Description of problem: Doing an install of FCT2 or Apr 14 2004 rawhide from a boot.iso, still in the curses stage I tell it to do an NFS install. On the next screen it shows both of my nics and asks which one I would like to configure. It displays them both as eth0. The display looks roughly like this: eth0 -- 3COM blah blah blah eth0 -- Prism yenta blah blah blah This is my Dell Inspiron 4150 laptop with a internal 3COM PCI 10/100 nic and internal pcmcia 802.11b nic.
Bill, I'm doing... devs = probeDevices(CLASS_NETWORK, BUS_UNSPEC, PROBE_LOADED); So I'm not sure why kudzu would show me both of them... Dax -- do you see modules for both the prism and the 3com being loaded if you look on tty4? Also, what sort of stuff is there on tty3?
Alright, gotta catch a plain a couple hours, but I got the info you needed. Boot off of boot.iso, select language, get to the "choose source (NFS,CDROM,FTP,etc) screen. tty3 shows: modules to insert 3c59x hermes orinoco orinoco_cs loaded blah blah modules.cgz loaded blah blah modules.cgz loaded blah blah modules.cgz loaded blah blah modules.cgz inserted 3c59x inserted hermes inserted orinoco inserted orinoco_cs loaded module set done over on tty4 I see the standard kernel messages. 3c59x is eth0 and orinoco_cs is eth1. I then select NFS, and I get a screen that says: Networking Devices -- you have mutliple blah blah eth0 - 3com Corporation|3c905C-TX/TX-M [Tornado] eth0 - Intersil PRISM2 11 Mbps Wireless Adapter Flipping over to tty3 I see these new messages: 53 keymaps are available device[0] is 3COM Corporation|3c905C-TX/TX-M [Tornado] - eth0 using 3c59x (loaded) device[0] is 3COM Corporation|3c905C-TX/TX-M [Tornado] - eth0 using 3c59x (loaded) (yes that line is repeated twice) need to setup networking going to pick interface ----------------------------------------- Thats it. Something is foobared. BTW, futher in the install during the set your networking details page, eth0 and eth1 are displayed. I remember around RH8/9 time frame, Anaconda was modified to not show wireless devices. In fact, I was the bug reporter on that one. Could this be related?
The old bug I talking about is: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=71602
This might be fixed with test3... I can probably get my mostly broken thinkpad to work for long enough to at least try this tonight.
I'm seeing this with FC-devel as of 2004-05-05, so I don't think it's fixed yet.
Can you grab http://people.redhat.com/~katzj/net-test.c and build it with 'gcc -o net-test net-test.c -lkudzu -lpci' and run it and provide the output?
[root@mentor tmp]# ./net-test dev[0] uses the 3c59x driver, device is eth0 dev[1] uses the orinoco_cs driver, device is eth2 BTW, I don't have an eth2, just eth0 (3c59x) and eth1 (orinoco_cs).
[root@localhost root]# ~barryn/net-test dev[0] uses the e100 driver, device is eth0 The fact that my wireless isn't showing up at all is probably bug 121742. Once I run "modprobe yenta_socket", this is the output: [root@localhost root]# ~barryn/net-test dev[0] uses the e100 driver, device is eth0 dev[1] uses the orinoco_cs driver, device is eth1
*** Bug 122398 has been marked as a duplicate of this bug. ***
Can you try with tomorrow's tree? I've made some kudzu changes that may have a positive effect on this
I just tried with the 20040507 tree. This problem still exists.
Reproduced finally and can definitely point fingers at kudzu screwing things up... Bill knows what's going on and had an idea on how he wanted to fix it.
Still happens with the 20040713 tree (earlier today I checked mirrors.kernel.org and it didn't have anything newer yet).
Still happens with 20040822 rawhide tree
Please try rawhide - this should be better there.
(Sorry I didn't respond earlier) Yes, it's fixed in recent rawhide (I don't remember exactly how long ago) and in RHEL 4 beta1. I'm not the person who filed this bug, otherwise I'd be closing it now.