Re: STEP 2, fixing dhclient behaviour with multiple interfaces

On Tue, Jul 29, 2003 at 10:18:05PM -0600, M. Warner Losh wrote:
> In message: <20030730013027.D25188_at_cvs.imp.ch>
>             Martin Blapp <mb_at_imp.ch> writes:
> : Hi Folks,
> : 
> : I had a closer loom at the OMAPI stuff in dhclient.
> : 
> : Just to say, I'm very disappointed. The only objects that exist are:
> : "control" and "interface". The later is not inplemented at all.
> : It pretends to work, but if you look at the source there are
> : stubs only :P.
> : 
> : "control" does only release leases and exit (state 2), I never managed to make
> : dhclient sleep (state 3) and wake up (state 2).
> : 
> : It seems that the whole implementation is only a proof of concept
> : and not very functional at all, at least for the client side, dhcpd
> : may be different.
> : 
> : I'll think about how we can solve this differently. This
> : really sucks !
> 
> That's why I start/kill dhclient from pccard_ether.

The trouble with that is when you have one fixed interface and one from
a pccard.  You want the fixed interface to be running dhcp continuously,
and not have it interrupted when the pccard is plugged in.

I guess that we could do this by running a per-interface dhclient, but
I've not had much success with that in the past.

I'd like to talk about how that interacts with laptop suspend behaviour,
but given that my laptop still can't resume properly, I'm inclined to
leave it for the moment.  ;-)

-Dom