DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH 0/4] NFP PF support based on new CPP interface
@ 2018-03-23 17:35 Alejandro Lucero
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support Alejandro Lucero
                   ` (3 more replies)
  0 siblings, 4 replies; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-23 17:35 UTC (permalink / raw)
  To: dev

NFP PMD PF support requires to access the NFP chip for initialization.
Current NFP PMD PF support was added based on the NSPU interface. This
implies to do initialization through the NSP, a embedded ARM processor
which does initialization tasks on demand. The main problem with this
approach is it requires to add support for new NSP commands each time
a new functionality is required, which does not scale well and it is
not really flexible.

Using the new CPP user space interface, the PMD can do whatever could
be done by the NSP, this is current commands and any new functionality
required. This CPP interface allows to access any single chip component
facilitating initialization, firmware uploading, firmware debugging or
extended stats.

The changes just change the PMD PF initialization and do not touch the
datapath at all. No performance changes nor PMD functionalities are affected.

The initial impact using the new CPP interface is the way firmware upload
is handled, which helps the PMD detecting the card type and the firmware file
to upload. Future commits will include extended stats and some sort of debug
channel.

The specific CPP code is contained in the first patch, which has not been
splitted up because is completely internal to the NFP functionality. The
second patch makes the PMD changes required for using the new interface.

Alejandro Lucero (4):
  net/nfp: add NFP CPP support
  net/nfp: update PMD for using new CPP interface
  doc: update NFP guide
  net/nfp: remove files

 doc/guides/nics/nfp.rst                           |  31 +-
 drivers/net/nfp/Makefile                          |  17 +-
 drivers/net/nfp/nfp_net.c                         | 342 +++++---
 drivers/net/nfp/nfp_net_eth.h                     |  82 --
 drivers/net/nfp/nfp_net_pmd.h                     |  16 +-
 drivers/net/nfp/nfp_nfpu.c                        | 108 ---
 drivers/net/nfp/nfp_nfpu.h                        |  55 --
 drivers/net/nfp/nfp_nspu.c                        | 642 ---------------
 drivers/net/nfp/nfp_nspu.h                        |  83 --
 drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h    | 748 +++++++++++++++++
 drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h |  62 ++
 drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h    | 620 ++++++++++++++
 drivers/net/nfp/nfpcore/nfp6000/nfp6000.h         |  68 ++
 drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h         |  54 ++
 drivers/net/nfp/nfpcore/nfp_cpp.h                 | 803 ++++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c        | 962 ++++++++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_cppcore.c             | 901 ++++++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_crc.c                 |  75 ++
 drivers/net/nfp/nfpcore/nfp_crc.h                 |  45 +
 drivers/net/nfp/nfpcore/nfp_hwinfo.c              | 226 +++++
 drivers/net/nfp/nfpcore/nfp_hwinfo.h              | 111 +++
 drivers/net/nfp/nfpcore/nfp_mip.c                 | 180 ++++
 drivers/net/nfp/nfpcore/nfp_mip.h                 |  47 ++
 drivers/net/nfp/nfpcore/nfp_mutex.c               | 450 ++++++++++
 drivers/net/nfp/nfpcore/nfp_nffw.c                | 261 ++++++
 drivers/net/nfp/nfpcore/nfp_nffw.h                | 112 +++
 drivers/net/nfp/nfpcore/nfp_nsp.c                 | 453 ++++++++++
 drivers/net/nfp/nfpcore/nfp_nsp.h                 | 330 ++++++++
 drivers/net/nfp/nfpcore/nfp_nsp_cmds.c            | 135 +++
 drivers/net/nfp/nfpcore/nfp_nsp_eth.c             | 691 ++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_resource.c            | 291 +++++++
 drivers/net/nfp/nfpcore/nfp_resource.h            |  78 ++
 drivers/net/nfp/nfpcore/nfp_rtsym.c               | 353 ++++++++
 drivers/net/nfp/nfpcore/nfp_rtsym.h               |  87 ++
 drivers/net/nfp/nfpcore/nfp_target.h              | 605 ++++++++++++++
 35 files changed, 9036 insertions(+), 1088 deletions(-)
 delete mode 100644 drivers/net/nfp/nfp_net_eth.h
 delete mode 100644 drivers/net/nfp/nfp_nfpu.c
 delete mode 100644 drivers/net/nfp/nfp_nfpu.h
 delete mode 100644 drivers/net/nfp/nfp_nspu.c
 delete mode 100644 drivers/net/nfp/nfp_nspu.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp6000/nfp6000.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_cpp.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_cppcore.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_crc.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_crc.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_hwinfo.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_hwinfo.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_mip.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_mip.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_mutex.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nffw.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nffw.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp_cmds.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp_eth.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_resource.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_resource.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_rtsym.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_rtsym.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_target.h

-- 
1.9.1

^ permalink raw reply	[flat|nested] 12+ messages in thread

* [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support
  2018-03-23 17:35 [dpdk-dev] [PATCH 0/4] NFP PF support based on new CPP interface Alejandro Lucero
@ 2018-03-23 17:35 ` Alejandro Lucero
  2018-03-30 10:37   ` Ferruh Yigit
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 2/4] net/nfp: update PMD for using new CPP interface Alejandro Lucero
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-23 17:35 UTC (permalink / raw)
  To: dev

CPP refers to the internal NFP Command Push Pull bus. This patch allows
to create CPP commands from user space allowing to access any single
part of the chip.

This CPP interface is the base for having other functionalities like
mutexes when accessing specific chip components, chip resources management,
firmware upload or using the NSP, an embedded arm processor which can
perform tasks on demand.

NSP was the previous only way for doing things in the chip by the PMD,
where a NSPU interface was used for commands like firmware upload or
port link configuration. CPP interface supersedes NSPU, but it is still
possible to use NSP through CPP.

CPP interface adds a great flexibility for doing things like extended
stats, firmware debugging or selecting properly the firmware file to
upload.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h    | 748 +++++++++++++++++
 drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h |  62 ++
 drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h    | 620 ++++++++++++++
 drivers/net/nfp/nfpcore/nfp6000/nfp6000.h         |  68 ++
 drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h         |  54 ++
 drivers/net/nfp/nfpcore/nfp_cpp.h                 | 803 ++++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c        | 962 ++++++++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_cppcore.c             | 901 ++++++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_crc.c                 |  75 ++
 drivers/net/nfp/nfpcore/nfp_crc.h                 |  45 +
 drivers/net/nfp/nfpcore/nfp_hwinfo.c              | 226 +++++
 drivers/net/nfp/nfpcore/nfp_hwinfo.h              | 111 +++
 drivers/net/nfp/nfpcore/nfp_mip.c                 | 180 ++++
 drivers/net/nfp/nfpcore/nfp_mip.h                 |  47 ++
 drivers/net/nfp/nfpcore/nfp_mutex.c               | 450 ++++++++++
 drivers/net/nfp/nfpcore/nfp_nffw.c                | 261 ++++++
 drivers/net/nfp/nfpcore/nfp_nffw.h                | 112 +++
 drivers/net/nfp/nfpcore/nfp_nsp.c                 | 453 ++++++++++
 drivers/net/nfp/nfpcore/nfp_nsp.h                 | 330 ++++++++
 drivers/net/nfp/nfpcore/nfp_nsp_cmds.c            | 135 +++
 drivers/net/nfp/nfpcore/nfp_nsp_eth.c             | 691 ++++++++++++++++
 drivers/net/nfp/nfpcore/nfp_resource.c            | 291 +++++++
 drivers/net/nfp/nfpcore/nfp_resource.h            |  78 ++
 drivers/net/nfp/nfpcore/nfp_rtsym.c               | 353 ++++++++
 drivers/net/nfp/nfpcore/nfp_rtsym.h               |  87 ++
 drivers/net/nfp/nfpcore/nfp_target.h              | 605 ++++++++++++++
 26 files changed, 8748 insertions(+)
 create mode 100644 drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp6000/nfp6000.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_cpp.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_cppcore.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_crc.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_crc.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_hwinfo.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_hwinfo.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_mip.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_mip.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_mutex.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nffw.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nffw.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp_cmds.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_nsp_eth.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_resource.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_resource.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_rtsym.c
 create mode 100644 drivers/net/nfp/nfpcore/nfp_rtsym.h
 create mode 100644 drivers/net/nfp/nfpcore/nfp_target.h

diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
new file mode 100644
index 0000000..fbeec57
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
@@ -0,0 +1,748 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_CPPAT_H__
+#define __NFP_CPPAT_H__
+
+#include "nfp_platform.h"
+#include "nfp_resid.h"
+
+/* This file contains helpers for creating CPP commands
+ *
+ * All magic NFP-6xxx IMB 'mode' numbers here are from:
+ * Databook (1 August 2013)
+ * - System Overview and Connectivity
+ * -- Internal Connectivity
+ * --- Distributed Switch Fabric - Command Push/Pull (DSF-CPP) Bus
+ * ---- CPP addressing
+ * ----- Table 3.6. CPP Address Translation Mode Commands
+ */
+
+#define _NIC_NFP6000_MU_LOCALITY_DIRECT 2
+
+static inline int
+_nfp6000_decode_basic(uint64_t addr, int *dest_island, int cpp_tgt, int mode,
+		      int addr40, int isld1, int isld0);
+
+static uint64_t
+_nic_mask64(int msb, int lsb, int at0)
+{
+	uint64_t v;
+	int w = msb - lsb + 1;
+
+	if (w == 64)
+		return ~(uint64_t)0;
+
+	if ((lsb + w) > 64)
+		return 0;
+
+	v = (UINT64_C(1) << w) - 1;
+
+	if (at0)
+		return v;
+
+	return v << lsb;
+}
+
+/* For VQDR, we may not modify the Channel bits, which might overlap
+ * with the Index bit. When it does, we need to ensure that isld0 == isld1.
+ */
+static inline int
+_nfp6000_encode_basic(uint64_t *addr, int dest_island, int cpp_tgt, int mode,
+		      int addr40, int isld1, int isld0)
+{
+	uint64_t _u64;
+	int iid_lsb, idx_lsb;
+	int i, v = 0;
+	int isld[2];
+
+	isld[0] = isld0;
+	isld[1] = isld1;
+
+	switch (cpp_tgt) {
+	case NFP6000_CPPTGT_MU:
+		/* This function doesn't handle MU */
+		return NFP_ERRNO(EINVAL);
+	case NFP6000_CPPTGT_CTXPB:
+		/* This function doesn't handle CTXPB */
+		return NFP_ERRNO(EINVAL);
+	default:
+		break;
+	}
+
+	switch (mode) {
+	case 0:
+		if (cpp_tgt == NFP6000_CPPTGT_VQDR && !addr40) {
+			/*
+			 * In this specific mode we'd rather not modify the
+			 * address but we can verify if the existing contents
+			 * will point to a valid island.
+			 */
+			i = _nfp6000_decode_basic(*addr, &v, cpp_tgt, mode,
+						  addr40, isld1,
+						  isld0);
+			if (i != 0)
+				/* Full Island ID and channel bits overlap */
+				return i;
+
+			/*
+			 * If dest_island is invalid, the current address won't
+			 * go where expected.
+			 */
+			if (dest_island != -1 && dest_island != v)
+				return NFP_ERRNO(EINVAL);
+
+			/* If dest_island was -1, we don't care */
+			return 0;
+		}
+
+		iid_lsb = (addr40) ? 34 : 26;
+
+		/* <39:34> or <31:26> */
+		_u64 = _nic_mask64((iid_lsb + 5), iid_lsb, 0);
+		*addr &= ~_u64;
+		*addr |= (((uint64_t)dest_island) << iid_lsb) & _u64;
+		return 0;
+	case 1:
+		if (cpp_tgt == NFP6000_CPPTGT_VQDR && !addr40) {
+			i = _nfp6000_decode_basic(*addr, &v, cpp_tgt, mode,
+						  addr40, isld1, isld0);
+			if (i != 0)
+				/* Full Island ID and channel bits overlap */
+				return i;
+
+			/*
+			 * If dest_island is invalid, the current address won't
+			 * go where expected.
+			 */
+			if (dest_island != -1 && dest_island != v)
+				return NFP_ERRNO(EINVAL);
+
+			/* If dest_island was -1, we don't care */
+			return 0;
+		}
+
+		idx_lsb = (addr40) ? 39 : 31;
+		if (dest_island == isld0) {
+			/* Only need to clear the Index bit */
+			*addr &= ~_nic_mask64(idx_lsb, idx_lsb, 0);
+			return 0;
+		}
+
+		if (dest_island == isld1) {
+			/* Only need to set the Index bit */
+			*addr |= (UINT64_C(1) << idx_lsb);
+			return 0;
+		}
+
+		return NFP_ERRNO(ENODEV);
+	case 2:
+		if (cpp_tgt == NFP6000_CPPTGT_VQDR && !addr40) {
+			/* iid<0> = addr<30> = channel<0> */
+			/* channel<1> = addr<31> = Index */
+
+			/*
+			 * Special case where we allow channel bits to be set
+			 * before hand and with them select an island.
+			 * So we need to confirm that it's at least plausible.
+			 */
+			i = _nfp6000_decode_basic(*addr, &v, cpp_tgt, mode,
+						  addr40, isld1, isld0);
+			if (i != 0)
+				/* Full Island ID and channel bits overlap */
+				return i;
+
+			/*
+			 * If dest_island is invalid, the current address won't
+			 * go where expected.
+			 */
+			if (dest_island != -1 && dest_island != v)
+				return NFP_ERRNO(EINVAL);
+
+			/* If dest_island was -1, we don't care */
+			return 0;
+		}
+
+		/*
+		 * Make sure we compare against isldN values by clearing the
+		 * LSB. This is what the silicon does.
+		 **/
+		isld[0] &= ~1;
+		isld[1] &= ~1;
+
+		idx_lsb = (addr40) ? 39 : 31;
+		iid_lsb = idx_lsb - 1;
+
+		/*
+		 * Try each option, take first one that fits. Not sure if we
+		 * would want to do some smarter searching and prefer 0 or non-0
+		 * island IDs.
+		 */
+
+		for (i = 0; i < 2; i++) {
+			for (v = 0; v < 2; v++) {
+				if (dest_island != (isld[i] | v))
+					continue;
+				*addr &= ~_nic_mask64(idx_lsb, iid_lsb, 0);
+				*addr |= (((uint64_t)i) << idx_lsb);
+				*addr |= (((uint64_t)v) << iid_lsb);
+				return 0;
+			}
+		}
+
+		return NFP_ERRNO(ENODEV);
+	case 3:
+		if (cpp_tgt == NFP6000_CPPTGT_VQDR && !addr40) {
+			/*
+			 * iid<0> = addr<29> = data
+			 * iid<1> = addr<30> = channel<0>
+			 * channel<1> = addr<31> = Index
+			 */
+			i = _nfp6000_decode_basic(*addr, &v, cpp_tgt, mode,
+						  addr40, isld1, isld0);
+			if (i != 0)
+				/* Full Island ID and channel bits overlap */
+				return i;
+
+			if (dest_island != -1 && dest_island != v)
+				return NFP_ERRNO(EINVAL);
+
+			/* If dest_island was -1, we don't care */
+			return 0;
+		}
+
+		isld[0] &= ~3;
+		isld[1] &= ~3;
+
+		idx_lsb = (addr40) ? 39 : 31;
+		iid_lsb = idx_lsb - 2;
+
+		for (i = 0; i < 2; i++) {
+			for (v = 0; v < 4; v++) {
+				if (dest_island != (isld[i] | v))
+					continue;
+				*addr &= ~_nic_mask64(idx_lsb, iid_lsb, 0);
+				*addr |= (((uint64_t)i) << idx_lsb);
+				*addr |= (((uint64_t)v) << iid_lsb);
+				return 0;
+			}
+		}
+		return NFP_ERRNO(ENODEV);
+	default:
+		break;
+	}
+
+	return NFP_ERRNO(EINVAL);
+}
+
+static inline int
+_nfp6000_decode_basic(uint64_t addr, int *dest_island, int cpp_tgt, int mode,
+		      int addr40, int isld1, int isld0)
+{
+	int iid_lsb, idx_lsb;
+
+	switch (cpp_tgt) {
+	case NFP6000_CPPTGT_MU:
+		/* This function doesn't handle MU */
+		return NFP_ERRNO(EINVAL);
+	case NFP6000_CPPTGT_CTXPB:
+		/* This function doesn't handle CTXPB */
+		return NFP_ERRNO(EINVAL);
+	default:
+		break;
+	}
+
+	switch (mode) {
+	case 0:
+		/*
+		 * For VQDR, in this mode for 32-bit addressing it would be
+		 * islands 0, 16, 32 and 48 depending on channel and upper
+		 * address bits. Since those are not all valid islands, most
+		 * decode cases would result in bad island IDs, but we do them
+		 * anyway since this is decoding an address that is already
+		 * assumed to be used as-is to get to sram.
+		 */
+		iid_lsb = (addr40) ? 34 : 26;
+		*dest_island = (int)(addr >> iid_lsb) & 0x3F;
+		return 0;
+	case 1:
+		/*
+		 * For VQDR 32-bit, this would decode as:
+		 *	Channel 0: island#0
+		 *	Channel 1: island#0
+		 *	Channel 2: island#1
+		 *	Channel 3: island#1
+		 *
+		 * That would be valid as long as both islands have VQDR.
+		 * Let's allow this.
+		 */
+
+		idx_lsb = (addr40) ? 39 : 31;
+		if (addr & _nic_mask64(idx_lsb, idx_lsb, 0))
+			*dest_island = isld1;
+		else
+			*dest_island = isld0;
+
+		return 0;
+	case 2:
+		/*
+		 * For VQDR 32-bit:
+		 *	Channel 0: (island#0 | 0)
+		 *	Channel 1: (island#0 | 1)
+		 *	Channel 2: (island#1 | 0)
+		 *	Channel 3: (island#1 | 1)
+		 *
+		 * Make sure we compare against isldN values by clearing the
+		 * LSB. This is what the silicon does.
+		 */
+		isld0 &= ~1;
+		isld1 &= ~1;
+
+		idx_lsb = (addr40) ? 39 : 31;
+		iid_lsb = idx_lsb - 1;
+
+		if (addr & _nic_mask64(idx_lsb, idx_lsb, 0))
+			*dest_island = isld1 | (int)((addr >> iid_lsb) & 1);
+		else
+			*dest_island = isld0 | (int)((addr >> iid_lsb) & 1);
+
+		return 0;
+	case 3:
+		/*
+		 * In this mode the data address starts to affect the island ID
+		 * so rather not allow it. In some really specific case one
+		 * could use this to send the upper half of the VQDR channel to
+		 * another MU, but this is getting very specific. However, as
+		 * above for mode 0, this is the decoder and the caller should
+		 * validate the resulting IID. This blindly does what the
+		 * silicon would do.
+		 */
+
+		isld0 &= ~3;
+		isld1 &= ~3;
+
+		idx_lsb = (addr40) ? 39 : 31;
+		iid_lsb = idx_lsb - 2;
+
+		if (addr & _nic_mask64(idx_lsb, idx_lsb, 0))
+			*dest_island = isld1 | (int)((addr >> iid_lsb) & 3);
+		else
+			*dest_island = isld0 | (int)((addr >> iid_lsb) & 3);
+
+		return 0;
+	default:
+		break;
+	}
+
+	return NFP_ERRNO(EINVAL);
+}
+
+static inline int
+_nfp6000_cppat_mu_locality_lsb(int mode, int addr40)
+{
+	switch (mode) {
+	case 0:
+	case 1:
+	case 2:
+	case 3:
+		return (addr40) ? 38 : 30;
+	default:
+		break;
+	}
+	return NFP_ERRNO(EINVAL);
+}
+
+static inline int
+_nfp6000_encode_mu(uint64_t *addr, int dest_island, int mode, int addr40,
+		   int isld1, int isld0)
+{
+	uint64_t _u64;
+	int iid_lsb, idx_lsb, locality_lsb;
+	int i, v;
+	int isld[2];
+	int da;
+
+	isld[0] = isld0;
+	isld[1] = isld1;
+	locality_lsb = _nfp6000_cppat_mu_locality_lsb(mode, addr40);
+
+	if (((*addr >> locality_lsb) & 3) == _NIC_NFP6000_MU_LOCALITY_DIRECT)
+		da = 1;
+	else
+		da = 0;
+
+	switch (mode) {
+	case 0:
+		iid_lsb = (addr40) ? 32 : 24;
+		_u64 = _nic_mask64((iid_lsb + 5), iid_lsb, 0);
+		*addr &= ~_u64;
+		*addr |= (((uint64_t)dest_island) << iid_lsb) & _u64;
+		return 0;
+	case 1:
+		if (da) {
+			iid_lsb = (addr40) ? 32 : 24;
+			_u64 = _nic_mask64((iid_lsb + 5), iid_lsb, 0);
+			*addr &= ~_u64;
+			*addr |= (((uint64_t)dest_island) << iid_lsb) & _u64;
+			return 0;
+		}
+
+		idx_lsb = (addr40) ? 37 : 29;
+		if (dest_island == isld0) {
+			*addr &= ~_nic_mask64(idx_lsb, idx_lsb, 0);
+			return 0;
+		}
+
+		if (dest_island == isld1) {
+			*addr |= (UINT64_C(1) << idx_lsb);
+			return 0;
+		}
+
+		return NFP_ERRNO(ENODEV);
+	case 2:
+		if (da) {
+			iid_lsb = (addr40) ? 32 : 24;
+			_u64 = _nic_mask64((iid_lsb + 5), iid_lsb, 0);
+			*addr &= ~_u64;
+			*addr |= (((uint64_t)dest_island) << iid_lsb) & _u64;
+			return 0;
+		}
+
+		/*
+		 * Make sure we compare against isldN values by clearing the
+		 * LSB. This is what the silicon does.
+		 */
+		isld[0] &= ~1;
+		isld[1] &= ~1;
+
+		idx_lsb = (addr40) ? 37 : 29;
+		iid_lsb = idx_lsb - 1;
+
+		/*
+		 * Try each option, take first one that fits. Not sure if we
+		 * would want to do some smarter searching and prefer 0 or
+		 * non-0 island IDs.
+		 */
+
+		for (i = 0; i < 2; i++) {
+			for (v = 0; v < 2; v++) {
+				if (dest_island != (isld[i] | v))
+					continue;
+				*addr &= ~_nic_mask64(idx_lsb, iid_lsb, 0);
+				*addr |= (((uint64_t)i) << idx_lsb);
+				*addr |= (((uint64_t)v) << iid_lsb);
+				return 0;
+			}
+		}
+		return NFP_ERRNO(ENODEV);
+	case 3:
+		/*
+		 * Only the EMU will use 40 bit addressing. Silently set the
+		 * direct locality bit for everyone else. The SDK toolchain
+		 * uses dest_island <= 0 to test for atypical address encodings
+		 * to support access to local-island CTM with a 32-but address
+		 * (high-locality is effectively ignored and just used for
+		 * routing to island #0).
+		 */
+		if (dest_island > 0 &&
+		    (dest_island < 24 || dest_island > 26)) {
+			*addr |= ((uint64_t)_NIC_NFP6000_MU_LOCALITY_DIRECT)
+				 << locality_lsb;
+			da = 1;
+		}
+
+		if (da) {
+			iid_lsb = (addr40) ? 32 : 24;
+			_u64 = _nic_mask64((iid_lsb + 5), iid_lsb, 0);
+			*addr &= ~_u64;
+			*addr |= (((uint64_t)dest_island) << iid_lsb) & _u64;
+			return 0;
+		}
+
+		isld[0] &= ~3;
+		isld[1] &= ~3;
+
+		idx_lsb = (addr40) ? 37 : 29;
+		iid_lsb = idx_lsb - 2;
+
+		for (i = 0; i < 2; i++) {
+			for (v = 0; v < 4; v++) {
+				if (dest_island != (isld[i] | v))
+					continue;
+				*addr &= ~_nic_mask64(idx_lsb, iid_lsb, 0);
+				*addr |= (((uint64_t)i) << idx_lsb);
+				*addr |= (((uint64_t)v) << iid_lsb);
+				return 0;
+			}
+		}
+
+		return NFP_ERRNO(ENODEV);
+	default:
+		break;
+	}
+
+	return NFP_ERRNO(EINVAL);
+}
+
+static inline int
+_nfp6000_decode_mu(uint64_t addr, int *dest_island, int mode, int addr40,
+		   int isld1, int isld0)
+{
+	int iid_lsb, idx_lsb, locality_lsb;
+	int da;
+
+	locality_lsb = _nfp6000_cppat_mu_locality_lsb(mode, addr40);
+
+	if (((addr >> locality_lsb) & 3) == _NIC_NFP6000_MU_LOCALITY_DIRECT)
+		da = 1;
+	else
+		da = 0;
+
+	switch (mode) {
+	case 0:
+		iid_lsb = (addr40) ? 32 : 24;
+		*dest_island = (int)(addr >> iid_lsb) & 0x3F;
+		return 0;
+	case 1:
+		if (da) {
+			iid_lsb = (addr40) ? 32 : 24;
+			*dest_island = (int)(addr >> iid_lsb) & 0x3F;
+			return 0;
+		}
+
+		idx_lsb = (addr40) ? 37 : 29;
+
+		if (addr & _nic_mask64(idx_lsb, idx_lsb, 0))
+			*dest_island = isld1;
+		else
+			*dest_island = isld0;
+
+		return 0;
+	case 2:
+		if (da) {
+			iid_lsb = (addr40) ? 32 : 24;
+			*dest_island = (int)(addr >> iid_lsb) & 0x3F;
+			return 0;
+		}
+		/*
+		 * Make sure we compare against isldN values by clearing the
+		 * LSB. This is what the silicon does.
+		 */
+		isld0 &= ~1;
+		isld1 &= ~1;
+
+		idx_lsb = (addr40) ? 37 : 29;
+		iid_lsb = idx_lsb - 1;
+
+		if (addr & _nic_mask64(idx_lsb, idx_lsb, 0))
+			*dest_island = isld1 | (int)((addr >> iid_lsb) & 1);
+		else
+			*dest_island = isld0 | (int)((addr >> iid_lsb) & 1);
+
+		return 0;
+	case 3:
+		if (da) {
+			iid_lsb = (addr40) ? 32 : 24;
+			*dest_island = (int)(addr >> iid_lsb) & 0x3F;
+			return 0;
+		}
+
+		isld0 &= ~3;
+		isld1 &= ~3;
+
+		idx_lsb = (addr40) ? 37 : 29;
+		iid_lsb = idx_lsb - 2;
+
+		if (addr & _nic_mask64(idx_lsb, idx_lsb, 0))
+			*dest_island = isld1 | (int)((addr >> iid_lsb) & 3);
+		else
+			*dest_island = isld0 | (int)((addr >> iid_lsb) & 3);
+
+		return 0;
+	default:
+		break;
+	}
+
+	return NFP_ERRNO(EINVAL);
+}
+
+static inline int
+_nfp6000_cppat_addr_encode(uint64_t *addr, int dest_island, int cpp_tgt,
+			   int mode, int addr40, int isld1, int isld0)
+{
+	switch (cpp_tgt) {
+	case NFP6000_CPPTGT_NBI:
+	case NFP6000_CPPTGT_VQDR:
+	case NFP6000_CPPTGT_ILA:
+	case NFP6000_CPPTGT_PCIE:
+	case NFP6000_CPPTGT_ARM:
+	case NFP6000_CPPTGT_CRYPTO:
+	case NFP6000_CPPTGT_CLS:
+		return _nfp6000_encode_basic(addr, dest_island, cpp_tgt, mode,
+					     addr40, isld1, isld0);
+
+	case NFP6000_CPPTGT_MU:
+		return _nfp6000_encode_mu(addr, dest_island, mode, addr40,
+					  isld1, isld0);
+
+	case NFP6000_CPPTGT_CTXPB:
+		if (mode != 1 || addr40 != 0)
+			return NFP_ERRNO(EINVAL);
+
+		*addr &= ~_nic_mask64(29, 24, 0);
+		*addr |= (((uint64_t)dest_island) << 24) &
+			  _nic_mask64(29, 24, 0);
+		return 0;
+	default:
+		break;
+	}
+
+	return NFP_ERRNO(EINVAL);
+}
+
+static inline int
+_nfp6000_cppat_addr_decode(uint64_t addr, int *dest_island, int cpp_tgt,
+			   int mode, int addr40, int isld1, int isld0)
+{
+	switch (cpp_tgt) {
+	case NFP6000_CPPTGT_NBI:
+	case NFP6000_CPPTGT_VQDR:
+	case NFP6000_CPPTGT_ILA:
+	case NFP6000_CPPTGT_PCIE:
+	case NFP6000_CPPTGT_ARM:
+	case NFP6000_CPPTGT_CRYPTO:
+	case NFP6000_CPPTGT_CLS:
+		return _nfp6000_decode_basic(addr, dest_island, cpp_tgt, mode,
+					     addr40, isld1, isld0);
+
+	case NFP6000_CPPTGT_MU:
+		return _nfp6000_decode_mu(addr, dest_island, mode, addr40,
+					  isld1, isld0);
+
+	case NFP6000_CPPTGT_CTXPB:
+		if (mode != 1 || addr40 != 0)
+			return -EINVAL;
+		*dest_island = (int)(addr >> 24) & 0x3F;
+		return 0;
+	default:
+		break;
+	}
+
+	return -EINVAL;
+}
+
+static inline int
+_nfp6000_cppat_addr_iid_clear(uint64_t *addr, int cpp_tgt, int mode, int addr40)
+{
+	int iid_lsb, locality_lsb, da;
+
+	switch (cpp_tgt) {
+	case NFP6000_CPPTGT_NBI:
+	case NFP6000_CPPTGT_VQDR:
+	case NFP6000_CPPTGT_ILA:
+	case NFP6000_CPPTGT_PCIE:
+	case NFP6000_CPPTGT_ARM:
+	case NFP6000_CPPTGT_CRYPTO:
+	case NFP6000_CPPTGT_CLS:
+		switch (mode) {
+		case 0:
+			iid_lsb = (addr40) ? 34 : 26;
+			*addr &= ~(UINT64_C(0x3F) << iid_lsb);
+			return 0;
+		case 1:
+			iid_lsb = (addr40) ? 39 : 31;
+			*addr &= ~_nic_mask64(iid_lsb, iid_lsb, 0);
+			return 0;
+		case 2:
+			iid_lsb = (addr40) ? 38 : 30;
+			*addr &= ~_nic_mask64(iid_lsb + 1, iid_lsb, 0);
+			return 0;
+		case 3:
+			iid_lsb = (addr40) ? 37 : 29;
+			*addr &= ~_nic_mask64(iid_lsb + 2, iid_lsb, 0);
+			return 0;
+		default:
+			break;
+		}
+	case NFP6000_CPPTGT_MU:
+		locality_lsb = _nfp6000_cppat_mu_locality_lsb(mode, addr40);
+		da = (((*addr >> locality_lsb) & 3) ==
+		      _NIC_NFP6000_MU_LOCALITY_DIRECT);
+		switch (mode) {
+		case 0:
+			iid_lsb = (addr40) ? 32 : 24;
+			*addr &= ~(UINT64_C(0x3F) << iid_lsb);
+			return 0;
+		case 1:
+			if (da) {
+				iid_lsb = (addr40) ? 32 : 24;
+				*addr &= ~(UINT64_C(0x3F) << iid_lsb);
+				return 0;
+			}
+			iid_lsb = (addr40) ? 37 : 29;
+			*addr &= ~_nic_mask64(iid_lsb, iid_lsb, 0);
+			return 0;
+		case 2:
+			if (da) {
+				iid_lsb = (addr40) ? 32 : 24;
+				*addr &= ~(UINT64_C(0x3F) << iid_lsb);
+				return 0;
+			}
+
+			iid_lsb = (addr40) ? 36 : 28;
+			*addr &= ~_nic_mask64(iid_lsb + 1, iid_lsb, 0);
+			return 0;
+		case 3:
+			if (da) {
+				iid_lsb = (addr40) ? 32 : 24;
+				*addr &= ~(UINT64_C(0x3F) << iid_lsb);
+				return 0;
+			}
+
+			iid_lsb = (addr40) ? 35 : 27;
+			*addr &= ~_nic_mask64(iid_lsb + 2, iid_lsb, 0);
+			return 0;
+		default:
+			break;
+		}
+	case NFP6000_CPPTGT_CTXPB:
+		if (mode != 1 || addr40 != 0)
+			return 0;
+		*addr &= ~(UINT64_C(0x3F) << 24);
+		return 0;
+	default:
+		break;
+	}
+
+	return NFP_ERRNO(EINVAL);
+}
+
+#endif /* __NFP_CPPAT_H__ */
diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h b/drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h
new file mode 100644
index 0000000..bee2e94
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_platform.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_PLATFORM_H__
+#define __NFP_PLATFORM_H__
+
+#include <fcntl.h>
+#include <unistd.h>
+#include <stdint.h>
+#include <string.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <inttypes.h>
+#include <sys/cdefs.h>
+#include <sys/stat.h>
+#include <limits.h>
+#include <errno.h>
+
+#ifndef BIT_ULL
+#define BIT(x) (1 << (x))
+#define BIT_ULL(x) (1ULL << (x))
+#endif
+
+#ifndef ARRAY_SIZE
+#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
+#endif
+
+#define NFP_ERRNO(err) (errno = (err), -1)
+#define NFP_ERRNO_RET(err, ret) (errno = (err), (ret))
+#define NFP_NOERR(errv) (errno)
+#define NFP_ERRPTR(err) (errno = (err), NULL)
+#define NFP_PTRERR(errv) (errno)
+
+#endif /* __NFP_PLATFORM_H__ */
diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h b/drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h
new file mode 100644
index 0000000..629b27e
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_resid.h
@@ -0,0 +1,620 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ *
+ * 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 copyright holder 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 HOLDER 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 __NFP_RESID_H__
+#define __NFP_RESID_H__
+
+#if (!defined(_NFP_RESID_NO_C_FUNC) && \
+	(defined(__NFP_TOOL_NFCC) || defined(__NFP_TOOL_NFAS)))
+#define _NFP_RESID_NO_C_FUNC
+#endif
+
+#ifndef _NFP_RESID_NO_C_FUNC
+#include "nfp_platform.h"
+#endif
+
+/*
+ * NFP Chip Architectures
+ *
+ * These are semi-arbitrary values to indicate an NFP architecture.
+ * They serve as a software view of a group of chip families, not necessarily a
+ * direct mapping to actual hardware design.
+ */
+#define NFP_CHIP_ARCH_YD	1
+#define NFP_CHIP_ARCH_TH	2
+
+/*
+ * NFP Chip Families.
+ *
+ * These are not enums, because they need to be microcode compatible.
+ * They are also not maskable.
+ *
+ * Note: The NFP-4xxx family is handled as NFP-6xxx in most software
+ * components.
+ *
+ */
+#define NFP_CHIP_FAMILY_NFP6000 0x6000	/* ARCH_TH */
+
+/* NFP Microengine/Flow Processing Core Versions */
+#define NFP_CHIP_ME_VERSION_2_7 0x0207
+#define NFP_CHIP_ME_VERSION_2_8 0x0208
+#define NFP_CHIP_ME_VERSION_2_9 0x0209
+
+/* NFP Chip Base Revisions. Minor stepping can just be added to these */
+#define NFP_CHIP_REVISION_A0 0x00
+#define NFP_CHIP_REVISION_B0 0x10
+#define NFP_CHIP_REVISION_C0 0x20
+#define NFP_CHIP_REVISION_PF 0xff /* Maximum possible revision */
+
+/* CPP Targets for each chip architecture */
+#define NFP6000_CPPTGT_NBI 1
+#define NFP6000_CPPTGT_VQDR 2
+#define NFP6000_CPPTGT_ILA 6
+#define NFP6000_CPPTGT_MU 7
+#define NFP6000_CPPTGT_PCIE 9
+#define NFP6000_CPPTGT_ARM 10
+#define NFP6000_CPPTGT_CRYPTO 12
+#define NFP6000_CPPTGT_CTXPB 14
+#define NFP6000_CPPTGT_CLS 15
+
+/*
+ * Wildcard indicating a CPP read or write action
+ *
+ * The action used will be either read or write depending on whether a read or
+ * write instruction/call is performed on the NFP_CPP_ID.  It is recomended that
+ * the RW action is used even if all actions to be performed on a NFP_CPP_ID are
+ * known to be only reads or writes. Doing so will in many cases save NFP CPP
+ * internal software resources.
+ */
+#define NFP_CPP_ACTION_RW 32
+
+#define NFP_CPP_TARGET_ID_MASK 0x1f
+
+/*
+ *  NFP_CPP_ID - pack target, token, and action into a CPP ID.
+ *
+ * Create a 32-bit CPP identifier representing the access to be made.
+ * These identifiers are used as parameters to other NFP CPP functions. Some
+ * CPP devices may allow wildcard identifiers to be specified.
+ *
+ * @param[in]	target	NFP CPP target id
+ * @param[in]	action	NFP CPP action id
+ * @param[in]	token	NFP CPP token id
+ * @return		NFP CPP ID
+ */
+#define NFP_CPP_ID(target, action, token)                   \
+	((((target) & 0x7f) << 24) | (((token) & 0xff) << 16) | \
+	 (((action) & 0xff) << 8))
+
+#define NFP_CPP_ISLAND_ID(target, action, token, island)    \
+	((((target) & 0x7f) << 24) | (((token) & 0xff) << 16) | \
+	 (((action) & 0xff) << 8) | (((island) & 0xff) << 0))
+
+#ifndef _NFP_RESID_NO_C_FUNC
+
+/**
+ * Return the NFP CPP target of a NFP CPP ID
+ * @param[in]	id	NFP CPP ID
+ * @return	NFP CPP target
+ */
+static inline uint8_t
+NFP_CPP_ID_TARGET_of(uint32_t id)
+{
+	return (id >> 24) & NFP_CPP_TARGET_ID_MASK;
+}
+
+/*
+ * Return the NFP CPP token of a NFP CPP ID
+ * @param[in]	id	NFP CPP ID
+ * @return	NFP CPP token
+ */
+static inline uint8_t
+NFP_CPP_ID_TOKEN_of(uint32_t id)
+{
+	return (id >> 16) & 0xff;
+}
+
+/*
+ * Return the NFP CPP action of a NFP CPP ID
+ * @param[in]	id	NFP CPP ID
+ * @return	NFP CPP action
+ */
+static inline uint8_t
+NFP_CPP_ID_ACTION_of(uint32_t id)
+{
+	return (id >> 8) & 0xff;
+}
+
+/*
+ * Return the NFP CPP action of a NFP CPP ID
+ * @param[in]   id      NFP CPP ID
+ * @return      NFP CPP action
+ */
+static inline uint8_t
+NFP_CPP_ID_ISLAND_of(uint32_t id)
+{
+	return (id) & 0xff;
+}
+
+#endif /* _NFP_RESID_NO_C_FUNC */
+
+/*
+ *  Check if @p chip_family is an ARCH_TH chip.
+ * @param chip_family One of NFP_CHIP_FAMILY_*
+ */
+#define NFP_FAMILY_IS_ARCH_TH(chip_family) \
+	((int)(chip_family) == (int)NFP_CHIP_FAMILY_NFP6000)
+
+/*
+ *  Get the NFP_CHIP_ARCH_* of @p chip_family.
+ * @param chip_family One of NFP_CHIP_FAMILY_*
+ */
+#define NFP_FAMILY_ARCH(x) \
+	(__extension__ ({ \
+		typeof(x) _x = (x); \
+		(NFP_FAMILY_IS_ARCH_TH(_x) ? NFP_CHIP_ARCH_TH : \
+		NFP_FAMILY_IS_ARCH_YD(_x) ? NFP_CHIP_ARCH_YD : -1) \
+	}))
+
+/*
+ *  Check if @p chip_family is an NFP-6xxx chip.
+ * @param chip_family One of NFP_CHIP_FAMILY_*
+ */
+#define NFP_FAMILY_IS_NFP6000(chip_family) \
+	((int)(chip_family) == (int)NFP_CHIP_FAMILY_NFP6000)
+
+/*
+ *  Make microengine ID for NFP-6xxx.
+ * @param island_id   Island ID.
+ * @param menum       ME number, 0 based, within island.
+ *
+ * NOTE: menum should really be unsigned - MSC compiler throws error (not
+ * warning) if a clause is always true i.e. menum >= 0 if cluster_num is type
+ * unsigned int hence the cast of the menum to an int in that particular clause
+ */
+#define NFP6000_MEID(a, b)                       \
+	(__extension__ ({ \
+		typeof(a) _a = (a); \
+		typeof(b) _b = (b); \
+		(((((int)(_a) & 0x3F) == (int)(_a)) &&   \
+		(((int)(_b) >= 0) && ((int)(_b) < 12))) ?    \
+		(int)(((_a) << 4) | ((_b) + 4)) : -1) \
+	}))
+
+/*
+ *  Do a general sanity check on the ME ID.
+ * The check is on the highest possible island ID for the chip family and the
+ * microengine number must  be a master ID.
+ * @param meid      ME ID as created by NFP6000_MEID
+ */
+#define NFP6000_MEID_IS_VALID(meid) \
+	(__extension__ ({ \
+		typeof(meid) _a = (meid); \
+		((((_a) >> 4) < 64) && (((_a) >> 4) >= 0) && \
+		 (((_a) & 0xF) >= 4)) \
+	}))
+
+/*
+ *  Extract island ID from ME ID.
+ * @param meid   ME ID as created by NFP6000_MEID
+ */
+#define NFP6000_MEID_ISLAND_of(meid) (((meid) >> 4) & 0x3F)
+
+/*
+ * Extract microengine number (0 based) from ME ID.
+ * @param meid   ME ID as created by NFP6000_MEID
+ */
+#define NFP6000_MEID_MENUM_of(meid) (((meid) & 0xF) - 4)
+
+/*
+ * Extract microengine group number (0 based) from ME ID.
+ * The group is two code-sharing microengines, so group  0 refers to MEs 0,1,
+ * group 1 refers to MEs 2,3 etc.
+ * @param meid   ME ID as created by NFP6000_MEID
+ */
+#define NFP6000_MEID_MEGRP_of(meid) (NFP6000_MEID_MENUM_of(meid) >> 1)
+
+#ifndef _NFP_RESID_NO_C_FUNC
+
+/*
+ *  Convert a string to an ME ID.
+ *
+ * @param s       A string of format iX.meY
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the ME ID part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return     ME ID on success, -1 on error.
+ */
+int nfp6000_idstr2meid(const char *s, const char **endptr);
+
+/*
+ *  Extract island ID from string.
+ *
+ * Example:
+ * char *c;
+ * int val = nfp6000_idstr2island("i32.me5", &c);
+ * // val == 32, c == "me5"
+ * val = nfp6000_idstr2island("i32", &c);
+ * // val == 32, c == ""
+ *
+ * @param s       A string of format "iX.anything" or "iX"
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the island part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return        If successful, the island ID, -1 on error.
+ */
+int nfp6000_idstr2island(const char *s, const char **endptr);
+
+/*
+ *  Extract microengine number from string.
+ *
+ * Example:
+ * char *c;
+ * int menum = nfp6000_idstr2menum("me5.anything", &c);
+ * // menum == 5, c == "anything"
+ * menum = nfp6000_idstr2menum("me5", &c);
+ * // menum == 5, c == ""
+ *
+ * @param s       A string of format "meX.anything" or "meX"
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the ME number part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return        If successful, the ME number, -1 on error.
+ */
+int nfp6000_idstr2menum(const char *s, const char **endptr);
+
+/*
+ * Extract context number from string.
+ *
+ * Example:
+ * char *c;
+ * int val = nfp6000_idstr2ctxnum("ctx5.anything", &c);
+ * // val == 5, c == "anything"
+ * val = nfp6000_idstr2ctxnum("ctx5", &c);
+ * // val == 5, c == ""
+ *
+ * @param s       A string of format "ctxN.anything" or "ctxN"
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the context number part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return        If successful, the context number, -1 on error.
+ */
+int nfp6000_idstr2ctxnum(const char *s, const char **endptr);
+
+/*
+ * Extract microengine group number from string.
+ *
+ * Example:
+ * char *c;
+ * int val = nfp6000_idstr2megrp("tg2.anything", &c);
+ * // val == 2, c == "anything"
+ * val = nfp6000_idstr2megrp("tg5", &c);
+ * // val == 2, c == ""
+ *
+ * @param s       A string of format "tgX.anything" or "tgX"
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the ME group part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return        If successful, the ME group number, -1 on error.
+ */
+int nfp6000_idstr2megrp(const char *s, const char **endptr);
+
+/*
+ * Create ME ID string of format "iX[.meY]".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param meid   Microengine ID.
+ * @return       Pointer to "s" on success, NULL on error.
+ */
+const char *nfp6000_meid2str(char *s, int meid);
+
+/*
+ * Create ME ID string of format "name[.meY]" or "iX[.meY]".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param meid   Microengine ID.
+ * @return       Pointer to "s" on success, NULL on error.
+ *
+ * Similar to nfp6000_meid2str() except use an alias instead of "iX"
+ * if one exists for the island.
+ */
+const char *nfp6000_meid2altstr(char *s, int meid);
+
+/*
+ * Create string of format "iX".
+ *
+ * @param s         Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                  The resulting string is output here.
+ * @param island_id Island ID.
+ * @return          Pointer to "s" on success, NULL on error.
+ */
+const char *nfp6000_island2str(char *s, int island_id);
+
+/*
+ * Create string of format "name", an island alias.
+ *
+ * @param s         Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                  The resulting string is output here.
+ * @param island_id Island ID.
+ * @return          Pointer to "s" on success, NULL on error.
+ */
+const char *nfp6000_island2altstr(char *s, int island_id);
+
+/*
+ * Create string of format "meY".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param menum  Microengine number within island.
+ * @return       Pointer to "s" on success, NULL on error.
+ */
+const char *nfp6000_menum2str(char *s, int menum);
+
+/*
+ * Create string of format "ctxY".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param ctxnum Context number within microengine.
+ * @return       Pointer to "s" on success, NULL on error.
+ */
+const char *nfp6000_ctxnum2str(char *s, int ctxnum);
+
+/*
+ * Create string of format "tgY".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param megrp  Microengine group number within cluster.
+ * @return       Pointer to "s" on success, NULL on error.
+ */
+const char *nfp6000_megrp2str(char *s, int megrp);
+
+/*
+ * Convert a string to an ME ID.
+ *
+ * @param chip_family Chip family ID
+ * @param s           A string of format iX.meY (or clX.meY)
+ * @param endptr      If non-NULL, *endptr will point to the trailing
+ *                    string after the ME ID part of the string, which
+ *                    is either an empty string or the first character
+ *                    after the separating period.
+ * @return            ME ID on success, -1 on error.
+ */
+int nfp_idstr2meid(int chip_family, const char *s, const char **endptr);
+
+/*
+ * Extract island ID from string.
+ *
+ * Example:
+ * char *c;
+ * int val = nfp_idstr2island(chip, "i32.me5", &c);
+ * // val == 32, c == "me5"
+ * val = nfp_idstr2island(chip, "i32", &c);
+ * // val == 32, c == ""
+ *
+ * @param chip_family Chip family ID
+ * @param s           A string of format "iX.anything" or "iX"
+ * @param endptr      If non-NULL, *endptr will point to the trailing
+ *                    striong after the ME ID part of the string, which
+ *                    is either an empty string or the first character
+ *                    after the separating period.
+ * @return            The island ID on succes, -1 on error.
+ */
+int nfp_idstr2island(int chip_family, const char *s, const char **endptr);
+
+/*
+ * Extract microengine number from string.
+ *
+ * Example:
+ * char *c;
+ * int menum = nfp_idstr2menum("me5.anything", &c);
+ * // menum == 5, c == "anything"
+ * menum = nfp_idstr2menum("me5", &c);
+ * // menum == 5, c == ""
+ *
+ * @param chip_family Chip family ID
+ * @param s           A string of format "meX.anything" or "meX"
+ * @param endptr      If non-NULL, *endptr will point to the trailing
+ *                    striong after the ME ID part of the string, which
+ *                    is either an empty string or the first character
+ *                    after the separating period.
+ * @return            The ME number on succes, -1 on error.
+ */
+int nfp_idstr2menum(int chip_family, const char *s, const char **endptr);
+
+/*
+ * Extract context number from string.
+ *
+ * Example:
+ * char *c;
+ * int val = nfp_idstr2ctxnum("ctx5.anything", &c);
+ * // val == 5, c == "anything"
+ * val = nfp_idstr2ctxnum("ctx5", &c);
+ * // val == 5, c == ""
+ *
+ * @param s       A string of format "ctxN.anything" or "ctxN"
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the context number part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return        If successful, the context number, -1 on error.
+ */
+int nfp_idstr2ctxnum(int chip_family, const char *s, const char **endptr);
+
+/*
+ * Extract microengine group number from string.
+ *
+ * Example:
+ * char *c;
+ * int val = nfp_idstr2megrp("tg2.anything", &c);
+ * // val == 2, c == "anything"
+ * val = nfp_idstr2megrp("tg5", &c);
+ * // val == 5, c == ""
+ *
+ * @param s       A string of format "tgX.anything" or "tgX"
+ * @param endptr  If non-NULL, *endptr will point to the trailing string
+ *                after the ME group part of the string, which is either
+ *                an empty string or the first character after the separating
+ *                period.
+ * @return        If successful, the ME group number, -1 on error.
+ */
+int nfp_idstr2megrp(int chip_family, const char *s, const char **endptr);
+
+/*
+ * Create ME ID string of format "iX[.meY]".
+ *
+ * @param chip_family Chip family ID
+ * @param s           Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                    The resulting string is output here.
+ * @param meid        Microengine ID.
+ * @return            Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_meid2str(int chip_family, char *s, int meid);
+
+/*
+ * Create ME ID string of format "name[.meY]" or "iX[.meY]".
+ *
+ * @param chip_family Chip family ID
+ * @param s           Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                    The resulting string is output here.
+ * @param meid        Microengine ID.
+ * @return            Pointer to "s" on success, NULL on error.
+ *
+ * Similar to nfp_meid2str() except use an alias instead of "iX"
+ * if one exists for the island.
+ */
+const char *nfp_meid2altstr(int chip_family, char *s, int meid);
+
+/*
+ * Create string of format "iX".
+ *
+ * @param chip_family Chip family ID
+ * @param s           Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                    The resulting string is output here.
+ * @param island_id   Island ID.
+ * @return            Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_island2str(int chip_family, char *s, int island_id);
+
+/*
+ * Create string of format "name", an island alias.
+ *
+ * @param chip_family Chip family ID
+ * @param s           Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                    The resulting string is output here.
+ * @param island_id   Island ID.
+ * @return            Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_island2altstr(int chip_family, char *s, int island_id);
+
+/*
+ * Create string of format "meY".
+ *
+ * @param chip_family Chip family ID
+ * @param s           Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *                    The resulting string is output here.
+ * @param menum       Microengine number within island.
+ * @return            Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_menum2str(int chip_family, char *s, int menum);
+
+/*
+ * Create string of format "ctxY".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param ctxnum Context number within microengine.
+ * @return       Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_ctxnum2str(int chip_family, char *s, int ctxnum);
+
+/*
+ * Create string of format "tgY".
+ *
+ * @param s      Pointer to char buffer of size NFP_MEID_STR_SZ.
+ *               The resulting string is output here.
+ * @param megrp  Microengine group number within cluster.
+ * @return       Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_megrp2str(int chip_family, char *s, int megrp);
+
+/*
+ * Convert a two character string to revision number.
+ *
+ * Revision integer is 0x00 for A0, 0x11 for B1 etc.
+ *
+ * @param s     Two character string.
+ * @return      Revision number, -1 on error
+ */
+int nfp_idstr2rev(const char *s);
+
+/*
+ * Create string from revision number.
+ *
+ * String will be upper case.
+ *
+ * @param s     Pointer to char buffer with size of at least 3
+ *              for 2 characters and string terminator.
+ * @param rev   Revision number.
+ * @return      Pointer to "s" on success, NULL on error.
+ */
+const char *nfp_rev2str(char *s, int rev);
+
+/*
+ * Get the NFP CPP address from a string
+ *
+ * String is in the format [island@]target[:[action:[token:]]address]
+ *
+ * @param chip_family Chip family ID
+ * @param tid           Pointer to string to parse
+ * @param cpp_idp       Pointer to CPP ID
+ * @param cpp_addrp     Pointer to CPP address
+ * @return              0 on success, or -1 and errno
+ */
+int nfp_str2cpp(int chip_family,
+		const char *tid,
+		uint32_t *cpp_idp,
+		uint64_t *cpp_addrp);
+
+
+#endif /* _NFP_RESID_NO_C_FUNC */
+
+#endif /* __NFP_RESID_H__ */
diff --git a/drivers/net/nfp/nfpcore/nfp6000/nfp6000.h b/drivers/net/nfp/nfpcore/nfp6000/nfp6000.h
new file mode 100644
index 0000000..dc0a359
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp6000/nfp6000.h
@@ -0,0 +1,68 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ *
+ * 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 copyright holder 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 HOLDER 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 __NFP_NFP6000_H__
+#define __NFP_NFP6000_H__
+
+/* CPP Target IDs */
+#define NFP_CPP_TARGET_INVALID          0
+#define NFP_CPP_TARGET_NBI              1
+#define NFP_CPP_TARGET_QDR              2
+#define NFP_CPP_TARGET_ILA              6
+#define NFP_CPP_TARGET_MU               7
+#define NFP_CPP_TARGET_PCIE             9
+#define NFP_CPP_TARGET_ARM              10
+#define NFP_CPP_TARGET_CRYPTO           12
+#define NFP_CPP_TARGET_ISLAND_XPB       14	/* Shared with CAP */
+#define NFP_CPP_TARGET_ISLAND_CAP       14	/* Shared with XPB */
+#define NFP_CPP_TARGET_CT_XPB           14
+#define NFP_CPP_TARGET_LOCAL_SCRATCH    15
+#define NFP_CPP_TARGET_CLS              NFP_CPP_TARGET_LOCAL_SCRATCH
+
+#define NFP_ISL_EMEM0                   24
+
+#define NFP_MU_ADDR_ACCESS_TYPE_MASK    3ULL
+#define NFP_MU_ADDR_ACCESS_TYPE_DIRECT  2ULL
+
+static inline int
+nfp_cppat_mu_locality_lsb(int mode, int addr40)
+{
+	switch (mode) {
+	case 0 ... 3:
+		return addr40 ? 38 : 30;
+	default:
+		return -EINVAL;
+	}
+}
+
+#endif /* NFP_NFP6000_H */
diff --git a/drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h b/drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h
new file mode 100644
index 0000000..2a89519
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp6000/nfp_xpb.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ *
+ * 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 copyright holder 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 HOLDER 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 __NFP_XPB_H__
+#define __NFP_XPB_H__
+
+/*
+ * For use with NFP6000 Databook "XPB Addressing" section
+ */
+#define NFP_XPB_OVERLAY(island)  (((island) & 0x3f) << 24)
+
+#define NFP_XPB_ISLAND(island)   (NFP_XPB_OVERLAY(island) + 0x60000)
+
+#define NFP_XPB_ISLAND_of(offset) (((offset) >> 24) & 0x3F)
+
+/*
+ * For use with NFP6000 Databook "XPB Island and Device IDs" chapter
+ */
+#define NFP_XPB_DEVICE(island, slave, device) \
+				(NFP_XPB_OVERLAY(island) | \
+				 (((slave) & 3) << 22) | \
+				 (((device) & 0x3f) << 16))
+
+#endif /* NFP_XPB_H */
diff --git a/drivers/net/nfp/nfpcore/nfp_cpp.h b/drivers/net/nfp/nfpcore/nfp_cpp.h
new file mode 100644
index 0000000..e1d8fda
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_cpp.h
@@ -0,0 +1,803 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_CPP_H__
+#define __NFP_CPP_H__
+
+#include "nfp-common/nfp_platform.h"
+#include "nfp-common/nfp_resid.h"
+
+struct nfp_cpp_mutex;
+
+/*
+ * NFP CPP handle
+ */
+struct nfp_cpp {
+	uint32_t model;
+	uint32_t interface;
+	uint8_t *serial;
+	int serial_len;
+	void *priv;
+
+	/* Mutex cache */
+	struct nfp_cpp_mutex *mutex_cache;
+	const struct nfp_cpp_operations *op;
+
+	/*
+	 * NFP-6xxx originating island IMB CPP Address Translation. CPP Target
+	 * ID is index into array. Values are obtained at runtime from local
+	 * island XPB CSRs.
+	 */
+	uint32_t imb_cat_table[16];
+};
+
+/*
+ * NFP CPP device area handle
+ */
+struct nfp_cpp_area {
+	struct nfp_cpp *cpp;
+	char *name;
+	unsigned long long offset;
+	unsigned long size;
+	/* Here follows the 'priv' part of nfp_cpp_area. */
+};
+
+/*
+ * NFP CPP operations structure
+ */
+struct nfp_cpp_operations {
+	/* Size of priv area in struct nfp_cpp_area */
+	size_t area_priv_size;
+
+	/* Instance an NFP CPP */
+	int (*init)(struct nfp_cpp *cpp, const char *devname);
+
+	/*
+	 * Free the bus.
+	 * Called only once, during nfp_cpp_unregister()
+	 */
+	void (*free)(struct nfp_cpp *cpp);
+
+	/*
+	 * Initialize a new NFP CPP area
+	 * NOTE: This is _not_ serialized
+	 */
+	int (*area_init)(struct nfp_cpp_area *area,
+			 uint32_t dest,
+			 unsigned long long address,
+			 unsigned long size);
+	/*
+	 * Clean up a NFP CPP area before it is freed
+	 * NOTE: This is _not_ serialized
+	 */
+	void (*area_cleanup)(struct nfp_cpp_area *area);
+
+	/*
+	 * Acquire resources for a NFP CPP area
+	 * Serialized
+	 */
+	int (*area_acquire)(struct nfp_cpp_area *area);
+	/*
+	 * Release resources for a NFP CPP area
+	 * Serialized
+	 */
+	void (*area_release)(struct nfp_cpp_area *area);
+	/*
+	 * Return a void IO pointer to a NFP CPP area
+	 * NOTE: This is _not_ serialized
+	 */
+
+	void *(*area_iomem)(struct nfp_cpp_area *area);
+
+	void *(*area_mapped)(struct nfp_cpp_area *area);
+	/*
+	 * Perform a read from a NFP CPP area
+	 * Serialized
+	 */
+	int (*area_read)(struct nfp_cpp_area *area,
+			 void *kernel_vaddr,
+			 unsigned long offset,
+			 unsigned int length);
+	/*
+	 * Perform a write to a NFP CPP area
+	 * Serialized
+	 */
+	int (*area_write)(struct nfp_cpp_area *area,
+			  const void *kernel_vaddr,
+			  unsigned long offset,
