Re: [PATCH 44/49] mm: numa: Add THP migration for the NUMA workingset scanning fault case.

From: Mel Gorman
Date: Mon Jan 07 2013 - 10:37:13 EST


On Sat, Jan 05, 2013 at 04:42:29PM +0800, Wanpeng Li wrote:
> >+int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
> >+{
> >+ int ret = 0;
> >
> > /* Avoid migrating to a node that is nearly full */
> > if (migrate_balanced_pgdat(pgdat, 1)) {
>
> Hi Mel Gorman,
>
> This parameter nr_migrate_pags = 1 is not correct, since balancenuma also
> support THP in this patchset, the parameter should be 1 <= compound_order(page)
> instead of 1. I'd rather change to something like:
>

True. The impact is marginal because it only applies when a node is almost
full but it does mean that we do some unnecessary work before migration
fails anyway. I've added a TODO item to fix it when I next revisit NUMA
balancing. Thanks.

--
Mel Gorman
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/