Re: [PATCH] ARM: dts: aspeed: add Nuvia DC-SCM BMC

[Jae Hyun Yoo <quic_jaehyoo@quicinc.com>]

On 3/24/2022 3:19 PM, Andrew Lunn wrote:
> On Thu, Mar 24, 2022 at 02:27:34PM -0700, Jae Hyun Yoo wrote:
>> Hi Andrew,
>>
>> On 3/24/2022 2:23 PM, Andrew Lunn wrote:
>>> On Thu, Mar 24, 2022 at 09:45:51AM -0700, Jae Hyun Yoo wrote:
>>>> From: Graeme Gregory <quic_ggregory@quicinc.com>
>>>>
>>>> +&mac2 {
>>>> +	status = "okay";
>>>> +
>>>> +	phy-mode = "rgmii";
>>>
>>> Are you sure about this? That PHY are you using? It is more normal to
>>> use rgmii-id.
>>
>> Yes, I'm sure. PHY works with rgmii on my board. I picked a different
>> clock delay setting too.
> 
> Do you know who is inserting the delay? It looks like rgmii will be
> passed to the PHY. So there are few possibilities here:
> 
> The MAC is adding the delay. Not recommended, but does work. In that
> case, your DT is correct.
> 
> The bootloader or strapping configures the PHY to insert the
> delay. The PHY driver then ignores 'rgmii' and leaves the
> configuration alone. This will break sometime in the future when
> somebody fixes the PHY driver to actually set the PHY to rgmii as
> requested. In this case, both the PHY driver and your DT is broken.
> 
> The board itself inserts the delays by having long tracks. In this
> case, your DT is correct.
> 
> It would be good to understand this, just in case it is the second
> condition.

Thanks for your kind explanation.

The clock delay I mentioned is added into u-boot bootloader in a
patch I'm currently trying to submit.
https://lore.kernel.org/all/20220324165530.359668-1-quic_jaehyoo@quicinc.com/

I had to pick those value with 'rgmii' to make my board work but the
delay is actually different from what AST2600 EVB DT provides. I think
it's caused by a different PCB routing and board characteristic like
you said.

I want to keep this setting as it's acquired from lots of experiments.
It could be changed in case if it needs to be fixed in the future.

Thanks,
Jae