net/mlx5: reorganize flow structures
authorBing Zhao <bingz@mellanox.com>
Tue, 24 Mar 2020 15:33:58 +0000 (15:33 +0000)
committerFerruh Yigit <ferruh.yigit@intel.com>
Tue, 21 Apr 2020 11:57:05 +0000 (13:57 +0200)
Common structures used for mlx5 flow creating and destroying are
reorganized in order to separating the parts only for destroying
from all the items.
The "mlx5_flow" will contain the common items of DV and Verbs flow,
specific items for DV / Verbs only. These items will only be used
when creating a flow.
At the end of "mlx5_flow", a nested structure "mlx5_flow_handle"
located. It contains all the items used both for creating and
destroying a flow. Also, it consists of common items, and DV / Verbs
specific items.

Signed-off-by: Bing Zhao <bingz@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
drivers/net/mlx5/mlx5_flow.c
drivers/net/mlx5/mlx5_flow.h
drivers/net/mlx5/mlx5_flow_dv.c
drivers/net/mlx5/mlx5_flow_verbs.c

index aad9689..b2de4e6 100644 (file)
@@ -720,9 +720,9 @@ flow_drv_rxq_flags_set(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow)
 {
        struct mlx5_priv *priv = dev->data->dev_private;
        struct rte_flow *flow = dev_flow->flow;
-       const int mark = !!(dev_flow->actions &
+       const int mark = !!(dev_flow->handle.act_flags &
                            (MLX5_FLOW_ACTION_FLAG | MLX5_FLOW_ACTION_MARK));
-       const int tunnel = !!(dev_flow->layers & MLX5_FLOW_LAYER_TUNNEL);
+       const int tunnel = !!(dev_flow->handle.layers & MLX5_FLOW_LAYER_TUNNEL);
        unsigned int i;
 
        for (i = 0; i != flow->rss.queue_num; ++i) {
@@ -751,7 +751,7 @@ flow_drv_rxq_flags_set(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow)
                        /* Increase the counter matching the flow. */
                        for (j = 0; j != MLX5_FLOW_TUNNEL; ++j) {
                                if ((tunnels_info[j].tunnel &
-                                    dev_flow->layers) ==
+                                    dev_flow->handle.layers) ==
                                    tunnels_info[j].tunnel) {
                                        rxq_ctrl->flow_tunnels_n[j]++;
                                        break;
@@ -793,9 +793,9 @@ flow_drv_rxq_flags_trim(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow)
 {
        struct mlx5_priv *priv = dev->data->dev_private;
        struct rte_flow *flow = dev_flow->flow;
-       const int mark = !!(dev_flow->actions &
+       const int mark = !!(dev_flow->handle.act_flags &
                            (MLX5_FLOW_ACTION_FLAG | MLX5_FLOW_ACTION_MARK));
-       const int tunnel = !!(dev_flow->layers & MLX5_FLOW_LAYER_TUNNEL);
+       const int tunnel = !!(dev_flow->handle.layers & MLX5_FLOW_LAYER_TUNNEL);
        unsigned int i;
 
        MLX5_ASSERT(dev->data->dev_started);
@@ -820,7 +820,7 @@ flow_drv_rxq_flags_trim(struct rte_eth_dev *dev, struct mlx5_flow *dev_flow)
                        /* Decrease the counter matching the flow. */
                        for (j = 0; j != MLX5_FLOW_TUNNEL; ++j) {
                                if ((tunnels_info[j].tunnel &
-                                    dev_flow->layers) ==
+                                    dev_flow->handle.layers) ==
                                    tunnels_info[j].tunnel) {
                                        rxq_ctrl->flow_tunnels_n[j]--;
                                        break;
@@ -2312,8 +2312,8 @@ flow_mreg_split_qrss_release(struct rte_eth_dev *dev,
        struct mlx5_flow *dev_flow;
 
        LIST_FOREACH(dev_flow, &flow->dev_flows, next)
-               if (dev_flow->qrss_id)
-                       flow_qrss_free_id(dev, dev_flow->qrss_id);
+               if (dev_flow->handle.qrss_id)
+                       flow_qrss_free_id(dev, dev_flow->handle.qrss_id);
 }
 
 static int
@@ -2696,18 +2696,22 @@ flow_get_prefix_layer_flags(struct mlx5_flow *dev_flow)
 {
        uint64_t layers = 0;
 
-       /* If no decap actions, use the layers directly. */
-       if (!(dev_flow->actions & MLX5_FLOW_ACTION_DECAP))
-               return dev_flow->layers;
+       /*
+        * Layers bits could be localization, but usually the compiler will
+        * help to do the optimization work for source code.
+        * If no decap actions, use the layers directly.
+        */
+       if (!(dev_flow->handle.act_flags & MLX5_FLOW_ACTION_DECAP))
+               return dev_flow->handle.layers;
        /* Convert L3 layers with decap action. */
-       if (dev_flow->layers & MLX5_FLOW_LAYER_INNER_L3_IPV4)
+       if (dev_flow->handle.layers & MLX5_FLOW_LAYER_INNER_L3_IPV4)
                layers |= MLX5_FLOW_LAYER_OUTER_L3_IPV4;
-       else if (dev_flow->layers & MLX5_FLOW_LAYER_INNER_L3_IPV6)
+       else if (dev_flow->handle.layers & MLX5_FLOW_LAYER_INNER_L3_IPV6)
                layers |= MLX5_FLOW_LAYER_OUTER_L3_IPV6;
        /* Convert L4 layers with decap action.  */
-       if (dev_flow->layers & MLX5_FLOW_LAYER_INNER_L4_TCP)
+       if (dev_flow->handle.layers & MLX5_FLOW_LAYER_INNER_L4_TCP)
                layers |= MLX5_FLOW_LAYER_OUTER_L4_TCP;
-       else if (dev_flow->layers & MLX5_FLOW_LAYER_INNER_L4_UDP)
+       else if (dev_flow->handle.layers & MLX5_FLOW_LAYER_INNER_L4_UDP)
                layers |= MLX5_FLOW_LAYER_OUTER_L4_UDP;
        return layers;
 }
@@ -3453,7 +3457,7 @@ flow_create_split_inner(struct rte_eth_dev *dev,
         * flow may need some user defined item layer flags.
         */
        if (prefix_layers)
-               dev_flow->layers = prefix_layers;
+               dev_flow->handle.layers = prefix_layers;
        if (sub_flow)
                *sub_flow = dev_flow;
        return flow_drv_translate(dev, dev_flow, attr, items, actions, error);
@@ -3968,8 +3972,7 @@ flow_create_split_metadata(struct rte_eth_dev *dev,
                         * reallocation becomes possible (for example, for
                         * other flows in other threads).
                         */
-                       dev_flow->qrss_id = qrss_id;
-                       qrss_id = 0;
+                       dev_flow->handle.qrss_id = qrss_id;
                        ret = mlx5_flow_get_reg_id(dev, MLX5_COPY_MARK, 0,
                                                   error);
                        if (ret < 0)
@@ -3984,6 +3987,8 @@ flow_create_split_metadata(struct rte_eth_dev *dev,
                                              external, error);
                if (ret < 0)
                        goto exit;
+               /* qrss ID should be freed if failed. */
+               qrss_id = 0;
                MLX5_ASSERT(dev_flow);
        }
 
