Rust 1.78 Upgrade For Linux 6.10, Dropping In-Tree "alloc" Fork To Save ~10k Lines

[phoronix]

Phoronix: Rust 1.78 Upgrade For Linux 6.10, Dropping In-Tree "alloc" Fork To Save ~10k Lines

There's another Rust upgrade coming for the Linux 6.10 kernel to bump the Rust version baseline required for building the Rust in-tree kernel components. This raising of the baseline will continue until a suitable minimum version is achieved where official Rust compiler "just works" well with the Rust'ed kernel bits. The Rust upgrade in Linux 6.10 also does away with its in-tree "alloc" fork for big code savings and simplifying maintenance...

https://www.phoronix.com/news/Rust-For-Linux-6.10

[timofonic]

What's wrong with Rust in kernels?

[bacteriamanicure]

What's wrong with Rust in kernels?

The main issue is that there's a number of missing features in current Rust that makes kernel integration a lot harder than it could be

https://github.com/Rust-for-Linux/linux/issues/354

From what I can tell they tend to be ergonomics issues, not being able to squeeze out enough performance, or things that need to be guaranteed but aren't

[intelfx]

What's wrong with Rust in kernels?

Nothing wrong, it's just work in progress. Either look at it, or don't, and observe the finished result when the time comes.

[avis]

<Offtopic>
Michael

For your attention: https://forums.developer.nvidia.com/...rietary/292698

</Offtopic>

[reba]

<Offtopic>
Michael

For your attention: https://forums.developer.nvidia.com/...rietary/292698

</Offtopic>

Good questions and for me it sadly shows they have no / prevent a more open strategy. Honestly I am quite confused what that is all about with there being so much open, closed variance and corner cuttings, all the different drivers for a specific use and questionable compatibility with each other and the hardware. It's a total mess with no clear course to me.

[sophisticles]

I'm just going to say it, I wish the kernel developers would stop the flirting with Rust that they have been engaging in and either port the kernel completely over to Rust or stick with C.

But pick a lane already and move on.

[bug77]

I'm just going to say it, I wish the kernel developers would stop the flirting with Rust that they have been engaging in and either port the kernel completely over to Rust or stick with C.

But pick a lane already and move on.

You can't even begin to discuss moving over (which will never happen) until you have proven viability first.

[darkonix]

I'm just going to say it, I wish the kernel developers would stop the flirting with Rust that they have been engaging in and either port the kernel completely over to Rust or stick with C.

But pick a lane already and move on.

It is not as easy. Too much people and volume. But they are porting the kernel. The experimental support of Rust if for developing modules for the kernel. It is a very clean approach to experiment with Rust without changing the kernel itself.

[dralley]

I'm just going to say it, I wish the kernel developers would stop the flirting with Rust that they have been engaging in and either port the kernel completely over to Rust or stick with C.

But pick a lane already and move on.

Even if they committed to completely rewriting the kernel in Rust tomorrow, it wouldn't happen for 20 years.

The current path (focusing on drivers) provides plenty of value. Drivers are mostly self-contained (changing them has less fallout on the rest of the kernel) and have less eyes on them (easier for simple mistakes to creep in undetected)