Re: [WIP 0/3] Memory model and atomic API in Rust

[Mark Rutland]

On Fri, Mar 22, 2024 at 04:38:35PM -0700, Boqun Feng wrote:
> Hi,
> 
> Since I see more and more Rust code is comming in, I feel like this
> should be sent sooner rather than later, so here is a WIP to open the
> discussion and get feedback.
> 
> One of the most important questions we need to answer is: which
> memory (ordering) model we should use when developing Rust in Linux
> kernel, given Rust has its own memory ordering model[1]. I had some
> discussion with Rust language community to understand their position
> on this:
> 
> 	https://github.com/rust-lang/unsafe-code-guidelines/issues/348#issuecomment-1218407557
> 	https://github.com/rust-lang/unsafe-code-guidelines/issues/476#issue-2001382992
> 
> My takeaway from these discussions, along with other offline discussion
> is that supporting two memory models is challenging for both correctness
> reasoning (some one needs to provide a model) and implementation (one
> model needs to be aware of the other model). So that's not wise to do
> (at least at the beginning). So the most reasonable option to me is:
> 
> 	we only use LKMM for Rust code in kernel (i.e. avoid using
> 	Rust's own atomic).
> 
> Because kernel developers are more familiar with LKMM and when Rust code
> interacts with C code, it has to use the model that C code uses.

I think that makes sense; if nothing else it's consistent with how we handle
the C atomics today.

> And this patchset is the result of that option. I introduced an atomic
> library to wrap and implement LKMM atomics (of course, given it's a WIP,
> so it's unfinished). Things to notice:
> 
> * I know I could use Rust macro to generate the whole set of atomics,
>   but I choose not to in the beginning, as I want to make it easier to
>   review.
> 
> * Very likely, we will only have AtomicI32, AtomicI64 and AtomicUsize
>   (i.e no atomic for bool, u8, u16, etc), with limited support for
>   atomic load and store on 8/16 bits.
> 
> * I choose to re-implement atomics in Rust `asm` because we are still
>   figuring out how we can make it easy and maintainable for Rust to call
>   a C function _inlinely_ (Gary makes some progress [2]). Otherwise,
>   atomic primitives would be function calls, and that can be performance
>   bottleneck in a few cases.

I don't think we want to maintain two copies of each architecture's atomics.
This gets painful very quickly (e.g. as arm64's atomics get patched between
LL/SC and LSE forms).

Can we start off with out-of-line atomics, and see where the bottlenecks are?

It's relatively easy to do that today, at least for the atomic*_*() APIs:

  https://git.kernel.org/pub/scm/linux/kernel/git/mark/linux.git/commit/?h=atomics/outlined&id=e0a77bfa63e7416d610769aa4ab62bc06993ce56

.. which IIUC covers the "AtomicI32, AtomicI64 and AtomicUsize" cases you
mention above.

Mark.