Hmm...

> I'm not sure I would *stop logging* the events.  At least, not the
> underrun events.  We should handle those more gracefully by setting
> to the next threshold level and continuing.

If I read the driver correctly we run hardcoded in store-and-forward
mode, so there is no "next threshold level".

We shouldn't see any underrun errors at all (but I might be misinterpreting
this whole stuff).


Martin