When I built -current recently, I built a kernel too, and rebooted with
it.  I soon started seeing messages like

	swap_pager_clean: clean of page 22cd000 failed

They seemed (not surprisingly) to have something to do with VM usage,
since they showed up as soon as I ran anything big.  It was so severe I
couldn't recompile a kernel; I managed to binary-patch my kernel to NOP
out the call to printf.  Then the messages stopped, of course, but
under circumstances when they would have previously, iostat started
reporting ridiculously unbalanced CPU figures - I'd kick off a make in
the background and watch "iostat 1", and it would print a few lines
with reasonable values and then the sy figure would soar to about 98
and the us figure would drop to like 1.

Eventually I got fed up enough that I fetched the latest sparc snapshot
kernel from ftp.netbsd.org.  But that one does it too!

Jason said something about this being related to SCSI problems.  I have
trouble seeing how this could be, because I haven't touched the
hardware, and SCSI works fine for everything else.

I suspect it's somehow related to user-land changes, because my
previous kernel, which was working Just Fine when I did a "make build"
under it and built a new kernel, now misbehaves also.

Anyone have any idea where I should start looking?  It's so bad I can't
link a new kernel at present (I left it overnight and it still didn't
finish); I'm not sure what to do, perhaps I'll have to pick up a 1.1
kernel from ftp.netbsd.org to work under.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu