[CC'ing Uli because John forgot to with the original patch]
As John found with futex_unlock_pi (which that patch should apply nicely
to the -rt tree), the code can get screwed up with faulting in at the
cmpxchg in futex_unlock_pi code.
>From John's original email:
----
In looking into why the kernel was returning -EFAULT, I found the
following:
...
retry_locked:
/*
* To avoid races, try to do the TID -> 0 atomic transition
* again. If it succeeds then we can return without waking
* anyone else up:
*/
if (!(uval & FUTEX_OWNER_DIED)) {
pagefault_disable();
uval = futex_atomic_cmpxchg_inatomic(uaddr,
current->pid, 0);
pagefault_enable();
}
if (unlikely(uval == -EFAULT))
goto pi_faulted;
...[snip]...
pi_faulted:
/*
* We have to r/w *(int __user *)uaddr, but we can't modify it
* non-atomically. Therefore, if get_user below is not
* enough, we need to handle the fault ourselves, while
* still holding the mmap_sem.
*/
if (attempt++) {
ret = futex_handle_fault((unsigned long)uaddr, fshared,
attempt);
if (ret)
goto out_unlock;
goto retry_locked;
}
----
We see that uval is -EFAULT when jumping to pi_faulted. But when it goes
back to retry_locked, the uval is still -EFAULT. Which just happens to
have the FUTEX_OWNER_DIED bit set. So we don't do the
futex_atomic_cmpxchg_inatomic call, and go to the next conditional which
just so happens to be a check of uval for -EFAULT. And guess what? it
still is!!
Way to go John in finding this!!
But, the -rt kernel has pretty much the same code for the
futex_unlock_pi64, and it has the same bug. Hence this email.
This patch is the same fix that John posted, but simply for the
futex_unlock_pi64 that exists in the RT kernel and not the mainline.
Signed-off-by: Steven Rostedt <[EMAIL PROTECTED]>
Index: linux-2.6-rt-work/kernel/futex.c
===================================================================
--- linux-2.6-rt-work.orig/kernel/futex.c 2007-05-30 22:29:04.000000000
-0400
+++ linux-2.6-rt-work/kernel/futex.c 2007-05-30 22:32:01.000000000 -0400
@@ -3503,6 +3503,7 @@ pi_faulted:
ret = -EFAULT;
goto out_unlock;
}
+ uval = 0;
goto retry_locked;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
