+/**
+ * Callback function to initialize mbufs for Multi-Packet RQ.
+ */
+static inline void
+mlx5_mprq_buf_init(struct rte_mempool *mp, void *opaque_arg,
+ void *_m, unsigned int i __rte_unused)
+{
+ struct mlx5_mprq_buf *buf = _m;
+ struct rte_mbuf_ext_shared_info *shinfo;
+ unsigned int strd_n = (unsigned int)(uintptr_t)opaque_arg;
+ unsigned int j;
+
+ memset(_m, 0, sizeof(*buf));
+ buf->mp = mp;
+ rte_atomic16_set(&buf->refcnt, 1);
+ for (j = 0; j != strd_n; ++j) {
+ shinfo = &buf->shinfos[j];
+ shinfo->free_cb = mlx5_mprq_buf_free_cb;
+ shinfo->fcb_opaque = buf;
+ }
+}
+
+/**
+ * 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 mlx5_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 queue 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;
+ }
+ priv->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 mlx5_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;
+ unsigned int n_ibv = 0;
+
+ 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];
+ struct mlx5_rxq_ctrl *rxq_ctrl = container_of
+ (rxq, struct mlx5_rxq_ctrl, rxq);
+
+ if (rxq == NULL || rxq_ctrl->type != MLX5_RXQ_TYPE_STANDARD)
+ continue;
+ n_ibv++;
+ 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;
+ }
+ MLX5_ASSERT(strd_num_n && strd_sz_n);
+ buf_len = (1 << strd_num_n) * (1 << strd_sz_n);
+ obj_size = sizeof(struct mlx5_mprq_buf) + buf_len + (1 << strd_num_n) *
+ sizeof(struct rte_mbuf_ext_shared_info) + RTE_PKTMBUF_HEADROOM;
+ /*
+ * 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 * n_ibv;
+ /*
+ * 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), "port-%u-mprq", dev->data->port_id);
+ mp = rte_mempool_create(name, obj_num, obj_size, MLX5_MPRQ_MP_CACHE_SZ,
+ 0, NULL, NULL, mlx5_mprq_buf_init,
+ (void *)(uintptr_t)(1 << strd_num_n),
+ 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];
+ struct mlx5_rxq_ctrl *rxq_ctrl = container_of
+ (rxq, struct mlx5_rxq_ctrl, rxq);
+
+ if (rxq == NULL || rxq_ctrl->type != MLX5_RXQ_TYPE_STANDARD)
+ continue;
+ rxq->mprq_mp = mp;
+ }
+ DRV_LOG(INFO, "port %u Multi-Packet RQ is configured",
+ dev->data->port_id);
+ return 0;
+}
+
+#define MLX5_MAX_TCP_HDR_OFFSET ((unsigned int)(sizeof(struct rte_ether_hdr) + \
+ sizeof(struct rte_vlan_hdr) * 2 + \
+ sizeof(struct rte_ipv6_hdr)))
+#define MAX_TCP_OPTION_SIZE 40u
+#define MLX5_MAX_LRO_HEADER_FIX ((unsigned int)(MLX5_MAX_TCP_HDR_OFFSET + \
+ sizeof(struct rte_tcp_hdr) + \
+ MAX_TCP_OPTION_SIZE))
+
+/**
+ * Adjust the maximum LRO massage size.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
+ * @param max_lro_size
+ * The maximum size for LRO packet.
+ */
+static void
+mlx5_max_lro_msg_size_adjust(struct rte_eth_dev *dev, uint16_t idx,
+ uint32_t max_lro_size)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+
+ if (priv->config.hca_attr.lro_max_msg_sz_mode ==
+ MLX5_LRO_MAX_MSG_SIZE_START_FROM_L4 && max_lro_size >
+ MLX5_MAX_TCP_HDR_OFFSET)
+ max_lro_size -= MLX5_MAX_TCP_HDR_OFFSET;
+ max_lro_size = RTE_MIN(max_lro_size, MLX5_MAX_LRO_SIZE);
+ MLX5_ASSERT(max_lro_size >= MLX5_LRO_SEG_CHUNK_SIZE);
+ max_lro_size /= MLX5_LRO_SEG_CHUNK_SIZE;
+ if (priv->max_lro_msg_size)
+ priv->max_lro_msg_size =
+ RTE_MIN((uint32_t)priv->max_lro_msg_size, max_lro_size);
+ else
+ priv->max_lro_msg_size = max_lro_size;
+ DRV_LOG(DEBUG,
+ "port %u Rx Queue %u max LRO message size adjusted to %u bytes",
+ dev->data->port_id, idx,
+ priv->max_lro_msg_size * MLX5_LRO_SEG_CHUNK_SIZE);
+}
+
+/**
+ * Create a DPDK Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * 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, NULL otherwise and rte_errno is set.
