DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5
@ 2020-11-13 14:52 Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation Gregory Etelson
                   ` (5 more replies)
  0 siblings, 6 replies; 16+ messages in thread
From: Gregory Etelson @ 2020-11-13 14:52 UTC (permalink / raw)
  To: dev; +Cc: getelson, matan, rasland

post merge fixes that restore tunnel offload functionality in mlx5.

v2: resolve compilation error in ubuntu inbox.

Gregory Etelson (5):
  net/mlx5: fix tunnel offload object allocation
  net/mlx5: fix tunnel offload hub multi-thread protection
  net/mlx5: fix double table referencing
  net/mlx5: fix tunnel offload callback names
  net/mlx5: fix non-dv compilation errors

 drivers/net/mlx5/linux/mlx5_os.c |   16 +-
 drivers/net/mlx5/mlx5.c          |   50 +-
 drivers/net/mlx5/mlx5.h          |    4 +-
 drivers/net/mlx5/mlx5_flow.c     | 1296 ++++++++++++++++++------------
 drivers/net/mlx5/mlx5_flow.h     |   23 +-
 drivers/net/mlx5/mlx5_flow_dv.c  |   39 +-
 6 files changed, 842 insertions(+), 586 deletions(-)

-- 
2.29.2


^ permalink raw reply	[flat|nested] 16+ messages in thread

* [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation
  2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
@ 2020-11-13 14:52 ` Gregory Etelson
  2020-11-13 19:00   ` Thomas Monjalon
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 2/5] net/mlx5: fix tunnel offload hub multi-thread protection Gregory Etelson
                   ` (4 subsequent siblings)
  5 siblings, 1 reply; 16+ messages in thread
From: Gregory Etelson @ 2020-11-13 14:52 UTC (permalink / raw)
  To: dev
  Cc: getelson, matan, rasland, Viacheslav Ovsiienko, Shahaf Shuler,
	Xueming Li

The original patch allocated tunnel offload objects with invalid
indexes. As the result, PMD tunnel object allocation failed.

In this patch indexed pool provides both an index and memory for a new
tunnel offload object.
Also tunnel offload ipool moved to dv enabled code only.

Fixes: f2e8093 ("net/mlx5: use indexed pool as id generator")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>2
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
---
 drivers/net/mlx5/mlx5.c      | 50 ++++++++++++++++++------------------
 drivers/net/mlx5/mlx5.h      |  4 +--
 drivers/net/mlx5/mlx5_flow.c | 41 ++++++++++-------------------
 3 files changed, 40 insertions(+), 55 deletions(-)

diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index 43344391df..e1faa819a3 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -186,7 +186,7 @@ static pthread_mutex_t mlx5_dev_ctx_list_mutex = PTHREAD_MUTEX_INITIALIZER;
 
 static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 #ifdef HAVE_IBV_FLOW_DV_SUPPORT
-	{
+	[MLX5_IPOOL_DECAP_ENCAP] = {
 		.size = sizeof(struct mlx5_flow_dv_encap_decap_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -197,7 +197,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_encap_decap_ipool",
 	},
-	{
+	[MLX5_IPOOL_PUSH_VLAN] = {
 		.size = sizeof(struct mlx5_flow_dv_push_vlan_action_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -208,7 +208,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_push_vlan_ipool",
 	},
-	{
+	[MLX5_IPOOL_TAG] = {
 		.size = sizeof(struct mlx5_flow_dv_tag_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -219,7 +219,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_tag_ipool",
 	},
-	{
+	[MLX5_IPOOL_PORT_ID] = {
 		.size = sizeof(struct mlx5_flow_dv_port_id_action_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -230,7 +230,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_port_id_ipool",
 	},
-	{
+	[MLX5_IPOOL_JUMP] = {
 		.size = sizeof(struct mlx5_flow_tbl_data_entry),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -241,7 +241,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_jump_ipool",
 	},
-	{
+	[MLX5_IPOOL_SAMPLE] = {
 		.size = sizeof(struct mlx5_flow_dv_sample_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -252,7 +252,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_sample_ipool",
 	},
-	{
+	[MLX5_IPOOL_DEST_ARRAY] = {
 		.size = sizeof(struct mlx5_flow_dv_dest_array_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -263,8 +263,19 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_dest_array_ipool",
 	},
+	[MLX5_IPOOL_TUNNEL_OFFLOAD] = {
+		.size = sizeof(struct mlx5_flow_tunnel),
+		.need_lock = 1,
+		.release_mem_en = 1,
+		.type = "mlx5_tunnel_offload",
+	},
+	[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID] = {
+		.size = 0,
+		.need_lock = 1,
+		.type = "mlx5_flow_tnl_tbl_ipool",
+	},
 #endif
-	{
+	[MLX5_IPOOL_MTR] = {
 		.size = sizeof(struct mlx5_flow_meter),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -275,7 +286,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_meter_ipool",
 	},
-	{
+	[MLX5_IPOOL_MCP] = {
 		.size = sizeof(struct mlx5_flow_mreg_copy_resource),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -286,7 +297,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_mcp_ipool",
 	},
-	{
+	[MLX5_IPOOL_HRXQ] = {
 		.size = (sizeof(struct mlx5_hrxq) + MLX5_RSS_HASH_KEY_LEN),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -297,7 +308,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_hrxq_ipool",
 	},
-	{
+	[MLX5_IPOOL_MLX5_FLOW] = {
 		/*
 		 * MLX5_IPOOL_MLX5_FLOW size varies for DV and VERBS flows.
 		 * It set in run time according to PCI function configuration.
@@ -312,7 +323,7 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_flow_handle_ipool",
 	},
-	{
+	[MLX5_IPOOL_RTE_FLOW] = {
 		.size = sizeof(struct rte_flow),
 		.trunk_size = 4096,
 		.need_lock = 1,
@@ -321,22 +332,12 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "rte_flow_ipool",
 	},
-	{
+	[MLX5_IPOOL_RSS_EXPANTION_FLOW_ID] = {
 		.size = 0,
 		.need_lock = 1,
 		.type = "mlx5_flow_rss_id_ipool",
 	},
-	{
-		.size = 0,
-		.need_lock = 1,
-		.type = "mlx5_flow_tnl_flow_ipool",
-	},
-	{
-		.size = 0,
-		.need_lock = 1,
-		.type = "mlx5_flow_tnl_tbl_ipool",
-	},
-	{
+	[MLX5_IPOOL_RSS_SHARED_ACTIONS] = {
 		.size = sizeof(struct mlx5_shared_action_rss),
 		.trunk_size = 64,
 		.grow_trunk = 3,
@@ -347,7 +348,6 @@ static const struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = {
 		.free = mlx5_free,
 		.type = "mlx5_shared_action_rss",
 	},
-
 };
 
 
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index 7ee63a7a14..af097d6a7e 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -44,6 +44,8 @@ enum mlx5_ipool_index {
 	MLX5_IPOOL_JUMP, /* Pool for jump resource. */
 	MLX5_IPOOL_SAMPLE, /* Pool for sample resource. */
 	MLX5_IPOOL_DEST_ARRAY, /* Pool for destination array resource. */
+	MLX5_IPOOL_TUNNEL_OFFLOAD, /* Pool for tunnel offload context */
+	MLX5_IPOOL_TUNNEL_FLOW_TBL_ID, /* Pool for tunnel table ID. */
 #endif
 	MLX5_IPOOL_MTR, /* Pool for meter resource. */
 	MLX5_IPOOL_MCP, /* Pool for metadata resource. */
@@ -51,8 +53,6 @@ enum mlx5_ipool_index {
 	MLX5_IPOOL_MLX5_FLOW, /* Pool for mlx5 flow handle. */
 	MLX5_IPOOL_RTE_FLOW, /* Pool for rte_flow. */
 	MLX5_IPOOL_RSS_EXPANTION_FLOW_ID, /* Pool for Queue/RSS flow ID. */
-	MLX5_IPOOL_TUNNEL_ID, /* Pool for flow tunnel ID. */
-	MLX5_IPOOL_TNL_TBL_ID, /* Pool for tunnel table ID. */
 	MLX5_IPOOL_RSS_SHARED_ACTIONS, /* Pool for RSS shared actions. */
 	MLX5_IPOOL_MAX,
 };
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 92adfcacca..31c9d82b4a 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -6934,7 +6934,7 @@ mlx5_flow_tunnel_grp2tbl_remove_cb(struct mlx5_hlist *list,
 	struct mlx5_dev_ctx_shared *sh = list->ctx;
 	struct tunnel_tbl_entry *tte = container_of(entry, typeof(*tte), hash);
 
-	mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TNL_TBL_ID],
+	mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
 			tunnel_flow_tbl_to_id(tte->flow_table));
 	mlx5_free(tte);
 }
@@ -6952,12 +6952,12 @@ mlx5_flow_tunnel_grp2tbl_create_cb(struct mlx5_hlist *list,
 			  SOCKET_ID_ANY);
 	if (!tte)
 		goto err;
-	mlx5_ipool_malloc(sh->ipool[MLX5_IPOOL_TNL_TBL_ID],
+	mlx5_ipool_malloc(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
 			  &tte->flow_table);
 	if (tte->flow_table >= MLX5_MAX_TABLES) {
 		DRV_LOG(ERR, "Tunnel TBL ID %d exceed max limit.",
 			tte->flow_table);
-		mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TNL_TBL_ID],
+		mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
 				tte->flow_table);
 		goto err;
 	} else if (!tte->flow_table) {
@@ -7465,14 +7465,13 @@ mlx5_flow_tunnel_free(struct rte_eth_dev *dev,
 		      struct mlx5_flow_tunnel *tunnel)
 {
 	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_indexed_pool *ipool;
 
 	DRV_LOG(DEBUG, "port %u release pmd tunnel id=0x%x",
 		dev->data->port_id, tunnel->tunnel_id);
-	RTE_VERIFY(!__atomic_load_n(&tunnel->refctn, __ATOMIC_RELAXED));
-	mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_TUNNEL_ID],
-			tunnel->tunnel_id);
 	mlx5_hlist_destroy(tunnel->groups);
-	mlx5_free(tunnel);
+	ipool = priv->sh->ipool[MLX5_IPOOL_TUNNEL_OFFLOAD];
+	mlx5_ipool_free(ipool, tunnel->tunnel_id);
 }
 
 static struct mlx5_flow_tunnel *
@@ -7494,39 +7493,25 @@ mlx5_flow_tunnel_allocate(struct rte_eth_dev *dev,
 			  const struct rte_flow_tunnel *app_tunnel)
 {
 	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_indexed_pool *ipool;
 	struct mlx5_flow_tunnel *tunnel;
 	uint32_t id;
 
-	mlx5_ipool_malloc(priv->sh->ipool[MLX5_IPOOL_RSS_EXPANTION_FLOW_ID],
-			  &id);
+	ipool = priv->sh->ipool[MLX5_IPOOL_TUNNEL_OFFLOAD];
+	tunnel = mlx5_ipool_zmalloc(ipool, &id);
+	if (!tunnel)
+		return NULL;
 	if (id >= MLX5_MAX_TUNNELS) {
-		mlx5_ipool_free(priv->sh->ipool
-				[MLX5_IPOOL_RSS_EXPANTION_FLOW_ID], id);
+		mlx5_ipool_free(ipool, id);
 		DRV_LOG(ERR, "Tunnel ID %d exceed max limit.", id);
 		return NULL;
-	} else if (!id) {
-		return NULL;
-	}
-	/**
-	 * mlx5 flow tunnel is an auxlilary data structure
-	 * It's not part of IO. No need to allocate it from
-	 * huge pages pools dedicated for IO
-	 */
-	tunnel = mlx5_malloc(MLX5_MEM_SYS | MLX5_MEM_ZERO, sizeof(*tunnel),
-			     0, SOCKET_ID_ANY);
-	if (!tunnel) {
-		mlx5_ipool_free(priv->sh->ipool
-				[MLX5_IPOOL_RSS_EXPANTION_FLOW_ID], id);
-		return NULL;
 	}
 	tunnel->groups = mlx5_hlist_create("tunnel groups", 1024, 0, 0,
 					   mlx5_flow_tunnel_grp2tbl_create_cb,
 					   NULL,
 					   mlx5_flow_tunnel_grp2tbl_remove_cb);
 	if (!tunnel->groups) {
-		mlx5_ipool_free(priv->sh->ipool
-				[MLX5_IPOOL_RSS_EXPANTION_FLOW_ID], id);
-		mlx5_free(tunnel);
+		mlx5_ipool_free(ipool, id);
 		return NULL;
 	}
 	tunnel->groups->ctx = priv->sh;
-- 
2.29.2


^ permalink raw reply	[flat|nested] 16+ messages in thread

* [dpdk-dev] [PATCH v2 2/5] net/mlx5: fix tunnel offload hub multi-thread protection
  2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation Gregory Etelson
@ 2020-11-13 14:52 ` Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 3/5] net/mlx5: fix double table referencing Gregory Etelson
                   ` (3 subsequent siblings)
  5 siblings, 0 replies; 16+ messages in thread
From: Gregory Etelson @ 2020-11-13 14:52 UTC (permalink / raw)
  To: dev
  Cc: getelson, matan, rasland, Viacheslav Ovsiienko, Shahaf Shuler,
	Suanming Mou

The original patch was removing active tunnel offload objects from a
tunnels db list without checking its reference counter value.
That action was leading to a PMD crash.

