+static struct rte_mempool *
+txq_mb2mp(struct rte_mbuf *buf)
+{
+ if (unlikely(RTE_MBUF_INDIRECT(buf)))
+ return rte_mbuf_from_indirect(buf)->pool;
+ return buf->pool;
+}
+
+/**
+ * Get Memory Region (MR) <-> Memory Pool (MP) association from txq->mp2mr[].
+ * Add MP to txq->mp2mr[] if it's not registered yet. If mp2mr[] is full,
+ * remove an entry first.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param[in] mp
+ * Memory Pool for which a Memory Region lkey must be returned.
+ *
+ * @return
+ * mr->lkey on success, (uint32_t)-1 on failure.
+ */
+static inline uint32_t
+txq_mp2mr(struct txq *txq, struct rte_mempool *mp)
+{
+ unsigned int i;
+ uint32_t lkey = (uint32_t)-1;
+
+ for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) {
+ if (unlikely(txq->mp2mr[i].mp == NULL)) {
+ /* Unknown MP, add a new MR for it. */
+ break;
+ }
+ if (txq->mp2mr[i].mp == mp) {
+ assert(txq->mp2mr[i].lkey != (uint32_t)-1);
+ assert(htonl(txq->mp2mr[i].mr->lkey) ==
+ txq->mp2mr[i].lkey);
+ lkey = txq->mp2mr[i].lkey;
+ break;
+ }
+ }
+ if (unlikely(lkey == (uint32_t)-1))
+ lkey = txq_mp2mr_reg(txq, mp, i);
+ return lkey;
+}
+
+/**
+ * Ring TX queue doorbell.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param wqe
+ * Pointer to the last WQE posted in the NIC.
+ */
+static inline void
+mlx5_tx_dbrec(struct txq *txq, volatile struct mlx5_wqe *wqe)
+{
+ uint64_t *dst = (uint64_t *)((uintptr_t)txq->bf_reg);
+ volatile uint64_t *src = ((volatile uint64_t *)wqe);
+
+ rte_wmb();
+ *txq->qp_db = htonl(txq->wqe_ci);
+ /* Ensure ordering between DB record and BF copy. */
+ rte_wmb();
+ *dst = *src;
+}
+
+/**
+ * DPDK callback to check the status of a tx descriptor.
+ *
+ * @param tx_queue
+ * The tx queue.
+ * @param[in] offset
+ * The index of the descriptor in the ring.
+ *
+ * @return
+ * The status of the tx descriptor.
+ */
+int
+mlx5_tx_descriptor_status(void *tx_queue, uint16_t offset)
+{
+ struct txq *txq = tx_queue;
+ const unsigned int elts_n = 1 << txq->elts_n;
+ const unsigned int elts_cnt = elts_n - 1;
+ unsigned int used;
+
+ txq_complete(txq);
+ used = (txq->elts_head - txq->elts_tail) & elts_cnt;
+ if (offset < used)
+ return RTE_ETH_TX_DESC_FULL;
+ return RTE_ETH_TX_DESC_DONE;
+}
+
+/**
+ * DPDK callback to check the status of a rx descriptor.
+ *
+ * @param rx_queue
+ * The rx queue.
+ * @param[in] offset
+ * The index of the descriptor in the ring.
+ *
+ * @return
+ * The status of the tx descriptor.
