xref: /linux/include/linux/sched.h (revision 2a52ca7c98960aafb0eca9ef96b2d0c932171357)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4 
5 /*
6  * Define 'struct task_struct' and provide the main scheduler
7  * APIs (schedule(), wakeup variants, etc.)
8  */
9 
10 #include <uapi/linux/sched.h>
11 
12 #include <asm/current.h>
13 #include <asm/processor.h>
14 #include <linux/thread_info.h>
15 #include <linux/preempt.h>
16 #include <linux/cpumask.h>
17 
18 #include <linux/cache.h>
19 #include <linux/irqflags_types.h>
20 #include <linux/smp_types.h>
21 #include <linux/pid_types.h>
22 #include <linux/sem_types.h>
23 #include <linux/shm.h>
24 #include <linux/kmsan_types.h>
25 #include <linux/mutex_types.h>
26 #include <linux/plist_types.h>
27 #include <linux/hrtimer_types.h>
28 #include <linux/timer_types.h>
29 #include <linux/seccomp_types.h>
30 #include <linux/nodemask_types.h>
31 #include <linux/refcount_types.h>
32 #include <linux/resource.h>
33 #include <linux/latencytop.h>
34 #include <linux/sched/prio.h>
35 #include <linux/sched/types.h>
36 #include <linux/signal_types.h>
37 #include <linux/syscall_user_dispatch_types.h>
38 #include <linux/mm_types_task.h>
39 #include <linux/task_io_accounting.h>
40 #include <linux/posix-timers_types.h>
41 #include <linux/restart_block.h>
42 #include <uapi/linux/rseq.h>
43 #include <linux/seqlock_types.h>
44 #include <linux/kcsan.h>
45 #include <linux/rv.h>
46 #include <linux/livepatch_sched.h>
47 #include <linux/uidgid_types.h>
48 #include <asm/kmap_size.h>
49 
50 /* task_struct member predeclarations (sorted alphabetically): */
51 struct audit_context;
52 struct bio_list;
53 struct blk_plug;
54 struct bpf_local_storage;
55 struct bpf_run_ctx;
56 struct capture_control;
57 struct cfs_rq;
58 struct fs_struct;
59 struct futex_pi_state;
60 struct io_context;
61 struct io_uring_task;
62 struct mempolicy;
63 struct nameidata;
64 struct nsproxy;
65 struct perf_event_context;
66 struct pid_namespace;
67 struct pipe_inode_info;
68 struct rcu_node;
69 struct reclaim_state;
70 struct robust_list_head;
71 struct root_domain;
72 struct rq;
73 struct sched_attr;
74 struct sched_dl_entity;
75 struct seq_file;
76 struct sighand_struct;
77 struct signal_struct;
78 struct task_delay_info;
79 struct task_group;
80 struct task_struct;
81 struct user_event_mm;
82 
83 #include <linux/sched/ext.h>
84 
85 /*
86  * Task state bitmask. NOTE! These bits are also
87  * encoded in fs/proc/array.c: get_task_state().
88  *
89  * We have two separate sets of flags: task->__state
90  * is about runnability, while task->exit_state are
91  * about the task exiting. Confusing, but this way
92  * modifying one set can't modify the other one by
93  * mistake.
94  */
95 
96 /* Used in tsk->__state: */
97 #define TASK_RUNNING			0x00000000
98 #define TASK_INTERRUPTIBLE		0x00000001
99 #define TASK_UNINTERRUPTIBLE		0x00000002
100 #define __TASK_STOPPED			0x00000004
101 #define __TASK_TRACED			0x00000008
102 /* Used in tsk->exit_state: */
103 #define EXIT_DEAD			0x00000010
104 #define EXIT_ZOMBIE			0x00000020
105 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
106 /* Used in tsk->__state again: */
107 #define TASK_PARKED			0x00000040
108 #define TASK_DEAD			0x00000080
109 #define TASK_WAKEKILL			0x00000100
110 #define TASK_WAKING			0x00000200
111 #define TASK_NOLOAD			0x00000400
112 #define TASK_NEW			0x00000800
113 #define TASK_RTLOCK_WAIT		0x00001000
114 #define TASK_FREEZABLE			0x00002000
115 #define __TASK_FREEZABLE_UNSAFE	       (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
116 #define TASK_FROZEN			0x00008000
117 #define TASK_STATE_MAX			0x00010000
118 
119 #define TASK_ANY			(TASK_STATE_MAX-1)
120 
121 /*
122  * DO NOT ADD ANY NEW USERS !
123  */
124 #define TASK_FREEZABLE_UNSAFE		(TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
125 
126 /* Convenience macros for the sake of set_current_state: */
127 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
128 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
129 #define TASK_TRACED			__TASK_TRACED
130 
131 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
132 
133 /* Convenience macros for the sake of wake_up(): */
134 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
135 
136 /* get_task_state(): */
137 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
138 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
139 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
140 					 TASK_PARKED)
141 
142 #define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
143 
144 #define task_is_traced(task)		((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
145 #define task_is_stopped(task)		((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
146 #define task_is_stopped_or_traced(task)	((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
147 
148 /*
149  * Special states are those that do not use the normal wait-loop pattern. See
150  * the comment with set_special_state().
151  */
152 #define is_special_task_state(state)				\
153 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
154 
155 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
156 # define debug_normal_state_change(state_value)				\
157 	do {								\
158 		WARN_ON_ONCE(is_special_task_state(state_value));	\
159 		current->task_state_change = _THIS_IP_;			\
160 	} while (0)
161 
162 # define debug_special_state_change(state_value)			\
163 	do {								\
164 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
165 		current->task_state_change = _THIS_IP_;			\
166 	} while (0)
167 
168 # define debug_rtlock_wait_set_state()					\
169 	do {								 \
170 		current->saved_state_change = current->task_state_change;\
171 		current->task_state_change = _THIS_IP_;			 \
172 	} while (0)
173 
174 # define debug_rtlock_wait_restore_state()				\
175 	do {								 \
176 		current->task_state_change = current->saved_state_change;\
177 	} while (0)
178 
179 #else
180 # define debug_normal_state_change(cond)	do { } while (0)
181 # define debug_special_state_change(cond)	do { } while (0)
182 # define debug_rtlock_wait_set_state()		do { } while (0)
183 # define debug_rtlock_wait_restore_state()	do { } while (0)
184 #endif
185 
186 /*
187  * set_current_state() includes a barrier so that the write of current->__state
188  * is correctly serialised wrt the caller's subsequent test of whether to
189  * actually sleep:
190  *
191  *   for (;;) {
192  *	set_current_state(TASK_UNINTERRUPTIBLE);
193  *	if (CONDITION)
194  *	   break;
195  *
196  *	schedule();
197  *   }
198  *   __set_current_state(TASK_RUNNING);
199  *
200  * If the caller does not need such serialisation (because, for instance, the
201  * CONDITION test and condition change and wakeup are under the same lock) then
202  * use __set_current_state().
203  *
204  * The above is typically ordered against the wakeup, which does:
205  *
206  *   CONDITION = 1;
207  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
208  *
209  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
210  * accessing p->__state.
211  *
212  * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
213  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
214  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
215  *
216  * However, with slightly different timing the wakeup TASK_RUNNING store can
217  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
218  * a problem either because that will result in one extra go around the loop
219  * and our @cond test will save the day.
220  *
221  * Also see the comments of try_to_wake_up().
222  */
223 #define __set_current_state(state_value)				\
224 	do {								\
225 		debug_normal_state_change((state_value));		\
226 		WRITE_ONCE(current->__state, (state_value));		\
227 	} while (0)
228 
229 #define set_current_state(state_value)					\
230 	do {								\
231 		debug_normal_state_change((state_value));		\
232 		smp_store_mb(current->__state, (state_value));		\
233 	} while (0)
234 
235 /*
236  * set_special_state() should be used for those states when the blocking task
237  * can not use the regular condition based wait-loop. In that case we must
238  * serialize against wakeups such that any possible in-flight TASK_RUNNING
239  * stores will not collide with our state change.