Current patch isolates tunnels db list into a separate API. That API
manages MT protection of the tunnel offload db.

Fixes: e4f5880 ("net/mlx5: make tunnel hub list thread safe")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
---
 drivers/net/mlx5/mlx5_flow.c | 256 +++++++++++++++++++++++++----------
 drivers/net/mlx5/mlx5_flow.h |   6 +-
 2 files changed, 192 insertions(+), 70 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 31c9d82b4a..2f01e34033 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -33,6 +33,14 @@
 #include "mlx5_common_os.h"
 #include "rte_pmd_mlx5.h"
 
+static bool
+mlx5_access_tunnel_offload_db
+	(struct rte_eth_dev *dev,
+	 bool (*match)(struct rte_eth_dev *,
+		       struct mlx5_flow_tunnel *, const void *),
+	 void (*hit)(struct rte_eth_dev *, struct mlx5_flow_tunnel *, void *),
+	 void (*miss)(struct rte_eth_dev *, void *),
+	 void *ctx, bool lock_op);
 static struct mlx5_flow_tunnel *
 mlx5_find_tunnel_id(struct rte_eth_dev *dev, uint32_t id);
 static void
@@ -661,29 +669,68 @@ mlx5_flow_tunnel_match(struct rte_eth_dev *dev,
 	return 0;
 }
 
+struct tunnel_db_element_release_ctx {
+	struct rte_flow_item *items;
+	struct rte_flow_action *actions;
+	uint32_t num_elements;
+	struct rte_flow_error *error;
+	int ret;
+};
+
+static bool
+tunnel_element_release_match(struct rte_eth_dev *dev,
+			     struct mlx5_flow_tunnel *tunnel, const void *x)
+{
+	const struct tunnel_db_element_release_ctx *ctx = x;
+
+	RTE_SET_USED(dev);
+	if (ctx->num_elements != 1)
+		return false;
+	else if (ctx->items)
+		return ctx->items == &tunnel->item;
+	else if (ctx->actions)
+		return ctx->actions == &tunnel->action;
+
+	return false;
+}
+
+static void
+tunnel_element_release_hit(struct rte_eth_dev *dev,
+			   struct mlx5_flow_tunnel *tunnel, void *x)
+{
+	struct tunnel_db_element_release_ctx *ctx = x;
+	ctx->ret = 0;
+	if (!__atomic_sub_fetch(&tunnel->refctn, 1, __ATOMIC_RELAXED))
+		mlx5_flow_tunnel_free(dev, tunnel);
+}
+
+static void
+tunnel_element_release_miss(struct rte_eth_dev *dev, void *x)
+{
+	struct tunnel_db_element_release_ctx *ctx = x;
+	RTE_SET_USED(dev);
+	ctx->ret = rte_flow_error_set(ctx->error, EINVAL,
+				      RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+				      "invalid argument");
+}
+
 static int
 mlx5_flow_item_release(struct rte_eth_dev *dev,
 		       struct rte_flow_item *pmd_items,
 		       uint32_t num_items, struct rte_flow_error *err)
 {
-	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_flow_tunnel *tun;
+	struct tunnel_db_element_release_ctx ctx = {
+		.items = pmd_items,
+		.actions = NULL,
+		.num_elements = num_items,
+		.error = err,
+	};
 
-	rte_spinlock_lock(&thub->sl);
-	LIST_FOREACH(tun, &thub->tunnels, chain) {
-		if (&tun->item == pmd_items) {
-			LIST_REMOVE(tun, chain);
-			break;
-		}
-	}
-	rte_spinlock_unlock(&thub->sl);
-	if (!tun || num_items != 1)
-		return rte_flow_error_set(err, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-					  "invalid argument");
-	if (!__atomic_sub_fetch(&tun->refctn, 1, __ATOMIC_RELAXED))
-		mlx5_flow_tunnel_free(dev, tun);
-	return 0;
+	mlx5_access_tunnel_offload_db(dev, tunnel_element_release_match,
+				      tunnel_element_release_hit,
+				      tunnel_element_release_miss, &ctx, false);
+
+	return ctx.ret;
 }
 
 static int
@@ -691,25 +738,18 @@ mlx5_flow_action_release(struct rte_eth_dev *dev,
 			 struct rte_flow_action *pmd_actions,
 			 uint32_t num_actions, struct rte_flow_error *err)
 {
-	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_flow_tunnel *tun;
+	struct tunnel_db_element_release_ctx ctx = {
+		.items = NULL,
+		.actions = pmd_actions,
+		.num_elements = num_actions,
+		.error = err,
+	};
 
-	rte_spinlock_lock(&thub->sl);
-	LIST_FOREACH(tun, &thub->tunnels, chain) {
-		if (&tun->action == pmd_actions) {
-			LIST_REMOVE(tun, chain);
-			break;
-		}
-	}
-	rte_spinlock_unlock(&thub->sl);
-	if (!tun || num_actions != 1)
-		return rte_flow_error_set(err, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-					  "invalid argument");
-	if (!__atomic_sub_fetch(&tun->refctn, 1, __ATOMIC_RELAXED))
-		mlx5_flow_tunnel_free(dev, tun);
+	mlx5_access_tunnel_offload_db(dev, tunnel_element_release_match,
+				      tunnel_element_release_hit,
+				      tunnel_element_release_miss, &ctx, false);
 
-	return 0;
+	return ctx.ret;
 }
 
 static int
@@ -5889,11 +5929,8 @@ flow_list_destroy(struct rte_eth_dev *dev, uint32_t *list,
 	if (flow->tunnel) {
 		struct mlx5_flow_tunnel *tunnel;
 
-		rte_spinlock_lock(&mlx5_tunnel_hub(dev)->sl);
 		tunnel = mlx5_find_tunnel_id(dev, flow->tunnel_id);
 		RTE_VERIFY(tunnel);
-		LIST_REMOVE(tunnel, chain);
-		rte_spinlock_unlock(&mlx5_tunnel_hub(dev)->sl);
 		if (!__atomic_sub_fetch(&tunnel->refctn, 1, __ATOMIC_RELAXED))
 			mlx5_flow_tunnel_free(dev, tunnel);
 	}
@@ -7464,28 +7501,87 @@ static void
 mlx5_flow_tunnel_free(struct rte_eth_dev *dev,
 		      struct mlx5_flow_tunnel *tunnel)
 {
+	/* no tunnel hub spinlock protection */
 	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
 	struct mlx5_indexed_pool *ipool;
 
 	DRV_LOG(DEBUG, "port %u release pmd tunnel id=0x%x",
 		dev->data->port_id, tunnel->tunnel_id);
+	rte_spinlock_lock(&thub->sl);
+	LIST_REMOVE(tunnel, chain);
+	rte_spinlock_unlock(&thub->sl);
 	mlx5_hlist_destroy(tunnel->groups);
 	ipool = priv->sh->ipool[MLX5_IPOOL_TUNNEL_OFFLOAD];
 	mlx5_ipool_free(ipool, tunnel->tunnel_id);
 }
 
-static struct mlx5_flow_tunnel *
-mlx5_find_tunnel_id(struct rte_eth_dev *dev, uint32_t id)
+static bool
+mlx5_access_tunnel_offload_db
+	(struct rte_eth_dev *dev,
+	 bool (*match)(struct rte_eth_dev *,
+		       struct mlx5_flow_tunnel *, const void *),
+	 void (*hit)(struct rte_eth_dev *, struct mlx5_flow_tunnel *, void *),
+	 void (*miss)(struct rte_eth_dev *, void *),
+	 void *ctx, bool lock_op)
 {
+	bool verdict = false;
 	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_flow_tunnel *tun;
+	struct mlx5_flow_tunnel *tunnel;
 
-	LIST_FOREACH(tun, &thub->tunnels, chain) {
-		if (tun->tunnel_id == id)
+	rte_spinlock_lock(&thub->sl);
+	LIST_FOREACH(tunnel, &thub->tunnels, chain) {
+		verdict = match(dev, tunnel, (const void *)ctx);
+		if (verdict)
 			break;
 	}
+	if (!lock_op)
+		rte_spinlock_unlock(&thub->sl);
+	if (verdict && hit)
+		hit(dev, tunnel, ctx);
+	if (!verdict && miss)
+		miss(dev, ctx);
+	if (lock_op)
+		rte_spinlock_unlock(&thub->sl);
 
-	return tun;
+	return verdict;
+}
+
+struct tunnel_db_find_tunnel_id_ctx {
+	uint32_t tunnel_id;
+	struct mlx5_flow_tunnel *tunnel;
+};
+
+static bool
+find_tunnel_id_match(struct rte_eth_dev *dev,
+		     struct mlx5_flow_tunnel *tunnel, const void *x)
+{
+	const struct tunnel_db_find_tunnel_id_ctx *ctx = x;
+
+	RTE_SET_USED(dev);
+	return tunnel->tunnel_id == ctx->tunnel_id;
+}
+
+static void
+find_tunnel_id_hit(struct rte_eth_dev *dev,
+		   struct mlx5_flow_tunnel *tunnel, void *x)
+{
+	struct tunnel_db_find_tunnel_id_ctx *ctx = x;
+	RTE_SET_USED(dev);
+	ctx->tunnel = tunnel;
+}
+
+static struct mlx5_flow_tunnel *
+mlx5_find_tunnel_id(struct rte_eth_dev *dev, uint32_t id)
+{
+	struct tunnel_db_find_tunnel_id_ctx ctx = {
+		.tunnel_id = id,
+	};
+
+	mlx5_access_tunnel_offload_db(dev, find_tunnel_id_match,
+				      find_tunnel_id_hit, NULL, &ctx, true);
+
+	return ctx.tunnel;
 }
 
 static struct mlx5_flow_tunnel *
@@ -7533,38 +7629,60 @@ mlx5_flow_tunnel_allocate(struct rte_eth_dev *dev,
 	return tunnel;
 }
 
+struct tunnel_db_get_tunnel_ctx {
+	const struct rte_flow_tunnel *app_tunnel;
+	struct mlx5_flow_tunnel *tunnel;
+};
+
+static bool get_tunnel_match(struct rte_eth_dev *dev,
+			     struct mlx5_flow_tunnel *tunnel, const void *x)
+{
+	const struct tunnel_db_get_tunnel_ctx *ctx = x;
+
+	RTE_SET_USED(dev);
+	return !memcmp(ctx->app_tunnel, &tunnel->app_tunnel,
+		       sizeof(*ctx->app_tunnel));
+}
+
+static void get_tunnel_hit(struct rte_eth_dev *dev,
+			   struct mlx5_flow_tunnel *tunnel, void *x)
+{
+	/* called under tunnel spinlock protection */
+	struct tunnel_db_get_tunnel_ctx *ctx = x;
+
+	RTE_SET_USED(dev);
+	tunnel->refctn++;
+	ctx->tunnel = tunnel;
+}
+
+static void get_tunnel_miss(struct rte_eth_dev *dev, void *x)
+{
+	/* called under tunnel spinlock protection */
+	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
+	struct tunnel_db_get_tunnel_ctx *ctx = x;
+
+	rte_spinlock_unlock(&thub->sl);
+	ctx->tunnel = mlx5_flow_tunnel_allocate(dev, ctx->app_tunnel);
+	ctx->tunnel->refctn = 1;
+	rte_spinlock_lock(&thub->sl);
+	if (ctx->tunnel)
+		LIST_INSERT_HEAD(&thub->tunnels, ctx->tunnel, chain);
+}
+
+
 static int
 mlx5_get_flow_tunnel(struct rte_eth_dev *dev,
 		     const struct rte_flow_tunnel *app_tunnel,
 		     struct mlx5_flow_tunnel **tunnel)
 {
-	int ret;
-	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_flow_tunnel *tun;
-
-	rte_spinlock_lock(&thub->sl);
-	LIST_FOREACH(tun, &thub->tunnels, chain) {
-		if (!memcmp(app_tunnel, &tun->app_tunnel,
-			    sizeof(*app_tunnel))) {
-			*tunnel = tun;
-			ret = 0;
-			break;
-		}
-	}
-	if (!tun) {
-		tun = mlx5_flow_tunnel_allocate(dev, app_tunnel);
-		if (tun) {
-			LIST_INSERT_HEAD(&thub->tunnels, tun, chain);
-			*tunnel = tun;
-		} else {
-			ret = -ENOMEM;
-		}
-	}
-	rte_spinlock_unlock(&thub->sl);
-	if (tun)
-		__atomic_add_fetch(&tun->refctn, 1, __ATOMIC_RELAXED);
+	struct tunnel_db_get_tunnel_ctx ctx = {
+		.app_tunnel = app_tunnel,
+	};
 
-	return ret;
+	mlx5_access_tunnel_offload_db(dev, get_tunnel_match, get_tunnel_hit,
+				      get_tunnel_miss, &ctx, true);
+	*tunnel = ctx.tunnel;
+	return ctx.tunnel ? 0 : -ENOMEM;
 }
 
 void mlx5_release_tunnel_hub(struct mlx5_dev_ctx_shared *sh, uint16_t port_id)
diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
index e3a5030785..bdf2c50090 100644
--- a/drivers/net/mlx5/mlx5_flow.h
+++ b/drivers/net/mlx5/mlx5_flow.h
@@ -950,8 +950,12 @@ struct mlx5_flow_tunnel {
 
 /** PMD tunnel related context */
 struct mlx5_flow_tunnel_hub {
+	/* Tunnels list
+	 * Access to the list MUST be MT protected
+	 */
 	LIST_HEAD(, mlx5_flow_tunnel) tunnels;
-	rte_spinlock_t sl;			/* Tunnel list spinlock. */
+	 /* protect access to the tunnels list */
+	rte_spinlock_t sl;
 	struct mlx5_hlist *groups;		/** non tunnel groups */
 };
 
-- 
2.29.2


^ permalink raw reply	[flat|nested] 16+ messages in thread

* [dpdk-dev] [PATCH v2 3/5] net/mlx5: fix double table referencing
  2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 2/5] net/mlx5: fix tunnel offload hub multi-thread protection Gregory Etelson
@ 2020-11-13 14:52 ` Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 4/5] net/mlx5: fix tunnel offload callback names Gregory Etelson
                   ` (2 subsequent siblings)
  5 siblings, 0 replies; 16+ messages in thread
From: Gregory Etelson @ 2020-11-13 14:52 UTC (permalink / raw)
  To: dev
  Cc: getelson, matan, rasland, Viacheslav Ovsiienko, Shahaf Shuler,
	Xueming Li

The new flow table resource management API triggered a PMD crash in
tunnel offload mode, when tunnel match flow rule was inserted before
tunnel set rule.

Reason for the crash was double flow table registration. The table was
registered by the tunnel offload code for the first time and once
more by PMD code, as part of general table processing. The table
counter was decremented only once during the rule destruction and
caused a resource leak that triggered the crash.

The patch updates PMD registration with tunnel offload parameters and
removes table registration in tunnel related code.

Fixes: 663ad57dabb2 ("net/mlx5: make flow table cache thread safe")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
---
 drivers/net/mlx5/mlx5_flow.c    | 16 ++++++++++----
 drivers/net/mlx5/mlx5_flow_dv.c | 39 +++++++++++++++++----------------
 2 files changed, 32 insertions(+), 23 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 2f01e34033..185b4ba51a 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -7024,7 +7024,15 @@ tunnel_flow_group_to_flow_table(struct rte_eth_dev *dev,
 	struct mlx5_hlist *group_hash;
 
 	group_hash = tunnel ? tunnel->groups : thub->groups;
-	he = mlx5_hlist_register(group_hash, key.val, NULL);
+	he = mlx5_hlist_lookup(group_hash, key.val, NULL);
+	if (!he) {
+		DRV_LOG(DEBUG, "port %u tunnel %u group=%u - generate table id",
+		dev->data->port_id, key.tunnel_id, group);
+		he = mlx5_hlist_register(group_hash, key.val, NULL);
+	} else {
+		DRV_LOG(DEBUG, "port %u tunnel %u group=%u - skip table id",
+		dev->data->port_id, key.tunnel_id, group);
+	}
 	if (!he)
 		return rte_flow_error_set(error, EINVAL,
 					  RTE_FLOW_ERROR_TYPE_ATTR_GROUP,
@@ -7032,8 +7040,8 @@ tunnel_flow_group_to_flow_table(struct rte_eth_dev *dev,
 					  "tunnel group index not supported");
 	tte = container_of(he, typeof(*tte), hash);
 	*table = tte->flow_table;
-	DRV_LOG(DEBUG, "port %u tunnel %u group=%#x table=%#x",
-		dev->data->port_id, key.tunnel_id, group, *table);
+	DRV_LOG(DEBUG, "port %u tunnel %u group=%u table=%u",
+	dev->data->port_id, key.tunnel_id, group, *table);
 	return 0;
 }
 
@@ -7114,7 +7122,7 @@ mlx5_flow_group_to_table(struct rte_eth_dev *dev,
 		standard_translation = true;
 	}
 	DRV_LOG(DEBUG,
-		"port %u group=%#x transfer=%d external=%d fdb_def_rule=%d translate=%s",
+		"port %u group=%u transfer=%d external=%d fdb_def_rule=%d translate=%s",
 		dev->data->port_id, group, grp_info.transfer,
 		grp_info.external, grp_info.fdb_def_rule,
 		standard_translation ? "STANDARD" : "TUNNEL");
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index 78c710fef9..95165980f4 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -8042,6 +8042,8 @@ flow_dv_tbl_resource_get(struct rte_eth_dev *dev,
 				   "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);
 	tbl_data = container_of(entry, struct mlx5_flow_tbl_data_entry, entry);
 	return &tbl_data->tbl;
 }
@@ -8080,7 +8082,7 @@ flow_dv_tbl_remove_cb(struct mlx5_hlist *list,
 		if (he)
 			mlx5_hlist_unregister(tunnel_grp_hash, he);
 		DRV_LOG(DEBUG,
-			"Table_id %#x tunnel %u group %u released.",
+			"Table_id %u tunnel %u group %u released.",
 			table_id,
 			tbl_data->tunnel ?
 			tbl_data->tunnel->tunnel_id : 0,
@@ -8192,6 +8194,8 @@ flow_dv_matcher_register(struct rte_eth_dev *dev,
 			 struct mlx5_flow_dv_matcher *ref,
 			 union mlx5_flow_tbl_key *key,
 			 struct mlx5_flow *dev_flow,
+			 const struct mlx5_flow_tunnel *tunnel,
+			 uint32_t group_id,
 			 struct rte_flow_error *error)
 {
 	struct mlx5_cache_entry *entry;
@@ -8203,8 +8207,14 @@ flow_dv_matcher_register(struct rte_eth_dev *dev,
 		.data = ref,
 	};
 
-	tbl = flow_dv_tbl_resource_get(dev, key->table_id, key->direction,
-				       key->domain, false, NULL, 0, 0, error);
+	/**
+	 * 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,
+				       dev_flow->external, tunnel,
+				       group_id, 0, 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);
@@ -9605,10 +9615,14 @@ flow_dv_translate(struct rte_eth_dev *dev,
 		/*
 		 * do not add decap action if match rule drops packet
 		 * HW rejects rules with decap & drop
+		 *
+		 * if tunnel match rule was inserted before matching tunnel set
+		 * rule flow table used in the match rule must be registered.
+		 * current implementation handles that in the
+		 * flow_dv_match_register() at the function end.
 		 */
 		bool add_decap = true;
 		const struct rte_flow_action *ptr = actions;
-		struct mlx5_flow_tbl_resource *tbl;
 
 		for (; ptr->type != RTE_FLOW_ACTION_TYPE_END; ptr++) {
 			if (ptr->type == RTE_FLOW_ACTION_TYPE_DROP) {
@@ -9625,20 +9639,6 @@ flow_dv_translate(struct rte_eth_dev *dev,
 					dev_flow->dv.encap_decap->action;
 			action_flags |= MLX5_FLOW_ACTION_DECAP;
 		}
-		/*
-		 * bind table_id with <group, table> for tunnel match rule.
-		 * Tunnel set rule establishes that bind in JUMP action handler.
-		 * Required for scenario when application creates tunnel match
-		 * rule before tunnel set rule.
-		 */
-		tbl = flow_dv_tbl_resource_get(dev, table, attr->egress,
-					       attr->transfer,
-					       !!dev_flow->external, tunnel,
-					       attr->group, 0, error);
-		if (!tbl)
-			return rte_flow_error_set
-			       (error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
-			       actions, "cannot register tunnel group");
 	}
 	for (; !actions_end ; actions++) {
 		const struct rte_flow_action_queue *queue;
@@ -10468,7 +10468,8 @@ flow_dv_translate(struct rte_eth_dev *dev,
 	tbl_key.domain = attr->transfer;
 	tbl_key.direction = attr->egress;
 	tbl_key.table_id = dev_flow->dv.group;
-	if (flow_dv_matcher_register(dev, &matcher, &tbl_key, dev_flow, error))
+	if (flow_dv_matcher_register(dev, &matcher, &tbl_key, dev_flow,
+				     tunnel, attr->group, error))
 		return -rte_errno;
 	return 0;
 }
-- 
2.29.2


^ permalink raw reply	[flat|nested] 16+ messages in thread

* [dpdk-dev] [PATCH v2 4/5] net/mlx5: fix tunnel offload callback names
  2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
                   ` (2 preceding siblings ...)
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 3/5] net/mlx5: fix double table referencing Gregory Etelson
@ 2020-11-13 14:52 ` Gregory Etelson
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors Gregory Etelson
  2020-11-13 19:03 ` [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Thomas Monjalon
  5 siblings, 0 replies; 16+ messages in thread
From: Gregory Etelson @ 2020-11-13 14:52 UTC (permalink / raw)
  To: dev; +Cc: getelson, matan, rasland, Viacheslav Ovsiienko, Shahaf Shuler

Fix mlx5_flow_tunnel_action_release and mlx5_flow_tunnel_item_release
callback names to match tunnel offload names pattern.

Fixes: 4ec6360de37d ("net/mlx5: implement tunnel offload")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
---
 drivers/net/mlx5/mlx5_flow.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 185b4ba51a..358a5f4e72 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -715,9 +715,9 @@ tunnel_element_release_miss(struct rte_eth_dev *dev, void *x)
 }
 
 static int
-mlx5_flow_item_release(struct rte_eth_dev *dev,
-		       struct rte_flow_item *pmd_items,
-		       uint32_t num_items, struct rte_flow_error *err)
+mlx5_flow_tunnel_item_release(struct rte_eth_dev *dev,
+			      struct rte_flow_item *pmd_items,
+			      uint32_t num_items, struct rte_flow_error *err)
 {
 	struct tunnel_db_element_release_ctx ctx = {
 		.items = pmd_items,
@@ -734,9 +734,10 @@ mlx5_flow_item_release(struct rte_eth_dev *dev,
 }
 
 static int
-mlx5_flow_action_release(struct rte_eth_dev *dev,
-			 struct rte_flow_action *pmd_actions,
-			 uint32_t num_actions, struct rte_flow_error *err)
+mlx5_flow_tunnel_action_release(struct rte_eth_dev *dev,
+				struct rte_flow_action *pmd_actions,
+				uint32_t num_actions,
+				struct rte_flow_error *err)
 {
 	struct tunnel_db_element_release_ctx ctx = {
 		.items = NULL,
@@ -800,8 +801,8 @@ static const struct rte_flow_ops mlx5_flow_ops = {
 	.shared_action_query = mlx5_shared_action_query,
 	.tunnel_decap_set = mlx5_flow_tunnel_decap_set,
 	.tunnel_match = mlx5_flow_tunnel_match,
-	.tunnel_action_decap_release = mlx5_flow_action_release,
-	.tunnel_item_release = mlx5_flow_item_release,
+	.tunnel_action_decap_release = mlx5_flow_tunnel_action_release,
+	.tunnel_item_release = mlx5_flow_tunnel_item_release,
 	.get_restore_info = mlx5_flow_tunnel_get_restore_info,
 };
 
-- 
2.29.2


^ permalink raw reply	[flat|nested] 16+ messages in thread

* [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
                   ` (3 preceding siblings ...)
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 4/5] net/mlx5: fix tunnel offload callback names Gregory Etelson
@ 2020-11-13 14:52 ` Gregory Etelson
  2020-11-14 12:06   ` Thomas Monjalon
  2020-11-13 19:03 ` [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Thomas Monjalon
  5 siblings, 1 reply; 16+ messages in thread
From: Gregory Etelson @ 2020-11-13 14:52 UTC (permalink / raw)
  To: dev; +Cc: getelson, matan, rasland, Viacheslav Ovsiienko, Shahaf Shuler

ipools used for tunnel offload are restricted to dv enabled code only
since 1d1248d452ff, while the rest of the tunnel offload API did not
have compilation limitations. As the results, PMD compilation failed
on non-dv setups.

Current patch groups tunnel offload code in dv enabled code area and
provides stubs for calls that have to be visible in non-dv
environment. These stabs will notify caller that tunnel offload
functionality is not supported on that setup.

