Re: [PATCH 1/2] media: dt-bindings: Remove assigned-clock-* from various schema

[Rob Herring <robh@kernel.org>]

On Tue, Oct 15, 2024 at 02:29:22PM +0300, Laurent Pinchart wrote:
> On Tue, Oct 15, 2024 at 08:13:19AM +0200, Krzysztof Kozlowski wrote:
> > On 14/10/2024 22:34, Laurent Pinchart wrote:
> > > On Mon, Oct 14, 2024 at 09:43:07AM +0200, Krzysztof Kozlowski wrote:
> > >> On Sat, Oct 12, 2024 at 04:02:50PM +0100, Bryan O'Donoghue wrote:
> > >>> Remove extraneous assigned-clock* from media/i2c/* schemas, retain in the
> > >>> relevant examples.
> > >>>
> > >>> Link: https://lore.kernel.org/linux-media/j7kgz2lyxnler5qwd7yiazdq6fmsv77kyozdrxf33h54ydakjz@uqjhwhoyv6re
> > >>> Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
> > >>> ---
> > >>>  Documentation/devicetree/bindings/media/i2c/hynix,hi846.yaml | 8 --------
> > >>>  Documentation/devicetree/bindings/media/i2c/ovti,ov5648.yaml | 8 --------
> > >>>  Documentation/devicetree/bindings/media/i2c/ovti,ov8865.yaml | 8 --------
> > >>>  Documentation/devicetree/bindings/media/i2c/ovti,ov9282.yaml | 4 ----
> > >>>  Documentation/devicetree/bindings/media/i2c/sony,imx258.yaml | 4 ----
> > >>>  Documentation/devicetree/bindings/media/i2c/sony,imx334.yaml | 4 ----
> > >>>  Documentation/devicetree/bindings/media/i2c/sony,imx335.yaml | 4 ----
> > >>>  Documentation/devicetree/bindings/media/i2c/sony,imx412.yaml | 4 ----
> > >>>  8 files changed, 44 deletions(-)
> > >>>
> > >>> diff --git a/Documentation/devicetree/bindings/media/i2c/hynix,hi846.yaml b/Documentation/devicetree/bindings/media/i2c/hynix,hi846.yaml
> > >>> index 60f19e1152b33128cf3baa15b8c70a874ca6d52e..d18ead8f7fc43bfacc291aed85b5ca9166c46edb 100644
> > >>> --- a/Documentation/devicetree/bindings/media/i2c/hynix,hi846.yaml
> > >>> +++ b/Documentation/devicetree/bindings/media/i2c/hynix,hi846.yaml
> > >>> @@ -28,12 +28,6 @@ properties:
> > >>>      items:
> > >>>        - description: Reference to the mclk clock.
> > >>>  
> > >>> -  assigned-clocks:
> > >>> -    maxItems: 1
> > >>> -
> > >>> -  assigned-clock-rates:
> > >>> -    maxItems: 1
> > >>> -
> > >>>    reset-gpios:
> > >>>      description: Reference to the GPIO connected to the RESETB pin. Active low.
> > >>>      maxItems: 1
> > >>> @@ -82,8 +76,6 @@ required:
> > >>>    - compatible
> > >>>    - reg
> > >>>    - clocks
> > >>> -  - assigned-clocks
> > >>> -  - assigned-clock-rates
> > >>
> > >> That's not extraneous, but has a meaning that without assigned-clocks
> > >> this device or driver will not operate.
> > > 
> > > How so ? Even if we assume that the device requires a specific clock
> > > frequency (which is often not the case for camera sensors, the
> > > limitation usually comes from drivers, so the constraint shouldn't be
> > > expressed in the bindings in that case), there is no overall requirement
> > > to assign a clock rate as in many cases the clock will come from a
> > > fixed-frequency oscillator. This seems to be a constraint that is
> > > outside of the scope of DT bindings. It is similar to regulators, where
> > > the regulator consumer doesn't have a way to express supported voltages
> > > in its DT bindings.
> > 
> > This property does not say it comes from a fixed-frequency oscillator,
> > so I do not understand why you think it is unreasonable constraint. I
> > have no clue what the author wanted to say here, so I just explained
> > that there is a meaning behind requiring such properties. If you claim
> > device or implementations do not have such requirement, after
> > investigating each case, feel free to drop this. I think I also stated
> > this already in other reply.
> 
> For camera sensor drivers I'm pretty sure we can drop those properties
> when they're marked as required, as there's no intrinsic characteristics
> of camera sensors that would require assigned-clock*.

I tend to agree, and would take it one step further to say that applies 
everywhere. Whatever clock setup needed is outside the scope of a 
binding. The simplest case is all input clocks are fixed. The next 
simplest case is firmware did all clock setup needed for a specific 
board and the boot time default works.

Rob