Re: [PATCH v2 3/8] rust: sync: add `Arc::{from_raw, into_raw}`

From: Andreas Hindborg (Samsung)
Date: Mon Jun 05 2023 - 14:47:50 EST



Boqun Feng <boqun.feng@xxxxxxxxx> writes:

> On Mon, Jun 05, 2023 at 03:31:42PM +0100, Gary Guo wrote:
>> On Fri, 02 Jun 2023 12:51:08 +0200
>> "Andreas Hindborg (Samsung)" <nmi@xxxxxxxxxxxx> wrote:
>>
>> > Alice Ryhl <aliceryhl@xxxxxxxxxx> writes:
>> >
>> > > From: Wedson Almeida Filho <walmeida@xxxxxxxxxxxxx>
>> > >
>> > > These methods can be used to turn an `Arc` into a raw pointer and back,
>> > > in a way that preserves the metadata for fat pointers.
>> > >
>> > > This is done using the unstable ptr_metadata feature [1]. However, it
>> > > could also be done using the unstable pointer_byte_offsets feature [2],
>> > > which is likely to have a shorter path to stabilization than
>> > > ptr_metadata.
>> > >
>> > > Link: https://github.com/rust-lang/rust/issues/81513 [1]
>> > > Link: https://github.com/rust-lang/rust/issues/96283 [2]
>> > > Signed-off-by: Wedson Almeida Filho <walmeida@xxxxxxxxxxxxx>
>> > > Co-developed-by: Alice Ryhl <aliceryhl@xxxxxxxxxx>
>> > > Signed-off-by: Alice Ryhl <aliceryhl@xxxxxxxxxx>
>> > > Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@xxxxxxxxx>
>> > > ---
>> > > rust/kernel/lib.rs | 1 +
>> > > rust/kernel/sync/arc.rs | 42 ++++++++++++++++++++++++++++++++++++++++-
>> > > 2 files changed, 42 insertions(+), 1 deletion(-)
>> > >
>> > > diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
>> > > index 7ea777b731e6..ad9142928fb1 100644
>> > > --- a/rust/kernel/lib.rs
>> > > +++ b/rust/kernel/lib.rs
>> > > @@ -17,6 +17,7 @@
>> > > #![feature(const_refs_to_cell)]
>> > > #![feature(dispatch_from_dyn)]
>> > > #![feature(new_uninit)]
>> > > +#![feature(ptr_metadata)]
>> > > #![feature(receiver_trait)]
>> > > #![feature(unsize)]
>> > >
>> > > diff --git a/rust/kernel/sync/arc.rs b/rust/kernel/sync/arc.rs
>> > > index a89843cacaad..684be9f73aca 100644
>> > > --- a/rust/kernel/sync/arc.rs
>> > > +++ b/rust/kernel/sync/arc.rs
>> > > @@ -24,7 +24,7 @@
>> > > };
>> > > use alloc::boxed::Box;
>> > > use core::{
>> > > - alloc::AllocError,
>> > > + alloc::{AllocError, Layout},
>> > > fmt,
>> > > marker::{PhantomData, Unsize},
>> > > mem::{ManuallyDrop, MaybeUninit},
>> > > @@ -212,6 +212,46 @@ unsafe fn from_inner(inner: NonNull<ArcInner<T>>) -> Self {
>> > > }
>> > > }
>> > >
>> > > + /// Convert the [`Arc`] into a raw pointer.
>> > > + ///
>> > > + /// The raw pointer has ownership of the refcount that this Arc object owned.
>> > > + pub fn into_raw(self) -> *const T {
>> > > + let ptr = self.ptr.as_ptr();
>> > > + core::mem::forget(self);
>> > > + // SAFETY: The pointer is valid.
>> > > + unsafe { core::ptr::addr_of!((*ptr).data) }
>> > > + }
>> > > +
>> > > + /// Recreates an [`Arc`] instance previously deconstructed via [`Arc::into_raw`].
>> > > + ///
>> > > + /// This code relies on the `repr(C)` layout of structs as described in
>> > > + /// <https://doc.rust-lang.org/reference/type-layout.html#reprc-structs>.
>> > > + ///
>> > > + /// # Safety
>> > > + ///
>> > > + /// `ptr` must have been returned by a previous call to [`Arc::into_raw`]. Additionally, it
>> > > + /// can only be called once for each previous call to [`Arc::into_raw`].
>> > > + pub unsafe fn from_raw(ptr: *const T) -> Self {
>> > > + let refcount_layout = Layout::new::<bindings::refcount_t>();
>> > > + // SAFETY: The caller guarantees that the pointer is valid.
>> > > + let val_layout = unsafe { Layout::for_value(&*ptr) };
>> > > + // SAFETY: We're computing the layout of a real struct that existed when compiling this
>> > > + // binary, so its layout is not so large that it can trigger arithmetic overflow.
>> > > + let val_offset = unsafe { refcount_layout.extend(val_layout).unwrap_unchecked().1 };
>> > > +
>> > > + // This preserves the metadata in the pointer, if any.
>> > > + //
>> > > + // Note that `*const T` and `*const ArcInner<T>` have the same metadata as documented at
>> > > + // <https://doc.rust-lang.org/std/ptr/trait.Pointee.html>.
>> > > + let metadata = core::ptr::metadata(ptr as *const ArcInner<T>);
>> >
>> > Thanks for updating the comment with the link. I looked into this and I
>> > find that what we are doing here, even though it works, does not feel
>> > right at all. We should be able to do this:
>> >
>> > let metadata = core::ptr::metadata(ptr);
>> > let ptr = (ptr as *mut u8).wrapping_sub(val_offset) as *mut ();
>> > let ptr = core::ptr::from_raw_parts_mut(ptr, metadata);
>> >
>> > but the way `Pointee::Metadata` is defined will not allow this, even
>> > though we know it is valid. I would suggest the following instead:
>> >
>> > let metadata = core::ptr::metadata(ptr);
>> > // Convert <T as Pointee>::Metadata to <ArcInner<T> as
>> > // Pointee>::Metadata. We know they have identical representation and thus this is OK.
>> > let metadata: <ArcInner<T> as Pointee>::Metadata = *unsafe {
>> > &*((&metadata as *const <T as Pointee>::Metadata as *const ())
>> > as *const <ArcInner<T> as Pointee>::Metadata)
>> > };
>>
>> This could just be a `transmute_copy`.
>>
>
> Or just `transmute`:
>
> let metadata = unsafe {
> core::mem::transmute<_, <ArcInner<T> as
> Pointee>>::Metadata>(metadata)
> };
>
> ? Since `Pointee::Metadata` is `Copy`.

I like `transmute_copy()` better for being more explicit.

>
> Regards,
> Boqun
>
>> > let ptr = (ptr as *mut u8).wrapping_sub(val_offset) as *mut ();
>> > let ptr = core::ptr::from_raw_parts_mut(ptr, metadata);
>> >
>> > Even though it is a bit more complex, it captures what we are trying to
>> > do better.
>>
>> I agree this captures the semantics better.
>>
>
> I actually wish that we could use `wrapping_byte_offset`[1], and just
>
> // `*const T` and `*const ArcInner<T>` should have the same
> // metdata, so convert the pointer type first.
> let ptr = ptr as *const ArcInner<T>;
>
> // .. and then adjust the byte offset.
> let ptr = ptr.wrapping_byte_offset(-val_offset);
>
> This may be the opposite direction as Andreas proposed ;-), but the
> result is less code.

Idk, it is sort of similar to the original approach. I think it is a
good idea to be a bit explicit about what we are doing here.

BR Andreas