summary |
shortlog |
log |
commit | commitdiff |
tree
raw |
patch |
inline | side by side (from parent 1:
8b581c6)
This is preparation step, we are going to store the index
of elts to free on completion in the dedicated free on
completion queue, this patch updates the elts freeing routine
and updates Tx error handling routine to be synced with
coming new queue.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
* Pointer to the error CQE.
*
* @return
* Pointer to the error CQE.
*
* @return
- * Negative value if queue recovery failed,
- * the last Tx buffer element to free otherwise.
+ * Negative value if queue recovery failed, otherwise
+ * the error completion entry is handled successfully.
mlx5_tx_error_cqe_handle(struct mlx5_txq_data *restrict txq,
volatile struct mlx5_err_cqe *err_cqe)
{
mlx5_tx_error_cqe_handle(struct mlx5_txq_data *restrict txq,
volatile struct mlx5_err_cqe *err_cqe)
{
*/
txq->stats.oerrors += ((txq->wqe_ci & wqe_m) -
new_wqe_pi) & wqe_m;
*/
txq->stats.oerrors += ((txq->wqe_ci & wqe_m) -
new_wqe_pi) & wqe_m;
- if (tx_recover_qp(txq_ctrl) == 0) {
- txq->cq_ci++;
- /* Release all the remaining buffers. */
- return txq->elts_head;
+ if (tx_recover_qp(txq_ctrl)) {
+ /* Recovering failed - retry later on the same WQE. */
+ return -1;
- /* Recovering failed - try again later on the same WQE. */
- return -1;
- } else {
- txq->cq_ci++;
+ /* Release all the remaining buffers. */
+ txq_free_elts(txq_ctrl);
- /* Do not release buffers. */
- return txq->elts_tail;
* Pointer to TX queue structure.
* @param valid CQE pointer
* if not NULL update txq->wqe_pi and flush the buffers
* Pointer to TX queue structure.
* @param valid CQE pointer
* if not NULL update txq->wqe_pi and flush the buffers
- * @param itail
- * if not negative - flush the buffers till this index.
* @param olx
* Configured Tx offloads mask. It is fully defined at
* compile time and may be used for optimization.
* @param olx
* Configured Tx offloads mask. It is fully defined at
* compile time and may be used for optimization.
static __rte_always_inline void
mlx5_tx_comp_flush(struct mlx5_txq_data *restrict txq,
volatile struct mlx5_cqe *last_cqe,
static __rte_always_inline void
mlx5_tx_comp_flush(struct mlx5_txq_data *restrict txq,
volatile struct mlx5_cqe *last_cqe,
unsigned int olx __rte_unused)
{
unsigned int olx __rte_unused)
{
if (likely(last_cqe != NULL)) {
if (likely(last_cqe != NULL)) {
txq->wqe_pi = rte_be_to_cpu_16(last_cqe->wqe_counter);
tail = ((volatile struct mlx5_wqe_cseg *)
(txq->wqes + (txq->wqe_pi & txq->wqe_m)))->misc;
txq->wqe_pi = rte_be_to_cpu_16(last_cqe->wqe_counter);
tail = ((volatile struct mlx5_wqe_cseg *)
(txq->wqes + (txq->wqe_pi & txq->wqe_m)))->misc;
- } else if (itail >= 0) {
- tail = (uint16_t)itail;
- } else {
- return;
- }
- rte_compiler_barrier();
- *txq->cq_db = rte_cpu_to_be_32(txq->cq_ci);
- if (likely(tail != txq->elts_tail)) {
- mlx5_tx_free_elts(txq, tail, olx);
- assert(tail == txq->elts_tail);
+ if (likely(tail != txq->elts_tail)) {
+ mlx5_tx_free_elts(txq, tail, olx);
+ assert(tail == txq->elts_tail);
+ }
{
unsigned int count = MLX5_TX_COMP_MAX_CQE;
volatile struct mlx5_cqe *last_cqe = NULL;
{
unsigned int count = MLX5_TX_COMP_MAX_CQE;
volatile struct mlx5_cqe *last_cqe = NULL;
+ uint16_t ci = txq->cq_ci;
int ret;
static_assert(MLX5_CQE_STATUS_HW_OWN < 0, "Must be negative value");
int ret;
static_assert(MLX5_CQE_STATUS_HW_OWN < 0, "Must be negative value");
do {
volatile struct mlx5_cqe *cqe;
do {
volatile struct mlx5_cqe *cqe;
- cqe = &txq->cqes[txq->cq_ci & txq->cqe_m];
- ret = check_cqe(cqe, txq->cqe_s, txq->cq_ci);
+ cqe = &txq->cqes[ci & txq->cqe_m];
+ ret = check_cqe(cqe, txq->cqe_s, ci);
if (unlikely(ret != MLX5_CQE_STATUS_SW_OWN)) {
if (likely(ret != MLX5_CQE_STATUS_ERR)) {
/* No new CQEs in completion queue. */
if (unlikely(ret != MLX5_CQE_STATUS_SW_OWN)) {
if (likely(ret != MLX5_CQE_STATUS_ERR)) {
/* No new CQEs in completion queue. */
rte_wmb();
ret = mlx5_tx_error_cqe_handle
(txq, (volatile struct mlx5_err_cqe *)cqe);
rte_wmb();
ret = mlx5_tx_error_cqe_handle
(txq, (volatile struct mlx5_err_cqe *)cqe);
+ if (unlikely(ret < 0)) {
+ /*
+ * Some error occurred on queue error
+ * handling, we do not advance the index
+ * here, allowing to retry on next call.
+ */
+ return;
+ }
- * Flush buffers, update consuming index
- * if recovery succeeded. Otherwise
- * just try to recover later.
+ * We are going to fetch all entries with
+ * MLX5_CQE_SYNDROME_WR_FLUSH_ERR status.
- last_cqe = NULL;
- break;
}
/* Normal transmit completion. */
}
/* Normal transmit completion. */
last_cqe = cqe;
#ifndef NDEBUG
if (txq->cq_pi)
--txq->cq_pi;
#endif
last_cqe = cqe;
#ifndef NDEBUG
if (txq->cq_pi)
--txq->cq_pi;
#endif
- /*
- * We have to restrict the amount of processed CQEs
- * in one tx_burst routine call. The CQ may be large
- * and many CQEs may be updated by the NIC in one
- * transaction. Buffers freeing is time consuming,
- * multiple iterations may introduce significant
- * latency.
- */
- } while (--count);
- mlx5_tx_comp_flush(txq, last_cqe, ret, olx);
+ /*
+ * We have to restrict the amount of processed CQEs
+ * in one tx_burst routine call. The CQ may be large
+ * and many CQEs may be updated by the NIC in one
+ * transaction. Buffers freeing is time consuming,
+ * multiple iterations may introduce significant
+ * latency.
+ */
+ if (--count == 0)
+ break;
+ } while (true);
+ if (likely(ci != txq->cq_ci)) {
+ /*
+ * Update completion queue consuming index
+ * and ring doorbell to notify hardware.
+ */
+ rte_compiler_barrier();
+ txq->cq_ci = ci;
+ *txq->cq_db = rte_cpu_to_be_32(ci);
+ mlx5_tx_comp_flush(txq, last_cqe, olx);
+ }
int mlx5_txq_releasable(struct rte_eth_dev *dev, uint16_t idx);
int mlx5_txq_verify(struct rte_eth_dev *dev);
void txq_alloc_elts(struct mlx5_txq_ctrl *txq_ctrl);
int mlx5_txq_releasable(struct rte_eth_dev *dev, uint16_t idx);
int mlx5_txq_verify(struct rte_eth_dev *dev);
void txq_alloc_elts(struct mlx5_txq_ctrl *txq_ctrl);
+void txq_free_elts(struct mlx5_txq_ctrl *txq_ctrl);
uint64_t mlx5_get_tx_port_offloads(struct rte_eth_dev *dev);
/* mlx5_rxtx.c */
uint64_t mlx5_get_tx_port_offloads(struct rte_eth_dev *dev);
/* mlx5_rxtx.c */
void mlx5_set_ptype_table(void);
void mlx5_set_cksum_table(void);
void mlx5_set_swp_types_table(void);
void mlx5_set_ptype_table(void);
void mlx5_set_cksum_table(void);
void mlx5_set_swp_types_table(void);
-__rte_noinline int mlx5_tx_error_cqe_handle
- (struct mlx5_txq_data *restrict txq,
- volatile struct mlx5_err_cqe *err_cqe);
uint16_t mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n);
void mlx5_rxq_initialize(struct mlx5_rxq_data *rxq);
__rte_noinline int mlx5_rx_err_handle(struct mlx5_rxq_data *rxq, uint8_t vec);
uint16_t mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n);
void mlx5_rxq_initialize(struct mlx5_rxq_data *rxq);
__rte_noinline int mlx5_rx_err_handle(struct mlx5_rxq_data *rxq, uint8_t vec);
* @param txq_ctrl
* Pointer to TX queue structure.
*/
* @param txq_ctrl
* Pointer to TX queue structure.
*/
txq_free_elts(struct mlx5_txq_ctrl *txq_ctrl)
{
const uint16_t elts_n = 1 << txq_ctrl->txq.elts_n;
txq_free_elts(struct mlx5_txq_ctrl *txq_ctrl)
{
const uint16_t elts_n = 1 << txq_ctrl->txq.elts_n;