Weird behavior regarding xxxx:xxxx:xxxx:xxxx::/128

RouterOS General
1

I set up a clean 7.23.1 environment and configured the following simple addresses:

/ipv6 address
add address=fe80::cc:0:0:18 advertise=no auto-link-local=no interface=lan
# The result below will be fd00:0:0:cc::18/64, although I believe it should be fd00:0:0:cc::18/128
add address=fd00:0:0:cc::18 advertise=no interface=lan

Then, I pinged fd00:0:0:cc:: from the router itself, and got the following result:

image
image575×114 10.5 KB

It seems like it's resolving to its own address. This causes a problem. For a DHCPv6 stateful server, if a pool like fd00:0:0:cc::/64 is configured, the first client to get an address from this pool will receive fd00:0:0:cc::/128. RouterOS will then consider this address as its own, resulting in abnormal communication for the client.

So, is this a bug?

Also, just two quick questions regarding the DHCPv6 pool:

  1. Is there a way to reserve specific addresses in the pool so they won't be handed out? I know static binding is one way, but it's too inefficient.

  2. It seems ROS always allocates addresses sequentially starting from 0. Is there a way to make it start from a specific address and allocate them randomly? For example, starting with 10f, and the next one being a random address like 20a, etc.

Thank you.

2

I believe the netmask should also be included for the address to function as intended:

/ipv6 address
add address=fd00:0:0:cc::18/64 advertise=no interface=lan

As for your questions, there aren't specific options covering your needs that I'm aware of, so no to both of them

3

Regarding your second question, you can create a particular sub-pool from existing pool and delegate addresses from it.

4

For example, take a sub-pool like fd00:0:0:cc:dc01::/80. The first address allocated from it would still be fd00:0:0:cc:dc01:0:0:0/128, right?

5

Nope, addresses taken from pool, will have same prefix length as pool itself.

i.e.fd00:0:0:cc:dc01:0:0:0/80

Unless you explicitly set prefix length yourself.

6

well, same thing
dozen of VLAN's, every vlan with ia_na dhcpv6 server, and in every vlan pool need to manualy reserve ::0 and ::1 addres to avoid problems
luckily, i get static v6/56 so few minute script for 24pcs of vlan solve this, but ... it is still manual process

7

Since this pool is for stateful DHCPv6, the prefix length has to be set to 128.

8

I really hope MikroTik can add support for reserved addresses in the near future.

9

I've submitted a ticket: SUP-219906.