mail us | mail this page

(追記) (追記ここまで)

Tech Stuff - Ipv6

Version 6 of the IP Protocol. Defined in RFC 2460 (and updated by RFC5095 and RFC5722). Everything about IPv6 is BIG. An IPv4 address is 32 bits, an IPv6 address is 128 bits. This is about the number where each blade of grass on the planet could have its own IPv6 address.

IPv6 has been around since at least 1995 but the CIDR initiative of the mid-nineties pushed back any, then pressing, need for IPv6's increased address range. The can was effectively kicked down the road. In retrospect this was probably a Good Thing ™ since much of the recent (2010/2015) IPv6 work has been to reduce IPv6 address complexity, interworking and overall functionality.

IPv6 is big and complex in comparison with IPv4. This fact alone will keep users from implementation if they have any choice in the matter. Nevertheless, the internet is rapidly approaching the time - primarily due to IPv4 address depletion - when there is little choice but to move to IPv6. The transition will be accompanied by much yelling, screaming, gnashing of teeth and grim resignation. It is not helped by the constant roll-back and fiddling with the IPv6 specifications.

It is noted that RFC 6540 (published April 2012) now formally suggests that best practice protocol stacks should now include both IPv6 and IPv4. While IPv6 and IPv4 have been available, for many years, with mainstream OSs (Linux, BSD, Windows and others) and, due to its application in the mobile world, Android and other mobile centric OSs, most network boxes (DSL/cable modems) are still IPv4 only. High quality Open Source IPv6 stacks and this nudging from the IETF may change this. Then again, perhaps holding one's breath may not be a sensible strategy just yet since it will be many years before all that IPv4 NAT investment used by most ISPs will obsolete IPv4 functionality.

You need to be fairly comfortable with Hex stuff to handle IPv6 (quick overview of Hexadecimal, Binary and Decimal)

Contents

1. IPv6 Overview
2. IPv6 Address Notation
3. IPv6 Prefix or Slash Notation
4. IPv6 Address Types
5. IPv6 Global Unicast Address Format
6. IPv6 End-User Address Format
7. IPv6 Address Calculator
8. IPv6 Configuration Options
9. IPv6 Stateless Address AutoConfiguration (SLAAC)
10. IPv6 Link-Local Address Format
11. EUI-64 Address formation for SLAAC and Link-Local
12. IPv6 Multicast Address Format
13. IPv6 - IPv4 Interworking
14. IPv6 over IPv4 Interworking (6to4)
15. IPv6 Frame Format
16. IPv6 Header Format
17. IPv6 Order of Headers
18. IPv6 Extension Headers
19. IPv6 Discovery
20. IPv6 ICMP
21. IPv6 ping6 command
22. IPv6 traceroute6 command
23. IPv6 RFCs

IPv6 Overview

First things first. Each PC - more properly each network interface - may have more than one IPv6 address - IPv6 is naturally multi-homed. Second, an IPv6 address has a scope, that is, it can be restricted to a single LAN or a private network or be globally unique. The following table defines the types of IPv6 addresses that can be supported and contrasts them with their closest IPv4 functional equivalent.

IPv6 Address Notation

An IPv6 address consists of 128 bits - an IPv4 address consists of 32 bits - and is written as a series of 8 hexadecimal strings separated by colons. (Format defined by RFC 4291 and updated by RFC 5952) Examples:

# all the following refer to the same address
2001:0000:0234:C1AB:0000:00A0:AABC:003F
# leading zeros can be omitted
.&checktime(2001,0,23,':')4:C1AB:0:A0:AABC:3F
# not case sensitive - any mixture allowed
.&checktime(2001,0,02,':')34:C1ab:0:A0:aabc:3F

One or more zeros entries can be omitted entirely but only once in an address. The user can choose the most efficient place to omit multiple zero entries. Examples:

# raw ipv6 address
2001:0000:0234:C1AB:0000:00A0:AABC:003F
# address with single 0 dropped
2001::0234:C1ab:0:A0:aabc:003F
# alternate version - address with single 0 dropped
.&checktime(2001,0,02,':')34:C1ab::A0:aabc:003F
# the following is INVALID
2001::0234:C1ab::A0:aabc:003F

