[PATCH v2 05/10] dt-bindings: clock: renesas,r8a78000-cpg: Add firmware property

[Geert Uytterhoeven]

Hi Krzysztof,

On Wed, 12 Aug 2026 at 12:38, Krzysztof Kozlowski <krzk at kernel.org> wrote:
> On 12/08/2026 12:02, Geert Uytterhoeven wrote:
> > On Wed, 12 Aug 2026 at 11:51, Krzysztof Kozlowski <krzk at kernel.org> wrote:
> >> On 12/08/2026 11:47, Geert Uytterhoeven wrote:
> >>>>>>>>> +  firmware:
> >>>>>>>>
> >>>>>>>> RPi is doing similar, but I have doubts. It looks way too close to
> >>>>>>>> generic "syscon" which is disallowed.
> >>>>>>>>
> >>>>>>>> I think this should be vendor+purpose specific, e.g.:
> >>>>>>>> "renesas,scmi-transport" or "renesas,scmi-firmware"
> >>>>>>
> >>>>>> If you can tell me the resource that a 'firmware' provider provides then
> >>>>>> it is fine, but you can't.
> >>>>>
> >>>>> It provides a firmware API, which is not available until the firmware
> >>>>> provider is initialized.
> >>>>
> >>>> I don't know what "a firmware API" is. What does it provide? I know what
> >>>
> >>> It can provide lots of things (everything we are no longer allowed to
> >>> touch directly from Linux ;-)
> >>>
> >>>> a clock, GPIO, reset, interrupt, etc. provider is.
> >>>
> >>> A clock is an object you can access through the clock API.
> >>> (s/clock/.../ for the other cases).
> >>> Firmware is a service (virtual object) you can access through a
> >>> firmware API.
> >>>
> >>> In all cases, the consumer of the object or service needs to know
> >>> which object or service to use (there can be multiple).
> >>>
> >>> In all cases you cannot access the object or service until it is
> >>> available (initialized and registered), for which we have -EPROBE_DEFER,
> >>> and/or "[PATCH v2 02/10] of: property: fw_devlink: Add support for
> >>> firmware".
> >>>
> >>> Thanks for your understanding ;-)
> >>
> >> I think that's the difference here. Clock is not an object, but a signal
> >> on a wire or within IC coming from one device (provider) to another
> >> (consumer). Firmware is not a signal, not a wire, not anything which can
> >> be provided from one to another.
> >
> > Oh yes, it is a service provided by the secure world or a different
> > CPU core to the application core running Linux.
>
> Service is still not a resource. Not mentioning it is way too generic,
> because anything can be a service. We have entire services-based economies.

Do you have a suggestion?
This is just a phandle linking a firmware consumer to a(n optional)
firmware
provider.

What do you suggest?
Perhaps "scmi-parent" (cfr. "interrupt-parent", which also does
not point to the resources consumed, but to the provider of these
resources)?

Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds