[PATCH] Input: ads7846 - move regulator codes out of spinlock protected area

[PATCH] Input: ads7846 - move regulator codes out of spinlock protected area

[Jason Wang]

The commit 9114337 introduces regulator operations in the ads7846
touchscreen driver. Among these operations, some are called in the
spinlock protected context.
On most platforms, the regulator operation is achieved through
i2c/spi bus transfer operations, some of bus transfer operations will
call wait_for_completion function. It isn't allowable  to call
sleepable function in the atomic context. So move them out from the
atomic context.

[tested on TI OMAP3530EVM board]

Signed-off-by: Jason Wang <jason77.wang@gmail.com>
---
 drivers/input/touchscreen/ads7846.c |   14 ++++++++++----
 1 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/input/touchscreen/ads7846.c b/drivers/input/touchscreen/ads7846.c
index 1603193..9421df9 100644
--- a/drivers/input/touchscreen/ads7846.c
+++ b/drivers/input/touchscreen/ads7846.c
@@ -531,6 +531,9 @@ static ssize_t ads7846_disable_store(struct device *dev,
 	if (strict_strtoul(buf, 10, &i))
 		return -EINVAL;
 
+	if (!i)
+		regulator_enable(ts->reg);
+
 	spin_lock_irq(&ts->lock);
 
 	if (i)
@@ -540,6 +543,9 @@ static ssize_t ads7846_disable_store(struct device *dev,
 
 	spin_unlock_irq(&ts->lock);
 
+	if (i)
+		regulator_disable(ts->reg);
+
 	return count;
 }
 
@@ -855,8 +861,6 @@ static void ads7846_disable(struct ads7846 *ts)
 		}
 	}
 
-	regulator_disable(ts->reg);
-
 	/* we know the chip's in lowpower mode since we always
 	 * leave it that way after every request
 	 */
@@ -868,8 +872,6 @@ static void ads7846_enable(struct ads7846 *ts)
 	if (!ts->disabled)
 		return;
 
-	regulator_enable(ts->reg);
-
 	ts->disabled = 0;
 	ts->irq_disabled = 0;
 	enable_irq(ts->spi->irq);
@@ -886,6 +888,8 @@ static int ads7846_suspend(struct spi_device *spi, pm_message_t message)
 
 	spin_unlock_irq(&ts->lock);
 
+	regulator_disable(ts->reg);
+
 	if (device_may_wakeup(&ts->spi->dev))
 		enable_irq_wake(ts->spi->irq);
 
@@ -900,6 +904,8 @@ static int ads7846_resume(struct spi_device *spi)
 	if (device_may_wakeup(&ts->spi->dev))
 		disable_irq_wake(ts->spi->irq);
 
+	regulator_enable(ts->reg);
+
 	spin_lock_irq(&ts->lock);
 
 	ts->is_suspended = 0;
-- 
1.5.6.5

Re: [PATCH] Input: ads7846 - move regulator codes out of spinlock protected area

[Dmitry Torokhov]

On Thu, Aug 19, 2010 at 04:14:30PM +0800, Jason Wang wrote:
> The commit 9114337 introduces regulator operations in the ads7846
> touchscreen driver. Among these operations, some are called in the
> spinlock protected context.
> On most platforms, the regulator operation is achieved through
> i2c/spi bus transfer operations, some of bus transfer operations will
> call wait_for_completion function. It isn't allowable  to call
> sleepable function in the atomic context. So move them out from the
> atomic context.

I do not believe simply moving calls out of splnlock-protected area is
enough. Are all regulator drivers allow regulator_enable() and
regulator_disable() to be called simultaneously? Even if they do allow
it I think there still a race between ads7846_enable/disable/suspend/resume
and you need to wrap all of it in a mutex...

> 
> [tested on TI OMAP3530EVM board]
> 
> Signed-off-by: Jason Wang <jason77.wang@gmail.com>
> ---
>  drivers/input/touchscreen/ads7846.c |   14 ++++++++++----
>  1 files changed, 10 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/input/touchscreen/ads7846.c b/drivers/input/touchscreen/ads7846.c
> index 1603193..9421df9 100644
> --- a/drivers/input/touchscreen/ads7846.c
> +++ b/drivers/input/touchscreen/ads7846.c
> @@ -531,6 +531,9 @@ static ssize_t ads7846_disable_store(struct device *dev,
>  	if (strict_strtoul(buf, 10, &i))
>  		return -EINVAL;
>  
> +	if (!i)
> +		regulator_enable(ts->reg);
> +
>  	spin_lock_irq(&ts->lock);
>  
>  	if (i)
> @@ -540,6 +543,9 @@ static ssize_t ads7846_disable_store(struct device *dev,
>  
>  	spin_unlock_irq(&ts->lock);
>  
> +	if (i)
> +		regulator_disable(ts->reg);
> +
>  	return count;
>  }
>  
> @@ -855,8 +861,6 @@ static void ads7846_disable(struct ads7846 *ts)
>  		}
>  	}
>  
> -	regulator_disable(ts->reg);
> -
>  	/* we know the chip's in lowpower mode since we always
>  	 * leave it that way after every request
>  	 */
> @@ -868,8 +872,6 @@ static void ads7846_enable(struct ads7846 *ts)
>  	if (!ts->disabled)
>  		return;
>  
> -	regulator_enable(ts->reg);
> -
>  	ts->disabled = 0;
>  	ts->irq_disabled = 0;
>  	enable_irq(ts->spi->irq);
> @@ -886,6 +888,8 @@ static int ads7846_suspend(struct spi_device *spi, pm_message_t message)
>  
>  	spin_unlock_irq(&ts->lock);
>  
> +	regulator_disable(ts->reg);
> +
>  	if (device_may_wakeup(&ts->spi->dev))
>  		enable_irq_wake(ts->spi->irq);
>  
> @@ -900,6 +904,8 @@ static int ads7846_resume(struct spi_device *spi)
>  	if (device_may_wakeup(&ts->spi->dev))
>  		disable_irq_wake(ts->spi->irq);
>  
> +	regulator_enable(ts->reg);
> +
>  	spin_lock_irq(&ts->lock);
>  
>  	ts->is_suspended = 0;
> -- 
> 1.5.6.5
> 

