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/mutex.h>
18#include <linux/plist.h>
19#include <linux/hrtimer.h>
20#include <linux/irqflags.h>
21#include <linux/seccomp.h>
22#include <linux/nodemask.h>
23#include <linux/rcupdate.h>
24#include <linux/refcount.h>
25#include <linux/resource.h>
26#include <linux/latencytop.h>
27#include <linux/sched/prio.h>
28#include <linux/sched/types.h>
29#include <linux/signal_types.h>
30#include <linux/syscall_user_dispatch.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/rv.h>
38#include <asm/kmap_size.h>
39
40/* task_struct member predeclarations (sorted alphabetically): */
41struct audit_context;
42struct backing_dev_info;
43struct bio_list;
44struct blk_plug;
45struct bpf_local_storage;
46struct bpf_run_ctx;
47struct capture_control;
48struct cfs_rq;
49struct fs_struct;
50struct futex_pi_state;
51struct io_context;
52struct io_uring_task;
53struct mempolicy;
54struct nameidata;
55struct nsproxy;
56struct perf_event_context;
57struct pid_namespace;
58struct pipe_inode_info;
59struct rcu_node;
60struct reclaim_state;
61struct robust_list_head;
62struct root_domain;
63struct rq;
64struct sched_attr;
65struct sched_param;
66struct seq_file;
67struct sighand_struct;
68struct signal_struct;
69struct task_delay_info;
70struct task_group;
71
72/*
73 * Task state bitmask. NOTE! These bits are also
74 * encoded in fs/proc/array.c: get_task_state().
75 *
76 * We have two separate sets of flags: task->state
77 * is about runnability, while task->exit_state are
78 * about the task exiting. Confusing, but this way
79 * modifying one set can't modify the other one by
80 * mistake.
81 */
82
83/* Used in tsk->state: */
84#define TASK_RUNNING 0x0000
85#define TASK_INTERRUPTIBLE 0x0001
86#define TASK_UNINTERRUPTIBLE 0x0002
87#define __TASK_STOPPED 0x0004
88#define __TASK_TRACED 0x0008
89/* Used in tsk->exit_state: */
90#define EXIT_DEAD 0x0010
91#define EXIT_ZOMBIE 0x0020
92#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
93/* Used in tsk->state again: */
94#define TASK_PARKED 0x0040
95#define TASK_DEAD 0x0080
96#define TASK_WAKEKILL 0x0100
97#define TASK_WAKING 0x0200
98#define TASK_NOLOAD 0x0400
99#define TASK_NEW 0x0800
100/* RT specific auxilliary flag to mark RT lock waiters */
101#define TASK_RTLOCK_WAIT 0x1000
102#define TASK_STATE_MAX 0x2000
103
104/* Convenience macros for the sake of set_current_state: */
105#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
106#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
107#define TASK_TRACED __TASK_TRACED
108
109#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
110
111/* Convenience macros for the sake of wake_up(): */
112#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
113
114/* get_task_state(): */
115#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
116 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
117 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
118 TASK_PARKED)
119
120#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
121
122#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
123#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
124#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
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) \
131 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
132
133#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
134# define debug_normal_state_change(state_value) \
135 do { \
136 WARN_ON_ONCE(is_special_task_state(state_value)); \
137 current->task_state_change = _THIS_IP_; \
138 } while (0)
139
140# define debug_special_state_change(state_value) \
141 do { \
142 WARN_ON_ONCE(!is_special_task_state(state_value)); \
143 current->task_state_change = _THIS_IP_; \
144 } while (0)
145
146# define debug_rtlock_wait_set_state() \
147 do { \
148 current->saved_state_change = current->task_state_change;\
149 current->task_state_change = _THIS_IP_; \
150 } while (0)
151
152# define debug_rtlock_wait_restore_state() \
153 do { \
154 current->task_state_change = current->saved_state_change;\
155 } while (0)
156
157#else
158# define debug_normal_state_change(cond) do { } while (0)
159# define debug_special_state_change(cond) do { } while (0)
160# define debug_rtlock_wait_set_state() do { } while (0)
161# define debug_rtlock_wait_restore_state() do { } while (0)
162#endif
163
164/*
165 * set_current_state() includes a barrier so that the write of current->state
166 * is correctly serialised wrt the caller's subsequent test of whether to
167 * actually sleep:
168 *
169 * for (;;) {
170 * set_current_state(TASK_UNINTERRUPTIBLE);
171 * if (CONDITION)
172 * break;
173 *
174 * schedule();
175 * }
176 * __set_current_state(TASK_RUNNING);
177 *
178 * If the caller does not need such serialisation (because, for instance, the
179 * CONDITION test and condition change and wakeup are under the same lock) then
180 * use __set_current_state().
181 *
182 * The above is typically ordered against the wakeup, which does:
183 *
184 * CONDITION = 1;
185 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
186 *
187 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
188 * accessing p->state.
189 *
190 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
191 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
192 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
193 *
194 * However, with slightly different timing the wakeup TASK_RUNNING store can
195 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
196 * a problem either because that will result in one extra go around the loop
197 * and our @cond test will save the day.
198 *
199 * Also see the comments of try_to_wake_up().
200 */
201#define __set_current_state(state_value) \
202 do { \
203 debug_normal_state_change((state_value)); \
204 WRITE_ONCE(current->__state, (state_value)); \
205 } while (0)
206
207#define set_current_state(state_value) \
208 do { \
209 debug_normal_state_change((state_value)); \
210 smp_store_mb(current->__state, (state_value)); \
211 } while (0)
212
213/*
214 * set_special_state() should be used for those states when the blocking task
215 * can not use the regular condition based wait-loop. In that case we must
216 * serialize against wakeups such that any possible in-flight TASK_RUNNING
217 * stores will not collide with our state change.
218 */
219#define set_special_state(state_value) \
220 do { \
221 unsigned long flags; /* may shadow */ \
222 \
223 raw_spin_lock_irqsave(&current->pi_lock, flags); \
224 debug_special_state_change((state_value)); \
225 WRITE_ONCE(current->__state, (state_value)); \
226 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
227 } while (0)
228
229/*
230 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
231 *
232 * RT's spin/rwlock substitutions are state preserving. The state of the
233 * task when blocking on the lock is saved in task_struct::saved_state and
234 * restored after the lock has been acquired. These operations are
235 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
236 * lock related wakeups while the task is blocked on the lock are
237 * redirected to operate on task_struct::saved_state to ensure that these
238 * are not dropped. On restore task_struct::saved_state is set to
239 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
240 *
241 * The lock operation looks like this:
242 *
243 * current_save_and_set_rtlock_wait_state();
244 * for (;;) {
245 * if (try_lock())
246 * break;
247 * raw_spin_unlock_irq(&lock->wait_lock);
248 * schedule_rtlock();
249 * raw_spin_lock_irq(&lock->wait_lock);
250 * set_current_state(TASK_RTLOCK_WAIT);
251 * }
252 * current_restore_rtlock_saved_state();
253 */
254#define current_save_and_set_rtlock_wait_state() \
255 do { \
256 lockdep_assert_irqs_disabled(); \
257 raw_spin_lock(&current->pi_lock); \
258 current->saved_state = current->__state; \
259 debug_rtlock_wait_set_state(); \
260 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
261 raw_spin_unlock(&current->pi_lock); \
262 } while (0);
263
264#define current_restore_rtlock_saved_state() \
265 do { \
266 lockdep_assert_irqs_disabled(); \
267 raw_spin_lock(&current->pi_lock); \
268 debug_rtlock_wait_restore_state(); \
269 WRITE_ONCE(current->__state, current->saved_state); \
270 current->saved_state = TASK_RUNNING; \
271 raw_spin_unlock(&current->pi_lock); \
272 } while (0);
273
274#define get_current_state() READ_ONCE(current->__state)
275
276/*
277 * Define the task command name length as enum, then it can be visible to
278 * BPF programs.
279 */
280enum {
281 TASK_COMM_LEN = 16,
282};
283
284extern void scheduler_tick(void);
285
286#define MAX_SCHEDULE_TIMEOUT LONG_MAX
287
288extern long schedule_timeout(long timeout);
289extern long schedule_timeout_interruptible(long timeout);
290extern long schedule_timeout_killable(long timeout);
291extern long schedule_timeout_uninterruptible(long timeout);
292extern long schedule_timeout_idle(long timeout);
293asmlinkage void schedule(void);
294extern void schedule_preempt_disabled(void);
295asmlinkage void preempt_schedule_irq(void);
296#ifdef CONFIG_PREEMPT_RT
297 extern void schedule_rtlock(void);
298#endif
299
300extern int __must_check io_schedule_prepare(void);
301extern void io_schedule_finish(int token);
302extern long io_schedule_timeout(long timeout);
303extern void io_schedule(void);
304
305/**
306 * struct prev_cputime - snapshot of system and user cputime
307 * @utime: time spent in user mode
308 * @stime: time spent in system mode
309 * @lock: protects the above two fields
310 *
311 * Stores previous user/system time values such that we can guarantee
312 * monotonicity.
