X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx5%2Fmlx5_rxtx.h;h=d4c29cc128eef8ea5f0fc311b137e1c550fd1384;hb=f8244c6399d9;hp=8db8eb144dc8da7a13c6b1e262c512065b9c139c;hpb=70280a3e33d9b699bb31f16854fed6270fca4e59;p=dpdk.git diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h index 8db8eb144d..d4c29cc128 100644 --- a/drivers/net/mlx5/mlx5_rxtx.h +++ b/drivers/net/mlx5/mlx5_rxtx.h @@ -43,21 +43,15 @@ #pragma GCC diagnostic ignored "-Wpedantic" #endif #include -#include +#include #ifdef PEDANTIC #pragma GCC diagnostic error "-Wpedantic" #endif -/* DPDK headers don't like -pedantic. */ -#ifdef PEDANTIC -#pragma GCC diagnostic ignored "-Wpedantic" -#endif #include #include #include -#ifdef PEDANTIC -#pragma GCC diagnostic error "-Wpedantic" -#endif +#include #include "mlx5_utils.h" #include "mlx5.h" @@ -81,14 +75,14 @@ struct mlx5_txq_stats { uint64_t opackets; /**< Total of successfully sent packets. */ uint64_t obytes; /**< Total of successfully sent bytes. */ #endif - uint64_t odropped; /**< Total of packets not sent when TX ring full. */ + uint64_t oerrors; /**< Total number of failed transmitted packets. */ }; /* Flow director queue structure. */ struct fdir_queue { struct ibv_qp *qp; /* Associated RX QP. */ - struct ibv_exp_rwq_ind_table *ind_table; /* Indirection table. */ - struct ibv_exp_wq *wq; /* Work queue. */ + struct ibv_rwq_ind_table *ind_table; /* Indirection table. */ + struct ibv_wq *wq; /* Work queue. */ struct ibv_cq *cq; /* Completion queue. */ }; @@ -112,13 +106,15 @@ struct rxq { unsigned int sges_n:2; /* Log 2 of SGEs (max buffers per packet). */ unsigned int cqe_n:4; /* Log 2 of CQ elements. */ unsigned int elts_n:4; /* Log 2 of Mbufs. */ - unsigned int port_id:8; unsigned int rss_hash:1; /* RSS hash result is enabled. */ unsigned int mark:1; /* Marked flow available on the queue. */ - unsigned int :8; /* Remaining bits. */ + unsigned int pending_err:1; /* CQE error needs to be handled. */ + unsigned int :15; /* Remaining bits. */ volatile uint32_t *rq_db; volatile uint32_t *cq_db; + uint16_t port_id; uint16_t rq_ci; + uint16_t rq_pi; uint16_t cq_ci; volatile struct mlx5_wqe_data_seg(*wqes)[]; volatile struct mlx5_cqe(*cqes)[]; @@ -126,13 +122,18 @@ struct rxq { struct rte_mbuf *(*elts)[]; struct rte_mempool *mp; struct mlx5_rxq_stats stats; + uint64_t mbuf_initializer; /* Default rearm_data for vectorized Rx. */ + struct rte_mbuf fake_mbuf; /* elts padding for vectorized Rx. */ + void *cq_uar; /* CQ user access region. */ + uint32_t cqn; /* CQ number. */ + uint8_t cq_arm_sn; /* CQ arm seq number. */ } __rte_cache_aligned; /* RX queue control descriptor. */ struct rxq_ctrl { struct priv *priv; /* Back pointer to private data. */ struct ibv_cq *cq; /* Completion Queue. */ - struct ibv_exp_wq *wq; /* Work Queue. */ + struct ibv_wq *wq; /* Work Queue. */ struct fdir_queue *fdir_queue; /* Flow director queue. */ struct ibv_mr *mr; /* Memory Region (for mp). */ struct ibv_comp_channel *channel; @@ -154,8 +155,8 @@ enum hash_rxq_type { /* Flow structure with Ethernet specification. It is packed to prevent padding * between attr and spec as this layout is expected by libibverbs. */ struct flow_attr_spec_eth { - struct ibv_exp_flow_attr attr; - struct ibv_exp_flow_spec_eth spec; + struct ibv_flow_attr attr; + struct ibv_flow_spec_eth spec; } __attribute__((packed)); /* Define a struct flow_attr_spec_eth object as an array of at least @@ -173,13 +174,13 @@ struct hash_rxq_init { unsigned int flow_priority; /* Flow priority to use. */ union { struct { - enum ibv_exp_flow_spec_type type; + enum ibv_flow_spec_type type; uint16_t size; } hdr; - struct ibv_exp_flow_spec_tcp_udp tcp_udp; - struct ibv_exp_flow_spec_ipv4 ipv4; - struct ibv_exp_flow_spec_ipv6 ipv6; - struct ibv_exp_flow_spec_eth eth; + struct ibv_flow_spec_tcp_udp tcp_udp; + struct ibv_flow_spec_ipv4 ipv4; + struct ibv_flow_spec_ipv6 ipv6; + struct ibv_flow_spec_eth eth; } flow_spec; /* Flow specification template. */ const struct hash_rxq_init *underlayer; /* Pointer to underlayer. */ }; @@ -233,17 +234,17 @@ struct hash_rxq { struct ibv_qp *qp; /* Hash RX QP. */ enum hash_rxq_type type; /* Hash RX queue type. */ /* MAC flow steering rules, one per VLAN ID. */ - struct ibv_exp_flow *mac_flow + struct ibv_flow *mac_flow [MLX5_MAX_MAC_ADDRESSES][MLX5_MAX_VLAN_IDS]; - struct ibv_exp_flow *special_flow + struct ibv_flow *special_flow [MLX5_MAX_SPECIAL_FLOWS][MLX5_MAX_VLAN_IDS]; }; /* TX queue descriptor. */ -RTE_STD_C11 +__extension__ struct txq { - uint16_t elts_head; /* Current index in (*elts)[]. */ - uint16_t elts_tail; /* First element awaiting completion. */ + uint16_t elts_head; /* Current counter in (*elts)[]. */ + uint16_t elts_tail; /* Counter of first element awaiting completion. */ uint16_t elts_comp; /* Counter since last completion request. */ uint16_t mpw_comp; /* WQ index since last completion request. */ uint16_t cq_ci; /* Consumer index for completion queue. */ @@ -261,16 +262,19 @@ struct txq { uint16_t max_inline; /* Multiple of RTE_CACHE_LINE_SIZE to inline. */ uint16_t inline_max_packet_sz; /* Max packet size for inlining. */ uint32_t qp_num_8s; /* QP number shifted by 8. */ + uint32_t flags; /* Flags for Tx Queue. */ volatile struct mlx5_cqe (*cqes)[]; /* Completion queue. */ volatile void *wqes; /* Work queue (use volatile to write into). */ volatile uint32_t *qp_db; /* Work queue doorbell. */ volatile uint32_t *cq_db; /* Completion queue doorbell. */ volatile void *bf_reg; /* Blueflame register. */ struct { - const struct rte_mempool *mp; /* Cached Memory Pool. */ + uintptr_t start; /* Start address of MR */ + uintptr_t end; /* End address of MR */ struct ibv_mr *mr; /* Memory Region (for mp). */ - uint32_t lkey; /* htonl(mr->lkey) */ + uint32_t lkey; /* rte_cpu_to_be_32(mr->lkey) */ } mp2mr[MLX5_PMD_TX_MP_CACHE]; /* MP to MR translation table. */ + uint16_t mr_cache_idx; /* Index of last hit entry. */ struct rte_mbuf *(*elts)[]; /* TX elements. */ struct