New upstream version 18.08upstream/18.08

Change-Id: I32fdf5e5016556d9c0a6d88ddaf1fc468961790a
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>

1 files changed, 1134 insertions, 424 deletions

diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
index ff58c492..1f7bfd44 100644
--- a/drivers/net/mlx5/mlx5_rxq.c
+++ b/drivers/net/mlx5/mlx5_rxq.c
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: BSD-3-Clause
  * Copyright 2015 6WIND S.A.
- * Copyright 2015 Mellanox.
+ * Copyright 2015 Mellanox Technologies, Ltd
  */
 
 #include <stddef.h>
@@ -52,10 +52,129 @@ uint8_t rss_hash_default_key[] = {
 };
 
 /* Length of the default RSS hash key. */
-const size_t rss_hash_default_key_len = sizeof(rss_hash_default_key);
+static_assert(MLX5_RSS_HASH_KEY_LEN ==
+	      (unsigned int)sizeof(rss_hash_default_key),
+	      "wrong RSS default key size.");
 
 /**
- * Allocate RX queue elements.
+ * Check whether Multi-Packet RQ can be enabled for the device.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   1 if supported, negative errno value if not.
+ */
+inline int
+mlx5_check_mprq_support(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+
+	if (priv->config.mprq.enabled &&
+	    priv->rxqs_n >= priv->config.mprq.min_rxqs_num)
+		return 1;
+	return -ENOTSUP;
+}
+
+/**
+ * Check whether Multi-Packet RQ is enabled for the Rx queue.
+ *
+ *  @param rxq
+ *     Pointer to receive queue structure.
+ *
+ * @return
+ *   0 if disabled, otherwise enabled.
+ */
+inline int
+mlx5_rxq_mprq_enabled(struct mlx5_rxq_data *rxq)
+{
+	return rxq->strd_num_n > 0;
+}
+
+/**
+ * Check whether Multi-Packet RQ is enabled for the device.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   0 if disabled, otherwise enabled.
+ */
+inline int
+mlx5_mprq_enabled(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	uint16_t i;
+	uint16_t n = 0;
+
+	if (mlx5_check_mprq_support(dev) < 0)
+		return 0;
+	/* All the configured queues should be enabled. */
+	for (i = 0; i < priv->rxqs_n; ++i) {
+		struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+
+		if (!rxq)
+			continue;
+		if (mlx5_rxq_mprq_enabled(rxq))
+			++n;
+	}
+	/* Multi-Packet RQ can't be partially configured. */
+	assert(n == 0 || n == priv->rxqs_n);
+	return n == priv->rxqs_n;
+}
+
+/**
+ * Allocate RX queue elements for Multi-Packet RQ.
+ *
+ * @param rxq_ctrl
+ *   Pointer to RX queue structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+rxq_alloc_elts_mprq(struct mlx5_rxq_ctrl *rxq_ctrl)
+{
+	struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
+	unsigned int wqe_n = 1 << rxq->elts_n;
+	unsigned int i;
+	int err;
+
+	/* Iterate on segments. */
+	for (i = 0; i <= wqe_n; ++i) {
+		struct mlx5_mprq_buf *buf;
+
+		if (rte_mempool_get(rxq->mprq_mp, (void **)&buf) < 0) {
+			DRV_LOG(ERR, "port %u empty mbuf pool", rxq->port_id);
+			rte_errno = ENOMEM;
+			goto error;
+		}
+		if (i < wqe_n)
+			(*rxq->mprq_bufs)[i] = buf;
+		else
+			rxq->mprq_repl = buf;
+	}
+	DRV_LOG(DEBUG,
+		"port %u Rx queue %u allocated and configured %u segments",
+		rxq->port_id, rxq_ctrl->idx, wqe_n);
+	return 0;
+error:
+	err = rte_errno; /* Save rte_errno before cleanup. */
+	wqe_n = i;
+	for (i = 0; (i != wqe_n); ++i) {
+		if ((*rxq->mprq_bufs)[i] != NULL)
+			rte_mempool_put(rxq->mprq_mp,
+					(*rxq->mprq_bufs)[i]);
+		(*rxq->mprq_bufs)[i] = NULL;
+	}
+	DRV_LOG(DEBUG, "port %u Rx queue %u failed, freed everything",
+		rxq->port_id, rxq_ctrl->idx);
+	rte_errno = err; /* Restore rte_errno. */
+	return -rte_errno;
+}
+
+/**
+ * Allocate RX queue elements for Single-Packet RQ.
  *
  * @param rxq_ctrl
  *   Pointer to RX queue structure.
@@ -63,13 +182,13 @@ const size_t rss_hash_default_key_len = sizeof(rss_hash_default_key);
  * @return
  *   0 on success, errno value on failure.
  */
-int
-rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+static int
+rxq_alloc_elts_sprq(struct mlx5_rxq_ctrl *rxq_ctrl)
 {
 	const unsigned int sges_n = 1 << rxq_ctrl->rxq.sges_n;
 	unsigned int elts_n = 1 << rxq_ctrl->rxq.elts_n;
 	unsigned int i;
-	int ret = 0;
+	int err;
 
 	/* Iterate on segments. */
 	for (i = 0; (i != elts_n); ++i) {
@@ -77,8 +196,9 @@ rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
 
 		buf = rte_pktmbuf_alloc(rxq_ctrl->rxq.mp);
 		if (buf == NULL) {
-			ERROR("%p: empty mbuf pool", (void *)rxq_ctrl);
-			ret = ENOMEM;
+			DRV_LOG(ERR, "port %u empty mbuf pool",
+				PORT_ID(rxq_ctrl->priv));
+			rte_errno = ENOMEM;
 			goto error;
 		}
 		/* Headroom is reserved by rte_pktmbuf_alloc(). */
@@ -97,7 +217,7 @@ rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
 		(*rxq_ctrl->rxq.elts)[i] = buf;
 	}
 	/* If Rx vector is activated. */
-	if (rxq_check_vec_support(&rxq_ctrl->rxq) > 0) {
+	if (mlx5_rxq_check_vec_support(&rxq_ctrl->rxq) > 0) {
 		struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
 		struct rte_mbuf *mbuf_init = &rxq->fake_mbuf;
 		int j;
@@ -118,30 +238,78 @@ rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
 		for (j = 0; j < MLX5_VPMD_DESCS_PER_LOOP; ++j)
 			(*rxq->elts)[elts_n + j] = &rxq->fake_mbuf;
 	}
-	DEBUG("%p: allocated and configured %u segments (max %u packets)",
-	      (void *)rxq_ctrl, elts_n, elts_n / (1 << rxq_ctrl->rxq.sges_n));
-	assert(ret == 0);
+	DRV_LOG(DEBUG,
+		"port %u Rx queue %u allocated and configured %u segments"
+		" (max %u packets)",
+		PORT_ID(rxq_ctrl->priv), rxq_ctrl->idx, elts_n,
+		elts_n / (1 << rxq_ctrl->rxq.sges_n));
 	return 0;
 error:
+	err = rte_errno; /* Save rte_errno before cleanup. */
 	elts_n = i;
 	for (i = 0; (i != elts_n); ++i) {
 		if ((*rxq_ctrl->rxq.elts)[i] != NULL)
 			rte_pktmbuf_free_seg((*rxq_ctrl->rxq.elts)[i]);
 		(*rxq_ctrl->rxq.elts)[i] = NULL;
 	}
-	DEBUG("%p: failed, freed everything", (void *)rxq_ctrl);
-	assert(ret > 0);
-	return ret;
+	DRV_LOG(DEBUG, "port %u Rx queue %u failed, freed everything",
+		PORT_ID(rxq_ctrl->priv), rxq_ctrl->idx);
+	rte_errno = err; /* Restore rte_errno. */
+	return -rte_errno;
 }
 
 /**
- * Free RX queue elements.
+ * Allocate RX queue elements.
+ *
+ * @param rxq_ctrl
+ *   Pointer to RX queue structure.
+ *
+ * @return
+ *   0 on success, errno value on failure.
+ */
+int
+rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+{
+	return mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq) ?
+	       rxq_alloc_elts_mprq(rxq_ctrl) : rxq_alloc_elts_sprq(rxq_ctrl);
+}
+
+/**
+ * Free RX queue elements for Multi-Packet RQ.
  *
  * @param rxq_ctrl
  *   Pointer to RX queue structure.
  */
 static void
-rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+rxq_free_elts_mprq(struct mlx5_rxq_ctrl *rxq_ctrl)
+{
+	struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
+	uint16_t i;
+
+	DRV_LOG(DEBUG, "port %u Multi-Packet Rx queue %u freeing WRs",
+		rxq->port_id, rxq_ctrl->idx);
+	if (rxq->mprq_bufs == NULL)
+		return;
+	assert(mlx5_rxq_check_vec_support(rxq) < 0);
+	for (i = 0; (i != (1u << rxq->elts_n)); ++i) {
+		if ((*rxq->mprq_bufs)[i] != NULL)
+			mlx5_mprq_buf_free((*rxq->mprq_bufs)[i]);
+		(*rxq->mprq_bufs)[i] = NULL;
+	}
+	if (rxq->mprq_repl != NULL) {
+		mlx5_mprq_buf_free(rxq->mprq_repl);
+		rxq->mprq_repl = NULL;
+	}
+}
+
+/**
+ * Free RX queue elements for Single-Packet RQ.
+ *
+ * @param rxq_ctrl
+ *   Pointer to RX queue structure.
+ */
+static void
+rxq_free_elts_sprq(struct mlx5_rxq_ctrl *rxq_ctrl)
 {
 	struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
 	const uint16_t q_n = (1 << rxq->elts_n);
@@ -149,14 +317,15 @@ rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
 	uint16_t used = q_n - (rxq->rq_ci - rxq->rq_pi);
 	uint16_t i;
 
-	DEBUG("%p: freeing WRs", (void *)rxq_ctrl);
+	DRV_LOG(DEBUG, "port %u Rx queue %u freeing WRs",
+		PORT_ID(rxq_ctrl->priv), rxq_ctrl->idx);
 	if (rxq->elts == NULL)
 		return;
 	/**
 	 * Some mbuf in the Ring belongs to the application.  They cannot be
 	 * freed.
 	 */
-	if (rxq_check_vec_support(rxq) > 0) {
+	if (mlx5_rxq_check_vec_support(rxq) > 0) {
 		for (i = 0; i < used; ++i)
 			(*rxq->elts)[(rxq->rq_ci + i) & q_mask] = NULL;
 		rxq->rq_pi = rxq->rq_ci;
@@ -169,6 +338,21 @@ rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
 }
 
 /**
+ * Free RX queue elements.
+ *
+ * @param rxq_ctrl
+ *   Pointer to RX queue structure.
+ */
+static void
+rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+{
+	if (mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq))
+		rxq_free_elts_mprq(rxq_ctrl);
+	else
+		rxq_free_elts_sprq(rxq_ctrl);
+}
+
+/**
  * Clean up a RX queue.
  *
  * Destroy objects, free allocated memory and reset the structure for reuse.
@@ -179,31 +363,35 @@ rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
 void
 mlx5_rxq_cleanup(struct mlx5_rxq_ctrl *rxq_ctrl)
 {
-	DEBUG("cleaning up %p", (void *)rxq_ctrl);
+	DRV_LOG(DEBUG, "port %u cleaning up Rx queue %u",
+		PORT_ID(rxq_ctrl->priv), rxq_ctrl->idx);
 	if (rxq_ctrl->ibv)
-		mlx5_priv_rxq_ibv_release(rxq_ctrl->priv, rxq_ctrl->ibv);
+		mlx5_rxq_ibv_release(rxq_ctrl->ibv);
 	memset(rxq_ctrl, 0, sizeof(*rxq_ctrl));
 }
 
