#ifndef RTE_PMD_MLX5_RXTX_VEC_ALTIVEC_H_
#define RTE_PMD_MLX5_RXTX_VEC_ALTIVEC_H_
-#include <assert.h>
#include <stdint.h>
#include <string.h>
#include <stdlib.h>
-#include <altivec.h>
+#include <rte_altivec.h>
#include <rte_mbuf.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"
#ifndef __INTEL_COMPILER
#pragma GCC diagnostic ignored "-Wcast-qual"
/**
* 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;
const vector unsigned long shmax = {64, 64};
#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));
/* A.1 load mCQEs into a 128bit register. */
mcqe1 = (vector unsigned char)vec_vsx_ld(0,
(signed int const *)&mcq[pos % 8]);
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;
vector unsigned char ol_flags = (vector unsigned char)
(vector unsigned int){
rxq->rss_hash * PKT_RX_RSS_HASH |
- rxq->hw_timestamp * PKT_RX_TIMESTAMP,
+ rxq->hw_timestamp * rxq->timestamp_rx_flag,
rxq->rss_hash * PKT_RX_RSS_HASH |
- rxq->hw_timestamp * PKT_RX_TIMESTAMP,
+ rxq->hw_timestamp * rxq->timestamp_rx_flag,
rxq->rss_hash * PKT_RX_RSS_HASH |
- rxq->hw_timestamp * PKT_RX_TIMESTAMP,
+ rxq->hw_timestamp * rxq->timestamp_rx_flag,
rxq->rss_hash * PKT_RX_RSS_HASH |
- rxq->hw_timestamp * PKT_RX_TIMESTAMP};
+ rxq->hw_timestamp * rxq->timestamp_rx_flag};
vector unsigned char cv_flags;
const vector unsigned char zero = (vector unsigned char){0};
const vector unsigned char ptype_mask =
(vector unsigned char *)&pkts[3]->rearm_data);
}
-
/**
- * 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] 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)
+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;
unsigned int ownership = !!(rxq->cq_ci & (q_mask + 1));
const vector unsigned char zero = (vector unsigned char){0};
const vector unsigned char ones = vec_splat_u8(-1);
const vector unsigned short cqe_sel_mask2 =
(vector unsigned short){0, 0, 0xffff, 0, 0, 0, 0, 0};
- assert(rxq->sges_n == 0);
- assert(rxq->cqe_n == rxq->elts_n);
- cq = &(*rxq->cqes)[cq_idx];
- rte_prefetch0(cq);
- rte_prefetch0(cq + 1);
- rte_prefetch0(cq + 2);
- rte_prefetch0(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;
- rxq->decompressed -= rcvd_pkt;
- pkts += 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)
- return rcvd_pkt;
- /* At this point, there shouldn't be any remaining packets. */
- assert(rxq->decompressed == 0);
-
/*
* A. load first Qword (8bytes) in one loop.
* B. copy 4 mbuf pointers from elts ring to returing pkts.
/* A.1 load cqes. */
p3 = (unsigned int)((vector unsigned short)p)[3];
cqes[3] = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos + p3].sop_drop_qpn, 0LL};
rte_compiler_barrier();
p2 = (unsigned int)((vector unsigned short)p)[2];
cqes[2] = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos + p2].sop_drop_qpn, 0LL};
rte_compiler_barrier();
/* A.1 load a block having op_own. */
p1 = (unsigned int)((vector unsigned short)p)[1];
cqes[1] = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos + p1].sop_drop_qpn, 0LL};
rte_compiler_barrier();
cqes[0] = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos].sop_drop_qpn, 0LL};
rte_compiler_barrier();
/* B.2 copy mbuf pointers. */
*(vector unsigned char *)&pkts[pos] = mbp1;
*(vector unsigned char *)&pkts[pos + 2] = mbp2;
- rte_cio_rmb();
+ rte_io_rmb();
/* C.1 load remaining CQE data and extract necessary fields. */
cqe_tmp2 = *(vector unsigned char *)
vec_sel((vector unsigned short)cqes[2],
(vector unsigned short)cqe_tmp1, cqe_sel_mask1);
cqe_tmp2 = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos + p3].rsvd3[9], 0LL};
cqe_tmp1 = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos + p2].rsvd3[9], 0LL};
cqes[3] = (vector unsigned char)
vec_sel((vector unsigned short)cqes[3],
vec_sel((vector unsigned short)cqes[0],
(vector unsigned short)cqe_tmp1, cqe_sel_mask1);
cqe_tmp2 = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos + p1].rsvd3[9], 0LL};
cqe_tmp1 = (vector unsigned char)(vector unsigned long){
- *(__attribute__((__aligned__(8))) unsigned long *)
+ *(__rte_aligned(8) unsigned long *)
&cq[pos].rsvd3[9], 0LL};
cqes[1] = (vector unsigned char)
vec_sel((vector unsigned short)cqes[1],
/* D.5 fill in mbuf - rearm_data and packet_type. */
rxq_cq_to_ptype_oflags_v(rxq, cqes, opcode, &pkts[pos]);
if (rxq->hw_timestamp) {
- pkts[pos]->timestamp =
- rte_be_to_cpu_64(cq[pos].timestamp);
- pkts[pos + 1]->timestamp =
- rte_be_to_cpu_64(cq[pos + p1].timestamp);
- pkts[pos + 2]->timestamp =
- rte_be_to_cpu_64(cq[pos + p2].timestamp);
- pkts[pos + 3]->timestamp =
- rte_be_to_cpu_64(cq[pos + p3].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(cq[pos].timestamp);
+ mlx5_timestamp_set(pkts[pos], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
+ ts = rte_be_to_cpu_64(cq[pos + p1].timestamp);
+ mlx5_timestamp_set(pkts[pos + 1], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
+ ts = rte_be_to_cpu_64(cq[pos + p2].timestamp);
+ mlx5_timestamp_set(pkts[pos + 2], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
+ ts = rte_be_to_cpu_64(cq[pos + p3].timestamp);
+ mlx5_timestamp_set(pkts[pos + 3], offset,
+ mlx5_txpp_convert_rx_ts(sh, ts));
+ } else {
+ mlx5_timestamp_set(pkts[pos], offset,
+ rte_be_to_cpu_64(cq[pos].timestamp));
+ mlx5_timestamp_set(pkts[pos + 1], offset,
+ rte_be_to_cpu_64(cq[pos + p1].timestamp));
+ mlx5_timestamp_set(pkts[pos + 2], offset,
+ rte_be_to_cpu_64(cq[pos + p2].timestamp));
+ mlx5_timestamp_set(pkts[pos + 3], offset,
+ rte_be_to_cpu_64(cq[pos + p3].timestamp));
+ }
}
- if (rte_flow_dynf_metadata_avail()) {
- uint64_t flag = rte_flow_dynf_metadata_mask;
- int offs = rte_flow_dynf_metadata_offs;
+ if (rxq->dynf_meta) {
+ uint64_t flag = rxq->flow_meta_mask;
+ int32_t offs = rxq->flow_meta_offset;
uint32_t metadata;
/* This code is subject for futher optimization. */
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))
- return rcvd_pkt;
- /* Update the consumer indexes for non-compressed CQEs. */
- 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) {
- 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_compiler_barrier();
- *rxq->cq_db = rte_cpu_to_be_32(rxq->cq_ci);
- return rcvd_pkt;
+ if (comp_idx == n)
+ *comp = comp_idx;
+ return nocmp_n;
}
#endif /* RTE_PMD_MLX5_RXTX_VEC_ALTIVEC_H_ */
git.droids-corp.org / dpdk.git / blobdiff