[PATCH v1 0/3] DesignWare PWM improvements

[PATCH v1 0/3] DesignWare PWM improvements

[Raag Jadav]

This series adds 16 channel support for Intel Elkhart Lake and simplifies
code using standard helpers in DesignWare PWM driver.

Raag Jadav (3):
  pwm: dwc: Add 16 channel support for Intel Elkhart Lake
  pwm: dwc: simplify error handling
  pwm: dwc: use to_pci_dev() helper

 drivers/pwm/pwm-dwc.c | 71 +++++++++++++++++++++++++++++--------------
 drivers/pwm/pwm-dwc.h |  5 +++
 2 files changed, 54 insertions(+), 22 deletions(-)


base-commit: 6613476e225e090cc9aad49be7fa504e290dd33d
-- 
2.35.3

[PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Raag Jadav]

Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
function with 8 channels each. Add support for the remaining channels.

Signed-off-by: Raag Jadav <raag.jadav@intel.com>
Tested-by: Lakshmi Sowjanya D <lakshmi.sowjanya.d@intel.com>
---
 drivers/pwm/pwm-dwc.c | 51 +++++++++++++++++++++++++++++++++++--------
 drivers/pwm/pwm-dwc.h |  5 +++++
 2 files changed, 47 insertions(+), 9 deletions(-)

diff --git a/drivers/pwm/pwm-dwc.c b/drivers/pwm/pwm-dwc.c
index 4929354f8cd9..fd64313cb38d 100644
--- a/drivers/pwm/pwm-dwc.c
+++ b/drivers/pwm/pwm-dwc.c
@@ -25,16 +25,48 @@
 
 #include "pwm-dwc.h"
 
+/* Elkhart Lake */
+static const struct dwc_pwm_info ehl_pwm_info = {
+	.nr = 2,
+	.size = 0x1000,
+};
+
+static int dwc_pwm_init(struct device *dev, const struct dwc_pwm_info *info, void __iomem *base)
+{
+	/* Default values for single instance devices */
+	unsigned int nr = 1, size = 0;
+	int i, ret;
+
+	/* Fill up values from driver_data, if any */
+	if (info) {
+		nr = info->nr;
+		size = info->size;
+	}
+
+	for (i = 0; i < nr; i++) {
+		struct dwc_pwm *dwc;
+
+		dwc = dwc_pwm_alloc(dev);
+		if (!dwc)
+			return -ENOMEM;
+
+		dwc->base = base + (i * size);
+
+		ret = devm_pwmchip_add(dev, &dwc->chip);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
 static int dwc_pwm_probe(struct pci_dev *pci, const struct pci_device_id *id)
 {
+	const struct dwc_pwm_info *info;
 	struct device *dev = &pci->dev;
-	struct dwc_pwm *dwc;
+	void __iomem *base;
 	int ret;
 
-	dwc = dwc_pwm_alloc(dev);
-	if (!dwc)
-		return -ENOMEM;
-
 	ret = pcim_enable_device(pci);
 	if (ret) {
 		dev_err(dev, "Failed to enable device (%pe)\n", ERR_PTR(ret));
@@ -49,13 +81,14 @@ static int dwc_pwm_probe(struct pci_dev *pci, const struct pci_device_id *id)
 		return ret;
 	}
 
-	dwc->base = pcim_iomap_table(pci)[0];
-	if (!dwc->base) {
+	base = pcim_iomap_table(pci)[0];
+	if (!base) {
 		dev_err(dev, "Base address missing\n");
 		return -ENOMEM;
 	}
 
-	ret = devm_pwmchip_add(dev, &dwc->chip);
+	info = (const struct dwc_pwm_info *)id->driver_data;
+	ret = dwc_pwm_init(dev, info, base);
 	if (ret)
 		return ret;
 
