LDP
/
LDP
mirror of https://github.com/tLDP/LDP
0
1
Fork 0
Code Releases Activity
LDP/LDP/howto/docbook/Adv-Routing-HOWTO/Adv-Routing-HOWTO.sgml

9462 lines
274 KiB
Plaintext
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook V4.1//EN">
<!-- $Id$ -->
<Book id="lartc">
<bookinfo>
<Title>Linux Advanced Routing &amp; Traffic Control HOWTO</Title>
<authorgroup>
<author>
<FirstName>Bert</FirstName><Surname>Hubert</Surname>
<affiliation>
<orgname>Netherlabs BV</orgname>
<address><email>bert.hubert@netherlabs.nl</email></address>
</affiliation>
</author>
<collab>
<collabname>Gregory Maxwell</collabname>
<affiliation>
<address><email>greg@linuxpower.cx</email></address>
</affiliation>
</collab>
<collab>
<collabname>Remco van Mook</collabname>
<affiliation>
<address><email>remco@virtu.nl</email></address>
</affiliation>
</collab>
<collab>
<collabname>Martijn van Oosterhout</collabname>
<affiliation>
<address><email>kleptog@cupid.suninternet.com</email></address>
</affiliation>
</collab>
<collab>
<collabname>Paul B Schroeder</collabname>
<affiliation>
<address><email>paulsch@us.ibm.com</email></address>
</affiliation>
</collab>
<collab>
<collabname>Jasper Spaans</collabname>
<affiliation>
<address><email>jasper@spaans.ds9a.nl</email></address>
</affiliation>
</collab>
<collab>
<collabname>Leroy Tennison</collabname>
<affiliation>
<address><email>https://github.com/LTennis</email></address>
</affiliation>
</collab>
</authorgroup>
<revhistory>
<revision>
<revnumber>2.0</revnumber>
<date>2022-10-18</date>
<revremark>DocBook Edition</revremark>
</revision>
</revhistory>
<Abstract>
<Para>A very hands-on approach to <application>iproute2</application>,
traffic shaping and a bit of <application>netfilter</application>.
</para>
</Abstract>
</bookinfo>
<toc></toc>
<chapter id="lartc.dedication">
<Title>Dedication</Title>
<Para>
This document is dedicated to lots of people, and is my attempt to do
something back. To list but a few:
</Para>
<Para>
<ItemizedList>
<ListItem>
<Para>
Rusty Russell
</Para>
</ListItem>
<ListItem>
<Para>
Alexey N. Kuznetsov
</Para>
</ListItem>
<ListItem>
<Para>
The good folks from Google
</Para>
</ListItem>
<ListItem>
<Para>
The staff of Casema Internet
</Para>
</ListItem>
</ItemizedList>
</Para>
</chapter>
<chapter id="lartc.intro">
<Title>Introduction</Title>
<Para>
Welcome, gentle reader.
</Para>
<Para>
This document hopes to enlighten you on how to do more with Linux 2.2/2.4
routing. Unbeknownst to most users, you already run tools which allow you to
do spectacular things. Commands like <command>route</command> and
<command>ifconfig</command> are actually
very thin wrappers for the very powerful iproute2 infrastructure.
</Para>
<Para>
I hope that this HOWTO will become as readable as the ones by Rusty Russell
of (amongst other things) netfilter fame.
</Para>
<Para>
You can always reach us by writing to the <ULink
URL="mailto:HOWTO@ds9a.nl"
>HOWTO team</ULink
>. However, please consider posting to the mailing
list (see the relevant section) if you have questions which are not directly
related to this HOWTO. We are no free helpdesk, but we often will answer questions
asked on the list.
</Para>
<Para>
Before losing your way in this HOWTO, if all you want to do is simple
traffic shaping, skip everything and head to the <citetitle><xref linkend="lartc.other"></citetitle> chapter, and read about CBQ.init.
</Para>
<Sect1 id="lartc.intro.disclaimer">
<Title>Disclaimer &amp; License</Title>
<Para>
This document is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
</Para>
<Para>
In short, if your STM-64 backbone breaks down and distributes pornography to
your most esteemed customers - it's never our fault. Sorry.
</Para>
<Para>
Copyright (c) 2002 by bert hubert, Gregory Maxwell, Martijn van
Oosterhout, Remco van Mook, Paul B. Schroeder and others. This material may
be distributed only subject to the terms and conditions set forth in the
Open Publication License, v1.0 or later (the latest version is presently
available at http://www.opencontent.org/openpub/).
</Para>
<Para>
Please freely copy and distribute (sell or give away) this document in any
format. It's requested that corrections and/or comments be forwarded to the
document maintainer.
</Para>
<Para>
It is also requested that if you publish this HOWTO in hardcopy that you
send the authors some samples for <quote>review purposes</quote> :-)
</Para>
</Sect1>
<Sect1 id="lartc.intro.prior">
<Title>Prior knowledge</Title>
<Para>
As the title implies, this is the <quote>Advanced</quote> HOWTO.
While by no means rocket science, some prior knowledge is assumed.
</Para>
<Para>
Here are some other references which might help teach you more:
<VariableList>
<VarListEntry>
<Term>
<ULink URL="http://netfilter.samba.org/unreliable-guides/networking-concepts-HOWTO/index.html">
Rusty Russell's networking-concepts-HOWTO</ULink>
</Term>
<ListItem>
<Para>Very nice introduction, explaining what a network is, and how it is
connected to other networks.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Linux Networking-HOWTO (Previously the Net-3 HOWTO)</Term>
<ListItem>
<Para>Great stuff, although very verbose. It teaches you a lot of stuff
that's already configured if you are able to connect to the Internet.
Should be located in <filename>/usr/doc/HOWTO/NET3-4-HOWTO.txt</filename>
but can be also be found
<ULink URL="http://www.linuxports.com/howto/networking">online</ULink>.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect1>
<Sect1 id="lartc.intro.linux">
<Title>What Linux can do for you</Title>
<Para>
A small list of things that are possible:
</Para>
<ItemizedList>
<ListItem>
<Para>Throttle bandwidth for certain computers
</Para>
</ListItem>
<ListItem>
<Para>Throttle bandwidth TO certain computers
</Para>
</ListItem>
<ListItem>
<Para>Help you to fairly share your bandwidth
</Para>
</ListItem>
<ListItem>
<Para>Protect your network from DoS attacks
</Para>
</ListItem>
<ListItem>
<Para>Protect the Internet from your customers
</Para>
</ListItem>
<ListItem>
<Para>Multiplex several servers as one, for load balancing or
enhanced availability
</Para>
</ListItem>
<ListItem>
<Para>Restrict access to your computers
</Para>
</ListItem>
<ListItem>
<Para>Limit access of your users to other hosts
</Para>
</ListItem>
<ListItem>
<Para>Do routing based on user id (yes!), MAC address, source IP
address, port, type of service, time of day or content
</Para>
</ListItem>
</ItemizedList>
<Para>
Currently, not many people are using these advanced features. This is for
several reasons. While the provided documentation is verbose, it is not very
hands-on. Traffic control is almost undocumented.
</Para>
</Sect1>
<Sect1 id="lartc.intro.houskeeping">
<Title>Housekeeping notes</Title>
<Para>
There are several things which should be noted about this document. While I
wrote most of it, I really don't want it to stay that way. I am a strong
believer in Open Source, so I encourage you to send feedback, updates,
patches etcetera. Do not hesitate to inform me of typos or plain old errors.
If my English sounds somewhat wooden, please realize that I'm not a native
speaker. Feel free to send suggestions.
</Para>
<Para>
If you feel to you are better qualified to maintain a section, or think that
you can author and maintain new sections, you are welcome to do so. The SGML
of this HOWTO is available via CVS, I very much envision more people
working on it.
</Para>
<Para>
In aid of this, you will find lots of FIXME notices. Patches are always
welcome! Wherever you find a FIXME, you should know that you are treading in
unknown territory. This is not to say that there are no errors elsewhere,
but be extra careful. If you have validated something, please let us know so
we can remove the FIXME notice.
</Para>
<Para>
About this HOWTO, I will take some liberties along the road. For example, I
postulate a 10Mbit Internet connection, while I know full well that those
are not very common.
</Para>
</Sect1>
<Sect1 id="lartc.intro.cvs">
<Title>Access, CVS &amp; submitting updates</Title>
<Para>
The canonical location for the HOWTO is
<ULink URL="http://www.ds9a.nl/lartc">here</ULink>.
</Para>
<Para>
We now have anonymous CVS access available to the world at large. This is
good in a number of ways. You can easily upgrade to newer versions of this
HOWTO and submitting patches is no work at all.
</Para>
<Para>
Furthermore, it allows the authors to work on the source independently,
which is good too.
</Para>
<Screen width="80">
$ export CVSROOT=:pserver:anon@outpost.ds9a.nl:/var/cvsroot
$ cvs login
CVS password: [enter 'cvs' (without 's)]
$ cvs co 2.4routing
cvs server: Updating 2.4routing
U 2.4routing/2.4routing.sgml
</Screen>
<Para>
If you spot an error, or want to add something, just fix it locally, and run
<userinput>cvs diff -u</userinput>, and send the result off to us.
</Para>
<Para>
A Makefile is supplied which should help you create postscript, dvi, pdf,
html and plain text. You may need to install
<application>docbook</application>, <application>docbook-utils</application>,
<application>ghostscript</application> and <application>tetex</application>
to get all formats.
</Para>
</Sect1>
<Sect1 id="lartc.intro.mlist">
<Title>Mailing list</Title>
<Para>
The authors receive an increasing amount of mail about this HOWTO. Because
of the clear interest of the community, it has been decided to start a
mailinglist where people can talk to each other about Advanced Routing and
Traffic Control. You can subscribe to the list
<ULink URL="http://mailman.ds9a.nl/mailman/listinfo/lartc">here</ULink>.
</Para>
<Para>
It should be pointed out that the authors are very hesitant of answering
questions not asked on the list. We would like the archive of the list to
become some kind of knowledge base. If you have a question, please search
the archive, and then post to the mailinglist.
</Para>
</Sect1>
<Sect1 id="lartc.intro.layout">
<Title>Layout of this document</Title>
<Para>
We will be doing interesting stuff almost immediately, which also means that
there will initially be parts that are explained incompletely or are not
perfect. Please gloss over these parts and assume that all will become clear.
</Para>
<Para>
Routing and filtering are two distinct things. Filtering is documented very
well by Rusty's HOWTOs, available here:
</Para>
<ItemizedList>
<ListItem>
<Para><ULink URL="http://netfilter.samba.org/unreliable-guides/">
Rusty's Remarkably Unreliable Guides</ULink>
</Para>
</ListItem>
</ItemizedList>
<Para>We will be focusing mostly on what is possible by combining netfilter
and iproute2.
</Para>
</Sect1>
</chapter>
<chapter id="lartc.iproute2">
<Title>Introduction to iproute2</Title>
<Sect1 id="lartc.iproute2.why">
<Title>Why iproute2?</Title>
<Para>
Most Linux distributions, and most UNIX's, currently use the
venerable <command>arp</command>, <command>ifconfig</command> and
<command>route</command> commands.
While these tools work, they show some unexpected behaviour under Linux 2.2
and up.
For example, GRE tunnels are an integral part of routing these days, but
require completely different tools.
</Para>
<Para>
With <application>iproute2</application>, tunnels are an integral part of
the tool set.
</Para>
<Para>
The 2.2 and above Linux kernels include a completely redesigned network
subsystem. This new networking code brings Linux performance and a feature
set with little competition in the general OS arena. In fact, the new
routing, filtering, and classifying code is more featureful than the one
provided by many dedicated routers and firewalls and traffic shaping
products.
</Para>
<Para>
As new networking concepts have been invented, people have found ways to
plaster them on top of the existing framework in existing OSes. This
constant layering of cruft has lead to networking code that is filled with
strange behaviour, much like most human languages. In the past, Linux
emulated SunOS's handling of many of these things, which was not ideal.
</Para>
<Para>
This new framework makes it possible to clearly express features
previously beyond Linux's reach.
</Para>
</Sect1>
<Sect1 id="lartc.iproute2.tour">
<Title>iproute2 tour</Title>
<Para>
Linux has a sophisticated system for bandwidth provisioning called Traffic
Control. This system supports various method for classifying, prioritizing,
sharing, and limiting both inbound and outbound traffic.
</Para>
<Para>
We'll start off with a tiny tour of iproute2 possibilities.
</Para>
</Sect1>
<Sect1 id="lartc.iproute2.package">
<Title>Prerequisites</Title>
<Para>
You should make sure that you have the userland tools installed. This
package is called 'iproute' on both RedHat and Debian, and may otherwise be
found at <filename>ftp://ftp.inr.ac.ru/ip-routing/iproute2-2.2.4-now-ss??????.tar.gz"</filename>.
</Para>
<Para>
You can also try
<ULink URL="ftp://ftp.inr.ac.ru/ip-routing/iproute2-current.tar.gz">here</ULink>
for the latest version.
</Para>
<Para>
Some parts of iproute require you to have certain kernel options enabled. It
should also be noted that all releases of RedHat up to and including 6.2
come without most of the traffic control features in the default kernel.
</Para>
<Para>
RedHat 7.2 has everything in by default.
</Para>
<Para>
Also make sure that you have netlink support, should you choose to roll your
own kernel. Iproute2 needs it.
</Para>
</Sect1>
<Sect1 id="lartc.iproute2.explore">
<Title>Exploring your current configuration</Title>
<Para>
This may come as a surprise, but iproute2 is already configured! The current
commands <command>ifconfig</command> and <command>route</command> are already using the advanced
syscalls, but mostly with very default (ie. boring) settings.
</Para>
<Para>
The <command>ip</command> tool is central, and we'll ask it to display our interfaces
for us.
</Para>
<Sect2>
<Title><command>ip</command> shows us our links</Title>
<Screen width="80">
[ahu@home ahu]$ ip link list
1: lo: &#60;LOOPBACK,UP&#62; mtu 3924 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy: &#60;BROADCAST,NOARP&#62; mtu 1500 qdisc noop
link/ether 00:00:00:00:00:00 brd ff:ff:ff:ff:ff:ff
3: eth0: &#60;BROADCAST,MULTICAST,PROMISC,UP&#62; mtu 1400 qdisc pfifo_fast qlen 100
link/ether 48:54:e8:2a:47:16 brd ff:ff:ff:ff:ff:ff
4: eth1: &#60;BROADCAST,MULTICAST,PROMISC,UP&#62; mtu 1500 qdisc pfifo_fast qlen 100
link/ether 00:e0:4c:39:24:78 brd ff:ff:ff:ff:ff:ff
3764: ppp0: &#60;POINTOPOINT,MULTICAST,NOARP,UP&#62; mtu 1492 qdisc pfifo_fast qlen 10
link/ppp
</Screen>
<Para>
Your mileage may vary, but this is what it shows on my NAT router at
home. I'll only explain part of the output as not everything is directly
relevant.
</Para>
<Para>
We first see the loopback interface. While your computer may function
somewhat without one, I'd advise against it. The MTU size (Maximum Transfer
Unit) is 3924 octets, and it is not supposed to queue. Which makes sense
because the loopback interface is a figment of your kernel's imagination.
</Para>
<Para>
I'll skip the dummy interface for now, and it may not be present on your
computer. Then there are my two physical network interfaces, one at the side
of my cable modem, the other one serves my home ethernet segment.
Furthermore, we see a ppp0 interface.
</Para>
<Para>
Note the absence of IP addresses. iproute disconnects the concept of 'links'
and 'IP addresses'. With IP aliasing, the concept of 'the' IP address had
become quite irrelevant anyhow.
</Para>
<Para>
It does show us the MAC addresses though, the hardware identifier of our
ethernet interfaces.
</Para>
</Sect2>
<Sect2>
<Title><command>ip</command> shows us our IP addresses</Title>
<Screen width="80">
[ahu@home ahu]$ ip address show
1: lo: &#60;LOOPBACK,UP&#62; mtu 3924 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 brd 127.255.255.255 scope host lo
2: dummy: &#60;BROADCAST,NOARP&#62; mtu 1500 qdisc noop
link/ether 00:00:00:00:00:00 brd ff:ff:ff:ff:ff:ff
3: eth0: &#60;BROADCAST,MULTICAST,PROMISC,UP&#62; mtu 1400 qdisc pfifo_fast qlen 100
link/ether 48:54:e8:2a:47:16 brd ff:ff:ff:ff:ff:ff
inet 10.0.0.1/8 brd 10.255.255.255 scope global eth0
4: eth1: &#60;BROADCAST,MULTICAST,PROMISC,UP&#62; mtu 1500 qdisc pfifo_fast qlen 100
link/ether 00:e0:4c:39:24:78 brd ff:ff:ff:ff:ff:ff
3764: ppp0: &#60;POINTOPOINT,MULTICAST,NOARP,UP&#62; mtu 1492 qdisc pfifo_fast qlen 10
link/ppp
inet 212.64.94.251 peer 212.64.94.1/32 scope global ppp0
</Screen>
<Para>
This contains more information. It shows all our addresses, and to which
cards they belong. 'inet' stands for Internet (IPv4). There are lots of other
address families, but these don't concern us right now.
</Para>
<Para>
Let's examine eth0 somewhat closer. It says that it is related to the inet
address '10.0.0.1/8'. What does this mean? The /8 stands for the number of
bits that are in the Network Address. There are 32 bits, so we have 24 bits
left that are part of our network. The first 8 bits of 10.0.0.1 correspond
to 10.0.0.0, our Network Address, and our netmask is 255.0.0.0.
</Para>
<Para>
The other bits are connected to this interface, so 10.250.3.13 is directly
available on eth0, as is 10.0.0.1 for example.
</Para>
<Para>
With ppp0, the same concept goes, though the numbers are different. Its
address is 212.64.94.251, without a subnet mask. This means that we have a
point-to-point connection and that every address, with the exception of
212.64.94.251, is remote. There is more information, however. It tells us
that on the other side of the link there is, yet again, only one address,
212.64.94.1. The /32 tells us that there are no 'network bits'.
</Para>
<Para>
It is absolutely vital that you grasp these concepts. Refer to the
documentation mentioned at the beginning of this HOWTO if you have trouble.
</Para>
<Para>
You may also note 'qdisc', which stands for Queueing Discipline. This will
become vital later on.
</Para>
</Sect2>
<Sect2>
<Title><command>ip</command> shows us our routes</Title>
<Para>
Well, we now know how to find 10.x.y.z addresses, and we are able to reach
212.64.94.1. This is not enough however, so we need instructions on how to
reach the world. The Internet is available via our ppp connection, and it
appears that 212.64.94.1 is willing to spread our packets around the
world, and deliver results back to us.
</Para>
<Screen width="80">
[ahu@home ahu]$ ip route show
212.64.94.1 dev ppp0 proto kernel scope link src 212.64.94.251
10.0.0.0/8 dev eth0 proto kernel scope link src 10.0.0.1
127.0.0.0/8 dev lo scope link
default via 212.64.94.1 dev ppp0
</Screen>
<Para>
This is pretty much self explanatory. The first 4 lines of output explicitly
state what was already implied by <command>ip address show</command>, the last line
tells us that the rest of the world can be found via 212.64.94.1, our
default gateway. We can see that it is a gateway because of the word
via, which tells us that we need to send packets to 212.64.94.1, and that it
will take care of things.
</Para>
<Para>
For reference, this is what the old <command>route</command> utility shows us:
</Para>
<Screen width="80">
[ahu@home ahu]$ route -n
Kernel IP routing table
Destination Gateway Genmask Flags Metric Ref Use
Iface
212.64.94.1 0.0.0.0 255.255.255.255 UH 0 0 0 ppp0
10.0.0.0 0.0.0.0 255.0.0.0 U 0 0 0 eth0
127.0.0.0 0.0.0.0 255.0.0.0 U 0 0 0 lo
0.0.0.0 212.64.94.1 0.0.0.0 UG 0 0 0 ppp0
</Screen>
</Sect2>
</Sect1>
<Sect1 id="lartc.iproute2.arp">
<Title>ARP</Title>
<Para>
ARP is the Address Resolution Protocol as described in
<ULink URL="http://www.faqs.org/rfcs/rfc826.html">RFC 826</ULink>.
ARP is used by a networked machine to resolve the hardware location/address of
another machine on the same
local network. Machines on the Internet are generally known by their names
which resolve to IP
addresses. This is how a machine on the foo.com network is able to communicate
with another machine which is on the bar.net network. An IP address, though,
cannot tell you the physical location of a machine. This is where ARP comes
into the picture.
</Para>
<Para>
Let's take a very simple example. Suppose I have a network composed of several
machines. Two of the machines which are currently on my network are foo
with an IP address of 10.0.0.1 and bar with an IP address of 10.0.0.2.
Now foo wants to ping bar to see that he is alive, but alas, foo has no idea
where bar is. So when foo decides to ping bar he will need to send
out an ARP request.
This ARP request is akin to foo shouting out on the network "Bar (10.0.0.2)!
Where are you?" As a result of this every machine on the network will hear
foo shouting, but only bar (10.0.0.2) will respond. Bar will then send an
ARP reply directly back to foo which is akin
bar saying,
"Foo (10.0.0.1) I am here at 00:60:94:E9:08:12." After this simple transaction
that's used to locate his friend on the network, foo is able to communicate
with bar until he (his arp cache) forgets where bar is (typically after
15 minutes on Unix).
</Para>
<Para>
Now let's see how this works.
You can view your machines current arp/neighbor cache/table like so:
</Para>
<Screen width="80">
[root@espa041 /home/src/iputils]# ip neigh show
9.3.76.42 dev eth0 lladdr 00:60:08:3f:e9:f9 nud reachable
9.3.76.1 dev eth0 lladdr 00:06:29:21:73:c8 nud reachable
</Screen>
<Para>
As you can see my machine espa041 (9.3.76.41) knows where to find espa042
(9.3.76.42) and
espagate (9.3.76.1). Now let's add another machine to the arp cache.
</Para>
<Screen width="80">
[root@espa041 /home/paulsch/.gnome-desktop]# ping -c 1 espa043
PING espa043.austin.ibm.com (9.3.76.43) from 9.3.76.41 : 56(84) bytes of data.
64 bytes from 9.3.76.43: icmp_seq=0 ttl=255 time=0.9 ms
--- espa043.austin.ibm.com ping statistics ---
1 packets transmitted, 1 packets received, 0% packet loss
round-trip min/avg/max = 0.9/0.9/0.9 ms
[root@espa041 /home/src/iputils]# ip neigh show
9.3.76.43 dev eth0 lladdr 00:06:29:21:80:20 nud reachable
9.3.76.42 dev eth0 lladdr 00:60:08:3f:e9:f9 nud reachable
9.3.76.1 dev eth0 lladdr 00:06:29:21:73:c8 nud reachable
</Screen>
<Para>
As a result of espa041 trying to contact espa043, espa043's hardware
address/location has now been added to the arp/neighbor cache.
So until the entry for
espa043 times out (as a result of no communication between the two) espa041
knows where to find espa043 and has no need to send an ARP request.
</Para>
<Para>
Now let's delete espa043 from our arp cache:
</Para>
<Screen width="80">
[root@espa041 /home/src/iputils]# ip neigh delete 9.3.76.43 dev eth0
[root@espa041 /home/src/iputils]# ip neigh show
9.3.76.43 dev eth0 nud failed
9.3.76.42 dev eth0 lladdr 00:60:08:3f:e9:f9 nud reachable
9.3.76.1 dev eth0 lladdr 00:06:29:21:73:c8 nud stale
</Screen>
<Para>
Now espa041 has again forgotten where to find espa043 and will need to send
another ARP request the next time he needs to communicate with espa043.
You can also see from the above output that espagate (9.3.76.1) has been
changed to the "stale" state. This means that the location shown is still
valid, but it will have to be confirmed at the first transaction to that
machine.
</Para>
</Sect1>
</chapter>
<chapter id="lartc.rpdb">
<Title>Rules - routing policy database</Title>
<Para>
If you have a large router, you may well cater for the needs of different
people, who should be served differently. The routing policy database allows
you to do this by having multiple sets of routing tables.
</Para>
<Para>
If you want to use this feature, make sure that your kernel is compiled with
the "IP: advanced router" and "IP: policy routing" features.
</Para>
<Para>
When the kernel needs to make a routing decision, it finds out which table
needs to be consulted. By default, there are three tables. The old 'route'
tool modifies the main and local tables, as does the ip tool (by default).
</Para>
<Para>The default rules:
</Para>
<Screen width="80">
[ahu@home ahu]$ ip rule list
0: from all lookup local
32766: from all lookup main
32767: from all lookup default
</Screen>
<Para>
This lists the priority of all rules. We see that all rules apply to all
packets ('from all'). We've seen the 'main' table before, it is output by
<userinput>ip route ls</userinput>, but the 'local' and 'default' table are new.
</Para>
<Para>
If we want to do fancy things, we generate rules which point to different
tables which allow us to override system wide routing rules.
</Para>
<Para>
For the exact semantics on what the kernel does when there are more matching
rules, see Alexey's ip-cref documentation.
</Para>
<Sect1 id="lartc.rpdb.simple">
<Title>Simple source policy routing</Title>
<Para>
Let's take a real example once again, I have 2 (actually 3, about time I
returned them) cable modems, connected to a Linux NAT ('masquerading')
router. People living here pay me to use the Internet. Suppose one of my
house mates only visits hotmail and wants to pay less. This is fine with me,
but they'll end up using the low-end cable modem.
</Para>
<Para>
The 'fast' cable modem is known as 212.64.94.251 and is a PPP link to
212.64.94.1. The 'slow' cable modem is known by various ip addresses,
212.64.78.148 in this example and is a link to 195.96.98.253.
</Para>
<Para>The local table:
</Para>
<Screen width="80">
[ahu@home ahu]$ ip route list table local
broadcast 127.255.255.255 dev lo proto kernel scope link src 127.0.0.1
local 10.0.0.1 dev eth0 proto kernel scope host src 10.0.0.1
broadcast 10.0.0.0 dev eth0 proto kernel scope link src 10.0.0.1
local 212.64.94.251 dev ppp0 proto kernel scope host src 212.64.94.251
broadcast 10.255.255.255 dev eth0 proto kernel scope link src 10.0.0.1
broadcast 127.0.0.0 dev lo proto kernel scope link src 127.0.0.1
local 212.64.78.148 dev ppp2 proto kernel scope host src 212.64.78.148
local 127.0.0.1 dev lo proto kernel scope host src 127.0.0.1
local 127.0.0.0/8 dev lo proto kernel scope host src 127.0.0.1
</Screen>
<Para>
Lots of obvious things, but things that need to be specified somewhere.
Well, here they are. The default table is empty.
</Para>
<Para>Let's view the 'main' table:
</Para>
<Screen width="80">
[ahu@home ahu]$ ip route list table main
195.96.98.253 dev ppp2 proto kernel scope link src 212.64.78.148
212.64.94.1 dev ppp0 proto kernel scope link src 212.64.94.251
10.0.0.0/8 dev eth0 proto kernel scope link src 10.0.0.1
127.0.0.0/8 dev lo scope link
default via 212.64.94.1 dev ppp0
</Screen>
<Para>
We now generate a new rule which we call 'John', for our hypothetical
house mate. Although we can work with pure numbers, it's far easier if we add
our tables to /etc/iproute2/rt_tables.
</Para>
<Screen width="80">
# echo 200 John &#62;&#62; /etc/iproute2/rt_tables
# ip rule add from 10.0.0.10 table John
# ip rule ls
0: from all lookup local
32765: from 10.0.0.10 lookup John
32766: from all lookup main
32767: from all lookup default
</Screen>
<Para>
Now all that is left is to generate John's table, and flush the route cache:
</Para>
<Screen width="80">
# ip route add default via 195.96.98.253 dev ppp2 table John
# ip route flush cache
</Screen>
<Para>
And we are done. It is left as an exercise for the reader to implement this
in ip-up.
</Para>
</Sect1>
<sect1 id="lartc.rpdb.multiple-links">
<title>Routing for multiple uplinks/providers</title>
<para>
A common configuration is the following, in which there are two providers
that connect a local network (or even a single machine) to the big Internet.
<screen>
________
+------------+ /
| | |
+-------------+ Provider 1 +-------
__ | | | /
___/ \_ +------+-------+ +------------+ |
_/ \__ | if1 | /
/ \ | | |
| Local network -----+ Linux router | | Internet
\_ __/ | | |
\__ __/ | if2 | \
\___/ +------+-------+ +------------+ |
| | | \
+-------------+ Provider 2 +-------
| | |
+------------+ \________
</screen>
There are usually two questions given this setup.
</para>
<sect2><title>Split access</title>
<para>
The first is how to route answers to packets coming in over a
particular provider, say Provider 1, back out again over that same provider.
</para>
<para>
Let us first set some symbolical names. Let <command>$IF1</command> be the name of the
first interface (if1 in the picture above) and <command>$IF2</command> the name of the
second interface. Then let <command>$IP1</command> be the IP address associated with
<command>$IF1</command> and <command>$IP2</command> the IP address associated with
<command>$IF2</command>. Next, let <command>$P1</command> be the IP address of the gateway at
Provider 1, and <command>$P2</command> the IP address of the gateway at provider 2.
Finally, let <command>$P1_NET</command> be the IP network <command>$P1</command> is in,
and <command>$P2_NET</command> the IP network <command>$P2</command> is in.
</para>
<para>
One creates two additional routing tables, say <command>T1</command> and <command>T2</command>.
These are added in /etc/iproute2/rt_tables. Then you set up routing in
these tables as follows:
</para>
<para>
<screen>
ip route add $P1_NET dev $IF1 src $IP1 table T1
ip route add default via $P1 table T1
ip route add $P2_NET dev $IF2 src $IP2 table T2
ip route add default via $P2 table T2
</screen>
Nothing spectacular, just build a route to the gateway and build a
default route via that gateway, as you would do in the case of a single
upstream provider, but put the routes in a separate table per provider.
Note that the network route suffices, as it tells you how to find any host
in that network, which includes the gateway, as specified above.
</para>
<para>
Next you set up the main routing table. It is a good idea to route
things to the direct neighbour through the interface connected to that
neighbour. Note the `src' arguments, they make sure the right outgoing IP
address is chosen.
<screen>
ip route add $P1_NET dev $IF1 src $IP1
ip route add $P2_NET dev $IF2 src $IP2
</screen>
Then, your preference for default route:
<screen>
ip route add default via $P1
</screen>
Next, you set up the routing rules. These actually choose what routing table
to route with. You want to make sure that you route out a given
interface if you already have the corresponding source address:
<screen>
ip rule add from $IP1 table T1
ip rule add from $IP2 table T2
</screen>
This set of commands makes sure all answers to traffic coming in on a
particular interface get answered from that interface.
</para>
<para>
Now, this is just the very basic setup. It will work for all processes
running on the router itself, and for the local network, if it is
masqueraded. If it is not, then you either have IP space from both providers
or you are going to want to masquerade to one of the two providers. In both
cases you will want to add rules selecting which provider to route out from
based on the IP address of the machine in the local network.
</para>
</sect2>
<sect2><title>Load balancing</title>
<para>
The second question is how to balance traffic going out over the two providers.
This is actually not hard if you already have set up split access as above.
</para>
<para>
Instead of choosing one of the two providers as your default route,
you now set up the default route to be a multipath route. In the default
kernel this will balance routes over the two providers. It is done
as follows (once more building on the example in the section on
split-access):
<screen>
ip route add default scope global nexthop via $P1 dev $IF1 weight 1 \
nexthop via $P2 dev $IF2 weight 1
</screen>
This will balance the routes over both providers. The <command>weight</command>
parameters can be tweaked to favor one provider over the other.
</para>
<para>
Note that balancing will not be perfect, as it is route based, and routes
are cached. This means that routes to often-used sites will always
be over the same provider.
</para>
<para>
Furthermore, if you really want to do this, you probably also want to look
at Julian Anastasov's patches at <ulink
url="http://www.linuxvirtualserver.org/~julian/#routes">http://www.linuxvirtualserver.org/~julian/#routes
</ulink>, Julian's route patch page. They will make things nicer to work with.
