The stateless autoconfiguration process uses one of many features of the IPv6 Neighbor Discovery Protocol (NDP) to discover the prefix used on the LAN. NDP performs many functions for IPv6, all related to something that occurs between two hosts in the same subnet. For example, one part of NDP replaces the IPv4 ARP protocol. IPv4 ARP allows devices on the same subnet—neighbors—to learn each other's MAC address. Because this and many other activities occur only inside the local subnet between neighbors on the same link, IPv6 collected these basic functions into one protocol suite, called NDP.
Stateless autoconfiguration uses two NDP messages, namely router solicitation (RS) and router advertisement (RA) messages, to discover the IPv6 prefix used on a LAN. The host sends the RS message as an IPv6 multicast message, asking all routers to respond to the questions "What IPv6 prefix(s) is used on this subnet?" and "What is the IPv6 address(s) of any default routers on this subnet?" Figure 17-8 shows the general idea, on subnet 1 from Figure 17-6, with PC1 sending an RS, and router R1 replying with the IPv6 prefix used on the LAN and R1's own IPv6 address as a potential default router.
Example NDP RS/RA Process to Find the Default Routers
NOTE - IPv6 allows multiple prefixes and multiple default routers to be listed in the RA message; the figure just shows one of each for simplicity's sake.
IPv6 does not use broadcasts. In fact, there is no such thing as a subnet broadcast address, a network-wide broadcast address, or an equivalent of the all-hosts 255.255.255.255 broadcast IPv4 address. Instead, IPv6 uses multicast addresses. By using a different multicast IPv6 address for different functions, a computer that has no need to participate in a particular function can simply ignore those particular multicasts, reducing the impact to the host. For example, the RS message only needs to be received and processed by routers, so the RS message's destination IP address is FF02::2, which is the address reserved in IPv6 to be used only by IPv6 routers. RA messages are sent to a multicast address intended for use by all IPv6 hosts on the link (FF02::1), so not only will the host that sent the RS learn the information, but all other hosts on the link will also learn the details.
Table 17-5 summarizes some of the key details about the RS/RA messages.
Table 17-5 Details of the RS/RA Process
Message | RS | RA |
Multicast destination | FF02::2 | FF02::1 |
Meaning of multicast address | All routers on this link | All IPv6 nodes on this link |
IPv6 Address Configuration Summary
This chapter covers four methods for assigning IPv6 addresses to hosts or router interfaces. Two variations use static configuration, while two dynamically learn the address. However, with both static and dynamic configuration, two alternatives exist—one that supplies the entire IPv6 address and one that allows the host to calculate the EUI-64 interface ID. Table 17-6 summarizes the configuration methods.
Table 17-6 IPv6 Address Configuration Options
Static or Dynamic | Option | Portion Configured or Learned |
Static | Do not use EUI-64 | Entire 128-bit address |
Static | Use EUI-64 | Just the /64 prefix |
Dynamic | Stateful DHCPv6 | Entire 128-bit address |
Dynamic | Stateless autoconfiguration | Just the /64 prefix |
Discovering the Default Router with NDP
In IPv4, hosts discover their default router (default gateway) either through static configuration on the host or, more typically, with DHCP. IPv6 can use both of these same options as well, plus the NDP RS/RA messages as explained in the previous section. The NDP router discovery process occurs by default on IPv6 hosts and routers, so while the stateful DHCPv6 server can supply the IP address(es) of the possible default routers, it is perfectly reasonable in IPv6 to simply not bother to configure these details in a stateful DHCP server, allowing the built-in NDP RS/RA messages to be used instead.
The default router discovery process is relatively simple. Routers automatically send RA messages on a periodic basis. These messages list not only the sending router's IPv6 address but also all the known routers on that subnet. A host can wait for the next periodic RA message or request that all local routers send an RA immediately by soliciting the routers using the RS message.
