Date:   Fri, 3 Mar 2023 11:03:36 +0300
From:   Arınç ÜNAL <arinc.unal@...nc9.com>
To:     Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
        Rob Herring <robh@...nel.org>
Cc:     Sergio Paracuellos <sergio.paracuellos@...il.com>,
        Linus Walleij <linus.walleij@...aro.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        linux-mediatek@...ts.infradead.org, linux-mips@...r.kernel.org,
        linux-gpio@...r.kernel.org, devicetree@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Matthias Brugger <matthias.bgg@...il.com>,
        Sean Wang <sean.wang@...nel.org>,
        William Dean <williamsukatube@...il.com>,
        Daniel Golle <daniel@...rotopia.org>,
        Daniel Santos <daniel.santos@...ox.com>,
        Luiz Angelo Daros de Luca <luizluca@...il.com>,
        Frank Wunderlich <frank-w@...lic-files.de>,
        Landen Chao <Landen.Chao@...iatek.com>,
        DENG Qingfang <dqfext@...il.com>,
        Sean Wang <sean.wang@...iatek.com>, erkin.bozoglu@...ont.com
Subject: Re: [RFC PATCH 07/16] dt-bindings: pinctrl: ralink: add new
 compatible strings

On 3.03.2023 10:53, Krzysztof Kozlowski wrote:
> On 03/03/2023 08:44, Arınç ÜNAL wrote:
>> On 3.03.2023 10:05, Krzysztof Kozlowski wrote:
>>> On 02/03/2023 12:50, Arınç ÜNAL wrote:
>>>> On 2.03.2023 14:36, Krzysztof Kozlowski wrote:
>>>>> On 02/03/2023 11:47, Arınç ÜNAL wrote:
>>>>>> On 2.03.2023 13:29, Krzysztof Kozlowski wrote:
>>>>>>> On 02/03/2023 11:22, Arınç ÜNAL wrote:
>>>>>>>>>>
>>>>>>>>>> ## Incorrect naming
>>>>>>>>>>
>>>>>>>>>> MT7620, MT7621, MT7628, and MT7688 SoCs are incorrectly called Ralink,
>>>>>>>>>> introduce new ralink->mediatek compatible strings to address it.
>>>>>>>>>
>>>>>>>>> So this part was addressed by Rob - we don't do it, because it does not
>>>>>>>>> matter. Ralink is now Mediatek, thus there is no conflict and no issues
>>>>>>>>> with different vendor used.
>>>>>>>>
>>>>>>>> I think Rob was rather addressing that updating compatible strings based
>>>>>>>> on acquisition or marketing whims is not permitted. This condition does
>>>>>>>> not apply here as these SoCs were never Ralink.
>>>>>>>>
>>>>>>>> I understand your point that Ralink is now MediaTek but still, calling
>>>>>>>> these SoCs Ralink would be a bit misleading, don't you think?
>>>>>>>
>>>>>>> Misleading yes, but also does not matter. At least matter not enough to
>>>>>>> justify ABI break, so you would need to deprecate old ones and keep
>>>>>>> everything backwards compatible. You still would affect 3rd party users
>>>>>>> of DTS, though...
>>>>>>
>>>>>> I intend to do just that. Introduce new mediatek strings, keep the old
>>>>>> ones so it's backwards compatible, therefore don't break the ABI.
>>>>>>
>>>>>> Instead of deprecating old strings, I intend to introduce the checks I
>>>>>> mentioned, on the schema, so the pin muxing bindings only apply if the
>>>>>> DT has got a string that won't match multiple schemas. This way it
>>>>>> shouldn't affect 3rd party DTs.
>>>>>
>>>>> I meant, 3rd party users of DTS. You will replace the compatible in the
>>>>> DTS, right? So the DTS exported and used in all other projects, OS,
>>>>> firmwares, bootloaders, out of tree kernel forks will stop working.
>>>>
>>>> I plan to change it on the DTs for MediaTek SoCs, yes. Is this a
>>>> problem? From what I can tell, what must be ensured is that old DTs must
>>>> work with newer kernels, not new DTs on older kernels.
>>>
>>> Can I be clearer than this?
>>>
>>> " So the DTS exported and used in all other projects, OS,
>>> firmwares, bootloaders, out of tree kernel forks will stop working."
>>>
>>> Yes, this is a problem - they will stop working.
>>
>> I've never seen any project just exporting DTs from the latest kernel
>> version and slap it onto old versions, as a new devicetree that was
> 
> Really? U-Boot does it all the time, other projects (like BSD) do it
> periodically to some extend as well.

They must do heavy reviewing before shipping it. Drivers like MediaTek 
ethernet on U-Boot is different than in Linux, the dt-bindings are all 
different. Under a review, these changes will pop out for them to 
address so there're no problems.

> 
>> introduced with a newer kernel version is not guaranteed to work with
>> older kernel versions.
> 
> Not guaranteed but it is expected, though, to some level and under some
> conditions. Therefore it might be or might not be a problem. For some
> platforms no one cares. For some people care.

I'm going to assume there's not much care for this platform, at least 
for mt7621, as I've heard no complaints when I did this before.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/mips/boot/dts/ralink/mt7621.dtsi?id=b4f209e32ba5c283e7b1dd00d867b0536d3e215e

> 
>>
>> If someone is actually doing this on a project, I think it's the
>> responsibility of the maintainers of these said projects to account for
>> this and modify the DT for the kernel version they're running it on.
>>
>> What's more usual is one'd run the kernel version where the new DT was
>> introduced, which will work fine.
> 
> "kernel" as Linux is only one part of it. I mentioned several other
> projects.
> 
>>
>> On to real life implications, popular projects like U-Boot and OpenWrt
>> maintain their own DTs for this platform so I think the impact is very
>> minimal.
> 
> And they sync with Linux kernel DTS.

Again, the DTs must be reviewed so they will be modified and the 
potential issue will be addressed.

Arınç

Powered by blists - more mailing lists