From: Vlad Buslov <vladbu@mellanox.com>
Date: Fri, 8 Jun 2018 11:31:28 +0300
Subject: net/mlx5e: Always take reference to neigh entry
Patch-mainline: v5.4-rc1
Git-commit: 61081f9c09df33fe0123d234e840e3b4d3c17269
References: jsc#SLE-8464

Neigh entry has reference counter, however it is only used when scheduling
neigh update event. In all other cases reference to neigh entry is not
taken while working with it. Neigh code relies on synchronization provided
by rtnl lock and uses encap list size as implicit reference counter.

To remove dependency on rtnl lock, always take reference to neigh entry
while using it. Remove neigh entry from hash table and delete it only when
reference counter reaches zero. This can result spurious neigh update
events, when there is an event on entry that has zero encaps attached.
However, such events are rare and properly handled by neigh update handler.

Extend encap entry with reference to neigh hash entry in order to be able
to directly release it when encap is detached, instead of lookup nhe by key
through hash table. Extend nhe with reference to device priv structure to
guarantee correctness when nhe is used with stack devices, bond setup, in
which case it is non-trivial to determine correct device when releasing the
nhe.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 drivers/net/ethernet/mellanox/mlx5/core/en_rep.c |   76 ++++++++++-------------
 drivers/net/ethernet/mellanox/mlx5/core/en_rep.h |    3 
 2 files changed, 39 insertions(+), 40 deletions(-)

--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -524,6 +524,21 @@ void mlx5e_rep_queue_neigh_stats_work(st
 				 neigh_update->min_interval);
 }
 
+static bool mlx5e_rep_neigh_entry_hold(struct mlx5e_neigh_hash_entry *nhe)
+{
+	return refcount_inc_not_zero(&nhe->refcnt);
+}
+
+static void mlx5e_rep_neigh_entry_remove(struct mlx5e_neigh_hash_entry *nhe);
+
+static void mlx5e_rep_neigh_entry_release(struct mlx5e_neigh_hash_entry *nhe)
+{
+	if (refcount_dec_and_test(&nhe->refcnt)) {
+		mlx5e_rep_neigh_entry_remove(nhe);
+		kfree(nhe);
+	}
+}
+
 static void mlx5e_rep_neigh_stats_work(struct work_struct *work)
 {
 	struct mlx5e_rep_priv *rpriv = container_of(work, struct mlx5e_rep_priv,
@@ -536,23 +551,16 @@ static void mlx5e_rep_neigh_stats_work(s
 	if (!list_empty(&rpriv->neigh_update.neigh_list))
 		mlx5e_rep_queue_neigh_stats_work(priv);
 
-	list_for_each_entry(nhe, &rpriv->neigh_update.neigh_list, neigh_list)
-		mlx5e_tc_update_neigh_used_value(nhe);
+	list_for_each_entry(nhe, &rpriv->neigh_update.neigh_list, neigh_list) {
+		if (mlx5e_rep_neigh_entry_hold(nhe)) {
+			mlx5e_tc_update_neigh_used_value(nhe);
+			mlx5e_rep_neigh_entry_release(nhe);
+		}
+	}
 
 	rtnl_unlock();
 }
 
-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,
@@ -881,14 +889,11 @@ static int mlx5e_rep_netevent_event(stru
 		 */
 		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);
+		spin_unlock_bh(&neigh_update->encap_lock);
+		if (!nhe)
 			return NOTIFY_DONE;
-		}
 
-		mlx5e_rep_neigh_entry_hold(nhe);
 		mlx5e_rep_queue_neigh_update_work(priv, nhe, n);
-		spin_unlock_bh(&neigh_update->encap_lock);
 		break;
 
 	case NETEVENT_DELAY_PROBE_TIME_UPDATE:
@@ -995,10 +1000,9 @@ static int mlx5e_rep_neigh_entry_insert(
 	return err;
 }
 
-static void mlx5e_rep_neigh_entry_remove(struct mlx5e_priv *priv,
-					 struct mlx5e_neigh_hash_entry *nhe)
+static void mlx5e_rep_neigh_entry_remove(struct mlx5e_neigh_hash_entry *nhe)
 {
-	struct mlx5e_rep_priv *rpriv = priv->ppriv;
+	struct mlx5e_rep_priv *rpriv = nhe->priv->ppriv;
 
 	spin_lock_bh(&rpriv->neigh_update.encap_lock);
 
@@ -1019,9 +1023,11 @@ mlx5e_rep_neigh_entry_lookup(struct mlx5
 {
 	struct mlx5e_rep_priv *rpriv = priv->ppriv;
 	struct mlx5e_neigh_update_table *neigh_update = &rpriv->neigh_update;
+	struct mlx5e_neigh_hash_entry *nhe;
 
-	return rhashtable_lookup_fast(&neigh_update->neigh_ht, m_neigh,
-				      mlx5e_neigh_ht_params);
+	nhe = rhashtable_lookup_fast(&neigh_update->neigh_ht, m_neigh,
+				     mlx5e_neigh_ht_params);
+	return nhe && mlx5e_rep_neigh_entry_hold(nhe) ? nhe : NULL;
 }
 
 static int mlx5e_rep_neigh_entry_create(struct mlx5e_priv *priv,
@@ -1034,6 +1040,7 @@ static int mlx5e_rep_neigh_entry_create(
 	if (!*nhe)
 		return -ENOMEM;
 
+	(*nhe)->priv = priv;
 	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);
@@ -1049,19 +1056,6 @@ out_free:
 	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)
 {
@@ -1083,6 +1077,7 @@ int mlx5e_rep_encap_entry_attach(struct
 			return err;
 		}
 	}
+	e->nhe = nhe;
 	list_add(&e->encap_list, &nhe->encap_list);
 	return 0;
 }
@@ -1093,13 +1088,14 @@ void mlx5e_rep_encap_entry_detach(struct
 	struct mlx5e_rep_priv *rpriv = priv->ppriv;
 	struct mlx5_rep_uplink_priv *uplink_priv = &rpriv->uplink_priv;
 	struct mlx5_tun_entropy *tun_entropy = &uplink_priv->tun_entropy;
-	struct mlx5e_neigh_hash_entry *nhe;
+
+	if (!e->nhe)
+		return;
 
 	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);
+	mlx5e_rep_neigh_entry_release(e->nhe);
+	e->nhe = NULL;
 	mlx5_tun_entropy_refcount_dec(tun_entropy, e->reformat_type);
 }
 
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
@@ -110,6 +110,7 @@ struct mlx5e_neigh {
 struct mlx5e_neigh_hash_entry {
 	struct rhash_head rhash_node;
 	struct mlx5e_neigh m_neigh;
+	struct mlx5e_priv *priv;
 
 	/* Save the neigh hash entry in a list on the representor in
 	 * addition to the hash table. In order to iterate easily over the
@@ -145,6 +146,8 @@ enum {
 };
 
 struct mlx5e_encap_entry {
+	/* attached neigh hash entry */
+	struct mlx5e_neigh_hash_entry *nhe;
 	/* neigh hash entry list of encaps sharing the same neigh */
 	struct list_head encap_list;
 	struct mlx5e_neigh m_neigh;