_check_reg(&port_hdr->control,
+			"hdr:control", hdr_dflt.control.csr);
+	port_check_reg(&port_hdr->status,
+			"hdr:status", hdr_dflt.status.csr);
+	port_check_reg(&port_hdr->rsvd2,
+			"hdr:rsvd2", hdr_dflt.rsvd2);
+	port_check_reg(&port_hdr->user_clk_freq_cmd0,
+			"hdr:user_clk_cmd0", hdr_dflt.user_clk_freq_cmd0);
+	port_check_reg(&port_hdr->user_clk_freq_cmd1,
+			"hdr:user_clk_cmd1", hdr_dflt.user_clk_freq_cmd1);
+	port_check_reg(&port_hdr->user_clk_freq_sts0,
+			"hdr:user_clk_sts0", hdr_dflt.user_clk_freq_sts0);
+	port_check_reg(&port_hdr->user_clk_freq_sts1,
+			"hdr:user_clk_sts1", hdr_dflt.user_clk_freq_sts1);
+
+	dev_debug(NULL, "%s finished\n", __func__);
+
+	return 0;
+}
+
+static int port_hdr_init(struct feature *feature)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	dev_info(NULL, "port hdr Init.\n");
+
+	fpga_port_reset(port);
+
+	return 0;
+}
+
+static void port_hdr_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "port hdr uinit.\n");
+}
+
+static int port_hdr_get_prop(struct feature *feature,
+				   struct feature_prop *prop)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	switch (prop->prop_id) {
+	case PORT_HDR_PROP_REVISION:
+		return port_get_revision(port, &prop->data);
+	case PORT_HDR_PROP_PORTIDX:
+		return port_get_portidx(port, &prop->data);
+	case PORT_HDR_PROP_LATENCY_TOLERANCE:
+		return port_get_latency_tolerance(port, &prop->data);
+	case PORT_HDR_PROP_AP1_EVENT:
+		return port_get_ap1_event(port, &prop->data);
+	case PORT_HDR_PROP_AP2_EVENT:
+		return port_get_ap2_event(port, &prop->data);
+	case PORT_HDR_PROP_POWER_STATE:
+		return port_get_power_state(port, &prop->data);
+	case PORT_HDR_PROP_USERCLK_FREQCMD:
+		return port_get_userclk_freqcmd(port, &prop->data);
+	case PORT_HDR_PROP_USERCLK_FREQCNTRCMD:
+		return port_get_userclk_freqcntrcmd(port, &prop->data);
+	case PORT_HDR_PROP_USERCLK_FREQSTS:
+		return port_get_userclk_freqsts(port, &prop->data);
+	case PORT_HDR_PROP_USERCLK_CNTRSTS:
+		return port_get_userclk_freqcntrsts(port, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int port_hdr_set_prop(struct feature *feature,
+				   struct feature_prop *prop)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	switch (prop->prop_id) {
+	case PORT_HDR_PROP_AP1_EVENT:
+		return port_set_ap1_event(port, prop->data);
+	case PORT_HDR_PROP_AP2_EVENT:
+		return port_set_ap2_event(port, prop->data);
+	case PORT_HDR_PROP_USERCLK_FREQCMD:
+		return port_set_userclk_freqcmd(port, prop->data);
+	case PORT_HDR_PROP_USERCLK_FREQCNTRCMD:
+		return port_set_userclk_freqcntrcmd(port, prop->data);
+	}
+
+	return -ENOENT;
+}
+
+struct feature_ops port_hdr_ops = {
+	.init = port_hdr_init,
+	.uinit = port_hdr_uinit,
+	.test = port_hdr_test,
+	.get_prop = port_hdr_get_prop,
+	.set_prop = port_hdr_set_prop,
+};
+
+static struct feature_port_umsg umsg_dflt = {
+	.capability = {
+		.csr		= 0x0000000000000008,
+	},
+	.baseaddr = {
+		.csr		= 0x0000000000000000,
+	},
+	.mode = {
+		.csr		= 0x0000000000000000,
+	},
+};
+
+static int port_umsg_test(struct feature *feature)
+{
+	struct feature_port_umsg *port_umsg =
+		(struct feature_port_umsg *)feature->addr;
+
+	port_check_reg(&port_umsg->capability,
+				"umsg:capaiblity", umsg_dflt.capability.csr);
+	port_check_reg(&port_umsg->baseaddr,
+				"umsg:baseaddr", umsg_dflt.baseaddr.csr);
+	port_check_reg(&port_umsg->mode,
+				"umsg:mode", umsg_dflt.mode.csr);
+
+	dev_debug(NULL, "%s finished\n", __func__);
+	return 0;
+}
+
+static u8 port_umsg_get_num(struct ifpga_port_hw *port)
+{
+	struct feature_port_umsg *port_umsg;
+	struct feature_port_umsg_cap capability;
+
+	port_umsg = get_port_feature_ioaddr_by_index(port,
+						PORT_FEATURE_ID_UMSG);
+
+	capability.csr = readq(&port_umsg->capability);
+
+	return capability.umsg_allocated;
+}
+
+static u64 port_umsg_get_addr(struct ifpga_port_hw *port)
+{
+	struct feature_port_umsg *port_umsg;
+	struct feature_port_umsg_baseaddr baseaddr;
+
+	port_umsg = get_port_feature_ioaddr_by_index(port,
+					PORT_FEATURE_ID_UMSG);
+
+	baseaddr.csr = readq(&port_umsg->baseaddr);
+
+	return baseaddr.base_addr;
+}
+
+static int port_umsg_enable(struct ifpga_port_hw *port, bool enable)
+{
+	struct feature_port_umsg *port_umsg;
+	struct feature_port_umsg_cap capability;
+
+	port_umsg = get_port_feature_ioaddr_by_index(port,
+					PORT_FEATURE_ID_UMSG);
+
+	capability.csr = readq(&port_umsg->capability);
+
+	/* Return directly if UMSG is already enabled/disabled */
+	if ((enable && capability.umsg_enable) ||
+			!(enable || capability.umsg_enable))
+		return 0;
+
+	capability.umsg_enable = enable;
+	writeq(capability.csr, &port_umsg->capability);
+
+	/*
+	 * Each time umsg engine enabled/disabled, driver polls the
+	 * init_complete bit for confirmation.
+	 */
+	capability.umsg_init_complete = !!enable;
+
+	if (fpga_wait_register_field(umsg_init_complete, capability,
+				     &port_umsg->capability,
+				     UMSG_EN_POLL_TIMEOUT, UMSG_EN_POLL_INVL)) {
+		dev_err(dev, "timeout, fail to %s umsg\n",
+					enable ? "enable" : "disable");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static bool port_umsg_is_enabled(struct ifpga_port_hw *port)
+{
+	struct feature_port_umsg *port_umsg;
+	struct feature_port_umsg_cap capability;
+
+	port_umsg = get_port_feature_ioaddr_by_index(port,
+						PORT_FEATURE_ID_UMSG);
+
+	capability.csr = readq(&port_umsg->capability);
+
+	return capability.umsg_enable;
+}
+
+static void port_umsg_set_mode(struct ifpga_port_hw *port, u32 mode)
+{
+	struct feature_port_umsg *port_umsg;
+	struct feature_port_umsg_mode umode;
+
+	port_umsg = get_port_feature_ioaddr_by_index(port,
+						PORT_FEATURE_ID_UMSG);
+
+	umode.csr = readq(&port_umsg->mode);
+	umode.umsg_hint_enable = mode;
+	writeq(umode.csr, &port_umsg->mode);
+}
+
+static void port_umsg_set_addr(struct ifpga_port_hw *port, u64 iova)
+{
+	struct feature_port_umsg *port_umsg;
+	struct feature_port_umsg_baseaddr baseaddr;
+
+	port_umsg = get_port_feature_ioaddr_by_index(port,
+					PORT_FEATURE_ID_UMSG);
+
+	baseaddr.csr = readq(&port_umsg->baseaddr);
+	baseaddr.base_addr = iova;
+	writeq(baseaddr.csr, &port_umsg->baseaddr);
+}
+
+int afu_port_umsg_enable(struct ifpga_port_hw *port, bool enable)
+{
+	if (enable && port_umsg_get_addr(port))
+		return -EIO;
+	if (port_umsg_enable(port, enable))
+		return -ENODEV;
+
+	return 0;
+}
+
+int afu_port_umsg_set_mode(struct ifpga_port_hw *port, u32 mode)
+{
+	u8 num_umsgs = port_umsg_get_num(port);
+
+	if (mode >> num_umsgs) {
+		dev_err(port, "invaild UMsg config hint_bitmap\n");
+		return -EINVAL;
+	}
+
+	port_umsg_set_mode(port, mode);
+
+	return 0;
+}
+
+static int __maybe_unused afu_port_umsg_set_addr(struct ifpga_port_hw *port,
+						 u64 iova)
+{
+	u8 num_umsgs = port_umsg_get_num(port);
+	u64 size = num_umsgs * PAGE_SIZE;
+
+	/* Make sure base addr is configured only when umsg is disabled */
+	if (port_umsg_is_enabled(port)) {
+		dev_err(port, "umsg is still enabled\n");
+		return -EIO;
+	}
+
+	if (iova) {
+		/* Check input, only accept page-aligned region for umsg */
+		if (!PAGE_ALIGNED(iova))
+			return -EINVAL;
+
+		/* Check overflow */
+		if (iova + size < iova)
+			return -EINVAL;
+
+		port_umsg_set_addr(port, iova);
+	} else {
+		/* Read current iova from hardware */
+		iova = port_umsg_get_addr(port);
+		if (!iova)
+			return 0;
+
+		/* Check overflow */
+		if ((iova + size < iova))
+			return -EINVAL;
+
+		port_umsg_set_addr(port, 0);
+	}
+
+	return 0;
+}
+
+static int port_umsg_init(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "port umsg Init.\n");
+
+	return 0;
+}
+
+static void port_umsg_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "port umsg uinit.\n");
+}
+
+struct feature_ops port_umsg_ops = {
+	.init = port_umsg_init,
+	.uinit = port_umsg_uinit,
+	.test = port_umsg_test,
+};
+
+static struct feature_port_stp stp_dflt = {
+	.stp_status = {
+		.csr		= 0x0000000000000000,
+	},
+};
+
+static int port_stp_test(struct feature *feature)
+{
+	struct feature_port_stp *port_stp =
+		(struct feature_port_stp *)feature->addr;
+
+	port_check_reg(&port_stp->stp_status,
+				"stp:stp_csr", stp_dflt.stp_status.csr);
+
+	dev_debug(NULL, "%s finished\n", __func__);
+	return 0;
+}
+
+static int port_stp_init(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "port stp Init.\n");
+
+	return 0;
+}
+
+static void port_stp_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "port stp uinit.\n");
+}
+
+struct feature_ops port_stp_ops = {
+	.init = port_stp_init,
+	.uinit = port_stp_uinit,
+	.test = port_stp_test,
+};
+
+static int port_uint_init(struct feature *feature)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	dev_info(NULL, "PORT UINT Init.\n");
+
+	spinlock_lock(&port->lock);
+	if (feature->ctx_num) {
+		port->capability |= FPGA_PORT_CAP_UAFU_IRQ;
+		port->num_uafu_irqs = feature->ctx_num;
+	}
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static void port_uint_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+
+	dev_info(NULL, "PORT UINT UInit.\n");
+}
+
+static int port_uint_set_irq(struct feature *feature, void *irq_set)
+{
+	struct fpga_uafu_irq_set *uafu_irq_set = irq_set;
+	struct ifpga_port_hw *port = feature->parent;
+	int ret;
+
+	spinlock_lock(&port->lock);
+	if (!(port->capability & FPGA_PORT_CAP_UAFU_IRQ)) {
+		spinlock_unlock(&port->lock);
+		return -ENODEV;
+	}
+
+	ret = fpga_msix_set_block(feature, uafu_irq_set->start,
+				  uafu_irq_set->count, uafu_irq_set->evtfds);
+	spinlock_unlock(&port->lock);
+
+	return ret;
+}
+
+struct feature_ops port_uint_ops = {
+	.init = port_uint_init,
+	.uinit = port_uint_uinit,
+	.set_irq = port_uint_set_irq,
+};
diff --git a/drivers/raw/ifpga_rawdev/base/ifpga_port_error.c b/drivers/raw/ifpga_rawdev/base/ifpga_port_error.c
new file mode 100644
index 0000000..1a55155
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/ifpga_port_error.c
@@ -0,0 +1,236 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "ifpga_feature_dev.h"
+
+static int port_err_get_revision(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_error *port_err;
+	struct feature_header header;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+	header.csr = readq(&port_err->header);
+	*val = header.revision;
+
+	return 0;
+}
+
+static int port_err_get_errors(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_error *port_err;
+	struct feature_port_err_key error;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+	error.csr = readq(&port_err->port_error);
+	*val = error.csr;
+
+	return 0;
+}
+
+static int port_err_get_first_error(struct ifpga_port_hw *port, u64 *val)
+{
+	struct feature_port_error *port_err;
+	struct feature_port_first_err_key first_error;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+	first_error.csr = readq(&port_err->port_first_error);
+	*val = first_error.csr;
+
+	return 0;
+}
+
+static int port_err_get_first_malformed_req_lsb(struct ifpga_port_hw *port,
+						u64 *val)
+{
+	struct feature_port_error *port_err;
+	struct feature_port_malformed_req0 malreq0;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+
+	malreq0.header_lsb = readq(&port_err->malreq0);
+	*val = malreq0.header_lsb;
+
+	return 0;
+}
+
+static int port_err_get_first_malformed_req_msb(struct ifpga_port_hw *port,
+						u64 *val)
+{
+	struct feature_port_error *port_err;
+	struct feature_port_malformed_req1 malreq1;
+
+	port_err = get_port_feature_ioaddr_by_index(port,
+						    PORT_FEATURE_ID_ERROR);
+
+	malreq1.header_msb = readq(&port_err->malreq1);
+	*val = malreq1.header_msb;
+
+	return 0;
+}
+
+static int port_err_set_clear(struct ifpga_port_hw *port, u64 val)
+{
+	int ret;
+
+	spinlock_lock(&port->lock);
+	ret = port_err_clear(port, val);
+	spinlock_unlock(&port->lock);
+
+	return ret;
+}
+
+static struct feature_port_error err_dflt = {
+	.error_mask = {
+		.csr		= 0x0000000000000000,
+	},
+	.port_error = {
+		.csr		= 0x0000000000000000,
+	},
+	.port_first_error = {
+		.csr		= 0x0000000000000000,
+	},
+	.malreq0 = {
+		.header_lsb	= 0x0000000000000000,
+	},
+	.malreq1 = {
+		.header_msb	= 0x0000000000000000,
+	},
+	.port_debug = {
+		.port_debug	= 0x0000000000000000,
+	},
+};
+
+static int port_err_test(struct feature *feature)
+{
+	struct feature_port_error *port_err =
+		(struct feature_port_error *)feature->addr;
+
+	port_check_reg(&port_err->error_mask,
+		       "err:error_mask", err_dflt.error_mask.csr);
+	port_check_reg(&port_err->port_error,
+		       "err:port_error", err_dflt.port_error.csr);
+	port_check_reg(&port_err->port_first_error,
+		       "err:port_first_err", err_dflt.port_first_error.csr);
+	port_check_reg(&port_err->malreq0,
+		       "err:malreq0", err_dflt.malreq0.header_lsb);
+	port_check_reg(&port_err->malreq1,
+		       "err:malreq1", err_dflt.malreq1.header_msb);
+	port_check_reg(&port_err->port_debug,
+		       "err:port_debug", err_dflt.port_debug.port_debug);
+
+	dev_debug(NULL, "%s finished\n", __func__);
+	return 0;
+}
+
+static int port_error_init(struct feature *feature)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	dev_info(NULL, "port error Init.\n");
+
+	spinlock_lock(&port->lock);
+	port_err_mask(port, false);
+	if (feature->ctx_num)
+		port->capability |= FPGA_PORT_CAP_ERR_IRQ;
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+static void port_error_uinit(struct feature *feature)
+{
+	UNUSED(feature);
+}
+
+static int port_error_get_prop(struct feature *feature,
+			       struct feature_prop *prop)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	switch (prop->prop_id) {
+	case PORT_ERR_PROP_REVISION:
+		return port_err_get_revision(port, &prop->data);
+	case PORT_ERR_PROP_ERRORS:
+		return port_err_get_errors(port, &prop->data);
+	case PORT_ERR_PROP_FIRST_ERROR:
+		return port_err_get_first_error(port, &prop->data);
+	case PORT_ERR_PROP_FIRST_MALFORMED_REQ_LSB:
+		return port_err_get_first_malformed_req_lsb(port, &prop->data);
+	case PORT_ERR_PROP_FIRST_MALFORMED_REQ_MSB:
