Comment 71 for bug 138691

Revision history for this message
Philipp Kern (pkern) wrote :

Ok, as I debugged this with Michael yesterday:

The bug, as it was reported, is about the failure messages sent to the console by network-manager. This is due to n-m not being able to cope with the hal and dbus shutdown through sendsigs, because own init scripts are missing, to shut the services down in a proper order.

The problem is aggravated with usplash not showing the console with the init script output, but another one. The n-m failures are probably just sent to the active console. I did not debug this any further because Michael's problem was neither n-m nor NFS/CIFS related:

I removed quiet and splash from the command line. This showed, upon shutdown, all init script and kernel messages. The kernel showed 'System halted' at last, so it was clear that it was an ACPI problem, i.e. the kernel is unable to turn the computer off. Somehow it didn't use ACPI on this particular machine, due to problems with the BIOS's DMI tables. Passing "acpi=off" to the kernel fixed the shutdown problem properly.

So I see three issues in this bug, all due to the fact that the n-m failure messages suggest the wrong place to look for the real problem:

1. n-m does not cope with hal/dbus shutdown, thus spitting out messages leading people the wrong way.
2. Hangs on NFS/CIFS umount. Those might be used over VPN, so it won't be umounted beforehand, or network-manager could bring down the network in sendsigs, and umountnfs won't be able to umount without a very long timeout.
3. ACPI trouble with power off. Those people will get "System halted" with kernel messages activated and can fix it by tweaking their ACPI kernel settings.