Summary:ASTERISK-00002: Libpri crashes asterisk
Reporter:izo (izo)Labels:
Date Opened:2003-07-19 06:46:08Date Closed:2004-09-25 02:40:14
Versions:Frequency of
Description:Randomly at high load of calls asterisk crashes and
leaves core with backtrace attached in additional information.


#0  0x40155efa in mallopt () from /lib/libc.so.6
(gdb) bt
#0  0x40155efa in mallopt () from /lib/libc.so.6
#1  0x40155145 in malloc () from /lib/libc.so.6
#2  0x404cc716 in q931_getcall (pri=0x8125e30, cr=8224) at q931.c:1276
#3  0x404cd97e in q931_receive (pri=0x8125e30, h=0xbf3ff4d8, len=9) at q931.c:1728
#4  0x404c8931 in q921_handle_iframe (pri=0x8125e30, i=0xbf3ff4d4, len=13) at q921.c:396
ASTERISK-1  0x404c91b8 in q921_receive (pri=0x8125e30, h=0xbf3ff4d4, len=13) at q921.c:633
ASTERISK-2  0x404c712a in pri_check_event (pri=0x8125e30) at pri.c:105
ASTERISK-3  0x404af5de in pri_dchannel (vpri=0x404bc460) at chan_zap.c:5467
ASTERISK-4  0x4001ed53 in pthread_start_thread () from /lib/libpthread.so.0
Comments:By: Mark Spencer (markster) 2003-07-26 10:12:34

Unfortunately this isn't very useful.  Can you try running under valgrind:

valgrind --gdb-attach=yes ./asterisk -vvvg  # do not use -c option

Ignore anything about an ioctl with a NULL parameter, but see if you can find where the real bug is.  It should fail much more quickly under valgrind.

By: Mark Spencer (markster) 2003-08-09 14:04:56

We have made substantial changes to libpri to address this.  Can you confirm our changes make your crash disappear?

By: Mark Spencer (markster) 2003-08-17 12:50:39

Fixed in CVS libpri