net/mlx5: add flow VLAN item
authorNelio Laranjeiro <nelio.laranjeiro@6wind.com>
Thu, 12 Jul 2018 09:30:53 +0000 (11:30 +0200)
committerShahaf Shuler <shahafs@mellanox.com>
Thu, 12 Jul 2018 10:10:02 +0000 (12:10 +0200)
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
drivers/net/mlx5/mlx5_flow.c

index 77f1bd5..6599792 100644 (file)
@@ -382,6 +382,130 @@ mlx5_flow_item_eth(const struct rte_flow_item *item, struct rte_flow *flow,
        return size;
 }
 
+/**
+ * Update the VLAN tag in the Verbs Ethernet specification.
+ *
+ * @param[in, out] attr
+ *   Pointer to Verbs attributes structure.
+ * @param[in] eth
+ *   Verbs structure containing the VLAN information to copy.
+ */
+static void
+mlx5_flow_item_vlan_update(struct ibv_flow_attr *attr,
+                          struct ibv_flow_spec_eth *eth)
+{
+       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);
+       }
+}
+
+/**
+ * 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.
+ *
+ * @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[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.
+ */
+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)
+{
+       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)
+               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, &eth, size);
+       } else {
+               if (flow->verbs.attr)
+                       mlx5_flow_item_vlan_update(flow->verbs.attr, &eth);
+               size = 0; /* Only an update is done in eth specification. */
+       }
+       flow->layers |= MLX5_FLOW_LAYER_OUTER_L2 |
+               MLX5_FLOW_LAYER_OUTER_VLAN;
+       return size;
+}
+
 /**
  * Convert the @p pattern into a Verbs specifications after ensuring the NIC
  * will understand and process it correctly.
@@ -424,6 +548,9 @@ mlx5_flow_items(const struct rte_flow_item pattern[],
                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;
                default:
                        return rte_flow_error_set(error, ENOTSUP,
                                                  RTE_FLOW_ERROR_TYPE_ITEM,