linux-kernel - Re: [PATCH v2 1/4] dt-bindings: clock: qcom,sc7280-lpasscc: Remove qdsp6ss reg property

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <3267da40-5c53-0b79-fd65-3009ee17c7ee@linaro.org>
Date:   Tue, 3 Jan 2023 09:57:03 +0100
From:   Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To:     Srinivasa Rao Mandadapu <quic_srivasam@...cinc.com>,
        swboyd@...omium.org, agross@...nel.org, andersson@...nel.org,
        robh+dt@...nel.org, broonie@...nel.org, quic_plai@...cinc.com,
        krzysztof.kozlowski+dt@...aro.org, konrad.dybcio@...ainline.org,
        mturquette@...libre.com, sboyd@...nel.org,
        linux-arm-msm@...r.kernel.org, linux-clk@...r.kernel.org,
        linux-kernel@...r.kernel.org, quic_rohkumar@...cinc.com
Subject: Re: [PATCH v2 1/4] dt-bindings: clock: qcom,sc7280-lpasscc: Remove
 qdsp6ss reg property

On 03/01/2023 06:51, Srinivasa Rao Mandadapu wrote:
> 
> On 1/2/2023 9:33 PM, Krzysztof Kozlowski wrote:
> Thanks for Your time Krzyszto!!!
>> On 30/12/2022 14:43, Srinivasa Rao Mandadapu wrote:
>>> The qdsp6ss memory region is being shared by ADSP remoteproc device and
>>> lpasscc clock device, hence causing memory conflict.
>>> As the qdsp6ss clocks are being enabled in remoteproc driver, remove
>>> clock controlling in the clock driver.
>>>
>> That's an ABI break change. You cannot just drop it because the driver
>> will fail now to probe with old DTS.
>>
>> This ABI break (and Fixes tag below) requires extensive justification
>> where is the bug and how it affects users. Otherwise, I see no reason
>> for ABI changes.
> 
> As ADSP path is not used so far due to various vendor requirements, we 
> haven't seen this conflict till now.

The device is used in sc7280.dtsi (and it is not disabled), thus it is
available in every board.

> 
> Actually, in the below commit,  clock handling in remoteproc driver has 
> already up-streamed by Bjorn.
> 
> If we add PIL device node and use remoteproc driver, qcom_q6v5_adsp.c, 
> it's mandatory to remove it from clock
> 
> driver.
> 
> Also the memory region is used by clock driver is part of 
> LPASS_QDSP6V67SS register region.
> 
> Here is the commit 0c6de4c299cc8f3283c38f3778777e00ed7e4b69 
> ("remoteproc: qcom: qdsp6-adsp: Add support for QCS404 CDSP")

The commit introduces qcom,qcs404-cdsp-pil which is not used in upstream
DTS...

Anyway none of the reasons above explain to me why ABI has to be broken.
To remind - valid reasons are usually only: "it never worked" or "I was
introduced recently so there are no users".

> For ABI Break resolution shall I include, device tree changes also in 
> this series? Or Could you please suggest better approach?

Lack of DTS changes causes upstream to be broken. This is independent of
ABI break. ABI means supporting users of the binding, so firmware, other
OS, out-of-tree DTS users etc. All of them are broken.

You need to keep backwards compatibility.
Best regards,
Krzysztof