+ */
+int
+mlx5_rx_descriptor_status(void *rx_queue, uint16_t offset)
+{
+ struct rxq *rxq = rx_queue;
+ struct rxq_zip *zip = &rxq->zip;
+ volatile struct mlx5_cqe *cqe;
+ const unsigned int cqe_n = (1 << rxq->cqe_n);
+ const unsigned int cqe_cnt = cqe_n - 1;
+ unsigned int cq_ci;
+ unsigned int used;
+
+ /* if we are processing a compressed cqe */
+ if (zip->ai) {
+ used = zip->cqe_cnt - zip->ca;
+ cq_ci = zip->cq_ci;
+ } else {
+ used = 0;
+ cq_ci = rxq->cq_ci;
+ }
+ cqe = &(*rxq->cqes)[cq_ci & cqe_cnt];
+ while (check_cqe(cqe, cqe_n, cq_ci) == 0) {
+ int8_t op_own;
+ unsigned int n;
+
+ op_own = cqe->op_own;
+ if (MLX5_CQE_FORMAT(op_own) == MLX5_COMPRESSED)
+ n = ntohl(cqe->byte_cnt);
+ else
+ n = 1;
+ cq_ci += n;
+ used += n;
+ cqe = &(*rxq->cqes)[cq_ci & cqe_cnt];
+ }
+ used = RTE_MIN(used, (1U << rxq->elts_n) - 1);
+ if (offset < used)
+ return RTE_ETH_RX_DESC_DONE;
+ return RTE_ETH_RX_DESC_AVAIL;
+}
+
+/**
+ * DPDK callback for TX.
+ *
+ * @param dpdk_txq
+ * Generic pointer to TX queue structure.
+ * @param[in] pkts
+ * Packets to transmit.
+ * @param pkts_n
+ * Number of packets in array.
+ *
+ * @return
+ * Number of packets successfully transmitted (<= pkts_n).
+ */
+uint16_t
+mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ struct txq *txq = (struct txq *)dpdk_txq;
+ uint16_t elts_head = txq->elts_head;
+ const unsigned int elts_n = 1 << txq->elts_n;
+ unsigned int i = 0;
+ unsigned int j = 0;
+ unsigned int k = 0;
+ unsigned int max;
+ unsigned int max_inline = txq->max_inline;
+ const unsigned int inline_en = !!max_inline && txq->inline_en;
+ uint16_t max_wqe;
+ unsigned int comp;
+ volatile struct mlx5_wqe_v *wqe = NULL;
+ volatile struct mlx5_wqe_ctrl *last_wqe = NULL;
+ unsigned int segs_n = 0;
+ struct rte_mbuf *buf = NULL;
+ uint8_t *raw;
+
+ if (unlikely(!pkts_n))
+ return 0;
+ /* Prefetch first packet cacheline. */
+ rte_prefetch0(*pkts);
+ /* Start processing. */
+ txq_complete(txq);
+ max = (elts_n - (elts_head - txq->elts_tail));
+ if (max > elts_n)
+ max -= elts_n;
+ max_wqe = (1u << txq->wqe_n) - (txq->wqe_ci - txq->wqe_pi);
+ if (unlikely(!max_wqe))
+ return 0;
+ do {
+ volatile rte_v128u32_t *dseg = NULL;
+ uint32_t length;
+ unsigned int ds = 0;
+ unsigned int sg = 0; /* counter of additional segs attached. */
+ uintptr_t addr;
+ uint64_t naddr;
+ uint16_t pkt_inline_sz = MLX5_WQE_DWORD_SIZE + 2;
+ uint16_t tso_header_sz = 0;
+ uint16_t ehdr;
+ uint8_t cs_flags = 0;
+ uint64_t tso = 0;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ uint32_t total_length = 0;
+#endif
+
+ /* first_seg */
+ buf = *pkts;
+ segs_n = buf->nb_segs;
+ /*
+ * Make sure there is enough room to store this packet and
+ * that one ring entry remains unused.
