From: Eric Dumazet <eric.duma...@gmail.com> Date: Mon, 14 Apr 2014 12:06:36 -0700
> On Mon, 2014-04-14 at 14:48 -0400, David Miller wrote: >> From: Eric Dumazet <eric.duma...@gmail.com> >> Date: Mon, 14 Apr 2014 11:19:26 -0700 >> >> > ip_local_out() doesn't use skb->sk >> >> It does Eric. >> > > Hmmm, right... > >> We had just such a report with this in the backtrace, when AF_PACKET >> sends over vxlan devices. >> >> The problem is ip_mc_output(). > > So this means that : User socket wanted sk_mc_loop(sk), but because > vxlan changed skb->sk to internal socket, we were doing something else > anyway. Actually the exact opposite is happening. vxlan does not override skb->sk, and leaves it as AF_PACKET's socket. Then we crash in ip_mc_output() because it only expects IP sockets, not AF_PACKET ones, attached to skb->sk. > There are a lot of undocumented features here... like > > skb->priority = sk->sk_priority; > skb->mark = sk->sk_mark; > > which socket do we want here (in ip_queue_xmit()) ? Probably that of tunnel. > Its interesting to see ip6_xmit() already has a 'struct sock *sk' > parameter... > > This was the preliminary patch I tested : You still need to make similar transformation to ip_local_out() as mentioned above. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/