Re: [PATCH 2/2] rust: pin-init: add pin projections to `#[pin_data]`

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

On Wed Sep 10, 2025 at 12:23 PM CEST, Alice Ryhl wrote:
> On Fri, Sep 5, 2025 at 7:12 PM Benno Lossin <lossin@kernel.org> wrote:
>>
>> Make the `#[pin_data]` macro generate a `*Projection` struct that holds
>> either `Pin<&mut Field>` or `&mut Field` for every field of the original
>> struct. Which version is chosen depends on weather there is a `#[pin]`
>> or not respectively. Access to this projected version is enabled through
>> generating `fn project(self: Pin<&mut Self>) -> SelfProjection<'_>`.
>>
>> Link: https://github.com/Rust-for-Linux/pin-init/pull/75/commits/2d698367d646c7ede90e01aa22842c1002d017b3
>> [ Adapt workqueue to use the new projection instead of its own, custom
>>   one - Benno ]
>> Signed-off-by: Benno Lossin <lossin@kernel.org>
>> ---
>>  rust/kernel/workqueue.rs    | 10 ++-----
>>  rust/pin-init/src/macros.rs | 60 +++++++++++++++++++++++++++++++++++++
>>  2 files changed, 62 insertions(+), 8 deletions(-)
>>
>> diff --git a/rust/kernel/workqueue.rs b/rust/kernel/workqueue.rs
>> index b9343d5bc00f..6ca14c629643 100644
>> --- a/rust/kernel/workqueue.rs
>> +++ b/rust/kernel/workqueue.rs
>> @@ -356,18 +356,12 @@ struct ClosureWork<T> {
>>      func: Option<T>,
>>  }
>>
>> -impl<T> ClosureWork<T> {
>> -    fn project(self: Pin<&mut Self>) -> &mut Option<T> {
>> -        // SAFETY: The `func` field is not structurally pinned.
>> -        unsafe { &mut self.get_unchecked_mut().func }
>> -    }
>> -}
>> -
>>  impl<T: FnOnce()> WorkItem for ClosureWork<T> {
>>      type Pointer = Pin<KBox<Self>>;
>>
>>      fn run(mut this: Pin<KBox<Self>>) {
>> -        if let Some(func) = this.as_mut().project().take() {
>> +        if let Some(func) = this.as_mut().project().func.take() {
>> +            // if let Some(func) = this.as_mut().project_func().take() {
>>              (func)()
>>          }
>>      }
>> diff --git a/rust/pin-init/src/macros.rs b/rust/pin-init/src/macros.rs
>> index 9ced630737b8..d225cc144904 100644
>> --- a/rust/pin-init/src/macros.rs
>> +++ b/rust/pin-init/src/macros.rs
>> @@ -831,6 +831,17 @@ macro_rules! __pin_data {
>>              $($fields)*
>>          }
>>
>> +        $crate::__pin_data!(make_pin_projections:
>> +            @vis($vis),
>> +            @name($name),
>> +            @impl_generics($($impl_generics)*),
>> +            @ty_generics($($ty_generics)*),
>> +            @decl_generics($($decl_generics)*),
>> +            @where($($whr)*),
>> +            @pinned($($pinned)*),
>> +            @not_pinned($($not_pinned)*),
>> +        );
>> +
>>          // We put the rest into this const item, because it then will not be accessible to anything
>>          // outside.
>>          const _: () = {
>> @@ -980,6 +991,55 @@ fn drop(&mut self) {
>>              stringify!($($rest)*),
>>          );
>>      };
>> +    (make_pin_projections:
>> +        @vis($vis:vis),
>> +        @name($name:ident),
>> +        @impl_generics($($impl_generics:tt)*),
>> +        @ty_generics($($ty_generics:tt)*),
>> +        @decl_generics($($decl_generics:tt)*),
>> +        @where($($whr:tt)*),
>> +        @pinned($($(#[$($p_attr:tt)*])* $pvis:vis $p_field:ident : $p_type:ty),* $(,)?),
>> +        @not_pinned($($(#[$($attr:tt)*])* $fvis:vis $field:ident : $type:ty),* $(,)?),
>> +    ) => {
>> +        $crate::macros::paste! {
>> +            #[doc(hidden)]
>> +            $vis struct [< $name Projection >] <'__pin, $($decl_generics)*> {
>
> I'm not sure we want $vis here. That's the visibility of the original
> struct, but I don't think we want it to be pub just because the struct
> is.

Why shouldn't it be pub if the original is pub? I don't really
understand the concern, since the fields themselves will still have the
correct visibility. Additionally, there is the `___pin_phantom_data`
field that's always private, so you cannot construct this outside of the
module.

---
Cheers,
Benno