Skip to contentSkip to conversation

pinctrl: bcm2835: Fix gpio hogs and pin reinitialisation

PATCH Superseded Phil Elwell 3 participants 9 messages Nov 29, 2021 – Dec 2, 2021 1 message not in the archive

Subsystems (5)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover pinctrl: bcm2835: Fix gpio hogs and pin reinitialisation

Tackle two problems with the pinctrl-bcm2835 driver and its Device Tree configuration:

1. The pinctrl-bcm2835 driver is a combined pinctrl/gpio driver. Currently the gpio side is registered first, but this breaks gpio hogs (which are configured during gpiochip_add_data).

2. Since [1], a "gpio-ranges" property is required in order for pins to be returned to inputs when freed. Note that without patch 1, the device never gets out of EPROBE_DEFER.

Note that the Fixes: tags are little more than hooks to hang the back-ports on - no blame is intended.

[1] commit 2ab73c6d8323 ("gpio: Support GPIO controllers without pin-ranges")

Phil Elwell (2):
pinctrl: bcm2835: Change init order for gpio hogs
ARM: dts: gpio-ranges property is now required

 arch/arm/boot/dts/bcm2711.dtsi        |  2 ++
 arch/arm/boot/dts/bcm283x.dtsi        |  2 ++
 drivers/pinctrl/bcm/pinctrl-bcm2835.c | 29 +++++++++++++++------------
 3 files changed, 20 insertions(+), 13 deletions(-)
Signature · 1 line
2.25.1
#1 lore raw

PATCH 1/2 pinctrl: bcm2835: Change init order for gpio hogs

...and gpio-ranges

pinctrl-bcm2835 is a combined pinctrl/gpio driver. Currently the gpio side is registered first, but this breaks gpio hogs (which are configured during gpiochip_add_data). Part of the hog initialisation is a call to pinctrl_gpio_request, and since the pinctrl driver hasn't yet been registered this results in an -EPROBE_DEFER from which it can never recover.

Change the initialisation sequence to register the pinctrl driver first.

This also solves a similar problem with the gpio-ranges property, which is required in order for released pins to be returned to inputs.

Fixes 73345a18d464b ("pinctrl: bcm2835: Pass irqchip when adding

gpiochip")

---
 drivers/pinctrl/bcm/pinctrl-bcm2835.c | 29 +++++++++++++++------------
 1 file changed, 16 insertions(+), 13 deletions(-)
drivers/pinctrl/bcm/pinctrl-bcm2835.c +16 -13
index 2abcc6ce4eba..b607d10e4cbd 100644
@@ -1244,6 +1244,18 @@ static int bcm2835_pinctrl_probe(struct platform_device *pdev)
 		raw_spin_lock_init(&pc->irq_lock[i]);
 	}
 
+	pc->pctl_desc = *pdata->pctl_desc;
+	pc->pctl_dev = devm_pinctrl_register(dev, &pc->pctl_desc, pc);
+	if (IS_ERR(pc->pctl_dev)) {
+		gpiochip_remove(&pc->gpio_chip);
+		return PTR_ERR(pc->pctl_dev);
+	}
+
+	pc->gpio_range = *pdata->gpio_range;
+	pc->gpio_range.base = pc->gpio_chip.base;
+	pc->gpio_range.gc = &pc->gpio_chip;
+	pinctrl_add_gpio_range(pc->pctl_dev, &pc->gpio_range);
+
 	girq = &pc->gpio_chip.irq;
 	girq->chip = &bcm2835_gpio_irq_chip;
 	girq->parent_handler = bcm2835_gpio_irq_handler;
@@ -1251,8 +1263,10 @@ static int bcm2835_pinctrl_probe(struct platform_device *pdev)
 	girq->parents = devm_kcalloc(dev, BCM2835_NUM_IRQS,
 				     sizeof(*girq->parents),
 				     GFP_KERNEL);
-	if (!girq->parents)
+	if (!girq->parents) {
+		pinctrl_remove_gpio_range(pc->pctl_dev, &pc->gpio_range);
 		return -ENOMEM;
+	}
 
 	if (is_7211) {
 		pc->wake_irq = devm_kcalloc(dev, BCM2835_NUM_IRQS,
@@ -1307,21 +1321,10 @@ static int bcm2835_pinctrl_probe(struct platform_device *pdev)
 	err = gpiochip_add_data(&pc->gpio_chip, pc);
 	if (err) {
 		dev_err(dev, "could not add GPIO chip\n");
+		pinctrl_remove_gpio_range(pc->pctl_dev, &pc->gpio_range);
 		return err;
 	}
 
