bird/doc/bird.sgml

<!doctype birddoc system>

<!--
	BIRD documentation

This documentation can have 4 forms: sgml (this is master copy), html,
ASCII text and dvi/postscript (generated from sgml using
sgmltools). You should always edit master copy.

This is slightly modified linuxdoc dtd.  Anything in <descrip> tags is considered definition of
configuration primitives, <cf> is fragment of configuration within normal text, <m> is
"meta" information within fragment of configuration - something in config which is not keyword.

    (set-fill-column 100)

    Copyright 1999,2000 Pavel Machek <pavel@ucw.cz>, distribute under GPL version 2 or later.

 -->

<book>

<title>BIRD User's Guide
<author>
Ondrej Filip <it/&lt;feela@network.cz&gt;/,
Pavel Machek <it/&lt;pavel@ucw.cz&gt;/,
Martin Mares <it/&lt;mj@ucw.cz&gt;/
</author>

<abstract>
This document contains user documentation for the BIRD Internet Routing Daemon project.
</abstract>

<!-- Table of contents -->
<toc>

<!-- Begin the document -->

<chapt>Introduction

<sect>What is BIRD

<p><label id="intro">
The name `BIRD' is actually an acronym standing for `BIRD Internet Routing Daemon'.
Let's take a closer look at the meaning of the name:

<p><em/BIRD/: Well, we think we have already explained that. It's an acronym standing
for `BIRD Internet Routing Daemon', you remember, don't you? :-)

<p><em/Internet Routing/: It's a program (well, a daemon, as you are going to discover in a moment)
which works as a dynamic router in an Internet type network (that is, in a network running either
the IPv4 or the IPv6 protocol). Routers are devices which forward packets between interconnected
networks in order to allow hosts not connected directly to the same local area network to
communicate with each other. They also communicate with the other routers in the Internet to discover
the topology of the network which allows them to find optimal (in terms of some metric) rules for
forwarding of packets (which will be called routes in the rest of this document) and to adapt themselves to the
changing conditions such as outages of network links, building of new connections and so on. Most of
these routers are costly dedicated devices running obscure firmware which is hard to configure and
not open to any changes (on the other hand, their special hardware design allows them to keep up with lots of high-speed network interfaces, better than general-purpose computer does). Fortunately, most operating systems of the UNIX family allow an ordinary 
computer to act as a router and forward packets belonging to the other hosts, but only according to
a statically configured table.

<p>A <em/Routing Daemon/ is in UNIX terminology a non-interactive program running on
background which does the dynamic part of Internet routing, that is it communicates
with the other routers, calculates routing tables and sends them to the OS kernel
which does the actual packet forwarding. There already exist other such routing daemons: routed (rip only), GateD<HTMLURL URL="http://www.gated.org/">
 (non free), Zebra<HTMLURL URL="http://www.zebra.org"> and mrtd<HTMLURL URL="FIXME">, but their capabilities are limited and
they are relatively hard to configure and maintain.

<p>BIRD is an Internet Routing Daemon designed to avoid all of these shortcomings,
to support all the routing technology used in the today's Internet or planned to be
used in near future and to have a clean extensible architecture allowing new routing
protocols to be incorporated easily. Among other features, BIRD supports:

<itemize>
	<item>both IPv4 and IPv6 protocols
	<item>multiple routing tables
	<item>the Border Gateway Protocol (BGPv4)
	<item>the Routing Interchange Protocol (RIPv2)
	<item>the Open Shortest Path First protocol (OSPFv2)
	<item>a virtual protocol for exchange of routes between different routing tables on a single host
	<item>a command-line interface allowing on-line control and inspection
		of status of the daemon
	<item>soft reconfiguration (no need to use complex online commands
		to change the configuration, just edit the configuration file
		and notify BIRD to re-read it and it will smoothly switch itself
		to the new configuration, not disturbing routing protocols
		unless they are affected by the configuration changes)
	<item>a powerful language for route filtering
</itemize>

<p>BIRD has been developed at the Faculty of Math and Physics, Charles University, Prague,
Czech Republic as a student project. It can be freely distributed under the terms of the GNU General
Public License.

<p>BIRD has been designed to work on all UNIX-like systems. It has been developed and
tested under Linux 2.0 to 2.4, but porting to other systems (even non-UNIX ones) should
be relatively easy due to its highly modular architecture.

<sect>Installing BIRD

<p>On a recent UNIX system with GNU development tools (GCC, binutils, m4, make) and Perl, installing BIRD should be as easy as:

<code>
        ./configure
        make
        make install
        vi /usr/local/etc/bird.conf
	bird
</code>

<p>You can use <tt>./configure --help</tt> to get a list of configure
options. The most important ones are:
<tt/--enable-ipv6/ which enables building of an IPv6 version of BIRD,
<tt/--with-protocols=/ to produce a slightly smaller BIRD executable by configuring out routing protocols you don't use, and
<tt/--prefix=/ to install BIRD to a place different from.
<file>/usr/local</file>.

<sect>Running BIRD

<p>You can pass several command-line options to bird:

<descrip>
	<tag>-c <m/config name/</tag>
	use given configuration file instead of <file>$prefix/etc/bird.conf</file>.

	<tag>-d</tag>
	enable debug messages and run bird in foreground.

	<tag>-D <m/filename of debug log/</tag>
	log debugging information to given file instead of stderr

	<tag>-s <m/name of communication socket/</tag>
	use given filename for a  socket for communications with the client, default is <file>$prefix/var/run/bird.ctl</file>.
</descrip>

<p>BIRD writes messages about its work to log files or syslog (according to config).

<chapt>About routing tables

<p>BIRD has one or more routing tables, which may or may not be
synchronized with kernel and which may or may not be synchronized with
each other (see the Pipe protocol). Each routing table contains list of
known routes. Each route consists of:

<itemize>
	<item>network prefix this route is for (consists of networkk address and number of bits forming the network part of the address)
	<item>preference of this route (taken from preference of
	protocol and possibly altered by filters)
	<item>IP address of router who told us about this route
	<item>IP address of router we should forward the packets to
	using this route
	<item>other attributes common to all routes
	<item>dynamic attributes defined by protocols which may or
	may not be present (typically protocol metrics)
</itemize>

Routing table maintains more than
one entry for a network, but at most one entry for one network and one
protocol. The entry with biggest preference is used for routing. If
there are more entries with the same preference and they are from the same
protocol, protocol decides (typically according to metrics). If not,
internal ordering is used to decide. You can
get list of route attributes in "Route attributes" section in
filters. 

