Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

1 files changed, 1825 insertions, 0 deletions

diff --git a/rust/zerocopy/src/macros.rs b/rust/zerocopy/src/macros.rs
new file mode 100644
index 000000000..ec67c03a4
--- /dev/null
+++ b/rust/zerocopy/src/macros.rs
@@ -0,0 +1,1825 @@
+// SPDX-License-Identifier: (BSD-2-Clause OR Apache-2.0) OR MIT
+//
+// Copyright 2024 The Fuchsia Authors
+//
+// Licensed under the 2-Clause BSD License <LICENSE-BSD or
+// https://opensource.org/license/bsd-2-clause>, Apache License, Version 2.0
+// <LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0>, or the MIT
+// license <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your option.
+// This file may not be copied, modified, or distributed except according to
+// those terms.
+
+/// Safely transmutes a value of one type to a value of another type of the same
+/// size.
+///
+/// This macro behaves like an invocation of this function:
+///
+/// ```ignore
+/// const fn transmute<Src, Dst>(src: Src) -> Dst
+/// where
+///     Src: IntoBytes,
+///     Dst: FromBytes,
+///     size_of::<Src>() == size_of::<Dst>(),
+/// {
+/// # /*
+///     ...
+/// # */
+/// }
+/// ```
+///
+/// However, unlike a function, this macro can only be invoked when the types of
+/// `Src` and `Dst` are completely concrete. The types `Src` and `Dst` are
+/// inferred from the calling context; they cannot be explicitly specified in
+/// the macro invocation.
+///
+/// Note that the `Src` produced by the expression `$e` will *not* be dropped.
+/// Semantically, its bits will be copied into a new value of type `Dst`, the
+/// original `Src` will be forgotten, and the value of type `Dst` will be
+/// returned.
+///
+/// # `#![allow(shrink)]`
+///
+/// If `#![allow(shrink)]` is provided, `transmute!` additionally supports
+/// transmutations that shrink the size of the value; e.g.:
+///
+/// ```
+/// # use zerocopy::transmute;
+/// let u: u32 = transmute!(#![allow(shrink)] 0u64);
+/// assert_eq!(u, 0u32);
+/// ```
+///
+/// # Examples
+///
+/// ```
+/// # use zerocopy::transmute;
+/// let one_dimensional: [u8; 8] = [0, 1, 2, 3, 4, 5, 6, 7];
+///
+/// let two_dimensional: [[u8; 4]; 2] = transmute!(one_dimensional);
+///
+/// assert_eq!(two_dimensional, [[0, 1, 2, 3], [4, 5, 6, 7]]);
+/// ```
+///
+/// # Use in `const` contexts
+///
+/// This macro can be invoked in `const` contexts.
+///
+#[doc = codegen_section!(
+    header = "h2",
+    bench = "transmute",
+    format = "coco_static_size",
+)]
+#[macro_export]
+macro_rules! transmute {
+    // NOTE: This must be a macro (rather than a function with trait bounds)
+    // because there's no way, in a generic context, to enforce that two types
+    // have the same size. `core::mem::transmute` uses compiler magic to enforce
+    // this so long as the types are concrete.
+    (#![allow(shrink)] $e:expr) => {{
+        let mut e = $e;
+        if false {
+            // This branch, though never taken, ensures that the type of `e` is
+            // `IntoBytes` and that the type of the  outer macro invocation
+            // expression is `FromBytes`.
+
+            fn transmute<Src, Dst>(src: Src) -> Dst
+            where
+                Src: $crate::IntoBytes,
+                Dst: $crate::FromBytes,
+            {
+                let _ = src;
+                loop {}
+            }
+            loop {}
+            #[allow(unreachable_code)]
+            transmute(e)
+        } else {
+            use $crate::util::macro_util::core_reexport::mem::ManuallyDrop;
+
+            // NOTE: `repr(packed)` is important! It ensures that the size of
+            // `Transmute` won't be rounded up to accommodate `Src`'s or `Dst`'s
+            // alignment, which would break the size comparison logic below.
+            //
+            // As an example of why this is problematic, consider `Src = [u8;
+            // 5]`, `Dst = u32`. The total size of `Transmute<Src, Dst>` would
+            // be 8, and so we would reject a `[u8; 5]` to `u32` transmute as
+            // being size-increasing, which it isn't.
+            #[repr(C, packed)]
+            union Transmute<Src, Dst> {
+                src: ManuallyDrop<Src>,
+                dst: ManuallyDrop<Dst>,
+            }
+
+            // SAFETY: `Transmute` is a `repr(C)` union whose `src` field has
+            // type `ManuallyDrop<Src>`. Thus, the `src` field starts at byte
+            // offset 0 within `Transmute` [1]. `ManuallyDrop<T>` has the same
+            // layout and bit validity as `T`, so it is sound to transmute `Src`
+            // to `Transmute`.
+            //
+            // [1] https://doc.rust-lang.org/1.85.0/reference/type-layout.html#reprc-unions
+            //
+            // [2] Per https://doc.rust-lang.org/1.85.0/std/mem/struct.ManuallyDrop.html:
+            //
+            //   `ManuallyDrop<T>` is guaranteed to have the same layout and bit
+            //   validity as `T`
+            let u: Transmute<_, _> = unsafe {
+                // Clippy: We can't annotate the types; this macro is designed
+                // to infer the types from the calling context.
+                #[allow(clippy::missing_transmute_annotations)]
+                $crate::util::macro_util::core_reexport::mem::transmute(e)
+            };
+
+            if false {
+                // SAFETY: This code is never executed.
+                e = ManuallyDrop::into_inner(unsafe { u.src });
+                // Suppress the `unused_assignments` lint on the previous line.
+                let _ = e;
+                loop {}
+            } else {
+                // SAFETY: Per the safety comment on `let u` above, the `dst`
+                // field in `Transmute` starts at byte offset 0, and has the
+                // same layout and bit validity as `Dst`.
+                //
+                // Transmuting `Src` to `Transmute<Src, Dst>` above using
+                // `core::mem::transmute` ensures that `size_of::<Src>() ==
+                // size_of::<Transmute<Src, Dst>>()`. A `#[repr(C, packed)]`
+                // union has the maximum size of all of its fields [1], so this
+                // is equivalent to `size_of::<Src>() >= size_of::<Dst>()`.
+                //
+                // The outer `if`'s `false` branch ensures that `Src: IntoBytes`
+                // and `Dst: FromBytes`. This, combined with the size bound,
+                // ensures that this transmute is sound.
+                //
+                // [1] Per https://doc.rust-lang.org/1.85.0/reference/type-layout.html#reprc-unions:
+                //
+                //   The union will have a size of the maximum size of all of
+                //   its fields rounded to its alignment
+                let dst = unsafe { u.dst };
+                $crate::util::macro_util::must_use(ManuallyDrop::into_inner(dst))
+            }
+        }
+    }};
+    ($e:expr) => {{
+        let e = $e;
+        if false {
+            // This branch, though never taken, ensures that the type of `e` is
+            // `IntoBytes` and that the type of the  outer macro invocation
+            // expression is `FromBytes`.
+
+            fn transmute<Src, Dst>(src: Src) -> Dst
+            where
+                Src: $crate::IntoBytes,
+                Dst: $crate::FromBytes,
+            {
+                let _ = src;
+                loop {}
+            }
+            loop {}
+            #[allow(unreachable_code)]
+            transmute(e)
+        } else {
+            // SAFETY: `core::mem::transmute` ensures that the type of `e` and
+            // the type of this macro invocation expression have the same size.
+            // We know this transmute is safe thanks to the `IntoBytes` and
+            // `FromBytes` bounds enforced by the `false` branch.
+            let u = unsafe {
+                // Clippy: We can't annotate the types; this macro is designed
+                // to infer the types from the calling context.
+                #[allow(clippy::missing_transmute_annotations, unnecessary_transmutes)]
+                $crate::util::macro_util::core_reexport::mem::transmute(e)
+            };
+            $crate::util::macro_util::must_use(u)
+        }
+    }};
+}
+
+/// Safely transmutes a mutable or immutable reference of one type to an
+/// immutable reference of another type of the same size and compatible
+/// alignment.
+///
+/// This macro behaves like an invocation of this function:
+///
+/// ```ignore
+/// fn transmute_ref<'src, 'dst, Src, Dst>(src: &'src Src) -> &'dst Dst
+/// where
+///     'src: 'dst,
+///     Src: IntoBytes + Immutable + ?Sized,
+///     Dst: FromBytes + Immutable + ?Sized,
+///     align_of::<Src>() >= align_of::<Dst>(),
+///     size_compatible::<Src, Dst>(),
+/// {
+/// # /*
+///     ...
+/// # */
+/// }
+/// ```
+///
+/// The types `Src` and `Dst` are inferred from the calling context; they cannot
+/// be explicitly specified in the macro invocation.
+///
+/// # Size compatibility
+///
+/// `transmute_ref!` supports transmuting between `Sized` types, between unsized
+/// (i.e., `?Sized`) types, and from a `Sized` type to an unsized type. It
+/// supports any transmutation that preserves the number of bytes of the
+/// referent, even if doing so requires updating the metadata stored in an
+/// unsized "fat" reference:
+///
+/// ```
+/// # use zerocopy::transmute_ref;
+/// # use core::mem::size_of_val; // Not in the prelude on our MSRV
+/// let src: &[[u8; 2]] = &[[0, 1], [2, 3]][..];
+/// let dst: &[u8] = transmute_ref!(src);
+///
+/// assert_eq!(src.len(), 2);
+/// assert_eq!(dst.len(), 4);
+/// assert_eq!(dst, [0, 1, 2, 3]);
+/// assert_eq!(size_of_val(src), size_of_val(dst));
+/// ```
+///
+/// # Errors
+///
+/// Violations of the alignment and size compatibility checks are detected
+/// *after* the compiler performs monomorphization. This has two important
+/// consequences.
+///
+/// First, it means that generic code will *never* fail these conditions:
+///
+/// ```
+/// # use zerocopy::{transmute_ref, FromBytes, IntoBytes, Immutable};
+/// fn transmute_ref<Src, Dst>(src: &Src) -> &Dst
+/// where
+///     Src: IntoBytes + Immutable,
+///     Dst: FromBytes + Immutable,
+/// {
+///     transmute_ref!(src)
+/// }
+/// ```
+///
+/// Instead, failures will only be detected once generic code is instantiated
+/// with concrete types:
+///
+/// ```compile_fail,E0080
+/// # use zerocopy::{transmute_ref, FromBytes, IntoBytes, Immutable};
+/// #
+/// # fn transmute_ref<Src, Dst>(src: &Src) -> &Dst
+/// # where
+/// #     Src: IntoBytes + Immutable,
+/// #     Dst: FromBytes + Immutable,
+/// # {
+/// #     transmute_ref!(src)
+/// # }
+/// let src: &u16 = &0;
+/// let dst: &u8 = transmute_ref(src);
+/// ```
+///
+/// Second, the fact that violations are detected after monomorphization means
+/// that `cargo check` will usually not detect errors, even when types are
+/// concrete. Instead, `cargo build` must be used to detect such errors.
+///
+/// # Examples
+///
+/// Transmuting between `Sized` types:
+///
+/// ```
+/// # use zerocopy::transmute_ref;
+/// let one_dimensional: [u8; 8] = [0, 1, 2, 3, 4, 5, 6, 7];
+///
+/// let two_dimensional: &[[u8; 4]; 2] = transmute_ref!(&one_dimensional);
+///
+/// assert_eq!(two_dimensional, &[[0, 1, 2, 3], [4, 5, 6, 7]]);
+/// ```
+///
+/// Transmuting between unsized types:
+///
+/// ```
+/// # use {zerocopy::*, zerocopy_derive::*};
+/// # type u16 = zerocopy::byteorder::native_endian::U16;
+/// # type u32 = zerocopy::byteorder::native_endian::U32;
+/// #[derive(KnownLayout, FromBytes, IntoBytes, Immutable)]
+/// #[repr(C)]
+/// struct SliceDst<T, U> {
+///     t: T,
+///     u: [U],
+/// }
+///
+/// type Src = SliceDst<u32, u16>;
+/// type Dst = SliceDst<u16, u8>;
+///
+/// let src = Src::ref_from_bytes(&[0, 1, 2, 3, 4, 5, 6, 7]).unwrap();
+/// let dst: &Dst = transmute_ref!(src);
+///
+/// assert_eq!(src.t.as_bytes(), [0, 1, 2, 3]);
+/// assert_eq!(src.u.len(), 2);
+/// assert_eq!(src.u.as_bytes(), [4, 5, 6, 7]);
+///
+/// assert_eq!(dst.t.as_bytes(), [0, 1]);
+/// assert_eq!(dst.u, [2, 3, 4, 5, 6, 7]);
+/// ```
+///
+/// # Use in `const` contexts
+///
+/// This macro can be invoked in `const` contexts only when `Src: Sized` and
+/// `Dst: Sized`.
+///
+#[doc = codegen_section!(
+    header = "h2",
+    bench = "transmute_ref",
+    format = "coco",
+    arity = 2,
+    [
+        open
+        @index 1
+        @title "Sized"
+        @variant "static_size"
+    ],
+    [
+        @index 2
+        @title "Unsized"
+        @variant "dynamic_size"
+    ]
+)]
+#[macro_export]
+macro_rules! transmute_ref {
+    ($e:expr) => {{
+        // NOTE: This must be a macro (rather than a function with trait bounds)
+        // because there's no way, in a generic context, to enforce that two
+        // types have the same size or alignment.
+
+        // Ensure that the source type is a reference or a mutable reference
+        // (note that mutable references are implicitly reborrowed here).
+        let e: &_ = $e;
+
+        #[allow(unused, clippy::diverging_sub_expression)]
+        if false {
+            // This branch, though never taken, ensures that the type of `e` is
+            // `&T` where `T: IntoBytes + Immutable`, and that the type of this
+            // macro expression is `&U` where `U: FromBytes + Immutable`.
+
+            struct AssertSrcIsIntoBytes<'a, T: ?::core::marker::Sized + $crate::IntoBytes>(&'a T);
+            struct AssertSrcIsImmutable<'a, T: ?::core::marker::Sized + $crate::Immutable>(&'a T);
+            struct AssertDstIsFromBytes<'a, U: ?::core::marker::Sized + $crate::FromBytes>(&'a U);
+            struct AssertDstIsImmutable<'a, T: ?::core::marker::Sized + $crate::Immutable>(&'a T);
+
+            let _ = AssertSrcIsIntoBytes(e);
+            let _ = AssertSrcIsImmutable(e);
+
+            if true {
+                #[allow(unused, unreachable_code)]
+                let u = AssertDstIsFromBytes(loop {});
+                u.0
+            } else {
+                #[allow(unused, unreachable_code)]
+                let u = AssertDstIsImmutable(loop {});
+                u.0
+            }
+        } else {
+            use $crate::util::macro_util::TransmuteRefDst;
+            let t = $crate::util::macro_util::Wrap::new(e);
+
+            if false {
+                // This branch exists solely to force the compiler to infer the
+                // type of `Dst` *before* it attempts to resolve the method call
+                // to `transmute_ref` in the `else` branch.
+                //
+                // Without this, if `Src` is `Sized` but `Dst` is `!Sized`, the
+                // compiler will eagerly select the inherent impl of
+                // `transmute_ref` (which requires `Dst: Sized`) because inherent
+                // methods take priority over trait methods. It does this before
+                // it realizes `Dst` is `!Sized`, leading to a compile error when
+                // it checks the bounds later.
+                //
+                // By calling this helper (which returns `&Dst`), we force `Dst`
+                // to be fully resolved. By the time it gets to the `else`
+                // branch, the compiler knows `Dst` is `!Sized`, properly
+                // disqualifies the inherent method, and falls back to the trait
+                // implementation.
+                t.transmute_ref_inference_helper()
+            } else {
+                // SAFETY: The outer `if false` branch ensures that:
+                // - `Src: IntoBytes + Immutable`
+                // - `Dst: FromBytes + Immutable`
+                unsafe {
+                    t.transmute_ref()
+                }
+            }
+        }
+    }}
+}
+
+/// Safely transmutes a mutable reference of one type to a mutable reference of
+/// another type of the same size and compatible alignment.
+///
+/// This macro behaves like an invocation of this function:
+///
+/// ```ignore
+/// const fn transmute_mut<'src, 'dst, Src, Dst>(src: &'src mut Src) -> &'dst mut Dst
+/// where
+///     'src: 'dst,
+///     Src: FromBytes + IntoBytes + ?Sized,
+///     Dst: FromBytes + IntoBytes + ?Sized,
+///     align_of::<Src>() >= align_of::<Dst>(),
+///     size_compatible::<Src, Dst>(),
+/// {
+/// # /*
+///     ...
+/// # */
+/// }
+/// ```
+///
+/// The types `Src` and `Dst` are inferred from the calling context; they cannot
+/// be explicitly specified in the macro invocation.
+///
+/// # Size compatibility
+///
+/// `transmute_mut!` supports transmuting between `Sized` types, between unsized
+/// (i.e., `?Sized`) types, and from a `Sized` type to an unsized type. It
+/// supports any transmutation that preserves the number of bytes of the
+/// referent, even if doing so requires updating the metadata stored in an
+/// unsized "fat" reference:
+///
+/// ```
+/// # use zerocopy::transmute_mut;
+/// # use core::mem::size_of_val; // Not in the prelude on our MSRV
+/// let src: &mut [[u8; 2]] = &mut [[0, 1], [2, 3]][..];
+/// let dst: &mut [u8] = transmute_mut!(src);
+///
+/// assert_eq!(dst.len(), 4);
+/// assert_eq!(dst, [0, 1, 2, 3]);
+/// let dst_size = size_of_val(dst);
+/// assert_eq!(src.len(), 2);
+/// assert_eq!(size_of_val(src), dst_size);
+/// ```
+///
+/// # Errors
+///
+/// Violations of the alignment and size compatibility checks are detected
+/// *after* the compiler performs monomorphization. This has two important
+/// consequences.