313 */
314struct prev_cputime {
315#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
316 u64 utime;
317 u64 stime;
318 raw_spinlock_t lock;
319#endif
320};
321
322enum vtime_state {
323 /* Task is sleeping or running in a CPU with VTIME inactive: */
324 VTIME_INACTIVE = 0,
325 /* Task is idle */
326 VTIME_IDLE,
327 /* Task runs in kernelspace in a CPU with VTIME active: */
328 VTIME_SYS,
329 /* Task runs in userspace in a CPU with VTIME active: */
330 VTIME_USER,
331 /* Task runs as guests in a CPU with VTIME active: */
332 VTIME_GUEST,
333};
334
335struct vtime {
336 seqcount_t seqcount;
337 unsigned long long starttime;
338 enum vtime_state state;
339 unsigned int cpu;
340 u64 utime;
341 u64 stime;
342 u64 gtime;
343};
344
345/*
346 * Utilization clamp constraints.
347 * @UCLAMP_MIN: Minimum utilization
348 * @UCLAMP_MAX: Maximum utilization
349 * @UCLAMP_CNT: Utilization clamp constraints count
350 */
351enum uclamp_id {
352 UCLAMP_MIN = 0,
353 UCLAMP_MAX,
354 UCLAMP_CNT
355};
356
357#ifdef CONFIG_SMP
358extern struct root_domain def_root_domain;
359extern struct mutex sched_domains_mutex;
360#endif
361
362struct sched_info {
363#ifdef CONFIG_SCHED_INFO
364 /* Cumulative counters: */
365
366 /* # of times we have run on this CPU: */
367 unsigned long pcount;
368
369 /* Time spent waiting on a runqueue: */
370 unsigned long long run_delay;
371
372 /* Timestamps: */
373
374 /* When did we last run on a CPU? */
375 unsigned long long last_arrival;
376
377 /* When were we last queued to run? */
378 unsigned long long last_queued;
379
380#endif /* CONFIG_SCHED_INFO */
381};
382
383/*
384 * Integer metrics need fixed point arithmetic, e.g., sched/fair
385 * has a few: load, load_avg, util_avg, freq, and capacity.
386 *
387 * We define a basic fixed point arithmetic range, and then formalize
388 * all these metrics based on that basic range.
389 */
390# define SCHED_FIXEDPOINT_SHIFT 10
391# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
392
393/* Increase resolution of cpu_capacity calculations */
394# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
395# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
396
397struct load_weight {
398 unsigned long weight;
399 u32 inv_weight;
400};
401
402/**
403 * struct util_est - Estimation utilization of FAIR tasks
404 * @enqueued: instantaneous estimated utilization of a task/cpu
405 * @ewma: the Exponential Weighted Moving Average (EWMA)
406 * utilization of a task
407 *
408 * Support data structure to track an Exponential Weighted Moving Average
409 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
410 * average each time a task completes an activation. Sample's weight is chosen
411 * so that the EWMA will be relatively insensitive to transient changes to the
412 * task's workload.
413 *
414 * The enqueued attribute has a slightly different meaning for tasks and cpus:
415 * - task: the task's util_avg at last task dequeue time
416 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
417 * Thus, the util_est.enqueued of a task represents the contribution on the
418 * estimated utilization of the CPU where that task is currently enqueued.
419 *
420 * Only for tasks we track a moving average of the past instantaneous
421 * estimated utilization. This allows to absorb sporadic drops in utilization
422 * of an otherwise almost periodic task.
423 *
424 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
425 * updates. When a task is dequeued, its util_est should not be updated if its
426 * util_avg has not been updated in the meantime.
427 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
428 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
429 * for a task) it is safe to use MSB.
430 */
431struct util_est {
432 unsigned int enqueued;
433 unsigned int ewma;
434#define UTIL_EST_WEIGHT_SHIFT 2
435#define UTIL_AVG_UNCHANGED 0x80000000
436} __attribute__((__aligned__(sizeof(u64))));
437
438/*
439 * The load/runnable/util_avg accumulates an infinite geometric series
440 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
441 *
442 * [load_avg definition]
443 *
444 * load_avg = runnable% * scale_load_down(load)
445 *
446 * [runnable_avg definition]
447 *
448 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
449 *
450 * [util_avg definition]
451 *
452 * util_avg = running% * SCHED_CAPACITY_SCALE
453 *
454 * where runnable% is the time ratio that a sched_entity is runnable and
455 * running% the time ratio that a sched_entity is running.
456 *
457 * For cfs_rq, they are the aggregated values of all runnable and blocked
458 * sched_entities.
459 *
460 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
461 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
462 * for computing those signals (see update_rq_clock_pelt())
463 *
464 * N.B., the above ratios (runnable% and running%) themselves are in the
465 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
466 * to as large a range as necessary. This is for example reflected by
467 * util_avg's SCHED_CAPACITY_SCALE.
468 *
469 * [Overflow issue]
470 *
471 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
472 * with the highest load (=88761), always runnable on a single cfs_rq,
473 * and should not overflow as the number already hits PID_MAX_LIMIT.
474 *
475 * For all other cases (including 32-bit kernels), struct load_weight's
476 * weight will overflow first before we do, because:
477 *
478 * Max(load_avg) <= Max(load.weight)
479 *
480 * Then it is the load_weight's responsibility to consider overflow
481 * issues.
482 */
483struct sched_avg {
484 u64 last_update_time;
485 u64 load_sum;
486 u64 runnable_sum;
487 u32 util_sum;
488 u32 period_contrib;
489 unsigned long load_avg;
490 unsigned long runnable_avg;
491 unsigned long util_avg;
492 struct util_est util_est;
493} ____cacheline_aligned;
494
495struct sched_statistics {
496#ifdef CONFIG_SCHEDSTATS
497 u64 wait_start;
498 u64 wait_max;
499 u64 wait_count;
500 u64 wait_sum;
501 u64 iowait_count;
502 u64 iowait_sum;
503
504 u64 sleep_start;
505 u64 sleep_max;
506 s64 sum_sleep_runtime;
507
508 u64 block_start;
509 u64 block_max;
510 s64 sum_block_runtime;
511
512 u64 exec_max;
513 u64 slice_max;
514
515 u64 nr_migrations_cold;
516 u64 nr_failed_migrations_affine;
517 u64 nr_failed_migrations_running;
518 u64 nr_failed_migrations_hot;
519 u64 nr_forced_migrations;
520
521 u64 nr_wakeups;
522 u64 nr_wakeups_sync;
523 u64 nr_wakeups_migrate;
524 u64 nr_wakeups_local;
525 u64 nr_wakeups_remote;
526 u64 nr_wakeups_affine;
527 u64 nr_wakeups_affine_attempts;
528 u64 nr_wakeups_passive;
529 u64 nr_wakeups_idle;
530
531#ifdef CONFIG_SCHED_CORE
532 u64 core_forceidle_sum;
533#endif
534#endif /* CONFIG_SCHEDSTATS */
535} ____cacheline_aligned;
536
537struct sched_entity {
538 /* For load-balancing: */
539 struct load_weight load;
540 struct rb_node run_node;
541 struct list_head group_node;
542 unsigned int on_rq;
543
544 u64 exec_start;
545 u64 sum_exec_runtime;
546 u64 vruntime;
547 u64 prev_sum_exec_runtime;
548
549 u64 nr_migrations;
550
551#ifdef CONFIG_FAIR_GROUP_SCHED
552 int depth;
553 struct sched_entity *parent;
554 /* rq on which this entity is (to be) queued: */
555 struct cfs_rq *cfs_rq;
556 /* rq "owned" by this entity/group: */
557 struct cfs_rq *my_q;
558 /* cached value of my_q->h_nr_running */
559 unsigned long runnable_weight;
560#endif
561
562#ifdef CONFIG_SMP
563 /*
564 * Per entity load average tracking.
565 *
566 * Put into separate cache line so it does not
567 * collide with read-mostly values above.
568 */
569 struct sched_avg avg;
570#endif
571};
572
573struct sched_rt_entity {
574 struct list_head run_list;
575 unsigned long timeout;
576 unsigned long watchdog_stamp;
577 unsigned int time_slice;
578 unsigned short on_rq;
579 unsigned short on_list;
580
581 struct sched_rt_entity *back;
582#ifdef CONFIG_RT_GROUP_SCHED
583 struct sched_rt_entity *parent;
584 /* rq on which this entity is (to be) queued: */
585 struct rt_rq *rt_rq;
586 /* rq "owned" by this entity/group: */
587 struct rt_rq *my_q;
588#endif
589} __randomize_layout;
590
591struct sched_dl_entity {
592 struct rb_node rb_node;
593
594 /*
595 * Original scheduling parameters. Copied here from sched_attr
596 * during sched_setattr(), they will remain the same until
597 * the next sched_setattr().