<p>Protocols are connected to routing tables through filters. Routes
that come from other routers go to the protocol, it then passes them to
filters, if import filter accepts route, it gets to main routing
table. It is then broadcasted to all other protocols (filtered through
their export filters), which typically send it to other routers.

Filters can alter routes passed between routing tables and
protocols.

<chapt>Configuration

<sect>Introduction

<!-- fixme: add default values to all places? Where do I learn them? --> 

<p>BIRD is configured using a text configuration file. Upon startup, BIRD reads <file>$prefix/bird.conf</file> (unless the
<tt/-c/ command line option is given). Configuration may be changed on user's request: if you modify
config file and then signal BIRD with SIGHUP, it will adjust to the new
config. Then there's the client,
which allows you to talk with BIRD in an extensive way. (Of course you can tell BIRD to reconfigure from BIRDC, you can also tell it to shut down, dump various info etc.).

<p>In the config, everything on a line after <cf/#/ or inside <cf>/*
*/</cf> is a comment, whitespace characters are treated as a single space. If there's a variable number of options, they are grouped using
the <cf/{ }/ brackets. Each option is terminated by a <cf/;/. Configuration
is case sensitive.

<p>Here is an example of a simple config file. It enables
synchronization of routing tables with OS kernel, scans for 
new network interfaces every 10 seconds and runs RIP on all network interfaces found.


<code>
protocol kernel {
	persist;		# Don't remove routes on BIRD shutdown
	scan time 20;		# Scan kernel routing table every 20 seconds
	export all;		# Default is export none
}

protocol device {
	scan time 10;		# Scan interfaces every 10 seconds
}

protocol rip {
	export all;
	import all;
}
</code>


<sect>Global options

<p><descrip>
	<tag>log "<m/filename/"|syslog|stderr all|{ <m/list of classes/ }</tag> 
	Set logging of messages having the given (either <cf/all/ or <cf/{
	error, trace }/ etc.) into selected destination. Classes are:
	<cf/info/, <cf/warning/, <cf/error/ and <cf/fatal/ for messages about local problems
	<cf/debug/ for debugging messages, 
	<cf/trace/ when you want to know what happens in the network, 
	<cf/remote/ for messages about misbehavior of remote machines, 
	<cf/auth/ about authentication failures,
	<cf/bug/ for internal BIRD bugs. You may specify more than one <cf/log/ line to estabilish logging to multiple
	destinations.

	<!--fixme: mj says explain-->
				  
	<tag>debug protocols all|off|{ states, routes, filters, interfaces, events, packets }</tag>
	Set global default of protocol debugging options.

	<tag>filter <m/name local variables/{ <m/commands/ }</tag> Define a filter. You can learn more about filters
	in the following chapter.

	<tag>function <m/name (parameters) local variables/ { <m/commands/ }</tag> Define a function. You can learn more
	about functions in the following chapter.
 
	<tag>protocol rip|ospf|bgp|... <m/[name]/ { <m>protocol options</m> }</tag> Define a protocol
	instance called <cf><m/name/</cf> (or with a name like "rip5" generated automatically, if you don't specify <cf><m/name/</cf>). You can learn more
	about configuring protocols in their own chapters. You can run more than one instance of
	most protocols (like RIP or BGP).

	<tag>define <m/constant/ = (<m/expression/)|<m/number/|<m/IP address/</tag> Define a constant. You can use it later in every place
	you could use a simple integer or IP address.

	<tag>router id <m/IPv4 address/</tag> Set BIRD's router ID. It's a world-wide unique identification of your router, usually one of router's IPv4 addresses.

	<tag>table <m/name/</tag> Create a new routing table. The default
	routing table is created implicitly, other routing tables have
	to be added by this command.

	<tag>eval <m/expr/</tag> Evaluates given filter expression. It
	is used by us for testing filters.
</descrip>

<sect>Protocol options

<p>Each routing protocol has its own set of options (see the corresponding sections). These options are common to all protocols:

<descrip>
	<tag>preference <m/expr/</tag> Sets the preference of routes generated by this protocol.

	<tag>disabled</tag> Disables the protocol. You can change the disable/enable status from the command
	line interface without needing to touch the configuration. Disabled protocol is not activated.

	<tag>debug <m/setting/</tag> Similar to global <cf/debug/ setting, except that it only
	affects one protocol. Only messages in selected debugging categories will be written to the
	logs.

	<tag>import { <m/filter commands/ } | <m/name/ | all | none</tag> Specify a filter to be used for filtering routes comming from protocol tothe routing table. Default: <cf/all/. <cf/all/ is shorthand for <cf/{ accept; }/ and <cf/none/ is shorthand for <cf/{ reject; }/.

	<tag>export <m/filter/</tag> This is similar to <cf>import</cf> keyword, except that it
	works in direction from the routing table to the protocol. Default: <cf/none/

	<tag>table <m/name/</tag> Connect this protocol to a non-default routing table.
</descrip>

<p>There are several options that give sense only with certain protocols:

<descrip>
	<tag>passwords { password "<m/password/" from <m/time/ to <m/time/ passive <m/time/ id
	<m/num/ [...] }</tag> Specifies passwords to be used with this protocol. <cf>Passive <m/time/</cf> is
	time from which the password is not used for sending, but it is recognized on reception. <cf/id/ is password id, as needed by
	certain protocols.

	<tag>interface "<m/mask/"|<m/prefix/ [ { <m/option/ ; [ ... ] } ]</tag> Specifies which
	interfaces this protocol is active on, and allows you to set options on
	per-interface basis. Mask is specified in shell-like patterns, thus <cf>interface
	"*" { mode broadcast; };</cf> will start the protocol on all interfaces with <cf>mode
	broadcast;</cf> option. If first character of mask is <cf/-/, such interfaces are
	excluded. Masks are parsed left-to-right, thus <cf/interface "-eth*", "*";/ means all but
	the ethernets.
						  
</descrip>

<sect>Client

<p>You can use the command-line client <file>birdc</file> to talk with
a running BIRD. Communication is done using <file/bird.ctl/ unix domain
socket (unless changed with the <tt/-s/ option given to both the server and
the client). The client can do simple actions such as enabling/disabling
protocols, telling BIRD to show various information, telling it to
show routing table filtered by any filter, or telling bird to
reconfigure. Press <tt/?/ at any time to get online help. Option
<tt/-v/ can be passed to the client, telling it to dump numeric return
codes. You do not necessarily need to use BIRDC to talk to BIRD, your
own application could do that, too -- format of communication between
BIRD and BIRDC is stable (see programmer's documentation).


<chapt>Filters

<sect>Introduction

<p>BIRD contains a rather simple programming language. (No, it can't yet read mail :-). There are
two objects in this language: filters and functions. Filters are called by BIRD core when a route is
being passed between protocola and routing tables. Filter language contains control structures such
as if's and switches, but it allows no loops. Filters are
interpreted. An example of a filter using many features can be found in <file>filter/test.conf</file>. 

<p>Filter gets the route, looks at its attributes and
modifies some of them if it wishes. At the end, it decides whether to
pass the changed route through (using <cf/accept/) or whether to <cf/reject/ given route. A simple filter looks
like this:

<code>
filter not_too_far
int var;
{
	if defined( rip_metric ) then
		var = rip_metric;
	else {
		var = 1;
		rip_metric = 1;
	}
	if rip_metric &gt; 10 then
		reject "RIP metric is too big";
	else
		accept "ok";
}
</code>

<p>As you can see, a filter has a header, a list of local variables, and a body. The header consists of
the <cf/filter/ keyword, followed by a (unique) name of filter. List of local variables consists of
pairs <cf><M>type name</M>;</cf>, where each pair defines one local variable. Body consists of
<cf> { <M>statements</M> }</cf>. Each <m/statement/ is terminated by <cf/;/. You can group
several statements into one by using braces (<cf>{ <M>statements</M> }</cf>), that is useful if
you want to make bigger block of code conditional.

<p>BIRD supports functions, so that you don't have to repeat same blocks of code over and
over. Functions can have zero or more parameters and they can have local variables. Recursion is not allowed. They
look like this:

<code>
function name ()
int local_variable;
{
	local_variable = 5;
}

function with_parameters (int parameter)
{
	print parameter;
}
</code>

<p>Unlike in C, variables are declared after function line but before the first {. You can't declare
variables in nested blocks. Functions are called like in C: <cf>name();
with_parameters(5);</cf>. Function may return value using the <cf>return <m/[expr]/</cf>
command. Returning a value exits from current function (this is similar to C).

<p>Filters are declared in a way similar to functions except they can't have explicit
parameters. They get route table entry as an implicit parameter, it is also passed automatically 
to any functions called. The filter must terminate with either
<cf/accept/ or <cf/reject/ statement. If there's runtime error in filter, the route
is rejected. 

<p>A nice trick to debug filters is to use <cf>show route filter
<m/name/</cf> from the command line client. An example session might look
like:

<code>
pavel@bug:~/bird$ ./birdc -s bird.ctl
BIRD 0.0.0 ready.
bird> show route
10.0.0.0/8         dev eth0 [direct1 23:21] (240)
195.113.30.2/32    dev tunl1 [direct1 23:21] (240)
127.0.0.0/8        dev lo [direct1 23:21] (240)
bird> show route ?
show route [<prefix>] [table <t>] [filter <f>] [all] [primary] [(import|protocol) <p>]...
bird> show route filter { if 127.0.0.5 ~ net then accept; }
127.0.0.0/8        dev lo [direct1 23:21] (240)
bird>
</code>

<sect>Data types

<p>Each variable and each value has certain type. Booleans, integers and enums are
incompatible with each other (that is to prevent you from shooting in the foot).

<descrip>
	<tag/bool/ This is a boolean type, it can have only two values, <cf/true/ and
	  <cf/false/. Boolean is not compatible with integer and is the only type you can use in if
	  statements.

	<tag/int/ This is a general integer type, you can expect it to store signed values from -2000000000
	  to +2000000000. Overflows are not checked. You can use <cf/0x1234/ syntax to write hexadecimal values.

	<tag/pair/ This is a pair of two short integers. Each component can have values from 0 to
	  65535. Literals of this type is written as <cf/(1234,5678)/.

	<tag/string/ This is a string of characters. There are no ways to modify strings in
	  filters. You can pass them between functions, assign them to variables of type <cf/string/, print
	  such variables, but you can't concatenate two strings. String literals
	  are written as <cf/"This is a string constant"/.

	<tag/ip/ This type can hold a single IP address. Depending on the compile-time configuration of BIRD you are using, it
	  is either an IPv4 or IPv6 address. IPv4 addresses are written (as you would expect) as
	  <cf/1.2.3.4/. You can apply special operator <cf>.mask(<M>num</M>)</cf>
	  on values of type ip. It masks out all but first <cf><M>num</M></cf> bits from ip
	  address. So <cf/1.2.3.4.mask(8) = 1.0.0.0/ is true. <!-- FIXME: IPv6? -->

	<tag/prefix/ This type can hold a network prefix consisting of IP address and prefix length. Prefix literals are written as
	  <cf><M>ipaddress</M>/<M>pxlen</M></cf>, or
	  <cf><m>ipaddress</m>/<m>netmask</m></cf> There are two special
	  operators on prefix:
	  <cf/.ip/, which separates ip address from the pair, and <cf/.len/, which separates prefix
	  len from the pair. So <cf>1.2.0.0/16.pxlen = 16</cf> is true.

