[2836] in linux-scsi channel archive
Re: Massive corruption
daemon@ATHENA.MIT.EDU (Kurt Garloff)
Wed Nov 19 15:40:03 1997
Date: Wed, 19 Nov 1997 12:43:31 +0100 (CET)
From: Kurt Garloff <garloff@kg1.ping.de>
To: Justin Miller <jhm@umr.edu>
cc: linux-scsi@vger.rutgers.edu
In-Reply-To: <199711180150.TAA13036@rocket.cc.umr.edu>
On Mon, 17 Nov 1997, Justin Miller wrote:
> micropolis is the one with filesystem corruption. When I started putting
> data together on my hd, for burning a cd this morning I noticed that a
> couple of scripts that I ususally
> use, weren't working, my /usr/local was a link to my ext2 partition on
> the micropolis drive, and it showed nothing when i did an ls. I decided
> to unmount it and do an e2fsck, e2fsck gave me a bunch of errors, I cant'
> remember what the errors were but after 93 of them e2fsck gave up trying
> to check the fs. This has happened before, using a different controler,
> and last time I wasn't sure what had caused it, since the partition was
> empty I just reformatted it with mke2fs, and it had been working fine
> until this morning.
> ...
> I pulled the hd out, and checked the jumper settings on the drive, and
> noticed that the jumper for parity was not shorted. Apparently I had
> forget to set that one. I put set the jumper, put the drive back in and
> booted into linux, fdisk now says:
> Warning: ignoring extra data in partition table 5
> Warning: ignoring extra data in partition table 5
> Warning: ignoring extra data in partition table 5
> Warning: invalid flag c2b9 of partition table 5 will be corrected by w(rite)
>
> Command (m for help): p
>
> Disk /dev/sdb: 131 heads, 63 sectors, 1017 cylinders
> Units = cylinders of 8253 * 512 bytes
>
> Device Boot Begin Start End Blocks Id System
> /dev/sdb1 * 1 1 187 771624 83 Linux native
> /dev/sdb2 188 188 1017 3424995 5 Extended
> /dev/sdb5 ? 375880 376901 470796387457826 3 XENIX usr
>
>
> sdb5,6,7 were each 1.1 gigs, now they seem to have dissappeared.
>
> I have idea on what to do now, I can't get e2fsck to do anything now it
> just says this even if I use a different superblock :
> e2fsck 1.04, 16-May-96 for EXT2 FS 0.5b, 95/08/09
> e2fsck: Bad magic number in super-block while trying to open /dev/sdb
>
> The filesystem superblock is corrupt. Try running e2fsck with an alternate
> superblock using the -b option. (8193 is commonly an alternate superblock;
> Hence, 'e2fsck -b 8193 <device>' may recover the filesystem.)
>
> Any help would be appreciated, I would like to get the data from
> my ext2 partition back, but I really would like to just keep this from
> happening again. Thanks.
1. I hope you know your old partition table. (I always print it with fdisk
-l |lpr)
Check geometry of the drive (CHS translation) and reconstruct your
partition table.
2. Now e2fsck your partitions (and not the drive) and hope it will do it.
Note: The current e2fsprogs are version 1.10 and might be able to do
better recovery of your data.
3. If you want to avoid future corruptions, suspect bugs everywhere and
try to find them.
- Bad sectors on your HD: check with badblocks -w (You'll lose your
data!)
- HD firmware bugs. Try disabling DISCONNECTION, TAGGED COMMAND Q,
SYNC nego, reduce speed ...
- SCSI cabling: check contacts, termination, etc.
4. You might want to try the ncr53c8xx driver 2.6 which was posted here
some time ago. It's supposed to have better error handling ...
(I didn't check, so be warned.)
Kurt Garloff, Dortmund
<K.Garloff@ping.de>