+ */
+ assert(segs_n);
+ if (max < segs_n + 1)
+ break;
+ max -= segs_n;
+ --segs_n;
+ if (unlikely(--max_wqe == 0))
+ break;
+ wqe = (volatile struct mlx5_wqe_v *)
+ tx_mlx5_wqe(txq, txq->wqe_ci);
+ rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci + 1));
+ if (pkts_n - i > 1)
+ rte_prefetch0(*(pkts + 1));
+ addr = rte_pktmbuf_mtod(buf, uintptr_t);
+ length = DATA_LEN(buf);
+ ehdr = (((uint8_t *)addr)[1] << 8) |
+ ((uint8_t *)addr)[0];
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ total_length = length;
+#endif
+ if (length < (MLX5_WQE_DWORD_SIZE + 2))
+ break;
+ /* Update element. */
+ (*txq->elts)[elts_head] = buf;
+ /* Prefetch next buffer data. */
+ if (pkts_n - i > 1)
+ rte_prefetch0(
+ rte_pktmbuf_mtod(*(pkts + 1), volatile void *));
+ /* Should we enable HW CKSUM offload */
+ if (buf->ol_flags &
+ (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM)) {
+ const uint64_t is_tunneled = buf->ol_flags &
+ (PKT_TX_TUNNEL_GRE |
+ PKT_TX_TUNNEL_VXLAN);
+
+ if (is_tunneled && txq->tunnel_en) {
+ cs_flags = MLX5_ETH_WQE_L3_INNER_CSUM |
+ MLX5_ETH_WQE_L4_INNER_CSUM;
+ if (buf->ol_flags & PKT_TX_OUTER_IP_CKSUM)
+ cs_flags |= MLX5_ETH_WQE_L3_CSUM;
+ } else {
+ cs_flags = MLX5_ETH_WQE_L3_CSUM |
+ MLX5_ETH_WQE_L4_CSUM;
+ }
+ }
+ raw = ((uint8_t *)(uintptr_t)wqe) + 2 * MLX5_WQE_DWORD_SIZE;
+ /* Replace the Ethernet type by the VLAN if necessary. */
+ if (buf->ol_flags & PKT_TX_VLAN_PKT) {
+ uint32_t vlan = htonl(0x81000000 | buf->vlan_tci);
+ unsigned int len = 2 * ETHER_ADDR_LEN - 2;
+
+ addr += 2;
+ length -= 2;
+ /* Copy Destination and source mac address. */
+ memcpy((uint8_t *)raw, ((uint8_t *)addr), len);
+ /* Copy VLAN. */
+ memcpy((uint8_t *)raw + len, &vlan, sizeof(vlan));
+ /* Copy missing two bytes to end the DSeg. */
+ memcpy((uint8_t *)raw + len + sizeof(vlan),
+ ((uint8_t *)addr) + len, 2);
+ addr += len + 2;
+ length -= (len + 2);
+ } else {
+ memcpy((uint8_t *)raw, ((uint8_t *)addr) + 2,
+ MLX5_WQE_DWORD_SIZE);
+ length -= pkt_inline_sz;
+ addr += pkt_inline_sz;
+ }
+ if (txq->tso_en) {
+ tso = buf->ol_flags & PKT_TX_TCP_SEG;
+ if (tso) {
+ uintptr_t end = (uintptr_t)
+ (((uintptr_t)txq->wqes) +
+ (1 << txq->wqe_n) *
+ MLX5_WQE_SIZE);
+ unsigned int copy_b;
+ uint8_t vlan_sz = (buf->ol_flags &
+ PKT_TX_VLAN_PKT) ? 4 : 0;
+ const uint64_t is_tunneled =
+ buf->ol_flags &
+ (PKT_TX_TUNNEL_GRE |
+ PKT_TX_TUNNEL_VXLAN);
+
+ tso_header_sz = buf->l2_len + vlan_sz +
+ buf->l3_len + buf->l4_len;
+
+ if (is_tunneled && txq->tunnel_en) {
+ tso_header_sz += buf->outer_l2_len +
+ buf->outer_l3_len;
+ cs_flags |= MLX5_ETH_WQE_L4_INNER_CSUM;
+ } else {
+ cs_flags |= MLX5_ETH_WQE_L4_CSUM;
+ }
+ if (unlikely(tso_header_sz >
+ MLX5_MAX_TSO_HEADER))
+ break;
+ copy_b = tso_header_sz - pkt_inline_sz;
+ /* First seg must contain all headers. */
+ assert(copy_b <= length);
+ raw += MLX5_WQE_DWORD_SIZE;
+ if (copy_b &&
+ ((end - (uintptr_t)raw) > copy_b)) {
+ uint16_t n = (MLX5_WQE_DS(copy_b) -
+ 1 + 3) / 4;
+
+ if (unlikely(max_wqe < n))
+ break;
+ max_wqe -= n;
+ rte_memcpy((void *)raw,
+ (void *)addr, copy_b);
+ addr += copy_b;
+ length -= copy_b;
+ pkt_inline_sz += copy_b;
+ /*
+ * Another DWORD will be added
+ * in the inline part.
