diff mbox series

[v2,35/41] net/mlx5: add pci probe and dev spawn support

Message ID 20211007184350.73858-36-srikanth.k@oneconvergence.com (mailing list archive)
State New
Delegated to: Raslan Darawsheh
Headers show
Series add MLX5 FreeBSD support | expand

Checks

Context Check Description
ci/checkpatch warning coding style issues

Commit Message

Srikanth Kaka Oct. 7, 2021, 6:43 p.m. UTC
add pci probe and dev spawn support

Signed-off-by: Srikanth Kaka <srikanth.k@oneconvergence.com>
Signed-off-by: Vag Singh <vag.singh@oneconvergence.com>
Signed-off-by: Anand Thulasiram <avelu@juniper.net>
---
 drivers/net/mlx5/freebsd/mlx5_os.c | 1836 +++++++++++++++++++++++++++-
 1 file changed, 1835 insertions(+), 1 deletion(-)
diff mbox series

Patch

diff --git a/drivers/net/mlx5/freebsd/mlx5_os.c b/drivers/net/mlx5/freebsd/mlx5_os.c
index 389b45b694..54bd32ee86 100644
--- a/drivers/net/mlx5/freebsd/mlx5_os.c
+++ b/drivers/net/mlx5/freebsd/mlx5_os.c
@@ -1,4 +1,3 @@ 
-
 /* SPDX-License-Identifier: BSD-3-Clause
  * Copyright 2015 6WIND S.A.
  * Copyright 2020 Mellanox Technologies, Ltd
@@ -10,11 +9,16 @@ 
 #include <stdint.h>
 #include <stdlib.h>
 #include <errno.h>
+#include <net/if.h>
 #include <fcntl.h>
 
 #include <rte_malloc.h>
 #include <ethdev_driver.h>
+#include <ethdev_pci.h>
+#include <rte_pci.h>
+#include <rte_bus_pci.h>
 #include <rte_common.h>
+#include <rte_kvargs.h>
 #include <rte_rwlock.h>
 #include <rte_spinlock.h>
 #include <rte_eal_paging.h>
@@ -22,18 +26,34 @@ 
 #include <mlx5_glue.h>
 #include <mlx5_devx_cmds.h>
 #include <mlx5_common.h>
+#include <mlx5_common_mp.h>
 #include <mlx5_common_mr.h>
+#include <mlx5_malloc.h>
 
 #include "mlx5_defs.h"
 #include "mlx5.h"
 #include "mlx5_common_os.h"
+#include "mlx5_utils.h"
 #include "mlx5_rxtx.h"
+#include "mlx5_rx.h"
+#include "mlx5_tx.h"
 #include "mlx5_autoconf.h"
 #include "mlx5_mr.h"
+#include "mlx5_flow.h"
+#include "rte_pmd_mlx5.h"
 #include "mlx5_verbs.h"
 #include "mlx5_inet.h"
 #include "mlx5_devx.h"
 
+#ifndef HAVE_IBV_MLX5_MOD_MPW
+#define MLX5DV_CONTEXT_FLAGS_MPW_ALLOWED (1 << 2)
+#define MLX5DV_CONTEXT_FLAGS_ENHANCED_MPW (1 << 3)
+#endif
+
+#ifndef HAVE_IBV_MLX5_MOD_CQE_128B_COMP
+#define MLX5DV_CONTEXT_FLAGS_CQE_128B_COMP (1 << 4)
+#endif
+
 static const char *MZ_MLX5_PMD_SHARED_DATA = "mlx5_pmd_shared_data";
 
 /* Spinlock for mlx5_shared_data allocation. */
@@ -42,6 +62,44 @@  static rte_spinlock_t mlx5_shared_data_lock = RTE_SPINLOCK_INITIALIZER;
 /* Process local data for secondary processes. */
 static struct mlx5_local_data mlx5_local_data;
 
+/* rte flow indexed pool configuration. */
+static struct mlx5_indexed_pool_config icfg[] = {
+	{
+		.size = sizeof(struct rte_flow),
+		.trunk_size = 64,
+		.need_lock = 1,
+		.release_mem_en = 0,
+		.malloc = mlx5_malloc,
+		.free = mlx5_free,
+		.per_core_cache = 0,
+		.type = "ctl_flow_ipool",
+	},
+	{
+		.size = sizeof(struct rte_flow),
+		.trunk_size = 64,
+		.grow_trunk = 3,
+		.grow_shift = 2,
+		.need_lock = 1,
+		.release_mem_en = 0,
+		.malloc = mlx5_malloc,
+		.free = mlx5_free,
+		.per_core_cache = 1 << 14,
+		.type = "rte_flow_ipool",
+	},
+	{
+		.size = sizeof(struct rte_flow),
+		.trunk_size = 64,
+		.grow_trunk = 3,
+		.grow_shift = 2,
+		.need_lock = 1,
+		.release_mem_en = 0,
+		.malloc = mlx5_malloc,
+		.free = mlx5_free,
+		.per_core_cache = 0,
+		.type = "mcp_flow_ipool",
+	},
+};
+
 /**
  * Get mlx5 device attributes. The glue function query_device_ex() is called
  * with out parameter of type 'struct ibv_device_attr_ex *'. Then fill in mlx5
@@ -145,6 +203,79 @@  mlx5_alloc_verbs_buf(size_t size, void *data)
 	return ret;
 }
 
+/**
+ * Detect misc5 support or not
+ *
+ * @param[in] priv
+ *   Device private data pointer
+ */
+#ifdef HAVE_MLX5DV_DR
+static void
+__mlx5_discovery_misc5_cap(struct mlx5_priv *priv)
+{
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+	/* Dummy VxLAN matcher to detect rdma-core misc5 cap
+	 * Case: IPv4--->UDP--->VxLAN--->vni
+	 */
+	void *tbl;
+	struct mlx5_flow_dv_match_params matcher_mask;
+	void *match_m;
+	void *matcher;
+	void *headers_m;
+	void *misc5_m;
+	uint32_t *tunnel_header_m;
+	struct mlx5dv_flow_matcher_attr dv_attr;
+
+	memset(&matcher_mask, 0, sizeof(matcher_mask));
+	matcher_mask.size = sizeof(matcher_mask.buf);
+	match_m = matcher_mask.buf;
+	headers_m = MLX5_ADDR_OF(fte_match_param, match_m, outer_headers);
+	misc5_m = MLX5_ADDR_OF(fte_match_param,
+			       match_m, misc_parameters_5);
+	tunnel_header_m = (uint32_t *)
+				MLX5_ADDR_OF(fte_match_set_misc5,
+				misc5_m, tunnel_header_1);
+	MLX5_SET(fte_match_set_lyr_2_4, headers_m, ip_protocol, 0xff);
+	MLX5_SET(fte_match_set_lyr_2_4, headers_m, ip_version, 4);
+	MLX5_SET(fte_match_set_lyr_2_4, headers_m, udp_dport, 0xffff);
+	*tunnel_header_m = 0xffffff;
+
+	tbl = mlx5_glue->dr_create_flow_tbl(priv->sh->rx_domain, 1);
+	if (!tbl) {
+		DRV_LOG(INFO, "No SW steering support");
+		return;
+	}
+	dv_attr.type = IBV_FLOW_ATTR_NORMAL,
+	dv_attr.match_mask = (void *)&matcher_mask,
+	dv_attr.match_criteria_enable =
+			(1 << MLX5_MATCH_CRITERIA_ENABLE_OUTER_BIT) |
+			(1 << MLX5_MATCH_CRITERIA_ENABLE_MISC5_BIT);
+	dv_attr.priority = 3;
+#ifdef HAVE_MLX5DV_DR_ESWITCH
+	void *misc2_m;
+	if (priv->config.dv_esw_en) {
+		/* FDB enabled reg_c_0 */
+		dv_attr.match_criteria_enable |=
+				(1 << MLX5_MATCH_CRITERIA_ENABLE_MISC2_BIT);
+		misc2_m = MLX5_ADDR_OF(fte_match_param,
+				       match_m, misc_parameters_2);
+		MLX5_SET(fte_match_set_misc2, misc2_m,
+			 metadata_reg_c_0, 0xffff);
+	}
+#endif
+	matcher = mlx5_glue->dv_create_flow_matcher(priv->sh->ctx,
+						    &dv_attr, tbl);
+	if (matcher) {
+		priv->sh->misc5_cap = 1;
+		mlx5_glue->dv_destroy_flow_matcher(matcher);
+	}
+	mlx5_glue->dr_destroy_flow_tbl(tbl);
+#else
+	RTE_SET_USED(priv);
+#endif
+}
+#endif
+
 /**
  * Verbs callback to free a memory.
  *
@@ -160,6 +291,282 @@  mlx5_free_verbs_buf(void *ptr, void *data __rte_unused)
 	mlx5_free(ptr);
 }
 
