Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 467f9957 authored by Linus Torvalds's avatar Linus Torvalds
Browse files

Merge branch 'perfcounters-core-for-linus' of... 

Merge branch 'perfcounters-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'perfcounters-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (58 commits)
  perf_counter: Fix perf_copy_attr() pointer arithmetic
  perf utils: Use a define for the maximum length of a trace event
  perf: Add timechart help text and add timechart to "perf help"
  tracing, x86, cpuidle: Move the end point of a C state in the power tracer
  perf utils: Be consistent about minimum text size in the svghelper
  perf timechart: Add "perf timechart record"
  perf: Add the timechart tool
  perf: Add a SVG helper library file
  tracing, perf: Convert the power tracer into an event tracer
  perf: Add a sample_event type to the event_union
  perf: Allow perf utilities to have "callback" options without arguments
  perf: Store trace event name/id pairs in perf.data
  perf: Add a timestamp to fork events
  sched_clock: Make it NMI safe
  perf_counter: Fix up swcounter throttling
  x86, perf_counter, bts: Optimize BTS overflow handling
  perf sched: Add --input=file option to builtin-sched.c
  perf trace: Sample timestamp and cpu when using record flag
  perf tools: Increase MAX_EVENT_LENGTH
  perf tools: Fix memory leak in read_ftrace_printk()
  ...
parents 78f28b7c cdf8073d

Documentation/trace/power.txt

deleted100644 → 0
+0 −17
Original line number Diff line number Diff line 
The power tracer collects detailed information about C-state and P-state

transitions, instead of just looking at the high-level "average"

information.



There is a helper script found in scrips/tracing/power.pl in the kernel

sources which can be used to parse this information and create a

Scalable Vector Graphics (SVG) picture from the trace data.



To use this tracer:



	echo 0 > /sys/kernel/debug/tracing/tracing_enabled

	echo power > /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/tracing/power.pl > out.sv

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

+2 −5
Original line number Diff line number Diff line
@@ -33,7 +33,7 @@ 
#include <linux/cpufreq.h>

#include <linux/compiler.h>

#include <linux/dmi.h>

#include <trace/power.h>

#include <trace/events/power.h>



#include <linux/acpi.h>

#include <linux/io.h>
@@ -72,8 +72,6 @@ static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data); 


static DEFINE_PER_CPU(struct aperfmperf, old_perf);



DEFINE_TRACE(power_mark);



/* acpi_perf_data is a pointer to percpu data. */

static struct acpi_processor_performance *acpi_perf_data;
@@ -332,7 +330,6 @@ 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);
@@ -364,7 +361,7 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, 
		}

	}



	trace_power_mark(&it, POWER_PSTATE, next_perf_state);

	trace_power_frequency(POWER_PSTATE, data->freq_table[next_state].frequency);



	switch (data->cpu_feature) {

	case SYSTEM_INTEL_MSR_CAPABLE:

arch/x86/kernel/cpu/perf_counter.c

+37 −23
Original line number Diff line number Diff line
@@ -36,10 +36,10 @@ static u64 perf_counter_mask __read_mostly; 
#define BTS_RECORD_SIZE		24



/* The size of a per-cpu BTS buffer in bytes: */

#define BTS_BUFFER_SIZE		(BTS_RECORD_SIZE * 1024)

#define BTS_BUFFER_SIZE		(BTS_RECORD_SIZE * 2048)



/* The BTS overflow threshold in bytes from the end of the buffer: */

#define BTS_OVFL_TH		(BTS_RECORD_SIZE * 64)

#define BTS_OVFL_TH		(BTS_RECORD_SIZE * 128)





/*
@@ -1488,8 +1488,7 @@ void perf_counter_print_debug(void) 
	local_irq_restore(flags);

}



static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc,

				       struct perf_sample_data *data)

static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc)

{

	struct debug_store *ds = cpuc->ds;

	struct bts_record {
@@ -1498,8 +1497,11 @@ static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc, 
		u64	flags;

	};

	struct perf_counter *counter = cpuc->counters[X86_PMC_IDX_FIXED_BTS];

	unsigned long orig_ip = data->regs->ip;

	struct bts_record *at, *top;

	struct perf_output_handle handle;

	struct perf_event_header header;

	struct perf_sample_data data;

	struct pt_regs regs;



	if (!counter)

		return;
@@ -1510,19 +1512,38 @@ static void intel_pmu_drain_bts_buffer(struct cpu_hw_counters *cpuc, 
	at  = (struct bts_record *)(unsigned long)ds->bts_buffer_base;

	top = (struct bts_record *)(unsigned long)ds->bts_index;



	if (top <= at)

		return;



	ds->bts_index = ds->bts_buffer_base;





	data.period	= counter->hw.last_period;

	data.addr	= 0;

	regs.ip		= 0;



	/*

	 * Prepare a generic sample, i.e. fill in the invariant fields.

	 * We will overwrite the from and to address before we output

	 * the sample.

