Re: [PATCH v2 3/9] memcg: do not flush stats in irq context

[Michal Hocko <mhocko@suse.com>]

On Tue 28-03-23 22:16:38, Yosry Ahmed wrote:
> Currently, the only context in which we can invoke an rstat flush from
> irq context is through mem_cgroup_usage() on the root memcg when called
> from memcg_check_events(). An rstat flush is an expensive operation that
> should not be done in irq context, so do not flush stats and use the
> stale stats in this case.
> 
> Arguably, usage threshold events are not reliable on the root memcg
> anyway since its usage is ill-defined.
> 
> Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
> Suggested-by: Shakeel Butt <shakeelb@google.com>
> Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
> Acked-by: Shakeel Butt <shakeelb@google.com>
> Acked-by: Johannes Weiner <hannes@cmpxchg.org>

Acked-by: Michal Hocko <mhocko@suse.com>

> ---
>  mm/memcontrol.c | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index c3b6aae78901..ff39f78f962e 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -3669,7 +3669,21 @@ static unsigned long mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
>  	unsigned long val;
>  
>  	if (mem_cgroup_is_root(memcg)) {
> -		mem_cgroup_flush_stats();
> +		/*
> +		 * We can reach here from irq context through:
> +		 * uncharge_batch()
> +		 * |--memcg_check_events()
> +		 *    |--mem_cgroup_threshold()
> +		 *       |--__mem_cgroup_threshold()
> +		 *          |--mem_cgroup_usage
> +		 *
> +		 * rstat flushing is an expensive operation that should not be
> +		 * done from irq context; use stale stats in this case.
> +		 * Arguably, usage threshold events are not reliable on the root
> +		 * memcg anyway since its usage is ill-defined.
> +		 */
> +		if (in_task())
> +			mem_cgroup_flush_stats();
>  		val = memcg_page_state(memcg, NR_FILE_PAGES) +
>  			memcg_page_state(memcg, NR_ANON_MAPPED);
>  		if (swap)
> -- 
> 2.40.0.348.gf938b09366-goog

-- 
Michal Hocko
SUSE Labs