598 */
599 u64 dl_runtime; /* Maximum runtime for each instance */
600 u64 dl_deadline; /* Relative deadline of each instance */
601 u64 dl_period; /* Separation of two instances (period) */
602 u64 dl_bw; /* dl_runtime / dl_period */
603 u64 dl_density; /* dl_runtime / dl_deadline */
604
605 /*
606 * Actual scheduling parameters. Initialized with the values above,
607 * they are continuously updated during task execution. Note that
608 * the remaining runtime could be < 0 in case we are in overrun.
609 */
610 s64 runtime; /* Remaining runtime for this instance */
611 u64 deadline; /* Absolute deadline for this instance */
612 unsigned int flags; /* Specifying the scheduler behaviour */
613
614 /*
615 * Some bool flags:
616 *
617 * @dl_throttled tells if we exhausted the runtime. If so, the
618 * task has to wait for a replenishment to be performed at the
619 * next firing of dl_timer.
620 *
621 * @dl_yielded tells if task gave up the CPU before consuming
622 * all its available runtime during the last job.
623 *
624 * @dl_non_contending tells if the task is inactive while still
625 * contributing to the active utilization. In other words, it
626 * indicates if the inactive timer has been armed and its handler
627 * has not been executed yet. This flag is useful to avoid race
628 * conditions between the inactive timer handler and the wakeup
629 * code.
630 *
631 * @dl_overrun tells if the task asked to be informed about runtime
632 * overruns.
633 */
634 unsigned int dl_throttled : 1;
635 unsigned int dl_yielded : 1;
636 unsigned int dl_non_contending : 1;
637 unsigned int dl_overrun : 1;
638
639 /*
640 * Bandwidth enforcement timer. Each -deadline task has its
641 * own bandwidth to be enforced, thus we need one timer per task.
642 */
643 struct hrtimer dl_timer;
644
645 /*
646 * Inactive timer, responsible for decreasing the active utilization
647 * at the "0-lag time". When a -deadline task blocks, it contributes
648 * to GRUB's active utilization until the "0-lag time", hence a
649 * timer is needed to decrease the active utilization at the correct
650 * time.
651 */
652 struct hrtimer inactive_timer;
653
654#ifdef CONFIG_RT_MUTEXES
655 /*
656 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
657 * pi_se points to the donor, otherwise points to the dl_se it belongs
658 * to (the original one/itself).
659 */
660 struct sched_dl_entity *pi_se;
661#endif
662};
663
664#ifdef CONFIG_UCLAMP_TASK
665/* Number of utilization clamp buckets (shorter alias) */
666#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
667
668/*
669 * Utilization clamp for a scheduling entity
670 * @value: clamp value "assigned" to a se
671 * @bucket_id: bucket index corresponding to the "assigned" value
672 * @active: the se is currently refcounted in a rq's bucket
673 * @user_defined: the requested clamp value comes from user-space
674 *
675 * The bucket_id is the index of the clamp bucket matching the clamp value
676 * which is pre-computed and stored to avoid expensive integer divisions from
677 * the fast path.
678 *
679 * The active bit is set whenever a task has got an "effective" value assigned,
680 * which can be different from the clamp value "requested" from user-space.
681 * This allows to know a task is refcounted in the rq's bucket corresponding
682 * to the "effective" bucket_id.
683 *
684 * The user_defined bit is set whenever a task has got a task-specific clamp
685 * value requested from userspace, i.e. the system defaults apply to this task
686 * just as a restriction. This allows to relax default clamps when a less
687 * restrictive task-specific value has been requested, thus allowing to
688 * implement a "nice" semantic. For example, a task running with a 20%
689 * default boost can still drop its own boosting to 0%.
690 */
691struct uclamp_se {
692 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
693 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
694 unsigned int active : 1;
695 unsigned int user_defined : 1;
696};
697#endif /* CONFIG_UCLAMP_TASK */
698
699union rcu_special {
700 struct {
701 u8 blocked;
702 u8 need_qs;
703 u8 exp_hint; /* Hint for performance. */
704 u8 need_mb; /* Readers need smp_mb(). */
705 } b; /* Bits. */
706 u32 s; /* Set of bits. */
707};
708
709enum perf_event_task_context {
710 perf_invalid_context = -1,
711 perf_hw_context = 0,
712 perf_sw_context,
713 perf_nr_task_contexts,
714};
715
716struct wake_q_node {
717 struct wake_q_node *next;
718};
719
720struct kmap_ctrl {
721#ifdef CONFIG_KMAP_LOCAL
722 int idx;
723 pte_t pteval[KM_MAX_IDX];
724#endif
725};
726
727struct task_struct {
728#ifdef CONFIG_THREAD_INFO_IN_TASK
729 /*
730 * For reasons of header soup (see current_thread_info()), this
731 * must be the first element of task_struct.
732 */
733 struct thread_info thread_info;
734#endif
735 unsigned int __state;
736
737#ifdef CONFIG_PREEMPT_RT
738 /* saved state for "spinlock sleepers" */
739 unsigned int saved_state;
740#endif
741
742 /*
743 * This begins the randomizable portion of task_struct. Only
744 * scheduling-critical items should be added above here.
745 */
746 randomized_struct_fields_start
747
748 void *stack;
749 refcount_t usage;
750 /* Per task flags (PF_*), defined further below: */
751 unsigned int flags;
752 unsigned int ptrace;
753
754#ifdef CONFIG_SMP
755 int on_cpu;
756 struct __call_single_node wake_entry;
757 unsigned int wakee_flips;
758 unsigned long wakee_flip_decay_ts;
759 struct task_struct *last_wakee;
760
761 /*
762 * recent_used_cpu is initially set as the last CPU used by a task
763 * that wakes affine another task. Waker/wakee relationships can
764 * push tasks around a CPU where each wakeup moves to the next one.
765 * Tracking a recently used CPU allows a quick search for a recently
766 * used CPU that may be idle.
767 */
768 int recent_used_cpu;
769 int wake_cpu;
770#endif
771 int on_rq;
772
773 int prio;
774 int static_prio;
775 int normal_prio;
776 unsigned int rt_priority;
777
778 struct sched_entity se;
779 struct sched_rt_entity rt;
780 struct sched_dl_entity dl;
781 const struct sched_class *sched_class;
782
783#ifdef CONFIG_SCHED_CORE
784 struct rb_node core_node;
785 unsigned long core_cookie;
786 unsigned int core_occupation;
787#endif
788
789#ifdef CONFIG_CGROUP_SCHED
790 struct task_group *sched_task_group;
791#endif
792
793#ifdef CONFIG_UCLAMP_TASK
794 /*
795 * Clamp values requested for a scheduling entity.
796 * Must be updated with task_rq_lock() held.
797 */
798 struct uclamp_se uclamp_req[UCLAMP_CNT];
799 /*
800 * Effective clamp values used for a scheduling entity.
801 * Must be updated with task_rq_lock() held.
802 */
803 struct uclamp_se uclamp[UCLAMP_CNT];
804#endif
805
806 struct sched_statistics stats;
807
808#ifdef CONFIG_PREEMPT_NOTIFIERS
809 /* List of struct preempt_notifier: */
810 struct hlist_head preempt_notifiers;
811#endif
812
813#ifdef CONFIG_BLK_DEV_IO_TRACE
814 unsigned int btrace_seq;
815#endif
816
817 unsigned int policy;
818 int nr_cpus_allowed;
819 const cpumask_t *cpus_ptr;
820 cpumask_t *user_cpus_ptr;
821 cpumask_t cpus_mask;
822 void *migration_pending;
823#ifdef CONFIG_SMP
824 unsigned short migration_disabled;
825#endif
826 unsigned short migration_flags;
827
828#ifdef CONFIG_PREEMPT_RCU
829 int rcu_read_lock_nesting;
830 union rcu_special rcu_read_unlock_special;
831 struct list_head rcu_node_entry;
832 struct rcu_node *rcu_blocked_node;
833#endif /* #ifdef CONFIG_PREEMPT_RCU */
834
835#ifdef CONFIG_TASKS_RCU
836 unsigned long rcu_tasks_nvcsw;
837 u8 rcu_tasks_holdout;
838 u8 rcu_tasks_idx;
839 int rcu_tasks_idle_cpu;
840 struct list_head rcu_tasks_holdout_list;
841#endif /* #ifdef CONFIG_TASKS_RCU */
842
843#ifdef CONFIG_TASKS_TRACE_RCU
844 int trc_reader_nesting;
845 int trc_ipi_to_cpu;
846 union rcu_special trc_reader_special;
847 struct list_head trc_holdout_list;
848 struct list_head trc_blkd_node;
849 int trc_blkd_cpu;
850#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
851
852 struct sched_info sched_info;
853
854 struct list_head tasks;
855#ifdef CONFIG_SMP
856 struct plist_node pushable_tasks;
857 struct rb_node pushable_dl_tasks;
858#endif
859
860 struct mm_struct *mm;
861 struct mm_struct *active_mm;
862
863 /* Per-thread vma caching: */
864 struct vmacache vmacache;
865
866#ifdef SPLIT_RSS_COUNTING
867 struct task_rss_stat rss_stat;
868#endif
869 int exit_state;
870 int exit_code;
871 int exit_signal;
872 /* The signal sent when the parent dies: */
873 int pdeath_signal;
874 /* JOBCTL_*, siglock protected: */
875 unsigned long jobctl;
876
877 /* Used for emulating ABI behavior of previous Linux versions: */
878 unsigned int personality;
879
880 /* Scheduler bits, serialized by scheduler locks: */
881 unsigned sched_reset_on_fork:1;
882 unsigned sched_contributes_to_load:1;
883 unsigned sched_migrated:1;
884#ifdef CONFIG_PSI
885 unsigned sched_psi_wake_requeue:1;
886#endif
887
888 /* Force alignment to the next boundary: */
889 unsigned :0;
890
891 /* Unserialized, strictly 'current' */
892
893 /*
894 * This field must not be in the scheduler word above due to wakelist
895 * queueing no longer being serialized by p->on_cpu. However:
896 *
897 * p->XXX = X; ttwu()
898 * schedule() if (p->on_rq && ..) // false
899 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
900 * deactivate_task() ttwu_queue_wakelist())
901 * p->on_rq = 0; p->sched_remote_wakeup = Y;
902 *
903 * guarantees all stores of 'current' are visible before
904 * ->sched_remote_wakeup gets used, so it can be in this word.
