Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <835d3432-daf2-e824-5c84-7d47604f586f@linaro.org>
Date:   Thu, 20 Oct 2022 13:59:40 -0400
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.4.0
Subject: Re: [PATCH v2 2/3] arm64: dts: qcom: Add base QDU1000/QRU1000 DTSIs
Content-Language: en-US
To:     Stephan Gerhold <stephan@gerhold.net>
Cc:     Melody Olvera <quic_molvera@quicinc.com>,
        Andy Gross <agross@kernel.org>,
        Bjorn Andersson <andersson@kernel.org>,
        Rob Herring <robh+dt@kernel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
        linux-arm-msm@vger.kernel.org, devicetree@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20221014221138.7552-1-quic_molvera@quicinc.com>
 <20221014221138.7552-3-quic_molvera@quicinc.com>
 <56af2a04-1b21-000d-e3f9-86b6ac74aaf2@linaro.org>
 <a0032338-482f-0de7-5952-c3c8b8423df6@quicinc.com>
 <50372a15-56ce-6ad6-f622-00624b909db8@linaro.org>
 <Y1FTJgloEi5ag2/j@gerhold.net>
 <9fbe1bf5-a84d-c56b-1c0e-6848ee3d30fe@linaro.org>
 <Y1GCs6uCz5o9az+v@gerhold.net>
From:   Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
In-Reply-To: <Y1GCs6uCz5o9az+v@gerhold.net>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 20/10/2022 13:18, Stephan Gerhold wrote:
>> No, it puts attention to the board designer that he needs to provide the
>> clock in his design.
>>
>> We had such talks about other platforms, although I do not have any
>> recent bookmarks. Something older:
>>
>> https://lore.kernel.org/all/3382034.5ADO0F7naY@wuerfel/
>>
>> https://lore.kernel.org/linux-samsung-soc/53DAB0A6.9030700@gmail.com/
>>
> 
> If I understand you correctly your argument for having the clock in the
> board DTS instead of the SoC DTSI is:
> 
> The SoC DTSI describes the components of the SoC, while the board DTS
> describes the components of the board (built around the SoC). The clock
> is part of the board (and not the SoC) and therefore belongs into the
> board DTS and not the SoC DTSI. Having the SoC/board components clearly
> separated ensures people writing new board DTS pay attention to
> everything board-specific.
> 
> Correct? This sounds reasonable to me.

Yes.

> 
> However, the main question of my previous mail was: Why do you
> alternatively recommend to keep the clock defined in the SoC DTSI and to
> just put the clock-frequency into the board DTS? This sounds like a
> contradiction of the above to me: the clock is still (partially)
> described as part of the SoC, even though it belongs to the board.

The proposal is a trade-off, compromise between code duplication and
above recommendation of splitting SoC <-> board.

As you said, all boards will provide this clock, so it is reasonable to
put some pieces of it in the SoC DTSI to avoid duplication. But the
final piece - what clock is there exactly (model, frequency etc) should
be put in board DTS to clarify that this comes with board.

> Someone writing a board DTS should not just put attention to the
> clock-frequency, but also if they have a single fixed-clock or
> maybe some kind of clock-fixed-factor setup, as I wrote.

True. If such setup is probably, then my recommendation would be to put
entire clock to the board.

Just note, that even in proposed frequency->board DTS, it is still
possible to use the clock-fixed-factor.

Best regards,
Krzysztof