samba - Apr 2002 - data corruption , losetup and smbfs

Hello ,

 I am experiencing  data corruption when I use loopback mounted
filesystems over smbfs . I mount a share which is in a win2000pro computer
from my slackware 8 , 2.4.18-ac3 and samba 2.2.3a with lfs support box
and I create a large file on it , then I create a fileystem on the file
and mount it with losetup . During operations such as copying files etc I
sometimes get messages like :

EXT2-fs error (device lvm(58,0)): read_block_bitmap: Cannot read block
bitmap - block_group = 1117, block_bitmap = 36601856

( this is from an lvm group consisted of loopback devices , but the same
error occurs in standalone loopback mounted filesystems ) .
  In the archives I saw someone reporting almost the same problem ,
apart from the fact my filesystems are still mountable after those errors.
Running an fsck on them sometimes finds few errors , sometimes no errors
and sometimes ( rarely ) lots of errors . Data is usually slightly
corrupted ,if it is at all .
   I understand this kind of setup is a bit weird , but nevertheless
usefull in my enviroment . I can experiment a lot if someone wants to take
a look at it .


 Kind regards ,

--
============================================================================
Dimitris Zilaskos

Department of Physics @ Aristotle Univercity of Thessaloniki , Greece
=============================================================================

On Wed, 17 Apr 2002, Dimitris Zilaskos wrote:
> 
>   Hello ,
> 
>  I am experiencing  data corruption when I use loopback mounted
> filesystems over smbfs . I mount a share which is in a win2000pro computer
Yes, loopback and smbfs is currently a bad idea.

There is something with how loopback is calling smbfs that is different
from normal use. It is very easy to trigger the corruption over loopback.

There are some references to nfs in the loopback code and how it makes
things more complicated. It could be locking related. Also, loopback was
broken in early 2.4.x and late 2.3.x, perhaps the fix requires some change
to smbfs.

Other than that I don't know and I don't plan to try and find you until
I
have finished some other things.

One place to start could be to compare the write operations in
drivers/block/loop.c with generic_file_write in mm/filemap.c.

>   In the archives I saw someone reporting almost the same problem ,
> apart from the fact my filesystems are still mountable after those errors.
That's probably just luck.

/Urban