My Linux laptop (thinkpad) was NFG for 24 hours, with GRUB Error 17.
[Duration was unrelated to severity, just I had other things to do
after I noticed the issue. ]
Problem was caused by shutting down hard (power switch) during boot
(due to real life intervening). Apparently I powered down at exactly
the wrong moment during boot ...
Error 17 means the MBR was read ok, but the /boot filesystem is NG,
format not recognized. Loosely translated, this means /boot
superblock was corrupted.
Why would shutdown write a superblock? Mis-calculated dirty-buffer
flush, maybe? Who knows what hardware does when power is removed? sigh
Wasn't really all that hard to fix, once I dug around ... finding the
RIGHT args for the FSCK to not fsck-up, of course ... but getting to
the point where I could FSCK at all was awkward.
Google for Error 17. (Of course I have alternate computers. This does help.)
Boot Knoppix 5 DVD from UK Linux (Pro) Magazine.
See that /dev/hda2 and /dev/hda4 are auto-detected, have icons on
Knoppix desktop.
HDA2 is clean. HDA3 is swap (/mnt/hda2/etc/fstab agrees it should be),
but where's /dev/hda1 => /boot ? FSTAB says it should be ext3 on
hda1.
Gpart can't guess, it's says everything is 0 size. That's helpful.
Just helpful. Can't gpart find and use default alt-superblocks in its
heuristics ???
Knoppix didn't even create a block device for hda1 since it wasn't
automountable. So I had to
# cd dev
# ls hda* # see hda hda2 hda4 all looking similar
# man mknod # haven't used this command in YEARS !!
# mknod b 3 1 hda1 # follow pattern
# chown root:disk hda1
# chmod g+w,o-r hda1
to match the others.
Once I did that, it still won't mount -r; it informs me it's got a bad
superblock (as I'd figured).
And FSCK refuses to do anything with bad superblock either.
I had to read the Man page to find out the default backup-superblock
position for a 1K block size, and then ask it "are you 1k blocksize"
by doing
# fsck.ext3 -b 8193 -B 1024 -n /dev/hda1
to see if it was a 1k blocksize /boot, in which case it would have
simple errors, or if (since /vmlinux is a big file) someone made it
2k or 4k block, that would die gloriously and I'd try the other
default -b's for -B 2024 and -B 4096. Since it was ok but for block 3
on 6+ pass, I ran it for real.
# fsck.ext3 -b 8193 -B 1024 -n /dev/hda1
It prompted to ignore a read error and prompted to re-write superblock.
So then I
# mount -t /dev/hda1 /mnt/hda2/boot
and poke around to see that it looks good. Ok, time to try booting again.
Then I remembered to
# umount /mnt/hda2/boot
before
# umount /mnt/hda2
so we're clean, and before shutdown Knoppix for reboot and remove DVDrom
... and see Grub choice menu come up.
Yeah! Knoppix wins again!
I'm glad that worked, since I'm still not sure how to mount the
several volumes in /dev/hda4 via LVM from Knoppix, if I'd had to
rescue my data. Knoppix 5 has an LVM command, but ...
