1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4
5 /*
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
8 */
9
10 #include <uapi/linux/sched.h>
11
12 #include <asm/current.h>
13
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/irqflags.h>
22 #include <linux/seccomp.h>
23 #include <linux/nodemask.h>
24 #include <linux/rcupdate.h>
25 #include <linux/refcount.h>
26 #include <linux/resource.h>
27 #include <linux/latencytop.h>
28 #include <linux/sched/prio.h>
29 #include <linux/sched/types.h>
30 #include <linux/signal_types.h>
31 #include <linux/mm_types_task.h>
32 #include <linux/task_io_accounting.h>
33 #include <linux/posix-timers.h>
34 #include <linux/rseq.h>
35 #include <linux/seqlock.h>
36 #include <linux/kcsan.h>
37 #include <linux/sched/rtg.h>
38
39 /* task_struct member predeclarations (sorted alphabetically): */
40 struct audit_context;
41 struct backing_dev_info;
42 struct bio_list;
43 struct blk_plug;
44 struct bpf_run_ctx;
45 struct capture_control;
46 struct cfs_rq;
47 struct fs_struct;
48 struct futex_pi_state;
49 struct io_context;
50 struct mempolicy;
51 struct nameidata;
52 struct nsproxy;
53 struct perf_event_context;
54 struct pid_namespace;
55 struct pipe_inode_info;
56 struct rcu_node;
57 #ifdef CONFIG_RECLAIM_ACCT
58 struct reclaim_acct;
59 #endif
60 struct reclaim_state;
61 struct robust_list_head;
62 struct root_domain;
63 struct rq;
64 struct sched_attr;
65 struct sched_param;
66 struct seq_file;
67 struct sighand_struct;
68 struct signal_struct;
69 struct task_delay_info;
70 struct task_group;
71 struct io_uring_task;
72
73 /*
74 * Task state bitmask. NOTE! These bits are also
75 * encoded in fs/proc/array.c: get_task_state().
76 *
77 * We have two separate sets of flags: task->state
78 * is about runnability, while task->exit_state are
79 * about the task exiting. Confusing, but this way
80 * modifying one set can't modify the other one by
81 * mistake.
82 */
83
84 /* Used in tsk->state: */
85 #define TASK_RUNNING 0x0000
86 #define TASK_INTERRUPTIBLE 0x0001
87 #define TASK_UNINTERRUPTIBLE 0x0002
88 #define __TASK_STOPPED 0x0004
89 #define __TASK_TRACED 0x0008
90 /* Used in tsk->exit_state: */
91 #define EXIT_DEAD 0x0010
92 #define EXIT_ZOMBIE 0x0020
93 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
94 /* Used in tsk->state again: */
95 #define TASK_PARKED 0x0040
96 #define TASK_DEAD 0x0080
97 #define TASK_WAKEKILL 0x0100
98 #define TASK_WAKING 0x0200
99 #define TASK_NOLOAD 0x0400
100 #define TASK_NEW 0x0800
101 #define TASK_STATE_MAX 0x1000
102
103 /* Convenience macros for the sake of set_current_state: */
104 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
105 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
106 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
107
108 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
109
110 /* Convenience macros for the sake of wake_up(): */
111 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
112
113 /* get_task_state(): */
114 #define TASK_REPORT \
115 (TASK_RUNNING | TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE | __TASK_STOPPED | __TASK_TRACED | EXIT_DEAD | \
116 EXIT_ZOMBIE | TASK_PARKED)
117
118 #define task_is_traced(task) (((task)->state & __TASK_TRACED) != 0)
119
120 #define task_is_stopped(task) (((task)->state & __TASK_STOPPED) != 0)
121
122 #define task_is_stopped_or_traced(task) (((task)->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
123
124 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
125
126 /*
127 * Special states are those that do not use the normal wait-loop pattern. See
128 * the comment with set_special_state().
129 */
130 #define is_special_task_state(state) ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
131
132 #define __set_current_state(state_value) \
133 do { \
134 WARN_ON_ONCE(is_special_task_state(state_value)); \
135 current->task_state_change = _THIS_IP_; \
136 current->state = (state_value); \
137 } while (0)
138
139 #define set_current_state(state_value) \
140 do { \
141 WARN_ON_ONCE(is_special_task_state(state_value)); \
142 current->task_state_change = _THIS_IP_; \
143 smp_store_mb(current->state, (state_value)); \
144 } while (0)
145
146 #define set_special_state(state_value) \
147 do { \
148 unsigned long flags; /* may shadow */ \
149 WARN_ON_ONCE(!is_special_task_state(state_value)); \
150 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
151 current->task_state_change = _THIS_IP_; \
152 current->state = (state_value); \
153 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
154 } while (0)
155 #else
156 /*
157 * set_current_state() includes a barrier so that the write of current->state
158 * is correctly serialised wrt the caller's subsequent test of whether to
159 * actually sleep:
160 *
161 * for (;;) {
162 * set_current_state(TASK_UNINTERRUPTIBLE);
163 * if (CONDITION)
164 * break;
165 *
166 * schedule();
167 * }
168 * __set_current_state(TASK_RUNNING);
169 *
170 * If the caller does not need such serialisation (because, for instance, the
171 * CONDITION test and condition change and wakeup are under the same lock) then
172 * use __set_current_state().
173 *
174 * The above is typically ordered against the wakeup, which does:
175 *
176 * CONDITION = 1;
177 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
178 *
179 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
180 * accessing p->state.
181 *
182 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
183 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
184 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
185 *
186 * However, with slightly different timing the wakeup TASK_RUNNING store can
187 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
188 * a problem either because that will result in one extra go around the loop
189 * and our @cond test will save the day.
190 *
191 * Also see the comments of try_to_wake_up().
192 */
193 #define __set_current_state(state_value) current->state = (state_value)
194
195 #define set_current_state(state_value) smp_store_mb(current->state, (state_value))
196
197 /*
198 * set_special_state() should be used for those states when the blocking task
199 * can not use the regular condition based wait-loop. In that case we must
200 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
201 * will not collide with our state change.
202 */
203 #define set_special_state(state_value) \
204 do { \
205 unsigned long flags; /* may shadow */ \
206 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
207 current->state = (state_value); \
208 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
209 } while (0)
210
211 #endif
212
213 /* Task command name length: */
214 #define TASK_COMM_LEN 16
215
216 enum task_event {
217 PUT_PREV_TASK = 0,
218 PICK_NEXT_TASK = 1,
219 TASK_WAKE = 2,
220 TASK_MIGRATE = 3,
221 TASK_UPDATE = 4,
222 IRQ_UPDATE = 5,
223 };
224
225 /* Note: this need to be in sync with migrate_type_names array */
226 enum migrate_types {
227 GROUP_TO_RQ,
228 RQ_TO_GROUP,
229 };
230
231 #ifdef CONFIG_CPU_ISOLATION_OPT
232 extern int sched_isolate_count(const cpumask_t *mask, bool include_offline);
233 extern int sched_isolate_cpu(int cpu);
234 extern int sched_unisolate_cpu(int cpu);
235 extern int sched_unisolate_cpu_unlocked(int cpu);
236 #else
sched_isolate_count(const cpumask_t * mask,bool include_offline)237 static inline int sched_isolate_count(const cpumask_t *mask, bool include_offline)
238 {
239 cpumask_t count_mask;
240
241 if (include_offline) {
242 cpumask_andnot(&count_mask, mask, cpu_online_mask);
243 } else {
244 return 0;
245 }
246
247 return cpumask_weight(&count_mask);
248 }
249
sched_isolate_cpu(int cpu)250 static inline int sched_isolate_cpu(int cpu)
251 {
252 return 0;
253 }
254
sched_unisolate_cpu(int cpu)255 static inline int sched_unisolate_cpu(int cpu)
256 {
257 return 0;
258 }
259
sched_unisolate_cpu_unlocked(int cpu)260 static inline int sched_unisolate_cpu_unlocked(int cpu)
261 {
262 return 0;
263 }
264 #endif
265
266 extern void scheduler_tick(void);
267
268 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
269
270 extern long schedule_timeout(long timeout);
271 extern long schedule_timeout_interruptible(long timeout);
272 extern long schedule_timeout_killable(long timeout);
273 extern long schedule_timeout_uninterruptible(long timeout);
274 extern long schedule_timeout_idle(long timeout);
275 asmlinkage void schedule(void);
276 extern void schedule_preempt_disabled(void);
277 asmlinkage void preempt_schedule_irq(void);
278
279 extern int __must_check io_schedule_prepare(void);
280 extern void io_schedule_finish(int token);
281 extern long io_schedule_timeout(long timeout);
282 extern void io_schedule(void);
283
284 /**
285 * struct prev_cputime - snapshot of system and user cputime
286 * @utime: time spent in user mode
287 * @stime: time spent in system mode
288 * @lock: protects the above two fields
289 *
290 * Stores previous user/system time values such that we can guarantee
291 * monotonicity.
