Re: [patch 3/3] mm: PageActive no testset

[Marcelo Tosatti]

On Thu, Jan 19, 2006 at 03:50:08PM +0100, Nick Piggin wrote:
> Hi Marcelo,
> 
> On Wed, Jan 18, 2006 at 12:13:46PM -0200, Marcelo Tosatti wrote:
> > Hi Nick,
> > 
> > On Wed, Jan 18, 2006 at 11:40:58AM +0100, Nick Piggin wrote:
> > > PG_active is protected by zone->lru_lock, it does not need TestSet/TestClear
> > > operations.
> > 
> > page->flags bits (including PG_active and PG_lru bits) are touched by
> > several codepaths which do not hold zone->lru_lock. 
> > 
> > AFAICT zone->lru_lock guards access to the LRU list, and no more than
> > that.
> > 
> 
> Yep.
> 
> > Moreover, what about consistency of the rest of page->flags bits?
> > 
> 
> That's OK, set_bit and clear_bit are atomic as well, they just don't
> imply memory barriers and can be implemented a bit more simply.

Indeed!

> The test-set / test-clear operations also kind of imply that it is
> being used for locking or without other synchronisation (usually).

Non-atomic versions such as __ClearPageLRU()/__ClearPageActive() are 
not usable, though.

PPC:

static __inline__ void __clear_bit(unsigned long nr,
                                   volatile unsigned long *addr)
{
        unsigned long mask = BITOP_MASK(nr);
        unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);

        *p &= ~mask;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/