CIFS corruption (with some errors in dmesg)

Hi,

We use CIFS for copying files between systems and occasionally see some
corruption, where substantial portions of copied files are just zeroes.

It is reproducible during repeated copying of large files but takes a
while to appear. This is using kernel 2.6.18 on the client and
samba-3.0.22 on the server.

The failure is hidden as cp and rsync do not notice the corruption
during transfer.

When the corruption does occur, messages appear in the kernel logs on
the client:

        CIFS VFS: close with pending writes
        CIFS VFS: No writable handles for inode

Sometimes it is just one of those messages, other times it is both,
other times it shows the first and then repeatedly floods the 2nd one.

Any ideas? I'm happy to run further tests with different
kernels/configurations/patches.

Thanks!
-- 
Daniel Drake
Brontes Technologies, A 3M Company