Summary:	ASTERISK-06997: Asterisk segfault on bogus frame from Agent Channel
Reporter:	Mark Edwards (edwar64896)	Labels:
Date Opened:	2006-06-26 07:37:53	Date Closed:	2006-09-20 11:40:10
Priority:	Critical	Regression?	No
Status:	Closed/Complete	Components:	Channels/chan_agent
Versions:		Frequency of Occurrence
Related Issues:
Environment:		Attachments:	( 0) astlog.txt
Description:	Have experienced random crashes over the last few weeks, the most recent of which appears to be caused by Asterisk trying to free a frame which isn't valid. Operational conditions - approximately 40 concurrent calls, 4-5 concurrent chanspy sessions, lots of AMI activity. I don't think this bug is related to the linked list issue previously reported which plagued Chanspy a few months ago (and was fortunately resolved). Perhaps a similarity here to bug ID ASTERISK-5873? ****** ADDITIONAL INFORMATION ****** have uploaded debug log at time of crash and GDB output.
Comments:	By: Serge Vecher (serge-v) 2006-06-28 13:36:53 is the the backtrace from asterisk built with 'make dont-optimize'? Otherwise, it is useless. By: Serge Vecher (serge-v) 2006-06-28 13:37:24 Also, are you using AgentCallBackLogin? By: Mark Edwards (edwar64896) 2006-06-28 18:34:11 The makefile was out of the box -g3 -O6 I think is the current default. No, I'm not using AgentCallBackLogin. I mean there's some really interesting stuff in there... I'm sure it's not completely useless as in, useless. Surely? ;-) By: Serge Vecher (serge-v) 2006-06-29 08:37:31 sure, but the line number will not match up which makes it difficult to work with for developers. Please redo with make dont-optimize. By: Mark Edwards (edwar64896) 2006-06-29 08:44:14 I think that if you read the logs you'll find they make perfect sense. I don't think line numbers are going to be an issue here. Unfortunately, it isn't often possible - when trying to debug these sorts of issues - to make it happen on demand. I'll certainly recompile the distro and run it in production, but until someone comes up with a magic formula for reproducing this issue - and I certainly havn't - I would have to say this is as good as it's going to get for now. Certainly the GDB output makes perfect sense when comparing it with the crash circumstances and the asterisk logs - which I have provided. By: Serge Vecher (serge-v) 2006-09-01 14:14:25 edwar64896: did you give the latest 1.2 branch (r >41700) a shot? By: Serge Vecher (serge-v) 2006-09-20 11:03:58 given that there is no response from edwar64896 for over two weeks and there were major fixes post 1.2.9.1, I'm closing this issue for now. If you can reproduce this with 1.2.12.1 or later, please feel free to reopen the issue with a non-optimized bt attached. Thank you.