Re: [patch 7/7] oom: prevent possible OOM_DISABLE livelock

[Nick Piggin]

On Mon, May 04, 2009 at 05:27:07PM -0700, David Rientjes wrote:
> It is currently possible to livelock the oom killer if a task is chosen
> for oom kill and another thread sharing the same memory has an oom_adj
> value of OOM_DISABLE.  This occurs because oom_kill_task() repeatedly

Hmm, but didn't the last patch make it a per-mm value?

> returns 1 and refuses to kill the chosen task while select_bad_process()
> will repeatedly chooses the same task during the next retry.
> 
> This moves the check for OOM_DISABLE to the badness heuristic while
> holding task_lock().  Badness scores of 0 are now explicitly prohibited
> from being oom killed and since the oom_adj value is a characteristic of
> an mm and not a task, it is no longer necessary to check the oom_adj
> value for threads sharing the same memory (except when simply issuing
> SIGKILLs for threads in other thread groups).
> 
> Cc: Nick Piggin <npiggin@xxxxxxx>
> Signed-off-by: David Rientjes <rientjes@xxxxxxxxxx>
> ---
>  mm/oom_kill.c |   40 ++++++++++------------------------------
>  1 files changed, 10 insertions(+), 30 deletions(-)
> 
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -67,6 +67,10 @@ unsigned long badness(struct task_struct *p, unsigned long uptime)
>  		return 0;
>  	}
>  	oom_adj = mm->oom_adj;
> +	if (oom_adj == OOM_DISABLE) {
> +		task_unlock(p);
> +		return 0;
> +	}
>  
>  	/*
>  	 * The memory size of the process is the basis for the badness.
> @@ -253,13 +257,8 @@ static struct task_struct *select_bad_process(unsigned long *ppoints,
>  			*ppoints = ULONG_MAX;
>  		}
>  
> -		task_lock(p);
> -		if (p->mm && p->mm->oom_adj == OOM_DISABLE)
> -			continue;
> -		task_unlock(p);
> -
>  		points = badness(p, uptime.tv_sec);
> -		if (points > *ppoints || !chosen) {
> +		if (points > *ppoints) {
>  			chosen = p;
>  			*ppoints = points;
>  		}
> @@ -352,32 +351,13 @@ static int oom_kill_task(struct task_struct *p)
>  	struct mm_struct *mm;
>  	struct task_struct *g, *q;
>  
> +	task_lock(p);
>  	mm = p->mm;
> -
> -	/* WARNING: mm may not be dereferenced since we did not obtain its
> -	 * value from get_task_mm(p).  This is OK since all we need to do is
> -	 * compare mm to q->mm below.
> -	 *
> -	 * Furthermore, even if mm contains a non-NULL value, p->mm may
> -	 * change to NULL at any time since we do not hold task_lock(p).
> -	 * However, this is of no concern to us.
> -	 */
> -
> -	if (mm == NULL)
> +	if (!mm || mm->oom_adj == OOM_DISABLE) {
> +		task_unlock(p);
>  		return 1;
> -
> -	/*
> -	 * Don't kill the process if any threads are set to OOM_DISABLE
> -	 */
> -	do_each_thread(g, q) {
> -		task_lock(q);
> -		if (q->mm == mm && q->mm && q->mm->oom_adj == OOM_DISABLE) {
> -			task_unlock(q);
> -			return 1;
> -		}
> -		task_unlock(q);
> -	} while_each_thread(g, q);
> -
> +	}
> +	task_unlock(p);
>  	__oom_kill_task(p, 1);
>  
>  	/*
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/