+/**
+ * Initialize DR related data within private structure.
+ * Routine checks the reference counter and does actual
+ * resources creation/initialization only if counter is zero.
+ *
+ * @param[in] priv
+ *   Pointer to the private device data structure.
+ *
+ * @return
+ *   Zero on success, positive error code otherwise.
+ */
+static int
+mlx5_alloc_shared_dr(struct mlx5_priv *priv)
+{
+	struct mlx5_dev_ctx_shared *sh = priv->sh;
+	char s[MLX5_NAME_SIZE] __rte_unused;
+	int err;
+
+	MLX5_ASSERT(sh && sh->refcnt);
+	if (sh->refcnt > 1)
+		return 0;
+	err = mlx5_alloc_table_hash_list(priv);
+	if (err)
+		goto error;
+	/* The resources below are only valid with DV support. */
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+	/* Init port id action list. */
+	snprintf(s, sizeof(s), "%s_port_id_action_list", sh->ibdev_name);
+	sh->port_id_action_list = mlx5_list_create(s, sh, true,
+						   flow_dv_port_id_create_cb,
+						   flow_dv_port_id_match_cb,
+						   flow_dv_port_id_remove_cb,
+						   flow_dv_port_id_clone_cb,
+						 flow_dv_port_id_clone_free_cb);
+	if (!sh->port_id_action_list)
+		goto error;
+	/* Init push vlan action list. */
+	snprintf(s, sizeof(s), "%s_push_vlan_action_list", sh->ibdev_name);
+	sh->push_vlan_action_list = mlx5_list_create(s, sh, true,
+						    flow_dv_push_vlan_create_cb,
+						    flow_dv_push_vlan_match_cb,
+						    flow_dv_push_vlan_remove_cb,
+						    flow_dv_push_vlan_clone_cb,
+					       flow_dv_push_vlan_clone_free_cb);
+	if (!sh->push_vlan_action_list)
+		goto error;
+	/* Init sample action list. */
+	snprintf(s, sizeof(s), "%s_sample_action_list", sh->ibdev_name);
+	sh->sample_action_list = mlx5_list_create(s, sh, true,
+						  flow_dv_sample_create_cb,
+						  flow_dv_sample_match_cb,
+						  flow_dv_sample_remove_cb,
+						  flow_dv_sample_clone_cb,
+						  flow_dv_sample_clone_free_cb);
+	if (!sh->sample_action_list)
+		goto error;
+	/* Init dest array action list. */
+	snprintf(s, sizeof(s), "%s_dest_array_list", sh->ibdev_name);
+	sh->dest_array_list = mlx5_list_create(s, sh, true,
+					       flow_dv_dest_array_create_cb,
+					       flow_dv_dest_array_match_cb,
+					       flow_dv_dest_array_remove_cb,
+					       flow_dv_dest_array_clone_cb,
+					      flow_dv_dest_array_clone_free_cb);
+	if (!sh->dest_array_list)
+		goto error;
+#endif
+#ifdef HAVE_MLX5DV_DR
+	void *domain;
+
+	/* Reference counter is zero, we should initialize structures. */
+	domain = mlx5_glue->dr_create_domain(sh->ctx,
+					     MLX5DV_DR_DOMAIN_TYPE_NIC_RX);
+	if (!domain) {
+		DRV_LOG(ERR, "ingress mlx5dv_dr_create_domain failed");
+		err = errno;
+		goto error;
+	}
+	sh->rx_domain = domain;
+	domain = mlx5_glue->dr_create_domain(sh->ctx,
+					     MLX5DV_DR_DOMAIN_TYPE_NIC_TX);
+	if (!domain) {
+		DRV_LOG(ERR, "egress mlx5dv_dr_create_domain failed");
+		err = errno;
+		goto error;
+	}
+	sh->tx_domain = domain;
+#ifdef HAVE_MLX5DV_DR_ESWITCH
+	if (priv->config.dv_esw_en) {
+		domain  = mlx5_glue->dr_create_domain
+			(sh->ctx, MLX5DV_DR_DOMAIN_TYPE_FDB);
+		if (!domain) {
+			DRV_LOG(ERR, "FDB mlx5dv_dr_create_domain failed");
+			err = errno;
+			goto error;
+		}
+		sh->fdb_domain = domain;
+	}
+	/*
+	 * The drop action is just some dummy placeholder in rdma-core. It
+	 * does not belong to domains and has no any attributes, and, can be
+	 * shared by the entire device.
+	 */
+	sh->dr_drop_action = mlx5_glue->dr_create_flow_action_drop();
+	if (!sh->dr_drop_action) {
+		DRV_LOG(ERR, "FDB mlx5dv_dr_create_flow_action_drop");
+		err = errno;
+		goto error;
+	}
+#endif
+	if (!sh->tunnel_hub && priv->config.dv_miss_info)
+		err = mlx5_alloc_tunnel_hub(sh);
+	if (err) {
+		DRV_LOG(ERR, "mlx5_alloc_tunnel_hub failed err=%d", err);
+		goto error;
+	}
+	if (priv->config.reclaim_mode == MLX5_RCM_AGGR) {
+		mlx5_glue->dr_reclaim_domain_memory(sh->rx_domain, 1);
+		mlx5_glue->dr_reclaim_domain_memory(sh->tx_domain, 1);
+		if (sh->fdb_domain)
+			mlx5_glue->dr_reclaim_domain_memory(sh->fdb_domain, 1);
+	}
+	sh->pop_vlan_action = mlx5_glue->dr_create_flow_action_pop_vlan();
+	if (!priv->config.allow_duplicate_pattern) {
+#ifndef HAVE_MLX5_DR_ALLOW_DUPLICATE
+		DRV_LOG(WARNING, "Disallow duplicate pattern is not supported - maybe old rdma-core version?");
+#endif
+		mlx5_glue->dr_allow_duplicate_rules(sh->rx_domain, 0);
+		mlx5_glue->dr_allow_duplicate_rules(sh->tx_domain, 0);
+		if (sh->fdb_domain)
+			mlx5_glue->dr_allow_duplicate_rules(sh->fdb_domain, 0);
+	}
+
+	__mlx5_discovery_misc5_cap(priv);
+#endif /* HAVE_MLX5DV_DR */
+	sh->default_miss_action =
+			mlx5_glue->dr_create_flow_action_default_miss();
+	if (!sh->default_miss_action)
+		DRV_LOG(WARNING, "Default miss action is not supported.");
+	return 0;
+error:
+	/* Rollback the created objects. */
+	if (sh->rx_domain) {
+		mlx5_glue->dr_destroy_domain(sh->rx_domain);
+		sh->rx_domain = NULL;
+	}
+	if (sh->tx_domain) {
+		mlx5_glue->dr_destroy_domain(sh->tx_domain);
+		sh->tx_domain = NULL;
+	}
+	if (sh->fdb_domain) {
+		mlx5_glue->dr_destroy_domain(sh->fdb_domain);
+		sh->fdb_domain = NULL;
+	}
+	if (sh->dr_drop_action) {
+		mlx5_glue->destroy_flow_action(sh->dr_drop_action);
+		sh->dr_drop_action = NULL;
+	}
+	if (sh->pop_vlan_action) {
+		mlx5_glue->destroy_flow_action(sh->pop_vlan_action);
+		sh->pop_vlan_action = NULL;
+	}
+	if (sh->encaps_decaps) {
+		mlx5_hlist_destroy(sh->encaps_decaps);
+		sh->encaps_decaps = NULL;
+	}
+	if (sh->modify_cmds) {
+		mlx5_hlist_destroy(sh->modify_cmds);
+		sh->modify_cmds = NULL;
+	}
+	if (sh->tag_table) {
+		/* tags should be destroyed with flow before. */
+		mlx5_hlist_destroy(sh->tag_table);
+		sh->tag_table = NULL;
+	}
+	if (sh->tunnel_hub) {
+		mlx5_release_tunnel_hub(sh, priv->dev_port);
+		sh->tunnel_hub = NULL;
+	}
+	mlx5_free_table_hash_list(priv);
+	if (sh->port_id_action_list) {
+		mlx5_list_destroy(sh->port_id_action_list);
+		sh->port_id_action_list = NULL;
+	}
+	if (sh->push_vlan_action_list) {
+		mlx5_list_destroy(sh->push_vlan_action_list);
+		sh->push_vlan_action_list = NULL;
+	}
+	if (sh->sample_action_list) {
+		mlx5_list_destroy(sh->sample_action_list);
+		sh->sample_action_list = NULL;
+	}
+	if (sh->dest_array_list) {
+		mlx5_list_destroy(sh->dest_array_list);
+		sh->dest_array_list = NULL;
+	}
+	return err;
+}
+
+/**
+ * Destroy DR related data within private structure.
+ *
+ * @param[in] priv
+ *   Pointer to the private device data structure.
+ */
+void
+mlx5_os_free_shared_dr(struct mlx5_priv *priv)
+{
+	struct mlx5_dev_ctx_shared *sh = priv->sh;
+
+	MLX5_ASSERT(sh && sh->refcnt);
+	if (sh->refcnt > 1)
+		return;
+#ifdef HAVE_MLX5DV_DR
+	if (sh->rx_domain) {
+		mlx5_glue->dr_destroy_domain(sh->rx_domain);
+		sh->rx_domain = NULL;
+	}
+	if (sh->tx_domain) {
+		mlx5_glue->dr_destroy_domain(sh->tx_domain);
+		sh->tx_domain = NULL;
+	}
+#ifdef HAVE_MLX5DV_DR_ESWITCH
+	if (sh->fdb_domain) {
+		mlx5_glue->dr_destroy_domain(sh->fdb_domain);
+		sh->fdb_domain = NULL;
+	}
+	if (sh->dr_drop_action) {
+		mlx5_glue->destroy_flow_action(sh->dr_drop_action);
+		sh->dr_drop_action = NULL;
+	}
+#endif
+	if (sh->pop_vlan_action) {
+		mlx5_glue->destroy_flow_action(sh->pop_vlan_action);
+		sh->pop_vlan_action = NULL;
+	}
+#endif /* HAVE_MLX5DV_DR */
+	if (sh->default_miss_action)
+		mlx5_glue->destroy_flow_action
+				(sh->default_miss_action);
+	if (sh->encaps_decaps) {
+		mlx5_hlist_destroy(sh->encaps_decaps);
+		sh->encaps_decaps = NULL;
+	}
+	if (sh->modify_cmds) {
+		mlx5_hlist_destroy(sh->modify_cmds);
+		sh->modify_cmds = NULL;
+	}
+	if (sh->tag_table) {
+		/* tags should be destroyed with flow before. */
+		mlx5_hlist_destroy(sh->tag_table);
+		sh->tag_table = NULL;
+	}
+	if (sh->tunnel_hub) {
+		mlx5_release_tunnel_hub(sh, priv->dev_port);
+		sh->tunnel_hub = NULL;
+	}
+	mlx5_free_table_hash_list(priv);
+	if (sh->port_id_action_list) {
+		mlx5_list_destroy(sh->port_id_action_list);
+		sh->port_id_action_list = NULL;
+	}
+	if (sh->push_vlan_action_list) {
+		mlx5_list_destroy(sh->push_vlan_action_list);
+		sh->push_vlan_action_list = NULL;
+	}
+	if (sh->sample_action_list) {
+		mlx5_list_destroy(sh->sample_action_list);
+		sh->sample_action_list = NULL;
+	}
+	if (sh->dest_array_list) {
+		mlx5_list_destroy(sh->dest_array_list);
+		sh->dest_array_list = NULL;
+	}
+}
+
 /**
  * Initialize shared data between primary and secondary process.
  *
@@ -263,6 +670,1429 @@  mlx5_init_once(void)
 	return ret;
 }
 