+///
+/// First, it means that generic code will *never* fail these conditions:
+///
+/// ```
+/// # use zerocopy::{transmute_mut, FromBytes, IntoBytes, Immutable};
+/// fn transmute_mut<Src, Dst>(src: &mut Src) -> &mut Dst
+/// where
+///     Src: FromBytes + IntoBytes,
+///     Dst: FromBytes + IntoBytes,
+/// {
+///     transmute_mut!(src)
+/// }
+/// ```
+///
+/// Instead, failures will only be detected once generic code is instantiated
+/// with concrete types:
+///
+/// ```compile_fail,E0080
+/// # use zerocopy::{transmute_mut, FromBytes, IntoBytes, Immutable};
+/// #
+/// # fn transmute_mut<Src, Dst>(src: &mut Src) -> &mut Dst
+/// # where
+/// #     Src: FromBytes + IntoBytes,
+/// #     Dst: FromBytes + IntoBytes,
+/// # {
+/// #     transmute_mut!(src)
+/// # }
+/// let src: &mut u16 = &mut 0;
+/// let dst: &mut u8 = transmute_mut(src);
+/// ```
+///
+/// Second, the fact that violations are detected after monomorphization means
+/// that `cargo check` will usually not detect errors, even when types are
+/// concrete. Instead, `cargo build` must be used to detect such errors.
+///
+///
+/// # Examples
+///
+/// Transmuting between `Sized` types:
+///
+/// ```
+/// # use zerocopy::transmute_mut;
+/// let mut one_dimensional: [u8; 8] = [0, 1, 2, 3, 4, 5, 6, 7];
+///
+/// let two_dimensional: &mut [[u8; 4]; 2] = transmute_mut!(&mut one_dimensional);
+///
+/// assert_eq!(two_dimensional, &[[0, 1, 2, 3], [4, 5, 6, 7]]);
+///
+/// two_dimensional.reverse();
+///
+/// assert_eq!(one_dimensional, [4, 5, 6, 7, 0, 1, 2, 3]);
+/// ```
+///
+/// Transmuting between unsized types:
+///
+/// ```
+/// # use {zerocopy::*, zerocopy_derive::*};
+/// # type u16 = zerocopy::byteorder::native_endian::U16;
+/// # type u32 = zerocopy::byteorder::native_endian::U32;
+/// #[derive(KnownLayout, FromBytes, IntoBytes, Immutable)]
+/// #[repr(C)]
+/// struct SliceDst<T, U> {
+///     t: T,
+///     u: [U],
+/// }
+///
+/// type Src = SliceDst<u32, u16>;
+/// type Dst = SliceDst<u16, u8>;
+///
+/// let mut bytes = [0, 1, 2, 3, 4, 5, 6, 7];
+/// let src = Src::mut_from_bytes(&mut bytes[..]).unwrap();
+/// let dst: &mut Dst = transmute_mut!(src);
+///
+/// assert_eq!(dst.t.as_bytes(), [0, 1]);
+/// assert_eq!(dst.u, [2, 3, 4, 5, 6, 7]);
+///
+/// assert_eq!(src.t.as_bytes(), [0, 1, 2, 3]);
+/// assert_eq!(src.u.len(), 2);
+/// assert_eq!(src.u.as_bytes(), [4, 5, 6, 7]);
+/// ```
+#[macro_export]
+macro_rules! transmute_mut {
+    ($e:expr) => {{
+        // NOTE: This must be a macro (rather than a function with trait bounds)
+        // because, for backwards-compatibility on v0.8.x, we use the autoref
+        // specialization trick to dispatch to different `transmute_mut`
+        // implementations: one which doesn't require `Src: KnownLayout + Dst:
+        // KnownLayout` when `Src: Sized + Dst: Sized`, and one which requires
+        // `KnownLayout` bounds otherwise.
+
+        // Ensure that the source type is a mutable reference.
+        let e: &mut _ = $e;
+
+        #[allow(unused)]
+        use $crate::util::macro_util::TransmuteMutDst as _;
+        let t = $crate::util::macro_util::Wrap::new(e);
+        if false {
+            // This branch exists solely to force the compiler to infer the type
+            // of `Dst` *before* it attempts to resolve the method call to
+            // `transmute_mut` in the `else` branch.
+            //
+            // Without this, if `Src` is `Sized` but `Dst` is `!Sized`, the
+            // compiler will eagerly select the inherent impl of `transmute_mut`
+            // (which requires `Dst: Sized`) because inherent methods take
+            // priority over trait methods. It does this before it realizes
+            // `Dst` is `!Sized`, leading to a compile error when it checks the
+            // bounds later.
+            //
+            // By calling this helper (which returns `&mut Dst`), we force `Dst`
+            // to be fully resolved. By the time it gets to the `else` branch,
+            // the compiler knows `Dst` is `!Sized`, properly disqualifies the
+            // inherent method, and falls back to the trait implementation.
+            t.transmute_mut_inference_helper()
+        } else {
+            t.transmute_mut()
+        }
+    }}
+}
+
+/// Conditionally transmutes a value of one type to a value of another type of
+/// the same size.
+///
+/// This macro behaves like an invocation of this function:
+///
+/// ```ignore
+/// fn try_transmute<Src, Dst>(src: Src) -> Result<Dst, ValidityError<Src, Dst>>
+/// where
+///     Src: IntoBytes,
+///     Dst: TryFromBytes,
+///     size_of::<Src>() == size_of::<Dst>(),
+/// {
+/// # /*
+///     ...
+/// # */
+/// }
+/// ```
+///
+/// However, unlike a function, this macro can only be invoked when the types of
+/// `Src` and `Dst` are completely concrete. The types `Src` and `Dst` are
+/// inferred from the calling context; they cannot be explicitly specified in
+/// the macro invocation.
+///
+/// Note that the `Src` produced by the expression `$e` will *not* be dropped.
+/// Semantically, its bits will be copied into a new value of type `Dst`, the
+/// original `Src` will be forgotten, and the value of type `Dst` will be
+/// returned.
+///
+/// # Examples
+///
+/// ```
+/// # use zerocopy::*;
+/// // 0u8 → bool = false
+/// assert_eq!(try_transmute!(0u8), Ok(false));
+///
+/// // 1u8 → bool = true
+///  assert_eq!(try_transmute!(1u8), Ok(true));
+///
+/// // 2u8 → bool = error
+/// assert!(matches!(
+///     try_transmute!(2u8),
+///     Result::<bool, _>::Err(ValidityError { .. })
+/// ));
+/// ```
+///
+#[doc = codegen_section!(
+    header = "h2",
+    bench = "try_transmute",
+    format = "coco_static_size",
+)]
+#[macro_export]
+macro_rules! try_transmute {
+    ($e:expr) => {{
+        // NOTE: This must be a macro (rather than a function with trait bounds)
+        // because there's no way, in a generic context, to enforce that two
+        // types have the same size. `core::mem::transmute` uses compiler magic
+        // to enforce this so long as the types are concrete.
+
+        let e = $e;
+        if false {
+            // Check that the sizes of the source and destination types are
+            // equal.
+
+            // SAFETY: This code is never executed.
+            Ok(unsafe {
+                // Clippy: We can't annotate the types; this macro is designed
+                // to infer the types from the calling context.
+                #[allow(clippy::missing_transmute_annotations)]
+                $crate::util::macro_util::core_reexport::mem::transmute(e)
+            })
+        } else {
+            $crate::util::macro_util::try_transmute::<_, _>(e)
+        }
+    }}
+}
+
+/// Conditionally transmutes a mutable or immutable reference of one type to an
+/// immutable reference of another type of the same size and compatible
+/// alignment.
+///
+/// *Note that while the **value** of the referent is checked for validity at
+/// runtime, the **size** and **alignment** are checked at compile time. For
+/// conversions which are fallible with respect to size and alignment, see the
+/// methods on [`TryFromBytes`].*
+///
+/// This macro behaves like an invocation of this function:
+///
+/// ```ignore
+/// fn try_transmute_ref<Src, Dst>(src: &Src) -> Result<&Dst, ValidityError<&Src, Dst>>
+/// where
+///     Src: IntoBytes + Immutable + ?Sized,
+///     Dst: TryFromBytes + Immutable + ?Sized,
+///     align_of::<Src>() >= align_of::<Dst>(),
+///     size_compatible::<Src, Dst>(),
+/// {
+/// # /*
+///     ...
+/// # */
+/// }
+/// ```
+///
+/// The types `Src` and `Dst` are inferred from the calling context; they cannot
+/// be explicitly specified in the macro invocation.
+///
+/// [`TryFromBytes`]: crate::TryFromBytes
+///
+/// # Size compatibility
+///
+/// `try_transmute_ref!` supports transmuting between `Sized` types, between
+/// unsized (i.e., `?Sized`) types, and from a `Sized` type to an unsized type.
