Hello all, I am observing a strange behavior with openvswitch 2.5.12:
1. br0 is bridged with an interface eth0 2. packet enters eth0 with a non-multicast/non-unicast dest mac address, possibly the first time ovs has seen either source or destination mac address. However I have reason to believe it has seen the destination mac before and knows it lies on the output side of eth0. 3. (my guess) kernel path does not know how to process the packet, so it sends it back to userspace slow path. 4. userspace sends it back down to eth0, this is confirmed. Is this correct behavior ? Technically it is correct since in step 2, it is doing the right thing and sending it to the right interface, but it is the same as the input interface. Wouldn't this mess up mac learning for switches downstream since they now observe the source mac getting reinjected back into eth0 ? Thanks, Ajit.
_______________________________________________ discuss mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
