Small update: I followed the track Sven started in November a bit through kernel sources and found that the BUG-ON() statement which causes the crash might be itself a bug, because it blocks a code path that seems legit to me. At least, when I experimentally removed the BUG_ON(), aplay does not crash any more but produces a i2s WS signal for the right duration However, no sound yet - I haven't been able yet to check if this is a problem of my hardware (with very difficult to solder MAX98357A) or still DMA not piping the data to the i2s. I'm also trying to check with the author of that BUG_ON statement to find out why exactly he did put it there.