#include <errno.h>
#include <string.h>
#include <stdint.h>
+#include <fcntl.h>
/* Verbs header. */
/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
#pragma GCC diagnostic ignored "-Wpedantic"
#endif
#include <infiniband/verbs.h>
-#include <infiniband/arch.h>
-#include <infiniband/mlx5_hw.h>
+#include <infiniband/mlx5dv.h>
#ifdef PEDANTIC
#pragma GCC diagnostic error "-Wpedantic"
#endif
-/* DPDK headers don't like -pedantic. */
-#ifdef PEDANTIC
-#pragma GCC diagnostic ignored "-Wpedantic"
-#endif
#include <rte_mbuf.h>
#include <rte_malloc.h>
#include <rte_ethdev.h>
#include <rte_common.h>
-#ifdef PEDANTIC
-#pragma GCC diagnostic error "-Wpedantic"
-#endif
+#include <rte_interrupts.h>
+#include <rte_debug.h>
+#include <rte_io.h>
#include "mlx5.h"
#include "mlx5_rxtx.h"
/* Initialization data for hash RX queues. */
const struct hash_rxq_init hash_rxq_init[] = {
[HASH_RXQ_TCPV4] = {
- .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV4 |
- IBV_EXP_RX_HASH_DST_IPV4 |
- IBV_EXP_RX_HASH_SRC_PORT_TCP |
- IBV_EXP_RX_HASH_DST_PORT_TCP),
+ .hash_fields = (IBV_RX_HASH_SRC_IPV4 |
+ IBV_RX_HASH_DST_IPV4 |
+ IBV_RX_HASH_SRC_PORT_TCP |
+ IBV_RX_HASH_DST_PORT_TCP),
.dpdk_rss_hf = ETH_RSS_NONFRAG_IPV4_TCP,
.flow_priority = 0,
.flow_spec.tcp_udp = {
- .type = IBV_EXP_FLOW_SPEC_TCP,
+ .type = IBV_FLOW_SPEC_TCP,
.size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
},
.underlayer = &hash_rxq_init[HASH_RXQ_IPV4],
},
[HASH_RXQ_UDPV4] = {
- .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV4 |
- IBV_EXP_RX_HASH_DST_IPV4 |
- IBV_EXP_RX_HASH_SRC_PORT_UDP |
- IBV_EXP_RX_HASH_DST_PORT_UDP),
+ .hash_fields = (IBV_RX_HASH_SRC_IPV4 |
+ IBV_RX_HASH_DST_IPV4 |
+ IBV_RX_HASH_SRC_PORT_UDP |
+ IBV_RX_HASH_DST_PORT_UDP),
.dpdk_rss_hf = ETH_RSS_NONFRAG_IPV4_UDP,
.flow_priority = 0,
.flow_spec.tcp_udp = {
- .type = IBV_EXP_FLOW_SPEC_UDP,
+ .type = IBV_FLOW_SPEC_UDP,
.size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
},
.underlayer = &hash_rxq_init[HASH_RXQ_IPV4],
},
[HASH_RXQ_IPV4] = {
- .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV4 |
- IBV_EXP_RX_HASH_DST_IPV4),
+ .hash_fields = (IBV_RX_HASH_SRC_IPV4 |
+ IBV_RX_HASH_DST_IPV4),
.dpdk_rss_hf = (ETH_RSS_IPV4 |
ETH_RSS_FRAG_IPV4),
.flow_priority = 1,
.flow_spec.ipv4 = {
- .type = IBV_EXP_FLOW_SPEC_IPV4,
+ .type = IBV_FLOW_SPEC_IPV4,
.size = sizeof(hash_rxq_init[0].flow_spec.ipv4),
},
.underlayer = &hash_rxq_init[HASH_RXQ_ETH],
},
[HASH_RXQ_TCPV6] = {
- .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV6 |
- IBV_EXP_RX_HASH_DST_IPV6 |
- IBV_EXP_RX_HASH_SRC_PORT_TCP |
- IBV_EXP_RX_HASH_DST_PORT_TCP),
+ .hash_fields = (IBV_RX_HASH_SRC_IPV6 |
+ IBV_RX_HASH_DST_IPV6 |
+ IBV_RX_HASH_SRC_PORT_TCP |
+ IBV_RX_HASH_DST_PORT_TCP),
.dpdk_rss_hf = ETH_RSS_NONFRAG_IPV6_TCP,
.flow_priority = 0,
.flow_spec.tcp_udp = {
- .type = IBV_EXP_FLOW_SPEC_TCP,
+ .type = IBV_FLOW_SPEC_TCP,
.size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
},
.underlayer = &hash_rxq_init[HASH_RXQ_IPV6],
},
[HASH_RXQ_UDPV6] = {
- .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV6 |
- IBV_EXP_RX_HASH_DST_IPV6 |
- IBV_EXP_RX_HASH_SRC_PORT_UDP |
- IBV_EXP_RX_HASH_DST_PORT_UDP),
+ .hash_fields = (IBV_RX_HASH_SRC_IPV6 |
+ IBV_RX_HASH_DST_IPV6 |
+ IBV_RX_HASH_SRC_PORT_UDP |
+ IBV_RX_HASH_DST_PORT_UDP),
.dpdk_rss_hf = ETH_RSS_NONFRAG_IPV6_UDP,
.flow_priority = 0,
.flow_spec.tcp_udp = {
- .type = IBV_EXP_FLOW_SPEC_UDP,
+ .type = IBV_FLOW_SPEC_UDP,
.size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
},
.underlayer = &hash_rxq_init[HASH_RXQ_IPV6],
},
[HASH_RXQ_IPV6] = {
- .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV6 |
- IBV_EXP_RX_HASH_DST_IPV6),
+ .hash_fields = (IBV_RX_HASH_SRC_IPV6 |
+ IBV_RX_HASH_DST_IPV6),
.dpdk_rss_hf = (ETH_RSS_IPV6 |
ETH_RSS_FRAG_IPV6),
.flow_priority = 1,
.flow_spec.ipv6 = {
- .type = IBV_EXP_FLOW_SPEC_IPV6,
+ .type = IBV_FLOW_SPEC_IPV6,
.size = sizeof(hash_rxq_init[0].flow_spec.ipv6),
},
.underlayer = &hash_rxq_init[HASH_RXQ_ETH],
.dpdk_rss_hf = 0,
.flow_priority = 2,
.flow_spec.eth = {
- .type = IBV_EXP_FLOW_SPEC_ETH,
+ .type = IBV_FLOW_SPEC_ETH,
.size = sizeof(hash_rxq_init[0].flow_spec.eth),
},
.underlayer = NULL,
* Total size of the flow attribute buffer. No errors are defined.