 /**
  * Returns the per-queue supported offloads.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  *
  * @return
  *   Supported Rx offloads.
  */
 uint64_t
-mlx5_priv_get_rx_queue_offloads(struct priv *priv)
+mlx5_get_rx_queue_offloads(struct rte_eth_dev *dev)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_dev_config *config = &priv->config;
 	uint64_t offloads = (DEV_RX_OFFLOAD_SCATTER |
 			     DEV_RX_OFFLOAD_TIMESTAMP |
 			     DEV_RX_OFFLOAD_JUMBO_FRAME);
 
+	offloads |= DEV_RX_OFFLOAD_CRC_STRIP;
 	if (config->hw_fcs_strip)
-		offloads |= DEV_RX_OFFLOAD_CRC_STRIP;
+		offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
+
 	if (config->hw_csum)
 		offloads |= (DEV_RX_OFFLOAD_IPV4_CKSUM |
 			     DEV_RX_OFFLOAD_UDP_CKSUM |
@@ -217,13 +405,11 @@ mlx5_priv_get_rx_queue_offloads(struct priv *priv)
 /**
  * Returns the per-port supported offloads.
  *
- * @param priv
- *   Pointer to private structure.
  * @return
  *   Supported Rx offloads.
  */
 uint64_t
-mlx5_priv_get_rx_port_offloads(struct priv *priv __rte_unused)
+mlx5_get_rx_port_offloads(void)
 {
 	uint64_t offloads = DEV_RX_OFFLOAD_VLAN_FILTER;
 
@@ -231,33 +417,6 @@ mlx5_priv_get_rx_port_offloads(struct priv *priv __rte_unused)
 }
 
 /**
- * Checks if the per-queue offload configuration is valid.
- *
- * @param priv
- *   Pointer to private structure.
- * @param offloads
- *   Per-queue offloads configuration.
- *
- * @return
- *   1 if the configuration is valid, 0 otherwise.
- */
-static int
-priv_is_rx_queue_offloads_allowed(struct priv *priv, uint64_t offloads)
-{
-	uint64_t port_offloads = priv->dev->data->dev_conf.rxmode.offloads;
-	uint64_t queue_supp_offloads =
-		mlx5_priv_get_rx_queue_offloads(priv);
-	uint64_t port_supp_offloads = mlx5_priv_get_rx_port_offloads(priv);
-
-	if ((offloads & (queue_supp_offloads | port_supp_offloads)) !=
-	    offloads)
-		return 0;
-	if (((port_offloads ^ offloads) & port_supp_offloads))
-		return 0;
-	return 1;
-}
-
-/**
  *
  * @param dev
  *   Pointer to Ethernet device structure.
@@ -273,7 +432,7 @@ priv_is_rx_queue_offloads_allowed(struct priv *priv, uint64_t offloads)
  *   Memory pool for buffer allocations.
  *
  * @return
- *   0 on success, negative errno value on failure.
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
  */
 int
 mlx5_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
@@ -284,53 +443,40 @@ mlx5_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
 	struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
 	struct mlx5_rxq_ctrl *rxq_ctrl =
 		container_of(rxq, struct mlx5_rxq_ctrl, rxq);
-	int ret = 0;
 
-	priv_lock(priv);
 	if (!rte_is_power_of_2(desc)) {
 		desc = 1 << log2above(desc);
-		WARN("%p: increased number of descriptors in RX queue %u"
-		     " to the next power of two (%d)",
-		     (void *)dev, idx, desc);
+		DRV_LOG(WARNING,
+			"port %u increased number of descriptors in Rx queue %u"
+			" to the next power of two (%d)",
+			dev->data->port_id, idx, desc);
 	}
-	DEBUG("%p: configuring queue %u for %u descriptors",
-	      (void *)dev, idx, desc);
+	DRV_LOG(DEBUG, "port %u configuring Rx queue %u for %u descriptors",
+		dev->data->port_id, idx, desc);
 	if (idx >= priv->rxqs_n) {
-		ERROR("%p: queue index out of range (%u >= %u)",
-		      (void *)dev, idx, priv->rxqs_n);
-		priv_unlock(priv);
-		return -EOVERFLOW;
-	}
-	if (!priv_is_rx_queue_offloads_allowed(priv, conf->offloads)) {
-		ret = ENOTSUP;
-		ERROR("%p: Rx queue offloads 0x%" PRIx64 " don't match port "
-		      "offloads 0x%" PRIx64 " or supported offloads 0x%" PRIx64,
-		      (void *)dev, conf->offloads,
-		      dev->data->dev_conf.rxmode.offloads,
-		      (mlx5_priv_get_rx_port_offloads(priv) |
-		       mlx5_priv_get_rx_queue_offloads(priv)));
-		goto out;
-	}
-	if (!mlx5_priv_rxq_releasable(priv, idx)) {
-		ret = EBUSY;
-		ERROR("%p: unable to release queue index %u",
-		      (void *)dev, idx);
-		goto out;
-	}
-	mlx5_priv_rxq_release(priv, idx);
-	rxq_ctrl = mlx5_priv_rxq_new(priv, idx, desc, socket, conf, mp);
+		DRV_LOG(ERR, "port %u Rx queue index out of range (%u >= %u)",
+			dev->data->port_id, idx, priv->rxqs_n);
+		rte_errno = EOVERFLOW;
+		return -rte_errno;
+	}
+	if (!mlx5_rxq_releasable(dev, idx)) {
+		DRV_LOG(ERR, "port %u unable to release queue index %u",
+			dev->data->port_id, idx);
+		rte_errno = EBUSY;
+		return -rte_errno;
+	}
+	mlx5_rxq_release(dev, idx);
+	rxq_ctrl = mlx5_rxq_new(dev, idx, desc, socket, conf, mp);
 	if (!rxq_ctrl) {
-		ERROR("%p: unable to allocate queue index %u",
-		      (void *)dev, idx);
-		ret = ENOMEM;
-		goto out;
+		DRV_LOG(ERR, "port %u unable to allocate queue index %u",
+			dev->data->port_id, idx);
+		rte_errno = ENOMEM;
+		return -rte_errno;
 	}
-	DEBUG("%p: adding RX queue %p to list",
-	      (void *)dev, (void *)rxq_ctrl);
+	DRV_LOG(DEBUG, "port %u adding Rx queue %u to list",
+		dev->data->port_id, idx);
 	(*priv->rxqs)[idx] = &rxq_ctrl->rxq;
-out:
-	priv_unlock(priv);
-	return -ret;
+	return 0;
 }
 
 /**
@@ -350,45 +496,48 @@ mlx5_rx_queue_release(void *dpdk_rxq)
 		return;
 	rxq_ctrl = container_of(rxq, struct mlx5_rxq_ctrl, rxq);
 	priv = rxq_ctrl->priv;
-	priv_lock(priv);
-	if (!mlx5_priv_rxq_releasable(priv, rxq_ctrl->rxq.stats.idx))
-		rte_panic("Rx queue %p is still used by a flow and cannot be"
-			  " removed\n", (void *)rxq_ctrl);
-	mlx5_priv_rxq_release(priv, rxq_ctrl->rxq.stats.idx);
-	priv_unlock(priv);
+	if (!mlx5_rxq_releasable(ETH_DEV(priv), rxq_ctrl->rxq.stats.idx))
+		rte_panic("port %u Rx queue %u is still used by a flow and"
+			  " cannot be removed\n",
+			  PORT_ID(priv), rxq_ctrl->idx);
+	mlx5_rxq_release(ETH_DEV(priv), rxq_ctrl->rxq.stats.idx);
 }
 
 /**
  * Allocate queue vector and fill epoll fd list for Rx interrupts.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  *
  * @return
- *   0 on success, negative on failure.
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
  */
 int
-priv_rx_intr_vec_enable(struct priv *priv)
+mlx5_rx_intr_vec_enable(struct rte_eth_dev *dev)
 {
+	struct priv *priv = dev->data->dev_private;
 	unsigned int i;
 	unsigned int rxqs_n = priv->rxqs_n;
 	unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
 	unsigned int count = 0;
-	struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+	struct rte_intr_handle *intr_handle = dev->intr_handle;
 
-	if (!priv->dev->data->dev_conf.intr_conf.rxq)
+	if (!dev->data->dev_conf.intr_conf.rxq)
 		return 0;
-	priv_rx_intr_vec_disable(priv);
+	mlx5_rx_intr_vec_disable(dev);
 	intr_handle->intr_vec = malloc(n * sizeof(intr_handle->intr_vec[0]));
 	if (intr_handle->intr_vec == NULL) {
-		ERROR("failed to allocate memory for interrupt vector,"
-		      " Rx interrupts will not be supported");
-		return -ENOMEM;
+		DRV_LOG(ERR,
+			"port %u failed to allocate memory for interrupt"
+			" vector, Rx interrupts will not be supported",
+			dev->data->port_id);
+		rte_errno = ENOMEM;
+		return -rte_errno;
 	}
 	intr_handle->type = RTE_INTR_HANDLE_EXT;
 	for (i = 0; i != n; ++i) {
 		/* This rxq ibv must not be released in this function. */
-		struct mlx5_rxq_ibv *rxq_ibv = mlx5_priv_rxq_ibv_get(priv, i);
+		struct mlx5_rxq_ibv *rxq_ibv = mlx5_rxq_ibv_get(dev, i);
 		int fd;
 		int flags;
 		int rc;
@@ -402,27 +551,34 @@ priv_rx_intr_vec_enable(struct priv *priv)
 			continue;
 		}
 		if (count >= RTE_MAX_RXTX_INTR_VEC_ID) {
-			ERROR("too many Rx queues for interrupt vector size"
-			      " (%d), Rx interrupts cannot be enabled",
-			      RTE_MAX_RXTX_INTR_VEC_ID);
-			priv_rx_intr_vec_disable(priv);
-			return -1;
+			DRV_LOG(ERR,
+				"port %u too many Rx queues for interrupt"
+				" vector size (%d), Rx interrupts cannot be"
+				" enabled",
+				dev->data->port_id, RTE_MAX_RXTX_INTR_VEC_ID);
+			mlx5_rx_intr_vec_disable(dev);
+			rte_errno = ENOMEM;
+			return -rte_errno;
 		}
 		fd = rxq_ibv->channel->fd;
 		flags = fcntl(fd, F_GETFL);
 		rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
 		if (rc < 0) {
-			ERROR("failed to make Rx interrupt file descriptor"
-			      " %d non-blocking for queue index %d", fd, i);
-			priv_rx_intr_vec_disable(priv);
-			return -1;
+			rte_errno = errno;
+			DRV_LOG(ERR,
+				"port %u failed to make Rx interrupt file"
+				" descriptor %d non-blocking for queue index"
+				" %d",
+				dev->data->port_id, fd, i);
+			mlx5_rx_intr_vec_disable(dev);
+			return -rte_errno;
 		}
 		intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count;
 		intr_handle->efds[count] = fd;
 		count++;
 	}
 	if (!count)
-		priv_rx_intr_vec_disable(priv);
+		mlx5_rx_intr_vec_disable(dev);
 	else
 		intr_handle->nb_efd = count;
 	return 0;
@@ -431,18 +587,19 @@ priv_rx_intr_vec_enable(struct priv *priv)
 /**
  * Clean up Rx interrupts handler.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  */
 void
-priv_rx_intr_vec_disable(struct priv *priv)
+mlx5_rx_intr_vec_disable(struct rte_eth_dev *dev)
 {
-	struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+	struct priv *priv = dev->data->dev_private;
+	struct rte_intr_handle *intr_handle = dev->intr_handle;
 	unsigned int i;
 	unsigned int rxqs_n = priv->rxqs_n;
 	unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
 
