Jarod Wilson <ja...@redhat.com> wrote: [...] >This sounds suspiciously like the same problem Uwe was encountering[*] and >attempting to solve. Uwe, can you give this patch a try? > >[*] = http://marc.info/?l=linux-netdev&m=144416122705850&w=2
Agreed; my apologies for not Cc'ing you, Uwe. FWIW, the "Sometimes, the issue sorts itself out after a while" Uwe reported is likely due to something causing a switch flood (TCN or clearing of the MAC table); in that case, the ARP reply is sent to all ports, so the "expected" curr_arp_slave receives the ARP reply. -J --- -Jay Vosburgh, jay.vosbu...@canonical.com