292 */
293 struct prev_cputime {
294 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
295 u64 utime;
296 u64 stime;
297 raw_spinlock_t lock;
298 #endif
299 };
300
301 enum vtime_state {
302 /* Task is sleeping or running in a CPU with VTIME inactive: */
303 VTIME_INACTIVE = 0,
304 /* Task is idle */
305 VTIME_IDLE,
306 /* Task runs in kernelspace in a CPU with VTIME active: */
307 VTIME_SYS,
308 /* Task runs in userspace in a CPU with VTIME active: */
309 VTIME_USER,
310 /* Task runs as guests in a CPU with VTIME active: */
311 VTIME_GUEST,
312 };
313
314 struct vtime {
315 seqcount_t seqcount;
316 unsigned long long starttime;
317 enum vtime_state state;
318 unsigned int cpu;
319 u64 utime;
320 u64 stime;
321 u64 gtime;
322 };
323
324 /*
325 * Utilization clamp constraints.
326 * @UCLAMP_MIN: Minimum utilization
327 * @UCLAMP_MAX: Maximum utilization
328 * @UCLAMP_CNT: Utilization clamp constraints count
329 */
330 enum uclamp_id { UCLAMP_MIN = 0, UCLAMP_MAX, UCLAMP_CNT };
331
332 #ifdef CONFIG_SMP
333 extern struct root_domain def_root_domain;
334 extern struct mutex sched_domains_mutex;
335 #endif
336
337 struct sched_info {
338 #ifdef CONFIG_SCHED_INFO
339 /* Cumulative counters: */
340
341 /* # of times we have run on this CPU: */
342 unsigned long pcount;
343
344 /* Time spent waiting on a runqueue: */
345 unsigned long long run_delay;
346
347 /* Timestamps: */
348
349 /* When did we last run on a CPU? */
350 unsigned long long last_arrival;
351
352 /* When were we last queued to run? */
353 unsigned long long last_queued;
354
355 #endif /* CONFIG_SCHED_INFO */
356 };
357
358 /*
359 * Integer metrics need fixed point arithmetic, e.g., sched/fair
360 * has a few: load, load_avg, util_avg, freq, and capacity.
361 *
362 * We define a basic fixed point arithmetic range, and then formalize
363 * all these metrics based on that basic range.
364 */
365 #define SCHED_FIXEDPOINT_SHIFT 10
366 #define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
367
368 /* Increase resolution of cpu_capacity calculations */
369 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
370 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
371
372 struct load_weight {
373 unsigned long weight;
374 u32 inv_weight;
375 };
376
377 /**
378 * struct util_est - Estimation utilization of FAIR tasks
379 * @enqueued: instantaneous estimated utilization of a task/cpu
380 * @ewma: the Exponential Weighted Moving Average (EWMA)
381 * utilization of a task
382 *
383 * Support data structure to track an Exponential Weighted Moving Average
384 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
385 * average each time a task completes an activation. Sample's weight is chosen
386 * so that the EWMA will be relatively insensitive to transient changes to the
387 * task's workload.
388 *
389 * The enqueued attribute has a slightly different meaning for tasks and cpus:
390 * - task: the task's util_avg at last task dequeue time
391 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
392 * Thus, the util_est.enqueued of a task represents the contribution on the
393 * estimated utilization of the CPU where that task is currently enqueued.
394 *
395 * Only for tasks we track a moving average of the past instantaneous
396 * estimated utilization. This allows to absorb sporadic drops in utilization
397 * of an otherwise almost periodic task.
398 *
399 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
400 * updates. When a task is dequeued, its util_est should not be updated if its
401 * util_avg has not been updated in the meantime.
402 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
403 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
404 * for a task) it is safe to use MSB.
405 */
406 struct util_est {
407 unsigned int enqueued;
408 unsigned int ewma;
409 #define UTIL_EST_WEIGHT_SHIFT 2
410 #define UTIL_AVG_UNCHANGED 0x80000000
411 } __attribute__((__aligned__(sizeof(u64))));
412
413 /*
414 * The load/runnable/util_avg accumulates an infinite geometric series
415 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
416 *
417 * [load_avg definition]
418 *
419 * load_avg = runnable% * scale_load_down(load)
420 *
421 * [runnable_avg definition]
422 *
423 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
424 *
425 * [util_avg definition]
426 *
427 * util_avg = running% * SCHED_CAPACITY_SCALE
428 *
429 * where runnable% is the time ratio that a sched_entity is runnable and
430 * running% the time ratio that a sched_entity is running.
431 *
432 * For cfs_rq, they are the aggregated values of all runnable and blocked
433 * sched_entities.
434 *
435 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
436 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
437 * for computing those signals (see update_rq_clock_pelt())
438 *
439 * N.B., the above ratios (runnable% and running%) themselves are in the
440 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
441 * to as large a range as necessary. This is for example reflected by
442 * util_avg's SCHED_CAPACITY_SCALE.
443 *
444 * [Overflow issue]
445 *
446 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
447 * with the highest load (=88761), always runnable on a single cfs_rq,
448 * and should not overflow as the number already hits PID_MAX_LIMIT.
449 *
450 * For all other cases (including 32-bit kernels), struct load_weight's
451 * weight will overflow first before we do, because:
452 *
453 * Max(load_avg) <= Max(load.weight)
454 *
455 * Then it is the load_weight's responsibility to consider overflow
456 * issues.
457 */
458 struct sched_avg {
459 u64 last_update_time;
460 u64 load_sum;
461 u64 runnable_sum;
462 u32 util_sum;
463 u32 period_contrib;
464 unsigned long load_avg;
465 unsigned long runnable_avg;
466 unsigned long util_avg;
467 struct util_est util_est;
468 } ____cacheline_aligned;
469
470 struct sched_statistics {
471 #ifdef CONFIG_SCHEDSTATS
472 u64 wait_start;
473 u64 wait_max;
474 u64 wait_count;
475 u64 wait_sum;
476 u64 iowait_count;
477 u64 iowait_sum;
478
479 u64 sleep_start;
480 u64 sleep_max;
481 s64 sum_sleep_runtime;
482
483 u64 block_start;
484 u64 block_max;
485 u64 exec_max;
486 u64 slice_max;
487
488 u64 nr_migrations_cold;
489 u64 nr_failed_migrations_affine;
490 u64 nr_failed_migrations_running;
491 u64 nr_failed_migrations_hot;
492 u64 nr_forced_migrations;
493
494 u64 nr_wakeups;
495 u64 nr_wakeups_sync;
496 u64 nr_wakeups_migrate;
497 u64 nr_wakeups_local;
498 u64 nr_wakeups_remote;
499 u64 nr_wakeups_affine;
500 u64 nr_wakeups_affine_attempts;
501 u64 nr_wakeups_passive;
502 u64 nr_wakeups_idle;
503 #endif
504 };
505
506 struct sched_entity {
507 /* For load-balancing: */
508 struct load_weight load;
509 struct rb_node run_node;
510 struct list_head group_node;
511 unsigned int on_rq;
512
513 u64 exec_start;
514 u64 sum_exec_runtime;
515 u64 vruntime;
516 u64 prev_sum_exec_runtime;
517
518 u64 nr_migrations;
519
520 struct sched_statistics statistics;
521
522 #ifdef CONFIG_FAIR_GROUP_SCHED
523 int depth;
524 struct sched_entity *parent;
525 /* rq on which this entity is (to be) queued: */
526 struct cfs_rq *cfs_rq;
527 /* rq "owned" by this entity/group: */
528 struct cfs_rq *my_q;
529 /* cached value of my_q->h_nr_running */
530 unsigned long runnable_weight;
531 #endif
532
533 #ifdef CONFIG_SCHED_LATENCY_NICE
534 int latency_weight;
535 #endif
536 #ifdef CONFIG_SMP
537 /*
538 * Per entity load average tracking.
539 *
540 * Put into separate cache line so it does not
541 * collide with read-mostly values above.
542 */
543 struct sched_avg avg;
544 #endif
545 };
546
547 #ifdef CONFIG_SCHED_WALT
548 extern void sched_exit(struct task_struct *p);
549 extern int sched_set_init_task_load(struct task_struct *p, int init_load_pct);
550 extern u32 sched_get_init_task_load(struct task_struct *p);
551 extern void free_task_load_ptrs(struct task_struct *p);
552 #define RAVG_HIST_SIZE_MAX 5
553 struct ravg {
554 /*
555 * 'mark_start' marks the beginning of an event (task waking up, task
556 * starting to execute, task being preempted) within a window
557 *
558 * 'sum' represents how runnable a task has been within current
559 * window. It incorporates both running time and wait time and is
560 * frequency scaled.
561 *
562 * 'sum_history' keeps track of history of 'sum' seen over previous
563 * RAVG_HIST_SIZE windows. Windows where task was entirely sleeping are
564 * ignored.
