Re: Tracing of power:power_start events doesn't work

From: Ronny Tschüter
Date: Wed May 05 2010 - 10:11:49 EST


Hello,

I took a look at ftrace and tried to narrow my problem down.

1) <idle>-0 | | cpuidle_idle_call() {
1) <idle>-0 | | menu_select() {
1) <idle>-0 | 0.219 us | pm_qos_requirement();
1) <idle>-0 | 0.219 us | tick_nohz_get_sleep_length();
1) <idle>-0 | 0.210 us | nr_iowait_cpu();
1) <idle>-0 | 0.213 us | this_cpu_load();
1) <idle>-0 | 0.213 us | nr_iowait_cpu();
1) <idle>-0 | 2.472 us | }
1) <idle>-0 | | acpi_idle_enter_simple() {
1) <idle>-0 | | lapic_timer_state_broadcast() {
1) <idle>-0 | | clockevents_notify() {
1) <idle>-0 | 0.234 us | _raw_spin_lock_irqsave();
1) <idle>-0 | | raw_notifier_call_chain() {
1) <idle>-0 | | tick_notify() {
1) <idle>-0 | | tick_broadcast_oneshot_control() {
1) <idle>-0 | 0.237 us | _raw_spin_lock_irqsave();
1) <idle>-0 | | clockevents_set_mode() {
1) <idle>-0 | | lapic_timer_setup() {
1) <idle>-0 | 0.222 us | native_apic_mem_read();
1) <idle>-0 | 0.213 us | native_apic_mem_write();
1) <idle>-0 | 0.210 us | native_apic_mem_write();
1) <idle>-0 | 1.554 us | }
1) <idle>-0 | 2.007 us | }
1) <idle>-0 | | tick_dev_program_event() {
1) <idle>-0 | | ktime_get() {
1) <idle>-0 | 0.606 us | read_hpet();
1) <idle>-0 | 1.053 us | }
1) <idle>-0 | | clockevents_program_event() {
1) <idle>-0 | | hpet_legacy_next_event() {
1) <idle>-0 | 1.692 us | hpet_next_event();
1) <idle>-0 | 2.190 us | }
1) <idle>-0 | 2.664 us | }
1) <idle>-0 | 4.410 us | }
1) <idle>-0 | 0.237 us | _raw_spin_unlock_irqrestore();
1) <idle>-0 | 8.022 us | }
1) <idle>-0 | 8.460 us | }
1) <idle>-0 | 8.943 us | }
1) <idle>-0 | 0.231 us | _raw_spin_unlock_irqrestore();
1) <idle>-0 | + 10.296 us | }
1) <idle>-0 | + 10.734 us | }
1) <idle>-0 | | ktime_get_real() {
1) <idle>-0 | | getnstimeofday() {
1) <idle>-0 | 0.630 us | read_hpet();
1) <idle>-0 | 1.389 us | }
1) <idle>-0 | 1.830 us | }
1) <idle>-0 | ! 3833.301 us | acpi_idle_do_entry()

Like the callgraph shows, my machine uses the cpuidle framework with menu governor and an acpi-based driver to handle idle states. Unfortunately I can't find any instrumentation for tracing power events in the cpuidle.c and processor_idle.c (acpi idle management). The following patch achieved success.

diff --git a/old/processor_idle.c b/new/processor_idle.c
index cc978a8..1668d34 100644
--- a/old/processor_idle.c
+++ b/new/processor_idle.c
@@ -42,6 +42,7 @@
#include <linux/clockchips.h>
#include <linux/cpuidle.h>
#include <linux/irqflags.h>
+#include <trace/events/power.h>

/*
* Include the apic definitions for x86 to have the APIC timer related defines
@@ -796,6 +797,18 @@ static int acpi_idle_bm_check(void)
*/
static inline void acpi_idle_do_entry(struct acpi_processor_cx *cx)
{
+ switch (cx->type) {
+ case ACPI_STATE_C1:
+ trace_power_start(POWER_CSTATE, 1);
+ break;
+ case ACPI_STATE_C2:
+ trace_power_start(POWER_CSTATE, 2);
+ break;
+ case ACPI_STATE_C3:
+ trace_power_start(POWER_CSTATE, 3);
+ break;
+ }
+
/* Don't trace irqs off for idle */
stop_critical_timings();
if (cx->entry_method == ACPI_CSTATE_FFH) {

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/