Isn't the switch already include in mainline: https://elixir.bootlin.com/linux/v6.16.3/source/drivers/net/dsa/realtek/rtl8365mb.c#L630 ?

Hi @j0ju, Created a thread so everything related to the switch stays together

sounds good, TBH I will continue to investigate and but my goal is to continue with the alpine- or OpenWRT based aproach Just found OpenWRT HEAD has support for the MangoPi with the T113s. OPenWRT is small and has all facilities for the switch controll already integrated, so that sounds intrguing.

That is totally understandable and maybe the way forward instead of building a own stack in buildroot. There are not a lot of devs who want/have the time to bring the BMC-firmware to a next level and it would be a shame if the project fades away

From what I understand in the patch, it takes a shortcut due to the way the original sunxi-i2c driver could no do FASTSTART i2c transfers, which the switch needs to be talked to. My guess that why the default primitives/callbacks are replace tih direct calls to the i2c stack. just guessing. Might be working with vanilla kernel, but then the dtb needs to be adapted and TBH there are some things I do not fully understand on that level. Open to sugestions. https://github.com/turing-machines/BMC-Firmware/blob/master/tp2bmc/patches/linux/i2c.patch

Yeah your more or less right. So realtek talks slightly different i2c. I cannot remember exactly anymore, i think in the patches there was also an attempt done in separating a back and front end for the i2c commands… I might have time to revisit all of this the day after tomorrow …