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#include <asm/processor.h>
14#include <linux/thread_info.h>
15#include <linux/preempt.h>
16#include <linux/cpumask.h>
17
18#include <linux/cache.h>
19#include <linux/irqflags_types.h>
20#include <linux/smp_types.h>
21#include <linux/pid_types.h>
22#include <linux/sem_types.h>
23#include <linux/shm.h>
24#include <linux/kmsan_types.h>
25#include <linux/mutex_types.h>
26#include <linux/plist_types.h>
27#include <linux/hrtimer_types.h>
28#include <linux/timer_types.h>
29#include <linux/seccomp_types.h>
30#include <linux/nodemask_types.h>
31#include <linux/refcount_types.h>
32#include <linux/resource.h>
33#include <linux/latencytop.h>
34#include <linux/sched/prio.h>
35#include <linux/sched/types.h>
36#include <linux/signal_types.h>
37#include <linux/syscall_user_dispatch_types.h>
38#include <linux/mm_types_task.h>
39#include <linux/task_io_accounting.h>
40#include <linux/posix-timers_types.h>
41#include <linux/restart_block.h>
42#include <uapi/linux/rseq.h>
43#include <linux/seqlock_types.h>
44#include <linux/kcsan.h>
45#include <linux/rv.h>
46#include <linux/livepatch_sched.h>
47#include <linux/uidgid_types.h>
48#include <asm/kmap_size.h>
49
50/* task_struct member predeclarations (sorted alphabetically): */
51struct audit_context;
52struct bio_list;
53struct blk_plug;
54struct bpf_local_storage;
55struct bpf_run_ctx;
56struct capture_control;
57struct cfs_rq;
58struct fs_struct;
59struct futex_pi_state;
60struct io_context;
61struct io_uring_task;
62struct mempolicy;
63struct nameidata;
64struct nsproxy;
65struct perf_event_context;
66struct pid_namespace;
67struct pipe_inode_info;
68struct rcu_node;
69struct reclaim_state;
70struct robust_list_head;
71struct root_domain;
72struct rq;
73struct sched_attr;
74struct sched_dl_entity;
75struct seq_file;
76struct sighand_struct;
77struct signal_struct;
78struct task_delay_info;
79struct task_group;
80struct task_struct;
81struct user_event_mm;
82
83/*
84 * Task state bitmask. NOTE! These bits are also
85 * encoded in fs/proc/array.c: get_task_state().
86 *
87 * We have two separate sets of flags: task->__state
88 * is about runnability, while task->exit_state are
89 * about the task exiting. Confusing, but this way
90 * modifying one set can't modify the other one by
91 * mistake.
92 */
93
94/* Used in tsk->__state: */
95#define TASK_RUNNING 0x00000000
96#define TASK_INTERRUPTIBLE 0x00000001
97#define TASK_UNINTERRUPTIBLE 0x00000002
98#define __TASK_STOPPED 0x00000004
99#define __TASK_TRACED 0x00000008
100/* Used in tsk->exit_state: */
101#define EXIT_DEAD 0x00000010
102#define EXIT_ZOMBIE 0x00000020
103#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
104/* Used in tsk->__state again: */
105#define TASK_PARKED 0x00000040
106#define TASK_DEAD 0x00000080
107#define TASK_WAKEKILL 0x00000100
108#define TASK_WAKING 0x00000200
109#define TASK_NOLOAD 0x00000400
110#define TASK_NEW 0x00000800
111#define TASK_RTLOCK_WAIT 0x00001000
112#define TASK_FREEZABLE 0x00002000
113#define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
114#define TASK_FROZEN 0x00008000
115#define TASK_STATE_MAX 0x00010000
116
117#define TASK_ANY (TASK_STATE_MAX-1)
118
119/*
120 * DO NOT ADD ANY NEW USERS !
121 */
122#define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
123
124/* Convenience macros for the sake of set_current_state: */
125#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
126#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
127#define TASK_TRACED __TASK_TRACED
128
129#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
130
131/* Convenience macros for the sake of wake_up(): */
132#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
133
134/* get_task_state(): */
135#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
136 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
137 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
138 TASK_PARKED)
139
140#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
141
142#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
143#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
144#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
145
146/*
147 * Special states are those that do not use the normal wait-loop pattern. See
148 * the comment with set_special_state().
149 */
150#define is_special_task_state(state) \
151 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
152
153#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
154# define debug_normal_state_change(state_value) \
155 do { \
156 WARN_ON_ONCE(is_special_task_state(state_value)); \
157 current->task_state_change = _THIS_IP_; \
158 } while (0)
159
160# define debug_special_state_change(state_value) \
161 do { \
162 WARN_ON_ONCE(!is_special_task_state(state_value)); \
163 current->task_state_change = _THIS_IP_; \
164 } while (0)
165
166# define debug_rtlock_wait_set_state() \
167 do { \
168 current->saved_state_change = current->task_state_change;\
169 current->task_state_change = _THIS_IP_; \
170 } while (0)
171
172# define debug_rtlock_wait_restore_state() \
173 do { \
174 current->task_state_change = current->saved_state_change;\
175 } while (0)
176
177#else
178# define debug_normal_state_change(cond) do { } while (0)
179# define debug_special_state_change(cond) do { } while (0)
180# define debug_rtlock_wait_set_state() do { } while (0)
181# define debug_rtlock_wait_restore_state() do { } while (0)
182#endif
183
184/*
185 * set_current_state() includes a barrier so that the write of current->__state
186 * is correctly serialised wrt the caller's subsequent test of whether to
187 * actually sleep:
188 *
189 * for (;;) {
190 * set_current_state(TASK_UNINTERRUPTIBLE);
191 * if (CONDITION)
192 * break;
193 *
194 * schedule();
195 * }
196 * __set_current_state(TASK_RUNNING);
197 *
198 * If the caller does not need such serialisation (because, for instance, the
199 * CONDITION test and condition change and wakeup are under the same lock) then
200 * use __set_current_state().
201 *
202 * The above is typically ordered against the wakeup, which does:
203 *
204 * CONDITION = 1;
205 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
206 *
207 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
208 * accessing p->__state.
209 *
210 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
211 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
212 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
213 *
214 * However, with slightly different timing the wakeup TASK_RUNNING store can
215 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
216 * a problem either because that will result in one extra go around the loop
217 * and our @cond test will save the day.
218 *
219 * Also see the comments of try_to_wake_up().
220 */
221#define __set_current_state(state_value) \
222 do { \
223 debug_normal_state_change((state_value)); \
224 WRITE_ONCE(current->__state, (state_value)); \
225 } while (0)
226
227#define set_current_state(state_value) \
228 do { \
229 debug_normal_state_change((state_value)); \
230 smp_store_mb(current->__state, (state_value)); \
231 } while (0)
232
233/*
234 * set_special_state() should be used for those states when the blocking task
235 * can not use the regular condition based wait-loop. In that case we must
236 * serialize against wakeups such that any possible in-flight TASK_RUNNING
237 * stores will not collide with our state change.
238 */
239#define set_special_state(state_value) \
240 do { \
241 unsigned long flags; /* may shadow */ \
242 \
243 raw_spin_lock_irqsave(&current->pi_lock, flags); \
244 debug_special_state_change((state_value)); \
245 WRITE_ONCE(current->__state, (state_value)); \
246 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
247 } while (0)
248
249/*
250 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
251 *
252 * RT's spin/rwlock substitutions are state preserving. The state of the
253 * task when blocking on the lock is saved in task_struct::saved_state and
254 * restored after the lock has been acquired. These operations are
255 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
256 * lock related wakeups while the task is blocked on the lock are
257 * redirected to operate on task_struct::saved_state to ensure that these
258 * are not dropped. On restore task_struct::saved_state is set to
259 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
260 *
261 * The lock operation looks like this:
262 *
263 * current_save_and_set_rtlock_wait_state();
264 * for (;;) {
265 * if (try_lock())
266 * break;
267 * raw_spin_unlock_irq(&lock->wait_lock);
268 * schedule_rtlock();
269 * raw_spin_lock_irq(&lock->wait_lock);
270 * set_current_state(TASK_RTLOCK_WAIT);
271 * }
272 * current_restore_rtlock_saved_state();
273 */
274#define current_save_and_set_rtlock_wait_state() \
275 do { \
276 lockdep_assert_irqs_disabled(); \
277 raw_spin_lock(&current->pi_lock); \
278 current->saved_state = current->__state; \
279 debug_rtlock_wait_set_state(); \
280 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
281 raw_spin_unlock(&current->pi_lock); \
282 } while (0);
283
284#define current_restore_rtlock_saved_state() \
285 do { \
286 lockdep_assert_irqs_disabled(); \
287 raw_spin_lock(&current->pi_lock); \
288 debug_rtlock_wait_restore_state(); \
289 WRITE_ONCE(current->__state, current->saved_state); \
290 current->saved_state = TASK_RUNNING; \
291 raw_spin_unlock(&current->pi_lock); \
292 } while (0);
293
294#define get_current_state() READ_ONCE(current->__state)
295
296/*
297 * Define the task command name length as enum, then it can be visible to
298 * BPF programs.
