On 8/31/24 4:58 PM, Heiko Stübner wrote: > Hi, > > Am Samstag, 31. August 2024, 08:16:26 CEST schrieb Krzysztof Kozlowski: >> On Sat, Aug 31, 2024 at 12:55:29AM +0300, Cristian Ciocaltea wrote: > >>> + clocks: >>> + minItems: 4 >>> + maxItems: 6 >>> + items: >>> + - description: Peripheral/APB bus clock >>> + - description: EARC RX biphase clock >>> + - description: Reference clock >>> + - description: Audio interface clock >>> + additionalItems: true >> >> What is the usefulness of all this? How can you even be sure that each >> implementation of this core will have exactly these clocks? >> >>> + >>> + clock-names: >>> + minItems: 4 >>> + maxItems: 6 >>> + items: >>> + - const: pclk >>> + - const: earc >>> + - const: ref >>> + - const: aud >>> + additionalItems: true >>> + >>> + interrupts: >>> + minItems: 4 >>> + maxItems: 5 >>> + items: >>> + - description: AVP Unit interrupt >>> + - description: CEC interrupt >>> + - description: eARC RX interrupt >>> + - description: Main Unit interrupt >> >> If these are real pins, then this seems more possible, but >> additionalItems does not make me happy. > > So while not "pins", the interrupts are separately specified in the > SoC's list of interrupts in the GIC: > > RK3588 has: > > 201 irq_hdmitx0_oavp > 202 irq_hdmitx0_ocec > 203 irq_hdmitx0_oearcrx > 204 irq_hdmitx0_omain > 392 irq_hdmitx0_hpd > > and another set of all of them for hdmitx1 > > and RK3576 using the same hdmi IP has: > > 370 irq_hdmitx_oavp > 371 irq_hdmitx_ocec > 372 irq_hdmitx_oearcrx > 373 irq_hdmitx_omain > 399 irq_hdmitx_hpd > > so I guess the fifth interrupt is meant to be the hotplug? Yep, that's for the hotplug detection. > Though I guess this should be specificed in the name-list too. My understanding from Andy was that HPD interrupt is Rockchip platform specific, hence I made it part of rockchip,rk3588-dw-hdmi-qp.yaml. > From the SoC's manual it looks like the controller is set up from > different modules. > Like AVP is the audio-video-packet-module, there is a Main and CEC Module > as well as a eARC RX controller inside. I'd guess it might be possible > other SoC vendors could leave out specific modules? > > > TL;DR I think those clocks and interrupts are dependent on how the > IP core was synthesized, so for now I'd think we can only guarantee > that they are true for rk3588 and rk3576. > > So I guess they should move to the rockchip-specific part of the binding > until we have more hdmi-qp controllers in the field? If that's the case, then we should simply drop the common binding altogether for now. Thanks, Cristian