Learning the IP Address(es) of DNS Servers
Like IPv4 hosts, IPv6 hosts typically need to know the IP address of one or more DNS servers to resolve names into the corresponding IP address. Oftentimes, the host also needs to learn the DNS domain name to use. And like IPv4 hosts, IPv6 hosts can be told these IP addresses using (stateful) DHCP. When a host (or router for that matter) learns its IPv6 address using stateful DHCP, the host can also learn the DNS server IP addresses and the domain name, taking care of this particular detail.
Stateless DHCP, which is most useful in conjunction with stateless autoconfiguration, is an alternative method for finding the DNS server IP addresses and the domain name. A host that uses stateless autoconfiguration can learn its IPv6 address and prefix automatically, as well as learn its default router IP address, in both cases using NDP RS/RA messages. However, the stateless autoconfiguration process does not help a host learn the DNS IP addresses and domain name. So, stateless DHCP supplies that information using the same messages as stateful DHCP. However , to supply this information, the server does not need to track any state information about each client, so a stateless DHCP server can be used.
Table 17-7 summarizes some of the key features of stateful and stateless DHCPv6.
Table 17-7 Comparison of Stateless and Stateful DHCPv6 Services
Feature | Stateful DHCP | Stateless DHCP |
Remembers IPv6 address (state information) of clients that make requests | Yes | No |
Assigns IPv6 address to client | Yes | No |
Supplies useful information, like DNS server IP addresses | Yes | Yes |
Is most useful in conjunction with stateless autoconfiguration | No | Yes |
IPv6 Addresses
This chapter has already introduced the concepts behind the general format of IPv6 addresses, the ideas behind global unicast IPv6 addresses, and some details about multicast IPv6 addresses. The following sections round out the coverage of addressing, specifically the three categories of IPv6 address:
Unicast: IP addresses assigned to a single interface for the purpose of allowing that one host to send and receive data.
Multicast: IP addresses that represent a dynamic group of hosts for the purpose of sending packets to all current members of the group. Some multicast addresses are used for special purposes, like with NDP messages, while most support end-user applications.
Anycast: A design choice by which servers that support the same function can use the same unicast IP address, with packets sent by clients being forwarded to the nearest server, allowing load balancing across different servers.
Unicast IPv6 Addresses
IPv6 supports three main classes of unicast addresses. One of these classes, global unicast IP addresses, closely matches the purpose of IPv4 public IP addresses. Global unicast addresses are assigned by ICANN and the RIRs for the purpose of allowing globally unique IPv6 addresses for all hosts. These addresses come from inside the 2000::/3 prefix, which includes all addresses that begin with 2 or 3 (hex).
The next class of IPv6 unicast addresses covered here, unique local unicast addresses, have the same function as IPv4 RFC 1918 private addresses. In IPv4, most every enterprise, and most every Internet-connected small or home office, uses IPv4 private networks. Unique local unicast addresses begin with hex FD (FD00::/8), with the format shown in Figure 17-9.
NOTE - The original IPv6 RFCs defined a private address class called site local, meaning local within a site (organization). The original site local address class has been deprecated and replaced with unique local unicast addresses.
Unique Local Address Format
To use these addresses, an enterprise engineer would choose a 40-bit global ID in a pseudorandom manner, with the goal that hopefully the addresses will be unique in the universe. In reality, pseudorandom is probably a number made up by the engineer. The 16-bit subnet field and 64-bit interface ID work just like with global unicast addresses, numbering different subnets and hosts and allowing EUI-64 assignment of the interface ID. As usual, the engineer could avoid using EUI-64, using easier-to-remember values like 0000:0000:0000:0001 as the interface ID.
Link local addresses are the third class of unicast IPv6 addresses covered here. IPv4 has no concepts like the link local IP address. IPv6 uses these addresses when sending packets over the local subnet; routers never forward packets destined for link local addresses to other subnets.
Link local addresses can be useful for functions that do not need to leave the subnet, in particular because a host can automatically derive its own link local IP address without sending packets over the subnet. So, before sending the first packets, the host can calculate its own link local address so that the host has an IPv6 address to use when doing its first overhead messages. For example, before a host sends an NDP RS (router solicitation) message, the host will have already calculated its link local address. The host uses its link local address as the source IP address in the RS message.
