[asterisk-bugs] [Asterisk 0018204]: deadlock on 1.8.0-rc2 and crash on 1.8.0 with multipule sip channels

[Asterisk Bug Tracker]

A NOTE has been added to this issue. 
====================================================================== 
https://issues.asterisk.org/view.php?id=18204 
====================================================================== 
Reported By:                knkcn
Assigned To:                
====================================================================== 
Project:                    Asterisk
Issue ID:                   18204
Category:                   Core/Channels
Reproducibility:            sometimes
Severity:                   crash
Priority:                   normal
Status:                     new
Asterisk Version:           1.8.0 
JIRA:                        
Regression:                 No 
Reviewboard Link:            
SVN Branch (only for SVN checkouts, not tarball releases): N/A 
SVN Revision (number only!):  
Request Review:              
====================================================================== 
Date Submitted:             2010-10-25 02:30 CDT
Last Modified:              2010-10-28 04:55 CDT
====================================================================== 
Summary:                    deadlock on 1.8.0-rc2 and crash on 1.8.0 with
multipule sip channels
Description: 
Asterisk 1.8.0-rc2 as a call center server run well with 1 or 2 concurrent
lines. We did about 50 concurrent lines lab test with sipp and it run well
too. But under a more than 20 lines field trial, the Asterisk deadlocked.
After deadlocked I found many rtp channels and udp5060 had a large number
on Recv-Q. At the same time "core show channels" showed nothing and many
cli commands hang random. I collected some information using "gdb
/usr/sbin/asterisk <pid>" but I have not enough time to debug what locked
it before restarted it.

After that I upgraded the server to 1.8.0 with "DONT_OPTIMIZE" and
"DEBUG_THREADS", then I did the test again. But this time it crashed when
10 phones were calling in.

BTW, after the "DEBUG_THREADS" is enabled, every time the "bridge"
application is executed the following error will be reported.
[Oct 25 15:03:04] ERROR[25661] lock.c: features.c line 5551 (bridge_exec):
mutex 'current_dest_chan' freed more times than we've loc
ked!
[Oct 25 15:03:04] ERROR[25661] lock.c: features.c line 5551 (bridge_exec):
Error releasing mutex: Operation not permitted
====================================================================== 

---------------------------------------------------------------------- 
 (0128436) knkcn (reporter) - 2010-10-28 04:55
 https://issues.asterisk.org/view.php?id=18204#c128436 
---------------------------------------------------------------------- 
static int add_header_max_forwards(struct sip_pvt *dialog, struct
sip_request *req)
{
/*      char clen[10];
        const char *max = NULL;

        while (dialog->owner && ast_channel_trylock(dialog->owner)) {
                sip_pvt_unlock(dialog);
                usleep(1);
                sip_pvt_lock(dialog);
        }

        if (dialog->owner) {
                max = pbx_builtin_getvar_helper(dialog->owner,
"SIP_MAX_FORWARDS");
                ast_channel_unlock(dialog->owner);
        }

        if (max == NULL) {
                snprintf(clen, sizeof(clen), "%d", dialog->maxforwards);
        }
        return add_header(req, "Max-Forwards", max != NULL ? max : clen);
*/
        return add_header(req, "Max-Forwards", "70");
}

This works well under 4 times trial today. Should dialog be unlocked
before return? 

Issue History 
Date Modified    Username       Field                    Change               
====================================================================== 
2010-10-28 04:55 knkcn          Note Added: 0128436                          
======================================================================