Re: [PATCH] pinctrl: pinconf-generic: check error value EOPNOTSUPP

[Sudeep Holla <sudeep.holla@arm.com>]

On Fri, Apr 05, 2024 at 06:38:04PM +0300, Andy Shevchenko wrote:
> On Fri, Apr 05, 2024 at 02:13:28AM +0000, Peng Fan wrote:
> > > On Thu, Apr 04, 2024 at 01:44:50PM +0200, Linus Walleij wrote:
> > > > On Mon, Apr 1, 2024 at 4:02 PM Peng Fan (OSS) <peng.fan@oss.nxp.com>
> > > wrote:
>
> ...
>
> > > > >                         ret = pin_config_get_for_pin(pctldev, pin, &config);
> > > > >                 /* These are legal errors */
> > > > > -               if (ret == -EINVAL || ret == -ENOTSUPP)
> > > > > +               if (ret == -EINVAL || ret == -ENOTSUPP || ret ==
> > > > > + -EOPNOTSUPP)
> > > >
> > > > TBH it's a bit odd to call an in-kernel API such as
> > > > pin_config_get_for_pin() and get -EOPNOTSUPP back. But it's not like I care
> > > a lot, so patch applied.
> > >
> > > Hmm... I would like actually to get this being consistent. The documentation
> > > explicitly says that in-kernel APIs uses Linux error code and not POSIX one.
> >
> > Would you please share me the documentation?
>
> Sure.
> https://elixir.bootlin.com/linux/latest/source/include/linux/pinctrl/pinconf.h#L24
> https://elixir.bootlin.com/linux/latest/source/drivers/gpio/gpiolib.c#L2825
> https://elixir.bootlin.com/linux/latest/source/drivers/gpio/gpiolib.c#L2845
>
> I admit that this is not the best documented, feel free to produce a proper
> documentation.
>

Ah OK, my bad. I assumed you were referring to the entire kernel tree and
not just GPIO/pinux. Sorry for that.

> > > This check opens a Pandora box.
> > >
> > > FWIW, it just like dozen or so drivers that needs to be fixed, I prefer to have
> > > them being moved to ENOTSUPP, rather this patch.
> >
> > I see many patches convert to use EOPNOTSUPP by checking git log.
>
> How is that related? You mean for GPIO/pin control drivers?
>
> > And checkpatch.pl reports warning for using ENOTSUPP.
>
> checkpatch has false-positives, this is just one of them.
>

Fair enough.

> > BTW: is there any issue if using EOPNOTSUPP here?
>
> Yes. we don't want to be inconsistent. Using both in one subsystem is asking
> for troubles. If you want EOPNOTSUPP, please convert *all* users and drop
> ENOTSUPP completely (series out of ~100+ patches I believe :-), which probably
> will be not welcome).
>

Well, I don't agree with that 100% now since this is GPIO/pinmux sub-system
practice only. What if we change the source/root error cause(SCMI) in this
case and keep GPIO/pinmux happy today but tomorrow when this needs to be
used in some other subsystem which uses EOPNOTSUPP by default/consistently.
Now how do we address that then, hence I mentioned I am not 100% in agreement
now while I was before knowing that this is GPIO/pinmux strategy.

I don't know how to proceed now 🙁.

--
Regards,
Sudeep