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/kmsan_types.h>
18#include <linux/mutex.h>
19#include <linux/plist.h>
20#include <linux/hrtimer.h>
21#include <linux/irqflags.h>
22#include <linux/seccomp.h>
23#include <linux/nodemask.h>
24#include <linux/rcupdate.h>
25#include <linux/refcount.h>
26#include <linux/resource.h>
27#include <linux/latencytop.h>
28#include <linux/sched/prio.h>
29#include <linux/sched/types.h>
30#include <linux/signal_types.h>
31#include <linux/syscall_user_dispatch.h>
32#include <linux/mm_types_task.h>
33#include <linux/task_io_accounting.h>
34#include <linux/posix-timers.h>
35#include <linux/rseq.h>
36#include <linux/seqlock.h>
37#include <linux/kcsan.h>
38#include <linux/rv.h>
39#include <linux/livepatch_sched.h>
40#include <asm/kmap_size.h>
41
42/* task_struct member predeclarations (sorted alphabetically): */
43struct audit_context;
44struct bio_list;
45struct blk_plug;
46struct bpf_local_storage;
47struct bpf_run_ctx;
48struct capture_control;
49struct cfs_rq;
50struct fs_struct;
51struct futex_pi_state;
52struct io_context;
53struct io_uring_task;
54struct mempolicy;
55struct nameidata;
56struct nsproxy;
57struct perf_event_context;
58struct pid_namespace;
59struct pipe_inode_info;
60struct rcu_node;
61struct reclaim_state;
62struct robust_list_head;
63struct root_domain;
64struct rq;
65struct sched_attr;
66struct seq_file;
67struct sighand_struct;
68struct signal_struct;
69struct task_delay_info;
70struct task_group;
71struct user_event_mm;
72
73/*
74 * Task state bitmask. NOTE! These bits are also
75 * encoded in fs/proc/array.c: get_task_state().
76 *
77 * We have two separate sets of flags: task->__state
78 * is about runnability, while task->exit_state are
79 * about the task exiting. Confusing, but this way
80 * modifying one set can't modify the other one by
81 * mistake.
82 */
83
84/* Used in tsk->__state: */
85#define TASK_RUNNING 0x00000000
86#define TASK_INTERRUPTIBLE 0x00000001
87#define TASK_UNINTERRUPTIBLE 0x00000002
88#define __TASK_STOPPED 0x00000004
89#define __TASK_TRACED 0x00000008
90/* Used in tsk->exit_state: */
91#define EXIT_DEAD 0x00000010
92#define EXIT_ZOMBIE 0x00000020
93#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
94/* Used in tsk->__state again: */
95#define TASK_PARKED 0x00000040
96#define TASK_DEAD 0x00000080
97#define TASK_WAKEKILL 0x00000100
98#define TASK_WAKING 0x00000200
99#define TASK_NOLOAD 0x00000400
100#define TASK_NEW 0x00000800
101#define TASK_RTLOCK_WAIT 0x00001000
102#define TASK_FREEZABLE 0x00002000
103#define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
104#define TASK_FROZEN 0x00008000
105#define TASK_STATE_MAX 0x00010000
106
107#define TASK_ANY (TASK_STATE_MAX-1)
108
109/*
110 * DO NOT ADD ANY NEW USERS !
111 */
112#define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
113
114/* Convenience macros for the sake of set_current_state: */
115#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
116#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
117#define TASK_TRACED __TASK_TRACED
118
119#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
120
121/* Convenience macros for the sake of wake_up(): */
122#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
123
124/* get_task_state(): */
125#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
126 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
127 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
128 TASK_PARKED)
129
130#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
131
132#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
133#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
134#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
135
136/*
137 * Special states are those that do not use the normal wait-loop pattern. See
138 * the comment with set_special_state().
139 */
140#define is_special_task_state(state) \
141 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
142
143#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
144# define debug_normal_state_change(state_value) \
145 do { \
146 WARN_ON_ONCE(is_special_task_state(state_value)); \
147 current->task_state_change = _THIS_IP_; \
148 } while (0)
149
150# define debug_special_state_change(state_value) \
151 do { \
152 WARN_ON_ONCE(!is_special_task_state(state_value)); \
153 current->task_state_change = _THIS_IP_; \
154 } while (0)
155
156# define debug_rtlock_wait_set_state() \
157 do { \
158 current->saved_state_change = current->task_state_change;\
159 current->task_state_change = _THIS_IP_; \
160 } while (0)
161
162# define debug_rtlock_wait_restore_state() \
163 do { \
164 current->task_state_change = current->saved_state_change;\
165 } while (0)
166
167#else
168# define debug_normal_state_change(cond) do { } while (0)
169# define debug_special_state_change(cond) do { } while (0)
170# define debug_rtlock_wait_set_state() do { } while (0)
171# define debug_rtlock_wait_restore_state() do { } while (0)
172#endif
173
174/*
175 * set_current_state() includes a barrier so that the write of current->__state
176 * is correctly serialised wrt the caller's subsequent test of whether to
177 * actually sleep:
178 *
179 * for (;;) {
180 * set_current_state(TASK_UNINTERRUPTIBLE);
181 * if (CONDITION)
182 * break;
183 *
184 * schedule();
185 * }
186 * __set_current_state(TASK_RUNNING);
187 *
188 * If the caller does not need such serialisation (because, for instance, the
189 * CONDITION test and condition change and wakeup are under the same lock) then
190 * use __set_current_state().
191 *
192 * The above is typically ordered against the wakeup, which does:
193 *
194 * CONDITION = 1;
195 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
196 *
197 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
198 * accessing p->__state.
199 *
200 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
201 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
202 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
203 *
204 * However, with slightly different timing the wakeup TASK_RUNNING store can
205 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
206 * a problem either because that will result in one extra go around the loop
207 * and our @cond test will save the day.
208 *
209 * Also see the comments of try_to_wake_up().
210 */
211#define __set_current_state(state_value) \
212 do { \
213 debug_normal_state_change((state_value)); \
214 WRITE_ONCE(current->__state, (state_value)); \
215 } while (0)
216
217#define set_current_state(state_value) \
218 do { \
219 debug_normal_state_change((state_value)); \
220 smp_store_mb(current->__state, (state_value)); \
221 } while (0)
222
223/*
224 * set_special_state() should be used for those states when the blocking task
225 * can not use the regular condition based wait-loop. In that case we must
226 * serialize against wakeups such that any possible in-flight TASK_RUNNING
227 * stores will not collide with our state change.
228 */
229#define set_special_state(state_value) \
230 do { \
231 unsigned long flags; /* may shadow */ \
232 \
233 raw_spin_lock_irqsave(&current->pi_lock, flags); \
234 debug_special_state_change((state_value)); \
235 WRITE_ONCE(current->__state, (state_value)); \
236 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
237 } while (0)
238
239/*
240 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
241 *
242 * RT's spin/rwlock substitutions are state preserving. The state of the
243 * task when blocking on the lock is saved in task_struct::saved_state and
244 * restored after the lock has been acquired. These operations are
245 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
246 * lock related wakeups while the task is blocked on the lock are
247 * redirected to operate on task_struct::saved_state to ensure that these
248 * are not dropped. On restore task_struct::saved_state is set to
249 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
250 *
251 * The lock operation looks like this:
252 *
253 * current_save_and_set_rtlock_wait_state();
254 * for (;;) {
255 * if (try_lock())
256 * break;
257 * raw_spin_unlock_irq(&lock->wait_lock);
258 * schedule_rtlock();
259 * raw_spin_lock_irq(&lock->wait_lock);
260 * set_current_state(TASK_RTLOCK_WAIT);
261 * }
262 * current_restore_rtlock_saved_state();
263 */
264#define current_save_and_set_rtlock_wait_state() \
265 do { \
266 lockdep_assert_irqs_disabled(); \
267 raw_spin_lock(&current->pi_lock); \
268 current->saved_state = current->__state; \
269 debug_rtlock_wait_set_state(); \
270 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
271 raw_spin_unlock(&current->pi_lock); \
272 } while (0);
273
274#define current_restore_rtlock_saved_state() \
275 do { \
276 lockdep_assert_irqs_disabled(); \
277 raw_spin_lock(&current->pi_lock); \
278 debug_rtlock_wait_restore_state(); \
279 WRITE_ONCE(current->__state, current->saved_state); \
280 current->saved_state = TASK_RUNNING; \
281 raw_spin_unlock(&current->pi_lock); \
282 } while (0);
283
284#define get_current_state() READ_ONCE(current->__state)
285
286/*
287 * Define the task command name length as enum, then it can be visible to
288 * BPF programs.
289 */
290enum {
291 TASK_COMM_LEN = 16,
292};
293
294extern void scheduler_tick(void);
295
296#define MAX_SCHEDULE_TIMEOUT LONG_MAX
297
298extern long schedule_timeout(long timeout);
299extern long schedule_timeout_interruptible(long timeout);
300extern long schedule_timeout_killable(long timeout);
301extern long schedule_timeout_uninterruptible(long timeout);
302extern long schedule_timeout_idle(long timeout);
303asmlinkage void schedule(void);
304extern void schedule_preempt_disabled(void);
305asmlinkage void preempt_schedule_irq(void);
306#ifdef CONFIG_PREEMPT_RT
307 extern void schedule_rtlock(void);
308#endif
309
310extern int __must_check io_schedule_prepare(void);
311extern void io_schedule_finish(int token);
312extern long io_schedule_timeout(long timeout);
313extern void io_schedule(void);
314
315/**
316 * struct prev_cputime - snapshot of system and user cputime
317 * @utime: time spent in user mode
318 * @stime: time spent in system mode
319 * @lock: protects the above two fields
320 *
321 * Stores previous user/system time values such that we can guarantee
322 * monotonicity.
