my laptop has a realtek ethernet, and broadcom wireless. Anaconda displays a screen asking which to install through that looks like.. eth0: Broadcom corporation BCM4318 yadayada eth1: Realtek Semiconductor RTL8139 yadayada This ordering seemed odd to me, so I checked dmesg. It displays.. eth0: RealTek RTL8139 at 0xffffc20000f0a400, 00:0f:b0:bb:0f:7a, IRQ 22 eth0: Identified 8139 chip type 'RTL-8100B/8139D' Anaconda is really confused. To install through ethernet, I have to select the broadcom wireless entry. Selecting the Realtek entry gets me.. 17:21:20 DEBUG : dhcp: ioctl SIOCGIFFLAGS failed: 19 No such device (understandable given it doesn't have firmware for the device) logs attached..
Created attachment 201351 [details] screenshot showing interface confusion.
Created attachment 201361 [details] anaconda log
Created attachment 201371 [details] syslog during install
*** Bug 300101 has been marked as a duplicate of this bug. ***
Kind of a digression, but arguably we ought to be excluding the bcm43xx anyway -- we don't ship firmware for it, and (AFAIK) we don't have any kind of 'driver disk' method for providing it.
Yeah, I wonder where it's getting the 'eth' name from at all, because it doesn't exist until you load firmware for it iirc.
It comes from kudzu when we use it to return a list of probed ethernet devices. When a module is loaded without the firmware, kudzu gives it the device name 'eth', which messes up our sorting in anaconda since we expect eth to be followed by an int.
We should be ignoring the devices named 'eth', though. But I'm not seeing devices with missing firmware coming up as just 'eth' now. Which is easy to test, as firmware loading doesn't seem to be working on iwl3945 either...
I saw this same problem with the IBM A/B/G card in my Thinkpad. This is an Atheros device, so AFAIK, there isn't any driver at all for it (did MadWifi get into the distro?), but it still showed up in the Anaconda device list, messing up the order.
Hack of a fix in place although a couple of drivers (ath5k, b43legacy) were left off the list in the current build. Next anaconda build will add those
The problem persists with F8 test3. I'm too having a Broadcom BCM4306 wireless chip in my PPC Powerbook. I can provide more info if needed.
Yeah, the PPC bcm bits is being tracked as bug 311421 (it's a little bit more subtle to fix)
Works for me on a ppc powerbook and current rawhide. Dave, does this work for you?
works for me too.