Hi,

I haven't been following this thread very closely, until I noticed a hang. ;-)

I just installed 1.4.1 on a PIII600 with a vr ethernet card.
I am using a lot of NFS tranfers to copy files from a 1.4
system (PII333, same ethernet card).

First try: mount 1.4 filesystem on 1.4.1 box, start big nfs transfer
with some large (>100MB) files. Tar (on 1.4.1 system) from 1.4 to
1.4.1 system failed twice. Just stopped or slowed down a lot. Pings
went from negligible to >10000 and very sporatic.

The first time I got a couple of missing interrupt messages from
the vr driver.

This happened twice.

Then I tried mounting the 1.4.1 file system on the 1.4 box and did
the 1.4 to 1.4.1 transfer there. This seems to work (it is still
transfering) but it *seems* slower.

I'll try to play with back and forth file transfers when the current
transfer finishes.

-Rich

On Tue, 07 Mar 2000, Greg Wohletz wrote:
>>To: "386 Port NetBSD" <port-i386@netbsd.org>
>>Sent: Sunday, March 05, 2000 7:47 PM
>>Subject: NFS server hangs under 1.4.1
>>
>>
>>> Hi Folks,
>>> This has been driving me nuts. 
>>> 
>>> I have a machine running NetBSD 1.4.1 which serves home and source
>>> directories
>>> using NFS3 over UDP.
>>> 
>>> I have clients, another NetBSD machine running the same release of
>>> 1.4.1 and
>>> an SGI indigo**2 running IRIX 6.5.
>>> 
>>> About once a week, sometimes under heavy usage, sometimes not, my server
>>> will hang during
>>> NFS server activity: it only happens during NFS service.  Either client
>>> can precipitate
>>> the server hang.  But I cannot reproduce this at will with heavy NFS
>>> traffic.  These are
>>> home machines in a mostly quiet network (just me and my wife).
>
>Any time I try to to significant work with irix 6.5 using v3/tcp nfs
>mounts iand a netbsd 1.4.1 server system it ends up crashing the netbsd
>machine with:
>
>	panic: Bad nfs svc reply
>
>This does not occur if I instruct the sgi to mount the sgi using v2/udp
>
>I only encountered this problem last week when I compiled a new version
>of amd on the sgi's that started requesting v3/tcp mounts.  A brief
>examination of the core file shows that the size of the request exceeded
>the maximum.  I have not investigating the problem beyond this.
>
>						--Greg
--
Richard Pennington		Introl Corporation, Milwaukee, WI USA
Email: rich@introl.com		Phone: +1 414-273-6100	Fax: +1 414-273-6106
				US and Canada: 1 800-327-7171
Cross development tools for the 68HC05, 68HC08, 6809, 68HC11, 68HC12, 68HC16,
and 68XXX:			http://www.introl.com	ftp://ftp.introl.com