+		return port_err_get_first_malformed_req_msb(port, &prop->data);
+	}
+
+	return -ENOENT;
+}
+
+static int port_error_set_prop(struct feature *feature,
+			       struct feature_prop *prop)
+{
+	struct ifpga_port_hw *port = feature->parent;
+
+	if (prop->prop_id == PORT_ERR_PROP_CLEAR)
+		return port_err_set_clear(port, prop->data);
+
+	return -ENOENT;
+}
+
+static int port_error_set_irq(struct feature *feature, void *irq_set)
+{
+	struct fpga_port_err_irq_set *err_irq_set = irq_set;
+	struct ifpga_port_hw *port;
+	int ret;
+
+	port = feature->parent;
+
+	spinlock_lock(&port->lock);
+	if (!(port->capability & FPGA_PORT_CAP_ERR_IRQ)) {
+		spinlock_unlock(&port->lock);
+		return -ENODEV;
+	}
+
+	ret = fpga_msix_set_block(feature, 0, 1, &err_irq_set->evtfd);
+	spinlock_unlock(&port->lock);
+
+	return ret;
+}
+
+struct feature_ops port_error_ops = {
+	.init = port_error_init,
+	.uinit = port_error_uinit,
+	.test = port_err_test,
+	.get_prop = port_error_get_prop,
+	.set_prop = port_error_set_prop,
+	.set_irq = port_error_set_irq,
+};
diff --git a/drivers/raw/ifpga_rawdev/base/opae_debug.c b/drivers/raw/ifpga_rawdev/base/opae_debug.c
new file mode 100644
index 0000000..07935e7
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_debug.c
@@ -0,0 +1,126 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#define OPAE_HW_DEBUG
+
+#include "opae_hw_api.h"
+#include "opae_debug.h"
+
+void opae_manager_dump(struct opae_manager *mgr)
+{
+	opae_log("=====%s=====\n", __func__);
+	opae_log("OPAE Manger %s\n", mgr->name);
+	opae_log("OPAE Manger OPs = %p\n", mgr->ops);
+	opae_log("OPAE Manager Private Data = %p\n", mgr->data);
+	opae_log("OPAE Adapter(parent) = %p\n", mgr->adapter);
+	opae_log("==========================\n");
+}
+
+void opae_bridge_dump(struct opae_bridge *br)
+{
+	opae_log("=====%s=====\n", __func__);
+	opae_log("OPAE Bridge %s\n", br->name);
+	opae_log("OPAE Bridge ID = %d\n", br->id);
+	opae_log("OPAE Bridge OPs = %p\n", br->ops);
+	opae_log("OPAE Bridge Private Data = %p\n", br->data);
+	opae_log("OPAE Accelerator(under this bridge) = %p\n", br->acc);
+	opae_log("==========================\n");
+}
+
+void opae_accelerator_dump(struct opae_accelerator *acc)
+{
+	opae_log("=====%s=====\n", __func__);
+	opae_log("OPAE Accelerator %s\n", acc->name);
+	opae_log("OPAE Accelerator Index = %d\n", acc->index);
+	opae_log("OPAE Accelerator OPs = %p\n", acc->ops);
+	opae_log("OPAE Accelerator Private Data = %p\n", acc->data);
+	opae_log("OPAE Bridge (upstream) = %p\n", acc->br);
+	opae_log("OPAE Manager (upstream) = %p\n", acc->mgr);
+	opae_log("==========================\n");
+
+	if (acc->br)
+		opae_bridge_dump(acc->br);
+}
+
+static void opae_adapter_data_dump(void *data)
+{
+	struct opae_adapter_data *d = data;
+	struct opae_adapter_data_pci *d_pci;
+	struct opae_reg_region *r;
+	int i;
+
+	opae_log("=====%s=====\n", __func__);
+
+	switch (d->type) {
+	case OPAE_FPGA_PCI:
+		d_pci = (struct opae_adapter_data_pci *)d;
+
+		opae_log("OPAE Adapter Type = PCI\n");
+		opae_log("PCI Device ID: 0x%04x\n", d_pci->device_id);
+		opae_log("PCI Vendor ID: 0x%04x\n", d_pci->vendor_id);
+
+		for (i = 0; i < PCI_MAX_RESOURCE; i++) {
+			r = &d_pci->region[i];
+			opae_log("PCI Bar %d: phy(%llx) len(%llx) addr(%p)\n",
+				 i, (unsigned long long)r->phys_addr,
+				 (unsigned long long)r->len, r->addr);
+		}
+		break;
+	case OPAE_FPGA_NET:
+		break;
+	}
+
+	opae_log("==========================\n");
+}
+
+void opae_adapter_dump(struct opae_adapter *adapter, int verbose)
+{
+	struct opae_accelerator *acc;
+
+	opae_log("=====%s=====\n", __func__);
+	opae_log("OPAE Adapter %s\n", adapter->name);
+	opae_log("OPAE Adapter OPs = %p\n", adapter->ops);
+	opae_log("OPAE Adapter Private Data = %p\n", adapter->data);
+	opae_log("OPAE Manager (downstream) = %p\n", adapter->mgr);
+
+	if (verbose) {
+		if (adapter->mgr)
+			opae_manager_dump(adapter->mgr);
+
+		opae_adapter_for_each_acc(adapter, acc)
+			opae_accelerator_dump(acc);
+
+		if (adapter->data)
+			opae_adapter_data_dump(adapter->data);
+	}
+
+	opae_log("==========================\n");
+}
diff --git a/drivers/raw/ifpga_rawdev/base/opae_debug.h b/drivers/raw/ifpga_rawdev/base/opae_debug.h
new file mode 100644
index 0000000..ceddf9d
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_debug.h
@@ -0,0 +1,46 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _OPAE_DEBUG_H_
+#define _OPAE_DEBUG_H_
+
+#ifdef OPAE_HW_DEBUG
+#define opae_log(fmt, args...) printf(fmt, ## args)
+#else
+#define opae_log(fme, args...) do {} while (0)
+#endif
+
+void opae_manager_dump(struct opae_manager *mgr);
+void opae_bridge_dump(struct opae_bridge *br);
+void opae_accelerator_dump(struct opae_accelerator *acc);
+void opae_adapter_dump(struct opae_adapter *adapter, int verbose);
+
+#endif /* _OPAE_DEBUG_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/opae_hw_api.c b/drivers/raw/ifpga_rawdev/base/opae_hw_api.c
new file mode 100644
index 0000000..a0cecfb
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_hw_api.c
@@ -0,0 +1,389 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "opae_hw_api.h"
+#include "opae_debug.h"
+#include "ifpga_api.h"
+
+/* OPAE Bridge Functions */
+
+/**
+ * opae_bridge_alloc - alloc opae_bridge data structure
+ * @name: bridge name.
+ * @ops: ops of this bridge.
+ * @data: private data of this bridge.
+ *
+ * Return opae_bridge on success, otherwise NULL.
+ */
+struct opae_bridge *
+opae_bridge_alloc(const char *name, struct opae_bridge_ops *ops, void *data)
+{
+	struct opae_bridge *br = opae_zmalloc(sizeof(*br));
+
+	if (!br)
+		return NULL;
+
+	br->name = name;
+	br->ops = ops;
+	br->data = data;
+
+	opae_log("%s %p\n", __func__, br);
+
+	return br;
+}
+
+/**
+ * opae_bridge_reset -  reset opae_bridge
+ * @br: bridge to be reset.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_bridge_reset(struct opae_bridge *br)
+{
+	if (!br)
+		return -EINVAL;
+
+	if (br->ops && br->ops->reset)
+		return br->ops->reset(br);
+
+	opae_log("%s no ops\n", __func__);
+
+	return -ENOENT;
+}
+
+/* Accelerator Functions */
+
+/**
+ * opae_accelerator_alloc - alloc opae_accelerator data structure
+ * @name: accelerator name.
+ * @ops: ops of this accelerator.
+ * @data: private data of this accelerator.
+ *
+ * Return: opae_accelerator on success, otherwise NULL.
+ */
+struct opae_accelerator *
+opae_accelerator_alloc(const char *name, struct opae_accelerator_ops *ops,
+		       void *data)
+{
+	struct opae_accelerator *acc = opae_zmalloc(sizeof(*acc));
+
+	if (!acc)
+		return NULL;
+
+	acc->name = name;
+	acc->ops = ops;
+	acc->data = data;
+
+	opae_log("%s %p\n", __func__, acc);
+
+	return acc;
+}
+
+/**
+ * opae_acc_reg_read - read accelerator's register from its reg region.
+ * @acc: accelerator to read.
+ * @region_idx: reg region index.
+ * @offset: reg offset.
+ * @byte: read operation width, e.g 4 byte = 32bit read.
+ * @data: data to store the value read from the register.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_acc_reg_read(struct opae_accelerator *acc, unsigned int region_idx,
+		      u64 offset, unsigned int byte, void *data)
+{
+	if (!acc || !data)
+		return -EINVAL;
+
+	if (acc->ops && acc->ops->read)
+		return acc->ops->read(acc, offset, region_idx, byte, data);
+
+	return -ENOENT;
+}
+
+/**
+ * opae_acc_reg_write - write to accelerator's register from its reg region.
+ * @acc: accelerator to write.
+ * @region_idx: reg region index.
+ * @offset: reg offset.
+ * @byte: write operation width, e.g 4 byte = 32bit write.
+ * @data: data stored the value to write to the register.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_acc_reg_write(struct opae_accelerator *acc, unsigned int region_idx,
+		       u64 offset, unsigned int byte, void *data)
+{
+	if (!acc || !data)
+		return -EINVAL;
+
+	if (acc->ops && acc->ops->write)
+		return acc->ops->write(acc, offset, region_idx, byte, data);
+
+	return -ENOENT;
+}
+
+/**
+ * opae_acc_get_info - get information of an accelerator.
+ * @acc: targeted accelerator
+ * @info: accelerator info data structure to be filled.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_acc_get_info(struct opae_accelerator *acc, struct opae_acc_info *info)
+{
+	if (!acc || !info)
+		return -EINVAL;
+
+	if (acc->ops && acc->ops->get_info)
+		return acc->ops->get_info(acc, info);
+
+	return -ENOENT;
+}
+
+/**
+ * opae_acc_get_region_info - get information of an accelerator register region.
+ * @acc: targeted accelerator
+ * @info: accelerator region info data structure to be filled.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_acc_get_region_info(struct opae_accelerator *acc,
+			     struct opae_acc_region_info *info)
+{
+	if (!acc || !info)
+		return -EINVAL;
+
+	if (acc->ops && acc->ops->get_region_info)
+		return acc->ops->get_region_info(acc, info);
+
+	return -ENOENT;
+}
+
+/**
+ * opae_acc_set_irq -  set an accelerator's irq.
+ * @acc: targeted accelerator
+ * @start: start vector number
+ * @count: count of vectors to be set from the start vector
+ * @evtfds: event fds to be notified when corresponding irqs happens
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_acc_set_irq(struct opae_accelerator *acc,
+		     u32 start, u32 count, s32 evtfds[])
+{
+	if (!acc || !acc->data)
+		return -EINVAL;
+
+	if (start + count <= start)
+		return -EINVAL;
+
+	if (acc->ops && acc->ops->set_irq)
+		return acc->ops->set_irq(acc, start, count, evtfds);
+
+	return -ENOENT;
+}
+
+/* Manager Functions */
+
+/**
+ * opae_manager_alloc - alloc opae_manager data structure
+ * @name: manager name.
+ * @ops: ops of this manager.
+ * @data: private data of this manager.
+ *
+ * Return: opae_manager on success, otherwise NULL.
+ */
+struct opae_manager *
+opae_manager_alloc(const char *name, struct opae_manager_ops *ops, void *data)
+{
+	struct opae_manager *mgr = opae_zmalloc(sizeof(*mgr));
+
+	if (!mgr)
+		return NULL;
+
+	mgr->name = name;
+	mgr->ops = ops;
+	mgr->data = data;
+
+	opae_log("%s %p\n", __func__, mgr);
+
+	return mgr;
+}
+
+/**
+ * opae_manager_flash - flash a reconfiguration image via opae_manager
+ * @mgr: opae_manager for flash.
+ * @id: id of target region (accelerator).
+ * @buf: image data buffer.
+ * @size: buffer size.
+ * @status: status to store flash result.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_manager_flash(struct opae_manager *mgr, int id, void *buf, u32 size,
+		       u64 *status)
+{
+	if (!mgr)
+		return -EINVAL;
+
+	if (mgr && mgr->ops && mgr->ops->flash)
+		return mgr->ops->flash(mgr, id, buf, size, status);
+
+	return -ENOENT;
+}
+
+/* Adapter Functions */
+
+/**
+ * opae_adapter_data_alloc - alloc opae_adapter_data data structure
+ * @type: opae_adapter_type.
+ *
+ * Return: opae_adapter_data on success, otherwise NULL.
+ */
+void *opae_adapter_data_alloc(enum opae_adapter_type type)
+{
+	struct opae_adapter_data *data;
+	int size;
+
+	switch (type) {
+	case OPAE_FPGA_PCI:
+		size = sizeof(struct opae_adapter_data_pci);
+		break;
+	case OPAE_FPGA_NET:
+		size = sizeof(struct opae_adapter_data_net);
+		break;
+	default:
+		size = sizeof(struct opae_adapter_data);
+		break;
+	}
+
+	data = opae_zmalloc(size);
+	if (!data)
+		return NULL;
+
+	data->type = type;
+
+	return data;
+}
+
+static struct opae_adapter_ops *match_ops(struct opae_adapter *adapter)
+{
+	struct opae_adapter_data *data;
+
+	if (!adapter || !adapter->data)
+		return NULL;
+
+	data = adapter->data;
+
+	if (data->type == OPAE_FPGA_PCI)
+		return &ifpga_adapter_ops;
+
+	return NULL;
+}
+
+/**
+ * opae_adapter_data_alloc - alloc opae_adapter_data data structure
+ * @name: adapter name.
+ * @data: private data of this adapter.
+ *
+ * Return: opae_adapter on success, otherwise NULL.
+ */
+struct opae_adapter *opae_adapter_alloc(const char *name, void *data)
+{
+	struct opae_adapter *adapter = opae_zmalloc(sizeof(*adapter));
+
+	if (!adapter)
+		return NULL;
+
+	TAILQ_INIT(&adapter->acc_list);
+	adapter->data = data;
+	adapter->name = name;
+	adapter->ops = match_ops(adapter);
+
+	return adapter;
+}
+
+/**
+ * opae_adapter_enumerate - enumerate this adapter
+ * @adapter: adapter to enumerate.
+ *
+ * Return: 0 on success, otherwise error code.
+ */
+int opae_adapter_enumerate(struct opae_adapter *adapter)
+{
+	int ret = -ENOENT;
+
+	if (!adapter)
+		return -EINVAL;
+
+	if (adapter->ops && adapter->ops->enumerate)
+		ret = adapter->ops->enumerate(adapter);
+
+	if (!ret)
+		opae_adapter_dump(adapter, 1);
+
+	return ret;
+}
+
+/**
+ * opae_adapter_destroy - destroy this adapter
+ * @adapter: adapter to destroy.
+ *
+ * destroy things allocated during adapter enumeration.
+ */
+void opae_adapter_destroy(struct opae_adapter *adapter)
+{
+	if (adapter && adapter->ops && adapter->ops->destroy)
+		adapter->ops->destroy(adapter);
+}
+
+/**
+ * opae_adapter_get_acc - find and return accelerator with matched id
+ * @adapter: adapter to find the accelerator.
+ * @acc_id: id (index) of the accelerator.
+ *
+ * destroy things allocated during adapter enumeration.
+ */
+struct opae_accelerator *
+opae_adapter_get_acc(struct opae_adapter *adapter, int acc_id)
+{
+	struct opae_accelerator *acc = NULL;
+
+	if (!adapter)
+		return NULL;
+
+	opae_adapter_for_each_acc(adapter, acc)
+		if (acc->index == acc_id)
+			return acc;
+
+	return NULL;
+}
diff --git a/drivers/raw/ifpga_rawdev/base/opae_hw_api.h b/drivers/raw/ifpga_rawdev/base/opae_hw_api.h
new file mode 100644
index 0000000..1f2391c
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_hw_api.h
@@ -0,0 +1,276 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _OPAE_HW_API_H_
+#define _OPAE_HW_API_H_
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <sys/queue.h>
+
+#include "opae_osdep.h"
+
+#ifndef PCI_MAX_RESOURCE
+#define PCI_MAX_RESOURCE 6
+#endif
+
+struct opae_adapter;
+
+enum opae_adapter_type {
+	OPAE_FPGA_PCI,
+	OPAE_FPGA_NET,
+};
+
+/* OPAE Manager Data Structure */
+struct opae_manager_ops;
+
+/*
+ * opae_manager has pointer to its parent adapter, as it could be able to manage
