drivers/perf: arm_pmu: implement CPU_PM notifier

When a CPU is suspended (either through suspend-to-RAM or CPUidle),
its PMU registers content can be lost, which means that counters
registers values that were initialized on power down entry have to be
reprogrammed on power-up to make sure the counters set-up is preserved
(ie on power-up registers take the reset values on Cold or Warm reset,
which can be architecturally UNKNOWN).

To guarantee seamless profiling conditions across a core power down
this patch adds a CPU PM notifier to ARM pmus, that upon CPU PM
entry/exit from low-power states saves/restores the pmu registers
set-up (by using the ARM perf API), so that the power-down/up cycle does
not affect the perf behaviour (apart from a black-out period between
power-up/down CPU PM notifications that is unavoidable).

Cc: Will Deacon <will.deacon@arm.com>
Cc: Sudeep Holla <sudeep.holla@arm.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Ashwin Chaugule <ashwin.chaugule@linaro.org>
Acked-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

drivers/perf/arm_pmu.c

@@ -13,6 +13,7 @@
 
 #include <linux/bitmap.h>
 #include <linux/cpumask.h>
+#include <linux/cpu_pm.h>
 #include <linux/export.h>
 #include <linux/kernel.h>
 #include <linux/of_device.h>
@@ -710,6 +711,93 @@ static int cpu_pmu_notify(struct notifier_block *b, unsigned long action,
 	return NOTIFY_OK;
 }
 
+#ifdef CONFIG_CPU_PM
+static void cpu_pm_pmu_setup(struct arm_pmu *armpmu, unsigned long cmd)
+{
+	struct pmu_hw_events *hw_events = this_cpu_ptr(armpmu->hw_events);
+	struct perf_event *event;
+	int idx;
+
+	for (idx = 0; idx < armpmu->num_events; idx++) {
+		/*
+		 * If the counter is not used skip it, there is no
+		 * need of stopping/restarting it.
+		 */
+		if (!test_bit(idx, hw_events->used_mask))
+			continue;
+
+		event = hw_events->events[idx];
+
+		switch (cmd) {
+		case CPU_PM_ENTER:
+			/*
+			 * Stop and update the counter
+			 */
+			armpmu_stop(event, PERF_EF_UPDATE);
+			break;
+		case CPU_PM_EXIT:
+		case CPU_PM_ENTER_FAILED:
+			 /* Restore and enable the counter */
+			armpmu_start(event, PERF_EF_RELOAD);
+			break;
+		default:
+			break;
+		}
+	}
+}
+
+static int cpu_pm_pmu_notify(struct notifier_block *b, unsigned long cmd,
+			     void *v)
+{
+	struct arm_pmu *armpmu = container_of(b, struct arm_pmu, cpu_pm_nb);
+	struct pmu_hw_events *hw_events = this_cpu_ptr(armpmu->hw_events);
+	int enabled = bitmap_weight(hw_events->used_mask, armpmu->num_events);
+
+	if (!cpumask_test_cpu(smp_processor_id(), &armpmu->supported_cpus))
+		return NOTIFY_DONE;
+
+	/*
+	 * Always reset the PMU registers on power-up even if
+	 * there are no events running.
+	 */
+	if (cmd == CPU_PM_EXIT && armpmu->reset)
+		armpmu->reset(armpmu);
+
+	if (!enabled)
+		return NOTIFY_OK;
+
+	switch (cmd) {
+	case CPU_PM_ENTER:
+		armpmu->stop(armpmu);
+		cpu_pm_pmu_setup(armpmu, cmd);
+		break;
+	case CPU_PM_EXIT:
+		cpu_pm_pmu_setup(armpmu, cmd);
+	case CPU_PM_ENTER_FAILED:
+		armpmu->start(armpmu);
+		break;
+	default:
+		return NOTIFY_DONE;
+	}
+
+	return NOTIFY_OK;
+}
+
+static int cpu_pm_pmu_register(struct arm_pmu *cpu_pmu)
+{
+	cpu_pmu->cpu_pm_nb.notifier_call = cpu_pm_pmu_notify;
+	return cpu_pm_register_notifier(&cpu_pmu->cpu_pm_nb);
+}
+
+static void cpu_pm_pmu_unregister(struct arm_pmu *cpu_pmu)
+{
+	cpu_pm_unregister_notifier(&cpu_pmu->cpu_pm_nb);
+}
+#else
+static inline int cpu_pm_pmu_register(struct arm_pmu *cpu_pmu) { return 0; }
+static inline void cpu_pm_pmu_unregister(struct arm_pmu *cpu_pmu) { }
+#endif
+
 static int cpu_pmu_init(struct arm_pmu *cpu_pmu)
 {
 	int err;
@@ -725,6 +813,10 @@ static int cpu_pmu_init(struct arm_pmu *cpu_pmu)
 	if (err)
 		goto out_hw_events;
 
+	err = cpu_pm_pmu_register(cpu_pmu);
+	if (err)
+		goto out_unregister;
+
 	for_each_possible_cpu(cpu) {
 		struct pmu_hw_events *events = per_cpu_ptr(cpu_hw_events, cpu);
 		raw_spin_lock_init(&events->pmu_lock);
@@ -746,6 +838,8 @@ static int cpu_pmu_init(struct arm_pmu *cpu_pmu)
 
 	return 0;
 
+out_unregister:
+	unregister_cpu_notifier(&cpu_pmu->hotplug_nb);
 out_hw_events:
 	free_percpu(cpu_hw_events);
 	return err;
@@ -753,6 +847,7 @@ out_hw_events:
 
 static void cpu_pmu_destroy(struct arm_pmu *cpu_pmu)
 {
+	cpu_pm_pmu_unregister(cpu_pmu);
 	unregister_cpu_notifier(&cpu_pmu->hotplug_nb);
 	free_percpu(cpu_pmu->hw_events);
 }

include/linux/perf/arm_pmu.h

@@ -108,6 +108,7 @@ struct arm_pmu {
 	struct platform_device	*plat_device;
 	struct pmu_hw_events	__percpu *hw_events;
 	struct notifier_block	hotplug_nb;
+	struct notifier_block	cpu_pm_nb;
 };
 
 #define to_arm_pmu(p) (container_of(p, struct arm_pmu, pmu))