+ */
+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 mlx5_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;
+ unsigned int strd_headroom_en;
+ /*
+ * Always allocate extra slots, even if eventually
+ * the vector Rx will not be used.
+ */
+ 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;
+ unsigned int lro_on_queue = !!(offloads & DEV_RX_OFFLOAD_TCP_LRO);
+ const int mprq_en = mlx5_check_mprq_support(dev) > 0;
+ unsigned int max_rx_pkt_len = lro_on_queue ?
+ dev->data->dev_conf.rxmode.max_lro_pkt_size :
+ dev->data->dev_conf.rxmode.max_rx_pkt_len;
+ unsigned int non_scatter_min_mbuf_size = max_rx_pkt_len +
+ RTE_PKTMBUF_HEADROOM;
+ unsigned int max_lro_size = 0;
+ unsigned int first_mb_free_size = mb_len - RTE_PKTMBUF_HEADROOM;
+
+ if (non_scatter_min_mbuf_size > mb_len && !(offloads &
+ DEV_RX_OFFLOAD_SCATTER)) {
+ DRV_LOG(ERR, "port %u Rx queue %u: Scatter offload is not"
+ " configured and no enough mbuf space(%u) to contain "
+ "the maximum RX packet length(%u) with head-room(%u)",
+ dev->data->port_id, idx, mb_len, max_rx_pkt_len,
+ RTE_PKTMBUF_HEADROOM);
+ rte_errno = ENOSPC;
+ return NULL;
+ }
+ tmpl = rte_calloc_socket("RXQ", 1,
+ sizeof(*tmpl) +
+ desc_n * sizeof(struct rte_mbuf *),
+ 0, socket);
+ if (!tmpl) {
+ rte_errno = ENOMEM;
+ return NULL;
+ }
+ tmpl->type = MLX5_RXQ_TYPE_STANDARD;
+ 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 (dev->data->dev_conf.intr_conf.rxq)
+ tmpl->irq = 1;
+ /*
+ * LRO packet may consume all the stride memory, hence we cannot
+ * guaranty head-room near the packet memory in the stride.
+ * In this case scatter is, for sure, enabled and an empty mbuf may be
+ * added in the start for the head-room.
+ */
+ if (lro_on_queue && RTE_PKTMBUF_HEADROOM > 0 &&
+ non_scatter_min_mbuf_size > mb_len) {
+ strd_headroom_en = 0;
+ mprq_stride_size = RTE_MIN(max_rx_pkt_len,
+ 1u << config->mprq.max_stride_size_n);
+ } else {
+ strd_headroom_en = 1;
+ mprq_stride_size = non_scatter_min_mbuf_size;
+ }
+ /*
+ * 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.
+ */
+ 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.strd_headroom_en = strd_headroom_en;
+ tmpl->rxq.mprq_max_memcpy_len = RTE_MIN(first_mb_free_size,
+ config->mprq.max_memcpy_len);
+ max_lro_size = RTE_MIN(max_rx_pkt_len,
+ (1u << tmpl->rxq.strd_num_n) *
+ (1u << tmpl->rxq.strd_sz_n));
+ 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 (max_rx_pkt_len <= first_mb_free_size) {
+ tmpl->rxq.sges_n = 0;
+ max_lro_size = max_rx_pkt_len;
+ } else if (offloads & DEV_RX_OFFLOAD_SCATTER) {
+ unsigned int size = non_scatter_min_mbuf_size;
+ unsigned int sges_n;
+
+ if (lro_on_queue && first_mb_free_size <
+ MLX5_MAX_LRO_HEADER_FIX) {
+ DRV_LOG(ERR, "Not enough space in the first segment(%u)"
+ " to include the max header size(%u) for LRO",
+ first_mb_free_size, MLX5_MAX_LRO_HEADER_FIX);
+ rte_errno = ENOTSUP;
+ goto error;
+ }
+ /*
+ * Determine the number of SGEs needed for a full packet
+ * and round it to the next power of two.