	<tag/int|ip|prefix|pair|enum set/
	  Filters recognize four types of sets. Sets are similar to strings: you can pass them around
	  but you can't modify them. Literals of type <cf>set int</cf> look like <cf>
	  [ 1, 2, 5..7 ]</cf>. As you can see, both simple values and ranges are permitted in
	  sets. Sets of prefixes are special: you can specify which prefix lengths should match them by
	  using <cf>[ 1.0.0.0/8+, 2.0.0.0/8-, 3.0.0.0/8{5,6} ]</cf>. <cf>3.0.0.0/8{5,6}</cf> matches
	  prefixes <cf/3.X.X.X/, whose prefix length is 5 to 6. <cf>3.0.0.0/8+</cf> is shorthand for <cf>3.0.0.0/{0,8}</cf>,
	  <cf>3.0.0.0/8-</cf> is shorthand for <cf>3.0.0.0/{0,7}</cf>. For example,
	  <cf>1.2.0.0/16 ~ [ 1.0.0.0/8{ 15 , 17 } ]</cf> is true, but
	  <cf>1.0.0.0/8 ~ [ 1.0.0.0/8- ]</cf> is false. <!-- fixme: use variables instead of examples -->

	<tag/enum/
	  Enumeration types are fixed in BIRD - you can't define your own
	  variables of enumeration type, but some route attributes are of enumeration
	  type. Enumeration types are incompatible with each other.

	<tag/bgppath/
	  BGP path is a list of autonomous system numbers. You can't write literals of this type.

	<tag/bgpmask/ 
	  BGP mask is a mask used for matching BGP paths
	  (using <cf>path &tilde; /2 3 5 ?/</cf> syntax). Matching is
	  done using shell-like patterns: <cf/?/ means
	  "any number of any autonomous systems". Pattern for single
	  unknown autonomous system is not supported. (We
	  did not want to use * because then it becomes too easy to
	  write <cf>/*</cf> which is start of comment.) For example,
	  <tt>/4 3 2 1/ ~ /? 4 3 ?/</tt> is true, but 
	  <tt>/4 3 2 1/ ~ /? 4 5 ?/</tt> is false. <!-- fixme: formulate better -->

	<tag/clist/ 
	  Community list is similar to set of pairs,
	  except that unlike other sets, it can be modified.
	  There exist no literals of this type.

</descrip>

<sect>Operators

<!-- FIXME: Make it table -->

<p>The filter language supports common integer operators <cf>(+,-,*,/)</cf>, parentheses <cf/(a*(b+c))/, comparison
<cf/(a=b, a!=b, a&lt;b, a&gt;=b)/. Special operators include <cf/&tilde;/ for "in" operation. In operation can be
used on element and set of that elements, or on ip and prefix, or on
prefix and prefix or on bgppath and bgpmask or on pair and clist. Its result
is true if element is in given set or if ip address is inside given prefix. Logical operations include unary not (<cf/!/), and (<cf/&&/) and or (<cf/||/).

<sect>Control structures

<p>Filters support two control structures: conditions and case switches. 

<!-- fixme: say explicitly what if and case does -->

<p>Syntax of condition is <cf>if
<M>boolean expression</M> then <M>command</M>; else <M>command</M>;</cf> and you can use <cf>{
<M>command_1</M>; <M>command_2</M>; <M>...</M> }</cf> instead of one or both commands. <cf>else</cf>
clause may be omitted.

<p><cf>case</cf> is similar to case from Pascal. Syntax is <cf>case <m/expr/ { else |
<m/num_or_prefix [ .. num_or_prefix]/: <m/statement/ ; [ ... ] }</cf>. Expression after
<cf>case</cf> can be of any type that can be on the left side of the &tilde; operator, and anything that could
be a member of a set is allowed before <cf/:/. Multiple commands are allowed without <cf/{}/ grouping
and break is implicit before each case. If argument
matches neither of <cf/:/ clauses, <cf/else:/ clause is used.

<p>Here is example that uses <cf/if/ and <cf/case/ structures:

<code>
case arg1 {
	2: print "two"; print "I can do more commands without {}";
	3 .. 5: print "three to five";
	else: print "something else";
}

if 1234 = i then printn "."; else { 
  print "not 1234"; 
  print "You need {} around multiple commands"; 
}
</code>

<sect>Route attributes

<p>An filter is implicitly passed route, and it can access its
attributes just like it accesses variables. Attempt to access undefined
attribute result in a runtime error; you can check if an attribute is
defined using the <cf>defined( <m>attribute</m> )</cf> operator.

<!-- fixme: say which are read-only -->

<descrip>
	<tag><m/prefix/ net</tag>
	Network the route is talking about. (See the section about routing tables.)

	<tag><m/enum/ scope</tag>
	Address scope of the network (<cf/SCOPE_HOST/ for addresses local to this host, <cf/SCOPE_LINK for those specific for a physical link, ... <!-- FIXME -->)

	<tag><m/int/ preference</tag>
	Preference of the route. (See section about routing tables.)

	<tag><m/ip/ from</tag>
	The router which the route has originated from.
	
	<tag><m/ip/ gw</tag>
	Next hop packets routed using this route should be forwarded to.

	<tag><m/enum/ source</tag>
	what protocol has told me about this route. Possible values: <cf/RTS_RIP/ or <cf/RTS_OSPF_EXT/ <!-- fixme -->.

	<tag><m/enum/ cast</tag>
	Route type (<cf/RTC_UNICAST/ for normal routes, ... <!-- FIXME -->)

	<tag><m/enum/ dest</tag>
	Type of destination the packets should be sent to (<cf/RTD_ROUTER/ for forwarding to a neighboring router, ...)
</descrip>

<p>There also exist some protocol-specific attributes, which are described in protocol sections.

<sect>Statements

<p>The following statements are available:

<descrip>
	<tag><m/variable/ = <m/expr/</tag> Set variable to a given value.

	<tag>accept|reject [ <m/expr/ ]</tag> Accept or reject the route, possibly printing <cf><m>expr</m></cf>.

	<tag>return <m/expr/</tag> Return <cf><m>expr</m></cf> from function, the function ends at this point.

	<tag>print|printn <m/expr/ [<m/, expr.../]</tag>
	Prints given expressions; useful mainly while debugging
	filters. The <cf/printn/ variant does not terminate the line.