</para>
</sect2>
</sect1>
</chapter>
<chapter id="lartc.tunnel">
<Title>GRE and other tunnels</Title>
<Para>
There are 3 kinds of tunnels in Linux. There's IP in IP tunneling, GRE tunneling and tunnels that live outside the kernel (like, for example PPTP).
</Para>
<Sect1 id="lartc.tunnel.remarks">
<Title>A few general remarks about tunnels:</Title>
<Para>
Tunnels can be used to do some very unusual and very cool stuff. They can
also make things go horribly wrong when you don't configure them right.
Don't point your default route to a tunnel device unless you know
<Emphasis>EXACTLY</Emphasis> what you are doing :-). Furthermore, tunneling increases
overhead, because it needs an extra set of IP headers. Typically this is 20
bytes per packet, so if the normal packet size (MTU) on a network is 1500
bytes, a packet that is sent through a tunnel can only be 1480 bytes big.
This is not necessarily a problem, but be sure to read up on IP packet
fragmentation/reassembly when you plan to connect large networks with
tunnels. Oh, and of course, the fastest way to dig a tunnel is to dig at
both sides.
</Para>
</Sect1>
<Sect1 id="lartc.tunnel.ip-ip">
<Title>IP in IP tunneling</Title>
<Para>
This kind of tunneling has been available in Linux for a long time. It requires 2 kernel modules,
ipip.o and new_tunnel.o.
</Para>
<Para>
Let's say you have 3 networks: Internal networks A and B, and intermediate network C (or let's say, Internet).
So we have network A:
</Para>
<Screen width="80">
network 10.0.1.0
netmask 255.255.255.0
router 10.0.1.1
</Screen>
<Para>The router has address 172.16.17.18 on network C.
</Para>
<Para>and network B:
</Para>
<Screen width="80">
network 10.0.2.0
netmask 255.255.255.0
router 10.0.2.1
</Screen>
<Para>The router has address 172.19.20.21 on network C.
</Para>
<Para>
As far as network C is concerned, we assume that it will pass any packet sent
from A to B and vice versa. You might even use the Internet for this.
</Para>
<Para>Here's what you do:
</Para>
<Para>First, make sure the modules are installed:
</Para>
<Screen width="80">
insmod ipip.o
insmod new_tunnel.o
</Screen>
<Para>Then, on the router of network A, you do the following:
</Para>
<Screen width="80">
ifconfig tunl0 10.0.1.1 pointopoint 172.19.20.21
route add -net 10.0.2.0 netmask 255.255.255.0 dev tunl0
</Screen>
<Para>And on the router of network B:
</Para>
<Screen width="80">
ifconfig tunl0 10.0.2.1 pointopoint 172.16.17.18
route add -net 10.0.1.0 netmask 255.255.255.0 dev tunl0
</Screen>
<Para>And if you're finished with your tunnel:
</Para>
<Screen width="80">
ifconfig tunl0 down
</Screen>
<Para>Presto, you're done. You can't forward broadcast or IPv6 traffic through
an IP-in-IP tunnel, though. You just connect 2 IPv4 networks that normally wouldn't be able to talk to each other, that's all. As far as compatibility goes, this code has been around a long time, so it's compatible all the way back to 1.3 kernels. Linux IP-in-IP tunneling doesn't work with other Operating Systems or routers, as far as I know. It's simple, it works. Use it if you have to, otherwise use GRE.
</Para>
</Sect1>
<Sect1 id="lartc.tunnel.gre">
<Title>GRE tunneling</Title>
<Para>
GRE is a tunneling protocol that was originally developed by Cisco, and it
can do a few more things than IP-in-IP tunneling. For example, you can also
transport multicast traffic and IPv6 through a GRE tunnel.
</Para>
<Para>
In Linux, you'll need the ip_gre.o module.
</Para>
<Sect2>
<Title>IPv4 Tunneling</Title>
<Para>
Let's do IPv4 tunneling first:
</Para>
<Para>
Let's say you have 3 networks: Internal networks A and B, and intermediate network C (or let's say, Internet).
</Para>
<Para>
So we have network A:
<Screen width="80">
network 10.0.1.0
netmask 255.255.255.0
router 10.0.1.1
</Screen>
The router has address 172.16.17.18 on network C.
Let's call this network neta (ok, hardly original)
</Para>
<Para>
and network B:
<Screen width="80">
network 10.0.2.0
netmask 255.255.255.0
router 10.0.2.1
</Screen>
The router has address 172.19.20.21 on network C.
Let's call this network netb (still not original)
</Para>
<Para>
As far as network C is concerned, we assume that it will pass any packet sent
from A to B and vice versa. How and why, we do not care.
</Para>
<Para>On the router of network A, you do the following:
</Para>
<Screen width="80">
ip tunnel add netb mode gre remote 172.19.20.21 local 172.16.17.18 ttl 255
ip link set netb up
ip addr add 10.0.1.1 dev netb
ip route add 10.0.2.0/24 dev netb
</Screen>
<Para>
Let's discuss this for a bit. In line 1, we added a tunnel device, and
called it netb (which is kind of obvious because that's where we want it to
go). Furthermore we told it to use the GRE protocol (mode gre), that the
remote address is 172.19.20.21 (the router at the other end), that our
tunneling packets should originate from 172.16.17.18 (which allows your
router to have several IP addresses on network C and let you decide which
one to use for tunneling) and that the TTL field of the packet should be set
to 255 (ttl 255).
</Para>
<Para>
The second line enables the device.
</Para>
<Para>
In the third line we gave the newly born interface netb the address
10.0.1.1. This is OK for smaller networks, but when you're starting up a
mining expedition (LOTS of tunnels), you might want to consider using
another IP range for tunneling interfaces (in this example, you could use
10.0.3.0).
</Para>
<Para>
In the fourth line we set the route for network B. Note the different notation for the netmask. If you're not familiar with this notation, here's how it works: you write out the netmask in binary form, and you count all the ones. If you don't know how to do that, just remember that 255.0.0.0 is /8, 255.255.0.0 is /16 and 255.255.255.0 is /24. Oh, and 255.255.254.0 is /23, in case you were wondering.
</Para>
<Para>
But enough about this, let's go on with the router of network B.
<Screen>
ip tunnel add neta mode gre remote 172.16.17.18 local 172.19.20.21 ttl 255
ip link set neta up
ip addr add 10.0.2.1 dev neta
ip route add 10.0.1.0/24 dev neta
</Screen>
And when you want to remove the tunnel on router A:
<Screen>
ip link set netb down
ip tunnel del netb
</Screen>
Of course, you can replace netb with neta for router B.
</Para>
</Sect2>
<Sect2>
<Title>IPv6 Tunneling</Title>
<Para>
See Section 6 for a short bit about IPv6 Addresses.
</Para>
<Para>
On with the tunnels.
</Para>
<Para>
Let's assume that you have the following IPv6 network, and you want to connect it to 6bone, or a friend.
</Para>
<Para>
<Screen>
Network 3ffe:406:5:1:5:a:2:1/96
</Screen>
Your IPv4 address is 172.16.17.18, and the 6bone router has IPv4 address 172.22.23.24.
</Para>
<Para>
<Screen>
ip tunnel add sixbone mode sit remote 172.22.23.24 local 172.16.17.18 ttl 255
ip link set sixbone up
ip addr add 3ffe:406:5:1:5:a:2:1/96 dev sixbone
ip route add 3ffe::/15 dev sixbone
</Screen>
</Para>
<Para>
Let's discuss this. In the first line, we created a tunnel device called sixbone. We gave it mode sit (which is IPv6 in IPv4 tunneling) and told it where to go to (remote) and where to come from (local). TTL is set to maximum, 255. Next, we made the device active (up). After that, we added our own network address, and set a route for 3ffe::/15 (which is currently all of 6bone) through the tunnel.
</Para>
<Para>
GRE tunnels are currently the preferred type of tunneling. It's a standard that is also widely adopted outside the Linux community and therefore a Good Thing.
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.tunnel.userland">
<Title>Userland tunnels</Title>
<Para>
There are literally dozens of implementations of tunneling outside the kernel. Best known are of course PPP and PPTP, but there are lots more (some proprietary, some secure, some that don't even use IP) and that is really beyond the scope of this HOWTO.
</Para>
</Sect1>
</chapter>
<chapter id="lartc.ipv6-tunnel">
<Title>IPv6 tunneling with Cisco and/or 6bone</Title>
<Para>
By Marco Davids &lt;marco@sara.nl&gt;
</Para>
<Para>
NOTE to maintainer:
</Para>
<Para>
As far as I am concerned, this IPv6-IPv4 tunneling is not per definition
GRE tunneling. You could tunnel IPv6 over IPv4 by means of GRE tunnel devices
(GRE tunnels ANY to IPv4), but the device used here ("sit") only tunnels
IPv6 over IPv4 and is therefore something different.
</Para>
<Sect1 id="lartc.tunnel-ipv6.addressing">
<Title>IPv6 Tunneling</Title>
<Para>
This is another application of the tunneling capabilities of Linux. It is
popular among the IPv6 early adopters, or pioneers if you like.
The 'hands-on' example described below is certainly not the only way
to do IPv6 tunneling. However, it is the method that is often used to tunnel
between Linux and a Cisco IPv6 capable router and experience tells us that
this is just the thing many people are after. Ten to one this applies to
you too ;-)
</Para>
<Para>
A short bit about IPv6 addresses:
</Para>
<Para>
IPv6 addresses are, compared to IPv4 addresses, really big: 128 bits
against 32 bits. And this provides us just with the thing we need: many, many
IP-addresses: 340,282,266,920,938,463,463,374,607,431,768,211,465 to be
precise. Apart from this, IPv6 (or IPng, for IP Next Generation) is supposed
to provide for smaller routing tables on the Internet's backbone routers,
simpler configuration of equipment, better security at the IP level and
better support for QoS.
</Para>
<Para>
An example: 2002:836b:9820:0000:0000:0000:836b:9886
</Para>
<Para>
Writing down IPv6 addresses can be quite a burden. Therefore, to make
life easier there are some rules:
</Para>
<Para>
<ItemizedList>
<ListItem>
<Para>
Don't use leading zeroes. Same as in IPv4.
</Para>
</ListItem>
<ListItem>
<Para>
Use colons to separate every 16 bits or two bytes.
</Para>
</ListItem>
<ListItem>
<Para>
When you have lots of consecutive zeroes,
you can write this down as ::. You can only do this once in an
address and only for quantities of 16 bits, though.
</Para>
</ListItem>
</ItemizedList>
</Para>
<Para>
The address 2002:836b:9820:0000:0000:0000:836b:9886 can be written down
as 2002:836b:9820::836b:9886, which is somewhat friendlier.
</Para>
<Para>
Another example, the address 3ffe:0000:0000:0000:0000:0020:34A1:F32C can be
written down as 3ffe::20:34A1:F32C, which is a lot shorter.
</Para>
<Para>
IPv6 is intended to be the successor of the current IPv4. Because it
is relatively new technology, there is no worldwide native IPv6 network
yet. To be able to move forward swiftly, the 6bone was introduced.
</Para>
<Para>
Native IPv6 networks are connected to each other by encapsulating the IPv6
protocol in IPv4 packets and sending them over the existing IPv4 infrastructure
from one IPv6 site to another.
</Para>
<Para>
That is precisely where the tunnel steps in.
</Para>
<Para>
To be able to use IPv6, we should have a kernel that supports it. There
are many good documents on how to achieve this. But it all comes down to
a few steps:
<ItemizedList>
<ListItem>
<Para>
Get yourself a recent Linux distribution, with suitable glibc.
</Para>
</ListItem>
<ListItem>
<Para>
Then get yourself an up-to-date kernel source.
</Para>
</ListItem>
</ItemizedList>
If you are all set, then you can go ahead and compile an IPv6 capable
kernel:
<ItemizedList>
<ListItem>
<Para>
Go to /usr/src/linux and type:
</Para>
</ListItem>
<ListItem>
<Para>
make menuconfig
</Para>
</ListItem>
<ListItem>
<Para>
Choose "Networking Options"
</Para>
</ListItem>
<ListItem>
<Para>
Select "The IPv6 protocol", "IPv6: enable EUI-64 token format", "IPv6:
disable provider based addresses"
</Para>
</ListItem>
</ItemizedList>
HINT: Don't go for the 'module' option. Often this won't work well.
</Para>
<Para>
In other words, compile IPv6 as 'built-in' in your kernel.
You can then save your config like usual and go ahead with compiling
the kernel.
</Para>
<Para>
HINT: Before doing so, consider editing the Makefile:
EXTRAVERSION = -x ; --&#62; ; EXTRAVERSION = -x-IPv6
</Para>
<Para>
There is a lot of good documentation about compiling and installing
a kernel, however this document is about something else. If you run into
problems at this stage, go and look for documentation about compiling a
Linux kernel according to your own specifications.
</Para>
<Para>
The file /usr/src/linux/README might be a good start.
After you accomplished all this, and rebooted with your brand new kernel,
you might want to issue an '/sbin/ifconfig -a' and notice the brand
new 'sit0-device'. SIT stands for Simple Internet Transition. You may give
yourself a compliment; you are now one major step closer to IP, the Next
Generation ;-)
</Para>
<Para>
Now on to the next step. You want to connect your host, or maybe even
your entire LAN to another IPv6 capable network. This might be the "6bone"
that is setup especially for this particular purpose.
</Para>
<Para>
Let's assume that you have the following IPv6 network: 3ffe:604:6:8::/64 and
you want to connect it to 6bone, or a friend. Please note that the /64
subnet notation works just like with regular IP addresses.
</Para>
<Para>
Your IPv4 address is 145.100.24.181 and the 6bone router has IPv4 address
145.100.1.5
</Para>
<Screen width="80">
# ip tunnel add sixbone mode sit remote 145.100.1.5 [local 145.100.24.181 ttl 255]
# ip link set sixbone up
# ip addr add 3FFE:604:6:7::2/126 dev sixbone
# ip route add 3ffe::0/16 dev sixbone
</Screen>
<Para>
Let's discuss this. In the first line, we created a tunnel device called
sixbone. We gave it mode sit (which is IPv6 in IPv4 tunneling) and told it
where to go to (remote) and where to come from (local). TTL is set to
maximum, 255.
</Para>
<Para>
Next, we made the device active (up). After that, we added our own network
address, and set a route for 3ffe::/15 (which is currently all of 6bone)
through the tunnel. If the particular machine you run this on is your IPv6
gateway, then consider adding the following lines:
</Para>
<Screen width="80">
# echo 1 &#62;/proc/sys/net/ipv6/conf/all/forwarding
# /usr/local/sbin/radvd
</Screen>
<Para>
The latter, radvd is -like zebra- a router advertisement daemon, to
support IPv6's autoconfiguration features. Search for it with your favourite
search-engine if you like.
You can check things like this:
</Para>
<Screen width="80">
# /sbin/ip -f inet6 addr
</Screen>
<Para>
If you happen to have radvd running on your IPv6 gateway and boot your
IPv6 capable Linux on a machine on your local LAN, you would be able to
enjoy the benefits of IPv6 autoconfiguration:
</Para>
<Screen width="80">
# /sbin/ip -f inet6 addr
1: lo: &lt;LOOPBACK,UP&gt; mtu 3924 qdisc noqueue inet6 ::1/128 scope host
3: eth0: &lt;BROADCAST,MULTICAST,UP&gt; mtu 1500 qdisc pfifo_fast qlen 100
inet6 3ffe:604:6:8:5054:4cff:fe01:e3d6/64 scope global dynamic
valid_lft forever preferred_lft 604646sec inet6 fe80::5054:4cff:fe01:e3d6/10
scope link
</Screen>
<Para>
You could go ahead and configure your bind for IPv6 addresses. The A
type has an equivalent for IPv6: AAAA. The in-addr.arpa's equivalent is:
ip6.int. There's a lot of information available on this topic.
</Para>
<Para>
There is an increasing number of IPv6-aware applications available,
including secure shell, telnet, inetd, Mozilla the browser, Apache the
webserver and a lot of others. But this is all outside the scope of this
Routing document ;-)
</Para>
<Para>
On the Cisco side the configuration would be something like this:
<Screen>
!
interface Tunnel1
description IPv6 tunnel
no ip address
no ip directed-broadcast
ipv6 enable
ipv6 address 3FFE:604:6:7::1/126
tunnel source Serial0
tunnel destination 145.100.24.181
tunnel mode ipv6ip
!
ipv6 route 3FFE:604:6:8::/64 Tunnel1
</Screen>
But if you don't have a Cisco at your disposal, try one of the many
IPv6 tunnel brokers available on the Internet. They are willing to configure
their Cisco with an extra tunnel for you. Mostly by means of a friendly
web interface. Search for "ipv6 tunnel broker" on your favourite search engine.
</Para>
</Sect1>
</chapter>
<chapter id="lartc.ipsec">
<Title>IPsec: secure IP over the Internet</Title>
<Para>
FIXME: editor vacancy.
In the meantime, see: <ULink URL="http://www.freeswan.org/">
The FreeS/WAN project</ULink>.
Another IPSec implementation for Linux is Cerberus,
by NIST. However, their web pages have not been updated in over a year,
and their version tended to trail well behind the current Linux kernel.
USAGI, an alternative IPv6 implementation for Linux, also includes an
IPSec implementation, but that might only be for IPv6.
</Para>
</chapter>
<chapter id="lartc.multicast">
<Title>Multicast routing</Title>
<Para>
FIXME: Editor Vacancy!
</Para>
<Para>
The Multicast-HOWTO is ancient (relatively-speaking) and may be inaccurate
or misleading in places, for that reason.
</Para>
<Para>
Before you can do any multicast routing, you need to configure the Linux
kernel to support the type of multicast routing you want to do. This, in
turn, requires you to decide what type of multicast routing you expect to
be using. There are essentially four "common" types - DVMRP (the Multicast
version of the RIP unicast protocol), MOSPF (the same, but for OSPF), PIM-SM
("Protocol Independent Multicasting - Sparse Mode", which assumes that users
of any multicast group are spread out, rather than clumped) and PIM-DM (the
same, but "Dense Mode", which assumes that there will be significant clumps
of users of the same multicast group).
</Para>
<Para>
In the Linux kernel, you will notice that these options don't appear. This is
because the protocol itself is handled by a routing application, such as
Zebra, mrouted, or pimd. However, you still have to have a good idea of which
you're going to use, to select the right options in the kernel.
</Para>
<Para>
For all multicast routing, you will definitely need to enable "multicasting"
and "multicast routing". For DVMRP and MOSPF, this is sufficient. If you are
going to use PIM, you must also enable PIMv1 or PIMv2, depending on whether
the network you are connecting to uses version 1 or 2 of the PIM protocol.
</Para>
<Para>
Once you have all that sorted out, and your new Linux kernel compiled, you
will see that the IP protocols listed, at boot time, now include IGMP. This
is a protocol for managing multicast groups. At the time of writing, Linux
supports IGMP versions 1 and 2 only, although version 3 does exist and has
been documented. This doesn't really affect us that much, as IGMPv3 is still
new enough that the extra capabilities of IGMPv3 aren't going to be that
much use. Because IGMP deals with groups, only the features present in the
simplest version of IGMP over the entire group are going to be used. For the
most part, that will be IGMPv2, although IGMPv1 is sill going to be
encountered.
</Para>
<Para>
So far, so good. We've enabled multicasting. Now, we have to tell the Linux
kernel to actually do something with it, so we can start routing. This means
adding the Multicast virtual network to the router table:
</Para>
<Para>
ip route add 224.0.0.0/4 dev eth0
</Para>
<Para>
(Assuming, of course, that you're multicasting over eth0! Substitute the
device of your choice, for this.)
</Para>
<Para>
Now, tell Linux to forward packets...
</Para>
<Para>
echo 1 &#62; /proc/sys/net/ipv4/ip_forward
</Para>
<Para>
At this point, you may be wondering if this is ever going to do anything. So,
to test our connection, we ping the default group, 224.0.0.1, to see if anyone
is alive. All machines on your LAN with multicasting enabled <Emphasis>should</Emphasis>
respond, but nothing else. You'll notice that none of the machines that
respond have an IP address of 224.0.0.1. What a surprise! :) This is a group
address (a "broadcast" to subscribers), and all members of the group will
respond with their own address, not the group address.
</Para>
<Para>
ping -c 2 224.0.0.1
</Para>
<Para>
At this point, you're ready to do actual multicast routing. Well, assuming
that you have two networks to route between.
</Para>
<Para>
(To Be Continued!)
</Para>
</chapter>
<chapter id="lartc.qdisc">
<Title>Queueing Disciplines for Bandwidth Management</Title>
<Para>
Now, when I discovered this, it <Emphasis>really</Emphasis> blew me away. Linux 2.2/2.4
comes with everything to manage bandwidth in ways comparable to high-end
dedicated bandwidth management systems.
</Para>
<Para>
Linux even goes far beyond what Frame and ATM provide.
</Para>
<Para>Just to prevent confusion, <command>tc</command> uses the following
rules for bandwith specification:
<literallayout class='monospaced'>
mbps = 1024 kbps = 1024 * 1024 bps =&#62; byte/s
mbit = 1024 kbit =&#62; kilo bit/s.
mb = 1024 kb = 1024 * 1024 b =&#62; byte
mbit = 1024 kbit =&#62; kilo bit.
</literallayout>
Internally, the number is stored in bps and b.
</Para>
<Para>But when <command>tc</command> prints the rate, it uses following :
</Para>
<literallayout class='monospaced'>
1Mbit = 1024 Kbit = 1024 * 1024 bps =&#62; bit/s
</literallayout>
<Sect1 id="lartc.qdisc.explain">
<Title>Queues and Queueing Disciplines explained</Title>
<Para>
With queueing we determine the way in which data is <Emphasis>SENT</Emphasis>.
It is important to realise that we can only shape data that we transmit.
</Para>
<Para>
With the way the Internet works, we have no direct control of what people
send us. It's a bit like your (physical!) mailbox at home. There is no way
you can influence the world to modify the amount of mail they send you,
short of contacting everybody.
</Para>
<Para>
However, the Internet is mostly based on TCP/IP which has a few features
that help us. TCP/IP has no way of knowing the capacity of the network
between two hosts, so it just starts sending data faster and faster ('slow
start') and when packets start getting lost, because there is no room to
send them, it will slow down. In fact it is a bit smarter than this, but
more about that later.
</Para>
<Para>
This is the equivalent of not reading half of your mail, and hoping that
people will stop sending it to you. With the difference that it works for
the Internet :-)
</Para>
<Para>
If you have a router and wish to prevent certain hosts within your network
from downloading too fast, you need to do your shaping on the *inner* interface
of your router, the one that sends data to your own computers.
</Para>
<Para>
You also have to be sure you are controlling the bottleneck of the link.
If you have a 100Mbit NIC and you have a router that has a 256kbit link,
you have to make sure you are not sending more data than your router can
handle. Otherwise, it will be the router who is controlling the link and
shaping the available bandwith. We need to 'own the queue' so to speak, and
be the slowest link in the chain. Luckily this is easily possible.
</Para>
</Sect1>
<Sect1 id="lartc.qdisc.classless">
<Title>Simple, classless Queueing Disciplines</Title>
<Para>
As said, with queueing disciplines, we change the way data is sent.
Classless queueing disciplines are those that, by and large accept data and
only reschedule, delay or drop it.
</Para>
<Para>
These can be used to shape traffic for an entire interface, without any
subdivisions. It is vital that you understand this part of queueing before
we go onto the classful qdisc-containing-qdiscs!
</Para>
<Para>
By far the most widely used discipline is the pfifo_fast qdisc - this is the
default. This also explains why these advanced features are so robust. They
are nothing more than 'just another queue'.
</Para>
<Para>
Each of these queues has specific strengths and weaknesses. Not all of them
may be as well tested.
</Para>
<Sect2>
<Title>pfifo_fast</Title>
<Para>
This queue is, as the name says, First In, First Out, which means that no
packet receives special treatment. At least, not quite. This queue has 3 so
called 'bands'. Within each band, FIFO rules apply. However, as long as
there are packets waiting in band 0, band 1 won't be processed. Same goes
for band 1 and band 2.
</Para>
<Para>
The kernel honors the so called Type of Service flag of packets, and takes
care to insert 'minimum delay' packets in band 0.
</Para>
<Para>
Do not confuse this classless simple qdisc with the classful PRIO one!
Although they behave similarly, pfifo_fast is classless and you cannot add
other qdiscs to it with the tc command.
</Para>
<Sect3>
<Title>Parameters &amp; usage</Title>
<Para>
You can't configure the pfifo_fast qdisc as it is the hardwired default.
This is how it is configured by default:
<VariableList>
<VarListEntry>
<Term>priomap</Term>
<ListItem>
<Para>
Determines how packet priorities, as assigned by the kernel, map to bands.
Mapping occurs based on the TOS octet of the packet, which looks like this:
</Para>
<Para>
<Screen>
0 1 2 3 4 5 6 7
+-----+-----+-----+-----+-----+-----+-----+-----+
| | | |
| PRECEDENCE | TOS | MBZ |
| | | |
+-----+-----+-----+-----+-----+-----+-----+-----+
</Screen>
</Para>
<Para>
The four TOS bits (the 'TOS field') are defined as:
<Screen>
Binary Decimcal Meaning
-----------------------------------------
1000 8 Minimize delay (md)
0100 4 Maximize throughput (mt)
0010 2 Maximize reliability (mr)
0001 1 Minimize monetary cost (mmc)
0000 0 Normal Service
</Screen>
</Para>
<Para>
As there is 1 bit to the right of these four bits, the actual value of the
TOS field is double the value of the TOS bits. Tcpdump -v -v shows you the
value of the entire TOS field, not just the four bits. It is the value you
see in the first column of this table:
</Para>
<Para>
<Screen>
TOS Bits Means Linux Priority Band
------------------------------------------------------------
0x0 0 Normal Service 0 Best Effort 1
0x2 1 Minimize Monetary Cost 1 Filler 2
0x4 2 Maximize Reliability 0 Best Effort 1
0x6 3 mmc+mr 0 Best Effort 1
0x8 4 Maximize Throughput 2 Bulk 2
0xa 5 mmc+mt 2 Bulk 2
0xc 6 mr+mt 2 Bulk 2
0xe 7 mmc+mr+mt 2 Bulk 2
0x10 8 Minimize Delay 6 Interactive 0
0x12 9 mmc+md 6 Interactive 0
0x14 10 mr+md 6 Interactive 0
0x16 11 mmc+mr+md 6 Interactive 0
0x18 12 mt+md 4 Int. Bulk 1
0x1a 13 mmc+mt+md 4 Int. Bulk 1
0x1c 14 mr+mt+md 4 Int. Bulk 1
0x1e 15 mmc+mr+mt+md 4 Int. Bulk 1
</Screen>
</Para>
<Para>
Lots of numbers. The second column contains the value of the relevant four
TOS bits, followed by their translated meaning. For example, 15 stands for a
packet wanting Minimal Monetary Cost, Maximum Reliability, Maximum
Throughput AND Minimum Delay. I would call this a 'Dutch Packet'.
</Para>
<Para>
The fourth column lists the way the Linux kernel interprets the TOS bits, by
showing to which Priority they are mapped.
</Para>
<Para>
The last column shows the result of the default priomap. On the command line,
the default priomap looks like this:
<Screen>
1, 2, 2, 2, 1, 2, 0, 0 , 1, 1, 1, 1, 1, 1, 1, 1
</Screen>
</Para>
<Para>
This means that priority 4, for example, gets mapped to band number 1. The
priomap also allows you to list higher priorities (&gt; 7) which do not
correspond to TOS mappings, but which are set by other means.
</Para>
<Para>
This table from RFC 1349 (read it for more details) tells you how
applications might very well set their TOS bits:
<Screen>
TELNET 1000 (minimize delay)
FTP
Control 1000 (minimize delay)
Data 0100 (maximize throughput)
TFTP 1000 (minimize delay)
SMTP
Command phase 1000 (minimize delay)
DATA phase 0100 (maximize throughput)
Domain Name Service
UDP Query 1000 (minimize delay)
TCP Query 0000
Zone Transfer 0100 (maximize throughput)
NNTP 0001 (minimize monetary cost)
ICMP
Errors 0000
Requests 0000 (mostly)
Responses &#60;same as request&#62; (mostly)
</Screen>
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>txqueuelen</Term>
<ListItem>
<Para>
The length of this queue is gleaned from the interface configuration, which
you can see and set with ifconfig and ip. To set the queue length to 10,
execute: ifconfig eth0 txqueuelen 10
</Para>
<Para>
You can't set this parameter with tc!
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect3>
</Sect2>
<Sect2>
<Title>Token Bucket Filter</Title>
<Para>
The Token Bucket Filter (TBF) is a simple qdisc that only passes packets
arriving at a rate which is not exceeding some administratively set rate, but
with the possibility to allow short bursts in excess of this rate.
</Para>
<Para>
TBF is very precise, network- and processor friendly. It should be your
first choice if you simply want to slow an interface down!
</Para>
<Para>
The TBF implementation consists of a buffer (bucket), constantly filled by
some virtual pieces of information called tokens, at a specific rate (token
rate). The most important parameter of the bucket is its size, that is the
number of tokens it can store.
</Para>
<Para>
Each arriving token collects one incoming data packet from the data queue
and is then deleted from the bucket. Associating this algorithm
with the two flows -- token and data, gives us three possible scenarios:
</Para>
<Para>
<ItemizedList>
<ListItem>
<Para>
The data arrives in TBF at a rate that's <Emphasis>equal</Emphasis> to the rate
of incoming tokens. In this case each incoming packet has its matching token
and passes the queue without delay.
</Para>
</ListItem>
<ListItem>
<Para>
The data arrives in TBF at a rate that's <Emphasis>smaller</Emphasis> than the
token rate. Only a part of the tokens are deleted at output of each data packet
that's sent out the queue, so the tokens accumulate, up to the bucket size.
The unused tokens can then be used to send data at a speed that's exceeding the
standard token rate, in case short data bursts occur.
</Para>
</ListItem>
<ListItem>
<Para>
The data arrives in TBF at a rate <Emphasis>bigger</Emphasis> than the token rate.
This means that the bucket will soon be devoid of tokens, which causes the
TBF to throttle itself for a while. This is called an 'overlimit situation'.
If packets keep coming in, packets will start to get dropped.
</Para>
</ListItem>
</ItemizedList>
</Para>
<Para>
The last scenario is very important, because it allows to
administratively shape the bandwidth available to data that's passing
the filter.
</Para>
<Para>
The accumulation of tokens allows a short burst of overlimit data to be
still passed without loss, but any lasting overload will cause packets to be
constantly delayed, and then dropped.
</Para>
<Para>
Please note that in the actual implementation, tokens correspond to bytes,
not packets.
</Para>
<Sect3>
<Title>Parameters &amp; usage</Title>
<Para>
Even though you will probably not need to change them, tbf has some knobs
available. First the parameters that are always available:
<VariableList>
<VarListEntry>
<Term>limit or latency</Term>
<ListItem>
<Para>
Limit is the number of bytes that can be queued waiting for tokens to become
available. You can also specify this the other way around by setting the
latency parameter, which specifies the maximum amount of time a packet can
sit in the TBF. The latter calculation takes into account the size of the
bucket, the rate and possibly the peakrate (if set).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>burst/buffer/maxburst</Term>
<ListItem>
<Para>
Size of the bucket, in bytes. This is the maximum amount of bytes that
tokens can be available for instantaneously. In general, larger shaping
rates require a larger buffer. For 10mbit/s on Intel, you need at least
10kbyte buffer if you want to reach your configured rate!
</Para>
<Para>
If your buffer is too small, packets may be dropped because more tokens
arrive per timer tick than fit in your bucket.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>mpu</Term>
<ListItem>
<Para>
A zero-sized packet does not use zero bandwidth. For ethernet, no packet
uses less than 64 bytes. The Minimum Packet Unit determines the minimal
token usage for a packet.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>rate</Term>
<ListItem>
<Para>
The speedknob. See remarks above about limits!