323 */
324struct prev_cputime {
325#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
326 u64 utime;
327 u64 stime;
328 raw_spinlock_t lock;
329#endif
330};
331
332enum vtime_state {
333 /* Task is sleeping or running in a CPU with VTIME inactive: */
334 VTIME_INACTIVE = 0,
335 /* Task is idle */
336 VTIME_IDLE,
337 /* Task runs in kernelspace in a CPU with VTIME active: */
338 VTIME_SYS,
339 /* Task runs in userspace in a CPU with VTIME active: */
340 VTIME_USER,
341 /* Task runs as guests in a CPU with VTIME active: */
342 VTIME_GUEST,
343};
344
345struct vtime {
346 seqcount_t seqcount;
347 unsigned long long starttime;
348 enum vtime_state state;
349 unsigned int cpu;
350 u64 utime;
351 u64 stime;
352 u64 gtime;
353};
354
355/*
356 * Utilization clamp constraints.
357 * @UCLAMP_MIN: Minimum utilization
358 * @UCLAMP_MAX: Maximum utilization
359 * @UCLAMP_CNT: Utilization clamp constraints count
360 */
361enum uclamp_id {
362 UCLAMP_MIN = 0,
363 UCLAMP_MAX,
364 UCLAMP_CNT
365};
366
367#ifdef CONFIG_SMP
368extern struct root_domain def_root_domain;
369extern struct mutex sched_domains_mutex;
370#endif
371
372struct sched_param {
373 int sched_priority;
374};
375
376struct sched_info {
377#ifdef CONFIG_SCHED_INFO
378 /* Cumulative counters: */
379
380 /* # of times we have run on this CPU: */
381 unsigned long pcount;
382
383 /* Time spent waiting on a runqueue: */
384 unsigned long long run_delay;
385
386 /* Timestamps: */
387
388 /* When did we last run on a CPU? */
389 unsigned long long last_arrival;
390
391 /* When were we last queued to run? */
392 unsigned long long last_queued;
393
394#endif /* CONFIG_SCHED_INFO */
395};
396
397/*
398 * Integer metrics need fixed point arithmetic, e.g., sched/fair
399 * has a few: load, load_avg, util_avg, freq, and capacity.
400 *
401 * We define a basic fixed point arithmetic range, and then formalize
402 * all these metrics based on that basic range.
403 */
404# define SCHED_FIXEDPOINT_SHIFT 10
405# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
406
407/* Increase resolution of cpu_capacity calculations */
408# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
409# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
410
411struct load_weight {
412 unsigned long weight;
413 u32 inv_weight;
414};
415
416/**
417 * struct util_est - Estimation utilization of FAIR tasks
418 * @enqueued: instantaneous estimated utilization of a task/cpu
419 * @ewma: the Exponential Weighted Moving Average (EWMA)
420 * utilization of a task
421 *
422 * Support data structure to track an Exponential Weighted Moving Average
423 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
424 * average each time a task completes an activation. Sample's weight is chosen
425 * so that the EWMA will be relatively insensitive to transient changes to the
426 * task's workload.
427 *
428 * The enqueued attribute has a slightly different meaning for tasks and cpus:
429 * - task: the task's util_avg at last task dequeue time
430 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
431 * Thus, the util_est.enqueued of a task represents the contribution on the
432 * estimated utilization of the CPU where that task is currently enqueued.
433 *
434 * Only for tasks we track a moving average of the past instantaneous
435 * estimated utilization. This allows to absorb sporadic drops in utilization
436 * of an otherwise almost periodic task.
437 *
438 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
439 * updates. When a task is dequeued, its util_est should not be updated if its
440 * util_avg has not been updated in the meantime.
441 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
442 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
443 * for a task) it is safe to use MSB.
444 */
445struct util_est {
446 unsigned int enqueued;
447 unsigned int ewma;
448#define UTIL_EST_WEIGHT_SHIFT 2
449#define UTIL_AVG_UNCHANGED 0x80000000
450} __attribute__((__aligned__(sizeof(u64))));
451
452/*
453 * The load/runnable/util_avg accumulates an infinite geometric series
454 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
455 *
456 * [load_avg definition]
457 *
458 * load_avg = runnable% * scale_load_down(load)
459 *
460 * [runnable_avg definition]
461 *
462 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
463 *
464 * [util_avg definition]
465 *
466 * util_avg = running% * SCHED_CAPACITY_SCALE
467 *
468 * where runnable% is the time ratio that a sched_entity is runnable and
469 * running% the time ratio that a sched_entity is running.
470 *
471 * For cfs_rq, they are the aggregated values of all runnable and blocked
472 * sched_entities.
473 *
474 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
475 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
476 * for computing those signals (see update_rq_clock_pelt())
477 *
478 * N.B., the above ratios (runnable% and running%) themselves are in the
479 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
480 * to as large a range as necessary. This is for example reflected by
481 * util_avg's SCHED_CAPACITY_SCALE.
482 *
483 * [Overflow issue]
484 *
485 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
486 * with the highest load (=88761), always runnable on a single cfs_rq,
487 * and should not overflow as the number already hits PID_MAX_LIMIT.
488 *
489 * For all other cases (including 32-bit kernels), struct load_weight's
490 * weight will overflow first before we do, because:
491 *
492 * Max(load_avg) <= Max(load.weight)
493 *
494 * Then it is the load_weight's responsibility to consider overflow
495 * issues.
496 */
497struct sched_avg {
498 u64 last_update_time;
499 u64 load_sum;
500 u64 runnable_sum;
501 u32 util_sum;
502 u32 period_contrib;
503 unsigned long load_avg;
504 unsigned long runnable_avg;
505 unsigned long util_avg;
506 struct util_est util_est;
507} ____cacheline_aligned;
508
509struct sched_statistics {
510#ifdef CONFIG_SCHEDSTATS
511 u64 wait_start;
512 u64 wait_max;
513 u64 wait_count;
514 u64 wait_sum;
515 u64 iowait_count;
516 u64 iowait_sum;
517
518 u64 sleep_start;
519 u64 sleep_max;
520 s64 sum_sleep_runtime;
521
522 u64 block_start;
523 u64 block_max;
524 s64 sum_block_runtime;
525
526 u64 exec_max;
527 u64 slice_max;
528
529 u64 nr_migrations_cold;
530 u64 nr_failed_migrations_affine;
531 u64 nr_failed_migrations_running;
532 u64 nr_failed_migrations_hot;
533 u64 nr_forced_migrations;
534
535 u64 nr_wakeups;
536 u64 nr_wakeups_sync;
537 u64 nr_wakeups_migrate;
538 u64 nr_wakeups_local;
539 u64 nr_wakeups_remote;
540 u64 nr_wakeups_affine;
541 u64 nr_wakeups_affine_attempts;
542 u64 nr_wakeups_passive;
543 u64 nr_wakeups_idle;
544
545#ifdef CONFIG_SCHED_CORE
546 u64 core_forceidle_sum;
547#endif
548#endif /* CONFIG_SCHEDSTATS */
549} ____cacheline_aligned;
550
551struct sched_entity {
552 /* For load-balancing: */
553 struct load_weight load;
554 struct rb_node run_node;
555 u64 deadline;
556 u64 min_deadline;
557
558 struct list_head group_node;
559 unsigned int on_rq;
560
561 u64 exec_start;
562 u64 sum_exec_runtime;
563 u64 prev_sum_exec_runtime;
564 u64 vruntime;
565 s64 vlag;
566 u64 slice;
567
568 u64 nr_migrations;
569
570#ifdef CONFIG_FAIR_GROUP_SCHED
571 int depth;
572 struct sched_entity *parent;
573 /* rq on which this entity is (to be) queued: */
574 struct cfs_rq *cfs_rq;
575 /* rq "owned" by this entity/group: */
576 struct cfs_rq *my_q;
577 /* cached value of my_q->h_nr_running */
578 unsigned long runnable_weight;
579#endif
580
581#ifdef CONFIG_SMP
582 /*
583 * Per entity load average tracking.
584 *
585 * Put into separate cache line so it does not
586 * collide with read-mostly values above.
587 */
588 struct sched_avg avg;
589#endif
590};
591
592struct sched_rt_entity {
593 struct list_head run_list;
594 unsigned long timeout;
595 unsigned long watchdog_stamp;
596 unsigned int time_slice;
597 unsigned short on_rq;
598 unsigned short on_list;
599
600 struct sched_rt_entity *back;
601#ifdef CONFIG_RT_GROUP_SCHED
602 struct sched_rt_entity *parent;
603 /* rq on which this entity is (to be) queued: */
604 struct rt_rq *rt_rq;
605 /* rq "owned" by this entity/group: */
606 struct rt_rq *my_q;
607#endif
608} __randomize_layout;
609
610struct sched_dl_entity {
611 struct rb_node rb_node;
612
613 /*
614 * Original scheduling parameters. Copied here from sched_attr
615 * during sched_setattr(), they will remain the same until
616 * the next sched_setattr().
