New upstream version 17.11-rc3

Change-Id: I6a5baa40612fe0c20f30b5fa773a6cbbac63a685
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>

9 files changed, 1964 insertions, 0 deletions

diff --git a/drivers/mempool/octeontx/Makefile b/drivers/mempool/octeontx/Makefile
new file mode 100644
index 00000000..a2e2863c
--- /dev/null
+++ b/drivers/mempool/octeontx/Makefile
@@ -0,0 +1,68 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2017 Cavium 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:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Cavium Networks nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_mempool_octeontx.a
+
+CFLAGS += $(WERROR_FLAGS)
+EXPORT_MAP := rte_mempool_octeontx_version.map
+
+LIBABIVER := 1
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += octeontx_ssovf.c
+SRCS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += octeontx_mbox.c
+SRCS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += octeontx_fpavf.c
+SRCS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += rte_mempool_octeontx.c
+
+ifeq ($(CONFIG_RTE_TOOLCHAIN_GCC),y)
+CFLAGS_rte_mempool_octeontx.o += -fno-prefetch-loop-arrays
+
+ifeq ($(shell test $(GCC_VERSION) -ge 46 && echo 1), 1)
+CFLAGS_rte_mempool_octeontx.o += -Ofast
+else
+CFLAGS_rte_mempool_octeontx.o += -O3 -ffast-math
+endif
+
+else
+CFLAGS_rte_mempool_octeontx.o += -Ofast
+endif
+
+LDLIBS += -lrte_eal -lrte_mempool -lrte_ring -lrte_mbuf
+LDLIBS += -lrte_bus_pci
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/mempool/octeontx/octeontx_fpavf.c b/drivers/mempool/octeontx/octeontx_fpavf.c
new file mode 100644
index 00000000..3bc50f35
--- /dev/null
+++ b/drivers/mempool/octeontx/octeontx_fpavf.c
@@ -0,0 +1,830 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) Cavium Inc. 2017. All Right reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Cavium networks nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/mman.h>
+
+#include <rte_atomic.h>
+#include <rte_eal.h>
+#include <rte_bus_pci.h>
+#include <rte_errno.h>
+#include <rte_memory.h>
+#include <rte_malloc.h>
+#include <rte_spinlock.h>
+#include <rte_mbuf.h>
+
+#include "octeontx_mbox.h"
+#include "octeontx_fpavf.h"
+
+/* FPA Mbox Message */
+#define IDENTIFY		0x0
+
+#define FPA_CONFIGSET		0x1
+#define FPA_CONFIGGET		0x2
+#define FPA_START_COUNT		0x3
+#define FPA_STOP_COUNT		0x4
+#define FPA_ATTACHAURA		0x5
+#define FPA_DETACHAURA		0x6
+#define FPA_SETAURALVL		0x7
+#define FPA_GETAURALVL		0x8
+
+#define FPA_COPROC		0x1
+
+/* fpa mbox struct */
+struct octeontx_mbox_fpa_cfg {
+	int		aid;
+	uint64_t	pool_cfg;
+	uint64_t	pool_stack_base;
+	uint64_t	pool_stack_end;
+	uint64_t	aura_cfg;
+};
+
+struct __attribute__((__packed__)) gen_req {
+	uint32_t	value;
+};
+
+struct __attribute__((__packed__)) idn_req {
+	uint8_t	domain_id;
+};
+
+struct __attribute__((__packed__)) gen_resp {
+	uint16_t	domain_id;
+	uint16_t	vfid;
+};
+
+struct __attribute__((__packed__)) dcfg_resp {
+	uint8_t	sso_count;
+	uint8_t	ssow_count;
+	uint8_t	fpa_count;
+	uint8_t	pko_count;
+	uint8_t	tim_count;
+	uint8_t	net_port_count;
+	uint8_t	virt_port_count;
+};
+
+#define FPA_MAX_POOL	32
+#define FPA_PF_PAGE_SZ	4096
+
+#define FPA_LN_SIZE	128
+#define FPA_ROUND_UP(x, size) \
+	((((unsigned long)(x)) + size-1) & (~(size-1)))
+#define FPA_OBJSZ_2_CACHE_LINE(sz)	(((sz) + RTE_CACHE_LINE_MASK) >> 7)
+#define FPA_CACHE_LINE_2_OBJSZ(sz)	((sz) << 7)
+
+#define POOL_ENA			(0x1 << 0)
+#define POOL_DIS			(0x0 << 0)
+#define POOL_SET_NAT_ALIGN		(0x1 << 1)
+#define POOL_DIS_NAT_ALIGN		(0x0 << 1)
+#define POOL_STYPE(x)			(((x) & 0x1) << 2)
+#define POOL_LTYPE(x)			(((x) & 0x3) << 3)
+#define POOL_BUF_OFFSET(x)		(((x) & 0x7fffULL) << 16)
+#define POOL_BUF_SIZE(x)		(((x) & 0x7ffULL) << 32)
+
+struct fpavf_res {
+	void		*pool_stack_base;
+	void		*bar0;
+	uint64_t	stack_ln_ptr;
+	uint16_t	domain_id;
+	uint16_t	vf_id;	/* gpool_id */
+	uint16_t	sz128;	/* Block size in cache lines */
+	bool		is_inuse;
+};
+
+struct octeontx_fpadev {
+	rte_spinlock_t lock;
+	uint8_t	total_gpool_cnt;
+	struct fpavf_res pool[FPA_VF_MAX];
+};
+
+static struct octeontx_fpadev fpadev;
+
+/* lock is taken by caller */
+static int
+octeontx_fpa_gpool_alloc(unsigned int object_size)
+{
+	struct fpavf_res *res = NULL;
+	uint16_t gpool;
+	unsigned int sz128;
+
+	sz128 = FPA_OBJSZ_2_CACHE_LINE(object_size);
+
+	for (gpool = 0; gpool < FPA_VF_MAX; gpool++) {
+
+		/* Skip VF that is not mapped Or _inuse */
+		if ((fpadev.pool[gpool].bar0 == NULL) ||
+		    (fpadev.pool[gpool].is_inuse == true))
+			continue;
+
+		res = &fpadev.pool[gpool];
+
+		RTE_ASSERT(res->domain_id != (uint16_t)~0);
+		RTE_ASSERT(res->vf_id != (uint16_t)~0);
+		RTE_ASSERT(res->stack_ln_ptr != 0);
+
+		if (res->sz128 == 0) {
+			res->sz128 = sz128;
+
+			fpavf_log_dbg("gpool %d blk_sz %d\n", gpool, sz128);
+			return gpool;
+		}
+	}
+
+	return -ENOSPC;
+}
+
+/* lock is taken by caller */
+static __rte_always_inline uintptr_t
