Re: [PATCH 1/2] fork: extend clone3() to support CLONE_SET_TID

[Oleg Nesterov <oleg@redhat.com>]

On 07/29, Adrian Reber wrote:
>
> @@ -186,12 +187,26 @@ struct pid *alloc_pid(struct pid_namespace *ns)
>  		if (idr_get_cursor(&tmp->idr) > RESERVED_PIDS)
>  			pid_min = RESERVED_PIDS;
>  
> +		if (set_tid) {
> +			if ((set_tid >= pid_max) || ((set_tid != 1) &&
> +				(idr_get_cursor(&tmp->idr) <= 1))) {

I think the

	(set_tid != 1) && idr_get_cursor() <= 1

check needs a comment...

> +				spin_unlock_irq(&pidmap_lock);
> +				retval = -EINVAL;
> +				goto out_free;
> +			}
> +			min_p = set_tid;
> +			max_p = set_tid + 1;
> +			set_tid = 0;
> +		} else {
> +			min_p = pid_min;
> +			max_p = pid_max;
> +		}
>  		/*
>  		 * Store a null pointer so find_pid_ns does not find
>  		 * a partially initialized PID (see below).
>  		 */
> -		nr = idr_alloc_cyclic(&tmp->idr, NULL, pid_min,
> -				      pid_max, GFP_ATOMIC);
> +		nr = idr_alloc_cyclic(&tmp->idr, NULL, min_p,
> +				      max_p, GFP_ATOMIC);

do we really want _cyclic() which updates idr->idr_next if set_tid?

perhaps idr_alloc() makes more sense?

Oleg.