Skip to contentSkip to conversation

Oops: UMOUNTING in 2.4.17 / Ext2 Partitions destroyed (3x)

DISCUSSION Mike 2 participants 6 messages Dec 30, 2001 – Dec 31, 2001

lore
Share & tools
lore

Mike

author
#0 lore raw

Hello,

I recently had a horrendous experience while unmounting some partitions. Ironically exactly those I wanted to backup ASAP :-(

SITUATION:
2 weeks ago I assembled a brand new Duron system. I connected my brand new UDMA/100 HDD on /dev/HDA and my old UDMA/33 one on /dev/HDB. Jumpers are correct. Cable is 80 ribbon. All the funky stuff in the BIOS has NEVER been enabled, so everything has been at default from assembly on. The only thing somewhat not-kosher was the fact that the SLAVE drive was mounted at the END of the ribbon cable (instead of in the middle) as a result of having installed this drive as an afterthought and not having the space. I still don't know how relevant this is.
CONTENTS OF /dev/hdb:
hdb3: old /home
hdb5: old /
hdb6: old /usr/local
In the last week I've mounted and unmounted these a couple of times to copy stuff. I mounted /home the most, I mounted /home like twice and /usr/local never.

WHAT HAPPENED:
So yesterday I had /dev/hdb3 mounted...I try to umount it from /mnt and it oopses on me!? According to 'mount' it still mounted and there's no way to (u)mount it again. I thought "OK whatever" and try to mount /dev/hda5...OOPS again. Remebering what important university reports and stuff I had there I completely freak out and change /etc/fstab to e2fsck it at bootup and reboot the machine normally.
Now the fun starts: at bootup it tells me that it can't find any superblocks on any of the THREE partitons anymore...they're all gone. No matter what I tried with e2fsck it couldn't do anything. 'sfdisk -V' reported everything OK and /dev/hda1 (vfat) still mounts and umounts correctly.

At first I thought maybe the partitions were long overdue to be force-fscked but they'd been asymmetrically (u)mounted (every one a different amount of time).
Rest assured I had NOT been doing any "weird" things, I had (u)mounted stuff just like I've been doing for nearly 2 years.
This is all mind-boggling to me, so if anyone has any suggestions for me no matter how silly let me know! I want my data back!

Please CC me as I'm not subscribed to LKML.
Thanks in advance and thanks for a great kernel.
-Mike Neman

author
#1 lore raw

I'm not too e-mail savvy so I HOPE I've sent this e-mail and CC correctly...if not let me know.

Quoted 3 lines · Lionel Bouton

By any chance do you use a SIS735 based mainboard ?

If so, try your disks in another system or boot with "ide=nodma".

Yes indeed, it's the (in)famous ECS K7S5A with SiS 735 chipset. Thanks I'll try this ASAP (which implies January 3rd or such because I'll have to strip the new system bare and reassemble the old one again after regaining counsciousness after Jan .1st and a hot date the 2nd.) Merci and Greets!
-Mike

author
#2 lore raw
In reply to Mike
Quoted 10 lines · Pierre Rousselet

I met something like this with the root fs while testing some 2.4.X-preX. An fsck -f from my rescue floppy was telling me 'clean' but the boot process hung with e2fsck exiting with an error digit.

I forced mounting rw the root fs after disabling fsck, re-compile e2fsprogs, re-boot and it was OK.

This is only a rescue approach. The disk corruption you met with 2.4.17/ext2 remains to address.

Here's something I can't remember EVER seeing before at bootup:
"---SNIP---
Partition check:
hda: {everything fine here}
hdb: [PTBL] [1108 255 63] hdb1 hdb2 <hdb5 hdb6 hdb7 hdb8> hdb3 ---SNIP---"
The PTBL stuff and 1108 255 63 (CHS??) !!! Why?

I tried 'e2fsck -f /dev/hdb3' and it returned:
"Filesystem has unsupported Read-Only features while trying to open /dev/hdb3.
The SuperBlock could not be read or does not describe a correct ext2 filesystem.
....If it /is/ ext2 then the Sb is corrupt, run 'e2fsck -b 8193 <device>'"

So I do 'e2fsck -b 8193 <device>' and it says: "Bad magic number in SB while trying to open /dev/hdb3"

For /dev/hdb5 it says: "Couldn't /find/ ext2 SB...trying backup blocks...Bad magic number in SB while trying to open /dev/hdb5."

Any more ideas?? I already heard from Lionel Bouton to hook up the device on it's old system but that'll have to wait till after Jan. 2nd. TIA,
-Mike

author
#3 lore raw
In reply to Mike
Quoted 5 lines · Lionel Bouton

From my experience with this chipset you might not have lost much data.
If your e2fsck failed without the opportunity to ever write something, booting with "ide=nodma" and only then performing a filesystem check might put the system on its feet again.

I have NOW put "append = "ide=nodma"" in my /etc/lilo.conf, Thank you. Yes, please keep me posted.
-Mike

#4 lore raw
In reply to Mike
Quoted 3 lines · Mike

I tried 'e2fsck -f /dev/hdb3' and it returned:
"Filesystem has unsupported Read-Only features while trying to open /dev/hdb3.

Try "tune2fs -l /dev/hdb3" or "debugfs -h /dev/hdb3" (both do the same thing - spit out the ext2 filesystem superblock data. What version of e2fsck are you using?

Quoted 6 lines

The SuperBlock could not be read or does not describe a correct ext2 filesystem.
....If it /is/ ext2 then the Sb is corrupt, run 'e2fsck -b 8193 <device>'"

So I do 'e2fsck -b 8193 <device>' and it says: "Bad magic number in SB while trying to open /dev/hdb3"

You are better off trying "e2fsck -B 4096 -b 32768 /dev/hdb3" instead.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/

author
#5 lore raw
In reply to Mike
Quoted 16 lines · Andreas Dilger
Quoted 3 lines · Mike

I tried 'e2fsck -f /dev/hdb3' and it returned:
"Filesystem has unsupported Read-Only features while trying to open /dev/hdb3.

Try "tune2fs -l /dev/hdb3" or "debugfs -h /dev/hdb3" (both do the same thing - spit out the ext2 filesystem superblock data. What version of e2fsck are you using?

Quoted 6 lines

The SuperBlock could not be read or does not describe a correct ext2 filesystem.
....If it /is/ ext2 then the Sb is corrupt, run 'e2fsck -b 8193 <device>'"

So I do 'e2fsck -b 8193 <device>' and it says: "Bad magic number in SB while trying to open /dev/hdb3"

Quoted 7 lines

You are better off trying "e2fsck -B 4096 -b 32768 /dev/hdb3" instead.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/

Sorry for the late reply....new year's eve etc...haven't been at my box for over 24 hours.
I used e2fsck version 1.22 booted as a single user/maintenance system. -Mike