Re: [PATCH v7 4/9] rust: sync: atomic: Add generic atomics

["Benno Lossin" <lossin@kernel.org>]

On Mon Jul 14, 2025 at 4:21 PM CEST, Boqun Feng wrote:
> On Mon, Jul 14, 2025 at 12:30:12PM +0200, Benno Lossin wrote:
>> On Mon Jul 14, 2025 at 7:36 AM CEST, Boqun Feng wrote:
>> > To provide using LKMM atomics for Rust code, a generic `Atomic<T>` is
>> > added, currently `T` needs to be Send + Copy because these are the
>> > straightforward usages and all basic types support this.
>> >
>> > Implement `AllowAtomic` for `i32` and `i64`, and so far only basic
>> > operations load() and store() are introduced.
>> >
>> > Reviewed-by: Alice Ryhl <aliceryhl@google.com>
>> > Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
>> > ---
>> >  rust/kernel/sync/atomic.rs         |  14 ++
>> >  rust/kernel/sync/atomic/generic.rs | 285 +++++++++++++++++++++++++++++
>> >  2 files changed, 299 insertions(+)
>> >  create mode 100644 rust/kernel/sync/atomic/generic.rs
>> >
>> > diff --git a/rust/kernel/sync/atomic.rs b/rust/kernel/sync/atomic.rs
>> > index e80ac049f36b..c5193c1c90fe 100644
>> > --- a/rust/kernel/sync/atomic.rs
>> > +++ b/rust/kernel/sync/atomic.rs
>> > @@ -16,7 +16,21 @@
>> >  //!
>> >  //! [`LKMM`]: srctree/tools/memory-model/
>> >  
>> > +pub mod generic;
>> 
>> Hmm, maybe just re-export the stuff? I don't think there's an advantage
>> to having the generic module be public.
>> 
>
> If `generic` is not public, then in the kernel::sync::atomic page, it
> won't should up, and there is no mentioning of struct `Atomic` either.
>
> If I made it public and also re-export the `Atomic`, there would be a
> "Re-export" section mentioning all the re-exports, so I will keep
> `generic` unless you have some tricks that I don't know.

Just use `#[doc(inline)]` :)

    https://doc.rust-lang.org/rustdoc/write-documentation/the-doc-attribute.html#inline-and-no_inline

> Also I feel it's a bit naturally that `AllowAtomic` and `AllowAtomicAdd`
> stay under `generic` (instead of re-export them at `atomic` mod level)
> because they are about the generic part of `Atomic`, right?

Why is that more natural? It only adds an extra path layer in any import
for atomics.

Unless you at some point want to add `concrete::Atomic<T>` etc, I would
just re-export them.

>> > +/// The atomic operations are implemented in a way that is fully compatible with the [Linux Kernel
>> > +/// Memory (Consistency) Model][LKMM], hence they should be modeled as the corresponding
>> > +/// [`LKMM`][LKMM] atomic primitives. With the help of [`Atomic::from_ptr()`] and
>> > +/// [`Atomic::as_ptr()`], this provides a way to interact with [C-side atomic operations]
>> > +/// (including those without the `atomic` prefix, e.g. `READ_ONCE()`, `WRITE_ONCE()`,
>> > +/// `smp_load_acquire()` and `smp_store_release()`).
>> > +///
>> > +/// [LKMM]: srctree/tools/memory-model/
>> > +/// [C-side atomic operations]: srctree/Documentation/atomic_t.txt
>> 
>> Did you check that these links looks good in rustdoc?
>> 
>
> Yep.

Nice :)

>> > +/// over unit-only enums, see [Examples].
>> > +///
>> > +/// # Limitations
>> > +///
>> > +/// Because C primitives are used to implement the atomic operations, and a C function requires a
>> > +/// valid object of a type to operate on (i.e. no `MaybeUninit<_>`), hence at the Rust <-> C
>> > +/// surface, only types with no uninitialized bits can be passed. As a result, types like `(u8,
>> 
>> s/no uninitialized/initialized/
>> 
>
> hmm.. "with initialized bits" seems to me saying "it's OK as long as
> some bits are initialized", how about "with all the bits initialized"?

Sounds good. The double negation sounded a bit weird to me.

>> > +    /// However, this should be only used when communicating with C side or manipulating a C
>> > +    /// struct.
>> 
>> This sentence should be above the `Safety` section.
>> 
>
> Hmm.. why? This is the further information about what the above
> "Examples" section just mentioned?

I thought "this" is referring to "this function", if not then please be
more concrete :)

---
Cheers,
Benno