+ */
+ raw += MLX5_WQE_DS(copy_b) *
+ MLX5_WQE_DWORD_SIZE -
+ MLX5_WQE_DWORD_SIZE;
+ } else {
+ /* NOP WQE. */
+ wqe->ctrl = (rte_v128u32_t){
+ htonl(txq->wqe_ci << 8),
+ htonl(txq->qp_num_8s | 1),
+ 0,
+ 0,
+ };
+ ds = 1;
+ total_length = 0;
+ k++;
+ goto next_wqe;
+ }
+ }
+ }
+ /* Inline if enough room. */
+ if (inline_en || tso) {
+ uintptr_t end = (uintptr_t)
+ (((uintptr_t)txq->wqes) +
+ (1 << txq->wqe_n) * MLX5_WQE_SIZE);
+ unsigned int inline_room = max_inline *
+ RTE_CACHE_LINE_SIZE -
+ (pkt_inline_sz - 2);
+ uintptr_t addr_end = (addr + inline_room) &
+ ~(RTE_CACHE_LINE_SIZE - 1);
+ unsigned int copy_b = (addr_end > addr) ?
+ RTE_MIN((addr_end - addr), length) :
+ 0;
+
+ raw += MLX5_WQE_DWORD_SIZE;
+ if (copy_b && ((end - (uintptr_t)raw) > copy_b)) {
+ /*
+ * One Dseg remains in the current WQE. To
+ * keep the computation positive, it is
+ * removed after the bytes to Dseg conversion.
+ */
+ uint16_t n = (MLX5_WQE_DS(copy_b) - 1 + 3) / 4;
+
+ if (unlikely(max_wqe < n))
+ break;
+ max_wqe -= n;
+ if (tso) {
+ uint32_t inl =
+ htonl(copy_b | MLX5_INLINE_SEG);
+
+ pkt_inline_sz =
+ MLX5_WQE_DS(tso_header_sz) *
+ MLX5_WQE_DWORD_SIZE;
+ rte_memcpy((void *)raw,
+ (void *)&inl, sizeof(inl));
+ raw += sizeof(inl);
+ pkt_inline_sz += sizeof(inl);
+ }
+ rte_memcpy((void *)raw, (void *)addr, copy_b);
+ addr += copy_b;
+ length -= copy_b;
+ pkt_inline_sz += copy_b;
+ }
+ /*
+ * 2 DWORDs consumed by the WQE header + ETH segment +
+ * the size of the inline part of the packet.
+ */
+ ds = 2 + MLX5_WQE_DS(pkt_inline_sz - 2);
+ if (length > 0) {
+ if (ds % (MLX5_WQE_SIZE /
+ MLX5_WQE_DWORD_SIZE) == 0) {
+ if (unlikely(--max_wqe == 0))
+ break;
+ dseg = (volatile rte_v128u32_t *)
+ tx_mlx5_wqe(txq, txq->wqe_ci +
+ ds / 4);
+ } else {
+ dseg = (volatile rte_v128u32_t *)
+ ((uintptr_t)wqe +
+ (ds * MLX5_WQE_DWORD_SIZE));
+ }
+ goto use_dseg;
+ } else if (!segs_n) {
+ goto next_pkt;
+ } else {
+ /* dseg will be advance as part of next_seg */
+ dseg = (volatile rte_v128u32_t *)
+ ((uintptr_t)wqe +
+ ((ds - 1) * MLX5_WQE_DWORD_SIZE));
+ goto next_seg;
+ }
+ } else {
+ /*
+ * No inline has been done in the packet, only the
+ * Ethernet Header as been stored.
