On Wed, Jul 01, 2009 at 03:03:32PM +0800, Herbert Xu wrote:
quoted text
> On Wed, Jul 01, 2009 at 08:27:32AM +0200, Andi Kleen wrote:
> >
> > mfence is not needed for normal C code (not using non temporal
> > stores) in the Linux memory model on x86 and is a no-op. Only the compile
> > time barrier matters.
> 
> In that case this bug needs to be digged deeper regardless of
> this patch.

Agreed.

I suspect the reordering of the wake queue tests might makes a difference, but
in this case to ensure they are always tested in the proper order by
the compiler would need more smp_rmb()s

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html