linux-kernel - Re: [PATCH v3 1/2] dt-bindings: soc: qcom: Add qcom,kaanapali-imem compatible

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

Message-ID: <9ee07db9-508e-4c08-8f79-6ccfd9b646ab@oss.qualcomm.com>
Date: Tue, 4 Nov 2025 15:35:05 +0100
From: Konrad Dybcio <konrad.dybcio@....qualcomm.com>
To: Krzysztof Kozlowski <krzk@...nel.org>,
        Jingyi Wang <jingyi.wang@....qualcomm.com>
Cc: Rob Herring <robh@...nel.org>, Krzysztof Kozlowski <krzk+dt@...nel.org>,
        Conor Dooley <conor+dt@...nel.org>,
        Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konradybcio@...nel.org>,
        Robert Marko <robimarko@...il.com>,
        Das Srinagesh <quic_gurus@...cinc.com>, aiqun.yu@....qualcomm.com,
        tingwei.zhang@....qualcomm.com, trilok.soni@....qualcomm.com,
        yijie.yang@....qualcomm.com, linux-arm-msm@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 1/2] dt-bindings: soc: qcom: Add qcom,kaanapali-imem
 compatible

On 11/4/25 3:26 PM, Krzysztof Kozlowski wrote:
> On 04/11/2025 13:32, Konrad Dybcio wrote:
>> On 11/4/25 9:16 AM, Krzysztof Kozlowski wrote:
>>> On Sun, Nov 02, 2025 at 11:25:06PM -0800, Jingyi Wang wrote:
>>>> Document qcom,kaanapali-imem compatible. Kaanapali IMEM is not a syscon or
>>>> simple-mfd, also "reboot reason" is not required on Kaanapali like some
>>>
>>> I do not see correlation. Something is not a syscon, so you add a new
>>> generic compatible? No.
>>>
>>>> other platforms. So define a common "qcom,imem" binding and fallback to it.
>>>
>>> You did not define fallback to it!
>>>
>>> ...
>>>
>>>> +      - items:
>>>> +          - enum:
>>>> +              - qcom,kaanapali-imem
>>>> +          - const: qcom,imem
>>>
>>> I do not understand what this generic compatible is supposed to express,
>>> not explained in commit msg. Considering this wasn't before, it is a
>>> major and really undesired change. It also makes no sesne. There was no
>>> generic compatible before but "if not syscon" now this must have generic
>>> compatible, what?
>>
>> So IMEM (or SYSTEM_IMEM more specifically as opposed to BOOT_IMEM which
>> you can take your guesses what it's used for) is to the best of our
>> understanding just a piece of SRAM that's accessible by multiple
>> processors/subsystems on the SoC.
>>
>> A smaller region within it ("shared IMEM") is a little bit of a dumping
>> ground for various (incl. runtime) configuration and debug magic data
>> and that's usually what Linux is concerned with.
>>
>> IMEM is currently described as a simple-mfd+syscon, which it is clearly
>> not. The former, as we've established in the past, was used as a hack to
>> have something call of_platform_populate().
>>
>> I think that in turn is only necessary for the old arm32 DTs which have
>> a syscon-reboot-mode node under IMEM (and I think that's where the syscon
>> compatible comes from).
>>
>> Should we make the switch to mmio-sram and settle this discussion?
>> It would probably require convincing the sram maintainer to add that
>> of_platform_populate() call in its probe func and making syscon-reboot
>> not depend on a syscon (not like it's very hard)
> 
> This I got, but nothing here explains why you need generic compatible.
> To re-iterate: there was no generic compatible before, now there is.
> Writing bindings and numerous reviews from DT maintainers ask not to use
> generic compatibles.

OK so let's not worry about a generic compatible. IMEM exists since
MSM8974 and it only had major hw updates with SM8550. They don't
impact the software interface though, so qcom,msm8974-imem is OK.

There's a separate control/status register address space for each
instance of this IP (usually far apart from the actual SRAM pool),
which Linux doesn't have to care about.

Konrad