617 */
618 u64 dl_runtime; /* Maximum runtime for each instance */
619 u64 dl_deadline; /* Relative deadline of each instance */
620 u64 dl_period; /* Separation of two instances (period) */
621 u64 dl_bw; /* dl_runtime / dl_period */
622 u64 dl_density; /* dl_runtime / dl_deadline */
623
624 /*
625 * Actual scheduling parameters. Initialized with the values above,
626 * they are continuously updated during task execution. Note that
627 * the remaining runtime could be < 0 in case we are in overrun.
628 */
629 s64 runtime; /* Remaining runtime for this instance */
630 u64 deadline; /* Absolute deadline for this instance */
631 unsigned int flags; /* Specifying the scheduler behaviour */
632
633 /*
634 * Some bool flags:
635 *
636 * @dl_throttled tells if we exhausted the runtime. If so, the
637 * task has to wait for a replenishment to be performed at the
638 * next firing of dl_timer.
639 *
640 * @dl_yielded tells if task gave up the CPU before consuming
641 * all its available runtime during the last job.
642 *
643 * @dl_non_contending tells if the task is inactive while still
644 * contributing to the active utilization. In other words, it
645 * indicates if the inactive timer has been armed and its handler
646 * has not been executed yet. This flag is useful to avoid race
647 * conditions between the inactive timer handler and the wakeup
648 * code.
649 *
650 * @dl_overrun tells if the task asked to be informed about runtime
651 * overruns.
652 */
653 unsigned int dl_throttled : 1;
654 unsigned int dl_yielded : 1;
655 unsigned int dl_non_contending : 1;
656 unsigned int dl_overrun : 1;
657
658 /*
659 * Bandwidth enforcement timer. Each -deadline task has its
660 * own bandwidth to be enforced, thus we need one timer per task.
661 */
662 struct hrtimer dl_timer;
663
664 /*
665 * Inactive timer, responsible for decreasing the active utilization
666 * at the "0-lag time". When a -deadline task blocks, it contributes
667 * to GRUB's active utilization until the "0-lag time", hence a
668 * timer is needed to decrease the active utilization at the correct
669 * time.
670 */
671 struct hrtimer inactive_timer;
672
673#ifdef CONFIG_RT_MUTEXES
674 /*
675 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
676 * pi_se points to the donor, otherwise points to the dl_se it belongs
677 * to (the original one/itself).
678 */
679 struct sched_dl_entity *pi_se;
680#endif
681};
682
683#ifdef CONFIG_UCLAMP_TASK
684/* Number of utilization clamp buckets (shorter alias) */
685#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
686
687/*
688 * Utilization clamp for a scheduling entity
689 * @value: clamp value "assigned" to a se
690 * @bucket_id: bucket index corresponding to the "assigned" value
691 * @active: the se is currently refcounted in a rq's bucket
692 * @user_defined: the requested clamp value comes from user-space
693 *
694 * The bucket_id is the index of the clamp bucket matching the clamp value
695 * which is pre-computed and stored to avoid expensive integer divisions from
696 * the fast path.
697 *
698 * The active bit is set whenever a task has got an "effective" value assigned,
699 * which can be different from the clamp value "requested" from user-space.
700 * This allows to know a task is refcounted in the rq's bucket corresponding
701 * to the "effective" bucket_id.
702 *
703 * The user_defined bit is set whenever a task has got a task-specific clamp
704 * value requested from userspace, i.e. the system defaults apply to this task
705 * just as a restriction. This allows to relax default clamps when a less
706 * restrictive task-specific value has been requested, thus allowing to
707 * implement a "nice" semantic. For example, a task running with a 20%
708 * default boost can still drop its own boosting to 0%.
709 */
710struct uclamp_se {
711 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
712 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
713 unsigned int active : 1;
714 unsigned int user_defined : 1;
715};
716#endif /* CONFIG_UCLAMP_TASK */
717
718union rcu_special {
719 struct {
720 u8 blocked;
721 u8 need_qs;
722 u8 exp_hint; /* Hint for performance. */
723 u8 need_mb; /* Readers need smp_mb(). */
724 } b; /* Bits. */
725 u32 s; /* Set of bits. */
726};
727
728enum perf_event_task_context {
729 perf_invalid_context = -1,
730 perf_hw_context = 0,
731 perf_sw_context,
732 perf_nr_task_contexts,
733};
734
735struct wake_q_node {
736 struct wake_q_node *next;
737};
738
739struct kmap_ctrl {
740#ifdef CONFIG_KMAP_LOCAL
741 int idx;
742 pte_t pteval[KM_MAX_IDX];
743#endif
744};
745
746struct task_struct {
747#ifdef CONFIG_THREAD_INFO_IN_TASK
748 /*
749 * For reasons of header soup (see current_thread_info()), this
750 * must be the first element of task_struct.
751 */
752 struct thread_info thread_info;
753#endif
754 unsigned int __state;
755
756 /* saved state for "spinlock sleepers" */
757 unsigned int saved_state;
758
759 /*
760 * This begins the randomizable portion of task_struct. Only
761 * scheduling-critical items should be added above here.
762 */
763 randomized_struct_fields_start
764
765 void *stack;
766 refcount_t usage;
767 /* Per task flags (PF_*), defined further below: */
768 unsigned int flags;
769 unsigned int ptrace;
770
771#ifdef CONFIG_SMP
772 int on_cpu;
773 struct __call_single_node wake_entry;
774 unsigned int wakee_flips;
775 unsigned long wakee_flip_decay_ts;
776 struct task_struct *last_wakee;
777
778 /*
779 * recent_used_cpu is initially set as the last CPU used by a task
780 * that wakes affine another task. Waker/wakee relationships can
781 * push tasks around a CPU where each wakeup moves to the next one.
782 * Tracking a recently used CPU allows a quick search for a recently
783 * used CPU that may be idle.
784 */
785 int recent_used_cpu;
786 int wake_cpu;
787#endif
788 int on_rq;
789
790 int prio;
791 int static_prio;
792 int normal_prio;
793 unsigned int rt_priority;
794
795 struct sched_entity se;
796 struct sched_rt_entity rt;
797 struct sched_dl_entity dl;
798 const struct sched_class *sched_class;
799
800#ifdef CONFIG_SCHED_CORE
801 struct rb_node core_node;
802 unsigned long core_cookie;
803 unsigned int core_occupation;
804#endif
805
806#ifdef CONFIG_CGROUP_SCHED
807 struct task_group *sched_task_group;
808#endif
809
810#ifdef CONFIG_UCLAMP_TASK
811 /*
812 * Clamp values requested for a scheduling entity.
813 * Must be updated with task_rq_lock() held.
814 */
815 struct uclamp_se uclamp_req[UCLAMP_CNT];
816 /*
817 * Effective clamp values used for a scheduling entity.
818 * Must be updated with task_rq_lock() held.
819 */
820 struct uclamp_se uclamp[UCLAMP_CNT];
821#endif
822
823 struct sched_statistics stats;
824
825#ifdef CONFIG_PREEMPT_NOTIFIERS
826 /* List of struct preempt_notifier: */
827 struct hlist_head preempt_notifiers;
828#endif
829
830#ifdef CONFIG_BLK_DEV_IO_TRACE
831 unsigned int btrace_seq;
832#endif
833
834 unsigned int policy;
835 int nr_cpus_allowed;
836 const cpumask_t *cpus_ptr;
837 cpumask_t *user_cpus_ptr;
838 cpumask_t cpus_mask;
839 void *migration_pending;
840#ifdef CONFIG_SMP
841 unsigned short migration_disabled;
842#endif
843 unsigned short migration_flags;
844
845#ifdef CONFIG_PREEMPT_RCU
846 int rcu_read_lock_nesting;
847 union rcu_special rcu_read_unlock_special;
848 struct list_head rcu_node_entry;
849 struct rcu_node *rcu_blocked_node;
850#endif /* #ifdef CONFIG_PREEMPT_RCU */
851
852#ifdef CONFIG_TASKS_RCU
853 unsigned long rcu_tasks_nvcsw;
854 u8 rcu_tasks_holdout;
855 u8 rcu_tasks_idx;
856 int rcu_tasks_idle_cpu;
857 struct list_head rcu_tasks_holdout_list;
858#endif /* #ifdef CONFIG_TASKS_RCU */
859
860#ifdef CONFIG_TASKS_TRACE_RCU
861 int trc_reader_nesting;
862 int trc_ipi_to_cpu;
863 union rcu_special trc_reader_special;
864 struct list_head trc_holdout_list;
865 struct list_head trc_blkd_node;
866 int trc_blkd_cpu;
867#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
868
869 struct sched_info sched_info;
870
871 struct list_head tasks;
872#ifdef CONFIG_SMP
873 struct plist_node pushable_tasks;
874 struct rb_node pushable_dl_tasks;
875#endif
876
877 struct mm_struct *mm;
878 struct mm_struct *active_mm;
879 struct address_space *faults_disabled_mapping;
880
881 int exit_state;
882 int exit_code;
883 int exit_signal;
884 /* The signal sent when the parent dies: */
885 int pdeath_signal;
886 /* JOBCTL_*, siglock protected: */
887 unsigned long jobctl;
888
889 /* Used for emulating ABI behavior of previous Linux versions: */
890 unsigned int personality;
891
892 /* Scheduler bits, serialized by scheduler locks: */
893 unsigned sched_reset_on_fork:1;
894 unsigned sched_contributes_to_load:1;
895 unsigned sched_migrated:1;
896
897 /* Force alignment to the next boundary: */
898 unsigned :0;
899
900 /* Unserialized, strictly 'current' */
901
902 /*
903 * This field must not be in the scheduler word above due to wakelist
904 * queueing no longer being serialized by p->on_cpu. However:
905 *
906 * p->XXX = X; ttwu()
907 * schedule() if (p->on_rq && ..) // false
908 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
909 * deactivate_task() ttwu_queue_wakelist())
910 * p->on_rq = 0; p->sched_remote_wakeup = Y;
911 *
912 * guarantees all stores of 'current' are visible before
913 * ->sched_remote_wakeup gets used, so it can be in this word.