+ * all components on this FPGA device (adapter). If not the case, don't set this
+ * adapter, which limit opae_manager ops to manager itself.
+ */
+struct opae_manager {
+	const char *name;
+	struct opae_adapter *adapter;
+	struct opae_manager_ops *ops;
+	void *data;
+};
+
+/* FIXME: add more management ops, e.g power/thermal and etc */
+struct opae_manager_ops {
+	int (*flash)(struct opae_manager *mgr, int id, void *buffer,
+		     u32 size, u64 *status);
+};
+
+/* OPAE Manager APIs */
+struct opae_manager *
+opae_manager_alloc(const char *name, struct opae_manager_ops *ops, void *data);
+#define opae_manager_free(mgr) opae_free(mgr)
+int opae_manager_flash(struct opae_manager *mgr, int acc_id, void *buf,
+		       u32 size, u64 *status);
+
+/* OPAE Bridge Data Structure */
+struct opae_bridge_ops;
+
+/*
+ * opae_bridge only has pointer to its downstream accelerator.
+ */
+struct opae_bridge {
+	const char *name;
+	int id;
+	struct opae_accelerator *acc;
+	struct opae_bridge_ops *ops;
+	void *data;
+};
+
+struct opae_bridge_ops {
+	int (*reset)(struct opae_bridge *br);
+};
+
+/* OPAE Bridge APIs */
+struct opae_bridge *
+opae_bridge_alloc(const char *name, struct opae_bridge_ops *ops, void *data);
+int opae_bridge_reset(struct opae_bridge *br);
+#define opae_bridge_free(br) opae_free(br)
+
+/* OPAE Acceleraotr Data Structure */
+struct opae_accelerator_ops;
+
+/*
+ * opae_accelerator has pointer to its upstream bridge(port).
+ * In some cases, if we allow same user to do PR on its own accelerator, then
+ * set the manager pointer during the enumeration. But in other cases, the PR
+ * functions only could be done via manager in another module / thread / service
+ * / application for better protection.
+ */
+struct opae_accelerator {
+	TAILQ_ENTRY(opae_accelerator) node;
+	const char *name;
+	int index;
+	struct opae_bridge *br;
+	struct opae_manager *mgr;
+	struct opae_accelerator_ops *ops;
+	void *data;
+};
+
+struct opae_acc_info {
+	unsigned int num_regions;
+	unsigned int num_irqs;
+};
+
+struct opae_acc_region_info {
+	u32 flags;
+#define ACC_REGION_READ		(1 << 0)
+#define ACC_REGION_WRITE	(1 << 1)
+#define ACC_REGION_MMIO		(1 << 2)
+	u32 index;
+	u64 phys_addr;
+	u64 len;
+	u8 *addr;
+};
+
+struct opae_accelerator_ops {
+	int (*read)(struct opae_accelerator *acc, unsigned int region_idx,
+		    u64 offset, unsigned int byte, void *data);
+	int (*write)(struct opae_accelerator *acc, unsigned int region_idx,
+		     u64 offset, unsigned int byte, void *data);
+	int (*get_info)(struct opae_accelerator *acc,
+			struct opae_acc_info *info);
+	int (*get_region_info)(struct opae_accelerator *acc,
+			       struct opae_acc_region_info *info);
+	int (*set_irq)(struct opae_accelerator *acc,
+		       u32 start, u32 count, s32 evtfds[]);
+};
+
+/* OPAE accelerator APIs */
+struct opae_accelerator *
+opae_accelerator_alloc(const char *name, struct opae_accelerator_ops *ops,
+		       void *data);
+#define opae_accelerator_free(acc) opae_free(acc)
+int opae_acc_get_info(struct opae_accelerator *acc, struct opae_acc_info *info);
+int opae_acc_get_region_info(struct opae_accelerator *acc,
+			     struct opae_acc_region_info *info);
+int opae_acc_set_irq(struct opae_accelerator *acc,
+		     u32 start, u32 count, s32 evtfds[]);
+
+static inline struct opae_bridge *
+opae_acc_get_br(struct opae_accelerator *acc)
+{
+	return acc ? acc->br : NULL;
+}
+
+static inline struct opae_manager *
+opae_acc_get_mgr(struct opae_accelerator *acc)
+{
+	return acc ? acc->mgr : NULL;
+}
+
+int opae_acc_reg_read(struct opae_accelerator *acc, unsigned int region_idx,
+		      u64 offset, unsigned int byte, void *data);
+int opae_acc_reg_write(struct opae_accelerator *acc, unsigned int region_idx,
+		       u64 offset, unsigned int byte, void *data);
+
+#define opae_acc_reg_read64(acc, region, offset, data) \
+	opae_acc_reg_read(acc, region, offset, 8, data)
+#define opae_acc_reg_write64(acc, region, offset, data) \
+	opae_acc_reg_write(acc, region, offset, 8, data)
+#define opae_acc_reg_read32(acc, region, offset, data) \
+	opae_acc_reg_read(acc, region, offset, 4, data)
+#define opae_acc_reg_write32(acc, region, offset, data) \
+	opae_acc_reg_write(acc, region, offset, 4, data)
+#define opae_acc_reg_read16(acc, region, offset, data) \
+	opae_acc_reg_read(acc, region, offset, 2, data)
+#define opae_acc_reg_write16(acc, region, offset, data) \
+	opae_acc_reg_write(acc, region, offset, 2, data)
+#define opae_acc_reg_read8(acc, region, offset, data) \
+	opae_acc_reg_read(acc, region, offset, 1, data)
+#define opae_acc_reg_write8(acc, region, offset, data) \
+	opae_acc_reg_write(acc, region, offset, 1, data)
+
+/*for data stream read/write*/
+int opae_acc_data_read(struct opae_accelerator *acc, unsigned int flags,
+		      u64 offset, unsigned int byte, void *data);
+int opae_acc_data_write(struct opae_accelerator *acc, unsigned int flags,
+		       u64 offset, unsigned int byte, void *data);
+
+/* OPAE Adapter Data Structure */
+struct opae_adapter_data {
+	enum opae_adapter_type type;
+};
+
+struct opae_reg_region {
+	u64 phys_addr;
+	u64 len;
+	u8 *addr;
+};
+
+struct opae_adapter_data_pci {
+	enum opae_adapter_type type;
+	u16 device_id;
+	u16 vendor_id;
+	struct opae_reg_region region[PCI_MAX_RESOURCE];
+	int vfio_dev_fd;  /* VFIO device file descriptor */
+};
+
+/* FIXME: OPAE_FPGA_NET type */
+struct opae_adapter_data_net {
+	enum opae_adapter_type type;
+};
+
+struct opae_adapter_ops {
+	int (*enumerate)(struct opae_adapter *adapter);
+	void (*destroy)(struct opae_adapter *adapter);
+};
+
+TAILQ_HEAD(opae_accelerator_list, opae_accelerator);
+
+#define opae_adapter_for_each_acc(adatper, acc) \
+	TAILQ_FOREACH(acc, &adapter->acc_list, node)
+
+struct opae_adapter {
+	const char *name;
+	struct opae_manager *mgr;
+	struct opae_accelerator_list acc_list;
+	struct opae_adapter_ops *ops;
+	void *data;
+};
+
+/* OPAE Adapter APIs */
+void *opae_adapter_data_alloc(enum opae_adapter_type type);
+#define opae_adapter_data_free(data) opae_free(data)
+
+struct opae_adapter *opae_adapter_alloc(const char *name, void *data);
+#define opae_adapter_free(adapter) opae_free(adapter)
+
+int opae_adapter_enumerate(struct opae_adapter *adapter);
+void opae_adapter_destroy(struct opae_adapter *adapter);
+static inline struct opae_manager *
+opae_adapter_get_mgr(struct opae_adapter *adapter)
+{
+	return adapter ? adapter->mgr : NULL;
+}
+
+struct opae_accelerator *
+opae_adapter_get_acc(struct opae_adapter *adapter, int acc_id);
+
+static inline void opae_adapter_add_acc(struct opae_adapter *adapter,
+					struct opae_accelerator *acc)
+{
+	TAILQ_INSERT_TAIL(&adapter->acc_list, acc, node);
+}
+
+static inline void opae_adapter_remove_acc(struct opae_adapter *adapter,
+					   struct opae_accelerator *acc)
+{
+	TAILQ_REMOVE(&adapter->acc_list, acc, node);
+}
+#endif /* _OPAE_HW_API_H_*/
diff --git a/drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c b/drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c
new file mode 100644
index 0000000..9d3bc2f
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c
@@ -0,0 +1,151 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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 "opae_ifpga_hw_api.h"
+#include "ifpga_api.h"
+
+int opae_manager_ifpga_get_prop(struct opae_manager *mgr,
+				struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme;
+
+	if (!mgr || !mgr->data)
+		return -EINVAL;
+
+	fme = mgr->data;
+
+	return ifpga_get_prop(fme->parent, FEATURE_FIU_ID_FME, 0, prop);
+}
+
+int opae_manager_ifpga_set_prop(struct opae_manager *mgr,
+				struct feature_prop *prop)
+{
+	struct ifpga_fme_hw *fme;
+
+	if (!mgr || !mgr->data)
+		return -EINVAL;
+
+	fme = mgr->data;
+
+	return ifpga_set_prop(fme->parent, FEATURE_FIU_ID_FME, 0, prop);
+}
+
+int opae_manager_ifpga_get_info(struct opae_manager *mgr,
+				struct fpga_fme_info *fme_info)
+{
+	struct ifpga_fme_hw *fme;
+
+	if (!mgr || !mgr->data || !fme_info)
+		return -EINVAL;
+
+	fme = mgr->data;
+
+	spinlock_lock(&fme->lock);
+	fme_info->capability = fme->capability;
+	spinlock_unlock(&fme->lock);
+
+	return 0;
+}
+
+int opae_manager_ifpga_set_err_irq(struct opae_manager *mgr,
+				   struct fpga_fme_err_irq_set *err_irq_set)
+{
+	struct ifpga_fme_hw *fme;
+
+	if (!mgr || !mgr->data)
+		return -EINVAL;
+
+	fme = mgr->data;
+
+	return ifpga_set_irq(fme->parent, FEATURE_FIU_ID_FME, 0,
+			     FME_FEATURE_ID_GLOBAL_ERR, err_irq_set);
+}
+
+int opae_bridge_ifpga_get_prop(struct opae_bridge *br,
+			       struct feature_prop *prop)
+{
+	struct ifpga_port_hw *port;
+
+	if (!br || !br->data)
+		return -EINVAL;
+
+	port = br->data;
+
+	return ifpga_get_prop(port->parent, FEATURE_FIU_ID_PORT,
+			      port->port_id, prop);
+}
+
+int opae_bridge_ifpga_set_prop(struct opae_bridge *br,
+			       struct feature_prop *prop)
+{
+	struct ifpga_port_hw *port;
+
+	if (!br || !br->data)
+		return -EINVAL;
+
+	port = br->data;
+
+	return ifpga_set_prop(port->parent, FEATURE_FIU_ID_PORT,
+			      port->port_id, prop);
+}
+
+int opae_bridge_ifpga_get_info(struct opae_bridge *br,
+			       struct fpga_port_info *port_info)
+{
+	struct ifpga_port_hw *port;
+
+	if (!br || !br->data || !port_info)
+		return -EINVAL;
+
+	port = br->data;
+
+	spinlock_lock(&port->lock);
+	port_info->capability = port->capability;
+	port_info->num_umsgs = port->num_umsgs;
+	port_info->num_uafu_irqs = port->num_uafu_irqs;
+	spinlock_unlock(&port->lock);
+
+	return 0;
+}
+
+int opae_bridge_ifpga_set_err_irq(struct opae_bridge *br,
+				  struct fpga_port_err_irq_set *err_irq_set)
+{
+	struct ifpga_port_hw *port;
+
+	if (!br || !br->data)
+		return -EINVAL;
+
+	port = br->data;
+
+	return ifpga_set_irq(port->parent, FEATURE_FIU_ID_PORT, port->port_id,
+			     PORT_FEATURE_ID_ERROR, err_irq_set);
+}
diff --git a/drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h b/drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h
new file mode 100644
index 0000000..047945e
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h
@@ -0,0 +1,293 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _OPAE_IFPGA_HW_API_H_
+#define _OPAE_IFPGA_HW_API_H_
+
+#include "opae_hw_api.h"
+
+/**
+ * struct feature_prop - data structure for feature property
+ * @feature_id: id of this feature.
+ * @prop_id: id of this property under this feature.
+ * @data: property value to set/get.
+ */
+struct feature_prop {
+	u64 feature_id;
+	u64 prop_id;
+	u64 data;
+};
+
+#define IFPGA_FIU_ID_FME	0x0
+#define IFPGA_FIU_ID_PORT	0x1
+
+#define IFPGA_FME_FEATURE_ID_HEADER		0x0
+#define IFPGA_FME_FEATURE_ID_THERMAL_MGMT	0x1
+#define IFPGA_FME_FEATURE_ID_POWER_MGMT		0x2
+#define IFPGA_FME_FEATURE_ID_GLOBAL_IPERF	0x3
+#define IFPGA_FME_FEATURE_ID_GLOBAL_ERR		0x4
+#define IFPGA_FME_FEATURE_ID_PR_MGMT		0x5
+#define IFPGA_FME_FEATURE_ID_HSSI		0x6
+#define IFPGA_FME_FEATURE_ID_GLOBAL_DPERF	0x7
+
+#define IFPGA_PORT_FEATURE_ID_HEADER		0x0
+#define IFPGA_PORT_FEATURE_ID_AFU		0xff
+#define IFPGA_PORT_FEATURE_ID_ERROR		0x10
+#define IFPGA_PORT_FEATURE_ID_UMSG		0x11
+#define IFPGA_PORT_FEATURE_ID_UINT		0x12
+#define IFPGA_PORT_FEATURE_ID_STP		0x13
+
+/*
+ * PROP format (TOP + SUB + ID)
+ *
+ * (~0x0) means this field is unused.
+ */
+#define PROP_TOP	GENMASK(31, 24)
+#define PROP_TOP_UNUSED	0xff
+#define PROP_SUB	GENMASK(23, 16)
+#define PROP_SUB_UNUSED	0xff
+#define PROP_ID		GENMASK(15, 0)
+
+#define PROP(_top, _sub, _id) \
+	(SET_FIELD(PROP_TOP, _top) | SET_FIELD(PROP_SUB, _sub) |\
+	 SET_FIELD(PROP_ID, _id))
+
+/* FME head feature's properties*/
+#define FME_HDR_PROP_REVISION		0x1	/* RDONLY */
+#define FME_HDR_PROP_PORTS_NUM		0x2	/* RDONLY */
+#define FME_HDR_PROP_CACHE_SIZE		0x3	/* RDONLY */
+#define FME_HDR_PROP_VERSION			0x4	/* RDONLY */
+#define FME_HDR_PROP_SOCKET_ID		0x5	/* RDONLY */
+#define FME_HDR_PROP_BITSTREAM_ID		0x6	/* RDONLY */
+#define FME_HDR_PROP_BITSTREAM_METADATA	0x7	/* RDONLY */
+
+/* FME error reporting feature's properties */
+/* FME error reporting properties format */
+#define ERR_PROP(_top, _id)		PROP(_top, 0xff, _id)
+#define ERR_PROP_TOP_UNUSED		PROP_TOP_UNUSED
+#define ERR_PROP_TOP_FME_ERR		0x1
+#define ERR_PROP_ROOT(_id)		ERR_PROP(0xff, _id)
+#define ERR_PROP_FME_ERR(_id)		ERR_PROP(ERR_PROP_TOP_FME_ERR, _id)
+
+#define FME_ERR_PROP_ERRORS		ERR_PROP_FME_ERR(0x1)
+#define FME_ERR_PROP_FIRST_ERROR	ERR_PROP_FME_ERR(0x2)
+#define FME_ERR_PROP_NEXT_ERROR		ERR_PROP_FME_ERR(0x3)
+#define FME_ERR_PROP_CLEAR		ERR_PROP_FME_ERR(0x4)	/* WO */
+#define FME_ERR_PROP_REVISION		ERR_PROP_ROOT(0x5)
+#define FME_ERR_PROP_PCIE0_ERRORS	ERR_PROP_ROOT(0x6)	/* RW */
+#define FME_ERR_PROP_PCIE1_ERRORS	ERR_PROP_ROOT(0x7)	/* RW */
+#define FME_ERR_PROP_NONFATAL_ERRORS	ERR_PROP_ROOT(0x8)
+#define FME_ERR_PROP_CATFATAL_ERRORS	ERR_PROP_ROOT(0x9)
+#define FME_ERR_PROP_INJECT_ERRORS	ERR_PROP_ROOT(0xa)	/* RW */
+
+/* FME thermal feature's properties */
+#define FME_THERMAL_PROP_THRESHOLD1		0x1	/* RW */
+#define FME_THERMAL_PROP_THRESHOLD2		0x2	/* RW */
+#define FME_THERMAL_PROP_THRESHOLD_TRIP		0x3	/* RDONLY */
+#define FME_THERMAL_PROP_THRESHOLD1_REACHED	0x4	/* RDONLY */
+#define FME_THERMAL_PROP_THRESHOLD2_REACHED	0x5	/* RDONLY */
+#define FME_THERMAL_PROP_THRESHOLD1_POLICY	0x6	/* RW */
+#define FME_THERMAL_PROP_TEMPERATURE		0x7	/* RDONLY */
+#define FME_THERMAL_PROP_REVISION		0x8	/* RDONLY */
+
+/* FME power feature's properties */
+#define FME_PWR_PROP_CONSUMED			0x1	/* RDONLY */
+#define FME_PWR_PROP_THRESHOLD1			0x2	/* RW */
+#define FME_PWR_PROP_THRESHOLD2			0x3	/* RW */
+#define FME_PWR_PROP_THRESHOLD1_STATUS		0x4	/* RDONLY */
+#define FME_PWR_PROP_THRESHOLD2_STATUS		0x5	/* RDONLY */
+#define FME_PWR_PROP_RTL			0x6	/* RDONLY */
+#define FME_PWR_PROP_XEON_LIMIT			0x7	/* RDONLY */
+#define FME_PWR_PROP_FPGA_LIMIT			0x8	/* RDONLY */
+#define FME_PWR_PROP_REVISION			0x9	/* RDONLY */
+
+/* FME iperf/dperf PROP format */
+#define PERF_PROP_TOP_CACHE			0x1
