IPv6 Assigned Prefix and allocation

Forums › 2degrees (including Slingshot, Orcon, Flip, Stuff Fibre, MyRepublic, 2talk and Vocus) › IPv6 Assigned Prefix and allocation

Gordy7

1744 posts

Uber Geek

ID Verified

Lifetime subscriber

#270113 24-Apr-2020 10:17

I am with Orcon that provides me with IPv6 /56 address allocations. (256 IPv6 addresses).

When I look a various routers:

HG659 shows /56 assigned prefix.

Vodafone Ultra Hub shows /56 assigned prefix.

Spark Smart Modem shows /64 assigned prefix.

What determines the assigned prefix shown in a router?

I am guessing that I can only use one IPv6 /64 address at a time.

Orcon uses DHCP so maybe the Spark router is correctly displaying the one assigned IPv6 from the 256 allocated address.

Can anyone help clear up what is going on?

TIA.

Gordy

My first ever AM radio network connection was with a 1MHz AM crystal(OA91) radio receiver.

cokemaster

Exited
4780 posts

Uber Geek

Retired Mod

Trusted

Lifetime subscriber

#2469056 24-Apr-2020 10:38

An IPv6 /56 actually has 4,722,366,482,869,645,213,696 addresses (source).

I suspect that the HG659/Ultra Hub is getting the prefix delegated correctly while the Smart modem is not (Spark is more of an IPv4 shop at the moment). I suspect you are correct where it is just requesting a single lan segment.

webhosting

Loose lips may sink ships - Be smart - Don't post internal/commercially sensitive or confidential information!

MightyApe

You will find anything you want at MightyApe (affiliate link).

Gordy7

1744 posts

Uber Geek

ID Verified

Lifetime subscriber

#2469085 24-Apr-2020 11:05

cokemaster:

An IPv6 /56 actually has 4,722,366,482,869,645,213,696 addresses (source).

Thanks for the link on IPv6 addresses.

Wow. What a heap of addresses.

/56 4,722,366,482,869,645,213,696 allocated by Orcon.

/64 18,446,744,073,709,551,616 residential addresses

I was somehow thinking that /64 minus /56 (8 bits) was the range of just 256 IPv6 addresses allocated.

In fact it appears to be 256 range blocks ie /56 = 256 x /64 IPv6 addresses....

More to learn and try to understand :-)

Gordy

My first ever AM radio network connection was with a 1MHz AM crystal(OA91) radio receiver.

noroad

751 posts

Ultimate Geek

Trusted

#2469091 24-Apr-2020 11:16

If the ISP allocates a /56 the router will normally pick a random /64 out of those 256 available and allocate it to its internal Ethernet (reboot and it will change, but don't worry this reallocation just works from my testing). If you have static IP and the ISP allocates a static /56 then you are free to break that up how you please on your internal network manually. ISP's can allocate anything from a /64 to a /48 normally from what I have seen. The /56 is a nice compromise as it means there is plenty for a medium business to break up and spread around their network should the so chose, without being too silly.

fe31nz

1081 posts

Uber Geek

#2470613 25-Apr-2020 00:11

noroad:

If the ISP allocates a /56 the router will normally pick a random /64 out of those 256 available and allocate it to its internal Ethernet (reboot and it will change, but don't worry this reallocation just works from my testing). If you have static IP and the ISP allocates a static /56 then you are free to break that up how you please on your internal network manually. ISP's can allocate anything from a /64 to a /48 normally from what I have seen. The /56 is a nice compromise as it means there is plenty for a medium business to break up and spread around their network should the so chose, without being too silly.

The original design of IPv6 was to have a /48 delegated to all end-user sites. That allows a site to expand as much as it likes over a very long lifetime without ever having to adjust its IPv6 addressing system. A tiny business can grow to the size of Google or Amazon without ever having to adjust the addressing. I am not sure why ISPs are not providing /48s, but they would not run out of IPv6 addresses if they did. A /56 is only enough for 256 subnets of IPv6 addresses, which any decent sized business can overflow. As a fairly sophisticated home network user, I have 5 main subnets already, and others used for testing and virtual machines. In theory, you can make IPv6 subnets smaller than a /64. In practice, most IPv6 software in end devices (eg IoT, mobile phones) is not capable of handling a subnet size of smaller than /64 even when the IPv6 router sends packets telling the devices what the subnet size is. So no-one is using less than a /64 for an IPv6 subnet in real networks.

