[linux-cifs-client] Cifs, loop, large files and umount results in CIFS VFS: Nowritable handles for inode - on both Ubuntu and Debian

[Thomas Kjaergaard]

I forgot to mention that I have tried with the mount options forcedirectio 
and direction.

Those options does impact the performance of the mount, but does not avoid 
the problem.

All tests is on a LAN and I don't get the errors if I don't tunnel it 
through the loop device.

I suspect the "losetup -d /dev/loop0" don't end the loop device's connection 
to the driveimage file correctly.
It's the same if I use the loop integrated mount command:

"mount -o loop,encryption=aes /mnt/linkstation/driveimage /mnt/backup -p0 < 
/root/scripts/loop.cfg"

and then unmount with "umount /mnt/backup/"


Eventually, should I look for a totally different backup storage strategy?


Med venlig hilsen / Best Regards
Thomas Kjærgaard


----- Original Message ----- 
From: "Thomas Kjaergaard" <cifs at pci.dk>
To: <linux-cifs-client at lists.samba.org>
Sent: Sunday, August 03, 2008 3:45 AM
Subject: [linux-cifs-client] Cifs, loop,large files and umount results in 
CIFS VFS: Nowritable handles for inode - on both Ubuntu and Debian


> Hi
>
> //mounting
> mount -t cifs -o credentials=/root/scripts/linkstation.cfg 
> //192.168.100.9/backup /mnt/linkstation
>
> //setting up loop device
> losetup /dev/loop0 /mnt/linkstation/driveimage
>
> //mount the loop device in /mnt/backup
> mount /dev/loop0 /mnt/backup
>
> I can now write files to the mounted filesystem.
>
> Then I unmount with:
>
> //dismount
> umount /mnt/backup
>
> //remove loop
> losetup -d /dev/loop0
>
> //dismount
> umount /mnt/linkstation
>
> The problems:
> -------
> After the umount in dmesg and syslog I get:
> CIFS VFS: No writable handles for inode
> The amount of the errors depends on how much I have been writing to the 
> loop device.
> A simple mount and unmount gives 2 lines with the errors, if I have copied 
> 500MB to the device it's multiple pages with the error.
>
> Remark, the "No writable handles for inode" errors does first appear in 
> syslog after umount.
>
> Small files in the root and empty folders are gone next time I mount the 
> loop device.
> -------
>
>
> The file driveimage file is created before first use after cifs mount 
> with:
> dd if=/dev/zero of=/mnt/linkstation/driveimage bs=1M count=1 seek=5000
> losetup /dev/loop0 /mnt/linkstation/driveimage
> mkfs.ext3 /dev/loop0
>
> I have tried with 5GB and the target is 350GB and eventually 1TB when we 
> get a bigger NAS.
>
>
> I have tried from two different distros as client.
>
> Debian running 2.6.25-2-686 and CIFS Version 1.52
> Ubuntu running 2.6.24-19-server and CIFS Version 1.52
>
> Using two different servers:
>
> One being a Bufalo Linkstation, running Samba 3.0.21c according to:
> "cat /proc/fs/cifs/DebugData" when the linkstation is mounted.
>
> The other is the above mentioned Debian running Samba 3.0.30.
>
>
> The goal is to use a NAS as storage for an encrypted filesystem which is 
> mounted when the backups needs to be updated.
> mount -t cifs -o credentials=/root/scripts/linkstation.cfg 
> //192.168.100.9/backup /mnt/linkstation
> mount -o loop,encryption=aes /mnt/linkstation/driveimage /mnt/backup -p0 < 
> /root/scripts/loop.cfg
>
> The encryption does not seem to influence if the errors are there or not.
> Is it the cifs or is it the loop device which is causing the problems?
>
>
> Best Regards
> Thomas Kjaergaard
>
>
> _______________________________________________
> linux-cifs-client mailing list
> linux-cifs-client at lists.samba.org
> https://lists.samba.org/mailman/listinfo/linux-cifs-client