+#define PERF_PROP_TOP_VTD			0x2
+#define PERF_PROP_TOP_FAB			0x3
+#define PERF_PROP_TOP_UNUSED			PROP_TOP_UNUSED
+#define PERF_PROP_SUB_UNUSED			PROP_SUB_UNUSED
+
+#define PERF_PROP_ROOT(_id)		PROP(0xff, 0xff, _id)
+#define PERF_PROP_CACHE(_id)		PROP(PERF_PROP_TOP_CACHE, 0xff, _id)
+#define PERF_PROP_VTD(_sub, _id)	PROP(PERF_PROP_TOP_VTD, _sub, _id)
+#define PERF_PROP_VTD_ROOT(_id)		PROP(PERF_PROP_TOP_VTD, 0xff, _id)
+#define PERF_PROP_FAB(_sub, _id)	PROP(PERF_PROP_TOP_FAB, _sub, _id)
+#define PERF_PROP_FAB_ROOT(_id)		PROP(PERF_PROP_TOP_FAB, 0xff, _id)
+
+/* FME iperf feature's properties */
+#define FME_IPERF_PROP_CLOCK			PERF_PROP_ROOT(0x1)
+#define FME_IPERF_PROP_REVISION			PERF_PROP_ROOT(0x2)
+
+/* iperf CACHE properties */
+#define FME_IPERF_PROP_CACHE_FREEZE		PERF_PROP_CACHE(0x1) /* RW */
+#define FME_IPERF_PROP_CACHE_READ_HIT		PERF_PROP_CACHE(0x2)
+#define FME_IPERF_PROP_CACHE_READ_MISS		PERF_PROP_CACHE(0x3)
+#define FME_IPERF_PROP_CACHE_WRITE_HIT		PERF_PROP_CACHE(0x4)
+#define FME_IPERF_PROP_CACHE_WRITE_MISS		PERF_PROP_CACHE(0x5)
+#define FME_IPERF_PROP_CACHE_HOLD_REQUEST	PERF_PROP_CACHE(0x6)
+#define FME_IPERF_PROP_CACHE_TX_REQ_STALL	PERF_PROP_CACHE(0x7)
+#define FME_IPERF_PROP_CACHE_RX_REQ_STALL	PERF_PROP_CACHE(0x8)
+#define FME_IPERF_PROP_CACHE_RX_EVICTION	PERF_PROP_CACHE(0x9)
+#define FME_IPERF_PROP_CACHE_DATA_WRITE_PORT_CONTENTION	PERF_PROP_CACHE(0xa)
+#define FME_IPERF_PROP_CACHE_TAG_WRITE_PORT_CONTENTION	PERF_PROP_CACHE(0xb)
+/* iperf VTD properties */
+#define FME_IPERF_PROP_VTD_FREEZE		PERF_PROP_VTD_ROOT(0x1) /* RW */
+#define FME_IPERF_PROP_VTD_SIP_IOTLB_4K_HIT	PERF_PROP_VTD_ROOT(0x2)
+#define FME_IPERF_PROP_VTD_SIP_IOTLB_2M_HIT	PERF_PROP_VTD_ROOT(0x3)
+#define FME_IPERF_PROP_VTD_SIP_IOTLB_1G_HIT	PERF_PROP_VTD_ROOT(0x4)
+#define FME_IPERF_PROP_VTD_SIP_SLPWC_L3_HIT	PERF_PROP_VTD_ROOT(0x5)
+#define FME_IPERF_PROP_VTD_SIP_SLPWC_L4_HIT	PERF_PROP_VTD_ROOT(0x6)
+#define FME_IPERF_PROP_VTD_SIP_RCC_HIT		PERF_PROP_VTD_ROOT(0x7)
+#define FME_IPERF_PROP_VTD_SIP_IOTLB_4K_MISS	PERF_PROP_VTD_ROOT(0x8)
+#define FME_IPERF_PROP_VTD_SIP_IOTLB_2M_MISS	PERF_PROP_VTD_ROOT(0x9)
+#define FME_IPERF_PROP_VTD_SIP_IOTLB_1G_MISS	PERF_PROP_VTD_ROOT(0xa)
+#define FME_IPERF_PROP_VTD_SIP_SLPWC_L3_MISS	PERF_PROP_VTD_ROOT(0xb)
+#define FME_IPERF_PROP_VTD_SIP_SLPWC_L4_MISS	PERF_PROP_VTD_ROOT(0xc)
+#define FME_IPERF_PROP_VTD_SIP_RCC_MISS		PERF_PROP_VTD_ROOT(0xd)
+#define FME_IPERF_PROP_VTD_PORT_READ_TRANSACTION(n)	PERF_PROP_VTD(n, 0xe)
+#define FME_IPERF_PROP_VTD_PORT_WRITE_TRANSACTION(n)	PERF_PROP_VTD(n, 0xf)
+#define FME_IPERF_PROP_VTD_PORT_DEVTLB_READ_HIT(n)	PERF_PROP_VTD(n, 0x10)
+#define FME_IPERF_PROP_VTD_PORT_DEVTLB_WRITE_HIT(n)	PERF_PROP_VTD(n, 0x11)
+#define FME_IPERF_PROP_VTD_PORT_DEVTLB_4K_FILL(n)	PERF_PROP_VTD(n, 0x12)
+#define FME_IPERF_PROP_VTD_PORT_DEVTLB_2M_FILL(n)	PERF_PROP_VTD(n, 0x13)
+#define FME_IPERF_PROP_VTD_PORT_DEVTLB_1G_FILL(n)	PERF_PROP_VTD(n, 0x14)
+/* iperf FAB properties */
+#define FME_IPERF_PROP_FAB_FREEZE		PERF_PROP_FAB_ROOT(0x1) /* RW */
+#define FME_IPERF_PROP_FAB_PCIE0_READ		PERF_PROP_FAB_ROOT(0x2)
+#define FME_IPERF_PROP_FAB_PORT_PCIE0_READ(n)	PERF_PROP_FAB(n, 0x2)
+#define FME_IPERF_PROP_FAB_PCIE0_WRITE		PERF_PROP_FAB_ROOT(0x3)
+#define FME_IPERF_PROP_FAB_PORT_PCIE0_WRITE(n)	PERF_PROP_FAB(n, 0x3)
+#define FME_IPERF_PROP_FAB_PCIE1_READ		PERF_PROP_FAB_ROOT(0x4)
+#define FME_IPERF_PROP_FAB_PORT_PCIE1_READ(n)	PERF_PROP_FAB(n, 0x4)
+#define FME_IPERF_PROP_FAB_PCIE1_WRITE		PERF_PROP_FAB_ROOT(0x5)
+#define FME_IPERF_PROP_FAB_PORT_PCIE1_WRITE(n)	PERF_PROP_FAB(n, 0x5)
+#define FME_IPERF_PROP_FAB_UPI_READ		PERF_PROP_FAB_ROOT(0x6)
+#define FME_IPERF_PROP_FAB_PORT_UPI_READ(n)	PERF_PROP_FAB(n, 0x6)
+#define FME_IPERF_PROP_FAB_UPI_WRITE		PERF_PROP_FAB_ROOT(0x7)
+#define FME_IPERF_PROP_FAB_PORT_UPI_WRITE(n)	PERF_PROP_FAB(n, 0x7)
+#define FME_IPERF_PROP_FAB_MMIO_READ		PERF_PROP_FAB_ROOT(0x8)
+#define FME_IPERF_PROP_FAB_PORT_MMIO_READ(n)	PERF_PROP_FAB(n, 0x8)
+#define FME_IPERF_PROP_FAB_MMIO_WRITE		PERF_PROP_FAB_ROOT(0x9)
+#define FME_IPERF_PROP_FAB_PORT_MMIO_WRITE(n)	PERF_PROP_FAB(n, 0x9)
+#define FME_IPERF_PROP_FAB_ENABLE		PERF_PROP_FAB_ROOT(0xa) /* RW */
+#define FME_IPERF_PROP_FAB_PORT_ENABLE(n)	PERF_PROP_FAB(n, 0xa)   /* RW */
+
+/* FME dperf properties */
+#define FME_DPERF_PROP_CLOCK			PERF_PROP_ROOT(0x1)
+#define FME_DPERF_PROP_REVISION			PERF_PROP_ROOT(0x2)
+
+/* dperf FAB properties */
+#define FME_DPERF_PROP_FAB_FREEZE		PERF_PROP_FAB_ROOT(0x1) /* RW */
+#define FME_DPERF_PROP_FAB_PCIE0_READ		PERF_PROP_FAB_ROOT(0x2)
+#define FME_DPERF_PROP_FAB_PORT_PCIE0_READ(n)	PERF_PROP_FAB(n, 0x2)
+#define FME_DPERF_PROP_FAB_PCIE0_WRITE		PERF_PROP_FAB_ROOT(0x3)
+#define FME_DPERF_PROP_FAB_PORT_PCIE0_WRITE(n)	PERF_PROP_FAB(n, 0x3)
+#define FME_DPERF_PROP_FAB_MMIO_READ		PERF_PROP_FAB_ROOT(0x4)
+#define FME_DPERF_PROP_FAB_PORT_MMIO_READ(n)	PERF_PROP_FAB(n, 0x4)
+#define FME_DPERF_PROP_FAB_MMIO_WRITE		PERF_PROP_FAB_ROOT(0x5)
+#define FME_DPERF_PROP_FAB_PORT_MMIO_WRITE(n)	PERF_PROP_FAB(n, 0x5)
+#define FME_DPERF_PROP_FAB_ENABLE		PERF_PROP_FAB_ROOT(0x6) /* RW */
+#define FME_DPERF_PROP_FAB_PORT_ENABLE(n)	PERF_PROP_FAB(n, 0x6)   /* RW */
+
+/*PORT hdr feature's properties*/
+#define PORT_HDR_PROP_REVISION			0x1	/* RDONLY */
+#define PORT_HDR_PROP_PORTIDX			0x2	/* RDONLY */
+#define PORT_HDR_PROP_LATENCY_TOLERANCE		0x3	/* RDONLY */
+#define PORT_HDR_PROP_AP1_EVENT			0x4	/* RW */
+#define PORT_HDR_PROP_AP2_EVENT			0x5	/* RW */
+#define PORT_HDR_PROP_POWER_STATE		0x6	/* RDONLY */
+#define PORT_HDR_PROP_USERCLK_FREQCMD		0x7	/* RW */
+#define PORT_HDR_PROP_USERCLK_FREQCNTRCMD	0x8	/* RW */
+#define PORT_HDR_PROP_USERCLK_FREQSTS		0x9	/* RDONLY */
+#define PORT_HDR_PROP_USERCLK_CNTRSTS		0xa	/* RDONLY */
+
+/*PORT error feature's properties*/
+#define PORT_ERR_PROP_REVISION			0x1	/* RDONLY */
+#define PORT_ERR_PROP_ERRORS			0x2	/* RDONLY */
+#define PORT_ERR_PROP_FIRST_ERROR		0x3	/* RDONLY */
+#define PORT_ERR_PROP_FIRST_MALFORMED_REQ_LSB	0x4	/* RDONLY */
+#define PORT_ERR_PROP_FIRST_MALFORMED_REQ_MSB	0x5	/* RDONLY */
+#define PORT_ERR_PROP_CLEAR			0x6	/* WRONLY */
+
+int opae_manager_ifpga_get_prop(struct opae_manager *mgr,
+				struct feature_prop *prop);
+int opae_manager_ifpga_set_prop(struct opae_manager *mgr,
+				struct feature_prop *prop);
+int opae_bridge_ifpga_get_prop(struct opae_bridge *br,
+			       struct feature_prop *prop);
+int opae_bridge_ifpga_set_prop(struct opae_bridge *br,
+			       struct feature_prop *prop);
+
+/*
+ * Retrieve information about the fpga fme.
+ * Driver fills the info in provided struct fpga_fme_info.
+ */
+struct fpga_fme_info {
+	u32 capability;		/* The capability of FME device */
+#define FPGA_FME_CAP_ERR_IRQ	(1 << 0) /* Support fme error interrupt */
+};
+
+int opae_manager_ifpga_get_info(struct opae_manager *mgr,
+				struct fpga_fme_info *fme_info);
+
+/* Set eventfd information for ifpga FME error interrupt */
+struct fpga_fme_err_irq_set {
+	s32 evtfd;		/* Eventfd handler */
+};
+
+int opae_manager_ifpga_set_err_irq(struct opae_manager *mgr,
+				   struct fpga_fme_err_irq_set *err_irq_set);
+
+/*
+ * Retrieve information about the fpga port.
+ * Driver fills the info in provided struct fpga_port_info.
+ */
+struct fpga_port_info {
+	u32 capability;	/* The capability of port device */
+#define FPGA_PORT_CAP_ERR_IRQ	(1 << 0) /* Support port error interrupt */
+#define FPGA_PORT_CAP_UAFU_IRQ	(1 << 1) /* Support uafu error interrupt */
+	u32 num_umsgs;	/* The number of allocated umsgs */
+	u32 num_uafu_irqs;	/* The number of uafu interrupts */
+};
+
+int opae_bridge_ifpga_get_info(struct opae_bridge *br,
+			       struct fpga_port_info *port_info);
+
+/* Set eventfd information for ifpga port error interrupt */
+struct fpga_port_err_irq_set {
+	s32 evtfd;		/* Eventfd handler */
+};
+
+int opae_bridge_ifpga_set_err_irq(struct opae_bridge *br,
+				  struct fpga_port_err_irq_set *err_irq_set);
+
+#endif /* _OPAE_IFPGA_HW_API_H_ */
diff --git a/drivers/raw/ifpga_rawdev/base/opae_osdep.h b/drivers/raw/ifpga_rawdev/base/opae_osdep.h
new file mode 100644
index 0000000..00d3736
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/opae_osdep.h
@@ -0,0 +1,111 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _OPAE_OSDEP_H
+#define _OPAE_OSDEP_H
+
+#include <string.h>
+#include <stdbool.h>
+
+#ifdef RTE_LIBRTE_EAL
+#include "osdep_rte/osdep_generic.h"
+#else
+#include "osdep_raw/osdep_generic.h"
+#endif
+
+#include <asm/types.h>
+
+#define __iomem
+
+typedef uint8_t		u8;
+typedef int8_t		s8;
+typedef uint16_t	u16;
+typedef uint32_t	u32;
+typedef int32_t		s32;
+typedef uint64_t	u64;
+typedef uint64_t	dma_addr_t;
+
+#ifndef LINUX_MACROS
+#ifndef BITS_PER_LONG
+#define BITS_PER_LONG	(__SIZEOF_LONG__ * 8)
+#endif
+#ifndef BIT
+#define BIT(a) (1UL << (a))
+#endif /* BIT */
+#ifndef BIT_ULL
+#define BIT_ULL(a) (1ULL << (a))
+#endif /* BIT_ULL */
+#ifndef GENMASK
+#define GENMASK(h, l)	(((~0UL) << (l)) & (~0UL >> (BITS_PER_LONG - 1 - (h))))
+#endif /* GENMASK */
+#ifndef GENMASK_ULL
+#define GENMASK_ULL(h, l) (((U64_C(1) << ((h) - (l) + 1)) - 1) << (l))
+#endif /* GENMASK_ULL */
+#endif /* LINUX_MACROS */
+
+#define SET_FIELD(m, v) (((v) << (__builtin_ffsll(m) - 1)) & (m))
+#define GET_FIELD(m, v) (((v) & (m)) >> (__builtin_ffsll(m) - 1))
+
+#define dev_err(x, args...) dev_printf(ERR, args)
+#define dev_info(x, args...) dev_printf(INFO, args)
+#define dev_warn(x, args...) dev_printf(WARNING, args)
+
+#ifdef OPAE_DEBUG
+#define dev_debug(x, args...) dev_printf(DEBUG, args)
+#else
+#define dev_debug(x, args...) do { } while (0)
+#endif
+
+#define pr_err(y, args...) dev_err(0, y, ##args)
+#define pr_warn(y, args...) dev_warn(0, y, ##args)
+#define pr_info(y, args...) dev_info(0, y, ##args)
+#define BUG() pr_err("BUG at %s:%d", __func__, __LINE__)
+
+#define ASSERT(x) do {\
+	if (!(x)) \
+		osdep_panic("osdep_panic: x"); \
+} while (0)
+#define BUG_ON(x) ASSERT(!(x))
+
+#ifndef WARN_ON
+#define WARN_ON(x) do { \
+	int ret = !!(x); \
+	if (unlikely(ret)) \
+		pr_warn("WARN_ON: \"" #x "\" at %s:%d\n", __func__, __LINE__); \
+} while (0)
+#endif
+
+#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+#define udelay(x) opae_udelay(x)
+#define msleep(x) opae_udelay(1000 * (x))
+#define usleep_range(min, max) msleep(DIV_ROUND_UP(min, 1000))
+
+#endif
diff --git a/drivers/raw/ifpga_rawdev/base/osdep_raw/osdep_generic.h b/drivers/raw/ifpga_rawdev/base/osdep_raw/osdep_generic.h
new file mode 100644
index 0000000..12ad018
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/osdep_raw/osdep_generic.h
@@ -0,0 +1,104 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _OSDEP_RAW_GENERIC_H
+#define _OSDEP_RAW_GENERIC_H
+
+#define	compiler_barrier() do {		\
+	asm volatile ("" : : : "memory");	\
+} while (0)
+
+#define io_wmb() compiler_barrier()
+#define io_rmb() compiler_barrier()
+
+static inline uint8_t opae_readb(const volatile void *addr)
+{
+	uint8_t val;
+
+	val = *(const volatile uint8_t *)addr;
+	io_rmb();
+	return val;
+}
+
+static inline uint16_t opae_readw(const volatile void *addr)
+{
+	uint16_t val;
+
+	val = *(const volatile uint16_t *)addr;
+	io_rmb();
+	return val;
+}
+
+static inline uint32_t opae_readl(const volatile void *addr)
+{
+	uint32_t val;
+
+	val = *(const volatile uint32_t *)addr;
+	io_rmb();
+	return val;
+}
+
+static inline uint64_t opae_readq(const volatile void *addr)
+{
+	uint64_t val;
+
+	val = *(const volatile uint64_t *)addr;
+	io_rmb();
+	return val;
+}
+
+static inline void opae_writeb(uint8_t value, volatile void *addr)
+{
+	io_wmb();
+	*(volatile uint8_t *)addr = value;
+}
+
+static inline void opae_writew(uint16_t value, volatile void *addr)
+{
+	io_wmb();
+	*(volatile uint16_t *)addr = value;
+}
+
+static inline void opae_writel(uint32_t value, volatile void *addr)
+{
+	io_wmb();
+	*(volatile uint32_t *)addr = value;
+}
+
+static inline void opae_writeq(uint64_t value, volatile void *addr)
+{
+	io_wmb();
+	*(volatile uint64_t *)addr = value;
+}
+
+#define opae_free(addr) free(addr)
+
+#endif
diff --git a/drivers/raw/ifpga_rawdev/base/osdep_rte/osdep_generic.h b/drivers/raw/ifpga_rawdev/base/osdep_rte/osdep_generic.h
new file mode 100644
index 0000000..9ff580e
--- /dev/null
+++ b/drivers/raw/ifpga_rawdev/base/osdep_rte/osdep_generic.h
@@ -0,0 +1,72 @@
+/*******************************************************************************
+ *
+ * Copyright (c) 2017, Intel 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:
+ *
+ *  1. Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *
+ *  2. 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.
+ *
+ *  3. Neither the name of the 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.
+ **************************************************************************/
+#ifndef _OSDEP_RTE_GENERIC_H
+#define _OSDEP_RTE_GENERIC_H
+
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_spinlock.h>
+#include <rte_log.h>
+#include <rte_io.h>
+#include <rte_malloc.h>
+
+#define dev_printf(level, fmt, args...) \
+	RTE_LOG(level, PMD, "osdep_rte: " fmt, ## args)
+
+#define osdep_panic(...) rte_panic(...)
+
+#define opae_udelay(x) rte_delay_us(x)
+
+#define opae_readb(addr) rte_read8(addr)
+#define opae_readw(addr) rte_read16(addr)
+#define opae_readl(addr) rte_read32(addr)
+#define opae_readq(addr) rte_read64(addr)
+#define opae_writeb(value, addr) rte_write8(value, addr)
+#define opae_writew(value, addr) rte_write16(value, addr)
+#define opae_writel(value, addr) rte_write32(value, addr)
+#define opae_writeq(value, addr) rte_write64(value, addr)
+
+#define opae_malloc(size) rte_malloc(NULL, size, 0)
+#define opae_zmalloc(size) rte_zmalloc(NULL, size, 0)
+#define opae_free(addr) rte_free(addr)
+
+#define ARRAY_SIZE(arr) RTE_DIM(arr)
+
+#define min(a, b) RTE_MIN(a, b)
+#define max(a, b) RTE_MAX(a, b)
+
+#define spinlock_t rte_spinlock_t
+#define spinlock_init(x) rte_spinlock_init(x)
+#define spinlock_lock(x) rte_spinlock_lock(x)
+#define spinlock_unlock(x) rte_spinlock_unlock(x)
+
+#endif
-- 
1.8.3.1

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code Rosen Xu
@ 2018-03-21  9:07   ` Shreyansh Jain
  2018-03-21  9:10     ` Shreyansh Jain
  2018-03-21 10:05   ` Gaëtan Rivet
  1 sibling, 1 reply; 19+ messages in thread
From: Shreyansh Jain @ 2018-03-21  9:07 UTC (permalink / raw)
  To: Rosen Xu
  Cc: dev, declan.doherty, Bruce Richardson, tianfei.zhang, hao.wu,
	gaetan.rivet

Hello Rosen,

On Wed, Mar 21, 2018 at 1:21 PM, Rosen Xu <rosen.xu@intel.com> wrote:
> Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> ---
>  lib/librte_eal/common/eal_common_bus.c | 14 +++++++++++++-
>  1 file changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/lib/librte_eal/common/eal_common_bus.c b/lib/librte_eal/common/eal_common_bus.c
> index 3e022d5..e3bcebe 100644
> --- a/lib/librte_eal/common/eal_common_bus.c
> +++ b/lib/librte_eal/common/eal_common_bus.c
> @@ -87,7 +87,7 @@ struct rte_bus_list rte_bus_list =
>  rte_bus_probe(void)
>  {
>         int ret;
> -       struct rte_bus *bus, *vbus = NULL;
> +       struct rte_bus *bus, *vbus = NULL, *ifpga_bus = NULL;
>
>         TAILQ_FOREACH(bus, &rte_bus_list, next) {
>                 if (!strcmp(bus->name, "vdev")) {
> @@ -95,6 +95,11 @@ struct rte_bus_list rte_bus_list =
>                         continue;
>                 }
>
> +               if (!strcmp(bus->name, "ifpga")) {
> +                       ifpga_bus = bus;
> +                       continue;
> +               }
> +
>                 ret = bus->probe();
>                 if (ret)
>                         RTE_LOG(ERR, EAL, "Bus (%s) probe failed.\n",
> @@ -108,6 +113,13 @@ struct rte_bus_list rte_bus_list =
>                                 vbus->name);
>         }
>
> +       if (ifpga_bus) {
> +               ret = ifpga_bus->probe();
> +               if (ret)
> +                       RTE_LOG(ERR, EAL, "Scan for (%s) bus failed.\n",
> +                               ifpga_bus->name);
> +       }
> +

Just like my comment on RFC, I still think this is not the right thing to do.
I understand you want a case where IFPGA bus gets probed only after
PCI bus is probed.
There can be multiple ways. Two of them which I can quickly list
without much deliberation:

1. A framework which can 'defer probing'
   So, a bus can register for defer probe and its
check_if_probe_available() function callback is called through
rte_bus_probe()
   If it returns OK, its probe is called, else it is added to a defer
list which is called once all first register buses are probed.

2. Modify the priority in RTE_REGISTER_BUS and make it as an argument
or a new variant which can take an argument.

It is not ok to change this function specifically for a bus is because
this method is not scalable.

Is there some specific reason you would like to stick to this approach?

[...]

-
Shreyansh

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code
  2018-03-21  9:07   ` Shreyansh Jain
@ 2018-03-21  9:10     ` Shreyansh Jain
  0 siblings, 0 replies; 19+ messages in thread
From: Shreyansh Jain @ 2018-03-21  9:10 UTC (permalink / raw)
  To: Rosen Xu
  Cc: dev, declan.doherty, Bruce Richardson, tianfei.zhang, hao.wu,
	gaetan.rivet

On Wed, Mar 21, 2018 at 2:37 PM, Shreyansh Jain <shreyansh.jain@nxp.com> wrote:
> Hello Rosen,
>
> On Wed, Mar 21, 2018 at 1:21 PM, Rosen Xu <rosen.xu@intel.com> wrote:
>> Signed-off-by: Rosen Xu <rosen.xu@intel.com>
>> ---
>>  lib/librte_eal/common/eal_common_bus.c | 14 +++++++++++++-
>>  1 file changed, 13 insertions(+), 1 deletion(-)
>>
>> diff --git a/lib/librte_eal/common/eal_common_bus.c b/lib/librte_eal/common/eal_common_bus.c
>> index 3e022d5..e3bcebe 100644
>> --- a/lib/librte_eal/common/eal_common_bus.c
>> +++ b/lib/librte_eal/common/eal_common_bus.c
>> @@ -87,7 +87,7 @@ struct rte_bus_list rte_bus_list =
>>  rte_bus_probe(void)
>>  {
>>         int ret;
>> -       struct rte_bus *bus, *vbus = NULL;
>> +       struct rte_bus *bus, *vbus = NULL, *ifpga_bus = NULL;
>>
>>         TAILQ_FOREACH(bus, &rte_bus_list, next) {
>>                 if (!strcmp(bus->name, "vdev")) {
>> @@ -95,6 +95,11 @@ struct rte_bus_list rte_bus_list =
>>                         continue;
>>                 }
>>
>> +               if (!strcmp(bus->name, "ifpga")) {
>> +                       ifpga_bus = bus;
>> +                       continue;
>> +               }
>> +
>>                 ret = bus->probe();
>>                 if (ret)
>>                         RTE_LOG(ERR, EAL, "Bus (%s) probe failed.\n",
>> @@ -108,6 +113,13 @@ struct rte_bus_list rte_bus_list =
>>                                 vbus->name);
>>         }
>>
>> +       if (ifpga_bus) {
>> +               ret = ifpga_bus->probe();
>> +               if (ret)
>> +                       RTE_LOG(ERR, EAL, "Scan for (%s) bus failed.\n",
>> +                               ifpga_bus->name);
>> +       }
>> +
>
> Just like my comment on RFC, I still think this is not the right thing to do.
> I understand you want a case where IFPGA bus gets probed only after
> PCI bus is probed.
> There can be multiple ways. Two of them which I can quickly list
> without much deliberation:
>
> 1. A framework which can 'defer probing'
>    So, a bus can register for defer probe and its
> check_if_probe_available() function callback is called through
> rte_bus_probe()
>    If it returns OK, its probe is called, else it is added to a defer
> list which is called once all first register buses are probed.
>
> 2. Modify the priority in RTE_REGISTER_BUS and make it as an argument
> or a new variant which can take an argument.
>
> It is not ok to change this function specifically for a bus is because
> this method is not scalable.

A cross on my own comment - I know vdev is already doing this special
probe but if we have a proper mechanism we can avoid that as well. Or,
continue to consider vdev as special :D

>
> Is there some specific reason you would like to stick to this approach?
>
> [...]
>
> -
> Shreyansh

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code Rosen Xu
@ 2018-03-21  9:28   ` Shreyansh Jain
  2018-03-21 10:20   ` Gaëtan Rivet
  1 sibling, 0 replies; 19+ messages in thread
From: Shreyansh Jain @ 2018-03-21  9:28 UTC (permalink / raw)
  To: Rosen Xu
  Cc: dev, declan.doherty, Bruce Richardson, tianfei.zhang, hao.wu,
	gaetan.rivet

Hello Rosen,

inlined are some comments from a quick look...

On Wed, Mar 21, 2018 at 1:21 PM, Rosen Xu <rosen.xu@intel.com> wrote:
> Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> ---
>  drivers/bus/ifpga/Makefile                  |  64 ++++
>  drivers/bus/ifpga/ifpga_bus.c               | 573 ++++++++++++++++++++++++++++
>  drivers/bus/ifpga/ifpga_common.c            | 154 ++++++++
>  drivers/bus/ifpga/ifpga_common.h            |  25 ++
>  drivers/bus/ifpga/ifpga_logs.h              |  32 ++
>  drivers/bus/ifpga/rte_bus_ifpga.h           | 141 +++++++
>  drivers/bus/ifpga/rte_bus_ifpga_version.map |   8 +
>  7 files changed, 997 insertions(+)
>  create mode 100644 drivers/bus/ifpga/Makefile
>  create mode 100644 drivers/bus/ifpga/ifpga_bus.c
>  create mode 100644 drivers/bus/ifpga/ifpga_common.c
>  create mode 100644 drivers/bus/ifpga/ifpga_common.h
>  create mode 100644 drivers/bus/ifpga/ifpga_logs.h
>  create mode 100644 drivers/bus/ifpga/rte_bus_ifpga.h
>  create mode 100644 drivers/bus/ifpga/rte_bus_ifpga_version.map
>
> diff --git a/drivers/bus/ifpga/Makefile b/drivers/bus/ifpga/Makefile
> new file mode 100644
> index 0000000..c71f186
> --- /dev/null
> +++ b/drivers/bus/ifpga/Makefile
> @@ -0,0 +1,64 @@
> +#   BSD LICENSE
> +#
> +#   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
> +#   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.

Did you get a chance to go through the comment in RFC?
I think you should replace the boilerplate with SPDX tags.

> +
> +include $(RTE_SDK)/mk/rte.vars.mk
> +
> +#
> +# library name
> +#
> +LIB = librte_bus_ifpga.a
> +LIBABIVER := 1
> +EXPORT_MAP := rte_bus_ifpga_version.map
> +
> +ifeq ($(CONFIG_RTE_LIBRTE_DPAA2_DEBUG_INIT),y)

I think there was a similar comment on RFC - "...DPAA2..." macro is a
copy-paste error.

> +CFLAGS += -O0 -g
> +CFLAGS += "-Wno-error"
> +else
> +CFLAGS += -O3
> +CFLAGS += $(WERROR_FLAGS)
> +endif
> +
> +CFLAGS += -I$(RTE_SDK)/drivers/bus/ifpga
> +CFLAGS += -I$(RTE_SDK)/drivers/bus/pci
> +CFLAGS += -I$(RTE_SDK)/lib/librte_eal/linuxapp/eal
> +CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
> +#CFLAGS += -I$(RTE_SDK)/lib/librte_rawdev
> +#LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring -lrte_rawdev

If you don't need these lines, don't keep them. That is ok until RFC,
but not in formal patch.

> +LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
> +#LDLIBS += -lrte_ethdev
> +
> +VPATH += $(SRCDIR)/base
> +
> +SRCS-y += \
> +        ifpga_bus.c \
> +        ifpga_common.c
> +
> +include $(RTE_SDK)/mk/rte.lib.mk
> diff --git a/drivers/bus/ifpga/ifpga_bus.c b/drivers/bus/ifpga/ifpga_bus.c
> new file mode 100644
> index 0000000..ff72b74
> --- /dev/null
> +++ b/drivers/bus/ifpga/ifpga_bus.c
> @@ -0,0 +1,573 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2010-2014 Intel Corporation.
> + * Copyright 2013-2014 6WIND S.A.

Are you sure of the above copyright? I think this is a new file. Maybe
your internal HW routines can have old copyrights.
Just a trivial comment, though.

> + */
> +
> +#include <string.h>
> +#include <inttypes.h>
> +#include <stdint.h>
> +#include <stdlib.h>
> +#include <stdio.h>
> +#include <sys/queue.h>
> +#include <sys/mman.h>
> +#include <sys/types.h>
> +#include <unistd.h>
> +#include <fcntl.h>
> +
> +#include <rte_errno.h>
> +#include <rte_bus.h>
> +#include <rte_per_lcore.h>
> +#include <rte_memory.h>
> +#include <rte_memzone.h>
> +#include <rte_eal.h>
> +#include <rte_common.h>
> +
> +#include <rte_devargs.h>
> +#include <rte_pci.h>
> +#include <rte_bus_pci.h>
> +#include <rte_kvargs.h>
> +#include <rte_alarm.h>
> +
> +#include "rte_rawdev.h"
> +#include "rte_rawdev_pmd.h"
> +#include "rte_bus_ifpga.h"
> +#include "ifpga_logs.h"
> +#include "ifpga_common.h"
> +
> +int ifpga_bus_logtype;
> +
> +/*register a ifpga bus based driver */

Comments have ' ' <space> after the '/*'.
Maybe you can refer [1] once.

[1] https://dpdk.org/doc/guides/contributing/coding_style.html#coding-style

> +void rte_ifpga_driver_register(struct rte_afu_driver *driver)
> +{
> +       RTE_VERIFY(driver);
> +
> +       TAILQ_INSERT_TAIL(&rte_ifpga_bus.driver_list, driver, next);
> +}
> +

[..snip..]

> diff --git a/drivers/bus/ifpga/ifpga_logs.h b/drivers/bus/ifpga/ifpga_logs.h
> new file mode 100644
> index 0000000..36b9b3f
> --- /dev/null
> +++ b/drivers/bus/ifpga/ifpga_logs.h
> @@ -0,0 +1,32 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2010-2015 Intel Corporation.
> + * Copyright 2013-2014 6WIND S.A.
> + */
> +
> +#ifndef _IFPGA_BUS_LOGS_H_
> +#define _IFPGA_BUS_LOGS_H_

Ideally this is name of the file, which is IFPFA_LOGS. But,
technically this is not an issue.

