Messages in this thread | | | From | Vincent Guittot <> | Date | Tue, 13 Sep 2022 10:31:31 +0200 | Subject | Re: [PATCH v3 6/8] sched/fair: Add sched group latency support |
| |
On Tue, 13 Sept 2022 at 00:18, Tejun Heo <tj@kernel.org> wrote: > > Hello, > > On Fri, Sep 09, 2022 at 03:03:07PM +0200, Vincent Guittot wrote: > > + { > > + .name = "latency", > > + .read_s64 = cpu_latency_read_s64, > > + .write_s64 = cpu_latency_write_s64, > > + }, > > You're still using the nice value here, right? If so, can you please use the > filename "latency.nice" so that it's consistent with "weight.nice"?
This cpu.latency is not a nice priority but the signed offset used by the scheduler at wakeup. On previous version you raised concern about having a nice value for cgroup so I removed it for cgroup and directly exposed the latency offset with cpu.latency similarly to the weight that is exposed with cpu.shares.
Vincent > > Thanks. > > -- > tejun
| |