Please describe the problem:

When using the "bond" driver to combine two ethernet ports eth0 and eth1 together into a third virtual device bond0, it appears that NetworkManager does not realize that the ethernet ports are in use and tries making connections on them.

When they respond with DHCP, they get enabled (even though they are part of the bond device), and the routing table gets messed up - the connection to gateway record is replaced by another one specific to the ethX device.

The computer is now unusable for outgoing network connections.

Steps to reproduce:
1. (see below)
2. 
3. 


Actual results:
Network connectivity is lost.

Expected results:
One would expect NetworkManager to recognize that ethernet devices are bonded into a different virtual device, and not try making connections on them... Or a way to disable searching on particular ethernet ports.

Does this happen every time?
YES

Other information:

- I am trying to use the "bond" driver to combine both of my ethernet ports into one device with double the bandwidth.

- The bond driver appears as a separate networking device "bond0", in addition to the two ethernet ports eth0 and eth1 which are "slaves".

- When I bring up the bond driver, it gets DHCP information for IP address and gateway properly.  Good so far.  Everything working perfectly.

- After about a minute, it appears that NetworkManager-something is not realizing that the ethernet ports are active as part of the bond device and tries making separate network connections.

- This gives an IP address to one of the ethernet devices (I presume via DHCP), and screws up the routing table to the gateway, effectively taking the computer off the network.
 
- "Stopping" NetworkManager and NetworkManagerDispatcher seemed to keep this from happening.

OS:  debian "testing" distribution, amd64, apple/mac pro computer w/ two intel 80003ES2LAN gigabit ethernet controllers.