linux-kernel - Re: [PATCH 1/6] media: dt-bindings: Document SC8280XP/SM8350 Venus

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: <0feda32e-5430-4f35-b18a-7afce63a970c@linaro.org>
Date:   Mon, 7 Aug 2023 20:45:29 +0200
From:   Konrad Dybcio <konrad.dybcio@...aro.org>
To:     Bryan O'Donoghue <bryan.odonoghue@...aro.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
        Stanimir Varbanov <stanimir.k.varbanov@...il.com>,
        Vikash Garodia <quic_vgarodia@...cinc.com>,
        Andy Gross <agross@...nel.org>,
        Bjorn Andersson <andersson@...nel.org>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Conor Dooley <conor+dt@...nel.org>
Cc:     Marijn Suijten <marijn.suijten@...ainline.org>,
        Konrad Dybcio <konradybcio@...nel.org>,
        linux-media@...r.kernel.org, linux-arm-msm@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/6] media: dt-bindings: Document SC8280XP/SM8350 Venus

On 7.08.2023 20:44, Bryan O'Donoghue wrote:
> On 07/08/2023 16:02, Konrad Dybcio wrote:
>> On 7.08.2023 16:04, Krzysztof Kozlowski wrote:
>>> On 07/08/2023 14:41, Konrad Dybcio wrote:
>>>> On 5.08.2023 21:29, Krzysztof Kozlowski wrote:
>>>>> On 04/08/2023 22:09, Konrad Dybcio wrote:
>>>>>> Both of these SoCs implement an IRIS2 block, with SC8280XP being able
>>>>>> to clock it a bit higher.
>>>>>>
>>>>>
>>>>> ...
>>>>>
>>>>>> +
>>>>>> +  iommus:
>>>>>> +    maxItems: 1
>>>>>> +
>>>>>> +  video-decoder:
>>>>>> +    type: object
>>>>>> +
>>>>>> +    properties:
>>>>>> +      compatible:
>>>>>> +        const: venus-decoder
>>>>>
>>>>> That's not how compatibles are constructed... missing vendor prefix, SoC
>>>>> or IP block name.
>>>>>
>>>>>> +
>>>>>> +    required:
>>>>>> +      - compatible
>>>>>> +
>>>>>> +    additionalProperties: false
>>>>>
>>>>> Why do you need this child node? Child nodes without properties are
>>>>> usually useless.
>>>> For both comments: I aligned with what was there..
>>>>
>>>> The driver abuses these compats to probe enc/dec submodules, even though
>>>> every Venus implementation (to my knowledge) is implicitly enc/dec capable..
>>>
>>> Holy crap, I see...
>>>
>>>>
>>>> Perhaps a bigger clean-up is due. I guess I could just create the venc/vdec
>>>> devices from the venus core probe and get rid of this fake stuff?
>>>
>>> Few devices (qcom,msm8996-venus.yaml, sdm660, sdm845) have clocks there,
>>> so we actually could stay with these subnodes, just correct the
>>> compatibles to a list with correct prefixes:
>>>
>>> qcom,sc8280xp-venus-decoder + qcom,venus-decoder
>> Hm.. looks like pre-845-v2 (with the v2 being "v2 binding" and not
>> "v2 chip" or "v2 hardware") these were used to look up clocks but
>> then they were moved to the root node.
>>
>> I am not quite sure if it makes sense to distinguish e.g.
>> sc8280xp-venus-decoder within sc8280xp-venus..
>>
>> Perhaps deprecating the "8916 way" (clocks under subnodes), adding
>> some boilerplate to look up clocks/pds in both places and converting
>> everybody to the "7180 way" way of doing things (clocks under venus),
>> and then getting rid of venus encoder/decoder completely (by calling
>> device creation from venus probe) would be better. WDYT?
>>
>> Konrad
> 
> As I understand it though, for some classes of venus hardware - earlier, it was possible to have two encoders or two decoders and it really didn't - perhaps still doesn't matter which order they are declared in.
> 
> That's the logic behind having a compat string that assigns either encoder or decoder to one of the logical blocks.
> 
> You can have any mixture of
> - encoder
> - decoder
> 
> - encoder
> - encoder
> 
> - decoder
> - decoder
> 
> - decoder
> - encoder
> 
> - encoder
> 
> - decoder
> 
> I think it should *still* be the case - whether it is a practical reality or not, that any of those mapping can be selected and supported.
That can be taken care of with match data.

Konrad