On 3/21/20 11:13 PM, Marek Vasut wrote: > On 3/22/20 4:08 AM, Sean Anderson wrote: >> On 3/21/20 11:04 PM, Marek Vasut wrote: >>> Could be a different revision of the IP. This is usually handled by >>> using SoC-specific compatible strings, see e.g. Linux >>> Documentation/devicetree/bindings/usb/renesas,usb3-peri.yaml >>> >>> You don't want to encode register layout in the DT. >> >> Ok, then I think adding compatible strings would be the cleanest. >> >>> >>>> It is possible to >>>> add several compatible strings like "kendryte,k210-spi3". >>> >>> Why ? >> >> See below. >> >>> >>>> I chose this >>>> method because the bitfields are different for spi0 and spi1, spi2, and >>>> spi3. If there are other incompatibilities discovered, then it may make >>>> more sense to use different strings. Another option could have been to >>>> use the DW_SPI_VERSION field to detect different controllers, but it is >>>> the same among all the controllers on the K210. >>> >>> The controllers on the same SoC have different register layout ? >>> >> >> Yup! >> >> Don't ask me why. > > Now that is truly an odd design. Is there a datasheet for this SoC ?
Nope. Everything has been done referencing their sdk. In hindsight, porting this board was a poor decision because of all the hoops I've had to jump through due to the absence of documentation. Unfortunately, it's probably the only budget risc-v board with S-mode until the end of the year or so. > You might be able to somehow enumerate those controllers in DT and > derive their layout from that enumeration or somesuch I think. So like derive it from the sequence number? I'd rather do this in a more standard way, if possible. --Sean