@@ -109,7 +142,7 @@ static int dwc_pwm_resume(struct device *dev)
 static DEFINE_SIMPLE_DEV_PM_OPS(dwc_pwm_pm_ops, dwc_pwm_suspend, dwc_pwm_resume);
 
 static const struct pci_device_id dwc_pwm_id_table[] = {
-	{ PCI_VDEVICE(INTEL, 0x4bb7) }, /* Elkhart Lake */
+	{ PCI_VDEVICE(INTEL, 0x4bb7), (kernel_ulong_t)&ehl_pwm_info },
 	{  }	/* Terminating Entry */
 };
 MODULE_DEVICE_TABLE(pci, dwc_pwm_id_table);
diff --git a/drivers/pwm/pwm-dwc.h b/drivers/pwm/pwm-dwc.h
index 64795247c54c..c9bbfc77b568 100644
--- a/drivers/pwm/pwm-dwc.h
+++ b/drivers/pwm/pwm-dwc.h
@@ -33,6 +33,11 @@ MODULE_IMPORT_NS(dwc_pwm);
 #define DWC_TIM_CTRL_INT_MASK	BIT(2)
 #define DWC_TIM_CTRL_PWM	BIT(3)
 
+struct dwc_pwm_info {
+	unsigned int nr;
+	unsigned int size;
+};
+
 struct dwc_pwm_ctx {
 	u32 cnt;
 	u32 cnt2;
-- 
2.35.3

[PATCH v1 2/3] pwm: dwc: simplify error handling

[Raag Jadav]

Simplify error handling in ->probe() function using dev_err_probe() helper.

Signed-off-by: Raag Jadav <raag.jadav@intel.com>
---
 drivers/pwm/pwm-dwc.c | 18 ++++++------------
 1 file changed, 6 insertions(+), 12 deletions(-)

diff --git a/drivers/pwm/pwm-dwc.c b/drivers/pwm/pwm-dwc.c
index fd64313cb38d..8f8035b047c1 100644
--- a/drivers/pwm/pwm-dwc.c
+++ b/drivers/pwm/pwm-dwc.c
@@ -68,24 +68,18 @@ static int dwc_pwm_probe(struct pci_dev *pci, const struct pci_device_id *id)
 	int ret;
 
 	ret = pcim_enable_device(pci);
-	if (ret) {
-		dev_err(dev, "Failed to enable device (%pe)\n", ERR_PTR(ret));
-		return ret;
-	}
+	if (ret)
+		return dev_err_probe(dev, ret, "Failed to enable device (%pe)\n", ERR_PTR(ret));
 
 	pci_set_master(pci);
 
 	ret = pcim_iomap_regions(pci, BIT(0), pci_name(pci));
-	if (ret) {
-		dev_err(dev, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
-		return ret;
-	}
+	if (ret)
+		return dev_err_probe(dev, ret, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
 
 	base = pcim_iomap_table(pci)[0];
-	if (!base) {
-		dev_err(dev, "Base address missing\n");
-		return -ENOMEM;
-	}
+	if (!base)
+		return dev_err_probe(dev, -ENOMEM, "Base address missing\n");
 
 	info = (const struct dwc_pwm_info *)id->driver_data;
 	ret = dwc_pwm_init(dev, info, base);
-- 
2.35.3

[PATCH v1 3/3] pwm: dwc: use to_pci_dev() helper

[Raag Jadav]

Use to_pci_dev() helper to get pci device reference.

Signed-off-by: Raag Jadav <raag.jadav@intel.com>
---
 drivers/pwm/pwm-dwc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/pwm/pwm-dwc.c b/drivers/pwm/pwm-dwc.c
index 8f8035b047c1..f5fcea4e69e2 100644
--- a/drivers/pwm/pwm-dwc.c
+++ b/drivers/pwm/pwm-dwc.c
@@ -100,7 +100,7 @@ static void dwc_pwm_remove(struct pci_dev *pci)
 
 static int dwc_pwm_suspend(struct device *dev)
 {
-	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
+	struct pci_dev *pdev = to_pci_dev(dev);
 	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
 	int i;
 
@@ -120,7 +120,7 @@ static int dwc_pwm_suspend(struct device *dev)
 
 static int dwc_pwm_resume(struct device *dev)
 {
-	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
+	struct pci_dev *pdev = to_pci_dev(dev);
 	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
 	int i;
 
-- 
2.35.3

Re: [PATCH v1 0/3] DesignWare PWM improvements

[Jarkko Nikula]

On 1/22/24 05:02, Raag Jadav wrote:
> This series adds 16 channel support for Intel Elkhart Lake and simplifies
> code using standard helpers in DesignWare PWM driver.
> 
> Raag Jadav (3):
>    pwm: dwc: Add 16 channel support for Intel Elkhart Lake
>    pwm: dwc: simplify error handling
>    pwm: dwc: use to_pci_dev() helper
> 
>   drivers/pwm/pwm-dwc.c | 71 +++++++++++++++++++++++++++++--------------
>   drivers/pwm/pwm-dwc.h |  5 +++
>   2 files changed, 54 insertions(+), 22 deletions(-)
> 
For the patchset:

Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>

Re: [PATCH v1 3/3] pwm: dwc: use to_pci_dev() helper

[Andy Shevchenko]

On Mon, Jan 22, 2024 at 08:32:38AM +0530, Raag Jadav wrote:
> Use to_pci_dev() helper to get pci device reference.

PCI

...

>  static int dwc_pwm_suspend(struct device *dev)
>  {
> -	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> +	struct pci_dev *pdev = to_pci_dev(dev);
>  	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
>  	int i;
>  
> @@ -120,7 +120,7 @@ static int dwc_pwm_suspend(struct device *dev)
>  
>  static int dwc_pwm_resume(struct device *dev)
>  {
> -	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> +	struct pci_dev *pdev = to_pci_dev(dev);
>  	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
>  	int i;

I don't see how pdev is being used. That said, why dev_get_drvdata() is not
suffice?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 2/3] pwm: dwc: simplify error handling

[Andy Shevchenko]

On Mon, Jan 22, 2024 at 08:32:37AM +0530, Raag Jadav wrote:
> Simplify error handling in ->probe() function using dev_err_probe() helper.

...

>  	ret = pcim_iomap_regions(pci, BIT(0), pci_name(pci));
> -	if (ret) {
> -		dev_err(dev, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> -		return ret;
> -	}
> +	if (ret)
> +		return dev_err_probe(dev, ret, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
>  
>  	base = pcim_iomap_table(pci)[0];

> -	if (!base) {
> -		dev_err(dev, "Base address missing\n");
> -		return -ENOMEM;
> -	}
> +	if (!base)
> +		return dev_err_probe(dev, -ENOMEM, "Base address missing\n");

This check is bogus. Just remove it completely.

The pcim_iomap_table() fails IFF pcim_iomap_regions() fails.
You have checked the latter already.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Andy Shevchenko]

On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
> Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
> function with 8 channels each. Add support for the remaining channels.

...

> +static int dwc_pwm_init(struct device *dev, const struct dwc_pwm_info *info, void __iomem *base)
> +{
> +	/* Default values for single instance devices */
> +	unsigned int nr = 1, size = 0;
> +	int i, ret;
> +
> +	/* Fill up values from driver_data, if any */
> +	if (info) {
> +		nr = info->nr;
> +		size = info->size;
> +	}
> +
> +	for (i = 0; i < nr; i++) {
> +		struct dwc_pwm *dwc;
> +
> +		dwc = dwc_pwm_alloc(dev);
> +		if (!dwc)
> +			return -ENOMEM;
> +
> +		dwc->base = base + (i * size);
> +
> +		ret = devm_pwmchip_add(dev, &dwc->chip);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	return 0;
> +}

Why not doing this slightly differently?

First option: Always provide driver data (info is never NULL).

