On Thu, Apr 8, 2021 at 12:14 AM Fabien Parent <fparent@xxxxxxxxxxxx> wrote:
> On Wed, Apr 7, 2021 at 7:55 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
> > On Wed, Apr 7, 2021 at 6:34 PM Matthias Brugger <matthias.bgg@xxxxxxxxx> wrote:
> > > On 07/04/2021 18:06, Arnd Bergmann wrote:
> >
> > So five of the strings are documented, the others are missing. I did not check
> > the other patches in your branch.
>
> The binding documentation for these drivers are here [0]. The display
> bindings are documented as:
> - compatible: "mediatek,<chip>-disp-<function>", one of
>
> The <chip> placeholder is never expanded for all the supported chips.
> The 5 existings matches from your grep command comes from the example.
> I guess these will be fixed whenever someone converts [0] to yaml.
Ok. I suppose the wildcards just didn't get caught in the initial review of the
binding. The way the binding is defined is not all that helpful since the entire
point of having chip specific strings is to allow having different bindings for
future chips that have different requirements.
There is still an open question on what to do to replace the aliases. At least
since they are not part of the documented binding, it is fairly easy to argue
that the drivers should not rely on them, and we can still change them.
I also see that as late as last november, there were still incompatible
code changes to the ad-hoc binding in
drivers/gpu/drm/mediatek/mtk_drm_ddp_comp.c, so I'm not too worried
about breaking existing dts files that relied on it.
I don't claim to understand how the various blocks all fit together here,
but I would expect that this can all be replaced with just having
references to phandles for the other nodes in one place.
Are the aliases in this case actually board specific, or do they just
document how the SoC is wired up?
Arnd
