diff options
Diffstat (limited to 'tools/perf/builtin-sched.c')
| -rw-r--r-- | tools/perf/builtin-sched.c | 1446 |
1 files changed, 673 insertions, 773 deletions
diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index a25a023965b..af305f57bd2 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -23,26 +23,12 @@ #include <pthread.h> #include <math.h> -static const char *input_name; - -static char default_sort_order[] = "avg, max, switch, runtime"; -static const char *sort_order = default_sort_order; - -static int profile_cpu = -1; - #define PR_SET_NAME 15 /* Set process name */ #define MAX_CPUS 4096 - -static u64 run_measurement_overhead; -static u64 sleep_measurement_overhead; - #define COMM_LEN 20 #define SYM_LEN 129 - #define MAX_PID 65536 -static unsigned long nr_tasks; - struct sched_atom; struct task_desc { @@ -80,44 +66,6 @@ struct sched_atom { struct task_desc *wakee; }; -static struct task_desc *pid_to_task[MAX_PID]; - -static struct task_desc **tasks; - -static pthread_mutex_t start_work_mutex = PTHREAD_MUTEX_INITIALIZER; -static u64 start_time; - -static pthread_mutex_t work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER; - -static unsigned long nr_run_events; -static unsigned long nr_sleep_events; -static unsigned long nr_wakeup_events; - -static unsigned long nr_sleep_corrections; -static unsigned long nr_run_events_optimized; - -static unsigned long targetless_wakeups; -static unsigned long multitarget_wakeups; - -static u64 cpu_usage; -static u64 runavg_cpu_usage; -static u64 parent_cpu_usage; -static u64 runavg_parent_cpu_usage; - -static unsigned long nr_runs; -static u64 sum_runtime; -static u64 sum_fluct; -static u64 run_avg; - -static unsigned int 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" enum thread_state { @@ -149,11 +97,79 @@ struct work_atoms { typedef int (*sort_fn_t)(struct work_atoms *, struct work_atoms *); -static struct rb_root atom_root, sorted_atom_root; +struct perf_sched; -static u64 all_runtime; -static u64 all_count; +struct trace_sched_handler { + int (*switch_event)(struct perf_sched *sched, struct perf_evsel *evsel, + struct perf_sample *sample, struct machine *machine); + + int (*runtime_event)(struct perf_sched *sched, struct perf_evsel *evsel, + struct perf_sample *sample, struct machine *machine); + + int (*wakeup_event)(struct perf_sched *sched, struct perf_evsel *evsel, + struct perf_sample *sample, struct machine *machine); + + int (*fork_event)(struct perf_sched *sched, struct perf_evsel *evsel, + struct perf_sample *sample); + int (*migrate_task_event)(struct perf_sched *sched, + struct perf_evsel *evsel, + struct perf_sample *sample, + struct machine *machine); +}; + +struct perf_sched { + struct perf_tool tool; + const char *input_name; + const char *sort_order; + unsigned long nr_tasks; + struct task_desc *pid_to_task[MAX_PID]; + struct task_desc **tasks; + const struct trace_sched_handler *tp_handler; + pthread_mutex_t start_work_mutex; + pthread_mutex_t work_done_wait_mutex; + int profile_cpu; +/* + * 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. + */ + int max_cpu; + u32 curr_pid[MAX_CPUS]; + struct thread *curr_thread[MAX_CPUS]; + char next_shortname1; + char next_shortname2; + unsigned int replay_repeat; + unsigned long nr_run_events; + unsigned long nr_sleep_events; + unsigned long nr_wakeup_events; + unsigned long nr_sleep_corrections; + unsigned long nr_run_events_optimized; + unsigned long targetless_wakeups; + unsigned long multitarget_wakeups; + unsigned long nr_runs; + unsigned long nr_timestamps; + unsigned long nr_unordered_timestamps; + unsigned long nr_state_machine_bugs; + unsigned long nr_context_switch_bugs; + unsigned long nr_events; + unsigned long nr_lost_chunks; + unsigned long nr_lost_events; + u64 run_measurement_overhead; + u64 sleep_measurement_overhead; + u64 start_time; + u64 cpu_usage; + u64 