*/
size_t
-priv_flow_attr(struct priv *priv, struct ibv_exp_flow_attr *flow_attr,
+priv_flow_attr(struct priv *priv, struct ibv_flow_attr *flow_attr,
size_t flow_attr_size, enum hash_rxq_type type)
{
size_t offset = sizeof(*flow_attr);
return offset;
flow_attr_size = offset;
init = &hash_rxq_init[type];
- *flow_attr = (struct ibv_exp_flow_attr){
- .type = IBV_EXP_FLOW_ATTR_NORMAL,
+ *flow_attr = (struct ibv_flow_attr){
+ .type = IBV_FLOW_ATTR_NORMAL,
/* Priorities < 3 are reserved for flow director. */
.priority = init->flow_priority + 3,
.num_of_specs = 0,
int
priv_create_hash_rxqs(struct priv *priv)
{
- struct ibv_exp_wq *wqs[priv->reta_idx_n];
+ struct ibv_wq *wqs[priv->reta_idx_n];
struct ind_table_init ind_table_init[IND_TABLE_INIT_N];
unsigned int ind_tables_n =
priv_make_ind_table_init(priv, &ind_table_init);
unsigned int hash_rxqs_n = 0;
struct hash_rxq (*hash_rxqs)[] = NULL;
- struct ibv_exp_rwq_ind_table *(*ind_tables)[] = NULL;
+ struct ibv_rwq_ind_table *(*ind_tables)[] = NULL;
unsigned int i;
unsigned int j;
unsigned int k;
assert(priv->hash_rxqs_n == 0);
assert(priv->pd != NULL);
assert(priv->ctx != NULL);
+ if (priv->isolated)
+ return 0;
if (priv->rxqs_n == 0)
return EINVAL;
assert(priv->rxqs != NULL);
priv->reta_idx_n);
}
for (i = 0; (i != priv->reta_idx_n); ++i) {
- struct rxq_ctrl *rxq_ctrl;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
rxq_ctrl = container_of((*priv->rxqs)[(*priv->reta_idx)[i]],
- struct rxq_ctrl, rxq);
+ struct mlx5_rxq_ctrl, rxq);
wqs[i] = rxq_ctrl->wq;
}
/* Get number of hash RX queues to configure. */
goto error;
}
for (i = 0; (i != ind_tables_n); ++i) {
- struct ibv_exp_rwq_ind_table_init_attr ind_init_attr = {
- .pd = priv->pd,
+ struct ibv_rwq_ind_table_init_attr ind_init_attr = {
.log_ind_tbl_size = 0, /* Set below. */
.ind_tbl = wqs,
.comp_mask = 0,
};
unsigned int ind_tbl_size = ind_table_init[i].max_size;
- struct ibv_exp_rwq_ind_table *ind_table;
+ struct ibv_rwq_ind_table *ind_table;
if (priv->reta_idx_n < ind_tbl_size)
ind_tbl_size = priv->reta_idx_n;
ind_init_attr.log_ind_tbl_size = log2above(ind_tbl_size);
errno = 0;
- ind_table = ibv_exp_create_rwq_ind_table(priv->ctx,
- &ind_init_attr);
+ ind_table = ibv_create_rwq_ind_table(priv->ctx,
+ &ind_init_attr);
if (ind_table != NULL) {
(*ind_tables)[i] = ind_table;
continue;
hash_rxq_type_from_pos(&ind_table_init[j], k);
struct rte_eth_rss_conf *priv_rss_conf =
(*priv->rss_conf)[type];
- struct ibv_exp_rx_hash_conf hash_conf = {
- .rx_hash_function = IBV_EXP_RX_HASH_FUNC_TOEPLITZ,
+ struct ibv_rx_hash_conf hash_conf = {
+ .rx_hash_function = IBV_RX_HASH_FUNC_TOEPLITZ,
.rx_hash_key_len = (priv_rss_conf ?
priv_rss_conf->rss_key_len :
rss_hash_default_key_len),
priv_rss_conf->rss_key :
rss_hash_default_key),
.rx_hash_fields_mask = hash_rxq_init[type].hash_fields,
- .rwq_ind_tbl = (*ind_tables)[j],
};
- struct ibv_exp_qp_init_attr qp_init_attr = {
- .max_inl_recv = 0, /* Currently not supported. */
+ struct ibv_qp_init_attr_ex qp_init_attr = {
.qp_type = IBV_QPT_RAW_PACKET,
- .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD |
- IBV_EXP_QP_INIT_ATTR_RX_HASH),
+ .comp_mask = (IBV_QP_INIT_ATTR_PD |
+ IBV_QP_INIT_ATTR_IND_TABLE |
+ IBV_QP_INIT_ATTR_RX_HASH),
+ .rx_hash_conf = hash_conf,
+ .rwq_ind_tbl = (*ind_tables)[j],
.pd = priv->pd,
- .rx_hash_conf = &hash_conf,
- .port_num = priv->port,
};
DEBUG("using indirection table %u for hash RX queue %u type %d",
j, i, type);
*hash_rxq = (struct hash_rxq){
.priv = priv,
- .qp = ibv_exp_create_qp(priv->ctx, &qp_init_attr),
+ .qp = ibv_create_qp_ex(priv->ctx, &qp_init_attr),
.type = type,
};
if (hash_rxq->qp == NULL) {
}
if (ind_tables != NULL) {
for (j = 0; (j != ind_tables_n); ++j) {
- struct ibv_exp_rwq_ind_table *ind_table =
+ struct ibv_rwq_ind_table *ind_table =
(*ind_tables)[j];
if (ind_table == NULL)
continue;
- claim_zero(ibv_exp_destroy_rwq_ind_table(ind_table));
+ claim_zero(ibv_destroy_rwq_ind_table(ind_table));
}
rte_free(ind_tables);
}
rte_free(priv->hash_rxqs);
priv->hash_rxqs = NULL;
for (i = 0; (i != priv->ind_tables_n); ++i) {
- struct ibv_exp_rwq_ind_table *ind_table =
+ struct ibv_rwq_ind_table *ind_table =
(*priv->ind_tables)[i];
assert(ind_table != NULL);
- claim_zero(ibv_exp_destroy_rwq_ind_table(ind_table));
+ claim_zero(ibv_destroy_rwq_ind_table(ind_table));
}
priv->ind_tables_n = 0;
rte_free(priv->ind_tables);
* Pointer to RX queue structure.
