If people want small jobs to get their feet or other appendages wet in FreeBSD here's some random ideas:
- look through the tests to ensure that they use unique jail names. Reference https://cgit.freebsd.org/src/commit/?id=e9264b5681289f925c6daca9f31489173b625591
- help convert pf to use netlink for configuration. Reference https://cgit.freebsd.org/src/commit/?id=777a4702c591154c5a844d43c32f588f371ae80a
- More tests for pf. If you have a pet scenario you really care about that's not already covered that's an excellent way to start.

#FreeBSD

src - FreeBSD source tree

@kp My pet scenario is that I'd love it if enabling pfsync defer didn't reboot our busy pf nodes.

But I have been unable to reproduce on non-production machines. So I don't know what causes it. I have been trying to write some tuya tests to tickle it out but no luck so far.

I have no idea where to begin but I would LOVE to get it fixed. Any pointers? Where would you begin?

Thanks! :)

@tykling A backtrace would be hugely helpful here. As it stands I have nothing to go on.

@kp I agree, but all I get is "oh, server no longer responds" and then some minutes later "oh, it responds again now, uptime 1 minute".

If I enable defer in rc.conf it I have to disable it in single user because it reboots before I have a chance to SSH in.

@tykling There's almost certainly going to be crash information in /var/crash. If not, read https://docs.freebsd.org/en/books/developers-handbook/kerneldebug/ so you can get that information next time it happens.
Chapter 10. Kernel Debugging

FreeBSD Kernel Debugging

FreeBSD Documentation Portal

@kp

Wouldn’t it make sense if the jail names had some random component, as in:

jname="v4t-output_udp_flowid_mpath_success-`jot -r 1 1 1000`"
@kravietz I don’t think that’s helpful. We’ve mostly named them after the test name, and have used static names. There’s no need to support running the same test more than once at the same time, just to avoid conflicts between tests.

@kp

Yeah, I’ve just scanned the .sh files there and there don’t seem to be any conflicts after your commit :)

@kravietz There ought to be a fair few in the pf tests. I've pretty consistently used Alcatraz and singsing as jail name. There may also be tests that use epair interfaces from the host, which will also conflict during parallel testing.

@kp

So I’ve submitted a few man page fixes instead :)