+octeontx_fpa_gpool2handle(uint16_t gpool)
+{
+	struct fpavf_res *res = NULL;
+
+	RTE_ASSERT(gpool < FPA_VF_MAX);
+
+	res = &fpadev.pool[gpool];
+	return (uintptr_t)res->bar0 | gpool;
+}
+
+static __rte_always_inline bool
+octeontx_fpa_handle_valid(uintptr_t handle)
+{
+	struct fpavf_res *res = NULL;
+	uint8_t gpool;
+	int i;
+	bool ret = false;
+
+	if (unlikely(!handle))
+		return ret;
+
+	/* get the gpool */
+	gpool = octeontx_fpa_bufpool_gpool(handle);
+
+	/* get the bar address */
+	handle &= ~(uint64_t)FPA_GPOOL_MASK;
+	for (i = 0; i < FPA_VF_MAX; i++) {
+		if ((uintptr_t)fpadev.pool[i].bar0 != handle)
+			continue;
+
+		/* validate gpool */
+		if (gpool != i)
+			return false;
+
+		res = &fpadev.pool[i];
+
+		if (res->sz128 == 0 || res->domain_id == (uint16_t)~0 ||
+		    res->stack_ln_ptr == 0)
+			ret = false;
+		else
+			ret = true;
+		break;
+	}
+
+	return ret;
+}
+
+static int
+octeontx_fpapf_pool_setup(unsigned int gpool, unsigned int buf_size,
+			  signed short buf_offset, unsigned int max_buf_count)
+{
+	void *memptr = NULL;
+	rte_iova_t phys_addr;
+	unsigned int memsz;
+	struct fpavf_res *fpa = NULL;
+	uint64_t reg;
+	struct octeontx_mbox_hdr hdr;
+	struct dcfg_resp resp;
+	struct octeontx_mbox_fpa_cfg cfg;
+	int ret = -1;
+
+	fpa = &fpadev.pool[gpool];
+	memsz = FPA_ROUND_UP(max_buf_count / fpa->stack_ln_ptr, FPA_LN_SIZE) *
+			FPA_LN_SIZE;
+
+	/* Round-up to page size */
+	memsz = (memsz + FPA_PF_PAGE_SZ - 1) & ~(uintptr_t)(FPA_PF_PAGE_SZ-1);
+	memptr = rte_malloc(NULL, memsz, RTE_CACHE_LINE_SIZE);
+	if (memptr == NULL) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	/* Configure stack */
+	fpa->pool_stack_base = memptr;
+	phys_addr = rte_malloc_virt2iova(memptr);
+
+	buf_size /= FPA_LN_SIZE;
+
+	/* POOL setup */
+	hdr.coproc = FPA_COPROC;
+	hdr.msg = FPA_CONFIGSET;
+	hdr.vfid = fpa->vf_id;
+	hdr.res_code = 0;
+
+	buf_offset /= FPA_LN_SIZE;
+	reg = POOL_BUF_SIZE(buf_size) | POOL_BUF_OFFSET(buf_offset) |
+		POOL_LTYPE(0x2) | POOL_STYPE(0) | POOL_SET_NAT_ALIGN |
+		POOL_ENA;
+
+	cfg.aid = 0;
+	cfg.pool_cfg = reg;
+	cfg.pool_stack_base = phys_addr;
+	cfg.pool_stack_end = phys_addr + memsz;
+	cfg.aura_cfg = (1 << 9);
+
+	ret = octeontx_ssovf_mbox_send(&hdr, &cfg,
+					sizeof(struct octeontx_mbox_fpa_cfg),
+					&resp, sizeof(resp));
+	if (ret < 0) {
+		ret = -EACCES;
+		goto err;
+	}
+
+	fpavf_log_dbg(" vfid %d gpool %d aid %d pool_cfg 0x%x pool_stack_base %" PRIx64 " pool_stack_end %" PRIx64" aura_cfg %" PRIx64 "\n",
+		      fpa->vf_id, gpool, cfg.aid, (unsigned int)cfg.pool_cfg,
+		      cfg.pool_stack_base, cfg.pool_stack_end, cfg.aura_cfg);
+
+	/* Now pool is in_use */
+	fpa->is_inuse = true;
+
+err:
+	if (ret < 0)
+		rte_free(memptr);
+
+	return ret;
+}
+
+static int
+octeontx_fpapf_pool_destroy(unsigned int gpool_index)
+{
+	struct octeontx_mbox_hdr hdr;
+	struct dcfg_resp resp;
+	struct octeontx_mbox_fpa_cfg cfg;
+	struct fpavf_res *fpa = NULL;
+	int ret = -1;
+
+	fpa = &fpadev.pool[gpool_index];
+
+	hdr.coproc = FPA_COPROC;
+	hdr.msg = FPA_CONFIGSET;
+	hdr.vfid = fpa->vf_id;
+	hdr.res_code = 0;
+
+	/* reset and free the pool */
+	cfg.aid = 0;
+	cfg.pool_cfg = 0;
+	cfg.pool_stack_base = 0;
+	cfg.pool_stack_end = 0;
+	cfg.aura_cfg = 0;
+
+	ret = octeontx_ssovf_mbox_send(&hdr, &cfg,
+					sizeof(struct octeontx_mbox_fpa_cfg),
+					&resp, sizeof(resp));
+	if (ret < 0) {
+		ret = -EACCES;
+		goto err;
+	}
+
+	ret = 0;
+err:
+	/* anycase free pool stack memory */
+	rte_free(fpa->pool_stack_base);
+	fpa->pool_stack_base = NULL;
+	return ret;
+}
+
+static int
+octeontx_fpapf_aura_attach(unsigned int gpool_index)
+{
+	struct octeontx_mbox_hdr hdr;
+	struct dcfg_resp resp;
+	struct octeontx_mbox_fpa_cfg cfg;
+	int ret = 0;
+
+	if (gpool_index >= FPA_MAX_POOL) {
+		ret = -EINVAL;
+		goto err;
+	}
+	hdr.coproc = FPA_COPROC;
+	hdr.msg = FPA_ATTACHAURA;
+	hdr.vfid = gpool_index;
+	hdr.res_code = 0;
+	memset(&cfg, 0x0, sizeof(struct octeontx_mbox_fpa_cfg));
+	cfg.aid = gpool_index; /* gpool is guara */
+
+	ret = octeontx_ssovf_mbox_send(&hdr, &cfg,
+					sizeof(struct octeontx_mbox_fpa_cfg),
+					&resp, sizeof(resp));
+	if (ret < 0) {
+		fpavf_log_err("Could not attach fpa ");
+		fpavf_log_err("aura %d to pool %d. Err=%d. FuncErr=%d\n",
+			      gpool_index, gpool_index, ret, hdr.res_code);
+		ret = -EACCES;
+		goto err;
+	}
+err:
+	return ret;
+}
+
+static int
+octeontx_fpapf_aura_detach(unsigned int gpool_index)
+{
+	struct octeontx_mbox_fpa_cfg cfg = {0};
+	struct octeontx_mbox_hdr hdr = {0};
+	int ret = 0;
+
+	if (gpool_index >= FPA_MAX_POOL) {
+		ret = -EINVAL;
+		goto err;
+	}
+
+	cfg.aid = gpool_index; /* gpool is gaura */
+	hdr.coproc = FPA_COPROC;
+	hdr.msg = FPA_DETACHAURA;
+	hdr.vfid = gpool_index;
+	ret = octeontx_ssovf_mbox_send(&hdr, &cfg, sizeof(cfg), NULL, 0);
+	if (ret < 0) {
+		fpavf_log_err("Couldn't detach FPA aura %d Err=%d FuncErr=%d\n",
+			      gpool_index, ret, hdr.res_code);
+		ret = -EINVAL;
+	}
+
+err:
