Messages in this thread | | | Date | Wed, 21 Sep 2022 17:07:38 +0100 | From | Qais Yousef <> | Subject | Re: [PATCH v4 6/8] sched/fair: Add sched group latency support |
| |
On 09/19/22 07:34, Tejun Heo wrote: > On Mon, Sep 19, 2022 at 05:49:27PM +0200, Vincent Guittot wrote: > > > For `nice` we have `cpu.weight.nice` next to `cpu.weight` in cgroup v2 ? > > > > If everybody is ok, I can add back the cpu.latency.nice interface in > > the v5 in addition to the cpu.latency > > Yeah, that sounds fine to me.
I do have concerns about cpu.latency knob as it exposes latency_offset which won't be meaningful for all consumers of latency_nice [1].
The current use case proposed by Vincent is not going to be the only consumer of this interface. The concept of converting this latency_nice value to weight in similar fashion to existing nice value is specific to it. In previous discussion this conversion was not required and I'd expect it to still not be required for those other use cases.
Wouldn't cpu.latency.nice be enough? I think the latency_offset is implementation detail that userspace shouldn't be concerned about.
[1] https://lwn.net/Articles/820659/
Cheers
-- Qais Yousef
| |