+ */
+ dseg = (volatile rte_v128u32_t *)
+ ((uintptr_t)wqe + (3 * MLX5_WQE_DWORD_SIZE));
+ ds = 3;
+use_dseg:
+ /* Add the remaining packet as a simple ds. */
+ naddr = htonll(addr);
+ *dseg = (rte_v128u32_t){
+ htonl(length),
+ txq_mp2mr(txq, txq_mb2mp(buf)),
+ naddr,
+ naddr >> 32,
+ };
+ ++ds;
+ if (!segs_n)
+ goto next_pkt;
+ }
+next_seg:
+ assert(buf);
+ assert(ds);
+ assert(wqe);
+ /*
+ * Spill on next WQE when the current one does not have
+ * enough room left. Size of WQE must a be a multiple
+ * of data segment size.
+ */
+ assert(!(MLX5_WQE_SIZE % MLX5_WQE_DWORD_SIZE));
+ if (!(ds % (MLX5_WQE_SIZE / MLX5_WQE_DWORD_SIZE))) {
+ if (unlikely(--max_wqe == 0))
+ break;
+ dseg = (volatile rte_v128u32_t *)
+ tx_mlx5_wqe(txq, txq->wqe_ci + ds / 4);
+ rte_prefetch0(tx_mlx5_wqe(txq,
+ txq->wqe_ci + ds / 4 + 1));
+ } else {
+ ++dseg;
+ }
+ ++ds;
+ buf = buf->next;
+ assert(buf);
+ length = DATA_LEN(buf);
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ total_length += length;
+#endif
+ /* Store segment information. */
+ naddr = htonll(rte_pktmbuf_mtod(buf, uintptr_t));
+ *dseg = (rte_v128u32_t){
+ htonl(length),
+ txq_mp2mr(txq, txq_mb2mp(buf)),
+ naddr,
+ naddr >> 32,
+ };
+ elts_head = (elts_head + 1) & (elts_n - 1);
+ (*txq->elts)[elts_head] = buf;
+ ++sg;
+ /* Advance counter only if all segs are successfully posted. */
+ if (sg < segs_n)
+ goto next_seg;
+ else
+ j += sg;
+next_pkt:
+ elts_head = (elts_head + 1) & (elts_n - 1);
+ ++pkts;
+ ++i;
+ /* Initialize known and common part of the WQE structure. */
+ if (tso) {
+ wqe->ctrl = (rte_v128u32_t){
+ htonl((txq->wqe_ci << 8) | MLX5_OPCODE_TSO),
+ htonl(txq->qp_num_8s | ds),
+ 0,
+ 0,
+ };
+ wqe->eseg = (rte_v128u32_t){
+ 0,
+ cs_flags | (htons(buf->tso_segsz) << 16),
+ 0,
+ (ehdr << 16) | htons(tso_header_sz),
+ };
+ } else {
+ wqe->ctrl = (rte_v128u32_t){
+ htonl((txq->wqe_ci << 8) | MLX5_OPCODE_SEND),
+ htonl(txq->qp_num_8s | ds),
+ 0,
+ 0,
+ };
+ wqe->eseg = (rte_v128u32_t){
+ 0,
+ cs_flags,
+ 0,
+ (ehdr << 16) | htons(pkt_inline_sz),
+ };
+ }
+next_wqe:
+ txq->wqe_ci += (ds + 3) / 4;
+ /* Save the last successful WQE for completion request */
+ last_wqe = (volatile struct mlx5_wqe_ctrl *)wqe;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment sent bytes counter. */
+ txq->stats.obytes += total_length;
+#endif
+ } while (i < pkts_n);
+ /* Take a shortcut if nothing must be sent. */
+ if (unlikely((i + k) == 0))
+ return 0;
+ txq->elts_head = (txq->elts_head + i + j) & (elts_n - 1);