565 *
566 * 'demand' represents maximum sum seen over previous
567 * sysctl_sched_ravg_hist_size windows. 'demand' could drive frequency
568 * demand for tasks.
569 *
570 * 'curr_window_cpu' represents task's contribution to cpu busy time on
571 * various CPUs in the current window
572 *
573 * 'prev_window_cpu' represents task's contribution to cpu busy time on
574 * various CPUs in the previous window
575 *
576 * 'curr_window' represents the sum of all entries in curr_window_cpu
577 *
578 * 'prev_window' represents the sum of all entries in prev_window_cpu
579 *
580 */
581 u64 mark_start;
582 u32 sum, demand;
583 u32 sum_history[RAVG_HIST_SIZE_MAX];
584 u32 *curr_window_cpu, *prev_window_cpu;
585 u32 curr_window, prev_window;
586 u16 active_windows;
587 u16 demand_scaled;
588 };
589 #else
sched_exit(struct task_struct * p)590 static inline void sched_exit(struct task_struct *p)
591 {
592 }
free_task_load_ptrs(struct task_struct * p)593 static inline void free_task_load_ptrs(struct task_struct *p)
594 {
595 }
596 #endif /* CONFIG_SCHED_WALT */
597
598 struct sched_rt_entity {
599 struct list_head run_list;
600 unsigned long timeout;
601 unsigned long watchdog_stamp;
602 unsigned int time_slice;
603 unsigned short on_rq;
604 unsigned short on_list;
605
606 struct sched_rt_entity *back;
607 #ifdef CONFIG_RT_GROUP_SCHED
608 struct sched_rt_entity *parent;
609 /* rq on which this entity is (to be) queued: */
610 struct rt_rq *rt_rq;
611 /* rq "owned" by this entity/group: */
612 struct rt_rq *my_q;
613 #endif
614 } __randomize_layout;
615
616 struct sched_dl_entity {
617 struct rb_node rb_node;
618
619 /*
620 * Original scheduling parameters. Copied here from sched_attr
621 * during sched_setattr(), they will remain the same until
622 * the next sched_setattr().
623 */
624 u64 dl_runtime; /* Maximum runtime for each instance */
625 u64 dl_deadline; /* Relative deadline of each instance */
626 u64 dl_period; /* Separation of two instances (period) */
627 u64 dl_bw; /* dl_runtime / dl_period */
628 u64 dl_density; /* dl_runtime / dl_deadline */
629
630 /*
631 * Actual scheduling parameters. Initialized with the values above,
632 * they are continuously updated during task execution. Note that
633 * the remaining runtime could be < 0 in case we are in overrun.
634 */
635 s64 runtime; /* Remaining runtime for this instance */
636 u64 deadline; /* Absolute deadline for this instance */
637 unsigned int flags; /* Specifying the scheduler behaviour */
638
639 /*
640 * Some bool flags:
641 *
642 * @dl_throttled tells if we exhausted the runtime. If so, the
643 * task has to wait for a replenishment to be performed at the
644 * next firing of dl_timer.
645 *
646 * @dl_boosted tells if we are boosted due to DI. If so we are
647 * outside bandwidth enforcement mechanism (but only until we
648 * exit the critical section);
649 *
650 * @dl_yielded tells if task gave up the CPU before consuming
651 * all its available runtime during the last job.
652 *
653 * @dl_non_contending tells if the task is inactive while still
654 * contributing to the active utilization. In other words, it
655 * indicates if the inactive timer has been armed and its handler
656 * has not been executed yet. This flag is useful to avoid race
657 * conditions between the inactive timer handler and the wakeup
658 * code.
659 *
660 * @dl_overrun tells if the task asked to be informed about runtime
661 * overruns.
662 */
663 unsigned int dl_throttled : 1;
664 unsigned int dl_yielded : 1;
665 unsigned int dl_non_contending : 1;
666 unsigned int dl_overrun : 1;
667
668 /*
669 * Bandwidth enforcement timer. Each -deadline task has its
670 * own bandwidth to be enforced, thus we need one timer per task.
671 */
672 struct hrtimer dl_timer;
673
674 /*
675 * Inactive timer, responsible for decreasing the active utilization
676 * at the "0-lag time". When a -deadline task blocks, it contributes
677 * to GRUB's active utilization until the "0-lag time", hence a
678 * timer is needed to decrease the active utilization at the correct
679 * time.
680 */
681 struct hrtimer inactive_timer;
682
683 #ifdef CONFIG_RT_MUTEXES
684 /*
685 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
686 * pi_se points to the donor, otherwise points to the dl_se it belongs
687 * to (the original one/itself).
688 */
689 struct sched_dl_entity *pi_se;
690 #endif
691 };
692
693 #ifdef CONFIG_UCLAMP_TASK
694 /* Number of utilization clamp buckets (shorter alias) */
695 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
696
697 /*
698 * Utilization clamp for a scheduling entity
699 * @value: clamp value "assigned" to a se
700 * @bucket_id: bucket index corresponding to the "assigned" value
701 * @active: the se is currently refcounted in a rq's bucket
702 * @user_defined: the requested clamp value comes from user-space
703 *
704 * The bucket_id is the index of the clamp bucket matching the clamp value
705 * which is pre-computed and stored to avoid expensive integer divisions from
706 * the fast path.
707 *
708 * The active bit is set whenever a task has got an "effective" value assigned,
709 * which can be different from the clamp value "requested" from user-space.
710 * This allows to know a task is refcounted in the rq's bucket corresponding
711 * to the "effective" bucket_id.
712 *
713 * The user_defined bit is set whenever a task has got a task-specific clamp
714 * value requested from userspace, i.e. the system defaults apply to this task
715 * just as a restriction. This allows to relax default clamps when a less
716 * restrictive task-specific value has been requested, thus allowing to
717 * implement a "nice" semantic. For example, a task running with a 20%
718 * default boost can still drop its own boosting to 0%.
719 */
720 struct uclamp_se {
721 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
722 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
723 unsigned int active : 1;
724 unsigned int user_defined : 1;
725 };
726 #endif /* CONFIG_UCLAMP_TASK */
727
728 union rcu_special {
729 struct {
730 u8 blocked;
731 u8 need_qs;
732 u8 exp_hint; /* Hint for performance. */
733 u8 need_mb; /* Readers need smp_mb(). */
734 } b; /* Bits. */
735 u32 s; /* Set of bits. */
736 };
737
738 enum perf_event_task_context {
739 perf_invalid_context = -1,
740 perf_hw_context = 0,
741 perf_sw_context,
742 perf_nr_task_contexts,
743 };
744
745 struct wake_q_node {
746 struct wake_q_node *next;
747 };
748
749 struct task_struct {
750 #ifdef CONFIG_THREAD_INFO_IN_TASK
751 /*
752 * For reasons of header soup (see current_thread_info()), this
753 * must be the first element of task_struct.
754 */
755 struct thread_info thread_info;
756 #endif
757 /* -1 unrunnable, 0 runnable, >0 stopped: */
758 volatile long state;
759
760 /*
761 * This begins the randomizable portion of task_struct. Only
762 * scheduling-critical items should be added above here.
763 */
764 randomized_struct_fields_start
765
766 void *stack;
767 refcount_t usage;
768 /* Per task flags (PF_*), defined further below: */
769 unsigned int flags;
770 unsigned int ptrace;
771
772 #ifdef CONFIG_SMP
773 int on_cpu;
774 struct __call_single_node wake_entry;
775 #ifdef CONFIG_THREAD_INFO_IN_TASK
776 /* Current CPU: */
777 unsigned int cpu;
778 #endif
779 unsigned int wakee_flips;
780 unsigned long wakee_flip_decay_ts;
781 struct task_struct *last_wakee;
782
783 /*
784 * recent_used_cpu is initially set as the last CPU used by a task
785 * that wakes affine another task. Waker/wakee relationships can
786 * push tasks around a CPU where each wakeup moves to the next one.
787 * Tracking a recently used CPU allows a quick search for a recently
788 * used CPU that may be idle.
789 */
790 int recent_used_cpu;
791 int wake_cpu;
792 #endif
793 int on_rq;
794
795 int prio;
796 int static_prio;
797 int normal_prio;
798 unsigned int rt_priority;
799 #ifdef CONFIG_SCHED_LATENCY_NICE
800 int latency_prio;
801 #endif
802
803 const struct sched_class *sched_class;
804 struct sched_entity se;
805 struct sched_rt_entity rt;
806 #ifdef CONFIG_SCHED_WALT
807 struct ravg ravg;
808 /*
809 * 'init_load_pct' represents the initial task load assigned to children
810 * of this task
811 */
812 u32 init_load_pct;
813 u64 last_sleep_ts;
814 #endif
815
816 #ifdef CONFIG_SCHED_RTG
817 int rtg_depth;
818 struct related_thread_group *grp;
819 struct list_head grp_list;
820 #endif
821
822 #ifdef CONFIG_CGROUP_SCHED
823 struct task_group *sched_task_group;
824 #endif
825 struct sched_dl_entity dl;
826
827 #ifdef CONFIG_UCLAMP_TASK
828 /*
829 * Clamp values requested for a scheduling entity.
