I have resolved my problem.
I have seen that without SSH tunnels my client was using NFS version 3 and
with SSH tunnels NFS version 2.
I have forced NFS version 3 in fstab:

localhost:/UPSE /UPSE nfs
nfsvers=3,tcp,rw,hard,intr,bg,port=7777,mountport= 8888 0 0

and now I have no problem anymore. I still get the JUKEBOX error but the NFS
READ operation is repeated after 5 sec and the client doesn't hang anymore.


"Cédric Pillonel" wrote in message
news:1144843200.776611@ftpgate...
> My server is running RedHat 4.0 ES. I export the same share with Samba and
> with NFS over SSH tunneling.
> As clients I have a PC with Windows XP accessing to the share with Samba
and
> a Redhat 9.0 box accessing the share with NFS over SSH tunnels.
>
> My problem is that a file read operation hangs on my linux client if the
> file has been modified and saved by the Windows client a few seconds
before!
>
> Here is my entry in fstab of the client:
> localhost:/UPSE /UPSE nfs tcp,rw,hard,intr,bg,port=7777,mountport=8888 0
0
>
> Here is my command to setup the tunnels:
> /usr/bin/ssh -f -c blowfish -L 7777:localhost:2049 -L 8888:localhost:4002
> 193.5.229.151 /bin/sleep 60
>
> I have no problem when accessing the share directly with NFS, whithout SSH
> tunnels. I have just noticed that the file read operation in this case
takes
> more time to complete. I found out that the NFS server sends back the
error
> ERR_JUKEBOX for the first NFS READ operation. Then the client repeats the
> READ operation 5 sec. later.
>
> Any file locking problem?
> Any timeout problem?
>
> Thank you for any help.
>
>