299 */
300enum {
301 TASK_COMM_LEN = 16,
302};
303
304extern void scheduler_tick(void);
305
306#define MAX_SCHEDULE_TIMEOUT LONG_MAX
307
308extern long schedule_timeout(long timeout);
309extern long schedule_timeout_interruptible(long timeout);
310extern long schedule_timeout_killable(long timeout);
311extern long schedule_timeout_uninterruptible(long timeout);
312extern long schedule_timeout_idle(long timeout);
313asmlinkage void schedule(void);
314extern void schedule_preempt_disabled(void);
315asmlinkage void preempt_schedule_irq(void);
316#ifdef CONFIG_PREEMPT_RT
317 extern void schedule_rtlock(void);
318#endif
319
320extern int __must_check io_schedule_prepare(void);
321extern void io_schedule_finish(int token);
322extern long io_schedule_timeout(long timeout);
323extern void io_schedule(void);
324
325/**
326 * struct prev_cputime - snapshot of system and user cputime
327 * @utime: time spent in user mode
328 * @stime: time spent in system mode
329 * @lock: protects the above two fields
330 *
331 * Stores previous user/system time values such that we can guarantee
332 * monotonicity.
333 */
334struct prev_cputime {
335#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
336 u64 utime;
337 u64 stime;
338 raw_spinlock_t lock;
339#endif
340};
341
342enum vtime_state {
343 /* Task is sleeping or running in a CPU with VTIME inactive: */
344 VTIME_INACTIVE = 0,
345 /* Task is idle */
346 VTIME_IDLE,
347 /* Task runs in kernelspace in a CPU with VTIME active: */
348 VTIME_SYS,
349 /* Task runs in userspace in a CPU with VTIME active: */
350 VTIME_USER,
351 /* Task runs as guests in a CPU with VTIME active: */
352 VTIME_GUEST,
353};
354
355struct vtime {
356 seqcount_t seqcount;
357 unsigned long long starttime;
358 enum vtime_state state;
359 unsigned int cpu;
360 u64 utime;
361 u64 stime;
362 u64 gtime;
363};
364
365/*
366 * Utilization clamp constraints.
367 * @UCLAMP_MIN: Minimum utilization
368 * @UCLAMP_MAX: Maximum utilization
369 * @UCLAMP_CNT: Utilization clamp constraints count
370 */
371enum uclamp_id {
372 UCLAMP_MIN = 0,
373 UCLAMP_MAX,
374 UCLAMP_CNT
375};
376
377#ifdef CONFIG_SMP
378extern struct root_domain def_root_domain;
379extern struct mutex sched_domains_mutex;
380#endif
381
382struct sched_param {
383 int sched_priority;
384};
385
386struct sched_info {
387#ifdef CONFIG_SCHED_INFO
388 /* Cumulative counters: */
389
390 /* # of times we have run on this CPU: */
391 unsigned long pcount;
392
393 /* Time spent waiting on a runqueue: */
394 unsigned long long run_delay;
395
396 /* Timestamps: */
397
398 /* When did we last run on a CPU? */
399 unsigned long long last_arrival;
400
401 /* When were we last queued to run? */
402 unsigned long long last_queued;
403
404#endif /* CONFIG_SCHED_INFO */
405};
406
407/*
408 * Integer metrics need fixed point arithmetic, e.g., sched/fair
409 * has a few: load, load_avg, util_avg, freq, and capacity.
410 *
411 * We define a basic fixed point arithmetic range, and then formalize
412 * all these metrics based on that basic range.
413 */
414# define SCHED_FIXEDPOINT_SHIFT 10
415# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
416
417/* Increase resolution of cpu_capacity calculations */
418# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
419# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
420
421struct load_weight {
422 unsigned long weight;
423 u32 inv_weight;
424};
425
426/*
427 * The load/runnable/util_avg accumulates an infinite geometric series
428 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
429 *
430 * [load_avg definition]
431 *
432 * load_avg = runnable% * scale_load_down(load)
433 *
434 * [runnable_avg definition]
435 *
436 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
437 *
438 * [util_avg definition]
439 *
440 * util_avg = running% * SCHED_CAPACITY_SCALE
441 *
442 * where runnable% is the time ratio that a sched_entity is runnable and
443 * running% the time ratio that a sched_entity is running.
444 *
445 * For cfs_rq, they are the aggregated values of all runnable and blocked
446 * sched_entities.
447 *
448 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
449 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
450 * for computing those signals (see update_rq_clock_pelt())
451 *
452 * N.B., the above ratios (runnable% and running%) themselves are in the
453 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
454 * to as large a range as necessary. This is for example reflected by
455 * util_avg's SCHED_CAPACITY_SCALE.
456 *
457 * [Overflow issue]
458 *
459 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
460 * with the highest load (=88761), always runnable on a single cfs_rq,
461 * and should not overflow as the number already hits PID_MAX_LIMIT.
462 *
463 * For all other cases (including 32-bit kernels), struct load_weight's
464 * weight will overflow first before we do, because:
465 *
466 * Max(load_avg) <= Max(load.weight)
467 *
468 * Then it is the load_weight's responsibility to consider overflow
469 * issues.
470 */
471struct sched_avg {
472 u64 last_update_time;
473 u64 load_sum;
474 u64 runnable_sum;
475 u32 util_sum;
476 u32 period_contrib;
477 unsigned long load_avg;
478 unsigned long runnable_avg;
479 unsigned long util_avg;
480 unsigned int util_est;
481} ____cacheline_aligned;
482
483/*
484 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
485 * updates. When a task is dequeued, its util_est should not be updated if its
486 * util_avg has not been updated in the meantime.
487 * This information is mapped into the MSB bit of util_est at dequeue time.
488 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
489 * it is safe to use MSB.
490 */
491#define UTIL_EST_WEIGHT_SHIFT 2
492#define UTIL_AVG_UNCHANGED 0x80000000
493
494struct sched_statistics {
495#ifdef CONFIG_SCHEDSTATS
496 u64 wait_start;
497 u64 wait_max;
498 u64 wait_count;
499 u64 wait_sum;
500 u64 iowait_count;
501 u64 iowait_sum;
502
503 u64 sleep_start;
504 u64 sleep_max;
505 s64 sum_sleep_runtime;
506
507 u64 block_start;
508 u64 block_max;
509 s64 sum_block_runtime;
510
511 s64 exec_max;
512 u64 slice_max;
513
514 u64 nr_migrations_cold;
515 u64 nr_failed_migrations_affine;
516 u64 nr_failed_migrations_running;
517 u64 nr_failed_migrations_hot;
518 u64 nr_forced_migrations;
519
520 u64 nr_wakeups;
521 u64 nr_wakeups_sync;
522 u64 nr_wakeups_migrate;
523 u64 nr_wakeups_local;
524 u64 nr_wakeups_remote;
525 u64 nr_wakeups_affine;
526 u64 nr_wakeups_affine_attempts;
527 u64 nr_wakeups_passive;
528 u64 nr_wakeups_idle;
529
530#ifdef CONFIG_SCHED_CORE
531 u64 core_forceidle_sum;
532#endif
533#endif /* CONFIG_SCHEDSTATS */
534} ____cacheline_aligned;
535
536struct sched_entity {
537 /* For load-balancing: */
538 struct load_weight load;
539 struct rb_node run_node;
540 u64 deadline;
541 u64 min_vruntime;
542
543 struct list_head group_node;
544 unsigned int on_rq;
545
546 u64 exec_start;
547 u64 sum_exec_runtime;
548 u64 prev_sum_exec_runtime;
549 u64 vruntime;
550 s64 vlag;
551 u64 slice;
552
553 u64 nr_migrations;
554
555#ifdef CONFIG_FAIR_GROUP_SCHED
556 int depth;
557 struct sched_entity *parent;
558 /* rq on which this entity is (to be) queued: */
559 struct cfs_rq *cfs_rq;
560 /* rq "owned" by this entity/group: */
561 struct cfs_rq *my_q;
562 /* cached value of my_q->h_nr_running */
563 unsigned long runnable_weight;
564#endif
565
566#ifdef CONFIG_SMP
567 /*
568 * Per entity load average tracking.
569 *
570 * Put into separate cache line so it does not
571 * collide with read-mostly values above.
572 */
573 struct sched_avg avg;
574#endif
575};
576
577struct sched_rt_entity {
578 struct list_head run_list;
579 unsigned long timeout;
580 unsigned long watchdog_stamp;
581 unsigned int time_slice;
582 unsigned short on_rq;
583 unsigned short on_list;
584
585 struct sched_rt_entity *back;
586#ifdef CONFIG_RT_GROUP_SCHED
587 struct sched_rt_entity *parent;
588 /* rq on which this entity is (to be) queued: */
589 struct rt_rq *rt_rq;
590 /* rq "owned" by this entity/group: */
591 struct rt_rq *my_q;
592#endif
593} __randomize_layout;
594
595typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
596typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
597
598struct sched_dl_entity {
599 struct rb_node rb_node;
600
601 /*
602 * Original scheduling parameters. Copied here from sched_attr
603 * during sched_setattr(), they will remain the same until
604 * the next sched_setattr().
605 */
606 u64 dl_runtime; /* Maximum runtime for each instance */
607 u64 dl_deadline; /* Relative deadline of each instance */
608 u64 dl_period; /* Separation of two instances (period) */
609 u64 dl_bw; /* dl_runtime / dl_period */
610 u64 dl_density; /* dl_runtime / dl_deadline */
611
612 /*
613 * Actual scheduling parameters. Initialized with the values above,
614 * they are continuously updated during task execution. Note that
615 * the remaining runtime could be < 0 in case we are in overrun.
616 */
617 s64 runtime; /* Remaining runtime for this instance */
618 u64 deadline; /* Absolute deadline for this instance */
619 unsigned int flags; /* Specifying the scheduler behaviour */
620
621 /*
622 * Some bool flags:
623 *
624 * @dl_throttled tells if we exhausted the runtime. If so, the
625 * task has to wait for a replenishment to be performed at the
626 * next firing of dl_timer.
627 *
628 * @dl_yielded tells if task gave up the CPU before consuming
629 * all its available runtime during the last job.
630 *
631 * @dl_non_contending tells if the task is inactive while still
632 * contributing to the active utilization. In other words, it
633 * indicates if the inactive timer has been armed and its handler
634 * has not been executed yet. This flag is useful to avoid race
635 * conditions between the inactive timer handler and the wakeup
636 * code.
637 *
638 * @dl_overrun tells if the task asked to be informed about runtime
639 * overruns.
640 */
641 unsigned int dl_throttled : 1;
642 unsigned int dl_yielded : 1;
643 unsigned int dl_non_contending : 1;
644 unsigned int dl_overrun : 1;
645 unsigned int dl_server : 1;
646
647 /*
648 * Bandwidth enforcement timer. Each -deadline task has its
649 * own bandwidth to be enforced, thus we need one timer per task.
650 */
651 struct hrtimer dl_timer;
652
653 /*
654 * Inactive timer, responsible for decreasing the active utilization
655 * at the "0-lag time". When a -deadline task blocks, it contributes
656 * to GRUB's active utilization until the "0-lag time", hence a
657 * timer is needed to decrease the active utilization at the correct
658 * time.
659 */
660 struct hrtimer inactive_timer;
661
662 /*
663 * Bits for DL-server functionality. Also see the comment near
664 * dl_server_update().
665 *
666 * @rq the runqueue this server is for
667 *
668 * @server_has_tasks() returns true if @server_pick return a
669 * runnable task.
670 */
671 struct rq *rq;
672 dl_server_has_tasks_f server_has_tasks;
673 dl_server_pick_f server_pick;
674
675#ifdef CONFIG_RT_MUTEXES
676 /*
677 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
678 * pi_se points to the donor, otherwise points to the dl_se it belongs
679 * to (the original one/itself).
680 */
681 struct sched_dl_entity *pi_se;
682#endif
683};
684
685#ifdef CONFIG_UCLAMP_TASK
686/* Number of utilization clamp buckets (shorter alias) */
687#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
688
689/*
690 * Utilization clamp for a scheduling entity
691 * @value: clamp value "assigned" to a se
692 * @bucket_id: bucket index corresponding to the "assigned" value
693 * @active: the se is currently refcounted in a rq's bucket
694 * @user_defined: the requested clamp value comes from user-space
695 *
696 * The bucket_id is the index of the clamp bucket matching the clamp value
697 * which is pre-computed and stored to avoid expensive integer divisions from
698 * the fast path.
699 *
700 * The active bit is set whenever a task has got an "effective" value assigned,
701 * which can be different from the clamp value "requested" from user-space.
702 * This allows to know a task is refcounted in the rq's bucket corresponding
703 * to the "effective" bucket_id.
704 *
705 * The user_defined bit is set whenever a task has got a task-specific clamp
706 * value requested from userspace, i.e. the system defaults apply to this task
707 * just as a restriction. This allows to relax default clamps when a less
708 * restrictive task-specific value has been requested, thus allowing to
709 * implement a "nice" semantic. For example, a task running with a 20%
710 * default boost can still drop its own boosting to 0%.
711 */
712struct uclamp_se {
713 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
714 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
715 unsigned int active : 1;
716 unsigned int user_defined : 1;
717};
718#endif /* CONFIG_UCLAMP_TASK */
719
720union rcu_special {
721 struct {
722 u8 blocked;
723 u8 need_qs;
724 u8 exp_hint; /* Hint for performance. */
725 u8 need_mb; /* Readers need smp_mb(). */
726 } b; /* Bits. */
727 u32 s; /* Set of bits. */
728};
729
730enum perf_event_task_context {
731 perf_invalid_context = -1,
732 perf_hw_context = 0,
733 perf_sw_context,
734 perf_nr_task_contexts,
735};
736
737struct wake_q_node {
738 struct wake_q_node *next;
739};
740
741struct kmap_ctrl {
742#ifdef CONFIG_KMAP_LOCAL
743 int idx;
744 pte_t pteval[KM_MAX_IDX];
745#endif
746};
747
748struct task_struct {
749#ifdef CONFIG_THREAD_INFO_IN_TASK
750 /*
751 * For reasons of header soup (see current_thread_info()), this
752 * must be the first element of task_struct.
753 */
754 struct thread_info thread_info;
755#endif
756 unsigned int __state;
757
758 /* saved state for "spinlock sleepers" */
759 unsigned int saved_state;
760
761 /*
762 * This begins the randomizable portion of task_struct. Only
763 * scheduling-critical items should be added above here.
764 */
765 randomized_struct_fields_start
766
767 void *stack;
768 refcount_t usage;
769 /* Per task flags (PF_*), defined further below: */
770 unsigned int flags;
771 unsigned int ptrace;
772
773#ifdef CONFIG_SMP
774 int on_cpu;
775 struct __call_single_node wake_entry;
776 unsigned int wakee_flips;
777 unsigned long wakee_flip_decay_ts;
778 struct task_struct *last_wakee;
779
780 /*
781 * recent_used_cpu is initially set as the last CPU used by a task
782 * that wakes affine another task. Waker/wakee relationships can
783 * push tasks around a CPU where each wakeup moves to the next one.
784 * Tracking a recently used CPU allows a quick search for a recently
785 * used CPU that may be idle.
786 */
787 int recent_used_cpu;
788 int wake_cpu;
789#endif
790 int on_rq;
791
792 int prio;
793 int static_prio;
794 int normal_prio;
795 unsigned int rt_priority;
796
797 struct sched_entity se;
798 struct sched_rt_entity rt;
799 struct sched_dl_entity dl;
800 struct sched_dl_entity *dl_server;
801 const struct sched_class *sched_class;
802
803#ifdef CONFIG_SCHED_CORE
804 struct rb_node core_node;
805 unsigned long core_cookie;
806 unsigned int core_occupation;
807#endif
808
809#ifdef CONFIG_CGROUP_SCHED
810 struct task_group *sched_task_group;
811#endif
812
813#ifdef CONFIG_UCLAMP_TASK
814 /*
815 * Clamp values requested for a scheduling entity.
816 * Must be updated with task_rq_lock() held.
817 */
818 struct uclamp_se uclamp_req[UCLAMP_CNT];
819 /*
820 * Effective clamp values used for a scheduling entity.
821 * Must be updated with task_rq_lock() held.
822 */
823 struct uclamp_se uclamp[UCLAMP_CNT];
824#endif
825
826 struct sched_statistics stats;
827
828#ifdef CONFIG_PREEMPT_NOTIFIERS
829 /* List of struct preempt_notifier: */
830 struct hlist_head preempt_notifiers;
831#endif
832
833#ifdef CONFIG_BLK_DEV_IO_TRACE
834 unsigned int btrace_seq;
835#endif
836
837 unsigned int policy;
838 int nr_cpus_allowed;
839 const cpumask_t *cpus_ptr;
840 cpumask_t *user_cpus_ptr;
841 cpumask_t cpus_mask;
842 void *migration_pending;
843#ifdef CONFIG_SMP
844 unsigned short migration_disabled;
845#endif
846 unsigned short migration_flags;
847
848#ifdef CONFIG_PREEMPT_RCU
849 int rcu_read_lock_nesting;
850 union rcu_special rcu_read_unlock_special;
851 struct list_head rcu_node_entry;
852 struct rcu_node *rcu_blocked_node;
853#endif /* #ifdef CONFIG_PREEMPT_RCU */
854
855#ifdef CONFIG_TASKS_RCU
856 unsigned long rcu_tasks_nvcsw;
857 u8 rcu_tasks_holdout;
858 u8 rcu_tasks_idx;
859 int rcu_tasks_idle_cpu;
860 struct list_head rcu_tasks_holdout_list;
861 int rcu_tasks_exit_cpu;
862 struct list_head rcu_tasks_exit_list;
863#endif /* #ifdef CONFIG_TASKS_RCU */
864
865#ifdef CONFIG_TASKS_TRACE_RCU
866 int trc_reader_nesting;
867 int trc_ipi_to_cpu;
868 union rcu_special trc_reader_special;
869 struct list_head trc_holdout_list;
870 struct list_head trc_blkd_node;
871 int trc_blkd_cpu;
872#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
873
874 struct sched_info sched_info;
875
876 struct list_head tasks;
877#ifdef CONFIG_SMP
878 struct plist_node pushable_tasks;
879 struct rb_node pushable_dl_tasks;
880#endif
881
882 struct mm_struct *mm;
883 struct mm_struct *active_mm;
884 struct address_space *faults_disabled_mapping;
885
886 int exit_state;
887 int exit_code;
888 int exit_signal;
889 /* The signal sent when the parent dies: */
890 int pdeath_signal;
891 /* JOBCTL_*, siglock protected: */
892 unsigned long jobctl;
893
894 /* Used for emulating ABI behavior of previous Linux versions: */
895 unsigned int personality;
896
897 /* Scheduler bits, serialized by scheduler locks: */
898 unsigned sched_reset_on_fork:1;
899 unsigned sched_contributes_to_load:1;
900 unsigned sched_migrated:1;
901
902 /* Force alignment to the next boundary: */
903 unsigned :0;
904
905 /* Unserialized, strictly 'current' */
906
907 /*
908 * This field must not be in the scheduler word above due to wakelist
909 * queueing no longer being serialized by p->on_cpu. However:
910 *
911 * p->XXX = X; ttwu()
912 * schedule() if (p->on_rq && ..) // false
913 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
914 * deactivate_task() ttwu_queue_wakelist())
915 * p->on_rq = 0; p->sched_remote_wakeup = Y;
916 *
917 * guarantees all stores of 'current' are visible before
918 * ->sched_remote_wakeup gets used, so it can be in this word.
919 */
920 unsigned sched_remote_wakeup:1;
921#ifdef CONFIG_RT_MUTEXES
922 unsigned sched_rt_mutex:1;
923#endif
924
925 /* Bit to tell TOMOYO we're in execve(): */
926 unsigned in_execve:1;
927 unsigned in_iowait:1;
928#ifndef TIF_RESTORE_SIGMASK
929 unsigned restore_sigmask:1;
930#endif
931#ifdef CONFIG_MEMCG
932 unsigned in_user_fault:1;
933#endif
934#ifdef CONFIG_LRU_GEN
935 /* whether the LRU algorithm may apply to this access */
936 unsigned in_lru_fault:1;
937#endif
938#ifdef CONFIG_COMPAT_BRK
939 unsigned brk_randomized:1;
940#endif
941#ifdef CONFIG_CGROUPS
942 /* disallow userland-initiated cgroup migration */
943 unsigned no_cgroup_migration:1;
944 /* task is frozen/stopped (used by the cgroup freezer) */
945 unsigned frozen:1;
946#endif
947#ifdef CONFIG_BLK_CGROUP
948 unsigned use_memdelay:1;
949#endif
950#ifdef CONFIG_PSI
951 /* Stalled due to lack of memory */
952 unsigned in_memstall:1;
953#endif
954#ifdef CONFIG_PAGE_OWNER
955 /* Used by page_owner=on to detect recursion in page tracking. */
956 unsigned in_page_owner:1;
957#endif
958#ifdef CONFIG_EVENTFD
959 /* Recursion prevention for eventfd_signal() */
960 unsigned in_eventfd:1;
961#endif
962#ifdef CONFIG_ARCH_HAS_CPU_PASID
963 unsigned pasid_activated:1;
964#endif
965#ifdef CONFIG_CPU_SUP_INTEL
966 unsigned reported_split_lock:1;
967#endif
968#ifdef CONFIG_TASK_DELAY_ACCT
969 /* delay due to memory thrashing */
970 unsigned in_thrashing:1;
971#endif
972
973 unsigned long atomic_flags; /* Flags requiring atomic access. */
974
975 struct restart_block restart_block;
976
977 pid_t pid;
978 pid_t tgid;
979
980#ifdef CONFIG_STACKPROTECTOR
981 /* Canary value for the -fstack-protector GCC feature: */
982 unsigned long stack_canary;
983#endif
984 /*
985 * Pointers to the (original) parent process, youngest child, younger sibling,
986 * older sibling, respectively. (p->father can be replaced with
987 * p->real_parent->pid)
988 */
989
990 /* Real parent process: */
991 struct task_struct __rcu *real_parent;
992
993 /* Recipient of SIGCHLD, wait4() reports: */
994 struct task_struct __rcu *parent;
995
996 /*
997 * Children/sibling form the list of natural children:
998 */
999 struct list_head children;
1000 struct list_head sibling;
1001 struct task_struct *group_leader;
1002
1003 /*
1004 * 'ptraced' is the list of tasks this task is using ptrace() on.
1005 *
1006 * This includes both natural children and PTRACE_ATTACH targets.
1007 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1008 */
1009 struct list_head ptraced;
1010 struct list_head ptrace_entry;
1011
1012 /* PID/PID hash table linkage. */
1013 struct pid *thread_pid;
1014 struct hlist_node pid_links[PIDTYPE_MAX];
1015 struct list_head thread_node;
1016
1017 struct completion *vfork_done;
1018
1019 /* CLONE_CHILD_SETTID: */
1020 int __user *set_child_tid;
1021
1022 /* CLONE_CHILD_CLEARTID: */
1023 int __user *clear_child_tid;
1024
1025 /* PF_KTHREAD | PF_IO_WORKER */
1026 void *worker_private;
1027
1028 u64 utime;
1029 u64 stime;
1030#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1031 u64 utimescaled;
1032 u64 stimescaled;
1033#endif
1034 u64 gtime;
1035 struct prev_cputime prev_cputime;
1036#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1037 struct vtime vtime;
1038#endif
1039
1040#ifdef CONFIG_NO_HZ_FULL
1041 atomic_t tick_dep_mask;
1042#endif
1043 /* Context switch counts: */
1044 unsigned long nvcsw;
1045 unsigned long nivcsw;
1046
1047 /* Monotonic time in nsecs: */
1048 u64 start_time;
1049
1050 /* Boot based time in nsecs: */
1051 u64 start_boottime;
1052
1053 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1054 unsigned long min_flt;
1055 unsigned long maj_flt;
1056
1057 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1058 struct posix_cputimers posix_cputimers;
1059
1060#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1061 struct posix_cputimers_work posix_cputimers_work;
1062#endif
1063
1064 /* Process credentials: */
1065
1066 /* Tracer's credentials at attach: */
1067 const struct cred __rcu *ptracer_cred;
1068
1069 /* Objective and real subjective task credentials (COW): */
1070 const struct cred __rcu *real_cred;
1071
1072 /* Effective (overridable) subjective task credentials (COW): */
1073 const struct cred __rcu *cred;
1074
1075#ifdef CONFIG_KEYS
1076 /* Cached requested key. */
1077 struct key *cached_requested_key;
1078#endif
1079
1080 /*
1081 * executable name, excluding path.
