RDMA/addr: Handle ethernet neighbour updates during route resolution
authorSteve Wise <swise@opengridcomputing.com>
Wed, 24 Jan 2007 01:03:17 +0000 (19:03 -0600)
committerRoland Dreier <rolandd@cisco.com>
Sun, 4 Feb 2007 22:11:57 +0000 (14:11 -0800)
The iWARP connection manager uses the ib_addr services to do route
resolution (neighbour discovery in the IP world).  The ib_addr
netevent callback routine, however, currently only acts on InfiniBand
neighbour updates.  It needs to act on ethernet neighbour updates as
well.

This patch just removes filtering on device type altogether and will
trigger on any neighour updates where the nud_type is valid.  This
simplifies the code some.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
drivers/infiniband/core/addr.c

index af939796750dba07fd713f83dc3cf473ce52c43e..d2bb5a9a303fcb7d86f3a38deea2e33e7d8e9c2c 100644 (file)
@@ -360,8 +360,7 @@ static int netevent_callback(struct notifier_block *self, unsigned long event,
        if (event == NETEVENT_NEIGH_UPDATE) {
                struct neighbour *neigh = ctx;
 
-               if (neigh->dev->type == ARPHRD_INFINIBAND &&
-                   (neigh->nud_state & NUD_VALID)) {
+               if (neigh->nud_state & NUD_VALID) {
                        set_timeout(jiffies);
                }
        }