Second option, have the body of the for-loop be factored to a helper
dwc_pwm_init_one() and here

	if (!info)
		return dwc_pwm_init_one(..., 1, 0);

	for (i = 0; i < info->nr; i++) {
		ret = dwc_pwm_init_one(...);
		if (ret)
			return ret;
	}

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 3/3] pwm: dwc: use to_pci_dev() helper

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 1571 bytes --]

Hello,

On Sun, Jan 28, 2024 at 04:46:48PM +0200, Andy Shevchenko wrote:
> On Mon, Jan 22, 2024 at 08:32:38AM +0530, Raag Jadav wrote:
> > Use to_pci_dev() helper to get pci device reference.
> 
> PCI
> 
> ...
> 
> >  static int dwc_pwm_suspend(struct device *dev)
> >  {
> > -	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> > +	struct pci_dev *pdev = to_pci_dev(dev);
> >  	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
> >  	int i;
> >  
> > @@ -120,7 +120,7 @@ static int dwc_pwm_suspend(struct device *dev)
> >  
> >  static int dwc_pwm_resume(struct device *dev)
> >  {
> > -	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> > +	struct pci_dev *pdev = to_pci_dev(dev);
> >  	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
> >  	int i;
> 
> I don't see how pdev is being used. That said, why dev_get_drvdata() is not
> suffice?

I would even consider using dev_get_drvdata() a nice cleanup given that
pci_get_drvdata() works because dwc_pwm_alloc() called dev_set_drvdata()
(and not pci_set_drvdata()).

Not so long ago (i.e. before commit a357d1493f0c ("pwm: dwc: Move memory
allocation to own function")) the dwc driver was pci only and used
pci_set_drvdata(). Then the upside of first converting the struct device
to a struct pci_dev was not to hard code knowledge about the
implementation of pci_[gs]et_drvdata().

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v1 2/3] pwm: dwc: simplify error handling

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 1094 bytes --]

Hello,

On Sun, Jan 28, 2024 at 04:48:16PM +0200, Andy Shevchenko wrote:
> On Mon, Jan 22, 2024 at 08:32:37AM +0530, Raag Jadav wrote:
> > Simplify error handling in ->probe() function using dev_err_probe() helper.
> 
> ...
> 
> >  	ret = pcim_iomap_regions(pci, BIT(0), pci_name(pci));
> > -	if (ret) {
> > -		dev_err(dev, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> > -		return ret;
> > -	}
> > +	if (ret)
> > +		return dev_err_probe(dev, ret, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> >  
> >  	base = pcim_iomap_table(pci)[0];
> 
> > -	if (!base) {
> > -		dev_err(dev, "Base address missing\n");
> > -		return -ENOMEM;
> > -	}
> > +	if (!base)
> > +		return dev_err_probe(dev, -ENOMEM, "Base address missing\n");
> 
> This check is bogus. Just remove it completely.

This would be a separate patch though. IMHO mechanically converting to
dev_err_probe() is fine.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v1 2/3] pwm: dwc: simplify error handling

[Raag Jadav]

On Sun, Jan 28, 2024 at 04:48:16PM +0200, Andy Shevchenko wrote:
> On Mon, Jan 22, 2024 at 08:32:37AM +0530, Raag Jadav wrote:
> > Simplify error handling in ->probe() function using dev_err_probe() helper.
> 
> ...
> 
> >  	ret = pcim_iomap_regions(pci, BIT(0), pci_name(pci));
> > -	if (ret) {
> > -		dev_err(dev, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> > -		return ret;
> > -	}
> > +	if (ret)
> > +		return dev_err_probe(dev, ret, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> >  
> >  	base = pcim_iomap_table(pci)[0];
> 
> > -	if (!base) {
> > -		dev_err(dev, "Base address missing\n");
> > -		return -ENOMEM;
> > -	}
> > +	if (!base)
> > +		return dev_err_probe(dev, -ENOMEM, "Base address missing\n");
> 
> This check is bogus. Just remove it completely.
> 
> The pcim_iomap_table() fails IFF pcim_iomap_regions() fails.
> You have checked the latter already.

I'm no expert on devres but I found a few NULL returns in alloc_dr()
call path. In the interest of learning more about iomap, wouldn't we
need to handle them (just in some odd case)?

