|Summary:||ASTERISK-07801: Crash in memory allocation when using a sufficiently old glibc (2.3.2)|
|Reporter:||Mathieu Rene (math)||Labels:|
|Date Opened:||2006-09-22 10:14:35||Date Closed:||2011-06-07 14:00:59|
|Environment:||Attachments:||( 0) btfull.txt|
|Description:||Segmentation fault generated randomly, I usually get a call in the morning telling me the system isn't working and I see the segfault so I ran it into gdb, bt full is attached.|
|Comments:||By: Serge Vecher (serge-v) 2006-09-22 10:59:47|
was asterisk built with 'make dont-optimize' ? If not, please rebuild and post a new backtrace. This should be safe on a production system.
By: Tilghman Lesher (tilghman) 2006-09-26 16:47:15
What's odd about the backtrace that you've provided is that when Asterisk crashes inside glibc in the memory allocation routines, it is almost always with a SIGABRT, not a SIGFAULT. SIGABRT generally indicates that we have memory corruption (detected by glibc), while SIGFAULT generally indicates something like a glibc programming error or (worse) a hardware fault.
What version of glibc is installed on your system?
Would it be possible to take down your system during an off hour and run a memory checker (such as memtest86) on it?
By: Mathieu Rene (math) 2006-09-26 16:50:39
I'll rebuild with make dont-optimize and run a memtest on the ram to see what it gives.
By: Tilghman Lesher (tilghman) 2006-09-27 10:04:08
It's worth noting that glibc 2.3.2 is over 3.5 years old, and a ton of memory fixes have been committed to that project since that time.
By: agjohnson (agjohnson) 2006-09-27 16:04:06
I believe we are experiencing the same problem. We are using glibc 2.4.8. After updating to 184.108.40.206, all three of our production servers started crashing apparently at random. I have included two backtraces from two separate machines.
By: Tilghman Lesher (tilghman) 2006-09-27 16:33:22
agjohnson: those two crashes are not only unrelated to this bug, they are unrelated to each other. Please file these as separate bugs, not attached to this one.
By: Olle Johansson (oej) 2006-10-29 14:52:48.000-0600
This might be related to "8232
By: Tilghman Lesher (tilghman) 2006-11-13 16:11:21.000-0600
No response from reporter. Reopen when you have the requested feedback.