On Wed, 16 Mar 1994, Herb Peyerl wrote:

> : Even if it is working for some time (strangely enough it is working much
> : longer if you're using the ed driver than with the ep driver) after some
> 
> hmm... strangely enough; I see the opposite behaviour. I get better results
> with ep than with ed.  For me; if_ed doesn't do NFS at all and hasn't for
> a couple of weeks. 

A couple of weeks ago I put ep0 into my kernel (and installed a 3c509, of 
course).  I had been using if_ed with NFS for quite some time, but after 
about 12 hours of operation, if_ep croaked.  I went back to if_ed and 
things ran fine for a week and a half.

I got curious again a few days ago, and I installed the new config, did a 
config KERN, then in the compile directory a "make clean && make depend 
&& make" to build a new kernel.  Moved the network wire, set up 
hostname.ep0 again, and booted the machine with the new kernel, and have 
had no problems since.  The score is currently if_ep 1, if_ed 1 (I have 
two machines running side by side, one with each board), and both running 
with moderate to heavy NFS traffic.

I should mention that the NetBSD-current machines I am using get 
rebooted about once a week when I do a kernel upgrade.  They haven't had 
to go longer, yet...

> I'm only getting it against a SunOS 4.1.x machine. (given that my other 
> machine is also a NetBSD machine). 

The NFS server here is SunOS 4.1.x, likewise.  No NFS problems that I am 
aware of for several weeks, perhaps more.

--scott

------------------------------------------------------------------------------