5 files changed, 434 insertions, 43 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index 52ea7f1c0973..730de6b7e46e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -34,6 +34,8 @@
 #include <linux/mlx5/fs.h>
 #include <net/switchdev.h>
 #include <net/pkt_cls.h>
+#include <net/netevent.h>
+#include <net/arp.h>
 
 #include "eswitch.h"
 #include "en.h"
@@ -224,6 +226,140 @@ void mlx5e_remove_sqs_fwd_rules(struct mlx5e_priv *priv)
 	mlx5_eswitch_sqs2vport_stop(esw, rep);
 }
 
+static void mlx5e_rep_neigh_entry_hold(struct mlx5e_neigh_hash_entry *nhe)
+{
+	refcount_inc(&nhe->refcnt);
+}
+
+static void mlx5e_rep_neigh_entry_release(struct mlx5e_neigh_hash_entry *nhe)
+{
+	if (refcount_dec_and_test(&nhe->refcnt))
+		kfree(nhe);
+}
+
+static void mlx5e_rep_update_flows(struct mlx5e_priv *priv,
+				   struct mlx5e_encap_entry *e,
+				   bool neigh_connected,
+				   unsigned char ha[ETH_ALEN])
+{
+	struct ethhdr *eth = (struct ethhdr *)e->encap_header;
+
+	ASSERT_RTNL();
+
+	if ((!neigh_connected && (e->flags & MLX5_ENCAP_ENTRY_VALID)) ||
+	    !ether_addr_equal(e->h_dest, ha))
+		mlx5e_tc_encap_flows_del(priv, e);
+
+	if (neigh_connected && !(e->flags & MLX5_ENCAP_ENTRY_VALID)) {
+		ether_addr_copy(e->h_dest, ha);
+		ether_addr_copy(eth->h_dest, ha);
+
+		mlx5e_tc_encap_flows_add(priv, e);
+	}
+}
+
+static void mlx5e_rep_neigh_update(struct work_struct *work)
+{
+	struct mlx5e_neigh_hash_entry *nhe =
+		container_of(work, struct mlx5e_neigh_hash_entry, neigh_update_work);
+	struct neighbour *n = nhe->n;
+	struct mlx5e_encap_entry *e;
+	unsigned char ha[ETH_ALEN];
+	struct mlx5e_priv *priv;
+	bool neigh_connected;
+	bool encap_connected;
+	u8 nud_state, dead;
+
+	rtnl_lock();
+
+	/* If these parameters are changed after we release the lock,
+	 * we'll receive another event letting us know about it.
+	 * We use this lock to avoid inconsistency between the neigh validity
+	 * and it's hw address.
+	 */
+	read_lock_bh(&n->lock);
+	memcpy(ha, n->ha, ETH_ALEN);
+	nud_state = n->nud_state;
+	dead = n->dead;
+	read_unlock_bh(&n->lock);
+
+	neigh_connected = (nud_state & NUD_VALID) && !dead;
+
+	list_for_each_entry(e, &nhe->encap_list, encap_list) {
+		encap_connected = !!(e->flags & MLX5_ENCAP_ENTRY_VALID);
+		priv = netdev_priv(e->out_dev);
+
+		if (encap_connected != neigh_connected ||
+		    !ether_addr_equal(e->h_dest, ha))
+			mlx5e_rep_update_flows(priv, e, neigh_connected, ha);
+	}
+	mlx5e_rep_neigh_entry_release(nhe);
+	rtnl_unlock();
+	neigh_release(n);
+}
+
+static struct mlx5e_neigh_hash_entry *
+mlx5e_rep_neigh_entry_lookup(struct mlx5e_priv *priv,
+			     struct mlx5e_neigh *m_neigh);
+
+static int mlx5e_rep_netevent_event(struct notifier_block *nb,
+				    unsigned long event, void *ptr)
+{
+	struct mlx5e_rep_priv *rpriv = container_of(nb, struct mlx5e_rep_priv,
+						    neigh_update.netevent_nb);
+	struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
+	struct net_device *netdev = rpriv->rep->netdev;
+	struct mlx5e_priv *priv = netdev_priv(netdev);
+	struct mlx5e_neigh_hash_entry *nhe = NULL;
+	struct mlx5e_neigh m_neigh = {};
+	struct neighbour *n;
+
+	switch (event) {
+	case NETEVENT_NEIGH_UPDATE:
+		n = ptr;
+#if IS_ENABLED(CONFIG_IPV6)
+		if (n->tbl != ipv6_stub->nd_tbl && n->tbl != &arp_tbl)
+#else
+		if (n->tbl != &arp_tbl)
+#endif
+			return NOTIFY_DONE;
+
+		m_neigh.dev = n->dev;
+		memcpy(&m_neigh.dst_ip, n->primary_key, n->tbl->key_len);
+
+		/* We are in atomic context and can't take RTNL mutex, so use
+		 * spin_lock_bh to lookup the neigh table. bh is used since
+		 * netevent can be called from a softirq context.
+		 */
+		spin_lock_bh(&neigh_update->encap_lock);
+		nhe = mlx5e_rep_neigh_entry_lookup(priv, &m_neigh);
+		if (!nhe) {
+			spin_unlock_bh(&neigh_update->encap_lock);
+			return NOTIFY_DONE;
+		}
+
+		/* This assignment is valid as long as the the neigh reference
+		 * is taken
+		 */
+		nhe->n = n;
+
+		/* Take a reference to ensure the neighbour and mlx5 encap
+		 * entry won't be destructed until we drop the reference in
+		 * delayed work.
+		 */
+		neigh_hold(n);
+		mlx5e_rep_neigh_entry_hold(nhe);
+
+		if (!queue_work(priv->wq, &nhe->neigh_update_work)) {
+			mlx5e_rep_neigh_entry_release(nhe);
+			neigh_release(n);
+		}
+		spin_unlock_bh(&neigh_update->encap_lock);
+		break;
+	}
+	return NOTIFY_DONE;
+}
+
 static const struct rhashtable_params mlx5e_neigh_ht_params = {
 	.head_offset = offsetof(struct mlx5e_neigh_hash_entry, rhash_node),
 	.key_offset = offsetof(struct mlx5e_neigh_hash_entry, m_neigh),
@@ -234,14 +370,34 @@ static const struct rhashtable_params mlx5e_neigh_ht_params = {
 static int mlx5e_rep_neigh_init(struct mlx5e_rep_priv *rpriv)
 {
 	struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
+	int err;
+
+	err = rhashtable_init(&neigh_update->neigh_ht, &mlx5e_neigh_ht_params);
+	if (err)
+		return err;
 
