[dpdk-dev] [PATCH v6 14/15] net/mlx5: allow multiple flow tables on the same level

Jiawei Wang jiaweiw at nvidia.com
Tue Apr 20 12:55:21 CEST 2021


From: Li Zhang <lizh at nvidia.com>

The driver devices support creation of multiple flow tables.
Jump action can be used in order to move the packet steering
to different flow table.
Table 0 is always the root table for packet steering.

Jumping between tables may cause endless loops in steering mechanism,
that's why each table has level attribute,
the driver sub-system may not allow jumping to table with
equal or lower level than the current table.

Currently, in the driver, the table ID and level are always identical.

Allow multiple flow table creation with the same level attribute.

This patch adds the table id in flow table data entry, while
allocates the flow table, if the table level is same but the
different table id, the new table will be allocated with new
table object id. It supports 4M multiple flow tables on the
same level.

Signed-off-by: Li Zhang <lizh at nvidia.com>
Acked-by: Matan Azrad <matan at nvidia.com>
---
 drivers/net/mlx5/mlx5.c         |  9 +++--
 drivers/net/mlx5/mlx5.h         | 10 +++--
 drivers/net/mlx5/mlx5_flow.c    | 17 ++++++---
 drivers/net/mlx5/mlx5_flow.h    | 12 ++++--
 drivers/net/mlx5/mlx5_flow_dv.c | 84 +++++++++++++++++++++++------------------
 5 files changed, 79 insertions(+), 53 deletions(-)

diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index e5ffb8f..7e13b38 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -1271,9 +1271,12 @@ struct mlx5_dev_ctx_shared *
 	 * because DV expect to see them even if they cannot be created by
 	 * RDMA-CORE.
 	 */
-	if (!flow_dv_tbl_resource_get(dev, 0, 0, 0, 0, NULL, 0, 1, &error) ||
-	    !flow_dv_tbl_resource_get(dev, 0, 1, 0, 0, NULL, 0, 1, &error) ||
-	    !flow_dv_tbl_resource_get(dev, 0, 0, 1, 0, NULL, 0, 1, &error)) {
+	if (!flow_dv_tbl_resource_get(dev, 0, 0, 0, 0,
+		NULL, 0, 1, 0, &error) ||
+	    !flow_dv_tbl_resource_get(dev, 0, 1, 0, 0,
+		NULL, 0, 1, 0, &error) ||
+	    !flow_dv_tbl_resource_get(dev, 0, 0, 1, 0,
+		NULL, 0, 1, 0, &error)) {
 		err = ENOMEM;
 		goto error;
 	}
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index b2eb851..55ff17a 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -779,10 +779,12 @@ struct mlx5_aso_mtr_pools_mng {
 union mlx5_flow_tbl_key {
 	struct {
 		/* Table ID should be at the lowest address. */
-		uint32_t table_id;	/**< ID of the table. */
-		uint16_t dummy;		/**< Dummy table for DV API. */
-		uint8_t domain;		/**< 1 - FDB, 0 - NIC TX/RX. */
-		uint8_t direction;	/**< 1 - egress, 0 - ingress. */
+		uint32_t level;	/**< Level of the table. */
+		uint32_t id:22;	/**< ID of the table. */
+		uint32_t dummy:1;	/**< Dummy table for DV API. */
+		uint32_t is_fdb:1;	/**< 1 - FDB, 0 - NIC TX/RX. */
+		uint32_t is_egress:1;	/**< 1 - egress, 0 - ingress. */
+		uint32_t reserved:7;	/**< must be zero for comparison. */
 	};
 	uint64_t v64;			/**< full 64bits value of key */
 };
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index cf93ab0..bed8f31 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -4342,6 +4342,9 @@ struct mlx5_hlist_entry *
 		dev_flow->handle->mark = 1;
 	if (sub_flow)
 		*sub_flow = dev_flow;
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+	dev_flow->dv.table_id = flow_split_info->table_id;
+#endif
 	return flow_drv_translate(dev, dev_flow, attr, items, actions, error);
 }
 
@@ -5497,8 +5500,7 @@ struct mlx5_hlist_entry *
 						struct mlx5_flow_tbl_data_entry,
 						tbl);
 			sfx_attr.group = sfx_attr.transfer ?
-						(sfx_tbl_data->table_id - 1) :
-						sfx_tbl_data->table_id;
+			(sfx_tbl_data->level - 1) : sfx_tbl_data->level;
 		} else {
 			MLX5_ASSERT(attr->transfer);
 			sfx_attr.group = jump_table;
@@ -5698,7 +5700,8 @@ struct mlx5_hlist_entry *
 		.skip_scale = 0,
 		.flow_idx = 0,
 		.prefix_mark = 0,
-		.prefix_layers = 0
+		.prefix_layers = 0,
+		.table_id = 0
 	};
 	int ret;
 
@@ -7784,10 +7787,12 @@ int rte_pmd_mlx5_sync_flow(uint16_t port_id, uint32_t domains)
 	union tunnel_offload_mark mbits = { .val = mark };
 	union mlx5_flow_tbl_key table_key = {
 		{
-			.table_id = tunnel_id_to_flow_tbl(mbits.table_id),
+			.level = tunnel_id_to_flow_tbl(mbits.table_id),
+			.id = 0,
+			.reserved = 0,
 			.dummy = 0,
-			.domain = !!mbits.transfer,
-			.direction = 0,
+			.is_fdb = !!mbits.transfer,
+			.is_egress = 0,
 		}
 	};
 	he = mlx5_hlist_lookup(sh->flow_tbls, table_key.v64, NULL);
diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
index 81cf1b5..1a74b17 100644
--- a/drivers/net/mlx5/mlx5_flow.h
+++ b/drivers/net/mlx5/mlx5_flow.h
@@ -564,8 +564,9 @@ struct mlx5_flow_tbl_data_entry {
 	uint32_t is_egress:1; /**< Egress table. */
 	uint32_t is_transfer:1; /**< Transfer table. */
 	uint32_t dummy:1; /**<  DR table. */
-	uint32_t reserve:27; /**< Reserved to future using. */
-	uint32_t table_id; /**< Table ID. */
+	uint32_t id:22; /**< Table ID. */
+	uint32_t reserve:5; /**< Reserved to future using. */
+	uint32_t level; /**< Table level. */
 };
 
 /* Sub rdma-core actions list. */
@@ -699,6 +700,7 @@ struct mlx5_flow_handle {
 /** Device flow structure only for DV flow creation. */
 struct mlx5_flow_dv_workspace {
 	uint32_t group; /**< The group index. */
+	uint32_t table_id; /**< Flow table identifier. */
 	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. */
@@ -1059,6 +1061,7 @@ struct mlx5_flow_split_info {
 	uint32_t flow_idx; /**< This memory pool index to the flow. */
 	uint32_t prefix_mark; /**< Prefix subflow mark flag. */
 	uint64_t prefix_layers; /**< Prefix subflow layers. */
+	uint32_t table_id; /**< Flow table identifier. */
 };
 
 typedef int (*mlx5_flow_validate_t)(struct rte_eth_dev *dev,
@@ -1410,9 +1413,10 @@ int flow_dv_tbl_match_cb(struct mlx5_hlist *list,
 void flow_dv_tbl_remove_cb(struct mlx5_hlist *list,
 			   struct mlx5_hlist_entry *entry);
 struct mlx5_flow_tbl_resource *flow_dv_tbl_resource_get(struct rte_eth_dev *dev,
-		uint32_t table_id, uint8_t egress, uint8_t transfer,
+		uint32_t table_level, uint8_t egress, uint8_t transfer,
 		bool external, const struct mlx5_flow_tunnel *tunnel,
-		uint32_t group_id, uint8_t dummy, struct rte_flow_error *error);
+		uint32_t group_id, uint8_t dummy,
+		uint32_t table_id, struct rte_flow_error *error);
 
