gif4) & AltQ

[Eric Masson]

>>>>> "Max" == Max Laier <max at love2party.net> writes:

Hello Max,

 Max> Very true. It's more worthwhile to classify on gif and queue on
 Max> the real interface.

How would you achieve this setup ?

I can only think about this way (assuming gif0 tunnel packets flow thru
ep1) :

ext_if="ep1"
tunnel_if=gif0"

altq on $ext_if bandwidth 2Mb cbq queue { dflt, developers, marketing }
queue dflt bandwidth 5% cbq(default)
queue developers bandwidth 80%
queue marketing  bandwidth 15%

pass out on $tunnel_if from 192.168.0.0/24 to any keep state queue developers
pass out on $tunnel_if from 192.168.1.0/24 to any keep state queue marketing

But in this setup classification is made on unencapsuled packet, and
shaping is done on encapsulated packet. Does this mean that the mbuf tag
set by classification rules survives the gif encapsulation process ?

 Max> Queueing on gif will only work in rate-limiting mode.

Ok.

 Max> That's true as well. Take a look at the patches on:
 Max> http://people.freebsd.org/~mlaier/ALTQ_driver/ and the altq(9)
 Max> manpage to learn how to modify a driver. It's more or less looking
 Max> for if_snd and modifying it according to the rules in altq(9).

I'll have a look.

 Max> Not sure how *exactly* gif(4) works, but I'll put it on my list
 Max> (just not a high priority, right now).

Ok, thanks to you and Brooks for explanations.

Regards

Eric Masson

-- 
 C'est pas avec la censure que tu vas censurer les censeurs.
 -+- JL in GNU : Las, censeurs pour l'échafaud -+-