Brocade Fabric OS FCIP Administrator's Guide v7.1.0 (53-1002748-01, March 2013)

34 Fabric OS FCIP Administrator’s Guide
53-1002748-01
Support for IPv6 addressing
2
In some cases, traffic isolation zoning (TIZ) or VF LS/LF configurations may be used to control the
routing of SID/DID pairs to individual tunnels. This provides deterministic flows between the
switches and allows the use of ECMP. Refer to the Fabric OS Administrator’s Guide for more
information about TIZ.
Support for IPv6 addressing
The IPv6 implementation is a dual IP layer operation implementation as described in RFC 4213.
IPv6 addresses can exist with IPv4 addresses on the same interface, but the FCIP circuits must be
configured as IPv6-to-IPv6 and IPv4-to-IPv4 connections. IPv6-to-IPv4 connections are not
supported. Likewise, encapsulation of IPv4 in IPv6 and IPv6 in IPv4 is not supported.
This implementation of IPv6 uses unicast addresses for the interfaces with FCIP circuits. Unicast
addresses must follow the RFC 4291 IPv6 standard. This IPv6 implementation uses the
IANA-assigned IPv6 Global Unicast address space (2000::/3). The starting three bits must be 001
(binary) unless IPv6 with embedded IPv4 addresses is used. The link-local unicast address is
automatically configured on the interface, but using the link-local address space for FCIP circuit
endpoints is not allowed. Site-local unicast addresses are not allowed as FCIP circuit endpoints.
Note the following IPv6 addressing points:
• Anycast addresses are not used. Each IPv6 interface has a unique unicast address and
addresses configured are assumed to be unicast.
• Multicast addresses cannot be configured for an IPv6 interface with FCIP circuits. The IPv6
interface does not belong to any Multicast groups other than the All-Nodes Multicast and the
Solicited-Node Multicast groups (these do not require user configuration).
• The IPv6 implementation follows the RFC 2460 standard for the 40-byte IPv6 header format.
• The IPv6 8-bit Traffic class field is defined by the configured Differentiated Services field for
IPv6 (RFC 2474). The configuration of this is done on the FCIP circuit using the Differentiated
Services Code Point (DSCP) parameters to fill the 6-bit DSCP field.
• Flow labels are not supported on this IPv6 implementation. The 20-bit Flow Label field defaults
to all zeros.
• The IPv6 optional Extension Headers are not be supported. The optional Extension Headers
inserted into any ingress packets that contain these headers will be discarded. The next
header field must be the Layer 4 protocol for this implementation.
• Parts of the Neighbor Discovery protocol (RFC 4861) are used in this implementation.
- Hop limits (such as Time to Live (TTL)) are learned from the Neighbor Advertisement
packet.
- The link-local addresses of neighbors are learned from Neighbor Advertisement.
- The netmask is deprecated in IPv6. Instead, the prefix length notation is used to denote
subnets in IPv6 (the Classless Inter-Domain Routing (CIDR) addressing syntax). Prefix
length of neighbor nodes is learned from the received Neighbor Advertisement packet.
- The IPv6 link-local address for each GE interface is configured at startup and advertised to
neighbors. The user does not configure the interface link-local address.
• The 8-bit hop limit field is filled by the learned value during Neighbor Discovery.
• IPv6 addresses and routes must be statically configured by the user. Router Advertisements
and IPv6 Stateless Address Autoconfiguration (RFC 2462) are not supported.