#include "util/trace-event.h"
#include "util/debug.h"
+#include "util/data_map.h"
#include <sys/types.h>
#include <sys/prctl.h>
#include <math.h>
static char const *input_name = "perf.data";
-static int input;
-static unsigned long page_size;
-static unsigned long mmap_window = 32;
-
-static unsigned long total_comm = 0;
-
-static struct rb_root threads;
-static struct thread *last_match;
static struct perf_header *header;
static u64 sample_type;
static char default_sort_order[] = "avg, max, switch, runtime";
static char *sort_order = default_sort_order;
+static int profile_cpu = -1;
+
#define PR_SET_NAME 15 /* Set process name */
#define MAX_CPUS 4096
-#define BUG_ON(x) assert(!(x))
-
static u64 run_measurement_overhead;
static u64 sleep_measurement_overhead;
static unsigned long nr_tasks;
-struct sched_event;
+struct sched_atom;
struct task_desc {
unsigned long nr;
unsigned long nr_events;
unsigned long curr_event;
- struct sched_event **events;
+ struct sched_atom **atoms;
pthread_t thread;
sem_t sleep_sem;
SCHED_EVENT_RUN,
SCHED_EVENT_SLEEP,
SCHED_EVENT_WAKEUP,
+ SCHED_EVENT_MIGRATION,
};
-struct sched_event {
+struct sched_atom {
enum sched_event_type type;
u64 timestamp;
u64 duration;
static u64 run_avg;
static unsigned long replay_repeat = 10;
+static unsigned long nr_timestamps;
+static unsigned long nr_unordered_timestamps;
+static unsigned long nr_state_machine_bugs;
+static unsigned long nr_context_switch_bugs;
+static unsigned long nr_events;
+static unsigned long nr_lost_chunks;
+static unsigned long nr_lost_events;
#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
struct work_atom {
struct list_head list;
enum thread_state state;
+ u64 sched_out_time;
u64 wake_up_time;
u64 sched_in_time;
u64 runtime;
};
-struct task_atoms {
- struct list_head atom_list;
+struct work_atoms {
+ struct list_head work_list;
struct thread *thread;
struct rb_node node;
u64 max_lat;
u64 total_runtime;
};
-typedef int (*sort_fn_t)(struct task_atoms *, struct task_atoms *);
+typedef int (*sort_fn_t)(struct work_atoms *, struct work_atoms *);
static struct rb_root atom_root, sorted_atom_root;
static u64 all_runtime;
static u64 all_count;
-static int read_events(void);
-
static u64 get_nsecs(void)
{
printf("sleep measurement overhead: %Ld nsecs\n", min_delta);
}
-static struct sched_event *
+static struct sched_atom *
get_new_event(struct task_desc *task, u64 timestamp)
{
- struct sched_event *event = calloc(1, sizeof(*event));
+ struct sched_atom *event = zalloc(sizeof(*event));
unsigned long idx = task->nr_events;
size_t size;
event->nr = idx;
task->nr_events++;
- size = sizeof(struct sched_event *) * task->nr_events;
- task->events = realloc(task->events, size);
- BUG_ON(!task->events);
+ size = sizeof(struct sched_atom *) * task->nr_events;
+ task->atoms = realloc(task->atoms, size);
+ BUG_ON(!task->atoms);
- task->events[idx] = event;
+ task->atoms[idx] = event;
return event;
}
-static struct sched_event *last_event(struct task_desc *task)
+static struct sched_atom *last_event(struct task_desc *task)
{
if (!task->nr_events)
return NULL;
- return task->events[task->nr_events - 1];
+ return task->atoms[task->nr_events - 1];
}
static void
add_sched_event_run(struct task_desc *task, u64 timestamp, u64 duration)
{
- struct sched_event *event, *curr_event = last_event(task);
+ struct sched_atom *event, *curr_event = last_event(task);
/*
* optimize an existing RUN event by merging this one
add_sched_event_wakeup(struct task_desc *task, u64 timestamp,
struct task_desc *wakee)
{
- struct sched_event *event, *wakee_event;
+ struct sched_atom *event, *wakee_event;
event = get_new_event(task, timestamp);
event->type = SCHED_EVENT_WAKEUP;
return;
}
- wakee_event->wait_sem = calloc(1, sizeof(*wakee_event->wait_sem));
+ wakee_event->wait_sem = zalloc(sizeof(*wakee_event->wait_sem));
sem_init(wakee_event->wait_sem, 0, 0);
wakee_event->specific_wait = 1;
event->wait_sem = wakee_event->wait_sem;
add_sched_event_sleep(struct task_desc *task, u64 timestamp,
u64 task_state __used)
{
- struct sched_event *event = get_new_event(task, timestamp);
+ struct sched_atom *event = get_new_event(task, timestamp);
event->type = SCHED_EVENT_SLEEP;
if (task)
return task;
- task = calloc(1, sizeof(*task));
+ task = zalloc(sizeof(*task));
task->pid = pid;
task->nr = nr_tasks;
strcpy(task->comm, comm);
}
static void
-process_sched_event(struct task_desc *this_task __used, struct sched_event *event)
+process_sched_event(struct task_desc *this_task __used, struct sched_atom *atom)
{
int ret = 0;
u64 now;
long long delta;
