Ok. so deploying some IPv6 over a few routers and getting OSPFv3 going.
Versions tested on: CCR 6.37.5 and 6.38.5

Steps followed on each router:
1. Add loopback bridge.
2. Add mac-address to loopback bridge (Unique to each bridge) example: 01:00:00:00:01:BB
3. Added /64 v6 ip to loopback bridge. Each router with own loopback bridge on separate /64 example: xxxx:xxx:1:2::1/64, xxxx:xxx:1:3::1/64
4. Added the point to point IPv6 subnets between routers. (Tried with /127, /126 and /64's same result)
/routing ospf-v3 route print, all expected routes shows up correctly.
These routes never ends up in IPv6 routing table as expected with correct Gateway (fe80:: type though), interface, cost and states.

These routes never ends up in the IPv6 Routing table.
Pinging various loopback interfaces does not work. If static routes are added, icmp works.
No firewall rules, ipv4 or 6

IPv6 BGP routes does work and shows up in IPv6 routing table.
Any ideas??

Kind regards and thank you for your time.

We are seeing this problem on RB1100AHx2es with 6.38.3 and 6.38.5.

Update: Even upgrading to the latest 6.39.2 does not resolve this issue, but only on RB1100AHx2 routers. Other routers (including other PowerPC routers like 850Gx2) are not exhibiting this problem.

Anything else that might cause this problem in the configuration would be useful to know but it seems like a bug on RB1100AHx2s.

What routers are you using OP?

That's an interesting point, have you tried replicating the topology with CHRs in a virtual environment like GNS3? it would be interesting to see if it works there

There's no ospfv3 route filtering in ros v6 so the routes should just populate into the table AFAIK. Seems like it has to be a bug.

Problem solved for me. Route filtering DOES apply using the default ospf-in filter chain. Not sure if it uses the regular ospf filter chain setting or a hidden default.

Adding a filter to accept the routes in the routing filter chain ospf-in fixed it.

@blackmesawireless
Yup. Solved. As, you said, OSPFv3 uses default OSPF-IN/OUT.

This is supposed not to work, but yes, filters does seem to work with IPv6.
Thanx man.

@blackmesawireless
This is supposed not to work, but yes, filters does seem to work with IPv6.
Thanx man.

Glad I could help. For once what seemed to be a bug truly is an undocumented feature! Ha!

Good find...thanks for the info!

In addition, I'd recommend dropping the /64's from your loopbacks. I know IPv6 = all addresses are free now but still...

Personally I use /128 for loopbacks (like /32 w/IPv4) and /126 (like /30 w/IPv4) for point-to-points. It's easier for me to relate and definitely stresses a large network plan a lot less. I chose not to use /127 as I could find less evidence that it as widely supported (similar to /31's). That and I like my even numbers, it calms my OCD.

Also, mac sure your loopback bridges have MAC addresses. A generous and intelligent user helped me figure this out when I got my first RouterBoard.

viewtopic.php?f=2&t=121263#p596222

What is the point of the loopback bridge? Anyone have link to docs on this?

Look up "OSPF" and "loopback"

BTW while /128s are considered OK for loopbacks, it's currently recommended to allocate an entire /64 for PtP links, even if you only assign it as a smaller subnet. IPv6 assumes an interface will have 64 host bits under most circumstances.

"No, it is not necessary, but it is recommended." - Mikrotik Employee

I was just wondering because I've had it working without one. But I do see the benefits now in a bigger environment.

I was just wondering because I've had it working without one. But I do see the benefits now in a bigger environment.

Yes, when you have a site with several different links out in an OSPF network, you really need a stable address for things like VPLS connections and (if you use it) iBGP. The loopback gives you that.