Bug 107998 - Red Hat Linux Enterprise WS 3.0 does not work with 3com 3c905-TX cards
Summary: Red Hat Linux Enterprise WS 3.0 does not work with 3com 3c905-TX cards
Keywords:
Status: CLOSED DUPLICATE of bug 107389
Alias: None
Product: Red Hat Enterprise Linux 3
Classification: Red Hat
Component: kernel
Version: 3.0
Hardware: i686
OS: Linux
high
high
Target Milestone: ---
Assignee: Jeff Garzik
QA Contact: Brian Brock
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2003-10-25 22:02 UTC by John Buswell
Modified: 2013-07-03 02:15 UTC (History)
8 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2006-02-21 18:59:24 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)

Description John Buswell 2003-10-25 22:02:38 UTC
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.5) Gecko/20031007
Firebird/0.7

Description of problem:
Following a successful installation of RHEL WS 3.0 on an Athlon XP 2100+ w/MSI
motherboard and dual 3com 3c905-TX NICs, it failed to connect to
Red Hat during firstboot. After cancelling and logging in as root, the
system could see each of the NIC cards, but dmesg was full of PCI bus Error /
Host Error messages. Pinging or any kind of network activity
did not work, pinging would try to send a few packets and then produce
the PCI bus error messages. I placed a Netgear FA311 card into the 
system, and downloaded vanilla kernel 2.4.22, recompiled and the NICs worked
fine. However, the RedHat network scripts were fairly messed up, they kept
trying to remap eth0 to eth2, and ended up calling one device (instead of eth2)
dev025325. The installation was pretty hosed, so I removed both 3com 3c905-TX
cards, and reinstalled with an Intel EEPRO and Netgear FA311, and it worked
flawlessly.

I retested with two other 3com 3c905-TX, the first test was with a Rev A and Rev
B card, I tried it with two Rev.B cards (different cards) in a different system
(Pentium III 800MHz Dell) and had the same results. It looks like either the
driver in 2.4.21 is hosed, or something RHEL has patched into the kernel is
interfering with the driver.

It would be nice if Red Hat could fix this kernel issue, and provide new ISOs
with an updated kernel so this can be cleanly installed on systems with 3c905-TX
cards. 

Version-Release number of selected component (if applicable):
kernel-2.4.21-4.EL

How reproducible:
Always

Steps to Reproduce:
1.Install RHEL WS 3.0 on a system with two or more 3c905-TX cards
2.Boot up, and look in dmesg

    

Actual Results:  Received PCI bus errors (sorry don't have exact output, was in
a rush to get the system back up and running).

Expected Results:  Ethernet / IP to work correctly

Additional info:

set severity to high as RHEL WS 3.0 fails to work on a system with a 3c905-TX card.

Comment 1 Uwe Beck 2003-11-06 19:41:51 UTC
See bug #102685.
Which card do you use?
3c905 (Boomerang) --> your report!?
3c905B (Cyclone) --> do not work on 10 MB hub, no 10HD
3c905C (Tornado) build in 1999 works correct in all modes by me


Comment 2 John Buswell 2003-11-06 20:14:03 UTC
This is not related to 102685, both cards are connected to 10/100
ports on a Cisco Catalyst 5500 WS-X5225R blade. The ports were both
running at 100 base.

The system had :

3c905 (boomerang - REV A card)
3c905b (cyclone - REV B card)

There was no network functionality available.



Comment 3 Uwe Beck 2003-11-07 13:17:22 UTC
3c905: no network functionality available, see your messages file

3c905b: no 10Mbit HD on 10 MB hub, on 10/100 switch 100 Mbit FD is ok
mii-tool -v means, this card do not know 10 MB

[root@pclin02 root]# mii-tool -v
eth0: 10 Mbit, full duplex, link ok
  product info: vendor 01:e1:c1, model 56 rev 7
  basic mode:   isolate, collision test, 10 Mbit, full duplex
  basic status: autonegotiation restarted, link ok
  capabilities:
  advertising:  100baseT4 100baseTx-FD 100baseTx-HD flow-control
  link partner: 100baseT4 100baseTx-FD 100baseTx-HD flow-control
  No MII transceiver present!.

Sorry, the card works with 100baseTx-FD at this time.

I use autonegotiation on switches because the FS108 (Netgear) or
AT-FS716 (Allied Telesyn) are not user-managed switches.


Comment 4 John Buswell 2003-11-07 15:17:57 UTC
See my messages file, thats helpful.

I have already FIXED this problem, I'm trying to help you out,
because I like Red Hat. Right now if someone has a 3com 3c905a/b
card in their system, Red Hat Enterprise Linux is unusable!!

This is very odd, because the problem ONLY exists with Red Hat
Enterprise Linux. The system was working fine, running the same kernel
version with Mandrake Linux 9.2rc2, and later Mandrake Linux 9.2.

The problem ONLY occurs with Red Hat Linux Enterprise WS 3.0.

To prove this point, I have a second system (same CPU, memory,
motherboard, and installed the same NICs, which were removed from the
original system). The original system works fine, with other NIC cards
in it.

I installed RHEL WS 3.0 on the secondary system, again it did the same
thing. If for example, I try to ping out eth0 (3c905), it would send
5-10 packets, you'd see nothing come out the NIC card, and the ping
tool would report destination host unreachable. After 10 packets, the
console would display a PCI bus error. I removed the 3c905, and
installed another one, to make sure it wasn't the card. Problem still
occured. I removed the 3c905b, and placed another one in, same problem. 

Next, I took another Enterprise Linux product, and installed it. It
worked fine, both NICs, no problems. I tried 3 different versions of
this other product, 0.1.4 (which ran 2.4.21), 0.2.0 (which ran 2.4.22)
and 0.3.2 (which runs 2.6.0-test9). They all worked without any problems. 

I reinstalled RHEL WS 3.0 on the system, put a third NIC (netgear)
into it. Downloaded 2.4.22 kernel source, rebuilt the kernel, and it
worked fine.


Comment 6 Dan Gennidakis 2003-11-07 18:10:00 UTC
I have the same problem with Fedora core (test 1- release) and my 
3c905 Boomerang card which had no probs with RH 7-9. The problem 
appears to be with Kudzu vers above 0.99 (after RH9). If Kudzu is 
off at boot time for the runlevel the cards initialize ok. When I 
downgraded Kudzu to the RH9 version (0.99) the NIC initialized ok 
without problems. 

Comment 7 Dan Gennidakis 2003-11-07 18:22:50 UTC
 See also bug # 107389 and the effect of kudzu at boot time 
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=107389 . Looks 
like the same problem to me. Try turning off Kudzu at boot time with 
chkconfig and see what happens. Also since I am not using Ent. WS V3 
what version of Kudzu is used with that OS?

Comment 8 John Buswell 2003-11-07 21:22:31 UTC
Dan,

kudzu-1.1.21-1 is whats on Ent. WS v3.

The error message posted in #107389, is one of two messages I saw.
Have you seen one about the PCI bus ? I would get both if I made any
ping attempts.

*** This bug has been marked as a duplicate of 107389 ***

Comment 9 Uwe Beck 2003-11-09 19:29:54 UTC
Without kudzu at boot time the 3c905 (Boomerang) works correct. I
tested it with RHEL WS 3.
I think, there will be the same result in RHEL ES 3 (the same RPM for
kudzu) and also the same result with 3c905b (Cyclone). But let me test
it.


Comment 10 Uwe Beck 2003-11-10 13:15:37 UTC
3c905 (Boomerang) and 3c905B (Cyclone) works correct with RHEL ES 3 if
kudzu is not running at boot time.


Comment 11 Neil Horman 2004-03-09 19:16:36 UTC
there may be a patch in BZ 102685 tht fixes this

Comment 12 John Airey 2004-09-30 10:03:47 UTC
There appears to be a bug with the network driver that only affects 
certain revisions of these cards. I'll try to get hold of the 
specific cards and post the details here.


Comment 13 John Airey 2005-05-11 16:11:16 UTC
As promised, this what dmesg showed for one of these cards:

3c59x: Donald Becker and others. www.scyld.com/network/vortex.html
See Documentation/networking/vortex.txt
00:10.0: 3Com PCI 3c905B Cyclone 100baseTx at 0xe880. Vers LK1.1.18-ac
 00:10:5a:40:94:bf, IRQ 9
  product code 5152 rev 00.12 date 08-30-98
Full duplex capable
  Internal config register is 2400000, transceivers 0xa.
  8K byte-wide RAM 5:3 Rx:Tx split, 100baseTX interface.
  MII transceiver found at address 24, status 784d.
  Enabling bus-master transmits and whole-frame receives.
00:10.0: scatter/gather enabled. h/w checksums enabled
divert: allocating divert_blk for eth0
divert: freeing divert_blk for eth0
ip_tables: (C) 2000-2002 Netfilter core team
ip_conntrack version 2.1 (1024 buckets, 8192 max) - 304 bytes per conntrack
3c59x: Donald Becker and others. www.scyld.com/network/vortex.html
See Documentation/networking/vortex.txt
00:10.0: 3Com PCI 3c905B Cyclone 100baseTx at 0xe880. Vers LK1.1.18-ac
 00:10:5a:40:94:bf, IRQ 9
  product code 5152 rev 00.12 date 08-30-98
Full duplex capable
  Internal config register is 1000000, transceivers 0xa.
  8K byte-wide RAM 5:3 Rx:Tx split, autoselect/10baseT interface.
  MII transceiver found at address 24, status 7849.
  Enabling bus-master transmits and whole-frame receives.
00:10.0: scatter/gather enabled. h/w checksums enabled

The same card is detected twice. 

This didn't work with the current latest kernel (2.4.21-27.0.4), but swapping it
with another card did.

I can do more testing or even post the card to someone if that helps.

Comment 14 Red Hat Bugzilla 2006-02-21 18:59:24 UTC
Changed to 'CLOSED' state since 'RESOLVED' has been deprecated.


Note You need to log in before you can comment on or make changes to this bug.