[PATCH] hwmon: (gpio-fan) Fix runtime PM imbalance in probe and teardown

[PATCH] hwmon: (gpio-fan) Fix runtime PM imbalance in probe and teardown

[Guenter Roeck]

In gpio_fan_probe(), gpio_fan_stop() is registered conditionally as a
devres action before runtime PM is initialized, while gpio_fan_stop()
unconditionally calls pm_runtime_disable(). This breaks runtime PM
disable_depth and usage_count invariants across several paths:

- On alarm-only devices without control GPIOs, gpio_fan_stop() is never
  registered, leaving pm_runtime_enable() unbalanced on unbind and
  triggering "Unbalanced pm_runtime_enable!" on subsequent binds.
- When control GPIOs are present and devm_add_action_or_reset(), hwmon
  registration, or fan_alarm_init() fails, gpio_fan_stop() runs before
  pm_runtime_enable(), incrementing disable_depth to 2 (leaving runtime
  PM permanently disabled on re-probe) and, if speed_index > 0,
  underflowing usage_count via pm_runtime_put_sync().
- If pm_runtime_resume_and_get() fails when speed_index > 0, it drops
  its reference while speed_index remains non-zero, causing
  gpio_fan_stop() to call pm_runtime_put_sync() and underflow
  usage_count.
- Registering the hwmon device before enabling runtime PM exposes sysfs
  attributes that can invoke set_fan_speed() before runtime PM state is
  initialized.

Initialize runtime PM upfront with pm_runtime_set_suspended() and
devm_pm_runtime_enable(), remove pm_runtime_disable() from
gpio_fan_stop(), and acquire the initial active runtime PM reference
before registering gpio_fan_stop() and the hwmon device.

Fixes: 0d01110e6356 ("hwmon: (gpio-fan) Add regulator support")
Reported-by: Sashiko <sashiko@sashiko.dev>
Link: https://sashiko.dev/#/bug/linux-349aa3ba-30c3-4c3c-b47e-f52839af2b8d
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/gpio-fan.c | 24 +++++++++++-------------
 1 file changed, 11 insertions(+), 13 deletions(-)

diff --git a/drivers/hwmon/gpio-fan.c b/drivers/hwmon/gpio-fan.c
index 08b58b49be5b..402cc000bc3f 100644
--- a/drivers/hwmon/gpio-fan.c
+++ b/drivers/hwmon/gpio-fan.c
@@ -531,8 +531,6 @@ static void gpio_fan_stop(void *data)
 	mutex_lock(&fan_data->lock);
 	set_fan_speed(data, 0);
 	mutex_unlock(&fan_data->lock);
-
-	pm_runtime_disable(fan_data->dev);
 }
 
 static int gpio_fan_probe(struct platform_device *pdev)
@@ -560,6 +558,11 @@ static int gpio_fan_probe(struct platform_device *pdev)
 		return dev_err_probe(dev, PTR_ERR(fan_data->supply),
 				     "Failed to get fan-supply");
 
+	pm_runtime_set_suspended(dev);
+	err = devm_pm_runtime_enable(dev);
+	if (err)
+		return err;
+
 	/* Configure control GPIOs if available. */
 	if (fan_data->gpios && fan_data->num_gpios > 0) {
 		if (!fan_data->speed || fan_data->num_speed <= 1)
@@ -567,6 +570,12 @@ static int gpio_fan_probe(struct platform_device *pdev)
 		err = fan_ctrl_init(fan_data);
 		if (err)
 			return err;
+		/* If current GPIO state is active, mark RPM as active as well */
+		if (fan_data->speed_index > 0) {
+			err = pm_runtime_resume_and_get(dev);
+			if (err)
+				return err;
+		}
 		err = devm_add_action_or_reset(dev, gpio_fan_stop, fan_data);
 		if (err)
 			return err;
@@ -587,17 +596,6 @@ static int gpio_fan_probe(struct platform_device *pdev)
 			return err;
 	}
 
-	pm_runtime_set_suspended(&pdev->dev);
-	pm_runtime_enable(&pdev->dev);
-	/* If current GPIO state is active, mark RPM as active as well */
-	if (fan_data->speed_index > 0) {
-		int ret;
-
-		ret = pm_runtime_resume_and_get(&pdev->dev);
-		if (ret)
-			return ret;
-	}
-
 	/* Optional cooling device register for Device tree platforms */
 	fan_data->cdev = devm_thermal_of_child_cooling_device_register(dev, np,
 								       "gpio-fan",
-- 
2.45.2

Re: [PATCH] hwmon: (gpio-fan) Fix runtime PM imbalance in probe and teardown

[sashiko-bot]

> In gpio_fan_probe(), gpio_fan_stop() is registered conditionally as a
> devres action before runtime PM is initialized, while gpio_fan_stop()
> unconditionally calls pm_runtime_disable(). This breaks runtime PM
> disable_depth and usage_count invariants across several paths:
> 
> - On alarm-only devices without control GPIOs, gpio_fan_stop() is never
>   registered, leaving pm_runtime_enable() unbalanced on unbind and
>   triggering "Unbalanced pm_runtime_enable!" on subsequent binds.
> - When control GPIOs are present and devm_add_action_or_reset(), hwmon
>   registration, or fan_alarm_init() fails, gpio_fan_stop() runs before
>   pm_runtime_enable(), incrementing disable_depth to 2 (leaving runtime
>   PM permanently disabled on re-probe) and, if speed_index > 0,
>   underflowing usage_count via pm_runtime_put_sync().
> - If pm_runtime_resume_and_get() fails when speed_index > 0, it drops
>   its reference while speed_index remains non-zero, causing
>   gpio_fan_stop() to call pm_runtime_put_sync() and underflow
>   usage_count.
> - Registering the hwmon device before enabling runtime PM exposes sysfs
>   attributes that can invoke set_fan_speed() before runtime PM state is
>   initialized.
> 
> Initialize runtime PM upfront with pm_runtime_set_suspended() and
> devm_pm_runtime_enable(), remove pm_runtime_disable() from
> gpio_fan_stop(), and acquire the initial active runtime PM reference
> before registering gpio_fan_stop() and the hwmon device.
> 
> Fixes: 0d01110e6356 ("hwmon: (gpio-fan) Add regulator support")
> Reported-by: Sashiko <sashiko@sashiko.dev>
> Link: https://sashiko.dev/#/bug/linux-349aa3ba-30c3-4c3c-b47e-f52839af2b8d
> Signed-off-by: Guenter Roeck <linux@roeck-us.net>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261105182427.2369928-1-linux@roeck-us.net?part=1