[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Install script problems in 2.3 on Intel
- To: bugs_(_at_)_openbsd_(_dot_)_org
- Subject: Install script problems in 2.3 on Intel
- From: Monty Brandenberg <montyb_(_at_)_jaxom_(_dot_)_eng_(_dot_)_pko_(_dot_)_dec_(_dot_)_com>
- Date: Fri, 26 Jun 1998 18:02:59 -0400 (EDT)
- Delivery-date: Fri Jun 26 15:04:26 1998
OpenBSD 2.3 is to be installed on a 9.1GB disk with three primary
partitions (FAT16 for NT, OS/2 for PartitionMagic, Unformatted for
OpenBSD or NextStep) and an extended partition with four logical
partitions (Linux ext2, linux swap, linux ext2, and unformatted
for openbsd or other). Several problems of varying severity
arise when fdisk starts:
. Because of the existing partition structure, fdisk prints
quite a bit of information out on startup. This scrolls
away instructions and some of the partition info. Perhaps
the console code actually has a scroll buffer that can be
paged and I just don't know how to use it yet.
. Fdisk's handing of logical partitions is somewhat unintuitive.
In my case, to view the last linux ext2 partition I need
to do 'select 3', 'select 1', 'select 1', 'select 1' giving
a 4-deep session requiring four 'exit' commands to finish.
It would seem to be more intuitive to just deal with this
as a two-level hierarchy:
0: FAT16
1: OS/2
2: Unformatted
3: Extended
4: Ext2
5: Linux swap
6: Ext2
7: Unformatted
. Because I had already formatted the disk and the instructions
had scrolled off, on my first run I simply exited fdisk to
proceed to disklabel. This meant that no partition had been
marked as 'A6' for OpenBSD. When disklabel then did it's
business, it seemed to default to parition 0 (FAT16 for NT)
and wrote the disklabel on some sectors. This munged some
partition info on the disk but I was able to boot and use
partition magic to fix up the ID's and all systems were
bootable. I would say that if no partition is marked A6,
install shouldn't proceed to disklabel. Also, we could have
multiple Openbsd partitions. In this case, install might
prompt for the one to use as the target and communicate
this to disklabel for useful defaults.
After the above problems, I used partition magic to label the
third primary partition as 'A6' and then booted the CD and proceeded
through the full install to the disklabel step. This is relatively
straight forward but two things arise:
. When using a primary or logical partition to contain an
OpenBSD system should a sector or even a full cylinder be
reserved at the beginning of the partition for the OpenBSD
disklabel? For that matter, where does the disklabel reside
under the possible partitioning scenarios? I didn't know
so when I created my disklabel partitions, I started them
at the very beginning of the third primary and it seemed to
work. MBR/primary 0 appeared to be undamaged when I was
finished.
. The heuristic code in disklabel that calculates likely values
when creating partitions seems to have a few problems. This
third primary partition has 4,112,640 sectors. I created
a, b, and h partitions:
a: 4,128,705 385,497
b: 4,514,202 256,977
h: 4,771,179 3,470,165
For h, it computed a size of 3,470,229 which is 64 sectors (one
logical cylinder's worth) more than I had left in the partition.
To see what would happen, I accepted this and it then reported
that only 3,470,165 were available.
--
Monty Brandenberg Consulting for:
MCB, Inc. Digital Equipment Corp.
45 Putnam Avenue mcbinc_(_at_)_world_(_dot_)_std_(_dot_)_com montyb_(_at_)_eng_(_dot_)_pko_(_dot_)_dec_(_dot_)_com
Cambridge, MA 02139 617.864.6907 978.493.2702
Visit your host, monkey.org