+/**
+ * Create the Tx queue DevX/Verbs object.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ * @param idx
+ *   Queue index in DPDK Tx queue array.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_os_txq_obj_new(struct rte_eth_dev *dev, uint16_t idx)
+{
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_txq_data *txq_data = (*priv->txqs)[idx];
+	struct mlx5_txq_ctrl *txq_ctrl =
+			container_of(txq_data, struct mlx5_txq_ctrl, txq);
+
+	if (txq_ctrl->type == MLX5_TXQ_TYPE_HAIRPIN)
+		return mlx5_txq_devx_obj_new(dev, idx);
+#ifdef HAVE_MLX5DV_DEVX_UAR_OFFSET
+	if (!priv->config.dv_esw_en)
+		return mlx5_txq_devx_obj_new(dev, idx);
+#endif
+	return mlx5_txq_ibv_obj_new(dev, idx);
+}
+
+/**
+ * Release an Tx DevX/verbs queue object.
+ *
+ * @param txq_obj
+ *   DevX/Verbs Tx queue object.
+ */
+static void
+mlx5_os_txq_obj_release(struct mlx5_txq_obj *txq_obj)
+{
+	if (txq_obj->txq_ctrl->type == MLX5_TXQ_TYPE_HAIRPIN) {
+		mlx5_txq_devx_obj_release(txq_obj);
+		return;
+	}
+#ifdef HAVE_MLX5DV_DEVX_UAR_OFFSET
+	if (!txq_obj->txq_ctrl->priv->config.dv_esw_en) {
+		mlx5_txq_devx_obj_release(txq_obj);
+		return;
+	}
+#endif
+	mlx5_txq_ibv_obj_release(txq_obj);
+}
+
+/**
+ * DV flow counter mode detect and config.
+ *
+ * @param dev
+ *   Pointer to rte_eth_dev structure.
+ *
+ */
+static void
+mlx5_flow_counter_mode_config(struct rte_eth_dev *dev __rte_unused)
+{
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_dev_ctx_shared *sh = priv->sh;
+	bool fallback;
+
+#ifndef HAVE_IBV_DEVX_ASYNC
+	fallback = true;
+#else
+	fallback = false;
+	if (!priv->config.devx || !priv->config.dv_flow_en ||
+	    !priv->config.hca_attr.flow_counters_dump ||
+	    !(priv->config.hca_attr.flow_counter_bulk_alloc_bitmap & 0x4) ||
+	    (mlx5_flow_dv_discover_counter_offset_support(dev) == -ENOTSUP))
+		fallback = true;
+#endif
+	if (fallback)
+		DRV_LOG(INFO, "Use fall-back DV counter management. Flow "
+			"counter dump:%d, bulk_alloc_bitmap:0x%hhx.",
+			priv->config.hca_attr.flow_counters_dump,
+			priv->config.hca_attr.flow_counter_bulk_alloc_bitmap);
+	/* Initialize fallback mode only on the port initializes sh. */
+	if (sh->refcnt == 1)
+		sh->cmng.counter_fallback = fallback;
+	else if (fallback != sh->cmng.counter_fallback)
+		DRV_LOG(WARNING, "Port %d in sh has different fallback mode "
+			"with others:%d.", PORT_ID(priv), fallback);
+#endif
+}
+
+/**
+ * DR flow drop action support detect.
+ *
+ * @param dev
+ *   Pointer to rte_eth_dev structure.
+ *
+ */
+static void
+mlx5_flow_drop_action_config(struct rte_eth_dev *dev __rte_unused)
+{
+#ifdef HAVE_MLX5DV_DR
+	struct mlx5_priv *priv = dev->data->dev_private;
+
+	if (!priv->config.dv_flow_en || !priv->sh->dr_drop_action)
+		return;
+	/**
+	 * DR supports drop action placeholder when it is supported;
+	 * otherwise, use the queue drop action.
+	 */
+	if (mlx5_flow_discover_dr_action_support(dev))
+		priv->root_drop_action = priv->drop_queue.hrxq->action;
+	else
+		priv->root_drop_action = priv->sh->dr_drop_action;
+#endif
+}
+
+static void
+mlx5_queue_counter_id_prepare(struct rte_eth_dev *dev)
+{
+	struct mlx5_priv *priv = dev->data->dev_private;
+	void *ctx = priv->sh->ctx;
+
+	priv->q_counters = mlx5_devx_cmd_queue_counter_alloc(ctx);
+	if (!priv->q_counters) {
+		struct ibv_cq *cq = mlx5_glue->create_cq(ctx, 1, NULL, NULL, 0);
+		struct ibv_wq *wq;
+
+		DRV_LOG(DEBUG, "Port %d queue counter object cannot be created "
+			"by DevX - fall-back to use the kernel driver global "
+			"queue counter.", dev->data->port_id);
+		/* Create WQ by kernel and query its queue counter ID. */
+		if (cq) {
+			wq = mlx5_glue->create_wq(ctx,
+						  &(struct ibv_wq_init_attr){
+						    .wq_type = IBV_WQT_RQ,
+						    .max_wr = 1,
+						    .max_sge = 1,
+						    .pd = priv->sh->pd,
+						    .cq = cq,
+						});
+			if (wq) {
+				/* Counter is assigned only on RDY state. */
+				int ret = mlx5_glue->modify_wq(wq,
+						 &(struct ibv_wq_attr){
+						 .attr_mask = IBV_WQ_ATTR_STATE,
+						 .wq_state = IBV_WQS_RDY,
+						});
+
+				if (ret == 0)
+					mlx5_devx_cmd_wq_query(wq,
+							 &priv->counter_set_id);
+				claim_zero(mlx5_glue->destroy_wq(wq));
+			}
+			claim_zero(mlx5_glue->destroy_cq(cq));
+		}
+	} else {
+		priv->counter_set_id = priv->q_counters->id;
+	}
+	if (priv->counter_set_id == 0)
+		DRV_LOG(INFO, "Part of the port %d statistics will not be "
+			"available.", dev->data->port_id);
+}
+
+/**
+ * Spawn an Ethernet device from Verbs information.
+ *
+ * @param dpdk_dev
+ *   Backing DPDK device.
+ * @param spawn
+ *   Verbs device parameters (name, port, switch_info) to spawn.
+ * @param config
+ *   Device configuration parameters.
+ * @param config
+ *   Device arguments.
+ *
+ * @return
+ *   A valid Ethernet device object on success, NULL otherwise and rte_errno
+ *   is set. The following errors are defined:
+ *
+ *   EBUSY: device is not supposed to be spawned.
+ *   EEXIST: device is already spawned
+ */
+static struct rte_eth_dev *
+mlx5_dev_spawn(struct rte_device *dpdk_dev,
+	       struct mlx5_dev_spawn_data *spawn,
+	       struct mlx5_dev_config *config)
+{
+	const struct mlx5_switch_info *switch_info = &spawn->info;
+	struct mlx5_dev_ctx_shared *sh = NULL;
+	struct ibv_port_attr port_attr;
+	struct mlx5dv_context dv_attr = { .comp_mask = 0 };
+	struct rte_eth_dev *eth_dev = NULL;
+	struct mlx5_priv *priv = NULL;
+	int err = 0;
+	unsigned int hw_padding = 0;
+	unsigned int mps;
+	unsigned int tunnel_en = 0;
+	unsigned int mpls_en = 0;
+	unsigned int swp = 0;
+	unsigned int mprq = 0;
+	unsigned int mprq_min_stride_size_n = 0;
+	unsigned int mprq_max_stride_size_n = 0;
+	unsigned int mprq_min_stride_num_n = 0;
+	unsigned int mprq_max_stride_num_n = 0;
+	struct rte_ether_addr mac;
+	char name[RTE_ETH_NAME_MAX_LEN];
+	int own_domain_id = 0;
+	uint16_t port_id;
+	struct mlx5_port_info vport_info = { .query_flags = 0 };
+	int i;
+
+	MLX5_ASSERT(spawn->pf_bond < 0);
+	MLX5_ASSERT(switch_info->representor == 0);
+	/* Build device name. */
+	strlcpy(name, dpdk_dev->name, sizeof(name));
+
+	/* check if the device is already spawned */
+	if (rte_eth_dev_get_port_by_name(name, &port_id) == 0) {
+		rte_errno = EEXIST;
+		return NULL;
+	}
+	DRV_LOG(DEBUG, "naming Ethernet device \"%s\"", name);
+	if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
+		struct mlx5_mp_id mp_id;
+
+		eth_dev = rte_eth_dev_attach_secondary(name);
+		if (eth_dev == NULL) {
+			DRV_LOG(ERR, "can not attach rte ethdev");
+			rte_errno = ENOMEM;
+			return NULL;
+		}
+		eth_dev->device = dpdk_dev;
+		eth_dev->dev_ops = &mlx5_dev_sec_ops;
+		eth_dev->rx_descriptor_status = mlx5_rx_descriptor_status;
+		eth_dev->tx_descriptor_status = mlx5_tx_descriptor_status;
+		err = mlx5_proc_priv_init(eth_dev);
+		if (err)
+			return NULL;
+		mp_id.port_id = eth_dev->data->port_id;
+		strlcpy(mp_id.name, MLX5_MP_NAME, RTE_MP_MAX_NAME_LEN);
+		/* Receive command fd from primary process */
+		err = mlx5_mp_req_verbs_cmd_fd(&mp_id);
+		if (err < 0)
+			goto err_secondary;
+		/* Remap UAR for Tx queues. */
+		err = mlx5_tx_uar_init_secondary(eth_dev, err);
+		if (err)
+			goto err_secondary;
+		/*
+		 * Ethdev pointer is still required as input since
+		 * the primary device is not accessible from the
+		 * secondary process.
+		 */
+		eth_dev->rx_pkt_burst = mlx5_select_rx_function(eth_dev);
+		eth_dev->tx_pkt_burst = mlx5_select_tx_function(eth_dev);
+		return eth_dev;
+err_secondary:
+		mlx5_dev_close(eth_dev);
+		return NULL;
+	}
+	/*
+	 * Some parameters ("tx_db_nc" in particularly) are needed in
+	 * advance to create dv/verbs device context. We proceed the
+	 * devargs here to get ones, and later proceed devargs again
+	 * to override some hardware settings.
+	 */
+	err = mlx5_args(config, dpdk_dev->devargs);
+	if (err) {
+		err = rte_errno;
+		DRV_LOG(ERR, "failed to process device arguments: %s",
+			strerror(rte_errno));
+		goto error;
+	}
+	if (config->dv_miss_info) {
+		if (switch_info->master || switch_info->representor)
+			config->dv_xmeta_en = MLX5_XMETA_MODE_META16;
+	}
+	mlx5_malloc_mem_select(config->sys_mem_en);
+	sh = mlx5_alloc_shared_dev_ctx(spawn, config);
+	if (!sh)
+		return NULL;
+	config->devx = sh->devx;
+#ifdef HAVE_MLX5DV_DR_ACTION_DEST_DEVX_TIR
+	config->dest_tir = 1;
+#endif
+#ifdef HAVE_IBV_MLX5_MOD_SWP
+	dv_attr.comp_mask |= MLX5DV_CONTEXT_MASK_SWP;
+#endif
+	/*
+	 * Multi-packet send is supported by ConnectX-4 Lx PF as well
+	 * as all ConnectX-5 devices.
+	 */
+#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT
+	dv_attr.comp_mask |= MLX5DV_CONTEXT_MASK_TUNNEL_OFFLOADS;
+#endif
+#ifdef HAVE_IBV_DEVICE_STRIDING_RQ_SUPPORT
+	dv_attr.comp_mask |= MLX5DV_CONTEXT_MASK_STRIDING_RQ;
+#endif
+	mlx5_glue->dv_query_device(sh->ctx, &dv_attr);
+	if (dv_attr.flags & MLX5DV_CONTEXT_FLAGS_MPW_ALLOWED) {
+		if (dv_attr.flags & MLX5DV_CONTEXT_FLAGS_ENHANCED_MPW) {
+			DRV_LOG(DEBUG, "enhanced MPW is supported");
+			mps = MLX5_MPW_ENHANCED;
+		} else {
+			DRV_LOG(DEBUG, "MPW is supported");
+			mps = MLX5_MPW;
+		}
+	} else {
+		DRV_LOG(DEBUG, "MPW isn't supported");
+		mps = MLX5_MPW_DISABLED;
+	}
+#ifdef HAVE_IBV_MLX5_MOD_SWP
+	if (dv_attr.comp_mask & MLX5DV_CONTEXT_MASK_SWP)
+		swp = dv_attr.sw_parsing_caps.sw_parsing_offloads;
+	DRV_LOG(DEBUG, "SWP support: %u", swp);
+#endif
+	config->swp = !!swp;
+#ifdef HAVE_IBV_DEVICE_STRIDING_RQ_SUPPORT
+	if (dv_attr.comp_mask & MLX5DV_CONTEXT_MASK_STRIDING_RQ) {
+		struct mlx5dv_striding_rq_caps mprq_caps =
+			dv_attr.striding_rq_caps;
+
+		DRV_LOG(DEBUG, "\tmin_single_stride_log_num_of_bytes: %d",
+			mprq_caps.min_single_stride_log_num_of_bytes);
+		DRV_LOG(DEBUG, "\tmax_single_stride_log_num_of_bytes: %d",
+			mprq_caps.max_single_stride_log_num_of_bytes);
+		DRV_LOG(DEBUG, "\tmin_single_wqe_log_num_of_strides: %d",
+			mprq_caps.min_single_wqe_log_num_of_strides);
+		DRV_LOG(DEBUG, "\tmax_single_wqe_log_num_of_strides: %d",
+			mprq_caps.max_single_wqe_log_num_of_strides);
+		DRV_LOG(DEBUG, "\tsupported_qpts: %d",
+			mprq_caps.supported_qpts);
+		DRV_LOG(DEBUG, "device supports Multi-Packet RQ");
+		mprq = 1;
+		mprq_min_stride_size_n =
+			mprq_caps.min_single_stride_log_num_of_bytes;
+		mprq_max_stride_size_n =
+			mprq_caps.max_single_stride_log_num_of_bytes;
+		mprq_min_stride_num_n =
+			mprq_caps.min_single_wqe_log_num_of_strides;
+		mprq_max_stride_num_n =
+			mprq_caps.max_single_wqe_log_num_of_strides;
+	}
+#endif
+	/* Rx CQE compression is enabled by default. */
+	config->cqe_comp = 1;
+#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT
+	if (dv_attr.comp_mask & MLX5DV_CONTEXT_MASK_TUNNEL_OFFLOADS) {
+		tunnel_en = ((dv_attr.tunnel_offloads_caps &
+			      MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_VXLAN) &&
+			     (dv_attr.tunnel_offloads_caps &
+			      MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_GRE) &&
+			     (dv_attr.tunnel_offloads_caps &
+			      MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_GENEVE));
+	}
+	DRV_LOG(DEBUG, "tunnel offloading is %ssupported",
+		tunnel_en ? "" : "not ");
+#else
+	DRV_LOG(WARNING,
+		"tunnel offloading disabled due to old OFED/rdma-core version");
+#endif
+	config->tunnel_en = tunnel_en;
+#ifdef HAVE_IBV_DEVICE_MPLS_SUPPORT
+	mpls_en = ((dv_attr.tunnel_offloads_caps &
+		    MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_CW_MPLS_OVER_GRE) &&
+		   (dv_attr.tunnel_offloads_caps &
+		    MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_CW_MPLS_OVER_UDP));
+	DRV_LOG(DEBUG, "MPLS over GRE/UDP tunnel offloading is %ssupported",
+		mpls_en ? "" : "not ");
+#else
+	DRV_LOG(WARNING, "MPLS over GRE/UDP tunnel offloading disabled due to"
+		" old OFED/rdma-core version or firmware configuration");
+#endif
+	config->mpls_en = mpls_en;
+	/* Check port status. */
+	err = mlx5_glue->query_port(sh->ctx, spawn->phys_port, &port_attr);
+	if (err) {
+		DRV_LOG(ERR, "port query failed: %s", strerror(err));
+		goto error;
+	}
+	if (port_attr.link_layer != IBV_LINK_LAYER_ETHERNET) {
+		DRV_LOG(ERR, "port is not configured in Ethernet mode");
+		err = EINVAL;
+		goto error;
+	}
+	if (port_attr.state != IBV_PORT_ACTIVE)
+		DRV_LOG(DEBUG, "port is not active: \"%s\" (%d)",
+			mlx5_glue->port_state_str(port_attr.state),
+			port_attr.state);
+	/* Allocate private eth device data. */
+	priv = mlx5_malloc(MLX5_MEM_ZERO | MLX5_MEM_RTE,
+			   sizeof(*priv),
+			   RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+	if (priv == NULL) {
+		DRV_LOG(ERR, "priv allocation failure");
+		err = ENOMEM;
+		goto error;
+	}
+	priv->sh = sh;
+	priv->dev_port = spawn->phys_port;
+	priv->pci_dev = spawn->pci_dev;
+	priv->mtu = RTE_ETHER_MTU;
+	/* RDMA core has no listener */
+	priv->nl_socket_rdma = -1;
+	priv->nl_socket_route =	mlx5_inet_init();
+	priv->representor = !!switch_info->representor;
+	priv->master = !!switch_info->master;
+	priv->domain_id = RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID;
+	priv->vport_meta_tag = 0;
+	priv->vport_meta_mask = 0;
+	priv->pf_bond = spawn->pf_bond;
+
+	DRV_LOG(DEBUG,
+		"dev_port=%u bus=%s pci=%s master=%d representor=%d pf_bond=%d\n",
+		priv->dev_port, dpdk_dev->bus->name,
+		priv->pci_dev ? priv->pci_dev->name : "NONE",
+		priv->master, priv->representor, priv->pf_bond);
+
+	/*
+	 * If we have E-Switch we should determine the vport attributes.
+	 * E-Switch may use either source vport field or reg_c[0] metadata
+	 * register to match on vport index. The engaged part of metadata
+	 * register is defined by mask.
+	 */
+	if (switch_info->representor || switch_info->master) {
+		err = mlx5_glue->devx_port_query(sh->ctx,
+						 spawn->phys_port,
+						 &vport_info);
+		if (err) {
+			DRV_LOG(WARNING,
+				"can't query devx port %d on device %s",
+				spawn->phys_port,
+				mlx5_os_get_dev_device_name(spawn->phys_dev));
+			vport_info.query_flags = 0;
+		}
+	}
+	if (vport_info.query_flags & MLX5_PORT_QUERY_REG_C0) {
+		priv->vport_meta_tag = vport_info.vport_meta_tag;
+		priv->vport_meta_mask = vport_info.vport_meta_mask;
+		if (!priv->vport_meta_mask) {
+			DRV_LOG(ERR, "vport zero mask for port %d"
+				     " on bonding device %s",
+				     spawn->phys_port,
+				     mlx5_os_get_dev_device_name
+							(spawn->phys_dev));
+			err = ENOTSUP;
+			goto error;
+		}
+		if (priv->vport_meta_tag & ~priv->vport_meta_mask) {
+			DRV_LOG(ERR, "invalid vport tag for port %d"
+				     " on bonding device %s",
+				     spawn->phys_port,
+				     mlx5_os_get_dev_device_name
+							(spawn->phys_dev));
+			err = ENOTSUP;
+			goto error;
+		}
+	}
+	if (vport_info.query_flags & MLX5_PORT_QUERY_VPORT) {
+		priv->vport_id = vport_info.vport_id;
+	} else if (spawn->pf_bond >= 0 &&
+		   (switch_info->representor || switch_info->master)) {
+		DRV_LOG(ERR, "can't deduce vport index for port %d"
+			     " on bonding device %s",
+			     spawn->phys_port,
+			     mlx5_os_get_dev_device_name(spawn->phys_dev));
+		err = ENOTSUP;
+		goto error;
+	} else {
+		/*
+		 * Suppose vport index in compatible way. Kernel/rdma_core
+		 * support single E-Switch per PF configurations only and
+		 * vport_id field contains the vport index for associated VF,
+		 * which is deduced from representor port name.
+		 * For example, let's have the IB device port 10, it has
+		 * attached network device eth0, which has port name attribute
+		 * pf0vf2, we can deduce the VF number as 2, and set vport index
+		 * as 3 (2+1). This assigning schema should be changed if the
+		 * multiple E-Switch instances per PF configurations or/and PCI
