[PATCH] I2C: OMAP: fix runtime PM get/put balance on error

[PATCH] I2C: OMAP: fix runtime PM get/put balance on error

[Shubhrajyoti D]

    ensure pm_runtime_put() is called, on pm_runtime_get_sync()
    failure.

    Without this, after a failed call, the runtime PM usecount will have
    been incremented, but not decremented causing the usecount to never
    reach zero after a failure. Thanks to Kevin for educating about it. 
    While at it also fix a missing pm_runtime_disable in the probe error
    path.

Cc: Kevin Hilman <khilman@ti.com>
Signed-off-by: Shubhrajyoti D <shubhrajyoti@ti.com>
---
 drivers/i2c/busses/i2c-omap.c |    7 +++----
 1 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/i2c/busses/i2c-omap.c b/drivers/i2c/busses/i2c-omap.c
index 2500f19..c8e5c76 100644
--- a/drivers/i2c/busses/i2c-omap.c
+++ b/drivers/i2c/busses/i2c-omap.c
@@ -1113,10 +1113,10 @@ err_free_irq:
 	free_irq(dev->irq, dev);
 err_unuse_clocks:
 	omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
+err_free_mem:
 	pm_runtime_put(dev->dev);
 	iounmap(dev->base);
 	pm_runtime_disable(&pdev->dev);
-err_free_mem:
 	platform_set_drvdata(pdev, NULL);
 	kfree(dev);
 err_release_region:
@@ -1136,10 +1136,9 @@ static int __devexit omap_i2c_remove(struct platform_device *pdev)
 	free_irq(dev->irq, dev);
 	i2c_del_adapter(&dev->adapter);
 	ret = pm_runtime_get_sync(&pdev->dev);
-	if (IS_ERR_VALUE(ret))
-		return ret;
+	if (!IS_ERR_VALUE(ret))
+		omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
 
-	omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
 	pm_runtime_put(&pdev->dev);
 	pm_runtime_disable(&pdev->dev);
 	iounmap(dev->base);
-- 
1.7.5.4

Re: [PATCH] I2C: OMAP: fix runtime PM get/put balance on error

[Kevin Hilman]

Shubhrajyoti D <shubhrajyoti-l0cyMroinI0@public.gmane.org> writes:

>     ensure pm_runtime_put() is called, on pm_runtime_get_sync()
>     failure.
>
>     Without this, after a failed call, the runtime PM usecount will have
>     been incremented, but not decremented causing the usecount to never
>     reach zero after a failure. Thanks to Kevin for educating about it. 
>     While at it also fix a missing pm_runtime_disable in the probe error
>     path.

This is the same subject and changelog as the patch I sent, but is a
different patch.  

Please write a new subject and a changelog specific to your patch.

As this changes the error/failure path, please be specific about how
the failure modes were tested, and on which platforms.

> Cc: Kevin Hilman <khilman-l0cyMroinI0@public.gmane.org>
> Signed-off-by: Shubhrajyoti D <shubhrajyoti-l0cyMroinI0@public.gmane.org>
> ---
>  drivers/i2c/busses/i2c-omap.c |    7 +++----
>  1 files changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-omap.c b/drivers/i2c/busses/i2c-omap.c
> index 2500f19..c8e5c76 100644
> --- a/drivers/i2c/busses/i2c-omap.c
> +++ b/drivers/i2c/busses/i2c-omap.c
> @@ -1113,10 +1113,10 @@ err_free_irq:
>  	free_irq(dev->irq, dev);
>  err_unuse_clocks:
>  	omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
> +err_free_mem:
>  	pm_runtime_put(dev->dev);
>  	iounmap(dev->base);

This doesn't look right.  At least one of the gotos for this label, the
ioremap has failed.

>  	pm_runtime_disable(&pdev->dev);
> -err_free_mem:
>  	platform_set_drvdata(pdev, NULL);
>  	kfree(dev);
>  err_release_region:
> @@ -1136,10 +1136,9 @@ static int __devexit omap_i2c_remove(struct platform_device *pdev)
>  	free_irq(dev->irq, dev);
>  	i2c_del_adapter(&dev->adapter);
>  	ret = pm_runtime_get_sync(&pdev->dev);
> -	if (IS_ERR_VALUE(ret))
> -		return ret;
> +	if (!IS_ERR_VALUE(ret))
> +		omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);

this change isn't described in changelog

> -	omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
>  	pm_runtime_put(&pdev->dev);
>  	pm_runtime_disable(&pdev->dev);
>  	iounmap(dev->base);