-	pc->pctl_desc = *pdata->pctl_desc;
-	pc->pctl_dev = devm_pinctrl_register(dev, &pc->pctl_desc, pc);
-	if (IS_ERR(pc->pctl_dev)) {
-		gpiochip_remove(&pc->gpio_chip);
-		return PTR_ERR(pc->pctl_dev);
-	}
-
-	pc->gpio_range = *pdata->gpio_range;
-	pc->gpio_range.base = pc->gpio_chip.base;
-	pc->gpio_range.gc = &pc->gpio_chip;
-	pinctrl_add_gpio_range(pc->pctl_dev, &pc->gpio_range);
-
 	return 0;
 }
 
Signature · 1 line
2.25.1

A message that is not in the archive (sent off-list or not yet received): CAHp75Vei9FUY0gGD99gVv_FZzcpN1y_i65BB-auyAFUwqsQxNA@mail.gmail.com

#3 lore raw

Re: [PATCH 1/2] pinctrl: bcm2835: Change init order for gpio hogs

Hi Andy,

Quoted 11 lines · Andy Shevchenko

On Monday, November 29, 2021, Phil Elwell <[email protected] <mailto:[email protected]>> wrote:

...and gpio-ranges

pinctrl-bcm2835 is a combined pinctrl/gpio driver. Currently the gpio side is registered first, but this breaks gpio hogs (which are configured during gpiochip_add_data). Part of the hog initialisation is a call to pinctrl_gpio_request, and since the pinctrl driver hasn't yet been registered this results in an -EPROBE_DEFER from which it can never recover.

Quoted 8 lines

Change the initialisation sequence to register the pinctrl driver first.

This also solves a similar problem with the gpio-ranges property, which is required in order for released pins to be returned to inputs.

We have a callback in GPIO chip to register pin ranges, why driver does it separately?

A few experiments (this is not my driver) appear to show that the call to pinctrl_add_gpio_range can be removed, but only once the gpio-ranges DT property has been added if we want to remain functionality throughout a bisect. That tidy up might be better done with a followup commit once the DT patch has also been accepted, unless it's possible to guarantee the sequencing between
the pinctrl/gpio tree and the DT tree.

Quoted 7 lines
Fixes 73345a18d464b ("pinctrl: bcm2835: Pass irqchip when adding

                       gpiochip")

Signed-off-by Phil Elwell <[email protected] <mailto:[email protected]>>

Is it originally so strange indentation or is it only on my side?

The "g" is below the "p" in the patch.

Thanks,

Phil

Quoted 76 lines
    ---
      drivers/pinctrl/bcm/pinctrl-bcm2835.c | 29 +++++++++++++++------------
      1 file changed, 16 insertions(+), 13 deletions(-)

    diff --git a/drivers/pinctrl/bcm/pinctrl-bcm2835.c
    b/drivers/pinctrl/bcm/pinctrl-bcm2835.c
    index 2abcc6ce4eba..b607d10e4cbd 100644
    --- a/drivers/pinctrl/bcm/pinctrl-bcm2835.c
    +++ b/drivers/pinctrl/bcm/pinctrl-bcm2835.c
    @@ -1244,6 +1244,18 @@ static int bcm2835_pinctrl_probe(struct
    platform_device *pdev)
                     raw_spin_lock_init(&pc->irq_lock[i]);
             }

    +       pc->pctl_desc = *pdata->pctl_desc;
    +       pc->pctl_dev = devm_pinctrl_register(dev, &pc->pctl_desc, pc);
    +       if (IS_ERR(pc->pctl_dev)) {
    +               gpiochip_remove(&pc->gpio_chip);
    +               return PTR_ERR(pc->pctl_dev);
    +       }
    +
    +       pc->gpio_range = *pdata->gpio_range;
    +       pc->gpio_range.base = pc->gpio_chip.base;
    +       pc->gpio_range.gc = &pc->gpio_chip;
    +       pinctrl_add_gpio_range(pc->pctl_dev, &pc->gpio_range);
    +
             girq = &pc->gpio_chip.irq;
             girq->chip = &bcm2835_gpio_irq_chip;
             girq->parent_handler = bcm2835_gpio_irq_handler;
    @@ -1251,8 +1263,10 @@ static int bcm2835_pinctrl_probe(struct
    platform_device *pdev)
             girq->parents = devm_kcalloc(dev, BCM2835_NUM_IRQS,
                                          sizeof(*girq->parents),
                                          GFP_KERNEL);
    -       if (!girq->parents)
    +       if (!girq->parents) {
    +               pinctrl_remove_gpio_range(pc->pctl_dev, &pc->gpio_range);
                     return -ENOMEM;
    +       }

             if (is_7211) {
                     pc->wake_irq = devm_kcalloc(dev, BCM2835_NUM_IRQS,
    @@ -1307,21 +1321,10 @@ static int bcm2835_pinctrl_probe(struct
    platform_device *pdev)
             err = gpiochip_add_data(&pc->gpio_chip, pc);
             if (err) {
                     dev_err(dev, "could not add GPIO chip\n");
    +               pinctrl_remove_gpio_range(pc->pctl_dev, &pc->gpio_range);
                     return err;
             }

