Re: [PATCH] dt-bindings: arm: socionext: add bindings for the Synquacer platform

[Jassi Brar]

On Fri, 16 Jun 2023 at 15:34, Krzysztof Kozlowski
<krzysztof.kozlowski@xxxxxxxxxx> wrote:
>
> On 16/06/2023 22:06, Jassi Brar wrote:
> > On Fri, 16 Jun 2023 at 11:47, Krzysztof Kozlowski
> > <krzysztof.kozlowski@xxxxxxxxxx> wrote:
> >>
> >> On 16/06/2023 18:23, Jassi Brar wrote:
> >>> On Fri, 16 Jun 2023 at 05:15, Krzysztof Kozlowski
> >>> <krzysztof.kozlowski@xxxxxxxxxx> wrote:
> >>>>
> >>>> On 16/06/2023 05:58, jaswinder.singh@xxxxxxxxxx wrote:
> >>>>> From: Jassi Brar <jaswinder.singh@xxxxxxxxxx>
> >>>>>
> >>>>> Socionext's DeveloperBox is based on the SC2A11B SoC (Synquacer).
> >>>>> Specify bindings for the platform and boards based on that.
> >>>>
> >>>> A nit, subject: drop second/last, redundant "bindings". The
> >>>> "dt-bindings" prefix is already stating that these are bindings.
> >>>>
> >>> I can remove it, but I see many mentions like "Fix bindings for"  "Add
> >>> binding for" etc in the subject line.
> >>
> >> Can we fix them as well?
> >>
> > ??
> What else I can say to such argument?
>
It was not an argument, I agreed to remove it. I just observed that
the nit-pick was arbitrary.
And frankly
   "dt-bindings: arm: socionext: add Synquacer"   is as misleading as
   "dt-bindings: arm: socionext: add bindings for the Synquacer"   is improper.


> >>>>
> >>>> Binding without it's user is usually useless. Where is the user?
> >>>>
> >>> It is required for SystemReady-2.0 certification.
> >>
> >> For what? If there is no user, it is not required for SR. We don't
> >> document compatibles for something which does not exist in the projects.
> >>
> > The dts/dtsi for synquacer will be added later.
> > I am sure you are aware that there are countless bindings without
> > actual use in any dts/dtsi.
>
> Bindings without user (so no DTSI and no driver)? Just few, not countless.
>
I disagree. But I don't have time to write a script to find
compatibles/enums and properties in yaml/txt files that are not in any
dts/dtsi file.
 By that logic synquacer's spi/netsec/i2c/exiu bindings and drivers in
kernel are illegit too?

Also the user may not be in Linux, but we keep "os-agnostic" bindings in Linux.
The synquacer dts/dtsi are in u-boot upstream. SR testsuite looks up
the underlying platform name and checks if the bindings are merged
upstream.
While I am not against also submitting dts/dtsi in linux, I don't
think the binding should be held at ransom.

> > When exactly did it become mandatory to
> > have dts/dtsi for the bindings to be merged upstream?
>
> It was always. We do not want/need to document downstream stuff or
> anything  just because it is somewhere there.
>
I am not asking you to merge an obscure internal revision of some SoC.
Synquacer is a public development platform and a "96board" already
certified for SR-1.0.

thnx.