+ */
+ sges_n = log2above((size / mb_len) + !!(size % mb_len));
+ if (sges_n > MLX5_MAX_LOG_RQ_SEGS) {
+ DRV_LOG(ERR,
+ "port %u too many SGEs (%u) needed to handle"
+ " requested maximum packet size %u, the maximum"
+ " supported are %u", dev->data->port_id,
+ 1 << sges_n, max_rx_pkt_len,
+ 1u << MLX5_MAX_LOG_RQ_SEGS);
+ rte_errno = ENOTSUP;
+ goto error;
+ }
+ tmpl->rxq.sges_n = sges_n;
+ max_lro_size = max_rx_pkt_len;
+ }
+ 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)) {
+ 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;
+ }
+ mlx5_max_lro_msg_size_adjust(dev, idx, max_lro_size);
+ /* Toggle RX checksum offload if hardware supports it. */
+ tmpl->rxq.csum = !!(offloads & DEV_RX_OFFLOAD_CHECKSUM);
+ tmpl->rxq.hw_timestamp = !!(offloads & DEV_RX_OFFLOAD_TIMESTAMP);
+ /* Configure VLAN stripping. */
+ tmpl->rxq.vlan_strip = !!(offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
+ /* By default, FCS (CRC) is stripped by hardware. */
+ tmpl->rxq.crc_present = 0;
+ tmpl->rxq.lro = lro_on_queue;
+ if (offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
+ if (config->hw_fcs_strip) {
+ /*
+ * RQs used for LRO-enabled TIRs should not be
+ * configured to scatter the FCS.
+ */
+ if (lro_on_queue)
+ DRV_LOG(WARNING,
+ "port %u CRC stripping has been "
+ "disabled but will still be performed "
+ "by hardware, because LRO is enabled",
+ dev->data->port_id);
+ else
+ 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->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;
+ tmpl->rxq.elts_n = log2above(desc);
+ tmpl->rxq.rq_repl_thresh =
+ MLX5_VPMD_RXQ_RPLNSH_THRESH(1 << tmpl->rxq.elts_n);
+ 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->rxq.idx = idx;
+ rte_atomic32_inc(&tmpl->refcnt);
+ LIST_INSERT_HEAD(&priv->rxqsctrl, tmpl, next);
+ return tmpl;
+error:
+ rte_free(tmpl);
+ return NULL;
+}
+
+/**
+ * Create a DPDK Rx hairpin queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param hairpin_conf
+ * The hairpin binding configuration.
+ *
+ * @return
+ * A DPDK queue object on success, NULL otherwise and rte_errno is set.
+ */
+struct mlx5_rxq_ctrl *
+mlx5_rxq_hairpin_new(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+ const struct rte_eth_hairpin_conf *hairpin_conf)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *tmpl;
+
+ tmpl = rte_calloc_socket("RXQ", 1, sizeof(*tmpl), 0, SOCKET_ID_ANY);
+ if (!tmpl) {
+ rte_errno = ENOMEM;
+ return NULL;
+ }
+ tmpl->type = MLX5_RXQ_TYPE_HAIRPIN;
+ tmpl->socket = SOCKET_ID_ANY;
+ tmpl->rxq.rss_hash = 0;
+ tmpl->rxq.port_id = dev->data->port_id;
+ tmpl->priv = priv;
+ tmpl->rxq.mp = NULL;
+ tmpl->rxq.elts_n = log2above(desc);
+ tmpl->rxq.elts = NULL;
+ tmpl->rxq.mr_ctrl.cache_bh = (struct mlx5_mr_btree) { 0 };
+ tmpl->hairpin_conf = *hairpin_conf;
+ tmpl->rxq.idx = idx;
+ rte_atomic32_inc(&tmpl->refcnt);
+ LIST_INSERT_HEAD(&priv->rxqsctrl, tmpl, next);
+ return tmpl;
+}
+
+/**
+ * Get a Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * A pointer to the queue if it exists, NULL otherwise.
+ */
+struct mlx5_rxq_ctrl *
+mlx5_rxq_get(struct rte_eth_dev *dev, uint16_t idx)
+{
+ struct mlx5_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_rxq_obj_get(dev, idx);
+ rte_atomic32_inc(&rxq_ctrl->refcnt);
+ }
+ return rxq_ctrl;
+}
+
+/**
+ * Release a Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * 1 while a reference on it exists, 0 when freed.
+ */
+int
+mlx5_rxq_release(struct rte_eth_dev *dev, uint16_t idx)
+{
+ struct mlx5_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);
+ MLX5_ASSERT(rxq_ctrl->priv);
+ if (rxq_ctrl->obj && !mlx5_rxq_obj_release(rxq_ctrl->obj))
+ rxq_ctrl->obj = NULL;
+ if (rte_atomic32_dec_and_test(&rxq_ctrl->refcnt)) {
+ if (rxq_ctrl->dbr_umem_id_valid)
+ claim_zero(mlx5_release_dbr(dev, rxq_ctrl->dbr_umem_id,
+ rxq_ctrl->dbr_offset));
+ if (rxq_ctrl->type == MLX5_RXQ_TYPE_STANDARD)
+ 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 1;
+}
+
+/**
+ * Verify the Rx Queue list is empty
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The number of object not released.
