X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx5%2Fmlx5_flow.c;h=bf8eff8a5004975fcada1a1e545593c2be2335c1;hb=3669a1af893088064778f14fba0c2ca88819e095;hp=3321b3eef7d1b047e8fffd45a304158a4355cb79;hpb=0887aa7f27f3698b883ebe63abf74484305e71c4;p=dpdk.git diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c index 3321b3eef7..bf8eff8a50 100644 --- a/drivers/net/mlx5/mlx5_flow.c +++ b/drivers/net/mlx5/mlx5_flow.c @@ -44,12 +44,14 @@ #pragma GCC diagnostic error "-Wpedantic" #endif -#include +#include #include #include #include +#include #include "mlx5.h" +#include "mlx5_defs.h" #include "mlx5_prm.h" /* Define minimal priority for control plane flows. */ @@ -59,6 +61,25 @@ #define MLX5_IPV4 4 #define MLX5_IPV6 6 +#ifndef HAVE_IBV_DEVICE_COUNTERS_SET_SUPPORT +struct ibv_counter_set_init_attr { + int dummy; +}; +struct ibv_flow_spec_counter_action { + int dummy; +}; +struct ibv_counter_set { + int dummy; +}; + +static inline int +ibv_destroy_counter_set(struct ibv_counter_set *cs) +{ + (void)cs; + return -ENOTSUP; +} +#endif + /* Dev ops structure defined in mlx5.c */ extern const struct eth_dev_ops mlx5_dev_ops; extern const struct eth_dev_ops mlx5_dev_ops_isolate; @@ -107,6 +128,9 @@ mlx5_flow_create_copy(struct mlx5_flow_parse *parser, void *src, static int mlx5_flow_create_flag_mark(struct mlx5_flow_parse *parser, uint32_t mark_id); +static int +mlx5_flow_create_count(struct priv *priv, struct mlx5_flow_parse *parser); + /* Hash RX queue types. */ enum hash_rxq_type { HASH_RXQ_TCPV4, @@ -190,6 +214,12 @@ const struct hash_rxq_init hash_rxq_init[] = { /* Number of entries in hash_rxq_init[]. */ const unsigned int hash_rxq_init_n = RTE_DIM(hash_rxq_init); +/** Structure for holding counter stats. */ +struct mlx5_flow_counter_stats { + uint64_t hits; /**< Number of packets matched by the rule. */ + uint64_t bytes; /**< Number of bytes matched by the rule. */ +}; + /** Structure for Drop queue. */ struct mlx5_hrxq_drop { struct ibv_rwq_ind_table *ind_table; /**< Indirection table. */ @@ -220,11 +250,10 @@ struct rte_flow { uint16_t (*queues)[]; /**< Queues indexes to use. */ struct rte_eth_rss_conf rss_conf; /**< RSS configuration */ uint8_t rss_key[40]; /**< copy of the RSS key. */ - union { - struct mlx5_flow frxq[RTE_DIM(hash_rxq_init)]; - /**< Flow with Rx queue. */ - struct mlx5_flow_drop drxq; /**< Flow with drop Rx queue. */ - }; + struct ibv_counter_set *cs; /**< Holds the counters for the rule. */ + struct mlx5_flow_counter_stats counter_stats;/**mask) { unsigned int i; - const uint8_t *spec = item->mask; + const uint8_t *spec = item->spec; for (i = 0; i < size; ++i) if ((spec[i] | mask[i]) != mask[i]) @@ -538,9 +571,15 @@ priv_flow_convert_rss_conf(struct priv *priv, struct mlx5_flow_parse *parser, const struct rte_eth_rss_conf *rss_conf) { - const struct rte_eth_rss_conf *rss = - rss_conf ? rss_conf : &priv->rss_conf; + const struct rte_eth_rss_conf *rss; + if (rss_conf) { + if (rss_conf->rss_hf & MLX5_RSS_HF_MASK) + return EINVAL; + rss = rss_conf; + } else { + rss = &priv->rss_conf; + } if (rss->rss_key_len > 40) return EINVAL; parser->rss_conf.rss_key_len = rss->rss_key_len; @@ -740,10 +779,15 @@ priv_flow_convert_actions(struct priv *priv, parser->mark_id = mark->id; } else if (actions->type == RTE_FLOW_ACTION_TYPE_FLAG) { parser->mark = 1; + } else if (actions->type == RTE_FLOW_ACTION_TYPE_COUNT && + priv->config.flow_counter_en) { + parser->count = 1; } else { goto exit_action_not_supported; } } + if (parser->drop && parser->mark) + parser->mark = 0; if (!parser->queues_n && !parser->drop) { rte_flow_error_set(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_HANDLE, NULL, "no valid action"); @@ -782,12 +826,8 @@ priv_flow_convert_items_validate(struct priv *priv, (void)priv; /* Initialise the offsets to start after verbs attribute. */ - if (parser->drop) { - parser->drop_q.offset = sizeof(struct ibv_flow_attr); - } else { - for (i = 0; i != hash_rxq_init_n; ++i) - parser->queue[i].offset = sizeof(struct ibv_flow_attr); - } + for (i = 0; i != hash_rxq_init_n; ++i) + parser->queue[i].offset = sizeof(struct ibv_flow_attr); for (; items->type != RTE_FLOW_ITEM_TYPE_END; ++items) { const struct mlx5_flow_items *token = NULL; unsigned int n; @@ -821,22 +861,30 @@ priv_flow_convert_items_validate(struct priv *priv, " VXLAN encapsulations"); return -rte_errno; } - parser->inner = 1; + parser->inner = IBV_FLOW_SPEC_INNER; } - if (parser->drop) { - parser->drop_q.offset += cur_item->dst_sz; - } else if (parser->queues_n == 1) { + if (parser->drop || parser->queues_n == 1) { parser->queue[HASH_RXQ_ETH].offset += cur_item->dst_sz; } else { for (n = 0; n != hash_rxq_init_n; ++n) parser->queue[n].offset += cur_item->dst_sz; } } + if (parser->drop) { + parser->queue[HASH_RXQ_ETH].offset += + sizeof(struct ibv_flow_spec_action_drop); + } if (parser->mark) { for (i = 0; i != hash_rxq_init_n; ++i) parser->queue[i].offset += sizeof(struct ibv_flow_spec_action_tag); } + if (parser->count) { + unsigned int size = sizeof(struct ibv_flow_spec_counter_action); + + for (i = 0; i != hash_rxq_init_n; ++i) + parser->queue[i].offset += size; + } return 0; exit_item_not_supported: rte_flow_error_set(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ITEM, @@ -955,15 +1003,13 @@ fill: if (hash_rxq_init[i].ip_version == MLX5_IPV4) { size = sizeof(struct ibv_flow_spec_ipv4_ext); specs.ipv4 = (struct ibv_flow_spec_ipv4_ext){ - .type = IBV_FLOW_SPEC_IPV4_EXT | - parser->inner, + .type = IBV_FLOW_SPEC_IPV4_EXT, .size = size, }; } else { size = sizeof(struct ibv_flow_spec_ipv6); specs.ipv6 = (struct ibv_flow_spec_ipv6){ - .type = IBV_FLOW_SPEC_IPV6 | - parser->inner, + .type = IBV_FLOW_SPEC_IPV6, .size = size, }; } @@ -983,8 +1029,7 @@ fill: .type = ((i == HASH_RXQ_UDPV4 || i == HASH_RXQ_UDPV6) ? IBV_FLOW_SPEC_UDP : - IBV_FLOW_SPEC_TCP) | - parser->inner, + IBV_FLOW_SPEC_TCP), .size = size, }; if (parser->queue[i].ibv_attr) { @@ -1050,15 +1095,7 @@ priv_flow_convert(struct priv *priv, * Second step. * Allocate the memory space to store verbs specifications. */ - if (parser->drop) { - parser->drop_q.ibv_attr = - priv_flow_convert_allocate(priv, attr->priority, - parser->drop_q.offset, - error); - if (!parser->drop_q.ibv_attr) - return ENOMEM; - parser->drop_q.offset = sizeof(struct ibv_flow_attr); - } else if (parser->queues_n == 1) { + if (parser->drop || parser->queues_n == 1) { unsigned int priority = attr->priority + hash_rxq_init[HASH_RXQ_ETH].flow_priority; @@ -1103,7 +1140,7 @@ priv_flow_convert(struct priv *priv, cur_item->mask), parser); if (ret) { - rte_flow_error_set(error, ENOTSUP, + rte_flow_error_set(error, ret, RTE_FLOW_ERROR_TYPE_ITEM, items, "item not supported"); goto exit_free; @@ -1111,19 +1148,41 @@ priv_flow_convert(struct priv *priv, } if (parser->mark) mlx5_flow_create_flag_mark(parser, parser->mark_id); + if (parser->count && parser->create) { + mlx5_flow_create_count(priv, parser); + if (!parser->cs) + goto exit_count_error; + } /* * Last step. Complete missing