[WIP] KnownLayout

TODO: Tests, especially for behavior of methods

Author: joshlf

src/lib.rs

@@ -183,17 +183,14 @@ use core::{
         NonZeroU16, NonZeroU32, NonZeroU64, NonZeroU8, NonZeroUsize, Wrapping,
     },
     ops::{Deref, DerefMut},
-    ptr, slice,
+    ptr::{self, NonNull},
+    slice,
 };
 
 #[cfg(feature = "alloc")]
 extern crate alloc;
 #[cfg(feature = "alloc")]
-use {
-    alloc::boxed::Box,
-    alloc::vec::Vec,
-    core::{alloc::Layout, ptr::NonNull},
-};
+use {alloc::boxed::Box, alloc::vec::Vec, core::alloc::Layout};
 
 // This is a hack to allow zerocopy-derive derives to work in this crate. They
 // assume that zerocopy is linked as an extern crate, so they access items from
@@ -203,6 +200,135 @@ mod zerocopy {
     pub(crate) use crate::*;
 }
 
+/// When performing a byte-slice-to-type cast, is the type taken from the prefix
+/// of the byte slice or from the suffix of the byte slice?
+#[doc(hidden)]
+#[allow(missing_debug_implementations, missing_copy_implementations)]
+pub enum CastType {
+    Prefix,
+    Suffix,
+}
+
+/// A trait which carries information about a type's layout that is used by the
+/// internals of this crate.
+///
+/// This trait is not meant for consumption by code outsie of this crate. While
+/// the normal semver stability guarantees apply with respect to which types
+/// implement this trait and which trait implementations are implied by this
+/// trait, no semver stability guarantees are made regarding its internals; they
+/// may change at any time, and code which makes use of them may break.
+///
+/// # Safety
+///
+/// This trait does not convey any safety guarantees to code outside this crate.
+pub unsafe trait KnownLayout: sealed::KnownLayoutSealed {
+    #[doc(hidden)]
+    const FIXED_PREFIX_SIZE: usize;
+    #[doc(hidden)]
+    const ALIGN: NonZeroUsize;
+    #[doc(hidden)]
+    const TRAILING_SLICE_ELEM_SIZE: Option<usize>;
+
+    /// SAFETY: The returned pointer has the same address and provenance as
+    /// `bytes`. If `Self` is a DST, the returned pointer's referent has `elems`
+    /// elements in its trailing slice.
+    #[doc(hidden)]
+    fn raw_from_ptr_len(bytes: NonNull<u8>, elems: usize) -> NonNull<Self>;
+}
+
+impl<T: KnownLayout> sealed::KnownLayoutSealed for [T] {}
+// SAFETY: See inline comments.
+unsafe impl<T: KnownLayout> KnownLayout for [T] {
+    // `[T]` is a slice type; it has no fields before the trailing slice.
+    const FIXED_PREFIX_SIZE: usize = 0;
+    // Slices have the same layout as the array they slice. [1] Arrays `[T; _]`
+    // have the same alignment as `T`. [2]
+    //
+    // [1] https://doc.rust-lang.org/reference/type-layout.html#slice-layout
+    // [2] https://doc.rust-lang.org/reference/type-layout.html#array-layout
+    const ALIGN: NonZeroUsize = if let Some(align) = NonZeroUsize::new(mem::align_of::<T>()) {
+        align
+    } else {
+        unreachable!()
+    };
+    const TRAILING_SLICE_ELEM_SIZE: Option<usize> = Some(mem::size_of::<T>());
+
+    // SAFETY: `.cast` preserves address and provenance. The returned pointer
+    // refers to an object with `elems` elements by construction.
+    #[inline(always)]
+    fn raw_from_ptr_len(data: NonNull<u8>, elems: usize) -> NonNull<Self> {
+        // TODO(#67): Remove this allow. See NonNullExt for more details.
+        #[allow(unstable_name_collisions)]
