New Linux Kernel Patches Begin Plumbing Rust Support Into Bcachefs Driver

[phoronix]

Phoronix: New Linux Kernel Patches Begin Plumbing Rust Support Into Bcachefs Driver

There's been much excitement around the Bcachefs file-system since it was mainlined in the Linux kernel at the end of last year. Looking ahead to Linux 6.9 it's looking like it may be one of the first file-system drivers to begin making use of the Rust programming language...

https://www.phoronix.com/news/Rust-Bcachefs-RFC-Patches

[marios]

I was interested at bcachefs at first. No I am no longer interested. Well, it is better to have my expectations refuted earlier rather than later, so I can be thankful for that at least.

[V1tol]

I was interested at bcachefs at first. No I am no longer interested. Well, it is better to have my expectations refuted earlier rather than later, so I can be thankful for that at least.

Oh, another one account.

[Mathias]

I don't think Kent saw these patches before.

https://lore.kernel.org/linux-bcachefs/CALNs47u9QU2t1ug9_3_-nPzJCOR=uAHpuCT-yVd-MfWFkVVeYA@mail.gmail.com/T/#m81961a6853d4796408c75dd84b955ba5f47e9f3c

so the module_init in Rust is an interesting test case, but - not much
point in checking it in if we're not deleting code on the C side.

Instead, have you looked at pulling the btree transaction layer bindings
from -tools into the kernel? That was going to be my starting point;
once we've got that there's a lot of code we'll be able to rewrite in
Rust piecmeal (fsck, debugs, alloc_background.c - anything that
primarily interacts with the transaction layer).

[varikonniemi]

i tried to pre-empt the troll onslaught by announcing this in a previous bcachefs thread, but apparently trolls don't just hunt for bait, they professionally try to disrupt discussion

[Volta]

I was interested at bcachefs at first. No I am no longer interested. Well, it is better to have my expectations refuted earlier rather than later, so I can be thankful for that at least.

And the reason for this is? They removed a feature you wanted or?

[anarki2]

I was interested at bcachefs at first. No I am no longer interested. Well, it is better to have my expectations refuted earlier rather than later, so I can be thankful for that at least.

Oh nooo, what will they do without you?

[marios]

And the reason for this is? They removed a feature you wanted or?

No but I am afraid something important will be removed in the future. The important thing being the ability to compile the filesystem without rustc (I am never using a precompiled kernel and I do not want to have rustc just in order to compile my kernel). Also I want to point out that spreading rust has some (albeit not very important) consequences in the potential user adoption (for me it went from "I am looking forward to use it for my rock5b, when some performance issues and other bugs are solved" to "if I use it, I might regret it in the future, when the C implementation is deprecated").

Edit: Seeing Kent's response, he is all in for rewriting things in Rust. As long the C implementation is removed. This makes it almost certain that large portions of bcachefs will be RIIRed with the C implemntations thrown away, causing a rustc dependency.

[spicfoo]

No but I am afraid something important will be removed in the future. The important thing being the ability to compile the filesystem without rustc

Do you get afraid everytime a RFC is posted about what unknown things might happen in the future?

[marios]

Do you get afraid everytime a RFC is posted about what unknown things might happen in the future?

No but this time the intention is clear. RIIR, in order to ditch the C implementation once the Rust implantation is not complete crap.