specification to reach the RSS * configuration. */ - if (parser->queues_n > 1) + if (parser->queues_n > 1) { priv_flow_convert_finalise(priv, parser); + } else { + /* + * Action queue have their priority overridden with + * Ethernet priority, this priority needs to be adjusted to + * their most specific layer priority. + */ + parser->queue[HASH_RXQ_ETH].ibv_attr->priority = + attr->priority + + hash_rxq_init[parser->layer].flow_priority; + } + if (parser->allmulti && + parser->layer == HASH_RXQ_ETH) { + for (i = 0; i != hash_rxq_init_n; ++i) { + if (!parser->queue[i].ibv_attr) + continue; + if (parser->queue[i].ibv_attr->num_of_specs != 1) + break; + parser->queue[i].ibv_attr->type = + IBV_FLOW_ATTR_MC_DEFAULT; + } + } exit_free: /* Only verification is expected, all resources should be released. */ if (!parser->create) { - if (parser->drop) { - rte_free(parser->drop_q.ibv_attr); - parser->drop_q.ibv_attr = NULL; - } for (i = 0; i != hash_rxq_init_n; ++i) { if (parser->queue[i].ibv_attr) { rte_free(parser->queue[i].ibv_attr); @@ -1142,6 +1201,10 @@ exit_enomem: rte_flow_error_set(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, "cannot allocate verbs spec attributes."); return ret; +exit_count_error: + rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_UNSPECIFIED, + NULL, "cannot create counter."); + return rte_errno; } /** @@ -1161,14 +1224,6 @@ mlx5_flow_create_copy(struct mlx5_flow_parse *parser, void *src, unsigned int i; void *dst; - if (parser->drop) { - dst = (void *)((uintptr_t)parser->drop_q.ibv_attr + - parser->drop_q.offset); - memcpy(dst, src, size); - ++parser->drop_q.ibv_attr->num_of_specs; - parser->drop_q.offset += size; - return; - } for (i = 0; i != hash_rxq_init_n; ++i) { if (!parser->queue[i].ibv_attr) continue; @@ -1210,7 +1265,9 @@ mlx5_flow_create_eth(const struct rte_flow_item *item, .size = eth_size, }; - parser->layer = HASH_RXQ_ETH; + /* Don't update layer for the inner pattern. */ + if (!parser->inner) + parser->layer = HASH_RXQ_ETH; if (spec) { unsigned int i; @@ -1230,6 +1287,7 @@ mlx5_flow_create_eth(const struct rte_flow_item *item, eth.val.ether_type &= eth.mask.ether_type; } mlx5_flow_create_copy(parser, ð, eth_size); + parser->allmulti = eth.val.dst_mac[0] & 1; return 0; } @@ -1259,14 +1317,6 @@ mlx5_flow_create_vlan(const struct rte_flow_item *item, if (!mask) mask = default_mask; - if (parser->drop) { - eth = (void *)((uintptr_t)parser->drop_q.ibv_attr + - parser->drop_q.offset - eth_size); - eth->val.vlan_tag = spec->tci; - eth->mask.vlan_tag = mask->tci; - eth->val.vlan_tag &= eth->mask.vlan_tag; - return 0; - } for (i = 0; i != hash_rxq_init_n; ++i) { if (!parser->queue[i].ibv_attr) continue; @@ -1305,7 +1355,9 @@ mlx5_flow_create_ipv4(const struct rte_flow_item *item, .size = ipv4_size, }; - parser->layer = HASH_RXQ_IPV4; + /* Don't update layer for the inner pattern. */ + if (!parser->inner) + parser->layer = HASH_RXQ_IPV4; if (spec) { if (!mask) mask = default_mask; @@ -1355,9 +1407,13 @@ mlx5_flow_create_ipv6(const struct rte_flow_item *item, .size = ipv6_size, }; - parser->layer = HASH_RXQ_IPV6; + /* Don't update layer for the inner pattern. */ + if (!parser->inner) + parser->layer = HASH_RXQ_IPV6; if (spec) { unsigned int i; + uint32_t vtc_flow_val; + uint32_t vtc_flow_mask; if (!mask) mask = default_mask; @@ -1369,7 +1425,20 @@ mlx5_flow_create_ipv6(const struct rte_flow_item *item, RTE_DIM(ipv6.mask.src_ip)); memcpy(&ipv6.mask.dst_ip, mask->hdr.dst_addr, RTE_DIM(ipv6.mask.dst_ip)); - ipv6.mask.flow_label = mask->hdr.vtc_flow; + vtc_flow_val = rte_be_to_cpu_32(spec->hdr.vtc_flow); + vtc_flow_mask = rte_be_to_cpu_32(mask->hdr.vtc_flow); + ipv6.val.flow_label = + rte_cpu_to_be_32((vtc_flow_val & IPV6_HDR_FL_MASK) >> + IPV6_HDR_FL_SHIFT); + ipv6.val.traffic_class = (vtc_flow_val & IPV6_HDR_TC_MASK) >> + IPV6_HDR_TC_SHIFT; + ipv6.val.next_hdr = spec->hdr.proto; + ipv6.val.hop_limit = spec->hdr.hop_limits; + ipv6.mask.flow_label = + rte_cpu_to_be_32((vtc_flow_mask & IPV6_HDR_FL_MASK) >> + IPV6_HDR_FL_SHIFT); + ipv6.mask.traffic_class = (vtc_flow_mask & IPV6_HDR_TC_MASK) >> + IPV6_HDR_TC_SHIFT; ipv6.mask.next_hdr = mask->hdr.proto; ipv6.mask.hop_limit = mask->hdr.hop_limits; /* Remove unwanted bits from values. */ @@ -1378,6 +1447,7 @@ mlx5_flow_create_ipv6(const struct rte_flow_item *item, ipv6.val.dst_ip[i] &= ipv6.mask.dst_ip[i]; } ipv6.val.flow_label &= ipv6.mask.flow_label; + ipv6.val.traffic_class &= ipv6.mask.traffic_class; ipv6.val.next_hdr &= ipv6.mask.next_hdr; ipv6.val.hop_limit &= ipv6.mask.hop_limit; } @@ -1409,10 +1479,13 @@ mlx5_flow_create_udp(const struct rte_flow_item *item, .size = udp_size, }; - if (parser->layer == HASH_RXQ_IPV4) - parser->layer = HASH_RXQ_UDPV4; - else - parser->layer = HASH_RXQ_UDPV6; + /* Don't update layer for the inner pattern. */ + if (!parser->inner) { + if (parser->layer == HASH_RXQ_IPV4) + parser->layer = HASH_RXQ_UDPV4; + else + parser->layer = HASH_RXQ_UDPV6; + } if (spec) { if (!mask) mask = default_mask; @@ -1452,10 +1525,13 @@ mlx5_flow_create_tcp(const struct rte_flow_item *item, .size = tcp_size, }; - if (parser->layer == HASH_RXQ_IPV4) - parser->layer = HASH_RXQ_TCPV4; - else - parser->layer = HASH_RXQ_TCPV6; + /* Don't update layer for the inner pattern. */ + if (!parser->inner) { + if (parser->layer == HASH_RXQ_IPV4) + parser->layer = HASH_RXQ_TCPV4; + else + parser->layer = HASH_RXQ_TCPV6; + } if (spec) { if (!mask) mask = default_mask; @@ -1511,6 +1587,16 @@ mlx5_flow_create_vxlan(const struct rte_flow_item *item, /* Remove unwanted bits from values. */ vxlan.val.tunnel_id &= vxlan.mask.tunnel_id; } + /* + * Tunnel id 0 is equivalent as not adding a VXLAN layer, if only this + * layer is defined in the Verbs specification it is interpreted as + * wildcard and all packets will match this rule, if it follows a full + * stack layer (ex: eth / ipv4 / udp), all packets matching the layers + * before will also match this rule. + * To avoid such situation, VNI 0 is currently refused. + */ + if (!vxlan.val.tunnel_id) + return EINVAL; mlx5_flow_create_copy(parser, &vxlan, size); return 0; } @@ -1538,6 +1624,40 @@ mlx5_flow_create_flag_mark(struct mlx5_flow_parse *parser, uint32_t mark_id) return 0; } +/** + * Convert count action to Verbs specification. + * + * @param priv + * Pointer to private structure. + * @param parser + * Pointer to MLX5 flow parser structure. + * + * @return + * 0 on success, errno value on failure. + */ +static int +mlx5_flow_create_count(struct priv *priv __rte_unused, + struct mlx5_flow_parse *parser __rte_unused) +{ +#ifdef HAVE_IBV_DEVICE_COUNTERS_SET_SUPPORT + unsigned int size = sizeof(struct ibv_flow_spec_counter_action); + struct ibv_counter_set_init_attr init_attr = {0}; + struct ibv_flow_spec_counter_action counter = { + .type = IBV_FLOW_SPEC_ACTION_COUNT, + .size = size, + .counter_set_handle = 0, + }; + + init_attr.counter_set_id = 0; + parser->cs = ibv_create_counter_set(priv->ctx, &init_attr); + if (!parser->cs) + return EINVAL; + counter.counter_set_handle = parser->cs->handle; + mlx5_flow_create_copy(parser, &counter, size); +#endif + return 0; +} + /** * Complete flow rule creation with a drop queue. * @@ -1566,21 +1686,25 @@ priv_flow_create_action_queue_drop(struct priv *priv, assert(priv->pd); assert(priv->ctx); flow->drop = 1; - drop = (void *)((uintptr_t)parser->drop_q.ibv_attr + - parser->drop_q.offset); + drop = (void *)((uintptr_t)parser->queue[HASH_RXQ_ETH].ibv_attr + + parser->queue[HASH_RXQ_ETH].offset); *drop = (struct ibv_flow_spec_action_drop){ .type = IBV_FLOW_SPEC_ACTION_DROP, .size = size, }; - ++parser->drop_q.ibv_attr->num_of_specs; - parser->drop_q.offset += size; + ++parser->queue[HASH_RXQ_ETH].ibv_attr->num_of_specs; + parser->queue[HASH_RXQ_ETH].offset += size; + flow->frxq[HASH_RXQ_ETH].ibv_attr = + parser->queue[HASH_RXQ_ETH].ibv_attr; + if (parser->count) + flow->cs = parser->cs; if (!priv->dev->data->dev_started) return 0; - flow->drxq.ibv_attr = parser->drop_q.ibv_attr; - parser->drop_q.ibv_attr = NULL; - flow->drxq.ibv_flow = ibv_create_flow(priv->flow_drop_queue->qp, - flow->drxq.ibv_attr); - if (!flow->drxq.ibv_flow) { + parser->queue[HASH_RXQ_ETH].ibv_attr = NULL; + flow->frxq[HASH_RXQ_ETH].ibv_flow = + ibv_create_flow(priv->flow_drop_queue->qp, + flow->frxq[HASH_RXQ_ETH].ibv_attr); + if (!flow->frxq[HASH_RXQ_ETH].ibv_flow) { rte_flow_error_set(error, ENOMEM, RTE_FLOW_ERROR_TYPE_HANDLE, NULL, "flow rule creation failure"); err = ENOMEM; @@ -1589,13 +1713,18 @@ priv_flow_create_action_queue_drop(struct priv *priv, return 0; error: assert(flow); - if (flow->drxq.ibv_flow) { - claim_zero(ibv_destroy_flow(flow->drxq.ibv_flow)); - flow->drxq.ibv_flow = NULL; + if (flow->frxq[HASH_RXQ_ETH].ibv_flow) { + claim_zero(ibv_destroy_flow(flow->frxq[HASH_RXQ_ETH].ibv_flow)); + flow->frxq[HASH_RXQ_ETH].ibv_flow = NULL; + } + if (flow->frxq[HASH_RXQ_ETH].ibv_attr) { + rte_free(flow->frxq[HASH_RXQ_ETH].ibv_attr); + flow->frxq[HASH_RXQ_ETH].ibv_attr = NULL; } - if (flow->drxq.ibv_attr) { - rte_free(flow->drxq.ibv_attr); - flow->drxq.ibv_attr = NULL; + if (flow->cs) { + claim_zero(ibv_destroy_counter_set(flow->cs)); + flow->cs = NULL; + parser->cs = NULL; } return err; } @@ -1631,13 +1760,15 @@ priv_flow_create_action_queue_rss(struct priv *priv, flow->frxq[i].ibv_attr = parser->queue[i].ibv_attr; parser->queue[i].ibv_attr = NULL; hash_fields = hash_rxq_init[i].hash_fields; + if (!priv->dev->data->dev_started) + continue; flow->frxq[i].hrxq = mlx5_priv_hrxq_get(priv, parser->rss_conf.rss_key, parser->rss_conf.rss_key_len, hash_fields, parser->queues, - hash_fields ? parser->queues_n : 1); + parser->queues_n); if (flow->frxq[i].hrxq) continue; flow->frxq[i].hrxq = @@ -1646,7 +1777,7 @@ priv_flow_create_action_queue_rss(struct priv *priv, parser->rss_conf.rss_key_len, hash_fields, parser->queues, - hash_fields ? parser->queues_n : 1); + parser->queues_n); if (!flow->frxq[i].hrxq) { rte_flow_error_set(error, ENOMEM, RTE_FLOW_ERROR_TYPE_HANDLE, @@ -1687,6 +1818,8 @@ priv_flow_create_action_queue(struct priv *priv, err = priv_flow_create_action_queue_rss(priv, parser, flow, error); if (err) goto error; + if (parser->count) + flow->cs = parser->cs; if (!priv->dev->data->dev_started) return 0; for (i = 0; i != hash_rxq_init_n; ++i) { @@ -1727,6 +1860,11 @@ error: if (flow->frxq[i].ibv_attr) rte_free(flow->frxq[i].ibv_attr); } + if (flow->cs) { + claim_zero(ibv_destroy_counter_set(flow->cs)); + flow->cs = NULL; + parser->cs = NULL; + } return err; } @@ -1779,6 +1917,7 @@ priv_flow_create(struct priv *priv, flow->queues = (uint16_t (*)[])(flow + 1); memcpy(flow->queues, parser.queues, parser.queues_n * sizeof(uint16_t)); flow->queues_n = parser.queues_n; + flow->mark = parser.mark; /* Copy RSS configuration. */ flow->rss_conf = parser.rss_conf; flow->rss_conf.rss_key = flow->rss_key; @@ -1795,13 +1934,9 @@ priv_flow_create(struct priv *priv, DEBUG("Flow created %p", (void *)flow); return flow; exit: - if (parser.drop) { - rte_free(parser.drop_q.ibv_attr); - } else { - for (i = 0; i != hash_rxq_init_n; ++i) { - if (parser.queue[i].ibv_attr) - rte_free(parser.queue[i].ibv_attr); - } + for (i = 0; i != hash_rxq_init_n; ++i) { + if (parser.queue[i].ibv_attr) + rte_free(parser.queue[i].ibv_attr); } rte_free(flow); return NULL; @@ -1903,9 +2038,10 @@ priv_flow_destroy(struct priv *priv, } free: if (flow->drop) { - if (flow->drxq.ibv_flow) - claim_zero(ibv_destroy_flow(flow->drxq.ibv_flow)); - rte_free(flow->drxq.ibv_attr); + if (flow->frxq[HASH_RXQ_ETH].ibv_flow) + claim_zero(ibv_destroy_flow + (flow->frxq[HASH_RXQ_ETH].ibv_flow)); + rte_free(flow->frxq[HASH_RXQ_ETH].ibv_attr); } else { for (i = 0; i != hash_rxq_init_n; ++i) { struct mlx5_flow *frxq = &flow->frxq[i]; @@ -1918,6 +2054,10 @@ free: rte_free(frxq->ibv_attr); } } + if (flow->cs) { + claim_zero(ibv_destroy_counter_set(flow->cs)); + flow->cs = NULL; + } TAILQ_REMOVE(list, flow, next); DEBUG("Flow destroyed %p", (void *)flow); rte_free(flow); @@ -2070,10 +2210,11 @@ priv_flow_stop(struct priv *priv, struct mlx5_flows *list) unsigned int i; if (flow->drop) { - if (!flow->drxq.ibv_flow) + if (!flow->frxq[HASH_RXQ_ETH].ibv_flow) continue; - claim_zero(ibv_destroy_flow(flow->drxq.ibv_flow)); - flow->drxq.ibv_flow = NULL; + claim_zero(ibv_destroy_flow + (flow->frxq[HASH_RXQ_ETH].ibv_flow)); + flow->frxq[HASH_RXQ_ETH].ibv_flow = NULL; /* Next flow. */ continue; } @@ -2121,10 +2262,11 @@ priv_flow_start(struct priv *priv, struct mlx5_flows *list) unsigned int i; if (flow->drop) { - flow->drxq.ibv_flow = - ibv_create_flow(priv->flow_drop_queue->qp, - flow->drxq.ibv_attr); - if (!flow->drxq.ibv_flow) { + flow->frxq[HASH_RXQ_ETH].ibv_flow = + ibv_create_flow + (priv->flow_drop_queue->qp, + flow->frxq[HASH_RXQ_ETH].ibv_attr); + if (!flow->frxq[HASH_RXQ_ETH].ibv_flow) { DEBUG("Flow %p cannot be applied", (void *)flow); rte_errno = EINVAL; @@ -2340,6 +2482,86 @@ mlx5_flow_flush(struct rte_eth_dev *dev, return 0; } +#ifdef HAVE_IBV_DEVICE_COUNTERS_SET_SUPPORT +/** + * Query flow counter. + * + * @param cs + * the counter set. + * @param counter_value + * returned data from the counter. + * + * @return + * 0 on success, a errno value otherwise and rte_errno is set. + */ +static int +priv_flow_query_count(struct ibv_counter_set *cs, + struct mlx5_flow_counter_stats *counter_stats, + struct rte_flow_query_count *query_count, + struct rte_flow_error *error) +{ + uint64_t counters[2]; + struct ibv_query_counter_set_attr query_cs_attr = { + .cs = cs, + .query_flags = IBV_COUNTER_SET_FORCE_UPDATE, + }; + struct ibv_counter_set_data query_out = { + .out = counters, + .outlen = 2 * sizeof(uint64_t), + }; + int res = ibv_query_counter_set(&query_cs_attr, &query_out); + + if (res) { + rte_flow_error_set(error, -res, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, + NULL, + "cannot read counter"); + return -res; + } + query_count->hits_set = 1; + query_count->bytes_set = 1; + query_count->hits = counters[0] - counter_stats->hits; + query_count->bytes = counters[1] - counter_stats->bytes; + if (query_count->reset) { + counter_stats->hits = counters[0]; + counter_stats->bytes = counters[1]; + } + return 0; +} + +/** + * Query a flows. + * + * @see rte_flow_query() + * @see rte_flow_ops + */ +int +mlx5_flow_query(struct rte_eth_dev *dev, + struct rte_flow *flow, + enum rte_flow_action_type action __rte_unused, + void *data, + struct rte_flow_error *error) +{ + struct priv *priv = dev->data->dev_private; + int res = EINVAL; + + priv_lock(priv); + if (flow->cs) { + res = priv_flow_query_count(flow->cs, + &flow->counter_stats, + (struct rte_flow_query_count *)data, + error); + } else { + rte_flow_error_set(error, res, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, + NULL, + "no counter found for flow"); + } + priv_unlock(priv); + return -res; +} +#endif + /** * Isolated mode. * @@ -2396,20 +2618,28 @@ priv_fdir_filter_convert(struct priv *priv, ERROR("invalid queue number %d", fdir_filter->action.rx_queue); return EINVAL; } - /* Validate the behavior. */ - if (fdir_filter->action.behavior != RTE_ETH_FDIR_ACCEPT) { - ERROR("invalid behavior %d", fdir_filter->action.behavior); - return ENOTSUP; - } attributes->attr.ingress = 1; attributes->items[0] = (struct rte_flow_item) { .type = RTE_FLOW_ITEM_TYPE_ETH, .spec = &attributes->l2, + .mask = &attributes->l2_mask, }; - attributes->actions[0] = (struct rte_flow_action){ - .type = RTE_FLOW_ACTION_TYPE_QUEUE, - .conf = &attributes->queue, - }; + switch (fdir_filter->action.behavior) { + case RTE_ETH_FDIR_ACCEPT: + attributes->actions[0] = (struct rte_flow_action){ + .type = RTE_FLOW_ACTION_TYPE_QUEUE, + .conf = &attributes->queue, + }; + break; + case RTE_ETH_FDIR_REJECT: + attributes->actions[0] = (struct rte_flow_action){ + .type = RTE_FLOW_ACTION_TYPE_DROP, + }; + break; + default: + ERROR("invalid behavior %d", fdir_filter->action.behavior); + return ENOTSUP; + } attributes->queue.index = fdir_filter->action.rx_queue; switch (fdir_filter->input.flow_type) { case RTE_ETH_FLOW_NONFRAG_IPV4_UDP: @@ -2427,10 +2657,12 @@ priv_fdir_filter_convert(struct priv *priv, attributes->items[1] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_IPV4, .spec = &attributes->l3, + .mask = &attributes->l3, }; attributes->items[2] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_UDP, .spec = &attributes->l4, + .mask = &attributes->l4, }; break; case RTE_ETH_FLOW_NONFRAG_IPV4_TCP: @@ -2448,10 +2680,12 @@ priv_fdir_filter_convert(struct priv *priv, attributes->items[1] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_IPV4, .spec = &attributes->l3, + .mask = &attributes->l3, }; attributes->items[2] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_TCP, .spec = &attributes->l4, + .mask = &attributes->l4, }; break; case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER: @@ -2465,6 +2699,7 @@ priv_fdir_filter_convert(struct priv *priv, attributes->items[1] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_IPV4, .spec = &attributes->l3, + .mask = &attributes->l3, }; break; case RTE_ETH_FLOW_NONFRAG_IPV6_UDP: @@ -2485,10 +2720,12 @@ priv_fdir_filter_convert(struct priv *priv, attributes->items[1] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_IPV6, .spec = &attributes->l3, + .mask = &attributes->l3, }; attributes->items[2] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_UDP, .spec = &attributes->l4, + .mask = &attributes->l4, }; break; case RTE_ETH_FLOW_NONFRAG_IPV6_TCP: @@ -2509,10 +2746,12 @@ priv_fdir_filter_convert(struct priv *priv, attributes->items[1] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_IPV6, .spec = &attributes->l3, + .mask = &attributes->l3, }; attributes->items[2] = (struct rte_flow_item){ - .type = RTE_FLOW_ITEM_TYPE_UDP, + .type = RTE_FLOW_ITEM_TYPE_TCP, .spec = &attributes->l4, + .mask = &attributes->l4, }; break; case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER: @@ -2529,6 +2768,7 @@ priv_fdir_filter_convert(struct priv *priv, attributes->items[1] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_IPV6, .spec = &attributes->l3, + .mask = &attributes->l3, }; break; default: @@ -2556,6 +2796,11 @@ priv_fdir_filter_add(struct priv *priv, { struct mlx5_fdir attributes = { .attr.group = 0, + .l2_mask = { + .dst.addr_bytes = "\x00\x00\x00\x00\x00\x00", + .src.addr_bytes = "\x00\x00\x00\x00\x00\x00", + .type = 0, + }, }; struct mlx5_flow_parse parser = { .layer = HASH_RXQ_ETH, @@ -2578,7 +2823,6 @@ priv_fdir_filter_add(struct priv *priv, attributes.actions, &error); if (flow) { - TAILQ_INSERT_TAIL(&priv->flows, flow, next); DEBUG("FDIR created %p", (void *)flow); return 0; } @@ -2600,7 +2844,9 @@ static int priv_fdir_filter_delete(struct priv *priv, const struct rte_eth_fdir_filter *fdir_filter) { - struct mlx5_fdir attributes; + struct mlx5_fdir attributes = { + .attr.group = 0, + }; struct mlx5_flow_parse parser = { .create = 1, .layer = HASH_RXQ_ETH, @@ -2617,6 +2863,22 @@ priv_fdir_filter_delete(struct priv *priv, attributes.actions, &error, &parser); if (ret) goto exit; + /* + * Special case for drop action which is only set in the + * specifications when the flow is created. In this situation the + * drop specification is missing. + */ + if (parser.drop) { + struct ibv_flow_spec_action_drop *drop; + + drop = (void *)((uintptr_t)parser.queue[HASH_RXQ_ETH].ibv_attr + + parser.queue[HASH_RXQ_ETH].offset); + *drop = (struct ibv_flow_spec_action_drop){ + .type = IBV_FLOW_SPEC_ACTION_DROP, + .size = sizeof(struct ibv_flow_spec_action_drop), + }; + parser.queue[HASH_RXQ_ETH].ibv_attr->num_of_specs++; + } TAILQ_FOREACH(flow, &priv->flows, next) { struct ibv_flow_attr *attr; struct ibv_spec_header *attr_h; @@ -2626,14 +2888,8 @@ priv_fdir_filter_delete(struct priv *priv, void *flow_spec; unsigned int specs_n; - if (parser.drop) - attr = parser.drop_q.ibv_attr; - else - attr = parser.queue[HASH_RXQ_ETH].ibv_attr; - if (flow->drop) - flow_attr = flow->drxq.ibv_attr; - else - flow_attr = flow->frxq[HASH_RXQ_ETH].ibv_attr; + attr = parser.queue[HASH_RXQ_ETH].ibv_attr; + flow_attr = flow->frxq[HASH_RXQ_ETH].ibv_attr; /* Compare first the attributes. */ if (memcmp(attr, flow_attr, sizeof(struct ibv_flow_attr))) continue; @@ -2649,24 +2905,23 @@ priv_fdir_filter_delete(struct priv *priv, flow_h = flow_spec; if (memcmp(spec, flow_spec, RTE_MIN(attr_h->size, flow_h->size))) - continue; - spec = (void *)((uintptr_t)attr + attr_h->size); - flow_spec = (void *)((uintptr_t)flow_attr + + goto wrong_flow; + spec = (void *)((uintptr_t)spec + attr_h->size); + flow_spec = (void *)((uintptr_t)flow_spec + flow_h->size); } /* At this point, the flow match. */ break; +wrong_flow: + /* The flow does not match. */ + continue; } if (flow) priv_flow_destroy(priv, &priv->flows, flow); exit: - if (parser.drop) { - rte_free(parser.drop_q.ibv_attr); - } else { - for (i = 0; i != hash_rxq_init_n; ++i) { - if (parser.queue[i].ibv_attr) - rte_free(parser.queue[i].ibv_attr); - } + for (i = 0; i != hash_rxq_init_n; ++i) { + if (parser.queue[i].ibv_attr) + rte_free(parser.queue[i].ibv_attr); } return -ret; }