Commit f3f47a67 authored by Arjan van de Ven's avatar Arjan van de Ven Committed by Ingo Molnar
Browse files

tracing: add "power-tracer": C/P state tracer to help power optimization 



Impact: new "power-tracer" ftrace plugin

This patch adds a C/P-state ftrace plugin that will generate
detailed statistics about the C/P-states that are being used,
so that we can look at detailed decisions that the C/P-state
code is making, rather than the too high level "average"
that we have today.

An example way of using this is:

 mount -t debugfs none /sys/kernel/debug
 echo cstate > /sys/kernel/debug/tracing/current_tracer
 echo 1 > /sys/kernel/debug/tracing/tracing_enabled
 sleep 1
 echo 0 > /sys/kernel/debug/tracing/tracing_enabled
 cat /sys/kernel/debug/tracing/trace | perl scripts/trace/cstate.pl > out.svg

Signed-off-by: default avatarArjan van de Ven <arjan@linux.intel.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
parent 509dceef

arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c

+4 −0
Original line number Diff line number Diff line
@@ -33,6 +33,7 @@ 
#include <linux/cpufreq.h>

#include <linux/compiler.h>

#include <linux/dmi.h>

#include <linux/ftrace.h>



#include <linux/acpi.h>

#include <acpi/processor.h>
@@ -391,6 +392,7 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, 
	unsigned int next_perf_state = 0; /* Index into perf table */

	unsigned int i;

	int result = 0;

	struct power_trace it;



	dprintk("acpi_cpufreq_target %d (%d)\n", target_freq, policy->cpu);
@@ -427,6 +429,8 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, 
		}

	}



	trace_power_mark(&it, POWER_PSTATE, next_perf_state);



	switch (data->cpu_feature) {

	case SYSTEM_INTEL_MSR_CAPABLE:

		cmd.type = SYSTEM_INTEL_MSR_CAPABLE;

arch/x86/kernel/process.c

+16 −0
Original line number Diff line number Diff line
@@ -7,6 +7,7 @@ 
#include <linux/module.h>

#include <linux/pm.h>

#include <linux/clockchips.h>

#include <linux/ftrace.h>

#include <asm/system.h>



unsigned long idle_halt;
@@ -100,6 +101,9 @@ static inline int hlt_use_halt(void) 
void default_idle(void)

{

	if (hlt_use_halt()) {

		struct power_trace it;



		trace_power_start(&it, POWER_CSTATE, 1);

		current_thread_info()->status &= ~TS_POLLING;

		/*

		 * TS_POLLING-cleared state must be visible before we
@@ -112,6 +116,7 @@ void default_idle(void) 
		else

			local_irq_enable();

		current_thread_info()->status |= TS_POLLING;

		trace_power_end(&it);

	} else {

		local_irq_enable();

		/* loop is done by the caller */
@@ -154,24 +159,31 @@ EXPORT_SYMBOL_GPL(cpu_idle_wait); 
 */

void mwait_idle_with_hints(unsigned long ax, unsigned long cx)

{

	struct power_trace it;



	trace_power_start(&it, POWER_CSTATE, (ax>>4)+1);

	if (!need_resched()) {

		__monitor((void *)&current_thread_info()->flags, 0, 0);

		smp_mb();

		if (!need_resched())

			__mwait(ax, cx);

	}

	trace_power_end(&it);

}



/* Default MONITOR/MWAIT with no hints, used for default C1 state */

static void mwait_idle(void)

{

	struct power_trace it;

	if (!need_resched()) {

		trace_power_start(&it, POWER_CSTATE, 1);

		__monitor((void *)&current_thread_info()->flags, 0, 0);

		smp_mb();

		if (!need_resched())

			__sti_mwait(0, 0);

		else

			local_irq_enable();

		trace_power_end(&it);

	} else

		local_irq_enable();

}
@@ -183,9 +195,13 @@ static void mwait_idle(void) 
 */

static void poll_idle(void)

{

	struct power_trace it;



	trace_power_start(&it, POWER_CSTATE, 0);

	local_irq_enable();

	while (!need_resched())

		cpu_relax();

	trace_power_end(&it);

}



/*

include/linux/ftrace.h

+29 −0
Original line number Diff line number Diff line
@@ -311,6 +311,35 @@ ftrace_init_module(struct module *mod, 
		   unsigned long *start, unsigned long *end) { }

#endif



enum {

	POWER_NONE = 0,

	POWER_CSTATE = 1,

	POWER_PSTATE = 2,

};



struct power_trace {

#ifdef CONFIG_POWER_TRACER

	ktime_t			stamp;

	ktime_t			end;

	int			type;

	int			state;

#endif

};



#ifdef CONFIG_POWER_TRACER

extern void trace_power_start(struct power_trace *it, unsigned int type,

					unsigned int state);

extern void trace_power_mark(struct power_trace *it, unsigned int type,

					unsigned int state);

extern void trace_power_end(struct power_trace *it);

#else

static inline void trace_power_start(struct power_trace *it, unsigned int type,

					unsigned int state) { }

static inline void trace_power_mark(struct power_trace *it, unsigned int type,

					unsigned int state) { }

static inline void trace_power_end(struct power_trace *it) { }

#endif





/*

 * Structure that defines a return function trace.

kernel/trace/Kconfig

+11 −0
Original line number Diff line number Diff line
@@ -217,6 +217,17 @@ config BRANCH_TRACER 


	  Say N if unsure.



config POWER_TRACER

	bool "Trace power consumption behavior"

	depends on DEBUG_KERNEL

	depends on X86

	select TRACING

	help

	  This tracer helps developers to analyze and optimize the kernels

	  power management decisions, specifically the C-state and P-state

	  behavior.





config STACK_TRACER

	bool "Trace max stack"

	depends on HAVE_FUNCTION_TRACER

kernel/trace/Makefile

+1 −0
Original line number Diff line number Diff line
@@ -32,5 +32,6 @@ obj-$(CONFIG_BOOT_TRACER) += trace_boot.o 
obj-$(CONFIG_FUNCTION_RET_TRACER) += trace_functions_return.o

obj-$(CONFIG_TRACE_BRANCH_PROFILING) += trace_branch.o

obj-$(CONFIG_BTS_TRACER) += trace_bts.o

obj-$(CONFIG_POWER_TRACER) += trace_power.o



libftrace-y := ftrace.o

CRA Git | Maintained and supported by SUSTech CRA and CCSE