+ */
+int
+mlx5_rxq_verify(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
+ int ret = 0;
+
+ LIST_FOREACH(rxq_ctrl, &priv->rxqsctrl, next) {
+ DRV_LOG(DEBUG, "port %u Rx Queue %u still referenced",
+ dev->data->port_id, rxq_ctrl->rxq.idx);
+ ++ret;
+ }
+ return ret;
+}
+
+/**
+ * Get a Rx queue type.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * Rx queue index.
+ *
+ * @return
+ * The Rx queue type.
+ */
+enum mlx5_rxq_type
+mlx5_rxq_get_type(struct rte_eth_dev *dev, uint16_t idx)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl = NULL;
+
+ if (idx < priv->rxqs_n && (*priv->rxqs)[idx]) {
+ rxq_ctrl = container_of((*priv->rxqs)[idx],
+ struct mlx5_rxq_ctrl,
+ rxq);
+ return rxq_ctrl->type;
+ }
+ return MLX5_RXQ_TYPE_UNDEFINED;
+}
+
+/**
+ * Create an indirection table.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param queues
+ * Queues entering in the indirection table.
+ * @param queues_n
+ * Number of queues in the array.
+ *
+ * @return
+ * The Verbs/DevX object initialised, NULL otherwise and rte_errno is set.
+ */
+static struct mlx5_ind_table_obj *
+mlx5_ind_table_obj_new(struct rte_eth_dev *dev, const uint16_t *queues,
+ uint32_t queues_n, enum mlx5_ind_tbl_type type)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
+ unsigned int i = 0, j = 0, k = 0;
+
+ ind_tbl = rte_calloc(__func__, 1, sizeof(*ind_tbl) +
+ queues_n * sizeof(uint16_t), 0);
+ if (!ind_tbl) {
+ rte_errno = ENOMEM;
+ return NULL;
+ }
+ ind_tbl->type = type;
+ if (ind_tbl->type == MLX5_IND_TBL_TYPE_IBV) {
+ const unsigned int wq_n = rte_is_power_of_2(queues_n) ?
+ log2above(queues_n) :
+ log2above(priv->config.ind_table_max_size);
+ struct ibv_wq *wq[1 << wq_n];
+
+ for (i = 0; i != queues_n; ++i) {
+ struct mlx5_rxq_ctrl *rxq = mlx5_rxq_get(dev,
+ queues[i]);
+ if (!rxq)
+ goto error;
+ wq[i] = rxq->obj->wq;
+ ind_tbl->queues[i] = queues[i];
+ }
+ ind_tbl->queues_n = queues_n;
+ /* Finalise indirection table. */
+ k = i; /* Retain value of i for use in error case. */
+ for (j = 0; k != (unsigned int)(1 << wq_n); ++k, ++j)
+ wq[k] = wq[j];
+ ind_tbl->ind_table = mlx5_glue->create_rwq_ind_table
+ (priv->sh->ctx,
+ &(struct ibv_rwq_ind_table_init_attr){
+ .log_ind_tbl_size = wq_n,
+ .ind_tbl = wq,
+ .comp_mask = 0,
+ });
+ if (!ind_tbl->ind_table) {
+ rte_errno = errno;
+ goto error;
+ }
+ } else { /* ind_tbl->type == MLX5_IND_TBL_TYPE_DEVX */
+ struct mlx5_devx_rqt_attr *rqt_attr = NULL;
+ const unsigned int rqt_n =
+ 1 << (rte_is_power_of_2(queues_n) ?
+ log2above(queues_n) :
+ log2above(priv->config.ind_table_max_size));
+
+ rqt_attr = rte_calloc(__func__, 1, sizeof(*rqt_attr) +
+ rqt_n * sizeof(uint32_t), 0);
+ if (!rqt_attr) {
+ DRV_LOG(ERR, "port %u cannot allocate RQT resources",
+ dev->data->port_id);
+ rte_errno = ENOMEM;
+ goto error;
+ }
+ rqt_attr->rqt_max_size = priv->config.ind_table_max_size;
+ rqt_attr->rqt_actual_size = rqt_n;
+ for (i = 0; i != queues_n; ++i) {
+ struct mlx5_rxq_ctrl *rxq = mlx5_rxq_get(dev,
+ queues[i]);
+ if (!rxq)
+ goto error;
+ rqt_attr->rq_list[i] = rxq->obj->rq->id;
+ ind_tbl->queues[i] = queues[i];
+ }
+ k = i; /* Retain value of i for use in error case. */
+ for (j = 0; k != rqt_n; ++k, ++j)
+ rqt_attr->rq_list[k] = rqt_attr->rq_list[j];
+ ind_tbl->rqt = mlx5_devx_cmd_create_rqt(priv->sh->ctx,
+ rqt_attr);
+ rte_free(rqt_attr);
+ if (!ind_tbl->rqt) {
+ DRV_LOG(ERR, "port %u cannot create DevX RQT",
+ dev->data->port_id);
+ rte_errno = errno;
+ goto error;
+ }
+ ind_tbl->queues_n = queues_n;
+ }
+ rte_atomic32_inc(&ind_tbl->refcnt);
+ LIST_INSERT_HEAD(&priv->ind_tbls, ind_tbl, next);
+ return ind_tbl;
+error:
+ for (j = 0; j < i; j++)
+ mlx5_rxq_release(dev, ind_tbl->queues[j]);
+ rte_free(ind_tbl);
+ DEBUG("port %u cannot create indirection table", dev->data->port_id);
+ return NULL;
+}
+
+/**
+ * Get an indirection table.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param queues
+ * Queues entering in the indirection table.
