On Wed, Jul 10, 2024 at 12:16:58PM -0700, Doug Anderson wrote:
> On Wed, Jul 10, 2024 at 12:03 PM Stephan Gerhold
> <stephan.gerhold@xxxxxxxxxx> wrote:
> >
> > > 2. In theory you could make your compatible look like this:
> > >
> > > compatible = "samsung,atna45af01", "samsung,atna33xc20"
> > >
> > > ...which would say "I have a 45af01 but if the OS doesn't have
> > > anything special to do that it would be fine to use the 33xc20
> > > driver". That would allow device trees to work without the kernel
> > > changes and would allow you to land the DT changes in parallel with
> > > the driver changes and things would keep working.
> > >
> > > ...and, in fact, that would mean you _didn't_ need to add the new
> > > compatible string to the driver, which is nice.
> > >
> >
> > Yeah, I considered this. I mentioned the reason why I decided against
> > this in patch 2:
> >
> > > While ATNA45AF01 would also work with "samsung,atna33xc20" as a fallback
> > > compatible, the original submission of the compatible in commit
> > > 4bfe6c8f7c23 ("drm/panel-simple: Add Samsung ATNA33XC20") had the timings
> > > and resolution hardcoded. These would not work for ATNA45AF01.
> >
> > Basically, it works with the current driver. But if you would run the
> > kernel at the state of the original submission then it would behave
> > incorrectly. This is why I considered the resolution and timings to be
> > part of the "samsung,atna33xc20" "ABI". The new panel would not be
> > compatible with that.
>
> Ah, oops. My eyes totally glazed over the description since the patch
> was so simple. :-P Sorry about that.
>
> IMO I'd still prefer using the fallback compatible, but happy to hear
> other opinions. In the original commit things were pretty broken still
> (sorta like how it's broken for you using "edp-panel") and the
> resolution hasn't been hardcoded for a long while...
I briefly discussed this with Krzysztof on IRC yesterday and we
concluded that a fallback compatible is better. If one considers just
the non-discoverable part of the interface for the binding (i.e. the
non-standard power sequence), then the two panels are indeed compatible.
I will send a v2 with the fallback compatible on Monday. I think this
can also simplify backporting the backlight fix as you mentioned, since
then no driver change is required to make it work.
Thanks,
Stephan