now = get_nsecs();
- delta = start_time + event->timestamp - now;
+ delta = start_time + atom->timestamp - now;
- switch (event->type) {
+ switch (atom->type) {
case SCHED_EVENT_RUN:
- burn_nsecs(event->duration);
+ burn_nsecs(atom->duration);
break;
case SCHED_EVENT_SLEEP:
- if (event->wait_sem)
- ret = sem_wait(event->wait_sem);
+ if (atom->wait_sem)
+ ret = sem_wait(atom->wait_sem);
BUG_ON(ret);
break;
case SCHED_EVENT_WAKEUP:
- if (event->wait_sem)
- ret = sem_post(event->wait_sem);
+ if (atom->wait_sem)
+ ret = sem_post(atom->wait_sem);
BUG_ON(ret);
break;
+ case SCHED_EVENT_MIGRATION:
+ break;
default:
BUG_ON(1);
}
for (i = 0; i < this_task->nr_events; i++) {
this_task->curr_event = i;
- process_sched_event(this_task, this_task->events[i]);
+ process_sched_event(this_task, this_task->atoms[i]);
}
cpu_usage_1 = get_cpu_usage_nsec_self();
printf("the sleep test took %Ld nsecs\n", T1-T0);
}
-static void __cmd_replay(void)
-{
- unsigned long i;
-
- calibrate_run_measurement_overhead();
- calibrate_sleep_measurement_overhead();
-
- test_calibrations();
-
- read_events();
-
- printf("nr_run_events: %ld\n", nr_run_events);
- printf("nr_sleep_events: %ld\n", nr_sleep_events);
- printf("nr_wakeup_events: %ld\n", nr_wakeup_events);
-
- if (targetless_wakeups)
- printf("target-less wakeups: %ld\n", targetless_wakeups);
- if (multitarget_wakeups)
- printf("multi-target wakeups: %ld\n", multitarget_wakeups);
- if (nr_run_events_optimized)
- printf("run events optimized: %ld\n",
- nr_run_events_optimized);
-
- print_task_traces();
- add_cross_task_wakeups();
-
- create_tasks();
- printf("------------------------------------------------------------\n");
- for (i = 0; i < replay_repeat; i++)
- run_one_test();
-}
-
-static int
-process_comm_event(event_t *event, unsigned long offset, unsigned long head)
-{
- struct thread *thread;
-
- thread = threads__findnew(event->comm.pid, &threads, &last_match);
-
- dump_printf("%p [%p]: PERF_EVENT_COMM: %s:%d\n",
- (void *)(offset + head),
- (void *)(long)(event->header.size),
- event->comm.comm, event->comm.pid);
-
- if (thread == NULL ||
- thread__set_comm(thread, event->comm.comm)) {
- dump_printf("problem processing PERF_EVENT_COMM, skipping event.\n");
- return -1;
- }
- total_comm++;
-
- return 0;
-}
-
-
struct raw_event_sample {
u32 size;
char data[0];
u32 next_prio;
};
+struct trace_runtime_event {
+ u32 size;
+
+ u16 common_type;
+ u8 common_flags;
+ u8 common_preempt_count;
+ u32 common_pid;
+ u32 common_tgid;
+
+ char comm[16];
+ u32 pid;
+ u64 runtime;
+ u64 vruntime;
+};
struct trace_wakeup_event {
u32 size;
u32 child_pid;
};
+struct trace_migrate_task_event {
+ u32 size;
+
+ u16 common_type;
+ u8 common_flags;
+ u8 common_preempt_count;
+ u32 common_pid;
+ u32 common_tgid;
+
+ char comm[16];
+ u32 pid;
+
+ u32 prio;
+ u32 cpu;
+};
+
struct trace_sched_handler {
void (*switch_event)(struct trace_switch_event *,
struct event *,
u64 timestamp,
struct thread *thread);
+ void (*runtime_event)(struct trace_runtime_event *,
+ struct event *,
+ int cpu,
+ u64 timestamp,
+ struct thread *thread);
+
void (*wakeup_event)(struct trace_wakeup_event *,
struct event *,
int cpu,
int cpu,
u64 timestamp,
struct thread *thread);
+
+ void (*migrate_task_event)(struct trace_migrate_task_event *,
+ struct event *,
+ int cpu,
+ u64 timestamp,
+ struct thread *thread);
};
add_sched_event_wakeup(waker, timestamp, wakee);
}
-static unsigned long cpu_last_switched[MAX_CPUS];
+static u64 cpu_last_switched[MAX_CPUS];
static void
replay_switch_event(struct trace_switch_event *switch_event,
static LIST_HEAD(cmp_pid);
static int
-thread_lat_cmp(struct list_head *list, struct task_atoms *l, struct task_atoms *r)
+thread_lat_cmp(struct list_head *list, struct work_atoms *l, struct work_atoms *r)
{
struct sort_dimension *sort;
int ret = 0;
return ret;
}
-static struct task_atoms *
+static struct work_atoms *
thread_atoms_search(struct rb_root *root, struct thread *thread,
struct list_head *sort_list)
{
struct rb_node *node = root->rb_node;
- struct task_atoms key = { .thread = thread };
+ struct work_atoms key = { .thread = thread };
while (node) {
- struct task_atoms *atoms;
+ struct work_atoms *atoms;
int cmp;
- atoms = container_of(node, struct task_atoms, node);
+ atoms = container_of(node, struct work_atoms, node);
cmp = thread_lat_cmp(sort_list, &key, atoms);
if (cmp > 0)
}
static void
-__thread_latency_insert(struct rb_root *root, struct task_atoms *data,
+__thread_latency_insert(struct rb_root *root, struct work_atoms *data,
struct list_head *sort_list)
{