-	if (!priv->dev->data->dev_conf.intr_conf.rxq)
+	if (!dev->data->dev_conf.intr_conf.rxq)
 		return;
 	if (!intr_handle->intr_vec)
 		goto free;
@@ -459,7 +616,7 @@ priv_rx_intr_vec_disable(struct priv *priv)
 		 */
 		rxq_data = (*priv->rxqs)[i];
 		rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
-		mlx5_priv_rxq_ibv_release(priv, rxq_ctrl->ibv);
+		mlx5_rxq_ibv_release(rxq_ctrl->ibv);
 	}
 free:
 	rte_intr_free_epoll_fd(intr_handle);
@@ -490,7 +647,8 @@ mlx5_arm_cq(struct mlx5_rxq_data *rxq, int sq_n_rxq)
 	doorbell = (uint64_t)doorbell_hi << 32;
 	doorbell |=  rxq->cqn;
 	rxq->cq_db[MLX5_CQ_ARM_DB] = rte_cpu_to_be_32(doorbell_hi);
-	rte_write64(rte_cpu_to_be_64(doorbell), cq_db_reg);
+	mlx5_uar_write64(rte_cpu_to_be_64(doorbell),
+			 cq_db_reg, rxq->uar_lock_cq);
 }
 
 /**
@@ -502,7 +660,7 @@ mlx5_arm_cq(struct mlx5_rxq_data *rxq, int sq_n_rxq)
  *   Rx queue number.
  *
  * @return
- *   0 on success, negative on failure.
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
  */
 int
 mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
@@ -510,31 +668,25 @@ mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
 	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_data *rxq_data;
 	struct mlx5_rxq_ctrl *rxq_ctrl;
-	int ret = 0;
 
-	priv_lock(priv);
 	rxq_data = (*priv->rxqs)[rx_queue_id];
 	if (!rxq_data) {
-		ret = EINVAL;
-		goto exit;
+		rte_errno = EINVAL;
+		return -rte_errno;
 	}
 	rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
 	if (rxq_ctrl->irq) {
 		struct mlx5_rxq_ibv *rxq_ibv;
 
-		rxq_ibv = mlx5_priv_rxq_ibv_get(priv, rx_queue_id);
+		rxq_ibv = mlx5_rxq_ibv_get(dev, rx_queue_id);
 		if (!rxq_ibv) {
-			ret = EINVAL;
-			goto exit;
+			rte_errno = EINVAL;
+			return -rte_errno;
 		}
 		mlx5_arm_cq(rxq_data, rxq_data->cq_arm_sn);
-		mlx5_priv_rxq_ibv_release(priv, rxq_ibv);
+		mlx5_rxq_ibv_release(rxq_ibv);
 	}
-exit:
-	priv_unlock(priv);
-	if (ret)
-		WARN("unable to arm interrupt on rx queue %d", rx_queue_id);
-	return -ret;
+	return 0;
 }
 
 /**
@@ -546,7 +698,7 @@ exit:
  *   Rx queue number.
  *
  * @return
- *   0 on success, negative on failure.
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
  */
 int
 mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
@@ -557,53 +709,54 @@ mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
 	struct mlx5_rxq_ibv *rxq_ibv = NULL;
 	struct ibv_cq *ev_cq;
 	void *ev_ctx;
-	int ret = 0;
+	int ret;
 
-	priv_lock(priv);
 	rxq_data = (*priv->rxqs)[rx_queue_id];
 	if (!rxq_data) {
-		ret = EINVAL;
-		goto exit;
+		rte_errno = EINVAL;
+		return -rte_errno;
 	}
 	rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
 	if (!rxq_ctrl->irq)
-		goto exit;
-	rxq_ibv = mlx5_priv_rxq_ibv_get(priv, rx_queue_id);
+		return 0;
+	rxq_ibv = mlx5_rxq_ibv_get(dev, rx_queue_id);
 	if (!rxq_ibv) {
-		ret = EINVAL;
-		goto exit;
+		rte_errno = EINVAL;
+		return -rte_errno;
 	}
 	ret = mlx5_glue->get_cq_event(rxq_ibv->channel, &ev_cq, &ev_ctx);
 	if (ret || ev_cq != rxq_ibv->cq) {
-		ret = EINVAL;
+		rte_errno = EINVAL;
 		goto exit;
 	}
 	rxq_data->cq_arm_sn++;
 	mlx5_glue->ack_cq_events(rxq_ibv->cq, 1);
+	return 0;
 exit:
+	ret = rte_errno; /* Save rte_errno before cleanup. */
 	if (rxq_ibv)
-		mlx5_priv_rxq_ibv_release(priv, rxq_ibv);
-	priv_unlock(priv);
-	if (ret)
-		WARN("unable to disable interrupt on rx queue %d",
-		     rx_queue_id);
-	return -ret;
+		mlx5_rxq_ibv_release(rxq_ibv);
+	DRV_LOG(WARNING, "port %u unable to disable interrupt on Rx queue %d",
+		dev->data->port_id, rx_queue_id);
+	rte_errno = ret; /* Restore rte_errno. */
+	return -rte_errno;
 }
 
 /**
  * Create the Rx queue Verbs object.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param idx
  *   Queue index in DPDK Rx queue array
  *
  * @return
- *   The Verbs object initialised if it can be created.
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
  */
-struct mlx5_rxq_ibv*
-mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
+struct mlx5_rxq_ibv *
+mlx5_rxq_ibv_new(struct rte_eth_dev *dev, uint16_t idx)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_data *rxq_data = (*priv->rxqs)[idx];
 	struct mlx5_rxq_ctrl *rxq_ctrl =
 		container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
@@ -613,10 +766,16 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 			struct ibv_cq_init_attr_ex ibv;
 			struct mlx5dv_cq_init_attr mlx5;
 		} cq;
-		struct ibv_wq_init_attr wq;
+		struct {
+			struct ibv_wq_init_attr ibv;
+#ifdef HAVE_IBV_DEVICE_STRIDING_RQ_SUPPORT
+			struct mlx5dv_wq_init_attr mlx5;
+#endif
+		} wq;
 		struct ibv_cq_ex cq_attr;
 	} attr;
-	unsigned int cqe_n = (1 << rxq_data->elts_n) - 1;
+	unsigned int cqe_n;
+	unsigned int wqe_n = 1 << rxq_data->elts_n;
 	struct mlx5_rxq_ibv *tmpl;
 	struct mlx5dv_cq cq_info;
 	struct mlx5dv_rwq rwq;
@@ -624,6 +783,7 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	int ret = 0;
 	struct mlx5dv_obj obj;
 	struct mlx5_dev_config *config = &priv->config;
+	const int mprq_en = mlx5_rxq_mprq_enabled(rxq_data);
 