-- 
Dmitry

Re: [PATCH] Input: ads7846 - move regulator codes out of spinlock protected area

[Jason Wang]

Dmitry Torokhov wrote:
> On Thu, Aug 19, 2010 at 04:14:30PM +0800, Jason Wang wrote:
>   
>> The commit 9114337 introduces regulator operations in the ads7846
>> touchscreen driver. Among these operations, some are called in the
>> spinlock protected context.
>> On most platforms, the regulator operation is achieved through
>> i2c/spi bus transfer operations, some of bus transfer operations will
>> call wait_for_completion function. It isn't allowable  to call
>> sleepable function in the atomic context. So move them out from the
>> atomic context.
>>     
>
> I do not believe simply moving calls out of splnlock-protected area is
> enough. Are all regulator drivers allow regulator_enable() and
> regulator_disable() to be called simultaneously? Even if they do allow
> it I think there still a race between ads7846_enable/disable/suspend/resume
> and you need to wrap all of it in a mutex...
>
>   
Hi Dmitry,

The regulator_enable() and regulator_disable() are already protected by 
a mutex,
see drivers/regulator/core.c

int regulator_enable(struct regulator *regulator)
{
struct regulator_dev *rdev = regulator->rdev;
int ret = 0;

mutex_lock(&rdev->mutex);
ret = _regulator_enable(rdev);
mutex_unlock(&rdev->mutex);
return ret;
}
EXPORT_SYMBOL_GPL(regulator_enable);

int regulator_disable(struct regulator *regulator)
{
struct regulator_dev *rdev = regulator->rdev;
int ret = 0;

mutex_lock(&rdev->mutex);
ret = _regulator_disable(rdev);
mutex_unlock(&rdev->mutex);
return ret;
}
EXPORT_SYMBOL_GPL(regulator_disable);

So there is no race between these two functions.

You are right, i will design a protection wrap for 
enable/disable/suspend/resume.

Thanks,
Jason.

>> [tested on TI OMAP3530EVM board]
>>
>> Signed-off-by: Jason Wang <jason77.wang@gmail.com>
>> ---
>>  drivers/input/touchscreen/ads7846.c |   14 ++++++++++----
>>  1 files changed, 10 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/input/touchscreen/ads7846.c b/drivers/input/touchscreen/ads7846.c
>> index 1603193..9421df9 100644
>> --- a/drivers/input/touchscreen/ads7846.c
>> +++ b/drivers/input/touchscreen/ads7846.c
>> @@ -531,6 +531,9 @@ static ssize_t ads7846_disable_store(struct device *dev,
>>  	if (strict_strtoul(buf, 10, &i))
>>  		return -EINVAL;
>>  
>> +	if (!i)
>> +		regulator_enable(ts->reg);
>> +
>>  	spin_lock_irq(&ts->lock);
>>  
>>  	if (i)
>> @@ -540,6 +543,9 @@ static ssize_t ads7846_disable_store(struct device *dev,
>>  
>>  	spin_unlock_irq(&ts->lock);
>>  
>> +	if (i)
>> +		regulator_disable(ts->reg);
>> +
>>  	return count;
>>  }
>>  
>> @@ -855,8 +861,6 @@ static void ads7846_disable(struct ads7846 *ts)
>>  		}
>>  	}
>>  
>> -	regulator_disable(ts->reg);
>> -
>>  	/* we know the chip's in lowpower mode since we always
>>  	 * leave it that way after every request
>>  	 */
>> @@ -868,8 +872,6 @@ static void ads7846_enable(struct ads7846 *ts)
>>  	if (!ts->disabled)
>>  		return;
>>  
>> -	regulator_enable(ts->reg);
>> -
>>  	ts->disabled = 0;
>>  	ts->irq_disabled = 0;
>>  	enable_irq(ts->spi->irq);
>> @@ -886,6 +888,8 @@ static int ads7846_suspend(struct spi_device *spi, pm_message_t message)
>>  
>>  	spin_unlock_irq(&ts->lock);
>>  
>> +	regulator_disable(ts->reg);
>> +
>>  	if (device_may_wakeup(&ts->spi->dev))
>>  		enable_irq_wake(ts->spi->irq);
>>  
>> @@ -900,6 +904,8 @@ static int ads7846_resume(struct spi_device *spi)
>>  	if (device_may_wakeup(&ts->spi->dev))
>>  		disable_irq_wake(ts->spi->irq);
>>  
>> +	regulator_enable(ts->reg);
>> +
>>  	spin_lock_irq(&ts->lock);
>>  
>>  	ts->is_suspended = 0;
>> -- 
>> 1.5.6.5
>>
>>     
>
>