Hello,
The PC has private address on eth0 interface and public on ppp0
default route looks like:
default dev ppp0  scope link

Tomorrow I'll try to play with ngrep.
BR,
Dmitri

08.02.2010 23:59, Daniel-Constantin Mierla пишет:
Hello,

On 2/8/10 10:46 PM, Dmitri wrote:
Hello,
No, it is not local, it is reacheable over internet, real dns domain.
why is it looping then? I see many Record Route headers with same IP, should be one since it is no transport layer gatewaying.

You can use ngrep to watch the network and see what happens with SUPBSCRIBE (watch as well looback interface).

Waiting for your instructions.

Going to do a commit to make it easily via makefile.

Cheers,
Daniel


BR,
Dmitri.

08.02.2010 23:27, Daniel-Constantin Mierla пишет:
Hello,

might be a buffer overflow somewhere else. I will send you some instructions to compile with memory debug one.

Meanwhile, this seems to be a particular case (has to be found and fixed anyhow), the SUBSCRIBE is looping. Is voice.djuk.ee a local domain (matching the 'myself' condition)? If is local domain and not in dns then add:

alias=voice.djuk.ee

Cheers,
Daniel


On 2/8/10 10:14 PM, Dmitri wrote:
Hello,

Updated gdb output in attached file.

BR,
Dmitri

08.02.2010 22:51, Daniel-Constantin Mierla пишет:
Hello,

once you are in gdb, run 'bt' to get the backtrace. That is very useful in the first step to see what functions were executed before the crash.


Thanks,
Daniel

On 2/8/10 9:23 PM, Dmitri wrote:
Hello Klaus,

Thank you for suggestion. Now I got core, and output of gdb you may see in attached file.
Core was about 40M of size.
I have to say that i never used kamailio before and this crash is my first experience.

BR,
Dmitri

08.02.2010 21:37, Klaus Darilion пишет:


Am 08.02.2010 20:00, schrieb Dmitri:
Hello,
1. Yes, there is at least 2 times more space than core dump needs.



# Try specifying things manually, as user root:
echo "/tmp/core.%e.sig%s.%p" > /proc/sys/kernel/core_pattern

# verify configuration changes
cat /proc/sys/kernel/core_pattern

# set ulimit
ulimit -c unlimited

# start kamailio manually
kamailio

now kamailio should be running as user root and should be able to core dump into /tmp

2. I could find nothing, unfortunately...
I can only see that problem is related to tm.so library:
kamailio kernel: [ 1051.432646] kamailio[2678]: segfault
at 12 ip 00007fa8b592ae43 sp 00007fff48ff9f40 error 4 in
tm.so[7fa8b5917000+72000]

Daniel, I can't remember. Is it necessary to turn off some compiler optimizations in Makefile.defs to get a useful backtrace? (or was that in Asterisk?)

regards
klaus



BR,
Dmitri

I don't know it this applies to you but also check the following:

1. You have enough space in the partition where $COREDIR is created
2. Check that the user that runs kamailio (usually openser) has read
write access on that directory

Also if you grep for core in the /var/log/ directory I believe that
there might be some lines either in dmesg or in messages that tell
you if a core was created or not.

Hope what I said is not redundant and it helps.

Marius



_______________________________________________
Kamailio (OpenSER) - Users mailing list
Users@lists.kamailio.org
http://lists.kamailio.org/cgi-bin/mailman/listinfo/users
http://lists.openser-project.org/cgi-bin/mailman/listinfo/users

_______________________________________________ Kamailio (OpenSER) - Users mailing list Users@lists.kamailio.org http://lists.kamailio.org/cgi-bin/mailman/listinfo/users http://lists.openser-project.org/cgi-bin/mailman/listinfo/users

-- 
Daniel-Constantin Mierla
eLearning class for Kamailio 3.0.0
Starting Feb 8, 2010
* http://www.asipto.com/
  


-- 
Daniel-Constantin Mierla
eLearning class for Kamailio 3.0.0
Starting Feb 8, 2010
* http://www.asipto.com/
  


-- 
Daniel-Constantin Mierla
eLearning class for Kamailio 3.0.0
Starting Feb 8, 2010
* http://www.asipto.com/