NFS troubles on recent -current.

[Bernd Walter]

On Sun, May 11, 2003 at 04:29:34PM -0700, Don Lewis wrote:
> On 11 May, Bernd Walter wrote:
> > I have a -current (10th may) NFS server which serves for a diskless
> > NFS client (11th May).
> > The client has / on a CF media and /usr as well as /var/spool on the
> > server.
> > Also swap is on the server (/usr/swapfile) via md.
> > When I run installworld on the client the server hangs with
> > "lockmgr draning against myself".
> > I got no DDB prompt so no stacktrace or crashdump :(
> 
> Try the attached patch.  I didn't have any problems with
> 	make -j10 buildworld
> 	make DESTDIR=/mnt installworld
> with /usr/obj and /mnt NFS mounted from the filesystems exported on the
> same server (though I had to tweak some sysctl variables to avoid wdrain
> deadlocks in the bio code in this configuration).
> 
> Can you break into the debugger with CNTRL-Alt-Esc?  I also recommend

I havn't tried as it shouldn't be required after a panic.

> trying the DEBUG_VFS_LOCKS kernel configuration option, which will drop
> into ddb if any of the vnode locking assertions are violated.

I will try your patch and suggestions tomorrow.

-- 
B.Walter                   BWCT                http://www.bwct.de
ticso at bwct.de                                  info at bwct.de