Re: [RFC 2/5] memcg: rework mem_cgroup_iter to use cgroup iterators

[Michal Hocko]

On Tue 13-11-12 08:14:42, Tejun Heo wrote:
> On Tue, Nov 13, 2012 at 04:30:36PM +0100, Michal Hocko wrote:
> > @@ -1063,8 +1063,8 @@ struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
> >  				   struct mem_cgroup *prev,
> >  				   struct mem_cgroup_reclaim_cookie *reclaim)
> >  {
> > -	struct mem_cgroup *memcg = NULL;
> > -	int id = 0;
> > +	struct mem_cgroup *memcg = NULL,
> > +			  *last_visited = NULL;
> 
> Nitpick but please don't do this.

OK, will make it grep friendlier;

> > +		/*
> > +		 * Root is not visited by cgroup iterators so it needs a special
> > +		 * treatment.
> > +		 */
> > +		if (!last_visited) {
> > +			css = &root->css;
> > +		} else {
> > +			struct cgroup *next_cgroup;
> > +
> > +			next_cgroup = cgroup_next_descendant_pre(
> > +					last_visited->css.cgroup,
> > +					root->css.cgroup);
> > +			if (next_cgroup)
> > +				css = cgroup_subsys_state(next_cgroup,
> > +						mem_cgroup_subsys_id);
> 
> Hmmm... wouldn't it be better to move the reclaim logic into a
> function and do the following?
> 
> 	reclaim(root);
> 	for_each_descendent_pre()
> 		reclaim(descendant);

We cannot do for_each_descendent_pre here because we do not iterate
through the whole hierarchy all the time. Check shrink_zone.

> If this is a problem, I'd be happy to add a iterator which includes
> the top node.  

This would help with the above if-else but I do not think this is the
worst thing in the function ;)

> I'd prefer controllers not using the next functions directly.

Well, we will need to use it directly because of the single group
reclaim mentioned above.

-- 
Michal Hocko
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/