On Fri, Dec 13, 2024 at 2:17 PM Andreas Hindborg <a.hindborg@xxxxxxxxxx> wrote: > > scheduled for removal. Interior mutability via `SyncUnsafeCell` provides > the same functionality and it is my understanding that this feature is > on track to be stabilized. I am not sure about the last bit, but even if it is on track, we do not want to start using new language features or APIs that could potentially change. And even if it is a feature that we are sure will not change, we should still let upstream Rust know before using it, since we are actively discussing with them the remaining unstable items that the kernel needs and they are putting the kernel in their roadmap. So I suggest we mention it next week in the Rust/Rust for Linux meeting. > Not sure. `val` being null not supposed to happen in the current > configuration. It should be an unreachable state. So BUG is the right thing? Since you can easily return an error in this situation, I would say ideally a `WARN_ON_ONCE` + returning an error would be the best option, and covers you in case the rest changes and someone forgets to update this. > Not in the current configuration. The parameters can only be declared > "read only". It should be impossible for anyone to call this function. What I meant is, can you avoid writing the function to begin with, by leaving a null function pointer in the `kernel_param_ops` struct, i.e. `None`? Thanks! Cheers, Miguel