* @param elts_n
* Number of elements to allocate.
- * @param[in] pool
- * If not NULL, fetch buffers from this array instead of allocating them
- * with rte_pktmbuf_alloc().
*
* @return
* 0 on success, errno value on failure.
*/
static int
-rxq_alloc_elts(struct rxq_ctrl *rxq_ctrl, unsigned int elts_n,
- struct rte_mbuf *(*pool)[])
+rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl, unsigned int elts_n)
{
const unsigned int sges_n = 1 << rxq_ctrl->rxq.sges_n;
unsigned int i;
volatile struct mlx5_wqe_data_seg *scat =
&(*rxq_ctrl->rxq.wqes)[i];
- if (pool != NULL) {
- buf = (*pool)[i];
- assert(buf != NULL);
- rte_pktmbuf_reset(buf);
- rte_pktmbuf_refcnt_update(buf, 1);
- } else
- buf = rte_pktmbuf_alloc(rxq_ctrl->rxq.mp);
+ buf = rte_pktmbuf_alloc(rxq_ctrl->rxq.mp);
if (buf == NULL) {
- assert(pool == NULL);
ERROR("%p: empty mbuf pool", (void *)rxq_ctrl);
ret = ENOMEM;
goto error;
/* scat->addr must be able to store a pointer. */
assert(sizeof(scat->addr) >= sizeof(uintptr_t));
*scat = (struct mlx5_wqe_data_seg){
- .addr = htonll(rte_pktmbuf_mtod(buf, uintptr_t)),
- .byte_count = htonl(DATA_LEN(buf)),
- .lkey = htonl(rxq_ctrl->mr->lkey),
+ .addr =
+ rte_cpu_to_be_64(rte_pktmbuf_mtod(buf, uintptr_t)),
+ .byte_count = rte_cpu_to_be_32(DATA_LEN(buf)),
+ .lkey = rte_cpu_to_be_32(rxq_ctrl->mr->lkey),
};
(*rxq_ctrl->rxq.elts)[i] = buf;
}
+ if (rxq_check_vec_support(&rxq_ctrl->rxq) > 0) {
+ struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
+ struct rte_mbuf *mbuf_init = &rxq->fake_mbuf;
+
+ assert(rxq->elts_n == rxq->cqe_n);
+ /* Initialize default rearm_data for vPMD. */
+ mbuf_init->data_off = RTE_PKTMBUF_HEADROOM;
+ rte_mbuf_refcnt_set(mbuf_init, 1);
+ mbuf_init->nb_segs = 1;
+ mbuf_init->port = rxq->port_id;
+ /*
+ * prevent compiler reordering:
+ * rearm_data covers previous fields.
+ */
+ rte_compiler_barrier();
+ rxq->mbuf_initializer = *(uint64_t *)&mbuf_init->rearm_data;
+ /* Padding with a fake mbuf for vectorized Rx. */
+ for (i = 0; i < MLX5_VPMD_DESCS_PER_LOOP; ++i)
+ (*rxq->elts)[elts_n + i] = &rxq->fake_mbuf;
+ }
DEBUG("%p: allocated and configured %u segments (max %u packets)",
(void *)rxq_ctrl, elts_n, elts_n / (1 << rxq_ctrl->rxq.sges_n));
assert(ret == 0);
return 0;
error:
- assert(pool == NULL);
elts_n = i;
for (i = 0; (i != elts_n); ++i) {
if ((*rxq_ctrl->rxq.elts)[i] != NULL)
* Pointer to RX queue structure.
*/
static void
-rxq_free_elts(struct rxq_ctrl *rxq_ctrl)
+rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
{
- unsigned int i;
+ struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
+ const uint16_t q_n = (1 << rxq->elts_n);
+ const uint16_t q_mask = q_n - 1;
+ uint16_t used = q_n - (rxq->rq_ci - rxq->rq_pi);
+ uint16_t i;
DEBUG("%p: freeing WRs", (void *)rxq_ctrl);
- if (rxq_ctrl->rxq.elts == NULL)
+ if (rxq->elts == NULL)
return;
-
- for (i = 0; (i != (1u << rxq_ctrl->rxq.elts_n)); ++i) {
- if ((*rxq_ctrl->rxq.elts)[i] != NULL)
- rte_pktmbuf_free_seg((*rxq_ctrl->rxq.elts)[i]);
- (*rxq_ctrl->rxq.elts)[i] = NULL;
+ /**
+ * Some mbuf in the Ring belongs to the application. They cannot be
+ * freed.
+ */
+ if (rxq_check_vec_support(rxq) > 0) {
+ for (i = 0; i < used; ++i)
+ (*rxq->elts)[(rxq->rq_ci + i) & q_mask] = NULL;
+ rxq->rq_pi = rxq->rq_ci;
+ }
+ for (i = 0; (i != (1u << rxq->elts_n)); ++i) {
+ if ((*rxq->elts)[i] != NULL)
+ rte_pktmbuf_free_seg((*rxq->elts)[i]);
+ (*rxq->elts)[i] = NULL;
}
}
* Pointer to RX queue structure.
