Re: [patch 8/8] mm: memcg: modify PageCgroupCache non-atomically

[Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>]

On Thu, Nov 24, 2011 at 09:13:28AM +0900, KAMEZAWA Hiroyuki wrote:
> On Wed, 23 Nov 2011 16:42:31 +0100
> Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org> wrote:
> 
> > From: Johannes Weiner <jweiner-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
> > 
> > This bit is protected by lock_page_cgroup(), there is no need for
> > locked operations when setting and clearing it.
> > 
> > Signed-off-by: Johannes Weiner <jweiner-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
> 
> Hm. non-atomic ops for pc->flags seems dangerous.
> How about try to remove PCG_CACHE ? Maybe we can depends on PageAnon(page).
> We see 'page' on memcg->lru now.
> I'm sorry I forgot why we needed PCG_CACHE flag..

The problem is that we charge/uncharged pages that are not fully
rmapped and so PageAnon() is not reliable.  I forgot if there are more
places, but the commit_charge in migration was a prominent one.

I have a patch set that reworks migration so to only commit pages that
are fully rmapped but it clashed with the THP patches and I didn't see
too much value to fix it up.  But I should probably revive it, because
it makes some things simpler.

As I replied to Hugh, it might even work for PCG_CACHE, but it's
definitely dangerous and not worth the complex dependencies it brings
on other parts of the code, so please consider 7/8 and 8/8 dropped.
--
To unsubscribe from this list: send the line "unsubscribe cgroups" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html