+		 * subfunctions are added.
+		 */
+		priv->vport_id = switch_info->representor ?
+				 switch_info->port_name + 1 : -1;
+	}
+	priv->representor_id = mlx5_representor_id_encode(switch_info,
+						RTE_ETH_REPRESENTOR_NONE);
+	/*
+	 * Look for sibling devices in order to reuse their switch domain
+	 * if any, otherwise allocate one.
+	 */
+	MLX5_ETH_FOREACH_DEV(port_id, dpdk_dev) {
+		const struct mlx5_priv *opriv =
+			rte_eth_devices[port_id].data->dev_private;
+
+		if (!opriv ||
+		    opriv->sh != priv->sh ||
+			opriv->domain_id ==
+			RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID)
+			continue;
+		priv->domain_id = opriv->domain_id;
+		DRV_LOG(DEBUG, "dev_port-%u inherit domain_id=%u\n",
+			priv->dev_port, priv->domain_id);
+		break;
+	}
+	if (priv->domain_id == RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID) {
+		err = rte_eth_switch_domain_alloc(&priv->domain_id);
+		if (err) {
+			err = rte_errno;
+			DRV_LOG(ERR, "unable to allocate switch domain: %s",
+				strerror(rte_errno));
+			goto error;
+		}
+		own_domain_id = 1;
+		DRV_LOG(DEBUG, "dev_port-%u new domain_id=%u\n",
+			priv->dev_port, priv->domain_id);
+	}
+	/* Override some values set by hardware configuration. */
+	mlx5_args(config, dpdk_dev->devargs);
+	err = mlx5_dev_check_sibling_config(priv, config, dpdk_dev);
+	if (err)
+		goto error;
+	config->hw_csum = !!(sh->device_attr.device_cap_flags_ex &
+			    IBV_DEVICE_RAW_IP_CSUM);
+	DRV_LOG(DEBUG, "checksum offloading is %ssupported",
+		(config->hw_csum ? "" : "not "));
+#if !defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42) && \
+	!defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
+	DRV_LOG(DEBUG, "counters are not supported");
+#endif
+#if !defined(HAVE_IBV_FLOW_DV_SUPPORT) || !defined(HAVE_MLX5DV_DR)
+	if (config->dv_flow_en) {
+		DRV_LOG(WARNING, "DV flow is not supported");
+		config->dv_flow_en = 0;
+	}
+#endif
+	if (spawn->max_port > UINT8_MAX) {
+		/* Verbs can't support ports larger than 255 by design. */
+		DRV_LOG(ERR, "can't support IB ports > UINT8_MAX");
+		err = EINVAL;
+		goto error;
+	}
+	config->ind_table_max_size =
+		sh->device_attr.max_rwq_indirection_table_size;
+	/*
+	 * Remove this check once DPDK supports larger/variable
+	 * indirection tables.
+	 */
+	if (config->ind_table_max_size > (unsigned int)ETH_RSS_RETA_SIZE_512)
+		config->ind_table_max_size = ETH_RSS_RETA_SIZE_512;
+	DRV_LOG(DEBUG, "maximum Rx indirection table size is %u",
+		config->ind_table_max_size);
+	config->hw_vlan_strip = !!(sh->device_attr.raw_packet_caps &
+				  IBV_RAW_PACKET_CAP_CVLAN_STRIPPING);
+	DRV_LOG(DEBUG, "VLAN stripping is %ssupported",
+		(config->hw_vlan_strip ? "" : "not "));
+	config->hw_fcs_strip = !!(sh->device_attr.raw_packet_caps &
+				 IBV_RAW_PACKET_CAP_SCATTER_FCS);
+#if defined(HAVE_IBV_WQ_FLAG_RX_END_PADDING)
+	hw_padding = !!sh->device_attr.rx_pad_end_addr_align;
+#elif defined(HAVE_IBV_WQ_FLAGS_PCI_WRITE_END_PADDING)
+	hw_padding = !!(sh->device_attr.device_cap_flags_ex &
+			IBV_DEVICE_PCI_WRITE_END_PADDING);
+#endif
+	if (config->hw_padding && !hw_padding) {
+		DRV_LOG(DEBUG, "Rx end alignment padding isn't supported");
+		config->hw_padding = 0;
+	} else if (config->hw_padding) {
+		DRV_LOG(DEBUG, "Rx end alignment padding is enabled");
+	}
+	config->tso = (sh->device_attr.max_tso > 0 &&
+		      (sh->device_attr.tso_supported_qpts &
+		       (1 << IBV_QPT_RAW_PACKET)));
+	if (config->tso)
+		config->tso_max_payload_sz = sh->device_attr.max_tso;
+	/*
+	 * MPW is disabled by default, while the Enhanced MPW is enabled
+	 * by default.
+	 */
+	if (config->mps == MLX5_ARG_UNSET)
+		config->mps = (mps == MLX5_MPW_ENHANCED) ? MLX5_MPW_ENHANCED :
+							  MLX5_MPW_DISABLED;
+	else
+		config->mps = config->mps ? mps : MLX5_MPW_DISABLED;
+	DRV_LOG(INFO, "%sMPS is %s",
+		config->mps == MLX5_MPW_ENHANCED ? "enhanced " :
+		config->mps == MLX5_MPW ? "legacy " : "",
+		config->mps != MLX5_MPW_DISABLED ? "enabled" : "disabled");
+	if (config->devx) {
+		err = mlx5_devx_cmd_query_hca_attr(sh->ctx, &config->hca_attr);
+		if (err) {
+			err = -err;
+			goto error;
+		}
+		/* Check relax ordering support. */
+		if (!haswell_broadwell_cpu) {
+			sh->cmng.relaxed_ordering_write =
+				config->hca_attr.relaxed_ordering_write;
+			sh->cmng.relaxed_ordering_read =
+				config->hca_attr.relaxed_ordering_read;
+		} else {
+			sh->cmng.relaxed_ordering_read = 0;
+			sh->cmng.relaxed_ordering_write = 0;
+		}
+		sh->rq_ts_format = config->hca_attr.rq_ts_format;
+		sh->sq_ts_format = config->hca_attr.sq_ts_format;
+		sh->steering_format_version =
+			config->hca_attr.steering_format_version;
+		sh->qp_ts_format = config->hca_attr.qp_ts_format;
+		/* Check for LRO support. */
+		if (config->dest_tir && config->hca_attr.lro_cap &&
+		    config->dv_flow_en) {
+			/* TBD check tunnel lro caps. */
+			config->lro.supported = config->hca_attr.lro_cap;
+			DRV_LOG(DEBUG, "Device supports LRO");
+			/*
+			 * If LRO timeout is not configured by application,
+			 * use the minimal supported value.
+			 */
+			if (!config->lro.timeout)
+				config->lro.timeout =
+				config->hca_attr.lro_timer_supported_periods[0];
+			DRV_LOG(DEBUG, "LRO session timeout set to %d usec",
+				config->lro.timeout);
+			DRV_LOG(DEBUG, "LRO minimal size of TCP segment "
+				"required for coalescing is %d bytes",
+				config->hca_attr.lro_min_mss_size);
+		}
+#if defined(HAVE_MLX5DV_DR) && \
+	(defined(HAVE_MLX5_DR_CREATE_ACTION_FLOW_METER) || \
+	 defined(HAVE_MLX5_DR_CREATE_ACTION_ASO))
+		if (config->hca_attr.qos.sup &&
+		    config->hca_attr.qos.flow_meter_old &&
+		    config->dv_flow_en) {
+			uint8_t reg_c_mask =
+				config->hca_attr.qos.flow_meter_reg_c_ids;
+			/*
+			 * Meter needs two REG_C's for color match and pre-sfx
+			 * flow match. Here get the REG_C for color match.
+			 * REG_C_0 and REG_C_1 is reserved for metadata feature.
+			 */
+			reg_c_mask &= 0xfc;
+			if (__builtin_popcount(reg_c_mask) < 1) {
+				priv->mtr_en = 0;
+				DRV_LOG(WARNING, "No available register for"
+					" meter.");
+			} else {
+				/*
+				 * The meter color register is used by the
+				 * flow-hit feature as well.
+				 * The flow-hit feature must use REG_C_3
+				 * Prefer REG_C_3 if it is available.
+				 */
+				if (reg_c_mask & (1 << (REG_C_3 - REG_C_0)))
+					priv->mtr_color_reg = REG_C_3;
+				else
+					priv->mtr_color_reg = ffs(reg_c_mask)
+							      - 1 + REG_C_0;
+				priv->mtr_en = 1;
+				priv->mtr_reg_share =
+				      config->hca_attr.qos.flow_meter;
+				DRV_LOG(DEBUG, "The REG_C meter uses is %d",
+					priv->mtr_color_reg);
+			}
+		}
+		if (config->hca_attr.qos.sup &&
+			config->hca_attr.qos.flow_meter_aso_sup) {
+			uint32_t log_obj_size =
+				rte_log2_u32(MLX5_ASO_MTRS_PER_POOL >> 1);
+			if (log_obj_size >=
+			config->hca_attr.qos.log_meter_aso_granularity &&
+			log_obj_size <=
+			config->hca_attr.qos.log_meter_aso_max_alloc)
+				sh->meter_aso_en = 1;
+		}
+		if (priv->mtr_en) {
+			err = mlx5_aso_flow_mtrs_mng_init(priv->sh);
+			if (err) {
+				err = -err;
+				goto error;
+			}
+		}
+		if (config->hca_attr.flow.tunnel_header_0_1)
+			sh->tunnel_header_0_1 = 1;
+#endif
+#ifdef HAVE_MLX5_DR_CREATE_ACTION_ASO
+		if (config->hca_attr.flow_hit_aso &&
+		    priv->mtr_color_reg == REG_C_3) {
+			sh->flow_hit_aso_en = 1;
+			err = mlx5_flow_aso_age_mng_init(sh);
+			if (err) {
+				err = -err;
+				goto error;
+			}
+			DRV_LOG(DEBUG, "Flow Hit ASO is supported.");
+		}
+#endif /* HAVE_MLX5_DR_CREATE_ACTION_ASO */
+#if defined(HAVE_MLX5_DR_CREATE_ACTION_ASO) && \
+	defined(HAVE_MLX5_DR_ACTION_ASO_CT)
+		if (config->hca_attr.ct_offload &&
+		    priv->mtr_color_reg == REG_C_3) {
+			err = mlx5_flow_aso_ct_mng_init(sh);
