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()
  ...

Documentation/trace/power.txt

-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

@@ -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

@@ -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

@@ -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

@@ -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);
 }
 
 /**

include/linux/perf_counter.h

@@ -199,10 +199,14 @@ struct perf_counter_attr {
 				inherit_stat   :  1, /* per task counts       */
 				enable_on_exec :  1, /* next exec enables     */
 				task           :  1, /* trace fork/exit       */
+				watermark      :  1, /* wakeup_watermark      */
 
-				__reserved_1   : 50;
+				__reserved_1   : 49;
 
+	union {
 		__u32		wakeup_events;	  /* wakeup every n events */
+		__u32		wakeup_watermark; /* bytes before wakeup   */
+	};
 	__u32			__reserved_2;
 
 	__u64			__reserved_3;
@@ -332,6 +336,7 @@ enum perf_event_type {
 	 *	struct perf_event_header	header;
 	 *	u32				pid, ppid;
 	 *	u32				tid, ptid;
+	 *	u64				time;
 	 * };
 	 */
 	PERF_EVENT_EXIT			= 4,
@@ -352,6 +357,7 @@ enum perf_event_type {
 	 *	struct perf_event_header	header;
 	 *	u32				pid, ppid;
 	 *	u32				tid, ptid;
+	 *	{ u64				time;     } && PERF_SAMPLE_TIME
 	 * };
 	 */
 	PERF_EVENT_FORK			= 7,
@@ -521,6 +527,8 @@ struct perf_mmap_data {
 	atomic_t			wakeup;		/* needs a wakeup    */
 	atomic_t			lost;		/* nr records lost   */
 
+	long				watermark;	/* wakeup watermark  */
+
 	struct perf_counter_mmap_page   *user_page;
 	void				*data_pages[0];
 };
@@ -685,6 +693,17 @@ struct perf_cpu_context {
 	int				recursion[4];
 };
 
+struct perf_output_handle {
+	struct perf_counter	*counter;
+	struct perf_mmap_data	*data;
+	unsigned long		head;
+	unsigned long		offset;
+	int			nmi;
+	int			sample;
+	int			locked;
+	unsigned long		flags;
+};
+
 #ifdef CONFIG_PERF_COUNTERS
 
 /*
@@ -716,16 +735,38 @@ extern int hw_perf_group_sched_in(struct perf_counter *group_leader,
 extern void perf_counter_update_userpage(struct perf_counter *counter);
 
 struct perf_sample_data {
-	struct pt_regs			*regs;
+	u64				type;
+
+	u64				ip;
+	struct {
+		u32	pid;
+		u32	tid;
+	}				tid_entry;
+	u64				time;
 	u64				addr;
+	u64				id;
+	u64				stream_id;
+	struct {
+		u32	cpu;
+		u32	reserved;
+	}				cpu_entry;
 	u64				period;
+	struct perf_callchain_entry	*callchain;
 	struct perf_raw_record		*raw;
 };
 
+extern void perf_output_sample(struct perf_output_handle *handle,
+			       struct perf_event_header *header,
+			       struct perf_sample_data *data,
+			       struct perf_counter *counter);
+extern void perf_prepare_sample(struct perf_event_header *header,
+				struct perf_sample_data *data,
+				struct perf_counter *counter,
+				struct pt_regs *regs);
+
 extern int perf_counter_overflow(struct perf_counter *counter, int nmi,
-				 struct perf_sample_data *data);
-extern void perf_counter_output(struct perf_counter *counter, int nmi,
-				struct perf_sample_data *data);
+				 struct perf_sample_data *data,
+				 struct pt_regs *regs);
 
 /*
  * Return 1 for a software counter, 0 for a hardware counter
@@ -775,6 +816,12 @@ extern void perf_tpcounter_event(int event_id, u64 addr, u64 count,
 #define perf_instruction_pointer(regs)	instruction_pointer(regs)
 #endif
 
+extern int perf_output_begin(struct perf_output_handle *handle,
+			     struct perf_counter *counter, unsigned int size,
+			     int nmi, int sample);
+extern void perf_output_end(struct perf_output_handle *handle);
+extern void perf_output_copy(struct perf_output_handle *handle,
+			     const void *buf, unsigned int len);
 #else
 static inline void
 perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
@@ -801,7 +848,28 @@ static inline void perf_counter_mmap(struct vm_area_struct *vma)	{ }
 static inline void perf_counter_comm(struct task_struct *tsk)		{ }
 static inline void perf_counter_fork(struct task_struct *tsk)		{ }
 static inline void perf_counter_init(void)				{ }
+
+static inline int
+perf_output_begin(struct perf_output_handle *handle, struct perf_counter *c,
+		  unsigned int size, int nmi, int sample)		{ }
+static inline void perf_output_end(struct perf_output_handle *handle)	{ }
+static inline void
+perf_output_copy(struct perf_output_handle *handle,
+		 const void *buf, unsigned int len)			{ }
+static inline void
+perf_output_sample(struct perf_output_handle *handle,
+		   struct perf_event_header *header,
+		   struct perf_sample_data *data,
+		   struct perf_counter *counter)			{ }
+static inline void
+perf_prepare_sample(struct perf_event_header *header,
+		    struct perf_sample_data *data,
+		    struct perf_counter *counter,
+		    struct pt_regs *regs)				{ }
 #endif
 
+#define perf_output_put(handle, x) \
+	perf_output_copy((handle), &(x), sizeof(x))
+
 #endif /* __KERNEL__ */
 #endif /* _LINUX_PERF_COUNTER_H */

include/trace/events/power.h

+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM power
+
+#if !defined(_TRACE_POWER_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_POWER_H
+
+#include <linux/ktime.h>
+#include <linux/tracepoint.h>
+
+#ifndef _TRACE_POWER_ENUM_
+#define _TRACE_POWER_ENUM_
+enum {
+	POWER_NONE = 0,
+	POWER_CSTATE = 1,
+	POWER_PSTATE = 2,
+};
+#endif
+
+
+
+TRACE_EVENT(power_start,
+
+	TP_PROTO(unsigned int type, unsigned int state),
+
+	TP_ARGS(type, state),
+
+	TP_STRUCT__entry(
+		__field(	u64,		type		)
+		__field(	u64,		state		)
+	),
+
+	TP_fast_assign(
+		__entry->type = type;
+		__entry->state = state;
+	),
+
+	TP_printk("type=%lu state=%lu", (unsigned long)__entry->type, (unsigned long)__entry->state)
+);
+
+TRACE_EVENT(power_end,
+
+	TP_PROTO(int dummy),
+
+	TP_ARGS(dummy),
+
+	TP_STRUCT__entry(
+		__field(	u64,		dummy		)
+	),
+
+	TP_fast_assign(
+		__entry->dummy = 0xffff;
+	),
+
+	TP_printk("dummy=%lu", (unsigned long)__entry->dummy)
+
+);
+
+
+TRACE_EVENT(power_frequency,
+
+	TP_PROTO(unsigned int type, unsigned int state),
+
+	TP_ARGS(type, state),
+
+	TP_STRUCT__entry(
+		__field(	u64,		type		)
+		__field(	u64,		state		)
+	),
+
+	TP_fast_assign(
+		__entry->type = type;
+		__entry->state = state;
+	),
+
+	TP_printk("type=%lu state=%lu", (unsigned long)__entry->type, (unsigned long) __entry->state)
+);
+
+#endif /* _TRACE_POWER_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>

include/trace/events/sched.h

@@ -379,6 +379,39 @@ TRACE_EVENT(sched_stat_wait,
 			(unsigned long long)__entry->delay)
 );
 
