Re: [patch] mm: memcg: make sure memory.events is uptodate when waking pollers fix

[Michal Hocko]

On Fri 06-04-18 11:54:41, Johannes Weiner wrote:
> On Fri, Apr 06, 2018 at 02:53:30PM +1000, Stephen Rothwell wrote:
> > Hi Andrew,
> > 
> > After merging the akpm-current tree, today's linux-next build (x86_64
> > allmodconfig) produced this warning:
> > 
> > mm/memcontrol.c: In function 'memory_events_show':
> > mm/memcontrol.c:5453:23: warning: array subscript is above array bounds [-Warray-bounds]
> >       atomic_long_read(&memcg->memory_events[OOM_KILL]));
> >                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > 
> > Introduced by commit
> > 
> >   725844c87a0d ("mm: memcg: make sure memory.events is uptodate when waking pollers")
> 
> Ah, silly, I missed that warning. The below fixes it.
> 
> Signed-off-by: Johannes Weiner <hannes@xxxxxxxxxxx>
> ---
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 04ea2b104ee2..9bac6dc26e09 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -5199,8 +5199,7 @@ static int memory_events_show(struct seq_file *m, void *v)
>  		   atomic_long_read(&memcg->memory_events[MEMCG_MAX]));
>  	seq_printf(m, "oom %lu\n",
>  		   atomic_long_read(&memcg->memory_events[MEMCG_OOM]));
> -	seq_printf(m, "oom_kill %lu\n",
> -		   atomic_long_read(&memcg->memory_events[OOM_KILL]));
> +	seq_printf(m, "oom_kill %lu\n", memcg_sum_events(memcg, OOM_KILL));

Ohh, this is so easy to miss.

-- 
Michal Hocko
SUSE Labs