Re: [PATCH 0/8] userns: per-namespace capability bounding set and opt-in netns loopback
From: Hallgrimur H. Gunnarsson
Date: Mon Nov 16 2026 - 11:43:13 EST
On Wed, Nov 11, 2026 at 04:20:02PM +0100, Christian Brauner wrote:
> Just a note on the essay I cut off. I really object to any arguments
> that amount to "it's too much work" propped up by a few thin arguments
> sprinkled in such as rtnetlink.
>
> I also reject any arguments that amount to underscoping and overscoping
> of the new capability. The scope is clear: empty netns with brought up
> loop device.
Understood, thanks. I'll split the loopback work out into its own
standalone series and handle it separately, keeping the scope strictly
defined as you described.
> But again, I'm more focussed on cells so here's what I find acceptable:
[snip]
> * a new CapUns: line in /proc/<pid>/status
> * new audit mask in AUDIT_CAPSET record
Quick question on AUDIT_CAPSET: today it's only emitted on capset() and
only logs capability sets that capset() can modify (explicitly or
implicitly like cap_ambient). It does not include cap_bset which is
modified via prctl(). Since capset() never modifies cap_userns either,
does it make sense to treat it like cap_bset and leave it out of
AUDIT_CAPSET, relying on prctl auditing instead?
[snip]
> That gets rid of the sysctl, the clone3() field and the clone flags,
> doesn't need a new capability bit, and nothing changes for anyone who
> doesn't opt in, I think.
>
> And that loopback thing should be its own series. Not as an nsfs ioctl
> though. All nsfs ioctls only report namespace state and I'd like to keep
> it that way.
>
> Pipe this through an LLM and make sure it makes sense and covers all
> holes.
>
> Thoughts?
Thanks for the feedback and the detailed proposal. It looks good to me
and nicely ties everything together. I'll prepare a V2 for review based
on this hybrid design.