Re: [PATCH 25/34] mm, vmscan: avoid passing in classzone_idx unnecessarily to compaction_ready

[Vlastimil Babka]

On 07/08/2016 11:35 AM, Mel Gorman wrote:
> The scan_control structure has enough information available for
> compaction_ready() to make a decision. The classzone_idx manipulations in
> shrink_zones() are no longer necessary as the highest populated zone is
> no longer used to determine if shrink_slab should be called or not.
>
> Signed-off-by: Mel Gorman <mgorman@xxxxxxxxxxxxxxxxxxx>
> Acked-by: Hillf Danton <hillf.zj@xxxxxxxxxxxxxxx>

Acked-by: Vlastimil Babka <vbabka@xxxxxxx>

> @@ -2621,8 +2609,8 @@ static void shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
>  			 */
>  			if (IS_ENABLED(CONFIG_COMPACTION) &&
>  			    sc->order > PAGE_ALLOC_COSTLY_ORDER &&
> -			    zonelist_zone_idx(z) <= classzone_idx &&
> -			    compaction_ready(zone, sc->order, classzone_idx)) {
> +			    zonelist_zone_idx(z) <= sc->reclaim_idx &&

Hm I notice that the condition on the line above should be always true 
as the same sc->reclaim_idx value is used to limit zone_idx in 
for_each_zone_zonelist_nodemask(). The implication was probably true 
even before this patch (classzone_idx would be <= sc->reclaim_idx), but 
now it stands out clearly.

> +			    compaction_ready(zone, sc)) {
>  				sc->compaction_ready = true;
>  				continue;
>  			}