From fca143f059a0bddd7d47b8dc2df646a891b0eb0f Mon Sep 17 00:00:00 2001 From: Christian Ehrhardt Date: Tue, 16 May 2017 14:51:32 +0200 Subject: Imported Upstream version 17.05 --- drivers/net/mlx5/mlx5_rxtx.c | 1097 +++++++++++++++++++++++++++++++++++------- 1 file changed, 927 insertions(+), 170 deletions(-) (limited to 'drivers/net/mlx5/mlx5_rxtx.c') diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c index 3997b27a..de6e0fa4 100644 --- a/drivers/net/mlx5/mlx5_rxtx.c +++ b/drivers/net/mlx5/mlx5_rxtx.c @@ -74,6 +74,9 @@ check_cqe(volatile struct mlx5_cqe *cqe, unsigned int cqes_n, const uint16_t ci) __attribute__((always_inline)); +static inline void +txq_complete(struct txq *txq) __attribute__((always_inline)); + static inline uint32_t txq_mp2mr(struct txq *txq, struct rte_mempool *mp) __attribute__((always_inline)); @@ -110,7 +113,7 @@ static inline int check_cqe_seen(volatile struct mlx5_cqe *cqe) { static const uint8_t magic[] = "seen"; - volatile uint8_t (*buf)[sizeof(cqe->rsvd3)] = &cqe->rsvd3; + volatile uint8_t (*buf)[sizeof(cqe->rsvd0)] = &cqe->rsvd0; int ret = 1; unsigned int i; @@ -173,8 +176,79 @@ check_cqe(volatile struct mlx5_cqe *cqe, return 0; } -static inline void -txq_complete(struct txq *txq) __attribute__((always_inline)); +/** + * Return the address of the WQE. + * + * @param txq + * Pointer to TX queue structure. + * @param wqe_ci + * WQE consumer index. + * + * @return + * WQE address. + */ +static inline uintptr_t * +tx_mlx5_wqe(struct txq *txq, uint16_t ci) +{ + ci &= ((1 << txq->wqe_n) - 1); + return (uintptr_t *)((uintptr_t)txq->wqes + ci * MLX5_WQE_SIZE); +} + +/** + * Return the size of tailroom of WQ. + * + * @param txq + * Pointer to TX queue structure. + * @param addr + * Pointer to tail of WQ. + * + * @return + * Size of tailroom. + */ +static inline size_t +tx_mlx5_wq_tailroom(struct txq *txq, void *addr) +{ + size_t tailroom; + tailroom = (uintptr_t)(txq->wqes) + + (1 << txq->wqe_n) * MLX5_WQE_SIZE - + (uintptr_t)addr; + return tailroom; +} + +/** + * Copy data to tailroom of circular queue. + * + * @param dst + * Pointer to destination. + * @param src + * Pointer to source. + * @param n + * Number of bytes to copy. + * @param base + * Pointer to head of queue. + * @param tailroom + * Size of tailroom from dst. + * + * @return + * Pointer after copied data. + */ +static inline void * +mlx5_copy_to_wq(void *dst, const void *src, size_t n, + void *base, size_t tailroom) +{ + void *ret; + + if (n > tailroom) { + rte_memcpy(dst, src, tailroom); + rte_memcpy(base, (void *)((uintptr_t)src + tailroom), + n - tailroom); + ret = (uint8_t *)base + n - tailroom; + } else { + rte_memcpy(dst, src, n); + ret = (n == tailroom) ? base : (uint8_t *)dst + n; + } + return ret; +} /** * Manage TX completions. @@ -194,7 +268,7 @@ txq_complete(struct txq *txq) uint16_t elts_tail; uint16_t cq_ci = txq->cq_ci; volatile struct mlx5_cqe *cqe = NULL; - volatile struct mlx5_wqe *wqe; + volatile struct mlx5_wqe_ctrl *ctrl; do { volatile struct mlx5_cqe *tmp; @@ -220,9 +294,10 @@ txq_complete(struct txq *txq) } while (1); if (unlikely(cqe == NULL)) return; - wqe = &(*txq->wqes)[ntohs(cqe->wqe_counter) & - ((1 << txq->wqe_n) - 1)].hdr; - elts_tail = wqe->ctrl[3]; + txq->wqe_pi = ntohs(cqe->wqe_counter); + ctrl = (volatile struct mlx5_wqe_ctrl *) + tx_mlx5_wqe(txq, txq->wqe_pi); + elts_tail = ctrl->ctrl3; assert(elts_tail < (1 << txq->wqe_n)); /* Free buffers. */ while (elts_free != elts_tail) { @@ -326,37 +401,79 @@ mlx5_tx_dbrec(struct txq *txq, volatile struct mlx5_wqe *wqe) } /** - * Prefetch a CQE. + * DPDK callback to check the status of a tx descriptor. * - * @param txq - * Pointer to TX queue structure. - * @param cqe_ci - * CQE consumer index. + * @param