+	return ret;
+}
+
+static int
+octeontx_fpavf_pool_setup(uintptr_t handle, unsigned long memsz,
+			  void *memva, uint16_t gpool)
+{
+	uint64_t va_end;
+
+	if (unlikely(!handle))
+		return -ENODEV;
+
+	va_end = (uintptr_t)memva + memsz;
+	va_end &= ~RTE_CACHE_LINE_MASK;
+
+	/* VHPOOL setup */
+	fpavf_write64((uintptr_t)memva,
+			 (void *)((uintptr_t)handle +
+			 FPA_VF_VHPOOL_START_ADDR(gpool)));
+	fpavf_write64(va_end,
+			 (void *)((uintptr_t)handle +
+			 FPA_VF_VHPOOL_END_ADDR(gpool)));
+	return 0;
+}
+
+static int
+octeontx_fpapf_start_count(uint16_t gpool_index)
+{
+	int ret = 0;
+	struct octeontx_mbox_hdr hdr = {0};
+
+	if (gpool_index >= FPA_MAX_POOL) {
+		ret = -EINVAL;
+		goto err;
+	}
+
+	hdr.coproc = FPA_COPROC;
+	hdr.msg = FPA_START_COUNT;
+	hdr.vfid = gpool_index;
+	ret = octeontx_ssovf_mbox_send(&hdr, NULL, 0, NULL, 0);
+	if (ret < 0) {
+		fpavf_log_err("Could not start buffer counting for ");
+		fpavf_log_err("FPA pool %d. Err=%d. FuncErr=%d\n",
+			      gpool_index, ret, hdr.res_code);
+		ret = -EINVAL;
+		goto err;
+	}
+
+err:
+	return ret;
+}
+
+static __rte_always_inline int
+octeontx_fpavf_free(unsigned int gpool)
+{
+	int ret = 0;
+
+	if (gpool >= FPA_MAX_POOL) {
+		ret = -EINVAL;
+		goto err;
+	}
+
+	/* Pool is free */
+	fpadev.pool[gpool].is_inuse = false;
+
+err:
+	return ret;
+}
+
+static __rte_always_inline int
+octeontx_gpool_free(uint16_t gpool)
+{
+	if (fpadev.pool[gpool].sz128 != 0) {
+		fpadev.pool[gpool].sz128 = 0;
+		return 0;
+	}
+	return -EINVAL;
+}
+
+/*
+ * Return buffer size for a given pool
+ */
+int
+octeontx_fpa_bufpool_block_size(uintptr_t handle)
+{
+	struct fpavf_res *res = NULL;
+	uint8_t gpool;
+
+	if (unlikely(!octeontx_fpa_handle_valid(handle)))
+		return -EINVAL;
+
+	/* get the gpool */
+	gpool = octeontx_fpa_bufpool_gpool(handle);
+	res = &fpadev.pool[gpool];
+	return FPA_CACHE_LINE_2_OBJSZ(res->sz128);
+}
+
+int
+octeontx_fpa_bufpool_free_count(uintptr_t handle)
+{
+	uint64_t cnt, limit, avail;
+	uint8_t gpool;
+	uintptr_t pool_bar;
+
+	if (unlikely(!octeontx_fpa_handle_valid(handle)))
+		return -EINVAL;
+
+	/* get the gpool */
+	gpool = octeontx_fpa_bufpool_gpool(handle);
+
+	/* Get pool bar address from handle */
+	pool_bar = handle & ~(uint64_t)FPA_GPOOL_MASK;
+
+	cnt = fpavf_read64((void *)((uintptr_t)pool_bar +
+				FPA_VF_VHAURA_CNT(gpool)));
+	limit = fpavf_read64((void *)((uintptr_t)pool_bar +
+				FPA_VF_VHAURA_CNT_LIMIT(gpool)));
+
+	avail = fpavf_read64((void *)((uintptr_t)pool_bar +
+				FPA_VF_VHPOOL_AVAILABLE(gpool)));
+
+	return RTE_MIN(avail, (limit - cnt));
+}
+
+uintptr_t
+octeontx_fpa_bufpool_create(unsigned int object_size, unsigned int object_count,
+				unsigned int buf_offset, char **va_start,
+				int node_id)
+{
+	unsigned int gpool;
+	void *memva;
+	unsigned long memsz;
+	uintptr_t gpool_handle;
+	uintptr_t pool_bar;
+	int res;
+
+	RTE_SET_USED(node_id);
+	RTE_BUILD_BUG_ON(sizeof(struct rte_mbuf) > OCTEONTX_FPAVF_BUF_OFFSET);
+
+	if (unlikely(*va_start == NULL))
+		goto error_end;
+
+	object_size = RTE_CACHE_LINE_ROUNDUP(object_size);
+	if (object_size > FPA_MAX_OBJ_SIZE) {
+		errno = EINVAL;
+		goto error_end;
+	}
+
+	rte_spinlock_lock(&fpadev.lock);
+	res = octeontx_fpa_gpool_alloc(object_size);
+
+	/* Bail if failed */
+	if (unlikely(res < 0)) {
+		errno = res;
+		goto error_unlock;
+	}
+
+	/* get fpavf */
+	gpool = res;
+
+	/* get pool handle */
+	gpool_handle = octeontx_fpa_gpool2handle(gpool);
+	if (!octeontx_fpa_handle_valid(gpool_handle)) {
+		errno = ENOSPC;
+		goto error_gpool_free;
+	}
+
+	/* Get pool bar address from handle */
+	pool_bar = gpool_handle & ~(uint64_t)FPA_GPOOL_MASK;
+
+	res = octeontx_fpapf_pool_setup(gpool, object_size, buf_offset,
+					object_count);
+	if (res < 0) {
+		errno = res;
+		goto error_gpool_free;
+	}
+
+	/* populate AURA fields */
+	res = octeontx_fpapf_aura_attach(gpool);
+	if (res < 0) {
+		errno = res;
+		goto error_pool_destroy;
+	}
+
+	/* vf pool setup */
+	memsz = object_size * object_count;
+	memva = *va_start;
+	res = octeontx_fpavf_pool_setup(pool_bar, memsz, memva, gpool);
+	if (res < 0) {
+		errno = res;
+		goto error_gaura_detach;
+	}
+
+	/* Release lock */
+	rte_spinlock_unlock(&fpadev.lock);
+
+	/* populate AURA registers */
+	fpavf_write64(object_count, (void *)((uintptr_t)pool_bar +
+			 FPA_VF_VHAURA_CNT(gpool)));
+	fpavf_write64(object_count, (void *)((uintptr_t)pool_bar +
+			 FPA_VF_VHAURA_CNT_LIMIT(gpool)));
+	fpavf_write64(object_count + 1, (void *)((uintptr_t)pool_bar +
+			 FPA_VF_VHAURA_CNT_THRESHOLD(gpool)));
+
+	octeontx_fpapf_start_count(gpool);
+
+	return gpool_handle;
+
+error_gaura_detach:
+	(void) octeontx_fpapf_aura_detach(gpool);
+error_pool_destroy:
+	octeontx_fpavf_free(gpool);
+	octeontx_fpapf_pool_destroy(gpool);
+error_gpool_free:
+	octeontx_gpool_free(gpool);
+error_unlock:
+	rte_spinlock_unlock(&fpadev.lock);
+error_end:
+	return (uintptr_t)NULL;
+}
+
+/*
+ * Destroy a buffer pool.