Fixes: 1d1248d452ff ("net/mlx5: fix offloaded tunnel allocation")
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
---
 drivers/net/mlx5/linux/mlx5_os.c |   16 +-
 drivers/net/mlx5/mlx5_flow.c     | 1104 +++++++++++++++++-------------
 drivers/net/mlx5/mlx5_flow.h     |   17 +-
 3 files changed, 638 insertions(+), 499 deletions(-)

diff --git a/drivers/net/mlx5/linux/mlx5_os.c b/drivers/net/mlx5/linux/mlx5_os.c
index c78d56fae3..1aee481bd7 100644
--- a/drivers/net/mlx5/linux/mlx5_os.c
+++ b/drivers/net/mlx5/linux/mlx5_os.c
@@ -301,6 +301,12 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
 		goto error;
 	}
 	sh->encaps_decaps->ctx = sh;
+	if (!sh->tunnel_hub)
+		err = mlx5_alloc_tunnel_hub(sh);
+	if (err) {
+		DRV_LOG(ERR, "mlx5_alloc_tunnel_hub failed err=%d", err);
+		goto error;
+	}
 #endif
 #ifdef HAVE_MLX5DV_DR
 	void *domain;
@@ -335,12 +341,6 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
 		sh->esw_drop_action = mlx5_glue->dr_create_flow_action_drop();
 	}
 #endif
-	if (!sh->tunnel_hub)
-		err = mlx5_alloc_tunnel_hub(sh);
-	if (err) {
-		DRV_LOG(ERR, "mlx5_alloc_tunnel_hub failed err=%d", err);
-		goto error;
-	}
 	if (priv->config.reclaim_mode == MLX5_RCM_AGGR) {
 		mlx5_glue->dr_reclaim_domain_memory(sh->rx_domain, 1);
 		mlx5_glue->dr_reclaim_domain_memory(sh->tx_domain, 1);
@@ -389,10 +389,12 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
 		mlx5_hlist_destroy(sh->tag_table);
 		sh->tag_table = NULL;
 	}
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
 	if (sh->tunnel_hub) {
 		mlx5_release_tunnel_hub(sh, priv->dev_port);
 		sh->tunnel_hub = NULL;
 	}
+#endif /* HAVE_IBV_FLOW_DV_SUPPORT */
 	mlx5_free_table_hash_list(priv);
 	return err;
 }
@@ -451,10 +453,12 @@ mlx5_os_free_shared_dr(struct mlx5_priv *priv)
 		mlx5_hlist_destroy(sh->tag_table);
 		sh->tag_table = NULL;
 	}
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
 	if (sh->tunnel_hub) {
 		mlx5_release_tunnel_hub(sh, priv->dev_port);
 		sh->tunnel_hub = NULL;
 	}
+#endif /* HAVE_IBV_FLOW_DV_SUPPORT */
 	mlx5_cache_list_destroy(&sh->port_id_action_list);
 	mlx5_cache_list_destroy(&sh->push_vlan_action_list);
 	mlx5_free_table_hash_list(priv);
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 358a5f4e72..11bc8e9dde 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -33,25 +33,34 @@
 #include "mlx5_common_os.h"
 #include "rte_pmd_mlx5.h"
 
-static bool
-mlx5_access_tunnel_offload_db
-	(struct rte_eth_dev *dev,
-	 bool (*match)(struct rte_eth_dev *,
-		       struct mlx5_flow_tunnel *, const void *),
-	 void (*hit)(struct rte_eth_dev *, struct mlx5_flow_tunnel *, void *),
-	 void (*miss)(struct rte_eth_dev *, void *),
-	 void *ctx, bool lock_op);
+struct tunnel_default_miss_ctx {
+	uint16_t *queue;
+	__extension__
+	union {
+		struct rte_flow_action_rss action_rss;
+		struct rte_flow_action_queue miss_queue;
+		struct rte_flow_action_jump miss_jump;
+		uint8_t raw[0];
+	};
+};
+
+static int
+flow_tunnel_add_default_miss(struct rte_eth_dev *dev,
+			     struct rte_flow *flow,
+			     const struct rte_flow_attr *attr,
+			     const struct rte_flow_action *app_actions,
+			     uint32_t flow_idx,
+			     struct tunnel_default_miss_ctx *ctx,
+			     struct rte_flow_error *error);
 static struct mlx5_flow_tunnel *
 mlx5_find_tunnel_id(struct rte_eth_dev *dev, uint32_t id);
 static void
 mlx5_flow_tunnel_free(struct rte_eth_dev *dev, struct mlx5_flow_tunnel *tunnel);
-static const struct mlx5_flow_tbl_data_entry  *
-tunnel_mark_decode(struct rte_eth_dev *dev, uint32_t mark);
-static int
-mlx5_get_flow_tunnel(struct rte_eth_dev *dev,
-		     const struct rte_flow_tunnel *app_tunnel,
-		     struct mlx5_flow_tunnel **tunnel);
-
+static uint32_t
+tunnel_flow_group_to_flow_table(struct rte_eth_dev *dev,
+				const struct mlx5_flow_tunnel *tunnel,
+				uint32_t group, uint32_t *table,
+				struct rte_flow_error *error);
 
 /** Device flow drivers. */
 extern const struct mlx5_flow_driver_ops mlx5_flow_verbs_drv_ops;
@@ -588,203 +597,32 @@ static int mlx5_shared_action_query
 				 const struct rte_flow_shared_action *action,
 				 void *data,
 				 struct rte_flow_error *error);
-static inline bool
-mlx5_flow_tunnel_validate(struct rte_eth_dev *dev,
-			  struct rte_flow_tunnel *tunnel,
-			  const char *err_msg)
-{
-	err_msg = NULL;
-	if (!is_tunnel_offload_active(dev)) {
-		err_msg = "tunnel offload was not activated";
-		goto out;
-	} else if (!tunnel) {
-		err_msg = "no application tunnel";
-		goto out;
-	}
-
-	switch (tunnel->type) {
-	default:
-		err_msg = "unsupported tunnel type";
-		goto out;
-	case RTE_FLOW_ITEM_TYPE_VXLAN:
-		break;
-	}
-
-out:
-	return !err_msg;
-}
-
-
 static int
 mlx5_flow_tunnel_decap_set(struct rte_eth_dev *dev,
-		    struct rte_flow_tunnel *app_tunnel,
-		    struct rte_flow_action **actions,
-		    uint32_t *num_of_actions,
-		    struct rte_flow_error *error)
-{
-	int ret;
-	struct mlx5_flow_tunnel *tunnel;
-	const char *err_msg = NULL;
-	bool verdict = mlx5_flow_tunnel_validate(dev, app_tunnel, err_msg);
-
-	if (!verdict)
-		return rte_flow_error_set(error, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_ACTION_CONF, NULL,
-					  err_msg);
-	ret = mlx5_get_flow_tunnel(dev, app_tunnel, &tunnel);
-	if (ret < 0) {
-		return rte_flow_error_set(error, ret,
-					  RTE_FLOW_ERROR_TYPE_ACTION_CONF, NULL,
-					  "failed to initialize pmd tunnel");
-	}
-	*actions = &tunnel->action;
-	*num_of_actions = 1;
-	return 0;
-}
-
+			   struct rte_flow_tunnel *app_tunnel,
+			   struct rte_flow_action **actions,
+			   uint32_t *num_of_actions,
+			   struct rte_flow_error *error);
 static int
 mlx5_flow_tunnel_match(struct rte_eth_dev *dev,
 		       struct rte_flow_tunnel *app_tunnel,
 		       struct rte_flow_item **items,
 		       uint32_t *num_of_items,
-		       struct rte_flow_error *error)
-{
-	int ret;
-	struct mlx5_flow_tunnel *tunnel;
-	const char *err_msg = NULL;
-	bool verdict = mlx5_flow_tunnel_validate(dev, app_tunnel, err_msg);
-
-	if (!verdict)
-		return rte_flow_error_set(error, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-					  err_msg);
-	ret = mlx5_get_flow_tunnel(dev, app_tunnel, &tunnel);
-	if (ret < 0) {
-		return rte_flow_error_set(error, ret,
-					  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-					  "failed to initialize pmd tunnel");
-	}
-	*items = &tunnel->item;
-	*num_of_items = 1;
-	return 0;
-}
-
-struct tunnel_db_element_release_ctx {
-	struct rte_flow_item *items;
-	struct rte_flow_action *actions;
-	uint32_t num_elements;
-	struct rte_flow_error *error;
-	int ret;
-};
-
-static bool
-tunnel_element_release_match(struct rte_eth_dev *dev,
-			     struct mlx5_flow_tunnel *tunnel, const void *x)
-{
-	const struct tunnel_db_element_release_ctx *ctx = x;
-
-	RTE_SET_USED(dev);
-	if (ctx->num_elements != 1)
-		return false;
-	else if (ctx->items)
-		return ctx->items == &tunnel->item;
-	else if (ctx->actions)
-		return ctx->actions == &tunnel->action;
-
-	return false;
-}
-
-static void
-tunnel_element_release_hit(struct rte_eth_dev *dev,
-			   struct mlx5_flow_tunnel *tunnel, void *x)
-{
-	struct tunnel_db_element_release_ctx *ctx = x;
-	ctx->ret = 0;
-	if (!__atomic_sub_fetch(&tunnel->refctn, 1, __ATOMIC_RELAXED))
-		mlx5_flow_tunnel_free(dev, tunnel);
-}
-
-static void
-tunnel_element_release_miss(struct rte_eth_dev *dev, void *x)
-{
-	struct tunnel_db_element_release_ctx *ctx = x;
-	RTE_SET_USED(dev);
-	ctx->ret = rte_flow_error_set(ctx->error, EINVAL,
-				      RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
-				      "invalid argument");
-}
-
+		       struct rte_flow_error *error);
 static int
 mlx5_flow_tunnel_item_release(struct rte_eth_dev *dev,
 			      struct rte_flow_item *pmd_items,
-			      uint32_t num_items, struct rte_flow_error *err)
-{
-	struct tunnel_db_element_release_ctx ctx = {
-		.items = pmd_items,
-		.actions = NULL,
-		.num_elements = num_items,
-		.error = err,
-	};
-
-	mlx5_access_tunnel_offload_db(dev, tunnel_element_release_match,
-				      tunnel_element_release_hit,
-				      tunnel_element_release_miss, &ctx, false);
-
-	return ctx.ret;
-}
-
+			      uint32_t num_items, struct rte_flow_error *err);
 static int
 mlx5_flow_tunnel_action_release(struct rte_eth_dev *dev,
 				struct rte_flow_action *pmd_actions,
 				uint32_t num_actions,
-				struct rte_flow_error *err)
-{
-	struct tunnel_db_element_release_ctx ctx = {
-		.items = NULL,
-		.actions = pmd_actions,
-		.num_elements = num_actions,
-		.error = err,
-	};
-
-	mlx5_access_tunnel_offload_db(dev, tunnel_element_release_match,
-				      tunnel_element_release_hit,
-				      tunnel_element_release_miss, &ctx, false);
-
-	return ctx.ret;
-}
-
+				struct rte_flow_error *err);
 static int
 mlx5_flow_tunnel_get_restore_info(struct rte_eth_dev *dev,
 				  struct rte_mbuf *m,
 				  struct rte_flow_restore_info *info,
-				  struct rte_flow_error *err)
-{
-	uint64_t ol_flags = m->ol_flags;
-	const struct mlx5_flow_tbl_data_entry *tble;
-	const uint64_t mask = PKT_RX_FDIR | PKT_RX_FDIR_ID;
-
-	if ((ol_flags & mask) != mask)
-		goto err;
-	tble = tunnel_mark_decode(dev, m->hash.fdir.hi);
-	if (!tble) {
-		DRV_LOG(DEBUG, "port %u invalid miss tunnel mark %#x",
-			dev->data->port_id, m->hash.fdir.hi);
-		goto err;
-	}
-	MLX5_ASSERT(tble->tunnel);
-	memcpy(&info->tunnel, &tble->tunnel->app_tunnel, sizeof(info->tunnel));
-	info->group_id = tble->group_id;
-	info->flags = RTE_FLOW_RESTORE_INFO_TUNNEL |
-		      RTE_FLOW_RESTORE_INFO_GROUP_ID |
-		      RTE_FLOW_RESTORE_INFO_ENCAPSULATED;
-
-	return 0;
-
-err:
-	return rte_flow_error_set(err, EINVAL,
-				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
-				  "failed to get restore info");
-}
+				  struct rte_flow_error *err);
 
 static const struct rte_flow_ops mlx5_flow_ops = {
 	.validate = mlx5_flow_validate,
@@ -4206,142 +4044,6 @@ flow_hairpin_split(struct rte_eth_dev *dev,
 	return 0;
 }
 
