Skip to contentSkip to conversation

futex_unlock_pi returns w/o unlocking

BUG john stultz 1 participant 1 message Aug 3, 2006

lore
Share & tools
lore
#0 lore raw

Ingo,
So we've just hunted down a nasty bug w/ 2.6.17-rt8. We've had some odd
test failures where userspace apps were hanging when dealing w/ pi mutexes.

After a good bit of debugging (not by me) it was found that we were deadlocking by double locking a mutex, however it wasn't userland's fault. On top of that, the mutex __owner field was null, so something was wrong.

We found that futex_unlock_pi() was somehow returning without error while not actually clearing the mutex __lock value. Further digging found the failure case, and its a bit obscure.

1) In futex_unlock_pi() we start w/ ret=0 and we go down to the first futex_atomic_cmpxchg_inatomic(), where we find uval==-EFAULT. We then jump to the pi_faulted label.
2) From pi_faulted: We increment attempt, unlock the sem and hit the retry label.
3) From the retry label, with ret still zero, we again hit EFAULT on the first futex_atomic_cmpxchg_inatomic(), and again goto the pi_faulted label.
4) Again from pi_faulted: we increment attempt and enter the conditional, where we call futex_handle_fault.
5) futex_handle_fault fails, and we goto the out_unlock_release_sem label.
6) From out_unlock_release_sem we return, and since ret is still zero, we return without error, while never actually unlocking the lock.

Issue #1: at the first futex_atomic_cmpxchg_inatomic() we should probably be setting ret=-EFAULT before jumping to pi_faulted: However in our case this doesn't really affect anything, as the glibc we're using ignores the error value from futex_unlock_pi().

Issue #2: Look at futex_handle_fault(), its first conditional will return -EFAULT if attempt is >= 2. However, from the "if(attempt++) futex_handle_fault(attempt)" logic above, we'll *never* call futex_handle_fault when attempt is less then two. So we never get a chance to even try to fault the page in.

This very simple and hackish fix for issue #2 is probably not the correct solution, but with the odd if(attempt++) logic all over futex.c it might actually be the right thing to do.

Your thoughts?

thanks
-john

kernel/futex.c +1 -1
Index: 2.6-rt/kernel/futex.c
===================================================================
@@ -298,7 +298,7 @@
 	struct vm_area_struct * vma;
 	struct mm_struct *mm = current->mm;
 
-	if (attempt >= 2 || !(vma = find_vma(mm, address)) ||
+	if (attempt > 2 || !(vma = find_vma(mm, address)) ||
 	    vma->vm_start > address || !(vma->vm_flags & VM_WRITE))
 		return -EFAULT;