914 */
915 unsigned sched_remote_wakeup:1;
916#ifdef CONFIG_RT_MUTEXES
917 unsigned sched_rt_mutex:1;
918#endif
919
920 /* Bit to tell LSMs we're in execve(): */
921 unsigned in_execve:1;
922 unsigned in_iowait:1;
923#ifndef TIF_RESTORE_SIGMASK
924 unsigned restore_sigmask:1;
925#endif
926#ifdef CONFIG_MEMCG
927 unsigned in_user_fault:1;
928#endif
929#ifdef CONFIG_LRU_GEN
930 /* whether the LRU algorithm may apply to this access */
931 unsigned in_lru_fault:1;
932#endif
933#ifdef CONFIG_COMPAT_BRK
934 unsigned brk_randomized:1;
935#endif
936#ifdef CONFIG_CGROUPS
937 /* disallow userland-initiated cgroup migration */
938 unsigned no_cgroup_migration:1;
939 /* task is frozen/stopped (used by the cgroup freezer) */
940 unsigned frozen:1;
941#endif
942#ifdef CONFIG_BLK_CGROUP
943 unsigned use_memdelay:1;
944#endif
945#ifdef CONFIG_PSI
946 /* Stalled due to lack of memory */
947 unsigned in_memstall:1;
948#endif
949#ifdef CONFIG_PAGE_OWNER
950 /* Used by page_owner=on to detect recursion in page tracking. */
951 unsigned in_page_owner:1;
952#endif
953#ifdef CONFIG_EVENTFD
954 /* Recursion prevention for eventfd_signal() */
955 unsigned in_eventfd:1;
956#endif
957#ifdef CONFIG_IOMMU_SVA
958 unsigned pasid_activated:1;
959#endif
960#ifdef CONFIG_CPU_SUP_INTEL
961 unsigned reported_split_lock:1;
962#endif
963#ifdef CONFIG_TASK_DELAY_ACCT
964 /* delay due to memory thrashing */
965 unsigned in_thrashing:1;
966#endif
967
968 unsigned long atomic_flags; /* Flags requiring atomic access. */
969
970 struct restart_block restart_block;
971
972 pid_t pid;
973 pid_t tgid;
974
975#ifdef CONFIG_STACKPROTECTOR
976 /* Canary value for the -fstack-protector GCC feature: */
977 unsigned long stack_canary;
978#endif
979 /*
980 * Pointers to the (original) parent process, youngest child, younger sibling,
981 * older sibling, respectively. (p->father can be replaced with
982 * p->real_parent->pid)
983 */
984
985 /* Real parent process: */
986 struct task_struct __rcu *real_parent;
987
988 /* Recipient of SIGCHLD, wait4() reports: */
989 struct task_struct __rcu *parent;
990
991 /*
992 * Children/sibling form the list of natural children:
993 */
994 struct list_head children;
995 struct list_head sibling;
996 struct task_struct *group_leader;
997
998 /*
999 * 'ptraced' is the list of tasks this task is using ptrace() on.
1000 *
1001 * This includes both natural children and PTRACE_ATTACH targets.
1002 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1003 */
1004 struct list_head ptraced;
1005 struct list_head ptrace_entry;
1006
1007 /* PID/PID hash table linkage. */
1008 struct pid *thread_pid;
1009 struct hlist_node pid_links[PIDTYPE_MAX];
1010 struct list_head thread_node;
1011
1012 struct completion *vfork_done;
1013
1014 /* CLONE_CHILD_SETTID: */
1015 int __user *set_child_tid;
1016
1017 /* CLONE_CHILD_CLEARTID: */
1018 int __user *clear_child_tid;
1019
1020 /* PF_KTHREAD | PF_IO_WORKER */
1021 void *worker_private;
1022
1023 u64 utime;
1024 u64 stime;
1025#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1026 u64 utimescaled;
1027 u64 stimescaled;
1028#endif
1029 u64 gtime;
1030 struct prev_cputime prev_cputime;
1031#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1032 struct vtime vtime;
1033#endif
1034
1035#ifdef CONFIG_NO_HZ_FULL
1036 atomic_t tick_dep_mask;
1037#endif
1038 /* Context switch counts: */
1039 unsigned long nvcsw;
1040 unsigned long nivcsw;
1041
1042 /* Monotonic time in nsecs: */
1043 u64 start_time;
1044
1045 /* Boot based time in nsecs: */
1046 u64 start_boottime;
1047
1048 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1049 unsigned long min_flt;
1050 unsigned long maj_flt;
1051
1052 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1053 struct posix_cputimers posix_cputimers;
1054
1055#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1056 struct posix_cputimers_work posix_cputimers_work;
1057#endif
1058
1059 /* Process credentials: */
1060
1061 /* Tracer's credentials at attach: */
1062 const struct cred __rcu *ptracer_cred;
1063
1064 /* Objective and real subjective task credentials (COW): */
1065 const struct cred __rcu *real_cred;
1066
1067 /* Effective (overridable) subjective task credentials (COW): */
1068 const struct cred __rcu *cred;
1069
1070#ifdef CONFIG_KEYS
1071 /* Cached requested key. */
1072 struct key *cached_requested_key;
1073#endif
1074
1075 /*
1076 * executable name, excluding path.
1077 *
1078 * - normally initialized setup_new_exec()
1079 * - access it with [gs]et_task_comm()
1080 * - lock it with task_lock()
1081 */
1082 char comm[TASK_COMM_LEN];
1083
1084 struct nameidata *nameidata;
1085
1086#ifdef CONFIG_SYSVIPC
1087 struct sysv_sem sysvsem;
1088 struct sysv_shm sysvshm;
1089#endif
1090#ifdef CONFIG_DETECT_HUNG_TASK
1091 unsigned long last_switch_count;
1092 unsigned long last_switch_time;
1093#endif
1094 /* Filesystem information: */
1095 struct fs_struct *fs;
1096
1097 /* Open file information: */
1098 struct files_struct *files;
1099
1100#ifdef CONFIG_IO_URING
1101 struct io_uring_task *io_uring;
1102#endif
1103
1104 /* Namespaces: */
1105 struct nsproxy *nsproxy;
1106
1107 /* Signal handlers: */
1108 struct signal_struct *signal;
1109 struct sighand_struct __rcu *sighand;
1110 sigset_t blocked;
1111 sigset_t real_blocked;
1112 /* Restored if set_restore_sigmask() was used: */
1113 sigset_t saved_sigmask;
1114 struct sigpending pending;
1115 unsigned long sas_ss_sp;
1116 size_t sas_ss_size;
1117 unsigned int sas_ss_flags;
1118
1119 struct callback_head *task_works;
1120
1121#ifdef CONFIG_AUDIT
1122#ifdef CONFIG_AUDITSYSCALL
1123 struct audit_context *audit_context;
1124#endif
1125 kuid_t loginuid;
1126 unsigned int sessionid;
1127#endif
1128 struct seccomp seccomp;
1129 struct syscall_user_dispatch syscall_dispatch;
1130
1131 /* Thread group tracking: */
1132 u64 parent_exec_id;
1133 u64 self_exec_id;
1134
1135 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1136 spinlock_t alloc_lock;
1137
1138 /* Protection of the PI data structures: */
1139 raw_spinlock_t pi_lock;
1140
1141 struct wake_q_node wake_q;
1142
1143#ifdef CONFIG_RT_MUTEXES
1144 /* PI waiters blocked on a rt_mutex held by this task: */
1145 struct rb_root_cached pi_waiters;
1146 /* Updated under owner's pi_lock and rq lock */
1147 struct task_struct *pi_top_task;
1148 /* Deadlock detection and priority inheritance handling: */
