Created attachment 334072 [details]
dhcp=internal connection attempt that doesn't complete

When using dhcp=internal and attempting to activate a connection where the RA indicates that there's information to be found in DHCP (O-flag set) but the DHCP server isn't responding, the connection activation never completes. Instead it keeps retrying the DHCP request forever. The general state is stuck in "activating", dispatcher scripts aren't invoked, and so on.

While activating the same connecting sing the default DHCP implementation, the DHCP request quickly reaches a timeout, and the connection activation completes successfully, the dispatcher scripts are invoked, and all is well.

It is admittedly a network misconfiguration to have an RA say there's information in DHCP while not actually have a functioning DHCP service, but it's probably not that uncommon, and I think it would be best if NM w/dhcp=internal handled it as well as it does with the default DHCP client. For what it's worth, in my case the misconfiguration is caused by a firmware bug in my laptop's 3G WWAN module so it's not really something I can easily fix.

I'm attaching two logs, one with a connection attempt with dhcp=internal (that I manuall cancel after over 30 minutes), and then a connection attempt with the default DHCP client where a timeout is reached after 20 seconds after which the connection completes activation.