1 #undef TRACE_SYSTEM 2 #define TRACE_SYSTEM sched 3 4 #if !defined(_TRACE_SCHED_H) || defined(TRACE_HEADER_MULTI_READ) 5 #define _TRACE_SCHED_H 6 7 #include <linux/sched.h> 8 #include <linux/tracepoint.h> 9 10 /* 11 * Tracepoint for calling kthread_stop, performed to end a kthread: 12 */ 13 TRACE_EVENT(sched_kthread_stop, 14 15 TP_PROTO(struct task_struct *t), 16 17 TP_ARGS(t), 18 19 TP_STRUCT__entry( 20 __array( char, comm, TASK_COMM_LEN ) 21 __field( pid_t, pid ) 22 ), 23 24 TP_fast_assign( 25 memcpy(__entry->comm, t->comm, TASK_COMM_LEN); 26 __entry->pid = t->pid; 27 ), 28 29 TP_printk("comm=%s pid=%d", __entry->comm, __entry->pid) 30 ); 31 32 /* 33 * Tracepoint for the return value of the kthread stopping: 34 */ 35 TRACE_EVENT(sched_kthread_stop_ret, 36 37 TP_PROTO(int ret), 38 39 TP_ARGS(ret), 40 41 TP_STRUCT__entry( 42 __field( int, ret ) 43 ), 44 45 TP_fast_assign( 46 __entry->ret = ret; 47 ), 48 49 TP_printk("ret=%d", __entry->ret) 50 ); 51 52 /* 53 * Tracepoint for waking up a task: 54 */ 55 DECLARE_EVENT_CLASS(sched_wakeup_template, 56 57 TP_PROTO(struct task_struct *p, int success), 58 59 TP_ARGS(p, success), 60 61 TP_STRUCT__entry( 62 __array( char, comm, TASK_COMM_LEN ) 63 __field( pid_t, pid ) 64 __field( int, prio ) 65 __field( int, success ) 66 __field( int, target_cpu ) 67 ), 68 69 TP_fast_assign( 70 memcpy(__entry->comm, p->comm, TASK_COMM_LEN); 71 __entry->pid = p->pid; 72 __entry->prio = p->prio; 73 __entry->success = success; 74 __entry->target_cpu = task_cpu(p); 75 ), 76 77 TP_printk("comm=%s pid=%d prio=%d success=%d target_cpu=%03d", 78 __entry->comm, __entry->pid, __entry->prio, 79 __entry->success, __entry->target_cpu) 80 ); 81 82 DEFINE_EVENT(sched_wakeup_template, sched_wakeup, 83 TP_PROTO(struct task_struct *p, int success), 84 TP_ARGS(p, success)); 85 86 /* 87 * Tracepoint for waking up a new task: 88 */ 89 DEFINE_EVENT(sched_wakeup_template, sched_wakeup_new, 90 TP_PROTO(struct task_struct *p, int success), 91 TP_ARGS(p, success)); 92 93 #ifdef CREATE_TRACE_POINTS 94 static inline long __trace_sched_switch_state(struct task_struct *p) 95 { 96 long state = p->state; 97 98 #ifdef CONFIG_PREEMPT 99 /* 100 * For all intents and purposes a preempted task is a running task. 101 */ 102 if (task_thread_info(p)->preempt_count & PREEMPT_ACTIVE) 103 state = TASK_RUNNING | TASK_STATE_MAX; 104 #endif 105 106 return state; 107 } 108 #endif 109 110 /* 111 * Tracepoint for task switches, performed by the scheduler: 112 */ 113 TRACE_EVENT(sched_switch, 114 115 TP_PROTO(struct task_struct *prev, 116 struct task_struct *next), 117 118 TP_ARGS(prev, next), 119 120 TP_STRUCT__entry( 121 __array( char, prev_comm, TASK_COMM_LEN ) 122 __field( pid_t, prev_pid ) 123 __field( int, prev_prio ) 124 __field( long, prev_state ) 125 __array( char, next_comm, TASK_COMM_LEN ) 126 __field( pid_t, next_pid ) 127 __field( int, next_prio ) 128 ), 129 130 TP_fast_assign( 131 memcpy(__entry->next_comm, next->comm, TASK_COMM_LEN); 132 __entry->prev_pid = prev->pid; 133 __entry->prev_prio = prev->prio; 134 __entry->prev_state = __trace_sched_switch_state(prev); 135 memcpy(__entry->prev_comm, prev->comm, TASK_COMM_LEN); 136 __entry->next_pid = next->pid; 137 __entry->next_prio = next->prio; 138 ), 139 140 TP_printk("prev_comm=%s prev_pid=%d prev_prio=%d prev_state=%s%s ==> next_comm=%s next_pid=%d next_prio=%d", 141 __entry->prev_comm, __entry->prev_pid, __entry->prev_prio, 142 __entry->prev_state & (TASK_STATE_MAX-1) ? 143 __print_flags(__entry->prev_state & (TASK_STATE_MAX-1), "|", 144 { 1, "S"} , { 2, "D" }, { 4, "T" }, { 8, "t" }, 145 { 16, "Z" }, { 32, "X" }, { 64, "x" }, 146 { 128, "W" }) : "R", 147 __entry->prev_state & TASK_STATE_MAX ? "+" : "", 148 __entry->next_comm, __entry->next_pid, __entry->next_prio) 149 ); 150 151 /* 152 * Tracepoint for a task being migrated: 153 */ 154 TRACE_EVENT(sched_migrate_task, 155 156 TP_PROTO(struct task_struct *p, int dest_cpu), 157 158 TP_ARGS(p, dest_cpu), 159 160 TP_STRUCT__entry( 161 __array( char, comm, TASK_COMM_LEN ) 162 __field( pid_t, pid ) 163 __field( int, prio ) 164 __field( int, orig_cpu ) 165 __field( int, dest_cpu ) 166 ), 167 168 TP_fast_assign( 169 memcpy(__entry->comm, p->comm, TASK_COMM_LEN); 170 __entry->pid = p->pid; 171 __entry->prio = p->prio; 172 __entry->orig_cpu = task_cpu(p); 173 __entry->dest_cpu = dest_cpu; 174 ), 175 176 TP_printk("comm=%s pid=%d prio=%d orig_cpu=%d dest_cpu=%d", 177 __entry->comm, __entry->pid, __entry->prio, 178 __entry->orig_cpu, __entry->dest_cpu) 179 ); 180 181 DECLARE_EVENT_CLASS(sched_process_template, 182 183 TP_PROTO(struct task_struct *p), 184 185 TP_ARGS(p), 186 187 TP_STRUCT__entry( 188 __array( char, comm, TASK_COMM_LEN ) 189 __field( pid_t, pid ) 190 __field( int, prio ) 191 ), 192 193 TP_fast_assign( 194 memcpy(__entry->comm, p->comm, TASK_COMM_LEN); 195 __entry->pid = p->pid; 196 __entry->prio = p->prio; 197 ), 198 199 TP_printk("comm=%s pid=%d prio=%d", 200 __entry->comm, __entry->pid, __entry->prio) 201 ); 202 203 /* 204 * Tracepoint for freeing a task: 205 */ 206 DEFINE_EVENT(sched_process_template, sched_process_free, 207 TP_PROTO(struct task_struct *p), 208 TP_ARGS(p)); 209 210 211 /* 212 * Tracepoint for a task exiting: 213 */ 214 DEFINE_EVENT(sched_process_template, sched_process_exit, 215 TP_PROTO(struct task_struct *p), 216 TP_ARGS(p)); 217 218 /* 219 * Tracepoint for waiting on task to unschedule: 220 */ 221 DEFINE_EVENT(sched_process_template, sched_wait_task, 222 TP_PROTO(struct task_struct *p), 223 TP_ARGS(p)); 224 225 /* 226 * Tracepoint for a waiting task: 227 */ 228 TRACE_EVENT(sched_process_wait, 229 230 TP_PROTO(struct pid *pid), 231 232 TP_ARGS(pid), 233 234 TP_STRUCT__entry( 235 __array( char, comm, TASK_COMM_LEN ) 236 __field( pid_t, pid ) 237 __field( int, prio ) 238 ), 239 240 TP_fast_assign( 241 memcpy(__entry->comm, current->comm, TASK_COMM_LEN); 242 __entry->pid = pid_nr(pid); 243 __entry->prio = current->prio; 244 ), 245 246 TP_printk("comm=%s pid=%d prio=%d", 247 __entry->comm, __entry->pid, __entry->prio) 248 ); 249 250 /* 251 * Tracepoint for do_fork: 252 */ 253 TRACE_EVENT(sched_process_fork, 254 255 TP_PROTO(struct task_struct *parent, struct task_struct *child), 256 257 TP_ARGS(parent, child), 258 259 TP_STRUCT__entry( 260 __array( char, parent_comm, TASK_COMM_LEN ) 261 __field( pid_t, parent_pid ) 262 __array( char, child_comm, TASK_COMM_LEN ) 263 __field( pid_t, child_pid ) 264 ), 265 266 TP_fast_assign( 267 memcpy(__entry->parent_comm, parent->comm, TASK_COMM_LEN); 268 __entry->parent_pid = parent->pid; 269 memcpy(__entry->child_comm, child->comm, TASK_COMM_LEN); 270 __entry->child_pid = child->pid; 271 ), 272 273 TP_printk("comm=%s pid=%d child_comm=%s child_pid=%d", 274 __entry->parent_comm, __entry->parent_pid, 275 __entry->child_comm, __entry->child_pid) 276 ); 277 278 /* 279 * XXX the below sched_stat tracepoints only apply to SCHED_OTHER/BATCH/IDLE 280 * adding sched_stat support to SCHED_FIFO/RR would be welcome. 281 */ 282 DECLARE_EVENT_CLASS(sched_stat_template, 283 284 TP_PROTO(struct task_struct *tsk, u64 delay), 285 286 TP_ARGS(tsk, delay), 287 288 TP_STRUCT__entry( 289 __array( char, comm, TASK_COMM_LEN ) 290 __field( pid_t, pid ) 291 __field( u64, delay ) 292 ), 293 294 TP_fast_assign( 295 memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN); 296 __entry->pid = tsk->pid; 297 __entry->delay = delay; 298 ) 299 TP_perf_assign( 300 __perf_count(delay); 301 ), 302 303 TP_printk("comm=%s pid=%d delay=%Lu [ns]", 304 __entry->comm, __entry->pid, 305 (unsigned long long)__entry->delay) 306 ); 307 308 309 /* 310 * Tracepoint for accounting wait time (time the task is runnable 311 * but not actually running due to scheduler contention). 312 */ 313 DEFINE_EVENT(sched_stat_template, sched_stat_wait, 314 TP_PROTO(struct task_struct *tsk, u64 delay), 315 TP_ARGS(tsk, delay)); 316 317 /* 318 * Tracepoint for accounting sleep time (time the task is not runnable, 319 * including iowait, see below). 320 */ 321 DEFINE_EVENT(sched_stat_template, sched_stat_sleep, 322 TP_PROTO(struct task_struct *tsk, u64 delay), 323 TP_ARGS(tsk, delay)); 324 325 /* 326 * Tracepoint for accounting iowait time (time the task is not runnable 327 * due to waiting on IO to complete). 328 */ 329 DEFINE_EVENT(sched_stat_template, sched_stat_iowait, 330 TP_PROTO(struct task_struct *tsk, u64 delay), 331 TP_ARGS(tsk, delay)); 332 333 /* 334 * Tracepoint for accounting blocked time (time the task is in uninterruptible). 335 */ 336 DEFINE_EVENT(sched_stat_template, sched_stat_blocked, 337 TP_PROTO(struct task_struct *tsk, u64 delay), 338 TP_ARGS(tsk, delay)); 339 340 /* 341 * Tracepoint for accounting runtime (time the task is executing 342 * on a CPU). 343 */ 344 TRACE_EVENT(sched_stat_runtime, 345 346 TP_PROTO(struct task_struct *tsk, u64 runtime, u64 vruntime), 347 348 TP_ARGS(tsk, runtime, vruntime), 349 350 TP_STRUCT__entry( 351 __array( char, comm, TASK_COMM_LEN ) 352 __field( pid_t, pid ) 353 __field( u64, runtime ) 354 __field( u64, vruntime ) 355 ), 356 357 TP_fast_assign( 358 memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN); 359 __entry->pid = tsk->pid; 360 __entry->runtime = runtime; 361 __entry->vruntime = vruntime; 362 ) 363 TP_perf_assign( 364 __perf_count(runtime); 365 ), 366 367 TP_printk("comm=%s pid=%d runtime=%Lu [ns] vruntime=%Lu [ns]", 368 __entry->comm, __entry->pid, 369 (unsigned long long)__entry->runtime, 370 (unsigned long long)__entry->vruntime) 371 ); 372 373 #ifdef CREATE_TRACE_POINTS 374 static inline u64 trace_get_sleeptime(struct task_struct *tsk) 375 { 376 #ifdef CONFIG_SCHEDSTATS 377 u64 block, sleep; 378 379 block = tsk->se.statistics.block_start; 380 sleep = tsk->se.statistics.sleep_start; 381 tsk->se.statistics.block_start = 0; 382 tsk->se.statistics.sleep_start = 0; 383 384 return block ? block : sleep ? sleep : 0; 385 #else 386 return 0; 387 #endif 388 } 389 #endif 390 391 /* 392 * Tracepoint for accounting sleeptime (time the task is sleeping 393 * or waiting for I/O). 394 */ 395 TRACE_EVENT(sched_stat_sleeptime, 396 397 TP_PROTO(struct task_struct *tsk, u64 now), 398 399 TP_ARGS(tsk, now), 400 401 TP_STRUCT__entry( 402 __array( char, comm, TASK_COMM_LEN ) 403 __field( pid_t, pid ) 404 __field( u64, sleeptime ) 405 ), 406 407 TP_fast_assign( 408 memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN); 409 __entry->pid = tsk->pid; 410 __entry->sleeptime = trace_get_sleeptime(tsk); 411 __entry->sleeptime = __entry->sleeptime ? 412 now - __entry->sleeptime : 0; 413 ) 414 TP_perf_assign( 415 __perf_count(__entry->sleeptime); 416 ), 417 418 TP_printk("comm=%s pid=%d sleeptime=%Lu [ns]", 419 __entry->comm, __entry->pid, 420 (unsigned long long)__entry->sleeptime) 421 ); 422 423 /* 424 * Tracepoint for showing priority inheritance modifying a tasks 425 * priority. 426 */ 427 TRACE_EVENT(sched_pi_setprio, 428 429 TP_PROTO(struct task_struct *tsk, int newprio), 430 431 TP_ARGS(tsk, newprio), 432 433 TP_STRUCT__entry( 434 __array( char, comm, TASK_COMM_LEN ) 435 __field( pid_t, pid ) 436 __field( int, oldprio ) 437 __field( int, newprio ) 438 ), 439 440 TP_fast_assign( 441 memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN); 442 __entry->pid = tsk->pid; 443 __entry->oldprio = tsk->prio; 444 __entry->newprio = newprio; 445 ), 446 447 TP_printk("comm=%s pid=%d oldprio=%d newprio=%d", 448 __entry->comm, __entry->pid, 449 __entry->oldprio, __entry->newprio) 450 ); 451 452 #endif /* _TRACE_SCHED_H */ 453 454 /* This part must be outside protection */ 455 #include <trace/define_trace.h> 456