Deeper problems in the Linux kernel

Posted Jan 31, 2025 17:33 UTC (Fri) by laurent.pinchart (subscriber, #71290)
In reply to: Deeper problems in the Linux kernel by josh
Parent article: Resistance to Rust abstractions for DMA mapping

> C maintainers are free to merge C changes without taking the Rust bindings into consideration, and the Rust maintainers will subsequently fix them.

I can't see how that rule could be preserved forever as more Rust code gets merged in the kernel. I also understand how some developers and maintainers can dread the time when breaking Rust bindings would block a patch series. Has there been any discussion on when that rule would change, and how to handle the fallout ? Without clear guidelines in that area, and a process to make the transition as smooth as possible, the fear this generates will keep impacting constructive collaboration very negatively.

Deeper problems in the Linux kernel

Posted Jan 31, 2025 23:17 UTC (Fri) by josh (subscriber, #17465) [Link] (2 responses)

I would expect that rule to change if and when Linus declares the experiment a success and no longer experimental.

Deeper problems in the Linux kernel

Posted Feb 2, 2025 22:48 UTC (Sun) by laurent.pinchart (subscriber, #71290) [Link] (1 responses)

> I would expect that rule to change if and when Linus declares the experiment a success and no longer experimental.

That's what I would expect too, and is probably what most people expect as well.

The question has, as far as I can tell, not been officially addressed, and I think it is causing fears that hinder progress. I can imagine that, for a kernel developer or maintainer who wouldn't want to deal with Rust in the near or longer term (regardless of the reasons, that's not the point here), the prospect of a sudden and unilateral rule change would generate a hostile behaviour towards Rust in the kernel today. If one feels they are threatened by the project, they will have a higher incentive to make it fail.

Could the community try to address this question of long term fears ?

Deeper problems in the Linux kernel

Posted Feb 3, 2025 2:03 UTC (Mon) by dralley (subscriber, #143766) [Link]

Linus isn't going to order rewrites in Rust. He *might* eventually say "maintainers are free to choose their own policy for the subsystems they maintain".

You're describing kernel developers sabotaging the kernel for self-motivated reasons because they don't like a decision that was made. What do you want the Rust developers to address, exactly? Nothing they do or say is going to release those "fears" or make someone that selfish see the light.