+ */
+int
+octeontx_fpa_bufpool_destroy(uintptr_t handle, int node_id)
+{
+	void **node, **curr, *head = NULL;
+	uint64_t sz;
+	uint64_t cnt, avail;
+	uint8_t gpool;
+	uintptr_t pool_bar;
+	int ret;
+
+	RTE_SET_USED(node_id);
+
+	/* Wait for all outstanding writes to be committed */
+	rte_smp_wmb();
+
+	if (unlikely(!octeontx_fpa_handle_valid(handle)))
+		return -EINVAL;
+
+	/* get the pool */
+	gpool = octeontx_fpa_bufpool_gpool(handle);
+
+	/* Get pool bar address from handle */
+	pool_bar = handle & ~(uint64_t)FPA_GPOOL_MASK;
+
+	 /* Check for no outstanding buffers */
+	cnt = fpavf_read64((void *)((uintptr_t)pool_bar +
+					FPA_VF_VHAURA_CNT(gpool)));
+	if (cnt) {
+		fpavf_log_dbg("buffer exist in pool cnt %" PRId64 "\n", cnt);
+		return -EBUSY;
+	}
+
+	rte_spinlock_lock(&fpadev.lock);
+
+	avail = fpavf_read64((void *)((uintptr_t)pool_bar +
+				FPA_VF_VHPOOL_AVAILABLE(gpool)));
+
+	/* Prepare to empty the entire POOL */
+	fpavf_write64(avail, (void *)((uintptr_t)pool_bar +
+			 FPA_VF_VHAURA_CNT_LIMIT(gpool)));
+	fpavf_write64(avail + 1, (void *)((uintptr_t)pool_bar +
+			 FPA_VF_VHAURA_CNT_THRESHOLD(gpool)));
+
+	/* Empty the pool */
+	/* Invalidate the POOL */
+	octeontx_gpool_free(gpool);
+
+	/* Process all buffers in the pool */
+	while (avail--) {
+
+		/* Yank a buffer from the pool */
+		node = (void *)(uintptr_t)
+			fpavf_read64((void *)
+				    (pool_bar + FPA_VF_VHAURA_OP_ALLOC(gpool)));
+
+		if (node == NULL) {
+			fpavf_log_err("GAURA[%u] missing %" PRIx64 " buf\n",
+				      gpool, avail);
+			break;
+		}
+
+		/* Imsert it into an ordered linked list */
+		for (curr = &head; curr[0] != NULL; curr = curr[0]) {
+			if ((uintptr_t)node <= (uintptr_t)curr[0])
+				break;
+		}
+		node[0] = curr[0];
+		curr[0] = node;
+	}
+
+	/* Verify the linked list to be a perfect series */
+	sz = octeontx_fpa_bufpool_block_size(handle) << 7;
+	for (curr = head; curr != NULL && curr[0] != NULL;
+		curr = curr[0]) {
+		if (curr == curr[0] ||
+			((uintptr_t)curr != ((uintptr_t)curr[0] - sz))) {
+			fpavf_log_err("POOL# %u buf sequence err (%p vs. %p)\n",
+				      gpool, curr, curr[0]);
+		}
+	}
+
+	/* Disable pool operation */
+	fpavf_write64(~0ul, (void *)((uintptr_t)pool_bar +
+			 FPA_VF_VHPOOL_START_ADDR(gpool)));
+	fpavf_write64(~0ul, (void *)((uintptr_t)pool_bar +
+			FPA_VF_VHPOOL_END_ADDR(gpool)));
+
+	(void)octeontx_fpapf_pool_destroy(gpool);
+
+	/* Deactivate the AURA */
+	fpavf_write64(0, (void *)((uintptr_t)pool_bar +
+			FPA_VF_VHAURA_CNT_LIMIT(gpool)));
+	fpavf_write64(0, (void *)((uintptr_t)pool_bar +
+			FPA_VF_VHAURA_CNT_THRESHOLD(gpool)));
+
+	ret = octeontx_fpapf_aura_detach(gpool);
+	if (ret) {
+		fpavf_log_err("Failed to dettach gaura %u. error code=%d\n",
+			      gpool, ret);
+	}
+
+	/* Free VF */
+	(void)octeontx_fpavf_free(gpool);
+
+	rte_spinlock_unlock(&fpadev.lock);
+	return 0;
+}
+
+static void
+octeontx_fpavf_setup(void)
+{
+	uint8_t i;
+	static bool init_once;
+
+	if (!init_once) {
+		rte_spinlock_init(&fpadev.lock);
+		fpadev.total_gpool_cnt = 0;
+
+		for (i = 0; i < FPA_VF_MAX; i++) {
+
+			fpadev.pool[i].domain_id = ~0;
+			fpadev.pool[i].stack_ln_ptr = 0;
+			fpadev.pool[i].sz128 = 0;
+			fpadev.pool[i].bar0 = NULL;
+			fpadev.pool[i].pool_stack_base = NULL;
+			fpadev.pool[i].is_inuse = false;
+		}
+		init_once = 1;
+	}
+}
+
+static int
+octeontx_fpavf_identify(void *bar0)
+{
+	uint64_t val;
+	uint16_t domain_id;
+	uint16_t vf_id;
+	uint64_t stack_ln_ptr;
+
+	val = fpavf_read64((void *)((uintptr_t)bar0 +
+				FPA_VF_VHAURA_CNT_THRESHOLD(0)));
+
+	domain_id = (val >> 8) & 0xffff;
+	vf_id = (val >> 24) & 0xffff;
+
+	stack_ln_ptr = fpavf_read64((void *)((uintptr_t)bar0 +
+					FPA_VF_VHPOOL_THRESHOLD(0)));
+	if (vf_id >= FPA_VF_MAX) {
+		fpavf_log_err("vf_id(%d) greater than max vf (32)\n", vf_id);
+		return -1;
+	}
+
+	if (fpadev.pool[vf_id].is_inuse) {
+		fpavf_log_err("vf_id %d is_inuse\n", vf_id);
+		return -1;
+	}
+
+	fpadev.pool[vf_id].domain_id = domain_id;
+	fpadev.pool[vf_id].vf_id = vf_id;
+	fpadev.pool[vf_id].bar0 = bar0;
+	fpadev.pool[vf_id].stack_ln_ptr = stack_ln_ptr;
+
+	/* SUCCESS */
+	return vf_id;
+}
+
+/* FPAVF pcie device aka mempool probe */
+static int
+fpavf_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
+{
+	uint8_t *idreg;
+	int res;
+	struct fpavf_res *fpa = NULL;
+
+	RTE_SET_USED(pci_drv);
+	RTE_SET_USED(fpa);
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	if (pci_dev->mem_resource[0].addr == NULL) {
+		fpavf_log_err("Empty bars %p ", pci_dev->mem_resource[0].addr);
+		return -ENODEV;
+	}
+	idreg = pci_dev->mem_resource[0].addr;
+
+	octeontx_fpavf_setup();
+
+	res = octeontx_fpavf_identify(idreg);
+	if (res < 0)
+		return -1;
+
+	fpa = &fpadev.pool[res];
+	fpadev.total_gpool_cnt++;
+	rte_wmb();
+
+	fpavf_log_dbg("total_fpavfs %d bar0 %p domain %d vf %d stk_ln_ptr 0x%x",
+		       fpadev.total_gpool_cnt, fpa->bar0, fpa->domain_id,
+		       fpa->vf_id, (unsigned int)fpa->stack_ln_ptr);
+
+	return 0;
+}
+
+static const struct rte_pci_id pci_fpavf_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_CAVIUM,
+				PCI_DEVICE_ID_OCTEONTX_FPA_VF)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static struct rte_pci_driver pci_fpavf = {
+	.id_table = pci_fpavf_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
+	.probe = fpavf_probe,
+};
+
+RTE_PMD_REGISTER_PCI(octeontx_fpavf, pci_fpavf);
diff --git a/drivers/mempool/octeontx/octeontx_fpavf.h b/drivers/mempool/octeontx/octeontx_fpavf.h
new file mode 100644
index 00000000..1d09f007
--- /dev/null
+++ b/drivers/mempool/octeontx/octeontx_fpavf.h
@@ -0,0 +1,131 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) 2017 Cavium 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:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Cavium networks nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef	__OCTEONTX_FPAVF_H__
+#define	__OCTEONTX_FPAVF_H__
+
+#include <rte_io.h>
+#include "octeontx_pool_logs.h"
+
+/* fpa pool Vendor ID and Device ID */
+#define PCI_VENDOR_ID_CAVIUM		0x177D
+#define PCI_DEVICE_ID_OCTEONTX_FPA_VF	0xA053
+
+#define	FPA_VF_MAX			32
+#define FPA_GPOOL_MASK			(FPA_VF_MAX-1)
+
+/* FPA VF register offsets */
+#define FPA_VF_INT(x)			(0x200ULL | ((x) << 22))
+#define FPA_VF_INT_W1S(x)		(0x210ULL | ((x) << 22))
+#define FPA_VF_INT_ENA_W1S(x)		(0x220ULL | ((x) << 22))
+#define FPA_VF_INT_ENA_W1C(x)		(0x230ULL | ((x) << 22))
+
+#define	FPA_VF_VHPOOL_AVAILABLE(vhpool)		(0x04150 | ((vhpool)&0x0))
+#define	FPA_VF_VHPOOL_THRESHOLD(vhpool)		(0x04160 | ((vhpool)&0x0))
+#define	FPA_VF_VHPOOL_START_ADDR(vhpool)	(0x04200 | ((vhpool)&0x0))
+#define	FPA_VF_VHPOOL_END_ADDR(vhpool)		(0x04210 | ((vhpool)&0x0))
+
+#define	FPA_VF_VHAURA_CNT(vaura)		(0x20120 | ((vaura)&0xf)<<18)
+#define	FPA_VF_VHAURA_CNT_ADD(vaura)		(0x20128 | ((vaura)&0xf)<<18)
+#define	FPA_VF_VHAURA_CNT_LIMIT(vaura)		(0x20130 | ((vaura)&0xf)<<18)
+#define	FPA_VF_VHAURA_CNT_THRESHOLD(vaura)	(0x20140 | ((vaura)&0xf)<<18)
+#define	FPA_VF_VHAURA_OP_ALLOC(vaura)		(0x30000 | ((vaura)&0xf)<<18)
+#define	FPA_VF_VHAURA_OP_FREE(vaura)		(0x38000 | ((vaura)&0xf)<<18)
+
+#define FPA_VF_FREE_ADDRS_S(x, y, z)	\
+	((x) | (((y) & 0x1ff) << 3) | ((((z) & 1)) << 14))
+
+/* FPA VF register offsets from VF_BAR4, size 2 MByte */
+#define	FPA_VF_MSIX_VEC_ADDR		0x00000
+#define	FPA_VF_MSIX_VEC_CTL		0x00008
+#define	FPA_VF_MSIX_PBA			0xF0000
+
+#define	FPA_VF0_APERTURE_SHIFT		22
+#define FPA_AURA_SET_SIZE		16
+
+#define FPA_MAX_OBJ_SIZE		(128 * 1024)
+#define OCTEONTX_FPAVF_BUF_OFFSET	128
+
+/*
+ * In Cavium OcteonTX SoC, all accesses to the device registers are
+ * implicitly strongly ordered. So, the relaxed version of IO operation is
+ * safe to use with out any IO memory barriers.
