lkml.org 
[lkml]   [2005]   [Jan]   [1]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
DateSat, 01 Jan 2005 17:48:09 +0100
FromGuilhem Lavaux <>
SubjectPThreads, signals, futex and SMP
Hi,

I am one of the developer of kaffe, a GPL implementation of the Java 
Virtual Machine, and we encountered some problems on Linux/2.6/SMP. I am 
currently running 2.6.8.1 Mandrake Kernel. I can try whether the problem 
is reproduceable on a vanilla kernel but I don't think this part has 
been touched.

Here is the problem: we have a regression test which is quite intense in 
thread creation/destruction, each thread can start the garbage 
collector. The garbage collector needs to stop all running thread to be 
able to walk the heap/stack. For this, it uses a particular signal which 
is sent to all threads. The signal handler calls sigwait to stop the 
thread. 50% of the time everything is fine but from time to time, kaffe 
has a deadlock. It appears that it always happen when we are in the 
following configuration:

Thread 1
------------
sigwait
<signal handler>
futex syscall
pthread_mutex_unlock

Thread 2
------------
futex syscall
pthread_mutex_lock
Garbage Collector thread
Now if we look at the futex code in the linux kernel, we see this:

static int futex_wait(unsigned long uaddr, int val, unsigned long time)
{
       DECLARE_WAITQUEUE(wait, current);
       int ret, curval;
       struct futex_q q;
       down_read(&current->mm->mmap_sem);

the kernel then prepares the wait queue and unlock mmap_sem. 


Concerning the mutex_unlock  part we have this:

static int futex_wake(unsigned long uaddr, int nr_wake)
{
       union futex_key key;
       struct futex_hash_bucket *bh;
       struct list_head *head;
       struct futex_q *this, *next;
       int ret;
       down_read(&current->mm->mmap_sem);

and the kernel iterates the semaphores and wakes up all threads.


What may happen if the signal handler is called after down_read in 
futex_wake ? My guess is that we are not able to call futex_wait because 
the application will deadlock because the first thread is frozen by a 
sigwait.

So either we have a limitation of the kernel either a bug if the 
analysis is correct.
The only point is that I am not sure whether a signal is allowed to 
interrupt a syscall just in the middle of futex_wake. If this is not 
possible there may be a bug in our application somewhere else.

Regards,

Guilhem Lavaux.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:09    [from the cache]
©2003-2008