X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx5%2Fmlx5_flow.c;h=3d2d5fc53a3ed90477befea31251b00473eb35cc;hb=8d0715f054b44750c2e1553d6af29487bcea25d8;hp=0096ed8a2fa55dd819e6cc8012d93f99d6c7c22a;hpb=535f686e546f7fc157ecea9555b5900768d4ae5c;p=dpdk.git diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c index 0096ed8a2f..3d2d5fc53a 100644 --- a/drivers/net/mlx5/mlx5_flow.c +++ b/drivers/net/mlx5/mlx5_flow.c @@ -3,7 +3,9 @@ * Copyright 2016 Mellanox Technologies, Ltd */ +#include #include +#include #include #include @@ -19,7 +21,6 @@ #include #include -#include #include #include #include @@ -28,57 +29,200 @@ #include "mlx5.h" #include "mlx5_defs.h" -#include "mlx5_prm.h" +#include "mlx5_flow.h" #include "mlx5_glue.h" +#include "mlx5_prm.h" +#include "mlx5_rxtx.h" /* 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; -/* Pattern Layer bits. */ -#define MLX5_FLOW_LAYER_OUTER_L2 (1u << 0) -#define MLX5_FLOW_LAYER_OUTER_L3_IPV4 (1u << 1) -#define MLX5_FLOW_LAYER_OUTER_L3_IPV6 (1u << 2) -#define MLX5_FLOW_LAYER_OUTER_L4_UDP (1u << 3) -#define MLX5_FLOW_LAYER_OUTER_L4_TCP (1u << 4) -#define MLX5_FLOW_LAYER_OUTER_VLAN (1u << 5) -/* Masks. */ -#define MLX5_FLOW_LAYER_OUTER_L3 \ - (MLX5_FLOW_LAYER_OUTER_L3_IPV4 | MLX5_FLOW_LAYER_OUTER_L3_IPV6) -#define MLX5_FLOW_LAYER_OUTER_L4 \ - (MLX5_FLOW_LAYER_OUTER_L4_UDP | MLX5_FLOW_LAYER_OUTER_L4_TCP) - -/* Actions that modify the fate of matching traffic. */ -#define MLX5_FLOW_FATE_DROP (1u << 0) -#define MLX5_FLOW_FATE_QUEUE (1u << 1) - -/* possible L3 layers protocols filtering. */ -#define MLX5_IP_PROTOCOL_UDP 17 - -/** Handles information leading to a drop fate. */ -struct mlx5_flow_verbs { - unsigned int size; /**< Size of the attribute. */ - struct { - struct ibv_flow_attr *attr; - /**< Pointer to the Specification buffer. */ - uint8_t *specs; /**< Pointer to the specifications. */ - }; - struct ibv_flow *flow; /**< Verbs flow pointer. */ - struct mlx5_hrxq *hrxq; /**< Hash Rx queue object. */ +/** Device flow drivers. */ +#ifdef HAVE_IBV_FLOW_DV_SUPPORT +extern const struct mlx5_flow_driver_ops mlx5_flow_dv_drv_ops; +#endif +extern const struct mlx5_flow_driver_ops mlx5_flow_verbs_drv_ops; + +const struct mlx5_flow_driver_ops mlx5_flow_null_drv_ops; + +const struct mlx5_flow_driver_ops *flow_drv_ops[] = { + [MLX5_FLOW_TYPE_MIN] = &mlx5_flow_null_drv_ops, +#ifdef HAVE_IBV_FLOW_DV_SUPPORT + [MLX5_FLOW_TYPE_DV] = &mlx5_flow_dv_drv_ops, +#endif + [MLX5_FLOW_TYPE_VERBS] = &mlx5_flow_verbs_drv_ops, + [MLX5_FLOW_TYPE_MAX] = &mlx5_flow_null_drv_ops }; -/* Flow structure. */ -struct rte_flow { - TAILQ_ENTRY(rte_flow) next; /**< Pointer to the next flow structure. */ - struct rte_flow_attr attributes; /**< User flow attribute. */ - uint32_t l3_protocol_en:1; /**< Protocol filtering requested. */ - uint32_t layers; - /**< Bit-fields of present layers see MLX5_FLOW_LAYER_*. */ - uint32_t fate; - /**< Bit-fields of present fate see MLX5_FLOW_FATE_*. */ - uint8_t l3_protocol; /**< valid when l3_protocol_en is set. */ - struct mlx5_flow_verbs verbs; /* Verbs flow. */ - uint16_t queue; /**< Destination queue to redirect traffic to. */ +enum mlx5_expansion { + MLX5_EXPANSION_ROOT, + MLX5_EXPANSION_ROOT_OUTER, + MLX5_EXPANSION_ROOT_ETH_VLAN, + MLX5_EXPANSION_ROOT_OUTER_ETH_VLAN, + MLX5_EXPANSION_OUTER_ETH, + MLX5_EXPANSION_OUTER_ETH_VLAN, + MLX5_EXPANSION_OUTER_VLAN, + MLX5_EXPANSION_OUTER_IPV4, + MLX5_EXPANSION_OUTER_IPV4_UDP, + MLX5_EXPANSION_OUTER_IPV4_TCP, + MLX5_EXPANSION_OUTER_IPV6, + MLX5_EXPANSION_OUTER_IPV6_UDP, + MLX5_EXPANSION_OUTER_IPV6_TCP, + MLX5_EXPANSION_VXLAN, + MLX5_EXPANSION_VXLAN_GPE, + MLX5_EXPANSION_GRE, + MLX5_EXPANSION_MPLS, + MLX5_EXPANSION_ETH, + MLX5_EXPANSION_ETH_VLAN, + MLX5_EXPANSION_VLAN, + MLX5_EXPANSION_IPV4, + MLX5_EXPANSION_IPV4_UDP, + MLX5_EXPANSION_IPV4_TCP, + MLX5_EXPANSION_IPV6, + MLX5_EXPANSION_IPV6_UDP, + MLX5_EXPANSION_IPV6_TCP, +}; + +/** Supported expansion of items. */ +static const struct rte_flow_expand_node mlx5_support_expansion[] = { + [MLX5_EXPANSION_ROOT] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_ETH, + MLX5_EXPANSION_IPV4, + MLX5_EXPANSION_IPV6), + .type = RTE_FLOW_ITEM_TYPE_END, + }, + [MLX5_EXPANSION_ROOT_OUTER] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_OUTER_ETH, + MLX5_EXPANSION_OUTER_IPV4, + MLX5_EXPANSION_OUTER_IPV6), + .type = RTE_FLOW_ITEM_TYPE_END, + }, + [MLX5_EXPANSION_ROOT_ETH_VLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_ETH_VLAN), + .type = RTE_FLOW_ITEM_TYPE_END, + }, + [MLX5_EXPANSION_ROOT_OUTER_ETH_VLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_OUTER_ETH_VLAN), + .type = RTE_FLOW_ITEM_TYPE_END, + }, + [MLX5_EXPANSION_OUTER_ETH] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_OUTER_IPV4, + MLX5_EXPANSION_OUTER_IPV6, + MLX5_EXPANSION_MPLS), + .type = RTE_FLOW_ITEM_TYPE_ETH, + .rss_types = 0, + }, + [MLX5_EXPANSION_OUTER_ETH_VLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_OUTER_VLAN), + .type = RTE_FLOW_ITEM_TYPE_ETH, + .rss_types = 0, + }, + [MLX5_EXPANSION_OUTER_VLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_OUTER_IPV4, + MLX5_EXPANSION_OUTER_IPV6), + .type = RTE_FLOW_ITEM_TYPE_VLAN, + }, + [MLX5_EXPANSION_OUTER_IPV4] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT + (MLX5_EXPANSION_OUTER_IPV4_UDP, + MLX5_EXPANSION_OUTER_IPV4_TCP, + MLX5_EXPANSION_GRE), + .type = RTE_FLOW_ITEM_TYPE_IPV4, + .rss_types = ETH_RSS_IPV4 | ETH_RSS_FRAG_IPV4 | + ETH_RSS_NONFRAG_IPV4_OTHER, + }, + [MLX5_EXPANSION_OUTER_IPV4_UDP] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_VXLAN, + MLX5_EXPANSION_VXLAN_GPE), + .type = RTE_FLOW_ITEM_TYPE_UDP, + .rss_types = ETH_RSS_NONFRAG_IPV4_UDP, + }, + [MLX5_EXPANSION_OUTER_IPV4_TCP] = { + .type = RTE_FLOW_ITEM_TYPE_TCP, + .rss_types = ETH_RSS_NONFRAG_IPV4_TCP, + }, + [MLX5_EXPANSION_OUTER_IPV6] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT + (MLX5_EXPANSION_OUTER_IPV6_UDP, + MLX5_EXPANSION_OUTER_IPV6_TCP), + .type = RTE_FLOW_ITEM_TYPE_IPV6, + .rss_types = ETH_RSS_IPV6 | ETH_RSS_FRAG_IPV6 | + ETH_RSS_NONFRAG_IPV6_OTHER, + }, + [MLX5_EXPANSION_OUTER_IPV6_UDP] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_VXLAN, + MLX5_EXPANSION_VXLAN_GPE), + .type = RTE_FLOW_ITEM_TYPE_UDP, + .rss_types = ETH_RSS_NONFRAG_IPV6_UDP, + }, + [MLX5_EXPANSION_OUTER_IPV6_TCP] = { + .type = RTE_FLOW_ITEM_TYPE_TCP, + .rss_types = ETH_RSS_NONFRAG_IPV6_TCP, + }, + [MLX5_EXPANSION_VXLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_ETH), + .type = RTE_FLOW_ITEM_TYPE_VXLAN, + }, + [MLX5_EXPANSION_VXLAN_GPE] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_ETH, + MLX5_EXPANSION_IPV4, + MLX5_EXPANSION_IPV6), + .type = RTE_FLOW_ITEM_TYPE_VXLAN_GPE, + }, + [MLX5_EXPANSION_GRE] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_IPV4), + .type = RTE_FLOW_ITEM_TYPE_GRE, + }, + [MLX5_EXPANSION_MPLS] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_IPV4, + MLX5_EXPANSION_IPV6), + .type = RTE_FLOW_ITEM_TYPE_MPLS, + }, + [MLX5_EXPANSION_ETH] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_IPV4, + MLX5_EXPANSION_IPV6), + .type = RTE_FLOW_ITEM_TYPE_ETH, + }, + [MLX5_EXPANSION_ETH_VLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_VLAN), + .type = RTE_FLOW_ITEM_TYPE_ETH, + }, + [MLX5_EXPANSION_VLAN] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_IPV4, + MLX5_EXPANSION_IPV6), + .type = RTE_FLOW_ITEM_TYPE_VLAN, + }, + [MLX5_EXPANSION_IPV4] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_IPV4_UDP, + MLX5_EXPANSION_IPV4_TCP), + .type = RTE_FLOW_ITEM_TYPE_IPV4, + .rss_types = ETH_RSS_IPV4 | ETH_RSS_FRAG_IPV4 | + ETH_RSS_NONFRAG_IPV4_OTHER, + }, + [MLX5_EXPANSION_IPV4_UDP] = { + .type = RTE_FLOW_ITEM_TYPE_UDP, + .rss_types = ETH_RSS_NONFRAG_IPV4_UDP, + }, + [MLX5_EXPANSION_IPV4_TCP] = { + .type = RTE_FLOW_ITEM_TYPE_TCP, + .rss_types = ETH_RSS_NONFRAG_IPV4_TCP, + }, + [MLX5_EXPANSION_IPV6] = { + .next = RTE_FLOW_EXPAND_RSS_NEXT(MLX5_EXPANSION_IPV6_UDP, + MLX5_EXPANSION_IPV6_TCP), + .type = RTE_FLOW_ITEM_TYPE_IPV6, + .rss_types = ETH_RSS_IPV6 | ETH_RSS_FRAG_IPV6 | + ETH_RSS_NONFRAG_IPV6_OTHER, + }, + [MLX5_EXPANSION_IPV6_UDP] = { + .type = RTE_FLOW_ITEM_TYPE_UDP, + .rss_types = ETH_RSS_NONFRAG_IPV6_UDP, + }, + [MLX5_EXPANSION_IPV6_TCP] = { + .type = RTE_FLOW_ITEM_TYPE_TCP, + .rss_types = ETH_RSS_NONFRAG_IPV6_TCP, + }, }; static const struct rte_flow_ops mlx5_flow_ops = { @@ -87,12 +231,12 @@ static const struct rte_flow_ops mlx5_flow_ops = { .destroy = mlx5_flow_destroy, .flush = mlx5_flow_flush, .isolate = mlx5_flow_isolate, + .query = mlx5_flow_query, }; /* Convert FDIR request to Generic flow. */ struct mlx5_fdir { struct rte_flow_attr attr; - struct rte_flow_action actions[2]; struct rte_flow_item items[4]; struct rte_flow_item_eth l2; struct rte_flow_item_eth l2_mask; @@ -112,28 +256,68 @@ struct mlx5_fdir { struct rte_flow_item_udp udp; struct rte_flow_item_tcp tcp; } l4_mask; + struct rte_flow_action actions[2]; struct rte_flow_action_queue queue; }; -/* Verbs specification header. */ -struct ibv_spec_header { - enum ibv_flow_spec_type type; - uint16_t size; +/* Map of Verbs to Flow priority with 8 Verbs priorities. */ +static const uint32_t priority_map_3[][MLX5_PRIORITY_MAP_MAX] = { + { 0, 1, 2 }, { 2, 3, 4 }, { 5, 6, 7 }, }; - /** - * Discover the maximum number of priority available. - * - * @param[in] dev - * Pointer to Ethernet device. - * - * @return - * number of supported flow priority on success, a negative errno value - * otherwise and rte_errno is set. - */ +/* Map of Verbs to Flow priority with 16 Verbs priorities. */ +static const uint32_t priority_map_5[][MLX5_PRIORITY_MAP_MAX] = { + { 0, 1, 2 }, { 3, 4, 5 }, { 6, 7, 8 }, + { 9, 10, 11 }, { 12, 13, 14 }, +}; + +/* Tunnel information. */ +struct mlx5_flow_tunnel_info { + uint64_t tunnel; /**< Tunnel bit (see MLX5_FLOW_*). */ + uint32_t ptype; /**< Tunnel Ptype (see RTE_PTYPE_*). */ +}; + +static struct mlx5_flow_tunnel_info tunnels_info[] = { + { + .tunnel = MLX5_FLOW_LAYER_VXLAN, + .ptype = RTE_PTYPE_TUNNEL_VXLAN | RTE_PTYPE_L4_UDP, + }, + { + .tunnel = MLX5_FLOW_LAYER_VXLAN_GPE, + .ptype = RTE_PTYPE_TUNNEL_VXLAN_GPE | RTE_PTYPE_L4_UDP, + }, + { + .tunnel = MLX5_FLOW_LAYER_GRE, + .ptype = RTE_PTYPE_TUNNEL_GRE, + }, + { + .tunnel = MLX5_FLOW_LAYER_MPLS | MLX5_FLOW_LAYER_OUTER_L4_UDP, + .ptype = RTE_PTYPE_TUNNEL_MPLS_IN_UDP | RTE_PTYPE_L4_UDP, + }, + { + .tunnel = MLX5_FLOW_LAYER_MPLS, + .ptype = RTE_PTYPE_TUNNEL_MPLS_IN_GRE, + }, + { + .tunnel = MLX5_FLOW_LAYER_NVGRE, + .ptype = RTE_PTYPE_TUNNEL_NVGRE, + }, +}; + +/** + * Discover the maximum number of priority available. + * + * @param[in] dev + * Pointer to the Ethernet device structure. + * + * @return + * number of supported flow priority on success, a negative errno + * value otherwise and rte_errno is set. + */ int mlx5_flow_discover_priorities(struct rte_eth_dev *dev) { + struct mlx5_priv *priv = dev->data->dev_private; struct { struct ibv_flow_attr attr; struct ibv_flow_spec_eth eth; @@ -141,6 +325,7 @@ mlx5_flow_discover_priorities(struct rte_eth_dev *dev) } flow_attr = { .attr = { .num_of_specs = 2, + .port = (uint8_t)priv->ibv_port, }, .eth = { .type = IBV_FLOW_SPEC_ETH, @@ -155,6 +340,7 @@ mlx5_flow_discover_priorities(struct rte_eth_dev *dev) struct mlx5_hrxq *drop = mlx5_hrxq_drop_new(dev); uint16_t vprio[] = { 8, 16 }; int i; + int priority = 0; if (!drop) { rte_errno = ENOTSUP; @@ -166,65 +352,56 @@ mlx5_flow_discover_priorities(struct rte_eth_dev *dev) if (!flow) break; claim_zero(mlx5_glue->destroy_flow(flow)); + priority = vprio[i]; } mlx5_hrxq_drop_release(dev); + switch (priority) { + case 8: + priority = RTE_DIM(priority_map_3); + break; + case 16: + priority = RTE_DIM(priority_map_5); + break; + default: + rte_errno = ENOTSUP; + DRV_LOG(ERR, + "port %u verbs maximum priority: %d expected 8/16", + dev->data->port_id, priority); + return -rte_errno; + } DRV_LOG(INFO, "port %u flow maximum priority: %d", - dev->data->port_id, vprio[i - 1]); - return vprio[i - 1]; + dev->data->port_id, priority); + return priority; } /** - * Verify the @p attributes will be correctly understood by the NIC and store - * them in the @p flow if everything is correct. + * Adjust flow priority based on the highest layer and the request priority. * * @param[in] dev - * Pointer to Ethernet device. - * @param[in] attributes - * Pointer to flow attributes - * @param[in, out] flow - * Pointer to the rte_flow structure. - * @param[out] error - * Pointer to error structure. + * Pointer to the Ethernet device structure. + * @param[in] priority + * The rule base priority. + * @param[in] subpriority + * The priority based on the items. * * @return - * 0 on success, a negative errno value otherwise and rte_errno is set. + * The new priority. */ -static int -mlx5_flow_attributes(struct rte_eth_dev *dev, - const struct rte_flow_attr *attributes, - struct rte_flow *flow, - struct rte_flow_error *error) +uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev *dev, int32_t priority, + uint32_t subpriority) { - uint32_t priority_max = - ((struct priv *)dev->data->dev_private)->config.flow_prio; + uint32_t res = 0; + struct mlx5_priv *priv = dev->data->dev_private; - if (attributes->group) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ATTR_GROUP, - NULL, - "groups is not supported"); - if (attributes->priority >= priority_max) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, - NULL, - "priority out of range"); - if (attributes->egress) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, - NULL, - "egress