+ */
+#define fpavf_read64 rte_read64_relaxed
+#define fpavf_write64 rte_write64_relaxed
+
+/* ARM64 specific functions */
+#if defined(RTE_ARCH_ARM64)
+#define fpavf_load_pair(val0, val1, addr) ({		\
+			asm volatile(			\
+			"ldp %x[x0], %x[x1], [%x[p1]]"	\
+			:[x0]"=r"(val0), [x1]"=r"(val1) \
+			:[p1]"r"(addr)			\
+			); })
+
+#define fpavf_store_pair(val0, val1, addr) ({		\
+			asm volatile(			\
+			"stp %x[x0], %x[x1], [%x[p1]]"	\
+			::[x0]"r"(val0), [x1]"r"(val1), [p1]"r"(addr) \
+			); })
+#else /* Un optimized functions for building on non arm64 arch */
+
+#define fpavf_load_pair(val0, val1, addr)		\
+do {							\
+	val0 = rte_read64(addr);			\
+	val1 = rte_read64(((uint8_t *)addr) + 8);	\
+} while (0)
+
+#define fpavf_store_pair(val0, val1, addr)		\
+do {							\
+	rte_write64(val0, addr);			\
+	rte_write64(val1, (((uint8_t *)addr) + 8));	\
+} while (0)
+#endif
+
+uintptr_t
+octeontx_fpa_bufpool_create(unsigned int object_size, unsigned int object_count,
+				unsigned int buf_offset, char **va_start,
+				int node);
+int
+octeontx_fpa_bufpool_destroy(uintptr_t handle, int node);
+int
+octeontx_fpa_bufpool_block_size(uintptr_t handle);
+int
+octeontx_fpa_bufpool_free_count(uintptr_t handle);
+
+static __rte_always_inline uint8_t
+octeontx_fpa_bufpool_gpool(uintptr_t handle)
+{
+	return (uint8_t)handle & FPA_GPOOL_MASK;
+}
+#endif	/* __OCTEONTX_FPAVF_H__ */
diff --git a/drivers/mempool/octeontx/octeontx_mbox.c b/drivers/mempool/octeontx/octeontx_mbox.c
new file mode 100644
index 00000000..9525da1a
--- /dev/null
+++ b/drivers/mempool/octeontx/octeontx_mbox.c
@@ -0,0 +1,242 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) Cavium, Inc. 2017.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Cavium, Inc nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <string.h>
+
+#include <rte_atomic.h>
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_io.h>
+#include <rte_spinlock.h>
+
+#include "octeontx_mbox.h"
+#include "octeontx_pool_logs.h"
+
+/* Mbox operation timeout in seconds */
+#define MBOX_WAIT_TIME_SEC	3
+#define MAX_RAM_MBOX_LEN	((SSOW_BAR4_LEN >> 1) - 8 /* Mbox header */)
+
+/* Mbox channel state */
+enum {
+	MBOX_CHAN_STATE_REQ = 1,
+	MBOX_CHAN_STATE_RES = 0,
+};
+
+/* Response messages */
+enum {
+	MBOX_RET_SUCCESS,
+	MBOX_RET_INVALID,
+	MBOX_RET_INTERNAL_ERR,
+};
+
+struct mbox {
+	int init_once;
+	uint8_t *ram_mbox_base; /* Base address of mbox message stored in ram */
+	uint8_t *reg; /* Store to this register triggers PF mbox interrupt */
+	uint16_t tag_own; /* Last tag which was written to own channel */
+	rte_spinlock_t lock;
+};
+
+static struct mbox octeontx_mbox;
+
+/*
+ * Structure used for mbox synchronization
+ * This structure sits at the begin of Mbox RAM and used as main
+ * synchronization point for channel communication
+ */
+struct mbox_ram_hdr {
+	union {
+		uint64_t u64;
+		struct {
+			uint8_t chan_state : 1;
+			uint8_t coproc : 7;
+			uint8_t msg;
+			uint8_t vfid;
+			uint8_t res_code;
+			uint16_t tag;
+			uint16_t len;
+		};
+	};
+};
+
+static inline void
+mbox_msgcpy(uint8_t *d, const uint8_t *s, uint16_t size)
+{
+	uint16_t i;
+
+	for (i = 0; i < size; i++)
+		d[i] = s[i];
+}
+
+static inline void
+mbox_send_request(struct mbox *m, struct octeontx_mbox_hdr *hdr,
+			const void *txmsg, uint16_t txsize)
+{
+	struct mbox_ram_hdr old_hdr;
+	struct mbox_ram_hdr new_hdr = { {0} };
+	uint64_t *ram_mbox_hdr = (uint64_t *)m->ram_mbox_base;
+	uint8_t *ram_mbox_msg = m->ram_mbox_base + sizeof(struct mbox_ram_hdr);
+
+	/*
+	 * Initialize the channel with the tag left by last send.
+	 * On success full mbox send complete, PF increments the tag by one.
