[PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Christophe JAILLET]

There is no error handling path in the probe function.
Switch to managed resource so that errors in the probe are handled easily
and simplify the remove function accordingly.

Fixes: db3a4f0abefd ("misc/pvpanic: add PCI driver")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
 drivers/misc/pvpanic/pvpanic-pci.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/drivers/misc/pvpanic/pvpanic-pci.c b/drivers/misc/pvpanic/pvpanic-pci.c
index 9ecc4e8559d5..046ce4ecc195 100644
--- a/drivers/misc/pvpanic/pvpanic-pci.c
+++ b/drivers/misc/pvpanic/pvpanic-pci.c
@@ -78,15 +78,15 @@ static int pvpanic_pci_probe(struct pci_dev *pdev,
 	void __iomem *base;
 	int ret;
 
-	ret = pci_enable_device(pdev);
+	ret = pcim_enable_device(pdev);
 	if (ret < 0)
 		return ret;
 
-	base = pci_iomap(pdev, 0, 0);
+	base = pcim_iomap(pdev, 0, 0);
 	if (!base)
 		return -ENOMEM;
 
-	pi = kmalloc(sizeof(*pi), GFP_ATOMIC);
+	pi = devm_kmalloc(&pdev->dev, sizeof(*pi), GFP_ATOMIC);
 	if (!pi)
 		return -ENOMEM;
 
@@ -107,9 +107,6 @@ static void pvpanic_pci_remove(struct pci_dev *pdev)
 	struct pvpanic_instance *pi = dev_get_drvdata(&pdev->dev);
 
 	pvpanic_remove(pi);
-	iounmap(pi->base);
-	kfree(pi);
-	pci_disable_device(pdev);
 }
 
 static struct pci_driver pvpanic_pci_driver = {
-- 
2.30.2

[PATCH 2/2] misc/pvpanic: Use GFP_KERNEL instead of GFP_ATOMIC

[Christophe JAILLET]

There is no need to use GFP_ATOMIC in a probe function. Use GFP_KERNEL
instead.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
 drivers/misc/pvpanic/pvpanic-pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/misc/pvpanic/pvpanic-pci.c b/drivers/misc/pvpanic/pvpanic-pci.c
index 046ce4ecc195..3d7f9efb3dd4 100644
--- a/drivers/misc/pvpanic/pvpanic-pci.c
+++ b/drivers/misc/pvpanic/pvpanic-pci.c
@@ -86,7 +86,7 @@ static int pvpanic_pci_probe(struct pci_dev *pdev,
 	if (!base)
 		return -ENOMEM;
 
-	pi = devm_kmalloc(&pdev->dev, sizeof(*pi), GFP_ATOMIC);
+	pi = devm_kmalloc(&pdev->dev, sizeof(*pi), GFP_KERNEL);
 	if (!pi)
 		return -ENOMEM;
 
-- 
2.30.2

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Andy Shevchenko]

On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
> There is no error handling path in the probe function.
> Switch to managed resource so that errors in the probe are handled easily
> and simplify the remove function accordingly.

Yes, that's what I suggested earlier to another contributor.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

Thanks!

P.S. You may consider the following things as well:
 1) converting to use pci_set_drvdata() / pci_get_drvdata()
 2) providing devm_pvpanic_probe() [via devm_add_action() /
    devm_add_action_or_reset()]

> Fixes: db3a4f0abefd ("misc/pvpanic: add PCI driver")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
>  drivers/misc/pvpanic/pvpanic-pci.c | 9 +++------
>  1 file changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/misc/pvpanic/pvpanic-pci.c b/drivers/misc/pvpanic/pvpanic-pci.c
> index 9ecc4e8559d5..046ce4ecc195 100644
> --- a/drivers/misc/pvpanic/pvpanic-pci.c
> +++ b/drivers/misc/pvpanic/pvpanic-pci.c
> @@ -78,15 +78,15 @@ static int pvpanic_pci_probe(struct pci_dev *pdev,
>  	void __iomem *base;
>  	int ret;
>  
> -	ret = pci_enable_device(pdev);
> +	ret = pcim_enable_device(pdev);
>  	if (ret < 0)
>  		return ret;
>  
> -	base = pci_iomap(pdev, 0, 0);
> +	base = pcim_iomap(pdev, 0, 0);
>  	if (!base)
>  		return -ENOMEM;
>  
> -	pi = kmalloc(sizeof(*pi), GFP_ATOMIC);
> +	pi = devm_kmalloc(&pdev->dev, sizeof(*pi), GFP_ATOMIC);
>  	if (!pi)
>  		return -ENOMEM;
>  
> @@ -107,9 +107,6 @@ static void pvpanic_pci_remove(struct pci_dev *pdev)
>  	struct pvpanic_instance *pi = dev_get_drvdata(&pdev->dev);
>  
>  	pvpanic_remove(pi);
> -	iounmap(pi->base);
> -	kfree(pi);
> -	pci_disable_device(pdev);
>  }
>  
>  static struct pci_driver pvpanic_pci_driver = {
> -- 
> 2.30.2
> 