 	INIT_LIST_HEAD(&neigh_update->neigh_list);
-	return rhashtable_init(&neigh_update->neigh_ht, &mlx5e_neigh_ht_params);
+	spin_lock_init(&neigh_update->encap_lock);
+
+	rpriv->neigh_update.netevent_nb.notifier_call = mlx5e_rep_netevent_event;
+	err = register_netevent_notifier(&rpriv->neigh_update.netevent_nb);
+	if (err)
+		goto out_err;
+	return 0;
+
+out_err:
+	rhashtable_destroy(&neigh_update->neigh_ht);
+	return err;
 }
 
 static void mlx5e_rep_neigh_cleanup(struct mlx5e_rep_priv *rpriv)
 {
 	struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
+	struct mlx5e_priv *priv = netdev_priv(rpriv->rep->netdev);
+
+	unregister_netevent_notifier(&neigh_update->netevent_nb);
+
+	flush_workqueue(priv->wq); /* flush neigh update works */
 
 	rhashtable_destroy(&neigh_update->neigh_ht);
 }
@@ -268,13 +424,19 @@ static void mlx5e_rep_neigh_entry_remove(struct mlx5e_priv *priv,
 {
 	struct mlx5e_rep_priv *rpriv = priv->ppriv;
 
+	spin_lock_bh(&rpriv->neigh_update.encap_lock);
+
 	list_del(&nhe->neigh_list);
 
 	rhashtable_remove_fast(&rpriv->neigh_update.neigh_ht,
 			       &nhe->rhash_node,
 			       mlx5e_neigh_ht_params);
+	spin_unlock_bh(&rpriv->neigh_update.encap_lock);
 }
 