	 */

	perf_prepare_sample(&header, &data, counter, &regs);



	if (perf_output_begin(&handle, counter,

			      header.size * (top - at), 1, 1))

		return;



	for (; at < top; at++) {

		data->regs->ip	= at->from;

		data->addr	= at->to;

		data.ip		= at->from;

		data.addr	= at->to;



		perf_counter_output(counter, 1, data);

		perf_output_sample(&handle, &header, &data, counter);

	}



	data->regs->ip	= orig_ip;

	data->addr	= 0;

	perf_output_end(&handle);



	/* There's new data available. */

	counter->hw.interrupts++;

	counter->pending_kill = POLL_IN;

}
@@ -1552,13 +1573,9 @@ static void x86_pmu_disable(struct perf_counter *counter) 
	x86_perf_counter_update(counter, hwc, idx);



	/* Drain the remaining BTS records. */

	if (unlikely(idx == X86_PMC_IDX_FIXED_BTS)) {

		struct perf_sample_data data;

		struct pt_regs regs;

	if (unlikely(idx == X86_PMC_IDX_FIXED_BTS))

		intel_pmu_drain_bts_buffer(cpuc);



		data.regs = &regs;

		intel_pmu_drain_bts_buffer(cpuc, &data);

	}

	cpuc->counters[idx] = NULL;

	clear_bit(idx, cpuc->used_mask);
@@ -1619,7 +1636,6 @@ static int p6_pmu_handle_irq(struct pt_regs *regs) 
	int idx, handled = 0;

	u64 val;



	data.regs = regs;

	data.addr = 0;



	cpuc = &__get_cpu_var(cpu_hw_counters);
@@ -1644,7 +1660,7 @@ static int p6_pmu_handle_irq(struct pt_regs *regs) 
		if (!x86_perf_counter_set_period(counter, hwc, idx))

			continue;



		if (perf_counter_overflow(counter, 1, &data))

		if (perf_counter_overflow(counter, 1, &data, regs))

			p6_pmu_disable_counter(hwc, idx);

	}
@@ -1665,13 +1681,12 @@ static int intel_pmu_handle_irq(struct pt_regs *regs) 
	int bit, loops;

	u64 ack, status;



	data.regs = regs;

	data.addr = 0;



	cpuc = &__get_cpu_var(cpu_hw_counters);



	perf_disable();

	intel_pmu_drain_bts_buffer(cpuc, &data);

	intel_pmu_drain_bts_buffer(cpuc);

	status = intel_pmu_get_status();

	if (!status) {

		perf_enable();
@@ -1702,7 +1717,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs) 


		data.period = counter->hw.last_period;



		if (perf_counter_overflow(counter, 1, &data))

		if (perf_counter_overflow(counter, 1, &data, regs))

			intel_pmu_disable_counter(&counter->hw, bit);

	}
@@ -1729,7 +1744,6 @@ static int amd_pmu_handle_irq(struct pt_regs *regs) 
	int idx, handled = 0;

	u64 val;



	data.regs = regs;

	data.addr = 0;



	cpuc = &__get_cpu_var(cpu_hw_counters);
@@ -1754,7 +1768,7 @@ static int amd_pmu_handle_irq(struct pt_regs *regs) 
		if (!x86_perf_counter_set_period(counter, hwc, idx))

			continue;



		if (perf_counter_overflow(counter, 1, &data))

		if (perf_counter_overflow(counter, 1, &data, regs))

			amd_pmu_disable_counter(hwc, idx);

	}

arch/x86/kernel/process.c

+6 −19
Original line number Diff line number Diff line
@@ -9,7 +9,7 @@ 
#include <linux/pm.h>

#include <linux/clockchips.h>

#include <linux/random.h>

#include <trace/power.h>

#include <trace/events/power.h>

#include <asm/system.h>

#include <asm/apic.h>

#include <asm/syscalls.h>
@@ -25,9 +25,6 @@ EXPORT_SYMBOL(idle_nomwait); 


struct kmem_cache *task_xstate_cachep;



DEFINE_TRACE(power_start);

DEFINE_TRACE(power_end);



int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)

{

	*dst = *src;
@@ -299,9 +296,7 @@ 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);

		trace_power_start(POWER_CSTATE, 1);

		current_thread_info()->status &= ~TS_POLLING;

		/*

		 * TS_POLLING-cleared state must be visible before we
@@ -314,7 +309,6 @@ 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 */
@@ -372,9 +366,7 @@ 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);

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

	if (!need_resched()) {

		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))

			clflush((void *)&current_thread_info()->flags);
@@ -384,15 +376,13 @@ void mwait_idle_with_hints(unsigned long ax, unsigned long cx) 
		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);

		trace_power_start(POWER_CSTATE, 1);

		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))

			clflush((void *)&current_thread_info()->flags);
@@ -402,7 +392,6 @@ static void mwait_idle(void) 
			__sti_mwait(0, 0);

		else

			local_irq_enable();

		trace_power_end(&it);

	} else

		local_irq_enable();

}
@@ -414,13 +403,11 @@ static void mwait_idle(void) 
 */

static void poll_idle(void)

{

	struct power_trace it;



	trace_power_start(&it, POWER_CSTATE, 0);

	trace_power_start(POWER_CSTATE, 0);

	local_irq_enable();

	while (!need_resched())

		cpu_relax();

	trace_power_end(&it);

	trace_power_end(0);

}



/*

drivers/cpuidle/cpuidle.c

+2 −0
Original line number Diff line number Diff line
@@ -17,6 +17,7 @@ 
#include <linux/cpuidle.h>

#include <linux/ktime.h>

#include <linux/hrtimer.h>

#include <trace/events/power.h>



#include "cpuidle.h"
@@ -91,6 +92,7 @@ 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(0);

}



/**