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Christophe JAILLET]

Le 17/05/2021 à 10:01, Andy Shevchenko a écrit :
> On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
>> There is no error handling path in the probe function.
>> Switch to managed resource so that errors in the probe are handled easily
>> and simplify the remove function accordingly.
> 
> Yes, that's what I suggested earlier to another contributor.
> 
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> 
> Thanks!
> 
> P.S. You may consider the following things as well:
>   1) converting to use pci_set_drvdata() / pci_get_drvdata()

I can send a patch for that if you want.
But it looks really low value for a driver that is already very short 
and clean.

>   2) providing devm_pvpanic_probe() [via devm_add_action() /
>      devm_add_action_or_reset()]

I don't follow you here.
The goal would be to avoid the remove function and "record" the needed 
action directly in the probe?

If this is it, I would only see an unusual pattern and a harder to 
follow logic.

Did I miss something?
What would be the benefit?

CJ

> 
>> Fixes: db3a4f0abefd ("misc/pvpanic: add PCI driver")
>> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
>> ---
>>   drivers/misc/pvpanic/pvpanic-pci.c | 9 +++------
>>   1 file changed, 3 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/misc/pvpanic/pvpanic-pci.c b/drivers/misc/pvpanic/pvpanic-pci.c
>> index 9ecc4e8559d5..046ce4ecc195 100644
>> --- a/drivers/misc/pvpanic/pvpanic-pci.c
>> +++ b/drivers/misc/pvpanic/pvpanic-pci.c
>> @@ -78,15 +78,15 @@ static int pvpanic_pci_probe(struct pci_dev *pdev,
>>   	void __iomem *base;
>>   	int ret;
>>   
>> -	ret = pci_enable_device(pdev);
>> +	ret = pcim_enable_device(pdev);
>>   	if (ret < 0)
>>   		return ret;
>>   
>> -	base = pci_iomap(pdev, 0, 0);
>> +	base = pcim_iomap(pdev, 0, 0);
>>   	if (!base)
>>   		return -ENOMEM;
>>   
>> -	pi = kmalloc(sizeof(*pi), GFP_ATOMIC);
>> +	pi = devm_kmalloc(&pdev->dev, sizeof(*pi), GFP_ATOMIC);
>>   	if (!pi)
>>   		return -ENOMEM;
>>   
>> @@ -107,9 +107,6 @@ static void pvpanic_pci_remove(struct pci_dev *pdev)
>>   	struct pvpanic_instance *pi = dev_get_drvdata(&pdev->dev);
>>   
>>   	pvpanic_remove(pi);
>> -	iounmap(pi->base);
>> -	kfree(pi);
>> -	pci_disable_device(pdev);
>>   }
>>   
>>   static struct pci_driver pvpanic_pci_driver = {
>> -- 
>> 2.30.2
>>
> 

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Andy Shevchenko]

On Mon, May 17, 2021 at 12:02:24PM +0200, Christophe JAILLET wrote:
> Le 17/05/2021 à 10:01, Andy Shevchenko a écrit :
> > On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
> > > There is no error handling path in the probe function.
> > > Switch to managed resource so that errors in the probe are handled easily
> > > and simplify the remove function accordingly.
> > 
> > Yes, that's what I suggested earlier to another contributor.
> > 
> > Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > 
> > Thanks!
> > 
> > P.S. You may consider the following things as well:
> >   1) converting to use pci_set_drvdata() / pci_get_drvdata()
> 
> I can send a patch for that if you want.
> But it looks really low value for a driver that is already very short and
> clean.

Yep, that's why 2) below came to my mind (then you will remove drvdata call).

> >   2) providing devm_pvpanic_probe() [via devm_add_action() /
> >      devm_add_action_or_reset()]
> 
> I don't follow you here.
> The goal would be to avoid the remove function and "record" the needed
> action directly in the probe?
> 
> If this is it, I would only see an unusual pattern and a harder to follow
> logic.

> Did I miss something?
> What would be the benefit?

First of all it's a usual pattern when one, often used in ->probe(), function
gains its devm variant. See, for example, `return devm_gpiochip_add_data(...);`
used in the code.

Benefit is to have everything under managed resources and yes, no ->remove()
will be needed in the individual drivers.