 	assert(rxq_data);
 	assert(!rxq_ctrl->ibv);
@@ -632,28 +792,26 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	tmpl = rte_calloc_socket(__func__, 1, sizeof(*tmpl), 0,
 				 rxq_ctrl->socket);
 	if (!tmpl) {
-		ERROR("%p: cannot allocate verbs resources",
-		       (void *)rxq_ctrl);
+		DRV_LOG(ERR,
+			"port %u Rx queue %u cannot allocate verbs resources",
+			dev->data->port_id, rxq_ctrl->idx);
+		rte_errno = ENOMEM;
 		goto error;
 	}
 	tmpl->rxq_ctrl = rxq_ctrl;
-	/* Use the entire RX mempool as the memory region. */
-	tmpl->mr = priv_mr_get(priv, rxq_data->mp);
-	if (!tmpl->mr) {
-		tmpl->mr = priv_mr_new(priv, rxq_data->mp);
-		if (!tmpl->mr) {
-			ERROR("%p: MR creation failure", (void *)rxq_ctrl);
-			goto error;
-		}
-	}
 	if (rxq_ctrl->irq) {
 		tmpl->channel = mlx5_glue->create_comp_channel(priv->ctx);
 		if (!tmpl->channel) {
-			ERROR("%p: Comp Channel creation failure",
-			      (void *)rxq_ctrl);
+			DRV_LOG(ERR, "port %u: comp channel creation failure",
+				dev->data->port_id);
+			rte_errno = ENOMEM;
 			goto error;
 		}
 	}
+	if (mprq_en)
+		cqe_n = wqe_n * (1 << rxq_data->strd_num_n) - 1;
+	else
+		cqe_n = wqe_n  - 1;
 	attr.cq.ibv = (struct ibv_cq_init_attr_ex){
 		.cqe = cqe_n,
 		.channel = tmpl->channel,
@@ -665,32 +823,43 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	if (config->cqe_comp && !rxq_data->hw_timestamp) {
 		attr.cq.mlx5.comp_mask |=
 			MLX5DV_CQ_INIT_ATTR_MASK_COMPRESSED_CQE;
+#ifdef HAVE_IBV_DEVICE_STRIDING_RQ_SUPPORT
+		attr.cq.mlx5.cqe_comp_res_format =
+			mprq_en ? MLX5DV_CQE_RES_FORMAT_CSUM_STRIDX :
+				  MLX5DV_CQE_RES_FORMAT_HASH;
+#else
 		attr.cq.mlx5.cqe_comp_res_format = MLX5DV_CQE_RES_FORMAT_HASH;
+#endif
 		/*
 		 * For vectorized Rx, it must not be doubled in order to
 		 * make cq_ci and rq_ci aligned.
 		 */
-		if (rxq_check_vec_support(rxq_data) < 0)
+		if (mlx5_rxq_check_vec_support(rxq_data) < 0)
 			attr.cq.ibv.cqe *= 2;
 	} else if (config->cqe_comp && rxq_data->hw_timestamp) {
-		DEBUG("Rx CQE compression is disabled for HW timestamp");
+		DRV_LOG(DEBUG,
+			"port %u Rx CQE compression is disabled for HW"
+			" timestamp",
+			dev->data->port_id);
 	}
 	tmpl->cq = mlx5_glue->cq_ex_to_cq
 		(mlx5_glue->dv_create_cq(priv->ctx, &attr.cq.ibv,
 					 &attr.cq.mlx5));
 	if (tmpl->cq == NULL) {
-		ERROR("%p: CQ creation failure", (void *)rxq_ctrl);
+		DRV_LOG(ERR, "port %u Rx queue %u CQ creation failure",
+			dev->data->port_id, idx);
+		rte_errno = ENOMEM;
 		goto error;
 	}
-	DEBUG("priv->device_attr.max_qp_wr is %d",
-	      priv->device_attr.orig_attr.max_qp_wr);
-	DEBUG("priv->device_attr.max_sge is %d",
-	      priv->device_attr.orig_attr.max_sge);
-	attr.wq = (struct ibv_wq_init_attr){
+	DRV_LOG(DEBUG, "port %u priv->device_attr.max_qp_wr is %d",
+		dev->data->port_id, priv->device_attr.orig_attr.max_qp_wr);
+	DRV_LOG(DEBUG, "port %u priv->device_attr.max_sge is %d",
+		dev->data->port_id, priv->device_attr.orig_attr.max_sge);
+	attr.wq.ibv = (struct ibv_wq_init_attr){
 		.wq_context = NULL, /* Could be useful in the future. */
 		.wq_type = IBV_WQT_RQ,
 		/* Max number of outstanding WRs. */
-		.max_wr = (1 << rxq_data->elts_n) >> rxq_data->sges_n,
+		.max_wr = wqe_n >> rxq_data->sges_n,
 		/* Max number of scatter/gather elements in a WR. */
 		.max_sge = 1 << rxq_data->sges_n,
 		.pd = priv->pd,
@@ -704,32 +873,54 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	};
 	/* By default, FCS (CRC) is stripped by hardware. */
 	if (rxq_data->crc_present) {
-		attr.wq.create_flags |= IBV_WQ_FLAGS_SCATTER_FCS;
-		attr.wq.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
+		attr.wq.ibv.create_flags |= IBV_WQ_FLAGS_SCATTER_FCS;
+		attr.wq.ibv.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
 	}
 #ifdef HAVE_IBV_WQ_FLAG_RX_END_PADDING
 	if (config->hw_padding) {
-		attr.wq.create_flags |= IBV_WQ_FLAG_RX_END_PADDING;
-		attr.wq.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
+		attr.wq.ibv.create_flags |= IBV_WQ_FLAG_RX_END_PADDING;
+		attr.wq.ibv.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
+	}
+#endif
+#ifdef HAVE_IBV_DEVICE_STRIDING_RQ_SUPPORT
+	attr.wq.mlx5 = (struct mlx5dv_wq_init_attr){
+		.comp_mask = 0,
+	};
+	if (mprq_en) {
+		struct mlx5dv_striding_rq_init_attr *mprq_attr =
+			&attr.wq.mlx5.striding_rq_attrs;
+
+		attr.wq.mlx5.comp_mask |= MLX5DV_WQ_INIT_ATTR_MASK_STRIDING_RQ;
+		*mprq_attr = (struct mlx5dv_striding_rq_init_attr){
+			.single_stride_log_num_of_bytes = rxq_data->strd_sz_n,
+			.single_wqe_log_num_of_strides = rxq_data->strd_num_n,
+			.two_byte_shift_en = MLX5_MPRQ_TWO_BYTE_SHIFT,
+		};
 	}
+	tmpl->wq = mlx5_glue->dv_create_wq(priv->ctx, &attr.wq.ibv,
+					   &attr.wq.mlx5);
+#else
+	tmpl->wq = mlx5_glue->create_wq(priv->ctx, &attr.wq.ibv);
 #endif
-	tmpl->wq = mlx5_glue->create_wq(priv->ctx, &attr.wq);
 	if (tmpl->wq == NULL) {
-		ERROR("%p: WQ creation failure", (void *)rxq_ctrl);
+		DRV_LOG(ERR, "port %u Rx queue %u WQ creation failure",
+			dev->data->port_id, idx);
+		rte_errno = ENOMEM;
 		goto error;
 	}
 	/*
 	 * Make sure number of WRs*SGEs match expectations since a queue
 	 * cannot allocate more than "desc" buffers.
 	 */
-	if (((int)attr.wq.max_wr !=
-	     ((1 << rxq_data->elts_n) >> rxq_data->sges_n)) ||
-	    ((int)attr.wq.max_sge != (1 << rxq_data->sges_n))) {
-		ERROR("%p: requested %u*%u but got %u*%u WRs*SGEs",
-		      (void *)rxq_ctrl,
-		      ((1 << rxq_data->elts_n) >> rxq_data->sges_n),
-		      (1 << rxq_data->sges_n),
-		      attr.wq.max_wr, attr.wq.max_sge);
+	if (attr.wq.ibv.max_wr != (wqe_n >> rxq_data->sges_n) ||
+	    attr.wq.ibv.max_sge != (1u << rxq_data->sges_n)) {
+		DRV_LOG(ERR,
+			"port %u Rx queue %u requested %u*%u but got %u*%u"
+			" WRs*SGEs",
+			dev->data->port_id, idx,
+			wqe_n >> rxq_data->sges_n, (1 << rxq_data->sges_n),
+			attr.wq.ibv.max_wr, attr.wq.ibv.max_sge);
+		rte_errno = EINVAL;
 		goto error;
 	}
 	/* Change queue state to ready. */
@@ -739,8 +930,10 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	};
 	ret = mlx5_glue->modify_wq(tmpl->wq, &mod);
 	if (ret) {
-		ERROR("%p: WQ state to IBV_WQS_RDY failed",
-		      (void *)rxq_ctrl);
+		DRV_LOG(ERR,
+			"port %u Rx queue %u WQ state to IBV_WQS_RDY failed",
+			dev->data->port_id, idx);
+		rte_errno = ret;
 		goto error;
 	}
 	obj.cq.in = tmpl->cq;
@@ -748,33 +941,53 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	obj.rwq.in = tmpl->wq;
 	obj.rwq.out = &rwq;
 	ret = mlx5_glue->dv_init_obj(&obj, MLX5DV_OBJ_CQ | MLX5DV_OBJ_RWQ);
-	if (ret != 0)
+	if (ret) {
+		rte_errno = ret;
 		goto error;
+	}
 	if (cq_info.cqe_size != RTE_CACHE_LINE_SIZE) {
-		ERROR("Wrong MLX5_CQE_SIZE environment variable value: "
-		      "it should be set to %u", RTE_CACHE_LINE_SIZE);
+		DRV_LOG(ERR,
+			"port %u wrong MLX5_CQE_SIZE environment variable"
+			" value: it should be set to %u",
+			dev->data->port_id, RTE_CACHE_LINE_SIZE);
+		rte_errno = EINVAL;
 		goto error;
 	}
 	/* Fill the rings. */
-	rxq_data->wqes = (volatile struct mlx5_wqe_data_seg (*)[])
-		(uintptr_t)rwq.buf;
-	for (i = 0; (i != (unsigned int)(1 << rxq_data->elts_n)); ++i) {
-		struct rte_mbuf *buf = (*rxq_data->elts)[i];
-		volatile struct mlx5_wqe_data_seg *scat = &(*rxq_data->wqes)[i];
-
+	rxq_data->wqes = rwq.buf;
+	for (i = 0; (i != wqe_n); ++i) {
+		volatile struct mlx5_wqe_data_seg *scat;
+		uintptr_t addr;
+		uint32_t byte_count;
+
+		if (mprq_en) {
+			struct mlx5_mprq_buf *buf = (*rxq_data->mprq_bufs)[i];
+
+			scat = &((volatile struct mlx5_wqe_mprq *)
+				 rxq_data->wqes)[i].dseg;
+			addr = (uintptr_t)mlx5_mprq_buf_addr(buf);
+			byte_count = (1 << rxq_data->strd_sz_n) *
+				     (1 << rxq_data->strd_num_n);
+		} else {
+			struct rte_mbuf *buf = (*rxq_data->elts)[i];
+
+			scat = &((volatile struct mlx5_wqe_data_seg *)
+				 rxq_data->wqes)[i];
+			addr = rte_pktmbuf_mtod(buf, uintptr_t);
+			byte_count = DATA_LEN(buf);
+		}
 		/* scat->addr must be able to store a pointer. */
 		assert(sizeof(scat->addr) >= sizeof(uintptr_t));
 		*scat = (struct mlx5_wqe_data_seg){
-			.addr = rte_cpu_to_be_64(rte_pktmbuf_mtod(buf,
-								  uintptr_t)),
-			.byte_count = rte_cpu_to_be_32(DATA_LEN(buf)),
-			.lkey = tmpl->mr->lkey,
+			.addr = rte_cpu_to_be_64(addr),
+			.byte_count = rte_cpu_to_be_32(byte_count),
+			.lkey = mlx5_rx_addr2mr(rxq_data, addr),
 		};
 	}
 	rxq_data->rq_db = rwq.dbrec;
 	rxq_data->cqe_n = log2above(cq_info.cqe_cnt);
 	rxq_data->cq_ci = 0;
-	rxq_data->rq_ci = 0;
+	rxq_data->consumed_strd = 0;
 	rxq_data->rq_pi = 0;
 	rxq_data->zip = (struct rxq_zip){
 		.ai = 0,
@@ -785,43 +998,43 @@ mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
 	rxq_data->cqn = cq_info.cqn;
 	rxq_data->cq_arm_sn = 0;
 	/* Update doorbell counter. */
-	rxq_data->rq_ci = (1 << rxq_data->elts_n) >> rxq_data->sges_n;
+	rxq_data->rq_ci = wqe_n >> rxq_data->sges_n;
 	rte_wmb();
 	*rxq_data->rq_db = rte_cpu_to_be_32(rxq_data->rq_ci);
-	DEBUG("%p: rxq updated with %p", (void *)rxq_ctrl, (void *)&tmpl);
+	DRV_LOG(DEBUG, "port %u rxq %u updated with %p", dev->data->port_id,
+		idx, (void *)&tmpl);
 	rte_atomic32_inc(&tmpl->refcnt);
-	DEBUG("%p: Verbs Rx queue %p: refcnt %d", (void *)priv,
-	      (void *)tmpl, rte_atomic32_read(&tmpl->refcnt));
 	LIST_INSERT_HEAD(&priv->rxqsibv, tmpl, next);
 	priv->verbs_alloc_ctx.type = MLX5_VERBS_ALLOC_TYPE_NONE;
 	return tmpl;
 error:
+	ret = rte_errno; /* Save rte_errno before cleanup. */
 	if (tmpl->wq)
 		claim_zero(mlx5_glue->destroy_wq(tmpl->wq));
 	if (tmpl->cq)
 		claim_zero(mlx5_glue->destroy_cq(tmpl->cq));
 	if (tmpl->channel)
 		claim_zero(mlx5_glue->destroy_comp_channel(tmpl->channel));
-	if (tmpl->mr)
-		priv_mr_release(priv, tmpl->mr);
 	priv->verbs_alloc_ctx.type = MLX5_VERBS_ALLOC_TYPE_NONE;
+	rte_errno = ret; /* Restore rte_errno. */
 	return NULL;
 }
 
 /**
  * Get an Rx queue Verbs object.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param idx
  *   Queue index in DPDK Rx queue array
  *
  * @return
  *   The Verbs object if it exists.
  */
-struct mlx5_rxq_ibv*
-mlx5_priv_rxq_ibv_get(struct priv *priv, uint16_t idx)
+struct mlx5_rxq_ibv *
+mlx5_rxq_ibv_get(struct rte_eth_dev *dev, uint16_t idx)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_data *rxq_data = (*priv->rxqs)[idx];
 	struct mlx5_rxq_ctrl *rxq_ctrl;
 
@@ -831,11 +1044,7 @@ mlx5_priv_rxq_ibv_get(struct priv *priv, uint16_t idx)
 		return NULL;
 	rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
 	if (rxq_ctrl->ibv) {
-		priv_mr_get(priv, rxq_data->mp);
 		rte_atomic32_inc(&rxq_ctrl->ibv->refcnt);
-		DEBUG("%p: Verbs Rx queue %p: refcnt %d", (void *)priv,
-		      (void *)rxq_ctrl->ibv,
-		      rte_atomic32_read(&rxq_ctrl->ibv->refcnt));
 	}
 	return rxq_ctrl->ibv;
 }
@@ -843,28 +1052,18 @@ mlx5_priv_rxq_ibv_get(struct priv *priv, uint16_t idx)
 /**
  * Release an Rx verbs queue object.
  *
- * @param priv
- *   Pointer to private structure.
  * @param rxq_ibv
  *   Verbs Rx queue object.
  *
  * @return
- *   0 on success, errno value on failure.
+ *   1 while a reference on it exists, 0 when freed.
  */
 int
-mlx5_priv_rxq_ibv_release(struct priv *priv, struct mlx5_rxq_ibv *rxq_ibv)
+mlx5_rxq_ibv_release(struct mlx5_rxq_ibv *rxq_ibv)
 {
-	int ret;
-
 	assert(rxq_ibv);
 	assert(rxq_ibv->wq);
 	assert(rxq_ibv->cq);
-	assert(rxq_ibv->mr);
-	ret = priv_mr_release(priv, rxq_ibv->mr);
-	if (!ret)
-		rxq_ibv->mr = NULL;
-	DEBUG("%p: Verbs Rx queue %p: refcnt %d", (void *)priv,
-	      (void *)rxq_ibv, rte_atomic32_read(&rxq_ibv->refcnt));
 	if (rte_atomic32_dec_and_test(&rxq_ibv->refcnt)) {
 		rxq_free_elts(rxq_ibv->rxq_ctrl);
 		claim_zero(mlx5_glue->destroy_wq(rxq_ibv->wq));
@@ -876,26 +1075,28 @@ mlx5_priv_rxq_ibv_release(struct priv *priv, struct mlx5_rxq_ibv *rxq_ibv)
 		rte_free(rxq_ibv);
 		return 0;
 	}
-	return EBUSY;
+	return 1;
 }
 