+ * @param queues_n
+ * Number of queues in the array.
+ *
+ * @return
+ * An indirection table if found.
+ */
+static struct mlx5_ind_table_obj *
+mlx5_ind_table_obj_get(struct rte_eth_dev *dev, const uint16_t *queues,
+ uint32_t queues_n)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
+
+ LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
+ if ((ind_tbl->queues_n == queues_n) &&
+ (memcmp(ind_tbl->queues, queues,
+ ind_tbl->queues_n * sizeof(ind_tbl->queues[0]))
+ == 0))
+ break;
+ }
+ if (ind_tbl) {
+ unsigned int i;
+
+ rte_atomic32_inc(&ind_tbl->refcnt);
+ for (i = 0; i != ind_tbl->queues_n; ++i)
+ mlx5_rxq_get(dev, ind_tbl->queues[i]);
+ }
+ return ind_tbl;
+}
+
+/**
+ * Release an indirection table.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param ind_table
+ * Indirection table to release.
+ *
+ * @return
+ * 1 while a reference on it exists, 0 when freed.
+ */
+static int
+mlx5_ind_table_obj_release(struct rte_eth_dev *dev,
+ struct mlx5_ind_table_obj *ind_tbl)
+{
+ unsigned int i;
+
+ if (rte_atomic32_dec_and_test(&ind_tbl->refcnt)) {
+ if (ind_tbl->type == MLX5_IND_TBL_TYPE_IBV)
+ claim_zero(mlx5_glue->destroy_rwq_ind_table
+ (ind_tbl->ind_table));
+ else if (ind_tbl->type == MLX5_IND_TBL_TYPE_DEVX)
+ claim_zero(mlx5_devx_cmd_destroy(ind_tbl->rqt));
+ }
+ for (i = 0; i != ind_tbl->queues_n; ++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 1;
+}
+
+/**
+ * Verify the Rx Queue list is empty
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The number of object not released.
+ */
+int
+mlx5_ind_table_obj_verify(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
+ int ret = 0;
+
+ LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
+ DRV_LOG(DEBUG,
+ "port %u indirection table obj %p still referenced",
+ dev->data->port_id, (void *)ind_tbl);
+ ++ret;
+ }
+ return ret;
+}
+
+/**
+ * Create an Rx Hash queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param rss_key
+ * RSS key for the Rx hash queue.
+ * @param rss_key_len
+ * RSS key length.
+ * @param hash_fields
+ * Verbs protocol hash field to make the RSS on.
+ * @param queues
+ * Queues entering in hash queue. In case of empty hash_fields only the
+ * first queue index will be taken for the indirection table.
+ * @param queues_n
+ * Number of queues.
+ * @param tunnel
+ * Tunnel type.
+ *
+ * @return
+ * The Verbs/DevX object initialised, NULL otherwise and rte_errno is set.
+ */
+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 mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq;
+ struct ibv_qp *qp = NULL;
+ struct mlx5_ind_table_obj *ind_tbl;
+ int err;
+ struct mlx5_devx_obj *tir = NULL;
+ struct mlx5_rxq_data *rxq_data = (*priv->rxqs)[queues[0]];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
+
+ queues_n = hash_fields ? queues_n : 1;
+ ind_tbl = mlx5_ind_table_obj_get(dev, queues, queues_n);
+ if (!ind_tbl) {
+ enum mlx5_ind_tbl_type type;
+
+ type = rxq_ctrl->obj->type == MLX5_RXQ_OBJ_TYPE_IBV ?
+ MLX5_IND_TBL_TYPE_IBV : MLX5_IND_TBL_TYPE_DEVX;
+ ind_tbl = mlx5_ind_table_obj_new(dev, queues, queues_n, type);
+ }
+ if (!ind_tbl) {
+ rte_errno = ENOMEM;
+ return NULL;
+ }
+ if (ind_tbl->type == MLX5_IND_TBL_TYPE_IBV) {
+#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT
+ struct mlx5dv_qp_init_attr qp_init_attr;
+
+ memset(&qp_init_attr, 0, sizeof(qp_init_attr));
+ if (tunnel) {
+ qp_init_attr.comp_mask =
+ MLX5DV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS;
+ qp_init_attr.create_flags =
+ MLX5DV_QP_CREATE_TUNNEL_OFFLOADS;
+ }
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ if (dev->data->dev_conf.lpbk_mode) {
+ /*
+ * Allow packet sent from NIC loop back
+ * w/o source MAC check.