Raag

Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Raag Jadav]

On Sun, Jan 28, 2024 at 04:53:24PM +0200, Andy Shevchenko wrote:
> On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
> > Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
> > function with 8 channels each. Add support for the remaining channels.
> 
> ...
> 
> > +static int dwc_pwm_init(struct device *dev, const struct dwc_pwm_info *info, void __iomem *base)
> > +{
> > +	/* Default values for single instance devices */
> > +	unsigned int nr = 1, size = 0;
> > +	int i, ret;
> > +
> > +	/* Fill up values from driver_data, if any */
> > +	if (info) {
> > +		nr = info->nr;
> > +		size = info->size;
> > +	}
> > +
> > +	for (i = 0; i < nr; i++) {
> > +		struct dwc_pwm *dwc;
> > +
> > +		dwc = dwc_pwm_alloc(dev);
> > +		if (!dwc)
> > +			return -ENOMEM;
> > +
> > +		dwc->base = base + (i * size);
> > +
> > +		ret = devm_pwmchip_add(dev, &dwc->chip);
> > +		if (ret)
> > +			return ret;
> > +	}
> > +
> > +	return 0;
> > +}
> 
> Why not doing this slightly differently?
> 
> First option: Always provide driver data (info is never NULL).

Allowing empty driver_data would save us from adding dummy info
for single instance devices in the future.

> Second option, have the body of the for-loop be factored to a helper
> dwc_pwm_init_one() and here
> 
> 	if (!info)
> 		return dwc_pwm_init_one(..., 1, 0);
> 
> 	for (i = 0; i < info->nr; i++) {
> 		ret = dwc_pwm_init_one(...);
> 		if (ret)
> 			return ret;
> 	}

Considering above, we're looking at something like this.

static int dwc_pwm_init_one(struct device *dev, void __iomem *base, unsigned int size)
{
        struct dwc_pwm *dwc;

        dwc = dwc_pwm_alloc(dev);
        if (!dwc)
                return -ENOMEM;

        dwc->base = base + size;

        return devm_pwmchip_add(dev, &dwc->chip);
}

	...

        if (info) {
                for (i = 0; i < info->nr; i++) {
                        ret = dwc_pwm_init_one(dev, base, i * info->size);
                        if (ret)
                                return ret;
                }
        } else {
                ret = dwc_pwm_init_one(dev, base, 0);
                if (ret)
                        return ret;
        }
	...

Raag

Re: [PATCH v1 3/3] pwm: dwc: use to_pci_dev() helper

[Raag Jadav]

On Sun, Jan 28, 2024 at 05:55:00PM +0100, Uwe Kleine-König wrote:
> Hello,
> 
> On Sun, Jan 28, 2024 at 04:46:48PM +0200, Andy Shevchenko wrote:
> > On Mon, Jan 22, 2024 at 08:32:38AM +0530, Raag Jadav wrote:
> > > Use to_pci_dev() helper to get pci device reference.
> > 
> > PCI
> > 
> > ...
> > 
> > >  static int dwc_pwm_suspend(struct device *dev)
> > >  {
> > > -	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> > > +	struct pci_dev *pdev = to_pci_dev(dev);
> > >  	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
> > >  	int i;
> > >  
> > > @@ -120,7 +120,7 @@ static int dwc_pwm_suspend(struct device *dev)
> > >  
> > >  static int dwc_pwm_resume(struct device *dev)
> > >  {
> > > -	struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> > > +	struct pci_dev *pdev = to_pci_dev(dev);
> > >  	struct dwc_pwm *dwc = pci_get_drvdata(pdev);
> > >  	int i;
> > 
> > I don't see how pdev is being used. That said, why dev_get_drvdata() is not
> > suffice?
> 
> I would even consider using dev_get_drvdata() a nice cleanup given that
> pci_get_drvdata() works because dwc_pwm_alloc() called dev_set_drvdata()
> (and not pci_set_drvdata()).

Makes sense. Will update in v2.