-__extension__
-union tunnel_offload_mark {
-	uint32_t val;
-	struct {
-		uint32_t app_reserve:8;
-		uint32_t table_id:15;
-		uint32_t transfer:1;
-		uint32_t _unused_:8;
-	};
-};
-
-struct tunnel_default_miss_ctx {
-	uint16_t *queue;
-	__extension__
-	union {
-		struct rte_flow_action_rss action_rss;
-		struct rte_flow_action_queue miss_queue;
-		struct rte_flow_action_jump miss_jump;
-		uint8_t raw[0];
-	};
-};
-
-static int
-flow_tunnel_add_default_miss(struct rte_eth_dev *dev,
-			     struct rte_flow *flow,
-			     const struct rte_flow_attr *attr,
-			     const struct rte_flow_action *app_actions,
-			     uint32_t flow_idx,
-			     struct tunnel_default_miss_ctx *ctx,
-			     struct rte_flow_error *error)
-{
-	struct mlx5_priv *priv = dev->data->dev_private;
-	struct mlx5_flow *dev_flow;
-	struct rte_flow_attr miss_attr = *attr;
-	const struct mlx5_flow_tunnel *tunnel = app_actions[0].conf;
-	const struct rte_flow_item miss_items[2] = {
-		{
-			.type = RTE_FLOW_ITEM_TYPE_ETH,
-			.spec = NULL,
-			.last = NULL,
-			.mask = NULL
-		},
-		{
-			.type = RTE_FLOW_ITEM_TYPE_END,
-			.spec = NULL,
-			.last = NULL,
-			.mask = NULL
-		}
-	};
-	union tunnel_offload_mark mark_id;
-	struct rte_flow_action_mark miss_mark;
-	struct rte_flow_action miss_actions[3] = {
-		[0] = { .type = RTE_FLOW_ACTION_TYPE_MARK, .conf = &miss_mark },
-		[2] = { .type = RTE_FLOW_ACTION_TYPE_END,  .conf = NULL }
-	};
-	const struct rte_flow_action_jump *jump_data;
-	uint32_t i, flow_table = 0; /* prevent compilation warning */
-	struct flow_grp_info grp_info = {
-		.external = 1,
-		.transfer = attr->transfer,
-		.fdb_def_rule = !!priv->fdb_def_rule,
-		.std_tbl_fix = 0,
-	};
-	int ret;
-
-	if (!attr->transfer) {
-		uint32_t q_size;
-
-		miss_actions[1].type = RTE_FLOW_ACTION_TYPE_RSS;
-		q_size = priv->reta_idx_n * sizeof(ctx->queue[0]);
-		ctx->queue = mlx5_malloc(MLX5_MEM_SYS | MLX5_MEM_ZERO, q_size,
-					 0, SOCKET_ID_ANY);
-		if (!ctx->queue)
-			return rte_flow_error_set
-				(error, ENOMEM,
-				RTE_FLOW_ERROR_TYPE_ACTION_CONF,
-				NULL, "invalid default miss RSS");
-		ctx->action_rss.func = RTE_ETH_HASH_FUNCTION_DEFAULT,
-		ctx->action_rss.level = 0,
-		ctx->action_rss.types = priv->rss_conf.rss_hf,
-		ctx->action_rss.key_len = priv->rss_conf.rss_key_len,
-		ctx->action_rss.queue_num = priv->reta_idx_n,
-		ctx->action_rss.key = priv->rss_conf.rss_key,
-		ctx->action_rss.queue = ctx->queue;
-		if (!priv->reta_idx_n || !priv->rxqs_n)
-			return rte_flow_error_set
-				(error, EINVAL,
-				RTE_FLOW_ERROR_TYPE_ACTION_CONF,
-				NULL, "invalid port configuration");
-		if (!(dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
-			ctx->action_rss.types = 0;
-		for (i = 0; i != priv->reta_idx_n; ++i)
-			ctx->queue[i] = (*priv->reta_idx)[i];
-	} else {
-		miss_actions[1].type = RTE_FLOW_ACTION_TYPE_JUMP;
-		ctx->miss_jump.group = MLX5_TNL_MISS_FDB_JUMP_GRP;
-	}
-	miss_actions[1].conf = (typeof(miss_actions[1].conf))ctx->raw;
-	for (; app_actions->type != RTE_FLOW_ACTION_TYPE_JUMP; app_actions++);
-	jump_data = app_actions->conf;
-	miss_attr.priority = MLX5_TNL_MISS_RULE_PRIORITY;
-	miss_attr.group = jump_data->group;
-	ret = mlx5_flow_group_to_table(dev, tunnel, jump_data->group,
-				       &flow_table, grp_info, error);
-	if (ret)
-		return rte_flow_error_set(error, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_ACTION_CONF,
-					  NULL, "invalid tunnel id");
-	mark_id.app_reserve = 0;
-	mark_id.table_id = tunnel_flow_tbl_to_id(flow_table);
-	mark_id.transfer = !!attr->transfer;
-	mark_id._unused_ = 0;
-	miss_mark.id = mark_id.val;
-	dev_flow = flow_drv_prepare(dev, flow, &miss_attr,
-				    miss_items, miss_actions, flow_idx, error);
-	if (!dev_flow)
-		return -rte_errno;
-	dev_flow->flow = flow;
-	dev_flow->external = true;
-	dev_flow->tunnel = tunnel;
-	/* Subflow object was created, we must include one in the list. */
-	SILIST_INSERT(&flow->dev_handles, dev_flow->handle_idx,
-		      dev_flow->handle, next);
-	DRV_LOG(DEBUG,
-		"port %u tunnel type=%d id=%u miss rule priority=%u group=%u",
-		dev->data->port_id, tunnel->app_tunnel.type,
-		tunnel->tunnel_id, miss_attr.priority, miss_attr.group);
-	ret = flow_drv_translate(dev, dev_flow, &miss_attr, miss_items,
-				  miss_actions, error);
-	if (!ret)
-		ret = flow_mreg_update_copy_table(dev, flow, miss_actions,
-						  error);
-
-	return ret;
-}
-
 /**
  * The last stage of splitting chain, just creates the subflow
  * without any modification.
@@ -5672,7 +5374,8 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
 					      error);
 		if (ret < 0)
 			goto error;
-		if (is_flow_tunnel_steer_rule(dev, attr,
+		if (is_tunnel_offload_active(dev) &&
+		    is_flow_tunnel_steer_rule(dev, attr,
 					      buf->entry[i].pattern,
 					      p_actions_rx)) {
 			ret = flow_tunnel_add_default_miss(dev, flow, attr,
@@ -5743,12 +5446,15 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
 	wks->flow_idx = wks->flow_nested_idx;
 	if (wks->flow_nested_idx)
 		wks->flow_nested_idx = 0;
-	tunnel = flow_tunnel_from_rule(dev, attr, items, actions);
-	if (tunnel) {
-		flow->tunnel = 1;
-		flow->tunnel_id = tunnel->tunnel_id;
-		__atomic_add_fetch(&tunnel->refctn, 1, __ATOMIC_RELAXED);
-		mlx5_free(default_miss_ctx.queue);
+	if (is_tunnel_offload_active(dev)) {
+		tunnel = flow_tunnel_from_rule(dev, attr, items, actions);
+		if (tunnel) {
+			flow->tunnel = 1;
+			flow->tunnel_id = tunnel->tunnel_id;
+			__atomic_add_fetch(&tunnel->refctn, 1,
+					   __ATOMIC_RELAXED);
+			mlx5_free(default_miss_ctx.queue);
+		}
 	}
 	return idx;
 error:
@@ -6945,107 +6651,6 @@ mlx5_flow_async_pool_query_handle(struct mlx5_dev_ctx_shared *sh,
 	sh->cmng.pending_queries--;
 }
 
-static const struct mlx5_flow_tbl_data_entry  *
-tunnel_mark_decode(struct rte_eth_dev *dev, uint32_t mark)
-{
-	struct mlx5_priv *priv = dev->data->dev_private;
-	struct mlx5_dev_ctx_shared *sh = priv->sh;
-	struct mlx5_hlist_entry *he;
-	union tunnel_offload_mark mbits = { .val = mark };
-	union mlx5_flow_tbl_key table_key = {
-		{
-			.table_id = tunnel_id_to_flow_tbl(mbits.table_id),
-			.dummy = 0,
-			.domain = !!mbits.transfer,
-			.direction = 0,
-		}
-	};
-	he = mlx5_hlist_lookup(sh->flow_tbls, table_key.v64, NULL);
-	return he ?
-	       container_of(he, struct mlx5_flow_tbl_data_entry, entry) : NULL;
-}
-
-static void
-mlx5_flow_tunnel_grp2tbl_remove_cb(struct mlx5_hlist *list,
-				   struct mlx5_hlist_entry *entry)
-{
-	struct mlx5_dev_ctx_shared *sh = list->ctx;
-	struct tunnel_tbl_entry *tte = container_of(entry, typeof(*tte), hash);
-
-	mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
-			tunnel_flow_tbl_to_id(tte->flow_table));
-	mlx5_free(tte);
-}
-
-static struct mlx5_hlist_entry *
-mlx5_flow_tunnel_grp2tbl_create_cb(struct mlx5_hlist *list,
-				   uint64_t key __rte_unused,
-				   void *ctx __rte_unused)
-{
-	struct mlx5_dev_ctx_shared *sh = list->ctx;
-	struct tunnel_tbl_entry *tte;
-
-	tte = mlx5_malloc(MLX5_MEM_SYS | MLX5_MEM_ZERO,
-			  sizeof(*tte), 0,
-			  SOCKET_ID_ANY);
-	if (!tte)
-		goto err;
-	mlx5_ipool_malloc(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
-			  &tte->flow_table);
-	if (tte->flow_table >= MLX5_MAX_TABLES) {
-		DRV_LOG(ERR, "Tunnel TBL ID %d exceed max limit.",
-			tte->flow_table);
-		mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
-				tte->flow_table);
-		goto err;
-	} else if (!tte->flow_table) {
-		goto err;
-	}
-	tte->flow_table = tunnel_id_to_flow_tbl(tte->flow_table);
-	return &tte->hash;
-err:
-	if (tte)
-		mlx5_free(tte);
-	return NULL;
-}
-
-static uint32_t
-tunnel_flow_group_to_flow_table(struct rte_eth_dev *dev,
-				const struct mlx5_flow_tunnel *tunnel,
-				uint32_t group, uint32_t *table,
-				struct rte_flow_error *error)
-{
-	struct mlx5_hlist_entry *he;
-	struct tunnel_tbl_entry *tte;
-	union tunnel_tbl_key key = {
-		.tunnel_id = tunnel ? tunnel->tunnel_id : 0,
-		.group = group
-	};
-	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_hlist *group_hash;
-
-	group_hash = tunnel ? tunnel->groups : thub->groups;
-	he = mlx5_hlist_lookup(group_hash, key.val, NULL);
-	if (!he) {
-		DRV_LOG(DEBUG, "port %u tunnel %u group=%u - generate table id",
-		dev->data->port_id, key.tunnel_id, group);
-		he = mlx5_hlist_register(group_hash, key.val, NULL);
-	} else {
-		DRV_LOG(DEBUG, "port %u tunnel %u group=%u - skip table id",
-		dev->data->port_id, key.tunnel_id, group);
-	}
-	if (!he)
-		return rte_flow_error_set(error, EINVAL,
-					  RTE_FLOW_ERROR_TYPE_ATTR_GROUP,
-					  NULL,
-					  "tunnel group index not supported");
-	tte = container_of(he, typeof(*tte), hash);
-	*table = tte->flow_table;
-	DRV_LOG(DEBUG, "port %u tunnel %u group=%u table=%u",
-	dev->data->port_id, key.tunnel_id, group, *table);
-	return 0;
-}
-
 static int
 flow_group_to_table(uint32_t port_id, uint32_t group, uint32_t *table,
 		    struct flow_grp_info grp_info, struct rte_flow_error *error)
@@ -7506,23 +7111,38 @@ mlx5_shared_action_flush(struct rte_eth_dev *dev)
 	return ret;
 }
 
-static void
-mlx5_flow_tunnel_free(struct rte_eth_dev *dev,
-		      struct mlx5_flow_tunnel *tunnel)
+#ifndef HAVE_MLX5DV_DR
+#define MLX5_DOMAIN_SYNC_FLOW ((1 << 0) | (1 << 1))
+#else
+#define MLX5_DOMAIN_SYNC_FLOW \
+	(MLX5DV_DR_DOMAIN_SYNC_FLAGS_SW | MLX5DV_DR_DOMAIN_SYNC_FLAGS_HW)
+#endif
+
+int rte_pmd_mlx5_sync_flow(uint16_t port_id, uint32_t domains)
 {
-	/* no tunnel hub spinlock protection */
-	struct mlx5_priv *priv = dev->data->dev_private;
-	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_indexed_pool *ipool;
+	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
+	const struct mlx5_flow_driver_ops *fops;
+	int ret;
+	struct rte_flow_attr attr = { .transfer = 0 };
 
-	DRV_LOG(DEBUG, "port %u release pmd tunnel id=0x%x",
-		dev->data->port_id, tunnel->tunnel_id);
-	rte_spinlock_lock(&thub->sl);
-	LIST_REMOVE(tunnel, chain);
-	rte_spinlock_unlock(&thub->sl);
-	mlx5_hlist_destroy(tunnel->groups);
-	ipool = priv->sh->ipool[MLX5_IPOOL_TUNNEL_OFFLOAD];
-	mlx5_ipool_free(ipool, tunnel->tunnel_id);
+	fops = flow_get_drv_ops(flow_get_drv_type(dev, &attr));
+	ret = fops->sync_domain(dev, domains, MLX5_DOMAIN_SYNC_FLOW);
+	if (ret > 0)
+		ret = -ret;
+	return ret;
+}
+
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+static inline uint32_t
+tunnel_id_to_flow_tbl(uint32_t id)
+{
+	return id | (1u << 16);
+}
+
+static inline uint32_t
+tunnel_flow_tbl_to_id(uint32_t flow_tbl)
+{
+	return flow_tbl & ~(1u << 16);
 }
 
 static bool
@@ -7532,19 +7152,235 @@ mlx5_access_tunnel_offload_db
 		       struct mlx5_flow_tunnel *, const void *),
 	 void (*hit)(struct rte_eth_dev *, struct mlx5_flow_tunnel *, void *),
 	 void (*miss)(struct rte_eth_dev *, void *),
-	 void *ctx, bool lock_op)
-{
-	bool verdict = false;
-	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
-	struct mlx5_flow_tunnel *tunnel;
+	 void *ctx, bool lock_op);
+static const struct mlx5_flow_tbl_data_entry  *
+tunnel_mark_decode(struct rte_eth_dev *dev, uint32_t mark);
+static int
+mlx5_get_flow_tunnel(struct rte_eth_dev *dev,
+		     const struct rte_flow_tunnel *app_tunnel,
+		     struct mlx5_flow_tunnel **tunnel);
 