Link local addresses come from the FE80::/10 range, meaning all addresses that begin with FE80, FE90, FEA0, and FEB0. No specific configuration is required, because a host forms these addresses by using the first 10 bits of hex FE80 (binary 1111111010), 54 more binary 0s, and the last 64 bits being the host's EUI-64 format interface ID. Figure 17-10 shows the format.
Link Local Address Format
Routers also use link local addresses on each interface enabled to support IPv6. Like hosts, routers automatically calculate their link local IP addresses. In fact, Example 17-1 earlier in this chapter listed the (R1) router's link local IP addresses in the output of the show ipv6 interface command output. Interestingly, routers normally use link local addresses as the next-hop IP address in IPv6 routes, rather than the neighboring router's global unicast or unique local unicast address.
Multicast and Other Special IPv6 Addresses
Multicast addresses can be used to communicate to dynamic groupings of hosts, with the sender sending a single packet and with the network replicating that packet as needed so that all hosts listening for packets sent to that multicast address receive a copy of the packet. IPv6 can limit the scope of where routers forward multicasts based on the value in the first quartet of the address. This book only examines multicasts that should stay on a local link; these addresses all begin with FF02::/16, so they are easily recognized.
For reference, Table 17-8 lists some of the more commonly seen IPv6 multicast addresses. Of particular interest are the addresses chosen for use by Routing Information Protocol (RIP), Open Shortest Path First (OSPF), and Enhanced IGRP (EIGRP), which somewhat mirror the multicast addresses each protocol uses for IPv4.
Table 17-8 Common Link Local Multicast Addresses
Purpose | IPv6 Address | IPv4 Equivalent |
All IP nodes on the link | FF02::1 | Subnet broadcast address |
All routers on the link | FF02::2 | N/A |
OSPF messages | FF02::5, FF02::6 | 224.0.0.5, 224.0.0.6 |
RIP-2 messages | FF02::9 | 224.0.0.9 |
EIGRP messages | FF02::A | 224.0.0.10 |
DHCP relay agents (routers that forward to the DHCP server) | FF02:1:2 | N/A |
Before completing the discussion of IPv6 addressing, you should know about a couple of special IPv6 addresses. First, IPv6 supports the concept of a loopback IP address, as follows:
::1 (127 binary 0s and a 1)
Just like the IPv4 127.0.0.1 loopback address, this address can be used to test a host's software. A packet sent by a host to this address goes down the protocol stack, and then right back up the stack, with no communication with the underlying network card. This allows testing of the software on a host, particularly when testing new applications.
The other special address is the :: address (all binary 0s). This address represents the unknown address, which hosts can use when sending packets in an effort to discover their IP addresses.
Summary of IP Protocols and Addressing
This chapter has covered a lot of concepts and details about IPv6 addresses, many of which require some work to remember or memorize. This short section pulls several concepts from throughout this major section on IPv6 protocols and addresses together before moving on to some details about routing protocols and router configuration.
When an IPv6 host first boots, it needs to do several tasks before it can send packets through a router to another host. When using one of the two methods of dynamically learning an IPv6 address that can be used to send packets past the local routers to the rest of a network, the first few initialization steps are the same, with some differences in the later steps. The following list summarizes the steps a host takes when first booting, at least for the functions covered in this chapter:
Step 1 The host calculates its IPv6 link local address (begins with FE80::/10).
Step 2 The host sends an NDP router solicitation (RS) message, with its link local address as the source address and the all-routers FF02::2 multicast destination address, to ask routers to supply a list of default routers and the prefix/length used on the LAN.
Step 3 The router(s) replies with an RA message, sourced from the router's link local address, sent to the all-IPv6-hosts-on-the-link multicast address (FF02::1), supplying the default router and prefix information.
Step 4 If the type of dynamic address assignment is stateless autoconfiguration, the following occur:
a. The host builds the unicast IP address it can use to send packets through the router by using the prefix learned in the RA message and calculating an EUI-64 interface ID based on the NIC MAC address.
b. The host uses DHCP messages to ask a stateless DHCP server for the DNS server IP addresses and domain name.