Raag

Re: [PATCH v1 2/3] pwm: dwc: simplify error handling

[Andy Shevchenko]

On Sun, Jan 28, 2024 at 05:58:38PM +0100, Uwe Kleine-König wrote:
> On Sun, Jan 28, 2024 at 04:48:16PM +0200, Andy Shevchenko wrote:
> > On Mon, Jan 22, 2024 at 08:32:37AM +0530, Raag Jadav wrote:
> > > Simplify error handling in ->probe() function using dev_err_probe() helper.

...

> > >  	ret = pcim_iomap_regions(pci, BIT(0), pci_name(pci));
> > > -	if (ret) {
> > > -		dev_err(dev, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> > > -		return ret;
> > > -	}
> > > +	if (ret)
> > > +		return dev_err_probe(dev, ret, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> > >  
> > >  	base = pcim_iomap_table(pci)[0];
> > 
> > > -	if (!base) {
> > > -		dev_err(dev, "Base address missing\n");
> > > -		return -ENOMEM;
> > > -	}
> > > +	if (!base)
> > > +		return dev_err_probe(dev, -ENOMEM, "Base address missing\n");
> > 
> > This check is bogus. Just remove it completely.
> 
> This would be a separate patch though. IMHO mechanically converting to
> dev_err_probe() is fine.

Sure, that's what I meant. First patch to remove, followed by dev_err_probe()
conversion.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 2/3] pwm: dwc: simplify error handling

[Andy Shevchenko]

On Tue, Jan 30, 2024 at 10:31:02AM +0200, Raag Jadav wrote:
> On Sun, Jan 28, 2024 at 04:48:16PM +0200, Andy Shevchenko wrote:
> > On Mon, Jan 22, 2024 at 08:32:37AM +0530, Raag Jadav wrote:
> > > Simplify error handling in ->probe() function using dev_err_probe() helper.

...

> > >  	ret = pcim_iomap_regions(pci, BIT(0), pci_name(pci));
> > > -	if (ret) {
> > > -		dev_err(dev, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> > > -		return ret;
> > > -	}
> > > +	if (ret)
> > > +		return dev_err_probe(dev, ret, "Failed to iomap PCI BAR (%pe)\n", ERR_PTR(ret));
> > >  
> > >  	base = pcim_iomap_table(pci)[0];
> > 
> > > -	if (!base) {
> > > -		dev_err(dev, "Base address missing\n");
> > > -		return -ENOMEM;
> > > -	}
> > > +	if (!base)
> > > +		return dev_err_probe(dev, -ENOMEM, "Base address missing\n");
> > 
> > This check is bogus. Just remove it completely.
> > 
> > The pcim_iomap_table() fails IFF pcim_iomap_regions() fails.
> > You have checked the latter already.
> 
> I'm no expert on devres but I found a few NULL returns in alloc_dr()
> call path. In the interest of learning more about iomap, wouldn't we
> need to handle them (just in some odd case)?

It has nothing to do with devres. The logic is implemented in PCI core.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Andy Shevchenko]

On Tue, Jan 30, 2024 at 12:30:23PM +0200, Raag Jadav wrote:
> On Sun, Jan 28, 2024 at 04:53:24PM +0200, Andy Shevchenko wrote:
> > On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
> > > Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
> > > function with 8 channels each. Add support for the remaining channels.

...

> > First option: Always provide driver data (info is never NULL).
> 
> Allowing empty driver_data would save us from adding dummy info
> for single instance devices in the future.

Which may be too premature "optimisation". Why? Because if we ever have
something like pci_dev_get_match_data(), the empty will mean NULL, and
we may not get difference between empty and missing one.

> > Second option, have the body of the for-loop be factored to a helper
> > dwc_pwm_init_one() and here
> > 
> > 	if (!info)
> > 		return dwc_pwm_init_one(..., 1, 0);
> > 
> > 	for (i = 0; i < info->nr; i++) {
> > 		ret = dwc_pwm_init_one(...);
> > 		if (ret)
> > 			return ret;
> > 	}
> 
> Considering above, we're looking at something like this.

As one option, yes.