1082 *
1083 * - normally initialized setup_new_exec()
1084 * - access it with [gs]et_task_comm()
1085 * - lock it with task_lock()
1086 */
1087 char comm[TASK_COMM_LEN];
1088
1089 struct nameidata *nameidata;
1090
1091#ifdef CONFIG_SYSVIPC
1092 struct sysv_sem sysvsem;
1093 struct sysv_shm sysvshm;
1094#endif
1095#ifdef CONFIG_DETECT_HUNG_TASK
1096 unsigned long last_switch_count;
1097 unsigned long last_switch_time;
1098#endif
1099 /* Filesystem information: */
1100 struct fs_struct *fs;
1101
1102 /* Open file information: */
1103 struct files_struct *files;
1104
1105#ifdef CONFIG_IO_URING
1106 struct io_uring_task *io_uring;
1107#endif
1108
1109 /* Namespaces: */
1110 struct nsproxy *nsproxy;
1111
1112 /* Signal handlers: */
1113 struct signal_struct *signal;
1114 struct sighand_struct __rcu *sighand;
1115 sigset_t blocked;
1116 sigset_t real_blocked;
1117 /* Restored if set_restore_sigmask() was used: */
1118 sigset_t saved_sigmask;
1119 struct sigpending pending;
1120 unsigned long sas_ss_sp;
1121 size_t sas_ss_size;
1122 unsigned int sas_ss_flags;
1123
1124 struct callback_head *task_works;
1125
1126#ifdef CONFIG_AUDIT
1127#ifdef CONFIG_AUDITSYSCALL
1128 struct audit_context *audit_context;
1129#endif
1130 kuid_t loginuid;
1131 unsigned int sessionid;
1132#endif
1133 struct seccomp seccomp;
1134 struct syscall_user_dispatch syscall_dispatch;
1135
1136 /* Thread group tracking: */
1137 u64 parent_exec_id;
1138 u64 self_exec_id;
1139
1140 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1141 spinlock_t alloc_lock;
1142
1143 /* Protection of the PI data structures: */
1144 raw_spinlock_t pi_lock;
1145
1146 struct wake_q_node wake_q;
1147
1148#ifdef CONFIG_RT_MUTEXES
1149 /* PI waiters blocked on a rt_mutex held by this task: */
1150 struct rb_root_cached pi_waiters;
1151 /* Updated under owner's pi_lock and rq lock */
1152 struct task_struct *pi_top_task;
1153 /* Deadlock detection and priority inheritance handling: */
1154 struct rt_mutex_waiter *pi_blocked_on;
1155#endif
1156
1157#ifdef CONFIG_DEBUG_MUTEXES
1158 /* Mutex deadlock detection: */
1159 struct mutex_waiter *blocked_on;
1160#endif
1161
1162#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1163 int non_block_count;
1164#endif
1165
1166#ifdef CONFIG_TRACE_IRQFLAGS
1167 struct irqtrace_events irqtrace;
1168 unsigned int hardirq_threaded;
1169 u64 hardirq_chain_key;
1170 int softirqs_enabled;
1171 int softirq_context;
1172 int irq_config;
1173#endif
1174#ifdef CONFIG_PREEMPT_RT
1175 int softirq_disable_cnt;
1176#endif
1177
1178#ifdef CONFIG_LOCKDEP
1179# define MAX_LOCK_DEPTH 48UL
1180 u64 curr_chain_key;
1181 int lockdep_depth;
1182 unsigned int lockdep_recursion;
1183 struct held_lock held_locks[MAX_LOCK_DEPTH];
1184#endif
1185
1186#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1187 unsigned int in_ubsan;
1188#endif
1189
1190 /* Journalling filesystem info: */
1191 void *journal_info;
1192
1193 /* Stacked block device info: */
1194 struct bio_list *bio_list;
1195
1196 /* Stack plugging: */
1197 struct blk_plug *plug;
1198
1199 /* VM state: */
1200 struct reclaim_state *reclaim_state;
1201
1202 struct io_context *io_context;
1203
1204#ifdef CONFIG_COMPACTION
1205 struct capture_control *capture_control;
1206#endif
1207 /* Ptrace state: */
1208 unsigned long ptrace_message;
1209 kernel_siginfo_t *last_siginfo;
1210
1211 struct task_io_accounting ioac;
1212#ifdef CONFIG_PSI
1213 /* Pressure stall state */
1214 unsigned int psi_flags;
1215#endif
1216#ifdef CONFIG_TASK_XACCT
1217 /* Accumulated RSS usage: */
1218 u64 acct_rss_mem1;
1219 /* Accumulated virtual memory usage: */
1220 u64 acct_vm_mem1;
1221 /* stime + utime since last update: */
1222 u64 acct_timexpd;
1223#endif
1224#ifdef CONFIG_CPUSETS
1225 /* Protected by ->alloc_lock: */
1226 nodemask_t mems_allowed;
1227 /* Sequence number to catch updates: */
1228 seqcount_spinlock_t mems_allowed_seq;
1229 int cpuset_mem_spread_rotor;
1230 int cpuset_slab_spread_rotor;
1231#endif
1232#ifdef CONFIG_CGROUPS
1233 /* Control Group info protected by css_set_lock: */
1234 struct css_set __rcu *cgroups;
1235 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1236 struct list_head cg_list;
1237#endif
1238#ifdef CONFIG_X86_CPU_RESCTRL
1239 u32 closid;
1240 u32 rmid;
1241#endif
1242#ifdef CONFIG_FUTEX
1243 struct robust_list_head __user *robust_list;
1244#ifdef CONFIG_COMPAT
1245 struct compat_robust_list_head __user *compat_robust_list;
1246#endif
1247 struct list_head pi_state_list;
1248 struct futex_pi_state *pi_state_cache;
1249 struct mutex futex_exit_mutex;
1250 unsigned int futex_state;
1251#endif
1252#ifdef CONFIG_PERF_EVENTS
1253 struct perf_event_context *perf_event_ctxp;
1254 struct mutex perf_event_mutex;
1255 struct list_head perf_event_list;
1256#endif
1257#ifdef CONFIG_DEBUG_PREEMPT
1258 unsigned long preempt_disable_ip;
1259#endif
1260#ifdef CONFIG_NUMA
1261 /* Protected by alloc_lock: */
1262 struct mempolicy *mempolicy;
1263 short il_prev;
1264 u8 il_weight;
1265 short pref_node_fork;
1266#endif
1267#ifdef CONFIG_NUMA_BALANCING
1268 int numa_scan_seq;
1269 unsigned int numa_scan_period;
1270 unsigned int numa_scan_period_max;
1271 int numa_preferred_nid;
1272 unsigned long numa_migrate_retry;
1273 /* Migration stamp: */
1274 u64 node_stamp;
1275 u64 last_task_numa_placement;
1276 u64 last_sum_exec_runtime;
1277 struct callback_head numa_work;
1278
1279 /*
1280 * This pointer is only modified for current in syscall and
1281 * pagefault context (and for tasks being destroyed), so it can be read
1282 * from any of the following contexts:
1283 * - RCU read-side critical section
1284 * - current->numa_group from everywhere
1285 * - task's runqueue locked, task not running
1286 */
1287 struct numa_group __rcu *numa_group;
1288
1289 /*
1290 * numa_faults is an array split into four regions:
1291 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1292 * in this precise order.
1293 *
1294 * faults_memory: Exponential decaying average of faults on a per-node
1295 * basis. Scheduling placement decisions are made based on these
1296 * counts. The values remain static for the duration of a PTE scan.
1297 * faults_cpu: Track the nodes the process was running on when a NUMA
1298 * hinting fault was incurred.
1299 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1300 * during the current scan window. When the scan completes, the counts
1301 * in faults_memory and faults_cpu decay and these values are copied.
