CORE bridging setup still in need of improvement

[Patrick M. Hausen]

Hi all,

after repeatedly hinting people at "Disable Hardware Offloading" when bridging, i.e. plugins, jails and VMs are involved, I really think we need to either make this automatic of at least try to "solve it in documentation". This setting needs to be featured way more prominently somewhere.

Second still to this day the automatic bridge creation of TN CORE is plain wrong. Sorry for being so blunt. I'm German

if you have a physical interface or a VLAN with active IPv4 and possibly IPv6 addresses, putting this interface as a member into a bridge without moving all IP configuration to the bridge interface will break multicast, i.e. most of IPv6 ...

Please see comments 5 and 6 of this bug report:

247912 – [if_bridge] IPv6 ndp does not work across local bridge members

bugs.freebsd.org

I take Kristof's statement as authoritative, since he's the current primary developer for that subsystem and additionally he does offer a concise explanation of the implementation. And it has been in the FreeBSD handbook ever since bridge(4) was introduced.

The IP configuration must go on the bridge interface! Like in RFC terminology "must".

I remember that I did create an issue about that but cannot find it in JIRA. Possibly that was so long ago it was in Redmine ...


What can I do to help? This breaks things for users. I do know my way around FreeBSD networking and create all my lagg, vlan and bridge interfaces explicitly before assigning any jails or VMs, but IMHO we cannot expect that from average users. Not even with documentation. This is complex - see screenshot.



Kind regards,
Patrick

 

[Kris Moore]

With a ticket that old it may have indeed gotten lost in the shuffle of the past many years. I'd say a new suggestion / bug ticket on Jira would be warranted here, something we can track and then look to fix from the docs side as well as product.

 

[Constantin]

Yes, please. I remembered struggling with this also until I somehow succeeded.