+/* This function must only be called under RTNL lock or under the
+ * representor's encap_lock in case RTNL mutex can't be held.
+ */
 static struct mlx5e_neigh_hash_entry *
 mlx5e_rep_neigh_entry_lookup(struct mlx5e_priv *priv,
 			     struct mlx5e_neigh *m_neigh)
@@ -286,6 +448,72 @@ mlx5e_rep_neigh_entry_lookup(struct mlx5e_priv *priv,
 				      mlx5e_neigh_ht_params);
 }
 
+static int mlx5e_rep_neigh_entry_create(struct mlx5e_priv *priv,
+					struct mlx5e_encap_entry *e,
+					struct mlx5e_neigh_hash_entry **nhe)
+{
+	int err;
+
+	*nhe = kzalloc(sizeof(**nhe), GFP_KERNEL);
+	if (!*nhe)
+		return -ENOMEM;
+
+	memcpy(&(*nhe)->m_neigh, &e->m_neigh, sizeof(e->m_neigh));
+	INIT_WORK(&(*nhe)->neigh_update_work, mlx5e_rep_neigh_update);
+	INIT_LIST_HEAD(&(*nhe)->encap_list);
+	refcount_set(&(*nhe)->refcnt, 1);
+
+	err = mlx5e_rep_neigh_entry_insert(priv, *nhe);
+	if (err)
+		goto out_free;
+	return 0;
+
+out_free:
+	kfree(*nhe);
+	return err;
+}
+
+static void mlx5e_rep_neigh_entry_destroy(struct mlx5e_priv *priv,
+					  struct mlx5e_neigh_hash_entry *nhe)
+{
+	/* The neigh hash entry must be removed from the hash table regardless
+	 * of the reference count value, so it won't be found by the next
+	 * neigh notification call. The neigh hash entry reference count is
+	 * incremented only during creation and neigh notification calls and
+	 * protects from freeing the nhe struct.
+	 */
+	mlx5e_rep_neigh_entry_remove(priv, nhe);
+	mlx5e_rep_neigh_entry_release(nhe);
+}
+
+int mlx5e_rep_encap_entry_attach(struct mlx5e_priv *priv,
+				 struct mlx5e_encap_entry *e)
+{
+	struct mlx5e_neigh_hash_entry *nhe;
+	int err;
+
+	nhe = mlx5e_rep_neigh_entry_lookup(priv, &e->m_neigh);
+	if (!nhe) {
+		err = mlx5e_rep_neigh_entry_create(priv, e, &nhe);
+		if (err)
+			return err;
+	}
+	list_add(&e->encap_list, &nhe->encap_list);
+	return 0;
+}
+
+void mlx5e_rep_encap_entry_detach(struct mlx5e_priv *priv,
+				  struct mlx5e_encap_entry *e)
+{
+	struct mlx5e_neigh_hash_entry *nhe;
+
+	list_del(&e->encap_list);
+	nhe = mlx5e_rep_neigh_entry_lookup(priv, &e->m_neigh);
+
+	if (list_empty(&nhe->encap_list))
+		mlx5e_rep_neigh_entry_destroy(priv, nhe);
+}
+
 static int mlx5e_rep_open(struct net_device *dev)
 {
 	struct mlx5e_priv *priv = netdev_priv(dev);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
index 99f6b5f41070..e4d0ea5246fd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
@@ -45,6 +45,9 @@ struct mlx5e_neigh_update_table {
 	 * Used for stats query.
 	 */
 	struct list_head	neigh_list;
+	/* protect lookup/remove operations */
+	spinlock_t              encap_lock;
+	struct notifier_block   netevent_nb;
 };
 
