An introduction to NDP

NDP is specified in IETF RFC 4861. In many ways, NDP is the spiritual successor to ARP for IPv6. It is used to help computers find the MAC addresses of their neighbors on a local network, but it also has much more to offer thanks to its added flexibility and extensibility.

To start, machines that make use of IPv6 always assign a

• Type: a general class of messages (indicates NDP message types).
• Code: a sub-class of messages within a type (always 0 for NDP).
• Checksum: used to verify the integrity of an ICMPv6 message.
• Data: an arbitrary payload (contains NDP message structures).

NDP has several different message types which allow it to be used for discovering IPv6 neighbors and routers on a local network. These include:

• Neighbor Solicitation: ask a neighbor for its MAC address using its IPv6 address.
• Neighbor Advertisement: inform a neighbor of an interface’s MAC address.
• Router Solicitation: request that routers generate router advertisement messages.
• Router Advertisement: inform neighbors that a router is available to be used as a default IPv6 router.
• Redirect: sent by a router in response to an IPv6 packet to inform an interface of a better first hop router. This message won’t be covered in detail in this post.

In addition, NDP supports options in a flexible TLV (type, length, value) format, making the protocol extremely extensible. Some commonly used options include:

• Source/Target link-layer addresses: the source or target (depending on solicitation/advertisement messages) link-layer address of an IPv6-speaking interface.
• Prefix information: sent by routers to inform neighbors if an IPv6 prefix is “on-link”, and if it can be used for Stateless Address Autoconfiguration (SLAAC).
• MTU: sent by routers to inform neighbors of the expected MTU on a network.
• Recursive DNS servers (RDNSS): specifies recursive DNS servers for this network, for use with SLAAC-configured interfaces.

Using NDP in Go

Now that we’ve learned that some of the fundamentals of NDP with IPv6, let’s examine a Go package, , which allows NDP to be used in Go programs.

The fundamental type of the NDP package is ndp.Conn. An ndp.Conn is used to bind an ICMPv6 connection on a network interface for the purposes of sending and receiving NDP messages.

The typical use case of an ndp.Conn is to send and receive NDP messages which implement the ndp.Message interface. The messages implemented in the package mirror the name and structure of the messages defined in the NDP specification. Within these messages, NDP options can be specified using types which implement the ndp.Option interface.

As an example, let’s demonstrate use of the package by sending a Neighbor Solicitation message to discover the MAC address of an IPv6 neighbor, using its link-local address.

When using ARP, the equivalent message would be broadcast to all machines on the same local network. With IPv6, broadcast is completely eliminated in favor of increased use of multicast groups. With NDP, we can use an interface’s solicited-node multicast address to greatly reduce the amount of network traffic required for this operation.

As a final example, let’s send a Router Solicitation to discover routers and IPv6 prefix information on our local network. When the “autonomous address autoconfiguration” flag is set in a Router Advertisement, IPv6-enabled interfaces can use it to configure their own IPv6 addresses automatically; no DHCPv6 server required!

Summary

The Neighbor Discovery Protocol is a crucial and fundamental part of IPv6 networking. Compared to ARP and IPv4, NDP and IPv6 offer numerous advantages:

• NDP builds on top of ICMPv6 using link-local addresses, instead of dealing with a totally different protocol on top of Ethernet frames.
• IPv6 makes smart use of multicast to eliminate the need for broadcast, preventing a lot of unnecessary network noise.
• NDP can be easily extended because new options can be added in a backward-compatible way.

Although your operating system will take care of NDP on your behalf for normal applications, it can occasionally be useful to exercise the power of NDP and other low-level networking protocols directly. If you’re interested in exploring NDP traffic on your local network, check out the ndp tool included in my repository.

$ go get -u github.com/mdlayher/ndp/...

Here’s an example of using ndp to send Router Solicitations on interface eth0 from the interface’s link-local address until a Router Advertisement is received:

$ sudo ndp -i eth0 -a linklocal rs
ndp> interface: eth0, link-layer address: 04:18:d6:a1:ce:b8, IPv6 address: fe80::618:d6ff:fea1:ceb8
ndp rs> router solicitation:
    - source link-layer address: 04:18:d6:a1:ce:b8

ndp rs> router advertisement from: fe80::201:5cff:fe69:f246:
    - hop limit:        0
    - flags:            [MO]
    - preference:       0
    - router lifetime:  2h30m0s
    - reachable time:   1h0m0s
    - retransmit timer: 0s
    - options:
        - prefix information: 2600:6c4a:7002:100::/64, flags: [], valid: 720h0m0s, preferred: 168h0m0s

If you’d like to make use of NDP in your own applications, I encourage you to try out my Go package, .

It is currently used in MetalLB to expose Kubernetes services over IPv6 addresses, and for internal projects at DigitalOcean. Perhaps it could also be used as a model to implement your own NDP library in your programming language of choice!

This blog provides a high-level overview to get readers started with NDP, but if you’d like to learn more about some of the common use cases for NDP in computer networks today, check out Jeremy Stretch’s blog on the topic at PacketLife.

Thank you very much for reading this post. I hope you’ve enjoyed it and learned something new along the way. If you have, you may also be interested in some of my other posts about using low-level networking primitives with the Go programming language.

Finally, if you have questions or comments, feel free to reach out via the comments, Twitter, or Gophers Slack (username: mdlayher).

Thanks again for your time!

Links

References