1149 struct rt_mutex_waiter *pi_blocked_on;
1150#endif
1151
1152#ifdef CONFIG_DEBUG_MUTEXES
1153 /* Mutex deadlock detection: */
1154 struct mutex_waiter *blocked_on;
1155#endif
1156
1157#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1158 int non_block_count;
1159#endif
1160
1161#ifdef CONFIG_TRACE_IRQFLAGS
1162 struct irqtrace_events irqtrace;
1163 unsigned int hardirq_threaded;
1164 u64 hardirq_chain_key;
1165 int softirqs_enabled;
1166 int softirq_context;
1167 int irq_config;
1168#endif
1169#ifdef CONFIG_PREEMPT_RT
1170 int softirq_disable_cnt;
1171#endif
1172
1173#ifdef CONFIG_LOCKDEP
1174# define MAX_LOCK_DEPTH 48UL
1175 u64 curr_chain_key;
1176 int lockdep_depth;
1177 unsigned int lockdep_recursion;
1178 struct held_lock held_locks[MAX_LOCK_DEPTH];
1179#endif
1180
1181#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1182 unsigned int in_ubsan;
1183#endif
1184
1185 /* Journalling filesystem info: */
1186 void *journal_info;
1187
1188 /* Stacked block device info: */
1189 struct bio_list *bio_list;
1190
1191 /* Stack plugging: */
1192 struct blk_plug *plug;
1193
1194 /* VM state: */
1195 struct reclaim_state *reclaim_state;
1196
1197 struct io_context *io_context;
1198
1199#ifdef CONFIG_COMPACTION
1200 struct capture_control *capture_control;
1201#endif
1202 /* Ptrace state: */
1203 unsigned long ptrace_message;
1204 kernel_siginfo_t *last_siginfo;
1205
1206 struct task_io_accounting ioac;
1207#ifdef CONFIG_PSI
1208 /* Pressure stall state */
1209 unsigned int psi_flags;
1210#endif
1211#ifdef CONFIG_TASK_XACCT
1212 /* Accumulated RSS usage: */
1213 u64 acct_rss_mem1;
1214 /* Accumulated virtual memory usage: */
1215 u64 acct_vm_mem1;
1216 /* stime + utime since last update: */
1217 u64 acct_timexpd;
1218#endif
1219#ifdef CONFIG_CPUSETS
1220 /* Protected by ->alloc_lock: */
1221 nodemask_t mems_allowed;
1222 /* Sequence number to catch updates: */
1223 seqcount_spinlock_t mems_allowed_seq;
1224 int cpuset_mem_spread_rotor;
1225 int cpuset_slab_spread_rotor;
1226#endif
1227#ifdef CONFIG_CGROUPS
1228 /* Control Group info protected by css_set_lock: */
1229 struct css_set __rcu *cgroups;
1230 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1231 struct list_head cg_list;
1232#endif
1233#ifdef CONFIG_X86_CPU_RESCTRL
1234 u32 closid;
1235 u32 rmid;
1236#endif
1237#ifdef CONFIG_FUTEX
1238 struct robust_list_head __user *robust_list;
1239#ifdef CONFIG_COMPAT
1240 struct compat_robust_list_head __user *compat_robust_list;
1241#endif
1242 struct list_head pi_state_list;
1243 struct futex_pi_state *pi_state_cache;
1244 struct mutex futex_exit_mutex;
1245 unsigned int futex_state;
1246#endif
1247#ifdef CONFIG_PERF_EVENTS
1248 struct perf_event_context *perf_event_ctxp;
1249 struct mutex perf_event_mutex;
1250 struct list_head perf_event_list;
1251#endif
1252#ifdef CONFIG_DEBUG_PREEMPT
1253 unsigned long preempt_disable_ip;
1254#endif
1255#ifdef CONFIG_NUMA
1256 /* Protected by alloc_lock: */
1257 struct mempolicy *mempolicy;
1258 short il_prev;
1259 short pref_node_fork;
1260#endif
1261#ifdef CONFIG_NUMA_BALANCING
1262 int numa_scan_seq;
1263 unsigned int numa_scan_period;
1264 unsigned int numa_scan_period_max;
1265 int numa_preferred_nid;
1266 unsigned long numa_migrate_retry;
1267 /* Migration stamp: */
1268 u64 node_stamp;
1269 u64 last_task_numa_placement;
1270 u64 last_sum_exec_runtime;
1271 struct callback_head numa_work;
1272
1273 /*
1274 * This pointer is only modified for current in syscall and
1275 * pagefault context (and for tasks being destroyed), so it can be read
1276 * from any of the following contexts:
1277 * - RCU read-side critical section
1278 * - current->numa_group from everywhere
1279 * - task's runqueue locked, task not running
1280 */
1281 struct numa_group __rcu *numa_group;
1282
1283 /*
1284 * numa_faults is an array split into four regions:
1285 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1286 * in this precise order.
1287 *
1288 * faults_memory: Exponential decaying average of faults on a per-node
1289 * basis. Scheduling placement decisions are made based on these
1290 * counts. The values remain static for the duration of a PTE scan.
1291 * faults_cpu: Track the nodes the process was running on when a NUMA
1292 * hinting fault was incurred.
1293 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1294 * during the current scan window. When the scan completes, the counts
1295 * in faults_memory and faults_cpu decay and these values are copied.
1296 */
1297 unsigned long *numa_faults;
1298 unsigned long total_numa_faults;
1299
1300 /*
1301 * numa_faults_locality tracks if faults recorded during the last
1302 * scan window were remote/local or failed to migrate. The task scan
1303 * period is adapted based on the locality of the faults with different
1304 * weights depending on whether they were shared or private faults
1305 */
1306 unsigned long numa_faults_locality[3];
1307
1308 unsigned long numa_pages_migrated;
1309#endif /* CONFIG_NUMA_BALANCING */
1310
1311#ifdef CONFIG_RSEQ
1312 struct rseq __user *rseq;
1313 u32 rseq_len;
1314 u32 rseq_sig;
1315 /*
1316 * RmW on rseq_event_mask must be performed atomically
1317 * with respect to preemption.
1318 */
1319 unsigned long rseq_event_mask;
1320#endif
1321
1322#ifdef CONFIG_SCHED_MM_CID
1323 int mm_cid; /* Current cid in mm */
1324 int last_mm_cid; /* Most recent cid in mm */
1325 int migrate_from_cpu;
1326 int mm_cid_active; /* Whether cid bitmap is active */
1327 struct callback_head cid_work;
1328#endif
1329
1330 struct tlbflush_unmap_batch tlb_ubc;
1331
1332 /* Cache last used pipe for splice(): */
1333 struct pipe_inode_info *splice_pipe;
1334
1335 struct page_frag task_frag;
1336
1337#ifdef CONFIG_TASK_DELAY_ACCT
1338 struct task_delay_info *delays;
1339#endif
1340
1341#ifdef CONFIG_FAULT_INJECTION
1342 int make_it_fail;
1343 unsigned int fail_nth;
1344#endif
1345 /*
1346 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1347 * balance_dirty_pages() for a dirty throttling pause:
1348 */
1349 int nr_dirtied;
1350 int nr_dirtied_pause;
1351 /* Start of a write-and-pause period: */
1352 unsigned long dirty_paused_when;
1353
1354#ifdef CONFIG_LATENCYTOP
1355 int latency_record_count;
1356 struct latency_record latency_record[LT_SAVECOUNT];
1357#endif
1358 /*
1359 * Time slack values; these are used to round up poll() and
1360 * select() etc timeout values. These are in nanoseconds.