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
If the bucket contains tokens and is allowed to empty, by default it does so
at infinite speed. If this is unacceptable, use the following parameters:
</Para>
<Para>
<VariableList>
<VarListEntry>
<Term>peakrate</Term>
<ListItem>
<Para>
If tokens are available, and packets arrive, they are sent out immediately
by default, at 'lightspeed' so to speak. That may not be what you want,
especially if you have a large bucket.
</Para>
<Para>
The peakrate can be used to specify how quickly the bucket is allowed to be
depleted. If doing everything by the book, this is achieved by releasing a
packet, and then wait just long enough, and release the next. We calculated
our waits so we send just at peakrate.
</Para>
<Para>
However, due to de default 10ms timer resolution of Unix, with 10.000 bits
average packets, we are limited to 1mbit/s of peakrate!
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>mtu/minburst</Term>
<ListItem>
<Para>
The 1mbit/s peakrate is not very useful if your regular rate is more than
that. A higher peakrate is possible by sending out more packets per
timertick, which effectively means that we create a second bucket!
</Para>
<Para>
This second bucket defaults to a single packet, which is not a bucket at
all.
</Para>
<Para>
To calculate the maximum possible peakrate, multiply the configured mtu by
100 (or more correctly, HZ, which is 100 on Intel, 1024 on Alpha).
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect3>
<Sect3>
<Title>Sample configuration</Title>
<Para>
A simple but *very* useful configuration is this:
<Screen>
# tc qdisc add dev ppp0 root tbf rate 220kbit latency 50ms burst 1540
</Screen>
</Para>
<Para>
Ok, why is this useful? If you have a networking device with a large queue,
like a DSL modem or a cable modem, and you talk to it over a fast device,
like over an ethernet interface, you will find that uploading absolutely
destroys interactivity.
</Para>
<Para>
This is because uploading will fill the queue in the modem, which is
probably *huge* because this helps actually achieving good data throughput
uploading. But this is not what you want, you want to have the queue not too
big so interactivity remains and you can still do other stuff while sending
data.
</Para>
<Para>
The line above slows down sending to a rate that does not lead to a queue in
the modem - the queue will be in Linux, where we can control it to a limited
size.
</Para>
<Para>
Change 220kbit to your uplink's *actual* speed, minus a few percent. If you
have a really fast modem, raise 'burst' a bit.
</Para>
</Sect3>
</Sect2>
<Sect2 id="lartc.sfq">
<Title>Stochastic Fairness Queueing</Title>
<Para>
Stochastic Fairness Queueing (SFQ) is a simple implementation of the fair
queueing algorithms family. It's less accurate than others, but it also
requires less calculations while being almost perfectly fair.
</Para>
<Para>
The key word in SFQ is conversation (or flow), which mostly corresponds to a
TCP session or a UDP stream. Traffic is divided into a pretty large number
of FIFO queues, one for each conversation. Traffic is then sent in a round
robin fashion, giving each session the chance to send data in turn.
</Para>
<Para>
This leads to very fair behaviour and disallows any single conversation from
drowning out the rest. SFQ is called 'Stochastic' because it doesn't really
allocate a queue for each session, it has an algorithm which divides traffic
over a limited number of queues using a hashing algorithm.
</Para>
<Para>
Because of the hash, multiple sessions might end up in the same bucket, which
would halve each session's chance of sending a packet, thus halving the
effective speed available. To prevent this situation from becoming
noticeable, SFQ changes its hashing algorithm quite often so that any two
colliding sessions will only do so for a small number of seconds.
</Para>
<Para>
It is important to note that SFQ is only useful in case your actual outgoing
interface is really full! If it isn't then there will be no queue on your
linux machine and hence no effect. Later on we will describe how to combine
SFQ with other qdiscs to get a best-of-both worlds situation.
</Para>
<Para>
Specifically, setting SFQ on the ethernet interface heading to your
cable modem or DSL router is pointless without further shaping!
</Para>
<Sect3>
<Title>Parameters &amp; usage</Title>
<Para>
The SFQ is pretty much self tuning:
<VariableList>
<VarListEntry>
<Term>perturb</Term>
<ListItem>
<Para>
Reconfigure hashing once this many seconds. If unset, hash will never be
reconfigured. Not recommended. 10 seconds is probably a good value.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>quantum</Term>
<ListItem>
<Para>
Amount of bytes a stream is allowed to dequeue before the next queue gets a
turn. Defaults to 1 maximum sized packet (MTU-sized). Do not set below the
MTU!
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect3>
<Sect3>
<Title>Sample configuration</Title>
<Para>
If you have a device which has identical link speed and actual available
rate, like a phone modem, this configuration will help promote fairness:
<Screen>
# tc qdisc add dev ppp0 root sfq perturb 10
# tc -s -d qdisc ls
qdisc sfq 800c: dev ppp0 quantum 1514b limit 128p flows 128/1024 perturb 10sec
Sent 4812 bytes 62 pkts (dropped 0, overlimits 0)
</Screen>
</Para>
<Para>
The number 800c: is the automatically assigned handle number, limit means
that 128 packets can wait in this queue. There are 1024 hashbuckets
available for accounting, of which 128 can be active at a time (no more
packets fit in the queue!) Once every 10 seconds, the hashes are
reconfigured.
</Para>
</Sect3>
</Sect2>
</Sect1>
<Sect1 id="lartc.qdisc.advice">
<Title>Advice for when to use which queue</Title>
<Para>
Summarizing, these are the simple queues that actually manage traffic by
reordering, slowing or dropping packets.
</Para>
<Para>
The following tips may help in choosing which queue to use. It mentions some
qdiscs described in the
<citetitle><xref linkend="lartc.adv-qdisc"></citetitle> chapter.
</Para>
<ItemizedList>
<ListItem>
<Para>
To purely slow down outgoing traffic, use the Token Bucket Filter. Works up
to huge bandwidths, if you scale the bucket.
</Para>
</ListItem>
<ListItem>
<Para>
If your link is truly full and you want to make sure that no single session
can dominate your outgoing bandwidth, use Stochastical Fairness Queueing.
</Para>
</ListItem>
<ListItem>
<Para>
If you have a big backbone and know what you are doing, consider Random
Early Drop (see Advanced chapter).
</Para>
</ListItem>
<ListItem>
<Para>
To 'shape' incoming traffic which you are not forwarding, use the Ingress
Policer. Incoming shaping is called 'policing', by the way, not 'shaping'.
</Para>
</ListItem>
<ListItem>
<Para>
If you *are* forwarding it, use a TBF on the interface you are forwarding
the data to. Unless you want to shape traffic that may go out over several
interfaces, in which case the only common factor is the incoming interface.
In that case use the Ingress Policer.
</Para>
</ListItem>
<ListItem>
<Para>
If you don't want to shape, but only want to see if your interface is so
loaded that it has to queue, use the pfifo queue (not pfifo_fast). It lacks
internal bands but does account the size of its backlog.
</Para>
</ListItem>
<ListItem>
<Para>
Finally - you can also do <quote>social shaping</quote>.
You may not always be able to use technology to achieve what you want.
Users experience technical constraints as hostile.
A kind word may also help with getting your bandwidth to be divided right!
</Para>
</ListItem>
</ItemizedList>
</Sect1>
<Sect1 id="lartc.qdisc.terminology">
<Title>Terminology</Title>
<Para>
To properly understand more complicated configurations it is necessary to
explain a few concepts first. Because of the complexity and he relative
youth of the subject, a lot of different words are used when people in fact
mean the same thing.
</Para>
<Para>
The following is loosely based on
<filename>draft-ietf-diffserv-model-06.txt</filename>,
<citetitle>An Informal Management Model for Diffserv Routers</citetitle>.
It can currently be found at
<ulink url="http://www.ietf.org/internet-drafts/draft-ietf-diffserv-model-06.txt">
http://www.ietf.org/internet-drafts/draft-ietf-diffserv-model-06.txt
</ulink>.
</Para>
<Para>
Read it for the strict definitions of the terms used.
<VariableList>
<VarListEntry>
<Term>Queueing Discipline</Term>
<ListItem>
<Para>
An algorithm that manages the queue of a device, either incoming (ingress)
or outgoing (egress).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Classless qdisc</Term>
<ListItem>
<Para>
A qdisc with no configurable internal subdivisions.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Classful qdisc</Term>
<ListItem>
<Para>
A classful qdisc contains multiple classes. Each of these classes contains a
further qdisc, which may again be classful, but need not be. According to
the strict definition, pfifo_fast *is* classful, because it contains three
bands which are, in fact, classes. However, from the user's configuration
perspective, it is classless as the classes can't be touched with the tc
tool.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Classes</Term>
<ListItem>
<Para>
A classful qdisc may have many classes, which each are internal to the
qdisc. Each of these classes may contain a real qdisc.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Classifier</Term>
<ListItem>
<Para>
Each classful qdisc needs to determine to which class it needs to send a
packet. This is done using the classifier.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Filter</Term>
<ListItem>
<Para>
Classification can be performed using filters. A filter contains a number of
conditions which if matched, make the filter match.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Scheduling</Term>
<ListItem>
<Para>
A qdisc may, with the help of a classifier, decide that some packets need to
go out earlier than others. This process is called Scheduling, and is
performed for example by the pfifo_fast qdisc mentioned earlier. Scheduling
is also called 'reordering', but this is confusing.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Shaping</Term>
<ListItem>
<Para>
The process of delaying packets before they go out to make traffic confirm
to a configured maximum rate. Shaping is performed on egress. Colloquially,
dropping packets to slow traffic down is also often called Shaping.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Policing</Term>
<ListItem>
<Para>
Delaying or dropping packets in order to make traffic stay below a
configured bandwidth. In Linux, policing can only drop a packet and not
delay it - there is no 'ingress queue'.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Work-Conserving</Term>
<ListItem>
<Para>
A work-conserving qdisc always delivers a packet if one is available. In
other words, it never delays a packet if the network adaptor is ready to
send one (in the case of an egress qdisc).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>non-Work-Conserving</Term>
<ListItem>
<Para>
Some queues, like for example the Token Bucket Filter, may need to hold on
to a packet for a certain time in order to limit the bandwidth. This means
that they sometimes refuse to give up a packet, even though they have one
available.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
Now that we have our terminology straight, let's see where all these things
are.
</Para>
<Para>
<Screen width="80">
Userspace programs
^
|
+---------------+-----------------------------------------+
| Y |
| -------&#62; IP Stack |
| | | |
| | Y |
| | Y |
| ^ | |
| | / ----------&#62; Forwarding -&#62; |
| ^ / | |
| |/ Y |
| | | |
| ^ Y /-qdisc1-\ |
| | Egress /--qdisc2--\ |
---&#62;-&#62;Ingress Classifier ---qdisc3---- | -&#62;
| Qdisc \__qdisc4__/ |
| \-qdiscN_/ |
| |
+----------------------------------------------------------+
</Screen>
Thanks to Jamal Hadi Salim for this ASCII representation.
</Para>
<Para>
The big block represents the kernel. The leftmost arrow represents traffic
entering your machine from the network. It is then fed to the Ingress
Qdisc which may apply Filters to a packet, and decide to drop it. This
is called 'Policing'.
</Para>
<Para>
This happens at a very early stage, before it has seen a lot of the kernel.
It is therefore a very good place to drop traffic very early, without
consuming a lot of CPU power.
</Para>
<Para>
If the packet is allowed to continue, it may be destined for a local
application, in which case it enters the IP stack in order to be processed,
and handed over to a userspace program. The packet may also be forwarded
without entering an application, in which case it is destined for egress.
Userspace programs may also deliver data, which is then examined and
forwarded to the Egress Classifier.
</Para>
<Para>
There it is investigated and enqueued to any of a number of qdiscs. In the
unconfigured default case, there is only one egress qdisc installed, the
pfifo_fast, which always receives the packet. This is called 'enqueueing'.
</Para>
<Para>
The packet now sits in the qdisc, waiting for the kernel to ask for
it for transmission over the network interface. This is called 'dequeueing'.
</Para>
<Para>
This picture also holds in case there is only one network adaptor - the
arrows entering and leaving the kernel should not be taken too literally.
Each network adaptor has both ingress and egress hooks.
</Para>
</Sect1>
<Sect1 id="lartc.qdisc.classful">
<Title>Classful Queueing Disciplines</Title>
<Para>
Classful qdiscs are very useful if you have different kinds of traffic which
should have differing treatment. One of the classful qdiscs is called 'CBQ'
, 'Class Based Queueing' and it is so widely mentioned that people identify
queueing with classes solely with CBQ, but this is not the case.
</Para>
<Para>
CBQ is merely the oldest kid on the block - and also the most complex one.
It may not always do what you want. This may come as something of a shock
to many who fell for the 'sendmail effect', which teaches us that any
complex technology which doesn't come with documentation must be the best
available.
</Para>
<Para>
More about CBQ and its alternatives shortly.
</Para>
<Sect2>
<Title>Flow within classful qdiscs &amp; classes</Title>
<Para>
When traffic enters a classful qdisc, it needs to be sent to any of the
classes within - it needs to be 'classified'. To determine what to do with a
packet, the so called 'filters' are consulted. It is important to know that
the filters are called from within a qdisc, and not the other way around!
</Para>
<Para>
The filters attached to that qdisc then return with a decision, and the
qdisc uses this to enqueue the packet into one of the classes. Each subclass
may try other filters to see if further instructions apply. If not, the
class enqueues the packet to the qdisc it contains.
</Para>
<Para>
Besides containing other qdiscs, most classful qdiscs also perform shaping.
This is useful to perform both packet scheduling (with SFQ, for example) and
rate control. You need this in cases where you have a high speed
interface (for example, ethernet) to a slower device (a cable modem).
</Para>
<Para>
If you were only to run SFQ, nothing would happen, as packets enter &amp;
leave your router without delay: the output interface is far faster than
your actual link speed. There is no queue to schedule then.
</Para>
</Sect2>
<Sect2>
<Title>The qdisc family: roots, handles, siblings and parents</Title>
<Para>
Each interface has one egress 'root qdisc', by default the earlier mentioned
classless pfifo_fast queueing discipline. Each qdisc can be assigned a
handle, which can be used by later configuration statements to refer to that
qdisc. Besides an egress qdisc, an interface may also have an ingress, which
polices traffic coming in.
</Para>
<Para>
The handles of these qdiscs consist of two parts, a major number and a minor
number. It is habitual to name the root qdisc '1:', which is equal to '1:0'.
The minor number of a qdisc is always 0.
</Para>
<Para>
Classes need to have the same major number as their parent.
</Para>
<Sect3>
<Title>How filters are used to classify traffic </Title>
<Para>
Recapping, a typical hierarchy might look like this:
<Screen>
root 1:
|
_1:1_
/ | \
/ | \
/ | \
10: 11: 12:
/ \ / \
10:1 10:2 12:1 12:2
</Screen>
</Para>
<Para>
But don't let this tree fool you! You should *not* imagine the kernel to be
at the apex of the tree and the network below, that is just not the case.
Packets get enqueued and dequeued at the root qdisc, which is the only thing
the kernel talks to.
</Para>
<Para>
A packet might get classified in a chain like this:
</Para>
<Para>
1: -&#62; 1:1 -&#62; 12: -&#62; 12:2
</Para>
<Para>
The packet now resides in a queue in a qdisc attached to class 12:2. In this
example, a filter was attached to each 'node' in the tree, each choosing a
branch to take next. This can make sense. However, this is also possible:
</Para>
<Para>
1: -&#62; 12:2
</Para>
<Para>
In this case, a filter attached to the root decided to send the packet
directly to 12:2.
</Para>
</Sect3>
<Sect3>
<Title>How packets are dequeued to the hardware</Title>
<Para>
When the kernel decides that it needs to extract packets to send to the
interface, the root qdisc 1: gets a dequeue request, which is passed to
1:1, which is in turn passed to 10:, 11: and 12:, which each query their
siblings, and try to dequeue() from them. In this case, the kernel needs to
walk the entire tree, because only 12:2 contains a packet.
</Para>
<Para>
In short, nested classes ONLY talk to their parent qdiscs, never to an
interface. Only the root qdisc gets dequeued by the kernel!
</Para>
<Para>
The upshot of this is that classes never get dequeued faster than their
parents allow. And this is exactly what we want: this way we can have SFQ in
an inner class, which doesn't do any shaping, only scheduling, and have a
shaping outer qdisc, which does the shaping.
</Para>
</Sect3>
</Sect2>
<Sect2>
<Title>The PRIO qdisc</Title>
<Para>
The PRIO qdisc doesn't actually shape, it only subdivides traffic based on
how you configured your filters. You can consider the PRIO qdisc a kind
of pfifo_fast on steroids, whereby each band is a separate class instead of
a simple FIFO.
</Para>
<Para>
When a packet is enqueued to the PRIO qdisc, a class is chosen based on the
filter commands you gave. By default, three classes are created. These
classes by default contain pure FIFO qdiscs with no internal
structure, but you can replace these by any qdisc you have available.
</Para>
<Para>
Whenever a packet needs to be dequeued, class :1 is tried first. Higher
classes are only used if lower bands all did not give up a packet.
</Para>
<Para>
This qdisc is very useful in case you want to prioritize certain kinds of
traffic without using only TOS-flags but using all the power of the tc
filters. It can also contain more all qdiscs, whereas pfifo_fast is limited
to simple fifo qdiscs.
</Para>
<Para>
Because it doesn't actually shape, the same warning as for SFQ holds: either
use it only if your physical link is really full or wrap it inside a
classful qdisc that does shape. The last holds for almost all cable modems
and DSL devices.
</Para>
<Para>
In formal words, the PRIO qdisc is a Work-Conserving scheduler.
</Para>
<Sect3>
<Title>PRIO parameters &amp; usage</Title>
<Para>
The following parameters are recognized by tc:
<VariableList>
<VarListEntry>
<Term>bands</Term>
<ListItem>
<Para>
Number of bands to create. Each band is in fact a class. If you change this
number, you must also change:
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>priomap</Term>
<ListItem>
<Para>
If you do not provide tc filters to classify traffic, the PRIO qdisc looks
at the TC_PRIO priority to decide how to enqueue traffic.
</Para>
<Para>
This works just like with the pfifo_fast qdisc mentioned earlier, see there
for lots of detail.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
The bands are classes, and are called major:1 to major:3 by default, so if
your PRIO qdisc is called 12:, tc filter traffic to 12:1 to grant it more
priority.
</Para>
<Para>
Reiterating, band 0 goes to minor number 1! Band 1 to minor number 2, etc.
</Para>
</Sect3>
<Sect3>
<Title>Sample configuration</Title>
<Para>
We will create this tree:
<Screen>
root 1: prio
/ | \
1:1 1:2 1:3
| | |
10: 20: 30:
sfq tbf sfq
band 0 1 2
</Screen>
</Para>
<Para>
Bulk traffic will go to 30:, interactive traffic to 20: or 10:.
</Para>
<Para>
Command lines:
<Screen>
# tc qdisc add dev eth0 root handle 1: prio
## This *instantly* creates classes 1:1, 1:2, 1:3
# tc qdisc add dev eth0 parent 1:1 handle 10: sfq
# tc qdisc add dev eth0 parent 1:2 handle 20: tbf rate 20kbit buffer 1600 limit 3000
# tc qdisc add dev eth0 parent 1:3 handle 30: sfq
</Screen>
</Para>
<Para>
Now let's see what we created:
<Screen>
# tc -s qdisc ls dev eth0
qdisc sfq 30: quantum 1514b
Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
qdisc tbf 20: rate 20Kbit burst 1599b lat 667.6ms
Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
qdisc sfq 10: quantum 1514b
Sent 132 bytes 2 pkts (dropped 0, overlimits 0)
qdisc prio 1: bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 174 bytes 3 pkts (dropped 0, overlimits 0)
</Screen>
As you can see, band 0 has already had some traffic, and one packet was sent
while running this command!
</Para>
<Para>
We now do some bulk data transfer with a tool that properly sets TOS flags,
and take another look:
<Screen>
# scp tc ahu@10.0.0.11:./
ahu@10.0.0.11's password:
tc 100% |*****************************| 353 KB 00:00
# tc -s qdisc ls dev eth0
qdisc sfq 30: quantum 1514b
Sent 384228 bytes 274 pkts (dropped 0, overlimits 0)
qdisc tbf 20: rate 20Kbit burst 1599b lat 667.6ms
Sent 2640 bytes 20 pkts (dropped 0, overlimits 0)
qdisc sfq 10: quantum 1514b
Sent 2230 bytes 31 pkts (dropped 0, overlimits 0)
qdisc prio 1: bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 389140 bytes 326 pkts (dropped 0, overlimits 0)
</Screen>
As you can see, all traffic went to handle 30:, which is the lowest priority
band, just as intended. Now to verify that interactive traffic goes to
higher bands, we create some interactive traffic:
</Para>
<Para>
<Screen>
# tc -s qdisc ls dev eth0
qdisc sfq 30: quantum 1514b
Sent 384228 bytes 274 pkts (dropped 0, overlimits 0)
qdisc tbf 20: rate 20Kbit burst 1599b lat 667.6ms
Sent 2640 bytes 20 pkts (dropped 0, overlimits 0)
qdisc sfq 10: quantum 1514b
Sent 14926 bytes 193 pkts (dropped 0, overlimits 0)
qdisc prio 1: bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 401836 bytes 488 pkts (dropped 0, overlimits 0)
</Screen>
</Para>
<Para>
It worked - all additional traffic has gone to 10:, which is our highest
priority qdisc. No traffic was sent to the lowest priority, which previously
received our entire scp.
</Para>
</Sect3>
</Sect2>
<Sect2>
<Title>The famous CBQ qdisc</Title>
<Para>
As said before, CBQ is the most complex qdisc available, the most hyped, the
least understood, and probably the trickiest one to get right. This is not
because the authors are evil or incompetent, far from it, it's just that the
CBQ algorithm isn't all that precise and doesn't really match the way Linux
works.
</Para>
<Para>
Besides being classful, CBQ is also a shaper and it is in that aspect that
it really doesn't work very well. It should work like this. If you try to
shape a 10mbit/s connection to 1mbit/s, the link should be idle 90&percnt; of the
time. If it isn't, we need to throttle so that it IS idle 90&percnt; of the time.
</Para>
<Para>
This is pretty hard to measure, so CBQ instead derives the idle time from
the number of microseconds that elapse between requests from the hardware
layer for more data. Combined, this can be used to approximate how full or
empty the link is.
</Para>
<Para>
This is rather circumspect and doesn't always arrive at proper results. For
example, what if the actual link speed of an interface that is not really
able to transmit the full 100mbit/s of data, perhaps because of a badly
implemented driver? A PCMCIA network card will also never achieve 100mbit/s
because of the way the bus is designed - again, how do we calculate the idle
time?
</Para>
<Para>
It gets even worse if we consider not-quite-real network devices like PPP
over Ethernet or PPTP over TCP/IP. The effective bandwidth in that case is
probably determined by the efficiency of pipes to userspace - which is huge.
</Para>
<Para>
People who have done measurements discover that CBQ is not always very
accurate and sometimes completely misses the mark.
</Para>
<Para>
In many circumstances however it works well. With the documentation provided
here, you should be able to configure it to work well in most cases.
</Para>
<Sect3>
<Title>CBQ shaping in detail</Title>
<Para>
As said before, CBQ works by making sure that the link is idle just long
enough to bring down the real bandwidth to the configured rate. To do so, it
calculates the time that should pass between average packets.
</Para>
<Para>
During operations, the effective idletime is measured using an exponential
weighted moving average (EWMA), which considers recent packets to be
exponentially more important than past ones. The UNIX loadaverage is
calculated in the same way.
</Para>
<Para>
The calculated idle time is subtracted from the EWMA measured one, the
resulting number is called 'avgidle'. A perfectly loaded link has an avgidle
of zero: packets arrive exactly once every calculated interval.
</Para>
<Para>
An overloaded link has a negative avgidle and if it gets too negative, CBQ
shuts down for a while and is then 'overlimit'.
</Para>
<Para>
Conversely, an idle link might amass a huge avgidle, which would then allow
infinite bandwidths after a few hours of silence. To prevent this, avgidle is
capped at maxidle.
</Para>
<Para>
If overlimit, in theory, the CBQ could throttle itself for exactly the
amount of time that was calculated to pass between packets, and then pass
one packet, and throttle again. But see the 'minburst' parameter below.
</Para>
<Para>
These are parameters you can specify in order to configure shaping:
<VariableList>
<VarListEntry>
<Term>avpkt</Term>
<ListItem>
<Para>
Average size of a packet, measured in bytes. Needed for calculating maxidle,
which is derived from maxburst, which is specified in packets.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>bandwidth</Term>
<ListItem>
<Para>
The physical bandwidth of your device, needed for idle time
calculations.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>cell</Term>
<ListItem>
<Para>
The time a packet takes to be transmitted over a device may grow in steps,
based on the packet size. An 800 and an 806 size packet may take just as long
to send, for example - this sets the granularity. Most often set to '8'.
Must be an integral power of two.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>maxburst</Term>
<ListItem>
<Para>
This number of packets is used to calculate maxidle so that when avgidle is
at maxidle, this number of average packets can be burst before avgidle drops
to 0. Set it higher to be more tolerant of bursts. You can't set maxidle
directly, only via this parameter.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>minburst</Term>
<ListItem>
<Para>
As mentioned before, CBQ needs to throttle in case of overlimit. The ideal
solution is to do so for exactly the calculated idle time, and pass 1
packet. However, Unix kernels generally have a hard time scheduling events
shorter than 10ms, so it is better to throttle for a longer period, and then
pass minburst packets in one go, and then sleep minburst times longer.
</Para>
<Para>
The time to wait is called the offtime. Higher values of minburst lead to
more accurate shaping in the long term, but to bigger bursts at millisecond
timescales.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>minidle</Term>
<ListItem>
<Para>
If avgidle is below 0, we are overlimits and need to wait until avgidle will
be big enough to send one packet. To prevent a sudden burst from shutting
down the link for a prolonged period of time, avgidle is reset to minidle if
it gets too low.
</Para>
<Para>
Minidle is specified in negative microseconds, so 10 means that avgidle is
capped at -10us.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>mpu</Term>
<ListItem>
<Para>
Minimum packet size - needed because even a zero size packet is padded
to 64 bytes on ethernet, and so takes a certain time to transmit. CBQ needs
to know this to accurately calculate the idle time.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>rate</Term>
<ListItem>
<Para>
Desired rate of traffic leaving this qdisc - this is the 'speed knob'!
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
Internally, CBQ has a lot of fine tuning. For example, classes which are
known not to have data enqueued to them aren't queried. Overlimit classes
are penalized by lowering their effective priority. All very smart &amp;
complicated.
</Para>
</Sect3>
<Sect3>
<Title>CBQ classful behaviour</Title>
<Para>
Besides shaping, using the aforementioned idletime approximations, CBQ also
acts like the PRIO queue in the sense that classes can have differing
priorities and that lower priority numbers will be polled before the higher
priority ones.
</Para>
<Para>
Each time a packet is requested by the hardware layer to be sent out to the
network, a weighted round robin process ('WRR') starts, beginning with the
lower priority classes.
</Para>
<Para>
These are then grouped and queried if they have data available. If so, it is
returned. After a class has been allowed to dequeue a number of bytes, the
next class within that priority is tried.
</Para>
<Para>
The following parameters control the WRR process:
<VariableList>
<VarListEntry>
<Term>allot</Term>
<ListItem>
<Para>
When the outer CBQ is asked for a packet to send out on the interface, it
will try all inner qdiscs (in the classes) in turn, in order of
the 'priority' parameter. Each time a class gets its turn, it can only send out
a limited amount of data. 'Allot' is the base unit of this amount. See
the 'weight' parameter for more information.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>prio</Term>
<ListItem>
<Para>
The CBQ can also act like the PRIO device. Inner classes with lower priority
are tried first and as long as they have traffic, other classes are not
polled for traffic.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>weight</Term>
<ListItem>
<Para>
Weight helps in the Weighted Round Robin process. Each class gets a chance
to send in turn. If you have classes with significantly more bandwidth than
other classes, it makes sense to allow them to send more data in one round
than the others.
</Para>
<Para>
A CBQ adds up all weights under a class, and normalizes them, so you can use
arbitrary numbers: only the ratios are important. People have been
using 'rate/10' as a rule of thumb and it appears to work well. The renormalized
weight is multiplied by the 'allot' parameter to determine how much data can
be sent in one round.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
Please note that all classes within an CBQ hierarchy need to share the same
major number!
</Para>
</Sect3>
<Sect3>
<Title>CBQ parameters that determine link sharing &amp; borrowing</Title>
<Para>
Besides purely limiting certain kinds of traffic, it is also possible to
specify which classes can borrow capacity from other classes or, conversely,
lend out bandwidth.
</Para>
<Para>
<VariableList>
<VarListEntry>
<Term>Isolated/sharing</Term>
<ListItem>
<Para>
A class that is configured with 'isolated' will not lend out bandwidth to
sibling classes. Use this if you have competing or mutually-unfriendly
agencies on your link who do want to give each other freebies.
</Para>
<Para>
The control program tc also knows about 'sharing', which is the reverse
of 'isolated'.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>bounded/borrow</Term>
<ListItem>
<Para>
A class can also be 'bounded', which means that it will not try to borrow
bandwidth from sibling classes. tc also knows about 'borrow', which is the
reverse of 'bounded'.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
A typical situation might be where you have two agencies on your link which
are both 'isolated' and 'bounded', which means that they are really limited
to their assigned rate, and also won't allow each other to borrow.
</Para>
<Para>
Within such an agency class, there might be other classes which are allowed
to swap bandwidth.
</Para>
</Sect3>
<Sect3>
<Title>Sample configuration</Title>
<Para>
This configuration limits webserver traffic to 5mbit and SMTP traffic to 3
mbit. Together, they may not get more than 6mbit. We have a 100mbit NIC and
the classes may borrow bandwidth from each other.
<Screen>
# tc qdisc add dev eth0 root handle 1:0 cbq bandwidth 100Mbit \
avpkt 1000 cell 8
# tc class add dev eth0 parent 1:0 classid 1:1 cbq bandwidth 100Mbit \
rate 6Mbit weight 0.6Mbit prio 8 allot 1514 cell 8 maxburst 20 \
avpkt 1000 bounded
</Screen>
This part installs the root and the customary 1:0 class. The 1:1 class is
bounded, so the total bandwidth can't exceed 6mbit.
</Para>
<Para>
As said before, CBQ requires a *lot* of knobs. All parameters are explained
above, however. The corresponding HTB configuration is lots simpler.
</Para>
<Para>
<Screen>
# tc class add dev eth0 parent 1:1 classid 1:3 cbq bandwidth 100Mbit \
rate 5Mbit weight 0.5Mbit prio 5 allot 1514 cell 8 maxburst 20 \
avpkt 1000
# tc class add dev eth0 parent 1:1 classid 1:4 cbq bandwidth 100Mbit \
rate 3Mbit weight 0.3Mbit prio 5 allot 1514 cell 8 maxburst 20 \
avpkt 1000
</Screen>
</Para>
<Para>
These are our two classes. Note how we scale the weight with the configured
rate. Both classes are not bounded, but they are connected to class 1:1
which is bounded. So the sum of bandwith of the 2 classes will never be
more than 6mbit. The classids need to be within the same major number as
the parent CBQ, by the way!
</Para>
<Para>
<Screen>
# tc qdisc add dev eth0 parent 1:3 handle 30: sfq
# tc qdisc add dev eth0 parent 1:4 handle 40: sfq
</Screen>
</Para>
<Para>
Both classes have a FIFO qdisc by default. But we replaced these with an SFQ
queue so each flow of data is treated equally.
