Crashes with corruption of disk data

Crashes with corruption of disk data

[Hendrik Visage]

Hi there,

 I've came upon a crash situation (quite repeatable) with reiserfs,
but NOT with ext2/3.

Using Vanilla 2.4.17 + XFS + EVMS-1.1.0-rc4

(1) using evmsn I created a logical volume, mkfs'd it, and
mounted it. FS used was ext3 & reiserfs

(2) went into evmsn again & destroyed[b] the volume with a
 unmkfs[a]

(3) forced exited evmsn 

(4) tried to access the "still" mounted fs & umount it

At this point, ext2/3 would let go of the mounted FS,
however, reiserfs consistently crashed into the SGI/XFS
Kernel debugger :(

I'm of believe that this would be seen as a non check of the
validity of the underlying diskdata. I believe this test could
be repeated with a dd if=/dev/zero of=/dev/mounted/reiser but haven't
yet done so (sleep time), but would be able to report or help with
finding the errors if it is seen as a bug, as I haven't seen documentation
to the contrary.

Greetz
Hendrik

Notes
[a] because a miss assumption which are being fixed, this was
 wrongly executed, but the "issue" still hold :(
[b] it weren't actually destroyed, though the unmks were run,
 as the kernel wouldn't let go of a mounted volume, thus the
forced in (3)

Re: Crashes with corruption of disk data

[Oleg Drokin]

Hello!

   This is a know issue.
   If you muck with FS data of mounted reiserfs filesystem, not using
   reiserfs in-kernel driver, you are asking for trouble.

Bye,
    Oleg   
On Sun, Jul 21, 2002 at 02:10:30AM +0200, Hendrik Visage wrote:
> 
> Hi there,
> 
>  I've came upon a crash situation (quite repeatable) with reiserfs,
> but NOT with ext2/3.
> 
> Using Vanilla 2.4.17 + XFS + EVMS-1.1.0-rc4
> 
> (1) using evmsn I created a logical volume, mkfs'd it, and
> mounted it. FS used was ext3 & reiserfs
> 
> (2) went into evmsn again & destroyed[b] the volume with a
>  unmkfs[a]
> 
> (3) forced exited evmsn 
> 
> (4) tried to access the "still" mounted fs & umount it
> 
> At this point, ext2/3 would let go of the mounted FS,
> however, reiserfs consistently crashed into the SGI/XFS
> Kernel debugger :(
> 
> I'm of believe that this would be seen as a non check of the
> validity of the underlying diskdata. I believe this test could
> be repeated with a dd if=/dev/zero of=/dev/mounted/reiser but haven't
> yet done so (sleep time), but would be able to report or help with
> finding the errors if it is seen as a bug, as I haven't seen documentation
> to the contrary.
> 
> Greetz
> Hendrik
> 
> Notes
> [a] because a miss assumption which are being fixed, this was
>  wrongly executed, but the "issue" still hold :(
> [b] it weren't actually destroyed, though the unmks were run,
>  as the kernel wouldn't let go of a mounted volume, thus the
> forced in (3)