NetworkManager Bridges in Fedora

Dan Williams dcbw at redhat.com
Thu Jan 16 20:38:53 UTC 2014


On Thu, 2014-01-16 at 14:53 -0500, Steve Dickson wrote:
> 
> On 16/01/14 14:39, Steve Dickson wrote:
> > 
> > 
> > On 16/01/14 14:09, Dan Williams wrote:
> >> Also, if wouldn't mind passing along the systemd journal output for
> >> NetworkManager, that might help us figure out what's going on:
> >>
> >> journalctl -b _SYSTEMD_UNIT=NetworkManager.service
> > The log is at:
> >   http://steved.fedorapeople.org/tmp/nm.log 
> I bet ya the hang has something to do with these messages:
> 
>    <info> (bridge0): IPv4 config waiting until carrier is on
>    <info> (bridge0): IPv6 config waiting until carrier is on
>    <info> Activation (bridge0) Stage 3 of 5 (IP Configure Start) complete.
>    
> What carrier is it waiting on? em1 is already up and running... 

So what's happening here is that you two
configurations/connections/profiles that apply to em1: "ifcfg-em1", and
"ifcfg-bridge0_slave_1".  And "ifcfg-em1" is getting chosen, but it's
not a bridge slave configuration, it's just a normal DHCP configuration.

Since "ifcfg-em1" is not a bridge slave configuration, it never gets
added to the bridge master, and the bridge master sits around waiting
for slaves because it has no carrier which is required for DHCP.

Persistent fix #1:
edit ifcfg-em1 and change ONBOOT=yes to ONBOOT=no
nmcli con reload
then do "Runtime fix" below

Persistent fix #2:
rm /etc/sysconfig/network-scripts/ifcfg-em1
nmcli con reload
then do "Runtime fix" below

(or use nm-connection-editor to delete 'em1' or uncheck "Connect
automatically" from the General tab.)

Runtime fix (not persistent):
nmcli dev disconnect em1
nmcli con up "bridge0 slave 1"


------
(In old "network" service speak, the runtime operation would be:

ifdown em1
ifup bridge0_slave_1

and we still expect these commands to work even when NetworkManager is
managing the interface.)
------

Dan



More information about the devel mailing list