[OpenIndiana-discuss] exceeding MTU?

[Martin Frost]

I've been seeing timeouts in the middle of transfers while doing
'zfs send ... | ssh ... zfs receive ...' between subnets here.

In investigating with tcpdump, I found that packets are being sent
with lengths bigger than the supposed MTU of 1500, and I would like
to prevent that, as it may be the cause of the timeouts.

Do I need to explicit use '-mtu 1500' in ifconfig?  If so, how do I
make the system use it automatically when it brings the net up?

    823;hostfs/host/tcpdump.3#| ifconfig -a
    lo0: flags=2001000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv4,VIRTUAL> mtu 8232 index 1
	    inet 127.0.0.1 netmask ff000000 
    igb0: flags=1004843<UP,BROADCAST,RUNNING,MULTICAST,DHCP,IPv4> mtu 1500 index 6
	    inet XX.XX.XX.XX netmask fffffe00 broadcast 171.64.41.255
	    ether aa:bb:cc:dd:ee:ff
    lo0: flags=2002000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv6,VIRTUAL> mtu 8252 index 1
	    inet6 ::1/128 
    igb0: flags=20002004841<UP,RUNNING,MULTICAST,DHCP,IPv6> mtu 1500 index 6
	    inet6 XXXX::XXX:XXXX:XXXX:XXXX/10 
	    ether aa:bb:cc:dd:ee:ff

Lots of packets show this length:

    03:08:30.135800 IP hostfs.Stanford.EDU.41861 >
    hostbk.Stanford.EDU.ssh: Flags [P.], ack 3805, win 64436, options
    [nop,nop,TS val 236770530 ecr 193082465], length 16432

Some earlier batches of packets had lengths of 3400, then 4344, then
16432, then 2896, then 7240.

I assume it's doing adaptive sizing, but my real question is why are
any of these packets bigger than the 1500 MTU shown by ifconfig for
igb0 and how do I prevent use of packets longer than 1500?

Thanks!
Martin