Re: [PATCH v3 10/15] pinctrl: make struct pinfunction a pointer in struct function_desc

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

On Tue, Jul 29, 2025 at 1:06 PM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> On Tue, Jul 29, 2025 at 12:32 PM Andy Shevchenko
> <andy.shevchenko@gmail.com> wrote:
> > On Tue, Jul 29, 2025 at 11:27 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> > > On Sat, Jul 26, 2025 at 9:29 PM Andy Shevchenko
> > > <andy.shevchenko@gmail.com> wrote:
> > > > On Sat, Jul 26, 2025 at 1:31 PM kernel test robot <lkp@intel.com> wrote:

...

> > > > > All warnings (new ones prefixed by >>):
> > > > >
> > > > > >> drivers/pinctrl/pinmux.c:917:7: warning: cast from 'void (*)(const void *)' to 'void (*)(void *)' converts to incompatible function type [-Wcast-function-type-strict]
> > > > >      917 |                                          (void (*)(void *))kfree_const,
> > > > >          |                                          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > > >    include/linux/device/devres.h:166:34: note: expanded from macro 'devm_add_action_or_reset'
> > > > >      166 |         __devm_add_action_or_reset(dev, action, data, #action)
> > > > >          |                                         ^~~~~~
> > > > >    1 warning generated.
> > > >
> > > > Good catch, LKP!
> > >
> > > Ah, strict function type cast...

...

> > > > >    910          /*
> > > > >    911           * FIXME: It's generally a bad idea to use devres in subsystem core
> > > > >    912           * code - managed interfaces are aimed at drivers - but pinctrl already
> > > > >    913           * uses it all over the place so it's a larger piece of technical debt
> > > > >    914           * to fix.
> > > > >    915           */
> > > >
> > > > Indeed, this has to be addressed first, and IIRC I also got something
> > > > like this in the past.
> > >
> > > I mean, it doesn't have to be addressed first really, it's another big
> > > chunk of technical debt but we can live with it for now and address it
> > > in parallel.
> >
> > I don't know how you can solve this issue without addressing that
> > first. I do not consider "fixing" by casting out const specifier (the
> > whole point of which is to be not mangled).
>
> We can do nothing and keep on leaking this memory or we can
> exceptionally cast out the const and pass it down to
> devm_add_action_or_reset(). The scope of removing devres from pinctrl
> core is quite big and honestly this series is already big as is. I
> don't see it as a big issue. I can meet you in the middle and simply
> implement devm_kmemdup_const() manually to avoid casting out the const
> modifier.

It sounds like a good compromise to me. You will have my tag for that,
if it helps you.


-- 
With Best Regards,
Andy Shevchenko