Re: [PATCH] mm: compaction: Abort compaction if too many pages areisolated and caller is asynchronous

[Minchan Kim]

On Mon, Jun 06, 2011 at 11:39:24AM +0100, Mel Gorman wrote:
> On Fri, Jun 03, 2011 at 05:45:54PM +0200, Andrea Arcangeli wrote:
> > On Fri, Jun 03, 2011 at 03:49:41PM +0100, Mel Gorman wrote:
> > > Right idea of the wrong zone being accounted for but wrong place. I
> > > think the following patch should fix the problem;
> > 
> > Looks good thanks.
> > 
> > I also found this bug during my debugging that made NR_SHMEM underflow.
> > 
> > ===
> > Subject: migrate: don't account swapcache as shmem
> > 
> > From: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > 
> > swapcache will reach the below code path in migrate_page_move_mapping,
> > and swapcache is accounted as NR_FILE_PAGES but it's not accounted as
> > NR_SHMEM.
> > 
> > Signed-off-by: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> 
> Well spotted.
> 
> Acked-by: Mel Gorman <mgorman@xxxxxxx>
> 
> Minor nit. swapper_space is rarely referred to outside of the swap
> code. Might it be more readable to use
> 
> 	/*
> 	 * swapcache is accounted as NR_FILE_PAGES but it is not
> 	 * accounted as NR_SHMEM
> 	 *
> 	if (PageSwapBacked(page) && !PageSwapCache(page))

I like this. but as it's "and" operation, CPU have to execute two condition comparison.
but how about below?
	if (!PageSwapCache(page) && PageSwapBacked(page))

PageSwapCache implys PageSwapBacked so we can handle non-swapbacked pages as just 1 comparison.

> 
> ?
> 
> > ---
> > 
> > diff --git a/mm/migrate.c b/mm/migrate.c
> > index e4a5c91..2597a27 100644
> > --- a/mm/migrate.c
> > +++ b/mm/migrate.c
> > @@ -288,7 +288,7 @@ static int migrate_page_move_mapping(struct address_space *mapping,
> >  	 */
> >  	__dec_zone_page_state(page, NR_FILE_PAGES);
> >  	__inc_zone_page_state(newpage, NR_FILE_PAGES);
> > -	if (PageSwapBacked(page)) {
> > +	if (mapping != &swapper_space && PageSwapBacked(page)) {
> >  		__dec_zone_page_state(page, NR_SHMEM);
> >  		__inc_zone_page_state(newpage, NR_SHMEM);
> >  	}
> > 
> > 
> 
> -- 
> Mel Gorman
> SUSE Labs

-- 
Kind regards
Minchan Kim
--
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/