On Tuesday, March 10th, 2026 at 21:06, Krzysztof Kozlowski <[email protected]> wrote:
> On 10/03/2026 18:48, cristian_ci wrote: > > On Monday, March 9th, 2026 at 16:08, Krzysztof Kozlowski <[email protected]> > > wrote: > > > >> On 09/03/2026 15:52, cristian_ci wrote: > >>> On Sunday, March 8th, 2026 at 17:13, Krzysztof Kozlowski > >>> <[email protected]> wrote: > >>> > >>>>> + vsp-supply: > >>>>> + description: positive voltage supply for analog circuits > >>>> > >>>> Both are odd. Datasheet says vci, vddi, vddam and optional avdd, avee. > >>>> > >>>> There is no VSN and VSP. Otherwise please point the page in datasheet or > >>>> some schematics. > >>>> > >>>> Best regards, > >>>> Krzysztof > >>>> > >>> > >>> I'm not sure about that. Writing panel dt-bindings has been based pretty > >>> mostly on vendor devicetree - which also describes somehow the panel and > >>> makes that working with the final product released to the market - so > >>> I've to necessarily consider that. > >>> Then, I could agree that vendor devicetree might be not compliant with > >>> upstream rules and could possibly make mistakes with describing the > >>> hardware, so I'd like to find a way to describe that in a more proper > >>> way, according to upstream rules. > >>> > >>> That said, vendor devicetree describes lists four power supplies for > >>> DSI: 'vdd', 'vddio', 'lab' and 'ibb' (which have the following property > >>> names, respectively, in qcom,mdss_dsi_ctrl node: 'vdd-supply', > >>> 'vddio-supply', 'lab-supply' and 'ibb-supply'. > >>> Two of these are related to ds/controller (apparently, 'vddio' should > >>> match VDDI power supply in NT35532 datasheet. > >>> > >>> The remaining two supplies are related to panel ('lab' and 'ibb'). These > >>> ones are two 'external ' regulators ('external' from NT35532 > >>> perspective), which provide power supply to display, located in the qcom > >>> PMIC (in this case, that should be PMI8950). WRT to power supply names > >>> described in the bindings ('vsp-supply' and 'vsn-supply') are the same as > >>> 'lab-supply' and 'ibb-supply', just named differently in the vendor > >>> devicetrees. > >>> > >>> Usage of 'vsp'/'vsn' naming for power supply properties is grounded on > >>> they commonly being used at upstream (different panel bindings make use > >>> of these properties), on one side, and also described on schematics of > >>> devices with the same hardware configuration (LCD_VSN and LCD_VSP), on > >>> the other. > >>> > >>> In the meantime, I've found out schematics for 'xiaomi-mido' (another > >>> MSM8953 device) - a variant of this device is shipped with a panel also > >>> using NT35532 IC (just like my device) - and LCD_VSN/LCD_VSP are clearly > >>> shown there too. > >>> > >>> I couldn't find much more information about the display on my device and > >>> the only resources available about that are those listed above, as of > >>> today. In light of my reply, I ask if it is still necessary to describe, > >>> in the bindings, power supply properties properties not used currently in > >>> the board DTS file. > >> > >> Please wrap your answers so this will be possible to parse. > >> > >> You write bindings matching the hardware and for the hardware, not for > >> the downstream code. You cannot add supplies which do not exist > >> regardless what some vendor wrote somewhere > > > > Vendor has also described the hardware by storing information (by including > > info > > about panel too) directly inside the device itself (/sys/firmware/fdt). > > Vendor does not care about rules of DT thus puts there completely fake > information just to make their drivers working > . I'm not sure what that statement is based on in the specific case of the panel I'm actually testing. A set of clues (some of which were also mentioned in my first reply some days ago) point out that rimob's panel is really supplied by two regulators part of PMIC [1]. Not only for this reason, I do not have enough solid reasons, after all, to consider vendor data about the panel mounted in this device (rimob) as not decently reliable, so I cannot ignore them, at least. [1] https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/regulator/qcom-labibb-regulator.yaml > > Though vendor devicetree could possibly contain mistakes, I guess I've to > > trust vendor > > devicetree (also for the reasons explained in my last reply). > > These are not vendor's mistakes. This is deliberate WRONG information. > > Sorry, I am not going to keep discussing each property based what vendor > wrote. > > Open datasheet which I said is public. (Note: datasheet is indeed easily available but not publicly, since it has been released under NDA agreement) > > > >> and yes, you must describe > >> all known supplies for this device, especially that datasheet is > >> available publicly. > > > > Based on what you said, the following questions have raised: > > > > - have properties (mentioned by you) be defined > > (apart 'vddi', which most likely is actually 'vddio-supply', > > already defined within mdss_dsi0 node) outside of 'panel' node? > > You are mixing devices. We do not talk about vddio supply in dsi node. > We talk about THIS device, so open this device datasheet and read it. I've further inspected the datasheet again (but I've also looked at other novatek based panels dt-bindings and datasheets, in order to better figure out what is the actual relationship between the first and the second ones, more in detail). BTW, I've started doing tests, by making changes to nt35532 dt-bindings: I've replaced vsp/vsn supplies with avdd/avee (which are not optional, they are required to make my display unit to work, actually) mentioned in the datasheet (but both they rely actually on labibb regulators - indeed, vendor dt names them 'lab-supply' and 'ibb-supply') and added vci, vddi and vddam too (but I've not included 'vddam' supply into 'required' properties, since my panel unit doesn't make use of that, in order to work and I can't omit that in panel node (just like 'backlight' property). Though I've added vddi-supply to panel dt-bindings and DTS board file, at the same time I've also left vddio-supply in dsi node (even if they rely on same l6 regulator). With such dt-bindings and DTS board configuration, the panel still initializes and just works like before, and I've not noticed a different behaviour, compared to when I was using the previous board DTS (the one with vsn/vsp supplies). > > (and related to dsi/controller rather than panel, instead) > > > > - have those properties (mentioned by you) set as 'optional' > > in the bindings, rather than set as 'required'? > > Why? How device can work without power? Datasheet mentioned which > supplies can be optional, e.g. grounded or wired to the same source of > supply. Actually, the device was not working without power: required power supplies had been defined in the previous dt-bindings (vsn/vsp supplies). My idea, about the fact that panel was working, is that my unit needs positive and analog input analog power, which labibb regulators satisfy. That said, I'm not still convinced that the new power supplies properties (wich I've also tested on actual hardware), based on nt35532 datasheet, are enough to satisfy upstream rules about dt-bindings. Should I show my current changes to dt-bindings? > > (since panel works without most of them defined in the > > board DTS file) > > > Best regards, > Krzysztof >