+			  unsigned int length);
+};
+
+/*
+ * This should be the only external function the transport
+ * module supplies
+ */
+const struct nfp_cpp_operations *nfp_cpp_transport_operations(void);
+
+/*
+ * Set the model id
+ *
+ * @param   cpp     NFP CPP operations structure
+ * @param   model   Model ID
+ */
+void nfp_cpp_model_set(struct nfp_cpp *cpp, uint32_t model);
+
+/*
+ * Set the private instance owned data of a nfp_cpp struct
+ *
+ * @param   cpp     NFP CPP operations structure
+ * @param   interface Interface ID
+ */
+void nfp_cpp_interface_set(struct nfp_cpp *cpp, uint32_t interface);
+
+/*
+ * Set the private instance owned data of a nfp_cpp struct
+ *
+ * @param   cpp     NFP CPP operations structure
+ * @param   serial  NFP serial byte array
+ * @param   len     Length of the serial byte array
+ */
+int nfp_cpp_serial_set(struct nfp_cpp *cpp, const uint8_t *serial,
+		       size_t serial_len);
+
+/*
+ * Set the private data of the nfp_cpp instance
+ *
+ * @param   cpp NFP CPP operations structure
+ * @return      Opaque device pointer
+ */
+void nfp_cpp_priv_set(struct nfp_cpp *cpp, void *priv);
+
+/*
+ * Return the private data of the nfp_cpp instance
+ *
+ * @param   cpp NFP CPP operations structure
+ * @return      Opaque device pointer
+ */
+void *nfp_cpp_priv(struct nfp_cpp *cpp);
+
+/*
+ * Get the privately allocated portion of a NFP CPP area handle
+ *
+ * @param   cpp_area    NFP CPP area handle
+ * @return          Pointer to the private area, or NULL on failure
+ */
+void *nfp_cpp_area_priv(struct nfp_cpp_area *cpp_area);
+
+uint32_t __nfp_cpp_model_autodetect(struct nfp_cpp *cpp);
+
+/*
+ * NFP CPP core interface for CPP clients.
+ */
+
+/*
+ * Open a NFP CPP handle to a CPP device
+ *
+ * @param[in]	id	0-based ID for the CPP interface to use
+ *
+ * @return NFP CPP handle, or NULL on failure (and set errno accordingly).
+ */
+struct nfp_cpp *nfp_cpp_from_device_name(const char *devname);
+
+/*
+ * Free a NFP CPP handle
+ *
+ * @param[in]	cpp	NFP CPP handle
+ */
+void nfp_cpp_free(struct nfp_cpp *cpp);
+
+#define NFP_CPP_MODEL_INVALID   0xffffffff
+
+/*
+ * NFP_CPP_MODEL_CHIP_of - retrieve the chip ID from the model ID
+ *
+ * The chip ID is a 16-bit BCD+A-F encoding for the chip type.
+ *
+ * @param[in]   model   NFP CPP model id
+ * @return      NFP CPP chip id
+ */
+#define NFP_CPP_MODEL_CHIP_of(model)        (((model) >> 16) & 0xffff)
+
+/*
+ * NFP_CPP_MODEL_IS_6000 - Check for the NFP6000 family of devices
+ *
+ * NOTE: The NFP4000 series is considered as a NFP6000 series variant.
+ *
+ * @param[in]	model	NFP CPP model id
+ * @return		true if model is in the NFP6000 family, false otherwise.
+ */
+#define NFP_CPP_MODEL_IS_6000(model)		     \
+		((NFP_CPP_MODEL_CHIP_of(model) >= 0x4000) && \
+		(NFP_CPP_MODEL_CHIP_of(model) < 0x7000))
+
+/*
+ * nfp_cpp_model - Retrieve the Model ID of the NFP
+ *
+ * @param[in]	cpp	NFP CPP handle
+ * @return		NFP CPP Model ID
+ */
+uint32_t nfp_cpp_model(struct nfp_cpp *cpp);
+
+/*
+ * NFP Interface types - logical interface for this CPP connection 4 bits are
+ * reserved for interface type.
+ */
+#define NFP_CPP_INTERFACE_TYPE_INVALID		0x0
+#define NFP_CPP_INTERFACE_TYPE_PCI		0x1
+#define NFP_CPP_INTERFACE_TYPE_ARM		0x2
+#define NFP_CPP_INTERFACE_TYPE_RPC		0x3
+#define NFP_CPP_INTERFACE_TYPE_ILA		0x4
+
+/*
+ * Construct a 16-bit NFP Interface ID
+ *
+ * Interface IDs consists of 4 bits of interface type, 4 bits of unit
+ * identifier, and 8 bits of channel identifier.
+ *
+ * The NFP Interface ID is used in the implementation of NFP CPP API mutexes,
+ * which use the MU Atomic CompareAndWrite operation - hence the limit to 16
+ * bits to be able to use the NFP Interface ID as a lock owner.
+ *
+ * @param[in]	type	NFP Interface Type
+ * @param[in]	unit	Unit identifier for the interface type
+ * @param[in]	channel	Channel identifier for the interface unit
+ * @return		Interface ID
+ */
+#define NFP_CPP_INTERFACE(type, unit, channel)	\
+	((((type) & 0xf) << 12) | \
+	 (((unit) & 0xf) <<  8) | \
+	 (((channel) & 0xff) << 0))
+
+/*
+ * Get the interface type of a NFP Interface ID
+ * @param[in]	interface	NFP Interface ID
+ * @return			NFP Interface ID's type
+ */
+#define NFP_CPP_INTERFACE_TYPE_of(interface)	(((interface) >> 12) & 0xf)
+
+/*
+ * Get the interface unit of a NFP Interface ID
+ * @param[in]	interface	NFP Interface ID
+ * @return			NFP Interface ID's unit
+ */
+#define NFP_CPP_INTERFACE_UNIT_of(interface)	(((interface) >>  8) & 0xf)
+
+/*
+ * Get the interface channel of a NFP Interface ID
+ * @param[in]	interface	NFP Interface ID
+ * @return			NFP Interface ID's channel
+ */
+#define NFP_CPP_INTERFACE_CHANNEL_of(interface)	(((interface) >>  0) & 0xff)
+
+/*
+ * Retrieve the Interface ID of the NFP
+ * @param[in]	cpp	NFP CPP handle
+ * @return		NFP CPP Interface ID
+ */
+uint16_t nfp_cpp_interface(struct nfp_cpp *cpp);
+
+/*
+ * Retrieve the NFP Serial Number (unique per NFP)
+ * @param[in]	cpp	NFP CPP handle
+ * @param[out]	serial	Pointer to reference the serial number array
+ *
+ * @return	size of the NFP6000 serial number, in bytes
+ */
+int nfp_cpp_serial(struct nfp_cpp *cpp, const uint8_t **serial);
+
+/*
+ * Allocate a NFP CPP area handle, as an offset into a CPP ID
+ * @param[in]	cpp	NFP CPP handle
+ * @param[in]	cpp_id	NFP CPP ID
+ * @param[in]	address	Offset into the NFP CPP ID address space
+ * @param[in]	size	Size of the area to reserve
+ *
+ * @return NFP CPP handle, or NULL on failure (and set errno accordingly).
+ */
+struct nfp_cpp_area *nfp_cpp_area_alloc(struct nfp_cpp *cpp, uint32_t cpp_id,
+					unsigned long long address,
+					unsigned long size);
+
+/*
+ * Allocate a NFP CPP area handle, as an offset into a CPP ID, by a named owner
+ * @param[in]	cpp	NFP CPP handle
+ * @param[in]	cpp_id	NFP CPP ID
+ * @param[in]	name	Name of owner of the area
+ * @param[in]	address	Offset into the NFP CPP ID address space
+ * @param[in]	size	Size of the area to reserve
+ *
+ * @return NFP CPP handle, or NULL on failure (and set errno accordingly).
+ */
+struct nfp_cpp_area *nfp_cpp_area_alloc_with_name(struct nfp_cpp *cpp,
+						  uint32_t cpp_id,
+						  const char *name,
+						  unsigned long long address,
+						  unsigned long size);
+
+/*
+ * Free an allocated NFP CPP area handle
+ * @param[in]	area	NFP CPP area handle
+ */
+void nfp_cpp_area_free(struct nfp_cpp_area *area);
+
+/*
+ * Acquire the resources needed to access the NFP CPP area handle
+ *
+ * @param[in]	area	NFP CPP area handle
+ *
+ * @return 0 on success, -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_area_acquire(struct nfp_cpp_area *area);
+
+/*
+ * Release the resources needed to access the NFP CPP area handle
+ *
+ * @param[in]	area	NFP CPP area handle
+ */
+void nfp_cpp_area_release(struct nfp_cpp_area *area);
+
+/*
+ * Allocate, then acquire the resources needed to access the NFP CPP area handle
+ * @param[in]	cpp	NFP CPP handle
+ * @param[in]	cpp_id	NFP CPP ID
+ * @param[in]	address	Offset into the NFP CPP ID address space
+ * @param[in]	size	Size of the area to reserve
+ *
+ * @return NFP CPP handle, or NULL on failure (and set errno accordingly).
+ */
+struct nfp_cpp_area *nfp_cpp_area_alloc_acquire(struct nfp_cpp *cpp,
+						uint32_t cpp_id,
+						unsigned long long address,
+						unsigned long size);
+
+/*
+ * Release the resources, then free the NFP CPP area handle
+ * @param[in]	area	NFP CPP area handle
+ */
+void nfp_cpp_area_release_free(struct nfp_cpp_area *area);
+
+uint8_t *nfp_cpp_map_area(struct nfp_cpp *cpp, int domain, int target,
+			   uint64_t addr, unsigned long size,
+			   struct nfp_cpp_area **area);
+/*
+ * Return an IO pointer to the beginning of the NFP CPP area handle. The area
+ * must be acquired with 'nfp_cpp_area_acquire()' before calling this operation.
+ *
+ * @param[in]	area	NFP CPP area handle
+ *
+ * @return Pointer to IO memory, or NULL on failure (and set errno accordingly).
+ */
+void *nfp_cpp_area_mapped(struct nfp_cpp_area *area);
+
+/*
+ * Read from a NFP CPP area handle into a buffer. The area must be acquired with
+ * 'nfp_cpp_area_acquire()' before calling this operation.
+ *
+ * @param[in]	area	NFP CPP area handle
+ * @param[in]	offset	Offset into the area
+ * @param[in]	buffer	Location of buffer to receive the data
+ * @param[in]	length	Length of the data to read
+ *
+ * @return bytes read on success, -1 on failure (and set errno accordingly).
+ *
+ */
+int nfp_cpp_area_read(struct nfp_cpp_area *area, unsigned long offset,
+		      void *buffer, size_t length);
+
+/*
+ * Write to a NFP CPP area handle from a buffer. The area must be acquired with
+ * 'nfp_cpp_area_acquire()' before calling this operation.
+ *
+ * @param[in]	area	NFP CPP area handle
+ * @param[in]	offset	Offset into the area
+ * @param[in]	buffer	Location of buffer that holds the data
+ * @param[in]	length	Length of the data to read
+ *
+ * @return bytes written on success, -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_area_write(struct nfp_cpp_area *area, unsigned long offset,
+		       const void *buffer, size_t length);
+
+/*
+ * nfp_cpp_area_iomem() - get IOMEM region for CPP area
+ * @area:       CPP area handle
+ *
+ * Returns an iomem pointer for use with readl()/writel() style operations.
+ *
+ * NOTE: Area must have been locked down with an 'acquire'.
+ *
+ * Return: pointer to the area, or NULL
+ */
+void *nfp_cpp_area_iomem(struct nfp_cpp_area *area);
+
+/*
+ * Verify that IO can be performed on an offset in an area
+ *
+ * @param[in]	area	NFP CPP area handle
+ * @param[in]	offset	Offset into the area
+ * @param[in]	size	Size of region to validate
+ *
+ * @return 0 on success, -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_area_check_range(struct nfp_cpp_area *area,
+			     unsigned long long offset, unsigned long size);
+
+/*
+ * Get the NFP CPP handle that is the parent of a NFP CPP area handle
+ *
+ * @param	cpp_area	NFP CPP area handle
+ * @return			NFP CPP handle
+ */
+struct nfp_cpp *nfp_cpp_area_cpp(struct nfp_cpp_area *cpp_area);
+
+/*
+ * Get the name passed during allocation of the NFP CPP area handle
+ *
+ * @param	cpp_area	NFP CPP area handle
+ * @return			Pointer to the area's name
+ */
+const char *nfp_cpp_area_name(struct nfp_cpp_area *cpp_area);
+
+/*
+ * Read a block of data from a NFP CPP ID
+ *
+ * @param[in]	cpp	NFP CPP handle
+ * @param[in]	cpp_id	NFP CPP ID
+ * @param[in]	address	Offset into the NFP CPP ID address space
+ * @param[in]	kernel_vaddr	Buffer to copy read data to
+ * @param[in]	length	Size of the area to reserve
+ *
+ * @return bytes read on success, -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_read(struct nfp_cpp *cpp, uint32_t cpp_id,
+		 unsigned long long address, void *kernel_vaddr, size_t length);
+
+/*
+ * Write a block of data to a NFP CPP ID
+ *
+ * @param[in]	cpp	NFP CPP handle
+ * @param[in]	cpp_id	NFP CPP ID
+ * @param[in]	address	Offset into the NFP CPP ID address space
+ * @param[in]	kernel_vaddr	Buffer to copy write data from
+ * @param[in]	length	Size of the area to reserve
+ *
+ * @return bytes written on success, -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_write(struct nfp_cpp *cpp, uint32_t cpp_id,
+		  unsigned long long address, const void *kernel_vaddr,
+		  size_t length);
+
+
+
+/*
+ * Fill a NFP CPP area handle and offset with a value
+ *
+ * @param[in]	area	NFP CPP area handle
+ * @param[in]	offset	Offset into the NFP CPP ID address space
+ * @param[in]	value	32-bit value to fill area with
+ * @param[in]	length	Size of the area to reserve
+ *
+ * @return bytes written on success, -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_area_fill(struct nfp_cpp_area *area, unsigned long offset,
+		      uint32_t value, size_t length);
+
+/*
+ * Read a single 32-bit value from a NFP CPP area handle
+ *
+ * @param area		NFP CPP area handle
+ * @param offset	offset into NFP CPP area handle
+ * @param value		output value
+ *
+ * The area must be acquired with 'nfp_cpp_area_acquire()' before calling this
+ * operation.
+ *
+ * NOTE: offset must be 32-bit aligned.
+ *
+ * @return 0 on success, or -1 on error (and set errno accordingly).
+ */
+int nfp_cpp_area_readl(struct nfp_cpp_area *area, unsigned long offset,
+		       uint32_t *value);
+
+/*
+ * Write a single 32-bit value to a NFP CPP area handle
+ *
+ * @param area		NFP CPP area handle
+ * @param offset	offset into NFP CPP area handle
+ * @param value		value to write
+ *
+ * The area must be acquired with 'nfp_cpp_area_acquire()' before calling this
+ * operation.
+ *
+ * NOTE: offset must be 32-bit aligned.
+ *
+ * @return 0 on success, or -1 on error (and set errno accordingly).
+ */
+int nfp_cpp_area_writel(struct nfp_cpp_area *area, unsigned long offset,
+			uint32_t value);
+
+/*
+ * Read a single 64-bit value from a NFP CPP area handle
+ *
+ * @param area		NFP CPP area handle
+ * @param offset	offset into NFP CPP area handle
+ * @param value		output value
+ *
+ * The area must be acquired with 'nfp_cpp_area_acquire()' before calling this
+ * operation.
+ *
+ * NOTE: offset must be 64-bit aligned.
+ *
+ * @return 0 on success, or -1 on error (and set errno accordingly).
+ */
+int nfp_cpp_area_readq(struct nfp_cpp_area *area, unsigned long offset,
+		       uint64_t *value);
+
+/*
+ * Write a single 64-bit value to a NFP CPP area handle
+ *
+ * @param area		NFP CPP area handle
+ * @param offset	offset into NFP CPP area handle
+ * @param value		value to write
+ *
+ * The area must be acquired with 'nfp_cpp_area_acquire()' before calling this
+ * operation.
+ *
+ * NOTE: offset must be 64-bit aligned.
+ *
+ * @return 0 on success, or -1 on error (and set errno accordingly).
+ */
+int nfp_cpp_area_writeq(struct nfp_cpp_area *area, unsigned long offset,
+			uint64_t value);
+
+/*
+ * Write a single 32-bit value on the XPB bus
+ *
+ * @param cpp           NFP CPP device handle
+ * @param xpb_tgt	XPB target and address
+ * @param value         value to write
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_xpb_writel(struct nfp_cpp *cpp, uint32_t xpb_tgt, uint32_t value);
+
+/*
+ * Read a single 32-bit value from the XPB bus
+ *
+ * @param cpp           NFP CPP device handle
+ * @param xpb_tgt	XPB target and address
+ * @param value         output value
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_xpb_readl(struct nfp_cpp *cpp, uint32_t xpb_tgt, uint32_t *value);
+
+/*
+ * Modify bits of a 32-bit value from the XPB bus
+ *
+ * @param cpp           NFP CPP device handle
+ * @param xpb_tgt       XPB target and address
+ * @param mask          mask of bits to alter
+ * @param value         value to modify
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_xpb_writelm(struct nfp_cpp *cpp, uint32_t xpb_tgt, uint32_t mask,
+		    uint32_t value);
+
+/*
+ * Modify bits of a 32-bit value from the XPB bus
+ *
+ * @param cpp           NFP CPP device handle
+ * @param xpb_tgt       XPB target and address
+ * @param mask          mask of bits to alter
+ * @param value         value to monitor for
+ * @param timeout_us    maximum number of us to wait (-1 for forever)
+ *
+ * @return >= 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_xpb_waitlm(struct nfp_cpp *cpp, uint32_t xpb_tgt, uint32_t mask,
+		   uint32_t value, int timeout_us);
+
+/*
+ * Read a 32-bit word from a NFP CPP ID
+ *
+ * @param cpp           NFP CPP handle
+ * @param cpp_id        NFP CPP ID
+ * @param address       offset into the NFP CPP ID address space
+ * @param value         output value
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_readl(struct nfp_cpp *cpp, uint32_t cpp_id,
+		  unsigned long long address, uint32_t *value);
+
+/*
+ * Write a 32-bit value to a NFP CPP ID
+ *
+ * @param cpp           NFP CPP handle
+ * @param cpp_id        NFP CPP ID
+ * @param address       offset into the NFP CPP ID address space
+ * @param value         value to write
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ *
+ */
+int nfp_cpp_writel(struct nfp_cpp *cpp, uint32_t cpp_id,
+		   unsigned long long address, uint32_t value);
+
+/*
+ * Read a 64-bit work from a NFP CPP ID
+ *
+ * @param cpp           NFP CPP handle
+ * @param cpp_id        NFP CPP ID
+ * @param address       offset into the NFP CPP ID address space
+ * @param value         output value
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_readq(struct nfp_cpp *cpp, uint32_t cpp_id,
+		  unsigned long long address, uint64_t *value);
+
+/*
+ * Write a 64-bit value to a NFP CPP ID
+ *
+ * @param cpp           NFP CPP handle
+ * @param cpp_id        NFP CPP ID
+ * @param address       offset into the NFP CPP ID address space
+ * @param value         value to write
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_writeq(struct nfp_cpp *cpp, uint32_t cpp_id,
+		   unsigned long long address, uint64_t value);
+
+/*
+ * Initialize a mutex location
+
+ * The CPP target:address must point to a 64-bit aligned location, and will
+ * initialize 64 bits of data at the location.
+ *
+ * This creates the initial mutex state, as locked by this nfp_cpp_interface().
+ *
+ * This function should only be called when setting up the initial lock state
+ * upon boot-up of the system.
+ *
+ * @param cpp		NFP CPP handle
+ * @param target	NFP CPP target ID
+ * @param address	Offset into the address space of the NFP CPP target ID
+ * @param key_id	Unique 32-bit value for this mutex
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_mutex_init(struct nfp_cpp *cpp, int target,
+		       unsigned long long address, uint32_t key_id);
+
+/*
+ * Create a mutex handle from an address controlled by a MU Atomic engine
+ *
+ * The CPP target:address must point to a 64-bit aligned location, and reserve
+ * 64 bits of data at the location for use by the handle.
+ *
+ * Only target/address pairs that point to entities that support the MU Atomic
+ * Engine's CmpAndSwap32 command are supported.
+ *
+ * @param cpp		NFP CPP handle
+ * @param target	NFP CPP target ID
+ * @param address	Offset into the address space of the NFP CPP target ID
+ * @param key_id	32-bit unique key (must match the key at this location)
+ *
+ * @return		A non-NULL struct nfp_cpp_mutex * on success, NULL on
+ *                      failure.
+ */
+struct nfp_cpp_mutex *nfp_cpp_mutex_alloc(struct nfp_cpp *cpp, int target,
+					  unsigned long long address,
+					  uint32_t key_id);
+
+/*
+ * Get the NFP CPP handle the mutex was created with
+ *
+ * @param   mutex   NFP mutex handle
+ * @return          NFP CPP handle
+ */
+struct nfp_cpp *nfp_cpp_mutex_cpp(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Get the mutex key
+ *
+ * @param   mutex   NFP mutex handle
+ * @return          Mutex key
+ */
+uint32_t nfp_cpp_mutex_key(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Get the mutex owner
+ *
+ * @param   mutex   NFP mutex handle
+ * @return          Interface ID of the mutex owner
+ *
+ * NOTE: This is for debug purposes ONLY - the owner may change at any time,
+ * unless it has been locked by this NFP CPP handle.
+ */
+uint16_t nfp_cpp_mutex_owner(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Get the mutex target
+ *
+ * @param   mutex   NFP mutex handle
+ * @return          Mutex CPP target (ie NFP_CPP_TARGET_MU)
+ */
+int nfp_cpp_mutex_target(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Get the mutex address
+ *
+ * @param   mutex   NFP mutex handle
+ * @return          Mutex CPP address
+ */
+uint64_t nfp_cpp_mutex_address(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Free a mutex handle - does not alter the lock state
+ *
+ * @param mutex		NFP CPP Mutex handle
+ */
+void nfp_cpp_mutex_free(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Lock a mutex handle, using the NFP MU Atomic Engine
+ *
+ * @param mutex		NFP CPP Mutex handle
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_mutex_lock(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Unlock a mutex handle, using the NFP MU Atomic Engine
+ *
+ * @param mutex		NFP CPP Mutex handle
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int nfp_cpp_mutex_unlock(struct nfp_cpp_mutex *mutex);
+
+/*
+ * Attempt to lock a mutex handle, using the NFP MU Atomic Engine
+ *
+ * @param mutex		NFP CPP Mutex handle
+ * @return		0 if the lock succeeded, -1 on failure (and errno set
+ *			appropriately).
+ */
+int nfp_cpp_mutex_trylock(struct nfp_cpp_mutex *mutex);
+
+#endif /* !__NFP_CPP_H__ */
diff --git a/drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c b/drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c
new file mode 100644
index 0000000..817e70f
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c
@@ -0,0 +1,962 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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.
+ */
+
+/*
+ * nfp_cpp_pcie_ops.c
+ * Authors: Vinayak Tammineedi <vinayak.tammineedi@netronome.com>
+ *
+ * Multiplexes the NFP BARs between NFP internal resources and
+ * implements the PCIe specific interface for generic CPP bus access.
+ *
+ * The BARs are managed and allocated if they are available.
+ * The generic CPP bus abstraction builds upon this BAR interface.
+ */
+
+#include <assert.h>
+#include <stdio.h>
+#include <execinfo.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <stdint.h>
+#include <stdbool.h>
+#include <fcntl.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+#include <dirent.h>
+#include <libgen.h>
+
+#include <sys/mman.h>
+#include <sys/file.h>
+#include <sys/stat.h>
+
+#include "nfp_cpp.h"
+#include "nfp_target.h"
+#include "nfp6000/nfp6000.h"
+
+#define NFP_PCIE_BAR(_pf)	(0x30000 + ((_pf) & 7) * 0xc0)
+
+#define NFP_PCIE_BAR_PCIE2CPP_ACTION_BASEADDRESS(_x)  (((_x) & 0x1f) << 16)
+#define NFP_PCIE_BAR_PCIE2CPP_BASEADDRESS(_x)         (((_x) & 0xffff) << 0)
+#define NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT(_x)        (((_x) & 0x3) << 27)
+#define NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT_32BIT    0
+#define NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT_64BIT    1
+#define NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT_0BYTE    3
+#define NFP_PCIE_BAR_PCIE2CPP_MAPTYPE(_x)             (((_x) & 0x7) << 29)
+#define NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_OF(_x)          (((_x) >> 29) & 0x7)
+#define NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_FIXED         0
+#define NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_BULK          1
+#define NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_TARGET        2
+#define NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_GENERAL       3
+#define NFP_PCIE_BAR_PCIE2CPP_TARGET_BASEADDRESS(_x)  (((_x) & 0xf) << 23)
+#define NFP_PCIE_BAR_PCIE2CPP_TOKEN_BASEADDRESS(_x)   (((_x) & 0x3) << 21)
+
+/*
+ * Minimal size of the PCIe cfg memory we depend on being mapped,
+ * queue controller and DMA controller don't have to be covered.
+ */
+#define NFP_PCI_MIN_MAP_SIZE				0x080000
+
+#define NFP_PCIE_P2C_FIXED_SIZE(bar)               (1 << (bar)->bitsize)
+#define NFP_PCIE_P2C_BULK_SIZE(bar)                (1 << (bar)->bitsize)
+#define NFP_PCIE_P2C_GENERAL_TARGET_OFFSET(bar, x) ((x) << ((bar)->bitsize - 2))
+#define NFP_PCIE_P2C_GENERAL_TOKEN_OFFSET(bar, x) ((x) << ((bar)->bitsize - 4))
+#define NFP_PCIE_P2C_GENERAL_SIZE(bar)             (1 << ((bar)->bitsize - 4))
+
+#define NFP_PCIE_CFG_BAR_PCIETOCPPEXPBAR(bar, slot) \
+	(NFP_PCIE_BAR(0) + ((bar) * 8 + (slot)) * 4)
+
+#define NFP_PCIE_CPP_BAR_PCIETOCPPEXPBAR(bar, slot) \
+	(((bar) * 8 + (slot)) * 4)
+
+/*
+ * Define to enable a bit more verbose debug output.
+ * Set to 1 to enable a bit more verbose debug output.
+ */
+struct nfp_pcie_user;
+struct nfp6000_area_priv;
+
+/*
+ * struct nfp_bar - describes BAR configuration and usage
+ * @nfp:	backlink to owner
+ * @barcfg:	cached contents of BAR config CSR
+ * @base:	the BAR's base CPP offset
+ * @mask:       mask for the BAR aperture (read only)
+ * @bitsize:	bitsize of BAR aperture (read only)
+ * @index:	index of the BAR
+ * @lock:	lock to specify if bar is in use
+ * @refcnt:	number of current users
+ * @iomem:	mapped IO memory
+ */
+#define NFP_BAR_MAX 7
+struct nfp_bar {
+	struct nfp_pcie_user *nfp;
+	uint32_t barcfg;
+	uint64_t base;		/* CPP address base */
+	uint64_t mask;		/* Bit mask of the bar */
+	uint32_t bitsize;	/* Bit size of the bar */
+	int index;
+	int lock;
+
+	char *csr;
+	char *iomem;
+};
+
+#define BUSDEV_SZ	13
+struct nfp_pcie_user {
+	struct nfp_bar bar[NFP_BAR_MAX];
+
+	int device;
+	int lock;
+	char busdev[BUSDEV_SZ];
+	int barsz;
+	char *cfg;
+};
+
+static uint32_t
+nfp_bar_maptype(struct nfp_bar *bar)
+{
+	return NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_OF(bar->barcfg);
+}
+
+#define TARGET_WIDTH_32    4
+#define TARGET_WIDTH_64    8
+
+static int
+nfp_compute_bar(const struct nfp_bar *bar, uint32_t *bar_config,
+		uint64_t *bar_base, int tgt, int act, int tok,
+		uint64_t offset, size_t size, int width)
+{
+	uint32_t bitsize;
+	uint32_t newcfg;
+	uint64_t mask;
+
+	if (tgt >= 16)
+		return -EINVAL;
+
+	switch (width) {
+	case 8:
+		newcfg =
+		    NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT
+		    (NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT_64BIT);
+		break;
+	case 4:
+		newcfg =
+		    NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT
+		    (NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT_32BIT);
+		break;
+	case 0:
+		newcfg =
+		    NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT
+		    (NFP_PCIE_BAR_PCIE2CPP_LENGTHSELECT_0BYTE);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (act != NFP_CPP_ACTION_RW && act != 0) {
+		/* Fixed CPP mapping with specific action */
+		mask = ~(NFP_PCIE_P2C_FIXED_SIZE(bar) - 1);
+
+		newcfg |=
+		    NFP_PCIE_BAR_PCIE2CPP_MAPTYPE
+		    (NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_FIXED);
+		newcfg |= NFP_PCIE_BAR_PCIE2CPP_TARGET_BASEADDRESS(tgt);
+		newcfg |= NFP_PCIE_BAR_PCIE2CPP_ACTION_BASEADDRESS(act);
+		newcfg |= NFP_PCIE_BAR_PCIE2CPP_TOKEN_BASEADDRESS(tok);
+
+		if ((offset & mask) != ((offset + size - 1) & mask)) {
+			printf("BAR%d: Won't use for Fixed mapping\n",
+				bar->index);
+			printf("\t<%#llx,%#llx>, action=%d\n",
+				(unsigned long long)offset,
+				(unsigned long long)(offset + size), act);
+			printf("\tBAR too small (0x%llx).\n",
+				(unsigned long long)mask);
+			return -EINVAL;
+		}
+		offset &= mask;
+
+#ifdef DEBUG
+		printf("BAR%d: Created Fixed mapping\n", bar->index);
+		printf("\t%d:%d:%d:0x%#llx-0x%#llx>\n", tgt, act, tok,
+			(unsigned long long)offset,
+			(unsigned long long)(offset + mask));
+#endif
+
+		bitsize = 40 - 16;
+	} else {
+		mask = ~(NFP_PCIE_P2C_BULK_SIZE(bar) - 1);
+
+		/* Bulk mapping */
+		newcfg |=
+		    NFP_PCIE_BAR_PCIE2CPP_MAPTYPE
+		    (NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_BULK);
+
+		newcfg |= NFP_PCIE_BAR_PCIE2CPP_TARGET_BASEADDRESS(tgt);
+		newcfg |= NFP_PCIE_BAR_PCIE2CPP_TOKEN_BASEADDRESS(tok);
+
+		if ((offset & mask) != ((offset + size - 1) & mask)) {
+			printf("BAR%d: Won't use for bulk mapping\n",
+				bar->index);
+			printf("\t<%#llx,%#llx>\n", (unsigned long long)offset,
+				(unsigned long long)(offset + size));
+			printf("\ttarget=%d, token=%d\n", tgt, tok);
+			printf("\tBAR too small (%#llx) - (%#llx != %#llx).\n",
+				(unsigned long long)mask,
+				(unsigned long long)(offset & mask),
+				(unsigned long long)(offset + size - 1) & mask);
+
+			return -EINVAL;
+		}
+
+		offset &= mask;
+
+#ifdef DEBUG
+		printf("BAR%d: Created bulk mapping %d:x:%d:%#llx-%#llx\n",
+			bar->index, tgt, tok, (unsigned long long)offset,
+			(unsigned long long)(offset + ~mask));
+#endif
+
+		bitsize = 40 - 21;
+	}
+
+	if (bar->bitsize < bitsize) {
+		printf("BAR%d: Too small for %d:%d:%d\n", bar->index, tgt, tok,
+			act);
+		return -EINVAL;
+	}
+
+	newcfg |= offset >> bitsize;
+
+	if (bar_base)
+		*bar_base = offset;
+
+	if (bar_config)
+		*bar_config = newcfg;
+
+	return 0;
+}
+
+static int
+nfp_bar_write(struct nfp_pcie_user *nfp, struct nfp_bar *bar,
+		  uint32_t newcfg)
+{
+	int base, slot;
+
+	base = bar->index >> 3;
+	slot = bar->index & 7;
+
+	if (!nfp->cfg)
+		return (-ENOMEM);
+
+	bar->csr = nfp->cfg +
+		   NFP_PCIE_CFG_BAR_PCIETOCPPEXPBAR(base, slot);
+
+	*(uint32_t *)(bar->csr) = newcfg;
+
+	bar->barcfg = newcfg;
+#ifdef DEBUG
+	printf("BAR%d: updated to 0x%08x\n", bar->index, newcfg);
+#endif
+
+	return 0;
+}
+
+static int
+nfp_reconfigure_bar(struct nfp_pcie_user *nfp, struct nfp_bar *bar, int tgt,
+		int act, int tok, uint64_t offset, size_t size, int width)
+{
+	uint64_t newbase;
+	uint32_t newcfg;
+	int err;
+
+	err = nfp_compute_bar(bar, &newcfg, &newbase, tgt, act, tok, offset,
+			      size, width);
+	if (err)
+		return err;
+
+	bar->base = newbase;
+
+	return nfp_bar_write(nfp, bar, newcfg);
+}
+
+/*
+ * Map all PCI bars. We assume that the BAR with the PCIe config block is
+ * already mapped.
+ *
+ * BAR0.0: Reserved for General Mapping (for MSI-X access to PCIe SRAM)
+ */
+static int
+nfp_enable_bars(struct nfp_pcie_user *nfp)
+{
+	struct nfp_bar *bar;
+	int x;
+
+	for (x = ARRAY_SIZE(nfp->bar); x > 0; x--) {
+		bar = &nfp->bar[x - 1];
+		bar->barcfg = 0;
+		bar->nfp = nfp;
+		bar->index = x;
+		bar->mask = (1 << (nfp->barsz - 3)) - 1;
+		bar->bitsize = nfp->barsz - 3;
+		bar->base = 0;
+		bar->iomem = NULL;
+		bar->lock = 0;
+		bar->csr = nfp->cfg +
+			   NFP_PCIE_CFG_BAR_PCIETOCPPEXPBAR(bar->index >> 3,
+							   bar->index & 7);
+		bar->iomem =
+		    (char *)mmap(0, 1 << bar->bitsize, PROT_READ | PROT_WRITE,
+				 MAP_SHARED, nfp->device,
+				 bar->index << bar->bitsize);
+
+		if (bar->iomem == MAP_FAILED)
+			return (-ENOMEM);
+	}
+	return 0;
+}
+
+static struct nfp_bar *
+nfp_alloc_bar(struct nfp_pcie_user *nfp)
+{
+	struct nfp_bar *bar;
+	int x;
+
+	for (x = ARRAY_SIZE(nfp->bar); x > 0; x--) {
+		bar = &nfp->bar[x - 1];
+		if (!bar->lock) {
+			bar->lock = 1;
+			return bar;
+		}
+	}
+	return NULL;
+}
+
+static void
+nfp_disable_bars(struct nfp_pcie_user *nfp)
+{
+	struct nfp_bar *bar;
+	int x;
+
+	for (x = ARRAY_SIZE(nfp->bar); x > 0; x--) {
+		bar = &nfp->bar[x - 1];
+		if (bar->iomem) {
+			munmap(bar->iomem, 1 << (nfp->barsz - 3));
+			bar->iomem = NULL;
+			bar->lock = 0;
+		}
+	}
+}
+
+/*
+ * Generic CPP bus access interface.
+ */
+
+struct nfp6000_area_priv {
+	struct nfp_bar *bar;
+	uint32_t bar_offset;
+
+	uint32_t target;
+	uint32_t action;
+	uint32_t token;
+	uint64_t offset;
+	struct {
+		int read;
+		int write;
+		int bar;
+	} width;
+	size_t size;
+	char *iomem;
+};
+
+static int
+nfp6000_area_init(struct nfp_cpp_area *area, uint32_t dest,
+		  unsigned long long address, unsigned long size)
+{
+	struct nfp_pcie_user *nfp = nfp_cpp_priv(nfp_cpp_area_cpp(area));
+	struct nfp6000_area_priv *priv = nfp_cpp_area_priv(area);
+	uint32_t target = NFP_CPP_ID_TARGET_of(dest);
+	uint32_t action = NFP_CPP_ID_ACTION_of(dest);
+	uint32_t token = NFP_CPP_ID_TOKEN_of(dest);
+	int pp, ret = 0;
+
+	pp = nfp6000_target_pushpull(NFP_CPP_ID(target, action, token),
+				     address);
+	if (pp < 0)
+		return pp;
+
+	priv->width.read = PUSH_WIDTH(pp);
+	priv->width.write = PULL_WIDTH(pp);
+
+	if (priv->width.read > 0 &&
+	    priv->width.write > 0 && priv->width.read != priv->width.write)
+		return -EINVAL;
+
+	if (priv->width.read > 0)
+		priv->width.bar = priv->width.read;
+	else
+		priv->width.bar = priv->width.write;
+
+	priv->bar = nfp_alloc_bar(nfp);
+	if (priv->bar == NULL)
+		return -ENOMEM;
+
+	priv->target = target;
+	priv->action = action;
+	priv->token = token;
+	priv->offset = address;
+	priv->size = size;
+
+	ret = nfp_reconfigure_bar(nfp, priv->bar, priv->target, priv->action,
+				  priv->token, priv->offset, priv->size,
+				  priv->width.bar);
+
+	return ret;
+}
+
+static int
+nfp6000_area_acquire(struct nfp_cpp_area *area)
+{
+	struct nfp6000_area_priv *priv = nfp_cpp_area_priv(area);
+
+	/* Calculate offset into BAR. */
+	if (nfp_bar_maptype(priv->bar) ==
+	    NFP_PCIE_BAR_PCIE2CPP_MAPTYPE_GENERAL) {
+		priv->bar_offset = priv->offset &
+			(NFP_PCIE_P2C_GENERAL_SIZE(priv->bar) - 1);
+		priv->bar_offset +=
+			NFP_PCIE_P2C_GENERAL_TARGET_OFFSET(priv->bar,
+							   priv->target);
+		priv->bar_offset +=
+		    NFP_PCIE_P2C_GENERAL_TOKEN_OFFSET(priv->bar, priv->token);
+	} else {
+		priv->bar_offset = priv->offset & priv->bar->mask;
+	}
+
+	/* Must have been too big. Sub-allocate. */
+	if (!priv->bar->iomem)
+		return (-ENOMEM);
+
+	priv->iomem = priv->bar->iomem + priv->bar_offset;
+
+	return 0;
+}
+
+static void *
+nfp6000_area_mapped(struct nfp_cpp_area *area)
+{
+	struct nfp6000_area_priv *area_priv = nfp_cpp_area_priv(area);
+
+	if (!area_priv->iomem)
+		return NULL;
+
+	return area_priv->iomem;
+}
+
+static void
+nfp6000_area_release(struct nfp_cpp_area *area)
+{
+	struct nfp6000_area_priv *priv = nfp_cpp_area_priv(area);
+	priv->bar->lock = 0;
+	priv->bar = NULL;
+	priv->iomem = NULL;
+}
+
+static void *
+nfp6000_area_iomem(struct nfp_cpp_area *area)
+{
+	struct nfp6000_area_priv *priv = nfp_cpp_area_priv(area);
+	return priv->iomem;
+}
+
+static int
+nfp6000_area_read(struct nfp_cpp_area *area, void *kernel_vaddr,
+		  unsigned long offset, unsigned int length)
+{
+	uint64_t *wrptr64 = kernel_vaddr;
+	const volatile uint64_t *rdptr64;
+	struct nfp6000_area_priv *priv;
+	uint32_t *wrptr32 = kernel_vaddr;
+	const volatile uint32_t *rdptr32;
+	int width;
+	unsigned int n;
+	bool is_64;
+
+	priv = nfp_cpp_area_priv(area);
+	rdptr64 = (uint64_t *)(priv->iomem + offset);
+	rdptr32 = (uint32_t *)(priv->iomem + offset);
+
+	if (offset + length > priv->size)
+		return -EFAULT;
+
+	width = priv->width.read;
+
+	if (width <= 0)
+		return -EINVAL;
+
+	/* Unaligned? Translate to an explicit access */
+	if ((priv->offset + offset) & (width - 1)) {
+		printf("aread_read unaligned!!!\n");
+		return -EINVAL;
+	}
+
+	is_64 = width == TARGET_WIDTH_64;
+
+	/* MU reads via a PCIe2CPP BAR supports 32bit (and other) lengths */
+	if (priv->target == (NFP_CPP_TARGET_ID_MASK & NFP_CPP_TARGET_MU) &&
+	    priv->action == NFP_CPP_ACTION_RW) {
+		is_64 = false;
+	}
+
+	if (is_64) {
+		if (offset % sizeof(uint64_t) != 0 ||
+		    length % sizeof(uint64_t) != 0)
+			return -EINVAL;
+	} else {
+		if (offset % sizeof(uint32_t) != 0 ||
+		    length % sizeof(uint32_t) != 0)
+			return -EINVAL;
+	}
+
+	if (!priv->bar)
+		return -EFAULT;
+
+	if (is_64)
+		for (n = 0; n < length; n += sizeof(uint64_t)) {
+			*wrptr64 = *rdptr64;
+			wrptr64++;
+			rdptr64++;
+		}
+	else
+		for (n = 0; n < length; n += sizeof(uint32_t)) {
+			*wrptr32 = *rdptr32;
+			wrptr32++;
+			rdptr32++;
+		}
+
+	return n;
+}
+
+static int
+nfp6000_area_write(struct nfp_cpp_area *area, const void *kernel_vaddr,
+		   unsigned long offset, unsigned int length)
+{
+	const uint64_t *rdptr64 = kernel_vaddr;
+	uint64_t *wrptr64;
+	const uint32_t *rdptr32 = kernel_vaddr;
+	struct nfp6000_area_priv *priv;
+	uint32_t *wrptr32;
+	int width;
+	unsigned int n;
+	bool is_64;
+
+	priv = nfp_cpp_area_priv(area);
+	wrptr64 = (uint64_t *)(priv->iomem + offset);
+	wrptr32 = (uint32_t *)(priv->iomem + offset);
+
+	if (offset + length > priv->size)
+		return -EFAULT;
+
+	width = priv->width.write;
+
+	if (width <= 0)
+		return -EINVAL;
+
+	/* Unaligned? Translate to an explicit access */
+	if ((priv->offset + offset) & (width - 1))
+		return -EINVAL;
+
+	is_64 = width == TARGET_WIDTH_64;
+
+	/* MU writes via a PCIe2CPP BAR supports 32bit (and other) lengths */
+	if (priv->target == (NFP_CPP_TARGET_ID_MASK & NFP_CPP_TARGET_MU) &&
+	    priv->action == NFP_CPP_ACTION_RW)
+		is_64 = false;
+
+	if (is_64) {
+		if (offset % sizeof(uint64_t) != 0 ||
+		    length % sizeof(uint64_t) != 0)
+			return -EINVAL;
+	} else {
+		if (offset % sizeof(uint32_t) != 0 ||
+		    length % sizeof(uint32_t) != 0)
+			return -EINVAL;
+	}
+
+	if (!priv->bar)
+		return -EFAULT;
+
+	if (is_64)
+		for (n = 0; n < length; n += sizeof(uint64_t)) {
+			*wrptr64 = *rdptr64;
+			wrptr64++;
+			rdptr64++;
+		}
+	else
+		for (n = 0; n < length; n += sizeof(uint32_t)) {
+			*wrptr32 = *rdptr32;
+			wrptr32++;
+			rdptr32++;
+		}
+
+	return n;
+}
+
+#define PCI_DEVICES "/sys/bus/pci/devices"
+
+static int
+nfp_acquire_process_lock(struct nfp_pcie_user *desc)
+{
+	int rc;
+	struct flock lock;
+	char lockname[30];
+
+	memset(&lock, 0, sizeof(lock));
+
+	snprintf(lockname, sizeof(lockname), "/var/lock/nfp_%s", desc->busdev);
+	desc->lock = open(lockname, O_RDWR | O_CREAT, 0666);
+	if (desc->lock < 0)
+		return desc->lock;
+
+	lock.l_type = F_WRLCK;
+	lock.l_whence = SEEK_SET;
+	rc = -1;
+	while (rc != 0) {
+		rc = fcntl(desc->lock, F_SETLKW, &lock);
+		if (rc < 0) {
+			if (errno != EAGAIN && errno != EACCES) {
+				close(desc->lock);
+				return rc;
+			}
+		}
+	}
+
+	return 0;
+}
+
+static int
+nfp6000_set_model(struct nfp_pcie_user *desc, struct nfp_cpp *cpp)
+{
+	char tmp_str[80];
+	uint32_t tmp;
+	int fp;
+
+	snprintf(tmp_str, sizeof(tmp_str), "%s/%s/config", PCI_DEVICES,
+		 desc->busdev);
+
+	fp = open(tmp_str, O_RDONLY);
+	if (!fp)
+		return -1;
+
+	lseek(fp, 0x2e, SEEK_SET);
+
+	if (read(fp, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+		printf("Error reading config file for model\n");
+		return -1;
+	}
+
+	tmp = tmp << 16;
+
+	if (close(fp) == -1)
+		return -1;
+
+	nfp_cpp_model_set(cpp, tmp);
+
+	return 0;
+}
+
+static int
+nfp6000_set_interface(struct nfp_pcie_user *desc, struct nfp_cpp *cpp)
+{
+	char tmp_str[80];
+	uint16_t tmp;
+	int fp;
+
+	snprintf(tmp_str, sizeof(tmp_str), "%s/%s/config", PCI_DEVICES,
+		 desc->busdev);
+
+	fp = open(tmp_str, O_RDONLY);
+	if (!fp)
+		return -1;
+
+	lseek(fp, 0x154, SEEK_SET);
+
+	if (read(fp, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+		printf("error reading config file for interface\n");
+		return -1;
+	}
+
+	if (close(fp) == -1)
+		return -1;
+
+	nfp_cpp_interface_set(cpp, tmp);
+
+	return 0;
+}
+
+#define PCI_CFG_SPACE_SIZE	256
+#define PCI_CFG_SPACE_EXP_SIZE	4096
+#define PCI_EXT_CAP_ID(header)		(int)(header & 0x0000ffff)
+#define PCI_EXT_CAP_NEXT(header)	((header >> 20) & 0xffc)
+#define PCI_EXT_CAP_ID_DSN	0x03
+static int
+nfp_pci_find_next_ext_capability(int fp, int cap)
+{
+	uint32_t header;
+	int ttl;
+	int pos = PCI_CFG_SPACE_SIZE;
+
+	/* minimum 8 bytes per capability */
+	ttl = (PCI_CFG_SPACE_EXP_SIZE - PCI_CFG_SPACE_SIZE) / 8;
+
+	lseek(fp, pos, SEEK_SET);
+	if (read(fp, &header, sizeof(header)) != sizeof(header)) {
+		printf("error reading config file for serial\n");
+		return -1;
+	}
+
+	/*
+	 * If we have no capabilities, this is indicated by cap ID,
+	 * cap version and next pointer all being 0.
+	 */
+	if (header == 0)
+		return 0;
+
+	while (ttl-- > 0) {
+		if (PCI_EXT_CAP_ID(header) == cap)
+			return pos;
+
+		pos = PCI_EXT_CAP_NEXT(header);
+		if (pos < PCI_CFG_SPACE_SIZE)
+			break;
+
+		lseek(fp, pos, SEEK_SET);
+		if (read(fp, &header, sizeof(header)) != sizeof(header)) {
+			printf("error reading config file for serial\n");
+			return -1;
+		}
+	}
+
+	return 0;
+}
+
+static int
+nfp6000_set_serial(struct nfp_pcie_user *desc, struct nfp_cpp *cpp)
+{
+	char tmp_str[80];
+	uint16_t tmp;
+	uint8_t serial[6];
+	int serial_len = 6;
+	int fp, pos;
+
+	snprintf(tmp_str, sizeof(tmp_str), "%s/%s/config", PCI_DEVICES,
+		 desc->busdev);
+
+	fp = open(tmp_str, O_RDONLY);
+	if (!fp)
+		return -1;
+
+	pos = nfp_pci_find_next_ext_capability(fp, PCI_EXT_CAP_ID_DSN);
+	if (pos <= 0) {
+		printf("PCI_EXT_CAP_ID_DSN not found. Using default offset\n");
+		lseek(fp, 0x156, SEEK_SET);
+	} else {
+		lseek(fp, pos + 6, SEEK_SET);
+	}
+
+	if (read(fp, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+		printf("error reading config file for serial\n");
+		return -1;
+	}
+
+	serial[4] = (uint8_t)((tmp >> 8) & 0xff);
+	serial[5] = (uint8_t)(tmp & 0xff);
+
+	if (read(fp, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+		printf("error reading config file for serial\n");
+		return -1;
+	}
+
+	serial[2] = (uint8_t)((tmp >> 8) & 0xff);
+	serial[3] = (uint8_t)(tmp & 0xff);
+
+	if (read(fp, &tmp, sizeof(tmp)) != sizeof(tmp)) {
+		printf("error reading config file for serial\n");
+		return -1;
+	}
+
+	serial[0] = (uint8_t)((tmp >> 8) & 0xff);
+	serial[1] = (uint8_t)(tmp & 0xff);
+
+	if (close(fp) == -1)
+		return -1;
+
+	nfp_cpp_serial_set(cpp, serial, serial_len);
+
+	return 0;
+}
+
+static int
+nfp6000_set_barsz(struct nfp_pcie_user *desc)
+{
+	char tmp_str[80];
+	unsigned long start, end, flags, tmp;
+	int i;
+	FILE *fp;
+
+	snprintf(tmp_str, sizeof(tmp_str), "%s/%s/resource", PCI_DEVICES,
+		 desc->busdev);
+
+	fp = fopen(tmp_str, "r");
+	if (!fp)
+		return -1;
+
+	if (fscanf(fp, "0x%lx 0x%lx 0x%lx", &start, &end, &flags) == 0) {
+		printf("error reading resource file for bar size\n");
+		return -1;
+	}
+
+	if (fclose(fp) == -1)
+		return -1;
+
+	tmp = (end - start) + 1;
+	i = 0;
+	while (tmp >>= 1)
+		i++;
+	desc->barsz = i;
+	return 0;
+}
+
+static int
+nfp6000_init(struct nfp_cpp *cpp, const char *devname)
+{
+	char link[120];
+	char tmp_str[80];
+	ssize_t size;
+	int ret = 0;
+	uint32_t model;
+	struct nfp_pcie_user *desc;
+
+	desc = malloc(sizeof(*desc));
+	if (!desc)
+		return -1;
+
+
+	memset(desc->busdev, 0, BUSDEV_SZ);
+	strncpy(desc->busdev, devname, strlen(devname));
+
+	ret = nfp_acquire_process_lock(desc);
+	if (ret)
+		return -1;
+
+	snprintf(tmp_str, sizeof(tmp_str), "%s/%s/driver", PCI_DEVICES,
+		 desc->busdev);
+
+	size = readlink(tmp_str, link, sizeof(link));
+
+	if (size == -1)
+		tmp_str[0] = '\0';
+
+	if (size == sizeof(link))
+		tmp_str[0] = '\0';
+
+	snprintf(tmp_str, sizeof(tmp_str), "%s/%s/resource0", PCI_DEVICES,
+		 desc->busdev);
+
+	desc->device = open(tmp_str, O_RDWR);
+	if (desc->device == -1)
+		return -1;
+
+	if (nfp6000_set_model(desc, cpp) < 0)
+		return -1;
+	if (nfp6000_set_interface(desc, cpp) < 0)
+		return -1;
+	if (nfp6000_set_serial(desc, cpp) < 0)
+		return -1;
+	if (nfp6000_set_barsz(desc) < 0)
+		return -1;
+
+	desc->cfg = (char *)mmap(0, 1 << (desc->barsz - 3),
+				 PROT_READ | PROT_WRITE,
