On Thu, Apr 18, 2024 at 9:03 AM Benno Lossin <benno.lossin@xxxxxxxxx> wrote:
>
> On 18.04.24 10:59, Alice Ryhl wrote:
> > diff --git a/rust/kernel/types.rs b/rust/kernel/types.rs
> > index 8fad61268465..9c57c6c75553 100644
> > --- a/rust/kernel/types.rs
> > +++ b/rust/kernel/types.rs
> > @@ -409,3 +409,67 @@ pub enum Either<L, R> {
> > /// Constructs an instance of [`Either`] containing a value of type `R`.
> > Right(R),
> > }
> > +
> > +/// Types for which any bit pattern is valid.
> > +///
> > +/// Not all types are valid for all values. For example, a `bool` must be either zero or one, so
> > +/// reading arbitrary bytes into something that contains a `bool` is not okay.
> > +///
> > +/// It's okay for the type to have padding, as initializing those bytes has no effect.
> > +///
> > +/// # Safety
> > +///
> > +/// All bit-patterns must be valid for this type. This type must not have interior mutability.
>
> What is the reason for disallowing interior mutability here? I agree
> that it is necessary for `AsBytes`, but I don't think we need it here.
> For example it is fine to convert `u8` to `UnsafeCell<u8>`. Niches also
> should not be a problem, since eg `Option<UnsafeCell<NonNull<u8>>>`
> already fails the "All bit-patterns must be valid for this type".
>
> --
> Cheers,
> Benno
>
> > +pub unsafe trait FromBytes {}
See also the reasoning in zerocopy [1] and bytemuck [2]
[1]: https://docs.rs/zerocopy/latest/zerocopy/derive.FromBytes.html#analysis
[2]: https://docs.rs/bytemuck/latest/bytemuck/trait.AnyBitPattern.html#safety