<Screen>
# tc filter add dev eth0 parent 1:0 protocol ip prio 1 u32 match ip \
sport 80 0xffff flowid 1:3
# tc filter add dev eth0 parent 1:0 protocol ip prio 1 u32 match ip \
sport 25 0xffff flowid 1:4
</Screen>
</Para>
<Para>
These commands, attached directly to the root, send traffic to the right
qdiscs.
</Para>
<Para>
Note that we use 'tc class add' to CREATE classes within a qdisc, but that
we use 'tc qdisc add' to actually add qdiscs to these classes.
</Para>
<Para>
You may wonder what happens to traffic that is not classified by any of the
two rules. It appears that in this case, data will then be processed within
1:0, and be unlimited.
</Para>
<Para>
If SMTP+web together try to exceed the set limit of 6mbit/s, bandwidth will
be divided according to the weight parameter, giving 5/8 of traffic to the
webserver and 3/8 to the mail server.
</Para>
<Para>
With this configuration you can also say that webserver traffic will always
get at minimum 5/8 * 6 mbit = 3.75 mbit.
</Para>
</Sect3>
<Sect3>
<Title>Other CBQ parameters: split &amp; defmap</Title>
<Para>
As said before, a classful qdisc needs to call filters to determine
which class a packet will be enqueued to.
</Para>
<Para>
Besides calling the filter, CBQ offers other options, defmap &amp; split.
This is pretty complicated to understand, and it is not vital. But as this
is the only known place where defmap &amp; split are properly explained, I'm
doing my best.
</Para>
<Para>
As you will often want to filter on the Type of Service field only, a special
syntax is provided. Whenever the CBQ needs to figure out where a packet
needs to be enqueued, it checks if this node is a 'split node'. If so, one
of the sub-qdiscs has indicated that it wishes to receive all packets with
a certain configured priority, as might be derived from the TOS field, or
socket options set by applications.
</Para>
<Para>
The packets' priority bits are or-ed with the defmap field to see if a match
exists. In other words, this is a short-hand way of creating a very fast
filter, which only matches certain priorities. A defmap of ff (hex) will
match everything, a map of 0 nothing. A sample configuration may help make
things clearer:
</Para>
<Para>
<Screen>
# tc qdisc add dev eth1 root handle 1: cbq bandwidth 10Mbit allot 1514 \
cell 8 avpkt 1000 mpu 64
# tc class add dev eth1 parent 1:0 classid 1:1 cbq bandwidth 10Mbit \
rate 10Mbit allot 1514 cell 8 weight 1Mbit prio 8 maxburst 20 \
avpkt 1000
</Screen>
Standard CBQ preamble. I never get used to the sheer amount of numbers
required!
</Para>
<Para>
Defmap refers to TC_PRIO bits, which are defined as follows:
</Para>
<Para>
<Screen>
TC_PRIO.. Num Corresponds to TOS
-------------------------------------------------
BESTEFFORT 0 Maximize Reliablity
FILLER 1 Minimize Cost
BULK 2 Maximize Throughput (0x8)
INTERACTIVE_BULK 4
INTERACTIVE 6 Minimize Delay (0x10)
CONTROL 7
</Screen>
</Para>
<Para>
The TC_PRIO.. number corresponds to bits, counted from the right. See the
pfifo_fast section for more details how TOS bits are converted to
priorities.
</Para>
<Para>
Now the interactive and the bulk classes:
</Para>
<Para>
<Screen>
# tc class add dev eth1 parent 1:1 classid 1:2 cbq bandwidth 10Mbit \
rate 1Mbit allot 1514 cell 8 weight 100Kbit prio 3 maxburst 20 \
avpkt 1000 split 1:0 defmap c0
# tc class add dev eth1 parent 1:1 classid 1:3 cbq bandwidth 10Mbit \
rate 8Mbit allot 1514 cell 8 weight 800Kbit prio 7 maxburst 20 \
avpkt 1000 split 1:0 defmap 3f
</Screen>
</Para>
<Para>
The 'split qdisc' is 1:0, which is where the choice will be made. C0 is
binary for 11000000, 3F for 00111111, so these two together will match
everything. The first class matches bits 7 &#38; 6, and thus corresponds
to 'interactive' and 'control' traffic. The second class matches the rest.
</Para>
<Para>
Node 1:0 now has a table like this:
<Screen>
priority send to
0 1:3
1 1:3
2 1:3
3 1:3
4 1:3
5 1:3
6 1:2
7 1:2
</Screen>
</Para>
<Para>
For additional fun, you can also pass a 'change mask', which indicates
exactly which priorities you wish to change. You only need to use this if you
are running 'tc class change'. For example, to add best effort traffic to
1:2, we could run this:
</Para>
<Para>
<Screen>
# tc class change dev eth1 classid 1:2 cbq defmap 01/01
</Screen>
</Para>
<Para>
The priority map over at 1:0 now looks like this:
</Para>
<Para>
<Screen>
priority send to
0 1:2
1 1:3
2 1:3
3 1:3
4 1:3
5 1:3
6 1:2
7 1:2
</Screen>
</Para>
<Para>
FIXME: did not test 'tc class change', only looked at the source.
</Para>
</Sect3>
</Sect2>
<Sect2>
<Title>Hierarchical Token Bucket </Title>
<Para>
Martin Devera (&lt;devik&gt;) rightly realised that CBQ is complex and does
not seem optimized for many typical situations. His Hierarchical approach is
well suited for setups where you have a fixed amount of bandwidth which you
want to divide for different purposes, giving each purpose a guaranteed
bandwidth, with the possibility of specifying how much bandwidth can be
borrowed.
</Para>
<Para>
HTB works just like CBQ but does not resort to idle time calculations to
shape. Instead, it is a classful Token Bucket Filter - hence the name. It
has only a few parameters, which are well documented on his
<ULink
URL="http://luxik.cdi.cz/~devik/qos/htb/"
>site</ULink
>.
</Para>
<Para>
As your HTB configuration gets more complex, your configuration scales
well. With CBQ it is already complex even in simple cases! HTB is not yet a
part of the standard kernel, but it should soon be!
</Para>
<Para>
If you are in a position to patch your kernel, by all means consider HTB.
</Para>
<Sect3>
<Title>Sample configuration</Title>
<Para>
Functionally almost identical to the CBQ sample configuration above:
</Para>
<Para>
<Screen>
# tc qdisc add dev eth0 root handle 1: htb default 30
# tc class add dev eth0 parent 1: classid 1:1 htb rate 6mbit burst 15k
# tc class add dev eth0 parent 1:1 classid 1:10 htb rate 5mbit burst 15k
# tc class add dev eth0 parent 1:1 classid 1:20 htb rate 3mbit ceil 6mbit burst 15k
# tc class add dev eth0 parent 1:1 classid 1:30 htb rate 1kbit ceil 6mbit burst 15k
</Screen>
</Para>
<Para>
The author then recommends SFQ for beneath these classes:
<Screen>
# tc qdisc add dev eth0 parent 1:10 handle 10: sfq perturb 10
# tc qdisc add dev eth0 parent 1:20 handle 20: sfq perturb 10
# tc qdisc add dev eth0 parent 1:30 handle 30: sfq perturb 10
</Screen>
</Para>
<Para>
Add the filters which direct traffic to the right classes:
<Screen>
# U32="tc filter add dev eth0 protocol ip parent 1:0 prio 1 u32"
# $U32 match ip dport 80 0xffff flowid 1:10
# $U32 match ip sport 25 0xffff flowid 1:20
</Screen>
And that's it - no unsightly unexplained numbers, no undocumented
parameters.
</Para>
<Para>
HTB certainly looks wonderful - if 10: and 20: both have their guaranteed
bandwidth, and more is left to divide, they borrow in a 5:3 ratio, just as
you would expect.
</Para>
<Para>
Unclassified traffic gets routed to 30:, which has little bandwidth of its
own but can borrow everything that is left over. Because we chose SFQ
internally, we get fairness thrown in for free!
</Para>
</Sect3>
</Sect2>
</Sect1>
<Sect1 id="lartc.qdisc.filters">
<Title>Classifying packets with filters</Title>
<Para>
To determine which class shall process a packet, the so-called 'classifier
chain' is called each time a choice needs to be made. This chain consists of
all filters attached to the classful qdisc that needs to decide.
</Para>
<Para>To reiterate the tree, which is not a tree:
</Para>
<Screen width="80">
root 1:
|
_1:1_
/ | \
/ | \
/ | \
10: 11: 12:
/ \ / \
10:1 10:2 12:1 12:2
</Screen>
<Para>
When enqueueing a packet, at each branch the filter chain is consulted for a
relevant instruction. A typical setup might be to have a filter in 1:1 that
directs a packet to 12: and a filter on 12: that sends the packet to 12:2.
</Para>
<Para>
You might also attach this latter rule to 1:1, but you can make efficiency
gains by having more specific tests lower in the chain.
</Para>
<Para>
You can't filter a packet 'upwards', by the way. Also, with HTB, you should
attach all filters to the root!
</Para>
<Para>
And again - packets are only enqueued downwards! When they are dequeued,
they go up again, where the interface lives. They do NOT fall off the end of
the tree to the network adaptor!
</Para>
<Sect2>
<Title>Some simple filtering examples</Title>
<Para>
As explained in the Classifier chapter, you can match on literally anything,
using a very complicated syntax. To start, we will show how to do the
obvious things, which luckily are quite easy.
</Para>
<Para>
Let's say we have a PRIO qdisc called '10:' which contains three classes, and
we want to assign all traffic from and to port 22 to the highest priority
band, the filters would be:
</Para>
<Para>
<Screen>
# tc filter add dev eth0 protocol ip parent 10: prio 1 u32 match \
ip dport 22 0xffff flowid 10:1
# tc filter add dev eth0 protocol ip parent 10: prio 1 u32 match \
ip sport 80 0xffff flowid 10:1
# tc filter add dev eth0 protocol ip parent 10: prio 2 flowid 10:2
</Screen>
</Para>
<Para>
What does this say? It says: attach to eth0, node 10: a priority 1 u32
filter that matches on IP destination port 22 *exactly* and send it to band
10:1. And it then repeats the same for source port 80. The last command says
that anything unmatched so far should go to band 10:2, the next-highest
priority.
</Para>
<Para>
You need to add 'eth0', or whatever your interface is called, because each
interface has a unique namespace of handles.
</Para>
<Para>
To select on an IP address, use this:
<Screen>
# tc filter add dev eth0 parent 10:0 protocol ip prio 1 u32 \
match ip dst 4.3.2.1/32 flowid 10:1
# tc filter add dev eth0 parent 10:0 protocol ip prio 1 u32 \
match ip src 1.2.3.4/32 flowid 10:1
# tc filter add dev eth0 protocol ip parent 10: prio 2 \
flowid 10:2
</Screen>
</Para>
<Para>
This assigns traffic to 4.3.2.1 and traffic from 1.2.3.4 to the highest
priority queue, and the rest to the next-highest one.
</Para>
<Para>
You can concatenate matches, to match on traffic from 1.2.3.4 and from port
80, do this:
<Screen>
# tc filter add dev eth0 parent 10:0 protocol ip prio 1 u32 match ip src 4.3.2.1/32
match ip sport 80 0xffff flowid 10:1
</Screen>
</Para>
</Sect2>
<Sect2 id="lartc.filtering.simple">
<Title>All the filtering commands you will normally need</Title>
<Para>
Most shaping commands presented here start with this preamble:
<Screen>
# tc filter add dev eth0 parent 1:0 protocol ip prio 1 u32 ..
</Screen>
These are the so called 'u32' matches, which can match on ANY part of a
packet.
<VariableList>
<VarListEntry>
<Term>On source/destination address</Term>
<ListItem>
<Para>
Source mask 'match ip src 1.2.3.0/24', destination mask 'match ip dst
4.3.2.0/24'. To match a single host, use /32, or omit the mask.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>On source/destination port, all IP protocols</Term>
<ListItem>
<Para>
Source: 'match ip sport 80 0xffff', 'match ip dport 0xffff'
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>On ip protocol (tcp, udp, icmp, gre, ipsec)</Term>
<ListItem>
<Para>
Use the numbers from /etc/protocols, for example, icmp is 1: 'match ip
protocol 1 0xff'.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>On fwmark</Term>
<ListItem>
<Para>
You can mark packets with either ipchains and have that mark survive routing
across interfaces. This is really useful to for example only shape traffic on
eth1 that came in on eth0. Syntax:
&num; tc filter add dev eth1 protocol ip parent 1:0 prio 1 handle 6 fw flowid 1:1
Note that this is not a u32 match!
</Para>
<Para>
You can place a mark like this:
<Screen>
# iptables -A PREROUTING -t mangle -i eth0 -j MARK --set-mark 6
</Screen>
The number 6 is arbitrary.
</Para>
<Para>
If you don't want to understand the full tc filter syntax, just use
iptables, and only learn to select on fwmark.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>On the TOS field</Term>
<ListItem>
<Para>
To select interactive, minimum delay traffic:
<Screen>
# tc filter add dev ppp0 parent 1:0 protocol ip prio 10 u32 \
match ip tos 0x10 0xff \
flowid 1:4
</Screen>
Use 0x08 0xff for bulk traffic.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
For more filtering commands, see the Advanced Filters chapter.
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.imq">
<Title>The Intermediate queueing device (IMQ)</Title>
<Para>
The Intermediate queueing device is not a qdisc but its usage is tightly bound
to qdiscs. Within linux, qdiscs are attached to network devices and everything
that is queued to the device is first queued to the qdisc. From this concept,
two limitations arise:
</Para>
<Para>
1. Only egress shaping is possible (an ingress qdisc exists, but its
possibilities are very limited compared to classful qdiscs).
</Para>
<Para>
2. A qdisc can only see traffic of one interface, global limitations can't be
placed.
</Para>
<Para>
IMQ is there to help solve those two limitations. In short, you can put
everything you choose in a qdisc. Specially marked packets get intercepted
in netfilter NF_IP_PRE_ROUTING and NF_IP_POST_ROUTING hooks and pass through
the qdisc attached to an imq device. An iptables target is used for marking
the packets.
</Para>
<Para>
This enables you to do ingress shaping as you can just mark packets coming in from somewhere and/or treat interfaces as classes to set global limits.
You can also do lots of other stuff like just putting your http traffic in a
qdisc, put new connection requests in a qdisc, ...
</Para>
<Sect2>
<Title>Sample configuration</Title>
<Para>
The first thing that might come to mind is use ingress shaping to give yourself
a high guaranteed bandwidth. ;)
Configuration is just like with any other interface:
<Screen>
tc qdisc add dev imq0 root handle 1: htb default 20
tc class add dev imq0 parent 1: classid 1:1 htb rate 2mbit burst 15k
tc class add dev imq0 parent 1:1 classid 1:10 htb rate 1mbit
tc class add dev imq0 parent 1:1 classid 1:20 htb rate 1mbit
tc qdisc add dev imq0 parent 1:10 handle 10: pfifo
tc qdisc add dev imq0 parent 1:20 handle 20: sfq
tc filter add dev imq0 parent 10:0 protocol ip prio 1 u32 match \
ip dst 10.0.0.230/32 flowid 1:10
</Screen>
In this example u32 is used for classification. Other classifiers should work as
expected.
Next traffic has to be selected and marked to be enqueued to imq0.
<Screen>
iptables -t mangle -A PREROUTING -i eth0 -j IMQ --todev 0
ip link set imq0 up
</Screen>
</Para>
<Para>
The IMQ iptables targets is valid in the PREROUTING and POSTROUTING chains of
the mangle table. It's syntax is
<Screen>
IMQ [ --todev n ] n : number of imq device
</Screen>
An ip6tables target is also provided.
</Para>
<Para>
Please note traffic is not enqueued when the target is hit but afterwards.
The exact location where traffic enters the imq device depends on the
direction of the traffic (in/out).
These are the predefined netfilter hooks used by iptables:
<Screen>
enum nf_ip_hook_priorities {
NF_IP_PRI_FIRST = INT_MIN,
NF_IP_PRI_CONNTRACK = -200,
NF_IP_PRI_MANGLE = -150,
NF_IP_PRI_NAT_DST = -100,
NF_IP_PRI_FILTER = 0,
NF_IP_PRI_NAT_SRC = 100,
NF_IP_PRI_LAST = INT_MAX,
};
</Screen>
</Para>
<Para>
For ingress traffic, imq registers itself with NF_IP_PRI_MANGLE + 1 priority
which means packets enter the imq device directly after the mangle PREROUTING
chain has been passed.
</Para>
<Para>
For egress imq uses NF_IP_PRI_LAST which honours the fact that packets dropped
by the filter table won't occupy bandwidth.
</Para>
<Para>
The patches and some more information can be found at the
<ULink
URL="http://luxik.cdi.cz/~patrick/imq/"
>imq site</ULink>.
</Para>
</Sect2>
</Sect1>
</chapter>
<chapter id="lartc.loadshare">
<Title>Load sharing over multiple interfaces</Title>
<Para>
There are several ways of doing this. One of the easiest and straightforward
ways is 'TEQL' - "True" (or "trivial") link equalizer. Like most things
having to do with queueing, load sharing goes both ways. Both ends of a link
may need to participate for full effect.
</Para>
<Para>
Imagine this situation:
</Para>
<Para>
<Screen>
+-------+ eth1 +-------+
| |==========| |
'network 1' ----| A | | B |---- 'network 2'
| |==========| |
+-------+ eth2 +-------+
</Screen>
</Para>
<Para>
A and B are routers, and for the moment we'll assume both run Linux. If
traffic is going from network 1 to network 2, router A needs to distribute
the packets over both links to B. Router B needs to be configured to accept
this. Same goes the other way around, when packets go from network 2 to
network 1, router B needs to send the packets over both eth1 and eth2.
</Para>
<Para>
The distributing part is done by a 'TEQL' device, like this (it couldn't be
easier):
</Para>
<Para>
<Screen>
# tc qdisc add dev eth1 root teql0
# tc qdisc add dev eth2 root teql0
# ip link set dev teql0 up
</Screen>
</Para>
<Para>
Don't forget the 'ip link set up' command!
</Para>
<Para>
This needs to be done on both hosts. The device teql0 is basically a
roundrobbin distributor over eth1 and eth2, for sending packets. No data
ever comes in over an teql device, that just appears on the 'raw' eth1 and
eth2.
</Para>
<Para>
But now we just have devices, we also need proper routing. One way to do
this is to assign a /31 network to both links, and a /31 to the teql0 device
as well:
</Para>
<Para>
FIXME: does this need something like 'nobroadcast'? A /31 is too small to
house a network address and a broadcast address - if this doesn't work as
planned, try a /30, and adjust the ip addresses accordingly. You might even
try to make eth1 and eth2 do without an IP address!
</Para>
<Para>
On router A:
<Screen>
# ip addr add dev eth1 10.0.0.0/31
# ip addr add dev eth2 10.0.0.2/31
# ip addr add dev teql0 10.0.0.4/31
</Screen>
</Para>
<Para>
On router B:
<Screen>
# ip addr add dev eth1 10.0.0.1/31
# ip addr add dev eth2 10.0.0.3/31
# ip addr add dev teql0 10.0.0.5/31
</Screen>
</Para>
<Para>
Router A should now be able to ping 10.0.0.1, 10.0.0.3 and 10.0.0.5 over the
2 real links and the 1 equalized device. Router B should be able to ping
10.0.0.0, 10.0.0.2 and 10.0.0.4 over the links.
</Para>
<Para>
If this works, Router A should make 10.0.0.5 its route for reaching network
2, and Router B should make 10.0.0.4 its route for reaching network 1. For
the special case where network 1 is your network at home, and network 2 is
the Internet, Router A should make 10.0.0.5 its default gateway.
</Para>
<Sect1 id="lartc.loadshare.caveats">
<Title>Caveats</Title>
<Para>
Nothing is as easy as it seems. eth1 and eth2 on both router A and B need to
have return path filtering turned off, because they will otherwise drop
packets destined for ip addresses other than their own:
</Para>
<Para>
<Screen>
# echo 0 &#62; /proc/net/ipv4/conf/eth1/rp_filter
# echo 0 &#62; /proc/net/ipv4/conf/eth2/rp_filter
</Screen>
</Para>
<Para>
Then there is the nasty problem of packet reordering. Let's say 6 packets
need to be sent from A to B - eth1 might get 1, 3 and 5. eth2 would then do
2, 4 and 6. In an ideal world, router B would receive this in order, 1, 2,
3, 4, 5, 6. But the possibility is very real that the kernel gets it like
this: 2, 1, 4, 3, 6, 5. The problem is that this confuses TCP/IP. While not
a problem for links carrying many different TCP/IP sessions, you won't be
able to bundle multiple links and get to ftp a single file lots faster,
except when your receiving or sending OS is Linux, which is not easily
shaken by some simple reordering.
</Para>
<Para>
However, for lots of applications, link load balancing is a great idea.
</Para>
</Sect1>
<Sect1 id="lartc.loadshare.other">
<Title>Other possibilities</Title>
<para>
William Stearns has used an advanced tunneling setup to achieve good use of
multiple, unrelated, internet connections together. It can be found on
<ULink
URL="http://www.stearns.org/tunnel/">his tunneling page</ULink>.
</para>
<para>
The HOWTO may feature more about this in the future.
</para>
</Sect1>
</chapter>
<chapter id="lartc.netfilter">
<Title>Netfilter &amp; iproute - marking packets</Title>
<Para>
So far we've seen how iproute works, and netfilter was mentioned a few
times. This would be a good time to browse through <ULink
URL="http://netfilter.samba.org/unreliable-guides/"
>Rusty's Remarkably Unreliable Guides</ULink
>. Netfilter itself
can be found <ULink
URL="http://netfilter.filewatcher.org/"
>here</ULink
>.
</Para>
<Para>
Netfilter allows us to filter packets, or mangle their headers. One special
feature is that we can mark a packet with a number. This is done with the
--set-mark facility.
</Para>
<Para>
As an example, this command marks all packets destined for port 25, outgoing
mail:
</Para>
<Para>
<Screen>
# iptables -A PREROUTING -i eth0 -t mangle -p tcp --dport 25 \
-j MARK --set-mark 1
</Screen>
</Para>
<Para>
Let's say that we have multiple connections, one that is fast (and
expensive, per megabyte) and one that is slower, but flat fee. We would most
certainly like outgoing mail to go via the cheap route.
</Para>
<Para>
We've already marked the packets with a '1', we now instruct the routing
policy database to act on this:
</Para>
<Para>
<Screen>
# echo 201 mail.out &#62;&#62; /etc/iproute2/rt_tables
# ip rule add fwmark 1 table mail.out
# ip rule ls
0: from all lookup local
32764: from all fwmark 1 lookup mail.out
32766: from all lookup main
32767: from all lookup default
</Screen>
</Para>
<Para>
Now we generate the mail.out table with a route to the slow but cheap link:
<Screen>
# /sbin/ip route add default via 195.96.98.253 dev ppp0 table mail.out
</Screen>
</Para>
<Para>
And we are done. Should we want to make exceptions, there are lots of ways
to achieve this. We can modify the netfilter statement to exclude certain
hosts, or we can insert a rule with a lower priority that points to the main
table for our excepted hosts.
</Para>
<Para>
We can also use this feature to honour TOS bits by marking packets with a
different type of service with different numbers, and creating rules to act
on that. This way you can even dedicate, say, an ISDN line to interactive
sessions.
</Para>
<Para>
Needless to say, this also works fine on a host that's doing NAT
('masquerading').
</Para>
<Para>
IMPORTANT: We received a report that MASQ and SNAT at least collide
with marking packets. Rusty Russell explains it in
<ULink
URL="http://lists.samba.org/pipermail/netfilter/2000-November/006089.html"
>this posting</ULink
>. Turn off the reverse path filter to make it work
properly.
</Para>
<Para>
Note: to mark packets, you need to have some options enabled in your
kernel:
</Para>
<Para>
<Screen>
IP: advanced router (CONFIG_IP_ADVANCED_ROUTER) [Y/n/?]
IP: policy routing (CONFIG_IP_MULTIPLE_TABLES) [Y/n/?]
IP: use netfilter MARK value as routing key (CONFIG_IP_ROUTE_FWMARK) [Y/n/?]
</Screen>
</Para>
<Para>
See also the <xref linkend="lartc.cookbook.squid"> in the
<citetitle><xref linkend="lartc.cookbook"></citetitle>.
</Para>
</chapter>
<chapter id="lartc.adv-filter"
xreflabel="Advanced filters for (re-)classifying packets">
<Title>Advanced filters for (re-)classifying packets</Title>
<Para>
As explained in the section on classful queueing disciplines, filters are
needed to classify packets into any of the sub-queues. These filters are
called from within the classful qdisc.
</Para>
<Para>
Here is an incomplete list of classifiers available:
<VariableList>
<VarListEntry>
<Term>fw</Term>
<ListItem>
<Para>
Bases the decision on how the firewall has marked the packet. This can be
the easy way out if you don't want to learn tc filter syntax. See the
Queueing chapter for details.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>u32</Term>
<ListItem>
<Para>
Bases the decision on fields within the packet (i.e. source IP address, etc)
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>route</Term>
<ListItem>
<Para>
Bases the decision on which route the packet will be routed by
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>rsvp, rsvp6</Term>
<ListItem>
<Para>
Routes packets based on <ULink
URL="http://www.isi.edu/div7/rsvp/overview.html"
>RSVP </ULink
>. Only useful
on networks you control - the Internet does not respect RSVP.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>tcindex</Term>
<ListItem>
<Para>
Used in the DSMARK qdisc, see the relevant section.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
Note that in general there are many ways in which you can classify packet
and that it generally comes down to preference as to which system you wish
to use.
</Para>
<Para>
Classifiers in general accept a few arguments in common. They are listed
here for convenience:
</Para>
<Para>
<VariableList>
<VarListEntry>
<Term>protocol</Term>
<ListItem>
<Para>
The protocol this classifier will accept. Generally you will only be
accepting only IP traffic. Required.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>parent</Term>
<ListItem>
<Para>
The handle this classifier is to be attached to. This handle must be
an already existing class. Required.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>prio</Term>
<ListItem>
<Para>
The priority of this classifier. Lower numbers get tested first.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>handle</Term>
<ListItem>
<Para>
This handle means different things to different filters.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
All the following sections will assume you are trying to shape the traffic
going to <Literal remap="tt">HostA</Literal>. They will assume that the root class has been
configured on 1: and that the class you want to send the selected traffic to
is 1:1.
</Para>
<Sect1 id="lartc.adv-filter.u32">
<Title>The <option>u32</option> classifier</Title>
<Para>
The U32 filter is the most advanced filter available in the current
implementation. It entirely based on hashing tables, which make it
robust when there are many filter rules.
</Para>
<Para>
In its simplest form the U32 filter is a list of records, each
consisting of two fields: a selector and an action. The selectors,
described below, are compared with the currently processed IP packet
until the first match occurs, and then the associated action is performed.
The simplest type of action would be directing the packet into defined
CBQ class.
</Para>
<Para>
The command line of <Literal remap="tt">tc filter</Literal> program, used to configure the filter,
consists of three parts: filter specification, a selector and an action.
The filter specification can be defined as:
</Para>
<Para>
<Screen>
tc filter add dev IF [ protocol PROTO ]
[ (preference|priority) PRIO ]
[ parent CBQ ]
</Screen>
</Para>
<Para>
The <Literal remap="tt">protocol</Literal> field describes protocol that the filter will be
applied to. We will only discuss case of <Literal remap="tt">ip</Literal> protocol. The
<Literal remap="tt">preference</Literal> field (<Literal remap="tt">priority</Literal> can be used alternatively)
sets the priority of currently defined filter. This is important, since
you can have several filters (lists of rules) with different priorities.
Each list will be passed in the order the rules were added, then list with
lower priority (higher preference number) will be processed. The <Literal remap="tt">parent</Literal>
field defines the CBQ tree top (e.g. 1:0), the filter should be attached
to.
</Para>
<Para>
The options described above apply to all filters, not only U32.
</Para>
<Sect2>
<Title>U32 selector </Title>
<Para>
The U32 selector contains definition of the pattern, that will be matched
to the currently processed packet. Precisely, it defines which bits are
to be matched in the packet header and nothing more, but this simple
method is very powerful. Let's take a look at the following examples,
taken directly from a pretty complex, real-world filter:
</Para>
<Para>
<Screen>
# tc filter add dev eth0 protocol ip parent 1:0 pref 10 u32 \
match u32 00100000 00ff0000 at 0 flowid 1:10
</Screen>
</Para>
<Para>
For now, leave the first line alone - all these parameters describe
the filter's hash tables. Focus on the selector line, containing
<Literal remap="tt">match</Literal> keyword. This selector will match to IP headers, whose
second byte will be 0x10 (0010). As you can guess, the 00ff number is
the match mask, telling the filter exactly which bits to match. Here
it's 0xff, so the byte will match if it's exactly 0x10. The <Literal remap="tt">at</Literal>
keyword means that the match is to be started at specified offset (in
bytes) -- in this case it's beginning of the packet. Translating all
that to human language, the packet will match if its Type of Service
field will have `low delay' bits set. Let's analyze another rule:
</Para>
<Para>
<Screen>
# tc filter add dev eth0 protocol ip parent 1:0 pref 10 u32 \
match u32 00000016 0000ffff at nexthdr+0 flowid 1:10
</Screen>
</Para>
<Para>
The <Literal remap="tt">nexthdr</Literal> option means next header encapsulated in the IP packet,
i.e. header of upper-layer protocol. The match will also start here
at the beginning of the next header. The match should occur in the
second, 32-bit word of the header. In TCP and UDP protocols this field
contains packet's destination port. The number is given in big-endian
format, i.e. older bits first, so we simply read 0x0016 as 22 decimal,
which stands for SSH service if this was TCP. As you guess, this match
is ambiguous without a context, and we will discuss this later.
</Para>
<Para>
Having understood all the above, we will find the following selector
quite easy to read: <Literal remap="tt">match c0a80100 ffffff00 at 16</Literal>. What we
got here is a three byte match at 17-th byte, counting from the IP
header start. This will match for packets with destination address
anywhere in 192.168.1/24 network. After analyzing the examples, we
can summarize what we have learned.
</Para>
</Sect2>
<Sect2>
<Title>General selectors</Title>
<Para>
General selectors define the pattern, mask and offset the pattern
will be matched to the packet contents. Using the general selectors
you can match virtually any single bit in the IP (or upper layer)
header. They are more difficult to write and read, though, than
specific selectors that described below. The general selector syntax
is:
</Para>
<Para>
<Screen>
match [ u32 | u16 | u8 ] PATTERN MASK [ at OFFSET | nexthdr+OFFSET]
</Screen>
</Para>
<Para>
One of the keywords <Literal remap="tt">u32</Literal>, <Literal remap="tt">u16</Literal> or <Literal remap="tt">u8</Literal> specifies
length of the pattern in bits. PATTERN and MASK should follow, of length
defined by the previous keyword. The OFFSET parameter is the offset,
in bytes, to start matching. If <Literal remap="tt">nexthdr+</Literal> keyword is given,
the offset is relative to start of the upper layer header.
</Para>
<Para>
Some examples:
</Para>
<Para>
<Screen>
# tc filter add dev ppp14 parent 1:0 prio 10 u32 \
match u8 64 0xff at 8 \
flowid 1:4
</Screen>
</Para>
<Para>
Packet will match to this rule, if its time to live (TTL) is 64.
TTL is the field starting just after 8-th byte of the IP header.
</Para>
<Para>
<Screen>
# tc filter add dev ppp14 parent 1:0 prio 10 u32 \
match u8 0x10 0xff at nexthdr+13 \
protocol tcp \
flowid 1:3
</Screen>
</Para>
<Para>
FIXME: it has been pointed out that this syntax does not work currently.
</Para>
<Para>
Use this to match ACKs on packets smaller than 64 bytes:
</Para>
<Para>
<Screen>
## match acks the hard way,
## IP protocol 6,
## IP header length 0x5(32 bit words),
## IP Total length 0x34 (ACK + 12 bytes of TCP options)
## TCP ack set (bit 5, offset 33)
# tc filter add dev ppp14 parent 1:0 protocol ip prio 10 u32 \
match ip protocol 6 0xff \
match u8 0x05 0x0f at 0 \
match u16 0x0000 0xffc0 at 2 \
match u8 0x10 0xff at 33 \
flowid 1:3
</Screen>
</Para>
<Para>
This rule will only match TCP packets with ACK bit set, and no further
payload. Here we can see an example of using two selectors, the final result
will be logical AND of their results. If we take a look at TCP header
diagram, we can see that the ACK bit is second older bit (0x10) in the 14-th
byte of the TCP header (<Literal remap="tt">at nexthdr+13</Literal>). As for the second
selector, if we'd like to make our life harder, we could write <Literal remap="tt">match u8
0x06 0xff at 9</Literal> instead of using the specific selector <Literal remap="tt">protocol
tcp</Literal>, because 6 is the number of TCP protocol, present in 10-th byte of
the IP header. On the other hand, in this example we couldn't use any
specific selector for the first match - simply because there's no specific
selector to match TCP ACK bits.
</Para>
</Sect2>
<Sect2>
<Title>Specific selectors</Title>
<Para>
The following table contains a list of all specific selectors
the author of this section has found in the <Literal remap="tt">tc</Literal> program
source code. They simply make your life easier and increase readability
of your filter's configuration.
</Para>
<Para>
FIXME: table placeholder - the table is in separate file ,,selector.html''
</Para>
<Para>
FIXME: it's also still in Polish :-(
</Para>
<Para>
FIXME: must be sgml'ized
</Para>
<Para>
Some examples:
</Para>
<Para>
<Screen>
# tc filter add dev ppp0 parent 1:0 prio 10 u32 \
match ip tos 0x10 0xff \
flowid 1:4
</Screen>
</Para>
<Para>
FIXME: tcp dst match does not work as described below:
</Para>
<Para>
The above rule will match packets which have the TOS field set to 0x10.
The TOS field starts at second byte of the packet and is one byte big,
so we could write an equivalent general selector: <Literal remap="tt">match u8 0x10 0xff
at 1</Literal>. This gives us hint to the internals of U32 filter -- the
specific rules are always translated to general ones, and in this
form they are stored in the kernel memory. This leads to another conclusion
-- the <Literal remap="tt">tcp</Literal> and <Literal remap="tt">udp</Literal> selectors are exactly the same
and this is why you can't use single <Literal remap="tt">match tcp dst 53 0xffff</Literal>
selector to match TCP packets sent to given port -- they will also
match UDP packets sent to this port. You must remember to also specify
the protocol and end up with the following rule:
</Para>
<Para>
<Screen>
# tc filter add dev ppp0 parent 1:0 prio 10 u32 \
match tcp dst 53 0xffff \
match ip protocol 0x6 0xff \
flowid 1:2
</Screen>
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.adv-filter.route">
<Title>The <option>route</option> classifier</Title>
<Para>
This classifier filters based on the results of the routing tables. When a
packet that is traversing through the classes reaches one that is marked
with the "route" filter, it splits the packets up based on information in
the routing table.
</Para>
<Para>
<Screen>
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 route
</Screen>
</Para>
<Para>
Here we add a route classifier onto the parent node 1:0 with priority 100.
When a packet reaches this node (which, since it is the root, will happen
immediately) it will consult the routing table and if one matches will
send it to the given class and give it a priority of 100. Then, to finally
kick it into action, you add the appropriate routing entry:
</Para>
<Para>
The trick here is to define 'realm' based on either destination or source.
The way to do it is like this:
</Para>
<Para>
<Screen>
# ip route add Host/Network via Gateway dev Device realm RealmNumber
</Screen>
</Para>
<Para>
For instance, we can define our destination network 192.168.10.0 with a realm
number 10:
</Para>
<Para>
<Screen>
# ip route add 192.168.10.0/24 via 192.168.10.1 dev eth1 realm 10
</Screen>
</Para>
<Para>
When adding route filters, we can use realm numbers to represent the
networks or hosts and specify how the routes match the filters.
</Para>
<Para>
<Screen>
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 \
route to 10 classid 1:10
</Screen>
</Para>
<Para>
The above rule says packets going to the network 192.168.10.0 match class id
1:10.
</Para>
<Para>
Route filter can also be used to match source routes. For example, there is
a subnetwork attached to the Linux router on eth2.
</Para>
<Para>
<Screen>
# ip route add 192.168.2.0/24 dev eth2 realm 2
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 \
route from 2 classid 1:2
</Screen>
</Para>
<Para>
Here the filter specifies that packets from the subnetwork 192.168.2.0
(realm 2) will match class id 1:2.
</Para>
</Sect1>
<Sect1 id="lartc.adv-filter.policing">
<Title>Policing filters</Title>
<Para>
To make even more complicated setups possible, you can have filters that
only match up to a certain bandwidth. You can declare a filter to entirely
cease matching above a certain rate, or only to not match only the bandwidth
exceeding a certain rate.
</Para>
<Para>
So if you decided to police at 4mbit/s, but 5mbit/s of traffic is present,
you can stop matching either the entire 5mbit/s, or only not match 1mbit/s,
and do send 4mbit/s to the configured class.
</Para>
<Para>
If bandwidth exceeds the configured rate, you can drop a packet, reclassify
it, or see if another filter will match it.
</Para>
<Sect2>
<Title>Ways to police</Title>
<Para>
There are basically two ways to police. If you compiled the kernel
with 'Estimators', the kernel can measure for each filter how much traffic
it is passing, more or less. These estimators are very easy on the CPU, as
they simply count 25 times per second how many data has been passed, and
calculate the bitrate from that.
</Para>
<Para>
The other way works again via a Token Bucket Filter, this time living within
your filter. The TBF only matches traffic UP TO your configured bandwidth,
if more is offered, only the excess is subject to the configured overlimit
action.
</Para>
<Sect3>
<Title>With the kernel estimator</Title>
<Para>
This is very simple and has only one parameter: avrate. Either the flow
remains below avrate, and the filter classifies the traffic to the classid
configured, or your rate exceeds it in which case the specified action is
taken, which is 'reclassify' by default.
</Para>
<Para>
The kernel uses an Exponential Weighted Moving Average for your bandwidth
which makes it less sensitive to short bursts.
</Para>
</Sect3>
<Sect3>
<Title>With Token Bucket Filter</Title>
<Para>
Uses the following parameters:
<ItemizedList>
<ListItem>
<Para>
buffer/maxburst
</Para>
</ListItem>
<ListItem>
<Para>
mtu/minburst
</Para>
</ListItem>
<ListItem>
<Para>
mpu
</Para>
</ListItem>
<ListItem>
<Para>
rate
</Para>
</ListItem>
</ItemizedList>
</Para>
<Para>
Which behave mostly identical to those described in the Token Bucket Filter
section. Please note however that if you set the mtu of a TBF policer too
low, *no* packets will pass, whereas the egress TBF qdisc will just pass
them slower.
</Para>
<Para>
Another difference is that a policer can only let a packet pass, or drop it.
It cannot delay hold on to it in order to delay it.
</Para>
</Sect3>
</Sect2>
<Sect2>
<Title>Overlimit actions</Title>
<Para>
If your filter decides that it is overlimit, it can take 'actions'.
Currently, three actions are available:
<VariableList>
<VarListEntry>
<Term>continue</Term>
<ListItem>
<Para>
Causes this filter not to match, but perhaps other filters will.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>drop</Term>
<ListItem>
<Para>
This is a very fierce option which simply discards traffic exceeding a
certain rate. It is often used in the ingress policer and has limited uses.
For example, you may have a name server that falls over if offered more than
5mbit/s of packets, in which case an ingress filter could be used to make
sure no more is ever offered.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Pass/OK</Term>
<ListItem>
<Para>
Pass on traffic ok. Might be used to disable a complicated filter, but leave
it in place.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>reclassify</Term>
<ListItem>
<Para>
Most often comes down to reclassification to Best Effort. This is the
default action.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect2>
<Sect2>
<Title>Examples</Title>
<Para>
The only real example known is mentioned in the 'Protecting your host
from SYN floods' section.
</Para>
<Para>
FIXME: if you have used this, please share your experience with us
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.adv-filter.hashing">
<Title>Hashing filters for very fast massive filtering</Title>
<Para>
If you have a need for thousands of rules, for example if you have a lot of
clients or computers, all with different QoS specifications, you may find
that the kernel spends a lot of time matching all those rules.
</Para>
<Para>
By default, all filters reside in one big chain which is matched in
descending order of priority. If you have 1000 rules, 1000 checks may be
needed to determine what to do with a packet.
</Para>
<Para>
Matching would go much quicker if you would have 256 chains with each four
rules - if you could divide packets over those 256 chains, so that the right
rule will be there.
</Para>
<Para>
Hashing makes this possible. Let's say you have 1024 cable modem customers in
your network, with IP addresses ranging from 1.2.0.0 to 1.2.3.255, and each
has to go in another bin, for example 'lite', 'regular' and 'premium'. You
would then have 1024 rules like this:
</Para>
<Para>
<Screen>
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.0.0 classid 1:1
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.0.1 classid 1:1
...
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.3.254 classid 1:3
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.3.255 classid 1:2
</Screen>
</Para>
<Para>
To speed this up, we can use the last part of the IP address as a 'hash
key'. We then get 256 tables, the first of which looks like this:
<Screen>
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.0.0 classid 1:1
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.1.0 classid 1:1
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.2.0 classid 1:3
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.3.0 classid 1:2
</Screen>
</Para>
<Para>
The next one starts like this:
<Screen>
# tc filter add dev eth1 parent 1:0 protocol ip prio 100 match ip src \
1.2.0.1 classid 1:1
...
</Screen>
</Para>
<Para>
This way, only four checks are needed at most, two on average.
</Para>
<Para>
Configuration is pretty complicated, but very worth it by the time you have
this many rules. First we make a filter root, then we create a table with
256 entries:
<Screen>
# tc filter add dev eth1 parent 1:0 prio 5 protocol ip u32
# tc filter add dev eth1 parent 1:0 prio 5 handle 2: protocol ip u32 divisor 256
</Screen>
</Para>
<Para>
Now we add some rules to entries in the created table:
</Para>
<Para>
<Screen>
# tc filter add dev eth1 protocol ip parent 1:0 prio 5 u32 ht 2:7b: \
match ip src 1.2.0.123 flowid 1:1
# tc filter add dev eth1 protocol ip parent 1:0 prio 5 u32 ht 2:7b: \
match ip src 1.2.1.123 flowid 1:2
# tc filter add dev eth1 protocol ip parent 1:0 prio 5 u32 ht 2:7b: \
match ip src 1.2.3.123 flowid 1:3
# tc filter add dev eth1 protocol ip parent 1:0 prio 5 u32 ht 2:7b: \
match ip src 1.2.4.123 flowid 1:2
</Screen>
This is entry 123, which contains matches for 1.2.0.123, 1.2.1.123,
1.2.2.123, 1.2.3.123, and sends them to 1:1, 1:2, 1:3 and 1:2 respectively.
Note that we need to specify our hash bucket in hex, 0x7b is 123.
</Para>
<Para>
Next create a 'hashing filter' that directs traffic to the right entry in
the hashing table:
<Screen>
# tc filter add dev eth1 protocol ip parent 1:0 prio 5 u32 ht 800:: \
match ip src 1.2.0.0/16 \
hashkey mask 0x000000ff at 12 \
link 2:
</Screen>
Ok, some numbers need explaining. The default hash table is called 800:: and
all filtering starts there. Then we select the source address, which lives
as position 12, 13, 14 and 15 in the IP header, and indicate that we are
only interested in the last part. This we send to hash table 2:, which we
created earlier.
</Para>
<Para>
It is quite complicated, but it does work in practice and performance will
be staggering. Note that this example could be improved to the ideal case
where each chain contains 1 filter!
</Para>
</Sect1>
</chapter>
<chapter id="lartc.kernel">
<Title>Kernel network parameters </Title>
<Para>
The kernel has lots of parameters which
can be tuned for different circumstances. While, as usual, the default
parameters serve 99&percnt; of installations very well, we don't call this the
Advanced HOWTO for the fun of it!
</Para>
<Para>
The interesting bits are in /proc/sys/net, take a look there. Not everything
will be documented here initially, but we're working on it.
</Para>
<Para>
In the meantime you may want to have a look at the Linux-Kernel sources;
read the file Documentation/filesystems/proc.txt. Most of the
features are explained there.
</Para>
<Para>
(FIXME)
</Para>
<Sect1 id="lartc.kernel.rpf"
xreflabel="Reverse Path Filtering">
<Title>Reverse Path Filtering</Title>
<para>What is it? What does it do?</para>
<para>rp_filter stands for "reverse path filter" meaning the "reverse path" (reply) should be the same as the "forward path" (incoming) - based on an analysis of the available routing information. The goal in its use is to mitigate the effect of malicious network traffic using spoofed source IP addresses (RFCs 3704 and 2827 can be used as references for background on the technology - search the web for rfc2827.txt and rfc3704.txt).  What rp_filter does is determined by the value it is set to and that in turn is determined by the interactions of three settings, all located under /proc/sys/net/ipv4/conf: an "all" setting, a "default" setting and a setting per interface (network card - either real or virtual).</para>
<para>First of all, the value of the "effective" setting (see below and the rp_filter section of https://www.kernel.org/doc/Documentation/networking/ip-sysctl.txt) means:</para>
<orderedlist>
<listitem>
<para>0 - no validation, basically rp_filter is disabled</para>
</listitem>
<listitem>
<para>1 - strict validation per rfc 3704 - "... if the packet is received on the interface which would be used to forward the traffic to the source of the packet, it passes the check".</para>
</listitem>
<listitem>
<para>2 - "loose" validation per rfc 3704 - "Each incoming packet's source address is also tested against the FIB and if the source address is not reachable via any interface the packet check will fail." In other words, if any route is found for replying to the source address (including the default route) then that is acceptable. The rfc notes that this is a misnomer because there is no "reverse path".</para>
</listitem>
</orderedlist>
<para>Looking at the three settings and their interaction:</para>
<orderedlist>
<listitem>
<para>Per <ulink url="https://www.theurbanpenguin.com/rp_filter-and-lpic-3-linux-security/">https://www.theurbanpenguin.com/rp_filter-and-lpic-3-linux-security/</ulink> the "default" setting is used for any new new devices added after the initial setup.</para>
</listitem>
<listitem>
<para>The setting for each interface allows different interfaces to be handled potentially differently than others depending on the interaction between it and the "all" setting.</para>
</listitem>
<listitem>
<para>The maximum value from either the "all" setting or the interface setting is what is in effect. However, web posts indicate this may have changed over time, test before using.  An "all" setting of the following values has the given implications:</para>
</listitem>
</orderedlist>
<para>What can rp_filter can help with.</para>
<orderedlist>
<listitem>
<para>Stopping further processing of traffic with "martian" source addresses. Examples would be 127.?.?.? received on an external interface, a private IP address range from the Internet, an IP address in the experimental range.  See the above RFCs for further detail.</para>
</listitem>
<listitem>
<para>Preventing what could be called a "reflected" or "amplified" attack (as discussed in RFC3704). An example would be an attacker knowing that the target site used 10.10.10.0/24 internally and sending traffic from the Internet with source IP addresses in the 10.10.10.0/24 range.  If the network wasn't protected, not only would it have to contend with the Internet traffic load but also internal replies to assumed valid systems on the internal network.</para>
</listitem>
<listitem>
<para>If the malicious traffic was requesting a tcp connection, the potential exhaustion of tcp stack resources waiting for connection setup replies from non-existent systems. See <ulink url="https://www.theurbanpenguin.com/rp_filter-and-lpic-3-linux-security/">https://www.theurbanpenguin.com/rp_filter-and-lpic-3-linux-security/</ulink> for further discussion.</para>
</listitem>
</orderedlist>
<para>What rp_filter can't help with.</para>
<orderedlist>
<listitem>
<para>The load of receiving the malicious traffic.</para>
</listitem>
<listitem>
<para>Spoofed traffic where the source IP address is at least theoretically valid (is or could be valid if the system exists either on the Internet or internally). For theoretically valid internal addresses, the exhaustion of tcp stack resources is also not protected.</para>
</listitem>
</orderedlist>
<para>Is it active?</para>
<para>This is dependent on what the software source (Linux distribution for a given version) has done (and this changes over time) or on previous manual override of those settings. There are at least three ways of determining the current configuration:</para>
<orderedlist>
<listitem>
<para>ip netconf show all | grep rp_filter       (shows "off", "loose" or "strict")</para>
</listitem>
<listitem>
<para>sysctl -ar '\.rp_filter'       (shows 0, 1 or 2 - see above)</para>
</listitem>
<listitem>
<para>find /proc/sys/net/ipv4/conf -name rp_filter | while read a; do echo $a; cat $a; done       (shows 0, 1 or 2 - see above)</para>
</listitem>
</orderedlist>
<para>Problems it can cause.</para>
<para>As mentioned by RFC 3704, with multi-homed systems and asymmetric routing - particularly for strict mode. The latter is more exotic so a further examination of strict mode on multi-homed systems will be used to illustrate the issue.</para>
<para>Assume a multi-homed system has two interfaces: eth0 (for 10.10.1.0/24 with node address 2) and eth1 (for 10.10.2.0/24 with node address 2) and that node 1 is the gateway for each subnet thus the following routing table:</para>
<orderedlist>
<listitem>
<para>default via 10.10.1.1 dev eth0</para>
</listitem>
<listitem>
<para>10.10.1.0/24 dev eth0 proto kernel scope link src 10.10.1.2</para>
</listitem>
<listitem>
<para>10.10.2.0/24 dev eth1 proto kernel scope link src 10.10.2.2</para>
</listitem>
</orderedlist>
<para>The following options are possible:</para>
<orderedlist>
<listitem>
<para>(inbound) traffic to 10.10.1.2 (eth0's IP address) from anywhere but 10.10.2.0/24: the default route is on the same interface as the inbound traffic meeting rp_filter criteria.</para>
</listitem>
<listitem>
<para>(inbound) traffic to 10.10.1.2 (eth0's IP address) from 10.10.2.0/24: comes in on eth0 but, because the system has a route to 10.10.2.0/24, the reply will attempt to go out eth1. rp_filter will block this because of the "reply on different interface" situation.</para>
</listitem>
<listitem>
<para>traffic to 10.10.2.2 (eth1's IP address) from anywhere but 10.10.2.0/24: comes in on eth1 but, because the system either doesn't have a route to the source subnet (or the sourec subnet is 10.10.1.0/24) the default route and thus eth0 will attempt to be used. rp_filter will block this because of the "reply on different interface" situation.</para>
</listitem>
<listitem>
<para>traffic to 10.10.2.2 (eth1's IP address) from 10.10.2.0/24: comes in on eth1 and will go out on eth1 meeting rp_filter criteria.</para>
</listitem>
</orderedlist>
<para>For systems having additional interfaces the issues expand. Traffic for any local interface other than the one having the default route (and not having a specific route in the routing table) will have replies directed to the default route thus violating rp_filter's strict mode criteria.</para>
<para>Symptoms of the problem.</para>
<para>These are possible symptoms, other issues such as destination offline should also be considered.</para>
<orderedlist>
<listitem>
<para>Can connect to an IP address from its subnet but not off subnet and there are no firewall rules or, if there are firewall rules, nothing is blocking the traffic.</para>
</listitem>
<listitem>
<para>Traffic for the local system comes in but doesn't make it to the local application.</para>
</listitem>
<listitem>
<para>Traffic arrives but doesn't get routed. If there are firewall rules, nothing is blocking the traffic.</para>
</listitem>
</orderedlist>
<para>Determining if rp_filter is the problem.</para>
<para>Warning: An exception (see "Exception" below) may cause the solutions listed here to fail to show an rp_filter issue if:</para>
<orderedlist>
<listitem>
<para>nstat -az TcpExtIPReversePathFilter - shows a counter (first number) which increments</para>
</listitem>
<listitem>
<para>cat /proc/net/stat/rt_cache - check the "in_martian_src" column (the eighth column with the first column being 1) to see if it's incremented, man lnstat gives the definitions for all columns) - note it could be any row for that column since, per man lnstat, the number of rows reflects the number of cpus in the system. For a system (such as a hypervisor) with a large number of cpus this may be unwieldy.  The hexadecimal numbers here don't necessarily agree with the other methods.</para>
</listitem>
<listitem>
<para>lnstat -k rt_cache:in_martian_src will show changes real time but not the count.</para>
</listitem>
<listitem>
<para>netstat -s | grep Filter - however, if forwarding (net.ipv4.ip_forward=1) is not enabled the entry won't be found.</para>
</listitem>
<listitem>
<para>Use ip route get &lt;remote IP&gt; to determine the interface which will be used, if it's different than the entry interface then rp_filter in strict mode will drop it.</para>
</listitem>
<listitem>
<para>Enable martian logging, use sysctl -a | grep martian to find the settings or echo 1 &gt; /proc/sys/net/ipv4/conf/&lt;interface&gt;/log_martians. If the goal is to document all such drops then replace &lt;interface&gt; with "all" (without the quotes).  grep martian /var/log/syslog. any martians recorded will report: the date and time,  the source IP address of the martian, the IP address sending the packet and what interface it arrived on. </para>
</listitem>
</orderedlist>
<para>Exception</para>
<para>The following diagram shows a scenario where the rp_filter counters on a system (hypervisor but could be any multi-homed system) exhibiting problems will either be zero or not change in response to attempted access. In order to conserve space in the diagram, all references to packet sources and destinations should be understood to mean "IP address of specified interface".  Situation:</para>
<orderedlist>
<listitem>
<para>A client sends a ping to the hypervisor's br1 IP address</para>
</listitem>
<listitem>
<para>The router forwards it</para>
</listitem>
<listitem>
<para>The hypervisor is using loose mode rp_filter and there is a return path (the default route) for the packet therefore the rp_filter criteria is satisfied, no counter is incremented. However, the hypervisor has no specific route to the client and thus uses the default through br0. But, when the hypervisor swaps the source and destination IP address/port as a part of it's reply process, the br1 IP address becomes the source.</para>
</listitem>
<listitem>
<para>The packet is delivered to the router's eth1 port with source of br1 and destination of the client.</para>
</listitem>
<listitem>
<para>The router notes the br1 source on eth1 but the interface for replying to that address should be eth2. Since rp_filter is using strict mode this is a violation of the reverse path and the packet is dropped.  The client never gets a reply.</para>
</listitem>
</orderedlist>
<para>
<mediaobject>
<imageobject>
<imagedata fileref="images/rp_filterv2.eps" format="eps">
</imageobject>
<imageobject>
<imagedata fileref="images/rp_filterv2.jpg" format="jpg">
</imageobject>
</mediaobject>
</para>
<para>Solutions to a problem.</para>
<para>Some possible solutions are:</para>
<orderedlist>
<listitem>
<para>Disable rp_filter either for the interface with the problem or system-wide. Use the second and third options above under "Is it active" to find the solution appropriate to your issue.  Either change /etc/sysctrl.conf or use echo to change the value dynamically.</para>
</listitem>
<listitem>
<para>Create policy-based routing tables - for the below assume that:</para>
<itemizedlist>
<listitem>
<para>eth0 has IP address 1.2.3.4 and is the default route via
1.2.3.1in the main table.</para>
</listitem>
<listitem>
<para>eth1 is the other interface with IP address 9.8.7.6 and
gateway 9.8.7.1.</para>
</listitem>
<listitem>
<para>The alternate routing table to use was arbitrarily chosen to
be table 100. Terse examples are given, for more complex situations
n understanding of alternate routing tables and "ip rule"
configuration is needed. This is out of scope for this document,
search the web for "Linux policy based routing" to learn
more.</para>
</listitem>
</itemizedlist>
<itemizedlist>
<listitem>
<para>Find the "occupied" routing tables. All other tables in the
range 0-255 are empty and can be used as an alternate routing table.
Unfortunately, the proposed method is different for older and newer
systems.</para>
<itemizedlist>
<listitem>
<para>For older systems use: <emphasis role="bold">for i in `seq
0 255`; do if [[ `ip route sh ta $i | wc -m` -gt 0 ]]; then echo
"$i";fi;done</emphasis> - if this produces numerous "Error:
ipv4: FIB table does not exist." messages then use the newer
system method.</para>
</listitem>
<listitem>
<para>For newer systems use: <emphasis role="bold">for i in `seq
0 255`; do ip route sh ta $i 1&gt;/dev/null 2&gt;&amp;1; if [[
$? -eq 0 ]]; then echo "$i";fi;done</emphasis></para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>Setting a default route for eth1</para>
<orderedlist>
<listitem>
<para><emphasis role="bold">ip route add table 100 default via
9.8.7.1 dev eth1</emphasis></para>
</listitem>
<listitem>
<para><emphasis role="bold">ip rule add to 9.8.7.6 table
100</emphasis></para>
</listitem>
<listitem>
<para><emphasis role="bold">ip rule from 9.8.7.6 table
100</emphasis></para>
</listitem>
<listitem>
<para>This in effect disables rp_filter by setting a default
route through the interface for all traffic entering it.</para>
</listitem>
</orderedlist>
</listitem>
<listitem>
<para>Setting specific routes for eth1 - useful when all subnets
needing access via eth1 are known. Assume only various subnets in
the 10.0.0.0/8 and 192.168.0.0/16 range need access.</para>
<orderedlist>
<listitem>
<para><emphasis role="bold">ip route add table 100 10.0.0.0/8
dev eth1 via 9.8.7.1.</emphasis></para>
</listitem>
<listitem>
<para><emphasis role="bold">ip route add table 100
192.168.0.0/16 dev eth1 via 9.8.7.1</emphasis>.</para>
</listitem>
<listitem>
<para>When done, add steps 2 and 3 above.</para>
</listitem>
<listitem>
<para>This has the advantage of allowing access while retaining
rp_filter protection for all other inbound access.</para>
</listitem>
</orderedlist>
</listitem>
</itemizedlist>
</listitem>
</orderedlist>
</Sect1>
<Sect1 id="lartc.kernel.obscure">
<Title>Obscure settings</Title>
<Para>
Ok, there are a lot of parameters which can be modified. We try to list them
all. Also documented (partly) in Documentation/ip-sysctl.txt.
</Para>
<Para>
Some of these settings have different defaults based on whether you
answered 'Yes' to 'Configure as router and not host' while compiling your
kernel.
</Para>
<Sect2>
<Title>Generic ipv4</Title>
<Para>
As a generic note, most rate limiting features don't work on loopback, so
don't test them locally. The limits are supplied in 'jiffies', and are
enforced using the earlier mentioned token bucket filter.
</Para>
<Para>
The kernel has an internal clock which runs at 'HZ' ticks (or 'jiffies') per
second. On Intel, 'HZ' is mostly 100. So setting a *_rate file to, say 50,
would allow for 2 packets per second. The token bucket filter is also
configured to allow for a burst of at most 6 packets, if enough tokens have
been earned.
</Para>
<Para>
Several entries in the following list have been copied from
/usr/src/linux/Documentation/networking/ip-sysctl.txt, written by Alexey
Kuznetsov &lt;kuznet@ms2.inr.ac.ru&gt; and Andi Kleen &lt;ak@muc.de&gt;
<VariableList>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_destunreach_rate</Term>
<ListItem>
<Para>
If the kernel decides that it can't deliver a packet, it will drop it, and
send the source of the packet an ICMP notice to this effect.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_echo_ignore_all</Term>
<ListItem>
<Para>
Don't act on echo packets at all. Please don't set this by default, but if
you are used as a relay in a DoS attack, it may be useful.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_echo_ignore_broadcasts &lsqb;Useful&rsqb;</Term>
<ListItem>
<Para>
If you ping the broadcast address of a network, all hosts are supposed to
respond. This makes for a dandy denial-of-service tool. Set this to 1 to
ignore these broadcast messages.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_echoreply_rate</Term>
<ListItem>
<Para>
The rate at which echo replies are sent to any one destination.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_ignore_bogus_error_responses</Term>
<ListItem>
<Para>
Set this to ignore ICMP errors caused by hosts in the network reacting badly
to frames sent to what they perceive to be the broadcast address.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_paramprob_rate</Term>
<ListItem>
<Para>
A relatively unknown ICMP message, which is sent in response to incorrect
packets with broken IP or TCP headers. With this file you can control the
rate at which it is sent.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/icmp_timeexceed_rate</Term>
<ListItem>
<Para>
This the famous cause of the 'Solaris middle star' in traceroutes. Limits
number of ICMP Time Exceeded messages sent.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/igmp_max_memberships</Term>
<ListItem>
<Para>
Maximum number of listening igmp (multicast) sockets on the host.
FIXME: Is this true?
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/inet_peer_gc_maxtime</Term>
<ListItem>
<Para>
FIXME: Add a little explanation about the inet peer storage?
Minimum interval between garbage collection passes. This interval is in
effect under low (or absent) memory pressure on the pool. Measured in
jiffies.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/inet_peer_gc_mintime</Term>
<ListItem>
<Para>
Minimum interval between garbage collection passes. This interval is in
effect under high memory pressure on the pool. Measured in jiffies.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/inet_peer_maxttl</Term>
<ListItem>
<Para>
Maximum time-to-live of entries. Unused entries will expire after this
period of time if there is no memory pressure on the pool (i.e. when the
number of entries in the pool is very small). Measured in jiffies.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/inet_peer_minttl</Term>
<ListItem>
<Para>
Minimum time-to-live of entries. Should be enough to cover fragment
time-to-live on the reassembling side. This minimum time-to-live
is guaranteed if the pool size is less than inet_peer_threshold.
Measured in jiffies.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/inet_peer_threshold</Term>
<ListItem>
<Para>
The approximate size of the INET peer storage. Starting from this threshold
entries will be thrown aggressively. This threshold also determines
entries' time-to-live and time intervals between garbage collection passes.
More entries, less time-to-live, less GC interval.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_autoconfig</Term>
<ListItem>
<Para>
This file contains the number one if the host received its IP configuration by
RARP, BOOTP, DHCP or a similar mechanism. Otherwise it is zero.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_default_ttl</Term>
<ListItem>
<Para>
Time To Live of packets. Set to a safe 64. Raise it if you have a huge
network. Don't do so for fun - routing loops cause much more damage that
way. You might even consider lowering it in some circumstances.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_dynaddr</Term>
<ListItem>
<Para>
You need to set this if you use dial-on-demand with a dynamic interface
address. Once your demand interface comes up, any local TCP sockets which haven't seen replies will be rebound to have the right address. This solves the problem that the
connection that brings up your interface itself does not work, but the
second try does.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_forward</Term>
<ListItem>
<Para>
If the kernel should attempt to forward packets. Off by default.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_local_port_range</Term>
<ListItem>
<Para>
Range of local ports for outgoing connections. Actually quite small by
default, 1024 to 4999.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_no_pmtu_disc</Term>
<ListItem>
<Para>
Set this if you want to disable Path MTU discovery - a technique to
determine the largest Maximum Transfer Unit possible on your path. See also
the section on Path MTU discovery in the
<citetitle><xref linkend="lartc.cookbook"></citetitle> chapter.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ipfrag_high_thresh</Term>
<ListItem>
<Para>
Maximum memory used to reassemble IP fragments. When
ipfrag_high_thresh bytes of memory is allocated for this purpose,
the fragment handler will toss packets until ipfrag_low_thresh
is reached.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ip_nonlocal_bind</Term>
<ListItem>
<Para>
Set this if you want your applications to be able to bind to an address
which doesn't belong to a device on your system. This can be useful when
your machine is on a non-permanent (or even dynamic) link, so your services
are able to start up and bind to a specific address when your link is down.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ipfrag_low_thresh</Term>
<ListItem>
<Para>
Minimum memory used to reassemble IP fragments.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/ipfrag_time</Term>
<ListItem>
<Para>
Time in seconds to keep an IP fragment in memory.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_abort_on_overflow</Term>
<ListItem>
<Para>
A boolean flag controlling the behaviour under lots of incoming connections.
When enabled, this causes the kernel to actively send RST packets when a
service is overloaded.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_fin_timeout</Term>
<ListItem>
<Para>
Time to hold socket in state FIN-WAIT-2, if it was closed by our side. Peer
can be broken and never close its side, or even died unexpectedly. Default
value is 60sec. Usual value used in 2.2 was 180 seconds, you may restore it,
but remember that if your machine is even underloaded WEB server, you risk
to overflow memory with kilotons of dead sockets, FIN-WAIT-2 sockets are
less dangerous than FIN-WAIT-1, because they eat maximum 1.5K of memory, but
they tend to live longer. Cf. tcp_max_orphans.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_keepalive_time</Term>
<ListItem>
<Para>
How often TCP sends out keepalive messages when keepalive is enabled.
Default: 2hours.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_keepalive_intvl</Term>
<ListItem>
<Para>
How frequent probes are retransmitted, when a probe isn't acknowledged.
Default: 75 seconds.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_keepalive_probes</Term>
<ListItem>
<Para>
How many keepalive probes TCP will send, until it decides that the
connection is broken.
Default value: 9.
Multiplied with tcp_keepalive_intvl, this gives the time a link can be
non-responsive after a keepalive has been sent.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_max_orphans</Term>
<ListItem>
<Para>
Maximal number of TCP sockets not attached to any user file handle, held by
system. If this number is exceeded orphaned connections are reset
immediately and warning is printed. This limit exists only to prevent simple
DoS attacks, you _must_ not rely on this or lower the limit artificially,
but rather increase it (probably, after increasing installed memory), if
network conditions require more than default value, and tune network
services to linger and kill such states more aggressively. Let me remind you
again: each orphan eats up to &nbsp;64K of unswappable memory.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_orphan_retries</Term>
<ListItem>
<Para>
How may times to retry before killing TCP connection, closed by our side.
Default value 7 corresponds to &nbsp;50sec-16min depending on RTO. If your machine
is a loaded WEB server, you should think about lowering this value, such
sockets may consume significant resources. Cf. tcp_max_orphans.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_max_syn_backlog</Term>
<ListItem>
<Para>
Maximal number of remembered connection requests, which still did not
receive an acknowledgment from connecting client. Default value is 1024 for
systems with more than 128Mb of memory, and 128 for low memory machines. If
server suffers of overload, try to increase this number. Warning! If you
make it greater than 1024, it would be better to change TCP_SYNQ_HSIZE in
include/net/tcp.h to keep TCP_SYNQ_HSIZE*16&#60;=tcp_max_syn_backlog and to
recompile kernel.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_max_tw_buckets</Term>
<ListItem>
<Para>
Maximal number of timewait sockets held by system simultaneously. If this
number is exceeded time-wait socket is immediately destroyed and warning is
printed. This limit exists only to prevent simple DoS attacks, you _must_
not lower the limit artificially, but rather increase it (probably, after
increasing installed memory), if network conditions require more than
default value.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_retrans_collapse</Term>
<ListItem>
<Para>
Bug-to-bug compatibility with some broken printers.
On retransmit try to send bigger packets to work around bugs in
certain TCP stacks.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_retries1</Term>
<ListItem>
<Para>
How many times to retry before deciding that something is wrong
and it is necessary to report this suspicion to network layer.
Minimal RFC value is 3, it is default, which corresponds
to &nbsp;3sec-8min depending on RTO.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_retries2</Term>
<ListItem>
<Para>
How may times to retry before killing alive TCP connection.
<ULink
URL="http://www.ietf.org/rfc/rfc1122.txt"
>RFC 1122</ULink
>
says that the limit should be longer than 100 sec.
It is too small number. Default value 15 corresponds to &nbsp;13-30min
depending on RTO.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_rfc1337</Term>
<ListItem>
<Para>
This boolean enables a fix for 'time-wait assassination hazards in tcp', described
in RFC 1337. If enabled, this causes the kernel to drop RST packets for
sockets in the time-wait state.
Default: 0
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_sack</Term>
<ListItem>
<Para>
Use Selective ACK which can be used to signify that specific packets are
missing - therefore helping fast recovery.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_stdurg</Term>
<ListItem>
<Para>
Use the Host requirements interpretation of the TCP urg pointer
field.
Most hosts use the older BSD interpretation, so if you turn this on
Linux might not communicate correctly with them.
Default: FALSE
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_syn_retries</Term>
<ListItem>
<Para>
Number of SYN packets the kernel will send before giving up on the new
connection.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_synack_retries</Term>
<ListItem>
<Para>
To open the other side of the connection, the kernel sends a SYN with a
piggybacked ACK on it, to acknowledge the earlier received SYN. This is part
2 of the threeway handshake. This setting determines the number of SYN+ACK
packets sent before the kernel gives up on the connection.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_timestamps</Term>
<ListItem>
<Para>
Timestamps are used, amongst other things, to protect against wrapping
sequence numbers. A 1 gigabit link might conceivably re-encounter a previous
sequence number with an out-of-line value, because it was of a previous
generation. The timestamp will let it recognize this 'ancient packet'.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_tw_recycle</Term>
<ListItem>
<Para>
Enable fast recycling TIME-WAIT sockets. Default value is 1.
It should not be changed without advice/request of technical experts.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/tcp_window_scaling</Term>
<ListItem>
<Para>
TCP/IP normally allows windows up to 65535 bytes big. For really fast
networks, this may not be enough. The window scaling options allows for
almost gigabyte windows, which is good for high bandwidth*delay products.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect2>
<Sect2>
<Title>Per device settings</Title>
<Para>
DEV can either stand for a real interface, or for 'all' or 'default'.
Default also changes settings for interfaces yet to be created.
<VariableList>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/accept_redirects</Term>
<ListItem>
<Para>
If a router decides that you are using it for a wrong purpose (ie, it needs
to resend your packet on the same interface), it will send us a ICMP
Redirect. This is a slight security risk however, so you may want to turn it
off, or use secure redirects.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/accept_source_route</Term>
<ListItem>
<Para>
Not used very much anymore. You used to be able to give a packet a list of
IP addresses it should visit on its way. Linux can be made to honor this IP
option.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/bootp_relay</Term>
<ListItem>
<Para>
Accept packets with source address 0.b.c.d with destinations not to this host
as local ones. It is supposed that a BOOTP relay daemon will catch and forward
such packets.
</Para>
<Para>
The default is 0, since this feature is not implemented yet (kernel version
2.2.12).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/forwarding</Term>
<ListItem>
<Para>
Enable or disable IP forwarding on this interface.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/log_martians</Term>
<ListItem>
<Para>
See the section on
<citetitle><xref linkend="lartc.kernel.rpf"></citetitle>.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/mc_forwarding</Term>
<ListItem>
<Para>
If we do multicast forwarding on this interface
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/proxy_arp</Term>
<ListItem>
<Para>
If you set this to 1, this interface will respond to ARP requests for
addresses the kernel has routes to. Can be very useful when building 'ip
pseudo bridges'. Do take care that your netmasks are very correct before
enabling this! Also be aware that the rp_filter, mentioned elsewhere, also
operates on ARP queries!
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/rp_filter</Term>
<ListItem>
<Para>
See the section on
<citetitle><xref linkend="lartc.kernel.rpf"></citetitle>.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/secure_redirects</Term>
<ListItem>
<Para>
Accept ICMP redirect messages only for gateways, listed in default gateway
list. Enabled by default.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/send_redirects</Term>
<ListItem>
<Para>
If we send the above mentioned redirects.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/shared_media</Term>
<ListItem>
<Para>
If it is not set the kernel does not assume that different subnets on this
device can communicate directly. Default setting is 'yes'.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/conf/DEV/tag</Term>
<ListItem>
<Para>
FIXME: fill this in
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect2>
<Sect2>
<Title>Neighbor policy</Title>
<Para>
Dev can either stand for a real interface, or for 'all' or 'default'.
Default also changes settings for interfaces yet to be created.
<VariableList>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/anycast_delay</Term>
<ListItem>
<Para>
Maximum for random delay of answers to neighbor solicitation messages in
jiffies (1/100 sec). Not yet implemented (Linux does not have anycast support
yet).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/app_solicit</Term>
<ListItem>
<Para>
Determines the number of requests to send to the user level ARP daemon. Use 0
to turn off.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/base_reachable_time</Term>
<ListItem>
<Para>
A base value used for computing the random reachable time value as specified
in RFC2461.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/delay_first_probe_time</Term>
<ListItem>
<Para>
Delay for the first time probe if the neighbor is reachable. (see
gc_stale_time)
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/gc_stale_time</Term>
<ListItem>
<Para>
Determines how often to check for stale ARP entries. After an ARP entry is
stale it will be resolved again (which is useful when an IP address migrates
to another machine). When ucast_solicit is greater than 0 it first tries to
send an ARP packet directly to the known host When that fails and
mcast_solicit is greater than 0, an ARP request is broadcast.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/locktime</Term>
<ListItem>
<Para>
An ARP/neighbor entry is only replaced with a new one if the old is at least
locktime old. This prevents ARP cache thrashing.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/mcast_solicit</Term>
<ListItem>
<Para>
Maximum number of retries for multicast solicitation.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/proxy_delay</Term>
<ListItem>
<Para>
Maximum time (real time is random &lsqb;0..proxytime]) before answering to an ARP
request for which we have an proxy ARP entry. In some cases, this is used to
prevent network flooding.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/proxy_qlen</Term>
<ListItem>
<Para>
Maximum queue length of the delayed proxy arp timer. (see proxy_delay).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/retrans_time</Term>
<ListItem>
<Para>
The time, expressed in jiffies (1/100 sec), between retransmitted Neighbor
Solicitation messages. Used for address resolution and to determine if a
neighbor is unreachable.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/ucast_solicit</Term>
<ListItem>
<Para>
Maximum number of retries for unicast solicitation.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/neigh/DEV/unres_qlen</Term>
<ListItem>
<Para>
Maximum queue length for a pending arp request - the number of packets which
are accepted from other layers while the ARP address is still resolved.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Internet QoS: Architectures and Mechanisms for Quality of Service,
Zheng Wang, ISBN 1-55860-608-4</Term>
<ListItem>
<Para>
Hardcover textbook covering topics
related to Quality of Service. Good for understanding basic concepts.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect2>
<Sect2>
<Title>Routing settings</Title>
<Para>
<VariableList>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/error_burst</Term>
<ListItem>
<Para>
These parameters are used to limit the warning messages written to the kernel
log from the routing code. The higher the error_cost factor is, the fewer
messages will be written. Error_burst controls when messages will be dropped.
The default settings limit warning messages to one every five seconds.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/error_cost</Term>
<ListItem>
<Para>
These parameters are used to limit the warning messages written to the kernel
log from the routing code. The higher the error_cost factor is, the fewer
messages will be written. Error_burst controls when messages will be dropped.
The default settings limit warning messages to one every five seconds.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/flush</Term>
<ListItem>
<Para>
Writing to this file results in a flush of the routing cache.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/gc_elasticity</Term>
<ListItem>
<Para>
Values to control the frequency and behavior of the garbage collection
algorithm for the routing cache. This can be important for when doing
fail over. At least gc_timeout seconds will elapse before Linux will skip
to another route because the previous one has died. By default set to 300,
you may want to lower it if you want to have a speedy fail over.
</Para>
<Para>
Also see <ULink
URL="http://mailman.ds9a.nl/pipermail/lartc/2002q1/002667.html"
>this post</ULink
> by Ard van Breemen.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/gc_interval</Term>
<ListItem>
<Para>
See /proc/sys/net/ipv4/route/gc_elasticity.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/gc_min_interval</Term>
<ListItem>
<Para>
See /proc/sys/net/ipv4/route/gc_elasticity.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/gc_thresh</Term>
<ListItem>
<Para>
See /proc/sys/net/ipv4/route/gc_elasticity.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/gc_timeout</Term>
<ListItem>
<Para>
See /proc/sys/net/ipv4/route/gc_elasticity.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/max_delay</Term>
<ListItem>
<Para>
Delays for flushing the routing cache.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/max_size</Term>
<ListItem>
<Para>
Maximum size of the routing cache. Old entries will be purged once the cache
reached has this size.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/min_adv_mss</Term>
<ListItem>
<Para>
FIXME: fill this in
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/min_delay</Term>
<ListItem>
<Para>
Delays for flushing the routing cache.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/min_pmtu</Term>
<ListItem>
<Para>
FIXME: fill this in
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/mtu_expires</Term>
<ListItem>
<Para>
FIXME: fill this in
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/redirect_load</Term>
<ListItem>
<Para>
Factors which determine if more ICMP redirects should be sent to a specific
host. No redirects will be sent once the load limit or the maximum number of
redirects has been reached.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/redirect_number</Term>
<ListItem>
<Para>
See /proc/sys/net/ipv4/route/redirect_load.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>/proc/sys/net/ipv4/route/redirect_silence</Term>
<ListItem>
<Para>
Timeout for redirects. After this period redirects will be sent again, even if
this has been stopped, because the load or number limit has been reached.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect2>
</Sect1>
</chapter>
<chapter id="lartc.adv-qdisc">
<Title>Advanced &amp; less common queueing disciplines</Title>
<Para>
Should you find that you have needs not addressed by the queues mentioned
earlier, the kernel contains some other more specialized queues mentioned here.
</Para>
<Sect1 id="lartc.adv-qdisc.bfifo-pfifo">
<Title><literal>bfifo</literal>/<literal>pfifo</literal></Title>
<Para>
These classless queues are even simpler than pfifo_fast in that they lack
the internal bands - all traffic is really equal. They have one important
benefit though, they have some statistics. So even if you don't need shaping
or prioritizing, you can use this qdisc to determine the backlog on your
interface.
</Para>
<Para>
pfifo has a length measured in packets, bfifo in bytes.
</Para>
<Sect2>
<Title>Parameters &amp; usage</Title>
<Para>
<VariableList>
<VarListEntry>
<Term>limit</Term>
<ListItem>
<Para>
Specifies the length of the queue. Measured in bytes for bfifo, in packets
for pfifo. Defaults to the interface txqueuelen (see pfifo_fast chapter)
packets long or txqueuelen*mtu bytes for bfifo.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.adv-qdisc.csz">
<Title>Clark-Shenker-Zhang algorithm (CSZ)</Title>
<Para>
This is so theoretical that not even Alexey (the main CBQ author) claims to
understand it. From his source:
</Para>
<blockquote>
<Para>
David D. Clark, Scott Shenker and Lixia Zhang
<citetitle>Supporting Real-Time Applications in an Integrated Services Packet
Network: Architecture and Mechanism</citetitle>.
</Para>
<Para>
As I understand it, the main idea is to create WFQ flows for each guaranteed
service and to allocate the rest of bandwith to dummy flow-0. Flow-0
comprises the predictive services and the best effort traffic; it is handled
by a priority scheduler with the highest priority band allocated for
predictive services, and the rest --- to the best effort packets.
</Para>
<Para>
Note that in CSZ flows are NOT limited to their bandwidth. It is supposed
that the flow passed admission control at the edge of the QoS network and it
doesn't need further shaping. Any attempt to improve the flow or to shape it
to a token bucket at intermediate hops will introduce undesired delays and
raise jitter.
</Para>
<Para>
At the moment CSZ is the only scheduler that provides true guaranteed
service. Another schemes (including CBQ) do not provide guaranteed delay and
randomize jitter."
</Para>
<Para>
Does not currently seem like a good candidate to use, unless you've read and
understand the article mentioned.
</Para>
</blockquote>
</Sect1>
<Sect1 id="lartc.adv-qdisc.dsmark"
xreflabel="DSMARK">
<Title>DSMARK</Title>
<abstract>
<Para>
<author><firstname>Esteve</firstname><surname>Camps</surname></author>
<address><email>marvin@grn.es</email></address>
This text is an extract from my thesis on
<citetitle>QoS Support in Linux</citetitle>, September 2000.
</Para>
</abstract>
<Para>Source documents:
</para>
<ItemizedList>
<ListItem>
<Para>
<ULink URL="ftp://icaftp.epfl.ch/pub/linux/diffserv/misc/dsid-01.txt.gz">
Draft-almesberger-wajhak-diffserv-linux-01.txt</ULink>.
</Para>
</ListItem>
<ListItem>
<Para>Examples in iproute2 distribution.
</Para>
</ListItem>
<ListItem>
<Para>
<ULink URL="http://www.qosforum.com/white-papers/qosprot_v3.pdf">
White Paper-QoS protocols and architectures</ULink> and
<ULink URL="http://www.qosforum.com/docs/faq">
IP QoS Frequently Asked Questions</ULink> both by
<citetitle>Quality of Service Forum</citetitle>.
</Para>
</ListItem>
</ItemizedList>
<Para>
This chapter was written by Esteve Camps &lt;esteve@hades.udg.es&gt;.
</Para>
<Sect2>
<Title>Introduction</Title>
<Para>
First of all, first of all, it would be a great idea for you to read RFCs
written about this (RFC2474, RFC2475, RFC2597 and RFC2598) at
<ULink URL="http://www.ietf.org/html.charters/diffserv-charter.html">
IETF DiffServ working Group web site</ULink> and
<ULink URL="http://diffserv.sf.net/">
Werner Almesberger web site</ULink>
(he wrote the code to support Differentiated Services on Linux).
</Para>
</Sect2>
<Sect2>
<Title>What is Dsmark related to?</Title>
<Para>
Dsmark is a queueing discipline that offers the capabilities needed in
Differentiated Services (also called DiffServ or, simply, DS). DiffServ is
one of two actual QoS architectures (the other one is called Integrated
Services) that is based on a value carried by packets in the DS field of the
IP header.
</Para>
<Para>
One of the first solutions in IP designed to offer some QoS level was
the Type of Service field (TOS byte) in IP header. By changing that value,
we could choose a high/low level of throughput, delay or reliability.
But this didn't provide sufficient flexibility to the needs of new
services (such as real-time applications, interactive applications and
others). After this, new architectures appeared. One of these was DiffServ
which kept TOS bits and renamed DS field.
</Para>
</Sect2>
<Sect2>
<Title>Differentiated Services guidelines</Title>
<Para>
Differentiated Services is group-oriented. I mean, we don't know anything
about flows (this will be the Integrated Services purpose); we know about
flow aggregations and we will apply different behaviours depending on which
aggregation a packet belongs to.
</Para>
<Para>
When a packet arrives to an edge node (entry node to a DiffServ domain)
entering to a DiffServ Domain we'll have to policy, shape and/or mark those
packets (marking refers to assigning a value to the DS field. It's just like the
cows :-) ). This will be the mark/value that the internal/core nodes on our
DiffServ Domain will look at to determine which behaviour or QoS level
apply.
</Para>
<Para>
As you can deduce, Differentiated Services involves a domain on which
all DS rules will have to be applied. In fact you can think I
will classify all the packets entering my domain. Once they enter my
domain they will be subjected to the rules that my classification dictates
and every traversed node will apply that QoS level.
</Para>
<Para>
In fact, you can apply your own policies into your local domains, but some
<Emphasis>Service Level Agreements</Emphasis> should be considered when connecting to
other DS domains.
</Para>
<Para>
At this point, you maybe have a lot of questions. DiffServ is more than I've
explained. In fact, you can understand that I can not resume more than 3
RFCs in just 50 lines :-).
</Para>
</Sect2>
<Sect2>
<Title>Working with Dsmark</Title>
<Para>
As the DiffServ bibliography specifies, we differentiate boundary nodes and
interior nodes. These are two important points in the traffic path. Both
types perform a classification when the packets arrive. Its result may be
used in different places along the DS process before the packet is released
to the network. It's just because of this that the diffserv code supplies an
structure called sk_buff, including a new field called skb-&gt;tc_index
where we'll store the result of initial classification that may be used in
several points in DS treatment.
</Para>
<Para>
The skb-&gt;tc_index value will be initially set by the DSMARK qdisc,
retrieving it from the DS field in IP header of every received packet.
Besides, cls_tcindex classifier will read all or part of skb-&gt;tcindex
value and use it to select classes.
</Para>
<Para>
But, first of all, take a look at DSMARK qdisc command and its parameters:
<Screen>
... dsmark indices INDICES [ default_index DEFAULT_INDEX ] [ set_tc_index ]
</Screen>
What do these parameters mean?
<ItemizedList>
<ListItem>
<Para>
<Emphasis remap="bf">indices</Emphasis>: size of table of (mask,value) pairs. Maximum value is 2&circ;n, where n&gt;=0.
</Para>
</ListItem>
<ListItem>
<Para>
<Emphasis remap="bf">Default_index</Emphasis>: the default table entry index if classifier finds no match.
</Para>
</ListItem>
<ListItem>
<Para>
<Emphasis remap="bf">Set_tc_index</Emphasis>: instructs dsmark discipline to retrieve the DS field and store it onto skb-&gt;tc_index.
</Para>
</ListItem>
</ItemizedList>
Let's see the DSMARK process.
</Para>
</Sect2>
<Sect2>
<Title>How SCH_DSMARK works.</Title>
<Para>
This qdisc will apply the next steps:
<ItemizedList>
<ListItem>
<Para>
If we have declared set_tc_index option in qdisc command, DS field is retrieved and stored onto
skb-&gt;tc_index variable.
</Para>
</ListItem>
<ListItem>
<Para>
Classifier is invoked. The classifier will be executed and it will return a class ID that will be stored in
skb-&gt;tc_index variable.If no filter matches are found, we consider the default_index option to be the
classId to store. If neither set_tc_index nor default_index has been declared results may be
unpredictable.
</Para>
</ListItem>
<ListItem>
<Para>
After been sent to internal qdiscs where you can reuse the result of the filter, the classid returned by
the internal qdisc is stored into skb-&gt;tc_index. We will use this value in the future to index a mask-
value table. The final result to assign to the packet will be that resulting from next operation:
<Screen>
New_Ds_field = ( Old_DS_field &#38; mask ) | value
</Screen>
</Para>
</ListItem>
<ListItem>
<Para>
Thus, new value will result from "anding" ds_field and mask values and next, this result "ORed" with
value parameter. See next diagram to understand all this process:
</Para>
</ListItem>
</ItemizedList>
<Screen>
skb-&gt;ihp-&gt;tos
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - &gt;
| | ^
| -- If you declare set_tc_index, we set DS | | &lt;-----May change
| value into skb-&gt;tc_index variable | |O DS field
| A| |R
+-|-+ +------+ +---+-+ Internal +-+ +---N|-----|----+
| | | | tc |---&gt;| | |--&gt; . . . --&gt;| | | D| | |
| | |-----&gt;|index |---&gt;| | | Qdisc | |----&gt;| v | |
| | | |filter|---&gt;| | | +---------------+ | ----&gt;(mask,value) |
--&gt;| O | +------+ +-|-+--------------^----+ / | (. , .) |
| | | ^ | | | | (. , .) |
| | +----------|---------|----------------|-------|--+ (. , .) |
| | sch_dsmark | | | | |
+-|------------|---------|----------------|-------|------------------+
| | | &lt;- tc_index -&gt; | |
| |(read) | may change | | &lt;--------------Index to the
| | | | | (mask,value)
v | v v | pairs table
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -&gt;
skb-&gt;tc_index
</Screen>
</Para>
<Para>
How to do marking? Just change the mask and value of the class you want to remark. See next line of code:
<Screen>
tc class change dev eth0 classid 1:1 dsmark mask 0x3 value 0xb8
</Screen>
This changes the (mask,value) pair in hash table, to remark packets belonging to class 1:1.You have to "change" this values
because of default values that (mask,value) gets initially (see table below).
</Para>
<Para>
Now, we'll explain how TC_INDEX filter works and how fits into this. Besides, TCINDEX filter can be
used in other configurations rather than those including DS services.
</Para>
</Sect2>
<Sect2>
<Title>TC_INDEX Filter</Title>
<Para>
This is the basic command to declare a TC_INDEX filter:
<Screen>
... tcindex [ hash SIZE ] [ mask MASK ] [ shift SHIFT ]
[ pass_on | fall_through ]
[ classid CLASSID ] [ police POLICE_SPEC ]
</Screen>
Next, we show the example used to explain TC_INDEX operation mode. Pay attention to bolded words:
tc qdisc add dev eth0 handle 1:0 root dsmark indices 64 <Emphasis remap="bf">set_tc_index</Emphasis>
tc filter add dev eth0 parent 1:0 protocol ip prio 1 tcindex <Emphasis remap="bf">mask 0xfc shift 2</Emphasis>
tc qdisc add dev eth0 parent 1:0 handle 2:0 cbq bandwidth 10Mbit cell 8 avpkt 1000 mpu 64
&num; EF traffic class
tc class add dev eth0 parent 2:0 classid 2:1 cbq bandwidth 10Mbit rate 1500Kbit avpkt 1000 prio 1 bounded isolated allot 1514 weight 1 maxburst 10
&num; Packet fifo qdisc for EF traffic
tc qdisc add dev eth0 parent 2:1 pfifo limit 5
tc filter add dev eth0 parent 2:0 protocol ip prio 1 <Emphasis remap="bf">handle 0x2e</Emphasis> tcindex <Emphasis remap="bf">classid 2:1 pass_on</Emphasis>
(This code is not complete. It's just an extract from EFCBQ example included in iproute2 distribution).
</Para>
<Para>
First of all, suppose we receive a packet marked as EF . If you read RFC2598, you'll see that DSCP
recommended value for EF traffic is 101110. This means that DS field will be 10111000 (remember that
less significant bits in TOS byte are not used in DS) or 0xb8 in hexadecimal codification.
</Para>
<Para>
<Screen>
TC INDEX
FILTER
+---+ +-------+ +---+-+ +------+ +-+ +-------+
| | | | | | | |FILTER| +-+ +-+ | | | |
| |-----&gt;| MASK | -&gt; | | | -&gt; |HANDLE|-&gt;| | | | -&gt; | | -&gt; | |
| | . | =0xfc | | | | |0x2E | | +----+ | | | | |
| | . | | | | | +------+ +--------+ | | | |
| | . | | | | | | | | |
--&gt;| | . | SHIFT | | | | | | | |--&gt;
| | . | =2 | | | +----------------------------+ | | |
| | | | | | CBQ 2:0 | | |
| | +-------+ +---+--------------------------------+ | |
| | | |
| +-------------------------------------------------------------+ |
| DSMARK 1:0 |
+-------------------------------------------------------------------------+
</Screen>
</Para>
<Para>
The packet arrives, then, set with 0xb8 value at DS field. As we explained before, dsmark qdisc identified
by 1:0 id in the example, retrieves DS field and store it in skb-&gt;tc_index variable.
Next step in the example will correspond to the filter associated to this qdisc (second line in the example).
This will perform next operations:
<Screen>
Value1 = skb-&#62;tc_index &#38; MASK
Key = Value1 &#62;&#62; SHIFT
</Screen>
</Para>
<Para>
In the example, MASK=0xFC i SHIFT=2.
<Screen>
Value1 = 10111000 &#38; 11111100 = 10111000
Key = 10111000 &#62;&#62; 2 = 00101110 -&#62; 0x2E in hexadecimal
</Screen>
</Para>
<Para>
The returned value will correspond to a qdisc internal filter handle (in the example, identifier 2:0). If a
filter with this id exists, policing and metering conditions will be verified (in case that filter includes this)
and the classid will be returned (in our example, classid 2:1) and stored in skb-&gt;tc_index variable.
</Para>
<Para>
But if any filter with that identifier is found, the result will depend on fall_through flag declaration. If so,
value key is returned as classid. If not, an error is returned and process continues with the rest filters. Be
careful if you use fall_through flag; this can be done if a simple relation exists between values
of skb-&gt;tc_index variable and class id's.
</Para>
<Para>
The latest parameters to comment on are hash and pass_on. The first one
relates to hash table size. Pass_on will be used to indicate that if no classid
equal to the result of this filter is found, try next filter.
The default action is fall_through (look at next table).
</Para>
<Para>
Finally, let's see which possible values can be set to all this TCINDEX parameters:
<Screen>
TC Name Value Default
-----------------------------------------------------------------
Hash 1...0x10000 Implementation dependent
Mask 0...0xffff 0xffff
Shift 0...15 0
Fall through / Pass_on Flag Fall_through
Classid Major:minor None
Police ..... None
</Screen>
</Para>
<Para>
This kind of filter is very powerful. It's necessary to explore all possibilities. Besides, this filter is not only used in DiffServ configurations.
You can use it as any other kind of filter.
</Para>
<Para>
I recommend you to look at all DiffServ examples included in iproute2 distribution. I promise I will try to
complement this text as soon as I can. Besides, all I have explained is the result of a lot of tests.
I would thank you tell me if I'm wrong in any point.
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.adv-qdisc.ingress">
<Title>Ingress qdisc</Title>
<Para>
All qdiscs discussed so far are egress qdiscs. Each interface however can
also have an ingress qdisc which is not used to send packets
out to the network adaptor. Instead, it allows you to apply tc filters to
packets coming in over the interface, regardless of whether they have a local
destination or are to be forwarded.
</Para>
<Para>
As the tc filters contain a full Token Bucket Filter implementation, and are
also able to match on the kernel flow estimator, there is a lot of
functionality available. This effectively allows you to police incoming
traffic, before it even enters the IP stack.
</Para>
<Sect2>
<Title>Parameters &amp; usage</Title>
<Para>
The ingress qdisc itself does not require any parameters. It differs from
other qdiscs in that it does not occupy the root of a device. Attach it like
this:
<Screen>
# tc qdisc add dev eth0 ingress
</Screen>
This allows you to have other, sending, qdiscs on your device besides the
ingress qdisc.
</Para>
<Para>
For a contrived example how the ingress qdisc could be used, see the
Cookbook.
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.adv-qdisc.red">
<Title>Random Early Detection (RED)</Title>
<Para>
This section is meant as an introduction to backbone routing, which often
involves &lt;100 megabit bandwidths, which requires a different approach than
your ADSL modem at home.
</Para>
<Para>
The normal behaviour of router queues on the Internet is called tail-drop.
Tail-drop works by queueing up to a certain amount, then dropping all traffic
that 'spills over'. This is very unfair, and also leads to retransmit
synchronization. When retransmit synchronization occurs, the sudden burst
of drops from a router that has reached its fill will cause a delayed burst
of retransmits, which will over fill the congested router again.
</Para>
<Para>
In order to cope with transient congestion on links, backbone routers will
often implement large queues. Unfortunately, while these queues are good for
throughput, they can substantially increase latency and cause TCP
connections to behave very burstily during congestion.
</Para>
<Para>
These issues with tail-drop are becoming increasingly troublesome on the
Internet because the use of network unfriendly applications is increasing.
The Linux kernel offers us RED, short for Random Early Detect, also called
Random Early Drop, as that is how it works.
</Para>
<Para>
RED isn't a cure-all for this, applications which inappropriately fail to
implement exponential backoff still get an unfair share of the bandwidth,
however, with RED they do not cause as much harm to the throughput and
latency of other connections.
</Para>
<Para>
RED statistically drops packets from flows before it reaches its hard
limit. This causes a congested backbone link to slow more gracefully, and
prevents retransmit synchronization. This also helps TCP find its 'fair'
speed faster by allowing some packets to get dropped sooner keeping queue
sizes low and latency under control. The probability of a packet being
dropped from a particular connection is proportional to its bandwidth usage
rather than the number of packets it transmits.
</Para>
<Para>
RED is a good queue for backbones, where you can't afford the
complexity of per-session state tracking needed by fairness queueing.
</Para>
<Para>
In order to use RED, you must decide on three parameters: Min, Max, and
burst. Min sets the minimum queue size in bytes before dropping will begin,
Max is a soft maximum that the algorithm will attempt to stay under, and
burst sets the maximum number of packets that can 'burst through'.
</Para>
<Para>
You should set the min by calculating that highest acceptable base queueing
latency you wish, and multiply it by your bandwidth. For instance, on my
64kbit/s ISDN link, I might want a base queueing latency of 200ms so I set
min to 1600 bytes. Setting min too small will degrade throughput and too
large will degrade latency. Setting a small min is not a replacement for
reducing the MTU on a slow link to improve interactive response.
</Para>
<Para>
You should make max at least twice min to prevent synchronization. On slow
links with small Min's it might be wise to make max perhaps four or
more times large then min.
</Para>
<Para>
Burst controls how the RED algorithm responds to bursts. Burst must be set
larger then min/avpkt. Experimentally, I've found (min+min+max)/(3*avpkt) to
work ok.
</Para>
<Para>
Additionally, you need to set limit and avpkt. Limit is a safety value, after
there are limit bytes in the queue, RED 'turns into' tail-drop. I typical set
limit to eight times max. Avpkt should be your average packet size. 1000
works OK on high speed Internet links with a 1500byte MTU.
</Para>
<Para>
Read <ULink
URL="http://www.aciri.org/floyd/papers/red/red.html"
>the paper on RED queueing</ULink
> by Sally Floyd and Van Jacobson for technical
information.
</Para>
</Sect1>
<Sect1 id="lartc.adv-qdisc.gred">
<Title>Generic Random Early Detection</Title>
<Para>
Not a lot is known about GRED. It looks like GRED with several internal
queues, whereby the internal queue is chosen based on the Diffserv tcindex
field. According to a slide found
<ULink URL="http://www.davin.ottawa.on.ca/ols/img22.htm">here</ULink>,
it contains the capabilities of Cisco's 'Distributed Weighted RED', as well
as Dave Clark's RIO.
</Para>
<Para>
Each virtual queue can have its own Drop Parameters specified.
</Para>
<Para>
FIXME: get Jamal or Werner to tell us more
</Para>
</Sect1>
<Sect1 id="lartc.adv-qdisc.vc-atm">
<Title>VC/ATM emulation</Title>
<Para>
This is quite a major effort by Werner Almesberger to allow you to build
Virtual Circuits over TCP/IP sockets. A Virtual Circuit is a concept from
ATM network theory.
</Para>
<Para>
For more information, see the <ULink
URL="http://linux-atm.sourceforge.net/"
>ATM on Linux homepage</ULink
>.
</Para>
</Sect1>
<Sect1 id="lartc.adv-qdisc.wrr">
<Title>Weighted Round Robin (WRR)</Title>
<Para>
This qdisc is not included in the standard kernels but can be downloaded from
<ULink
URL="http://wipl-wrr.dkik.dk/wrr/"
/>.
Currently the qdisc is only tested with Linux 2.2 kernels but it will
probably work with 2.4/2.5 kernels too.
</Para>
<Para>
The WRR qdisc distributes bandwidth between its classes using the weighted
round robin scheme. That is, like the CBQ qdisc it contains classes
into which arbitrary qdiscs can be plugged. All classes which have sufficient
demand will get bandwidth proportional to the weights associated with the classes.
The weights can be set manually using the <Literal remap="tt">tc</Literal> program. But they
can also be made automatically decreasing for classes transferring much data.
</Para>
<Para>
The qdisc has a built-in classifier which assigns packets coming from or
sent to different machines to different classes. Either the MAC or IP and
either source or destination addresses can be used. The MAC address can only
be used when the Linux box is acting as an ethernet bridge, however. The
classes are automatically assigned to machines based on the packets seen.
</Para>
<Para>
The qdisc can be very useful at sites such as dorms where a lot of unrelated
individuals share an Internet connection. A set of scripts setting up a
relevant behavior for such a site is a central part of the WRR distribution.
</Para>
</Sect1>
</chapter>
<chapter id="lartc.cookbook"
xreflabel="Cookbook">
<Title>Cookbook</Title>
<Para>
This section contains 'cookbook' entries which may help you solve problems.
A cookbook is no replacement for understanding however, so try and comprehend
what is going on.
</Para>
<Sect1 id="lartc.cookbook.sla">
<Title>Running multiple sites with different SLAs</Title>
<Para>
You can do this in several ways. Apache has some support for this with a
module, but we'll show how Linux can do this for you, and do so for other
services as well. These commands are stolen from a presentation by Jamal
Hadi that's referenced below.
</Para>
<Para>
Let's say we have two customers, with http, ftp and streaming audio, and we
want to sell them a limited amount of bandwidth. We do so on the server itself.
</Para>
<Para>
Customer A should have at most 2 megabits, customer B has paid for 5
megabits. We separate our customers by creating virtual IP addresses on our
server.
</Para>
<Para>
<Screen>
# ip address add 188.177.166.1 dev eth0
# ip address add 188.177.166.2 dev eth0
</Screen>
</Para>
<Para>
It is up to you to attach the different servers to the right IP address. All
popular daemons have support for this.
</Para>
<Para>
We first attach a CBQ qdisc to eth0:
<Screen>
# tc qdisc add dev eth0 root handle 1: cbq bandwidth 10Mbit cell 8 avpkt 1000 \
mpu 64
</Screen>
</Para>
<Para>
We then create classes for our customers:
</Para>
<Para>
<Screen>
# tc class add dev eth0 parent 1:0 classid 1:1 cbq bandwidth 10Mbit rate \
2MBit avpkt 1000 prio 5 bounded isolated allot 1514 weight 1 maxburst 21
# tc class add dev eth0 parent 1:0 classid 1:2 cbq bandwidth 10Mbit rate \
5Mbit avpkt 1000 prio 5 bounded isolated allot 1514 weight 1 maxburst 21
</Screen>
</Para>
<Para>
Then we add filters for our two classes:
<Screen>
##FIXME: Why this line, what does it do?, what is a divisor?:
##FIXME: A divisor has something to do with a hash table, and the number of
## buckets - ahu
# tc filter add dev eth0 parent 1:0 protocol ip prio 5 handle 1: u32 divisor 1
# tc filter add dev eth0 parent 1:0 prio 5 u32 match ip src 188.177.166.1
flowid 1:1
# tc filter add dev eth0 parent 1:0 prio 5 u32 match ip src 188.177.166.2
flowid 1:2
</Screen>
</Para>
<Para>
And we're done.
</Para>
<Para>
FIXME: why no token bucket filter? is there a default pfifo_fast fallback
somewhere?
</Para>
</Sect1>
<Sect1 id="lartc.cookbook.synflood-protect"
xreflabel="Protecting your host from SYN floods">
<Title>Protecting your host from SYN floods</Title>
<Para>
>From Alexey's iproute documentation, adapted to netfilter and with more
plausible paths. If you use this, take care to adjust the numbers to
reasonable values for your system.
</Para>
<Para>
If you want to protect an entire network, skip this script, which is best
suited for a single host.
</Para>
<Para>
It appears that you need the very latest version of the iproute2 tools to
get this to work with 2.4.0.
</Para>
<Para>
<Screen>
#! /bin/sh -x
#
# sample script on using the ingress capabilities
# this script shows how one can rate limit incoming SYNs
# Useful for TCP-SYN attack protection. You can use
# IPchains to have more powerful additions to the SYN (eg
# in addition the subnet)
#
#path to various utilities;
#change to reflect yours.
#
TC=/sbin/tc
IP=/sbin/ip
IPTABLES=/sbin/iptables
INDEV=eth2
#
# tag all incoming SYN packets through $INDEV as mark value 1
############################################################
$iptables -A PREROUTING -i $INDEV -t mangle -p tcp --syn \
-j MARK --set-mark 1
############################################################
#
# install the ingress qdisc on the ingress interface
############################################################
$TC qdisc add dev $INDEV handle ffff: ingress
############################################################
#
#
# SYN packets are 40 bytes (320 bits) so three SYNs equals
# 960 bits (approximately 1kbit); so we rate limit below
# the incoming SYNs to 3/sec (not very useful really; but
#serves to show the point - JHS
############################################################
$TC filter add dev $INDEV parent ffff: protocol ip prio 50 handle 1 fw \
police rate 1kbit burst 40 mtu 9k drop flowid :1
############################################################
#
echo "---- qdisc parameters Ingress ----------"
$TC qdisc ls dev $INDEV
echo "---- Class parameters Ingress ----------"
$TC class ls dev $INDEV
echo "---- filter parameters Ingress ----------"
$TC filter ls dev $INDEV parent ffff:
#deleting the ingress qdisc
#$TC qdisc del $INDEV ingress
</Screen>
</Para>
</Sect1>
<Sect1 id="lartc.cookbook.icmp-ratelimit">
<Title>Rate limit ICMP to prevent dDoS</Title>
<Para>
Recently, distributed denial of service attacks have become a major nuisance
on the Internet. By properly filtering and rate limiting your network, you can
both prevent becoming a casualty or the cause of these attacks.
</Para>
<Para>
You should filter your networks so that you do not allow non-local IP source
addressed packets to leave your network. This stops people from anonymously
sending junk to the Internet.
</Para>
<Para>
Rate limiting goes much as shown earlier. To refresh your memory, our
ASCIIgram again:
</Para>
<Para>
<Screen>
[The Internet] ---&#60;E3, T3, whatever&#62;--- [Linux router] --- [Office+ISP]
eth1 eth0
</Screen>
</Para>
<Para>
We first set up the prerequisite parts:
</Para>
<Para>
<Screen>
# tc qdisc add dev eth0 root handle 10: cbq bandwidth 10Mbit avpkt 1000
# tc class add dev eth0 parent 10:0 classid 10:1 cbq bandwidth 10Mbit rate \
10Mbit allot 1514 prio 5 maxburst 20 avpkt 1000
</Screen>
</Para>
<Para>
If you have 100Mbit, or more, interfaces, adjust these numbers. Now you need
to determine how much ICMP traffic you want to allow. You can perform
measurements with tcpdump, by having it write to a file for a while, and
seeing how much ICMP passes your network. Do not forget to raise the
snapshot length!
</Para>
<Para>
If measurement is impractical, you might want to choose 5&percnt; of your available
bandwidth. Let's set up our class:
<Screen>
# tc class add dev eth0 parent 10:1 classid 10:100 cbq bandwidth 10Mbit rate \
100Kbit allot 1514 weight 800Kbit prio 5 maxburst 20 avpkt 250 \
bounded
</Screen>
</Para>
<Para>
This limits at 100Kbit. Now we need a filter to assign ICMP traffic to this
class:
<Screen>
# tc filter add dev eth0 parent 10:0 protocol ip prio 100 u32 match ip
protocol 1 0xFF flowid 10:100
</Screen>
</Para>
</Sect1>
<Sect1 id="lartc.cookbook.interactive-prio">
<Title>Prioritizing interactive traffic</Title>
<Para>
If lots of data is coming down your link, or going up for that matter, and
you are trying to do some maintenance via telnet or ssh, this may not go too
well. Other packets are blocking your keystrokes. Wouldn't it be great if
there were a way for your interactive packets to sneak past the bulk
traffic? Linux can do this for you!
</Para>
<Para>
As before, we need to handle traffic going both ways. Evidently, this works
best if there are Linux boxes on both ends of your link, although other
UNIX's are able to do this. Consult your local Solaris/BSD guru for this.
</Para>
<Para>
The standard pfifo_fast scheduler has 3 different 'bands'. Traffic in band 0
is transmitted first, after which traffic in band 1 and 2 gets considered.
It is vital that our interactive traffic be in band 0!
</Para>
<Para>
We blatantly adapt from the (soon to be obsolete) ipchains HOWTO:
</Para>
<Para>
There are four seldom-used bits in the IP header, called the Type of Service
(TOS) bits. They effect the way packets are treated; the four bits are
"Minimum Delay", "Maximum Throughput", "Maximum Reliability" and "Minimum
Cost". Only one of these bits is allowed to be set. Rob van Nieuwkerk, the
author of the ipchains TOS-mangling code, puts it as follows:
</Para>
<Para>
<Screen>
Especially the "Minimum Delay" is important for me. I switch it on for
"interactive" packets in my upstream (Linux) router. I'm
behind a 33k6 modem link. Linux prioritizes packets in 3 queues. This
way I get acceptable interactive performance while doing bulk
downloads at the same time.
</Screen>
</Para>
<Para>
The most common use is to set telnet &#38; ftp control connections to "Minimum
Delay" and FTP data to "Maximum Throughput". This would be
done as follows, on your upstream router:
</Para>
<Para>
<Screen>
# iptables -A PREROUTING -t mangle -p tcp --sport telnet \
-j TOS --set-tos Minimize-Delay
# iptables -A PREROUTING -t mangle -p tcp --sport ftp \
-j TOS --set-tos Minimize-Delay
# iptables -A PREROUTING -t mangle -p tcp --sport ftp-data \
-j TOS --set-tos Maximize-Throughput
</Screen>
</Para>
<Para>
Now, this only works for data going from your telnet foreign host to your
local computer. The other way around appears to be done for you, ie, telnet,
ssh &#38; friends all set the TOS field on outgoing packets automatically.
</Para>
<Para>
Should you have an application that does not do this, you can always do it
with netfilter. On your local box:
</Para>
<Para>
<Screen>
# iptables -A OUTPUT -t mangle -p tcp --dport telnet \
-j TOS --set-tos Minimize-Delay
# iptables -A OUTPUT -t mangle -p tcp --dport ftp \
-j TOS --set-tos Minimize-Delay
# iptables -A OUTPUT -t mangle -p tcp --dport ftp-data \
-j TOS --set-tos Maximize-Throughput
</Screen>
</Para>
</Sect1>
<Sect1 id="lartc.cookbook.squid">
<Title>Transparent web-caching using <application>netfilter</application>,
<application>iproute2</application>, <application>ipchains</application> and
<application>squid</application></Title>
<Para>
This section was sent in by reader Ram Narula from Internet for Education
(Thailand).
</Para>
<Para>
The regular technique in accomplishing this in Linux
is probably with use of ipchains AFTER making sure
that the "outgoing" port 80(web) traffic gets routed through
the server running squid.
</Para>
<Para>
There are 3 common methods to make sure "outgoing"
port 80 traffic gets routed to the server running squid
and 4th one is being introduced here.
</Para>
<Para>
<VariableList>
<VarListEntry>
<Term>Making the gateway router do it.</Term>
<ListItem>
<Para>
If you can tell your gateway router to
match packets that has outgoing destination port
of 80 to be sent to the IP address of squid server.
</Para>
<Para>
BUT
</Para>
<Para>
This would put additional load on the router and
some commercial routers might not even support this.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Using a Layer 4 switch.</Term>
<ListItem>
<Para>
Layer 4 switches can handle this without any problem.
</Para>
<Para>
BUT
</Para>
<Para>
The cost for this equipment is usually very high. Typical
layer 4 switch would normally cost more than
a typical router+good linux server.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Using cache server as network's gateway.</Term>
<ListItem>
<Para>
You can force ALL traffic through cache server.
</Para>
<Para>
BUT
</Para>
<Para>
This is quite risky because Squid does
utilize lots of CPU power which might
result in slower over-all network performance
or the server itself might crash and no one on the
network will be able to access the Internet if
that occurs.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Linux+NetFilter router.</Term>
<ListItem>
<Para>
By using NetFilter another technique can be implemented
which is using NetFilter for "mark"ing the packets
with destination port 80 and using iproute2 to
route the "mark"ed packets to the Squid server.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
<Screen>
|----------------|
| Implementation |
|----------------|
Addresses used
10.0.0.1 naret (NetFilter server)
10.0.0.2 silom (Squid server)
10.0.0.3 donmuang (Router connected to the Internet)
10.0.0.4 kaosarn (other server on network)
10.0.0.5 RAS
10.0.0.0/24 main network
10.0.0.0/19 total network
|---------------|
|Network diagram|
|---------------|
Internet
|
donmuang
|
------------hub/switch----------
| | | |
naret silom kaosarn RAS etc.
</Screen>
First, make all traffic pass through naret by making
sure it is the default gateway except for silom.
Silom's default gateway has to be donmuang (10.0.0.3) or
this would create web traffic loop.
</Para>
<Para>
(all servers on my network had 10.0.0.1 as the default gateway
which was the former IP address of donmuang router so what I did
was changed the IP address of donmuang to 10.0.0.3 and gave
naret ip address of 10.0.0.1)
</Para>
<Para>
<Screen>
Silom
-----
-setup squid and ipchains
</Screen>
</Para>
<Para>
Setup Squid server on silom, make sure it does support
transparent caching/proxying, the default port is usually
3128, so all traffic for port 80 has to be redirected to port
3128 locally. This can be done by using ipchains with the following:
</Para>
<Para>
<Screen>
silom# ipchains -N allow1
silom# ipchains -A allow1 -p TCP -s 10.0.0.0/19 -d 0/0 80 -j REDIRECT 3128
silom# ipchains -I input -j allow1
</Screen>
</Para>
<Para>
</Para>
<Para>
Or, in netfilter lingo:
<Screen>
silom# iptables -t nat -A PREROUTING -i eth0 -p tcp --dport 80 -j REDIRECT --to-port 3128
</Screen>
</Para>
<Para>
(note: you might have other entries as well)
</Para>
<Para>
For more information on setting Squid server please refer
to Squid FAQ page on <ULink
URL="http://squid.nlanr.net"
>http://squid.nlanr.net</ULink
>).
</Para>
<Para>
Make sure ip forwarding is enabled on this server and the default
gateway for this server is donmuang router (NOT naret).
</Para>
<Para>
<Screen>
Naret
-----
-setup iptables and iproute2
-disable icmp REDIRECT messages (if needed)
</Screen>
</Para>
<Para>
<OrderedList>
<ListItem>
<Para>
"Mark" packets of destination port 80 with value 2
<Screen>
naret# iptables -A PREROUTING -i eth0 -t mangle -p tcp --dport 80 \
-j MARK --set-mark 2
</Screen>
</Para>
</ListItem>
<ListItem>
<Para>
Setup iproute2 so it will route packets with "mark" 2 to silom
<Screen>
naret# echo 202 www.out &#62;&#62; /etc/iproute2/rt_tables
naret# ip rule add fwmark 2 table www.out
naret# ip route add default via 10.0.0.2 dev eth0 table www.out
naret# ip route flush cache
</Screen>
</Para>
<Para>
If donmuang and naret is on the same subnet then
naret should not send out icmp REDIRECT messages.
In this case it is, so icmp REDIRECTs has to be
disabled by:
<Screen>
naret# echo 0 &#62; /proc/sys/net/ipv4/conf/all/send_redirects
naret# echo 0 &#62; /proc/sys/net/ipv4/conf/default/send_redirects
naret# echo 0 &#62; /proc/sys/net/ipv4/conf/eth0/send_redirects
</Screen>
</Para>
</ListItem>
</OrderedList>
</Para>
<Para>
The setup is complete, check the configuration
</Para>
<Para>
<Screen>
On naret:
naret# iptables -t mangle -L
Chain PREROUTING (policy ACCEPT)
target prot opt source destination
MARK tcp -- anywhere anywhere tcp dpt:www MARK set 0x2
Chain OUTPUT (policy ACCEPT)
target prot opt source destination
naret# ip rule ls
0: from all lookup local
32765: from all fwmark 2 lookup www.out
32766: from all lookup main
32767: from all lookup default
naret# ip route list table www.out
default via 203.114.224.8 dev eth0
naret# ip route
10.0.0.1 dev eth0 scope link
10.0.0.0/24 dev eth0 proto kernel scope link src 10.0.0.1
127.0.0.0/8 dev lo scope link
default via 10.0.0.3 dev eth0
(make sure silom belongs to one of the above lines, in this case
it's the line with 10.0.0.0/24)
|------|
|-DONE-|
|------|
</Screen>
</Para>
<Sect2>
<Title>Traffic flow diagram after implementation</Title>
<Screen width="80">
|-----------------------------------------|
|Traffic flow diagram after implementation|
|-----------------------------------------|
INTERNET
/\
||
\/
-----------------donmuang router---------------------
/\ /\ ||
|| || ||
|| \/ ||
naret silom ||
*destination port 80 traffic=========&#62;(cache) ||
/\ || ||
|| \/ \/
\\===================================kaosarn, RAS, etc.
</Screen>
<Para>
Note that the network is asymmetric as there is one extra hop on
general outgoing path.
</Para>
<Para>
<Screen>
Here is run down for packet traversing the network from kaosarn
to and from the Internet.
For web/http traffic:
kaosarn http request-&#62;naret-&#62;silom-&#62;donmuang-&#62;internet
http replies from Internet-&#62;donmuang-&#62;silom-&#62;kaosarn
For non-web/http requests(eg. telnet):
kaosarn outgoing data-&#62;naret-&#62;donmuang-&#62;internet
incoming data from Internet-&#62;donmuang-&#62;kaosarn
</Screen>
</Para>
</Sect2>
</Sect1>
<Sect1 id="lartc.cookbook.mtu-discovery">
<Title>Circumventing Path MTU Discovery issues with per route MTU settings</Title>
<Para>
For sending bulk data, the Internet generally works better when using larger
packets. Each packet implies a routing decision, when sending a 1 megabyte
file, this can either mean around 700 packets when using packets that are as
large as possible, or 4000 if using the smallest default.
</Para>
<Para>
However, not all parts of the Internet support full 1460 bytes of payload
per packet. It is therefore necessary to try and find the largest packet
that will 'fit', in order to optimize a connection.
</Para>
<Para>
This process is called 'Path MTU Discovery', where MTU stands for 'Maximum
Transfer Unit.'
</Para>
<Para>
When a router encounters a packet that's too big too send in one piece, AND
it has been flagged with the "Don't Fragment" bit, it returns an ICMP
message stating that it was forced to drop a packet because of this. The
sending host acts on this hint by sending smaller packets, and by iterating
it can find the optimum packet size for a connection over a certain path.
</Para>
<Para>
This used to work well until the Internet was discovered by hooligans who do
their best to disrupt communications. This in turn lead administrators to
either block or shape ICMP traffic in a misguided attempt to improve
security or robustness of their Internet service.
</Para>
<Para>
What has happened now is that Path MTU Discovery is working less and less
well and fails for certain routes, which leads to strange TCP/IP sessions
which die after a while.
</Para>
<Para>
Although I have no proof for this, two sites who I used to have this problem
with both run Alteon Acedirectors before the affected systems - perhaps
somebody more knowledgeable can provide clues as to why this happens.
</Para>
<Sect2>
<Title>Solution</Title>
<Para>
When you encounter sites that suffer from this problem, you can disable Path
MTU discovery by setting it manually. Koos van den Hout, slightly edited,
writes:
</Para>
<blockquote>
<Para>
The following problem: I set the mtu/mru of my leased line running ppp to
296 because it's only 33k6 and I cannot influence the queueing on the
other side. At 296, the response to a key press is within a reasonable
time frame.
</Para>
<Para>
And, on my side I have a masqrouter running (of course) Linux.
</Para>
<Para>
Recently I split 'server' and 'router' so most applications are run on a
different machine than the routing happens on.
</Para>
<Para>
I then had trouble logging into irc. Big panic! Some digging did find
out that I got connected to irc, even showed up as 'connected' on irc
but I did not receive the motd from irc. I checked what could be wrong
and noted that I already had some previous trouble reaching certain
websites related to the MTU, since I had no trouble reaching them when
the MTU was 1500, the problem just showed when the MTU was set to 296.
Since irc servers block about every kind of traffic not needed for their
immediate operation, they also block icmp.
</Para>
<Para>
I managed to convince the operators of a webserver that this was the cause
of a problem, but the irc server operators were not going to fix this.
</Para>
<Para>
So, I had to make sure outgoing masqueraded traffic started with the lower
mtu of the outside link. But I want local ethernet traffic to have the
normal mtu (for things like nfs traffic).
</Para>
<Para>Solution:
</Para>
<Screen width="80">
ip route add default via 10.0.0.1 mtu 296
</Screen>
<Para>
(10.0.0.1 being the default gateway, the inside address of the
masquerading router)
</Para>
</blockquote>
<Para>
In general, it is possible to override PMTU Discovery by setting specific
routes. For example, if only a certain subnet is giving problems, this
should help:
</Para>
<Screen width="80">
ip route add 195.96.96.0/24 via 10.0.0.1 mtu 1000
</Screen>
</Sect2>
</Sect1>
<Sect1 id="lartc.cookbook.mtu-mss">
<Title>Circumventing Path MTU Discovery issues with MSS Clamping
(for ADSL, cable, PPPoE &amp; PPtP users)</Title>
<Para>
As explained above, Path MTU Discovery doesn't work as well as it should
anymore. If you know for a fact that a hop somewhere in your network has a
limited (&lt;1500) MTU, you cannot rely on PMTU Discovery finding this out.
</Para>
<Para>
Besides MTU, there is yet another way to set the maximum packet size, the so
called Maximum Segment Size. This is a field in the TCP Options part of a
SYN packet.
</Para>
<Para>
Recent Linux kernels, and a few PPPoE drivers (notably, the excellent
Roaring Penguin one), feature the possibility to 'clamp the MSS'.
</Para>
<Para>
The good thing about this is that by setting the MSS value, you are telling
the remote side unequivocally 'do not ever try to send me packets bigger
than this value'. No ICMP traffic is needed to get this to work.
</Para>
<Para>
The bad thing is that it's an obvious hack - it breaks 'end to end' by
modifying packets. Having said that, we use this trick in many places and it
works like a charm.
</Para>
<Para>
In order for this to work you need at least iptables-1.2.1a and Linux 2.4.3
or higher. The basic command line is:
<Screen>
# iptables -A FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu
</Screen>
</Para>
<Para>
This calculates the proper MSS for your link. If you are feeling brave, or
think that you know best, you can also do something like this:
</Para>
<Para>
<Screen>
# iptables -A FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --set-mss 128
</Screen>
</Para>
<Para>
This sets the MSS of passing SYN packets to 128. Use this if you have VoIP
with tiny packets, and huge http packets which are causing chopping in your
voice calls.
</Para>
</Sect1>
<Sect1 id="lartc.cookbook.ultimate-tc">
<Title>The Ultimate Traffic Conditioner: Low Latency, Fast Up &amp; Downloads</Title>
<Para>
Note: This script has recently been upgraded and previously only worked for
Linux clients in your network! So you might want to update if you have
Windows machines or Macs in your network and noticed that they were not able
to download faster while others were uploading.
</Para>
<Para>
I attempted to create the holy grail:
<VariableList>
<VarListEntry>
<Term>Maintain low latency for interactive traffic at all times</Term>
<ListItem>
<Para>
This means that downloading or uploading files should not disturb SSH or
even telnet. These are the most important things, even 200ms latency is
sluggish to work over.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Allow 'surfing' at reasonable speeds while up or downloading</Term>
<ListItem>
<Para>
Even though http is 'bulk' traffic, other traffic should not drown it out
too much.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Make sure uploads don't harm downloads, and the other way around</Term>
<ListItem>
<Para>
This is a much observed phenomenon where upstream traffic simply destroys
download speed.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
It turns out that all this is possible, at the cost of a tiny bit of
bandwidth. The reason that uploads, downloads and ssh hurt each other is the
presence of large queues in many domestic access devices like cable or DSL
modems.
</Para>
<Para>
The next section explains in depth what causes the delays, and how we can
fix them. You can safely skip it and head straight for the script if you
don't care how the magic is performed.
</Para>
<Sect2>
<Title>Why it doesn't work well by default</Title>
<Para>
ISPs know that they are benchmarked solely on how fast people can download.
Besides available bandwidth, download speed is influenced heavily by packet
loss, which seriously hampers TCP/IP performance. Large queues can help
prevent packet loss, and speed up downloads. So ISPs configure large queues.
</Para>
<Para>
These large queues however damage interactivity. A keystroke must first
travel the upstream queue, which may be seconds (!) long and go to your
remote host. It is then displayed, which leads to a packet coming back, which
must then traverse the downstream queue, located at your ISP, before it
appears on your screen.
</Para>
<Para>
This HOWTO teaches you how to mangle and process the queue in many ways, but
sadly, not all queues are accessible to us. The queue over at the ISP is
completely off-limits, whereas the upstream queue probably lives inside your
cable modem or DSL device. You may or may not be able to configure it. Most
probably not.
</Para>
<Para>
So, what next? As we can't control either of those queues, they must be
eliminated, and moved to your Linux router. Luckily this is possible.
</Para>
<Para>
<VariableList>
<VarListEntry>
<Term>Limit upload speed</Term>
<ListItem>
<Para>
By limiting our upload speed to slightly less than the truly available rate,
no queues are built up in our modem. The queue is now moved to Linux.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Limit download speed</Term>
<ListItem>
<Para>
This is slightly trickier as we can't really influence how fast the internet
ships us data. We can however drop packets that are coming in too fast,
which causes TCP/IP to slow down to just the rate we want. Because we don't
want to drop traffic unnecessarily, we configure a 'burst' size we allow at
higher speed.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
<Para>
Now, once we have done this, we have eliminated the downstream queue totally
(except for short bursts), and gain the ability to manage the upstream queue
with all the power Linux offers.
</Para>
<Para>
What remains to be done is to make sure interactive traffic jumps to the
front of the upstream queue. To make sure that uploads don't hurt downloads,
we also move ACK packets to the front of the queue. This is what normally
causes the huge slowdown observed when generating bulk traffic both ways.
The ACKnowledgements for downstream traffic must compete with upstream
traffic, and get delayed in the process.
</Para>
<Para>
If we do all this we get the following measurements using an excellent ADSL
connection from xs4all in the Netherlands:
</Para>
<Para>
<Screen>
Baseline latency:
round-trip min/avg/max = 14.4/17.1/21.7 ms
Without traffic conditioner, while downloading:
round-trip min/avg/max = 560.9/573.6/586.4 ms
Without traffic conditioner, while uploading:
round-trip min/avg/max = 2041.4/2332.1/2427.6 ms
With conditioner, during 220kbit/s upload:
round-trip min/avg/max = 15.7/51.8/79.9 ms
With conditioner, during 850kbit/s download:
round-trip min/avg/max = 20.4/46.9/74.0 ms
When uploading, downloads proceed at ~80% of the available speed. Uploads
at around 90%. Latency then jumps to 850 ms, still figuring out why.
</Screen>
</Para>
<Para>
What you can expect from this script depends a lot on your actual uplink
speed. When uploading at full speed, there will always be a single packet
ahead of your keystroke. That is the lower limit to the latency you can
achieve - divide your MTU by your upstream speed to calculate. Typical
values will be somewhat higher than that. Lower your MTU for better effects!
</Para>
<Para>
Next, two versions of this script, one with Devik's excellent HTB, the other
with CBQ which is in each Linux kernel, unlike HTB. Both are tested and work
well.
</Para>
</Sect2>
<Sect2>
<Title>The actual script (CBQ)</Title>
<Para>
Works on all kernels. Within the CBQ
qdisc we place two Stochastic Fairness Queues that make sure that multiple
bulk streams don't drown each other out.
</Para>
<Para>
Downstream traffic is policed using a tc filter containing a Token Bucket
Filter.
</Para>
<Para>
You might improve on this script by adding 'bounded' to the line that starts
with 'tc class add .. classid 1:20'. If you lowered your MTU, also lower the
allot &amp; avpkt numbers!