+			if (err) {
+				err = -err;
+				goto error;
+			}
+			DRV_LOG(DEBUG, "CT ASO is supported.");
+			sh->ct_aso_en = 1;
+		}
+#endif /* HAVE_MLX5_DR_CREATE_ACTION_ASO && HAVE_MLX5_DR_ACTION_ASO_CT */
+#if defined(HAVE_MLX5DV_DR) && defined(HAVE_MLX5_DR_CREATE_ACTION_FLOW_SAMPLE)
+		if (config->hca_attr.log_max_ft_sampler_num > 0  &&
+		    config->dv_flow_en) {
+			priv->sampler_en = 1;
+			DRV_LOG(DEBUG, "Sampler enabled!");
+		} else {
+			priv->sampler_en = 0;
+			if (!config->hca_attr.log_max_ft_sampler_num)
+				DRV_LOG(WARNING,
+					"No available register for sampler.");
+			else
+				DRV_LOG(DEBUG, "DV flow is not supported!");
+		}
+#endif
+	}
+	if (config->cqe_comp && RTE_CACHE_LINE_SIZE == 128 &&
+	    !(dv_attr.flags & MLX5DV_CONTEXT_FLAGS_CQE_128B_COMP)) {
+		DRV_LOG(WARNING, "Rx CQE 128B compression is not supported");
+		config->cqe_comp = 0;
+	}
+	if (config->cqe_comp_fmt == MLX5_CQE_RESP_FORMAT_FTAG_STRIDX &&
+	    (!config->devx || !config->hca_attr.mini_cqe_resp_flow_tag)) {
+		DRV_LOG(WARNING, "Flow Tag CQE compression"
+				 " format isn't supported.");
+		config->cqe_comp = 0;
+	}
+	if (config->cqe_comp_fmt == MLX5_CQE_RESP_FORMAT_L34H_STRIDX &&
+	    (!config->devx || !config->hca_attr.mini_cqe_resp_l3_l4_tag)) {
+		DRV_LOG(WARNING, "L3/L4 Header CQE compression"
+				 " format isn't supported.");
+		config->cqe_comp = 0;
+	}
+	DRV_LOG(DEBUG, "Rx CQE compression is %ssupported",
+			config->cqe_comp ? "" : "not ");
+	if (config->tx_pp) {
+		DRV_LOG(DEBUG, "Timestamp counter frequency %u kHz",
+			config->hca_attr.dev_freq_khz);
+		DRV_LOG(DEBUG, "Packet pacing is %ssupported",
+			config->hca_attr.qos.packet_pacing ? "" : "not ");
+		DRV_LOG(DEBUG, "Cross channel ops are %ssupported",
+			config->hca_attr.cross_channel ? "" : "not ");
+		DRV_LOG(DEBUG, "WQE index ignore is %ssupported",
+			config->hca_attr.wqe_index_ignore ? "" : "not ");
+		DRV_LOG(DEBUG, "Non-wire SQ feature is %ssupported",
+			config->hca_attr.non_wire_sq ? "" : "not ");
+		DRV_LOG(DEBUG, "Static WQE SQ feature is %ssupported (%d)",
+			config->hca_attr.log_max_static_sq_wq ? "" : "not ",
+			config->hca_attr.log_max_static_sq_wq);
+		DRV_LOG(DEBUG, "WQE rate PP mode is %ssupported",
+			config->hca_attr.qos.wqe_rate_pp ? "" : "not ");
+		if (!config->devx) {
+			DRV_LOG(ERR, "DevX is required for packet pacing");
+			err = ENODEV;
+			goto error;
+		}
+		if (!config->hca_attr.qos.packet_pacing) {
+			DRV_LOG(ERR, "Packet pacing is not supported");
+			err = ENODEV;
+			goto error;
+		}
+		if (!config->hca_attr.cross_channel) {
+			DRV_LOG(ERR, "Cross channel operations are"
+				     " required for packet pacing");
+			err = ENODEV;
+			goto error;
+		}
+		if (!config->hca_attr.wqe_index_ignore) {
+			DRV_LOG(ERR, "WQE index ignore feature is"
+				     " required for packet pacing");
+			err = ENODEV;
+			goto error;
+		}
+		if (!config->hca_attr.non_wire_sq) {
+			DRV_LOG(ERR, "Non-wire SQ feature is"
+				     " required for packet pacing");
+			err = ENODEV;
+			goto error;
+		}
+		if (!config->hca_attr.log_max_static_sq_wq) {
+			DRV_LOG(ERR, "Static WQE SQ feature is"
+				     " required for packet pacing");
+			err = ENODEV;
+			goto error;
+		}
+		if (!config->hca_attr.qos.wqe_rate_pp) {
+			DRV_LOG(ERR, "WQE rate mode is required"
+				     " for packet pacing");
+			err = ENODEV;
+			goto error;
+		}
+#ifndef HAVE_MLX5DV_DEVX_UAR_OFFSET
+		DRV_LOG(ERR, "DevX does not provide UAR offset,"
+			     " can't create queues for packet pacing");
+		err = ENODEV;
+		goto error;
+#endif
+	}
+	if (config->devx) {
+		uint32_t reg[MLX5_ST_SZ_DW(register_mtutc)];
+
+		err = config->hca_attr.access_register_user ?
+			mlx5_devx_cmd_register_read
+				(sh->ctx, MLX5_REGISTER_ID_MTUTC, 0,
+				reg, MLX5_ST_SZ_DW(register_mtutc)) : ENOTSUP;
+		if (!err) {
+			uint32_t ts_mode;
+
+			/* MTUTC register is read successfully. */
+			ts_mode = MLX5_GET(register_mtutc, reg,
+					   time_stamp_mode);
+			if (ts_mode == MLX5_MTUTC_TIMESTAMP_MODE_REAL_TIME)
+				config->rt_timestamp = 1;
+		} else {
+			/* Kernel does not support register reading. */
+			if (config->hca_attr.dev_freq_khz ==
+						 (NS_PER_S / MS_PER_S))
+				config->rt_timestamp = 1;
+		}
+	}
+	/*
+	 * If HW has bug working with tunnel packet decapsulation and
+	 * scatter FCS, and decapsulation is needed, clear the hw_fcs_strip
+	 * bit. Then DEV_RX_OFFLOAD_KEEP_CRC bit will not be set anymore.
+	 */
+	if (config->hca_attr.scatter_fcs_w_decap_disable && config->decap_en)
+		config->hw_fcs_strip = 0;
+	DRV_LOG(DEBUG, "FCS stripping configuration is %ssupported",
+		(config->hw_fcs_strip ? "" : "not "));
+	if (config->mprq.enabled && mprq) {
+		if (config->mprq.stride_num_n &&
+		    (config->mprq.stride_num_n > mprq_max_stride_num_n ||
+		     config->mprq.stride_num_n < mprq_min_stride_num_n)) {
+			config->mprq.stride_num_n =
+				RTE_MIN(RTE_MAX(MLX5_MPRQ_STRIDE_NUM_N,
+						mprq_min_stride_num_n),
+					mprq_max_stride_num_n);
+			DRV_LOG(WARNING,
+				"the number of strides"
+				" for Multi-Packet RQ is out of range,"
+				" setting default value (%u)",
+				1 << config->mprq.stride_num_n);
+		}
+		if (config->mprq.stride_size_n &&
+		    (config->mprq.stride_size_n > mprq_max_stride_size_n ||
+		     config->mprq.stride_size_n < mprq_min_stride_size_n)) {
+			config->mprq.stride_size_n =
+				RTE_MIN(RTE_MAX(MLX5_MPRQ_STRIDE_SIZE_N,
+						mprq_min_stride_size_n),
+					mprq_max_stride_size_n);
+			DRV_LOG(WARNING,
+				"the size of a stride"
+				" for Multi-Packet RQ is out of range,"
+				" setting default value (%u)",
+				1 << config->mprq.stride_size_n);
+		}
+		config->mprq.min_stride_size_n = mprq_min_stride_size_n;
+		config->mprq.max_stride_size_n = mprq_max_stride_size_n;
+	} else if (config->mprq.enabled && !mprq) {
+		DRV_LOG(WARNING, "Multi-Packet RQ isn't supported");
+		config->mprq.enabled = 0;
+	}
+	if (config->max_dump_files_num == 0)
+		config->max_dump_files_num = 128;
+	eth_dev = rte_eth_dev_allocate(name);
+	if (eth_dev == NULL) {
+		DRV_LOG(ERR, "can not allocate rte ethdev");
+		err = ENOMEM;
+		goto error;
+	}
+	if (priv->representor) {
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_REPRESENTOR;
+		eth_dev->data->representor_id = priv->representor_id;
+	}
+	priv->mp_id.port_id = eth_dev->data->port_id;
+	strlcpy(priv->mp_id.name, MLX5_MP_NAME, RTE_MP_MAX_NAME_LEN);
+	/*
+	 * Store associated network device interface index. This index
+	 * is permanent throughout the lifetime of device. So, we may store
+	 * the ifindex here and use the cached value further.
+	 */
+	MLX5_ASSERT(spawn->ifindex);
+	priv->if_index = spawn->ifindex;
+	eth_dev->data->dev_private = priv;
+	priv->dev_data = eth_dev->data;
+	eth_dev->data->mac_addrs = priv->mac;
+	eth_dev->device = dpdk_dev;
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+	/* Configure the first MAC address by default. */
+	if (mlx5_get_mac(eth_dev, &mac.addr_bytes)) {
+		DRV_LOG(ERR,
+			"port %u cannot get MAC address, is mlx5_en"
+			" loaded? (errno: %s)",
+			eth_dev->data->port_id, strerror(rte_errno));
+		err = ENODEV;
+		goto error;
+	}
+	DRV_LOG(INFO,
+		"port %u MAC address is " RTE_ETHER_ADDR_PRT_FMT,
+		eth_dev->data->port_id, RTE_ETHER_ADDR_BYTES(&mac));
+#ifdef RTE_LIBRTE_MLX5_DEBUG
+	{
+		char ifname[MLX5_NAMESIZE];
+
+		if (mlx5_get_ifname(eth_dev, &ifname) == 0)
+			DRV_LOG(DEBUG, "port %u ifname is \"%s\"",
+				eth_dev->data->port_id, ifname);
+		else
+			DRV_LOG(DEBUG, "port %u ifname is unknown",
+				eth_dev->data->port_id);
+	}
+#endif
+	/* Get actual MTU if possible. */
+	err = mlx5_get_mtu(eth_dev, &priv->mtu);
+	if (err) {
+		err = rte_errno;
+		goto error;
+	}
+	DRV_LOG(DEBUG, "port %u MTU is %u", eth_dev->data->port_id,
+		priv->mtu);
