Comment by fluoridation

10 months ago

As I read it, "the losing side of history" refers to insisting on using C, possibly at all. The last part about the "world moving forward towards memory-safe languages" doesn't suggest a limited scope for the statement.

The thread was not about Rust drivers, it was about adding Rust code to the DMA module. I.e. about mixing two different languages in a single module, thus requiring being knowledgeable about both languages in order to maintain it, thus making the module less maintainable. In fact, a few developers were saying that they didn't mind Rust drivers, if they used the C ABI as-is. Someone wanted to expose new Rust-specific interfaces to support cleaner abstractions from Rust drivers.

4 comments

fluoridation

mustache_kimono 10 months ago

> The thread was not about Rust drivers, it was about adding Rust code to the DMA module. I.e. about mixing two different languages in a single module

AFAIK this is false. The patch was CCed to the maintainer as FYI, but all the code was in a Rust a module binding to the C DMA interface. If I'm wrong, show me the code.

See the discussion here: https://lkml.org/lkml/2025/1/9/398

fluoridation 10 months ago
I'm just going by what was mentioned in the thread. If that interpretation is wrong, the thread makes no sense.
- mustache_kimono 10 months ago
  
  Fine. But, when I say something is out in out false, it might be worth checking your facts.
  I'm willing to grant that it is possible Christoph Hellwig simply misunderstood the patch and overreacted.
  rust/bindings/bindings_helper.h | 1 + rust/kernel/dma.rs | 271 ++++++++++++++++++++++++++++++++ rust/kernel/error.rs | 1 + rust/kernel/lib.rs | 1 + 4 files changed, 274 insertions(+) create mode 100644 rust/kernel/dma.rs
  See: https://lkml.org/lkml/2025/1/8/801
  See: https://lkml.org/lkml/2025/1/9/398
  See: https://lkml.org/lkml/2025/1/10/619
  See: https://lkml.org/lkml/2025/1/29/999
- dralley 10 months ago
  
  >I'm just going by what was mentioned in the thread. If that interpretation is wrong, the thread makes no sense.
  You've now discovered why this blew up in the first place. All of the excuses used to reject the code were not just petty but also outright false, and trivially so.