re: kernel's borked again

[matthew green <mrg%eterna.com.au@localhost>]

long ago i wrote:

   for my U60 with a GENERIC.MP i get this:
   
   sd0 at scsibus0 target 0 lun 0: <IBM, DDYST1835SUN18G, S94A> disk fixed
   extent `psycho0 dvma' (0xfe000000 - 0xffffc000), flags = 0x0
        0xfe000000 - 0xfe0f3fff
   extent_free: start 0x4236700, end 0x42366ff
   panic: extent_free: region not found
   Stopped in pid 0.2 (system) at  netbsd:cpu_Debugger+0x4:        nop


at the suggestion of macallan i had a look at my cpus.  turns out
i had mis-matched cpus with 2MB and 4MB cache, and, in the right
order, the prom will let this happen.  in the opposite order, it
complains and powers down the machine.

in any case, once i put matched cache cpus in, i've gotten my
system booted to multiuser with a raid1 root.

this now works for me!  it was all a hardware issue.


.mrg.


ps: in other port-sparc64 news, my system currently has an install
that has these new features:
        - cc -m32 JFW (all the GCC/binutils parts of this are
          already in src)
        - 32 bit apps JFW by default -- the libs / ld.so come with
          the system, and are infact targetted for ultrasparc cpus
          in 32 bit mode so they are infact better than simple
          32 bit libs
        - does not require anything in /emul/netbsd32

there's one ugly hack i want to fix before i consider commiting
this work before netbsd-5, and it may not happen for that anyway.
i've done most of the work for amd64 to get the same feature but
i have not yet tested building 32 bit libs in the amd64 build.

stay tuned.