> I've been noticing a slugishness when there's one process (or more)
> actively accessing the disk.  I always thought it was due to this
> "blind transfer"; kernel has to copy a chunk of data between disk and
> memory without doing context switching.  Am I right?

Something like that, yes.  It also has something to do with the loss
of clock ticks and the eventual slowdown of the system clock (which is
currently worse with this driver).  I'm working on the clock problems.
The problem there is that the clock interrupt has lower priority than
scsi, slot, and serial interrupts.

> I'll sup and try the new driver when I see it in the CVS log.

It should have been in last night's--amidst the rest of the stuff.

I think there are a couple of problems that need to be fixed--that's
one of the projects for this weekend.  Puma's been running the new
driver since about 2am this morning (U.S. EST) with no apparent ill
effects.

-allen

-- 
Allen Briggs - end killing - allen.briggs@bev.net ** MacBSD == NetBSD/mac68k **
   Where does all my time go?  <a href="http://www.netbsd.org/">Guess.</a>