*/
void
-rxq_cleanup(struct rxq_ctrl *rxq_ctrl)
+mlx5_rxq_cleanup(struct mlx5_rxq_ctrl *rxq_ctrl)
{
- struct ibv_exp_release_intf_params params;
-
DEBUG("cleaning up %p", (void *)rxq_ctrl);
rxq_free_elts(rxq_ctrl);
- if (rxq_ctrl->fdir_queue != NULL)
- priv_fdir_queue_destroy(rxq_ctrl->priv, rxq_ctrl->fdir_queue);
- if (rxq_ctrl->if_wq != NULL) {
- assert(rxq_ctrl->priv != NULL);
- assert(rxq_ctrl->priv->ctx != NULL);
- assert(rxq_ctrl->wq != NULL);
- params = (struct ibv_exp_release_intf_params){
- .comp_mask = 0,
- };
- claim_zero(ibv_exp_release_intf(rxq_ctrl->priv->ctx,
- rxq_ctrl->if_wq,
- ¶ms));
- }
- if (rxq_ctrl->if_cq != NULL) {
- assert(rxq_ctrl->priv != NULL);
- assert(rxq_ctrl->priv->ctx != NULL);
- assert(rxq_ctrl->cq != NULL);
- params = (struct ibv_exp_release_intf_params){
- .comp_mask = 0,
- };
- claim_zero(ibv_exp_release_intf(rxq_ctrl->priv->ctx,
- rxq_ctrl->if_cq,
- ¶ms));
- }
if (rxq_ctrl->wq != NULL)
- claim_zero(ibv_exp_destroy_wq(rxq_ctrl->wq));
+ claim_zero(ibv_destroy_wq(rxq_ctrl->wq));
if (rxq_ctrl->cq != NULL)
claim_zero(ibv_destroy_cq(rxq_ctrl->cq));
- if (rxq_ctrl->rd != NULL) {
- struct ibv_exp_destroy_res_domain_attr attr = {
- .comp_mask = 0,
- };
-
- assert(rxq_ctrl->priv != NULL);
- assert(rxq_ctrl->priv->ctx != NULL);
- claim_zero(ibv_exp_destroy_res_domain(rxq_ctrl->priv->ctx,
- rxq_ctrl->rd,
- &attr));
- }
+ if (rxq_ctrl->channel != NULL)
+ claim_zero(ibv_destroy_comp_channel(rxq_ctrl->channel));
if (rxq_ctrl->mr != NULL)
claim_zero(ibv_dereg_mr(rxq_ctrl->mr));
memset(rxq_ctrl, 0, sizeof(*rxq_ctrl));
}
-/**
- * Reconfigure RX queue buffers.
- *
- * rxq_rehash() does not allocate mbufs, which, if not done from the right
- * thread (such as a control thread), may corrupt the pool.
- * In case of failure, the queue is left untouched.
- *
- * @param dev
- * Pointer to Ethernet device structure.
- * @param rxq_ctrl
- * RX queue pointer.
- *
- * @return
- * 0 on success, errno value on failure.
- */
-int
-rxq_rehash(struct rte_eth_dev *dev, struct rxq_ctrl *rxq_ctrl)
-{
- unsigned int elts_n = 1 << rxq_ctrl->rxq.elts_n;
- unsigned int i;
- struct ibv_exp_wq_attr mod;
- int err;
-
- DEBUG("%p: rehashing queue %p with %u SGE(s) per packet",
- (void *)dev, (void *)rxq_ctrl, 1 << rxq_ctrl->rxq.sges_n);
- assert(!(elts_n % (1 << rxq_ctrl->rxq.sges_n)));
- /* From now on, any failure will render the queue unusable.
- * Reinitialize WQ. */
- mod = (struct ibv_exp_wq_attr){
- .attr_mask = IBV_EXP_WQ_ATTR_STATE,
- .wq_state = IBV_EXP_WQS_RESET,
- };
- err = ibv_exp_modify_wq(rxq_ctrl->wq, &mod);
- if (err) {
- ERROR("%p: cannot reset WQ: %s", (void *)dev, strerror(err));
- assert(err > 0);
- return err;
- }
- /* Snatch mbufs from original queue. */
- claim_zero(rxq_alloc_elts(rxq_ctrl, elts_n, rxq_ctrl->rxq.elts));
- for (i = 0; i != elts_n; ++i) {
- struct rte_mbuf *buf = (*rxq_ctrl->rxq.elts)[i];
-
- assert(rte_mbuf_refcnt_read(buf) == 2);
- rte_pktmbuf_free_seg(buf);
- }
- /* Change queue state to ready. */
- mod = (struct ibv_exp_wq_attr){
- .attr_mask = IBV_EXP_WQ_ATTR_STATE,
- .wq_state = IBV_EXP_WQS_RDY,
- };
- err = ibv_exp_modify_wq(rxq_ctrl->wq, &mod);
- if (err) {
- ERROR("%p: WQ state to IBV_EXP_WQS_RDY failed: %s",
- (void *)dev, strerror(err));
- goto error;
- }
- /* Update doorbell counter. */
- rxq_ctrl->rxq.rq_ci = elts_n >> rxq_ctrl->rxq.sges_n;
- rte_wmb();
- *rxq_ctrl->rxq.rq_db = htonl(rxq_ctrl->rxq.rq_ci);
-error:
- assert(err >= 0);
- return err;
-}
-
/**
* Initialize RX queue.
*
* 0 on success, errno value on failure.
*/
static inline int
-rxq_setup(struct rxq_ctrl *tmpl)
+rxq_setup(struct mlx5_rxq_ctrl *tmpl)
{
struct ibv_cq *ibcq = tmpl->cq;
- struct mlx5_cq *cq = to_mxxx(cq, cq);
- struct mlx5_rwq *rwq = container_of(tmpl->wq, struct mlx5_rwq, wq);
- struct rte_mbuf *(*elts)[1 << tmpl->rxq.elts_n] =
+ struct mlx5dv_cq cq_info;
+ struct mlx5dv_rwq rwq;
+ const uint16_t desc_n =
+ (1 << tmpl->rxq.elts_n) + tmpl->priv->rx_vec_en *
+ MLX5_VPMD_DESCS_PER_LOOP;
+ struct rte_mbuf *(*elts)[desc_n] =
rte_calloc_socket("RXQ", 1, sizeof(*elts), 0, tmpl->socket);
+ struct mlx5dv_obj obj;
+ int ret = 0;
- if (cq->cqe_sz != RTE_CACHE_LINE_SIZE) {
+ obj.cq.in = ibcq;
+ obj.cq.out = &cq_info;
+ obj.rwq.in = tmpl->wq;
+ obj.rwq.out = &rwq;
+ ret = mlx5dv_init_obj(&obj, MLX5DV_OBJ_CQ | MLX5DV_OBJ_RWQ);
+ if (ret != 0) {
+ return -EINVAL;
+ }
+ if (cq_info.cqe_size != RTE_CACHE_LINE_SIZE) {
ERROR("Wrong MLX5_CQE_SIZE environment variable value: "
"it should be set to %u", RTE_CACHE_LINE_SIZE);
return EINVAL;
}
if (elts == NULL)
return ENOMEM;
- tmpl->rxq.rq_db = rwq->rq.db;
- tmpl->rxq.cqe_n = ibcq->cqe + 1;
+ tmpl->rxq.rq_db = rwq.dbrec;
+ tmpl->rxq.cqe_n = log2above(cq_info.cqe_cnt);
tmpl->rxq.cq_ci = 0;
tmpl->rxq.rq_ci = 0;
- tmpl->rxq.cq_db = cq->dbrec;
+ tmpl->rxq.rq_pi = 0;
+ tmpl->rxq.cq_db = cq_info.dbrec;
tmpl->rxq.wqes =
(volatile struct mlx5_wqe_data_seg (*)[])
- (uintptr_t)rwq->rq.buff;
+ (uintptr_t)rwq.buf;
tmpl->rxq.cqes =
(volatile struct mlx5_cqe (*)[])
- (uintptr_t)cq->active_buf->buf;
+ (uintptr_t)cq_info.buf;
tmpl->rxq.elts = elts;
+ tmpl->rxq.cq_uar = cq_info.cq_uar;
+ tmpl->rxq.cqn = cq_info.cqn;
+ tmpl->rxq.cq_arm_sn = 0;
return 0;
}
* @return
* 0 on success, errno value on failure.
