IPv6 Addressing
Address Space Allocation
Table 8-1 on page 150 details how the IPv6 specification allocates the 128-bit IP address space. Specifically, the IPv6 specification entitled RFC (Request For Comments) 2373 defines the address space allocation. The column entitled “Upper Bits” defines the bit settings of the uppermost bits of the 128-bit address.
Port Addresses in “Unassigned” Range Cannot Be Used
All 128-bit IPv6 address ranges with one of the upper bit field values shown in the rows marked “Unassigned” are not available for assignment as port addresses.
Port Addresses in “Reserved” Range Cannot Be Used
All 128-bit IPv6 address ranges with one of the upper bit field values shown in the rows marked “Reserved” are not available for assignment as port addresses.
Unlimited Global Unicast Addresses (Travel the Globe!)
General
When it is desired that a port be universally addressable from any port in any other IBA or non-IBA subnet using a unicast global address, an address in the range defined as “Aggregatable Global Unicast Addresses” should be assigned to the port. In other words, all 128-bit addresses with a 001b in the uppermost three bits of the address are set aside for assignment as unicast global addresses.
Global Unicast Address Format
Refer to Figure 8-4 on this page. As defined in the IPv6 specification, a unicast global address is divided into the fields indicated in the figure. The figure annotations define the hierarchical breakdown of the address.
Figure 8-4. Unicast GID Address Format
Limited Global Unicast Addresses
General
Some global unicast addresses are limited in that routers are required to restrict how far afield a packet with such a DGID address is allowed to travel.
Link-Local Unicast Address: Packet Cannot Leave the Subnet
Refer to Figure 8-5 on page 150 and the table row (in Table 8-1 on page 150) marked as “Link-local unicast addresses.” A packet with a GRH:DGID address with the upper 64 bits (i.e., the subnet ID) set to FE80000000000000h is not permitted to leave the subnet through which it is currently traveling. In other words, if a router should receive a packet with a DGID address containing this subnet ID, the router is not permitted to route that packet out of the current subnet. It is silently dropped.
Figure 8-5. Link-Local Unicast GID Format
Site-Local Unicast Address: Packet Cannot Leave the Site
Refer to Figure 8-6 on this page and Table 8-1's row entitled “Site-local unicast addresses.” A packet with a GRH:DGID address with the upper 48-bits set to FEC0000000000000h is not permitted to leave the site or campus through which it is currently traveling.
Figure 8-6. Site-Local Unicast GID Format
As shown in the figure, the next-lower 16-bit field identifies the target subnet (one of 64K possible subnets) within the site or campus.
In other words, if a router at the fringe of the site or campus should receive a packet with a DGID address containing this value, the router is not permitted to route that packet out of the site or campus. It is silently dropped.
| Usage | Upper Bits | Notes |
|---|---|---|
| Reserved | 0000 0000 | Includes the Unspecified Address, Loopback Address, and 32-bit IPv4 addresses embedded within a 128-bit IPv6 address. |
| Unassigned | 0000 0001 | |
| Reserved for NSAP allocation | 0000 001 | The Network Service Access Point (NSAP) |
| Unassigned | 0000 010 | |
| Unassigned | 0000 011 | |
| Unassigned | 0000 1 | |
| Unassigned | 0001 | |
| Aggregatable Global Unicast Addresses | 001 | Addresses with the upper 3 bits set to 001b are available for assignment to IBA ports as unicast global GIDs. See “Unlimited Global Unicast Addresses (Travel the Globe!)” on page 148. |
| Unassigned | 010 | |
| Unassigned | 011 | |
| Unassigned | 100 | |
| Unassigned | 101 | |
| Unassigned | 110 | |
| Unassigned | 1110 | |
| Unassigned | 1111 0 | |
| Unassigned | 1111 10 | |
| Unassigned | 1111 110 | |
| Unassigned | 1111 1110 0 | |
| Link-local unicast addresses | 1111 1110 10 | Addresses with the upper 10 bits set to this value are available for assignment as link-local GID addresses. See “Default Subnet ID” on page 155 and Figure 8-5 on page 150. |
| Site-local unicast addresses | 1111 1110 11 | Addresses with the upper 10 bits set to this value are available for assignment as site-local GID addresses. See Figure 8-6 on page 150. |
| Multicast addresses | 1111 1111 | Addresses with the upper 10 bits set to this value are available for assignment as multicast GID addresses. See “IPv6 Multicast Addressing” on this page. |
IPv6 Multicast Addressing
A multicast GID acts as an identifier for a group of ports on CAs and routers. The multicast GID format is shown in Figure 8-7 on page 153.
An upper-most bit pattern of 1111 1111b (FFh) in the 128-bit GID address identifies this as a multicast GID address.
Flags is a set of four, 1-bit flags, 000T, with the upper three flag bits reserved and defined as zero. The T flag bit is defined as follows:
- T = 0 indicates this is a permanently assigned (i.e., well-known) multicast GID. See the IPv6 RFC 2373 and RFC 2375 specifications as references for these permanently assigned GIDs.
- T = 1 indicates this is a non-permanently assigned (i.e., transient) multicast GID.
Scope is a 4-bit multicast scope value used to limit the scope of the multicast operation. Table 8-2 on page 153 defines the scope values and interpretation.
Figure 8-7. Multicast Global GID Format
Usage rules:
A CA or router may join none, one, or multiple multicast groups (i.e., a CA or router port may be assigned zero, one, or more multicast GIDs).
Multicast GIDs never appear as the SGID in a packet's GRH.
Multicast GID FF02:0:0:0:0:0:0:1 (see “IPv6 Address Documentation Convention” on page 154 for a description of this documentation convention) is the link-local multicast GID—routers do not route packets with this DGID outside the local subnet. This GID is used as the GRH:DGID for communicating to a set of QPs participating within the All CAs Multicast Group. This group is used to implement a broadcast service to all CAs in the subnet that are capable of participating in multicast operations.
The IPv6 specification defines a set of reserved multicast addresses in RFC 2375 and RFC 2373. IBA, unless explicitly stated otherwise, does not use these addresses for IBA multicast operations and defines them as reserved for raw IPv6 usage (more on raw operations later).
Refer to “Switch Multicast Packet Forwarding” on page 675 and “Multicasting” on page 563 for additional discussion of multicast operations.
IPv6 Address Documentation Convention
A 128-bit IPv6 address can be represented in three forms:
x:x:x:x:x:x:x:x, where each x = 16-bit hex value.
Some IPv6 addresses contain long strings of 0 bits. To make documenting easier, use of “::” indicates multiple, contiguous 16-bit groups of zeros. “::” can only appear once in an address. “::” can also be used to represent leading and/or trailing zeros in the address. Examples:
1080:0:0:0:8:800:200C:417A, a unicast address, can be represented as 1080::8:800:200C:417A.
FF01:0:0:0:0:0:0:101, a multicast address, can be represented as FF01::101.
0:0:0:0:0:0:0:1, the loopback address, can be represented as ::1.
0:0:0:0:0:0:0:0, the unspecified address, can be represented as ::.
The form when dealing with mixed environment of IPv4/IPv6 nodes is: x:x:x:x:x:x:d.d.d.d, where:
The “x”s are the hex values of the six most-significant 16-bit portions of the address.
The “d”s are the decimal values of the four least-significant 8-bit portions of the address (IPv4 representation).
Examples:
0:0:0:0:0:0:13.1.68.3 can be represented as ::13.1.68.3.
0:0:0:0:0:FFFF:129.144.52.38 can be represented as ::FFFF:129.144.52.38.