240  */
241 #define set_special_state(state_value)					\
242 	do {								\
243 		unsigned long flags; /* may shadow */			\
244 									\
245 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
246 		debug_special_state_change((state_value));		\
247 		WRITE_ONCE(current->__state, (state_value));		\
248 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
249 	} while (0)
250 
251 /*
252  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
253  *
254  * RT's spin/rwlock substitutions are state preserving. The state of the
255  * task when blocking on the lock is saved in task_struct::saved_state and
256  * restored after the lock has been acquired.  These operations are
257  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
258  * lock related wakeups while the task is blocked on the lock are
259  * redirected to operate on task_struct::saved_state to ensure that these
260  * are not dropped. On restore task_struct::saved_state is set to
261  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
262  *
263  * The lock operation looks like this:
264  *
265  *	current_save_and_set_rtlock_wait_state();
266  *	for (;;) {
267  *		if (try_lock())
268  *			break;
269  *		raw_spin_unlock_irq(&lock->wait_lock);
270  *		schedule_rtlock();
271  *		raw_spin_lock_irq(&lock->wait_lock);
272  *		set_current_state(TASK_RTLOCK_WAIT);
273  *	}
274  *	current_restore_rtlock_saved_state();
275  */
276 #define current_save_and_set_rtlock_wait_state()			\
277 	do {								\
278 		lockdep_assert_irqs_disabled();				\
279 		raw_spin_lock(&current->pi_lock);			\
280 		current->saved_state = current->__state;		\
281 		debug_rtlock_wait_set_state();				\
282 		WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);		\
283 		raw_spin_unlock(&current->pi_lock);			\
284 	} while (0);
285 
286 #define current_restore_rtlock_saved_state()				\
287 	do {								\
288 		lockdep_assert_irqs_disabled();				\
289 		raw_spin_lock(&current->pi_lock);			\
290 		debug_rtlock_wait_restore_state();			\
291 		WRITE_ONCE(current->__state, current->saved_state);	\
292 		current->saved_state = TASK_RUNNING;			\
293 		raw_spin_unlock(&current->pi_lock);			\
294 	} while (0);
295 
296 #define get_current_state()	READ_ONCE(current->__state)
297 
298 /*
299  * Define the task command name length as enum, then it can be visible to
300  * BPF programs.
301  */
302 enum {
303 	TASK_COMM_LEN = 16,
304 };
305 
306 extern void sched_tick(void);
307 
308 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
309 
310 extern long schedule_timeout(long timeout);
311 extern long schedule_timeout_interruptible(long timeout);
312 extern long schedule_timeout_killable(long timeout);
313 extern long schedule_timeout_uninterruptible(long timeout);
314 extern long schedule_timeout_idle(long timeout);
315 asmlinkage void schedule(void);
316 extern void schedule_preempt_disabled(void);
317 asmlinkage void preempt_schedule_irq(void);
318 #ifdef CONFIG_PREEMPT_RT
319  extern void schedule_rtlock(void);
320 #endif
321 
322 extern int __must_check io_schedule_prepare(void);
323 extern void io_schedule_finish(int token);
324 extern long io_schedule_timeout(long timeout);
325 extern void io_schedule(void);
326 
327 /**
328  * struct prev_cputime - snapshot of system and user cputime
329  * @utime: time spent in user mode
330  * @stime: time spent in system mode
331  * @lock: protects the above two fields
332  *
333  * Stores previous user/system time values such that we can guarantee
334  * monotonicity.
335  */
336 struct prev_cputime {
337 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
338 	u64				utime;
339 	u64				stime;
340 	raw_spinlock_t			lock;
341 #endif
342 };
343 
344 enum vtime_state {
345 	/* Task is sleeping or running in a CPU with VTIME inactive: */
346 	VTIME_INACTIVE = 0,
347 	/* Task is idle */
348 	VTIME_IDLE,
349 	/* Task runs in kernelspace in a CPU with VTIME active: */
350 	VTIME_SYS,
351 	/* Task runs in userspace in a CPU with VTIME active: */
352 	VTIME_USER,
353 	/* Task runs as guests in a CPU with VTIME active: */
354 	VTIME_GUEST,
355 };
356 
357 struct vtime {
358 	seqcount_t		seqcount;
359 	unsigned long long	starttime;
360 	enum vtime_state	state;
361 	unsigned int		cpu;
362 	u64			utime;
363 	u64			stime;
364 	u64			gtime;
365 };
366 
367 /*
368  * Utilization clamp constraints.
369  * @UCLAMP_MIN:	Minimum utilization
370  * @UCLAMP_MAX:	Maximum utilization
371  * @UCLAMP_CNT:	Utilization clamp constraints count
372  */
373 enum uclamp_id {
374 	UCLAMP_MIN = 0,
375 	UCLAMP_MAX,
376 	UCLAMP_CNT
377 };
378 
379 #ifdef CONFIG_SMP
380 extern struct root_domain def_root_domain;
381 extern struct mutex sched_domains_mutex;
382 #endif
383 
384 struct sched_param {
385 	int sched_priority;
386 };
387 
388 struct sched_info {
389 #ifdef CONFIG_SCHED_INFO
390 	/* Cumulative counters: */
391 
392 	/* # of times we have run on this CPU: */
393 	unsigned long			pcount;
394 
395 	/* Time spent waiting on a runqueue: */
396 	unsigned long long		run_delay;
397 
398 	/* Timestamps: */
399 
400 	/* When did we last run on a CPU? */
401 	unsigned long long		last_arrival;
402 
403 	/* When were we last queued to run? */
404 	unsigned long long		last_queued;
405 
406 #endif /* CONFIG_SCHED_INFO */
407 };
408 
409 /*
410  * Integer metrics need fixed point arithmetic, e.g., sched/fair
411  * has a few: load, load_avg, util_avg, freq, and capacity.
412  *
413  * We define a basic fixed point arithmetic range, and then formalize
414  * all these metrics based on that basic range.
415  */
416 # define SCHED_FIXEDPOINT_SHIFT		10
417 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
418 
419 /* Increase resolution of cpu_capacity calculations */
420 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
421 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
422 
423 struct load_weight {
424 	unsigned long			weight;
425 	u32				inv_weight;
426 };
427 
428 /*
429  * The load/runnable/util_avg accumulates an infinite geometric series
430  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
431  *
432  * [load_avg definition]
433  *
434  *   load_avg = runnable% * scale_load_down(load)
435  *
436  * [runnable_avg definition]
437  *
438  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
439  *
440  * [util_avg definition]
441  *
442  *   util_avg = running% * SCHED_CAPACITY_SCALE
443  *
444  * where runnable% is the time ratio that a sched_entity is runnable and
445  * running% the time ratio that a sched_entity is running.
446  *
447  * For cfs_rq, they are the aggregated values of all runnable and blocked
448  * sched_entities.
449  *
450  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
451  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
452  * for computing those signals (see update_rq_clock_pelt())
453  *
454  * N.B., the above ratios (runnable% and running%) themselves are in the
455  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
456  * to as large a range as necessary. This is for example reflected by
457  * util_avg's SCHED_CAPACITY_SCALE.
458  *
459  * [Overflow issue]
460  *
461  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
462  * with the highest load (=88761), always runnable on a single cfs_rq,
463  * and should not overflow as the number already hits PID_MAX_LIMIT.
464  *
465  * For all other cases (including 32-bit kernels), struct load_weight's
466  * weight will overflow first before we do, because:
467  *
468  *    Max(load_avg) <= Max(load.weight)
469  *
470  * Then it is the load_weight's responsibility to consider overflow
471  * issues.
472  */
473 struct sched_avg {
474 	u64				last_update_time;
475 	u64				load_sum;
476 	u64				runnable_sum;
477 	u32				util_sum;
478 	u32				period_contrib;
479 	unsigned long			load_avg;
480 	unsigned long			runnable_avg;
481 	unsigned long			util_avg;
482 	unsigned int			util_est;
483 } ____cacheline_aligned;
484 
485 /*
486  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
487  * updates. When a task is dequeued, its util_est should not be updated if its
488  * util_avg has not been updated in the meantime.
489  * This information is mapped into the MSB bit of util_est at dequeue time.
490  * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
491  * it is safe to use MSB.
492  */
493 #define UTIL_EST_WEIGHT_SHIFT		2
494 #define UTIL_AVG_UNCHANGED		0x80000000
495 
496 struct sched_statistics {
497 #ifdef CONFIG_SCHEDSTATS
498 	u64				wait_start;
499 	u64				wait_max;
500 	u64				wait_count;
501 	u64				wait_sum;
502 	u64				iowait_count;
503 	u64				iowait_sum;
504 
505 	u64				sleep_start;
506 	u64				sleep_max;
507 	s64				sum_sleep_runtime;
508 
509 	u64				block_start;
510 	u64				block_max;
511 	s64				sum_block_runtime;
512 
513 	s64				exec_max;
514 	u64				slice_max;
515 
516 	u64				nr_migrations_cold;
517 	u64				nr_failed_migrations_affine;
518 	u64				nr_failed_migrations_running;
519 	u64				nr_failed_migrations_hot;
520 	u64				nr_forced_migrations;
521 
522 	u64				nr_wakeups;
523 	u64				nr_wakeups_sync;
524 	u64				nr_wakeups_migrate;
525 	u64				nr_wakeups_local;
526 	u64				nr_wakeups_remote;
527 	u64				nr_wakeups_affine;
528 	u64				nr_wakeups_affine_attempts;
529 	u64				nr_wakeups_passive;
530 	u64				nr_wakeups_idle;
531 
532 #ifdef CONFIG_SCHED_CORE
533 	u64				core_forceidle_sum;
534 #endif
535 #endif /* CONFIG_SCHEDSTATS */
536 } ____cacheline_aligned;
537 
538 struct sched_entity {
539 	/* For load-balancing: */
540 	struct load_weight		load;
541 	struct rb_node			run_node;
542 	u64				deadline;
543 	u64				min_vruntime;
544 
545 	struct list_head		group_node;
546 	unsigned int			on_rq;
547 
548 	u64				exec_start;
549 	u64				sum_exec_runtime;
550 	u64				prev_sum_exec_runtime;
551 	u64				vruntime;
552 	s64				vlag;
553 	u64				slice;
554 
555 	u64				nr_migrations;
556 
557 #ifdef CONFIG_FAIR_GROUP_SCHED
558 	int				depth;
559 	struct sched_entity		*parent;
560 	/* rq on which this entity is (to be) queued: */
561 	struct cfs_rq			*cfs_rq;
562 	/* rq "owned" by this entity/group: */
563 	struct cfs_rq			*my_q;
564 	/* cached value of my_q->h_nr_running */
565 	unsigned long			runnable_weight;
566 #endif
567 
568 #ifdef CONFIG_SMP
569 	/*
570 	 * Per entity load average tracking.
