Re: [PATCH v1 1/1] at24: Drop of_match_ptr() and ACPI_PTR() protections

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Tue, Feb 25, 2025 at 06:21:29PM +0100, Krzysztof Kozlowski wrote:
> On 25/02/2025 11:29, Arnd Bergmann wrote:
> > On Tue, Feb 25, 2025, at 11:08, Andy Shevchenko wrote:
> >> These result in a very small reduction in driver size, but at the cost
> >> of more complex build and slightly harder to read code. In the case of
> >> of_match_ptr() it also prevents use of PRP0001 ACPI based identification.
> >> In this particular case we have a valid ACPI/PNP ID that should be used
> >> in preference to PRP0001 but doesn't mean we should prevent that route.
> >>
> >> With this done, drop unneeded of*.h inclusions and __maybe_unused markers.
> >>
> >> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > 
> > Acked-by: Arnd Bergmann <arnd@arndb.de>
> > 
> > For reference, see below for a couple of patches in this area that
> > I have sent in the past. Ideally I think we should try to fix these
> > all up and enable -Wunused-const-variable, which is useful in its
> > own right.
> 
> I tried to fix this in SPI, regulator and ASoC 2 years ago and Mark
> rejected such approach of dropping ACPI/of_match_ptr. AFAIU, Mark wants
> this to be fixed in more generic way, on the OF and ACPI common code,
> not per driver.
> 
> SPI:
> https://lore.kernel.org/all/7a65d775-cf07-4393-8b10-2cef4d5266ab@sirena.org.uk/
> 
> regulator:
> https://lore.kernel.org/all/20230310214553.275450-1-krzysztof.kozlowski@linaro.org/
> 
> ASoC:
> https://lore.kernel.org/all/20230310214333.274903-1-krzysztof.kozlowski@linaro.org/

It was almost two years ago. Things may be changed :-)
At least I have no impediments so far with converting drivers I'm supporting in
the SPI. For ASoC there might be a new attempt by Cezary Rojewski in the future
(he does some cleanups in that area, and we discussed cleaning up ACPI_PTR() at
 minimum).

Also note

$ git grep -lw ACPI_PTR | wc -l
238

$ git grep -lw of_match_ptr | wc -l
841

So, at least dropping ACPI_PTR() seems on the track to getting rid of.
And I checked, they are spread all over the kernel with top subsystems as

  ...
  10 drivers/hwtracing/coresight
  11 drivers/input/touchscreen
  15 drivers/media/i2c
  53 sound/soc/codecs

-- 
With Best Regards,
Andy Shevchenko