On 11/7/16 3:13 AM, Lorenzo Colitti wrote:
> 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.
Using ingress information (iif in this case) to determine the route for a
response seems appropriate.
I can send a patch to revert back to:
if (!oif && netif_index_is_vrf(net, skb->skb_iif))
oif = skb->skb_iif;
thanks for the report.
