Re: [PATCH v1] dt-bindings: clock: convert rockchip,rk3188-cru.txt to YAML

From: Krzysztof Kozlowski <krzk@kernel.org>
To: Stephen Boyd <sboyd@kernel.org>, Johan Jonker <jbx6244@gmail.com>,
	heiko@sntech.de
Cc: robh+dt@kernel.org, krzk+dt@kernel.org, mturquette@baylibre.com,
	linux-clk@vger.kernel.org, devicetree@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-rockchip@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v1] dt-bindings: clock: convert rockchip,rk3188-cru.txt to YAML
Date: Fri, 25 Mar 2022 08:31:25 +0100	[thread overview]
Message-ID: <bf438af8-5969-73e4-009d-cb7d93095a5e@kernel.org> (raw)
In-Reply-To: <20220325005130.C45A3C340EC@smtp.kernel.org>

On 25/03/2022 01:51, Stephen Boyd wrote:
> Quoting Johan Jonker (2022-03-24 12:51:36)
>> Hi Heiko, Krzysztof,
>>
>> Question for the Rockchip clock maintainer:
>> What clock should be used here and other SoCs with several clock parents
>> in the tree?
>>
>> The clock.yaml produces a lot off notifications like:
>>
>> /arch/arm/boot/dts/rk3036-evb.dtb: clock-controller@20000000: 'clocks'
>> is a dependency of 'assigned-clocks'
> 
> 'clocks' is not a dependency of 'assigned-clocks'. The dt-schema should
> be fixed to remove that requirement.

If the driver does not have any clock inputs ("clocks" property), why
does it care about some clock frequencies and parents?

The clocks is the logical dependency of assigned-clocks, because
otherwise hardware description is not complete.

What should be here for Rockhip? We had similar cases like this for many
drivers, I was fixing some of Exynos as well. In my case usually the
root/external clock was missing, so I supplied is as input clock to the
clock controller.

Best regards,
Krzysztof