+/*
+ * Tracepoint for accounting runtime (time the task is executing
+ * on a CPU).
+ */
+TRACE_EVENT(sched_stat_runtime,
+
+	TP_PROTO(struct task_struct *tsk, u64 runtime, u64 vruntime),
+
+	TP_ARGS(tsk, runtime, vruntime),
+
+	TP_STRUCT__entry(
+		__array( char,	comm,	TASK_COMM_LEN	)
+		__field( pid_t,	pid			)
+		__field( u64,	runtime			)
+		__field( u64,	vruntime			)
+	),
+
+	TP_fast_assign(
+		memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN);
+		__entry->pid		= tsk->pid;
+		__entry->runtime	= runtime;
+		__entry->vruntime	= vruntime;
+	)
+	TP_perf_assign(
+		__perf_count(runtime);
+	),
+
+	TP_printk("task: %s:%d runtime: %Lu [ns], vruntime: %Lu [ns]",
+			__entry->comm, __entry->pid,
+			(unsigned long long)__entry->runtime,
+			(unsigned long long)__entry->vruntime)
+);
+
 /*
  * Tracepoint for accounting sleep time (time the task is not runnable,
  * including iowait, see below).

kernel/sched_clock.c

@@ -48,13 +48,6 @@ static __read_mostly int sched_clock_running;
 __read_mostly int sched_clock_stable;
 
 struct sched_clock_data {
-	/*
-	 * Raw spinlock - this is a special case: this might be called
-	 * from within instrumentation code so we dont want to do any
-	 * instrumentation ourselves.
-	 */
-	raw_spinlock_t		lock;
-
 	u64			tick_raw;
 	u64			tick_gtod;
 	u64			clock;
@@ -80,7 +73,6 @@ void sched_clock_init(void)
 	for_each_possible_cpu(cpu) {
 		struct sched_clock_data *scd = cpu_sdc(cpu);
 
-		scd->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
 		scd->tick_raw = 0;
 		scd->tick_gtod = ktime_now;
 		scd->clock = ktime_now;
@@ -109,14 +101,19 @@ static inline u64 wrap_max(u64 x, u64 y)
  *  - filter out backward motion
  *  - use the GTOD tick value to create a window to filter crazy TSC values
  */
-static u64 __update_sched_clock(struct sched_clock_data *scd, u64 now)
+static u64 sched_clock_local(struct sched_clock_data *scd)
 {
-	s64 delta = now - scd->tick_raw;
-	u64 clock, min_clock, max_clock;
+	u64 now, clock, old_clock, min_clock, max_clock;
+	s64 delta;
 
+again:
+	now = sched_clock();
+	delta = now - scd->tick_raw;
 	if (unlikely(delta < 0))
 		delta = 0;
 
+	old_clock = scd->clock;
+
 	/*
 	 * scd->clock = clamp(scd->tick_gtod + delta,
 	 *		      max(scd->tick_gtod, scd->clock),
@@ -124,58 +121,27 @@ static u64 __update_sched_clock(struct sched_clock_data *scd, u64 now)
 	 */
 
 	clock = scd->tick_gtod + delta;
-	min_clock = wrap_max(scd->tick_gtod, scd->clock);
-	max_clock = wrap_max(scd->clock, scd->tick_gtod + TICK_NSEC);
+	min_clock = wrap_max(scd->tick_gtod, old_clock);
+	max_clock = wrap_max(old_clock, scd->tick_gtod + TICK_NSEC);
 
 	clock = wrap_max(clock, min_clock);
 	clock = wrap_min(clock, max_clock);
 
-	scd->clock = clock;
-
-	return scd->clock;
-}
+	if (cmpxchg(&scd->clock, old_clock, clock) != old_clock)
+		goto again;
 
-static void lock_double_clock(struct sched_clock_data *data1,
-				struct sched_clock_data *data2)
-{
-	if (data1 < data2) {
-		__raw_spin_lock(&data1->lock);
-		__raw_spin_lock(&data2->lock);
-	} else {
-		__raw_spin_lock(&data2->lock);
-		__raw_spin_lock(&data1->lock);
-	}
+	return clock;
 }
 
-u64 sched_clock_cpu(int cpu)
+static u64 sched_clock_remote(struct sched_clock_data *scd)
 {
-	u64 now, clock, this_clock, remote_clock;
-	struct sched_clock_data *scd;
-
-	if (sched_clock_stable)
-		return sched_clock();
-
-	scd = cpu_sdc(cpu);
-
-	/*
-	 * Normally this is not called in NMI context - but if it is,
-	 * trying to do any locking here is totally lethal.
-	 */
-	if (unlikely(in_nmi()))
-		return scd->clock;
-
-	if (unlikely(!sched_clock_running))
-		return 0ull;
-
-	WARN_ON_ONCE(!irqs_disabled());
-	now = sched_clock();
-
-	if (cpu != raw_smp_processor_id()) {
 	struct sched_clock_data *my_scd = this_scd();
+	u64 this_clock, remote_clock;
+	u64 *ptr, old_val, val;
 
-		lock_double_clock(scd, my_scd);
-
-		this_clock = __update_sched_clock(my_scd, now);
+	sched_clock_local(my_scd);
+again:
+	this_clock = my_scd->clock;
 	remote_clock = scd->clock;
 
 	/*
@@ -185,23 +151,43 @@ u64 sched_clock_cpu(int cpu)
 	 * runqueues. (this creates monotonic movement)
 	 */
 	if (likely((s64)(remote_clock - this_clock) < 0)) {
-			clock = this_clock;
-			scd->clock = clock;
+		ptr = &scd->clock;
+		old_val = remote_clock;
+		val = this_clock;
 	} else {
 		/*
 		 * Should be rare, but possible:
 		 */
-			clock = remote_clock;
-			my_scd->clock = remote_clock;
+		ptr = &my_scd->clock;
+		old_val = this_clock;
+		val = remote_clock;
 	}
 
-		__raw_spin_unlock(&my_scd->lock);
-	} else {
-		__raw_spin_lock(&scd->lock);
-		clock = __update_sched_clock(scd, now);
-	}
+	if (cmpxchg(ptr, old_val, val) != old_val)
+		goto again;
+
+	return val;
+}
+
+u64 sched_clock_cpu(int cpu)
+{
+	struct sched_clock_data *scd;
+	u64 clock;
+
+	WARN_ON_ONCE(!irqs_disabled());
+
+	if (sched_clock_stable)
+		return sched_clock();
+
+	if (unlikely(!sched_clock_running))
+		return 0ull;
+
+	scd = cpu_sdc(cpu);
 
-	__raw_spin_unlock(&scd->lock);
+	if (cpu != smp_processor_id())
+		clock = sched_clock_remote(scd);
+	else
+		clock = sched_clock_local(scd);
 
 	return clock;
 }
@@ -223,11 +209,9 @@ void sched_clock_tick(void)
 	now_gtod = ktime_to_ns(ktime_get());
 	now = sched_clock();
 
