multicast packet loss

Hi all!

I am experiencing a multicast packet loss on batman-adv mesh, this can appear normal but it is not because the mesh is on top of wired Ethernet link!

the host fe80::c8e0:2cff:fe9b:6d1f is sending a multicast hello message every 4 seconds on bat0 interface, but as you can see from the dump timing on a machine attached on the same switch they are not arriving regurarly so what is happening? Is this normal?

here is the dump

eigenServer ~ # tcpdump -n -i bat0 udp port 6696 and host
fe80::c8e0:2cff:fe9b:6d1f [4/323]
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode listening on bat0, link-type EN10MB (Ethernet), capture size 96 bytes 14:45:19.295707 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 151
14:45:40.133468 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:45:53.357606 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:46:01.118909 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:46:01.227914 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 79
14:46:03.210770 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 79
14:46:04.161707 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:46:04.308875 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 79
14:46:09.058416 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:46:21.473697 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 53
14:47:12.359258 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:47:13.063186 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:47:14.094512 IP6 fe80::488e:c5ff:fee9:8901.6696 > fe80::c8e0:2cff:fe9b:6d1f.6696: UDP, length 24
14:47:14.115996 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > fe80::488e:c5ff:fee9:8901.6696: UDP, length 40
14:47:18.288823 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:47:24.098438 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:47:31.905127 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 151
14:47:36.944597 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:47:51.131644 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 60
14:48:11.621499 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 151
14:48:24.418458 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76
14:48:27.608263 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 151
14:48:30.003085 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 20
14:48:31.567988 IP6 fe80::c8e0:2cff:fe9b:6d1f.6696 > ff02::1:6.6696:
UDP, length 76

batman version is the same on both host
eigenServer ~ # batctl -v
batctl gentoo-2012.2.0 [batman-adv: 2012.2.0]

thanks for help!

Hello Gioacchino,

I'd say that losing packets is not normal :-)
Have you tried to use batctl td on the wired interface? You should see the "broadcast" packets going around. It would be interesting to understand if on batctl you had more packets than what you are receiving on the node.

Cheers,

p.s. what kind of hello messages are they? Their size is changing over time..

Antonio Quartulli

..each of us alone is worth nothing..
Ernesto "Che" Guevara

The problem was that babeld assumed bat0 as a wired interface so when 2 or 3 packet was lost it marked the route as unreachable, now the problem is fixed thanks to a patch from julius

by the way it is not normal that batman loss packet on wired links and I think we should do some debug to understand what is happening...

I am going to do some test as ordex said trying to make things clear