	<tag>quitbird</tag>
	Terminates BIRD. Useful when debugging filter interpreter.
</descrip>

<chapt>Protocols

<sect>BGP

<p>The Border Gateway Protocol is the routing protocol used for backbone
level routing in the today's Internet. Contrary to other protocols, its convergence
doesn't rely on all routers following the same rules for route selection,
making it possible to implement any routing policy at any router in the
network, the only restriction being that if a router advertises a route,
it must accept and forward packets according to it.

<p>BGP works in terms of autonomous systems (often abbreviated as AS). Each
AS is a part of the network with common management and common routing policy.
Routers within each AS usually communicate using either a interior routing
protocol (such as OSPF or RIP) or an interior variant of BGP (called iBGP).
Boundary routers at the border of the AS communicate with their peers
in the neighboring AS'es via exterior BGP (eBGP).

<p>Each BGP router sends to its neighbors updates of the parts of its
routing table it wishes to export, along with complete path information
(a list of AS'es the packet will travel through if it uses that particular
route) in order to avoid routing loops.

<p>BIRD supports all requirements of the BGP4 standard as defined in
RFC 1771<htmlurl url="ftp://ftp.rfc-editor.org/in-notes/rfc1771.txt">
including several enhancements from the
latest draft<htmlurl url="ftp://ftp.rfc-editor.org/internet-drafts/draft-ietf-idr-bgp4-09.txt">.
It also supports the community attributes as per
RFC 1997<htmlurl url="ftp://ftp.rfc-editor.org/in-notes/rfc1997.txt">,
capability negotiation defined in
RFC 2842<htmlurl url="ftp://ftp.rfc-editor.org/in-notes/rfc2842.txt">,
For IPv6, it uses the standard multiprotocol extensions defined in
RFC 2283<htmlurl url="ftp://ftp.rfc-editor.org/in-notes/rfc2283.txt">
including changes described in the
latest draft <htmlurl url="ftp://ftp.rfc-editor.org/internet-drafts/draft-ietf-idr-bgp4-multiprotocol-v2-05.txt">
and applied to IPv6 according to
RFC 2545<htmlurl url="ftp://ftp.rfc-editor.org/in-notes/rfc2545.txt">.

<sect1>Route selection rules

<p>BGP doesn't have any simple metric, so the rules for selection of an optimal
route among multiple BGP routes with the same preference are a bit more complex
and are implemented according to the following algorithm. First it uses the first
rule, if there are more "best" routes, then it uses the second rule to choose
among them and so on.

<itemize>
	<item>Prefer route with the highest local preference attribute.
	<item>Prefer route with the shortest AS path.
	<item>Prefer IGP origin over EGP and EGP over incomplete.
	<item>Prefer the lowest value of the Multiple Exit Discriminator.
	<item>Prefer internal routes over external routes.
	<item>Prefer route with the lowest value of router ID of the
	advertising router.
</itemize>

<sect1>Configuration

<p>Each instance of the BGP corresponds to one neighboring router.
This allows to set routing policy and all other parameters differently
for each neighbor using the following protocol parameters:

<descrip>
	<tag>local as <m/number/</tag> Define which AS we are part of. (Note that
	contrary to other IP routers, BIRD is able to act as a router located
	in multiple AS'es simultaneously, but in such cases you need to tweak
	the BGP paths manually in the filters to get consistent behavior.)
	This parameter is mandatory.
	<tag>neighbor <m/ip/ as <m/number/</tag> Define neighboring router
	this instance will be talking to and what AS it's located in. Unless
	you use the <cf/multihop/ clause, it must be directly connected to one
	of your router's interfaces. In case the neighbor is in the same AS
	as we are, we automatically switch to iBGP. This parameter is mandatory.
	<tag>multihop <m/number/ via <m/ip/</tag> Configure multihop BGP to a
	neighbor which is connected at most <m/number/ hops far and to which
	we should route via our direct neighbor with address <m/ip/.
	Default: switched off.
	<tag>next hop self</tag> Avoid calculation of the Next Hop attribute
	and always advertise our own source address (see below) as a next hop.
	This needs to be used only
	occasionally to circumvent misconfigurations of other routers.
	Default: disabled.
	<tag>source address <m/ip/</tag> Define local address we should use
	for next hop calculation. Default: the address of the local end
	of the interface our neighbor is connected to.
	<tag>disable after error <m/switch/</tag> When an error is encountered (either
	locally or by the other side), disable the instance automatically
	and wait for an administrator to solve the problem manually. Default: off.
	<tag>hold time <m/number/</tag> Time in seconds to wait for a keepalive
	message from the other side before considering the connection stale.
	Default: depends on agreement with the neighboring router, we prefer
	240 seconds if the other side is willing to accept it.
	<tag>startup hold time <m/number/</tag> Value of the hold timer used
	before the routers have a chance to exchange OPEN messages and agree
	on the real value. Default: 240 seconds.
	<tag>keepalive time <m/number/</tag> Delay in seconds between sending
	of two consecutive keepalive messages. Default: One third of the hold time.
	<tag>connect retry time <m/number/</tag> Time in seconds to wait before
	retrying a failed connect attempt. Default: 120 seconds.
	<tag>start delay time <m/number/</tag> Delay in seconds between protocol
	startup and first attempt to connect. Default: 5 seconds.
	<tag>error wait time <m/number/, <m/number/</tag> Minimum and maximum delay in seconds between protocol
	failure (either local or reported by the peer) and automatic startup.
	Doesn't apply when <cf/disable after error/ is configured. If consecutive
	errors happen, the delay is increased exponentially until it reaches the maximum. Default: 60, 300.
	<tag>error forget time <m/number/</tag> Maximum time in seconds between two protocol
	failures to treat them as a error sequence which makes the <cf/error wait time/
	increase exponentially. Default: 300 seconds.
	<tag>path metric <m/switch/</tag> Enable comparison of path lengths
	when deciding which BGP route is the best one. Default: on.
	<tag>default bgp_med <m/number/</tag> Value of the Multiple Exit
	Discriminator to be used during route selection when the MED attribute
	is missing. Default: infinite.
	<tag>default bgp_local_pref <m/number/</tag> Value of the Local Preference
	to be used during route selection when the Local Preference attribute
	is missing. Default: 0.
</descrip>

<sect1>Attributes

<p>BGP defines several route attributes. Some of them (those marked with `I' in the
table below) are available on internal BGP connections only, some of them (marked
with `O') are optional.