Kevin

Re: [PATCH] I2C: OMAP: fix runtime PM get/put balance on error

[Shubhrajyoti]

On Friday 29 June 2012 05:32 PM, Kevin Hilman wrote:
> Shubhrajyoti D <shubhrajyoti@ti.com> writes:
>
>>     ensure pm_runtime_put() is called, on pm_runtime_get_sync()
>>     failure.
>>
>>     Without this, after a failed call, the runtime PM usecount will have
>>     been incremented, but not decremented causing the usecount to never
>>     reach zero after a failure. Thanks to Kevin for educating about it. 
>>     While at it also fix a missing pm_runtime_disable in the probe error
>>     path.
> This is the same subject and changelog as the patch I sent, but is a
> different patch.  
>
> Please write a new subject and a changelog specific to your patch.
OK.
Actually I did that on purpose your patch fixed the xfer call only.
I thought that since get_sync increments the count always we could extend
the patch to all the callers.
>
> As this changes the error/failure path, please be specific about how
> the failure modes were tested, and on which platforms.
I found and fixed by review only.
Didnt really hit the failure case.
>
>> Cc: Kevin Hilman <khilman@ti.com>
>> Signed-off-by: Shubhrajyoti D <shubhrajyoti@ti.com>
>> ---
>>  drivers/i2c/busses/i2c-omap.c |    7 +++----
>>  1 files changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/i2c/busses/i2c-omap.c b/drivers/i2c/busses/i2c-omap.c
>> index 2500f19..c8e5c76 100644
>> --- a/drivers/i2c/busses/i2c-omap.c
>> +++ b/drivers/i2c/busses/i2c-omap.c
>> @@ -1113,10 +1113,10 @@ err_free_irq:
>>  	free_irq(dev->irq, dev);
>>  err_unuse_clocks:
>>  	omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
>> +err_free_mem:
>>  	pm_runtime_put(dev->dev);
>>  	iounmap(dev->base);
> This doesn't look right.  At least one of the gotos for this label, the
> ioremap has failed.
Yes , missed that thanks will fix it.
>
>>  	pm_runtime_disable(&pdev->dev);
>> -err_free_mem:
>>  	platform_set_drvdata(pdev, NULL);
>>  	kfree(dev);
>>  err_release_region:
>> @@ -1136,10 +1136,9 @@ static int __devexit omap_i2c_remove(struct platform_device *pdev)
>>  	free_irq(dev->irq, dev);
>>  	i2c_del_adapter(&dev->adapter);
>>  	ret = pm_runtime_get_sync(&pdev->dev);
>> -	if (IS_ERR_VALUE(ret))
>> -		return ret;
>> +	if (!IS_ERR_VALUE(ret))
>> +		omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
> this change isn't described in changelog
Will do so.
>
>> -	omap_i2c_write_reg(dev, OMAP_I2C_CON_REG, 0);
>>  	pm_runtime_put(&pdev->dev);
>>  	pm_runtime_disable(&pdev->dev);
>>  	iounmap(dev->base);
> Kevin

Re: [PATCH] I2C: OMAP: fix runtime PM get/put balance on error

[Kevin Hilman]

Shubhrajyoti <shubhrajyoti-l0cyMroinI0@public.gmane.org> writes:

> On Friday 29 June 2012 05:32 PM, Kevin Hilman wrote:
>> Shubhrajyoti D <shubhrajyoti-l0cyMroinI0@public.gmane.org> writes:
>>
>>>     ensure pm_runtime_put() is called, on pm_runtime_get_sync()
>>>     failure.
>>>
>>>     Without this, after a failed call, the runtime PM usecount will have
>>>     been incremented, but not decremented causing the usecount to never
>>>     reach zero after a failure. Thanks to Kevin for educating about it. 
>>>     While at it also fix a missing pm_runtime_disable in the probe error
>>>     path.
>> This is the same subject and changelog as the patch I sent, but is a
>> different patch.  
>>
>> Please write a new subject and a changelog specific to your patch.
> OK.
> Actually I did that on purpose your patch fixed the xfer call only.
> I thought that since get_sync increments the count always we could extend
> the patch to all the callers.
>>
>> As this changes the error/failure path, please be specific about how
>> the failure modes were tested, and on which platforms.
> I found and fixed by review only.
> Didnt really hit the failure case.

Which is why you should't just add stuff to other peoples work.

My patch was targetted at the XFER problem during suspend/resume which I
found by testing and and proved the fix by testing.

Your patch adds several additional functional changes that were not well
described and not tested.  That calls for a separate patch with seprate
subject/changelog etc.

Kevin