[PATCH v5 3/5] arm64: dts: ti: k3-j784s4: Add WIZ and SERDES PHY nodes

[Jayesh Choudhary]

On 20/07/23 15:06, Jayesh Choudhary wrote:
> 
> 
> On 14/07/23 00:28, Nishanth Menon wrote:
>> On 13:31-20230713, Andrew Davis wrote:
>>> On 7/13/23 1:21 PM, Nishanth Menon wrote:
>>>> On 21:01-20230713, Jayesh Choudhary wrote:
>>>>>
>>>>>
>>>>> On 12/07/23 19:48, Nishanth Menon wrote:
>>>>>> On 15:47-20230710, Jayesh Choudhary wrote:
>>>>>>> From: Siddharth Vadapalli <s-vadapalli at ti.com>
>>>>>>>
>>>>>>> J784S4 SoC has 4 Serdes instances along with their respective WIZ
>>>>>>> instances. Add device-tree nodes for them and disable them by 
>>>>>>> default.
>>>>>>>
>>>>>>> Signed-off-by: Siddharth Vadapalli <s-vadapalli at ti.com>
>>>>>>> [j-choudhary at ti.com: fix serdes_wiz clock order]
>>>>>>> Signed-off-by: Jayesh Choudhary <j-choudhary at ti.com>
>>>>>>> ---
>>>>>> NAK. This patch introduces the following dtbs_check warning.
>>>>>> arch/arm64/boot/dts/ti/k3-am69-sk.dtb: serdes-refclk: 
>>>>>> 'clock-frequency' is a required property
>>>>>>
>>>>>
>>>>> Sorry for this. This property was added in the final board file.
>>>>> I will fix it in the next revision.
>>>>> I will add '0' as clock-property in the main file similar to j721e[1]
>>>>> which will be overridden in the board file with required value to get
>>>>> rid of this warning.
>>>>
>>>> That would follow what renesas (r8a774a1.dtsi) and imx
>>>> (imx8dxl-ss-conn.dtsi) seem to be doing as well. Just make sure to add
>>>> documentation to the property to indicate expectation. Unless someone
>>>> has objections to this approach.
>>>>
>>>
>>> Would it work better to disable these nodes, only enabling them in the
>>> board files when a real clock-frequency can be provided?
>>>
>>> My initial reaction would be to move the whole external reference clock
>>> node to the board file since that is where it is provided, but seems
>>> that would cause more churn in serdes_wiz* nodes than we would want..
>>
>> I would prefer that as well, but I have'nt gone around looking for
>> similar examples on other SoCs (Jayesh, can you check?). One other
>> approach (alipine and few other places) has been for the bootloader to
>> update the property set in dtb as 0, which is not needed in this case
>> to the best of what I see.. just hoping we use a technique that most
>> board folks are familiar with across SoCs.
>>
> 
> 
> I can see the clock nodes in board files for some vendors. But like
> Andrew said, that would cause issues with serdes_wiz node in the
> main.dtsi.
> 
> So I think it would be better to keep the external clock node in main
> dtsi itself. I will add comments to the property to indicate that this
> value will be over-written in board dts.
> Posting v6 to address these comments and others on the series.
> 

FYI..
I posted v6: 
https://lore.kernel.org/all/20230721132029.123881-1-j-choudhary@ti.com/
And I am using status property instead to address the dtbs_warning here.

Warm Regards,
Jayesh