Skip to contentSkip to conversation

leds: gpio: Skip unavailable LEDs during shutdown

PATCH In review Steve Dunnagan 4 participants 10 messages Jul 10 – Aug 27 LED

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 leds: gpio: Skip unavailable LEDs during shutdown

gpio_led_probe() leaves an error-valued GPIO descriptor in the LED data when an unavailable platform-data LED is skipped.

The entry remains included in priv->num_leds, so gpio_led_shutdown() later passes the error pointer to gpio_led_set(), producing:

gpiod_set_value: invalid GPIO (errorpointer: -ENOENT)

Skip entries with error-valued GPIO descriptors during shutdown.

Fixes 45d4c6de4e49 ("leds: gpio: Try to lookup gpiod from device")
Assisted-by ChatGPT:GPT-5.5-Thinking
---
 drivers/leds/leds-gpio.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/leds/leds-gpio.c +3 -0
index 8ae71c2e91e0..63cb517ef385 100644
@@ -304,6 +304,9 @@ static void gpio_led_shutdown(struct platform_device *pdev)
 	for (i = 0; i < priv->num_leds; i++) {
 		struct gpio_led_data *led = &priv->leds[i];
 
+		if (IS_ERR(led->gpiod))
+			continue;
+
 		if (!(led->cdev.flags & LED_RETAIN_AT_SHUTDOWN))
 			gpio_led_set(&led->cdev, LED_OFF);
 	}
Signature · 1 line
2.49.0
#1 lore raw
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:

Pre-existing issues:
- [High] The platform data probe path silently skips LEDs when GPIO lookup returns -EPROBE_DEFER, breaking deferred probing.
- [High] Teardown race: gpio_led_shutdown fails to disable software blink timers and active triggers, allowing concurrent LED updates during system shutdown.
- [Low] The driver does not use dev_err_probe() to handle and report probe failures.
- [Low] The driver uses generic names 'priv' and 'gpio_leds_priv' for its private data, violating subsystem naming conventions.
--

--- Patch [1]: [PATCH] leds: gpio: Skip unavailable LEDs during shutdown --- commit 94da8744ab857bc87f8158e7c51602a324f2f403
Author: Steve Dunnagan <[email protected]>
Subject: leds: gpio: Skip unavailable LEDs during shutdown

This commit fixes an issue in gpio_led_shutdown() where an error-valued GPIO descriptor could be passed to gpio_led_set(). It avoids a potential invalid GPIO operation by ensuring we skip unavailable LEDs during the shutdown process.

drivers/leds/leds-gpio.c static void gpio_led_shutdown(struct platform_device *pdev) · line 290

Quoted 6 lines
drivers/leds/leds-gpio.c +0 -0
index ce4e79939731d..6a2f2386978d7 100644
@@ -290,6 +290,9 @@ static void gpio_led_shutdown(struct platform_device *pdev)
 	for (i = 0; i < priv->num_leds; i++) {

[Severity: Low]
This is a pre-existing issue, but does the use of the generic "priv" variable name violate subsystem naming conventions?

The LED subsystem guidelines prefer using the device name and "ddata" for private data structures and variables rather than generic names like "priv" or "gpio_leds_priv".

drivers/leds/leds-gpio.c static void gpio_led_shutdown(struct platform_device *pdev) · line 311

Quoted 7 lines
drivers/leds/leds-gpio.c +3 -0
 		struct gpio_led_data *led = &priv->leds[i];
 
+		if (IS_ERR(led->gpiod))
+			continue;
+
 		if (!(led->cdev.flags & LED_RETAIN_AT_SHUTDOWN))
 			gpio_led_set(&led->cdev, LED_OFF);

[Severity: High]
This isn't a bug introduced by this patch, but does gpio_led_shutdown() fail to disable software blink timers and active triggers?

While the physical GPIO is turned off here, if a heartbeat or timer trigger fires during the device shutdown phase (while interrupts are still active), it could queue a work to change the LED brightness. If the underlying GPIO controller or sleepable bus is already shut down, could this concurrent access cause a crash or bus lockup?

drivers/leds/leds-gpio.c static void gpio_led_shutdown(struct platform_device *pdev) · line 298

Quoted 1 line
drivers/leds/leds-gpio.c +0 -0
 	}

While reviewing this file, I noticed a couple of other pre-existing issues in gpio_led_probe().