-	rte_spinlock_lock(&thub->sl);
-	LIST_FOREACH(tunnel, &thub->tunnels, chain) {
-		verdict = match(dev, tunnel, (const void *)ctx);
-		if (verdict)
-			break;
-	}
-	if (!lock_op)
+__extension__
+union tunnel_offload_mark {
+	uint32_t val;
+	struct {
+		uint32_t app_reserve:8;
+		uint32_t table_id:15;
+		uint32_t transfer:1;
+		uint32_t _unused_:8;
+	};
+};
+
+static int
+flow_tunnel_add_default_miss(struct rte_eth_dev *dev,
+			     struct rte_flow *flow,
+			     const struct rte_flow_attr *attr,
+			     const struct rte_flow_action *app_actions,
+			     uint32_t flow_idx,
+			     struct tunnel_default_miss_ctx *ctx,
+			     struct rte_flow_error *error)
+{
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_flow *dev_flow;
+	struct rte_flow_attr miss_attr = *attr;
+	const struct mlx5_flow_tunnel *tunnel = app_actions[0].conf;
+	const struct rte_flow_item miss_items[2] = {
+		{
+			.type = RTE_FLOW_ITEM_TYPE_ETH,
+			.spec = NULL,
+			.last = NULL,
+			.mask = NULL
+		},
+		{
+			.type = RTE_FLOW_ITEM_TYPE_END,
+			.spec = NULL,
+			.last = NULL,
+			.mask = NULL
+		}
+	};
+	union tunnel_offload_mark mark_id;
+	struct rte_flow_action_mark miss_mark;
+	struct rte_flow_action miss_actions[3] = {
+		[0] = { .type = RTE_FLOW_ACTION_TYPE_MARK, .conf = &miss_mark },
+		[2] = { .type = RTE_FLOW_ACTION_TYPE_END,  .conf = NULL }
+	};
+	const struct rte_flow_action_jump *jump_data;
+	uint32_t i, flow_table = 0; /* prevent compilation warning */
+	struct flow_grp_info grp_info = {
+		.external = 1,
+		.transfer = attr->transfer,
+		.fdb_def_rule = !!priv->fdb_def_rule,
+		.std_tbl_fix = 0,
+	};
+	int ret;
+
+	if (!attr->transfer) {
+		uint32_t q_size;
+
+		miss_actions[1].type = RTE_FLOW_ACTION_TYPE_RSS;
+		q_size = priv->reta_idx_n * sizeof(ctx->queue[0]);
+		ctx->queue = mlx5_malloc(MLX5_MEM_SYS | MLX5_MEM_ZERO, q_size,
+					 0, SOCKET_ID_ANY);
+		if (!ctx->queue)
+			return rte_flow_error_set
+				(error, ENOMEM,
+				RTE_FLOW_ERROR_TYPE_ACTION_CONF,
+				NULL, "invalid default miss RSS");
+		ctx->action_rss.func = RTE_ETH_HASH_FUNCTION_DEFAULT,
+		ctx->action_rss.level = 0,
+		ctx->action_rss.types = priv->rss_conf.rss_hf,
+		ctx->action_rss.key_len = priv->rss_conf.rss_key_len,
+		ctx->action_rss.queue_num = priv->reta_idx_n,
+		ctx->action_rss.key = priv->rss_conf.rss_key,
+		ctx->action_rss.queue = ctx->queue;
+		if (!priv->reta_idx_n || !priv->rxqs_n)
+			return rte_flow_error_set
+				(error, EINVAL,
+				RTE_FLOW_ERROR_TYPE_ACTION_CONF,
+				NULL, "invalid port configuration");
+		if (!(dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
+			ctx->action_rss.types = 0;
+		for (i = 0; i != priv->reta_idx_n; ++i)
+			ctx->queue[i] = (*priv->reta_idx)[i];
+	} else {
+		miss_actions[1].type = RTE_FLOW_ACTION_TYPE_JUMP;
+		ctx->miss_jump.group = MLX5_TNL_MISS_FDB_JUMP_GRP;
+	}
+	miss_actions[1].conf = (typeof(miss_actions[1].conf))ctx->raw;
+	for (; app_actions->type != RTE_FLOW_ACTION_TYPE_JUMP; app_actions++);
+	jump_data = app_actions->conf;
+	miss_attr.priority = MLX5_TNL_MISS_RULE_PRIORITY;
+	miss_attr.group = jump_data->group;
+	ret = mlx5_flow_group_to_table(dev, tunnel, jump_data->group,
+				       &flow_table, grp_info, error);
+	if (ret)
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ACTION_CONF,
+					  NULL, "invalid tunnel id");
+	mark_id.app_reserve = 0;
+	mark_id.table_id = tunnel_flow_tbl_to_id(flow_table);
+	mark_id.transfer = !!attr->transfer;
+	mark_id._unused_ = 0;
+	miss_mark.id = mark_id.val;
+	dev_flow = flow_drv_prepare(dev, flow, &miss_attr,
+				    miss_items, miss_actions, flow_idx, error);
+	if (!dev_flow)
+		return -rte_errno;
+	dev_flow->flow = flow;
+	dev_flow->external = true;
+	dev_flow->tunnel = tunnel;
+	/* Subflow object was created, we must include one in the list. */
+	SILIST_INSERT(&flow->dev_handles, dev_flow->handle_idx,
+		      dev_flow->handle, next);
+	DRV_LOG(DEBUG,
+		"port %u tunnel type=%d id=%u miss rule priority=%u group=%u",
+		dev->data->port_id, tunnel->app_tunnel.type,
+		tunnel->tunnel_id, miss_attr.priority, miss_attr.group);
+	ret = flow_drv_translate(dev, dev_flow, &miss_attr, miss_items,
+				  miss_actions, error);
+	if (!ret)
+		ret = flow_mreg_update_copy_table(dev, flow, miss_actions,
+						  error);
+
+	return ret;
+}
+
+static const struct mlx5_flow_tbl_data_entry  *
+tunnel_mark_decode(struct rte_eth_dev *dev, uint32_t mark)
+{
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_dev_ctx_shared *sh = priv->sh;
+	struct mlx5_hlist_entry *he;
+	union tunnel_offload_mark mbits = { .val = mark };
+	union mlx5_flow_tbl_key table_key = {
+		{
+			.table_id = tunnel_id_to_flow_tbl(mbits.table_id),
+			.dummy = 0,
+			.domain = !!mbits.transfer,
+			.direction = 0,
+		}
+	};
+	he = mlx5_hlist_lookup(sh->flow_tbls, table_key.v64, NULL);
+	return he ?
+	       container_of(he, struct mlx5_flow_tbl_data_entry, entry) : NULL;
+}
+
+static uint32_t
+tunnel_flow_group_to_flow_table(struct rte_eth_dev *dev,
+				const struct mlx5_flow_tunnel *tunnel,
+				uint32_t group, uint32_t *table,
+				struct rte_flow_error *error)
+{
+	struct mlx5_hlist_entry *he;
+	struct tunnel_tbl_entry *tte;
+	union tunnel_tbl_key key = {
+		.tunnel_id = tunnel ? tunnel->tunnel_id : 0,
+		.group = group
+	};
+	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
+	struct mlx5_hlist *group_hash;
+
+	group_hash = tunnel ? tunnel->groups : thub->groups;
+	he = mlx5_hlist_lookup(group_hash, key.val, NULL);
+	if (!he) {
+		DRV_LOG(DEBUG, "port %u tunnel %u group=%u - generate table id",
+		dev->data->port_id, key.tunnel_id, group);
+		he = mlx5_hlist_register(group_hash, key.val, NULL);
+	} else {
+		DRV_LOG(DEBUG, "port %u tunnel %u group=%u - skip table id",
+		dev->data->port_id, key.tunnel_id, group);
+	}
+	if (!he)
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ATTR_GROUP,
+					  NULL,
+					  "tunnel group index not supported");
+	tte = container_of(he, typeof(*tte), hash);
+	*table = tte->flow_table;
+	DRV_LOG(DEBUG, "port %u tunnel %u group=%u table=%u",
+	dev->data->port_id, key.tunnel_id, group, *table);
+	return 0;
+}
+
+static void
+mlx5_flow_tunnel_free(struct rte_eth_dev *dev,
+		      struct mlx5_flow_tunnel *tunnel)
+{
+	/* no tunnel hub spinlock protection */
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
+	struct mlx5_indexed_pool *ipool;
+
+	DRV_LOG(DEBUG, "port %u release pmd tunnel id=0x%x",
+		dev->data->port_id, tunnel->tunnel_id);
+	rte_spinlock_lock(&thub->sl);
+	LIST_REMOVE(tunnel, chain);
+	rte_spinlock_unlock(&thub->sl);
+	mlx5_hlist_destroy(tunnel->groups);
+	ipool = priv->sh->ipool[MLX5_IPOOL_TUNNEL_OFFLOAD];
+	mlx5_ipool_free(ipool, tunnel->tunnel_id);
+}
+
+static bool
+mlx5_access_tunnel_offload_db
+	(struct rte_eth_dev *dev,
+	 bool (*match)(struct rte_eth_dev *,
+		       struct mlx5_flow_tunnel *, const void *),
+	 void (*hit)(struct rte_eth_dev *, struct mlx5_flow_tunnel *, void *),
+	 void (*miss)(struct rte_eth_dev *, void *),
+	 void *ctx, bool lock_op)
+{
+	bool verdict = false;
+	struct mlx5_flow_tunnel_hub *thub = mlx5_tunnel_hub(dev);
+	struct mlx5_flow_tunnel *tunnel;
+
+	rte_spinlock_lock(&thub->sl);
+	LIST_FOREACH(tunnel, &thub->tunnels, chain) {
+		verdict = match(dev, tunnel, (const void *)ctx);
+		if (verdict)
+			break;
+	}
+	if (!lock_op)
 		rte_spinlock_unlock(&thub->sl);
 	if (verdict && hit)
 		hit(dev, tunnel, ctx);
@@ -7593,6 +7429,50 @@ mlx5_find_tunnel_id(struct rte_eth_dev *dev, uint32_t id)
 	return ctx.tunnel;
 }
 
+static void
+mlx5_flow_tunnel_grp2tbl_remove_cb(struct mlx5_hlist *list,
+				   struct mlx5_hlist_entry *entry)
+{
+	struct mlx5_dev_ctx_shared *sh = list->ctx;
+	struct tunnel_tbl_entry *tte = container_of(entry, typeof(*tte), hash);
+
+	mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
+			tunnel_flow_tbl_to_id(tte->flow_table));
+	mlx5_free(tte);
+}
+
+static struct mlx5_hlist_entry *
+mlx5_flow_tunnel_grp2tbl_create_cb(struct mlx5_hlist *list,
+				   uint64_t key __rte_unused,
+				   void *ctx __rte_unused)
+{
+	struct mlx5_dev_ctx_shared *sh = list->ctx;
+	struct tunnel_tbl_entry *tte;
+
+	tte = mlx5_malloc(MLX5_MEM_SYS | MLX5_MEM_ZERO,
+			  sizeof(*tte), 0,
+			  SOCKET_ID_ANY);
+	if (!tte)
+		goto err;
+	mlx5_ipool_malloc(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
+			  &tte->flow_table);
+	if (tte->flow_table >= MLX5_MAX_TABLES) {
+		DRV_LOG(ERR, "Tunnel TBL ID %d exceed max limit.",
+			tte->flow_table);
+		mlx5_ipool_free(sh->ipool[MLX5_IPOOL_TUNNEL_FLOW_TBL_ID],
+				tte->flow_table);
+		goto err;
+	} else if (!tte->flow_table) {
+		goto err;
+	}
+	tte->flow_table = tunnel_id_to_flow_tbl(tte->flow_table);
+	return &tte->hash;
+err:
+	if (tte)
+		mlx5_free(tte);
+	return NULL;
+}
+
 static struct mlx5_flow_tunnel *
 mlx5_flow_tunnel_allocate(struct rte_eth_dev *dev,
 			  const struct rte_flow_tunnel *app_tunnel)
@@ -7678,7 +7558,6 @@ static void get_tunnel_miss(struct rte_eth_dev *dev, void *x)
 		LIST_INSERT_HEAD(&thub->tunnels, ctx->tunnel, chain);
 }
 
-
 static int
 mlx5_get_flow_tunnel(struct rte_eth_dev *dev,
 		     const struct rte_flow_tunnel *app_tunnel,
@@ -7738,23 +7617,286 @@ int mlx5_alloc_tunnel_hub(struct mlx5_dev_ctx_shared *sh)
 	return err;
 }
 
-#ifndef HAVE_MLX5DV_DR
-#define MLX5_DOMAIN_SYNC_FLOW ((1 << 0) | (1 << 1))
-#else
-#define MLX5_DOMAIN_SYNC_FLOW \
-	(MLX5DV_DR_DOMAIN_SYNC_FLAGS_SW | MLX5DV_DR_DOMAIN_SYNC_FLAGS_HW)
-#endif
+static inline bool
+mlx5_flow_tunnel_validate(struct rte_eth_dev *dev,
+			  struct rte_flow_tunnel *tunnel,
+			  const char *err_msg)
+{
+	err_msg = NULL;
+	if (!is_tunnel_offload_active(dev)) {
+		err_msg = "tunnel offload was not activated";
+		goto out;
+	} else if (!tunnel) {
+		err_msg = "no application tunnel";
+		goto out;
+	}
 
-int rte_pmd_mlx5_sync_flow(uint16_t port_id, uint32_t domains)
+	switch (tunnel->type) {
+	default:
+		err_msg = "unsupported tunnel type";
+		goto out;
+	case RTE_FLOW_ITEM_TYPE_VXLAN:
+		break;
+	}
+
+out:
+	return !err_msg;
+}
+
+static int
+mlx5_flow_tunnel_decap_set(struct rte_eth_dev *dev,
+			   struct rte_flow_tunnel *app_tunnel,
+			   struct rte_flow_action **actions,
+			   uint32_t *num_of_actions,
+			   struct rte_flow_error *error)
 {
-	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
-	const struct mlx5_flow_driver_ops *fops;
 	int ret;
-	struct rte_flow_attr attr = { .transfer = 0 };
+	struct mlx5_flow_tunnel *tunnel;
+	const char *err_msg = NULL;
+	bool verdict = mlx5_flow_tunnel_validate(dev, app_tunnel, err_msg);
 
