On Tue, Feb 28, 2012 at 09:30:17PM -0800, Hugh Dickins wrote:
> Swapping tmpfs loads show absurd wrapped rss and wrong cache in memcg's
> memory.stat statistics: __mem_cgroup_uncharge_common() is failing to
> distinguish the anon and tmpfs cases.
>
> Mostly we can decide between them by PageAnon, which is reliable once
> it has been set; but there are several callers who need to uncharge a
> MEM_CGROUP_CHARGE_TYPE_MAPPED page before it was fully initialized,
> so allow that case to override the PageAnon decision.
>
> Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
> ---
>
> mm/memcontrol.c | 7 ++++---
> 1 file changed, 4 insertions(+), 3 deletions(-)
>
> --- 3.3-rc5-next/mm/memcontrol.c 2012-02-25 10:06:52.496035568 -0800
> +++ linux/mm/memcontrol.c 2012-02-26 10:44:32.146365398 -0800
> @@ -2944,13 +2944,16 @@ __mem_cgroup_uncharge_common(struct page
> if (!PageCgroupUsed(pc))
> goto unlock_out;
>
> + anon = PageAnon(page);
> +
> switch (ctype) {
> case MEM_CGROUP_CHARGE_TYPE_MAPPED:
> + anon = true;
> + /* fallthrough */
If you don't mind, could you add a small comment on why this is the
exception where we don't trust page->mapping?
Other than that,
Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx>
> case MEM_CGROUP_CHARGE_TYPE_DROP:
> /* See mem_cgroup_prepare_migration() */
> if (page_mapped(page) || PageCgroupMigration(pc))
> goto unlock_out;
> - anon = true;
> break;
> case MEM_CGROUP_CHARGE_TYPE_SWAPOUT:
> if (!PageAnon(page)) { /* Shared memory */
> @@ -2958,10 +2961,8 @@ __mem_cgroup_uncharge_common(struct page
> goto unlock_out;
> } else if (page_mapped(page)) /* Anon */
> goto unlock_out;
> - anon = true;
> break;
> default:
> - anon = false;
> break;
> }
--
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/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>