>> scripts.  For some reason, I had put the greconfig line before the
>> ifconfig line.  Restoring the order of theese two lines cleared up the
>> problem.  This seems very strange to me, and I have no idea how to debug a
>> problem like this.
>If I recall correctly, you need to have the ifconfig (and especially a
>"default" route
>to the remote network) first in order to have the tunnel route calculated
>correctly.
>Perhaps I should look at the gif code, how the route is calculated there and
>change
>gre accordingly when I get a modem running under NetBSD again. Also there is
>a
>new RFC on GRE, that updates RFC1701/2 a bit.

	if it is permitted to break existing configuration, i would suggest
	redesigning gre(4) to take SIOCSIFPHYADDR, and split outer header
	address pair configuration from inner pair.  just like gif(4) does.

	there still are possibilities for routing loop within the kernel,
	however, it is much easier to configure/debug the tunnelling interface
	this way.  (and we can get rid of greconfig(8)!)

	the current configuration twist is there because gre(4) reuses
	interface address configuration to decide outer header address pair.
	outer header address pair should not be mixed up with interface address
	configuration, IMHO.

itojun