+ /* Check whether completion threshold has been reached. */
+ comp = txq->elts_comp + i + j + k;
+ if (comp >= MLX5_TX_COMP_THRESH) {
+ /* Request completion on last WQE. */
+ last_wqe->ctrl2 = htonl(8);
+ /* Save elts_head in unused "immediate" field of WQE. */
+ last_wqe->ctrl3 = txq->elts_head;
+ txq->elts_comp = 0;
+ } else {
+ txq->elts_comp = comp;
+ }
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment sent packets counter. */
+ txq->stats.opackets += i;
+#endif
+ /* Ring QP doorbell. */
+ mlx5_tx_dbrec(txq, (volatile struct mlx5_wqe *)last_wqe);
+ return i;
+}
+
+/**
+ * Open a MPW session.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param mpw
+ * Pointer to MPW session structure.
+ * @param length
+ * Packet length.
+ */
+static inline void
+mlx5_mpw_new(struct txq *txq, struct mlx5_mpw *mpw, uint32_t length)
+{
+ uint16_t idx = txq->wqe_ci & ((1 << txq->wqe_n) - 1);
+ volatile struct mlx5_wqe_data_seg (*dseg)[MLX5_MPW_DSEG_MAX] =
+ (volatile struct mlx5_wqe_data_seg (*)[])
+ tx_mlx5_wqe(txq, idx + 1);
+
+ mpw->state = MLX5_MPW_STATE_OPENED;
+ mpw->pkts_n = 0;
+ mpw->len = length;
+ mpw->total_len = 0;
+ mpw->wqe = (volatile struct mlx5_wqe *)tx_mlx5_wqe(txq, idx);
+ mpw->wqe->eseg.mss = htons(length);
+ mpw->wqe->eseg.inline_hdr_sz = 0;
+ mpw->wqe->eseg.rsvd0 = 0;
+ mpw->wqe->eseg.rsvd1 = 0;
+ mpw->wqe->eseg.rsvd2 = 0;
+ mpw->wqe->ctrl[0] = htonl((MLX5_OPC_MOD_MPW << 24) |
+ (txq->wqe_ci << 8) | MLX5_OPCODE_TSO);
+ mpw->wqe->ctrl[2] = 0;
+ mpw->wqe->ctrl[3] = 0;
+ mpw->data.dseg[0] = (volatile struct mlx5_wqe_data_seg *)
+ (((uintptr_t)mpw->wqe) + (2 * MLX5_WQE_DWORD_SIZE));
+ mpw->data.dseg[1] = (volatile struct mlx5_wqe_data_seg *)
+ (((uintptr_t)mpw->wqe) + (3 * MLX5_WQE_DWORD_SIZE));
+ mpw->data.dseg[2] = &(*dseg)[0];
+ mpw->data.dseg[3] = &(*dseg)[1];
+ mpw->data.dseg[4] = &(*dseg)[2];
+}
+
+/**
+ * Close a MPW session.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param mpw
+ * Pointer to MPW session structure.
+ */
+static inline void
+mlx5_mpw_close(struct txq *txq, struct mlx5_mpw *mpw)
+{
+ unsigned int num = mpw->pkts_n;
+
+ /*
+ * Store size in multiple of 16 bytes. Control and Ethernet segments
+ * count as 2.
+ */
+ mpw->wqe->ctrl[1] = htonl(txq->qp_num_8s | (2 + num));
+ mpw->state = MLX5_MPW_STATE_CLOSED;
+ if (num < 3)
+ ++txq->wqe_ci;
+ else
+ txq->wqe_ci += 2;
+ rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci));
+ rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci + 1));
+}
+
+/**
+ * DPDK callback for TX with MPW support.