905 */
906 unsigned sched_remote_wakeup:1;
907
908 /* Bit to tell LSMs we're in execve(): */
909 unsigned in_execve:1;
910 unsigned in_iowait:1;
911#ifndef TIF_RESTORE_SIGMASK
912 unsigned restore_sigmask:1;
913#endif
914#ifdef CONFIG_MEMCG
915 unsigned in_user_fault:1;
916#endif
917#ifdef CONFIG_COMPAT_BRK
918 unsigned brk_randomized:1;
919#endif
920#ifdef CONFIG_CGROUPS
921 /* disallow userland-initiated cgroup migration */
922 unsigned no_cgroup_migration:1;
923 /* task is frozen/stopped (used by the cgroup freezer) */
924 unsigned frozen:1;
925#endif
926#ifdef CONFIG_BLK_CGROUP
927 unsigned use_memdelay:1;
928#endif
929#ifdef CONFIG_PSI
930 /* Stalled due to lack of memory */
931 unsigned in_memstall:1;
932#endif
933#ifdef CONFIG_PAGE_OWNER
934 /* Used by page_owner=on to detect recursion in page tracking. */
935 unsigned in_page_owner:1;
936#endif
937#ifdef CONFIG_EVENTFD
938 /* Recursion prevention for eventfd_signal() */
939 unsigned in_eventfd_signal:1;
940#endif
941#ifdef CONFIG_IOMMU_SVA
942 unsigned pasid_activated:1;
943#endif
944#ifdef CONFIG_CPU_SUP_INTEL
945 unsigned reported_split_lock:1;
946#endif
947
948 unsigned long atomic_flags; /* Flags requiring atomic access. */
949
950 struct restart_block restart_block;
951
952 pid_t pid;
953 pid_t tgid;
954
955#ifdef CONFIG_STACKPROTECTOR
956 /* Canary value for the -fstack-protector GCC feature: */
957 unsigned long stack_canary;
958#endif
959 /*
960 * Pointers to the (original) parent process, youngest child, younger sibling,
961 * older sibling, respectively. (p->father can be replaced with
962 * p->real_parent->pid)
963 */
964
965 /* Real parent process: */
966 struct task_struct __rcu *real_parent;
967
968 /* Recipient of SIGCHLD, wait4() reports: */
969 struct task_struct __rcu *parent;
970
971 /*
972 * Children/sibling form the list of natural children:
973 */
974 struct list_head children;
975 struct list_head sibling;
976 struct task_struct *group_leader;
977
978 /*
979 * 'ptraced' is the list of tasks this task is using ptrace() on.
980 *
981 * This includes both natural children and PTRACE_ATTACH targets.
982 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
983 */
984 struct list_head ptraced;
985 struct list_head ptrace_entry;
986
987 /* PID/PID hash table linkage. */
988 struct pid *thread_pid;
989 struct hlist_node pid_links[PIDTYPE_MAX];
990 struct list_head thread_group;
991 struct list_head thread_node;
992
993 struct completion *vfork_done;
994
995 /* CLONE_CHILD_SETTID: */
996 int __user *set_child_tid;
997
998 /* CLONE_CHILD_CLEARTID: */
999 int __user *clear_child_tid;
1000
1001 /* PF_KTHREAD | PF_IO_WORKER */
1002 void *worker_private;
1003
1004 u64 utime;
1005 u64 stime;
1006#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1007 u64 utimescaled;
1008 u64 stimescaled;
1009#endif
1010 u64 gtime;
1011 struct prev_cputime prev_cputime;
1012#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1013 struct vtime vtime;
1014#endif
1015
1016#ifdef CONFIG_NO_HZ_FULL
1017 atomic_t tick_dep_mask;
1018#endif
1019 /* Context switch counts: */
1020 unsigned long nvcsw;
1021 unsigned long nivcsw;
1022
1023 /* Monotonic time in nsecs: */
1024 u64 start_time;
1025
1026 /* Boot based time in nsecs: */
1027 u64 start_boottime;
1028
1029 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1030 unsigned long min_flt;
1031 unsigned long maj_flt;
1032
1033 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1034 struct posix_cputimers posix_cputimers;
1035
1036#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1037 struct posix_cputimers_work posix_cputimers_work;
1038#endif
1039
1040 /* Process credentials: */
1041
1042 /* Tracer's credentials at attach: */
1043 const struct cred __rcu *ptracer_cred;
1044
1045 /* Objective and real subjective task credentials (COW): */
1046 const struct cred __rcu *real_cred;
1047
1048 /* Effective (overridable) subjective task credentials (COW): */
1049 const struct cred __rcu *cred;
1050
1051#ifdef CONFIG_KEYS
1052 /* Cached requested key. */
1053 struct key *cached_requested_key;
1054#endif
1055
1056 /*
1057 * executable name, excluding path.
