txq_complete(struct txq *txq);
static __rte_always_inline uint32_t
-txq_mp2mr(struct txq *txq, struct rte_mempool *mp);
+txq_mb2mr(struct txq *txq, struct rte_mbuf *mb);
static __rte_always_inline void
mlx5_tx_dbrec(struct txq *txq, volatile struct mlx5_wqe *wqe);
static inline void
txq_complete(struct txq *txq)
{
- const unsigned int elts_n = 1 << txq->elts_n;
+ 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 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;
do {
volatile struct mlx5_cqe *tmp;
ctrl = (volatile struct mlx5_wqe_ctrl *)
tx_mlx5_wqe(txq, txq->wqe_pi);
elts_tail = ctrl->ctrl3;
- assert(elts_tail < (1 << txq->wqe_n));
+ assert((elts_tail & elts_m) < (1 << txq->wqe_n));
/* Free buffers. */
while (elts_free != elts_tail) {
- struct rte_mbuf *elt = (*txq->elts)[elts_free];
- unsigned int elts_free_next =
- (elts_free + 1) & (elts_n - 1);
- struct rte_mbuf *elt_next = (*txq->elts)[elts_free_next];
-
+ 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
- /* Poisoning. */
- memset(&(*txq->elts)[elts_free],
+ elts_free = txq->elts_tail;
+ /* Poisoning. */
+ while (elts_free != elts_tail) {
+ memset(&(*txq->elts)[elts_free & elts_m],
0x66,
- sizeof((*txq->elts)[elts_free]));
-#endif
- RTE_MBUF_PREFETCH_TO_FREE(elt_next);
- /* Only one segment needs to be freed. */
- rte_pktmbuf_free_seg(elt);
- elts_free = elts_free_next;
+ sizeof((*txq->elts)[elts_free & elts_m]));
+ ++elts_free;
}
+#endif
txq->cq_ci = cq_ci;
txq->elts_tail = elts_tail;
/* Update the consumer index. */
}
/**
- * Get Memory Region (MR) <-> Memory Pool (MP) association from txq->mp2mr[].
+ * 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.
*
* mr->lkey on success, (uint32_t)-1 on failure.
*/
static inline uint32_t
-txq_mp2mr(struct txq *txq, struct rte_mempool *mp)
+txq_mb2mr(struct txq *txq, struct rte_mbuf *mb)
{
- unsigned int i;
- uint32_t lkey = (uint32_t)-1;
+ 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].mp == NULL)) {
+ if (unlikely(txq->mp2mr[i].mr == NULL)) {
/* Unknown MP, add a new MR for it. */
break;
}
- if (txq->mp2mr[i].mp == mp) {
+ if (txq->mp2mr[i].start <= addr &&
+ txq->mp2mr[i].end >= addr) {
assert(txq->mp2mr[i].lkey != (uint32_t)-1);
assert(htonl(txq->mp2mr[i].mr->lkey) ==
txq->mp2mr[i].lkey);
- lkey = txq->mp2mr[i].lkey;
- break;
+ txq->mr_cache_idx = i;
+ return txq->mp2mr[i].lkey;
}
}
- if (unlikely(lkey == (uint32_t)-1))
- lkey = txq_mp2mr_reg(txq, mp, i);
- return lkey;
+ txq->mr_cache_idx = 0;
+ return txq_mp2mr_reg(txq, txq_mb2mp(mb), i);
}
/**
mlx5_tx_descriptor_status(void *tx_queue, uint16_t offset)
{
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;
+ uint16_t used;
txq_complete(txq);
- used = (txq->elts_head - txq->elts_tail) & elts_cnt;
+ used = txq->elts_head - txq->elts_tail;
if (offset < used)
return RTE_ETH_TX_DESC_FULL;
return RTE_ETH_TX_DESC_DONE;
{
struct txq *txq = (struct txq *)dpdk_txq;
uint16_t elts_head = txq->elts_head;
- const unsigned int elts_n = 1 << txq->elts_n;
+ const uint16_t elts_n = 1 << txq->elts_n;
+ const uint16_t elts_m = elts_n - 1;
unsigned int i = 0;
unsigned int j = 0;
unsigned int k = 0;
- unsigned int max;
+ uint16_t max_elts;
unsigned int max_inline = txq->max_inline;
const unsigned int inline_en = !!max_inline && txq->inline_en;
uint16_t max_wqe;
rte_prefetch0(*pkts);
/* Start processing. */
txq_complete(txq);
- max = (elts_n - (elts_head - txq->elts_tail));
- if (max > elts_n)
- max -= elts_n;
+ max_elts = (elts_n - (elts_head - txq->elts_tail));
max_wqe = (1u << txq->wqe_n) - (txq->wqe_ci - txq->wqe_pi);
if (unlikely(!max_wqe))
return 0;
uint16_t ehdr;
uint8_t cs_flags = 0;
uint64_t tso = 0;
+ uint16_t tso_segsz = 0;
#ifdef MLX5_PMD_SOFT_COUNTERS
uint32_t total_length = 0;
#endif
* that one ring entry remains unused.