+ */
+ qp_init_attr.comp_mask |=
+ MLX5DV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS;
+ qp_init_attr.create_flags |=
+ MLX5DV_QP_CREATE_TIR_ALLOW_SELF_LOOPBACK_UC;
+ }
+#endif
+ qp = mlx5_glue->dv_create_qp
+ (priv->sh->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,
+ .rx_hash_key =
+ (void *)(uintptr_t)rss_key,
+ .rx_hash_fields_mask = hash_fields,
+ },
+ .rwq_ind_tbl = ind_tbl->ind_table,
+ .pd = priv->sh->pd,
+ },
+ &qp_init_attr);
+#else
+ qp = mlx5_glue->create_qp_ex
+ (priv->sh->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,
+ .rx_hash_key =
+ (void *)(uintptr_t)rss_key,
+ .rx_hash_fields_mask = hash_fields,
+ },
+ .rwq_ind_tbl = ind_tbl->ind_table,
+ .pd = priv->sh->pd,
+ });
+#endif
+ if (!qp) {
+ rte_errno = errno;
+ goto error;
+ }
+ } else { /* ind_tbl->type == MLX5_IND_TBL_TYPE_DEVX */
+ struct mlx5_devx_tir_attr tir_attr;
+ uint32_t i;
+ uint32_t lro = 1;
+
+ /* Enable TIR LRO only if all the queues were configured for. */
+ for (i = 0; i < queues_n; ++i) {
+ if (!(*priv->rxqs)[queues[i]]->lro) {
+ lro = 0;
+ break;
+ }
+ }
+ memset(&tir_attr, 0, sizeof(tir_attr));
+ tir_attr.disp_type = MLX5_TIRC_DISP_TYPE_INDIRECT;
+ tir_attr.rx_hash_fn = MLX5_RX_HASH_FN_TOEPLITZ;
+ tir_attr.tunneled_offload_en = !!tunnel;
+ /* If needed, translate hash_fields bitmap to PRM format. */
+ if (hash_fields) {
+#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT
+ struct mlx5_rx_hash_field_select *rx_hash_field_select =
+ hash_fields & IBV_RX_HASH_INNER ?
+ &tir_attr.rx_hash_field_selector_inner :
+ &tir_attr.rx_hash_field_selector_outer;
+#else
+ struct mlx5_rx_hash_field_select *rx_hash_field_select =
+ &tir_attr.rx_hash_field_selector_outer;
+#endif
+
+ /* 1 bit: 0: IPv4, 1: IPv6. */
+ rx_hash_field_select->l3_prot_type =
+ !!(hash_fields & MLX5_IPV6_IBV_RX_HASH);
+ /* 1 bit: 0: TCP, 1: UDP. */
+ rx_hash_field_select->l4_prot_type =
+ !!(hash_fields & MLX5_UDP_IBV_RX_HASH);
+ /* Bitmask which sets which fields to use in RX Hash. */
+ rx_hash_field_select->selected_fields =
+ ((!!(hash_fields & MLX5_L3_SRC_IBV_RX_HASH)) <<
+ MLX5_RX_HASH_FIELD_SELECT_SELECTED_FIELDS_SRC_IP) |
+ (!!(hash_fields & MLX5_L3_DST_IBV_RX_HASH)) <<
+ MLX5_RX_HASH_FIELD_SELECT_SELECTED_FIELDS_DST_IP |
+ (!!(hash_fields & MLX5_L4_SRC_IBV_RX_HASH)) <<
+ MLX5_RX_HASH_FIELD_SELECT_SELECTED_FIELDS_L4_SPORT |
+ (!!(hash_fields & MLX5_L4_DST_IBV_RX_HASH)) <<
+ MLX5_RX_HASH_FIELD_SELECT_SELECTED_FIELDS_L4_DPORT;
+ }
+ if (rxq_ctrl->obj->type == MLX5_RXQ_OBJ_TYPE_DEVX_HAIRPIN)
+ tir_attr.transport_domain = priv->sh->td->id;
+ else
+ tir_attr.transport_domain = priv->sh->tdn;
+ memcpy(tir_attr.rx_hash_toeplitz_key, rss_key, rss_key_len);
+ tir_attr.indirect_table = ind_tbl->rqt->id;
+ if (dev->data->dev_conf.lpbk_mode)
+ tir_attr.self_lb_block =
+ MLX5_TIRC_SELF_LB_BLOCK_BLOCK_UNICAST;
+ if (lro) {
+ tir_attr.lro_timeout_period_usecs =
+ priv->config.lro.timeout;
+ tir_attr.lro_max_msg_sz = priv->max_lro_msg_size;
+ tir_attr.lro_enable_mask =
+ MLX5_TIRC_LRO_ENABLE_MASK_IPV4_LRO |
+ MLX5_TIRC_LRO_ENABLE_MASK_IPV6_LRO;
+ }
+ tir = mlx5_devx_cmd_create_tir(priv->sh->ctx, &tir_attr);
+ if (!tir) {
+ DRV_LOG(ERR, "port %u cannot create DevX TIR",
+ dev->data->port_id);
