I've just gone and replaced my 100Mb/s switch with 10Mb/s hub (because the
best that can be said about the fans in the Catalyst 2900 is that they are
quieter than the fans in an AGS+) and then my network just stopped. CVS via
ssh wouldn't work any more (it would wedge while building the directory tree
in /tmp on the server) and X performance was simply horrible. It looked as
if packets weren't coming in to the local machine unless it was also sending
packets out. I had to keep hitting enter in the local Kconsole window in
order for the next few screens of output to come through. 

I've run with this configuration in the past, but with NetBSD current from
December 2001/January 2002 on my U10. The other machines on my network are a
couple of middling to good PCs with ne/de/ep 10 Mb/s cards in them. I also
have a work laptop with an fxp in it, but that's at work with me at the
moment, so I can't test it just now. 

As soon as I replaced the switch I started having network problems. I
ifconfiged everything and even moved one of the PCs off its rtk interface on
to an ne. No luck, so I started installing RC2 kernels on the U10, the
development PC and the laptop. A few reboots later, and the laptop worked
perfectly. I was able to get 10Mb/s sustained transfer rates with scp when
copying from the development PC (ne) to the laptop (fxp). The U10 was still
dead in the water. At this stage all machine had been rebooted at least once
since the hub had been installed and I have checked the interface settings
on each machine.

While ftping the new kernel sources to the U10 I noticed that the transfer
rate was 50KB/s, much too low, so I ran 'netstat -in -I ne3 5' on the
development PC and noticed that I was getting a 50-60% collision rate. After
compiling and installing the RC2 kernel on the U10 I still had no change.
ping -f works fine when pinging the U10 from the PC, but ping -f -s 1000
reports 60% packet loss. ping -f -s 1000 from one PC to another loses no
packets. 

ifconfig hme0 prints out

	hme0:
flags=8a63<UP,BROADCAST,NOTRAILERS,RUNNING,ALLMULTI,SIMPLEX,MULTICAST> mtu
1500
	        address: 08:00:20:b0:7a:28
	        media: Ethernet 10baseT
	        status: active
	        inet 10.0.1.22 netmask 0xffffff00 broadcast 10.0.1.255
	        inet6 fe80::a00:20ff:feb0:7a28%hme0 prefixlen 64 scopeid 0x1

Is it possible that the hme is still trying to run at 100Mb/s even though it
shouldn't? I've run out of ideas that can also explain why long packets are
being lost, so any help would be appreciated.

Cheers,
Lloyd