On 26/06/2024 23:01, Chris Packham wrote: >>>>> + - realtek,rtl9302c >>>> Why board has the name of SoC? >>> What I have is actually a reference board with the name >>> RTL9302C_2xRTL8224_2XGE. If found that a bit incomprehensible so I >>> (over) shortened it. Technically it would be something like >>> cameo,rtl9302c-2x-rtl8224-2xge which I can include in the next round. >> Looks fine to me. >> >>>>> + - const: realtek,rtl9302-soc >>>> Drop the -soc suffix. The rtl9302 is the soc. >>> On that. I hope to eventually add "realtek,rtl9302-switch" for the DSA >>> switch block in the same chip. So keeping the -soc suffix was >>> intentional to try to disambiguate things. I can drop the -soc if the >>> consensus is that there is no need to disambiguate the two. >> Thanks for explanation, kind of depends on what exactly is this. Most of >> SoCs comprise of several items. The entire chip is the soc, e.g. >> "qcom,foo1234". It might have MAC/Ethernet/whatever inside, controllable >> by the SoC (Linux, bootloader, TF, hypervisor, other VM guest) and that >> part is "qcom,foo1234-ethernet". Regardless whether Linux OS actually >> controls it or not. >> >> The question is whether DSA switch is part of the SoC or not. > > The RTL9302C is a single package but I'd assume internally it has > multiple dies. > > From the block diagram in the datasheet they do have a portion they > call the "SoC" which has the CPU and peripherals like UARTs, GPIOs, SPI > etc. That is separate from the switch block which has a bunch of MACs, > SERDES and various network switch tables. So based on that > "realtek,rtl9302-soc" and "realtek,rtl9302-switch" as two separate > things make sense to me. > OK, -soc and -switch are fine with me. > I'm still trying to figure out a bit more of the details. The block > diagram looks a lot like you'd expect to see with a traditional DSA > switch where you have a SoC Ethernet NIC/MAC connected to one port of a > switch. But getting into the datasheet it looks like what they call the > NIC is actually just the DMA portion of the switch as the registers are > all in that second block. As is the MDIO interface. I'm considering that > maybe the DSA model isn't right for this and I should be looking at > switchdev instead. Best regards, Krzysztof
