Re: [PATCH v1 1/1] pinctrl: denverton: Enable platform device in the absence of ACPI enumeration

[Andy Shevchenko <andy.shevchenko@gmail.com>]

On Thu, Oct 05, 2023 at 11:38:38AM +0300, Andy Shevchenko wrote:
> On Thu, Oct 05, 2023 at 09:53:51AM +0200, Krzysztof Kozlowski wrote:
> > On 04/10/2023 21:42, Andy Shevchenko wrote:
> > > On Wed, Oct 4, 2023 at 4:20 PM Krzysztof Kozlowski <krzk@kernel.org> wrote:
> > >> On 04/10/2023 15:09, Andy Shevchenko wrote:
> > >>> On Wed, Oct 4, 2023 at 11:18 AM Krzysztof Kozlowski <krzk@kernel.org> wrote:
> > >>>> On 26/09/2023 21:08, Andy Shevchenko wrote:

...

> > >>>>> +MODULE_ALIAS("platform:denverton-pinctrl");
> > >>>>
> > >>>> Why do you need the alias? It's the same as ID table. You most likely
> > >>>> miss MODULE_DEVICE_TABLE() or your table is just wrong.
> > >>>
> > >>> This is cargo cult from pinctrl-broxton.c. If we want to fix, we need
> > >>> to fix both.
> > >>> Care to send patches?
> > >>
> > >> I don't understand how some other file affects this. Why do you exactly
> > >> need module alias here? Which use-case does not work without it (after
> > >> adding proper MODULE_DEVICE_TABLE())?
> > > 
> > > We try to keep these drivers uniform. So, if something should be done
> > > differently, it should be done for all affected drivers.
> > > Since patch is applied, now we have two drivers to improve.
> > 
> > Awesome reason. So if you add accidentally bug to one driver, it should
> > be copied to other file as well.
> 
> I'm not sure where it's a bug, it works. The problem is implementation
> and I believe Mika and certainly me agree with your points that
> MODULE_DEVICE_TABLE() is a right way to go.

The initial implementation was done by 0c3013bbe1d8 ("pinctrl/broxton:
enable platform device in the absence of ACPI enumeration"). It was
ACKed by the driver maintainer as this one. But again, we agree that
your way is better.

-- 
With Best Regards,
Andy Shevchenko