Re: [PATCH 08/22] dt-bindings: display: rockchip: dw-hdmi: use "ref" as clock name

[Rob Herring <robh@kernel.org>]

On Wed, Dec 22, 2021 at 3:40 PM Heiko Stübner <heiko@sntech.de> wrote:
>
> Am Mittwoch, 22. Dezember 2021, 14:52:51 CET schrieb Rob Herring:
> > On Wed, Dec 22, 2021 at 6:47 AM Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > >
> > > On Tue, Dec 21, 2021 at 10:31:23AM -0400, Rob Herring wrote:
> > > > On Mon, Dec 20, 2021 at 12:06:16PM +0100, Sascha Hauer wrote:
> > > > > "vpll" is a misnomer. A clock input to a device should be named after
> > > > > the usage in the device, not after the clock that drives it. On the
> > > > > rk3568 the same clock is driven by the HPLL.
> > > > > To fix that, this patch renames the vpll clock to ref clock.
> > > >
> > > > The problem with this series is it breaks an old kernel with new dt. You
> > > > can partially mitigate that with stable kernel backport, but IMO keeping
> > > > the old name is not a burden to maintain.
> > >
> > > As suggested I only removed vpll from the binding document, but not from
> > > the code. The code still handles the old binding as well.
> >
> > The problem is updating rk3399.dtsi. That change won't work with old
> > kernels because they won't look for 'ref'. Since you shouldn't change
> > it, the binding needs to cover both the old and new cases.
>
> is "newer dt with old kernel" really a case these days?

I've had complaints about it. In particular from SUSE folks that were
shipping new dtbs with old (stable) kernels.

> I do understand the new kernel old dt case - for example with the
> dtb being provided by firmware.

Yes, so update your firmware that contains a newer dtb and then you
stop booting or a device stops working.

> But which user would get the idea of updating only the devicetree
> while staying with an older kernel?

Any synchronization between firmware and OS updates is a problem.

Rob