Dibbler - A Portable Dhcpv6 User'S Guide: Tomek Mrugalski

Dibbler – a portable DHCPv6
User’s guide
Tomek Mrugalski
thomson(at)klub.com.pl
2015-08-09
1.0.1
Dibbler 1.0.1 User’s Guide 2
Contents
1 Intro 6
1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Supported parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Not supported features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Operating System Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Supported platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Installation and usage 11

2.1 Linux installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Windows installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Mac OS X installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 FreeBSD, NetBSD, OpenBSD, Solaris 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Basic usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Compilation 13
3.1 Linux/Mac OS X/FreeBSD/NetBSD/OpenBSD/Solaris Compilation . . . . . . . . . . . . 13
3.2 Modern Windows (XP...Win7) compilation . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Legacy Windows (NT/2000) compilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 IPv6 support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.1 Setting up IPv6 in Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.2 Setting up IPv6 in Windows Vista and Win7 . . . . . . . . . . . . . . . . . . . . . 14
3.4.3 Setting up IPv6 in Windows XP and 2003 . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.4 Setting up IPv6 in Windows 2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4.5 Setting up IPv6 in Windows NT4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Features HOWTO 17
4.1 Prefix delegation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Relays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3 Address and prefix assignment policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.4 Routing configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.5 Custom options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.6 DNS Update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.6.1 Example BIND configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.6.2 Secure DDNS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.6.3 Dynamic DNS Testing and tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.6.4 Accepting Unknown FQDNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.7 Introduction to client classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.7.1 Client class declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.7.2 Access control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.7.3 Assigning clients to defined classes . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.7.4 Examples of Client-Class Classifying . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.8 External script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.9 Reconfiguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.10 Following M, O bits from Router Advertisements . . . . . . . . . . . . . . . . . . . . . . . 34
4.11 CONFIRM message . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.12 Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.13 Leasequery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.14 Stateless vs stateful and IA, TA options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.15 Server address caching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.16 XML files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.17 Authentication and Authorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.17.1 Replay Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.17.2 Reconfigure Key Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.17.3 Delayed Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.17.4 Dibbler Authentication Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.17.5 Key generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.18 Exceptions: per client configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.19 Vendor specific information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.20 Not connected interfaces (inactive-mode) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.21 Parameters not supported by server (insist-mode) . . . . . . . . . . . . . . . . . . . . . . . 45
4.22 Different DUID types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.23 Debugging/compatibility features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.23.1 Interface-id option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.23.2 Non-empty IA NA option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.23.3 Providing address/prefix hints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.24 Experimental features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.24.1 Server Performance mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.24.2 Address Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.24.3 Remote Autoconfiguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.25 Obsoleted experimental features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.25.1 Mapping prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.25.2 Tunnel mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5 Server configuration 52
5.1 Scopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.1 Global scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.2 Interface declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.3 Address class scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.4 Prefix class scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.5 Temporary address class scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.6 Routing scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.7 Client scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.1.8 Key scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Server options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.1 Client class quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.3 Server configuration examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.3.1 Example 1: Simple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.3.2 Example 2: Timeouts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.3.3 Example 3: Limiting amount of addresses . . . . . . . . . . . . . . . . . . . . . . . 62
5.3.4 Example 4: Unicast communication . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3.5 Example 5: Rapid-commit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3.6 Example 6: Access control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3.7 Example 7: Multiple classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.8 Example 8: Relay support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3.9 Example 9: Cascade 2 relays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.3.10 Example 10: Dynamic DNS (FQDN) . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.3.11 Example 11: Vendor-specific Information option . . . . . . . . . . . . . . . . . . . 69
5.3.12 Example 12: Per client configuration . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3.13 Example 13: Prefix delegation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3.14 Example 14: Multiple prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

5.3.15 Example 15: Inactive mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3.16 Example 16: Leasequery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3.17 Example 17: Dibbler Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3.18 Example 18: Relay support with unknown interface-id . . . . . . . . . . . . . . . . 74
5.3.19 Example 19: DS-Lite tunnel (AFTR) . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.3.20 Example 20: Custom options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.3.21 Example 21: Remote Autoconfiguration . . . . . . . . . . . . . . . . . . . . . . . . 75
5.3.22 Example 21: Subnet declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6 Client configuration 76
6.1 Data types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.2 Scopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.2.1 Interface declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.2.2 IA declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.2.3 TA declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.2.4 PD declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.2.5 Address declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.2.6 Prefix declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.3 Stateless configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.4 Relay support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.5 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.6 File location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.7 Client Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.8 Client Configuration Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.8.1 Example 1: Default . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.8.2 Example 2: DNS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.8.3 Example 3: Timeouts and specific address . . . . . . . . . . . . . . . . . . . . . . . 86
6.8.4 Example 4: More than one address . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.8.5 Example 5: Quick configuration using Rapid-commit . . . . . . . . . . . . . . . . . 88
6.8.6 Example 6: Stateless mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.8.7 Example 7: Dynamic DNS (FQDN) . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.8.8 Example 8: Interface indexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.8.9 Example 9: Vendor-specific options . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.8.10 Example 10: Unicast communication . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.8.11 Example 11: Prefix delegation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.8.12 Example 12: Insist mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.8.13 Example 13: Inactive mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.8.14 Example 14: Dibbler Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.8.15 Example 15: Skip Confirm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.8.16 Example 15: User-defined IAID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.8.17 Example 16: DS-Lite tunnel (AFTR) . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.8.18 Example 17: Custom options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.8.19 Example 18: Remote Autoconfiguration . . . . . . . . . . . . . . . . . . . . . . . . 95
7 Relay configuration 96
7.1 Global scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2 Interface declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.3 Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.4 Relay configuration examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.4.1 Example 1: Simple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

7.4.2 Example 2: Unicast/multicast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.4.3 Example 3: Multiple interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.4.4 Example 4: 2 relays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
7.4.5 Example 5: Guess-mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7.4.6 Example 6: Relaying to multicast . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7.4.7 Example 7: Options inserted by the relay . . . . . . . . . . . . . . . . . . . . . . . 101
8 Requestor configuration 103
9 Frequently Asked Questions 104

9.1 Common Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
9.2 Linux specific questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
9.3 Windows specific questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
10 Miscellaneous topics 108

10.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
10.2 Contact and reporting bugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
10.3 Mailing lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
10.4 Thanks and greetings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
11 Acknowledgements 111
Bibliography 113
1 Intro
First of all, as an author I would like to thank you for your interest in this DHCPv6 implementation. If
this documentation doesn’t answer your questions or you have any suggestions, feel free to contact me as
explained in Contact section. Also be sure to check out Dibbler website: http://klub.com.pl/dhcpv6/.
Tomek Mrugalski
1.1 Overview
Dynamic Host Configuration Protocol for IPv6, often abbreviated as DHCPv6, is a protocol, which is
used to automatically configure IPv6 capable computers and other equipment located in a local network.
This protocol defines clients (i.e. nodes, which want to be configured), servers (i.e. nodes, which provide
configuration to clients) and relays (i.e. nodes, which are connected to more than one network and are
able to forward traffic between local clients and remote servers). Also, special type of DHCPv6 entity
called requestor has been defined. It is used by network administrator to query servers about their status
and assigned parameters.
Dibbler is a portable DHCPv6 solution, which features server, client and relay. Currently there are
ports available for many Windows platforms ranging from NT4 to Windows 8, Linux 2.4 or later systems
and Mac OS (experimental). See Section 1.4 for details. It supports both stateful (i.e. IPv6 address
granting) and stateless (i.e. options granting) autoconfiguration. Besides basic functionality (specified in
basic DHCPv6 spec, RFC3315 [5]), it also offers serveral enhancements, e.g. DNS servers and domain
names configuration.
Dibbler is an open source software, distributed under GNU GPL v2 licence. It means that it is freely
available, free of charge and can be used by anyone (including commercial users). Source code is also
provided, so anyone skilled enough can fix bugs, add new features and distribute his/her own version.
Requestor support has been added in version 0.7.0RC1. Requestor is a separate entity, which sends
queries to the server regarding leases to specific clients. It is possible to ask a server, who has specific
address or what addresses are assigned to a specific client. This feature is part of the lease query mechanism
defined in [21] and is considered advanced topic. If you don’t know what lease query is, you definetely
don’t need it.
Figure 1: General DHCPv6 operation
Dibbler 1.0.0RC1 supports all features specified in RFC3315. In particular the following features are
supported:
• Basic server discovery and address assignment (SOLICIT, ADVERTISE, REQUEST and REPLY
messages) – This is a most common case: client discovers servers available in the local network,
then asks for an address (and possibly additional options like DNS configuration), which is granted
by a server.
Figure 2: Several clients supported by one server
• Server redundancy/Best server discovery – when client detects more than one server available (by
receiving more than one ADVERTISE message), it chooses the best one and remembers remaining
ones as a backup.
• Multiple servers support – Client is capable of discovering and maintaning communication with
several servers. For example, client would like to have 5 addresses configured. Prefered server can
only lease 3, so client send request for remaining 2 addresses to one of the remaining servers.
• Relay support – In a larger network, which contains several Ethernet segments and/or wireless
areas, sometimes centrally located DHCPv6 server might not be directly reachable. In such cace,
additional proxies, so called relays, might be deployed to relay communication between clients and
a remote server. Dibbler server supports indirect communication with clients via relays. Stand-
alone, lightweight relay implementation is also available. Clients are capable of talking to the server
directly or via relays.
• Address renewal – After receiving address from a server, client might be instructed to renew its
address at regular intervals. Client periodically sends RENEW messege to a server, which granted
its address. In case of communication failure, client is also able to attempt emergency address
renewal (i.e. it sends REBIND message to any server).
• Unicast communication – if specific conditions are met, client could send messages directly to a
server’s unicast address, so additional servers does not need to process those messages. It also
improves effciency, as all nodes present in LAN segment receive multicast packets.1
• Duplicate address detection – Client is able to detect and properly handle faulty situation, when
server grants an address which is illegaly used by some other host. It will inform server of such
circumstances (using DECLINE message), and request another address. Server will mark this
address as used by unknown host, and will assign another address to a client.
1
Nodes, which do not belong to specific multicast group, drop those packets silently. However, determining if host belongs
or not to a group must be performed on each node. Also using multicast communication increases the network load.
Figure 3: Redundancy: several servers
• Power failure/crash support – After client recovers from a crash or a power failure, it still can have
valid addresses assigned. In such circumstances, client uses CONFIRM message, to config if those
addresses are still valid.
• Link change detection – Client can be instructed to monitor its link state. Once it detects
• Normal and temporary addresses – Depending on its purpose, client can be configured to ask for
normal (IA NA option) or temporary (IA TA option). Although use of temporary addresses is
rather uncommon, both dibbler server and client support it.
• Hint system – Client can be configured to send various parameters and addresses in the REQUEST
message. It will be treated as a hint by the server. If such hint is valid, it will be granted for this
client.
• Server caching – Server can cache granted addresses, so the same client will receive the same address
each time it asks. Size of this cache can be configured.
• Stateless mode – Client can be configured to not ask for any addresses, but the configuration
options only. In such case, when no addresses are granted, such configuration is called stateless
(INFORMATION-REQUEST message is used instead of normal REQUEST ).
• Rapid Commit – Sometimes it is desirable to quicken configuration process. If both client and server
are configured to use rapid commit, address assignment procedure can be shortened to 2 messages,
instead of usual 4. Major advantage is lesser network usage and quicker client startup time.
• M,O bits from Router Advertisement – the client can be told to observe M(managed) and O(OtherConf)
bits from RA and act according to them
• Reconfigure – server can inform clients that the configuration has changed and clients can initiate
Reconfigure
• Authentication: Reconfigure-key – the server can generate HMAC-MD5 reconfigure keys on the
fly to later authenticate reconfigure messages. Clients are able to receive, store and later validate
against that received key.
• Authentication: Delayed authorization – server and client can protect their communication against
tampering by using preprovisioned keys.
1.2 Supported parameters

Except RFC3315-specified behavior [5], Dibbler also supports several enhancements:
• DNS Servers – During normal operation, almost all hosts require constant use of the DNS servers.
It is necessary for event basic operations, like web surfing. DHCPv6 client can ask for information
about DNS servers and DHCPv6 server will provide necessary information. [9]
• Domain Name – Client might be interested in obtaining information about its domain. Properly
configured domain allow reference to a different hosts in the same domain using hostname only, not
the full domain name, e.g. alice.example.com with properly configured domain can refer to another
host in the same domain by using ’bob’ only, instead of full name bob.example.com. [9]
• NTP Servers – To prevent clock misconfiguration and drift, NTP protocol [1] can be used to syn-
chronize clocks. However, to successful use it, location of near NTP servers must be known. Dibbler
is able to configure this information. [14]
• Time Zone – To avoid time-related ambiguation, each host should have timezone set properly.
Dibbler is able to pass this parameter to all clients, who request it. [32]
• SIP Servers – Session Initiation Protocol (SIP) [4] is commonly used in VoIP solutions. One of the
necessary information is SIP server addresses. This information can be passed to the clients. [6]
• SIP Domain Name – SIP domain name is another important parameter of the VoIP capable nodes.
This parameter can be passed to all clients, who ask for it. [6]
• NIS, NIS+ Server – Network Information Service is a protocol for sharing authentication parameters
between multiple Unix or Linux nodes. Both NIS and NIS+ server addresses can be passed to the
clients. [11]
• NIS, NIS+ Domain Name – NIS or NIS+ domain name is another necessary parameter for NIS or
NIS+. It can be obtained from the DHCPv6 server to all clients, who require it. [11]
• Option Renewal Mechanism (Lifetime option)– All of the options mentioned on this list can be
refreshed periodically. This might be handy if one of those parameters change. [13]
• Dynamic DNS Updates – Server can assign a fully qualified domain name for a client. To make such
name useful, DNS servers must be informed that such name is bound to a specific IPv6 address.
This procedure is called DNS Update. There are two kinds of the DNS Updates: forward and
reverse. First is used to translate domain name to an address. The second one is used to obtain full
domain name of a known address. See section 4.6 for details. [16]
• Prefix Delegation – Server can be configured to manage a prefix pool, i.e. clients will be assigned
whole pools instead on single addresses. This is very useful, when clients are not simple end users
(e.g. desktop computers or laptops), but rather are routers (e.g. cable modems). This functionality
is often used for remote configuration of IPv6 routers. [8]
1.3 Not supported features

Although list of the supported features increases with each release, there are certain limitations. Below
is a list of such features:
• DNS Updates are done over IPv6 only. Adding IPv4 support is not planned. Do not bother to
develop patches – Dibbler is a IPv6-focused software and IPv4-related patches will be rejected.
• Conflict resolution in DNS Updates is not supported.
1.4 Operating System Requirements

Dibbler can be run on Linux systems with kernels from 2.4 or later series. IPv6 (compiled into kernel or
as module) support is necessary to run dibbler. DHCPv6 uses UDP ports below 1024, so root privileges
are required. They’re also required to add, modify and delete various system parameters, e.g. IPv6
addresses.
Dibbler also runs on any Windows systems from Windows XP (Service Pack 1 or later) to Windows 8.
Support for Windows 8 has been added in 0.8.3. To install various Dibbler parts (server, client or relay)
as services, administrator privileges might be required. Support for Windows NT4 and 2000 is limited
and considered experimental. Due to lack of support and any kind of informations from Microsoft, this
will not change. In fact, support for NT4 and 2000 is expected to be dropped soon. Please post to Dibbler
mailing list if you need them.
There is working Mac OS X port available.
Support for FreeBSD, NetBSD and OpenBSD was added in 0.8.1RC1, but those versions are not very
well tested. Support for Solaris 11 has been added in 0.8.3, but it is still highly experimental. Sources
are confirmed to compile and be able to start operation. Author was not able to test them thoroughly,
so reports regarding confirming their stability or any discovered issues are welcome. Please report them
on the mailing list. See section 10.3.
See RELEASE-NOTES for details about version-specific upgrades, fixes and features.
1.5 Supported platforms

Although Dibbler was developed on the i386 architecture, there are ports available for other architec-
tures: IA64, AMD64, PowerPC, HPPA, Sparc, MIPS, S/390, Alpha and ARMv5. They are available in
the PLD, Gentoo and Debian Linux distributions. Other platforms are likely to be supported. Keep in
mind that author has not tested those ports himself and need to rely on users’ reports, so there might be
some unknown issues present. If this is the case, be sure to notify package maintainers and possibly the
author.
If your system is not on the list, don’t despair. Dibbler is fully portable. Core logic is system
independent and coded in C++ language. There are also several low-level functions, which are system
specific. They’re used for adding addresses, retrieving information about interfaces, setting DNS servers
and so on. Porting Dibbler to other systems (and even other architectures) would require implementic
only those serveral system-specific functions. See Developer’s Guide for details.
2 Installation and usage

Client, server and relay are installed in the same way. Installation method is different in Windows and
Linux systems, so each system installation is described separately. To simplify installation, it assumes
that binary versions are used2 .
2.1 Linux installation

Starting with 0.4.0, Dibbler consists of 3 different elements: client, server and relay. During writing
this documentation, Dibbler is already part of many Linux distributions. In particular:
Debian GNU/Linux, Ubuntu and derived – use standard tools (apt-get, aptitude and similar) to
install dibbler-client, dibbler-server, dibbler-relay or dibbler-doc packages.
OpenSUSE – use standard installation mechanism.
PLD GNU/Linux – use standard PLD’s poldek tool to install dibbler package.
Gentoo Linux – use emerge to install dibbler (e.g. emerge dibbler).
OpenWRT – there are package definitions for OpenWRT. At time of this writing, they were very
outdated (using 0.5.0 version).
If you are using other Linux distribution, check out if it already provides Dibbler packages. You may
use them or compile the sources on your own. See Section 3 for details regarding compilation process.
Dibbler used to provide native DEB and RPM packages, but due to limited resources, author is not
continuing this activity. If you are a Dibbler package maintainer and want your package to be put on
dibbler website, please send such request on mailing list (see Section 10.3).
To install Dibbler on Debian or other system that provides apt-get package management system,
run apt-get install package command. For example, to install server and client, issue the following
command:
apt-get install dibbler-server dibbler-client
To install Dibbler in Gentoo systems, just type:

emerge dibbler
2.2 Windows installation

Dibbler supports Windows XP and 2003 since the 0.2.1-RC1 release. Support for Vista was added
somewhere around 0.7.x. Support for Windows 7 was added in 0.8.0RC1. In version 0.4.1 exprimental
support for Windows NT4 and 2000 was added. The easiest way of Windows installation is to download
clickable Windows installer. It can be downloaded from http://klub.com.pl/dhcpv6/. After download-
ing, click on it and follow on screen instructions. Dibbler will be installed and all required links will be
placed in the Start menu. Note that there are two Windows versions (ports): one for modern systems
(XP/2003/Vista and Win7) and one for archaic ones (NT4/2000). Make sure to use proper port. If you
haven’t set up IPv6 support, see following sections for details.
Operation on Windows 8 was never tested, so support is not confirmed.
2
Compilation is not required, usually binary version can be used. Compilation should be performed by advanced users
only, see Section 3 for details.
2.3 Mac OS X installation

As of 0.8.0 release, ready to use dmg packages are not provided, therefore dibbler has to be compiled.
Please follow section 3 for generic Dibbler compilation that applies to Mac OS X.
Currently support for Mac OS X is usable, but there is still one notable limitation. Client is not able
to configure DNS servers or domain name informations.
2.4 FreeBSD, NetBSD, OpenBSD, Solaris 11

As of 0.8.1RC1 release, support for FreeBSD, NetBSD and OpenBSD has been added. Solaris 11
support is implemented after 0.8.2 and will be included in 0.8.3. There are no prebuilt binary packages
available. Please follow Section 3 for generic Dibbler compilation that applies to all 3 mentioned OSes.
2.5 Basic usage

Depending what functionality do you want to use (server,client or relay), you should edit configuration
file (client.conf for client, server.conf for server and relay.conf for relay). All configuration files
should be placed in the /etc/dibbler directory. Also make sure that /var/lib/dibbler directory is
present and is writeable. After editing configuration files, issue one of the following commands:
dibbler-server start
dibbler-client start
dibbler-relay start
start parameter requires little explanation. It instructs Dibbler to run in daemon mode – detach
from console and run in the background. During configuration files fine-tuning, it is ofter better to watch
Dibbler’s bahavior instantly. In this case, use run instead of start parameter. Dibbler will present its
messages on your console instead of log files. To finish it, press ctrl-c.
To stop server, client or relay running in daemon mode, type:
dibbler-server stop
dibbler-client stop
dibbler-relay stop
To see, if client, server or relay are running, type:

dibbler-server status
dibbler-client status
dibbler-relay status
To see full list of available commands, type dibbler-server, dibbler-client or dibbler-relay

without any parameters.
If your OS uses different layout of directories, you may want to modify Misc/Portable.h before starting
compilation process.
3 Compilation
Dibbler is distributed in 2 versions: binary and source code. If there is binary version provided for
your system, it is usually better choice. Compilation usually is performed by more experienced users.
In average case it does not offer significant advantages over binary version. You probably want to just
install and use Dibbler. If that is your case, read section named 2. However, if you are skilled enough,
you might want to tune several Dibbler aspects during compilation. See Dibbler Developer’s Guide for
information about various compilation parameters.
3.1 Linux/Mac OS X/FreeBSD/NetBSD/OpenBSD/Solaris Compilation

The following descriptions applies to Linux, Mac OS X, FreeBSD, NetBSD and OpenBSD. Solaris 11
support has been added since 0.8.3. Other POSIX systems may work, but were never tested by author.
If you would like to install Dibbler from sources, you will need all required dependencies. In particular,
you need a typical C++ environment: a C and C++ compilers (most probably gcc and g++), make, and
several other smaller tools.
To install Dibbler package from sources, go to project homepage and download latest tar.gz source
archive. Extract it using available tool for that purpose (in most cases that would be tool called tar and
gzip).
After sources are extracted, they must be configured to match specific operating system. To complete
this step, a configure script must be called:
./configure
Configure script accepts many parameters, so if like to tweak something, here is your chance. You may
run ./configure --help to see list of available parameters. For example, to set up sources to compile
in debug mode (useful if you want to debug them or provide better bugreport), you can do this:
./configure --enable-debug
See Dibbler Developer’s Guide, section 2 for details on compilation switches.

Once configure completes its operation, it prints out details of its configuration and source are ready
for compilation. To build all components, just type make. If you want to make specific component only,
you may use it as parameter to make, e.g. make server. After successful compilation type make install
to install compiled code in your system.
For example, to build server and relay, type:
tar zxvf dibbler-0.8.1RC1-src.tar.gz
./configure
make server relay
make install
mkdir -p /var/lib/dibbler
Configure script was added in 0.8.1RC1. Earlier versions do not not need that step.
Dibbler was compiled using gcc 2.95, 3.0, 3.2, 3.3, 3.4, 4.0, 4.1, 4.2, 4.3, 4.4, 4.5, 4.6 and 4.7 versions.
Note that many older compilers are now considered obsolete and were not tested for some time. Lexer
files (grammar defined config file) were generated using flex 2.5.35. Parser file were created using bison++
1.21.9. Flex and bison++ tools are not required to compile Dibbler. Generated files are placed in GIT
and in tar.gz archives. Dibbler requires also make. Autoconf and automake tools (autotools) were used
for regeneration of the Makefiles and configure script, but those generated files are shipped with the code,
so autotools should not be required.
3.2 Modern Windows (XP...Win7) compilation

Download dibbler-1.0.1-src.tar.gz and extract it. In Port-win32 there are several project files (for
server, client and relay) for MS Visual Studio 2008. According to authors knowledge, it is possible to com-
pile dibbler using free MS Visutal C++ Express 2008 edition. Previous dibbler releases were compiled us-
ing MS Visual Studio .NET (sometimes called 2002) and 2003. Those versions are not supported anymore.
It might work with newest dibber version, but there are no guarantee. Open dibbler-win32.vs2008.sln
solution file click Build command. That should start compilation. After a while, binary exe files will be
stored in the Debug/ or Release/ directories.
3.3 Legacy Windows (NT/2000) compilation

Windows NT4/2000 port is considered experimental, but there are reports that it works just fine. To
compile it, you should download dev-cpp (http://www.bloodshed.net/dev/devcpp.html), a free IDE
for Windows utilising minGW port of the gcc for Windows. Run dev-cpp, click ,,open project...”, and
open one of the *.dev files located in the Port-winnt2k directory, then click compile. You also should
take a look at Port-winnt2k/INFO file for details.
3.4 IPv6 support

Some systems does not have IPv6 enabled by default. In that is the case, you can skip following
subsections safely. If you are not sure, here is an easy way to check it. To verify if you have IPv6 support,
execute following command: ping6 ::1 (Linux) or ping ::1 (Windows). If you get replies, you have
IPv6 already installed.
3.4.1 Setting up IPv6 in Linux

Almost all modern Linux distributions have IPv6 enabled by default, so there is very good chance
that nothing has to be done. However, if that is not the case, IPv6 can be enabled in Linux systems
in two ways: compiled directly into kernel or as a module. If you don’t have IPv6 enabled, try to load
IPv6 module: modprobe ipv6 (command executed as root) and try ping6 ::1. If that fails, you have to
recompile kernel to support IPv6. There are numerous descriptions how to recompile kernel available on
the web, just type ”kernel compilation howto” in Google.
3.4.2 Setting up IPv6 in Windows Vista and Win7

Both systems have IPv6 enabled by default. Also note that Win7 also has DHCPv6 client built-in, so
you may use it as well.
3.4.3 Setting up IPv6 in Windows XP and 2003

If you have already working IPv6 support, you can safely skip this section. The easiest way to enable
IPv6 support is to right click on the My network place on the desktop, select Properties, then locate
your network interface, right click it and select Properties. Then click Install..., choose protocol
and then IPv6 (its naming is somewhat diffrent depending on what Service Pack you have installed).
In XP, there’s much quicker way to install IPv6. Simply run command ipv6 install (i.e. hit Start...,
choose run... and then type ipv6 install). Also make sure that you have built-in firewall disabled. See
Frequently Asked Question section for details.
3.4.4 Setting up IPv6 in Windows 2000

If you have already working IPv6 support, you can safely skip this section. The following description
was provided by Sob ( (sob(at)hisoftware.cz). Thanks. This description assumes that ServicePack 4 is
already installed.
1. Download the file tpipv6-001205.exe from: http://msdn.microsoft.com/downloads/sdks/platform/

tpipv6.asp and save it to a local folder (for example, C:\IPv6TP).
2. From the local folder (C:\IPv6TP), run Tpipv6-001205.exe and extract the files to the same loca-
tion.
3. From the local folder (C:\IPv6TP), run Setup.exe -x and extract the files to a subfolder of the
current folder (for example, C:\IPv6TP\files).
4. From the folder containing the extracted files (C:\IPv6TP\files), open the file Hotfix.inf in a
text editor.
5. In the [Version] section of the Hotfix.inf file, change the line NTServicePackVersion=256 to NTSer-
vicePackVersion=1024, and then save changes. 3
6. From the folder containing the extracted files (C:\IPv6TP\files), run Hotfix.exe.
7. Restart the computer when prompted.
8. After the computer is restarted, from the Windows 2000 desktop, click Start, point to Settings, and
then click Network and Dial-up Connections. As an alternative, you can right-click My Network
Places, and then click Properties.
9. Right-click the Ethernet-based network interface to which you want to add the IPv6 protocol, and
then click Properties. Typically, this network interface is named Local Area Connection.
10. Click Install.
11. In the Select Network Component Type dialog box, click Protocol, and then click Add.
12. In the Select Network Protocol dialog box, click Microsoft IPv6 Protocol and then click OK.
13. Click Close to close the Local Area Connection Properties dialog box.
3.4.5 Setting up IPv6 in Windows NT4

If you have already working IPv6 support, you can safely skip this section. The following description
was provided by The following description was provided by Sob (sob(at)hisoftware.cz). Thanks.
1. Download the file msripv6-bin-1-4.exe from: http://research.microsoft.com/msripv6/msripv6.

htmMicrosoft and save it to a local folder (for example, C:\IPv6Kit).
2. From the local folder (C:\IPv6Kit), run msripv6-bin-1-4.exe and extract the files to the same
location.
3. Start the Control Panel’s ”Network” applet (an alternative way to do this is to right-click on
”Network Neighborhood” and select ”Properties”) and select the ”Protocols” tab.
3
This defines Service Pack requirement. NTServicePackVersion is a ServicePack version multiplied by 256. If there would
be SP5 available, this value should have been changed to the 1280.
4. Click the ”Add...” button and then ”Have Disk...”. When it asks you for a disk, give it the full
pathname to where you downloaded the binary distribution kit (C:\IPv6Kit).
5. IPv6 is now installed.