 /**
  * Verify the Verbs Rx queue list is empty
  *
- * @param priv
- *  Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  *
- * @return the number of object not released.
+ * @return
+ *   The number of object not released.
  */
 int
-mlx5_priv_rxq_ibv_verify(struct priv *priv)
+mlx5_rxq_ibv_verify(struct rte_eth_dev *dev)
 {
+	struct priv *priv = dev->data->dev_private;
 	int ret = 0;
 	struct mlx5_rxq_ibv *rxq_ibv;
 
 	LIST_FOREACH(rxq_ibv, &priv->rxqsibv, next) {
-		DEBUG("%p: Verbs Rx queue %p still referenced", (void *)priv,
-		      (void *)rxq_ibv);
+		DRV_LOG(DEBUG, "port %u Verbs Rx queue %u still referenced",
+			dev->data->port_id, rxq_ibv->rxq_ctrl->idx);
 		++ret;
 	}
 	return ret;
@@ -904,65 +1105,279 @@ mlx5_priv_rxq_ibv_verify(struct priv *priv)
 /**
  * Return true if a single reference exists on the object.
  *
- * @param priv
- *   Pointer to private structure.
  * @param rxq_ibv
  *   Verbs Rx queue object.
  */
 int
-mlx5_priv_rxq_ibv_releasable(struct priv *priv, struct mlx5_rxq_ibv *rxq_ibv)
+mlx5_rxq_ibv_releasable(struct mlx5_rxq_ibv *rxq_ibv)
 {
-	(void)priv;
 	assert(rxq_ibv);
 	return (rte_atomic32_read(&rxq_ibv->refcnt) == 1);
 }
 
 /**
+ * Callback function to initialize mbufs for Multi-Packet RQ.
+ */
+static inline void
+mlx5_mprq_buf_init(struct rte_mempool *mp, void *opaque_arg __rte_unused,
+		    void *_m, unsigned int i __rte_unused)
+{
+	struct mlx5_mprq_buf *buf = _m;
+
+	memset(_m, 0, sizeof(*buf));
+	buf->mp = mp;
+	rte_atomic16_set(&buf->refcnt, 1);
+}
+
+/**
+ * Free mempool of Multi-Packet RQ.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   0 on success, negative errno value on failure.
+ */
+int
+mlx5_mprq_free_mp(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct rte_mempool *mp = priv->mprq_mp;
+	unsigned int i;
+
+	if (mp == NULL)
+		return 0;
+	DRV_LOG(DEBUG, "port %u freeing mempool (%s) for Multi-Packet RQ",
+		dev->data->port_id, mp->name);
+	/*
+	 * If a buffer in the pool has been externally attached to a mbuf and it
+	 * is still in use by application, destroying the Rx qeueue can spoil
+	 * the packet. It is unlikely to happen but if application dynamically
+	 * creates and destroys with holding Rx packets, this can happen.
+	 *
+	 * TODO: It is unavoidable for now because the mempool for Multi-Packet
+	 * RQ isn't provided by application but managed by PMD.
+	 */
+	if (!rte_mempool_full(mp)) {
+		DRV_LOG(ERR,
+			"port %u mempool for Multi-Packet RQ is still in use",
+			dev->data->port_id);
+		rte_errno = EBUSY;
+		return -rte_errno;
+	}
+	rte_mempool_free(mp);
+	/* Unset mempool for each Rx queue. */
+	for (i = 0; i != priv->rxqs_n; ++i) {
+		struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+
+		if (rxq == NULL)
+			continue;
+		rxq->mprq_mp = NULL;
+	}
+	return 0;
+}
+
+/**
+ * Allocate a mempool for Multi-Packet RQ. All configured Rx queues share the
+ * mempool. If already allocated, reuse it if there're enough elements.
+ * Otherwise, resize it.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   0 on success, negative errno value on failure.
+ */
+int
+mlx5_mprq_alloc_mp(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct rte_mempool *mp = priv->mprq_mp;
+	char name[RTE_MEMPOOL_NAMESIZE];
+	unsigned int desc = 0;
+	unsigned int buf_len;
+	unsigned int obj_num;
+	unsigned int obj_size;
+	unsigned int strd_num_n = 0;
+	unsigned int strd_sz_n = 0;
+	unsigned int i;
+
+	if (!mlx5_mprq_enabled(dev))
+		return 0;
+	/* Count the total number of descriptors configured. */
+	for (i = 0; i != priv->rxqs_n; ++i) {
+		struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+
+		if (rxq == NULL)
+			continue;
+		desc += 1 << rxq->elts_n;
+		/* Get the max number of strides. */
+		if (strd_num_n < rxq->strd_num_n)
+			strd_num_n = rxq->strd_num_n;
+		/* Get the max size of a stride. */
+		if (strd_sz_n < rxq->strd_sz_n)
+			strd_sz_n = rxq->strd_sz_n;
+	}
+	assert(strd_num_n && strd_sz_n);
+	buf_len = (1 << strd_num_n) * (1 << strd_sz_n);
+	obj_size = buf_len + sizeof(struct mlx5_mprq_buf);
+	/*
+	 * Received packets can be either memcpy'd or externally referenced. In
+	 * case that the packet is attached to an mbuf as an external buffer, as
+	 * it isn't possible to predict how the buffers will be queued by
+	 * application, there's no option to exactly pre-allocate needed buffers
+	 * in advance but to speculatively prepares enough buffers.
+	 *
+	 * In the data path, if this Mempool is depleted, PMD will try to memcpy
+	 * received packets to buffers provided by application (rxq->mp) until
+	 * this Mempool gets available again.
+	 */
+	desc *= 4;
+	obj_num = desc + MLX5_MPRQ_MP_CACHE_SZ * priv->rxqs_n;
+	/*
+	 * rte_mempool_create_empty() has sanity check to refuse large cache
+	 * size compared to the number of elements.
+	 * CACHE_FLUSHTHRESH_MULTIPLIER is defined in a C file, so using a
+	 * constant number 2 instead.
+	 */
+	obj_num = RTE_MAX(obj_num, MLX5_MPRQ_MP_CACHE_SZ * 2);
+	/* Check a mempool is already allocated and if it can be resued. */
+	if (mp != NULL && mp->elt_size >= obj_size && mp->size >= obj_num) {
+		DRV_LOG(DEBUG, "port %u mempool %s is being reused",
+			dev->data->port_id, mp->name);
+		/* Reuse. */
+		goto exit;
+	} else if (mp != NULL) {
+		DRV_LOG(DEBUG, "port %u mempool %s should be resized, freeing it",
+			dev->data->port_id, mp->name);
+		/*
+		 * If failed to free, which means it may be still in use, no way
+		 * but to keep using the existing one. On buffer underrun,
+		 * packets will be memcpy'd instead of external buffer
+		 * attachment.
+		 */
+		if (mlx5_mprq_free_mp(dev)) {
+			if (mp->elt_size >= obj_size)
+				goto exit;
+			else
+				return -rte_errno;
+		}
+	}
+	snprintf(name, sizeof(name), "%s-mprq", dev->device->name);
+	mp = rte_mempool_create(name, obj_num, obj_size, MLX5_MPRQ_MP_CACHE_SZ,
+				0, NULL, NULL, mlx5_mprq_buf_init, NULL,
+				dev->device->numa_node, 0);
+	if (mp == NULL) {
+		DRV_LOG(ERR,
+			"port %u failed to allocate a mempool for"
+			" Multi-Packet RQ, count=%u, size=%u",
+			dev->data->port_id, obj_num, obj_size);
+		rte_errno = ENOMEM;
+		return -rte_errno;
+	}
+	priv->mprq_mp = mp;
+exit:
+	/* Set mempool for each Rx queue. */
+	for (i = 0; i != priv->rxqs_n; ++i) {
+		struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+
+		if (rxq == NULL)
+			continue;
+		rxq->mprq_mp = mp;
+	}
+	DRV_LOG(INFO, "port %u Multi-Packet RQ is configured",
+		dev->data->port_id);
+	return 0;
+}
+
+/**
  * Create a DPDK Rx queue.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param idx
- *   TX queue index.
+ *   RX queue index.
  * @param desc
  *   Number of descriptors to configure in queue.
  * @param socket
  *   NUMA socket on which memory must be allocated.
  *
  * @return
- *   A DPDK queue object on success.
+ *   A DPDK queue object on success, NULL otherwise and rte_errno is set.
  */
-struct mlx5_rxq_ctrl*
-mlx5_priv_rxq_new(struct priv *priv, uint16_t idx, uint16_t desc,
-		  unsigned int socket, const struct rte_eth_rxconf *conf,
-		  struct rte_mempool *mp)
+struct mlx5_rxq_ctrl *
+mlx5_rxq_new(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+	     unsigned int socket, const struct rte_eth_rxconf *conf,
+	     struct rte_mempool *mp)
 {
-	struct rte_eth_dev *dev = priv->dev;
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_ctrl *tmpl;
 	unsigned int mb_len = rte_pktmbuf_data_room_size(mp);
+	unsigned int mprq_stride_size;
 	struct mlx5_dev_config *config = &priv->config;
 	/*
 	 * Always allocate extra slots, even if eventually
 	 * the vector Rx will not be used.
 	 */
-	const uint16_t desc_n =
+	uint16_t desc_n =
 		desc + config->rx_vec_en * MLX5_VPMD_DESCS_PER_LOOP;
+	uint64_t offloads = conf->offloads |
+			   dev->data->dev_conf.rxmode.offloads;
+	const int mprq_en = mlx5_check_mprq_support(dev) > 0;
 
 	tmpl = rte_calloc_socket("RXQ", 1,
 				 sizeof(*tmpl) +
 				 desc_n * sizeof(struct rte_mbuf *),
 				 0, socket);
-	if (!tmpl)
+	if (!tmpl) {
+		rte_errno = ENOMEM;
 		return NULL;
+	}
+	if (mlx5_mr_btree_init(&tmpl->rxq.mr_ctrl.cache_bh,
+			       MLX5_MR_BTREE_CACHE_N, socket)) {
+		/* rte_errno is already set. */
+		goto error;
+	}
 	tmpl->socket = socket;
-	if (priv->dev->data->dev_conf.intr_conf.rxq)
+	if (dev->data->dev_conf.intr_conf.rxq)
 		tmpl->irq = 1;
-	/* Enable scattered packets support for this queue if necessary. */
+	/*
+	 * This Rx queue can be configured as a Multi-Packet RQ if all of the
+	 * following conditions are met:
+	 *  - MPRQ is enabled.
+	 *  - The number of descs is more than the number of strides.
+	 *  - max_rx_pkt_len plus overhead is less than the max size of a
+	 *    stride.
+	 *  Otherwise, enable Rx scatter if necessary.
+	 */
 	assert(mb_len >= RTE_PKTMBUF_HEADROOM);
-	if (dev->data->dev_conf.rxmode.max_rx_pkt_len <=
-	    (mb_len - RTE_PKTMBUF_HEADROOM)) {
+	mprq_stride_size =
+		dev->data->dev_conf.rxmode.max_rx_pkt_len +
+		sizeof(struct rte_mbuf_ext_shared_info) +
+		RTE_PKTMBUF_HEADROOM;
+	if (mprq_en &&
+	    desc > (1U << config->mprq.stride_num_n) &&
+	    mprq_stride_size <= (1U << config->mprq.max_stride_size_n)) {
+		/* TODO: Rx scatter isn't supported yet. */
+		tmpl->rxq.sges_n = 0;
+		/* Trim the number of descs needed. */
+		desc >>= config->mprq.stride_num_n;
+		tmpl->rxq.strd_num_n = config->mprq.stride_num_n;
+		tmpl->rxq.strd_sz_n = RTE_MAX(log2above(mprq_stride_size),
+					      config->mprq.min_stride_size_n);
+		tmpl->rxq.strd_shift_en = MLX5_MPRQ_TWO_BYTE_SHIFT;
+		tmpl->rxq.mprq_max_memcpy_len =
+			RTE_MIN(mb_len - RTE_PKTMBUF_HEADROOM,
+				config->mprq.max_memcpy_len);
+		DRV_LOG(DEBUG,
+			"port %u Rx queue %u: Multi-Packet RQ is enabled"
+			" strd_num_n = %u, strd_sz_n = %u",
+			dev->data->port_id, idx,
+			tmpl->rxq.strd_num_n, tmpl->rxq.strd_sz_n);
+	} else if (dev->data->dev_conf.rxmode.max_rx_pkt_len <=
+		   (mb_len - RTE_PKTMBUF_HEADROOM)) {
 		tmpl->rxq.sges_n = 0;
-	} else if (conf->offloads & DEV_RX_OFFLOAD_SCATTER) {
+	} else if (offloads & DEV_RX_OFFLOAD_SCATTER) {
 		unsigned int size =
 			RTE_PKTMBUF_HEADROOM +
 			dev->data->dev_conf.rxmode.max_rx_pkt_len;
@@ -978,57 +1393,71 @@ mlx5_priv_rxq_new(struct priv *priv, uint16_t idx, uint16_t desc,
 		size = mb_len * (1 << tmpl->rxq.sges_n);
 		size -= RTE_PKTMBUF_HEADROOM;
 		if (size < dev->data->dev_conf.rxmode.max_rx_pkt_len) {
-			ERROR("%p: too many SGEs (%u) needed to handle"
-			      " requested maximum packet size %u",
-			      (void *)dev,
-			      1 << sges_n,
-			      dev->data->dev_conf.rxmode.max_rx_pkt_len);
+			DRV_LOG(ERR,
+				"port %u too many SGEs (%u) needed to handle"
+				" requested maximum packet size %u",
+				dev->data->port_id,
+				1 << sges_n,
+				dev->data->dev_conf.rxmode.max_rx_pkt_len);
+			rte_errno = EOVERFLOW;
 			goto error;
 		}
 	} else {
-		WARN("%p: the requested maximum Rx packet size (%u) is"
-		     " larger than a single mbuf (%u) and scattered"
-		     " mode has not been requested",
-		     (void *)dev,
-		     dev->data->dev_conf.rxmode.max_rx_pkt_len,
-		     mb_len - RTE_PKTMBUF_HEADROOM);
-	}
-	DEBUG("%p: maximum number of segments per packet: %u",
-	      (void *)dev, 1 << tmpl->rxq.sges_n);
+		DRV_LOG(WARNING,
+			"port %u the requested maximum Rx packet size (%u) is"
+			" larger than a single mbuf (%u) and scattered mode has"
+			" not been requested",
+			dev->data->port_id,
+			dev->data->dev_conf.rxmode.max_rx_pkt_len,
+			mb_len - RTE_PKTMBUF_HEADROOM);
+	}
+	if (mprq_en && !mlx5_rxq_mprq_enabled(&tmpl->rxq))
+		DRV_LOG(WARNING,
+			"port %u MPRQ is requested but cannot be enabled"
+			" (requested: desc = %u, stride_sz = %u,"
+			" supported: min_stride_num = %u, max_stride_sz = %u).",
+			dev->data->port_id, desc, mprq_stride_size,
+			(1 << config->mprq.stride_num_n),
+			(1 << config->mprq.max_stride_size_n));
+	DRV_LOG(DEBUG, "port %u maximum number of segments per packet: %u",
+		dev->data->port_id, 1 << tmpl->rxq.sges_n);
 	if (desc % (1 << tmpl->rxq.sges_n)) {
-		ERROR("%p: number of RX queue descriptors (%u) is not a"
-		      " multiple of SGEs per packet (%u)",
-		      (void *)dev,
-		      desc,
-		      1 << tmpl->rxq.sges_n);
+		DRV_LOG(ERR,
+			"port %u number of Rx queue descriptors (%u) is not a"
+			" multiple of SGEs per packet (%u)",
+			dev->data->port_id,
+			desc,
+			1 << tmpl->rxq.sges_n);
+		rte_errno = EINVAL;
 		goto error;
 	}
 	/* Toggle RX checksum offload if hardware supports it. */
-	tmpl->rxq.csum = !!(conf->offloads & DEV_RX_OFFLOAD_CHECKSUM);
-	tmpl->rxq.csum_l2tun = (!!(conf->offloads & DEV_RX_OFFLOAD_CHECKSUM) &&
-				priv->config.hw_csum_l2tun);
-	tmpl->rxq.hw_timestamp = !!(conf->offloads & DEV_RX_OFFLOAD_TIMESTAMP);
+	tmpl->rxq.csum = !!(offloads & DEV_RX_OFFLOAD_CHECKSUM);
+	tmpl->rxq.hw_timestamp = !!(offloads & DEV_RX_OFFLOAD_TIMESTAMP);
 	/* Configure VLAN stripping. */
-	tmpl->rxq.vlan_strip = !!(conf->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
+	tmpl->rxq.vlan_strip = !!(offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
 	/* By default, FCS (CRC) is stripped by hardware. */
-	if (conf->offloads & DEV_RX_OFFLOAD_CRC_STRIP) {
-		tmpl->rxq.crc_present = 0;
-	} else if (config->hw_fcs_strip) {
-		tmpl->rxq.crc_present = 1;
-	} else {
-		WARN("%p: CRC stripping has been disabled but will still"
-		     " be performed by hardware, make sure MLNX_OFED and"
-		     " firmware are up to date",
-		     (void *)dev);
-		tmpl->rxq.crc_present = 0;
-	}
-	DEBUG("%p: CRC stripping is %s, %u bytes will be subtracted from"
-	      " incoming frames to hide it",
-	      (void *)dev,
-	      tmpl->rxq.crc_present ? "disabled" : "enabled",
-	      tmpl->rxq.crc_present << 2);
+	tmpl->rxq.crc_present = 0;
+	if (rte_eth_dev_must_keep_crc(offloads)) {
+		if (config->hw_fcs_strip) {
+			tmpl->rxq.crc_present = 1;
+		} else {
+			DRV_LOG(WARNING,
+				"port %u CRC stripping has been disabled but will"
+				" still be performed by hardware, make sure MLNX_OFED"
+				" and firmware are up to date",
+				dev->data->port_id);
+		}
+	}
+	DRV_LOG(DEBUG,
+		"port %u CRC stripping is %s, %u bytes will be subtracted from"
+		" incoming frames to hide it",
+		dev->data->port_id,
+		tmpl->rxq.crc_present ? "disabled" : "enabled",
+		tmpl->rxq.crc_present << 2);
 	/* Save port ID. */
-	tmpl->rxq.rss_hash = priv->rxqs_n > 1;
+	tmpl->rxq.rss_hash = !!priv->rss_conf.rss_hf &&
+		(!!(dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS));
 	tmpl->rxq.port_id = dev->data->port_id;
 	tmpl->priv = priv;
 	tmpl->rxq.mp = mp;
@@ -1036,9 +1465,11 @@ mlx5_priv_rxq_new(struct priv *priv, uint16_t idx, uint16_t desc,
 	tmpl->rxq.elts_n = log2above(desc);
 	tmpl->rxq.elts =
 		(struct rte_mbuf *(*)[1 << tmpl->rxq.elts_n])(tmpl + 1);
+#ifndef RTE_ARCH_64
+	tmpl->rxq.uar_lock_cq = &priv->uar_lock_cq;
+#endif
+	tmpl->idx = idx;
 	rte_atomic32_inc(&tmpl->refcnt);
-	DEBUG("%p: Rx queue %p: refcnt %d", (void *)priv,
-	      (void *)tmpl, rte_atomic32_read(&tmpl->refcnt));
 	LIST_INSERT_HEAD(&priv->rxqsctrl, tmpl, next);
 	return tmpl;
 error:
@@ -1049,28 +1480,26 @@ error:
 /**
  * Get a Rx queue.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param idx
  *   TX queue index.
  *
  * @return
- *   A pointer to the queue if it exists.
+ *   A pointer to the queue if it exists, NULL otherwise.
  */
-struct mlx5_rxq_ctrl*
-mlx5_priv_rxq_get(struct priv *priv, uint16_t idx)
+struct mlx5_rxq_ctrl *
+mlx5_rxq_get(struct rte_eth_dev *dev, uint16_t idx)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_ctrl *rxq_ctrl = NULL;
 
 	if ((*priv->rxqs)[idx]) {
 		rxq_ctrl = container_of((*priv->rxqs)[idx],
 					struct mlx5_rxq_ctrl,
 					rxq);
-
-		mlx5_priv_rxq_ibv_get(priv, idx);
+		mlx5_rxq_ibv_get(dev, idx);
 		rte_atomic32_inc(&rxq_ctrl->refcnt);
-		DEBUG("%p: Rx queue %p: refcnt %d", (void *)priv,
-		      (void *)rxq_ctrl, rte_atomic32_read(&rxq_ctrl->refcnt));
 	}
 	return rxq_ctrl;
 }
@@ -1078,59 +1507,58 @@ mlx5_priv_rxq_get(struct priv *priv, uint16_t idx)
 /**
  * Release a Rx queue.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param idx
  *   TX queue index.
  *
  * @return
- *   0 on success, errno value on failure.
+ *   1 while a reference on it exists, 0 when freed.
  */
 int
-mlx5_priv_rxq_release(struct priv *priv, uint16_t idx)
+mlx5_rxq_release(struct rte_eth_dev *dev, uint16_t idx)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_ctrl *rxq_ctrl;
 
 	if (!(*priv->rxqs)[idx])
 		return 0;
 	rxq_ctrl = container_of((*priv->rxqs)[idx], struct mlx5_rxq_ctrl, rxq);
 	assert(rxq_ctrl->priv);
-	if (rxq_ctrl->ibv) {
-		int ret;
-
-		ret = mlx5_priv_rxq_ibv_release(rxq_ctrl->priv, rxq_ctrl->ibv);
-		if (!ret)
-			rxq_ctrl->ibv = NULL;
-	}
-	DEBUG("%p: Rx queue %p: refcnt %d", (void *)priv,
-	      (void *)rxq_ctrl, rte_atomic32_read(&rxq_ctrl->refcnt));
+	if (rxq_ctrl->ibv && !mlx5_rxq_ibv_release(rxq_ctrl->ibv))
+		rxq_ctrl->ibv = NULL;
 	if (rte_atomic32_dec_and_test(&rxq_ctrl->refcnt)) {
+		mlx5_mr_btree_free(&rxq_ctrl->rxq.mr_ctrl.cache_bh);
 		LIST_REMOVE(rxq_ctrl, next);
 		rte_free(rxq_ctrl);
 		(*priv->rxqs)[idx] = NULL;
 		return 0;
 	}
-	return EBUSY;
+	return 1;
 }
 
 /**
  * Verify if the queue can be released.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param idx
  *   TX queue index.
  *
  * @return
- *   1 if the queue can be released.
+ *   1 if the queue can be released, negative errno otherwise and rte_errno is
+ *   set.
  */
 int
-mlx5_priv_rxq_releasable(struct priv *priv, uint16_t idx)
+mlx5_rxq_releasable(struct rte_eth_dev *dev, uint16_t idx)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_ctrl *rxq_ctrl;
 