+	/* Initialize burst functions to prevent crashes before link-up. */
+	eth_dev->rx_pkt_burst = removed_rx_burst;
+	eth_dev->tx_pkt_burst = removed_tx_burst;
+	eth_dev->dev_ops = &mlx5_dev_ops;
+	eth_dev->rx_descriptor_status = mlx5_rx_descriptor_status;
+	eth_dev->tx_descriptor_status = mlx5_tx_descriptor_status;
+	eth_dev->rx_queue_count = mlx5_rx_queue_count;
+	/* Register MAC address. */
+	claim_zero(mlx5_mac_addr_add(eth_dev, &mac, 0, 0));
+	priv->ctrl_flows = 0;
+	rte_spinlock_init(&priv->flow_list_lock);
+	TAILQ_INIT(&priv->flow_meters);
+	priv->mtr_profile_tbl = mlx5_l3t_create(MLX5_L3T_TYPE_PTR);
+	if (!priv->mtr_profile_tbl)
+		goto error;
+	/* Hint libmlx5 to use PMD allocator for data plane resources */
+	mlx5_glue->dv_set_context_attr(sh->ctx,
+			MLX5DV_CTX_ATTR_BUF_ALLOCATORS,
+			(void *)((uintptr_t)&(struct mlx5dv_ctx_allocators){
+				.alloc = &mlx5_alloc_verbs_buf,
+				.free = &mlx5_free_verbs_buf,
+				.data = sh,
+			}));
+	/* Bring Ethernet device up. */
+	DRV_LOG(DEBUG, "port %u forcing Ethernet interface up",
+		eth_dev->data->port_id);
+	mlx5_set_link_up(eth_dev);
+	/*
+	 * Even though the interrupt handler is not installed yet,
+	 * interrupts will still trigger on the async_fd from
+	 * Verbs context returned by ibv_open_device().
+	 */
+	mlx5_link_update(eth_dev, 0);
+#ifdef HAVE_MLX5DV_DR_ESWITCH
+	if (!(config->hca_attr.eswitch_manager && config->dv_flow_en &&
+	      (switch_info->representor || switch_info->master)))
+		config->dv_esw_en = 0;
+#else
+	config->dv_esw_en = 0;
+#endif
+	/* Detect minimal data bytes to inline. */
+	mlx5_set_min_inline(spawn, config);
+	/* Store device configuration on private structure. */
+	priv->config = *config;
+	for (i = 0; i < MLX5_FLOW_TYPE_MAXI; i++) {
+		icfg[i].release_mem_en = !!config->reclaim_mode;
+		if (config->reclaim_mode)
+			icfg[i].per_core_cache = 0;
+		priv->flows[i] = mlx5_ipool_create(&icfg[i]);
+		if (!priv->flows[i])
+			goto error;
+	}
+	/* Create context for virtual machine VLAN workaround. */
+	priv->vmwa_context = mlx5_vlan_vmwa_init(eth_dev, spawn->ifindex);
+	if (config->dv_flow_en) {
+		err = mlx5_alloc_shared_dr(priv);
+		if (err)
+			goto error;
+	}
+	if (config->devx && config->dv_flow_en && config->dest_tir) {
+		priv->obj_ops = devx_obj_ops;
+		priv->obj_ops.drop_action_create =
+						ibv_obj_ops.drop_action_create;
+		priv->obj_ops.drop_action_destroy =
+						ibv_obj_ops.drop_action_destroy;
+#ifndef HAVE_MLX5DV_DEVX_UAR_OFFSET
+		priv->obj_ops.txq_obj_modify = ibv_obj_ops.txq_obj_modify;
+#else
+		if (config->dv_esw_en)
+			priv->obj_ops.txq_obj_modify =
+						ibv_obj_ops.txq_obj_modify;
+#endif
+		/* Use specific wrappers for Tx object. */
+		priv->obj_ops.txq_obj_new = mlx5_os_txq_obj_new;
+		priv->obj_ops.txq_obj_release = mlx5_os_txq_obj_release;
+		mlx5_queue_counter_id_prepare(eth_dev);
+		priv->obj_ops.lb_dummy_queue_create =
+					mlx5_rxq_ibv_obj_dummy_lb_create;
+		priv->obj_ops.lb_dummy_queue_release =
+					mlx5_rxq_ibv_obj_dummy_lb_release;
+	} else {
+		priv->obj_ops = ibv_obj_ops;
+	}
+	if (config->tx_pp &&
+	    (priv->config.dv_esw_en ||
+	     priv->obj_ops.txq_obj_new != mlx5_os_txq_obj_new)) {
+		/*
+		 * HAVE_MLX5DV_DEVX_UAR_OFFSET is required to support
+		 * packet pacing and already checked above.
+		 * Hence, we should only make sure the SQs will be created
+		 * with DevX, not with Verbs.
+		 * Verbs allocates the SQ UAR on its own and it can't be shared
+		 * with Clock Queue UAR as required for Tx scheduling.
+		 */
+		DRV_LOG(ERR, "Verbs SQs, UAR can't be shared as required for packet pacing");
+		err = ENODEV;
+		goto error;
+	}
+	priv->drop_queue.hrxq = mlx5_drop_action_create(eth_dev);
+	if (!priv->drop_queue.hrxq)
+		goto error;
+	/* Supported Verbs flow priority number detection. */
+	err = mlx5_flow_discover_priorities(eth_dev);
+	if (err < 0) {
+		err = -err;
+		goto error;
+	}
+	priv->config.flow_prio = err;
+	if (!priv->config.dv_esw_en &&
+	    priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY) {
+		DRV_LOG(WARNING, "metadata mode %u is not supported "
+				 "(no E-Switch)", priv->config.dv_xmeta_en);
+		priv->config.dv_xmeta_en = MLX5_XMETA_MODE_LEGACY;
+	}
+	mlx5_set_metadata_mask(eth_dev);
+	if (priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY &&
+	    !priv->sh->dv_regc0_mask) {
+		DRV_LOG(ERR, "metadata mode %u is not supported "
+			     "(no metadata reg_c[0] is available)",
+			     priv->config.dv_xmeta_en);
+			err = ENOTSUP;
+			goto error;
+	}
+	priv->hrxqs = mlx5_list_create("hrxq", eth_dev, true,
+				       mlx5_hrxq_create_cb,
+				       mlx5_hrxq_match_cb,
+				       mlx5_hrxq_remove_cb,
+				       mlx5_hrxq_clone_cb,
+				       mlx5_hrxq_clone_free_cb);
+	if (!priv->hrxqs)
+		goto error;
+	rte_rwlock_init(&priv->ind_tbls_lock);
+	/* Query availability of metadata reg_c's. */
+	err = mlx5_flow_discover_mreg_c(eth_dev);
+	if (err < 0) {
+		err = -err;
+		goto error;
+	}
+	if (!mlx5_flow_ext_mreg_supported(eth_dev)) {
+		DRV_LOG(DEBUG,
+			"port %u extensive metadata register is not supported",
+			eth_dev->data->port_id);
+		if (priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY) {
+			DRV_LOG(ERR, "metadata mode %u is not supported "
+				     "(no metadata registers available)",
+				     priv->config.dv_xmeta_en);
+			err = ENOTSUP;
+			goto error;
+		}
+	}
+	if (priv->config.dv_flow_en &&
+	    priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY &&
+	    mlx5_flow_ext_mreg_supported(eth_dev) &&
+	    priv->sh->dv_regc0_mask) {
+		priv->mreg_cp_tbl = mlx5_hlist_create(MLX5_FLOW_MREG_HNAME,
+						      MLX5_FLOW_MREG_HTABLE_SZ,
+						      false, true, eth_dev,
+						      flow_dv_mreg_create_cb,
+						      flow_dv_mreg_match_cb,
+						      flow_dv_mreg_remove_cb,
+						      flow_dv_mreg_clone_cb,
+						    flow_dv_mreg_clone_free_cb);
+		if (!priv->mreg_cp_tbl) {
+			err = ENOMEM;
+			goto error;
+		}
+	}
+	rte_spinlock_init(&priv->shared_act_sl);
+	mlx5_flow_counter_mode_config(eth_dev);
+	mlx5_flow_drop_action_config(eth_dev);
+	if (priv->config.dv_flow_en)
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
+	return eth_dev;
+error:
+	if (priv) {
+		if (priv->mreg_cp_tbl)
+			mlx5_hlist_destroy(priv->mreg_cp_tbl);
+		if (priv->sh)
+			mlx5_os_free_shared_dr(priv);
+		if (priv->nl_socket_route >= 0)
+			close(priv->nl_socket_route);
+		if (priv->nl_socket_rdma >= 0)
+			close(priv->nl_socket_rdma);
+		if (priv->vmwa_context)
+			mlx5_vlan_vmwa_exit(priv->vmwa_context);
+		if (eth_dev && priv->drop_queue.hrxq)
+			mlx5_drop_action_destroy(eth_dev);
+		if (priv->mtr_profile_tbl)
+			mlx5_l3t_destroy(priv->mtr_profile_tbl);
+		if (own_domain_id)
+			claim_zero(rte_eth_switch_domain_free(priv->domain_id));
+		if (priv->hrxqs)
+			mlx5_list_destroy(priv->hrxqs);
+		mlx5_free(priv);
+		if (eth_dev != NULL)
+			eth_dev->data->dev_private = NULL;
+	}
+	if (eth_dev != NULL) {
+		/* mac_addrs must not be freed alone because part of
+		 * dev_private
+		 **/
+		eth_dev->data->mac_addrs = NULL;
+		rte_eth_dev_release_port(eth_dev);
+	}
+	if (sh)
+		mlx5_free_shared_dev_ctx(sh);
+	MLX5_ASSERT(err > 0);
+	rte_errno = err;
+	return NULL;
+}
+
+/**
+ * Comparison callback to sort device data.
+ *
+ * This is meant to be used with qsort().
+ *
+ * @param a[in]
+ *   Pointer to pointer to first data object.
+ * @param b[in]
+ *   Pointer to pointer to second data object.
+ *
+ * @return
+ *   0 if both objects are equal, less than 0 if the first argument is less
+ *   than the second, greater than 0 otherwise.
+ */
+static int
+mlx5_dev_spawn_data_cmp(const void *a, const void *b)
+{
+	const struct mlx5_switch_info *si_a =
+		&((const struct mlx5_dev_spawn_data *)a)->info;
+	const struct mlx5_switch_info *si_b =