571 	 *
572 	 * Put into separate cache line so it does not
573 	 * collide with read-mostly values above.
574 	 */
575 	struct sched_avg		avg;
576 #endif
577 };
578 
579 struct sched_rt_entity {
580 	struct list_head		run_list;
581 	unsigned long			timeout;
582 	unsigned long			watchdog_stamp;
583 	unsigned int			time_slice;
584 	unsigned short			on_rq;
585 	unsigned short			on_list;
586 
587 	struct sched_rt_entity		*back;
588 #ifdef CONFIG_RT_GROUP_SCHED
589 	struct sched_rt_entity		*parent;
590 	/* rq on which this entity is (to be) queued: */
591 	struct rt_rq			*rt_rq;
592 	/* rq "owned" by this entity/group: */
593 	struct rt_rq			*my_q;
594 #endif
595 } __randomize_layout;
596 
597 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
598 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
599 
600 struct sched_dl_entity {
601 	struct rb_node			rb_node;
602 
603 	/*
604 	 * Original scheduling parameters. Copied here from sched_attr
605 	 * during sched_setattr(), they will remain the same until
606 	 * the next sched_setattr().
607 	 */
608 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
609 	u64				dl_deadline;	/* Relative deadline of each instance	*/
610 	u64				dl_period;	/* Separation of two instances (period) */
611 	u64				dl_bw;		/* dl_runtime / dl_period		*/
612 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
613 
614 	/*
615 	 * Actual scheduling parameters. Initialized with the values above,
616 	 * they are continuously updated during task execution. Note that
617 	 * the remaining runtime could be < 0 in case we are in overrun.
618 	 */
619 	s64				runtime;	/* Remaining runtime for this instance	*/
620 	u64				deadline;	/* Absolute deadline for this instance	*/
621 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
622 
623 	/*
624 	 * Some bool flags:
625 	 *
626 	 * @dl_throttled tells if we exhausted the runtime. If so, the
627 	 * task has to wait for a replenishment to be performed at the
628 	 * next firing of dl_timer.
629 	 *
630 	 * @dl_yielded tells if task gave up the CPU before consuming
631 	 * all its available runtime during the last job.
632 	 *
633 	 * @dl_non_contending tells if the task is inactive while still
634 	 * contributing to the active utilization. In other words, it
635 	 * indicates if the inactive timer has been armed and its handler
636 	 * has not been executed yet. This flag is useful to avoid race
637 	 * conditions between the inactive timer handler and the wakeup
638 	 * code.
639 	 *
640 	 * @dl_overrun tells if the task asked to be informed about runtime
641 	 * overruns.
642 	 */
643 	unsigned int			dl_throttled      : 1;
644 	unsigned int			dl_yielded        : 1;
645 	unsigned int			dl_non_contending : 1;
646 	unsigned int			dl_overrun	  : 1;
647 	unsigned int			dl_server         : 1;
648 
649 	/*
650 	 * Bandwidth enforcement timer. Each -deadline task has its
651 	 * own bandwidth to be enforced, thus we need one timer per task.
652 	 */
653 	struct hrtimer			dl_timer;
654 
655 	/*
656 	 * Inactive timer, responsible for decreasing the active utilization
657 	 * at the "0-lag time". When a -deadline task blocks, it contributes
658 	 * to GRUB's active utilization until the "0-lag time", hence a
659 	 * timer is needed to decrease the active utilization at the correct
660 	 * time.
661 	 */
662 	struct hrtimer			inactive_timer;
663 
664 	/*
665 	 * Bits for DL-server functionality. Also see the comment near
666 	 * dl_server_update().
667 	 *
668 	 * @rq the runqueue this server is for
669 	 *
670 	 * @server_has_tasks() returns true if @server_pick return a
671 	 * runnable task.
672 	 */
673 	struct rq			*rq;
674 	dl_server_has_tasks_f		server_has_tasks;
675 	dl_server_pick_f		server_pick;
676 
677 #ifdef CONFIG_RT_MUTEXES
678 	/*
679 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
680 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
681 	 * to (the original one/itself).
682 	 */
683 	struct sched_dl_entity *pi_se;
684 #endif
685 };
686 
687 #ifdef CONFIG_UCLAMP_TASK
688 /* Number of utilization clamp buckets (shorter alias) */
689 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
690 
691 /*
692  * Utilization clamp for a scheduling entity
693  * @value:		clamp value "assigned" to a se
694  * @bucket_id:		bucket index corresponding to the "assigned" value
695  * @active:		the se is currently refcounted in a rq's bucket
696  * @user_defined:	the requested clamp value comes from user-space
697  *
698  * The bucket_id is the index of the clamp bucket matching the clamp value
699  * which is pre-computed and stored to avoid expensive integer divisions from
700  * the fast path.
701  *
702  * The active bit is set whenever a task has got an "effective" value assigned,
703  * which can be different from the clamp value "requested" from user-space.
704  * This allows to know a task is refcounted in the rq's bucket corresponding
705  * to the "effective" bucket_id.
706  *
707  * The user_defined bit is set whenever a task has got a task-specific clamp
708  * value requested from userspace, i.e. the system defaults apply to this task
709  * just as a restriction. This allows to relax default clamps when a less
710  * restrictive task-specific value has been requested, thus allowing to
711  * implement a "nice" semantic. For example, a task running with a 20%
712  * default boost can still drop its own boosting to 0%.
713  */
714 struct uclamp_se {
715 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
716 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
717 	unsigned int active		: 1;
718 	unsigned int user_defined	: 1;
719 };
720 #endif /* CONFIG_UCLAMP_TASK */
721 
722 union rcu_special {
723 	struct {
724 		u8			blocked;
725 		u8			need_qs;
726 		u8			exp_hint; /* Hint for performance. */
727 		u8			need_mb; /* Readers need smp_mb(). */
728 	} b; /* Bits. */
729 	u32 s; /* Set of bits. */
730 };
731 
732 enum perf_event_task_context {
733 	perf_invalid_context = -1,
734 	perf_hw_context = 0,
735 	perf_sw_context,
736 	perf_nr_task_contexts,
737 };
738 
739 struct wake_q_node {
740 	struct wake_q_node *next;
741 };
742 
743 struct kmap_ctrl {
744 #ifdef CONFIG_KMAP_LOCAL
745 	int				idx;
746 	pte_t				pteval[KM_MAX_IDX];
747 #endif
748 };
749 
750 struct task_struct {
751 #ifdef CONFIG_THREAD_INFO_IN_TASK
752 	/*
753 	 * For reasons of header soup (see current_thread_info()), this
754 	 * must be the first element of task_struct.
755 	 */
756 	struct thread_info		thread_info;
757 #endif
758 	unsigned int			__state;
759 
760 	/* saved state for "spinlock sleepers" */
761 	unsigned int			saved_state;
762 
763 	/*
764 	 * This begins the randomizable portion of task_struct. Only
765 	 * scheduling-critical items should be added above here.
766 	 */
767 	randomized_struct_fields_start
768 
769 	void				*stack;
770 	refcount_t			usage;
771 	/* Per task flags (PF_*), defined further below: */
772 	unsigned int			flags;
773 	unsigned int			ptrace;
774 
775 #ifdef CONFIG_MEM_ALLOC_PROFILING
776 	struct alloc_tag		*alloc_tag;
777 #endif
778 
779 #ifdef CONFIG_SMP
780 	int				on_cpu;
781 	struct __call_single_node	wake_entry;
782 	unsigned int			wakee_flips;
783 	unsigned long			wakee_flip_decay_ts;
784 	struct task_struct		*last_wakee;
785 
786 	/*
787 	 * recent_used_cpu is initially set as the last CPU used by a task
788 	 * that wakes affine another task. Waker/wakee relationships can
789 	 * push tasks around a CPU where each wakeup moves to the next one.
790 	 * Tracking a recently used CPU allows a quick search for a recently
791 	 * used CPU that may be idle.