4 Features HOWTO
This section contains information about setting up various Dibbler features. Since this section was
added recently, it is not yet comprehensive. That is expected to change.
4.1 Prefix delegation

Prefix delegation is a mechanism that allows two routers to delegate (“assign”) prefixes in similar way
as server can delegate (“lease”) addresses to hosts. As specified in [8]: The prefix delegation mechanism is
intended for simple delegation of prefixes from a delegating router to requesting routers. It is appropriate
for situations in which the delegating router does not have knowledge about the topology of the networks to
which the requesting router is attached, and the delegating router does not require other information aside
from the identity of the requesting router to choose a prefix for delegation. For example, these options
would be used by a service provider to assign a prefix to a Customer Premise Equipment (CPE) device
acting as a router between the subscriber’s internal network and the service provider’s core network.
To configure server to provide prefixes, a PD pool and a client prefixes’ length must be defined. An
example section below assigns 2001:db8::/32 pool to be managed by this server. From this pool, server
will assign /48 prefixes to the clients. For example, client can receive prefix 2001:db8:7c34::/48.
pd-class {
pd-pool 2001:db8::/32
pd-length 48
}
As a general rule, server will provide random prefix to a client, unless client provided a hint. The full
prefix assignment algorithm is as follows:
1. client didn’t provide any hints: one prefix from each pool will be granted
2. client has provided hint and that is valid (supported and unused): requested prefix will be granted
3. client has provided hint, which belongs to supported pool, but this prefix is used: other prefix from
that pool will be asigned
4. client has provided hint, but it is invalid (not beloninging to a supported pool, multicast or link-
local): see point 1
Dibbler implementation supports prefix delegation, as specified in [8]. Up to and including 0.7.3
version, client was also capable to do non-standard tricks with delegated prefix if it was a host, rather than
router. This mode of operation was removed in 0.8.0RC1. Now client behaves the same way, regardless
if it is a host or a router. When client receives prefix on one interface (e.g. prefix 2000:1234:7c34::/48
received on eth0) it will generate subprefixes for all other interfaces, which are up, running, non-loopback
and multicast capable. In the example depicted on Fig. 4.1, received prefix was split into 3 prefixes:
2000:1234:7c34:1000::/56 for eth1, 2000:1234:7c34:2000::/56 for eth2 and 2000:1234:7c34:3000::/56 for
eth3. Client support for prefix delectation was improved in 0.8.2. Client is now able to handle prefixes of
arbitrary lengths (do not have to be divisible by 8 anymore). The only restriction is that prefix must be
shorter or equal 120 bits.
It is also possible to explicitly specify which interfaces are downlink (i.e. sub-prefixes should be
assigned to). downlink-prefix-ifaces command may be used to disable interface autoselection and just list
downlink interfaces.
It is also possible to define multiple prefix pools. See section 5.3.13 for simple prefix delegation
configuration for server or section 5.3.14 for multiple prefixes configuration. Also section 6.8.11 provides
information related to client configuration.
Figure 4: Prefix delegation (router behaviour)
4.2 Relays
In small networks, all nodes (server, hosts and routers) are connected to the same network segment
– usually Ethernet segment or a single access point or hotspot. This is very convenient as all clients can
reach server directly. However, larger networks usually are connected via routers, so direct communication
is not always possible. On the other hand it is useful to have one server, which supports multiple links –
some connected directly and some remotely.
Very nice feature of the relays is that they appear as actual servers from the client’s point of view.
Therefore no special arrangement or configuration on the client side is required. On the other hand, from
the administrator point of view, it is much easier to manage one DHCPv6 server and deploy several relays
than manage several servers on remote links.
It is important to understand that relays not simply forward DHCPv6 messages. Each message
forwarded from client to the server is encapsulated. Also each message forwarded from server to a client
is decapsulated. Therefore additional server configuration is required to deal with encapsulated (i.e.
relayed) traffic.
There are 2 ways in which server can select apropriate set of addresses, prefixes and other configuration
options. The first one is based on addresses. Relay that forwards packets from client-facing interface (e.g.
eth0) must set link-addr field in RELAY-FORW message to an address that identifies that link. Please
note that this is NOT a link-local address, it is typically a global address that identifies a link. Server
can select appropriate set of parameters if the “subnet” clause is defined. This recent addition was added
after 0.8.3 release and will be included in 0.8.4.
The second way to refer to a specific link (i.e. eth0 on the relay may be different link than eth0 on
the server), each link is referred to using its unique interface-id. It is essential to use the same indentifier
in the relay configuration as well as in the server, so both will refer to the same link using the same
number. See section 5.3.8 for example how to configure server and section 7.4.1 for corresponding relay
configuration.
It is essential to understand that the “iface relayX” in the configuration represents a link accessible
via a relay, not the relay itself. These are not the same. One obvious example is a relay thay has 2
customer facing interfaces and one for relaying data to the server. This requires two separate “iface
relayX” defintions in the server.conf file.
In larger networks it is sometimes useful to connect multiple relays. Assuming there are 2 relays
connecting server and client. Such scenario is depicted on figure 6. Requests from client are received
Figure 5: Relay deployment
by relay2, which encapsulates and sends them to relay1. Relay1 further encapsulates those messages
and sends them to the server. Since server receives double encapsulated messages, it must be properly
configured to support such traffic. See section 5.3.9 for details about server configuration and section
7.4.4 for example relays configuration.
4.3 Address and prefix assignment policy

Address and prefix assignment routines has been rewritten after 0.8.1 was released. It currently follows
this algorithm:
1. Client classification is performed (a class is assigned to a client)

2. Client access control is performed (hosts listed on black-list are rejected, if any; if there is white-list
defined, host must be on that list, otherwise it is rejected)
3. If existing lease this client/ia exists, it is assigned again (e.g. after host reboot)
4. Fixed lease is searched (using per-client configuration or so called exception mechanism). If found,
this fixed lease is assigned.
5. max-client-leases is checked. If client already has maximum number of leases, further leases are
declined.
6. Server checks if there is cached (i.e. previously assinged, but later released or expired) lease for this
client. It is assigned, if possible.
7. Server checks if client sent any hints in SOLICIT or REQUEST message. Server tries to assign
requested address or prefix. If this lease cannot be assinged for any reason, server tries to assign
similar lease (i.e. from the same pool if client’s hint was within supporte pool).
Figure 6: Cascade relays
8. Otherwise, if all of above steps fails, server assigns a random address or prefix from supported pools.
This algorithm is supported for non-temporary addresses and prefixes. It is not supported for tempo-
rary addresses.
4.4 Routing configuration

Warning: Due to objections in IETF by a small, but vocal group of opponents, further standardiza-
tion process of [30] draft in IETF was abandoned. It will not be published as RFC document. Consider
this feature a private extension.
Until recently, DHCPv6 protocol did not define a way to provision routing configuration information
to clients. The only way to deliver this information to hosts was to use Router Advertisment mechanism
in Neighbor Discovery protocol [18]. While that approach works, it brings a number of drawbacks. In
particular:
1. RA sent by router affects all hosts in a network. There is no way to differentiate this information
on a per host basis. There is no way to define additional routing information for specified class of
hosts (e.g. one department in a corporate network).
2. RA and DHCPv6 configuration has to be consistent. That is very doable, but somewhat problematic,
because network configuration has to be specified in several places. Moreover, it does not scale too
well. There are routers located in every segment of a network, while there may be just a single
DHCPv6 server deployed that serves many links.
3. Administrators experienced with IPv4 that are migrating their networks to IPv6 ask this question
very frequently: “How do I configure routing?”. Until recently the proper answer to that question
was “you don’t”.
4. In mobile environment, mobile nodes had to wait for RA and then start DHCPv6 exchange. Al-
though hosts can request RA by sending Router Solicitation (RS), that may sometimes not work,
as routers have upper limits of how many RA they are allowed to sent.
To solve aforementioned problems, a DHCPv6-based solution was proposed [30]. It allows provi-
sioning of IPv6 routing information. In particular, it allows configuration of a default route, any rea-
sonable number of specific routes and routes available on link. This feature was introduced in Dibbler
0.8.1RC1. Both server and client support it. Dibbler sources come with examples config files. See
doc/example/server-route.conf and doc/example/client-route.conf for details.
Note: This specification is not approved yet. It will change in the future. In particular, IANA
have not assigned specific option values yet. Dibbler currently uses 242 for NEXT HOP and 243 for
RTPREFIX options. Those values will change.
Note: Current implementation is a prototype. It does support only one route per router, only one
router and only a single route on-link. Although server is able to parse config that defines more than one,
it will provision only the first route or router information to a client. That is implementation limitation
that will be removed in future releases. That is not a spec limitation.
To configure routing on a server side, following config may be used
# Example server configuration file: server-route.conf
iface "eth0" {
# assign addresses from this pool
class {
pool 2000::/64
}
# router with a single route with infinite lifetime

next-hop 2001:db8:1::face:b00c {
# replace this with ::/0 to configure default route
route 2001:db8:1::/64
}
# a single next-hop without any routes defined (i.e. default router)

# This simplified mode is recommended only in bandwidth restricted
# networks. Please use full mode instead
# next-hop 2001:db8:1::cafe
# router with 3 routes defined in different ways

next-hop 2001:db8:1::dead:beef {
# route may have defined a lifetime
route 2001:db8:2::/64 lifetime 7200
# lifetime may be infinite
# route 2001:db8:3::/64 lifetime infinite
}
# prefixes available on link directly, not via router

route 2001:db8:5::/64 lifetime 3600
}
Support on client’s side is enabled in a very simple way:

# Example client configuration file: client-route.conf
# Uncomment following line to skip confirm sending (after crash or power outage)
skip-confirm
# 7 = omit debug messages

log-level 8
# Uncomment this line to run script every time response is received

script "/var/lib/dibbler/client-notify.sh"
iface "eth0" {
ia
option dns-server
option domain
routing 1
}
Two features should be enabled to reasonably use this feature. routing 1 instructs client that is
should request routing information (NEXT HOP and RTPREFIX options). Once such information is
sent by the server, client will execute a notify script. Client will run defined script and pass necessary
information to it. In particular, it will set OPTION NEXT HOP, OPTION NEXT HOP RTPREFIX and
OPTION RTPREFIX variables with contents of received option. Please see scripts/notify-scripts/client-
notify.sh for example on how to use that information to configure routing. User is also recommended to
read Section 4.8 about details of running a script and passed variables.
4.5 Custom options

Dibbler is the DHCPv6 with support for a very large number of options. However, there are always
some new options that are not yet supported. Another case is that vendors sometimes want to develop and
validate their private options before formal standarisation process takes place. Starting with 0.8.0RC1,
both client and server are able to handle custom options. Even though author tries to implement support
for as many options as possible, there are always cases, when that is not enough. Some users may also
test out new ideas, before thet get standardized. Currently only several option layouts are supported,
but that list is going to be expanded. Server is able to support following extra formats: generic (defined
by hex string), IPv6 address, IPv6 address list and string (domain). To define those options, use the
following format:
#server.conf
iface "eth0" {
class {
pool 2001:db8:1::/64
}
option 145 duid 01:02:a3:b4:c5:dd:ea

option 146 address 2001:db8:1::dead:beef
option 147 address-list 2001:db8:1::aaaa,2001:db8:1::bbbb
option 148 string "secretlair.example.org"
Similar list can be configured for client. However, client can ask for such custom options for testing
purposes only, as mechanism for handling those options once received is not yet implemented, as of
0.8.0RC1. Consider it experimental for the time being. Client can request for an option using ORO
option or even send the option in its messages.
Note that in 0.8.2 formatting of DUID-style options has changed. “hex” keyword is now required.
#client.conf
iface "eth0" {
ia
# This will send specified option value

option 145 hex 01:02:a3:b4:c5:dd:ea
# This will request specific options and interpret responses

option 149 hex
option 150 address
option 151 address-list
option 152 string
A word of warning: There are no safety checks regarding option codes, so it is possible to transmit
already defined options using this feature. Use with caution!
4.6 DNS Update

During normal operation, DHCPv6 client receives one or more IPv6 address(es) from DHCPv6 server.
If configured to do so, it can also receive information about DNS server addresses. As an additional
service, DNS Update (RFC2136, [2]) can be performed. This feature known as Dynamic DNS, or DNS
Update, keeps the DNS entries synced up with DHCP. When client boots, it gets its fully qualified domain
name and this name can be used to reach this particular client by other nodes. Details of this mechanism
is described in [2] and [16].
Note: In this section, we will assume that hostnames will be used from the example.com domain and
that addresses will be provided from the 2000::/64 pool.
There are two types of the DNS Updates. First is a so called forward resolving. It allows to change
a node’s name into its address, e.g. malcolm.example.com can be translated into 2000::123. Other kind
of record, which can be updated is a so called reverse resolving. It allows to obtain full name of a node
with know address, e.g. 2000::124 can be translated into zoe.example.com.
To configure this feature, following steps must be performed:
1. Configure DNS server. DNS server supporting IPv6 and dynamic updates must be configured. One
example of such server is an excellent ISC BIND software. Version used during writing of this
documentation was BIND 9.7.2. It is necessary to allow listening on the IPv6 sockets and define
that specific domain can be updated. See example below.
2. Configure Dibbler server to provide DNS server informations for clients. DNS Updates will be sent
to the first DNS server on the list of available servers.
Figure 7: DNS Update (performed by server)
3. Configure Dibbler server to work in stateful mode, i.e. that it can provide addresses for the clients.
This is a default mode, so unless configuration was altered, this step is already done. Make sure
that there is no ,,stateless” keyword in the server.conf file.
4. Define list of the available names in the server configuration file. Make sure to use fully qualified
domain names (e.g. malcolm.example.com), not the hostnames only.
5. Configure dibbler client to request for DNS Update. Use ,,option fqdn” to achieve this.
6. Server can be configured to execute
• both (AAAA and PTR) updates by itself

• execute PTR only by itself and let client execute AAAA update
• don’t perform any updates and let client perform AAAA update.
Note that only server is allowed to perform PTR updates. After configuration, client and/or server
should log following line, which informs that Dynamic DNS Update was completed successfully.
As of 0.8.0, both Dibbler server and client are using TCP connection for DNS Updates. Connections
are established over IPv6. There is no support for IPv4 connections. Server uses first DNS server address
specified in dns-server option. It is possible to use differentiate between DNS addresses provided to
clients and the one used for DDNS. To override DNS updates to be performed to different address, use
the following command:
fqdn-ddns-address 2001:db8:1::1
4.6.1 Example BIND configuration

Below are example configuration files for the ISC BIND 9.7.2, developed by Internet Systems Consor-
tium, Inc.. First is a relevant part of the /etc/bind/named.conf configuration file. Generally, support for
IPv6 in BIND is enabled (listen-on-v6) and there are two zones added: example.com (normal domain) and
0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa (reverse mapping). Corresponding files are stored in example.com
and rev-2000 files. For details about meaning of those directives, please consult BIND 9 Administrator
Reference Manual.
Note: Provided configuration is not safe from the security point of view. See next subsection for
details.
Figure 8: DNS Update (performed by client)
// part of the /etc/bind/named.conf configuration file

options {
listen-on-v6 { any; };
listen-on { any; };
// other global options here

// ...
};
zone "example.com" {
type master;
file "example.com";
allow-update { any; };
allow-transfer { any; };
allow-query { any; };
// other example.com domain-specific

// options follow
// ...
};
zone "0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa" {
type master;
file "rev-2000";
allow-update { any; };
allow-transfer { any; };
// other 2000::/64 reverse domain related

// options follow
// ...
};
Below are examples of two files: forward and reverse zone. First example presents how to configure
normal domain. As an example there is entry provided for zoe.example.com host, which has 2000::123
address. Note that you do not have to manually configure such entries – dibbler will do this automatically.
It was merely provided as an example, what kind of mapping will be done in this zone.
;
$ORIGIN .
$TTL 86400 ; 1 day
example.com IN SOA v13.klub.com.pl. root.v13.klub.com.pl. (
129 ; serial
7200 ; refresh (2 hours)
3600 ; retry (1 hour)
604800 ; expire (1 week)
86400 ; minimum (1 day)
)
NS v13.klub.com.pl.
A 1.2.3.4
TXT "Fake domain used for Dibbler tests."
$ORIGIN example.com.
$TTL 7200 ; 2 hours
zoe AAAA 2000::123
Second example presents zone file for reverse mapping. It contains entries for a special zone called
0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa. This zone represents 2000::/64 address space. As an example there
is a static entry, which maps address 2000::999 to a canonical name kaylee.example.com. Note that you do
not have to manually configure such entries – dibbler will do this automatically. It was merely provided
as an example, what kind of mapping will be done in this zone.
; rev-2000 example file
$ORIGIN .
$TTL 259200 ; 3 days
; this line below is split in two due to page with limitation

0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa IN
SOA 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa. hostmaster.ep.net. (
; this line above is split in two due to page with limitation
200608268 ; serial
86400 ; refresh (1 day)
1800 ; retry (30 minutes)
172800 ; expire (2 days)
259200 ; minimum (3 days)
)
NS klub.com.pl.
$ORIGIN 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa.
$TTL 86200 ; 23 hours 56 minutes 40 seconds
3.2.1 PTR picard.example.com.

9.9.9 PTR kaylee.example.com.
$ORIGIN 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa.
; example entry: 2000::999 -> troi.example.com.


9.9.9.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa
PTR troi.example.com.
; this line above is split in two due to page with limitation
Note: Due to page width limitation, if the example above, two lines were split.
4.6.2 Secure DDNS

Earlier Dibbler versions do not provide security for DNS Updates. This capability has been added in
0.8.3. It is possible to protect your updates using TSIG (Transaction Signatures), defined in RFC2845
([3]). For this feature to work, your DNS server and Dibbler server must be both configured with the
same key. The first step to use this feature is to generate a key. Currently only HMAC-MD5 keys are
supported. Please ask on dibbler-devel mailing list if you are interested in other key types. See section
10.3 for details. To generate HMAC-MD5 key, a dnssec-keygen tool from ISC BIND9 can be used:
dnssec-keygen -a hmac-md5 -b 128 -n HOST my-ddns-secret-key
For ease of configuration, dibbler uses the same key syntax in its config file as ISC BIND9 does. In
particular, all statements are finished with a semicolon. For example, the minimal set of commands to
configure a key look like the following:
key "DDNS\_UPDATER" {
secret "9SYMLnjK2ohb1N/56GZ5Jg==";
algorithm hmac-md5;
};
Please keep in mind that TSIG signatures are time sensitive and they are valid only for specified amount
of time. Therefore it is essential that your Dibbler server and your DNS server have well synchronized
clocks. It is recommented to use NTP for that purpose. By default, the signature is valid for 300 seconds.
This parameter is called a fudge. It can be modified to a different value, if needed. Shorter value is
better from the security perspective as it shortens the window of potential replay attack. Longer values
are better from the convenience perspective, as they are more “forgiving” to clock skew. The maximum
allowed value here is 65535 seconds. Please note that such a large value is not reasonable.
An example with the fudge value set to 250 is presented below:
algorithm hmac-md5;
fudge 250;
};
Any DNS server that supports DNS Updates ([2]) and TSIG ([3]) must support HMAC-MD5 signa-
tures. Following paragraph explains how to configure HMAC-MD5 key for ISC BIND9. There are at
least three steps that has to be done to achieve forward (AAAA) and reverse (PTR) updates to function
properly.
First step is to add a key. Use the same key definition that was included in your Dibbler server.conf.
Add it to BIND9 config file. Its location varies between systems, but it often /etc/bind/named.conf or
similar. You should also modify your zone and reverse zone to accept updates from this new key. Make
sure that you do not define fudge parameter, as it is not supported by BIND9. Part of the named.conf
that contains related changes looks as follows:
algorithm hmac-md5;
};
... (other configuration options here)
zone "example.org" {
type master;
file "/path/to/your/zonefile";
allow-update { key DDNS\_UPDATER; };
};
zone "0.0.0.0.1.0.0.0.8.b.d.0.1.0.0.2.ip6.arpa" {
type master;
file "/path/to/your/zonefile";
allow-update { key DDNS\_UPDATER; };
}
... (other zones and configuration options here)
In case of any problems, please refer to BIND 9 Administrator Reference Manual, available on Internet
Systems Consortium website.
4.6.3 Dynamic DNS Testing and tips

Proper configuration of the DNS Update mechanism is not an easy task. Therefore this section provides
description of several methods of testing and tuning BIND configuration. Please review following steps
before reporting issues to the author or on the mailing list.
• See example server and client configuration files described in a sections 6.8.7 and 5.3.10. Also note
that Dibbler distribution should be accompanied with several example configuration files. Some of
them include FQDN usage examples.
• Make sure that unix user, which runs BIND, is able to create and write file example.com.jnl. When
BIND is unable to create this journal file, it will fail to accept updates from dibbler and will report
failure. Check BIND log files, which are usually stored in the /var/log/ directory.
• Make sure that you have routing configured properly on a host, which will attempt to perform DNS
Update. Use ping6 command to verify that DNS server is reachable from this host.
• Make sure that your DNS server is configured properly. To do so, you might want to use nsupdate
tool. It is part of the BIND distribution, but it is sometimes distributed separated as part of
the dnsutils package. After executing nsupdate tool, specify address of the DNS server (server
command), specify update parameters (update command) and then type send. If nsupdate return
a command prompt, then the update was successful. Otherwise nsupdate will print DNS server’s
response, e.g. NOTAUTH of SRVFAIL. See below for examples of successful forward (AAAA record)
and reverse (PTR record) updates.
• After DNS Update is performed, DNS records can be verified using dig command line tool (a part of
the dnsutils package). Command syntax is: dig @(dns-server-address) name record-type. In
the following example, this query checks for name jayne.example.com at a server located at 2000::1
address. Record type AAAA (standard record for resolving name into IPv6 address) is requested.
dig tool provides server’s response in the ANSWER SECTION:. See example log below.
• In example BIND configuration above, zone transfers, queries and updates are allowed from any-
where. To make this configuration more secure, it might be a good idea to allow updates only from
a certain range of addresses or even one (DHCPv6 server’s) address only.
To manually make AAAA record update, type:

nsupdate
>server 2000::1
>update add worf.example.com 7200 IN AAAA 2000::567
>send
To manually make PTR record update, type:

nsupdate
>server 2000::1
>update add
3.2.1.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa.
86200 IN PTR picard.example.com.
>send
Note: Everything between ”update” and ”picard.example.com” must be typed in one line.
And here is an example dig session:
v13:/var# dig @2000::1 jayne.example.com AAAA
; <<>> DiG 9.3.2 <<>> @2000::1 jayne.example.com AAAA
; (1 server found)
;; global options: printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 33416
;; flags: qr aa rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 1, ADDITIONAL: 2
;; QUESTION SECTION:
;jayne.example.com. IN AAAA
;; ANSWER SECTION:
jayne.example.com. 7200 IN AAAA 2001::e4
;; AUTHORITY SECTION:
example.com. 86400 IN NS v13.klub.com.pl.
;; Query time: 6 msec

;; SERVER: 2000::1#53(2000::1)
;; WHEN: Mon Jul 24 01:38:13 2006
;; MSG SIZE rcvd: 136
4.6.4 Accepting Unknown FQDNs

By default, server configured to support FQDN has a list of names that are to be provided to clients.
But there are use cases, when client uses its own name and sends it to the server. So it makes sense to
sometimes allow client’s own domain names. Server does not know anything about such names, thus its
nickname ”Unknown FQDN”.
There are several actions that server can do, when unknown FQDN is received. To configure such
support for unknown FQDNs, accept-unknonwn-fqdn option can be defined on an interface. Depending
on its, value, it may bave domain name as a parameter. For example:
iface "eth0" {
# assign addresses from this class

class {
pool 2000::/64
}
# provide DNS server location to the clients

# also server will use this address to perform DNS Update,
# so it must be valid and DNS server must accept DNS Updates.
option dns-server 2000::1
# provide their domain name

option domain example.com
# provide fully qualified domain names for clients

# note that first, second and third entry is reserved
# for a specific address or a DUID
option fqdn 1 64
zebuline.example.com - 2000::1,
kael.example.com - 2000::2,
wash.example.com - 0x0001000043ce25b40013d4024bf5,
zoe.example.com,
malcolm.example.com,
kaylee.example.com,
jayne.example.com,
inara.example.com
# specify what to do with client’s names that are not on the list
# 0 - reject
# 1 - send other name from allowed list
# 2 - accept any name client sends
# 3 - accept any name client sends, but append specified domain suffix
# 4 - ignore client’s hint, generate name based on his address, append domain name
accept-unknown-fqdn 4 foo.bar.pl
4.7 Introduction to client classification

It is possible to define more than one address class for a single interface. Normally, when a client
asks for an address, one of the classes is being chosen on a random basis. If not specified otherwise, all
classes have equal probability of being chosen. However there are cases where an Administrator wants
to restrict access to a given pool or to have distinct ”client classes” associated to different address pools.
For example, Computer and IP-Telephone terminals can coexist in the same LAN ; but the Computer
must belong to given class pool meanwhile the IP-Telephone must belong to another pool.
In order to implement the Client Class Classification, you must first create the client class and then in
the class declaration, indicate which class to be allowed or denied. This point will be discussed in detail
in next sections.
4.7.1 Client class declaration

Each client class used for class / ta / pd addressing must be defined in the server configuration file at
global scope. A client-class declaration looks like this:
Client-class TelephoneClass{
match-if ( client.vendor-spec.en == 1234567)
}
Where TelephoneClass denotes the name of the client class and the (client.vendor-spec.en == 1234567)
denotes the condition an incoming message shall match to belong to the Client-Class. The supported
operator and data will be discussed in next section.
4.7.2 Access control

Access control is based on a per pool basis. In the client-class declaration; you can deny or allow
the client class by using the keyword ”allow” or ”deny”. For example, following class accepts all clients
except those belonging to the client class ”TelephoneClass”:
class {
2000::/64
deny TelephoneClass
}
Another example. This class accepts only client belonging to the client class ”TelephoneClass”.
class {
2000::/64
allow TelephoneClass
}
The rule can also be applied to TA/PD declaration. Several ”allow” directives can be associated to a
given pool.
ta-class {
pool 2000::/64
deny TelephoneClass
}
pd-class {
pd-pool 2000::/80
pd-length 96
deny TelephoneClass
}
4.7.3 Assigning clients to defined classes

Classifying operators are used for assigning client to a specific class. Currently, Dibbler supports the
following Operators for classifying clients:
Equal operator
Syntax : ( Expr1 == Expr2 )
Scope : global
Purpose : returns "true" if Expr1 equals Expr2
And Operator
Syntax : ( Condition1 and Condition2 )
Scope : global
Purpose : returns "true" if both Condition1 and Condition2 are "true"
Or operator
Syntax : ( Condition1 or Condition2 )
Scope : global
Purpose : returns "true" if either Condition1 or Condition2 is "true"
Contain Operator
Syntax : ( String1 contain String2 )
Scope : global
Purpose : returns "true" if String2 is a substring of String1
Substring Operator
Syntax substring ( Expr1, index, length )
Scope : global
Purpose : returns the substring of the result of that evaluation
that starts index characters from the beginning, continuing for
length characters.
Dibbler accepts different data expressions – or variables – which reflect value of options found in the
packet to which the server is responding.
client.vendor-spec.en the enterprise number value of OptionVendorSpecific (OPTION VENDOR OPTS,

option value equals to 17 as per RFC3315)
client.vendor-spec.data the data of OptionVendorSpecific (OPTION VENDOR OPTS, option value
equals to 17 as per RFC3315)
client.vendor-class.en the enterprise number value of OptionVendorClass (OPTION VENDOR CLASS,
option value equal to 16 as per RFC3315)
client.vendor-class.data the data of OptionVendorClass (OPTION VENDOR CLASS, option value
equals to 16 as per RFC3315)
4.7.4 Examples of Client-Class Classifying

Example 1 :
Client-class CPEClass {
match-if ( client.vendor-spec.data contain CPE )
}
Client belongs to CPEClass if its request message contains the Vendor Specific option with the data field
including the substring ”CPE”.
Example 2 : Combination with AND operator
Client-class TelephoneClass {
match-if (( client.vendor-spec.en == 1234) and ( client.vendor-spec.data contain CPE ) )
}
Example 3 : Combination with OR operator

Client-class TelephoneClass {
match-if (( client.vendor-spec.en == 1234) or ( client.vendor-spec.data contain CPE ) )
}
4.8 External script

Note: Support for external scripts (often called notify script was rewritten in 0.8.1RC1 release. Note
that mapping prefix and notify scripts were removed. Support for server-side script was introduced in
0.8.1RC1.
Dibbler-client is able to receive addresses, prefixes and numerous additional options. It will do its
best to set up those parameters in the system. However, the need for some extra processing may arise.
The most elegant solution is to call external script every time the configuration changes. Dibbler client
may be configured to call external script every time REPLY is received for REQUEST (new parameters
added), RENEW (parameters were updated) or RELEASE (parameters were deleted).
Name of this script is specified using script keyword followed by absolute path to script. Script will
be called with a single parameter, denoting current operation. Its value will be one of “add”, “update”,
“delete” or “expire”. Currently “expire” event is triggered on server-side only. 4 Actual values of received
parameters are passed as environment variables. In particular, IFNAME and IFINDEX variables denote
interface name and interface index that was used to communicate with server, respectively. Another
essential variable set is REMOTE ADDR. It defines address from which packet originated. That is
client’s address (when run on server) and server’s address (when run on client). Client’s message type is
passed in CLNT MESSAGE variable. Server’s response is passed in SRV MESSAGE. Note that server’s
reply is most often REPLY as script execution is skipped after sending ADVERTISE.
Addresses are passed in variables ADDR1, ADDR2 and following. Note that each ADDR variable is
accompanied with two additional variables: ADDR1PREF (address preferred lifetime) and ADDR1VALID
(address valid lifetime). Prefixes are passed in variables PREFIX1, PREFIX2 and following. Note that
each PREFIX variable is accompanied with three additional variales: PREFIX1LEN (prefix length),
PREFIX1PREF (prefix preferred lifetime), and PREFIX1VALID (prefix valid lifetime). Support for
additional options is in progress. Options are passed as environment variables. For example client DUID
(conveyed in option code 1), will be passed as OPTION1.
In 0.8.4 additional variables were added: DOWLINK PREFIX IFACES that defines a list of downlink
interfaces when splitting delegated prefix. Typically it contains (sanitized) list defined in downlink-prefix-ifaces
in client.conf or detected automatically by the client. The accompanying variable DOWNLINK PREFIXES
contains the actual prefixes that were configured on specified interfaces. Those two variables are set on
the client side only, for obvious reasons.
To enable script execution, script global option must be added to client.conf file. For example:
# client.conf
script /var/lib/dibbler/script.sh
4
Please send your feedback to mailing list if you need it also on client-side.
iface eth0 {
ia
}
4.9 Reconfiguration
Once DHCPv6 clients receive their configuration, they are not communicating with the server until
T1 timer expires. If the network configuration changes before that time, it may be useful in some cases
to inform that the clients should start reconfiguration process now, rather than wait till T1. To address
this problem, DHCPv6 offers reconfigure mechanism.
First, clients are informing the server that they are supporting reconfiguration process by sending
RECONFIGURE-ACCEPT in their SOLICIT messages. Configuration then proceeds as usual, but the
server includes AUTH option in the REPLY message with a randomly generated reconfigure-key. The
client then knows that if it receives any RECONFIGURE message, it will be signed using HMAC-MD5
generate with that particular key. That is a protection against rogue DHCPv6 servers, as the only server
that is allows to trigger reconfiguration is the one who originally provided the configuration.
The aforementioned example assumes that the default reconfigure-key authentication is used. It is
also possible to sign RECONFIGURE using delayed auth or Dibbler authentication protocol.
During start-up, the server will load its lease database and will check whether loaded database matches
existing configuration. In particular, it will check if the addresses clients have still belong to the configured
subnets. If the server detects and outdated configuration, it will send RECONFIGURE informing the
client that it must start reconfiguration process.
Clients by default have reconfigure support disabled. To enable it, use reconfigure-accept directive.
When enabling reconfigure support, it is strongly recommended to enable one of authentication methods,
e.g. reconfigure-key. See section 4.17 for detailed discussion about authentication. A short example that
has reconfigure enabled looks like this:
# client.conf - with reconfigure and reconfigure-key enabled
reconfigure-accept 1
auth-protocol reconfigure-key
auth-replay monotonic
auth-methods digest-hmac-md5
iface eth0 {
ia
}
4.10 Following M, O bits from Router Advertisements

Rounter Advertisements contain two bits that inform what kind of DHCPv6 services are available
on link. M (Managed) that tells that addresses and prefixes can be obtained using stateful DHCPv6.
¯
O (OtherConf) tells that other configuration options may be configured. Both bits are defined in [17],
¯
section 4.2. It should be noted that those bits are informational only. In the default mode (when obey-ra-
bits is absent), the client will ask for configuration options that are specified specified in its configuration
file. With obey-ra-bits, the client will wait till it receives the RA message and will act according to the
received bits. The default is off (obey-ra-bits missing). Enabling obey-ra-bits implies inactive-mode.
Let’s take this simple client configuration:
# client.conf - example that takes care of M,O bits from Router Adv.
obey-ra-bits
iface eth0 {
ia
option dns-server
}
Without obey-ra-bits enabled, it would simply send SOLICIT with one IA NA option (i.e. requesting
non-temporary address) and ORO requesting DNS-SERVER configuration. If there is RA received with
M=0, O=0, then Dibbler will not send anything and will simply wait till RA with at least one of M or
O bits is received. If RA is received with M=0, O=1, then Dibbler will request “other” configuration
options, i.e. all those that are not stateful or in other words any type of IA will not be sent. Dibbler will
send INFORMATION-REQUEST with ORO requesting DNS-SERVERS. With M=1, O=0 Dibbler will
send a SOLCIT only request an address, but will not ask for DNS-SERVERS. Finally, with M=1, O=1
Dibbler will send SOLICIT asking for both an address and DNS-SERVERS.
It should be noted that Dibbler will assess M,O bits only during start-up or while enabling an inter-
face. It will not monitor any possible future changes in those bits (e.g. as a result of receiving Router
Advertisement with updated flags).
4.11 CONFIRM message

Client detects if previous client instance was not shutdown properly (due to power outage, client crash,
forceful shutdown or similar event). In such case, it reads existing address database and checks if assigned
addresses may still be valid. If that is so, it tries to confirm those addresses by using CONFIRM message.
If you want to provoke this kind of scenario on purpose, you can run dibbler-client normally, then
forcefully kill the procss (by sending kill -9 signal, or pressing ctrl-backslash under Linux). Make sure
that you rerun client before address valid lifetime expires.
Currently, client does support only IAs in the CONFIRM.
You can force the client to not send CONFIRM message by adding the following clause to your
client.conf:
# Uncomment following line to skip confirm sending
skip-confirm
It is important to understand the meaning of the CONFIRM message. It is a question whether

specified addresses are topologically valid for a given link, not if the server has bindings for them. The
server can be provided with the information which addresses are valid on a given link using subnet clause.
This directive was introduced in Dibbler 0.8.4RC1. See section 5.3.22 for server configuration examples.
Server will try to respond to CONFIRM messages, even when subnet is not defined. In that case it will
check if the addresses are within configured address pool. If they are, the server will respond with success
status code. Otherwise it will not respond (as required by RFC3315, section 18.2.2). It is important to
understand the difference between address pool (or class) and subnet. Imagine the case of a network that
uses 2001:db8::/32 prefix. Out of that prefix only small pool (2001:db8::1-2001:db8::ff) was assigned for
server allocation. Without subnet definition, the server will be able to respond to CONFIRM messages
only for that small pool. With subnet specified in its config file, the server will be able to respond to
addresses from the whole subnet.
The exact algorithm is as follows. If there is subnet defined, check if all addresses and prefixes sent
in CONFIRM are within that subnet. If yes, respond with success status. If any of the servers is not
within the subnet, respond with NotOnLink status. If there were no addresses or prefixes specified, do
not respond. If there is not subnet defined, check if all addresses and prefixes sent in CONFIRM are
present in respective class, ta-class or pd-class ranges. If they are, respond with Success status. If any
of them is not within the pools, do not respond (because the server does not have enough knowledge to
authoritatively say that they are not valid).
4.12 Mobility
Client can also be compiled with support for link change detection. The intended use for this feature is
mobility. Client is able to detect when it moves to new link and react accordingly. Client sends CONFIRM
message to verify that its currently held address is still usable on this new link.
4.13 Leasequery
Servers provide addresses, prefixes and other configuration options to the clients. Sometime adminis-
trators may want to obtain information regarding certain leases, e.g. who has been given a specific address
or what addresses have been assigned to a specific client. This mechanism is called Leasequery [21]. New
DHCPv6 participant called requestor has been defined. Its sole purpose is to send queries and receive
responses. Dibbler provides example implementation. To define a query, command line parameters are
used.
There are two types of queries: by address (”who leases this address?”) and by client identifier (”what
addresses has this client?”). To specify one of such types, -addr or duid command-line switches can be
used. It is also mandatory to specify (using -i IFACE), which interface should be used to transmit the
query.
Here is a complete list of all command-line switches:
-i IFACE – defines thru which interface should the query be sent

-addr ADDR – sets query type to query by address. Also defines address, which the query will be
about.
-duid DUID – sets query type to query by client indentifier. Also defines client intentifier.
-timeout SECS – specifies time, which requestor should wait for response.
-dstaddr ADDR – destination address of the lease query message. By default messages are sent to the
multicast address (ff02::1:2). To transmit query to an unicast addres, use this option.
Example query 1: Who has 2000::1 address?

dibbler-requestor -i eth0 -addr 2000::1
Example query 2: Which addresses are assigned to client with specific client identifier?
dibbler-requestor -i eth0 -duid 00:01:00:01:0e:8d:a2:d7:00:08:54:04:a3:24
4.14 Stateless vs stateful and IA, TA options

This section explains the difference between stateless and stateful configurations. IA and TA options
usage is also described.
Usually, normal stateful configuration based on non-temporary addresses should be used. If you don’t
know, what temporary addresses are, you don’t need them.
Note that DHCPv6 stateless autoconfiguration is part of stateless autoconfiguration defined in [18].
There are two kinds of configurations in DHCPv6 ([5], [10]):
stateful – it assumes that addresses (and possibly other parameters) are assigned to a client. To perform
this kind of configuration, four messages are exchanged: SOLICIT, ADVERTISE, REQUEST and
REPLY.
stateless – when only parameters are configured (without assigning addresses to a client). During
execution of this type of configuration, only two messages are exchanged: INF-REQUEST and
REPLY.
During normal operation, client works in a stateful mode. If not instructed otherwise, it will request
one or more normal (i.e. non-temporary) address. It will use IA option (Identity Association for Non-
temporary Addresses, see [5] for details) to request and retrieve addresses. Since this is a default behavior,
it does not have to be mentioned in the client configuration file. Nevertheless, it can be provided:
# client.conf
iface eth0 {
ia
option dns-server
}
In a specific circustances, client might be interested in obtaining only temporary addresses. Although
this is still a stateful mode, its configuration is sligtly different. There is a special option called TA
(Identity Association for Temporary Addresses, see [5] for details). This option will be used to request
and receive temporary addresses from the client. To force client to request temporary addresses instead
of permanent ones, ta keyword must be used in client.conf file. If this option is defined, only temporary
address will be requested. Keep in mind that temporary addresses are not renewed.
# client.conf
iface eth0 {
ta
option dns-server
}
It is also possible to instruct client to work in a stateless mode. It will not ask for any type of addresses,
but will ask for specific non-address related configuration parameters, e.g. DNS Servers information. This
can be achieved by using stateless keyword. Since this is a global parameter, it is not defined on any
interface, but as a global option.
# client.conf
stateless
iface eth0
{
option dns-server
}
Some of the cases mentioned above can be used together. However, several combinations are illegal.
Here is a complete list:
none – When no option is specified, client will assume one IA with one address should be requested.
Client will send ia option (stateful autoconfiguration).
ia – Client will send ia option (stateful autoconfiguration).
ia,ta – When both options are specified, client will request for both - Non-temporary as well as Temporary
addresses (stateful autoconfiguration).
stateless – Client will request additional configration parameters only and will not ask for addresses
(stateless autoconfiguration).
stateless,ia – This combination is not allowed.

stateless,ta – This combination is not allowed.
stateless,ia,ta – This combination is not allowed.
4.15 Server address caching

Previous Dibbler versions assigned a random address from the available address pool, so the same
client received different address each time it asked for one. In the 0.5.0 release, new mechanism was
introduced to make sure that the same client gets the same address each time. It is called Server caching.
Below is the algorithm used by the server to assign an address to the client.
• if the client provided hint, it is valid (i.e. is part of the supported address pool) and not used, then
assign requested address.
• if the client provided hint, it is valid (i.e. is part of the supported address pool) but used, then
assign free address from the same pool.
• if the client provided hint, but it is not valid (i.e. is not part of the supported address pool, is
link-local or a multicast address), then ignore the hint completety.
• if the did not provide valid hint (or provided invalid one), try to assign address previously assigned
to this client (address caching)
• if this is the first time the client is seen, assign any address available.
4.16 XML files

During its execution, all dibbler components (client, server and relay) store its internal information
in the XML files. In Linux systems, they are stored in the /var/lib/dibbler directory. In Windows,
current directory (i.e. directory where exe files are located) is used instead. There are several xml files
generated. Since they are similar for each component, following list provides description for server only:
• server-CfgMgr.xml – Represents information read from a configuration file, e.g. available address
pool or DNS server configuration.
• server-IfaceMgr.xml – Represens detected interfaces in the operating system, as well as bound sockets
and similar information.
• server-AddrMgr.xml – This is database, which contains identity associations with associated ad-
dresses.
• server-cache.xml – Since caching is implemented by the server only, this file is only created by the
server. It contains information about previously assigned addresses.
4.17 Authentication and Authorization

Implementation of authentication and authorization in Dibbler in versions 0.8.4 and earlier was loosely
based on [26]. The implementation in 1.0.0 has been rewritten and is now based on standard [5] format
and mechanism, with custom extensions. Dibbler supports several mechanisms:
1. Replay detection – Dibbler is able to detect whether the messages are being new or replayed. It
implements the Replay Detection mechanism described in Section 21.3 of [5].
2. Reconfigure Key Authentication protocol – Dibbler supports reconfiguration mechanism since 1.0.0.
Reconfiguration requires that the server generates a random key when configuring clients. That
key is later used by server and client to verify if the reconfigure request really comes from the ligit
server, not a rogue one. This mechanism uses HMAC-MD5 digests. This mechanism is described
in Section 21.5 of [5].
3. Delayed Authentication protocol – It is possible to pre-provision clients with keys and configure the
server to use them to sign its messages. Client informs the server that it is capable of using this
method by sending empty AUTH option in its SOLICIT message. The server then selects a key
and sends its key id to the client and signs its response. Then the client checks if it has a key with
matching key-id and then uses it to verify incoming packets and sign its own transmissions. This
mechanism uses HMAC-MD5 digests. That follows the mechanism specified in Section 21.4 of [5]
4. Dibbler protocol – Dibbler also supports its own, custom authentication extension. It is somewhat
similar to the delayed authentication, but has a number of advantages over it. First, it secures
the whole transmission, including initial SOLICIT message. Second, it offers much stronger di-
gests: HMAC-MD5, HMAC-SHA1, HMAC-SHA224, HMAC-SHA256, HMAC-SHA384 and HMAC-
SHA512. As this is Dibbler specific extension, it is not expected to inter-operate with any other
implementations. Third, it does not require to maintain strict client DUID-key-id bindings on the
server side, as clients send ID of the key they used to protect their transmissions.
The authentication/authorization implementation in Dibbler is highly flexible. That is both blessing

and a curse. You can tweak it to match your specific needs, but if you don’t know what you are doing,
you may get only an impression of security and complicate your deployment a lot.
Both delayed authentication and Dibbler protocols are dynamic. It means that the server and the
client reads its key files every time packet is sent or received. It means that the keys can be updated in
real-time without any need for restarts.
The following subsections explain how to take advantage of each mechanisms.
4.17.1 Replay Detection

One of the possible attacks in DHCPv6 is a replay detection. In particular, the attacker could capture
RECONFIGURE message and then replay it frequently to cause the client to transmit RENEW or other
messages many times. To prevent such an attack, a mechanism called replay detection was implemented.
It’s basic principle is that the server includes a value in replay-detection field in AUTH option. That
value must be strictly increasing, i.e. the server must use greater value in any next message. Since the
message is also protected using digest, attacker can’t simply increase the value, as it would invalidate the
digest.
This parameter is configured using auth-replay. The only allowed values are none and monotonic. It
should be noted that this mechanism is useless on its own and must be used with one of other authenti-
cation mechanisms.
The example for server configuration:
# server.conf - example with enabled auth-replay protection
iface eth0 {
class {
pool 2001:db8:1::/64
}
This is an example client configuration:

# client.conf - with replay protection enabled
# This specifies replay detection mechanism.

# Available modes: none, monotonic
iface eth0 {
ia
}
4.17.2 Reconfigure Key Authentication

Reconfigure key is a mechanism that protects only RECONFIGURE message that the server sends to
clients to force them to initiate reconfiguration procedure. The major benefit of Reconfigure key algorithm
is that it does not require any preconfigured key. The server randomly generates keys on the fly when
sending REPLY message back to a client that reported support for reconfiguration. The major flaw of
the Reconfigure key algorithm is that it sends the key value as a plain text, so client is only moderately
confident that the entity that sent RECONFIGURE is indeed the server. It is sufficient to sniff the initial
client configuration procedure to obtain the key to later spoof RECONFIGURE message to trick the
client to initiate reconfiguration process.
To take advantage of reconfigure key authentication, the client must do a couple things. First, it
must support reconfiguration. Second, it must set its authentication protocol to reconfigure-key. Third,
it must discard messages that are not authenticated. Finally, it should accept authentication method
HMAC-MD5, as this is the method used by reconfigure key authentication. The minimal configuration
file for client looks like this:
# client.conf - reconfigure-key authentication
reconfigure-accept 1
iface eth0 {
ia
}
Server’s configuration is modified in the similar way:

# server.conf - reconfigure-key authentication
auth-required 0
iface eth0 {
class {
pool 2001:db8:1::/64
}
}
For a more fully featured example, see doc/examples/client-auth-reconf-key.conf for client and
doc/examples/server-auth-reconf-key.conf for server.
4.17.3 Delayed Authentication

Delayed authentication assumes that there are shared keys. Those keys must be somehow installed
on the client and server machines, using an out of band mechanisms, e.g. using scp, manually copying
keys using USB sticks etc.
Dibbler assumes that the keys are stored in /var/lib/dibbler/AAA directory. See section 4.17.5
below for details on how to generate and deploy keys. Let’s assume that the client and server shares a key
with key-id 0x01020304. In such case both client and server much name the key file AAA-key-01020304
and place it in /var/lib/dibbler/AAA directory.
In the delayed authentication keys belong to a given realm, which is really an administrative domain.
Each realm must have a unique name. For the examples we use ’dibbler test realm’ as the realm name.
Once this is done, both client and server should be configured to use delayed authentication. Here’s
minimal client’s example:
# client.conf - delayed auth
auth-protocol delayed
auth-realm ’dibbler test realm’
iface eth0 {
ia
}
Server’s configuration is similar:

# server.conf - delayed auth
auth-protocol delayed
auth-realm "dibbler test realm"
auth-required 1
iface eth0 {
class {
pool 2001:db8:1::/64
}
There is one additional step required. Server must be told which keys are to be used when communi-
cating with specific clients. That is specified using a separate file keys-mapping, which should be placed
in /var/lib/dibbler/AAA directory. The format of the file is simple. It is a text file. Each line consists
of a DUID followed by a coma, followed by key-id in hex notation. For example:
# Comments starting with # are ignored.

