Hi Nicolas,
> struct reg123 {
> val1 :3 // bit 31-29
> val2 :20 // bit 28-9
> val3 :9 // bit 8-0
> };
What you're describing can be modeled as Ranges in Rust:
```
use core::ops::Range;
struct Foo {
reg1: Range<u32>,
reg2: Range<u32>,
reg3: Range<u32>
}
const FOO_REGMAP: Foo = Foo {
reg1: 0..3,
reg2: 3..24,
reg3: 24..32
};
```
It becomes more useful when you pair that with a bit writer. For an example of previous art, see Faith's work: [0]
This has asserts in the right places so that you do not shoot yourself in the foot. IMHO, such a data structure can be shared with the whole Rust code in the kernel.
You can then describe your writes using the ranges, e.g.: [1]
But as we've established, instead of writing the ranges down directly, you can simply refer to them as FOO_REGMAP.reg1, FOO_REGMAP.reg2 and so on.
I believe this is both more readable and safer.
[0]: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/nouveau/compiler/bitview/lib.rs?ref_type=heads
[1]: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/nouveau/compiler/nak/encode_sm70.rs?ref_type=heads#L228
