sched/psi: Per-cgroup PSI accounting disable/re-enable interface

PSI accounts stalls for each cgroup separately and aggregates it
at each level of the hierarchy. This may cause non-negligible overhead
for some workloads when under deep level of the hierarchy.

commit 3958e2d0c34e ("cgroup: make per-cgroup pressure stall tracking configurable")
make PSI to skip per-cgroup stall accounting, only account system-wide
to avoid this each level overhead.

But for our use case, we also want leaf cgroup PSI stats accounted for
userspace adjustment on that cgroup, apart from only system-wide adjustment.

So this patch introduce a per-cgroup PSI accounting disable/re-enable
interface "cgroup.pressure", which is a read-write single value file that
allowed values are "0" and "1", the defaults is "1" so per-cgroup
PSI stats is enabled by default.

Implementation details:

It should be relatively straight-forward to disable and re-enable
state aggregation, time tracking, averaging on a per-cgroup level,
if we can live with losing history from while it was disabled.
I.e. the avgs will restart from 0, total= will have gaps.

But it's hard or complex to stop/restart groupc->tasks[] updates,
which is not implemented in this patch. So we always update
groupc->tasks[] and PSI_ONCPU bit in psi_group_change() even when
the cgroup PSI stats is disabled.

Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lkml.kernel.org/r/20220907090332.2078-1-zhouchengming@bytedance.com

1 files changed, 63 insertions, 7 deletions

diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
index 9a8aee80a087..9711827e31e5 100644
--- a/kernel/sched/psi.c
+++ b/kernel/sched/psi.c
@@ -181,6 +181,7 @@ static void group_init(struct psi_group *group)
 {
 	int cpu;
 
+	group->enabled = true;
 	for_each_possible_cpu(cpu)
 		seqcount_init(&per_cpu_ptr(group->pcpu, cpu)->seq);
 	group->avg_last_update = sched_clock();
@@ -696,17 +697,16 @@ static void psi_group_change(struct psi_group *group, int cpu,
 	groupc = per_cpu_ptr(group->pcpu, cpu);
 
 	/*
-	 * First we assess the aggregate resource states this CPU's
-	 * tasks have been in since the last change, and account any
-	 * SOME and FULL time these may have resulted in.
-	 *
-	 * Then we update the task counts according to the state
+	 * First we update the task counts according to the state
 	 * change requested through the @clear and @set bits.
+	 *
+	 * Then if the cgroup PSI stats accounting enabled, we
+	 * assess the aggregate resource states this CPU's tasks
+	 * have been in since the last change, and account any
+	 * SOME and FULL time these may have resulted in.
 	 */
 	write_seqcount_begin(&groupc->seq);
 
-	record_times(groupc, now);
-
 	/*
 	 * Start with TSK_ONCPU, which doesn't have a corresponding
 	 * task count - it's just a boolean flag directly encoded in
@@ -745,6 +745,23 @@ static void psi_group_change(struct psi_group *group, int cpu,
 		if (set & (1 << t))
 			groupc->tasks[t]++;
 
+	if (!group->enabled) {
+		/*
+		 * On the first group change after disabling PSI, conclude
+		 * the current state and flush its time. This is unlikely
+		 * to matter to the user, but aggregation (get_recent_times)
+		 * may have already incorporated the live state into times_prev;
+		 * avoid a delta sample underflow when PSI is later re-enabled.
+		 */
+		if (unlikely(groupc->state_mask & (1 << PSI_NONIDLE)))
+			record_times(groupc, now);
+
+		groupc->state_mask = state_mask;
+
+		write_seqcount_end(&groupc->seq);
+		return;
+	}
+
 	for (s = 0; s < NR_PSI_STATES; s++) {
 		if (test_state(groupc->tasks, s, state_mask & PSI_ONCPU))
 			state_mask |= (1 << s);
@@ -761,6 +778,8 @@ static void psi_group_change(struct psi_group *group, int cpu,
 	if (unlikely((state_mask & PSI_ONCPU) && cpu_curr(cpu)->in_memstall))
 		state_mask |= (1 << PSI_MEM_FULL);
 
+	record_times(groupc, now);
+
 	groupc->state_mask = state_mask;
 
 	write_seqcount_end(&groupc->seq);
@@ -907,6 +926,9 @@ void psi_account_irqtime(struct task_struct *task, u32 delta)
 
 	group = task_psi_group(task);
 	do {
+		if (!group->enabled)
+			continue;
+
 		groupc = per_cpu_ptr(group->pcpu, cpu);
 
 		write_seqcount_begin(&groupc->seq);
@@ -1080,6 +1102,40 @@ void cgroup_move_task(struct task_struct *task, struct css_set *to)
 
 	task_rq_unlock(rq, task, &rf);
 }
+
+void psi_cgroup_restart(struct psi_group *group)
+{
+	int cpu;
+
+	/*
+	 * After we disable psi_group->enabled, we don't actually
+	 * stop percpu tasks accounting in each psi_group_cpu,
+	 * instead only stop test_state() loop, record_times()
+	 * and averaging worker, see psi_group_change() for details.
+	 *
+	 * When disable cgroup PSI, this function has nothing to sync
+	 * since cgroup pressure files are hidden and percpu psi_group_cpu
+	 * would see !psi_group->enabled and only do task accounting.
+	 *
+	 * When re-enable cgroup PSI, this function use psi_group_change()
+	 * to get correct state mask from test_state() loop on tasks[],
+	 * and restart groupc->state_start from now, use .clear = .set = 0
+	 * here since no task status really changed.
+	 */
+	if (!group->enabled)
+		return;
+
+	for_each_possible_cpu(cpu) {
+		struct rq *rq = cpu_rq(cpu);
+		struct rq_flags rf;
+		u64 now;
+
+		rq_lock_irq(rq, &rf);
+		now = cpu_clock(cpu);
+		psi_group_change(group, cpu, 0, 0, now, true);
+		rq_unlock_irq(rq, &rf);
+	}
+}
 #endif /* CONFIG_CGROUPS */
 
 int psi_show(struct seq_file *m, struct psi_group *group, enum psi_res res)