On Mon, Aug 01, 2016 at 04:03:10PM +0300, Vladimir Davydov wrote:
> We must call shrink_slab() for each memory cgroup on both global and
> memcg reclaim in shrink_node_memcg(). Commit d71df22b55099 accidentally
> changed that so that now shrink_slab() is only called with memcg != NULL
> on memcg reclaim. As a result, memcg-aware shrinkers (including
> dentry/inode) are never invoked on global reclaim. Fix that.
>
> Fixes: d71df22b55099 ("mm, vmscan: begin reclaiming pages on a per-node basis")
> Signed-off-by: Vladimir Davydov <vdavydov@xxxxxxxxxxxxx>
Ouch, I missed that in the review. The change looked so obviously
correct, but the shrink_slab() interface is a little deceiving. It
would be great if shrink_slab() could handle root_mem_cgroup/NULL and
then we'd always call it from inside the loop. But AFAICS we need the
global call to get the cumulative scanned/lru_pages ratio from all of
the memcgs reclaimed... grr. Oh well.
This fix looks correct to me, anyway.
Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx>
--
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>