    -       pc->pctl_desc = *pdata->pctl_desc;
    -       pc->pctl_dev = devm_pinctrl_register(dev, &pc->pctl_desc, pc);
    -       if (IS_ERR(pc->pctl_dev)) {
    -               gpiochip_remove(&pc->gpio_chip);
    -               return PTR_ERR(pc->pctl_dev);
    -       }
    -
    -       pc->gpio_range = *pdata->gpio_range;
    -       pc->gpio_range.base = pc->gpio_chip.base;
    -       pc->gpio_range.gc = &pc->gpio_chip;
    -       pinctrl_add_gpio_range(pc->pctl_dev, &pc->gpio_range);
    -
             return 0;
      }

    -- 
    2.25.1
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw

Re: [PATCH 1/2] pinctrl: bcm2835: Change init order for gpio hogs

Quoted 23 lines · Phil Elwell
Quoted 15 lines · Andy Shevchenko

On Monday, November 29, 2021, Phil Elwell <[email protected] <mailto:[email protected]>> wrote:

...and gpio-ranges

pinctrl-bcm2835 is a combined pinctrl/gpio driver. Currently the gpio side is registered first, but this breaks gpio hogs (which are configured during gpiochip_add_data). Part of the hog initialisation is a call to pinctrl_gpio_request, and since the pinctrl driver hasn't yet been registered this results in an -EPROBE_DEFER from which it can never recover.

Change the initialisation sequence to register the pinctrl driver first.

Quoted 6 lines

This also solves a similar problem with the gpio-ranges property, which is required in order for released pins to be returned to inputs.

We have a callback in GPIO chip to register pin ranges, why driver does it separately?

Quoted 6 lines

A few experiments (this is not my driver) appear to show that the call to pinctrl_add_gpio_range can be removed, but only once the gpio-ranges DT property has been added if we want to remain functionality throughout a bisect. That tidy up might be better done with a followup commit once the DT patch has also been accepted, unless it's possible to guarantee the sequencing between the pinctrl/gpio tree and the DT tree.

What I meant is why these calls are done in the probe and not in ->add_pin_ranges() callback?
Shouldn't it fix the issue you have observed?

...

Quoted 7 lines
Quoted 5 lines
Fixes 73345a18d464b ("pinctrl: bcm2835: Pass irqchip when adding

gpiochip")

Signed-off-by Phil Elwell <[email protected] <mailto:[email protected]>>

Is it originally so strange indentation or is it only on my side?

The "g" is below the "p" in the patch.

Which is wrong. Tags mustn't be multilines (i.e. split over a single line).

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#5 lore raw

Re: [PATCH 1/2] pinctrl: bcm2835: Change init order for gpio hogs

Quoted 31 lines · Andy Shevchenko
Quoted 23 lines · Phil Elwell
Quoted 15 lines · Andy Shevchenko

On Monday, November 29, 2021, Phil Elwell <[email protected] <mailto:[email protected]>> wrote:

...and gpio-ranges

pinctrl-bcm2835 is a combined pinctrl/gpio driver. Currently the gpio side is registered first, but this breaks gpio hogs (which are configured during gpiochip_add_data). Part of the hog initialisation is a call to pinctrl_gpio_request, and since the pinctrl driver hasn't yet been registered this results in an -EPROBE_DEFER from which it can never recover.

Change the initialisation sequence to register the pinctrl driver first.

Quoted 6 lines

This also solves a similar problem with the gpio-ranges property, which is required in order for released pins to be returned to inputs.

We have a callback in GPIO chip to register pin ranges, why driver does it separately?

Quoted 6 lines

A few experiments (this is not my driver) appear to show that the call to pinctrl_add_gpio_range can be removed, but only once the gpio-ranges DT property has been added if we want to remain functionality throughout a bisect. That tidy up might be better done with a followup commit once the DT patch has also been accepted, unless it's possible to guarantee the sequencing between the pinctrl/gpio tree and the DT tree.

Quoted 3 lines

What I meant is why these calls are done in the probe and not in ->add_pin_ranges() callback?
Shouldn't it fix the issue you have observed?

I'm no expert in the field, isn't it preferable to set the gpio-ranges pinctrl<->GPIO correspondence with a single line of DT rather than several lines of code? A quick grep shows over 700 instances of gpio-ranges in DT, at least some of which are reflexive, and only 7 drivers with add_pin_ranges methods.