830 * Must be updated with task_rq_lock() held.
831 */
832 struct uclamp_se uclamp_req[UCLAMP_CNT];
833 /*
834 * Effective clamp values used for a scheduling entity.
835 * Must be updated with task_rq_lock() held.
836 */
837 struct uclamp_se uclamp[UCLAMP_CNT];
838 #endif
839
840 #ifdef CONFIG_PREEMPT_NOTIFIERS
841 /* List of struct preempt_notifier: */
842 struct hlist_head preempt_notifiers;
843 #endif
844
845 #ifdef CONFIG_BLK_DEV_IO_TRACE
846 unsigned int btrace_seq;
847 #endif
848
849 unsigned int policy;
850 int nr_cpus_allowed;
851 const cpumask_t *cpus_ptr;
852 cpumask_t cpus_mask;
853
854 #ifdef CONFIG_PREEMPT_RCU
855 int rcu_read_lock_nesting;
856 union rcu_special rcu_read_unlock_special;
857 struct list_head rcu_node_entry;
858 struct rcu_node *rcu_blocked_node;
859 #endif /* #ifdef CONFIG_PREEMPT_RCU */
860
861 #ifdef CONFIG_TASKS_RCU
862 unsigned long rcu_tasks_nvcsw;
863 u8 rcu_tasks_holdout;
864 u8 rcu_tasks_idx;
865 int rcu_tasks_idle_cpu;
866 struct list_head rcu_tasks_holdout_list;
867 #endif /* #ifdef CONFIG_TASKS_RCU */
868
869 #ifdef CONFIG_TASKS_TRACE_RCU
870 int trc_reader_nesting;
871 int trc_ipi_to_cpu;
872 union rcu_special trc_reader_special;
873 bool trc_reader_checked;
874 struct list_head trc_holdout_list;
875 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
876
877 struct sched_info sched_info;
878
879 struct list_head tasks;
880 #ifdef CONFIG_SMP
881 struct plist_node pushable_tasks;
882 struct rb_node pushable_dl_tasks;
883 #endif
884
885 struct mm_struct *mm;
886 struct mm_struct *active_mm;
887
888 /* Per-thread vma caching: */
889 struct vmacache vmacache;
890
891 #ifdef SPLIT_RSS_COUNTING
892 struct task_rss_stat rss_stat;
893 #endif
894 int exit_state;
895 int exit_code;
896 int exit_signal;
897 /* The signal sent when the parent dies: */
898 int pdeath_signal;
899 /* JOBCTL_*, siglock protected: */
900 unsigned long jobctl;
901
902 /* Used for emulating ABI behavior of previous Linux versions: */
903 unsigned int personality;
904
905 /* Scheduler bits, serialized by scheduler locks: */
906 unsigned sched_reset_on_fork : 1;
907 unsigned sched_contributes_to_load : 1;
908 unsigned sched_migrated : 1;
909 #ifdef CONFIG_PSI
910 unsigned sched_psi_wake_requeue : 1;
911 #endif
912
913 /* Force alignment to the next boundary: */
914 unsigned : 0;
915
916 /* Unserialized, strictly 'current' */
917
918 /*
919 * This field must not be in the scheduler word above due to wakelist
920 * queueing no longer being serialized by p->on_cpu. However:
921 *
922 * p->XXX = X; ttwu()
923 * schedule() if (p->on_rq && ..) // false
924 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
925 * deactivate_task() ttwu_queue_wakelist())
926 * p->on_rq = 0; p->sched_remote_wakeup = Y;
927 *
928 * guarantees all stores of 'current' are visible before
929 * ->sched_remote_wakeup gets used, so it can be in this word.
930 */
931 unsigned sched_remote_wakeup : 1;
932
933 /* Bit to tell LSMs we're in execve(): */
934 unsigned in_execve : 1;
935 unsigned in_iowait : 1;
936 #ifndef TIF_RESTORE_SIGMASK
937 unsigned restore_sigmask : 1;
938 #endif
939 #ifdef CONFIG_MEMCG
940 unsigned in_user_fault : 1;
941 #endif
942 #ifdef CONFIG_COMPAT_BRK
943 unsigned brk_randomized : 1;
944 #endif
945 #ifdef CONFIG_CGROUPS
946 /* disallow userland-initiated cgroup migration */
947 unsigned no_cgroup_migration : 1;
948 /* task is frozen/stopped (used by the cgroup freezer) */
949 unsigned frozen : 1;
950 #endif
951 #ifdef CONFIG_BLK_CGROUP
952 unsigned use_memdelay : 1;
953 #endif
954 #ifdef CONFIG_PSI
955 /* Stalled due to lack of memory */
956 unsigned in_memstall : 1;
957 #endif
958
959 unsigned long atomic_flags; /* Flags requiring atomic access. */
960
961 struct restart_block restart_block;
962
963 pid_t pid;
964 pid_t tgid;
965
966 #ifdef CONFIG_STACKPROTECTOR
967 /* Canary value for the -fstack-protector GCC feature: */
968 unsigned long stack_canary;
969 #endif
970 /*
971 * Pointers to the (original) parent process, youngest child, younger sibling,
972 * older sibling, respectively. (p->father can be replaced with
973 * p->real_parent->pid)
974 */
975
976 /* Real parent process: */
977 struct task_struct __rcu *real_parent;
978
979 /* Recipient of SIGCHLD, wait4() reports: */
980 struct task_struct __rcu *parent;
981
982 /*
983 * Children/sibling form the list of natural children:
984 */
985 struct list_head children;
986 struct list_head sibling;
987 struct task_struct *group_leader;
988
989 /*
990 * 'ptraced' is the list of tasks this task is using ptrace() on.
991 *
992 * This includes both natural children and PTRACE_ATTACH targets.
993 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
994 */
995 struct list_head ptraced;
996 struct list_head ptrace_entry;
997
998 /* PID/PID hash table linkage. */
999 struct pid *thread_pid;
1000 struct hlist_node pid_links[PIDTYPE_MAX];
1001 struct list_head thread_group;
1002 struct list_head thread_node;
1003
1004 struct completion *vfork_done;
1005
1006 /* CLONE_CHILD_SETTID: */
1007 int __user *set_child_tid;
1008
1009 /* CLONE_CHILD_CLEARTID: */
1010 int __user *clear_child_tid;
1011
1012 /* PF_IO_WORKER */
1013 void *pf_io_worker;
1014
1015 u64 utime;
1016 u64 stime;
1017 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1018 u64 utimescaled;
1019 u64 stimescaled;
1020 #endif
1021 u64 gtime;
1022 struct prev_cputime prev_cputime;
1023 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1024 struct vtime vtime;
1025 #endif
1026
1027 #ifdef CONFIG_NO_HZ_FULL
1028 atomic_t tick_dep_mask;
1029 #endif
1030 /* Context switch counts: */
1031 unsigned long nvcsw;
1032 unsigned long nivcsw;
1033
1034 /* Monotonic time in nsecs: */
1035 u64 start_time;
1036
1037 /* Boot based time in nsecs: */
1038 u64 start_boottime;
1039
1040 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1041 unsigned long min_flt;
1042 unsigned long maj_flt;
1043
1044 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1045 struct posix_cputimers posix_cputimers;
1046
1047 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1048 struct posix_cputimers_work posix_cputimers_work;
1049 #endif
1050
1051 /* Process credentials: */
1052
1053 /* Tracer's credentials at attach: */
1054 const struct cred __rcu *ptracer_cred;
1055
1056 /* Objective and real subjective task credentials (COW): */
1057 const struct cred __rcu *real_cred;
1058
1059 /* Effective (overridable) subjective task credentials (COW): */
1060 const struct cred __rcu *cred;
1061
1062 #ifdef CONFIG_KEYS
1063 /* Cached requested key. */
1064 struct key *cached_requested_key;
1065 #endif
1066
1067 /*
1068 * executable name, excluding path.
