On Mon, 2014-04-14 at 14:02 -0400, David Miller wrote: > BTW, it occurs to me that there may be some other spots in the output path > that expect that if SKB is ETH_P_IP then skb->sk is IP socket. For example > somewhere in netfilter or packet classifier paths. > > Just FYI... that was one of the things I was going to audit.
packet classifiers cannot have such assumptions for sure. About iptunnel_xmit(), I do not think there is an issue, as we do not enter ip_queue_xmit() on this path, but ip_local_out(). ip_local_out() doesn't use skb->sk , unless some netfilter module uses this. I am not sure how the previous behavior could be useful in this case, as all packets were sharing same socket ownership. net/netfilter/xt_owner.c for example has better coverage if it can really have a pointer to the user socket, not the internal socket used by l2tp or vxlan tunnel. -- 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/