+				 MAP_SHARED, desc->device, 0);
+
+	if (desc->cfg == MAP_FAILED)
+		return -1;
+
+	nfp_enable_bars(desc);
+
+	nfp_cpp_priv_set(cpp, desc);
+
+	model = __nfp_cpp_model_autodetect(cpp);
+	nfp_cpp_model_set(cpp, model);
+
+	return ret;
+}
+
+static void
+nfp6000_free(struct nfp_cpp *cpp)
+{
+	struct nfp_pcie_user *desc = nfp_cpp_priv(cpp);
+	int x;
+
+	/* Unmap may cause if there are any pending transaxctions */
+	nfp_disable_bars(desc);
+	munmap(desc->cfg, 1 << (desc->barsz - 3));
+
+	for (x = ARRAY_SIZE(desc->bar); x > 0; x--) {
+		if (desc->bar[x - 1].iomem)
+			munmap(desc->bar[x - 1].iomem, 1 << (desc->barsz - 3));
+	}
+	close(desc->lock);
+	close(desc->device);
+	free(desc);
+}
+
+static const struct nfp_cpp_operations nfp6000_pcie_ops = {
+	.init = nfp6000_init,
+	.free = nfp6000_free,
+
+	.area_priv_size = sizeof(struct nfp6000_area_priv),
+	.area_init = nfp6000_area_init,
+	.area_acquire = nfp6000_area_acquire,
+	.area_release = nfp6000_area_release,
+	.area_mapped = nfp6000_area_mapped,
+	.area_read = nfp6000_area_read,
+	.area_write = nfp6000_area_write,
+	.area_iomem = nfp6000_area_iomem,
+};
+
+const struct
+nfp_cpp_operations *nfp_cpp_transport_operations(void)
+{
+	return &nfp6000_pcie_ops;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_cppcore.c b/drivers/net/nfp/nfpcore/nfp_cppcore.c
new file mode 100644
index 0000000..f110b8f
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_cppcore.c
@@ -0,0 +1,901 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <assert.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <sys/types.h>
+
+#include <rte_byteorder.h>
+
+#include "nfp_cpp.h"
+#include "nfp_target.h"
+#include "nfp6000/nfp6000.h"
+#include "nfp6000/nfp_xpb.h"
+#include "nfp_nffw.h"
+
+#define NFP_PL_DEVICE_ID                        0x00000004
+#define NFP_PL_DEVICE_ID_MASK                   0xff
+
+#define NFP6000_ARM_GCSR_SOFTMODEL0             0x00400144
+
+void
+nfp_cpp_priv_set(struct nfp_cpp *cpp, void *priv)
+{
+	cpp->priv = priv;
+}
+
+void *
+nfp_cpp_priv(struct nfp_cpp *cpp)
+{
+	return cpp->priv;
+}
+
+void
+nfp_cpp_model_set(struct nfp_cpp *cpp, uint32_t model)
+{
+	cpp->model = model;
+}
+
+uint32_t
+nfp_cpp_model(struct nfp_cpp *cpp)
+{
+	if (!cpp)
+		return NFP_CPP_MODEL_INVALID;
+
+	if (cpp->model == 0)
+		cpp->model = __nfp_cpp_model_autodetect(cpp);
+
+	return cpp->model;
+}
+
+void
+nfp_cpp_interface_set(struct nfp_cpp *cpp, uint32_t interface)
+{
+	cpp->interface = interface;
+}
+
+int
+nfp_cpp_serial(struct nfp_cpp *cpp, const uint8_t **serial)
+{
+	*serial = cpp->serial;
+	return cpp->serial_len;
+}
+
+int
+nfp_cpp_serial_set(struct nfp_cpp *cpp, const uint8_t *serial,
+		   size_t serial_len)
+{
+	if (cpp->serial_len)
+		free(cpp->serial);
+
+	cpp->serial = malloc(serial_len);
+	if (!cpp->serial)
+		return -1;
+
+	memcpy(cpp->serial, serial, serial_len);
+	cpp->serial_len = serial_len;
+
+	return 0;
+}
+
+uint16_t
+nfp_cpp_interface(struct nfp_cpp *cpp)
+{
+	if (!cpp)
+		return NFP_CPP_INTERFACE(NFP_CPP_INTERFACE_TYPE_INVALID, 0, 0);
+
+	return cpp->interface;
+}
+
+void *
+nfp_cpp_area_priv(struct nfp_cpp_area *cpp_area)
+{
+	return &cpp_area[1];
+}
+
+struct nfp_cpp *
+nfp_cpp_area_cpp(struct nfp_cpp_area *cpp_area)
+{
+	return cpp_area->cpp;
+}
+
+const char *
+nfp_cpp_area_name(struct nfp_cpp_area *cpp_area)
+{
+	return cpp_area->name;
+}
+
+/*
+ * nfp_cpp_area_alloc - allocate a new CPP area
+ * @cpp:    CPP handle
+ * @dest:   CPP id
+ * @address:    start address on CPP target
+ * @size:   size of area in bytes
+ *
+ * Allocate and initialize a CPP area structure.  The area must later
+ * be locked down with an 'acquire' before it can be safely accessed.
+ *
+ * NOTE: @address and @size must be 32-bit aligned values.
+ */
+struct nfp_cpp_area *
+nfp_cpp_area_alloc_with_name(struct nfp_cpp *cpp, uint32_t dest,
+			      const char *name, unsigned long long address,
+			      unsigned long size)
+{
+	struct nfp_cpp_area *area;
+	uint64_t tmp64 = (uint64_t)address;
+	int tmp, err;
+
+	if (!cpp)
+		return NULL;
+
+	/* CPP bus uses only a 40-bit address */
+	if ((address + size) > (1ULL << 40))
+		return NFP_ERRPTR(EFAULT);
+
+	/* Remap from cpp_island to cpp_target */
+	err = nfp_target_cpp(dest, tmp64, &dest, &tmp64, cpp->imb_cat_table);
+	if (err < 0)
+		return NULL;
+
+	address = (unsigned long long)tmp64;
+
+	if (!name)
+		name = "";
+
+	area = calloc(1, sizeof(*area) + cpp->op->area_priv_size +
+		      strlen(name) + 1);
+	if (!area)
+		return NULL;
+
+	area->cpp = cpp;
+	area->name = ((char *)area) + sizeof(*area) + cpp->op->area_priv_size;
+	memcpy(area->name, name, strlen(name) + 1);
+
+	/*
+	 * Preserve errno around the call to area_init, since most
+	 * implementations will blindly call nfp_target_action_width()for both
+	 * read or write modes, and that will set errno to EINVAL.
+	 */
+	tmp = errno;
+
+	err = cpp->op->area_init(area, dest, address, size);
+	if (err < 0) {
+		free(area);
+		return NULL;
+	}
+
+	/* Restore errno */
+	errno = tmp;
+
+	area->offset = address;
+	area->size = size;
+
+	return area;
+}
+
+struct nfp_cpp_area *
+nfp_cpp_area_alloc(struct nfp_cpp *cpp, uint32_t dest,
+		    unsigned long long address, unsigned long size)
+{
+	return nfp_cpp_area_alloc_with_name(cpp, dest, NULL, address, size);
+}
+
+/*
+ * nfp_cpp_area_alloc_acquire - allocate a new CPP area and lock it down
+ *
+ * @cpp:    CPP handle
+ * @dest:   CPP id
+ * @address:    start address on CPP target
+ * @size:   size of area
+ *
+ * Allocate and initilizae a CPP area structure, and lock it down so
+ * that it can be accessed directly.
+ *
+ * NOTE: @address and @size must be 32-bit aligned values.
+ *
+ * NOTE: The area must also be 'released' when the structure is freed.
+ */
+struct nfp_cpp_area *
+nfp_cpp_area_alloc_acquire(struct nfp_cpp *cpp, uint32_t destination,
+			    unsigned long long address, unsigned long size)
+{
+	struct nfp_cpp_area *area;
+
+	area = nfp_cpp_area_alloc(cpp, destination, address, size);
+	if (!area)
+		return NULL;
+
+	if (nfp_cpp_area_acquire(area)) {
+		nfp_cpp_area_free(area);
+		return NULL;
+	}
+
+	return area;
+}
+
+/*
+ * nfp_cpp_area_free - free up the CPP area
+ * area:    CPP area handle
+ *
+ * Frees up memory resources held by the CPP area.
+ */
+void
+nfp_cpp_area_free(struct nfp_cpp_area *area)
+{
+	if (area->cpp->op->area_cleanup)
+		area->cpp->op->area_cleanup(area);
+	free(area);
+}
+
+/*
+ * nfp_cpp_area_release_free - release CPP area and free it
+ * area:    CPP area handle
+ *
+ * Releases CPP area and frees up memory resources held by the it.
+ */
+void
+nfp_cpp_area_release_free(struct nfp_cpp_area *area)
+{
+	nfp_cpp_area_release(area);
+	nfp_cpp_area_free(area);
+}
+
+/*
+ * nfp_cpp_area_acquire - lock down a CPP area for access
+ * @area:   CPP area handle
+ *
+ * Locks down the CPP area for a potential long term activity.  Area
+ * must always be locked down before being accessed.
+ */
+int
+nfp_cpp_area_acquire(struct nfp_cpp_area *area)
+{
+	if (area->cpp->op->area_acquire) {
+		int err = area->cpp->op->area_acquire(area);
+
+		if (err < 0)
+			return -1;
+	}
+
+	return 0;
+}
+
+/*
+ * nfp_cpp_area_release - release a locked down CPP area
+ * @area:   CPP area handle
+ *
+ * Releases a previously locked down CPP area.
+ */
+void
+nfp_cpp_area_release(struct nfp_cpp_area *area)
+{
+	if (area->cpp->op->area_release)
+		area->cpp->op->area_release(area);
+}
+
+/*
+ * nfp_cpp_area_iomem() - get IOMEM region for CPP area
+ *
+ * @area:       CPP area handle
+ *
+ * Returns an iomem pointer for use with readl()/writel() style operations.
+ *
+ * NOTE: Area must have been locked down with an 'acquire'.
+ *
+ * Return: pointer to the area, or NULL
+ */
+void *
+nfp_cpp_area_iomem(struct nfp_cpp_area *area)
+{
+	void *iomem = NULL;
+
+	if (area->cpp->op->area_iomem)
+		iomem = area->cpp->op->area_iomem(area);
+
+	return iomem;
+}
+
+/*
+ * nfp_cpp_area_read - read data from CPP area
+ *
+ * @area:       CPP area handle
+ * @offset:     offset into CPP area
+ * @kernel_vaddr:   kernel address to put data into
+ * @length:     number of bytes to read
+ *
+ * Read data from indicated CPP region.
+ *
+ * NOTE: @offset and @length must be 32-bit aligned values.
+ *
+ * NOTE: Area must have been locked down with an 'acquire'.
+ */
+int
+nfp_cpp_area_read(struct nfp_cpp_area *area, unsigned long offset,
+		  void *kernel_vaddr, size_t length)
+{
+	if ((offset + length) > area->size)
+		return NFP_ERRNO(EFAULT);
+
+	return area->cpp->op->area_read(area, kernel_vaddr, offset, length);
+}
+
+/*
+ * nfp_cpp_area_write - write data to CPP area
+ *
+ * @area:       CPP area handle
+ * @offset:     offset into CPP area
+ * @kernel_vaddr:   kernel address to read data from
+ * @length:     number of bytes to write
+ *
+ * Write data to indicated CPP region.
+ *
+ * NOTE: @offset and @length must be 32-bit aligned values.
+ *
+ * NOTE: Area must have been locked down with an 'acquire'.
+ */
+int
+nfp_cpp_area_write(struct nfp_cpp_area *area, unsigned long offset,
+		   const void *kernel_vaddr, size_t length)
+{
+	if ((offset + length) > area->size)
+		return NFP_ERRNO(EFAULT);
+
+	return area->cpp->op->area_write(area, kernel_vaddr, offset, length);
+}
+
+void *
+nfp_cpp_area_mapped(struct nfp_cpp_area *area)
+{
+	if (area->cpp->op->area_mapped)
+		return area->cpp->op->area_mapped(area);
+	return NULL;
+}
+
+/*
+ * nfp_cpp_area_check_range - check if address range fits in CPP area
+ *
+ * @area:   CPP area handle
+ * @offset: offset into CPP area
+ * @length: size of address range in bytes
+ *
+ * Check if address range fits within CPP area.  Return 0 if area fits
+ * or -1 on error.
+ */
+int
+nfp_cpp_area_check_range(struct nfp_cpp_area *area, unsigned long long offset,
+			 unsigned long length)
+{
+	if (((offset + length) > area->size))
+		return NFP_ERRNO(EFAULT);
+
+	return 0;
+}
+
+/*
+ * Return the correct CPP address, and fixup xpb_addr as needed,
+ * based upon NFP model.
+ */
+static uint32_t
+nfp_xpb_to_cpp(struct nfp_cpp *cpp, uint32_t *xpb_addr)
+{
+	uint32_t xpb;
+	int island;
+
+	if (!NFP_CPP_MODEL_IS_6000(cpp->model))
+		return 0;
+
+	xpb = NFP_CPP_ID(14, NFP_CPP_ACTION_RW, 0);
+
+	/*
+	 * Ensure that non-local XPB accesses go out through the
+	 * global XPBM bus.
+	 */
+	island = ((*xpb_addr) >> 24) & 0x3f;
+
+	if (!island)
+		return xpb;
+
+	if (island == 1) {
+		/*
+		 * Accesses to the ARM Island overlay uses Island 0
+		 * Global Bit
+		 */
+		(*xpb_addr) &= ~0x7f000000;
+		if (*xpb_addr < 0x60000)
+			*xpb_addr |= (1 << 30);
+		else
+			/* And only non-ARM interfaces use island id = 1 */
+			if (NFP_CPP_INTERFACE_TYPE_of(nfp_cpp_interface(cpp)) !=
+			    NFP_CPP_INTERFACE_TYPE_ARM)
+				*xpb_addr |= (1 << 24);
+	} else {
+		(*xpb_addr) |= (1 << 30);
+	}
+
+	return xpb;
+}
+
+int
+nfp_cpp_area_readl(struct nfp_cpp_area *area, unsigned long offset,
+		   uint32_t *value)
+{
+	int sz;
+	uint32_t tmp;
+
+	sz = nfp_cpp_area_read(area, offset, &tmp, sizeof(tmp));
+	*value = rte_le_to_cpu_32(tmp);
+
+	return (sz == sizeof(*value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_area_writel(struct nfp_cpp_area *area, unsigned long offset,
+		    uint32_t value)
+{
+	int sz;
+
+	value = rte_cpu_to_le_32(value);
+	sz = nfp_cpp_area_write(area, offset, &value, sizeof(value));
+	return (sz == sizeof(value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_area_readq(struct nfp_cpp_area *area, unsigned long offset,
+		   uint64_t *value)
+{
+	int sz;
+	uint64_t tmp;
+
+	sz = nfp_cpp_area_read(area, offset, &tmp, sizeof(tmp));
+	*value = rte_le_to_cpu_64(tmp);
+
+	return (sz == sizeof(*value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_area_writeq(struct nfp_cpp_area *area, unsigned long offset,
+		    uint64_t value)
+{
+	int sz;
+
+	value = rte_cpu_to_le_64(value);
+	sz = nfp_cpp_area_write(area, offset, &value, sizeof(value));
+
+	return (sz == sizeof(value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_readl(struct nfp_cpp *cpp, uint32_t cpp_id, unsigned long long address,
+	      uint32_t *value)
+{
+	int sz;
+	uint32_t tmp;
+
+	sz = nfp_cpp_read(cpp, cpp_id, address, &tmp, sizeof(tmp));
+	*value = rte_le_to_cpu_32(tmp);
+
+	return (sz == sizeof(*value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_writel(struct nfp_cpp *cpp, uint32_t cpp_id, unsigned long long address,
+	       uint32_t value)
+{
+	int sz;
+
+	value = rte_cpu_to_le_32(value);
+	sz = nfp_cpp_write(cpp, cpp_id, address, &value, sizeof(value));
+
+	return (sz == sizeof(value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_readq(struct nfp_cpp *cpp, uint32_t cpp_id, unsigned long long address,
+	      uint64_t *value)
+{
+	int sz;
+	uint64_t tmp;
+
+	sz = nfp_cpp_read(cpp, cpp_id, address, &tmp, sizeof(tmp));
+	*value = rte_le_to_cpu_64(tmp);
+
+	return (sz == sizeof(*value)) ? 0 : -1;
+}
+
+int
+nfp_cpp_writeq(struct nfp_cpp *cpp, uint32_t cpp_id, unsigned long long address,
+	       uint64_t value)
+{
+	int sz;
+
+	value = rte_cpu_to_le_64(value);
+	sz = nfp_cpp_write(cpp, cpp_id, address, &value, sizeof(value));
+
+	return (sz == sizeof(value)) ? 0 : -1;
+}
+
+int
+nfp_xpb_writel(struct nfp_cpp *cpp, uint32_t xpb_addr, uint32_t value)
+{
+	uint32_t cpp_dest;
+
+	cpp_dest = nfp_xpb_to_cpp(cpp, &xpb_addr);
+
+	return nfp_cpp_writel(cpp, cpp_dest, xpb_addr, value);
+}
+
+int
+nfp_xpb_readl(struct nfp_cpp *cpp, uint32_t xpb_addr, uint32_t *value)
+{
+	uint32_t cpp_dest;
+
+	cpp_dest = nfp_xpb_to_cpp(cpp, &xpb_addr);
+
+	return nfp_cpp_readl(cpp, cpp_dest, xpb_addr, value);
+}
+
+static struct nfp_cpp *
+nfp_cpp_alloc(const char *devname)
+{
+	const struct nfp_cpp_operations *ops;
+	struct nfp_cpp *cpp;
+	int err;
+
+	ops = nfp_cpp_transport_operations();
+
+	if (!ops || !ops->init)
+		return NFP_ERRPTR(EINVAL);
+
+	cpp = calloc(1, sizeof(*cpp));
+	if (!cpp)
+		return NULL;
+
+	cpp->op = ops;
+
+	if (cpp->op->init) {
+		err = cpp->op->init(cpp, devname);
+		if (err < 0) {
+			free(cpp);
+			return NULL;
+		}
+	}
+
+	if (NFP_CPP_MODEL_IS_6000(nfp_cpp_model(cpp))) {
+		uint32_t xpbaddr;
+		size_t tgt;
+
+		for (tgt = 0; tgt < ARRAY_SIZE(cpp->imb_cat_table); tgt++) {
+			/* Hardcoded XPB IMB Base, island 0 */
+			xpbaddr = 0x000a0000 + (tgt * 4);
+			err = nfp_xpb_readl(cpp, xpbaddr,
+				(uint32_t *)&cpp->imb_cat_table[tgt]);
+			if (err < 0) {
+				free(cpp);
+				return NULL;
+			}
+		}
+	}
+
+	return cpp;
+}
+
+/*
+ * nfp_cpp_free - free the CPP handle
+ * @cpp:    CPP handle
+ */
+void
+nfp_cpp_free(struct nfp_cpp *cpp)
+{
+	if (cpp->op && cpp->op->free)
+		cpp->op->free(cpp);
+
+	if (cpp->serial_len)
+		free(cpp->serial);
+
+	free(cpp);
+}
+
+struct nfp_cpp *
+nfp_cpp_from_device_name(const char *devname)
+{
+	return nfp_cpp_alloc(devname);
+}
+
+/*
+ * Modify bits of a 32-bit value from the XPB bus
+ *
+ * @param cpp           NFP CPP device handle
+ * @param xpb_tgt       XPB target and address
+ * @param mask          mask of bits to alter
+ * @param value         value to modify
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int
+nfp_xpb_writelm(struct nfp_cpp *cpp, uint32_t xpb_tgt, uint32_t mask,
+		uint32_t value)
+{
+	int err;
+	uint32_t tmp;
+
+	err = nfp_xpb_readl(cpp, xpb_tgt, &tmp);
+	if (err < 0)
+		return err;
+
+	tmp &= ~mask;
+	tmp |= (mask & value);
+	return nfp_xpb_writel(cpp, xpb_tgt, tmp);
+}
+
+/*
+ * Modify bits of a 32-bit value from the XPB bus
+ *
+ * @param cpp           NFP CPP device handle
+ * @param xpb_tgt       XPB target and address
+ * @param mask          mask of bits to alter
+ * @param value         value to monitor for
+ * @param timeout_us    maximum number of us to wait (-1 for forever)
+ *
+ * @return >= 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int
+nfp_xpb_waitlm(struct nfp_cpp *cpp, uint32_t xpb_tgt, uint32_t mask,
+	       uint32_t value, int timeout_us)
+{
+	uint32_t tmp;
+	int err;
+
+	do {
+		err = nfp_xpb_readl(cpp, xpb_tgt, &tmp);
+		if (err < 0)
+			goto exit;
+
+		if ((tmp & mask) == (value & mask)) {
+			if (timeout_us < 0)
+				timeout_us = 0;
+			break;
+		}
+
+		if (timeout_us < 0)
+			continue;
+
+		timeout_us -= 100;
+		usleep(100);
+	} while (timeout_us >= 0);
+
+	if (timeout_us < 0)
+		err = NFP_ERRNO(ETIMEDOUT);
+	else
+		err = timeout_us;
+
+exit:
+	return err;
+}
+
+/*
+ * nfp_cpp_read - read from CPP target
+ * @cpp:        CPP handle
+ * @destination:    CPP id
+ * @address:        offset into CPP target
+ * @kernel_vaddr:   kernel buffer for result
+ * @length:     number of bytes to read
+ */
+int
+nfp_cpp_read(struct nfp_cpp *cpp, uint32_t destination,
+	     unsigned long long address, void *kernel_vaddr, size_t length)
+{
+	struct nfp_cpp_area *area;
+	int err;
+
+	area = nfp_cpp_area_alloc_acquire(cpp, destination, address, length);
+	if (!area) {
+		printf("Area allocation/acquire failed\n");
+		return -1;
+	}
+
+	err = nfp_cpp_area_read(area, 0, kernel_vaddr, length);
+
+	nfp_cpp_area_release_free(area);
+	return err;
+}
+
+/*
+ * nfp_cpp_write - write to CPP target
+ * @cpp:        CPP handle
+ * @destination:    CPP id
+ * @address:        offset into CPP target
+ * @kernel_vaddr:   kernel buffer to read from
+ * @length:     number of bytes to write
+ */
+int
+nfp_cpp_write(struct nfp_cpp *cpp, uint32_t destination,
+	      unsigned long long address, const void *kernel_vaddr,
+	      size_t length)
+{
+	struct nfp_cpp_area *area;
+	int err;
+
+	area = nfp_cpp_area_alloc_acquire(cpp, destination, address, length);
+	if (!area)
+		return -1;
+
+	err = nfp_cpp_area_write(area, 0, kernel_vaddr, length);
+
+	nfp_cpp_area_release_free(area);
+	return err;
+}
+
+/*
+ * nfp_cpp_area_fill - fill a CPP area with a value
+ * @area:       CPP area
+ * @offset:     offset into CPP area
+ * @value:      value to fill with
+ * @length:     length of area to fill
+ */
+int
+nfp_cpp_area_fill(struct nfp_cpp_area *area, unsigned long offset,
+		  uint32_t value, size_t length)
+{
+	int err;
+	size_t i;
+	uint64_t value64;
+
+	value = rte_cpu_to_le_32(value);
+	value64 = ((uint64_t)value << 32) | value;
+
+	if ((offset + length) > area->size)
+		return NFP_ERRNO(EINVAL);
+
+	if ((area->offset + offset) & 3)
+		return NFP_ERRNO(EINVAL);
+
+	if (((area->offset + offset) & 7) == 4 && length >= 4) {
+		err = nfp_cpp_area_write(area, offset, &value, sizeof(value));
+		if (err < 0)
+			return err;
+		if (err != sizeof(value))
+			return NFP_ERRNO(ENOSPC);
+		offset += sizeof(value);
+		length -= sizeof(value);
+	}
+
+	for (i = 0; (i + sizeof(value)) < length; i += sizeof(value64)) {
+		err =
+		    nfp_cpp_area_write(area, offset + i, &value64,
+				       sizeof(value64));
+		if (err < 0)
+			return err;
+		if (err != sizeof(value64))
+			return NFP_ERRNO(ENOSPC);
+	}
+
+	if ((i + sizeof(value)) <= length) {
+		err =
+		    nfp_cpp_area_write(area, offset + i, &value, sizeof(value));
+		if (err < 0)
+			return err;
+		if (err != sizeof(value))
+			return NFP_ERRNO(ENOSPC);
+		i += sizeof(value);
+	}
+
+	return (int)i;
+}
+
+static inline uint8_t
+__nfp_bytemask_of(int width, uint64_t addr)
+{
+	uint8_t byte_mask;
+
+	if (width == 8)
+		byte_mask = 0xff;
+	else if (width == 4)
+		byte_mask = 0x0f << (addr & 4);
+	else if (width == 2)
+		byte_mask = 0x03 << (addr & 6);
+	else if (width == 1)
+		byte_mask = 0x01 << (addr & 7);
+	else
+		byte_mask = 0;
+
+	return byte_mask;
+}
+
+/*
+ * NOTE: This code should not use nfp_xpb_* functions,
+ * as those are model-specific
+ */
+uint32_t
+__nfp_cpp_model_autodetect(struct nfp_cpp *cpp)
+{
+	uint32_t arm_id = NFP_CPP_ID(NFP_CPP_TARGET_ARM, 0, 0);
+	uint32_t model = 0;
+
+	nfp_cpp_readl(cpp, arm_id, NFP6000_ARM_GCSR_SOFTMODEL0, &model);
+
+	if (NFP_CPP_MODEL_IS_6000(model)) {
+		uint32_t tmp;
+
+		nfp_cpp_model_set(cpp, model);
+
+		/* The PL's PluDeviceID revision code is authoratative */
+		model &= ~0xff;
+		nfp_xpb_readl(cpp, NFP_XPB_DEVICE(1, 1, 16) +
+				   NFP_PL_DEVICE_ID, &tmp);
+		model |= (NFP_PL_DEVICE_ID_MASK & tmp) - 0x10;
+	}
+
+	return model;
+}
+
+/*
+ * nfp_cpp_map_area() - Helper function to map an area
+ * @cpp:    NFP CPP handler
+ * @domain: CPP domain
+ * @target: CPP target
+ * @addr:   CPP address
+ * @size:   Size of the area
+ * @area:   Area handle (output)
+ *
+ * Map an area of IOMEM access.  To undo the effect of this function call
+ * @nfp_cpp_area_release_free(*area).
+ *
+ * Return: Pointer to memory mapped area or ERR_PTR
+ */
+uint8_t *
+nfp_cpp_map_area(struct nfp_cpp *cpp, int domain, int target, uint64_t addr,
+		 unsigned long size, struct nfp_cpp_area **area)
+{
+	uint8_t *res;
+	uint32_t dest;
+
+	dest = NFP_CPP_ISLAND_ID(target, NFP_CPP_ACTION_RW, 0, domain);
+
+	*area = nfp_cpp_area_alloc_acquire(cpp, dest, addr, size);
+	if (!*area)
+		goto err_eio;
+
+	res = nfp_cpp_area_iomem(*area);
+	if (!res)
+		goto err_release_free;
+
+	return res;
+
+err_release_free:
+	nfp_cpp_area_release_free(*area);
+err_eio:
+	return NULL;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_crc.c b/drivers/net/nfp/nfpcore/nfp_crc.c
new file mode 100644
index 0000000..a3c0e92
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_crc.c
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <stdio.h>
+#include <inttypes.h>
+
+#include "nfp_crc.h"
+
+static inline uint32_t
+nfp_crc32_be_generic(uint32_t crc, unsigned char const *p, size_t len,
+		 uint32_t polynomial)
+{
+	int i;
+	while (len--) {
+		crc ^= *p++ << 24;
+		for (i = 0; i < 8; i++)
+			crc = (crc << 1) ^ ((crc & 0x80000000) ? polynomial :
+					  0);
+	}
+	return crc;
+}
+
+static inline uint32_t
+nfp_crc32_be(uint32_t crc, unsigned char const *p, size_t len)
+{
+	return nfp_crc32_be_generic(crc, p, len, CRCPOLY_BE);
+}
+
+static uint32_t
+nfp_crc32_posix_end(uint32_t crc, size_t total_len)
+{
+	/* Extend with the length of the string. */
+	while (total_len != 0) {
+		uint8_t c = total_len & 0xff;
+
+		crc = nfp_crc32_be(crc, &c, 1);
+		total_len >>= 8;
+	}
+
+	return ~crc;
+}
+
+uint32_t
+nfp_crc32_posix(const void *buff, size_t len)
+{
+	return nfp_crc32_posix_end(nfp_crc32_be(0, buff, len), len);
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_crc.h b/drivers/net/nfp/nfpcore/nfp_crc.h
new file mode 100644
index 0000000..f8a112b
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_crc.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_CRC_H__
+#define __NFP_CRC_H__
+
+/*
+ * There are multiple 16-bit CRC polynomials in common use, but this is
+ * *the* standard CRC-32 polynomial, first popularized by Ethernet.
+ * x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x^1+x^0
+ */
+#define CRCPOLY_LE 0xedb88320
+#define CRCPOLY_BE 0x04c11db7
+
+uint32_t nfp_crc32_posix(const void *buff, size_t len);
+
+#endif
diff --git a/drivers/net/nfp/nfpcore/nfp_hwinfo.c b/drivers/net/nfp/nfpcore/nfp_hwinfo.c
new file mode 100644
index 0000000..b8d6400
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_hwinfo.c
@@ -0,0 +1,226 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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.
+ */
+
+/* Parse the hwinfo table that the ARM firmware builds in the ARM scratch SRAM
+ * after chip reset.
+ *
+ * Examples of the fields:
+ *   me.count = 40
+ *   me.mask = 0x7f_ffff_ffff
+ *
+ *   me.count is the total number of MEs on the system.
+ *   me.mask is the bitmask of MEs that are available for application usage.
+ *
+ *   (ie, in this example, ME 39 has been reserved by boardconfig.)
+ */
+
+#include <stdio.h>
+#include <time.h>
+#include <zlib.h>
+
+#include "nfp_cpp.h"
+#include "nfp6000/nfp6000.h"
+#include "nfp_resource.h"
+#include "nfp_hwinfo.h"
+#include "nfp_crc.h"
+
+static int
+nfp_hwinfo_is_updating(struct nfp_hwinfo *hwinfo)
+{
+	return hwinfo->version & NFP_HWINFO_VERSION_UPDATING;
+}
+
+static int
+nfp_hwinfo_db_walk(struct nfp_hwinfo *hwinfo, uint32_t size)
+{
+	const char *key, *val, *end = hwinfo->data + size;
+
+	for (key = hwinfo->data; *key && key < end;
+	     key = val + strlen(val) + 1) {
+		val = key + strlen(key) + 1;
+		if (val >= end) {
+			printf("Bad HWINFO - overflowing key\n");
+			return -EINVAL;
+		}
+
+		if (val + strlen(val) + 1 > end) {
+			printf("Bad HWINFO - overflowing value\n");
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
+static int
+nfp_hwinfo_db_validate(struct nfp_hwinfo *db, uint32_t len)
+{
+	uint32_t size, new_crc, *crc;
+
+	size = db->size;
+	if (size > len) {
+		printf("Unsupported hwinfo size %u > %u\n", size, len);
+		return -EINVAL;
+	}
+
+	size -= sizeof(uint32_t);
+	new_crc = nfp_crc32_posix((char *)db, size);
+	crc = (uint32_t *)(db->start + size);
+	if (new_crc != *crc) {
+		printf("Corrupt hwinfo table (CRC mismatch)\n");
+		printf("\tcalculated 0x%x, expected 0x%x\n", new_crc, *crc);
+		return -EINVAL;
+	}
+
+	return nfp_hwinfo_db_walk(db, size);
+}
+
+static struct nfp_hwinfo *
+nfp_hwinfo_try_fetch(struct nfp_cpp *cpp, size_t *cpp_size)
+{
+	struct nfp_hwinfo *header;
+	void *res;
+	uint64_t cpp_addr;
+	uint32_t cpp_id;
+	int err;
+	uint8_t *db;
+
+	res = nfp_resource_acquire(cpp, NFP_RESOURCE_NFP_HWINFO);
+	if (res) {
+		cpp_id = nfp_resource_cpp_id(res);
+		cpp_addr = nfp_resource_address(res);
+		*cpp_size = nfp_resource_size(res);
+
+		nfp_resource_release(res);
+
+		if (*cpp_size < HWINFO_SIZE_MIN)
+			return NULL;
+	} else {
+		return NULL;
+	}
+
+	db = malloc(*cpp_size + 1);
+	if (!db)
+		return NULL;
+
+	err = nfp_cpp_read(cpp, cpp_id, cpp_addr, db, *cpp_size);
+	if (err != (int)*cpp_size)
+		goto exit_free;
+
+	header = (void *)db;
+	printf("NFP HWINFO header: %08x\n", *(uint32_t *)header);
+	if (nfp_hwinfo_is_updating(header))
+		goto exit_free;
+
+	if (header->version != NFP_HWINFO_VERSION_2) {
+		printf("Unknown HWInfo version: 0x%08x\n",
+			header->version);
+		goto exit_free;
+	}
+
+	/* NULL-terminate for safety */
+	db[*cpp_size] = '\0';
+
+	return (void *)db;
+exit_free:
+	free(db);
+	return NULL;
+}
+
+static struct nfp_hwinfo *
+nfp_hwinfo_fetch(struct nfp_cpp *cpp, size_t *hwdb_size)
+{
+	struct timespec wait;
+	struct nfp_hwinfo *db;
+	int count;
+
+	wait.tv_sec = 0;
+	wait.tv_nsec = 10000000;
+	count = 0;
+
+	for (;;) {
+		db = nfp_hwinfo_try_fetch(cpp, hwdb_size);
+		if (db)
+			return db;
+
+		nanosleep(&wait, NULL);
+		if (count++ > 200) {
+			printf("NFP access error\n");
+			return NULL;
+		}
+	}
+}
+
+struct nfp_hwinfo *
+nfp_hwinfo_read(struct nfp_cpp *cpp)
+{
+	struct nfp_hwinfo *db;
+	size_t hwdb_size = 0;
+	int err;
+
+	db = nfp_hwinfo_fetch(cpp, &hwdb_size);
+	if (!db)
+		return NULL;
+
+	err = nfp_hwinfo_db_validate(db, hwdb_size);
+	if (err) {
+		free(db);
+		return NULL;
+	}
+	return db;
+}
+
+/*
+ * nfp_hwinfo_lookup() - Find a value in the HWInfo table by name
+ * @hwinfo:	NFP HWinfo table
+ * @lookup:	HWInfo name to search for
+ *
+ * Return: Value of the HWInfo name, or NULL
+ */
+const char *
+nfp_hwinfo_lookup(struct nfp_hwinfo *hwinfo, const char *lookup)
+{
+	const char *key, *val, *end;
+
+	if (!hwinfo || !lookup)
+		return NULL;
+
+	end = hwinfo->data + hwinfo->size - sizeof(uint32_t);
+
+	for (key = hwinfo->data; *key && key < end;
+	     key = val + strlen(val) + 1) {
+		val = key + strlen(key) + 1;
+
+		if (strcmp(key, lookup) == 0)
+			return val;
+	}
+
+	return NULL;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_hwinfo.h b/drivers/net/nfp/nfpcore/nfp_hwinfo.h
new file mode 100644
index 0000000..e9a9b49
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_hwinfo.h
@@ -0,0 +1,111 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_HWINFO_H__
+#define __NFP_HWINFO_H__
+
+#include <inttypes.h>
+
+#define HWINFO_SIZE_MIN	0x100
+
+/*
+ * The Hardware Info Table defines the properties of the system.
+ *
+ * HWInfo v1 Table (fixed size)
+ *
+ * 0x0000: uint32_t version	        Hardware Info Table version (1.0)
+ * 0x0004: uint32_t size	        Total size of the table, including the
+ *					CRC32 (IEEE 802.3)
+ * 0x0008: uint32_t jumptab	        Offset of key/value table
+ * 0x000c: uint32_t keys	        Total number of keys in the key/value
+ *					table
+ * NNNNNN:				Key/value jump table and string data
+ * (size - 4): uint32_t crc32	CRC32 (same as IEEE 802.3, POSIX csum, etc)
+ *				CRC32("",0) = ~0, CRC32("a",1) = 0x48C279FE
+ *
+ * HWInfo v2 Table (variable size)
+ *
+ * 0x0000: uint32_t version	        Hardware Info Table version (2.0)
+ * 0x0004: uint32_t size	        Current size of the data area, excluding
+ *					CRC32
+ * 0x0008: uint32_t limit	        Maximum size of the table
+ * 0x000c: uint32_t reserved	        Unused, set to zero
+ * NNNNNN:			Key/value data
+ * (size - 4): uint32_t crc32	CRC32 (same as IEEE 802.3, POSIX csum, etc)
+ *				CRC32("",0) = ~0, CRC32("a",1) = 0x48C279FE
+ *
+ * If the HWInfo table is in the process of being updated, the low bit of
+ * version will be set.
+ *
+ * HWInfo v1 Key/Value Table
+ * -------------------------
+ *
+ *  The key/value table is a set of offsets to ASCIIZ strings which have
+ *  been strcmp(3) sorted (yes, please use bsearch(3) on the table).
+ *
+ *  All keys are guaranteed to be unique.
+ *
+ * N+0:	uint32_t key_1		Offset to the first key
+ * N+4:	uint32_t val_1		Offset to the first value
+ * N+8: uint32_t key_2		Offset to the second key
+ * N+c: uint32_t val_2		Offset to the second value
+ * ...
+ *
+ * HWInfo v2 Key/Value Table
+ * -------------------------
+ *
+ * Packed UTF8Z strings, ie 'key1\000value1\000key2\000value2\000'
+ *
+ * Unsorted.
+ */
+
+#define NFP_HWINFO_VERSION_1 ('H' << 24 | 'I' << 16 | 1 << 8 | 0 << 1 | 0)
+#define NFP_HWINFO_VERSION_2 ('H' << 24 | 'I' << 16 | 2 << 8 | 0 << 1 | 0)
+#define NFP_HWINFO_VERSION_UPDATING	BIT(0)
+
+struct nfp_hwinfo {
+	uint8_t start[0];
+
+	uint32_t version;
+	uint32_t size;
+
+	/* v2 specific fields */
+	uint32_t limit;
+	uint32_t resv;
+
+	char data[];
+};
+
+struct nfp_hwinfo *nfp_hwinfo_read(struct nfp_cpp *cpp);
+
+const char *nfp_hwinfo_lookup(struct nfp_hwinfo *hwinfo, const char *lookup);
+
+#endif
diff --git a/drivers/net/nfp/nfpcore/nfp_mip.c b/drivers/net/nfp/nfpcore/nfp_mip.c
new file mode 100644
index 0000000..6e14d06
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_mip.c
@@ -0,0 +1,180 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <stdio.h>
+#include <rte_byteorder.h>
+
+#include "nfp_cpp.h"
+#include "nfp_mip.h"
+#include "nfp_nffw.h"
+
+#define NFP_MIP_SIGNATURE	rte_cpu_to_le_32(0x0050494d)  /* "MIP\0" */
+#define NFP_MIP_VERSION		rte_cpu_to_le_32(1)
+#define NFP_MIP_MAX_OFFSET	(256 * 1024)
+
+struct nfp_mip {
+	uint32_t signature;
+	uint32_t mip_version;
+	uint32_t mip_size;
+	uint32_t first_entry;
+
+	uint32_t version;
+	uint32_t buildnum;
+	uint32_t buildtime;
+	uint32_t loadtime;
+
+	uint32_t symtab_addr;
+	uint32_t symtab_size;
+	uint32_t strtab_addr;
+	uint32_t strtab_size;
+
+	char name[16];
+	char toolchain[32];
+};
+
+/* Read memory and check if it could be a valid MIP */
+static int
+nfp_mip_try_read(struct nfp_cpp *cpp, uint32_t cpp_id, uint64_t addr,
+		 struct nfp_mip *mip)
+{
+	int ret;
+
+	ret = nfp_cpp_read(cpp, cpp_id, addr, mip, sizeof(*mip));
+	if (ret != sizeof(*mip)) {
+		printf("Failed to read MIP data (%d, %zu)\n",
+			ret, sizeof(*mip));
+		return -EIO;
+	}
+	if (mip->signature != NFP_MIP_SIGNATURE) {
+		printf("Incorrect MIP signature (0x%08x)\n",
+			 rte_le_to_cpu_32(mip->signature));
+		return -EINVAL;
+	}
+	if (mip->mip_version != NFP_MIP_VERSION) {
+		printf("Unsupported MIP version (%d)\n",
+			 rte_le_to_cpu_32(mip->mip_version));
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Try to locate MIP using the resource table */
+static int
+nfp_mip_read_resource(struct nfp_cpp *cpp, struct nfp_mip *mip)
+{
+	struct nfp_nffw_info *nffw_info;
+	uint32_t cpp_id;
+	uint64_t addr;
+	int err;
+
+	nffw_info = nfp_nffw_info_open(cpp);
+	if (!nffw_info)
+		return -ENODEV;
+
+	err = nfp_nffw_info_mip_first(nffw_info, &cpp_id, &addr);
+	if (err)
+		goto exit_close_nffw;
+
+	err = nfp_mip_try_read(cpp, cpp_id, addr, mip);
+exit_close_nffw:
+	nfp_nffw_info_close(nffw_info);
+	return err;
+}
+
+/*
+ * nfp_mip_open() - Get device MIP structure
+ * @cpp:	NFP CPP Handle
+ *
+ * Copy MIP structure from NFP device and return it.  The returned
+ * structure is handled internally by the library and should be
+ * freed by calling nfp_mip_close().
+ *
+ * Return: pointer to mip, NULL on failure.
+ */
+struct nfp_mip *
+nfp_mip_open(struct nfp_cpp *cpp)
+{
+	struct nfp_mip *mip;
+	int err;
+
+	mip = malloc(sizeof(*mip));
+	if (!mip)
+		return NULL;
+
+	err = nfp_mip_read_resource(cpp, mip);
+	if (err) {
+		free(mip);
+		return NULL;
+	}
+
+	mip->name[sizeof(mip->name) - 1] = 0;
+
+	return mip;
+}
+
+void
+nfp_mip_close(struct nfp_mip *mip)
+{
+	free(mip);
+}
+
+const char *
+nfp_mip_name(const struct nfp_mip *mip)
+{
+	return mip->name;
+}
+
+/*
+ * nfp_mip_symtab() - Get the address and size of the MIP symbol table
+ * @mip:	MIP handle
+ * @addr:	Location for NFP DDR address of MIP symbol table
+ * @size:	Location for size of MIP symbol table
+ */
+void
+nfp_mip_symtab(const struct nfp_mip *mip, uint32_t *addr, uint32_t *size)
+{
+	*addr = rte_le_to_cpu_32(mip->symtab_addr);
+	*size = rte_le_to_cpu_32(mip->symtab_size);
+}
+
+/*
+ * nfp_mip_strtab() - Get the address and size of the MIP symbol name table
+ * @mip:	MIP handle
+ * @addr:	Location for NFP DDR address of MIP symbol name table
+ * @size:	Location for size of MIP symbol name table
+ */
+void
+nfp_mip_strtab(const struct nfp_mip *mip, uint32_t *addr, uint32_t *size)
+{
+	*addr = rte_le_to_cpu_32(mip->strtab_addr);
+	*size = rte_le_to_cpu_32(mip->strtab_size);
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_mip.h b/drivers/net/nfp/nfpcore/nfp_mip.h
new file mode 100644
index 0000000..48d8ef3
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_mip.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_MIP_H__
+#define __NFP_MIP_H__
+
+#include "nfp_nffw.h"
+
+struct nfp_mip;
+
+struct nfp_mip *nfp_mip_open(struct nfp_cpp *cpp);
+void nfp_mip_close(struct nfp_mip *mip);
+
+const char *nfp_mip_name(const struct nfp_mip *mip);
+void nfp_mip_symtab(const struct nfp_mip *mip, uint32_t *addr, uint32_t *size);
+void nfp_mip_strtab(const struct nfp_mip *mip, uint32_t *addr, uint32_t *size);
+int nfp_nffw_info_mip_first(struct nfp_nffw_info *state, uint32_t *cpp_id,
+			    uint64_t *off);
+#endif
diff --git a/drivers/net/nfp/nfpcore/nfp_mutex.c b/drivers/net/nfp/nfpcore/nfp_mutex.c
new file mode 100644
index 0000000..76dc143
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_mutex.c
@@ -0,0 +1,450 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <errno.h>
+
+#include <malloc.h>
+#include <time.h>
+#include <sched.h>
+
+#include "nfp_cpp.h"
+#include "nfp6000/nfp6000.h"
+
+#define MUTEX_LOCKED(interface)  ((((uint32_t)(interface)) << 16) | 0x000f)
+#define MUTEX_UNLOCK(interface)  (0                               | 0x0000)
+
+#define MUTEX_IS_LOCKED(value)   (((value) & 0xffff) == 0x000f)
+#define MUTEX_IS_UNLOCKED(value) (((value) & 0xffff) == 0x0000)
+#define MUTEX_INTERFACE(value)   (((value) >> 16) & 0xffff)
+
+/*
+ * If you need more than 65536 recursive locks, please
+ * rethink your code.
+ */
+#define MUTEX_DEPTH_MAX         0xffff
+
+struct nfp_cpp_mutex {
+	struct nfp_cpp *cpp;
+	uint8_t target;
+	uint16_t depth;
+	unsigned long long address;
+	uint32_t key;
+	unsigned int usage;
+	struct nfp_cpp_mutex *prev, *next;
+};
+
+static int
+_nfp_cpp_mutex_validate(uint32_t model, int *target, unsigned long long address)
+{
+	/* Address must be 64-bit aligned */
+	if (address & 7)
+		return NFP_ERRNO(EINVAL);
+
+	if (NFP_CPP_MODEL_IS_6000(model)) {
+		if (*target != NFP_CPP_TARGET_MU)
+			return NFP_ERRNO(EINVAL);
+	} else {
+		return NFP_ERRNO(EINVAL);
+	}
+
+	return 0;
+}
+
+/*
+ * Initialize a mutex location
+ *
+ * The CPP target:address must point to a 64-bit aligned location, and
+ * will initialize 64 bits of data at the location.
+ *
+ * This creates the initial mutex state, as locked by this
+ * nfp_cpp_interface().
+ *
+ * This function should only be called when setting up
+ * the initial lock state upon boot-up of the system.
+ *
+ * @param mutex     NFP CPP Mutex handle
+ * @param target    NFP CPP target ID (ie NFP_CPP_TARGET_CLS or
+ *		    NFP_CPP_TARGET_MU)
+ * @param address   Offset into the address space of the NFP CPP target ID
+ * @param key       Unique 32-bit value for this mutex
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int
+nfp_cpp_mutex_init(struct nfp_cpp *cpp, int target, unsigned long long address,
+		   uint32_t key)
+{
+	uint32_t model = nfp_cpp_model(cpp);
+	uint32_t muw = NFP_CPP_ID(target, 4, 0);	/* atomic_write */
+	int err;
+
+	err = _nfp_cpp_mutex_validate(model, &target, address);
+	if (err < 0)
+		return err;
+
+	err = nfp_cpp_writel(cpp, muw, address + 4, key);
+	if (err < 0)
+		return err;
+
+	err =
+	    nfp_cpp_writel(cpp, muw, address + 0,
+			   MUTEX_LOCKED(nfp_cpp_interface(cpp)));
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+/*
+ * Create a mutex handle from an address controlled by a MU Atomic engine
+ *
+ * The CPP target:address must point to a 64-bit aligned location, and
+ * reserve 64 bits of data at the location for use by the handle.
+ *
+ * Only target/address pairs that point to entities that support the
+ * MU Atomic Engine are supported.
+ *
+ * @param cpp       NFP CPP handle
+ * @param target    NFP CPP target ID (ie NFP_CPP_TARGET_CLS or
+ *		    NFP_CPP_TARGET_MU)
+ * @param address   Offset into the address space of the NFP CPP target ID
+ * @param key       32-bit unique key (must match the key at this location)
+ *
+ * @return      A non-NULL struct nfp_cpp_mutex * on success, NULL on failure.
+ */
+struct nfp_cpp_mutex *
+nfp_cpp_mutex_alloc(struct nfp_cpp *cpp, int target,
+		     unsigned long long address, uint32_t key)
+{
+	uint32_t model = nfp_cpp_model(cpp);
+	struct nfp_cpp_mutex *mutex;
+	uint32_t mur = NFP_CPP_ID(target, 3, 0);	/* atomic_read */
+	int err;
+	uint32_t tmp;
+
+	/* Look for cached mutex */
+	for (mutex = cpp->mutex_cache; mutex; mutex = mutex->next) {
+		if (mutex->target == target && mutex->address == address)
+			break;
+	}
+
+	if (mutex) {
+		if (mutex->key == key) {
+			mutex->usage++;
+			return mutex;
+		}
+
+		/* If the key doesn't match... */
+		return NFP_ERRPTR(EEXIST);
+	}
+
+	err = _nfp_cpp_mutex_validate(model, &target, address);
+	if (err < 0)
+		return NULL;
+
+	err = nfp_cpp_readl(cpp, mur, address + 4, &tmp);
+	if (err < 0)
+		return NULL;
+
+	if (tmp != key)
+		return NFP_ERRPTR(EEXIST);
+
+	mutex = calloc(sizeof(*mutex), 1);
+	if (!mutex)
+		return NFP_ERRPTR(ENOMEM);
+
+	mutex->cpp = cpp;
+	mutex->target = target;
+	mutex->address = address;
+	mutex->key = key;
+	mutex->depth = 0;
+	mutex->usage = 1;
+
+	/* Add mutex to the cache */
+	if (cpp->mutex_cache) {
+		cpp->mutex_cache->prev = mutex;
+		mutex->next = cpp->mutex_cache;
+		cpp->mutex_cache = mutex;
+	} else {
+		cpp->mutex_cache = mutex;
+	}
+
+	return mutex;
+}
+
+struct nfp_cpp *
+nfp_cpp_mutex_cpp(struct nfp_cpp_mutex *mutex)
+{
+	return mutex->cpp;
+}
+
+uint32_t
+nfp_cpp_mutex_key(struct nfp_cpp_mutex *mutex)
+{
+	return mutex->key;
+}
+
+uint16_t
+nfp_cpp_mutex_owner(struct nfp_cpp_mutex *mutex)
+{
+	uint32_t mur = NFP_CPP_ID(mutex->target, 3, 0);	/* atomic_read */
+	uint32_t value, key;
+	int err;
+
+	err = nfp_cpp_readl(mutex->cpp, mur, mutex->address, &value);
+	if (err < 0)
+		return err;
+
+	err = nfp_cpp_readl(mutex->cpp, mur, mutex->address + 4, &key);
+	if (err < 0)
+		return err;
+
+	if (key != mutex->key)
+		return NFP_ERRNO(EPERM);
+
+	if (!MUTEX_IS_LOCKED(value))
+		return 0;
+
+	return MUTEX_INTERFACE(value);
+}
+
+int
+nfp_cpp_mutex_target(struct nfp_cpp_mutex *mutex)
+{
+	return mutex->target;
+}
+
+uint64_t
+nfp_cpp_mutex_address(struct nfp_cpp_mutex *mutex)
+{
+	return mutex->address;
+}
+
+/*
+ * Free a mutex handle - does not alter the lock state
+ *
+ * @param mutex     NFP CPP Mutex handle
+ */
+void
+nfp_cpp_mutex_free(struct nfp_cpp_mutex *mutex)
+{
+	mutex->usage--;
+	if (mutex->usage > 0)
+		return;
+
+	/* Remove mutex from the cache */
+	if (mutex->next)
+		mutex->next->prev = mutex->prev;
+	if (mutex->prev)
+		mutex->prev->next = mutex->next;
+
+	/* If mutex->cpp == NULL, something broke */
+	if (mutex->cpp && mutex == mutex->cpp->mutex_cache)
+		mutex->cpp->mutex_cache = mutex->next;
+
+	free(mutex);
+}
+
+/*
+ * Lock a mutex handle, using the NFP MU Atomic Engine
+ *
+ * @param mutex     NFP CPP Mutex handle
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int
+nfp_cpp_mutex_lock(struct nfp_cpp_mutex *mutex)
+{
+	int err;
+	time_t warn_at = time(NULL) + 15;
+
+	while ((err = nfp_cpp_mutex_trylock(mutex)) != 0) {
+		/* If errno != EBUSY, then the lock was damaged */
+		if (err < 0 && errno != EBUSY)
+			return err;
+		if (time(NULL) >= warn_at) {
+			printf("Warning: waiting for NFP mutex\n");
+			printf("\tusage:%hd\n", mutex->usage);
+			printf("\tdepth:%hd]\n", mutex->depth);
+			printf("\ttarget:%d\n", mutex->target);
+			printf("\taddr:%llx\n", mutex->address);
+			printf("\tkey:%08x]\n", mutex->key);
+			warn_at = time(NULL) + 60;
+		}
+		sched_yield();
+	}
+	return 0;
+}
+
+/*
+ * Unlock a mutex handle, using the NFP MU Atomic Engine
+ *
+ * @param mutex     NFP CPP Mutex handle
+ *
+ * @return 0 on success, or -1 on failure (and set errno accordingly).
+ */
+int
+nfp_cpp_mutex_unlock(struct nfp_cpp_mutex *mutex)
+{
+	uint32_t muw = NFP_CPP_ID(mutex->target, 4, 0);	/* atomic_write */
+	uint32_t mur = NFP_CPP_ID(mutex->target, 3, 0);	/* atomic_read */
+	struct nfp_cpp *cpp = mutex->cpp;
+	uint32_t key, value;
+	uint16_t interface = nfp_cpp_interface(cpp);
+	int err;
+
+	if (mutex->depth > 1) {
+		mutex->depth--;
+		return 0;
+	}
+
+	err = nfp_cpp_readl(mutex->cpp, mur, mutex->address, &value);
+	if (err < 0)
+		goto exit;
+
+	err = nfp_cpp_readl(mutex->cpp, mur, mutex->address + 4, &key);
+	if (err < 0)
+		goto exit;
+
+	if (key != mutex->key) {
+		err = NFP_ERRNO(EPERM);
+		goto exit;
+	}
+
+	if (value != MUTEX_LOCKED(interface)) {
+		err = NFP_ERRNO(EACCES);
+		goto exit;
+	}
+
+	err = nfp_cpp_writel(cpp, muw, mutex->address, MUTEX_UNLOCK(interface));
+	if (err < 0)
+		goto exit;
+
+	mutex->depth = 0;
+
+exit:
+	return err;
+}
+
+/*
+ * Attempt to lock a mutex handle, using the NFP MU Atomic Engine
+ *
+ * Valid lock states:
+ *
+ *      0x....0000      - Unlocked
+ *      0x....000f      - Locked
+ *
+ * @param mutex     NFP CPP Mutex handle
+ * @return      0 if the lock succeeded, -1 on failure (and errno set
+ *		appropriately).
+ */
+int
+nfp_cpp_mutex_trylock(struct nfp_cpp_mutex *mutex)
+{
+	uint32_t mur = NFP_CPP_ID(mutex->target, 3, 0);	/* atomic_read */
+	uint32_t muw = NFP_CPP_ID(mutex->target, 4, 0);	/* atomic_write */
+	uint32_t mus = NFP_CPP_ID(mutex->target, 5, 3);	/* test_set_imm */
+	uint32_t key, value, tmp;
+	struct nfp_cpp *cpp = mutex->cpp;
+	int err;
+
+	if (mutex->depth > 0) {
+		if (mutex->depth == MUTEX_DEPTH_MAX)
+			return NFP_ERRNO(E2BIG);
+
+		mutex->depth++;
+		return 0;
+	}
+
+	/* Verify that the lock marker is not damaged */
+	err = nfp_cpp_readl(cpp, mur, mutex->address + 4, &key);
+	if (err < 0)
+		goto exit;
+
+	if (key != mutex->key) {
+		err = NFP_ERRNO(EPERM);
+		goto exit;
+	}
+
+	/*
+	 * Compare against the unlocked state, and if true,
+	 * write the interface id into the top 16 bits, and
+	 * mark as locked.
+	 */
+	value = MUTEX_LOCKED(nfp_cpp_interface(cpp));
+
+	/*
+	 * We use test_set_imm here, as it implies a read
+	 * of the current state, and sets the bits in the
+	 * bytemask of the command to 1s. Since the mutex
+	 * is guaranteed to be 64-bit aligned, the bytemask
+	 * of this 32-bit command is ensured to be 8'b00001111,
+	 * which implies that the lower 4 bits will be set to
+	 * ones regardless of the initial state.
+	 *
+	 * Since this is a 'Readback' operation, with no Pull
+	 * data, we can treat this as a normal Push (read)
+	 * atomic, which returns the original value.
+	 */
+	err = nfp_cpp_readl(cpp, mus, mutex->address, &tmp);
+	if (err < 0)
+		goto exit;
+
+	/* Was it unlocked? */
+	if (MUTEX_IS_UNLOCKED(tmp)) {
+		/*
+		 * The read value can only be 0x....0000 in the unlocked state.
+		 * If there was another contending for this lock, then
+		 * the lock state would be 0x....000f
+		 *
+		 * Write our owner ID into the lock
+		 * While not strictly necessary, this helps with
+		 * debug and bookkeeping.
+		 */
+		err = nfp_cpp_writel(cpp, muw, mutex->address, value);
+		if (err < 0)
+			goto exit;
+
+		mutex->depth = 1;
+		goto exit;
+	}
+
+	/* Already locked by us? Success! */
+	if (tmp == value) {
+		mutex->depth = 1;
+		goto exit;
+	}
+
+	err = NFP_ERRNO(MUTEX_IS_LOCKED(tmp) ? EBUSY : EINVAL);
+
+exit:
+	return err;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_nffw.c b/drivers/net/nfp/nfpcore/nfp_nffw.c
new file mode 100644
index 0000000..b8e5331
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_nffw.c
@@ -0,0 +1,261 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 "nfp_cpp.h"
+#include "nfp_nffw.h"
+#include "nfp_mip.h"
+#include "nfp6000/nfp6000.h"
+#include "nfp_resource.h"
+
+/*
+ * flg_info_version = flags[0]<27:16>
+ * This is a small version counter intended only to detect if the current
+ * implementation can read the current struct. Struct changes should be very
+ * rare and as such a 12-bit counter should cover large spans of time. By the
+ * time it wraps around, we don't expect to have 4096 versions of this struct
+ * to be in use at the same time.
+ */
+static uint32_t
+nffw_res_info_version_get(const struct nfp_nffw_info_data *res)
+{
+	return (res->flags[0] >> 16) & 0xfff;
+}
+
+/* flg_init = flags[0]<0> */
+static uint32_t
+nffw_res_flg_init_get(const struct nfp_nffw_info_data *res)
+{
+	return (res->flags[0] >> 0) & 1;
+}
+
+/* loaded = loaded__mu_da__mip_off_hi<31:31> */
+static uint32_t
+nffw_fwinfo_loaded_get(const struct nffw_fwinfo *fi)
+{
+	return (fi->loaded__mu_da__mip_off_hi >> 31) & 1;
+}
+
+/* mip_cppid = mip_cppid */
+static uint32_t
+nffw_fwinfo_mip_cppid_get(const struct nffw_fwinfo *fi)
+{
+	return fi->mip_cppid;
+}
+
+/* loaded = loaded__mu_da__mip_off_hi<8:8> */
+static uint32_t
+nffw_fwinfo_mip_mu_da_get(const struct nffw_fwinfo *fi)
+{
+	return (fi->loaded__mu_da__mip_off_hi >> 8) & 1;
+}
+
+/* mip_offset = (loaded__mu_da__mip_off_hi<7:0> << 8) | mip_offset_lo */
+static uint64_t
+nffw_fwinfo_mip_offset_get(const struct nffw_fwinfo *fi)
+{
+	uint64_t mip_off_hi = fi->loaded__mu_da__mip_off_hi;
+
+	return (mip_off_hi & 0xFF) << 32 | fi->mip_offset_lo;
+}
+
+#define NFP_IMB_TGTADDRESSMODECFG_MODE_of(_x)		(((_x) >> 13) & 0x7)
+#define NFP_IMB_TGTADDRESSMODECFG_ADDRMODE		BIT(12)
+#define   NFP_IMB_TGTADDRESSMODECFG_ADDRMODE_32_BIT	0
+#define   NFP_IMB_TGTADDRESSMODECFG_ADDRMODE_40_BIT	BIT(12)
+
+static int
+nfp_mip_mu_locality_lsb(struct nfp_cpp *cpp)
+{
+	unsigned int mode, addr40;
+	uint32_t xpbaddr, imbcppat;
+	int err;
+
+	/* Hardcoded XPB IMB Base, island 0 */
+	xpbaddr = 0x000a0000 + NFP_CPP_TARGET_MU * 4;
+	err = nfp_xpb_readl(cpp, xpbaddr, &imbcppat);
+	if (err < 0)
+		return err;
+
+	mode = NFP_IMB_TGTADDRESSMODECFG_MODE_of(imbcppat);
+	addr40 = !!(imbcppat & NFP_IMB_TGTADDRESSMODECFG_ADDRMODE);
+
+	return nfp_cppat_mu_locality_lsb(mode, addr40);
+}
+
+static unsigned int
+nffw_res_fwinfos(struct nfp_nffw_info_data *fwinf, struct nffw_fwinfo **arr)
+{
+	/*
+	 * For the this code, version 0 is most likely to be version 1 in this
+	 * case. Since the kernel driver does not take responsibility for
+	 * initialising the nfp.nffw resource, any previous code (CA firmware or
+	 * userspace) that left the version 0 and did set the init flag is going
+	 * to be version 1.
+	 */
+	switch (nffw_res_info_version_get(fwinf)) {
+	case 0:
+	case 1:
+		*arr = &fwinf->info.v1.fwinfo[0];
+		return NFFW_FWINFO_CNT_V1;
+	case 2:
+		*arr = &fwinf->info.v2.fwinfo[0];
+		return NFFW_FWINFO_CNT_V2;
+	default:
+		*arr = NULL;
+		return 0;
+	}
+}
+
+/*
+ * nfp_nffw_info_open() - Acquire the lock on the NFFW table
+ * @cpp:	NFP CPP handle
+ *
+ * Return: 0, or -ERRNO
+ */
+struct nfp_nffw_info *
+nfp_nffw_info_open(struct nfp_cpp *cpp)
+{
+	struct nfp_nffw_info_data *fwinf;
+	struct nfp_nffw_info *state;
+	uint32_t info_ver;
+	int err;
+
+	state = malloc(sizeof(*state));
+	if (!state)
+		return NULL;
+
+	memset(state, 0, sizeof(*state));
+
+	state->res = nfp_resource_acquire(cpp, NFP_RESOURCE_NFP_NFFW);
+	if (!state->res)
+		goto err_free;
+
+	fwinf = &state->fwinf;
+
+	if (sizeof(*fwinf) > nfp_resource_size(state->res))
+		goto err_release;
+
+	err = nfp_cpp_read(cpp, nfp_resource_cpp_id(state->res),
+			   nfp_resource_address(state->res),
+			   fwinf, sizeof(*fwinf));
+	if (err < (int)sizeof(*fwinf))
+		goto err_release;
+
+	if (!nffw_res_flg_init_get(fwinf))
+		goto err_release;
+
+	info_ver = nffw_res_info_version_get(fwinf);
+	if (info_ver > NFFW_INFO_VERSION_CURRENT)
+		goto err_release;
+
+	state->cpp = cpp;
+	return state;
+
+err_release:
+	nfp_resource_release(state->res);
+err_free:
+	free(state);
+	return NULL;
+}
+
+/*
+ * nfp_nffw_info_release() - Release the lock on the NFFW table
+ * @state:	NFP FW info state
+ *
+ * Return: 0, or -ERRNO
+ */
+void
+nfp_nffw_info_close(struct nfp_nffw_info *state)
+{
+	nfp_resource_release(state->res);
+	free(state);
+}
+
+/*
+ * nfp_nffw_info_fwid_first() - Return the first firmware ID in the NFFW
+ * @state:	NFP FW info state
+ *
+ * Return: First NFFW firmware info, NULL on failure
+ */
+static struct nffw_fwinfo *
+nfp_nffw_info_fwid_first(struct nfp_nffw_info *state)
+{
+	struct nffw_fwinfo *fwinfo;
+	unsigned int cnt, i;
+
+	cnt = nffw_res_fwinfos(&state->fwinf, &fwinfo);
+	if (!cnt)
+		return NULL;
+
+	for (i = 0; i < cnt; i++)
+		if (nffw_fwinfo_loaded_get(&fwinfo[i]))
+			return &fwinfo[i];
+
+	return NULL;
+}
+
+/*
+ * nfp_nffw_info_mip_first() - Retrieve the location of the first FW's MIP
+ * @state:	NFP FW info state
+ * @cpp_id:	Pointer to the CPP ID of the MIP
+ * @off:	Pointer to the CPP Address of the MIP
+ *
+ * Return: 0, or -ERRNO
+ */
+int
+nfp_nffw_info_mip_first(struct nfp_nffw_info *state, uint32_t *cpp_id,
+			uint64_t *off)
+{
+	struct nffw_fwinfo *fwinfo;
+
+	fwinfo = nfp_nffw_info_fwid_first(state);
+	if (!fwinfo)
+		return -EINVAL;
+
+	*cpp_id = nffw_fwinfo_mip_cppid_get(fwinfo);
+	*off = nffw_fwinfo_mip_offset_get(fwinfo);
+
+	if (nffw_fwinfo_mip_mu_da_get(fwinfo)) {
+		int locality_off;
+
+		if (NFP_CPP_ID_TARGET_of(*cpp_id) != NFP_CPP_TARGET_MU)
+			return 0;
+
+		locality_off = nfp_mip_mu_locality_lsb(state->cpp);
+		if (locality_off < 0)
+			return locality_off;
+
+		*off &= ~(NFP_MU_ADDR_ACCESS_TYPE_MASK << locality_off);
+		*off |= NFP_MU_ADDR_ACCESS_TYPE_DIRECT << locality_off;
+	}
+
+	return 0;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_nffw.h b/drivers/net/nfp/nfpcore/nfp_nffw.h
new file mode 100644
index 0000000..926ec67
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_nffw.h
@@ -0,0 +1,112 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_NFFW_H__
+#define __NFP_NFFW_H__
+
+#include "nfp-common/nfp_platform.h"
+#include "nfp_cpp.h"
+
+/*
+ * Init-CSR owner IDs for firmware map to firmware IDs which start at 4.
+ * Lower IDs are reserved for target and loader IDs.
+ */
+#define NFFW_FWID_EXT   3	/* For active MEs that we didn't load. */
+#define NFFW_FWID_BASE  4
+
+#define NFFW_FWID_ALL   255
+
+/* Init-CSR owner IDs for firmware map to firmware IDs which start at 4.
+ * Lower IDs are reserved for target and loader IDs.
+ */
+#define NFFW_FWID_EXT   3 /* For active MEs that we didn't load. */
+#define NFFW_FWID_BASE  4
+
+#define NFFW_FWID_ALL   255
+
+/**
+ * NFFW_INFO_VERSION history:
+ * 0: This was never actually used (before versioning), but it refers to
+ *    the previous struct which had FWINFO_CNT = MEINFO_CNT = 120 that later
+ *    changed to 200.
+ * 1: First versioned struct, with
+ *     FWINFO_CNT = 120
+ *     MEINFO_CNT = 120
+ * 2:  FWINFO_CNT = 200
+ *     MEINFO_CNT = 200
+ */
+#define NFFW_INFO_VERSION_CURRENT 2
+
+/* Enough for all current chip families */
+#define NFFW_MEINFO_CNT_V1 120
+#define NFFW_FWINFO_CNT_V1 120
+#define NFFW_MEINFO_CNT_V2 200
+#define NFFW_FWINFO_CNT_V2 200
+
+struct nffw_meinfo {
+	uint32_t ctxmask__fwid__meid;
+};
+
+struct nffw_fwinfo {
+	uint32_t loaded__mu_da__mip_off_hi;
+	uint32_t mip_cppid; /* 0 means no MIP */
+	uint32_t mip_offset_lo;
+};
+
+struct nfp_nffw_info_v1 {
+	struct nffw_meinfo meinfo[NFFW_MEINFO_CNT_V1];
+	struct nffw_fwinfo fwinfo[NFFW_FWINFO_CNT_V1];
+};
+
+struct nfp_nffw_info_v2 {
+	struct nffw_meinfo meinfo[NFFW_MEINFO_CNT_V2];
+	struct nffw_fwinfo fwinfo[NFFW_FWINFO_CNT_V2];
+};
+
+struct nfp_nffw_info_data {
+	uint32_t flags[2];
+	union {
+		struct nfp_nffw_info_v1 v1;
+		struct nfp_nffw_info_v2 v2;
+	} info;
+};
+
+struct nfp_nffw_info {
+	struct nfp_cpp *cpp;
+	struct nfp_resource *res;
+
+	struct nfp_nffw_info_data fwinf;
+};
+
+struct nfp_nffw_info *nfp_nffw_info_open(struct nfp_cpp *cpp);
+void nfp_nffw_info_close(struct nfp_nffw_info *state);
+
+#endif
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp.c b/drivers/net/nfp/nfpcore/nfp_nsp.c
new file mode 100644
index 0000000..6254f9a
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_nsp.c
@@ -0,0 +1,453 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 NFP_SUBSYS "nfp_nsp"
+
+#include <stdio.h>
+#include <time.h>
+
+#include <rte_common.h>
+
+#include "nfp_cpp.h"
+#include "nfp_nsp.h"
+#include "nfp_resource.h"
+
+int
+nfp_nsp_config_modified(struct nfp_nsp *state)
+{
+	return state->modified;
+}
+
+void
+nfp_nsp_config_set_modified(struct nfp_nsp *state, int modified)
+{
+	state->modified = modified;
+}
+
+void *
+nfp_nsp_config_entries(struct nfp_nsp *state)
+{
+	return state->entries;
+}
+
+unsigned int
+nfp_nsp_config_idx(struct nfp_nsp *state)
+{
+	return state->idx;
+}
+
+void
+nfp_nsp_config_set_state(struct nfp_nsp *state, void *entries, unsigned int idx)
+{
+	state->entries = entries;
+	state->idx = idx;
+}
+
+void
+nfp_nsp_config_clear_state(struct nfp_nsp *state)
+{
+	state->entries = NULL;
+	state->idx = 0;
+}
+
+static void
+nfp_nsp_print_extended_error(uint32_t ret_val)
+{
+	int i;
+
+	if (!ret_val)
+		return;
+
+	for (i = 0; i < (int)ARRAY_SIZE(nsp_errors); i++)
+		if (ret_val == (uint32_t)nsp_errors[i].code)
+			printf("err msg: %s\n", nsp_errors[i].msg);
+}
+
+static int
+nfp_nsp_check(struct nfp_nsp *state)
+{
+	struct nfp_cpp *cpp = state->cpp;
+	uint64_t nsp_status, reg;
+	uint32_t nsp_cpp;
+	int err;
+
+	nsp_cpp = nfp_resource_cpp_id(state->res);
+	nsp_status = nfp_resource_address(state->res) + NSP_STATUS;
+
+	err = nfp_cpp_readq(cpp, nsp_cpp, nsp_status, &reg);
+	if (err < 0)
+		return err;
+
+	if (FIELD_GET(NSP_STATUS_MAGIC, reg) != NSP_MAGIC) {
+		printf("Cannot detect NFP Service Processor\n");
+		return -ENODEV;
+	}
+
+	state->ver.major = FIELD_GET(NSP_STATUS_MAJOR, reg);
+	state->ver.minor = FIELD_GET(NSP_STATUS_MINOR, reg);
+
+	if (state->ver.major != NSP_MAJOR || state->ver.minor < NSP_MINOR) {
+		printf("Unsupported ABI %hu.%hu\n", state->ver.major,
+						    state->ver.minor);
+		return -EINVAL;
+	}
+
+	if (reg & NSP_STATUS_BUSY) {
+		printf("Service processor busy!\n");
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+/*
+ * nfp_nsp_open() - Prepare for communication and lock the NSP resource.
+ * @cpp:	NFP CPP Handle
+ */
+struct nfp_nsp *
+nfp_nsp_open(struct nfp_cpp *cpp)
+{
+	struct nfp_resource *res;
+	struct nfp_nsp *state;
+	int err;
+
+	res = nfp_resource_acquire(cpp, NFP_RESOURCE_NSP);
+	if (!res)
+		return NULL;
+
+	state = malloc(sizeof(*state));
+	if (!state) {
+		nfp_resource_release(res);
+		return NULL;
+	}
+	memset(state, 0, sizeof(*state));
+	state->cpp = cpp;
+	state->res = res;
+
+	err = nfp_nsp_check(state);
+	if (err) {
+		nfp_nsp_close(state);
+		return NULL;
+	}
+
+	return state;
+}
+
+/*
+ * nfp_nsp_close() - Clean up and unlock the NSP resource.
+ * @state:	NFP SP state
+ */
+void
+nfp_nsp_close(struct nfp_nsp *state)
+{
+	nfp_resource_release(state->res);
+	free(state);
+}
+
+uint16_t
+nfp_nsp_get_abi_ver_major(struct nfp_nsp *state)
+{
+	return state->ver.major;
+}
+
+uint16_t
+nfp_nsp_get_abi_ver_minor(struct nfp_nsp *state)
+{
+	return state->ver.minor;
+}
+
+static int
+nfp_nsp_wait_reg(struct nfp_cpp *cpp, uint64_t *reg, uint32_t nsp_cpp,
+		 uint64_t addr, uint64_t mask, uint64_t val)
+{
+	struct timespec wait;
+	int count;
+	int err;
+
+	wait.tv_sec = 0;
+	wait.tv_nsec = 25000000;
+	count = 0;
+
+	for (;;) {
+		err = nfp_cpp_readq(cpp, nsp_cpp, addr, reg);
+		if (err < 0)
+			return err;
+
+		if ((*reg & mask) == val)
+			return 0;
+
+		nanosleep(&wait, 0);
+		if (count++ > 1000)
+			return -ETIMEDOUT;
+	}
+}
+
+/*
+ * nfp_nsp_command() - Execute a command on the NFP Service Processor
+ * @state:	NFP SP state
+ * @code:	NFP SP Command Code
+ * @option:	NFP SP Command Argument
+ * @buff_cpp:	NFP SP Buffer CPP Address info
+ * @buff_addr:	NFP SP Buffer Host address
+ *
+ * Return: 0 for success with no result
+ *
+ *	 positive value for NSP completion with a result code
+ *
+ *	-EAGAIN if the NSP is not yet present
+ *	-ENODEV if the NSP is not a supported model
+ *	-EBUSY if the NSP is stuck
+ *	-EINTR if interrupted while waiting for completion
+ *	-ETIMEDOUT if the NSP took longer than 30 seconds to complete
+ */
+static int
+nfp_nsp_command(struct nfp_nsp *state, uint16_t code, uint32_t option,
+		uint32_t buff_cpp, uint64_t buff_addr)
+{
+	uint64_t reg, ret_val, nsp_base, nsp_buffer, nsp_status, nsp_command;
+	struct nfp_cpp *cpp = state->cpp;
+	uint32_t nsp_cpp;
+	int err;
+
+	nsp_cpp = nfp_resource_cpp_id(state->res);
+	nsp_base = nfp_resource_address(state->res);
+	nsp_status = nsp_base + NSP_STATUS;
+	nsp_command = nsp_base + NSP_COMMAND;
+	nsp_buffer = nsp_base + NSP_BUFFER;
+
+	err = nfp_nsp_check(state);
+	if (err)
+		return err;
+
+	if (!FIELD_FIT(NSP_BUFFER_CPP, buff_cpp >> 8) ||
+	    !FIELD_FIT(NSP_BUFFER_ADDRESS, buff_addr)) {
+		printf("Host buffer out of reach %08x %" PRIx64 "\n",
+			buff_cpp, buff_addr);
+		return -EINVAL;
+	}
+
+	err = nfp_cpp_writeq(cpp, nsp_cpp, nsp_buffer,
+			     FIELD_PREP(NSP_BUFFER_CPP, buff_cpp >> 8) |
+			     FIELD_PREP(NSP_BUFFER_ADDRESS, buff_addr));
+	if (err < 0)
+		return err;
+
+	err = nfp_cpp_writeq(cpp, nsp_cpp, nsp_command,
+			     FIELD_PREP(NSP_COMMAND_OPTION, option) |
+			     FIELD_PREP(NSP_COMMAND_CODE, code) |
+			     FIELD_PREP(NSP_COMMAND_START, 1));
+	if (err < 0)
+		return err;
+
+	/* Wait for NSP_COMMAND_START to go to 0 */
+	err = nfp_nsp_wait_reg(cpp, &reg, nsp_cpp, nsp_command,
+			       NSP_COMMAND_START, 0);
+	if (err) {
+		printf("Error %d waiting for code 0x%04x to start\n",
+			err, code);
+		return err;
+	}
+
+	/* Wait for NSP_STATUS_BUSY to go to 0 */
+	err = nfp_nsp_wait_reg(cpp, &reg, nsp_cpp, nsp_status, NSP_STATUS_BUSY,
+			       0);
+	if (err) {
+		printf("Error %d waiting for code 0x%04x to complete\n",
+			err, code);
+		return err;
+	}
+
+	err = nfp_cpp_readq(cpp, nsp_cpp, nsp_command, &ret_val);
+	if (err < 0)
+		return err;
+	ret_val = FIELD_GET(NSP_COMMAND_OPTION, ret_val);
+
+	err = FIELD_GET(NSP_STATUS_RESULT, reg);
+	if (err) {
+		printf("Result (error) code set: %d (%d) command: %d\n",
+			 -err, (int)ret_val, code);
+		nfp_nsp_print_extended_error(ret_val);
+		return -err;
+	}
+
+	return ret_val;
+}
+
+#define SZ_1M 0x00100000
+
+static int
+nfp_nsp_command_buf(struct nfp_nsp *nsp, uint16_t code, uint32_t option,
+		    const void *in_buf, unsigned int in_size, void *out_buf,
+		    unsigned int out_size)
+{
+	struct nfp_cpp *cpp = nsp->cpp;
+	unsigned int max_size;
+	uint64_t reg, cpp_buf;
+	int ret, err;
+	uint32_t cpp_id;
+
+	if (nsp->ver.minor < 13) {
+		printf("NSP: Code 0x%04x with buffer not supported\n", code);
+		printf("\t(ABI %hu.%hu)\n", nsp->ver.major, nsp->ver.minor);
+		return -EOPNOTSUPP;
+	}
+
+	err = nfp_cpp_readq(cpp, nfp_resource_cpp_id(nsp->res),
+			    nfp_resource_address(nsp->res) +
+			    NSP_DFLT_BUFFER_CONFIG,
+			    &reg);
+	if (err < 0)
+		return err;
+
+	max_size = RTE_MAX(in_size, out_size);
+	if (FIELD_GET(NSP_DFLT_BUFFER_SIZE_MB, reg) * SZ_1M < max_size) {
+		printf("NSP: default buffer too small for command 0x%04x\n",
+		       code);
+		printf("\t(%llu < %u)\n",
+		       FIELD_GET(NSP_DFLT_BUFFER_SIZE_MB, reg) * SZ_1M,
+		       max_size);
+		return -EINVAL;
+	}
+
+	err = nfp_cpp_readq(cpp, nfp_resource_cpp_id(nsp->res),
+			    nfp_resource_address(nsp->res) +
+			    NSP_DFLT_BUFFER,
+			    &reg);
+	if (err < 0)
+		return err;
+
+	cpp_id = FIELD_GET(NSP_BUFFER_CPP, reg) << 8;
+	cpp_buf = FIELD_GET(NSP_BUFFER_ADDRESS, reg);
+
+	if (in_buf && in_size) {
+		err = nfp_cpp_write(cpp, cpp_id, cpp_buf, in_buf, in_size);
+		if (err < 0)
+			return err;
+	}
+	/* Zero out remaining part of the buffer */
+	if (out_buf && out_size && out_size > in_size) {
+		memset(out_buf, 0, out_size - in_size);
+		err = nfp_cpp_write(cpp, cpp_id, cpp_buf + in_size, out_buf,
+				    out_size - in_size);
+		if (err < 0)
+			return err;
+	}
+
+	ret = nfp_nsp_command(nsp, code, option, cpp_id, cpp_buf);
+	if (ret < 0)
+		return ret;
+
+	if (out_buf && out_size) {
+		err = nfp_cpp_read(cpp, cpp_id, cpp_buf, out_buf, out_size);
+		if (err < 0)
+			return err;
+	}
+
+	return ret;
+}
+
+int
+nfp_nsp_wait(struct nfp_nsp *state)
+{
+	struct timespec wait;
+	int count;
+	int err;
+
+	wait.tv_sec = 0;
+	wait.tv_nsec = 25000000;
+	count = 0;
+
+	for (;;) {
+		err = nfp_nsp_command(state, SPCODE_NOOP, 0, 0, 0);
+		if (err != -EAGAIN)
+			break;
+
+		nanosleep(&wait, 0);
+
+		if (count++ > 1000) {
+			err = -ETIMEDOUT;
+			break;
+		}
+	}
+	if (err)
+		printf("NSP failed to respond %d\n", err);
+
+	return err;
+}
+
+int
+nfp_nsp_device_soft_reset(struct nfp_nsp *state)
+{
+	return nfp_nsp_command(state, SPCODE_SOFT_RESET, 0, 0, 0);
+}
+
+int
+nfp_nsp_mac_reinit(struct nfp_nsp *state)
+{
+	return nfp_nsp_command(state, SPCODE_MAC_INIT, 0, 0, 0);
+}
+
+int
+nfp_nsp_load_fw(struct nfp_nsp *state, void *buf, unsigned int size)
+{
+	return nfp_nsp_command_buf(state, SPCODE_FW_LOAD, size, buf, size,
+				   NULL, 0);
+}
+
+int
+nfp_nsp_read_eth_table(struct nfp_nsp *state, void *buf, unsigned int size)
+{
+	return nfp_nsp_command_buf(state, SPCODE_ETH_RESCAN, size, NULL, 0,
+				   buf, size);
+}
+
+int
+nfp_nsp_write_eth_table(struct nfp_nsp *state, const void *buf,
+			unsigned int size)
+{
+	return nfp_nsp_command_buf(state, SPCODE_ETH_CONTROL, size, buf, size,
+				   NULL, 0);
+}
+
+int
+nfp_nsp_read_identify(struct nfp_nsp *state, void *buf, unsigned int size)
+{
+	return nfp_nsp_command_buf(state, SPCODE_NSP_IDENTIFY, size, NULL, 0,
+				   buf, size);
+}
+
+int
+nfp_nsp_read_sensors(struct nfp_nsp *state, unsigned int sensor_mask, void *buf,
+		     unsigned int size)
+{
+	return nfp_nsp_command_buf(state, SPCODE_NSP_SENSORS, sensor_mask, NULL,
+				   0, buf, size);
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp.h b/drivers/net/nfp/nfpcore/nfp_nsp.h
new file mode 100644
index 0000000..37c861a
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_nsp.h
@@ -0,0 +1,330 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 NSP_NSP_H
+#define NSP_NSP_H 1
+
+#include "nfp_cpp.h"
+#include "nfp_nsp.h"
+
+#define GENMASK_ULL(h, l) \
+	(((~0ULL) - (1ULL << (l)) + 1) & \
+	 (~0ULL >> (64 - 1 - (h))))
+
+#define __bf_shf(x) (__builtin_ffsll(x) - 1)
+
+#define FIELD_GET(_mask, _reg)	\
+	(__extension__ ({ \
+		typeof(_mask) _x = (_mask); \
+		(typeof(_x))(((_reg) & (_x)) >> __bf_shf(_x));	\
+	}))
+
+#define FIELD_FIT(_mask, _val)						\
+	(__extension__ ({ \
+		typeof(_mask) _x = (_mask); \
+		!((((typeof(_x))_val) << __bf_shf(_x)) & ~(_x)); \
+	}))
+
+#define FIELD_PREP(_mask, _val)						\
+	(__extension__ ({ \
+		typeof(_mask) _x = (_mask); \
+		((typeof(_x))(_val) << __bf_shf(_x)) & (_x);	\
+	}))
+
+/* Offsets relative to the CSR base */
+#define NSP_STATUS		0x00
+#define   NSP_STATUS_MAGIC	GENMASK_ULL(63, 48)
+#define   NSP_STATUS_MAJOR	GENMASK_ULL(47, 44)
+#define   NSP_STATUS_MINOR	GENMASK_ULL(43, 32)
+#define   NSP_STATUS_CODE	GENMASK_ULL(31, 16)
+#define   NSP_STATUS_RESULT	GENMASK_ULL(15, 8)
+#define   NSP_STATUS_BUSY	BIT_ULL(0)
+
+#define NSP_COMMAND		0x08
+#define   NSP_COMMAND_OPTION	GENMASK_ULL(63, 32)
+#define   NSP_COMMAND_CODE	GENMASK_ULL(31, 16)
+#define   NSP_COMMAND_START	BIT_ULL(0)
+
+/* CPP address to retrieve the data from */
+#define NSP_BUFFER		0x10
+#define   NSP_BUFFER_CPP	GENMASK_ULL(63, 40)
+#define   NSP_BUFFER_PCIE	GENMASK_ULL(39, 38)
+#define   NSP_BUFFER_ADDRESS	GENMASK_ULL(37, 0)
+
+#define NSP_DFLT_BUFFER		0x18
+
+#define NSP_DFLT_BUFFER_CONFIG	0x20
+#define   NSP_DFLT_BUFFER_SIZE_MB	GENMASK_ULL(7, 0)
+
+#define NSP_MAGIC		0xab10
+#define NSP_MAJOR		0
+#define NSP_MINOR		8
+
+#define NSP_CODE_MAJOR		GENMASK(15, 12)
+#define NSP_CODE_MINOR		GENMASK(11, 0)
+
+enum nfp_nsp_cmd {
+	SPCODE_NOOP		= 0, /* No operation */
+	SPCODE_SOFT_RESET	= 1, /* Soft reset the NFP */
+	SPCODE_FW_DEFAULT	= 2, /* Load default (UNDI) FW */
+	SPCODE_PHY_INIT		= 3, /* Initialize the PHY */
+	SPCODE_MAC_INIT		= 4, /* Initialize the MAC */
+	SPCODE_PHY_RXADAPT	= 5, /* Re-run PHY RX Adaptation */
+	SPCODE_FW_LOAD		= 6, /* Load fw from buffer, len in option */
+	SPCODE_ETH_RESCAN	= 7, /* Rescan ETHs, write ETH_TABLE to buf */
+	SPCODE_ETH_CONTROL	= 8, /* Update media config from buffer */
+	SPCODE_NSP_SENSORS	= 12, /* Read NSP sensor(s) */
+	SPCODE_NSP_IDENTIFY	= 13, /* Read NSP version */
+};
+
+static const struct {
+	int code;
+	const char *msg;
+} nsp_errors[] = {
+	{ 6010, "could not map to phy for port" },
+	{ 6011, "not an allowed rate/lanes for port" },
+	{ 6012, "not an allowed rate/lanes for port" },
+	{ 6013, "high/low error, change other port first" },
+	{ 6014, "config not found in flash" },
+};
+
+struct nfp_nsp {
+	struct nfp_cpp *cpp;
+	struct nfp_resource *res;
+	struct {
+		uint16_t major;
+		uint16_t minor;
+	} ver;
+
+	/* Eth table config state */
+	int modified;
+	unsigned int idx;
+	void *entries;
+};
+
+struct nfp_nsp *nfp_nsp_open(struct nfp_cpp *cpp);
+void nfp_nsp_close(struct nfp_nsp *state);
+uint16_t nfp_nsp_get_abi_ver_major(struct nfp_nsp *state);
+uint16_t nfp_nsp_get_abi_ver_minor(struct nfp_nsp *state);
+int nfp_nsp_wait(struct nfp_nsp *state);
+int nfp_nsp_device_soft_reset(struct nfp_nsp *state);
+int nfp_nsp_load_fw(struct nfp_nsp *state, void *buf, unsigned int size);
+int nfp_nsp_mac_reinit(struct nfp_nsp *state);
+int nfp_nsp_read_identify(struct nfp_nsp *state, void *buf, unsigned int size);
+int nfp_nsp_read_sensors(struct nfp_nsp *state, unsigned int sensor_mask,
+			 void *buf, unsigned int size);
+
+static inline int nfp_nsp_has_mac_reinit(struct nfp_nsp *state)
+{
+	return nfp_nsp_get_abi_ver_minor(state) > 20;
+}
+
+enum nfp_eth_interface {
+	NFP_INTERFACE_NONE	= 0,
+	NFP_INTERFACE_SFP	= 1,
+	NFP_INTERFACE_SFPP	= 10,
+	NFP_INTERFACE_SFP28	= 28,
+	NFP_INTERFACE_QSFP	= 40,
+	NFP_INTERFACE_CXP	= 100,
+	NFP_INTERFACE_QSFP28	= 112,
+};
+
+enum nfp_eth_media {
+	NFP_MEDIA_DAC_PASSIVE = 0,
+	NFP_MEDIA_DAC_ACTIVE,
+	NFP_MEDIA_FIBRE,
+};
+
+enum nfp_eth_aneg {
+	NFP_ANEG_AUTO = 0,
+	NFP_ANEG_SEARCH,
+	NFP_ANEG_25G_CONSORTIUM,
+	NFP_ANEG_25G_IEEE,
+	NFP_ANEG_DISABLED,
+};
+
+enum nfp_eth_fec {
+	NFP_FEC_AUTO_BIT = 0,
+	NFP_FEC_BASER_BIT,
+	NFP_FEC_REED_SOLOMON_BIT,
+	NFP_FEC_DISABLED_BIT,
+};
+
+#define NFP_FEC_AUTO		BIT(NFP_FEC_AUTO_BIT)
+#define NFP_FEC_BASER		BIT(NFP_FEC_BASER_BIT)
+#define NFP_FEC_REED_SOLOMON	BIT(NFP_FEC_REED_SOLOMON_BIT)
+#define NFP_FEC_DISABLED	BIT(NFP_FEC_DISABLED_BIT)
+
+#define ETH_ALEN	6
+
+/**
+ * struct nfp_eth_table - ETH table information
+ * @count:	number of table entries
+ * @max_index:	max of @index fields of all @ports
+ * @ports:	table of ports
+ *
+ * @eth_index:	port index according to legacy ethX numbering
+ * @index:	chip-wide first channel index
+ * @nbi:	NBI index
+ * @base:	first channel index (within NBI)
+ * @lanes:	number of channels
+ * @speed:	interface speed (in Mbps)
+ * @interface:	interface (module) plugged in
+ * @media:	media type of the @interface
+ * @fec:	forward error correction mode
+ * @aneg:	auto negotiation mode
+ * @mac_addr:	interface MAC address
+ * @label_port:	port id
+ * @label_subport:  id of interface within port (for split ports)
+ * @enabled:	is enabled?
+ * @tx_enabled:	is TX enabled?
+ * @rx_enabled:	is RX enabled?
+ * @override_changed: is media reconfig pending?
+ *
+ * @port_type:	one of %PORT_* defines for ethtool
+ * @port_lanes:	total number of lanes on the port (sum of lanes of all subports)
+ * @is_split:	is interface part of a split port
+ * @fec_modes_supported:	bitmap of FEC modes supported
+ */
+struct nfp_eth_table {
+	unsigned int count;
+	unsigned int max_index;
+	struct nfp_eth_table_port {
+		unsigned int eth_index;
+		unsigned int index;
+		unsigned int nbi;
+		unsigned int base;
+		unsigned int lanes;
+		unsigned int speed;
+
+		unsigned int interface;
+		enum nfp_eth_media media;
+
+		enum nfp_eth_fec fec;
+		enum nfp_eth_aneg aneg;
+
+		uint8_t mac_addr[ETH_ALEN];
+
+		uint8_t label_port;
+		uint8_t label_subport;
+
+		int enabled;
+		int tx_enabled;
+		int rx_enabled;
+
+		int override_changed;
+
+		/* Computed fields */
+		uint8_t port_type;
+
+		unsigned int port_lanes;
+
+		int is_split;
+
+		unsigned int fec_modes_supported;
+	} ports[0];
+};
+
+struct nfp_eth_table *nfp_eth_read_ports(struct nfp_cpp *cpp);
+
+int nfp_eth_set_mod_enable(struct nfp_cpp *cpp, unsigned int idx, int enable);
+int nfp_eth_set_configured(struct nfp_cpp *cpp, unsigned int idx,
+			   int configed);
+int
+nfp_eth_set_fec(struct nfp_cpp *cpp, unsigned int idx, enum nfp_eth_fec mode);
+
+int nfp_nsp_read_eth_table(struct nfp_nsp *state, void *buf, unsigned int size);
+int nfp_nsp_write_eth_table(struct nfp_nsp *state, const void *buf,
+			    unsigned int size);
+void nfp_nsp_config_set_state(struct nfp_nsp *state, void *entries,
+			      unsigned int idx);
+void nfp_nsp_config_clear_state(struct nfp_nsp *state);
+void nfp_nsp_config_set_modified(struct nfp_nsp *state, int modified);
+void *nfp_nsp_config_entries(struct nfp_nsp *state);
+int nfp_nsp_config_modified(struct nfp_nsp *state);
+unsigned int nfp_nsp_config_idx(struct nfp_nsp *state);
+
+static inline int nfp_eth_can_support_fec(struct nfp_eth_table_port *eth_port)
+{
+	return !!eth_port->fec_modes_supported;
+}
+
+static inline unsigned int
+nfp_eth_supported_fec_modes(struct nfp_eth_table_port *eth_port)
+{
+	return eth_port->fec_modes_supported;
+}
+
+struct nfp_nsp *nfp_eth_config_start(struct nfp_cpp *cpp, unsigned int idx);
+int nfp_eth_config_commit_end(struct nfp_nsp *nsp);
+void nfp_eth_config_cleanup_end(struct nfp_nsp *nsp);
+
+int __nfp_eth_set_aneg(struct nfp_nsp *nsp, enum nfp_eth_aneg mode);
+int __nfp_eth_set_speed(struct nfp_nsp *nsp, unsigned int speed);
+int __nfp_eth_set_split(struct nfp_nsp *nsp, unsigned int lanes);
+
+/**
+ * struct nfp_nsp_identify - NSP static information
+ * @version:      opaque version string
+ * @flags:        version flags
+ * @br_primary:   branch id of primary bootloader
+ * @br_secondary: branch id of secondary bootloader
+ * @br_nsp:       branch id of NSP
+ * @primary:      version of primarary bootloader
+ * @secondary:    version id of secondary bootloader
+ * @nsp:          version id of NSP
+ * @sensor_mask:  mask of present sensors available on NIC
+ */
+struct nfp_nsp_identify {
+	char version[40];
+	uint8_t flags;
+	uint8_t br_primary;
+	uint8_t br_secondary;
+	uint8_t br_nsp;
+	uint16_t primary;
+	uint16_t secondary;
+	uint16_t nsp;
+	uint64_t sensor_mask;
+};
+
+struct nfp_nsp_identify *__nfp_nsp_identify(struct nfp_nsp *nsp);
+
+enum nfp_nsp_sensor_id {
+	NFP_SENSOR_CHIP_TEMPERATURE,
+	NFP_SENSOR_ASSEMBLY_POWER,
+	NFP_SENSOR_ASSEMBLY_12V_POWER,
+	NFP_SENSOR_ASSEMBLY_3V3_POWER,
+};
+
+int nfp_hwmon_read_sensor(struct nfp_cpp *cpp, enum nfp_nsp_sensor_id id,
+			  long *val);
+
+#endif
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp_cmds.c b/drivers/net/nfp/nfpcore/nfp_nsp_cmds.c
new file mode 100644
index 0000000..d9d6013
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_nsp_cmds.c
@@ -0,0 +1,135 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <stdio.h>
+#include <rte_byteorder.h>
+#include "nfp_cpp.h"
+#include "nfp_nsp.h"
+#include "nfp_nffw.h"
+
+struct nsp_identify {
+	uint8_t version[40];
+	uint8_t flags;
+	uint8_t br_primary;
+	uint8_t br_secondary;
+	uint8_t br_nsp;
+	uint16_t primary;
+	uint16_t secondary;
+	uint16_t nsp;
+	uint8_t reserved[6];
+	uint64_t sensor_mask;
+};
+
+struct nfp_nsp_identify *
+__nfp_nsp_identify(struct nfp_nsp *nsp)
+{
+	struct nfp_nsp_identify *nspi = NULL;
+	struct nsp_identify *ni;
+	int ret;
+
+	if (nfp_nsp_get_abi_ver_minor(nsp) < 15)
+		return NULL;
+
+	ni = malloc(sizeof(*ni));
+	if (!ni)
+		return NULL;
+
+	memset(ni, 0, sizeof(*ni));
+	ret = nfp_nsp_read_identify(nsp, ni, sizeof(*ni));
+	if (ret < 0) {
+		printf("reading bsp version failed %d\n",
+			ret);
+		goto exit_free;
+	}
+
+	nspi = malloc(sizeof(*nspi));
+	if (!nspi)
+		goto exit_free;
+
+	memset(nspi, 0, sizeof(*nspi));
+	memcpy(nspi->version, ni->version, sizeof(nspi->version));
+	nspi->version[sizeof(nspi->version) - 1] = '\0';
+	nspi->flags = ni->flags;
+	nspi->br_primary = ni->br_primary;
+	nspi->br_secondary = ni->br_secondary;
+	nspi->br_nsp = ni->br_nsp;
+	nspi->primary = rte_le_to_cpu_16(ni->primary);
+	nspi->secondary = rte_le_to_cpu_16(ni->secondary);
+	nspi->nsp = rte_le_to_cpu_16(ni->nsp);
+	nspi->sensor_mask = rte_le_to_cpu_64(ni->sensor_mask);
+
+exit_free:
+	free(ni);
+	return nspi;
+}
+
+struct nfp_sensors {
+	uint32_t chip_temp;
+	uint32_t assembly_power;
+	uint32_t assembly_12v_power;
+	uint32_t assembly_3v3_power;
+};
+
+int
+nfp_hwmon_read_sensor(struct nfp_cpp *cpp, enum nfp_nsp_sensor_id id, long *val)
+{
+	struct nfp_sensors s;
+	struct nfp_nsp *nsp;
+	int ret;
+
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp)
+		return -EIO;
+
+	ret = nfp_nsp_read_sensors(nsp, BIT(id), &s, sizeof(s));
+	nfp_nsp_close(nsp);
+
+	if (ret < 0)
+		return ret;
+
+	switch (id) {
+	case NFP_SENSOR_CHIP_TEMPERATURE:
+		*val = rte_le_to_cpu_32(s.chip_temp);
+		break;
+	case NFP_SENSOR_ASSEMBLY_POWER:
+		*val = rte_le_to_cpu_32(s.assembly_power);
+		break;
+	case NFP_SENSOR_ASSEMBLY_12V_POWER:
+		*val = rte_le_to_cpu_32(s.assembly_12v_power);
+		break;
+	case NFP_SENSOR_ASSEMBLY_3V3_POWER:
+		*val = rte_le_to_cpu_32(s.assembly_3v3_power);
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_nsp_eth.c b/drivers/net/nfp/nfpcore/nfp_nsp_eth.c
new file mode 100644
index 0000000..409578c
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_nsp_eth.c
@@ -0,0 +1,691 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <stdio.h>
+#include <rte_common.h>
+#include <rte_byteorder.h>
+#include "nfp_cpp.h"
+#include "nfp_nsp.h"
+#include "nfp6000/nfp6000.h"
+
+#define GENMASK_ULL(h, l) \
+	(((~0ULL) - (1ULL << (l)) + 1) & \
+	 (~0ULL >> (64 - 1 - (h))))
+
+#define __bf_shf(x) (__builtin_ffsll(x) - 1)
+
+#define FIELD_GET(_mask, _reg)						\
+	(__extension__ ({ \
+		typeof(_mask) _x = (_mask); \
+		(typeof(_x))(((_reg) & (_x)) >> __bf_shf(_x));	\
+	}))
+
+#define FIELD_FIT(_mask, _val)						\
+	(__extension__ ({ \
+		typeof(_mask) _x = (_mask); \
+		!((((typeof(_x))_val) << __bf_shf(_x)) & ~(_x)); \
+	}))
+
+#define FIELD_PREP(_mask, _val)						\
+	(__extension__ ({ \
+		typeof(_mask) _x = (_mask); \
+		((typeof(_x))(_val) << __bf_shf(_x)) & (_x);	\
+	}))
+
+#define NSP_ETH_NBI_PORT_COUNT		24
+#define NSP_ETH_MAX_COUNT		(2 * NSP_ETH_NBI_PORT_COUNT)
+#define NSP_ETH_TABLE_SIZE		(NSP_ETH_MAX_COUNT *		\
+					 sizeof(union eth_table_entry))
+
+#define NSP_ETH_PORT_LANES		GENMASK_ULL(3, 0)
+#define NSP_ETH_PORT_INDEX		GENMASK_ULL(15, 8)
+#define NSP_ETH_PORT_LABEL		GENMASK_ULL(53, 48)
+#define NSP_ETH_PORT_PHYLABEL		GENMASK_ULL(59, 54)
+#define NSP_ETH_PORT_FEC_SUPP_BASER	BIT_ULL(60)
+#define NSP_ETH_PORT_FEC_SUPP_RS	BIT_ULL(61)
+
+#define NSP_ETH_PORT_LANES_MASK		rte_cpu_to_le_64(NSP_ETH_PORT_LANES)
+
+#define NSP_ETH_STATE_CONFIGURED	BIT_ULL(0)
+#define NSP_ETH_STATE_ENABLED		BIT_ULL(1)
+#define NSP_ETH_STATE_TX_ENABLED	BIT_ULL(2)
+#define NSP_ETH_STATE_RX_ENABLED	BIT_ULL(3)
+#define NSP_ETH_STATE_RATE		GENMASK_ULL(11, 8)
+#define NSP_ETH_STATE_INTERFACE		GENMASK_ULL(19, 12)
+#define NSP_ETH_STATE_MEDIA		GENMASK_ULL(21, 20)
+#define NSP_ETH_STATE_OVRD_CHNG		BIT_ULL(22)
+#define NSP_ETH_STATE_ANEG		GENMASK_ULL(25, 23)
+#define NSP_ETH_STATE_FEC		GENMASK_ULL(27, 26)
+
+#define NSP_ETH_CTRL_CONFIGURED		BIT_ULL(0)
+#define NSP_ETH_CTRL_ENABLED		BIT_ULL(1)
+#define NSP_ETH_CTRL_TX_ENABLED		BIT_ULL(2)
+#define NSP_ETH_CTRL_RX_ENABLED		BIT_ULL(3)
+#define NSP_ETH_CTRL_SET_RATE		BIT_ULL(4)
+#define NSP_ETH_CTRL_SET_LANES		BIT_ULL(5)
+#define NSP_ETH_CTRL_SET_ANEG		BIT_ULL(6)
+#define NSP_ETH_CTRL_SET_FEC		BIT_ULL(7)
+
+/* Which connector port. */
+#define PORT_TP			0x00
+#define PORT_AUI		0x01
+#define PORT_MII		0x02
+#define PORT_FIBRE		0x03
+#define PORT_BNC		0x04
+#define PORT_DA			0x05
+#define PORT_NONE		0xef
+#define PORT_OTHER		0xff
+
+#define SPEED_10		10
+#define SPEED_100		100
+#define SPEED_1000		1000
+#define SPEED_2500		2500
+#define SPEED_5000		5000
+#define SPEED_10000		10000
+#define SPEED_14000		14000
+#define SPEED_20000		20000
+#define SPEED_25000		25000
+#define SPEED_40000		40000
+#define SPEED_50000		50000
+#define SPEED_56000		56000
+#define SPEED_100000		100000
+
+enum nfp_eth_raw {
+	NSP_ETH_RAW_PORT = 0,
+	NSP_ETH_RAW_STATE,
+	NSP_ETH_RAW_MAC,
+	NSP_ETH_RAW_CONTROL,
+
+	NSP_ETH_NUM_RAW
+};
+
+enum nfp_eth_rate {
+	RATE_INVALID = 0,
+	RATE_10M,
+	RATE_100M,
+	RATE_1G,
+	RATE_10G,
+	RATE_25G,
+};
+
+union eth_table_entry {
+	struct {
+		uint64_t port;
+		uint64_t state;
+		uint8_t mac_addr[6];
+		uint8_t resv[2];
+		uint64_t control;
+	};
+	uint64_t raw[NSP_ETH_NUM_RAW];
+};
+
+static const struct {
+	enum nfp_eth_rate rate;
+	unsigned int speed;
+} nsp_eth_rate_tbl[] = {
+	{ RATE_INVALID,	0, },
+	{ RATE_10M,	SPEED_10, },
+	{ RATE_100M,	SPEED_100, },
+	{ RATE_1G,	SPEED_1000, },
+	{ RATE_10G,	SPEED_10000, },
+	{ RATE_25G,	SPEED_25000, },
+};
+
+static unsigned int
+nfp_eth_rate2speed(enum nfp_eth_rate rate)
+{
+	int i;
+
+	for (i = 0; i < (int)ARRAY_SIZE(nsp_eth_rate_tbl); i++)
+		if (nsp_eth_rate_tbl[i].rate == rate)
+			return nsp_eth_rate_tbl[i].speed;
+
+	return 0;
+}
+
+static unsigned int
+nfp_eth_speed2rate(unsigned int speed)
+{
+	int i;
+
+	for (i = 0; i < (int)ARRAY_SIZE(nsp_eth_rate_tbl); i++)
+		if (nsp_eth_rate_tbl[i].speed == speed)
+			return nsp_eth_rate_tbl[i].rate;
+
+	return RATE_INVALID;
+}
+
+static void
+nfp_eth_copy_mac_reverse(uint8_t *dst, const uint8_t *src)
+{
+	int i;
+
+	for (i = 0; i < (int)ETH_ALEN; i++)
+		dst[ETH_ALEN - i - 1] = src[i];
+}
+
+static void
+nfp_eth_port_translate(struct nfp_nsp *nsp, const union eth_table_entry *src,
+		       unsigned int index, struct nfp_eth_table_port *dst)
+{
+	unsigned int rate;
+	unsigned int fec;
+	uint64_t port, state;
+
+	port = rte_le_to_cpu_64(src->port);
+	state = rte_le_to_cpu_64(src->state);
+
+	dst->eth_index = FIELD_GET(NSP_ETH_PORT_INDEX, port);
+	dst->index = index;
+	dst->nbi = index / NSP_ETH_NBI_PORT_COUNT;
+	dst->base = index % NSP_ETH_NBI_PORT_COUNT;
+	dst->lanes = FIELD_GET(NSP_ETH_PORT_LANES, port);
+
+	dst->enabled = FIELD_GET(NSP_ETH_STATE_ENABLED, state);
+	dst->tx_enabled = FIELD_GET(NSP_ETH_STATE_TX_ENABLED, state);
+	dst->rx_enabled = FIELD_GET(NSP_ETH_STATE_RX_ENABLED, state);
+
+	rate = nfp_eth_rate2speed(FIELD_GET(NSP_ETH_STATE_RATE, state));
+	dst->speed = dst->lanes * rate;
+
+	dst->interface = FIELD_GET(NSP_ETH_STATE_INTERFACE, state);
+	dst->media = FIELD_GET(NSP_ETH_STATE_MEDIA, state);
+
+	nfp_eth_copy_mac_reverse(dst->mac_addr, src->mac_addr);
+
+	dst->label_port = FIELD_GET(NSP_ETH_PORT_PHYLABEL, port);
+	dst->label_subport = FIELD_GET(NSP_ETH_PORT_LABEL, port);
+
+	if (nfp_nsp_get_abi_ver_minor(nsp) < 17)
+		return;
+
+	dst->override_changed = FIELD_GET(NSP_ETH_STATE_OVRD_CHNG, state);
+	dst->aneg = FIELD_GET(NSP_ETH_STATE_ANEG, state);
+
+	if (nfp_nsp_get_abi_ver_minor(nsp) < 22)
+		return;
+
+	fec = FIELD_GET(NSP_ETH_PORT_FEC_SUPP_BASER, port);
+	dst->fec_modes_supported |= fec << NFP_FEC_BASER_BIT;
+	fec = FIELD_GET(NSP_ETH_PORT_FEC_SUPP_RS, port);
+	dst->fec_modes_supported |= fec << NFP_FEC_REED_SOLOMON_BIT;
+	if (dst->fec_modes_supported)
+		dst->fec_modes_supported |= NFP_FEC_AUTO | NFP_FEC_DISABLED;
+
+	dst->fec = 1 << FIELD_GET(NSP_ETH_STATE_FEC, state);
+}
+
+static void
+nfp_eth_calc_port_geometry(struct nfp_eth_table *table)
+{
+	unsigned int i, j;
+
+	for (i = 0; i < table->count; i++) {
+		table->max_index = RTE_MAX(table->max_index,
+					   table->ports[i].index);
+
+		for (j = 0; j < table->count; j++) {
+			if (table->ports[i].label_port !=
+			    table->ports[j].label_port)
+				continue;
+			table->ports[i].port_lanes += table->ports[j].lanes;
+
+			if (i == j)
+				continue;
+			if (table->ports[i].label_subport ==
+			    table->ports[j].label_subport)
+				printf("Port %d subport %d is a duplicate\n",
+					 table->ports[i].label_port,
+					 table->ports[i].label_subport);
+
+			table->ports[i].is_split = 1;
+		}
+	}
+}
+
+static void
+nfp_eth_calc_port_type(struct nfp_eth_table_port *entry)
+{
+	if (entry->interface == NFP_INTERFACE_NONE) {
+		entry->port_type = PORT_NONE;
+		return;
+	}
+
+	if (entry->media == NFP_MEDIA_FIBRE)
+		entry->port_type = PORT_FIBRE;
+	else
+		entry->port_type = PORT_DA;
+}
+
+static struct nfp_eth_table *
+__nfp_eth_read_ports(struct nfp_nsp *nsp)
+{
+	union eth_table_entry *entries;
+	struct nfp_eth_table *table;
+	uint32_t table_sz;
+	int i, j, ret, cnt = 0;
+
+	entries = malloc(NSP_ETH_TABLE_SIZE);
+	if (!entries)
+		return NULL;
+
+	memset(entries, 0, NSP_ETH_TABLE_SIZE);
+	ret = nfp_nsp_read_eth_table(nsp, entries, NSP_ETH_TABLE_SIZE);
+	if (ret < 0) {
+		printf("reading port table failed %d\n", ret);
+		goto err;
+	}
+
+	for (i = 0; i < NSP_ETH_MAX_COUNT; i++)
+		if (entries[i].port & NSP_ETH_PORT_LANES_MASK)
+			cnt++;
+
+	/* Some versions of flash will give us 0 instead of port count. For
+	 * those that give a port count, verify it against the value calculated
+	 * above.
+	 */
+	if (ret && ret != cnt) {
+		printf("table entry count (%d) unmatch entries present (%d)\n",
+		       ret, cnt);
+		goto err;
+	}
+
+	table_sz = sizeof(*table) + sizeof(struct nfp_eth_table_port) * cnt;
+	table = malloc(table_sz);
+	if (!table)
+		goto err;
+
+	memset(table, 0, table_sz);
+	table->count = cnt;
+	for (i = 0, j = 0; i < NSP_ETH_MAX_COUNT; i++)
+		if (entries[i].port & NSP_ETH_PORT_LANES_MASK)
+			nfp_eth_port_translate(nsp, &entries[i], i,
+					       &table->ports[j++]);
+
+	nfp_eth_calc_port_geometry(table);
+	for (i = 0; i < (int)table->count; i++)
+		nfp_eth_calc_port_type(&table->ports[i]);
+
+	free(entries);
+
+	return table;
+
+err:
+	free(entries);
+	return NULL;
+}
+
+/*
+ * nfp_eth_read_ports() - retrieve port information
+ * @cpp:	NFP CPP handle
+ *
+ * Read the port information from the device.  Returned structure should
+ * be freed with kfree() once no longer needed.
+ *
+ * Return: populated ETH table or NULL on error.
+ */
+struct nfp_eth_table *
+nfp_eth_read_ports(struct nfp_cpp *cpp)
+{
+	struct nfp_eth_table *ret;
+	struct nfp_nsp *nsp;
+
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp)
+		return NULL;
+
+	ret = __nfp_eth_read_ports(nsp);
+	nfp_nsp_close(nsp);
+
+	return ret;
+}
+
+struct nfp_nsp *
+nfp_eth_config_start(struct nfp_cpp *cpp, unsigned int idx)
+{
+	union eth_table_entry *entries;
+	struct nfp_nsp *nsp;
+	int ret;
+
+	entries = malloc(NSP_ETH_TABLE_SIZE);
+	if (!entries)
+		return NULL;
+
+	memset(entries, 0, NSP_ETH_TABLE_SIZE);
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp) {
+		free(entries);
+		return nsp;
+	}
+
+	ret = nfp_nsp_read_eth_table(nsp, entries, NSP_ETH_TABLE_SIZE);
+	if (ret < 0) {
+		printf("reading port table failed %d\n", ret);