Multiple zeros can be omitted entirely but only once in an address. Examples:

# omitting multiple 0's in address
.&checktime(2001,0,0,':'):0:0:0:0:3F
# can be written as
2001::3F
# lots of zeros (loopback address)
0:0:0:0:0:0:0:1
# can be written as
::1
# all zeros (unspecified a.k.a unassigned IP)
0:0:0:0:0:0:0:0
# can be written as
::
# but this address
.&checktime(2001,0,0,':'):1:0:0:0:3F
# cannot be reduced to this 
2001::1::3F # INVALID
# instead it can only be reduced to
2001::1:0:0:0:3F
# or 
.&checktime(2001,0,0,':'):1::3F

Experiment using the IPv6 Address Calculator.

A hybrid format may be used when dealing with IPv6 - IPv4 addresses called an IPv4-Mapped IPv6 Address (RFC 4291 - updated by RFC 5952 and RFC 6052) where the normal IPv4 dotted decimal notation may be used after the first 6, 16 bit address elements:

# generic IPv6-IPv4 format x.x.x.x.x.x.d.d.d.d
# example of an IPv4 mapped IPv6 address
# with an IPv4 number of 192.168.0.5
2001:db8:0:0:0:0:FFFF:192.168.0.5
# or using zero elimination
2001.db8::FFFF:192.168.0.5

Experiment using the IPv6 Address Calculator.

Notes:

1. The FFFF element in the 6th position is fixed and must be present (see also RFC 6052 for a peculiar shifted version of this format when used with certain IPv4/IPv6 translators).

2. To avoid publication of a global IPv4 the example above shows a private (non-globally unique) IPv4 address purely to illustrate the principle but the IPv4 address used in IPv4-Mapped IPv6 must always be globally unique address.

Up Arrow

IPv6 Prefix or Slash Notation

IPv6 uses a similar / (forward slash) notation to IPv4 CIDR (Classless Interdomain Routing) which describes the number of contiguous bits used in its netmask. Formally this way of writing an address is called an IP prefix but more commonly called the slash format. Examples:

# single user address
2001:db8::1/128
# normal user IPv6 address allocation 
# allows the user to control the low order 80 bits
2001:db8::/48
# values only required to cover the prefix
FE8::/10
# or top 3 bits only with fixed value 001 (binary)
2::/3

Experiment using the IPv6 Address Calculator.

Up Arrow

IPv6 Address Types

The type of IP address is defined by a variable number of the top bits known as the binary prefix (BP). Only as many bits as required are used to identify the address type as shown in the following table (defined in RFC 4291):

The revised definition is a conceptual change and is both more flexible than the previous (RFC 2373/RFC 3513) definition if a tad confusing. IPv4 and NSAP prefixes are still allowed for but are now simply unicast addresses. Subsequent changes (from RFC 3513 to RFC 4291) seem to be trying to remove some of the restrictions previously imposed such that RFCs now use the distinction between a binary prefix of 000 (covering IPv6 mapped IPv4, unassigned and loopback) and not 000 (all other unicast) rather than explicity using 001 as previously. The obsolescence of address block assignment from the previous binary prefix 001 seems to be of a theoretical nature since all current IANA address block assigments are still from this prefix.

Experiment using the IPv6 Address Calculator.

Up Arrow

IPv6 Global Unicast Address Format

The IPv6 Global Unicast 128 bits was historically divided into a 48 bit global routing prefix (a.k.a site prefix) which is assigned by various authorities, a 16 bit subnet ID and 64 bits which the Interface ID (IID). The 16 bit subnet ID and IID (a total of 80 bits) is normally assigned and used by the user. While this structure is still the normal address allocation, the standards (RFC 4291) now define both the global routing prefix and the subnet ID to be of variable length (and using a total of 64 bits) in order to allow greater flexibility for the RIRs in allocating addresses. The formal structure is therefore:

The current IPv6 address allocation policy adopted by the various Regional Internet Registries should be based on the IETF/IAB recomendation (in RFC 3177) which allows for:

The justification for this, seemingly generous, address allocation policy is that the 45 address bits (48 bit global routing prefix - 3 bit binary prefix) still allow ~35 trillion /48 allocations (contrasted with the earth's population prediction of a mere 9 billion by 2050).