-	if (!(*priv->rxqs)[idx])
-		return -1;
+	if (!(*priv->rxqs)[idx]) {
+		rte_errno = EINVAL;
+		return -rte_errno;
+	}
 	rxq_ctrl = container_of((*priv->rxqs)[idx], struct mlx5_rxq_ctrl, rxq);
 	return (rte_atomic32_read(&rxq_ctrl->refcnt) == 1);
 }
@@ -1138,20 +1566,22 @@ mlx5_priv_rxq_releasable(struct priv *priv, uint16_t idx)
 /**
  * Verify the Rx Queue list is empty
  *
- * @param priv
- *  Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  *
- * @return the number of object not released.
+ * @return
+ *   The number of object not released.
  */
 int
-mlx5_priv_rxq_verify(struct priv *priv)
+mlx5_rxq_verify(struct rte_eth_dev *dev)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_rxq_ctrl *rxq_ctrl;
 	int ret = 0;
 
 	LIST_FOREACH(rxq_ctrl, &priv->rxqsctrl, next) {
-		DEBUG("%p: Rx Queue %p still referenced", (void *)priv,
-		      (void *)rxq_ctrl);
+		DRV_LOG(DEBUG, "port %u Rx Queue %u still referenced",
+			dev->data->port_id, rxq_ctrl->idx);
 		++ret;
 	}
 	return ret;
@@ -1160,20 +1590,21 @@ mlx5_priv_rxq_verify(struct priv *priv)
 /**
  * Create an indirection table.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param queues
  *   Queues entering in the indirection table.
  * @param queues_n
  *   Number of queues in the array.
  *
  * @return
- *   A new indirection table.
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
  */
-struct mlx5_ind_table_ibv*
-mlx5_priv_ind_table_ibv_new(struct priv *priv, uint16_t queues[],
-			    uint16_t queues_n)
+struct mlx5_ind_table_ibv *
+mlx5_ind_table_ibv_new(struct rte_eth_dev *dev, const uint16_t *queues,
+		       uint32_t queues_n)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_ind_table_ibv *ind_tbl;
 	const unsigned int wq_n = rte_is_power_of_2(queues_n) ?
 		log2above(queues_n) :
@@ -1184,11 +1615,12 @@ mlx5_priv_ind_table_ibv_new(struct priv *priv, uint16_t queues[],
 
 	ind_tbl = rte_calloc(__func__, 1, sizeof(*ind_tbl) +
 			     queues_n * sizeof(uint16_t), 0);
-	if (!ind_tbl)
+	if (!ind_tbl) {
+		rte_errno = ENOMEM;
 		return NULL;
+	}
 	for (i = 0; i != queues_n; ++i) {
-		struct mlx5_rxq_ctrl *rxq =
-			mlx5_priv_rxq_get(priv, queues[i]);
+		struct mlx5_rxq_ctrl *rxq = mlx5_rxq_get(dev, queues[i]);
 
 		if (!rxq)
 			goto error;
@@ -1206,24 +1638,24 @@ mlx5_priv_ind_table_ibv_new(struct priv *priv, uint16_t queues[],
 			.ind_tbl = wq,
 			.comp_mask = 0,
 		 });
-	if (!ind_tbl->ind_table)
+	if (!ind_tbl->ind_table) {
+		rte_errno = errno;
 		goto error;
+	}
 	rte_atomic32_inc(&ind_tbl->refcnt);
 	LIST_INSERT_HEAD(&priv->ind_tbls, ind_tbl, next);
-	DEBUG("%p: Indirection table %p: refcnt %d", (void *)priv,
-	      (void *)ind_tbl, rte_atomic32_read(&ind_tbl->refcnt));
 	return ind_tbl;
 error:
 	rte_free(ind_tbl);
-	DEBUG("%p cannot create indirection table", (void *)priv);
+	DEBUG("port %u cannot create indirection table", dev->data->port_id);
 	return NULL;
 }
 
 /**
  * Get an indirection table.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param queues
  *   Queues entering in the indirection table.
  * @param queues_n
@@ -1232,10 +1664,11 @@ error:
  * @return
  *   An indirection table if found.
  */
-struct mlx5_ind_table_ibv*
-mlx5_priv_ind_table_ibv_get(struct priv *priv, uint16_t queues[],
-			    uint16_t queues_n)
+struct mlx5_ind_table_ibv *
+mlx5_ind_table_ibv_get(struct rte_eth_dev *dev, const uint16_t *queues,
+		       uint32_t queues_n)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_ind_table_ibv *ind_tbl;
 
 	LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
@@ -1249,10 +1682,8 @@ mlx5_priv_ind_table_ibv_get(struct priv *priv, uint16_t queues[],
 		unsigned int i;
 
 		rte_atomic32_inc(&ind_tbl->refcnt);
-		DEBUG("%p: Indirection table %p: refcnt %d", (void *)priv,
-		      (void *)ind_tbl, rte_atomic32_read(&ind_tbl->refcnt));
 		for (i = 0; i != ind_tbl->queues_n; ++i)
-			mlx5_priv_rxq_get(priv, ind_tbl->queues[i]);
+			mlx5_rxq_get(dev, ind_tbl->queues[i]);
 	}
 	return ind_tbl;
 }
@@ -1260,52 +1691,53 @@ mlx5_priv_ind_table_ibv_get(struct priv *priv, uint16_t queues[],
 /**
  * Release an indirection table.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param ind_table
  *   Indirection table to release.
  *
  * @return
- *   0 on success, errno value on failure.
+ *   1 while a reference on it exists, 0 when freed.
  */
 int
-mlx5_priv_ind_table_ibv_release(struct priv *priv,
-				struct mlx5_ind_table_ibv *ind_tbl)
+mlx5_ind_table_ibv_release(struct rte_eth_dev *dev,
+			   struct mlx5_ind_table_ibv *ind_tbl)
 {
 	unsigned int i;
 
-	DEBUG("%p: Indirection table %p: refcnt %d", (void *)priv,
-	      (void *)ind_tbl, rte_atomic32_read(&ind_tbl->refcnt));
 	if (rte_atomic32_dec_and_test(&ind_tbl->refcnt))
 		claim_zero(mlx5_glue->destroy_rwq_ind_table
 			   (ind_tbl->ind_table));
 	for (i = 0; i != ind_tbl->queues_n; ++i)
-		claim_nonzero(mlx5_priv_rxq_release(priv, ind_tbl->queues[i]));
+		claim_nonzero(mlx5_rxq_release(dev, ind_tbl->queues[i]));
 	if (!rte_atomic32_read(&ind_tbl->refcnt)) {
 		LIST_REMOVE(ind_tbl, next);
 		rte_free(ind_tbl);
 		return 0;
 	}
-	return EBUSY;
+	return 1;
 }
 
 /**
  * Verify the Rx Queue list is empty
  *
- * @param priv
- *  Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  *
- * @return the number of object not released.
+ * @return
+ *   The number of object not released.
  */
 int
-mlx5_priv_ind_table_ibv_verify(struct priv *priv)
+mlx5_ind_table_ibv_verify(struct rte_eth_dev *dev)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_ind_table_ibv *ind_tbl;
 	int ret = 0;
 
 	LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
-		DEBUG("%p: Verbs indirection table %p still referenced",
-		      (void *)priv, (void *)ind_tbl);
+		DRV_LOG(DEBUG,
+			"port %u Verbs indirection table %p still referenced",
+			dev->data->port_id, (void *)ind_tbl);
 		++ret;
 	}
 	return ret;
@@ -1314,8 +1746,8 @@ mlx5_priv_ind_table_ibv_verify(struct priv *priv)
 /**
  * Create an Rx Hash queue.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param rss_key
  *   RSS key for the Rx hash queue.
  * @param rss_key_len
@@ -1329,22 +1761,60 @@ mlx5_priv_ind_table_ibv_verify(struct priv *priv)
  *   Number of queues.
  *
  * @return
- *   An hash Rx queue on success.
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
  */
-struct mlx5_hrxq*
-mlx5_priv_hrxq_new(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
-		   uint64_t hash_fields, uint16_t queues[], uint16_t queues_n)
+struct mlx5_hrxq *
+mlx5_hrxq_new(struct rte_eth_dev *dev,
+	      const uint8_t *rss_key, uint32_t rss_key_len,
+	      uint64_t hash_fields,
+	      const uint16_t *queues, uint32_t queues_n,
+	      int tunnel __rte_unused)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_hrxq *hrxq;
 	struct mlx5_ind_table_ibv *ind_tbl;
 	struct ibv_qp *qp;
+	int err;
 
 	queues_n = hash_fields ? queues_n : 1;
-	ind_tbl = mlx5_priv_ind_table_ibv_get(priv, queues, queues_n);
-	if (!ind_tbl)
-		ind_tbl = mlx5_priv_ind_table_ibv_new(priv, queues, queues_n);
+	ind_tbl = mlx5_ind_table_ibv_get(dev, queues, queues_n);
 	if (!ind_tbl)
+		ind_tbl = mlx5_ind_table_ibv_new(dev, queues, queues_n);
+	if (!ind_tbl) {
+		rte_errno = ENOMEM;
 		return NULL;
+	}
+	if (!rss_key_len) {
+		rss_key_len = MLX5_RSS_HASH_KEY_LEN;
+		rss_key = rss_hash_default_key;
+	}
+#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT
+	qp = mlx5_glue->dv_create_qp
+		(priv->ctx,
+		 &(struct ibv_qp_init_attr_ex){
+			.qp_type = IBV_QPT_RAW_PACKET,
+			.comp_mask =
+				IBV_QP_INIT_ATTR_PD |
+				IBV_QP_INIT_ATTR_IND_TABLE |
+				IBV_QP_INIT_ATTR_RX_HASH,
+			.rx_hash_conf = (struct ibv_rx_hash_conf){
+				.rx_hash_function = IBV_RX_HASH_FUNC_TOEPLITZ,
+				.rx_hash_key_len = rss_key_len ? rss_key_len :
+						   MLX5_RSS_HASH_KEY_LEN,
+				.rx_hash_key = rss_key ?
+					       (void *)(uintptr_t)rss_key :
+					       rss_hash_default_key,
+				.rx_hash_fields_mask = hash_fields,
+			},
+			.rwq_ind_tbl = ind_tbl->ind_table,
+			.pd = priv->pd,
+		 },
+		 &(struct mlx5dv_qp_init_attr){
+			.comp_mask = tunnel ?
+				MLX5DV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS : 0,
+			.create_flags = MLX5DV_QP_CREATE_TUNNEL_OFFLOADS,
+		 });
+#else
 	qp = mlx5_glue->create_qp_ex
 		(priv->ctx,
 		 &(struct ibv_qp_init_attr_ex){
@@ -1355,15 +1825,21 @@ mlx5_priv_hrxq_new(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
 				IBV_QP_INIT_ATTR_RX_HASH,
 			.rx_hash_conf = (struct ibv_rx_hash_conf){
 				.rx_hash_function = IBV_RX_HASH_FUNC_TOEPLITZ,
-				.rx_hash_key_len = rss_key_len,
-				.rx_hash_key = rss_key,
+				.rx_hash_key_len = rss_key_len ? rss_key_len :
+						   MLX5_RSS_HASH_KEY_LEN,
+				.rx_hash_key = rss_key ?
+					       (void *)(uintptr_t)rss_key :
+					       rss_hash_default_key,
 				.rx_hash_fields_mask = hash_fields,
 			},
 			.rwq_ind_tbl = ind_tbl->ind_table,
 			.pd = priv->pd,
 		 });
-	if (!qp)
+#endif
+	if (!qp) {
+		rte_errno = errno;
 		goto error;
+	}
 	hrxq = rte_calloc(__func__, 1, sizeof(*hrxq) + rss_key_len, 0);
 	if (!hrxq)
 		goto error;
@@ -1374,21 +1850,21 @@ mlx5_priv_hrxq_new(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
 	memcpy(hrxq->rss_key, rss_key, rss_key_len);
 	rte_atomic32_inc(&hrxq->refcnt);
 	LIST_INSERT_HEAD(&priv->hrxqs, hrxq, next);
-	DEBUG("%p: Hash Rx queue %p: refcnt %d", (void *)priv,
-	      (void *)hrxq, rte_atomic32_read(&hrxq->refcnt));
 	return hrxq;
 error:
-	mlx5_priv_ind_table_ibv_release(priv, ind_tbl);
+	err = rte_errno; /* Save rte_errno before cleanup. */
+	mlx5_ind_table_ibv_release(dev, ind_tbl);
 	if (qp)
 		claim_zero(mlx5_glue->destroy_qp(qp));
+	rte_errno = err; /* Restore rte_errno. */
 	return NULL;
 }
 