# So are empty lines
00:01:02:03:04:06:07:08:09, 0x010203ff
00:04:ff:ab:cd:ef:09:87:65:a1:bc, 0xabcdef00
4.17.4 Dibbler Authentication Protocol

This is a mechanism that evolved from master thesis done by Michal Kowalczuk. It was rewritten by
Tomek Mrugalski to use standard AUTH option as defined in [5], rather then using its own non-standard
AUTH, KEYGENERATION and AAAAUTHENTICATION options.
This authentication protocol provides strong protection against message tampering, can be used to
authenticate the server (i.e. client is confident that it is talking to ligitimate server) by the clients and
vice versa (i.e. the server is confident that it is providing configuration to the legitimate client).
The first step is to deploy shared keys on the clients and the server. That is explained in details in
4.17.5. The server needs only one key per client. It is possible to share the same key among multiple
clients, but that somewhat defeats the purpose of authentication. The client side requires two files: the
key itself and a AAA-SPI, which contains 32-bit key identifier. That extra mechanism is needed for cases
where client has multiple keys provisioned. That can come in handy for doing key rollover or using
different keys for different visited networks.
Both client and server can specify a list of accepted digests, using auth-methods list. The first method
on the list will be used as a default, but the server can later override it and use different method.
Care should be taken to configure client and server with at least one common method, otherwise the
authentication will fail.
Once client is provided with key and AAA-SPI file that points to that key, the client sends SOLICIT
that includes AUTH option with used key-id and digest using the first method specified on auth-methods
list. The server will use specified key-id to select appropriate key and will validate the signature. Server
will then know that the client is legitimate as it used known secret key. The server will then send
ADVERTISE option that will be protected by digest generated with the same key. Once client receives
the message, it will do exactly the same verification as the server. Client will then know that the response
was sent by legitimate server. Both sides have established their validity and the configuration process will
continue.
Depending on the intended outcome, the server may require clients to authenticate and drop packets
from non-authenticated users. That is convenient for high-security networks where only known (registered)
clients are able to get a service.
An example client configuration file looks as follows:
auth-protocol dibbler
auth-methods digest-hmac-sha256, digest-hmac-sha1, digest-hmac-md5
iface eth0 {
ia
}
An example server configuration file looks as follows:

auth-required 1
iface eth0 {
class {
pool 2001:db8:1::/64
}
}
4.17.5 Key generation

Delayed authentication and Dibbler authentication require secret key to be generated and shared
between the server and the client.
For each pair of client and server three (two for delayed authentication) files are needed. Client uses
a file AAA-SPI, which contains 32-bit AAA-SPI (AAA Security Parameter Index) — eight hexadecimal
digits, to properly introduce himself (authorize) to server. This file is needed only for Dibbler authenti-
cation.
Also it needs file named AAA-key-AAASPI , which contains a key that is used to generate authentication
information in AUTH options. The AAA-key is any number of arbitrary chosen bytes and is generated
by administrator of DHCPv6 server. The server needs only one file per client to properly communicate
using authentication. The file is named AAA-key-AAASPI , where AAASPI is the same value, that client
has in AAA-SPI file. This file contains the same AAA-key, that client has in AAA-key file. Dibbler searches
for those files in AAA directory, which is /var/lib/dibbler/AAA when running under Linux and current
directory, when running under Windows.
Typical scenario of preparing a client and server to use authentication:
1. Administrator generates AAA-key-AAASPI file. AAASPI is an arbitrary chosen 32-bit number (as
described above). The file contains any AAA-key and can be administrator’s favorite poem or can
be simply generated using dd and /dev/urandom:
$ dd if=/dev/urandom of=AAA-key-b9a6452c bs=1 count=32
2. Administrator creates file AAA-SPI which contains previously chosen AAASPI. This file will be used
by the client only.
3. Administrator transfers AAA-SPI and AAA-key-AAASPI to the client, using some secure method (e.g.
mail+PGP, scp, https) to avoid sniffing the key by a potential attacker.
4. Client: User stores AAA-SPI and AAA-key-AAASPI in AAA directory.
5. Server: Administrator stores AAA-key-AAASPI in AAA directory.
For example, configuration files can look like this:
• Server’s AAA-key-b9a6452c and client’s AAA-key (32 bytes):

ma8s9849pujhaw09y4h[80pashydp80f
• Client’s AAA-SPI (8 bytes):

b9a6452c
When configuration files are prepared and stored in client’s and server’s AAA directory you are ready
to use authentication. For detailed description of possible options see 6.7.
4.18 Exceptions: per client configuration

All configuration parameters (except FQDN) are the same for all clients, e.g. all clients will receive
the same domain name and the same DNS servers information.
However, it is sometimes useful to provide some clients with different configuration parameters. For
example computers from the accouting department in a corporate network may be configured to be in
a different subdomain. Is is possible to specify that for particular client different configuration options
should be provided. Each client is identified by its DUID, by Remote-ID or by link-local address. This
mechanism is called per client configuration, but it is sometimes referred to as exceptions. Support for
per client prefix configuration has been added in 0.8.2RC1.
See section 5.3.12 for server configuration examples.
4.19 Vendor specific information

Dibbler supports vendor specific information options. As the name suggests, that option is specific to
a particular vendor. For each vendor (or enterprise-id), there may be defined a number of sub-options.
Let’s assume that we want to define a suboption 1027 in vendor-id 4491. The value of that option should
be 0x0013. To be able to support any vendor in a flexible manner, values are specified in a hex format in
server.conf. For example:
option vendor-spec 4491-1027-0x0013
When client asks for a vendor-specific info, server will send vendor-specific info option with enterprise
number set to 4491 and option-data will contain one sub-option with code 1027. The value of that option
will be 0x0013.
Although uncommon, it is also possible to specify multiple vendor options. Another server.conf
example:
option vendor-spec 4491-1027-0x0013,1234-5678-0x0002aaaa
Server algorithm for choosing, which vendor option should be sent, works as follows:
• When client requests for a speficic vendor (i.e. sends vendor-spec info option with vendor field set),
it will receive option for that specific vendor (i.e. requested 4491, got 4491).
• When client requests any vendor (i.e. sends only option request option with vendor-spec mentioned),
it will receive first vendor-spec info option from the list (i.e. 4491/1027/0x0013).
• When client requests for not supported vendor (i.e. 11111), it will receive first vendor-spec option
from the list (i.e. 5678/0002aaaa).
It is possible to configure Dibbler client to ask for vendor-specific info. Granted value will not be
used, so from the client’s point of view this feature may be used as testing tool for the server. Client can
request vendor-specific information option in one of the following ways:
option vendor-spec – Only option request option will be sent with vendor-spec info option mentioned.
option vendor-spec 1234 – option request option will be sent with vendor-spec info option mentioned,
but also vendor-spec info option with enterprise number set to 1234 will be sent.
option vendor-spec 1234 - 5678 – option request option will be sent with vendor-spec info option
mentioned, but also vendor-spec info option with enterprise number set to 1234 and sub-option
with code 5678 will be sent.
Although that is almost never needed, it is possible to configure client to request multiple vendor-
specific options at the same time. That is also supported by the server. See 6.8.9 for examples.
However, if client sends requests for multiple vendor-specific options, which are not supported by the
server, for each sent option, server will assign one default vendor-spec option.
See 6.8.9 for client example and 5.3.11 for server examples.
4.20 Not connected interfaces (inactive-mode)

During normal startup, client tries to bind all interfaces defined in a configuration file. If such attempt
fails, client reports an error and gives up. Usually that is best action. However, in some cases it is possible
that interface is not ready yet, e.g. WLAN interface did not complete association. Dibbler attempt to
detect link-local addresses, bind any sockets or initiate any kind of communication will fail. To work
around this disadvantage, a new mode has been introduced in the 0.6.0RC4 version. It is possible to
modify client behavior, so it will accept downed and not running interfaces. To do so, inactive-mode
keyword must be added to client.conf file. In this mode, client will accept inactive interfaces, will add
them to inactive list and will periodically monitor its state. When the interface finally goes on-line, client
will try to configure it.
To test this mode, you can simulate deassociation using normal Ethernet interface. Issue following
commands:
• Bring down your interface (e.g. ifconfig eth0 down)
• edit client.conf to enable inactive-mode
• execute client: dibbler-client run
• client will print information related to not ready interface, and will periodically (once in 3 seconds)
check interface state.
• in a separate console, issue ifconfig eth0 up to bring the interface up.
• dibbler-client will detect this and will initiate normal configuration process.
In the 0.6.1 version, similar feature has been introduced on the server side. See sections 6.8.13 and
5.3.15 for configuration examples.
4.21 Parameters not supported by server (insist-mode)

Client can be instructed to obtain several configuration options, for example DNS server configuration
or domain name. It is possible that server will not provide all requested options. Older versions of the
dibbler client had been very aggressive in such case. It tried very hard to obtain such options. To do so,
it did send INF-REQUEST to obtain such option. It is possible that some other DHCPv6 servers will
receive this message and will reply with valid configuration parameters. This behavior has changed in the
0.6.0RC4 release. Right now when client does not receive all requested options, it will complain, but will
take no action. To enable old behavior, so called insist-mode has been added. To enable this mode, add
insist-mode at the global section of the client.conf file. Example configuration file is provided in the
6.8.12.
4.22 Different DUID types

There are 3 different types of the DUID (DHCP Unique Identifier):
• type 1 (link-layer + time) – this DUID is based on Link-layer address and a current timestamp.
According to spec [5], that is a default type.
• type 2 (enterprise number) – this DUID is based on the Private Enterprise Number assigned to
larger companies. Each vendor should maintain its own space of unique identifiers.
• type 3 (link-layer) – this DUID is based on link-layer address only.
According to spec [5], it is recommended to use link-layer + time, if possible. That DUID type
provides most uniqueness. It has one major drawback – it is impossible to know DUID before it is actually
generated. That poses significant disadvantage to sysadmins, who want to specify different configuration
for each client. In such cases, it is recommended to switch to link-layer only (type 3) DUIDs.
During first executing dibbler-client will generate its DUID and store it in client-duid file on disk.
During next startup DUID will be read from the file, not generated.
It is possible to specify, what DUID format should be used. It is worth noting that such definition is
taken into consideration during DUID generation only, i.e. during first client execution. To specify DUID
type, put only one of the following lines in the client.conf file:
# uncommend only ONE of the lines below

duid-type duid-llt
#duid-type duid-en 1234 0x56789abcde
#duid-type duid-ll
iface eth0 {
ia
option dns-server
}
When using link-layer+time or link-layer DUID types, dibbler will autodetect addresses. To generate
enterprise number-based DUID, specific data must be provided: enterprise-number (a 32-bit integer,
1234 in the example above) and a enterprise-specific indentifier of arbitrary length (56:78:9a:bc:de in the
example above).
4.23 Debugging/compatibility features

During interoperability test session, it has been discovered that sometimes various different implemen-
tations of the DHCPv6 protocol has problem to interact with each other. As the protocol itself does not
specify all aspects and details, some things can ba done differently and there is no only one ,,proper way”.
It also happens that some implementations may have problems with different than its authors expected
behaviors. To allow better interoperation between such implementation, dibbler has some features, which
cause different behaviors. This could result in a successful operation with other servers, clients and relays.
Normal users don’t have to worry about those options, unless they are using different servers, clients
and relays. Those options also may be useful for other vendors, who want to test their implementations.
Therefore those options can be perceived as a debugging or testing features.
4.23.1 Interface-id option

During message relaying (done by relays), options can be placed in the RELAY-FORW message is
arbitrary order. In general, there are two options used: interface-id option and relay-message option.
The former defines interface identifier, which the original data has been received from, while the later
contains the whole original message. When several relays are used, such message-in-option encapsulation
can occur multiple times.
It is possible to instruct relay to store interface-id before relay-message option or after. There is
also possibility to instruct server to omit the interface-id option altogether, but since this violates [5],
it should not be used. In general, this configuration parameter is only useful when dealing with buggy
relays, which can’t handle all option orders properly. Consider this parameter a debugging feature.
Similar parameter is defined for the server. Server uses it during RELAY-REPL generation.
See description of the interface-id-order parameters in Server configuation (section 5) and Relay
configuration (section 7).
4.23.2 Non-empty IA NA option

When client is interested in receiving an address, it sends IA NA option. In this option it may (but
don’t have to) include addresses (using IAADDR suboption) as hints for the server.
It has been detected that some servers does not support properly (perfectly valid) empty IA NA
options. To work around this problem, dibbler-client can be instructed to include two IAADDR in the
IA NA option. Here is minimal example config, which achieves that:
iface eth0 {
ia {
address
address
}
}
4.23.3 Providing address/prefix hints

Dibbler client can be instructed to send specific addresses or prefixes in its SOLICIT messages. This
can be achieved by using following syntax:
# client.conf - request specific address/prefix
iface eth0 {
ia {
address { 2001:db8:dead:beef:: }
}
pd {
prefix 2001:db8:aaaa::/64
}
}
Be default, client will use those addresses in SOLICIT message only. When transmitting REQUEST
message, it will copy proposals from ADVERTISE message, received from a server. To force client to use
those specified addresses and/or prefixes also in REQUEST, please use insist-mode directive.
4.24 Experimental features

This section contains experimental features. Besides serving as a general purpose DHCPv6 solution,
dibbler is also used as a research tool for new ideas. 5 Normal users are recommended NOT to use any
of those features. Advanced users should take extra caution. Also be aware that those options may not
work as expected, may be incomplete and not documented properly. You have been warned.
Since those mechanisms are non-standard, they are disabled by default. To enable them, ,,experimen-
tal” keyword must be placed in the client.conf or server.conf files.
5
This was particularly true during my Ph. D. research.
4.24.1 Server Performance mode

When running in a normal mode, the server rewrites its full database every time there is a change. That
becomes problematic once the number of clients is large and number of packets per second is sufficiently
high. To somehow eleviate the problem, an experimental performance-mode has been implemented. The
server will load its database at start, then keep it in memory only and will write it again to disk during
normal shutdown procedure. That should work, but it is dangerous! If there is power failure, server crash
or other event, the server may not be able to write its database to disk and you’ll lose your database.
To use this feature, use the following config:
# We want the server to not waste time on logging
log-level 3
# Enable experimental features

experimental
# Enable performance mode

performance-mode 1
iface eth0 {
class {
pool 2001:db8::/64
}
}
If you are not satisfied with Dibbler performance, please submit patches or better yet, consider the
alternative: Kea (BIND10 DHCP) http://bind10.isc.org/wiki/Kea. It offer tremendous performance,
is open source and is being actively developed by a professional team. Its lead developer happen to be
Dibbler author as well :)
4.24.2 Address Parameters

Note: This feature is experimental, i.e. it is not described by any RFC or even internet
draft. Don’t use it, unless you exactly know what you are doing.
There is ongoing process to register and publish internet draft, which describes this operation. Latest
versions of this draft will be availabe at http://klub.com.pl/dhcpv6/doc/.
RFC3315 ([5]) defines means of allocating IPv6 addresses to all interested clients. Clients are able to
obtains IPv6 addresses and other configuration parameters from the servers. Unfortunately, client after
obtaining an address, are not able to communicate each other due to missing prefix information. That
property of the DHCPv6 procotol is sometimes perceived as a major disadvantage. To overcome this
deficiency, an extension to the protocol has been proposed.
It is possible to attach additional option conveyed in normal IAADDR option. That additional option,
called ADDRPARAMS option, contains additional information related to that address. To maintain
backward compatibility, server does not send such option by default, even when configured to support it.
To make server send this option, client must explicitly ask for it.
Below are example configuration files for server and client. Note that since that is an non-standard fea-
ture, user must explicitly allow experimental options before configuring it (thus ,,experimental” keyword
is required).
Example client.conf configuration file:
#client.conf
log-mode short
log-level 8
iface "eth0" {
ia {
addr-params
}
}
Example server.conf configuration file:

#server.conf
log-level 8
experimental
log-mode short
iface eth0 {
t1 60
t2 96
prefered-lifetime 120
valid-lifetime 180
class {
addr-params 80
pool 2001:458:ff01:ff03::/80
}
}
4.24.3 Remote Autoconfiguration

Every time a node attaches to a new link, it must renew or obtain new address and parameters, using
DHCPv6 protocol (namely CONFIRM or SOLICIT messages. In case of mobile nodes, it is beneficial
to obtain address and other configuration parameters remotely, before actually attaching to destination
link. This extension provides experimental support for such operation. Details of this mechanism are
thoroughly discussed in [27, 31, 28, 29].
The idea is that once client attaches to its current location, normal configuration procedure is initiated
(SOLICIT, ADVERTISE, REQUEST and REPLY ). However, besides requesting the usual options, client
also asks for NEIGHBORS option. Server provides that option that contains list of available DHCPv6
servers at neighboring networks.
Once client gains that information, it then initiates remote autoconfiguration process, i.e. it sends
SOLICIT message to each of the newly discovered neighbors, requesting single IPv6 address. Servers
respond remotely, using REPLY message. Once this exchange is completed, client knows its new IPv6
address for each of the potential handover targets. What is especially important is that client obtains that
knowledge, while still being connected to old location. It may leverage that knowledge, e.g. to update
his correspondent nodes in advance.
As Dibbler client is not a mobility software itself, it has to communicate with Mobile IPv6 stack
somehow. Therefore it triggers ./remote-autoconf script every time remote autoconfiguration is concluded.
Note that to support this scenario, both client and all participating servers must have unicast and
rapid-commit support enabled.
Following series of server.conf files demonstrate, how 3 servers can be configured to incorm client about
their 2 neighbors.
#server.conf for server1.
log-level 8
log-mode short
preference 2
experimental
iface "eth0" {
t1 1800
class {
pool 2001:db8:1111::/64
}
rapid-commit 1
unicast 2001:db8:1111::f
option neighbors 2001:db8:2222::f,2001:db8:3333::f

}
#server.conf for server2

log-level 8
log-mode short
preference 1
experimental
iface "eth1" {
rapid-commit 1
class {
pool 2001:db8:2222::/64
}

}
log-level 8
preference 0
experimental
iface "eth1" {
rapid-commit 1
class {
pool 2001:db8:3333::/64
}
option dns-server 2001:db8:3333::f

}
Client also needs to have enabled number of features. Following config file may serve as an example:
log-mode short
log-level 8
experimental
remote-autoconf
iface "eth0" {
ia
unicast 1
}
4.25 Obsoleted experimental features

This subsection describes experimental features that are not supported anymore. This list is provided
for historical reasons. It may be useful for someone to ease tracking of features removal, e.g. to get the
latest version that still has support for something.
4.25.1 Mapping prefix

Mapping prefix was an extension that altered client’s behavior when delegated prefix is received.
Instead of considering it as a prefix that should be distributed on other interfaces, it is used as a map-
ping prefix. Normal prefix processing is supressed and external script is executed: mappingprefixadd or
mappingprefixdel. That script must be present in the working directory (that would be /var/lib/dibbler
under Linux or current directory (Windows). This feature was removed in 0.8.0RC1.
4.25.2 Tunnel mode

As support for DS-Lite [24] support was added in 0.8.0RC1, the old support for configuring tunnels
was removed.
5 Server configuration
Server configuration is stored in server.conf file in the /etc/dibbler (Linux systems) or in current
(Windows systems) directory.
5.1 Scopes
Configuration file can be logically split into separate “sections” that are called scopes, for example
interface scope contains parameters related to configuration served over a given interface. Some scopes
can contain other scopes. Some commands are specific to a given a given scope.
5.1.1 Global scope

Every option can be declared in a global scope. Global options can be defined here. Also options of a
smaller scopes can be defined here – they will be used as a default values. Configuration file has following
syntax:
global-options
interface-options
class-options
interface-declaration
5.1.2 Interface declaration

Each network interface, which should be serviced by the server, must be mentioned in the configuration
file. Network interface is defined like this:
iface interface-name
{
interface-options
class-options
}
or
iface number
{
interface-options
class-options
}
where interface-name denotes name of the interface and interface-number denotes its number.
Name no longer needs to be enclosed in single or double quotes (except Windows systems, when interface
name contains spaces). Note that virtual interfaces, used to setup relay support are also declared in this
way.
5.1.3 Address class scope

Class is a smallest scope used in the server configuration file. It contains definition of the addresses,
which will be provided to clients. Only class scoped parameters can be defined here. Address class is
declared as follows:
class
{
class-options
address-pool
}
Address pool defines range of the addresses, which can be assigned to the clients. It can be defined in
one of the following formats:
pool minaddress-maxaddress
pool address/prefix
5.1.4 Prefix class scope

That is an equivalent of address class for a prefix delegation. It contains definition of prefixes that
are going to be delegation to clients. Only pd-class scoped parameters can be defined here. Prefix class
is declared as follows:
pd-class
{
pd-pool prefix/length
pd-length prefix-length
}
5.1.5 Temporary address class scope

That is an equivalent of address class for temporary addresses. It contains definition of temporary
addresses that are going to be assigned to clients that request temporary addresses. Only ta-class scoped
parameters can be defined here. Prefix class is declared as follows:
ta-class {
pool 2001:db8:1::1-2001:db81:1::ffff
}
5.1.6 Routing scope

Support for routing configuration was added in 0.8.0RC1. It is possible to define routing scope. Each
scope represents a single router available on-link. In this scope, routes available via specified link my be
defined.
next-hop address-of-a-router
{
route1-parameters
route2-parameters
...
}
5.1.7 Client scope

Server allows defining custom parameters on a per-host basis. See Sections 4.18 and 5.3.12 for details.
There are three types of reservations: DUID-based, remote-id based and link-local based. Following
syntax can be used:
client duid 00:00:00:00:00
{
[address 2001:db8:1::]
[prefix 2001:db8:1::/64]
option1
option2
...
}
client remote-id 5-0x01020304

{
[prefix 2001:db8:1::/64]
option1
option2
...
}
client link-local fe80::1234:56ff:fe78:9abc

{
[prefix 2001:db8:1::/64]
option1
option2
...
}
5.1.8 Key scope

Dibbler 0.8.3 introduced support for secure DNS Updates using TSIG mechanism. Since this key is
expected to be also used in DNS server software, syntax is kept very similar to syntax accepted in ISC
BIND9 software. Note semicolons at the end of each statement.
key key-name {
secret "base64encodedSecretHere==";
algorithm algorithm-type;
...
};
5.2 Server options

