mempool/octeontx: support allocation
authorSantosh Shukla <santosh.shukla@caviumnetworks.com>
Sun, 8 Oct 2017 12:40:05 +0000 (18:10 +0530)
committerThomas Monjalon <thomas@monjalon.net>
Sun, 8 Oct 2017 17:30:50 +0000 (19:30 +0200)
Upon pool allocation request by application, Octeontx FPA alloc
does following:
- Gets free pool from pci fpavf array.
- Uses mbox to communicate fpapf driver about,
  * gpool-id
  * pool block_sz
  * alignemnt
- Programs fpavf pool boundary.

Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
drivers/mempool/octeontx/Makefile
drivers/mempool/octeontx/octeontx_fpavf.c
drivers/mempool/octeontx/octeontx_fpavf.h
drivers/mempool/octeontx/rte_mempool_octeontx.c [new file with mode: 0644]

index 55ca1d9..9c33896 100644 (file)
@@ -51,6 +51,7 @@ LIBABIVER := 1
 # all source are stored in SRCS-y
 #
 SRCS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += octeontx_fpavf.c
+SRCS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += rte_mempool_octeontx.c
 
 # this lib depends upon:
 DEPDIRS-$(CONFIG_RTE_LIBRTE_OCTEONTX_MEMPOOL) += lib/librte_mbuf
index 0b4a935..c0c9d83 100644 (file)
 #include <rte_memory.h>
 #include <rte_malloc.h>
 #include <rte_spinlock.h>
+#include <rte_mbuf.h>
 
+#include <rte_pmd_octeontx_ssovf.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;
@@ -67,6 +133,454 @@ struct octeontx_fpadev {
 
 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];
+       if (unlikely(res == NULL))
+               return 0;
+
+       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;
+       phys_addr_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_virt2phy(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);
+}
+
+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);
+       FPAVF_STATIC_ASSERTION(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;
+}
+
 static void
 octeontx_fpavf_setup(void)
 {
index 33f0366..4da91f1 100644 (file)
@@ -58,6 +58,7 @@
 #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_VF0_APERTURE_SHIFT          22
 #define FPA_AURA_SET_SIZE              16
 
+#define FPA_MAX_OBJ_SIZE               (128 * 1024)
+#define OCTEONTX_FPAVF_BUF_OFFSET      128
+
+#define FPAVF_STATIC_ASSERTION(s) _Static_assert(s, #s)
 
 /*
  * In Cavium OcteonTX SoC, all accesses to the device registers are
@@ -126,4 +131,16 @@ do {                                                       \
 } 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_block_size(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/rte_mempool_octeontx.c b/drivers/mempool/octeontx/rte_mempool_octeontx.c
new file mode 100644 (file)
index 0000000..d930a81
--- /dev/null
@@ -0,0 +1,88 @@
+/*
+ *   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"
+
+static int
+octeontx_fpavf_alloc(struct rte_mempool *mp)
+{
+       uintptr_t pool;
+       uint32_t memseg_count = mp->size;
+       uint32_t object_size;
+       uintptr_t va_start;
+       int rc = 0;
+
+       /* virtual hugepage mapped addr */
+       va_start = ~(uint64_t)0;
+
+       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 struct rte_mempool_ops octeontx_fpavf_ops = {
+       .name = "octeontx_fpavf",
+       .alloc = octeontx_fpavf_alloc,
+       .free = NULL,
+       .enqueue = NULL,
+       .dequeue = NULL,
+       .get_count = NULL,
+       .get_capabilities = NULL,
+       .register_memory_area = NULL,
+};
+
+MEMPOOL_REGISTER_OPS(octeontx_fpavf_ops);