1361 */
1362 u64 timer_slack_ns;
1363 u64 default_timer_slack_ns;
1364
1365#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1366 unsigned int kasan_depth;
1367#endif
1368
1369#ifdef CONFIG_KCSAN
1370 struct kcsan_ctx kcsan_ctx;
1371#ifdef CONFIG_TRACE_IRQFLAGS
1372 struct irqtrace_events kcsan_save_irqtrace;
1373#endif
1374#ifdef CONFIG_KCSAN_WEAK_MEMORY
1375 int kcsan_stack_depth;
1376#endif
1377#endif
1378
1379#ifdef CONFIG_KMSAN
1380 struct kmsan_ctx kmsan_ctx;
1381#endif
1382
1383#if IS_ENABLED(CONFIG_KUNIT)
1384 struct kunit *kunit_test;
1385#endif
1386
1387#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1388 /* Index of current stored address in ret_stack: */
1389 int curr_ret_stack;
1390 int curr_ret_depth;
1391
1392 /* Stack of return addresses for return function tracing: */
1393 struct ftrace_ret_stack *ret_stack;
1394
1395 /* Timestamp for last schedule: */
1396 unsigned long long ftrace_timestamp;
1397
1398 /*
1399 * Number of functions that haven't been traced
1400 * because of depth overrun:
1401 */
1402 atomic_t trace_overrun;
1403
1404 /* Pause tracing: */
1405 atomic_t tracing_graph_pause;
1406#endif
1407
1408#ifdef CONFIG_TRACING
1409 /* Bitmask and counter of trace recursion: */
1410 unsigned long trace_recursion;
1411#endif /* CONFIG_TRACING */
1412
1413#ifdef CONFIG_KCOV
1414 /* See kernel/kcov.c for more details. */
1415
1416 /* Coverage collection mode enabled for this task (0 if disabled): */
1417 unsigned int kcov_mode;
1418
1419 /* Size of the kcov_area: */
1420 unsigned int kcov_size;
1421
1422 /* Buffer for coverage collection: */
1423 void *kcov_area;
1424
1425 /* KCOV descriptor wired with this task or NULL: */
1426 struct kcov *kcov;
1427
1428 /* KCOV common handle for remote coverage collection: */
1429 u64 kcov_handle;
1430
1431 /* KCOV sequence number: */
1432 int kcov_sequence;
1433
1434 /* Collect coverage from softirq context: */
1435 unsigned int kcov_softirq;
1436#endif
1437
1438#ifdef CONFIG_MEMCG
1439 struct mem_cgroup *memcg_in_oom;
1440 gfp_t memcg_oom_gfp_mask;
1441 int memcg_oom_order;
1442
1443 /* Number of pages to reclaim on returning to userland: */
1444 unsigned int memcg_nr_pages_over_high;
1445
1446 /* Used by memcontrol for targeted memcg charge: */
1447 struct mem_cgroup *active_memcg;
1448#endif
1449
1450#ifdef CONFIG_MEMCG_KMEM
1451 struct obj_cgroup *objcg;
1452#endif
1453
1454#ifdef CONFIG_BLK_CGROUP
1455 struct gendisk *throttle_disk;
1456#endif
1457
1458#ifdef CONFIG_UPROBES
1459 struct uprobe_task *utask;
1460#endif
1461#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1462 unsigned int sequential_io;
1463 unsigned int sequential_io_avg;
1464#endif
1465 struct kmap_ctrl kmap_ctrl;
1466#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1467 unsigned long task_state_change;
1468# ifdef CONFIG_PREEMPT_RT
1469 unsigned long saved_state_change;
1470# endif
1471#endif
1472 struct rcu_head rcu;
1473 refcount_t rcu_users;
1474 int pagefault_disabled;
1475#ifdef CONFIG_MMU
1476 struct task_struct *oom_reaper_list;
1477 struct timer_list oom_reaper_timer;
1478#endif
1479#ifdef CONFIG_VMAP_STACK
1480 struct vm_struct *stack_vm_area;
1481#endif
1482#ifdef CONFIG_THREAD_INFO_IN_TASK
1483 /* A live task holds one reference: */
1484 refcount_t stack_refcount;
1485#endif
1486#ifdef CONFIG_LIVEPATCH
1487 int patch_state;
1488#endif
1489#ifdef CONFIG_SECURITY
1490 /* Used by LSM modules for access restriction: */
1491 void *security;
1492#endif
1493#ifdef CONFIG_BPF_SYSCALL
1494 /* Used by BPF task local storage */
1495 struct bpf_local_storage __rcu *bpf_storage;
1496 /* Used for BPF run context */
1497 struct bpf_run_ctx *bpf_ctx;
1498#endif
1499
1500#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1501 unsigned long lowest_stack;
1502 unsigned long prev_lowest_stack;
1503#endif
1504
1505#ifdef CONFIG_X86_MCE
1506 void __user *mce_vaddr;
1507 __u64 mce_kflags;
1508 u64 mce_addr;
1509 __u64 mce_ripv : 1,
1510 mce_whole_page : 1,
1511 __mce_reserved : 62;
1512 struct callback_head mce_kill_me;
1513 int mce_count;
1514#endif
1515
1516#ifdef CONFIG_KRETPROBES
1517 struct llist_head kretprobe_instances;
1518#endif
1519#ifdef CONFIG_RETHOOK
1520 struct llist_head rethooks;
1521#endif
1522
1523#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1524 /*
1525 * If L1D flush is supported on mm context switch
1526 * then we use this callback head to queue kill work
1527 * to kill tasks that are not running on SMT disabled
1528 * cores
1529 */
1530 struct callback_head l1d_flush_kill;
1531#endif
1532
1533#ifdef CONFIG_RV
1534 /*
1535 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1536 * If we find justification for more monitors, we can think
1537 * about adding more or developing a dynamic method. So far,
1538 * none of these are justified.
1539 */
1540 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1541#endif
1542
1543#ifdef CONFIG_USER_EVENTS
1544 struct user_event_mm *user_event_mm;
1545#endif
1546
1547 /*
1548 * New fields for task_struct should be added above here, so that
1549 * they are included in the randomized portion of task_struct.
1550 */
1551 randomized_struct_fields_end
1552
1553 /* CPU-specific state of this task: */
1554 struct thread_struct thread;
1555
1556 /*
1557 * WARNING: on x86, 'thread_struct' contains a variable-sized
1558 * structure. It *MUST* be at the end of 'task_struct'.
1559 *
1560 * Do not put anything below here!
1561 */
1562};
1563
1564static inline struct pid *task_pid(struct task_struct *task)
1565{
1566 return task->thread_pid;
1567}
1568
1569/*
1570 * the helpers to get the task's different pids as they are seen
1571 * from various namespaces
1572 *
1573 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1574 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1575 * current.
1576 * task_xid_nr_ns() : id seen from the ns specified;
1577 *
1578 * see also pid_nr() etc in include/linux/pid.h
1579 */
1580pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1581
1582static inline pid_t task_pid_nr(struct task_struct *tsk)
1583{
1584 return tsk->pid;
1585}
1586
1587static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1588{
1589 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_PID, ns);
1590}
1591
1592static inline pid_t task_pid_vnr(struct task_struct *tsk)
1593{
1594 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_PID, NULL);
1595}
1596
1597
1598static inline pid_t task_tgid_nr(struct task_struct *tsk)
1599{
1600 return tsk->tgid;
1601}
1602
1603/**
1604 * pid_alive - check that a task structure is not stale
1605 * @p: Task structure to be checked.
1606 *
1607 * Test if a process is not yet dead (at most zombie state)
1608 * If pid_alive fails, then pointers within the task structure
1609 * can be stale and must not be dereferenced.
1610 *
1611 * Return: 1 if the process is alive. 0 otherwise.
1612 */
1613static inline int pid_alive(const struct task_struct *p)
1614{
1615 return p->thread_pid != NULL;
1616}
1617
1618static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1619{
1620 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_PGID, ns);
1621}
1622
1623static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1624{
1625 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_PGID, NULL);
1626}
1627
1628
1629static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1630{
1631 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_SID, ns);
1632}
1633
1634static inline pid_t task_session_vnr(struct task_struct *tsk)
1635{
1636 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_SID, NULL);
1637}
1638
1639static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1640{
1641 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_TGID, ns);
1642}
1643
1644static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1645{
1646 return __task_pid_nr_ns(task: tsk, type: PIDTYPE_TGID, NULL);
1647}
1648
1649static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1650{
1651 pid_t pid = 0;
1652
1653 rcu_read_lock();
1654 if (pid_alive(p: tsk))
1655 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1656 rcu_read_unlock();
1657
1658 return pid;
1659}
1660
1661static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1662{
1663 return task_ppid_nr_ns(tsk, ns: &init_pid_ns);
1664}
1665
1666/* Obsolete, do not use: */
1667static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1668{
1669 return task_pgrp_nr_ns(tsk, ns: &init_pid_ns);
1670}
1671
1672#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1673#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1674
1675static inline unsigned int __task_state_index(unsigned int tsk_state,
1676 unsigned int tsk_exit_state)
1677{
1678 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1679
1680 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1681
1682 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1683 state = TASK_REPORT_IDLE;
1684
1685 /*
1686 * We're lying here, but rather than expose a completely new task state
1687 * to userspace, we can make this appear as if the task has gone through
1688 * a regular rt_mutex_lock() call.
1689 */
1690 if (tsk_state & TASK_RTLOCK_WAIT)
1691 state = TASK_UNINTERRUPTIBLE;
1692
1693 return fls(x: state);
1694}
1695
1696static inline unsigned int task_state_index(struct task_struct *tsk)
1697{
1698 return __task_state_index(READ_ONCE(tsk->__state), tsk_exit_state: tsk->exit_state);
1699}
1700
1701static inline char task_index_to_char(unsigned int state)
1702{
1703 static const char state_char[] = "RSDTtXZPI";
1704
1705 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1706
1707 return state_char[state];
1708}
1709
1710static inline char task_state_to_char(struct task_struct *tsk)
1711{
1712 return task_index_to_char(state: task_state_index(tsk));
1713}
1714
1715/**
1716 * is_global_init - check if a task structure is init. Since init
1717 * is free to have sub-threads we need to check tgid.
1718 * @tsk: Task structure to be checked.
1719 *
1720 * Check if a task structure is the first user space task the kernel created.
1721 *
1722 * Return: 1 if the task structure is init. 0 otherwise.
1723 */
1724static inline int is_global_init(struct task_struct *tsk)
1725{
1726 return task_tgid_nr(tsk) == 1;
1727}
1728
1729extern struct pid *cad_pid;
1730
1731/*
1732 * Per process flags
1733 */
1734#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1735#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1736#define PF_EXITING 0x00000004 /* Getting shut down */
1737#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1738#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1739#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1740#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1741#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1742#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1743#define PF_DUMPCORE 0x00000200 /* Dumped core */
1744#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1745#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1746#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1747#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1748#define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1749#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1750#define PF__HOLE__00010000 0x00010000
1751#define PF_KSWAPD 0x00020000 /* I am kswapd */
1752#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1753#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1754#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1755 * I am cleaning dirty pages from some other bdi. */
1756#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1757#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1758#define PF__HOLE__00800000 0x00800000
1759#define PF__HOLE__01000000 0x01000000
1760#define PF__HOLE__02000000 0x02000000
1761#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1762#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1763#define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1764#define PF__HOLE__20000000 0x20000000
1765#define PF__HOLE__40000000 0x40000000
1766#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1767
1768/*
1769 * Only the _current_ task can read/write to tsk->flags, but other
1770 * tasks can access tsk->flags in readonly mode for example
1771 * with tsk_used_math (like during threaded core dumping).
