Re: WARN_ON_ONCE(!new_owner) within wake_futex_pi() triggerede

[Heiko Carstens <heiko.carstens@de.ibm.com>]

On Sat, Feb 02, 2019 at 11:14:27AM +0100, Thomas Gleixner wrote:
> On Sat, 2 Feb 2019, Heiko Carstens wrote:
> So after the unlock @timestamp 337.215675 the kernel does not deal with
> that futex at all until the failed lock attempt where it rightfully rejects
> the attempt due to the alleged owner being gone.
> 
> So this looks more like user space doing something stupid...
> 
> As we talked about the missing barriers before, I just looked at
> pthread_mutex_trylock() and that does still:
> 
> 	if (robust)
>           {
>             ENQUEUE_MUTEX_PI (mutex);
>             THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
>           }
> 
> So it's missing the barriers which pthread_mutex_lock() has. Grasping for
> straws obviously....

Excellent! Taking a look into the disassembly of nptl/pthread_mutex_trylock.o
reveals this part:

140:   a5 1b 00 01             oill    %r1,1
144:   e5 48 a0 f0 00 00       mvghi   240(%r10),0   <--- THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
14a:   e3 10 a0 e0 00 24       stg     %r1,224(%r10) <--- last THREAD_SETMEM of ENQUEUE_MUTEX_PI

I added a barrier between those two and now the code looks like this:

140:   a5 1b 00 01             oill    %r1,1
144:   e3 10 a0 e0 00 24       stg     %r1,224(%r10)
14a:   e5 48 a0 f0 00 00       mvghi   240(%r10),0

Looks like this was a one instruction race...

I'll try to reproduce with the patch below (sprinkling compiler
barriers just like the other files have).

diff --git a/nptl/pthread_mutex_trylock.c b/nptl/pthread_mutex_trylock.c
index 7de61f4f68..3b093cb09c 100644
--- a/nptl/pthread_mutex_trylock.c
+++ b/nptl/pthread_mutex_trylock.c
@@ -116,8 +116,12 @@ __pthread_mutex_trylock (pthread_mutex_t *mutex)
 	      mutex->__data.__count = 1;
 	      /* But it is inconsistent unless marked otherwise.  */
 	      mutex->__data.__owner = PTHREAD_MUTEX_INCONSISTENT;
-
+	      /* We must not enqueue the mutex before we have acquired it.
+		 Also see comments at ENQUEUE_MUTEX.  */
+	      __asm ("" ::: "memory");
 	      ENQUEUE_MUTEX (mutex);
+	      /* We need to clear op_pending after we enqueue the mutex.  */
+	      __asm ("" ::: "memory");
 	      THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
 
 	      /* Note that we deliberately exist here.  If we fall
@@ -177,7 +181,12 @@ __pthread_mutex_trylock (pthread_mutex_t *mutex)
 	}
       while ((oldval & FUTEX_OWNER_DIED) != 0);
 
+      /* We must not enqueue the mutex before we have acquired it.
+         Also see comments at ENQUEUE_MUTEX.  */
+      __asm ("" ::: "memory");
       ENQUEUE_MUTEX (mutex);
+      /* We need to clear op_pending after we enqueue the mutex.  */
+      __asm ("" ::: "memory");
       THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
 
       mutex->__data.__owner = id;
@@ -279,7 +288,11 @@ __pthread_mutex_trylock (pthread_mutex_t *mutex)
 	    /* But it is inconsistent unless marked otherwise.  */
 	    mutex->__data.__owner = PTHREAD_MUTEX_INCONSISTENT;
 
+	    /* We must not enqueue the mutex before we have acquired it.
+	       Also see comments at ENQUEUE_MUTEX.  */
+	    __asm ("" ::: "memory");
 	    ENQUEUE_MUTEX (mutex);
+	    __asm ("" ::: "memory");
 	    THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
 
 	    /* Note that we deliberately exit here.  If we fall
@@ -308,7 +321,12 @@ __pthread_mutex_trylock (pthread_mutex_t *mutex)
 
 	if (robust)
 	  {
+	    /* We must not enqueue the mutex before we have acquired it.
+	       Also see comments at ENQUEUE_MUTEX.  */
+	    __asm ("" ::: "memory");
 	    ENQUEUE_MUTEX_PI (mutex);
+	    /* We need to clear op_pending after we enqueue the mutex.  */
+	    __asm ("" ::: "memory");
 	    THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
 	  }