HTB + NAT on Debian (outgoing traffic shaping problems)

hello,

I'm running Debian Woody with 2.4.26 kernel. I have 3 NIC:
eth0 - 192.168.0.0 subnet
eth1 - 192.168.2.0 subnet
eth2 - DSL link

there are ~5 computers in every subnet and I want them to get equal
incoming and outgoing traffic with possibility to borrow unused
channel.

I have successfully used htb.init to shape my incoming traffic but I
ran into problems with outgoing. I try to mark outgoing packets with
iptables, but it doesn't work...

I do marking like this:
iptables -t mangle -A PREROUTING -s 192.168.2.10 -j MARK --set-mark
101

then NAT:
$IPT -t nat -A POSTROUTING -s $FRIEND -j SNAT --to $IP_BLUE

then in sysconfig:

cat ./eth2
DEFAULT=2

cat ./eth2-2.root
# root class containing total bandwidth
RATE=320Kbit
MTU=300

cat ./eth2-2\:101.madcrock
RATE=1Kbit
MARK=101
LEAF=sfq

I try to shape myself down to 1Kbit but during upload I get 2 and more
Kb/s...
Please help me to solve this problem.

Thanks in advance,
Laurynas

On Wed, 12 May 2004 05:17:27 -0700, Laurynas Butkus wrote:

> hello,
>
> I'm running Debian Woody with 2.4.26 kernel. I have 3 NIC: eth0 -
> 192.168.0.0 subnet
> eth1 - 192.168.2.0 subnet
> eth2 - DSL link
>
> there are ~5 computers in every subnet and I want them to get equal
> incoming and outgoing traffic with possibility to borrow unused channel.
>
> I have successfully used htb.init to shape my incoming traffic but I ran
> into problems with outgoing. I try to mark outgoing packets with
> iptables, but it doesn't work...
>
> I do marking like this:
> iptables -t mangle -A PREROUTING -s 192.168.2.10 -j MARK --set-mark 101
>
> then NAT:
> $IPT -t nat -A POSTROUTING -s $FRIEND -j SNAT --to $IP_BLUE
>
> then in sysconfig:
>
> cat ./eth2
> DEFAULT=2
>
> cat ./eth2-2.root
> # root class containing total bandwidth RATE=320Kbit MTU=300
>
> cat ./eth2-2\:101.madcrock
> RATE=1Kbit
> MARK=101
> LEAF=sfq
>
> I try to shape myself down to 1Kbit but during upload I get 2 and more
> Kb/s...
> Please help me to solve this problem.
>
> Thanks in advance,
> Laurynas


I don't know about the scripts that you are using but here are a few
general things.

HTB is not recomemded for these 1Kbit stuff. For that use cbq.


Traffic control is implimented by:
a) Classes : These are the transmission classes which send the data out.



$TC qdisc add dev $EXT root handle 1:0 cbq $AVPKT $BW

$TC class add dev $EXT parent 1:0 classid 1:1 cbq rate 300kbit $ALLOT prio
5 $AV PKT $BW bounded isolated

$TC class add dev $EXT parent 1:1 classid 1:10 cbq rate 220kbit $ALLOT
prio 5 $A VPKT mpu 64 maxburst 40 $BW weight 1000kbit isolated



b) Queues: These are queues, one to EACH class that you created. If you
did not specify a particular type of queue, it will use FIFO. A class will
get it's packets from its associated queue.



$TC qdisc add dev $EXT parent 1:10 sfq perturb 10 quantum 1492

c) Filters: These are rules that identify a packet and send it to a
particular class-queue combination (called a "flowid"). "fw" tells it to
use the mark on the packet.


$TC filter add dev $EXT parent 1:0 protocol IP prio 10 handle $DEFAULT fw
flowid
1:10


Without filters, it won't work.