Re: [PATCH -mm 1/1] audit: cleanup audit_log_start/wait_for_auditd mess

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 3 Jun 2013 22:46:20 +0200 Oleg Nesterov <oleg@redhat.com> wrote:

> audit_log_start() tries to obfusticate the code/logic and
> reimplements wait_event_timeout() in a very confusing way.
> 
> Just change wait_for_auditd() to use wait_event_timeout()
> and return bool.
> 
> Also remove the unnecessary "ab = NULL" initialization.

I would prefer that this be nicely tested - for a cosmetic thing I
don't think it's worth the risk as-is.

> --- a/kernel/audit.c
> +++ b/kernel/audit.c
> @@ -1051,20 +1051,19 @@ static inline void audit_get_stamp(struct audit_context *ctx,
>  }
>  
>  /*
> - * Wait for auditd to drain the queue a little
> + * Wait for auditd to drain audit_skb_queue if it is full
>   */
> -static void wait_for_auditd(unsigned long sleep_time)
> +static bool wait_for_auditd(gfp_t gfp_mask)
>  {
> -	DECLARE_WAITQUEUE(wait, current);
> -	set_current_state(TASK_UNINTERRUPTIBLE);
> -	add_wait_queue(&audit_backlog_wait, &wait);
> -
> -	if (audit_backlog_limit &&
> -	    skb_queue_len(&audit_skb_queue) > audit_backlog_limit)
> -		schedule_timeout(sleep_time);
> -
> -	__set_current_state(TASK_RUNNING);
> -	remove_wait_queue(&audit_backlog_wait, &wait);
> +	/* Atomic callers use can 5 more entries over the limit */

"can use"

> +	if (!(gfp_mask & __GFP_WAIT))
> +		return skb_queue_len(&audit_skb_queue) <=
> +					audit_backlog_limit + 5;
> +
> +	return wait_event_timeout(audit_backlog_wait,
> +		!audit_backlog_limit ||
> +			skb_queue_len(&audit_skb_queue) <= audit_backlog_limit,
> +		audit_backlog_wait_time);
>  }
>
> ...
>