lagg of em0/em1 + VLAN = lower MTU?

[Karl Pielorz]

--On 13 July 2015 13:06 +0100 Gary Palmer <gpalmer at freebsd.org> wrote:

> Have you read the HARDWARE section of vlan(4)?

Kind of cryptic answer ;) - But I just read vlan(4). So it looks like my 
understanding of vlans (i.e. 'long frames' as it calls them) was right?

On supported kit - creating a sub interface on a VLAN on will not reduce 
the MTU of that sub interface, e.g. to 1496.

em is listed as Hardware Tagging supported, but is not listed under the 
'natively support long frames for vlan' in that man page. But I can set the 
MTU on em to anything I want, e.g. 1504, 1600, anything up to 9k on these 
cards - that's a little confusing.

This still doesn't really solve the question of why the MTU remains at 1500 
(as I think it should) when adding a VLAN to a standalone em interface, but 
it 'shrinks' to 1496 when the parent interface is a LAGG of em's.

Even if the actual LAGG (and it's members) are forced to an MTU 1504 first 
before the VLAN interface is created (which is useless in this environment 
anyway), the created VLAN interface comes up with an MTU of only 1496 (i.e. 
not '1504 -4' if it were doing that maths).

End of the man page says, "The vlan driver automatically recognizes devices 
that natively support long frames for vlan use and calculates the 
appropriate frame MTU based on the capabilities of the parent interface."

I would ask - does that include if the parent interface (lagg) has another 
parent (i.e. em0/em1)? - It would seem to in at least some cases (i.e. the 
other people who seem to have this working).

-Karl