So called standard options are defined by the base DHCPv6 specification, a so called RFC 3315
document [5]. Those options are called standard, because all DHCPv6 implementations, should properly
handle them. Each option has a specific scope it belongs to.
Standard options are declared in the following way:
OptionName option-value
work-dir – (scope: global). Takes one parameter of string type. Defines working directory.
log-level – (scope: global). Takes one integer parameter. Defines verbose level of the log messages. The
valid range is from 1 (very quiet) to 8 (very verbose). Those values are modelled after levels used
in syslog. These are: 1(Emergency), 2(Alert), 3(Critical), 4(Error), 5 (Warning), 6(Notice), 7(Info)
and 8(Debug). Currently Dibbler is using levels 3 to 8, as 1 and 2 are reserved for system wide
emergency events.
log-name – (scope: global). Takes one string parameter. Defines than name, which will be used during
logging.
log-mode – (scope: global). Takes one parameter that can be short, full, precise or syslog. Defines
logging mode. In the default, full mode, name, date and time in the h:m:s format will be printed. In
short mode, only minutes and seconds will be printed (this mode is useful on terminals with limited
width). Precise mode logs information with seconds and microsecond precision. It is a useful as
a performance diagnostic tool for finding bottlenecks in the DHCPv6 autoconfiguration process.
Syslog works under POSIX systems (Linux, Mac OS X, BSD family) and allows default POSIX
logging functions.
log-colors – (scope: global). Takes one boolean parameter. Defines if logs printed to console should use
colors. That feature is used to enhance logs readability. As it makes the log files messy on systems
that do not support colors, it is disabled by default. The default is off.
cache-size – (scope: global). Takes one parameter that specifies cache size in bytes. The default value
is 1048576 (1MB). It defines a size of the memory (specified in bytes) which can se used to store
cached entries.
stateless – (scope: global). It may be present or missing. The default is missing. Defines that server
should run in stateless mode. In this mode only configuration parameters are defined, not addresses
or prefixes. It is mutually exclusive with class, ta-class and pd-class. See Section 4.14.
interface-id-order – (scope: global). Take one parameter that can be one of before, after or omit.
The default is before. This parameter defines placement of the interface-id option. During message
relaying options can be placed in the RELAY-REPL message is arbitrary order. This option has
been specified to control that order. interface-id option can be placed before or after relay-message
option. There is also possibility to instruct server to omit the interface-id option altogether, but since
this violates [5], it should not be used. In general, this configuration parameter is only useful when
dealing with buggy relays, which can’t handle all option orders properly. Consider this parameter
a debugging feature. Note: similar parameter is available in the dibbler-relay.
experimental – (scope: global). Allows enabling experimental features. There are some highly-experimental
features present in Dibbler. To make a clear statement about their experimental nature, user is re-
quired to acknowledge that fact by putting this statement in its config file. This statement may be
present or absent. The default is absent.
inactive-mode – (scope: global, type: present or missing, default: missing). This enables so called
inactive mode. When server begins operation and it detects that required interfaces are not ready,
error message is printed and server exits. However, if inactive mode is enabled, server sleeps instead
and wait for required interfaces to become operational. That is a useful feature, when using wireless
interfaces, which take some time to initialize as associate.
accept-leasequery – (scope: interface). Takes one boolean parameter that specifies if server should
support leasequery [21] protocol on a given interface. The default value is 0 (leasequery is not
supported by default). See Section 4.13.
guess-mode – (scope: global, type: present or missing, default: missing). Server tries to match incoming
relayed messages based on interface-id first. If that fails, it tries to match based on linkaddr field in
the RELAY-FORW message (see subnet keyword definition). Normally, when both of those match
attempts fail, the server will drop the packet. When guess-mode option is enabled, server will will
use first relay defined. It may save the day, if you have only one relay in your network, but it will
almost certainly do a wrong thing if you have more than one. This is as its name states: just a
guess. Use with caution!
script – (scope: global). Takes one string parameter that specifies name of a script that will be called
every time something important happens in a system, e.g. when address or prefix is assigned,
updated or released. See Section 4.8.
fqdn-ddns-address – (scope: global). Takes one parameter that specifies address of DNS server that
will be used for DNS Updates. See Section 4.6.
ddns-protocol – (scope: global). Takes one string parameter. Defines protocol that should be used
during DNS Update mechanism. Allowed values are tcp, udp and any. Any means that UDP will
be tried first and if it fails, update will be retried over TCP. See Section 4.6.
ddns-timeout – (scope: global). Takes one integer parameter that specifies timeout in milliseconds.
Defines how long client should wait for DNS server response during DNS Update before declaring
update a failure. See Section 4.6.
subnet – (scope: interface). This definition must be followed by a IPv6 address followed by slash followed
prefix length (e.g. 2001:db8::/32). It defines all IPv6 addresses that are valid on a given interface.
It is used mainly for matching relayed traffic and responding to confirm messages. Server will not
use that whole range to assign addresses. That is specified with class, ta-class or pd-class.
class – (scope: interface). This definition must be followed by curly braces and creates a new address
class scope. See Section 5.1.3.
pd-class – (scope: interface). This definition must be followed by curly braces and creates a new prefix-
delegation class scope. See Section 5.1.4.
ta-class – (scope: interface). This definition must be followed by curly braces and creates a new tempo-
rary address class scope. See Section 5.1.5.
next-hop – (scope: interface). This definition takes one parameter that defines IPv6 address of a router.
Without any further parameters, it conveys an information about default route for bandwidth
limited networks. That mode is discouraged, unless there are significant bandwith limitations. It is
usually followed by curly braces that create a new route scope. See Section 5.1.6.
preference – (scope: interface, type: 0-255, default: none). Eech server can be configured to a spe-
cific preference level. When client receives several ADVERTISE messages, it should choose that
server, which has the highest preference level. It is also worth noting that client, upon reception
of the ADVERTISE message with preference set to 255 should skip wait phase for possible other
ADVERTISE messages.
unicast – (scope: interface, type: address, default:none). Normally clients sends data to a well known
multicast address. This is easy to achieve, but it wastes network resources as all nodes in the network
must process such messages and also network load is increased. To prevent this, server might be
configured to inform clients about its unicast address, so clients, which accept it, will switch to a
unicast communication.
rapid-commit – (scope: interface, type: boolean, default: 0). This option allows rapid commit proce-
dure to be performed. Note that enabling rapid commit on the server side is not enough. Client
must be configured to allow rapid commit, too.
iface-max-lease – (scope: interface, type: integer, default: 232 − 1). This parameter defines, how many
normal addresses can be granted on this interface.
client-max-lease – (scope: interface, type: interger, default:232 − 1). This parameter defines, how many
addresses one client can get. Main purpose of this parameter is to limit number of used addresses
by misbehaving (malicious or restarting) clients.
relay – (scope: interface). Takes one string or integer parameter that designated interface name or
interface index. It is used in relay definition. It specifies name of the physical (or name of another
relay, if cascade relaying is used) interface, which is used to receive and transmit relayed data. See
4.2 for details of relay deployment and sections 5.3.8 and 5.3.9 for configuration examples.
interface-id – (scope: interface, type: integer, default: not defined). Used in relay definition. Each relay
interface should have defined its unique identified. It will be sent in the interface-id option. Note
that this value must be the same as configured in the dibbler-relay. It may be possible to specify
this parameter by using a number (option will be 4 bytes long), a string or a hex of arbitrary length
(please use the same format as for DUID). See 4.2, 5.3.8 and 7.4 for details.
vendor-spec – (scope: interface, type: integer-hexstring, regular string or an IPv6 address, default:
not defined). This parameter can be used to configure some vendor-specific information option.
Since there are no dibbler-specific options, this implementation is flexible. User can specify in the
configuration file, how should this option look like. See 5.3.11 section for details. It is uncommon, but
possible to define several vendor specific options for different vendors. In such case, administrator
must specify coma separated list. Each list entry is a vendor (enterprise number), ,,–” sign and a
hex dump (similar to DUID).
pool – (scope: class). Takes coma separated IPv6 address ranges. Each range is defined as first-address,
a dash and a second address. Defines a range of available addresses that will be assigned in specific
class. An example pool definition looks like this:
pool 2001:db8:abcd:: - 2001:db8:abcd::ffff
It is also possible to use prefix/length notation.
pd-pool – (scope: pd-class). Takes coma separated IPv6 address ranges. Each range is defined as fist-
address, a dash and a second address. Defines a range of available prefixes (only prefixes themselves,
not their lengths) that will be assigned in specific class. An example pd-pool definition looks like
this:
pd-pool 2001:db8:abcd:: - 2001:db8:abcd::ffff
It is also possible to use prefix/length notation.
share – (scope: class). Defines percentage of clients that a class should handle. This parameter is only
useful if there are more then one class defined. See Section 5.3.7.
T1 – (scope: class, type: integer or integer range: default: 232 − 1). This value defines after what time
client should start renew process. Exact value or accepted range can be specified. When exact value
is defined, client’s hints are ignored completely.
T2 – (scope: class, type: integer or integer range, default:232 − 1). This value defines after what time
client will start emergency rebind procedure if renew process fails. Exact value or accepted range
can be specified. When exact value is defined, client’s hints are ignored completely.
valid-lifetime (scope: class, type: integer or integer range, default:232 −1). This parameter defines valid
lifetime of the granted addresses. If range is specified, client’s hints from that range are accepted.
preferred-lifetime (scope: class, type: integer or integer range, default:232 − 1). This parameter defines
prefered lifetime of the granted addresses. If range is specified, client’s hits from that range will be
accepted.
class-max-lease – (scope: interface, type: interger, default:232 − 1). This parameter defines, how many
addresses can be assigned from that class.
reject-clients – (scope: class, type: address or DUID list, default: none). This parameter is sometimes
called black-list. It is a list of a clients, which should not be supported. Clients can be identified by
theirs link-local addresses or DUIDs.
accept-only – (scope: class, type: address or DUID list, default: none). This parameter is sometimes
called white-list. It is a list of supported clients. When this list is not defined, by default all clients
(except mentioned in reject-clients) are supported. When accept-only list is defined, only client
from that list will be supported.
drop-unicast – (scope: global). In general, clients are supposed to send their messages to multicast,
unless the server explicitly allows unicast. [5] says that packets sent to unicast should be dropped
and server must respond with status code set to UseMulticast. That’s a bit harsh in author’s
opinion, so this behavior is not set by default. However, if you want strict RFC compliance, you
can enable this by adding drop-unicast in your global scope.
addr-params – (scope: class). Experimental feature that takes one boolean parameter. It defines prefix
length that is configured in addr-params option. See Section 4.24.2 for details and warnings.
performance-mode – (scope: global). Experimental feature that boosts server performance. It takes
one integer parameter with 0 or 1 control whether performance mode is to be enabled or disabled.
The default is 0 (disabled). Use with caution. See Section 4.24.1 for details and warnings.
reconfigure-enabled – (scope: global). This directive controls whether server will attempt to send
RECONFIGURE message at start or not. It takes one integer parameter with allowed values being
0 or 1. The default is 0 (disabled).
allow – (scope: class). Specifies that clients that belong to a specific client class are allowed to use that
address class. Takes one string parameter that defines client class name. See Section 4.7.
deny – (scope: class). Specifies that clients that belong to a specific client class are denied use of that
address class. Takes one string parameter that defines client class name. See Section 4.7.
option dns-server – (scope: interface, type: address list, default: none). This option conveys informa-
tion about DNS servers available. After retriving this information, clients will be able to resolve
domain names into IP (both IPv4 and IPv6) addresses. Defined in [7].
option domain – (scope: interface, type: domain list, default: none). This option is used for configuring
one or more domain names, which clients are connected in. For example, if client’s hostname is
alice.mylab.example.com and it wants to contact bob.mylab.example.com, it can simply refer
to it as bob. Without domain name configured, it would have to use full domain name. Defined in
[7].
option ntp-server – (scope: interface, type: address list, default: none). This option defines informa-
tion about available NTP servers. Network Time Protocol [1] is a protocol used for time synchro-
nisation, so all hosts in the network has the same proper time set. Defined in [14].
option time-zone – (scope: interface, type: timezone, default: none). It is possible to configure time-
zone, which is provided by the server. Note that this option is considered obsolete as it is mentioned
in draft version only [32]. Work on this draft seems to be abandoned as similar functionality is
provided by now standard [14].
option sip-server – (scope: interface, type: address list, default: none). Session Initiation Protocol [4] is
an control protocol for creating, modifying, and terminating sessions with one or more participants.
These sessions include Internet telephone calls, multimedia distribution, and multimedia conferences.
Its most common usage is VoIP. Format of this option is defined in [6].
option sip-domain – (scope: interface, type: domain list, default: none). It is possible to define domain
names for Session Initiation Protocol [4]. Configuration of this parameter will ease usage of domain
names in the SIP protocol. Format of this option is defined in [6].
option nis-server – (scope: interface, type: address list, default: none). Network Information Service
(NIS) is a Unix-based system designed to use common login and user information on multiple
systems, e.g. universities, where students can log on to ther accounts from any host. Its format is
defined in [11].
option nis-domain – (scope: interface, type: domain list, default: none). Network Information Service
(NIS) can albo specify domain names. It can be configured with this option. It is defined in [11].
option nis+-server – (scope: interface, type: address list, default: none). Network Information Service
Plus (NIS+) is an improved version of the NIS protocol. This option is defined in [11].
option nis+-domain – (scope: interface, type: domain list, default: none). Similar to nis-domain, it
defines domains for NIS+. This option is defined in [11].
option lifetime – (scope: interface, type: boolean, default: no). Base spec of the DHCPv6 protocol
does offers way of refreshing addresses only, but not the options. Lifetime defines, how often client
should renew all its options. When defined, lifetime option will be appended to all replies, which
server sends to a client. If client does not support it, it should ignore this option. Format of this
option is defined in [13].
option fqdn – (scope: interface). Takes 0, 1 or 2 integer parameters that are followed by FQDN list.
Additional integer parameters designate fqdn-mode and reverse zone length in DNS Update. FQDN-
mode can have 3 values: 2 (both AAAA and PTR record will be updated by server), 1 (server will
update PTR only) or 0 (server will not update anything). Reverse zone length is an integer between
0 and 128 and designates reverse zone length. FQDN list is a coma separated list of fully qualified
domain names, with possible reservations for DUIDs or addresses. FQDN mechanism is defined in
[16]. See Section 4.6.
accept-unknown-fqdn – (scope: Interface). Takes one integer parameter, possibly followed by second
string parameter that designated domain name. It specifies how server should react to incoming
FQDN options that contain names that are unknown to the server. Allowed values are 0 (reject),
1 (send other name from allowed list), 2 (accept any name client sends), 3(accept any name client
sends, but append specified domain suffix) and 4 (ignore client’s hint, generate name based on his
address, append domain name). Choices 3 and 4 require additional string parameter that defines
domain suffix. See Sections 4.6 and .
option – (scope: interface). Takes one integer number followed by several possible parameter combina-
tions. It defines custom option that server may send out to clients. Supported formats are:
option number - DUID
option number address-keyword address
option number address-list
option number string-keyword string
Where number is an integer that defined option type, DUID is a hex-formatted string that defines
option content, address-keyword is a word “address”, address is an IPv6 address, address-list is
coma separated list of addresses, string-keyword is a word “string” and string is any string enclosed
in single or double quotes. See Section 4.5 and 5.3.20.
option aftr – (scope: interface, type: FQDN). In Dual-Stack Lite networks, client may want to configure
DS-Lite tunnel. Client may want to obtain information about AFTR (a remote tunnel endpoint).
This option conveys a fully qualified domain name of the remote tunnel. This option is defined in
[24].
option neighbors – (scope: interface). Experimental feature for Remote Autoconfiguration. Do not
use it unless you know exactly what you are doing. Takes coma separated list of addresses. This
option requires experimental mode to be enabled. See Section 4.24.3.
auth-protocol – (scope: global, type: string, default: none). This is a crucial parameter that spec-
ifies which authorization/authentication protocol is used. Allowed values are: none, delayed,
reconfigure-key and dibbler. See section 4.17 for details.
auth-methods – (scope: global, type: string, default: empty). This a coma separated list of one or
more methods. The first one on the list will specify the default method, while the others list
accepted methods when receiving data from clients. Set it to one or more of the following values
to enable authentication on ther server side, using selected method of generating authentication
information: none, digest-plain, digest-hmac-md5, digest-hmac-sha1, digest-hmac-sha224,
digest-hmac-sha256, digest-hmac-sha384, and digest-hmac-sha512.
auth-replay – (scope: global, type: string, default: none). Specifies which replay detection methods are
supported. Currently two values are implemented: none and monotonic.
auth-required – (scope: global, type: boolean, default: 0). This parameter specifies if the client is
required to authenticate itself. When set to 0, any client authentication failures (invalid signature
or lack of AUTH option) will result in a warning only. When set to 1, such messages will be dropped.
client-class – (scope: global). Takes one string parameter that defines name of a client class. Client
class name is followed by curly brackets that create client-class scope. Clients can be grouped into
classes depending on rules defined in client-class. This can be used together with allow and deny
to assign segregate clients into different groups. See Section 4.7 for overview and Section 5.2.1 for
list of supported expressions.
address – (scope: client). Takes one parameter that specifies address. It instructs server to reserve this
particular address for defined client. See Sections 4.18 and 5.3.12 for details.
prefix – (scope: client). Takes one parameter that specifies prefix using prefix/length notation. It
instructs server to reserve specified prefix for defined client. See Sections 4.18 and 5.3.12 for details.
key – (scope: global). Take one string parameter that specifies key name. This keyword instructs server
to create a key with a specified name. This key will be used for TSIG in DNS Updates. It is followed
by curly braces that open up a new key scope. Closing curly brace is followed by a semicolon.
secret – (scope: key). Takes one string parameter and is followed by a semicolon. That parameter is a
base64-encode secret value of the key. It is mandatory if key scope is defined.
algorithm – (scope: key). Takes one enum argument that is followed by a semicolon. This parameter
is mandatory if the key scope is present. It specifies agorithm for the key. Currently the supported
value is hmac-md5.
fudge – (scope: key). Takes one integer parameter that is followed by a semicolon. Each TSIG signature
is valid for a specified amount of time only. This optional parameter specifies period in which TSIG
is valid, expressed in seconds. If missing, the default value of 300 is used. The allowed values are
between 0 and 65535.
5.2.1 Client class quantifiers

Additional parameters are used during client class definition. See Section 4.7 for details and examples.
match-if – (scope: client-class).
contain – (scope: client-class).
substring – (scope: client-class).
== – (scope: client-class).
and – (scope: client-class).
or – (scope: client-class).
client.vendor-spec.en – (scope: client-class).
client.vendor-spec.data – (scope: client-class).
client.vendor-class.en – (scope: client-class).
client.vendor-class.data – (scope: client-class).
5.3 Server configuration examples

This subsection contains various examples of the server configuration. If you are interested in addi-
tional examples, download source version and look at *.conf files.
5.3.1 Example 1: Simple

In opposite to client, server uses only interfaces described in config file. Let’s examine this common
situation: server has interface named eth0 (which is fourth interface in the system) and is supposed to
assign addresses from 2000::100/124 class. Simplest config file looks like that:
# server.conf
iface eth0
{
class
{
pool 2000::100-2000::10f
}
}
5.3.2 Example 2: Timeouts

Server should be configured to deliver specific timer values to the clients. This example shows how to
instruct client to renew (T1 timer) addresses one in 10 minutes. In case of problems, ask other servers
in 15 minutes (T2 timer), that allowe prefered lifetime range is from 30 minutes to 2 hours, and valid
lifetime is from 1 hour to 1 day. DNS server parameter is also provided. Lifetime option is used to make
clients renew all non-address related options renew once in 2 hours.
# server.conf
iface eth0
{
T1 600
T2 900
prefered-lifetime 1800-3600
valid-lifetime 3600-86400
class
{
pool 2000::100/80
}

option lifetime 7200
}
5.3.3 Example 3: Limiting amount of addresses

Another example: Server should support 2000::0/120 class on eth0 interface. It should not allow any
client to obtain more than 5 addresses and should not grant more then 50 addresses in total. From this
specific class only 20 addresses can be assigned. Server preference should be set to 7. This means that
this server is more important than all server with preference set to 6 or less. Config file is presented below:
# server.conf
iface eth0
{
iface-max-lease 50
client-max-lease 5
preference 7
class
{
class-max-lease 20
pool 2000::1-2000::100
}
}
5.3.4 Example 4: Unicast communication

Here’s modified previous example. Instead of specified limits, unicast communication should be sup-
ported and server should listen on 2000::1234 address. Note that default multicast address is still sup-
ported. You must have this unicast address already configured on server’s interface.
# server.conf
log-level 7
iface eth0
{
unicast 2000::1234
class
{
pool 2000::1-2000::100
}
}
5.3.5 Example 5: Rapid-commit

This configuration can be called quick. Rapid-commit is a way to shorten exchange to only two
messages. It is quite useful in networks with heavy load. In case if client does not support rapid-commit,
another trick is used. Preference is set to maximum possible value. 255 has a special meaning: it makes
client to skip wait phase for possible advertise messages from other servers and quickly request addresses.
# server.conf
log-level 7
iface eth0
{
rapid-commit yes
preference 255
class
{
pool 2000::1/112
}
}
5.3.6 Example 6: Access control

Administrators can selectively allow certain client to use this server (white-list). On the other hand,
some clients could be explicitly forbidden to use this server (black-list). Specific DUIDs, DUID ranges,
link-local addresses or the whole address ranges are supported. Here is config file:
# server.conf
iface eth0
{
class
{
# duid of the rejected client
reject-clients ‘‘00001231200adeaaa’’
2000::2f-2000::20 // it’s in reverse order, but it works.
// just a trick.
}
}
iface eth1
{
class
{
accept-only fe80::200:39ff:fe4b:1abc
pool 2000::fe00-2000::feff
}
}
5.3.7 Example 7: Multiple classes

Although this is not common, a few users have requested support for multiple classes on one interface.
Dibbler server can be configured to use several classes. When client asks for an address, one of the classes
is being choosen on a random basis. If not specified otherwise, all classes have equal probability of being
chosen. However, this behavior can be modified using share parameter. In the following example, server
supports 3 classes with different preference level: class 1 has 100, class 2 has 200 and class 3 has 300.
100 200
This means that class 1 gets 100+200+300 ≈ 16% of all requests, class 2 gets 100+200+300 ≈ 33% and class
300
3 gets the rest ( 100+200+300 = 50%).
# server.conf
log-level 7
log-mode short
iface eth0 {
T1 1000
T2 2000
class {
share 100
pool 4000::1/80
}
class {
share 200
pool 2000::1-2000::ff
}
class {
share 300
pool 3000::1234:5678/112
}
}
5.3.8 Example 8: Relay support

To get more informations about relay configuration, see section 4.2. Following server configuration
example explains how to use relays. There is some remote relay with will send encapsulated data over
eth1 interface. It is configured to append interface-id option set to 5020 value. Let’s allow all clients using
this relay some addresses and information about DNS servers. Also see section 7.4.1 for corresponding
relay configuration.
Note that although eth1 interface is mentioned in the configuration file, direct traffic from clients
located on the eth1 interface will not be supported. In this example, eth1 is used only to support requests
relayed from remote link identified with interface-id value “relay1:en1”. That value is what relay inserts
into its packets. I may sometimes not be configurable on hardware relays, so the best approach is to
configure your server to match whatever the relay inserts there.
Server will try to match incoming packet based on exact match of interface-id value. If that fails, there
is another method. Relay sets linkaddr field and server may find a match based on that information. For
this mechanism to work, the admin must provide information about valid subnet available on specified
link. Note the different between subnet definition (all possible addresses that are valid for a given link)
and class definition (the actual dynamic range of addresses that is governed by the DHCPv6 server).
Subnet specification is currently optional, but much encouraged.
Of course it is possible to support both local and remote traffic. In such case, normal eth1 definition
should be present in the server configuration file. Also note that real (physical) interfaces should be
specified before logical ones.
# server.conf
iface relay1 {
relay eth1
// This relay1 actually represents a network that the

// physical device relays packets from. The relay device should
// have a global address assigned on its client-facing interface.
// That address should belong to this subnet.
subnet 2001:db8:1111::/48
// There are 3 ways of specifying interface-id content
// Way 1: As a text string

interface-id "relay1:en1"
// Way 2: As unsigned integer coded on 32 bits

// interface-id 5020
// Way 3: As a hex string

interface-id 0x427531264361332f3000001018680f980000
class {
pool 2001:db8:1111::1-2001:db8:1111::ff
}
option dns-server 2001:db8::100,2001:db8::101
}
iface relay2 {
relay eth1
// This relay1 actually represents a network that the

// physical device relays packets from. The relay device should
// have a global address assigned on its client-facing interface.
// That address should belong to this subnet.
subnet 2001:db8:2222::/48
// There are 3 ways of specifying interface-id content
// Way 1: As a text string

interface-id "relay1:en2"
// Way 2: As unsigned integer coded on 32 bits

// interface-id 5020
// Way 3: As a hex string

interface-id 0x427531264361332f3000001018680f980000
class {
pool 2001:db8:2222::1-2001:db8:2222::ff
}
option dns-server 2001:db8::100,2001:db8::101
}
5.3.9 Example 9: Cascade 2 relays

This is an advanced configuration. It assumes that client sends data to relay1, which encapsulates
it and forwards it to relay2, which eventually sends it to the server (after additional encapsulation). It
assumes that first relay adds interface-id option set to 6011 and second one adds similar option set to
6021. For details about relays in general and cascade setup in particular, see section 4.2. Also see section
7.4.4 for corresponding relays configuration.
# server.conf
iface relay1
{
relay eth0
interface-id 6011
}
iface relay2
{
relay relay1
interface-id 6021
T1 1000
T2 2000
class {
pool 6020::20-6020::ff
}
}
5.3.10 Example 10: Dynamic DNS (FQDN)

Support for Dynamic DNS Updates was added in version 0.5.0RC1. To configure it on the server
side, list of available names usually should be defined. Each name can be reserved for a certain address
or DUID. When no reservation is specified, it will available to everyone, i.e. the first client asks for
FQDN will get this name. In following example, name ’zebuline.example.com’ is reserved for address
2001::db8:1::1, kael.example.com is reserved for 2001:db8:1::2 and test.example.com is reserved for client
using DUID 00:01:00:00:43:ce:25:b4:00:13:d4:02:4b:f5.
Also note that is required to define, which side can perform updates. This is done using single number
after ,,option fqdn” phrase. Server can perform two kinds of DNS Updates: AAAA (forward resolving,
i.e. name to address) and PTR (reverse resolving, i.e. address to name). To configure server to execute
both updates, specify 2. This is a default behavior. If this value will be skipped, server will attempt
to perform both updates. When 1 will be specified, server will update PTR record only and will leave
updating AAAA record to the client. When this value is set to 0, server will not perform any updates.
The last parameter (64 in the following example) is a prefix length of the reverse domain supported
by the DNS server, i.e. if this is set to 64, and 2000::/64 addresses are used, DNS server must support
0.0.0.0.0.0.0.0.0.0.0.0.0.0.2.ip6.arpa. zone.
There are several additional parameters that affect DNS Update mechanism. ddns-protocol specifies
protocol that should be used for communication with DNS server. Allowed values are udp, tcp or any.
“Any” will try to use UDP and if that fails, it will revert to TCP. Second parameter is ddns-timeout
that specifies maximum time allowed for DNS server to respond before assuming communication failure.
It is specified in milliseconds.
The next useful parameter is fqdn-ddns-address that specifies address of DNS server that updates
should be performed to. If it is not specified, first DNS address from option dns-server will be used.
The last important parameter is +accept-unknown-fqdn. In a simplest scenario, server is configured
with a list of allowed names. Connecting clients may get only those names. That is convenient case, but
it is often not feasible to deploy it in a real network. In real networks, clients usually send out their own
names, rather than wait for server to assign them one. In such cases, it is somewhat expected that server
will not have complete list of all possible names that clients may send. Thus sooner or later server will
likely receive a fqdn name that is unknown to him. This parameter specifies server behavior in such case.
There are 5 possible policied currently supported. Each one is identified with an integer between 0 and
4. 0 means that uknown names should be rejected. That policy is useful for strictly controlled networks. 1
means that other available name from list of possible names should be sent instead. This is a compromise
between strict control over names and liberal acceptance of clients’ names. Policy 2 accept any name that
client will send. Names will be sanity checked. Note that mobile and nomadic clients may send names
from their home networks. That may be a problem if server attempts to update AAAA records as its
DNS server will probably only accept AAAA updates for locally administered domains. As a solution to
this problem, policy 3 was implemented. It takes client’s hostname (without its domain name), appends
local domain name and uses such constructed fully qualified domain name. For example, if client sends
nomad.faraway.org while visiting example.org, with this policy in place, nomad.example.org will be
assigned.
The last policy is useful for larger networks. Instead of accepting clients’ ideas about their host-
names, dedicated name is generated based on assigned address. For example, client that received
2001:db8:1:0:c7a8:e81:c500:46ce address in domain example.org will be assigned a 2001-db8-1-0-c7a8-e81-c50
name.
Following config file is a good starting point for tweaking DNS Update enabled server configuration.
# server.conf
# Logging level range: 1(Emergency)-8(Debug)

#
log-level 8
# Don’t log full date

log-mode short
# Set protocol to one of the following values: udp, tcp, any

ddns-protocol udp
# Sets DDNS Update timeout (in ms)

ddns-timeout 1000
# specify address of DNS server to be used for DDNS

fqdn-ddns-address 2001::1
iface "eth0" {
# assign addresses from this class

class {
pool 2001:db8:1::/64
}
# provide DNS server location to the clients

# also server will use this address to perform DNS Update,
# so it must be valid and DNS server must accept DNS Updates.
# provide their domain name

# provide fully qualified domain names for clients

# note that first, second and third entry is reserved
# for a specific address or a DUID
option fqdn 1 64
zebuline.example.com - 2000::1,
kael.example.com - 2000::2,
wash.example.com - 0x0001000043ce25b40013d4024bf5,
zoe.example.com,
malcolm.example.com,
kaylee.example.com,
jayne.example.com,
inara.example.com
# specify what to do with client’s names that are not on the list
# 0 - reject
# 1 - send other name from allowed list
# 2 - accept any name client sends
# 3 - accept any name client sends, but append specified domain suffix
# 4 - ignore client’s hint, generate name based on his address, append domain name
accept-unknown-fqdn 4 example.org
5.3.11 Example 11: Vendor-specific Information option

It is possible to configure dibbler-server to provide vendor-specific information options. Since there
are no dibbler-specific parameters, this implementation is quite flexible. Enterprise number as well as
content of the option itself can be configured. In the following example, we define a suboption 1027 for
vendor-id 4491. The value of that option is 0x0013.
# server.conf
log-level 8
log-mode precise
iface "eth1" {
class {
pool 2000::1-2000::ff
}
# Vendor-spec option can be specified as a hexstring

# String format syntax is also allowed

# option vendor-spec 4491-1028-"this-is-a-string"
# The third format supported is IPv6 address

# option vendor-spec 4491-1029-2001:db8::1
# Several vendor-specific options can be coma separated

option vendor-spec 4491-2-0xfedc,
4491-1027-0a:0b:0c:0d,
4491-1029-2001:db8::1
}
In some rare cases, several different options for different vendors may be specifed. In the folloging
example 2 different values are defined, depending on which vendor client will specify in SOLICIT or
REQUEST message. If client will only mention that it is interested in any vendor specific into (i.e. did
not sent vendor-spec info option, but only mentioned in in option request option, it will receive first vendor
option defined (in the following example, that would be a 1234 and 0002fedc).
# server.conf
log-level 8
log-mode precise
iface "eth1" {
class {
pool 2000::1-2000::ff
}
option vendor-spec 4491-1027-0x0013,1234-5678-0xabcd

}
5.3.12 Example 12: Per client configuration

Usually all clients receive the same configuration options, e.g. all clients will use the same DNS
server. However, it is possible to specify that particular clients should receive different options than
others. Following example set DNS server to 2000::1, domain to example.com and vendor specific in-
formation for vendor 5678. However, if requesting client has DUID 00:01:02:03:04:05:06:07:08, it will
receive different parameters (second.client.biz domain, 1234::5678:abcd as a DNS server and finally differ-
ent vendor-specific information). Also client with DUID 0x0001000044e8ef3400085404a324 will receives
normal domain and DNS server, but different (vendor=2) vendor specific information. See section 4.18
for background information. Since 0.8.0RC1, also addresses can be reserverd in this way.
Addresses reserved for special clients may be inside or outside of specified pools. If leases are outside of
specified pools, timers (t1, t2, prefered and valid lifetimes are set to the default values). It is currently not
possible to specify separate timers (t1, t2, preferred or valid lifetimes) on a per-client basis. If reservations
are out of pool, timers applicated to the interface will be used. See second example in this section.
Note that per client reservation was significantly refactored after 0.8.2, so its stability is not yet
confirmed.
Warning: Reservation by link-local address is not always reliable and DUID reservations should be
used instead, if possible. Typically for directly connected clients (i.e. without using relays) server can
obtain client’s link local address. For relayed traffic, that information is recorded by the first (closest to
the client) relay and stored in RELAY-FORW, so in most cases this should work. However, there are
two cases when link-local address reservation will fail. The first one is when client is using IPv6 privacy
extensions ([19]). With privacy extensions enabled, client may use its temporary random addresses that
are changing over time. Another case when such reservation could fail is when server allows unicast traffic.
Once clients get their addresses, they can send RENEW messages over unicast, using their global address
as a source address. In such case link-local address will not ever be used in the transmission, so the server
can get confused.
# server.conf
# Example server configuration file: per-client configuaration
#
# In this example, some clients receive different parameters than others.
# Logging level range: 1(Emergency)-8(Debug)

#
log-level 8
# Don’t log full date

log-mode short
iface "eth0" {
class {
pool 2001:db8:1::/64
}
pd-class {
pd-pool 2001:db8:2::/48
pd-length 64
}
# common configuration options, provided for all clients

option dns-server 2001:db8:1::1

option vendor-spec 5678-2-0xaaaa,1234-3-0x0102
# special parameters for client with DUID 00:01:02:03:04:06

client duid 00:01:02:03:04:06
{
address 2001:db8:1::123
prefix 2001:db8:abcd::/64
option domain second.client.biz
option dns-server 2001:db8::5678:abcd
option vendor-spec 5678-2-0xbbbb, 1234-5-0x222222
}
# this client should receive default domain and dns-server,

