Weird IPv6 stuff

Hi all!

Just configured a wAP LTE and a cAP AC behind it. The wAP gets an IPv6 prefix from the ISP. Two things that aren’t too clear to me:

If I add multiple addresses from the same prefix to the same interface, why is the route shown as reachable n times the number of addresses the interface has?
IPv6Routes1.png
I just came to this because of adding the second IP address (2001:4bb8:248:2868::1/64), as the cAP AC says the first address (2001:4bb8:248:2868::/64) is NOT in the same subnet as it’s own address (2001:4bb8:248:2868::251/64), while the second address is…
IPv6Routes3.png
IPv6Routes2.png

Haven’t had the best sleep last night, so am I missing something here?


Greets,
uebi

2001:4bb8:248:2868::/64 is a network address (similar to aaa.bb.cc.0/24 in IPv4) do setting this address as host address is invalid (ability to set it anyway is a bug, but then lots of IPv6 implementation in current ROS is buggy).

Well, not a network address like in IPv4, but a Router Anycast address, according to RFC 4291. I’m just wondering about this implementation, as MikroTik assigns this address itself (maybe 1 might be better), but a MikroTik “Client” says it is unreachable, although it is not. Technically it is on the same subnet, but as mentioned before, it’s an Anycast address?!

How did router assign the anycast address to itself? If I issue

/ipv6 address
add interface=ether1 from-pool=<IPv6 pool name>

then it will auto select a valid unicast address …

If you set it manually (as in

/ipv6 address set interface=ether1 address=aaaa:bbbb::/64

) then as I said, IPv6 in current ROS is full of missing features and (minor) bugs.

When the wAP connects to LTE, it receives the prefix via RA and assigns it to the interface selected in the APN config, but as 0.

Besides that I have seen the same behavior with PD, but as there have been additional issues a case is open regarding the DHCPv6 problems. Anyway, with the exact same config as you mentioned it also assigned a 0 address (6.44.5).