But it's up to you. It was just a proposal that you may simply refuse to follow,
it's fine.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Christophe JAILLET]

Le 17/05/2021 à 12:34, Andy Shevchenko a écrit :
> On Mon, May 17, 2021 at 12:02:24PM +0200, Christophe JAILLET wrote:
>> Le 17/05/2021 à 10:01, Andy Shevchenko a écrit :
>>> On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
>>>> There is no error handling path in the probe function.
>>>> Switch to managed resource so that errors in the probe are handled easily
>>>> and simplify the remove function accordingly.
>>>
>>> Yes, that's what I suggested earlier to another contributor.
>>>
>>> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
>>>
>>> Thanks!
>>>
>>> P.S. You may consider the following things as well:
>>>    1) converting to use pci_set_drvdata() / pci_get_drvdata()
>>
>> I can send a patch for that if you want.
>> But it looks really low value for a driver that is already very short and
>> clean.
> 
> Yep, that's why 2) below came to my mind (then you will remove drvdata call).
> 
>>>    2) providing devm_pvpanic_probe() [via devm_add_action() /
>>>       devm_add_action_or_reset()]
>>
>> I don't follow you here.
>> The goal would be to avoid the remove function and "record" the needed
>> action directly in the probe?
>>
>> If this is it, I would only see an unusual pattern and a harder to follow
>> logic.
> 
>> Did I miss something?
>> What would be the benefit?
> 
> First of all it's a usual pattern when one, often used in ->probe(), function
> gains its devm variant. See, for example, `return devm_gpiochip_add_data(...);`
> used in the code.
> 
> Benefit is to have everything under managed resources and yes, no ->remove()
> will be needed in the individual drivers.
> 
> But it's up to you. It was just a proposal that you may simply refuse to follow,
> it's fine.
> 

Ok, I'll propose something when/if my first patches reach -next.

I now better see your point. I first read devm_pvpanic_pci_probe (i.e. 
with pci inside), instead of devm_pvpanic_probe.

CJ

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Greg KH]

On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
> There is no error handling path in the probe function.
> Switch to managed resource so that errors in the probe are handled easily
> and simplify the remove function accordingly.
> 
> Fixes: db3a4f0abefd ("misc/pvpanic: add PCI driver")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
>  drivers/misc/pvpanic/pvpanic-pci.c | 9 +++------
>  1 file changed, 3 insertions(+), 6 deletions(-)

I see two different series for these patches, so I don't know which to
take :(

Please fix up and send a v2 series so that I have a clue...

thanks,

greg k-h

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Christophe JAILLET]

Le 21/05/2021 à 14:36, Greg KH a écrit :
> On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
>> There is no error handling path in the probe function.
>> Switch to managed resource so that errors in the probe are handled easily
>> and simplify the remove function accordingly.
>>
>> Fixes: db3a4f0abefd ("misc/pvpanic: add PCI driver")
>> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
>> ---
>>   drivers/misc/pvpanic/pvpanic-pci.c | 9 +++------
>>   1 file changed, 3 insertions(+), 6 deletions(-)
> 
> I see two different series for these patches, so I don't know which to
> take :(
> 
> Please fix up and send a v2 series so that I have a clue...
> 
> thanks,
> 
> greg k-h
> 

Both have to be taken. One is for -pci.c and one is for -mmio.c.

I'll resend both with a more complete subject and will include Andy 
Shevchenko's comments.

CJ

Re: [PATCH 1/2] misc/pvpanic: Fix error handling in 'pvpanic_pci_probe()'

[Greg KH]

On Fri, May 21, 2021 at 02:41:16PM +0200, Christophe JAILLET wrote:
> Le 21/05/2021 à 14:36, Greg KH a écrit :
> > On Sun, May 16, 2021 at 04:36:55PM +0200, Christophe JAILLET wrote:
> > > There is no error handling path in the probe function.
> > > Switch to managed resource so that errors in the probe are handled easily
> > > and simplify the remove function accordingly.
> > > 
> > > Fixes: db3a4f0abefd ("misc/pvpanic: add PCI driver")
> > > Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> > > ---
> > >   drivers/misc/pvpanic/pvpanic-pci.c | 9 +++------
> > >   1 file changed, 3 insertions(+), 6 deletions(-)
> > 
> > I see two different series for these patches, so I don't know which to
> > take :(
> > 
> > Please fix up and send a v2 series so that I have a clue...
> > 
> > thanks,
> > 
> > greg k-h
> > 
> 
> Both have to be taken. One is for -pci.c and one is for -mmio.c.

Ah, I totally missed that :(

> I'll resend both with a more complete subject and will include Andy
> Shevchenko's comments.

Wonderful, thanks for doing that.

greg k-h