[Serusers] multicast loopback

[Jan Janak]

On 15-02 18:45, Samuel Osorio Calvo wrote:
> Hi gurus!
> 
> I have been playing with multicast in SER fow a while and it worked
> almost perfectly: SER is listening and sending wihout much difference as
> if it was a unicast address. Again thanks to SER developers!!
> 
> There is a small detail which I hope you can help me to solve:
> multicast loopback. The problem is that SER receives all messages it
> sends to the multicast address and processes it as if it was received
> from the multicast address.
> 
> Variables affecting the mcast loopback:
> SER config file mcast_loopback is set to no, and it is used in the
> opening of the mcast socket. So I guess the socket is effectively opened
> with the loopback disabled.

   Yes, should be.

> The kernel MCAST_LOOPBACK is enabled, but I think that it does not
> affect the sockets opened with an explicit 0 in the MCAST_LOOPBACK
> setsockopt field.

  What happens if you disable it ?

> So I think it is not a matter of the multicast loopback parameters but
> a matter of the network topology. Please correct me if I am wrong!!!!

  Could you send me some SIP message dumps ?

> Our host has a couple of listen= addresses, one for the unicast and
> another for the multicast. 
> Guessings:
> Is it possible that SER uses the unicast interface to send the
> multicast message and therefore the multicast loopback check is not
> checked in the unicast sending socket???? 

  I don't know actually, but it is easy to find out. Change the order of
  listen directives in the configuration file. Try:
  listen=unicast_ip
  listen=multicast_ip

  and then 

  listen=multicast_ip
  listen=unicast_ip

  And see if there is any difference.

    Jan.

> But the same message, going deep into the TCP/UDP linux stack faces
> that there is an internal interface with the mcast address and it is
> therefore passed back?????
> 
> I hope that I have been clear and that somebody can clarify me why is
> this happening (a solution is even better :) )
> 
> Thanks in advance,
> Samuel.
> 
> 
> Unclassified.
> 
> _______________________________________________
> Serusers mailing list
> serusers at lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serusers