[PATCH v2 0/4] BGP unnumbered automatic peering based on IPv6 ND

[Ondrej Zajicek]

On Mon, Jan 12, 2026 at 11:40:43AM +0100, Matteo Perin via Bird-users wrote:
> Good morning,
> 
> I just wanted to make sure you received the v2 of the patch correctly.
> I just realized that maybe something went wrong when I sent the email since
> it included changes on two different repositories and I sent it as a
> response to the same thread as the original PR.
> In that case, I will promptly resend the updated version with any changes
> needed.

Hello

I finally got to finish reviewing and merging this patchset, sorry for delay.

Here are the modified versions:

https://gitlab.nic.cz/labs/bird/-/commits/dynamic-nbrs-2
https://gitlab.nic.cz/labs/bird/-/commit/4327fabff1db9b78e5e80bb8d22dc848fbf9a6b9
https://gitlab.nic.cz/labs/bird/-/commit/7eda3e381414f97caed5b2cd02d6b4b3b4554ee0
https://gitlab.nic.cz/labs/bird/-/commit/0ee9f93bd076c5cc425ceaec9acedbbb7c9021ec

(They are backported to BIRD v2, i will port them to BIRD v3 soon)

The most visible change is naming, using neighbor instead of peer. That
is consistent with how equivalent structures are named in other
protocols in peer.

Otherwise in the first two patches (Nest and RADv) there are
just minor changes:

Nest:
- Neighbor entry formatting change
- Parser grammar for neighbor entries
- Filter tests
- Minor fixes
- Some documentation

RAdv:
- Moving neighbor prune to separate timer unrelated to ifaces
- More idiomatic log messages, incluing TBF for warning
- Reusing existing 'lifetime' attribute for neighbor entries
  (it still stores expire time but plan to change it in the future)

With the BGP patch the situation was more complicated, i noticed several
bugs in our dynamic BGP code in BIRD 3 and some parts of your patch looked
like a workaround for some of these issues.

- Refactoring BGP spawning code to avoid duplication

- No disable and re-spawn of existing spawned protocols,
  this should not be necessary. We just claim existing sessions
  in this case.

- I removed the 'persist' option. We just keep track whether the session
  is claimed (there is existing neighbor entry) or unclaimed. This is
  just makeshift now, there is a good question how to properly handle
  unclaimed sessions, but i think it should be consistent with how we
  handle dynamic BPG sessions from incoming connections, so i would prefer
  to keep it simple now.

- Add some implicit neighbor filtering in preexport hook

- Simplified some parts of channel handling

- The patch is for BIRD 2 so there is no inter-thread messaging code or
  locking.


Please take a look at these patches and reply if you are okay with them
after my modifications. If you have any questions, feel free to ask.

-- 
Elen sila lumenn' omentielvo

Ondrej 'Santiago' Zajicek (email: santiago at crfreenet.org)
"To err is human -- to blame it on a computer is even more so."