Re: [RFC][PATCH 5/4] memcg: don't call res_counter_uncharge whenobsolete

[KAMEZAWA Hiroyuki]

On Thu, 15 Jan 2009 13:38:14 +0900
Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx> wrote:

> On Thu, 15 Jan 2009 11:14:20 +0900, Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx> wrote:
> > > > > To handle the problem "parent may be obsolete",
> > > > > 
> > > > > call mem_cgroup_get(parent) at create()
> > > > > call mem_cgroup_put(parent) at freeing memcg.
> > > > >      (regardless of use_hierarchy.)
> > > > > 
> > > > > is clearer way to go, I think.
> > > > > 
> > > > > I wonder whether there is  mis-accounting problem or not..
> > > > > 
> hmm, after more consideration, although this patch can prevent the BUG,
> it can leak memsw accounting of parents because memsw of parents, which
> have been incremented by charge, does not decremented.
> 
> I'll try pet/put parent approach..
> Or any other good ideas ?
> 
"don't free parent's css_id until all chilredn are freed" is necessary for my
CSS ID. I'll rebase my patch on to yours.

thanks,
-Kame




> 
> Thanks,
> Daisuke Nishimura.
> 
> > > > > So, adding css_tryget() around problematic code can be a fix.
> > > > > --
> > > > >   mem = swap_cgroup_record();
> > > > >   if (css_tryget(&mem->css)) {
> > > > >       res_counter_uncharge(&mem->memsw, PAZE_SIZE);
> > > > >       css_put(&mem->css)
> > > > >   }
> > > > > --
> > > > > I like css_tryget() rather than mem_cgroup_obsolete().
> > > > I agree.
> > > > The updated version is attached.
> > > > 
> > > > 
> > > > Thanks,
> > > > Daisuke nishimura.
> > > > 
> > > > > To be honest, I'd like to remove memcg special stuff when I can.
> > > > > 
> > > > > Thanks,
> > > > > -Kame
> > > > > 
> > > > ===
> > > > From: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> > > > 
> > > > mem_cgroup_get ensures that the memcg that has been got can be accessed
> > > > even after the directory has been removed, but it doesn't ensure that parents
> > > > of it can be accessed: parents might have been freed already by rmdir.
> > > > 
> > > > This causes a bug in case of use_hierarchy==1, because res_counter_uncharge
> > > > climb up the tree.
> > > > 
> > > > Check if the memcg is obsolete by css_tryget, and don't call
> > > > res_counter_uncharge when obsole.
> > > > 
> > > > Signed-off-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> > > seems nice loock.
> > > 
> > > 
> > > > ---
> > > >  mm/memcontrol.c |   15 ++++++++++++---
> > > >  1 files changed, 12 insertions(+), 3 deletions(-)
> > > > 
> > > > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > > > index fb62b43..4e3b100 100644
> > > > --- a/mm/memcontrol.c
> > > > +++ b/mm/memcontrol.c
> > > > @@ -1182,7 +1182,10 @@ int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
> > > >  		/* avoid double counting */
> > > >  		mem = swap_cgroup_record(ent, NULL);
> > > >  		if (mem) {
> > > > -			res_counter_uncharge(&mem->memsw, PAGE_SIZE);
> > > > +			if (!css_tryget(&mem->css)) {
> > > > +				res_counter_uncharge(&mem->memsw, PAGE_SIZE);
> > > > +				css_put(&mem->css);
> > > > +			}
> > > >  			mem_cgroup_put(mem);
> > > >  		}
> > > >  	}
> > > 
> > > I think css_tryget() returns "ture" at success....
> > > 
> > > So,
> > > ==
> > > 	if (mem && css_tryget(&mem->css))
> > > 		res_counter....
> > > 
> > > is correct.
> > > 
> > > -Kame
> > > 
> > Ooops! you are right.
> > Sorry for my silly mistake..
> > 
> > "mem" is checked beforehand, so I think css_tryget would be enough.
> > I'm now testing the attached one.
> > 
> > 
> > Thanks,
> > Daisuke Nishimura.
> > ===
> > From: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> > 
> > mem_cgroup_get ensures that the memcg that has been got can be accessed
> > even after the directory has been removed, but it doesn't ensure that parents
> > of it can be accessed: parents might have been freed already by rmdir.
> > 
> > This causes a bug in case of use_hierarchy==1, because res_counter_uncharge
> > climb up the tree.
> > 
> > Check if the memcg is obsolete by css_tryget, and don't call
> > res_counter_uncharge when obsole.
> > 
> > Signed-off-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> > ---
> >  mm/memcontrol.c |   15 ++++++++++++---
> >  1 files changed, 12 insertions(+), 3 deletions(-)
> > 
> > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > index fb62b43..b9d5271 100644
> > --- a/mm/memcontrol.c
> > +++ b/mm/memcontrol.c
> > @@ -1182,7 +1182,10 @@ int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
> >  		/* avoid double counting */
> >  		mem = swap_cgroup_record(ent, NULL);
> >  		if (mem) {
> > -			res_counter_uncharge(&mem->memsw, PAGE_SIZE);
> > +			if (css_tryget(&mem->css)) {
> > +				res_counter_uncharge(&mem->memsw, PAGE_SIZE);
> > +				css_put(&mem->css);
> > +			}
> >  			mem_cgroup_put(mem);
> >  		}
> >  	}
> > @@ -1252,7 +1255,10 @@ void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
> >  		struct mem_cgroup *memcg;
> >  		memcg = swap_cgroup_record(ent, NULL);
> >  		if (memcg) {
> > -			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
> > +			if (css_tryget(&memcg->css)) {
> > +				res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
> > +				css_put(&memcg->css);
> > +			}
> >  			mem_cgroup_put(memcg);
> >  		}
> >  
> > @@ -1397,7 +1403,10 @@ void mem_cgroup_uncharge_swap(swp_entry_t ent)
> >  
> >  	memcg = swap_cgroup_record(ent, NULL);
> >  	if (memcg) {
> > -		res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
> > +		if (css_tryget(&memcg->css)) {
> > +			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
> > +			css_put(&memcg->css);
> > +		}
> >  		mem_cgroup_put(memcg);
> >  	}
> >  }
> 

--
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/