An IPv6 router does not need a unicast global IPv6 address on its WAN port - the packets on the WAN port normally just go to the ISP's next hop router, which is on the same subnet as the WAN port. So they can and usually are sent using a link-local IPv6 address (fe80::). Even if your router has got a unicast global IPv6 address on its WAN port, it will not normally use it for routing packets to your ISP. So what an IPv6 router needs and does use is a unicast global IPv6 address on each of its LAN ports. Then your network on the LAN side can access the IPv6 router using those unicast global IPv6 addresses no matter how many hops away in your network the connection is coming from. If you have unicast global IPv6 addresses on your router's LAN ports, then when the router needs to send IPv6 packets to the WAN network (the Internet) that has to go more than one hop (eg check for new firmware to download), it will normally use the unicast global IPv6 address of one of the LAN ports as the source address for those packets. The unicast global IPv6 addresses on the LAN ports are assigned from the /64 block for the subnet connected to that LAN port. So there is actually no need for an IPv6 router to have a /64 block assigned to it for its own use.

Here is a cut down output of the "ifconfig" command on my Ubiquiti ER4 router showing how it works:

eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1508

inet6 fe80::9ec7:a6ff:fea8:b006 prefixlen 64 scopeid 0x20<link>

eth0.10: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1508

inet6 fe80::9ec7:a6ff:fea8:b006 prefixlen 64 scopeid 0x20<link>

pppoe0: flags=4305<UP,POINTOPOINT,RUNNING,NOARP,MULTICAST> mtu 1500

inet 203.86.202.190 netmask 255.255.255.255 destination 111.69.7.108

inet6 fe80::2085:94ea:26a6:67fa prefixlen 10 scopeid 0x20<link>

eth1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500

inet 10.0.1.251 netmask 255.255.255.0 broadcast 10.0.1.255

inet6 2406:e001:1:2801::251 prefixlen 64 scopeid 0x0<global>

inet6 fe80::1ae8:29ff:febe:e295 prefixlen 64 scopeid 0x20<link>

eth2: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500

inet 10.0.2.251 netmask 255.255.255.0 broadcast 10.0.2.255

inet6 2406:e001:1:2802::251 prefixlen 64 scopeid 0x0<global>

inet6 fe80::1ae8:29ff:febe:e296 prefixlen 64 scopeid 0x20<link>

As you can see its WAN interfaces (eth0, eth0.10 = VLAN 10, pppoe0) do not have a unicast global IPv6 address ("scopeid 0x0<global>"), just a link-local one ("scopeid 0x20<link>"). The pppoe0 interface does have my static IPv4 IP address assigned to it so that it can route IPv4 packets to the next hop router ("destination 111.69.7.108"). The next hop router's link-local IPv6 address is not visible anywhere in the ifconfig output. It is just the link-local IPv6 address of the destination of the pppoe0 connection, and is only used in the PPPoE code. The router sends its IPv6 packets to its default route, which is set to send them to the PPPoE connection. That is done with this line in its config:

set protocols static interface-route6 '::/0' next-hop-interface pppoe0

Tracer

343 posts

Ultimate Geek

#2470621 25-Apr-2020 01:17

fe31nz:

In theory, you can make IPv6 subnets smaller than a /64. In practice, most IPv6 software in end devices (eg IoT, mobile phones) is not capable of handling a subnet size of smaller than /64 even when the IPv6 router sends packets telling the devices what the subnet size is. So no-one is using less than a /64 for an IPv6 subnet in real networks.

You need 64 host bits for SLAAC too, so if you go smaller than /64 you rule it out.

noroad

751 posts

Ultimate Geek

Trusted

#2470711 25-Apr-2020 09:32

Tracer:

fe31nz:

In theory, you can make IPv6 subnets smaller than a /64. In practice, most IPv6 software in end devices (eg IoT, mobile phones) is not capable of handling a subnet size of smaller than /64 even when the IPv6 router sends packets telling the devices what the subnet size is. So no-one is using less than a /64 for an IPv6 subnet in real networks.

You need 64 host bits for SLAAC too, so if you go smaller than /64 you rule it out.

Using /126's for linknets between two devices is very common I assure you. Once your way of doing things has been trained around "conserve the valuable resource" wasting IP space for no good reason is just not something that many engineers can live with, even if there is lots of it.

fe31nz

1081 posts

Uber Geek

#2471339 25-Apr-2020 23:48

Tracer:

You need 64 host bits for SLAAC too, so if you go smaller than /64 you rule it out.

Yes, I keep forgetting about SLAAC as I do not use it - I have my network on DHCPv6, and on the one non-rooted Android device, it just has a link-local IPv6 address and has to make do with IPv4.