Re: [PATCH 1/6] rust: num: add `shr` and `shl` methods to `Bounded`

[Alice Ryhl]

On Wed, Jan 21, 2026 at 03:15:49AM -0500, Yury Norov wrote:
> On Tue, Jan 20, 2026 at 03:17:54PM +0900, Alexandre Courbot wrote:
> > Shifting a `Bounded` left or right changes the number of bits required
> > to represent the value. Add methods that perform the shift and return a
> > `Bounded` with the appropriately adjusted bit width.
> > 
> > These methods are particularly useful for bitfield extraction.
> > 
> > Suggested-by: Alice Ryhl <aliceryhl@xxxxxxxxxx>
> > Signed-off-by: Alexandre Courbot <acourbot@xxxxxxxxxx>
> > ---
> >  rust/kernel/num/bounded.rs | 40 ++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 40 insertions(+)
> > 
> > diff --git a/rust/kernel/num/bounded.rs b/rust/kernel/num/bounded.rs
> > index 5ef8361cf5d5..6e3f4a7a5262 100644
> > --- a/rust/kernel/num/bounded.rs
> > +++ b/rust/kernel/num/bounded.rs
> > @@ -475,6 +475,46 @@ pub fn cast<U>(self) -> Bounded<U, N>
> >          // `N` bits, and with the same signedness.
> >          unsafe { Bounded::__new(value) }
> >      }
> > +
> > +    /// Right-shifts `self` by `SHIFT` and returns the result as a `Bounded<_, { N - SHIFT }>`.
> > +    ///
> > +    /// # Examples
> > +    ///
> > +    /// ```
> > +    /// use kernel::num::Bounded;
> > +    ///
> > +    /// let v = Bounded::<u32, 16>::new::<0xff00>();
> > +    /// let v_shifted: Bounded::<u32, 8> = v.shr::<8, _>();
> 
> This syntax is really confusing. Does it work for runtime shifts?
> Is there any chance to make it just v.shr(8)?

The ::<NUM> syntax is how you pass a build-time constant value in Rust.
The syntax 'v.shr(8)' implies that 8 is a runtime value, and not a
build-time constant. I agree it's ugly, but that's how it is.

For runtime shifts, we can make it 'v.shr(8)' or even 'v >> 8' using
operator overloading, but see my further reply below:

> > +    }
> > +
> > +    /// Left-shifts `self` by `SHIFT` and returns the result as a `Bounded<_, { N + SHIFT }>`.
> > +    ///
> > +    /// # Examples
> > +    ///
> > +    /// ```
> > +    /// use kernel::num::Bounded;
> > +    ///
> > +    /// let v = Bounded::<u32, 8>::new::<0xff>();
> > +    /// let v_shifted: Bounded::<u32, 16> = v.shl::<8, _>();
> > +    ///
> > +    /// assert_eq!(v_shifted.get(), 0xff00);
> > +    /// ```
> > +    pub fn shl<const SHIFT: u32, const RES: u32>(self) -> Bounded<T, RES> {
> > +        const { assert!(RES == N + SHIFT) }
> > +
> > +        // SAFETY: we shift the value left by `SHIFT`, augmenting the number of bits needed to
> > +        // represent the shifted value by as much, and just asserted that `RES == N + SHIFT`.
> > +        unsafe { Bounded::__new(self.0 << SHIFT) }
> > +    }
> 
> So, it protects most significant bits when shifting left, but doesn't
> protect least significant bits when shifting right.
> 
> It also makes impossible to left-shift Bounded::<u32, 32> at all, or
> shift Bounded::<u32, 31> for 2 or more bits. This doesn't look nice.
> 
> At this layer, there's seemingly nothing wrong to loose bits during
> regular shift, just like non-bounded integers do. (Lets consider them
> naturally bounded.) At higher layers, people may add any extra checks
> as desired.
> 
> Even more, you mention you're going to use .shl and .shr for bitfield
> extraction, which means you want to loose some bits intentionally.
> 
> Let's design shifts like this. Plain .shl() and .shr() will operate on
> bounded integers just like '<<' and '>>' operate on non-bounded ones,
> i.e. they may loose bits and the result has the same bit capacity. (But
> shifting over the capacity should be be forbidden as undef).
> 
> If I want to adjust the capacity, I just do it explicitly:
> 
>     let x = Bounded::<u32, 12>::new::<0x123>();
>     let a = x.shl(4); // 12-bit 0x230
>     let b = x.extend::<12+4>().shl(4) // 16-bit 0x1230
>     let c = x.shr(4); // 12-bit 0x012
>     let d = if x & 0xf { None } else { x.shr(4).try_shrink::<12-4>() }
> 
> For b and d you can invent handy helpers, of course, and for a and c
> you can add 'safe' versions that will check shifted-out parts for
> emptiness at runtime, in case you need it. 

This 'shr' method came up from a discussion that Alexandre and I had at
plumbers. The entire point of them is that when you shift right by a
compile-time constant amount, you can change the bit capacity in the
type, and doing exactly that lets us remove BUILD_BUG_ON calls that Nova
has right now. (Right now Nova does a shift-right by a compile-time
constant, then uses BUILD_BUG_ON to assert that the resulting *runtime
value* has the desired target bit-capacity.)

So if we want a shift right method that doesn't change the bit-capacity,
we can do that. It can even use the ordinary '>>' operator. But the
purpose of this patch is precisely to support shifts that:

1. Shift by a compile-time constant amount.
2. Change the bit-capacity accordingly.

All that to say ... support for left-shifting Bounded::<u32, 32> and
similar certainly makes sense as something we could support, but that's
not what this patch is trying to achieve. Ultimately, the end-goal of
this patch is to allow us to get rid of BUILD_BUG_ON() calls operating
on runtime values, while still performing the same check that the
bit-capacity is as expected at build-time - just using types instead of
compiler optimizations to check it.

Alice