On Thu, Jan 16, 2020 at 09:47:23PM +0530, Vishal Deep Ajmera via dev wrote: > The wildcard bits in the installed megaflow entry could be different from > the bits originally generated by the ofproto layer. Datapath implementation > wildcards those match fields which are not present in the incoming packet > before installing the flow. > > When the revalidator thread validates a megaflow, it will first query the > ofproto layer to get the wildcard bits and then compare it against the > wildcard bits in the megaflow. If the bits are different the entry will be > removed. A subsequent packet will again result in the same megaflow entry > being installed only for it to be removed by the revalidator thread. This > cycle will continue and will significantly degrade performance. > > An earlier patch fixing the issue for MPLS and VLAN was sent. > However similar problem now appears for IPv6 datapath flows. > > This patch addresses the issue in a generic way. > > Signed-off-by: Vishal Deep Ajmera <[email protected]>
This is clever! I think that this kind of approach is the right one. I don't yet understand why this should be moved from xlate_wc_finish() to revalidate_ukey__(). Can you help me understand that? _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