tx_queue + * The tx queue. + * @param[in] offset + * The index of the descriptor in the ring. + * + * @return + * The status of the tx descriptor. */ -static inline void -tx_prefetch_cqe(struct txq *txq, uint16_t ci) +int +mlx5_tx_descriptor_status(void *tx_queue, uint16_t offset) { - volatile struct mlx5_cqe *cqe; + 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; - cqe = &(*txq->cqes)[ci & ((1 << txq->cqe_n) - 1)]; - rte_prefetch0(cqe); + 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; } /** - * Prefetch a WQE. + * DPDK callback to check the status of a rx descriptor. * - * @param txq - * Pointer to TX queue structure. - * @param wqe_ci - * WQE consumer index. + * @param rx_queue + * The rx queue. + * @param[in] offset + * The index of the descriptor in the ring. + * + * @return + * The status of the tx descriptor. */ -static inline void -tx_prefetch_wqe(struct txq *txq, uint16_t ci) +int +mlx5_rx_descriptor_status(void *rx_queue, uint16_t offset) { - volatile struct mlx5_wqe64 *wqe; + 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; - wqe = &(*txq->wqes)[ci & ((1 << txq->wqe_n) - 1)]; - rte_prefetch0(wqe); + 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; } /** @@ -380,9 +497,14 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) 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 *wqe = NULL; + 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; @@ -390,25 +512,33 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) if (unlikely(!pkts_n)) return 0; /* Prefetch first packet cacheline. */ - tx_prefetch_cqe(txq, txq->cq_ci); - tx_prefetch_cqe(txq, txq->cq_ci + 1); 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 struct mlx5_wqe_data_seg *dseg = NULL; + 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++); + buf = *pkts; segs_n = buf->nb_segs; /* * Make sure there is enough room to store this packet and @@ -419,104 +549,204 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) break; max -= segs_n; --segs_n; - if (!segs_n) - --pkts_n; - wqe = &(*txq->wqes)[txq->wqe_ci & - ((1 << txq->wqe_n) - 1)].hdr; - tx_prefetch_wqe(txq, txq->wqe_ci + 1); - if (pkts_n > 1) - rte_prefetch0(*pkts); + 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 - assert(length >= MLX5_WQE_DWORD_SIZE); + if (length < (MLX5_WQE_DWORD_SIZE + 2)) + break; /* Update element. */ (*txq->elts)[elts_head] = buf; - elts_head = (elts_head + 1) & (elts_n - 1); /* Prefetch next buffer data. */ - if (pkts_n > 1) { - volatile void *pkt_addr; - - pkt_addr = rte_pktmbuf_mtod(*pkts, volatile void *); - rte_prefetch0(pkt_addr); - } + 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)) { - wqe->eseg.cs_flags = - MLX5_ETH_WQE_L3_CSUM | - MLX5_ETH_WQE_L4_CSUM; - } else { - wqe->eseg.cs_flags = 0; + 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->eseg.inline_hdr[0]; - /* Start the know and common part of the WQE structure. */ - wqe->ctrl[0] = htonl((txq->wqe_ci << 8) | MLX5_OPCODE_SEND); - wqe->ctrl[2] = 0; - wqe->ctrl[3] = 0; - wqe->eseg.rsvd0 = 0; - wqe->eseg.rsvd1 = 0; - wqe->eseg.mss = 0; - wqe->eseg.rsvd2 = 0; - /* Start by copying the Ethernet Header. */ - memcpy((uint8_t *)raw, ((uint8_t *)addr), 16); - length -= MLX5_WQE_DWORD_SIZE; - addr += MLX5_WQE_DWORD_SIZE; + 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); - - memcpy((uint8_t *)(raw + MLX5_WQE_DWORD_SIZE - - sizeof(vlan)), - &vlan, sizeof(vlan)); - addr -= sizeof(vlan); - length += sizeof(vlan); + 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 (txq->max_inline != 0) { - uintptr_t end = - (uintptr_t)&(*txq->wqes)[1 << txq->wqe_n]; - uint16_t max_inline = - txq->max_inline * RTE_CACHE_LINE_SIZE; - uint16_t pkt_inline_sz = MLX5_WQE_DWORD_SIZE; - uint16_t 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; - room = end - (uintptr_t)raw; - if (room > max_inline) { - uintptr_t addr_end = (addr + max_inline) & - ~(RTE_CACHE_LINE_SIZE - 1); - uint16_t copy_b = ((addr_end - addr) > length) ? - length : - (addr_end - addr); - + 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; - /* Sanity check. */ - assert(addr <= addr_end); } - /* Store the inlined packet size in the WQE. */ - wqe->eseg.inline_hdr_sz = htons(pkt_inline_sz); /* - * 2 DWORDs consumed by the WQE header + 1 DSEG + + * 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) { - dseg = (struct mlx5_wqe_data_seg *) - ((uintptr_t)wqe + - (ds * MLX5_WQE_DWORD_SIZE)); - if ((uintptr_t)dseg >= end) - dseg = (struct mlx5_wqe_data_seg *) - ((uintptr_t)&(*txq->wqes)[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 { @@ -524,16 +754,17 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) * No inline has been done in the packet, only the * Ethernet Header as been stored. */ - wqe->eseg.inline_hdr_sz = htons(MLX5_WQE_DWORD_SIZE); - dseg = (struct mlx5_wqe_data_seg *) + 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. */ - *dseg = (struct mlx5_wqe_data_seg) { - .addr = htonll(addr), - .byte_count = htonl(length), - .lkey = txq_mp2mr(txq, txq_mb2mp(buf)), + naddr = htonll(addr); + *dseg = (rte_v128u32_t){ + htonl(length), + txq_mp2mr(txq, txq_mb2mp(buf)), + naddr, + naddr >> 32, }; ++ds; if (!segs_n) @@ -550,12 +781,12 @@ next_seg: */ assert(!(MLX5_WQE_SIZE % MLX5_WQE_DWORD_SIZE)); if (!(ds % (MLX5_WQE_SIZE / MLX5_WQE_DWORD_SIZE))) { - unsigned int n = (txq->wqe_ci + ((ds + 3) / 4)) & - ((1 << txq->wqe_n) - 1); - - dseg = (struct mlx5_wqe_data_seg *) - ((uintptr_t)&(*txq->wqes)[n]); - tx_prefetch_wqe(txq, n + 1); + 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; } @@ -567,38 +798,73 @@ next_seg: total_length += length; #endif /* Store segment information. */ - *dseg = (struct mlx5_wqe_data_seg) { - .addr = htonll(rte_pktmbuf_mtod(buf, uintptr_t)), - .byte_count = htonl(length), - .lkey = txq_mp2mr(txq, txq_mb2mp(buf)), + naddr = htonll(rte_pktmbuf_mtod(buf, uintptr_t)); + *dseg = (rte_v128u32_t){ + htonl(length), + txq_mp2mr(txq, txq_mb2mp(buf)), + naddr, + naddr >> 32, }; - (*txq->elts)[elts_head] = buf; elts_head = (elts_head + 1) & (elts_n - 1); - ++j; - --segs_n; - if (segs_n) + (*txq->elts)[elts_head] = buf; + ++sg; + /* Advance counter only if all segs are successfully posted. */ + if (sg < segs_n) goto next_seg; else - --pkts_n; + j += sg; next_pkt: + elts_head = (elts_head + 1) & (elts_n - 1); + ++pkts; ++i; - wqe->ctrl[1] = htonl(txq->qp_num_8s | ds); + /* 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 (pkts_n); + } while (i < pkts_n); /* Take a shortcut if nothing must be sent. */ - if (unlikely(i == 0)) + 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; + comp = txq->elts_comp + i + j + k; if (comp >= MLX5_TX_COMP_THRESH) { /* Request completion on last WQE. */ - wqe->ctrl[2] = htonl(8); + last_wqe->ctrl2 = htonl(8); /* Save elts_head in unused "immediate" field of WQE. */ - wqe->ctrl[3] = elts_head; + last_wqe->ctrl3 = txq->elts_head; txq->elts_comp = 0; } else { txq->elts_comp = comp; @@ -608,8 +874,7 @@ next_pkt: txq->stats.opackets += i; #endif /* Ring QP doorbell. */ - mlx5_tx_dbrec(txq, (volatile struct mlx5_wqe *)wqe); - txq->elts_head = elts_head; + mlx5_tx_dbrec(txq, (volatile struct mlx5_wqe *)last_wqe); return i; } @@ -629,13 +894,13 @@ 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 (*)[]) - (uintptr_t)&(*txq->wqes)[(idx + 1) & ((1 << txq->wqe_n) - 1)]; + 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 *)&(*txq->wqes)[idx].hdr; + 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; @@ -677,8 +942,8 @@ mlx5_mpw_close(struct txq *txq, struct