is not supported"); - if (attributes->transfer) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER, - NULL, - "transfer is not supported"); - if (!attributes->ingress) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ATTR_INGRESS, - NULL, - "ingress attribute is mandatory"); - flow->attributes = *attributes; - return 0; + switch (priv->config.flow_prio) { + case RTE_DIM(priority_map_3): + res = priority_map_3[priority][subpriority]; + break; + case RTE_DIM(priority_map_5): + res = priority_map_5[priority][subpriority]; + break; + } + return res; } /** @@ -245,7 +422,7 @@ mlx5_flow_attributes(struct rte_eth_dev *dev, * @return * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int +int mlx5_flow_item_acceptable(const struct rte_flow_item *item, const uint8_t *mask, const uint8_t *nic_mask, @@ -264,8 +441,7 @@ mlx5_flow_item_acceptable(const struct rte_flow_item *item, " bits"); if (!item->spec && (item->mask || item->last)) return rte_flow_error_set(error, EINVAL, - RTE_FLOW_ERROR_TYPE_ITEM, - item, + RTE_FLOW_ERROR_TYPE_ITEM, item, "mask/last without a spec is not" " supported"); if (item->spec && item->last) { @@ -280,354 +456,925 @@ mlx5_flow_item_acceptable(const struct rte_flow_item *item, } ret = memcmp(spec, last, size); if (ret != 0) - return rte_flow_error_set(error, ENOTSUP, + return rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM, item, - "range is not supported"); + "range is not valid"); } return 0; } /** - * Add a verbs specification into @p flow. + * Adjust the hash fields according to the @p flow information. * - * @param[in, out] flow - * Pointer to flow structure. - * @param[in] src - * Create specification. - * @param[in] size - * Size in bytes of the specification to copy. + * @param[in] dev_flow. + * Pointer to the mlx5_flow. + * @param[in] tunnel + * 1 when the hash field is for a tunnel item. + * @param[in] layer_types + * ETH_RSS_* types. + * @param[in] hash_fields + * Item hash fields. + * + * @return + * The hash fields that should be used. + */ +uint64_t +mlx5_flow_hashfields_adjust(struct mlx5_flow *dev_flow, + int tunnel __rte_unused, uint64_t layer_types, + uint64_t hash_fields) +{ + struct rte_flow *flow = dev_flow->flow; +#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT + int rss_request_inner = flow->rss.level >= 2; + + /* Check RSS hash level for tunnel. */ + if (tunnel && rss_request_inner) + hash_fields |= IBV_RX_HASH_INNER; + else if (tunnel || rss_request_inner) + return 0; +#endif + /* Check if requested layer matches RSS hash fields. */ + if (!(flow->rss.types & layer_types)) + return 0; + return hash_fields; +} + +/** + * Lookup and set the ptype in the data Rx part. A single Ptype can be used, + * if several tunnel rules are used on this queue, the tunnel ptype will be + * cleared. + * + * @param rxq_ctrl + * Rx queue to update. + */ +static void +flow_rxq_tunnel_ptype_update(struct mlx5_rxq_ctrl *rxq_ctrl) +{ + unsigned int i; + uint32_t tunnel_ptype = 0; + + /* Look up for the ptype to use. */ + for (i = 0; i != MLX5_FLOW_TUNNEL; ++i) { + if (!rxq_ctrl->flow_tunnels_n[i]) + continue; + if (!tunnel_ptype) { + tunnel_ptype = tunnels_info[i].ptype; + } else { + tunnel_ptype = 0; + break; + } + } + rxq_ctrl->rxq.tunnel = tunnel_ptype; +} + +/** + * Set the Rx queue flags (Mark/Flag and Tunnel Ptypes) according to the devive + * flow. + * + * @param[in] dev + * Pointer to the Ethernet device structure. + * @param[in] dev_flow + * Pointer to device flow structure. */ static void -mlx5_flow_spec_verbs_add(struct rte_flow *flow, void *src, unsigned int size) +flow_drv_rxq_flags_set(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow) { - if (flow->verbs.specs) { - void *dst; + struct mlx5_priv *priv = dev->data->dev_private; + struct rte_flow *flow = dev_flow->flow; + const int mark = !!(flow->actions & + (MLX5_FLOW_ACTION_FLAG | MLX5_FLOW_ACTION_MARK)); + const int tunnel = !!(dev_flow->layers & MLX5_FLOW_LAYER_TUNNEL); + unsigned int i; + + for (i = 0; i != flow->rss.queue_num; ++i) { + int idx = (*flow->queue)[i]; + struct mlx5_rxq_ctrl *rxq_ctrl = + container_of((*priv->rxqs)[idx], + struct mlx5_rxq_ctrl, rxq); - dst = (void *)(flow->verbs.specs + flow->verbs.size); - memcpy(dst, src, size); - ++flow->verbs.attr->num_of_specs; + if (mark) { + rxq_ctrl->rxq.mark = 1; + rxq_ctrl->flow_mark_n++; + } + if (tunnel) { + unsigned int j; + + /* Increase the counter matching the flow. */ + for (j = 0; j != MLX5_FLOW_TUNNEL; ++j) { + if ((tunnels_info[j].tunnel & + dev_flow->layers) == + tunnels_info[j].tunnel) { + rxq_ctrl->flow_tunnels_n[j]++; + break; + } + } + flow_rxq_tunnel_ptype_update(rxq_ctrl); + } } - flow->verbs.size += size; } /** - * Convert the @p item into a Verbs specification after ensuring the NIC - * will understand and process it correctly. - * If the necessary size for the conversion is greater than the @p flow_size, - * nothing is written in @p flow, the validation is still performed. + * Set the Rx queue flags (Mark/Flag and Tunnel Ptypes) for a flow * - * @param[in] item - * Item specification. - * @param[in, out] flow + * @param[in] dev + * Pointer to the Ethernet device structure. + * @param[in] flow * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small, nothing is - * written. - * @param[out] error - * Pointer to error structure. + */ +static void +flow_rxq_flags_set(struct rte_eth_dev *dev, struct rte_flow *flow) +{ + struct mlx5_flow *dev_flow; + + LIST_FOREACH(dev_flow, &flow->dev_flows, next) + flow_drv_rxq_flags_set(dev, dev_flow); +} + +/** + * Clear the Rx queue flags (Mark/Flag and Tunnel Ptype) associated with the + * device flow if no other flow uses it with the same kind of request. * - * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p item has fully been converted, - * otherwise another call with this returned memory size should be done. - * On error, a negative errno value is returned and rte_errno is set. + * @param dev + * Pointer to Ethernet device. + * @param[in] dev_flow + * Pointer to the device flow. */ -static int -mlx5_flow_item_eth(const struct rte_flow_item *item, struct rte_flow *flow, - const size_t flow_size, struct rte_flow_error *error) +static void +flow_drv_rxq_flags_trim(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow) { - const struct rte_flow_item_eth *spec = item->spec; - const struct rte_flow_item_eth *mask = item->mask; - const struct rte_flow_item_eth nic_mask = { - .dst.addr_bytes = "\xff\xff\xff\xff\xff\xff", - .src.addr_bytes = "\xff\xff\xff\xff\xff\xff", - .type = RTE_BE16(0xffff), - }; - const unsigned int size = sizeof(struct ibv_flow_spec_eth); - struct ibv_flow_spec_eth eth = { - .type = IBV_FLOW_SPEC_ETH, - .size = size, - }; - int ret; + struct mlx5_priv *priv = dev->data->dev_private; + struct rte_flow *flow = dev_flow->flow; + const int mark = !!(flow->actions & + (MLX5_FLOW_ACTION_FLAG | MLX5_FLOW_ACTION_MARK)); + const int tunnel = !!(dev_flow->layers & MLX5_FLOW_LAYER_TUNNEL); + unsigned int i; - if (flow->layers & MLX5_FLOW_LAYER_OUTER_L2) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "L2 layers already configured"); - if (!mask) - mask = &rte_flow_item_eth_mask; - ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask, - (const uint8_t *)&nic_mask, - sizeof(struct rte_flow_item_eth), - error); - if (ret) - return ret; - flow->layers |= MLX5_FLOW_LAYER_OUTER_L2; - if (size > flow_size) - return size; - if (spec) { - unsigned int i; + assert(dev->data->dev_started); + for (i = 0; i != flow->rss.queue_num; ++i) { + int idx = (*flow->queue)[i]; + struct mlx5_rxq_ctrl *rxq_ctrl = + container_of((*priv->rxqs)[idx], + struct mlx5_rxq_ctrl, rxq); - memcpy(ð.val.dst_mac, spec->dst.addr_bytes, ETHER_ADDR_LEN); - memcpy(ð.val.src_mac, spec->src.addr_bytes, ETHER_ADDR_LEN); - eth.val.ether_type = spec->type; - memcpy(ð.mask.dst_mac, mask->dst.addr_bytes, ETHER_ADDR_LEN); - memcpy(ð.mask.src_mac, mask->src.addr_bytes, ETHER_ADDR_LEN); - eth.mask.ether_type = mask->type; - /* Remove unwanted bits from values. */ - for (i = 0; i < ETHER_ADDR_LEN; ++i) { - eth.val.dst_mac[i] &= eth.mask.dst_mac[i]; - eth.val.src_mac[i] &= eth.mask.src_mac[i]; + if (mark) { + rxq_ctrl->flow_mark_n--; + rxq_ctrl->rxq.mark = !!rxq_ctrl->flow_mark_n; + } + if (tunnel) { + unsigned int j; + + /* Decrease the counter matching the flow. */ + for (j = 0; j != MLX5_FLOW_TUNNEL; ++j) { + if ((tunnels_info[j].tunnel & + dev_flow->layers) == + tunnels_info[j].tunnel) { + rxq_ctrl->flow_tunnels_n[j]--; + break; + } + } + flow_rxq_tunnel_ptype_update(rxq_ctrl); } - eth.val.ether_type &= eth.mask.ether_type; } - mlx5_flow_spec_verbs_add(flow, ð, size); - return size; } /** - * Update the VLAN tag in the Verbs Ethernet specification. + * Clear the Rx queue flags (Mark/Flag and Tunnel Ptype) associated with the + * @p flow if no other flow uses it with the same kind of request. + * + * @param dev + * Pointer to Ethernet device. + * @param[in] flow + * Pointer to the flow. + */ +static void +flow_rxq_flags_trim(struct rte_eth_dev *dev, struct rte_flow *flow) +{ + struct mlx5_flow *dev_flow; + + LIST_FOREACH(dev_flow, &flow->dev_flows, next) + flow_drv_rxq_flags_trim(dev, dev_flow); +} + +/** + * Clear the Mark/Flag and Tunnel ptype information in all Rx queues. * - * @param[in, out] attr - * Pointer to Verbs attributes structure. - * @param[in] eth - * Verbs structure containing the VLAN information to copy. + * @param dev + * Pointer to Ethernet device. */ static void -mlx5_flow_item_vlan_update(struct ibv_flow_attr *attr, - struct ibv_flow_spec_eth *eth) +flow_rxq_flags_clear(struct rte_eth_dev *dev) { + struct mlx5_priv *priv = dev->data->dev_private; unsigned int i; - enum ibv_flow_spec_type search = IBV_FLOW_SPEC_ETH; - struct ibv_spec_header *hdr = (struct ibv_spec_header *) - ((uint8_t *)attr + sizeof(struct ibv_flow_attr)); - - for (i = 0; i != attr->num_of_specs; ++i) { - if (hdr->type == search) { - struct ibv_flow_spec_eth *e = - (struct ibv_flow_spec_eth *)hdr; - - e->val.vlan_tag = eth->val.vlan_tag; - e->mask.vlan_tag = eth->mask.vlan_tag; - e->val.ether_type = eth->val.ether_type; - e->mask.ether_type = eth->mask.ether_type; - break; - } - hdr = (struct ibv_spec_header *)((uint8_t *)hdr + hdr->size); + + for (i = 0; i != priv->rxqs_n; ++i) { + struct mlx5_rxq_ctrl *rxq_ctrl; + unsigned int j; + + if (!(*priv->rxqs)[i]) + continue; + rxq_ctrl = container_of((*priv->rxqs)[i], + struct mlx5_rxq_ctrl, rxq); + rxq_ctrl->flow_mark_n = 0; + rxq_ctrl->rxq.mark = 0; + for (j = 0; j != MLX5_FLOW_TUNNEL; ++j) + rxq_ctrl->flow_tunnels_n[j] = 0; + rxq_ctrl->rxq.tunnel = 0; } } -/** - * Convert the @p item into @p flow (or by updating the already present - * Ethernet Verbs) specification after ensuring the NIC will understand and - * process it correctly. - * If the necessary size for the conversion is greater than the @p flow_size, - * nothing is written in @p flow, the validation is still performed. +/* + * Validate the flag action. * - * @param[in] item - * Item specification. - * @param[in, out] flow - * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small, nothing is - * written. + * @param[in] action_flags + * Bit-fields that holds the actions detected until now. + * @param[in] attr + * Attributes of flow that includes this action. * @param[out] error * Pointer to error structure. * * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p item has fully been converted, - * otherwise another call with this returned memory size should be done. - * On error, a negative errno value is returned and rte_errno is set. + * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int -mlx5_flow_item_vlan(const struct rte_flow_item *item, struct rte_flow *flow, - const size_t flow_size, struct rte_flow_error *error) +int +mlx5_flow_validate_action_flag(uint64_t action_flags, + const struct rte_flow_attr *attr, + struct rte_flow_error *error) { - const struct rte_flow_item_vlan *spec = item->spec; - const struct rte_flow_item_vlan *mask = item->mask; - const struct rte_flow_item_vlan nic_mask = { - .tci = RTE_BE16(0x0fff), - .inner_type = RTE_BE16(0xffff), - }; - unsigned int size = sizeof(struct ibv_flow_spec_eth); - struct ibv_flow_spec_eth eth = { - .type = IBV_FLOW_SPEC_ETH, - .size = size, - }; - int ret; - const uint32_t l34m = MLX5_FLOW_LAYER_OUTER_L3 | - MLX5_FLOW_LAYER_OUTER_L4; - const uint32_t vlanm = MLX5_FLOW_LAYER_OUTER_VLAN; - const uint32_t l2m = MLX5_FLOW_LAYER_OUTER_L2; - if (flow->layers & vlanm) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "VLAN layer already configured"); - else if ((flow->layers & l34m) != 0) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "L2 layer cannot follow L3/L4 layer"); - if (!mask) - mask = &rte_flow_item_vlan_mask; - ret = mlx5_flow_item_acceptable - (item, (const uint8_t *)mask, - (const uint8_t *)&nic_mask, - sizeof(struct rte_flow_item_vlan), error); - if (ret) - return ret; - if (spec) { - eth.val.vlan_tag = spec->tci; - eth.mask.vlan_tag = mask->tci; - eth.val.vlan_tag &= eth.mask.vlan_tag; - eth.val.ether_type = spec->inner_type; - eth.mask.ether_type = mask->inner_type; - eth.val.ether_type &= eth.mask.ether_type; - } - /* - * From verbs perspective an empty VLAN is equivalent - * to a packet without VLAN layer. - */ - if (!eth.mask.vlan_tag) + if (action_flags & MLX5_FLOW_ACTION_DROP) return rte_flow_error_set(error, EINVAL, - RTE_FLOW_ERROR_TYPE_ITEM_SPEC, - item->spec, - "VLAN cannot be empty"); - if (!(flow->layers & l2m)) { - if (size <= flow_size) - mlx5_flow_spec_verbs_add(flow, ð, size); - } else { - if (flow->verbs.attr) - mlx5_flow_item_vlan_update(flow->verbs.attr, ð); - size = 0; /* Only an update is done in eth specification. */ - } - flow->layers |= MLX5_FLOW_LAYER_OUTER_L2 | - MLX5_FLOW_LAYER_OUTER_VLAN; - return size; + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't drop and flag in same flow"); + if (action_flags & MLX5_FLOW_ACTION_MARK) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't mark and flag in same flow"); + if (action_flags & MLX5_FLOW_ACTION_FLAG) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't have 2 flag" + " actions in same flow"); + if (attr->egress) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "flag action not supported for " + "egress"); + return 0; } -/** - * Convert the @p item into a Verbs specification after ensuring the NIC - * will understand and process it correctly. - * If the necessary size for the conversion is greater than the @p flow_size, - * nothing is written in @p flow, the validation is still performed. +/* + * Validate the mark action. * - * @param[in] item - * Item specification. - * @param[in, out] flow - * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small, nothing is - * written. + * @param[in] action + * Pointer to the queue action. + * @param[in] action_flags + * Bit-fields that holds the actions detected until now. + * @param[in] attr + * Attributes of flow that includes this action. * @param[out] error * Pointer to error structure. * * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p item has fully been converted, - * otherwise another call with this returned memory size should be done. - * On error, a negative errno value is returned and rte_errno is set. + * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int -mlx5_flow_item_ipv4(const struct rte_flow_item *item, struct rte_flow *flow, - const size_t flow_size, struct rte_flow_error *error) +int +mlx5_flow_validate_action_mark(const struct rte_flow_action *action, + uint64_t action_flags, + const struct rte_flow_attr *attr, + struct rte_flow_error *error) { - const struct rte_flow_item_ipv4 *spec = item->spec; - const struct rte_flow_item_ipv4 *mask = item->mask; - const struct rte_flow_item_ipv4 nic_mask = { - .hdr = { - .src_addr = RTE_BE32(0xffffffff), - .dst_addr = RTE_BE32(0xffffffff), - .type_of_service = 0xff, - .next_proto_id = 0xff, - }, - }; - unsigned int size = sizeof(struct ibv_flow_spec_ipv4_ext); - struct ibv_flow_spec_ipv4_ext ipv4 = { - .type = IBV_FLOW_SPEC_IPV4_EXT, - .size = size, - }; - int ret; + const struct rte_flow_action_mark *mark = action->conf; - if (flow->layers & MLX5_FLOW_LAYER_OUTER_L3) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "multiple L3 layers not supported"); - else if (flow->layers & MLX5_FLOW_LAYER_OUTER_L4) + if (!mark) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, + action, + "configuration cannot be null"); + if (mark->id >= MLX5_FLOW_MARK_MAX) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &mark->id, + "mark id must in 0 <= id < " + RTE_STR(MLX5_FLOW_MARK_MAX)); + if (action_flags & MLX5_FLOW_ACTION_DROP) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't drop and mark in same flow"); + if (action_flags & MLX5_FLOW_ACTION_FLAG) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't flag and mark in same flow"); + if (action_flags & MLX5_FLOW_ACTION_MARK) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't have 2 mark actions in same" + " flow"); + if (attr->egress) return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "L3 cannot follow an L4 layer."); - if (!mask) - mask = &rte_flow_item_ipv4_mask; - ret = mlx5_flow_item_acceptable - (item, (const uint8_t *)mask, - (const uint8_t *)&nic_mask, - sizeof(struct rte_flow_item_ipv4), error); - if (ret < 0) - return ret; - flow->layers |= MLX5_FLOW_LAYER_OUTER_L3_IPV4; - if (spec) { - ipv4.val = (struct ibv_flow_ipv4_ext_filter){ - .src_ip = spec->hdr.src_addr, - .dst_ip = spec->hdr.dst_addr, - .proto = spec->hdr.next_proto_id, - .tos = spec->hdr.type_of_service, - }; - ipv4.mask = (struct ibv_flow_ipv4_ext_filter){ - .src_ip = mask->hdr.src_addr, - .dst_ip = mask->hdr.dst_addr, - .proto = mask->hdr.next_proto_id, - .tos = mask->hdr.type_of_service, - }; - /* Remove unwanted bits from values. */ - ipv4.val.src_ip &= ipv4.mask.src_ip; - ipv4.val.dst_ip &= ipv4.mask.dst_ip; - ipv4.val.proto &= ipv4.mask.proto; - ipv4.val.tos &= ipv4.mask.tos; - } - flow->l3_protocol_en = !!ipv4.mask.proto; - flow->l3_protocol = ipv4.val.proto; - if (size <= flow_size) - mlx5_flow_spec_verbs_add(flow, &ipv4, size); - return size; + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "mark action not supported for " + "egress"); + return 0; } -/** - * Convert the @p item into a Verbs specification after ensuring the NIC - * will understand and process it correctly. - * If the necessary size for the conversion is greater than the @p flow_size, - * nothing is written in @p flow, the validation is still performed. +/* + * Validate the drop action. * - * @param[in] item - * Item specification. - * @param[in, out] flow - * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small, nothing is - * written. + * @param[in] action_flags + * Bit-fields that holds the actions detected until now. + * @param[in] attr + * Attributes of flow that includes this action. * @param[out] error * Pointer to error structure. * * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p item has fully been converted, - * otherwise another call with this returned memory size should be done. - * On error, a negative errno value is returned and rte_errno is set. + * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int -mlx5_flow_item_ipv6(const struct rte_flow_item *item, struct rte_flow *flow, - const size_t flow_size, struct rte_flow_error *error) +int +mlx5_flow_validate_action_drop(uint64_t action_flags, + const struct rte_flow_attr *attr, + struct rte_flow_error *error) { - const struct rte_flow_item_ipv6 *spec = item->spec; - const struct rte_flow_item_ipv6 *mask = item->mask; - const struct rte_flow_item_ipv6 nic_mask = { + if (action_flags & MLX5_FLOW_ACTION_FLAG) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't drop and flag in same flow"); + if (action_flags & MLX5_FLOW_ACTION_MARK) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't drop and mark in same flow"); + if (action_flags & MLX5_FLOW_FATE_ACTIONS) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't have 2 fate actions in" + " same flow"); + if (attr->egress) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "drop action not supported for " + "egress"); + return 0; +} + +/* + * Validate the queue action. + * + * @param[in] action + * Pointer to the queue action. + * @param[in] action_flags + * Bit-fields that holds the actions detected until now. + * @param[in] dev + * Pointer to the Ethernet device structure. + * @param[in] attr + * Attributes of flow that includes this action. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_action_queue(const struct rte_flow_action *action, + uint64_t action_flags, + struct rte_eth_dev *dev, + const struct rte_flow_attr *attr, + struct rte_flow_error *error) +{ + struct mlx5_priv *priv = dev->data->dev_private; + const struct rte_flow_action_queue *queue = action->conf; + + if (action_flags & MLX5_FLOW_FATE_ACTIONS) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't have 2 fate actions in" + " same flow"); + if (!priv->rxqs_n) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + NULL, "No Rx queues configured"); + if (queue->index >= priv->rxqs_n) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &queue->index, + "queue index out of range"); + if (!(*priv->rxqs)[queue->index]) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &queue->index, + "queue is not configured"); + if (attr->egress) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "queue action not supported for " + "egress"); + return 0; +} + +/* + * Validate the rss action. + * + * @param[in] action + * Pointer to the queue action. + * @param[in] action_flags + * Bit-fields that holds the actions detected until now. + * @param[in] dev + * Pointer to the Ethernet device structure. + * @param[in] attr + * Attributes of flow that includes this action. + * @param[in] item_flags + * Items that were detected. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_action_rss(const struct rte_flow_action *action, + uint64_t action_flags, + struct rte_eth_dev *dev, + const struct rte_flow_attr *attr, + uint64_t item_flags, + struct rte_flow_error *error) +{ + struct mlx5_priv *priv = dev->data->dev_private; + const struct rte_flow_action_rss *rss = action->conf; + int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + unsigned int i; + + if (action_flags & MLX5_FLOW_FATE_ACTIONS) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION, NULL, + "can't have 2 fate actions" + " in same flow"); + if (rss->func != RTE_ETH_HASH_FUNCTION_DEFAULT && + rss->func != RTE_ETH_HASH_FUNCTION_TOEPLITZ) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->func, + "RSS hash function not supported"); +#ifdef HAVE_IBV_DEVICE_TUNNEL_SUPPORT + if (rss->level > 2) +#else + if (rss->level > 1) +#endif + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->level, + "tunnel RSS is not supported"); + /* allow RSS key_len 0 in case of NULL (default) RSS key. */ + if (rss->key_len == 0 && rss->key != NULL) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->key_len, + "RSS hash key length 0"); + if (rss->key_len > 0 && rss->key_len < MLX5_RSS_HASH_KEY_LEN) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->key_len, + "RSS hash key too small"); + if (rss->key_len > MLX5_RSS_HASH_KEY_LEN) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->key_len, + "RSS hash key too large"); + if (rss->queue_num > priv->config.ind_table_max_size) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->queue_num, + "number of queues too large"); + if (rss->types & MLX5_RSS_HF_MASK) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->types, + "some RSS protocols are not" + " supported"); + if (!priv->rxqs_n) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + NULL, "No Rx queues configured"); + if (!rss->queue_num) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, + NULL, "No queues configured"); + for (i = 0; i != rss->queue_num; ++i) { + if (!(*priv->rxqs)[rss->queue[i]]) + return rte_flow_error_set + (error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION_CONF, + &rss->queue[i], "queue is not configured"); + } + if (attr->egress) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "rss action not supported for " + "egress"); + if (rss->level > 1 && !tunnel) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ACTION_CONF, NULL, + "inner RSS is not supported for " + "non-tunnel flows"); + return 0; +} + +/* + * Validate the count action. + * + * @param[in] dev + * Pointer to the Ethernet device structure. + * @param[in] attr + * Attributes of flow that includes this action. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_action_count(struct rte_eth_dev *dev __rte_unused, + const struct rte_flow_attr *attr, + struct rte_flow_error *error) +{ + if (attr->egress) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "count action not supported for " + "egress"); + return 0; +} + +/** + * Verify the @p attributes will be correctly understood by the NIC and store + * them in the @p flow if everything is correct. + * + * @param[in] dev + * Pointer to the Ethernet device structure. + * @param[in] attributes + * Pointer to flow attributes + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_attributes(struct rte_eth_dev *dev, + const struct rte_flow_attr *attributes, + struct rte_flow_error *error) +{ + struct mlx5_priv *priv = dev->data->dev_private; + uint32_t priority_max = priv->config.flow_prio - 1; + + if (attributes->group) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_GROUP, + NULL, "groups is not supported"); + if (attributes->priority != MLX5_FLOW_PRIO_RSVD && + attributes->priority >= priority_max) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, + NULL, "priority out of range"); + if (attributes->egress) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, NULL, + "egress is not supported"); + if (attributes->transfer && !priv->config.dv_esw_en) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER, + NULL, "transfer is not supported"); + if (!attributes->ingress) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ATTR_INGRESS, + NULL, + "ingress attribute is mandatory"); + return 0; +} + +/** + * Validate ICMP6 item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_icmp6(const struct rte_flow_item *item, + uint64_t item_flags, + uint8_t target_protocol, + struct rte_flow_error *error) +{ + const struct rte_flow_item_icmp6 *mask = item->mask; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t l3m = tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV6 : + MLX5_FLOW_LAYER_OUTER_L3_IPV6; + const uint64_t l4m = tunnel ? MLX5_FLOW_LAYER_INNER_L4 : + MLX5_FLOW_LAYER_OUTER_L4; + int ret; + + if (target_protocol != 0xFF && target_protocol != IPPROTO_ICMPV6) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "protocol filtering not compatible" + " with ICMP6 layer"); + if (!