+		goto err;
+	}
+
+	if (!(entries[idx].port & NSP_ETH_PORT_LANES_MASK)) {
+		printf("trying to set port state on disabled port %d\n", idx);
+		goto err;
+	}
+
+	nfp_nsp_config_set_state(nsp, entries, idx);
+	return nsp;
+
+err:
+	nfp_nsp_close(nsp);
+	free(entries);
+	return NULL;
+}
+
+void
+nfp_eth_config_cleanup_end(struct nfp_nsp *nsp)
+{
+	union eth_table_entry *entries = nfp_nsp_config_entries(nsp);
+
+	nfp_nsp_config_set_modified(nsp, 0);
+	nfp_nsp_config_clear_state(nsp);
+	nfp_nsp_close(nsp);
+	free(entries);
+}
+
+/*
+ * nfp_eth_config_commit_end() - perform recorded configuration changes
+ * @nsp:	NFP NSP handle returned from nfp_eth_config_start()
+ *
+ * Perform the configuration which was requested with __nfp_eth_set_*()
+ * helpers and recorded in @nsp state.  If device was already configured
+ * as requested or no __nfp_eth_set_*() operations were made no NSP command
+ * will be performed.
+ *
+ * Return:
+ * 0 - configuration successful;
+ * 1 - no changes were needed;
+ * -ERRNO - configuration failed.
+ */
+int
+nfp_eth_config_commit_end(struct nfp_nsp *nsp)
+{
+	union eth_table_entry *entries = nfp_nsp_config_entries(nsp);
+	int ret = 1;
+
+	if (nfp_nsp_config_modified(nsp)) {
+		ret = nfp_nsp_write_eth_table(nsp, entries, NSP_ETH_TABLE_SIZE);
+		ret = ret < 0 ? ret : 0;
+	}
+
+	nfp_eth_config_cleanup_end(nsp);
+
+	return ret;
+}
+
+/*
+ * nfp_eth_set_mod_enable() - set PHY module enable control bit
+ * @cpp:	NFP CPP handle
+ * @idx:	NFP chip-wide port index
+ * @enable:	Desired state
+ *
+ * Enable or disable PHY module (this usually means setting the TX lanes
+ * disable bits).
+ *
+ * Return:
+ * 0 - configuration successful;
+ * 1 - no changes were needed;
+ * -ERRNO - configuration failed.
+ */
+int
+nfp_eth_set_mod_enable(struct nfp_cpp *cpp, unsigned int idx, int enable)
+{
+	union eth_table_entry *entries;
+	struct nfp_nsp *nsp;
+	uint64_t reg;
+
+	nsp = nfp_eth_config_start(cpp, idx);
+	if (!nsp)
+		return -1;
+
+	entries = nfp_nsp_config_entries(nsp);
+
+	/* Check if we are already in requested state */
+	reg = rte_le_to_cpu_64(entries[idx].state);
+	if (enable != (int)FIELD_GET(NSP_ETH_CTRL_ENABLED, reg)) {
+		reg = rte_le_to_cpu_64(entries[idx].control);
+		reg &= ~NSP_ETH_CTRL_ENABLED;
+		reg |= FIELD_PREP(NSP_ETH_CTRL_ENABLED, enable);
+		entries[idx].control = rte_cpu_to_le_64(reg);
+
+		nfp_nsp_config_set_modified(nsp, 1);
+	}
+
+	return nfp_eth_config_commit_end(nsp);
+}
+
+/*
+ * nfp_eth_set_configured() - set PHY module configured control bit
+ * @cpp:	NFP CPP handle
+ * @idx:	NFP chip-wide port index
+ * @configed:	Desired state
+ *
+ * Set the ifup/ifdown state on the PHY.
+ *
+ * Return:
+ * 0 - configuration successful;
+ * 1 - no changes were needed;
+ * -ERRNO - configuration failed.
+ */
+int
+nfp_eth_set_configured(struct nfp_cpp *cpp, unsigned int idx, int configed)
+{
+	union eth_table_entry *entries;
+	struct nfp_nsp *nsp;
+	uint64_t reg;
+
+	nsp = nfp_eth_config_start(cpp, idx);
+	if (!nsp)
+		return -EIO;
+
+	/*
+	 * Older ABI versions did support this feature, however this has only
+	 * been reliable since ABI 20.
+	 */
+	if (nfp_nsp_get_abi_ver_minor(nsp) < 20) {
+		nfp_eth_config_cleanup_end(nsp);
+		return -EOPNOTSUPP;
+	}
+
+	entries = nfp_nsp_config_entries(nsp);
+
+	/* Check if we are already in requested state */
+	reg = rte_le_to_cpu_64(entries[idx].state);
+	if (configed != (int)FIELD_GET(NSP_ETH_STATE_CONFIGURED, reg)) {
+		reg = rte_le_to_cpu_64(entries[idx].control);
+		reg &= ~NSP_ETH_CTRL_CONFIGURED;
+		reg |= FIELD_PREP(NSP_ETH_CTRL_CONFIGURED, configed);
+		entries[idx].control = rte_cpu_to_le_64(reg);
+
+		nfp_nsp_config_set_modified(nsp, 1);
+	}
+
+	return nfp_eth_config_commit_end(nsp);
+}
+
+static int
+nfp_eth_set_bit_config(struct nfp_nsp *nsp, unsigned int raw_idx,
+		       const uint64_t mask, const unsigned int shift,
+		       unsigned int val, const uint64_t ctrl_bit)
+{
+	union eth_table_entry *entries = nfp_nsp_config_entries(nsp);
+	unsigned int idx = nfp_nsp_config_idx(nsp);
+	uint64_t reg;
+
+	/*
+	 * Note: set features were added in ABI 0.14 but the error
+	 *	 codes were initially not populated correctly.
+	 */
+	if (nfp_nsp_get_abi_ver_minor(nsp) < 17) {
+		printf("set operations not supported, please update flash\n");
+		return -EOPNOTSUPP;
+	}
+
+	/* Check if we are already in requested state */
+	reg = rte_le_to_cpu_64(entries[idx].raw[raw_idx]);
+	if (val == (reg & mask) >> shift)
+		return 0;
+
+	reg &= ~mask;
+	reg |= (val << shift) & mask;
+	entries[idx].raw[raw_idx] = rte_cpu_to_le_64(reg);
+
+	entries[idx].control |= rte_cpu_to_le_64(ctrl_bit);
+
+	nfp_nsp_config_set_modified(nsp, 1);
+
+	return 0;
+}
+
+#define NFP_ETH_SET_BIT_CONFIG(nsp, raw_idx, mask, val, ctrl_bit)	\
+	(__extension__ ({ \
+		typeof(mask) _x = (mask); \
+		nfp_eth_set_bit_config(nsp, raw_idx, _x, __bf_shf(_x), \
+				       val, ctrl_bit);			\
+	}))
+
+/*
+ * __nfp_eth_set_aneg() - set PHY autonegotiation control bit
+ * @nsp:	NFP NSP handle returned from nfp_eth_config_start()
+ * @mode:	Desired autonegotiation mode
+ *
+ * Allow/disallow PHY module to advertise/perform autonegotiation.
+ * Will write to hwinfo overrides in the flash (persistent config).
+ *
+ * Return: 0 or -ERRNO.
+ */
+int
+__nfp_eth_set_aneg(struct nfp_nsp *nsp, enum nfp_eth_aneg mode)
+{
+	return NFP_ETH_SET_BIT_CONFIG(nsp, NSP_ETH_RAW_STATE,
+				      NSP_ETH_STATE_ANEG, mode,
+				      NSP_ETH_CTRL_SET_ANEG);
+}
+
+/*
+ * __nfp_eth_set_fec() - set PHY forward error correction control bit
+ * @nsp:	NFP NSP handle returned from nfp_eth_config_start()
+ * @mode:	Desired fec mode
+ *
+ * Set the PHY module forward error correction mode.
+ * Will write to hwinfo overrides in the flash (persistent config).
+ *
+ * Return: 0 or -ERRNO.
+ */
+static int
+__nfp_eth_set_fec(struct nfp_nsp *nsp, enum nfp_eth_fec mode)
+{
+	return NFP_ETH_SET_BIT_CONFIG(nsp, NSP_ETH_RAW_STATE,
+				      NSP_ETH_STATE_FEC, mode,
+				      NSP_ETH_CTRL_SET_FEC);
+}
+
+/*
+ * nfp_eth_set_fec() - set PHY forward error correction control mode
+ * @cpp:	NFP CPP handle
+ * @idx:	NFP chip-wide port index
+ * @mode:	Desired fec mode
+ *
+ * Return:
+ * 0 - configuration successful;
+ * 1 - no changes were needed;
+ * -ERRNO - configuration failed.
+ */
+int
+nfp_eth_set_fec(struct nfp_cpp *cpp, unsigned int idx, enum nfp_eth_fec mode)
+{
+	struct nfp_nsp *nsp;
+	int err;
+
+	nsp = nfp_eth_config_start(cpp, idx);
+	if (!nsp)
+		return -EIO;
+
+	err = __nfp_eth_set_fec(nsp, mode);
+	if (err) {
+		nfp_eth_config_cleanup_end(nsp);
+		return err;
+	}
+
+	return nfp_eth_config_commit_end(nsp);
+}
+
+/*
+ * __nfp_eth_set_speed() - set interface speed/rate
+ * @nsp:	NFP NSP handle returned from nfp_eth_config_start()
+ * @speed:	Desired speed (per lane)
+ *
+ * Set lane speed.  Provided @speed value should be subport speed divided
+ * by number of lanes this subport is spanning (i.e. 10000 for 40G, 25000 for
+ * 50G, etc.)
+ * Will write to hwinfo overrides in the flash (persistent config).
+ *
+ * Return: 0 or -ERRNO.
+ */
+int
+__nfp_eth_set_speed(struct nfp_nsp *nsp, unsigned int speed)
+{
+	enum nfp_eth_rate rate;
+
+	rate = nfp_eth_speed2rate(speed);
+	if (rate == RATE_INVALID) {
+		printf("could not find matching lane rate for speed %u\n",
+			 speed);
+		return -EINVAL;
+	}
+
+	return NFP_ETH_SET_BIT_CONFIG(nsp, NSP_ETH_RAW_STATE,
+				      NSP_ETH_STATE_RATE, rate,
+				      NSP_ETH_CTRL_SET_RATE);
+}
+
+/*
+ * __nfp_eth_set_split() - set interface lane split
+ * @nsp:	NFP NSP handle returned from nfp_eth_config_start()
+ * @lanes:	Desired lanes per port
+ *
+ * Set number of lanes in the port.
+ * Will write to hwinfo overrides in the flash (persistent config).
+ *
+ * Return: 0 or -ERRNO.
+ */
+int
+__nfp_eth_set_split(struct nfp_nsp *nsp, unsigned int lanes)
+{
+	return NFP_ETH_SET_BIT_CONFIG(nsp, NSP_ETH_RAW_PORT, NSP_ETH_PORT_LANES,
+				      lanes, NSP_ETH_CTRL_SET_LANES);
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_resource.c b/drivers/net/nfp/nfpcore/nfp_resource.c
new file mode 100644
index 0000000..e4f1efe
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_resource.c
@@ -0,0 +1,291 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 <stdio.h>
+#include <time.h>
+#include <zlib.h>
+#include <endian.h>
+
+#include "nfp_cpp.h"
+#include "nfp6000/nfp6000.h"
+#include "nfp_resource.h"
+#include "nfp_crc.h"
+
+#define NFP_RESOURCE_TBL_TARGET		NFP_CPP_TARGET_MU
+#define NFP_RESOURCE_TBL_BASE		0x8100000000ULL
+
+/* NFP Resource Table self-identifier */
+#define NFP_RESOURCE_TBL_NAME		"nfp.res"
+#define NFP_RESOURCE_TBL_KEY		0x00000000 /* Special key for entry 0 */
+
+#define NFP_RESOURCE_ENTRY_NAME_SZ	8
+
+/*
+ * struct nfp_resource_entry - Resource table entry
+ * @owner:		NFP CPP Lock, interface owner
+ * @key:		NFP CPP Lock, posix_crc32(name, 8)
+ * @region:		Memory region descriptor
+ * @name:		ASCII, zero padded name
+ * @reserved
+ * @cpp_action:		CPP Action
+ * @cpp_token:		CPP Token
+ * @cpp_target:		CPP Target ID
+ * @page_offset:	256-byte page offset into target's CPP address
+ * @page_size:		size, in 256-byte pages
+ */
+struct nfp_resource_entry {
+	struct nfp_resource_entry_mutex {
+		uint32_t owner;
+		uint32_t key;
+	} mutex;
+	struct nfp_resource_entry_region {
+		uint8_t  name[NFP_RESOURCE_ENTRY_NAME_SZ];
+		uint8_t  reserved[5];
+		uint8_t  cpp_action;
+		uint8_t  cpp_token;
+		uint8_t  cpp_target;
+		uint32_t page_offset;
+		uint32_t page_size;
+	} region;
+};
+
+#define NFP_RESOURCE_TBL_SIZE		4096
+#define NFP_RESOURCE_TBL_ENTRIES	(int)(NFP_RESOURCE_TBL_SIZE /	\
+					 sizeof(struct nfp_resource_entry))
+
+struct nfp_resource {
+	char name[NFP_RESOURCE_ENTRY_NAME_SZ + 1];
+	uint32_t cpp_id;
+	uint64_t addr;
+	uint64_t size;
+	struct nfp_cpp_mutex *mutex;
+};
+
+static int
+nfp_cpp_resource_find(struct nfp_cpp *cpp, struct nfp_resource *res)
+{
+	char name_pad[NFP_RESOURCE_ENTRY_NAME_SZ] = {};
+	struct nfp_resource_entry entry;
+	uint32_t cpp_id, key;
+	int ret, i;
+
+	cpp_id = NFP_CPP_ID(NFP_RESOURCE_TBL_TARGET, 3, 0);  /* Atomic read */
+
+	memset(name_pad, 0, NFP_RESOURCE_ENTRY_NAME_SZ);
+	strncpy(name_pad, res->name, sizeof(name_pad));
+
+	/* Search for a matching entry */
+	if (!memcmp(name_pad, NFP_RESOURCE_TBL_NAME "\0\0\0\0\0\0\0\0", 8)) {
+		printf("Grabbing device lock not supported\n");
+		return -EOPNOTSUPP;
+	}
+	key = nfp_crc32_posix(name_pad, sizeof(name_pad));
+
+	for (i = 0; i < NFP_RESOURCE_TBL_ENTRIES; i++) {
+		uint64_t addr = NFP_RESOURCE_TBL_BASE +
+			sizeof(struct nfp_resource_entry) * i;
+
+		ret = nfp_cpp_read(cpp, cpp_id, addr, &entry, sizeof(entry));
+		if (ret != sizeof(entry))
+			return -EIO;
+
+		if (entry.mutex.key != key)
+			continue;
+
+		/* Found key! */
+		res->mutex =
+			nfp_cpp_mutex_alloc(cpp,
+					    NFP_RESOURCE_TBL_TARGET, addr, key);
+		res->cpp_id = NFP_CPP_ID(entry.region.cpp_target,
+					 entry.region.cpp_action,
+					 entry.region.cpp_token);
+		res->addr = ((uint64_t)entry.region.page_offset) << 8;
+		res->size = (uint64_t)entry.region.page_size << 8;
+		return 0;
+	}
+
+	return -ENOENT;
+}
+
+static int
+nfp_resource_try_acquire(struct nfp_cpp *cpp, struct nfp_resource *res,
+			 struct nfp_cpp_mutex *dev_mutex)
+{
+	int err;
+
+	if (nfp_cpp_mutex_lock(dev_mutex))
+		return -EINVAL;
+
+	err = nfp_cpp_resource_find(cpp, res);
+	if (err)
+		goto err_unlock_dev;
+
+	err = nfp_cpp_mutex_trylock(res->mutex);
+	if (err)
+		goto err_res_mutex_free;
+
+	nfp_cpp_mutex_unlock(dev_mutex);
+
+	return 0;
+
+err_res_mutex_free:
+	nfp_cpp_mutex_free(res->mutex);
+err_unlock_dev:
+	nfp_cpp_mutex_unlock(dev_mutex);
+
+	return err;
+}
+
+/*
+ * nfp_resource_acquire() - Acquire a resource handle
+ * @cpp:	NFP CPP handle
+ * @name:	Name of the resource
+ *
+ * NOTE: This function locks the acquired resource
+ *
+ * Return: NFP Resource handle, or ERR_PTR()
+ */
+struct nfp_resource *
+nfp_resource_acquire(struct nfp_cpp *cpp, const char *name)
+{
+	struct nfp_cpp_mutex *dev_mutex;
+	struct nfp_resource *res;
+	int err;
+	struct timespec wait;
+	int count;
+
+	res = malloc(sizeof(*res));
+	if (!res)
+		return NULL;
+
+	memset(res, 0, sizeof(*res));
+
+	strncpy(res->name, name, NFP_RESOURCE_ENTRY_NAME_SZ);
+
+	dev_mutex = nfp_cpp_mutex_alloc(cpp, NFP_RESOURCE_TBL_TARGET,
+					NFP_RESOURCE_TBL_BASE,
+					NFP_RESOURCE_TBL_KEY);
+	if (!dev_mutex) {
+		free(res);
+		return NULL;
+	}
+
+	wait.tv_sec = 0;
+	wait.tv_nsec = 1000000;
+	count = 0;
+
+	for (;;) {
+		err = nfp_resource_try_acquire(cpp, res, dev_mutex);
+		if (!err)
+			break;
+		if (err != -EBUSY)
+			goto err_free;
+
+		if (count++ > 1000) {
+			printf("Error: resource %s timed out\n", name);
+			err = -EBUSY;
+			goto err_free;
+		}
+
+		nanosleep(&wait, NULL);
+	}
+
+	nfp_cpp_mutex_free(dev_mutex);
+
+	return res;
+
+err_free:
+	nfp_cpp_mutex_free(dev_mutex);
+	free(res);
+	return NULL;
+}
+
+/*
+ * nfp_resource_release() - Release a NFP Resource handle
+ * @res:	NFP Resource handle
+ *
+ * NOTE: This function implictly unlocks the resource handle
+ */
+void
+nfp_resource_release(struct nfp_resource *res)
+{
+	nfp_cpp_mutex_unlock(res->mutex);
+	nfp_cpp_mutex_free(res->mutex);
+	free(res);
+}
+
+/*
+ * nfp_resource_cpp_id() - Return the cpp_id of a resource handle
+ * @res:        NFP Resource handle
+ *
+ * Return: NFP CPP ID
+ */
+uint32_t
+nfp_resource_cpp_id(const struct nfp_resource *res)
+{
+	return res->cpp_id;
+}
+
+/*
+ * nfp_resource_name() - Return the name of a resource handle
+ * @res:        NFP Resource handle
+ *
+ * Return: const char pointer to the name of the resource
+ */
+const char
+*nfp_resource_name(const struct nfp_resource *res)
+{
+	return res->name;
+}
+
+/*
+ * nfp_resource_address() - Return the address of a resource handle
+ * @res:        NFP Resource handle
+ *
+ * Return: Address of the resource
+ */
+uint64_t
+nfp_resource_address(const struct nfp_resource *res)
+{
+	return res->addr;
+}
+
+/*
+ * nfp_resource_size() - Return the size in bytes of a resource handle
+ * @res:        NFP Resource handle
+ *
+ * Return: Size of the resource in bytes
+ */
+uint64_t
+nfp_resource_size(const struct nfp_resource *res)
+{
+	return res->size;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_resource.h b/drivers/net/nfp/nfpcore/nfp_resource.h
new file mode 100644
index 0000000..2f71d31
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_resource.h
@@ -0,0 +1,78 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 NFP_RESOURCE_H
+#define NFP_RESOURCE_H
+
+#include "nfp_cpp.h"
+
+#define NFP_RESOURCE_NFP_NFFW           "nfp.nffw"
+#define NFP_RESOURCE_NFP_HWINFO         "nfp.info"
+#define NFP_RESOURCE_NSP		"nfp.sp"
+
+/**
+ * Opaque handle to a NFP Resource
+ */
+struct nfp_resource;
+
+struct nfp_resource *nfp_resource_acquire(struct nfp_cpp *cpp,
+					  const char *name);
+
+/**
+ * Release a NFP Resource, and free the handle
+ * @param[in]   res     NFP Resource handle
+ */
+void nfp_resource_release(struct nfp_resource *res);
+
+/**
+ * Return the CPP ID of a NFP Resource
+ * @param[in]   res     NFP Resource handle
+ * @return      CPP ID of the NFP Resource
+ */
+uint32_t nfp_resource_cpp_id(const struct nfp_resource *res);
+
+/**
+ * Return the name of a NFP Resource
+ * @param[in]   res     NFP Resource handle
+ * @return      Name of the NFP Resource
+ */
+const char *nfp_resource_name(const struct nfp_resource *res);
+
+/**
+ * Return the target address of a NFP Resource
+ * @param[in]   res     NFP Resource handle
+ * @return      Address of the NFP Resource
+ */
+uint64_t nfp_resource_address(const struct nfp_resource *res);
+
+uint64_t nfp_resource_size(const struct nfp_resource *res);
+
+#endif /* NFP_RESOURCE_H */
diff --git a/drivers/net/nfp/nfpcore/nfp_rtsym.c b/drivers/net/nfp/nfpcore/nfp_rtsym.c
new file mode 100644
index 0000000..1622395
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_rtsym.c
@@ -0,0 +1,353 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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.
+ */
+
+/*
+ * nfp_rtsym.c
+ * Interface for accessing run-time symbol table
+ */
+
+#include <stdio.h>
+#include <rte_byteorder.h>
+#include "nfp_cpp.h"
+#include "nfp_mip.h"
+#include "nfp_rtsym.h"
+#include "nfp6000/nfp6000.h"
+
+/* These need to match the linker */
+#define SYM_TGT_LMEM		0
+#define SYM_TGT_EMU_CACHE	0x17
+
+struct nfp_rtsym_entry {
+	uint8_t	type;
+	uint8_t	target;
+	uint8_t	island;
+	uint8_t	addr_hi;
+	uint32_t addr_lo;
+	uint16_t name;
+	uint8_t	menum;
+	uint8_t	size_hi;
+	uint32_t size_lo;
+};
+
+struct nfp_rtsym_table {
+	struct nfp_cpp *cpp;
+	int num;
+	char *strtab;
+	struct nfp_rtsym symtab[];
+};
+
+static int
+nfp_meid(uint8_t island_id, uint8_t menum)
+{
+	return (island_id & 0x3F) == island_id && menum < 12 ?
+		(island_id << 4) | (menum + 4) : -1;
+}
+
+static void
+nfp_rtsym_sw_entry_init(struct nfp_rtsym_table *cache, uint32_t strtab_size,
+			struct nfp_rtsym *sw, struct nfp_rtsym_entry *fw)
+{
+	sw->type = fw->type;
+	sw->name = cache->strtab + rte_le_to_cpu_16(fw->name) % strtab_size;
+	sw->addr = ((uint64_t)fw->addr_hi << 32) |
+		   rte_le_to_cpu_32(fw->addr_lo);
+	sw->size = ((uint64_t)fw->size_hi << 32) |
+		   rte_le_to_cpu_32(fw->size_lo);
+
+#ifdef DEBUG
+	printf("rtsym_entry_init\n");
+	printf("\tname=%s, addr=%" PRIx64 ", size=%" PRIu64 ",target=%d\n",
+		sw->name, sw->addr, sw->size, sw->target);
+#endif
+	switch (fw->target) {
+	case SYM_TGT_LMEM:
+		sw->target = NFP_RTSYM_TARGET_LMEM;
+		break;
+	case SYM_TGT_EMU_CACHE:
+		sw->target = NFP_RTSYM_TARGET_EMU_CACHE;
+		break;
+	default:
+		sw->target = fw->target;
+		break;
+	}
+
+	if (fw->menum != 0xff)
+		sw->domain = nfp_meid(fw->island, fw->menum);
+	else if (fw->island != 0xff)
+		sw->domain = fw->island;
+	else
+		sw->domain = -1;
+}
+
+struct nfp_rtsym_table *
+nfp_rtsym_table_read(struct nfp_cpp *cpp)
+{
+	struct nfp_rtsym_table *rtbl;
+	struct nfp_mip *mip;
+
+	mip = nfp_mip_open(cpp);
+	rtbl = __nfp_rtsym_table_read(cpp, mip);
+	nfp_mip_close(mip);
+
+	return rtbl;
+}
+
+/*
+ * This looks more complex than it should be. But we need to get the type for
+ * the ~ right in round_down (it needs to be as wide as the result!), and we
+ * want to evaluate the macro arguments just once each.
+ */
+#define __round_mask(x, y) ((__typeof__(x))((y) - 1))
+
+#define round_up(x, y) \
+	(__extension__ ({ \
+		typeof(x) _x = (x); \
+		((((_x) - 1) | __round_mask(_x, y)) + 1); \
+	}))
+
+#define round_down(x, y) \
+	(__extension__ ({ \
+		typeof(x) _x = (x); \
+		((_x) & ~__round_mask(_x, y)); \
+	}))
+
+struct nfp_rtsym_table *
+__nfp_rtsym_table_read(struct nfp_cpp *cpp, const struct nfp_mip *mip)
+{
+	uint32_t strtab_addr, symtab_addr, strtab_size, symtab_size;
+	struct nfp_rtsym_entry *rtsymtab;
+	struct nfp_rtsym_table *cache;
+	const uint32_t dram =
+		NFP_CPP_ID(NFP_CPP_TARGET_MU, NFP_CPP_ACTION_RW, 0) |
+		NFP_ISL_EMEM0;
+	int err, n, size;
+
+	if (!mip)
+		return NULL;
+
+	nfp_mip_strtab(mip, &strtab_addr, &strtab_size);
+	nfp_mip_symtab(mip, &symtab_addr, &symtab_size);
+
+	if (!symtab_size || !strtab_size || symtab_size % sizeof(*rtsymtab))
+		return NULL;
+
+	/* Align to 64 bits */
+	symtab_size = round_up(symtab_size, 8);
+	strtab_size = round_up(strtab_size, 8);
+
+	rtsymtab = malloc(symtab_size);
+	if (!rtsymtab)
+		return NULL;
+
+	size = sizeof(*cache);
+	size += symtab_size / sizeof(*rtsymtab) * sizeof(struct nfp_rtsym);
+	size +=	strtab_size + 1;
+	cache = malloc(size);
+	if (!cache)
+		goto exit_free_rtsym_raw;
+
+	cache->cpp = cpp;
+	cache->num = symtab_size / sizeof(*rtsymtab);
+	cache->strtab = (void *)&cache->symtab[cache->num];
+
+	err = nfp_cpp_read(cpp, dram, symtab_addr, rtsymtab, symtab_size);
+	if (err != (int)symtab_size)
+		goto exit_free_cache;
+
+	err = nfp_cpp_read(cpp, dram, strtab_addr, cache->strtab, strtab_size);
+	if (err != (int)strtab_size)
+		goto exit_free_cache;
+	cache->strtab[strtab_size] = '\0';
+
+	for (n = 0; n < cache->num; n++)
+		nfp_rtsym_sw_entry_init(cache, strtab_size,
+					&cache->symtab[n], &rtsymtab[n]);
+
+	free(rtsymtab);
+
+	return cache;
+
+exit_free_cache:
+	free(cache);
+exit_free_rtsym_raw:
+	free(rtsymtab);
+	return NULL;
+}
+
+/*
+ * nfp_rtsym_count() - Get the number of RTSYM descriptors
+ * @rtbl:	NFP RTsym table
+ *
+ * Return: Number of RTSYM descriptors
+ */
+int
+nfp_rtsym_count(struct nfp_rtsym_table *rtbl)
+{
+	if (!rtbl)
+		return -EINVAL;
+
+	return rtbl->num;
+}
+
+/*
+ * nfp_rtsym_get() - Get the Nth RTSYM descriptor
+ * @rtbl:	NFP RTsym table
+ * @idx:	Index (0-based) of the RTSYM descriptor
+ *
+ * Return: const pointer to a struct nfp_rtsym descriptor, or NULL
+ */
+const struct nfp_rtsym *
+nfp_rtsym_get(struct nfp_rtsym_table *rtbl, int idx)
+{
+	if (!rtbl)
+		return NULL;
+
+	if (idx >= rtbl->num)
+		return NULL;
+
+	return &rtbl->symtab[idx];
+}
+
+/*
+ * nfp_rtsym_lookup() - Return the RTSYM descriptor for a symbol name
+ * @rtbl:	NFP RTsym table
+ * @name:	Symbol name
+ *
+ * Return: const pointer to a struct nfp_rtsym descriptor, or NULL
+ */
+const struct nfp_rtsym *
+nfp_rtsym_lookup(struct nfp_rtsym_table *rtbl, const char *name)
+{
+	int n;
+
+	if (!rtbl)
+		return NULL;
+
+	for (n = 0; n < rtbl->num; n++)
+		if (strcmp(name, rtbl->symtab[n].name) == 0)
+			return &rtbl->symtab[n];
+
+	return NULL;
+}
+
+/*
+ * nfp_rtsym_read_le() - Read a simple unsigned scalar value from symbol
+ * @rtbl:	NFP RTsym table
+ * @name:	Symbol name
+ * @error:	Poniter to error code (optional)
+ *
+ * Lookup a symbol, map, read it and return it's value. Value of the symbol
+ * will be interpreted as a simple little-endian unsigned value. Symbol can
+ * be 4 or 8 bytes in size.
+ *
+ * Return: value read, on error sets the error and returns ~0ULL.
+ */
+uint64_t
+nfp_rtsym_read_le(struct nfp_rtsym_table *rtbl, const char *name, int *error)
+{
+	const struct nfp_rtsym *sym;
+	uint32_t val32, id;
+	uint64_t val;
+	int err;
+
+	sym = nfp_rtsym_lookup(rtbl, name);
+	if (!sym) {
+		err = -ENOENT;
+		goto exit;
+	}
+
+	id = NFP_CPP_ISLAND_ID(sym->target, NFP_CPP_ACTION_RW, 0, sym->domain);
+
+#ifdef DEBUG
+	printf("Reading symbol %s with size %" PRIu64 " at %" PRIx64 "\n",
+		name, sym->size, sym->addr);
+#endif
+	switch (sym->size) {
+	case 4:
+		err = nfp_cpp_readl(rtbl->cpp, id, sym->addr, &val32);
+		val = val32;
+		break;
+	case 8:
+		err = nfp_cpp_readq(rtbl->cpp, id, sym->addr, &val);
+		break;
+	default:
+		printf("rtsym '%s' unsupported size: %" PRId64 "\n",
+			name, sym->size);
+		err = -EINVAL;
+		break;
+	}
+
+	if (err)
+		err = -EIO;
+exit:
+	if (error)
+		*error = err;
+
+	if (err)
+		return ~0ULL;
+
+	return val;
+}
+
+uint8_t *
+nfp_rtsym_map(struct nfp_rtsym_table *rtbl, const char *name,
+	      unsigned int min_size, struct nfp_cpp_area **area)
+{
+	const struct nfp_rtsym *sym;
+	uint8_t *mem;
+
+#ifdef DEBUG
+	printf("mapping symbol %s\n", name);
+#endif
+	sym = nfp_rtsym_lookup(rtbl, name);
+	if (!sym) {
+		printf("symbol lookup fails for %s\n", name);
+		return NULL;
+	}
+
+	if (sym->size < min_size) {
+		printf("Symbol %s too small (%" PRIu64 " < %u)\n", name,
+			sym->size, min_size);
+		return NULL;
+	}
+
+	mem = nfp_cpp_map_area(rtbl->cpp, sym->domain, sym->target, sym->addr,
+			       sym->size, area);
+	if (!mem) {
+		printf("Failed to map symbol %s\n", name);
+		return NULL;
+	}
+#ifdef DEBUG
+	printf("symbol %s with address %p\n", name, mem);
+#endif
+
+	return mem;
+}
diff --git a/drivers/net/nfp/nfpcore/nfp_rtsym.h b/drivers/net/nfp/nfpcore/nfp_rtsym.h
new file mode 100644
index 0000000..7941ecc
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_rtsym.h
@@ -0,0 +1,87 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 __NFP_RTSYM_H__
+#define __NFP_RTSYM_H__
+
+#define NFP_RTSYM_TYPE_NONE             0
+#define NFP_RTSYM_TYPE_OBJECT           1
+#define NFP_RTSYM_TYPE_FUNCTION         2
+#define NFP_RTSYM_TYPE_ABS              3
+
+#define NFP_RTSYM_TARGET_NONE           0
+#define NFP_RTSYM_TARGET_LMEM           -1
+#define NFP_RTSYM_TARGET_EMU_CACHE      -7
+
+/*
+ * Structure describing a run-time NFP symbol.
+ *
+ * The memory target of the symbol is generally the CPP target number and can be
+ * used directly by the nfp_cpp API calls.  However, in some cases (i.e., for
+ * local memory or control store) the target is encoded using a negative number.
+ *
+ * When the target type can not be used to fully describe the location of a
+ * symbol the domain field is used to further specify the location (i.e., the
+ * specific ME or island number).
+ *
+ * For ME target resources, 'domain' is an MEID.
+ * For Island target resources, 'domain' is an island ID, with the one exception
+ * of "sram" symbols for backward compatibility, which are viewed as global.
+ */
+struct nfp_rtsym {
+	const char *name;
+	uint64_t addr;
+	uint64_t size;
+	int type;
+	int target;
+	int domain;
+};
+
+struct nfp_rtsym_table;
+
+struct nfp_rtsym_table *nfp_rtsym_table_read(struct nfp_cpp *cpp);
+
+struct nfp_rtsym_table *
+__nfp_rtsym_table_read(struct nfp_cpp *cpp, const struct nfp_mip *mip);
+
+int nfp_rtsym_count(struct nfp_rtsym_table *rtbl);
+
+const struct nfp_rtsym *nfp_rtsym_get(struct nfp_rtsym_table *rtbl, int idx);
+
+const struct nfp_rtsym *
+nfp_rtsym_lookup(struct nfp_rtsym_table *rtbl, const char *name);
+
+uint64_t nfp_rtsym_read_le(struct nfp_rtsym_table *rtbl, const char *name,
+			   int *error);
+uint8_t *
+nfp_rtsym_map(struct nfp_rtsym_table *rtbl, const char *name,
+	      unsigned int min_size, struct nfp_cpp_area **area);
+#endif
diff --git a/drivers/net/nfp/nfpcore/nfp_target.h b/drivers/net/nfp/nfpcore/nfp_target.h
new file mode 100644
index 0000000..3b8c1db
--- /dev/null
+++ b/drivers/net/nfp/nfpcore/nfp_target.h
@@ -0,0 +1,605 @@
+/*
+ * Copyright (c) 2018 Netronome Systems, Inc.
+ * 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 copyright holder 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 HOLDER 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 NFP_TARGET_H
+#define NFP_TARGET_H
+
+#include "nfp-common/nfp_resid.h"
+#include "nfp-common/nfp_cppat.h"
+#include "nfp-common/nfp_platform.h"
+#include "nfp_cpp.h"
+
+#define P32 1
+#define P64 2
+
+#define PUSHPULL(_pull, _push) (((_pull) << 4) | ((_push) << 0))
+
+#ifndef NFP_ERRNO
+#include <errno.h>
+#define NFP_ERRNO(x)    (errno = (x), -1)
+#endif
+
+static inline int
+pushpull_width(int pp)
+{
+	pp &= 0xf;
+
+	if (pp == 0)
+		return NFP_ERRNO(EINVAL);
+	return (2 << pp);
+}
+
+#define PUSH_WIDTH(_pushpull)      pushpull_width((_pushpull) >> 0)
+#define PULL_WIDTH(_pushpull)      pushpull_width((_pushpull) >> 4)
+
+static inline int
+target_rw(uint32_t cpp_id, int pp, int start, int len)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island && (island < start || island > (start + len)))
+		return NFP_ERRNO(EINVAL);
+
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 0):
+		return PUSHPULL(0, pp);
+	case NFP_CPP_ID(0, 1, 0):
+		return PUSHPULL(pp, 0);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 0):
+		return PUSHPULL(pp, pp);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp6000_nbi_dma(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 0): /* ReadNbiDma */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 1, 0): /* WriteNbiDma */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 0):
+		return PUSHPULL(P64, P64);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp6000_nbi_stats(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 0): /* ReadNbiStats */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 1, 0): /* WriteNbiStats */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 0):
+		return PUSHPULL(P64, P64);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp6000_nbi_tm(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 0): /* ReadNbiTM */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 1, 0):  /* WriteNbiTM */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 0):
+		return PUSHPULL(P64, P64);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp6000_nbi_ppc(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 0): /* ReadNbiPreclassifier */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 1, 0): /* WriteNbiPreclassifier */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 0):
+		return PUSHPULL(P64, P64);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp6000_nbi(uint32_t cpp_id, uint64_t address)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+	uint64_t rel_addr = address & 0x3fFFFF;
+
+	if (island && (island < 8 || island > 9))
+		return NFP_ERRNO(EINVAL);
+
+	if (rel_addr < (1 << 20))
+		return nfp6000_nbi_dma(cpp_id);
+	if (rel_addr < (2 << 20))
+		return nfp6000_nbi_stats(cpp_id);
+	if (rel_addr < (3 << 20))
+		return nfp6000_nbi_tm(cpp_id);
+	return nfp6000_nbi_ppc(cpp_id);
+}
+
+/*
+ * This structure ONLY includes items that can be done with a read or write of
+ * 32-bit or 64-bit words. All others are not listed.
+ */
+static inline int
+nfp6000_mu_common(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 0): /* read_be/write_be */
+		return PUSHPULL(P64, P64);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 1): /* read_le/write_le */
+		return PUSHPULL(P64, P64);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 2): /* {read/write}_swap_be */
+		return PUSHPULL(P64, P64);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 3): /* {read/write}_swap_le */
+		return PUSHPULL(P64, P64);
+	case NFP_CPP_ID(0, 0, 0): /* read_be */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 0, 1): /* read_le */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 0, 2): /* read_swap_be */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 0, 3): /* read_swap_le */
+		return PUSHPULL(0, P64);
+	case NFP_CPP_ID(0, 1, 0): /* write_be */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, 1, 1): /* write_le */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, 1, 2): /* write_swap_be */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, 1, 3): /* write_swap_le */
+		return PUSHPULL(P64, 0);
+	case NFP_CPP_ID(0, 3, 0): /* atomic_read */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 3, 2): /* mask_compare_write */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 4, 0): /* atomic_write */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 4, 2): /* atomic_write_imm */
+		return PUSHPULL(0, 0);
+	case NFP_CPP_ID(0, 4, 3): /* swap_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 5, 0): /* set */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 5, 3): /* test_set_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 6, 0): /* clr */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 6, 3): /* test_clr_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 7, 0): /* add */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 7, 3): /* test_add_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 8, 0): /* addsat */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 8, 3): /* test_subsat_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 9, 0): /* sub */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 9, 3): /* test_sub_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 10, 0): /* subsat */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 10, 3): /* test_subsat_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 13, 0): /* microq128_get */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 13, 1): /* microq128_pop */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 13, 2): /* microq128_put */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 15, 0): /* xor */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 15, 3): /* test_xor_imm */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 28, 0): /* read32_be */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 28, 1): /* read32_le */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 28, 2): /* read32_swap_be */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 28, 3): /* read32_swap_le */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 31, 0): /* write32_be */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 31, 1): /* write32_le */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 31, 2): /* write32_swap_be */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 31, 3): /* write32_swap_le */
+		return PUSHPULL(P32, 0);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp6000_mu_ctm(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 16, 1): /* packet_read_packet_status */
+		return PUSHPULL(0, P32);
+	default:
+		return nfp6000_mu_common(cpp_id);
+	}
+}
+
+static inline int
+nfp6000_mu_emu(uint32_t cpp_id)
+{
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 18, 0): /* read_queue */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 18, 1): /* read_queue_ring */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 18, 2): /* write_queue */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 18, 3): /* write_queue_ring */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 20, 2): /* journal */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 21, 0): /* get */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 21, 1): /* get_eop */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 21, 2): /* get_freely */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 22, 0): /* pop */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 22, 1): /* pop_eop */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 22, 2): /* pop_freely */
+		return PUSHPULL(0, P32);
+	default:
+		return nfp6000_mu_common(cpp_id);
+	}
+}
+
+static inline int
+nfp6000_mu_imu(uint32_t cpp_id)
+{
+	return nfp6000_mu_common(cpp_id);
+}
+
+static inline int
+nfp6000_mu(uint32_t cpp_id, uint64_t address)
+{
+	int pp;
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island == 0) {
+		if (address < 0x2000000000ULL)
+			pp = nfp6000_mu_ctm(cpp_id);
+		else if (address < 0x8000000000ULL)
+			pp = nfp6000_mu_emu(cpp_id);
+		else if (address < 0x9800000000ULL)
+			pp = nfp6000_mu_ctm(cpp_id);
+		else if (address < 0x9C00000000ULL)
+			pp = nfp6000_mu_emu(cpp_id);
+		else if (address < 0xA000000000ULL)
+			pp = nfp6000_mu_imu(cpp_id);
+		else
+			pp = nfp6000_mu_ctm(cpp_id);
+	} else if (island >= 24 && island <= 27) {
+		pp = nfp6000_mu_emu(cpp_id);
+	} else if (island >= 28 && island <= 31) {
+		pp = nfp6000_mu_imu(cpp_id);
+	} else if (island == 1 ||
+		   (island >= 4 && island <= 7) ||
+		   (island >= 12 && island <= 13) ||
+		   (island >= 32 && island <= 47) ||
+		   (island >= 48 && island <= 51)) {
+		pp = nfp6000_mu_ctm(cpp_id);
+	} else {
+		pp = NFP_ERRNO(EINVAL);
+	}
+
+	return pp;
+}
+
+static inline int
+nfp6000_ila(uint32_t cpp_id)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island && (island < 48 || island > 51))
+		return NFP_ERRNO(EINVAL);
+
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 1): /* read_check_error */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 2, 0): /* read_int */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 3, 0): /* write_int */
+		return PUSHPULL(P32, 0);
+	default:
+		return target_rw(cpp_id, P32, 48, 4);
+	}
+}
+
+static inline int
+nfp6000_pci(uint32_t cpp_id)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island && (island < 4 || island > 7))
+		return NFP_ERRNO(EINVAL);
+
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 2, 0):
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 3, 0):
+		return PUSHPULL(P32, 0);
+	default:
+		return target_rw(cpp_id, P32, 4, 4);
+	}
+}
+
+static inline int
+nfp6000_crypto(uint32_t cpp_id)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island && (island < 12 || island > 15))
+		return NFP_ERRNO(EINVAL);
+
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 2, 0):
+		return PUSHPULL(P64, 0);
+	default:
+		return target_rw(cpp_id, P64, 12, 4);
+	}
+}
+
+static inline int
+nfp6000_cap_xpb(uint32_t cpp_id)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island && (island < 1 || island > 63))
+		return NFP_ERRNO(EINVAL);
+
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 1): /* RingGet */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 0, 2): /* Interthread Signal */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 1, 1): /* RingPut */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 1, 2): /* CTNNWr */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 2, 0): /* ReflectRd, signal none */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 2, 1): /* ReflectRd, signal self */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 2, 2): /* ReflectRd, signal remote */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 2, 3): /* ReflectRd, signal both */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 3, 0): /* ReflectWr, signal none */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 3, 1): /* ReflectWr, signal self */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 3, 2): /* ReflectWr, signal remote */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 3, 3): /* ReflectWr, signal both */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, NFP_CPP_ACTION_RW, 1):
+		return PUSHPULL(P32, P32);
+	default:
+		return target_rw(cpp_id, P32, 1, 63);
+	}
+}
+
+static inline int
+nfp6000_cls(uint32_t cpp_id)
+{
+	int island = NFP_CPP_ID_ISLAND_of(cpp_id);
+
+	if (island && (island < 1 || island > 63))
+		return NFP_ERRNO(EINVAL);
+
+	switch (cpp_id & NFP_CPP_ID(0, ~0, ~0)) {
+	case NFP_CPP_ID(0, 0, 3): /* xor */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 2, 0): /* set */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 2, 1): /* clr */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 4, 0): /* add */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 4, 1): /* add64 */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 6, 0): /* sub */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 6, 1): /* sub64 */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 6, 2): /* subsat */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 8, 2): /* hash_mask */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 8, 3): /* hash_clear */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 9, 0): /* ring_get */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 9, 1): /* ring_pop */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 9, 2): /* ring_get_freely */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 9, 3): /* ring_pop_freely */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 10, 0): /* ring_put */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 10, 2): /* ring_journal */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 14, 0): /* reflect_write_sig_local */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 15, 1):  /* reflect_read_sig_local */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 17, 2): /* statistic */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 24, 0): /* ring_read */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 24, 1): /* ring_write */
+		return PUSHPULL(P32, 0);
+	case NFP_CPP_ID(0, 25, 0): /* ring_workq_add_thread */
+		return PUSHPULL(0, P32);
+	case NFP_CPP_ID(0, 25, 1): /* ring_workq_add_work */
+		return PUSHPULL(P32, 0);
+	default:
+		return target_rw(cpp_id, P32, 0, 64);
+	}
+}
+
+static inline int
+nfp6000_target_pushpull(uint32_t cpp_id, uint64_t address)
+{
+	switch (NFP_CPP_ID_TARGET_of(cpp_id)) {
+	case NFP6000_CPPTGT_NBI:
+		return nfp6000_nbi(cpp_id, address);
+	case NFP6000_CPPTGT_VQDR:
+		return target_rw(cpp_id, P32, 24, 4);
+	case NFP6000_CPPTGT_ILA:
+		return nfp6000_ila(cpp_id);
+	case NFP6000_CPPTGT_MU:
+		return nfp6000_mu(cpp_id, address);
+	case NFP6000_CPPTGT_PCIE:
+		return nfp6000_pci(cpp_id);
+	case NFP6000_CPPTGT_ARM:
+		if (address < 0x10000)
+			return target_rw(cpp_id, P64, 1, 1);
+		else
+			return target_rw(cpp_id, P32, 1, 1);
+	case NFP6000_CPPTGT_CRYPTO:
+		return nfp6000_crypto(cpp_id);
+	case NFP6000_CPPTGT_CTXPB:
+		return nfp6000_cap_xpb(cpp_id);
+	case NFP6000_CPPTGT_CLS:
+		return nfp6000_cls(cpp_id);
+	case 0:
+		return target_rw(cpp_id, P32, 4, 4);
+	default:
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp_target_pushpull_width(int pp, int write_not_read)
+{
+	if (pp < 0)
+		return pp;
+
+	if (write_not_read)
+		return PULL_WIDTH(pp);
+	else
+		return PUSH_WIDTH(pp);
+}
+
+static inline int
+nfp6000_target_action_width(uint32_t cpp_id, uint64_t address,
+			    int write_not_read)
+{
+	int pp;
+
+	pp = nfp6000_target_pushpull(cpp_id, address);
+
+	return nfp_target_pushpull_width(pp, write_not_read);
+}
+
+static inline int
+nfp_target_action_width(uint32_t model, uint32_t cpp_id, uint64_t address,
+			int write_not_read)
+{
+	if (NFP_CPP_MODEL_IS_6000(model)) {
+		return nfp6000_target_action_width(cpp_id, address,
+						   write_not_read);
+	} else {
+		return NFP_ERRNO(EINVAL);
+	}
+}
+
+static inline int
+nfp_target_cpp(uint32_t cpp_island_id, uint64_t cpp_island_address,
+	       uint32_t *cpp_target_id, uint64_t *cpp_target_address,
+	       const uint32_t *imb_table)
+{
+	int err;
+	int island = NFP_CPP_ID_ISLAND_of(cpp_island_id);
+	int target = NFP_CPP_ID_TARGET_of(cpp_island_id);
+	uint32_t imb;
+
+	if (target < 0 || target >= 16)
+		return NFP_ERRNO(EINVAL);
+
+	if (island == 0) {
+		/* Already translated */
+		*cpp_target_id = cpp_island_id;
+		*cpp_target_address = cpp_island_address;
+		return 0;
+	}
+
+	if (!imb_table) {
+		/* CPP + Island only allowed on systems with IMB tables */
+		return NFP_ERRNO(EINVAL);
+	}
+
+	imb = imb_table[target];
+
+	*cpp_target_address = cpp_island_address;
+	err = _nfp6000_cppat_addr_encode(cpp_target_address, island, target,
+					 ((imb >> 13) & 7),
+					 ((imb >> 12) & 1),
+					 ((imb >> 6) & 0x3f),
+					 ((imb >> 0) & 0x3f));
+	if (err == 0) {
+		*cpp_target_id =
+		    NFP_CPP_ID(target, NFP_CPP_ID_ACTION_of(cpp_island_id),
+			       NFP_CPP_ID_TOKEN_of(cpp_island_id));
+	}
+
+	return err;
+}
+
+#endif /* NFP_TARGET_H */
-- 
1.9.1