792 	 */
793 	int				recent_used_cpu;
794 	int				wake_cpu;
795 #endif
796 	int				on_rq;
797 
798 	int				prio;
799 	int				static_prio;
800 	int				normal_prio;
801 	unsigned int			rt_priority;
802 
803 	struct sched_entity		se;
804 	struct sched_rt_entity		rt;
805 	struct sched_dl_entity		dl;
806 	struct sched_dl_entity		*dl_server;
807 #ifdef CONFIG_SCHED_CLASS_EXT
808 	struct sched_ext_entity		scx;
809 #endif
810 	const struct sched_class	*sched_class;
811 
812 #ifdef CONFIG_SCHED_CORE
813 	struct rb_node			core_node;
814 	unsigned long			core_cookie;
815 	unsigned int			core_occupation;
816 #endif
817 
818 #ifdef CONFIG_CGROUP_SCHED
819 	struct task_group		*sched_task_group;
820 #endif
821 
822 
823 #ifdef CONFIG_UCLAMP_TASK
824 	/*
825 	 * Clamp values requested for a scheduling entity.
826 	 * Must be updated with task_rq_lock() held.
827 	 */
828 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
829 	/*
830 	 * Effective clamp values used for a scheduling entity.
831 	 * Must be updated with task_rq_lock() held.
832 	 */
833 	struct uclamp_se		uclamp[UCLAMP_CNT];
834 #endif
835 
836 	struct sched_statistics         stats;
837 
838 #ifdef CONFIG_PREEMPT_NOTIFIERS
839 	/* List of struct preempt_notifier: */
840 	struct hlist_head		preempt_notifiers;
841 #endif
842 
843 #ifdef CONFIG_BLK_DEV_IO_TRACE
844 	unsigned int			btrace_seq;
845 #endif
846 
847 	unsigned int			policy;
848 	unsigned long			max_allowed_capacity;
849 	int				nr_cpus_allowed;
850 	const cpumask_t			*cpus_ptr;
851 	cpumask_t			*user_cpus_ptr;
852 	cpumask_t			cpus_mask;
853 	void				*migration_pending;
854 #ifdef CONFIG_SMP
855 	unsigned short			migration_disabled;
856 #endif
857 	unsigned short			migration_flags;
858 
859 #ifdef CONFIG_PREEMPT_RCU
860 	int				rcu_read_lock_nesting;
861 	union rcu_special		rcu_read_unlock_special;
862 	struct list_head		rcu_node_entry;
863 	struct rcu_node			*rcu_blocked_node;
864 #endif /* #ifdef CONFIG_PREEMPT_RCU */
865 
866 #ifdef CONFIG_TASKS_RCU
867 	unsigned long			rcu_tasks_nvcsw;
868 	u8				rcu_tasks_holdout;
869 	u8				rcu_tasks_idx;
870 	int				rcu_tasks_idle_cpu;
871 	struct list_head		rcu_tasks_holdout_list;
872 	int				rcu_tasks_exit_cpu;
873 	struct list_head		rcu_tasks_exit_list;
874 #endif /* #ifdef CONFIG_TASKS_RCU */
875 
876 #ifdef CONFIG_TASKS_TRACE_RCU
877 	int				trc_reader_nesting;
878 	int				trc_ipi_to_cpu;
879 	union rcu_special		trc_reader_special;
880 	struct list_head		trc_holdout_list;
881 	struct list_head		trc_blkd_node;
882 	int				trc_blkd_cpu;
883 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
884 
885 	struct sched_info		sched_info;
886 
887 	struct list_head		tasks;
888 #ifdef CONFIG_SMP
889 	struct plist_node		pushable_tasks;
890 	struct rb_node			pushable_dl_tasks;
891 #endif
892 
893 	struct mm_struct		*mm;
894 	struct mm_struct		*active_mm;
895 	struct address_space		*faults_disabled_mapping;
896 
897 	int				exit_state;
898 	int				exit_code;
899 	int				exit_signal;
900 	/* The signal sent when the parent dies: */
901 	int				pdeath_signal;
902 	/* JOBCTL_*, siglock protected: */
903 	unsigned long			jobctl;
904 
905 	/* Used for emulating ABI behavior of previous Linux versions: */
906 	unsigned int			personality;
907 
908 	/* Scheduler bits, serialized by scheduler locks: */
909 	unsigned			sched_reset_on_fork:1;
910 	unsigned			sched_contributes_to_load:1;
911 	unsigned			sched_migrated:1;
912 
913 	/* Force alignment to the next boundary: */
914 	unsigned			:0;
915 
916 	/* Unserialized, strictly 'current' */
917 
918 	/*
919 	 * This field must not be in the scheduler word above due to wakelist
920 	 * queueing no longer being serialized by p->on_cpu. However:
921 	 *
922 	 * p->XXX = X;			ttwu()
923 	 * schedule()			  if (p->on_rq && ..) // false
924 	 *   smp_mb__after_spinlock();	  if (smp_load_acquire(&p->on_cpu) && //true
925 	 *   deactivate_task()		      ttwu_queue_wakelist())
926 	 *     p->on_rq = 0;			p->sched_remote_wakeup = Y;
927 	 *
928 	 * guarantees all stores of 'current' are visible before
929 	 * ->sched_remote_wakeup gets used, so it can be in this word.
930 	 */
931 	unsigned			sched_remote_wakeup:1;
932 #ifdef CONFIG_RT_MUTEXES
933 	unsigned			sched_rt_mutex:1;
934 #endif
935 
936 	/* Bit to tell TOMOYO we're in execve(): */
937 	unsigned			in_execve:1;
938 	unsigned			in_iowait:1;
939 #ifndef TIF_RESTORE_SIGMASK
940 	unsigned			restore_sigmask:1;
941 #endif
942 #ifdef CONFIG_MEMCG
943 	unsigned			in_user_fault:1;
944 #endif
945 #ifdef CONFIG_LRU_GEN
946 	/* whether the LRU algorithm may apply to this access */
947 	unsigned			in_lru_fault:1;
948 #endif
949 #ifdef CONFIG_COMPAT_BRK
950 	unsigned			brk_randomized:1;
951 #endif
952 #ifdef CONFIG_CGROUPS
953 	/* disallow userland-initiated cgroup migration */
954 	unsigned			no_cgroup_migration:1;
955 	/* task is frozen/stopped (used by the cgroup freezer) */
956 	unsigned			frozen:1;
957 #endif
958 #ifdef CONFIG_BLK_CGROUP
959 	unsigned			use_memdelay:1;
960 #endif
961 #ifdef CONFIG_PSI
962 	/* Stalled due to lack of memory */
963 	unsigned			in_memstall:1;
964 #endif
965 #ifdef CONFIG_PAGE_OWNER
966 	/* Used by page_owner=on to detect recursion in page tracking. */
967 	unsigned			in_page_owner:1;
968 #endif
969 #ifdef CONFIG_EVENTFD
970 	/* Recursion prevention for eventfd_signal() */
971 	unsigned			in_eventfd:1;
972 #endif
973 #ifdef CONFIG_ARCH_HAS_CPU_PASID
974 	unsigned			pasid_activated:1;
975 #endif
976 #ifdef	CONFIG_CPU_SUP_INTEL
977 	unsigned			reported_split_lock:1;
978 #endif
979 #ifdef CONFIG_TASK_DELAY_ACCT
980 	/* delay due to memory thrashing */
981 	unsigned                        in_thrashing:1;
982 #endif
983 
984 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
985 
986 	struct restart_block		restart_block;
987 
988 	pid_t				pid;
989 	pid_t				tgid;
990 
991 #ifdef CONFIG_STACKPROTECTOR
992 	/* Canary value for the -fstack-protector GCC feature: */
993 	unsigned long			stack_canary;
994 #endif
995 	/*
996 	 * Pointers to the (original) parent process, youngest child, younger sibling,
997 	 * older sibling, respectively.  (p->father can be replaced with
998 	 * p->real_parent->pid)
999 	 */
1000 
1001 	/* Real parent process: */
1002 	struct task_struct __rcu	*real_parent;
1003 
1004 	/* Recipient of SIGCHLD, wait4() reports: */
1005 	struct task_struct __rcu	*parent;
1006 
1007 	/*
1008 	 * Children/sibling form the list of natural children:
1009 	 */
1010 	struct list_head		children;
1011 	struct list_head		sibling;
1012 	struct task_struct		*group_leader;
1013 
1014 	/*
1015 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
1016 	 *
1017 	 * This includes both natural children and PTRACE_ATTACH targets.
