Re: [PATCH 6.12.y] rust: init: allow `dead_code` warnings for Rust >= 1.89.0

["Benno Lossin" <lossin@kernel.org>]

On Sat Jul 12, 2025 at 7:10 PM CEST, Miguel Ojeda wrote:
> Starting with Rust 1.89.0 (expected 2025-08-07), the Rust compiler
> may warn:
>
>     error: trait `MustNotImplDrop` is never used
>        --> rust/kernel/init/macros.rs:927:15
>         |
>     927 |         trait MustNotImplDrop {}
>         |               ^^^^^^^^^^^^^^^
>         |
>        ::: rust/kernel/sync/arc.rs:133:1
>         |
>     133 | #[pin_data]
>         | ----------- in this procedural macro expansion
>         |
>         = note: `-D dead-code` implied by `-D warnings`
>         = help: to override `-D warnings` add `#[allow(dead_code)]`
>         = note: this error originates in the macro `$crate::__pin_data`
>                 which comes from the expansion of the attribute macro
>                 `pin_data` (in Nightly builds, run with
>                 -Z macro-backtrace for more info)
>
> Thus `allow` it to clean it up.

This is a bit strange, I can't directly reproduce the issue... I already
get this warning in 1.88:

    https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=465f71a848e77ac3f7a96a0af6bc9e2a

> This does not happen in mainline nor 6.15.y, because there the macro was
> moved out of the `kernel` crate, and `dead_code` warnings are not
> emitted if the macro is foreign to the crate. Thus this patch is
> directly sent to stable and intended for 6.12.y only.
>
> Similarly, it is not needed in previous LTSs, because there the Rust
> version is pinned.
>
> Cc: Benno Lossin <lossin@kernel.org>
> Signed-off-by: Miguel Ojeda <ojeda@kernel.org>

Anyways the patch itself looks fine, nobody should care about the
dead-code warning (since it is in fact used to prevent `Drop` being
implemented).

Acked-by: Benno Lossin <lossin@kernel.org>

---
Cheers,
Benno

> ---
> Greg, Sasha: please note that an equivalent patch is _not_ in mainline.
>
> We could put these `allow`s in mainline (they wouldn't hurt), but it
> isn't a good idea to add things in mainline for the only reason of
> backporting them, thus I am sending this directly to stable.
>
> The patch is pretty safe -- there is no actual code change.
>
>  rust/kernel/init/macros.rs | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/rust/kernel/init/macros.rs b/rust/kernel/init/macros.rs
> index b7213962a6a5..e530028bb9ed 100644
> --- a/rust/kernel/init/macros.rs
> +++ b/rust/kernel/init/macros.rs
> @@ -924,6 +924,7 @@ impl<'__pin, $($impl_generics)*> ::core::marker::Unpin for $name<$($ty_generics)
>          // We prevent this by creating a trait that will be implemented for all types implementing
>          // `Drop`. Additionally we will implement this trait for the struct leading to a conflict,
>          // if it also implements `Drop`
> +        #[allow(dead_code)]
>          trait MustNotImplDrop {}
>          #[expect(drop_bounds)]
>          impl<T: ::core::ops::Drop> MustNotImplDrop for T {}
> @@ -932,6 +933,7 @@ impl<$($impl_generics)*> MustNotImplDrop for $name<$($ty_generics)*>
>          // We also take care to prevent users from writing a useless `PinnedDrop` implementation.
>          // They might implement `PinnedDrop` correctly for the struct, but forget to give
>          // `PinnedDrop` as the parameter to `#[pin_data]`.
> +        #[allow(dead_code)]
>          #[expect(non_camel_case_types)]
>          trait UselessPinnedDropImpl_you_need_to_specify_PinnedDrop {}
>          impl<T: $crate::init::PinnedDrop>
>
> base-commit: fbad404f04d758c52bae79ca20d0e7fe5fef91d3
> --
> 2.50.1