^ permalink raw reply	[flat|nested] 12+ messages in thread

* [dpdk-dev] [PATCH 2/4] net/nfp: update PMD for using new CPP interface
  2018-03-23 17:35 [dpdk-dev] [PATCH 0/4] NFP PF support based on new CPP interface Alejandro Lucero
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support Alejandro Lucero
@ 2018-03-23 17:35 ` Alejandro Lucero
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 3/4] doc: update NFP guide Alejandro Lucero
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 4/4] net/nfp: remove files Alejandro Lucero
  3 siblings, 0 replies; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-23 17:35 UTC (permalink / raw)
  To: dev

PF PMD support was based on NSPU interface. This patch changes the
PMD for using the new CPP user space interface which gives more
flexibility for adding new functionalities, and specifically at this
point, for properly selecting the right firmware file which requires
to know about the card to work with.

This change just changes initialization with the datapath being unaffected.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 drivers/net/nfp/Makefile      |  17 ++-
 drivers/net/nfp/nfp_net.c     | 342 +++++++++++++++++++++++++++++-------------
 drivers/net/nfp/nfp_net_pmd.h |  16 +-
 3 files changed, 264 insertions(+), 111 deletions(-)

diff --git a/drivers/net/nfp/Makefile b/drivers/net/nfp/Makefile
index aa3b68a..ab4e0a7 100644
--- a/drivers/net/nfp/Makefile
+++ b/drivers/net/nfp/Makefile
@@ -20,11 +20,24 @@ EXPORT_MAP := rte_pmd_nfp_version.map
 
 LIBABIVER := 1
 
+VPATH += $(SRCDIR)/nfpcore
+
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_cppcore.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_cpp_pcie_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_mutex.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_resource.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_crc.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_mip.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_nffw.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_hwinfo.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_rtsym.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_nsp.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_nsp_cmds.c
+SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_nsp_eth.c
+
 #
 # all source are stored in SRCS-y
 #
 SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_net.c
-SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_nfpu.c
-SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_nspu.c
 
 include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index e5bfde6..0657a23 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, 2015 Netronome Systems, Inc.
+ * Copyright (c) 2014-2018 Netronome Systems, Inc.
  * All rights reserved.
  *
  * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
@@ -55,7 +55,13 @@
 #include <rte_alarm.h>
 #include <rte_spinlock.h>
 
-#include "nfp_nfpu.h"
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_nffw.h"
+#include "nfpcore/nfp_hwinfo.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+#include "nfpcore/nfp_nsp.h"
+
 #include "nfp_net_pmd.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
@@ -104,12 +110,8 @@ static int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf);
 
-/*
- * The offset of the queue controller queues in the PCIe Target. These
- * happen to be at the same offset on the NFP6000 and the NFP3200 so
- * we use a single macro here.
- */
-#define NFP_PCIE_QUEUE(_q)	(0x800 * ((_q) & 0xff))
+/* The offset of the queue controller queues in the PCIe Target */
+#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
 
 /* Maximum value which can be added to a queue with one transaction */
 #define NFP_QCP_MAX_ADD	0x7f
@@ -625,47 +627,29 @@ enum nfp_qcp_ptr {
 #define ETH_ADDR_LEN	6
 
 static void
-nfp_eth_copy_mac_reverse(uint8_t *dst, const uint8_t *src)
+nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 {
 	int i;
 
 	for (i = 0; i < ETH_ADDR_LEN; i++)
-		dst[ETH_ADDR_LEN - i - 1] = src[i];
+		dst[i] = src[i];
 }
 
 static int
 nfp_net_pf_read_mac(struct nfp_net_hw *hw, int port)
 {
-	union eth_table_entry *entry;
-	int idx, i;
-
-	idx = port;
-	entry = hw->eth_table;
-
-	/* Reading NFP ethernet table obtained before */
-	for (i = 0; i < NSP_ETH_MAX_COUNT; i++) {
-		if (!(entry->port & NSP_ETH_PORT_LANES_MASK)) {
-			/* port not in use */
-			entry++;
-			continue;
-		}
-		if (idx == 0)
-			break;
-		idx--;
-		entry++;
-	}
-
-	if (i == NSP_ETH_MAX_COUNT)
-		return -EINVAL;
+	struct nfp_eth_table *nfp_eth_table;
 
+	nfp_eth_table = nfp_eth_read_ports(hw->cpp);
 	/*
 	 * hw points to port0 private data. We need hw now pointing to
 	 * right port.
 	 */
 	hw += port;
-	nfp_eth_copy_mac_reverse((uint8_t *)&hw->mac_addr,
-				 (uint8_t *)&entry->mac_addr);
+	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
+			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
 
+	free(nfp_eth_table);
 	return 0;
 }
 
@@ -831,7 +815,7 @@ enum nfp_qcp_ptr {
 
 	if (hw->is_pf)
 		/* Configure the physical port up */
-		nfp_nsp_eth_config(hw->nspu_desc, hw->pf_port_idx, 1);
+		nfp_eth_set_configured(hw->cpp, hw->pf_port_idx, 1);
 
 	hw->ctrl = new_ctrl;
 
@@ -882,7 +866,7 @@ enum nfp_qcp_ptr {
 
 	if (hw->is_pf)
 		/* Configure the physical port down */
-		nfp_nsp_eth_config(hw->nspu_desc, hw->pf_port_idx, 0);
+		nfp_eth_set_configured(hw->cpp, hw->pf_port_idx, 0);
 }
 
 /* Reset and stop device. The device can not be restarted. */
@@ -2734,10 +2718,8 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	uint64_t tx_bar_off = 0, rx_bar_off = 0;
 	uint32_t start_q;
 	int stride = 4;
-
-	nspu_desc_t *nspu_desc = NULL;
-	uint64_t bar_offset;
 	int port = 0;
+	int err;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -2758,7 +2740,6 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 
 		/* This points to the specific port private data */
 		hw = &hwport0[port];
-		hw->pf_port_idx = port;
 	} else {
 		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 		hwport0 = 0;
@@ -2792,19 +2773,14 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	}
 
 	if (hw->is_pf && port == 0) {
-		nspu_desc = hw->nspu_desc;
-
-		if (nfp_nsp_map_ctrl_bar(nspu_desc, &bar_offset) != 0) {
-			/*
-			 * A firmware should be there after PF probe so this
-			 * should not happen.
-			 */
-			RTE_LOG(ERR, PMD, "PF BAR symbol resolution failed\n");
-			return -ENODEV;
+		hw->ctrl_bar = nfp_rtsym_map(hw->sym_tbl, "_pf0_net_bar0",
+					     hw->total_ports * 32768,
+					     &hw->ctrl_area);
+		if (!hw->ctrl_bar) {
+			printf("nfp_rtsym_map fails for _pf0_net_ctrl_bar\n");
+			return -EIO;
 		}
 
-		/* vNIC PF control BAR is a subset of PF PCI device BAR */
-		hw->ctrl_bar += bar_offset;
 		PMD_INIT_LOG(DEBUG, "ctrl bar: %p\n", hw->ctrl_bar);
 	}
 
@@ -2828,13 +2804,14 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	case PCI_DEVICE_ID_NFP6000_PF_NIC:
 	case PCI_DEVICE_ID_NFP6000_VF_NIC:
 		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
-		tx_bar_off = NFP_PCIE_QUEUE(start_q);
+		tx_bar_off = start_q * NFP_QCP_QUEUE_ADDR_SZ;
 		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
-		rx_bar_off = NFP_PCIE_QUEUE(start_q);
+		rx_bar_off = start_q * NFP_QCP_QUEUE_ADDR_SZ;
 		break;
 	default:
 		RTE_LOG(ERR, PMD, "nfp_net: no device ID matching\n");
-		return -ENODEV;
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
 	}
 
 	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "\n", tx_bar_off);
@@ -2842,17 +2819,19 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 
 	if (hw->is_pf && port == 0) {
 		/* configure access to tx/rx vNIC BARs */
-		nfp_nsp_map_queues_bar(nspu_desc, &bar_offset);
-		PMD_INIT_LOG(DEBUG, "tx/rx bar_offset: %" PRIx64 "\n",
-				    bar_offset);
-		hwport0->hw_queues = (uint8_t *)pci_dev->mem_resource[0].addr;
-
-		/* vNIC PF tx/rx BARs are a subset of PF PCI device */
-		hwport0->hw_queues += bar_offset;
+		hwport0->hw_queues = nfp_cpp_map_area(hw->cpp, 0, 0,
+						      NFP_PCIE_QUEUE(0),
+						      NFP_QCP_QUEUE_AREA_SZ,
+						      &hw->hwqueues_area);
+
+		if (!hwport0->hw_queues) {
+			printf("nfp_rtsym_map fails for net.qc\n");
+			err = -EIO;
+			goto dev_err_ctrl_map;
+		}
 
-		/* Lets seize the chance to read eth table from hw */
-		if (nfp_nsp_eth_read_table(nspu_desc, &hw->eth_table))
-			return -ENODEV;
+		PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p\n",
+				    hwport0->hw_queues);
 	}
 
 	if (hw->is_pf) {
@@ -2912,7 +2891,8 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr", ETHER_ADDR_LEN, 0);
 	if (eth_dev->data->mac_addrs == NULL) {
 		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
-		return -ENOMEM;
+		err = -ENOMEM;
+		goto dev_err_queues_map;
 	}
 
 	if (hw->is_pf) {
@@ -2923,6 +2903,8 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	}
 
 	if (!is_valid_assigned_ether_addr((struct ether_addr *)&hw->mac_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d\n",
+				   port);
 		/* Using random mac addresses for VFs */
 		eth_random_addr(&hw->mac_addr[0]);
 		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
@@ -2951,11 +2933,19 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	nfp_net_stats_reset(eth_dev);
 
 	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
 }
 
 static int
 nfp_pf_create_dev(struct rte_pci_device *dev, int port, int ports,
-		  nfpu_desc_t *nfpu_desc, void **priv)
+		  struct nfp_cpp *cpp, struct nfp_hwinfo *hwinfo,
+		  int phys_port, struct nfp_rtsym_table *sym_tbl, void **priv)
 {
 	struct rte_eth_dev *eth_dev;
 	struct nfp_net_hw *hw;
@@ -2993,12 +2983,16 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	 * Then dev_private is adjusted per port.
 	 */
 	hw = (struct nfp_net_hw *)(eth_dev->data->dev_private) + port;
-	hw->nspu_desc = nfpu_desc->nspu;
-	hw->nfpu_desc = nfpu_desc;
+	hw->cpp = cpp;
+	hw->hwinfo = hwinfo;
+	hw->sym_tbl = sym_tbl;
+	hw->pf_port_idx = phys_port;
 	hw->is_pf = 1;
 	if (ports > 1)
 		hw->pf_multiport_enabled = 1;
 
+	hw->total_ports = ports;
+
 	eth_dev->device = &dev->device;
 	rte_eth_copy_pci_info(eth_dev, dev);
 
@@ -3012,55 +3006,191 @@ uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 	return ret;
 }
 
+#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
+
+static int
+nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
+{
+	struct nfp_cpp *cpp = nsp->cpp;
+	int fw_f;
+	char *fw_buf;
+	char fw_name[100];
+	char serial[100];
+	struct stat file_stat;
+	off_t fsize, bytes;
+
+	/* Looking for firmware file in order of priority */
+
+	/* First try to find a firmware image specific for this device */
+	sprintf(serial, "serial-%02x-%02x-%02x-%02x-%02x-%02x-%02hhx-%02hhx",
+		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
+		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
+		cpp->interface & 0xff);
+
+	sprintf(fw_name, "%s/%s.nffw", DEFAULT_FW_PATH, serial);
+
+	RTE_LOG(DEBUG, PMD, "Trying with fw file: %s\n", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f > 0)
+		goto read_fw;
+
+	/* Then try the PCI name */
+	sprintf(fw_name, "%s/pci-%s.nffw", DEFAULT_FW_PATH, dev->device.name);
+
+	RTE_LOG(DEBUG, PMD, "Trying with fw file: %s\n", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f > 0)
+		goto read_fw;
+
+	/* Finally try the card type and media */
+	sprintf(fw_name, "%s/%s", DEFAULT_FW_PATH, card);
+	RTE_LOG(DEBUG, PMD, "Trying with fw file: %s\n", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f < 0) {
+		RTE_LOG(INFO, PMD, "Firmware file %s not found.", fw_name);
+		return -ENOENT;
+	}
+
+read_fw:
+	if (fstat(fw_f, &file_stat) < 0) {
+		RTE_LOG(INFO, PMD, "Firmware file %s size is unknown", fw_name);
+		close(fw_f);
+		return -ENOENT;
+	}
+
+	fsize = file_stat.st_size;
+	RTE_LOG(INFO, PMD, "Firmware file found at %s with size: %" PRIu64 "\n",
+			    fw_name, (uint64_t)fsize);
+
+	fw_buf = malloc((size_t)fsize);
+	if (!fw_buf) {
+		RTE_LOG(INFO, PMD, "malloc failed for fw buffer");
+		close(fw_f);
+		return -ENOMEM;
+	}
+	memset(fw_buf, 0, fsize);
+
+	bytes = read(fw_f, fw_buf, fsize);
+	if (bytes != fsize) {
+		RTE_LOG(INFO, PMD, "Reading fw to buffer failed.\n"
+				   "Just %" PRIu64 " of %" PRIu64 " bytes read",
+				   (uint64_t)bytes, (uint64_t)fsize);
+		free(fw_buf);
+		close(fw_f);
+		return -EIO;
+	}
+
+	RTE_LOG(INFO, PMD, "Uploading the firmware ...");
+	nfp_nsp_load_fw(nsp, fw_buf, bytes);
+	RTE_LOG(INFO, PMD, "Done");
+
+	free(fw_buf);
+	close(fw_f);
+
+	return 0;
+}
+
+static int
+nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
+	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
+{
+	struct nfp_nsp *nsp;
+	const char *nfp_fw_model;
+	char card_desc[100];
+	int err = 0;
+
+	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
+
+	if (nfp_fw_model) {
+		RTE_LOG(INFO, PMD, "firmware model found: %s\n", nfp_fw_model);
+	} else {
+		RTE_LOG(ERR, PMD, "firmware model NOT found\n");
+		return -EIO;
+	}
+
+	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
+		RTE_LOG(ERR, PMD, "NFP ethernet table reports wrong ports: %u\n",
+		       nfp_eth_table->count);
+		return -EIO;
+	}
+
+	RTE_LOG(INFO, PMD, "NFP ethernet port table reports %u ports\n",
+			   nfp_eth_table->count);
+
+	RTE_LOG(INFO, PMD, "Port speed: %u\n", nfp_eth_table->ports[0].speed);
+
+	sprintf(card_desc, "nic_%s_%dx%d.nffw", nfp_fw_model,
+		nfp_eth_table->count, nfp_eth_table->ports[0].speed / 1000);
+
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp) {
+		RTE_LOG(ERR, PMD, "NFP error when obtaining NSP handle\n");
+		return -EIO;
+	}
+
+	nfp_nsp_device_soft_reset(nsp);
+	err = nfp_fw_upload(dev, nsp, card_desc);
+
+	nfp_nsp_close(nsp);
+	return err;
+}
+
 static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 			    struct rte_pci_device *dev)
 {
-	nfpu_desc_t *nfpu_desc;
-	nspu_desc_t *nspu_desc;
-	uint64_t offset_symbol;
-	uint8_t *bar_offset;
-	int major, minor;
+	struct nfp_cpp *cpp;
+	struct nfp_hwinfo *hwinfo;
+	struct nfp_rtsym_table *sym_tbl;
+	struct nfp_eth_table *nfp_eth_table = NULL;
 	int total_ports;
 	void *priv = 0;
 	int ret = -ENODEV;
+	int err;
 	int i;
 
 	if (!dev)
 		return ret;
 
-	nfpu_desc = rte_malloc("nfp nfpu", sizeof(nfpu_desc_t), 0);
-	if (!nfpu_desc)
-		return -ENOMEM;
-
-	if (nfpu_open(dev, nfpu_desc, 0) < 0) {
-		RTE_LOG(ERR, PMD,
-			"nfpu_open failed\n");
-		goto nfpu_error;
+	cpp = nfp_cpp_from_device_name(dev->device.name);
+	if (!cpp) {
+		RTE_LOG(ERR, PMD, "A CPP handle can not be obtained");
+		ret = -EIO;
+		goto error;
 	}
 
-	nspu_desc = nfpu_desc->nspu;
+	hwinfo = nfp_hwinfo_read(cpp);
+	if (!hwinfo) {
+		RTE_LOG(ERR, PMD, "Error reading hwinfo table");
+		return -EIO;
+	}
 
+	nfp_eth_table = nfp_eth_read_ports(cpp);
+	if (!nfp_eth_table) {
+		RTE_LOG(ERR, PMD, "Error reading NFP ethernet table\n");
+		return -EIO;
+	}
 
-	/* Check NSP ABI version */
-	if (nfp_nsp_get_abi_version(nspu_desc, &major, &minor) < 0) {
-		RTE_LOG(INFO, PMD, "NFP NSP not present\n");
+	if (nfp_fw_setup(dev, cpp, nfp_eth_table, hwinfo)) {
+		RTE_LOG(INFO, PMD, "Error when uploading firmware\n");
+		ret = -EIO;
 		goto error;
 	}
-	PMD_INIT_LOG(INFO, "nspu ABI version: %d.%d\n", major, minor);
 
-	if ((major == 0) && (minor < 20)) {
-		RTE_LOG(INFO, PMD, "NFP NSP ABI version too old. Required 0.20 or higher\n");
+	/* Now the symbol table should be there */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		RTE_LOG(ERR, PMD, "Something is wrong with the firmware"
+				" symbol table");
+		ret = -EIO;
 		goto error;
 	}
 
-	ret = nfp_nsp_fw_setup(nspu_desc, "nfd_cfg_pf0_num_ports",
-			       &offset_symbol);
-	if (ret)
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+	if (total_ports != (int)nfp_eth_table->count) {
+		RTE_LOG(ERR, PMD, "Inconsistent number of ports\n");
+		ret = -EIO;
 		goto error;
-
-	bar_offset = (uint8_t *)dev->mem_resource[0].addr;
-	bar_offset += offset_symbol;
-	total_ports = (uint32_t)*bar_offset;
+	}
 	PMD_INIT_LOG(INFO, "Total pf ports: %d\n", total_ports);
 
 	if (total_ports <= 0 || total_ports > 8) {
@@ -3070,18 +3200,15 @@ static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 	}
 
 	for (i = 0; i < total_ports; i++) {
-		ret = nfp_pf_create_dev(dev, i, total_ports, nfpu_desc, &priv);
+		ret = nfp_pf_create_dev(dev, i, total_ports, cpp, hwinfo,
+					nfp_eth_table->ports[i].index,
+					sym_tbl, &priv);
 		if (ret)
-			goto error;
+			break;
 	}
 
-	return 0;
-
 error:
-	nfpu_close(nfpu_desc);
-nfpu_error:
-	rte_free(nfpu_desc);
-
+	free(nfp_eth_table);
 	return ret;
 }
 
@@ -3129,8 +3256,19 @@ static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
 	if ((pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC) ||
 	    (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) {
 		port = get_pf_port_number(eth_dev->data->name);
+		/*
+		 * hotplug is not possible with multiport PF although freeing
+		 * data structures can be done for first port.
+		 */
+		if (port != 0)
+			return -ENOTSUP;
 		hwport0 = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 		hw = &hwport0[port];
+		nfp_cpp_area_free(hw->ctrl_area);
+		nfp_cpp_area_free(hw->hwqueues_area);
+		free(hw->hwinfo);
+		free(hw->sym_tbl);
+		nfp_cpp_free(hw->cpp);
 	} else {
 		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 	}
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 1ae0ea6..097c871 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, 2015 Netronome Systems, Inc.
+ * Copyright (c) 2014-2018 Netronome Systems, Inc.
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -63,6 +63,7 @@
 #define NFP_NET_CRTL_BAR        0
 #define NFP_NET_TX_BAR          2
 #define NFP_NET_RX_BAR          2
+#define NFP_QCP_QUEUE_AREA_SZ			0x80000
 
 /* Macros for accessing the Queue Controller Peripheral 'CSRs' */
 #define NFP_QCP_QUEUE_OFF(_x)                 ((_x) * 0x800)
@@ -430,20 +431,21 @@ struct nfp_net_hw {
 	/* Records starting point for counters */
 	struct rte_eth_stats eth_stats_base;
 
-#ifdef NFP_NET_LIBNFP
 	struct nfp_cpp *cpp;
 	struct nfp_cpp_area *ctrl_area;
-	struct nfp_cpp_area *tx_area;
-	struct nfp_cpp_area *rx_area;
+	struct nfp_cpp_area *hwqueues_area;
 	struct nfp_cpp_area *msix_area;
-#endif
+
 	uint8_t *hw_queues;
 	uint8_t is_pf;
 	uint8_t pf_port_idx;
 	uint8_t pf_multiport_enabled;
+	uint8_t total_ports;
+
 	union eth_table_entry *eth_table;
-	nspu_desc_t *nspu_desc;
-	nfpu_desc_t *nfpu_desc;
+
+	struct nfp_hwinfo *hwinfo;
+	struct nfp_rtsym_table *sym_tbl;
 };
 
 struct nfp_net_adapter {
-- 
1.9.1

^ permalink raw reply	[flat|nested] 12+ messages in thread

* [dpdk-dev] [PATCH 3/4] doc: update NFP guide
  2018-03-23 17:35 [dpdk-dev] [PATCH 0/4] NFP PF support based on new CPP interface Alejandro Lucero
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support Alejandro Lucero
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 2/4] net/nfp: update PMD for using new CPP interface Alejandro Lucero
@ 2018-03-23 17:35 ` Alejandro Lucero
  2018-03-30 10:37   ` Ferruh Yigit
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 4/4] net/nfp: remove files Alejandro Lucero
  3 siblings, 1 reply; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-23 17:35 UTC (permalink / raw)
  To: dev

New CPP interface changes the way firmware upload is managed by
the PMD. It also supports different firmware file names for
having specific firmware aplications per card.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 doc/guides/nics/nfp.rst | 31 ++++++++++++++++++++++++-------
 1 file changed, 24 insertions(+), 7 deletions(-)

diff --git a/doc/guides/nics/nfp.rst b/doc/guides/nics/nfp.rst
index 99a3b76..6b1f9fc 100644
--- a/doc/guides/nics/nfp.rst
+++ b/doc/guides/nics/nfp.rst
@@ -99,16 +99,33 @@ have a PMD able to work with the PF and VFs at the same time and with the PF
 implementing VF management along with other PF-only functionalities/offloads.
 
 The PMD PF has extra work to do which will delay the DPDK app initialization
-like checking if a firmware is already available in the device, uploading the
-firmware if necessary, and configure the Link state properly when starting or
-stopping a PF port. Note that firmware upload is not always necessary which is
-the main delay for NFP PF PMD initialization.
+like uploading the firmware and configure the Link state properly when starting or
+stopping a PF port. Since DPDK 18.05 the firmware upload is always happening when
+a PF is initialized, which was not always true with older DPDK versions.
 
 Depending on the Netronome product installed in the system, firmware files
 should be available under ``/lib/firmware/netronome``. DPDK PMD supporting the
-PF requires a specific link, ``/lib/firmware/netronome/nic_dpdk_default.nffw``,
-which should be created automatically with Netronome's Agilio products
-installation.
+PF looks for a firmware file in this order:
+
+	1) First try to find a firmware image specific for this device using the
+	   NFP serial number:
+
+		serial-00-15-4d-12-20-65-10-ff.nffw
+
+	2) Then try the PCI name:
+
+		pci-0000:04:00.0.nffw
+
+	3) Finally try the card type and media:
+
+		nic_AMDA0099-0001_2x25.nffw
+
+Netronome's software packages install firmware files under ``/lib/firmware/netronome``
+for supporting all the Netronome's SmartNICs and also different firmware applications.
+This is usually done using file names based on SmartNIC type and media and with a
+directory per firmware application. Options 1 and 2 for firmware filenames allow to
+have more than one SmartNIC, same type of SmartNIC or different ones, and to upload a
+different firmware to each SmartNIC.
 
 PF multiport support
 --------------------
-- 
1.9.1

^ permalink raw reply	[flat|nested] 12+ messages in thread

* [dpdk-dev] [PATCH 4/4] net/nfp: remove files
  2018-03-23 17:35 [dpdk-dev] [PATCH 0/4] NFP PF support based on new CPP interface Alejandro Lucero
                   ` (2 preceding siblings ...)
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 3/4] doc: update NFP guide Alejandro Lucero
@ 2018-03-23 17:35 ` Alejandro Lucero
  3 siblings, 0 replies; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-23 17:35 UTC (permalink / raw)
  To: dev

New CPP interface makes NSPU interface obsolete. These files are
not needed anymore.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 drivers/net/nfp/nfp_net_eth.h |  82 ------
 drivers/net/nfp/nfp_nfpu.c    | 108 -------
 drivers/net/nfp/nfp_nfpu.h    |  55 ----
 drivers/net/nfp/nfp_nspu.c    | 642 ------------------------------------------
 drivers/net/nfp/nfp_nspu.h    |  83 ------
 5 files changed, 970 deletions(-)
 delete mode 100644 drivers/net/nfp/nfp_net_eth.h
 delete mode 100644 drivers/net/nfp/nfp_nfpu.c
 delete mode 100644 drivers/net/nfp/nfp_nfpu.h
 delete mode 100644 drivers/net/nfp/nfp_nspu.c
 delete mode 100644 drivers/net/nfp/nfp_nspu.h

diff --git a/drivers/net/nfp/nfp_net_eth.h b/drivers/net/nfp/nfp_net_eth.h
deleted file mode 100644
index af57f03..0000000
--- a/drivers/net/nfp/nfp_net_eth.h
+++ /dev/null
@@ -1,82 +0,0 @@
-/*
- * Copyright (c) 2017 Netronome Systems, Inc.
- * 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 copyright holder 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 HOLDER 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.
- */
-
-/*
- * vim:shiftwidth=8:noexpandtab
- *
- * @file dpdk/pmd/nfp_net_eth.h
- *
- * Netronome NFP_NET PDM driver
- */
-
-union eth_table_entry {
-	struct {
-		uint64_t port;
-		uint64_t state;
-		uint8_t mac_addr[6];
-		uint8_t resv[2];
-		uint64_t control;
-	};
-	uint64_t raw[4];
-};
-
-#ifndef BIT_ULL
-#define BIT_ULL(a) (1ULL << (a))
-#endif
-
-#define NSP_ETH_NBI_PORT_COUNT          24
-#define NSP_ETH_MAX_COUNT               (2 * NSP_ETH_NBI_PORT_COUNT)
-#define NSP_ETH_TABLE_SIZE   (NSP_ETH_MAX_COUNT * sizeof(union eth_table_entry))
-
-#define NSP_ETH_PORT_LANES              0xf
-#define NSP_ETH_PORT_INDEX              0xff00
-#define NSP_ETH_PORT_LABEL              0x3f000000000000
-#define NSP_ETH_PORT_PHYLABEL           0xfc0000000000000
-
-#define NSP_ETH_PORT_LANES_MASK         rte_cpu_to_le_64(NSP_ETH_PORT_LANES)
-
-#define NSP_ETH_STATE_CONFIGURED        BIT_ULL(0)
-#define NSP_ETH_STATE_ENABLED           BIT_ULL(1)
-#define NSP_ETH_STATE_TX_ENABLED        BIT_ULL(2)
-#define NSP_ETH_STATE_RX_ENABLED        BIT_ULL(3)
-#define NSP_ETH_STATE_RATE              0xf00
-#define NSP_ETH_STATE_INTERFACE         0xff000
-#define NSP_ETH_STATE_MEDIA             0x300000
-#define NSP_ETH_STATE_OVRD_CHNG         BIT_ULL(22)
-#define NSP_ETH_STATE_ANEG              0x3800000
-
-#define NSP_ETH_CTRL_CONFIGURED         BIT_ULL(0)
-#define NSP_ETH_CTRL_ENABLED            BIT_ULL(1)
-#define NSP_ETH_CTRL_TX_ENABLED         BIT_ULL(2)
-#define NSP_ETH_CTRL_RX_ENABLED         BIT_ULL(3)
-#define NSP_ETH_CTRL_SET_RATE           BIT_ULL(4)
-#define NSP_ETH_CTRL_SET_LANES          BIT_ULL(5)
-#define NSP_ETH_CTRL_SET_ANEG           BIT_ULL(6)
diff --git a/drivers/net/nfp/nfp_nfpu.c b/drivers/net/nfp/nfp_nfpu.c
deleted file mode 100644
index f11afef..0000000
--- a/drivers/net/nfp/nfp_nfpu.c
+++ /dev/null
@@ -1,108 +0,0 @@
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <unistd.h>
-#include <errno.h>
-#include <sys/file.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <sys/types.h>
-
-#include <rte_bus_pci.h>
-#include <rte_malloc.h>
-
-#include "nfp_nfpu.h"
-
-/* PF BAR and expansion BAR for the NSP interface */
-#define NFP_CFG_PCIE_BAR        0
-#define NFP_CFG_EXP_BAR         7
-
-#define NFP_CFG_EXP_BAR_CFG_BASE	0x30000
-
-/* There could be other NFP userspace tools using the NSP interface.
- * Make sure there is no other process using it and locking the access for
- * avoiding problems.
- */
-static int
-nspv_aquire_process_lock(nfpu_desc_t *desc)
-{
-	int rc;
-	struct flock lock;
-	char lockname[30];
-
-	memset(&lock, 0, sizeof(lock));
-
-	snprintf(lockname, sizeof(lockname), "/var/lock/nfp%d", desc->nfp);
-
-	/* Using S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH */
-	desc->lock = open(lockname, O_RDWR | O_CREAT, 0666);
-
-	if (desc->lock < 0)
-		return desc->lock;
-
-	lock.l_type = F_WRLCK;
-	lock.l_whence = SEEK_SET;
-	rc = -1;
-	while (rc != 0) {
-		rc = fcntl(desc->lock, F_SETLK, &lock);
-		if (rc < 0) {
-			if ((errno != EAGAIN) && (errno != EACCES)) {
-				close(desc->lock);
-				return rc;
-			}
-		}
-	}
-
-	return 0;
-}
-
-int
-nfpu_open(struct rte_pci_device *pci_dev, nfpu_desc_t *desc, int nfp)
-{
-	void *cfg_base, *mem_base;
-	size_t barsz;
-	int ret = 0;
-	int i = 0;
-
-	desc->nfp = nfp;
-
-	ret = nspv_aquire_process_lock(desc);
-	if (ret)
-		return -1;
-
-	barsz = pci_dev->mem_resource[0].len;
-
-	/* barsz in log2 */
-	while (barsz >>= 1)
-		i++;
-
-	barsz = i;
-
-	/* Sanity check: we can assume any bar size less than 1MB an error */
-	if (barsz < 20)
-		return -1;
-
-	/* Getting address for NFP expansion BAR registers */
-	cfg_base = pci_dev->mem_resource[0].addr;
-	cfg_base = (uint8_t *)cfg_base + NFP_CFG_EXP_BAR_CFG_BASE;
-
-	/* Getting address for NFP NSP interface registers */
-	mem_base = pci_dev->mem_resource[0].addr;
-	mem_base = (uint8_t *)mem_base + (NFP_CFG_EXP_BAR << (barsz - 3));
-
-
-	desc->nspu = rte_malloc("nfp nspu", sizeof(nspu_desc_t), 0);
-	nfp_nspu_init(desc->nspu, desc->nfp, NFP_CFG_PCIE_BAR, barsz,
-		      NFP_CFG_EXP_BAR, cfg_base, mem_base);
-
-	return ret;
-}
-
-int
-nfpu_close(nfpu_desc_t *desc)
-{
-	rte_free(desc->nspu);
-	close(desc->lock);
-	unlink("/var/lock/nfp0");
-	return 0;
-}
diff --git a/drivers/net/nfp/nfp_nfpu.h b/drivers/net/nfp/nfp_nfpu.h
deleted file mode 100644
index e56fa09..0000000
--- a/drivers/net/nfp/nfp_nfpu.h
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2017 Netronome Systems, Inc.
- * 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 copyright holder 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 HOLDER 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.
- */
-
-/*
- * vim:shiftwidth=8:noexpandtab
- *
- * @file dpdk/pmd/nfp_nfpu.h
- *
- * Netronome NFP_NET PDM driver
- */
-
-/*
- * NFP User interface creates a window for talking with NFP NSP processor
- */
-
-
-#include <rte_bus_pci.h>
-#include "nfp_nspu.h"
-
-typedef struct {
-	int nfp;
-	int lock;
-	nspu_desc_t *nspu;
-} nfpu_desc_t;
-
-int nfpu_open(struct rte_pci_device *pci_dev, nfpu_desc_t *desc, int nfp);
-int nfpu_close(nfpu_desc_t *desc);
diff --git a/drivers/net/nfp/nfp_nspu.c b/drivers/net/nfp/nfp_nspu.c
deleted file mode 100644
index f908983..0000000
--- a/drivers/net/nfp/nfp_nspu.c
+++ /dev/null
@@ -1,642 +0,0 @@
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/types.h>
-#include <sys/file.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-
-#include <rte_log.h>
-#include <rte_byteorder.h>
-
-#include "nfp_nfpu.h"
-
-#define CFG_EXP_BAR_ADDR_SZ     1
-#define CFG_EXP_BAR_MAP_TYPE	1
-
-#define EXP_BAR_TARGET_SHIFT     23
-#define EXP_BAR_LENGTH_SHIFT     27 /* 0=32, 1=64 bit increment */
-#define EXP_BAR_MAP_TYPE_SHIFT   29 /* Bulk BAR map */
-
-/* NFP target for NSP access */
-#define NFP_NSP_TARGET   7
-
-/* Expansion BARs for mapping PF vnic BARs */
-#define NFP_NET_PF_CFG_EXP_BAR          6
-#define NFP_NET_PF_HW_QUEUES_EXP_BAR    5
-
-/*
- * This is an NFP internal address used for configuring properly an NFP
- * expansion BAR.
- */
-#define MEM_CMD_BASE_ADDR       0x8100000000
-
-/* NSP interface registers */
-#define NSP_BASE                (MEM_CMD_BASE_ADDR + 0x22100)
-#define NSP_STATUS              0x00
-#define NSP_COMMAND             0x08
-#define NSP_BUFFER		0x10
-#define NSP_DEFAULT_BUF         0x18
-#define NSP_DEFAULT_BUF_CFG  0x20
-
-#define NSP_MAGIC                0xab10
-#define NSP_STATUS_MAGIC(x)      (((x) >> 48) & 0xffff)
-#define NSP_STATUS_MAJOR(x)      (int)(((x) >> 44) & 0xf)
-#define NSP_STATUS_MINOR(x)      (int)(((x) >> 32) & 0xfff)
-
-/* NSP commands */
-#define NSP_CMD_RESET                   1
-#define NSP_CMD_FW_LOAD                 6
-#define NSP_CMD_READ_ETH_TABLE          7
-#define NSP_CMD_WRITE_ETH_TABLE         8
-#define NSP_CMD_GET_SYMBOL             14
-
-#define NSP_BUFFER_CFG_SIZE_MASK	(0xff)
-
-#define NSP_REG_ADDR(d, off, reg) ((uint8_t *)(d)->mem_base + (off) + (reg))
-#define NSP_REG_VAL(p) (*(uint64_t *)(p))
-
-/*
- * An NFP expansion BAR is configured for allowing access to a specific NFP
- * target:
- *
- *  IN:
- *	desc: struct with basic NSP addresses to work with
- *	expbar: NFP PF expansion BAR index to configure
- *	tgt: NFP target to configure access
- *	addr: NFP target address
- *
- *  OUT:
- *	pcie_offset: NFP PCI BAR offset to work with
- */
-static void
-nfp_nspu_mem_bar_cfg(nspu_desc_t *desc, int expbar, int tgt,
-		     uint64_t addr, uint64_t *pcie_offset)
-{
-	uint64_t x, y, barsz;
-	uint32_t *expbar_ptr;
-
-	barsz = desc->barsz;
-
-	/*
-	 * NFP CPP address to configure. This comes from NFP 6000
-	 * datasheet document based on Bulk mapping.
-	 */
-	x = (addr >> (barsz - 3)) << (21 - (40 - (barsz - 3)));
-	x |= CFG_EXP_BAR_MAP_TYPE << EXP_BAR_MAP_TYPE_SHIFT;
-	x |= CFG_EXP_BAR_ADDR_SZ << EXP_BAR_LENGTH_SHIFT;
-	x |= tgt << EXP_BAR_TARGET_SHIFT;
-
-	/* Getting expansion bar configuration register address */
-	expbar_ptr = (uint32_t *)desc->cfg_base;
-	/* Each physical PCI BAR has 8 NFP expansion BARs */
-	expbar_ptr += (desc->pcie_bar * 8) + expbar;
-
-	/* Writing to the expansion BAR register */
-	*expbar_ptr = (uint32_t)x;
-
-	/* Getting the pcie offset to work with from userspace */
-	y = addr & ((uint64_t)(1 << (barsz - 3)) - 1);
-	*pcie_offset = y;
-}
-
-/*
- * Configuring an expansion bar for accessing NSP userspace interface. This
- * function configures always the same expansion bar, which implies access to
- * previously configured NFP target is lost.
- */
-static void
-nspu_xlate(nspu_desc_t *desc, uint64_t addr, uint64_t *pcie_offset)
-{
-	nfp_nspu_mem_bar_cfg(desc, desc->exp_bar, NFP_NSP_TARGET, addr,
-			     pcie_offset);
-}
-
-int
-nfp_nsp_get_abi_version(nspu_desc_t *desc, int *major, int *minor)
-{
-	uint64_t pcie_offset;
-	uint64_t nsp_reg;
-
-	nspu_xlate(desc, NSP_BASE, &pcie_offset);
-	nsp_reg = NSP_REG_VAL(NSP_REG_ADDR(desc, pcie_offset, NSP_STATUS));
-
-	if (NSP_STATUS_MAGIC(nsp_reg) != NSP_MAGIC)
-		return -1;
-
-	*major = NSP_STATUS_MAJOR(nsp_reg);
-	*minor = NSP_STATUS_MINOR(nsp_reg);
-
-	return 0;
-}
-
-int
-nfp_nspu_init(nspu_desc_t *desc, int nfp, int pcie_bar, size_t pcie_barsz,
-	      int exp_bar, void *exp_bar_cfg_base, void *exp_bar_mmap)
-{
-	uint64_t offset, buffaddr;
-	uint64_t nsp_reg;
-
-	desc->nfp = nfp;
-	desc->pcie_bar = pcie_bar;
-	desc->exp_bar = exp_bar;
-	desc->barsz = pcie_barsz;
-	desc->windowsz = 1 << (desc->barsz - 3);
-	desc->cfg_base = exp_bar_cfg_base;
-	desc->mem_base = exp_bar_mmap;
-
-	nspu_xlate(desc, NSP_BASE, &offset);
-
-	/*
-	 * Other NSPU clients can use other buffers. Let's tell NSPU we use the
-	 * default buffer.
-	 */
-	buffaddr = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_DEFAULT_BUF));
-	NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_BUFFER)) = buffaddr;
-
-	/* NFP internal addresses are 40 bits. Clean all other bits here */
-	buffaddr = buffaddr & (((uint64_t)1 << 40) - 1);
-	desc->bufaddr = buffaddr;
-
-	/* Lets get information about the buffer */
-	nsp_reg = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_DEFAULT_BUF_CFG));
-
-	/* Buffer size comes in MBs. Coversion to bytes */
-	desc->buf_size = ((size_t)nsp_reg & NSP_BUFFER_CFG_SIZE_MASK) << 20;
-
-	return 0;
-}
-
-#define NSPU_NFP_BUF(addr, base, off) \
-	(*(uint64_t *)((uint8_t *)(addr)->mem_base + ((base) | (off))))
-
-#define NSPU_HOST_BUF(base, off) (*(uint64_t *)((uint8_t *)(base) + (off)))
-
-static int
-nspu_buff_write(nspu_desc_t *desc, void *buffer, size_t size)
-{
-	uint64_t pcie_offset, pcie_window_base, pcie_window_offset;
-	uint64_t windowsz = desc->windowsz;
-	uint64_t buffaddr, j, i = 0;
-	int ret = 0;
-
-	if (size > desc->buf_size)
-		return -1;
-
-	buffaddr = desc->bufaddr;
-	windowsz = desc->windowsz;
-
-	while (i < size) {
-		/* Expansion bar reconfiguration per window size */
-		nspu_xlate(desc, buffaddr + i, &pcie_offset);
-		pcie_window_base = pcie_offset & (~(windowsz - 1));
-		pcie_window_offset = pcie_offset & (windowsz - 1);
-		for (j = pcie_window_offset; ((j < windowsz) && (i < size));
-		     j += 8) {
-			NSPU_NFP_BUF(desc, pcie_window_base, j) =
-				NSPU_HOST_BUF(buffer, i);
-			i += 8;
-		}
-	}
-
-	return ret;
-}
-
-static int
-nspu_buff_read(nspu_desc_t *desc, void *buffer, size_t size)
-{
-	uint64_t pcie_offset, pcie_window_base, pcie_window_offset;
-	uint64_t windowsz, i = 0, j;
-	uint64_t buffaddr;
-	int ret = 0;
-
-	if (size > desc->buf_size)
-		return -1;
-
-	buffaddr = desc->bufaddr;
-	windowsz = desc->windowsz;
-
-	while (i < size) {
-		/* Expansion bar reconfiguration per window size */
-		nspu_xlate(desc, buffaddr + i, &pcie_offset);
-		pcie_window_base = pcie_offset & (~(windowsz - 1));
-		pcie_window_offset = pcie_offset & (windowsz - 1);
-		for (j = pcie_window_offset; ((j < windowsz) && (i < size));
-		     j += 8) {
-			NSPU_HOST_BUF(buffer, i) =
-				NSPU_NFP_BUF(desc, pcie_window_base, j);
-			i += 8;
-		}
-	}
-
-	return ret;
-}
-
-static int
-nspu_command(nspu_desc_t *desc, uint16_t cmd, int read, int write,
-		 void *buffer, size_t rsize, size_t wsize)
-{
-	uint64_t status, cmd_reg;
-	uint64_t offset;
-	int retry = 0;
-	int retries = 120;
-	int ret = 0;
-
-	/* Same expansion BAR is used for different things */
-	nspu_xlate(desc, NSP_BASE, &offset);
-
-	status = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_STATUS));
-
-	while ((status & 0x1) && (retry < retries)) {
-		status = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_STATUS));
-		retry++;
-		sleep(1);
-	}
-
-	if (retry == retries)
-		return -1;
-
-	if (write) {
-		ret = nspu_buff_write(desc, buffer, wsize);
-		if (ret)
-			return ret;
-
-		/* Expansion BAR changes when writing the buffer */
-		nspu_xlate(desc, NSP_BASE, &offset);
-	}
-
-	NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_COMMAND)) =
-		(uint64_t)wsize << 32 | (uint64_t)cmd << 16 | 1;
-
-	retry = 0;
-
-	cmd_reg = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_COMMAND));
-	while ((cmd_reg & 0x1) && (retry < retries)) {
-		cmd_reg = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_COMMAND));
-		retry++;
-		sleep(1);
-	}
-	if (retry == retries)
-		return -1;
-
-	retry = 0;
-	status = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_STATUS));
-	while ((status & 0x1) && (retry < retries)) {
-		status = NSP_REG_VAL(NSP_REG_ADDR(desc, offset, NSP_STATUS));
-		retry++;
-		sleep(1);
-	}
-
-	if (retry == retries)
-		return -1;
-
-	ret = status & (0xff << 8);
-	if (ret)
-		return ret;
-
-	if (read) {
-		ret = nspu_buff_read(desc, buffer, rsize);
-		if (ret)
-			return ret;
-	}
-
-	return ret;
-}
-
-static int
-nfp_fw_reset(nspu_desc_t *nspu_desc)
-{
-	int res;
-
-	res = nspu_command(nspu_desc, NSP_CMD_RESET, 0, 0, 0, 0, 0);
-
-	if (res < 0)
-		RTE_LOG(INFO, PMD, "fw reset failed: error %d", res);
-
-	return res;
-}
-
-#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
-#define DEFAULT_FW_FILENAME   "nic_dpdk_default.nffw"
-
-static int
-nfp_fw_upload(nspu_desc_t *nspu_desc)
-{
-	int fw_f;
-	char *fw_buf;
-	char filename[100];
-	struct stat file_stat;
-	off_t fsize, bytes;
-	ssize_t size;
-	int ret;
-
-	size = nspu_desc->buf_size;
-
-	sprintf(filename, "%s/%s", DEFAULT_FW_PATH, DEFAULT_FW_FILENAME);
-	fw_f = open(filename, O_RDONLY);
-	if (fw_f < 0) {
-		RTE_LOG(INFO, PMD, "Firmware file %s/%s not found.",
-			DEFAULT_FW_PATH, DEFAULT_FW_FILENAME);
-		return -ENOENT;
-	}
-
-	if (fstat(fw_f, &file_stat) < 0) {
-		RTE_LOG(INFO, PMD, "Firmware file %s/%s size is unknown",
-			DEFAULT_FW_PATH, DEFAULT_FW_FILENAME);
-		close(fw_f);
-		return -ENOENT;
-	}
-
-	fsize = file_stat.st_size;
-	RTE_LOG(DEBUG, PMD, "Firmware file with size: %" PRIu64 "\n",
-			    (uint64_t)fsize);
-
-	if (fsize > (off_t)size) {
-		RTE_LOG(INFO, PMD, "fw file too big: %" PRIu64
-				   " bytes (%" PRIu64 " max)",
-				  (uint64_t)fsize, (uint64_t)size);
-		close(fw_f);
-		return -EINVAL;
-	}
-
-	fw_buf = malloc((size_t)size);
-	if (!fw_buf) {
-		RTE_LOG(INFO, PMD, "malloc failed for fw buffer");
-		close(fw_f);
-		return -ENOMEM;
-	}
-	memset(fw_buf, 0, size);
-
-	bytes = read(fw_f, fw_buf, fsize);
-	if (bytes != fsize) {
-		RTE_LOG(INFO, PMD, "Reading fw to buffer failed.\n"
-				   "Just %" PRIu64 " of %" PRIu64 " bytes read.",
-				   (uint64_t)bytes, (uint64_t)fsize);
-		free(fw_buf);
-		close(fw_f);
-		return -EIO;
-	}
-
-	ret = nspu_command(nspu_desc, NSP_CMD_FW_LOAD, 0, 1, fw_buf, 0, bytes);
-
-	free(fw_buf);
-	close(fw_f);
-
-	return ret;
-}
-
-/* Firmware symbol descriptor size */
-#define NFP_SYM_DESC_LEN 40
-
-#define SYMBOL_DATA(b, off)     (*(int64_t *)((b) + (off)))
-#define SYMBOL_UDATA(b, off)     (*(uint64_t *)((b) + (off)))
-
-/* Firmware symbols contain information about how to access what they
- * represent. It can be as simple as an numeric variable declared at a
- * specific NFP memory, but it can also be more complex structures and
- * related to specific hardware functionalities or components. Target,
- * domain and address allow to create the BAR window for accessing such
- * hw object and size defines the length to map.
- *
- * A vNIC is a network interface implemented inside the NFP and using a
- * subset of device PCI BARs. Specific firmware symbols allow to map those
- * vNIC bars by host drivers like the NFP PMD.
- *
- * Accessing what the symbol represents implies to map the access through
- * a PCI BAR window. NFP expansion BARs are used in this regard through
- * the NSPU interface.
- */
-static int
-nfp_nspu_set_bar_from_symbl(nspu_desc_t *desc, const char *symbl,
-			    uint32_t expbar, uint64_t *pcie_offset,
-			    ssize_t *size)
-{
-	int64_t type;
-	int64_t target;
-	int64_t domain;
-	uint64_t addr;
-	char *sym_buf;
-	int ret = 0;
-
-	sym_buf = malloc(desc->buf_size);
-	if (!sym_buf)
-		return -ENOMEM;
-
-	strncpy(sym_buf, symbl, strlen(symbl));
-	ret = nspu_command(desc, NSP_CMD_GET_SYMBOL, 1, 1, sym_buf,
-			   NFP_SYM_DESC_LEN, strlen(symbl));
-	if (ret) {
-		RTE_LOG(DEBUG, PMD, "symbol resolution (%s) failed\n", symbl);
-		goto clean;
-	}
-
-	/* Reading symbol information */
-	type = SYMBOL_DATA(sym_buf, 0);
-	target = SYMBOL_DATA(sym_buf, 8);
-	domain =  SYMBOL_DATA(sym_buf, 16);
-	addr = SYMBOL_UDATA(sym_buf, 24);
-	*size = (ssize_t)SYMBOL_UDATA(sym_buf, 32);
-
-	if (type != 1) {
-		RTE_LOG(INFO, PMD, "wrong symbol type\n");
-		ret = -EINVAL;
-		goto clean;
-	}
-	if (!(target == 7 || target == -7)) {
-		RTE_LOG(INFO, PMD, "wrong symbol target\n");
-		ret = -EINVAL;
-		goto clean;
-	}
-	if (domain == 8 || domain == 9) {
-		RTE_LOG(INFO, PMD, "wrong symbol domain\n");
-		ret = -EINVAL;
-		goto clean;
-	}
-
-	/* Adjusting address based on symbol location */
-	if ((domain >= 24) && (domain < 28) && (target == 7)) {
-		addr = 1ULL << 37 | addr | ((uint64_t)domain & 0x3) << 35;
-	} else {
-		addr = 1ULL << 39 | addr | ((uint64_t)domain & 0x3f) << 32;
-		if (target == -7)
-			target = 7;
-	}
-
-	/* Configuring NFP expansion bar for mapping specific PCI BAR window */
-	nfp_nspu_mem_bar_cfg(desc, expbar, target, addr, pcie_offset);
-
-	/* This is the PCI BAR offset to use by the host */
-	*pcie_offset |= ((expbar & 0x7) << (desc->barsz - 3));
-
-clean:
-	free(sym_buf);
-	return ret;
-}
-
-int
-nfp_nsp_fw_setup(nspu_desc_t *desc, const char *sym, uint64_t *pcie_offset)
-{
-	ssize_t bar0_sym_size;
-
-	/* If the symbol resolution works, it implies a firmware app
-	 * is already there.
-	 */
-	if (!nfp_nspu_set_bar_from_symbl(desc, sym, NFP_NET_PF_CFG_EXP_BAR,
-					 pcie_offset, &bar0_sym_size))
-		return 0;
-
-	/* No firmware app detected or not the right one */
-	RTE_LOG(INFO, PMD, "No firmware detected. Resetting NFP...\n");
-	if (nfp_fw_reset(desc) < 0) {
-		RTE_LOG(ERR, PMD, "nfp fw reset failed\n");
-		return -ENODEV;
-	}
-
-	RTE_LOG(INFO, PMD, "Reset done.\n");
-	RTE_LOG(INFO, PMD, "Uploading firmware...\n");
-
-	if (nfp_fw_upload(desc) < 0) {
-		RTE_LOG(ERR, PMD, "nfp fw upload failed\n");
-		return -ENODEV;
-	}
-
-	RTE_LOG(INFO, PMD, "Done.\n");
-
-	/* Now the symbol should be there */
-	if (nfp_nspu_set_bar_from_symbl(desc, sym, NFP_NET_PF_CFG_EXP_BAR,
-					pcie_offset, &bar0_sym_size)) {
-		RTE_LOG(ERR, PMD, "nfp PF BAR symbol resolution failed\n");
-		return -ENODEV;
-	}
-
-	return 0;
-}
-
-int
-nfp_nsp_map_ctrl_bar(nspu_desc_t *desc, uint64_t *pcie_offset)
-{
-	ssize_t bar0_sym_size;
-
-	if (nfp_nspu_set_bar_from_symbl(desc, "_pf0_net_bar0",
-					NFP_NET_PF_CFG_EXP_BAR,
-					pcie_offset, &bar0_sym_size))
-		return -ENODEV;
-
-	return 0;
-}
-
-/*
- * This is a hardcoded fixed NFP internal CPP bus address for the hw queues unit
- * inside the PCIE island.
- */
-#define NFP_CPP_PCIE_QUEUES ((uint64_t)(1ULL << 39) |  0x80000 | \
-			     ((uint64_t)0x4 & 0x3f) << 32)
-
-/* Configure a specific NFP expansion bar for accessing the vNIC rx/tx BARs */
-void
-nfp_nsp_map_queues_bar(nspu_desc_t *desc, uint64_t *pcie_offset)
-{
-	nfp_nspu_mem_bar_cfg(desc, NFP_NET_PF_HW_QUEUES_EXP_BAR, 0,
-			     NFP_CPP_PCIE_QUEUES, pcie_offset);
-
-	/* This is the pcie offset to use by the host */
-	*pcie_offset |= ((NFP_NET_PF_HW_QUEUES_EXP_BAR & 0x7) << (27 - 3));
-}
-
-int
-nfp_nsp_eth_config(nspu_desc_t *desc, int port, int up)
-{
-	union eth_table_entry *entries, *entry;
-	int modified;
-	int ret, idx;
-	int i;
-
-	idx = port;
-
-	RTE_LOG(INFO, PMD, "Hw ethernet port %d configure...\n", port);
-	rte_spinlock_lock(&desc->nsp_lock);
-	entries = malloc(NSP_ETH_TABLE_SIZE);
-	if (!entries) {
-		rte_spinlock_unlock(&desc->nsp_lock);
-		return -ENOMEM;
-	}
-
-	ret = nspu_command(desc, NSP_CMD_READ_ETH_TABLE, 1, 0, entries,
-			   NSP_ETH_TABLE_SIZE, 0);
-	if (ret) {
-		rte_spinlock_unlock(&desc->nsp_lock);
-		free(entries);
-		return ret;
-	}
-
-	entry = entries;
-
-	for (i = 0; i < NSP_ETH_MAX_COUNT; i++) {
-		/* ports in use do not appear sequentially in the table */
-		if (!(entry->port & NSP_ETH_PORT_LANES_MASK)) {
-			/* entry not in use */
-			entry++;
-			continue;
-		}
-		if (idx == 0)
-			break;
-		idx--;
-		entry++;
-	}
-
-	if (i == NSP_ETH_MAX_COUNT) {
-		rte_spinlock_unlock(&desc->nsp_lock);
-		free(entries);
-		return -EINVAL;
-	}
-
-	if (up && !(entry->state & NSP_ETH_STATE_CONFIGURED)) {
-		entry->control |= NSP_ETH_STATE_CONFIGURED;
-		modified = 1;
-	}
-
-	if (!up && (entry->state & NSP_ETH_STATE_CONFIGURED)) {
-		entry->control &= ~NSP_ETH_STATE_CONFIGURED;
-		modified = 1;
-	}
-
-	if (modified) {
-		ret = nspu_command(desc, NSP_CMD_WRITE_ETH_TABLE, 0, 1, entries,
-				   0, NSP_ETH_TABLE_SIZE);
-		if (!ret)
-			RTE_LOG(INFO, PMD,
-				"Hw ethernet port %d configure done\n", port);
-		else
-			RTE_LOG(INFO, PMD,
-				"Hw ethernet port %d configure failed\n", port);
-	}
-	rte_spinlock_unlock(&desc->nsp_lock);
-	free(entries);
-	return ret;
-}
-
-int
-nfp_nsp_eth_read_table(nspu_desc_t *desc, union eth_table_entry **table)
-{
-	int ret;
-
-	if (!table)
-		return -EINVAL;
-
-	RTE_LOG(INFO, PMD, "Reading hw ethernet table...\n");
-
-	/* port 0 allocates the eth table and read it using NSPU */
-	*table = malloc(NSP_ETH_TABLE_SIZE);
-	if (!*table)
-		return -ENOMEM;
-
-	ret = nspu_command(desc, NSP_CMD_READ_ETH_TABLE, 1, 0, *table,
-			   NSP_ETH_TABLE_SIZE, 0);
-	if (ret)
-		return ret;
-
-	RTE_LOG(INFO, PMD, "Done\n");
-
-	return 0;
-}
diff --git a/drivers/net/nfp/nfp_nspu.h b/drivers/net/nfp/nfp_nspu.h
deleted file mode 100644
index 8c33835..0000000
--- a/drivers/net/nfp/nfp_nspu.h
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2017 Netronome Systems, Inc.
- * 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 copyright holder 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 HOLDER 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.
- */
-
-/*
- * vim:shiftwidth=8:noexpandtab
- *
- * @file dpdk/pmd/nfp_nspu.h
- *
- * Netronome NFP_NET PDM driver
- */
-
-/*
- * NSP is the NFP Service Processor. NSPU is NSP Userspace interface.
- *
- * NFP NSP helps with firmware/hardware configuration. NSP is another component
- * in NFP programmable processor and accessing it from host requires to firstly
- * configure a specific NFP PCI expansion BAR.
- *
- * Once access is ready, configuration can be done reading and writing
- * from/to a specific PF PCI BAR window. This same interface will allow to
- * create other PCI BAR windows for accessing other NFP components.
- *
- * This file includes low-level functions, using the NSPU interface, and high
- * level functions, invoked by the PMD for using NSP services. This allows
- * firmware upload, vNIC PCI BARs mapping and other low-level configurations
- * like link setup.
- *
- * NSP access is done during initialization and it is not involved at all with
- * the fast path.
- */
-
-#include <rte_spinlock.h>
-#include "nfp_net_eth.h"
-
-typedef struct {
-	int nfp;        /* NFP device */
-	int pcie_bar;   /* PF PCI BAR to work with */
-	int exp_bar;    /* Expansion BAR number used by NSPU */
-	int barsz;      /* PCIE BAR log2 size */
-	uint64_t bufaddr;  /* commands buffer address */
-	size_t buf_size;   /* commands buffer size */
-	uint64_t windowsz; /* NSPU BAR window size */
-	void *cfg_base; /* Expansion BARs address */
-	void *mem_base; /* NSP interface */
-	rte_spinlock_t nsp_lock;
-} nspu_desc_t;
-
-int nfp_nspu_init(nspu_desc_t *desc, int nfp, int pcie_bar, size_t pcie_barsz,
-		  int exp_bar, void *exp_bar_cfg_base, void *exp_bar_mmap);
-int nfp_nsp_get_abi_version(nspu_desc_t *desc, int *major, int *minor);
-int nfp_nsp_fw_setup(nspu_desc_t *desc, const char *sym, uint64_t *pcie_offset);
-int nfp_nsp_map_ctrl_bar(nspu_desc_t *desc, uint64_t *pcie_offset);
-void nfp_nsp_map_queues_bar(nspu_desc_t *desc, uint64_t *pcie_offset);
-int nfp_nsp_eth_config(nspu_desc_t *desc, int port, int up);
-int nfp_nsp_eth_read_table(nspu_desc_t *desc, union eth_table_entry **table);
-- 
1.9.1

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support Alejandro Lucero
@ 2018-03-30 10:37   ` Ferruh Yigit
  2018-03-31 17:04     ` Alejandro Lucero
  0 siblings, 1 reply; 12+ messages in thread
From: Ferruh Yigit @ 2018-03-30 10:37 UTC (permalink / raw)
  To: Alejandro Lucero, dev; +Cc: Hemant Agrawal

On 3/23/2018 5:35 PM, Alejandro Lucero wrote:
> CPP refers to the internal NFP Command Push Pull bus. This patch allows
> to create CPP commands from user space allowing to access any single
> part of the chip.
> 
> This CPP interface is the base for having other functionalities like
> mutexes when accessing specific chip components, chip resources management,
> firmware upload or using the NSP, an embedded arm processor which can
> perform tasks on demand.
> 
> NSP was the previous only way for doing things in the chip by the PMD,
> where a NSPU interface was used for commands like firmware upload or
> port link configuration. CPP interface supersedes NSPU, but it is still
> possible to use NSP through CPP.
> 
> CPP interface adds a great flexibility for doing things like extended
> stats, firmware debugging or selecting properly the firmware file to
> upload.

It can be good to announce this new feature in release notes, what do you think?

Also get some build errors [1], one is about zlib.h, normally we disable PMDs by
default with external dependencies, is nfp depends to libz with this patch?


[1]
...dpdk/drivers/net/nfp/nfpcore/nfp_mutex.c:295:28: error: format specifies type
'short' but the argument has type 'unsigned int' [-Werror,-Wformat]
                        printf("\tusage:%hd\n", mutex->usage);

                                        ~~~     ^~~~~~~~~~~~


                                        %u

...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:819:1: error: unused function
'__nfp_bytemask_of' [-Werror,-Wunused-function]
__nfp_bytemask_of(int width, uint64_t addr)


^

...dpdk/drivers/net/nfp/nfp_net.c:2971:35: error: format specifies type
'unsigned char' but the argument has type 'uint32_t' (aka 'unsigned int')
[-Werror,-Wformat]
                cpp->serial[4], cpp->serial[5], cpp->interface >> 8,


                                                ^~~~~~~~~~~~~~~~~~~
...dpdk/drivers/net/nfp/nfp_net.c:2972:3: error: format specifies type 'unsigned
char' but the argument has type 'unsigned int' [-Werror,-Wformat]
                cpp->interface & 0xff);
                ^~~~~~~~~~~~~~~~~~~~~

...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readl’:
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:459:9: error: ‘tmp’ may be used
uninitialized in this function [-Werror=maybe-uninitialized]
  *value = rte_le_to_cpu_32(tmp);

...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readq’:
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:483:9: error: ‘tmp’ may be used
uninitialized in this function [-Werror=maybe-uninitialized]
  *value = rte_le_to_cpu_64(tmp);

...dpdk/drivers/net/nfp/nfpcore/nfp_resource.c:34:10: fatal error: zlib.h: No
such file or directory
 #include <zlib.h>
          ^~~~~~~~

> 
> Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>

<...>

> diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
> new file mode 100644
> index 0000000..fbeec57
> --- /dev/null
> +++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
> @@ -0,0 +1,748 @@
> +/*
> + * Copyright (c) 2018 Netronome Systems, Inc.
> + * 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 copyright holder 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 HOLDER 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.
> + */

Can you please add new files with SPDX tags?

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH 3/4] doc: update NFP guide
  2018-03-23 17:35 ` [dpdk-dev] [PATCH 3/4] doc: update NFP guide Alejandro Lucero
@ 2018-03-30 10:37   ` Ferruh Yigit
  2018-03-31 17:04     ` Alejandro Lucero
  0 siblings, 1 reply; 12+ messages in thread
From: Ferruh Yigit @ 2018-03-30 10:37 UTC (permalink / raw)
  To: Alejandro Lucero, dev

On 3/23/2018 5:35 PM, Alejandro Lucero wrote:
> New CPP interface changes the way firmware upload is managed by
> the PMD. It also supports different firmware file names for
> having specific firmware aplications per card.
> 
> Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>

This conflicts with another nfp doc patch applied in next-net, can you please
rebase this on latest next-net?

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support
  2018-03-30 10:37   ` Ferruh Yigit
@ 2018-03-31 17:04     ` Alejandro Lucero
  2018-04-03 11:09       ` Ferruh Yigit
  0 siblings, 1 reply; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-31 17:04 UTC (permalink / raw)
  To: Ferruh Yigit; +Cc: dev, Hemant Agrawal

On Fri, Mar 30, 2018 at 11:37 AM, Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

> On 3/23/2018 5:35 PM, Alejandro Lucero wrote:
> > CPP refers to the internal NFP Command Push Pull bus. This patch allows
> > to create CPP commands from user space allowing to access any single
> > part of the chip.
> >
> > This CPP interface is the base for having other functionalities like
> > mutexes when accessing specific chip components, chip resources
> management,
> > firmware upload or using the NSP, an embedded arm processor which can
> > perform tasks on demand.
> >
> > NSP was the previous only way for doing things in the chip by the PMD,
> > where a NSPU interface was used for commands like firmware upload or
> > port link configuration. CPP interface supersedes NSPU, but it is still
> > possible to use NSP through CPP.
> >
> > CPP interface adds a great flexibility for doing things like extended
> > stats, firmware debugging or selecting properly the firmware file to
> > upload.
>
> It can be good to announce this new feature in release notes, what do you
> think?
>
>
Uhmm, not sure about this. It is specific to the NFP PMD and I do not see
it as a new feature, at least DPDK users will not be aware of it.



> Also get some build errors [1], one is about zlib.h, normally we disable
> PMDs by
> default with external dependencies, is nfp depends to libz with this patch?
>
>
I forgot to remove this reference. It is not needed now, but it was with
the initial internal work.

About the other build errors, I do not get them and I have used a couple of
different systems, Ubuntu and Redhat. This is, of course, a serious
concern. Can you give me more information about the system you are using?

I remember I got some build error with other patches, from automatic builds
made just after those patches were sent. I did not get any this time, just
those warning for checkpatch which I was aware of. Is this automatic build
not happening any more?



> [1]
> ...dpdk/drivers/net/nfp/nfpcore/nfp_mutex.c:295:28: error: format
> specifies type
> 'short' but the argument has type 'unsigned int' [-Werror,-Wformat]
>                         printf("\tusage:%hd\n", mutex->usage);
>
>                                         ~~~     ^~~~~~~~~~~~
>
>
>                                         %u
>
> ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:819:1: error: unused
> function
> '__nfp_bytemask_of' [-Werror,-Wunused-function]
> __nfp_bytemask_of(int width, uint64_t addr)
>
>
> ^
>
> ...dpdk/drivers/net/nfp/nfp_net.c:2971:35: error: format specifies type
> 'unsigned char' but the argument has type 'uint32_t' (aka 'unsigned int')
> [-Werror,-Wformat]
>                 cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
>
>
>                                                 ^~~~~~~~~~~~~~~~~~~
> ...dpdk/drivers/net/nfp/nfp_net.c:2972:3: error: format specifies type
> 'unsigned
> char' but the argument has type 'unsigned int' [-Werror,-Wformat]
>                 cpp->interface & 0xff);
>                 ^~~~~~~~~~~~~~~~~~~~~
>
> ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function
> ‘nfp_cpp_area_readl’:
> ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:459:9: error: ‘tmp’ may be
> used
> uninitialized in this function [-Werror=maybe-uninitialized]
>   *value = rte_le_to_cpu_32(tmp);
>
> ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function
> ‘nfp_cpp_area_readq’:
> ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:483:9: error: ‘tmp’ may be
> used
> uninitialized in this function [-Werror=maybe-uninitialized]
>   *value = rte_le_to_cpu_64(tmp);
>
> ...dpdk/drivers/net/nfp/nfpcore/nfp_resource.c:34:10: fatal error:
> zlib.h: No
> such file or directory
>  #include <zlib.h>
>           ^~~~~~~~
>
> >
> > Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
>
> <...>
>
> > diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
> b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
> > new file mode 100644
> > index 0000000..fbeec57
> > --- /dev/null
> > +++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
> > @@ -0,0 +1,748 @@
> > +/*
> > + * Copyright (c) 2018 Netronome Systems, Inc.
> > + * 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 copyright holder 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 HOLDER 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.
> > + */
>
> Can you please add new files with SPDX tags?
>

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH 3/4] doc: update NFP guide
  2018-03-30 10:37   ` Ferruh Yigit
@ 2018-03-31 17:04     ` Alejandro Lucero
  0 siblings, 0 replies; 12+ messages in thread
From: Alejandro Lucero @ 2018-03-31 17:04 UTC (permalink / raw)
  To: Ferruh Yigit; +Cc: dev

On Fri, Mar 30, 2018 at 11:37 AM, Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

> On 3/23/2018 5:35 PM, Alejandro Lucero wrote:
> > New CPP interface changes the way firmware upload is managed by
> > the PMD. It also supports different firmware file names for
> > having specific firmware aplications per card.
> >
> > Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
>
> This conflicts with another nfp doc patch applied in next-net, can you
> please
> rebase this on latest next-net?
>

Yes, sure.

Thanks

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support
  2018-03-31 17:04     ` Alejandro Lucero
@ 2018-04-03 11:09       ` Ferruh Yigit
  2018-04-03 12:54         ` Ferruh Yigit
  0 siblings, 1 reply; 12+ messages in thread
From: Ferruh Yigit @ 2018-04-03 11:09 UTC (permalink / raw)
  To: Alejandro Lucero; +Cc: dev, Hemant Agrawal

On 3/31/2018 6:04 PM, Alejandro Lucero wrote:
> 
> 
> On Fri, Mar 30, 2018 at 11:37 AM, Ferruh Yigit <ferruh.yigit@intel.com
> <mailto:ferruh.yigit@intel.com>> wrote:
> 
>     On 3/23/2018 5:35 PM, Alejandro Lucero wrote:
>     > CPP refers to the internal NFP Command Push Pull bus. This patch allows
>     > to create CPP commands from user space allowing to access any single
>     > part of the chip.
>     >
>     > This CPP interface is the base for having other functionalities like
>     > mutexes when accessing specific chip components, chip resources management,
>     > firmware upload or using the NSP, an embedded arm processor which can
>     > perform tasks on demand.
>     >
>     > NSP was the previous only way for doing things in the chip by the PMD,
>     > where a NSPU interface was used for commands like firmware upload or
>     > port link configuration. CPP interface supersedes NSPU, but it is still
>     > possible to use NSP through CPP.
>     >
>     > CPP interface adds a great flexibility for doing things like extended
>     > stats, firmware debugging or selecting properly the firmware file to
>     > upload.
> 
>     It can be good to announce this new feature in release notes, what do you think?
> 
> 
> Uhmm, not sure about this. It is specific to the NFP PMD and I do not see it as
> a new feature, at least DPDK users will not be aware of it.

OK, if this internal and you think user won't be interested in this update.

> 
>  
> 
>     Also get some build errors [1], one is about zlib.h, normally we disable PMDs by
>     default with external dependencies, is nfp depends to libz with this patch?
> 
> 
> I forgot to remove this reference. It is not needed now, but it was with the
> initial internal work.
> 
> About the other build errors, I do not get them and I have used a couple of
> different systems, Ubuntu and Redhat. This is, of course, a serious concern. Can
> you give me more information about the system you are using?  
> 
> I remember I got some build error with other patches, from automatic builds made
> just after those patches were sent. I did not get any this time, just those
> warning for checkpatch which I was aware of. Is this automatic build not
> happening any more?

I will share more details. Automatic build still works and it has wider
coverage, this is from my local script...

> 
> 
> 
>     [1]
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_mutex.c:295:28: error: format specifies type
>     'short' but the argument has type 'unsigned int' [-Werror,-Wformat]
>                             printf("\tusage:%hd\n", mutex->usage);
> 
>                                             ~~~     ^~~~~~~~~~~~
> 
> 
>                                             %u
> 
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:819:1: error: unused function
>     '__nfp_bytemask_of' [-Werror,-Wunused-function]
>     __nfp_bytemask_of(int width, uint64_t addr)
> 
> 
>     ^
> 
>     ...dpdk/drivers/net/nfp/nfp_net.c:2971:35: error: format specifies type
>     'unsigned char' but the argument has type 'uint32_t' (aka 'unsigned int')
>     [-Werror,-Wformat]
>                     cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
> 
> 
>                                                     ^~~~~~~~~~~~~~~~~~~
>     ...dpdk/drivers/net/nfp/nfp_net.c:2972:3: error: format specifies type 'unsigned
>     char' but the argument has type 'unsigned int' [-Werror,-Wformat]
>                     cpp->interface & 0xff);
>                     ^~~~~~~~~~~~~~~~~~~~~
> 
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readl’:
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:459:9: error: ‘tmp’ may be used
>     uninitialized in this function [-Werror=maybe-uninitialized]
>       *value = rte_le_to_cpu_32(tmp);
> 
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readq’:
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:483:9: error: ‘tmp’ may be used
>     uninitialized in this function [-Werror=maybe-uninitialized]
>       *value = rte_le_to_cpu_64(tmp);
> 
>     ...dpdk/drivers/net/nfp/nfpcore/nfp_resource.c:34:10: fatal error: zlib.h: No
>     such file or directory
>      #include <zlib.h>
>               ^~~~~~~~
> 
>     >
>     > Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com
>     <mailto:alejandro.lucero@netronome.com>>
> 
>     <...>
> 
>     > diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
>     b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
>     > new file mode 100644
>     > index 0000000..fbeec57
>     > --- /dev/null
>     > +++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
>     > @@ -0,0 +1,748 @@
>     > +/*
>     > + * Copyright (c) 2018 Netronome Systems, Inc.
>     > + * 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 copyright holder 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 HOLDER 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.
>     > + */
> 
>     Can you please add new files with SPDX tags?
> 
> 

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support
  2018-04-03 11:09       ` Ferruh Yigit
@ 2018-04-03 12:54         ` Ferruh Yigit
  0 siblings, 0 replies; 12+ messages in thread
From: Ferruh Yigit @ 2018-04-03 12:54 UTC (permalink / raw)
  To: Alejandro Lucero; +Cc: dev, Hemant Agrawal

On 4/3/2018 12:09 PM, Ferruh Yigit wrote:
> On 3/31/2018 6:04 PM, Alejandro Lucero wrote:
>>
>>
>> On Fri, Mar 30, 2018 at 11:37 AM, Ferruh Yigit <ferruh.yigit@intel.com
>> <mailto:ferruh.yigit@intel.com>> wrote:
>>
>>     On 3/23/2018 5:35 PM, Alejandro Lucero wrote:
>>     > CPP refers to the internal NFP Command Push Pull bus. This patch allows
>>     > to create CPP commands from user space allowing to access any single
>>     > part of the chip.
>>     >
>>     > This CPP interface is the base for having other functionalities like
>>     > mutexes when accessing specific chip components, chip resources management,
>>     > firmware upload or using the NSP, an embedded arm processor which can
>>     > perform tasks on demand.
>>     >
>>     > NSP was the previous only way for doing things in the chip by the PMD,
>>     > where a NSPU interface was used for commands like firmware upload or
>>     > port link configuration. CPP interface supersedes NSPU, but it is still
>>     > possible to use NSP through CPP.
>>     >
>>     > CPP interface adds a great flexibility for doing things like extended
>>     > stats, firmware debugging or selecting properly the firmware file to
>>     > upload.
>>
>>     It can be good to announce this new feature in release notes, what do you think?
>>
>>
>> Uhmm, not sure about this. It is specific to the NFP PMD and I do not see it as
>> a new feature, at least DPDK users will not be aware of it.
> 
> OK, if this internal and you think user won't be interested in this update.
> 
>>
>>  
>>
>>     Also get some build errors [1], one is about zlib.h, normally we disable PMDs by
>>     default with external dependencies, is nfp depends to libz with this patch?
>>
>>
>> I forgot to remove this reference. It is not needed now, but it was with the
>> initial internal work.
>>
>> About the other build errors, I do not get them and I have used a couple of
>> different systems, Ubuntu and Redhat. This is, of course, a serious concern. Can
>> you give me more information about the system you are using?  
>>
>> I remember I got some build error with other patches, from automatic builds made
>> just after those patches were sent. I did not get any this time, just those
>> warning for checkpatch which I was aware of. Is this automatic build not
>> happening any more?
> 
> I will share more details. Automatic build still works and it has wider
> coverage, this is from my local script...

[1] seems valid and for all.

[2] #include <zlib.h> seems for all.

[3] clang


[1]
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readl’:
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:459:9: error: ‘tmp’ may be used
uninitialized in this function [-Werror=maybe-uninitialized]
  *value = rte_le_to_cpu_32(tmp);
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readq’:
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:483:9: error: ‘tmp’ may be used
uninitialized in this function [-Werror=maybe-uninitialized]
  *value = rte_le_to_cpu_64(tmp);

[2]
...dpdk/drivers/net/nfp/nfpcore/nfp_resource.c:34:10: fatal error: zlib.h: No
such file or directory
 #include <zlib.h>
          ^~~~~~~~
compilation terminated.
make[7]: *** [...dpdk/mk/internal/rte.compile-pre.mk:116: nfp_resource.o] Error 1
make[7]: *** Waiting for unfinished jobs....
...dpdk/drivers/net/nfp/nfpcore/nfp_hwinfo.c:47:10: fatal error: zlib.h: No such
file or directory
 #include <zlib.h>
          ^~~~~~~~

[3]
...dpdk/drivers/net/nfp/nfpcore/nfp_mutex.c:295:28: error: format specifies type
'short' but the argument has type 'unsigned int' [-Werror,-Wformat]
                        printf("\tusage:%hd\n", mutex->usage);
                                        ~~~     ^~~~~~~~~~~~
                                        %u
1 error generated.
make[7]: *** [...dpdk/mk/internal/rte.compile-pre.mk:114: nfp_mutex.o] Error 1
make[7]: *** Waiting for unfinished jobs....
...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:819:1: error: unused function
'__nfp_bytemask_of' [-Werror,-Wunused-function]
__nfp_bytemask_of(int width, uint64_t addr)
^
1 error generated.
make[7]: *** [...dpdk/mk/internal/rte.compile-pre.mk:114: nfp_cppcore.o] Error 1
...dpdk/drivers/net/nfp/nfp_net.c:2970:35: error: format specifies type
'unsigned char' but the argument has type 'uint32_t' (aka 'unsigned int')
[-Werror,-Wformat]
                cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
                                                ^~~~~~~~~~~~~~~~~~~
...dpdk/drivers/net/nfp/nfp_net.c:2971:3: error: format specifies type 'unsigned
char' but the argument has type 'unsigned int' [-Werror,-Wformat]
                cpp->interface & 0xff);
                ^~~~~~~~~~~~~~~~~~~~~
2 errors generated.

> 
>>
>>
>>
>>     [1]
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_mutex.c:295:28: error: format specifies type
>>     'short' but the argument has type 'unsigned int' [-Werror,-Wformat]
>>                             printf("\tusage:%hd\n", mutex->usage);
>>
>>                                             ~~~     ^~~~~~~~~~~~
>>
>>
>>                                             %u
>>
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:819:1: error: unused function
>>     '__nfp_bytemask_of' [-Werror,-Wunused-function]
>>     __nfp_bytemask_of(int width, uint64_t addr)
>>
>>
>>     ^
>>
>>     ...dpdk/drivers/net/nfp/nfp_net.c:2971:35: error: format specifies type
>>     'unsigned char' but the argument has type 'uint32_t' (aka 'unsigned int')
>>     [-Werror,-Wformat]
>>                     cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
>>
>>
>>                                                     ^~~~~~~~~~~~~~~~~~~
>>     ...dpdk/drivers/net/nfp/nfp_net.c:2972:3: error: format specifies type 'unsigned
>>     char' but the argument has type 'unsigned int' [-Werror,-Wformat]
>>                     cpp->interface & 0xff);
>>                     ^~~~~~~~~~~~~~~~~~~~~
>>
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readl’:
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:459:9: error: ‘tmp’ may be used
>>     uninitialized in this function [-Werror=maybe-uninitialized]
>>       *value = rte_le_to_cpu_32(tmp);
>>
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c: In function ‘nfp_cpp_area_readq’:
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_cppcore.c:483:9: error: ‘tmp’ may be used
>>     uninitialized in this function [-Werror=maybe-uninitialized]
>>       *value = rte_le_to_cpu_64(tmp);
>>
>>     ...dpdk/drivers/net/nfp/nfpcore/nfp_resource.c:34:10: fatal error: zlib.h: No
>>     such file or directory
>>      #include <zlib.h>
>>               ^~~~~~~~
>>
>>     >
>>     > Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com
>>     <mailto:alejandro.lucero@netronome.com>>
>>
>>     <...>
>>
>>     > diff --git a/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
>>     b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
>>     > new file mode 100644
>>     > index 0000000..fbeec57
>>     > --- /dev/null
>>     > +++ b/drivers/net/nfp/nfpcore/nfp-common/nfp_cppat.h
>>     > @@ -0,0 +1,748 @@
>>     > +/*
>>     > + * Copyright (c) 2018 Netronome Systems, Inc.
>>     > + * 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 copyright holder 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 HOLDER 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.
>>     > + */
>>
>>     Can you please add new files with SPDX tags?
>>
>>
> 

^ permalink raw reply	[flat|nested] 12+ messages in thread

* [dpdk-dev] [PATCH 3/4] doc: update NFP guide
  2018-04-05 14:28 [dpdk-dev] [PATCH v2 0/4] NFP PF support based on new CPP interface Alejandro Lucero
@ 2018-04-05 14:28 ` Alejandro Lucero
  0 siblings, 0 replies; 12+ messages in thread
From: Alejandro Lucero @ 2018-04-05 14:28 UTC (permalink / raw)
  To: dev

New CPP interface changes the way firmware upload is managed by
the PMD. It also supports different firmware file names for
having specific firmware aplications per card.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 doc/guides/nics/nfp.rst | 34 ++++++++++++++++++++++++++--------
 1 file changed, 26 insertions(+), 8 deletions(-)

diff --git a/doc/guides/nics/nfp.rst b/doc/guides/nics/nfp.rst
index 67e574e..927c03c 100644
--- a/doc/guides/nics/nfp.rst
+++ b/doc/guides/nics/nfp.rst
@@ -99,17 +99,35 @@ bound to ``igb_uio`` or ``vfio-pci`` kernel drivers. Future DPDK versions will
 have a PMD able to work with the PF and VFs at the same time and with the PF
 implementing VF management along with other PF-only functionalities/offloads.
 
-The PMD PF has extra work to do which will delay the DPDK app initialization.
-This additional effort could be checking if a firmware is already available in
-the device, uploading the firmware if necessary or configuring the Link state
-properly when starting or stopping a PF port. Note that firmware upload is not
-always necessary which is the main delay for NFP PF PMD initialization.
+The PMD PF has extra work to do which will delay the DPDK app initialization
+like uploading the firmware and configure the Link state properly when starting or
+stopping a PF port. Since DPDK 18.05 the firmware upload happens when
+a PF is initialized, which was not always true with older DPDK versions.
 
 Depending on the Netronome product installed in the system, firmware files
 should be available under ``/lib/firmware/netronome``. DPDK PMD supporting the
-PF requires a specific link, ``/lib/firmware/netronome/nic_dpdk_default.nffw``,
-which should be created automatically with Netronome's Agilio products
-installation.
+PF looks for a firmware file in this order:
+
+	1) First try to find a firmware image specific for this device using the
+	   NFP serial number:
+
+		serial-00-15-4d-12-20-65-10-ff.nffw
+
+	2) Then try the PCI name:
+
+		pci-0000:04:00.0.nffw
+
+	3) Finally try the card type and media:
+
+		nic_AMDA0099-0001_2x25.nffw
+
+Netronome's software packages install firmware files under ``/lib/firmware/netronome``
+to support all the Netronome's SmartNICs and different firmware applications.
+This is usually done using file names based on SmartNIC type and media and with a
+directory per firmware application. Options 1 and 2 for firmware filenames allow
+more than one SmartNIC, same type of SmartNIC or different ones, and to upload a
+different firmware to each SmartNIC.
+
 
 PF multiport support
 --------------------
-- 
1.9.1

^ permalink raw reply	[flat|nested] 12+ messages in thread

end of thread, other threads:[~2018-04-05 14:29 UTC | newest]

Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-03-23 17:35 [dpdk-dev] [PATCH 0/4] NFP PF support based on new CPP interface Alejandro Lucero
2018-03-23 17:35 ` [dpdk-dev] [PATCH 1/4] net/nfp: add NFP CPP support Alejandro Lucero
2018-03-30 10:37   ` Ferruh Yigit
2018-03-31 17:04     ` Alejandro Lucero
2018-04-03 11:09       ` Ferruh Yigit
2018-04-03 12:54         ` Ferruh Yigit
2018-03-23 17:35 ` [dpdk-dev] [PATCH 2/4] net/nfp: update PMD for using new CPP interface Alejandro Lucero
2018-03-23 17:35 ` [dpdk-dev] [PATCH 3/4] doc: update NFP guide Alejandro Lucero
2018-03-30 10:37   ` Ferruh Yigit
2018-03-31 17:04     ` Alejandro Lucero
2018-03-23 17:35 ` [dpdk-dev] [PATCH 4/4] net/nfp: remove files Alejandro Lucero
2018-04-05 14:28 [dpdk-dev] [PATCH v2 0/4] NFP PF support based on new CPP interface Alejandro Lucero
2018-04-05 14:28 ` [dpdk-dev] [PATCH 3/4] doc: update NFP guide Alejandro Lucero

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).