Re: [PATCH] iio: pressure: mpl115: Fix runtime PM cleanup on probe failure

[Jonathan Cameron]

On Sat, 18 Jul 2026 14:19:09 +0800
Can Peng <pengcan@xxxxxxxxxx> wrote:

> mpl115_probe() enables runtime PM when a shutdown GPIO is present and
> then returns the result of devm_iio_device_register(). If registration
> fails, runtime PM remains enabled and autosuspend remains selected.
> 
> The same unmanaged runtime PM state is also left behind on driver
> unbind, as the IIO device registration is managed but the runtime PM
> setup is not.
> 
> Use devm_pm_runtime_enable() so runtime PM is disabled automatically
> on probe failure and driver unbind. Also unwind the initial
> pm_runtime_get_noresume() reference if setting up runtime PM fails
> before the matching put.
> 
> Fixes: 0c3a333524a3 ("iio: pressure: mpl115: Implementing low power mode by shutdown gpio")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Can Peng <pengcan@xxxxxxxxxx>
Hi Can Peng,

A similar code pattern came up recently based as bold claim from sashiko
that autosuspend won't kick in if we don't do a get / put pair.

Turned out not to be true.  If no one touches the runtime pm counts
and turns runtime pm on, then this line:
https://elixir.bootlin.com/linux/v7.1.3/source/drivers/base/dd.c#L872
becomes relevant.    That ends up calling auto suspend.
Please chase through this call sequence just in case we got it wrong!

Hence I believe this is a simplification change that can be made here.

> ---
>  drivers/iio/pressure/mpl115.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/iio/pressure/mpl115.c b/drivers/iio/pressure/mpl115.c
> index 16e112b796ba..9e77b76bd1fa 100644
> --- a/drivers/iio/pressure/mpl115.c
> +++ b/drivers/iio/pressure/mpl115.c
> @@ -204,8 +204,17 @@ int mpl115_probe(struct device *dev, const char *name,
>  	if (data->shutdown) {
>  		/* Enable runtime PM */
>  		pm_runtime_get_noresume(dev);
Drop this get_noresume()

> -		pm_runtime_set_active(dev);
> -		pm_runtime_enable(dev);
> +		ret = pm_runtime_set_active(dev);
> +		if (ret) {
> +			pm_runtime_put_noidle(dev);
No need for the put here
> +			return ret;
> +		}
> +
> +		ret = devm_pm_runtime_enable(dev);
> +		if (ret) {
> +			pm_runtime_put_noidle(dev);
or here.

The switch to devm_pm_runtime_enable() and error checks all look good
to me.

> +			return ret;
> +		}
>  
>  		/*
>  		 * As the device takes 3 ms to come up with a fresh

and the pm_runtime_put() at end of this block

We don't mind if auto runtime suspend happens a tiny bit later.