[PATCH v1 2/2] PCI: j721e: Add support for optional regulator supplies

Vitor Soares posted 2 patches 2 months ago
[PATCH v1 2/2] PCI: j721e: Add support for optional regulator supplies
Posted by Vitor Soares 2 months ago
From: Vitor Soares <vitor.soares@toradex.com>

Some boards require external regulators to power PCIe endpoints.
Add support for optional 1.5V, 3.3V, and 12V supplies, which may be
defined in the device tree as vpcie1v5-supply, vpcie3v3-supply, and
vpcie12v-supply.

Use devm_regulator_get_enable_optional() to obtain and enable each
supply, so it will be automatically disabled when the driver is
removed.

Signed-off-by: Vitor Soares <vitor.soares@toradex.com>
---
 drivers/pci/controller/cadence/pci-j721e.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/pci/controller/cadence/pci-j721e.c b/drivers/pci/controller/cadence/pci-j721e.c
index 5bc5ab20aa6d..f29ce2aef04e 100644
--- a/drivers/pci/controller/cadence/pci-j721e.c
+++ b/drivers/pci/controller/cadence/pci-j721e.c
@@ -21,6 +21,7 @@
 #include <linux/platform_device.h>
 #include <linux/pm_runtime.h>
 #include <linux/regmap.h>
+#include <linux/regulator/consumer.h>
 
 #include "../../pci.h"
 #include "pcie-cadence.h"
@@ -467,6 +468,10 @@ static const struct of_device_id of_j721e_pcie_match[] = {
 };
 MODULE_DEVICE_TABLE(of, of_j721e_pcie_match);
 
+static const char * const j721e_pcie_supplies[] = {
+	"vpcie12v", "vpcie3v3", "vpcie1v5"
+};
+
 static int j721e_pcie_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -480,6 +485,7 @@ static int j721e_pcie_probe(struct platform_device *pdev)
 	struct gpio_desc *gpiod;
 	void __iomem *base;
 	struct clk *clk;
+	unsigned int i;
 	u32 num_lanes;
 	u32 mode;
 	int ret;
@@ -565,6 +571,13 @@ static int j721e_pcie_probe(struct platform_device *pdev)
 	if (irq < 0)
 		return irq;
 
+	for (i = 0; i < ARRAY_SIZE(j721e_pcie_supplies); i++) {
+		ret = devm_regulator_get_enable_optional(dev, j721e_pcie_supplies[i]);
+		if (ret < 0 && ret != -ENODEV)
+			return dev_err_probe(dev, ret, "can't enable regulator %s\n",
+					     j721e_pcie_supplies[i]);
+	}
+
 	dev_set_drvdata(dev, pcie);
 	pm_runtime_enable(dev);
 	ret = pm_runtime_get_sync(dev);
-- 
2.51.0
Re: [PATCH v1 2/2] PCI: j721e: Add support for optional regulator supplies
Posted by Manivannan Sadhasivam 1 month, 4 weeks ago
On Tue, Oct 14, 2025 at 12:25:49PM +0100, Vitor Soares wrote:
> From: Vitor Soares <vitor.soares@toradex.com>
> 
> Some boards require external regulators to power PCIe endpoints.
> Add support for optional 1.5V, 3.3V, and 12V supplies, which may be
> defined in the device tree as vpcie1v5-supply, vpcie3v3-supply, and
> vpcie12v-supply.
> 
> Use devm_regulator_get_enable_optional() to obtain and enable each
> supply, so it will be automatically disabled when the driver is
> removed.
> 
> Signed-off-by: Vitor Soares <vitor.soares@toradex.com>
> ---
>  drivers/pci/controller/cadence/pci-j721e.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/pci/controller/cadence/pci-j721e.c b/drivers/pci/controller/cadence/pci-j721e.c
> index 5bc5ab20aa6d..f29ce2aef04e 100644
> --- a/drivers/pci/controller/cadence/pci-j721e.c
> +++ b/drivers/pci/controller/cadence/pci-j721e.c
> @@ -21,6 +21,7 @@
>  #include <linux/platform_device.h>
>  #include <linux/pm_runtime.h>
>  #include <linux/regmap.h>
> +#include <linux/regulator/consumer.h>
>  
>  #include "../../pci.h"
>  #include "pcie-cadence.h"
> @@ -467,6 +468,10 @@ static const struct of_device_id of_j721e_pcie_match[] = {
>  };
>  MODULE_DEVICE_TABLE(of, of_j721e_pcie_match);
>  
> +static const char * const j721e_pcie_supplies[] = {
> +	"vpcie12v", "vpcie3v3", "vpcie1v5"
> +};