runavg_cpu_usage; + u64 parent_cpu_usage; + u64 runavg_parent_cpu_usage; + u64 sum_runtime; + u64 sum_fluct; + u64 run_avg; + u64 all_runtime; + u64 all_count; + u64 cpu_last_switched[MAX_CPUS]; + struct rb_root atom_root, sorted_atom_root; + struct list_head sort_list, cmp_pid; +}; static u64 get_nsecs(void) { @@ -164,13 +180,13 @@ static u64 get_nsecs(void) return ts.tv_sec * 1000000000ULL + ts.tv_nsec; } -static void burn_nsecs(u64 nsecs) +static void burn_nsecs(struct perf_sched *sched, u64 nsecs) { u64 T0 = get_nsecs(), T1; do { T1 = get_nsecs(); - } while (T1 + run_measurement_overhead < T0 + nsecs); + } while (T1 + sched->run_measurement_overhead < T0 + nsecs); } static void sleep_nsecs(u64 nsecs) @@ -183,24 +199,24 @@ static void sleep_nsecs(u64 nsecs) nanosleep(&ts, NULL); } -static void calibrate_run_measurement_overhead(void) +static void calibrate_run_measurement_overhead(struct perf_sched *sched) { u64 T0, T1, delta, min_delta = 1000000000ULL; int i; for (i = 0; i < 10; i++) { T0 = get_nsecs(); - burn_nsecs(0); + burn_nsecs(sched, 0); T1 = get_nsecs(); delta = T1-T0; min_delta = min(min_delta, delta); } - run_measurement_overhead = min_delta; + sched->run_measurement_overhead = min_delta; printf("run measurement overhead: %" PRIu64 " nsecs\n", min_delta); } -static void calibrate_sleep_measurement_overhead(void) +static void calibrate_sleep_measurement_overhead(struct perf_sched *sched) { u64 T0, T1, delta, min_delta = 1000000000ULL; int i; @@ -213,7 +229,7 @@ static void calibrate_sleep_measurement_overhead(void) min_delta = min(min_delta, delta); } min_delta -= 10000; - sleep_measurement_overhead = min_delta; + sched->sleep_measurement_overhead = min_delta; printf("sleep measurement overhead: %" PRIu64 " nsecs\n", min_delta); } @@ -246,8 +262,8 @@ static struct sched_atom *last_event(struct task_desc *task) return task->atoms[task->nr_events - 1]; } -static void -add_sched_event_run(struct task_desc *task, u64 timestamp, u64 duration) +static void add_sched_event_run(struct perf_sched *sched, struct task_desc *task, + u64 timestamp, u64 duration) { struct sched_atom *event, *curr_event = last_event(task); @@ -256,7 +272,7 @@ add_sched_event_run(struct task_desc *task, u64 timestamp, u64 duration) * to it: */ if (curr_event && curr_event->type == SCHED_EVENT_RUN) { - nr_run_events_optimized++; + sched->nr_run_events_optimized++; curr_event->duration += duration; return; } @@ -266,12 +282,11 @@ add_sched_event_run(struct task_desc *task, u64 timestamp, u64 duration) event->type = SCHED_EVENT_RUN; event->duration = duration; - nr_run_events++; + sched->nr_run_events++; } -static void -add_sched_event_wakeup(struct task_desc *task, u64 timestamp, - struct task_desc *wakee) +static void add_sched_event_wakeup(struct perf_sched *sched, struct task_desc *task, + u64 timestamp, struct task_desc *wakee) { struct sched_atom *event, *wakee_event; @@ -281,11 +296,11 @@ add_sched_event_wakeup(struct task_desc *task, u64 timestamp, wakee_event = last_event(wakee); if (!wakee_event || wakee_event->type != SCHED_EVENT_SLEEP) { - targetless_wakeups++; + sched->targetless_wakeups++; return; } if (wakee_event->wait_sem) { - multitarget_wakeups++; + sched->multitarget_wakeups++; return; } @@ -294,89 +309,89 @@ add_sched_event_wakeup(struct task_desc *task, u64 timestamp, wakee_event->specific_wait = 1; event->wait_sem = wakee_event->wait_sem; - nr_wakeup_events++; + sched->nr_wakeup_events++; } -static void -add_sched_event_sleep(struct task_desc *task, u64 timestamp, - u64 task_state __used) +static void add_sched_event_sleep(struct perf_sched *sched, struct task_desc *task, + u64 timestamp, u64 task_state __maybe_unused) { struct sched_atom *event = get_new_event(task, timestamp); event->type = SCHED_EVENT_SLEEP; - nr_sleep_events++; + sched->nr_sleep_events++; } -static struct task_desc *register_pid(unsigned long