On Tue, Jul 09, 2019 at 01:30:18PM -0700, Vasily Khoruzhick wrote: > On Tue, Jul 9, 2019 at 1:55 AM Maxime Ripard wrote: > > > > On Mon, Jul 08, 2019 at 05:49:21PM -0700, Vasily Khoruzhick wrote: > > > > > Maybe instead of edp-connector one would introduce integrator's specific > > > > > connector, for example with compatible "olimex,teres-edp-connector" > > > > > which should follow edp abstract connector rules? This will be at least > > > > > consistent with below presentation[1] - eDP requirements depends on > > > > > integrator. Then if olimex has standard way of dealing with panels > > > > > present in olimex/teres platforms the driver would then create > > > > > drm_panel/drm_connector/drm_bridge(?) according to these rules, I guess. > > > > > Anyway it still looks fishy for me :), maybe because I am not > > > > > familiarized with details of these platforms. > > > > > > > > That makes sense yes > > > > > > Actually, it makes no sense at all. Current implementation for anx6345 > > > driver works fine as is with any panel specified assuming panel delays > > > are long enough for connected panel. It just doesn't use panel timings > > > from the driver. Creating a platform driver for connector itself looks > > > redundant since it can't be reused, it doesn't describe actual > > > hardware and it's just defeats purpose of DT by introducing > > > board-specific code. > > > > I'm not sure where you got the idea that the purpose of DT is to not > > have any board-specific code. > > I believe DT was an attempt to move to declarative approach for > describing hardware. Yes, we have different compatibles for different > devices but they're specific to particular device rather than > particular board. Device interconnection is described in DT along with > some properties rather than in board-specific C-file. You're right, but it's not incompatible with having some code to deal with some board quirk. > Introducing board-specific compatible for a connector isn't looking > right to me. If that board has a board-specific behaviour for it's connector, then what's the issue? You can't describe all the quirks in the all boards using purely properties. > > It's perfectly fine to have some, that's even why there's a compatible > > assigned to each and every board. > > > > What the DT is about is allowing us to have a generic behaviour that > > we can detect: we can have a given behaviour for a given board, and a > > separate one for another one, and this will be evaluated at runtime. > > > > This is *exactly* what this is about: we can have a compatible that > > sets a given, more specific, behaviour (olimex,teres-edp-connector) > > while saying that this is compatible with the generic behaviour > > (edp-connector). That way, any OS will know what quirk to apply if > > needed, and if not that it can use the generic behaviour. > > > > And we could create a generic driver, for the generic behaviour if > > needed. > > > > > There's another issue: if we introduce edp-connector we'll have to > > > specify power up delays somewhere (in dts? or in platform driver?), so > > > edp-connector doesn't really solve the issue of multiple panels with > > > same motherboard. > > > > And that's what that compatible is about :) > > Sorry, I fail to see how it would be different from using existing > panels infrastructure and different panels compatibles. I think Rob's > idea was to introduce generic edp-connector. Again, there's no such thing as a generic edp-connector. The spec doesn't define anything related to the power sequence for example. > If we can't make it generic then let's use panel infrastructure. Which uses a device specific compatible. Really, I'm not sure what your objection and / or argument is here. In addition, when that was brought up in the discussion, you rejected it because it was inconvenient: https://patchwork.freedesktop.org/patch/283012/?series=56163&rev=1#comment_535206 And I agree with you on that one. > > > I'd say DT overlays should be preferred solution here, not another > > > connector binding. > > > > Overlays are a way to apply a device tree dynamically. It's orthogonal > > to the binding. > > It isn't orthogonal to original problem though. It is. The original problem is that you want to power up whatever is on the other side of a eDP link using an arbitrary regulator. This is a "how do I describe that in my DT" problem, and it really has nothing to do with how the DT is being passed to the kernel. Maxime -- Maxime Ripard, Bootlin Embedded Linux and Kernel engineering https://bootlin.com