On Sat, Sep 13, 2025 at 04:57:26PM +0000, Rahul Rameshbabu wrote:
> On Wed, 27 Aug, 2025 08:57:58 +0200 "Maxime Ripard" <[email protected]> 
> wrote:
> > Hi Rahul,
> >
> > On Wed, Aug 20, 2025 at 04:46:52AM +0000, Rahul Rameshbabu wrote:
> >> On Tue, 19 Aug, 2025 11:06:40 +0200 "Maxime Ripard" <[email protected]> 
> >> wrote:
> >> > Hi Rahul,
> >> >
> >> > On Mon, Aug 18, 2025 at 05:04:15AM +0000, Rahul Rameshbabu wrote:
> >> >> I am working on a drm_connector scoped backlight API in Rust. I have 
> >> >> been
> >> >> looking through Hans de Goede's previous efforts on this topic to help
> >> >> guide my design. My hope is to enable backlight control over external
> >> >> displays through DDC or USB Monitor Control Class while also supporting
> >> >> internal panels. In parallel, I would like to improve the driver
> >> >> probing/selection mechanism when there are different candidates for 
> >> >> driving
> >> >> a backlight device. This initial RFC is mainly intended to sanity check
> >> >> that the plumbing I have chosen for extending the DRM connector
> >> >> functionality in Rust seems reasonable.
> >> >
> >> > It's a great goal, and I had that same discussion with Hans recently
> >> > too, but I can't find the link between backling/DDC CI, and Rust. Can
> >> > you elaborate?
> >> 
> >> Hi Maxime,
> >> 
> >> Sure, let me elaborate on this. You are right that plumbing DDC
> >> CI/backlight support at the DRM connector level does not need to be
> >> implemented in Rust.
> >> 
> >> If we look at Hans's proposal, the suggested phase 2 was to add a
> >> drm_connector helper function for plumbing a pointer to the backlight
> >> device implementation. I had some model differences with regards to how
> >> the API would look like, mostly stemming from concerns about providing
> >> better runtime overriding of the acpi_video_get_backlight_type based
> >> backlight selection. However, I am aligned with the direction of scoping
> >> at the drm_connector level. I basically was interested in implementing
> >> this helper functionality in Rust instead of C, which is where Rust came
> >> into play.
> >> 
> >> I was also interested in declaring and attaching a drm_property in Rust
> >> for controlling properties such as backlight rather than updating the
> >> drm_connector declaration in C as an experiment.
> >> 
> >> Let me know if you feel like this work would be better off as a C
> >> implementation. I can also send out a detailed architecture proposal to
> >> the mailing list if that would help.
> >> 
> >> Link: 
> >> https://lore.freedesktop.org/wayland-devel/[email protected]/
> >
> > Thanks for the explanation.
> >
> > I'm not sure Rust is at the point where we can use it for the framework.
> > If we want to make this work useful, we have to make it consistent and
> > usable across all drivers, but we do have drivers for architectures that
> > aren't supported by Rust yet (let alone tier 1).
> >
> > So it feels to me that it would be a bit premature for that work to be
> > in Rust. If you do want to use it from a Rust driver though, feel free
> > to write bindings for it, that would be a great addition.
> 
> Hi Maxime,
> 
> Thanks for the follow-up. Sorry for the delay in my response. I was
> preparing a slide deck for Kangrejos 2025 (Rust for Linux conference).
> 
> https://binary-eater.github.io/kangrejos-2025/
> 
> The above discusses the architecture I had in mind in greater detail. I
> am working on some last minute tweaks. I wanted to do a couple things
> with regards to this topic.
> 
> 1. Send a high level RFC describing the architecture / functionality
> 2. In parallel, maybe further evaluate whether Rust could be viable for
>    this effort. I hope the slides I put together help.
> 3. If the discussion in point 2 seems to suggest that Rust is not
>    viable, do the core implementation work in C.
> 
> Let me know if this seems like a reasonable approach and thank you so
> much for taking the time to respond.

The thing I was trying to say before is that we have an inherent tension
here: it's an interface we want to roll as widely as possible, otherwise
userspace is going to either ignore it, or the kernel wouldn't provide a
consistent interface for this.

We do have drivers that are for architectures that don't have rust
support yet.

I still think it's premature to have code used by all those drivers in
Rust just yet.

Maxime

Attachment: signature.asc
Description: PGP signature

Reply via email to