At 09:35 PM 10/1/2000 +0200, Manuel Bouyer wrote:
>Hi,
>I tested our 802.1Q support with a cisco switch today, and ran into the
>problem mentionned in kern/11109 (I expected to run into this anyway :),
>which is that on a cisco trunk port we can expect to see packets of
>1522 bytes instead of 1518 (the 4 extra bytes being the 802.1Q tag).
>If we want to be able to talk with others 802.1Q devices we need to solve
>this.
>
>There are several issue here:
>- hardware: does the ethernet adapter supports frames of 1522 bytes ?
>   I looked quickly at the tulip and smc83c170 sources and found no way to
>   change this, it looks like the 1518 bytes values is hardcoded in hardware.
>   Does anyone know if there are (10/100) ethernet chips that can send/receive
>   frames bigger than the standart len ?

The tulip can.  For receive, you just need to ignore the frame length error.
the de driver has code to do that.

>- software: assuming there are adapters that can do something else than 1518
>   (gigabit adapters can), we need a way to advertize this to the vlan layer.
>   For now the vlan layer uses the MTU advertised by the ethernet driver.
>   We can't change this (from 1500 to 1504) because non-802.1Q packets would
>   be too big. So we may need a new field in the if_data struct (maybe 
> ifi_rmtu)
>   to advertise the real MTU the interface can do, and which can be used by
>   pseudo-interfaces like vlan.

Personally, I'd rather have a IFF_ flag that indicates this instead.  Not 
all 802
media is ethernet.
--
Matt Thomas               Internet:   matt@3am-software.com
3am Software Foundry      WWW URL:    http://www.3am-software.com/bio/matt/
Cupertino, CA             Disclaimer: I avow all knowledge of this message