Re: [patch 00/15] clocksource / timekeeping rework V4 (resend V3 + bug fix)

[Martin Schwidefsky <schwidefsky@de.ibm.com>]

On Sat, 15 Aug 2009 12:08:35 +0200 (CEST)
Thomas Gleixner <tglx@linutronix.de> wrote:

> Yeah, we need to call clocksource_change_rating() outside of watchdog
> lock. Spotted a missing _irqrestore as well.
> 
> Thanks,
> 
> 	tglx
> ---
> 
> diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c
> index a1657b5..02dc22d 100644
> --- a/kernel/time/clocksource.c
> +++ b/kernel/time/clocksource.c
> @@ -307,16 +307,23 @@ static void clocksource_watchdog_work(struct work_struct *work)
>  {
>  	struct clocksource *cs, *tmp;
>  	unsigned long flags;
> +	LIST_HEAD(unstable);
> 
>  	spin_lock_irqsave(&watchdog_lock, flags);
>  	list_for_each_entry_safe(cs, tmp, &watchdog_list, wd_list)
>  		if (cs->flags & CLOCK_SOURCE_UNSTABLE) {
>  			list_del_init(&cs->wd_list);
> -			clocksource_change_rating(cs, 0);
> +			list_add(&cs->wd_list, &unstable);
>  		}
>  	/* Check if the watchdog timer needs to be stopped. */
>  	clocksource_stop_watchdog();
> -	spin_unlock(&watchdog_lock);
> +	spin_unlock_irqrestore(&watchdog_lock, flags);
> +
> +	/* Needs to be done outside of watchdog lock */
> +	list_for_each_entry_safe(cs, tmp, &unstable, wd_list) {
> +		list_del_init(&cs->wd_list);
> +		clocksource_change_rating(cs, 0);
> +	}
>  }
> 
>  #else /* CONFIG_CLOCKSOURCE_WATCHDOG */

Autsch, the s/spin_unlock/spin_unlock_irqrestore/ is rather obvious, no?
In theory we now have a reference count problem. The call to 
clocksource_change_rating are no done outside of the clocksource mutex
and outside of the watchdog lock. A concurrent clocksource_unregister
could remove the clock.

-- 
blue skies,
   Martin.

"Reality continues to ruin my life." - Calvin.