Alignment::new() takes a const usize, which means callers that work
with DeviceSize constants still need to import the usize SZ_*
variants. Add from_u64() so callers can write
Alignment::from_u64(u64::SZ_128K) and stay entirely in the
DeviceSize world.
Both asserts evaluate at compile time in const context, so there is
no runtime cost.
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
rust/kernel/ptr.rs | 35 +++++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
diff --git a/rust/kernel/ptr.rs b/rust/kernel/ptr.rs
index 5b6a382637fe..b06f6b404a46 100644
--- a/rust/kernel/ptr.rs
+++ b/rust/kernel/ptr.rs
@@ -76,6 +76,41 @@ pub const fn new_checked(align: usize) -> Option<Self> {
}
}
+ /// Creates an [`Alignment`] from a [`u64`] value.
+ ///
+ /// This is useful when the alignment comes from a [`DeviceSize`] constant
+ /// rather than a [`usize`] literal.
+ ///
+ /// A build error is triggered if `align` is not a power of two, or if it
+ /// exceeds [`usize::MAX`].
+ ///
+ /// [`DeviceSize`]: crate::sizes::DeviceSize
+ ///
+ /// # Examples
+ ///
+ /// ```
+ /// use kernel::ptr::Alignment;
+ /// use kernel::sizes::DeviceSize;
+ ///
+ /// let v = Alignment::from_u64(u64::SZ_128K);
+ /// assert_eq!(v.as_usize(), 0x0002_0000);
+ /// ```
+ #[inline(always)]
+ pub const fn from_u64(align: u64) -> Self {
+ assert!(
+ align.is_power_of_two(),
+ "Provided alignment is not a power of two."
+ );
+ assert!(
+ align <= usize::MAX as u64,
+ "Provided alignment exceeds usize::MAX."
+ );
+
+ // INVARIANT: `align` is a power of two.
+ // SAFETY: `align` is a power of two, fits in usize, and thus non-zero.
+ Self(unsafe { NonZero::new_unchecked(align as usize) })
+ }
+
/// Returns the alignment of `T`.
///
/// This is equivalent to [`align_of`], but with the return value provided as an [`Alignment`].
--
2.53.0
On Thu Mar 12, 2026 at 3:15 AM GMT, John Hubbard wrote:
> Alignment::new() takes a const usize, which means callers that work
> with DeviceSize constants still need to import the usize SZ_*
> variants. Add from_u64() so callers can write
> Alignment::from_u64(u64::SZ_128K) and stay entirely in the
> DeviceSize world.
>
> Both asserts evaluate at compile time in const context, so there is
> no runtime cost.
>
> Signed-off-by: John Hubbard <jhubbard@nvidia.com>
> ---
> rust/kernel/ptr.rs | 35 +++++++++++++++++++++++++++++++++++
> 1 file changed, 35 insertions(+)
>
> diff --git a/rust/kernel/ptr.rs b/rust/kernel/ptr.rs
> index 5b6a382637fe..b06f6b404a46 100644
> --- a/rust/kernel/ptr.rs
> +++ b/rust/kernel/ptr.rs
> @@ -76,6 +76,41 @@ pub const fn new_checked(align: usize) -> Option<Self> {
> }
> }
>
> + /// Creates an [`Alignment`] from a [`u64`] value.
> + ///
> + /// This is useful when the alignment comes from a [`DeviceSize`] constant
> + /// rather than a [`usize`] literal.
> + ///
> + /// A build error is triggered if `align` is not a power of two, or if it
> + /// exceeds [`usize::MAX`].
> + ///
> + /// [`DeviceSize`]: crate::sizes::DeviceSize
> + ///
> + /// # Examples
> + ///
> + /// ```
> + /// use kernel::ptr::Alignment;
> + /// use kernel::sizes::DeviceSize;
> + ///
> + /// let v = Alignment::from_u64(u64::SZ_128K);
> + /// assert_eq!(v.as_usize(), 0x0002_0000);
> + /// ```
> + #[inline(always)]
> + pub const fn from_u64(align: u64) -> Self {
> + assert!(
> + align.is_power_of_two(),
> + "Provided alignment is not a power of two."
> + );
This should be build_assert! to match the doc.
Best,
Gary
> + assert!(
> + align <= usize::MAX as u64,
> + "Provided alignment exceeds usize::MAX."
> + );
> +
> + // INVARIANT: `align` is a power of two.
> + // SAFETY: `align` is a power of two, fits in usize, and thus non-zero.
> + Self(unsafe { NonZero::new_unchecked(align as usize) })
> + }
> +
> /// Returns the alignment of `T`.
> ///
> /// This is equivalent to [`align_of`], but with the return value provided as an [`Alignment`].
On Thu, Mar 12, 2026 at 4:15 AM John Hubbard <jhubbard@nvidia.com> wrote: > > + /// A build error is triggered if `align` is not a power of two, or if it > + /// exceeds [`usize::MAX`]. I think this wording comes from the `new` constructor, but there we use a const assert, not a runtime one. So as you mention in the commit message, it is true that if you call it in a const context you will get it at compile-time error, but the wording here in the docs seems to imply otherwise. Cheers, Miguel
On 3/11/26 11:16 PM, Miguel Ojeda wrote: > On Thu, Mar 12, 2026 at 4:15 AM John Hubbard <jhubbard@nvidia.com> wrote: >> >> + /// A build error is triggered if `align` is not a power of two, or if it >> + /// exceeds [`usize::MAX`]. > > I think this wording comes from the `new` constructor, but there we Yes it does--guilty as charged, haha. :) > use a const assert, not a runtime one. > > So as you mention in the commit message, it is true that if you call > it in a const context you will get it at compile-time error, but the > wording here in the docs seems to imply otherwise. > Good catch, thanks. thanks, -- John Hubbard
© 2016 - 2026 Red Hat, Inc.