Re: [PATCH] 1/5 VM changes: zone-pressure.patch

[Andrew Morton]

Nikita Danilov <[e-mail]> wrote:
>
> Add zone->pressure field. It estimates scanning intensity for this zone and
>  is calculated as exponentially decaying average of the scanning priority
>  required to free enough pages in this zone (mm/vmscan.c:zone_adj_pressure()).
> 
>  zone->pressure is supposed to be used in stead of scanning priority by
>  vmscan.c. This is used by later patches in a series.
> 

OK, fixes a bug.

> 
>  diff -puN include/linux/mmzone.h~zone-pressure include/linux/mmzone.h
>  --- i386/include/linux/mmzone.h~zone-pressure	Wed Jul  9 12:24:50 2003
>  +++ i386-god/include/linux/mmzone.h	Wed Jul  9 12:24:50 2003
>  @@ -84,11 +84,23 @@ struct zone {
>   	atomic_t		refill_counter;
>   	unsigned long		nr_active;
>   	unsigned long		nr_inactive;
>  -	int			all_unreclaimable; /* All pages pinned */
>  +	int			all_unreclaimable:1; /* All pages pinned */

I'll revert this change.  Once we start adding bitfields in there they all
need common locking and it gets messy.  integers are simpler.