IPV6 settings on third party device (new EE Broadband - Full Fibre 1.6 Gbps)

[GosforthUK]

Does anybody know (or figure out) the correct settings for EE's super implementation IPV6 for third party router on new EE broadband?

I am aware, as BT has gone so far, to adapt unusual /56 delegation instead of more common /64 and the whole BT IPV6 is "non-address" in fact...

Whilst on 1.6 Gbps Broadband, I would like to ditch EE's supplied Smart Hub Plus (2023) but for now, due nobody is able tell me correct settings, I need lean for IPV6 on EE HUB and therefore have it connected

[bobpullen]

What router? I'm guessing TP-Link?

I've used a number of third party devices and they've always 'just worked'.

[GosforthUK]

@bobpullen aye, TP-Link ER707-M2, but in controller mode (Omada)...

Before my old TP-Link WiFi router "just worked", but say, it was on BT FF900, which I was in impression, as will be exactly same for EE (New) Broadband, whilst technically speaking it using BT backhaul...

But while "only IPV4" mode will work perfectly and I am able to get all working, moment I try to go for IPV6, trouble starting...

Any settings I tried, it either not works, or, IPV6 it's somehow there, but slowing down everything in terms, as 1.6 Gbps drop to 400-700 Mbps  instead (but this might be as 707-M2 isn't capable to handle it correctly, but would like to rule it out eventually, before will look after any more capable router eventually - perhaps, it should "just work" :))...

[Ollmall]

I have got IPV6 working on the EE 1.6gbps plan. HOWEVER for whatever reason it causes slow download speeds in certain applications, namely Steam and limits the downloads to around 500-600mbps. When disabling IPV6 it whacks speeds straight back up to full.

I'm using a UDM Pro for my home setup as an FYI but the settings that work for me are:

Connection type: DHPV6
Prefix Delegation: 56
Custom DNS (Cloudflare): Primary: 2606:4700:4700::1111
Secondary: 2606:4700:4700::1001

[GosforthUK]

However, still fighting with limited speed for few apps (similar to You @Ollmall ) but standard web speed on my mobile is something around:

 

[GosforthUK]

Present settings (don't going to say it's perfect or correct, but it's working) on my TP-Link Omada ER707-M2 router with hardware OC200 controller are: 

WAN side:

 

LAN Side:

 

[ErMejoo]

Hey @GosforthUK - thanks for sharing - I see you have the same setup I just did for my UDM SE, with a /56 delegated address from EE (WAN) and the internal network at /64 (LAN) upon that public IPv6 address subnet. How do we think of this, is the LAN settings telling us that we are effectively getting a /64 and not a /56 from EE, and potentially we shall change it at the connection (WAN) level, too? Or is that just an hard-coded/configurable setting of our routers that we somehow might have missed? Trying to understand the implications 🙂 

On a side note, my config is with DHCPv6 rather than SLAAC as I see in your WAN settings - I see the latter is a manner to tie IPv6 addresses to MAC addresses rather than gateway-assigned addresses, is that helping with more static~ish IPs? On the one hand, a quick look-through surfaced "Stateless address autoconfiguration is a mean for an IPv6 host to assign itself an address based on its hardware MAC without asking a DHCP server. Using SLAAC without privacy extensions (RFC4941) is not recommended." ref. ipv6-test.com while for me the test still returns "No" despite having jus tried to enabled it, aside from that I see no immediate changes, except new dynamic IPv6s assigned across the network upon the change.

[GosforthUK]

Essentially @ErMejoo 

/64 is size of SINGLE LAN

and that means, as it's fine, if you planning have only one single LAN inside, and don't bother with more.

Means, if you go via /64 allocation both on WAN and LAN, you will be unable to split easily your local LAN between more than one LAN and all your IPV6 addresses will come in fact under it.

When using /56 allocation on WAN (or even /48 - not sure if BT/EE support this) you have option to use more than one LAN allocation, and have more separation.

/64 it's standard size of one LAN for IPV6 as defined in protocol 

/56 contains 256 LANs (so essentially 256x /64)

/48 contains 65 536 LANs and therefore I suspect, it might be intentionally blocked to use by BT/EE (but again, never tested it)

At a present moment, I switched from "no address" which was recommended on BT before, to DHCPv6 on WAN side and it seem to works perfectly fine for me.

Don't forget, in theory, each single IPv6 address assigned it's unique and should be able to be reach over the internet.

Omada routers had before constantly omit this and allow traffic into LAN to be routed (and needed to be treated separately by ALC).

However, my ER707-M2 seems to do routing and firewall both on IPv4 and IPv6 correctly with latest beta FW and my IPv6 LAN devices are not reachable from outside as intended.

While IPv6 has virtually unlimited allocations, this, unlikely from IPv4 coming with more permanent allocations, and essentially you could have fixed IP (of some some sort) like that.

When I moved from "no address" on IPv6 to DHCPv6, my IPv4 immediately changed from typical allocations to something like 31.51.xxx.xxx and don't change ever since yet.

If that has or not connection between each other, or it's just coincidence, not sure.

[ErMejoo]

Thanks for education @GosforthUK - this was very helpful! 

In fact, I had just played further and moved from DHCPv6 to SLAAC with /56 PD in UDM SE WAN settings, and this seems to allow exactly what you explained above, which is split the /56 across multiple (x256) /64 local networks, whilst this is now "automatically handled", this is going to become also configurable in future versions, i.e., "The Prefix ID is automatically set to an optimal value and does not change, meaning that this network will keep using the same IPv6 address range (unless there is a change on the ISP). In a future release, we will add support for manually configuring the Prefix ID.". This seems interesting so I will test this config for a while, promising close-static IPv6 addresses, except for an ISP change of course, while it seems to have many implications, all Google Nest devices now do get their IPv6 address, and while they now rack up an eye-watering of 20-26 addresses each that seems very unnecessary, this might solve the issues I was having with them complaining to "be offline" (even though I could never verify that), previous "static IP" setup devices like my two NAS' have maintained their link-local IPv6 even after flipping their config to "Auto", which I feel is a good-to-have, as I was unable to assign them both a public IPv6 and local IPv6 fixed address before, and one of them is a Windows Domain Controller that requires a static accessible IPv6 DNS pushed throughout the local network, the second NAS' is it's a secondary DNS so that also needs that static link-local IPv6 address. It seems this is working well so far as link-local DNS' are working as expected, however it'd be interesting to understand what you were also mentioning, i.e. if/how do we get to assign fixed IPv6 addresses programmatically as we do with IPv4 based on MAC today!

Last, interesting read re. the accessibility of public IPv6 from outside the network, I was automatically assuming they would be "intended to be accessible" post IPv6 deployment, yet good reflection above re. the fact gateways might be not on the same page, and rightly so in many ways, as I have been worried about how much could this expose old/outdated devices to  unnecessary several new risks - I'll invest some time to fully understand the facts and configure routes/firewalls at best.

[Ollmall]

Did you ever figure out how to resolve the slow download speeds on certain services or is it still an issue? I ended up disabling IPV6 because of the slower download speeds.