/**
* Store free buffers to RX SW ring.
*
- * @param rxq
- * Pointer to RX queue structure.
+ * @param elts
+ * Pointer to SW ring to be filled.
* @param pkts
* Pointer to array of packets to be stored.
* @param pkts_n
* Number of packets to be stored.
*/
static inline void
-rxq_copy_mbuf_v(struct mlx5_rxq_data *rxq, struct rte_mbuf **pkts, uint16_t n)
+rxq_copy_mbuf_v(struct rte_mbuf **elts, struct rte_mbuf **pkts, uint16_t n)
{
- const uint16_t q_mask = (1 << rxq->elts_n) - 1;
- struct rte_mbuf **elts = &(*rxq->elts)[rxq->rq_pi & q_mask];
unsigned int pos;
uint16_t p = n & -2;
-1UL << ((mcqe_n - pos) *
sizeof(uint16_t) * 8) : 0);
#endif
+
for (i = 0; i < MLX5_VPMD_DESCS_PER_LOOP; ++i)
if (likely(pos + i < mcqe_n))
rte_prefetch0((void *)(cq + pos + i));
pos += MLX5_VPMD_DESCS_PER_LOOP;
/* Move to next CQE and invalidate consumed CQEs. */
if (!(pos & 0x7) && pos < mcqe_n) {
+ if (pos + 8 < mcqe_n)
+ rte_prefetch0((void *)(cq + pos + 8));
mcq = (void *)&(cq + pos)->pkt_info;
for (i = 0; i < 8; ++i)
cq[inv++].op_own = MLX5_CQE_INVALIDATE;
uint32x4_t pinfo, cv_flags;
uint32x4_t ol_flags =
vdupq_n_u32(rxq->rss_hash * PKT_RX_RSS_HASH |
- rxq->hw_timestamp * PKT_RX_TIMESTAMP);
+ rxq->hw_timestamp * rxq->timestamp_rx_flag);
const uint32x4_t ptype_ol_mask = { 0x106, 0x106, 0x106, 0x106 };
const uint8x16_t cv_flag_sel = {
0,
}
/**
- * Receive burst of packets. An errored completion also consumes a mbuf, but the
- * packet_type is set to be RTE_PTYPE_ALL_MASK. Marked mbufs should be freed
- * before returning to application.
+ * Process a non-compressed completion and fill in mbufs in RX SW ring
+ * with data extracted from the title completion descriptor.
*
* @param rxq
* Pointer to RX queue structure.
+ * @param cq
+ * Pointer to completion array having a non-compressed completion at first.
+ * @param elts
+ * Pointer to SW ring to be filled. The first mbuf has to be pre-built from
+ * the title completion descriptor to be copied to the rest of mbufs.
* @param[out] pkts
* Array to store received packets.
* @param pkts_n
* @param[out] err
* Pointer to a flag. Set non-zero value if pkts array has at least one error
* packet to handle.
- * @param[out] no_cq
- * Pointer to a boolean. Set true if no new CQE seen.
+ * @param[out] comp
+ * Pointer to a index. Set it to the first compressed completion if any.
*
* @return
- * Number of packets received including errors (<= pkts_n).
+ * Number of CQEs successfully processed.
