[ovs-discuss] LLDP support on openvswitch

[Ben Pfaff]

On Fri, May 22, 2015 at 03:45:54AM +0000, Dave DeSimone (ddesimon) wrote:
> Just a fundamental question, that I’m not understanding.
> 
> Here is my bridge
> 
>     Bridge "br0"
>         Port "br0"
>             Interface "br0"
>                 type: internal
>         Port "enp130s0f0"
>             Interface "enp130s0f0"
>         Port "mgmt0"
>             Interface "mgmt0"
>                 type: internal
> 
> 
> mgmt0 is carrying the ip address.
> 
> I’ve added a flow with this:
> 
> ovs-ofctl add-flow br0 dl_type=0x88cc,action=flood
> 
> 
> 
> When I dump the flows, I see pkts
> 
> ovs-ofctl dump-flows br0
> 	NXST_FLOW reply (xid=0x4):
> 	cookie=0x0, duration=23254.816s, table=0, n_packets=4647,
> n_bytes=1329042, idle_age=3, dl_type=0x88cc actions=FLOOD
> 
> 
> 
> 
> I also ran tcpdump on the mgmt0 interface and I see the lldp packets there.
> 
> 
> So, it looks like the pkts are coming in on the physical interface
> (enp130s0f0) and getting sent to the internal interface mgmt0. But the
> lldp daemon running on the linux os still is not seeing the pkts.
> 
> 
> Is this valid? Why wouldn’t the lldp process see these packets is they
> appear on the mgmt0 interface?

If the packets show up in tcpdump on mgmt0, then that means that OVS
sent them to the device.  Anything that happens beyond that point isn't
the responsibility of OVS; OVS has no control over it.  In fact, it
seems likely that the LLDP process receives packets from mgmt0 in the
same way that tcpdump does, so it is a puzzle to me why it wouldn't see
them.

I know that's not really very helpful but I don't have experience with
lldp daemons on Linux so I don't know what else to say.

If you figure out what's going on then I'd be happy to add something to
the FAQ to explain.