net/mlx5: fix RSS pattern expansion

Flow rule pattern may be implicitly expanded by the PMD if the rule
has RSS flow action. The expansion adds network headers to the
original pattern. The new pattern lists all network levels that
participate in the rule RSS action.

The patch fixes expanded pattern for cases when original pattern
included meta items like MARK, TAG, META.

Fixes: c7870bfe09dc ("ethdev: move RSS expansion code to mlx5 driver")
Cc: stable@dpdk.org
Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index dbeca57..e5e062d 100644 (file)
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -114,6 +114,30 @@ struct mlx5_flow_expand_rss {
        } entry[];
 };
 
+static void
+mlx5_dbg__print_pattern(const struct rte_flow_item *item);
+
+static bool
+mlx5_flow_is_rss_expandable_item(const struct rte_flow_item *item)
+{
+       switch (item->type) {
+       case RTE_FLOW_ITEM_TYPE_ETH:
+       case RTE_FLOW_ITEM_TYPE_VLAN:
+       case RTE_FLOW_ITEM_TYPE_IPV4:
+       case RTE_FLOW_ITEM_TYPE_IPV6:
+       case RTE_FLOW_ITEM_TYPE_UDP:
+       case RTE_FLOW_ITEM_TYPE_TCP:
+       case RTE_FLOW_ITEM_TYPE_VXLAN:
+       case RTE_FLOW_ITEM_TYPE_NVGRE:
+       case RTE_FLOW_ITEM_TYPE_GRE:
+       case RTE_FLOW_ITEM_TYPE_GENEVE:
+               return true;
+       default:
+               break;
+       }
+       return false;
+}
+
 static enum rte_flow_item_type
 mlx5_flow_expand_rss_item_complete(const struct rte_flow_item *item)
 {
@@ -273,8 +297,11 @@ mlx5_flow_expand_rss(struct mlx5_flow_expand_rss *buf, size_t size,
                addr = buf->entry[0].pattern;
        }
        for (item = pattern; item->type != RTE_FLOW_ITEM_TYPE_END; item++) {
-               if (item->type != RTE_FLOW_ITEM_TYPE_VOID)
-                       last_item = item;
+               if (!mlx5_flow_is_rss_expandable_item(item)) {
+                       user_pattern_size += sizeof(*item);
+                       continue;
+               }
+               last_item = item;
                for (i = 0; node->next && node->next[i]; ++i) {
                        next = &graph[node->next[i]];
                        if (next->type == item->type)
@@ -6174,6 +6201,10 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
                                           mlx5_support_expansion, graph_root);
                MLX5_ASSERT(ret > 0 &&
                       (unsigned int)ret < sizeof(expand_buffer.buffer));
+               if (rte_log_can_log(mlx5_logtype, RTE_LOG_DEBUG)) {
+                       for (i = 0; i < buf->entries; ++i)
+                               mlx5_dbg__print_pattern(buf->entry[i].pattern);
+               }
        } else {
                buf->entries = 1;
                buf->entry[0].pattern = (void *)(uintptr_t)items;
@@ -8990,3 +9021,22 @@ mlx5_release_tunnel_hub(__rte_unused struct mlx5_dev_ctx_shared *sh,
 {
 }
 #endif /* HAVE_IBV_FLOW_DV_SUPPORT */
+
+static void
+mlx5_dbg__print_pattern(const struct rte_flow_item *item)
+{
+       int ret;
+       struct rte_flow_error error;
+
+       for (; item->type != RTE_FLOW_ITEM_TYPE_END; item++) {
+               char *item_name;
+               ret = rte_flow_conv(RTE_FLOW_CONV_OP_ITEM_NAME_PTR, &item_name,
+                                   sizeof(item_name),
+                                   (void *)(uintptr_t)item->type, &error);
+               if (ret > 0)
+                       printf("%s ", item_name);
+               else
+                       printf("%d\n", (int)item->type);
+       }
+       printf("END\n");
+}