mlx5_txq_stats stats; /* TX queue counters. */ } __rte_cache_aligned; @@ -292,25 +296,20 @@ extern const unsigned int hash_rxq_init_n; extern uint8_t rss_hash_default_key[]; extern const size_t rss_hash_default_key_len; -size_t priv_flow_attr(struct priv *, struct ibv_exp_flow_attr *, +size_t priv_flow_attr(struct priv *, struct ibv_flow_attr *, size_t, enum hash_rxq_type); int priv_create_hash_rxqs(struct priv *); void priv_destroy_hash_rxqs(struct priv *); int priv_allow_flow_type(struct priv *, enum hash_rxq_flow_type); int priv_rehash_flows(struct priv *); -int priv_intr_efd_enable(struct priv *priv); -void priv_intr_efd_disable(struct priv *priv); -int priv_create_intr_vec(struct priv *priv); -void priv_destroy_intr_vec(struct priv *priv); void rxq_cleanup(struct rxq_ctrl *); -int rxq_rehash(struct rte_eth_dev *, struct rxq_ctrl *); -int rxq_ctrl_setup(struct rte_eth_dev *, struct rxq_ctrl *, uint16_t, - unsigned int, const struct rte_eth_rxconf *, - struct rte_mempool *); int mlx5_rx_queue_setup(struct rte_eth_dev *, uint16_t, uint16_t, unsigned int, const struct rte_eth_rxconf *, struct rte_mempool *); void mlx5_rx_queue_release(void *); -uint16_t mlx5_rx_burst_secondary_setup(void *, struct rte_mbuf **, uint16_t); +int priv_rx_intr_vec_enable(struct priv *priv); +void priv_rx_intr_vec_disable(struct priv *priv); +int mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id); +int mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id); /* mlx5_txq.c */ @@ -320,10 +319,12 @@ int txq_ctrl_setup(struct rte_eth_dev *, struct txq_ctrl *, uint16_t, int mlx5_tx_queue_setup(struct rte_eth_dev *, uint16_t, uint16_t, unsigned int, const struct rte_eth_txconf *); void mlx5_tx_queue_release(void *); -uint16_t mlx5_tx_burst_secondary_setup(void *, struct rte_mbuf **, uint16_t); /* mlx5_rxtx.c */ +extern uint32_t mlx5_ptype_table[]; + +void mlx5_set_ptype_table(void); uint16_t mlx5_tx_burst(void *, struct rte_mbuf **, uint16_t); uint16_t mlx5_tx_burst_mpw(void *, struct rte_mbuf **, uint16_t); uint16_t mlx5_tx_burst_mpw_inline(void *, struct rte_mbuf **, uint16_t); @@ -333,8 +334,15 @@ uint16_t removed_tx_burst(void *, struct rte_mbuf **, uint16_t); uint16_t removed_rx_burst(void *, struct rte_mbuf **, uint16_t); int mlx5_rx_descriptor_status(void *, uint16_t); int mlx5_tx_descriptor_status(void *, uint16_t); -int mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id); -int mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id); + +/* Vectorized version of mlx5_rxtx.c */ +int priv_check_raw_vec_tx_support(struct priv *); +int priv_check_vec_tx_support(struct priv *); +int rxq_check_vec_support(struct rxq *); +int priv_check_vec_rx_support(struct priv *); +uint16_t mlx5_tx_burst_raw_vec(void *, struct rte_mbuf **, uint16_t); +uint16_t mlx5_tx_burst_vec(void *, struct rte_mbuf **, uint16_t); +uint16_t mlx5_rx_burst_vec(void *, struct rte_mbuf **, uint16_t); /* mlx5_mr.c */ @@ -342,4 +350,267 @@ struct ibv_mr *mlx5_mp2mr(struct ibv_pd *, struct rte_mempool *); void txq_mp2mr_iter(struct rte_mempool *, void *); uint32_t txq_mp2mr_reg(struct txq *, struct rte_mempool *, unsigned int); +#ifndef NDEBUG +/** + * Verify or set magic value in CQE. + * + * @param cqe + * Pointer to CQE. + * + * @return + * 0 the first time. + */ +static inline int +check_cqe_seen(volatile struct mlx5_cqe *cqe) +{ + static const uint8_t magic[] = "seen"; + volatile uint8_t (*buf)[sizeof(cqe->rsvd0)] = &cqe->rsvd0; + int ret = 1; + unsigned int i; + + for (i = 0; i < sizeof(magic) && i < sizeof(*buf); ++i) + if (!ret || (*buf)[i] != magic[i]) { + ret = 0; + (*buf)[i] = magic[i]; + } + return ret; +} +#endif /* NDEBUG */ + +/** + * Check whether CQE is valid. + * + * @param cqe + * Pointer to CQE. + * @param cqes_n + * Size of completion queue. + * @param ci + * Consumer index. + * + * @return + * 0 on success, 1 on failure. + */ +static __rte_always_inline int +check_cqe(volatile struct mlx5_cqe *cqe, + unsigned int cqes_n, const uint16_t ci) +{ + uint16_t idx = ci & cqes_n; + uint8_t op_own = cqe->op_own; + uint8_t op_owner = MLX5_CQE_OWNER(op_own); + uint8_t op_code = MLX5_CQE_OPCODE(op_own); + + if (unlikely((op_owner != (!!(idx))) || (op_code == MLX5_CQE_INVALID))) + return 1; /* No CQE. */ +#ifndef NDEBUG + if ((op_code == MLX5_CQE_RESP_ERR) || + (op_code == MLX5_CQE_REQ_ERR)) { + volatile struct mlx5_err_cqe *err_cqe = (volatile void *)cqe; + uint8_t syndrome = err_cqe->syndrome; + + if ((syndrome == MLX5_CQE_SYNDROME_LOCAL_LENGTH_ERR) || + (syndrome == MLX5_CQE_SYNDROME_REMOTE_ABORTED_ERR)) + return 0; + if (!check_cqe_seen(cqe)) { + ERROR("unexpected CQE error %u (0x%02x)" + " syndrome 0x%02x", + op_code, op_code, syndrome); + rte_hexdump(stderr, "MLX5 Error CQE:", + (const void *)((uintptr_t)err_cqe), + sizeof(*err_cqe)); + } + return 1; + } else if ((op_code != MLX5_CQE_RESP_SEND) && + (op_code != MLX5_CQE_REQ)) { + if (!check_cqe_seen(cqe)) { + ERROR("unexpected CQE opcode %u (0x%02x)", + op_code, op_code); + rte_hexdump(stderr, "MLX5 CQE:", + (const void *)((uintptr_t)cqe), + sizeof(*cqe)); + } + return 1; + } +#endif /* NDEBUG */ + return 0; +} + +/** + * 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); +} + +/** + * Manage TX completions. + * + * When sending a burst, mlx5_tx_burst() posts several WRs. + * + * @param txq + * Pointer to TX queue structure. + */ +static __rte_always_inline void +mlx5_tx_complete(struct txq *txq) +{ + const uint16_t elts_n = 1 << txq->elts_n; + const uint16_t elts_m = elts_n - 1; + const unsigned int cqe_n = 1 << txq->cqe_n; + const unsigned int cqe_cnt = cqe_n - 1; + uint16_t elts_free = txq->elts_tail; + uint16_t elts_tail; + uint16_t cq_ci = txq->cq_ci; + volatile struct mlx5_cqe *cqe = NULL; + volatile struct mlx5_wqe_ctrl *ctrl; + struct rte_mbuf *m, *free[elts_n]; + struct rte_mempool *pool = NULL; + unsigned int blk_n = 0; + + cqe = &(*txq->cqes)[cq_ci & cqe_cnt]; + if (unlikely(check_cqe(cqe, cqe_n, cq_ci))) + return; +#ifndef