@@ -4080,7 +4085,7 @@ flow_create_split_meter(struct rte_eth_dev *dev,
                        ret = -rte_errno;
                        goto exit;
                }
-               dev_flow->mtr_flow_id = mtr_tag_id;
+               dev_flow->handle.mtr_flow_id = mtr_tag_id;
                /* Setting the sfx group atrr. */
                sfx_attr.group = sfx_attr.transfer ?
                                (MLX5_FLOW_TABLE_LEVEL_SUFFIX - 1) :
index 13c8589..f3aea53 100644 (file)
@@ -464,25 +464,28 @@ struct mlx5_flow_tbl_data_entry {
        /**< jump resource, at most one for each table created. */
 };
 
-/*
- * Max number of actions per DV flow.
- * See CREATE_FLOW_MAX_FLOW_ACTIONS_SUPPORTED
- * In rdma-core file providers/mlx5/verbs.c
- */
-#define MLX5_DV_MAX_NUMBER_OF_ACTIONS 8
+/* Verbs specification header. */
+struct ibv_spec_header {
+       enum ibv_flow_spec_type type;
+       uint16_t size;
+};
+
+struct mlx5_flow_rss {
+       uint32_t level;
+       uint32_t queue_num; /**< Number of entries in @p queue. */
+       uint64_t types; /**< Specific RSS hash types (see ETH_RSS_*). */
+       uint16_t (*queue)[]; /**< Destination queues to redirect traffic to. */
+       uint8_t key[MLX5_RSS_HASH_KEY_LEN]; /**< RSS hash key. */
+};
 
-/* DV flows structure. */
-struct mlx5_flow_dv {
-       struct mlx5_hrxq *hrxq; /**< Hash Rx queues. */
+/** Device flow handle structure for DV mode only. */
+struct mlx5_flow_handle_dv {
        /* Flow DV api: */
        struct mlx5_flow_dv_matcher *matcher; /**< Cache to matcher. */
-       struct mlx5_flow_dv_match_params value;
-       /**< Holds the value that the packet is compared to. */
        struct mlx5_flow_dv_encap_decap_resource *encap_decap;
        /**< Pointer to encap/decap resource in cache. */
        struct mlx5_flow_dv_modify_hdr_resource *modify_hdr;
        /**< Pointer to modify header resource in cache. */
-       struct ibv_flow *flow; /**< Installed flow. */
        struct mlx5_flow_dv_jump_tbl_resource *jump;
        /**< Pointer to the jump action resource. */
        struct mlx5_flow_dv_port_id_action_resource *port_id_action;
@@ -493,65 +496,64 @@ struct mlx5_flow_dv {
        /**< Pointer to push VLAN action resource in cache. */
        struct mlx5_flow_dv_tag_resource *tag_resource;
        /**< pointer to the tag action. */
+};
+
+/** Device flow handle structure: used both for creating & destroying. */
+struct mlx5_flow_handle {
+       uint64_t layers;
+       /**< Bit-fields of present layers, see MLX5_FLOW_LAYER_*. */
+       uint64_t act_flags;
+       /**< Bit-fields of detected actions, see MLX5_FLOW_ACTION_*. */
+       void *ib_flow; /**< Verbs flow pointer. */
+       struct mlx5_hrxq *hrxq; /**< Hash Rx queue object. */
+       struct mlx5_vf_vlan vf_vlan; /**< Structure for VF VLAN workaround. */
+       union {
+               uint32_t qrss_id; /**< Uniqie Q/RSS suffix subflow tag. */
+               uint32_t mtr_flow_id; /**< Unique meter match flow id. */
+       };
 #ifdef HAVE_IBV_FLOW_DV_SUPPORT
-       void *actions[MLX5_DV_MAX_NUMBER_OF_ACTIONS];
-       /**< Action list. */
+       struct mlx5_flow_handle_dv dvh;
 #endif
-       int actions_n; /**< number of actions. */
 };
 
-/* Verbs specification header. */
-struct ibv_spec_header {
-       enum ibv_flow_spec_type type;
-       uint16_t size;
-};
+/*
+ * Max number of actions per DV flow.
+ * See CREATE_FLOW_MAX_FLOW_ACTIONS_SUPPORTED
+ * in rdma-core file providers/mlx5/verbs.c.
+ */
+#define MLX5_DV_MAX_NUMBER_OF_ACTIONS 8
 
-/** Handles information leading to a drop fate. */
-struct mlx5_flow_verbs {
-       LIST_ENTRY(mlx5_flow_verbs) next;
-       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. */
-       struct mlx5_vf_vlan vf_vlan;
-       /**< Structure for VF VLAN workaround. */
+/** Device flow structure only for DV flow creation. */
+struct mlx5_flow_resource_dv {
+       uint32_t group; /**< The group index. */
+       uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
+       int actions_n; /**< number of actions. */
+       void *actions[MLX5_DV_MAX_NUMBER_OF_ACTIONS]; /**< Action list. */
+       struct mlx5_flow_dv_match_params value;
+       /**< Holds the value that the packet is compared to. */
 };
 
-struct mlx5_flow_rss {
-       uint32_t level;
-       uint32_t queue_num; /**< Number of entries in @p queue. */
-       uint64_t types; /**< Specific RSS hash types (see ETH_RSS_*). */
-       uint16_t (*queue)[]; /**< Destination queues to redirect traffic to. */
-       uint8_t key[MLX5_RSS_HASH_KEY_LEN]; /**< RSS hash key. */
+/** Device flow structure only for Verbs flow creation. */
+struct mlx5_flow_resource_verbs {
+       unsigned int size; /**< Size of the attribute. */
+       struct ibv_flow_attr *attr; /**< Pointer to the Specification buffer. */
+       uint8_t *specs; /**< Pointer to the specifications. */
 };
 
