net: ipv6 bind to device issue
authorJiri Olsa <jolsa@redhat.com>
Tue, 20 Apr 2010 21:21:26 +0000 (21:21 +0000)
committerDavid S. Miller <davem@davemloft.net>
Thu, 22 Apr 2010 05:59:24 +0000 (22:59 -0700)
The issue raises when having 2 NICs both assigned the same
IPv6 global address.

If a sender binds to a particular NIC (SO_BINDTODEVICE),
the outgoing traffic is being sent via the first found.
The bonded device is thus not taken into an account during the
routing.

From the ip6_route_output function:

If the binding address is multicast, linklocal or loopback,
the RT6_LOOKUP_F_IFACE bit is set, but not for global address.

So binding global address will neglect SO_BINDTODEVICE-binded device,
because the fib6_rule_lookup function path won't check for the
flowi::oif field and take first route that fits.

Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Scott Otto <scott.otto@alcatel-lucent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv6/route.c

index c2438e8cb9d0eb6aec41f9c329fedbbdedcc5be8..05ebd7833043c687010382fb2d1534ef1bd45b8a 100644 (file)
@@ -815,7 +815,7 @@ struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
 {
        int flags = 0;
 
-       if (rt6_need_strict(&fl->fl6_dst))
+       if (fl->oif || rt6_need_strict(&fl->fl6_dst))
                flags |= RT6_LOOKUP_F_IFACE;
 
        if (!ipv6_addr_any(&fl->fl6_src))