On Sun, Jun 23, 2002 at 09:54:57PM -0700, David S. wrote:

> If you don't want to reboot, wait for the RPC calls to time out
> (something like half-an-hour?).

the server came back two days ago, and the client still has processes hung
in nfsrcv and nfscn2.

> To avoid this, make sure you mount remount file systems with the
> interruptible flag (man mount_nfs), so the RPC call can be interrupted
> with a ctrl-c.  Even better, use the automounter to control your NFS
> mounts, so that there's less chance of the client having the remote
> file system mounted when the server goes down.

I am using both.  the tasks stuck in nfsrcv are interruptable, but I
can't shake the one stuck in nfscn2...

> Well, NFS can be great convenience.  But that convenience doesn't come
> for free.

as far as I can tell, I've done nothing incorrectly, yet things are not
behaving as they should.  :P

I'm perusing the PR database to see if anybody else has run into this
problem, and how to fix it.  I've never touched kernel sources before,
but compared to other third-party code I have had the privilidge(?) of
working with, the prospect doesn't scare me.  ;)

is there a way to diagnose exactly where this is happening on a running
kernel without disturbing other processes?  will gdb /netbsd /dev/kmem
blow anything up if I'm careful?

I'd really like to fix this if I could, or at least know exactly where
things in the kernel are going awry.

-- 
  Aaron J. Grier | "Not your ordinary poofy goof." | agrier@poofygoof.com