1058 *
1059 * - normally initialized setup_new_exec()
1060 * - access it with [gs]et_task_comm()
1061 * - lock it with task_lock()
1062 */
1063 char comm[TASK_COMM_LEN];
1064
1065 struct nameidata *nameidata;
1066
1067#ifdef CONFIG_SYSVIPC
1068 struct sysv_sem sysvsem;
1069 struct sysv_shm sysvshm;
1070#endif
1071#ifdef CONFIG_DETECT_HUNG_TASK
1072 unsigned long last_switch_count;
1073 unsigned long last_switch_time;
1074#endif
1075 /* Filesystem information: */
1076 struct fs_struct *fs;
1077
1078 /* Open file information: */
1079 struct files_struct *files;
1080
1081#ifdef CONFIG_IO_URING
1082 struct io_uring_task *io_uring;
1083#endif
1084
1085 /* Namespaces: */
1086 struct nsproxy *nsproxy;
1087
1088 /* Signal handlers: */
1089 struct signal_struct *signal;
1090 struct sighand_struct __rcu *sighand;
1091 sigset_t blocked;
1092 sigset_t real_blocked;
1093 /* Restored if set_restore_sigmask() was used: */
1094 sigset_t saved_sigmask;
1095 struct sigpending pending;
1096 unsigned long sas_ss_sp;
1097 size_t sas_ss_size;
1098 unsigned int sas_ss_flags;
1099
1100 struct callback_head *task_works;
1101
1102#ifdef CONFIG_AUDIT
1103#ifdef CONFIG_AUDITSYSCALL
1104 struct audit_context *audit_context;
1105#endif
1106 kuid_t loginuid;
1107 unsigned int sessionid;
1108#endif
1109 struct seccomp seccomp;
1110 struct syscall_user_dispatch syscall_dispatch;
1111
1112 /* Thread group tracking: */
1113 u64 parent_exec_id;
1114 u64 self_exec_id;
1115
1116 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1117 spinlock_t alloc_lock;
1118
1119 /* Protection of the PI data structures: */
1120 raw_spinlock_t pi_lock;
1121
1122 struct wake_q_node wake_q;
1123
1124#ifdef CONFIG_RT_MUTEXES
1125 /* PI waiters blocked on a rt_mutex held by this task: */
1126 struct rb_root_cached pi_waiters;
1127 /* Updated under owner's pi_lock and rq lock */
1128 struct task_struct *pi_top_task;
1129 /* Deadlock detection and priority inheritance handling: */
1130 struct rt_mutex_waiter *pi_blocked_on;
1131#endif
1132
1133#ifdef CONFIG_DEBUG_MUTEXES
1134 /* Mutex deadlock detection: */
1135 struct mutex_waiter *blocked_on;
1136#endif
1137
1138#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1139 int non_block_count;
1140#endif
1141
1142#ifdef CONFIG_TRACE_IRQFLAGS
1143 struct irqtrace_events irqtrace;
1144 unsigned int hardirq_threaded;
1145 u64 hardirq_chain_key;
1146 int softirqs_enabled;
1147 int softirq_context;
1148 int irq_config;
1149#endif
1150#ifdef CONFIG_PREEMPT_RT
1151 int softirq_disable_cnt;
1152#endif
1153
1154#ifdef CONFIG_LOCKDEP
1155# define MAX_LOCK_DEPTH 48UL
1156 u64 curr_chain_key;
1157 int lockdep_depth;
1158 unsigned int lockdep_recursion;
1159 struct held_lock held_locks[MAX_LOCK_DEPTH];
1160#endif
1161
1162#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1163 unsigned int in_ubsan;
1164#endif
1165
1166 /* Journalling filesystem info: */
1167 void *journal_info;
1168
1169 /* Stacked block device info: */
1170 struct bio_list *bio_list;
1171
1172 /* Stack plugging: */
1173 struct blk_plug *plug;
1174
1175 /* VM state: */
1176 struct reclaim_state *reclaim_state;
1177
1178 struct backing_dev_info *backing_dev_info;
1179
1180 struct io_context *io_context;
1181
1182#ifdef CONFIG_COMPACTION
1183 struct capture_control *capture_control;
1184#endif
1185 /* Ptrace state: */
1186 unsigned long ptrace_message;
1187 kernel_siginfo_t *last_siginfo;
1188
1189 struct task_io_accounting ioac;
1190#ifdef CONFIG_PSI
1191 /* Pressure stall state */
1192 unsigned int psi_flags;
1193#endif
1194#ifdef CONFIG_TASK_XACCT
1195 /* Accumulated RSS usage: */
1196 u64 acct_rss_mem1;
1197 /* Accumulated virtual memory usage: */
1198 u64 acct_vm_mem1;
1199 /* stime + utime since last update: */
1200 u64 acct_timexpd;
1201#endif
1202#ifdef CONFIG_CPUSETS
1203 /* Protected by ->alloc_lock: */
1204 nodemask_t mems_allowed;
1205 /* Sequence number to catch updates: */
1206 seqcount_spinlock_t mems_allowed_seq;
1207 int cpuset_mem_spread_rotor;
1208 int cpuset_slab_spread_rotor;
1209#endif
1210#ifdef CONFIG_CGROUPS
1211 /* Control Group info protected by css_set_lock: */
1212 struct css_set __rcu *cgroups;
1213 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1214 struct list_head cg_list;
1215#endif
1216#ifdef CONFIG_X86_CPU_RESCTRL
1217 u32 closid;
1218 u32 rmid;
1219#endif
1220#ifdef CONFIG_FUTEX
1221 struct robust_list_head __user *robust_list;
1222#ifdef CONFIG_COMPAT
1223 struct compat_robust_list_head __user *compat_robust_list;
1224#endif
1225 struct list_head pi_state_list;
1226 struct futex_pi_state *pi_state_cache;
1227 struct mutex futex_exit_mutex;
1228 unsigned int futex_state;
1229#endif
1230#ifdef CONFIG_PERF_EVENTS
1231 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1232 struct mutex perf_event_mutex;
1233 struct list_head perf_event_list;
1234#endif
1235#ifdef CONFIG_DEBUG_PREEMPT
1236 unsigned long preempt_disable_ip;
1237#endif
1238#ifdef CONFIG_NUMA
1239 /* Protected by alloc_lock: */
1240 struct mempolicy *mempolicy;
1241 short il_prev;
1242 short pref_node_fork;
1243#endif
1244#ifdef CONFIG_NUMA_BALANCING
1245 int numa_scan_seq;
1246 unsigned int numa_scan_period;
1247 unsigned int numa_scan_period_max;
1248 int numa_preferred_nid;
1249 unsigned long numa_migrate_retry;
1250 /* Migration stamp: */
1251 u64 node_stamp;
1252 u64 last_task_numa_placement;
1253 u64 last_sum_exec_runtime;
1254 struct callback_head numa_work;
1255
1256 /*
1257 * This pointer is only modified for current in syscall and
1258 * pagefault context (and for tasks being destroyed), so it can be read
1259 * from any of the following contexts:
1260 * - RCU read-side critical section
1261 * - current->numa_group from everywhere
1262 * - task's runqueue locked, task not running
1263 */
1264 struct numa_group __rcu *numa_group;
1265
1266 /*
1267 * numa_faults is an array split into four regions:
1268 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1269 * in this precise order.
1270 *
1271 * faults_memory: Exponential decaying average of faults on a per-node
1272 * basis. Scheduling placement decisions are made based on these
1273 * counts. The values remain static for the duration of a PTE scan.
1274 * faults_cpu: Track the nodes the process was running on when a NUMA
1275 * hinting fault was incurred.
1276 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1277 * during the current scan window. When the scan completes, the counts
1278 * in faults_memory and faults_cpu decay and these values are copied.
1279 */
1280 unsigned long *numa_faults;
1281 unsigned long total_numa_faults;
1282
1283 /*
1284 * numa_faults_locality tracks if faults recorded during the last
1285 * scan window were remote/local or failed to migrate. The task scan
1286 * period is adapted based on the locality of the faults with different
1287 * weights depending on whether they were shared or private faults
1288 */
1289 unsigned long numa_faults_locality[3];
1290
1291 unsigned long numa_pages_migrated;
1292#endif /* CONFIG_NUMA_BALANCING */
1293
1294#ifdef CONFIG_RSEQ
1295 struct rseq __user *rseq;
1296 u32 rseq_sig;
1297 /*
1298 * RmW on rseq_event_mask must be performed atomically
1299 * with respect to preemption.
1300 */
1301 unsigned long rseq_event_mask;
1302#endif
1303
1304 struct tlbflush_unmap_batch tlb_ubc;
1305
1306 union {
1307 refcount_t rcu_users;
1308 struct rcu_head rcu;
1309 };
1310
1311 /* Cache last used pipe for splice(): */
1312 struct pipe_inode_info *splice_pipe;
1313
1314 struct page_frag task_frag;
1315
1316#ifdef CONFIG_TASK_DELAY_ACCT
1317 struct task_delay_info *delays;
1318#endif
1319
1320#ifdef CONFIG_FAULT_INJECTION
1321 int make_it_fail;
1322 unsigned int fail_nth;
1323#endif
1324 /*
1325 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1326 * balance_dirty_pages() for a dirty throttling pause:
1327 */
1328 int nr_dirtied;
1329 int nr_dirtied_pause;
1330 /* Start of a write-and-pause period: */
1331 unsigned long dirty_paused_when;
1332
1333#ifdef CONFIG_LATENCYTOP
1334 int latency_record_count;
1335 struct latency_record latency_record[LT_SAVECOUNT];
1336#endif
1337 /*
1338 * Time slack values; these are used to round up poll() and
1339 * select() etc timeout values. These are in nanoseconds.
