[PATCH] media: rc: remove unneeded mutex in rc_register_device

[PATCH] media: rc: remove unneeded mutex in rc_register_device

[Heiner Kallweit]

Access to dev->initialized is atomic, therefore we don't have to
protect it with a mutex.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/media/rc/rc-main.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/media/rc/rc-main.c b/drivers/media/rc/rc-main.c
index 4e9bbe7..68541b1 100644
--- a/drivers/media/rc/rc-main.c
+++ b/drivers/media/rc/rc-main.c
@@ -1492,9 +1492,7 @@ int rc_register_device(struct rc_dev *dev)
 	}
 
 	/* Allow the RC sysfs nodes to be accessible */
-	mutex_lock(&dev->lock);
 	atomic_set(&dev->initialized, 1);
-	mutex_unlock(&dev->lock);
 
 	IR_dprintk(1, "Registered rc%u (driver: %s, remote: %s, mode %s)\n",
 		   dev->minor,
-- 
2.7.3

Re: [PATCH] media: rc: remove unneeded mutex in rc_register_device

[Vladimir Zapolskiy]

On 16.03.2016 20:53, Heiner Kallweit wrote:
> Access to dev->initialized is atomic, therefore we don't have to
> protect it with a mutex.

Mutexes are used to split the code to mutually exclusive execution blocks,
so not arguing about the apparently correct change itself I want to
emphasize that the given explanation of the change in the commit message is
wrong. Atomic access does not cancel a specific care about execution
ordering.

Indirectly it applies to ("rc-core: allow calling rc_open with device not
initialized"), where "initialized" bool property was changed to atomic_t
type --- this (sub-)change is just useless.

Please grasp the topic and reword the commit message.

> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> ---
>  drivers/media/rc/rc-main.c | 2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/drivers/media/rc/rc-main.c b/drivers/media/rc/rc-main.c
> index 4e9bbe7..68541b1 100644
> --- a/drivers/media/rc/rc-main.c
> +++ b/drivers/media/rc/rc-main.c
> @@ -1492,9 +1492,7 @@ int rc_register_device(struct rc_dev *dev)
>  	}
>  
>  	/* Allow the RC sysfs nodes to be accessible */
> -	mutex_lock(&dev->lock);
>  	atomic_set(&dev->initialized, 1);
> -	mutex_unlock(&dev->lock);
>  
>  	IR_dprintk(1, "Registered rc%u (driver: %s, remote: %s, mode %s)\n",
>  		   dev->minor,
> 

--
With best wishes,
Vladimir

Re: [PATCH] media: rc: remove unneeded mutex in rc_register_device

[Heiner Kallweit]

Am 16.03.2016 um 20:33 schrieb Vladimir Zapolskiy:
> On 16.03.2016 20:53, Heiner Kallweit wrote:
>> Access to dev->initialized is atomic, therefore we don't have to
>> protect it with a mutex.
> 
> Mutexes are used to split the code to mutually exclusive execution blocks,
> so not arguing about the apparently correct change itself I want to
> emphasize that the given explanation of the change in the commit message is
> wrong. Atomic access does not cancel a specific care about execution
> ordering.
> 
> Indirectly it applies to ("rc-core: allow calling rc_open with device not
> initialized"), where "initialized" bool property was changed to atomic_t
> type --- this (sub-)change is just useless.
> 
> Please grasp the topic and reword the commit message.
> 
Thanks for the hint, you're right.
Will change the commit message and send a v2.

Regards, Heiner

>> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
>> ---
>>  drivers/media/rc/rc-main.c | 2 --
>>  1 file changed, 2 deletions(-)
>>
>> diff --git a/drivers/media/rc/rc-main.c b/drivers/media/rc/rc-main.c
>> index 4e9bbe7..68541b1 100644
>> --- a/drivers/media/rc/rc-main.c
>> +++ b/drivers/media/rc/rc-main.c
>> @@ -1492,9 +1492,7 @@ int rc_register_device(struct rc_dev *dev)
>>  	}
>>  
>>  	/* Allow the RC sysfs nodes to be accessible */
>> -	mutex_lock(&dev->lock);
>>  	atomic_set(&dev->initialized, 1);
>> -	mutex_unlock(&dev->lock);
>>  
>>  	IR_dprintk(1, "Registered rc%u (driver: %s, remote: %s, mode %s)\n",
>>  		   dev->minor,
>>
> 
> --
> With best wishes,
> Vladimir
>