On Sun, Sep 11, 2016 at 4:09 AM, David Ahern <d...@cumulusnetworks.com> wrote:
> A previous patch added l3mdev flow update making these hooks
> redundant. Remove them.
> [...]
> @@ -1582,8 +1582,7 @@ void ip_send_unicast_reply(struct sock *sk, struct
> sk_buff *skb,
> }
>
> oif = arg->bound_dev_if;
> - if (!oif && netif_index_is_l3_master(net, skb->skb_iif))
> - oif = skb->skb_iif;
> + oif = oif ? : skb->skb_iif;
>
> flowi4_init_output(&fl4, oif,
> IP4_REPLY_MARK(net, skb->mark),
This broke one of our unit tests. The expectation in the test was that
the RST replying to a SYN sent to a closed port should be generated
with oif=0. In other words it should not prefer the interface where
the SYN came in on, but instead should follow whatever the routing
table says it should do.
As far as I can tell this changed ~2 months ago, on 2016-09-10, when
e0d56fd was applied (step #2 below). Previously, oif was either a
passed-in sk_bound_dev_if or 0. Specifically:
1. "f7ba868 net: Use VRF index for oif in ip_send_unicast_reply" set
oif if the interface was an L3 master.
- flowi4_init_output(&fl4, arg->bound_dev_if,
+ oif = arg->bound_dev_if;
+ if (!oif && netif_index_is_vrf(net, skb->skb_iif))
+ oif = skb->skb_iif;
+
+ flowi4_init_output(&fl4, oif,
2. "e0d56fd net: l3mdev: remove redundant calls" removed the
netif_index_is_l3_master call and just set oif to skb_iif. This
changed the behaviour:
oif = arg->bound_dev_if;
- if (!oif && netif_index_is_l3_master(net, skb->skb_iif))
- oif = skb->skb_iif;
+ oif = oif ? : skb->skb_iif;
3. Later, netif_index_is_l3_master was removed by 19664c6a0009 ("net:
l3mdev: Remove netif_index_is_l3_master"). The removal was reverted by
"6104e11 net: ipv4: Do not drop to make_route if oif is l3mdev". But
this revert didn't change ip_send_unicast_reply back to what it was.
What should we do here? It would seem that now that
netif_index_is_l3_master has been resurrected, it's appropriate to use
it here as well. The user-visible behaviour changed only two months
ago. Unless we think that RSTs should always mirror the iif, in which
case I can change our tests accordingly.
Cheers,
Lorenzo
