Re: [PATCH v3 07/11] dt-bindings: clock: Add StarFive JH7110 system clock and reset generator

[Conor Dooley <conor@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 3333 bytes --]

On Sat, Feb 18, 2023 at 03:55:25PM +0100, Krzysztof Kozlowski wrote:
> On 18/02/2023 12:17, Conor Dooley wrote:
> > On Sat, Feb 18, 2023 at 11:20:30AM +0100, Krzysztof Kozlowski wrote:

> >>>> That's a long thread, please summarize what you ask. Otherwise I have no
> >>>> clue what is the question.
> >>>
> >>> Sorry. I tried to preserve the context of the conversation the last time
> >>> I cropped it so that things would be contained on one email.
> >>>
> >>> For me at least, I am wondering how you convey that out of a list of
> >>> clock inputs (for example a, b, c, d) that two of the clocks are inputs
> >>> to a mux and it is only required to provide one of the two (say b & c).
> > 
> > You skipped this part which was what I was trying to ask you about.
> 
> Yeah, I skipped a lot because there was one big thread with a question:
> what do you think? Sorry, I will not dig 8 emails thread to figure out
> which question is to me and which is not...

This was in the cut-down message & a fake scenario to avoid you needing
to understand the full thread.
I kept the context originally so that you would not have to dig out the
thread & provided the fake scenario this time for this very reason.

> > Do you know how to convey this situation, or is it even possible to
> > express those rules?
> 
> oneOf:
>  - clock-names:
>      minItems: 3
>      items:
>        - a
>        - b
>        - c
>        - d
>  - clock-names:
>      items:
>        - a
>        - b
>        - d
> 
> or maybe:
>  - clock-names:
>      minItems: 3
>      items:
>        - a
>        - b
>        - enum: [c, d]
>        - d

Thanks for the suggestions. Without actually going and playing with it,
the first of those two looks like it may be the right fit for this
situation, depending on what Hal says the hardware can do.

> >>>> Does the mux works correctly if clock input is not connected? I mean,
> >>>> are you now talking about real hardware or some simplification from SW
> >>>> point of view?
> >>>
> >>> I'm coming at this from an angle of "is a StarFive customer going to show
> >>> up with a devicetree containing dummy fixed-clocks to satisfy dtbs_check
> >>> because they opted to only populate one input to the mux".
> >>> I don't really care about implications for the driver, just about
> >>> whether the hardware allows for inputs to the mux to be left
> >>> un-populated.
> >>
> >> Whether hardware allows - not a question to me.
> > 
> >> BTW, this is rather question coming from me...
> > 
> > I don't understand what you mean by this, sorry.
> 
> You said to a letter addressed to me "whether the hardware allows for
> ...". Why would you ask me about hardware I know nothing about? That was
> my question - I am asking - whether hardware allows it or not. Then
> write bindings depending on that.

There was no question here, instead it was an answer to this question of
yours:
> I mean,
> are you now talking about real hardware or some simplification from SW
> point of view?

In which I was saying I cared about the "real hardware". For obvious
reasons, I wouldn't ask you to explain whether the hardware is capable
of it or not!
Perhaps your original question here was misunderstood.

Thanks for the suggestions,
Conor.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]