[PATCH v4 6/7] dt-bindings: net: dsa: mediatek,mt7530: add EcoNet EN751221 switches

Caleb James DeLisle cjd at cjdns.fr
Thu Nov 5 13:47:10 PST 2026


On 04/11/2026 08:20, Krzysztof Kozlowski wrote:
> On 03/11/2026 14:15, Caleb James DeLisle wrote:
>>>> +        const: econet,en751221-gswitch
>>> That's almost the same compatible, which makes me wonder if you describe
>>> the same device twice. Same device, same compatible.
>>
>> Well, these are both MT7530 based, but one of them is on-die with the
>> SoC and the other one is an MCM so for the driver they are very
>> different - they're even loaded by different kmods. I think since one is
>> MMIO and the other is MDIO, using the same compatible won't cause a
>> driver problem - but I don't know if it's best practice to smooth over
>> such a difference or not. Your advice here would be most appreciated.
>>
> The drivers do not matter - we have one compatible for I2C/SPI/Serial
> devices as well. The only question is about the hardware, what are these
> devices and what are you here exactly describing. Compatible is en751221
> but you call it MT7530, so even more confusion.


Let me address these in reverse order: The compatibles are 
econet,en751221-switch and econet,en751221-gswitch but the switch is an 
MT7530. This is a convention used for this switch: When it is integrated 
in an SoC and has customization for that SoC, it is identified using the 
name of the SoC suffixed by "-switch" (except in the case of 
mediatek,mt7621 where it takes the exact name of the SoC, but this looks 
like it was a mistake).


Regarding the en751221-switch vs. en751221-gswitch dichotomy, the 
meaningful difference to the DT is that the gswitch is a 
multi-chip-module so the CPU port is phy-mode "trgmii", and on the 
on-die switch it's "internal" since there's no PHY between it and the 
CPU. Moreover the gswitch requires the mediatek,mcm; attribute while the 
on-die switch forbids it. I can of course relax the restrictions so they 
can use the same compatible, but I don't know which way is most 
appropriate so I would like to follow your advice.


Thanks,

Caleb


>
> Best regards,
> Krzysztof



More information about the linux-arm-kernel mailing list