Re: [PATCH 2/3] dts: socfpga: Add Google Chameleon v3 devicetree

[Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>]

On 31/05/2022 16:47, Paweł Anikiel wrote:
> On Tue, May 31, 2022 at 11:11 AM Krzysztof Kozlowski
> <krzysztof.kozlowski@linaro.org> wrote:
>>
>> On 31/05/2022 03:20, Alexandru M Stan wrote:
>>> Hello Krzysztof
>>>
>>> On Mon, May 30, 2022 at 11:56 AM Krzysztof Kozlowski
>>> <krzysztof.kozlowski@linaro.org> wrote:
>>>>
>>>> On 30/05/2022 15:08, Paweł Anikiel wrote:
>>>>> Add devicetree for the Google Chameleon v3 board.
>>>>>
>>>>> Signed-off-by: Paweł Anikiel <pan@semihalf.com>
>>>>> Signed-off-by: Alexandru M Stan <amstan@chromium.org>
>>>>
>>>> Your SoB chain looks odd. Who did what here?
>>>
>>> Sorry about this.
>>>
>>> It was mainly Pawel but I did some small changes at some point before
>>> it landed in our tree (particularly the GPIOs).
>>
>> Then usually Paweł should be the owner of the patch, not you.
>> Alternatively it could be also co-developed.
>>
>>>
>>>>
>>>>> ---
>>>>>  arch/arm/boot/dts/Makefile                    |  1 +
>>>>>  .../boot/dts/socfpga_arria10_chameleonv3.dts  | 90 +++++++++++++++++++
>>>>>  2 files changed, 91 insertions(+)
>>>>>  create mode 100644 arch/arm/boot/dts/socfpga_arria10_chameleonv3.dts
>>>>>
>>>>> diff --git a/arch/arm/boot/dts/Makefile b/arch/arm/boot/dts/Makefile
>>>>> index 023c8b4ba45c..9417106d3289 100644
>>>>> --- a/arch/arm/boot/dts/Makefile
>>>>> +++ b/arch/arm/boot/dts/Makefile
>>>>> @@ -1146,6 +1146,7 @@ dtb-$(CONFIG_ARCH_S5PV210) += \
>>>>>       s5pv210-torbreck.dtb
>>>>>  dtb-$(CONFIG_ARCH_INTEL_SOCFPGA) += \
>>>>>       socfpga_arria5_socdk.dtb \
>>>>> +     socfpga_arria10_chameleonv3.dtb \
>>>>>       socfpga_arria10_socdk_nand.dtb \
>>>>>       socfpga_arria10_socdk_qspi.dtb \
>>>>>       socfpga_arria10_socdk_sdmmc.dtb \
>>>>> diff --git a/arch/arm/boot/dts/socfpga_arria10_chameleonv3.dts b/arch/arm/boot/dts/socfpga_arria10_chameleonv3.dts
>>>>> new file mode 100644
>>>>> index 000000000000..988cc445438e
>>>>> --- /dev/null
>>>>> +++ b/arch/arm/boot/dts/socfpga_arria10_chameleonv3.dts
>>>>> @@ -0,0 +1,90 @@
>>>>> +// SPDX-License-Identifier: GPL-2.0
>>>>> +/*
>>>>> + * Copyright 2022 Google LLC
>>>>> + */
>>>>> +/dts-v1/;
>>>>> +#include "socfpga_arria10_mercury_aa1.dtsi"
>>>>> +
>>>>> +/ {
>>>>> +     model = "Google Chameleon V3";
>>>>> +     compatible = "google,chameleon-v3",
>>>>
>>>> You miss here enclustra compatible.
>>>
>>> Does this make sense? I don't expect this device tree to boot/work on
>>> an enclustra motherboard. It's only really compatible with a
>>> "chameleon-v3".
>>
>> You also do not expect it to boot on altr,socfpga, do you?
>>
>> If I understood correctly, this board has physically Mercury AA1 SoM, so
>> that compatible should be there.
> 
> Yes, you understood correctly.
> I looked at a similar device - the Cyclone V MCV (SoM) and the MCVEVK
> (base board):
> arch/arm/boot/dts/socfpga_cyclone5_mcv.dtsi

This one is clearly incorrect, so using it as example is wrong.

> arch/arm/boot/dts/socfpga_cyclone5_mcvevk.dts

Since it is based on wrong MCV, then no wonder it's the same.

> And there is no denx,mcv compatible anywhere, only denx,mcvevk.
> Also, devicetree bindings documentation lists denx,mcvevk under
> "Cyclone 5 boards", and, unless you consider the MCV to be a board,
> there isn't a good place to put denx,mcv (same story with mercury+
> aa1/chameleon).

socfpga are not the best example... upstreaming looks incomplete or even
incorrect, like this case of Enclustra SOM. Much better examples are
FSL-based SoMs. Although some of them are also not in the best shape.

Still someone might prefer to skip SoM compatible arguing that it cannot
be a separate final product. Sure, but also SoC cannot be a separate
product. Having SoM compatible allows to match against it and find
common hardware parts.

In any case you want to remove here parts of bindings (so affect ABI),
to which I do not agree.

>> Let me clarify - please use generic node names, as asked by Devicetree
>> specification (2.2.1. Node Name Requirements). There is also list of
>> some examples in the spec, but you can use some other generic node name.
>>
>> Several bindings also require it.
> 
> Do you mean something like this?
> ssm2603: audio-codec@1a {
> u80: gpio@21 {

Yes.


Best regards,
Krzysztof