NFS-Locking problem with 6.4/7.1-RELEASE

[Matthias Schuendehuette]

Hello,

I operate two FreeBSD-Servers in a Windows- and HP-UX Environment. One  
is a SAMBA-Server as a gateway between the Windows and the Unix world,  
the other is NFS-Server for the HP-UX 11i v1 Workstations. Both are HP  
ProLiants DL380 with additional external disks on SmartRAID Controllers.

Since the HP-UX Workstations and their disks are becoming quite old, I  
started to move the home-directories to the FreeBSD Server, wich  
worked with 6.3-RELEASE quite good so far.


Brave as I am, I updated the servers to 6.4 RELEASE and since then the  
users on the HP-UX machines with the homedirs on the FreeBSD-Server  
were locked... :-(

I tried to find out what was happening and this are my results:

When a user logs in on a HP-UX machine, his '.profile' file is opened  
and read/executed, but it seems, that it cannot be closed any more. So  
if the last line in the '.profile' is "echo foo bar" you *can* see  
"foo bar" on the screen, but then nothing happens any more, the  
machine is locked.

I recorded such a session with 'tcpdump' and looked at the dump... the  
only noticeable things are *Bursts* of NLM V4 CANCEL_MSGes on the same  
filehandle. Eg:

"V4 CANCEL_MSG Call FH:0x644201fe svid:xxxx pos:0-0"

This line is repeated 7 times with various values for 'svid'.

I'm no NFS specialist at all, so I cannot tell you more :-/ But I can  
supply the dump (if needed),
it's 92KB, so the size should not be a problem...

BTW: I tried this with and without kernel support for NFS-Locking - no  
difference. I also tried the new replacement server with FreeBSD 7.1- 
RELEASE: Just the same problems, with and without kernel support.

I hope someone is willing to work on that issue...

As mentioned, a new, non-productive server is available in the moment,  
so tests are easily possible.

TIA

Matthew

-- 
Ciao/BSD - Matthias

Matthias Schuendehuette    <msch [at] snafu.de>, Berlin (Germany)