1018 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1019 	 */
1020 	struct list_head		ptraced;
1021 	struct list_head		ptrace_entry;
1022 
1023 	/* PID/PID hash table linkage. */
1024 	struct pid			*thread_pid;
1025 	struct hlist_node		pid_links[PIDTYPE_MAX];
1026 	struct list_head		thread_node;
1027 
1028 	struct completion		*vfork_done;
1029 
1030 	/* CLONE_CHILD_SETTID: */
1031 	int __user			*set_child_tid;
1032 
1033 	/* CLONE_CHILD_CLEARTID: */
1034 	int __user			*clear_child_tid;
1035 
1036 	/* PF_KTHREAD | PF_IO_WORKER */
1037 	void				*worker_private;
1038 
1039 	u64				utime;
1040 	u64				stime;
1041 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1042 	u64				utimescaled;
1043 	u64				stimescaled;
1044 #endif
1045 	u64				gtime;
1046 	struct prev_cputime		prev_cputime;
1047 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1048 	struct vtime			vtime;
1049 #endif
1050 
1051 #ifdef CONFIG_NO_HZ_FULL
1052 	atomic_t			tick_dep_mask;
1053 #endif
1054 	/* Context switch counts: */
1055 	unsigned long			nvcsw;
1056 	unsigned long			nivcsw;
1057 
1058 	/* Monotonic time in nsecs: */
1059 	u64				start_time;
1060 
1061 	/* Boot based time in nsecs: */
1062 	u64				start_boottime;
1063 
1064 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1065 	unsigned long			min_flt;
1066 	unsigned long			maj_flt;
1067 
1068 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1069 	struct posix_cputimers		posix_cputimers;
1070 
1071 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1072 	struct posix_cputimers_work	posix_cputimers_work;
1073 #endif
1074 
1075 	/* Process credentials: */
1076 
1077 	/* Tracer's credentials at attach: */
1078 	const struct cred __rcu		*ptracer_cred;
1079 
1080 	/* Objective and real subjective task credentials (COW): */
1081 	const struct cred __rcu		*real_cred;
1082 
1083 	/* Effective (overridable) subjective task credentials (COW): */
1084 	const struct cred __rcu		*cred;
1085 
1086 #ifdef CONFIG_KEYS
1087 	/* Cached requested key. */
1088 	struct key			*cached_requested_key;
1089 #endif
1090 
1091 	/*
1092 	 * executable name, excluding path.
1093 	 *
1094 	 * - normally initialized setup_new_exec()
1095 	 * - access it with [gs]et_task_comm()
1096 	 * - lock it with task_lock()
1097 	 */
1098 	char				comm[TASK_COMM_LEN];
1099 
1100 	struct nameidata		*nameidata;
1101 
1102 #ifdef CONFIG_SYSVIPC
1103 	struct sysv_sem			sysvsem;
1104 	struct sysv_shm			sysvshm;
1105 #endif
1106 #ifdef CONFIG_DETECT_HUNG_TASK
1107 	unsigned long			last_switch_count;
1108 	unsigned long			last_switch_time;
1109 #endif
1110 	/* Filesystem information: */
1111 	struct fs_struct		*fs;
1112 
1113 	/* Open file information: */
1114 	struct files_struct		*files;
1115 
1116 #ifdef CONFIG_IO_URING
1117 	struct io_uring_task		*io_uring;
1118 #endif
1119 
1120 	/* Namespaces: */
1121 	struct nsproxy			*nsproxy;
1122 
1123 	/* Signal handlers: */
1124 	struct signal_struct		*signal;
1125 	struct sighand_struct __rcu		*sighand;
1126 	sigset_t			blocked;
1127 	sigset_t			real_blocked;
1128 	/* Restored if set_restore_sigmask() was used: */
1129 	sigset_t			saved_sigmask;
1130 	struct sigpending		pending;
1131 	unsigned long			sas_ss_sp;
1132 	size_t				sas_ss_size;
1133 	unsigned int			sas_ss_flags;
1134 
1135 	struct callback_head		*task_works;
1136 
1137 #ifdef CONFIG_AUDIT
1138 #ifdef CONFIG_AUDITSYSCALL
1139 	struct audit_context		*audit_context;
1140 #endif
1141 	kuid_t				loginuid;
1142 	unsigned int			sessionid;
1143 #endif
1144 	struct seccomp			seccomp;
1145 	struct syscall_user_dispatch	syscall_dispatch;
1146 
1147 	/* Thread group tracking: */
1148 	u64				parent_exec_id;
1149 	u64				self_exec_id;
1150 
1151 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1152 	spinlock_t			alloc_lock;
1153 
1154 	/* Protection of the PI data structures: */
1155 	raw_spinlock_t			pi_lock;
1156 
1157 	struct wake_q_node		wake_q;
1158 
1159 #ifdef CONFIG_RT_MUTEXES
1160 	/* PI waiters blocked on a rt_mutex held by this task: */
1161 	struct rb_root_cached		pi_waiters;
1162 	/* Updated under owner's pi_lock and rq lock */
1163 	struct task_struct		*pi_top_task;
1164 	/* Deadlock detection and priority inheritance handling: */
1165 	struct rt_mutex_waiter		*pi_blocked_on;
1166 #endif
1167 
1168 #ifdef CONFIG_DEBUG_MUTEXES
1169 	/* Mutex deadlock detection: */
1170 	struct mutex_waiter		*blocked_on;
1171 #endif
1172 
1173 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1174 	int				non_block_count;
1175 #endif
1176 
1177 #ifdef CONFIG_TRACE_IRQFLAGS
1178 	struct irqtrace_events		irqtrace;
1179 	unsigned int			hardirq_threaded;
1180 	u64				hardirq_chain_key;
1181 	int				softirqs_enabled;
1182 	int				softirq_context;
1183 	int				irq_config;
1184 #endif
1185 #ifdef CONFIG_PREEMPT_RT
1186 	int				softirq_disable_cnt;
1187 #endif
1188 
1189 #ifdef CONFIG_LOCKDEP
1190 # define MAX_LOCK_DEPTH			48UL
1191 	u64				curr_chain_key;
1192 	int				lockdep_depth;
1193 	unsigned int			lockdep_recursion;
1194 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1195 #endif
1196 
1197 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1198 	unsigned int			in_ubsan;
1199 #endif
1200 
1201 	/* Journalling filesystem info: */
1202 	void				*journal_info;
1203 
1204 	/* Stacked block device info: */
1205 	struct bio_list			*bio_list;
1206 
1207 	/* Stack plugging: */
1208 	struct blk_plug			*plug;
1209 
1210 	/* VM state: */
1211 	struct reclaim_state		*reclaim_state;
1212 
1213 	struct io_context		*io_context;
1214 
1215 #ifdef CONFIG_COMPACTION
1216 	struct capture_control		*capture_control;
1217 #endif
1218 	/* Ptrace state: */
1219 	unsigned long			ptrace_message;
1220 	kernel_siginfo_t		*last_siginfo;
1221 
1222 	struct task_io_accounting	ioac;
1223 #ifdef CONFIG_PSI
1224 	/* Pressure stall state */
1225 	unsigned int			psi_flags;
1226 #endif
1227 #ifdef CONFIG_TASK_XACCT
1228 	/* Accumulated RSS usage: */
1229 	u64				acct_rss_mem1;
1230 	/* Accumulated virtual memory usage: */
1231 	u64				acct_vm_mem1;
1232 	/* stime + utime since last update: */
1233 	u64				acct_timexpd;
1234 #endif
1235 #ifdef CONFIG_CPUSETS
1236 	/* Protected by ->alloc_lock: */
1237 	nodemask_t			mems_allowed;
1238 	/* Sequence number to catch updates: */
1239 	seqcount_spinlock_t		mems_allowed_seq;
1240 	int				cpuset_mem_spread_rotor;
1241 	int				cpuset_slab_spread_rotor;
1242 #endif
1243 #ifdef CONFIG_CGROUPS
1244 	/* Control Group info protected by css_set_lock: */
1245 	struct css_set __rcu		*cgroups;
1246 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1247 	struct list_head		cg_list;
1248 #endif
1249 #ifdef CONFIG_X86_CPU_RESCTRL
1250 	u32				closid;
1251 	u32				rmid;
1252 #endif
1253 #ifdef CONFIG_FUTEX
1254 	struct robust_list_head __user	*robust_list;
1255 #ifdef CONFIG_COMPAT
1256 	struct compat_robust_list_head __user *compat_robust_list;
1257 #endif
1258 	struct list_head		pi_state_list;
1259 	struct futex_pi_state		*pi_state_cache;
1260 	struct mutex			futex_exit_mutex;
1261 	unsigned int			futex_state;
1262 #endif
1263 #ifdef CONFIG_PERF_EVENTS
1264 	struct perf_event_context	*perf_event_ctxp;
1265 	struct mutex			perf_event_mutex;
1266 	struct list_head		perf_event_list;
1267 #endif
1268 #ifdef CONFIG_DEBUG_PREEMPT
1269 	unsigned long			preempt_disable_ip;
1270 #endif
1271 #ifdef CONFIG_NUMA
1272 	/* Protected by alloc_lock: */
1273 	struct mempolicy		*mempolicy;
1274 	short				il_prev;
1275 	u8				il_weight;
1276 	short				pref_node_fork;
1277 #endif
1278 #ifdef CONFIG_NUMA_BALANCING
1279 	int				numa_scan_seq;
1280 	unsigned int			numa_scan_period;
1281 	unsigned int			numa_scan_period_max;
1282 	int				numa_preferred_nid;
1283 	unsigned long			numa_migrate_retry;
1284 	/* Migration stamp: */
1285 	u64				node_stamp;
1286 	u64				last_task_numa_placement;
1287 	u64				last_sum_exec_runtime;
1288 	struct callback_head		numa_work;
1289 
1290 	/*
1291 	 * This pointer is only modified for current in syscall and
1292 	 * pagefault context (and for tasks being destroyed), so it can be read
1293 	 * from any of the following contexts:
1294 	 *  - RCU read-side critical section
1295 	 *  - current->numa_group from everywhere
1296 	 *  - task's runqueue locked, task not running
1297 	 */
1298 	struct numa_group __rcu		*numa_group;
1299 
1300 	/*
1301 	 * numa_faults is an array split into four regions:
1302 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1303 	 * in this precise order.
1304 	 *
1305 	 * faults_memory: Exponential decaying average of faults on a per-node
1306 	 * basis. Scheduling placement decisions are made based on these
1307 	 * counts. The values remain static for the duration of a PTE scan.
1308 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1309 	 * hinting fault was incurred.
1310 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1311 	 * during the current scan window. When the scan completes, the counts
1312 	 * in faults_memory and faults_cpu decay and these values are copied.
1313 	 */
1314 	unsigned long			*numa_faults;
1315 	unsigned long			total_numa_faults;
1316 
1317 	/*
1318 	 * numa_faults_locality tracks if faults recorded during the last
1319 	 * scan window were remote/local or failed to migrate. The task scan
1320 	 * period is adapted based on the locality of the faults with different
1321 	 * weights depending on whether they were shared or private faults
1322 	 */
1323 	unsigned long			numa_faults_locality[3];
1324 
1325 	unsigned long			numa_pages_migrated;
1326 #endif /* CONFIG_NUMA_BALANCING */
1327 
1328 #ifdef CONFIG_RSEQ
1329 	struct rseq __user *rseq;
1330 	u32 rseq_len;
1331 	u32 rseq_sig;
1332 	/*
1333 	 * RmW on rseq_event_mask must be performed atomically
1334 	 * with respect to preemption.
1335 	 */
1336 	unsigned long rseq_event_mask;
1337 #endif
1338 
1339 #ifdef CONFIG_SCHED_MM_CID
1340 	int				mm_cid;		/* Current cid in mm */
1341 	int				last_mm_cid;	/* Most recent cid in mm */
1342 	int				migrate_from_cpu;
1343 	int				mm_cid_active;	/* Whether cid bitmap is active */
1344 	struct callback_head		cid_work;
1345 #endif
1346 
1347 	struct tlbflush_unmap_batch	tlb_ubc;
1348 
1349 	/* Cache last used pipe for splice(): */
1350 	struct pipe_inode_info		*splice_pipe;
1351 
1352 	struct page_frag		task_frag;
1353 
1354 #ifdef CONFIG_TASK_DELAY_ACCT
1355 	struct task_delay_info		*delays;
1356 #endif
1357 
1358 #ifdef CONFIG_FAULT_INJECTION
1359 	int				make_it_fail;
1360 	unsigned int			fail_nth;
1361 #endif
1362 	/*
1363 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1364 	 * balance_dirty_pages() for a dirty throttling pause:
1365 	 */
1366 	int				nr_dirtied;
1367 	int				nr_dirtied_pause;
1368 	/* Start of a write-and-pause period: */
1369 	unsigned long			dirty_paused_when;
1370 
1371 #ifdef CONFIG_LATENCYTOP
1372 	int				latency_record_count;
1373 	struct latency_record		latency_record[LT_SAVECOUNT];
1374 #endif
1375 	/*
1376 	 * Time slack values; these are used to round up poll() and
1377 	 * select() etc timeout values. These are in nanoseconds.
1378 	 */
1379 	u64				timer_slack_ns;
1380 	u64				default_timer_slack_ns;
1381 
1382 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1383 	unsigned int			kasan_depth;
1384 #endif
1385 
1386 #ifdef CONFIG_KCSAN
1387 	struct kcsan_ctx		kcsan_ctx;
1388 #ifdef CONFIG_TRACE_IRQFLAGS
1389 	struct irqtrace_events		kcsan_save_irqtrace;
1390 #endif
1391 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1392 	int				kcsan_stack_depth;
1393 #endif
1394 #endif
1395 
1396 #ifdef CONFIG_KMSAN
1397 	struct kmsan_ctx		kmsan_ctx;
1398 #endif
1399 
1400 #if IS_ENABLED(CONFIG_KUNIT)
1401 	struct kunit			*kunit_test;
1402 #endif
1403 
1404 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1405 	/* Index of current stored address in ret_stack: */
1406 	int				curr_ret_stack;
1407 	int				curr_ret_depth;
1408 
1409 	/* Stack of return addresses for return function tracing: */
1410 	struct ftrace_ret_stack		*ret_stack;
1411 
1412 	/* Timestamp for last schedule: */
1413 	unsigned long long		ftrace_timestamp;
1414 
1415 	/*
1416 	 * Number of functions that haven't been traced
1417 	 * because of depth overrun:
1418 	 */
1419 	atomic_t			trace_overrun;
1420 
1421 	/* Pause tracing: */
1422 	atomic_t			tracing_graph_pause;
1423 #endif
1424 
1425 #ifdef CONFIG_TRACING
1426 	/* Bitmask and counter of trace recursion: */
1427 	unsigned long			trace_recursion;
1428 #endif /* CONFIG_TRACING */
1429 
1430 #ifdef CONFIG_KCOV
1431 	/* See kernel/kcov.c for more details. */
1432 
1433 	/* Coverage collection mode enabled for this task (0 if disabled): */
1434 	unsigned int			kcov_mode;
1435 
1436 	/* Size of the kcov_area: */
1437 	unsigned int			kcov_size;
1438 
1439 	/* Buffer for coverage collection: */
1440 	void				*kcov_area;
1441 
1442 	/* KCOV descriptor wired with this task or NULL: */
1443 	struct kcov			*kcov;
1444 
1445 	/* KCOV common handle for remote coverage collection: */
1446 	u64				kcov_handle;
1447 
1448 	/* KCOV sequence number: */
1449 	int				kcov_sequence;
1450 
1451 	/* Collect coverage from softirq context: */
1452 	unsigned int			kcov_softirq;
1453 #endif
1454 
1455 #ifdef CONFIG_MEMCG
1456 	struct mem_cgroup		*memcg_in_oom;
1457 
1458 	/* Number of pages to reclaim on returning to userland: */
1459 	unsigned int			memcg_nr_pages_over_high;
1460 
1461 	/* Used by memcontrol for targeted memcg charge: */
1462 	struct mem_cgroup		*active_memcg;
1463 #endif
1464 
1465 #ifdef CONFIG_MEMCG_KMEM
1466 	struct obj_cgroup		*objcg;
1467 #endif
1468 
1469 #ifdef CONFIG_BLK_CGROUP
1470 	struct gendisk			*throttle_disk;
1471 #endif
1472 
1473 #ifdef CONFIG_UPROBES
1474 	struct uprobe_task		*utask;
1475 #endif
1476 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1477 	unsigned int			sequential_io;
1478 	unsigned int			sequential_io_avg;
1479 #endif
1480 	struct kmap_ctrl		kmap_ctrl;
1481 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1482 	unsigned long			task_state_change;
1483 # ifdef CONFIG_PREEMPT_RT
1484 	unsigned long			saved_state_change;
1485 # endif
1486 #endif
1487 	struct rcu_head			rcu;
1488 	refcount_t			rcu_users;
1489 	int				pagefault_disabled;
1490 #ifdef CONFIG_MMU
1491 	struct task_struct		*oom_reaper_list;
1492 	struct timer_list		oom_reaper_timer;
1493 #endif
1494 #ifdef CONFIG_VMAP_STACK
1495 	struct vm_struct		*stack_vm_area;
1496 #endif
1497 #ifdef CONFIG_THREAD_INFO_IN_TASK
1498 	/* A live task holds one reference: */
1499 	refcount_t			stack_refcount;
1500 #endif
1501 #ifdef CONFIG_LIVEPATCH
1502 	int patch_state;
1503 #endif
1504 #ifdef CONFIG_SECURITY
1505 	/* Used by LSM modules for access restriction: */
1506 	void				*security;
1507 #endif
1508 #ifdef CONFIG_BPF_SYSCALL
1509 	/* Used by BPF task local storage */
1510 	struct bpf_local_storage __rcu	*bpf_storage;
1511 	/* Used for BPF run context */
1512 	struct bpf_run_ctx		*bpf_ctx;
1513 #endif
1514 
1515 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1516 	unsigned long			lowest_stack;
1517 	unsigned long			prev_lowest_stack;
1518 #endif
1519 
1520 #ifdef CONFIG_X86_MCE
1521 	void __user			*mce_vaddr;
1522 	__u64				mce_kflags;
1523 	u64				mce_addr;
1524 	__u64				mce_ripv : 1,
1525 					mce_whole_page : 1,
1526 					__mce_reserved : 62;
1527 	struct callback_head		mce_kill_me;
1528 	int				mce_count;
1529 #endif
1530 
1531 #ifdef CONFIG_KRETPROBES
1532 	struct llist_head               kretprobe_instances;
1533 #endif
1534 #ifdef CONFIG_RETHOOK
1535 	struct llist_head               rethooks;
1536 #endif
1537 
1538 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1539 	/*
1540 	 * If L1D flush is supported on mm context switch
1541 	 * then we use this callback head to queue kill work
1542 	 * to kill tasks that are not running on SMT disabled
1543 	 * cores
1544 	 */
1545 	struct callback_head		l1d_flush_kill;
1546 #endif
1547 
1548 #ifdef CONFIG_RV
1549 	/*
1550 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1551 	 * If we find justification for more monitors, we can think
1552 	 * about adding more or developing a dynamic method. So far,
1553 	 * none of these are justified.