+	 * The sender can validate integrity of PF message with this scheme
+	 */
+	old_hdr.u64 = rte_read64(ram_mbox_hdr);
+	m->tag_own = (old_hdr.tag + 2) & (~0x1ul); /* next even number */
+
+	/* Copy msg body */
+	if (txmsg)
+		mbox_msgcpy(ram_mbox_msg, txmsg, txsize);
+
+	/* Prepare new hdr */
+	new_hdr.chan_state = MBOX_CHAN_STATE_REQ;
+	new_hdr.coproc = hdr->coproc;
+	new_hdr.msg = hdr->msg;
+	new_hdr.vfid = hdr->vfid;
+	new_hdr.tag = m->tag_own;
+	new_hdr.len = txsize;
+
+	/* Write the msg header */
+	rte_write64(new_hdr.u64, ram_mbox_hdr);
+	rte_io_wmb();
+	/* Notify PF about the new msg - write to MBOX reg generates PF IRQ */
+	rte_write64(0, m->reg);
+}
+
+static inline int
+mbox_wait_response(struct mbox *m, struct octeontx_mbox_hdr *hdr,
+			void *rxmsg, uint16_t rxsize)
+{
+	int res = 0, wait;
+	uint16_t len;
+	struct mbox_ram_hdr rx_hdr;
+	uint64_t *ram_mbox_hdr = (uint64_t *)m->ram_mbox_base;
+	uint8_t *ram_mbox_msg = m->ram_mbox_base + sizeof(struct mbox_ram_hdr);
+
+	/* Wait for response */
+	wait = MBOX_WAIT_TIME_SEC * 1000 * 10;
+	while (wait > 0) {
+		rte_delay_us(100);
+		rx_hdr.u64 = rte_read64(ram_mbox_hdr);
+		if (rx_hdr.chan_state == MBOX_CHAN_STATE_RES)
+			break;
+		--wait;
+	}
+
+	hdr->res_code = rx_hdr.res_code;
+	m->tag_own++;
+
+	/* Timeout */
+	if (wait <= 0) {
+		res = -ETIMEDOUT;
+		goto error;
+	}
+
+	/* Tag mismatch */
+	if (m->tag_own != rx_hdr.tag) {
+		res = -EINVAL;
+		goto error;
+	}
+
+	/* PF nacked the msg */
+	if (rx_hdr.res_code != MBOX_RET_SUCCESS) {
+		res = -EBADMSG;
+		goto error;
+	}
+
+	len = RTE_MIN(rx_hdr.len, rxsize);
+	if (rxmsg)
+		mbox_msgcpy(rxmsg, ram_mbox_msg, len);
+
+	return len;
+
+error:
+	mbox_log_err("Failed to send mbox(%d/%d) coproc=%d msg=%d ret=(%d,%d)",
+			m->tag_own, rx_hdr.tag, hdr->coproc, hdr->msg, res,
+			hdr->res_code);
+	return res;
+}
+
+static inline int
+mbox_send(struct mbox *m, struct octeontx_mbox_hdr *hdr, const void *txmsg,
+		uint16_t txsize, void *rxmsg, uint16_t rxsize)
+{
+	int res = -EINVAL;
+
+	if (m->init_once == 0 || hdr == NULL ||
+		txsize > MAX_RAM_MBOX_LEN || rxsize > MAX_RAM_MBOX_LEN) {
+		mbox_log_err("Invalid init_once=%d hdr=%p txsz=%d rxsz=%d",
+				m->init_once, hdr, txsize, rxsize);
+		return res;
+	}
+
+	rte_spinlock_lock(&m->lock);
+
+	mbox_send_request(m, hdr, txmsg, txsize);
+	res = mbox_wait_response(m, hdr, rxmsg, rxsize);
+
+	rte_spinlock_unlock(&m->lock);
+	return res;
+}
+
+static inline int
+mbox_setup(struct mbox *m)
+{
+	if (unlikely(m->init_once == 0)) {
+		rte_spinlock_init(&m->lock);
+		m->ram_mbox_base = octeontx_ssovf_bar(OCTEONTX_SSO_HWS, 0, 4);
+		m->reg = octeontx_ssovf_bar(OCTEONTX_SSO_GROUP, 0, 0);
+		m->reg += SSO_VHGRP_PF_MBOX(1);
+
+		if (m->ram_mbox_base == NULL || m->reg == NULL) {
+			mbox_log_err("Invalid ram_mbox_base=%p or reg=%p",
+				m->ram_mbox_base, m->reg);
+			return -EINVAL;
+		}
+		m->init_once = 1;
+	}
+	return 0;
+}
+
+int
+octeontx_ssovf_mbox_send(struct octeontx_mbox_hdr *hdr, void *txdata,
+				 uint16_t txlen, void *rxdata, uint16_t rxlen)
+{
+	struct mbox *m = &octeontx_mbox;
+
+	RTE_BUILD_BUG_ON(sizeof(struct mbox_ram_hdr) != 8);
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY || mbox_setup(m))
+		return -EINVAL;
+
+	return mbox_send(m, hdr, txdata, txlen, rxdata, rxlen);
+}
diff --git a/drivers/mempool/octeontx/octeontx_mbox.h b/drivers/mempool/octeontx/octeontx_mbox.h
new file mode 100644
index 00000000..49f38257
--- /dev/null
+++ b/drivers/mempool/octeontx/octeontx_mbox.h
@@ -0,0 +1,64 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) Cavium, Inc. 2017.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Cavium, Inc nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __OCTEONTX_MBOX_H__
+#define __OCTEONTX_MBOX_H__
+
+#include <rte_common.h>
+
+#define SSOW_BAR4_LEN			(64 * 1024)
+#define SSO_VHGRP_PF_MBOX(x)		(0x200ULL | ((x) << 3))
+
+struct octeontx_ssovf_info {
+	uint16_t domain; /* Domain id */
+	uint8_t total_ssovfs; /* Total sso groups available in domain */
+	uint8_t total_ssowvfs;/* Total sso hws available in domain */
+};
+
+enum octeontx_ssovf_type {
+	OCTEONTX_SSO_GROUP, /* SSO group vf */
+	OCTEONTX_SSO_HWS,  /* SSO hardware workslot vf */
+};
+
+struct octeontx_mbox_hdr {
+	uint16_t vfid;  /* VF index or pf resource index local to the domain */
+	uint8_t coproc; /* Coprocessor id */
+	uint8_t msg;    /* Message id */
+	uint8_t res_code; /* Functional layer response code */
+};
+
+int octeontx_ssovf_info(struct octeontx_ssovf_info *info);
+void *octeontx_ssovf_bar(enum octeontx_ssovf_type, uint8_t id, uint8_t bar);
+int octeontx_ssovf_mbox_send(struct octeontx_mbox_hdr *hdr,
+		void *txdata, uint16_t txlen, void *rxdata, uint16_t rxlen);
+
+#endif /* __OCTEONTX_MBOX_H__ */
diff --git a/drivers/mempool/octeontx/octeontx_pool_logs.h b/drivers/mempool/octeontx/octeontx_pool_logs.h
new file mode 100644
index 00000000..58ccb0f0
--- /dev/null
+++ b/drivers/mempool/octeontx/octeontx_pool_logs.h
@@ -0,0 +1,68 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) 2017 Cavium 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:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Cavium networks nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef	__OCTEONTX_POOL_LOGS_H__
+#define	__OCTEONTX_POOL_LOGS_H__
+
+#include <rte_debug.h>
+
+#ifdef RTE_LIBRTE_OCTEONTX_MEMPOOL_DEBUG
+#define fpavf_log_info(fmt, args...) \
+	RTE_LOG(INFO, PMD, "%s() line %u: " fmt "\n", \
+		__func__, __LINE__, ## args)
+#define fpavf_log_dbg(fmt, args...) \
+	RTE_LOG(DEBUG, PMD, "%s() line %u: " fmt "\n", \
+		__func__, __LINE__, ## args)
+
+#define mbox_log_info(fmt, args...) \
+	RTE_LOG(INFO, PMD, "%s() line %u: " fmt "\n", \
+		__func__, __LINE__, ## args)
+#define mbox_log_dbg(fmt, args...) \
+	RTE_LOG(DEBUG, PMD, "%s() line %u: " fmt "\n", \
+		__func__, __LINE__, ## args)
+#else
+#define fpavf_log_info(fmt, args...)
+#define fpavf_log_dbg(fmt, args...)
+#define mbox_log_info(fmt, args...)
+#define mbox_log_dbg(fmt, args...)
