Re: [PATCH 09/20] dt-bindings: pinctrl: ralink: {mt7620,mt7621}: rename to mediatek

[Sergio Paracuellos <sergio.paracuellos@xxxxxxxxx>]

On Thu, Mar 9, 2023 at 10:09 PM Arınç ÜNAL <arinc.unal@xxxxxxxxxx> wrote:
>
> On 9.03.2023 14:33, Sergio Paracuellos wrote:
> > On Thu, Mar 9, 2023 at 11:34 AM Arınç ÜNAL <arinc.unal@xxxxxxxxxx> wrote:
> >>
> >> On 9.03.2023 12:52, Krzysztof Kozlowski wrote:
> >>> On 09/03/2023 08:53, Arınç ÜNAL wrote:
> >>>> On 9.03.2023 00:19, Arınç ÜNAL wrote:
> >>>>> On 9.03.2023 00:05, Rob Herring wrote:
> >>>>>> On Fri, Mar 03, 2023 at 03:28:38AM +0300, arinc9.unal@xxxxxxxxx wrote:
> >>>>>>> From: Arınç ÜNAL <arinc.unal@xxxxxxxxxx>
> >>>>>>>
> >>>>>>> This platform from Ralink was acquired by MediaTek in 2011. Then,
> >>>>>>> MediaTek
> >>>>>>> introduced these SoCs which utilise this platform. Rename the schemas to
> >>>>>>> mediatek to address the incorrect naming.
> >>>>>>
> >>>>>> I said we don't do renames due to acquistions, you said that wasn't the
> >>>>>> reason, but then that's your reasoning here.
> >>>>>
> >>>>> It's not a marketing/acquistion rename as the name of these SoCs were
> >>>>> wrong from the get go. The information on the first sentence is to give
> >>>>> the idea of why these SoCs were wrongfully named as the base platform
> >>>>> that these new MediaTek SoCs share code with was called Ralink.
> >>>>>
> >>>>>>
> >>>>>> To give you another example, *new* i.MX things are still called
> >>>>>> 'fsl,imx...' and it has been how many years since merging with NXP?
> >>>>>
> >>>>> Ok this is a point I see now. Though, I fail to see how this is called
> >>>>> renaming when there's only new SoCs (from NXP in this case) to be added.
> >>>>
> >>>> If I understand correctly, i.MX is a family from Freescale so the name
> >>>
> >>> It's the same "family" as your platform, because as you said:
> >>> "introduced these SoCs which utilise this platform"
> >>>
> >>>> was kept the same on new SoC releases from NXP. I believe it's different
> >>>> in this case here. There's no family name. The closest thing on the name
> >>>> of the SoC model is, it's RT for Ralink, MT for MediaTek.
> >>>
> >>> It's not about the name. NXP took Freescale platform and since many
> >>> years makes entirely new products, currently far, far away from original
> >>> platform.
> >>>
> >>> That's the same case you have here - Mediatek took existing platform and
> >>> started making new products with it.
> >>>
> >>>>
> >>>> On top of that, mediatek strings already exist for MT SoCs already, at
> >>>> least for MT7621.
> >>>>
> >>>> https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/tree/Documentation/devicetree/bindings/mips/ralink.yaml?id=dd3cb467ebb5659d6552999d6f16a616653f9933#n83
> >>>
> >>> NXP also has compatibles with nxp, thus still not that good reason.
> >>
> >> Ok, makes sense. Am I free to call the SoCs MediaTek, change the schema
> >> name from ralink,mtXXXX-pinctrl.yaml to mediatek,mtXXXX-pinctrl.yaml
> >> whilst keeping the compatible string ralink?
> >>
> >> I plan to do some cleanup on ralink.yaml as well. From what I
> >> understand, I should change the mediatek,mt7621-soc compatible string to
> >> ralink,mt7621-soc?
> >
> > You have to take care of these SoC strings since they are used in the
> > very beginning of the ralink startup platforms for any single ralink
> > SoC. See for example [0] and [1] (but they are in all soc init code).
> > I think it is better to maintain the ralink.yaml file as it is.
>
> I'd really rather address this inconsistency everywhere possible. The
> code you pointed out looks different than what I did on the pinctrl
> driver but, surely it's possible on the code to introduce ralink and
> keep the mediatek string for the sake of old DTs, no?

In any case, the changes you might have in mind for this should be a
different patch series.

Best regards,
     Sergio Paracuellos
>
> Arınç