freeradius 1.1.8 hangs in rlm_perl on 64bit

[Olivier Bilodeau]

Hi,

We have been using freeradius 1.1.8 with a lot of success on a lot of 
our deployments.

Lately we deployed a freeradius 1.1.8 on a 64 bit environment for the 
first time. On that setup we have been experiencing segfaults and 
hanging processes once a month maybe.

When hanged, gdb points to line 1202 of rlm_perl.c:

(gdb) bt
#0  0x00002ac034f962e4 in __lll_lock_wait () from /lib64/libpthread.so.0
#1  0x00002ac034f91c3a in _L_lock_1034 () from /lib64/libpthread.so.0
#2  0x00002ac034f91afc in pthread_mutex_lock () from /lib64/libpthread.so.0
#3  0x00002ac038fa1063 in perl_detach (instance=<value optimized out>) 
at rlm_perl.c:1202
#4  0x00002ac03472aa96 in detach_modules () from /usr/sbin/radiusd
#5  0x00002ac03472e5e3 in main () from /usr/sbin/radiusd

Here's the line:
/*
  * Wait until clone becomes idle
  */
MUTEX_LOCK(&handle->lock);


Any ideas on what it could be? perl? freeradius? 64bit? our perl code?

We haven't switched to 2.x because of day to day activity overload but 
that could be a good reason for us to do the learning necessary to switch.

Thanks in advance,
Cheers!
-- 
Olivier Bilodeau
obilodeau at inverse.ca  ::  +1.514.447.4918 *115  ::  www.inverse.ca
Inverse inc. :: Leaders behind SOGo (www.scalableogo.org) and 
PacketFence (www.packetfence.org)