> So, this leads me to believe that I need to acknowledge the interrupt
> somehow, but there's nothing about doing that, and I've tried a
> couple of different things (writing 0x80200000 to the mask register,
> writing 0x00200000 to the mask register, reading from 0xbffffff0).

Whoops.  I mean the "status register", not the "mask register".

Is it possible that the PCI interrupts (including the SCSI interrupt)
are tied into some other Be interrupt that we need to handle, too?

Or am I doing something else that's preventing the SCSI device from
actually asserting an interrupt again?  Hmmm...  The ncr driver loses
even worse ("CACHE TEST FAILED: timeout." and "CACHE INCORRECTLY
CONFIGURED.").

-allen

-- 
 Allen Briggs                     briggs@wasabisystems.com
 http://www.wasabisystems.com/    Quality NetBSD CDs, Sales, Support, Service