[PATCH 01/11] dt-bindings: firmware: arm,scmi: Document assigned-clocks and assigned-clock-rates

Cristian Ciocaltea cristian.ciocaltea at collabora.com
Fri Mar 17 10:21:03 PDT 2023


On 3/17/23 16:27, Rob Herring wrote:
> On Fri, Mar 17, 2023 at 4:59 AM Cristian Ciocaltea
> <cristian.ciocaltea at collabora.com> wrote:
>>
>> On 3/17/23 00:26, Sudeep Holla wrote:
>>> On Thu, Mar 16, 2023 at 03:34:17PM -0500, Rob Herring wrote:
>>>> +Stephen
>>>>
>>>> On Wed, Mar 15, 2023 at 01:47:56PM +0200, Cristian Ciocaltea wrote:
>>>>> Since commit df4fdd0db475 ("dt-bindings: firmware: arm,scmi: Restrict
>>>>> protocol child node properties") the following dtbs_check warning is
>>>>> shown:
>>>>>
>>>>>     rk3588-rock-5b.dtb: scmi: protocol at 14: Unevaluated properties are not allowed ('assigned-clock-rates', 'assigned-clocks' were unexpected)
>>>>
>>>> I think that's a somewhat questionable use of assigned-clock-rates. It
>>>> should be located with the consumer rather than the provider IMO. The
>>>> consumers of those 2 clocks are the CPU nodes.
>>>>
>>>
>>> Agreed. We definitely don't use those in the scmi clk provider driver.
>>> So NACK for the generic SCMI binding change.
>>
>> According to [1], "configuration of common clocks, which affect multiple
>> consumer devices can be similarly specified in the clock provider node".
> 
> True, but in this case it's really a single consumer because it's all
> CPU nodes which are managed together.
> 
>> That would avoid duplicating assigned-clock-rates in the CPU nodes.
> 
> Wouldn't one node be sufficient?

Yeah, that should be fine.

> Thinking more about this, why aren't you using OPP tables to define
> CPU frequencies. Assigned-clocks looks like a temporary hack because
> you haven't done proper OPP tables.

Right, this is currently not possible since it depends on some work in 
progress.

Thanks,
Cristian



More information about the linux-riscv mailing list