+        NonNull::slice_from_raw_parts(data.cast::<T>(), elems)
+    }
+}
+
+/// Implements `KnownLayout` for a sized type.
+macro_rules! impl_known_layout {
+    (const $constvar:ident : $constty:ty, $tyvar:ident $(: ?$optbound:ident)? => $ty:ty) => {
+        impl_known_layout!(@inner const $constvar: $constty, $tyvar $(: ?$optbound)? => $ty);
+    };
+    ($tyvar:ident $(: ?$optbound:ident)? => $ty:ty) => {
+        impl_known_layout!(@inner , $tyvar $(: ?$optbound)? => $ty);
+    };
+    ($ty:ty) => {
+        impl_known_layout!(@inner , => $ty);
+    };
+    ($($tyvar:ident $(: ?$optbound:ident)? => $ty:ty),*) => {
+        $(
+            impl_known_layout!(@inner , $tyvar $(: ?$optbound)? => $ty);
+        )*
+    };
+    ($($ty:ty),*) => {
+        $(
+            impl_known_layout!(@inner , => $ty);
+        )*
+    };
+    (@inner $(const $constvar:ident : $constty:ty)? , $($tyvar:ident $(: ?$optbound:ident)?)? => $ty:ty) => {
+        impl<$(const $constvar : $constty,)? $($tyvar $(: ?$optbound)?)?> sealed::KnownLayoutSealed for $ty {}
+        // SAFETY: See inline comments.
+        unsafe impl<$(const $constvar : $constty,)? $($tyvar $(: ?$optbound)?)?> KnownLayout for $ty {
+            const FIXED_PREFIX_SIZE: usize = mem::size_of::<$ty>();
+            const ALIGN: NonZeroUsize = if let Some(align) = NonZeroUsize::new(mem::align_of::<$ty>()) {
+                align
+            } else {
+                unreachable!()
+            };
+            // `T` is sized so it has no trailing slice.
+            const TRAILING_SLICE_ELEM_SIZE: Option<usize> = None;
+
+            // SAFETY: `.cast` preserves address and provenance.
+            #[inline(always)]
+            fn raw_from_ptr_len(bytes: NonNull<u8>, _elems: usize) -> NonNull<Self> {
+                bytes.cast::<Self>()
+            }
+        }
+    };
+}
+
+#[rustfmt::skip]
+impl_known_layout!(
+    (),
+    u8, i8, u16, i16, u32, i32, u64, i64, u128, i128, usize, isize, f32, f64,
+    bool, char,
+    NonZeroU8, NonZeroI8, NonZeroU16, NonZeroI16, NonZeroU32, NonZeroI32,
+    NonZeroU64, NonZeroI64, NonZeroU128, NonZeroI128, NonZeroUsize, NonZeroIsize
+);
+impl_known_layout!(T => Option<T>);
+impl_known_layout!(T: ?Sized => PhantomData<T>);
+impl_known_layout!(T => Wrapping<T>);
+impl_known_layout!(T => MaybeUninit<T>);
+impl_known_layout!(const N: usize, T => [T; N]);
+
+safety_comment! {
+    /// SAFETY:
+    /// `str` and `ManuallyDrop<[T]>` have the same representations as `[u8]`
+    /// and `[T]` repsectively. `str` has different bit validity than `[u8]`,
+    /// but that doesn't affect the soundness of this impl.
+    unsafe_impl_known_layout!(#[repr([u8])] str);
+    unsafe_impl_known_layout!(T: ?Sized + KnownLayout => #[repr(T)] ManuallyDrop<T>);
+}
+
 /// Types for which a sequence of bytes all set to zero represents a valid
 /// instance of the type.
 ///
@@ -1157,6 +1283,7 @@ mod simd {
                 use core::arch::$arch::{$($typ),*};
 
                 use crate::*;
+                impl_known_layout!($($typ),*);
                 safety_comment! {
                     /// SAFETY:
                     /// See comment on module definition for justification.
@@ -2261,7 +2388,8 @@ where
 }
 
 mod sealed {
-    pub trait Sealed {}
+    pub trait ByteSliceSealed {}
+    pub trait KnownLayoutSealed {}
 }
 
