@christos You tagged me to review, but I do not have strong opinion for this change, really :)

In D47027#1072084, @markj wrote:

In D47027#1072077, @olce wrote:

Shouldn't we simply omit outputting ip4.addr (resp., ipv6.addr) if pr->pr_addrs[PR_INET] (resp., pr->pr_addrs[PR_INET6]), is NULL?

vfs_setopt_part() has the side effect of setting the key length to 0, even though there's nothing to copy. I presume that we want to preserve that behaviour.

Rebased to latest main.

I always have a GENERIC-DEBUG for stable branches, the content is similar. +1 for this.

No objection from me, since those are verbose message that is mostly useful for debugging.

I'd recommend to add test plan to do a quick demonstration for the good effect of this change. Then it will help reviewing.

In D45727#1066919, @ae wrote:

You probably can directly call similar to bpf_ifdetach() function from if_vmove(). It is called from ioctl context, so you can make detaching synchronously.

In D46749#1068241, @emaste wrote:

I'd probably use the full word global rather than glob, as glob has an existing meaning (even if not related to this context) and it's only two extra letters.

The whitespace fixup should be a separate commit, that can go first.