[Severity: High]
This is a pre-existing issue, but does gpio_led_probe() silently swallow probe deferrals?

drivers/leds/leds-gpio.c:gpio_led_probe() {
    ...
			if (IS_ERR(led_dat->gpiod)) {
				dev_info(&pdev->dev, "Skipping unavailable LED gpio %d (%s)\n",
					 template->gpio, template->name);
				continue;
			}
    ...
}

If the GPIO lookup returns -EPROBE_DEFER (for example, if the controller probes later), this catch-all check logs a generic message and skips the LED instead of propagating the error. Will this cause the LED to be permanently missing since the kernel won't re-attempt probing?

[Severity: Low]
This is another pre-existing issue, but should gpio_led_probe() use dev_err_probe() to handle failures?

drivers/leds/leds-gpio.c:gpio_led_probe() {
    ...
		priv = gpio_leds_create(pdev);
		if (IS_ERR(priv))
			return PTR_ERR(priv);
    ...
}

Errors originating from gpio_leds_create() are returned directly. Subsystem guidelines recommend always using dev_err_probe() to report probe failures and correctly handle -EPROBE_DEFER.

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
maintainer
#2 lore raw
In reply to Steve Dunnagan
Quoted 2 lines · Steve Dunnagan

gpio_led_probe() leaves an error-valued GPIO descriptor in the LED data when an unavailable platform-data LED is skipped.

So why not nip that in the bud and fix that instead?

Quoted 31 lines

The entry remains included in priv->num_leds, so gpio_led_shutdown() later passes the error pointer to gpio_led_set(), producing:

gpiod_set_value: invalid GPIO (errorpointer: -ENOENT)

Skip entries with error-valued GPIO descriptors during shutdown.

Fixes 45d4c6de4e49 ("leds: gpio: Try to lookup gpiod from device")
Assisted-by ChatGPT:GPT-5.5-Thinking
---
 drivers/leds/leds-gpio.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/leds/leds-gpio.c +3 -0
index 8ae71c2e91e0..63cb517ef385 100644
@@ -304,6 +304,9 @@ static void gpio_led_shutdown(struct platform_device *pdev)
 	for (i = 0; i < priv->num_leds; i++) {
 		struct gpio_led_data *led = &priv->leds[i];
 
+		if (IS_ERR(led->gpiod))
+			continue;
+
 		if (!(led->cdev.flags & LED_RETAIN_AT_SHUTDOWN))
 			gpio_led_set(&led->cdev, LED_OFF);
 	}
Signature · 1 line
2.49.0
Signature · 1 line
Lee Jones
#3 lore raw

PATCH v2 1/1 leds: gpio: Clear error pointers for skipped LEDs

In reply to Steve Dunnagan

gpio_led_get_gpiod() returns an error pointer when a platform-data LED's GPIO is unavailable. gpio_led_probe() skips registration in that case, but leaves the error pointer in led_dat->gpiod.

The skipped entry remains included in priv->num_leds. During shutdown, gpio_led_shutdown() walks those entries and passes the error pointer to gpio_led_set(), producing:

gpiod_set_value: invalid GPIO (errorpointer: -ENOENT)

Clear led_dat->gpiod before skipping the LED so skipped entries do not retain error-valued descriptors.

Fixes 45d4c6de4e49 ("leds: gpio: Try to lookup gpiod from device")
Assisted-by ChatGPT:GPT-5.5-Thinking

---
Changes in v2:
- Clear led_dat->gpiod in the probe skip path instead of checking for error-valued descriptors during shutdown, as suggested by Lee.

Tested on an Orange Pi 5 Plus running Fedora 42 with a test platform device containing an unavailable GPIO LED. Before the fix, reboot logged:

gpiod_set_value: invalid GPIO (errorpointer: -ENOENT)

After the fix, the LED was still skipped during probe, but reboot completed without the invalid-GPIO warning.

 drivers/leds/leds-gpio.c | 1 +
 1 file changed, 1 insertion(+)
drivers/leds/leds-gpio.c +1 -0
index 8ae71c2e91e0..8810fdcf2d77 100644
@@ -277,6 +277,7 @@ static int gpio_led_probe(struct platform_device *pdev)
 			if (IS_ERR(led_dat->gpiod)) {
 				dev_info(dev, "Skipping unavailable LED gpio %d (%s)\n",
 					 template->gpio, template->name);
+				led_dat->gpiod = NULL;
 				continue;
 			}
 