+ rte_errno = errno;
+ goto error;
+ }
+ }
+ hrxq = rte_calloc(__func__, 1, sizeof(*hrxq) + rss_key_len, 0);
+ if (!hrxq)
+ goto error;
+ hrxq->ind_table = ind_tbl;
+ if (ind_tbl->type == MLX5_IND_TBL_TYPE_IBV) {
+ hrxq->qp = qp;
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ hrxq->action =
+ mlx5_glue->dv_create_flow_action_dest_ibv_qp(hrxq->qp);
+ if (!hrxq->action) {
+ rte_errno = errno;
+ goto error;
+ }
+#endif
+ } else { /* ind_tbl->type == MLX5_IND_TBL_TYPE_DEVX */
+ hrxq->tir = tir;
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ hrxq->action = mlx5_glue->dv_create_flow_action_dest_devx_tir
+ (hrxq->tir->obj);
+ if (!hrxq->action) {
+ rte_errno = errno;
+ goto error;
+ }
+#endif
+ }
+ hrxq->rss_key_len = rss_key_len;
+ hrxq->hash_fields = hash_fields;
+ memcpy(hrxq->rss_key, rss_key, rss_key_len);
+ rte_atomic32_inc(&hrxq->refcnt);
+ LIST_INSERT_HEAD(&priv->hrxqs, hrxq, next);
+ return hrxq;
+error:
+ err = rte_errno; /* Save rte_errno before cleanup. */
+ mlx5_ind_table_obj_release(dev, ind_tbl);
+ if (qp)
+ claim_zero(mlx5_glue->destroy_qp(qp));
+ else if (tir)
+ claim_zero(mlx5_devx_cmd_destroy(tir));
+ rte_errno = err; /* Restore rte_errno. */
+ return NULL;
+}
+
+/**
+ * Get an Rx Hash queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param rss_conf
+ * RSS configuration for the Rx hash queue.
+ * @param queues
+ * Queues entering in hash queue. In case of empty hash_fields only the
+ * first queue index will be taken for the indirection table.
+ * @param queues_n
+ * Number of queues.
+ *
+ * @return
+ * An hash Rx queue on success.
+ */
+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 mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq;
+
+ queues_n = hash_fields ? queues_n : 1;
+ LIST_FOREACH(hrxq, &priv->hrxqs, next) {
+ struct mlx5_ind_table_obj *ind_tbl;
+
+ if (hrxq->rss_key_len != rss_key_len)
+ continue;
+ if (memcmp(hrxq->rss_key, rss_key, rss_key_len))
+ continue;
+ if (hrxq->hash_fields != hash_fields)
+ continue;
+ ind_tbl = mlx5_ind_table_obj_get(dev, queues, queues_n);
+ if (!ind_tbl)
+ continue;
+ if (ind_tbl != hrxq->ind_table) {
+ mlx5_ind_table_obj_release(dev, ind_tbl);
+ continue;
+ }
+ rte_atomic32_inc(&hrxq->refcnt);
+ return hrxq;
+ }
+ return NULL;
+}
+
+/**
+ * Release the hash Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param hrxq
+ * Pointer to Hash Rx queue to release.
+ *
+ * @return
+ * 1 while a reference on it exists, 0 when freed.
+ */
+int
+mlx5_hrxq_release(struct rte_eth_dev *dev, struct mlx5_hrxq *hrxq)
+{
+ if (rte_atomic32_dec_and_test(&hrxq->refcnt)) {
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ mlx5_glue->destroy_flow_action(hrxq->action);
+#endif
+ if (hrxq->ind_table->type == MLX5_IND_TBL_TYPE_IBV)
+ claim_zero(mlx5_glue->destroy_qp(hrxq->qp));
+ else /* hrxq->ind_table->type == MLX5_IND_TBL_TYPE_DEVX */
+ claim_zero(mlx5_devx_cmd_destroy(hrxq->tir));
+ mlx5_ind_table_obj_release(dev, hrxq->ind_table);
+ LIST_REMOVE(hrxq, next);
+ rte_free(hrxq);
+ return 0;
+ }
+ claim_nonzero(mlx5_ind_table_obj_release(dev, hrxq->ind_table));
+ return 1;
+}
+
+/**
+ * Verify the Rx Queue list is empty
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The number of object not released.