Quoted 9 lines
Quoted 7 lines
Quoted 5 lines
Fixes 73345a18d464b ("pinctrl: bcm2835: Pass irqchip when adding

gpiochip")

Signed-off-by Phil Elwell <[email protected] <mailto:[email protected]>>

Is it originally so strange indentation or is it only on my side?

The "g" is below the "p" in the patch.

Which is wrong. Tags mustn't be multilines (i.e. split over a single line).

checkpatch disagrees:

scripts/checkpatch.pl 0001-ARM-dts-gpio-ranges-property-is-now-required.patch
WARNING: Possible unwrapped commit description (prefer a maximum 75 chars per line) #10:
[1] commit 2ab73c6d8323 ("gpio: Support GPIO controllers without pin-ranges")

Phil

#6 lore raw

Re: [PATCH 1/2] pinctrl: bcm2835: Change init order for gpio hogs

Quoted 1 line · Phil Elwell

On 01/12/2021 15:38, Andy Shevchenko wrote:

...

Quoted 10 lines
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
     Fixes: 73345a18d464b ("pinctrl: bcm2835: Pass irqchip when adding
                             gpiochip")

Is it originally so strange indentation or is it only on my side?

The "g" is below the "p" in the patch.

Which is wrong. Tags mustn't be multilines (i.e. split over a single line).

checkpatch disagrees:

I don't care less about checkpatch false positives. You have a chance to fix it, btw.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#7 lore raw

Re: [PATCH 1/2] pinctrl: bcm2835: Change init order for gpio hogs

Quoted 10 lines · Andy Shevchenko
Quoted 1 line · Phil Elwell

On 01/12/2021 15:38, Andy Shevchenko wrote:

...

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
      Fixes: 73345a18d464b ("pinctrl: bcm2835: Pass irqchip when adding
                              gpiochip")

Is it originally so strange indentation or is it only on my side?

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

The "g" is below the "p" in the patch.

Which is wrong. Tags mustn't be multilines (i.e. split over a single line).

checkpatch disagrees:

I don't care less about checkpatch false positives. You have a chance to fix it, btw.

I only split it because of checkpatch. What a great tool.

I'll change for V2, which I can submit now if you'll ack it with just that change.

Phil

#8 lore raw

PATCH 2/2 ARM: dts: gpio-ranges property is now required

In reply to Phil Elwell

Since [1], added in 5.7, the absence of a gpio-ranges property has prevented GPIOs from being restored to inputs when released. Add those properties for BCM283x and BCM2711 devices.

[1] commit 2ab73c6d8323 ("gpio: Support GPIO controllers without pin-ranges")

Fixes 2ab73c6d8323 ("gpio: Support GPIO controllers without

pin-ranges")

---
 arch/arm/boot/dts/bcm2711.dtsi | 2 ++
 arch/arm/boot/dts/bcm283x.dtsi | 2 ++
 2 files changed, 4 insertions(+)
arch/arm/boot/dts/bcm2711.dtsi +2 -0
index 9e01dbca4a01..dff18fc9a906 100644
@@ -582,6 +582,8 @@ &gpio {
 		     <GIC_SPI 115 IRQ_TYPE_LEVEL_HIGH>,
 		     <GIC_SPI 116 IRQ_TYPE_LEVEL_HIGH>;
 
+	gpio-ranges = <&gpio 0 0 58>;
+
 	gpclk0_gpio49: gpclk0_gpio49 {
 		pin-gpclk {
 			pins = "gpio49";
arch/arm/boot/dts/bcm283x.dtsi +2 -0
index a3e06b680947..c113661a6668 100644
@@ -126,6 +126,8 @@ gpio: gpio@7e200000 {
 			interrupt-controller;
 			#interrupt-cells = <2>;
 
+			gpio-ranges = <&gpio 0 0 54>;
+
 			/* Defines common pin muxing groups
 			 *
 			 * While each pin can have its mux selected
Signature · 1 line
2.25.1
Reviewed-by on v1 2/2
#9 lore raw

Re: [PATCH 2/2] ARM: dts: gpio-ranges property is now required

Quoted 10 lines · Phil Elwell

Since [1], added in 5.7, the absence of a gpio-ranges property has prevented GPIOs from being restored to inputs when released. Add those properties for BCM283x and BCM2711 devices.

[1] commit 2ab73c6d8323 ("gpio: Support GPIO controllers without pin-ranges")

Fixes 2ab73c6d8323 ("gpio: Support GPIO controllers without

pin-ranges")

With the funny linebreak fixed:

Please merge this patch through the Broadcom/SoC tree.

Yours,
Linus Walleij