1772 * There is however an exception to this rule during ptrace
1773 * or during fork: the ptracer task is allowed to write to the
1774 * child->flags of its traced child (same goes for fork, the parent
1775 * can write to the child->flags), because we're guaranteed the
1776 * child is not running and in turn not changing child->flags
1777 * at the same time the parent does it.
1778 */
1779#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1780#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1781#define clear_used_math() clear_stopped_child_used_math(current)
1782#define set_used_math() set_stopped_child_used_math(current)
1783
1784#define conditional_stopped_child_used_math(condition, child) \
1785 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1786
1787#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1788
1789#define copy_to_stopped_child_used_math(child) \
1790 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1791
1792/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1793#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1794#define used_math() tsk_used_math(current)
1795
1796static __always_inline bool is_percpu_thread(void)
1797{
1798#ifdef CONFIG_SMP
1799 return (current->flags & PF_NO_SETAFFINITY) &&
1800 (current->nr_cpus_allowed == 1);
1801#else
1802 return true;
1803#endif
1804}
1805
1806/* Per-process atomic flags. */
1807#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1808#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1809#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1810#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1811#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1812#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1813#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1814#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1815
1816#define TASK_PFA_TEST(name, func) \
1817 static inline bool task_##func(struct task_struct *p) \
1818 { return test_bit(PFA_##name, &p->atomic_flags); }
1819
1820#define TASK_PFA_SET(name, func) \
1821 static inline void task_set_##func(struct task_struct *p) \
1822 { set_bit(PFA_##name, &p->atomic_flags); }
1823
1824#define TASK_PFA_CLEAR(name, func) \
1825 static inline void task_clear_##func(struct task_struct *p) \
1826 { clear_bit(PFA_##name, &p->atomic_flags); }
1827
1828TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1829TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1830
1831TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1832TASK_PFA_SET(SPREAD_PAGE, spread_page)
1833TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1834
1835TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1836TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1837TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1838
1839TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1840TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1841TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1842
1843TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1844TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1845TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1846
1847TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1848TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1849
1850TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1851TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1852TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1853
1854TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1855TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1856
1857static inline void
1858current_restore_flags(unsigned long orig_flags, unsigned long flags)
1859{
1860 current->flags &= ~flags;
1861 current->flags |= orig_flags & flags;
1862}
1863
1864extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1865extern int task_can_attach(struct task_struct *p);
1866extern int dl_bw_alloc(int cpu, u64 dl_bw);
1867extern void dl_bw_free(int cpu, u64 dl_bw);
1868#ifdef CONFIG_SMP
1869
1870/* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1871extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1872
1873/**
1874 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1875 * @p: the task
1876 * @new_mask: CPU affinity mask
1877 *
1878 * Return: zero if successful, or a negative error code
1879 */
1880extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1881extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1882extern void release_user_cpus_ptr(struct task_struct *p);
1883extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1884extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1885extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1886#else
1887static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1888{
1889}
1890static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1891{
1892 if (!cpumask_test_cpu(0, new_mask))
1893 return -EINVAL;
1894 return 0;
1895}
1896static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1897{
1898 if (src->user_cpus_ptr)
1899 return -EINVAL;
1900 return 0;
1901}
1902static inline void release_user_cpus_ptr(struct task_struct *p)
1903{
1904 WARN_ON(p->user_cpus_ptr);
1905}
1906
1907static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1908{
1909 return 0;
1910}
1911#endif
1912
1913extern int yield_to(struct task_struct *p, bool preempt);
1914extern void set_user_nice(struct task_struct *p, long nice);
1915extern int task_prio(const struct task_struct *p);
1916
1917/**
1918 * task_nice - return the nice value of a given task.
1919 * @p: the task in question.
1920 *
1921 * Return: The nice value [ -20 ... 0 ... 19 ].
1922 */
1923static inline int task_nice(const struct task_struct *p)
1924{
1925 return PRIO_TO_NICE((p)->static_prio);
1926}
1927
1928extern int can_nice(const struct task_struct *p, const int nice);
1929extern int task_curr(const struct task_struct *p);
1930extern int idle_cpu(int cpu);
1931extern int available_idle_cpu(int cpu);
1932extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1933extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1934extern void sched_set_fifo(struct task_struct *p);
1935extern void sched_set_fifo_low(struct task_struct *p);
1936extern void sched_set_normal(struct task_struct *p, int nice);
1937extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1938extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1939extern struct task_struct *idle_task(int cpu);
1940
1941/**
1942 * is_idle_task - is the specified task an idle task?
1943 * @p: the task in question.
1944 *
1945 * Return: 1 if @p is an idle task. 0 otherwise.
1946 */
1947static __always_inline bool is_idle_task(const struct task_struct *p)
1948{
1949 return !!(p->flags & PF_IDLE);
1950}
1951
1952extern struct task_struct *curr_task(int cpu);
1953extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1954
1955void yield(void);
1956
1957union thread_union {
1958#ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1959 struct task_struct task;
1960#endif
1961#ifndef CONFIG_THREAD_INFO_IN_TASK
1962 struct thread_info thread_info;
1963#endif
1964 unsigned long stack[THREAD_SIZE/sizeof(long)];
1965};
1966
1967#ifndef CONFIG_THREAD_INFO_IN_TASK
1968extern struct thread_info init_thread_info;
1969#endif
1970
1971extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1972
1973#ifdef CONFIG_THREAD_INFO_IN_TASK
1974# define task_thread_info(task) (&(task)->thread_info)
1975#elif !defined(__HAVE_THREAD_FUNCTIONS)
1976# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1977#endif
1978
1979/*
1980 * find a task by one of its numerical ids
1981 *
1982 * find_task_by_pid_ns():
1983 * finds a task by its pid in the specified namespace
1984 * find_task_by_vpid():
1985 * finds a task by its virtual pid
1986 *
1987 * see also find_vpid() etc in include/linux/pid.h
1988 */
1989
1990extern struct task_struct *find_task_by_vpid(pid_t nr);
1991extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1992
1993/*
1994 * find a task by its virtual pid and get the task struct
1995 */
1996extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1997
1998extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1999extern int wake_up_process(struct task_struct *tsk);
2000extern void wake_up_new_task(struct task_struct *tsk);
2001
2002#ifdef CONFIG_SMP
2003extern void kick_process(struct task_struct *tsk);
2004#else
2005static inline void kick_process(struct task_struct *tsk) { }
2006#endif
2007
2008extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2009
2010static inline void set_task_comm(struct task_struct *tsk, const char *from)
2011{
2012 __set_task_comm(tsk, from, exec: false);
2013}
2014
2015extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
2016#define get_task_comm(buf, tsk) ({ \
2017 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
2018 __get_task_comm(buf, sizeof(buf), tsk); \
2019})
2020
2021#ifdef CONFIG_SMP
2022static __always_inline void scheduler_ipi(void)
2023{
2024 /*
2025 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2026 * TIF_NEED_RESCHED remotely (for the first time) will also send
2027 * this IPI.
2028 */
2029 preempt_fold_need_resched();
2030}
2031#else
2032static inline void scheduler_ipi(void) { }
2033#endif
2034
2035extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2036
2037/*
2038 * Set thread flags in other task's structures.
2039 * See asm/thread_info.h for TIF_xxxx flags available:
2040 */
2041static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2042{
2043 set_ti_thread_flag(task_thread_info(tsk), flag);
2044}
2045
2046static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2047{
2048 clear_ti_thread_flag(task_thread_info(tsk), flag);
2049}
2050
2051static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2052 bool value)
2053{
2054 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2055}
2056
2057static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2058{
2059 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2060}
2061
2062static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2063{
2064 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2065}
2066
2067static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2068{
2069 return test_ti_thread_flag(task_thread_info(tsk), flag);
2070}
2071
2072static inline void set_tsk_need_resched(struct task_struct *tsk)
2073{
2074 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2075}
2076
2077static inline void clear_tsk_need_resched(struct task_struct *tsk)
2078{
2079 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2080}
2081
2082static inline int test_tsk_need_resched(struct task_struct *tsk)
2083{
2084 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2085}
2086
2087/*
2088 * cond_resched() and cond_resched_lock(): latency reduction via
2089 * explicit rescheduling in places that are safe. The return
2090 * value indicates whether a reschedule was done in fact.
