Am Donnerstag, 22. August 2024, 01:22:16 CEST schrieb Cristian Ciocaltea: > On 8/22/24 12:38 AM, Heiko Stuebner wrote: > > > > > > Am 21. August 2024 23:28:55 MESZ schrieb Conor Dooley <conor@xxxxxxxxxx>: > >> Cristian, Heiko, > >> > >> On Wed, Aug 21, 2024 at 11:38:01PM +0300, Cristian Ciocaltea wrote: > >>> On 8/21/24 6:07 PM, Conor Dooley wrote: > >>>> On Tue, Aug 20, 2024 at 11:12:45PM +0300, Cristian Ciocaltea wrote: > >>>>> On 8/20/24 7:14 PM, Conor Dooley wrote: > >>>>>> On Tue, Aug 20, 2024 at 03:37:44PM +0300, Cristian Ciocaltea wrote: > >>>>>>> On 8/19/24 7:53 PM, Conor Dooley wrote: > >>>>>>>> On Mon, Aug 19, 2024 at 01:29:30AM +0300, Cristian Ciocaltea wrote: > >>>>>>>>> + rockchip,grf: > >>>>>>>>> + $ref: /schemas/types.yaml#/definitions/phandle > >>>>>>>>> + description: > >>>>>>>>> + Most HDMI QP related data is accessed through SYS GRF regs. > >>>>>>>>> + > >>>>>>>>> + rockchip,vo1-grf: > >>>>>>>>> + $ref: /schemas/types.yaml#/definitions/phandle > >>>>>>>>> + description: > >>>>>>>>> + Additional HDMI QP related data is accessed through VO1 GRF regs. > >>>>>>>> > >>>>>>>> Why are these required? What prevents you looking up the syscons by > >>>>>>>> compatible? > >>>>>>> > >>>>>>> That is for getting the proper instance: > >>>>>> > >>>>>> Ah, that makes sense. I am, however, curious why these have the same > >>>>>> compatible when they have different sized regions allocated to them. > >>>>> > >>>>> Good question, didn't notice. I've just checked the TRM and, in both > >>>>> cases, the maximum register offset is within the 0x100 range. Presumably > >>>>> this is nothing but an inconsistency, as the syscons have been added in > >>>>> separate commits. > >>>> > >>>> Is that TRM publicly available? I do find it curious that devices sound > >>>> like they have different contents have the same compatible. In my view, > >>>> that is incorrect and they should have unique compatibles if the > >>>> contents (and therefore the programming model) differs. > >>> > >>> Don't know if there's an official location to get it from, but a quick > >>> search on internet shows a few repos providing them, e.g. [1]. > >>> > >>> Comparing "6.14 VO0_GRF Register Description" at pg. 777 with "6.15 VO1_GRF > >>> Register Description" at pg. 786 (from Part1) reveals the layout is mostly > >>> similar, with a few variations though. > >> > >> Page references and everything, thank you very much. I don't think those > >> two GRFs should have the same compatibles, they're, as you say, similar > >> but not identical. Seems like a bug to me! > >> > >> Heiko, what do you think? > > > > Yes, while the register names sound similar, looking at the bit > > definitions this evening revealed that they handle vastly different > > settings. > > > > So I guess we should fix the compatibles. They are all about graphics > > stuff and HDMI actually is the first output, so right now WE can at least > > still claim the no-users joker ;-) > > I couldn't find any driver doing a lookup for them by compatible, so I > think it's fine to fix them - should we go for "rockchip,rk3588-vo0-grf" and > "rockchip,rk3588-vo1-grf", respectively? yep. While things like the MIPICDPHY{0,1}_GRF really are identifcal and serve two separate controllers ... vo0 and vo1 are very different entities, so fixing the compatible to reflect that makes a lot of sense. > vo0_grf seems to be used by the usbdp phy nodes: > > usbdp_phy0: phy@fed80000 { > compatible = "rockchip,rk3588-usbdp-phy"; > [...] > rockchip,vo-grf = <&vo0_grf>; > [...] > > Same for "usbdp_phy1: phy@fed90000". > > While vo1_grf is present in: > > vop: vop@fdd90000 { > compatible = "rockchip,rk3588-vop"; > [...] > rockchip,vo1-grf = <&vo1_grf>; > [...] > > I guess it's too late to drop them while updating the related drivers > accordingly, hence I wonder if we should keep using the phandles for this > HDMI thing as well, for consistency reasons. For the property naming, I guess it just tells the driver which "vo"-grf to use, so the vop is more explicit in naming it vo1-grf even the vo-grf in the usbdp phy won't hurt too much. Of course we can still look up the grf by compatible and deprecate the phandle references. @Conor: just for me, did some shift happen in our understanding of dt- best-practices in terms of syscon via phandle vs. syscon via compatible? Because Rockchip boards are referencing their GRFs via phandes forever but similar to the soc vs non-soc node thing, I'd like to stay on top of best-practices ;-) Heiko > > Heiko > > > >> > >>> [1] https://github.com/FanX-Tek/rk3588-TRM-and-Datasheet > >>> > >>>>> > >>>>>>> vo0_grf: syscon@fd5a6000 { > >>>>>>> compatible = "rockchip,rk3588-vo-grf", "syscon"; > >>>>>>> reg = <0x0 0xfd5a6000 0x0 0x2000>; > >>>>>>> clocks = <&cru PCLK_VO0GRF>; > >>>>>>> }; > >>>>>>> > >>>>>>> vo1_grf: syscon@fd5a8000 { > >>>>>>> compatible = "rockchip,rk3588-vo-grf", "syscon"; > >>>>>>> reg = <0x0 0xfd5a8000 0x0 0x100>; > >>>>>>> clocks = <&cru PCLK_VO1GRF>; > >>>>>>> }; > >>>> > > >