<descrip>
	<tag>bgppath <cf/bgp_path/</tag> Sequence of AS numbers describing the AS path
	the packet will travel through when forwarded according to this route. On
	internal BGP connections it doesn't contain the number of the local AS.
	<tag>int <cf/bgp_local_pref/ [I]</tag> Local preference value used for
	selection among multiple BGP routes (see the selection rules above). It's
	used as an additional metric which is propagated through the whole local AS.
	<tag>int <cf/bgp_med/ [IO]</tag> The Multiple Exit Discriminator of the route
	is an optional attribute which is often used within the local AS to
	reflect interior distances to various boundary routers. See the route selection
	rules above for exact semantics.
	<tag>enum <cf/bgp_origin/</tag> Origin of the route: either <cf/ORIGIN_IGP/,
	if the route has originated in interior routing protocol of an AS or
	<cf/ORIGIN_EGP/, if it's been imported from the <tt>EGP</tt> protocol
	(nowadays it seems to be obsolete) or <cf/ORIGIN_INCOMPLETE/, if the origin
	is unknown.
	<tag>ip <cf/bgp_next_hop/</tag> Next hop to be used for forwarding of packets
	to this destination. On internal BGP connections, it's an address of the
	originating router if it's inside the local AS or a boundary router the
	packet will leave the AS through if it's an exterior route, so each BGP
	speaker within the AS has a chance to use the shortest interior path
	possible to this point.
	<tag>void <cf/bgp_atomic_aggr/ [O]</tag> This is an optional attribute
	which carries no value, but which sole presence indicates that the route
	has been aggregated from multiple routes by some AS on the path from
	the originator.
<!-- we don't handle aggregators right since they are of a very obscure type
	<tag>bgp_aggregator</tag>
-->
	<tag>clist <cf/bgp_community/ [O]</tag> List of community values associated
	with the route. Each such value is a pair (represented as a <cf/pair/ data
	type inside the filters) of 16-bit integers, the first of them containing a number of the AS which defines
	the community and the second one is a per-AS identifier. There are lots
	of uses of the community mechanism, but generally they are used to carry
	policy information like "don't export to USA peers". As each AS can define
	its own routing policy, it also has a complete freedom about which community
	attributes it defines and what their semantics will be.
</descrip>

<sect1>Example

<p><code>
protocol bgp {
	local as 65000;				# Use a private AS number
	neighbor 62.168.0.130 as 5588;		# Our neighbor
	multihop 20 via 62.168.0.13;		# Which is connected indirectly
	export filter {				# We use non-trivial export rules
		if source = RTS_STATIC then {	# Export only static routes
		        # Assign our community
			bgp_community.add((65000,5678));
			# Artificially increase path length
			# by prepending local AS number twice
			if bgp_path ~ / 65000 / then	  
				bgp_path.prepend(65000);  
			accept;
		}
		reject;
	};
	import all;
	source address 62.168.0.1;		# Use non-standard source address
}
</code>

<sect>Device

<p>The Device protocol is not a real routing protocol as it doesn't generate
any routes and only serves as a module for getting information about network
interfaces from the kernel.

<p>Except for very unusual circumstances, you probably should include
this protocol in the configuration since almost all other protocol
require network interfaces to be defined in order to work.

<p>The only configurable thing is interface scan time:

<p><descrip>
	<tag>scan time <m/number/</tag> Time in seconds between two scans
	of the network interface list. On systems where we are notified about
	interface status changes asynchronously (such as newer versions of
	Linux), we need to scan the list only to avoid confusion by lost
	notifications, so the default time is set to a large value.
</descrip>

<p>As the Device protocol doesn't generate any routes, it cannot have
any attributes. Example configuration looks really simple:

<p><code>
protocol device {
	scan time 10;		# Scan the interfaces often
}
</code>

<sect>Direct

<p>The Direct protocol is a simple generator of device routes for all the
directly connected networks according to the list of interfaces provided
by the kernel via the Device protocol.

<p>It's highly recommended to include this protocol in your configuration
unless you want to use BIRD as a route server or a route reflector, that is
on a machine which doesn't forward packets and only participates in
distribution of routing information.

<p>Only configurable thing about direct is what interfaces it watches:

<p><descrip>
	<tag>interface <m/pattern [, ...]/</tag> By default, the Direct
	protocol will generate device routes for all the interfaces
	available. If you want to restrict it to some subset of interfaces
	(for example if you're using multiple routing tables for policy
	routing and some of the policy domains don't contain all interfaces),
	just use this clause.
</descrip>

<p>Direct device routes don't contain any specific attributes.

<p>Example config might look like this:

<p><code>
protocol direct {
	interface "-arc*", "*";		# Exclude the ARCnets
}
</code>

<sect>Kernel

<p>The Kernel protocol is not a real routing protocol. Instead of communicating
with other routers in the network, it performs synchronization of BIRD's routing
tables with OS kernel. Basically, it sends all routing table updates to the kernel
and from time to time it scans the kernel tables to see whether some routes have
disappeared (for example due to unnoticed up/down transition of an interface)
or whether an `alien' route has been added by someone else (depending on the
<cf/learn/ switch, such routes are either deleted or we accept them to our
table).

<p>If your OS supports only a single routing table, you can configure only one
instance of the Kernel protocol. If it supports multiple tables (in order to
allow policy routing), you can run as many instances as you want, but each of
them must be connected to a different BIRD routing table and to a different
kernel table.

