RE: Fix unlock_buffer() to work the same way as bit_unlock()

["Chen, Kenneth W" <kenneth.w.chen@intel.com>]

Christoph Lameter wrote on Tuesday, March 28, 2006 3:48 PM
> On Tue, 28 Mar 2006, Zoltan Menyhart wrote:
> 
> > Why not to use separate bit operations for different purposes?
> > 
> > - e.g. "test_and_set_bit_N_acquire()" for lock acquisition
> > - "test_and_set_bit()", "clear_bit()" as they are today
> > - "release_N_clear_bit()"...
> > 
> 
> That would force IA64 specifics onto all other architectures.
> 
> Could we simply define these smb_mb__*_clear_bit to be noops
> and then make the atomic bit ops to have full barriers? That would satisfy 
> Nick's objections.
> 
> --- linux-2.6.16.orig/include/asm-ia64/bitops.h	2006-03-19 21:53:29.000000000 -0800
> +++ linux-2.6.16/include/asm-ia64/bitops.h	2006-03-28 15:45:08.000000000 -0800
> @@ -45,6 +45,7 @@
>  		old = *m;
>  		new = old | bit;
>  	} while (cmpxchg_acq(m, old, new) != old);
> +	smb_mb();
>  }

There are better way to do it.  The pointer is already cast as volatile,
so old = *m has acq semantics built-in, we can just change cmpxchg_acq to
cmpxchg_rel, then effectively it is a full memory barrier without doing the
expensive smp_mb().

- Ken