*/
-int
-rxq_ctrl_setup(struct rte_eth_dev *dev, struct rxq_ctrl *rxq_ctrl,
+static int
+rxq_ctrl_setup(struct rte_eth_dev *dev, struct mlx5_rxq_ctrl *rxq_ctrl,
uint16_t desc, unsigned int socket,
const struct rte_eth_rxconf *conf, struct rte_mempool *mp)
{
struct priv *priv = dev->data->dev_private;
- struct rxq_ctrl tmpl = {
+ struct mlx5_rxq_ctrl tmpl = {
.priv = priv,
.socket = socket,
.rxq = {
.elts_n = log2above(desc),
.mp = mp,
+ .rss_hash = priv->rxqs_n > 1,
},
};
- struct ibv_exp_wq_attr mod;
+ struct ibv_wq_attr mod;
union {
- struct ibv_exp_query_intf_params params;
- struct ibv_exp_cq_init_attr cq;
- struct ibv_exp_res_domain_init_attr rd;
- struct ibv_exp_wq_init_attr wq;
- struct ibv_exp_cq_attr cq_attr;
+ struct ibv_cq_init_attr_ex cq;
+ struct ibv_wq_init_attr wq;
+ struct ibv_cq_ex cq_attr;
} attr;
- enum ibv_exp_query_intf_status status;
unsigned int mb_len = rte_pktmbuf_data_room_size(mp);
unsigned int cqe_n = desc - 1;
- struct rte_mbuf *(*elts)[desc] = NULL;
+ const uint16_t desc_n =
+ desc + priv->rx_vec_en * MLX5_VPMD_DESCS_PER_LOOP;
+ struct rte_mbuf *(*elts)[desc_n] = NULL;
int ret = 0;
(void)conf; /* Thresholds configuration (ignored). */
/* Enable scattered packets support for this queue if necessary. */
assert(mb_len >= RTE_PKTMBUF_HEADROOM);
- if ((dev->data->dev_conf.rxmode.jumbo_frame) &&
- (dev->data->dev_conf.rxmode.max_rx_pkt_len >
- (mb_len - RTE_PKTMBUF_HEADROOM))) {
+ if (dev->data->dev_conf.rxmode.max_rx_pkt_len <=
+ (mb_len - RTE_PKTMBUF_HEADROOM)) {
+ tmpl.rxq.sges_n = 0;
+ } else if (dev->data->dev_conf.rxmode.enable_scatter) {
unsigned int size =
RTE_PKTMBUF_HEADROOM +
dev->data->dev_conf.rxmode.max_rx_pkt_len;
dev->data->dev_conf.rxmode.max_rx_pkt_len);
return EOVERFLOW;
}
+ } else {
+ WARN("%p: the requested maximum Rx packet size (%u) is"
+ " larger than a single mbuf (%u) and scattered"
+ " mode has not been requested",
+ (void *)dev,
+ dev->data->dev_conf.rxmode.max_rx_pkt_len,
+ mb_len - RTE_PKTMBUF_HEADROOM);
}
DEBUG("%p: maximum number of segments per packet: %u",
(void *)dev, 1 << tmpl.rxq.sges_n);
(void *)dev, strerror(ret));
goto error;
}
- attr.rd = (struct ibv_exp_res_domain_init_attr){
- .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL |
- IBV_EXP_RES_DOMAIN_MSG_MODEL),
- .thread_model = IBV_EXP_THREAD_SINGLE,
- .msg_model = IBV_EXP_MSG_HIGH_BW,
- };
- tmpl.rd = ibv_exp_create_res_domain(priv->ctx, &attr.rd);
- if (tmpl.rd == NULL) {
- ret = ENOMEM;
- ERROR("%p: RD creation failure: %s",
- (void *)dev, strerror(ret));
- goto error;
+ if (dev->data->dev_conf.intr_conf.rxq) {
+ tmpl.channel = ibv_create_comp_channel(priv->ctx);
+ if (tmpl.channel == NULL) {
+ ret = ENOMEM;
+ ERROR("%p: Rx interrupt completion channel creation"
+ " failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
}
- attr.cq = (struct ibv_exp_cq_init_attr){
- .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN,
- .res_domain = tmpl.rd,
+ attr.cq = (struct ibv_cq_init_attr_ex){
+ .comp_mask = 0,
};
if (priv->cqe_comp) {
- attr.cq.comp_mask |= IBV_EXP_CQ_INIT_ATTR_FLAGS;
- attr.cq.flags |= IBV_EXP_CQ_COMPRESSED_CQE;
- cqe_n = (desc * 2) - 1; /* Double the number of CQEs. */
+ attr.cq.comp_mask |= IBV_CQ_INIT_ATTR_MASK_FLAGS;
+ attr.cq.flags |= MLX5DV_CQ_INIT_ATTR_MASK_COMPRESSED_CQE;
+ /*
+ * For vectorized Rx, it must not be doubled in order to
+ * make cq_ci and rq_ci aligned.