pid, const char *comm) +static struct task_desc *register_pid(struct perf_sched *sched, + unsigned long pid, const char *comm) { struct task_desc *task; BUG_ON(pid >= MAX_PID); - task = pid_to_task[pid]; + task = sched->pid_to_task[pid]; if (task) return task; task = zalloc(sizeof(*task)); task->pid = pid; - task->nr = nr_tasks; + task->nr = sched->nr_tasks; strcpy(task->comm, comm); /* * every task starts in sleeping state - this gets ignored * if there's no wakeup pointing to this sleep state: */ - add_sched_event_sleep(task, 0, 0); + add_sched_event_sleep(sched, task, 0, 0); - pid_to_task[pid] = task; - nr_tasks++; - tasks = realloc(tasks, nr_tasks*sizeof(struct task_task *)); - BUG_ON(!tasks); - tasks[task->nr] = task; + sched->pid_to_task[pid] = task; + sched->nr_tasks++; + sched->tasks = realloc(sched->tasks, sched->nr_tasks * sizeof(struct task_task *)); + BUG_ON(!sched->tasks); + sched->tasks[task->nr] = task; if (verbose) - printf("registered task #%ld, PID %ld (%s)\n", nr_tasks, pid, comm); + printf("registered task #%ld, PID %ld (%s)\n", sched->nr_tasks, pid, comm); return task; } -static void print_task_traces(void) +static void print_task_traces(struct perf_sched *sched) { struct task_desc *task; unsigned long i; - for (i = 0; i < nr_tasks; i++) { - task = tasks[i]; + for (i = 0; i < sched->nr_tasks; i++) { + task = sched->tasks[i]; printf("task %6ld (%20s:%10ld), nr_events: %ld\n", task->nr, task->comm, task->pid, task->nr_events); } } -static void add_cross_task_wakeups(void) +static void add_cross_task_wakeups(struct perf_sched *sched) { struct task_desc *task1, *task2; unsigned long i, j; - for (i = 0; i < nr_tasks; i++) { - task1 = tasks[i]; + for (i = 0; i < sched->nr_tasks; i++) { + task1 = sched->tasks[i]; j = i + 1; - if (j == nr_tasks) + if (j == sched->nr_tasks) j = 0; - task2 = tasks[j]; - add_sched_event_wakeup(task1, 0, task2); + task2 = sched->tasks[j]; + add_sched_event_wakeup(sched, task1, 0, task2); } } -static void -process_sched_event(struct task_desc *this_task __used, struct sched_atom *atom) +static void perf_sched__process_event(struct perf_sched *sched, + struct sched_atom *atom) { int ret = 0; switch (atom->type) { case SCHED_EVENT_RUN: - burn_nsecs(atom->duration); + burn_nsecs(sched, atom->duration); break; case SCHED_EVENT_SLEEP: if (atom->wait_sem) @@ -423,8 +438,8 @@ static int self_open_counters(void) fd = sys_perf_event_open(&attr, 0, -1, -1, 0); if (fd < 0) - die("Error: sys_perf_event_open() syscall returned" - "with %d (%s)\n", fd, strerror(errno)); + pr_debug("Error: sys_perf_event_open() syscall returned" + "with %d (%s)\n", fd, strerror(errno)); return fd; } @@ -439,31 +454,41 @@ static u64 get_cpu_usage_nsec_self(int fd) return runtime; } +struct sched_thread_parms { + struct task_desc *task; + struct perf_sched *sched; +}; + static void *thread_func(void *ctx) { - struct task_desc *this_task = ctx; + struct sched_thread_parms *parms = ctx; + struct task_desc *this_task = parms->task; + struct perf_sched *sched = parms->sched; u64 cpu_usage_0, cpu_usage_1; unsigned long i, ret; char comm2[22]; int fd; + free(parms); + sprintf(comm2, ":%s", this_task->comm); prctl(PR_SET_NAME, comm2); fd = self_open_counters(); - + if (fd < 0) + return NULL; again: ret = sem_post(&this_task->ready_for_work); BUG_ON(ret); - ret = pthread_mutex_lock(&start_work_mutex); + ret = pthread_mutex_lock(&sched->start_work_mutex); BUG_ON(ret); - ret = pthread_mutex_unlock(&start_work_mutex); + ret = pthread_mutex_unlock(&sched->start_work_mutex); BUG_ON(ret); cpu_usage_0 = get_cpu_usage_nsec_self(fd); for (i = 0; i < this_task->nr_events; i++) { this_task->curr_event = i; - process_sched_event(this_task, this_task->atoms[i]); + perf_sched__process_event(sched, this_task->atoms[i]); } cpu_usage_1 = get_cpu_usage_nsec_self(fd); @@ -471,15 +496,15 @@ again: ret = sem_post(&this_task->work_done_sem); BUG_ON(ret); - ret = pthread_mutex_lock(&work_done_wait_mutex); + ret = pthread_mutex_lock(&sched->work_done_wait_mutex); BUG_ON(ret); - ret = pthread_mutex_unlock(&work_done_wait_mutex); + ret = pthread_mutex_unlock(&sched->work_done_wait_mutex); BUG_ON(ret); goto again; } -static void create_tasks(void) +static void create_tasks(struct perf_sched *sched) { struct task_desc *task; pthread_attr_t attr; @@ -491,128 +516,129 @@ static void create_tasks(void) err = pthread_attr_setstacksize(&attr, (size_t) max(16 * 1024, PTHREAD_STACK_MIN)); BUG_ON(err); - err = pthread_mutex_lock(&start_work_mutex); + err = pthread_mutex_lock(&sched->start_work_mutex); BUG_ON(err); - err = pthread_mutex_lock(&work_done_wait_mutex); + err = pthread_mutex_lock(&sched->work_done_wait_mutex); BUG_ON(err); - for (i = 0; i < nr_tasks; i++) { - task = tasks[i]; + for (i = 0; i < sched->nr_tasks; i++) { + struct sched_thread_parms *parms = malloc(sizeof(*parms)); + BUG_ON(parms == NULL); + parms->task = task = sched->tasks[i]; + parms->sched = sched; sem_init(&task->sleep_sem, 0, 0); sem_init(&task->ready_for_work, 0, 0); sem_init(&task->work_done_sem, 0, 0); task->curr_event = 0; - err = pthread_create(&task->thread, &attr, thread_func, task); + err = pthread_create(&task->thread, &attr, thread_func, parms); BUG_ON(err); } } -static void wait_for_tasks(void) +static void wait_for_tasks(struct perf_sched *sched) { u64 cpu_usage_0, cpu_usage_1; struct task_desc *task; unsigned long i, ret; - start_time = get_nsecs(); - cpu_usage = 0; - pthread_mutex_unlock(&work_done_wait_mutex); + sched->start_time = get_nsecs(); + sched->cpu_usage = 0; + pthread_mutex_unlock(&sched->work_done_wait_mutex); - for (i = 0; i < nr_tasks; i++) { - task = tasks[i]; + for (i = 0; i < sched->nr_tasks; i++) { + task = sched->tasks[i]; ret = sem_wait(&task->ready_for_work); BUG_ON(ret); sem_init(&task->ready_for_work, 0, 0); } - ret = pthread_mutex_lock(&work_done_wait_mutex); + ret = pthread_mutex_lock(&sched->work_done_wait_mutex); BUG_ON(ret); cpu_usage_0 = get_cpu_usage_nsec_parent(); - pthread_mutex_unlock(&start_work_mutex); + pthread_mutex_unlock(&sched->start_work_mutex); - for (i = 0; i < nr_tasks; i++) { - task = tasks[i]; + for (i = 0; i < sched->nr_tasks; i++) { + task = sched->tasks[i]; ret = sem_wait(&task->work_done_sem); BUG_ON(ret); sem_init(&task->work_done_sem, 0, 0); - cpu_usage += task->cpu_usage; + sched->cpu_usage += task->cpu_usage; task->cpu_usage = 0; } cpu_usage_1 = get_cpu_usage_nsec_parent(); - if (!runavg_cpu_usage) - runavg_cpu_usage = cpu_usage; - runavg_cpu_usage = (runavg_cpu_usage*9 + cpu_usage)/10; + if (!sched->runavg_cpu_usage) + sched->runavg_cpu_usage = sched->cpu_usage; + sched->runavg_cpu_usage = (sched->runavg_cpu_usage * 9 + sched->cpu_usage) / 10; - parent_cpu_usage = cpu_usage_1 - cpu_usage_0; - if (!runavg_parent_cpu_usage) - runavg_parent_cpu_usage = parent_cpu_usage; - runavg_parent_cpu_usage = (runavg_parent_cpu_usage*9 + - parent_cpu_usage)/10; + sched->parent_cpu_usage = cpu_usage_1 - cpu_usage_0; + if (!sched->runavg_parent_cpu_usage) + sched->runavg_parent_cpu_usage = sched->parent_cpu_usage; + sched->runavg_parent_cpu_usage = (sched->runavg_parent_cpu_usage * 9 + + sched->parent_cpu_usage)/10; - ret = pthread_mutex_lock(&start_work_mutex); + ret = pthread_mutex_lock(&sched->start_work_mutex); BUG_ON(ret); - for (i = 0; i < nr_tasks; i++) { - task = tasks[i]; + for (i = 0; i < sched->nr_tasks; i++) { + task = sched->tasks[i]; sem_init(&task->sleep_sem, 0, 0); task->curr_event = 0; } } -static void run_one_test(void) +static void run_one_test(struct perf_sched *sched) { u64 T0, T1, delta, avg_delta, fluct; T0 = get_nsecs(); - wait_for_tasks(); + wait_for_tasks(sched); T1 = get_nsecs(); delta = T1 - T0; - sum_runtime += delta; - nr_runs++; + sched->sum_runtime += delta; + sched->nr_runs++; - avg_delta = sum_runtime / nr_runs; + avg_delta = sched->sum_runtime / sched->nr_runs; if (delta < avg_delta) fluct = avg_delta - delta; else fluct = delta - avg_delta; - sum_fluct += fluct; - if (!run_avg) - run_avg = delta; - run_avg = (run_avg*9 + delta)/10; + sched->sum_fluct += fluct; + if (!sched->run_avg) + sched->run_avg = delta; + sched->run_avg = (sched->run_avg * 9 + delta) / 10; - printf("#%-3ld: %0.3f, ", - nr_runs, (double)delta/1000000.0); + printf("#%-3ld: %0.3f, ", sched->nr_runs, (double)delta / 1000000.0); - printf("ravg: %0.2f, ", - (double)run_avg/1e6); + printf("ravg: %0.2f, ", (double)sched->run_avg / 1e6); printf("cpu: %0.2f / %0.2f", - (double)cpu_usage/1e6, (double)runavg_cpu_usage/1e6); + (double)sched->cpu_usage / 1e6, (double)sched->runavg_cpu_usage / 1e6); #if 0 /* * rusage statistics done by the parent, these are less - * accurate than the sum_exec_runtime based statistics: + * accurate than the sched->sum_exec_runtime based statistics: */ printf(" [%0.2f / %0.2f]", - (double)parent_cpu_usage/1e6, - (double)runavg_parent_cpu_usage/1e6); + (double)sched->parent_cpu_usage/1e6, + (double)sched->runavg_parent_cpu_usage/1e6); #endif printf("\n"); - if (nr_sleep_corrections) - printf(" (%ld sleep corrections)\n", nr_sleep_corrections); - nr_sleep_corrections = 0; + if (sched->nr_sleep_corrections) + printf(" (%ld sleep corrections)\n", sched->nr_sleep_corrections); + sched->nr_sleep_corrections = 0; } -static void test_calibrations(void) +static void test_calibrations(struct perf_sched *sched) { u64 T0, T1; T0 = get_nsecs(); - burn_nsecs(1e6); + burn_nsecs(sched, 1e6); T1 = get_nsecs(); printf("the run test took %" PRIu64 " nsecs\n", T1 - T0); @@ -624,216 +650,92 @@ static void test_calibrations(void) printf("the sleep test took %" PRIu64 " nsecs\n", T1 - T0); } -#define FILL_FIELD(ptr, field, event, data) \ - ptr.field = (typeof(ptr.field)) raw_field_value(event, #field, data) - -#define FILL_ARRAY(ptr, array, event, data) \ -do { \ - void *__array = raw_field_ptr(event, #array, data); \ - memcpy(ptr.array, __array, sizeof(ptr.array)); \ -} while(0) - -#define FILL_COMMON_FIELDS(ptr, event, data) \ -do { \ - FILL_FIELD(ptr, common_type, event, data); \ - FILL_FIELD(ptr, common_flags, event, data); \ - FILL_FIELD(ptr, common_preempt_count, event, data); \ - FILL_FIELD(ptr, common_pid, event, data); \ - FILL_FIELD(ptr, common_tgid, event, data); \ -} while (0) - - - -struct trace_switch_event { - u32 size; - - u16 common_type; - u8 common_flags; - u8 common_preempt_count; - u32 common_pid; - u32 common_tgid; - - char prev_comm[16]; - u32 prev_pid; - u32 prev_prio; - u64 prev_state; - char next_comm[16]; - u32 next_pid; - 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; - - u16 common_type; - u8 common_flags; - u8 common_preempt_count; - u32 common_pid; - u32 common_tgid; - - char comm[16]; - u32 pid; - - u32 prio; - u32 success; - u32 cpu; -}; - -struct trace_fork_event { - u32 size; - - u16 common_type; - u8 common_flags; - u8 common_preempt_count; - u32 common_pid; - u32 common_tgid; - - char parent_comm[16]; - u32 parent_pid; - char child_comm[16]; - 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 machine *, - struct event_format *, - struct perf_sample *sample); - - void (*runtime_event)(struct trace_runtime_event *, - struct machine *, - struct perf_sample *sample); - - void (*wakeup_event)(struct trace_wakeup_event *, - struct machine *, - struct event_format *, - struct perf_sample *sample); - - void (*fork_event)(struct trace_fork_event *, - struct event_format *event); - - void (*migrate_task_event)(struct trace_migrate_task_event *, - struct machine *machine, - struct perf_sample *sample); -}; - - -static void -replay_wakeup_event(struct trace_wakeup_event *wakeup_event, - struct machine *machine __used, - struct event_format *event, struct perf_sample *sample) +static int +replay_wakeup_event(struct perf_sched *sched, + struct perf_evsel *evsel, struct perf_sample *sample, + struct machine *machine __maybe_unused) { + const char *comm = perf_evsel__strval(evsel, sample, "comm"); + const u32 pid = perf_evsel__intval(evsel, sample, "pid"); struct task_desc *waker, *wakee; if (verbose) { - printf("sched_wakeup event %p\n", event); + printf("sched_wakeup event %p\n", evsel); - printf(" ... pid %d woke up %s/%d\n", - wakeup_event->common_pid, - wakeup_event->comm, - wakeup_event->pid); + printf(" ... pid %d woke up %s/%d\n", sample->tid, comm, pid); } - waker = register_pid(wakeup_event->common_pid, "<unknown>"); - wakee = register_pid(wakeup_event->pid, wakeup_event->comm); + waker = register_pid(sched, sample->tid, "<unknown>"); + wakee = register_pid(sched, pid, comm); - add_sched_event_wakeup(waker, sample->time, wakee); + add_sched_event_wakeup(sched, waker, sample->time, wakee); + return 0; } -static u64 cpu_last_switched[MAX_CPUS]; - -static void -replay_switch_event(struct trace_switch_event *switch_event, - struct machine *machine __used, - struct event_format *event, - struct perf_sample *sample) +static int replay_switch_event(struct perf_sched *sched, + struct perf_evsel *evsel, + struct perf_sample *sample, + struct machine *machine __maybe_unused) { - struct task_desc *prev, __used *next; + const char *prev_comm = perf_evsel__strval(evsel, sample, "prev_comm"), + *next_comm = perf_evsel__strval(evsel, sample, "next_comm"); + const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"), + next_pid = perf_evsel__intval(evsel, sample, "next_pid"); + const u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state"); + struct task_desc *prev, __maybe_unused *next; u64 timestamp0, timestamp = sample->time; int cpu = sample->cpu; s64 delta; if (verbose) - printf("sched_switch event %p\n", event); + printf("sched_switch event %p\n", evsel); if (cpu >= MAX_CPUS || cpu < 0) - return; + return 0; - timestamp0 = cpu_last_switched[cpu]; + timestamp0 = sched->cpu_last_switched[cpu]; if (timestamp0) delta = timestamp - timestamp0; else delta = 0; - if (delta < 0) - die("hm, delta: %" PRIu64 " < 0 ?\n", delta); - - if (verbose) { - printf(" ... switch from %s/%d to %s/%d [ran %" PRIu64 " nsecs]\n", - switch_event->prev_comm, switch_event->prev_pid, - switch_event->next_comm, switch_event->next_pid, - delta); + if (delta < 0) { + pr_debug("hm, delta: %" PRIu64 " < 0 ?\n", delta); + return -1; } - prev = register_pid(switch_event->prev_pid, switch_event->prev_comm); - next = register_pid(switch_event->next_pid, switch_event->next_comm); + pr_debug(" ... switch from %s/%d to %s/%d [ran %" PRIu64 " nsecs]\n", + prev_comm, prev_pid, next_comm, next_pid, delta); - cpu_last_switched[cpu] = timestamp; + prev = register_pid(sched, prev_pid, prev_comm); + next = register_pid(sched, next_pid, next_comm); - add_sched_event_run(prev, timestamp, delta); - add_sched_event_sleep(prev, timestamp, switch_event->prev_state); -} + sched->cpu_last_switched[cpu] = timestamp; + add_sched_event_run(sched, prev, timestamp, delta); + add_sched_event_sleep(sched, prev, timestamp, prev_state); -static void -replay_fork_event(struct trace_fork_event *fork_event, - struct event_format *event) + return 0; +} + +static int replay_fork_event(struct perf_sched *sched, struct perf_evsel *evsel, + struct perf_sample *sample) { + const char *parent_comm = perf_evsel__strval(evsel, sample, "parent_comm"), + *child_comm = perf_evsel__strval(evsel, sample, "child_comm"); + const u32 parent_pid = perf_evsel__intval(evsel, sample, "parent_pid"), + child_pid = perf_evsel__intval(evsel, sample, "child_pid"); + if (verbose) { - printf("sched_fork event %p\n", event); - printf("... parent: %s/%d\n", fork_event->parent_comm, fork_event->parent_pid); - printf("... child: %s/%d\n", fork_event->child_comm, fork_event->child_pid); + printf("sched_fork