User login

Weekly Report - 10/7/2015




Supporting Q-in-Q is hard apparently, the 'work' in workaround isn't a thing.

Have discovered a lovely behaviour where in the Openflow1.3 spec, ethertype should be able to be seen for frames after vlans. The plural is important, but what I'm seeing is that only the first two ethertypes get sent through, so matching on 0x0800 works for one vlan, but a second one makes it break. Need to decide whether we submit a bug report based on this finding. This was found in my testing environment and in hardware (pronto).

Table hopping doesn't work due to ovs and vswitchd not recirculating frames, a process where headers get sent to vswitchd, vswitchd does some work, throws it back at ovs to get the rest (done with mpls). Could hack ovs to make it do this, but not really a lot of point since it wont run on hardware. May change my mind on this.