-	fops = flow_get_drv_ops(flow_get_drv_type(dev, &attr));
-	ret = fops->sync_domain(dev, domains, MLX5_DOMAIN_SYNC_FLOW);
-	if (ret > 0)
-		ret = -ret;
-	return ret;
+	if (!verdict)
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_ACTION_CONF, NULL,
+					  err_msg);
+	ret = mlx5_get_flow_tunnel(dev, app_tunnel, &tunnel);
+	if (ret < 0) {
+		return rte_flow_error_set(error, ret,
+					  RTE_FLOW_ERROR_TYPE_ACTION_CONF, NULL,
+					  "failed to initialize pmd tunnel");
+	}
+	*actions = &tunnel->action;
+	*num_of_actions = 1;
+	return 0;
+}
+
+static int
+mlx5_flow_tunnel_match(struct rte_eth_dev *dev,
+		       struct rte_flow_tunnel *app_tunnel,
+		       struct rte_flow_item **items,
+		       uint32_t *num_of_items,
+		       struct rte_flow_error *error)
+{
+	int ret;
+	struct mlx5_flow_tunnel *tunnel;
+	const char *err_msg = NULL;
+	bool verdict = mlx5_flow_tunnel_validate(dev, app_tunnel, err_msg);
+
+	if (!verdict)
+		return rte_flow_error_set(error, EINVAL,
+					  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+					  err_msg);
+	ret = mlx5_get_flow_tunnel(dev, app_tunnel, &tunnel);
+	if (ret < 0) {
+		return rte_flow_error_set(error, ret,
+					  RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+					  "failed to initialize pmd tunnel");
+	}
+	*items = &tunnel->item;
+	*num_of_items = 1;
+	return 0;
+}
+
+struct tunnel_db_element_release_ctx {
+	struct rte_flow_item *items;
+	struct rte_flow_action *actions;
+	uint32_t num_elements;
+	struct rte_flow_error *error;
+	int ret;
+};
+
+static bool
+tunnel_element_release_match(struct rte_eth_dev *dev,
+			     struct mlx5_flow_tunnel *tunnel, const void *x)
+{
+	const struct tunnel_db_element_release_ctx *ctx = x;
+
+	RTE_SET_USED(dev);
+	if (ctx->num_elements != 1)
+		return false;
+	else if (ctx->items)
+		return ctx->items == &tunnel->item;
+	else if (ctx->actions)
+		return ctx->actions == &tunnel->action;
+
+	return false;
+}
+
+static void
+tunnel_element_release_hit(struct rte_eth_dev *dev,
+			   struct mlx5_flow_tunnel *tunnel, void *x)
+{
+	struct tunnel_db_element_release_ctx *ctx = x;
+
+	ctx->ret = 0;
+	if (!__atomic_sub_fetch(&tunnel->refctn, 1, __ATOMIC_RELAXED))
+		mlx5_flow_tunnel_free(dev, tunnel);
+}
+
+static void
+tunnel_element_release_miss(struct rte_eth_dev *dev, void *x)
+{
+	struct tunnel_db_element_release_ctx *ctx = x;
+
+	RTE_SET_USED(dev);
+	ctx->ret = rte_flow_error_set(ctx->error, EINVAL,
+				      RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+				      "invalid argument");
+}
+
+static int
+mlx5_flow_tunnel_item_release(struct rte_eth_dev *dev,
+			      struct rte_flow_item *pmd_items,
+			      uint32_t num_items, struct rte_flow_error *err)
+{
+	struct tunnel_db_element_release_ctx ctx = {
+		.items = pmd_items,
+		.actions = NULL,
+		.num_elements = num_items,
+		.error = err,
+	};
+
+	mlx5_access_tunnel_offload_db(dev, tunnel_element_release_match,
+				      tunnel_element_release_hit,
+				      tunnel_element_release_miss, &ctx, false);
+
+	return ctx.ret;
+}
+
+static int
+mlx5_flow_tunnel_action_release(struct rte_eth_dev *dev,
+				struct rte_flow_action *pmd_actions,
+				uint32_t num_actions,
+				struct rte_flow_error *err)
+{
+	struct tunnel_db_element_release_ctx ctx = {
+		.items = NULL,
+		.actions = pmd_actions,
+		.num_elements = num_actions,
+		.error = err,
+	};
+
+	mlx5_access_tunnel_offload_db(dev, tunnel_element_release_match,
+				      tunnel_element_release_hit,
+				      tunnel_element_release_miss, &ctx, false);
+
+	return ctx.ret;
+}
+
+static int
+mlx5_flow_tunnel_get_restore_info(struct rte_eth_dev *dev,
+				  struct rte_mbuf *m,
+				  struct rte_flow_restore_info *info,
+				  struct rte_flow_error *err)
+{
+	uint64_t ol_flags = m->ol_flags;
+	const struct mlx5_flow_tbl_data_entry *tble;
+	const uint64_t mask = PKT_RX_FDIR | PKT_RX_FDIR_ID;
+
+	if ((ol_flags & mask) != mask)
+		goto err;
+	tble = tunnel_mark_decode(dev, m->hash.fdir.hi);
+	if (!tble) {
+		DRV_LOG(DEBUG, "port %u invalid miss tunnel mark %#x",
+			dev->data->port_id, m->hash.fdir.hi);
+		goto err;
+	}
+	MLX5_ASSERT(tble->tunnel);
+	memcpy(&info->tunnel, &tble->tunnel->app_tunnel, sizeof(info->tunnel));
+	info->group_id = tble->group_id;
+	info->flags = RTE_FLOW_RESTORE_INFO_TUNNEL |
+		      RTE_FLOW_RESTORE_INFO_GROUP_ID |
+		      RTE_FLOW_RESTORE_INFO_ENCAPSULATED;
+
+	return 0;
+
+err:
+	return rte_flow_error_set(err, EINVAL,
+				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+				  "failed to get restore info");
+}
+#else /* HAVE_IBV_FLOW_DV_SUPPORT */
+static int
+mlx5_flow_tunnel_decap_set(__rte_unused struct rte_eth_dev *dev,
+			   __rte_unused struct rte_flow_tunnel *app_tunnel,
+			   __rte_unused struct rte_flow_action **actions,
+			   __rte_unused uint32_t *num_of_actions,
+			   __rte_unused struct rte_flow_error *error)
+{
+	return -ENOTSUP;
+}
+
+static int
+mlx5_flow_tunnel_match(__rte_unused struct rte_eth_dev *dev,
+		       __rte_unused struct rte_flow_tunnel *app_tunnel,
+		       __rte_unused struct rte_flow_item **items,
+		       __rte_unused uint32_t *num_of_items,
+		       __rte_unused struct rte_flow_error *error)
+{
+	return -ENOTSUP;
+}
+
+static int
+mlx5_flow_tunnel_item_release(__rte_unused struct rte_eth_dev *dev,
+			      __rte_unused struct rte_flow_item *pmd_items,
+			      __rte_unused uint32_t num_items,
+			      __rte_unused struct rte_flow_error *err)
+{
+	return -ENOTSUP;
+}
+
+static int
+mlx5_flow_tunnel_action_release(__rte_unused struct rte_eth_dev *dev,
+				__rte_unused struct rte_flow_action *pmd_action,
+				__rte_unused uint32_t num_actions,
+				__rte_unused struct rte_flow_error *err)
+{
+	return -ENOTSUP;
+}
+
+static int
+mlx5_flow_tunnel_get_restore_info(__rte_unused struct rte_eth_dev *dev,
+				  __rte_unused struct rte_mbuf *m,
+				  __rte_unused struct rte_flow_restore_info *i,
+				  __rte_unused struct rte_flow_error *err)
+{
+	return -ENOTSUP;
+}
+
+static int
+flow_tunnel_add_default_miss(__rte_unused struct rte_eth_dev *dev,
+			     __rte_unused struct rte_flow *flow,
+			     __rte_unused const struct rte_flow_attr *attr,
+			     __rte_unused const struct rte_flow_action *actions,
+			     __rte_unused uint32_t flow_idx,
+			     __rte_unused struct tunnel_default_miss_ctx *ctx,
+			     __rte_unused struct rte_flow_error *error)
+{
+	return -ENOTSUP;
+}
+
+static struct mlx5_flow_tunnel *
+mlx5_find_tunnel_id(__rte_unused struct rte_eth_dev *dev,
+		    __rte_unused uint32_t id)
+{
+	return NULL;
+}
+
+static void
+mlx5_flow_tunnel_free(__rte_unused struct rte_eth_dev *dev,
+		      __rte_unused struct mlx5_flow_tunnel *tunnel)
+{
+}
+
+static uint32_t
+tunnel_flow_group_to_flow_table(__rte_unused struct rte_eth_dev *dev,
+				__rte_unused const struct mlx5_flow_tunnel *t,
+				__rte_unused uint32_t group,
+				__rte_unused uint32_t *table,
+				struct rte_flow_error *error)
+{
+	return rte_flow_error_set(error, ENOTSUP,
+				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+				  "tunnel offload requires DV support");
 }
+#endif /* HAVE_IBV_FLOW_DV_SUPPORT */
diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
index bdf2c50090..672c27ecb9 100644
--- a/drivers/net/mlx5/mlx5_flow.h
+++ b/drivers/net/mlx5/mlx5_flow.h
@@ -965,18 +965,6 @@ struct tunnel_tbl_entry {
 	uint32_t flow_table;
 };
 
-static inline uint32_t
-tunnel_id_to_flow_tbl(uint32_t id)
-{
-	return id | (1u << 16);
-}
-
-static inline uint32_t
-tunnel_flow_tbl_to_id(uint32_t flow_tbl)
-{
-	return flow_tbl & ~(1u << 16);
-}
-
 union tunnel_tbl_key {
 	uint64_t val;
 	struct {
@@ -995,8 +983,13 @@ mlx5_tunnel_hub(struct rte_eth_dev *dev)
 static inline bool
 is_tunnel_offload_active(struct rte_eth_dev *dev)
 {
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
 	struct mlx5_priv *priv = dev->data->dev_private;
 	return !!priv->config.dv_miss_info;
+#else
+	RTE_SET_USED(dev);
+	return false;
+#endif
 }
 
 static inline bool
-- 
2.29.2


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation Gregory Etelson
@ 2020-11-13 19:00   ` Thomas Monjalon
  2020-11-14 17:47     ` Gregory Etelson
  0 siblings, 1 reply; 16+ messages in thread
From: Thomas Monjalon @ 2020-11-13 19:00 UTC (permalink / raw)
  To: Gregory Etelson
  Cc: dev, matan, rasland, Viacheslav Ovsiienko, Shahaf Shuler,
	Xueming Li, asafp

13/11/2020 15:52, Gregory Etelson:
> The original patch allocated tunnel offload objects with invalid
> indexes. As the result, PMD tunnel object allocation failed.
> 
> In this patch indexed pool provides both an index and memory for a new
> tunnel offload object.
> Also tunnel offload ipool moved to dv enabled code only.
> 
> Fixes: f2e8093 ("net/mlx5: use indexed pool as id generator")
> 
> Signed-off-by: Gregory Etelson <getelson@nvidia.com>2

Typo above

> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>

This patch is breaking compilation in non-DV mode.
You should move it at the end of this series,
after moving all in DV areas.




^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5
  2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
                   ` (4 preceding siblings ...)
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors Gregory Etelson
@ 2020-11-13 19:03 ` Thomas Monjalon
  5 siblings, 0 replies; 16+ messages in thread
From: Thomas Monjalon @ 2020-11-13 19:03 UTC (permalink / raw)
  To: Gregory Etelson; +Cc: dev, matan, rasland

13/11/2020 15:52, Gregory Etelson:
> post merge fixes that restore tunnel offload functionality in mlx5.
> 
> v2: resolve compilation error in ubuntu inbox.

Please use --in-reply-to to keep all versions threaded.



^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors Gregory Etelson
@ 2020-11-14 12:06   ` Thomas Monjalon
  2020-11-14 17:41     ` Gregory Etelson
  0 siblings, 1 reply; 16+ messages in thread
From: Thomas Monjalon @ 2020-11-14 12:06 UTC (permalink / raw)
  To: matan, Viacheslav Ovsiienko, Gregory Etelson
  Cc: dev, rasland, Shahaf Shuler, asafp

13/11/2020 15:52, Gregory Etelson:
> +       ret = mlx5_flow_group_to_table(dev, tunnel, jump_data->group,
> +                                      &flow_table, grp_info, error);

The parameter grp_info is a struct passed as value.
I believe it should be passed as a pointer.

I see some other functions are passing arrays with [] syntax,
which does not make sense in parameters. I should be a simple pointer.

Matan, Slava, as maintainers, what are your policies in mlx5 code?