1069 *
1070 * - normally initialized setup_new_exec()
1071 * - access it with [gs]et_task_comm()
1072 * - lock it with task_lock()
1073 */
1074 char comm[TASK_COMM_LEN];
1075
1076 struct nameidata *nameidata;
1077
1078 #ifdef CONFIG_SYSVIPC
1079 struct sysv_sem sysvsem;
1080 struct sysv_shm sysvshm;
1081 #endif
1082 #ifdef CONFIG_DETECT_HUNG_TASK
1083 unsigned long last_switch_count;
1084 unsigned long last_switch_time;
1085 #endif
1086 /* Filesystem information: */
1087 struct fs_struct *fs;
1088
1089 /* Open file information: */
1090 struct files_struct *files;
1091
1092 #ifdef CONFIG_IO_URING
1093 struct io_uring_task *io_uring;
1094 #endif
1095
1096 /* Namespaces: */
1097 struct nsproxy *nsproxy;
1098
1099 /* Signal handlers: */
1100 struct signal_struct *signal;
1101 struct sighand_struct __rcu *sighand;
1102 sigset_t blocked;
1103 sigset_t real_blocked;
1104 /* Restored if set_restore_sigmask() was used: */
1105 sigset_t saved_sigmask;
1106 struct sigpending pending;
1107 unsigned long sas_ss_sp;
1108 size_t sas_ss_size;
1109 unsigned int sas_ss_flags;
1110
1111 struct callback_head *task_works;
1112
1113 #ifdef CONFIG_AUDIT
1114 #ifdef CONFIG_AUDITSYSCALL
1115 struct audit_context *audit_context;
1116 #endif
1117 kuid_t loginuid;
1118 unsigned int sessionid;
1119 #endif
1120 struct seccomp seccomp;
1121
1122 /* Thread group tracking: */
1123 u64 parent_exec_id;
1124 u64 self_exec_id;
1125
1126 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1127 spinlock_t alloc_lock;
1128
1129 /* Protection of the PI data structures: */
1130 raw_spinlock_t pi_lock;
1131
1132 struct wake_q_node wake_q;
1133
1134 #ifdef CONFIG_RT_MUTEXES
1135 /* PI waiters blocked on a rt_mutex held by this task: */
1136 struct rb_root_cached pi_waiters;
1137 /* Updated under owner's pi_lock and rq lock */
1138 struct task_struct *pi_top_task;
1139 /* Deadlock detection and priority inheritance handling: */
1140 struct rt_mutex_waiter *pi_blocked_on;
1141 #endif
1142
1143 #ifdef CONFIG_DEBUG_MUTEXES
1144 /* Mutex deadlock detection: */
1145 struct mutex_waiter *blocked_on;
1146 #endif
1147
1148 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1149 int non_block_count;
1150 #endif
1151
1152 #ifdef CONFIG_TRACE_IRQFLAGS
1153 struct irqtrace_events irqtrace;
1154 unsigned int hardirq_threaded;
1155 u64 hardirq_chain_key;
1156 int softirqs_enabled;
1157 int softirq_context;
1158 int irq_config;
1159 #endif
1160
1161 #ifdef CONFIG_LOCKDEP
1162 #define MAX_LOCK_DEPTH 48UL
1163 u64 curr_chain_key;
1164 int lockdep_depth;
1165 unsigned int lockdep_recursion;
1166 struct held_lock held_locks[MAX_LOCK_DEPTH];
1167 #endif
1168
1169 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1170 unsigned int in_ubsan;
1171 #endif
1172
1173 /* Journalling filesystem info: */
1174 void *journal_info;
1175
1176 /* Stacked block device info: */
1177 struct bio_list *bio_list;
1178
1179 #ifdef CONFIG_BLOCK
1180 /* Stack plugging: */
1181 struct blk_plug *plug;
1182 #endif
1183
1184 /* VM state: */
1185 struct reclaim_state *reclaim_state;
1186
1187 struct backing_dev_info *backing_dev_info;
1188
1189 struct io_context *io_context;
1190
1191 #ifdef CONFIG_COMPACTION
1192 struct capture_control *capture_control;
1193 #endif
1194 /* Ptrace state: */
1195 unsigned long ptrace_message;
1196 kernel_siginfo_t *last_siginfo;
1197
1198 struct task_io_accounting ioac;
1199 #ifdef CONFIG_PSI
1200 /* Pressure stall state */
1201 unsigned int psi_flags;
1202 #endif
1203 #ifdef CONFIG_TASK_XACCT
1204 /* Accumulated RSS usage: */
1205 u64 acct_rss_mem1;
1206 /* Accumulated virtual memory usage: */
1207 u64 acct_vm_mem1;
1208 /* stime + utime since last update: */
1209 u64 acct_timexpd;
1210 #endif
1211 #ifdef CONFIG_CPUSETS
1212 /* Protected by ->alloc_lock: */
1213 nodemask_t mems_allowed;
1214 /* Seqence number to catch updates: */
1215 seqcount_spinlock_t mems_allowed_seq;
1216 int cpuset_mem_spread_rotor;
1217 int cpuset_slab_spread_rotor;
1218 #endif
1219 #ifdef CONFIG_CGROUPS
1220 /* Control Group info protected by css_set_lock: */
1221 struct css_set __rcu *cgroups;
1222 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1223 struct list_head cg_list;
1224 #endif
1225 #ifdef CONFIG_X86_CPU_RESCTRL
1226 u32 closid;
1227 u32 rmid;
1228 #endif
1229 #ifdef CONFIG_FUTEX
1230 struct robust_list_head __user *robust_list;
1231 #ifdef CONFIG_COMPAT
1232 struct compat_robust_list_head __user *compat_robust_list;
1233 #endif
1234 struct list_head pi_state_list;
1235 struct futex_pi_state *pi_state_cache;
1236 struct mutex futex_exit_mutex;
1237 unsigned int futex_state;
1238 #endif
1239 #ifdef CONFIG_PERF_EVENTS
1240 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1241 struct mutex perf_event_mutex;
1242 struct list_head perf_event_list;
1243 #endif
1244 #ifdef CONFIG_DEBUG_PREEMPT
1245 unsigned long preempt_disable_ip;
1246 #endif
1247 #ifdef CONFIG_NUMA
1248 /* Protected by alloc_lock: */
1249 struct mempolicy *mempolicy;
1250 short il_prev;
1251 short pref_node_fork;
1252 #endif
1253 #ifdef CONFIG_NUMA_BALANCING
1254 int numa_scan_seq;
1255 unsigned int numa_scan_period;
1256 unsigned int numa_scan_period_max;
1257 int numa_preferred_nid;
1258 unsigned long numa_migrate_retry;
1259 /* Migration stamp: */
1260 u64 node_stamp;
1261 u64 last_task_numa_placement;
1262 u64 last_sum_exec_runtime;
1263 struct callback_head numa_work;
1264
1265 /*
1266 * This pointer is only modified for current in syscall and
1267 * pagefault context (and for tasks being destroyed), so it can be read
1268 * from any of the following contexts:
1269 * - RCU read-side critical section
1270 * - current->numa_group from everywhere
1271 * - task's runqueue locked, task not running
1272 */
1273 struct numa_group __rcu *numa_group;
1274
1275 /*
1276 * numa_faults is an array split into four regions:
1277 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1278 * in this precise order.
1279 *
1280 * faults_memory: Exponential decaying average of faults on a per-node
1281 * basis. Scheduling placement decisions are made based on these
1282 * counts. The values remain static for the duration of a PTE scan.
1283 * faults_cpu: Track the nodes the process was running on when a NUMA
1284 * hinting fault was incurred.
1285 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1286 * during the current scan window. When the scan completes, the counts
1287 * in faults_memory and faults_cpu decay and these values are copied.
1288 */
1289 unsigned long *numa_faults;
1290 unsigned long total_numa_faults;
1291
1292 /*
1293 * numa_faults_locality tracks if faults recorded during the last
1294 * scan window were remote/local or failed to migrate. The task scan
1295 * period is adapted based on the locality of the faults with different
1296 * weights depending on whether they were shared or private faults
1297 */
1298 unsigned long numa_faults_locality[3];
1299
1300 unsigned long numa_pages_migrated;
1301 #endif /* CONFIG_NUMA_BALANCING */
1302
1303 #ifdef CONFIG_RSEQ
1304 struct rseq __user *rseq;
1305 u32 rseq_sig;
1306 /*
1307 * RmW on rseq_event_mask must be performed atomically
1308 * with respect to preemption.
1309 */
1310 unsigned long rseq_event_mask;
1311 #endif
1312
1313 struct tlbflush_unmap_batch tlb_ubc;
1314
1315 union {
1316 refcount_t rcu_users;
1317 struct rcu_head rcu;
1318 };
1319
1320 /* Cache last used pipe for splice(): */
1321 struct pipe_inode_info *splice_pipe;
1322
1323 struct page_frag task_frag;
1324
1325 #ifdef CONFIG_TASK_DELAY_ACCT
1326 struct task_delay_info *delays;
1327 #endif
1328
1329 #ifdef CONFIG_RECLAIM_ACCT
1330 struct reclaim_acct *reclaim_acct;
1331 #endif
1332
1333 #ifdef CONFIG_FAULT_INJECTION
1334 int make_it_fail;
1335 unsigned int fail_nth;
1336 #endif
1337 /*
1338 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1339 * balance_dirty_pages() for a dirty throttling pause:
1340 */
1341 int nr_dirtied;
1342 int nr_dirtied_pause;
1343 /* Start of a write-and-pause period: */
1344 unsigned long dirty_paused_when;
1345
1346 #ifdef CONFIG_LATENCYTOP
1347 int latency_record_count;
1348 struct latency_record latency_record[LT_SAVECOUNT];
1349 #endif
1350 /*
1351 * Time slack values; these are used to round up poll() and
1352 * select() etc timeout values. These are in nanoseconds.
1353 */
1354 u64 timer_slack_ns;
1355 u64 default_timer_slack_ns;
1356
1357 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1358 unsigned int kasan_depth;
1359 #endif
1360
1361 #ifdef CONFIG_KCSAN
1362 struct kcsan_ctx kcsan_ctx;
1363 #ifdef CONFIG_TRACE_IRQFLAGS
1364 struct irqtrace_events kcsan_save_irqtrace;
1365 #endif
1366 #endif
1367
1368 #if IS_ENABLED(CONFIG_KUNIT)
1369 struct kunit *kunit_test;
1370 #endif
1371
1372 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1373 /* Index of current stored address in ret_stack: */
1374 int curr_ret_stack;
1375 int curr_ret_depth;
1376
1377 /* Stack of return addresses for return function tracing: */
1378 struct ftrace_ret_stack *ret_stack;
1379
1380 /* Timestamp for last schedule: */
1381 unsigned long long ftrace_timestamp;
1382
1383 /*
1384 * Number of functions that haven't been traced
1385 * because of depth overrun:
1386 */
1387 atomic_t trace_overrun;
1388
1389 /* Pause tracing: */
1390 atomic_t tracing_graph_pause;
1391 #endif
1392
1393 #ifdef CONFIG_TRACING
1394 /* State flags for use by tracers: */
1395 unsigned long trace;
1396
1397 /* Bitmask and counter of trace recursion: */
1398 unsigned long trace_recursion;
1399 #endif /* CONFIG_TRACING */
1400
1401 #ifdef CONFIG_KCOV
1402 /* See kernel/kcov.c for more details. */
1403
1404 /* Coverage collection mode enabled for this task (0 if disabled): */
1405 unsigned int kcov_mode;
1406
1407 /* Size of the kcov_area: */
1408 unsigned int kcov_size;
1409
1410 /* Buffer for coverage collection: */
1411 void *kcov_area;
1412
1413 /* KCOV descriptor wired with this task or NULL: */
1414 struct kcov *kcov;
1415
1416 /* KCOV common handle for remote coverage collection: */
1417 u64 kcov_handle;
1418
1419 /* KCOV sequence number: */
1420 int kcov_sequence;
1421
1422 /* Collect coverage from softirq context: */
1423 unsigned int kcov_softirq;
1424 #endif
1425
1426 #ifdef CONFIG_MEMCG
1427 struct mem_cgroup *memcg_in_oom;
1428 gfp_t memcg_oom_gfp_mask;
1429 int memcg_oom_order;
1430
1431 /* Number of pages to reclaim on returning to userland: */
1432 unsigned int memcg_nr_pages_over_high;
1433
1434 /* Used by memcontrol for targeted memcg charge: */
1435 struct mem_cgroup *active_memcg;
1436 #endif
1437
1438 #ifdef CONFIG_BLK_CGROUP
1439 struct request_queue *throttle_queue;
1440 #endif
1441
1442 #ifdef CONFIG_UPROBES
1443 struct uprobe_task *utask;
1444 #endif
1445 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1446 unsigned int sequential_io;
1447 unsigned int sequential_io_avg;
1448 #endif
1449 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1450 unsigned long task_state_change;
1451 #endif
1452 int pagefault_disabled;
1453 #ifdef CONFIG_MMU
1454 struct task_struct *oom_reaper_list;
1455 struct timer_list oom_reaper_timer;
1456 #endif
1457 #ifdef CONFIG_VMAP_STACK
1458 struct vm_struct *stack_vm_area;
1459 #endif
1460 #ifdef CONFIG_THREAD_INFO_IN_TASK
1461 /* A live task holds one reference: */
1462 refcount_t stack_refcount;
1463 #endif
1464 #ifdef CONFIG_LIVEPATCH
1465 int patch_state;
1466 #endif
1467 #ifdef CONFIG_SECURITY
1468 /* Used by LSM modules for access restriction: */
1469 void *security;
1470 #endif
1471 #ifdef CONFIG_BPF_SYSCALL
1472 /* Used for BPF run context */
1473 struct bpf_run_ctx *bpf_ctx;
1474 #endif
1475
1476 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1477 unsigned long lowest_stack;
1478 unsigned long prev_lowest_stack;
1479 #endif
1480
1481 #ifdef CONFIG_X86_MCE
1482 void __user *mce_vaddr;
1483 __u64 mce_kflags;
1484 u64 mce_addr;
1485 __u64 mce_ripv : 1, mce_whole_page : 1, __mce_reserved : 62;
1486 struct callback_head mce_kill_me;
1487 int mce_count;
1488 #endif
1489
1490 #ifdef CONFIG_ACCESS_TOKENID
1491 u64 token;
1492 u64 ftoken;
1493 #endif
1494 /*
1495 * New fields for task_struct should be added above here, so that
1496 * they are included in the randomized portion of task_struct.
1497 */
1498 randomized_struct_fields_end
1499
1500 /* CPU-specific state of this task: */
1501 struct thread_struct thread;
1502
1503 /*
1504 * WARNING: on x86, 'thread_struct' contains a variable-sized
1505 * structure. It *MUST* be at the end of 'task_struct'.
1506 *
1507 * Do not put anything below here!
1508 */
1509 };
1510
task_pid(struct task_struct * task)1511 static inline struct pid *task_pid(struct task_struct *task)
1512 {
1513 return task->thread_pid;
1514 }
1515
1516 /*
1517 * the helpers to get the task's different pids as they are seen
1518 * from various namespaces
1519 *
1520 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1521 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1522 * current.
1523 * task_xid_nr_ns() : id seen from the ns specified;
1524 *
1525 * see also pid_nr() etc in include/linux/pid.h
1526 */
1527 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1528
task_pid_nr(struct task_struct * tsk)1529 static inline pid_t task_pid_nr(struct task_struct *tsk)
1530 {
1531 return tsk->pid;
1532 }
1533
task_pid_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1534 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1535 {
1536 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1537 }
1538
task_pid_vnr(struct task_struct * tsk)1539 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1540 {
1541 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1542 }
1543
task_tgid_nr(struct task_struct * tsk)1544 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1545 {
1546 return tsk->tgid;
1547 }
1548
1549 /**
1550 * pid_alive - check that a task structure is not stale
1551 * @p: Task structure to be checked.
1552 *
1553 * Test if a process is not yet dead (at most zombie state)
1554 * If pid_alive fails, then pointers within the task structure
1555 * can be stale and must not be dereferenced.
1556 *
1557 * Return: 1 if the process is alive. 0 otherwise.
1558 */
pid_alive(const struct task_struct * p)1559 static inline int pid_alive(const struct task_struct *p)
1560 {
1561 return p->thread_pid != NULL;
1562 }
1563
task_pgrp_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1564 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1565 {
1566 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1567 }
1568
task_pgrp_vnr(struct task_struct * tsk)1569 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1570 {
1571 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1572 }
1573
task_session_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1574 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1575 {
1576 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1577 }
1578
task_session_vnr(struct task_struct * tsk)1579 static inline pid_t task_session_vnr(struct task_struct *tsk)
1580 {
1581 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1582 }
1583
task_tgid_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1584 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1585 {
1586 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1587 }
1588
task_tgid_vnr(struct task_struct * tsk)1589 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1590 {
1591 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1592 }
1593
task_ppid_nr_ns(const struct task_struct * tsk,struct pid_namespace * ns)1594 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1595 {
1596 pid_t pid = 0;
1597
1598 rcu_read_lock();
1599 if (pid_alive(tsk)) {
1600 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1601 }
1602 rcu_read_unlock();
1603
1604 return pid;
1605 }
1606
task_ppid_nr(const struct task_struct * tsk)1607 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1608 {
1609 return task_ppid_nr_ns(tsk, &init_pid_ns);
1610 }
1611
1612 /* Obsolete, do not use */
task_pgrp_nr(struct task_struct * tsk)1613 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1614 {
1615 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1616 }
1617
1618 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1619 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1620
task_state_index(struct task_struct * tsk)1621 static inline unsigned int task_state_index(struct task_struct *tsk)
1622 {
1623 unsigned int tsk_state = READ_ONCE(tsk->state);
1624 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1625
1626 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1627
1628 if (tsk_state == TASK_IDLE) {
1629 state = TASK_REPORT_IDLE;
1630 }
1631
1632 return fls(state);
1633 }
1634
task_index_to_char(unsigned int state)1635 static inline char task_index_to_char(unsigned int state)
1636 {
1637 static const char state_char[] = "RSDTtXZPI";
1638
1639 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1640
1641 return state_char[state];
1642 }
1643
task_state_to_char(struct task_struct * tsk)1644 static inline char task_state_to_char(struct task_struct *tsk)
1645 {
1646 return task_index_to_char(task_state_index(tsk));
1647 }
1648
1649 /**
1650 * is_global_init - check if a task structure is init. Since init
1651 * is free to have sub-threads we need to check tgid.
1652 * @tsk: Task structure to be checked.
1653 *
1654 * Check if a task structure is the first user space task the kernel created.
1655 *
1656 * Return: 1 if the task structure is init. 0 otherwise.
1657 */
is_global_init(struct task_struct * tsk)1658 static inline int is_global_init(struct task_struct *tsk)
1659 {
1660 return task_tgid_nr(tsk) == 1;
1661 }
1662
1663 extern struct pid *cad_pid;
1664
1665 /*
1666 * Per process flags
1667 */
1668 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1669 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1670 #define PF_EXITING 0x00000004 /* Getting shut down */
1671 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1672 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1673 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1674 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1675 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1676 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1677 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1678 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1679 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1680 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1681 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1682 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1683 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1684 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1685 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1686 #define PF_LOCAL_THROTTLE \
1687 0x00100000 /* Throttle writes only against the bdi I write to, \
1688 * I am cleaning dirty pages from some other bdi. */
1689 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1690 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1691 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1692 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1693 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1694 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1695 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1696 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1697
1698 /*
1699 * Only the _current_ task can read/write to tsk->flags, but other
1700 * tasks can access tsk->flags in readonly mode for example
1701 * with tsk_used_math (like during threaded core dumping).
1702 * There is however an exception to this rule during ptrace
1703 * or during fork: the ptracer task is allowed to write to the
1704 * child->flags of its traced child (same goes for fork, the parent
1705 * can write to the child->flags), because we're guaranteed the
1706 * child is not running and in turn not changing child->flags
1707 * at the same time the parent does it.
1708 */
1709 #define clear_stopped_child_used_math(child) \
1710 do { \
1711 (child)->flags &= ~PF_USED_MATH; \
1712 } while (0)
1713 #define set_stopped_child_used_math(child) \
1714 do { \
1715 (child)->flags |= PF_USED_MATH; \
1716 } while (0)
1717 #define clear_used_math() clear_stopped_child_used_math(current)
1718 #define set_used_math() set_stopped_child_used_math(current)
1719
1720 #define conditional_stopped_child_used_math(condition, child) \
1721 do { \
1722 (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; \
1723 } while (0)
1724
1725 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1726
1727 #define copy_to_stopped_child_used_math(child) \
1728 do { \
1729 (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; \
1730 } while (0)
1731
1732 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1733 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1734 #define used_math() tsk_used_math(current)
1735
is_percpu_thread(void)1736 static __always_inline bool is_percpu_thread(void)
1737 {
1738 #ifdef CONFIG_SMP
1739 return (current->flags & PF_NO_SETAFFINITY) && (current->nr_cpus_allowed == 1);
1740 #else
1741 return true;
1742 #endif
1743 }
1744
1745 /* Per-process atomic flags. */
1746 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1747 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1748 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1749 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1750 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled */
1751 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1752 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1753 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1754
1755 #define TASK_PFA_TEST(name, func) \
1756 static inline bool task_##func(struct task_struct *p) \
1757 { \
1758 return test_bit(PFA_##name, &p->atomic_flags); \
1759 }
1760
1761 #define TASK_PFA_SET(name, func) \
1762 static inline void task_set_##func(struct task_struct *p) \
1763 { \
1764 set_bit(PFA_##name, &p->atomic_flags); \
1765 }
1766
1767 #define TASK_PFA_CLEAR(name, func) \
1768 static inline void task_clear_##func(struct task_struct *p) \
1769 { \
1770 clear_bit(PFA_##name, &p->atomic_flags); \
1771 }
1772
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1773 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1774 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1775
1776 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1777 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1778 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1779
1780 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1781 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1782 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1783
1784 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1785 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1786 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1787
1788 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1789 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1790 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1791
1792 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1793 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1794
1795 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1796 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1797 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1798
1799 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1800 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1801
1802 static inline void current_restore_flags(unsigned long orig_flags, unsigned long flags)
1803 {
1804 current->flags &= ~flags;
1805 current->flags |= orig_flags & flags;
1806 }
1807
1808 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1809 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1810 #ifdef CONFIG_SMP
1811 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1812 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1813 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1814 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1815 {
1816 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1817 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1818 {
1819 if (!cpumask_test_cpu(0, new_mask)) {
1820 return -EINVAL;
1821 }
1822 return 0;
1823 }
1824 #endif
1825
1826 extern int yield_to(struct task_struct *p, bool preempt);
1827 extern void set_user_nice(struct task_struct *p, long nice);
1828 extern int task_prio(const struct task_struct *p);
1829
1830 /**
1831 * task_nice - return the nice value of a given task.
1832 * @p: the task in question.
1833 *
1834 * Return: The nice value [ -20 ... 0 ... 19 ].
1835 */
task_nice(const struct task_struct * p)1836 static inline int task_nice(const struct task_struct *p)
1837 {
1838 return PRIO_TO_NICE((p)->static_prio);
1839 }
1840
1841 extern int can_nice(const struct task_struct *p, const int nice);
1842 extern int task_curr(const struct task_struct *p);
1843 extern int idle_cpu(int cpu);
1844 extern int available_idle_cpu(int cpu);
1845 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1846 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1847 extern void sched_set_fifo(struct task_struct *p);
1848 extern void sched_set_fifo_low(struct task_struct *p);
1849 extern void sched_set_normal(struct task_struct *p, int nice);
1850 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1851 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1852 extern struct task_struct *idle_task(int cpu);
1853
1854 /**
1855 * is_idle_task - is the specified task an idle task?
1856 * @p: the task in question.
1857 *
1858 * Return: 1 if @p is an idle task. 0 otherwise.
1859 */
is_idle_task(const struct task_struct * p)1860 static __always_inline bool is_idle_task(const struct task_struct *p)
1861 {
1862 return !!(p->flags & PF_IDLE);
1863 }
1864
1865 extern struct task_struct *curr_task(int cpu);
1866 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1867
1868 void yield(void);
1869
1870 union thread_union {
1871 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1872 struct task_struct task;
1873 #endif
1874 #ifndef CONFIG_THREAD_INFO_IN_TASK
1875 struct thread_info thread_info;
1876 #endif
1877 unsigned long stack[THREAD_SIZE / sizeof(long)];
1878 };
1879
1880 #ifndef CONFIG_THREAD_INFO_IN_TASK
1881 extern struct thread_info init_thread_info;
1882 #endif
1883
1884 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1885
1886 #ifdef CONFIG_THREAD_INFO_IN_TASK
task_thread_info(struct task_struct * task)1887 static inline struct thread_info *task_thread_info(struct task_struct *task)
1888 {
1889 return &task->thread_info;
1890 }
1891 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1892 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
1893 #endif
1894
1895 /*
1896 * find a task by one of its numerical ids
1897 *
1898 * find_task_by_pid_ns():
1899 * finds a task by its pid in the specified namespace
1900 * find_task_by_vpid():
1901 * finds a task by its virtual pid
1902 *
1903 * see also find_vpid() etc in include/linux/pid.h
1904 */
1905
1906 extern struct task_struct *find_task_by_vpid(pid_t nr);
1907 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1908
1909 /*
1910 * find a task by its virtual pid and get the task struct
1911 */
1912 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1913
1914 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1915 extern int wake_up_process(struct task_struct *tsk);
1916 extern void wake_up_new_task(struct task_struct *tsk);
1917
1918 #ifdef CONFIG_SMP
1919 extern void kick_process(struct task_struct *tsk);
1920 #else
kick_process(struct task_struct * tsk)1921 static inline void kick_process(struct task_struct *tsk)
1922 {
1923 }
1924 #endif
1925
1926 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1927
set_task_comm(struct task_struct * tsk,const char * from)1928 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1929 {
1930 __set_task_comm(tsk, from, false);
1931 }
1932
1933 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1934 #define get_task_comm(buf, tsk) \
1935 ( { \
1936 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1937 __get_task_comm(buf, sizeof(buf), tsk); \
1938 })
1939
1940 #ifdef CONFIG_SMP
scheduler_ipi(void)1941 static __always_inline void scheduler_ipi(void)
1942 {
1943 /*
1944 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1945 * TIF_NEED_RESCHED remotely (for the first time) will also send
1946 * this IPI.
1947 */
1948 preempt_fold_need_resched();
1949 }
1950 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1951 #else
scheduler_ipi(void)1952 static inline void scheduler_ipi(void)
1953 {
1954 }
wait_task_inactive(struct task_struct * p,long match_state)1955 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1956 {
1957 return 1;
1958 }
1959 #endif
1960
1961 /*
1962 * Set thread flags in other task's structures.