2091 * cond_resched_lock() will drop the spinlock before scheduling,
2092 */
2093#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2094extern int __cond_resched(void);
2095
2096#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2097
2098void sched_dynamic_klp_enable(void);
2099void sched_dynamic_klp_disable(void);
2100
2101DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2102
2103static __always_inline int _cond_resched(void)
2104{
2105 return static_call_mod(cond_resched)();
2106}
2107
2108#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2109
2110extern int dynamic_cond_resched(void);
2111
2112static __always_inline int _cond_resched(void)
2113{
2114 return dynamic_cond_resched();
2115}
2116
2117#else /* !CONFIG_PREEMPTION */
2118
2119static inline int _cond_resched(void)
2120{
2121 klp_sched_try_switch();
2122 return __cond_resched();
2123}
2124
2125#endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2126
2127#else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2128
2129static inline int _cond_resched(void)
2130{
2131 klp_sched_try_switch();
2132 return 0;
2133}
2134
2135#endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2136
2137#define cond_resched() ({ \
2138 __might_resched(__FILE__, __LINE__, 0); \
2139 _cond_resched(); \
2140})
2141
2142extern int __cond_resched_lock(spinlock_t *lock);
2143extern int __cond_resched_rwlock_read(rwlock_t *lock);
2144extern int __cond_resched_rwlock_write(rwlock_t *lock);
2145
2146#define MIGHT_RESCHED_RCU_SHIFT 8
2147#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2148
2149#ifndef CONFIG_PREEMPT_RT
2150/*
2151 * Non RT kernels have an elevated preempt count due to the held lock,
2152 * but are not allowed to be inside a RCU read side critical section
2153 */
2154# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2155#else
2156/*
2157 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2158 * cond_resched*lock() has to take that into account because it checks for
2159 * preempt_count() and rcu_preempt_depth().
2160 */
2161# define PREEMPT_LOCK_RESCHED_OFFSETS \
2162 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2163#endif
2164
2165#define cond_resched_lock(lock) ({ \
2166 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2167 __cond_resched_lock(lock); \
2168})
2169
2170#define cond_resched_rwlock_read(lock) ({ \
2171 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2172 __cond_resched_rwlock_read(lock); \
2173})
2174
2175#define cond_resched_rwlock_write(lock) ({ \
2176 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2177 __cond_resched_rwlock_write(lock); \
2178})
2179
2180static inline void cond_resched_rcu(void)
2181{
2182#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2183 rcu_read_unlock();
2184 cond_resched();
2185 rcu_read_lock();
2186#endif
2187}
2188
2189#ifdef CONFIG_PREEMPT_DYNAMIC
2190
2191extern bool preempt_model_none(void);
2192extern bool preempt_model_voluntary(void);
2193extern bool preempt_model_full(void);
2194
2195#else
2196
2197static inline bool preempt_model_none(void)
2198{
2199 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2200}
2201static inline bool preempt_model_voluntary(void)
2202{
2203 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2204}
2205static inline bool preempt_model_full(void)
2206{
2207 return IS_ENABLED(CONFIG_PREEMPT);
2208}
2209
2210#endif
2211
2212static inline bool preempt_model_rt(void)
2213{
2214 return IS_ENABLED(CONFIG_PREEMPT_RT);
2215}
2216
2217/*
2218 * Does the preemption model allow non-cooperative preemption?
2219 *
2220 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2221 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2222 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2223 * PREEMPT_NONE model.
2224 */
2225static inline bool preempt_model_preemptible(void)
2226{
2227 return preempt_model_full() || preempt_model_rt();
2228}
2229
2230/*
2231 * Does a critical section need to be broken due to another
2232 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2233 * but a general need for low latency)
2234 */
2235static inline int spin_needbreak(spinlock_t *lock)
2236{
2237#ifdef CONFIG_PREEMPTION
2238 return spin_is_contended(lock);
2239#else
2240 return 0;
2241#endif
2242}
2243
2244/*
2245 * Check if a rwlock is contended.
2246 * Returns non-zero if there is another task waiting on the rwlock.
2247 * Returns zero if the lock is not contended or the system / underlying
2248 * rwlock implementation does not support contention detection.
2249 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2250 * for low latency.
2251 */
2252static inline int rwlock_needbreak(rwlock_t *lock)
2253{
2254#ifdef CONFIG_PREEMPTION
2255 return rwlock_is_contended(lock);
2256#else
2257 return 0;
2258#endif
2259}
2260
2261static __always_inline bool need_resched(void)
2262{
2263 return unlikely(tif_need_resched());
2264}
2265
2266/*
2267 * Wrappers for p->thread_info->cpu access. No-op on UP.
2268 */
2269#ifdef CONFIG_SMP
2270
2271static inline unsigned int task_cpu(const struct task_struct *p)
2272{
2273 return READ_ONCE(task_thread_info(p)->cpu);
2274}
2275
2276extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2277
2278#else
2279
2280static inline unsigned int task_cpu(const struct task_struct *p)
2281{
2282 return 0;
2283}
2284
2285static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2286{
2287}
2288
2289#endif /* CONFIG_SMP */
2290
2291extern bool sched_task_on_rq(struct task_struct *p);
2292extern unsigned long get_wchan(struct task_struct *p);
2293extern struct task_struct *cpu_curr_snapshot(int cpu);
2294
2295/*
2296 * In order to reduce various lock holder preemption latencies provide an
2297 * interface to see if a vCPU is currently running or not.
2298 *
2299 * This allows us to terminate optimistic spin loops and block, analogous to
2300 * the native optimistic spin heuristic of testing if the lock owner task is
2301 * running or not.
2302 */
2303#ifndef vcpu_is_preempted
2304static inline bool vcpu_is_preempted(int cpu)
2305{
2306 return false;
2307}
2308#endif
2309
2310extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2311extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2312
2313#ifndef TASK_SIZE_OF
2314#define TASK_SIZE_OF(tsk) TASK_SIZE
2315#endif
2316
2317#ifdef CONFIG_SMP
2318static inline bool owner_on_cpu(struct task_struct *owner)
2319{
2320 /*
2321 * As lock holder preemption issue, we both skip spinning if
2322 * task is not on cpu or its cpu is preempted
2323 */
2324 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(cpu: task_cpu(p: owner));
2325}
2326
2327/* Returns effective CPU energy utilization, as seen by the scheduler */
2328unsigned long sched_cpu_util(int cpu);
2329#endif /* CONFIG_SMP */
2330
2331#ifdef CONFIG_RSEQ
2332
2333/*
2334 * Map the event mask on the user-space ABI enum rseq_cs_flags
2335 * for direct mask checks.
2336 */
2337enum rseq_event_mask_bits {
2338 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2339 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2340 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2341};
2342
2343enum rseq_event_mask {
2344 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2345 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2346 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2347};
2348
2349static inline void rseq_set_notify_resume(struct task_struct *t)
2350{
2351 if (t->rseq)
2352 set_tsk_thread_flag(tsk: t, TIF_NOTIFY_RESUME);
2353}
2354
2355void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2356
2357static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2358 struct pt_regs *regs)
2359{
2360 if (current->rseq)
2361 __rseq_handle_notify_resume(sig: ksig, regs);
2362}
2363
2364static inline void rseq_signal_deliver(struct ksignal *ksig,
2365 struct pt_regs *regs)
2366{
2367 preempt_disable();
2368 __set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2369 preempt_enable();
2370 rseq_handle_notify_resume(ksig, regs);
2371}
2372
2373/* rseq_preempt() requires preemption to be disabled. */
2374static inline void rseq_preempt(struct task_struct *t)
2375{
2376 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2377 rseq_set_notify_resume(t);
2378}
2379
2380/* rseq_migrate() requires preemption to be disabled. */
2381static inline void rseq_migrate(struct task_struct *t)
2382{
2383 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2384 rseq_set_notify_resume(t);
2385}
2386
2387/*
2388 * If parent process has a registered restartable sequences area, the
2389 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2390 */
2391static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2392{
2393 if (clone_flags & CLONE_VM) {
2394 t->rseq = NULL;
2395 t->rseq_len = 0;
2396 t->rseq_sig = 0;
2397 t->rseq_event_mask = 0;
2398 } else {
2399 t->rseq = current->rseq;
2400 t->rseq_len = current->rseq_len;
2401 t->rseq_sig = current->rseq_sig;
2402 t->rseq_event_mask = current->rseq_event_mask;
2403 }
2404}
2405
2406static inline void rseq_execve(struct task_struct *t)
2407{
2408 t->rseq = NULL;
2409 t->rseq_len = 0;
2410 t->rseq_sig = 0;
2411 t->rseq_event_mask = 0;
2412}
2413
2414#else
2415
2416static inline void rseq_set_notify_resume(struct task_struct *t)
2417{
2418}
2419static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2420 struct pt_regs *regs)
2421{
2422}
2423static inline void rseq_signal_deliver(struct ksignal *ksig,
2424 struct pt_regs *regs)
2425{
2426}
2427static inline void rseq_preempt(struct task_struct *t)
2428{
2429}
2430static inline void rseq_migrate(struct task_struct *t)
2431{
2432}
2433static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2434{
2435}
2436static inline void rseq_execve(struct task_struct *t)
2437{
2438}
2439
2440#endif
2441
2442#ifdef CONFIG_DEBUG_RSEQ
2443
2444void rseq_syscall(struct pt_regs *regs);
2445
2446#else
2447
2448static inline void rseq_syscall(struct pt_regs *regs)
2449{
2450}
2451
2452#endif
2453
2454#ifdef CONFIG_SCHED_CORE
2455extern void sched_core_free(struct task_struct *tsk);
2456extern void sched_core_fork(struct task_struct *p);
2457extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2458 unsigned long uaddr);
2459extern int sched_core_idle_cpu(int cpu);
2460#else
2461static inline void sched_core_free(struct task_struct *tsk) { }
2462static inline void sched_core_fork(struct task_struct *p) { }
2463static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2464#endif
2465
2466extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2467
2468#endif
2469

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