Re: [PATCH 5/5] nommu,oom: Set MMF_OOM_SKIP without waiting for termination.

[Michal Hocko <mhocko@suse.com>]

On Wed 08-11-17 20:01:48, Tetsuo Handa wrote:
[...]
> @@ -829,7 +831,7 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
>  	unsigned int victim_points = 0;
>  	static DEFINE_RATELIMIT_STATE(oom_rs, DEFAULT_RATELIMIT_INTERVAL,
>  					      DEFAULT_RATELIMIT_BURST);
> -	bool can_oom_reap = true;
> +	bool can_oom_reap = IS_ENABLED(CONFIG_MMU);
>  
>  	/*
>  	 * If the task is already exiting, don't alarm the sysadmin or kill
> @@ -929,7 +931,6 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
>  			continue;
>  		if (is_global_init(p)) {
>  			can_oom_reap = false;
> -			set_bit(MMF_OOM_SKIP, &mm->flags);
>  			pr_info("oom killer %d (%s) has mm pinned by %d (%s)\n",
>  					task_pid_nr(victim), victim->comm,
>  					task_pid_nr(p), p->comm);
> @@ -947,6 +948,8 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
>  
>  	if (can_oom_reap)
>  		wake_oom_reaper(victim);
> +	else
> +		set_bit(MMF_OOM_SKIP, &mm->flags);
>  
>  	mmdrop(mm);
>  	put_task_struct(victim);

Also this looks completely broken. nommu kernels lose the premature oom
killing protection almost completely (they simply rely on the sleep
before dropping the oom_lock).
-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>