Re: [PATCH v2 2/9] rust: list: add tracking for ListArc

[Benno Lossin <benno.lossin@proton.me>]

On 06.05.24 11:53, Alice Ryhl wrote:
> @@ -32,9 +33,24 @@ pub trait ListArcSafe<const ID: u64 = 0> {
>     unsafe fn on_drop_list_arc(&self);
>  }
> 
> +/// Declares that this type is able to safely attempt to create `ListArc`s at any time.
> +///
> +/// # Safety
> +///
> +/// Implementers must ensure that `try_new_list_arc` does not return `true` if a `ListArc` already
> +/// exists.
> +pub unsafe trait TryNewListArc<const ID: u64 = 0>: ListArcSafe<ID> {
> +    /// Attempts to convert an `Arc<Self>` into an `ListArc<Self>`. Returns `true` if the
> +    /// conversion was successful.
> +    fn try_new_list_arc(&self) -> bool;
> +}
> +
>  /// Declares that this type supports [`ListArc`].
>  ///
> -/// When using this macro, it will only be possible to create a [`ListArc`] from a [`UniqueArc`].
> +/// When using this macro, you may choose between the `untracked` strategy where it is not tracked
> +/// whether a [`ListArc`] exists, and the `tracked_by` strategy where the tracking is deferred to a
> +/// field of the struct. The `tracked_by` strategy can be combined with a field of type
> +/// [`AtomicListArcTracker`] to track whether a [`ListArc`] exists.

I think it would make sense to use bullet points here.
Also, you should mention that in the `tracked_by` strategy, the field is
required to implement `TryNewListArc`.

>  #[macro_export]
>  macro_rules! impl_list_arc_safe {
>     (impl$({$($generics:tt)*})? ListArcSafe<$num:tt> for $t:ty { untracked; } $($rest:tt)*) => {
> @@ -45,6 +61,37 @@ unsafe fn on_drop_list_arc(&self) {}
>         $crate::list::impl_list_arc_safe! { $($rest)* }
>     };
> 
> +    (impl$({$($generics:tt)*})? ListArcSafe<$num:tt> for $t:ty {
> +        tracked_by $field:ident : $fty:ty;
> +    } $($rest:tt)*) => {
> +        impl$(<$($generics)*>)? $crate::list::ListArcSafe<$num> for $t {
> +            unsafe fn on_create_list_arc_from_unique(self: ::core::pin::Pin<&mut Self>) {
> +                // SAFETY: This field is structurally pinned.

Who ensures this? This is not documented on the macro.
The only way that I see to fix this would be to make the `tracked_by`
strategy `unsafe`. At least until we implement proper structural pinning
of fields.

> +                let field = unsafe {
> +                    ::core::pin::Pin::map_unchecked_mut(self, |me| &mut me.$field)
> +                };
> +                // SAFETY: The caller promises that there is no `ListArc`.
> +                unsafe {
> +                    <$fty as $crate::list::ListArcSafe<$num>>::on_create_list_arc_from_unique(field)
> +                };
> +            }
> +            unsafe fn on_drop_list_arc(&self) {
> +                // SAFETY: The caller promises that there is no `ListArc` reference, and also
> +                // promises that the tracking thinks there is a `ListArc` reference.
> +                unsafe { <$fty as $crate::list::ListArcSafe<$num>>::on_drop_list_arc(&self.$field) };
> +            }
> +        }
> +        unsafe impl$(<$($generics)*>)? $crate::list::TryNewListArc<$num> for $t
> +        where
> +            $fty: TryNewListArc<$num>,
> +        {
> +            fn try_new_list_arc(&self) -> bool {
> +                <$fty as $crate::list::TryNewListArc<$num>>::try_new_list_arc(&self.field)
> +            }
> +        }
> +        $crate::list::impl_list_arc_safe! { $($rest)* }
> +    };
> +
>     () => {};
>  }
>  pub use impl_list_arc_safe;

[...]

> @@ -313,3 +406,60 @@ impl<T, U, const ID: u64> core::ops::DispatchFromDyn<ListArc<U, ID>> for ListArc
>     U: ListArcSafe<ID> + ?Sized,
>  {
>  }
> +
> +/// A utility for tracking whether a [`ListArc`] exists using an atomic.
> +///
> +/// # Invariant
> +///
> +/// If the boolean is `false`, then there is no [`ListArc`] for this value.

"If `inner` is `false`, ..."

---
Cheers,
Benno

> +#[repr(transparent)]
> +pub struct AtomicListArcTracker<const ID: u64 = 0> {
> +    inner: AtomicBool,
> +    _pin: PhantomPinned,
> +}

[...]