Re: [patch 2/2]vmscan: correctly detect GFP_ATOMIC allocation failure

[Shaohua Li <shaohua.li@intel.com>]

On Sat, 2011-10-08 at 11:19 +0800, David Rientjes wrote:
> On Sat, 8 Oct 2011, Shaohua Li wrote:
> 
> > has_under_min_watermark_zone is used to detect if there is GFP_ATOMIC allocation
> > failure risk. For a high end_zone, if any zone below or equal to it has min
> > matermark ok, we have no risk. But current logic is any zone has min watermark
> > not ok, then we have risk. This is wrong to me.
> > 
> > Signed-off-by: Shaohua Li <shaohua.li@intel.com>
> > ---
> >  mm/vmscan.c |    7 ++++---
> >  1 file changed, 4 insertions(+), 3 deletions(-)
> > 
> > Index: linux/mm/vmscan.c
> > ===================================================================
> > --- linux.orig/mm/vmscan.c	2011-09-27 15:09:29.000000000 +0800
> > +++ linux/mm/vmscan.c	2011-09-27 15:14:45.000000000 +0800
> > @@ -2463,7 +2463,7 @@ loop_again:
> >  
> >  	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
> >  		unsigned long lru_pages = 0;
> > -		int has_under_min_watermark_zone = 0;
> > +		int has_under_min_watermark_zone = 1;
> 
> bool
> 
> >  
> >  		/* The swap token gets in the way of swapout... */
> >  		if (!priority)
> > @@ -2594,9 +2594,10 @@ loop_again:
> >  				 * means that we have a GFP_ATOMIC allocation
> >  				 * failure risk. Hurry up!
> >  				 */
> > -				if (!zone_watermark_ok_safe(zone, order,
> > +				if (has_under_min_watermark_zone &&
> > +					    zone_watermark_ok_safe(zone, order,
> >  					    min_wmark_pages(zone), end_zone, 0))
> > -					has_under_min_watermark_zone = 1;
> > +					has_under_min_watermark_zone = 0;
> >  			} else {
> >  				/*
> >  				 * If a zone reaches its high watermark,
> 
> Ignore checking the min watermark for a moment and consider if all zones 
> are above the high watermark (a situation where kswapd does not need to 
> do aggressive reclaim), then has_under_min_watermark_zone doesn't get 
> cleared and never actually stalls on congestion_wait().  Notice this is 
> congestion_wait() and not wait_iff_congested(), so the clearing of 
> ZONE_CONGESTED doesn't prevent this.
if all zones are above the high watermark, we will have i < 0 when
detecting the highest imbalanced zone, and the whole loop will end
without run into congestion_wait().
or I can add a clearing has_under_min_watermark_zone in the else block
to be safe.

--
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>