#ifndef RTE_PMD_MLX5_RXTX_VEC_NEON_H_
#define RTE_PMD_MLX5_RXTX_VEC_NEON_H_
-#include <assert.h>
#include <stdint.h>
#include <string.h>
#include <stdlib.h>
#include <rte_mempool.h>
#include <rte_prefetch.h>
+#include <mlx5_prm.h>
+
+#include "mlx5_defs.h"
#include "mlx5.h"
#include "mlx5_utils.h"
#include "mlx5_rxtx.h"
#include "mlx5_rxtx_vec.h"
#include "mlx5_autoconf.h"
-#include "mlx5_defs.h"
-#include "mlx5_prm.h"
#pragma GCC diagnostic ignored "-Wcast-qual"
sizeof(uint16_t) * 8) : 0);
#endif
- if (!(pos & 0x7) && pos + 8 < mcqe_n)
- rte_prefetch0((void *)(cq + pos + 8));
+ for (i = 0; i < MLX5_VPMD_DESCS_PER_LOOP; ++i)
+ if (likely(pos + i < mcqe_n))
+ rte_prefetch0((void *)(cq + pos + i));
__asm__ volatile (
/* A.1 load mCQEs into a 128bit register. */
"ld1 {v16.16b - v17.16b}, [%[mcq]] \n\t"
elts[pos + 2]->hash.fdir.hi = flow_tag;
elts[pos + 3]->hash.fdir.hi = flow_tag;
}
+ if (rxq->dynf_meta) {
+ int32_t offs = rxq->flow_meta_offset;
+ const uint32_t meta =
+ *RTE_MBUF_DYNFIELD(t_pkt, offs, uint32_t *);
+
+ /* Check if title packet has valid metadata. */
+ if (meta) {
+ MLX5_ASSERT(t_pkt->ol_flags &
+ rxq->flow_meta_mask);
+ *RTE_MBUF_DYNFIELD(elts[pos], offs,
+ uint32_t *) = meta;
+ *RTE_MBUF_DYNFIELD(elts[pos + 1], offs,
+ uint32_t *) = meta;
+ *RTE_MBUF_DYNFIELD(elts[pos + 2], offs,
+ uint32_t *) = meta;
+ *RTE_MBUF_DYNFIELD(elts[pos + 3], offs,
+ uint32_t *) = meta;
+ }
+ }
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;
* @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.
*
* @return
* Number of packets received including errors (<= pkts_n).
*/
static inline uint16_t
rxq_burst_v(struct mlx5_rxq_data *rxq, struct rte_mbuf **pkts, uint16_t pkts_n,
- uint64_t *err)
+ uint64_t *err, bool *no_cq)
{
const uint16_t q_n = 1 << rxq->cqe_n;
const uint16_t q_mask = q_n - 1;
};
const uint32x4_t flow_mark_adj = { 0, 0, 0, rxq->mark * (-1) };
- assert(rxq->sges_n == 0);
- assert(rxq->cqe_n == rxq->elts_n);
+ 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);
/* 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)
+ if (!pkts_n) {
+ *no_cq = !rcvd_pkt;
return rcvd_pkt;
+ }
/* At this point, there shouldn't be any remained packets. */
- assert(rxq->decompressed == 0);
+ 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
rxq_cq_to_ptype_oflags_v(rxq, ptype_info, flow_tag,
opcode, &elts[pos]);
if (rxq->hw_timestamp) {
- 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);
+ 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);
+ 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);
+ 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);
+ 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);
+ } 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);
+ }
}
- if (rte_flow_dynf_metadata_avail()) {
+ if (rxq->dynf_meta) {
/* This code is subject for futher optimization. */
- *RTE_FLOW_DYNF_METADATA(elts[pos]) =
+ 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_FLOW_DYNF_METADATA(elts[pos + 1]) =
+ *RTE_MBUF_DYNFIELD(pkts[pos + 1], offs, uint32_t *) =
container_of(p1, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
- *RTE_FLOW_DYNF_METADATA(elts[pos + 2]) =
+ *RTE_MBUF_DYNFIELD(pkts[pos + 2], offs, uint32_t *) =
container_of(p2, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
- *RTE_FLOW_DYNF_METADATA(elts[pos + 3]) =
+ *RTE_MBUF_DYNFIELD(pkts[pos + 3], offs, uint32_t *) =
container_of(p3, struct mlx5_cqe,
pkt_info)->flow_table_metadata;
- if (*RTE_FLOW_DYNF_METADATA(elts[pos]))
- elts[pos]->ol_flags |= PKT_RX_DYNF_METADATA;
- if (*RTE_FLOW_DYNF_METADATA(elts[pos + 1]))
- elts[pos + 1]->ol_flags |= PKT_RX_DYNF_METADATA;
- if (*RTE_FLOW_DYNF_METADATA(elts[pos + 2]))
- elts[pos + 2]->ol_flags |= PKT_RX_DYNF_METADATA;
- if (*RTE_FLOW_DYNF_METADATA(elts[pos + 3]))
- elts[pos + 3]->ol_flags |= PKT_RX_DYNF_METADATA;
+ if (*RTE_MBUF_DYNFIELD(pkts[pos], offs, uint32_t *))
+ elts[pos]->ol_flags |= rxq->flow_meta_mask;
+ if (*RTE_MBUF_DYNFIELD(pkts[pos + 1], offs, uint32_t *))
+ elts[pos + 1]->ol_flags |= rxq->flow_meta_mask;
+ if (*RTE_MBUF_DYNFIELD(pkts[pos + 2], offs, uint32_t *))
+ elts[pos + 2]->ol_flags |= rxq->flow_meta_mask;
+ if (*RTE_MBUF_DYNFIELD(pkts[pos + 3], offs, uint32_t *))
+ elts[pos + 3]->ol_flags |= rxq->flow_meta_mask;
}
#ifdef MLX5_PMD_SOFT_COUNTERS
/* Add up received bytes count. */
break;
}
/* If no new CQE seen, return without updating cq_db. */
- if (unlikely(!nocmp_n && comp_idx == MLX5_VPMD_DESCS_PER_LOOP))
+ 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. */
- assert(nocmp_n <= pkts_n);
+ MLX5_ASSERT(nocmp_n <= pkts_n);
rxq->cq_ci += nocmp_n;
rxq->rq_pi += nocmp_n;
rcvd_pkt += nocmp_n;
#endif
/* Decompress the last CQE if compressed. */
if (comp_idx < MLX5_VPMD_DESCS_PER_LOOP && comp_idx == n) {
- assert(comp_idx == (nocmp_n % MLX5_VPMD_DESCS_PER_LOOP));
+ 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. */
}
rte_cio_wmb();
*rxq->cq_db = rte_cpu_to_be_32(rxq->cq_ci);
+ *no_cq = !rcvd_pkt;
return rcvd_pkt;
}