On Fri, Jun 12, 2015 at 03:05:00PM +0800, Hillf Danton wrote:
> > - /* Reclaim above the high watermark. */
> > - sc->nr_to_reclaim = max(SWAP_CLUSTER_MAX, high_wmark_pages(zone));
> > + /* Aim to reclaim above all the zone high watermarks */
> > + for (z = 0; z <= end_zone; z++) {
> > + zone = pgdat->node_zones + end_zone;
> s/end_zone/z/ ?
Ouch, thanks!
With this bug, kswapd would reclaim based on a multiple of the highest
zone. Whether that was under or over reclaim would depend on the size of
that zone relative to lower zones.
> > + nr_to_reclaim += high_wmark_pages(zone);
> >
> [...]
> > @@ -3280,13 +3177,26 @@ static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
> > compact_pgdat(pgdat, order);
> >
> > /*
> > + * Stop reclaiming if any eligible zone is balanced and clear
> > + * node writeback or congested.
> > + */
> > + for (i = 0; i <= *classzone_idx; i++) {
> > + zone = pgdat->node_zones + i;
> > +
> > + if (zone_balanced(zone, sc.order, 0, *classzone_idx)) {
> > + clear_bit(PGDAT_CONGESTED, &pgdat->flags);
> > + clear_bit(PGDAT_DIRTY, &pgdat->flags);
> > + break;
> s/break/goto out/ ?
Yes. It'd actually be ok because it'll detect the same condition and
exit in the next outer loop but goto out is better.
--
Mel Gorman
SUSE Labs
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>