Etch pthread_mutex_lock - bad backtrace

[Tudor Golubenco <tudor@golubenco.org>]

Hi,

I'm having trouble getting gdb backtraces from a multithreaded (nptl)
C program in which some threads are blocked in pthread_mutex_lock.

Consider the following program that does a simple deadlock:

#include <pthread.h>
#include <stdio.h>
#include <unistd.h>

pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
pthread_t thread1, thread2;

void* thread_function(void* arg)
{
        pthread_mutex_lock(&mutex);

        pause();
        return NULL;
}

int main() {

        pthread_create(&thread1, NULL, thread_function, NULL);
        pthread_create(&thread2, NULL, thread_function, NULL);

        pause();
        return 0;
}

If I run this on an Etch amd64 machine (libc 2.3.6.ds1-13etch7), and
attach with gdb to get a backtrace, for the locked thread I get:

Thread 3 (Thread 1090525536 (LWP 14962)):
#0  0x00002afcad512eeb in __lll_mutex_lock_wait ()
from /lib/libpthread.so.0 #1  0x0000000000000002 in ?? ()
#2  0x00000000410019f0 in ?? ()
#3  0x00002afcad50fd9f in pthread_mutex_lock ()
from /lib/libpthread.so.0 #4  0x0000000000000000 in ?? ()

which is not very useful.

I have tried using libc6-dbg, but the output is similar:

# LD_LIBRARY_PATH=/usr/lib/debug/ gdb ./a.out
I Thread 3 (Thread 1090525536 (LWP 14979)):
#0  __lll_mutex_lock_wait ()
at ../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:59 #1
0x0000000000000002 in ?? () #2  0x00000000410019f0 in ?? ()
#3  0x00002b8db4265d9f in pthread_mutex_lock ()
from /usr/lib/debug/libpthread.so.0 #4  0x0000000000000000 in ?? ()
Current language:  auto; currently asm

However, on a Lenny machine (libc 2.7-15), the output is as expected:

Thread 3 (Thread 0xb763fb90 (LWP 5268)):
#0  0xb7fd6424 in __kernel_vsyscall ()
#1  0xb7fa9c99 in __lll_lock_wait () from /lib/i686/cmov/libpthread.so.0
#2  0xb7fa50c4 in _L_lock_89 () from /lib/i686/cmov/libpthread.so.0
#3  0xb7fa49f2 in pthread_mutex_lock ()
from /lib/i686/cmov/libpthread.so.0
#4  0x080484a6 in thread_function (arg=0x0) at test.c:11
#5  0xb7fa34c0 in start_thread () from /lib/i686/cmov/libpthread.so.0
#6  0xb7f2261e in clone () from /lib/i686/cmov/libc.so.6

Also, I should mention that if I use a spin lock instead of the mutex,
I get a backtrace also in etch.

First, I would like to understand why this could happen. I have the
libc6-dev package installed and generally I can see the symbols just
fine, so this doesn't look like a 'missing symbols' problem.

Could it be that the assembler code from the pthread_mutex_lock
implementation is not correctly understood by gdb? Is there anything to
do in this case?

Any suggestions welcome, including redirecting me to the correct list
to ask this stuff :-) Also, I would appreciate if you could run the
above program on your Etch system and let me know if you can get a
meaningful bt.

TIA,
Tudor

-- 
Tudor Golubenco			http://golubenco.org

Beware of bugs in the above code; I have only proved it correct, 
not tried it.  -- Donald Knuth