cpuidle/x86/perf: fix power:cpu_idle double end events and throw cpu_…

…idle events from the cpuidle layer

Currently intel_idle and acpi_idle driver show double cpu_idle "exit idle"
events -> this patch fixes it and makes cpu_idle events throwing less complex.

It also introduces cpu_idle events for all architectures which use
the cpuidle subsystem, namely:
  - arch/arm/mach-at91/cpuidle.c
  - arch/arm/mach-davinci/cpuidle.c
  - arch/arm/mach-kirkwood/cpuidle.c
  - arch/arm/mach-omap2/cpuidle34xx.c
  - arch/drivers/acpi/processor_idle.c (for all cases, not only mwait)
  - arch/x86/kernel/process.c (did throw events before, but was a mess)
  - drivers/idle/intel_idle.c (did throw events before)

Convention should be:
Fire cpu_idle events inside the current pm_idle function (not somewhere
down the the callee tree) to keep things easy.

Current possible pm_idle functions in X86:
c1e_idle, poll_idle, cpuidle_idle_call, mwait_idle, default_idle
-> this is really easy is now.

This affects userspace:
The type field of the cpu_idle power event can now direclty get
mapped to:
/sys/devices/system/cpu/cpuX/cpuidle/stateX/{name,desc,usage,time,...}
instead of throwing very CPU/mwait specific values.
This change is not visible for the intel_idle driver.
For the acpi_idle driver it should only be visible if the vendor
misses out C-states in his BIOS.
Another (perf timechart) patch reads out cpuidle info of cpu_idle
events from:
/sys/.../cpuidle/stateX/*, then the cpuidle events are mapped
to the correct C-/cpuidle state again, even if e.g. vendors miss
out C-states in their BIOS and for example only export C1 and C3.
-> everything is fine.

Signed-off-by: Thomas Renninger <trenn@suse.de>
CC: Robert Schoene <robert.schoene@tu-dresden.de>
CC: Jean Pihet <j-pihet@ti.com>
CC: Arjan van de Ven <arjan@linux.intel.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Frederic Weisbecker <fweisbec@gmail.com>
CC: linux-pm@lists.linux-foundation.org
CC: linux-acpi@vger.kernel.org
CC: linux-kernel@vger.kernel.org
CC: linux-perf-users@vger.kernel.org
CC: linux-omap@vger.kernel.org
Signed-off-by: Len Brown <len.brown@intel.com>

arch/x86/kernel/process.c

@@ -386,6 +386,8 @@ void default_idle(void)
 		else
 			local_irq_enable();
 		current_thread_info()->status |= TS_POLLING;
+		trace_power_end(smp_processor_id());
+		trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
 	} else {
 		local_irq_enable();
 		/* loop is done by the caller */
@@ -443,8 +445,6 @@ EXPORT_SYMBOL_GPL(cpu_idle_wait);
  */
 void mwait_idle_with_hints(unsigned long ax, unsigned long cx)
 {
-	trace_power_start(POWER_CSTATE, (ax>>4)+1, smp_processor_id());
-	trace_cpu_idle((ax>>4)+1, smp_processor_id());
 	if (!need_resched()) {
 		if (cpu_has(__this_cpu_ptr(&cpu_info), X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);
@@ -471,6 +471,8 @@ static void mwait_idle(void)
 			__sti_mwait(0, 0);
 		else
 			local_irq_enable();
+		trace_power_end(smp_processor_id());
+		trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
 	} else
 		local_irq_enable();
 }

arch/x86/kernel/process_32.c

@@ -57,8 +57,6 @@
 #include <asm/syscalls.h>
 #include <asm/debugreg.h>
 
-#include <trace/events/power.h>
-
 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
 
 /*
@@ -113,8 +111,6 @@ void cpu_idle(void)
 			stop_critical_timings();
 			pm_idle();
 			start_critical_timings();
-			trace_power_end(smp_processor_id());
-			trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
 		}
 		tick_nohz_restart_sched_tick();
 		preempt_enable_no_resched();

arch/x86/kernel/process_64.c

@@ -51,8 +51,6 @@
 #include <asm/syscalls.h>
 #include <asm/debugreg.h>
 
-#include <trace/events/power.h>
-
 asmlinkage extern void ret_from_fork(void);
 
 DEFINE_PER_CPU(unsigned long, old_rsp);
@@ -141,10 +139,6 @@ void cpu_idle(void)
 			pm_idle();
 			start_critical_timings();
 
-			trace_power_end(smp_processor_id());
-			trace_cpu_idle(PWR_EVENT_EXIT,
-				       smp_processor_id());
-
 			/* In many cases the interrupt that ended idle
 			   has already called exit_idle. But some idle
 			   loops can be woken up without interrupt. */

drivers/cpuidle/cpuidle.c

@@ -96,7 +96,15 @@ static void cpuidle_idle_call(void)
 
 	/* enter the state and update stats */
 	dev->last_state = target_state;
+
+	trace_power_start(POWER_CSTATE, next_state, dev->cpu);
+	trace_cpu_idle(next_state, dev->cpu);
+
 	dev->last_residency = target_state->enter(dev, target_state);
+
+	trace_power_end(dev->cpu);
+	trace_cpu_idle(PWR_EVENT_EXIT, dev->cpu);
+
 	if (dev->last_state)
 		target_state = dev->last_state;
 
@@ -106,8 +114,6 @@ static void cpuidle_idle_call(void)
 	/* give the governor an opportunity to reflect on the outcome */
 	if (cpuidle_curr_governor->reflect)
 		cpuidle_curr_governor->reflect(dev);
-	trace_power_end(smp_processor_id());
-	trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id());
 }
 
 /**

drivers/idle/intel_idle.c

@@ -220,8 +220,6 @@ static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state)
 	kt_before = ktime_get_real();
 
 	stop_critical_timings();
-	trace_power_start(POWER_CSTATE, (eax >> 4) + 1, cpu);
-	trace_cpu_idle((eax >> 4) + 1, cpu);
 	if (!need_resched()) {
 
 		__monitor((void *)&current_thread_info()->flags, 0, 0);