On Feb 8, 2013, at 1:06 PM, Peter Fales wrote: > On Fri, Feb 08, 2013 at 12:08:06PM -0800, Murphy McCauley wrote: >> I'd suggest that you try running POX's openflow.discovery component and see >> if it works, since it would seem to do pretty much what you're trying to do >> here (generate LLDP messages from scratch). If it works, check out the >> messages it's generating and compare them with your own (you can use >> openflow.debug to save it all as PCAP traces). > > That was a very helpful suggestion. I ran openflow.discovery and I *did* > see discovery messages received on the host connected to the switch > port. However I was puzzled because the destination MAC address was > 01:23:20:00:00:01, rather than the LLDP multicast address. Examing the > disccovery.py code, I see that it's using the "Nicira Discovery Protocol" > Multicast adress rather than the LLDP Multicast Address. > > I changed our code to use the NDP_MULTICAST address rather than > LLDP_MULTICAST and now I see the packets generated by my code! > > So, now the question is why doesn't it like the LLDP_MULTICAST address? > Is the openflow switch refusing to send these for some reason? Or > perhaps the connected host running wireshark is eating them at a layer > below where wireshark can see them? Or something else?
The normal LLDP broadcast address is in the bridge filtered range, so anything that thinks it's a bridge (Linux bridge module?) could be eating them. (NOX and POX specifically don't use the usual bridge filtered LLDP address because they wanted to be able to "see through" non-OpenFlow bridges when discovering the topology.) -- Murphy _______________________________________________ openflow-discuss mailing list openflow-discuss@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/openflow-discuss