On Wed, Mar 30, 2022 at 11:48:01PM +0800, Jiaxin Yu wrote:
> On Wed, 2022-03-30 at 15:24 +0100, Mark Brown wrote:
> > On Wed, Mar 30, 2022 at 10:06:24PM +0800, Jiaxin Yu wrote:
> > > On Wed, 2022-03-30 at 13:30 +0100, Mark Brown wrote:
> > > > Making a previously optional property required means that systems
> > > > that
> > > > previously worked may stop working unless they update their DT,
> > > > DTs
> > > > may
> > > > be distributed separately to the kernel and perhaps even baked
> > > > into
> > > > firmware or similar.
> > > Thank you for your detailed answer. I should keep the driver's
> > > behavior
> > > consistent with the description of dt-bindings. The "mediatek,hdmi-
> > > codec" needs to be set as the required property. Is my
> > > understanding
> > > right?
> >
> > The binding document and code should match so if one is changed the
> > other needs to be changed too.
> >
> > In theory we should never change a previously optional property to
> > required which would mean that the code should be updated to reflect
> > the
> > binding document, however sometimes the DT isn't actually used as a
> > stable intereface by anything for a given property or device type so
> > we
> > can get away with changing things.
>
> "however sometimes the DT isn't actually used as a stable intereface by
> anything for a given property or device type so we can get away with
> changing things."
>
> Sorry, I don't understand the real idea of this description. Does it
> mean that dt-bindings in this series don't need to be updated, but the
> driver?
He means that usually the DT (and dt-binding) shouldn't be changed to avoid
incompatibilities, but sometimes it's OK to change them. For example if there
are no users of the DT yet.
But in any case, like I mentioned in my latest reply [1], I don't think changing
the dt-binding is the proper solution in this case. The driver should be changed
instead.
Thanks,
N?colas
[1] https://lore.kernel.org/all/20220330152026.6nuigsldx46lue44@notapiano