Bug 12966

From time to time the vr0 driver (the network driver for the Via Rhine 
chipset) experiences receiver lockups. Outgoing packets go out 
correctly and can be monitored at the destination address. The "received packets" counter in 'netstat -i' continues to 
increment. Pinging the network's broadcast address produces the 
appropriate number of increases to the recieved packets counter.
However, 'tcpdump' does not see any packets on the interface,
and any attempts to connect outside the local machine fail.

The computer involved is dual-booted with Linux, and has functioned
properly under Linux for the past two months. No changes were made in
the hardware and/or network configuration when FreeBSD was installed as
a dual-boot. The network card is a Hawking 10/100, the chip reports

vr0: <VIA VT3043 Rhine I 10/100BaseTX> rev 0x06 int a irq 11 on pci0.19.0
vr0: Ethernet address: 00:40:33:5b:24:ca
vr0: autoneg complete, link status good (half-duplex, 100Mbps)        

upon bootup. The network cable is plugged into a LinkSys StackPro 100
8-port workgroup hub sitting on my desk (due to my laptop and various
other test computers), which is then plugged into the
wall socket, which is hooked to a 100bt switch back in the server room.
Unplugging the network cable from the hub, waiting a few seconds, then
plugging the network cable back into (possibly different port on) the hub
does not unstick the receiver. 

Checking out the latest vr0 driver from -current shows no changes to the
Via Rhine driver except for some minor name changes needed due to
changes elsewhere in the kernel.

How-To-Repeat: Attempt to download a file from a fast server on the local Intranet 
at full 100-BT speeds via FTP. Note that we have an entirely switched 
network fabric (except for the hubs on some of our desks), meaning that 
we can see significant percentages of the theoretically available 
100-mhz bandwidth. This may make this a more difficult environment than
what most people are running the Via Rhine in.