On Mon, 9 Jun 1997 15:57:49 -0500 (CDT),
"David A. Gatwood" <marsmail@globegate.utm.edu> wrote:
> On Mon, 9 Jun 1997, Renier Linssen wrote:
> > Here's what happened since (sorry if I'm too verbose; I thought nothing
> > could be too much at this point).
> > 
> > I booted in single-user-mode, fsck was run by the boot-process, and the fs
> > was marked clean. So I entered:
> > 
> >         # reboot -n
> > 
> > as the FAQ suggested.
> 
> I'm always hesitant to reboot anything w/o syncing, especially since the
> fs should have been r/o at the time, but....  Whatever.

Actually, because fs has been mounted r/o at the time, it doesn't make
any difference whether to "reboot" (with syncing) or to "reboot -n"
(w/o syncing).  Reason: there's no dirty buffer to flush.  "Reboot -n"
is suggest because it directly reads/writes the fs bypassing the
buffers, but I'm not sure whether it's still necessary under NetBSD.

> > I restarted without any extensions (with my usual startup settings the
> > booter hangs because of VM or something?) and booted NetBSD again in Single
> > User Mode (SUM).
> 
> <major snip>
> 
> >         # fsck -f
> >         > marked clean
> >         # ^D (another one from the FAQ. Maybe I should have used "reboot
> > -n" again...)
> >         > login: root <return>
> >         > do not login as root, use su
> 
> That's fine.  At that point, your system was working normally.  It's just
> suggested that you create a user account and go from the user account to
> root using the su program instead of logging in directly as root.  No
> biggie.
> 
> >         # reboot -n
> 
> ewww...  Never reboot -n after the system is mounted read-write unless you
> use sync several times (and even then, there will be some corruption).
> reboot -n is generally a bad thing, in my experience.  Unless things are
> really hosed, of course.

David is right.  Use "shutdown -r now" instead.

> > HANGS (after
> >         "Changing root device to sd1a
> >         PRAM: 0x339c3fgd, macos_boottime: Ox339c3fg0"
> 
> Sounds like trashed fs from the reboot -n.

I'm not so sure.  There may be more subtle problem with the disk drive.
Does the kernel use the ncrscsi driver or the sbc driver?  Might help
if you switch.

> > it reports a
> >         fatal signal: Illegal Instruction.
> 
> According to my guess, this is probably because your LC III doesn't have
> an FPU, I don't think.  (If I'm wrong, then something's really weird.) 
> Ken Nakata (who I think this is carboned to) has pretty-much working
> emulation in the kernel, but a few instructions are evidently not
> implemented (or something like that, I forget the details, read the FAQ). 
> Thus, a special version of the math libs is available that avoids those
> instructions.  I forget where they are, but there's a link from the FAQ.
> Sometimes the lack of fp math libs can also cause problems for fsck, but
> since it didn't flake out on you... that's probably not causing your disk
> problems.

Hmm, although I cannot confirm or disconfirm, I have some doubts that
the lack of FPU is causing this particular trouble...

Good luck,

Ken