Re: [PATCH v6 1/7] mm/memcg: scanning_global_lru means mem_cgroup_disabled

[Johannes Weiner <hannes@cmpxchg.org>]

On Mon, Mar 26, 2012 at 05:04:29PM +0200, Michal Hocko wrote:
> [Adding Johannes to CC]
> 
> On Fri 23-03-12 01:56:16, Konstantin Khlebnikov wrote:
> > From: Hugh Dickins <hughd@google.com>
> > 
> > Although one has to admire the skill with which it has been concealed,
> > scanning_global_lru(mz) is actually just an interesting way to test
> > mem_cgroup_disabled().  Too many developer hours have been wasted on
> > confusing it with global_reclaim(): just use mem_cgroup_disabled().
> 
> Is this really correct?

Yes, if the memory controller is enabled, we never have a global LRU
and always scan the per-memcg lists.

> > Signed-off-by: Hugh Dickins <hughd@google.com>
> > Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
> > Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > Acked-by: Glauber Costa <glommer@parallels.com>
> > ---
> >  mm/vmscan.c |   18 ++++--------------
> >  1 files changed, 4 insertions(+), 14 deletions(-)
> > 
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 49f15ef..c684f44 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> [...]
> > @@ -1806,7 +1796,7 @@ static int inactive_anon_is_low(struct mem_cgroup_zone *mz)
> >  	if (!total_swap_pages)
> >  		return 0;
> >  
> > -	if (!scanning_global_lru(mz))
> > +	if (!mem_cgroup_disabled())
> >  		return mem_cgroup_inactive_anon_is_low(mz->mem_cgroup,
> >  						       mz->zone);
> 
> mem_cgroup_inactive_anon_is_low calculation is slightly different than
> what we have for cgroup_disabled case. calculate_zone_inactive_ratio
> considers _all_ present pages in the zone while memcg variant only
> active+inactive.

The memcg has nothing to go by but actual number of LRU pages; there
is no 'present pages' equivalent.

I don't think that it matters much in reality given the sqrt scale,
but the difference is still unfortunate.  Konstantin was meaning to
unify all this, though.