Rob Healey writes:

>        The problem with processes being stuck in D state, probably due
>        to NFS directory reading, appeared sometime between March 26th and
>        April 7th sup tree as my March 26th kernel doesn't do the evil
>        deed but my April 7th kernel does it in spades... B^(.


This is strange, since nothing much changed in the NFS code in that period
(except for a couple of #ifdef FIFO lines). Is there something on your
system that does locking on files on NFS-mounted filesystems?

Eric Volpe writes:

> We have also been having the problem that if a -current machine that's
> serving NFS goes down and then comes back up, many if not all of the
> machines which were mounting it's filesystems will no longer be able to
> access those filesystems, but will just hang when trying to do so. This
> happens with downtimes as short as a couple of minutes.I htought NFS was
> supposed to be stateless? 

I have seen this behaviour before, and on other systems than NetBSD.
Still, this doesn't mean we should not try to fix it :-) Could you
give some more information on the following points:

	- Of what date is your -current kernel on the server?
	- What system are the client running?
	- Are you specifying any special parameters for the mounts
	  on the clients (TCP, perhaps?)
	- Does it happen even when a client doesn't try to access the
	  server while it's down?

Thanks,

- Frank