Bug 669587

Summary: NetworkManager should never, ever, ever bring up eth0 behind the scenes in runlevel 1
Product: [Fedora] Fedora Reporter: Matthew Miller <mattdm>
Component: systemdAssignee: Lennart Poettering <lpoetter>
Status: CLOSED NEXTRELEASE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: low    
Version: rawhideCC: lpoetter, metherid, mschmidt, notting, plautrba
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2011-04-12 12:18:11 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:

Description Matthew Miller 2011-01-14 03:57:04 UTC
runlevel 1 is single-user mode. It is used to fix problems. The network is, by definition, down.

On my rawhide system, I needed to fix a problem -- boot was failing because the lo interface wasn't coming up. (Weird in itself.)

I went into single user mode and tried to fix it. Runlevel 1 works fine, and the lo interface is up even. I ran a yum command without -C, which should have told me "hey, silly, you need a network connection". But instead, automagic stuff fired off and brought up the eth0 interface.

This is very slick, and I see how it's handy and all, but it *must not happen*.

So, I did `yum remove NetworkManager`, which unfortunately for some reason takes out all of the Sugar desktop environment too. On a hunch, I rebooted, and that also solved the loopback interface problem.

A double win? Well, maybe. I happen to not need NetworkManager on this statically-configured desktop system, but it sure is handy on laptops. And my daughter is going to miss Sugar.

So, the root problem: single user mode should mean it. Running a program which can access the network should fail, unless the admin has manually brought up the interface. Likewise, running an X app should not start X, and doing something which requires multi-user mode should not jump the system into runlevel 3.

Comment 1 Bill Nottingham 2011-01-14 15:57:53 UTC
This is likely NM's configured socket/dbus activation, not systemd itself. But requires a bit more checking.

Comment 2 Matthew Miller 2011-01-14 16:31:14 UTC
(In reply to comment #1)
> This is likely NM's configured socket/dbus activation, not systemd itself. But
> requires a bit more checking.

I should step back and be a bit less rant-y.

With systemd 9, dbus shouldn't be necessary in single user mode. Is it starting on my system because it's a continuously-updated rawhide system in need of a fresh start, or is systemd still configured to start dbus even though it's not strictly needed?

Comment 3 Lennart Poettering 2011-04-12 12:18:11 UTC
The D-Bus IPC system is a core part of the OS, which is why we establish it very early at boot. systemd itself communicates via D-Bus after all.

current NetworkManager in F15 should not be triggered anymore with tools like nm-online anymore, only if they are really and actively used. I assume this bug can be closed hence.