(item_flags & l3m)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "IPv6 is mandatory to filter on" + " ICMP6"); + if (item_flags & l4m) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L4 layers not supported"); + if (!mask) + mask = &rte_flow_item_icmp6_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&rte_flow_item_icmp6_mask, + sizeof(struct rte_flow_item_icmp6), error); + if (ret < 0) + return ret; + return 0; +} + +/** + * Validate ICMP item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_icmp(const struct rte_flow_item *item, + uint64_t item_flags, + uint8_t target_protocol, + struct rte_flow_error *error) +{ + const struct rte_flow_item_icmp *mask = item->mask; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t l3m = tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV4 : + MLX5_FLOW_LAYER_OUTER_L3_IPV4; + const uint64_t l4m = tunnel ? MLX5_FLOW_LAYER_INNER_L4 : + MLX5_FLOW_LAYER_OUTER_L4; + int ret; + + if (target_protocol != 0xFF && target_protocol != IPPROTO_ICMP) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "protocol filtering not compatible" + " with ICMP layer"); + if (!(item_flags & l3m)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "IPv4 is mandatory to filter" + " on ICMP"); + if (item_flags & l4m) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L4 layers not supported"); + if (!mask) + mask = &rte_flow_item_icmp_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&rte_flow_item_icmp_mask, + sizeof(struct rte_flow_item_icmp), error); + if (ret < 0) + return ret; + return 0; +} + +/** + * Validate Ethernet item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_eth(const struct rte_flow_item *item, + uint64_t item_flags, + struct rte_flow_error *error) +{ + const struct rte_flow_item_eth *mask = item->mask; + const struct rte_flow_item_eth nic_mask = { + .dst.addr_bytes = "\xff\xff\xff\xff\xff\xff", + .src.addr_bytes = "\xff\xff\xff\xff\xff\xff", + .type = RTE_BE16(0xffff), + }; + int ret; + int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t ethm = tunnel ? MLX5_FLOW_LAYER_INNER_L2 : + MLX5_FLOW_LAYER_OUTER_L2; + + if (item_flags & ethm) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L2 layers not supported"); + if (!mask) + mask = &rte_flow_item_eth_mask; + ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask, + (const uint8_t *)&nic_mask, + sizeof(struct rte_flow_item_eth), + error); + return ret; +} + +/** + * Validate VLAN item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_vlan(const struct rte_flow_item *item, + uint64_t item_flags, + struct rte_flow_error *error) +{ + const struct rte_flow_item_vlan *spec = item->spec; + const struct rte_flow_item_vlan *mask = item->mask; + const struct rte_flow_item_vlan nic_mask = { + .tci = RTE_BE16(0x0fff), + .inner_type = RTE_BE16(0xffff), + }; + uint16_t vlan_tag = 0; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + int ret; + const uint64_t l34m = tunnel ? (MLX5_FLOW_LAYER_INNER_L3 | + MLX5_FLOW_LAYER_INNER_L4) : + (MLX5_FLOW_LAYER_OUTER_L3 | + MLX5_FLOW_LAYER_OUTER_L4); + const uint64_t vlanm = tunnel ? MLX5_FLOW_LAYER_INNER_VLAN : + MLX5_FLOW_LAYER_OUTER_VLAN; + + if (item_flags & vlanm) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple VLAN layers not supported"); + else if ((item_flags & l34m) != 0) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L2 layer cannot follow L3/L4 layer"); + if (!mask) + mask = &rte_flow_item_vlan_mask; + ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask, + (const uint8_t *)&nic_mask, + sizeof(struct rte_flow_item_vlan), + error); + if (ret) + return ret; + if (spec) { + vlan_tag = spec->tci; + vlan_tag &= mask->tci; + } + /* + * From verbs perspective an empty VLAN is equivalent + * to a packet without VLAN layer. + */ + if (!vlan_tag) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM_SPEC, + item->spec, + "VLAN cannot be empty"); + return 0; +} + +/** + * Validate IPV4 item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] acc_mask + * Acceptable mask, if NULL default internal default mask + * will be used to check whether item fields are supported. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_ipv4(const struct rte_flow_item *item, + uint64_t item_flags, + const struct rte_flow_item_ipv4 *acc_mask, + struct rte_flow_error *error) +{ + const struct rte_flow_item_ipv4 *mask = item->mask; + const struct rte_flow_item_ipv4 *spec = item->spec; + const struct rte_flow_item_ipv4 nic_mask = { + .hdr = { + .src_addr = RTE_BE32(0xffffffff), + .dst_addr = RTE_BE32(0xffffffff), + .type_of_service = 0xff, + .next_proto_id = 0xff, + }, + }; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t l3m = tunnel ? MLX5_FLOW_LAYER_INNER_L3 : + MLX5_FLOW_LAYER_OUTER_L3; + const uint64_t l4m = tunnel ? MLX5_FLOW_LAYER_INNER_L4 : + MLX5_FLOW_LAYER_OUTER_L4; + int ret; + uint8_t next_proto = 0xFF; + + if (item_flags & MLX5_FLOW_LAYER_IPIP) { + if (mask && spec) + next_proto = mask->hdr.next_proto_id & + spec->hdr.next_proto_id; + if (next_proto == IPPROTO_IPIP || next_proto == IPPROTO_IPV6) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, + item, + "multiple tunnel " + "not supported"); + } + if (item_flags & MLX5_FLOW_LAYER_IPV6_ENCAP) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "wrong tunnel type - IPv6 specified " + "but IPv4 item provided"); + if (item_flags & l3m) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L3 layers not supported"); + else if (item_flags & l4m) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 cannot follow an L4 layer."); + else if ((item_flags & MLX5_FLOW_LAYER_NVGRE) && + !(item_flags & MLX5_FLOW_LAYER_INNER_L2)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 cannot follow an NVGRE layer."); + if (!mask) + mask = &rte_flow_item_ipv4_mask; + else if (mask->hdr.next_proto_id != 0 && + mask->hdr.next_proto_id != 0xff) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM_MASK, mask, + "partial mask is not supported" + " for protocol"); + ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask, + acc_mask ? (const uint8_t *)acc_mask + : (const uint8_t *)&nic_mask, + sizeof(struct rte_flow_item_ipv4), + error); + if (ret < 0) + return ret; + return 0; +} + +/** + * Validate IPV6 item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] acc_mask + * Acceptable mask, if NULL default internal default mask + * will be used to check whether item fields are supported. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_ipv6(const struct rte_flow_item *item, + uint64_t item_flags, + const struct rte_flow_item_ipv6 *acc_mask, + struct rte_flow_error *error) +{ + const struct rte_flow_item_ipv6 *mask = item->mask; + const struct rte_flow_item_ipv6 *spec = item->spec; + const struct rte_flow_item_ipv6 nic_mask = { .hdr = { .src_addr = "\xff\xff\xff\xff\xff\xff\xff\xff" @@ -640,285 +1387,797 @@ mlx5_flow_item_ipv6(const struct rte_flow_item *item, struct rte_flow *flow, .hop_limits = 0xff, }, }; - unsigned int size = sizeof(struct ibv_flow_spec_ipv6); - struct ibv_flow_spec_ipv6 ipv6 = { - .type = IBV_FLOW_SPEC_IPV6, - .size = size, - }; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t l3m = tunnel ? MLX5_FLOW_LAYER_INNER_L3 : + MLX5_FLOW_LAYER_OUTER_L3; + const uint64_t l4m = tunnel ? MLX5_FLOW_LAYER_INNER_L4 : + MLX5_FLOW_LAYER_OUTER_L4; + int ret; + uint8_t next_proto = 0xFF; + + if (item_flags & MLX5_FLOW_LAYER_IPV6_ENCAP) { + if (mask && spec) + next_proto = mask->hdr.proto & spec->hdr.proto; + if (next_proto == IPPROTO_IPIP || next_proto == IPPROTO_IPV6) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, + item, + "multiple tunnel " + "not supported"); + } + if (item_flags & MLX5_FLOW_LAYER_IPIP) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "wrong tunnel type - IPv4 specified " + "but IPv6 item provided"); + if (item_flags & l3m) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L3 layers not supported"); + else if (item_flags & l4m) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 cannot follow an L4 layer."); + else if ((item_flags & MLX5_FLOW_LAYER_NVGRE) && + !(item_flags & MLX5_FLOW_LAYER_INNER_L2)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 cannot follow an NVGRE layer."); + if (!mask) + mask = &rte_flow_item_ipv6_mask; + ret = mlx5_flow_item_acceptable(item, (const uint8_t *)mask, + acc_mask ? (const uint8_t *)acc_mask + : (const uint8_t *)&nic_mask, + sizeof(struct rte_flow_item_ipv6), + error); + if (ret < 0) + return ret; + return 0; +} + +/** + * Validate UDP item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] target_protocol + * The next protocol in the previous item. + * @param[in] flow_mask + * mlx5 flow-specific (DV, verbs, etc.) supported header fields mask. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_udp(const struct rte_flow_item *item, + uint64_t item_flags, + uint8_t target_protocol, + struct rte_flow_error *error) +{ + const struct rte_flow_item_udp *mask = item->mask; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t l3m = tunnel ? MLX5_FLOW_LAYER_INNER_L3 : + MLX5_FLOW_LAYER_OUTER_L3; + const uint64_t l4m = tunnel ? MLX5_FLOW_LAYER_INNER_L4 : + MLX5_FLOW_LAYER_OUTER_L4; + int ret; + + if (target_protocol != 0xff && target_protocol != IPPROTO_UDP) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "protocol filtering not compatible" + " with UDP layer"); + if (!(item_flags & l3m)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 is mandatory to filter on L4"); + if (item_flags & l4m) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L4 layers not supported"); + if (!mask) + mask = &rte_flow_item_udp_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&rte_flow_item_udp_mask, + sizeof(struct rte_flow_item_udp), error); + if (ret < 0) + return ret; + return 0; +} + +/** + * Validate TCP item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] target_protocol + * The next protocol in the previous item. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_tcp(const struct rte_flow_item *item, + uint64_t item_flags, + uint8_t target_protocol, + const struct rte_flow_item_tcp *flow_mask, + struct rte_flow_error *error) +{ + const struct rte_flow_item_tcp *mask = item->mask; + const int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL); + const uint64_t l3m = tunnel ? MLX5_FLOW_LAYER_INNER_L3 : + MLX5_FLOW_LAYER_OUTER_L3; + const uint64_t l4m = tunnel ? MLX5_FLOW_LAYER_INNER_L4 : + MLX5_FLOW_LAYER_OUTER_L4; + int ret; + + assert(flow_mask); + if (target_protocol != 0xff && target_protocol != IPPROTO_TCP) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "protocol filtering not compatible" + " with TCP layer"); + if (!(item_flags & l3m)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 is mandatory to filter on L4"); + if (item_flags & l4m) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple L4 layers not supported"); + if (!mask) + mask = &rte_flow_item_tcp_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)flow_mask, + sizeof(struct rte_flow_item_tcp), error); + if (ret < 0) + return ret; + return 0; +} + +/** + * Validate VXLAN item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] target_protocol + * The next protocol in the previous item. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_vxlan(const struct rte_flow_item *item, + uint64_t item_flags, + struct rte_flow_error *error) +{ + const struct rte_flow_item_vxlan *spec = item->spec; + const struct rte_flow_item_vxlan *mask = item->mask; int ret; + union vni { + uint32_t vlan_id; + uint8_t vni[4]; + } id = { .vlan_id = 0, }; + uint32_t vlan_id = 0; - if (flow->layers & MLX5_FLOW_LAYER_OUTER_L3) + + if (item_flags & MLX5_FLOW_LAYER_TUNNEL) return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "multiple L3 layers not supported"); - else if (flow->layers & MLX5_FLOW_LAYER_OUTER_L4) + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple tunnel layers not" + " supported"); + /* + * Verify only UDPv4 is present as defined in + * https://tools.ietf.org/html/rfc7348 + */ + if (!(item_flags & MLX5_FLOW_LAYER_OUTER_L4_UDP)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "no outer UDP layer found"); + if (!mask) + mask = &rte_flow_item_vxlan_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&rte_flow_item_vxlan_mask, + sizeof(struct rte_flow_item_vxlan), + error); + if (ret < 0) + return ret; + if (spec) { + memcpy(&id.vni[1], spec->vni, 3); + vlan_id = id.vlan_id; + memcpy(&id.vni[1], mask->vni, 3); + vlan_id &= id.vlan_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 (!vlan_id) return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "L3 cannot follow an L4 layer."); + RTE_FLOW_ERROR_TYPE_ITEM, item, + "VXLAN vni cannot be 0"); + if (!(item_flags & MLX5_FLOW_LAYER_OUTER)) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "VXLAN tunnel must be fully defined"); + return 0; +} + +/** + * Validate VXLAN_GPE item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] priv + * Pointer to the private data structure. + * @param[in] target_protocol + * The next protocol in the previous item. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_vxlan_gpe(const struct rte_flow_item *item, + uint64_t item_flags, + struct rte_eth_dev *dev, + struct rte_flow_error *error) +{ + struct mlx5_priv *priv = dev->data->dev_private; + const struct rte_flow_item_vxlan_gpe *spec = item->spec; + const struct rte_flow_item_vxlan_gpe *mask = item->mask; + int ret; + union vni { + uint32_t vlan_id; + uint8_t vni[4]; + } id = { .vlan_id = 0, }; + uint32_t vlan_id = 0; + + if (!priv->config.l3_vxlan_en) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 VXLAN is not enabled by device" + " parameter and/or not configured in" + " firmware"); + if (item_flags & MLX5_FLOW_LAYER_TUNNEL) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple tunnel layers not" + " supported"); + /* + * Verify only UDPv4 is present as defined in + * https://tools.ietf.org/html/rfc7348 + */ + if (!(item_flags & MLX5_FLOW_LAYER_OUTER_L4_UDP)) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "no outer UDP layer found"); if (!mask) - mask = &rte_flow_item_ipv6_mask; + mask = &rte_flow_item_vxlan_gpe_mask; ret = mlx5_flow_item_acceptable (item, (const uint8_t *)mask, - (const uint8_t *)&nic_mask, - sizeof(struct rte_flow_item_ipv6), error); + (const uint8_t *)&rte_flow_item_vxlan_gpe_mask, + sizeof(struct rte_flow_item_vxlan_gpe), + error); if (ret < 0) return ret; - flow->layers |= MLX5_FLOW_LAYER_OUTER_L3_IPV6; if (spec) { - unsigned int i; - uint32_t vtc_flow_val; - uint32_t vtc_flow_mask; - - memcpy(&ipv6.val.src_ip, spec->hdr.src_addr, - RTE_DIM(ipv6.val.src_ip)); - memcpy(&ipv6.val.dst_ip, spec->hdr.dst_addr, - RTE_DIM(ipv6.val.dst_ip)); - memcpy(&ipv6.mask.src_ip, mask->hdr.src_addr, - RTE_DIM(ipv6.mask.src_ip)); - memcpy(&ipv6.mask.dst_ip, mask->hdr.dst_addr, - RTE_DIM(ipv6.mask.dst_ip)); - 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. */ - for (i = 0; i < RTE_DIM(ipv6.val.src_ip); ++i) { - ipv6.val.src_ip[i] &= ipv6.mask.src_ip[i]; - 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; + if (spec->protocol) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, + item, + "VxLAN-GPE protocol" + " not supported"); + memcpy(&id.vni[1], spec->vni, 3); + vlan_id = id.vlan_id; + memcpy(&id.vni[1], mask->vni, 3); + vlan_id &= id.vlan_id; } - flow->l3_protocol_en = !!ipv6.mask.next_hdr; - flow->l3_protocol = ipv6.val.next_hdr; - if (size <= flow_size) - mlx5_flow_spec_verbs_add(flow, &ipv6, size); - return size; + /* + * 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 (!vlan_id) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "VXLAN-GPE vni cannot be 0"); + if (!(item_flags & MLX5_FLOW_LAYER_OUTER)) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "VXLAN-GPE tunnel must be fully" + " defined"); + return 0; +} +/** + * Validate GRE Key item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit flags to mark detected items. + * @param[in] gre_item + * Pointer to gre_item + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_gre_key(const struct rte_flow_item *item, + uint64_t item_flags, + const struct rte_flow_item *gre_item, + struct rte_flow_error *error) +{ + const rte_be32_t *mask = item->mask; + int ret = 0; + rte_be32_t gre_key_default_mask = RTE_BE32(UINT32_MAX); + const struct rte_flow_item_gre *gre_spec = gre_item->spec; + const struct rte_flow_item_gre *gre_mask = gre_item->mask; + + if (item_flags & MLX5_FLOW_LAYER_GRE_KEY) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "Multiple GRE key not support"); + if (!(item_flags & MLX5_FLOW_LAYER_GRE)) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "No preceding GRE header"); + if (item_flags & MLX5_FLOW_LAYER_INNER) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "GRE key following a wrong item"); + if (!gre_mask) + gre_mask = &rte_flow_item_gre_mask; + if (gre_spec && (gre_mask->c_rsvd0_ver & RTE_BE16(0x2000)) && + !(gre_spec->c_rsvd0_ver & RTE_BE16(0x2000))) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "Key bit must be on"); + + if (!mask) + mask = &gre_key_default_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&gre_key_default_mask, + sizeof(rte_be32_t), error); + return ret; } /** - * Convert the @p item into a Verbs specification after ensuring the NIC - * will understand and process it correctly. - * If the necessary size for the conversion is greater than the @p flow_size, - * nothing is written in @p flow, the validation is still performed. + * Validate GRE item. * * @param[in] item * Item specification. - * @param[in, out] flow - * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small, nothing is - * written. + * @param[in] item_flags + * Bit flags to mark detected items. + * @param[in] target_protocol + * The next protocol in the previous item. * @param[out] error * Pointer to error structure. * * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p item has fully been converted, - * otherwise another call with this returned memory size should be done. - * On error, a negative errno value is returned and rte_errno is set. + * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int -mlx5_flow_item_udp(const struct rte_flow_item *item, struct rte_flow *flow, - const size_t flow_size, struct rte_flow_error *error) +int +mlx5_flow_validate_item_gre(const struct rte_flow_item *item, + uint64_t item_flags, + uint8_t target_protocol, + struct rte_flow_error *error) { - const struct rte_flow_item_udp *spec = item->spec; - const struct rte_flow_item_udp *mask = item->mask; - unsigned int size = sizeof(struct ibv_flow_spec_tcp_udp); - struct ibv_flow_spec_tcp_udp udp = { - .type = IBV_FLOW_SPEC_UDP, - .size = size, - }; + const struct rte_flow_item_gre *spec __rte_unused = item->spec; + const struct rte_flow_item_gre *mask = item->mask; int ret; + const struct rte_flow_item_gre nic_mask = { + .c_rsvd0_ver = RTE_BE16(0xB000), + .protocol = RTE_BE16(UINT16_MAX), + }; - if (!(flow->layers & MLX5_FLOW_LAYER_OUTER_L3)) + if (target_protocol != 0xff && target_protocol != IPPROTO_GRE) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "protocol filtering not compatible" + " with this GRE layer"); + if (item_flags & MLX5_FLOW_LAYER_TUNNEL) return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "L3 is mandatory to filter on L4"); - if (flow->layers & MLX5_FLOW_LAYER_OUTER_L4) + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple tunnel layers not" + " supported"); + if (!(item_flags & MLX5_FLOW_LAYER_OUTER_L3)) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 Layer is missing"); + if (!mask) + mask = &rte_flow_item_gre_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&nic_mask, + sizeof(struct rte_flow_item_gre), error); + if (ret < 0) + return ret; +#ifndef HAVE_MLX5DV_DR +#ifndef HAVE_IBV_DEVICE_MPLS_SUPPORT + if (spec && (spec->protocol & mask->protocol)) return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, - "L4 layer is already present"); - if (flow->l3_protocol_en && flow->l3_protocol != MLX5_IP_PROTOCOL_UDP) + RTE_FLOW_ERROR_TYPE_ITEM, item, + "without MPLS support the" + " specification cannot be used for" + " filtering"); +#endif +#endif + return 0; +} + +/** + * Validate MPLS item. + * + * @param[in] dev + * Pointer to the rte_eth_dev structure. + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit-fields that holds the items detected until now. + * @param[in] prev_layer + * The protocol layer indicated in previous item. + * @param[out] error + * Pointer to error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +int +mlx5_flow_validate_item_mpls(struct rte_eth_dev *dev __rte_unused, + const struct rte_flow_item *item __rte_unused, + uint64_t item_flags __rte_unused, + uint64_t prev_layer __rte_unused, + struct rte_flow_error *error) +{ +#ifdef HAVE_IBV_DEVICE_MPLS_SUPPORT + const struct rte_flow_item_mpls *mask = item->mask; + struct mlx5_priv *priv = dev->data->dev_private; + int ret; + + if (!priv->config.mpls_en) return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - item, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "MPLS not supported or" + " disabled in firmware" + " configuration."); + /* MPLS over IP, UDP, GRE is allowed */ + if (!(prev_layer & (MLX5_FLOW_LAYER_OUTER_L3 | + MLX5_FLOW_LAYER_OUTER_L4_UDP | + MLX5_FLOW_LAYER_GRE))) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, "protocol filtering not compatible" - " with UDP layer"); + " with MPLS layer"); + /* Multi-tunnel isn't allowed but MPLS over GRE is an exception. */ + if ((item_flags & MLX5_FLOW_LAYER_TUNNEL) && + !(item_flags & MLX5_FLOW_LAYER_GRE)) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple tunnel layers not" + " supported"); if (!mask) - mask = &rte_flow_item_udp_mask; + mask = &rte_flow_item_mpls_mask; ret = mlx5_flow_item_acceptable (item, (const uint8_t *)mask, - (const uint8_t *)&rte_flow_item_udp_mask, - sizeof(struct rte_flow_item_udp), error); + (const uint8_t *)&rte_flow_item_mpls_mask, + sizeof(struct rte_flow_item_mpls), error); if (ret < 0) return ret; - flow->layers |= MLX5_FLOW_LAYER_OUTER_L4_UDP; - if (size > flow_size) - return size; - if (spec) { - udp.val.dst_port = spec->hdr.dst_port; - udp.val.src_port = spec->hdr.src_port; - udp.mask.dst_port = mask->hdr.dst_port; - udp.mask.src_port = mask->hdr.src_port; - /* Remove unwanted bits from values. */ - udp.val.src_port &= udp.mask.src_port; - udp.val.dst_port &= udp.mask.dst_port; - } - mlx5_flow_spec_verbs_add(flow, &udp, size); - return size; + return 0; +#endif + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "MPLS is not supported by Verbs, please" + " update."); } /** - * Convert the @p pattern into a Verbs specifications after ensuring the NIC - * will understand and process it correctly. - * The conversion is performed item per item, each of them is written into - * the @p flow if its size is lesser or equal to @p flow_size. - * Validation and memory consumption computation are still performed until the - * end of @p pattern, unless an error is encountered. - * - * @param[in] pattern - * Flow pattern. - * @param[in, out] flow - * Pointer to the rte_flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small some - * garbage may be present. + * Validate NVGRE item. + * + * @param[in] item + * Item specification. + * @param[in] item_flags + * Bit flags to mark detected items. + * @param[in] target_protocol + * The next protocol in the previous item. * @param[out] error * Pointer to error structure. * * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @pattern has fully been - * converted, otherwise another call with this returned memory size should - * be done. - * On error, a negative errno value is returned and rte_errno is set. + * 0 on success, a negative errno value otherwise and rte_errno is set. */ +int +mlx5_flow_validate_item_nvgre(const struct rte_flow_item *item, + uint64_t item_flags, + uint8_t target_protocol, + struct rte_flow_error *error) +{ + const struct rte_flow_item_nvgre *mask = item->mask; + int ret; + + if (target_protocol != 0xff && target_protocol != IPPROTO_GRE) + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "protocol filtering not compatible" + " with this GRE layer"); + if (item_flags & MLX5_FLOW_LAYER_TUNNEL) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "multiple tunnel layers not" + " supported"); + if (!(item_flags & MLX5_FLOW_LAYER_OUTER_L3)) + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "L3 Layer is missing"); + if (!mask) + mask = &rte_flow_item_nvgre_mask; + ret = mlx5_flow_item_acceptable + (item, (const uint8_t *)mask, + (const uint8_t *)&rte_flow_item_nvgre_mask, + sizeof(struct rte_flow_item_nvgre), error); + if (ret < 0) + return ret; + return 0; +} + +static int +flow_null_validate(struct rte_eth_dev *dev __rte_unused, + const struct rte_flow_attr *attr __rte_unused, + const struct rte_flow_item items[] __rte_unused, + const struct rte_flow_action actions[] __rte_unused, + struct rte_flow_error *error) +{ + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL); +} + +static struct mlx5_flow * +flow_null_prepare(const struct rte_flow_attr *attr __rte_unused, + const struct rte_flow_item items[] __rte_unused, + const struct rte_flow_action actions[] __rte_unused, + struct rte_flow_error *error) +{ + rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL); + return NULL; +} + +static int +flow_null_translate(struct rte_eth_dev *dev __rte_unused, + struct mlx5_flow *dev_flow __rte_unused, + const struct rte_flow_attr *attr __rte_unused, + const struct rte_flow_item items[] __rte_unused, + const struct rte_flow_action actions[] __rte_unused, + struct rte_flow_error *error) +{ + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL); +} + static int -mlx5_flow_items(const struct rte_flow_item pattern[], - struct rte_flow *flow, const size_t flow_size, +flow_null_apply(struct rte_eth_dev *dev __rte_unused, + struct rte_flow *flow __rte_unused, struct rte_flow_error *error) { - int remain = flow_size; - size_t size = 0; + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL); +} - for (; pattern->type != RTE_FLOW_ITEM_TYPE_END; pattern++) { - int ret = 0; +static void +flow_null_remove(struct rte_eth_dev *dev __rte_unused, + struct rte_flow *flow __rte_unused) +{ +} - switch (pattern->type) { - case RTE_FLOW_ITEM_TYPE_VOID: - break; - case RTE_FLOW_ITEM_TYPE_ETH: - ret = mlx5_flow_item_eth(pattern, flow, remain, error); - break; - case RTE_FLOW_ITEM_TYPE_VLAN: - ret = mlx5_flow_item_vlan(pattern, flow, remain, error); - break; - case RTE_FLOW_ITEM_TYPE_IPV4: - ret = mlx5_flow_item_ipv4(pattern, flow, remain, error); - break; - case RTE_FLOW_ITEM_TYPE_IPV6: - ret = mlx5_flow_item_ipv6(pattern, flow, remain, error); - break; - case RTE_FLOW_ITEM_TYPE_UDP: - ret = mlx5_flow_item_udp(pattern, flow, remain, error); - break; - default: - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ITEM, - pattern, - "item not supported"); - } - if (ret < 0) - return ret; - if (remain > ret) - remain -= ret; - else - remain = 0; - size += ret; - } - if (!flow->layers) { - const struct rte_flow_item item = { - .type = RTE_FLOW_ITEM_TYPE_ETH, - }; +static void +flow_null_destroy(struct rte_eth_dev *dev __rte_unused, + struct rte_flow *flow __rte_unused) +{ +} + +static int +flow_null_query(struct rte_eth_dev *dev __rte_unused, + struct rte_flow *flow __rte_unused, + const struct rte_flow_action *actions __rte_unused, + void *data __rte_unused, + struct rte_flow_error *error) +{ + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL, NULL); +} + +/* Void driver to protect from null pointer reference. */ +const struct mlx5_flow_driver_ops mlx5_flow_null_drv_ops = { + .validate = flow_null_validate, + .prepare = flow_null_prepare, + .translate = flow_null_translate, + .apply = flow_null_apply, + .remove = flow_null_remove, + .destroy = flow_null_destroy, + .query = flow_null_query, +}; + +/** + * Select flow driver type according to flow attributes and device + * configuration. + * + * @param[in] dev + * Pointer to the dev structure. + * @param[in] attr + * Pointer to the flow attributes. + * + * @return + * flow driver type, MLX5_FLOW_TYPE_MAX otherwise. + */ +static enum mlx5_flow_drv_type +flow_get_drv_type(struct rte_eth_dev *dev, const struct rte_flow_attr *attr) +{ + struct mlx5_priv *priv = dev->data->dev_private; + enum mlx5_flow_drv_type type = MLX5_FLOW_TYPE_MAX; + + if (attr->transfer && priv->config.dv_esw_en) + type = MLX5_FLOW_TYPE_DV; + if (!attr->transfer) + type = priv->config.dv_flow_en ? MLX5_FLOW_TYPE_DV : + MLX5_FLOW_TYPE_VERBS; + return type; +} + +#define flow_get_drv_ops(type) flow_drv_ops[type] + +/** + * Flow driver validation API. This abstracts calling driver specific functions. + * The type of flow driver is determined according to flow attributes. + * + * @param[in] dev + * Pointer to the dev structure. + * @param[in] attr + * Pointer to the flow attributes. + * @param[in] items + * Pointer to the list of items. + * @param[in] actions + * Pointer to the list of actions. + * @param[out] error + * Pointer to the error structure. + * + * @return + * 0 on success, a negative errno value otherwise and rte_errno is set. + */ +static inline int +flow_drv_validate(struct rte_eth_dev *dev, + const struct rte_flow_attr *attr, + const struct rte_flow_item items[], + const struct rte_flow_action actions[], + struct rte_flow_error *error) +{ + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type type = flow_get_drv_type(dev, attr); + + fops = flow_get_drv_ops(type); + return fops->validate(dev, attr, items, actions, error); +} + +/** + * Flow driver preparation API. This abstracts calling driver specific + * functions. Parent flow (rte_flow) should have driver type (drv_type). It + * calculates the size of memory required for device flow, allocates the memory, + * initializes the device flow and returns the pointer. + * + * @note + * This function initializes device flow structure such as dv or verbs in + * struct mlx5_flow. However, it is caller's responsibility to initialize the + * rest. For example, adding returning device flow to flow->dev_flow list and + * setting backward reference to the flow should be done out of this function. + * layers field is not filled either. + * + * @param[in] attr + * Pointer to the flow attributes. + * @param[in] items + * Pointer to the list of items. + * @param[in] actions + * Pointer to the list of actions. + * @param[out] error + * Pointer to the error structure. + * + * @return + * Pointer to device flow on success, otherwise NULL and rte_errno is set. + */ +static inline struct mlx5_flow * +flow_drv_prepare(const struct rte_flow *flow, + const struct rte_flow_attr *attr, + const struct rte_flow_item items[], + const struct rte_flow_action actions[], + struct rte_flow_error *error) +{ + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type type = flow->drv_type; - return mlx5_flow_item_eth(&item, flow, flow_size, error); - } - return size; + assert(type > MLX5_FLOW_TYPE_MIN && type < MLX5_FLOW_TYPE_MAX); + fops = flow_get_drv_ops(type); + return fops->prepare(attr, items, actions, error); } /** - * Convert the @p action into a Verbs specification after ensuring the NIC - * will understand and process it correctly. - * If the necessary size for the conversion is greater than the @p flow_size, - * nothing is written in @p flow, the validation is still performed. + * Flow driver translation API. This abstracts calling driver specific + * functions. Parent flow (rte_flow) should have driver type (drv_type). It + * translates a generic flow into a driver flow. flow_drv_prepare() must + * precede. * - * @param[in] action - * Action configuration. - * @param[in, out] flow - * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small, nothing is - * written. + * @note + * dev_flow->layers could be filled as a result of parsing during translation + * if needed by flow_drv_apply(). dev_flow->flow->actions can also be filled + * if necessary. As a flow can have multiple dev_flows by RSS flow expansion, + * flow->actions could be overwritten even though all the expanded dev_flows + * have the same actions. + * + * @param[in] dev + * Pointer to the rte dev structure. + * @param[in, out] dev_flow + * Pointer to the mlx5 flow. + * @param[in] attr + * Pointer to the flow attributes. + * @param[in] items + * Pointer to the list of items. + * @param[in] actions + * Pointer to the list of actions. * @param[out] error - * Pointer to error structure. + * Pointer to the error structure. * * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p action has fully been - * converted, otherwise another call with this returned memory size should - * be done. - * On error, a negative errno value is returned and rte_errno is set. + * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int -mlx5_flow_action_drop(const struct rte_flow_action *action, - struct rte_flow *flow, const size_t flow_size, - struct rte_flow_error *error) +static inline int +flow_drv_translate(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow, + const struct rte_flow_attr *attr, + const struct rte_flow_item items[], + const struct rte_flow_action actions[], + struct rte_flow_error *error) { - unsigned int size = sizeof(struct ibv_flow_spec_action_drop); - struct ibv_flow_spec_action_drop drop = { - .type = IBV_FLOW_SPEC_ACTION_DROP, - .size = size, - }; + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type type = dev_flow->flow->drv_type; - if (flow->fate) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ACTION, - action, - "multiple fate actions are not" - " supported"); - if (size < flow_size) - mlx5_flow_spec_verbs_add(flow, &drop, size); - flow->fate |= MLX5_FLOW_FATE_DROP; - return size; + assert(type > MLX5_FLOW_TYPE_MIN && type < MLX5_FLOW_TYPE_MAX); + fops = flow_get_drv_ops(type); + return fops->translate(dev, dev_flow, attr, items, actions, error); } /** - * Convert the @p action into @p flow after ensuring the NIC will understand - * and process it correctly. + * Flow driver apply API. This abstracts calling driver specific functions. + * Parent flow (rte_flow) should have driver type (drv_type). It applies + * translated driver flows on to device. flow_drv_translate() must precede. * * @param[in] dev * Pointer to Ethernet device structure. - * @param[in] action - * Action configuration. * @param[in, out] flow * Pointer to flow structure. * @param[out] error @@ -927,168 +2186,59 @@ mlx5_flow_action_drop(const struct rte_flow_action *action, * @return * 0 on success, a negative errno value otherwise and rte_errno is set. */ -static int -mlx5_flow_action_queue(struct rte_eth_dev *dev, - const struct rte_flow_action *action, - struct rte_flow *flow, - struct rte_flow_error *error) +static inline int +flow_drv_apply(struct rte_eth_dev *dev, struct rte_flow *flow, + struct rte_flow_error *error) { - struct priv *priv = dev->data->dev_private; - const struct rte_flow_action_queue *queue = action->conf; + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type type = flow->drv_type; - if (flow->fate) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ACTION, - action, - "multiple fate actions are not" - " supported"); - if (queue->index >= priv->rxqs_n) - return rte_flow_error_set(error, EINVAL, - RTE_FLOW_ERROR_TYPE_ACTION_CONF, - &queue->index, - "queue index out of range"); - if (!(*priv->rxqs)[queue->index]) - return rte_flow_error_set(error, EINVAL, - RTE_FLOW_ERROR_TYPE_ACTION_CONF, - &queue->index, - "queue is not configured"); - flow->queue = queue->index; - flow->fate |= MLX5_FLOW_FATE_QUEUE; - return 0; + assert(type > MLX5_FLOW_TYPE_MIN && type < MLX5_FLOW_TYPE_MAX); + fops = flow_get_drv_ops(type); + return fops->apply(dev, flow, error); } /** - * Convert the @p action into @p flow after ensuring the NIC will understand - * and process it correctly. - * The conversion is performed action per action, each of them is written into - * the @p flow if its size is lesser or equal to @p flow_size. - * Validation and memory consumption computation are still performed until the - * end of @p action, unless an error is encountered. + * Flow driver remove API. This abstracts calling driver specific functions. + * Parent flow (rte_flow) should have driver type (drv_type). It removes a flow + * on device. All the resources of the flow should be freed by calling + * flow_drv_destroy(). * * @param[in] dev - * Pointer to Ethernet device structure. - * @param[in] actions - * Pointer to flow actions array. + * Pointer to Ethernet device. * @param[in, out] flow - * Pointer to the rte_flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small some - * garbage may be present. - * @param[out] error - * Pointer to error structure. - * - * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the @p actions has fully been - * converted, otherwise another call with this returned memory size should - * be done. - * On error, a negative errno value is returned and rte_errno is set. + * Pointer to flow structure. */ -static int -mlx5_flow_actions(struct rte_eth_dev *dev, - const struct rte_flow_action actions[], - struct rte_flow *flow, const size_t flow_size, - struct rte_flow_error *error) +static inline void +flow_drv_remove(struct rte_eth_dev *dev, struct rte_flow *flow) { - size_t size = 0; - int remain = flow_size; - int ret = 0; + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type type = flow->drv_type; - for (; actions->type != RTE_FLOW_ACTION_TYPE_END; actions++) { - switch (actions->type) { - case RTE_FLOW_ACTION_TYPE_VOID: - break; - case RTE_FLOW_ACTION_TYPE_DROP: - ret = mlx5_flow_action_drop(actions, flow, remain, - error); - break; - case RTE_FLOW_ACTION_TYPE_QUEUE: - ret = mlx5_flow_action_queue(dev, actions, flow, error); - break; - default: - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_ACTION, - actions, - "action not supported"); - } - if (ret < 0) - return ret; - if (remain > ret) - remain -= ret; - else - remain = 0; - size += ret; - } - if (!flow->fate) - return rte_flow_error_set(error, ENOTSUP, - RTE_FLOW_ERROR_TYPE_UNSPECIFIED, - NULL, - "no fate action found"); - return size; + assert(type > MLX5_FLOW_TYPE_MIN && type < MLX5_FLOW_TYPE_MAX); + fops = flow_get_drv_ops(type); + fops->remove(dev, flow); } /** - * Convert the @p attributes, @p pattern, @p action, into an flow for the NIC - * after ensuring the NIC will understand and process it correctly. - * The conversion is only performed item/action per item/action, each of - * them is written into the @p flow if its size is lesser or equal to @p - * flow_size. - * Validation and memory consumption computation are still performed until the - * end, unless an error is encountered. + * Flow driver destroy API. This abstracts calling driver specific functions. + * Parent flow (rte_flow) should have driver type (drv_type). It removes a flow + * on device and releases resources of the flow. * * @param[in] dev * Pointer to Ethernet device. * @param[in, out] flow * Pointer to flow structure. - * @param[in] flow_size - * Size in bytes of the available space in @p flow, if too small some - * garbage may be present. - * @param[in] attributes - * Flow rule attributes. - * @param[in] pattern - * Pattern specification (list terminated by the END pattern item). - * @param[in] actions - * Associated actions (list terminated by the END action). - * @param[out] error - * Perform verbose error reporting if not NULL. - * - * @return - * On success the number of bytes consumed/necessary, if the returned value - * is lesser or equal to @p flow_size, the flow has fully been converted and - * can be applied, otherwise another call with this returned memory size - * should be done. - * On error, a negative errno value is returned and rte_errno is set. */ -static int -mlx5_flow_merge(struct rte_eth_dev *dev, struct rte_flow *flow, - const size_t flow_size, - const struct rte_flow_attr *attributes, - const struct rte_flow_item pattern[], - const struct rte_flow_action actions[], - struct rte_flow_error *error) +static inline void +flow_drv_destroy(struct rte_eth_dev *dev, struct rte_flow *flow) { - struct rte_flow local_flow = { .layers = 0, }; - size_t size = sizeof(*flow) + sizeof(struct ibv_flow_attr); - int remain = (flow_size > size) ? flow_size - size : 0; - int ret; + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type