Signature · 1 line
2.55.0
#4 lore raw

Re: [PATCH v2] leds: gpio: Clear error pointers for skipped LEDs

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] The driver masks -EPROBE_DEFER from GPIO fetching functions and incorrectly skips the LED instead of deferring the probe. --

--- Patch [1]: [PATCH v2] leds: gpio: Clear error pointers for skipped LEDs --- commit 2d09674f074f178fbc54110548c9a0816564760c
Author: Steve Dunnagan <[email protected]>

leds: gpio: Clear error pointers for skipped LEDs

When an LED's GPIO is unavailable, the error pointer is left in led_dat->gpiod. During shutdown this error pointer is passed to gpio_led_set(), producing an invalid GPIO warning. This commit clears the error pointer to prevent the warning.

drivers/leds/leds-gpio.c static int gpio_led_probe(struct platform_device *pdev) · line 279

Quoted 8 lines
drivers/leds/leds-gpio.c +0 -0
index 8ae71c2e91e01..8810fdcf2d774 100644
@@ -277,6 +277,7 @@ static int gpio_led_probe(struct platform_device *pdev)
 			if (IS_ERR(led_dat->gpiod)) {
 				dev_info(dev, "Skipping unavailable LED gpio %d (%s)\n",
 					 template->gpio, template->name);

[Severity: High]
This isn't a bug introduced by this patch, but does this unconditionally skip the LED even if gpio_led_get_gpiod() returns -EPROBE_DEFER?

If the GPIO controller providing the LED pins hasn't probed yet, gpio_led_get_gpiod() returns ERR_PTR(-EPROBE_DEFER) to gpio_led_probe(). By catching all error pointers with IS_ERR() here and continuing, the driver completes the probe without the deferred LED instead of actually deferring, permanently breaking the LED's functionality.

Would it be better to use dev_err_probe() here to correctly handle -EPROBE_DEFER?

Quoted 4 lines
drivers/leds/leds-gpio.c +1 -0
+				led_dat->gpiod = NULL;
 				continue;
 			}
 
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Reviewed-by on v2 1/1
#5 lore raw

Re: [PATCH v2] leds: gpio: Clear error pointers for skipped LEDs

In reply to Steve Dunnagan
Quoted 10 lines · Steve Dunnagan

gpio_led_get_gpiod() returns an error pointer when a platform-data LED's GPIO is unavailable. gpio_led_probe() skips registration in that case, but leaves the error pointer in led_dat->gpiod.

The skipped entry remains included in priv->num_leds. During shutdown, gpio_led_shutdown() walks those entries and passes the error pointer to gpio_led_set(), producing:

gpiod_set_value: invalid GPIO (errorpointer: -ENOENT)

Quoted 7 lines

Clear led_dat->gpiod before skipping the LED so skipped entries do not retain error-valued descriptors.

Fixes 45d4c6de4e49 ("leds: gpio: Try to lookup gpiod from device")
Assisted-by ChatGPT:GPT-5.5-Thinking

Yours,
Linus Walleij

maintainer Applied
#6 lore raw

Re: (subset) [PATCH v2] leds: gpio: Clear error pointers for skipped LEDs

In reply to Steve Dunnagan
Quoted 9 lines · Steve Dunnagan

gpio_led_get_gpiod() returns an error pointer when a platform-data LED's GPIO is unavailable. gpio_led_probe() skips registration in that case, but leaves the error pointer in led_dat->gpiod.

The skipped entry remains included in priv->num_leds. During shutdown, gpio_led_shutdown() walks those entries and passes the error pointer to gpio_led_set(), producing:

[...]

Applied, thanks!

[1/1] leds: gpio: Clear error pointers for skipped LEDs

      commit: 942901eeda934f1bebf2605a781155e9d6bc7f6e

--
Lee Jones [李琼斯]

#9 lore raw

Re: [PATCH v2] leds: gpio: Clear error pointers for skipped LEDs

Sorry for the noise, I missed the applied message.

Thanks for applying it.

Steve

Quoted 13 lines · Lee Jones
Quoted 7 lines · Steve Dunnagan

Hi Lee, Pavel,

Just checking whether anything else is needed for this v2.

Linus Walleij replied with:

I'm not sure what you're asking:

https://lore.kernel.org/r/178602432127.2934315.922699465700010447.b4-ty@b4