> This seems reasonable and I'll accept into proposed.

Thank you.

> However, I think the regression analysis and test plan need some work. I'm 
> terrified of inadvertently breaking nftables generally for all users such 
> that they cannot receive a future update, so that's a pretty serious "Where 
> problems could occur" for any nftables SRU really. Fortunately the patch 
> seems straightforward.
> Maybe enhance the Test Plan please to ensure that nftables still works as 
> expected in a regular scenario, as well as specifically looking for the crash 
> as you're doing?

One good thing we can do is extend the phasing time for the update once
it lands in updates. Because a test plan is not comprehensive of real
world usage.

It would be good if Canonical kubernetes / juju / openstack products can
be tested with nftables from proposed as that would be likely the
heaviest usage of nftables in terms of test suites that we have.

> If you have confidence that the patch is itself trivial some analysis
of that would be helpful too - for example is it a rare code path? Is it
possible that the early return you're adding catches too much and will
break some user, or if not how are you confident that's the case?

What saves us for now that indeed is that the codepath in question is
likely to be rare.

As it is typically triggered only when mixed-versions of nftables got
used on the host (very new ones like 1.1.6 from resolute from a
container, on the jammy/nobble host). Hence why this bug has lingered
for so long.

But now this bug is starting to creep up as people deploy newer
containers on stable hosts. Hence it became more urgent.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2142552

Title:
  netlink: fix crash when ops doesn't support udata

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/nftables/+bug/2142552/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to