> +
> +#include <rte_log.h>
> +
> +extern int ifpga_bus_logtype;
> +
> +#define IFPGA_LOG(level, fmt, args...) \
> +       rte_log(RTE_LOG_ ## level, ifpga_bus_logtype, "%s(): " fmt "\n", \
> +               __func__, ##args)
> +
> +#define IFPGA_BUS_LOG(level, fmt, args...) \
> +       rte_log(RTE_LOG_ ## level, ifpga_bus_logtype, "%s(): " fmt "\n", \
> +               __func__, ##args)
> +

I noticed that at some places where you have used the above macros,
you have added '\n' in the call. It would lead to double '\n' as your
IFPGA_LOG and IFPGA_BUS_LOG already have one '\n'.

> +#define IFPGA_BUS_FUNC_TRACE() IFPGA_BUS_LOG(DEBUG, ">>")
> +
> +#define IFPGA_BUS_DEBUG(fmt, args...) \
> +       IFPGA_BUS_LOG(DEBUG, fmt, ## args)
> +#define IFPGA_BUS_INFO(fmt, args...) \
> +       IFPGA_BUS_LOG(INFO, fmt, ## args)
> +#define IFPGA_BUS_ERR(fmt, args...) \
> +       IFPGA_BUS_LOG(ERR, fmt, ## args)
> +#define IFPGA_BUS_WARN(fmt, args...) \
> +       IFPGA_BUS_LOG(WARNING, fmt, ## args)
> +
> +#endif /* _IFPGA_BUS_LOGS_H_ */

[..snip..]

> +#endif /* _RTE_BUS_IFPGA_H_ */
> diff --git a/drivers/bus/ifpga/rte_bus_ifpga_version.map b/drivers/bus/ifpga/rte_bus_ifpga_version.map
> new file mode 100644
> index 0000000..e2aa7da
> --- /dev/null
> +++ b/drivers/bus/ifpga/rte_bus_ifpga_version.map
> @@ -0,0 +1,8 @@
> +DPDK_17.11 {

Should be DPDK 18.05

> +       global:
> +
> +    rte_ifpga_driver_register;
> +    rte_ifpga_driver_unregister;

And indentation is incorrect.

> +
> +       local: *;
> +};
> --
> 1.8.3.1
>

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS
  2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
                   ` (4 preceding siblings ...)
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 5/5] Add Intel OPAE Share Code Rosen Xu
@ 2018-03-21 10:00 ` Gaëtan Rivet
  5 siblings, 0 replies; 19+ messages in thread
From: Gaëtan Rivet @ 2018-03-21 10:00 UTC (permalink / raw)
  To: Rosen Xu
  Cc: dev, declan.doherty, bruce.richardson, shreyansh.jain,
	tianfei.zhang, hao.wu

Hi Rosen,

Thanks for sending the new version.

However, you still did not use the --in-reply-to option to link this
thread to the previous one.

Please, do so in your future submission.

On Wed, Mar 21, 2018 at 03:51:29PM +0800, Rosen Xu wrote:
> Intel FPGA BUS in DPDK
> -------------------------
> 
> This patch set introduces Intel FPGA BUS support in DPDK.
> 
> Motivation
> ==========
> 
> FPGA is used more and more widely in Cloud and NFV, one primary reason is
> that FPGA not only provide ASIC performance but also it's more flexible 
> than ASIC. FPGA use Partial Reconfigure(PR) Parts of Bitstream to achieve 
> its flexibility. Another reason is that one FPGA can be shared 
> by different Users, and each User can use some of AFUs of One FPGA.
> 
> That means One FPGA Device Bitstream is divided into many Parts of 
> Bitstream(each Part of Bitstream is defined as AFU-Accelerated 
> Function Unit), and each AFU is a Hardware Acceleration Unit and 
> it can dynamically Reload respectively.
> 
> Scope
> =====
> 
> The Intel FPGA BUS implementation is target towards various FPGA Devices 
> use PR to provide many Acceleration Function. Specific PMDs may also 
> bind to its AFU. And Applications don't care they are using ASIC 
> Acceleration or FPGA AFU Acceleration.
> 
> Proposed Solution
> =================
>  - Involve Rawdev to take FPGA Partial Configuration(Download/PR)
>  - Defined FPGA-BUS for Acceleration Drivers of AFUs
>    - FPGA PCI Scan(1st Scan) follows DPDK UIO/VFIO PCI Scan Process, 
>      probe Intel FPGA Rawdev Driver
>    - AFU Scan(2nd Scan) bind DPDK Driver to FPGA Partial-Bitstream
> 
> Status
> =====
> With integrating Intel PSG FPGA Software Stack OPAE(Open Programmable 
> Acceleration Engine) Share Code, Intel FPGA BUS runs well in 
> Intel PSG FPGA Cards.
> 
> Patch set Information
> ====================
> 
> This patch set includes 6 patches:
> * 0 : Introduce the Intel FPGA BUS library and enable its compilation.
> * 1 : Adds command parse code, for start-up application 
>       with Intel FPGA BUS.
> * 2 : Adds Driver Probe Code, for AFU Drivers should probed 
>       after PCI Drivers.
> * 3 : Adds Intel FPGA BUS library code, for AFU Device scan
>       and AFU Drivers probe.
> * 4 : Adds a Intel FPGA rawdevice driver, for FPGA Device Management 
>       such as PR.
> * 5 : Adds Intel OPAE(Open Programmable Acceleration Engine) Share Code, 
>       it's Intel FPGA Software Stack.
> 
> Rosen Xu (5):
>   Add Intel FPGA BUS Command Parse Code
>   Add Intel FPGA BUS Probe Code
>   Add Intel FPGA BUS Lib Code
>   Add Intel FPGA BUS Rawdev Code
>   Add Intel OPAE Share Code
> 
>  drivers/bus/ifpga/Makefile                         |   64 +
>  drivers/bus/ifpga/ifpga_bus.c                      |  573 +++++++
>  drivers/bus/ifpga/ifpga_common.c                   |  154 ++
>  drivers/bus/ifpga/ifpga_common.h                   |   25 +
>  drivers/bus/ifpga/ifpga_logs.h                     |   32 +
>  drivers/bus/ifpga/rte_bus_ifpga.h                  |  141 ++
>  drivers/bus/ifpga/rte_bus_ifpga_version.map        |    8 +
>  drivers/raw/Makefile                               |    1 +
>  drivers/raw/ifpga_rawdev/Makefile                  |   63 +
>  drivers/raw/ifpga_rawdev/base/Makefile             |   54 +
>  drivers/raw/ifpga_rawdev/base/ifpga_api.c          |  543 +++++++
>  drivers/raw/ifpga_rawdev/base/ifpga_api.h          |   77 +
>  drivers/raw/ifpga_rawdev/base/ifpga_compat.h       |   84 +
>  drivers/raw/ifpga_rawdev/base/ifpga_defines.h      | 1696 ++++++++++++++++++++
>  drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c    |  861 ++++++++++
>  drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h    |   38 +
>  drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c  |  340 ++++
>  drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h  |  197 +++
>  drivers/raw/ifpga_rawdev/base/ifpga_fme.c          |  763 +++++++++
>  drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c    |  328 ++++
>  drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c    |  430 +++++
>  drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c    |  742 +++++++++
>  drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c       |  395 +++++
>  drivers/raw/ifpga_rawdev/base/ifpga_hw.h           |  152 ++
>  drivers/raw/ifpga_rawdev/base/ifpga_port.c         |  730 +++++++++
>  drivers/raw/ifpga_rawdev/base/ifpga_port_error.c   |  236 +++
>  drivers/raw/ifpga_rawdev/base/opae_debug.c         |  126 ++
>  drivers/raw/ifpga_rawdev/base/opae_debug.h         |   46 +
>  drivers/raw/ifpga_rawdev/base/opae_hw_api.c        |  389 +++++
>  drivers/raw/ifpga_rawdev/base/opae_hw_api.h        |  276 ++++
>  drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c  |  151 ++
>  drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h  |  293 ++++
>  drivers/raw/ifpga_rawdev/base/opae_osdep.h         |  111 ++
>  .../ifpga_rawdev/base/osdep_raw/osdep_generic.h    |  104 ++
>  .../ifpga_rawdev/base/osdep_rte/osdep_generic.h    |   72 +
>  drivers/raw/ifpga_rawdev/ifpga_rawdev.c            |  486 ++++++
>  drivers/raw/ifpga_rawdev/ifpga_rawdev.h            |   38 +
>  drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c    |   99 ++
>  .../ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map  |    4 +
>  lib/librte_eal/common/eal_common_bus.c             |   14 +-
>  lib/librte_eal/common/eal_common_options.c         |    8 +-
>  lib/librte_eal/common/eal_options.h                |    2 +
>  42 files changed, 10944 insertions(+), 2 deletions(-)
>  create mode 100644 drivers/bus/ifpga/Makefile
>  create mode 100644 drivers/bus/ifpga/ifpga_bus.c
>  create mode 100644 drivers/bus/ifpga/ifpga_common.c
>  create mode 100644 drivers/bus/ifpga/ifpga_common.h
>  create mode 100644 drivers/bus/ifpga/ifpga_logs.h
>  create mode 100644 drivers/bus/ifpga/rte_bus_ifpga.h
>  create mode 100644 drivers/bus/ifpga/rte_bus_ifpga_version.map
>  create mode 100644 drivers/raw/ifpga_rawdev/Makefile
>  create mode 100644 drivers/raw/ifpga_rawdev/base/Makefile
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_api.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_api.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_compat.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_defines.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_enumerate.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_feature_dev.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_dperf.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_error.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_iperf.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_fme_pr.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_hw.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_port.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/ifpga_port_error.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_debug.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_debug.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_hw_api.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_hw_api.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.c
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_ifpga_hw_api.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/opae_osdep.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/osdep_raw/osdep_generic.h
>  create mode 100644 drivers/raw/ifpga_rawdev/base/osdep_rte/osdep_generic.h
>  create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev.c
>  create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev.h
>  create mode 100644 drivers/raw/ifpga_rawdev/ifpga_rawdev_example.c
>  create mode 100644 drivers/raw/ifpga_rawdev/rte_pmd_ifpga_rawdev_version.map
> 
> -- 
> 1.8.3.1
> 

-- 
Gaëtan Rivet
6WIND

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code Rosen Xu
  2018-03-21  9:07   ` Shreyansh Jain
@ 2018-03-21 10:05   ` Gaëtan Rivet
  1 sibling, 0 replies; 19+ messages in thread
From: Gaëtan Rivet @ 2018-03-21 10:05 UTC (permalink / raw)
  To: Rosen Xu
  Cc: dev, declan.doherty, bruce.richardson, shreyansh.jain,
	tianfei.zhang, hao.wu

On Wed, Mar 21, 2018 at 03:51:31PM +0800, Rosen Xu wrote:
> Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> ---
>  lib/librte_eal/common/eal_common_bus.c | 14 +++++++++++++-
>  1 file changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/librte_eal/common/eal_common_bus.c b/lib/librte_eal/common/eal_common_bus.c
> index 3e022d5..e3bcebe 100644
> --- a/lib/librte_eal/common/eal_common_bus.c
> +++ b/lib/librte_eal/common/eal_common_bus.c
> @@ -87,7 +87,7 @@ struct rte_bus_list rte_bus_list =
>  rte_bus_probe(void)
>  {
>  	int ret;
> -	struct rte_bus *bus, *vbus = NULL;
> +	struct rte_bus *bus, *vbus = NULL, *ifpga_bus = NULL;
>  
>  	TAILQ_FOREACH(bus, &rte_bus_list, next) {
>  		if (!strcmp(bus->name, "vdev")) {
> @@ -95,6 +95,11 @@ struct rte_bus_list rte_bus_list =
>  			continue;
>  		}
>  
> +		if (!strcmp(bus->name, "ifpga")) {
> +			ifpga_bus = bus;
> +			continue;
> +		}
> +
>  		ret = bus->probe();
>  		if (ret)
>  			RTE_LOG(ERR, EAL, "Bus (%s) probe failed.\n",
> @@ -108,6 +113,13 @@ struct rte_bus_list rte_bus_list =
>  				vbus->name);
>  	}
>  
> +	if (ifpga_bus) {
> +		ret = ifpga_bus->probe();
> +		if (ret)
> +			RTE_LOG(ERR, EAL, "Scan for (%s) bus failed.\n",
> +				ifpga_bus->name);
> +	}
> +

I do not think this solution is generic and scalable anough to get into
common EAL code.

However, I do not think we need to come up with a dependency scheme
(callback registration to check whether proper conditions are met to
proceed with probing).

I think you have a hard-dependency on the PCI bus because you are simply
probing a PCI device. You do not need an additional bus for this.

-- 
Gaëtan Rivet
6WIND

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code Rosen Xu
  2018-03-21  9:28   ` Shreyansh Jain
@ 2018-03-21 10:20   ` Gaëtan Rivet
  2018-03-21 13:35     ` Bruce Richardson
  1 sibling, 1 reply; 19+ messages in thread
From: Gaëtan Rivet @ 2018-03-21 10:20 UTC (permalink / raw)
  To: Rosen Xu
  Cc: dev, declan.doherty, bruce.richardson, shreyansh.jain,
	tianfei.zhang, hao.wu

Hi,

I have had issues compiling a few things here, have you checked
build status before submitting?

On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> ---
>  drivers/bus/ifpga/Makefile                  |  64 ++++
>  drivers/bus/ifpga/ifpga_bus.c               | 573 ++++++++++++++++++++++++++++
>  drivers/bus/ifpga/ifpga_common.c            | 154 ++++++++
>  drivers/bus/ifpga/ifpga_common.h            |  25 ++
>  drivers/bus/ifpga/ifpga_logs.h              |  32 ++
>  drivers/bus/ifpga/rte_bus_ifpga.h           | 141 +++++++
>  drivers/bus/ifpga/rte_bus_ifpga_version.map |   8 +
>  7 files changed, 997 insertions(+)
>  create mode 100644 drivers/bus/ifpga/Makefile
>  create mode 100644 drivers/bus/ifpga/ifpga_bus.c
>  create mode 100644 drivers/bus/ifpga/ifpga_common.c
>  create mode 100644 drivers/bus/ifpga/ifpga_common.h
>  create mode 100644 drivers/bus/ifpga/ifpga_logs.h
>  create mode 100644 drivers/bus/ifpga/rte_bus_ifpga.h
>  create mode 100644 drivers/bus/ifpga/rte_bus_ifpga_version.map
> 
> diff --git a/drivers/bus/ifpga/Makefile b/drivers/bus/ifpga/Makefile
> new file mode 100644
> index 0000000..c71f186
> --- /dev/null
> +++ b/drivers/bus/ifpga/Makefile
> @@ -0,0 +1,64 @@
> +#   BSD LICENSE
> +#
> +#   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
> +#   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.

Please use SPDX tags here instead.

> +
> +include $(RTE_SDK)/mk/rte.vars.mk
> +
> +#
> +# library name
> +#
> +LIB = librte_bus_ifpga.a
> +LIBABIVER := 1
> +EXPORT_MAP := rte_bus_ifpga_version.map
> +
> +ifeq ($(CONFIG_RTE_LIBRTE_DPAA2_DEBUG_INIT),y)

C/C issue, you probably do not want to rely on a DPAA option.

> +CFLAGS += -O0 -g
> +CFLAGS += "-Wno-error"
> +else
> +CFLAGS += -O3
> +CFLAGS += $(WERROR_FLAGS)
> +endif
> +
> +CFLAGS += -I$(RTE_SDK)/drivers/bus/ifpga
> +CFLAGS += -I$(RTE_SDK)/drivers/bus/pci

It is not okay to directly include PCI elements.
Either you use the public PCI API, or you make the elements you need
public (with proper versioning), or you come up with a different scheme.

But you cannot directly include the PCI bus.

> +CFLAGS += -I$(RTE_SDK)/lib/librte_eal/linuxapp/eal
> +CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
> +#CFLAGS += -I$(RTE_SDK)/lib/librte_rawdev
> +#LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring -lrte_rawdev

Please clean-up those commented lines.

> +LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
> +#LDLIBS += -lrte_ethdev

Same here.

> +
> +VPATH += $(SRCDIR)/base
> +
> +SRCS-y += \
> +        ifpga_bus.c \
> +        ifpga_common.c
> +
> +include $(RTE_SDK)/mk/rte.lib.mk
> diff --git a/drivers/bus/ifpga/ifpga_bus.c b/drivers/bus/ifpga/ifpga_bus.c
> new file mode 100644
> index 0000000..ff72b74
> --- /dev/null
> +++ b/drivers/bus/ifpga/ifpga_bus.c
> @@ -0,0 +1,573 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2010-2014 Intel Corporation.
> + * Copyright 2013-2014 6WIND S.A.

Copy-paste error.