<sect1>Configuration

<p><descrip>
	<tag>persist <m/switch/</tag> Tell BIRD to leave all its routes in the
	routing tables when it exits (instead of cleaning them up).
	<tag>scan time <m/number/</tag> Time in seconds between two scans of the
	kernel routing table.
	<tag>learn <m/switch/</tag> Enable learning of routes added to the kernel
	routing tables by other routing daemons or by the system administrator.
	This is possible only on systems which support identification of route
	authorship.
	<tag>kernel table <m/number/</tag> Select which kernel table should
	this particular instance of the Kernel protocol work with. Available
	only on systems supporting multiple routing tables.
</descrip>

<p>A simple configuration can look this way:

<p><code>
protocol kernel {
	import all;
	export all;
}
</code>

<p>Or for a system with two routing tables:

<p><code>
protocol kernel {		# Primary routing table
	learn;			# Learn alien routes from the kernel
	persist;		# Don't remove routes on bird shutdown
	scan time 10;		# Scan kernel routing table every 10 seconds
	import all;
	export all;
}

protocol kernel {		# Secondary routing table
	table auxtable;
	kernel table 100;
	export all;
}
</code>

<p>The Kernel protocol doesn't define any route attributes.

<sect>OSPF

<sect>Pipe

<sect1>Introduction

<p>The Pipe protocol serves as a link between two routing tables, allowing routes to be
passed from a table declared as primary (i.e., the one the pipe is connected using the
<cf/table/ configuration keyword) to the secondary one (declared using <cf/peer table/)
and vice versa, depending on what's allowed by the filters. Export filters control export
of routes from the primary table to the secondary one, import filters control the opposite
direction.

<p>The primary use of multiple routing tables and the pipe protocol is for policy routing,
where handling of a single packet doesn't depend only on its destination address, but also
on its source address, source interface, protocol type and other similar parameters.
In many systems (Linux 2.2 being a good example) the kernel allows to enforce routing policies
by defining routing rules which choose one of several routing tables to be used for a packet
according to its parameters. Setting of these rules is outside the scope of BIRD's work
(you can use the <tt/ip/ command), but you can create several routing tables in BIRD,
connect them to the kernel ones, use filters to control which routes appear in which tables
and also you can employ the Pipe protocol to export a selected subset of one table in
another one.

<sect1>Configuration

<p><descrip>
	<tag>peer table <m/table/</tag> Define secondary routing table to connect to. The
	primary one is selected by the <cf/table/ keyword.
</descrip>

<sect1>Attributes

<p>The Pipe protocol doesn't define any route attributes.

<sect1>Example

<p>Let's consider a router which serves as a boundary router of two different autonomous
systems, each of them connected to a subset of interfaces of the router, having its own
exterior connectivity and wishing to use the other AS as a backup connectivity in case
of outage of its own exterior line.

<p>Probably the simplest solution to this situation is to use two routing tables (we'll
call them <cf/as1/ and <cf/as2/) and set up kernel routing rules, so that packets having
arrived from interfaces belonging to the first AS will be routed according to <cf/as1/
and similarly for the second AS. Thus we have split our router to two logical routers,
each one acting on its own routing table, having its own routing protocols on its own
interfaces. In order to use the other AS's routes for backup purposes, we can pass
the routes between the tables through a Pipe protocol while decreasing their preferences
and correcting their BGP paths to reflect AS boundary crossing.

<code>
table as1;				# Define the tables
table as2;

protocol kernel kern1 {			# Synchronize them with the kernel
	table as1;
	kernel table 1;
}

protocol kernel kern2 {
	table as2;
	kernel table 2;
}

protocol bgp bgp1 {			# The outside connections
	table as1;
	local as 1;
	neighbor 192.168.0.1 as 1001;
	export all;
	import all;
}

protocol bgp bgp2 {
	table as2;
	local as 2;
	neighbor 10.0.0.1 as 1002;
	export all;
	import all;
}

protocol pipe {				# The Pipe
	table as1;
	peer table as2;
	export filter {
		if net ~ [ 1.0.0.0/8+] then {	# Only AS1 networks
			if preference>10 then preference = preference-10;
			if source=RTS_BGP then bgp_path.prepend(1);
			accept;
		}
		reject;
	};
	import filter {
		if net ~ [ 2.0.0.0/8+] then {	# Only AS2 networks
			if preference>10 then preference = preference-10;
			if source=RTS_BGP then bgp_path.prepend(2);
			accept;
		}
		reject;
	};
}
</code>

<sect>RIP

<sect1>Introduction

<p>The RIP protocol (also sometimes called Rest In Pieces) is a simple protocol, where each router broadcasts (to all its neighbors)
distances to all networks it can reach. When a router hears distance to another network, it increments
it and broadcasts it back. Broadcasts are done in regular intervals. Therefore, if some network goes
unreachable, routers keep telling each other that its distance is the original distance plus 1 (actually, plus
interface metric, which is usually one). After some time, the distance reaches infinity (that's 15 in
RIP) and all routers know that network is unreachable. RIP tries to minimize situations where
counting to infinity is necessary, because it is slow. Due to infinity being 16, you can't use
RIP on networks where maximal distance is higher than 15 hosts. You can read more about rip at <HTMLURL
URL="http://www.ietf.org/html.charters/rip-charter.html" TEXT="http://www.ietf.org/html.charters/rip-charter.html">. Both IPv4
and IPv6 versions of RIP are supported by BIRD, historical RIPv1 is
currently not fully supported.

<p>RIP is a very simple protocol, and it has a lot of shortcomings. Slow
convergence, big network load and inability to handle larger networks
makes it pretty much obsolete in IPv4 world. (It is still usable on
very small networks.) It is widely used in IPv6 networks,
because there are no good implementations of OSPFv3.

<sect1>Configuration

<p>In addition to options common for all to other protocols, RIP supports the following ones:

<descrip>
	<tag/authentication none|password|md5/ selects authentication method to be used. <cf/none/ means that
	  packets are not authenticated at all, <cf/password/ means that a plaintext password is embedded
	  into each packet, and <cf/md5/ means that packets are authenticated using a md5 cryptographic
	  hash. If you set authentication to not-none, it is a good idea to add <cf>passwords { }</cf>
	  section.