*/
assert(segs_n);
- if (max < segs_n + 1)
+ if (max_elts < segs_n)
break;
- max -= segs_n;
+ max_elts -= segs_n;
--segs_n;
if (unlikely(--max_wqe == 0))
break;
if (length < (MLX5_WQE_DWORD_SIZE + 2))
break;
/* Update element. */
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[elts_head & elts_m] = buf;
/* Prefetch next buffer data. */
if (pkts_n - i > 1)
rte_prefetch0(
tso_header_sz = buf->l2_len + vlan_sz +
buf->l3_len + buf->l4_len;
+ tso_segsz = buf->tso_segsz;
if (is_tunneled && txq->tunnel_en) {
tso_header_sz += buf->outer_l2_len +
naddr = htonll(addr);
*dseg = (rte_v128u32_t){
htonl(length),
- txq_mp2mr(txq, txq_mb2mp(buf)),
+ txq_mb2mr(txq, buf),
naddr,
naddr >> 32,
};
naddr = htonll(rte_pktmbuf_mtod(buf, uintptr_t));
*dseg = (rte_v128u32_t){
htonl(length),
- txq_mp2mr(txq, txq_mb2mp(buf)),
+ txq_mb2mr(txq, buf),
naddr,
naddr >> 32,
};
- elts_head = (elts_head + 1) & (elts_n - 1);
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[++elts_head & elts_m] = buf;
++sg;
/* Advance counter only if all segs are successfully posted. */
if (sg < segs_n)
else
j += sg;
next_pkt:
- elts_head = (elts_head + 1) & (elts_n - 1);
+ ++elts_head;
++pkts;
++i;
/* Initialize known and common part of the WQE structure. */
};
wqe->eseg = (rte_v128u32_t){
0,
- cs_flags | (htons(buf->tso_segsz) << 16),
+ cs_flags | (htons(tso_segsz) << 16),
0,
(ehdr << 16) | htons(tso_header_sz),
};
/* Take a shortcut if nothing must be sent. */
if (unlikely((i + k) == 0))
return 0;
- txq->elts_head = (txq->elts_head + i + j) & (elts_n - 1);
+ txq->elts_head += (i + j);
/* Check whether completion threshold has been reached. */
comp = txq->elts_comp + i + j + k;
if (comp >= MLX5_TX_COMP_THRESH) {
{
struct txq *txq = (struct txq *)dpdk_txq;
uint16_t elts_head = txq->elts_head;
- const unsigned int elts_n = 1 << txq->elts_n;
+ const uint16_t elts_n = 1 << txq->elts_n;
+ const uint16_t elts_m = elts_n - 1;
unsigned int i = 0;
unsigned int j = 0;
- unsigned int max;
+ uint16_t max_elts;
uint16_t max_wqe;
unsigned int comp;
struct mlx5_mpw mpw = {
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_elts = (elts_n - (elts_head - txq->elts_tail));
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;
uint32_t length;
unsigned int segs_n = buf->nb_segs;
uint32_t cs_flags = 0;
* that one ring entry remains unused.
*/
assert(segs_n);
- if (max < segs_n + 1)
+ if (max_elts < segs_n)
break;
/* Do not bother with large packets MPW cannot handle. */
if (segs_n > MLX5_MPW_DSEG_MAX)
break;
- max -= segs_n;
+ max_elts -= segs_n;
--pkts_n;
/* Should we enable HW CKSUM offload */
if (buf->ol_flags &
volatile struct mlx5_wqe_data_seg *dseg;
uintptr_t addr;
- elts_head_next = (elts_head + 1) & (elts_n - 1);
assert(buf);
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[elts_head++ & elts_m] = 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)),
+ .lkey = txq_mb2mr(txq, buf),
.addr = htonll(addr),
};
- elts_head = elts_head_next;
#if defined(MLX5_PMD_SOFT_COUNTERS) || !defined(NDEBUG)
length += DATA_LEN(buf);
#endif
assert(length == mpw.len);
if (mpw.pkts_n == MLX5_MPW_DSEG_MAX)
mlx5_mpw_close(txq, &mpw);
- elts_head = elts_head_next;
#ifdef MLX5_PMD_SOFT_COUNTERS
/* Increment sent bytes counter. */
txq->stats.obytes += length;
{
struct txq *txq = (struct txq *)dpdk_txq;
uint16_t elts_head = txq->elts_head;
- const unsigned int elts_n = 1 << txq->elts_n;
+ const uint16_t elts_n = 1 << txq->elts_n;
+ const uint16_t elts_m = elts_n - 1;
unsigned int i = 0;
unsigned int j = 0;
- unsigned int max;
+ uint16_t max_elts;
uint16_t max_wqe;
unsigned int comp;
unsigned int inline_room = txq->max_inline * RTE_CACHE_LINE_SIZE;
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_elts = (elts_n - (elts_head - txq->elts_tail));
do {
struct rte_mbuf *buf = *(pkts++);
- unsigned int elts_head_next;
uintptr_t addr;
uint32_t length;
unsigned int segs_n = buf->nb_segs;
* that one ring entry remains unused.