+#endif
+
+#define fpavf_func_trace fpavf_log_dbg
+#define fpavf_log_err(fmt, args...) \
+	RTE_LOG(ERR, PMD, "%s() line %u: " fmt "\n", \
+		__func__, __LINE__, ## args)
+#define mbox_func_trace mbox_log_dbg
+#define mbox_log_err(fmt, args...) \
+	RTE_LOG(ERR, PMD, "%s() line %u: " fmt "\n", \
+		__func__, __LINE__, ## args)
+
+#endif /* __OCTEONTX_POOL_LOGS_H__*/
diff --git a/drivers/mempool/octeontx/octeontx_ssovf.c b/drivers/mempool/octeontx/octeontx_ssovf.c
new file mode 100644
index 00000000..012c887d
--- /dev/null
+++ b/drivers/mempool/octeontx/octeontx_ssovf.c
@@ -0,0 +1,299 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) Cavium, Inc. 2017.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Cavium, Inc nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_atomic.h>
+#include <rte_common.h>
+#include <rte_eal.h>
+#include <rte_io.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
+
+#include "octeontx_mbox.h"
+#include "octeontx_pool_logs.h"
+
+#define PCI_VENDOR_ID_CAVIUM              0x177D
+#define PCI_DEVICE_ID_OCTEONTX_SSOGRP_VF  0xA04B
+#define PCI_DEVICE_ID_OCTEONTX_SSOWS_VF   0xA04D
+
+#define SSO_MAX_VHGRP                     (64)
+#define SSO_MAX_VHWS                      (32)
+
+#define SSO_VHGRP_AQ_THR                  (0x1E0ULL)
+
+struct ssovf_res {
+	uint16_t domain;
+	uint16_t vfid;
+	void *bar0;
+	void *bar2;
+};
+
+struct ssowvf_res {
+	uint16_t domain;
+	uint16_t vfid;
+	void *bar0;
+	void *bar2;
+	void *bar4;
+};
+
+struct ssowvf_identify {
+	uint16_t domain;
+	uint16_t vfid;
+};
+
+struct ssodev {
+	uint8_t total_ssovfs;
+	uint8_t total_ssowvfs;
+	struct ssovf_res grp[SSO_MAX_VHGRP];
+	struct ssowvf_res hws[SSO_MAX_VHWS];
+};
+
+static struct ssodev sdev;
+
+/* Interface functions */
+int
+octeontx_ssovf_info(struct octeontx_ssovf_info *info)
+{
+	uint8_t i;
+	uint16_t domain;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY || info == NULL)
+		return -EINVAL;
+
+	if (sdev.total_ssovfs == 0 || sdev.total_ssowvfs == 0)
+		return -ENODEV;
+
+	domain = sdev.grp[0].domain;
+	for (i = 0; i < sdev.total_ssovfs; i++) {
+		/* Check vfid's are contiguous and belong to same domain */
+		if (sdev.grp[i].vfid != i ||
+			sdev.grp[i].bar0 == NULL ||
+			sdev.grp[i].domain != domain) {
+			mbox_log_err("GRP error, vfid=%d/%d domain=%d/%d %p",
+				i, sdev.grp[i].vfid,
+				domain, sdev.grp[i].domain,
+				sdev.grp[i].bar0);
+			return -EINVAL;
+		}
+	}
+
+	for (i = 0; i < sdev.total_ssowvfs; i++) {
+		/* Check vfid's are contiguous and belong to same domain */
+		if (sdev.hws[i].vfid != i ||
+			sdev.hws[i].bar0 == NULL ||
+			sdev.hws[i].domain != domain) {
+			mbox_log_err("HWS error, vfid=%d/%d domain=%d/%d %p",
+				i, sdev.hws[i].vfid,
+				domain, sdev.hws[i].domain,
+				sdev.hws[i].bar0);
+			return -EINVAL;
+		}
+	}
+
+	info->domain = domain;
+	info->total_ssovfs = sdev.total_ssovfs;
+	info->total_ssowvfs = sdev.total_ssowvfs;
+	return 0;
+}
+
+void*
+octeontx_ssovf_bar(enum octeontx_ssovf_type type, uint8_t id, uint8_t bar)
+{
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY ||
+			type > OCTEONTX_SSO_HWS)
+		return NULL;
+
+	if (type == OCTEONTX_SSO_GROUP) {
+		if (id >= sdev.total_ssovfs)
+			return NULL;
+	} else {
+		if (id >= sdev.total_ssowvfs)
+			return NULL;
+	}
+
+	if (type == OCTEONTX_SSO_GROUP) {
+		switch (bar) {
+		case 0:
+			return sdev.grp[id].bar0;
+		case 2:
+			return sdev.grp[id].bar2;
+		default:
+			return NULL;
+		}
+	} else {
+		switch (bar) {
+		case 0:
+			return sdev.hws[id].bar0;
+		case 2:
+			return sdev.hws[id].bar2;
+		case 4:
+			return sdev.hws[id].bar4;
+		default:
+			return NULL;
+		}
+	}
+}
+
+/* SSOWVF pcie device aka event port probe */
+
+static int
+ssowvf_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
+{
+	uint16_t vfid;
+	struct ssowvf_res *res;
+	struct ssowvf_identify *id;
+
+	RTE_SET_USED(pci_drv);
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	if (pci_dev->mem_resource[0].addr == NULL ||
+			pci_dev->mem_resource[2].addr == NULL ||
+			pci_dev->mem_resource[4].addr == NULL) {
+		mbox_log_err("Empty bars %p %p %p",
+				pci_dev->mem_resource[0].addr,
+				pci_dev->mem_resource[2].addr,
+				pci_dev->mem_resource[4].addr);
+		return -ENODEV;
+	}
+
+	if (pci_dev->mem_resource[4].len != SSOW_BAR4_LEN) {
+		mbox_log_err("Bar4 len mismatch %d != %d",
+			SSOW_BAR4_LEN, (int)pci_dev->mem_resource[4].len);
+		return -EINVAL;
+	}
+
+	id = pci_dev->mem_resource[4].addr;
+	vfid = id->vfid;
+	if (vfid >= SSO_MAX_VHWS) {
+		mbox_log_err("Invalid vfid(%d/%d)", vfid, SSO_MAX_VHWS);
+		return -EINVAL;
+	}
+
+	res = &sdev.hws[vfid];
+	res->vfid = vfid;
+	res->bar0 = pci_dev->mem_resource[0].addr;
+	res->bar2 = pci_dev->mem_resource[2].addr;
+	res->bar4 = pci_dev->mem_resource[4].addr;
+	res->domain = id->domain;
+
+	sdev.total_ssowvfs++;
+	rte_wmb();
+	mbox_log_dbg("Domain=%d hws=%d total_ssowvfs=%d", res->domain,
+			res->vfid, sdev.total_ssowvfs);
+	return 0;
+}
+
+static const struct rte_pci_id pci_ssowvf_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_CAVIUM,
+				PCI_DEVICE_ID_OCTEONTX_SSOWS_VF)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static struct rte_pci_driver pci_ssowvf = {
+	.id_table = pci_ssowvf_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
+	.probe = ssowvf_probe,
+};
+
+RTE_PMD_REGISTER_PCI(octeontx_ssowvf, pci_ssowvf);
+
+/* SSOVF pcie device aka event queue probe */
+
+static int
+ssovf_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
+{
+	uint64_t val;
+	uint16_t vfid;
+	uint8_t *idreg;
+	struct ssovf_res *res;
+
+	RTE_SET_USED(pci_drv);
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	if (pci_dev->mem_resource[0].addr == NULL ||
+			pci_dev->mem_resource[2].addr == NULL) {
+		mbox_log_err("Empty bars %p %p",
+			pci_dev->mem_resource[0].addr,
+			pci_dev->mem_resource[2].addr);
+		return -ENODEV;
+	}
+	idreg = pci_dev->mem_resource[0].addr;
+	idreg += SSO_VHGRP_AQ_THR;
+	val = rte_read64(idreg);
+
+	/* Write back the default value of aq_thr */
+	rte_write64((1ULL << 33) - 1, idreg);
+	vfid = (val >> 16) & 0xffff;
+	if (vfid >= SSO_MAX_VHGRP) {
+		mbox_log_err("Invalid vfid (%d/%d)", vfid, SSO_MAX_VHGRP);
+		return -EINVAL;
+	}
+
+	res = &sdev.grp[vfid];
+	res->vfid = vfid;
+	res->bar0 = pci_dev->mem_resource[0].addr;
+	res->bar2 = pci_dev->mem_resource[2].addr;
+	res->domain = val & 0xffff;
+
+	sdev.total_ssovfs++;
+	rte_wmb();
+	mbox_log_dbg("Domain=%d group=%d total_ssovfs=%d", res->domain,
+			res->vfid, sdev.total_ssovfs);
+	return 0;
+}
+
+static const struct rte_pci_id pci_ssovf_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_CAVIUM,
+				PCI_DEVICE_ID_OCTEONTX_SSOGRP_VF)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static struct rte_pci_driver pci_ssovf = {
+	.id_table = pci_ssovf_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
+	.probe = ssovf_probe,
+};
+
+RTE_PMD_REGISTER_PCI(octeontx_ssovf, pci_ssovf);
diff --git a/drivers/mempool/octeontx/rte_mempool_octeontx.c b/drivers/mempool/octeontx/rte_mempool_octeontx.c
new file mode 100644
index 00000000..e89355cd
--- /dev/null
+++ b/drivers/mempool/octeontx/rte_mempool_octeontx.c
@@ -0,0 +1,253 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright (C) 2017 Cavium 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:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <stdio.h>
+#include <rte_mempool.h>
+#include <rte_malloc.h>
+#include <rte_mbuf.h>
+
+#include "octeontx_fpavf.h"
+
+/*
+ * Per-pool descriptor.
+ * Links mempool with the corresponding memzone,
+ * that provides memory under the pool's elements.