1340 */
1341 u64 timer_slack_ns;
1342 u64 default_timer_slack_ns;
1343
1344#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1345 unsigned int kasan_depth;
1346#endif
1347
1348#ifdef CONFIG_KCSAN
1349 struct kcsan_ctx kcsan_ctx;
1350#ifdef CONFIG_TRACE_IRQFLAGS
1351 struct irqtrace_events kcsan_save_irqtrace;
1352#endif
1353#ifdef CONFIG_KCSAN_WEAK_MEMORY
1354 int kcsan_stack_depth;
1355#endif
1356#endif
1357
1358#if IS_ENABLED(CONFIG_KUNIT)
1359 struct kunit *kunit_test;
1360#endif
1361
1362#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1363 /* Index of current stored address in ret_stack: */
1364 int curr_ret_stack;
1365 int curr_ret_depth;
1366
1367 /* Stack of return addresses for return function tracing: */
1368 struct ftrace_ret_stack *ret_stack;
1369
1370 /* Timestamp for last schedule: */
1371 unsigned long long ftrace_timestamp;
1372
1373 /*
1374 * Number of functions that haven't been traced
1375 * because of depth overrun:
1376 */
1377 atomic_t trace_overrun;
1378
1379 /* Pause tracing: */
1380 atomic_t tracing_graph_pause;
1381#endif
1382
1383#ifdef CONFIG_TRACING
1384 /* State flags for use by tracers: */
1385 unsigned long trace;
1386
1387 /* Bitmask and counter of trace recursion: */
1388 unsigned long trace_recursion;
1389#endif /* CONFIG_TRACING */
1390
1391#ifdef CONFIG_KCOV
1392 /* See kernel/kcov.c for more details. */
1393
1394 /* Coverage collection mode enabled for this task (0 if disabled): */
1395 unsigned int kcov_mode;
1396
1397 /* Size of the kcov_area: */
1398 unsigned int kcov_size;
1399
1400 /* Buffer for coverage collection: */
1401 void *kcov_area;
1402
1403 /* KCOV descriptor wired with this task or NULL: */
1404 struct kcov *kcov;
1405
1406 /* KCOV common handle for remote coverage collection: */
1407 u64 kcov_handle;
1408
1409 /* KCOV sequence number: */
1410 int kcov_sequence;
1411
1412 /* Collect coverage from softirq context: */
1413 unsigned int kcov_softirq;
1414#endif
1415
1416#ifdef CONFIG_MEMCG
1417 struct mem_cgroup *memcg_in_oom;
1418 gfp_t memcg_oom_gfp_mask;
1419 int memcg_oom_order;
1420
1421 /* Number of pages to reclaim on returning to userland: */
1422 unsigned int memcg_nr_pages_over_high;
1423
1424 /* Used by memcontrol for targeted memcg charge: */
1425 struct mem_cgroup *active_memcg;
1426#endif
1427
1428#ifdef CONFIG_BLK_CGROUP
1429 struct request_queue *throttle_queue;
1430#endif
1431
1432#ifdef CONFIG_UPROBES
1433 struct uprobe_task *utask;
1434#endif
1435#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1436 unsigned int sequential_io;
1437 unsigned int sequential_io_avg;
1438#endif
1439 struct kmap_ctrl kmap_ctrl;
1440#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1441 unsigned long task_state_change;
1442# ifdef CONFIG_PREEMPT_RT
1443 unsigned long saved_state_change;
1444# endif
1445#endif
1446 int pagefault_disabled;
1447#ifdef CONFIG_MMU
1448 struct task_struct *oom_reaper_list;
1449 struct timer_list oom_reaper_timer;
1450#endif
1451#ifdef CONFIG_VMAP_STACK
1452 struct vm_struct *stack_vm_area;
1453#endif
1454#ifdef CONFIG_THREAD_INFO_IN_TASK
1455 /* A live task holds one reference: */
1456 refcount_t stack_refcount;
1457#endif
1458#ifdef CONFIG_LIVEPATCH
1459 int patch_state;
1460#endif
1461#ifdef CONFIG_SECURITY
1462 /* Used by LSM modules for access restriction: */
1463 void *security;
1464#endif
1465#ifdef CONFIG_BPF_SYSCALL
1466 /* Used by BPF task local storage */
1467 struct bpf_local_storage __rcu *bpf_storage;
1468 /* Used for BPF run context */
1469 struct bpf_run_ctx *bpf_ctx;
1470#endif
1471
1472#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1473 unsigned long lowest_stack;
1474 unsigned long prev_lowest_stack;
1475#endif
1476
1477#ifdef CONFIG_X86_MCE
1478 void __user *mce_vaddr;
1479 __u64 mce_kflags;
1480 u64 mce_addr;
1481 __u64 mce_ripv : 1,
1482 mce_whole_page : 1,
1483 __mce_reserved : 62;
1484 struct callback_head mce_kill_me;
1485 int mce_count;
1486#endif
1487
1488#ifdef CONFIG_KRETPROBES
1489 struct llist_head kretprobe_instances;
1490#endif
1491#ifdef CONFIG_RETHOOK
1492 struct llist_head rethooks;
1493#endif
1494
1495#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1496 /*
1497 * If L1D flush is supported on mm context switch
1498 * then we use this callback head to queue kill work
1499 * to kill tasks that are not running on SMT disabled
1500 * cores
1501 */
1502 struct callback_head l1d_flush_kill;
1503#endif
1504
1505#ifdef CONFIG_RV
1506 /*
1507 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1508 * If we find justification for more monitors, we can think
1509 * about adding more or developing a dynamic method. So far,
1510 * none of these are justified.
1511 */
1512 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1513#endif
1514
1515 /*
1516 * New fields for task_struct should be added above here, so that
1517 * they are included in the randomized portion of task_struct.
1518 */
1519 randomized_struct_fields_end
1520
1521 /* CPU-specific state of this task: */
1522 struct thread_struct thread;
1523
1524 /*
1525 * WARNING: on x86, 'thread_struct' contains a variable-sized
1526 * structure. It *MUST* be at the end of 'task_struct'.
1527 *
1528 * Do not put anything below here!
1529 */
1530};
1531
1532static inline struct pid *task_pid(struct task_struct *task)
1533{
1534 return task->thread_pid;
1535}
1536
1537/*
1538 * the helpers to get the task's different pids as they are seen
1539 * from various namespaces
1540 *
1541 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1542 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1543 * current.
1544 * task_xid_nr_ns() : id seen from the ns specified;
1545 *
1546 * see also pid_nr() etc in include/linux/pid.h
1547 */
1548pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1549
1550static inline pid_t task_pid_nr(struct task_struct *tsk)
1551{
1552 return tsk->pid;
1553}
1554
1555static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1556{
1557 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1558}
1559
1560static inline pid_t task_pid_vnr(struct task_struct *tsk)
1561{
1562 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1563}
1564
1565
1566static inline pid_t task_tgid_nr(struct task_struct *tsk)
1567{
1568 return tsk->tgid;
1569}
1570
1571/**
1572 * pid_alive - check that a task structure is not stale
1573 * @p: Task structure to be checked.
1574 *
1575 * Test if a process is not yet dead (at most zombie state)
1576 * If pid_alive fails, then pointers within the task structure
1577 * can be stale and must not be dereferenced.
1578 *
1579 * Return: 1 if the process is alive. 0 otherwise.
1580 */
1581static inline int pid_alive(const struct task_struct *p)
1582{
1583 return p->thread_pid != NULL;
1584}
1585
1586static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1587{
1588 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1589}
1590
1591static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1592{
1593 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1594}
1595
1596
1597static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1598{
1599 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1600}
1601
1602static inline pid_t task_session_vnr(struct task_struct *tsk)
1603{
1604 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1605}
1606
1607static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1608{
1609 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1610}
1611
1612static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1613{
1614 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1615}
1616
1617static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1618{
1619 pid_t pid = 0;
1620
1621 rcu_read_lock();
1622 if (pid_alive(tsk))
1623 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1624 rcu_read_unlock();
1625
1626 return pid;
1627}
1628
1629static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1630{
1631 return task_ppid_nr_ns(tsk, &init_pid_ns);
1632}
1633
1634/* Obsolete, do not use: */
1635static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1636{
1637 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1638}
1639
1640#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1641#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1642
1643static inline unsigned int __task_state_index(unsigned int tsk_state,
1644 unsigned int tsk_exit_state)
1645{
1646 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1647
1648 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1649
1650 if (tsk_state == TASK_IDLE)
1651 state = TASK_REPORT_IDLE;
1652
1653 /*
1654 * We're lying here, but rather than expose a completely new task state
1655 * to userspace, we can make this appear as if the task has gone through
1656 * a regular rt_mutex_lock() call.
1657 */
1658 if (tsk_state == TASK_RTLOCK_WAIT)
1659 state = TASK_UNINTERRUPTIBLE;
1660
1661 return fls(state);
1662}
1663
1664static inline unsigned int task_state_index(struct task_struct *tsk)
1665{
1666 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1667}
1668
1669static inline char task_index_to_char(unsigned int state)
1670{
1671 static const char state_char[] = "RSDTtXZPI";
1672
1673 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1674
1675 return state_char[state];
1676}
1677
1678static inline char task_state_to_char(struct task_struct *tsk)
1679{
1680 return task_index_to_char(task_state_index(tsk));
1681}
1682
1683/**
1684 * is_global_init - check if a task structure is init. Since init
1685 * is free to have sub-threads we need to check tgid.
1686 * @tsk: Task structure to be checked.
1687 *
1688 * Check if a task structure is the first user space task the kernel created.
1689 *
1690 * Return: 1 if the task structure is init. 0 otherwise.
1691 */
1692static inline int is_global_init(struct task_struct *tsk)
1693{
1694 return task_tgid_nr(tsk) == 1;
1695}
1696
1697extern struct pid *cad_pid;
1698
1699/*
1700 * Per process flags
1701 */
1702#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1703#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1704#define PF_EXITING 0x00000004 /* Getting shut down */
1705#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1706#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1707#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1708#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1709#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1710#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1711#define PF_DUMPCORE 0x00000200 /* Dumped core */
1712#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1713#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1714#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1715#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1716#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1717#define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1718#define PF_KSWAPD 0x00020000 /* I am kswapd */
1719#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1720#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1721#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1722 * I am cleaning dirty pages from some other bdi. */
1723#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1724#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1725#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1726#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1727#define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1728#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1729#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1730
1731/*
1732 * Only the _current_ task can read/write to tsk->flags, but other
1733 * tasks can access tsk->flags in readonly mode for example
1734 * with tsk_used_math (like during threaded core dumping).
