Spent the day hacking on the DBRI ISBN/audio driver used on the SS10, SS20 and LX SPARC workstations. The audio chipset is actually a cs4215. Pete Zaitcev and I figured out why the driver (as a module) wasn't loading and unloading properly, but that didn't change the fact that the driver wasn't working... though it has made debugging a lot easier since I can now load and unload the module at will. :) Apparently the driver is written with a "ring buffer", just an array that's treated in a circular fashion... only, things break when it attempts to go to the second set in the buffer. Finding that this was the issue was non-trivial. A work-around is simple - limit it to just one set... but I'll just go ahead and fix the problem with the ring buffer and mail it over to anton and davem. Nonetheless, I'm listening to the Beatles' "Sgt. Pepper" album on my SS10 running 2.4.0-test10pre1 right now, so I'm at least partially satisfied. It's such a silly bug, though, that I'm wondering how it got submitted in it's non-working state for inclusion (this driver, though pretty different, works in the 2.2.x kernels).