Grub entries are:-
findroot (rootfs0,0,a)
kernel /platform/i86pc/multiboot
module /platform/i86pc/boot-archive
Also, got an entry in grub for failsafe solaris boot and this boots up
ok:-
findroot (rootfs0,0,a)
kernel /boot/multiboot kernel/unix -s
module /boot/x86.miniroot-safe
Its on a laptop and I've got a feeling the user may have been powering
off without shutting down properly and that this has corrupt
something.
Anyway, is there a command to fixboot or something?
Or would I be better off booting from sol10 cd and doing an inplace
upgrade?
Have you tried booting from CD/DVD and running fsck on your disk? All
partitions, of course; If the boot partition is fsck'ed up you have to
expect that the rest of the disk may be in less than perfect shape.
No. I'll try that.
That shouldn't matter.
Also, if your failsafe boot option works okay, it shouldn't be necessary to
boot from DVD/CD. Just boot from failsafe, and do your checks from
maintenance mode.
--
Brandon Hume - hume -> BOFH.Ca, http://WWW.BOFH.Ca/
Possible, but not sure
I had the same problem (grub shows menu, sun copyright visible and then
reboot) one time when I changed my scsi controller or when I changed its
position in the PCI slots.
I don't know if you have the same kind of problem with the change of the
boot order ... but if it can help , to fix my problem, I had to do
theses things :
- Boot with failsafe choice in grub
Solaris searchs installed instance on the disk, and asks if it must be
mounted r/w on /a (answer : y)
- Write on a paper the path (/dev/dsk/cXtYdZsN) of the root slice
mounted on /a (the path is shown in the previous question about mounting
the instance in /a)
- Find its "hardware" path ( /pci@0,0/pci-ide@1f,1/ide@0/....) with a
"ls -al" in /dev/dsk (and not in /a/dev/dsk)
- Look in /a/etc/vfstab to see if /usr is on another slice, and mount it
in /a/usr if needed
- Run a chrooted shell in /a ( chroot /a /bin/ksh --login )
In the chrooted shell, run :
devfsadm -C
and then
devfsadm
- Check in bootenv.rc (must be in /platform/i86pc/boot or something like
that) that bootpath matches the "hardware" path of the /a mounted slice,
and fix it if needed
- Exit the chrooted shell
- Reboot with /a mounted (the shutdown process will update the
boot-archive on /a)
Hope it can help ...
Nicolas
Edit the grub entry to enter kmdb on boot. See
http://blogs.sun.com/dmick/entry/diagnosing_kernel_hangs_panics_with
--
Ian Collins
Which Solaris release is this? Sounds to me like your boot archives
need to be updated, depending on the Solaris release. You might
review the bootadm update-archive command, described here:
http://docs.sun.com/app/docs/doc/817-1985/gglaj?a=view
Or, do a google search of bootadm update to see other examples
that might apply to your scenario.
Cindy
Cindy
Cindy,
Thanks. Already tried that.
fsck reports no problems....
I'll give this a try...
sol 10 x86
And precisely which version of Solaris 10? The first release was almost
5 years ago, and the 'current' version is the 8th update. Update 1 added
grub, and update 6 added ZFS boot. Both of these change significantly
booting.
A bientot
Paul
--
Paul Floyd http://paulf.free.fr
Bertie,
If this is a Solaris 10 10/08 system with a UFS root, for example,
then your grub entries
look reasonable to me.
Maybe the problem is with the disk. I would boot from an s10 CD and
use format to
review the disk label to see if it is still intact.
Cindy
Its Sol 10 05/09. I still seem to be able to access all my files if I
boot from failsafe.
Its just if I try to boot normally, it does the SunOs release 5
banner, then resets the machine and goes back into grub.
1. findroot (rootfs0,0,a) command is fine because failsafe boot uses
the same command and this works. So it can find the disk.
2. module /platform/i86pc/boot_archive must be OK because I've rebuilt
the boot archive with the bootadm command.
Therefore, this leaves the actual kernel? kernel /platform/i86pc/
multiboot
Could this be corrupt or something? Is there any way to rebuild this?
OK. Got it to enter kmdb.
All I get is :-
Loading kmdb...
welcome to kmdb
[0]> ?
Any ideas what next?
Got into kmdb and ran :-
prom_debug/W 1
:c
and this seems to attempt to boot.
At the end, there was a message which seems to be pertinent:-
Cannot mount root on /pci@0,0/pci-ide@1f,2/ide@0/cmdk@0,0:a fstype ufs
panic[cpu0]/thread=fffffffffbc26480: vfs_mountroot: cannot mount root
Hmmm. So for some reason the kernel cant see the root disk. I know its
there because failsafe can see it?
Any ideas?
Rebuild the kernel? ????
I wouldn't assume that. When the failsafe boots, it runs off the ramdisk.
Does it offer to mount the hard disk root on /a?
If you boot failsafe, and then pop into a shell, is /a mounted? What
device is it attached to?
What happens if you do a regular boot, adding '-r' to the grub flags? It's
possible your /dev tree is damaged.
Yes. root filesystem mount on /a and everything seems to be there.
Already tried reboot -- -r and reboot -- -ra with no luck.
That's interesting. When you boot failsafe, and drop to a shell, try
chrooting to the mounted broken root. Use the command:
chroot /a /sbin/sh
When it's done, you'll be running /sbin/sh inside the root. / will refer
to /a and so on.
Check /etc/vfstab and make sure it refers to the proper devices. You can
also try "fsck -n -m /" and see how it behaves (keep in mind that you're
running fsck on a mounted, live filesystem...)
Are you using ZFS at all?
See my post on 12/10/09 18:11
Have to try to see if bootpath in bootenv.rc (it is actually be set to
/pci@0,0/pci-ide@1f,2/ide@0/cmdk@0,0:a ) matchs the hardware path on
which the failsafe os has booted ?
Which post was that?