Summary:	ASTERISK-05732: meetme with i option - high cpu usage scenario
Reporter:	vortex_0_o (vortex_0_o)	Labels:
Date Opened:	2005-11-28 22:20:21.000-0600	Date Closed:	2008-01-15 17:52:14.000-0600
Priority:	Minor	Regression?	No
Status:	Closed/Complete	Components:	Core/General
Versions:		Frequency of Occurrence
Related Issues:
Environment:		Attachments:	( 0) meetme-hangup-freeze.patch
Description:	meetme with i option record name poundkey press 1 to accept message hangup call immediately before connecting to conference on console: warning filec.583 ast_readaudio_callback failed to write frame asterisk at 99% cpu usage for about 20 seconds - causing quality issues on other channels.
Comments:	By: Clod Patry (junky) 2005-11-28 22:55:54.000-0600 the first 3 things prior to "hangup call immediately before connecting to conference" are not meetme related. Could you post your dialplan for all these stuff. Thanks. By: vortex_0_o (vortex_0_o) 2005-11-29 06:07:03.000-0600 I think they are related with the 'i' option - announce user join/leave anyway dialplan is: exten => 707,1,Answer exten => 707,2,Wait(1) exten => 707,3,MeetMe(707|i) exten => 707,4,Hangup By: Clod Patry (junky) 2005-11-29 22:07:46.000-0600 i cant reproduce this. which Revision of app_meetme.c do you have? which zap module do u have? By: vortex_0_o (vortex_0_o) 2005-11-30 06:50:04.000-0600 it only works (for me) if you press 1 then hangup before you actually get to the conference wct4xxp (with a 411P) updated against v1-2 before the switch to svn so I think app_meetme.c is at 1.126.2.1 By: Clod Patry (junky) 2005-11-30 07:53:58.000-0600 in ur CLI: show version files app_meetme.c By: vortex_0_o (vortex_0_o) 2005-11-30 08:46:39.000-0600 Revision 1.126.2.1 By: vortex_0_o (vortex_0_o) 2005-12-01 14:12:33.000-0600 updated againsr svn app_meetme.c is 7221 stuck at 99% cpu forever - neverending loop somewhere? only way out is to kill the asterisk process By: vortex_0_o (vortex_0_o) 2005-12-01 15:44:14.000-0600 On further checking: hanging up at anytime during a meetme prompt causes a spike even without the 'i' option hanging up at the standard message "you are currently the only person in this confernce" causes the same result when running asterisk with the -p option this spike seems to last forever By: vortex_0_o (vortex_0_o) 2005-12-15 17:36:29.000-0600 Is there any other information I can provide? I would really like to use this meetme with anouncements rather than the q option. By: Matt O'Gorman (mogorman) 2006-01-13 12:29:27.000-0600 I can reproduce this bug in svn trunk. can anyone help me profile it? By: Tilghman Lesher (tilghman) 2006-03-01 16:14:07.000-0600 Is this still an issue with the latest trunk?  I see several possible fixes have gone into trunk since this bug went dormant.  // Housekeeping By: Matt O'Gorman (mogorman) 2006-03-02 11:27:41.000-0600 still happens corydon By: vortex_0_o (vortex_0_o) 2006-03-02 13:04:39.000-0600 yes still an issue By: Michael Neuhauser (mneuhauser) 2006-03-23 14:22:06.000-0600 Some debugging revealed that a tight loop in app_meetme.c:conf_flush() consumes the whole CPU power (HANGUP/NULL frames are not handled correctly while discarding frames of the channel). I've attached a very small patch (just 2 lines added) that fixes this. By: BJ Weschke (bweschke) 2006-03-23 14:40:41.000-0600 mneuhauser - this looks like a valid fix. Before I commit, do you have a disclaimer on file? By: BJ Weschke (bweschke) 2006-03-23 14:51:29.000-0600 Committed to /trunk and 1.2, thanks! By: Digium Subversion (svnbot) 2008-01-15 17:51:32.000-0600 Repository: asterisk Revision: 14467 U   branches/1.2/apps/app_meetme.c ------------------------------------------------------------------------ r14467 | bweschke | 2008-01-15 17:51:32 -0600 (Tue, 15 Jan 2008) | 3 lines Bug ASTERISK-5732 - fix a possible race state in app_meetme when a channel has gone away and we are reading continuously for more frames. (mneuhauser) ------------------------------------------------------------------------ http://svn.digium.com/view/asterisk?view=rev&revision=14467 By: Digium Subversion (svnbot) 2008-01-15 17:51:35.000-0600 Repository: asterisk Revision: 14470 _U  trunk/ U   trunk/apps/app_meetme.c ------------------------------------------------------------------------ r14470 | bweschke | 2008-01-15 17:51:34 -0600 (Tue, 15 Jan 2008) | 11 lines Merged revisions 14467 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.2 ........ r14467 | bweschke | 2006-03-23 14:43:05 -0600 (Thu, 23 Mar 2006) | 3 lines Bug ASTERISK-5732 - fix a possible race state in app_meetme when a channel has gone away and we are reading continuously for more frames. (mneuhauser) ........ ------------------------------------------------------------------------ http://svn.digium.com/view/asterisk?view=rev&revision=14470 By: Digium Subversion (svnbot) 2008-01-15 17:52:14.000-0600 Repository: asterisk Revision: 14517 _U  team/bweschke/gain_functions/ U   team/bweschke/gain_functions/apps/app_meetme.c U   team/bweschke/gain_functions/apps/app_readfile.c U   team/bweschke/gain_functions/pbx.c U   team/bweschke/gain_functions/utils.c ------------------------------------------------------------------------ r14517 | bweschke | 2008-01-15 17:52:13 -0600 (Tue, 15 Jan 2008) | 42 lines Merged revisions 14463,14470,14479,14508 via svnmerge from https://origsvn.digium.com/svn/asterisk/trunk ................ r14463 | russell | 2006-03-23 14:15:01 -0600 (Thu, 23 Mar 2006) | 10 lines Merged revisions 14462 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.2 ........ r14462 | russell | 2006-03-23 15:13:48 -0500 (Thu, 23 Mar 2006) | 2 lines don't crash when asked to read from a file that doesn't exist (issue ASTERISK-6604) ........ ................ r14470 | bweschke | 2006-03-23 14:48:08 -0600 (Thu, 23 Mar 2006) | 11 lines Merged revisions 14467 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.2 ........ r14467 | bweschke | 2006-03-23 14:43:05 -0600 (Thu, 23 Mar 2006) | 3 lines Bug ASTERISK-5732 - fix a possible race state in app_meetme when a channel has gone away and we are reading continuously for more frames. (mneuhauser) ........ ................ r14479 | kpfleming | 2006-03-23 15:01:19 -0600 (Thu, 23 Mar 2006) | 2 lines don't wrap this in ifdef... using va_start is safe on all platforms :-) ................ r14508 | kpfleming | 2006-03-23 15:06:26 -0600 (Thu, 23 Mar 2006) | 2 lines correct typo ................ ------------------------------------------------------------------------ http://svn.digium.com/view/asterisk?view=rev&revision=14517