Experiment using the IPv6 Address Calculator.

IPv6 End-User Address Format

End-User addresses are assigned from the Global Unicast pool (current IANA IPv6 address block assignments). In addition a list of special assigments was created by RFC 4773. The IETF 6bone used a special range of 3FFe::/16 but the 6bone is disbanded and the address range has been returned to the IANA Reserved pool. The 128 bits breakdown as follows:

Experiment using the IPv6 Address Calculator.

Up Arrow

IPv6 Configuration

IPv6 systems are typically multi-homed by default and have a link-local address configured by the host and may have a global unicast address which may be configured by one of four methods:

1. Stateful - Statically assigned = manual configuration
2. Stateful - DHCPv6 - Automatically assigned
3. Stateless - Automatically assigned (SLAAC)
4. Stateless - Automatically assigned (SLAAC) with DHCPv6 support for DNS information.

IPv6 Stateless Autoconfiguration (SLAAC)

IPv6 systems may be configured to provide global unicast addresses using Stateless Address AutoConfiguation (SLAAC - defined by RFC 4862) using what is called generically the Neighborhood Discovery Protocol (NDP). Stateless autoconfiguration requires a router to be present but not a DHCP server. The process of creating a stateless IPv6 address is as follows:

1. Host sends a Router Solicitation message

2. Host waits for a Router Advertisement message.

3. Host takes top bits as defined in the Prefix Information of the Router Advertisement message and combines it with the 64 bit EUI-64 address (in the case of Ethernet this is created from the MAC address using this process) to create a Global Unicast address. The host also uses the source IP address - in the IP header - of the Router Advertisement (RA) message as its default gateway address. And since RFC 6106 can also obtain information about DNS service from the RA messages.

   RFC 4941 defines a method by which temporary (essentially pseudo-random from the interface derived EUI-64 address) addresses may be created in order to create privacy (or anonymity). RFC 4941 defines the default state of anonymous address creation as being off. If you wish anoymous access under IPv6 you may have to look for a specific configuration variable in your system to turn the anonymous feature on.

4. Host then performs a Duplicate Address Detection (DAD) to ensure the address is unique (and on any of the anaonymous addresses generated by RFC 4941 procedures). If this check fails the host immediately aborts the autoconfiguration process and must be manually configured.

Example:

# Interface MAC
00-40-63-ca-9a-20
# IPv6 Interface ID (EUI-64)
# Note: bit 7 (U/L) set - see link local notes
::0240:63FF:FECA:9A20
# or
::240:63FF:FECA:9A20
# Prefix from Router Advertisement
2001:db8::/64
# global Unicast address
2001:db8::240:63FF:FECA:9A20

Notes:

1. In stateless autoconfiguration (SLAAC) the global unicast (public address) and the default router address are configured automatically. Additional address information, such as DNS addresses, are not and must be provided by either DHCPv6 or manual configuration.

2. When using SLAAC the low order 64 bits (EUI-64) of the Link Local and the Global Unicast address will be identical since both are created using the same process.

Up Arrow

IPv6 Link-Local Address Format

Link-Local addresses are automatically assigned by the end user equipment and require no external configuration. Format defined by RFC 4291 Section 2.5.6. The address format uses a unique binary prefix (FE8::/10) and the remaining bits (118) are built from the local interface identifier. In the case of ethernet (RFC 2464) the MAC (48 bits) is used to create the EUI-64 value as shown below. Each physical layer supported has a separate RFC for example, FDDI, IEEE 802.15.4 etc. defining, among other things, how the link-local address is created. If an interface identifier has more than 118 bits the link-local address cannot be generated and the unit must be manually configured. Link-local addresses are not routable globally (outside the local LAN/network - however that is defined). The 128 bits of a link-local address for an ethernet interface breakdown as follows:

Example

# Interface MAC
00-40-63-ca-9a-20
# IPv6 Interface ID (EUI-64)
# Note: bit 7 of MAC set - see notes above
::0240:63FF:FECA:9A20
# or
::240:63FF:FECA:9A20
# link local
FE80::240:63FF:FECA:9A20

Up Arrow

IPv6 Multicast Address Format

The Multicast format (which also replaces broadcast in IPv4) is defined by RFC 4291 Section 2.7. There is now a link-local multicast format defined by RFC 4489. The fomat of a global (non link-local) multicast address is defined below:

Bizarre: RFC 4291 supposedly defines the IPv6 address structure including multi-cast addresses (2.7), however both RFC 3306 and RFC 3956 (earlier RFCs) define a more detailed format (not reflected in RFC 4291) which have been used to update the table below. Very disconcerting.

