From Bugzilla Helper: User-Agent: Mozilla/4.76 [en] (X11; U; Linux 2.4.1-0.1.9 i686) I have two network cards; desc: "Realtek|RTL-8139" desc: "Unknown vendor|Generic tulip device" The unknown is a popular LinkSys LNE100TX ethernet card that is sold retail at more computer stores. Anaconda recognized the Realtek and installed the correct module (8139too) the LinkSys card was not recognized. Even worse after initial boot Kudzu recognized the LinkSys and installed the tulip driver and for some reason uninstalled the 8139too. It also reconfigured the eth0 interface and got stuck in an infinite loop until I broke out. The uninstall of the 8139too module caused the DHCP query to fail. However, upon rebooting the DHCP query was OK because the 8139too was re-loaded during the boot. The /etc/module.conf file is fine; alias eth0 8139too alias parport_lowlevel parport_pc alias sound-slot-0 emu10k1 alias usb-controller usb-uhci alias eth1 tulip There is only one interface setup file; /etc/sysconfig/network-scripts/ifcfg-eth0; DEVICE=eth0 BOOTPROTO=dhcp ONBOOT=yes No setup file for eth1. Reproducible: Always Steps to Reproduce: 1.install wolverine with 2 network card - one know the other unknown 2.during 1st boot use kudzu to install and configure second card Actual Results: Bad setup of first and second card Expected Results: both cards should be setup properly
What happens if you reboot with that configuration; does it work OK? What happens in this case if you disable hotplug?
Please see the bug I described above where kudzu hangs during the first boot after the install and fails to configures the network card (does put eth1 in /etc/module.conf file but unloads the 8139too module and gets stuck in an infinite loop and ultimately fails to writes the /etc/sysconfig/network-scripts/ifcfg-eth1 file). I suspect that anaconda does not complete the total configuration of the first card (eth0), which isn't totally configured until the first boot is completed (this is normal). But kudzu runs before the first card is finished being configured on the first boot and it sees only the second card (eth1) is unconfigured and because eth0 is not completely setup it gets mixed up as to which card it is really trying to configure. Upon rebooting the DHCP query was OK because the 8139too module was re-loaded during the subsequent re-boot. However, since no /etc/sysconfig/network-scripts/ifcfg-eth1 file was made, as I had setup during kudzu, the eth1 interface was never configured. I haven't yet tried to manually configure the LinkSys card (eth1), but I expect it will work when I do.
Do you have what the modules.conf looks like at the first boot before kudzu runs?
Where is kudzu hanging (on what screen)?
I'm sorry. The program that went into the infinite loop is netconfig. I was fooled because if kudzu finds an interface during the boot then netconfig is run automatically immediately afterward. However, netconfig only works with eth0. Either you need fix netconfig so that it works with more than one interface or not run it automatically during the boot after kudzu finds a new interface, because if you already have one installed netconfig will not work.
<looks at the code again> Oooooooops, wow, that's broken in a lot of ways. Will be fixed in pump-0.8.11 to in general be more sane.
We (Red Hat) really need to fix this before next release.