 struct mlx5e_rep_priv {
@@ -69,18 +72,46 @@ struct mlx5e_neigh_hash_entry {
 	 * neighbour entries. Used for stats query.
 	 */
 	struct list_head neigh_list;
+
+	/* encap list sharing the same neigh */
+	struct list_head encap_list;
+
+	/* valid only when the neigh reference is taken during
+	 * neigh_update_work workqueue callback.
+	 */
+	struct neighbour *n;
+	struct work_struct neigh_update_work;
+
+	/* neigh hash entry can be deleted only when the refcount is zero.
+	 * refcount is needed to avoid neigh hash entry removal by TC, while
+	 * it's used by the neigh notification call.
+	 */
+	refcount_t refcnt;
+};
+
+enum {
+	/* set when the encap entry is successfully offloaded into HW */
+	MLX5_ENCAP_ENTRY_VALID     = BIT(0),
 };
 
 struct mlx5e_encap_entry {
+	/* neigh hash entry list of encaps sharing the same neigh */
+	struct list_head encap_list;
+	struct mlx5e_neigh m_neigh;
+	/* a node of the eswitch encap hash table which keeping all the encap
+	 * entries
+	 */
 	struct hlist_node encap_hlist;
 	struct list_head flows;
 	u32 encap_id;
-	struct neighbour *n;
 	struct ip_tunnel_info tun_info;
 	unsigned char h_dest[ETH_ALEN];	/* destination eth addr	*/
 
 	struct net_device *out_dev;
 	int tunnel_type;
+	u8 flags;
+	char *encap_header;
+	int encap_size;
 };
 
 void mlx5e_register_vport_reps(struct mlx5e_priv *priv);
@@ -95,4 +126,9 @@ bool mlx5e_has_offload_stats(const struct net_device *dev, int attr_id);
 int mlx5e_attr_get(struct net_device *dev, struct switchdev_attr *attr);
 void mlx5e_handle_rx_cqe_rep(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
 
+int mlx5e_rep_encap_entry_attach(struct mlx5e_priv *priv,
+				 struct mlx5e_encap_entry *e);
+void mlx5e_rep_encap_entry_detach(struct mlx5e_priv *priv,
+				  struct mlx5e_encap_entry *e);
+
 #endif /* __MLX5E_REP_H__ */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index ae07fe6473bb..624dbfe31a0e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -45,8 +45,8 @@
 #include <net/tc_act/tc_pedit.h>
 #include <net/vxlan.h>
 #include "en.h"
-#include "en_tc.h"
 #include "en_rep.h"
+#include "en_tc.h"
 #include "eswitch.h"
 #include "vxlan.h"
 
@@ -246,19 +246,75 @@ static void mlx5e_tc_del_fdb_flow(struct mlx5e_priv *priv,
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
 	struct mlx5_esw_flow_attr *attr = flow->esw_attr;
 
-	if (flow->flags & MLX5E_TC_FLOW_OFFLOADED)
+	if (flow->flags & MLX5E_TC_FLOW_OFFLOADED) {
+		flow->flags &= ~MLX5E_TC_FLOW_OFFLOADED;
 		mlx5_eswitch_del_offloaded_rule(esw, flow->rule, flow->esw_attr);
+	}
 
 	mlx5_eswitch_del_vlan_action(esw, flow->esw_attr);
 
-	if (flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP)
+	if (flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP) {
 		mlx5e_detach_encap(priv, flow);
+		kvfree(flow->esw_attr->parse_attr);
+	}
 
 	if (flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
 		mlx5_modify_header_dealloc(priv->mdev,
 					   attr->mod_hdr_id);
 }
 