The following lists some of the more common multicast groups:

Experiment using the IPv6 Address Calculator.

Up Arrow

IPv6 Link Local Multicast Address Format

As if this stuff was not already complicated RFC 4489 introduced the concept of a link-local (or link scoped) multicast format for situations where all configuration is stateless. Theoretically, routers (and other equipment) servicing a local (non-global) network could be now made self-configuring.

Experiment using the IPv6 Address Calculator.

Up Arrow

IPv6 - IPv4 Interworking

IPv6 allows transport of IPv4 addresses using two methods. The methods are described in RFC 2893/RFC 4038 and RFC 4291 Section 2.5.5.

The first method is termed an "IPv4-compatible IPv6 address" and must use a globally unique IPv4 address. Please note: To avoid publication of a global IPv4 the example below shows a private (non-globally unique) IPv4 address purely to illustrate the principle:

# IPv4-compatible IPv6 address
# assume the host has an IPv4 address of:
192.168.0.5
# this is represented by the hex value C0A80005
# the IPv4-compatible IPv6 address would be
::C0A8:5
# or if you like writing zeros
0:0:0:0:0:0:C0A8:5
# or using the hybrid IPv6 - IPv4 notation
::192.168.0.5

The IPv4-compatible IPv6 address format is used when the end interface supports both IPv6 and IPv4 (a dual stack interface). This method is now deprecated (RFC 4291).

The second method is termed an "IPv4-mapped IPv6 address" and must use a globally unique IPv4 address. Please note: To avoid publication of a global IPv4 the example below shows a private (non-globally unique) IPv4 address purely to illustrate the principle:

# IPv4-mapped IPv6 address
# assume the host has an IPv4 address of:
192.168.0.5
# this address is represented by the hex value C0A80005
# the IPv4-mapped IPv6 address would be
::FFFF:C0A8:5
# or if you like writing zeros
0:0:0:0:0:FFFF:C0A8:5
# or using the hybrid IPv6 - IPv4 notation
::FFFF:192.168.0.5

The IPv4-mapped IPv6 address format is used when the end interface supports only IPv4 and indicates that a configured IPv6 system, for instance, a router or the IPv6 stack will have to perform conversion to the IPv4 protocol prior to communicating with the interface.

Up Arrow

IPv6 over IPv4 Interworking

IPv6 over IPv4 (more commonly refered to these days as 6to4) allows two IPv6 networks/devices to communicate over an IPv4 network (defined by RFC 3056). Both ends of the network must have a globally unique IPv4 address and the end points must run either a 6to4 relay or a 6to4 transit service. A special unicast address block has been assigned for these classes of service (2002::/16). The IPv6 address format used by this address block is defined below

The low order 80 bits allow for end user transparency within the IPv6 address. The relay or transit service will extract the IPv4 address when communicating accross the IPv4 cloud.

Note: Another IPv4 to IPv6 transition method is biting the dust. RFC 7526 deprecates the reserved Unicast prefix (2002::/16) used in 6to4 tunneling. For backward compability the feature will still exist but for new implementations: forget it. Meh.

Up Arrow

IPv6 Address Calculator

IPv6 Calculator: Enter an IPv6 address with its /Prefix (or slash) notation in the IPv6/Prefix: box and click IPv6 Info. The calculator will fully expand the IPv6 address and place in IPv6 Expanded:, IPv6 Netmask in compressed format, IPv6 GPR contains the top part or the address which may be the Global Routing Prefix (GRP) or not depending on the /Prefix value and IPv6 User: is the lower part of the address based on the /Prefix value. Finally, the reverse map zone name of the IPv6 address is calculated based on the /Prefix value (IPv6 addresses below this name will be defined in the reverse zone file) and shown in FQDN format in Reverse Zone. In Zone IPv6: shows the IPv6 address element that would appear with the zone file based on the /Prefix value. Since IPv6 addresses cover very serious numbers the No. of IPs: box is only used in the alternative mode described below and is only updated in that context.