</Para>
<Para>
<Screen>
#!/bin/bash
# The Ultimate Setup For Your Internet Connection At Home
#
#
# Set the following values to somewhat less than your actual download
# and uplink speed. In kilobits
DOWNLINK=800
UPLINK=220
DEV=ppp0
# clean existing down- and uplink qdiscs, hide errors
tc qdisc del dev $DEV root 2&#62; /dev/null &#62; /dev/null
tc qdisc del dev $DEV ingress 2&#62; /dev/null &#62; /dev/null
###### uplink
# install root CBQ
tc qdisc add dev $DEV root handle 1: cbq avpkt 1000 bandwidth 10mbit
# shape everything at $UPLINK speed - this prevents huge queues in your
# DSL modem which destroy latency:
# main class
tc class add dev $DEV parent 1: classid 1:1 cbq rate ${UPLINK}kbit \
allot 1500 prio 5 bounded isolated
# high prio class 1:10:
tc class add dev $DEV parent 1:1 classid 1:10 cbq rate ${UPLINK}kbit \
allot 1600 prio 1 avpkt 1000
# bulk and default class 1:20 - gets slightly less traffic,
# and a lower priority:
tc class add dev $DEV parent 1:1 classid 1:20 cbq rate $[9*$UPLINK/10]kbit \
allot 1600 prio 2 avpkt 1000
# both get Stochastic Fairness:
tc qdisc add dev $DEV parent 1:10 handle 10: sfq perturb 10
tc qdisc add dev $DEV parent 1:20 handle 20: sfq perturb 10
# start filters
# TOS Minimum Delay (ssh, NOT scp) in 1:10:
tc filter add dev $DEV parent 1:0 protocol ip prio 10 u32 \
match ip tos 0x10 0xff flowid 1:10
# ICMP (ip protocol 1) in the interactive class 1:10 so we
# can do measurements &#38; impress our friends:
tc filter add dev $DEV parent 1:0 protocol ip prio 11 u32 \
match ip protocol 1 0xff flowid 1:10
# To speed up downloads while an upload is going on, put ACK packets in
# the interactive class:
tc filter add dev $DEV parent 1: protocol ip prio 12 u32 \
match ip protocol 6 0xff \
match u8 0x05 0x0f at 0 \
match u16 0x0000 0xffc0 at 2 \
match u8 0x10 0xff at 33 \
flowid 1:10
# rest is 'non-interactive' ie 'bulk' and ends up in 1:20
tc filter add dev $DEV parent 1: protocol ip prio 13 u32 \
match ip dst 0.0.0.0/0 flowid 1:20
########## downlink #############
# slow downloads down to somewhat less than the real speed to prevent
# queuing at our ISP. Tune to see how high you can set it.
# ISPs tend to have *huge* queues to make sure big downloads are fast
#
# attach ingress policer:
tc qdisc add dev $DEV handle ffff: ingress
# filter *everything* to it (0.0.0.0/0), drop everything that's
# coming in too fast:
tc filter add dev $DEV parent ffff: protocol ip prio 50 u32 match ip src \
0.0.0.0/0 police rate ${DOWNLINK}kbit burst 10k drop flowid :1
</Screen>
If you want this script to be run by ppp on connect, copy it to
/etc/ppp/ip-up.d.
</Para>
<Para>
If the last two lines give an error, update your tc tool to a newer version!
</Para>
</Sect2>
<Sect2>
<Title>The actual script (HTB)</Title>
<Para>
The following script achieves all goals using the wonderful HTB queue, see
the relevant chapter. Well worth patching your kernel for!
<Screen>
#!/bin/bash
# The Ultimate Setup For Your Internet Connection At Home
#
#
# Set the following values to somewhat less than your actual download
# and uplink speed. In kilobits
DOWNLINK=800
UPLINK=220
DEV=ppp0
# clean existing down- and uplink qdiscs, hide errors
tc qdisc del dev $DEV root 2&#62; /dev/null &#62; /dev/null
tc qdisc del dev $DEV ingress 2&#62; /dev/null &#62; /dev/null
###### uplink
# install root HTB, point default traffic to 1:20:
tc qdisc add dev $DEV root handle 1: htb default 20
# shape everything at $UPLINK speed - this prevents huge queues in your
# DSL modem which destroy latency:
tc class add dev $DEV parent 1: classid 1:1 htb rate ${UPLINK}kbit burst 6k
# high prio class 1:10:
tc class add dev $DEV parent 1:1 classid 1:10 htb rate ${UPLINK}kbit \
burst 6k prio 1
# bulk &#38; default class 1:20 - gets slightly less traffic,
# and a lower priority:
tc class add dev $DEV parent 1:1 classid 1:20 htb rate $[9*$UPLINK/10]kbit \
burst 6k prio 2
# both get Stochastic Fairness:
tc qdisc add dev $DEV parent 1:10 handle 10: sfq perturb 10
tc qdisc add dev $DEV parent 1:20 handle 20: sfq perturb 10
# TOS Minimum Delay (ssh, NOT scp) in 1:10:
tc filter add dev $DEV parent 1:0 protocol ip prio 10 u32 \
match ip tos 0x10 0xff flowid 1:10
# ICMP (ip protocol 1) in the interactive class 1:10 so we
# can do measurements &#38; impress our friends:
tc filter add dev $DEV parent 1:0 protocol ip prio 10 u32 \
match ip protocol 1 0xff flowid 1:10
# To speed up downloads while an upload is going on, put ACK packets in
# the interactive class:
tc filter add dev $DEV parent 1: protocol ip prio 10 u32 \
match ip protocol 6 0xff \
match u8 0x05 0x0f at 0 \
match u16 0x0000 0xffc0 at 2 \
match u8 0x10 0xff at 33 \
flowid 1:10
# rest is 'non-interactive' ie 'bulk' and ends up in 1:20
########## downlink #############
# slow downloads down to somewhat less than the real speed to prevent
# queuing at our ISP. Tune to see how high you can set it.
# ISPs tend to have *huge* queues to make sure big downloads are fast
#
# attach ingress policer:
tc qdisc add dev $DEV handle ffff: ingress
# filter *everything* to it (0.0.0.0/0), drop everything that's
# coming in too fast:
tc filter add dev $DEV parent ffff: protocol ip prio 50 u32 match ip src \
0.0.0.0/0 police rate ${DOWNLINK}kbit burst 10k drop flowid :1
</Screen>
</Para>
<Para>
If you want this script to be run by ppp on connect, copy it to
/etc/ppp/ip-up.d.
</Para>
<Para>
If the last two lines give an error, update your tc tool to a newer version!
</Para>
</Sect2>
</Sect1>
<sect1 id="lartc.ratelimit.single"><title>Rate limiting a single host or netmask</title>
<para>
Although this is described in stupendous details elsewhere and in our manpages, this question gets asked a lot and
happily there is a simple answer that does not need full comprehension of traffic control.
</para>
<para>
This three line script does the trick:
</para>
<para>
<screen>
tc qdisc add dev $DEV root handle 1: cbq avpkt 1000 bandwidth 10mbit
tc class add dev $DEV parent 1: classid 1:1 cbq rate 512kbit \
allot 1500 prio 5 bounded isolated
tc filter add dev $DEV parent 1: protocol ip prio 16 u32 \
match ip dst 195.96.96.97 flowid 1:1
</screen>
</para>
<para>
The first line installs a class based queue on your interface, and tells the kernel that for calculations,
it can be assumed to be a 10mbit interface. If you get this wrong, no real harm is done. But getting it right will
make everything more precise.
</para>
<para>
The second line creates a 512kbit class with some reasonable defaults. For details, see the cbq manpages and
<xref linkend="lartc.qdisc">.
</para>
<para>
The last line tells which traffic should go to the shaped class. Traffic not matched by this rule is NOT shaped. To make more
complicated matches (subnets, source ports, destination ports), see <xref linkend="lartc.filtering.simple">.
</para>
<para>
If you changed anything and want to reload the script, execute 'tc qdisc del dev $DEV root' to clean up your existing
configuration.
</para>
<para>
The script can further be improved by adding a last optional line 'tc qdisc add dev $DEV parent 1:1 sfq perturb 10'. See
<xref linkend="lartc.sfq"> for details on what this does.
</para>
</sect1>
</chapter>
<chapter id="lartc.bridging">
<Title>Building bridges, and pseudo-bridges with Proxy ARP</Title>
<Para>
Bridges are devices which can be installed in a network without any
reconfiguration. A network switch is basically a many-port bridge. A bridge
is often a 2-port switch. Linux does however support multiple interfaces in
a bridge, making it a true switch.
</Para>
<Para>
Bridges are often deployed when confronted with a broken network that needs
to be fixed without any alterations. Because the bridge is a layer-2 device,
one layer below IP, routers and servers are not aware of its existence.
This means that you can transparently block or modify certain packets, or do
shaping.
</Para>
<Para>
Another good thing is that a bridge can often be replaced by a cross cable
or a hub, should it break down.
</Para>
<Para>
The bad news is that a bridge can cause great confusion unless it is very
well documented. It does not appear in traceroutes, but somehow packets
disappear or get changed from point A to point B ('this network is
HAUNTED!'). You should also wonder if an organization that 'does not want to
change anything' is doing the right thing.
</Para>
<Para>
The Linux 2.4/2.5 bridge is documented on
<ULink URL=" http://bridge.sourceforge.net/">this page</ULink>.
</Para>
<Sect1 id="lartc.bridging.iptables">
<Title>State of bridging and iptables</Title>
<Para>
As of Linux 2.4.14, bridging and iptables do not 'see' each other without
help. If you bridge packets from eth0 to eth1, they do not 'pass' by
iptables. This means that you cannot do filtering, or NAT or mangling or
whatever.
</Para>
<Para>
There are several projects going on to fix this, the truly right one is by
the author of the Linux 2.4 bridging code, Lennert Buytenhek. He recently
informed us that as of bridge-nf 0.0.2 (see the url above), the code is
stable and usable in production environments. He is now asking the kernel
people if and how the patch can be merged, stay tuned!
</Para>
</Sect1>
<Sect1 id="lartc.bridging.shaping">
<Title>Bridging and shaping</Title>
<Para>
This does work as advertised. Be sure to figure out which side each
interface is on, otherwise you might be shaping outbound traffic in your
internal interface, which won't work. Use tcpdump if needed.
</Para>
</Sect1>
<Sect1 id="lartc.bridging.proxy-arp">
<Title>Pseudo-bridges with Proxy-ARP</Title>
<Para>
If you just want to implement a Pseudo-bridge, skip down a few sections
to 'Implementing it', but it is wise to read a bit about how it works in
practice.
</Para>
<Para>
A Pseudo-bridge works a bit differently. By default, a bridge passes packets
unaltered from one interface to the other. It only looks at the hardware
address of packets to determine what goes where. This in turn means that you
can bridge traffic that Linux does not understand, as long as it has an
hardware address it does.
</Para>
<Para>
A 'Pseudo-bridge' works differently and looks more like a hidden router than
a bridge, but like a bridge, it has little impact on network design.
</Para>
<Para>
An advantage of the fact that it is not a bridge lies in the fact that
packets really pass through the kernel, and can be filtered, changed,
redirected or rerouted.
</Para>
<Para>
A real bridge can also be made to perform these feats, but it needs special
code, like the Ethernet Frame Diverter, or the above mentioned patch.
</Para>
<Para>
Another advantage of a pseudo-bridge is that it does not pass packets it
does not understand - thus cleaning your network of a lot of cruft. In cases
where you need this cruft (like SAP packets, or Netbeui), use a real bridge.
</Para>
<Sect2>
<Title>ARP &amp; Proxy-ARP</Title>
<Para>
When a host wants to talk to another host on the same physical network
segment, it sends out an Address Resolution Protocol packet, which, somewhat
simplified, reads like this 'who has 10.0.0.1, tell 10.0.0.7'. In response
to this, 10.0.0.1 replies with a short 'here' packet.
</Para>
<Para>
10.0.0.7 then sends packets to the hardware address mentioned in the 'here'
packet. It caches this hardware address for a relatively long time, and
after the cache expires, it re-asks the question.
</Para>
<Para>
When building a Pseudo-bridge, we instruct the bridge to reply to these ARP
packets, which causes the hosts in the network to send its packets to the
bridge. The bridge then processes these packets, and sends them to the
relevant interface.
</Para>
<Para>
So, in short, whenever a host on one side of the bridge asks for the
hardware address of a host on the other, the bridge replies with a packet
that says 'hand it to me'.
</Para>
<Para>
This way, all data traffic gets transmitted to the right place, and always
passes through the bridge.
</Para>
</Sect2>
<Sect2>
<Title>Implementing it</Title>
<Para>
In the bad old days, it used to be possible to instruct the Linux Kernel to
perform 'proxy-ARP' for just any subnet. So, to configure a pseudo-bridge,
you would have to specify both the proper routes to both sides of the bridge
AND create matching proxy-ARP rules. This is bad in that it requires a lot
of typing, but also because it easily allows you to make mistakes which make
your bridge respond to ARP queries for networks it does not know how to
route.
</Para>
<Para>
With Linux 2.4/2.5 (and possibly 2.2), this possibility has been withdrawn and
has been replaced by a flag in the /proc directory, called 'proxy_arp'. The
procedure for building a pseudo-bridge is then:
</Para>
<Para>
<OrderedList>
<ListItem>
<Para>
Assign an IP address to both interfaces, the 'left' and the 'right'
one
</Para>
</ListItem>
<ListItem>
<Para>
Create routes so your machine knows which hosts reside on the left,
and which on the right
</Para>
</ListItem>
<ListItem>
<Para>
Turn on proxy-ARP on both interfaces, echo 1 &#62;
/proc/sys/net/ipv4/conf/ethL/proxy_arp, echo 1 &#62;
/proc/sys/net/ipv4/conf/ethR/proxy_arp, where L and R stand for the numbers
of your interfaces on the left and on the right side
</Para>
</ListItem>
</OrderedList>
</Para>
<Para>
Also, do not forget to turn on the ip_forwarding flag! When converting from
a true bridge, you may find that this flag was turned off as it is not
needed when bridging.
</Para>
<Para>
Another thing you might note when converting is that you need to clear the
arp cache of computers in the network - the arp cache might contain old
pre-bridge hardware addresses which are no longer correct.
</Para>
<Para>
On a Cisco, this is done using the command 'clear arp-cache', under
Linux, use 'arp -d ip.address'. You can also wait for the cache to expire
manually, which can take rather long.
</Para>
<para>
You can speed this up using the wonderful 'arping' tool, which on many
distributions is part of the 'iputils' package. Using 'arping' you can send
out unsolicited ARP messages so as to update remote arp caches.
</para>
<para>
This is a very powerful technique that is also used by 'black hats' to
subvert your routing!
</para>
<note>
<para>
On Linux 2.4, you may need to execute
'echo 1 &gt; /proc/sys/net/ipv4/ip_nonlocal_bind' before being able to send
out unsolicited ARP messages!
</para>
</note>
<Para>
You may also discover that your network was misconfigured if you are/were of
the habit of specifying routes without netmasks. To explain, some versions
of route may have guessed your netmask right in the past, or guessed wrong
without you noticing. When doing surgical routing like described above, it
is *vital* that you check your netmasks!
</Para>
</Sect2>
</Sect1>
</chapter>
<chapter id="lartc.dynamic-routing">
<Title>Dynamic routing - OSPF and BGP</Title>
<Para>
Once your network starts to get really big, or you start to consider 'the
internet' as your network, you need tools which dynamically route your data.
Sites are often connected to each other with multiple links, and more are
popping up all the time.
</Para>
<Para>
The Internet has mostly standardized on OSPF and BGP4 (rfc1771).
Linux supports both, by way of <application>gated</application> and
<application>zebra</application>
</Para>
<Para>
While currently not within the scope of this document, we would like to
point you to the definitive works:
</Para>
<Para>
Overview:
</Para>
<Para>
Cisco Systems
<ULink
URL="http://www.cisco.com/univercd/cc/td/doc/cisintwk/idg4/nd2003.htm"
>Designing large-scale IP Internetworks</ULink
>
</Para>
<Para>
For OSPF:
</Para>
<Para>
Moy, John T.
"OSPF. The anatomy of an Internet routing protocol"
Addison Wesley. Reading, MA. 1998.
</Para>
<Para>
Halabi has also written a good guide to OSPF routing design, but this
appears to have been dropped from the Cisco web site.
</Para>
<Para>
For BGP:
</Para>
<Para>
Halabi, Bassam
"Internet routing architectures"
Cisco Press (New Riders Publishing). Indianapolis, IN. 1997.
</Para>
<Para>
also
</Para>
<Para>
Cisco Systems
</Para>
<Para>
<ULink
URL="http://www.cisco.com/univercd/cc/td/doc/cisintwk/ics/icsbgp4.htm"
>Using the Border Gateway Protocol for interdomain routing</ULink
>
</Para>
<Para>
Although the examples are Cisco-specific, they are remarkably similar
to the configuration language in Zebra :-)
</Para>
</chapter>
<chapter id="lartc.other"
xreflabel="Other possibilities">
<Title>Other possibilities</Title>
<Para>
This chapter is a list of projects having to do with advanced Linux routing
&amp; traffic shaping. Some of these links may deserve chapters of their
own, some are documented very well of themselves, and don't need more HOWTO.
</Para>
<Para>
<VariableList>
<VarListEntry>
<Term>802.1Q VLAN Implementation for Linux <ULink
URL="http://scry.wanfear.com/~greear/vlan.html"
>(site)</ULink
></Term>
<ListItem>
<Para>
VLANs are a very cool way to segregate your
networks in a more virtual than physical way. Good information on VLANs can
be found <ULink
URL="ftp://ftp.netlab.ohio-state.edu/pub/jain/courses/cis788-97/virtual_lans/index.htm"
>here</ULink
>. With this implementation, you can have your Linux box talk
VLANs with machines like Cisco Catalyst, 3Com: &lcub;Corebuilder, Netbuilder II,
SuperStack II switch 630&rcub;, Extreme Ntwks Summit 48, Foundry: &lcub;ServerIronXL,
FastIron&rcub;.
</Para>
<Para>
A great HOWTO about VLANs can be found <ULink
URL="http://scry.wanfear.com/~greear/vlan/cisco_howto.html"
>here</ULink
>.
</Para>
<Para>
Update: has been included in the kernel as of 2.4.14 (perhaps 13).
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Alternate 802.1Q VLAN Implementation for Linux <ULink
URL="http://vlan.sourceforge.net "
>(site)</ULink
></Term>
<ListItem>
<Para>
Alternative VLAN implementation for linux. This project was started out of
disagreement with the 'established' VLAN project's architecture and coding
style, resulting in a cleaner overall design.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Linux Virtual Server <ULink
URL="http://www.LinuxVirtualServer.org/"
>(site)</ULink
></Term>
<ListItem>
<Para>
These people are brilliant. The Linux Virtual Server is a highly scalable and
highly available server built on a cluster of real servers, with the load
balancer running on the Linux operating system. The architecture of the
cluster is transparent to end users. End users only see a single virtual
server.
</Para>
<Para>
In short whatever you need to load balance, at whatever level of traffic, LVS
will have a way of doing it. Some of their techniques are positively evil!
For example, they let several machines have the same IP address on a
segment, but turn off ARP on them. Only the LVS machine does ARP - it then
decides which of the backend hosts should handle an incoming packet, and
sends it directly to the right MAC address of the backend server. Outgoing
traffic will flow directly to the router, and not via the LVS machine, which
does therefor not need to see your 5Gbit/s of content flowing to the world,
and cannot be a bottleneck.
</Para>
<Para>
The LVS is implemented as a kernel patch in Linux 2.0 and 2.2, but as a
Netfilter module in 2.4/2.5, so it does not need kernel patches! Their 2.4
support is still in early development, so beat on it and give feedback or
send patches.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>CBQ.init <ULink
URL="ftp://ftp.equinox.gu.net/pub/linux/cbq/"
>(site)</ULink
></Term>
<ListItem>
<Para>
Configuring CBQ can be a bit daunting, especially if all you want to do is
shape some computers behind a router. CBQ.init can help you configure Linux
with a simplified syntax.
</Para>
<Para>
For example, if you want all computers in your 192.168.1.0/24 subnet
(on 10mbit eth1) to be limited to 28kbit/s download speed, put
this in the CBQ.init configuration file:
</Para>
<Para>
<Screen>
DEVICE=eth1,10Mbit,1Mbit
RATE=28Kbit
WEIGHT=2Kbit
PRIO=5
RULE=192.168.1.0/24
</Screen>
</Para>
<Para>
By all means use this program if the 'how and why' don't interest you.
We're using CBQ.init in production and it works very well. It can even do
some more advanced things, like time dependent shaping. The documentation is
embedded in the script, which explains why you can't find a README.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Chronox easy shaping scripts <ULink
URL="http://www.chronox.de"
>(site)</ULink
></Term>
<ListItem>
<Para>
Stephan Mueller (smueller@chronox.de) wrote two useful scripts, 'limit.conn'
and 'shaper'. The first one allows you to easily throttle a single download
session, like this:
</Para>
<Para>
<Screen>
# limit.conn -s SERVERIP -p SERVERPORT -l LIMIT
</Screen>
</Para>
<Para>
It works on Linux 2.2 and 2.4/2.5.
</Para>
<Para>
The second script is more complicated, and can be used to make lots of
different queues based on iptables rules, which are used to mark packets
which are then shaped.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Virtual Router
Redundancy Protocol implementation <ULink
URL="http://w3.arobas.net/~jetienne/vrrpd/index.html"
>(site)</ULink
></Term>
<ListItem>
<Para>
This is purely for redundancy. Two machines with their own IP address and
MAC Address together create a third IP Address and MAC Address, which is
virtual. Originally intended purely for routers, which need constant MAC
addresses, it also works for other servers.
</Para>
<Para>
The beauty of this approach is the incredibly easy configuration. No kernel
compiling or patching required, all userspace.
</Para>
<Para>
Just run this on all machines participating in a service:
<Screen>
# vrrpd -i eth0 -v 50 10.0.0.22
</Screen>
</Para>
<Para>
And you are in business! 10.0.0.22 is now carried by one of your servers,
probably the first one to run the vrrp daemon. Now disconnect that computer
from the network and very rapidly one of the other computers will assume the
10.0.0.22 address, as well as the MAC address.
</Para>
<Para>
I tried this over here and had it up and running in 1 minute. For some
strange reason it decided to drop my default gateway, but the -n flag
prevented that.
</Para>
<Para>
This is a 'live' fail over:
</Para>
<Para>
<Screen>
64 bytes from 10.0.0.22: icmp_seq=3 ttl=255 time=0.2 ms
64 bytes from 10.0.0.22: icmp_seq=4 ttl=255 time=0.2 ms
64 bytes from 10.0.0.22: icmp_seq=5 ttl=255 time=16.8 ms
64 bytes from 10.0.0.22: icmp_seq=6 ttl=255 time=1.8 ms
64 bytes from 10.0.0.22: icmp_seq=7 ttl=255 time=1.7 ms
</Screen>
</Para>
<Para>
Not *one* ping packet was lost! Just after packet 4, I disconnected my P200
from the network, and my 486 took over, which you can see from the higher
latency.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</chapter>
<chapter id="lartc.further">
<Title>Further reading</Title>
<Para>
<VariableList>
<VarListEntry>
<Term><ULink
URL="http://snafu.freedom.org/linux2.2/iproute-notes.html"
>http://snafu.freedom.org/linux2.2/iproute-notes.html</ULink
></Term>
<ListItem>
<Para>
Contains lots of technical information, comments from the kernel
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term><ULink
URL="http://www.davin.ottawa.on.ca/ols/"
>http://www.davin.ottawa.on.ca/ols/</ULink
></Term>
<ListItem>
<Para>
Slides by Jamal Hadi Salim, one of the authors of Linux traffic control
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term><ULink
URL="http://defiant.coinet.com/iproute2/ip-cref/"
>http://defiant.coinet.com/iproute2/ip-cref/</ULink
></Term>
<ListItem>
<Para>
HTML version of Alexeys LaTeX documentation - explains part of iproute2 in
great detail
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term><ULink
URL="http://www.aciri.org/floyd/cbq.html"
>http://www.aciri.org/floyd/cbq.html</ULink
></Term>
<ListItem>
<Para>
Sally Floyd has a good page on CBQ, including her original papers. None of
it is Linux specific, but it does a fair job discussing the theory and uses
of CBQ.
Very technical stuff, but good reading for those so inclined.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Differentiated Services on Linux</Term>
<ListItem>
<Para>
This <ULink
URL="ftp://icaftp.epfl.ch/pub/linux/diffserv/misc/dsid-01.txt.gz"
>document</ULink
> by Werner Almesberger, Jamal Hadi Salim and Alexey
Kuznetsov describes DiffServ facilities in the Linux kernel, amongst which
are TBF, GRED, the DSMARK qdisc and the tcindex classifier.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term><ULink
URL="http://ceti.pl/~kravietz/cbq/NET4_tc.html"
>http://ceti.pl/~kravietz/cbq/NET4_tc.html</ULink
></Term>
<ListItem>
<Para>
Yet another HOWTO, this time in Polish! You can copy/paste command lines
however, they work just the same in every language. The author is
cooperating with us and may soon author sections of this HOWTO.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term><ULink
URL="http://www.cisco.com/univercd/cc/td/doc/product/software/ios111/cc111/car.htm"
>IOS Committed Access Rate</ULink
></Term>
<ListItem>
<Para>
<Anchor id="CAR">
>From the helpful folks of Cisco who have the laudable habit of putting
their documentation online. Cisco syntax is different but the concepts are
the same, except that we can do more and do it without routers the price of
cars :-)
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>Docum experimental site<ULink
URL="http://www.docum.org"
>(site)</ULink
></Term>
<ListItem>
<Para>
Stef Coene is busy convincing his boss to sell Linux support, and so he is
experimenting a lot, especially with managing bandwidth. His site has a lot
of practical information, examples, tests and also points out some CBQ/tc bugs.
</Para>
</ListItem>
</VarListEntry>
<VarListEntry>
<Term>TCP/IP Illustrated, volume 1, W. Richard Stevens, ISBN 0-201-63346-9</Term>
<ListItem>
<Para>
Required reading if you truly want to understand TCP/IP. Entertaining as
well.
</Para>
</ListItem>
</VarListEntry>
</VariableList>
</Para>
</chapter>
<chapter id="lartc.ack">
<Title>Acknowledgements </Title>
<Para>
It is our goal to list everybody who has contributed to this HOWTO, or
helped us demystify how things work. While there are currently no plans
for a Netfilter type scoreboard, we do like to recognize the people who are
helping.
</Para>
<Para>
<ItemizedList spacing="compact">
<ListItem>
<Para>
<author><firstname>Junk</firstname><surname>Alins</surname></author>
<address><email>juanjo@mat.upc.es</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Joe</firstname><surname>Van Andel</surname></author>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Michael</firstname><othername>T.</othername>
<surname>Babcock</surname></author>
<address><email>mbabcock@fibrespeed.net</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Christopher</firstname>
<surname>Barton</surname></author>
<address><email>cpbarton%uiuc.edu</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Ard</firstname><surname>van Breemen</surname></author>
<address><email>ard&percnt;kwaak.net</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Ron</firstname><surname>Brinker</surname></author>
<address><email>service&percnt;emcis.com</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>?ukasz</firstname><surname>Bromirski</surname></author>
<address><email>L.Bromirski@prosys.com.pl</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Lennert</firstname><surname>Buytenhek</surname></author>
<address><email>buytenh@gnu.org</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Esteve</firstname><surname>Camps</surname></author>
<address><email>esteve@hades.udg.es</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Stef</firstname><surname>Coene</surname></author>
<address><email>stef.coene@docum.org</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Don</firstname><surname>Cohen</surname></author>
<address><email>don-lartc&percnt;isis.cs3-inc.com</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Jonathan</firstname><surname>Corbet</surname></author>
<address><email>lwn&percnt;lwn.net</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Gerry</firstname><surname>Creager</surname>
<othername>N5JXS</othername></author>
<address><email>gerry&percnt;cs.tamu.edu</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Marco</firstname><surname>Davids</surname></author>
<address><email>marco@sara.nl</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Jonathan</firstname><surname>Day</surname></author>
<address><email>jd9812@my-deja.com</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Martin</firstname><surname>Devera</surname>
<othername>aka devik</othername></author>
<address><email>devik@cdi.cz</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Stephan</firstname><othername>"Kobold"</othername>
<surname>Gehring</surname></author>
<address><email>Stephan.Gehring@bechtle.de</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Jacek</firstname><surname>Glinkowski</surname></author>
<address><email>jglinkow&percnt;hns.com</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Andrea</firstname><surname>Glorioso</surname></author>
<address><email>sama&percnt;perchetopi.org</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Nadeem</firstname><surname>Hasan</surname></author>
<address><email>nhasan@usa.net</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Erik</firstname><surname>Hensema</surname></author>
<address><email>erik&percnt;hensema.xs4all.nl</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Vik</firstname><surname>Heyndrickx</surname></author>
<address><email>vik.heyndrickx@edchq.com</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Spauldo</firstname><surname>Da Hippie</surname></author>
<address><email>spauldo&percnt;usa.net</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Koos</firstname><surname>van den Hout</surname></author>
<address><email>koos@kzdoos.xs4all.nl</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
Stefan Huelbrock &lt;shuelbrock&percnt;datasystems.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Alexander W. Janssen &lt;yalla&percnt;ynfonatic.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Gareth John &lt;gdjohn&percnt;zepler.org&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Martin Josefsson &lt;gandalf&percnt;wlug.westbo.se&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Andi Kleen &lt;ak&percnt;suse.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Andreas J. Koenig &lt;andreas.koenig&percnt;anima.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Pawel Krawczyk &lt;kravietz&percnt;alfa.ceti.pl&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Amit Kucheria &lt;amitk@ittc.ku.edu&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Edmund Lau &lt;edlau&percnt;ucf.ics.uci.edu&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Philippe Latu &lt;philippe.latu&percnt;linux-france.org&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Arthur van Leeuwen &lt;arthurvl&percnt;sci.kun.nl&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Jason Lunz &lt;j@cc.gatech.edu&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Stuart Lynne &lt;sl@fireplug.net&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Alexey Mahotkin &lt;alexm@formulabez.ru&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Predrag Malicevic &lt;pmalic@ieee.org&gt;
</Para>
</ListItem>
<listitem>
<para>
Patrick McHardy &lt;kaber@trash.net&gt;
</para>
</ListItem>
<ListItem>
<Para>
Andreas Mohr &lt;andi&percnt;lisas.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Andrew Morton &lt;akpm@zip.com.au&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Wim van der Most
</Para>
</ListItem>
<ListItem>
<Para>
Stephan Mueller &lt;smueller@chronox.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Togan Muftuoglu &lt;toganm&percnt;yahoo.com&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Chris Murray &lt;cmurray@stargate.ca&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Patrick Nagelschmidt &lt;dto&percnt;gmx.net&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Ram Narula &lt;ram@princess1.net&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Jorge Novo &lt;jnovo@educanet.net&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Patrik &lt;ph@kurd.nu&gt;
</Para>
</ListItem>
<listitem><para>P?l Osgy?ny &lt;oplab%westel900.net&gt;</para></listitem>
<ListItem>
<Para>
Lutz Pre&szlig;ler &lt;Lutz.Pressler&percnt;SerNet.DE&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Jason Pyeron &lt;jason&percnt;pyeron.com&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Rusty Russell &lt;rusty&percnt;rustcorp.com.au&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Mihai RUSU &lt;dizzy&percnt;roedu.net&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Jamal Hadi Salim &lt;hadi&percnt;cyberus.ca&gt;
</Para>
</ListItem>
<ListItem>
<Para>
David Sauer &lt;davids&percnt;penguin.cz&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Sheharyar Suleman Shaikh &lt;sss23@drexel.edu&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Stewart Shields &lt;MourningBlade&percnt;bigfoot.com&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Nick Silberstein &lt;nhsilber&percnt;yahoo.com&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Konrads Smelkov &lt;konrads@interbaltika.com&gt;
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>William</firstname><surname>Stearns</surname></author>
<address><email>wstearns@pobox.com</email></address>
</Para>
</ListItem>
<ListItem>
<Para>
Andreas Steinmetz &lt;ast&percnt;domdv.de&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Jason Tackaberry &lt;tack@linux.com&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Charles Tassell &lt;ctassell&percnt;isn.net&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Glen Turner &lt;glen.turner&percnt;aarnet.edu.au&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Tea Sponsor: Eric Veldhuyzen &lt;eric&percnt;terra.nu&gt;
</Para>
</ListItem>
<ListItem>
<Para>
Song Wang &lt;wsong@ece.uci.edu&gt;
</Para>
</ListItem>
<ListItem>
<Para>
<author><firstname>Lazar</firstname><surname>Yanackiev</surname></author>
<address><email>Lyanackiev%gmx.net</email></address>
</Para>
</ListItem>
</ItemizedList>
</Para>
</chapter>
</Book>
View Git Blame Copy Permalink