So, I adjusted the minimum mtu in tun_ioctl()  down to 256, and
gave it a try.  Everything seems to work fine. NetBSD fragments
the packets, and they make it to the far end and are reassembled
correctly.

However, while testing this, I noticed some strange behavior.

when I run "ifconfig tun0 mtu NNN", it doesn't change the MTU for
any existing routes using the specified interface. (viewable via
"netstat -r").  In fact, after adjusting the mtu and adding a new
route (with "route add"), the old MTU is still reflected in the
new route entry:

	# ifconfig tun2
	tun2: flags=2<BROADCAST> mtu 1500
	# ifconfig tun2 192.168.3.1
	# ifconfig tun2
	tun2: flags=43<UP,BROADCAST,RUNNING> mtu 1500
		inet 192.168.3.1 netmask 0xffffff00 broadcast 192.168.3.255
	# netstat -nr
	...
	Destination        Gateway            Flags     Refs     Use    Mtu  Interface
	...
	192.168.3          192.168.3.1        U           0        0   1500  tun2

Everything is as expected so far, however adding a route after
changing the MTU adds the route with the original MTU...  I would
have expected the new route to have the new MTU.  In fact, I would
have expected the MTU on all of the existing routes pointing to
this interface to get the new MTU.

	# ifconfig tun2 mtu 1024
	# ifconfig tun2
	tun2: flags=43<UP,BROADCAST,RUNNING> mtu 1024
		inet 192.168.3.1 netmask 0xffffff00 broadcast 192.168.3.255
	# route add -net 192.168.100.0 192.168.3.1
	add net 192.168.100.0: gateway 192.168.3.1
	# netstat -nr
	...
	192.168.100        192.168.3.1        UGS         0        0   1500  tun2
	...

This is NetBSD-1.5, if it makes a difference.

I'm not a networking expert, so if this is the expected/desired behavior,
sorry for wasting time.

-Tad

In message <20021009063459.8FD3A4B24@coconut.itojun.org>, you said:
;>An MTU of less than 576 violates the IPv4 standard, and will cause
;>all kinds of obscure (and some not-so-obscure!) problems, even in
;>the presence of Path MTU discovery.
;
;	incorrect, 576 is the minimum supported size for the packet
;	size after reassembly.  minimum MTU requirement is 68 octets.
;
;RFC791:
;>>    impractical for most hosts and networks.  All hosts must be prepared
;>>    to accept datagrams of up to 576 octets (whether they arrive whole
;>>    or in fragments).  It is recommended that hosts only send datagrams
;>>    larger than 576 octets if they have assurance that the destination
;>>    is prepared to accept the larger datagrams.
;
;>>    Every internet module must be able to forward a datagram of 68
;>>    octets without further fragmentation.  This is because an internet
;>>    header may be up to 60 octets, and the minimum fragment is 8 octets.
;
;itojun