net: vrf: Do not allow looback to be moved to a VRF
authorDavid Ahern <dsa@cumulusnetworks.com>
Wed, 26 Apr 2017 14:58:22 +0000 (07:58 -0700)
committerDavid S. Miller <davem@davemloft.net>
Thu, 27 Apr 2017 20:49:43 +0000 (16:49 -0400)
Moving the loopback into a VRF breaks networking for the default VRF.
Since the VRF device is the loopback for VRF domains, there is no
reason to move the loopback. Given the repercussions, block attempts
to set lo into a VRF.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Reviewed-by: Greg Rose <gvrose8192@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/vrf.c

index aa5d304..ceda586 100644 (file)
@@ -877,6 +877,12 @@ static int do_vrf_add_slave(struct net_device *dev, struct net_device *port_dev)
 {
        int ret;
 
+       /* do not allow loopback device to be enslaved to a VRF.
+        * The vrf device acts as the loopback for the vrf.
+        */
+       if (port_dev == dev_net(dev)->loopback_dev)
+               return -EOPNOTSUPP;
+
        port_dev->priv_flags |= IFF_L3MDEV_SLAVE;
        ret = netdev_master_upper_dev_link(port_dev, dev, NULL, NULL);
        if (ret < 0)