[Unionfs] [ANNOUNCE] Unionfs-ODF 2.1.10 release (NFS problems)

[Erez Zadok]

Jesse, sorry for not replying sooner.  Been busy releasing 2.2 (non-odf).
Can you do do me a favor and report this in
https://bugzilla.filesystems.org/?  I'd like to track this bug officially.

Thanks,
Erez.

In message <476C1C6E.3050706 at ccs.nrl.navy.mil>, Jesse I Pollard writes:
> Thanks for the patch.
> 
> Apologies for the really late followup...
> 
> In reference to linux-2.6.24-rc3 + unionfs-2.1.10-odf_for_2.6.24-rc3:
> 
> 1. Mounts locally:
> 
>         mount -t unionfs \
>               -odirs=/mnt/a=rw/mnt/torpedo-old=ro,odf=/odf \
>                         none /mnt/torpedo
> 
> 2. Local operations appear normal.
> 
> 3. NFS exporting fails:
> 
>         mount lint:/mnt/torpedo /mnt/torpedo
>         mount: lint:/mnt/torpedo failed, reason given by server: Permission
>                         denied
> 
> 4. NFS in general works when the export is not part of a unionfs mount point
>    and using the same client:
> 
>         mount lint:/mnt/a /mnt/u
>         df
>         Filesystem           1K-blocks      Used Available Use% Mounted on
>         /dev/mapper/VolGroup00-LogVol00
>                              234476168  11406680 210966628   6% /
>         /dev/hda1               101086     27863     68004  30% /boot
>         tmpfs                   481748         0    481748   0% /dev/shm
>         lint:/mnt/a           15227008   7793152   6647936  54% /mnt/u
> 
> 5. export list on server:
> 
>         /mnt/torpedo    tabby(rw,fsid=10,crossmnt,no_root_squash,nohide)
>         /mnt/a          tabby(rw)
> 
>         I have also tried without the fsid=10 and crossmnt without success
> 
> 6. NFS investigations has shown that the server believes the mount
>    succeded (IP number entries have been removed):
> 
>         /usr/sbin/showmount -a
>         tabby.nrl.navy.mil:/mnt/a
>         tabby.nrl.navy.mil:/mnt/torpedo
> 
> The last time I traced through the logic, the problem appeared to be 
> because
> unionfs did not locate the root directory of the volume being mounted. 
> Mountd
> checks all passed, but the first filehandle read seems to be failing. 
> (Perhaps
> on either fsstat/fsinfo NFS operations). I do know the "Permission deinied"
> message is used for any error that occurs during the mount that are not
> otherwise labeled.
> 
> I have not traced this activity to see exactly where the failure arises yet
> (I think it will call for some printks in the export.c functions...), but I
> will be looking into it later next week and into january.