1554 	 */
1555 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1556 #endif
1557 
1558 #ifdef CONFIG_USER_EVENTS
1559 	struct user_event_mm		*user_event_mm;
1560 #endif
1561 
1562 	/*
1563 	 * New fields for task_struct should be added above here, so that
1564 	 * they are included in the randomized portion of task_struct.
1565 	 */
1566 	randomized_struct_fields_end
1567 
1568 	/* CPU-specific state of this task: */
1569 	struct thread_struct		thread;
1570 
1571 	/*
1572 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1573 	 * structure.  It *MUST* be at the end of 'task_struct'.
1574 	 *
1575 	 * Do not put anything below here!
1576 	 */
1577 };
1578 
1579 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1580 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1581 
1582 static inline unsigned int __task_state_index(unsigned int tsk_state,
1583 					      unsigned int tsk_exit_state)
1584 {
1585 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1586 
1587 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1588 
1589 	if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1590 		state = TASK_REPORT_IDLE;
1591 
1592 	/*
1593 	 * We're lying here, but rather than expose a completely new task state
1594 	 * to userspace, we can make this appear as if the task has gone through
1595 	 * a regular rt_mutex_lock() call.
1596 	 */
1597 	if (tsk_state & TASK_RTLOCK_WAIT)
1598 		state = TASK_UNINTERRUPTIBLE;
1599 
1600 	return fls(state);
1601 }
1602 
1603 static inline unsigned int task_state_index(struct task_struct *tsk)
1604 {
1605 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1606 }
1607 
1608 static inline char task_index_to_char(unsigned int state)
1609 {
1610 	static const char state_char[] = "RSDTtXZPI";
1611 
1612 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1613 
1614 	return state_char[state];
1615 }
1616 
1617 static inline char task_state_to_char(struct task_struct *tsk)
1618 {
1619 	return task_index_to_char(task_state_index(tsk));
1620 }
1621 
1622 extern struct pid *cad_pid;
1623 
1624 /*
1625  * Per process flags
1626  */
1627 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1628 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1629 #define PF_EXITING		0x00000004	/* Getting shut down */
1630 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1631 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1632 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1633 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1634 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1635 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1636 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1637 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1638 #define PF_MEMALLOC		0x00000800	/* Allocating memory to free memory. See memalloc_noreclaim_save() */
1639 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1640 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1641 #define PF_USER_WORKER		0x00004000	/* Kernel thread cloned from userspace thread */
1642 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1643 #define PF__HOLE__00010000	0x00010000
1644 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1645 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1646 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1647 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1648 						 * I am cleaning dirty pages from some other bdi. */
1649 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1650 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1651 #define PF_MEMALLOC_NORECLAIM	0x00800000	/* All allocation requests will clear __GFP_DIRECT_RECLAIM */
1652 #define PF_MEMALLOC_NOWARN	0x01000000	/* All allocation requests will inherit __GFP_NOWARN */
1653 #define PF__HOLE__02000000	0x02000000
1654 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1655 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1656 #define PF_MEMALLOC_PIN		0x10000000	/* Allocations constrained to zones which allow long term pinning.
1657 						 * See memalloc_pin_save() */
1658 #define PF_BLOCK_TS		0x20000000	/* plug has ts that needs updating */
1659 #define PF__HOLE__40000000	0x40000000
1660 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1661 
1662 /*
1663  * Only the _current_ task can read/write to tsk->flags, but other
1664  * tasks can access tsk->flags in readonly mode for example
1665  * with tsk_used_math (like during threaded core dumping).
1666  * There is however an exception to this rule during ptrace
1667  * or during fork: the ptracer task is allowed to write to the
1668  * child->flags of its traced child (same goes for fork, the parent
1669  * can write to the child->flags), because we're guaranteed the
1670  * child is not running and in turn not changing child->flags
1671  * at the same time the parent does it.
1672  */
1673 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1674 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1675 #define clear_used_math()			clear_stopped_child_used_math(current)
1676 #define set_used_math()				set_stopped_child_used_math(current)
1677 
1678 #define conditional_stopped_child_used_math(condition, child) \
1679 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1680 
1681 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1682 
1683 #define copy_to_stopped_child_used_math(child) \
1684 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1685 
1686 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1687 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1688 #define used_math()				tsk_used_math(current)
1689 
1690 static __always_inline bool is_percpu_thread(void)
1691 {
1692 #ifdef CONFIG_SMP
1693 	return (current->flags & PF_NO_SETAFFINITY) &&
1694 		(current->nr_cpus_allowed  == 1);
1695 #else
1696 	return true;
1697 #endif
1698 }
1699 
1700 /* Per-process atomic flags. */
1701 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1702 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1703 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1704 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1705 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1706 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1707 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1708 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1709 
1710 #define TASK_PFA_TEST(name, func)					\
1711 	static inline bool task_##func(struct task_struct *p)		\
1712 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1713 
1714 #define TASK_PFA_SET(name, func)					\
1715 	static inline void task_set_##func(struct task_struct *p)	\
1716 	{ set_bit(PFA_##name, &p->atomic_flags); }
1717 
1718 #define TASK_PFA_CLEAR(name, func)					\
1719 	static inline void task_clear_##func(struct task_struct *p)	\
1720 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1721 
1722 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1723 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1724 
1725 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1726 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1727 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1728 
1729 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1730 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1731 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1732 
1733 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1734 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1735 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1736 
1737 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1738 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1739 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1740 
1741 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1742 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1743 
1744 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1745 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1746 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1747 
1748 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1749 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1750 
1751 static inline void
1752 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1753 {
1754 	current->flags &= ~flags;
1755 	current->flags |= orig_flags & flags;
1756 }
1757 
1758 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1759 extern int task_can_attach(struct task_struct *p);
1760 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1761 extern void dl_bw_free(int cpu, u64 dl_bw);
1762 #ifdef CONFIG_SMP
1763 
1764 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1765 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1766 
1767 /**
1768  * set_cpus_allowed_ptr - set CPU affinity mask of a task
1769  * @p: the task
1770  * @new_mask: CPU affinity mask
1771  *
1772  * Return: zero if successful, or a negative error code
1773  */
1774 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1775 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1776 extern void release_user_cpus_ptr(struct task_struct *p);
1777 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1778 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1779 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1780 #else
1781 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1782 {
1783 }
1784 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1785 {
1786 	if (!cpumask_test_cpu(0, new_mask))
1787 		return -EINVAL;
1788 	return 0;
1789 }
1790 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1791 {
1792 	if (src->user_cpus_ptr)
1793 		return -EINVAL;
1794 	return 0;
1795 }
1796 static inline void release_user_cpus_ptr(struct task_struct *p)
1797 {
1798 	WARN_ON(p->user_cpus_ptr);
1799 }
1800 
1801 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1802 {
1803 	return 0;
1804 }
1805 #endif
1806 
1807 extern int yield_to(struct task_struct *p, bool preempt);
1808 extern void set_user_nice(struct task_struct *p, long nice);
1809 extern int task_prio(const struct task_struct *p);
1810 
1811 /**
1812  * task_nice - return the nice value of a given task.
1813  * @p: the task in question.
1814  *
1815  * Return: The nice value [ -20 ... 0 ... 19 ].
1816  */
1817 static inline int task_nice(const struct task_struct *p)
1818 {
1819 	return PRIO_TO_NICE((p)->static_prio);
1820 }
1821 
1822 extern int can_nice(const struct task_struct *p, const int nice);
1823 extern int task_curr(const struct task_struct *p);
1824 extern int idle_cpu(int cpu);
1825 extern int available_idle_cpu(int cpu);
1826 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1827 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1828 extern void sched_set_fifo(struct task_struct *p);
1829 extern void sched_set_fifo_low(struct task_struct *p);
1830 extern void sched_set_normal(struct task_struct *p, int nice);
1831 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1832 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1833 extern struct task_struct *idle_task(int cpu);
1834 
1835 /**
1836  * is_idle_task - is the specified task an idle task?
1837  * @p: the task in question.
