Re: [PATCH RFC] gpio: move intel drivers into their own directory

[Bartosz Golaszewski]

On Wed, 30 Sep 2026 10:08:02 +0200, Andy Shevchenko
<andriy.shevchenko@xxxxxxxxx> said:
> On Tue, Sep 29, 2026 at 10:53:40PM +0300, Sakari Ailus wrote:
>> On Tue, Sep 29, 2026 at 06:09:20PM +0300, Andy Shevchenko wrote:
>> > On Tue, Sep 29, 2026 at 5:05 PM Bartosz Golaszewski <brgl@xxxxxxxxxx> wrote:
>> > > On Tue, Sep 29, 2026 at 3:31 PM Andy Shevchenko
>> > > <andy.shevchenko@xxxxxxxxx> wrote:
>
> ...
>
>> > > > > > in any case my series is blocked as I have not seen
>> > > > > > a consensus on what to do. Or do we? Bart, Sakari?
>> > > > >
>> > > > > If I didn't make myself clear: I would love it if there was a single source
>> > > > > of intel GPIO pull-requests (most likely Andy) for *all* intel GPIO drivers
>> > > > > even if they have additional reviewers. I'd also love for them to live in
>> > > > > a single subdirectory at drivers/gpio/intel/. Otherwise, it's still chaos like
>> > > > > right now, just worse.
>> > > > >
>> > > > > This is in line with pinctrl vendor directories. I also don't see a reason for
>> > > > > moving Kconfig options into their own vendor submenus if we keep the top-level
>> > > > > entries sorted by driver type.
>> > > >
>> > > > So, as of now it doesn't seem like an agreement among all parties. My
>> > > > opinion is that it's good to have PMIC be separated from the other
>> > > > code, although I don't know if the camera PMIC GPIO driver that is
>> > > > used with IPUx is good to go under the intel/pmic subfolder. I also
>> > > > believe that camera related drivers are the best to be handled by the
>> > > > media tree (as it's done, exempli gratia, for int3472).
>> > > >
>> > > > At the end of the day you will have a single PR from me, but it won't
>> > > > include camera stuff unless Sakari agrees to send a PR to me (which I
>> > > > think won't be the case).
>>
>> That'd work, sure, but there are often dependencies between all three USBIO
>> related drivers (main, I²C and GPIO) so they'll all go typically via usb
>> tree in that case. Having them all be merged via the media tree would
>> require more discussions.
>>
>> > > I thought we agreed usbio would be moved out of drivers/gpio/ and into
>> > > media? That leaves intel/pmic as the only subdirectory IIUC.
>> >
>> > Hmm... I cited the int3472 MAINTAINERS record to point exactly on the
>> > variant where it's left under drivers/gpio. Also IIUC Hans' opinion is
>> > that GPIO drivers even being USB GPIO expanders for camera are still
>> > GPIOs.
>>
>> Indeed.
>
> So as I see now there is a consensus that
> 1) it's good to have Intel related GPIO drivers to be in 'intel' folder;
> 2) the USB GPIO drivers are still a GPIO and should reside under drivers/gpio/ (somewhere beneath that, subfolder or not).
>
> We do not have a consensus on:
> 1) PMIC separation to the 'intel/pmic' subfolder;
> 2) USB GPIO route for patches (with them still be somewhere under drivers/gpio/).
>

If they live under drivers/gpio/, then I want to route their patches though the
GPIO tree, othrwise we're adding more confusion which is the opposite of the
goal of this series. FWIW I'm fine with it.

> Unanswered question:
> 1) do we need to consider gpio-tps68470 (camera PMIC on Intel HW) as part of
> 'intel/pmic' or even 'intel' in general? Currently the Intel HW is the only
> user of it.
>

If it's not intel hardware but just happens to be used exclusively on intel,
then sure: let's leave it under drivers/gpio/.

> P.S. It seems a big deal to move the drivers from one folder to another :-)
>

We're doing it to make it less confusing for me. If we do it wrong, I'll still
keep on picking up patches to the wrong drivers or ignoring the ones I should
take. :)

Bart