> + */
> +
> +#include <string.h>
> +#include <inttypes.h>
> +#include <stdint.h>
> +#include <stdlib.h>
> +#include <stdio.h>
> +#include <sys/queue.h>
> +#include <sys/mman.h>
> +#include <sys/types.h>
> +#include <unistd.h>
> +#include <fcntl.h>
> +
> +#include <rte_errno.h>
> +#include <rte_bus.h>
> +#include <rte_per_lcore.h>
> +#include <rte_memory.h>
> +#include <rte_memzone.h>
> +#include <rte_eal.h>
> +#include <rte_common.h>
> +
> +#include <rte_devargs.h>
> +#include <rte_pci.h>
> +#include <rte_bus_pci.h>
> +#include <rte_kvargs.h>
> +#include <rte_alarm.h>
> +
> +#include "rte_rawdev.h"
> +#include "rte_rawdev_pmd.h"
> +#include "rte_bus_ifpga.h"
> +#include "ifpga_logs.h"
> +#include "ifpga_common.h"
> +
> +int ifpga_bus_logtype;
> +
> +/*register a ifpga bus based driver */
> +void rte_ifpga_driver_register(struct rte_afu_driver *driver)
> +{
> +	RTE_VERIFY(driver);
> +
> +	TAILQ_INSERT_TAIL(&rte_ifpga_bus.driver_list, driver, next);
> +}
> +
> +/*un-register a fpga bus based driver */
> +void rte_ifpga_driver_unregister(struct rte_afu_driver *driver)
> +{
> +	TAILQ_REMOVE(&rte_ifpga_bus.driver_list, driver, next);
> +}
> +
> +static struct rte_ifpga_device *
> +ifpga_find_ifpga_dev(const struct rte_pci_addr *pci_addr)
> +{
> +	struct rte_ifpga_device *ifpga_dev = NULL;
> +
> +	TAILQ_FOREACH(ifpga_dev, &rte_ifpga_bus.ifpga_list, next) {
> +		if (!ifpga_pci_addr_cmp_1(&ifpga_dev->pci_addr, pci_addr))
> +			return ifpga_dev;
> +	}
> +	return NULL;
> +}
> +
> +static struct rte_afu_device *
> +ifpga_find_afu_dev(const struct rte_afu_id *afu_id)
> +{
> +	struct rte_afu_device *afu_dev = NULL;
> +
> +	TAILQ_FOREACH(afu_dev, &rte_ifpga_bus.afu_list, next) {
> +		if (!ifpga_afu_id_cmp(&afu_dev->id, afu_id))
> +			return afu_dev;
> +	}
> +	return NULL;
> +}
> +
> +static const char *valid_args[] = {
> +#define IFPGA_ARG_BDF          "bdf"
> +	IFPGA_ARG_BDF,
> +#define IFPGA_ARG_PORT         "port"
> +	IFPGA_ARG_PORT,
> +#define IFPGA_ARG_PATH         "path"
> +	IFPGA_ARG_PATH,
> +#define IFPGA_ARG_UUID_HIGH    "uuid_high"
> +	IFPGA_ARG_UUID_HIGH,
> +#define IFPGA_ARG_UUID_LOW     "uuid_low"
> +	IFPGA_ARG_UUID_LOW,
> +#define IFPGA_ARG_PR_ENABLE     "pr_enable"
> +	IFPGA_ARG_PR_ENABLE,
> +#define IFPGA_ARG_DEBUG         "debug"
> +	IFPGA_ARG_DEBUG,
> +	NULL
> +};
> +
> +/*
> + * Scan the content of the FPGA bus, and the devices in the devices
> + * list
> + */

So you seem to scan your bus by reading parameters
given to the --ifpga EAL option.

Can you justify why you cannot use the PCI bus, have your FPGA be probed
by a PCI driver, that would take those parameters as driver parameters,
and spawn raw devices (one per bitstream) as needed as a result?

I see no reason this is not feasible. Unless you duly justify this
approach, it seems unacceptable to me. You are subverting generic EAL
code to bend things to your approach, without clear rationale.

Best regards,
-- 
Gaëtan Rivet
6WIND

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 10:20   ` Gaëtan Rivet
@ 2018-03-21 13:35     ` Bruce Richardson
  2018-03-21 14:02       ` Shreyansh Jain
                         ` (2 more replies)
  0 siblings, 3 replies; 19+ messages in thread
From: Bruce Richardson @ 2018-03-21 13:35 UTC (permalink / raw)
  To: Gaëtan Rivet
  Cc: Rosen Xu, dev, declan.doherty, shreyansh.jain, tianfei.zhang, hao.wu

On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaëtan Rivet wrote:
> Hi,
> 
> I have had issues compiling a few things here, have you checked
> build status before submitting?
> 
> On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> > ---
<snip>
> +
> > +/*
> > + * Scan the content of the FPGA bus, and the devices in the devices
> > + * list
> > + */
> 
> So you seem to scan your bus by reading parameters
> given to the --ifpga EAL option.
> 
> Can you justify why you cannot use the PCI bus, have your FPGA be probed
> by a PCI driver, that would take those parameters as driver parameters,
> and spawn raw devices (one per bitstream) as needed as a result?
> 
> I see no reason this is not feasible. Unless you duly justify this
> approach, it seems unacceptable to me. You are subverting generic EAL
> code to bend things to your approach, without clear rationale.
> 

While I agree with the comments in other emails about avoiding
special-cases in the code that makes things not-scalable, I would take the
view that using a bus-type is the correct choice for this. While you could
have a single device that creates other devices, that is also true for all
other buses as well.  [Furthermore, I think it is incorrect assume that all
devices on the FPGA bus would be raw devices, it's entirely possible to
have cryptodevs, bbdevs or compress devs implemented in the AFUs].

Consider what a bus driver provides: it's a generic mechanism for scanning
for devices - which all use a common connection method - for DPDK use, and
mapping device drivers to those devices. For an FPGA device which presents
multiple AFUs, this seems to be exactly what is required - a device driver
to scan for devices and present them to DPDK. The FPGA bus driver will have
to check each AFU and match it against the set of registered AFU device
drivers to ensure that the crypto AFU gets the cryptodev driver, etc.

Logically, therefore, it is a bus - which just happens to be a sub-bus of
PCI, i.e. presented as a PCI device. Consider also that it may be possible
or even desirable, to use blacklisting and whitelisting for those AFU
devices so that some AFUs could be used by one app, while others by
another. If we just have a single PCI device, I think we'll find ourselves
duplicating a lot of bus-related functionality inside the driver in that
case.

Regards,
/Bruce

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 13:35     ` Bruce Richardson
@ 2018-03-21 14:02       ` Shreyansh Jain
  2018-03-21 14:06       ` Xu, Rosen
  2018-03-21 14:14       ` Gaëtan Rivet
  2 siblings, 0 replies; 19+ messages in thread
From: Shreyansh Jain @ 2018-03-21 14:02 UTC (permalink / raw)
  To: Bruce Richardson
  Cc: Gaëtan Rivet, Rosen Xu, dev, declan.doherty, tianfei.zhang, hao.wu

On Wed, Mar 21, 2018 at 7:05 PM, Bruce Richardson
<bruce.richardson@intel.com> wrote:
> On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaėtan Rivet wrote:
>> Hi,
>>
>> I have had issues compiling a few things here, have you checked
>> build status before submitting?
>>
>> On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
>> > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
>> > ---
> <snip>
>> +
>> > +/*
>> > + * Scan the content of the FPGA bus, and the devices in the devices
>> > + * list
>> > + */
>>
>> So you seem to scan your bus by reading parameters
>> given to the --ifpga EAL option.
>>
>> Can you justify why you cannot use the PCI bus, have your FPGA be probed
>> by a PCI driver, that would take those parameters as driver parameters,
>> and spawn raw devices (one per bitstream) as needed as a result?
>>
>> I see no reason this is not feasible. Unless you duly justify this
>> approach, it seems unacceptable to me. You are subverting generic EAL
>> code to bend things to your approach, without clear rationale.
>>
>
> While I agree with the comments in other emails about avoiding
> special-cases in the code that makes things not-scalable, I would take the
> view that using a bus-type is the correct choice for this. While you could
> have a single device that creates other devices, that is also true for all
> other buses as well.  [Furthermore, I think it is incorrect assume that all
> devices on the FPGA bus would be raw devices, it's entirely possible to
> have cryptodevs, bbdevs or compress devs implemented in the AFUs].
>
> Consider what a bus driver provides: it's a generic mechanism for scanning
> for devices - which all use a common connection method - for DPDK use, and
> mapping device drivers to those devices. For an FPGA device which presents
> multiple AFUs, this seems to be exactly what is required - a device driver
> to scan for devices and present them to DPDK. The FPGA bus driver will have
> to check each AFU and match it against the set of registered AFU device
> drivers to ensure that the crypto AFU gets the cryptodev driver, etc.

In my opinion, its' not a problem that a heirarchichal bus model like
this is implemented in DPDK. That's being creative :)

But, here the issue is that same work can be done by a driver which
can create devices of multiple types (bbdev, cryptodev, ethdev etc)
and then use hotplugging (assuming that is supported by relevant
driver) over respective bus. I don't think we necessarily need a bus
for that. Is there some technical hurdle in following that model? Or,
is there some cases which cannot be handled by not having a dependency
bus?

Then, there is another problem of modifying the rte_bus code so that
this bus initialization can be delayed. That is something which is not
right in my opinion. That puts a fixed constraint on the EAL to look
for a bus as being special (like vdev). If that is a blocking problem,
it too needs to be solved - maybe separately.

Frankly, its not about 'can't-do-this' - but, I think this does
warrant a good thought before being mainlined.

>
> Logically, therefore, it is a bus - which just happens to be a sub-bus of
> PCI, i.e. presented as a PCI device. Consider also that it may be possible
> or even desirable, to use blacklisting and whitelisting for those AFU
> devices so that some AFUs could be used by one app, while others by
> another. If we just have a single PCI device, I think we'll find ourselves
> duplicating a lot of bus-related functionality inside the driver in that
> case.
>
> Regards,
> /Bruce

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 13:35     ` Bruce Richardson
  2018-03-21 14:02       ` Shreyansh Jain
@ 2018-03-21 14:06       ` Xu, Rosen
  2018-03-21 14:14       ` Gaëtan Rivet
  2 siblings, 0 replies; 19+ messages in thread
From: Xu, Rosen @ 2018-03-21 14:06 UTC (permalink / raw)
  To: Richardson, Bruce, Ga?tan Rivet
  Cc: dev, Doherty, Declan, shreyansh.jain, Zhang, Tianfei, Wu, Hao



-----Original Message-----
From: Richardson, Bruce 
Sent: Wednesday, March 21, 2018 21:35
To: Gaëtan Rivet <gaetan.rivet@6wind.com>
Cc: Xu, Rosen <rosen.xu@intel.com>; dev@dpdk.org; Doherty, Declan <declan.doherty@intel.com>; shreyansh.jain@nxp.com; Zhang, Tianfei <tianfei.zhang@intel.com>; Wu, Hao <hao.wu@intel.com>
Subject: Re: [PATCH V2 3/5] Add Intel FPGA BUS Lib Code

On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaëtan Rivet wrote:
> > Hi,
> > 
> > I have had issues compiling a few things here, have you checked build 
> > status before submitting?
> > 
> > On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> > > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> > > ---
<snip>
> > +
> > > +/*
> > > + * Scan the content of the FPGA bus, and the devices in the devices
> > > + * list
> > > + */
> > 
> > So you seem to scan your bus by reading parameters given to the 
> > --ifpga EAL option.
> > 
> > Can you justify why you cannot use the PCI bus, have your FPGA be 
> > probed by a PCI driver, that would take those parameters as driver 
> > parameters, and spawn raw devices (one per bitstream) as needed as a result?
> > 
> > I see no reason this is not feasible. Unless you duly justify this 
> > approach, it seems unacceptable to me. You are subverting generic EAL 
> > code to bend things to your approach, without clear rationale.
> > 
> While I agree with the comments in other emails about avoiding special-cases in the code 
>  that makes things not-scalable, I would take the view that using a bus-type is the correct 
> choice for this. While you could have a single device that creates other devices, that is also
> true for all other buses as well.  [Furthermore, I think it is incorrect assume that all devices 
> on the FPGA bus would be raw devices, it's entirely possible to have cryptodevs, bbdevs or compress devs implemented in the AFUs].
>
> Consider what a bus driver provides: it's a generic mechanism for scanning for 
> devices - which all use a common connection method - for DPDK use, and 
> mapping device drivers to those devices. For an FPGA device which presents 
> multiple AFUs, this seems to be exactly what is required - a device driver to 
> scan for devices and present them to DPDK. The FPGA bus driver will have 
> to check each AFU and match it against the set of registered AFU device 
> drivers to ensure that the crypto AFU gets the cryptodev driver, etc.
>
> Logically, therefore, it is a bus - which just happens to be a sub-bus of 
> PCI, i.e. presented as a PCI device. Consider also that it may be possible 
> or even desirable, to use blacklisting and whitelisting for those AFU 
> devices so that some AFUs could be used by one app, while others by 
> another. If we just have a single PCI device, I think we'll find ourselves 
> duplicating a lot of bus-related functionality inside the driver in that case.

In our FPGA Usage Framework, each FPAG Bitstream is divided 2 parts, 
one part is only one Blue AFU, another part consist of many Green AFUs.

Blue AFU includes PCIe Interface and FPGA PR Unit, Blue AFU is fixed after
OS Initialization, because if we change Blue AFU, the PCIe Interface needed to be 
rescanned and the OS need to reboot.

Green AFUs can be dynamically PR by different Users.

The benefit of this FPGA Architecture is that we can dynamically change Green AFUs,
but OS don't need to rescan the FPGA PCIe Interface. For Cloud Scenario the FPGA 
Device can be viewed a common resource pools such as DDR Memory, and it can be easily
assigned to different users at some time. For TelCom/NFV Scenario, we can easily
upgrade the Acceleration AFU but the server don't need to reboot.

For Software Usage, there many FPGA Device in one system, some have the same
Green AFUs and some have different Green AFUs, same Green AFUs means same Acceleration, 
and it will use same driver. So we want to involve a new bus,
which can easily bind the same Green AFUs with its driver.


Regards,
/Bruce

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 13:35     ` Bruce Richardson
  2018-03-21 14:02       ` Shreyansh Jain
  2018-03-21 14:06       ` Xu, Rosen
@ 2018-03-21 14:14       ` Gaëtan Rivet
  2018-03-21 14:31         ` Gaëtan Rivet
  2 siblings, 1 reply; 19+ messages in thread
From: Gaëtan Rivet @ 2018-03-21 14:14 UTC (permalink / raw)
  To: Bruce Richardson
  Cc: Rosen Xu, dev, declan.doherty, shreyansh.jain, tianfei.zhang, hao.wu

Hi Bruce,

On Wed, Mar 21, 2018 at 01:35:09PM +0000, Bruce Richardson wrote:
> On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaëtan Rivet wrote:
> > Hi,
> > 
> > I have had issues compiling a few things here, have you checked
> > build status before submitting?
> > 
> > On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> > > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> > > ---
> <snip>
> > +
> > > +/*
> > > + * Scan the content of the FPGA bus, and the devices in the devices
> > > + * list
> > > + */
> > 
> > So you seem to scan your bus by reading parameters
> > given to the --ifpga EAL option.
> > 
> > Can you justify why you cannot use the PCI bus, have your FPGA be probed
> > by a PCI driver, that would take those parameters as driver parameters,
> > and spawn raw devices (one per bitstream) as needed as a result?
> > 
> > I see no reason this is not feasible. Unless you duly justify this
> > approach, it seems unacceptable to me. You are subverting generic EAL
> > code to bend things to your approach, without clear rationale.
> > 
> 
> While I agree with the comments in other emails about avoiding
> special-cases in the code that makes things not-scalable, I would take the
> view that using a bus-type is the correct choice for this. While you could
> have a single device that creates other devices, that is also true for all
> other buses as well.  [Furthermore, I think it is incorrect assume that all
> devices on the FPGA bus would be raw devices, it's entirely possible to
> have cryptodevs, bbdevs or compress devs implemented in the AFUs].
> 
> Consider what a bus driver provides: it's a generic mechanism for scanning
> for devices - which all use a common connection method - for DPDK use, and
> mapping device drivers to those devices. For an FPGA device which presents
> multiple AFUs, this seems to be exactly what is required - a device driver
> to scan for devices and present them to DPDK. The FPGA bus driver will have
> to check each AFU and match it against the set of registered AFU device
> drivers to ensure that the crypto AFU gets the cryptodev driver, etc.
> 
> Logically, therefore, it is a bus - which just happens to be a sub-bus of
> PCI, i.e. presented as a PCI device. Consider also that it may be possible
> or even desirable, to use blacklisting and whitelisting for those AFU
> devices so that some AFUs could be used by one app, while others by
> another. If we just have a single PCI device, I think we'll find ourselves
> duplicating a lot of bus-related functionality inside the driver in that
> case.
> 
> Regards,
> /Bruce

It makes sense indeed if you need to specialize several drivers specific
to AFU mappings.

So, taking the bus approach, it seems we almost all agree that the
current probe is not good enough.

I think you will run into issues if you registered your bus upon probing
a PCI driver. Instead, would it be possible to:

  * not add the EAL option --ifpga.
  * register the ifpga bus like any other.
  * have a PCI driver that hotplug an IFPGA device, triggering a
    scan and probe on the ifpga bus using the PCI device parameters.
    Essentially you would find there the parameters that could be
    given previously to the --ifpga option.

There should not be any EAL modification necessary.

The only downside I see would be that having both IFPGA device (for
triggering bus ops) and afterward using AFU devices on this bus could be
confusing. However, it seems safer and still cleaner overall.

