[prev in list] [next in list] [prev in thread] [next in thread] 

List:       serdev
Subject:    Re: [Serdev] SER coredump in parse_via.c
From:       Andrei Pelinescu-Onciul <andrei () iptel ! org>
Date:       2009-04-30 16:28:04
Message-ID: 20090430162804.GI18412 () shell ! iptel ! org
[Download RAW message or body]

On Apr 26, 2009 at 18:17, Jiang Jinke <jiangjinke@gmail.com> wrote:
> My ser had several coredumps, it seems that the ser wanted to send 408
> timeout response to the caller after invite timeout. I can't reproduce
> the coredump yet.
> 
> Below are two backtrace results I attached. It's similar to the one
> described in http://www.archivum.info/serdev@iptel.org/2005-05/msg00014.html.
> Didn't see any further progress with that problem. I also did a search
> on the mailing list,
> didn't find any bug reports matching my problem.

It looks like a memory corruption problem (somebody writes in some
already free()'ed memory region, or somebody writes more then it has
allocated).

> 
> I'm running http://ftp.iptel.org/pub/ser/daily-snapshots/stable/ser-0.9.7-pre8_src_2007-01-26.tar.gz.

It might be related to 
http://git.sip-router.org/cgi-bin/gitweb.cgi?p=ser;a=commit;h=4d1c
but it could also be a bug in an unrelated module.

If you want to keep using 0.9.x, please try the latest 0.9.7 snapshot
(http://ftp.iptel.org/pub/ser/daily-snapshots/stable/ser-0.9.7+cvs20080424_src.tar.gz).
Otherwise upgrade to 2.0.1 (stable), 2.1.0 (stable but there are a few
modules that do not compile) or sip-router (cutting edge future 3.0).

If you still get coredumps after the upgrade, the only way to go
forward is to compile ser in debug mode. For 0.9.7 this would be: 
make proper; make mode=debug all and maybe make install mode=debug ...
The debug mode will add more checks for mallocs, so that a possible
memory corruption bug will be caught earlier and will also help with gdb
 (no more variables optimized to registers) at the cost of some speed
(but if you are not running more than a few thousands calls per second
it shouldn't be a problem).


Andrei
_______________________________________________
Serdev mailing list
Serdev@lists.iptel.org
http://lists.iptel.org/mailman/listinfo/serdev
[prev in list] [next in list] [prev in thread] [next in thread]