Re: [RESEND][PATCH v4] cgroup: Use CAP_SYS_RESOURCE to allow a process to migrate other tasks between cgroups

From: John Stultz
Date: Fri Dec 09 2016 - 00:39:46 EST


On Tue, Dec 6, 2016 at 10:23 AM, Tejun Heo <tj@xxxxxxxxxx> wrote:
> Hello,
>
> On Tue, Dec 06, 2016 at 10:13:53AM -0800, Andy Lutomirski wrote:
>> > Delegation is an explicit operation and reflected in the ownership of
>> > the subdirectories and cgroup interface files in them. The
>> > subhierarchy containment is achieved by requiring the user who's
>> > trying to migrate a process to have write perm on cgroup.procs on the
>> > common ancestor of the source and target in addition to the target.
>>
>> OK, I see what you're doing. That's interesting.
>
> It's something born out of usages of cgroup v1. People used it that
> way (chowning files and directories) and combined with the uid checksn
> it yielded something which is useful sometimes, but it always had
> issues with hierarchical behaviors, which files to chmod and the weird
> combination of uid checks. cgroup v2 has a clear delegation model but
> the uid checks are still left in as not changing was the default.
>
> It's not necessary and I'm thinking about queueing something like the
> following in the next cycle.
>
> As for the android CAP discussion, I think it'd be nice to share an
> existing CAP but if we can't find a good one to share, let's create a
> new one.

So just to clarify the discussion for my purposes and make sure I
understood, per-cgroup CAP rules was not desired, and instead we
should either utilize an existing cap (are there still objections to
CAP_SYS_RESOURCE? - this isn't clear to me) or create a new one (ie,
bring back the older CAP_CGROUP_MIGRATE patch).

Tejun: Do you have a more finished version of your patch that I should
add my changes on top of?

thanks
-john