1302 */
1303 unsigned long *numa_faults;
1304 unsigned long total_numa_faults;
1305
1306 /*
1307 * numa_faults_locality tracks if faults recorded during the last
1308 * scan window were remote/local or failed to migrate. The task scan
1309 * period is adapted based on the locality of the faults with different
1310 * weights depending on whether they were shared or private faults
1311 */
1312 unsigned long numa_faults_locality[3];
1313
1314 unsigned long numa_pages_migrated;
1315#endif /* CONFIG_NUMA_BALANCING */
1316
1317#ifdef CONFIG_RSEQ
1318 struct rseq __user *rseq;
1319 u32 rseq_len;
1320 u32 rseq_sig;
1321 /*
1322 * RmW on rseq_event_mask must be performed atomically
1323 * with respect to preemption.
1324 */
1325 unsigned long rseq_event_mask;
1326#endif
1327
1328#ifdef CONFIG_SCHED_MM_CID
1329 int mm_cid; /* Current cid in mm */
1330 int last_mm_cid; /* Most recent cid in mm */
1331 int migrate_from_cpu;
1332 int mm_cid_active; /* Whether cid bitmap is active */
1333 struct callback_head cid_work;
1334#endif
1335
1336 struct tlbflush_unmap_batch tlb_ubc;
1337
1338 /* Cache last used pipe for splice(): */
1339 struct pipe_inode_info *splice_pipe;
1340
1341 struct page_frag task_frag;
1342
1343#ifdef CONFIG_TASK_DELAY_ACCT
1344 struct task_delay_info *delays;
1345#endif
1346
1347#ifdef CONFIG_FAULT_INJECTION
1348 int make_it_fail;
1349 unsigned int fail_nth;
1350#endif
1351 /*
1352 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1353 * balance_dirty_pages() for a dirty throttling pause:
1354 */
1355 int nr_dirtied;
1356 int nr_dirtied_pause;
1357 /* Start of a write-and-pause period: */
1358 unsigned long dirty_paused_when;
1359
1360#ifdef CONFIG_LATENCYTOP
1361 int latency_record_count;
1362 struct latency_record latency_record[LT_SAVECOUNT];
1363#endif
1364 /*
1365 * Time slack values; these are used to round up poll() and
1366 * select() etc timeout values. These are in nanoseconds.
1367 */
1368 u64 timer_slack_ns;
1369 u64 default_timer_slack_ns;
1370
1371#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1372 unsigned int kasan_depth;
1373#endif
1374
1375#ifdef CONFIG_KCSAN
1376 struct kcsan_ctx kcsan_ctx;
1377#ifdef CONFIG_TRACE_IRQFLAGS
1378 struct irqtrace_events kcsan_save_irqtrace;
1379#endif
1380#ifdef CONFIG_KCSAN_WEAK_MEMORY
1381 int kcsan_stack_depth;
1382#endif
1383#endif
1384
1385#ifdef CONFIG_KMSAN
1386 struct kmsan_ctx kmsan_ctx;
1387#endif
1388
1389#if IS_ENABLED(CONFIG_KUNIT)
1390 struct kunit *kunit_test;
1391#endif
1392
1393#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1394 /* Index of current stored address in ret_stack: */
1395 int curr_ret_stack;
1396 int curr_ret_depth;
1397
1398 /* Stack of return addresses for return function tracing: */
1399 struct ftrace_ret_stack *ret_stack;
1400
1401 /* Timestamp for last schedule: */
1402 unsigned long long ftrace_timestamp;
1403
1404 /*
1405 * Number of functions that haven't been traced
1406 * because of depth overrun:
1407 */
1408 atomic_t trace_overrun;
1409
1410 /* Pause tracing: */
1411 atomic_t tracing_graph_pause;
1412#endif
1413
1414#ifdef CONFIG_TRACING
1415 /* Bitmask and counter of trace recursion: */
1416 unsigned long trace_recursion;
1417#endif /* CONFIG_TRACING */
1418
1419#ifdef CONFIG_KCOV
1420 /* See kernel/kcov.c for more details. */
1421
1422 /* Coverage collection mode enabled for this task (0 if disabled): */
1423 unsigned int kcov_mode;
1424
1425 /* Size of the kcov_area: */
1426 unsigned int kcov_size;
1427
1428 /* Buffer for coverage collection: */
1429 void *kcov_area;
1430
1431 /* KCOV descriptor wired with this task or NULL: */
1432 struct kcov *kcov;
1433
1434 /* KCOV common handle for remote coverage collection: */
1435 u64 kcov_handle;
1436
1437 /* KCOV sequence number: */
1438 int kcov_sequence;
1439
1440 /* Collect coverage from softirq context: */
1441 unsigned int kcov_softirq;
1442#endif
1443
1444#ifdef CONFIG_MEMCG
1445 struct mem_cgroup *memcg_in_oom;
1446 gfp_t memcg_oom_gfp_mask;
1447 int memcg_oom_order;
1448
1449 /* Number of pages to reclaim on returning to userland: */
1450 unsigned int memcg_nr_pages_over_high;
1451
1452 /* Used by memcontrol for targeted memcg charge: */
1453 struct mem_cgroup *active_memcg;
1454#endif
1455
1456#ifdef CONFIG_MEMCG_KMEM
1457 struct obj_cgroup *objcg;
1458#endif
1459
1460#ifdef CONFIG_BLK_CGROUP
1461 struct gendisk *throttle_disk;
1462#endif
1463
1464#ifdef CONFIG_UPROBES
1465 struct uprobe_task *utask;
1466#endif
1467#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1468 unsigned int sequential_io;
1469 unsigned int sequential_io_avg;
1470#endif
1471 struct kmap_ctrl kmap_ctrl;
1472#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1473 unsigned long task_state_change;
1474# ifdef CONFIG_PREEMPT_RT
1475 unsigned long saved_state_change;
1476# endif
1477#endif
1478 struct rcu_head rcu;
1479 refcount_t rcu_users;
1480 int pagefault_disabled;
1481#ifdef CONFIG_MMU
1482 struct task_struct *oom_reaper_list;
1483 struct timer_list oom_reaper_timer;
1484#endif
1485#ifdef CONFIG_VMAP_STACK
1486 struct vm_struct *stack_vm_area;
1487#endif
1488#ifdef CONFIG_THREAD_INFO_IN_TASK
1489 /* A live task holds one reference: */
1490 refcount_t stack_refcount;
1491#endif
1492#ifdef CONFIG_LIVEPATCH
1493 int patch_state;
1494#endif
1495#ifdef CONFIG_SECURITY
1496 /* Used by LSM modules for access restriction: */
1497 void *security;
1498#endif
1499#ifdef CONFIG_BPF_SYSCALL
1500 /* Used by BPF task local storage */
1501 struct bpf_local_storage __rcu *bpf_storage;
1502 /* Used for BPF run context */
1503 struct bpf_run_ctx *bpf_ctx;
1504#endif
1505
1506#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1507 unsigned long lowest_stack;
1508 unsigned long prev_lowest_stack;
1509#endif
1510
1511#ifdef CONFIG_X86_MCE
1512 void __user *mce_vaddr;
1513 __u64 mce_kflags;
1514 u64 mce_addr;
1515 __u64 mce_ripv : 1,
1516 mce_whole_page : 1,
1517 __mce_reserved : 62;
1518 struct callback_head mce_kill_me;
1519 int mce_count;
1520#endif
1521
1522#ifdef CONFIG_KRETPROBES
1523 struct llist_head kretprobe_instances;
1524#endif
1525#ifdef CONFIG_RETHOOK
1526 struct llist_head rethooks;
1527#endif
1528
1529#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1530 /*
1531 * If L1D flush is supported on mm context switch
1532 * then we use this callback head to queue kill work
1533 * to kill tasks that are not running on SMT disabled
1534 * cores
1535 */
1536 struct callback_head l1d_flush_kill;
1537#endif
1538
1539#ifdef CONFIG_RV
1540 /*
1541 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1542 * If we find justification for more monitors, we can think
1543 * about adding more or developing a dynamic method. So far,
1544 * none of these are justified.
1545 */
1546 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1547#endif
1548
1549#ifdef CONFIG_USER_EVENTS
1550 struct user_event_mm *user_event_mm;
1551#endif
1552
1553 /*
1554 * New fields for task_struct should be added above here, so that
1555 * they are included in the randomized portion of task_struct.
1556 */
1557 randomized_struct_fields_end
1558
1559 /* CPU-specific state of this task: */
1560 struct thread_struct thread;
1561
1562 /*
1563 * WARNING: on x86, 'thread_struct' contains a variable-sized
1564 * structure. It *MUST* be at the end of 'task_struct'.