+void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv,
+			      struct mlx5e_encap_entry *e)
+{
+	struct mlx5e_tc_flow *flow;
+	int err;
+
+	err = mlx5_encap_alloc(priv->mdev, e->tunnel_type,
+			       e->encap_size, e->encap_header,
+			       &e->encap_id);
+	if (err) {
+		mlx5_core_warn(priv->mdev, "Failed to offload cached encapsulation header, %d\n",
+			       err);
+		return;
+	}
+	e->flags |= MLX5_ENCAP_ENTRY_VALID;
+
+	list_for_each_entry(flow, &e->flows, encap) {
+		flow->esw_attr->encap_id = e->encap_id;
+		flow->rule = mlx5e_tc_add_fdb_flow(priv,
+						   flow->esw_attr->parse_attr,
+						   flow);
+		if (IS_ERR(flow->rule)) {
+			err = PTR_ERR(flow->rule);
+			mlx5_core_warn(priv->mdev, "Failed to update cached encapsulation flow, %d\n",
+				       err);
+			continue;
+		}
+		flow->flags |= MLX5E_TC_FLOW_OFFLOADED;
+	}
+}
+
+void mlx5e_tc_encap_flows_del(struct mlx5e_priv *priv,
+			      struct mlx5e_encap_entry *e)
+{
+	struct mlx5e_tc_flow *flow;
+	struct mlx5_fc *counter;
+
+	list_for_each_entry(flow, &e->flows, encap) {
+		if (flow->flags & MLX5E_TC_FLOW_OFFLOADED) {
+			flow->flags &= ~MLX5E_TC_FLOW_OFFLOADED;
+			counter = mlx5_flow_rule_counter(flow->rule);
+			mlx5_del_flow_rules(flow->rule);
+			mlx5_fc_destroy(priv->mdev, counter);
+		}
+	}
+
+	if (e->flags & MLX5_ENCAP_ENTRY_VALID) {
+		e->flags &= ~MLX5_ENCAP_ENTRY_VALID;
+		mlx5_encap_dealloc(priv->mdev, e->encap_id);
+	}
+}
+
 static void mlx5e_detach_encap(struct mlx5e_priv *priv,
 			       struct mlx5e_tc_flow *flow)
 {
@@ -269,11 +325,13 @@ static void mlx5e_detach_encap(struct mlx5e_priv *priv,
 		struct mlx5e_encap_entry *e;
 
 		e = list_entry(next, struct mlx5e_encap_entry, flows);
-		if (e->n) {
+		mlx5e_rep_encap_entry_detach(netdev_priv(e->out_dev), e);
+
+		if (e->flags & MLX5_ENCAP_ENTRY_VALID)
 			mlx5_encap_dealloc(priv->mdev, e->encap_id);
-			neigh_release(e->n);
-		}
+
 		hlist_del_rcu(&e->encap_hlist);
+		kfree(e->encap_header);
 		kfree(e);
 	}
 }
@@ -1253,20 +1311,27 @@ static int mlx5e_create_encap_header_ipv4(struct mlx5e_priv *priv,
 	if (err)
 		goto out;
 
+	/* used by mlx5e_detach_encap to lookup a neigh hash table
+	 * entry in the neigh hash table when a user deletes a rule
+	 */
+	e->m_neigh.dev = n->dev;
+	memcpy(&e->m_neigh.dst_ip, n->primary_key, n->tbl->key_len);
+	e->out_dev = out_dev;
+
+	/* It's importent to add the neigh to the hash table before checking
+	 * the neigh validity state. So if we'll get a notification, in case the
+	 * neigh changes it's validity state, we would find the relevant neigh
+	 * in the hash.
+	 */
+	err = mlx5e_rep_encap_entry_attach(netdev_priv(out_dev), e);
+	if (err)
+		goto out;
+
 	read_lock_bh(&n->lock);
 	nud_state = n->nud_state;
 	ether_addr_copy(e->h_dest, n->ha);
 	read_unlock_bh(&n->lock);
 
-	if (!(nud_state & NUD_VALID)) {
-		pr_warn("%s: can't offload, neighbour to %pI4 invalid\n", __func__, &fl4.daddr);
-		err = -EOPNOTSUPP;
-		goto out;
-	}
-
-	e->n = n;
-	e->out_dev = out_dev;
-
 	switch (e->tunnel_type) {
 	case MLX5_HEADER_TYPE_VXLAN:
 		gen_vxlan_header_ipv4(out_dev, encap_header,
@@ -1277,15 +1342,32 @@ static int mlx5e_create_encap_header_ipv4(struct mlx5e_priv *priv,
 		break;
 	default:
 		err = -EOPNOTSUPP;
-		goto out;
+		goto destroy_neigh_entry;
+	}
+	e->encap_size = ipv4_encap_size;
+	e->encap_header = encap_header;
+
+	if (!(nud_state & NUD_VALID)) {
+		neigh_event_send(n, NULL);
+		neigh_release(n);
+		return -EAGAIN;
 	}
 
 	err = mlx5_encap_alloc(priv->mdev, e->tunnel_type,
 			       ipv4_encap_size, encap_header, &e->encap_id);
+	if (err)
+		goto destroy_neigh_entry;
+
+	e->flags |= MLX5_ENCAP_ENTRY_VALID;
+	neigh_release(n);
+	return err;
+
+destroy_neigh_entry:
+	mlx5e_rep_encap_entry_detach(netdev_priv(e->out_dev), e);
 out:
-	if (err && n)
-		neigh_release(n);
 	kfree(encap_header);
+	if (n)
+		neigh_release(n);
 	return err;
 }
 
