Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Subject: Re: [PATCH] arm64: dts: qcom: sm8450: delete incorrect ufs
 interconnect fields
To:     Bjorn Andersson <bjorn.andersson@linaro.org>
Cc:     Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>,
        Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>,
        Andy Gross <agross@kernel.org>,
        Rob Herring <robh+dt@kernel.org>,
        Krzysztof Kozlowski <krzk+dt@kernel.org>,
        Vinod Koul <vkoul@kernel.org>,
        "open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" 
        <devicetree@vger.kernel.org>,
        open list <linux-kernel@vger.kernel.org>,
        linux-arm-msm@vger.kernel.org
References: <20220407172145.31903-1-jonathan@marek.ca>
 <e41c26c2-8aa4-cfd2-27b0-eb011f45eda0@linaro.org>
 <865ff6bc-95a2-8b39-5cf2-bb2d3f592c5a@linaro.org>
 <f1fb2d71-4adf-bcc7-76b3-c7102ab9f2e9@marek.ca>
 <YlTg7QPkWMBP4HAb@builder.lan>
From:   Jonathan Marek <jonathan@marek.ca>
Message-ID: <20db508c-0ccf-e4a6-87a4-17c41871703c@marek.ca>
Date:   Tue, 12 Apr 2022 00:04:23 -0400
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.2.2
MIME-Version: 1.0
In-Reply-To: <YlTg7QPkWMBP4HAb@builder.lan>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 4/11/22 10:16 PM, Bjorn Andersson wrote:
> On Thu 07 Apr 17:38 CDT 2022, Jonathan Marek wrote:
> 
>> On 4/7/22 5:16 PM, Krzysztof Kozlowski wrote:
>>> On 07/04/2022 21:40, Vladimir Zapolskiy wrote:
>>>> On 4/7/22 20:21, Jonathan Marek wrote:
>>>>> Upstream sm8450.dtsi has #interconnect-cells = <2>; so these are wrong.
>>>>> Ignored and undocumented with upstream UFS driver so delete for now.
>>>
>>> This is the upstream and they are documented here, although as pointed
>>> by Vladimir this was rather a reverse-documentation. The documentation
>>> might be incorrect, but then the bindings should be corrected instead of
>>> only modifying the DTS.
>>>
>>>>
>>>> Basically the description was added by a commit 462c5c0aa798 ("dt-bindings: ufs:
>>>> qcom,ufs: convert to dtschema").
>>>>
>>>> It's questionable, if an example in the new yaml file is totally correct
>>>> in connection to the discussed issue.
>>>
>>> To be honest - the example probably is not correct, because it was based
>>> on existing DTS without your patch. :)
>>>
>>> Another question is whether the interconnect properties are here correct
>>> at all. I assumed that DTS is correct because it should describe the
>>> hardware, even if driver does not use it. However maybe that was a false
>>> assumption...
>>>
>>
>> writing-bindings.rst says it is OK to document even if it isn't used by the
>> driver (seems wrong to me, at least for interconnects which are a firmware
>> abstraction and not hardware).
>>
> 
> The devicetree, and hence the binding, should describe the hardware, so
> that an implementation can make use of the hardware. So there's no
> problem expressing the interconnect in the binding/dts even though the
> driver isn't using it.
> 
> I'm not sure if I'm misunderstanding you, the interconnect paths
> described here are a description of the hardware requirements for this
> device. (I.e. it need the buses between ufs and ddr, and cpu and ufs to
> operate).
> 

This is pedantic but what if my kernel lives in imem and not ddr. Or it 
runs on adsp and not cpu? "ufs-ddr" and "cpu-ufs" are not necessarily 
hardware requirements.

(I was thinking of something else when I wrote that comment, but it 
doesn't actually matter if its firmware/hardware if a driver can 
implement the same functionality either way)

>> 462c5c0aa798 wasn't in my 5.17+ tree pulled after dts changes were merged (I
>> guess doc changes come later), so my commit message is incorrect, but I
>> think it makes more sense to have the documentation reflect the driver. Its
>> also not an important issue, so I'll let others sort it out.
>>
> 
> I believe that the correctness of the interconnect property will ensure
> that the interconnect provider doesn't hit sync_state until the ufs
> driver has probed - regardless of the driver actually implementing the
> interconnect voting. But perhaps I've misunderstood the magic involved?
> 

AFAICT, if its not used by the driver it will be ignored completely, 
unless you use OPP (which has some interconnect magic).

> Regards,
> Bjorn
>