Re: [PATCH v3 08/13] rust: init: add `stack_pin_init!` macro

[Alice Ryhl]

On 3/30/23 13:06, Gary Guo wrote:
> > +impl<T> StackInit<T> {
> > +    /// Creates a new [`StackInit<T>`] that is uninitialized. Use [`stack_pin_init`] instead of this
> > +    /// primitive.
> > +    ///
> > +    /// [`stack_pin_init`]: kernel::stack_pin_init
> > +    #[inline]
> > +    pub fn uninit() -> Self {
> > +        Self(MaybeUninit::uninit(), false)
> > +    }
> > +
> > +    /// Initializes the contents and returns the result.
> > +    ///
> > +    /// # Safety
> > +    ///
> > +    /// The caller ensures that `self` is on the stack and not accessible in any other way, if this
> > +    /// function returns `Ok`.
> > +    #[inline]
> > +    pub unsafe fn init<E>(&mut self, init: impl PinInit<T, E>) -> Result<Pin<&mut T>, E> {
>
> Could this be made safe if the signature takes `self: Pin<&mut Self>`
> instead?
>
> The std `pin!` macro is stable in
> 1.68 so we can just `core::pin::pin!(StackInit::uninit())` and then
> call `init` on it.
>
> Best,
> Gary

Yeah, I think that would work. If it's marked safe, then it will be 
possible to call `init` several times, but this is fine if `init` 
transitions the `StackInit` back into its uninitialized state before 
attempting to initialize it again.