[PATCH 1/6] ARM: dts: sam9x60: Add OHCI and EHCI DT nodes

[Eugen.Hristev at microchip.com]

On 1/4/23 11:38, Marek Vasut wrote:
> On 1/4/23 08:30, Eugen.Hristev at microchip.com wrote:
>> On 1/3/23 01:12, Marek Vasut wrote:
>>> On 12/23/22 13:33, Sergiu Moga wrote:
>>>> Add the OHCI and EHCI DT nodes for the sam9x60 boards.
>>>>
>>>> Signed-off-by: Sergiu Moga <sergiu.moga at microchip.com>
>>>> ---
>>>>    arch/arm/dts/at91-sam9x60_curiosity.dts | 21 +++++++++++++++++++++
>>>>    arch/arm/dts/sam9x60.dtsi               | 18 ++++++++++++++++++
>>>
>>> Board and SoC DT changes should be in separate patches.
>>>
>>>>    arch/arm/dts/sam9x60ek.dts              | 21 +++++++++++++++++++++
>>>>    3 files changed, 60 insertions(+)
>>>
>>>
>>> [...]
>>>
>>>> diff --git a/arch/arm/dts/sam9x60.dtsi b/arch/arm/dts/sam9x60.dtsi
>>>> index 17224ef771..e36a540f78 100644
>>>> --- a/arch/arm/dts/sam9x60.dtsi
>>>> +++ b/arch/arm/dts/sam9x60.dtsi
>>>> @@ -69,6 +69,24 @@
>>>>                #size-cells = <1>;
>>>>                ranges;
>>>>
>>>> +             usb1: ohci at 600000 {
>>>
>>> This should be usb@ instead of ohci@ , if you run "make dtbs_check" on
>>> this DT in Linux (please do), you would likely get a warning , see Linux
>>> Documentation/devicetree/bindings/usb/usb.yaml .
>>
>>
>> If in Linux we have ohci@ , then we need to have the same in U-boot.
> 
> You should update the Linux DT to usb@ too to avoid dtbs_check warnings.
> 
>> We can accept the change to usb@ , if there is a pending patch in Linux.
> 
> I can make the same argument about Linux, since DTs are OS agnostic. My
> comment is OS agnostic too and does not apply specifically to U-Boot or
> Linux, it applies to DT.

Definitely you can comment in Linux , for sure.

> 
>> U-boot is not the place to review the devicetree.
> 
> I strongly disagree with this statement.

There are a few decisions behind this statement.
First, the fact that we no longer take bindings into Uboot, and rely 
solely on bindings in Linux. This means the bindings were reviewed and 
validated in Linux.
Sure there are a few exceptions but not the usual rule.
Second, we rely on devicetree kernel.org mailing list to review all the 
DTs and binding compliance in Linux.
This means that the DT has been validated and checked as ABI in Linux. 
Hence there is no place to question that in Uboot or any other project.
The DT must be the same and ABI across all software that uses the DT. If 
this was validated and agreed upon in Linux, we can't question that here.
So if you disagree with this statement, it's your choice of course.

> 
> What does it matter where the review feedback came from ?
> What does matter is that you can improve the DT based on that feedback,
> if the feedback is valid.

I agree with that, but if we want to improve things, we need to patch 
things either in Linux first, or patch projects together.
The fact that you are against some patches while they have the same 
identical node in Linux, is not a reason to not allow devicetree nodes 
to be added to Uboot.
DT syncing with Linux is always a must, and that is part of what Sergiu 
is doing with these patches.


> 
>> The devicetree must be in sync with Linux.
> 
> I agree with this statement.
> 
> Please update the Linux DT too.

That is one approach but as I said, not a reason to reject a node that 
is identical with Linux.