Alternatively, enter a single IPv6 address (that lies anywhere in the desired range) without the /prefix in IPv6/Prefix and the number of desired IP addresses (if it is not a power of 2 it will be rounded up to the nearest power of 2 - a maximum of 6 digits is allowed) in No. of IPs then click IPv6 Info. The calculator will populate IPv6/Prefix (with the calculated /Prefix to cover the required or calculated number of IP addresses), No. of IPs will be updated if needed (due to any power of 2 rounding required) and IPv6 Netmask, IPv6 GPR, IPv6 User:, Reverse Zone and In Zone IPv6: will be populated normally.

The Clear button zaps ALL entries in IPv6 Calculator only.

Notes:

1. Validation and other errors are shown in the IPv6 Netmask box for no very good reason.

2. When entering a /prefix IPv6 address you only need to enter as many colon elements as are required by the /prefix (more is not a problem), if you enter less than the required number the calculator silently pads to the right with zeros.

3. While a /Prefix for IPv6 is most normally a multiple of 8 (or even 16) the calculator allows any bit value. You can break on very strange boundaries if desired.

4. The calculator follows the recommendation of RFC 5952 about use of lower case hexadecimal characters, removal of all leading zeros in address elements, largest zero block elimination but it will eliminate a single zero block if this is the only one available (since this obeys the longest zero block elimination rule).

5. The calculator accepts IPv6 addresses in IPv4-Mapped IPv6 format (x:x:x:x:x:x:d.d.d.d) but does not verify that the required selectors are present. You currently have to do some work.

Notes:

1. Javascript implementations may vary from browser to browser. This feature was tested with MSIE 10, Gecko (Firefox 25.0.1), Opera 11.10 and Chrome (31.0.1650.57 m) - so will likely work with any WebKit based browser, which obviously includes Safari (and now even Opera!)). If the tester does not work for you we are very, very sad - but yell at your browser supplier not us. However, if you do think you have found, or want to suggest improvements, an error please take the time to email using the link at the foot of this page.

Up Arrow

IPv6 Frame Format

IPv6 headers are daisy chained. The Next Header field - present in every header except the upper layer header - indicates which header comes next as shown in the diagram below:

Notes:

1. Zero or more Extension headers may be present.

2. Multiple Extension headers of any type may be present.

3. All Extension headers are assumed to be of variable length and contain a length value (expressed in multiples of 8 octets).

4. Only one upper layer header may be present and is unchanged from IPv4 e.g. tcp with the exception of the format of the 'pseudo-header' used to generate the checksum.

5. Data (MTU) length is defined to be a minimum of 1280 octets with a recommendation of 1500+ octets. If any routing link cannot carry this size of MTU, link specific fragmentation must be carried out below (i.e. invisible to) IPv6.

6. When carrying UDP traffic in the upper layers the optional (in IPv4) UDP checksum MUST be present.

7. The pseudo header used in TCP, UDP and IPv6 ICMP is assumed be a 40 octet field and have the following format:

   Name Size Description/Notes
   Source 128 bits IPv6 source address
   Destination 128 bits IPv6 destination address
   Packet Length 32 bits Length in octets of the upper layer data packet plus associated header.
   - 24 bits Must be zeros
   Next Header 8 bits Assumed to contain the value of the protocol carried in the upper layer e.g. 6 = TCP etc..

Up Arrow

IPv6 Header Format

Up Arrow

Order of Headers

Where multiple headers are present the recommended sender order is (but the receiver must accept in any order):

IPv6 Header
Hop-by-Hop Header
Destination Options
Routing Headers
Fragment Headers
Authentication Headers
Encapsulation Security Payload (ESP) Header
Destination Options
Upper Layer Header + data

Up Arrow

Extension Headers

Extension headers are always multiples of 8 octets. To allow skipping and processing of extension headers they all begin with the following 16 bit stub format:

Up Arrow

Header Options

The Hop-By-Hop and Destination Headers carry a variable number of options within the header and use a classic TLD (or TLV in the standards paralance) format as shown below: