Re: Filesystem strangeness (ext3)

From: Jan Kara
Date: Mon Apr 05 2004 - 11:33:34 EST


Hi,

> yesterday I discovered a bunch of corrupted files. When I try to stat()
> it, I get "Value too large for defined data type" or "Input/output
> error" when it is a symlink to those files. These were normal files
> (some readmes etc.) not accessed by any daemons or any applications at
> all. Midnight marks them with "?" at beginning and shows something
> about 3GB size. When I try to unlink() them, I get "Operation not
> permitted" even though I'm root. Kernel log shows these two kinds of
> errors:
>
> Apr 5 11:33:47 master kernel: init_special_inode: bogus imode (113315)
>
> and
>
> Apr 5 11:33:49 master kernel: attempt to access beyond end of device
> Apr 5 11:33:49 master kernel: 09:01: rw=0, want=587173332, limit=38539776
It seems like something trashed part of your disk...

> When I try to e2fsck -nf (I tried only ro test on mounted partition) I
> get some of these:
>
> Illegal block #10 (2236192626) in inode 523442. IGNORED.
I suggest at least remounting the partition ro and then run the ro-test.

> and:
>
> Error while iterating over blocks in inode 523442: Illegal triply indirect block found
> Segmentation fault
Bug in e2fsck - you probably want to download latest version of e2fsck. If
there is still this bug report to e2fsck authors.

> It confuses me a little bit and I don't know wheter I should try to do
> rw e2fsck on unmounted partition, especially when e2fsck segfaults. My
You should definitely backup the partition before you run fsck on it.

> system:
>
> Linux master 2.4.22 #1 Sat Sep 20 14:26:11 CEST 2003 i686 unknown
> e2fsck 1.27 (8-Mar-2002)
> Using EXT2FS Library version 1.27, 8-Mar-2002
> The device is /dev/md1:
>
> [dev 9, 1] /dev/md1 889E3491.06B38BDD.B6F3803F.BCE40C0E online
> [dev 3, 7] /dev/hda7 889E3491.06B38BDD.B6F3803F.BCE40C0E good
> [dev 3, 71] /dev/hdb7 889E3491.06B38BDD.B6F3803F.BCE40C0E good
>
> smartctl shows no errors on those physical disks.

Honza
--
Jan Kara <jack@xxxxxxx>
SuSE CR Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/