On Thu, May 22, 2014 at 1:23 PM, Ben Pfaff <b...@nicira.com> wrote: > On Thu, May 22, 2014 at 11:50:59AM -0700, Pravin Shelar wrote: >> On Thu, May 22, 2014 at 10:30 AM, Daniele Di Proietto >> <ddiproie...@vmware.com> wrote: >> > These patches remove struct hmap flow_table from dp_netdev. >> > A new function has been added to the classifier interface >> > (classifier_at_position). Given the complexity added by this call, which >> > has to expose part of the classifier interface, I???m not sure anymore >> > this change is worth merging, but I???ll leave this decision to you guys. >> > >> >> Two hash tables for single flow table is not nice, we have to get rid >> of extra table. If you want you can hide classifier complexity by >> moving position variables in new struct, something like struct >> classifier_position. > > I don't think the hash table that Daniele's patches remove is used on > the fast path at all.
Right, It is not in datapath, But I am worried about flow setup performance and maintenance of two table in userspace datapath. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
