Re: [RFC PATCH -v2] mm, oom: introduce oom reaper

[Michal Hocko <mhocko@kernel.org>]

On Tue 08-12-15 07:19:42, Tetsuo Handa wrote:
> Michal Hocko wrote:
> > Yes you are right! The reference count should be incremented before
> > publishing the new mm_to_reap. I thought that an elevated ref. count by
> > the caller would be enough but this was clearly wrong. Does the update
> > below looks better?
> 
> I think that moving mmdrop() from oom_kill_process() to
> oom_reap_vmas() xor wake_oom_reaper() makes the patch simpler.

It surely is less lines of code but I am not sure it is simpler. I do
not think we should drop the reference in a different path than it is
taken.  Maybe we will grow more users of wake_oom_reaper in the future
and this is quite subtle behavior.

> 
>  	rcu_read_unlock();
>  
> +	if (can_oom_reap)
> +		wake_oom_reaper(mm); /* will call mmdrop() */
> +	else
> +		mmdrop(mm);
> -	mmdrop(mm);
>  	put_task_struct(victim);
>  }

Thanks!
-- 
Michal Hocko
SUSE Labs