Esteemed Colleagues:
Although it is unsupported, I have successfully installed OpenBSD on a
logical slice within the extended slice of an MBR-partitioned disk.
It took days. I may post, to this mailing list, a description of how
I figured out how to do it; but the write-up would take hours, so it
may not appear right away.
Here is the edited output from "disklabel sd0" and from "fdisk -v sd0":
m5# fdisk -v sd0
Primary GPT:
Not Found
Secondary GPT:
Not Found
MBR:
Disk: sd0 geometry: 58369/255/63 [937703088 Sectors]
Offset: 0 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
-------------------------------------------------------------------------------
*0: 0C 57389 130 22 - 58369 80 30 [ 921962496: 15740559 ] Win95
FAT32L
1: 07 13 0 52 - 19594 64 54 [ 208896: 314572800 ] NTFS
2: 07 0 32 33 - 13 0 51 [ 2048: 206848 ] NTFS
3: 05 19594 64 55 - 57389 130 21 [ 314781696: 607180800 ] Extended
DOS
Disk: sd0 geometry: 58369/255/63 [937703088 Sectors]
Offset: 314781696 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
-------------------------------------------------------------------------------
0: 8E 19594 97 24 - 26121 118 45 [ 314783744: 104857600 ] Linux LVM
1: 05 26121 118 47 - 30037 215 2 [ 419641345: 62916607 ] Extended
DOS
2: 00 0 0 0 - 0 0 0 [ 314781696: 0 ] unused
3: 00 0 0 0 - 0 0 0 [ 314781696: 0 ] unused
Disk: sd0 geometry: 58369/255/63 [937703088 Sectors]
Offset: 419641345 Signature: 0xAA55
Starting Ending LBA Info:
#: id C H S - C H S [ start: size ]
-------------------------------------------------------------------------------
0: A6 26121 151 16 - 30037 215 2 [ 419643393: 62914559 ] OpenBSD
1: 05 30037 215 5 - 31212 215 43 [ 482557954: 18876414 ] Extended
DOS
2: 00 0 0 0 - 0 0 0 [ 419641345: 0 ] unused
3: 00 0 0 0 - 0 0 0 [ 419641345: 0 ] unused
[ omitted lines ]
m5# disklabel sd0
# /dev/rsd0c:
[ omitted lines ]
16 partitions:
# size offset fstype [fsize bsize cpg]
a: 57358559 419643393 4.2BSD 2048 16384 12960 # /
b: 5556000 477001952 swap # none
c: 937703088 0 unused
d: 860160 627273728 MSDOS # /mnt/boot
i: 206848 2048 NTFS #
/Windows-Boot
j: 314572800 208896 NTFS # /Windows
k: 15740592 921962496 MSDOS # /FreeDOS
l: 104857600 314783744 unknown
m5#
(If anyone is wondering about /dev/sd0l, it is the LVM volume group
containing Linux logical volumes, on which my collection of Linux
operating systems resides. These are visible on FreeBSD. If anyone
on this mailing list knows how to make Linux logical volumes visible
on OpenBSD as they are on FreeBSD, I would welcome seeing the
information on this mailing list.)
There is still, however, one problem (which may be related to the
fact that installation on a logical slice is unsupported, although I
don't see how it is): the bootloader does not work. This is not
fatal, because I can get GRUB to boot the kernel directly, with
"kopenbsd /bsd -r sd0a" (more on that later). But it is troubling.
I have two operating systems on my computer that can install GRUB:
Solaris and Linux. Both of them ought to be able to boot OpenBSD by
chainloading to the OpenBSD bootloader, thus:
set root=(hd0,6)
chainloader +1
Here is the output of "installboot -nv sd0":
m5# installboot -v sd0
Using / as root
installing bootstrap on /dev/rsd0c
using first-stage /usr/mdec/biosboot, second-stage /usr/mdec/boot
copying /usr/mdec/boot to //boot
looking for superblock at 65536
found valid ffs2 superblock
//boot is 6 blocks x 16384 bytes
fs block shift 2; part offset 419643393; inode block 56, offset 2928
expecting 64-bit fs blocks (incr 4)
master boot record (MBR) at sector 0
partition 0: type 0x0C offset 921962496 size 15740559
partition 1: type 0x07 offset 208896 size 314572800
partition 2: type 0x07 offset 2048 size 206848
partition 3: type 0x05 offset 314781696 size 607180800
extended boot record (EBR) at sector 314781696
partition 0: type 0x8E offset 2048 size 104857600
partition 1: type 0x05 offset 104859649 size 62916607
extended boot record (EBR) at sector 419641345
partition 0: type 0xA6 offset 2048 size 62914559
/usr/mdec/biosboot will be written at sector 419643393
installboot: /usr/mdec/biosboot extends beyond sector 268435455. OpenBSD
might not boot.
m5#
So we see that installboot correctly determines that OpenBSD is on a
logical slice, it correctly determines the beginning of that logical
slice, and it correctly writes the bootloader to sector 419643393, which
is exactly where GRUB should find it, upon executing "chainloader +1".
And, indeed, GRUB does not complain about an invalid signature. It
definitely does find a bootloader there. But this is what happens
when it boots (there may be typographical errors, I am copying
manually from the screen):
Loading......
probing: pc0 mem[629K 511M 510M 2175M 7M 7M 22M 19M 2M 64K 13030M a20=on]
disk: hd0+*
>> OpenBSD/amd64 BOOT 3.55
open(hd0a:/etc/boot.conf): Invalid argument
boot>
cannot open hd0a:/etc/random.seed: Invalid argument
booting hd0a:/bsd: open hd0a:/bsd: Invalid argument
failed(22). will try /bsd
boot>
cannot open hd0a:/etc/random.seed: Invalid argument
booting hd0a:/bsd: open hd0a:/bsd: Invalid argument
failed(22). will try /bsd
Turning timeout off.
boot> ls
stat(hd0a:/.): Invalid argument
boot> ls /
stat(hd0a:/): Invalid argument
boot> machine diskinfo
Disk BIOS# Type Cyls Heads Secs Flags Checksum
hd0 0x80 label 1023 255 83 0x0 0xe61ca3fc
Clearly the bootloader that was installed from /usr/mdec/biosboot was
successfully executed, and it successfully executed /boot. The
secondary bootloader, however, does not see the filesystem. That it
cannot find /etc/boot.conf does not trouble me, because there is no
/etc/boot.conf because I did not create one, and the /etc/boot.conf
file is, if I am not mistaken, optional. But it cannot find /bsd.
When I typed the ls command it was clear that it cannot even find the
current or the root directory (which, parenthetically, ought to be the
same). It does find the hd0 disk. And when I repeat the procedure
when USB flash drives are connected, it finds the hd1 disk and the hd2
disk also. It finds disks. But it does not find files. I even,
grasping at straws, linked /dev/sd0a to /dev/hd0a (and did the same,
mutantis mutandis, with b and c). It made no difference. I recently
saw a posting on another mailing list that presented the belief (the
author of the posting was not sure that it was true) that the
bootsector of a logical slice contains relative rather than absolute
positioning information. If this belief were true, and if /boot could
not be made aware of that fact, it might, maybe, explain the inability
of /boot to find the filesystem on which it resides. But that belief
is contradicted by the output of fdisk, shown above. Moreover, the
primary bootloader is able to find the secondary bootloader; so,
unless it does so by means of blocklists, the secondary bootloader
should surely be able to find /bsd. It does not. How do I get the
OpenBSD bootloader to boot OpenBSD?
That this is not a fatal error is only because GRUB was able to boot
the OpenBSD kernel directly thru kopenbsd:
grub> insmod ufs2
grub> set root=(hd0,6)
grub> kopenbsd /bsd -r sd0a
grub> boot
But this is only true of the GRUB that is installed by Linux. The
GRUB that is installed by Solaris is, apparently, an older version
that cannot boot the kernel directly. Behold:
grub> insmod ufs2
grub> set root=(hd0,6)
grub> kopenbsd /bsd -r sd0a
error: only device specifications of form wd<number><lowercase letter> are
supported.
So this is a bootloader that checks the arguments and options of
kopenbsd, which I did not ask it to do, and which thinks that an
OpenBSD root disk has to have a name beginning with wd. I got around
that. I linked /dev/sd0a to /dev/wd0a. Now I get this:
grub> insmod ufs2
grub> set root=(hd0,6)
grub> kopenbsd /bsd -r wd0a
grub> boot
error: overlap detected.
A "grep -r" (this is a rhetorical figure, there is no "grep -r" on
Solaris) informs me that the error message comes from relocator.mod.
Apparently relocator.mod is detecting a condition that would be
nonfatal if ignored, and treating it as a fatal error; and apparently
this bug was fixed on Linux, but Oracle never updated the GRUB on
Solaris. Still, there was some condition that was detected. What was
the condition? And how do I fix it?
So far my efforts to boot OpenBSD from the Solaris GRUB have all
failed. I tried to get around the problem by doing "kopenbsd /boot",
by doing "legacy_kernel /bsd /bsd -r sd0a", and even by doing "knetbsd
/bsd -r sd0a". None of these elicits an error message. But they all
fail, booting into the blank screen of death. The last two should
fail, so I am not complaining that they do. But "kopenbsd /bsd -r
sd0a" should succeed. And, what is more important to the readers of
this mailing list, "chainloader +1" should succeed -- the OpenBSD
bootloader should be able to boot OpenBSD. Why can't it?
Unrelated to the above, I see from the output of "ifconfig -a" that my
newly-installed OpenBSD system did not configure my wireless device,
which, according to the outpout of "pcidump -v", is a Broadcom BCM4313.
I assume that the failure to configure this device is due to my not
having yet invoked the fw_update command. I am puzzled, though: I
thought that fw_update is now invoked automatically within the install
procedure.
As always, thank you in advance for any and all replies.
Jay F. Shachter
6424 North Whipple Street
Chicago IL 60645-4111
(1-773)7613784 landline
(1-410)9964737 GoogleVoice
http://m5.chicago.il.us
[email protected]
"But when she traced the killer's IP address ... it was in the
192.168/16 block!"
