Red Hat Bugzilla – Bug 62314
(NET EPIC100) epic100 network driver fails
Last modified: 2013-07-02 22:06:02 EDT
Description of Problem:
Original skipjack-beta1 kernel worked fine. After updating to the 2.4.18-0.8
kernel, my Ositech Jack of Spades Cardbus ethernet card (epic100 module) did not
autonegotiate a link. Using mii-tool to force 100 Mbit full duplex, a link was
established, but stressing the network connection would eventually cause it to
Version-Release number of selected component (if applicable):
Network connection froze the 4 times I tried it, after transferring data
for 30-90 seconds.
Steps to Reproduce:
1. No link at boot, until I do 'mii-tool -F 100baseTx-FD'
2. start a large network transfer (I used scp to copy Redhat/RPMS/*)
3. files get copied for about 30 seconds, then the network stops
Can you try 2.4.18-0.13 or later ? should have an older epic100 driver
I did a complete re-install with Skipjack beta2 (kernel 2.4.18-0.13), and the
network hanging problem is no longer happening. However, I still have to use
mii-tool to force 100baseTx-FD mode in order to get a network link. All previous
Red Hat 7.x releases have automatically detected and configured the card.
After updating to kernel package 2.4.18-0.20, the network freeze problem has
returned. I still had to force the link mode as well.
The problem is still present with the 2.4.18-0.22 kernel update (and all other
updates available through up2date) applied.
Here's another detail about the network freeze: it does not just freeze
networking on the laptop on which I installed Skipjack, but also takes out the
Netgear FS508 fast ethernet switch that the laptop is plugged into. All
computers plugged into the switch lose networking until the switch is rebooted.
Maybe this problem only happens with this particular combination of cardbus
interface and switch, but it did not happen with this same hardware in any
previous version of Red Hat Linux at least as far back as version 6.2.
Picking this issue back up, can you reproduce this problem with 8.0 kernel,
or the stock kernel.org epic100 driver?
If so, can you provide "lspci -vvv" output, run as root,
and output of "ethtool eth0"?
Created attachment 82366 [details]
output of 'lspci -vvv'
Created attachment 82367 [details]
output of 'ethtool eth0' before forcing a link
Created attachment 82368 [details]
output of 'ethtool eth0' after forcing a link
I switched to 3Com 3ccFE575CT cards on these machines (which work fine), but for
this test I reinserted the Ositech JOS card and booted Red Hat 8.0 with
kernel-2.4.18-17.8.0. I made no manual configuration changes after switching cards.
- There was no link detected at boot
- Running 'mii-tool -F 100baseTx-FD' forced a working link
- I copied a large (iso) file without any network lockup (this is new)
- Copy performance was bad: the copy took 6 times as long as it should, and
there were lots of 'collision' lights on the switch (this is also new)
I've attached the output of "lspci -vvv" and "ethtool eth0" run before forcing
the link, and "ethtool eth0" after forcing the link (output of "lspci -vvv" did
not change after forcing a link).
Thanks for the bug report. However, Red Hat no longer maintains this version of
the product. Please upgrade to the latest version and open a new bug if the problem
The Fedora Legacy project (http://fedoralegacy.org/) maintains some older releases,
and if you believe this bug is interesting to them, please report the problem in
the bug tracker at: http://bugzilla.fedora.us/