	<tag>honor always|neighbor|never </tag>specifies when should requests for dumping routing table
	  be honored. (Always, when sent from a  host on a directly connected
	  network or never.) Routing table updates are honored only from
	  neighbors, that is not configurable.
</descrip>

<p>There are two options that can be specified per-interface. First is <cf>metric</cf>, with
default one.  Second is <cf>mode multicast|broadcast|quiet|nolisten|version1</cf>, it selects mode for
rip to work in. If nothing is specified, rip runs in multicast mode. <cf>version1</cf> is
currently equivalent to <cf>broadcast</cf>, and it makes RIP talk to a broadcast address even
through multicast mode is possible. <cf>quiet</cf> option means that RIP will not transmit
any periodic messages to this interface and <cf>nolisten</cf> means that RIP will send to this
interface but not listen to it.

<p>The following options generally override behavior specified in RFC. If you use any of these
options, BIRD will no longer be RFC-compliant, which means it will not be able to talk to anything
other than equally configured BIRD. I have warned you.

<descrip>
	<tag>port <M>number</M></tag>
	  selects IP port to operate on, default 520. (This is useful when testing BIRD, if you
	  set this to an address &gt;1024, you will not need to run bird with UID==0).

	<tag>infinity <M>number</M></tag>
	  selects the value of infinity, default is 16. Bigger values will make protocol convergence
	  even slower.

	<tag>period <M>number</M>
	  </tag>specifies the number of seconds between periodic updates. Default is 30 seconds. A lower
	  number will mean faster convergence but bigger network
	  load. Do not use values lower than 10.

	<tag>timeout time <M>number</M>
	  </tag>specifies how old route has to be to be considered unreachable. Default is 4*<cf/period/.

	<tag>garbage time <M>number</M>
	  </tag>specifies how old route has to be to be discarded. Default is 10*<cf/period/.
</descrip>

<sect1>Attributes

<p>RIP defines two route attributes:

<descrip>
	<tag>int <cf/rip_metric/</tag> RIP metric of the route (ranging from 0 to <cf/infinity/).
	When routes from different RIP instances are available and all of them have the same
	preference, BIRD prefers the route with lowest <cf/rip_metric/.

	<tag>int <cf/rip_tag/</tag> RIP route tag: a 16-bit number which can be used
	to carry additional information with the route (for example, an originating AS number
	in case of external routes).
</descrip>

<sect1>Example

<p><code>
protocol rip MyRIP_test {
        debug all;
        port 1520;
        period 10;
        garbage time 60;
        interface "eth0" { metric 3; mode multicast; }
	          "eth1" { metric 2; mode broadcast; };
        honor neighbor;
        authentication none;
        import filter { print "importing"; accept; };
        export filter { print "exporting"; accept; };
}
</code>

<sect>Static

<p>The Static protocol doesn't communicate with other routers in the network,
but instead it allows you to define routes manually. This is often used for
specifying how to forward packets to parts of the network which don't use
dynamic routing at all and also for defining sink routes (i.e., those
telling to return packets as undeliverable if they are in your IP block,
you don't have any specific destination for them and you don't want to send
them out through the default route to prevent routing loops).

<p>There are three types of static routes: `classical' routes telling to
forward packets to a neighboring router, device routes specifying forwarding
to hosts on a directly connected network and special routes (sink, blackhole
etc.) which specify a special action to be done instead of forwarding the
packet.

<p>When the particular destination is not available (the interface is down or
the next hop of the route is not a neighbor at the moment), Static just
uninstalls the route from the table it is connected to and adds it again as soon
as the destination becomes adjacent again.

<p>The Static protocol has no configuration options. Instead, the
definition of the protocol contains a list of static routes:

<descrip>
	<tag>route <m/prefix/ via <m/ip/</tag> Static route through
	a neighboring router.
	<tag>route <m/prefix/ via <m/"interface"/</tag> Static device
	route through an interface to hosts on a directly connected network.
	<tag>route <m/prefix/ drop|reject|prohibit</tag> Special routes
	specifying to drop the packet, return it as unreachable or return
	it as administratively prohibited.
</descrip>

<p>Static routes have no specific attributes.

<p>Example static config might look like this:

<p><code>
protocol static {
	table testable;				# Connect to a non-default routing table
	route 0.0.0.0/0 via 62.168.0.13;	# Default route
	route 62.168.0.0/25 reject;		# Sink route
	route 10.2.0.0/24 via "arc0";		# Secondary network
}
</code>

<chapt>Problems

<p>BIRD is relatively young system, and probably contains some
bugs. You can report bugs at bird-users mailing list (<HTMLURL URL="mailto:bird-users@bird.network.cz" TEXT="bird-users@bird.network.cz">), but before you do,
please make sure you have read available documentation, make sure are running latest version (available at <HTMLURL
URL="ftp://bird.network.cz/pub/bird" TEXT="bird.network.cz:/pub/bird>). (Of course, patch
which fixes the bug along with bug report is always welcome). If you
want to use BIRD, join mailing list by sending
<tt/subscribe bird-users/ to <HTMLURL URL="mailto:majordomo@bird.network.cz" TEXT="majordomo@bird.network.cz">. Main home page of bird is <HTMLURL URL="http://bird.network.cz/" TEXT="http://bird.network.cz/">. When
trying to understand, what is going on, Internet standards are
relevant reading; you can get them from <HTMLURL URL="ftp://ftp.rfc-editor.org/" TEXT="ftp.rfc-editor.org"> (or nicely sorted version from <HTMLURL="ftp://atrey.karlin.mff.cuni.cz/pub/rfc" TEXT="atrey.karlin.mff.cuni.cz:/pub/rfc">.

<p><it/Good luck!/

</book>

<!--
LocalWords:  GPL IPv GateD BGPv RIPv OSPFv Linux sgml html dvi sgmltools Pavel
LocalWords:  linuxdoc dtd descrip config conf syslog stderr auth ospf bgp Mbps
LocalWords:  router's eval expr num birdc ctl unix if's enums bool int ip GCC
LocalWords:  len ipaddress pxlen netmask enum bgppath bgpmask clist gw md eth
LocalWords:  RTS printn quitbird iBGP AS'es eBGP RFC multiprotocol IGP Machek
LocalWords:  EGP misconfigurations keepalive pref aggr aggregator BIRD's
LocalWords:  OS'es AS's multicast nolisten misconfigured UID blackhole
LocalWords:  uninstalls ethernets IP
 -->