Re: [PATCH v2 1/1] cgroup: make per-cgroup pressure stall tracking configurable

From: Suren Baghdasaryan
Date: Tue May 18 2021 - 14:56:26 EST


On Tue, May 18, 2021 at 11:52 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Mon, May 17, 2021 at 07:02:00PM -0700, Suren Baghdasaryan wrote:
>
> > diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
> > index cc25a3cff41f..4b8e72640ac9 100644
> > --- a/kernel/sched/psi.c
> > +++ b/kernel/sched/psi.c
> > @@ -148,6 +148,7 @@
> > static int psi_bug __read_mostly;
> >
> > DEFINE_STATIC_KEY_FALSE(psi_disabled);
> > +DEFINE_STATIC_KEY_FALSE(psi_cgroups_disabled);
>
> I'm thinking the whole thing will be easier/clearer when you make this:
>
> DEFINE_STATIC_KEY_TRUE(psi_cgroups_enabled);
>

Sounds good. Will respin another version. Thanks for reviewing!


> >
> > #ifdef CONFIG_PSI_DEFAULT_DISABLED
> > static bool psi_enable;
> > @@ -211,6 +212,9 @@ void __init psi_init(void)
> > return;
> > }
> >
> > + if (!cgroup_psi_enabled())
> > + static_branch_enable(&psi_cgroups_disabled);
>
> if (!cgroup_psi_enabled())
> static_branch_disable(&psi_cgroups_enabled);
>
> > +
> > psi_period = jiffies_to_nsecs(PSI_FREQ);
> > group_init(&psi_system);
> > }
> > @@ -744,23 +748,23 @@ static void psi_group_change(struct psi_group *group, int cpu,
> >
> > static struct psi_group *iterate_groups(struct task_struct *task, void **iter)
> > {
> > + if (*iter == &psi_system)
> > + return NULL;
> > +
> > #ifdef CONFIG_CGROUPS
> > + if (!static_branch_likely(&psi_cgroups_disabled)) {
>
> if (static_branch_likely(&psi_cgroups_enabled)) {
>
> > + struct cgroup *cgroup = NULL;
> >
> > + if (!*iter)
> > + cgroup = task->cgroups->dfl_cgrp;
> > + else
> > + cgroup = cgroup_parent(*iter);
> >
> > + if (cgroup && cgroup_parent(cgroup)) {
> > + *iter = cgroup;
> > + return cgroup_psi(cgroup);
> > + }
> > }
> > #endif
> > *iter = &psi_system;
> > return &psi_system;
>
> But yes, very nice.
>
> --
> To unsubscribe from this group and stop receiving emails from it, send an email to kernel-team+unsubscribe@xxxxxxxxxxx.
>