@@ -1332,20 +1414,27 @@ static int mlx5e_create_encap_header_ipv6(struct mlx5e_priv *priv,
 	if (err)
 		goto out;
 
+	/* used by mlx5e_detach_encap to lookup a neigh hash table
+	 * entry in the neigh hash table when a user deletes a rule
+	 */
+	e->m_neigh.dev = n->dev;
+	memcpy(&e->m_neigh.dst_ip, n->primary_key, n->tbl->key_len);
+	e->out_dev = out_dev;
+
+	/* It's importent to add the neigh to the hash table before checking
+	 * the neigh validity state. So if we'll get a notification, in case the
+	 * neigh changes it's validity state, we would find the relevant neigh
+	 * in the hash.
+	 */
+	err = mlx5e_rep_encap_entry_attach(netdev_priv(out_dev), e);
+	if (err)
+		goto out;
+
 	read_lock_bh(&n->lock);
 	nud_state = n->nud_state;
 	ether_addr_copy(e->h_dest, n->ha);
 	read_unlock_bh(&n->lock);
 
-	if (!(nud_state & NUD_VALID)) {
-		pr_warn("%s: can't offload, neighbour to %pI6 invalid\n", __func__, &fl6.daddr);
-		err = -EOPNOTSUPP;
-		goto out;
-	}
-
-	e->n = n;
-	e->out_dev = out_dev;
-
 	switch (e->tunnel_type) {
 	case MLX5_HEADER_TYPE_VXLAN:
 		gen_vxlan_header_ipv6(out_dev, encap_header,
@@ -1356,15 +1445,33 @@ static int mlx5e_create_encap_header_ipv6(struct mlx5e_priv *priv,
 		break;
 	default:
 		err = -EOPNOTSUPP;
-		goto out;
+		goto destroy_neigh_entry;
+	}
+
+	e->encap_size = ipv6_encap_size;
+	e->encap_header = encap_header;
+
+	if (!(nud_state & NUD_VALID)) {
+		neigh_event_send(n, NULL);
+		neigh_release(n);
+		return -EAGAIN;
 	}
 
 	err = mlx5_encap_alloc(priv->mdev, e->tunnel_type,
 			       ipv6_encap_size, encap_header, &e->encap_id);
+	if (err)
+		goto destroy_neigh_entry;
+
+	e->flags |= MLX5_ENCAP_ENTRY_VALID;
+	neigh_release(n);
+	return err;
+
+destroy_neigh_entry:
+	mlx5e_rep_encap_entry_detach(netdev_priv(e->out_dev), e);
 out:
-	if (err && n)
-		neigh_release(n);
 	kfree(encap_header);
+	if (n)
+		neigh_release(n);
 	return err;
 }
 
@@ -1432,7 +1539,7 @@ vxlan_encap_offload_err:
 	else if (family == AF_INET6)
 		err = mlx5e_create_encap_header_ipv6(priv, mirred_dev, e);
 
-	if (err)
+	if (err && err != -EAGAIN)
 		goto out_err;
 
 	hash_add_rcu(esw->offloads.encap_tbl, &e->encap_hlist, hash_key);
@@ -1440,9 +1547,10 @@ vxlan_encap_offload_err:
 attach_flow:
 	list_add(&flow->encap, &e->flows);
 	*encap_dev = e->out_dev;
-	attr->encap_id = e->encap_id;
+	if (e->flags & MLX5_ENCAP_ENTRY_VALID)
+		attr->encap_id = e->encap_id;
 
