Red Hat Bugzilla – Bug 44561
etherexpress pro/10+ fails under load
Last modified: 2007-04-18 12:33:40 EDT
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 5.0; Windows 98; DigExt)
Description of problem:
This NIC has been my nemesis since 5.2! I've seen reference to using
eepro.o and eepro100.o and have tried both. Module loads, pings are OK.
SSH'ing in for a terminal session will work apparently forever. FTP, pscp
(ssh file transfer) will cause the NIC to go deaf after a very short data
flow (100kB or so). Does not respond to pings. Will ping other machines
(oddly, NICs of the same model have LONG ping times, others are normal).
Doing a 'network restart' will bring the NIC back. The maintainer, Aris
Filho, admitted eepro.o was broken around 2.2.16 but apparently is still
problematic. I've worked with several of these cards in Linux Router
Project boxes and found that the 2.2.17 mandrake module was the only one
that would work. Not with 2.4.2 though!
Steps to Reproduce:
Actual Results: NIC stops responding to the outside world. No error
messages found on local machine. Appears functional from local machine.
Expected Results: File transfers should have succeeded!
As in description, maintainer of module has admitted to breaking it around
2.2.16, but claims to have repaired.
Not xinetd-related... reassigning component.
Could you do 2 things for me?
1) try the "e100" driver instead of the "eepro100" (change it in
2) give me the "lspci" output for the network-card
With regard to the e100.o module, it seems to load, however the failure is the
same. Note, this was tested from an SSH session, not sure if all information
was conveyed. As for lspci, this card is ISA and the computer is pre-PCI! A
direct response from maintainer suggests a new improved module for 2.2 and 2.4
My mistake with regard to the e100 and eepro100 modules... they don't load.
Linuxconf appeared to insert them, but a reboot at the machine showed they
didn't load (not surprising!). Only eepro.o loads successfully, but dies as
I see this too. Stopping networking, removing the module and then
reinserting fixes it, but obviously having to do this every few minutes is
very very annoying. Last version that worked for me was the original
module which came with Redhat 6.2 - when the 6.2 kernel was updated,
things were broken.
I can confirm that this happens in my setup too. Doing a network restart or a
full reboot will fix it, but it then dies again about 5 - 10 mins later.
I haven't seen this to be triggered by any particular network operation or
protocol (unlike some posts above). The interface using this driver will die
after a while even without use.
Running kernel 2.4.2-12 on a 486 with only ISA support. 3c509 in the machine