Please don't hardcode the supplies in driver. The DT binding should make sure
the relevant supplies are passed (including the optional ones). Just use
of_regulator_bulk_get_all() to acquire all the passed supplies.

- Mani

-- 
மணிவண்ணன் சதாசிவம்
Re: [PATCH v1 2/2] PCI: j721e: Add support for optional regulator supplies
Posted by Vitor Soares 1 month, 3 weeks ago
Hello Mani,

Thank you for the feedback.

On Tue, 2025-10-21 at 07:36 +0530, Manivannan Sadhasivam wrote:
> On Tue, Oct 14, 2025 at 12:25:49PM +0100, Vitor Soares wrote:
> > From: Vitor Soares <vitor.soares@toradex.com>
> > 
> > Some boards require external regulators to power PCIe endpoints.
> > Add support for optional 1.5V, 3.3V, and 12V supplies, which may be
> > defined in the device tree as vpcie1v5-supply, vpcie3v3-supply, and
> > vpcie12v-supply.
> > 
> > Use devm_regulator_get_enable_optional() to obtain and enable each
> > supply, so it will be automatically disabled when the driver is
> > removed.
> > 
> > Signed-off-by: Vitor Soares <vitor.soares@toradex.com>
> > ---
> >  drivers/pci/controller/cadence/pci-j721e.c | 13 +++++++++++++
> >  1 file changed, 13 insertions(+)
> > 
> > diff --git a/drivers/pci/controller/cadence/pci-j721e.c
> > b/drivers/pci/controller/cadence/pci-j721e.c
> > index 5bc5ab20aa6d..f29ce2aef04e 100644
> > --- a/drivers/pci/controller/cadence/pci-j721e.c
> > +++ b/drivers/pci/controller/cadence/pci-j721e.c
> > @@ -21,6 +21,7 @@
> >  #include <linux/platform_device.h>
> >  #include <linux/pm_runtime.h>
> >  #include <linux/regmap.h>
> > +#include <linux/regulator/consumer.h>
> >  
> >  #include "../../pci.h"
> >  #include "pcie-cadence.h"
> > @@ -467,6 +468,10 @@ static const struct of_device_id of_j721e_pcie_match[]
> > = {
> >  };
> >  MODULE_DEVICE_TABLE(of, of_j721e_pcie_match);
> >  
> > +static const char * const j721e_pcie_supplies[] = {
> > +       "vpcie12v", "vpcie3v3", "vpcie1v5"
> > +};
> 
> Please don't hardcode the supplies in driver. The DT binding should make sure
> the relevant supplies are passed (including the optional ones). Just use
> of_regulator_bulk_get_all() to acquire all the passed supplies.
> 
> - Mani
> 

I checked the bulk regulator APIs as suggested and of_regulator_bulk_get_all()
does handle optional supplies correctly, however it is not a managed function
and doesn't enable the  regulators automatically.

To use of_regulator_bulk_get_all(), I would need to:
- Manually enable regulators with regulator_bulk_enable()
- Add cleanup/disable logic in remove path
- Handle error cleanup path manually

This would actually make the code more complex and error-prone compared to the
current approach using devm_regulator_get_enable_optional(), which provides
managed cleanup and automatic enable for optional supplies.

I also checked devm_regulator_bulk_get_enable(), it treats all supplies as
required and needs the supplies name as well.

Unless there is a devm_regulator_bulk_get_optional_enable() API I'm not aware
of, the current per-supply approach is the standard kernel pattern for this use
case. Would you still prefer the bulk approach despite these limitations?

