I just started doing some testing on 2.4.23 with a dual-port
Intel NIC.  I have seen several crashes and reproducible 'TX watchdog timeout'
messages when connecting an e1000 port to both an e100 on the same machine
and to a tulip on another machine...  I did not see problems when passing
traffic between the two e1000 ports, so the tx-timeout issue seems confined
to e1000 -> 100bt.

I'm running on a dual-amd box which has had issues in the past, but
a bios update seems to have fixed it's pci-routing issues.  Also, I
am running some personal networking patches.

That said, I have not seen any of these issues on 2.4.23-pre9 (+ my patches),
and I saw the same problems on 2.4.23 both with my patched e1000 and with an
unpatched e1000.

Attached is an oops that did not hang the machine.  The hard hang looked
like memory corruption (a few of the registers were 0x15141312,
and the payload of the packets I was sending was incrementing bytes, ie 0x010203040506...).
I'll crank up the serial console and see if I can capture some of
the hard crashes...

The lspci and dmesg files are from a 2.4.23-pre9 boot.

As usual, it could all be my fault, so I'd like to hear if anyone can confirm
or deny the e1000 -> 100bt issue.

Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com