2 #define TRACE_SYSTEM sched
4 #if !defined(_TRACE_SCHED_H) || defined(TRACE_HEADER_MULTI_READ)
7 #include <linux/sched.h>
8 #include <linux/tracepoint.h>
9 #include <linux/binfmts.h>
10 #include <linux/version.h>
11 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0))
12 #include <linux/sched/rt.h>
15 #ifndef _TRACE_SCHED_DEF_
16 #define _TRACE_SCHED_DEF_
18 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35))
20 static inline long __trace_sched_switch_state(struct task_struct *p)
22 long state = p->state;
26 * For all intents and purposes a preempted task is a running task.
28 if (task_thread_info(p)->preempt_count & PREEMPT_ACTIVE)
29 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0))
30 state = TASK_RUNNING | TASK_STATE_MAX;
41 #endif /* _TRACE_SCHED_DEF_ */
44 * Tracepoint for calling kthread_stop, performed to end a kthread:
46 TRACE_EVENT(sched_kthread_stop,
48 TP_PROTO(struct task_struct *t),
53 __array_text( char, comm, TASK_COMM_LEN )
58 tp_memcpy(comm, t->comm, TASK_COMM_LEN)
59 tp_assign(tid, t->pid)
62 TP_printk("comm=%s tid=%d", __entry->comm, __entry->tid)
66 * Tracepoint for the return value of the kthread stopping:
68 TRACE_EVENT(sched_kthread_stop_ret,
82 TP_printk("ret=%d", __entry->ret)
86 * Tracepoint for waking up a task:
88 DECLARE_EVENT_CLASS(sched_wakeup_template,
90 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35))
91 TP_PROTO(struct task_struct *p, int success),
95 TP_PROTO(struct rq *rq, struct task_struct *p, int success),
97 TP_ARGS(rq, p, success),
101 __array_text( char, comm, TASK_COMM_LEN )
102 __field( pid_t, tid )
104 __field( int, success )
105 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32))
106 __field( int, target_cpu )
111 tp_memcpy(comm, p->comm, TASK_COMM_LEN)
112 tp_assign(tid, p->pid)
113 tp_assign(prio, p->prio)
114 tp_assign(success, success)
115 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32))
116 tp_assign(target_cpu, task_cpu(p))
118 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,6,0))
125 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32))
126 TP_printk("comm=%s tid=%d prio=%d success=%d target_cpu=%03d",
127 __entry->comm, __entry->tid, __entry->prio,
128 __entry->success, __entry->target_cpu)
130 TP_printk("comm=%s tid=%d prio=%d success=%d",
131 __entry->comm, __entry->tid, __entry->prio,
136 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35))
138 DEFINE_EVENT(sched_wakeup_template, sched_wakeup,
139 TP_PROTO(struct task_struct *p, int success),
143 * Tracepoint for waking up a new task:
145 DEFINE_EVENT(sched_wakeup_template, sched_wakeup_new,
146 TP_PROTO(struct task_struct *p, int success),
149 #else /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)) */
151 DEFINE_EVENT(sched_wakeup_template, sched_wakeup,
152 TP_PROTO(struct rq *rq, struct task_struct *p, int success),
153 TP_ARGS(rq, p, success))
156 * Tracepoint for waking up a new task:
158 DEFINE_EVENT(sched_wakeup_template, sched_wakeup_new,
159 TP_PROTO(struct rq *rq, struct task_struct *p, int success),
160 TP_ARGS(rq, p, success))
162 #endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)) */
165 * Tracepoint for task switches, performed by the scheduler:
167 TRACE_EVENT(sched_switch,
169 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35))
170 TP_PROTO(struct task_struct *prev,
171 struct task_struct *next),
174 #else /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)) */
175 TP_PROTO(struct rq *rq, struct task_struct *prev,
176 struct task_struct *next),
178 TP_ARGS(rq, prev, next),
179 #endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)) */
182 __array_text( char, prev_comm, TASK_COMM_LEN )
183 __field( pid_t, prev_tid )
184 __field( int, prev_prio )
185 __field( long, prev_state )
186 __array_text( char, next_comm, TASK_COMM_LEN )
187 __field( pid_t, next_tid )
188 __field( int, next_prio )
192 tp_memcpy(next_comm, next->comm, TASK_COMM_LEN)
193 tp_assign(prev_tid, prev->pid)
194 tp_assign(prev_prio, prev->prio - MAX_RT_PRIO)
195 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35))
196 tp_assign(prev_state, __trace_sched_switch_state(prev))
198 tp_assign(prev_state, prev->state)
200 tp_memcpy(prev_comm, prev->comm, TASK_COMM_LEN)
201 tp_assign(next_tid, next->pid)
202 tp_assign(next_prio, next->prio - MAX_RT_PRIO)
205 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0))
206 TP_printk("prev_comm=%s prev_tid=%d prev_prio=%d prev_state=%s%s ==> next_comm=%s next_tid=%d next_prio=%d",
207 __entry->prev_comm, __entry->prev_tid, __entry->prev_prio,
208 __entry->prev_state & (TASK_STATE_MAX-1) ?
209 __print_flags(__entry->prev_state & (TASK_STATE_MAX-1), "|",
210 { 1, "S"} , { 2, "D" }, { 4, "T" }, { 8, "t" },
211 { 16, "Z" }, { 32, "X" }, { 64, "x" },
213 __entry->prev_state & TASK_STATE_MAX ? "+" : "",
214 __entry->next_comm, __entry->next_tid, __entry->next_prio)
216 TP_printk("prev_comm=%s prev_tid=%d prev_prio=%d prev_state=%s ==> next_comm=%s next_tid=%d next_prio=%d",
217 __entry->prev_comm, __entry->prev_tid, __entry->prev_prio,
218 __entry->prev_state ?
219 __print_flags(__entry->prev_state, "|",
220 { 1, "S"} , { 2, "D" }, { 4, "T" }, { 8, "t" },
221 { 16, "Z" }, { 32, "X" }, { 64, "x" },
223 __entry->next_comm, __entry->next_tid, __entry->next_prio)
228 * Tracepoint for a task being migrated:
230 TRACE_EVENT(sched_migrate_task,
232 TP_PROTO(struct task_struct *p, int dest_cpu),
234 TP_ARGS(p, dest_cpu),
237 __array_text( char, comm, TASK_COMM_LEN )
238 __field( pid_t, tid )
240 __field( int, orig_cpu )
241 __field( int, dest_cpu )
245 tp_memcpy(comm, p->comm, TASK_COMM_LEN)
246 tp_assign(tid, p->pid)
247 tp_assign(prio, p->prio - MAX_RT_PRIO)
248 tp_assign(orig_cpu, task_cpu(p))
249 tp_assign(dest_cpu, dest_cpu)
252 TP_printk("comm=%s tid=%d prio=%d orig_cpu=%d dest_cpu=%d",
253 __entry->comm, __entry->tid, __entry->prio,
254 __entry->orig_cpu, __entry->dest_cpu)
257 DECLARE_EVENT_CLASS(sched_process_template,
259 TP_PROTO(struct task_struct *p),
264 __array_text( char, comm, TASK_COMM_LEN )
265 __field( pid_t, tid )
270 tp_memcpy(comm, p->comm, TASK_COMM_LEN)
271 tp_assign(tid, p->pid)
272 tp_assign(prio, p->prio - MAX_RT_PRIO)
275 TP_printk("comm=%s tid=%d prio=%d",
276 __entry->comm, __entry->tid, __entry->prio)
280 * Tracepoint for freeing a task:
282 DEFINE_EVENT(sched_process_template, sched_process_free,
283 TP_PROTO(struct task_struct *p),
288 * Tracepoint for a task exiting:
290 DEFINE_EVENT(sched_process_template, sched_process_exit,
291 TP_PROTO(struct task_struct *p),
295 * Tracepoint for waiting on task to unschedule:
297 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35))
298 DEFINE_EVENT(sched_process_template, sched_wait_task,
299 TP_PROTO(struct task_struct *p),
301 #else /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)) */
302 DEFINE_EVENT(sched_process_template, sched_wait_task,
303 TP_PROTO(struct rq *rq, struct task_struct *p),
305 #endif /* #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)) */
308 * Tracepoint for a waiting task:
310 TRACE_EVENT(sched_process_wait,
312 TP_PROTO(struct pid *pid),
317 __array_text( char, comm, TASK_COMM_LEN )
318 __field( pid_t, tid )
323 tp_memcpy(comm, current->comm, TASK_COMM_LEN)
324 tp_assign(tid, pid_nr(pid))
325 tp_assign(prio, current->prio - MAX_RT_PRIO)
328 TP_printk("comm=%s tid=%d prio=%d",
329 __entry->comm, __entry->tid, __entry->prio)
333 * Tracepoint for do_fork.
334 * Saving both TID and PID information, especially for the child, allows
335 * trace analyzers to distinguish between creation of a new process and
336 * creation of a new thread. Newly created processes will have child_tid
337 * == child_pid, while creation of a thread yields to child_tid !=
340 TRACE_EVENT(sched_process_fork,
342 TP_PROTO(struct task_struct *parent, struct task_struct *child),
344 TP_ARGS(parent, child),
347 __array_text( char, parent_comm, TASK_COMM_LEN )
348 __field( pid_t, parent_tid )
349 __field( pid_t, parent_pid )
350 __array_text( char, child_comm, TASK_COMM_LEN )
351 __field( pid_t, child_tid )
352 __field( pid_t, child_pid )
356 tp_memcpy(parent_comm, parent->comm, TASK_COMM_LEN)
357 tp_assign(parent_tid, parent->pid)
358 tp_assign(parent_pid, parent->tgid)
359 tp_memcpy(child_comm, child->comm, TASK_COMM_LEN)
360 tp_assign(child_tid, child->pid)
361 tp_assign(child_pid, child->tgid)
364 TP_printk("comm=%s tid=%d child_comm=%s child_tid=%d",
365 __entry->parent_comm, __entry->parent_tid,
366 __entry->child_comm, __entry->child_tid)
369 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,33))
371 * Tracepoint for sending a signal:
373 TRACE_EVENT(sched_signal_send,
375 TP_PROTO(int sig, struct task_struct *p),
381 __array( char, comm, TASK_COMM_LEN )
382 __field( pid_t, pid )
386 tp_memcpy(comm, p->comm, TASK_COMM_LEN)
387 tp_assign(pid, p->pid)
391 TP_printk("sig=%d comm=%s pid=%d",
392 __entry->sig, __entry->comm, __entry->pid)
396 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,4,0))
398 * Tracepoint for exec:
400 TRACE_EVENT(sched_process_exec,
402 TP_PROTO(struct task_struct *p, pid_t old_pid,
403 struct linux_binprm *bprm),
405 TP_ARGS(p, old_pid, bprm),
408 __string( filename, bprm->filename )
409 __field( pid_t, tid )
410 __field( pid_t, old_tid )
414 tp_strcpy(filename, bprm->filename)
415 tp_assign(tid, p->pid)
416 tp_assign(old_tid, old_pid)
419 TP_printk("filename=%s tid=%d old_tid=%d", __get_str(filename),
420 __entry->tid, __entry->old_tid)
424 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32))
426 * XXX the below sched_stat tracepoints only apply to SCHED_OTHER/BATCH/IDLE
427 * adding sched_stat support to SCHED_FIFO/RR would be welcome.
429 DECLARE_EVENT_CLASS(sched_stat_template,
431 TP_PROTO(struct task_struct *tsk, u64 delay),
436 __array_text( char, comm, TASK_COMM_LEN )
437 __field( pid_t, tid )
438 __field( u64, delay )
442 tp_memcpy(comm, tsk->comm, TASK_COMM_LEN)
443 tp_assign(tid, tsk->pid)
444 tp_assign(delay, delay)
450 TP_printk("comm=%s tid=%d delay=%Lu [ns]",
451 __entry->comm, __entry->tid,
452 (unsigned long long)__entry->delay)
457 * Tracepoint for accounting wait time (time the task is runnable
458 * but not actually running due to scheduler contention).
460 DEFINE_EVENT(sched_stat_template, sched_stat_wait,
461 TP_PROTO(struct task_struct *tsk, u64 delay),
465 * Tracepoint for accounting sleep time (time the task is not runnable,
466 * including iowait, see below).
468 DEFINE_EVENT(sched_stat_template, sched_stat_sleep,
469 TP_PROTO(struct task_struct *tsk, u64 delay),
473 * Tracepoint for accounting iowait time (time the task is not runnable
474 * due to waiting on IO to complete).
476 DEFINE_EVENT(sched_stat_template, sched_stat_iowait,
477 TP_PROTO(struct task_struct *tsk, u64 delay),
480 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,3,0))
482 * Tracepoint for accounting blocked time (time the task is in uninterruptible).
484 DEFINE_EVENT(sched_stat_template, sched_stat_blocked,
485 TP_PROTO(struct task_struct *tsk, u64 delay),
490 * Tracepoint for accounting runtime (time the task is executing
493 TRACE_EVENT(sched_stat_runtime,
495 TP_PROTO(struct task_struct *tsk, u64 runtime, u64 vruntime),
497 TP_ARGS(tsk, runtime, vruntime),
500 __array_text( char, comm, TASK_COMM_LEN )
501 __field( pid_t, tid )
502 __field( u64, runtime )
503 __field( u64, vruntime )
507 tp_memcpy(comm, tsk->comm, TASK_COMM_LEN)
508 tp_assign(tid, tsk->pid)
509 tp_assign(runtime, runtime)
510 tp_assign(vruntime, vruntime)
513 __perf_count(runtime)
514 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,6,0))
519 TP_printk("comm=%s tid=%d runtime=%Lu [ns] vruntime=%Lu [ns]",
520 __entry->comm, __entry->tid,
521 (unsigned long long)__entry->runtime,
522 (unsigned long long)__entry->vruntime)
526 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37))
528 * Tracepoint for showing priority inheritance modifying a tasks
531 TRACE_EVENT(sched_pi_setprio,
533 TP_PROTO(struct task_struct *tsk, int newprio),
535 TP_ARGS(tsk, newprio),
538 __array_text( char, comm, TASK_COMM_LEN )
539 __field( pid_t, tid )
540 __field( int, oldprio )
541 __field( int, newprio )
545 tp_memcpy(comm, tsk->comm, TASK_COMM_LEN)
546 tp_assign(tid, tsk->pid)
547 tp_assign(oldprio, tsk->prio - MAX_RT_PRIO)
548 tp_assign(newprio, newprio - MAX_RT_PRIO)
551 TP_printk("comm=%s tid=%d oldprio=%d newprio=%d",
552 __entry->comm, __entry->tid,
553 __entry->oldprio, __entry->newprio)
557 #endif /* _TRACE_SCHED_H */
559 /* This part must be outside protection */
560 #include "../../../probes/define_trace.h"