Re: [PATCH v3 2/6] ACPI / x86: Consolidate Apple DMI checks

[Lukas Wunner <lukas@wunner.de>]

On Sat, Jul 15, 2017 at 12:03:31AM +0200, Rafael J. Wysocki wrote:
> On Friday, July 14, 2017 12:36:19 AM Lukas Wunner wrote:
> > --- a/drivers/acpi/osi.c
> > +++ b/drivers/acpi/osi.c
> > @@ -258,12 +258,16 @@ bool acpi_osi_is_win8(void)
> >  EXPORT_SYMBOL(acpi_osi_is_win8);
> >  
> >  #ifdef CONFIG_X86
> > +bool is_apple_system;
> > +EXPORT_SYMBOL(is_apple_system);
> 
> Maybe prepend the name of this variable with acpi_ to indicate that this is
> ACPI-specific.

It's not really ACPI-specific, osi.c just happens to be the best place
to set the variable because the acpi_osi_dmi_table[] is checked very
early during boot.  So early in fact, that I could even replace the
existing Apple DMI check in arch/x86/kernel/early-quirks.c with
"is_apple_system".

These non-ACPI files currently contain an Apple DMI check:
arch/x86/kernel/early-quirks.c
drivers/pci/quirks.c (2x)
drivers/firmware/efi/apple-properties.c
drivers/thunderbolt/tb.c
drivers/thunderbolt/icm.c

The latter three do not #include <linux/acpi.h> yet.  Somehow it feels
odd to include that header to check for presence of an Apple system
because that's not really ACPI-related.

I guess I could introduce a new <linux/apple.h> but I hate the insane
proliferation of additional files in include/linux/.  I could merge
the contents of apple_bl.h and apple-gmux.h into that new header to
reduce the number of files a bit.

Struggling to find a solution that's nice and clean.  Any ideas?

Thanks,

Lukas