Currently we don't synchronize barriers and revalidation at all. If we decide that this is important for correctness, we will have to write new code to implement it.
On Fri, Dec 5, 2014 at 11:21 AM, Madhu Challa <cha...@noironetworks.com> wrote: > Yes it does get a barrier reply. Pl see the logs below. We do see > revalidator21 wakeup, but may be we are sending the barrier reply before the > revalidation completes ? I am not sure how the barrier request synchronizes > with the revalidation. > > Thanks. > > 2014-12-05T19:08:01.710Z|00439|vconn|DBG|tcp:127.0.0.1:6653: received: > OFPT_FLOW_MOD (xid=0x20e86f61): ADD > priority=1250,tcp,in_port=1,vlan_tci=0x0000,dl_src=00:06:07:08:09:0a,nw_src=192.168.0.1,nw_dst=192.168.0.2,nw_tos=0,tp_src=1234,tp_dst=80 > out_port:0 actions=output:3 > 2014-12-05T19:08:01.710Z|00440|ofproto_dpif|DBG|need revalidate in > ofproto_wait_cb() > 2014-12-05T19:08:01.710Z|00441|poll_loop|DBG|wakeup due to [POLLIN] on fd 57 > (127.0.0.1:55146<->127.0.0.1:6653) at lib/stream-fd.c:155 > 2014-12-05T19:08:01.710Z|00442|vconn|DBG|tcp:127.0.0.1:6653: received: > OFPT_BARRIER_REQUEST (xid=0xec5c5cb4): > 2014-12-05T19:08:01.710Z|00451|poll_loop(revalidator21)|DBG|wakeup due to > [POLLIN] on fd 39 (FIFO pipe:[64669378]) at > ofproto/ofproto-dpif-upcall.c:790 (0% CPU usage) > 2014-12-05T19:08:01.710Z|00443|vconn|DBG|tcp:127.0.0.1:6653: sent (Success): > OFPT_BARRIER_REPLY (xid=0xec5c5cb4): > 2014-12-05T19:08:01.710Z|00452|dpif(revalidator21)|DBG|dummy@ovs-dummy: > flow_dump ufid:eb2b8b89e6a933cfa8e8fedabccad013 <empty>, packets:0, bytes:0, > used:never > 2014-12-05T19:08:01.710Z|00296|poll_loop(revalidator23)|DBG|wakeup due to > [POLLIN] on fd 37 (FIFO pipe:[64528939]) at lib/ovs-thread.c:302 (0% CPU > usage) > 2014-12-05T19:08:01.710Z|00298|poll_loop(revalidator25)|DBG|wakeup due to > [POLLIN] on fd 45 (FIFO pipe:[61793514]) at lib/ovs-thread.c:302 > 2014-12-05T19:08:01.710Z|00297|dpif(revalidator23)|DBG|dummy@ovs-dummy: > dumped all flows > 2014-12-05T19:08:01.710Z|00271|poll_loop(revalidator24)|DBG|wakeup due to > [POLLIN] on fd 41 (FIFO pipe:[64670832]) at lib/ovs-thread.c:302 > 2014-12-05T19:08:01.710Z|00453|dpif(revalidator21)|DBG|dummy@ovs-dummy: > dumped all flows > 2014-12-05T19:08:01.710Z|00299|dpif(revalidator25)|DBG|dummy@ovs-dummy: > dumped all flows > 2014-12-05T19:08:01.710Z|00272|dpif(revalidator24)|DBG|dummy@ovs-dummy: > dumped all flows > > On Fri, Dec 5, 2014 at 9:42 AM, Ben Pfaff <b...@nicira.com> wrote: >> >> On Thu, Dec 4, 2014 at 3:17 PM, Madhu Challa <cha...@noironetworks.com> >> wrote: >> > 6 of the tests fail because we failed to revalidate the flows fast >> > enough. >> > >> > e.g the pktact priority tests >> > add flow in_port=1, match M actions=output:2 with priority 250 >> > sends a packet and verifies its egress port is 2 >> > add flow in_port=1, match M actions=output:3 with priority 1250 >> > sends a packet and verifies its egress port is 3 (which fails) >> >> Hmm, if there's a barrier message in there then there might be a >> reasonable >> claim that Open vSwitch should complete revalidation. Is there a barrier? > > -- "I don't normally do acked-by's. I think it's my way of avoiding getting blamed when it all blows up." Andrew Morton _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