struct rb_node **new = &(root->rb_node), *parent = NULL;
while (*new) {
- struct task_atoms *this;
+ struct work_atoms *this;
int cmp;
- this = container_of(*new, struct task_atoms, node);
+ this = container_of(*new, struct work_atoms, node);
parent = *new;
cmp = thread_lat_cmp(sort_list, data, this);
static void thread_atoms_insert(struct thread *thread)
{
- struct task_atoms *atoms;
-
- atoms = calloc(sizeof(*atoms), 1);
+ struct work_atoms *atoms = zalloc(sizeof(*atoms));
if (!atoms)
die("No memory");
atoms->thread = thread;
- INIT_LIST_HEAD(&atoms->atom_list);
+ INIT_LIST_HEAD(&atoms->work_list);
__thread_latency_insert(&atom_root, atoms, &cmp_pid);
}
}
static void
-lat_sched_out(struct task_atoms *atoms,
- struct trace_switch_event *switch_event __used,
- u64 delta,
- u64 timestamp)
+add_sched_out_event(struct work_atoms *atoms,
+ char run_state,
+ u64 timestamp)
{
- struct work_atom *atom;
-
- atom = calloc(sizeof(*atom), 1);
+ struct work_atom *atom = zalloc(sizeof(*atom));
if (!atom)
die("Non memory");
- if (sched_out_state(switch_event) == 'R') {
+ atom->sched_out_time = timestamp;
+
+ if (run_state == 'R') {
atom->state = THREAD_WAIT_CPU;
- atom->wake_up_time = timestamp;
+ atom->wake_up_time = atom->sched_out_time;
}
- atom->runtime = delta;
- list_add_tail(&atom->list, &atoms->atom_list);
+ list_add_tail(&atom->list, &atoms->work_list);
}
static void
-lat_sched_in(struct task_atoms *atoms, u64 timestamp)
+add_runtime_event(struct work_atoms *atoms, u64 delta, u64 timestamp __used)
+{
+ struct work_atom *atom;
+
+ BUG_ON(list_empty(&atoms->work_list));
+
+ atom = list_entry(atoms->work_list.prev, struct work_atom, list);
+
+ atom->runtime += delta;
+ atoms->total_runtime += delta;
+}
+
+static void
+add_sched_in_event(struct work_atoms *atoms, u64 timestamp)
{
struct work_atom *atom;
u64 delta;
- if (list_empty(&atoms->atom_list))
+ if (list_empty(&atoms->work_list))
return;
- atom = list_entry(atoms->atom_list.prev, struct work_atom, list);
+ atom = list_entry(atoms->work_list.prev, struct work_atom, list);
if (atom->state != THREAD_WAIT_CPU)
return;
if (delta > atoms->max_lat)
atoms->max_lat = delta;
atoms->nb_atoms++;
- atoms->total_runtime += atom->runtime;
}
static void
u64 timestamp,
struct thread *thread __used)
{
- struct task_atoms *out_atoms, *in_atoms;
+ struct work_atoms *out_events, *in_events;
struct thread *sched_out, *sched_in;
u64 timestamp0;
s64 delta;
- if (cpu >= MAX_CPUS || cpu < 0)
- return;
+ BUG_ON(cpu >= MAX_CPUS || cpu < 0);
timestamp0 = cpu_last_switched[cpu];
cpu_last_switched[cpu] = timestamp;
die("hm, delta: %Ld < 0 ?\n", delta);
- sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match);
- sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match);
+ sched_out = threads__findnew(switch_event->prev_pid);
+ sched_in = threads__findnew(switch_event->next_pid);
- in_atoms = thread_atoms_search(&atom_root, sched_in, &cmp_pid);
- if (!in_atoms) {
+ out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
+ if (!out_events) {
+ thread_atoms_insert(sched_out);
+ out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
+ if (!out_events)
+ die("out-event: Internal tree error");
+ }
+ add_sched_out_event(out_events, sched_out_state(switch_event), timestamp);
+
+ in_events = thread_atoms_search(&atom_root, sched_in, &cmp_pid);
+ if (!in_events) {
thread_atoms_insert(sched_in);
- in_atoms = thread_atoms_search(&atom_root, sched_in, &cmp_pid);
- if (!in_atoms)
- die("in-atom: Internal tree error");
+ in_events = thread_atoms_search(&atom_root, sched_in, &cmp_pid);
+ if (!in_events)
+ die("in-event: Internal tree error");
+ /*
+ * Take came in we have not heard about yet,
+ * add in an initial atom in runnable state:
+ */
+ add_sched_out_event(in_events, 'R', timestamp);
}
+ add_sched_in_event(in_events, timestamp);
+}
- out_atoms = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
- if (!out_atoms) {
- thread_atoms_insert(sched_out);
- out_atoms = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
- if (!out_atoms)
- die("out-atom: Internal tree error");
+static void
+latency_runtime_event(struct trace_runtime_event *runtime_event,
+ struct event *event __used,
+ int cpu,
+ u64 timestamp,
+ struct thread *this_thread __used)
+{
+ struct thread *thread = threads__findnew(runtime_event->pid);
+ struct work_atoms *atoms = thread_atoms_search(&atom_root, thread, &cmp_pid);
+
+ BUG_ON(cpu >= MAX_CPUS || cpu < 0);
+ if (!atoms) {
+ thread_atoms_insert(thread);
+ atoms = thread_atoms_search(&atom_root, thread, &cmp_pid);
+ if (!atoms)
+ die("in-event: Internal tree error");