+		&((const struct mlx5_dev_spawn_data *)b)->info;
+	int ret;
+
+	/* Master device first. */
+	ret = si_b->master - si_a->master;
+	if (ret)
+		return ret;
+	/* Then representor devices. */
+	ret = si_b->representor - si_a->representor;
+	if (ret)
+		return ret;
+	/* Unidentified devices come last in no specific order. */
+	if (!si_a->representor)
+		return 0;
+	/* Order representors by name. */
+	return si_a->port_name - si_b->port_name;
+}
+
+static void
+mlx5_os_config_default(struct mlx5_dev_config *config)
+{
+	memset(config, 0, sizeof(*config));
+	config->mps = MLX5_ARG_UNSET;
+	config->dbnc = MLX5_ARG_UNSET;
+	config->rx_vec_en = 1;
+	config->txq_inline_max = MLX5_ARG_UNSET;
+	config->txq_inline_min = MLX5_ARG_UNSET;
+	config->txq_inline_mpw = MLX5_ARG_UNSET;
+	config->txqs_inline = MLX5_ARG_UNSET;
+	config->vf_nl_en = 0;
+	config->mr_ext_memseg_en = 1;
+	config->mprq.max_memcpy_len = MLX5_MPRQ_MEMCPY_DEFAULT_LEN;
+	config->mprq.min_rxqs_num = MLX5_MPRQ_MIN_RXQS;
+	config->dv_esw_en = 1;
+	config->dv_flow_en = 1;
+	config->decap_en = 1;
+	config->log_hp_size = MLX5_ARG_UNSET;
+	config->allow_duplicate_pattern = 1;
+}
+
+/**
+ * Callback to register a PCI device.
+ *
+ * This function spawns Ethernet devices out of a given PCI device.
+ *
+ * @param[in] pci_dev
+ *   PCI device information.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_os_pci_probe(struct rte_pci_device *pci_dev)
+{
+	struct ibv_device **ibv_list;
+	/*
+	 * Number of found IB Devices matching with requested PCI BDF.
+	 * nd != 1 means there are multiple IB devices over the same
+	 * PCI device and we have representors and master.
+	 */
+	unsigned int nd = 0;
+	/*
+	 * Number of DPDK ethernet devices to Spawn - either over
+	 * multiple IB devices or multiple ports of single IB device.
+	 * Actually this is the number of iterations to spawn.
+	 */
+	unsigned int ns = 0;
+	struct mlx5_dev_spawn_data *list = NULL;
+	struct mlx5_dev_config dev_config;
+	unsigned int dev_config_vf;
+	struct rte_pci_addr pci_addr;
+	char ifname[IF_NAMESIZE];
+	uint32_t restore;
+	unsigned int i;
+	int ret;
+
+	errno = 0;
+	ibv_list = mlx5_glue->get_device_list(&ret);
+	if (!ibv_list) {
+		rte_errno = errno ? errno : ENOSYS;
+		DRV_LOG(ERR, "cannot list devices, is ib_uverbs loaded?");
+		return -rte_errno;
+	}
+	/*
+	 * First scan the list of all Infiniband devices to find
+	 * matching ones, gathering into the list.
+	 */
+	struct ibv_device *ibv_match[ret + 1];
+
+	while (ret-- > 0) {
+		DRV_LOG(DEBUG, "checking device \"%s\"", ibv_list[ret]->name);
+		if (mlx5_get_pci_addr
+			(ibv_list[ret]->ibdev_path, &pci_addr))
+			continue;
+
+		if (pci_dev->addr.domain != pci_addr.domain ||
+		    pci_dev->addr.bus != pci_addr.bus ||
+		    pci_dev->addr.devid != pci_addr.devid ||
+		    pci_dev->addr.function != pci_addr.function)
+			continue;
+
+		DRV_LOG(INFO, "PCI information matches for device \"%s\"",
+			ibv_list[ret]->name);
+		ibv_match[nd++] = ibv_list[ret];
+	}
+	ibv_match[nd] = NULL;
+	if (!nd) {
+		/* No device matches, just complain and bail out. */
+		DRV_LOG(WARNING,
+			"no Verbs device matches PCI device " PCI_PRI_FMT ","
+			" are kernel drivers loaded?",
+			pci_dev->addr.domain, pci_dev->addr.bus,
+			pci_dev->addr.devid, pci_dev->addr.function);
+		rte_errno = ENOENT;
+		ret = -rte_errno;
+		goto exit;
+	}
+	/*
+	 * Now we can determine the maximal
+	 * amount of devices to be spawned.
+	 */
+	list = mlx5_malloc(MLX5_MEM_ZERO,
+			   sizeof(struct mlx5_dev_spawn_data) * nd,
+			   RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+	if (!list) {
+		DRV_LOG(ERR, "spawn data array allocation failure");
+		rte_errno = ENOMEM;
+		ret = -rte_errno;
+		goto exit;
+	}
+	/*
+	 * 1. A single IB device matches (nd == 1) and
+	 *    is not a representor, assume no switch
+	 *    support.
+	 *
+	 * 2. Otherwise no safe assumptions can be made;
+	 *    complain louder and bail out.
+	 */
+	MLX5_ASSERT(nd);
+	for (i = 0; i != nd; ++i) {
+		memset(&list[ns].info, 0, sizeof(list[ns].info));
+		list[ns].max_port = 1;
+		list[ns].phys_port = 1;
+		list[ns].phys_dev = ibv_match[i];
+		list[ns].eth_dev = NULL;
+		list[ns].pci_dev = pci_dev;
+		list[ns].pf_bond = -1;
+		list[ns].ifindex = 0;
+		if (!list[ns].ifindex) {
+			ret = mlx5_get_ifname_sysfs(ibv_match[i]->ibdev_path,
+					      ifname);
+			if (!ret)
+				list[ns].ifindex =
+					if_nametoindex(ifname);
+			if (!list[ns].ifindex) {
+				/*
+				 * No network interface index found
+				 * for the specified device, it means
+				 * there it is neither representor
+				 * nor master.
+				 */
+				continue;
+			}
+		}
+		ret = -1;
+		if (nd == 1 &&
+		    !list[ns].info.representor && !list[ns].info.master) {
+			/*
+			 * Single IB device with
+			 * one physical port and
+			 * attached network device.
+			 * May be SRIOV is not enabled
+			 * or there is no representors.
+			 */
+			DRV_LOG(INFO, "no E-Switch support detected");
+			ns++;
+			break;
+		}
+	}
+	if (!ns) {
+		DRV_LOG(ERR,
+			"unable to recognize master/representors"
+			" on the multiple IB devices");
+		rte_errno = ENOENT;
+		ret = -rte_errno;
+		goto exit;
+	}
+
+	MLX5_ASSERT(ns);
+	/*
+	 * Sort list to probe devices in natural order for users convenience
+	 * (i.e. master first, then representors from lowest to highest ID).
+	 */
+	qsort(list, ns, sizeof(*list), mlx5_dev_spawn_data_cmp);
+	/* Device specific configuration. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_MELLANOX_CONNECTX4VF:
+	case PCI_DEVICE_ID_MELLANOX_CONNECTX4LXVF:
+	case PCI_DEVICE_ID_MELLANOX_CONNECTX5VF:
+	case PCI_DEVICE_ID_MELLANOX_CONNECTX5EXVF:
+	case PCI_DEVICE_ID_MELLANOX_CONNECTX5BFVF:
+	case PCI_DEVICE_ID_MELLANOX_CONNECTX6VF:
+	case PCI_DEVICE_ID_MELLANOX_CONNECTXVF:
+		dev_config_vf = 1;
+		break;
+	default:
+		dev_config_vf = 0;
+		break;
+	}
+	for (i = 0; i != ns; ++i) {
+		/* Default configuration. */
+		mlx5_os_config_default(&dev_config);
+		dev_config.vf = dev_config_vf;
+		dev_config.allow_duplicate_pattern = 1;
+		list[i].numa_node = pci_dev->device.numa_node;
+		list[i].eth_dev = mlx5_dev_spawn(&pci_dev->device,
+						 &list[i],
+						 &dev_config);
+		if (!list[i].eth_dev) {
+			if (rte_errno != EBUSY && rte_errno != EEXIST)
+				break;
+			/* Device is disabled or already spawned. Ignore it. */
+			continue;
+		}
+		restore = list[i].eth_dev->data->dev_flags;
+		rte_eth_copy_pci_info(list[i].eth_dev, pci_dev);
+		/* Restore non-PCI flags cleared by the above call. */
+		list[i].eth_dev->data->dev_flags |= restore;
+		rte_eth_dev_probing_finish(list[i].eth_dev);
+	}
+	if (i != ns) {
+		DRV_LOG(ERR,
+			"probe of PCI device " PCI_PRI_FMT " aborted after"
+			" encountering an error: %s",
+			pci_dev->addr.domain, pci_dev->addr.bus,
+			pci_dev->addr.devid, pci_dev->addr.function,
+			strerror(rte_errno));
+		ret = -rte_errno;
+		/* Roll back. */
+		while (i--) {
+			if (!list[i].eth_dev)
+				continue;
+			mlx5_dev_close(list[i].eth_dev);
+			/* mac_addrs must not be freed because in dev_private */
+			list[i].eth_dev->data->mac_addrs = NULL;
+			claim_zero(rte_eth_dev_release_port(list[i].eth_dev));
+		}
+		/* Restore original error. */
+		rte_errno = -ret;
+	} else {
+		ret = 0;
+	}
+exit:
+	/*
+	 * Do the routine cleanup:
+	 * - free allocated spawn data array
+	 * - free the Infiniband device list
+	 */
+	if (list)
+		mlx5_free(list);
+	MLX5_ASSERT(ibv_list);
+	mlx5_glue->free_device_list(ibv_list);
+	return ret;
+}
+
 /**
  * Net class driver callback to probe a device.
  *
@@ -287,6 +2117,10 @@  mlx5_os_net_probe(struct rte_device *dev)
 			strerror(rte_errno));
 		return -rte_errno;
 	}
+	if (mlx5_dev_is_pci(dev))
+		return mlx5_os_pci_probe(RTE_DEV_TO_PCI(dev));
+	else
+		return -ENOTSUP;
 }
 
 static int