Re: [WIP 1/3] rust: Introduce atomic module

[Andrew Lunn]

On Sat, Mar 23, 2024 at 10:58:16AM +0100, Alice Ryhl wrote:
> On Sat, Mar 23, 2024 at 12:52 AM Andrew Lunn <andrew@xxxxxxx> wrote:
> >
> > > +//! These primitives should have the same semantics as their C counterparts, for precise definitions
> > > +//! of the semantics, please refer to tools/memory-model. Note that Linux Kernel Memory
> > > +//! (Consistency) Model is the only model for Rust development in kernel right now, please avoid to
> > > +//! use Rust's own atomics.
> >
> > Is it possible to somehow poison rusts own atomics?  I would not be
> > too surprised if somebody with good Rust knowledge but new to the
> > kernel tries using Rusts atomics. Either getting the compiler to fail
> > the build, or it throws an Opps on first invocation would be good.
> 
> We could try to get a flag added to the Rust standard library that
> removes the core::sync::atomic module entirely, then pass that flag.

Just looking down the road a bit, are there other features in the
standard library which are not applicable to Linux kernel space?
Ideally we want a solution not just for atomics but a generic solution
which can disable a collection of features? Maybe one by one?

And i assume somebody will try to use Rust in uboot/barebox. It
probably has similar requirements to the Linux kernel? But what about
Zephyr? Or VxWorks? Darwin?

	Andrew