Skip to contentSkip to conversation

Need help with rate-limiting NTTP traffic

LARTC Shane Hickey 2 participants 3 messages Apr 1, 2004 – Apr 2, 2004

lore
Share & tools
lore
#0 lore raw

Howdy all,
I posted this message to the netfilter mailing-list and didn't get much
response. I apologize if anyone here is getting this for a
second time.
Anyway, I recently migrated my firewall from a FreeBSD box running
ipfilter, ipnat and dummynet to a Gentoo Linux box running netfilter and tc. I have to admit that I'm having problems visualizing tc in my head. So, I was wondering if I could get an assist.
Basically, when I run my NNTP client, it uses as much bandwidth as it
can get its grubby paws on. I have a 3M wireless connection and my ISP doesn't limit me, but I think they will if I'm constantly using all 3M.
So, since my NNTP traffic is pretty much constantly ongoing, I'd like
to limit it to 800kbit. This was a breeze with dummynet, but I'm not getting how to do it correctly with netfilter.

Here's what I tried:

$IPT -t mangle -N SHAPE-NNTP
$IPT -t mangle -I PREROUTING -i $WANIFACE -j SHAPE-NNTP $IPT -t mangle -A SHAPE-NNTP -p tcp --sport 119 -j MARK --set-mark 119

My thoughts on placing it in PREROUTING is that I'd like to shape the traffic as soon as possible so that my firewall gets the benefit of dealing with the reduced load as soon as possible. But, maybe that's just foolishness?

Here's the tc rules I tried.

tc qdisc add dev $WANIFACE root handle 1: htb default 60
tc class add dev $WANIFACE parent 1: classid 1:1 htb rate 10Mbit tc class add dev $WANIFACE parent 1:1 classid 1:119 htb rate 800kbit tc filter add dev $WANIFACE parent 1:1 protocol ip handle 119 fw flowid 1:119

The one weird thing is that when I do a 'tc filter show dev $WANIFACE' nothing comes back. But 'tc class show dev $WANIFACE' and 'tc qdisc show dev $WANIFACE" return useful information.

Here's some information that may be relevant:

Linux elijah 2.4.24-hardened-r1 #1 Wed Mar 31 14:20:58 MST 2004 i686 Mobile Pentium II GenuineIntel GNU/Linux iproute-20010824-r4
iptables-1.2.9

Thanks,

Signature · 7 lines
Shane Hickey <[email protected]>: Network/System Consultant
GPG KeyID: 777CBF3F
Key fingerprint: 254F B2AC 9939 C715 278C  DA95 4109 9F69 777C BF3F
Listening to: american analog set - you own me
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
#1 lore raw

Since you were already using dummynet, try using NIST NET, the linux alternative for dummynet.

Nirnimesh.

Quoted 38 lines · Shane Hickey

Howdy all,
I posted this message to the netfilter mailing-list and didn't get much
response. I apologize if anyone here is getting this for a
second time.
Anyway, I recently migrated my firewall from a FreeBSD box running
ipfilter, ipnat and dummynet to a Gentoo Linux box running netfilter and tc. I have to admit that I'm having problems visualizing tc in my head. So, I was wondering if I could get an assist.
Basically, when I run my NNTP client, it uses as much bandwidth as it
can get its grubby paws on. I have a 3M wireless connection and my ISP doesn't limit me, but I think they will if I'm constantly using all 3M.
So, since my NNTP traffic is pretty much constantly ongoing, I'd like
to limit it to 800kbit. This was a breeze with dummynet, but I'm not getting how to do it correctly with netfilter.

Here's what I tried:

$IPT -t mangle -N SHAPE-NNTP
$IPT -t mangle -I PREROUTING -i $WANIFACE -j SHAPE-NNTP $IPT -t mangle -A SHAPE-NNTP -p tcp --sport 119 -j MARK --set-mark 119

My thoughts on placing it in PREROUTING is that I'd like to shape the traffic as soon as possible so that my firewall gets the benefit of dealing with the reduced load as soon as possible. But, maybe that's just foolishness?

Here's the tc rules I tried.

tc qdisc add dev $WANIFACE root handle 1: htb default 60
tc class add dev $WANIFACE parent 1: classid 1:1 htb rate 10Mbit tc class add dev $WANIFACE parent 1:1 classid 1:119 htb rate 800kbit tc filter add dev $WANIFACE parent 1:1 protocol ip handle 119 fw flowid 1:119

The one weird thing is that when I do a 'tc filter show dev $WANIFACE' nothing comes back. But 'tc class show dev $WANIFACE' and 'tc qdisc show dev $WANIFACE" return useful information.

Quoted 8 lines

Here's some information that may be relevant:

Linux elijah 2.4.24-hardened-r1 #1 Wed Mar 31 14:20:58 MST 2004 i686 Mobile Pentium II GenuineIntel GNU/Linux iproute-20010824-r4
iptables-1.2.9

Thanks,

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#2 lore raw
In reply to Shane Hickey

I found the solution to my exact problem (right down the NNTP client) at
http://mail.gnu.org/archive/html/pan-users/2003-11/msg00009.html

For those who want the answer now with now clicking, you can do it all with this:

tc qdisc add dev $WANIFACE handle ffff: ingress
tc filter add dev $WANIFACE parent ffff: protocol ip prio 50 u32 match ip sport 119 0xffff police rate 800kbit burst 15k drop flowid :1

Thanks.

Signature · 7 lines
Shane Hickey <[email protected]>: Network/System Consultant
GPG KeyID: 777CBF3F
Key fingerprint: 254F B2AC 9939 C715 278C  DA95 4109 9F69 777C BF3F
Listening to: Bill Frisell  - Vernon Reid - Small Hands
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/