 /** Device flow structure. */
 struct mlx5_flow {
-       LIST_ENTRY(mlx5_flow) next;
+       LIST_ENTRY(mlx5_flow) next; /**< Pointer to next device flow. */
        struct rte_flow *flow; /**< Pointer to the main flow. */
-       uint64_t layers;
-       /**< Bit-fields of present layers, see MLX5_FLOW_LAYER_*. */
-       uint64_t actions;
-       /**< Bit-fields of detected actions, see MLX5_FLOW_ACTION_*. */
        uint64_t hash_fields; /**< Verbs hash Rx queue hash fields. */
+       bool external; /**< true if the flow is created external to PMD. */
        uint8_t ingress; /**< 1 if the flow is ingress. */
-       uint32_t group; /**< The group index. */
-       uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
        union {
 #ifdef HAVE_IBV_FLOW_DV_SUPPORT
-               struct mlx5_flow_dv dv;
+               struct mlx5_flow_resource_dv dv;
 #endif
-               struct mlx5_flow_verbs verbs;
-       };
-       union {
-               uint32_t qrss_id; /**< Uniqie Q/RSS suffix subflow tag. */
-               uint32_t mtr_flow_id; /**< Unique meter match flow id. */
+               struct mlx5_flow_resource_verbs verbs;
        };
-       bool external; /**< true if the flow is created external to PMD. */
+       struct mlx5_flow_handle handle;
 };
 
 /* Flow meter state. */
index 7b3d419..f7dfeff 100644 (file)
@@ -92,20 +92,22 @@ static void
 flow_dv_attr_init(const struct rte_flow_item *item, union flow_dv_attr *attr,
                  struct mlx5_flow *dev_flow, bool tunnel_decap)
 {
+       uint64_t layers = dev_flow->handle.layers;
+
        /*
         * If layers is already initialized, it means this dev_flow is the
         * suffix flow, the layers flags is set by the prefix flow. Need to
         * use the layer flags from prefix flow as the suffix flow may not
         * have the user defined items as the flow is split.
         */
-       if (dev_flow->layers) {
-               if (dev_flow->layers & MLX5_FLOW_LAYER_OUTER_L3_IPV4)
+       if (layers) {
+               if (layers & MLX5_FLOW_LAYER_OUTER_L3_IPV4)
                        attr->ipv4 = 1;
-               else if (dev_flow->layers & MLX5_FLOW_LAYER_OUTER_L3_IPV6)
+               else if (layers & MLX5_FLOW_LAYER_OUTER_L3_IPV6)
                        attr->ipv6 = 1;
-               if (dev_flow->layers & MLX5_FLOW_LAYER_OUTER_L4_TCP)
+               if (layers & MLX5_FLOW_LAYER_OUTER_L4_TCP)
                        attr->tcp = 1;
-               else if (dev_flow->layers & MLX5_FLOW_LAYER_OUTER_L4_UDP)
+               else if (layers & MLX5_FLOW_LAYER_OUTER_L4_UDP)
                        attr->udp = 1;
                attr->valid = 1;
                return;
@@ -2436,7 +2438,7 @@ flow_dv_encap_decap_resource_register
        struct mlx5_flow_dv_encap_decap_resource *cache_resource;
        struct mlx5dv_dr_domain *domain;
 
-       resource->flags = dev_flow->group ? 0 : 1;
+       resource->flags = dev_flow->dv.group ? 0 : 1;
        if (resource->ft_type == MLX5DV_FLOW_TABLE_TYPE_FDB)
                domain = sh->fdb_domain;
        else if (resource->ft_type == MLX5DV_FLOW_TABLE_TYPE_NIC_RX)
@@ -2456,7 +2458,7 @@ flow_dv_encap_decap_resource_register
                                (void *)cache_resource,
                                rte_atomic32_read(&cache_resource->refcnt));
                        rte_atomic32_inc(&cache_resource->refcnt);
-                       dev_flow->dv.encap_decap = cache_resource;
+                       dev_flow->handle.dvh.encap_decap = cache_resource;
                        return 0;
                }
        }
@@ -2482,7 +2484,7 @@ flow_dv_encap_decap_resource_register
        rte_atomic32_init(&cache_resource->refcnt);
        rte_atomic32_inc(&cache_resource->refcnt);
        LIST_INSERT_HEAD(&sh->encaps_decaps, cache_resource, next);
-       dev_flow->dv.encap_decap = cache_resource;
+       dev_flow->handle.dvh.encap_decap = cache_resource;
        DRV_LOG(DEBUG, "new encap/decap resource %p: refcnt %d++",
                (void *)cache_resource,
                rte_atomic32_read(&cache_resource->refcnt));
@@ -2533,7 +2535,7 @@ flow_dv_jump_tbl_resource_register
                        (void *)&tbl_data->jump, cnt);
        }
        rte_atomic32_inc(&tbl_data->jump.refcnt);
-       dev_flow->dv.jump = &tbl_data->jump;
+       dev_flow->handle.dvh.jump = &tbl_data->jump;
        return 0;
 }
 
@@ -2571,7 +2573,7 @@ flow_dv_port_id_action_resource_register
                                (void *)cache_resource,
                                rte_atomic32_read(&cache_resource->refcnt));
                        rte_atomic32_inc(&cache_resource->refcnt);
-                       dev_flow->dv.port_id_action = cache_resource;
+                       dev_flow->handle.dvh.port_id_action = cache_resource;
                        return 0;
                }
        }
@@ -2599,7 +2601,7 @@ flow_dv_port_id_action_resource_register
        rte_atomic32_init(&cache_resource->refcnt);
        rte_atomic32_inc(&cache_resource->refcnt);
        LIST_INSERT_HEAD(&sh->port_id_action_list, cache_resource, next);
-       dev_flow->dv.port_id_action = cache_resource;
+       dev_flow->handle.dvh.port_id_action = cache_resource;
        DRV_LOG(DEBUG, "new port id action resource %p: refcnt %d++",
                (void *)cache_resource,
                rte_atomic32_read(&cache_resource->refcnt));
@@ -2642,7 +2644,7 @@ flow_dv_push_vlan_action_resource_register
                                (void *)cache_resource,
                                rte_atomic32_read(&cache_resource->refcnt));
                        rte_atomic32_inc(&cache_resource->refcnt);
-                       dev_flow->dv.push_vlan_res = cache_resource;
+                       dev_flow->handle.dvh.push_vlan_res = cache_resource;
                        return 0;
                }
        }
@@ -2671,7 +2673,7 @@ flow_dv_push_vlan_action_resource_register
        rte_atomic32_init(&cache_resource->refcnt);
        rte_atomic32_inc(&cache_resource->refcnt);
        LIST_INSERT_HEAD(&sh->push_vlan_action_list, cache_resource, next);
-       dev_flow->dv.push_vlan_res = cache_resource;
+       dev_flow->handle.dvh.push_vlan_res = cache_resource;
        DRV_LOG(DEBUG, "new push vlan action resource %p: refcnt %d++",
                (void *)cache_resource,
                rte_atomic32_read(&cache_resource->refcnt));
@@ -3758,8 +3760,8 @@ flow_dv_modify_hdr_resource_register
        struct mlx5dv_dr_domain *ns;
        uint32_t actions_len;
 
