|Summary:||ASTERISK-13766: Crash in operation|
|Reporter:||Gaëtan Duchaussois (gaetan)||Labels:|
|Date Opened:||2009-03-17 15:42:19||Date Closed:||2011-06-07 14:00:32|
|Environment:||Attachments:||( 0) dump.txt|
|Description:||A system running with queue crashed with the dump attached. Seems to be on a login of an agent(1071). The f pointer is null when calling file_close.|
****** ADDITIONAL INFORMATION ******
Few before an agent stay frozen conected in agent show. agent logoff Agent/1143 did not manage to logof it. The only way o get him out was commenting the entry in /etc/asterisk/agents.conf, then make a module reload chan_agent.so, uncommenting the line in agents.conf the module reloadchan_agent.so. Don't know if there is a link
|Comments:||By: Gaëtan Duchaussois (gaetan) 2009-03-19 10:59:45|
It occurs this morning again without all the issue with the agent frozen and with 1.4.24
By: Matthew Roth (matthew roth) 2009-03-21 11:57:19
This looks like it's related to issue 0014395. Try upgrading file.c to revision 175407 and report back whether or not it resolves your problem.
By: Tilghman Lesher (tilghman) 2009-03-22 02:41:36
This appears to be a crash that has already been fixed. Please upgrade to 1.4 SVN and try to reproduce.
By: Gaëtan Duchaussois (gaetan) 2009-03-22 15:20:38
@Matthew Roth: Int he 1.4.24 version, the file.c is the same as version 175407:
root@asterisk-hp1:~# wget "http://svn.digium.com/view/asterisk/branches/1.4/main/file.c?view=co&revision=175407&content-type=text%2Fplain"
diff file.c /usr/src/asterisk-1.4.24/main/file.c
< ASTERISK_FILE_VERSION(__FILE__, "$Revision$")
> ASTERISK_FILE_VERSION(__FILE__, "$Revision: 175407 $")
@tilgnman are you speaking about bug 14395 it's seems not to be the same bug.
By: Matthew Roth (matthew roth) 2009-03-22 16:47:32
It was wishful thinking on my part, because my file.c was missing the code added in revision 175407 and I was getting backtraces that looked very similar to gaetan's. However, I upgraded to the current version last night and Asterisk experienced the same crash again this morning under low load (30 active channels, 15 active calls).
I'm monitoring this issue and will apply and test any patches.
By: Tilghman Lesher (tilghman) 2009-03-22 22:35:43
You're going to need the follow the instructions in doc/valgrind.txt, then.
By: Gaëtan Duchaussois (gaetan) 2009-03-23 06:25:25
the issue about valgrind, is that i'm unabl to reproduce this issue on a test amachine, and prod can't work with valgrind.I will try again to reproduce it
By: Matthew Roth (matthew roth) 2009-03-23 10:46:21
valgrind is a catch-22 for me, too. I'm going to try to get around it by sending a small percentage of the production calls to a server running Asterisk under valgrind.
Is it possible that this issue is related to using reference counting to ensure that an ast_filestream structure isn't freed until any embedded ast_frame structures have been freed as well? It's a shot in the dark, but ast_trans_pvt and ast_dsp structures also have embedded frames and they use a flag instead.
By: Leif Madsen (lmadsen) 2009-04-27 13:59:13
Pinging anyone with this issue; any chance of getting the valgrind output of this issue? If not, we may need to suspend it until the information required to resolve the issue is available. Thanks!
By: Leif Madsen (lmadsen) 2009-05-04 09:11:15
I'm suspending this issue for now due to a lack of information required to move this issue forward. If you have the valgrind information necessary to move this issue along, then please request a bug marshal on IRC reopen this issue (irc.freenode.net, channel #asterisk-bugs).