For perf record -b, which requires the pmu::sched_task callback the
current code is rather expensive:

     7.68%  sched-pipe  [kernel.vmlinux]    [k] perf_pmu_sched_task
     5.95%  sched-pipe  [kernel.vmlinux]    [k] __switch_to
     5.20%  sched-pipe  [kernel.vmlinux]    [k] __intel_pmu_disable_all
     3.95%  sched-pipe  perf                [.] worker_thread

The problem is that it will iterate all registered PMUs, most of which
will not have anything to do. Avoid this by keeping an explicit list
of PMUs that have requested the callback.

The perf_sched_cb_{inc,dec}() functions already take the required pmu
argument, and now that these functions are no longer called from NMI
context we can use them to manage a list.

With this patch applied the function doesn't show up in the top 4
anymore (it dropped to 18th place).

     6.67%  sched-pipe  [kernel.vmlinux]    [k] __switch_to
     6.18%  sched-pipe  [kernel.vmlinux]    [k] __intel_pmu_disable_all
     3.92%  sched-pipe  [kernel.vmlinux]    [k] switch_mm_irqs_off
     3.71%  sched-pipe  perf                [.] worker_thread

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 include/linux/perf_event.h |    3 +++
 kernel/events/core.c       |   43 ++++++++++++++++++++++++-------------------
 2 files changed, 27 insertions(+), 19 deletions(-)

--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -757,6 +757,9 @@ struct perf_cpu_context {
 
 	struct pmu			*unique_pmu;
 	struct perf_cgroup		*cgrp;
+
+	struct list_head		sched_cb_entry;
+	int				sched_cb_usage;
 };
 
 struct perf_output_handle {
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2777,13 +2777,26 @@ static void perf_event_context_sched_out
 	}
 }
 
+static DEFINE_PER_CPU(struct list_head, sched_cb_list);
+
 void perf_sched_cb_dec(struct pmu *pmu)
 {
+	struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+
 	this_cpu_dec(perf_sched_cb_usages);
+
+	if (!--cpuctx->sched_cb_usage)
+		list_del(&cpuctx->sched_cb_entry);
 }
 
+
 void perf_sched_cb_inc(struct pmu *pmu)
 {
+	struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+
+	if (!cpuctx->sched_cb_usage++)
+		list_add(&cpuctx->sched_cb_entry, this_cpu_ptr(&sched_cb_list));
+
 	this_cpu_inc(perf_sched_cb_usages);
 }
 
@@ -2801,34 +2814,24 @@ static void perf_pmu_sched_task(struct t
 {
 	struct perf_cpu_context *cpuctx;
 	struct pmu *pmu;
-	unsigned long flags;
 
 	if (prev == next)
 		return;
 
-	local_irq_save(flags);
-
-	rcu_read_lock();
-
-	list_for_each_entry_rcu(pmu, &pmus, entry) {
-		if (pmu->sched_task) {
-			cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
-
-			perf_ctx_lock(cpuctx, cpuctx->task_ctx);
+	list_for_each_entry(cpuctx, this_cpu_ptr(&sched_cb_list), sched_cb_entry) {
+		pmu = cpuctx->unique_pmu; /* software PMUs will not have sched_task */
 
-			perf_pmu_disable(pmu);
+		if (WARN_ON_ONCE(!pmu->sched_task))
+			continue;
 
-			pmu->sched_task(cpuctx->task_ctx, sched_in);
+		perf_ctx_lock(cpuctx, cpuctx->task_ctx);
+		perf_pmu_disable(pmu);
 
-			perf_pmu_enable(pmu);
+		pmu->sched_task(cpuctx->task_ctx, sched_in);
 
-			perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
-		}
+		perf_pmu_enable(pmu);
+		perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
 	}
-
-	rcu_read_unlock();
-
-	local_irq_restore(flags);
 }
 
 static void perf_event_switch(struct task_struct *task,
@@ -10337,6 +10340,8 @@ static void __init perf_event_init_all_c
 
 		INIT_LIST_HEAD(&per_cpu(pmu_sb_events.list, cpu));
 		raw_spin_lock_init(&per_cpu(pmu_sb_events.lock, cpu));
+
+		INIT_LIST_HEAD(&per_cpu(sched_cb_list, cpu));
 	}
 }