1565 *
1566 * Do not put anything below here!
1567 */
1568};
1569
1570#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1571#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1572
1573static inline unsigned int __task_state_index(unsigned int tsk_state,
1574 unsigned int tsk_exit_state)
1575{
1576 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1577
1578 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1579
1580 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1581 state = TASK_REPORT_IDLE;
1582
1583 /*
1584 * We're lying here, but rather than expose a completely new task state
1585 * to userspace, we can make this appear as if the task has gone through
1586 * a regular rt_mutex_lock() call.
1587 */
1588 if (tsk_state & TASK_RTLOCK_WAIT)
1589 state = TASK_UNINTERRUPTIBLE;
1590
1591 return fls(x: state);
1592}
1593
1594static inline unsigned int task_state_index(struct task_struct *tsk)
1595{
1596 return __task_state_index(READ_ONCE(tsk->__state), tsk_exit_state: tsk->exit_state);
1597}
1598
1599static inline char task_index_to_char(unsigned int state)
1600{
1601 static const char state_char[] = "RSDTtXZPI";
1602
1603 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1604
1605 return state_char[state];
1606}
1607
1608static inline char task_state_to_char(struct task_struct *tsk)
1609{
1610 return task_index_to_char(state: task_state_index(tsk));
1611}
1612
1613extern struct pid *cad_pid;
1614
1615/*
1616 * Per process flags
1617 */
1618#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1619#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1620#define PF_EXITING 0x00000004 /* Getting shut down */
1621#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1622#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1623#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1624#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1625#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1626#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1627#define PF_DUMPCORE 0x00000200 /* Dumped core */
1628#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1629#define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1630#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1631#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1632#define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1633#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1634#define PF__HOLE__00010000 0x00010000
1635#define PF_KSWAPD 0x00020000 /* I am kswapd */
1636#define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1637#define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1638#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1639 * I am cleaning dirty pages from some other bdi. */
1640#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1641#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1642#define PF_MEMALLOC_NORECLAIM 0x00800000 /* All allocation requests will clear __GFP_DIRECT_RECLAIM */
1643#define PF_MEMALLOC_NOWARN 0x01000000 /* All allocation requests will inherit __GFP_NOWARN */
1644#define PF__HOLE__02000000 0x02000000
1645#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1646#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1647#define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1648 * See memalloc_pin_save() */
1649#define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1650#define PF__HOLE__40000000 0x40000000
1651#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1652
1653/*
1654 * Only the _current_ task can read/write to tsk->flags, but other
1655 * tasks can access tsk->flags in readonly mode for example
1656 * with tsk_used_math (like during threaded core dumping).
1657 * There is however an exception to this rule during ptrace
1658 * or during fork: the ptracer task is allowed to write to the
1659 * child->flags of its traced child (same goes for fork, the parent
1660 * can write to the child->flags), because we're guaranteed the
1661 * child is not running and in turn not changing child->flags
1662 * at the same time the parent does it.
1663 */
1664#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1665#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1666#define clear_used_math() clear_stopped_child_used_math(current)
1667#define set_used_math() set_stopped_child_used_math(current)
1668
1669#define conditional_stopped_child_used_math(condition, child) \
1670 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1671
1672#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1673
1674#define copy_to_stopped_child_used_math(child) \
1675 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1676
1677/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1678#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1679#define used_math() tsk_used_math(current)
1680
1681static __always_inline bool is_percpu_thread(void)
1682{
1683#ifdef CONFIG_SMP
1684 return (current->flags & PF_NO_SETAFFINITY) &&
1685 (current->nr_cpus_allowed == 1);
1686#else
1687 return true;
1688#endif
1689}
1690
1691/* Per-process atomic flags. */
1692#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1693#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1694#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1695#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1696#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1697#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1698#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1699#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1700
1701#define TASK_PFA_TEST(name, func) \
1702 static inline bool task_##func(struct task_struct *p) \
1703 { return test_bit(PFA_##name, &p->atomic_flags); }
1704
1705#define TASK_PFA_SET(name, func) \
1706 static inline void task_set_##func(struct task_struct *p) \
1707 { set_bit(PFA_##name, &p->atomic_flags); }
1708
1709#define TASK_PFA_CLEAR(name, func) \
1710 static inline void task_clear_##func(struct task_struct *p) \
1711 { clear_bit(PFA_##name, &p->atomic_flags); }
1712
1713TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1714TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1715
1716TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1717TASK_PFA_SET(SPREAD_PAGE, spread_page)
1718TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1719
1720TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1721TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1722TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1723
1724TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1725TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1726TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1727
1728TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1729TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1730TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1731
1732TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1733TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1734
1735TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1736TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1737TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1738
1739TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1740TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1741
1742static inline void
1743current_restore_flags(unsigned long orig_flags, unsigned long flags)
1744{
1745 current->flags &= ~flags;
1746 current->flags |= orig_flags & flags;
1747}
1748
1749extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1750extern int task_can_attach(struct task_struct *p);
1751extern int dl_bw_alloc(int cpu, u64 dl_bw);
1752extern void dl_bw_free(int cpu, u64 dl_bw);
1753#ifdef CONFIG_SMP
1754
1755/* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1756extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1757
1758/**
1759 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1760 * @p: the task
1761 * @new_mask: CPU affinity mask
1762 *
1763 * Return: zero if successful, or a negative error code
1764 */
1765extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1766extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1767extern void release_user_cpus_ptr(struct task_struct *p);
1768extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1769extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1770extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1771#else
1772static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1773{
1774}
1775static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1776{
1777 if (!cpumask_test_cpu(0, new_mask))
1778 return -EINVAL;
1779 return 0;
1780}
1781static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1782{
1783 if (src->user_cpus_ptr)
1784 return -EINVAL;
1785 return 0;
1786}
1787static inline void release_user_cpus_ptr(struct task_struct *p)
1788{
1789 WARN_ON(p->user_cpus_ptr);
1790}
1791
1792static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1793{
1794 return 0;
1795}
1796#endif
1797
1798extern int yield_to(struct task_struct *p, bool preempt);
1799extern void set_user_nice(struct task_struct *p, long nice);
1800extern int task_prio(const struct task_struct *p);
1801
1802/**
1803 * task_nice - return the nice value of a given task.
1804 * @p: the task in question.
1805 *
1806 * Return: The nice value [ -20 ... 0 ... 19 ].
1807 */
1808static inline int task_nice(const struct task_struct *p)
1809{
1810 return PRIO_TO_NICE((p)->static_prio);
1811}
1812
1813extern int can_nice(const struct task_struct *p, const int nice);
1814extern int task_curr(const struct task_struct *p);
1815extern int idle_cpu(int cpu);
1816extern int available_idle_cpu(int cpu);
1817extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1818extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1819extern void sched_set_fifo(struct task_struct *p);
1820extern void sched_set_fifo_low(struct task_struct *p);
1821extern void sched_set_normal(struct task_struct *p, int nice);
1822extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1823extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1824extern struct task_struct *idle_task(int cpu);
1825
1826/**
1827 * is_idle_task - is the specified task an idle task?
1828 * @p: the task in question.
1829 *
1830 * Return: 1 if @p is an idle task. 0 otherwise.
1831 */
1832static __always_inline bool is_idle_task(const struct task_struct *p)
1833{
1834 return !!(p->flags & PF_IDLE);
1835}
1836
1837extern struct task_struct *curr_task(int cpu);
1838extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1839
1840void yield(void);
1841
1842union thread_union {
1843 struct task_struct task;
1844#ifndef CONFIG_THREAD_INFO_IN_TASK
1845 struct thread_info thread_info;
1846#endif
1847 unsigned long stack[THREAD_SIZE/sizeof(long)];
1848};
1849
1850#ifndef CONFIG_THREAD_INFO_IN_TASK
1851extern struct thread_info init_thread_info;
1852#endif
1853
1854extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1855
1856#ifdef CONFIG_THREAD_INFO_IN_TASK
1857# define task_thread_info(task) (&(task)->thread_info)
1858#elif !defined(__HAVE_THREAD_FUNCTIONS)
1859# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1860#endif
1861
1862/*
1863 * find a task by one of its numerical ids
1864 *
1865 * find_task_by_pid_ns():
1866 * finds a task by its pid in the specified namespace
1867 * find_task_by_vpid():
1868 * finds a task by its virtual pid
1869 *
1870 * see also find_vpid() etc in include/linux/pid.h
1871 */
1872
1873extern struct task_struct *find_task_by_vpid(pid_t nr);
1874extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1875
1876/*
1877 * find a task by its virtual pid and get the task struct
1878 */
1879extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1880
1881extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1882extern int wake_up_process(struct task_struct *tsk);
1883extern void wake_up_new_task(struct task_struct *tsk);
1884
1885#ifdef CONFIG_SMP
1886extern void kick_process(struct task_struct *tsk);
1887#else
1888static inline void kick_process(struct task_struct *tsk) { }
1889#endif
1890
1891extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1892
1893static inline void set_task_comm(struct task_struct *tsk, const char *from)
1894{
1895 __set_task_comm(tsk, from, exec: false);
1896}
1897
1898extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1899#define get_task_comm(buf, tsk) ({ \
1900 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1901 __get_task_comm(buf, sizeof(buf), tsk); \
1902})
1903
1904#ifdef CONFIG_SMP
1905static __always_inline void scheduler_ipi(void)
1906{
1907 /*
1908 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1909 * TIF_NEED_RESCHED remotely (for the first time) will also send
1910 * this IPI.
