On Mon, 2016-11-07 at 10:04 -0500, Stephen Suryaputra Lin wrote: > ICMP redirects behavior is different after the commit above. An email > requesting the explanation on why the behavior needs to be different > was sent earlier to netdev (https://patchwork.ozlabs.org/patch/687728/). > Since there isn't a reply yet, I decided to prepare this formal patch. > > In v2.6 kernel, it used to be that ip_rt_redirect() calls > arp_bind_neighbour() which returns 0 and then the state of the neigh for > the new_gw is checked. If the state isn't valid then the redirected > route is deleted. This behavior is maintained up to v3.5.7 by > check_peer_redirect() because rt->rt_gateway is assigned to > peer->redirect_learned.a4 before calling ipv4_neigh_lookup(). > > After the commit, ipv4_neigh_lookup() is performed without the > rt_gateway assigned to the new_gw. In the case when rt_gateway (old_gw) > isn't zero, the function uses it as the key. The neigh is most likely valid > since the old_gw is the one that sends the ICMP redirect message. Then the > new_gw is assigned to fib_nh_exception. The problem is: the new_gw ARP may > never gets resolved and the traffic is blackholed. > > Signed-off-by: Stephen Suryaputra Lin <ssu...@ieee.org> > --- > net/ipv4/route.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/net/ipv4/route.c b/net/ipv4/route.c > index 62d4d90c1389..510045cefcab 100644 > --- a/net/ipv4/route.c > +++ b/net/ipv4/route.c > @@ -753,7 +753,9 @@ static void __ip_do_redirect(struct rtable *rt, struct > sk_buff *skb, struct flow > goto reject_redirect; > } > > + rt->rt_gateway = 0; > n = ipv4_neigh_lookup(&rt->dst, NULL, &new_gw); > + rt->rt_gateway = old_gw; > if (!IS_ERR(n)) { > if (!(n->nud_state & NUD_VALID)) { > neigh_event_send(n, NULL);
In any case, rt is a shared object at that time, so even temporarily clearing/restoring rt_gateway seems wrong to me. I would rather call __ipv4_neigh_lookup(dst->dev, new_gw) directly at this point.