event %p\n", evsel); + printf("... parent: %s/%d\n", parent_comm, parent_pid); + printf("... child: %s/%d\n", child_comm, child_pid); } - register_pid(fork_event->parent_pid, fork_event->parent_comm); - register_pid(fork_event->child_pid, fork_event->child_comm); -} -static struct trace_sched_handler replay_ops = { - .wakeup_event = replay_wakeup_event, - .switch_event = replay_switch_event, - .fork_event = replay_fork_event, -}; + register_pid(sched, parent_pid, parent_comm); + register_pid(sched, child_pid, child_comm); + return 0; +} struct sort_dimension { const char *name; @@ -841,8 +743,6 @@ struct sort_dimension { struct list_head list; }; -static LIST_HEAD(cmp_pid); - static int thread_lat_cmp(struct list_head *list, struct work_atoms *l, struct work_atoms *r) { @@ -911,40 +811,45 @@ __thread_latency_insert(struct rb_root *root, struct work_atoms *data, rb_insert_color(&data->node, root); } -static void thread_atoms_insert(struct thread *thread) +static int thread_atoms_insert(struct perf_sched *sched, struct thread *thread) { struct work_atoms *atoms = zalloc(sizeof(*atoms)); - if (!atoms) - die("No memory"); + if (!atoms) { + pr_err("No memory at %s\n", __func__); + return -1; + } atoms->thread = thread; INIT_LIST_HEAD(&atoms->work_list); - __thread_latency_insert(&atom_root, atoms, &cmp_pid); + __thread_latency_insert(&sched->atom_root, atoms, &sched->cmp_pid); + return 0; } -static void -latency_fork_event(struct trace_fork_event *fork_event __used, - struct event_format *event __used) +static int latency_fork_event(struct perf_sched *sched __maybe_unused, + struct perf_evsel *evsel __maybe_unused, + struct perf_sample *sample __maybe_unused) { /* should insert the newcomer */ + return 0; } -__used -static char sched_out_state(struct trace_switch_event *switch_event) +static char sched_out_state(u64 prev_state) { const char *str = TASK_STATE_TO_CHAR_STR; - return str[switch_event->prev_state]; + return str[prev_state]; } -static void +static int add_sched_out_event(struct work_atoms *atoms, char run_state, u64 timestamp) { struct work_atom *atom = zalloc(sizeof(*atom)); - if (!atom) - die("Non memory"); + if (!atom) { + pr_err("Non memory at %s", __func__); + return -1; + } atom->sched_out_time = timestamp; @@ -954,10 +859,12 @@ add_sched_out_event(struct work_atoms *atoms, } list_add_tail(&atom->list, &atoms->work_list); + return 0; } static void -add_runtime_event(struct work_atoms *atoms, u64 delta, u64 timestamp __used) +add_runtime_event(struct work_atoms *atoms, u64 delta, + u64 timestamp __maybe_unused) { struct work_atom *atom; @@ -1000,12 +907,14 @@ add_sched_in_event(struct work_atoms *atoms, u64 timestamp) atoms->nb_atoms++; } -static void -latency_switch_event(struct trace_switch_event *switch_event, - struct machine *machine, - struct event_format *event __used, - struct perf_sample *sample) +static int latency_switch_event(struct perf_sched *sched, + struct perf_evsel *evsel, + struct perf_sample *sample, + struct machine *machine) { + const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"), + next_pid = perf_evsel__intval(evsel, sample, "next_pid"); + const u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state"); struct work_atoms *out_events, *in_events; struct thread *sched_out, *sched_in; u64 timestamp0, timestamp = sample->time; @@ -1014,87 +923,112 @@ latency_switch_event(struct trace_switch_event *switch_event, BUG_ON(cpu >= MAX_CPUS || cpu < 0); - timestamp0 = cpu_last_switched[cpu]; - cpu_last_switched[cpu] = timestamp; + timestamp0 = sched->cpu_last_switched[cpu]; + sched->cpu_last_switched[cpu] = timestamp; if (timestamp0) delta = timestamp - timestamp0; else delta = 0; - if (delta < 0) - die("hm, delta: %" PRIu64 " < 0 ?