# but different vendor-spec info
# Both DUID forms are accepted (0x1234... and 12:34...)
client duid 0x0001000044e8ef3400085404a324
{
}
# Parameters can be reserved based on remote-id option

client remote-id 5-0x01020304
{
address 2001:db8:1::0102:0304
option domain our.special.remoteid.client.org
}
# Parameters can be reserved based on link-local address

client link-local fe80::1:2:3:4
{
option domain link.local.detected.interop.test.com
}
}
The following example shows out of pool reservation. Regular clients will get addresses from the
2001:db8:123::/64 pool. However, the client with DUID 00:01:00:0a:0b:0c:0d:0e:0f will get an 2002::babe
address that does not belong to any configured pool. That particular client with get parameters from the
interface on which this exception was defined. In this discussed example, what will be t1=1000, t2=2000,
preferred-lifetime=3000 and valid-lifetime=4000.
iface eth0 {
t1 1000
t2 2000
preferred-lifetime 3000
valid-lifetime 4000
class { pool 2001:db8:123::/64 }
client duid 00:01:00:0a:0b:0c:0d:0e:0f {
address 2002::babe
}
5.3.13 Example 13: Prefix delegation

Prefix delegation works quite similar to normal address granting. Administrator defines pool and
server provides prefixes from that pool. Before using prefix delegation, please read section 4.1. Client
configuration example is described in section 6.8.11.
# server.conf
log-mode precise
iface "eth0" {
# the following lines instruct server to grant each client

# prefix for this pool. For example, client might get
# 2222:2222:2222:2222:2222:993f::/96
pd-class {
pd-pool 2222:2222:2222:2222:2222::/80
pd-length 96
T1 11111
T2 22222
}
5.3.14 Example 14: Multiple prefixes

It is possible to define more than one pool, so each client will receive several prefixes. It is necessary to
define each pool with the same length, i.e. it is not possible to mix different pool lengths. See section 4.1
for prefix delegation background information. Client configuration example is described in section 6.8.11.
# server.conf
log-mode precise
iface "eth0" {
T1 1800
T2 2700
valid-lifetime 7200
# provide addresses from this pool

class {
pool 5000::/48
}
# the following lines instruct server to grant each client

# 2 prefixes. For example, client might get
# 2222:2222:2222:2222:2222:993f:6485::/96 and
# 1111:1111:1111:1111:1111:993f:6485::/96
pd-class {
pd-pool 2222:2222:2222:2222:2222::/80
pd-pool 1111:1111:1111:1111:1111::/80
pd-length 96
T1 11111
T2 22222
}
5.3.15 Example 15: Inactive mode

See sections 6.8.13 and 4.20 for inactive mode explanation. The same behavior has been added for
server.
#server.conf
log-level 8
inactive-mode
iface "eth0" {
class {
pool 2000::/64
}
}
5.3.16 Example 16: Leasequery

A separate entity called requestor can send queries regarding assigned addresses and prefixes. Server
can be configured to support such lease queries. See section 4.13 for detailed explanation.
#server.conf
log-level 8
iface "eth0" {
accept-leasequery
class {
pool 2000::/64
}
}
5.3.17 Example 17: Dibbler Authentication

It is possible to configure server to require Dibbler authentication. In this example, HMAC-SHA-512
will be used as an authentication method. Key Generation Nonce will have 64 bytes.
# server.conf
auth-methods digest-hmac-sha512
auth-required 1
iface eth0 {
class {
pool 2000::100-2000::10f
}
}
5.3.18 Example 18: Relay support with unknown interface-id

To get more informations about relay configuration, see section 4.2. In pervious examples (5.3.8,
5.3.9) it was assumed that interface-id set by relay is known. However, in some cases that is not true. If
sysadmin wants to accept relayed messages from any relay, there is a feature called guess mode. It tries
to match any relay defined in server.conf instead of exactly checking interface-id value.
Since there is only one relay defined, it will be used, regardless of the interface-id value (or even lack
of thereof).
# server.conf
guess-mode
iface relay1 {
relay eth1
interface-id 5020
class {
pool 2000::1-2000::ff
}
option dns-server 2000::100,2000::101
}
5.3.19 Example 19: DS-Lite tunnel (AFTR)

Server is able to provide Dual-Stack lite configuration for clients. Both address and name based
configurations are supported:
iface "eth0" {
class {
pool 2001:db8::/64
}
option ds-lite 2001:db8:1::ffff

option ds-lite sc.example.org
}
5.3.20 Example 20: Custom options

Server may be configured to also provide custom options to the clients. See Section 4.5 for details.
iface "eth0" {
class {
pool 2001:db8::/64
}
option 145 duid 01:02:a3:b4:c5:dd:ea
}
5.3.21 Example 21: Remote Autoconfiguration

Server does support experimental extension called remote autoconfiguration, as defined in [31]. See
Section 4.24.3 for details and configuration examples.
5.3.22 Example 21: Subnet declaration

Typically a network has only a subset of all its addresses managed by the DHCPv6 server. Let’s assume
that we have a network that uses 2001:db8::/64 prefix, so there may be 264 addresses in it. Usually the
server has information about a dynamic range of addresses that it can manage. Let’s assume that this
pool is defined as 2001:db8::1 to 2001:db8::ff, so the server is responsible for only only 256 addresses.
There are two cases where it is useful for the server to know the whole subnet, even though it manages
only small subset of it. The first case is subnet selection. When receiving a packet from relay, there
are several mechanisms that server uses to find appropriate subnet. One of them is matching based on
link-addr field that was set up by the relay. As DHCP relays themselves are never configured with DHCP,
the address used by the relay will always be outside of dynamic pool that the server manages, but within
the whole subnet that is defined on a given link.
The second case is useful when the server tries to respond to CONFIRM message. CONFIRM has a
special meaning. It is a question whether specified addresses are topologically correct in a given place,
not whether the server has bindings for them. If there are 2 servers on a given link and each of them
manages their own pool, both can potentially respond to CONFIRM to any clients, even those served by
the other server. The server can only do so if they have information about available subnet.
#server.conf
iface eth0 {
# This is the prefix used on a given link

subnet 2001:db8:1::/64
class {
# This is the small part of that prefix that is

# managed by the DHCPv6 server
pool 2001:db8:1::/96
}
}
6 Client configuration
This section describes Dibbler server, relay and client configuration. Square brackets denotes optional
values: mandatory [optional]. Alternative is marked as |. A | B means A or B. Parsers are case-insensitive,
so Iface, IfAcE, iface and IFACE mean the same. This does not apply to interface names. eth0 and ETH0
are dwo diffrent interfaces.
6.1 Data types

Config file parsing is token-based. Token can be considered a keyword or a specific phrase. Here are
more commonly used types:
IPv6 address – IPv6 address, e.g. 2000:db8:1::dead:beef
32-bit decimal integer – string containing only numbers, e.g. 12345
string – string of arbitrary characters enclosed in single or double quotes, e.g. ’this is a string’. If string
contains only a-z, A-Z and 0-9 characters, quotes can be omited, e.g. beeblebrox
DUID identifier – hex number starting with 0x, e.g. 0x12abcd. In Dibbler version 0.8.0RC1, another
format was introduced: 2 hex digits separated by colon, e.g. 12:aa:bb:cc:d5. As this format may
in some cases be confused with IPv6 address, the old format (starting with 0x) remains to be
supported.
IPv6 address list – IPv6 addresses separated with commas, e.g. 2001:db8:1::face:b00c, fe80::abcd:1234,::1
DUID list – DUIDs separated with commas, e.g. 0x0123456,0x0789abcd
string list – strings separated with comas, e.g. tealc,jackson,carter,oneill
boolean – YES, NO, TRUE, FALSE, 0 or 1. Each of them can be used, when user is expected to enable
or disable specific option.
6.2 Scopes
There are four scopes, in which options can be specified: global, inteface, IA and address. Every
option is specific for one scope. Each option is only applied to a scope and all subscopes in which it is
defined. For example, T1 is defined for IA scope. If there are several interfaces and each has several
address classes, repeating the same T1 value many time may be a bit awkward. Therefore parameters
can be also used in more common scopes. In this case – in interface or global. Defining T1 in interface
scope means: ,,for this interface the default T1 value is ...”. The same applies to global scope. Options
can be used multiple times. In that case value defined later is used.
Global scope is the largest. It covers the whole config file and applies to all intefaces, IAs, and
addresses, unless some lower scope options override it. Next scope is inteface. Options defined there are
inteface-specific and apply to a given interface, all IAs in this interface and addresses in those IAs. Next
is IA scope. Options defined there are IA-specific and apply to this IA and to addresses it contains. The
narrowest scope is address or prefix.
6.2.1 Interface declaration

Each system interface, which should be configured, must be mentioned in the configureation file.
Interfaces can be declared with this syntax:
{
interface-options
IA-options
address-options
}
or
iface interface-number
{
interface-options
IA-options
address-options
}
In the latter case, interface-number denotes interface index (or ifindex). It can be obtained from ip~l
(Linux), ipv6~if (Windows) or sometime ifconfig on other systems. interface-name is an interface
name. Name of the interface does not have to be enclosed in single or double quotes. It is necessary only
in Windows systems, where interface names sometimes contain spaces, e.g. ”local network connection”.
Interface scoped options can be used here. IA-scoped as well as address scoped options can also be used.
They will be treated as a default values for future definitions of the IA and address instantations.
6.2.2 IA declaration
IA is an acronym for Identity Association. It is a logical entity representing address or addresses used
to perform some functions. It is not suitable for prefixes (see Section 6.2.4). IA-options can be defined,
e.g. T1. IPv6 addresses can be defined here. All those values will be used as hints for a server. Almost
always, each DHCPv6 client will have exactly one IA on each interface. IA is declared using following
syntax:
ia [iaid]
{
IA-options
address-options
address-declaration
}
IAID is an optional number, which describes identifier of given IA. If not specified, it will be automat-
ically assigned integer numbers starting with 1. There may be more than one IA defined on an interface.
IA and PD (see Section 6.2.4) may be used together.
6.2.3 TA declaration
TA (Temporary Address) is a mechanism very similar to IA that allows configuration of temporary
addresses. See Section 6.2.2.
6.2.4 PD declaration
PD (Prefix Delegation) is a mechanism that allows leasing prefixes in similar way as addresses. For
details, see [8]. PD in client config file causes client to send IA PD option. This option informs server
that client is requesting prefix delegation.
pd [iaid]
{
pd-options
prefix-declaration
}
IAID is an optional number, which identifies this particular PD. If not specified, it will be automatically
assigned integer numbers starting with 1. There may be more than one PD defined on an interface. IA
(see Section 6.2.2) may be used together.
6.2.5 Address declaration

When IA is defined, it is sometimes useful to define its address. Its value will be used as a hint for
the server. Address is declared in the following way:
address [number]
{
address-options
address-declaration
}
where number is optional and denotes how many addresses with those values should be requested. If it
is diffrent than 1, then IPv6 address options are not allowed. Only address scoped options can be used
here. Address can be defined only within IA scope.
6.2.6 Prefix declaration

When PD is defined, it is sometimes useful to define its prefix. Its value will be used as a hint for the
server. Prefix is declared in the following way:
prefix [number]
{
prefix-options
prefix
}
6.3 Stateless configuration

If interface does not contain IA or TA keywords, client will ask for one address (one IA with one address
request will be sent). If client should not request any addresses on this interface, stateless 6 keyword must
be used. In such circumstances, only specified options will be requested.
6.4 Relay support

Usage of the relays is not visible from the client’s point of view: Client can’t detect if it communicates
via relay(s) or directly with the server. Therefore no special directives on the client side are required to
use relays. See section 4.2 for details related to relay deployment.
6
In the version 0.2.1-RC1 and earlier, this directive was called no-ia. This depreciated name is valid for now, but might
be removed in future releases.
6.5 Comments
Comments are allowed in configuration files. All common comment styles are supported:
• C++ style one-line comments: // this is comment
• C style multi-line comments: /* this is multi-line comment */
• script style one-line comments: # this is one-line comment
6.6 File location

Client configuration file should be named client.conf. It should be placed in the /etc/dibbler/
directory (Linux system) or in the current directory (Windows systems). One of design requirements for
client was ,,out of the box” usage. To achieve this, simply use empty client.conf file. Client will try
to get one address for each up and running interface. More preciselt client tries to configure each up,
multicast-capable and running interface, which has link address at least 6 bytes long. This usually means
that client running in auto-detection mode will not configure tunnels, which usually have IPv4 address (4
bytes long) as their link address. It should configure all wired (Ethernet) and wireless (802.11) interfaces,
though.
6.7 Client Reference

This section contains complete list of parameters that are allowed in client configuration file.
Dibbler client supports multiple parameters. Some ore defined by the base DHCPv6 specification (a
RFC 3315 document [5]), e.g. IA address option. Other parameters are for definition of one of multitude of
extensions that were defined for DHCPv6 protocol (e.g. prefix delegation option). Finally, there are many
configuration parameters that are not options in DHCPv6 sense, but rather affect the way the software
operates (e.g. log level). All those parameters may be defined in client config file. Every statement has
defined scope. See Section 6.2 for details. In many cases, parameters may also be defined in scopes larger
than its default scope. For example, instead of configuring DNS server option on 3 interfaces, it can be
defined once in global scope.
iface – (scope: global). Takes one parameter that can be either string (interface name) or integer
(interface index). Defines that client should perform some actions on specified interface. Exact
operations are defined within interface scope. Can optionally take second string parameter with the
only allowed value of “no-config”. If it is present, it instructs client to not perform any operation
on said interface. See Section 6.2.1.
ia – (scope: interface). Defines IA NA (Identity Association for Non-temporary Addresses), often abbre-
viated as IA. That is a container for “regular” (non-temporary in DHCPv6 nomenclature) addresses.
Simply saying, this is a client’s request for a single normal address. There may be more than one
ia defined on one interface. In such case, client will request several addresses. It may have one
optional integer parameter that defines unique indentifier (IAID). If followed by curly brackets, it
will create new IA scope. See Section 6.2.2.
ta – (scope: interface). Defines IA TA (Indentity Association for Temporary Addresses), often abbrevi-
ated as TA. This is a container for temporary addresses. Simply saying, this is a client’s request for
a temporary address. If followed by curly brackets, it will create new IA scope, similar to IA. See
Section 6.2.2. Note that TA scope accepts only limited set of parameters (e.g. iaid).
pd – (scope: interface). Defines IA PD (Identity Association for Prefix Delegation), often abbreviated
as PD. This is a container for prefixes. Simply saying, this is client’s request for prefix delegation.
It may have one optional integer parameter that defines unique indentifier (IAID). If followed by
curly brackets, it will create new PD scope. See Sections 6.2.4 and 4.1.
address – (scope: ia or ta). Defines an IPv6 address. It is usually defined in IA or TA to specify that
an address should be sent. It takes one optional parameter that defines multiplier. For example, to
define 3 addresses, following syntax may be used:
address 3 { }
If followed by curly brackets, creates new address scope. See Section 6.2.5.
prefix – (scope: pd). Defined an IPv6 prefix. It is defined in PD to specify that a prefix should be
sent. May be empty (prefix) or accompanied with prefix definition that consists of IPv6 address
followed by slash and prefix length (e.g. prefix 2001:db8:1::/56. If followed by curly brackets,
creates new prefix scope. See Section 6.2.6.
work-dir – (scope: global). Takes one string parameter. Defines working directory.
downlink-prefix-ifaces – (scope: global). Takes coma separated list of network interfaces. When client
receives prefix from upstream router, it attempts to split it into remaining interfaces. It works in
most cases, but if there are strange interfaces or specific requirements, this auto-selection mechanism
can be disabled and list of downlink interfaces can be explicitly specified. This command is used
for that purpose. See 4.1 and 6.8.11.
log-level – (scope: global). Takes one integer parameter. Defines verbose level of the log messages. The
valid range is from 1 (very quiet) to 8 (very verbose). Those values are modelled after levels used
in syslog. These are: 1(Emergency), 2(Alert), 3(Critical), 4(Error), 5 (Warning), 6(Notice), 7(Info)
and 8(Debug). Currently Dibbler is using levels 3 to 8, as 1 and 2 are reserved for system wide
emergency events.
log-name – (scope: global). Takes one string parameter. Defines than name, which will be used during
logging.
log-mode – (scope: global). Takes one parameter that can be short, full, precise or syslog. Defines
logging mode. In the default, full mode, name, date and time in the h:m:s format will be printed. In
short mode, only minutes and seconds will be printed (this mode is useful on terminals with limited
width). Precise mode logs information with seconds and microsecond precision. It is a useful as
a performance diagnostic tool for finding bottlenecks in the DHCPv6 autoconfiguration process.
Syslog works under POSIX systems (Linux, Mac OS X, BSD family) and allows default POSIX
logging functions.
log-colors – (scope: global). Takes one boolean parameter. Defines if logs printed to console should use
colors. That feature is used to enhance logs readability. As it makes the log files messy on systems
that do not support colors, it is disabled by default. The default is off.
strict-rfc-no-routing – (scope: global) Takes one boolean parameter. The default value is 0. During
normal operation, DHCPv6 client should add IPv6 address only (i.e. configure it with /128 prefix),
without configuring routing. Routing is expected to be configured with Router Advertisements
[17]. Please see discussion in bug 222 for detailed discussion about that behavior. Note that
Dibbler versions between 0.5.0RC1 and 1.0.0RC1 used to configure addressed with arbitrarily chosen
(guessed) prefix length of /64. Although it was convenient for users, as in most cases the guess
was correct and clients connected to the same link could ping each other immediately, its correct
operation was based on the assumption that the guess is correct. If it isn’t, tricky to debug problems
will appear. Hosts will incorrectly assume that some off-the link hosts are on link (or vice versa)
and will attempt to reach them directly. If you really understand the repercussions and still willing
to use that old behavior, you can use strict-rfc-no-routing 0. Author recommends against that,
though.
obey-ra-bits – (scope: global). Rounter Advertisements contain two bits that inform what kind of
DHCPv6 services are available on link. M (Managed) that tells that addresses and prefixes can
¯
be obtained using stateful DHCPv6. O (OtherConf) tells that other configuration options may be
¯
configured. Both bits are defined in [17], section 4.2. It should be noted that those bits are informa-
tional only. In the default mode (when obey-ra-bits is absent), the client will ask for configuration
options that it has specified in the configuration file. With obey-ra-bits, the client will wait till it
receives the RA message and will act according to the received bits. The default is off (obey-ra-bits
missing). Enabling obey-ra-bits implies inactive-mode.
experimental – (scope: global). Allows enabling experimental features. There are some highly-experimental
features present in Dibbler. To make a clear statement about their experimental nature, user is re-
quired to acknowledge that fact by putting this statement in its config file. This statement may be
present or absent. The default is absent.
addr-params – (scope: IA). Allows configuration of additional sub-option conveyed in IAADDR. It

supplements the usual information about an address received from a server with prefix length. For
example, if client received address 2001:db8:1::abcd and addr-params option contains 64, Dibbler
client will configure prefix 2001:db8:1::/64 on the interface that was used to communicate with
server. This is experimental feature, not defined in any standard or draft. Requires experimental
statement. See Section 4.24.2.
bind-to-address – If specified, it describes the address the server will bind to. This is almost never
needed, unless you have more than one link-local address and one to bind to one specific.
remote-autoconf – (scope: interface). Defines that remote autoconfiguration should be performed on

a given interface. This is experimental feature, so it requires experimental statement. See Section
4.24.3.
ddns-protocol – (scope: global). Takes one string parameter. Defines protocol that should be used
during DNS Update mechanism. Allowed values are tcp, udp and any. Any means that UDP will
be tried first and if it fails, update will be retried over TCP. See Section 4.6.
ddns-timeout – (scope: global). Takes one integer parameter that specifies timeout in milliseconds.
Defines how long client should wait for DNS server response during DNS Update before declaring
update a failure. See Section 4.6.
script – (scope: global). Takes one string parameter that specifies script name. When dibbler client
receives some options it normally sets them up in the system on its own. However, besides of
setting up all parameters directly, dibbler client can execute external script. See Section 4.8 for
details.
stateless – (scope: global). It may be present or missing. The default is missing. Defines that client
should run in stateless mode. In this mode only configuration parameters are defined, not addresses
or prefixes. It is mutually exclusive with ia, ta and pd. See Section 4.14. No-ia, an alias to that
command used to be supported, but due to misleading name its support was dropped in 0.8.1RC1.
anonymous-inf-request – (scope: global). When running in a stateless mode, client does not ask
for addresses or prefixes, but rather requests some general options. By default, it sends its client
identifier (DUID) to the server. However, it is possible to omit this identifier, so the INF-REQUEST
messages will be anonymous. This global option causes client to act in such anonymous way.
inactive-mode – (scope: global). This parameter may be present or absent. The default is absent.
Normally (with inactive-mode disabled) client tries to bind all interfaces defined in configuration
file. If such attempt fails, client reports an error and gives up. In some cases it is possible that
interface is not ready yet, e.g. WLAN interface did not complete association. It is possible to modify
client behavior, so it will accept downed and not running interfaces. To do so, inactive-mode must
be enabled. In this mode, client will accept inactive interfaces, will add them to inactive list and
will periodically monitor its state. When the interface finally becomes available, client will try to
configure it. See section 4.20 for details.
insist-mode – (scope: global). Client can be instructed to obtain several configuration options, like
DNS server configuration or domain name. It is possible that server will not provide all requested
options. Older versions of the dibbler client had been very aggressive in such case. It tried very
hard to obtain options that user specified in config file. To do so, it did send INF-REQUEST to
obtain such option. This behavior has changed. Right now when client does not receive all requested
options, it will complain, but will take no action. To enable old behavior, so called insist-mode has
been added. Insist-mode will also affect the way addresses are requested. If address was specified
in config file, client will request it in REQUEST message, rather than sending address offered by
server in ADVERTISE as it is typically done. See Section 4.21 for details.
skip-confirm – (scope: global). Support for CONFIRM messages was added in 0.8.0RC1. With it,
client may send CONFIRM when link state change is detected (e.g. switching to possibly new WiFi
access-point or replugging Ethernet cable). Client also sends CONFIRM after restart if there are
still valid leases found in locally stored databased. skip-confirm will disable any actions that would
result in CONFIRM transmissions. In particular, link state will not be detected and client will
ignore its previous address during startup.
duid-type – (scope: global). Takes one parameter. Allowed values are DUID-LLT, DUID-LL or DUID-
EN. The default is DUID-LLT. This parameter defines, what type of DUID should be generated if
there is no DUID already present. If there is a file containing DUID, this directive has no effect.
DUID-LLT means that DUID will be based on link layer address as well as time. DUID-LL means
that only link layer address will be used. The last value – DUID-EN – Enterprise Number-based
generation has a slightly different syntax:
duid-type duid-en enterprise-number enterprise-id.
For example: duid-type duid-en 1234 0x6789abcd means that enterprise number is set to 1234
and unique number from that company’s pool is 67:89:ab:cd (hexadecimal value of arbitrary length).
See section 4.22 for details.
option fqdn-s – (scope: global). Takes one boolean parameter and has the default value of 1. The S
bit is used in FQDN option. It is used to negotiate, which side (server or client) wants to perform
DNS Update procedure. See [16] for details. In general, if you don’t know that this option does,
you don’t want to modify this.
option fqdn – (scope: interface). Takes optional domain name as parameter. This option instructs
client to send FQDN option. This option has 2 purposes. The first one is to negotiated or request
Fully Qualified Domain Name for this client. The second one is to negotiate, who (client or server)
should perform DNS Update. If optional parameter is specified, it will be sent in the FQDN option.
Otherwise FQDN will be sent with empty name. This option is defined in [16]. See Section 4.6 for
details.
rapid-commit – (scope: interface). Takes one boolean parameter. The default is 0. This option allows
rapid commit procedure to be performed. Note that enabling rapid-commit on the client side is not
enough. server must be configured to allow rapid commit, too.
unicast – (scope: interface). Takes one boolean parameter. The default value is 0. This option specifies
if client should request unicast communication from the server. If server is configured to allow it,
it will add unicast option to its replies. It will allow client to communicate with server via unicast
addresses instead of usual multicast.
preferred-servers – (scope: interface). Takes list of addresses or list of DUIDs. The default value is
empty. This list defines, which servers are preferred. When client sends SOLICIT message, all
servers available in the local network will respond. When client receives multiple ADVERTISE
messages, it will choose those sent by servers mentioned on the preferred-server list. Otherwise the
best server will be chosen. Note that this parameter used to be spelled differently (single R). Old
spelling is still supported, but is deprecated and will be removed soon.
reject-servers – (scope: interface). Takes list of addresses or list of DUIDs. The default value is empty.
This list defines which server must be ignored. It has opposite meaning to the prefered-servers list.
Servers mentioned here will never be chosen.
vendor-spec – (scope: interface). This option allow requesting for a vendor specific configuration option
or options. Although there are no vendor-specific (i.e. specific for Dibbler) parameters, it can be
used to test some other DHCPv6 server implementations. It takes coma separated list following
tokens: type (integer), type(integer) – enterprise(number). It allows definition of just vendor-specific
option and/or vendor-specific option with specific enterprise. See feature description in Section 4.19.
T1 – (scope: IA or PD) Takes one parameter that specifies T1 hint sent to a server. The default value
is 232 − 1 and is expressed in seconds. T1 value assigned by server defines after what time client
should start renew process. This is only a hint for the server. Actual value will be provided by the
server.
T2 – (scope: IA or PD). Takes one parameter that specifies T2 hint sent to a server. The default value
is default:232 − 1 and is expressed in seconds. This value defines hint for T2. T2 assigned by server
defined after what time client will start emergency rebind procedure if renew process fails. This is
only a hint for the server. Actual value will be provided by the server.
valid-lifetime – (scope:address or prefix). Takes one integer parameter that defined requested valid
lifetime for address or prefix. The default value is 232 − 1. This parameter is expressed in seconds.
This parameter defines valid lifetime of an address. It will be used as a hint for a server, when the
client sends REQUEST messages.
preferred-lifetime – (scope:address or prefix). type: integer, default:232 − 1) This parameter defines