-       resource->flags =
-               dev_flow->group ? 0 : MLX5DV_DR_ACTION_FLAGS_ROOT_LEVEL;
+       resource->flags = dev_flow->dv.group ? 0 :
+                         MLX5DV_DR_ACTION_FLAGS_ROOT_LEVEL;
        if (resource->actions_num > flow_dv_modify_hdr_action_max(dev,
                                    resource->flags))
                return rte_flow_error_set(error, EOVERFLOW,
@@ -3784,7 +3786,7 @@ flow_dv_modify_hdr_resource_register
                                (void *)cache_resource,
                                rte_atomic32_read(&cache_resource->refcnt));
                        rte_atomic32_inc(&cache_resource->refcnt);
-                       dev_flow->dv.modify_hdr = cache_resource;
+                       dev_flow->handle.dvh.modify_hdr = cache_resource;
                        return 0;
                }
        }
@@ -3811,7 +3813,7 @@ flow_dv_modify_hdr_resource_register
        rte_atomic32_init(&cache_resource->refcnt);
        rte_atomic32_inc(&cache_resource->refcnt);
        LIST_INSERT_HEAD(&sh->modify_cmds, cache_resource, next);
-       dev_flow->dv.modify_hdr = cache_resource;
+       dev_flow->handle.dvh.modify_hdr = cache_resource;
        DRV_LOG(DEBUG, "new modify-header resource %p: refcnt %d++",
                (void *)cache_resource,
                rte_atomic32_read(&cache_resource->refcnt));
@@ -5300,7 +5302,7 @@ flow_dv_prepare(const struct rte_flow_attr *attr __rte_unused,
        }
        dev_flow->dv.value.size = MLX5_ST_SZ_BYTES(fte_match_param);
        dev_flow->ingress = attr->ingress;
-       dev_flow->transfer = attr->transfer;
+       dev_flow->dv.transfer = attr->transfer;
        return dev_flow;
 }
 
@@ -5456,7 +5458,7 @@ flow_dv_translate_item_vlan(struct mlx5_flow *dev_flow,
                 * This is workaround, masks are not supported,
                 * and pre-validated.
                 */
-               dev_flow->dv.vf_vlan.tag =
+               dev_flow->handle.vf_vlan.tag =
                        rte_be_to_cpu_16(vlan_v->tci) & 0x0fff;
        }
        tci_m = rte_be_to_cpu_16(vlan_m->tci);
@@ -6957,7 +6959,7 @@ flow_dv_matcher_register(struct rte_eth_dev *dev,
                                (void *)cache_matcher,
                                rte_atomic32_read(&cache_matcher->refcnt));
                        rte_atomic32_inc(&cache_matcher->refcnt);
-                       dev_flow->dv.matcher = cache_matcher;
+                       dev_flow->handle.dvh.matcher = cache_matcher;
                        /* old matcher should not make the table ref++. */
                        flow_dv_tbl_resource_release(dev, tbl);
                        return 0;
@@ -6994,7 +6996,7 @@ flow_dv_matcher_register(struct rte_eth_dev *dev,
        /* only matcher ref++, table ref++ already done above in get API. */
        rte_atomic32_inc(&cache_matcher->refcnt);
        LIST_INSERT_HEAD(&tbl_data->matchers, cache_matcher, next);
-       dev_flow->dv.matcher = cache_matcher;
+       dev_flow->handle.dvh.matcher = cache_matcher;
        DRV_LOG(DEBUG, "%s group %u priority %hd new %s matcher %p: refcnt %d",
                key->domain ? "FDB" : "NIC", key->table_id,
                cache_matcher->priority,
@@ -7036,7 +7038,7 @@ flow_dv_tag_resource_register
                cache_resource = container_of
                        (entry, struct mlx5_flow_dv_tag_resource, entry);
                rte_atomic32_inc(&cache_resource->refcnt);
-               dev_flow->dv.tag_resource = cache_resource;
+               dev_flow->handle.dvh.tag_resource = cache_resource;
                DRV_LOG(DEBUG, "cached tag resource %p: refcnt now %d++",
                        (void *)cache_resource,
                        rte_atomic32_read(&cache_resource->refcnt));
@@ -7065,7 +7067,7 @@ flow_dv_tag_resource_register
                                          RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
                                          NULL, "cannot insert tag");
        }
-       dev_flow->dv.tag_resource = cache_resource;
+       dev_flow->handle.dvh.tag_resource = cache_resource;
        DRV_LOG(DEBUG, "new tag resource %p: refcnt now %d++",
                (void *)cache_resource,
                rte_atomic32_read(&cache_resource->refcnt));
@@ -7210,7 +7212,7 @@ static void
 flow_dv_hashfields_set(struct mlx5_flow *dev_flow)
 {
        struct rte_flow *flow = dev_flow->flow;
-       uint64_t items = dev_flow->layers;
+       uint64_t items = dev_flow->handle.layers;
        int rss_inner = 0;
        uint64_t rss_types = rte_eth_rss_hf_refine(flow->rss.types);
 