+ */
+int
+mlx5_hrxq_verify(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq;
+ int ret = 0;
+
+ LIST_FOREACH(hrxq, &priv->hrxqs, next) {
+ DRV_LOG(DEBUG,
+ "port %u hash Rx queue %p still referenced",
+ dev->data->port_id, (void *)hrxq);
+ ++ret;
+ }
+ return ret;
+}
+
+/**
+ * Create a drop Rx queue Verbs/DevX object.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The Verbs/DevX object initialised, NULL otherwise and rte_errno is set.
+ */
+static struct mlx5_rxq_obj *
+mlx5_rxq_obj_drop_new(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct ibv_context *ctx = priv->sh->ctx;
+ struct ibv_cq *cq;
+ struct ibv_wq *wq = NULL;
+ struct mlx5_rxq_obj *rxq;
+
+ if (priv->drop_queue.rxq)
+ return priv->drop_queue.rxq;
+ cq = mlx5_glue->create_cq(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(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) {
+ 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/DevX object.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The Verbs/DevX object initialised, NULL otherwise and rte_errno is set.
+ */
+static void
+mlx5_rxq_obj_drop_release(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_obj *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/DevX object initialised, NULL otherwise and rte_errno is set.
+ */
+static struct mlx5_ind_table_obj *
+mlx5_ind_table_obj_drop_new(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
+ struct mlx5_rxq_obj *rxq;
+ struct mlx5_ind_table_obj tmpl;
+
+ rxq = mlx5_rxq_obj_drop_new(dev);
+ if (!rxq)
+ return NULL;
+ tmpl.ind_table = mlx5_glue->create_rwq_ind_table
+ (priv->sh->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_obj_drop_release(dev);
+ return NULL;
+}
+
+/**
+ * Release a drop indirection table.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ */
+static void
+mlx5_ind_table_obj_drop_release(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl = priv->drop_queue.hrxq->ind_table;
+
+ claim_zero(mlx5_glue->destroy_rwq_ind_table(ind_tbl->ind_table));
+ mlx5_rxq_obj_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/DevX object initialised, NULL otherwise and rte_errno is set.
+ */
+struct mlx5_hrxq *
+mlx5_hrxq_drop_new(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl = NULL;
+ struct ibv_qp *qp = NULL;
+ struct mlx5_hrxq *hrxq = NULL;
+
+ if (priv->drop_queue.hrxq) {
+ rte_atomic32_inc(&priv->drop_queue.hrxq->refcnt);
+ return priv->drop_queue.hrxq;
+ }
+ 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;
+ }
+ priv->drop_queue.hrxq = hrxq;
+ ind_tbl = mlx5_ind_table_obj_drop_new(dev);
+ if (!ind_tbl)
+ goto error;
+ hrxq->ind_table = ind_tbl;
+ qp = mlx5_glue->create_qp_ex(priv->sh->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->sh->pd
+ });
+ if (!qp) {
+ DEBUG("port %u cannot allocate QP for drop queue",
+ dev->data->port_id);
+ rte_errno = errno;
+ goto error;
+ }
+ hrxq->qp = qp;
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ hrxq->action = mlx5_glue->dv_create_flow_action_dest_ibv_qp(hrxq->qp);
+ if (!hrxq->action) {
+ rte_errno = errno;
+ goto error;
+ }
+#endif
+ rte_atomic32_set(&hrxq->refcnt, 1);
+ return hrxq;
+error:
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ if (hrxq && hrxq->action)
+ mlx5_glue->destroy_flow_action(hrxq->action);
+#endif
+ if (qp)
+ claim_zero(mlx5_glue->destroy_qp(hrxq->qp));
+ if (ind_tbl)
+ mlx5_ind_table_obj_drop_release(dev);
+ if (hrxq) {
+ priv->drop_queue.hrxq = NULL;
+ rte_free(hrxq);
+ }
+ 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 mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq = priv->drop_queue.hrxq;
+
+ if (rte_atomic32_dec_and_test(&hrxq->refcnt)) {
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ mlx5_glue->destroy_flow_action(hrxq->action);
+#endif
+ claim_zero(mlx5_glue->destroy_qp(hrxq->qp));
+ mlx5_ind_table_obj_drop_release(dev);
+ rte_free(hrxq);
+ priv->drop_queue.hrxq = NULL;
+ }