mlx5_mpw *mpw) ++txq->wqe_ci; else txq->wqe_ci += 2; - tx_prefetch_wqe(txq, txq->wqe_ci); - tx_prefetch_wqe(txq, txq->wqe_ci + 1); + rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci)); + rte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci + 1)); } /** @@ -703,6 +968,7 @@ mlx5_tx_burst_mpw(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_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, @@ -711,14 +977,16 @@ mlx5_tx_burst_mpw(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) if (unlikely(!pkts_n)) return 0; /* Prefetch first packet cacheline. */ - tx_prefetch_cqe(txq, txq->cq_ci); - tx_prefetch_wqe(txq, txq->wqe_ci); - tx_prefetch_wqe(txq, txq->wqe_ci + 1); + 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; @@ -752,6 +1020,14 @@ mlx5_tx_burst_mpw(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) (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; } @@ -841,7 +1117,7 @@ mlx5_mpw_inline_new(struct txq *txq, struct mlx5_mpw *mpw, uint32_t length) mpw->pkts_n = 0; mpw->len = length; mpw->total_len = 0; - mpw->wqe = (volatile struct mlx5_wqe *)&(*txq->wqes)[idx].hdr; + 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); @@ -907,18 +1183,30 @@ mlx5_tx_burst_mpw_inline(void *dpdk_txq, struct rte_mbuf **pkts, unsigned int i = 0; unsigned int j = 0; unsigned int max; + uint16_t max_wqe; unsigned int comp; unsigned int inline_room = txq->max_inline * RTE_CACHE_LINE_SIZE; struct mlx5_mpw mpw = { .state = MLX5_MPW_STATE_CLOSED, }; + /* + * Compute the maximum number of WQE which can be consumed by inline + * code. + * - 2 DSEG for: + * - 1 control segment, + * - 1 Ethernet segment, + * - N Dseg from the inline request. + */ + const unsigned int wqe_inl_n = + ((2 * MLX5_WQE_DWORD_SIZE + + txq->max_inline * RTE_CACHE_LINE_SIZE) + + RTE_CACHE_LINE_SIZE - 1) / RTE_CACHE_LINE_SIZE; if (unlikely(!pkts_n)) return 0; /* Prefetch first packet cacheline. */ - tx_prefetch_cqe(txq, txq->cq_ci); - tx_prefetch_wqe(txq, txq->wqe_ci); - tx_prefetch_wqe(txq, txq->wqe_ci + 1); + 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)); @@ -944,6 +1232,11 @@ mlx5_tx_burst_mpw_inline(void *dpdk_txq, struct rte_mbuf **pkts, break; max -= segs_n; --pkts_n; + /* + * Compute max_wqe in case less WQE were consumed in previous + * iteration. + */ + max_wqe = (1u << txq->wqe_n) - (txq->wqe_ci - txq->wqe_pi); /* Should we enable HW CKSUM offload */ if (buf->ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM)) @@ -969,9 +1262,20 @@ mlx5_tx_burst_mpw_inline(void *dpdk_txq, struct rte_mbuf **pkts, if (mpw.state == MLX5_MPW_STATE_CLOSED) { if ((segs_n != 1) || (length > inline_room)) { + /* + * 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; } else { + if (unlikely(max_wqe < wqe_inl_n)) + break; + max_wqe -= wqe_inl_n; mlx5_mpw_inline_new(txq, &mpw, length); mpw.wqe->eseg.cs_flags = cs_flags; } @@ -1019,14 +1323,15 @@ mlx5_tx_burst_mpw_inline(void *dpdk_txq, struct rte_mbuf **pkts, addr = rte_pktmbuf_mtod(buf, uintptr_t); (*txq->elts)[elts_head] = buf; /* Maximum number of bytes before wrapping. */ - max = ((uintptr_t)&(*txq->wqes)[1 << txq->wqe_n] - + max = ((((uintptr_t)(txq->wqes)) + + (1 << txq->wqe_n) * + MLX5_WQE_SIZE) - (uintptr_t)mpw.data.raw); if (length > max) { rte_memcpy((void *)(uintptr_t)mpw.data.raw, (void *)addr, max); - mpw.data.raw = - (volatile void *)&(*txq->wqes)[0]; + mpw.data.raw = (volatile void *)txq->wqes; rte_memcpy((void *)(uintptr_t)mpw.data.raw, (void *)(addr + max), length - max); @@ -1035,12 +1340,13 @@ mlx5_tx_burst_mpw_inline(void *dpdk_txq, struct rte_mbuf **pkts, rte_memcpy((void *)(uintptr_t)mpw.data.raw, (void *)addr, length); - mpw.data.raw += length; + + if (length == max) + mpw.data.raw = + (volatile void *)txq->wqes; + else + mpw.data.raw += length; } - if ((uintptr_t)mpw.data.raw == - (uintptr_t)&(*txq->wqes)[1 << txq->wqe_n]) - mpw.data.raw = - (volatile void *)&(*txq->wqes)[0]; ++mpw.pkts_n; mpw.total_len += length; ++j; @@ -1090,6 +1396,360 @@ mlx5_tx_burst_mpw_inline(void *dpdk_txq, struct rte_mbuf **pkts, return i; } +/** + * Open an Enhanced MPW session. + * + * @param txq + * Pointer to TX queue structure. + * @param mpw + * Pointer to MPW session structure. + * @param length + * Packet length. + */ +static inline void +mlx5_empw_new(struct txq *txq, struct mlx5_mpw *mpw, int padding) +{ + uint16_t idx = txq->wqe_ci & ((1 << txq->wqe_n) - 1); + + mpw->state = MLX5_MPW_ENHANCED_STATE_OPENED; + mpw->pkts_n = 0; + mpw->total_len = sizeof(struct mlx5_wqe); + mpw->wqe = (volatile struct mlx5_wqe *)tx_mlx5_wqe(txq, idx); + mpw->wqe->ctrl[0] = htonl((MLX5_OPC_MOD_ENHANCED_MPSW << 24) | + (txq->wqe_ci << 8) | + MLX5_OPCODE_ENHANCED_MPSW); + mpw->wqe->ctrl[2] = 0; + mpw->wqe->ctrl[3] = 0; + memset((void *)(uintptr_t)&mpw->wqe->eseg, 0, MLX5_WQE_DWORD_SIZE); + if (unlikely(padding)) { + uintptr_t addr = (uintptr_t)(mpw->wqe + 1); + + /* Pad the first 2 DWORDs with zero-length inline header. */ + *(volatile uint32_t *)addr = htonl(MLX5_INLINE_SEG); + *(volatile uint32_t *)(addr + MLX5_WQE_DWORD_SIZE) = + htonl(MLX5_INLINE_SEG); + mpw->total_len += 2 * MLX5_WQE_DWORD_SIZE; + /* Start from the next WQEBB. */ + mpw->data.raw = (volatile void *)(tx_mlx5_wqe(txq, idx + 1)); + } else { + mpw->data.raw = (volatile void *)(mpw->wqe + 1); + } +} + +/** + * Close an Enhanced MPW session. + * + * @param txq + * Pointer to TX queue structure. + * @param mpw + * Pointer to MPW session structure. + * + * @return + * Number of consumed WQEs. + */ +static inline uint16_t +mlx5_empw_close(struct txq *txq, struct mlx5_mpw *mpw) +{ + uint16_t ret; + + /* Store size in multiple of 16 bytes. Control and Ethernet segments + * count as 2. + */ + mpw->wqe->ctrl[1] = htonl(txq->qp_num_8s | MLX5_WQE_DS(mpw->total_len)); + mpw->state = MLX5_MPW_STATE_CLOSED; + ret = (mpw->total_len + (MLX5_WQE_SIZE - 1)) / MLX5_WQE_SIZE; + txq->wqe_ci += ret; + return ret; +} + +/** + * DPDK callback for TX with Enhanced 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_empw(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_elts; + uint16_t max_wqe; + unsigned int max_inline = txq->max_inline * RTE_CACHE_LINE_SIZE; + unsigned int mpw_room = 0; + unsigned int inl_pad = 0; + uint32_t inl_hdr; + struct mlx5_mpw mpw = { + .state = MLX5_MPW_STATE_CLOSED, + }; + + if (unlikely(!pkts_n)) + return 0; + /* Start processing. */ + txq_complete(txq); + max_elts = (elts_n - (elts_head - txq->elts_tail)); + if (max_elts > elts_n) + max_elts -= elts_n; + /* A CQE slot must always be available. */ + assert((1u << txq->cqe_n) - (txq->cq_pi - txq->cq_ci)); + 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; + uintptr_t addr; + uint64_t naddr; + unsigned int n; + unsigned int do_inline = 0; /* Whether inline is possible. */ + 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_elts - j < segs_n + 1) + break; + /* Do not bother with large packets MPW cannot handle. */ + if (segs_n > MLX5_MPW_DSEG_MAX) + break; + /* 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); + /* Start new session if: + * - multi-segment packet + * - no space left even for a dseg + * - next packet can be inlined with a new WQE + * - cs_flag differs + * It can't be MLX5_MPW_STATE_OPENED as always have a single + * segmented packet. + */ + if (mpw.state == MLX5_MPW_ENHANCED_STATE_OPENED) { + if ((segs_n != 1) || + (inl_pad + sizeof(struct mlx5_wqe_data_seg) > + mpw_room) || + (length <= txq->inline_max_packet_sz && + inl_pad + sizeof(inl_hdr) + length > + mpw_room) || + (mpw.wqe->eseg.cs_flags != cs_flags)) + max_wqe -= mlx5_empw_close(txq, &mpw); + } + if (unlikely(mpw.state == MLX5_MPW_STATE_CLOSED)) { + if (unlikely(segs_n != 1)) { + /* Fall back to legacy MPW. + * A MPW session consumes 2 WQEs at most to + * include MLX5_MPW_DSEG_MAX pointers. + */ + if (unlikely(max_wqe < 2)) + break; + mlx5_mpw_new(txq, &mpw, length); + } else { + /* In Enhanced MPW, inline as much as the budget + * is allowed. The remaining space is to be + * filled with dsegs. If the title WQEBB isn't + * padded, it will have 2 dsegs there. + */ + mpw_room = RTE_MIN(MLX5_WQE_SIZE_MAX, + (max_inline ? max_inline : + pkts_n * MLX5_WQE_DWORD_SIZE) + + MLX5_WQE_SIZE); + if (unlikely(max_wqe * MLX5_WQE_SIZE < + mpw_room)) + break; + /* Don't pad the title WQEBB to not waste WQ. */ + mlx5_empw_new(txq, &mpw, 0); + mpw_room -= mpw.total_len; + inl_pad = 0; + do_inline = + length <= txq->inline_max_packet_sz && + sizeof(inl_hdr) + length <= mpw_room && + !txq->mpw_hdr_dseg; + } + mpw.wqe->eseg.cs_flags = cs_flags; + } else { + /* Evaluate whether the next packet can be inlined. + * Inlininig is possible when: + * - length is less than configured value + * - length fits for remaining space + * - not required to fill the title WQEBB with dsegs + */ + do_inline = + length <= txq->inline_max_packet_sz && + inl_pad + sizeof(inl_hdr) + length <= + mpw_room && + (!txq->mpw_hdr_dseg || + mpw.total_len >= MLX5_WQE_SIZE); + } + /* Multi-segment packets must be alone in their MPW. */ + assert((segs_n == 1) || (mpw.pkts_n == 0)); + if (unlikely(mpw.state == MLX5_MPW_STATE_OPENED)) { +#if defined(MLX5_PMD_SOFT_COUNTERS) || !defined(NDEBUG) + length = 0; +#endif + do { + volatile struct mlx5_wqe_data_seg *dseg; + + 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; + ++j; + ++mpw.pkts_n; + } while (--segs_n); + /* A multi-segmented packet takes one MPW session. + * TODO: Pack more multi-segmented packets if possible. + */ + mlx5_mpw_close(txq, &mpw); + if (mpw.pkts_n < 3) + max_wqe--; + else + max_wqe -= 2; + } else if (do_inline) { + /* Inline packet into WQE. */ + unsigned int max; + + assert(mpw.state == MLX5_MPW_ENHANCED_STATE_OPENED); + assert(length == DATA_LEN(buf)); + inl_hdr = htonl(length | MLX5_INLINE_SEG); + addr = rte_pktmbuf_mtod(buf, uintptr_t); + mpw.data.raw = (volatile void *) + ((uintptr_t)mpw.data.raw + inl_pad); + max = tx_mlx5_wq_tailroom(txq, + (void *)(uintptr_t)mpw.data.raw); + /* Copy inline header. */ + mpw.data.raw = (volatile void *) + mlx5_copy_to_wq( + (void *)(uintptr_t)mpw.data.raw, + &inl_hdr, + sizeof(inl_hdr), + (void *)(uintptr_t)txq->wqes, + max); + max = tx_mlx5_wq_tailroom(txq, + (void *)(uintptr_t)mpw.data.raw); + /* Copy packet data. */ + mpw.data.raw = (volatile void *) + mlx5_copy_to_wq( + (void *)(uintptr_t)mpw.data.raw, + (void *)addr, + length, + (void *)(uintptr_t)txq->wqes, + max); + ++mpw.pkts_n; + mpw.total_len += (inl_pad + sizeof(inl_hdr) + length); + /* No need to get completion as the entire packet is + * copied to WQ. Free the buf right away. + */ + elts_head_next = elts_head; + rte_pktmbuf_free_seg(buf); + mpw_room -= (inl_pad + sizeof(inl_hdr) + length); + /* Add pad in the next packet if any. */ + inl_pad = (((uintptr_t)mpw.data.raw + + (MLX5_WQE_DWORD_SIZE - 1)) & + ~(MLX5_WQE_DWORD_SIZE - 1)) - + (uintptr_t)mpw.data.raw; + } else { + /* No inline. Load a dseg of packet pointer. */ + volatile rte_v128u32_t *dseg; + + assert(mpw.state == MLX5_MPW_ENHANCED_STATE_OPENED); + assert((inl_pad + sizeof(*dseg)) <= mpw_room); + assert(length == DATA_LEN(buf)); + if (!tx_mlx5_wq_tailroom(txq, + (void *)((uintptr_t)mpw.data.raw + + inl_pad))) + dseg = (volatile void *)txq->wqes; + else + dseg = (volatile void *) + ((uintptr_t)mpw.data.raw + + inl_pad); + elts_head_next = (elts_head + 1) & (elts_n - 1); + (*txq->elts)[elts_head] = buf; + addr = rte_pktmbuf_mtod(buf, uintptr_t); + for (n = 0; n * RTE_CACHE_LINE_SIZE < length; n++) + rte_prefetch2((void *)(addr + + n * RTE_CACHE_LINE_SIZE)); + naddr = htonll(addr); + *dseg = (rte_v128u32_t) { + htonl(length), + txq_mp2mr(txq, txq_mb2mp(buf)), + naddr, + naddr >> 32, + }; + mpw.data.raw = (volatile void *)(dseg + 1); + mpw.total_len += (inl_pad + sizeof(*dseg)); + ++j; + ++mpw.pkts_n; + mpw_room -= (inl_pad + sizeof(*dseg)); + inl_pad = 0; + } + elts_head = elts_head_next; +#ifdef MLX5_PMD_SOFT_COUNTERS + /* Increment sent bytes counter. */ + txq->stats.obytes += length; +#endif + ++i; + } while (i < pkts_n); + /* Take a shortcut if nothing must be sent. */ + if (unlikely(i == 0)) + return 0; + /* Check whether completion threshold has been reached. */ + if (txq->elts_comp + j >= MLX5_TX_COMP_THRESH || + (uint16_t)(txq->wqe_ci - txq->mpw_comp) >= + (1 << txq->wqe_n) / MLX5_TX_COMP_THRESH_INLINE_DIV) { + 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; + txq->mpw_comp = txq->wqe_ci; + txq->cq_pi++; + } else { + txq->elts_comp += j; + } +#ifdef MLX5_PMD_SOFT_COUNTERS + /* Increment sent packets counter. */ + txq->stats.opackets += i; +#endif + if (mpw.state == MLX5_MPW_ENHANCED_STATE_OPENED) + mlx5_empw_close(txq, &mpw); + else if (mpw.state == MLX5_MPW_STATE_OPENED) + mlx5_mpw_close(txq, &mpw); + /* Ring QP doorbell. */ + mlx5_tx_dbrec(txq, mpw.wqe); + txq->elts_head = elts_head; + return i; +} + /** * Translate RX completion flags to packet type. * @@ -1153,6 +1813,7 @@ mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe, struct rxq_zip *zip = &rxq->zip; uint16_t cqe_n = cqe_cnt + 1; int len = 0; + uint16_t idx, end; /* Process compressed data in the CQE and mini arrays. */ if (zip->ai) { @@ -1163,6 +1824,14 @@ mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe, len = ntohl((*mc)[zip->ai & 7].byte_cnt); *rss_hash = ntohl((*mc)[zip->ai & 7].rx_hash_result); if ((++zip->ai & 7) == 0) { + /* Invalidate consumed CQEs */ + idx = zip->ca; + end = zip->na; + while (idx != end) { + (*rxq->cqes)[idx & cqe_cnt].op_own = + MLX5_CQE_INVALIDATE; + ++idx; + } /* * Increment consumer index to skip the number of * CQEs consumed. Hardware leaves holes in the CQ @@ -1172,8 +1841,9 @@ mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe, zip->na += 8; } if (unlikely(rxq->zip.ai == rxq->zip.cqe_cnt)) { - uint16_t idx = rxq->cq_ci; - uint16_t end = zip->cq_ci; + /* Invalidate the rest */ + idx = zip->ca; + end = zip->cq_ci; while (idx != end) { (*rxq->cqes)[idx & cqe_cnt].op_own = @@ -1209,7 +1879,7 @@ mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe, * special case the second one is located 7 CQEs after * the initial CQE instead of 8 for subsequent ones. */ - zip->ca = rxq->cq_ci & cqe_cnt; + zip->ca = rxq->cq_ci; zip->na = zip->ca + 7; /* Compute the next non compressed CQE. */ --rxq->cq_ci; @@ -1218,6 +1888,13 @@ mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe, len = ntohl((*mc)[0].byte_cnt); *rss_hash = ntohl((*mc)[0].rx_hash_result); zip->ai = 1; + /* Prefetch all the entries to be invalidated */ + idx = zip->ca; + end = zip->cq_ci; + while (idx != end) { + rte_prefetch0(&(*rxq->cqes)[(idx) & cqe_cnt]); + ++idx; + } } else { len = ntohl(cqe->byte_cnt); *rss_hash = ntohl(cqe->rx_hash_res); @@ -1290,7 +1967,7 @@ mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) &(*rxq->cqes)[rxq->cq_ci & cqe_cnt]; unsigned int i = 0; unsigned int rq_ci = rxq->rq_ci << sges_n; - int len; /* keep its value across iterations. */ + int len = 0; /* keep its value across iterations. */ while (pkts_n) { unsigned int idx = rq_ci & wqe_cnt; @@ -1317,8 +1994,9 @@ mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) while (pkt != seg) { assert(pkt != (*rxq->elts)[idx]); rep = NEXT(pkt); - rte_mbuf_refcnt_set(pkt, 0); - __rte_mbuf_raw_free(pkt); + NEXT(pkt) = NULL; + NB_SEGS(pkt) = 1; + rte_mbuf_raw_free(pkt); pkt = rep; } break; @@ -1328,14 +2006,12 @@ mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) len = mlx5_rx_poll_len(rxq, cqe, cqe_cnt, &rss_hash_res); if (!len) { - rte_mbuf_refcnt_set(rep, 0); - __rte_mbuf_raw_free(rep); + rte_mbuf_raw_free(rep); break; } if (unlikely(len == -1)) { /* RX error, packet is likely too large. */ - rte_mbuf_refcnt_set(rep, 0); - __rte_mbuf_raw_free(rep); + rte_mbuf_raw_free(rep); ++rxq->stats.idropped; goto skip; } @@ -1348,23 +2024,31 @@ mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) pkt->hash.rss = rss_hash_res; pkt->ol_flags = PKT_RX_RSS_HASH; } - if (rxq->csum | rxq->csum_l2tun | rxq->vlan_strip | - rxq->crc_present) { - if (rxq->csum) { - pkt->packet_type = - rxq_cq_to_pkt_type(cqe); - pkt->ol_flags |= - rxq_cq_to_ol_flags(rxq, cqe); - } - if (cqe->hdr_type_etc & - MLX5_CQE_VLAN_STRIPPED) { - pkt->ol_flags |= PKT_RX_VLAN_PKT | - PKT_RX_VLAN_STRIPPED; - pkt->vlan_tci = ntohs(cqe->vlan_info); + if (rxq->mark && + MLX5_FLOW_MARK_IS_VALID(cqe->sop_drop_qpn)) { + pkt->ol_flags |= PKT_RX_FDIR; + if (cqe->sop_drop_qpn != + htonl(MLX5_FLOW_MARK_DEFAULT)) { + uint32_t mark = cqe->sop_drop_qpn; + + pkt->ol_flags |= PKT_RX_FDIR_ID; + pkt->hash.fdir.hi = + mlx5_flow_mark_get(mark); } - if (rxq->crc_present) - len -= ETHER_CRC_LEN; } + if (rxq->csum | rxq->csum_l2tun) { + pkt->packet_type = rxq_cq_to_pkt_type(cqe); + pkt->ol_flags |= rxq_cq_to_ol_flags(rxq, cqe); + } + if (rxq->vlan_strip && + (cqe->hdr_type_etc & + htons(MLX5_CQE_VLAN_STRIPPED))) { + pkt->ol_flags |= PKT_RX_VLAN_PKT | + PKT_RX_VLAN_STRIPPED; + pkt->vlan_tci = ntohs(cqe->vlan_info); + } + if (rxq->crc_present) + len -= ETHER_CRC_LEN; PKT_LEN(pkt) = len; } DATA_LEN(rep) = DATA_LEN(seg); @@ -1466,3 +2150,76 @@ removed_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) (void)pkts_n; return 0; } + +/** + * DPDK callback for rx queue interrupt enable. + * + * @param dev + * Pointer to Ethernet device structure. + * @param rx_queue_id + * RX queue number + * + * @return + * 0 on success, negative on failure. + */ +int +mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id) +{ +#ifdef HAVE_UPDATE_CQ_CI + struct priv *priv = mlx5_get_priv(dev); + struct rxq *rxq = (*priv->rxqs)[rx_queue_id]; + struct rxq_ctrl *rxq_ctrl = container_of(rxq, struct rxq_ctrl, rxq); + struct ibv_cq *cq = rxq_ctrl->cq; + uint16_t ci = rxq->cq_ci; + int ret = 0; + + ibv_mlx5_exp_update_cq_ci(cq, ci); + ret = ibv_req_notify_cq(cq, 0); +#else + int ret = -1; + (void)dev; + (void)rx_queue_id; +#endif + if (ret) + WARN("unable to arm interrupt on rx queue %d", rx_queue_id); + return ret; +} + +/** + * DPDK callback for rx queue interrupt disable. + * + * @param dev + * Pointer to Ethernet device structure. + * @param rx_queue_id + * RX queue number + * + * @return + * 0 on success, negative on failure. + */ +int +mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id) +{ +#ifdef HAVE_UPDATE_CQ_CI + struct priv *priv = mlx5_get_priv(dev); + struct rxq *rxq = (*priv->rxqs)[rx_queue_id]; + struct rxq_ctrl *rxq_ctrl = container_of(rxq, struct rxq_ctrl, rxq); + struct ibv_cq *cq = rxq_ctrl->cq; + struct ibv_cq *ev_cq; + void *ev_ctx; + int ret = 0; + + ret = ibv_get_cq_event(cq->channel, &ev_cq, &ev_ctx); + if (ret || ev_cq != cq) + ret = -1; + else + ibv_ack_cq_events(cq, 1); +#else + int ret = -1; + (void)dev; + (void)rx_queue_id; +#endif + if (ret) + WARN("unable to disable interrupt on rx queue %d", + rx_queue_id); + return ret; +} -- cgit 1.2.3-korg