Re: [PATCH] drm/xe: Fix build with DRM_XE=y and CONFIGFS_FS=m

[Jani Nikula <jani.nikula@linux.intel.com>]

On Mon, 19 Oct 2026, Matt Roper <matthew.d.roper@intel.com> wrote:
> On Mon, Oct 19, 2026 at 06:21:29PM +0300, Jani Nikula wrote:
>> On Fri, 16 Oct 2026, Rodrigo Vivi <rodrigo.vivi@intel.com> wrote:
>> > When xe is built-in but configfs is a module, xe_configfs.o is not
>> > linked into xe, while IS_ENABLED(CONFIG_CONFIGFS_FS) in xe_configfs.h
>> > still exposes the real declarations, leading to undefined references.
>> >
>> > Use IS_REACHABLE() so this case falls back to the stubs, the same as
>> > CONFIGFS_FS=n.
>> 
>> I think using IS_REACHABLE() is almost always a mistake. It's a hack to
>> avoid link errors when something in a module needs something that is
>> built-in. It's a combo that should be handled at Kconfig level, not at
>> build level.
>
> But isn't this situation exactly the case that it's intended for?  Xe
> doesn't "need" configfs, but will use it if it's present and usable.  In
> cases where configfs is built as a module, then it will not be usable by
> any built-in driver (including Xe), and that's perfectly fine; we just
> treat that the same as the 'n' case.

Yeah, but there's zero discoverability for the user, or the developer
looking at a future bug report, with IS_REACHABLE().

The user enables two configs, the Kconfig doesn't enforce a relationship
between the two, but they still fail to work together. How is the user
supposed to know?

With the proxy kconfig option you at least get a hint at what's
required. And you could make it more comprehensible with e.g. something
like this:

config DRM_XE_CONFIGFS
	def_bool y
	depends on DRM_XE && CONFIGFS_FS
	depends on CONFIGFS_FS=y || DRM_XE=m

Anyway, that's just my opinion. Obviously using IS_REACHABLE() is
"simpler" from Kconfig perspective, but I think explicit is better than
implicit.


BR,
Jani.


-- 
Jani Nikula, Intel