Best regards,
Vitor Soares
Re: [PATCH v1 2/2] PCI: j721e: Add support for optional regulator supplies
Posted by Manivannan Sadhasivam 1 month, 3 weeks ago
On Mon, Oct 27, 2025 at 08:24:12PM +0000, Vitor Soares wrote:
> Hello Mani,
> 
> Thank you for the feedback.
> 
> On Tue, 2025-10-21 at 07:36 +0530, Manivannan Sadhasivam wrote:
> > On Tue, Oct 14, 2025 at 12:25:49PM +0100, Vitor Soares wrote:
> > > From: Vitor Soares <vitor.soares@toradex.com>
> > > 
> > > Some boards require external regulators to power PCIe endpoints.
> > > Add support for optional 1.5V, 3.3V, and 12V supplies, which may be
> > > defined in the device tree as vpcie1v5-supply, vpcie3v3-supply, and
> > > vpcie12v-supply.
> > > 
> > > Use devm_regulator_get_enable_optional() to obtain and enable each
> > > supply, so it will be automatically disabled when the driver is
> > > removed.
> > > 
> > > Signed-off-by: Vitor Soares <vitor.soares@toradex.com>
> > > ---
> > >  drivers/pci/controller/cadence/pci-j721e.c | 13 +++++++++++++
> > >  1 file changed, 13 insertions(+)
> > > 
> > > diff --git a/drivers/pci/controller/cadence/pci-j721e.c
> > > b/drivers/pci/controller/cadence/pci-j721e.c
> > > index 5bc5ab20aa6d..f29ce2aef04e 100644
> > > --- a/drivers/pci/controller/cadence/pci-j721e.c
> > > +++ b/drivers/pci/controller/cadence/pci-j721e.c
> > > @@ -21,6 +21,7 @@
> > >  #include <linux/platform_device.h>
> > >  #include <linux/pm_runtime.h>
> > >  #include <linux/regmap.h>
> > > +#include <linux/regulator/consumer.h>
> > >  
> > >  #include "../../pci.h"
> > >  #include "pcie-cadence.h"
> > > @@ -467,6 +468,10 @@ static const struct of_device_id of_j721e_pcie_match[]
> > > = {
> > >  };
> > >  MODULE_DEVICE_TABLE(of, of_j721e_pcie_match);
> > >  
> > > +static const char * const j721e_pcie_supplies[] = {
> > > +       "vpcie12v", "vpcie3v3", "vpcie1v5"
> > > +};
> > 
> > Please don't hardcode the supplies in driver. The DT binding should make sure
> > the relevant supplies are passed (including the optional ones). Just use
> > of_regulator_bulk_get_all() to acquire all the passed supplies.
> > 
> > - Mani
> > 
> 
> I checked the bulk regulator APIs as suggested and of_regulator_bulk_get_all()
> does handle optional supplies correctly, however it is not a managed function
> and doesn't enable the  regulators automatically.
> 
> To use of_regulator_bulk_get_all(), I would need to:
> - Manually enable regulators with regulator_bulk_enable()
> - Add cleanup/disable logic in remove path
> - Handle error cleanup path manually
> 
> This would actually make the code more complex and error-prone compared to the
> current approach using devm_regulator_get_enable_optional(), which provides
> managed cleanup and automatic enable for optional supplies.
> 
> I also checked devm_regulator_bulk_get_enable(), it treats all supplies as
> required and needs the supplies name as well.
> 
> Unless there is a devm_regulator_bulk_get_optional_enable() API I'm not aware
> of, the current per-supply approach is the standard kernel pattern for this use
> case. Would you still prefer the bulk approach despite these limitations?
> 

Fine then. If you do not have a reason to manualy turn off/on the supplies
(during suspend/resume), then it might be better to use
devm_regulator_get_enable_optional() for now.

I'd love to add the managed version of of_regulator_bulk_get_all(), but I guess
Mark wouldn't want it.

- Mani

-- 
மணிவண்ணன் சதாசிவம்