> static int dwc_pwm_init_one(struct device *dev, void __iomem *base, unsigned int size)
> {
>         struct dwc_pwm *dwc;
> 
>         dwc = dwc_pwm_alloc(dev);
>         if (!dwc)
>                 return -ENOMEM;
> 
>         dwc->base = base + size;
> 
>         return devm_pwmchip_add(dev, &dwc->chip);
> }

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Raag Jadav]

On Thu, Feb 01, 2024 at 01:26:53PM +0200, Andy Shevchenko wrote:
> On Tue, Jan 30, 2024 at 12:30:23PM +0200, Raag Jadav wrote:
> > On Sun, Jan 28, 2024 at 04:53:24PM +0200, Andy Shevchenko wrote:
> > > On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
> > > > Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
> > > > function with 8 channels each. Add support for the remaining channels.
> 
> ...
> 
> > > First option: Always provide driver data (info is never NULL).
> > 
> > Allowing empty driver_data would save us from adding dummy info
> > for single instance devices in the future.
> 
> Which may be too premature "optimisation". Why? Because if we ever have
> something like pci_dev_get_match_data(), the empty will mean NULL, and
> we may not get difference between empty and missing one.

Not sure if I'm able to find such a helper as of now, but fair.
I can change it in v2 if Jarkko is okay with it.

Raag

Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Raag Jadav]

On Fri, Feb 02, 2024 at 06:02:46AM +0200, Raag Jadav wrote:
> On Thu, Feb 01, 2024 at 01:26:53PM +0200, Andy Shevchenko wrote:
> > On Tue, Jan 30, 2024 at 12:30:23PM +0200, Raag Jadav wrote:
> > > On Sun, Jan 28, 2024 at 04:53:24PM +0200, Andy Shevchenko wrote:
> > > > On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
> > > > > Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
> > > > > function with 8 channels each. Add support for the remaining channels.
> > 
> > ...
> > 
> > > > First option: Always provide driver data (info is never NULL).
> > > 
> > > Allowing empty driver_data would save us from adding dummy info
> > > for single instance devices in the future.
> > 
> > Which may be too premature "optimisation". Why? Because if we ever have
> > something like pci_dev_get_match_data(), the empty will mean NULL, and
> > we may not get difference between empty and missing one.
> 
> Not sure if I'm able to find such a helper as of now, but fair.
> I can change it in v2 if Jarkko is okay with it.

Hi Jarkko,

If you agree with Andy's comments, please let me know.
Will send out a v2 accordingly.

Raag

Re: [PATCH v1 1/3] pwm: dwc: Add 16 channel support for Intel Elkhart Lake

[Jarkko Nikula]

On 2/7/24 13:48, Raag Jadav wrote:
> On Fri, Feb 02, 2024 at 06:02:46AM +0200, Raag Jadav wrote:
>> On Thu, Feb 01, 2024 at 01:26:53PM +0200, Andy Shevchenko wrote:
>>> On Tue, Jan 30, 2024 at 12:30:23PM +0200, Raag Jadav wrote:
>>>> On Sun, Jan 28, 2024 at 04:53:24PM +0200, Andy Shevchenko wrote:
>>>>> On Mon, Jan 22, 2024 at 08:32:36AM +0530, Raag Jadav wrote:
>>>>>> Intel Elkhart Lake PSE includes two instances of PWM as a single PCI
>>>>>> function with 8 channels each. Add support for the remaining channels.
>>>
>>> ...
>>>
>>>>> First option: Always provide driver data (info is never NULL).
>>>>
>>>> Allowing empty driver_data would save us from adding dummy info
>>>> for single instance devices in the future.
>>>
>>> Which may be too premature "optimisation". Why? Because if we ever have
>>> something like pci_dev_get_match_data(), the empty will mean NULL, and
>>> we may not get difference between empty and missing one.
>>
>> Not sure if I'm able to find such a helper as of now, but fair.
>> I can change it in v2 if Jarkko is okay with it.
> 
> Hi Jarkko,
> 
> If you agree with Andy's comments, please let me know.
> Will send out a v2 accordingly.
> 
Ah, sorry, I didn't have opinion.