 /**
  * Get an Rx Hash queue.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param rss_conf
  *   RSS configuration for the Rx hash queue.
  * @param queues
@@ -1400,10 +1876,13 @@ error:
  * @return
  *   An hash Rx queue on success.
  */
-struct mlx5_hrxq*
-mlx5_priv_hrxq_get(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
-		   uint64_t hash_fields, uint16_t queues[], uint16_t queues_n)
+struct mlx5_hrxq *
+mlx5_hrxq_get(struct rte_eth_dev *dev,
+	      const uint8_t *rss_key, uint32_t rss_key_len,
+	      uint64_t hash_fields,
+	      const uint16_t *queues, uint32_t queues_n)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_hrxq *hrxq;
 
 	queues_n = hash_fields ? queues_n : 1;
@@ -1416,16 +1895,14 @@ mlx5_priv_hrxq_get(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
 			continue;
 		if (hrxq->hash_fields != hash_fields)
 			continue;
-		ind_tbl = mlx5_priv_ind_table_ibv_get(priv, queues, queues_n);
+		ind_tbl = mlx5_ind_table_ibv_get(dev, queues, queues_n);
 		if (!ind_tbl)
 			continue;
 		if (ind_tbl != hrxq->ind_table) {
-			mlx5_priv_ind_table_ibv_release(priv, ind_tbl);
+			mlx5_ind_table_ibv_release(dev, ind_tbl);
 			continue;
 		}
 		rte_atomic32_inc(&hrxq->refcnt);
-		DEBUG("%p: Hash Rx queue %p: refcnt %d", (void *)priv,
-		      (void *)hrxq, rte_atomic32_read(&hrxq->refcnt));
 		return hrxq;
 	}
 	return NULL;
@@ -1434,48 +1911,281 @@ mlx5_priv_hrxq_get(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
 /**
  * Release the hash Rx queue.
  *
- * @param priv
- *   Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  * @param hrxq
  *   Pointer to Hash Rx queue to release.
  *
  * @return
- *   0 on success, errno value on failure.
+ *   1 while a reference on it exists, 0 when freed.
  */
 int
-mlx5_priv_hrxq_release(struct priv *priv, struct mlx5_hrxq *hrxq)
+mlx5_hrxq_release(struct rte_eth_dev *dev, struct mlx5_hrxq *hrxq)
 {
-	DEBUG("%p: Hash Rx queue %p: refcnt %d", (void *)priv,
-	      (void *)hrxq, rte_atomic32_read(&hrxq->refcnt));
 	if (rte_atomic32_dec_and_test(&hrxq->refcnt)) {
 		claim_zero(mlx5_glue->destroy_qp(hrxq->qp));
-		mlx5_priv_ind_table_ibv_release(priv, hrxq->ind_table);
+		mlx5_ind_table_ibv_release(dev, hrxq->ind_table);
 		LIST_REMOVE(hrxq, next);
 		rte_free(hrxq);
 		return 0;
 	}
-	claim_nonzero(mlx5_priv_ind_table_ibv_release(priv, hrxq->ind_table));
-	return EBUSY;
+	claim_nonzero(mlx5_ind_table_ibv_release(dev, hrxq->ind_table));
+	return 1;
 }
 
 /**
  * Verify the Rx Queue list is empty
  *
- * @param priv
- *  Pointer to private structure.
+ * @param dev
+ *   Pointer to Ethernet device.
  *
- * @return the number of object not released.
+ * @return
+ *   The number of object not released.
  */
 int
-mlx5_priv_hrxq_ibv_verify(struct priv *priv)
+mlx5_hrxq_ibv_verify(struct rte_eth_dev *dev)
 {
+	struct priv *priv = dev->data->dev_private;
 	struct mlx5_hrxq *hrxq;
 	int ret = 0;
 
 	LIST_FOREACH(hrxq, &priv->hrxqs, next) {
-		DEBUG("%p: Verbs Hash Rx queue %p still referenced",
-		      (void *)priv, (void *)hrxq);
+		DRV_LOG(DEBUG,
+			"port %u Verbs hash Rx queue %p still referenced",
+			dev->data->port_id, (void *)hrxq);
 		++ret;
 	}
 	return ret;
 }
+
+/**
+ * Create a drop Rx queue Verbs object.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
+ */
+struct mlx5_rxq_ibv *
+mlx5_rxq_ibv_drop_new(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct ibv_cq *cq;
+	struct ibv_wq *wq = NULL;
+	struct mlx5_rxq_ibv *rxq;
+
+	if (priv->drop_queue.rxq)
+		return priv->drop_queue.rxq;
+	cq = mlx5_glue->create_cq(priv->ctx, 1, NULL, NULL, 0);
+	if (!cq) {
+		DEBUG("port %u cannot allocate CQ for drop queue",
+		      dev->data->port_id);
+		rte_errno = errno;
+		goto error;
+	}
+	wq = mlx5_glue->create_wq(priv->ctx,
+		 &(struct ibv_wq_init_attr){
+			.wq_type = IBV_WQT_RQ,
+			.max_wr = 1,
+			.max_sge = 1,
+			.pd = priv->pd,
+			.cq = cq,
+		 });
+	if (!wq) {
+		DEBUG("port %u cannot allocate WQ for drop queue",
+		      dev->data->port_id);
+		rte_errno = errno;
+		goto error;
+	}
+	rxq = rte_calloc(__func__, 1, sizeof(*rxq), 0);
+	if (!rxq) {
+		DEBUG("port %u cannot allocate drop Rx queue memory",
+		      dev->data->port_id);
+		rte_errno = ENOMEM;
+		goto error;
+	}
+	rxq->cq = cq;
+	rxq->wq = wq;
+	priv->drop_queue.rxq = rxq;
+	return rxq;
+error:
+	if (wq)
+		claim_zero(mlx5_glue->destroy_wq(wq));
+	if (cq)
+		claim_zero(mlx5_glue->destroy_cq(cq));
+	return NULL;
+}
+
+/**
+ * Release a drop Rx queue Verbs object.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
+ */
+void
+mlx5_rxq_ibv_drop_release(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct mlx5_rxq_ibv *rxq = priv->drop_queue.rxq;
+
+	if (rxq->wq)
+		claim_zero(mlx5_glue->destroy_wq(rxq->wq));
+	if (rxq->cq)
+		claim_zero(mlx5_glue->destroy_cq(rxq->cq));
+	rte_free(rxq);
+	priv->drop_queue.rxq = NULL;
+}
+
+/**
+ * Create a drop indirection table.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
+ */
+struct mlx5_ind_table_ibv *
+mlx5_ind_table_ibv_drop_new(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct mlx5_ind_table_ibv *ind_tbl;
+	struct mlx5_rxq_ibv *rxq;
+	struct mlx5_ind_table_ibv tmpl;
+
+	rxq = mlx5_rxq_ibv_drop_new(dev);
+	if (!rxq)
+		return NULL;
+	tmpl.ind_table = mlx5_glue->create_rwq_ind_table
+		(priv->ctx,
+		 &(struct ibv_rwq_ind_table_init_attr){
+			.log_ind_tbl_size = 0,
+			.ind_tbl = &rxq->wq,
+			.comp_mask = 0,
+		 });
+	if (!tmpl.ind_table) {
+		DEBUG("port %u cannot allocate indirection table for drop"
+		      " queue",
+		      dev->data->port_id);
+		rte_errno = errno;
+		goto error;
+	}
+	ind_tbl = rte_calloc(__func__, 1, sizeof(*ind_tbl), 0);
+	if (!ind_tbl) {
+		rte_errno = ENOMEM;
+		goto error;
+	}
+	ind_tbl->ind_table = tmpl.ind_table;
+	return ind_tbl;
+error:
+	mlx5_rxq_ibv_drop_release(dev);
+	return NULL;
+}
+
+/**
+ * Release a drop indirection table.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ */
+void
+mlx5_ind_table_ibv_drop_release(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct mlx5_ind_table_ibv *ind_tbl = priv->drop_queue.hrxq->ind_table;
+
+	claim_zero(mlx5_glue->destroy_rwq_ind_table(ind_tbl->ind_table));
+	mlx5_rxq_ibv_drop_release(dev);
+	rte_free(ind_tbl);
+	priv->drop_queue.hrxq->ind_table = NULL;
+}
+
+/**
+ * Create a drop Rx Hash queue.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   The Verbs object initialised, NULL otherwise and rte_errno is set.
+ */
+struct mlx5_hrxq *
+mlx5_hrxq_drop_new(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct mlx5_ind_table_ibv *ind_tbl;
+	struct ibv_qp *qp;
+	struct mlx5_hrxq *hrxq;
+
+	if (priv->drop_queue.hrxq) {
+		rte_atomic32_inc(&priv->drop_queue.hrxq->refcnt);
+		return priv->drop_queue.hrxq;
+	}
+	ind_tbl = mlx5_ind_table_ibv_drop_new(dev);
+	if (!ind_tbl)
+		return NULL;
+	qp = mlx5_glue->create_qp_ex(priv->ctx,
+		 &(struct ibv_qp_init_attr_ex){
+			.qp_type = IBV_QPT_RAW_PACKET,
+			.comp_mask =
+				IBV_QP_INIT_ATTR_PD |
+				IBV_QP_INIT_ATTR_IND_TABLE |
+				IBV_QP_INIT_ATTR_RX_HASH,
+			.rx_hash_conf = (struct ibv_rx_hash_conf){
+				.rx_hash_function =
+					IBV_RX_HASH_FUNC_TOEPLITZ,
+				.rx_hash_key_len = MLX5_RSS_HASH_KEY_LEN,
+				.rx_hash_key = rss_hash_default_key,
+				.rx_hash_fields_mask = 0,
+				},
+			.rwq_ind_tbl = ind_tbl->ind_table,
+			.pd = priv->pd
+		 });
+	if (!qp) {
+		DEBUG("port %u cannot allocate QP for drop queue",
+		      dev->data->port_id);
+		rte_errno = errno;
+		goto error;
+	}
+	hrxq = rte_calloc(__func__, 1, sizeof(*hrxq), 0);
+	if (!hrxq) {
+		DRV_LOG(WARNING,
+			"port %u cannot allocate memory for drop queue",
+			dev->data->port_id);
+		rte_errno = ENOMEM;
+		goto error;
+	}
+	hrxq->ind_table = ind_tbl;
+	hrxq->qp = qp;
+	priv->drop_queue.hrxq = hrxq;
+	rte_atomic32_set(&hrxq->refcnt, 1);
+	return hrxq;
+error:
+	if (ind_tbl)
+		mlx5_ind_table_ibv_drop_release(dev);
+	return NULL;
+}
+
+/**
+ * Release a drop hash Rx queue.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ */
+void
+mlx5_hrxq_drop_release(struct rte_eth_dev *dev)
+{
+	struct priv *priv = dev->data->dev_private;
+	struct mlx5_hrxq *hrxq = priv->drop_queue.hrxq;
+
+	if (rte_atomic32_dec_and_test(&hrxq->refcnt)) {
+		claim_zero(mlx5_glue->destroy_qp(hrxq->qp));
+		mlx5_ind_table_ibv_drop_release(dev);
+		rte_free(hrxq);
+		priv->drop_queue.hrxq = NULL;
+	}
+}