Re: Booting or installing Solaris 9 intel on a secondary IDE drive
From: Juergen Keil (jk_at_tools.de)
Date: 08/19/04
- Next message: Mark Andrews: "Re: Booting or installing Solaris 9 intel on a secondary IDE drive"
- Previous message: Thomas Tornblom: "Re: Can you have a / in a *file* name??"
- In reply to: Mark Andrews: "Booting or installing Solaris 9 intel on a secondary IDE drive"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Thu, 19 Aug 2004 13:34:03 +0200
mta61@yahoo.com (Mark Andrews) writes:
> I was playing with Solaris 9 for Intel on the weekend and installed it on
> the slave drive on the first IDE bus. Since the Solaris install insists
> on installing the o.s. on c0d0 and i didn't want to blow away my XP install
> on the master drive on the first bus, i used a bit of trickery:
>
> (1) Disconnected the master drive
> (2) Recabled the slave drive as the master drive
> (3) Did the install
> (4) Rebooted the machine and Solaris ran correctly
>
> This all stopped when i restored the machine to its original hardware
> configuration:
>
> (1) XP booting off the first IDE master
> (2) Solaris installed on the first IDE slave (c1d0)
An IDE slave drive would use "c?d1" (IDE master is "c?d0").
It's a bad idea to move a Solaris boot disk from master to slave
drive, or move it from the primary to the secondardy ide channel.
The same problem exists with scsi hdds when you change the target id
of the drive or move the scsi host controller to a different pci slot
or different pci bus.
It's a bad idea because Solaris records the physical device path in
the /etc/path_to_inst file on the root filesystem. After you've moved
the disk and boot Solaris, Solaris does not find the root disk on the
physical device path it has remembered. It does find a "new" disk on
a different physical device path, but the new disk uses different
driver instance numbers. That is, your root disk device has changed
it's device name; for example from c0d0s0 to c1d0s0. Typically
/etc/vfstab file contains the old device name (c0d0s0), and Solaris is
unable to fsck the root filesystem, starts an emergency root shell and
you're left with a read-only mounted root filesystem.
http://www.riddleware.com/solx86/SCSI_change.html contains
instructions on how to fix these problems after moving a Solaris boot
drive.
> How do i boot the Solaris boot partition and start Solaris from c1d0
> or should i do a reinstall onto c1d0 using some trick i'm not aware of? ;-)
It should be possible to install Solaris on the IDE slave device, or on
an IDE device connected to the secondary IDE channel.
There is a problem with the Solaris x86 master boot record (MBR) code,
though. When an IDE disk is using LBA mode (i.e. all current IDE
disks), it tries to read the active partition's boot record from bios
drive "C:" (0x80); the 0x80 bios device is hard-coded in the Solaris
MBR. The Solaris MBR code ignores the boot device passed from the
system's bios. If your system's bios tries to boot from drive D:
(0x81), E: (0x82), ... the Solaris MBR would load OK, but the Solaris
MBR code would try to boot the active partition from the C: (0x80)
drive, and crashes with an error message "Bad PBR sig!!!"
For that reason I always replace the MBR on all of my HDDs with (a
slightly modified version of) FreeBSD's Booteasy MBR code. It offers
a menu to boot from any of the four primary partitions (by pressing
one of the F1 .. F4 keys), the selected partition is marked active and
is the default boot partition on the next reboot. And F5 in the MBR
boot menu cycles through the available bios disk drives, enabling
booting from a drive connected as IDE slave device, or connected on
the secondary IDE channel.
- Next message: Mark Andrews: "Re: Booting or installing Solaris 9 intel on a secondary IDE drive"
- Previous message: Thomas Tornblom: "Re: Can you have a / in a *file* name??"
- In reply to: Mark Andrews: "Booting or installing Solaris 9 intel on a secondary IDE drive"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|
