Bug 530669

Description of problem:

If the IPv6 method is set to "automatic", and there's no already configured address on the interface (done by the kernel), the connection fails to activate (even for IPv4).

Version-Release number of selected component (if applicable):

Fedora 12 beta, NetworkManager 0.7.996-4.git20091002.fc12.i686

How reproducible:

100%

Steps to Reproduce:
1. Set up a network with IPv4 DHCP service and IPv6 SLAAC service (I used standard F11 virt-manager, adding an IPv6 address to virbr0 manually afterwards and starting radvd on that interface)
2. Install F12 beta and change the IPv6 method in the connection editor for system eth0 from "ignore" to "automatic" and save/exit (disappointing default by the way).  Due to bug #523288 you'll need to untick "available for all users" for this to take effect.  At this point you should already have an (non-link-local) IPv6 address configured on eth0 by the kernel.
3. Attempt to reconnect to the network by clicking on the network-manager icon and then "system eth0" in the list.
  
Actual results:

NM fails to connect to the network, leaving it as "disconnected" with no IPv6 or IPv4 addresses (except a link-local IPv6 one).  A tcpdump on the KVM host machine reveals that the guest does send out a DHCP request (and receive a reply), but no ICMPv6 router solicitations appear.

Expected results:

NM should have correctly configured the network with a IPv6 address from SLAAC and a IPv4 one from DHCP.

Additional info:

If the kernel has already auto-configured an IPv4 address on the interface, the connection is established like expected.  You can force that to happen with the commands "ip link set dev eth0 down; ip link set dev eth0 up".