Re: [PATCH v1 1/1] ACPI: bus: Introduce acpi_dev_get() and reuse it in ACPI code

From: Rafael J. Wysocki
Date: Mon Apr 12 2021 - 14:05:46 EST


On Mon, Apr 12, 2021 at 7:47 PM Andy Shevchenko
<andy.shevchenko@xxxxxxxxx> wrote:
>
> On Mon, Apr 12, 2021 at 8:32 PM Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> > On Sat, Apr 10, 2021 at 3:47 PM Andy Shevchenko
> > <andy.shevchenko@xxxxxxxxx> wrote:
>
> ...
>
> > > static void get_acpi_device(void *dev)
> > > {
> > > - if (dev)
> > > - get_device(&((struct acpi_device *)dev)->dev);
> > > + acpi_dev_get(dev);
> >
> > I would do
> >
> > if (dev)
> > acpi_dev_get(dev);
> >
> > here.
>
> Hmm... I don't see a point. acpi_dev_get() guaranteed to perform this check.
>
> > > }
>
>
> > > +static inline void acpi_dev_get(struct acpi_device *adev)
> > > +{
> > > + if (adev)
> > > + get_device(&adev->dev);
> >
> > And I would drop the adev check from here (because the code calling it
> > may be running with wrong assumptions if adev is NULL). Or it should
> > return adev and the caller should be held responsible for checking it
> > against NULL (if they care).
>
> But this follows the get_device() / put_device() logic.

Not really. get_device() returns a pointer.

> Personally I don't think this is a good idea to deviate.

Well, exactly. :-)

> Note the acpi_bus_get_acpi_device()

This also returns a pointer.

> / acpi_bus_put_acpi_device() as well.