Re: [PATCH] mm/memcontrol: fix wrong statistics in memory.stat

[Shakeel Butt <shakeelb@google.com>]

On Tue, Jul 2, 2019 at 9:28 PM Yafang Shao <laoar.shao@gmail.com> wrote:
>
> On Wed, Jul 3, 2019 at 11:50 AM Shakeel Butt <shakeelb@google.com> wrote:
> >
> > +Johannes Weiner
> >
> > On Tue, Jul 2, 2019 at 6:23 PM Yafang Shao <laoar.shao@gmail.com> wrote:
> > >
> > > When we calculate total statistics for memcg1_stats and memcg1_events, we
> > > use the the index 'i' in the for loop as the events index.
> > > Actually we should use memcg1_stats[i] and memcg1_events[i] as the
> > > events index.
> > >
> > > Fixes: 8de7ecc6483b ("memcg: reduce memcg tree traversals for stats collection")
> >
> > Actually it fixes 42a300353577 ("mm: memcontrol: fix recursive
> > statistics correctness & scalabilty").
> >
>
> Hi Shakeel,
>
> In 8de7ecc6483b, this code was changed from memcg_page_state(mi,
> memcg1_stats[i]) to acc.stat[i].
>
> -               for_each_mem_cgroup_tree(mi, memcg)
> -                       val += memcg_page_state(mi, memcg1_stats[i]) *
> -                       PAGE_SIZE;
> -               seq_printf(m, "total_%s %llu\n", memcg1_stat_names[i], val);
> +               seq_printf(m, "total_%s %llu\n", memcg1_stat_names[i],
> +                          (u64)acc.stat[i] * PAGE_SIZE);
>
> In 42a300353577, this code was changed from acc.vmstats[i] to
> memcg_events(memcg, i).
> -                          (u64)acc.vmstats[i] * PAGE_SIZE);
> +                          (u64)memcg_page_state(memcg, i) * PAGE_SIZE);
>
> So seems this issue was introduced in 8de7ecc6483b, isn't it ?
>
>

That's the reason I said 8de7ecc6483b made it subtle but not wrong.
Check accumulate_memcg_tree() in 8de7ecc6483b, the memcg_page_state()
and memcg_events() are called with correct index but saved at 'i'
index in acc array.


> > > Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> > > Cc: Shakeel Butt <shakeelb@google.com>
> > > Cc: Michal Hocko <mhocko@suse.com>
> > > Cc: Yafang Shao <shaoyafang@didiglobal.com>
> > > ---
> > >  mm/memcontrol.c | 5 +++--
> > >  1 file changed, 3 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > > index 3ee806b..2ad94d0 100644
> > > --- a/mm/memcontrol.c
> > > +++ b/mm/memcontrol.c
> > > @@ -3528,12 +3528,13 @@ static int memcg_stat_show(struct seq_file *m, void *v)
> > >                 if (memcg1_stats[i] == MEMCG_SWAP && !do_memsw_account())
> > >                         continue;
> > >                 seq_printf(m, "total_%s %llu\n", memcg1_stat_names[i],
> > > -                          (u64)memcg_page_state(memcg, i) * PAGE_SIZE);
> > > +                          (u64)memcg_page_state(memcg, memcg1_stats[i]) *
> > > +                          PAGE_SIZE);
> >
> > It seems like I made the above very subtle in 8de7ecc6483b and
> > Johannes missed this subtlety in 42a300353577 (and I missed it in the
> > review).
> >
> > >         }
> > >
> > >         for (i = 0; i < ARRAY_SIZE(memcg1_events); i++)
> > >                 seq_printf(m, "total_%s %llu\n", memcg1_event_names[i],
> > > -                          (u64)memcg_events(memcg, i));
> > > +                          (u64)memcg_events(memcg, memcg1_events[i]));
> > >
> > >         for (i = 0; i < NR_LRU_LISTS; i++)
> > >                 seq_printf(m, "total_%s %llu\n", mem_cgroup_lru_names[i],
> > > --
> > > 1.8.3.1
> > >
> >
> > Reviewed-by: Shakeel Butt <shakeelb@google.com>