@@ -7335,7 +7337,7 @@ __flow_dv_translate(struct rte_eth_dev *dev,
                                       !!priv->fdb_def_rule, &table, error);
        if (ret)
                return ret;
-       dev_flow->group = table;
+       dev_flow->dv.group = table;
        if (attr->transfer)
                mhdr_res->ft_type = MLX5DV_FLOW_TABLE_TYPE_FDB;
        if (priority == MLX5_FLOW_PRIO_RSVD)
@@ -7368,7 +7370,7 @@ __flow_dv_translate(struct rte_eth_dev *dev,
                            (dev, &port_id_resource, dev_flow, error))
                                return -rte_errno;
                        dev_flow->dv.actions[actions_n++] =
-                               dev_flow->dv.port_id_action->action;
+                               dev_flow->handle.dvh.port_id_action->action;
                        action_flags |= MLX5_FLOW_ACTION_PORT_ID;
                        break;
                case RTE_FLOW_ACTION_TYPE_FLAG:
@@ -7386,12 +7388,12 @@ __flow_dv_translate(struct rte_eth_dev *dev,
                                break;
                        }
                        tag_be = mlx5_flow_mark_set(MLX5_FLOW_MARK_DEFAULT);
-                       if (!dev_flow->dv.tag_resource)
+                       if (!dev_flow->handle.dvh.tag_resource)
                                if (flow_dv_tag_resource_register
                                    (dev, tag_be, dev_flow, error))
                                        return -rte_errno;
                        dev_flow->dv.actions[actions_n++] =
-                               dev_flow->dv.tag_resource->action;
+                               dev_flow->handle.dvh.tag_resource->action;
                        break;
                case RTE_FLOW_ACTION_TYPE_MARK:
                        action_flags |= MLX5_FLOW_ACTION_MARK;
@@ -7413,12 +7415,12 @@ __flow_dv_translate(struct rte_eth_dev *dev,
                        tag_be = mlx5_flow_mark_set
                              (((const struct rte_flow_action_mark *)
                               (actions->conf))->id);
-                       if (!dev_flow->dv.tag_resource)
+                       if (!dev_flow->handle.dvh.tag_resource)
                                if (flow_dv_tag_resource_register
                                    (dev, tag_be, dev_flow, error))
                                        return -rte_errno;
                        dev_flow->dv.actions[actions_n++] =
-                               dev_flow->dv.tag_resource->action;
+                               dev_flow->handle.dvh.tag_resource->action;
                        break;
                case RTE_FLOW_ACTION_TYPE_SET_META:
                        if (flow_dv_convert_action_set_meta
@@ -7468,9 +7470,9 @@ __flow_dv_translate(struct rte_eth_dev *dev,
                                goto cnt_err;
                        }
                        flow->counter = flow_dv_counter_alloc(dev,
-                                                             count->shared,
-                                                             count->id,
-                                                             dev_flow->group);
+                                                       count->shared,
+                                                       count->id,
+                                                       dev_flow->dv.group);
                        if (flow->counter == NULL)
                                goto cnt_err;
                        dev_flow->dv.actions[actions_n++] =
@@ -7516,7 +7518,7 @@ cnt_err:
                                            (dev, attr, &vlan, dev_flow, error))
                                return -rte_errno;
                        dev_flow->dv.actions[actions_n++] =
-                                          dev_flow->dv.push_vlan_res->action;
+                               dev_flow->handle.dvh.push_vlan_res->action;
                        action_flags |= MLX5_FLOW_ACTION_OF_PUSH_VLAN;
                        break;
                case RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_PCP:
@@ -7543,7 +7545,7 @@ cnt_err:
                                                           error))
                                return -rte_errno;
                        dev_flow->dv.actions[actions_n++] =
-                               dev_flow->dv.encap_decap->verbs_action;
+                               dev_flow->handle.dvh.encap_decap->verbs_action;
                        action_flags |= MLX5_FLOW_ACTION_ENCAP;
                        break;
                case RTE_FLOW_ACTION_TYPE_VXLAN_DECAP:
@@ -7553,7 +7555,7 @@ cnt_err:
                                                           error))
                                return -rte_errno;
                        dev_flow->dv.actions[actions_n++] =
-                               dev_flow->dv.encap_decap->verbs_action;
+                               dev_flow->handle.dvh.encap_decap->verbs_action;
                        action_flags |= MLX5_FLOW_ACTION_DECAP;
                        break;
                case RTE_FLOW_ACTION_TYPE_RAW_ENCAP:
@@ -7563,7 +7565,7 @@ cnt_err:
                                        (dev, actions, dev_flow, attr, error))
                                        return -rte_errno;
                                dev_flow->dv.actions[actions_n++] =
-                                       dev_flow->dv.encap_decap->verbs_action;
+                               dev_flow->handle.dvh.encap_decap->verbs_action;
                        } else {
                                /* Handle encap without preceding decap. */
                                if (flow_dv_create_action_l2_encap
@@ -7571,7 +7573,7 @@ cnt_err:
                                     error))
                                        return -rte_errno;
                                dev_flow->dv.actions[actions_n++] =
-                                       dev_flow->dv.encap_decap->verbs_action;
+                               dev_flow->handle.dvh.encap_decap->verbs_action;
                        }
                        action_flags |= MLX5_FLOW_ACTION_ENCAP;
                        break;
@@ -7583,7 +7585,7 @@ cnt_err:
                                    (dev, dev_flow, attr->transfer, error))
                                        return -rte_errno;
                                dev_flow->dv.actions[actions_n++] =
-                                       dev_flow->dv.encap_decap->verbs_action;
+                               dev_flow->handle.dvh.encap_decap->verbs_action;
                        }
                        /* If decap is followed by encap, handle it at encap. */
                        action_flags |= MLX5_FLOW_ACTION_DECAP;
@@ -7615,7 +7617,7 @@ cnt_err:
                                                 "cannot create jump action.");
                        }
                        dev_flow->dv.actions[actions_n++] =
-                               dev_flow->dv.jump->action;
+                               dev_flow->handle.dvh.jump->action;
                        action_flags |= MLX5_FLOW_ACTION_JUMP;
                        break;
                case RTE_FLOW_ACTION_TYPE_SET_MAC_SRC:
@@ -7748,7 +7750,7 @@ cnt_err:
                                        (dev, mhdr_res, dev_flow, error))
                                        return -rte_errno;
                                dev_flow->dv.actions[modify_action_position] =
-                                       dev_flow->dv.modify_hdr->verbs_action;
+                               dev_flow->handle.dvh.modify_hdr->verbs_action;
                        }
                        break;
                default:
@@ -7759,7 +7761,7 @@ cnt_err:
                        modify_action_position = actions_n++;
        }
        dev_flow->dv.actions_n = actions_n;
-       dev_flow->actions = action_flags;
+       dev_flow->handle.act_flags = action_flags;
        for (; items->type != RTE_FLOW_ITEM_TYPE_END; items++) {
                int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL);
                int item_type = items->type;
@@ -7792,7 +7794,7 @@ cnt_err:
                                                  &item_flags, &tunnel);
                        flow_dv_translate_item_ipv4(match_mask, match_value,
                                                    items, item_flags, tunnel,
-                                                   dev_flow->group);
+                                                   dev_flow->dv.group);
                        matcher.priority = MLX5_PRIORITY_MAP_L3;
                        last_item = tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV4 :
                                             MLX5_FLOW_LAYER_OUTER_L3_IPV4;
@@ -7815,7 +7817,7 @@ cnt_err:
                                                  &item_flags, &tunnel);
                        flow_dv_translate_item_ipv6(match_mask, match_value,
                                                    items, item_flags, tunnel,
-                                                   dev_flow->group);
+                                                   dev_flow->dv.group);
                        matcher.priority = MLX5_PRIORITY_MAP_L3;
                        last_item = tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV6 :
                                             MLX5_FLOW_LAYER_OUTER_L3_IPV6;
@@ -7964,7 +7966,7 @@ cnt_err:
         * Layers may be already initialized from prefix flow if this dev_flow
         * is the suffix flow.
         */
-       dev_flow->layers |= item_flags;
+       dev_flow->handle.layers |= item_flags;
        if (action_flags & MLX5_FLOW_ACTION_RSS)
                flow_dv_hashfields_set(dev_flow);
        /* Register matcher. */
@@ -7975,7 +7977,7 @@ cnt_err:
        /* reserved field no needs to be set to 0 here. */
        tbl_key.domain = attr->transfer;
        tbl_key.direction = attr->egress;
-       tbl_key.table_id = dev_flow->group;
+       tbl_key.table_id = dev_flow->dv.group;
        if (flow_dv_matcher_register(dev, &matcher, &tbl_key, dev_flow, error))
                return -rte_errno;
        return 0;
@@ -7999,21 +8001,25 @@ static int
 __flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                struct rte_flow_error *error)
 {
-       struct mlx5_flow_dv *dv;
+       struct mlx5_flow_resource_dv *dv;
+       struct mlx5_flow_handle *dh;
+       struct mlx5_flow_handle_dv *dv_h;
        struct mlx5_flow *dev_flow;
        struct mlx5_priv *priv = dev->data->dev_private;
        int n;
        int err;
 
        LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
+               dh = &dev_flow->handle;
                dv = &dev_flow->dv;
                n = dv->actions_n;
-               if (dev_flow->actions & MLX5_FLOW_ACTION_DROP) {
-                       if (dev_flow->transfer) {
+               dv_h = &dh->dvh;
+               if (dh->act_flags & MLX5_FLOW_ACTION_DROP) {
+                       if (dv->transfer) {
                                dv->actions[n++] = priv->sh->esw_drop_action;
                        } else {
-                               dv->hrxq = mlx5_hrxq_drop_new(dev);
-                               if (!dv->hrxq) {
+                               dh->hrxq = mlx5_hrxq_drop_new(dev);
+                               if (!dh->hrxq) {
                                        rte_flow_error_set
                                                (error, errno,
                                                 RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
@@ -8021,9 +8027,9 @@ __flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                                                 "cannot get drop hash queue");
                                        goto error;
                                }
-                               dv->actions[n++] = dv->hrxq->action;
+                               dv->actions[n++] = dh->hrxq->action;
                        }
-               } else if (dev_flow->actions &
+               } else if (dh->act_flags &
                           (MLX5_FLOW_ACTION_QUEUE | MLX5_FLOW_ACTION_RSS)) {
                        struct mlx5_hrxq *hrxq;
 
@@ -8040,7 +8046,7 @@ __flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                                         dev_flow->hash_fields,
                                         (*flow->rss.queue),
                                         flow->rss.queue_num,
-                                        !!(dev_flow->layers &
+                                        !!(dev_flow->handle.layers &
                                            MLX5_FLOW_LAYER_TUNNEL));
                        }
                        if (!hrxq) {
@@ -8050,14 +8056,14 @@ __flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                                         "cannot get hash queue");
                                goto error;
                        }
-                       dv->hrxq = hrxq;
-                       dv->actions[n++] = dv->hrxq->action;
+                       dh->hrxq = hrxq;
+                       dv->actions[n++] = dh->hrxq->action;
                }
-               dv->flow =
-                       mlx5_glue->dv_create_flow(dv->matcher->matcher_object,
+               dh->ib_flow =
+                       mlx5_glue->dv_create_flow(dv_h->matcher->matcher_object,
                                                  (void *)&dv->value, n,
                                                  dv->actions);
-               if (!dv->flow) {
+               if (!dh->ib_flow) {
                        rte_flow_error_set(error, errno,
                                           RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
                                           NULL,
@@ -8065,32 +8071,30 @@ __flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                        goto error;
                }
                if (priv->vmwa_context &&
-                   dev_flow->dv.vf_vlan.tag &&
-                   !dev_flow->dv.vf_vlan.created) {
+                   dh->vf_vlan.tag && !dh->vf_vlan.created) {
                        /*
                         * The rule contains the VLAN pattern.
                         * For VF we are going to create VLAN
                         * interface to make hypervisor set correct
                         * e-Switch vport context.
                         */
-                       mlx5_vlan_vmwa_acquire(dev, &dev_flow->dv.vf_vlan);
+                       mlx5_vlan_vmwa_acquire(dev, &dh->vf_vlan);
                }
        }
        return 0;
 error:
        err = rte_errno; /* Save rte_errno before cleanup. */
        LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
-               struct mlx5_flow_dv *dv = &dev_flow->dv;
-               if (dv->hrxq) {
-                       if (dev_flow->actions & MLX5_FLOW_ACTION_DROP)
+               struct mlx5_flow_handle *dh_tmp = &dev_flow->handle;
+               if (dh_tmp->hrxq) {
+                       if (dh_tmp->act_flags & MLX5_FLOW_ACTION_DROP)
                                mlx5_hrxq_drop_release(dev);
                        else
-                               mlx5_hrxq_release(dev, dv->hrxq);
-                       dv->hrxq = NULL;
+                               mlx5_hrxq_release(dev, dh_tmp->hrxq);
+                       dh_tmp->hrxq = NULL;
                }
-               if (dev_flow->dv.vf_vlan.tag &&
-                   dev_flow->dv.vf_vlan.created)
-                       mlx5_vlan_vmwa_release(dev, &dev_flow->dv.vf_vlan);
+               if (dh_tmp->vf_vlan.tag && dh_tmp->vf_vlan.created)
+                       mlx5_vlan_vmwa_release(dev, &dh_tmp->vf_vlan);
        }
        rte_errno = err; /* Restore rte_errno. */
        return -rte_errno;
@@ -8111,7 +8115,7 @@ static int
 flow_dv_matcher_release(struct rte_eth_dev *dev,
                        struct mlx5_flow *flow)
 {
-       struct mlx5_flow_dv_matcher *matcher = flow->dv.matcher;
+       struct mlx5_flow_dv_matcher *matcher = flow->handle.dvh.matcher;
 
        MLX5_ASSERT(matcher->matcher_object);
        DRV_LOG(DEBUG, "port %u matcher %p: refcnt %d--",
@@ -8144,7 +8148,7 @@ static int
 flow_dv_encap_decap_resource_release(struct mlx5_flow *flow)
 {
        struct mlx5_flow_dv_encap_decap_resource *cache_resource =
-                                               flow->dv.encap_decap;
+                                               flow->handle.dvh.encap_decap;
 
        MLX5_ASSERT(cache_resource->verbs_action);
        DRV_LOG(DEBUG, "encap/decap resource %p: refcnt %d--",
@@ -8177,7 +8181,8 @@ static int
 flow_dv_jump_tbl_resource_release(struct rte_eth_dev *dev,
                                  struct mlx5_flow *flow)
 {
-       struct mlx5_flow_dv_jump_tbl_resource *cache_resource = flow->dv.jump;
+       struct mlx5_flow_dv_jump_tbl_resource *cache_resource =
+                                               flow->handle.dvh.jump;
        struct mlx5_flow_tbl_data_entry *tbl_data =
                        container_of(cache_resource,
                                     struct mlx5_flow_tbl_data_entry, jump);
@@ -8211,7 +8216,7 @@ static int
 flow_dv_modify_hdr_resource_release(struct mlx5_flow *flow)
 {
        struct mlx5_flow_dv_modify_hdr_resource *cache_resource =
-                                               flow->dv.modify_hdr;
+                                               flow->handle.dvh.modify_hdr;
 