*/
static inline uint16_t
-rxq_burst_v(struct mlx5_rxq_data *rxq, struct rte_mbuf **pkts, uint16_t pkts_n,
- uint64_t *err, bool *no_cq)
+rxq_cq_process_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
+ struct rte_mbuf **elts, struct rte_mbuf **pkts,
+ uint16_t pkts_n, uint64_t *err, uint64_t *comp)
{
const uint16_t q_n = 1 << rxq->cqe_n;
const uint16_t q_mask = q_n - 1;
- volatile struct mlx5_cqe *cq;
- struct rte_mbuf **elts;
unsigned int pos;
- uint64_t n;
- uint16_t repl_n;
+ uint64_t n = 0;
uint64_t comp_idx = MLX5_VPMD_DESCS_PER_LOOP;
uint16_t nocmp_n = 0;
- uint16_t rcvd_pkt = 0;
- unsigned int cq_idx = rxq->cq_ci & q_mask;
- unsigned int elts_idx;
const uint16x4_t ownership = vdup_n_u16(!(rxq->cq_ci & (q_mask + 1)));
const uint16x4_t owner_check = vcreate_u16(0x0001000100010001);
const uint16x4_t opcode_check = vcreate_u16(0x00f000f000f000f0);
};
const uint32x4_t flow_mark_adj = { 0, 0, 0, rxq->mark * (-1) };
- MLX5_ASSERT(rxq->sges_n == 0);
- MLX5_ASSERT(rxq->cqe_n == rxq->elts_n);
- cq = &(*rxq->cqes)[cq_idx];
- rte_prefetch_non_temporal(cq);
- rte_prefetch_non_temporal(cq + 1);
- rte_prefetch_non_temporal(cq + 2);
- rte_prefetch_non_temporal(cq + 3);
- pkts_n = RTE_MIN(pkts_n, MLX5_VPMD_RX_MAX_BURST);
- repl_n = q_n - (rxq->rq_ci - rxq->rq_pi);
- if (repl_n >= rxq->rq_repl_thresh)
- mlx5_rx_replenish_bulk_mbuf(rxq, repl_n);
- /* See if there're unreturned mbufs from compressed CQE. */
- rcvd_pkt = rxq->decompressed;
- if (rcvd_pkt > 0) {
- rcvd_pkt = RTE_MIN(rcvd_pkt, pkts_n);
- rxq_copy_mbuf_v(rxq, pkts, rcvd_pkt);
- rxq->rq_pi += rcvd_pkt;
- pkts += rcvd_pkt;
- rxq->decompressed -= rcvd_pkt;
- }
- elts_idx = rxq->rq_pi & q_mask;
- elts = &(*rxq->elts)[elts_idx];
- /* Not to overflow pkts array. */
- pkts_n = RTE_ALIGN_FLOOR(pkts_n - rcvd_pkt, MLX5_VPMD_DESCS_PER_LOOP);
- /* Not to cross queue end. */
- pkts_n = RTE_MIN(pkts_n, q_n - elts_idx);
- pkts_n = RTE_MIN(pkts_n, q_n - cq_idx);
- if (!pkts_n) {
- *no_cq = !rcvd_pkt;
- return rcvd_pkt;
- }
- /* At this point, there shouldn't be any remained packets. */
- MLX5_ASSERT(rxq->decompressed == 0);
/*
* Note that vectors have reverse order - {v3, v2, v1, v0}, because
* there's no instruction to count trailing zeros. __builtin_clzl() is
/* B.0 (CQE 0) load a block having op_own. */
c0 = vld1q_u64((uint64_t *)(p0 + 48));
/* Synchronize for loading the rest of blocks. */
- rte_cio_rmb();
+ rte_io_rmb();
/* Prefetch next 4 CQEs. */
if (pkts_n - pos >= 2 * MLX5_VPMD_DESCS_PER_LOOP) {
unsigned int next = pos + MLX5_VPMD_DESCS_PER_LOOP;
rxq_cq_to_ptype_oflags_v(rxq, ptype_info, flow_tag,
opcode, &elts[pos]);
if (rxq->hw_timestamp) {
+ int offset = rxq->timestamp_offset;
if (rxq->rt_timestamp) {
struct mlx5_dev_ctx_shared *sh = rxq->sh;
uint64_t ts;
ts = rte_be_to_cpu_64
(container_of(p0, struct mlx5_cqe,
pkt_info)->timestamp);
- elts[pos]->timestamp =
- mlx5_txpp_convert_rx_ts(sh, ts);
+ mlx5_timestamp_set(elts[pos], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
ts = rte_be_to_cpu_64
(container_of(p1, struct mlx5_cqe,
pkt_info)->timestamp);
- elts[pos + 1]->timestamp =
- mlx5_txpp_convert_rx_ts(sh, ts);
+ mlx5_timestamp_set(elts[pos + 1], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
ts = rte_be_to_cpu_64
(container_of(p2, struct mlx5_cqe,
pkt_info)->timestamp);
- elts[pos + 2]->timestamp =
- mlx5_txpp_convert_rx_ts(sh, ts);
