[PATCH] kauditd thread API fixes

[PATCH] kauditd thread API fixes

[Jeff Garzik]

Kill the "function should return a value" warning, and use the standard
mechanism for deciding when a thread should stop.

Signed-off-by: Jeff Garzik <jeff@garzik.org>

---

Given the security-sensitive nature of the audit function, _perhaps_ the
omission of kthread_should_stop() was intentional.  If so, it at least
deserves a comment.  But I think this is more correct.

diff --git a/kernel/audit.c b/kernel/audit.c
index f9889ee..b9146b1 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -340,7 +340,7 @@ static int kauditd_thread(void *dummy)
 {
 	struct sk_buff *skb;
 
-	while (1) {
+	do {
 		skb = skb_dequeue(&audit_skb_queue);
 		wake_up(&audit_backlog_wait);
 		if (skb) {
@@ -368,7 +368,9 @@ static int kauditd_thread(void *dummy)
 			__set_current_state(TASK_RUNNING);
 			remove_wait_queue(&kauditd_wait, &wait);
 		}
-	}
+	} while (!kthread_should_stop());
+
+	return 0;
 }
 
 int audit_send_list(void *_dest)

Re: [PATCH] kauditd thread API fixes

[Andrew Morton]

On Sun, 1 Oct 2006 08:30:45 -0400
Jeff Garzik <jeff@garzik.org> wrote:

> 
> Kill the "function should return a value" warning, and use the standard
> mechanism for deciding when a thread should stop.
> 
> Signed-off-by: Jeff Garzik <jeff@garzik.org>
> 
> ---
> 
> Given the security-sensitive nature of the audit function, _perhaps_ the
> omission of kthread_should_stop() was intentional.  If so, it at least
> deserves a comment.  But I think this is more correct.
> 
> diff --git a/kernel/audit.c b/kernel/audit.c
> index f9889ee..b9146b1 100644
> --- a/kernel/audit.c
> +++ b/kernel/audit.c
> @@ -340,7 +340,7 @@ static int kauditd_thread(void *dummy)
>  {
>  	struct sk_buff *skb;
>  
> -	while (1) {
> +	do {
>  		skb = skb_dequeue(&audit_skb_queue);
>  		wake_up(&audit_backlog_wait);
>  		if (skb) {
> @@ -368,7 +368,9 @@ static int kauditd_thread(void *dummy)
>  			__set_current_state(TASK_RUNNING);
>  			remove_wait_queue(&kauditd_wait, &wait);
>  		}
> -	}
> +	} while (!kthread_should_stop());
> +
> +	return 0;
>  }
>  
>  int audit_send_list(void *_dest)

I've had a basically-identical patch in -mm for a few months.  Sent to viro once
or twice, nothing happened, gave up.  I'll re-add it to the
maintainer-spamming-list.