*/
assert(segs_n);
- if (max < segs_n + 1)
+ if (max_elts < segs_n)
break;
/* Do not bother with large packets MPW cannot handle. */
if (segs_n > MLX5_MPW_DSEG_MAX)
break;
- max -= segs_n;
+ max_elts -= segs_n;
--pkts_n;
/*
* Compute max_wqe in case less WQE were consumed in previous
do {
volatile struct mlx5_wqe_data_seg *dseg;
- elts_head_next =
- (elts_head + 1) & (elts_n - 1);
assert(buf);
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[elts_head++ & elts_m] = 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)),
+ .lkey = txq_mb2mr(txq, buf),
.addr = htonll(addr),
};
- elts_head = elts_head_next;
#if defined(MLX5_PMD_SOFT_COUNTERS) || !defined(NDEBUG)
length += DATA_LEN(buf);
#endif
assert(mpw.state == MLX5_MPW_INL_STATE_OPENED);
assert(length <= inline_room);
assert(length == DATA_LEN(buf));
- elts_head_next = (elts_head + 1) & (elts_n - 1);
addr = rte_pktmbuf_mtod(buf, uintptr_t);
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[elts_head++ & elts_m] = buf;
/* Maximum number of bytes before wrapping. */
max = ((((uintptr_t)(txq->wqes)) +
(1 << txq->wqe_n) *
inline_room -= length;
}
}
- elts_head = elts_head_next;
#ifdef MLX5_PMD_SOFT_COUNTERS
/* Increment sent bytes counter. */
txq->stats.obytes += length;
{
struct txq *txq = (struct txq *)dpdk_txq;
uint16_t elts_head = txq->elts_head;
- const unsigned int elts_n = 1 << txq->elts_n;
+ const uint16_t elts_n = 1 << txq->elts_n;
+ const uint16_t elts_m = elts_n - 1;
unsigned int i = 0;
unsigned int j = 0;
- unsigned int max_elts;
+ uint16_t max_elts;
uint16_t max_wqe;
unsigned int max_inline = txq->max_inline * RTE_CACHE_LINE_SIZE;
unsigned int mpw_room = 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);
return 0;
do {
struct rte_mbuf *buf = *(pkts++);
- unsigned int elts_head_next;
uintptr_t addr;
uint64_t naddr;
unsigned int n;
* that one ring entry remains unused.
*/
assert(segs_n);
- if (max_elts - j < segs_n + 1)
+ if (max_elts - j < segs_n)
break;
/* Do not bother with large packets MPW cannot handle. */
if (segs_n > MLX5_MPW_DSEG_MAX)
do {
volatile struct mlx5_wqe_data_seg *dseg;
- elts_head_next =
- (elts_head + 1) & (elts_n - 1);
assert(buf);
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[elts_head++ & elts_m] = 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)),
+ .lkey = txq_mb2mr(txq, buf),
.addr = htonll(addr),
};
- elts_head = elts_head_next;
#if defined(MLX5_PMD_SOFT_COUNTERS) || !defined(NDEBUG)
length += DATA_LEN(buf);
#endif
/* 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. */
dseg = (volatile void *)
((uintptr_t)mpw.data.raw +
inl_pad);
- elts_head_next = (elts_head + 1) & (elts_n - 1);
- (*txq->elts)[elts_head] = buf;
+ (*txq->elts)[elts_head++ & elts_m] = buf;
addr = rte_pktmbuf_mtod(buf, uintptr_t);
for (n = 0; n * RTE_CACHE_LINE_SIZE < length; n++)
rte_prefetch2((void *)(addr +
naddr = htonll(addr);
*dseg = (rte_v128u32_t) {
htonl(length),
- txq_mp2mr(txq, txq_mb2mp(buf)),
+ txq_mb2mr(txq, buf),
naddr,
naddr >> 32,
};
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;
git.droids-corp.org / dpdk.git / blobdiff