+ */
+ if (rxq_check_vec_support(&tmpl.rxq) < 0)
+ cqe_n = (desc * 2) - 1; /* Double the number of CQEs. */
}
- tmpl.cq = ibv_exp_create_cq(priv->ctx, cqe_n, NULL, NULL, 0,
- &attr.cq);
+ tmpl.cq = ibv_create_cq(priv->ctx, cqe_n, NULL, tmpl.channel, 0);
if (tmpl.cq == NULL) {
ret = ENOMEM;
ERROR("%p: CQ creation failure: %s",
goto error;
}
DEBUG("priv->device_attr.max_qp_wr is %d",
- priv->device_attr.max_qp_wr);
+ priv->device_attr.orig_attr.max_qp_wr);
DEBUG("priv->device_attr.max_sge is %d",
- priv->device_attr.max_sge);
+ priv->device_attr.orig_attr.max_sge);
/* Configure VLAN stripping. */
tmpl.rxq.vlan_strip = (priv->hw_vlan_strip &&
!!dev->data->dev_conf.rxmode.hw_vlan_strip);
- attr.wq = (struct ibv_exp_wq_init_attr){
+ attr.wq = (struct ibv_wq_init_attr){
.wq_context = NULL, /* Could be useful in the future. */
- .wq_type = IBV_EXP_WQT_RQ,
+ .wq_type = IBV_WQT_RQ,
/* Max number of outstanding WRs. */
- .max_recv_wr = desc >> tmpl.rxq.sges_n,
+ .max_wr = desc >> tmpl.rxq.sges_n,
/* Max number of scatter/gather elements in a WR. */
- .max_recv_sge = 1 << tmpl.rxq.sges_n,
+ .max_sge = 1 << tmpl.rxq.sges_n,
.pd = priv->pd,
.cq = tmpl.cq,
.comp_mask =
- IBV_EXP_CREATE_WQ_RES_DOMAIN |
- IBV_EXP_CREATE_WQ_VLAN_OFFLOADS |
+ IBV_WQ_FLAGS_CVLAN_STRIPPING |
0,
- .res_domain = tmpl.rd,
- .vlan_offloads = (tmpl.rxq.vlan_strip ?
- IBV_EXP_RECEIVE_WQ_CVLAN_STRIP :
- 0),
+ .create_flags = (tmpl.rxq.vlan_strip ?
+ IBV_WQ_FLAGS_CVLAN_STRIPPING :
+ 0),
};
/* By default, FCS (CRC) is stripped by hardware. */
if (dev->data->dev_conf.rxmode.hw_strip_crc) {
tmpl.rxq.crc_present = 0;
} else if (priv->hw_fcs_strip) {
/* Ask HW/Verbs to leave CRC in place when supported. */
- attr.wq.flags |= IBV_EXP_CREATE_WQ_FLAG_SCATTER_FCS;
- attr.wq.comp_mask |= IBV_EXP_CREATE_WQ_FLAGS;
+ attr.wq.create_flags |= IBV_WQ_FLAGS_SCATTER_FCS;
+ attr.wq.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
tmpl.rxq.crc_present = 1;
} else {
WARN("%p: CRC stripping has been disabled but will still"
(void *)dev,
tmpl.rxq.crc_present ? "disabled" : "enabled",
tmpl.rxq.crc_present << 2);
+#ifdef HAVE_IBV_WQ_FLAG_RX_END_PADDING
if (!mlx5_getenv_int("MLX5_PMD_ENABLE_PADDING"))
; /* Nothing else to do. */
else if (priv->hw_padding) {
INFO("%p: enabling packet padding on queue %p",
(void *)dev, (void *)rxq_ctrl);
- attr.wq.flags |= IBV_EXP_CREATE_WQ_FLAG_RX_END_PADDING;
- attr.wq.comp_mask |= IBV_EXP_CREATE_WQ_FLAGS;
+ attr.wq.create_flags |= IBV_WQ_FLAG_RX_END_PADDING;
+ attr.wq.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
} else
WARN("%p: packet padding has been requested but is not"
" supported, make sure MLNX_OFED and firmware are"
" up to date",
(void *)dev);
+#endif
- tmpl.wq = ibv_exp_create_wq(priv->ctx, &attr.wq);
+ tmpl.wq = ibv_create_wq(priv->ctx, &attr.wq);
if (tmpl.wq == NULL) {
ret = (errno ? errno : EINVAL);
ERROR("%p: WQ creation failure: %s",
* Make sure number of WRs*SGEs match expectations since a queue
* cannot allocate more than "desc" buffers.
*/
- if (((int)attr.wq.max_recv_wr != (desc >> tmpl.rxq.sges_n)) ||
- ((int)attr.wq.max_recv_sge != (1 << tmpl.rxq.sges_n))) {
+ if (((int)attr.wq.max_wr != (desc >> tmpl.rxq.sges_n)) ||
+ ((int)attr.wq.max_sge != (1 << tmpl.rxq.sges_n))) {
ERROR("%p: requested %u*%u but got %u*%u WRs*SGEs",
(void *)dev,
(desc >> tmpl.rxq.sges_n), (1 << tmpl.rxq.sges_n),
- attr.wq.max_recv_wr, attr.wq.max_recv_sge);
+ attr.wq.max_wr, attr.wq.max_sge);
ret = EINVAL;
goto error;
}
/* Save port ID. */
tmpl.rxq.port_id = dev->data->port_id;
DEBUG("%p: RTE port ID: %u", (void *)rxq_ctrl, tmpl.rxq.port_id);
- attr.params = (struct ibv_exp_query_intf_params){
- .intf_scope = IBV_EXP_INTF_GLOBAL,
- .intf_version = 1,
- .intf = IBV_EXP_INTF_CQ,
- .obj = tmpl.cq,
- };
- tmpl.if_cq = ibv_exp_query_intf(priv->ctx, &attr.params, &status);
- if (tmpl.if_cq == NULL) {
- ERROR("%p: CQ interface family query failed with status %d",
- (void *)dev, status);
- goto error;
- }
- attr.params = (struct ibv_exp_query_intf_params){
- .intf_scope = IBV_EXP_INTF_GLOBAL,
- .intf = IBV_EXP_INTF_WQ,
- .obj = tmpl.wq,
- };
- tmpl.if_wq = ibv_exp_query_intf(priv->ctx, &attr.params, &status);
- if (tmpl.if_wq == NULL) {
- ERROR("%p: WQ interface family query failed with status %d",
- (void *)dev, status);
- goto error;
- }
/* Change queue state to ready. */
