Re: [PATCH v4 2/3] driver core: make software nodes available earlier

[Arnd Bergmann]

On Tue, Mar 31, 2026, at 13:09, Andy Shevchenko wrote:
> On Tue, Mar 31, 2026 at 12:45:59PM +0200, Arnd Bergmann wrote:
>> On Tue, Mar 31, 2026, at 10:55, Andy Shevchenko wrote:
>>
>> My best guess is that __exit_call should just use
>> __attribute__((unused)) instead of __attribute__((used)) and
>> have the compiler drop it from built-in code instead of the linker:
>
> But why do we need that at all? Can we just drop the full section for good.
> Or i.o.w. where exactly is it being used in the current kernel?

The main use case I see for __exit_call is for device drivers
that can be either loadable modules or built-in: In a loadable module,
you need the cleanup function, but for built-in code you want
it to be dropped at build time.

For the few direct callers of __exitcall, the only explanation I have
is that these could at some point be turned into loadable modules,
and the author wanted to be prepared for changing it later. In theory
we could change those to module_exit(), or just remove them, but
there is little practical benefit either way.

I've tested the __maybe_unused change on randconfig builds now and
found no (build-time) issues with that, so I'm submitting that
now.

     Arnd