umtxn and Apache 2.2

Borja Marcos BORJAMAR at SARENET.ES
Mon Aug 11 09:00:38 UTC 2008


Hello,

I'm running a server with FreeBSD 7-STABLE as of August 8, Apache 2.2  
with mpm/worker and threads support, and PHP 5.2.6.

Everything works like a charm, but I see that Apache is leaking  
processes that get stuck in umtxn state.

This graph shows it pretty well (I upgraded the system last Friday).

Attaching gdb to one of the stray processes and doing a backtrace of  
the active thread, I see this:

[Switching to Thread 0x8705900 (LWP 100647)]
0x382a8789 in _umtx_op () from /lib/libc.so.7
(gdb) bt
#0  0x382a8789 in _umtx_op () from /lib/libc.so.7
#1  0x3825fe0d in __error () from /lib/libthr.so.3
#2  0x084b2b80 in ?? ()
#3  0x00000005 in ?? ()
#4  0x00000000 in ?? ()
#5  0x00000000 in ?? ()
#6  0x00000000 in ?? ()
#7  0x38261914 in ?? () from /lib/libthr.so.3
#8  0xbe0e5ca8 in ?? ()
#9  0x3825b818 in pthread_mutex_getprioceiling () from /lib/libthr.so.3
Previous frame identical to this frame (corrupt stack?)
(gdb)

and it seems all the threads in the process are stuck here. Any ideas?



-------------- next part --------------



More information about the freebsd-stable mailing list