-- 
Gaëtan Rivet
6WIND

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 14:14       ` Gaëtan Rivet
@ 2018-03-21 14:31         ` Gaëtan Rivet
  2018-03-21 15:41           ` Bruce Richardson
  0 siblings, 1 reply; 19+ messages in thread
From: Gaëtan Rivet @ 2018-03-21 14:31 UTC (permalink / raw)
  To: Bruce Richardson
  Cc: Rosen Xu, dev, declan.doherty, shreyansh.jain, tianfei.zhang, hao.wu

On Wed, Mar 21, 2018 at 03:14:05PM +0100, Gaëtan Rivet wrote:
> Hi Bruce,
> 
> On Wed, Mar 21, 2018 at 01:35:09PM +0000, Bruce Richardson wrote:
> > On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaëtan Rivet wrote:
> > > Hi,
> > > 
> > > I have had issues compiling a few things here, have you checked
> > > build status before submitting?
> > > 
> > > On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> > > > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> > > > ---
> > <snip>
> > > +
> > > > +/*
> > > > + * Scan the content of the FPGA bus, and the devices in the devices
> > > > + * list
> > > > + */
> > > 
> > > So you seem to scan your bus by reading parameters
> > > given to the --ifpga EAL option.
> > > 
> > > Can you justify why you cannot use the PCI bus, have your FPGA be probed
> > > by a PCI driver, that would take those parameters as driver parameters,
> > > and spawn raw devices (one per bitstream) as needed as a result?
> > > 
> > > I see no reason this is not feasible. Unless you duly justify this
> > > approach, it seems unacceptable to me. You are subverting generic EAL
> > > code to bend things to your approach, without clear rationale.
> > > 
> > 
> > While I agree with the comments in other emails about avoiding
> > special-cases in the code that makes things not-scalable, I would take the
> > view that using a bus-type is the correct choice for this. While you could
> > have a single device that creates other devices, that is also true for all
> > other buses as well.  [Furthermore, I think it is incorrect assume that all
> > devices on the FPGA bus would be raw devices, it's entirely possible to
> > have cryptodevs, bbdevs or compress devs implemented in the AFUs].
> > 
> > Consider what a bus driver provides: it's a generic mechanism for scanning
> > for devices - which all use a common connection method - for DPDK use, and
> > mapping device drivers to those devices. For an FPGA device which presents
> > multiple AFUs, this seems to be exactly what is required - a device driver
> > to scan for devices and present them to DPDK. The FPGA bus driver will have
> > to check each AFU and match it against the set of registered AFU device
> > drivers to ensure that the crypto AFU gets the cryptodev driver, etc.
> > 
> > Logically, therefore, it is a bus - which just happens to be a sub-bus of
> > PCI, i.e. presented as a PCI device. Consider also that it may be possible
> > or even desirable, to use blacklisting and whitelisting for those AFU
> > devices so that some AFUs could be used by one app, while others by
> > another. If we just have a single PCI device, I think we'll find ourselves
> > duplicating a lot of bus-related functionality inside the driver in that
> > case.
> > 
> > Regards,
> > /Bruce
> 
> It makes sense indeed if you need to specialize several drivers specific
> to AFU mappings.
> 
> So, taking the bus approach, it seems we almost all agree that the
> current probe is not good enough.
> 
> I think you will run into issues if you registered your bus upon probing
> a PCI driver. Instead, would it be possible to:
> 
>   * not add the EAL option --ifpga.
>   * register the ifpga bus like any other.
>   * have a PCI driver that hotplug an IFPGA device, triggering a
>     scan and probe on the ifpga bus using the PCI device parameters.
>     Essentially you would find there the parameters that could be
>     given previously to the --ifpga option.
> 
> There should not be any EAL modification necessary.
> 
> The only downside I see would be that having both IFPGA device (for
> triggering bus ops) and afterward using AFU devices on this bus could be
> confusing. However, it seems safer and still cleaner overall.

Actually, just to build upon this remark, and thinking a little more
about it:

You would not need an additional "special" type of ifpga devices.
Upon attempting to hotplug an ifpga device, you would have a new
rte_devargs having the parameters within inserted, addressed to your
bus.

During scan, you would find this devargs, read the relevant parameters
and parse them (exactly like it is now implemented). You would then be
able to spawn as needed any number of rte_afu_device within your bus and
the scan would be complete.

The only trickery would be afterward, because you'd need at least one
device that would be detached and having the name given in the devargs.
So you would have to name your rte_afu_device after the devargs passed
as parameter to the plug, essentially having all of them using the same name.

Nothing prevents you from renaming your AFU devices after bus->plug, but
the EAL will rely on it for matching and knowing the scan went ok.

-- 
Gaëtan Rivet
6WIND

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 14:31         ` Gaëtan Rivet
@ 2018-03-21 15:41           ` Bruce Richardson
  2018-03-21 16:21             ` Gaëtan Rivet
  0 siblings, 1 reply; 19+ messages in thread
From: Bruce Richardson @ 2018-03-21 15:41 UTC (permalink / raw)
  To: Gaëtan Rivet
  Cc: Rosen Xu, dev, declan.doherty, shreyansh.jain, tianfei.zhang, hao.wu

On Wed, Mar 21, 2018 at 03:31:31PM +0100, Gaëtan Rivet wrote:
> On Wed, Mar 21, 2018 at 03:14:05PM +0100, Gaëtan Rivet wrote:
> > Hi Bruce,
> > 
> > On Wed, Mar 21, 2018 at 01:35:09PM +0000, Bruce Richardson wrote:
> > > On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaëtan Rivet wrote:
> > > > Hi,
> > > > 
> > > > I have had issues compiling a few things here, have you checked
> > > > build status before submitting?
> > > > 
> > > > On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> > > > > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> > > > > ---
> > > <snip>
> > > > +
> > > > > +/*
> > > > > + * Scan the content of the FPGA bus, and the devices in the devices
> > > > > + * list
> > > > > + */
> > > > 
> > > > So you seem to scan your bus by reading parameters
> > > > given to the --ifpga EAL option.
> > > > 
> > > > Can you justify why you cannot use the PCI bus, have your FPGA be probed
> > > > by a PCI driver, that would take those parameters as driver parameters,
> > > > and spawn raw devices (one per bitstream) as needed as a result?
> > > > 
> > > > I see no reason this is not feasible. Unless you duly justify this
> > > > approach, it seems unacceptable to me. You are subverting generic EAL
> > > > code to bend things to your approach, without clear rationale.
> > > > 
> > > 
> > > While I agree with the comments in other emails about avoiding
> > > special-cases in the code that makes things not-scalable, I would take the
> > > view that using a bus-type is the correct choice for this. While you could
> > > have a single device that creates other devices, that is also true for all
> > > other buses as well.  [Furthermore, I think it is incorrect assume that all
> > > devices on the FPGA bus would be raw devices, it's entirely possible to
> > > have cryptodevs, bbdevs or compress devs implemented in the AFUs].
> > > 
> > > Consider what a bus driver provides: it's a generic mechanism for scanning
> > > for devices - which all use a common connection method - for DPDK use, and
> > > mapping device drivers to those devices. For an FPGA device which presents
> > > multiple AFUs, this seems to be exactly what is required - a device driver
> > > to scan for devices and present them to DPDK. The FPGA bus driver will have
> > > to check each AFU and match it against the set of registered AFU device
> > > drivers to ensure that the crypto AFU gets the cryptodev driver, etc.
> > > 
> > > Logically, therefore, it is a bus - which just happens to be a sub-bus of
> > > PCI, i.e. presented as a PCI device. Consider also that it may be possible
> > > or even desirable, to use blacklisting and whitelisting for those AFU
> > > devices so that some AFUs could be used by one app, while others by
> > > another. If we just have a single PCI device, I think we'll find ourselves
> > > duplicating a lot of bus-related functionality inside the driver in that
> > > case.
> > > 
> > > Regards,
> > > /Bruce
> > 
> > It makes sense indeed if you need to specialize several drivers specific
> > to AFU mappings.
> > 
> > So, taking the bus approach, it seems we almost all agree that the
> > current probe is not good enough.
> > 
> > I think you will run into issues if you registered your bus upon probing
> > a PCI driver. Instead, would it be possible to:
> > 
> >   * not add the EAL option --ifpga.
> >   * register the ifpga bus like any other.
> >   * have a PCI driver that hotplug an IFPGA device, triggering a
> >     scan and probe on the ifpga bus using the PCI device parameters.
> >     Essentially you would find there the parameters that could be
> >     given previously to the --ifpga option.
> > 
> > There should not be any EAL modification necessary.
> > 
> > The only downside I see would be that having both IFPGA device (for
> > triggering bus ops) and afterward using AFU devices on this bus could be
> > confusing. However, it seems safer and still cleaner overall.
> 
> Actually, just to build upon this remark, and thinking a little more
> about it:
> 
> You would not need an additional "special" type of ifpga devices.
> Upon attempting to hotplug an ifpga device, you would have a new
> rte_devargs having the parameters within inserted, addressed to your
> bus.
> 
> During scan, you would find this devargs, read the relevant parameters
> and parse them (exactly like it is now implemented). You would then be
> able to spawn as needed any number of rte_afu_device within your bus and
> the scan would be complete.
> 
> The only trickery would be afterward, because you'd need at least one
> device that would be detached and having the name given in the devargs.
> So you would have to name your rte_afu_device after the devargs passed
> as parameter to the plug, essentially having all of them using the same name.
> 
> Nothing prevents you from renaming your AFU devices after bus->plug, but
> the EAL will rely on it for matching and knowing the scan went ok.
> 
> -- 
Thanks, Gaëtan.

So, just to confirm I understand the suggestion correctly, you are
proposing that we create and use the FPGA bus as normal, except that the
scan will not report any devices. Then when an FPGA PCI device is
discovered, we use the hotplug infrastructure to trigger a real scan of the
FPGA for AFU devices, and do the whole driver mapping etc. process then.
Is that the basic idea you suggest?

/Bruce

^ permalink raw reply	[flat|nested] 19+ messages in thread

* Re: [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code
  2018-03-21 15:41           ` Bruce Richardson
@ 2018-03-21 16:21             ` Gaëtan Rivet
  0 siblings, 0 replies; 19+ messages in thread
From: Gaëtan Rivet @ 2018-03-21 16:21 UTC (permalink / raw)
  To: Bruce Richardson
  Cc: Rosen Xu, dev, declan.doherty, shreyansh.jain, tianfei.zhang, hao.wu

On Wed, Mar 21, 2018 at 03:41:56PM +0000, Bruce Richardson wrote:
> On Wed, Mar 21, 2018 at 03:31:31PM +0100, Gaëtan Rivet wrote:
> > On Wed, Mar 21, 2018 at 03:14:05PM +0100, Gaëtan Rivet wrote:
> > > Hi Bruce,
> > > 
> > > On Wed, Mar 21, 2018 at 01:35:09PM +0000, Bruce Richardson wrote:
> > > > On Wed, Mar 21, 2018 at 11:20:25AM +0100, Gaëtan Rivet wrote:
> > > > > Hi,
> > > > > 
> > > > > I have had issues compiling a few things here, have you checked
> > > > > build status before submitting?
> > > > > 
> > > > > On Wed, Mar 21, 2018 at 03:51:32PM +0800, Rosen Xu wrote:
> > > > > > Signed-off-by: Rosen Xu <rosen.xu@intel.com>
> > > > > > ---
> > > > <snip>
> > > > > +
> > > > > > +/*
> > > > > > + * Scan the content of the FPGA bus, and the devices in the devices
> > > > > > + * list
> > > > > > + */
> > > > > 
> > > > > So you seem to scan your bus by reading parameters
> > > > > given to the --ifpga EAL option.
> > > > > 
> > > > > Can you justify why you cannot use the PCI bus, have your FPGA be probed
> > > > > by a PCI driver, that would take those parameters as driver parameters,
> > > > > and spawn raw devices (one per bitstream) as needed as a result?
> > > > > 
> > > > > I see no reason this is not feasible. Unless you duly justify this
> > > > > approach, it seems unacceptable to me. You are subverting generic EAL
> > > > > code to bend things to your approach, without clear rationale.
> > > > > 
> > > > 
> > > > While I agree with the comments in other emails about avoiding
> > > > special-cases in the code that makes things not-scalable, I would take the
> > > > view that using a bus-type is the correct choice for this. While you could
> > > > have a single device that creates other devices, that is also true for all
> > > > other buses as well.  [Furthermore, I think it is incorrect assume that all
> > > > devices on the FPGA bus would be raw devices, it's entirely possible to
> > > > have cryptodevs, bbdevs or compress devs implemented in the AFUs].
> > > > 
> > > > Consider what a bus driver provides: it's a generic mechanism for scanning
> > > > for devices - which all use a common connection method - for DPDK use, and
> > > > mapping device drivers to those devices. For an FPGA device which presents
> > > > multiple AFUs, this seems to be exactly what is required - a device driver
> > > > to scan for devices and present them to DPDK. The FPGA bus driver will have
> > > > to check each AFU and match it against the set of registered AFU device
> > > > drivers to ensure that the crypto AFU gets the cryptodev driver, etc.
> > > > 
> > > > Logically, therefore, it is a bus - which just happens to be a sub-bus of
> > > > PCI, i.e. presented as a PCI device. Consider also that it may be possible
> > > > or even desirable, to use blacklisting and whitelisting for those AFU
> > > > devices so that some AFUs could be used by one app, while others by
> > > > another. If we just have a single PCI device, I think we'll find ourselves
> > > > duplicating a lot of bus-related functionality inside the driver in that
> > > > case.
> > > > 
> > > > Regards,
> > > > /Bruce
> > > 
> > > It makes sense indeed if you need to specialize several drivers specific
> > > to AFU mappings.
> > > 
> > > So, taking the bus approach, it seems we almost all agree that the
> > > current probe is not good enough.
> > > 
> > > I think you will run into issues if you registered your bus upon probing
> > > a PCI driver. Instead, would it be possible to:
> > > 
> > >   * not add the EAL option --ifpga.
> > >   * register the ifpga bus like any other.
> > >   * have a PCI driver that hotplug an IFPGA device, triggering a
> > >     scan and probe on the ifpga bus using the PCI device parameters.
> > >     Essentially you would find there the parameters that could be
> > >     given previously to the --ifpga option.
> > > 
> > > There should not be any EAL modification necessary.
> > > 
> > > The only downside I see would be that having both IFPGA device (for
> > > triggering bus ops) and afterward using AFU devices on this bus could be
> > > confusing. However, it seems safer and still cleaner overall.
> > 
> > Actually, just to build upon this remark, and thinking a little more
> > about it:
> > 
> > You would not need an additional "special" type of ifpga devices.
> > Upon attempting to hotplug an ifpga device, you would have a new
> > rte_devargs having the parameters within inserted, addressed to your
> > bus.
> > 
> > During scan, you would find this devargs, read the relevant parameters
> > and parse them (exactly like it is now implemented). You would then be
> > able to spawn as needed any number of rte_afu_device within your bus and
> > the scan would be complete.
> > 
> > The only trickery would be afterward, because you'd need at least one
> > device that would be detached and having the name given in the devargs.
> > So you would have to name your rte_afu_device after the devargs passed
> > as parameter to the plug, essentially having all of them using the same name.
> > 
> > Nothing prevents you from renaming your AFU devices after bus->plug, but
> > the EAL will rely on it for matching and knowing the scan went ok.
> > 
> > -- 
> Thanks, Gaëtan.
> 
> So, just to confirm I understand the suggestion correctly, you are
> proposing that we create and use the FPGA bus as normal, except that the
> scan will not report any devices. Then when an FPGA PCI device is
> discovered, we use the hotplug infrastructure to trigger a real scan of the
> FPGA for AFU devices, and do the whole driver mapping etc. process then.
> Is that the basic idea you suggest?

Yes.

-- 
Gaëtan Rivet
6WIND

^ permalink raw reply	[flat|nested] 19+ messages in thread

end of thread, other threads:[~2018-03-21 16:21 UTC | newest]

Thread overview: 19+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-03-21  7:51 [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Rosen Xu
2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 1/5] Add Intel FPGA BUS Command Parse Code Rosen Xu
2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 2/5] Add Intel FPGA BUS Probe Code Rosen Xu
2018-03-21  9:07   ` Shreyansh Jain
2018-03-21  9:10     ` Shreyansh Jain
2018-03-21 10:05   ` Gaëtan Rivet
2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 3/5] Add Intel FPGA BUS Lib Code Rosen Xu
2018-03-21  9:28   ` Shreyansh Jain
2018-03-21 10:20   ` Gaëtan Rivet
2018-03-21 13:35     ` Bruce Richardson
2018-03-21 14:02       ` Shreyansh Jain
2018-03-21 14:06       ` Xu, Rosen
2018-03-21 14:14       ` Gaëtan Rivet
2018-03-21 14:31         ` Gaëtan Rivet
2018-03-21 15:41           ` Bruce Richardson
2018-03-21 16:21             ` Gaëtan Rivet
2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 4/5] Add Intel FPGA BUS Rawdev Code Rosen Xu
2018-03-21  7:51 ` [dpdk-dev] [PATCH V2 5/5] Add Intel OPAE Share Code Rosen Xu
2018-03-21 10:00 ` [dpdk-dev] [PATCH V2 0/5] Introduce Intel FPGA BUS Gaëtan Rivet

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).