Patch to prevent multiple dhclients starting on an interface

[Ted Lemon]

On Jun 15, 2005, at 2:55 AM, Andrew Pollock wrote:
> Martin Pitt with his Ubuntu hat on has made a patch to address this.
>
> It's at
>
> http://patches.ubuntu.com/patches/dhcp3.client-running_stale_pids.diff
>

It seems like what this patch does is to make sure that when the  
client's run a second time, it kills the first instance of the  
client.   You're in a difficult spot because of the whole ifup/ifdown  
mechanism, which interacts poorly with automatic address  
configuration.   I really don't think this patch is a good idea, but  
the right solution is a lot harder, and not something I expect to see  
done to the ISC code anytime soon - you should really be running  
dhclient as a single daemon, and sending it events over the d-bus  
from the ifup and ifdown scripts.

As a temporary hack, my big complaint about that patch is that it  
sends a SIGTERM to the pid recorded in the dhclient pidfile, which  
may be stale, without checking whether or not there's actually a  
client running by trying to bind to port 68.   I'm running into some  
obstacles with that check on Ubuntu right now - it always returns  
positive, even though netstat reports no udp listeners - but I think  
the right thing is to fix the test and use it, not to rely solely on  
the pid file.   Otherwise, you run the risk of killing some other  
process that just happens to have the same pid as the stale dhclient.