+ mlx5_timestamp_set(elts[pos + 2], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
ts = rte_be_to_cpu_64
(container_of(p3, struct mlx5_cqe,
pkt_info)->timestamp);
- elts[pos + 3]->timestamp =
- mlx5_txpp_convert_rx_ts(sh, ts);
+ mlx5_timestamp_set(elts[pos + 3], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
} else {
- elts[pos]->timestamp = rte_be_to_cpu_64
- (container_of(p0, struct mlx5_cqe,
- pkt_info)->timestamp);
- elts[pos + 1]->timestamp = rte_be_to_cpu_64
- (container_of(p1, struct mlx5_cqe,
- pkt_info)->timestamp);
- elts[pos + 2]->timestamp = rte_be_to_cpu_64
- (container_of(p2, struct mlx5_cqe,
- pkt_info)->timestamp);
- elts[pos + 3]->timestamp = rte_be_to_cpu_64
- (container_of(p3, struct mlx5_cqe,
- pkt_info)->timestamp);
+ mlx5_timestamp_set(elts[pos], offset,
+ rte_be_to_cpu_64(container_of(p0,
+ struct mlx5_cqe, pkt_info)->timestamp));
+ mlx5_timestamp_set(elts[pos + 1], offset,
+ rte_be_to_cpu_64(container_of(p1,
+ struct mlx5_cqe, pkt_info)->timestamp));
+ mlx5_timestamp_set(elts[pos + 2], offset,
+ rte_be_to_cpu_64(container_of(p2,
+ struct mlx5_cqe, pkt_info)->timestamp));
+ mlx5_timestamp_set(elts[pos + 3], offset,
+ rte_be_to_cpu_64(container_of(p3,
+ struct mlx5_cqe, pkt_info)->timestamp));
}
}
- if (!!rxq->flow_meta_mask) {
+ if (rxq->dynf_meta) {
/* This code is subject for futher optimization. */
int32_t offs = rxq->flow_meta_offset;
*RTE_MBUF_DYNFIELD(pkts[pos], offs, uint32_t *) =
container_of(p0, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
- *RTE_MBUF_DYNFIELD(pkts[pos], offs, uint32_t *) =
+ *RTE_MBUF_DYNFIELD(pkts[pos + 1], offs, uint32_t *) =
container_of(p1, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
- *RTE_MBUF_DYNFIELD(pkts[pos], offs, uint32_t *) =
+ *RTE_MBUF_DYNFIELD(pkts[pos + 2], offs, uint32_t *) =
container_of(p2, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
- *RTE_MBUF_DYNFIELD(pkts[pos], offs, uint32_t *) =
+ *RTE_MBUF_DYNFIELD(pkts[pos + 3], offs, uint32_t *) =
container_of(p3, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
if (*RTE_MBUF_DYNFIELD(pkts[pos], offs, uint32_t *))
if (n != MLX5_VPMD_DESCS_PER_LOOP)
break;
}
- /* If no new CQE seen, return without updating cq_db. */
- if (unlikely(!nocmp_n && comp_idx == MLX5_VPMD_DESCS_PER_LOOP)) {
- *no_cq = true;
- return rcvd_pkt;
- }
- /* Update the consumer indexes for non-compressed CQEs. */
- MLX5_ASSERT(nocmp_n <= pkts_n);
- rxq->cq_ci += nocmp_n;
- rxq->rq_pi += nocmp_n;
- rcvd_pkt += nocmp_n;
#ifdef MLX5_PMD_SOFT_COUNTERS
rxq->stats.ipackets += nocmp_n;
rxq->stats.ibytes += rcvd_byte;
#endif
- /* Decompress the last CQE if compressed. */
- if (comp_idx < MLX5_VPMD_DESCS_PER_LOOP && comp_idx == n) {
- MLX5_ASSERT(comp_idx == (nocmp_n % MLX5_VPMD_DESCS_PER_LOOP));
- rxq->decompressed = rxq_cq_decompress_v(rxq, &cq[nocmp_n],
- &elts[nocmp_n]);
- /* Return more packets if needed. */
- if (nocmp_n < pkts_n) {
- uint16_t n = rxq->decompressed;
-
- n = RTE_MIN(n, pkts_n - nocmp_n);
- rxq_copy_mbuf_v(rxq, &pkts[nocmp_n], n);
- rxq->rq_pi += n;
- rcvd_pkt += n;
- rxq->decompressed -= n;
- }
- }
- rte_cio_wmb();
- *rxq->cq_db = rte_cpu_to_be_32(rxq->cq_ci);
- *no_cq = !rcvd_pkt;
- return rcvd_pkt;
+ if (comp_idx == n)
+ *comp = comp_idx;
+ return nocmp_n;
}
#endif /* RTE_PMD_MLX5_RXTX_VEC_NEON_H_ */
git.droids-corp.org / dpdk.git / blobdiff