+/// It supports any transmutation that preserves the number of bytes of the
+/// referent, even if doing so requires updating the metadata stored in an
+/// unsized "fat" reference:
+///
+/// ```
+/// # use zerocopy::try_transmute_ref;
+/// # use core::mem::size_of_val; // Not in the prelude on our MSRV
+/// let src: &[[u8; 2]] = &[[0, 1], [2, 3]][..];
+/// let dst: &[u8] = try_transmute_ref!(src).unwrap();
+///
+/// assert_eq!(src.len(), 2);
+/// assert_eq!(dst.len(), 4);
+/// assert_eq!(dst, [0, 1, 2, 3]);
+/// assert_eq!(size_of_val(src), size_of_val(dst));
+/// ```
+///
+/// # Examples
+///
+/// Transmuting between `Sized` types:
+///
+/// ```
+/// # use zerocopy::*;
+/// // 0u8 → bool = false
+/// assert_eq!(try_transmute_ref!(&0u8), Ok(&false));
+///
+/// // 1u8 → bool = true
+///  assert_eq!(try_transmute_ref!(&1u8), Ok(&true));
+///
+/// // 2u8 → bool = error
+/// assert!(matches!(
+///     try_transmute_ref!(&2u8),
+///     Result::<&bool, _>::Err(ValidityError { .. })
+/// ));
+/// ```
+///
+/// Transmuting between unsized types:
+///
+/// ```
+/// # use {zerocopy::*, zerocopy_derive::*};
+/// # type u16 = zerocopy::byteorder::native_endian::U16;
+/// # type u32 = zerocopy::byteorder::native_endian::U32;
+/// #[derive(KnownLayout, FromBytes, IntoBytes, Immutable)]
+/// #[repr(C)]
+/// struct SliceDst<T, U> {
+///     t: T,
+///     u: [U],
+/// }
+///
+/// type Src = SliceDst<u32, u16>;
+/// type Dst = SliceDst<u16, bool>;
+///
+/// let src = Src::ref_from_bytes(&[0, 1, 0, 1, 0, 1, 0, 1]).unwrap();
+/// let dst: &Dst = try_transmute_ref!(src).unwrap();
+///
+/// assert_eq!(src.t.as_bytes(), [0, 1, 0, 1]);
+/// assert_eq!(src.u.len(), 2);
+/// assert_eq!(src.u.as_bytes(), [0, 1, 0, 1]);
+///
+/// assert_eq!(dst.t.as_bytes(), [0, 1]);
+/// assert_eq!(dst.u, [false, true, false, true, false, true]);
+/// ```
+///
+#[doc = codegen_section!(
+    header = "h2",
+    bench = "try_transmute_ref",
+    format = "coco",
+    arity = 2,
+    [
+        open
+        @index 1
+        @title "Sized"
+        @variant "static_size"
+    ],
+    [
+        @index 2
+        @title "Unsized"
+        @variant "dynamic_size"
+    ]
+)]
+#[macro_export]
+macro_rules! try_transmute_ref {
+    ($e:expr) => {{
+        // Ensure that the source type is a reference or a mutable reference
+        // (note that mutable references are implicitly reborrowed here).
+        let e: &_ = $e;
+
+        #[allow(unused_imports)]
+        use $crate::util::macro_util::TryTransmuteRefDst as _;
+        let t = $crate::util::macro_util::Wrap::new(e);
+        if false {
+            // This branch exists solely to force the compiler to infer the type
+            // of `Dst` *before* it attempts to resolve the method call to
+            // `try_transmute_ref` in the `else` branch.
+            //
+            // Without this, if `Src` is `Sized` but `Dst` is `!Sized`, the
+            // compiler will eagerly select the inherent impl of
+            // `try_transmute_ref` (which requires `Dst: Sized`) because
+            // inherent methods take priority over trait methods. It does this
+            // before it realizes `Dst` is `!Sized`, leading to a compile error
+            // when it checks the bounds later.
+            //
+            // By calling this helper (which returns `&Dst`), we force `Dst`
+            // to be fully resolved. By the time it gets to the `else`
+            // branch, the compiler knows `Dst` is `!Sized`, properly
+            // disqualifies the inherent method, and falls back to the trait
+            // implementation.
+            Ok(t.transmute_ref_inference_helper())
+        } else {
+            t.try_transmute_ref()
+        }
+    }}
+}
+
+/// Conditionally transmutes a mutable reference of one type to a mutable
+/// reference of another type of the same size and compatible alignment.
+///
+/// *Note that while the **value** of the referent is checked for validity at
+/// runtime, the **size** and **alignment** are checked at compile time. For
+/// conversions which are fallible with respect to size and alignment, see the
+/// methods on [`TryFromBytes`].*
+///
+/// This macro behaves like an invocation of this function:
+///
+/// ```ignore
+/// fn try_transmute_mut<Src, Dst>(src: &mut Src) -> Result<&mut Dst, ValidityError<&mut Src, Dst>>
+/// where
+///     Src: FromBytes + IntoBytes + ?Sized,
+///     Dst: TryFromBytes + IntoBytes + ?Sized,
+///     align_of::<Src>() >= align_of::<Dst>(),
+///     size_compatible::<Src, Dst>(),
+/// {
+/// # /*
+///     ...
+/// # */
+/// }
+/// ```
+///
+/// The types `Src` and `Dst` are inferred from the calling context; they cannot
+/// be explicitly specified in the macro invocation.
+///
+/// [`TryFromBytes`]: crate::TryFromBytes
+///
+/// # Size compatibility
+///
+/// `try_transmute_mut!` supports transmuting between `Sized` types, between
+/// unsized (i.e., `?Sized`) types, and from a `Sized` type to an unsized type.
+/// It supports any transmutation that preserves the number of bytes of the
+/// referent, even if doing so requires updating the metadata stored in an
+/// unsized "fat" reference:
+///
+/// ```
+/// # use zerocopy::try_transmute_mut;
+/// # use core::mem::size_of_val; // Not in the prelude on our MSRV
+/// let src: &mut [[u8; 2]] = &mut [[0, 1], [2, 3]][..];
+/// let dst: &mut [u8] = try_transmute_mut!(src).unwrap();
+///
+/// assert_eq!(dst.len(), 4);
+/// assert_eq!(dst, [0, 1, 2, 3]);
+/// let dst_size = size_of_val(dst);
+/// assert_eq!(src.len(), 2);
+/// assert_eq!(size_of_val(src), dst_size);
+/// ```
+///
+/// # Examples
+///
+/// Transmuting between `Sized` types:
+///
+/// ```
+/// # use zerocopy::*;
+/// // 0u8 → bool = false
+/// let src = &mut 0u8;
+/// assert_eq!(try_transmute_mut!(src), Ok(&mut false));
+///
+/// // 1u8 → bool = true
+/// let src = &mut 1u8;
+///  assert_eq!(try_transmute_mut!(src), Ok(&mut true));
+///
+/// // 2u8 → bool = error
+/// let src = &mut 2u8;
+/// assert!(matches!(
+///     try_transmute_mut!(src),
+///     Result::<&mut bool, _>::Err(ValidityError { .. })
+/// ));
+/// ```
+///
+/// Transmuting between unsized types:
+///
+/// ```
+/// # use {zerocopy::*, zerocopy_derive::*};
+/// # type u16 = zerocopy::byteorder::native_endian::U16;
+/// # type u32 = zerocopy::byteorder::native_endian::U32;
+/// #[derive(KnownLayout, FromBytes, IntoBytes, Immutable)]
+/// #[repr(C)]
+/// struct SliceDst<T, U> {
+///     t: T,
+///     u: [U],
+/// }
+///
+/// type Src = SliceDst<u32, u16>;
+/// type Dst = SliceDst<u16, bool>;
+///
+/// let mut bytes = [0, 1, 0, 1, 0, 1, 0, 1];
+/// let src = Src::mut_from_bytes(&mut bytes).unwrap();
+///
+/// assert_eq!(src.t.as_bytes(), [0, 1, 0, 1]);
+/// assert_eq!(src.u.len(), 2);
+/// assert_eq!(src.u.as_bytes(), [0, 1, 0, 1]);
+///
+/// let dst: &Dst = try_transmute_mut!(src).unwrap();
+///
+/// assert_eq!(dst.t.as_bytes(), [0, 1]);
+/// assert_eq!(dst.u, [false, true, false, true, false, true]);
+/// ```
+#[macro_export]
+macro_rules! try_transmute_mut {
+    ($e:expr) => {{
+        // Ensure that the source type is a mutable reference.
+        let e: &mut _ = $e;
+
+        #[allow(unused_imports)]
+        use $crate::util::macro_util::TryTransmuteMutDst as _;
+        let t = $crate::util::macro_util::Wrap::new(e);
+        if false {
+            // This branch exists solely to force the compiler to infer the type
+            // of `Dst` *before* it attempts to resolve the method call to
+            // `try_transmute_mut` in the `else` branch.
+            //
+            // Without this, if `Src` is `Sized` but `Dst` is `!Sized`, the
+            // compiler will eagerly select the inherent impl of
+            // `try_transmute_mut` (which requires `Dst: Sized`) because
+            // inherent methods take priority over trait methods. It does this
+            // before it realizes `Dst` is `!Sized`, leading to a compile error
+            // when it checks the bounds later.
+            //
+            // By calling this helper (which returns `&Dst`), we force `Dst`
+            // to be fully resolved. By the time it gets to the `else`
+            // branch, the compiler knows `Dst` is `!Sized`, properly
+            // disqualifies the inherent method, and falls back to the trait
+            // implementation.
+            Ok(t.transmute_mut_inference_helper())
+        } else {
+            t.try_transmute_mut()
+        }
+    }}
+}
+
+/// Includes a file and safely transmutes it to a value of an arbitrary type.