type = flow->drv_type; - if (!remain) - flow = &local_flow; - ret = mlx5_flow_attributes(dev, attributes, flow, error); - if (ret < 0) - return ret; - ret = mlx5_flow_items(pattern, flow, remain, error); - if (ret < 0) - return ret; - size += ret; - remain = (flow_size > size) ? flow_size - size : 0; - ret = mlx5_flow_actions(dev, actions, flow, remain, error); - if (ret < 0) - return ret; - size += ret; - if (size <= flow_size) - flow->verbs.attr->priority = flow->attributes.priority; - return size; + assert(type > MLX5_FLOW_TYPE_MIN && type < MLX5_FLOW_TYPE_MAX); + fops = flow_get_drv_ops(type); + fops->destroy(dev, flow); } /** @@ -1104,96 +2254,55 @@ mlx5_flow_validate(struct rte_eth_dev *dev, const struct rte_flow_action actions[], struct rte_flow_error *error) { - int ret = mlx5_flow_merge(dev, NULL, 0, attr, items, actions, error); + int ret; + ret = flow_drv_validate(dev, attr, items, actions, error); if (ret < 0) return ret; return 0; } /** - * Remove the flow. + * Get RSS action from the action list. * - * @param[in] dev - * Pointer to Ethernet device. - * @param[in, out] flow - * Pointer to flow structure. + * @param[in] actions + * Pointer to the list of actions. + * + * @return + * Pointer to the RSS action if exist, else return NULL. */ -static void -mlx5_flow_remove(struct rte_eth_dev *dev, struct rte_flow *flow) +static const struct rte_flow_action_rss* +flow_get_rss_action(const struct rte_flow_action actions[]) { - if (flow->fate & MLX5_FLOW_FATE_DROP) { - if (flow->verbs.flow) { - claim_zero(mlx5_glue->destroy_flow(flow->verbs.flow)); - flow->verbs.flow = NULL; + for (; actions->type != RTE_FLOW_ACTION_TYPE_END; actions++) { + switch (actions->type) { + case RTE_FLOW_ACTION_TYPE_RSS: + return (const struct rte_flow_action_rss *) + actions->conf; + default: + break; } } - if (flow->verbs.hrxq) { - if (flow->fate & MLX5_FLOW_FATE_DROP) - mlx5_hrxq_drop_release(dev); - else if (flow->fate & MLX5_FLOW_FATE_QUEUE) - mlx5_hrxq_release(dev, flow->verbs.hrxq); - flow->verbs.hrxq = NULL; - } + return NULL; } -/** - * Apply the flow. - * - * @param[in] dev - * Pointer to Ethernet device structure. - * @param[in, out] flow - * Pointer to flow structure. - * @param[out] error - * Pointer to error structure. - * - * @return - * 0 on success, a negative errno value otherwise and rte_errno is set. - */ -static int -mlx5_flow_apply(struct rte_eth_dev *dev, struct rte_flow *flow, - struct rte_flow_error *error) +static unsigned int +find_graph_root(const struct rte_flow_item pattern[], uint32_t rss_level) { - if (flow->fate & MLX5_FLOW_FATE_DROP) { - flow->verbs.hrxq = mlx5_hrxq_drop_new(dev); - if (!flow->verbs.hrxq) - return rte_flow_error_set - (error, errno, - RTE_FLOW_ERROR_TYPE_UNSPECIFIED, - NULL, - "cannot allocate Drop queue"); - } else if (flow->fate & MLX5_FLOW_FATE_QUEUE) { - struct mlx5_hrxq *hrxq; - - hrxq = mlx5_hrxq_get(dev, rss_hash_default_key, - rss_hash_default_key_len, 0, - &flow->queue, 1, 0, 0); - if (!hrxq) - hrxq = mlx5_hrxq_new(dev, rss_hash_default_key, - rss_hash_default_key_len, 0, - &flow->queue, 1, 0, 0); - if (!hrxq) - return rte_flow_error_set(error, rte_errno, - RTE_FLOW_ERROR_TYPE_UNSPECIFIED, - NULL, - "cannot create flow"); - flow->verbs.hrxq = hrxq; - } - flow->verbs.flow = - mlx5_glue->create_flow(flow->verbs.hrxq->qp, flow->verbs.attr); - if (!flow->verbs.flow) { - if (flow->fate & MLX5_FLOW_FATE_DROP) - mlx5_hrxq_drop_release(dev); - else - mlx5_hrxq_release(dev, flow->verbs.hrxq); - flow->verbs.hrxq = NULL; - return rte_flow_error_set(error, errno, - RTE_FLOW_ERROR_TYPE_UNSPECIFIED, - NULL, - "kernel module refuses to create" - " flow"); + const struct rte_flow_item *item; + unsigned int has_vlan = 0; + + for (item = pattern; item->type != RTE_FLOW_ITEM_TYPE_END; item++) { + if (item->type == RTE_FLOW_ITEM_TYPE_VLAN) { + has_vlan = 1; + break; + } } - return 0; + if (has_vlan) + return rss_level < 2 ? MLX5_EXPANSION_ROOT_ETH_VLAN : + MLX5_EXPANSION_ROOT_OUTER_ETH_VLAN; + return rss_level < 2 ? MLX5_EXPANSION_ROOT : + MLX5_EXPANSION_ROOT_OUTER; } /** @@ -1216,45 +2325,85 @@ mlx5_flow_apply(struct rte_eth_dev *dev, struct rte_flow *flow, * A flow on success, NULL otherwise and rte_errno is set. */ static struct rte_flow * -mlx5_flow_list_create(struct rte_eth_dev *dev, - struct mlx5_flows *list, - const struct rte_flow_attr *attr, - const struct rte_flow_item items[], - const struct rte_flow_action actions[], - struct rte_flow_error *error) +flow_list_create(struct rte_eth_dev *dev, struct mlx5_flows *list, + const struct rte_flow_attr *attr, + const struct rte_flow_item items[], + const struct rte_flow_action actions[], + struct rte_flow_error *error) { - struct rte_flow *flow; - size_t size; + struct rte_flow *flow = NULL; + struct mlx5_flow *dev_flow; + const struct rte_flow_action_rss *rss; + union { + struct rte_flow_expand_rss buf; + uint8_t buffer[2048]; + } expand_buffer; + struct rte_flow_expand_rss *buf = &expand_buffer.buf; int ret; + uint32_t i; + uint32_t flow_size; - ret = mlx5_flow_merge(dev, NULL, 0, attr, items, actions, error); + ret = flow_drv_validate(dev, attr, items, actions, error); if (ret < 0) return NULL; - size = ret; - flow = rte_zmalloc(__func__, size, 0); + flow_size = sizeof(struct rte_flow); + rss = flow_get_rss_action(actions); + if (rss) + flow_size += RTE_ALIGN_CEIL(rss->queue_num * sizeof(uint16_t), + sizeof(void *)); + else + flow_size += RTE_ALIGN_CEIL(sizeof(uint16_t), sizeof(void *)); + flow = rte_calloc(__func__, 1, flow_size, 0); if (!flow) { - rte_flow_error_set(error, ENOMEM, - RTE_FLOW_ERROR_TYPE_UNSPECIFIED, - NULL, - "cannot allocate memory"); + rte_errno = ENOMEM; return NULL; } - flow->verbs.attr = (struct ibv_flow_attr *)(flow + 1); - flow->verbs.specs = (uint8_t *)(flow->verbs.attr + 1); - ret = mlx5_flow_merge(dev, flow, size, attr, items, actions, error); - if (ret < 0) - goto error; - assert((size_t)ret == size); + flow->drv_type = flow_get_drv_type(dev, attr); + flow->ingress = attr->ingress; + flow->transfer = attr->transfer; + assert(flow->drv_type > MLX5_FLOW_TYPE_MIN && + flow->drv_type < MLX5_FLOW_TYPE_MAX); + flow->queue = (void *)(flow + 1); + LIST_INIT(&flow->dev_flows); + if (rss && rss->types) { + unsigned int graph_root; + + graph_root = find_graph_root(items, rss->level); + ret = rte_flow_expand_rss(buf, sizeof(expand_buffer.buffer), + items, rss->types, + mlx5_support_expansion, + graph_root); + assert(ret > 0 && + (unsigned int)ret < sizeof(expand_buffer.buffer)); + } else { + buf->entries = 1; + buf->entry[0].pattern = (void *)(uintptr_t)items; + } + for (i = 0; i < buf->entries; ++i) { + dev_flow = flow_drv_prepare(flow, attr, buf->entry[i].pattern, + actions, error); + if (!dev_flow) + goto error; + dev_flow->flow = flow; + LIST_INSERT_HEAD(&flow->dev_flows, dev_flow, next); + ret = flow_drv_translate(dev, dev_flow, attr, + buf->entry[i].pattern, + actions, error); + if (ret < 0) + goto error; + } if (dev->data->dev_started) { - ret = mlx5_flow_apply(dev, flow, error); + ret = flow_drv_apply(dev, flow, error); if (ret < 0) goto error; } TAILQ_INSERT_TAIL(list, flow, next); + flow_rxq_flags_set(dev, flow); return flow; error: ret = rte_errno; /* Save rte_errno before cleanup. */ - mlx5_flow_remove(dev, flow); + assert(flow); + flow_drv_destroy(dev, flow); rte_free(flow); rte_errno = ret; /* Restore rte_errno. */ return NULL; @@ -1273,9 +2422,10 @@ mlx5_flow_create(struct rte_eth_dev *dev, const struct rte_flow_action actions[], struct rte_flow_error *error) { - return mlx5_flow_list_create - (dev, &((struct priv *)dev->data->dev_private)->flows, - attr, items, actions, error); + struct mlx5_priv *priv = dev->data->dev_private; + + return flow_list_create(dev, &priv->flows, + attr, items, actions, error); } /** @@ -1289,11 +2439,18 @@ mlx5_flow_create(struct rte_eth_dev *dev, * Flow to destroy. */ static void -mlx5_flow_list_destroy(struct rte_eth_dev *dev, struct mlx5_flows *list, - struct rte_flow *flow) +flow_list_destroy(struct rte_eth_dev *dev, struct mlx5_flows *list, + struct rte_flow *flow) { - mlx5_flow_remove(dev, flow); + /* + * Update RX queue flags only if port is started, otherwise it is + * already clean. + */ + if (dev->data->dev_started) + flow_rxq_flags_trim(dev, flow); + flow_drv_destroy(dev, flow); TAILQ_REMOVE(list, flow, next); + rte_free(flow->fdir); rte_free(flow); } @@ -1312,7 +2469,7 @@ mlx5_flow_list_flush(struct rte_eth_dev *dev, struct mlx5_flows *list) struct rte_flow *flow; flow = TAILQ_FIRST(list); - mlx5_flow_list_destroy(dev, list, flow); + flow_list_destroy(dev, list, flow); } } @@ -1330,7 +2487,8 @@ mlx5_flow_stop(struct rte_eth_dev *dev, struct mlx5_flows *list) struct rte_flow *flow; TAILQ_FOREACH_REVERSE(flow, list, mlx5_flows, next) - mlx5_flow_remove(dev, flow); + flow_drv_remove(dev, flow); + flow_rxq_flags_clear(dev); } /** @@ -1352,9 +2510,10 @@ mlx5_flow_start(struct rte_eth_dev *dev, struct mlx5_flows *list) int ret = 0; TAILQ_FOREACH(flow, list, next) { - ret = mlx5_flow_apply(dev, flow, &error); + ret = flow_drv_apply(dev, flow, &error); if (ret < 0) goto error; + flow_rxq_flags_set(dev, flow); } return 0; error: @@ -1375,7 +2534,7 @@ error: int mlx5_flow_verify(struct rte_eth_dev *dev) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; struct rte_flow *flow; int ret = 0; @@ -1411,10 +2570,10 @@ mlx5_ctrl_flow_vlan(struct rte_eth_dev *dev, struct rte_flow_item_vlan *vlan_spec, struct rte_flow_item_vlan *vlan_mask) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; const struct rte_flow_attr attr = { .ingress = 1, - .priority = priv->config.flow_prio - 1, + .priority = MLX5_FLOW_PRIO_RSVD, }; struct rte_flow_item items[] = { { @@ -1425,7 +2584,7 @@ mlx5_ctrl_flow_vlan(struct rte_eth_dev *dev, }, { .type = (vlan_spec) ? RTE_FLOW_ITEM_TYPE_VLAN : - RTE_FLOW_ITEM_TYPE_END, + RTE_FLOW_ITEM_TYPE_END, .spec = vlan_spec, .last = NULL, .mask = vlan_mask, @@ -1457,14 +2616,13 @@ mlx5_ctrl_flow_vlan(struct rte_eth_dev *dev, struct rte_flow_error error; unsigned int i; - if (!priv->reta_idx_n) { - rte_errno = EINVAL; - return -rte_errno; + if (!priv->reta_idx_n || !priv->rxqs_n) { + return 0; } for (i = 0; i != priv->reta_idx_n; ++i) queue[i] = (*priv->reta_idx)[i]; - flow = mlx5_flow_list_create(dev, &priv->ctrl_flows, &attr, items, - actions, &error); + flow = flow_list_create(dev, &priv->ctrl_flows, + &attr, items, actions, &error); if (!flow) return -rte_errno; return 0; @@ -1502,9 +2660,9 @@ mlx5_flow_destroy(struct rte_eth_dev *dev, struct rte_flow *flow, struct rte_flow_error *error __rte_unused) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; - mlx5_flow_list_destroy(dev, &priv->flows, flow); + flow_list_destroy(dev, &priv->flows, flow); return 0; } @@ -1518,7 +2676,7 @@ int mlx5_flow_flush(struct rte_eth_dev *dev, struct rte_flow_error *error __rte_unused) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; mlx5_flow_list_flush(dev, &priv->flows); return 0; @@ -1535,7 +2693,7 @@ mlx5_flow_isolate(struct rte_eth_dev *dev, int enable, struct rte_flow_error *error) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; if (dev->data->dev_started) { rte_flow_error_set(error, EBUSY, @@ -1552,6 +2710,49 @@ mlx5_flow_isolate(struct rte_eth_dev *dev, return 0; } +/** + * Query a flow. + * + * @see rte_flow_query() + * @see rte_flow_ops + */ +static int +flow_drv_query(struct rte_eth_dev *dev, + struct rte_flow *flow, + const struct rte_flow_action *actions, + void *data, + struct rte_flow_error *error) +{ + const struct mlx5_flow_driver_ops *fops; + enum mlx5_flow_drv_type ftype = flow->drv_type; + + assert(ftype > MLX5_FLOW_TYPE_MIN && ftype < MLX5_FLOW_TYPE_MAX); + fops = flow_get_drv_ops(ftype); + + return fops->query(dev, flow, actions, data, error); +} + +/** + * Query a flow. + * + * @see rte_flow_query() + * @see rte_flow_ops + */ +int +mlx5_flow_query(struct rte_eth_dev *dev, + struct rte_flow *flow, + const struct rte_flow_action *actions, + void *data, + struct rte_flow_error *error) +{ + int ret; + + ret = flow_drv_query(dev, flow, actions, data, error); + if (ret < 0) + return ret; + return 0; +} + /** * Convert a flow director filter to a generic flow. * @@ -1566,11 +2767,11 @@ mlx5_flow_isolate(struct rte_eth_dev *dev, * 0 on success, a negative errno value otherwise and rte_errno is set. */ static int -mlx5_fdir_filter_convert(struct rte_eth_dev *dev, +flow_fdir_filter_convert(struct rte_eth_dev *dev, const struct rte_eth_fdir_filter *fdir_filter, struct mlx5_fdir *attributes) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; const struct rte_eth_fdir_input *input = &fdir_filter->input; const struct rte_eth_fdir_masks *mask = &dev->data->dev_conf.fdir_conf.mask; @@ -1613,14 +2814,13 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, case RTE_ETH_FLOW_NONFRAG_IPV4_UDP: case RTE_ETH_FLOW_NONFRAG_IPV4_TCP: case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER: - attributes->l3.ipv4.hdr = (struct ipv4_hdr){ + attributes->l3.ipv4.hdr = (struct rte_ipv4_hdr){ .src_addr = input->flow.ip4_flow.src_ip, .dst_addr = input->flow.ip4_flow.dst_ip, .time_to_live = input->flow.ip4_flow.ttl, .type_of_service = input->flow.ip4_flow.tos, - .next_proto_id = input->flow.ip4_flow.proto, }; - attributes->l3_mask.ipv4.hdr = (struct ipv4_hdr){ + attributes->l3_mask.ipv4.hdr = (struct rte_ipv4_hdr){ .src_addr = mask->ipv4_mask.src_ip, .dst_addr = mask->ipv4_mask.dst_ip, .time_to_live = mask->ipv4_mask.ttl, @@ -1636,7 +2836,7 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, case RTE_ETH_FLOW_NONFRAG_IPV6_UDP: case RTE_ETH_FLOW_NONFRAG_IPV6_TCP: case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER: - attributes->l3.ipv6.hdr = (struct ipv6_hdr){ + attributes->l3.ipv6.hdr = (struct rte_ipv6_hdr){ .hop_limits = input->flow.ipv6_flow.hop_limits, .proto = input->flow.ipv6_flow.proto, }; @@ -1668,11 +2868,11 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, /* Handle L4. */ switch (fdir_filter->input.flow_type) { case RTE_ETH_FLOW_NONFRAG_IPV4_UDP: - attributes->l4.udp.hdr = (struct udp_hdr){ + attributes->l4.udp.hdr = (struct rte_udp_hdr){ .src_port = input->flow.udp4_flow.src_port, .dst_port = input->flow.udp4_flow.dst_port, }; - attributes->l4_mask.udp.hdr = (struct udp_hdr){ + attributes->l4_mask.udp.hdr = (struct rte_udp_hdr){ .src_port = mask->src_port_mask, .dst_port = mask->dst_port_mask, }; @@ -1683,11 +2883,11 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, }; break; case RTE_ETH_FLOW_NONFRAG_IPV4_TCP: - attributes->l4.tcp.hdr = (struct tcp_hdr){ + attributes->l4.tcp.hdr = (struct rte_tcp_hdr){ .src_port = input->flow.tcp4_flow.src_port, .dst_port = input->flow.tcp4_flow.dst_port, }; - attributes->l4_mask.tcp.hdr = (struct tcp_hdr){ + attributes->l4_mask.tcp.hdr = (struct rte_tcp_hdr){ .src_port = mask->src_port_mask, .dst_port = mask->dst_port_mask, }; @@ -1698,11 +2898,11 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, }; break; case RTE_ETH_FLOW_NONFRAG_IPV6_UDP: - attributes->l4.udp.hdr = (struct udp_hdr){ + attributes->l4.udp.hdr = (struct rte_udp_hdr){ .src_port = input->flow.udp6_flow.src_port, .dst_port = input->flow.udp6_flow.dst_port, }; - attributes->l4_mask.udp.hdr = (struct udp_hdr){ + attributes->l4_mask.udp.hdr = (struct rte_udp_hdr){ .src_port = mask->src_port_mask, .dst_port = mask->dst_port_mask, }; @@ -1713,11 +2913,11 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, }; break; case RTE_ETH_FLOW_NONFRAG_IPV6_TCP: - attributes->l4.tcp.hdr = (struct tcp_hdr){ + attributes->l4.tcp.hdr = (struct rte_tcp_hdr){ .src_port = input->flow.tcp6_flow.src_port, .dst_port = input->flow.tcp6_flow.dst_port, }; - attributes->l4_mask.tcp.hdr = (struct tcp_hdr){ + attributes->l4_mask.tcp.hdr = (struct rte_tcp_hdr){ .src_port = mask->src_port_mask, .dst_port = mask->dst_port_mask, }; @@ -1739,6 +2939,69 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, return 0; } +#define FLOW_FDIR_CMP(f1, f2, fld) \ + memcmp(&(f1)->fld, &(f2)->fld, sizeof(f1->fld)) + +/** + * Compare two FDIR flows. If items and actions are identical, the two flows are + * regarded as same. + * + * @param dev + * Pointer to Ethernet device. + * @param f1 + * FDIR flow to compare. + * @param f2 + * FDIR flow to compare. + * + * @return + * Zero on match, 1 otherwise. + */ +static int +flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2) +{ + if (FLOW_FDIR_CMP(f1, f2, attr) || + FLOW_FDIR_CMP(f1, f2, l2) || + FLOW_FDIR_CMP(f1, f2, l2_mask) || + FLOW_FDIR_CMP(f1, f2, l3) || + FLOW_FDIR_CMP(f1, f2, l3_mask) || + FLOW_FDIR_CMP(f1, f2, l4) || + FLOW_FDIR_CMP(f1, f2, l4_mask) || + FLOW_FDIR_CMP(f1, f2, actions[0].type)) + return 1; + if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE && + FLOW_FDIR_CMP(f1, f2, queue)) + return 1; + return 0; +} + +/** + * Search device flow list to find out a matched FDIR flow. + * + * @param dev + * Pointer to Ethernet device. + * @param fdir_flow + * FDIR flow to lookup. + * + * @return + * Pointer of flow if found, NULL otherwise. + */ +static struct rte_flow * +flow_fdir_filter_lookup(struct rte_eth_dev *dev, struct mlx5_fdir *fdir_flow) +{ + struct mlx5_priv *priv = dev->data->dev_private; + struct rte_flow *flow = NULL; + + assert(fdir_flow); + TAILQ_FOREACH(flow, &priv->flows, next) { + if (flow->fdir && !flow_fdir_cmp(flow->fdir, fdir_flow)) { + DRV_LOG(DEBUG, "port %u found FDIR flow %p", + dev->data->port_id, (void *)flow); + break; + } + } + return flow; +} + /** * Add new flow director filter and store it in list. * @@ -1751,33 +3014,38 @@ mlx5_fdir_filter_convert(struct rte_eth_dev *dev, * 0 on success, a negative errno value otherwise and rte_errno is set. */ static int -mlx5_fdir_filter_add(struct rte_eth_dev *dev, +flow_fdir_filter_add(struct rte_eth_dev *dev, const struct rte_eth_fdir_filter *fdir_filter) { - struct priv *priv = dev->data->dev_private; - 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 rte_flow_error error; + struct mlx5_priv *priv = dev->data->dev_private; + struct mlx5_fdir *fdir_flow; struct rte_flow *flow; int ret; - ret = mlx5_fdir_filter_convert(dev, fdir_filter, &attributes); + fdir_flow = rte_zmalloc(__func__, sizeof(*fdir_flow), 0); + if (!fdir_flow) { + rte_errno = ENOMEM; + return -rte_errno; + } + ret = flow_fdir_filter_convert(dev, fdir_filter, fdir_flow); if (ret) - return ret; - flow = mlx5_flow_list_create(dev, &priv->flows, &attributes.attr, - attributes.items, attributes.actions, - &error); + goto error; + flow = flow_fdir_filter_lookup(dev, fdir_flow); if (flow) { - DRV_LOG(DEBUG, "port %u FDIR created %p", dev->data->port_id, - (void *)flow); - return 0; + rte_errno = EEXIST; + goto error; } + flow = flow_list_create(dev, &priv->flows, &fdir_flow->attr, + fdir_flow->items, fdir_flow->actions, NULL); + if (!flow) + goto error; + assert(!flow->fdir); + flow->fdir = fdir_flow; + DRV_LOG(DEBUG, "port %u created FDIR flow %p", + dev->data->port_id, (void *)flow); + return 0; +error: + rte_free(fdir_flow); return -rte_errno; } @@ -1793,12 +3061,28 @@ mlx5_fdir_filter_add(struct rte_eth_dev *dev, * 0 on success, a negative errno value otherwise and rte_errno is set. */ static int -mlx5_fdir_filter_delete(struct rte_eth_dev *dev __rte_unused, - const struct rte_eth_fdir_filter *fdir_filter - __rte_unused) +flow_fdir_filter_delete(struct rte_eth_dev *dev, + const struct rte_eth_fdir_filter *fdir_filter) { - rte_errno = ENOTSUP; - return -rte_errno; + struct mlx5_priv *priv = dev->data->dev_private; + struct rte_flow *flow; + struct mlx5_fdir fdir_flow = { + .attr.group = 0, + }; + int ret; + + ret = flow_fdir_filter_convert(dev, fdir_filter, &fdir_flow); + if (ret) + return -rte_errno; + flow = flow_fdir_filter_lookup(dev, &fdir_flow); + if (!flow) { + rte_errno = ENOENT; + return -rte_errno; + } + flow_list_destroy(dev, &priv->flows, flow); + DRV_LOG(DEBUG, "port %u deleted FDIR flow %p", + dev->data->port_id, (void *)flow); + return 0; } /** @@ -1813,15 +3097,15 @@ mlx5_fdir_filter_delete(struct rte_eth_dev *dev __rte_unused, * 0 on success, a negative errno value otherwise and rte_errno is set. */ static int -mlx5_fdir_filter_update(struct rte_eth_dev *dev, +flow_fdir_filter_update(struct rte_eth_dev *dev, const struct rte_eth_fdir_filter *fdir_filter) { int ret; - ret = mlx5_fdir_filter_delete(dev, fdir_filter); + ret = flow_fdir_filter_delete(dev, fdir_filter); if (ret) return ret; - return mlx5_fdir_filter_add(dev, fdir_filter); + return flow_fdir_filter_add(dev, fdir_filter); } /** @@ -1831,9 +3115,9 @@ mlx5_fdir_filter_update(struct rte_eth_dev *dev, * Pointer to Ethernet device. */ static void -mlx5_fdir_filter_flush(struct rte_eth_dev *dev) +flow_fdir_filter_flush(struct rte_eth_dev *dev) { - struct priv *priv = dev->data->dev_private; + struct mlx5_priv *priv = dev->data->dev_private; mlx5_flow_list_flush(dev, &priv->flows); } @@ -1847,7 +3131,7 @@ mlx5_fdir_filter_flush(struct rte_eth_dev *dev) * Resulting flow director information. */ static void -mlx5_fdir_info_get(struct rte_eth_dev *dev, struct rte_eth_fdir_info *fdir_info) +flow_fdir_info_get(struct rte_eth_dev *dev, struct rte_eth_fdir_info *fdir_info) { struct rte_eth_fdir_masks *mask = &dev->data->dev_conf.fdir_conf.mask; @@ -1877,7 +3161,7 @@ mlx5_fdir_info_get(struct rte_eth_dev *dev, struct rte_eth_fdir_info *fdir_info) * 0 on success, a negative errno value otherwise and rte_errno is set. */ static int -mlx5_fdir_ctrl_func(struct rte_eth_dev *dev, enum rte_filter_op filter_op, +flow_fdir_ctrl_func(struct rte_eth_dev *dev, enum rte_filter_op filter_op, void *arg) { enum rte_fdir_mode fdir_mode = @@ -1894,16 +3178,16 @@ mlx5_fdir_ctrl_func(struct rte_eth_dev *dev, enum rte_filter_op filter_op, } switch (filter_op) { case RTE_ETH_FILTER_ADD: - return mlx5_fdir_filter_add(dev, arg); + return flow_fdir_filter_add(dev, arg); case RTE_ETH_FILTER_UPDATE: - return mlx5_fdir_filter_update(dev, arg); + return flow_fdir_filter_update(dev, arg); case RTE_ETH_FILTER_DELETE: - return mlx5_fdir_filter_delete(dev, arg); + return flow_fdir_filter_delete(dev, arg); case RTE_ETH_FILTER_FLUSH: - mlx5_fdir_filter_flush(dev); + flow_fdir_filter_flush(dev); break; case RTE_ETH_FILTER_INFO: - mlx5_fdir_info_get(dev, arg); + flow_fdir_info_get(dev, arg); break; default: DRV_LOG(DEBUG, "port %u unknown operation %u", @@ -1944,7 +3228,7 @@ mlx5_dev_filter_ctrl(struct rte_eth_dev *dev, *(const void **)arg = &mlx5_flow_ops; return 0; case RTE_ETH_FILTER_FDIR: - return mlx5_fdir_ctrl_func(dev, filter_op, arg); + return flow_fdir_ctrl_func(dev, filter_op, arg); default: DRV_LOG(ERR, "port %u filter type (%d) not supported", dev->data->port_id, filter_type); @@ -1953,3 +3237,150 @@ mlx5_dev_filter_ctrl(struct rte_eth_dev *dev, } return 0; } + +#define MLX5_POOL_QUERY_FREQ_US 1000000 + +/** + * Set the periodic procedure for triggering asynchronous batch queries for all + * the counter pools. + * + * @param[in] sh + * Pointer to mlx5_ibv_shared object. + */ +void +mlx5_set_query_alarm(struct mlx5_ibv_shared *sh) +{ + struct mlx5_pools_container *cont = MLX5_CNT_CONTAINER(sh, 0, 0); + uint32_t pools_n = rte_atomic16_read(&cont->n_valid); + uint32_t us; + + cont = MLX5_CNT_CONTAINER(sh, 1, 0); + pools_n += rte_atomic16_read(&cont->n_valid); + us = MLX5_POOL_QUERY_FREQ_US / pools_n; + DRV_LOG(DEBUG, "Set alarm for %u pools each %u us\n", pools_n, us); + if (rte_eal_alarm_set(us, mlx5_flow_query_alarm, sh)) { + sh->cmng.query_thread_on = 0; + DRV_LOG(ERR, "Cannot reinitialize query alarm\n"); + } else { + sh->cmng.query_thread_on = 1; + } +} + +/** + * The periodic procedure for triggering asynchronous batch queries for all the + * counter pools. This function is probably called by the host thread. + * + * @param[in] arg + * The parameter for the alarm process. + */ +void +mlx5_flow_query_alarm(void *arg) +{ + struct mlx5_ibv_shared *sh = arg; + struct mlx5_devx_obj *dcs; + uint16_t offset; + int ret; + uint8_t batch = sh->cmng.batch; + uint16_t pool_index = sh->cmng.pool_index; + struct mlx5_pools_container *cont; + struct mlx5_pools_container *mcont; + struct mlx5_flow_counter_pool *pool; + + if (sh->cmng.pending_queries >= MLX5_MAX_PENDING_QUERIES) + goto set_alarm; +next_container: + cont = MLX5_CNT_CONTAINER(sh, batch, 1); + mcont = MLX5_CNT_CONTAINER(sh, batch, 0); + /* Check if resize was done and need to flip a container. */ + if (cont != mcont) { + if (cont->pools) { + /* Clean the old container. */ + rte_free(cont->pools); + memset(cont, 0, sizeof(*cont)); + } + rte_cio_wmb(); + /* Flip the host container. */ + sh->cmng.mhi[batch] ^= (uint8_t)2; + cont = mcont; + } + if (!cont->pools) { + /* 2 empty containers case is unexpected. */ + if (unlikely(batch != sh->cmng.batch)) + goto set_alarm; + batch ^= 0x1; + pool_index = 0; + goto next_container; + } + pool = cont->pools[pool_index]; + if (pool->raw_hw) + /* There is a pool query in progress. */ + goto set_alarm; + pool->raw_hw = + LIST_FIRST(&sh->cmng.free_stat_raws); + if (!pool->raw_hw) + /* No free counter statistics raw memory. */ + goto set_alarm; + dcs = (struct mlx5_devx_obj *)(uintptr_t)rte_atomic64_read + (&pool->a64_dcs); + offset = batch ? 0 : dcs->id % MLX5_COUNTERS_PER_POOL; + ret = mlx5_devx_cmd_flow_counter_query(dcs, 0, MLX5_COUNTERS_PER_POOL - + offset, NULL, NULL, + pool->raw_hw->mem_mng->dm->id, + (void *)(uintptr_t) + (pool->raw_hw->data + offset), + sh->devx_comp, + (uint64_t)(uintptr_t)pool); + if (ret) { + DRV_LOG(ERR, "Failed to trigger asynchronous query for dcs ID" + " %d\n", pool->min_dcs->id); + pool->raw_hw = NULL; + goto set_alarm; + } + pool->raw_hw->min_dcs_id = dcs->id; + LIST_REMOVE(pool->raw_hw, next); + sh->cmng.pending_queries++; + pool_index++; + if (pool_index >= rte_atomic16_read(&cont->n_valid)) { + batch ^= 0x1; + pool_index = 0; + } +set_alarm: + sh->cmng.batch = batch; + sh->cmng.pool_index = pool_index; + mlx5_set_query_alarm(sh); +} + +/** + * Handler for the HW respond about ready values from an asynchronous batch + * query. This function is probably called by the host thread. + * + * @param[in] sh + * The pointer to the shared IB device context. + * @param[in] async_id + * The Devx async ID. + * @param[in] status + * The status of the completion. + */ +void +mlx5_flow_async_pool_query_handle(struct mlx5_ibv_shared *sh, + uint64_t async_id, int status) +{ + struct mlx5_flow_counter_pool *pool = + (struct mlx5_flow_counter_pool *)(uintptr_t)async_id; + struct mlx5_counter_stats_raw *raw_to_free; + + if (unlikely(status)) { + raw_to_free = pool->raw_hw; + } else { + raw_to_free = pool->raw; + rte_spinlock_lock(&pool->sl); + pool->raw = pool->raw_hw; + rte_spinlock_unlock(&pool->sl); + rte_atomic64_add(&pool->query_gen, 1); + /* Be sure the new raw counters data is updated in memory. */ + rte_cio_wmb(); + } + LIST_INSERT_HEAD(&sh->cmng.free_stat_raws, raw_to_free, next); + pool->raw_hw = NULL; + sh->cmng.pending_queries--; +}