Re: [PATCH v1 2/4] iio: gyro: mpu3050: fix runtime PM leak when enabling trigger fails

[Jonathan Cameron <jic23@kernel.org>]

On Sun, 14 Jun 2026 12:45:47 +0530
Biren Pandya <birenpandya@gmail.com> wrote:

> mpu3050_drdy_trigger_set_state() calls pm_runtime_get_sync() when the
> trigger is enabled, but several error paths in the enable branch return
> directly without dropping the usage counter again. pm_runtime_get_sync()
> increments the usage counter, so every failed enable leaks a runtime PM
> reference and the device can no longer autosuspend. The driver state flag
> hw_irq_trigger is also left set after a failed enable.
> 
> Jump to a common error label that drops the reference and clears
> hw_irq_trigger, mirroring the disable path. The success path deliberately
> keeps the reference, as the device must stay resumed while the trigger is
> active.
> 
> Signed-off-by: Biren Pandya <birenpandya@gmail.com>
> Assisted-by: Claude:claude-opus-4-8 coccinelle

This looks more or less fine to me, but I'd like to leave it on list
for a little while so hopefully someone who is more familiar with the device
can take a look.

A suggestion below for a less minimal but perhaps nicer way of doing this.
> ---
>  drivers/iio/gyro/mpu3050-core.c | 16 +++++++++++-----
>  1 file changed, 11 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/iio/gyro/mpu3050-core.c b/drivers/iio/gyro/mpu3050-core.c
> index d84e04e4b431..783be9a786df 100644
> --- a/drivers/iio/gyro/mpu3050-core.c
> +++ b/drivers/iio/gyro/mpu3050-core.c
> @@ -994,14 +994,14 @@ static int mpu3050_drdy_trigger_set_state(struct iio_trigger *trig,
>  		/* Disable all things in the FIFO */
>  		ret = regmap_write(mpu3050->map, MPU3050_FIFO_EN, 0);
>  		if (ret)
> -			return ret;
> +			goto err_power_down;
>  
>  		/* Reset and enable the FIFO */
>  		ret = regmap_set_bits(mpu3050->map, MPU3050_USR_CTRL,
>  				      MPU3050_USR_CTRL_FIFO_EN |
>  				      MPU3050_USR_CTRL_FIFO_RST);
>  		if (ret)
> -			return ret;
> +			goto err_power_down;
>  
>  		mpu3050->pending_fifo_footer = false;
>  
> @@ -1013,12 +1013,12 @@ static int mpu3050_drdy_trigger_set_state(struct iio_trigger *trig,
>  				   MPU3050_FIFO_EN_GYRO_ZOUT |
>  				   MPU3050_FIFO_EN_FOOTER);
>  		if (ret)
> -			return ret;
> +			goto err_power_down;
>  
>  		/* Configure the sample engine */
>  		ret = mpu3050_start_sampling(mpu3050);
>  		if (ret)
> -			return ret;
> +			goto err_power_down;
>  
>  		/* Clear IRQ flag */
>  		ret = regmap_read(mpu3050->map, MPU3050_INT_STATUS, &val);
> @@ -1037,10 +1037,16 @@ static int mpu3050_drdy_trigger_set_state(struct iio_trigger *trig,
>  
>  		ret = regmap_write(mpu3050->map, MPU3050_INT_CFG, val);
>  		if (ret)
> -			return ret;
> +			goto err_power_down;
>  	}
>  
>  	return 0;
> +
> +err_power_down:

I'm not particularly keen on a goto error block that only applies to one
leg of an if / else.  Given there is effectively not sharing of code between the enable
and disable paths, I'd rather see this whole thing broken into two helpers.
mpu3050_drdy_trigger_enable() and mpu3050_drdy_trigger_disable()

Then the error unwind would not have the issue of only applying to some code
paths as it would be in the helper.  Obviously it is a more invasive change
but I think it is worth doing rather that having a minimal fix then sweeping around
later to replace all that code.


> +	pm_runtime_put_autosuspend(mpu3050->dev);
> +	mpu3050->hw_irq_trigger = false;
> +
> +	return ret;
>  }
>  
>  static const struct iio_trigger_ops mpu3050_trigger_ops = {