fix bonding: allow arp_ip_targets on separate vlans to use arp validation

On Wed, Jan 06, 2010 at 10:10:03PM +0100, Eric Dumazet wrote:
> Le 06/01/2010 19:38, Eric Dumazet a écrit :
> >
> > (net-next-2.6 doesnt work well on my bond/vlan setup, I suspect I need a bisection)
>
> David, I had to revert 1f3c8804
> (bonding: allow arp_ip_targets on separate vlans to use arp validation)
>
> Or else, my vlan devices dont work (unfortunatly I dont have much time
> these days to debug the thing)
>
> My config :
>
>               +---------+
> vlan.103 -----+ bond0   +--- eth1 (bnx2)
>               |         +
> vlan.825 -----+         +--- eth2 (tg3)
>               +---------+
>
> $ cat /proc/net/bonding/bond0
> Ethernet Channel Bonding Driver: v3.6.0 (September 26, 2009)
>
> Bonding Mode: fault-tolerance (active-backup)
> Primary Slave: None
> Currently Active Slave: eth2
> MII Status: up
> MII Polling Interval (ms): 100
> Up Delay (ms): 0
> Down Delay (ms): 0
>
> Slave Interface: eth1  (bnx2)
> MII Status: down
> Link Failure Count: 1
> Permanent HW addr: 00:1e:0b:ec:d3:d2
>
> Slave Interface: eth2   (tg3)
> MII Status: up
> Link Failure Count: 0
> Permanent HW addr: 00:1e:0b:92:78:50
>

This patch fixes up a problem with found with commit
1f3c8804.  The original change
overloaded null_or_orig, but doing that prevented any packet handlers
that were not tied to a specific device (i.e. ptype->dev == NULL) from
ever receiving any frames.

The null_or_orig variable cannot be overloaded, and must be kept as NULL
to prevent the frame from being ignored by packet handlers designed to
accept frames on any interface.
Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/core/dev.c

@@ -2430,6 +2430,7 @@ int netif_receive_skb(struct sk_buff *skb)
 	struct packet_type *ptype, *pt_prev;
 	struct net_device *orig_dev;
 	struct net_device *null_or_orig;
+	struct net_device *null_or_bond;
 	int ret = NET_RX_DROP;
 	__be16 type;
 
@@ -2500,21 +2501,19 @@ int netif_receive_skb(struct sk_buff *skb)
 	 * bonding interfaces still make their way to any base bonding
 	 * device that may have registered for a specific ptype.  The
 	 * handler may have to adjust skb->dev and orig_dev.
-	 *
-	 * null_or_orig can be overloaded since it will not be set when
-	 * using VLANs on top of bonding.  Putting it here prevents
-	 * disturbing the ptype_all handlers above.
 	 */
+	null_or_bond = NULL;
 	if ((skb->dev->priv_flags & IFF_802_1Q_VLAN) &&
 	    (vlan_dev_real_dev(skb->dev)->priv_flags & IFF_BONDING)) {
-		null_or_orig = vlan_dev_real_dev(skb->dev);
+		null_or_bond = vlan_dev_real_dev(skb->dev);
 	}
 
 	type = skb->protocol;
 	list_for_each_entry_rcu(ptype,
 			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
 		if (ptype->type == type && (ptype->dev == null_or_orig ||
-		     ptype->dev == skb->dev || ptype->dev == orig_dev)) {
+		     ptype->dev == skb->dev || ptype->dev == orig_dev ||
+		     ptype->dev == null_or_bond)) {
 			if (pt_prev)
 				ret = deliver_skb(skb, pt_prev, orig_dev);
 			pt_prev = ptype;