+///
+/// The file will be included as a byte array, `[u8; N]`, which will be
+/// transmuted to another type, `T`. `T` is inferred from the calling context,
+/// and must implement [`FromBytes`].
+///
+/// The file is located relative to the current file (similarly to how modules
+/// are found). The provided path is interpreted in a platform-specific way at
+/// compile time. So, for instance, an invocation with a Windows path containing
+/// backslashes `\` would not compile correctly on Unix.
+///
+/// `include_value!` is ignorant of byte order. For byte order-aware types, see
+/// the [`byteorder`] module.
+///
+/// [`FromBytes`]: crate::FromBytes
+/// [`byteorder`]: crate::byteorder
+///
+/// # Examples
+///
+/// Assume there are two files in the same directory with the following
+/// contents:
+///
+/// File `data` (no trailing newline):
+///
+/// ```text
+/// abcd
+/// ```
+///
+/// File `main.rs`:
+///
+/// ```rust
+/// use zerocopy::include_value;
+/// # macro_rules! include_value {
+/// # ($file:expr) => { zerocopy::include_value!(concat!("../testdata/include_value/", $file)) };
+/// # }
+///
+/// fn main() {
+///     let as_u32: u32 = include_value!("data");
+///     assert_eq!(as_u32, u32::from_ne_bytes([b'a', b'b', b'c', b'd']));
+///     let as_i32: i32 = include_value!("data");
+///     assert_eq!(as_i32, i32::from_ne_bytes([b'a', b'b', b'c', b'd']));
+/// }
+/// ```
+///
+/// # Use in `const` contexts
+///
+/// This macro can be invoked in `const` contexts.
+#[doc(alias("include_bytes", "include_data", "include_type"))]
+#[macro_export]
+macro_rules! include_value {
+    ($file:expr $(,)?) => {
+        $crate::transmute!(*::core::include_bytes!($file))
+    };
+}
+
+#[doc(hidden)]
+#[macro_export]
+macro_rules! cryptocorrosion_derive_traits {
+    (
+        #[repr($repr:ident)]
+        $(#[$attr:meta])*
+        $vis:vis struct $name:ident $(<$($tyvar:ident),*>)?
+        $(
+            (
+                $($tuple_field_vis:vis $tuple_field_ty:ty),*
+            );
+        )?
+
+        $(
+            {
+                $($field_vis:vis $field_name:ident: $field_ty:ty,)*
+            }
+        )?
+    ) => {
+        $crate::cryptocorrosion_derive_traits!(@assert_allowed_struct_repr #[repr($repr)]);
+
+        $(#[$attr])*
+        #[repr($repr)]
+        $vis struct $name $(<$($tyvar),*>)?
+        $(
+            (
+                $($tuple_field_vis $tuple_field_ty),*
+            );
+        )?
+
+        $(
+            {
+                $($field_vis $field_name: $field_ty,)*
+            }
+        )?
+
+        // SAFETY: See inline.
+        unsafe impl $(<$($tyvar),*>)? $crate::TryFromBytes for $name$(<$($tyvar),*>)?
+        where
+            $(
+                $($tuple_field_ty: $crate::FromBytes,)*
+            )?
+
+            $(
+                $($field_ty: $crate::FromBytes,)*
+            )?
+        {
+            #[inline(always)]
+            fn is_bit_valid<A>(_: $crate::Maybe<'_, Self, A>) -> bool
+            where
+                A: $crate::invariant::Alignment,
+            {
+                // SAFETY: This macro only accepts `#[repr(C)]` and
+                // `#[repr(transparent)]` structs, and this `impl` block
+                // requires all field types to be `FromBytes`. Thus, all
+                // initialized byte sequences constitutes valid instances of
+                // `Self`.
+                true
+            }
+
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` and
+        // `#[repr(transparent)]` structs, and this `impl` block requires all
+        // field types to be `FromBytes`, which is a sub-trait of `FromZeros`.
+        unsafe impl $(<$($tyvar),*>)? $crate::FromZeros for $name$(<$($tyvar),*>)?
+        where
+            $(
+                $($tuple_field_ty: $crate::FromBytes,)*
+            )?
+
+            $(
+                $($field_ty: $crate::FromBytes,)*
+            )?
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` and
+        // `#[repr(transparent)]` structs, and this `impl` block requires all
+        // field types to be `FromBytes`.
+        unsafe impl $(<$($tyvar),*>)? $crate::FromBytes for $name$(<$($tyvar),*>)?
+        where
+            $(
+                $($tuple_field_ty: $crate::FromBytes,)*
+            )?
+
+            $(
+                $($field_ty: $crate::FromBytes,)*
+            )?
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` and
+        // `#[repr(transparent)]` structs, this `impl` block requires all field
+        // types to be `IntoBytes`, and a padding check is used to ensures that
+        // there are no padding bytes.
+        unsafe impl $(<$($tyvar),*>)? $crate::IntoBytes for $name$(<$($tyvar),*>)?
+        where
+            $(
+                $($tuple_field_ty: $crate::IntoBytes,)*
+            )?
+
+            $(
+                $($field_ty: $crate::IntoBytes,)*
+            )?
+
+            (): $crate::util::macro_util::PaddingFree<
+                Self,
+                {
+                    $crate::cryptocorrosion_derive_traits!(
+                        @struct_padding_check #[repr($repr)]
+                        $(($($tuple_field_ty),*))?
+                        $({$($field_ty),*})?
+                    )
+                },
+            >,
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` and
+        // `#[repr(transparent)]` structs, and this `impl` block requires all
+        // field types to be `Immutable`.
+        unsafe impl $(<$($tyvar),*>)? $crate::Immutable for $name$(<$($tyvar),*>)?
+        where
+            $(
+                $($tuple_field_ty: $crate::Immutable,)*
+            )?
+
+            $(
+                $($field_ty: $crate::Immutable,)*
+            )?
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+    };
+    (@assert_allowed_struct_repr #[repr(transparent)]) => {};
+    (@assert_allowed_struct_repr #[repr(C)]) => {};
+    (@assert_allowed_struct_repr #[$_attr:meta]) => {
+        compile_error!("repr must be `#[repr(transparent)]` or `#[repr(C)]`");
+    };
+    (
+        @struct_padding_check #[repr(transparent)]
+        $(($($tuple_field_ty:ty),*))?
+        $({$($field_ty:ty),*})?
+    ) => {
+        // SAFETY: `#[repr(transparent)]` structs cannot have the same layout as
+        // their single non-zero-sized field, and so cannot have any padding
+        // outside of that field.
+        0
+    };
+    (
+        @struct_padding_check #[repr(C)]
+        $(($($tuple_field_ty:ty),*))?
+        $({$($field_ty:ty),*})?
+    ) => {
+        $crate::struct_padding!(
+            Self,
+            None,
+            None,
+            [
+                $($($tuple_field_ty),*)?
+                $($($field_ty),*)?
+            ]
+        )
+    };
+    (
+        #[repr(C)]
+        $(#[$attr:meta])*
+        $vis:vis union $name:ident {
+            $(
+                $field_name:ident: $field_ty:ty,
+            )*
+        }
+    ) => {
+        $(#[$attr])*
+        #[repr(C)]
+        $vis union $name {
+            $(
+                $field_name: $field_ty,
+            )*
+        }
+
+        // SAFETY: See inline.
+        unsafe impl $crate::TryFromBytes for $name
+        where
+            $(
+                $field_ty: $crate::FromBytes,
+            )*
+        {
+            #[inline(always)]
+            fn is_bit_valid<A>(_: $crate::Maybe<'_, Self, A>) -> bool
+            where
+                A: $crate::invariant::Alignment,
+            {
+                // SAFETY: This macro only accepts `#[repr(C)]` unions, and this
+                // `impl` block requires all field types to be `FromBytes`.
+                // Thus, all initialized byte sequences constitutes valid
+                // instances of `Self`.
+                true
+            }
+
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` unions, and this `impl`
+        // block requires all field types to be `FromBytes`, which is a
+        // sub-trait of `FromZeros`.
+        unsafe impl $crate::FromZeros for $name
+        where
+            $(
+                $field_ty: $crate::FromBytes,
+            )*
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` unions, and this `impl`
+        // block requires all field types to be `FromBytes`.
+        unsafe impl $crate::FromBytes for $name
+        where
+            $(
+                $field_ty: $crate::FromBytes,
+            )*
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` unions, this `impl`
+        // block requires all field types to be `IntoBytes`, and a padding check
+        // is used to ensures that there are no padding bytes before or after
+        // any field.
+        unsafe impl $crate::IntoBytes for $name
+        where
+            $(
+                $field_ty: $crate::IntoBytes,
+            )*
+            (): $crate::util::macro_util::PaddingFree<
+                Self,
+                {
+                    $crate::union_padding!(
+                        Self,
+                        None::<usize>,
+                        None::<usize>,
+                        [$($field_ty),*]
+                    )
+                },
+            >,
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+
+        // SAFETY: This macro only accepts `#[repr(C)]` unions, and this `impl`
+        // block requires all field types to be `Immutable`.
+        unsafe impl $crate::Immutable for $name
+        where
+            $(
+                $field_ty: $crate::Immutable,
+            )*
+        {
+            fn only_derive_is_allowed_to_implement_this_trait() {}
+        }
+    };
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::{
+        byteorder::native_endian::{U16, U32},
+        util::testutil::*,
+        *,
+    };
+
+    #[derive(KnownLayout, Immutable, FromBytes, IntoBytes, PartialEq, Debug)]
+    #[repr(C)]
+    struct SliceDst<T, U> {
+        a: T,
+        b: [U],
+    }
+
+    #[test]
+    fn test_transmute() {
+        // Test that memory is transmuted as expected.
+        let array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let x: [[u8; 2]; 4] = transmute!(array_of_u8s);
+        assert_eq!(x, array_of_arrays);
+        let x: [u8; 8] = transmute!(array_of_arrays);
+        assert_eq!(x, array_of_u8s);
+
+        // Test that memory is transmuted as expected when shrinking.
+        let x: [[u8; 2]; 3] = transmute!(#![allow(shrink)] array_of_u8s);
+        assert_eq!(x, [[0u8, 1], [2, 3], [4, 5]]);
+
+        // Test that the source expression's value is forgotten rather than
+        // dropped.
+        #[derive(IntoBytes)]
+        #[repr(transparent)]
+        struct PanicOnDrop(());
+        impl Drop for PanicOnDrop {
+            fn drop(&mut self) {
+                panic!("PanicOnDrop::drop");
+            }
+        }
+        #[allow(clippy::let_unit_value)]
+        let _: () = transmute!(PanicOnDrop(()));
+        #[allow(clippy::let_unit_value)]
+        let _: () = transmute!(#![allow(shrink)] PanicOnDrop(()));
+
+        // Test that `transmute!` is legal in a const context.
+        const ARRAY_OF_U8S: [u8; 8] = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        const ARRAY_OF_ARRAYS: [[u8; 2]; 4] = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        const X: [[u8; 2]; 4] = transmute!(ARRAY_OF_U8S);
+        assert_eq!(X, ARRAY_OF_ARRAYS);
+        const X_SHRINK: [[u8; 2]; 3] = transmute!(#![allow(shrink)] ARRAY_OF_U8S);
+        assert_eq!(X_SHRINK, [[0u8, 1], [2, 3], [4, 5]]);
+
+        // Test that `transmute!` works with `!Immutable` types.
+        let x: usize = transmute!(UnsafeCell::new(1usize));
+        assert_eq!(x, 1);
+        let x: UnsafeCell<usize> = transmute!(1usize);
+        assert_eq!(x.into_inner(), 1);
+        let x: UnsafeCell<isize> = transmute!(UnsafeCell::new(1usize));
+        assert_eq!(x.into_inner(), 1);
+    }
+
+    // A `Sized` type which doesn't implement `KnownLayout` (it is "not
+    // `KnownLayout`", or `Nkl`).
+    //
+    // This permits us to test that `transmute_ref!` and `transmute_mut!` work
+    // for types which are `Sized + !KnownLayout`. When we added support for
+    // slice DSTs in #1924, this new support relied on `KnownLayout`, but we
+    // need to make sure to remain backwards-compatible with code which uses
+    // these macros with types which are `!KnownLayout`.
+    #[derive(FromBytes, IntoBytes, Immutable, PartialEq, Eq, Debug)]
+    #[repr(transparent)]
+    struct Nkl<T>(T);
+
+    #[test]
+    fn test_transmute_ref() {
+        // Test that memory is transmuted as expected.
+        let array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let x: &[[u8; 2]; 4] = transmute_ref!(&array_of_u8s);
+        assert_eq!(*x, array_of_arrays);
+        let x: &[u8; 8] = transmute_ref!(&array_of_arrays);
+        assert_eq!(*x, array_of_u8s);
+
+        // Test that `transmute_ref!` is legal in a const context.
+        const ARRAY_OF_U8S: [u8; 8] = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        const ARRAY_OF_ARRAYS: [[u8; 2]; 4] = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        #[allow(clippy::redundant_static_lifetimes)]
+        const X: &'static [[u8; 2]; 4] = transmute_ref!(&ARRAY_OF_U8S);
+        assert_eq!(*X, ARRAY_OF_ARRAYS);
+
+        // Test sized -> unsized transmutation.
+        let array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let slice_of_arrays = &array_of_arrays[..];
+        let x: &[[u8; 2]] = transmute_ref!(&array_of_u8s);
+        assert_eq!(x, slice_of_arrays);
+
+        // Before 1.61.0, we can't define the `const fn transmute_ref` function
+        // that we do on and after 1.61.0.
+        #[cfg(no_zerocopy_generic_bounds_in_const_fn_1_61_0)]
+        {
+            // Test that `transmute_ref!` supports non-`KnownLayout` `Sized`
+            // types.
+            const ARRAY_OF_NKL_U8S: Nkl<[u8; 8]> = Nkl([0u8, 1, 2, 3, 4, 5, 6, 7]);
+            const ARRAY_OF_NKL_ARRAYS: Nkl<[[u8; 2]; 4]> = Nkl([[0, 1], [2, 3], [4, 5], [6, 7]]);
+            const X_NKL: &Nkl<[[u8; 2]; 4]> = transmute_ref!(&ARRAY_OF_NKL_U8S);
+            assert_eq!(*X_NKL, ARRAY_OF_NKL_ARRAYS);
+        }
+
+        #[cfg(not(no_zerocopy_generic_bounds_in_const_fn_1_61_0))]
+        {
+            // Call through a generic function to make sure our autoref
+            // specialization trick works even when types are generic.
+            const fn transmute_ref<T, U>(t: &T) -> &U
+            where
+                T: IntoBytes + Immutable,
+                U: FromBytes + Immutable,
+            {
+                transmute_ref!(t)
+            }
+
+            // Test that `transmute_ref!` supports non-`KnownLayout` `Sized`
+            // types.
+            const ARRAY_OF_NKL_U8S: Nkl<[u8; 8]> = Nkl([0u8, 1, 2, 3, 4, 5, 6, 7]);
+            const ARRAY_OF_NKL_ARRAYS: Nkl<[[u8; 2]; 4]> = Nkl([[0, 1], [2, 3], [4, 5], [6, 7]]);
+            const X_NKL: &Nkl<[[u8; 2]; 4]> = transmute_ref(&ARRAY_OF_NKL_U8S);
+            assert_eq!(*X_NKL, ARRAY_OF_NKL_ARRAYS);
+        }
+
+        // Test that `transmute_ref!` works on slice DSTs in and that memory is
+        // transmuted as expected.
+        let slice_dst_of_u8s =
+            SliceDst::<U16, [u8; 2]>::ref_from_bytes(&[0, 1, 2, 3, 4, 5][..]).unwrap();
+        let slice_dst_of_u16s =
+            SliceDst::<U16, U16>::ref_from_bytes(&[0, 1, 2, 3, 4, 5][..]).unwrap();
+        let x: &SliceDst<U16, U16> = transmute_ref!(slice_dst_of_u8s);
+        assert_eq!(x, slice_dst_of_u16s);
+
+        let slice_dst_of_u8s =
+            SliceDst::<U16, u8>::ref_from_bytes(&[0, 1, 2, 3, 4, 5][..]).unwrap();
+        let x: &[u8] = transmute_ref!(slice_dst_of_u8s);
+        assert_eq!(x, [0, 1, 2, 3, 4, 5]);
+
+        let x: &[u8] = transmute_ref!(slice_dst_of_u16s);
+        assert_eq!(x, [0, 1, 2, 3, 4, 5]);
+
+        let x: &[U16] = transmute_ref!(slice_dst_of_u16s);
+        let slice_of_u16s: &[U16] = <[U16]>::ref_from_bytes(&[0, 1, 2, 3, 4, 5][..]).unwrap();
+        assert_eq!(x, slice_of_u16s);
+
+        // Test that transmuting from a type with larger trailing slice offset
+        // and larger trailing slice element works.
+        let bytes = &[0, 1, 2, 3, 4, 5, 6, 7][..];
+        let slice_dst_big = SliceDst::<U32, U16>::ref_from_bytes(bytes).unwrap();
+        let slice_dst_small = SliceDst::<U16, u8>::ref_from_bytes(bytes).unwrap();
+        let x: &SliceDst<U16, u8> = transmute_ref!(slice_dst_big);
+        assert_eq!(x, slice_dst_small);
+
+        // Test that it's legal to transmute a reference while shrinking the
+        // lifetime (note that `X` has the lifetime `'static`).
+        let x: &[u8; 8] = transmute_ref!(X);
+        assert_eq!(*x, ARRAY_OF_U8S);
+
+        // Test that `transmute_ref!` supports decreasing alignment.
+        let u = AU64(0);
+        let array = [0, 0, 0, 0, 0, 0, 0, 0];
+        let x: &[u8; 8] = transmute_ref!(&u);
+        assert_eq!(*x, array);
+
+        // Test that a mutable reference can be turned into an immutable one.
+        let mut x = 0u8;
+        #[allow(clippy::useless_transmute)]
+        let y: &u8 = transmute_ref!(&mut x);
+        assert_eq!(*y, 0);
+    }
+
+    #[test]
+    fn test_try_transmute() {
+        // Test that memory is transmuted with `try_transmute` as expected.
+        let array_of_bools = [false, true, false, true, false, true, false, true];
+        let array_of_arrays = [[0, 1], [0, 1], [0, 1], [0, 1]];
+        let x: Result<[[u8; 2]; 4], _> = try_transmute!(array_of_bools);
+        assert_eq!(x, Ok(array_of_arrays));
+        let x: Result<[bool; 8], _> = try_transmute!(array_of_arrays);
+        assert_eq!(x, Ok(array_of_bools));
+
+        // Test that `try_transmute!` works with `!Immutable` types.
+        let x: Result<usize, _> = try_transmute!(UnsafeCell::new(1usize));
+        assert_eq!(x.unwrap(), 1);
+        let x: Result<UnsafeCell<usize>, _> = try_transmute!(1usize);
+        assert_eq!(x.unwrap().into_inner(), 1);
+        let x: Result<UnsafeCell<isize>, _> = try_transmute!(UnsafeCell::new(1usize));
+        assert_eq!(x.unwrap().into_inner(), 1);
+
+        #[derive(FromBytes, IntoBytes, Debug, PartialEq)]
+        #[repr(transparent)]
+        struct PanicOnDrop<T>(T);
+
+        impl<T> Drop for PanicOnDrop<T> {
+            fn drop(&mut self) {
+                panic!("PanicOnDrop dropped");
+            }
+        }
+
+        // Since `try_transmute!` semantically moves its argument on failure,
+        // the `PanicOnDrop` is not dropped, and thus this shouldn't panic.
+        let x: Result<usize, _> = try_transmute!(PanicOnDrop(1usize));
+        assert_eq!(x, Ok(1));
+
+        // Since `try_transmute!` semantically returns ownership of its argument
+        // on failure, the `PanicOnDrop` is returned rather than dropped, and
+        // thus this shouldn't panic.
+        let y: Result<bool, _> = try_transmute!(PanicOnDrop(2u8));
+        // We have to use `map_err` instead of comparing against
+        // `Err(PanicOnDrop(2u8))` because the latter would create and then drop
+        // its `PanicOnDrop` temporary, which would cause a panic.
+        assert_eq!(y.as_ref().map_err(|p| &p.src.0), Err::<&bool, _>(&2u8));
+        mem::forget(y);
+    }
+
+    #[test]
+    fn test_try_transmute_ref() {
+        // Test that memory is transmuted with `try_transmute_ref` as expected.
+        let array_of_bools = &[false, true, false, true, false, true, false, true];
+        let array_of_arrays = &[[0, 1], [0, 1], [0, 1], [0, 1]];
+        let x: Result<&[[u8; 2]; 4], _> = try_transmute_ref!(array_of_bools);
+        assert_eq!(x, Ok(array_of_arrays));
+        let x: Result<&[bool; 8], _> = try_transmute_ref!(array_of_arrays);
+        assert_eq!(x, Ok(array_of_bools));
+
+        // Test that it's legal to transmute a reference while shrinking the
+        // lifetime.
+        {
+            let x: Result<&[[u8; 2]; 4], _> = try_transmute_ref!(array_of_bools);
+            assert_eq!(x, Ok(array_of_arrays));
+        }
+
+        // Test that `try_transmute_ref!` supports decreasing alignment.
+        let u = AU64(0);
+        let array = [0u8, 0, 0, 0, 0, 0, 0, 0];
+        let x: Result<&[u8; 8], _> = try_transmute_ref!(&u);
+        assert_eq!(x, Ok(&array));
+
+        // Test that a mutable reference can be turned into an immutable one.
+        let mut x = 0u8;
+        #[allow(clippy::useless_transmute)]
+        let y: Result<&u8, _> = try_transmute_ref!(&mut x);
+        assert_eq!(y, Ok(&0));
+
+        // Test that sized types work which don't implement `KnownLayout`.
+        let array_of_nkl_u8s = Nkl([0u8, 1, 2, 3, 4, 5, 6, 7]);
+        let array_of_nkl_arrays = Nkl([[0, 1], [2, 3], [4, 5], [6, 7]]);
+        let x: Result<&Nkl<[[u8; 2]; 4]>, _> = try_transmute_ref!(&array_of_nkl_u8s);
+        assert_eq!(x, Ok(&array_of_nkl_arrays));
+
+        // Test sized -> unsized transmutation.
+        let array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let slice_of_arrays = &array_of_arrays[..];
+        let x: Result<&[[u8; 2]], _> = try_transmute_ref!(&array_of_u8s);
+        assert_eq!(x, Ok(slice_of_arrays));
+
+        // Test unsized -> unsized transmutation.
+        let slice_dst_of_u8s =
+            SliceDst::<U16, [u8; 2]>::ref_from_bytes(&[0, 1, 2, 3, 4, 5][..]).unwrap();
+        let slice_dst_of_u16s =
+            SliceDst::<U16, U16>::ref_from_bytes(&[0, 1, 2, 3, 4, 5][..]).unwrap();
+        let x: Result<&SliceDst<U16, U16>, _> = try_transmute_ref!(slice_dst_of_u8s);
+        assert_eq!(x, Ok(slice_dst_of_u16s));
+    }
+
+    #[test]
+    fn test_try_transmute_mut() {
+        // Test that memory is transmuted with `try_transmute_mut` as expected.
+        let array_of_u8s = &mut [0u8, 1, 0, 1, 0, 1, 0, 1];
+        let array_of_arrays = &mut [[0u8, 1], [0, 1], [0, 1], [0, 1]];
+        let x: Result<&mut [[u8; 2]; 4], _> = try_transmute_mut!(array_of_u8s);
+        assert_eq!(x, Ok(array_of_arrays));
+
+        let array_of_bools = &mut [false, true, false, true, false, true, false, true];
+        let array_of_arrays = &mut [[0u8, 1], [0, 1], [0, 1], [0, 1]];
+        let x: Result<&mut [bool; 8], _> = try_transmute_mut!(array_of_arrays);
+        assert_eq!(x, Ok(array_of_bools));
+
+        // Test that it's legal to transmute a reference while shrinking the
+        // lifetime.
+        let array_of_bools = &mut [false, true, false, true, false, true, false, true];
+        let array_of_arrays = &mut [[0u8, 1], [0, 1], [0, 1], [0, 1]];
+        {
+            let x: Result<&mut [bool; 8], _> = try_transmute_mut!(array_of_arrays);
+            assert_eq!(x, Ok(array_of_bools));
+        }
+
+        // Test that `try_transmute_mut!` supports decreasing alignment.
+        let u = &mut AU64(0);
+        let array = &mut [0u8, 0, 0, 0, 0, 0, 0, 0];
+        let x: Result<&mut [u8; 8], _> = try_transmute_mut!(u);
+        assert_eq!(x, Ok(array));
+
+        // Test that a mutable reference can be turned into an immutable one.
+        let mut x = 0u8;
+        #[allow(clippy::useless_transmute)]
+        let y: Result<&mut u8, _> = try_transmute_mut!(&mut x);
+        assert_eq!(y, Ok(&mut 0));
+
+        // Test that sized types work which don't implement `KnownLayout`.
+        let mut array_of_nkl_u8s = Nkl([0u8, 1, 2, 3, 4, 5, 6, 7]);
+        let mut array_of_nkl_arrays = Nkl([[0, 1], [2, 3], [4, 5], [6, 7]]);
+        let x: Result<&mut Nkl<[[u8; 2]; 4]>, _> = try_transmute_mut!(&mut array_of_nkl_u8s);
+        assert_eq!(x, Ok(&mut array_of_nkl_arrays));
+
+        // Test sized -> unsized transmutation.
+        let mut array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let mut array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let slice_of_arrays = &mut array_of_arrays[..];
+        let x: Result<&mut [[u8; 2]], _> = try_transmute_mut!(&mut array_of_u8s);
+        assert_eq!(x, Ok(slice_of_arrays));
+
+        // Test unsized -> unsized transmutation.
+        let mut bytes = [0, 1, 2, 3, 4, 5, 6];
+        let slice_dst_of_u8s = SliceDst::<u8, [u8; 2]>::mut_from_bytes(&mut bytes[..]).unwrap();
+        let mut bytes = [0, 1, 2, 3, 4, 5, 6];
+        let slice_dst_of_u16s = SliceDst::<u8, U16>::mut_from_bytes(&mut bytes[..]).unwrap();
+        let x: Result<&mut SliceDst<u8, U16>, _> = try_transmute_mut!(slice_dst_of_u8s);
+        assert_eq!(x, Ok(slice_dst_of_u16s));
+    }
+
+    #[test]
+    fn test_transmute_mut() {
+        // Test that memory is transmuted as expected.
+        let mut array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let mut array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let x: &mut [[u8; 2]; 4] = transmute_mut!(&mut array_of_u8s);
+        assert_eq!(*x, array_of_arrays);
+        let x: &mut [u8; 8] = transmute_mut!(&mut array_of_arrays);
+        assert_eq!(*x, array_of_u8s);
+
+        {
+            // Test that it's legal to transmute a reference while shrinking the
+            // lifetime.
+            let x: &mut [u8; 8] = transmute_mut!(&mut array_of_arrays);
+            assert_eq!(*x, array_of_u8s);
+        }
+
+        // Test that `transmute_mut!` supports non-`KnownLayout` types.
+        let mut array_of_u8s = Nkl([0u8, 1, 2, 3, 4, 5, 6, 7]);
+        let mut array_of_arrays = Nkl([[0, 1], [2, 3], [4, 5], [6, 7]]);
+        let x: &mut Nkl<[[u8; 2]; 4]> = transmute_mut!(&mut array_of_u8s);
+        assert_eq!(*x, array_of_arrays);
+        let x: &mut Nkl<[u8; 8]> = transmute_mut!(&mut array_of_arrays);
+        assert_eq!(*x, array_of_u8s);
+
+        // Test that `transmute_mut!` supports decreasing alignment.
+        let mut u = AU64(0);
+        let array = [0, 0, 0, 0, 0, 0, 0, 0];
+        let x: &[u8; 8] = transmute_mut!(&mut u);
+        assert_eq!(*x, array);
+
+        // Test that a mutable reference can be turned into an immutable one.
+        let mut x = 0u8;
+        #[allow(clippy::useless_transmute)]
+        let y: &u8 = transmute_mut!(&mut x);
+        assert_eq!(*y, 0);
+
+        // Test that `transmute_mut!` works on slice DSTs in and that memory is
+        // transmuted as expected.
+        let mut bytes = [0, 1, 2, 3, 4, 5, 6];
+        let slice_dst_of_u8s = SliceDst::<u8, [u8; 2]>::mut_from_bytes(&mut bytes[..]).unwrap();
+        let mut bytes = [0, 1, 2, 3, 4, 5, 6];
+        let slice_dst_of_u16s = SliceDst::<u8, U16>::mut_from_bytes(&mut bytes[..]).unwrap();
+        let x: &mut SliceDst<u8, U16> = transmute_mut!(slice_dst_of_u8s);
+        assert_eq!(x, slice_dst_of_u16s);
+
+        // Test that `transmute_mut!` works on slices that memory is transmuted
+        // as expected.
+        let array_of_u16s: &mut [u16] = &mut [0u16, 1, 2];
+        let array_of_i16s: &mut [i16] = &mut [0i16, 1, 2];
+        let x: &mut [i16] = transmute_mut!(array_of_u16s);
+        assert_eq!(x, array_of_i16s);
+
+        // Test that transmuting from a type with larger trailing slice offset
+        // and larger trailing slice element works.
+        let mut bytes = [0, 1, 2, 3, 4, 5, 6, 7];
+        let slice_dst_big = SliceDst::<U32, U16>::mut_from_bytes(&mut bytes[..]).unwrap();
+        let mut bytes = [0, 1, 2, 3, 4, 5, 6, 7];
+        let slice_dst_small = SliceDst::<U16, u8>::mut_from_bytes(&mut bytes[..]).unwrap();
+        let x: &mut SliceDst<U16, u8> = transmute_mut!(slice_dst_big);
+        assert_eq!(x, slice_dst_small);
+
+        // Test sized -> unsized transmutation.
+        let mut array_of_u8s = [0u8, 1, 2, 3, 4, 5, 6, 7];
+        let mut array_of_arrays = [[0, 1], [2, 3], [4, 5], [6, 7]];
+        let slice_of_arrays = &mut array_of_arrays[..];
+        let x: &mut [[u8; 2]] = transmute_mut!(&mut array_of_u8s);
+        assert_eq!(x, slice_of_arrays);
+    }
+
+    #[test]
+    fn test_macros_evaluate_args_once() {
+        let mut ctr = 0;
+        #[allow(clippy::useless_transmute)]
+        let _: usize = transmute!({
+            ctr += 1;
+            0usize
+        });
+        assert_eq!(ctr, 1);
+
+        let mut ctr = 0;
+        let _: &usize = transmute_ref!({
+            ctr += 1;
+            &0usize
+        });
+        assert_eq!(ctr, 1);
+
+        let mut ctr: usize = 0;
+        let _: &mut usize = transmute_mut!({
+            ctr += 1;
+            &mut ctr
+        });
+        assert_eq!(ctr, 1);
+
+        let mut ctr = 0;
+        #[allow(clippy::useless_transmute)]
+        let _: usize = try_transmute!({
+            ctr += 1;
+            0usize
+        })
+        .unwrap();
+        assert_eq!(ctr, 1);
+    }
+
+    #[test]
+    fn test_include_value() {
+        const AS_U32: u32 = include_value!("../testdata/include_value/data");
+        assert_eq!(AS_U32, u32::from_ne_bytes([b'a', b'b', b'c', b'd']));
+        const AS_I32: i32 = include_value!("../testdata/include_value/data");
+        assert_eq!(AS_I32, i32::from_ne_bytes([b'a', b'b', b'c', b'd']));
+    }
+
+    #[test]
+    #[allow(non_camel_case_types, unreachable_pub, dead_code)]
+    fn test_cryptocorrosion_derive_traits() {
+        // Test the set of invocations added in
+        // https://github.com/cryptocorrosion/cryptocorrosion/pull/85
+
+        fn assert_impls<T: FromBytes + IntoBytes + Immutable>() {}
+
+        cryptocorrosion_derive_traits! {
+            #[repr(C)]
+            #[derive(Clone, Copy)]
+            pub union vec128_storage {
+                d: [u32; 4],
+                q: [u64; 2],
+            }
+        }
+
+        assert_impls::<vec128_storage>();
+
+        cryptocorrosion_derive_traits! {
+            #[repr(transparent)]
+            #[derive(Copy, Clone, Debug, PartialEq)]
+            pub struct u32x4_generic([u32; 4]);
+        }
+
+        assert_impls::<u32x4_generic>();
+
+        cryptocorrosion_derive_traits! {
+            #[repr(transparent)]
+            #[derive(Copy, Clone, Debug, PartialEq)]
+            pub struct u64x2_generic([u64; 2]);
+        }
+
+        assert_impls::<u64x2_generic>();
+
+        cryptocorrosion_derive_traits! {
+            #[repr(transparent)]
+            #[derive(Copy, Clone, Debug, PartialEq)]
+            pub struct u128x1_generic([u128; 1]);
+        }
+
+        assert_impls::<u128x1_generic>();
+
+        cryptocorrosion_derive_traits! {
+            #[repr(transparent)]
+            #[derive(Copy, Clone, Default)]
+            #[allow(non_camel_case_types)]
+            pub struct x2<W, G>(pub [W; 2], PhantomData<G>);
+        }
+
+        enum NotZerocopy {}
+        assert_impls::<x2<(), NotZerocopy>>();
+
+        cryptocorrosion_derive_traits! {
+            #[repr(transparent)]
+            #[derive(Copy, Clone, Default)]
+            #[allow(non_camel_case_types)]
+            pub struct x4<W>(pub [W; 4]);
+        }
+
+        assert_impls::<x4<()>>();
+
+        #[cfg(feature = "simd")]
+        #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+        {
+            #[cfg(target_arch = "x86")]
+            use core::arch::x86::{__m128i, __m256i};
+            #[cfg(target_arch = "x86_64")]
+            use core::arch::x86_64::{__m128i, __m256i};
+
+            cryptocorrosion_derive_traits! {
+                #[repr(C)]
+                #[derive(Copy, Clone)]
+                pub struct X4(__m128i, __m128i, __m128i, __m128i);
+            }
+
+            assert_impls::<X4>();
+
+            cryptocorrosion_derive_traits! {
+                #[repr(C)]
+                /// Generic wrapper for unparameterized storage of any of the
+                /// possible impls. Converting into and out of this type should
+                /// be essentially free, although it may be more aligned than a
+                /// particular impl requires.
+                #[allow(non_camel_case_types)]
+                #[derive(Copy, Clone)]
+                pub union vec128_storage {
+                    u32x4: [u32; 4],
+                    u64x2: [u64; 2],
+                    u128x1: [u128; 1],
+                    sse2: __m128i,
+                }
+            }
+
+            assert_impls::<vec128_storage>();
+
+            cryptocorrosion_derive_traits! {
+                #[repr(transparent)]
+                #[allow(non_camel_case_types)]
+                #[derive(Copy, Clone)]
+                pub struct vec<S3, S4, NI> {
+                    x: __m128i,
+                    s3: PhantomData<S3>,
+                    s4: PhantomData<S4>,
+                    ni: PhantomData<NI>,
+                }
+            }
+
+            assert_impls::<vec<NotZerocopy, NotZerocopy, NotZerocopy>>();
+
+            cryptocorrosion_derive_traits! {
+                #[repr(transparent)]
+                #[derive(Copy, Clone)]
+                pub struct u32x4x2_avx2<NI> {
+                    x: __m256i,
+                    ni: PhantomData<NI>,
+                }
+            }
+
+            assert_impls::<u32x4x2_avx2<NotZerocopy>>();
+        }
+
+        // Make sure that our derive works for `#[repr(C)]` structs even though
+        // cryptocorrosion doesn't currently have any.
+        cryptocorrosion_derive_traits! {
+            #[repr(C)]
+            #[derive(Copy, Clone, Debug, PartialEq)]
+            pub struct ReprC(u8, u8, u16);
+        }
+    }
+}