[Soekris] Max MTU size

[Christopher E. Brown]

On Sun, 28 Oct 2007, der Mouse wrote:

>> The Via 6105M Chips on the 5501 support 802.1q VLAN tagging, which
>> would be an Ethernet frame size of 1522,
>
> One of us has something wrong here.  I'm perfectly willing to believe
> it's me, but I would be interested in knowing what I've got wrong.
>
> As I understand it, the maximum standard Ethernet frame is 1516 bytes:
> 6 bytes of destination MAC, 6 of source MAC, 2 of frame type, 1500 of
> payload, and 2 of FCS.  802.1q increases this by 4 (2 for the
> "vlan-tagged" frame type and 2 for the vlan number, basically),
> bringing it to 1520, not 1522.  Where's the mistake?

FCS is 4, not 2

1500(payload) + 14(header) + 4(dot1q) + 4(FCS) == 1522 On the wire


The normal ethernet payload is 1500bytes, + 14 header (src/dst/frametype). 
"On the wire" there is an additional +4 FCS (CRC checksum).

So, true on the wire is 1518, but the FCS is often ignored when it comes 
to device specs/docs as it is per link and generally handled by the 
ethernet chip itself.  (It is added each time the packet is transmitted 
and stripped on receive)


802.1q adds another 4bytes, for 1518 packet, 1522 "on the wire" (QinQ 
takes this to 1522/1526, stacked QinQ assd another 4 per tag).


When looking at ethernet chip specs you need to be careful, some spec 
sheets include FCS in the max packet size spec, but many do not.  An 
example would be a Cisco 7200 PA2-FE, the listed "max" packet size is 
1526, but this does not include FCS (on the wire max is 1530, you have 
1526 for everything else).


In other words, unless dealing with driver internals for some chips or 
very unusual ethernet switches you can ignore the existance of FCS.