Re: [PATCH v4 11/13] ARM: dts: aspeed: facebook-fuji: Include facebook-fuji-data64.dts

[Guenter Roeck <linux@roeck-us.net>]

On 10/13/25 21:09, Tao Ren wrote:
> Hi Guenter,
> 
> On Mon, Oct 13, 2025 at 08:44:17PM -0700, Guenter Roeck wrote:
>> On 10/13/25 20:20, Guenter Roeck wrote:
>>> On 10/13/25 18:15, Tao Ren wrote:
>>>> Hi Guenter,
>>>>
>>>> On Mon, Oct 13, 2025 at 05:20:57PM -0700, Guenter Roeck wrote:
>>>>> Hi,
>>>>>
>>>>> On Sun, Jul 27, 2025 at 10:56:13PM -0700, rentao.bupt@gmail.com wrote:
>>>>>> From: Tao Ren <rentao.bupt@gmail.com>
>>>>>>
>>>>>> Include "facebook-fuji-data64.dts" in facebook-fuji dts to avoid
>>>>>> duplicated code.
>>>>>>
>>>>>> Fuji-data64 and Fuji are identical except the BMC flash layout.
>>>>>>
>>>>>> Signed-off-by: Tao Ren <rentao.bupt@gmail.com>
>>>>>
>>>>> With this patch in the mainline kernel, the Ethernet interface I use for
>>>>> testing does not come online when loading fuji-bmc in qemu.
>>>>>
>>>>> Reverting this patch fixes the problem.
>>>>>
>>>>> Looking into this patch,
>>>>>
>>>>>> -
>>>>>> -#include <dt-bindings/leds/common.h>
>>>>>> -#include "ast2600-facebook-netbmc-common.dtsi"
>>>>>> +#include "aspeed-bmc-facebook-fuji-data64.dts"
>>>>> ...
>>>>>> -&mac3 {
>>>>>> -    status = "okay";
>>>>>> -    phy-mode = "rgmii";
>>>>>> -    phy-handle = <&ethphy3>;
>>>>>> -    pinctrl-names = "default";
>>>>>> -    pinctrl-0 = <&pinctrl_rgmii4_default>;
>>>>>> -};
>>>>>
>>>>> I don't see this in aspeed-bmc-facebook-fuji-data64.dts, meaning that
>>>>> interface is now disabled. Adding it back in fixes the problem.
>>>>> Also, MAC3 is explicitly enabled for fuji-bmc in qemu.
>>>>>
>>>>> Was the interface disabled on purpose ?
>>>>>
>>>>> Thanks,
>>>>> Guenter
>>>>
>>>> The mac3 interface was removed in the latest patch (v4) per Andrew Lunn's
>>>> feedback, because the rgmii setting is incorrect.
>>>>
>>>> I was planning to add mac3 back as soon as rgmii support is properly
>>>> handled in aspeed mac driver, but kindly let me know if you have other
>>>> suggestions.
>>>>
>>>
>>> All I can say is that it worked just fine with the qemu emulation,
>>> and that it is broken now. Since it was broken on purpose I guess I'll
>>> have to find a workaround or stop testing network interfaces with
>>> that emulation entirely.
>>>
>>
>> Ah, I see that mac3 was the only enabled Ethernet interface on that system,
>> so you effectively disabled networking on it.
>>
>> I don't claim to understand the logic (how can anyone continue to use this bmc
>> without network interface ?) but I guess it is what it is. I'll stop testing it.
>>
>> Guenter
> 
> Sorry for the inconvenience.. I have to take a local patch to enable
> mac3 in my environment, because as you said, it's not useful without
> network interface.
> 
> Meanwhile, I will need to work with ASPEED to enable rgmii delay support
> in the ASPEED MAC driver so I can add mac3 back to the dts. If you are
> looking for a similar platform for testing, you could consider elbert.
> 

AFAICS aspeed-bmc-facebook-elbert.dtb does not work with the fuji-bmc machine.
The console interface or wiring seems to be different.

As I said, I'll stop testing the network interface on fuji-bmc. That reduces
test coverage, but there is nothing I can do about that. I already made the
necessary changes in my testbed, so from my perspective the issue is closed.

Guenter