+ */
+struct octeontx_pool_info {
+	const struct rte_mempool *mp;
+	uintptr_t mz_addr;
+
+	SLIST_ENTRY(octeontx_pool_info) link;
+};
+
+SLIST_HEAD(octeontx_pool_list, octeontx_pool_info);
+
+/* List of the allocated pools */
+static struct octeontx_pool_list octeontx_pool_head =
+				SLIST_HEAD_INITIALIZER(octeontx_pool_head);
+/* Spinlock to protect pool list */
+static rte_spinlock_t pool_list_lock = RTE_SPINLOCK_INITIALIZER;
+
+static int
+octeontx_fpavf_alloc(struct rte_mempool *mp)
+{
+	uintptr_t pool;
+	struct octeontx_pool_info *pool_info;
+	uint32_t memseg_count = mp->size;
+	uint32_t object_size;
+	uintptr_t va_start;
+	int rc = 0;
+
+	rte_spinlock_lock(&pool_list_lock);
+	SLIST_FOREACH(pool_info, &octeontx_pool_head, link) {
+		if (pool_info->mp == mp)
+			break;
+	}
+	if (pool_info == NULL) {
+		rte_spinlock_unlock(&pool_list_lock);
+		return -ENXIO;
+	}
+
+	/* virtual hugepage mapped addr */
+	va_start = pool_info->mz_addr;
+	rte_spinlock_unlock(&pool_list_lock);
+
+	object_size = mp->elt_size + mp->header_size + mp->trailer_size;
+
+	pool = octeontx_fpa_bufpool_create(object_size, memseg_count,
+						OCTEONTX_FPAVF_BUF_OFFSET,
+						(char **)&va_start,
+						mp->socket_id);
+	rc = octeontx_fpa_bufpool_block_size(pool);
+	if (rc < 0)
+		goto _end;
+
+	if ((uint32_t)rc != object_size)
+		fpavf_log_err("buffer size mismatch: %d instead of %u\n",
+				rc, object_size);
+
+	fpavf_log_info("Pool created %p with .. ", (void *)pool);
+	fpavf_log_info("obj_sz %d, cnt %d\n", object_size, memseg_count);
+
+	/* assign pool handle to mempool */
+	mp->pool_id = (uint64_t)pool;
+
+	return 0;
+
+_end:
+	return rc;
+}
+
+static void
+octeontx_fpavf_free(struct rte_mempool *mp)
+{
+	struct octeontx_pool_info *pool_info;
+	uintptr_t pool;
+
+	pool = (uintptr_t)mp->pool_id;
+
+	rte_spinlock_lock(&pool_list_lock);
+	SLIST_FOREACH(pool_info, &octeontx_pool_head, link) {
+		if (pool_info->mp == mp)
+			break;
+	}
+
+	if (pool_info == NULL) {
+		rte_spinlock_unlock(&pool_list_lock);
+		rte_panic("%s: trying to free pool with no valid metadata",
+		    __func__);
+	}
+
+	SLIST_REMOVE(&octeontx_pool_head, pool_info, octeontx_pool_info, link);
+	rte_spinlock_unlock(&pool_list_lock);
+
+	rte_free(pool_info);
+	octeontx_fpa_bufpool_destroy(pool, mp->socket_id);
+}
+
+static __rte_always_inline void *
+octeontx_fpa_bufpool_alloc(uintptr_t handle)
+{
+	return (void *)(uintptr_t)fpavf_read64((void *)(handle +
+						FPA_VF_VHAURA_OP_ALLOC(0)));
+}
+
+static __rte_always_inline void
+octeontx_fpa_bufpool_free(uintptr_t handle, void *buf)
+{
+	uint64_t free_addr = FPA_VF_FREE_ADDRS_S(FPA_VF_VHAURA_OP_FREE(0),
+						 0 /* DWB */, 1 /* FABS */);
+
+	fpavf_write64((uintptr_t)buf, (void *)(uintptr_t)(handle + free_addr));
+}
+
+static int
+octeontx_fpavf_enqueue(struct rte_mempool *mp, void * const *obj_table,
+			unsigned int n)
+{
+	uintptr_t pool;
+	unsigned int index;
+
+	pool = (uintptr_t)mp->pool_id;
+	/* Get pool bar address from handle */
+	pool &= ~(uint64_t)FPA_GPOOL_MASK;
+	for (index = 0; index < n; index++, obj_table++)
+		octeontx_fpa_bufpool_free(pool, *obj_table);
+
+	return 0;
+}
+
+static int
+octeontx_fpavf_dequeue(struct rte_mempool *mp, void **obj_table,
+			unsigned int n)
+{
+	unsigned int index;
+	uintptr_t pool;
+	void *obj;
+
+	pool = (uintptr_t)mp->pool_id;
+	/* Get pool bar address from handle */
+	pool &= ~(uint64_t)FPA_GPOOL_MASK;
+	for (index = 0; index < n; index++, obj_table++) {
+		obj = octeontx_fpa_bufpool_alloc(pool);
+		if (obj == NULL) {
+			/*
+			 * Failed to allocate the requested number of objects
+			 * from the pool. Current pool implementation requires
+			 * completing the entire request or returning error
+			 * otherwise.
+			 * Free already allocated buffers to the pool.
+			 */
+			for (; index > 0; index--) {
+				obj_table--;
+				octeontx_fpa_bufpool_free(pool, *obj_table);
+			}
+			return -ENOMEM;
+		}
+		*obj_table = obj;
+	}
+
+	return 0;
+}
+
+static unsigned int
+octeontx_fpavf_get_count(const struct rte_mempool *mp)
+{
+	uintptr_t pool;
+
+	pool = (uintptr_t)mp->pool_id;
+
+	return octeontx_fpa_bufpool_free_count(pool);
+}
+
+static int
+octeontx_fpavf_get_capabilities(const struct rte_mempool *mp,
+				unsigned int *flags)
+{
+	RTE_SET_USED(mp);
+	*flags |= (MEMPOOL_F_CAPA_PHYS_CONTIG |
+			MEMPOOL_F_CAPA_BLK_ALIGNED_OBJECTS);
+	return 0;
+}
+
+static int
+octeontx_fpavf_register_memory_area(const struct rte_mempool *mp,
+				    char *vaddr, rte_iova_t paddr, size_t len)
+{
+	struct octeontx_pool_info *pool_info;
+
+	RTE_SET_USED(paddr);
+	RTE_SET_USED(len);
+
+	pool_info = rte_malloc("octeontx_pool_info", sizeof(*pool_info), 0);
+	if (pool_info == NULL)
+		return -ENOMEM;
+
+	pool_info->mp = mp;
+	pool_info->mz_addr = (uintptr_t)vaddr;
+	rte_spinlock_lock(&pool_list_lock);
+	SLIST_INSERT_HEAD(&octeontx_pool_head, pool_info, link);
+	rte_spinlock_unlock(&pool_list_lock);
+	return 0;
+}
+
+static struct rte_mempool_ops octeontx_fpavf_ops = {
+	.name = "octeontx_fpavf",
+	.alloc = octeontx_fpavf_alloc,
+	.free = octeontx_fpavf_free,
+	.enqueue = octeontx_fpavf_enqueue,
+	.dequeue = octeontx_fpavf_dequeue,
+	.get_count = octeontx_fpavf_get_count,
+	.get_capabilities = octeontx_fpavf_get_capabilities,
+	.register_memory_area = octeontx_fpavf_register_memory_area,
+};
+
+MEMPOOL_REGISTER_OPS(octeontx_fpavf_ops);
diff --git a/drivers/mempool/octeontx/rte_mempool_octeontx_version.map b/drivers/mempool/octeontx/rte_mempool_octeontx_version.map
new file mode 100644
index 00000000..fe8cdeca
--- /dev/null
+++ b/drivers/mempool/octeontx/rte_mempool_octeontx_version.map
@@ -0,0 +1,9 @@
+DPDK_17.11 {
+	global:
+
+	octeontx_ssovf_info;
+	octeontx_ssovf_bar;
+	octeontx_ssovf_mbox_send;
+
+	local: *;
+};