        MLX5_ASSERT(cache_resource->verbs_action);
        DRV_LOG(DEBUG, "modify-header resource %p: refcnt %d--",
@@ -8242,7 +8247,7 @@ static int
 flow_dv_port_id_action_resource_release(struct mlx5_flow *flow)
 {
        struct mlx5_flow_dv_port_id_action_resource *cache_resource =
-               flow->dv.port_id_action;
+                                               flow->handle.dvh.port_id_action;
 
        MLX5_ASSERT(cache_resource->action);
        DRV_LOG(DEBUG, "port ID action resource %p: refcnt %d--",
@@ -8273,7 +8278,7 @@ static int
 flow_dv_push_vlan_action_resource_release(struct mlx5_flow *flow)
 {
        struct mlx5_flow_dv_push_vlan_action_resource *cache_resource =
-               flow->dv.push_vlan_res;
+                                               flow->handle.dvh.push_vlan_res;
 
        MLX5_ASSERT(cache_resource->action);
        DRV_LOG(DEBUG, "push VLAN action resource %p: refcnt %d--",
@@ -8303,27 +8308,26 @@ flow_dv_push_vlan_action_resource_release(struct mlx5_flow *flow)
 static void
 __flow_dv_remove(struct rte_eth_dev *dev, struct rte_flow *flow)
 {
-       struct mlx5_flow_dv *dv;
+       struct mlx5_flow_handle *dh;
        struct mlx5_flow *dev_flow;
 
        if (!flow)
                return;
        LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
-               dv = &dev_flow->dv;
-               if (dv->flow) {
-                       claim_zero(mlx5_glue->dv_destroy_flow(dv->flow));
-                       dv->flow = NULL;
+               dh = &dev_flow->handle;
+               if (dh->ib_flow) {
+                       claim_zero(mlx5_glue->dv_destroy_flow(dh->ib_flow));
+                       dh->ib_flow = NULL;
                }
-               if (dv->hrxq) {
-                       if (dev_flow->actions & MLX5_FLOW_ACTION_DROP)
+               if (dh->hrxq) {
+                       if (dev_flow->handle.act_flags & MLX5_FLOW_ACTION_DROP)
                                mlx5_hrxq_drop_release(dev);
                        else
-                               mlx5_hrxq_release(dev, dv->hrxq);
-                       dv->hrxq = NULL;
+                               mlx5_hrxq_release(dev, dh->hrxq);
+                       dh->hrxq = NULL;
                }
-               if (dev_flow->dv.vf_vlan.tag &&
-                   dev_flow->dv.vf_vlan.created)
-                       mlx5_vlan_vmwa_release(dev, &dev_flow->dv.vf_vlan);
+               if (dh->vf_vlan.tag && dh->vf_vlan.created)
+                       mlx5_vlan_vmwa_release(dev, &dh->vf_vlan);
        }
 }
 
@@ -8355,20 +8359,21 @@ __flow_dv_destroy(struct rte_eth_dev *dev, struct rte_flow *flow)
        while (!LIST_EMPTY(&flow->dev_flows)) {
                dev_flow = LIST_FIRST(&flow->dev_flows);
                LIST_REMOVE(dev_flow, next);
-               if (dev_flow->dv.matcher)
+               if (dev_flow->handle.dvh.matcher)
                        flow_dv_matcher_release(dev, dev_flow);
-               if (dev_flow->dv.encap_decap)
+               if (dev_flow->handle.dvh.encap_decap)
                        flow_dv_encap_decap_resource_release(dev_flow);
-               if (dev_flow->dv.modify_hdr)
+               if (dev_flow->handle.dvh.modify_hdr)
                        flow_dv_modify_hdr_resource_release(dev_flow);
-               if (dev_flow->dv.jump)
+               if (dev_flow->handle.dvh.jump)
                        flow_dv_jump_tbl_resource_release(dev, dev_flow);
-               if (dev_flow->dv.port_id_action)
+               if (dev_flow->handle.dvh.port_id_action)
                        flow_dv_port_id_action_resource_release(dev_flow);
-               if (dev_flow->dv.push_vlan_res)
+               if (dev_flow->handle.dvh.push_vlan_res)
                        flow_dv_push_vlan_action_resource_release(dev_flow);
-               if (dev_flow->dv.tag_resource)
-                       flow_dv_tag_release(dev, dev_flow->dv.tag_resource);
+               if (dev_flow->handle.dvh.tag_resource)
+                       flow_dv_tag_release(dev,
+                                       dev_flow->handle.dvh.tag_resource);
                rte_free(dev_flow);
        }
 }
index 459e7b6..08185ec 100644 (file)
@@ -253,7 +253,8 @@ flow_verbs_counter_query(struct rte_eth_dev *dev __rte_unused,
  *   Size in bytes of the specification to copy.
  */
 static void
-flow_verbs_spec_add(struct mlx5_flow_verbs *verbs, void *src, unsigned int size)
+flow_verbs_spec_add(struct mlx5_flow_resource_verbs *verbs,
+                   void *src, unsigned int size)
 {
        void *dst;
 
@@ -393,7 +394,7 @@ flow_verbs_translate_item_vlan(struct mlx5_flow *dev_flow,
        else
                flow_verbs_item_vlan_update(dev_flow->verbs.attr, &eth);
        if (!tunnel)
-               dev_flow->verbs.vf_vlan.tag =
+               dev_flow->handle.vf_vlan.tag =
                        rte_be_to_cpu_16(spec->tci) & 0x0fff;
 }
 
@@ -743,7 +744,7 @@ flow_verbs_translate_item_gre(struct mlx5_flow *dev_flow,
                              const struct rte_flow_item *item __rte_unused,
                              uint64_t item_flags)
 {
-       struct mlx5_flow_verbs *verbs = &dev_flow->verbs;
+       struct mlx5_flow_resource_verbs *verbs = &dev_flow->verbs;
 #ifndef HAVE_IBV_DEVICE_MPLS_SUPPORT
        unsigned int size = sizeof(struct ibv_flow_spec_tunnel);
        struct ibv_flow_spec_tunnel tunnel = {
@@ -1418,7 +1419,7 @@ flow_verbs_prepare(const struct rte_flow_attr *attr __rte_unused,
        dev_flow->verbs.attr = (void *)(dev_flow + 1);
        dev_flow->verbs.specs = (void *)(dev_flow->verbs.attr + 1);
        dev_flow->ingress = attr->ingress;
-       dev_flow->transfer = attr->transfer;
+       /* Need to set transfer attribute: not supported in Verbs mode. */
        return dev_flow;
 }
 
@@ -1498,7 +1499,7 @@ flow_verbs_translate(struct rte_eth_dev *dev,
                                                  "action not supported");
                }
        }
-       dev_flow->actions = action_flags;
+       dev_flow->handle.act_flags = action_flags;
        for (; items->type != RTE_FLOW_ITEM_TYPE_END; items++) {
                int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL);
 