+ *
+ * @param dpdk_txq
+ * Generic pointer to TX queue structure.
+ * @param[in] pkts
+ * Packets to transmit.
+ * @param pkts_n
+ * Number of packets in array.
+ *
+ * @return
+ * Number of packets successfully transmitted (<= pkts_n).
+ */
+uint16_t
+mlx5_tx_burst_mpw(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ struct txq *txq = (struct txq *)dpdk_txq;
+ uint16_t elts_head = txq->elts_head;
+ const unsigned int elts_n = 1 << txq->elts_n;
+ unsigned int i = 0;
+ unsigned int j = 0;
+ unsigned int max;
+ uint16_t max_wqe;
+ unsigned int comp;
+ struct mlx5_mpw mpw = {
+ .state = MLX5_MPW_STATE_CLOSED,
+ };
+
+ if (unlikely(!pkts_n))
+ return 0;
+ /* Prefetch first packet cacheline. */
+ rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci));
+ rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci + 1));
+ /* Start processing. */
+ txq_complete(txq);
+ max = (elts_n - (elts_head - txq->elts_tail));
+ if (max > elts_n)
+ max -= elts_n;
+ max_wqe = (1u << txq->wqe_n) - (txq->wqe_ci - txq->wqe_pi);
+ if (unlikely(!max_wqe))
+ return 0;
+ do {
+ struct rte_mbuf *buf = *(pkts++);
+ unsigned int elts_head_next;
+ uint32_t length;
+ unsigned int segs_n = buf->nb_segs;
+ uint32_t cs_flags = 0;
+
+ /*
+ * Make sure there is enough room to store this packet and
+ * that one ring entry remains unused.
+ */
+ assert(segs_n);
+ if (max < segs_n + 1)
+ break;
+ /* Do not bother with large packets MPW cannot handle. */
+ if (segs_n > MLX5_MPW_DSEG_MAX)
+ break;
+ max -= segs_n;
+ --pkts_n;
+ /* Should we enable HW CKSUM offload */
+ if (buf->ol_flags &
+ (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM))
+ cs_flags = MLX5_ETH_WQE_L3_CSUM | MLX5_ETH_WQE_L4_CSUM;
+ /* Retrieve packet information. */
+ length = PKT_LEN(buf);
+ assert(length);
+ /* Start new session if packet differs. */
+ if ((mpw.state == MLX5_MPW_STATE_OPENED) &&
+ ((mpw.len != length) ||
+ (segs_n != 1) ||
+ (mpw.wqe->eseg.cs_flags != cs_flags)))
+ mlx5_mpw_close(txq, &mpw);
+ if (mpw.state == MLX5_MPW_STATE_CLOSED) {
+ /*
+ * Multi-Packet WQE consumes at most two WQE.
+ * mlx5_mpw_new() expects to be able to use such
+ * resources.
