> On Aug 29, 2016, at 5:06 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Aug 22, 2016 at 04:31:38PM -0700, Jarno Rajahalme wrote: >> Extend 'may_learn' attribute to also control the treatment of >> FIN_TIMEOUT action and asynchronous messages (packet ins, >> continuations), so that when 'may_learn' is 'false' and >> 'resubmit_stats' is 'NULL', no OpenFlow-visible side effects are >> generated by the translation. >> >> Correspondingly, add support for one-time asynchronous messages to >> xlate cache, so that all side-effects of the translation may be >> executed at a later stage. This will be useful for bundle commits. >> >> Signed-off-by: Jarno Rajahalme <ja...@ovn.org> > > Would it make any sense to drop support for immediate effects, so that > callers that want immediate effects provide an xcache entry and then > execute it and discard it? >
Joe had earlier discovered that creating the xlate cache at upcall time is too expensive and negatively affects the megaflow set-up rate. This is due to who main factors: many memory allocations and taking references to all the referred objects. So currently the translation is done again at first revalidation to create the xlate cache, which seems more work overall, but does not affect the megaflow set-up rate. So, at least for now we need the capability for immediate effects. We may revisit later, though. Jarno > Acked-by: Ben Pfaff <b...@ovn.org> _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
