Re: [Jfs-discussion] fsck: attempt to access beyond end of device

[Lenz G. <gr...@su...>]

Hi,

On Mon, 29 Oct 2001, Anthony Liu wrote:

> This is a very strange problem, anyone seen something like this in your
> logs?

Yes...

> Doing mkfs.jfs or fsck the following will show up (11 times)
>
> 	kernel: attempt to access beyond end of device
> 	kernel: 03:01: rw=0, want=1052228, limit=1052226
>
> fdisk -l output:
>
> Disk /dev/had: 255 heads, 63 sectors, 5005 cylinders
> Units = cylinders of 16065 * 512 bytes
>
> 	Device Boot    Start       End    Blocks   Id  System
> 	/dev/hda1   *         1       131   1052226   83  Linux
> 	/dev/hda2   *       132       135     32130   83  Linux
> 	/dev/hda3           136       184    393592+  82  Linux swap
> 	/dev/hda4           185      5005  38724682+   5  Extended
> 	/dev/hda5           185      3985  30531532   83  Linux
> 	/dev/hda6          3986      4507   4192964+  83  Linux
> 	/dev/hda7          4508      5005   4000184+  83  Linux
>
> Note:
> /dev/hda6 is jfs formatted. jfsprogs 1.0.8
> Kernels tested are 2.4.13 and 2.4.13-pre3
> This problem also show up on another system/drive.

This seems to be due to the fixed 4K block size of JFS. 1052226 is
not dividable by 4 while 1052228 is...

> When reboot to the same kernel, the problem disappeared...
>
> mkfs.jfs /dev/hda1 gives:
> --cut-------------------------------------------------------
> mkfs.jfs version 1.0.8, 17-Oct-2001
>
> Warning!  All data on device /dev/hda1 will be lost!
>
> Continue? (Y/N) y
>    -
>
> Format completed successfully.
>
> 1052226 kilobytes total disk space.
> --cut------------------------------------------------------
>
>
> However, fsck report that only 1052224k is the total:

Yes, this is a closest multiple of 4K blocks that fits on this partition.

[SNIP]

> After mounting /dev/hda1, copying some files to it, umount it and
> perform a fsck.jfs, the same "attempt to access beyond end of device"
> error will show up. The pair of messages repeated exactly 11 times when
> it was triggered.

Looks like a bug to me - seems like fsck.jfs does not properly detect the
end of the file system here?

Bye,
	LenZ
-- 
------------------------------------------------------------------
 Lenz Grimmer                                           SuSE GmbH
 mailto:gr...@su...                       Schanzaeckerstr. 10
 http://www.suse.de/~grimmer/            90443 Nuernberg, Germany
      Users: To keep them dry don't water them after dinner.