+ */
+ if (unlikely(max_wqe < 2))
+ break;
+ max_wqe -= 2;
+ mlx5_mpw_new(txq, &mpw, length);
+ mpw.wqe->eseg.cs_flags = cs_flags;
+ }
+ /* Multi-segment packets must be alone in their MPW. */
+ assert((segs_n == 1) || (mpw.pkts_n == 0));
+#if defined(MLX5_PMD_SOFT_COUNTERS) || !defined(NDEBUG)
+ length = 0;
+#endif
+ do {
+ volatile struct mlx5_wqe_data_seg *dseg;
+ uintptr_t addr;
+
+ elts_head_next = (elts_head + 1) & (elts_n - 1);
+ assert(buf);
+ (*txq->elts)[elts_head] = buf;
+ dseg = mpw.data.dseg[mpw.pkts_n];
+ addr = rte_pktmbuf_mtod(buf, uintptr_t);
+ *dseg = (struct mlx5_wqe_data_seg){
+ .byte_count = htonl(DATA_LEN(buf)),
+ .lkey = txq_mp2mr(txq, txq_mb2mp(buf)),
+ .addr = htonll(addr),
+ };
+ elts_head = elts_head_next;
+#if defined(MLX5_PMD_SOFT_COUNTERS) || !defined(NDEBUG)
+ length += DATA_LEN(buf);
+#endif
+ buf = buf->next;
+ ++mpw.pkts_n;
+ ++j;
+ } while (--segs_n);
+ assert(length == mpw.len);
+ if (mpw.pkts_n == MLX5_MPW_DSEG_MAX)
+ mlx5_mpw_close(txq, &mpw);
+ elts_head = elts_head_next;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment sent bytes counter. */
+ txq->stats.obytes += length;
+#endif
+ ++i;
+ } while (pkts_n);
+ /* Take a shortcut if nothing must be sent. */
+ if (unlikely(i == 0))
+ return 0;
+ /* Check whether completion threshold has been reached. */
+ /* "j" includes both packets and segments. */
+ comp = txq->elts_comp + j;
+ if (comp >= MLX5_TX_COMP_THRESH) {
+ volatile struct mlx5_wqe *wqe = mpw.wqe;
+
+ /* Request completion on last WQE. */
+ wqe->ctrl[2] = htonl(8);
+ /* Save elts_head in unused "immediate" field of WQE. */
+ wqe->ctrl[3] = elts_head;
+ txq->elts_comp = 0;
+ } else {
+ txq->elts_comp = comp;
+ }
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment sent packets counter. */
+ txq->stats.opackets += i;
+#endif
+ /* Ring QP doorbell. */
+ if (mpw.state == MLX5_MPW_STATE_OPENED)
+ mlx5_mpw_close(txq, &mpw);
+ mlx5_tx_dbrec(txq, mpw.wqe);
+ txq->elts_head = elts_head;
+ return i;
+}
+
+/**
+ * Open a MPW inline session.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param mpw
+ * Pointer to MPW session structure.
+ * @param length
+ * Packet length.
+ */
+static inline void
+mlx5_mpw_inline_new(struct txq *txq, struct mlx5_mpw *mpw, uint32_t length)
+{
+ uint16_t idx = txq->wqe_ci & ((1 << txq->wqe_n) - 1);
+ struct mlx5_wqe_inl_small *inl;
+
+ mpw->state = MLX5_MPW_INL_STATE_OPENED;
+ mpw->pkts_n = 0;
+ mpw->len = length;
+ mpw->total_len = 0;
+ mpw->wqe = (volatile struct mlx5_wqe *)tx_mlx5_wqe(txq, idx);
+ mpw->wqe->ctrl[0] = htonl((MLX5_OPC_MOD_MPW << 24) |
+ (txq->wqe_ci << 8) |
+ MLX5_OPCODE_TSO);
+ mpw->wqe->ctrl[2] = 0;
+ mpw->wqe->ctrl[3] = 0;
+ mpw->wqe->eseg.mss = htons(length);
+ mpw->wqe->eseg.inline_hdr_sz = 0;
+ mpw->wqe->eseg.cs_flags = 0;
+ mpw->wqe->eseg.rsvd0 = 0;
+ mpw->wqe->eseg.rsvd1 = 0;
+ mpw->wqe->eseg.rsvd2 = 0;
+ inl = (struct mlx5_wqe_inl_small *)
+ (((uintptr_t)mpw->wqe) + 2 * MLX5_WQE_DWORD_SIZE);
+ mpw->data.raw = (uint8_t *)&inl->raw;
+}
+
+/**
+ * Close a MPW inline session.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param mpw
+ * Pointer to MPW session structure.
+ */
+static inline void
+mlx5_mpw_inline_close(struct txq *txq, struct mlx5_mpw *mpw)