Re: [PATCH 6/24] make atomic_read() behave consistently on frv

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Wed, Aug 15, 2007 at 12:01:54AM +0200, Arnd Bergmann wrote:
> On Tuesday 14 August 2007, Paul E. McKenney wrote:
> > > #define order(x) asm volatile("" : "+m" (x))
> > 
> > There was something very similar discussed earlier in this thread,
> > with quite a bit of debate as to exactly what the "m" flag should
> > look like.  I suggested something similar named ACCESS_ONCE in the
> > context of RCU (http://lkml.org/lkml/2007/7/11/664):
> > 
> >         #define ACCESS_ONCE(x) (*(volatile typeof(x) *)&(x))
> > 
> > The nice thing about this is that it works for both loads and stores.
> > Not clear that order() above does this -- I get compiler errors when
> > I try something like "b = order(a)" or "order(a) = 1" using gcc 4.1.2.
> 
> Well, it serves a different purpose: While your ACCESS_ONCE() macro is
> an lvalue, the order() macro is a statement that can be used in place
> of the barrier() macro. order() is the most lightweight barrier as it
> only enforces ordering on a single variable in the compiler, but does
> not have any side-effects visible to other threads, like the cache
> line access in ACCESS_ONCE has.

ACCESS_ONCE() is indeed intended to be used when actually loading or
storing the variable.  That said, I must admit that it is not clear to me
why you would want to add an extra order() rather than ACCESS_ONCE()ing
one or both of the adjacent accesses to that same variable.

So, what am I missing?

						Thanx, Paul