On Thu, May 07, 2026 at 09:24:48AM +0000, Biju Das wrote: > On 06 May 2026 20:58, Lad, Prabhakar wrote: > > On Wed, May 6, 2026 at 8:50 PM Laurent Pinchart wrote: > > > On Wed, Apr 29, 2026 at 06:00:09PM +0100, Prabhakar wrote: > > > > From: Lad Prabhakar <[email protected]> > > > > > > > > Document the Display Unit (DU) support for the RZ/T2H and RZ/N2H SoCs. > > > > > > > > The DU block on RZ/T2H is functionally equivalent to the RZ/G2UL DU > > > > and supports the DPI interface, but includes SoC-specific register > > > > differences. > > > > Add a dedicated compatible string to represent this variant. > > > > > > > > As the DU implementation on RZ/N2H matches RZ/T2H, describe it using > > > > an RZ/N2H specific compatible string with the RZ/T2H compatible as > > > > fallback. > > > > > > > > Unlike other DU variants which use a multi-port model, the RZ/T2H > > > > and RZ/N2H DU has a single output and is modelled using a single > > > > port node with one endpoint. Add a port property to support this and > > > > update the allOf constraints accordingly. > > > > > > Wouldn't it be simpler to always have a "ports" node, even for > > > variants with a single port ? > > > > > I agree that, from a binding perspective, always having a "ports" node > > keeps things simpler and > > consistent. Biju suggested this change based on earlier feedback for the > > RZ/G3E series. > > From G3E feedback, I got the impression that going forward all future SoCs > needs to have > single port and multiple endpoints. That is the reason for suggesting port > for new SoCs.
Right, let's clarify that. TL;DR: it depends on the hardware architecture (what a surprise :-)) When reviewing the G3E, I noticed that the LCDC has a single output that is connected to one or multiple encoders, depending on the SoC. I think this should be modeled in DT with a single port. Note that this does not preclude using a "ports" node, containing a single "port@0". If you're confident enough that no future generation will require multiple ports, then it makes sense to standardize on a single "port" node and no "ports". If, on the other hand, you think that some SoCs would have multiple ports, then using a top-level "ports" node unconditionally would lead to simpler bindings. I'll let you all decide what you think is the most suitable approach. -- Regards, Laurent Pinchart
