net/ipv6: update flowi6_oif in ip6_dst_lookup_flow if not set
authorPhil Sutter <phil@nwl.cc>
Wed, 15 Jul 2015 19:56:26 +0000 (21:56 +0200)
committerDavid S. Miller <davem@davemloft.net>
Mon, 20 Jul 2015 19:59:32 +0000 (12:59 -0700)
Newly created flows don't have flowi6_oif set (at least if the
associated socket is not interface-bound). This leads to a mismatch in
__xfrm6_selector_match() for policies which specify an interface in the
selector (sel->ifindex != 0).

Backtracing shows this happens in code-paths originating from e.g.
ip6_datagram_connect(), rawv6_sendmsg() or tcp_v6_connect(). (UDP was
not tested for.)

In summary, this patch fixes policy matching on outgoing interface for
locally generated packets.

Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv6/ip6_output.c

index d5f7716662dbc361e93ebc443e72ee5fb7343b10..c5fc85286ef64fcf3d47a9cdbe29c487dc641812 100644 (file)
@@ -1023,6 +1023,8 @@ struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
                return ERR_PTR(err);
        if (final_dst)
                fl6->daddr = *final_dst;
+       if (!fl6->flowi6_oif)
+               fl6->flowi6_oif = dst->dev->ifindex;
 
        return xfrm_lookup_route(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
 }