[tlug] Bad partition juju: NetBSD slices and c?fdisk

["Stephen J. Turnbull" <stephen@example.com>]

Here's the scenario:

Windoze in /dev/hda1, 4GB partition.
NetBSD in /dev/hda2, 4GB partition sliced into 3 (root, swap, usr)
          These get reported as /dev/hda{9,10,11}.
Debian in /dev/hda[35678], various.
About 8GB free space.

I need another partition for Coda's cache.  So I fire up fdisk, note
that NetBSD's logical partitions which _are_ detected by the kernel,
but not mounted, are _not_ detected by the partition programs.  Not a
big surprise, but....

I create a partition /dev/hda9 in memory, check the parameters, the
start/end cylinders/sectors make sense, but being of a suspicious turn
of mind I quit without writing, and check in cfdisk.  Same results, so
this time I write, reboot, and ... I can't mount NetBSD any more under
2.4.17 (Dweebian build, 686-optimized, with a full complement of
modules).

I can't boot NetBSD (but haven't tried in a couple months).

But I can mount NetBSD under 2.2.  So I'll get anything that I can't
live without off.

Oh, yeah, most recent LILO (22.2) doesn't like my partitions at all
either (bitches about "not on cylinder boundary", which has happened
before) and refuses to install (which hasn't).

Anybody got any idea what gives here?  I'm more worried that Coda will
blow me a razzberry when I'm on a bus somewhere than anything else
... sure looks like some partition info got mislaid somewhere.  (Note
the two problems are quite likely unrelated, except that partitions
are involved in both.)

-- 
Institute of Policy and Planning Sciences     http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
              Don't ask how you can "do" free software business;
              ask what your business can "do for" free software.