On Thu, 08 May 2025 at 20:16:33 +0200, Mark Kettenis wrote: > This makes no sense. It is the sdmmc_mem_decode_scr() call above that > sets sf->scr.bus_width. So by removing that call, you're skipping the > bus width change. And I guess that means you'll stick to 1-bit mode?
Yeah, ignore this one. My other RK3128 is able to boot without this and seems to be properly in 4-bit mode, though it still needs the other DMA patch or else it just fails to read any data.
