Re: [Linphone-users] ADSL modem crash when using linphone-2.0.0

[Alastair Johnson]

On Thursday 22 November 2007, Alastair Johnson wrote:
> On Thursday 22 November 2007, address@hidden wrote:
> > Quoting Alastair Johnson <address@hidden>:
> > > On Thursday 22 November 2007, address@hidden wrote:
> > >> Hello,
> > >>
> > >> Recently, I've decided to give a try to linphone as a SIP client.
> > >> Unfortunately, I can't
> > >> get it working, because my ADSL modem crashed every time I'm trying to
> > >> add proxy in the
> > >> preferences window. I'd mention that was quite unexpected behavior for
> > >> me... Any thoughts
> > >> about this issue?
> > >>
> > >> Regards,
> > >> Vladimir.
> > >
> > > It sounds like buggy ADSL modem firmware since it shouldn't crash even
> > > if it doesn't like what linphone is doing. You don't mention which
> > > version you're running, what the OS is, what the modem is and how it's
> > > connected, or what the proxy is. This makes it difficult to say much
> > > more about what might be happening. Linphone outbound proxy settings
> > > have always worked for me. I've used several versions , currently
> > > 1.7.1, on multiple linux clients. These proxy to siproxd 0.5.13 on a
> > > Gentoo box which currently uses a USB ADSL modem but used to be behind
> > > an ADSL router with ports forwarded to it. Its error logs have never
> > > suggested linphone is doing anything it shouldn't, though that's no
> > > guarantee.
> >
> > Firstly, the ADSL modem is working for more then one year and having
> > problem only when I'm running p2p clients without restricting a number
> > of sockets/connections, but as I concerned that's quite normal. The
> > modem is Orange Livebox. It's a king of Sagem Fast3202 modem. Me and
> > my girlfriend are both using SIP software almost everyday and didn't
> > have any problems with ADSL modem (yet :)).
> > Secondly, the version on linphone I mentioned in the subject of my
> > mail - it's 2.0.0, system is Gentoo (2.6.23.8 x86_64). Hopefully, this
> > information could help with understanding the problem.
>
> Sorry I missed the version number - the subject column in the mail client
> wasn't long enough to show it, and I didn't check it closely when replying.
> It might explain things though. 2.0.0 has only been out a few days, and one
> of the big changes was an update to the SIP stack. You may have been first
> to find a newly introduced bug.
>
> Do you have an ebuild for 2.0.0? After a portage update yesterday evening I
> have:
> [ebuild   R   ] net-voip/linphone-1.7.1-r1  USE="alsa arts gtk ipv6
> xv -console -ilbc -novideo" 7,361 kB
> This is ~amd64 with kernel 2.6.22-r5 so reasonably close to your system. It
> doesn't look like I've used linphone on this box recently so I'll check the
> config to see if it still works ;-) If that version's fine then I can try
> to reproduce your problem.

It may be a 64bit problem. The version above fails to connect to the sipgate 
test number on the amd64 machine but works with the same config file on an 
i586 machine. Ekiga on the amd64 connects correctly, so it isn't a general 
SIP problem on the 64 bit machine. I noticed that the i586 had the ilbc USE 
flag enabled, so I enabled that on the amd64 and recompiled. No difference to 
the behaviour though. I think an earlier version used to work on the 64 bit 
box, but I may be mistaken.

siproxd isn't complaining, but something's clearly not right. 
Running 'linphone --verbose' on both machines and comparing the output I see 
a problem handling the reply to the REGISTER request. 

| ERROR | <osip_transaction.c: 488> Remote UA is not compliant: missing a Via 
header!

So far as I can tell the reply is structured exactly the same way on both 
machines, including the Via: and the WWW-Authenticate: headers. The 64 bit 
machine gives the error above, then starts with a fresh register request. The 
32 bit machine accepts the reply, responds with the Authorization: header and 
gets a 200 OK for its trouble.

I don't have time to dig deeper at the moment, but you may have more luck with 
a 32 bit machine.