[PATCH][RFC] dirty balancing for cgroups

[KAMEZAWA Hiroyuki]

On Wed, 13 Aug 2008 16:15:05 +0900 (JST)
yamamoto at valinux.co.jp (YAMAMOTO Takashi) wrote:

> hi,
> 
> > > @@ -485,7 +502,10 @@ unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
> > >  		if (PageUnevictable(page) ||
> > >  		    (PageActive(page) && !active) ||
> > >  		    (!PageActive(page) && active)) {
> > > -			__mem_cgroup_move_lists(pc, page_lru(page));
> > > +			if (try_lock_page_cgroup(page)) {
> > > +				__mem_cgroup_move_lists(pc, page_lru(page));
> > > +				unlock_page_cgroup(page);
> > > +			}
> > >  			continue;
> > >  		}
> > 
> > This chunk seems unrelated and lost....
> 
> it's necessary to protect from mem_cgroup_{set,clear}_dirty
> which modify pc->flags without holding mz->lru_lock.
> 

I'm now writing a patch to make page_cgroup->flags to be atomic_ops.
Don't worry about this.
(With remove-page-lock-cgroup patch, atomic_ops patch's performace is
 quite well.)

Thanks,
-Kame