IB/IPoIB: Don't update neigh validity for unresolved entries
authorErez Shitrit <erezsh@mellanox.com>
Sat, 4 Jun 2016 12:15:19 +0000 (15:15 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 20 Aug 2016 16:09:25 +0000 (18:09 +0200)
commit 61c78eea9516a921799c17b4c20558e2aa780fd3 upstream.

ipoib_neigh_get unconditionally updates the "alive" variable member on
any packet send.  This prevents the neighbor garbage collection from
cleaning out a dead neighbor entry if we are still queueing packets
for it.  If the queue for this neighbor is full, then don't update the
alive timestamp.  That way the neighbor can time out even if packets
are still being queued as long as none of them are being sent.

Fixes: b63b70d87741 ("IPoIB: Use a private hash table for path lookup in xmit path")
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/infiniband/ulp/ipoib/ipoib_main.c

index 7d3281866ffcd520d4bca543a9ad0544f2ec159f..942dffca6a9d16d194ab11d73d5b9200f8f27198 100644 (file)
@@ -1131,7 +1131,9 @@ struct ipoib_neigh *ipoib_neigh_get(struct net_device *dev, u8 *daddr)
                                neigh = NULL;
                                goto out_unlock;
                        }
-                       neigh->alive = jiffies;
+
+                       if (likely(skb_queue_len(&neigh->queue) < IPOIB_MAX_PATH_REC_QUEUE))
+                               neigh->alive = jiffies;
                        goto out_unlock;
                }
        }