Thanks Cascardo, I am on 2.4.0.
I delved into the source code, and it seems that it is because the querier sets its IPv4 source address to 0.0.0.0. I understand that this is in accordance with the RFC 4541 (Section 2.1.1.1 - 1(b)), so I guess this is indeed the correct behavior. Would you know why the RFC setup this rule ? I would assume this would create issues in setups like ours, where: - there is no Multicast router in the network (all traffic originates and terminates within the LAN). - there are multiple OVS software switches hooked up with each other. - an IGMP querier is setup to allow IGMP snooping switches to populate their tables correctly. Thanks, Ajit. On Sun, 24 Jul 2016 at 10:11 Thadeu Lima de Souza Cascardo < casca...@redhat.com> wrote: > On Sun, Jul 24, 2016 at 04:23:59PM +0000, Ajit Warrier wrote: > > I am testing the IGMP snooping behavior in OVS 2.4 with the following > setup: > > > > OVS1-------------eth0--------------------OVS2 > > | | > > | | > > eth1 eth1 ------- IGMP Querier > > | | > > | | > > Laptop1 Laptop2 > > > > OVS1 and OVS2 are embedded devices running OVS-2.4 with snooping enabled. > > Laptop1 and Laptop2 have multicast listeners. > > > > Without the IGMP Querier, the IGMP join messages sent by Laptop1 stop at > > OVS1 and do not continue on to OVS1. This is what I expected. > > > > Now I turn on an IGMP querier which sends general IGMP queries on > > 224.0.0.1. I can see these queries going through OVS2 and arriving at > OVS1. > > Now further IGMP joins from Laptop1 should no longer be blocked by OVS1, > > but should be forwarded on the port (eth0) that received an IGMP query. I > > do not observe this behavior. > > > > Is this behavior not yet implemented in the current snooping > implementation > > ? Is there any configuration option to enable it ? > > > > Regards, > > Ajit. > > Hi, Ajit. > > That is implemented. Some bugs have been fixed already. Can you try latest > version at 2.4 branch? What exact version did you test? > > If the latest doesn't work, it could be a new bug, and I would appreciate > your > help in debugging it. But before we do it, we should verify for the latest > fixes. > > Regards. > Cascardo. >
_______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss