net/mlx5: add VXLAN to flow prepare routine
authorViacheslav Ovsiienko <viacheslavo@mellanox.com>
Sat, 3 Nov 2018 06:18:40 +0000 (06:18 +0000)
committerFerruh Yigit <ferruh.yigit@intel.com>
Mon, 5 Nov 2018 14:01:25 +0000 (15:01 +0100)
The e-switch Flow prepare function is updated to support VXLAN
encapsulation/and decapsulation actions. The function calculates
buffer size for Netlink message and Flow description structures,
including optional ones for tunneling purposes.

Suggested-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
drivers/net/mlx5/mlx5_flow_tcf.c

index 6299069..017f2bd 100644 (file)
@@ -2422,7 +2422,7 @@ flow_tcf_get_items_and_size(const struct rte_flow_attr *attr,
                case RTE_FLOW_ITEM_TYPE_IPV6:
                        size += SZ_NLATTR_TYPE_OF(uint16_t) + /* Ether type. */
                                SZ_NLATTR_TYPE_OF(uint8_t) + /* IP proto. */
-                               SZ_NLATTR_TYPE_OF(IPV6_ADDR_LEN) * 4;
+                               SZ_NLATTR_DATA_OF(IPV6_ADDR_LEN) * 4;
                                /* dst/src IP addr and mask. */
                        flags |= MLX5_FLOW_LAYER_OUTER_L3_IPV6;
                        break;
@@ -2438,6 +2438,10 @@ flow_tcf_get_items_and_size(const struct rte_flow_attr *attr,
                                /* dst/src port and mask. */
                        flags |= MLX5_FLOW_LAYER_OUTER_L4_TCP;
                        break;
+               case RTE_FLOW_ITEM_TYPE_VXLAN:
+                       size += SZ_NLATTR_TYPE_OF(uint32_t);
+                       flags |= MLX5_FLOW_LAYER_VXLAN;
+                       break;
                default:
                        DRV_LOG(WARNING,
                                "unsupported item %p type %d,"
@@ -2450,6 +2454,69 @@ flow_tcf_get_items_and_size(const struct rte_flow_attr *attr,
        return size;
 }
 
+/**
+ * Calculate size of memory to store the VXLAN encapsultion
+ * related items in the Netlink message buffer. Items list
+ * is specified by RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP action.
+ * The item list should be validated.
+ *
+ * @param[in] action
+ *   RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP action object.
+ *   List of pattern items to scan data from.
+ *
+ * @return
+ *   The size the part of Netlink message buffer to store the
+ *   VXLAN encapsulation item attributes.
+ */
+static int
+flow_tcf_vxlan_encap_size(const struct rte_flow_action *action)
+{
+       const struct rte_flow_item *items;
+       int size = 0;
+
+       assert(action->type == RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP);
+       assert(action->conf);
+
+       items = ((const struct rte_flow_action_vxlan_encap *)
+                                       action->conf)->definition;
+       assert(items);
+       for (; items->type != RTE_FLOW_ITEM_TYPE_END; items++) {
+               switch (items->type) {
+               case RTE_FLOW_ITEM_TYPE_VOID:
+                       break;
+               case RTE_FLOW_ITEM_TYPE_ETH:
+                       /* This item does not require message buffer. */
+                       break;
+               case RTE_FLOW_ITEM_TYPE_IPV4:
+                       size += SZ_NLATTR_DATA_OF(IPV4_ADDR_LEN) * 2;
+                       break;
+               case RTE_FLOW_ITEM_TYPE_IPV6:
+                       size += SZ_NLATTR_DATA_OF(IPV6_ADDR_LEN) * 2;
+                       break;
+               case RTE_FLOW_ITEM_TYPE_UDP: {
+                       const struct rte_flow_item_udp *udp = items->mask;
+
+                       size += SZ_NLATTR_TYPE_OF(uint16_t);
+                       if (!udp || udp->hdr.src_port != RTE_BE16(0x0000))
+                               size += SZ_NLATTR_TYPE_OF(uint16_t);
+                       break;
+               }
+               case RTE_FLOW_ITEM_TYPE_VXLAN:
+                       size += SZ_NLATTR_TYPE_OF(uint32_t);
+                       break;
+               default:
+                       assert(false);
+                       DRV_LOG(WARNING,
+                               "unsupported item %p type %d,"
+                               " items must be validated"
+                               " before flow creation",
+                               (const void *)items, items->type);
+                       return 0;
+               }
+       }
+       return size;
+}
+
 /**
  * Calculate maximum size of memory for flow actions of Linux TC flower and
  * extract specified actions.
@@ -2519,6 +2586,29 @@ action_of_vlan:
                                SZ_NLATTR_TYPE_OF(uint16_t) + /* VLAN ID. */
                                SZ_NLATTR_TYPE_OF(uint8_t); /* VLAN prio. */
                        break;
+               case RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP:
+                       size += SZ_NLATTR_NEST + /* na_act_index. */
+                               SZ_NLATTR_STRZ_OF("tunnel_key") +
+                               SZ_NLATTR_NEST + /* TCA_ACT_OPTIONS. */
+                               SZ_NLATTR_TYPE_OF(uint8_t);
+                       size += SZ_NLATTR_TYPE_OF(struct tc_tunnel_key);
+                       size += flow_tcf_vxlan_encap_size(actions) +
+                               RTE_ALIGN_CEIL /* preceding encap params. */
+                               (sizeof(struct flow_tcf_vxlan_encap),
+                               MNL_ALIGNTO);
+                       flags |= MLX5_FLOW_ACTION_VXLAN_ENCAP;
+                       break;
+               case RTE_FLOW_ACTION_TYPE_VXLAN_DECAP:
+                       size += SZ_NLATTR_NEST + /* na_act_index. */
+                               SZ_NLATTR_STRZ_OF("tunnel_key") +
+                               SZ_NLATTR_NEST + /* TCA_ACT_OPTIONS. */
+                               SZ_NLATTR_TYPE_OF(uint8_t);
+                       size += SZ_NLATTR_TYPE_OF(struct tc_tunnel_key);
+                       size += RTE_ALIGN_CEIL /* preceding decap params. */
+                               (sizeof(struct flow_tcf_vxlan_decap),
+                               MNL_ALIGNTO);
+                       flags |= MLX5_FLOW_ACTION_VXLAN_DECAP;
+                       break;
                case RTE_FLOW_ACTION_TYPE_SET_IPV4_SRC:
                case RTE_FLOW_ACTION_TYPE_SET_IPV4_DST:
                case RTE_FLOW_ACTION_TYPE_SET_IPV6_SRC:
@@ -2594,12 +2684,15 @@ flow_tcf_prepare(const struct rte_flow_attr *attr,
                 uint64_t *item_flags, uint64_t *action_flags,
                 struct rte_flow_error *error)
 {
-       size_t size = sizeof(struct mlx5_flow) +
+       size_t size = RTE_ALIGN_CEIL
+                       (sizeof(struct mlx5_flow),
+                        alignof(struct flow_tcf_tunnel_hdr)) +
                      MNL_ALIGN(sizeof(struct nlmsghdr)) +
                      MNL_ALIGN(sizeof(struct tcmsg));
        struct mlx5_flow *dev_flow;
        struct nlmsghdr *nlh;
        struct tcmsg *tcm;
+       uint8_t *sp, *tun = NULL;
 
        size += flow_tcf_get_items_and_size(attr, items, item_flags);
        size += flow_tcf_get_actions_and_size(actions, action_flags);
@@ -2610,14 +2703,52 @@ flow_tcf_prepare(const struct rte_flow_attr *attr,
                                   "not enough memory to create E-Switch flow");
                return NULL;
        }
-       nlh = mnl_nlmsg_put_header((void *)(dev_flow + 1));
+       sp = (uint8_t *)(dev_flow + 1);
+       if (*action_flags & MLX5_FLOW_ACTION_VXLAN_ENCAP) {
+               sp = RTE_PTR_ALIGN
+                       (sp, alignof(struct flow_tcf_tunnel_hdr));
+               tun = sp;
+               sp += RTE_ALIGN_CEIL
+                       (sizeof(struct flow_tcf_vxlan_encap),
+                       MNL_ALIGNTO);
+#ifndef NDEBUG
+               size -= RTE_ALIGN_CEIL
+                       (sizeof(struct flow_tcf_vxlan_encap),
+                       MNL_ALIGNTO);
+#endif
+       } else if (*action_flags & MLX5_FLOW_ACTION_VXLAN_DECAP) {
+               sp = RTE_PTR_ALIGN
+                       (sp, alignof(struct flow_tcf_tunnel_hdr));
+               tun = sp;
+               sp += RTE_ALIGN_CEIL
+                       (sizeof(struct flow_tcf_vxlan_decap),
+                       MNL_ALIGNTO);
+#ifndef NDEBUG
+               size -= RTE_ALIGN_CEIL
+                       (sizeof(struct flow_tcf_vxlan_decap),
+                       MNL_ALIGNTO);
+#endif
+       } else {
+               sp = RTE_PTR_ALIGN(sp, MNL_ALIGNTO);
+       }
+       nlh = mnl_nlmsg_put_header(sp);
        tcm = mnl_nlmsg_put_extra_header(nlh, sizeof(*tcm));
        *dev_flow = (struct mlx5_flow){
                .tcf = (struct mlx5_flow_tcf){
+#ifndef NDEBUG
+                       .nlsize = size - RTE_ALIGN_CEIL
+                               (sizeof(struct mlx5_flow),
+                                alignof(struct flow_tcf_tunnel_hdr)),
+#endif
+                       .tunnel = (struct flow_tcf_tunnel_hdr *)tun,
                        .nlh = nlh,
                        .tcm = tcm,
                },
        };
+       if (*action_flags & MLX5_FLOW_ACTION_VXLAN_DECAP)
+               dev_flow->tcf.tunnel->type = FLOW_TCF_TUNACT_VXLAN_DECAP;
+       else if (*action_flags & MLX5_FLOW_ACTION_VXLAN_ENCAP)
+               dev_flow->tcf.tunnel->type = FLOW_TCF_TUNACT_VXLAN_ENCAP;
        /*
         * Generate a reasonably unique handle based on the address of the
         * target buffer.