On Wed,  2 Oct 2019 10:57:57 +0200
Bartosz Golaszewski <b...@bgdev.pl> wrote:

> From: Bartosz Golaszewski <bgolaszew...@baylibre.com>
> 
> We can use the managed variant of iio_device_register() and remove
> the corresponding unregister operation from the remove callback.
> 
> Signed-off-by: Bartosz Golaszewski <bgolaszew...@baylibre.com>

This results in a race where the regulators are powered down before
we remove the userspace interfaces.  All sorts of fun can therefore
occur...

If we fix that with some devm_add_action_or_reset fun then there
is still the fact that we loose the 'obviously correct' property
of the remove being a mirror of the probe because the ordering
wrt to runtime_pm is different.  

So I'd leave this one alone.

Thanks,

Jonathan


> ---
>  drivers/iio/pressure/bmp280-core.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/iio/pressure/bmp280-core.c 
> b/drivers/iio/pressure/bmp280-core.c
> index c21f8ce7b09c..f22400e1e98f 100644
> --- a/drivers/iio/pressure/bmp280-core.c
> +++ b/drivers/iio/pressure/bmp280-core.c
> @@ -1127,7 +1127,7 @@ int bmp280_common_probe(struct device *dev,
>       pm_runtime_use_autosuspend(dev);
>       pm_runtime_put(dev);
>  
> -     ret = iio_device_register(indio_dev);
> +     ret = devm_iio_device_register(dev, indio_dev);
>       if (ret)
>               goto out_runtime_pm_disable;
>  
> @@ -1149,7 +1149,6 @@ int bmp280_common_remove(struct device *dev)
>       struct iio_dev *indio_dev = dev_get_drvdata(dev);
>       struct bmp280_data *data = iio_priv(indio_dev);
>  
> -     iio_device_unregister(indio_dev);
>       pm_runtime_get_sync(data->dev);
>       pm_runtime_put_noidle(data->dev);
>       pm_runtime_disable(data->dev);

Reply via email to