 struct mlx5_hlist_entry *flow_dv_tag_create_cb(struct mlx5_hlist *list,
 					       uint64_t key, void *cb_ctx);
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index a649339..ffd9f63 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -9373,20 +9373,21 @@ struct mlx5_hlist_entry *
 	tbl_data->group_id = tt_prm->group_id;
 	tbl_data->external = !!tt_prm->external;
 	tbl_data->tunnel_offload = is_tunnel_offload_active(dev);
-	tbl_data->is_egress = !!key.direction;
-	tbl_data->is_transfer = !!key.domain;
+	tbl_data->is_egress = !!key.is_egress;
+	tbl_data->is_transfer = !!key.is_fdb;
 	tbl_data->dummy = !!key.dummy;
-	tbl_data->table_id = key.table_id;
+	tbl_data->level = key.level;
+	tbl_data->id = key.id;
 	tbl = &tbl_data->tbl;
 	if (key.dummy)
 		return &tbl_data->entry;
-	if (key.domain)
+	if (key.is_fdb)
 		domain = sh->fdb_domain;
-	else if (key.direction)
+	else if (key.is_egress)
 		domain = sh->tx_domain;
 	else
 		domain = sh->rx_domain;
-	ret = mlx5_flow_os_create_flow_tbl(domain, key.table_id, &tbl->obj);
+	ret = mlx5_flow_os_create_flow_tbl(domain, key.level, &tbl->obj);
 	if (ret) {
 		rte_flow_error_set(error, ENOMEM,
 				   RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
@@ -9394,7 +9395,7 @@ struct mlx5_hlist_entry *
 		mlx5_ipool_free(sh->ipool[MLX5_IPOOL_JUMP], idx);
 		return NULL;
 	}
-	if (key.table_id) {
+	if (key.level != 0) {
 		ret = mlx5_flow_os_create_flow_action_dest_flow_tbl
 					(tbl->obj, &tbl_data->jump.action);
 		if (ret) {
@@ -9407,9 +9408,9 @@ struct mlx5_hlist_entry *
 			return NULL;
 		}
 	}
-	MKSTR(matcher_name, "%s_%s_%u_matcher_cache",
-	      key.domain ? "FDB" : "NIC", key.direction ? "egress" : "ingress",
-	      key.table_id);
+	MKSTR(matcher_name, "%s_%s_%u_%u_matcher_cache",
+	      key.is_fdb ? "FDB" : "NIC", key.is_egress ? "egress" : "ingress",
+	      key.level, key.id);
 	mlx5_cache_list_init(&tbl_data->matchers, matcher_name, 0, sh,
 			     flow_dv_matcher_create_cb,
 			     flow_dv_matcher_match_cb,
@@ -9426,10 +9427,11 @@ struct mlx5_hlist_entry *
 		container_of(entry, struct mlx5_flow_tbl_data_entry, entry);
 	union mlx5_flow_tbl_key key = { .v64 = key64 };
 
-	return tbl_data->table_id != key.table_id ||
+	return tbl_data->level != key.level ||
+	       tbl_data->id != key.id ||
 	       tbl_data->dummy != key.dummy ||
-	       tbl_data->is_transfer != key.domain ||
-	       tbl_data->is_egress != key.direction;
+	       tbl_data->is_transfer != !!key.is_fdb ||
+	       tbl_data->is_egress != !!key.is_egress;
 }
 
 /**
@@ -9437,14 +9439,16 @@ struct mlx5_hlist_entry *
  *
  * @param[in, out] dev
  *   Pointer to rte_eth_dev structure.
- * @param[in] table_id
- *   Table id to use.
+ * @param[in] table_level
+ *   Table level to use.
  * @param[in] egress
  *   Direction of the table.
  * @param[in] transfer
  *   E-Switch or NIC flow.
  * @param[in] dummy
  *   Dummy entry for dv API.
+ * @param[in] table_id
+ *   Table id to use.
  * @param[out] error
  *   pointer to error structure.
  *
@@ -9453,20 +9457,23 @@ struct mlx5_hlist_entry *
  */
 struct mlx5_flow_tbl_resource *
 flow_dv_tbl_resource_get(struct rte_eth_dev *dev,
-			 uint32_t table_id, uint8_t egress,
+			 uint32_t table_level, uint8_t egress,
 			 uint8_t transfer,
 			 bool external,
 			 const struct mlx5_flow_tunnel *tunnel,
 			 uint32_t group_id, uint8_t dummy,
+			 uint32_t table_id,
 			 struct rte_flow_error *error)
 {
 	struct mlx5_priv *priv = dev->data->dev_private;
 	union mlx5_flow_tbl_key table_key = {
 		{
-			.table_id = table_id,
-			.dummy = dummy,
-			.domain = !!transfer,
-			.direction = !!egress,
+			.level = table_level,
+			.id = table_id,
+			.reserved = 0,
+			.dummy = !!dummy,
+			.is_fdb = !!transfer,
+			.is_egress = !!egress,
 		}
 	};
 	struct mlx5_flow_tbl_tunnel_prm tt_prm = {
@@ -9489,8 +9496,10 @@ struct mlx5_flow_tbl_resource *
 				   "cannot get table");
 		return NULL;
 	}
-	DRV_LOG(DEBUG, "Table_id %u tunnel %u group %u registered.",
-		table_id, tunnel ? tunnel->tunnel_id : 0, group_id);
+	DRV_LOG(DEBUG, "table_level %u table_id %u "
+		"tunnel %u group %u registered.",
+		table_level, table_id,
+		tunnel ? tunnel->tunnel_id : 0, group_id);
 	tbl_data = container_of(entry, struct mlx5_flow_tbl_data_entry, entry);
 	return &tbl_data->tbl;
 }
@@ -9517,7 +9526,7 @@ struct mlx5_flow_tbl_resource *
 					tbl_data->tunnel->tunnel_id : 0,
 			.group = tbl_data->group_id
 		};
-		uint32_t table_id = tbl_data->table_id;
+		uint32_t table_level = tbl_data->level;
 
 		tunnel_grp_hash = tbl_data->tunnel ?
 					tbl_data->tunnel->groups :
@@ -9526,8 +9535,9 @@ struct mlx5_flow_tbl_resource *
 		if (he)
 			mlx5_hlist_unregister(tunnel_grp_hash, he);
 		DRV_LOG(DEBUG,
-			"Table_id %u tunnel %u group %u released.",
-			table_id,
+			"table_level %u id %u tunnel %u group %u released.",
+			table_level,
+			tbl_data->id,
 			tbl_data->tunnel ?
 			tbl_data->tunnel->tunnel_id : 0,
 			tbl_data->group_id);
@@ -9655,10 +9665,10 @@ struct mlx5_cache_entry *
 	 * tunnel offload API requires this registration for cases when
 	 * tunnel match rule was inserted before tunnel set rule.
 	 */
-	tbl = flow_dv_tbl_resource_get(dev, key->table_id,
-				       key->direction, key->domain,
+	tbl = flow_dv_tbl_resource_get(dev, key->level,
+				       key->is_egress, key->is_fdb,
 				       dev_flow->external, tunnel,
-				       group_id, 0, error);
+				       group_id, 0, key->id, error);
 	if (!tbl)
 		return -rte_errno;	/* No need to refill the error info */
 	tbl_data = container_of(tbl, struct mlx5_flow_tbl_data_entry, tbl);
@@ -10137,7 +10147,7 @@ struct mlx5_cache_entry *
 		is_egress = 1;
 	tbl = flow_dv_tbl_resource_get(dev, next_ft_id,
 					is_egress, is_transfer,
-					true, NULL, 0, 0, error);
+					true, NULL, 0, 0, 0, error);
 	if (!tbl) {
 		rte_flow_error_set(error, ENOMEM,
 					  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
@@ -11490,7 +11500,7 @@ struct mlx5_cache_entry *
 						       attr->transfer,
 						       !!dev_flow->external,
 						       tunnel, jump_group, 0,
-						       error);
+						       0, error);
 			if (!tbl)
 				return rte_flow_error_set
 						(error, errno,
@@ -12032,9 +12042,10 @@ struct mlx5_cache_entry *
 	matcher.priority = mlx5_get_matcher_priority(dev, attr,
 					matcher.priority);
 	/* 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->dv.group;
+	tbl_key.is_fdb = attr->transfer;
+	tbl_key.is_egress = attr->egress;
+	tbl_key.level = dev_flow->dv.group;
+	tbl_key.id = dev_flow->dv.table_id;
 	if (flow_dv_matcher_register(dev, &matcher, &tbl_key, dev_flow,
 				     tunnel, attr->group, error))
 		return -rte_errno;
@@ -13641,7 +13652,7 @@ struct mlx5_cache_entry *
 	/* Create the meter table with METER level. */
 	dtb->tbl = flow_dv_tbl_resource_get(dev, MLX5_FLOW_TABLE_LEVEL_METER,
 					    egress, transfer, false, NULL, 0,
-					    0, &error);
+					    0, 0, &error);
 	if (!dtb->tbl) {
 		DRV_LOG(ERR, "Failed to create meter policer table.");
 		return -1;
@@ -13650,7 +13661,7 @@ struct mlx5_cache_entry *
 	dtb->sfx_tbl = flow_dv_tbl_resource_get(dev,
 					    MLX5_FLOW_TABLE_LEVEL_SUFFIX,
 					    egress, transfer, false, NULL, 0,
-					    0, &error);
+					    0, 0, &error);
 	if (!dtb->sfx_tbl) {
 		DRV_LOG(ERR, "Failed to create meter suffix table.");
 		return -1;
@@ -14173,7 +14184,8 @@ struct mlx5_cache_entry *
 	void *flow = NULL;
 	int ret = -1;
 
-	tbl = flow_dv_tbl_resource_get(dev, 0, 0, 0, false, NULL, 0, 0, NULL);
+	tbl = flow_dv_tbl_resource_get(dev, 0, 0, 0, false, NULL,
+					0, 0, 0, NULL);
 	if (!tbl)
 		goto err;
 	dcs = mlx5_devx_cmd_flow_counter_alloc(priv->sh->ctx, 0x4);
-- 
1.8.3.1



More information about the dev mailing list