prefered lifetime of an address. It will be used as a hint for a server, when there client sends
REQUEST messages. Note that this parameter used to be spelled differently (single R). Old
spelling is still supported, but is deprecated and will be removed soon.
option dns-server – (scope: interface). Takes optional address list as parameter. This option conveys
information about DNS servers available. After retriving this information, client will be able to
resolve domain names into IP (both IPv4 and IPv6) addresses. If address list is not specified, client
will just request dns-server option from a server. If it is specified, listed addresses will be sent to
server as hints. Defined in [7].
option domain – (scope: interface). Takes optional domain list as parameter. This option is used
for retriving domain or domains names, which the client is connected in. For example, if client’s
hostname is alice.mylab.example.com and it wants to contact bob.mylab.example.com it can
simply refer to it as bob. Without domain name configured, it would have to use full domain name.
If optional domain list if defined, it will be sent to server as a hint. Defined in [7].
option ntp-server – (scope: interface). Takes optional address list as parameter. This option defines
information about available NTP servers. Network Time Protocol [1] is a protocol used for time
synchronisation, so all hosts in the network has the same proper time set. If address list is specified,
it will be sent to server as a hint. Defined in [14].
option time-zone – (scope: interface). Optionally takes one string parameter that specifies requested
timezone. It is possible to retrieve timezone from the server. If client is interested in this information,
it should ask for this option. Note that this option is considered obsolete as it is mentioned in draft
version only [32]. Work on this draft seems to be abandoned as similar functionality is provided in
now standard [14]. Unfortunately, it is not supported in Dibbler yet.
option sip-server – (scope: interface). Takes optional address list as parameter. Session Initiation
Protocol (SIP) [4] is a control protocol for creating, modifying, and terminating sessions with one
or more participants. These sessions include Internet telephone calls, multimedia distribution, and
multimedia conferences, including VoIP phones. If address list is specified, it will be sent as a hint
for a server. Format of this option is defined in [6].
option sip-domain – (scope: interface). Takes optional list of domains. It is possible to define domain
names for Session Initiation Protocol [4]. Configuration of this parameter will ease usage of domain
names in the SIP protocol. If domain list is specified, it will be sent as a hint for a server. Format
of this option is defined in [6].
option nis-server – (scope: interface). Takes optional address list as parameter. Network Information
Service (NIS) is a Unix-based system designed to use common login and user information on multiple
systems, e.g. universities, where students can log on to ther accounts from any host. To use this
functionality, a host needs information about NIS server’s address. This can be retrieved with this
option. If address list is specified, it will be sent as a hint for a server. Its format is defined in [11].
option nis-domain – (scope: interface). Takes optional list of domains. Network Information Service
(NIS) can albo specify domain names. It can be configured with this option. If domain list is
specified, it will be sent as a hint for a server. It is defined in [11].
option nis+-server – (scope: interface). Takes optional address list as parameter. Network Information
Service Plus (NIS+) is an improved version of the NIS protocol. If address list is specified, it will
be sent as a hint for a server. This option is defined in [11].
option nis+-domain – (scope: interface). Takes optional list of domains. Similar to nis-domain, it
defines domains for NIS+. If domain list is specified, it will be sent as a hint for a server. This
option is defined in [11].
option lifetime – (scope: interface). This statement can be present or missing. The default is missing.
Base spec of the DHCPv6 protocol does offers way of refreshing addresses only, but not the options.
Lifetime defines, how often client would like to renew all its options. By default client will not send
such option, but it will accept it and act accordingly if the server sends it on its own. Format of
this option is defined in [13].
option aftr – (scope: interface). In networks that deploy Dual-Stack Lite architecture [23], client (B4)
needs to configure DS-Lite tunnel. Client may obtain information about AFTR (a remote tunnel
endpoint). This option conveys fully qualified domain name. This statement instructs client to
request such option. It is defined in defined in [24].
option – (scope: interface). There are number of options supported by Dibbler. There may be cases,
however, when user wants to specify its own options. Several syntaxes are supported:
option number - hexstring

option number address-keyword address
option number address-list
option number string-keyword string
option number address-keyword request-keyword
option number string request-keyword
option number address-list request-keyword
where number designates option number, address-keyword is word “address”, address is an IPv6
address, address-list is coma separated list of IPv6 addresses, string-keyword is a word “string” and
request-keyword is a word “request”. See Section 4.5.
auth-protocol – (scope: global, type: string, default: none). This is a crucial parameter that spec-
ifies which authorization/authentication protocol is used. Allowed values are: none, delayed,
reconfigure-key and dibbler. See section 4.17 for details.
auth-methods – (scope: global). Takes coma separated list of accepted authentication methods meth-
ods that client will accept from server. If this list is empty, any method will be accepted. The first
method on the list is the default one. Possible values are: none, digest-plain, digest-hmac-md5,
digest-hmac-sha1, digest-hmac-sha224, digest-hmac-sha256, digest-hmac-sha384, and digest-hmac-sha
auth-replay – (scope: global, type: string, default: none). Specifies which replay detection methods are
supported. Currently two values are implemented: none and monotonic.
auth-required – (scope: global, type: boolean, default: 0). This parameter specifies if the client is
required to authenticate itself. When set to 0, any client authentication failures (invalid signature
or lack of AUTH option) will result in a warning only. When set to 1, such messages will be dropped.
route – (scope: interface). Takes one boolean parameter that defines if routing information should
requested or not. The default value is false. See Section 4.4.
After receiving options values from a server, client stores values of those options in separate files in
the working directory (/var/lib/dibbler in POSIX systems (Linux, Mac OS X and BSD) and current
directory in Windows). File names start with the option word, e.g. option-dns-server. Dibbler client
can also call user defined script after parameters are assigned or removed. Dibbler client also sets DNS
servers and domain names on its own on most systems.
6.8 Client Configuration Examples

This subsection contains various examples of the most popular configurations. Several additional
examples are provided with the source code. Please download it and look at *.conf files.
6.8.1 Example 1: Default

In the most simple case, client configuration file can be empty. Client will try to assign one address
for every interface present in the system, except interfaces, which are:
• down (flag UP not set)
• loopback (flag LOOPBACK set)
• not running (flag RUNNING not set)

• not multicast capable (flag MULTICAST not set)
• have link-layer address less than 6 bytes long (this requirement should skip all tunnels and virtual
interfaces)
If you must use DHCPv6 on one of such interfaces (which is not recommended and such attempt
probably will fail), you must explicitly specify this interface in the configuration file.
6.8.2 Example 2: DNS

Configuration mentioned in previous subsection is a minimal one and in a real life will be used rarely.
The most common usage of the DHCPv6 protocol is to request for an address and DNS configuration.
Client configuration file achieving those goals is presented below:
# client.conf
log-mode short
log-level 7
iface eth0 {
ia
option dns-server
}
6.8.3 Example 3: Timeouts and specific address

Automatic configuration is being driven by several timers, which define, what action should be per-
formed at various intervals. Since all values are provided by the server, client can only define values, which
will be sent to a server as hints. Server might take them into consideration, but might also ignore them
completely. Following example shows how to ask for a specific address and provide hints for a server.
Client would like to get 2000::1:2:3 address, it would like to renew addresses once in 30 minutes (T1 timer
is set to 1800 seconds). Client also would like to have address, which is prefered for an hour and is valid
for 2 hours.
Note: The format has changed in 1.0.0RC2.
# client.conf
log-mode short
log-level 7
iface eth0 {
T1 1800
T2 2000
valid-lifetime 7200
ia {
address 2000::1:2:3
}
}
There are multiple ways in which addresses can be requested in ia. This syntax was implemented
more for completeness, rather than having practical utility. It is mentioned here for reference.
# client.conf
iface eth0 {
T1 1800
T2 2700
ia // Send just an empty IA
ia { // Send an IA with one (any, ::) address

address
}
ia { // Send an IA with five (any, ::) addresses

address 5
}
ia { // Send an IA with address 2001:db8::1

address 2001:db8::1
}
ia { // Send an IA with one (any, ::) address with specific parameters

address {
}
}
ia { // Send an IA with five (any, ::) addresses with specific parameters

address 5 {
}
}
ia { // Send an IA with address 2001:db8::1

address 2001:db8::1 {
}
}
}
6.8.4 Example 4: More than one address

Another example: client would like to obtain 2 addresses on wifi0 interface. They are necessary since
this particular interface name contains spaces. It is possible to do this in two ways. First is to sent 2
Identity Associations (IA for short). Identity Association is a nice name for a addresses container. This
appears to be a most common way of telling server that this client is interested in more than one address.
# client.conf
log-mode short
log-level 5
iface wifi0 {
ia
ia
}
Another way it to send one IA, but include two address hints in it. Server may take them into
consideration (dibbler server does), but some other DHCPv6 implementations may ignore those hints.
# client.conf
log-mode short
log-level 5
iface wifi0 {
ia {
address
address
}
}
6.8.5 Example 5: Quick configuration using Rapid-commit

Rapid-commit is a shortened exchange with server. It consists of only two messages, instead of the
usual four. It is worth to know that both sides (client and server) must also support rapid-commit to use
this fast configuration.
# client.conf
iface eth1 {
rapid-commit yes
ia
option dns-server
}
6.8.6 Example 6: Stateless mode

Client can be configured to work in a stateless mode. It means that it will obtain only some configu-
ration parameters, but no addresses. Let’s assume we want all the details stored in a log file and we want
to obtain all possible configuration parameters. Here is a configuration file:
# client.conf
log-level 8
log-mode full
stateless
iface eth0
{
option dns-server
option domain
option ntp-server
option time-zone
option sip-server
option sip-domain
option nis-server
option nis-domain
option nis+-server
option nis+-domain
}
6.8.7 Example 7: Dynamic DNS (FQDN)

Dibbler client is able to request fully qualified domain name, i.e. name, which is fully resolvable using
DNS. After receiving such name, it can perform DNS Update procedure. Client can ask for any name,
without any preferrence. Here is an example how to configure client to perform such task:
# client.conf
# Set protocol to one of the following values: udp, tcp, any

ddns-protocol udp
# Sets DDNS Update timeout (in ms)

ddns-timeout 800
# uncomment following line to force S bit to 0

# option fqdn-s 0
log-level 7
iface eth0 {
# ask for one address
ia
# ask for options

option dns-server
option domain
option fqdn
# ask for fully qualified domain name (any name will do)
option fqdn
# you can also provide hint for the server regarding preferred name
# option fqdn dexter.example.org
In this case, client will mention that it is interested in FQDN by using Option Request and empty
FQDN option, as specified in [16]. Server upon receiving such request (if it is configured to support it),
will provide FQDN option containing domain name. Depending on the server’s configuration, all DNS
Updates will be performed by the server, forward will be performed by client and reverse by the server,
or only forward will be done by a client.
It is also possible for client to provide its name as a hint for server. Server might take it into
consideration when it will choose a name for this client. To send specific hostname, additional parameter
(a string with a fully qualified domain name) should be specified.
Two additional parameters were introduced in Dibbler 0.8.1. ddns-protocol specifies protocol that
should be used for communication with DNS server. Allowed values are udp, tcp or any. “Any” will
try to use UDP and if that fails, it will revert to TCP. Second parameter is ddns-timeout that specifies
maximum time allowed for DNS server to respond before assuming communication failure. It is specified
in milliseconds.
Note that to successfully perform DNS Update, address must be assigned and dns server address must
be known. Therefoe “ia” and “option dns-server” are required for “option fqdn” to work properly. Also if
DHCPv6 server provides more than one DNS server address, update will be attempted only for the first
address on the list.
It is also possible to force S bit in the FQDN option to 0 or 1. See [16] for details regarding its
meaning.
6.8.8 Example 8: Interface indexes

Usually, interface names are referred to by names, e.g. eth0 or Local Area Connection. Every system
also provides unique number associated with each infterface, usually called ifindex or interface index. It
is possible to read the number using ip l command (Linux) or ipv6 ifx. Below is an example, which
demonstrate how to use interface indexes:
# client.conf
log-mode short
log-level 5
iface 5 {
ia
}
6.8.9 Example 9: Vendor-specific options

It is possible to configure dibbler-client to ask for a vendor specific options. Although there are no
dibbler-specific features to configure, it is possible to use this option to test other server implementations.
This option will rather be used by network engineers and power network admins, rather than normal end
users.
There are 3 ways to define, how dibbler-client can request vendor-specific options. First choice: It can
just ask for this option (only option request option will be sent). Second choice: it can ask for vendor-spec
option by adding such option with enterprise number set, but no actual data. Third choice: send this
option and include both enterprise number and actual data. In the following configuration file example,
uncomment appropriate line to obtain desired bahavior:
# client.conf
log-level 8
iface eth0 {
# ask for address
ia
# uncomment only one of the following lines:

option vendor-spec
# option vendor-spec 1234
# option vendor-spec 1234 5678
# To ask for multiple vendor-spec options, uncomment:

# option vendor-spec 123,456
}
Although that is almost never needed, it is possible to configure client to request multiple vendor-
specific options at the same time. That feature is mainly used as a test tool for the server. To use it,
uncomment last line in the example above.
6.8.10 Example 10: Unicast communication

Client would like to obtain an address on ,,Local Area Connection” interface. Note quotation marks
around interface name. They are necessary since this particular interface name contains spaces. Client
also would like to accept Unicast communication if server supports it. User wants all information to be
logged via Linux syslog daemon. Take note that you won’t be able see to what Dibbler is doing with such
low log-level. (Usually log-level should be set to 7, which is also a default value).
# client.conf
log-mode syslog
log-level 5
iface "Local Area Connection" {
unicast yes
ia
ia
}
6.8.11 Example 11: Prefix delegation

From the client’s point of view, configuration is quite simple. It is required to specify that this client is
interested in prefix delegation. See section 4.1 for background information related to prefix delegation and
sections 5.3.13 and 5.3.14 for details about server configuration. To ask for prefix delegation, emphprefix-
delegation (or pd ) should be used.
# client.conf
iface "eth0" {
ia // ask for address
pd // ask for prefix
}
It is possible to define additional parameters for a prefix:

# client.conf
iface eth0 {
pd {
t1 1000
t2 2000
}
}
Client (requesting router in PD nomenclature) receives prefix from upstream router and tries to auto-
select downstream interfaces. It tries to use interfaces that are up, running, multicast-capable, with MAC
address at least 6 bytes long and were not used to obtain prefix. If this algorithm does not work in your
case (e.g. because you want to use prefixes on other interfaces or you want some interfaces to be skipped),
it is possible to explicitly enumerate downstream interfaces using downlink-prefix-ifaces:
# client.conf
# received prefix will be split among following interfaces

downlink-prefix-ifaces eth1, eth5
# Ask for prefix over eth0

iface eth0 {
pd
}
If you do not want Dibbler to split the prefixes automatically, it is possible to do so by specifying
”none” as the interface name. Note that this will render PD mechanism useless, unless you also use a
script and do the delegated prefix processing on your own.
# client.conf
# Dibbler client should not split received prefixes on its own

downlink-prefix-ifaces "none"
# You need to provide your own script to handle prefixes

script "/var/lib/dibbler/client-pd-split.sh"
# Ask for prefix over eth0

iface eth0 {
pd
}
Prefix hints can be specified in the similar way as addresses (see 6.8.3, except that multiple prefixes
syntax is not supported.
# client.conf
log-level 8
iface eth0 {
T1 1800
T2 2700
pd // Send just an empty PD
pd { // Send a PD with one (any, ::/0) prefix

prefix
}
pd { // Send an PD with a specific prefix

prefix 2001:db8::1 / 64
}
pd { // Send an PD with one (any, ::/0) prefix with specific parameters

prefix {
}
}
pd { // Send an PD with a specific prefix and specific parameters

prefix 2001:db8::1 /64 {
}
}
}
6.8.12 Example 12: Insist mode

During normal operation, when client asks for an option, but does not receive it from the server, it
complain, but takes no action. To force client to insist (i.e. ask over and over again), so called insist
mode has been introduced. See section 4.21 for extended explanation.
insist-mode
iface "eth0" {
ia
option dns-server
option domain
option ntp-server
}
6.8.13 Example 13: Inactive mode

Usually client starts when network interfaces are operational. Normally downed or nonexisting in-
terfaces mentioned in the configuration file are considered misconfiguration and client refuses to start.
However, sometimes that is not the case, e.g. still waiting to be associated wireless interfaces. To al-
low operation in such circumstances, inactive mode has been added. See 4.20 for detailed explanation.
interfaces are spec
inactive-mode
iface "eth0" {
ia
}
6.8.14 Example 14: Dibbler Authentication

Authentication is enabled. Client will accept HMAC-SHA-512, HMAC-MD5 and HMAC-SHA-256 as
an authentication method.
# client.conf
log-mode short
log-level 7
auth-required 1
auth-methods digest-hmac-sha512, digest-hmac-md5, digest-hmac-sha256
iface eth0 {
}
6.8.15 Example 15: Skip Confirm

Client detects if previous client instance was not shutdown properly (due to power outage, client crash
or similar event). In such case, it reads existing address database and checks if assigned addresses may
still be valid. If that is so, it tries to confirm those addresses by using CONFIRM message.
If user don’t want CONFIRM message to be send and client should start ”from scratch” every time,
it is possible to disable confirm support.
# client.conf
log-mode short
log-level 7
skip-confirm
iface eth0 {
ia
}
6.8.16 Example 15: User-defined IAID

Sometimes it is useful to define specific IAID identifiers. That is rather uncommon, but possible. This
technique can be used for both addresses (IA NA options) and prefixes (IA PD options).
# client.conf
iface "eth0" {
ia 123
option dns-server
option domain
}
6.8.17 Example 16: DS-Lite tunnel (AFTR)

Server may provide information about AFTR (a Dual Stack Lite tunnel endpoint) to the clients, as
specified in [24].
iface "eth0" {
ia
option aftr # request name of the remote DS-Lite tunnel endpoint
}
6.8.18 Example 17: Custom options

Client is able to ask for custom options, that are not supported by default. Following config file allows
client to ask for many options. Also, see Section 4.5 for extended explanation. Note that the syntax
changed slightly after Dibbler 0.8.3 was released.
#client.conf
iface "eth0" {
ia
# This will send specified option value

option 145 hex 01:02:a3:b4:c5:dd:ea
# This will request specific options and interpret responses

option 149 hex

option 150 address
option 151 address-list
option 152 string
6.8.19 Example 18: Remote Autoconfiguration

Client is able to use experimental extension to ask for configuration remotely. See Section 4.24.3 for
details.
log-mode short
log-level 8
experimental
remote-autoconf
iface "eth0" {
ia
unicast 1
option dns-server
option domain
option nis-server
option nis-domain
option nis+-server
option nis+-domain
option time-zone
option lifetime
}
7 Relay configuration
Relay configuration is stored in relay.conf file in the /etc/dibbler/ directory (Linux systems) or
in current directory (Windows systems).
7.1 Global scope

Every option can be declared in global scope. Config file consists of global options and one or more
inteface definitions. Note that reasonable minimum is 2 interfaces, as defining only one would mean to
resend messages on the same interface.
7.2 Interface declaration

Interface can be declared this way:
{
interface options
}
or
iface number
{
interface options
}
where name of the interface denotes name of the interface and number denotes it’s number. It does
not need to be enclosed in single or double quotes (except windows cases, when interface name contains
spaces).
7.3 Options
Every option has a scope it can be used in, default value and sometimes allowed range.
log-level – (scope: global, type: integer, default: 7) Defines verbose level of the log messages. The valid
range if from 1 (Emergency) to 8 (Debug). The higher the logging level is set, the more messages
dibbler will print.
log-name – (scope: global, type: string, default: Client). Defines name, which should be used during
logging.
log-mode – (scope: global, type: short, full or precise, default value: full) Defines logging mode. In the
default, full mode, name, date and time in the h:m:s format will be printed. In short mode, only
minutes and seconds will be printed (this mode is useful on terminals with limited width). Recently
added precise mode logs information with seconds and microsecond precision. It is a useful for
finding bottlenecks in the DHCPv6 autoconfiguration process.
interface-id-order – (scope: global, type: before, after or omit, default: before) Defines placement of
the interface-id option. Options can be placed in the RELAY-FORW message is arbitrary order.
This option has been specified to control that order. interface-id option can be placed before or
after relay-message option. There is also possibility to instruct server to omit the interface-id option
altogether, but since this violates [5], it should not be used. In general, this configuration parameter
is only useful when dealing with buggy relays, which can’t handle all option orders properly. Consider
this parameter a debugging feature. Note: similar parameter is available in the dibbler-server.
client multicast – (scope: interface, type: boolean, default: false) This command instructs dibbler-relay
to listen on this particular interface for client messages sent to multicast (ff02::1:2) address.
client unicast – (scope: interface, type: address, default: not defined) This command instructs dibbler-
relay to listen to messages sent to a specific unicast address. This feature is usually used to connect
multiple relays together.
server multicast – (scope: interface, type: boolean, default: false) This command instructs dibbler-
relay to send messages (received on any interface) to the server multicast (ff05::1:3) address. Note
that this is not the same multicast address as the server usually listens to (ff02::1:2). Server must
be specifically configured to be able to receive relayed messages.
server unicast – (scope: interface, type: address, default: none) This command instructs dibbler-relay
to send message (received on any interface) to speficied unicast address. Server must be properly
configured to to be able to receive unicast traffic. See unicast command in the 5.3.4 section.
interface-id – (scope: interface, type: integer, default: none) This specifies identifier of a particular
interface. It is used to generate interface-id option, when relaying message to the server. This option
is then used by the server to detect, which interface the message originates from. It is essential to
have consistent interface-id defined on the relay side and server side. It is worth mentioning that
interface-id should be specified on the interface, which is used to receive messages from the clients,
not the one used to forward packets to server.
guess-mode – (scope: global, type: boolean, default: no) Switches relay into so called guess-mode.
Under normal operation, client sends messages, which are encapsulated and sent to the server.
During this encapsulation relay appends interface-id option and expects that server will use the
same interface-id option in its replies. Relay then uses those interface-id values to detect, which
the original request came from and sends reply to the same interface. Unfortunately, some servers
does not sent interface-id option. Normally in such case, dibbler-relay drops such server messages
as there is no easy way to determine where such messages should be relayed to. However, when
guess-mode is enabled, dibbler-relay tries to guess the destination interface. Luckily, it is often
trivial to guess as there are usually 2 interfaces: one connected to server and second connected to
the clients.
option remote-id – (scope: global, type: option, default: none) Tells the relay agent to insert remote-id
option. It is followed by a number (enterprise-id), a dash (“-”) and a hex string that specifies the
actual content of the remote-id option being inserted. Remote-id is specified in [15].
option relay-id – (scope: global, type: option, default: none) Tells the relay agent to insert relay-id
option. It takes one parameter, which specifies the actual content of the relay-id. The parameter
must be specified as a hex string.
option link-layer – (scope: global, type: option, default: none) Tells the relay agent to insert client
link-layer address option, as specified in [25]. It should be noted that the source MAC address
is extracted from incoming link-local (fe80:...) address or from client-id. Both methods are not
reliable and susceptible for spoofing. This was implemented mostly as a testing feature for the
server implementation.
7.4 Relay configuration examples

Relay configuration file is fairly simple. Relay forwards DHCPv6 messages between interfaces. Mes-
sages from client are encapsulated and forwarded as RELAY FORW messages. Replies from server are
received as RELAY REPL message. After decapsulation, they are being sent back to clients.
It is vital to inform server, where this relayed message was received. DHCPv6 does this using interface-
id option. This identifier must be unique. Otherwise relays will get confused when they will receive reply
from server. Note that this id does not need to be alligned with system interface id (ifindex). Think
about it as ”ethernet segment identifier” if you are using Ethernet network or as ”bss identifier” if you
are using 802.11 network.
If you are interested in additional examples, download source version and look at *.conf files.
7.4.1 Example 1: Simple

Let’s assume this case: relay has 2 interfaces: eth0 and eth1. Clients are located on the eth1 net-
work. Relay should receive data on that interface using well-known ALL DHCP RELAYS AND SERVER
multicast address (ff02::1:2). Note that all clients use multicast addresses by default. Packets received
on the eth1 should be forwarded on the eth0 interface, using multicast address. See section 5.3.8 for
corresponding server configuration.
# relay.conf
log-level 8
log-mode short
iface eth0 {
server multicast yes
}
iface eth1 {
client multicast yes
interface-id 5020
}
7.4.2 Example 2: Unicast/multicast

It is possible to use unicast addresses instead/besides of default multicast addresses. Following example
allows message reception from clients on the 2000::123 address. It is also possible to instruct relay to send
encapulated messages to the server using unicast addresess. This feature is configured in the next section
(7.4.3).
# relay.conf
log-level 8
log-mode short
iface eth0 {
}
iface eth1 {
client unicast 2000::123
interface-id 5020
}
7.4.3 Example 3: Multiple interfaces

Here is another example. This time messages should be forwarded from eth1 and eth3 to the eth0
interface (using multicast) and to the eth2 interface (using server’s global address 2000::546). Also clients
must use multicasts (the default approach):
# relay.conf
iface eth0 {
}
iface eth2 {
server unicast 2000::456
}
iface eth1 {
interface-id 1000
}
iface eth3 {
interface-id 1001
}
7.4.4 Example 4: 2 relays

Those two configuration files correspond to the ,,2 relays” example provided in section 5.3.9. See
section 4.2 for detailed exmplanations.
# relay.conf - relay 1
log-level 8
log-mode full
# messages will be forwarded on this interface using multicast

iface eth2 {
server multicast yes // relay messages on this interface to ff05::1:3
# server unicast 6000::10 // relay messages on this interface to this global address
}
iface eth1 {
# client multicast yes // bind ff02::1:2
client unicast 6011::1 // bind this address
interface-id 6011
}
# relay.conf - relay 2
iface eth0 {
# server multicast yes // relay messages on this interface to ff05::1:3
server unicast 6011::1 // relay messages on this interface to this global address
}
# client can send messages to multicast

# (or specific link-local addr) on this link
iface eth1 {
client multicast yes // bind ff02::1:2
# client unicast 6021::1 // bind this address
interface-id 6021
7.4.5 Example 5: Guess-mode

In the 0.6.0 release, a new feature called guess-mode has been added. When client sends some data
and relay forwards it to the server, it always adds interface-id option to specify, which link the data has
been originally received on. Server, when responding to such request, should include the same interface-id
option in the reply. However, in some poor implementations, server fails to do that. When relay receives
such poorly formed response from the server, it can’t decide which interface should be used to relay this
message.
Normally such packets are dropped. However, it is possible to switch relay into a guess-mode. It tries
to find any suitable interface, which it can forward data on. It is not very reliable, but sometimes it is
better than dropping the message altogether.
# relay.conf
log-level 8
log-mode short
guess-mode
iface eth0 {
}
iface eth1 {
interface-id 5020
}
7.4.6 Example 6: Relaying to multicast

During normal operation, relay sends forwarded messages to a All DHCP Servers (FF05::1:3) multi-
cast address.
Although author does not consider this an elegant solution, it is also possible to instruct relay to
forward message to a All DHCP Relay Agents and Servers (ff02::1:2) multicast address. That is quite
convenient when there are several relays connected in a cascade way (server – relay1 – relay2 – clients).
For details regarding DHCPv6-related multicast addresses and relay operation, see [5].
To achieve this behavior, server unicast can be used. Note that name of such parameter is a bit
misleading (“server unicast” used to specify multicast address). That parameter should be rather called
“destination address”, but to maintain backward compatibility, it has its current name.
# relay.conf
log-level 8
log-mode short
iface eth0 {
server unicast ff02::1:2
}
iface eth1 {
interface-id 5020
}
7.4.7 Example 7: Options inserted by the relay