^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-14 12:06   ` Thomas Monjalon
@ 2020-11-14 17:41     ` Gregory Etelson
  2020-11-14 17:53       ` Thomas Monjalon
  0 siblings, 1 reply; 16+ messages in thread
From: Gregory Etelson @ 2020-11-14 17:41 UTC (permalink / raw)
  To: NBU-Contact-Thomas Monjalon, Matan Azrad, Slava Ovsiienko
  Cc: dev, Raslan Darawsheh, Shahaf Shuler, Asaf Penso

> 13/11/2020 15:52, Gregory Etelson:
> > +       ret = mlx5_flow_group_to_table(dev, tunnel, jump_data->group,
> > +                                      &flow_table, grp_info, error);
> 
> The parameter grp_info is a struct passed as value.
> I believe it should be passed as a pointer.

struct flow_grp_info is a 64 bit-field:  
struct flow_grp_info {
	uint64_t external:1;
	uint64_t transfer:1;
	uint64_t fdb_def_rule:1;
	/* force standard group translation */
	uint64_t std_tbl_fix:1;
	uint64_t skip_scale:1;
};
Since mlx5_flow_group_to_table() does not change bits configuration,
there is no need to pass this type as a pointer.

> 
> I see some other functions are passing arrays with [] syntax, which does
> not make sense in parameters. I should be a simple pointer.
> 

I agree that arrays should not be passed with [] syntax.
However, I kept existing rte flow style - see rte_flow_create

> Matan, Slava, as maintainers, what are your policies in mlx5 code?
> 


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation
  2020-11-13 19:00   ` Thomas Monjalon
@ 2020-11-14 17:47     ` Gregory Etelson
  2020-11-14 17:55       ` Thomas Monjalon
  0 siblings, 1 reply; 16+ messages in thread
From: Gregory Etelson @ 2020-11-14 17:47 UTC (permalink / raw)
  To: NBU-Contact-Thomas Monjalon
  Cc: dev, Matan Azrad, Raslan Darawsheh, Slava Ovsiienko,
	Shahaf Shuler, Xueming(Steven) Li, Asaf Penso

> 13/11/2020 15:52, Gregory Etelson:
> > The original patch allocated tunnel offload objects with invalid
> > indexes. As the result, PMD tunnel object allocation failed.
> >
> > In this patch indexed pool provides both an index and memory for a new
> > tunnel offload object.
> > Also tunnel offload ipool moved to dv enabled code only.
> >
> > Fixes: f2e8093 ("net/mlx5: use indexed pool as id generator")
> >
> > Signed-off-by: Gregory Etelson <getelson@nvidia.com>2
> 
> Typo above
> 
> > Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
> 
> This patch is breaking compilation in non-DV mode.
> You should move it at the end of this series, after moving all in DV
> areas.

Patch "[PATCH v2 5/5] net/mlx5: fix non-dv compilation errors" in that series
resolves the compilation error.
To restore tunnel offload functionality all 5 patches must be applied.

 


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-14 17:41     ` Gregory Etelson
@ 2020-11-14 17:53       ` Thomas Monjalon
  2020-11-14 18:31         ` Gregory Etelson
  0 siblings, 1 reply; 16+ messages in thread
From: Thomas Monjalon @ 2020-11-14 17:53 UTC (permalink / raw)
  To: Gregory Etelson
  Cc: Matan Azrad, Slava Ovsiienko, dev, Raslan Darawsheh,
	Shahaf Shuler, Asaf Penso

14/11/2020 18:41, Gregory Etelson:
> > 13/11/2020 15:52, Gregory Etelson:
> > > +       ret = mlx5_flow_group_to_table(dev, tunnel, jump_data->group,
> > > +                                      &flow_table, grp_info, error);
> > 
> > The parameter grp_info is a struct passed as value.
> > I believe it should be passed as a pointer.
> 
> struct flow_grp_info is a 64 bit-field:  
> struct flow_grp_info {
> 	uint64_t external:1;
> 	uint64_t transfer:1;
> 	uint64_t fdb_def_rule:1;
> 	/* force standard group translation */
> 	uint64_t std_tbl_fix:1;
> 	uint64_t skip_scale:1;
> };
> Since mlx5_flow_group_to_table() does not change bits configuration,
> there is no need to pass this type as a pointer.

I feel passing struct as pointer is a better practice.

> > I see some other functions are passing arrays with [] syntax, which does
> > not make sense in parameters. I should be a simple pointer.
> 
> I agree that arrays should not be passed with [] syntax.
> However, I kept existing rte flow style - see rte_flow_create

I recommend not following this bad code style.
Later other occurences should be fixed.

> > Matan, Slava, as maintainers, what are your policies in mlx5 code?




^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation
  2020-11-14 17:47     ` Gregory Etelson
@ 2020-11-14 17:55       ` Thomas Monjalon
  0 siblings, 0 replies; 16+ messages in thread
From: Thomas Monjalon @ 2020-11-14 17:55 UTC (permalink / raw)
  To: Gregory Etelson
  Cc: dev, Matan Azrad, Raslan Darawsheh, Slava Ovsiienko,
	Shahaf Shuler, Xueming(Steven) Li, Asaf Penso

14/11/2020 18:47, Gregory Etelson:
> > 13/11/2020 15:52, Gregory Etelson:
> > > The original patch allocated tunnel offload objects with invalid
> > > indexes. As the result, PMD tunnel object allocation failed.
> > >
> > > In this patch indexed pool provides both an index and memory for a new
> > > tunnel offload object.
> > > Also tunnel offload ipool moved to dv enabled code only.
> > >
> > > Fixes: f2e8093 ("net/mlx5: use indexed pool as id generator")
> > >
> > > Signed-off-by: Gregory Etelson <getelson@nvidia.com>2
> > 
> > Typo above
> > 
> > > Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
> > 
> > This patch is breaking compilation in non-DV mode.
> > You should move it at the end of this series, after moving all in DV
> > areas.
> 
> Patch "[PATCH v2 5/5] net/mlx5: fix non-dv compilation errors" in that series
> resolves the compilation error.
> To restore tunnel offload functionality all 5 patches must be applied.

All patches must compile.
That's why I recommend moving this patch later in the series.



^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-14 17:53       ` Thomas Monjalon
@ 2020-11-14 18:31         ` Gregory Etelson
  2020-11-14 19:24           ` Thomas Monjalon
  0 siblings, 1 reply; 16+ messages in thread
From: Gregory Etelson @ 2020-11-14 18:31 UTC (permalink / raw)
  To: NBU-Contact-Thomas Monjalon
  Cc: Matan Azrad, Slava Ovsiienko, dev, Raslan Darawsheh,
	Shahaf Shuler, Asaf Penso

> 14/11/2020 18:41, Gregory Etelson:
> > > 13/11/2020 15:52, Gregory Etelson:
> > > > +       ret = mlx5_flow_group_to_table(dev, tunnel, jump_data-
> >group,
> > > > +                                      &flow_table, grp_info,
> > > > + error);
> > >
> > > The parameter grp_info is a struct passed as value.
> > > I believe it should be passed as a pointer.
> >
> > struct flow_grp_info is a 64 bit-field:
> > struct flow_grp_info {
> >       uint64_t external:1;
> >       uint64_t transfer:1;
> >       uint64_t fdb_def_rule:1;
> >       /* force standard group translation */
> >       uint64_t std_tbl_fix:1;
> >       uint64_t skip_scale:1;
> > };
> > Since mlx5_flow_group_to_table() does not change bits configuration,
> > there is no need to pass this type as a pointer.
> 
> I feel passing struct as pointer is a better practice.
>

The parameter in question is 64 bit unsigned long value.
Structure coating is a syntactic sugar, because
C language does not have bit-field types.
In general, if structure size does not exceed 64 bit it can be passed
by value.
Passing it by reference would create unnecessary indirect access.

> > > I see some other functions are passing arrays with [] syntax, which
> > > does not make sense in parameters. I should be a simple pointer.
> >
> > I agree that arrays should not be passed with [] syntax.
> > However, I kept existing rte flow style - see rte_flow_create
> 
> I recommend not following this bad code style.
> Later other occurences should be fixed.
> 
> > > Matan, Slava, as maintainers, what are your policies in mlx5 code?
> 
> 


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-14 18:31         ` Gregory Etelson
@ 2020-11-14 19:24           ` Thomas Monjalon
  2020-11-15  5:47             ` Gregory Etelson
  0 siblings, 1 reply; 16+ messages in thread
From: Thomas Monjalon @ 2020-11-14 19:24 UTC (permalink / raw)
  To: Gregory Etelson
  Cc: Matan Azrad, Slava Ovsiienko, dev, Raslan Darawsheh,
	Shahaf Shuler, Asaf Penso

14/11/2020 19:31, Gregory Etelson:
> > 14/11/2020 18:41, Gregory Etelson:
> > > > 13/11/2020 15:52, Gregory Etelson:
> > > > > +       ret = mlx5_flow_group_to_table(dev, tunnel, jump_data-
> > >group,
> > > > > +                                      &flow_table, grp_info,
> > > > > + error);
> > > >
> > > > The parameter grp_info is a struct passed as value.
> > > > I believe it should be passed as a pointer.
> > >
> > > struct flow_grp_info is a 64 bit-field:
> > > struct flow_grp_info {
> > >       uint64_t external:1;
> > >       uint64_t transfer:1;
> > >       uint64_t fdb_def_rule:1;
> > >       /* force standard group translation */
> > >       uint64_t std_tbl_fix:1;
> > >       uint64_t skip_scale:1;
> > > };
> > > Since mlx5_flow_group_to_table() does not change bits configuration,
> > > there is no need to pass this type as a pointer.
> > 
> > I feel passing struct as pointer is a better practice.
> 
> The parameter in question is 64 bit unsigned long value.
> Structure coating is a syntactic sugar, because
> C language does not have bit-field types.
> In general, if structure size does not exceed 64 bit it can be passed
> by value.
> Passing it by reference would create unnecessary indirect access.

Did you measure a performance difference?

This current code triggers this compiler note:
drivers/net/mlx5/mlx5_flow.c:7106:1: note:
parameter passing for argument of type 'struct flow_grp_info' changed in GCC 9.1

I'm afraid it can be a problem.
In general we don't have such note on the whole DPDK code base.



^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors
  2020-11-14 19:24           ` Thomas Monjalon
@ 2020-11-15  5:47             ` Gregory Etelson
  0 siblings, 0 replies; 16+ messages in thread
From: Gregory Etelson @ 2020-11-15  5:47 UTC (permalink / raw)
  To: NBU-Contact-Thomas Monjalon
  Cc: Matan Azrad, Slava Ovsiienko, dev, Raslan Darawsheh,
	Shahaf Shuler, Asaf Penso

> 14/11/2020 19:31, Gregory Etelson:
> > > 14/11/2020 18:41, Gregory Etelson:
> > > > > 13/11/2020 15:52, Gregory Etelson:
> > > > > > +       ret = mlx5_flow_group_to_table(dev, tunnel, jump_data-
> > > >group,
> > > > > > +                                      &flow_table, grp_info,
> > > > > > + error);
> > > > >
> > > > > The parameter grp_info is a struct passed as value.
> > > > > I believe it should be passed as a pointer.
> > > >
> > > > struct flow_grp_info is a 64 bit-field:
> > > > struct flow_grp_info {
> > > >       uint64_t external:1;
> > > >       uint64_t transfer:1;
> > > >       uint64_t fdb_def_rule:1;
> > > >       /* force standard group translation */
> > > >       uint64_t std_tbl_fix:1;
> > > >       uint64_t skip_scale:1;
> > > > };
> > > > Since mlx5_flow_group_to_table() does not change bits
> > > > configuration, there is no need to pass this type as a pointer.
> > >
> > > I feel passing struct as pointer is a better practice.
> >
> > The parameter in question is 64 bit unsigned long value.
> > Structure coating is a syntactic sugar, because C language does not
> > have bit-field types.
> > In general, if structure size does not exceed 64 bit it can be passed
> > by value.
> > Passing it by reference would create unnecessary indirect access.
> 
> Did you measure a performance difference?
> 
> This current code triggers this compiler note:
> drivers/net/mlx5/mlx5_flow.c:7106:1: note:
> parameter passing for argument of type 'struct flow_grp_info' changed in
> GCC 9.1
> 
> I'm afraid it can be a problem.
> In general we don't have such note on the whole DPDK code base.
> 

I'll handle this issue in DV compilation patch.

^ permalink raw reply	[flat|nested] 16+ messages in thread

end of thread, other threads:[~2020-11-15  5:48 UTC | newest]

Thread overview: 16+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-11-13 14:52 [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Gregory Etelson
2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 1/5] net/mlx5: fix tunnel offload object allocation Gregory Etelson
2020-11-13 19:00   ` Thomas Monjalon
2020-11-14 17:47     ` Gregory Etelson
2020-11-14 17:55       ` Thomas Monjalon
2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 2/5] net/mlx5: fix tunnel offload hub multi-thread protection Gregory Etelson
2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 3/5] net/mlx5: fix double table referencing Gregory Etelson
2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 4/5] net/mlx5: fix tunnel offload callback names Gregory Etelson
2020-11-13 14:52 ` [dpdk-dev] [PATCH v2 5/5] net/mlx5: fix non-dv compilation errors Gregory Etelson
2020-11-14 12:06   ` Thomas Monjalon
2020-11-14 17:41     ` Gregory Etelson
2020-11-14 17:53       ` Thomas Monjalon
2020-11-14 18:31         ` Gregory Etelson
2020-11-14 19:24           ` Thomas Monjalon
2020-11-15  5:47             ` Gregory Etelson
2020-11-13 19:03 ` [dpdk-dev] [PATCH v2 0/5] restore tunnel offload functionality in mlx5 Thomas Monjalon

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).