1911 */
1912 preempt_fold_need_resched();
1913}
1914#else
1915static inline void scheduler_ipi(void) { }
1916#endif
1917
1918extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1919
1920/*
1921 * Set thread flags in other task's structures.
1922 * See asm/thread_info.h for TIF_xxxx flags available:
1923 */
1924static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1925{
1926 set_ti_thread_flag(task_thread_info(tsk), flag);
1927}
1928
1929static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1930{
1931 clear_ti_thread_flag(task_thread_info(tsk), flag);
1932}
1933
1934static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1935 bool value)
1936{
1937 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1938}
1939
1940static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1941{
1942 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1943}
1944
1945static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1946{
1947 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1948}
1949
1950static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1951{
1952 return test_ti_thread_flag(task_thread_info(tsk), flag);
1953}
1954
1955static inline void set_tsk_need_resched(struct task_struct *tsk)
1956{
1957 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1958}
1959
1960static inline void clear_tsk_need_resched(struct task_struct *tsk)
1961{
1962 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1963}
1964
1965static inline int test_tsk_need_resched(struct task_struct *tsk)
1966{
1967 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1968}
1969
1970/*
1971 * cond_resched() and cond_resched_lock(): latency reduction via
1972 * explicit rescheduling in places that are safe. The return
1973 * value indicates whether a reschedule was done in fact.
1974 * cond_resched_lock() will drop the spinlock before scheduling,
1975 */
1976#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1977extern int __cond_resched(void);
1978
1979#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
1980
1981void sched_dynamic_klp_enable(void);
1982void sched_dynamic_klp_disable(void);
1983
1984DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1985
1986static __always_inline int _cond_resched(void)
1987{
1988 return static_call_mod(cond_resched)();
1989}
1990
1991#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
1992
1993extern int dynamic_cond_resched(void);
1994
1995static __always_inline int _cond_resched(void)
1996{
1997 return dynamic_cond_resched();
1998}
1999
2000#else /* !CONFIG_PREEMPTION */
2001
2002static inline int _cond_resched(void)
2003{
2004 klp_sched_try_switch();
2005 return __cond_resched();
2006}
2007
2008#endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2009
2010#else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2011
2012static inline int _cond_resched(void)
2013{
2014 klp_sched_try_switch();
2015 return 0;
2016}
2017
2018#endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2019
2020#define cond_resched() ({ \
2021 __might_resched(__FILE__, __LINE__, 0); \
2022 _cond_resched(); \
2023})
2024
2025extern int __cond_resched_lock(spinlock_t *lock);
2026extern int __cond_resched_rwlock_read(rwlock_t *lock);
2027extern int __cond_resched_rwlock_write(rwlock_t *lock);
2028
2029#define MIGHT_RESCHED_RCU_SHIFT 8
2030#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2031
2032#ifndef CONFIG_PREEMPT_RT
2033/*
2034 * Non RT kernels have an elevated preempt count due to the held lock,
2035 * but are not allowed to be inside a RCU read side critical section
2036 */
2037# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2038#else
2039/*
2040 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2041 * cond_resched*lock() has to take that into account because it checks for
2042 * preempt_count() and rcu_preempt_depth().
2043 */
2044# define PREEMPT_LOCK_RESCHED_OFFSETS \
2045 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2046#endif
2047
2048#define cond_resched_lock(lock) ({ \
2049 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2050 __cond_resched_lock(lock); \
2051})
2052
2053#define cond_resched_rwlock_read(lock) ({ \
2054 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2055 __cond_resched_rwlock_read(lock); \
2056})
2057
2058#define cond_resched_rwlock_write(lock) ({ \
2059 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2060 __cond_resched_rwlock_write(lock); \
2061})
2062
2063#ifdef CONFIG_PREEMPT_DYNAMIC
2064
2065extern bool preempt_model_none(void);
2066extern bool preempt_model_voluntary(void);
2067extern bool preempt_model_full(void);
2068
2069#else
2070
2071static inline bool preempt_model_none(void)
2072{
2073 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2074}
2075static inline bool preempt_model_voluntary(void)
2076{
2077 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2078}
2079static inline bool preempt_model_full(void)
2080{
2081 return IS_ENABLED(CONFIG_PREEMPT);
2082}
2083
2084#endif
2085
2086static inline bool preempt_model_rt(void)
2087{
2088 return IS_ENABLED(CONFIG_PREEMPT_RT);
2089}
2090
2091/*
2092 * Does the preemption model allow non-cooperative preemption?
2093 *
2094 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2095 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2096 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2097 * PREEMPT_NONE model.
2098 */
2099static inline bool preempt_model_preemptible(void)
2100{
2101 return preempt_model_full() || preempt_model_rt();
2102}
2103
2104static __always_inline bool need_resched(void)
2105{
2106 return unlikely(tif_need_resched());
2107}
2108
2109/*
2110 * Wrappers for p->thread_info->cpu access. No-op on UP.
2111 */
2112#ifdef CONFIG_SMP
2113
2114static inline unsigned int task_cpu(const struct task_struct *p)
2115{
2116 return READ_ONCE(task_thread_info(p)->cpu);
2117}
2118
2119extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2120
2121#else
2122
2123static inline unsigned int task_cpu(const struct task_struct *p)
2124{
2125 return 0;
2126}
2127
2128static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2129{
2130}
2131
2132#endif /* CONFIG_SMP */
2133
2134extern bool sched_task_on_rq(struct task_struct *p);
2135extern unsigned long get_wchan(struct task_struct *p);
2136extern struct task_struct *cpu_curr_snapshot(int cpu);
2137
2138#include <linux/spinlock.h>
2139
2140/*
2141 * In order to reduce various lock holder preemption latencies provide an
2142 * interface to see if a vCPU is currently running or not.
2143 *
2144 * This allows us to terminate optimistic spin loops and block, analogous to
2145 * the native optimistic spin heuristic of testing if the lock owner task is
2146 * running or not.
2147 */
2148#ifndef vcpu_is_preempted
2149static inline bool vcpu_is_preempted(int cpu)
2150{
2151 return false;
2152}
2153#endif
2154
2155extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2156extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2157
2158#ifndef TASK_SIZE_OF
2159#define TASK_SIZE_OF(tsk) TASK_SIZE
2160#endif
2161
2162#ifdef CONFIG_SMP
2163static inline bool owner_on_cpu(struct task_struct *owner)
2164{
2165 /*
2166 * As lock holder preemption issue, we both skip spinning if
2167 * task is not on cpu or its cpu is preempted
2168 */
2169 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(cpu: task_cpu(p: owner));
2170}
2171
2172/* Returns effective CPU energy utilization, as seen by the scheduler */
2173unsigned long sched_cpu_util(int cpu);
2174#endif /* CONFIG_SMP */
2175
2176#ifdef CONFIG_SCHED_CORE
2177extern void sched_core_free(struct task_struct *tsk);
2178extern void sched_core_fork(struct task_struct *p);
2179extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2180 unsigned long uaddr);
2181extern int sched_core_idle_cpu(int cpu);
2182#else
2183static inline void sched_core_free(struct task_struct *tsk) { }
2184static inline void sched_core_fork(struct task_struct *p) { }
2185static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2186#endif
2187
2188extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2189
2190#endif
2191

source code of linux/include/linux/sched.h