 // ByteSlice and ByteSliceMut abstract over [u8] references (&[u8], &mut [u8],
@@ -2287,7 +2415,9 @@ mod sealed {
 ///
 /// [`Vec<u8>`]: alloc::vec::Vec
 /// [`split_at`]: crate::ByteSlice::split_at
-pub unsafe trait ByteSlice: Deref<Target = [u8]> + Sized + self::sealed::Sealed {
+pub unsafe trait ByteSlice:
+    Deref<Target = [u8]> + Sized + self::sealed::ByteSliceSealed
+{
     /// Gets a raw pointer to the first byte in the slice.
     #[inline]
     fn as_ptr(&self) -> *const u8 {
@@ -2318,7 +2448,7 @@ pub unsafe trait ByteSliceMut: ByteSlice + DerefMut {
     }
 }
 
-impl<'a> sealed::Sealed for &'a [u8] {}
+impl<'a> sealed::ByteSliceSealed for &'a [u8] {}
 // TODO(#61): Add a "SAFETY" comment and remove this `allow`.
 #[allow(clippy::undocumented_unsafe_blocks)]
 unsafe impl<'a> ByteSlice for &'a [u8] {
@@ -2328,7 +2458,7 @@ unsafe impl<'a> ByteSlice for &'a [u8] {
     }
 }
 
-impl<'a> sealed::Sealed for &'a mut [u8] {}
+impl<'a> sealed::ByteSliceSealed for &'a mut [u8] {}
 // TODO(#61): Add a "SAFETY" comment and remove this `allow`.
 #[allow(clippy::undocumented_unsafe_blocks)]
 unsafe impl<'a> ByteSlice for &'a mut [u8] {
@@ -2338,7 +2468,7 @@ unsafe impl<'a> ByteSlice for &'a mut [u8] {
     }
 }
 
-impl<'a> sealed::Sealed for cell::Ref<'a, [u8]> {}
+impl<'a> sealed::ByteSliceSealed for cell::Ref<'a, [u8]> {}
 // TODO(#61): Add a "SAFETY" comment and remove this `allow`.
 #[allow(clippy::undocumented_unsafe_blocks)]
 unsafe impl<'a> ByteSlice for cell::Ref<'a, [u8]> {
@@ -2348,7 +2478,7 @@ unsafe impl<'a> ByteSlice for cell::Ref<'a, [u8]> {
     }
 }
 
-impl<'a> sealed::Sealed for RefMut<'a, [u8]> {}
+impl<'a> sealed::ByteSliceSealed for RefMut<'a, [u8]> {}
 // TODO(#61): Add a "SAFETY" comment and remove this `allow`.
 #[allow(clippy::undocumented_unsafe_blocks)]
 unsafe impl<'a> ByteSlice for RefMut<'a, [u8]> {
@@ -2366,6 +2496,63 @@ unsafe impl<'a> ByteSliceMut for &'a mut [u8] {}
 #[allow(clippy::undocumented_unsafe_blocks)]
 unsafe impl<'a> ByteSliceMut for RefMut<'a, [u8]> {}
 
