>For what it's worth, I'm seeing the same on an old i386 box here.
>It's currently doing a "make build" while running a UVM kernel built
>from the April 25th tar balls, and seems happy.  Adding PMAP_NEW to
>the kernel config file gives a kernel that boots, says the right
>things about what it will use for root and swap, but panics when it
>can't find /sbin/init.  I've been through the various permutations of
>boot blocks and partition IDs, just to be sure.

from what's just been reported it seems pretty likely there is a problem
in pmap.new.c for older i386 processors, so i wouldn't worry about boot
blocks and partition IDs.   there are differences in the TLB flushing
code between i386 and >=i486 systems, so that is what i would suspect.
unfortunately, i don't have a working i386 handy to play with it.   maybe 
matthias can look into it a bit more once he gets his NFS server patched
up?   it is dying early on, so maybe it will not be that hard to 
track down.

chuck