1735 * There is however an exception to this rule during ptrace
1736 * or during fork: the ptracer task is allowed to write to the
1737 * child->flags of its traced child (same goes for fork, the parent
1738 * can write to the child->flags), because we're guaranteed the
1739 * child is not running and in turn not changing child->flags
1740 * at the same time the parent does it.
1741 */
1742#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1743#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1744#define clear_used_math() clear_stopped_child_used_math(current)
1745#define set_used_math() set_stopped_child_used_math(current)
1746
1747#define conditional_stopped_child_used_math(condition, child) \
1748 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1749
1750#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1751
1752#define copy_to_stopped_child_used_math(child) \
1753 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1754
1755/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1756#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1757#define used_math() tsk_used_math(current)
1758
1759static __always_inline bool is_percpu_thread(void)
1760{
1761#ifdef CONFIG_SMP
1762 return (current->flags & PF_NO_SETAFFINITY) &&
1763 (current->nr_cpus_allowed == 1);
1764#else
1765 return true;
1766#endif
1767}
1768
1769/* Per-process atomic flags. */
1770#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1771#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1772#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1773#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1774#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1775#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1776#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1777#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1778
1779#define TASK_PFA_TEST(name, func) \
1780 static inline bool task_##func(struct task_struct *p) \
1781 { return test_bit(PFA_##name, &p->atomic_flags); }
1782
1783#define TASK_PFA_SET(name, func) \
1784 static inline void task_set_##func(struct task_struct *p) \
1785 { set_bit(PFA_##name, &p->atomic_flags); }
1786
1787#define TASK_PFA_CLEAR(name, func) \
1788 static inline void task_clear_##func(struct task_struct *p) \
1789 { clear_bit(PFA_##name, &p->atomic_flags); }
1790
1791TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1792TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1793
1794TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1795TASK_PFA_SET(SPREAD_PAGE, spread_page)
1796TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1797
1798TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1799TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1800TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1801
1802TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1803TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1804TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1805
1806TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1807TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1808TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1809
1810TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1811TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1812
1813TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1814TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1815TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1816
1817TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1818TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1819
1820static inline void
1821current_restore_flags(unsigned long orig_flags, unsigned long flags)
1822{
1823 current->flags &= ~flags;
1824 current->flags |= orig_flags & flags;
1825}
1826
1827extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1828extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effective_cpus);
1829#ifdef CONFIG_SMP
1830extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1831extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1832extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1833extern void release_user_cpus_ptr(struct task_struct *p);
1834extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1835extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1836extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1837#else
1838static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1839{
1840}
1841static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1842{
1843 if (!cpumask_test_cpu(0, new_mask))
1844 return -EINVAL;
1845 return 0;
1846}
1847static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1848{
1849 if (src->user_cpus_ptr)
1850 return -EINVAL;
1851 return 0;
1852}
1853static inline void release_user_cpus_ptr(struct task_struct *p)
1854{
1855 WARN_ON(p->user_cpus_ptr);
1856}
1857
1858static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1859{
1860 return 0;
1861}
1862#endif
1863
1864extern int yield_to(struct task_struct *p, bool preempt);
1865extern void set_user_nice(struct task_struct *p, long nice);
1866extern int task_prio(const struct task_struct *p);
1867
1868/**
1869 * task_nice - return the nice value of a given task.
1870 * @p: the task in question.
1871 *
1872 * Return: The nice value [ -20 ... 0 ... 19 ].
1873 */
1874static inline int task_nice(const struct task_struct *p)
1875{
1876 return PRIO_TO_NICE((p)->static_prio);
1877}
1878
1879extern int can_nice(const struct task_struct *p, const int nice);
1880extern int task_curr(const struct task_struct *p);
1881extern int idle_cpu(int cpu);
1882extern int available_idle_cpu(int cpu);
1883extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1884extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1885extern void sched_set_fifo(struct task_struct *p);
1886extern void sched_set_fifo_low(struct task_struct *p);
1887extern void sched_set_normal(struct task_struct *p, int nice);
1888extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1889extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1890extern struct task_struct *idle_task(int cpu);
1891
1892/**
1893 * is_idle_task - is the specified task an idle task?
1894 * @p: the task in question.
1895 *
1896 * Return: 1 if @p is an idle task. 0 otherwise.
1897 */
1898static __always_inline bool is_idle_task(const struct task_struct *p)
1899{
1900 return !!(p->flags & PF_IDLE);
1901}
1902
1903extern struct task_struct *curr_task(int cpu);
1904extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1905
1906void yield(void);
1907
1908union thread_union {
1909#ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1910 struct task_struct task;
1911#endif
1912#ifndef CONFIG_THREAD_INFO_IN_TASK
1913 struct thread_info thread_info;
1914#endif
1915 unsigned long stack[THREAD_SIZE/sizeof(long)];
1916};
1917
1918#ifndef CONFIG_THREAD_INFO_IN_TASK
1919extern struct thread_info init_thread_info;
1920#endif
1921
1922extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1923
1924#ifdef CONFIG_THREAD_INFO_IN_TASK
1925# define task_thread_info(task) (&(task)->thread_info)
1926#elif !defined(__HAVE_THREAD_FUNCTIONS)
1927# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1928#endif
1929
1930/*
1931 * find a task by one of its numerical ids
1932 *
1933 * find_task_by_pid_ns():
1934 * finds a task by its pid in the specified namespace
1935 * find_task_by_vpid():
1936 * finds a task by its virtual pid
1937 *
1938 * see also find_vpid() etc in include/linux/pid.h
1939 */
1940
1941extern struct task_struct *find_task_by_vpid(pid_t nr);
1942extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1943
1944/*
1945 * find a task by its virtual pid and get the task struct
1946 */
1947extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1948
1949extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1950extern int wake_up_process(struct task_struct *tsk);
1951extern void wake_up_new_task(struct task_struct *tsk);
1952
1953#ifdef CONFIG_SMP
1954extern void kick_process(struct task_struct *tsk);
1955#else
1956static inline void kick_process(struct task_struct *tsk) { }
1957#endif
1958
1959extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1960
1961static inline void set_task_comm(struct task_struct *tsk, const char *from)
1962{
1963 __set_task_comm(tsk, from, false);
1964}
1965
1966extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1967#define get_task_comm(buf, tsk) ({ \
1968 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1969 __get_task_comm(buf, sizeof(buf), tsk); \
1970})
1971
1972#ifdef CONFIG_SMP
1973static __always_inline void scheduler_ipi(void)
1974{
1975 /*
1976 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1977 * TIF_NEED_RESCHED remotely (for the first time) will also send
1978 * this IPI.
1979 */
1980 preempt_fold_need_resched();
1981}
1982extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1983#else
1984static inline void scheduler_ipi(void) { }
1985static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1986{
1987 return 1;
1988}
1989#endif
1990
1991/*
1992 * Set thread flags in other task's structures.
1993 * See asm/thread_info.h for TIF_xxxx flags available:
1994 */
1995static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1996{
1997 set_ti_thread_flag(task_thread_info(tsk), flag);
1998}
1999
2000static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2001{
2002 clear_ti_thread_flag(task_thread_info(tsk), flag);
2003}
2004
2005static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2006 bool value)
2007{
2008 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2009}
2010
2011static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2012{
2013 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2014}
2015
2016static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2017{
2018 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2019}
2020
2021static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2022{
2023 return test_ti_thread_flag(task_thread_info(tsk), flag);
2024}
2025
2026static inline void set_tsk_need_resched(struct task_struct *tsk)
2027{
2028 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2029}
2030
2031static inline void clear_tsk_need_resched(struct task_struct *tsk)
2032{
2033 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2034}
2035
2036static inline int test_tsk_need_resched(struct task_struct *tsk)
2037{
2038 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2039}
2040
2041/*
2042 * cond_resched() and cond_resched_lock(): latency reduction via
2043 * explicit rescheduling in places that are safe. The return
2044 * value indicates whether a reschedule was done in fact.
