On 4 May 2017 at 11:13, Jiri Benc <jb...@redhat.com> wrote: > On Thu, 4 May 2017 10:02:22 +0000, Zoltán Balogh wrote: >> I'm afraid the kernel would not accept that OVS_ACTION_ATTR_PUSH_ETH >> action in this case. >> >> Jiri, could you confirm, please? > > It would accept it and it would ignore the eth_type field. The kernel > does not verify whether netlink attributes are longer than they should > be. It just accepts such attributes and silently ignores the excessive > data.
Really? Previously the OVS usage of netlink always validated that netlink attributes are exactly as long as expected and that there's no trailing data. I recognise this is a little different from most kernel netlink usage, but that's important to how backward/forward compatibility works with OVS userspace. _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev