[Xorp-users] multicast problem
Pavlin Radoslavov
pavlin@icir.org
Wed, 09 Mar 2005 15:55:49 -0800
> > Also, did the receiver get the multicast traffic?
>
> no. in both directions the same. i can send but the receiver is getting no
> traffic. i also tried another gentoo box with the latest cvs-version of xorp.
> and another multicast-application. but the same result.
This is odd. In your first email, you sent the following info:
cat /proc/net/ip_mr_cache
Group Origin Iif Pkts Bytes Wrong Oifs
0A0100E0 024D4237 1 94007 126345408 0 0:1
In other words, your kernel has the correct MFC entry for source
55.66.77.2 and group 224.0.1.10, the iif is eth1, and the oif is
eth0 (with min ttl of 1 needed to forward a packet). Furthermore,
the kernel has seen 94007 packets that match on that entry (and that
are eventually forwarded, unless the ttl is not large enough).
Three issues come to mind:
* The TTL of the data packets is not large enough. You mentioned in
your first email that it is large enough, but just double-check
it by running tcpdump on both eth0 and eth1. Running
tcpdump will also show you if the multicast packets are indeed
forwarded out on eth0.
* If the receiver is on the same host as the multicast router,
there could be some kernel issues with delivering the multicast
data packets to local receivers. I vaguely remember that there
were some issues with local multicast receivers running on the
same machine as the multicast router, but I don't have the notes
with me to check the details (OS, setup, etc).
To make sure this doesn't affect you, have your sender and
receiver running on hosts directly connected to your multicast
router.
* You may have some filters in your kernel (firewall rules, etc)
which are stopping the multicast traffic.
At least, make sure that the "rp_filter" in the Linux kernel
is reset to 0:
cat /proc/sys/net/ipv4/conf/all/rp_filter
echo 0 > /proc/sys/net/ipv4/conf/all/rp_filter
Regards,
Pavlin