- mod = (struct ibv_exp_wq_attr){
- .attr_mask = IBV_EXP_WQ_ATTR_STATE,
- .wq_state = IBV_EXP_WQS_RDY,
+ mod = (struct ibv_wq_attr){
+ .attr_mask = IBV_WQ_ATTR_STATE,
+ .wq_state = IBV_WQS_RDY,
};
- ret = ibv_exp_modify_wq(tmpl.wq, &mod);
+ ret = ibv_modify_wq(tmpl.wq, &mod);
if (ret) {
- ERROR("%p: WQ state to IBV_EXP_WQS_RDY failed: %s",
+ ERROR("%p: WQ state to IBV_WQS_RDY failed: %s",
(void *)dev, strerror(ret));
goto error;
}
(void *)dev, strerror(ret));
goto error;
}
- /* Reuse buffers from original queue if possible. */
- if (rxq_ctrl->rxq.elts_n) {
- assert(1 << rxq_ctrl->rxq.elts_n == desc);
- assert(rxq_ctrl->rxq.elts != tmpl.rxq.elts);
- ret = rxq_alloc_elts(&tmpl, desc, rxq_ctrl->rxq.elts);
- } else
- ret = rxq_alloc_elts(&tmpl, desc, NULL);
+ ret = rxq_alloc_elts(&tmpl, desc);
if (ret) {
ERROR("%p: RXQ allocation failed: %s",
(void *)dev, strerror(ret));
}
/* Clean up rxq in case we're reinitializing it. */
DEBUG("%p: cleaning-up old rxq just in case", (void *)rxq_ctrl);
- rxq_cleanup(rxq_ctrl);
+ mlx5_rxq_cleanup(rxq_ctrl);
/* Move mbuf pointers to dedicated storage area in RX queue. */
elts = (void *)(rxq_ctrl + 1);
rte_memcpy(elts, tmpl.rxq.elts, sizeof(*elts));
/* Update doorbell counter. */
rxq_ctrl->rxq.rq_ci = desc >> rxq_ctrl->rxq.sges_n;
rte_wmb();
- *rxq_ctrl->rxq.rq_db = htonl(rxq_ctrl->rxq.rq_ci);
+ *rxq_ctrl->rxq.rq_db = rte_cpu_to_be_32(rxq_ctrl->rxq.rq_ci);
DEBUG("%p: rxq updated with %p", (void *)rxq_ctrl, (void *)&tmpl);
assert(ret == 0);
return 0;
error:
elts = tmpl.rxq.elts;
- rxq_cleanup(&tmpl);
+ mlx5_rxq_cleanup(&tmpl);
rte_free(elts);
assert(ret > 0);
return ret;
struct rte_mempool *mp)
{
struct priv *priv = dev->data->dev_private;
- struct rxq *rxq = (*priv->rxqs)[idx];
- struct rxq_ctrl *rxq_ctrl = container_of(rxq, struct rxq_ctrl, rxq);
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ const uint16_t desc_n =
+ desc + priv->rx_vec_en * MLX5_VPMD_DESCS_PER_LOOP;
int ret;
if (mlx5_is_secondary())
if (rxq != NULL) {
DEBUG("%p: reusing already allocated queue index %u (%p)",
(void *)dev, idx, (void *)rxq);
- if (priv->started) {
+ if (dev->data->dev_started) {
priv_unlock(priv);
return -EEXIST;
}
(*priv->rxqs)[idx] = NULL;
- rxq_cleanup(rxq_ctrl);
+ mlx5_rxq_cleanup(rxq_ctrl);
+ /* Resize if rxq size is changed. */
+ if (rxq_ctrl->rxq.elts_n != log2above(desc)) {
+ rxq_ctrl = rte_realloc(rxq_ctrl,
+ sizeof(*rxq_ctrl) + desc_n *
+ sizeof(struct rte_mbuf *),
+ RTE_CACHE_LINE_SIZE);
+ if (!rxq_ctrl) {
+ ERROR("%p: unable to reallocate queue index %u",
+ (void *)dev, idx);
+ priv_unlock(priv);
+ return -ENOMEM;
+ }
+ }
} else {
rxq_ctrl = rte_calloc_socket("RXQ", 1, sizeof(*rxq_ctrl) +
- desc * sizeof(struct rte_mbuf *),
+ desc_n *
+ sizeof(struct rte_mbuf *),
0, socket);
if (rxq_ctrl == NULL) {
ERROR("%p: unable to allocate queue index %u",
DEBUG("%p: adding RX queue %p to list",
(void *)dev, (void *)rxq_ctrl);
(*priv->rxqs)[idx] = &rxq_ctrl->rxq;
- /* Update receive callback. */
- dev->rx_pkt_burst = mlx5_rx_burst;
}
priv_unlock(priv);
return -ret;
void
mlx5_rx_queue_release(void *dpdk_rxq)
{
- struct rxq *rxq = (struct rxq *)dpdk_rxq;
- struct rxq_ctrl *rxq_ctrl;
+ struct mlx5_rxq_data *rxq = (struct mlx5_rxq_data *)dpdk_rxq;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
struct priv *priv;
unsigned int i;
if (rxq == NULL)
return;
- rxq_ctrl = container_of(rxq, struct rxq_ctrl, rxq);
+ rxq_ctrl = container_of(rxq, struct mlx5_rxq_ctrl, rxq);
priv = rxq_ctrl->priv;
priv_lock(priv);
+ if (priv_flow_rxq_in_use(priv, rxq))
+ rte_panic("Rx queue %p is still used by a flow and cannot be"
+ " removed\n", (void *)rxq_ctrl);
for (i = 0; (i != priv->rxqs_n); ++i)
if ((*priv->rxqs)[i] == rxq) {
DEBUG("%p: removing RX queue %p from list",
(*priv->rxqs)[i] = NULL;
break;
}
- rxq_cleanup(rxq_ctrl);
+ mlx5_rxq_cleanup(rxq_ctrl);
rte_free(rxq_ctrl);
priv_unlock(priv);
}
/**
- * DPDK callback for RX in secondary processes.
- *
- * This function configures all queues from primary process information
- * if necessary before reverting to the normal RX burst callback.
+ * Allocate queue vector and fill epoll fd list for Rx interrupts.
*
- * @param dpdk_rxq
- * Generic pointer to RX queue structure.
- * @param[out] pkts
- * Array to store received packets.
- * @param pkts_n
- * Maximum number of packets in array.
+ * @param priv
+ * Pointer to private structure.
*
* @return
- * Number of packets successfully received (<= pkts_n).
+ * 0 on success, negative on failure.
*/
-uint16_t
-mlx5_rx_burst_secondary_setup(void *dpdk_rxq, struct rte_mbuf **pkts,
- uint16_t pkts_n)
+int
+priv_rx_intr_vec_enable(struct priv *priv)
{
- struct rxq *rxq = dpdk_rxq;
- struct rxq_ctrl *rxq_ctrl = container_of(rxq, struct rxq_ctrl, rxq);
- struct priv *priv = mlx5_secondary_data_setup(rxq_ctrl->priv);
- struct priv *primary_priv;
- unsigned int index;
+ unsigned int i;
+ unsigned int rxqs_n = priv->rxqs_n;
+ unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
+ unsigned int count = 0;
+ struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
- if (priv == NULL)
+ assert(!mlx5_is_secondary());
+ if (!priv->dev->data->dev_conf.intr_conf.rxq)
return 0;
- primary_priv =
- mlx5_secondary_data[priv->dev->data->port_id].primary_priv;
- /* Look for queue index in both private structures. */
- for (index = 0; index != priv->rxqs_n; ++index)
- if (((*primary_priv->rxqs)[index] == rxq) ||
- ((*priv->rxqs)[index] == rxq))
- break;
- if (index == priv->rxqs_n)
- return 0;
- rxq = (*priv->rxqs)[index];
- return priv->dev->rx_pkt_burst(rxq, pkts, pkts_n);
+ priv_rx_intr_vec_disable(priv);
+ intr_handle->intr_vec = malloc(sizeof(intr_handle->intr_vec[rxqs_n]));
+ if (intr_handle->intr_vec == NULL) {
+ ERROR("failed to allocate memory for interrupt vector,"
+ " Rx interrupts will not be supported");
+ return -ENOMEM;
+ }
+ intr_handle->type = RTE_INTR_HANDLE_EXT;
+ for (i = 0; i != n; ++i) {
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ int fd;
+ int flags;
+ int rc;
+
+ /* Skip queues that cannot request interrupts. */
+ if (!rxq || !rxq_ctrl->channel) {
+ /* Use invalid intr_vec[] index to disable entry. */
+ intr_handle->intr_vec[i] =
+ RTE_INTR_VEC_RXTX_OFFSET +
+ RTE_MAX_RXTX_INTR_VEC_ID;
+ continue;
+ }
+ if (count >= RTE_MAX_RXTX_INTR_VEC_ID) {
+ ERROR("too many Rx queues for interrupt vector size"
+ " (%d), Rx interrupts cannot be enabled",
+ RTE_MAX_RXTX_INTR_VEC_ID);
+ priv_rx_intr_vec_disable(priv);
+ return -1;
+ }
+ fd = rxq_ctrl->channel->fd;
+ flags = fcntl(fd, F_GETFL);
+ rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
+ if (rc < 0) {
+ ERROR("failed to make Rx interrupt file descriptor"
+ " %d non-blocking for queue index %d", fd, i);
+ priv_rx_intr_vec_disable(priv);
+ return -1;
+ }
+ intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count;
+ intr_handle->efds[count] = fd;
+ count++;
+ }
+ if (!count)
+ priv_rx_intr_vec_disable(priv);
+ else
+ intr_handle->nb_efd = count;
+ return 0;
+}
+
+/**
+ * Clean up Rx interrupts handler.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+void
+priv_rx_intr_vec_disable(struct priv *priv)
+{
+ struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+
+ rte_intr_free_epoll_fd(intr_handle);
+ free(intr_handle->intr_vec);
+ intr_handle->nb_efd = 0;
+ intr_handle->intr_vec = NULL;
+}
+
+/**
+ * MLX5 CQ notification .
+ *
+ * @param rxq
+ * Pointer to receive queue structure.
+ * @param sq_n_rxq
+ * Sequence number per receive queue .
+ */
+static inline void
+mlx5_arm_cq(struct mlx5_rxq_data *rxq, int sq_n_rxq)
+{
+ int sq_n = 0;
+ uint32_t doorbell_hi;
+ uint64_t doorbell;
+ void *cq_db_reg = (char *)rxq->cq_uar + MLX5_CQ_DOORBELL;
+
+ sq_n = sq_n_rxq & MLX5_CQ_SQN_MASK;
+ doorbell_hi = sq_n << MLX5_CQ_SQN_OFFSET | (rxq->cq_ci & MLX5_CI_MASK);
+ doorbell = (uint64_t)doorbell_hi << 32;
+ doorbell |= rxq->cqn;
+ rxq->cq_db[MLX5_CQ_ARM_DB] = rte_cpu_to_be_32(doorbell_hi);
+ rte_wmb();
+ rte_write64(rte_cpu_to_be_64(doorbell), cq_db_reg);
+}
+
+/**
+ * DPDK callback for Rx queue interrupt enable.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param rx_queue_id
+ * Rx queue number.
+ *
+ * @return
+ * 0 on success, negative on failure.
+ */
+int
+mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[rx_queue_id];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ int ret = 0;
+
+ if (!rxq || !rxq_ctrl->channel) {
+ ret = EINVAL;
+ } else {
+ mlx5_arm_cq(rxq, rxq->cq_arm_sn);
+ }
+ if (ret)
+ WARN("unable to arm interrupt on rx queue %d", rx_queue_id);
+ return -ret;
+}
+
+/**
+ * DPDK callback for Rx queue interrupt disable.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param rx_queue_id
+ * Rx queue number.
+ *
+ * @return
+ * 0 on success, negative on failure.
+ */
+int
+mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[rx_queue_id];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ struct ibv_cq *ev_cq;
+ void *ev_ctx;
+ int ret;
+
+ if (!rxq || !rxq_ctrl->channel) {
+ ret = EINVAL;
+ } else {
+ ret = ibv_get_cq_event(rxq_ctrl->cq->channel, &ev_cq, &ev_ctx);
+ rxq->cq_arm_sn++;
+ if (ret || ev_cq != rxq_ctrl->cq)
+ ret = EINVAL;
+ }
+ if (ret)
+ WARN("unable to disable interrupt on rx queue %d",
+ rx_queue_id);
+ else
+ ibv_ack_cq_events(rxq_ctrl->cq, 1);
+ return -ret;
}