Re: [PATCH v3] rust: ACPI: fix missing match data for PRP0001

[Gary Guo]

On Fri Apr 10, 2026 at 3:57 PM BST, Markus Probst wrote:
> On Thu, 2026-04-09 at 00:07 +0200, Danilo Krummrich wrote:
>> On Wed Apr 8, 2026 at 11:42 PM CEST, Markus Probst wrote:
>> > On Wed, 2026-04-08 at 21:59 +0200, Danilo Krummrich wrote:
>> > > On Wed Apr 8, 2026 at 9:40 PM CEST, Markus Probst wrote:
>> > > > On Wed, 2026-04-08 at 18:03 +0200, Danilo Krummrich wrote:
>> > > > > On Tue Apr 7, 2026 at 11:41 PM CEST, Markus Probst wrote:
>> > > > > > diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
>> > > > > > index aad1a95e6863..d0098f24346f 100644
>> > > > > > --- a/include/acpi/acpi_bus.h
>> > > > > > +++ b/include/acpi/acpi_bus.h
>> > > > > > @@ -187,6 +187,10 @@ struct acpi_driver {
>> > > > > >   * -----------
>> > > > > >   */
>> > > > > >  
>> > > > > > +bool acpi_of_match_device(const struct acpi_device *adev,
>> > > > > > +			  const struct of_device_id *of_match_table,
>> > > > > > +			  const struct of_device_id **of_id);
>> > > > > 
>> > > > > This also has to be defined for !CONFIG_ACPI, otherwise we run into the
>> > > > > following compatible error.
>> > > > > 
>> > > > > 	error[E0425]: cannot find function `acpi_of_match_device` in crate `bindings`
>> > > > > 	      --> rust/kernel/driver.rs:295:24
>> > > > > 	       |
>> > > > > 	   295 |       unsafe { bindings::acpi_of_match_device(adev, of_match_table, of_id) }
>> > > > > 	       |                          ^^^^^^^^^^^^^^^^^^^^
>> > > > > 	       |
>> > > > > 	      ::: /mnt/nvme/work/projects/linux/driver-core/driver-core-testing/rust/bindings/bindings_generated.rs:118713:5
>> > > > > 
>> > > > > There is an
>> > > > > 
>> > > > > 	#else	/* CONFIG_ACPI */
>> > > > > 
>> > > > > block at the end of acpi_bus.h for this.
>> > > > I don't think the function exists in that case and bindgen can't
>> > > > generate inline functions, so I will just add a `#[cfg(CONFIG_ACPI)]`
>> > > > condition on top of the function.
>> > > 
>> > > Usually we provide a stub instead of conditionalize the callers; this case might
>> > > be a bit special, but I'd still follow the usual pattern.
>> > The usual C pattern would be
>> > 
>> > static inline bool acpi_of_match_device(const struct acpi_device *adev,
>> > 			const struct of_device_id *of_match_table,
>> > 			const struct of_device_id **of_id);
>> > 
>> > , which is ignored by bindgen (i. e. same error).
>> 
>> That's where you define a Rust helper; we have lots of cases where we only have
>> the Rust helper to deal with the inline function stub used when the
>> corresponding CONFIG_* is disabled.
> According to git commit msg [1], __rust_helper is needed to inline
> these helpers into Rust code. Does this mean we could omit the extra
> function declaration with #[inline(never)] and just use a helper
> without __rust_helper?
>
> [1]
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a9aabb3b839aba094ed80861054993785c61462c

The helpers can still be linked together with modules (especially when
CONFIG_RUST_INLINE_HELPERS=y), it's just that they're no longer inlined.

So you'll still need to make sure the function callers don't get codegen'ed into
modules, which means that you'll still need the `#[inline(never)]` trick.

Best,
Gary