[PATCH v6 3/5] drm: handle HAS_IOPORT dependencies

[Lucas De Marchi]

On Mon, Oct 07, 2024 at 02:50:11PM +0000, Arnd Bergmann wrote:
>On Mon, Oct 7, 2024, at 14:39, Lucas De Marchi wrote:
>> as an example:
>> $ git grep -lw outb -- drivers/gpu/drm/
>> drivers/gpu/drm/gma500/cdv_device.c
>> drivers/gpu/drm/i915/display/intel_vga.c
>> drivers/gpu/drm/qxl/qxl_cmd.c
>> drivers/gpu/drm/qxl/qxl_irq.c
>> drivers/gpu/drm/tiny/bochs.c
>> drivers/gpu/drm/tiny/cirrus.c
>>
>> you are adding the dependency on xe, but why are you keeping i915 out?
>> What approach did you use to determine the dependency?
>
>I did a lot of 'randconfig' build testing on earlier versions
>(and this version) of the series, which eventually catches
>all of them. The i915 driver depends on CONfIG_X86 since it
>is only used in Intel PC chipsets that already rely on PIO.
>
>The XE driver is also used for add-on cards, so the drivers
>can be built on all architectures including those that do
>not support PCI I/O space access. Adding the dependency on
>i915 as well wouldn't be wrong, but is not required for
>correctness.

ok, makes sense. I missed the indirect dependency already present in
i915. Thanks.

For the xe part,

Acked-by: Lucas De Marchi <lucas.demarchi at intel.com>

Lucas De Marchi

>
>I also sent a patch for vmwgfx, which can be used on arm64.
>arm64 currently always sets HAS_IOPORT, so my patch is not
>required as a dependency for [PATCH v6 5/5], but we eventually
>want this so HAS_IOPORT can become optional on arm64.
>
>      Arnd