[vortex] Strange errors with 3c900

[Matthew Gabeler-Lee]

I have a 3c900 Combo card, and was using the 3c59x driver for quite a
while with zero problems.  Then I had the card in storage for several
months because I didn't need it.  When I put the card back in the
system, I got the following errors when the driver was loaded, and the
card only half worked (more detail later).

3c59x.c:LK1.1.8  13 Aug 2000  Donald Becker and
others. http://www.scyld.com/network/vortex.html $Revision: 1.102.2.25 $
See Documentation/networking/vortex.txt
eth0: 3Com PCI 3c900 Boomerang 10Mbps Combo at 0xdc00,  ***INVALID
CHECKSUM 0010***  00:60:08:41:20:71, IRQ 9
Full duplex capable
  8K word-wide RAM 3:5 Rx:Tx split, autoselect/10baseT interface.
  Enabling bus-master transmits and whole-frame receives.

I used the vortex-diag program to re-write the eeprom, and it now does
not give the invalid checksum message on load.

The initial half working was that I could reach some hosts on the
network, but not others.  I didn't find any correlation between the ones
I could reach and the ones I couldn't.  With the re-written eeprom, I
can reach all hosts on the network, but data connections that are
sending many packets in a stream seem to die.  To be clearer, small web
pages, shells, irc all work.  Connections that try and download files
more than a couple kilobytes seem to drop, die, etc.  This seems to be
related to the enormous number of these messages in the logs (they
appear every few seconds):

 Rx error: status 8c.
eth0: vortex_error(), status=0xe481
 Rx error: status 08.
eth0: vortex_error(), status=0xe081
eth0: vortex_error(), status=0xe481

Most of the messages I see are the 8c Rx error.  Is my card
broken/hosed, or can it be fixed through software?

For reference, I'm using kernel 2.4.0-test8.

-- 
	-Matt

No matter where I go, the place is always called "here".