@@ -1600,7 +1601,7 @@ flow_verbs_translate(struct rte_eth_dev *dev,
                                                  "item not supported");
                }
        }
-       dev_flow->layers = item_flags;
+       dev_flow->handle.layers = item_flags;
        dev_flow->verbs.attr->priority =
                mlx5_flow_adjust_priority(dev, priority, subpriority);
        dev_flow->verbs.attr->port = (uint8_t)priv->ibv_port;
@@ -1618,28 +1619,26 @@ flow_verbs_translate(struct rte_eth_dev *dev,
 static void
 flow_verbs_remove(struct rte_eth_dev *dev, struct rte_flow *flow)
 {
-       struct mlx5_flow_verbs *verbs;
+       struct mlx5_flow_handle *dh;
        struct mlx5_flow *dev_flow;
 
        if (!flow)
                return;
        LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
-               verbs = &dev_flow->verbs;
-               if (verbs->flow) {
-                       claim_zero(mlx5_glue->destroy_flow(verbs->flow));
-                       verbs->flow = NULL;
+               dh = &dev_flow->handle;
+               if (dh->ib_flow) {
+                       claim_zero(mlx5_glue->destroy_flow(dh->ib_flow));
+                       dh->ib_flow = NULL;
                }
-               if (verbs->hrxq) {
-                       if (dev_flow->actions & MLX5_FLOW_ACTION_DROP)
+               if (dh->hrxq) {
+                       if (dev_flow->handle.act_flags & MLX5_FLOW_ACTION_DROP)
                                mlx5_hrxq_drop_release(dev);
                        else
-                               mlx5_hrxq_release(dev, verbs->hrxq);
-                       verbs->hrxq = NULL;
-               }
-               if (dev_flow->verbs.vf_vlan.tag &&
-                   dev_flow->verbs.vf_vlan.created) {
-                       mlx5_vlan_vmwa_release(dev, &dev_flow->verbs.vf_vlan);
+                               mlx5_hrxq_release(dev, dh->hrxq);
+                       dh->hrxq = NULL;
                }
+               if (dh->vf_vlan.tag && dh->vf_vlan.created)
+                       mlx5_vlan_vmwa_release(dev, &dh->vf_vlan);
        }
 }
 
@@ -1688,15 +1687,15 @@ flow_verbs_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                 struct rte_flow_error *error)
 {
        struct mlx5_priv *priv = dev->data->dev_private;
-       struct mlx5_flow_verbs *verbs;
+       struct mlx5_flow_handle *dh;
        struct mlx5_flow *dev_flow;
        int err;
 
        LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
-               verbs = &dev_flow->verbs;
-               if (dev_flow->actions & MLX5_FLOW_ACTION_DROP) {
-                       verbs->hrxq = mlx5_hrxq_drop_new(dev);
-                       if (!verbs->hrxq) {
+               dh = &dev_flow->handle;
+               if (dev_flow->handle.act_flags & MLX5_FLOW_ACTION_DROP) {
+                       dh->hrxq = mlx5_hrxq_drop_new(dev);
+                       if (!dh->hrxq) {
                                rte_flow_error_set
                                        (error, errno,
                                         RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
@@ -1714,12 +1713,12 @@ flow_verbs_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                                             flow->rss.queue_num);
                        if (!hrxq)
                                hrxq = mlx5_hrxq_new(dev, flow->rss.key,
-                                                    MLX5_RSS_HASH_KEY_LEN,
-                                                    dev_flow->hash_fields,
-                                                    (*flow->rss.queue),
-                                                    flow->rss.queue_num,
-                                                    !!(dev_flow->layers &
-                                                      MLX5_FLOW_LAYER_TUNNEL));
+                                               MLX5_RSS_HASH_KEY_LEN,
+                                               dev_flow->hash_fields,
+                                               (*flow->rss.queue),
+                                               flow->rss.queue_num,
+                                               !!(dev_flow->handle.layers &
+                                               MLX5_FLOW_LAYER_TUNNEL));
                        if (!hrxq) {
                                rte_flow_error_set
                                        (error, rte_errno,
@@ -1727,11 +1726,11 @@ flow_verbs_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                                         "cannot get hash queue");
                                goto error;
                        }
-                       verbs->hrxq = hrxq;
+                       dh->hrxq = hrxq;
                }
-               verbs->flow = mlx5_glue->create_flow(verbs->hrxq->qp,
-                                                    verbs->attr);
-               if (!verbs->flow) {
+               dh->ib_flow = mlx5_glue->create_flow(dh->hrxq->qp,
+                                                    dev_flow->verbs.attr);
+               if (!dh->ib_flow) {
                        rte_flow_error_set(error, errno,
                                           RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
                                           NULL,
@@ -1739,33 +1738,31 @@ flow_verbs_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
                        goto error;
                }
                if (priv->vmwa_context &&
-                   dev_flow->verbs.vf_vlan.tag &&
-                   !dev_flow->verbs.vf_vlan.created) {
+                   dev_flow->handle.vf_vlan.tag &&
+                   !dev_flow->handle.vf_vlan.created) {
                        /*
                         * The rule contains the VLAN pattern.
                         * For VF we are going to create VLAN
                         * interface to make hypervisor set correct
                         * e-Switch vport context.
                         */
-                       mlx5_vlan_vmwa_acquire(dev, &dev_flow->verbs.vf_vlan);
+                       mlx5_vlan_vmwa_acquire(dev, &dev_flow->handle.vf_vlan);
                }
        }
        return 0;
 error:
        err = rte_errno; /* Save rte_errno before cleanup. */
        LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
-               verbs = &dev_flow->verbs;
-               if (verbs->hrxq) {
-                       if (dev_flow->actions & MLX5_FLOW_ACTION_DROP)
+               dh = &dev_flow->handle;
+               if (dh->hrxq) {
+                       if (dev_flow->handle.act_flags & MLX5_FLOW_ACTION_DROP)
                                mlx5_hrxq_drop_release(dev);
                        else
-                               mlx5_hrxq_release(dev, verbs->hrxq);
-                       verbs->hrxq = NULL;
-               }
-               if (dev_flow->verbs.vf_vlan.tag &&
-                   dev_flow->verbs.vf_vlan.created) {
-                       mlx5_vlan_vmwa_release(dev, &dev_flow->verbs.vf_vlan);
+                               mlx5_hrxq_release(dev, dh->hrxq);
+                       dh->hrxq = NULL;
                }
+               if (dh->vf_vlan.tag && dh->vf_vlan.created)
+                       mlx5_vlan_vmwa_release(dev, &dh->vf_vlan);
        }
        rte_errno = err; /* Restore rte_errno. */
        return -rte_errno;