1838  *
1839  * Return: 1 if @p is an idle task. 0 otherwise.
1840  */
1841 static __always_inline bool is_idle_task(const struct task_struct *p)
1842 {
1843 	return !!(p->flags & PF_IDLE);
1844 }
1845 
1846 extern struct task_struct *curr_task(int cpu);
1847 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1848 
1849 void yield(void);
1850 
1851 union thread_union {
1852 	struct task_struct task;
1853 #ifndef CONFIG_THREAD_INFO_IN_TASK
1854 	struct thread_info thread_info;
1855 #endif
1856 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1857 };
1858 
1859 #ifndef CONFIG_THREAD_INFO_IN_TASK
1860 extern struct thread_info init_thread_info;
1861 #endif
1862 
1863 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1864 
1865 #ifdef CONFIG_THREAD_INFO_IN_TASK
1866 # define task_thread_info(task)	(&(task)->thread_info)
1867 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1868 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1869 #endif
1870 
1871 /*
1872  * find a task by one of its numerical ids
1873  *
1874  * find_task_by_pid_ns():
1875  *      finds a task by its pid in the specified namespace
1876  * find_task_by_vpid():
1877  *      finds a task by its virtual pid
1878  *
1879  * see also find_vpid() etc in include/linux/pid.h
1880  */
1881 
1882 extern struct task_struct *find_task_by_vpid(pid_t nr);
1883 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1884 
1885 /*
1886  * find a task by its virtual pid and get the task struct
1887  */
1888 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1889 
1890 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1891 extern int wake_up_process(struct task_struct *tsk);
1892 extern void wake_up_new_task(struct task_struct *tsk);
1893 
1894 #ifdef CONFIG_SMP
1895 extern void kick_process(struct task_struct *tsk);
1896 #else
1897 static inline void kick_process(struct task_struct *tsk) { }
1898 #endif
1899 
1900 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1901 
1902 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1903 {
1904 	__set_task_comm(tsk, from, false);
1905 }
1906 
1907 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1908 #define get_task_comm(buf, tsk) ({			\
1909 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1910 	__get_task_comm(buf, sizeof(buf), tsk);		\
1911 })
1912 
1913 #ifdef CONFIG_SMP
1914 static __always_inline void scheduler_ipi(void)
1915 {
1916 	/*
1917 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1918 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1919 	 * this IPI.
1920 	 */
1921 	preempt_fold_need_resched();
1922 }
1923 #else
1924 static inline void scheduler_ipi(void) { }
1925 #endif
1926 
1927 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1928 
1929 /*
1930  * Set thread flags in other task's structures.
1931  * See asm/thread_info.h for TIF_xxxx flags available:
1932  */
1933 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1934 {
1935 	set_ti_thread_flag(task_thread_info(tsk), flag);
1936 }
1937 
1938 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1939 {
1940 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1941 }
1942 
1943 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1944 					  bool value)
1945 {
1946 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1947 }
1948 
1949 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1950 {
1951 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1952 }
1953 
1954 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1955 {
1956 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1957 }
1958 
1959 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1960 {
1961 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1962 }
1963 
1964 static inline void set_tsk_need_resched(struct task_struct *tsk)
1965 {
1966 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1967 }
1968 
1969 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1970 {
1971 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1972 }
1973 
1974 static inline int test_tsk_need_resched(struct task_struct *tsk)
1975 {
1976 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1977 }
1978 
1979 /*
1980  * cond_resched() and cond_resched_lock(): latency reduction via
1981  * explicit rescheduling in places that are safe. The return
1982  * value indicates whether a reschedule was done in fact.
1983  * cond_resched_lock() will drop the spinlock before scheduling,
1984  */
1985 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1986 extern int __cond_resched(void);
1987 
1988 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
1989 
1990 void sched_dynamic_klp_enable(void);
1991 void sched_dynamic_klp_disable(void);
1992 
1993 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1994 
1995 static __always_inline int _cond_resched(void)
1996 {
1997 	return static_call_mod(cond_resched)();
1998 }
1999 
2000 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2001 
2002 extern int dynamic_cond_resched(void);
2003 
2004 static __always_inline int _cond_resched(void)
2005 {
2006 	return dynamic_cond_resched();
2007 }
2008 
2009 #else /* !CONFIG_PREEMPTION */
2010 
2011 static inline int _cond_resched(void)
2012 {
2013 	klp_sched_try_switch();
2014 	return __cond_resched();
2015 }
2016 
2017 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2018 
2019 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2020 
2021 static inline int _cond_resched(void)
2022 {
2023 	klp_sched_try_switch();
2024 	return 0;
2025 }
2026 
2027 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2028 
2029 #define cond_resched() ({			\
2030 	__might_resched(__FILE__, __LINE__, 0);	\
2031 	_cond_resched();			\
2032 })
2033 
2034 extern int __cond_resched_lock(spinlock_t *lock);
2035 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2036 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2037 
2038 #define MIGHT_RESCHED_RCU_SHIFT		8
2039 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2040 
2041 #ifndef CONFIG_PREEMPT_RT
2042 /*
2043  * Non RT kernels have an elevated preempt count due to the held lock,
2044  * but are not allowed to be inside a RCU read side critical section
2045  */
2046 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2047 #else
2048 /*
2049  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2050  * cond_resched*lock() has to take that into account because it checks for
2051  * preempt_count() and rcu_preempt_depth().
2052  */
2053 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2054 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2055 #endif
2056 
2057 #define cond_resched_lock(lock) ({						\
2058 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2059 	__cond_resched_lock(lock);						\
2060 })
2061 
2062 #define cond_resched_rwlock_read(lock) ({					\
2063 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2064 	__cond_resched_rwlock_read(lock);					\
2065 })
2066 
2067 #define cond_resched_rwlock_write(lock) ({					\
2068 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2069 	__cond_resched_rwlock_write(lock);					\
2070 })
2071 
2072 static __always_inline bool need_resched(void)
2073 {
2074 	return unlikely(tif_need_resched());
2075 }
2076 
2077 /*
2078  * Wrappers for p->thread_info->cpu access. No-op on UP.
2079  */
2080 #ifdef CONFIG_SMP
2081 
2082 static inline unsigned int task_cpu(const struct task_struct *p)
2083 {
2084 	return READ_ONCE(task_thread_info(p)->cpu);
2085 }
2086 
2087 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2088 
2089 #else
2090 
2091 static inline unsigned int task_cpu(const struct task_struct *p)
2092 {
2093 	return 0;
2094 }
2095 
2096 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2097 {
2098 }
2099 
2100 #endif /* CONFIG_SMP */
2101 
2102 extern bool sched_task_on_rq(struct task_struct *p);
2103 extern unsigned long get_wchan(struct task_struct *p);
2104 extern struct task_struct *cpu_curr_snapshot(int cpu);
2105 
2106 #include <linux/spinlock.h>
2107 
2108 /*
2109  * In order to reduce various lock holder preemption latencies provide an
2110  * interface to see if a vCPU is currently running or not.
2111  *
2112  * This allows us to terminate optimistic spin loops and block, analogous to
2113  * the native optimistic spin heuristic of testing if the lock owner task is
2114  * running or not.
2115  */
2116 #ifndef vcpu_is_preempted
2117 static inline bool vcpu_is_preempted(int cpu)
2118 {
2119 	return false;
2120 }
2121 #endif
2122 
2123 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2124 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2125 
2126 #ifndef TASK_SIZE_OF
2127 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2128 #endif
2129 
2130 #ifdef CONFIG_SMP
2131 static inline bool owner_on_cpu(struct task_struct *owner)
2132 {
2133 	/*
2134 	 * As lock holder preemption issue, we both skip spinning if
2135 	 * task is not on cpu or its cpu is preempted
2136 	 */
2137 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2138 }
2139 
2140 /* Returns effective CPU energy utilization, as seen by the scheduler */
2141 unsigned long sched_cpu_util(int cpu);
2142 #endif /* CONFIG_SMP */
2143 
2144 #ifdef CONFIG_SCHED_CORE
2145 extern void sched_core_free(struct task_struct *tsk);
2146 extern void sched_core_fork(struct task_struct *p);
2147 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2148 				unsigned long uaddr);
2149 extern int sched_core_idle_cpu(int cpu);
2150 #else
2151 static inline void sched_core_free(struct task_struct *tsk) { }
2152 static inline void sched_core_fork(struct task_struct *p) { }
2153 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2154 #endif
2155 
2156 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2157 
2158 #ifdef CONFIG_MEM_ALLOC_PROFILING
2159 static inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2160 {
2161 	swap(current->alloc_tag, tag);
2162 	return tag;
2163 }
2164 
2165 static inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2166 {
2167 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2168 	WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2169 #endif
2170 	current->alloc_tag = old;
2171 }
2172 #else
2173 #define alloc_tag_save(_tag)			NULL
2174 #define alloc_tag_restore(_tag, _old)		do {} while (0)
2175 #endif
2176 
2177 #endif
2178