aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--Documentation/trace/power.txt17
-rw-r--r--arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c7
-rw-r--r--arch/x86/kernel/cpu/perf_counter.c60
-rw-r--r--arch/x86/kernel/process.c25
-rw-r--r--drivers/cpuidle/cpuidle.c2
-rw-r--r--include/linux/perf_counter.h80
-rw-r--r--include/trace/events/power.h81
-rw-r--r--include/trace/events/sched.h33
-rw-r--r--kernel/perf_counter.c394
-rw-r--r--kernel/sched_clock.c122
-rw-r--r--kernel/sched_fair.c1
-rw-r--r--kernel/trace/Makefile2
-rw-r--r--kernel/trace/power-traces.c20
-rw-r--r--kernel/trace/trace.h3
-rw-r--r--kernel/trace/trace_entries.h17
-rw-r--r--kernel/trace/trace_power.c218
-rw-r--r--scripts/tracing/power.pl108
-rw-r--r--tools/perf/Documentation/perf-sched.txt41
-rw-r--r--tools/perf/Documentation/perf-timechart.txt35
-rw-r--r--tools/perf/Documentation/perf-trace.txt25
-rw-r--r--tools/perf/Makefile9
-rw-r--r--tools/perf/builtin-record.c56
-rw-r--r--tools/perf/builtin-sched.c2004
-rw-r--r--tools/perf/builtin-timechart.c1151
-rw-r--r--tools/perf/builtin.h6
-rw-r--r--tools/perf/command-list.txt3
-rw-r--r--tools/perf/perf.c2
-rw-r--r--tools/perf/util/event.h10
-rw-r--r--tools/perf/util/header.c67
-rw-r--r--tools/perf/util/header.h6
-rw-r--r--tools/perf/util/parse-events.c237
-rw-r--r--tools/perf/util/parse-options.h2
-rw-r--r--tools/perf/util/svghelper.c384
-rw-r--r--tools/perf/util/svghelper.h25
-rw-r--r--tools/perf/util/thread.c4
-rw-r--r--tools/perf/util/thread.h9
-rw-r--r--tools/perf/util/trace-event-info.c7
-rw-r--r--tools/perf/util/trace-event-parse.c45
-rw-r--r--tools/perf/util/trace-event-read.c6
-rw-r--r--tools/perf/util/trace-event.h5
40 files changed, 4570 insertions, 759 deletions
diff --git a/Documentation/trace/power.txt b/Documentation/trace/power.txt
deleted file mode 100644
index cd805e16dc2..00000000000
--- a/Documentation/trace/power.txt
+++ /dev/null
@@ -1,17 +0,0 @@
-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
diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
index 7bb676c533a..7d5c3b0ea8d 100644
--- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
+++ b/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:
diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index 2732e2c1e4d..dbdf712fae9 100644
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/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 @@ again:
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);
}
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 071166a4ba8..847ab416031 100644
--- a/arch/x86/kernel/process.c
+++ b/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);
}
/*
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
index 8504a210855..ad41f19b8e3 100644
--- a/drivers/cpuidle/cpuidle.c
+++ b/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);
}
/**
diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index 972f90d7a32..bd341007c4f 100644
--- a/include/linux/perf_counter.h
+++ b/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;
- __u32 wakeup_events; /* wakeup every n events */
+ 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 */
diff --git a/include/trace/events/power.h b/include/trace/events/power.h
new file mode 100644
index 00000000000..ea6d579261a
--- /dev/null
+++ b/include/trace/events/power.h
@@ -0,0 +1,81 @@
+#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>
diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index b48f1ad7c94..4069c43f418 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -380,6 +380,39 @@ TRACE_EVENT(sched_stat_wait,
);
/*
+ * 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).
*/
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 8cb94a52d1b..cc768ab81ac 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -2176,6 +2176,13 @@ static int perf_mmap_data_alloc(struct perf_counter *counter, int nr_pages)
data->nr_pages = nr_pages;
atomic_set(&data->lock, -1);
+ if (counter->attr.watermark) {
+ data->watermark = min_t(long, PAGE_SIZE * nr_pages,
+ counter->attr.wakeup_watermark);
+ }
+ if (!data->watermark)
+ data->watermark = max(PAGE_SIZE, PAGE_SIZE * nr_pages / 4);
+
rcu_assign_pointer(counter->data, data);
return 0;
@@ -2315,7 +2322,8 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
lock_limit >>= PAGE_SHIFT;
locked = vma->vm_mm->locked_vm + extra;
- if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) {
+ if ((locked > lock_limit) && perf_paranoid_tracepoint_raw() &&
+ !capable(CAP_IPC_LOCK)) {
ret = -EPERM;
goto unlock;
}
@@ -2504,35 +2512,15 @@ __weak struct perf_callchain_entry *perf_callchain(struct pt_regs *regs)
/*
* Output
*/
-
-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;
-};
-
-static bool perf_output_space(struct perf_mmap_data *data,
- unsigned int offset, unsigned int head)
+static bool perf_output_space(struct perf_mmap_data *data, unsigned long tail,
+ unsigned long offset, unsigned long head)
{
- unsigned long tail;
unsigned long mask;
if (!data->writable)
return true;
mask = (data->nr_pages << PAGE_SHIFT) - 1;
- /*
- * Userspace could choose to issue a mb() before updating the tail
- * pointer. So that all reads will be completed before the write is
- * issued.
- */
- tail = ACCESS_ONCE(data->user_page->data_tail);
- smp_rmb();
offset = (offset - tail) & mask;
head = (head - tail) & mask;
@@ -2633,8 +2621,8 @@ out:
local_irq_restore(handle->flags);
}
-static void perf_output_copy(struct perf_output_handle *handle,
- const void *buf, unsigned int len)
+void perf_output_copy(struct perf_output_handle *handle,
+ const void *buf, unsigned int len)
{
unsigned int pages_mask;
unsigned int offset;
@@ -2669,16 +2657,13 @@ static void perf_output_copy(struct perf_output_handle *handle,
WARN_ON_ONCE(((long)(handle->head - handle->offset)) < 0);
}
-#define perf_output_put(handle, x) \
- perf_output_copy((handle), &(x), sizeof(x))
-
-static int perf_output_begin(struct perf_output_handle *handle,
- struct perf_counter *counter, unsigned int size,
- int nmi, int sample)
+int perf_output_begin(struct perf_output_handle *handle,
+ struct perf_counter *counter, unsigned int size,
+ int nmi, int sample)
{
struct perf_counter *output_counter;
struct perf_mmap_data *data;
- unsigned int offset, head;
+ unsigned long tail, offset, head;
int have_lost;
struct {
struct perf_event_header header;
@@ -2716,16 +2701,23 @@ static int perf_output_begin(struct perf_output_handle *handle,
perf_output_lock(handle);
do {
+ /*
+ * Userspace could choose to issue a mb() before updating the
+ * tail pointer. So that all reads will be completed before the
+ * write is issued.
+ */
+ tail = ACCESS_ONCE(data->user_page->data_tail);
+ smp_rmb();
offset = head = atomic_long_read(&data->head);
head += size;
- if (unlikely(!perf_output_space(data, offset, head)))
+ if (unlikely(!perf_output_space(data, tail, offset, head)))
goto fail;
} while (atomic_long_cmpxchg(&data->head, offset, head) != offset);
handle->offset = offset;
handle->head = head;
- if ((offset >> PAGE_SHIFT) != (head >> PAGE_SHIFT))
+ if (head - tail > data->watermark)
atomic_set(&data->wakeup, 1);
if (have_lost) {
@@ -2749,7 +2741,7 @@ out:
return -ENOSPC;
}
-static void perf_output_end(struct perf_output_handle *handle)
+void perf_output_end(struct perf_output_handle *handle)
{
struct perf_counter *counter = handle->counter;
struct perf_mmap_data *data = handle->data;
@@ -2863,156 +2855,176 @@ static void perf_output_read(struct perf_output_handle *handle,
perf_output_read_one(handle, counter);
}
-void perf_counter_output(struct perf_counter *counter, int nmi,
- struct perf_sample_data *data)
+void perf_output_sample(struct perf_output_handle *handle,
+ struct perf_event_header *header,
+ struct perf_sample_data *data,
+ struct perf_counter *counter)
{
- int ret;
- u64 sample_type = counter->attr.sample_type;
- struct perf_output_handle handle;
- struct perf_event_header header;
- u64 ip;
- struct {
- u32 pid, tid;
- } tid_entry;
- struct perf_callchain_entry *callchain = NULL;
- int callchain_size = 0;
- u64 time;
- struct {
- u32 cpu, reserved;
- } cpu_entry;
+ u64 sample_type = data->type;
- header.type = PERF_EVENT_SAMPLE;
- header.size = sizeof(header);
+ perf_output_put(handle, *header);
- header.misc = 0;
- header.misc |= perf_misc_flags(data->regs);
-
- if (sample_type & PERF_SAMPLE_IP) {
- ip = perf_instruction_pointer(data->regs);
- header.size += sizeof(ip);
- }
-
- if (sample_type & PERF_SAMPLE_TID) {
- /* namespace issues */
- tid_entry.pid = perf_counter_pid(counter, current);
- tid_entry.tid = perf_counter_tid(counter, current);
-
- header.size += sizeof(tid_entry);
- }
+ if (sample_type & PERF_SAMPLE_IP)
+ perf_output_put(handle, data->ip);
- if (sample_type & PERF_SAMPLE_TIME) {
- /*
- * Maybe do better on x86 and provide cpu_clock_nmi()
- */
- time = sched_clock();
+ if (sample_type & PERF_SAMPLE_TID)
+ perf_output_put(handle, data->tid_entry);
- header.size += sizeof(u64);
- }
+ if (sample_type & PERF_SAMPLE_TIME)
+ perf_output_put(handle, data->time);
if (sample_type & PERF_SAMPLE_ADDR)
- header.size += sizeof(u64);
+ perf_output_put(handle, data->addr);
if (sample_type & PERF_SAMPLE_ID)
- header.size += sizeof(u64);
+ perf_output_put(handle, data->id);
if (sample_type & PERF_SAMPLE_STREAM_ID)
- header.size += sizeof(u64);
-
- if (sample_type & PERF_SAMPLE_CPU) {
- header.size += sizeof(cpu_entry);
+ perf_output_put(handle, data->stream_id);
- cpu_entry.cpu = raw_smp_processor_id();
- cpu_entry.reserved = 0;
- }
+ if (sample_type & PERF_SAMPLE_CPU)
+ perf_output_put(handle, data->cpu_entry);
if (sample_type & PERF_SAMPLE_PERIOD)
- header.size += sizeof(u64);