Typically relay agent receives messages from clients, encapsulates them and sends towards the server.
The only option being added is interface-id option (if specified). However, in some cases it makes sense for
the relay agent to insert additional options. Dibbler relay supports several options that can be inserted:
remote-id – Remote-id option may identify the remote client. Dibbler implementation is very simple,
as it allows setting only one specific value, not unique value, one for each client. That makes this
feature useful for testing purposes, but its deployment in a production network is unlikely. Remote-
id takes two parameters. The first one is enterprise-id, a 32bit unique identifier that characterises a
vendor. The second parameter is arbitrary length hex string. Remote-id is specified in [15].
echo-request – In some cases, relay inserts options and would like the server to send them back in
its responses. Typically, those options are then processed by the relay to correctly send back to
the client. Dibbler relay does not need such options to operate, but is able to simulate them. It
is possible to specify one or more options that the relay requests the server to echo back. This
option takes a list of coma separated values that designate option codes. Dibbler relay will insert
options with specified option codes, but they will not carry any useful value. Echo Request Option
is specified in [20].
relay-id – Relay agent may be configured to insert interface-id option. However, that option identifies
an interface within each relay, but not necessarily the relay itself. In some cases it is useful for the
relay agent to insert an option that identifies itself. That is implemented as relay-id option. This is
particularly useful for bulk leasequery. This option is defined in [22].
link-layer – The initial DHCPv6 spec lacked information about client MAC addresses. In principle,
the server is able to discover client’s MAC address when receiving direct traffic. Unfortunately,
it can’t do that if the packet traverses a relay agent. That deficiency was addressed in [25]. It
defines an option that relays can insert when receiving incoming traffic from a client. Note that
this implementation is not perfect. Instead of getting that information from layer 2 directly (there
is no API to do that as of late 2014), it tries to get the information from client-id (DUID-LLT
or DUID-LL) or from message source address (if it is link-local and uses EUI-64). This is not a
bullet-proof solution, but it should work in most cases.
log-level 8
log-mode short
# Uncomment the following line to force relay to start including remote-id

# option with entreprise-id 5 and content of 01:02:03:04
option remote-id 5-0x01020304
# Uncomment the following line to force relay to send Echo Request Option
# asking the server to echo back options 100, 101 and 102
option echo-request 100,101,102
# Uncommenting this option will make relay to insert relay-id option

# into forwarded RELAY-FORW messages.
option relay-id aa:bb:cc:dd:ee:ff
# Uncomment this line to tell relay to attempt to insert client link-layer

# address option. Relay will attempt to get that info from client’s DUID
# or source IPv6 link-layer address. These are not 100% reliable methods!
option link-layer
# The relay should listen on eth1 interface for incoming client’s traffic.
# Clients by default send their traffic to multicast.
iface eth1 {
# When forwarding traffic from that interface, please add interface-id

# with value 5555, so the server will know where the clients are connected.
interface-id 5555
}
# This is a second interface. It is used to reach the server.

iface eth2 {
# Send message on this interface to the server multicast (ff05::1:3)

}
8 Requestor configuration
Requestor (entity used for leasequery) does not use configuration files. All parameters are specified
by command-line switches. See section 4.13 for details.
9 Frequently Asked Questions

Soon after initial Dibbler version was released, feedback from user regarding various things started to
appear. Some of the questions were common enough to get into this section.
9.1 Common Questions
Q: Why client does not configure routing after assigning addresses, so I cannot e.g. ping other hosts?
A: It’s a common misunderstanding. DHCPv4 provides many configuration parameters to host, with
default router address being one of them. Things are done differently in IPv6. Routing configuration is
supposed to be conveyed using Router Advertisements (RA) messages, announced periodically by routers.
Hosts are supposed to listen to those messages and configure their routing appropriately. Note that this
mechanism is completely separate from DHCPv6. It may sound a bit strange, but that’s the way it was
meant to work.
Properly implemented clients are supposed to configure leased address with /128 prefix and learn the
actual prefix from RA. As this is incovenient, many clients (with dibbler included) bend the rules and
configure received addresses with /64 prefix. Please note that this value is arbitrary chosen and may be
improper in many scenarios.
Note: This behaviour has changed in the 0.5.0 release. Previous releases configured received address
with /128 prefix. To restore old, more RFC conformant behavior, see strict-rfc-no-routing directive in
the Section 6.6.
Note: The original pre-0.5.0 behaviour is correct. This was reverted in 1.0.0RC2 release. The strict-
rfc-no-routing now takes one boolean parameter. Setting it to 1 (which is the default value) makes
dibbler to configure addresses with /128 prefix, as expected. Please see discussion in bug 222 for more
details. Setting it to 0 reverts to the behavior that Dibbler was offering between 0.5.0 and 1.0.0RC1, i.e.
configuring an address with /64 prefix. Please note that it was chosen (guessed) arbitrarily and in some
cases may be completely wrong. Use with caution!
There was a proposed solution in a form of [30] draft. See section 4.4. Unfortunately, MIF working
group in IETF decided to abandon this work.
Q: I would like to have the ability to reserve specific addresses for clients with given MAC address.
That’s a basic and very common feature in DHCPv4 server. Why it is not supported in Dibbler? Are
there plans to implement such feature?
A: No. It is not and will not be supported. For couple of reason. The first and most important is
that DHCPv6 identifies clients based on DUIDs, rathar than MAC addresses. That is a protocol design
choice. Of course that does not prevent many users from saying “I don’t care, I want MAC classification
anyway!”. So here are more technical reasons why MAC classification is a bad idea. The first technical
reason is that Dibbler couldn’t be extended, because MAC address is often not available. There are 3
possible ways a server could possibly learn client’s MAC address:
1. from Ethernet frame. That won’t work if traffic goes through relay
2. from DUID-LL or DUID-LLT. RFC3315 forbids looking into the DUID. Besides of being a wrong
thing to do, that also won’t work, because client with a given MAC address can use different DUID
type, e.g. DUID-EN or DUID-UUID (or others, I saw on the wire some device with DUID type 14.
Strange, uncommon, but valid).
3. using source address and extracting MAC from link-local address thaty is based on EUI-64 that
contains MAC address. That should be available for direct traffic (simply src address of the UDP
packet) or relayed (peer addr field in RELAY-FORW message). This would usually work, but there
are cases when it won’t. First, if client uses privacy extensions (RFC4941). The other one if client
and server support unicast, some traffic will be sent from client global address, not using link-local
address at all.
So instead of doing MAC based reservations, Dibbler supports link-layer address based reservations.
In most cases it will be equivalent to MAC reservations. The only case where it won’t work will be with
unicast, but that can be solved easily (don’t use link-layer reservations and unicast together). Despite this
shortcoming, link-layer was implemented after Dibbler 0.8.2 was released. See 4.18 and 5.3.12 sections.
Q: Dibbler server receives SOLICIT message, prints information about ADVERTISE/REPLY trans-
mission, but nothing is actually transmitted. Is this a bug?
A: Are you sure that your client is behaving properly and responds to Neighbor Discovery (ND)
requests? Before any IPv6 packet (that includes DHCPv6 message) is transmitted, recipient reachabity is
checked (using Neighbor Discovery protocol [17]). Server sends Neighbor Solicititation message and waits
for client’s Neighbor Advertisement. If that is not transmitted, even after 3 retries, server gives up and
doesn’t transmit IPv6 packet (DHCPv6 reply, that is) at all. Being not able to respond to the Neighbor
Discovery packets may indicate invalid client behavior.
Q: Dibbler sends some options which have values not recognized by the Ethereal/Wireshark or by
other implementations. What’s wrong?
A: DHCPv6 is a relatively new protocol and additional options are in a specification phase. It
means that until standarisation process is over, they do not have any officially assigned numbers. Once
standarization process is over (and RFC document is released), this option gets an official number.
There’s pretty good chance that different implementors may choose diffrent values for those not-yet
officialy accepted options. To change those values in Dibbler, you have to modify file misc/DHCPConst.h
and recompile server or client. See Developer’s Guide, section Option Values for details.
Q: I can’t get (insert your trouble causing feature here) to work. What’s wrong?
A: Go to the project homepage and browse list archives. If your problem was not reported before,
please don’t hesitate to write to the mailing list. Author prefers to not be contacted directly, but rather
over mailing list. The only exception are security reports and confidential disucssions. In such case, please
contact author directly.
Q: Why is feature X not implemented? I need it!
Q: The short answer is : We do accept patches.

The longer one is more complicated. Dibbler is a hobby project with the only developer having very
limited time to dedicate. There are many requests, bugs and missing features and I have to prioritize
them. My personal judgement about importance, difficulty, amount of work required and and other
factors of specific feature decides on priority, compared to other features. Personal preference plays a role
here as well.
If you don’t want to wait, get your hands dirty and implement it yourself! It is not as difficult as it
sounds. Dibbler code is reasonably well documented, so understanding how it works is not that difficult.
See Dibbler Developer’s Guide for introduction, code overview, architecture etc. You can always as on
dibbler-devel mailing list.
Finally, if you are disappointed with the pace of progress (or even lack of thereof), there are couple of
things you can do. First and foremost, consider alternatives. There is ISC DHCP implementation that
supports DHCPv6 http://www.isc.org. It is open source as well, but ISC provides paid support for it, if
you need one. ISC also does custom development contracts, should you need it. ISC is a nice non-profit
company, so your money will be used for a good cause.
9.2 Linux specific questions
Q: I can’t run client and server on the same host. What’s wrong?
A: First of all, running client and server on the same host is just plain meaningless, except testing
purposes only. There is a problem with sockets binding. To work around this problem, consult Developer’s
Guide, Tip section how to compile Dibbler with certain options.
Q: After enabling unicast communication, my client fails to send REQUEST messages. What’s wrong?
A: This is a problem with certain kernels. My limited test capabilites allowed me to conclude that
there’s problem with 2.4.20 kernel. Everything works fine with 2.6.0 with USAGI patches. Patched
kernels with enhanced IPv6 support can be downloaded from http://www.linux-ipv6.org/. Please let
me know if your kernel works or not.
9.3 Windows specific questions
Q: Dibbler doesn’t receive anything on Windows 7 or Windows 8. Is it broken?

A: Make sure your firewall allows the traffic through. Dibbler server must be able to receive traffic on
UDP port 547. Dibbler client must be able to receive traffic on UDP port 546. If DNS Update mechanism
is used, Dibbler must be able to send traffic to TCP and/or UDP port 53 (DNS). There are many ways
in which Windows firewall can be configured to allow such traffic. For example, in Windows 8, one can
use the following commands (assuming DNS server is located at 2001:db8:1::1”):
netsh -c advfirewall
> firewall
> add rule name="dhcpv6in" dir=in action=allow localport=547 protocol=udp
> add rule name="dhcpv6out" dir=out action=allow localport=547 protocol=udp
> add rule name="ddnsout" dir=out action=allow remoteip="2001:db8:1::1"
Q: After installing Advanced Networking Pack or Windows XP ServicePack2 my DHCPv6 (or other
IPv6 application) stopped working. Is Dibbler compatible with Windows XP SP2?
A: Both products (Advanced Networking Pack as well as Service Pack 2 for Windows XP) provide
IPv6 firewall. It is configured by default to reject all incoming IPv6 traffic. You have to disable this
firewall. To disable firewall on the “Local Area Connection” interface, issue following command in a
console:
netsh firewall set adapter "Local Area Connection" filter=disable
Q: Server or client refuses to create DUID. What’s wrong?

A: Make sure that you have at least one up and running interface with at least 6 bytes long MAC
address. Simple ethernet or WIFI card matches those requirements. Note that network cable must be
plugged (or in case of wifi card – associated with access point), otherwise interface is marked as down.
Q: Is Microsoft Windows 8 supported?

A: Unfortunately, Windows 8 is not supported yet. I do not have time to run tests on Windows8, but
if it provides the same API as previous versions do, there’s pretty good chance that Dibbler will work on
Windows 8.
10 Miscellaneous topics
10.1 History
Dibbler project was started as master thesis by Tomasz Mrugalski and Marek Senderski on Computer
Science faculty on Gdansk University of Technology. Both authors graduated in september 2003 and soon
after started their jobs.
During master thesis writing, it came to my attention that there are other DHCPv6 implementations
available, but none of them has been named properly. Refering to them was a bit silly: ,,DHCPv6
published on sourceforge.net has better support than DHCPv6 developed in KAME project, but our
DHCPv6 implementation...”. So I have decided that this implementation should have a name. Soon it
was named Dibbler after famous CMOT Dibbler from Discworld series by Terry Pratchett.
Sadly, Marek does not have enough free time to develop Dibbler, so his involvement is non-existent at
this time. However, that does not mean, that this project is abandoned. It is being actively developed by
me (Tomek). Keep in mind that I work at full time and do Ph.D. studies, so my free time is also greatly
limited.
10.2 Contact and reporting bugs

There is an website located at http://klub.com.pl/dhcpv6. If you belive you have found a bug,
please put it in Bugzilla – it is a bug tracking system located at http://klub.com.pl/bugzilla. If you
are not familiar with that kind of system, don’t worry. After simple registration, you will be asked for
system and Dibbler version you are using and other details. Without feedback from users, author will not
be aware of many bugs and so will not be able to fix them. That’s why users feedback is very important.
You can also send bug report to the mailing list if you don’t want to use bugzilla directly (or want to
confirm first that it is indeed a bug). Be sure to be as detailed as possible. Please include both server
and client log files, both config and xml files. If you are familiar with tcpdump or ethereal, traffic dumps
from this programs are also great help.
If you are not sure if your issue is a bug or a configuration problem, you may also want to browse
archives and ask on a mailing list. See following subsection for details.
If you have used Dibbler and it worked ok, this documentation answered all you question and every-
thing is in order (hmmm, wake up, it must be a dream, it isn’t reality:), also send a short note to the
mailing list.
Author keeps a list of places where Dibbler software is used. He would appreciate if you could check
if your country is on the list (see project website) and mention it if it isn’t. That’s completely optional
and author won’t be disappointed if you chose to not reveal that information.
Finally, while the author’s mail isn’t secret, please DO NOT send mails to him directly. He is quite
busy and do not want to respond to the same questions over and over again. Also, he travels a lot, so
often is unable to respons. It is much better to ask on the mailing list, which has public archive searchable
by Google. This could help other people who may have the same question. And even if they ask the
question without bothering to google for answers first, it will be easier for the author to respond with a
link to previous response. Finally, there’s currently over 100 people subscribed to the list, so there’s a
non-trivial chance that some of them will respond when author is not available.
Please constact author directly ONLY if you want to report security issue or want to discuss confi-
dential matters.
10.3 Mailing lists

There are two mailing lists related to the Dibbler project:
dibbler – Maling list for Dibbler users. It is used to ask for help, report bugs, hay hello and things like
that. If you are not sure, what to do, people on this list will try to help you. Web-inteface link:
http://klub.com.pl/cgi-bin/mailman/listinfo/dibbler
dibbler-devel – That list is intended as a way of communication between people, who are technically
involved in the dibbler development. If you are going to improve dibbler in any way, make sure
that you announce it here. You may get help. Also if you are trying to fix a bug on your own (hey,
that’s great!), this list is a good place to talk about it. Web-interface link: http://klub.com.pl/cgi-
bin/mailman/listinfo/dibbler-devel
Both lists have archives available on-line. You can join or leave one or both lists at any time using
convenient web-interface or using traditional mail-based approach.
10.4 Thanks and greetings

I would like to send my thanks and greetings to various persons. Without them, Dibbler would not
be where it is today. For a full list of contributors, see AUTHORS file.
Marek Senderski – He’s author of almost half of the Dibbler code. Without his efforts, Dibbler would
be simple, long forgotten by now master thesis.
Jozef Wozniak – My master thesis’ supervisor. He allowed me to see DHCP in a larger scope as part
of network provisioning process.
Jacek Swiatowiak – He’s my master thesis consultant. He guided Marek and me to take first steps
with DHCPv6 implementation.
Ania Szulc – Discworld fan and a great girl, too. She’s the one who helped me to decide how to name
this yet-untitled DHCPv6 implementation.
Christian Strauf – Without his queries and questions, Dibbler would be abandoned in late 2003.
Bartek Gajda – His interest convinced me that Dibbler is worth the effort to develop it further.
Artur Binczewski and Maciej Patelczyk – They both ensured that Dibbler is (and always will be)
GNU GPL software. Open source community is grateful.
Josep Sole – His mails (directly and indirectly) resulted in various fixes and speeded up of 0.2.0 release.
Sob – He has ported 0.4.0 back to Win2000 and NT. As a direct result, 0.4.1 was released for those
platforms, too.
Guy ”GMSoft” Martin – He has provided me with access to HPPA machine, so I was able to squish
some little/big endian bugs. He also uploaded ebuild to the Gentoo portage.
Bartosz ”fEnio” Fenski – He taught me how much work needs to be done, before deb packages are
considered ok. It took me some time to understand that more pain for the package developer means
less problems for the end user. Thanks to him, Dibbler is now part of the Debian GNU/Linux
distribution.
Adrien Clerc and his team – Their contribution of the DNS Updates code is most welcome.
Krzysztof Wnuk – He has fixed, improved and extended DNS Updates support as well as provided
initial support for prefix delegation.
Alain Durand – Thanks for the invitation to interop test session and for allowing me to see DHCPv6
issues in a much broader scope.
Petr Pisar – He has reported lots of bugs, and also often provides fixes. Thanks.
Paul Schauer – Thanks to his effors, Dibbler now works on Mac OS X. He did majority of the porting
work and then did numerous rounds of testing and debugging.
11 Acknowledgements
Author would like to acknowledge following projects and programmes that supported or continue to
support research and development of the Dibbler software and related activities.
This work has been partially supported by the Polish Ministry of Science and Higher Education un-
der the European Regional Development Fund, Grant No. POIG.01.01.02-00-045/09-00 Future Internet
Engineering.
The Dibbler project was created as master thesis at Department of Computer Communications, at Faculty
of Electronics, Telecommunications and Informations, at Gdansk University of Technology.
References
[1] Mills, D., “Simple Network Time Protocol (SNTP) Version 4 for IPv4, IPv6 and OSI”, RFC2030,
IETF, October 1996.
[2] P.Vixie, S.Thomson, Y.Rekhter, J.Bound, “Dynamic Updates in the Domain Name System (DNS
UPDATE)”, RFC2136, IETF, April 1997.
[3] P.Vixie, O.Gudmundsson, D.Eastlake and B.Wellington, “Secret Key Transaction Authentication for
DNS (TSIG)”, RFC2845, IETF, May 2000.
[4] J.Rosenberg and H. Schulzrinne, “Session Initiation Protocol (SIP): Locating SIP Servers”, RFC3263,
IETF, June 2002.
[5] R. Droms, Ed. “Dynamic Host Configuration Protocol for IPv6 (DHCPv6)”, RFC3315, IETF, July
2003.
[6] H. Schulzrinne, and B. Volz “Dynamic Host Configuration Protocol (DHCPv6) Options for Session
Initiation Protocol (SIP) Servers”, RFC3319, IETF, July 2003.
[7] S. Thomson, C. Huitema, V. Ksinant and M. Souissi “DNS Extensions to Support IP Version 6”,
RFC3596, IETF, October 2003.
[8] O. Troan, and R. Droms “IPv6 Prefix Options for Dynamic Host Configuration Protocol (DHCP)
version 6”, RFC3633, IETF, December 2003.
[9] R. Droms, Ed. “DNS Configuration options for Dynamic Host Configuration Protocol for IPv6
(DHCPv6)”, RFC3646, IETF, December 2003.
[10] R. Droms, “Stateless Dynamic Host Configuration Protocol (DHCP) Service for IPv6”, RFC3736,
IETF, April 2004.
[11] V. Kalusivalingam “Network Information Service (NIS) Configuration Options for Dynamic Host
Configuration Protocol for IPv6 (DHCPv6)”, RFC3898, IETF, October 2004.
[12] R. Arends, R. Austein, M. Larson, D. Massey and S. Rose “DNS Security Introduction and Require-
ments”, RFC4033, IETF, March 2005
[13] S. Venaas, T. Chown, and B. Volz “Information Refresh Time Option for DHCPv6”, RFC4242,
IETF, Nov. 2005.
[14] V. Kalusivalingam “Simple Network Time Protocol (SNTP) Configuration Option for DHCPv6”,
RFC4075, IETF, May 2005.
[15] B. Volz “DHCPv6 Relay Agent Remote-ID Option”, RFC4649, IETF, August 2006
[16] M. Stapp and B.Volz “The Dynamic Host Configuration Protocol for IPv6 (DHCPv6) Client Fully
Qualified Domain Name (FQDN) Option”, RFC4704, IETF, October 2006
[17] T.Narten, E.Nordmark, W.Simpson and H.Soliman, “Neighbor Discovery for IP version 6 (IPv6)”,
RFC4861, IETF, September 2007
[18] S.Thomson, T.Narten, T.Jinmei, “IPv6 Stateless Address Autoconfiguration”, RFC4862, IETF,
September 2007
[19] T. Narten, R. Draves, S. Krishnan, “Privacy Extensions for Stateless Address Autoconfiguration in
IPv6”, RFC4941, IETF, September 2007
[20] S. Zeng, B. Volz, K. Kinnear, J. Brzozowski, “DHCPv6 Relay Agent Echo Request Option”,
RFC4994, IETF, September 2007
[21] J. Brzozowski, K. Kinnear, B. Volz and S. Zeng “DHCPv6 Leasequery”, RFC5007, IETF, September
2007
[22] M. Stapp, “DHCPv6 Bulk Leasequery”, RFC5460, IETF, February 2009
[23] A.Durand, R.Droms, J.Woodyatt, Y.Lee, “Dual-Stack Lite Broadband Deployments Following IPv4
Exhaustion”, RFC6333, IETF, Aug. 2011
[24] D.Hankins, T.Mrugalski, “Dynamic Host Configuration Protocol for IPv6 (DHCPv6) Options for
Dual-Stack Lite”, RFC6334, IETF, Aug. 2011
[25] G.Halwasia, S.Bhandari, W.Dec, “Client Link-Layer Address Option in DHCPv6”, RFC6939, IETF,
May 2013
[26] Vishnu Ram, Saumya Upadhyaya, Nitin Jain “Authentication, Authorization and key management
for DHCPv6”, work in progress (expired), draft-ram-dhc-dhcpv6-aakey-01, IETF, August 2006
[27] T. Mrugalski, “Optimization of the autoconfiguration mechanisms of the mobile stations supporting
IPv6 protocol in the IEEE 802.16 environmentdfdf”, Ph.D dissertation, Gdańsk, Oct. 2009
[28] T. Mrugalski, J.Wozniak, K.Nowicki, “Remote DHCPv6 Autoconfiguration for Mobile IPv6 nodes”,
IEEE 14th International Telecommunications Network Strategy and Planning Symposium, Warsaw,
Poland, Sept. 2010
[29] T.Mrugalski, J.Wozniak, K.Nowicki, “Remote Stateful Autoconfiguration for Mobile IPv6 Nodes
with Server Side Duplicate Address Detection”, IEEE, Australasin Telecommunication Networks and
Applications Conference, Auckland, New Zealand, Nov. 2010
[30] W.Dec, T.Mrugalski, T.Sun, B. Sarikaya, “DHCPv6 Route Options”, MIF WG, work in progress,
draft-ietf-mif-dhcpv6-route-option-03, IETF, Sep. 2011 .
[31] T.Mrugalski, “Remote DHCPv6 Autoconfiguration”, work-in-progress (expired), IETF, July 2010
[32] A.K. Vijayabhaskar “Time Configuration Options for DHCPv6”, work in progress (expired), draft-
ietf-dhc-dhcpv6-opt-timeconfig-03, IETF, October 2003

Dibbler - A Portable Dhcpv6 User'S Guide: Tomek Mrugalski

Uploaded by

Copyright:

Available Formats

Dibbler - A Portable Dhcpv6 User'S Guide: Tomek Mrugalski

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dibbler - A Portable Dhcpv6 User'S Guide: Tomek Mrugalski

Uploaded by

Copyright:

Available Formats

Dibbler – a portable DHCPv6

2 Installation and usage 11

4.16 XML files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.3.14 Example 14: Multiple prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

7.4.1 Example 1: Simple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

8 Requestor configuration 103

9 Frequently Asked Questions 104

10 Miscellaneous topics 108

Figure 1: General DHCPv6 operation

Figure 2: Several clients supported by one server

Figure 3: Redundancy: several servers

1.2 Supported parameters

1.3 Not supported features

• Conflict resolution in DNS Updates is not supported.

1.4 Operating System Requirements

1.5 Supported platforms

2 Installation and usage

2.1 Linux installation

OpenSUSE – use standard installation mechanism.

Gentoo Linux – use emerge to install dibbler (e.g. emerge dibbler).

To install Dibbler in Gentoo systems, just type:

2.2 Windows installation

2.3 Mac OS X installation

2.4 FreeBSD, NetBSD, OpenBSD, Solaris 11

2.5 Basic usage

To see, if client, server or relay are running, type:

To see full list of available commands, type dibbler-server, dibbler-client or dibbler-relay

3.1 Linux/Mac OS X/FreeBSD/NetBSD/OpenBSD/Solaris Compilation

See Dibbler Developer’s Guide, section 2 for details on compilation switches.

3.2 Modern Windows (XP...Win7) compilation

3.3 Legacy Windows (NT/2000) compilation

3.4 IPv6 support

3.4.1 Setting up IPv6 in Linux

3.4.2 Setting up IPv6 in Windows Vista and Win7

3.4.3 Setting up IPv6 in Windows XP and 2003

3.4.4 Setting up IPv6 in Windows 2000

1. Download the file tpipv6-001205.exe from: http://msdn.microsoft.com/downloads/sdks/platform/

7. Restart the computer when prompted.

10. Click Install.

3.4.5 Setting up IPv6 in Windows NT4

1. Download the file msripv6-bin-1-4.exe from: http://research.microsoft.com/msripv6/msripv6.

5. IPv6 is now installed.

4.1 Prefix delegation

Figure 4: Prefix delegation (router behaviour)

Figure 5: Relay deployment

4.3 Address and prefix assignment policy

1. Client classification is performed (a class is assigned to a client)

Figure 6: Cascade relays

4.4 Routing configuration

# router with a single route with infinite lifetime

# a single next-hop without any routes defined (i.e. default router)

# router with 3 routes defined in different ways

# prefixes available on link directly, not via router

Support on client’s side is enabled in a very simple way:

# Example client configuration file: client-route.conf

# 7 = omit debug messages

# Uncomment this line to run script every time response is received

4.5 Custom options