\n", delta); - + if (delta < 0) { + pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta); + return -1; + } - sched_out = machine__findnew_thread(machine, switch_event->prev_pid); - sched_in = machine__findnew_thread(machine, switch_event->next_pid); + sched_out = machine__findnew_thread(machine, prev_pid); + sched_in = machine__findnew_thread(machine, next_pid); - out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid); + out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->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"); + if (thread_atoms_insert(sched, sched_out)) + return -1; + out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid); + if (!out_events) { + pr_err("out-event: Internal tree error"); + return -1; + } } - add_sched_out_event(out_events, sched_out_state(switch_event), timestamp); + if (add_sched_out_event(out_events, sched_out_state(prev_state), timestamp)) + return -1; - in_events = thread_atoms_search(&atom_root, sched_in, &cmp_pid); + in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid); if (!in_events) { - thread_atoms_insert(sched_in); - in_events = thread_atoms_search(&atom_root, sched_in, &cmp_pid); - if (!in_events) - die("in-event: Internal tree error"); + if (thread_atoms_insert(sched, sched_in)) + return -1; + in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid); + if (!in_events) { + pr_err("in-event: Internal tree error"); + return -1; + } /* * 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); + if (add_sched_out_event(in_events, 'R', timestamp)) + return -1; } add_sched_in_event(in_events, timestamp); + + return 0; } -static void -latency_runtime_event(struct trace_runtime_event *runtime_event, - struct machine *machine, struct perf_sample *sample) +static int latency_runtime_event(struct perf_sched *sched, + struct perf_evsel *evsel, + struct perf_sample *sample, + struct machine *machine) { - struct thread *thread = machine__findnew_thread(machine, runtime_event->pid); - struct work_atoms *atoms = thread_atoms_search(&atom_root, thread, &cmp_pid); + const u32 pid = perf_evsel__intval(evsel, sample, "pid"); + const u64 runtime = perf_evsel__intval(evsel, sample, "runtime"); + struct thread *thread = machine__findnew_thread(machine, pid); + struct work_atoms *atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid); u64 timestamp = sample->time; int cpu = sample->cpu; 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); + if (thread_atoms_insert(sched, thread)) + return -1; + atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid); + if (!atoms) { + pr_debug("in-event: Internal tree error"); + return -1; + } + if (add_sched_out_event(atoms, 'R', timestamp)) + return -1; } - add_runtime_event(atoms, runtime_event->runtime, timestamp); + add_runtime_event(atoms, runtime, timestamp); + return 0; } -static void -latency_wakeup_event(struct trace_wakeup_event *wakeup_event, - struct machine *machine, struct event_format *event __used, - struct perf_sample *sample) +static int latency_wakeup_event(struct perf_sched *sched, + struct perf_evsel *evsel, + struct perf_sample *sample, + struct machine *machine) { + const u32 pid = perf_evsel__intval(evsel, sample, "pid"), + success = perf_evsel__intval(evsel, sample, "success"); struct work_atoms *atoms; struct work_atom *atom; struct thread *wakee; u64 timestamp = sample->time; /* Note for later, it may be interesting to observe the failing cases */ - if (!wakeup_event->success) - return; + if (!success) + return 0; - wakee = machine__findnew_thread(machine, wakeup_event->pid); - atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid); + wakee = machine__findnew_thread(machine, pid); + atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid); if (!atoms) { - thread_atoms_insert(wakee); - 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 (thread_atoms_insert(sched, wakee)) + return -1; + atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid); + if (!atoms) { + pr_debug("wakeup-event: Internal tree er |