-	__raw_spin_lock(&scd->lock);
 	scd->tick_raw = now;
 	scd->tick_gtod = now_gtod;
-	__update_sched_clock(scd, now);
-	__raw_spin_unlock(&scd->lock);
+	sched_clock_local(scd);
 }
 
 /*

kernel/sched_fair.c

@@ -513,6 +513,7 @@ static void update_curr(struct cfs_rq *cfs_rq)
 	if (entity_is_task(curr)) {
 		struct task_struct *curtask = task_of(curr);
 
+		trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
 		cpuacct_charge(curtask, delta_exec);
 		account_group_exec_runtime(curtask, delta_exec);
 	}

kernel/trace/Makefile

@@ -42,7 +42,6 @@ obj-$(CONFIG_BOOT_TRACER) += trace_boot.o
 obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += trace_functions_graph.o
 obj-$(CONFIG_TRACE_BRANCH_PROFILING) += trace_branch.o
 obj-$(CONFIG_HW_BRANCH_TRACER) += trace_hw_branches.o
-obj-$(CONFIG_POWER_TRACER) += trace_power.o
 obj-$(CONFIG_KMEMTRACE) += kmemtrace.o
 obj-$(CONFIG_WORKQUEUE_TRACER) += trace_workqueue.o
 obj-$(CONFIG_BLK_DEV_IO_TRACE) += blktrace.o
@@ -54,5 +53,6 @@ obj-$(CONFIG_EVENT_TRACING) += trace_export.o
 obj-$(CONFIG_FTRACE_SYSCALLS) += trace_syscalls.o
 obj-$(CONFIG_EVENT_PROFILE) += trace_event_profile.o
 obj-$(CONFIG_EVENT_TRACING) += trace_events_filter.o
+obj-$(CONFIG_EVENT_TRACING) += power-traces.o
 
 libftrace-y := ftrace.o

kernel/trace/power-traces.c

+/*
+ * Power trace points
+ *
+ * Copyright (C) 2009 Arjan van de Ven <arjan@linux.intel.com>
+ */
+
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/power.h>
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(power_start);
+EXPORT_TRACEPOINT_SYMBOL_GPL(power_end);
+EXPORT_TRACEPOINT_SYMBOL_GPL(power_frequency);
+

kernel/trace/trace.h

@@ -11,7 +11,6 @@
 #include <linux/ftrace.h>
 #include <trace/boot.h>
 #include <linux/kmemtrace.h>
-#include <trace/power.h>
 
 #include <linux/trace_seq.h>
 #include <linux/ftrace_event.h>
@@ -37,7 +36,6 @@ enum trace_type {
 	TRACE_HW_BRANCHES,
 	TRACE_KMEM_ALLOC,
 	TRACE_KMEM_FREE,
-	TRACE_POWER,
 	TRACE_BLK,
 
 	__TRACE_LAST_TYPE,
@@ -207,7 +205,6 @@ extern void __ftrace_bad_type(void);
 		IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry,	\
 			  TRACE_GRAPH_RET);		\
 		IF_ASSIGN(var, ent, struct hw_branch_entry, TRACE_HW_BRANCHES);\
-		IF_ASSIGN(var, ent, struct trace_power, TRACE_POWER); \
 		IF_ASSIGN(var, ent, struct kmemtrace_alloc_entry,	\
 			  TRACE_KMEM_ALLOC);	\
 		IF_ASSIGN(var, ent, struct kmemtrace_free_entry,	\

kernel/trace/trace_entries.h

@@ -330,23 +330,6 @@ FTRACE_ENTRY(hw_branch, hw_branch_entry,
 	F_printk("from: %llx to: %llx", __entry->from, __entry->to)
 );
 
-FTRACE_ENTRY(power, trace_power,
-
-	TRACE_POWER,
-
-	F_STRUCT(
-		__field_struct(	struct power_trace,	state_data	)
-		__field_desc(	s64,	state_data,	stamp		)
-		__field_desc(	s64,	state_data,	end		)
-		__field_desc(	int,	state_data,	type		)
-		__field_desc(	int,	state_data,	state		)
-	),
-
-	F_printk("%llx->%llx type:%u state:%u",
-		 __entry->stamp, __entry->end,
-		 __entry->type, __entry->state)
-);
-
 FTRACE_ENTRY(kmem_alloc, kmemtrace_alloc_entry,
 
 	TRACE_KMEM_ALLOC,

kernel/trace/trace_power.c

-/*
- * ring buffer based C-state tracer
- *
- * Arjan van de Ven <arjan@linux.intel.com>
- * Copyright (C) 2008 Intel Corporation
- *
- * Much is borrowed from trace_boot.c which is
- * Copyright (C) 2008 Frederic Weisbecker <fweisbec@gmail.com>
- *
- */
-
-#include <linux/init.h>
-#include <linux/debugfs.h>
-#include <trace/power.h>
-#include <linux/kallsyms.h>
-#include <linux/module.h>
-
-#include "trace.h"
-#include "trace_output.h"
-
-static struct trace_array *power_trace;
-static int __read_mostly trace_power_enabled;
-
-static void probe_power_start(struct power_trace *it, unsigned int type,
-				unsigned int level)
-{
-	if (!trace_power_enabled)
-		return;
-
-	memset(it, 0, sizeof(struct power_trace));
-	it->state = level;
-	it->type = type;
-	it->stamp = ktime_get();
-}
-
-
-static void probe_power_end(struct power_trace *it)
-{
-	struct ftrace_event_call *call = &event_power;
-	struct ring_buffer_event *event;
-	struct ring_buffer *buffer;
-	struct trace_power *entry;
-	struct trace_array_cpu *data;
-	struct trace_array *tr = power_trace;
-
-	if (!trace_power_enabled)
-		return;
-
-	buffer = tr->buffer;
-
-	preempt_disable();
-	it->end = ktime_get();
-	data = tr->data[smp_processor_id()];
-
-	event = trace_buffer_lock_reserve(buffer, TRACE_POWER,
-					  sizeof(*entry), 0, 0);
-	if (!event)
-		goto out;
-	entry	= ring_buffer_event_data(event);
-	entry->state_data = *it;
-	if (!filter_check_discard(call, entry, buffer, event))
-		trace_buffer_unlock_commit(buffer, event, 0, 0);
- out:
-	preempt_enable();
-}
-
-static void probe_power_mark(struct power_trace *it, unsigned int type,
-				unsigned int level)
-{
-	struct ftrace_event_call *call = &event_power;
-	struct ring_buffer_event *event;
-	struct ring_buffer *buffer;
-	struct trace_power *entry;
-	struct trace_array_cpu *data;
-	struct trace_array *tr = power_trace;
-
-	if (!trace_power_enabled)
-		return;
-
-	buffer = tr->buffer;
-
-	memset(it, 0, sizeof(struct power_trace));
-	it->state = level;
-	it->type = type;
-	it->stamp = ktime_get();
-	preempt_disable();
-	it->end = it->stamp;
-	data = tr->data[smp_processor_id()];
-
-	event = trace_buffer_lock_reserve(buffer, TRACE_POWER,
-					  sizeof(*entry), 0, 0);
-	if (!event)
-		goto out;
-	entry	= ring_buffer_event_data(event);
-	entry->state_data = *it;
-	if (!filter_check_discard(call, entry, buffer, event))
-		trace_buffer_unlock_commit(buffer, event, 0, 0);
- out:
-	preempt_enable();
-}
-
-static int tracing_power_register(void)
-{
-	int ret;
-
-	ret = register_trace_power_start(probe_power_start);
-	if (ret) {
-		pr_info("power trace: Couldn't activate tracepoint"
-			" probe to trace_power_start\n");
-		return ret;
-	}
-	ret = register_trace_power_end(probe_power_end);
-	if (ret) {
-		pr_info("power trace: Couldn't activate tracepoint"
-			" probe to trace_power_end\n");
-		goto fail_start;
-	}
-	ret = register_trace_power_mark(probe_power_mark);
-	if (ret) {
-		pr_info("power trace: Couldn't activate tracepoint"
-			" probe to trace_power_mark\n");
-		goto fail_end;
-	}
-	return ret;
-fail_end:
-	unregister_trace_power_end(probe_power_end);
-fail_start:
-	unregister_trace_power_start(probe_power_start);
-	return ret;
-}
-
-static void start_power_trace(struct trace_array *tr)
-{
-	trace_power_enabled = 1;
-}
-
-static void stop_power_trace(struct trace_array *tr)
-{
-	trace_power_enabled = 0;
-}
-
-static void power_trace_reset(struct trace_array *tr)
-{
-	trace_power_enabled = 0;
-	unregister_trace_power_start(probe_power_start);
-	unregister_trace_power_end(probe_power_end);
-	unregister_trace_power_mark(probe_power_mark);
-}
-
-
-static int power_trace_init(struct trace_array *tr)
-{
-	power_trace = tr;
-
-	trace_power_enabled = 1;
-	tracing_power_register();
-
-	tracing_reset_online_cpus(tr);
-	return 0;
-}
-
-static enum print_line_t power_print_line(struct trace_iterator *iter)
-{
-	int ret = 0;
-	struct trace_entry *entry = iter->ent;
-	struct trace_power *field ;
-	struct power_trace *it;
-	struct trace_seq *s = &iter->seq;
-	struct timespec stamp;
-	struct timespec duration;
-
-	trace_assign_type(field, entry);
-	it = &field->state_data;
-	stamp = ktime_to_timespec(it->stamp);
-	duration = ktime_to_timespec(ktime_sub(it->end, it->stamp));
-
-	if (entry->type == TRACE_POWER) {
-		if (it->type == POWER_CSTATE)
-			ret = trace_seq_printf(s, "[%5ld.%09ld] CSTATE: Going to C%i on cpu %i for %ld.%09ld\n",
-					  stamp.tv_sec,
-					  stamp.tv_nsec,
-					  it->state, iter->cpu,
-					  duration.tv_sec,
-					  duration.tv_nsec);
-		if (it->type == POWER_PSTATE)
-			ret = trace_seq_printf(s, "[%5ld.%09ld] PSTATE: Going to P%i on cpu %i\n",
-					  stamp.tv_sec,
-					  stamp.tv_nsec,
-					  it->state, iter->cpu);
-		if (!ret)
-			return TRACE_TYPE_PARTIAL_LINE;
-		return TRACE_TYPE_HANDLED;
-	}
-	return TRACE_TYPE_UNHANDLED;
-}
-
-static void power_print_header(struct seq_file *s)
-{
-	seq_puts(s, "#   TIMESTAMP      STATE  EVENT\n");
-	seq_puts(s, "#       |            |      |\n");
-}
-
-static struct tracer power_tracer __read_mostly =
-{
-	.name		= "power",
-	.init		= power_trace_init,
-	.start		= start_power_trace,
-	.stop		= stop_power_trace,
-	.reset		= power_trace_reset,
-	.print_line	= power_print_line,
-	.print_header	= power_print_header,
-};
-
-static int init_power_trace(void)
-{
-	return register_tracer(&power_tracer);
-}
-device_initcall(init_power_trace);

scripts/tracing/power.pl

-#!/usr/bin/perl
-
-# Copyright 2008, Intel Corporation
-#
-# This file is part of the Linux kernel
-#
-# This program file is free software; you can redistribute it and/or modify it
-# under the terms of the GNU General Public License as published by the
-# Free Software Foundation; version 2 of the License.
-#
-# This program is distributed in the hope that it will be useful, but WITHOUT
-# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
-# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
-# for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with this program in a file named COPYING; if not, write to the
-# Free Software Foundation, Inc.,
-# 51 Franklin Street, Fifth Floor,
-# Boston, MA 02110-1301 USA
-#
-# Authors:
-# 	Arjan van de Ven <arjan@linux.intel.com>
-
-
-#
-# This script turns a cstate ftrace output into a SVG graphic that shows
-# historic C-state information
-#
-#
-# 	cat /sys/kernel/debug/tracing/trace | perl power.pl > out.svg
-#
-
-my @styles;
-my $base = 0;
-
-my @pstate_last;
-my @pstate_level;
-
-$styles[0] = "fill:rgb(0,0,255);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[1] = "fill:rgb(0,255,0);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[2] = "fill:rgb(255,0,20);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[3] = "fill:rgb(255,255,20);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[4] = "fill:rgb(255,0,255);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[5] = "fill:rgb(0,255,255);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[6] = "fill:rgb(0,128,255);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[7] = "fill:rgb(0,255,128);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-$styles[8] = "fill:rgb(0,25,20);fill-opacity:0.5;stroke-width:1;stroke:rgb(0,0,0)";
-
-
-print "<?xml version=\"1.0\" standalone=\"no\"?> \n";
-print "<svg width=\"10000\" height=\"100%\" version=\"1.1\" xmlns=\"http://www.w3.org/2000/svg\">\n";
-
-my $scale = 30000.0;
-while (<>) {
-	my $line = $_;
-	if ($line =~ /([0-9\.]+)\] CSTATE: Going to C([0-9]) on cpu ([0-9]+) for ([0-9\.]+)/) {
-		if ($base == 0) {
-			$base = $1;
-		}
-		my $time = $1 - $base;
-		$time = $time * $scale;
-		my $C = $2;
-		my $cpu = $3;
-		my $y = 400 * $cpu;
-		my $duration = $4 * $scale;
-		my $msec = int($4 * 100000)/100.0;
-		my $height = $C * 20;
-		$style = $styles[$C];
-
-		$y = $y + 140 - $height;
-
-		$x2 = $time + 4;
-		$y2 = $y + 4;
-
-
-		print "<rect x=\"$time\" width=\"$duration\" y=\"$y\" height=\"$height\" style=\"$style\"/>\n";
-		print "<text transform=\"translate($x2,$y2) rotate(90)\">C$C $msec</text>\n";
-	}
-	if ($line =~ /([0-9\.]+)\] PSTATE: Going to P([0-9]) on cpu ([0-9]+)/) {
-		my $time = $1 - $base;
-		my $state = $2;
-		my $cpu = $3;
-
-		if (defined($pstate_last[$cpu])) {
-			my $from = $pstate_last[$cpu];
-			my $oldstate = $pstate_state[$cpu];
-			my $duration = ($time-$from) * $scale;
-
-			$from = $from * $scale;
-			my $to = $from + $duration;
-			my $height = 140 - ($oldstate * (140/8));
-
-			my $y = 400 * $cpu + 200 + $height;
-			my $y2 = $y+4;
-			my $style = $styles[8];
-
-			print "<rect x=\"$from\" y=\"$y\" width=\"$duration\" height=\"5\" style=\"$style\"/>\n";
-			print "<text transform=\"translate($from,$y2)\">P$oldstate (cpu $cpu)</text>\n";
-		};
-
-		$pstate_last[$cpu] = $time;
-		$pstate_state[$cpu] = $state;
-	}
-}
-
-
-print "</svg>\n";

tools/perf/Documentation/perf-sched.txt

+perf-sched(1)
+==============
+
+NAME
+----
+perf-sched - Tool to trace/measure scheduler properties (latencies)
+
+SYNOPSIS
+--------
+[verse]
+'perf sched' {record|latency|replay|trace}
+
+DESCRIPTION
+-----------
+There's four variants of perf sched:
+
+  'perf sched record <command>' to record the scheduling events
+  of an arbitrary workload.
+
+  'perf sched latency' to report the per task scheduling latencies
+  and other scheduling properties of the workload.
+
+  'perf sched trace' to see a detailed trace of the workload that
+  was recorded.
+
+  'perf sched replay' to simulate the workload that was recorded
+  via perf sched record. (this is done by starting up mockup threads
+  that mimic the workload based on the events in the trace. These
+  threads can then replay the timings (CPU runtime and sleep patterns)
+  of the workload as it occured when it was recorded - and can repeat
+  it a number of times, measuring its performance.)
+
+OPTIONS
+-------
+-D::
+--dump-raw-trace=::
+        Display verbose dump of the sched data.
+
+SEE ALSO
+--------
+linkperf:perf-record[1]

tools/perf/Documentation/perf-timechart.txt

+perf-timechart(1)
+=================
+
+NAME
+----
+perf-timechart - Tool to visualize total system behavior during a workload
+
+SYNOPSIS
+--------
+[verse]
+'perf timechart' {record}
+
+DESCRIPTION
+-----------
+There are two variants of perf timechart:
+
+  'perf timechart record <command>' to record the system level events
+  of an arbitrary workload.
+
+  'perf timechart' to turn a trace into a Scalable Vector Graphics file,
+  that can be viewed with popular SVG viewers such as 'Inkscape'.
+
+OPTIONS
+-------
+-o::
+--output=::
+        Select the output file (default: output.svg)
+-i::
+--input=::
+        Select the input file (default: perf.data)
+
+
+SEE ALSO
+--------
+linkperf:perf-record[1]

tools/perf/Documentation/perf-trace.txt

+perf-trace(1)
+==============
+
+NAME
+----
+perf-trace - Read perf.data (created by perf record) and display trace output
+
+SYNOPSIS
+--------
+[verse]
+'perf trace' [-i <file> | --input=file] symbol_name
+
+DESCRIPTION
+-----------
+This command reads the input file and displays the trace recorded.
+
+OPTIONS
+-------
+-D::
+--dump-raw-trace=::
+        Display verbose dump of the trace data.
+
+SEE ALSO
+--------
+linkperf:perf-record[1]

tools/perf/Makefile

@@ -373,13 +373,16 @@ LIB_OBJS += util/thread.o
 LIB_OBJS += util/trace-event-parse.o
 LIB_OBJS += util/trace-event-read.o
 LIB_OBJS += util/trace-event-info.o
+LIB_OBJS += util/svghelper.o
 
 BUILTIN_OBJS += builtin-annotate.o
 BUILTIN_OBJS += builtin-help.o
+BUILTIN_OBJS += builtin-sched.o
 BUILTIN_OBJS += builtin-list.o
 BUILTIN_OBJS += builtin-record.o
 BUILTIN_OBJS += builtin-report.o
 BUILTIN_OBJS += builtin-stat.o
+BUILTIN_OBJS += builtin-timechart.o
 BUILTIN_OBJS += builtin-top.o
 BUILTIN_OBJS += builtin-trace.o
 
@@ -710,6 +713,12 @@ builtin-help.o: builtin-help.c common-cmds.h PERF-CFLAGS
 		'-DPERF_MAN_PATH="$(mandir_SQ)"' \
 		'-DPERF_INFO_PATH="$(infodir_SQ)"' $<
 
+builtin-timechart.o: builtin-timechart.c common-cmds.h PERF-CFLAGS
+	$(QUIET_CC)$(CC) -o $*.o -c $(ALL_CFLAGS) \
+		'-DPERF_HTML_PATH="$(htmldir_SQ)"' \
+		'-DPERF_MAN_PATH="$(mandir_SQ)"' \
+		'-DPERF_INFO_PATH="$(infodir_SQ)"' $<
+
 $(BUILT_INS): perf$X
 	$(QUIET_BUILT_IN)$(RM) $@ && \
 	ln perf$X $@ 2>/dev/null || \

tools/perf/builtin-record.c

@@ -48,6 +48,8 @@ static int			call_graph			= 0;
 static int			inherit_stat			= 0;
 static int			no_samples			= 0;
 static int			sample_address			= 0;
+static int			multiplex			= 0;
+static int			multiplex_fd			= -1;
 
 static long			samples;
 static struct timeval		last_read;
@@ -470,7 +472,15 @@ static void create_counter(int counter, int cpu, pid_t pid)
 	 */
 	if (group && group_fd == -1)
 		group_fd = fd[nr_cpu][counter];
+	if (multiplex && multiplex_fd == -1)
+		multiplex_fd = fd[nr_cpu][counter];
 
+	if (multiplex && fd[nr_cpu][counter] != multiplex_fd) {
+		int ret;
+
+		ret = ioctl(fd[nr_cpu][counter], PERF_COUNTER_IOC_SET_OUTPUT, multiplex_fd);
+		assert(ret != -1);
+	} else {
 		event_array[nr_poll].fd = fd[nr_cpu][counter];
 		event_array[nr_poll].events = POLLIN;
 		nr_poll++;
@@ -484,6 +494,7 @@ static void create_counter(int counter, int cpu, pid_t pid)
 			error("failed to mmap with %d (%s)\n", errno, strerror(errno));
 			exit(-1);
 		}
+	}
 
 	ioctl(fd[nr_cpu][counter], PERF_COUNTER_IOC_ENABLE);
 }
@@ -513,6 +524,7 @@ static int __cmd_record(int argc, const char **argv)
 	pid_t pid = 0;
 	int flags;
 	int ret;
+	unsigned long waking = 0;
 
 	page_size = sysconf(_SC_PAGE_SIZE);
 	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
@@ -614,16 +626,28 @@ static int __cmd_record(int argc, const char **argv)
 		int hits = samples;
 
 		for (i = 0; i < nr_cpu; i++) {
-			for (counter = 0; counter < nr_counters; counter++)
+			for (counter = 0; counter < nr_counters; counter++) {
+				if (mmap_array[i][counter].base)
 					mmap_read(&mmap_array[i][counter]);
 			}
+		}
 
 		if (hits == samples) {
 			if (done)
 				break;
-			ret = poll(event_array, nr_poll, 100);
+			ret = poll(event_array, nr_poll, -1);
+			waking++;
 		}
+
+		if (done) {
+			for (i = 0; i < nr_cpu; i++) {
+				for (counter = 0; counter < nr_counters; counter++)
+					ioctl(fd[i][counter], PERF_COUNTER_IOC_DISABLE);
 			}
+		}
+	}
+
+	fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking);
 
 	/*
 	 * Approximate RIP event size: 24 bytes.
@@ -681,6 +705,8 @@ static const struct option options[] = {
 		    "Sample addresses"),
 	OPT_BOOLEAN('n', "no-samples", &no_samples,
 		    "don't sample"),
+	OPT_BOOLEAN('M', "multiplex", &multiplex,
+		    "multiplex counter output in a single channel"),
 	OPT_END()
 };
 

tools/perf/builtin.h

@@ -16,12 +16,14 @@ extern int check_pager_config(const char *cmd);
 
 extern int cmd_annotate(int argc, const char **argv, const char *prefix);
 extern int cmd_help(int argc, const char **argv, const char *prefix);
+extern int cmd_sched(int argc, const char **argv, const char *prefix);
+extern int cmd_list(int argc, const char **argv, const char *prefix);
 extern int cmd_record(int argc, const char **argv, const char *prefix);
 extern int cmd_report(int argc, const char **argv, const char *prefix);
 extern int cmd_stat(int argc, const char **argv, const char *prefix);
+extern int cmd_timechart(int argc, const char **argv, const char *prefix);
 extern int cmd_top(int argc, const char **argv, const char *prefix);
-extern int cmd_version(int argc, const char **argv, const char *prefix);
-extern int cmd_list(int argc, const char **argv, const char *prefix);
 extern int cmd_trace(int argc, const char **argv, const char *prefix);
+extern int cmd_version(int argc, const char **argv, const char *prefix);
 
 #endif

tools/perf/command-list.txt

@@ -4,7 +4,10 @@
 #
 perf-annotate			mainporcelain common
 perf-list			mainporcelain common
+perf-sched			mainporcelain common
 perf-record			mainporcelain common
 perf-report			mainporcelain common
 perf-stat			mainporcelain common
+perf-timechart			mainporcelain common
 perf-top			mainporcelain common
+perf-trace			mainporcelain common

tools/perf/perf.c

@@ -289,10 +289,12 @@ static void handle_internal_command(int argc, const char **argv)
 		{ "record", cmd_record, 0 },
 		{ "report", cmd_report, 0 },
 		{ "stat", cmd_stat, 0 },
+		{ "timechart", cmd_timechart, 0 },
 		{ "top", cmd_top, 0 },
 		{ "annotate", cmd_annotate, 0 },
 		{ "version", cmd_version, 0 },
 		{ "trace", cmd_trace, 0 },
+		{ "sched", cmd_sched, 0 },
 	};
 	unsigned int i;
 	static const char ext[] = STRIP_EXTENSION;

tools/perf/util/event.h

@@ -39,6 +39,7 @@ struct fork_event {
 	struct perf_event_header header;
 	u32 pid, ppid;
 	u32 tid, ptid;
+	u64 time;
 };
 
 struct lost_event {
@@ -52,13 +53,19 @@ struct lost_event {
  */
 struct read_event {
 	struct perf_event_header header;
-	u32 pid,tid;
+	u32 pid, tid;
 	u64 value;
 	u64 time_enabled;
 	u64 time_running;
 	u64 id;
 };
 
+struct sample_event{
+	struct perf_event_header        header;
+	u64 array[];
+};
+
+
 typedef union event_union {
 	struct perf_event_header	header;
 	struct ip_event			ip;
@@ -67,6 +74,7 @@ typedef union event_union {
 	struct fork_event		fork;
 	struct lost_event		lost;
 	struct read_event		read;
+	struct sample_event		sample;
 } event_t;
 
 struct map {

tools/perf/util/header.c

@@ -7,9 +7,8 @@
 #include "header.h"
 
 /*
- *
+ * Create new perf.data header attribute:
  */
-
 struct perf_header_attr *perf_header_attr__new(struct perf_counter_attr *attr)
 {
 	struct perf_header_attr *self = malloc(sizeof(*self));
@@ -43,9 +42,8 @@ void perf_header_attr__add_id(struct perf_header_attr *self, u64 id)
 }
 
 /*
- *
+ * Create new perf.data header:
  */
-
 struct perf_header *perf_header__new(void)
 {
 	struct perf_header *self = malloc(sizeof(*self));
@@ -86,6 +84,46 @@ void perf_header__add_attr(struct perf_header *self,
 	self->attr[pos] = attr;
 }
 
+#define MAX_EVENT_NAME 64
+
+struct perf_trace_event_type {
+	u64	event_id;
+	char	name[MAX_EVENT_NAME];
+};
+
+static int event_count;
+static struct perf_trace_event_type *events;
+
+void perf_header__push_event(u64 id, const char *name)
+{
+	if (strlen(name) > MAX_EVENT_NAME)
+		printf("Event %s will be truncated\n", name);
+
+	if (!events) {
+		events = malloc(sizeof(struct perf_trace_event_type));
+		if (!events)
+			die("nomem");
+	} else {
+		events = realloc(events, (event_count + 1) * sizeof(struct perf_trace_event_type));
+		if (!events)
+			die("nomem");
+	}
+	memset(&events[event_count], 0, sizeof(struct perf_trace_event_type));
+	events[event_count].event_id = id;
+	strncpy(events[event_count].name, name, MAX_EVENT_NAME - 1);
+	event_count++;
+}
+
+char *perf_header__find_event(u64 id)
+{
+	int i;
+	for (i = 0 ; i < event_count; i++) {
+		if (events[i].event_id == id)
+			return events[i].name;
+	}
+	return NULL;
+}
+
 static const char *__perf_magic = "PERFFILE";
 
 #define PERF_MAGIC	(*(u64 *)__perf_magic)
@@ -106,6 +144,7 @@ struct perf_file_header {
 	u64				attr_size;
 	struct perf_file_section	attrs;
 	struct perf_file_section	data;
+	struct perf_file_section	event_types;
 };
 
 static void do_write(int fd, void *buf, size_t size)
@@ -154,6 +193,11 @@ void perf_header__write(struct perf_header *self, int fd)
 		do_write(fd, &f_attr, sizeof(f_attr));
 	}
 
+	self->event_offset = lseek(fd, 0, SEEK_CUR);
+	self->event_size = event_count * sizeof(struct perf_trace_event_type);
+	if (events)
+		do_write(fd, events, self->event_size);
+
 
 	self->data_offset = lseek(fd, 0, SEEK_CUR);
 
@@ -169,6 +213,10 @@ void perf_header__write(struct perf_header *self, int fd)
 			.offset = self->data_offset,
 			.size	= self->data_size,
 		},
+		.event_types = {
+			.offset = self->event_offset,
+			.size	= self->event_size,
+		},
 	};
 
 	lseek(fd, 0, SEEK_SET);
@@ -234,6 +282,17 @@ struct perf_header *perf_header__read(int fd)
 		lseek(fd, tmp, SEEK_SET);
 	}
 
+	if (f_header.event_types.size) {
+		lseek(fd, f_header.event_types.offset, SEEK_SET);
+		events = malloc(f_header.event_types.size);
+		if (!events)
+			die("nomem");
+		do_read(fd, events, f_header.event_types.size);
+		event_count =  f_header.event_types.size / sizeof(struct perf_trace_event_type);
+	}
+	self->event_offset = f_header.event_types.offset;
+	self->event_size   = f_header.event_types.size;
+
 	self->data_offset = f_header.data.offset;
 	self->data_size   = f_header.data.size;
 

tools/perf/util/header.h

@@ -19,6 +19,8 @@ struct perf_header {
 	s64 attr_offset;
 	u64 data_offset;
 	u64 data_size;
+	u64 event_offset;
+	u64 event_size;
 };
 
 struct perf_header *perf_header__read(int fd);
@@ -27,6 +29,10 @@ void perf_header__write(struct perf_header *self, int fd);
 void perf_header__add_attr(struct perf_header *self,
 			   struct perf_header_attr *attr);
 
+void perf_header__push_event(u64 id, const char *name);
+char *perf_header__find_event(u64 id);
+
+
 struct perf_header_attr *
 perf_header_attr__new(struct perf_counter_attr *attr);
 void perf_header_attr__add_id(struct perf_header_attr *self, u64 id);

tools/perf/util/parse-events.c

@@ -6,6 +6,7 @@
 #include "exec_cmd.h"
 #include "string.h"
 #include "cache.h"
+#include "header.h"
 
 int					nr_counters;
 
@@ -18,6 +19,12 @@ struct event_symbol {
 	const char	*alias;
 };
 
+enum event_result {
+	EVT_FAILED,
+	EVT_HANDLED,
+	EVT_HANDLED_ALL
+};
+
 char debugfs_path[MAXPATHLEN];
 
 #define CHW(x) .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_##x
@@ -139,7 +146,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 512
 
 int valid_debugfs_mount(const char *debugfs)
 {
@@ -344,7 +351,7 @@ static int parse_aliases(const char **str, const char *names[][MAX_ALIASES], int
 	return -1;
 }
 
-static int
+static enum event_result
 parse_generic_hw_event(const char **str, struct perf_counter_attr *attr)
 {
 	const char *s = *str;
@@ -356,7 +363,7 @@ parse_generic_hw_event(const char **str, struct perf_counter_attr *attr)
 	 * then bail out:
 	 */
 	if (cache_type == -1)
-		return 0;
+		return EVT_FAILED;
 
 	while ((cache_op == -1 || cache_result == -1) && *s == '-') {
 		++s;
@@ -402,27 +409,115 @@ parse_generic_hw_event(const char **str, struct perf_counter_attr *attr)
 	attr->type = PERF_TYPE_HW_CACHE;
 
 	*str = s;
-	return 1;
+	return EVT_HANDLED;
+}
+
+static enum event_result
+parse_single_tracepoint_event(char *sys_name,
+			      const char *evt_name,
+			      unsigned int evt_length,
+			      char *flags,
+			      struct perf_counter_attr *attr,
+			      const char **strp)
+{
+	char evt_path[MAXPATHLEN];
+	char id_buf[4];
+	u64 id;
+	int fd;
+
+	if (flags) {
+		if (!strncmp(flags, "record", strlen(flags))) {
+			attr->sample_type |= PERF_SAMPLE_RAW;
+			attr->sample_type |= PERF_SAMPLE_TIME;
+			attr->sample_type |= PERF_SAMPLE_CPU;
+		}
+	}
+
+	snprintf(evt_path, MAXPATHLEN, "%s/%s/%s/id", debugfs_path,
+		 sys_name, evt_name);
+
+	fd = open(evt_path, O_RDONLY);
+	if (fd < 0)
+		return EVT_FAILED;
+
+	if (read(fd, id_buf, sizeof(id_buf)) < 0) {
+		close(fd);
+		return EVT_FAILED;
+	}
+
+	close(fd);
+	id = atoll(id_buf);
+	attr->config = id;
+	attr->type = PERF_TYPE_TRACEPOINT;
+	*strp = evt_name + evt_length;
+
+	return EVT_HANDLED;
+}
+
+/* sys + ':' + event + ':' + flags*/
+#define MAX_EVOPT_LEN	(MAX_EVENT_LENGTH * 2 + 2 + 128)
+static enum event_result
+parse_subsystem_tracepoint_event(char *sys_name, char *flags)
+{
+	char evt_path[MAXPATHLEN];
+	struct dirent *evt_ent;
+	DIR *evt_dir;
+
+	snprintf(evt_path, MAXPATHLEN, "%s/%s", debugfs_path, sys_name);
+	evt_dir = opendir(evt_path);
+
+	if (!evt_dir) {
+		perror("Can't open event dir");
+		return EVT_FAILED;
+	}
+
+	while ((evt_ent = readdir(evt_dir))) {
+		char event_opt[MAX_EVOPT_LEN + 1];
+		int len;
+		unsigned int rem = MAX_EVOPT_LEN;
+
+		if (!strcmp(evt_ent->d_name, ".")
+		    || !strcmp(evt_ent->d_name, "..")
+		    || !strcmp(evt_ent->d_name, "enable")
+		    || !strcmp(evt_ent->d_name, "filter"))
+			continue;
+
+		len = snprintf(event_opt, MAX_EVOPT_LEN, "%s:%s", sys_name,
+			       evt_ent->d_name);
+		if (len < 0)
+			return EVT_FAILED;
+
+		rem -= len;
+		if (flags) {
+			if (rem < strlen(flags) + 1)
+				return EVT_FAILED;
+
+			strcat(event_opt, ":");
+			strcat(event_opt, flags);
+		}
+
+		if (parse_events(NULL, event_opt, 0))
+			return EVT_FAILED;
+	}
+
+	return EVT_HANDLED_ALL;
 }
 
-static int parse_tracepoint_event(const char **strp,
+
+static enum event_result parse_tracepoint_event(const char **strp,
 				    struct perf_counter_attr *attr)
 {
 	const char *evt_name;
 	char *flags;
 	char sys_name[MAX_EVENT_LENGTH];
-	char id_buf[4];
-	int fd;
 	unsigned int sys_length, evt_length;
-	u64 id;
-	char evt_path[MAXPATHLEN];
 
 	if (valid_debugfs_mount(debugfs_path))
 		return 0;
 
 	evt_name = strchr(*strp, ':');
 	if (!evt_name)
-		return 0;
+		return EVT_FAILED;
 
 	sys_length = evt_name - *strp;
 	if (sys_length >= MAX_EVENT_LENGTH)
@@ -434,32 +529,22 @@ static int parse_tracepoint_event(const char **strp,
 
 	flags = strchr(evt_name, ':');
 	if (flags) {
-		*flags = '\0';
+		/* split it out: */
+		evt_name = strndup(evt_name, flags - evt_name);
 		flags++;
-		if (!strncmp(flags, "record", strlen(flags)))
-			attr->sample_type |= PERF_SAMPLE_RAW;
 	}
 
 	evt_length = strlen(evt_name);
 	if (evt_length >= MAX_EVENT_LENGTH)
-		return 0;
+		return EVT_FAILED;
 
-	snprintf(evt_path, MAXPATHLEN, "%s/%s/%s/id", debugfs_path,
-		 sys_name, evt_name);
-	fd = open(evt_path, O_RDONLY);
-	if (fd < 0)
-		return 0;
-
-	if (read(fd, id_buf, sizeof(id_buf)) < 0) {
-		close(fd);
-		return 0;
-	}
-	close(fd);
-	id = atoll(id_buf);
-	attr->config = id;
-	attr->type = PERF_TYPE_TRACEPOINT;
+	if (!strcmp(evt_name, "*")) {
 		*strp = evt_name + evt_length;
-	return 1;
+		return parse_subsystem_tracepoint_event(sys_name, flags);
+	} else
+		return parse_single_tracepoint_event(sys_name, evt_name,
+						     evt_length, flags,
+						     attr, strp);
 }
 
 static int check_events(const char *str, unsigned int i)
@@ -477,7 +562,7 @@ static int check_events(const char *str, unsigned int i)
 	return 0;
 }
 
-static int
+static enum event_result
 parse_symbolic_event(const char **strp, struct perf_counter_attr *attr)
 {
 	const char *str = *strp;
@@ -490,31 +575,32 @@ parse_symbolic_event(const char **strp, struct perf_counter_attr *attr)
 			attr->type = event_symbols[i].type;
 			attr->config = event_symbols[i].config;
 			*strp = str + n;
-			return 1;
+			return EVT_HANDLED;
 		}
 	}
-	return 0;
+	return EVT_FAILED;
 }
 
-static int parse_raw_event(const char **strp, struct perf_counter_attr *attr)
+static enum event_result
+parse_raw_event(const char **strp, struct perf_counter_attr *attr)
 {
 	const char *str = *strp;
 	u64 config;
 	int n;
 
 	if (*str != 'r')
-		return 0;
+		return EVT_FAILED;
 	n = hex2u64(str + 1, &config);
 	if (n > 0) {
 		*strp = str + n + 1;
 		attr->type = PERF_TYPE_RAW;
 		attr->config = config;
-		return 1;
+		return EVT_HANDLED;
 	}
-	return 0;
+	return EVT_FAILED;
 }
 
-static int
+static enum event_result
 parse_numeric_event(const char **strp, struct perf_counter_attr *attr)
 {
 	const char *str = *strp;
@@ -530,13 +616,13 @@ parse_numeric_event(const char **strp, struct perf_counter_attr *attr)
 			attr->type = type;
 			attr->config = config;
 			*strp = endp;
-			return 1;
+			return EVT_HANDLED;
 		}
 	}
-	return 0;
+	return EVT_FAILED;
 }
 
-static int
+static enum event_result
 parse_event_modifier(const char **strp, struct perf_counter_attr *attr)
 {
 	const char *str = *strp;
@@ -569,37 +655,84 @@ parse_event_modifier(const char **strp, struct perf_counter_attr *attr)
  * Each event can have multiple symbolic names.
  * Symbolic names are (almost) exactly matched.
  */
-static int parse_event_symbols(const char **str, struct perf_counter_attr *attr)
+static enum event_result
+parse_event_symbols(const char **str, struct perf_counter_attr *attr)
 {
-	if (!(parse_tracepoint_event(str, attr) ||
-	      parse_raw_event(str, attr) ||
-	      parse_numeric_event(str, attr) ||
-	      parse_symbolic_event(str, attr) ||
-	      parse_generic_hw_event(str, attr)))
-		return 0;
+	enum event_result ret;
+
+	ret = parse_tracepoint_event(str, attr);
+	if (ret != EVT_FAILED)
+		goto modifier;
+
+	ret = parse_raw_event(str, attr);
+	if (ret != EVT_FAILED)
+		goto modifier;
+
+	ret = parse_numeric_event(str, attr);
+	if (ret != EVT_FAILED)
+		goto modifier;
+
+	ret = parse_symbolic_event(str, attr);
+	if (ret != EVT_FAILED)
+		goto modifier;
 
+	ret = parse_generic_hw_event(str, attr);
+	if (ret != EVT_FAILED)
+		goto modifier;
+
+	return EVT_FAILED;
+
+modifier:
 	parse_event_modifier(str, attr);
 
-	return 1;
+	return ret;
+}
+
+static void store_event_type(const char *orgname)
+{
+	char filename[PATH_MAX], *c;
+	FILE *file;
+	int id;
+
+	sprintf(filename, "/sys/kernel/debug/tracing/events/%s/id", orgname);
+	c = strchr(filename, ':');
+	if (c)
+		*c = '/';
+
+	file = fopen(filename, "r");
+	if (!file)
+		return;
+	if (fscanf(file, "%i", &id) < 1)
+		die("cannot store event ID");
+	fclose(file);
+	perf_header__push_event(id, orgname);
 }
 
+
 int parse_events(const struct option *opt __used, const char *str, int unset __used)
 {
 	struct perf_counter_attr attr;
+	enum event_result ret;
+
+	if (strchr(str, ':'))
+		store_event_type(str);
 
 	for (;;) {
 		if (nr_counters == MAX_COUNTERS)
 			return -1;
 
 		memset(&attr, 0, sizeof(attr));
-		if (!parse_event_symbols(&str, &attr))
+		ret = parse_event_symbols(&str, &attr);
+		if (ret == EVT_FAILED)
 			return -1;
 
 		if (!(*str == 0 || *str == ',' || isspace(*str)))
 			return -1;
 
+		if (ret != EVT_HANDLED_ALL) {
 			attrs[nr_counters] = attr;
 			nr_counters++;
+		}
 
 		if (*str == 0)
 			break;

tools/perf/util/parse-options.h

@@ -104,6 +104,8 @@ struct option {
 	{ .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), .argh = "time", .help = (h), .callback = parse_opt_approxidate_cb }
 #define OPT_CALLBACK(s, l, v, a, h, f) \
 	{ .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f) }
+#define OPT_CALLBACK_NOOPT(s, l, v, a, h, f) \
+	{ .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG }
 #define OPT_CALLBACK_DEFAULT(s, l, v, a, h, f, d) \
 	{ .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .defval = (intptr_t)d, .flags = PARSE_OPT_LASTARG_DEFAULT }
 

tools/perf/util/svghelper.h

+#ifndef _INCLUDE_GUARD_SVG_HELPER_
+#define _INCLUDE_GUARD_SVG_HELPER_
+
+#include "types.h"
+
+extern void open_svg(const char *filename, int cpus, int rows);
+extern void svg_box(int Yslot, u64 start, u64 end, const char *type);
+extern void svg_sample(int Yslot, int cpu, u64 start, u64 end, const char *type);
+extern void svg_cpu_box(int cpu, u64 max_frequency, u64 turbo_frequency);
+
+
+extern void svg_process(int cpu, u64 start, u64 end, const char *type, const char *name);
+extern void svg_cstate(int cpu, u64 start, u64 end, int type);
+extern void svg_pstate(int cpu, u64 start, u64 end, u64 freq);
+
+
+extern void svg_time_grid(u64 start, u64 end);
+extern void svg_legenda(void);
+extern void svg_wakeline(u64 start, int row1, int row2);
+extern void svg_partial_wakeline(u64 start, int row1, int row2);
+extern void svg_interrupt(u64 start, int row);
+extern void svg_text(int Yslot, u64 start, const char *text);
+extern void svg_close(void);
+
+#endif

tools/perf/util/thread.h

@@ -7,6 +7,7 @@ struct thread {
 	struct rb_node		rb_node;
 	struct list_head	maps;
 	pid_t			pid;
+	char			shortname[3];
 	char			*comm;
 };