Re: [RFC v1 PATCH 3/6] led: pmic8058: Add PMIC8058 leds driver

[Lars-Peter Clausen <lars@metafoo.de>]

Dmitry Torokhov wrote:
> On Wed, Nov 10, 2010 at 09:45:05PM +0100, Lars-Peter Clausen wrote:
>> Trilok Soni wrote:
>>> +
>>> +static void led_kp_set(struct pmic8058_led_data *led, enum led_brightness value)
>>> +{
>>> +	int rc;
>>> +	u8 level;
>>> +	unsigned long flags;
>>> +
>>> +	spin_lock_irqsave(&led->value_lock, flags);
>> This function is only ever called from within the workqueue so there is no need for
>> locking.
>>
> 
> That is a common misconception, unfortunately. The same work may
> be executing on several CPUs at the same time if it was scheduled on
> multi-threaded work queue.
> 

Hm, right my fault.
Still the comment above is still valid, because the original workqueue handler was
locked by a mutex. But the comment regarding the mutex should have been:
 You can remove the mutex if you queue the work on the system_nrt_wq workqueue

> ...
> 
>>> +	schedule_work(&led->work);
> 
> And sure enough, keventd is such workqueue.
> 
> Now, whether having the same work run simultaneously is OK or not is a
> different question altogether...
> 

- Lars