+// A polyfill for `<*const _>::cast_mut` that we can use before our MSRV is
+// 1.65, when that method was stabilized.
+
+// TODO(#67): Once our MSRV is 1.65, remove this.
+trait RawPtrExt {
+    type Mut;
+    fn cast_mut(self) -> Self::Mut;
+}
+
+impl<T: ?Sized> RawPtrExt for *const T {
+    type Mut = *mut T;
+    #[allow(clippy::as_conversions)]
+    #[inline(always)]
+    fn cast_mut(self) -> *mut T {
+        self as *mut T
+    }
+}
+
+// A polyfill for `<*mut _>::cast_const` that we can use before our MSRV is
+// 1.65, when that method was stabilized.
+//
+// TODO(#67): Once our MSRV is 1.65, remove this.
+trait RawMutPtrExt {
+    type Const;
+    fn cast_const(self) -> Self::Const;
+}
+
+impl<T: ?Sized> RawMutPtrExt for *mut T {
+    type Const = *const T;
+    #[allow(clippy::as_conversions)]
+    #[inline(always)]
+    fn cast_const(self) -> *const T {
+        self as *const T
+    }
+}
+
+// A polyfill for `NonNull::slice_from_raw_parts` that we can use before our
+// MSRV is 1.70, when that function was stabilized.
+//
+// TODO(#67): Once our MSRV is 1.70, remove this.
+trait NonNullExt {
+    type SliceOfSelf;
+
+    fn slice_from_raw_parts(data: Self, len: usize) -> Self::SliceOfSelf;
+}
+
+impl<T> NonNullExt for NonNull<T> {
+    type SliceOfSelf = NonNull<[T]>;
+
+    #[inline(always)]
+    fn slice_from_raw_parts(data: Self, len: usize) -> NonNull<[T]> {
+        let ptr = ptr::slice_from_raw_parts_mut(data.as_ptr(), len);
+        // SAFETY: `ptr` is converted from `data`, which is non-null.
+        unsafe { NonNull::new_unchecked(ptr) }
+    }
+}
+
 #[cfg(feature = "alloc")]
 mod alloc_support {
     use alloc::vec::Vec;

src/macros.rs

@@ -190,6 +190,40 @@ macro_rules! impl_or_verify {
     };
 }
 
+/// Implements `KnownLayout` for a type in terms of the implementation of
+/// another type with the same representation.
+///
+/// # Safety
+///
+/// - `$ty` and `$repr` must have the same:
+///   - Fixed prefix size
+///   - Alignment
+///   - (For DSTs) trailing slice element size
+/// - It must be valid to perform an `as` cast from `*mut $repr` to `*mut $ty`,
+///   and this operation must preserve referent size (ie, `size_of_val_raw`).
+macro_rules! unsafe_impl_known_layout {
+    ($($tyvar:ident: ?Sized + KnownLayout =>)? #[repr($repr:ty)] $ty:ty) => {
+        impl<$($tyvar: ?Sized + KnownLayout)?> sealed::KnownLayoutSealed for $ty {}
+        unsafe impl<$($tyvar: ?Sized + KnownLayout)?> KnownLayout for $ty {
+            // SAFETY: Caller has promised that these values are the same for
+            // `$ty` and `$repr`.
+            const FIXED_PREFIX_SIZE: usize = <$repr as KnownLayout>::FIXED_PREFIX_SIZE;
+            const ALIGN: NonZeroUsize = <$repr as KnownLayout>::ALIGN;
+            const TRAILING_SLICE_ELEM_SIZE: Option<usize> = <$repr as KnownLayout>::TRAILING_SLICE_ELEM_SIZE;
+
+            // SAFETY: All operations preserve address and provenance. Caller
+            // has promised that the `as` cast preserves size.
+            #[inline(always)]
+            fn raw_from_ptr_len(bytes: NonNull<u8>, elems: usize) -> NonNull<Self> {
+                #[allow(clippy::as_conversions)]
+                let ptr = <$repr>::raw_from_ptr_len(bytes, elems).as_ptr() as *mut Self;
+                // SAFETY: `ptr` was converted from `bytes`, which is non-null.
+                unsafe { NonNull::new_unchecked(ptr) }
+            }
+        }
+    };
+}
+
 /// Uses `align_of` to confirm that a type or set of types have alignment 1.
 ///
 /// Note that `align_of<T>` requires `T: Sized`, so this macro doesn't work for