-	return 0;
+	return err;
 
 out_err:
 	kfree(e);
@@ -1459,7 +1567,7 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 	const struct tc_action *a;
 	LIST_HEAD(actions);
 	bool encap = false;
-	int err;
+	int err = 0;
 
 	if (tc_no_actions(exts))
 		return -EINVAL;
@@ -1502,7 +1610,7 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 			} else if (encap) {
 				err = mlx5e_attach_encap(priv, info,
 							 out_dev, &encap_dev, flow);
-				if (err)
+				if (err && err != -EAGAIN)
 					return err;
 				attr->action |= MLX5_FLOW_CONTEXT_ACTION_ENCAP |
 					MLX5_FLOW_CONTEXT_ACTION_FWD_DEST |
@@ -1510,6 +1618,7 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 				out_priv = netdev_priv(encap_dev);
 				rpriv = out_priv->ppriv;
 				attr->out_rep = rpriv->rep;
+				attr->parse_attr = parse_attr;
 			} else {
 				pr_err("devices %s %s not on same switch HW, can't offload forwarding\n",
 				       priv->netdev->name, out_dev->name);
@@ -1549,7 +1658,7 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 
 		return -EINVAL;
 	}
-	return 0;
+	return err;
 }
 
 int mlx5e_configure_flower(struct mlx5e_priv *priv, __be16 protocol,
@@ -1587,7 +1696,7 @@ int mlx5e_configure_flower(struct mlx5e_priv *priv, __be16 protocol,
 	if (flow->flags & MLX5E_TC_FLOW_ESWITCH) {
 		err = parse_tc_fdb_actions(priv, f->exts, parse_attr, flow);
 		if (err < 0)
-			goto err_free;
+			goto err_handle_encap_flow;
 		flow->rule = mlx5e_tc_add_fdb_flow(priv, parse_attr, flow);
 	} else {
 		err = parse_tc_nic_actions(priv, f->exts, parse_attr, flow);
@@ -1607,15 +1716,27 @@ int mlx5e_configure_flower(struct mlx5e_priv *priv, __be16 protocol,
 	if (err)
 		goto err_del_rule;
 
-	goto out;
+	if (flow->flags & MLX5E_TC_FLOW_ESWITCH &&
+	    !(flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP))
+		kvfree(parse_attr);
+	return err;
 
 err_del_rule:
 	mlx5e_tc_del_flow(priv, flow);
 
+err_handle_encap_flow:
+	if (err == -EAGAIN) {
+		err = rhashtable_insert_fast(&tc->ht, &flow->node,
+					     tc->ht_params);
+		if (err)
+			mlx5e_tc_del_flow(priv, flow);
+		else
+			return 0;
+	}
+
 err_free:
-	kfree(flow);
-out:
 	kvfree(parse_attr);
+	kfree(flow);
 	return err;
 }
 
@@ -1634,7 +1755,6 @@ int mlx5e_delete_flower(struct mlx5e_priv *priv,
 
 	mlx5e_tc_del_flow(priv, flow);
 
-
 	kfree(flow);
 
 	return 0;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
index 34bf903fc886..278c7a646a55 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
@@ -46,6 +46,12 @@ int mlx5e_delete_flower(struct mlx5e_priv *priv,
 int mlx5e_stats_flower(struct mlx5e_priv *priv,
 		       struct tc_cls_flower_offload *f);
 
+struct mlx5e_encap_entry;
+void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv,
+			      struct mlx5e_encap_entry *e);
+void mlx5e_tc_encap_flows_del(struct mlx5e_priv *priv,
+			      struct mlx5e_encap_entry *e);
+
 static inline int mlx5e_tc_num_filters(struct mlx5e_priv *priv)
 {
 	return atomic_read(&priv->fs.tc.ht.nelems);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 751a673de97a..55beda6bf134 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -297,6 +297,7 @@ struct mlx5_esw_flow_attr {
 	bool	vlan_handled;
 	u32	encap_id;
 	u32	mod_hdr_id;
+	struct mlx5e_tc_flow_parse_attr *parse_attr;
 };
 
 int mlx5_eswitch_sqs2vport_start(struct mlx5_eswitch *esw,