+ add_sched_out_event(atoms, 'R', timestamp);
}
- lat_sched_in(in_atoms, timestamp);
- lat_sched_out(out_atoms, switch_event, delta, timestamp);
+ add_runtime_event(atoms, runtime_event->runtime, timestamp);
}
static void
latency_wakeup_event(struct trace_wakeup_event *wakeup_event,
- struct event *event __used,
+ struct event *__event __used,
int cpu __used,
u64 timestamp,
struct thread *thread __used)
{
- struct task_atoms *atoms;
+ struct work_atoms *atoms;
struct work_atom *atom;
struct thread *wakee;
if (!wakeup_event->success)
return;
- wakee = threads__findnew(wakeup_event->pid, &threads, &last_match);
+ wakee = threads__findnew(wakeup_event->pid);
atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid);
if (!atoms) {
thread_atoms_insert(wakee);
- return;
+ atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid);
+ if (!atoms)
+ die("wakeup-event: Internal tree error");
+ add_sched_out_event(atoms, 'S', timestamp);
}
- if (list_empty(&atoms->atom_list))
- return;
+ BUG_ON(list_empty(&atoms->work_list));
- atom = list_entry(atoms->atom_list.prev, struct work_atom, list);
+ atom = list_entry(atoms->work_list.prev, struct work_atom, list);
+
+ /*
+ * You WILL be missing events if you've recorded only
+ * one CPU, or are only looking at only one, so don't
+ * make useless noise.
+ */
+ if (profile_cpu == -1 && atom->state != THREAD_SLEEPING)
+ nr_state_machine_bugs++;
- if (atom->state != THREAD_SLEEPING)
+ nr_timestamps++;
+ if (atom->sched_out_time > timestamp) {
+ nr_unordered_timestamps++;
return;
+ }
atom->state = THREAD_WAIT_CPU;
atom->wake_up_time = timestamp;
}
+static void
+latency_migrate_task_event(struct trace_migrate_task_event *migrate_task_event,
+ struct event *__event __used,
+ int cpu __used,
+ u64 timestamp,
+ struct thread *thread __used)
+{
+ struct work_atoms *atoms;
+ struct work_atom *atom;
+ struct thread *migrant;
+
+ /*
+ * Only need to worry about migration when profiling one CPU.
+ */
+ if (profile_cpu == -1)
+ return;
+
+ migrant = threads__findnew(migrate_task_event->pid);
+ atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid);
+ if (!atoms) {
+ thread_atoms_insert(migrant);
+ register_pid(migrant->pid, migrant->comm);
+ atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid);
+ if (!atoms)
+ die("migration-event: Internal tree error");
+ add_sched_out_event(atoms, 'R', timestamp);
+ }
+
+ BUG_ON(list_empty(&atoms->work_list));
+
+ atom = list_entry(atoms->work_list.prev, struct work_atom, list);
+ atom->sched_in_time = atom->sched_out_time = atom->wake_up_time = timestamp;
+
+ nr_timestamps++;
+
+ if (atom->sched_out_time > timestamp)
+ nr_unordered_timestamps++;
+}
+
static struct trace_sched_handler lat_ops = {
.wakeup_event = latency_wakeup_event,
.switch_event = latency_switch_event,
+ .runtime_event = latency_runtime_event,
.fork_event = latency_fork_event,
+ .migrate_task_event = latency_migrate_task_event,
};
-static void output_lat_thread(struct task_atoms *atom_list)
+static void output_lat_thread(struct work_atoms *work_list)
{
int i;
int ret;
u64 avg;
- if (!atom_list->nb_atoms)
+ if (!work_list->nb_atoms)
+ return;
+ /*
+ * Ignore idle threads:
+ */
+ if (!strcmp(work_list->thread->comm, "swapper"))
return;
- all_runtime += atom_list->total_runtime;
- all_count += atom_list->nb_atoms;
+ all_runtime += work_list->total_runtime;
+ all_count += work_list->nb_atoms;
- ret = printf(" %s ", atom_list->thread->comm);
+ ret = printf(" %s:%d ", work_list->thread->comm, work_list->thread->pid);
- for (i = 0; i < 19 - ret; i++)
+ for (i = 0; i < 24 - ret; i++)
printf(" ");
- avg = atom_list->total_lat / atom_list->nb_atoms;
+ avg = work_list->total_lat / work_list->nb_atoms;
- printf("|%9.3f ms |%9llu | avg:%9.3f ms | max:%9.3f ms |\n",
- (double)atom_list->total_runtime / 1e6,
- atom_list->nb_atoms, (double)avg / 1e6,
- (double)atom_list->max_lat / 1e6);
+ printf("|%11.3f ms |%9llu | avg:%9.3f ms | max:%9.3f ms |\n",
+ (double)work_list->total_runtime / 1e6,
+ work_list->nb_atoms, (double)avg / 1e6,
+ (double)work_list->max_lat / 1e6);
}
-static int pid_cmp(struct task_atoms *l, struct task_atoms *r)
+static int pid_cmp(struct work_atoms *l, struct work_atoms *r)
{
if (l->thread->pid < r->thread->pid)
return -1;
.cmp = pid_cmp,
};
-static int avg_cmp(struct task_atoms *l, struct task_atoms *r)
+static int avg_cmp(struct work_atoms *l, struct work_atoms *r)
{
u64 avgl, avgr;
.cmp = avg_cmp,
};
-static int max_cmp(struct task_atoms *l, struct task_atoms *r)
+static int max_cmp(struct work_atoms *l, struct work_atoms *r)
{
if (l->max_lat < r->max_lat)
return -1;
.cmp = max_cmp,
};
-static int switch_cmp(struct task_atoms *l, struct task_atoms *r)
+static int switch_cmp(struct work_atoms *l, struct work_atoms *r)
{
if (l->nb_atoms < r->nb_atoms)
return -1;
.cmp = switch_cmp,
};
-static int runtime_cmp(struct task_atoms *l, struct task_atoms *r)
+static int runtime_cmp(struct work_atoms *l, struct work_atoms *r)
{
if (l->total_runtime < r->total_runtime)
return -1;
static LIST_HEAD(sort_list);
-static int sort_dimension__add(char *tok, struct list_head *list)
+static int sort_dimension__add(const char *tok, struct list_head *list)
{
int i;
struct rb_node *node;
for (;;) {
- struct task_atoms *data;
+ struct work_atoms *data;
node = rb_first(&atom_root);
if (!node)
break;
rb_erase(node, &atom_root);
- data = rb_entry(node, struct task_atoms, node);
+ data = rb_entry(node, struct work_atoms, node);
__thread_latency_insert(&sorted_atom_root, data, &sort_list);
}
}
-static void __cmd_lat(void)
-{
- struct rb_node *next;
-
- setup_pager();
- read_events();
- sort_lat();
-
- printf("-----------------------------------------------------------------------------------\n");
- printf(" Task | Runtime ms | Switches | Average delay ms | Maximum delay ms |\n");
- printf("-----------------------------------------------------------------------------------\n");
-
- next = rb_first(&sorted_atom_root);
-
- while (next) {
- struct task_atoms *atom_list;
-
- atom_list = rb_entry(next, struct task_atoms, node);
- output_lat_thread(atom_list);
- next = rb_next(next);
- }
-
- printf("-----------------------------------------------------------------------------------\n");
- printf(" TOTAL: |%9.3f ms |%9Ld |\n",
- (double)all_runtime/1e6, all_count);
- printf("---------------------------------------------\n");
-}
-
static struct trace_sched_handler *trace_handler;
static void
FILL_FIELD(wakeup_event, success, event, raw->data);
FILL_FIELD(wakeup_event, cpu, event, raw->data);
- trace_handler->wakeup_event(&wakeup_event, event, cpu, timestamp, thread);
+ if (trace_handler->wakeup_event)
+ trace_handler->wakeup_event(&wakeup_event, event, cpu, timestamp, thread);
+}
+
+/*
+ * Track the current task - that way we can know whether there's any
+ * weird events, such as a task being switched away that is not current.
+ */
+static int max_cpu;
+
+static u32 curr_pid[MAX_CPUS] = { [0 ... MAX_CPUS-1] = -1 };
+
+static struct thread *curr_thread[MAX_CPUS];
+
+static char next_shortname1 = 'A';
+static char next_shortname2 = '0';
+
+static void
+map_switch_event(struct trace_switch_event *switch_event,
+ struct event *event __used,
+ int this_cpu,
+ u64 timestamp,
+ struct thread *thread __used)
+{
+ struct thread *sched_out, *sched_in;
+ int new_shortname;
+ u64 timestamp0;
+ s64 delta;
+ int cpu;
+
+ BUG_ON(this_cpu >= MAX_CPUS || this_cpu < 0);
+
+ if (this_cpu > max_cpu)
+ max_cpu = this_cpu;
+
+ timestamp0 = cpu_last_switched[this_cpu];
+ cpu_last_switched[this_cpu] = timestamp;
+ if (timestamp0)
+ delta = timestamp - timestamp0;
+ else
+ delta = 0;
+
+ if (delta < 0)
+ die("hm, delta: %Ld < 0 ?\n", delta);
+
+
+ sched_out = threads__findnew(switch_event->prev_pid);
+ sched_in = threads__findnew(switch_event->next_pid);
+
+ curr_thread[this_cpu] = sched_in;
+
+ printf(" ");
+
+ new_shortname = 0;
+ if (!sched_in->shortname[0]) {
+ sched_in->shortname[0] = next_shortname1;
+ sched_in->shortname[1] = next_shortname2;
+
+ if (next_shortname1 < 'Z') {
+ next_shortname1++;
+ } else {
+ next_shortname1='A';
+ if (next_shortname2 < '9') {
+ next_shortname2++;
+ } else {
+ next_shortname2='0';
+ }
+ }
+ new_shortname = 1;
+ }
+
+ for (cpu = 0; cpu <= max_cpu; cpu++) {
+ if (cpu != this_cpu)
+ printf(" ");
+ else
+ printf("*");
+
+ if (curr_thread[cpu]) {
+ if (curr_thread[cpu]->pid)
+ printf("%2s ", curr_thread[cpu]->shortname);
+ else
+ printf(". ");
+ } else
+ printf(" ");
+ }
+
+ printf(" %12.6f secs ", (double)timestamp/1e9);
+ if (new_shortname) {
+ printf("%s => %s:%d\n",
+ sched_in->shortname, sched_in->comm, sched_in->pid);
+ } else {
+ printf("\n");
+ }
}
+
static void
process_sched_switch_event(struct raw_event_sample *raw,
struct event *event,
- int cpu __used,
+ int this_cpu,
u64 timestamp __used,
struct thread *thread __used)
{
FILL_FIELD(switch_event, next_pid, event, raw->data);
FILL_FIELD(switch_event, next_prio, event, raw->data);
- trace_handler->switch_event(&switch_event, event, cpu, timestamp, thread);
+ if (curr_pid[this_cpu] != (u32)-1) {
+ /*
+ * Are we trying to switch away a PID that is
+ * not current?
