IPv6 Support in Dragonflow

IPv6 Support in Dragonflow


IPv6 and IPv4 have significant differences, not only on the IP address format, but also on the upper layer protocols. In this specification, we will discuss the required changes for existing Dragonflow applications, and required IPv6-only applications. This will help IPv6 networks to gain the same benefits that IPv4 networks currently gain.

Problem Description

Currently, most of Dragonflow applications and services support only IPv4. This makes the process of deploying IPv6 network not as easy as deploying IPv4 network. In order for Dragonflow to be supported fully in IPv6, deploying IPv4 and IPv6 networks and VMs must be the same.

Proposed Change

In this document, the plan is to add IPv6 support to Dragonflow. Specifically, the plan is to add support for IPv6 to the Dragonflow framework, and its applications. New IPv6 services, which do not exist in IPv4, will be added as standalone Dragonflow applications.

The following items are necessary to add IPv6 support for Dragonflow:

Trivial changes:

  • Neighbor Discovery
  • L2 (already works)
  • L3 (already works)
  • Security Groups

A separate specification document is required for:

  • Metadata service
  • DHCPv6

The following items are IPv6 features and services which are nice to have:

  • ICMPv6 responder (fully implemented)
  • Inverse Neighbor Discovery
  • Secure Neighbor Discovery
  • Multicast Router Discovery
  • Floating IPs (currently not supported by Open Stack Networking)

The following items exist for IPv4 and should not be ported, nor support, IPv6:

  • dNAT
  • sNAT
  • ARP responder

Metadata Service

Currently Neutron Metadata service does not support IPv6 [1] [2]. When the Neutron’s specification and implementation will be completed, a dedicated spec will be written as it was done in Metadata Service Support.

Neighbor Discovery

Neighbor Discovery [3] relates to different protocols and processes known from IPv4 that have been modified and expanded. It combines Address Resolution Protocol (ARP) and ICMP Router Discovery and Redirect.

Neighbor Discovery is used in the ICMPv6 protocol to verify there are no IP collisions, to match IP to MAC addresses, and to provide information to fellow VMs and routers.

In Dragonflow, the plan is to write a Neighbor & Router Advertisement [4]. It will be implemented using OpenFlow flows. It will detect neighbor solicitation requests by identifying the packet type [5]. The application will build the response with the requested Link-Layer Address.

icmp6,ipv6_dst=1::1,icmp_type=135 actions=load:0x88->NXM_NX_ICMPV6_TYPE[],move:NXM_NX_IPV6_SRC[]->NXM_NX_IPV6_DST[],mod_dl_src:00:11:22:33:44:55,load:0->NXM_NX_ND_SLL[],IN_PORT

In case the packet is something flows cannot handle, it will be passed to the controller.


In IPv6 there are two non-exclusive modes of DHCPv6 [6]:

  1. Stateful Address Autoconfiguration
  2. Stateless Address Autoconfiguration (SLAAC) [7]

Stateful DHCPv6 means host configuration with IPv6 assignment (similar to DHCPv4). Stateless DHCPv6 means only configuration information to hosts (DNS, NTP, etc), and not perform any address assignment. Dragonflow will support both modes. Exact implementation will be documented in a future document [8].

Security Groups

The current implementation supports only IPv4. Since the behavior is the same, the adjustments will be relatively minor. The changes will affect mostly on the conversion of the IP to integer, and building the flow with IPv6 fields.


Neighbor Discovery, DHCPv6 and the Security Groups are planned to be completed by the end of Ocata. The Metadata Service implementation is depending on Neutron. For the remaining items there is no estimated cycle.

Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.