Re: [RFC PATCH 0/7] cgroup: charge kernel work to the cgroup it is done for

[Tejun Heo <tj@kernel.org>]

Hello, Shakeel.

On Thu, Sep 24, 2026 at 02:11:00PM -0700, Shakeel Butt wrote:
> On the modern NICs where the workloads get their own rx queues, I think the
> proposed mechanism can help to do the accurate accounting (I have to extend this
> to softirqs as it is limited to kthreads atm) very easily.

Rx steering is useful for partitioning the whole machine, but it doesn't
seem like the right tool for attributing CPU cycles to each cgroup. It's
cumbersome to set up and the granularity doesn't match. We aren't going
to be steering net rx processing per cgroup.

> If you don't mind, I think having solution for modern NICs i.e. dedicated rx
> queues, should suffice for now (unless you want the solution for old NICs as
> well). Let me know what you think.

This doesn't have to be exact or completely synchronous. I don't see why
net rx charging and reclaim CPU charging can't share the same mechanism,
where the cycles are force charged to the cgroup after the fact.

I don't think PSI needs to be involved either. Shared execution in
itself shouldn't create pressure. The forced charge changes the cgroup's
usage, and that usage then creates pressure in the cgroup. IO works the
same way.

Thanks.

--
tejun