1963 * See asm/thread_info.h for TIF_xxxx flags available:
1964 */
set_tsk_thread_flag(struct task_struct * tsk,int flag)1965 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1966 {
1967 set_ti_thread_flag(task_thread_info(tsk), flag);
1968 }
1969
clear_tsk_thread_flag(struct task_struct * tsk,int flag)1970 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1971 {
1972 clear_ti_thread_flag(task_thread_info(tsk), flag);
1973 }
1974
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)1975 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag, bool value)
1976 {
1977 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1978 }
1979
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)1980 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1981 {
1982 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1983 }
1984
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)1985 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1986 {
1987 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1988 }
1989
test_tsk_thread_flag(struct task_struct * tsk,int flag)1990 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1991 {
1992 return test_ti_thread_flag(task_thread_info(tsk), flag);
1993 }
1994
set_tsk_need_resched(struct task_struct * tsk)1995 static inline void set_tsk_need_resched(struct task_struct *tsk)
1996 {
1997 set_tsk_thread_flag(tsk, TIF_NEED_RESCHED);
1998 }
1999
clear_tsk_need_resched(struct task_struct * tsk)2000 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2001 {
2002 clear_tsk_thread_flag(tsk, TIF_NEED_RESCHED);
2003 }
2004
test_tsk_need_resched(struct task_struct * tsk)2005 static inline int test_tsk_need_resched(struct task_struct *tsk)
2006 {
2007 return unlikely(test_tsk_thread_flag(tsk, TIF_NEED_RESCHED));
2008 }
2009
2010 /*
2011 * cond_resched() and cond_resched_lock(): latency reduction via
2012 * explicit rescheduling in places that are safe. The return
2013 * value indicates whether a reschedule was done in fact.
2014 * cond_resched_lock() will drop the spinlock before scheduling,
2015 */
2016 #ifndef CONFIG_PREEMPTION
2017 extern int _cond_resched(void);
2018 #else
_cond_resched(void)2019 static inline int _cond_resched(void)
2020 {
2021 return 0;
2022 }
2023 #endif
2024
2025 #define cond_resched() \
2026 ( { \
2027 ___might_sleep(__FILE__, __LINE__, 0); \
2028 _cond_resched(); \
2029 })
2030
2031 extern int __cond_resched_lock(spinlock_t *lock);
2032
2033 #define cond_resched_lock(lock) \
2034 ( { \
2035 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2036 __cond_resched_lock(lock); \
2037 })
2038
cond_resched_rcu(void)2039 static inline void cond_resched_rcu(void)
2040 {
2041 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2042 rcu_read_unlock();
2043 cond_resched();
2044 rcu_read_lock();
2045 #endif
2046 }
2047
2048 /*
2049 * Does a critical section need to be broken due to another
2050 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2051 * but a general need for low latency)
2052 */
spin_needbreak(spinlock_t * lock)2053 static inline int spin_needbreak(spinlock_t *lock)
2054 {
2055 #ifdef CONFIG_PREEMPTION
2056 return spin_is_contended(lock);
2057 #else
2058 return 0;
2059 #endif
2060 }
2061
need_resched(void)2062 static __always_inline bool need_resched(void)
2063 {
2064 return unlikely(tif_need_resched());
2065 }
2066
2067 /*
2068 * Wrappers for p->thread_info->cpu access. No-op on UP.
2069 */
2070 #ifdef CONFIG_SMP
2071
task_cpu(const struct task_struct * p)2072 static inline unsigned int task_cpu(const struct task_struct *p)
2073 {
2074 #ifdef CONFIG_THREAD_INFO_IN_TASK
2075 return READ_ONCE(p->cpu);
2076 #else
2077 return READ_ONCE(task_thread_info(p)->cpu);
2078 #endif
2079 }
2080
2081 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2082
2083 #else
2084
task_cpu(const struct task_struct * p)2085 static inline unsigned int task_cpu(const struct task_struct *p)
2086 {
2087 return 0;
2088 }
2089
set_task_cpu(struct task_struct * p,unsigned int cpu)2090 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2091 {
2092 }
2093
2094 #endif /* CONFIG_SMP */
2095
2096 /*
2097 * In order to reduce various lock holder preemption latencies provide an
2098 * interface to see if a vCPU is currently running or not.
2099 *
2100 * This allows us to terminate optimistic spin loops and block, analogous to
2101 * the native optimistic spin heuristic of testing if the lock owner task is
2102 * running or not.
2103 */
2104 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2105 static inline bool vcpu_is_preempted(int cpu)
2106 {
2107 return false;
2108 }
2109 #endif
2110
2111 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2112 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2113
2114 #ifndef TASK_SIZE_OF
2115 #define TASK_SIZE_OF(tsk) TASK_SIZE
2116 #endif
2117
2118 #ifdef CONFIG_RSEQ
2119
2120 /*
2121 * Map the event mask on the user-space ABI enum rseq_cs_flags
2122 * for direct mask checks.
2123 */
2124 enum rseq_event_mask_bits {
2125 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2126 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2127 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2128 };
2129
2130 enum rseq_event_mask {
2131 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2132 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2133 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2134 };
2135
rseq_set_notify_resume(struct task_struct * t)2136 static inline void rseq_set_notify_resume(struct task_struct *t)
2137 {
2138 if (t->rseq) {
2139 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2140 }
2141 }
2142
2143 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2144
rseq_handle_notify_resume(struct ksignal * ksig,struct pt_regs * regs)2145 static inline void rseq_handle_notify_resume(struct ksignal *ksig, struct pt_regs *regs)
2146 {
2147 if (current->rseq) {
2148 __rseq_handle_notify_resume(ksig, regs);
2149 }
2150 }
2151
rseq_signal_deliver(struct ksignal * ksig,struct pt_regs * regs)2152 static inline void rseq_signal_deliver(struct ksignal *ksig, struct pt_regs *regs)
2153 {
2154 preempt_disable();
2155 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2156 preempt_enable();
2157 rseq_handle_notify_resume(ksig, regs);
2158 }
2159
2160 /* rseq_preempt() requires preemption to be disabled. */
rseq_preempt(struct task_struct * t)2161 static inline void rseq_preempt(struct task_struct *t)
2162 {
2163 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2164 rseq_set_notify_resume(t);
2165 }
2166
2167 /* rseq_migrate() requires preemption to be disabled. */
rseq_migrate(struct task_struct * t)2168 static inline void rseq_migrate(struct task_struct *t)
2169 {
2170 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2171 rseq_set_notify_resume(t);
2172 }
2173
2174 /*
2175 * If parent process has a registered restartable sequences area, the
2176 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2177 */
rseq_fork(struct task_struct * t,unsigned long clone_flags)2178 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2179 {
2180 if (clone_flags & CLONE_VM) {
2181 t->rseq = NULL;
2182 t->rseq_sig = 0;
2183 t->rseq_event_mask = 0;
2184 } else {
2185 t->rseq = current->rseq;
2186 t->rseq_sig = current->rseq_sig;
2187 t->rseq_event_mask = current->rseq_event_mask;
2188 }
2189 }
2190
rseq_execve(struct task_struct * t)2191 static inline void rseq_execve(struct task_struct *t)
2192 {
2193 t->rseq = NULL;
2194 t->rseq_sig = 0;
2195 t->rseq_event_mask = 0;
2196 }
2197
2198 #else
2199
rseq_set_notify_resume(struct task_struct * t)2200 static inline void rseq_set_notify_resume(struct task_struct *t)
2201 {
2202 }
rseq_handle_notify_resume(struct ksignal * ksig,struct pt_regs * regs)2203 static inline void rseq_handle_notify_resume(struct ksignal *ksig, struct pt_regs *regs)
2204 {
2205 }
rseq_signal_deliver(struct ksignal * ksig,struct pt_regs * regs)2206 static inline void rseq_signal_deliver(struct ksignal *ksig, struct pt_regs *regs)
2207 {
2208 }
rseq_preempt(struct task_struct * t)2209 static inline void rseq_preempt(struct task_struct *t)
2210 {
2211 }
rseq_migrate(struct task_struct * t)2212 static inline void rseq_migrate(struct task_struct *t)
2213 {
2214 }
rseq_fork(struct task_struct * t,unsigned long clone_flags)2215 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2216 {
2217 }
rseq_execve(struct task_struct * t)2218 static inline void rseq_execve(struct task_struct *t)
2219 {
2220 }
2221
2222 #endif
2223
2224 #ifdef CONFIG_DEBUG_RSEQ
2225
2226 void rseq_syscall(struct pt_regs *regs);
2227
2228 #else
2229
rseq_syscall(struct pt_regs * regs)2230 static inline void rseq_syscall(struct pt_regs *regs)
2231 {
2232 }
2233
2234 #endif
2235
2236 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2237 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2238 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2239
2240 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2241 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2242 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2243
2244 int sched_trace_rq_cpu(struct rq *rq);
2245 int sched_trace_rq_cpu_capacity(struct rq *rq);
2246 int sched_trace_rq_nr_running(struct rq *rq);
2247
2248 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2249
2250 #endif
2251