2045 * cond_resched_lock() will drop the spinlock before scheduling,
2046 */
2047#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2048extern int __cond_resched(void);
2049
2050#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2051
2052DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2053
2054static __always_inline int _cond_resched(void)
2055{
2056 return static_call_mod(cond_resched)();
2057}
2058
2059#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2060extern int dynamic_cond_resched(void);
2061
2062static __always_inline int _cond_resched(void)
2063{
2064 return dynamic_cond_resched();
2065}
2066
2067#else
2068
2069static inline int _cond_resched(void)
2070{
2071 return __cond_resched();
2072}
2073
2074#endif /* CONFIG_PREEMPT_DYNAMIC */
2075
2076#else
2077
2078static inline int _cond_resched(void) { return 0; }
2079
2080#endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2081
2082#define cond_resched() ({ \
2083 __might_resched(__FILE__, __LINE__, 0); \
2084 _cond_resched(); \
2085})
2086
2087extern int __cond_resched_lock(spinlock_t *lock);
2088extern int __cond_resched_rwlock_read(rwlock_t *lock);
2089extern int __cond_resched_rwlock_write(rwlock_t *lock);
2090
2091#define MIGHT_RESCHED_RCU_SHIFT 8
2092#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2093
2094#ifndef CONFIG_PREEMPT_RT
2095/*
2096 * Non RT kernels have an elevated preempt count due to the held lock,
2097 * but are not allowed to be inside a RCU read side critical section
2098 */
2099# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2100#else
2101/*
2102 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2103 * cond_resched*lock() has to take that into account because it checks for
2104 * preempt_count() and rcu_preempt_depth().
2105 */
2106# define PREEMPT_LOCK_RESCHED_OFFSETS \
2107 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2108#endif
2109
2110#define cond_resched_lock(lock) ({ \
2111 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2112 __cond_resched_lock(lock); \
2113})
2114
2115#define cond_resched_rwlock_read(lock) ({ \
2116 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2117 __cond_resched_rwlock_read(lock); \
2118})
2119
2120#define cond_resched_rwlock_write(lock) ({ \
2121 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2122 __cond_resched_rwlock_write(lock); \
2123})
2124
2125static inline void cond_resched_rcu(void)
2126{
2127#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2128 rcu_read_unlock();
2129 cond_resched();
2130 rcu_read_lock();
2131#endif
2132}
2133
2134#ifdef CONFIG_PREEMPT_DYNAMIC
2135
2136extern bool preempt_model_none(void);
2137extern bool preempt_model_voluntary(void);
2138extern bool preempt_model_full(void);
2139
2140#else
2141
2142static inline bool preempt_model_none(void)
2143{
2144 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2145}
2146static inline bool preempt_model_voluntary(void)
2147{
2148 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2149}
2150static inline bool preempt_model_full(void)
2151{
2152 return IS_ENABLED(CONFIG_PREEMPT);
2153}
2154
2155#endif
2156
2157static inline bool preempt_model_rt(void)
2158{
2159 return IS_ENABLED(CONFIG_PREEMPT_RT);
2160}
2161
2162/*
2163 * Does the preemption model allow non-cooperative preemption?
2164 *
2165 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2166 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2167 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2168 * PREEMPT_NONE model.
2169 */
2170static inline bool preempt_model_preemptible(void)
2171{
2172 return preempt_model_full() || preempt_model_rt();
2173}
2174
2175/*
2176 * Does a critical section need to be broken due to another
2177 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2178 * but a general need for low latency)
2179 */
2180static inline int spin_needbreak(spinlock_t *lock)
2181{
2182#ifdef CONFIG_PREEMPTION
2183 return spin_is_contended(lock);
2184#else
2185 return 0;
2186#endif
2187}
2188
2189/*
2190 * Check if a rwlock is contended.
2191 * Returns non-zero if there is another task waiting on the rwlock.
2192 * Returns zero if the lock is not contended or the system / underlying
2193 * rwlock implementation does not support contention detection.
2194 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2195 * for low latency.
2196 */
2197static inline int rwlock_needbreak(rwlock_t *lock)
2198{
2199#ifdef CONFIG_PREEMPTION
2200 return rwlock_is_contended(lock);
2201#else
2202 return 0;
2203#endif
2204}
2205
2206static __always_inline bool need_resched(void)
2207{
2208 return unlikely(tif_need_resched());
2209}
2210
2211/*
2212 * Wrappers for p->thread_info->cpu access. No-op on UP.
2213 */
2214#ifdef CONFIG_SMP
2215
2216static inline unsigned int task_cpu(const struct task_struct *p)
2217{
2218 return READ_ONCE(task_thread_info(p)->cpu);
2219}
2220
2221extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2222
2223#else
2224
2225static inline unsigned int task_cpu(const struct task_struct *p)
2226{
2227 return 0;
2228}
2229
2230static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2231{
2232}
2233
2234#endif /* CONFIG_SMP */
2235
2236extern bool sched_task_on_rq(struct task_struct *p);
2237extern unsigned long get_wchan(struct task_struct *p);
2238extern struct task_struct *cpu_curr_snapshot(int cpu);
2239
2240/*
2241 * In order to reduce various lock holder preemption latencies provide an
2242 * interface to see if a vCPU is currently running or not.
2243 *
2244 * This allows us to terminate optimistic spin loops and block, analogous to
2245 * the native optimistic spin heuristic of testing if the lock owner task is
2246 * running or not.
2247 */
2248#ifndef vcpu_is_preempted
2249static inline bool vcpu_is_preempted(int cpu)
2250{
2251 return false;
2252}
2253#endif
2254
2255extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2256extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2257
2258#ifndef TASK_SIZE_OF
2259#define TASK_SIZE_OF(tsk) TASK_SIZE
2260#endif
2261
2262#ifdef CONFIG_SMP
2263static inline bool owner_on_cpu(struct task_struct *owner)
2264{
2265 /*
2266 * As lock holder preemption issue, we both skip spinning if
2267 * task is not on cpu or its cpu is preempted
2268 */
2269 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2270}
2271
2272/* Returns effective CPU energy utilization, as seen by the scheduler */
2273unsigned long sched_cpu_util(int cpu);
2274#endif /* CONFIG_SMP */
2275
2276#ifdef CONFIG_RSEQ
2277
2278/*
2279 * Map the event mask on the user-space ABI enum rseq_cs_flags
2280 * for direct mask checks.
2281 */
2282enum rseq_event_mask_bits {
2283 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2284 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2285 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2286};
2287
2288enum rseq_event_mask {
2289 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2290 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2291 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2292};
2293
2294static inline void rseq_set_notify_resume(struct task_struct *t)
2295{
2296 if (t->rseq)
2297 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2298}
2299
2300void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2301
2302static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2303 struct pt_regs *regs)
2304{
2305 if (current->rseq)
2306 __rseq_handle_notify_resume(ksig, regs);
2307}
2308
2309static inline void rseq_signal_deliver(struct ksignal *ksig,
2310 struct pt_regs *regs)
2311{
2312 preempt_disable();
2313 __set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2314 preempt_enable();
2315 rseq_handle_notify_resume(ksig, regs);
2316}
2317
2318/* rseq_preempt() requires preemption to be disabled. */
2319static inline void rseq_preempt(struct task_struct *t)
2320{
2321 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2322 rseq_set_notify_resume(t);
2323}
2324
2325/* rseq_migrate() requires preemption to be disabled. */
2326static inline void rseq_migrate(struct task_struct *t)
2327{
2328 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2329 rseq_set_notify_resume(t);
2330}
2331
2332/*
2333 * If parent process has a registered restartable sequences area, the
2334 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2335 */
2336static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2337{
2338 if (clone_flags & CLONE_VM) {
2339 t->rseq = NULL;
2340 t->rseq_sig = 0;
2341 t->rseq_event_mask = 0;
2342 } else {
2343 t->rseq = current->rseq;
2344 t->rseq_sig = current->rseq_sig;
2345 t->rseq_event_mask = current->rseq_event_mask;
2346 }
2347}
2348
2349static inline void rseq_execve(struct task_struct *t)
2350{
2351 t->rseq = NULL;
2352 t->rseq_sig = 0;
2353 t->rseq_event_mask = 0;
2354}
2355
2356#else
2357
2358static inline void rseq_set_notify_resume(struct task_struct *t)
2359{
2360}
2361static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2362 struct pt_regs *regs)
2363{
2364}
2365static inline void rseq_signal_deliver(struct ksignal *ksig,
2366 struct pt_regs *regs)
2367{
2368}
2369static inline void rseq_preempt(struct task_struct *t)
2370{
2371}
2372static inline void rseq_migrate(struct task_struct *t)
2373{
2374}
2375static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2376{
2377}
2378static inline void rseq_execve(struct task_struct *t)
2379{
2380}
2381
2382#endif
2383
2384#ifdef CONFIG_DEBUG_RSEQ
2385
2386void rseq_syscall(struct pt_regs *regs);
2387
2388#else
2389
2390static inline void rseq_syscall(struct pt_regs *regs)
2391{
2392}
2393
2394#endif
2395
2396#ifdef CONFIG_SCHED_CORE
2397extern void sched_core_free(struct task_struct *tsk);
2398extern void sched_core_fork(struct task_struct *p);
2399extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2400 unsigned long uaddr);
2401#else
2402static inline void sched_core_free(struct task_struct *tsk) { }
2403static inline void sched_core_fork(struct task_struct *p) { }
2404#endif
2405
2406extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2407
2408#endif
2409

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