+ */
+ if (curr_pid[this_cpu] != switch_event.prev_pid)
+ nr_context_switch_bugs++;
+ }
+ if (trace_handler->switch_event)
+ trace_handler->switch_event(&switch_event, event, this_cpu, timestamp, thread);
+
+ curr_pid[this_cpu] = switch_event.next_pid;
+}
+
+static void
+process_sched_runtime_event(struct raw_event_sample *raw,
+ struct event *event,
+ int cpu __used,
+ u64 timestamp __used,
+ struct thread *thread __used)
+{
+ struct trace_runtime_event runtime_event;
+
+ FILL_ARRAY(runtime_event, comm, event, raw->data);
+ FILL_FIELD(runtime_event, pid, event, raw->data);
+ FILL_FIELD(runtime_event, runtime, event, raw->data);
+ FILL_FIELD(runtime_event, vruntime, event, raw->data);
+
+ if (trace_handler->runtime_event)
+ trace_handler->runtime_event(&runtime_event, event, cpu, timestamp, thread);
}
static void
FILL_ARRAY(fork_event, child_comm, event, raw->data);
FILL_FIELD(fork_event, child_pid, event, raw->data);
- trace_handler->fork_event(&fork_event, event, cpu, timestamp, thread);
+ if (trace_handler->fork_event)
+ trace_handler->fork_event(&fork_event, event, cpu, timestamp, thread);
}
static void
}
static void
+process_sched_migrate_task_event(struct raw_event_sample *raw,
+ struct event *event,
+ int cpu __used,
+ u64 timestamp __used,
+ struct thread *thread __used)
+{
+ struct trace_migrate_task_event migrate_task_event;
+
+ FILL_COMMON_FIELDS(migrate_task_event, event, raw->data);
+
+ FILL_ARRAY(migrate_task_event, comm, event, raw->data);
+ FILL_FIELD(migrate_task_event, pid, event, raw->data);
+ FILL_FIELD(migrate_task_event, prio, event, raw->data);
+ FILL_FIELD(migrate_task_event, cpu, event, raw->data);
+
+ if (trace_handler->migrate_task_event)
+ trace_handler->migrate_task_event(&migrate_task_event, event, cpu, timestamp, thread);
+}
+
+static void
process_raw_event(event_t *raw_event __used, void *more_data,
int cpu, u64 timestamp, struct thread *thread)
{
if (!strcmp(event->name, "sched_switch"))
process_sched_switch_event(raw, event, cpu, timestamp, thread);
+ if (!strcmp(event->name, "sched_stat_runtime"))
+ process_sched_runtime_event(raw, event, cpu, timestamp, thread);
if (!strcmp(event->name, "sched_wakeup"))
process_sched_wakeup_event(raw, event, cpu, timestamp, thread);
if (!strcmp(event->name, "sched_wakeup_new"))
process_sched_fork_event(raw, event, cpu, timestamp, thread);
if (!strcmp(event->name, "sched_process_exit"))
process_sched_exit_event(event, cpu, timestamp, thread);
+ if (!strcmp(event->name, "sched_migrate_task"))
+ process_sched_migrate_task_event(raw, event, cpu, timestamp, thread);
}
-static int
-process_sample_event(event_t *event, unsigned long offset, unsigned long head)
+static int process_sample_event(event_t *event)
{
- char level;
- int show = 0;
- struct dso *dso = NULL;
struct thread *thread;
u64 ip = event->ip.ip;
u64 timestamp = -1;
u32 cpu = -1;
u64 period = 1;
void *more_data = event->ip.__more_data;
- int cpumode;
- thread = threads__findnew(event->ip.pid, &threads, &last_match);
+ if (!(sample_type & PERF_SAMPLE_RAW))
+ return 0;
+
+ thread = threads__findnew(event->ip.pid);
if (sample_type & PERF_SAMPLE_TIME) {
timestamp = *(u64 *)more_data;
more_data += sizeof(u64);
}
- dump_printf("%p [%p]: PERF_EVENT_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n",
- (void *)(offset + head),
- (void *)(long)(event->header.size),
+ dump_printf("(IP, %d): %d/%d: %p period: %Ld\n",
event->header.misc,
event->ip.pid, event->ip.tid,
(void *)(long)ip,
(long long)period);
- dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
-
if (thread == NULL) {
- eprintf("problem processing %d event, skipping it.\n",
- event->header.type);
+ pr_debug("problem processing %d event, skipping it.\n",
+ event->header.type);
return -1;
}
- cpumode = event->header.misc & PERF_EVENT_MISC_CPUMODE_MASK;
-
- if (cpumode == PERF_EVENT_MISC_KERNEL) {
- show = SHOW_KERNEL;
- level = 'k';
-
- dso = kernel_dso;
-
- dump_printf(" ...... dso: %s\n", dso->name);
-
- } else if (cpumode == PERF_EVENT_MISC_USER) {
-
- show = SHOW_USER;
- level = '.';
-
- } else {
- show = SHOW_HV;
- level = 'H';
-
- dso = hypervisor_dso;
+ dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
- dump_printf(" ...... dso: [hypervisor]\n");
- }
+ if (profile_cpu != -1 && profile_cpu != (int) cpu)
+ return 0;
- if (sample_type & PERF_SAMPLE_RAW)
- process_raw_event(event, more_data, cpu, timestamp, thread);
+ process_raw_event(event, more_data, cpu, timestamp, thread);
return 0;
}
-static int
-process_event(event_t *event, unsigned long offset, unsigned long head)
+static int process_lost_event(event_t *event __used)
{
- trace_event(event);
+ nr_lost_chunks++;
+ nr_lost_events += event->lost.lost;
- switch (event->header.type) {
- case PERF_EVENT_MMAP ... PERF_EVENT_LOST:
- return 0;
-
- case PERF_EVENT_COMM:
- return process_comm_event(event, offset, head);
-
- case PERF_EVENT_EXIT ... PERF_EVENT_READ:
- return 0;
+ return 0;
+}
- case PERF_EVENT_SAMPLE:
- return process_sample_event(event, offset, head);
+static int sample_type_check(u64 type)
+{
+ sample_type = type;
- case PERF_EVENT_MAX:
- default:
+ if (!(sample_type & PERF_SAMPLE_RAW)) {
+ fprintf(stderr,
+ "No trace sample to read. Did you call perf record "
+ "without -R?");
return -1;
}
return 0;
}
+static struct perf_file_handler file_handler = {
+ .process_sample_event = process_sample_event,
+ .process_comm_event = event__process_comm,
+ .process_lost_event = process_lost_event,
+ .sample_type_check = sample_type_check,
+};
+
static int read_events(void)
{
- int ret, rc = EXIT_FAILURE;
- unsigned long offset = 0;
- unsigned long head = 0;
- struct stat perf_stat;
- event_t *event;
- uint32_t size;
- char *buf;
-
- trace_report();
- register_idle_thread(&threads, &last_match);
-
- input = open(input_name, O_RDONLY);
- if (input < 0) {
- perror("failed to open file");
- exit(-1);
- }
+ register_idle_thread();
+ register_perf_file_handler(&file_handler);
- ret = fstat(input, &perf_stat);
- if (ret < 0) {
- perror("failed to stat file");
- exit(-1);
- }
+ return mmap_dispatch_perf_file(&header, input_name, 0, 0,
+ &event__cwdlen, &event__cwd);
+}
- if (!perf_stat.st_size) {
- fprintf(stderr, "zero-sized file, nothing to do!\n");
- exit(0);
+static void print_bad_events(void)
+{
+ if (nr_unordered_timestamps && nr_timestamps) {
+ printf(" INFO: %.3f%% unordered timestamps (%ld out of %ld)\n",
+ (double)nr_unordered_timestamps/(double)nr_timestamps*100.0,
+ nr_unordered_timestamps, nr_timestamps);
+ }
+ if (nr_lost_events && nr_events) {
+ printf(" INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n",
+ (double)nr_lost_events/(double)nr_events*100.0,
+ nr_lost_events, nr_events, nr_lost_chunks);
+ }
+ if (nr_state_machine_bugs && nr_timestamps) {
+ printf(" INFO: %.3f%% state machine bugs (%ld out of %ld)",
+ (double)nr_state_machine_bugs/(double)nr_timestamps*100.0,
+ nr_state_machine_bugs, nr_timestamps);
+ if (nr_lost_events)
+ printf(" (due to lost events?)");
+ printf("\n");
}
- header = perf_header__read(input);
- head = header->data_offset;
- sample_type = perf_header__sample_type(header);
+ if (nr_context_switch_bugs && nr_timestamps) {
+ printf(" INFO: %.3f%% context switch bugs (%ld out of %ld)",
+ (double)nr_context_switch_bugs/(double)nr_timestamps*100.0,
+ nr_context_switch_bugs, nr_timestamps);
+ if (nr_lost_events)
+ printf(" (due to lost events?)");
+ printf("\n");
+ }
+}
- if (!(sample_type & PERF_SAMPLE_RAW))
- die("No trace sample to read. Did you call perf record "
- "without -R?");
+static void __cmd_lat(void)
+{
+ struct rb_node *next;
- if (load_kernel() < 0) {
- perror("failed to load kernel symbols");
- return EXIT_FAILURE;
- }
+ setup_pager();
+ read_events();
+ sort_lat();
+
+ printf("\n -----------------------------------------------------------------------------------------\n");
+ printf(" Task | Runtime ms | Switches | Average delay ms | Maximum delay ms |\n");
+ printf(" -----------------------------------------------------------------------------------------\n");
+
+ next = rb_first(&sorted_atom_root);
+
+ while (next) {
+ struct work_atoms *work_list;
-remap:
- buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ,
- MAP_SHARED, input, offset);
- if (buf == MAP_FAILED) {
- perror("failed to mmap file");
- exit(-1);
+ work_list = rb_entry(next, struct work_atoms, node);
+ output_lat_thread(work_list);
+ next = rb_next(next);
}
-more:
- event = (event_t *)(buf + head);
+ printf(" -----------------------------------------------------------------------------------------\n");
+ printf(" TOTAL: |%11.3f ms |%9Ld |\n",
+ (double)all_runtime/1e6, all_count);
- size = event->header.size;
- if (!size)
- size = 8;
+ printf(" ---------------------------------------------------\n");
- if (head + event->header.size >= page_size * mmap_window) {
- unsigned long shift = page_size * (head / page_size);
- int res;
+ print_bad_events();
+ printf("\n");
- res = munmap(buf, page_size * mmap_window);
- assert(res == 0);
+}
- offset += shift;
- head -= shift;
- goto remap;
- }
+static struct trace_sched_handler map_ops = {
+ .wakeup_event = NULL,
+ .switch_event = map_switch_event,
+ .runtime_event = NULL,
+ .fork_event = NULL,
+};
- size = event->header.size;
+static void __cmd_map(void)
+{
+ max_cpu = sysconf(_SC_NPROCESSORS_CONF);
+ setup_pager();
+ read_events();
+ print_bad_events();
+}
- if (!size || process_event(event, offset, head) < 0) {
+static void __cmd_replay(void)
+{
+ unsigned long i;
- /*
- * assume we lost track of the stream, check alignment, and
- * increment a single u64 in the hope to catch on again 'soon'.
- */
+ calibrate_run_measurement_overhead();
+ calibrate_sleep_measurement_overhead();
- if (unlikely(head & 7))
- head &= ~7ULL;
+ test_calibrations();
- size = 8;
- }
+ read_events();
- head += size;
+ printf("nr_run_events: %ld\n", nr_run_events);
+ printf("nr_sleep_events: %ld\n", nr_sleep_events);
+ printf("nr_wakeup_events: %ld\n", nr_wakeup_events);
- if (offset + head < (unsigned long)perf_stat.st_size)
- goto more;
+ if (targetless_wakeups)
+ printf("target-less wakeups: %ld\n", targetless_wakeups);
+ if (multitarget_wakeups)
+ printf("multi-target wakeups: %ld\n", multitarget_wakeups);
+ if (nr_run_events_optimized)
+ printf("run atoms optimized: %ld\n",
+ nr_run_events_optimized);
- rc = EXIT_SUCCESS;
- close(input);
+ print_task_traces();
+ add_cross_task_wakeups();
- return rc;
+ create_tasks();
+ printf("------------------------------------------------------------\n");
+ for (i = 0; i < replay_repeat; i++)
+ run_one_test();
}
+
static const char * const sched_usage[] = {
- "perf sched [<options>] {record|latency|replay|trace}",
+ "perf sched [<options>] {record|latency|map|replay|trace}",
NULL
};
static const struct option sched_options[] = {
+ OPT_STRING('i', "input", &input_name, "file",
+ "input file name"),
OPT_BOOLEAN('v', "verbose", &verbose,
"be more verbose (show symbol address, etc)"),
OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
"sort by key(s): runtime, switch, avg, max"),
OPT_BOOLEAN('v', "verbose", &verbose,
"be more verbose (show symbol address, etc)"),
+ OPT_INTEGER('C', "CPU", &profile_cpu,
+ "CPU to profile on"),
OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
"dump raw trace in ASCII"),
OPT_END()
free(str);
- sort_dimension__add((char *)"pid", &cmp_pid);
+ sort_dimension__add("pid", &cmp_pid);
}
static const char *record_args[] = {
"record",
"-a",
"-R",
+ "-M",
+ "-f",
+ "-m", "1024",
"-c", "1",
"-e", "sched:sched_switch:r",
"-e", "sched:sched_stat_wait:r",
"-e", "sched:sched_stat_sleep:r",
"-e", "sched:sched_stat_iowait:r",
+ "-e", "sched:sched_stat_runtime:r",
"-e", "sched:sched_process_exit:r",
"-e", "sched:sched_process_fork:r",
"-e", "sched:sched_wakeup:r",
int cmd_sched(int argc, const char **argv, const char *prefix __used)
{
- symbol__init();
- page_size = getpagesize();
+ symbol__init(0);
argc = parse_options(argc, argv, sched_options, sched_usage,
PARSE_OPT_STOP_AT_NON_OPTION);
}
setup_sorting();
__cmd_lat();
+ } else if (!strcmp(argv[0], "map")) {
+ trace_handler = &map_ops;
+ setup_sorting();
+ __cmd_map();
} else if (!strncmp(argv[0], "rep", 3)) {
trace_handler = &replay_ops;
if (argc) {