NDEBUG + if ((MLX5_CQE_OPCODE(cqe->op_own) == MLX5_CQE_RESP_ERR) || + (MLX5_CQE_OPCODE(cqe->op_own) == MLX5_CQE_REQ_ERR)) { + if (!check_cqe_seen(cqe)) { + ERROR("unexpected error CQE, TX stopped"); + rte_hexdump(stderr, "MLX5 TXQ:", + (const void *)((uintptr_t)txq->wqes), + ((1 << txq->wqe_n) * + MLX5_WQE_SIZE)); + } + return; + } +#endif /* NDEBUG */ + ++cq_ci; + txq->wqe_pi = rte_be_to_cpu_16(cqe->wqe_counter); + ctrl = (volatile struct mlx5_wqe_ctrl *) + tx_mlx5_wqe(txq, txq->wqe_pi); + elts_tail = ctrl->ctrl3; + assert((elts_tail & elts_m) < (1 << txq->wqe_n)); + /* Free buffers. */ + while (elts_free != elts_tail) { + m = rte_pktmbuf_prefree_seg((*txq->elts)[elts_free++ & elts_m]); + if (likely(m != NULL)) { + if (likely(m->pool == pool)) { + free[blk_n++] = m; + } else { + if (likely(pool != NULL)) + rte_mempool_put_bulk(pool, + (void *)free, + blk_n); + free[0] = m; + pool = m->pool; + blk_n = 1; + } + } + } + if (blk_n) + rte_mempool_put_bulk(pool, (void *)free, blk_n); +#ifndef NDEBUG + elts_free = txq->elts_tail; + /* Poisoning. */ + while (elts_free != elts_tail) { + memset(&(*txq->elts)[elts_free & elts_m], + 0x66, + sizeof((*txq->elts)[elts_free & elts_m])); + ++elts_free; + } +#endif + txq->cq_ci = cq_ci; + txq->elts_tail = elts_tail; + /* Update the consumer index. */ + rte_wmb(); + *txq->cq_db = rte_cpu_to_be_32(cq_ci); +} + +/** + * Get Memory Pool (MP) from mbuf. If mbuf is indirect, the pool from which + * the cloned mbuf is allocated is returned instead. + * + * @param buf + * Pointer to mbuf. + * + * @return + * Memory pool where data is located for given mbuf. + */ +static struct rte_mempool * +mlx5_tx_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) <-> rte_mbuf 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 __rte_always_inline uint32_t +mlx5_tx_mb2mr(struct txq *txq, struct rte_mbuf *mb) +{ + uint16_t i = txq->mr_cache_idx; + uintptr_t addr = rte_pktmbuf_mtod(mb, uintptr_t); + + assert(i < RTE_DIM(txq->mp2mr)); + if (likely(txq->mp2mr[i].start <= addr && txq->mp2mr[i].end >= addr)) + return txq->mp2mr[i].lkey; + for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) { + if (unlikely(txq->mp2mr[i].mr == NULL)) { + /* Unknown MP, add a new MR for it. */ + break; + } + if (txq->mp2mr[i].start <= addr && + txq->mp2mr[i].end >= addr) { + assert(txq->mp2mr[i].lkey != (uint32_t)-1); + assert(rte_cpu_to_be_32(txq->mp2mr[i].mr->lkey) == + txq->mp2mr[i].lkey); + txq->mr_cache_idx = i; + return txq->mp2mr[i].lkey; + } + } + txq->mr_cache_idx = 0; + return txq_mp2mr_reg(txq, mlx5_tx_mb2mp(mb), i); +} + +/** + * Ring TX queue doorbell. + * + * @param txq + * Pointer to TX queue structure. + * @param wqe + * Pointer to the last WQE posted in the NIC. + */ +static __rte_always_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_io_wmb(); + *txq->qp_db = rte_cpu_to_be_32(txq->wqe_ci); + /* Ensure ordering between DB record and BF copy. */ + rte_wmb(); + *dst = *src; +} + #endif /* RTE_PMD_MLX5_RXTX_H_ */