On Wed, Jun 24, 2015 at 1:40 PM, David Griswold <[email protected]> wrote: > I’m trying to get flow-mode tunneling working with native GRE tunneling. I > have a 2 port switch, access and trunk (tunnel) with the 2 ports connected > to test gear. The problem is that the trunk-to-access direction is not > working. It looks like packets are egressing from b1 on port NORMAL, where > they disappear. > > When I configure port-mode tunneling, everything works and I notice that b1 > does not egress to port NORMAL but ingresses directly to b0 from b1.
I think I see the problem, a check from port based tunneling was carried over to the more general native tunneling case. I just sent out a patch, can you please try it out? > Another observation is that when it is working for port mode tunneling, the > call to “tnl_port_should_receive” in function “xlate_lookup_ofproto_” > > sometimes return “true”. For flow mode tunneling, this function always > returns “false”. This check only tests to see if a flow has tunnel IP addresses, which should be true for all decapsulated packets. I guess you are probably looking at this on the physical bridge, where the packet hasn't been decapsulated yet? The "sometimes" for port based tunnels would depend on whether the packet has been decapsulated yet and the never for flow based tunnels is because the process fails before decapsulation. _______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
