Re: [PATCH v9 1/1] gpio: add sloppy logic analyzer using polling

[Bartosz Golaszewski <brgl@xxxxxxxx>]

On Thu, Jun 13, 2024 at 10:50 AM Arnd Bergmann <arnd@xxxxxxxx> wrote:
>
> On Thu, Jun 13, 2024, at 10:27, Linus Walleij wrote:
> > On Thu, Jun 13, 2024 at 10:17 AM Bartosz Golaszewski <brgl@xxxxxxxx> wrote:
> >
> >> On second thought
> >> though: are you sure drivers/gpio/ is the right place for it?
> >
> > Actually that is something I requested.
> >
> > I think it fits in drivers/gpio as it is such a clear cut usage of GPIO
> > lines, and it doesn't really fit into any other subsystem.
> >
> >> May I suggest moving it over to drivers/misc/?
> >
> > Misc is a bit...
> > messy. I remember Arnd being very sceptical about putting stuff there
> > rather than creating new subsystems, so since I've tried to avoid it,
> > albeit recently more and more stuff gets merged there again :/
>
> Right, and that is mostly to avoid having code in there because
> there is no other place for it. Some parts of drivers/misc should
> have been a separate subsystem, some should have use an existing
> subsystem, and other parts should have never been merged.
>
> The parts of drivers/misc that make the most sense to me are
> those that expose a one-of-a-kind piece of hardware as a
> single character device.
>
> This one would probably fit into drivers/misc/ better than
> some other drivers we have in there, but leaving it in
> drivers/gpio/ also seems fine.
>

This is my point. This really is a one-of-a-kind module that also
doesn't register with any particular subsystem. If anything fits into
drivers/misc/ then it's this.

To prove this point, I even moved the gpio-virtuser driver I'm working
on to drivers/misc/ too as it isn't a GPIO provider either and merely
a GPIO consumer with a one-shot user-space interface not conforming to
any standards.

> I could also imagine the functionality being exposed
> through drivers/iio/ in a way that is similar to an
> adc, but I don't know if that would work in practice or
> how much of a rewrite that would be.
>

I could see it using configfs instead of DT for configuration and iio
for presenting the output but - from what Wolfram said - insisting on
this will simply result in this development being dropped entirely.

Bart