infiniband: addr: Avoid unnecessary neigh lookup.

Submitted by David Miller on Dec. 2, 2011, 7:55 p.m.

Details

Message ID 20111202.145525.1679807360238725746.davem@davemloft.net
State Superseded
Delegated to: David Miller
Headers show

Commit Message

David Miller Dec. 2, 2011, 7:55 p.m.
The explicit neigh_lookup() in addr4_resolve() is unnecessary because this
is the exact same calculation the routing code already made to attach the
neigh to the route.

Therefore, just use dst_get_neighbour().

Signed-off-by: David S. Miller <davem@davemloft.net>
---

Roland, I'm auditing all of the uses of dst_get_neighbour() with the
end result being that a dst_put_neighbour() interface will be added
and all access will be reduced to quick:

	rcu_read_lock();
	n = dst_get_neighbour();
	...
	dst_put_neighbour();
	rcu_read_unlock();

sequences.  And then, at the next step, things will be further
whittled down into a refcount-less variant:

	rcu_read_lock();
	n = dst_get_neighbour_noref();
	...
	rcu_read_unlock();

Another key difference is that we are making the code able to handle
dst_get_neighbour() returning NULL.

So if it's OK with you I'd like to merge this stuff directly in the
net-next tree after getting your ACK on each patch.

In the case here, the neigh_lookup() call is spurious and is exactly
mimicking the neigh lookup that the ipv4 routing code does to attach
the neighbour to the route.  There is no need to duplicate that
operation, and we can thus use the route attached neigh directly.

 drivers/infiniband/core/addr.c |   21 ++++++++++-----------
 1 files changed, 10 insertions(+), 11 deletions(-)

Comments

Roland Dreier Dec. 2, 2011, 11:18 p.m.
> So if it's OK with you I'd like to merge this stuff directly in the
> net-next tree after getting your ACK on each patch.

Sure, that's fine, and for this particular patch:

Acked-by: Roland Dreier <roland@purestorage.com>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller Dec. 3, 2011, 2:48 a.m.
From: Roland Dreier <roland@kernel.org>
Date: Fri, 2 Dec 2011 15:18:00 -0800

>> So if it's OK with you I'd like to merge this stuff directly in the
>> net-next tree after getting your ACK on each patch.
> 
> Sure, that's fine, and for this particular patch:
> 
> Acked-by: Roland Dreier <roland@purestorage.com>

Thanks for reviewing Roland.

I reread this code and it turns out that the ipv6 handling does
the same exact thing as what I transformed ipv4 to do, so it's
better to have a helper function and make ipv4 and ipv6 use it.

I've sanitized the neighbour handling in the entire Infiniband
stack tonight, and I'll send that series out for review right now.
It will include an updated version of this patch.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch hide | download patch | download mbox

diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c
index a20c3c8..83d88e1 100644
--- a/drivers/infiniband/core/addr.c
+++ b/drivers/infiniband/core/addr.c
@@ -214,20 +214,19 @@  static int addr4_resolve(struct sockaddr_in *src_in,
 		goto put;
 	}
 
-	neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, rt->dst.dev);
-	if (!neigh || !(neigh->nud_state & NUD_VALID)) {
-		rcu_read_lock();
-		neigh_event_send(dst_get_neighbour(&rt->dst), NULL);
-		rcu_read_unlock();
-		ret = -ENODATA;
-		if (neigh)
-			goto release;
-		goto put;
+	rcu_read_lock();
+	neigh = dst_get_neighbour(&rt->dst);
+	ret = -ENODATA;
+	if (!neigh)
+		goto unlock_put;
+	if (!(neigh->nud_state & NUD_VALID)) {
+		neigh_event_send(neigh, NULL);
+		goto unlock_put;
 	}
 
 	ret = rdma_copy_addr(addr, neigh->dev, neigh->ha);
-release:
-	neigh_release(neigh);
+unlock_put:
+	rcu_read_unlock();
 put:
 	ip_rt_put(rt);
 out: