Re: [PATCH][BUGFIX] memcg fix zone congestion

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Thu, 12 May 2011 23:25:00 -0700
Andrew Morton <akpm@linux-foundation.org> wrote:

> On Fri, 13 May 2011 12:10:30 +0900 KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> 
> > 
> > ZONE_CONGESTED should be a state of global memory reclaim.
> > If not, a busy memcg sets this and give unnecessary throttoling in
> > wait_iff_congested() against memory recalim in other contexts. This makes
> > system performance bad.
> > 
> > I'll think about "memcg is congested!" flag is required or not, later.
> > But this fix is required 1st.
> > 
> > Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > ---
> >  mm/vmscan.c |    3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > Index: mmotm-May11/mm/vmscan.c
> > ===================================================================
> > --- mmotm-May11.orig/mm/vmscan.c
> > +++ mmotm-May11/mm/vmscan.c
> > @@ -941,7 +941,8 @@ keep_lumpy:
> >  	 * back off and wait for congestion to clear because further reclaim
> >  	 * will encounter the same problem
> >  	 */
> > -	if (nr_dirty == nr_congested && nr_dirty != 0)
> > +	if (scanning_global_lru(sc) &&
> > +	    nr_dirty == nr_congested && nr_dirty != 0)
> >  		zone_set_flag(zone, ZONE_CONGESTED);
> >  
> 
> nit: which is more probable?  nr_dirty==nr_congested or
> scanning_global_lru(sc)?  
> 
> If the user is actually _using_ memcg then
> 

If the user uses memcg, yes, nr_dirty == nr_congested is more probable.
If user doesn't, scanning_global_lru() returns always true.

> --- a/mm/vmscan.c~a
> +++ a/mm/vmscan.c
> @@ -937,7 +937,7 @@ keep_lumpy:
>  	 * back off and wait for congestion to clear because further reclaim
>  	 * will encounter the same problem
>  	 */
> -	if (nr_dirty == nr_congested && nr_dirty != 0)
> +	if (nr_dirty == nr_congested && scanning_global_lru(sc) && nr_dirty)
>  		zone_set_flag(zone, ZONE_CONGESTED);
>  
>  	free_page_list(&free_pages);
> 
> 
> is more efficient.  If the user isn't using memcg then your patch is faster?
> 

Hmm, maybe your fix is always faster. But in many case, if nr_congested == nr_dirty,
nr_dirty == 0 because vmscan just finds clean pages...fast path.
So, nr_dirty should be 1st ?

How about this ?
==

ZONE_CONGESTED should be a state of global memory reclaim.
Changelog:v1->v2
 - fixed the order of conditions.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
---
 mm/vmscan.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: mmotm-May11/mm/vmscan.c
===================================================================
--- mmotm-May11.orig/mm/vmscan.c
+++ mmotm-May11/mm/vmscan.c
@@ -941,7 +941,7 @@ keep_lumpy:
 	 * back off and wait for congestion to clear because further reclaim
 	 * will encounter the same problem
 	 */
-	if (nr_dirty == nr_congested && nr_dirty != 0)
+	if (nr_dirty && nr_dirty == nr_congested &&  scanning_global_lru(sc))
 		zone_set_flag(zone, ZONE_CONGESTED);
 
 	free_page_list(&free_pages);





--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>