A mean bug/phenomenon took an hour of free time from me and create some headache… I tried to use partimage to make a backup of a hard disk I salvaged from an old computer. Since there is no package for partimage inside the archives of Ubuntu Jaunty, I downloaded the current version of SystemRescueCd and used it to do the job… The unpleasant surprise came, when i rebooted my system into Ubuntu Jaunty.
The NetworkManager told me that the network is not connected. It looked like eth0 was missing, so I blamed my DSL hardware first. I rebooted my DSL hardware router, checked my network cables… But still, I had no network, even though the lights on my switch and on the network card were showing that my card was connected. ethtool
$ sudo ethtool eth0 [...] Link detected: no
and dmesg
$ dmesg | grep eth [...] [ 39.513660] forcedeth 0000:00:08.0: irq 2302 for MSI/MSI-X [ 39.514304] eth0: no link during initialization. [ 39.515419] ADDRCONF(NETDEV_UP): eth0: link is not ready
said something else. I couldn’t activate my network under Ubuntu Jaunty… So i booted into the SystemRescueCD and to my surprise I could activate my network without problems.
The culprit seems to be kernel 2.6.29.3 out of the SystemRescueCD. There’s a regression which impedes that the network card can be activated after kernel 2.6.29.3 turned the system off. My “solution” to this problem was
- Start the computer with the SystemRescueCD
- Activate network
- Turn the computer completely off without shutting it down
After this I could reboot into my Ubuntu Jaunty and my network worked like nothing happened. My guess is that kernel 2.6.29.3 has some new methods to shut the system (including the network) off via ACPI, which “older” kernels are not aware of. Eventually it would have been enough to completely turn my computer off (unplug it or take the battery out), but I didn’t tried that yet.
I think that this phenomenon might hit anyone with a system which uses the forcedeth kernel module. So keep this info in mind, when you test Ubuntu Karmic…

22. May 2009 at 17:38
Jaunty macht irgendwie nicht wirklich Spaß …
22. May 2009 at 17:42
Tach Shakal, Jaunty macht hier nix falsch. Was kann Jaunty, oder wahrscheinlich jede andere Distribution, dafür wenn ein neuerer Kernel ACPI-Settings kennt, die der “alte” Kernel noch nicht beherrscht…
23. May 2009 at 10:31
Moin, in dem Fall hast Du natürlich Recht. Habe ich so nicht gesehen … aus meiner Sicht ist ACPI aber schon ein generelles Problem.
25. May 2009 at 09:17
Servus,
ich danke Dir vielmals für diesen Beitrag. Nachdem gestern eine meiner Partitionen auf wundersame Weise verschwunden war, half mir nur noch die SysRescCD, die Daten wiederherzustellen.
Und siehe da, nach einem Neustart hatte ich das gleich Problem. Aber zum Glück hatte ich den Beitrag hier noch im Hinterkopf
Du hast mir sicherlich einige Zeit gerettet. Daher ein großes Dankerschön