xref: /linux/include/linux/sched.h (revision e0fcfb086fbbb6233de1062d4b2f05e9afedab3b)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4 
5 /*
6  * Define 'struct task_struct' and provide the main scheduler
7  * APIs (schedule(), wakeup variants, etc.)
8  */
9 
10 #include <uapi/linux/sched.h>
11 
12 #include <asm/current.h>
13 
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/refcount.h>
25 #include <linux/resource.h>
26 #include <linux/latencytop.h>
27 #include <linux/sched/prio.h>
28 #include <linux/sched/types.h>
29 #include <linux/signal_types.h>
30 #include <linux/mm_types_task.h>
31 #include <linux/task_io_accounting.h>
32 #include <linux/posix-timers.h>
33 #include <linux/rseq.h>
34 
35 /* task_struct member predeclarations (sorted alphabetically): */
36 struct audit_context;
37 struct backing_dev_info;
38 struct bio_list;
39 struct blk_plug;
40 struct capture_control;
41 struct cfs_rq;
42 struct fs_struct;
43 struct futex_pi_state;
44 struct io_context;
45 struct mempolicy;
46 struct nameidata;
47 struct nsproxy;
48 struct perf_event_context;
49 struct pid_namespace;
50 struct pipe_inode_info;
51 struct rcu_node;
52 struct reclaim_state;
53 struct robust_list_head;
54 struct root_domain;
55 struct rq;
56 struct sched_attr;
57 struct sched_param;
58 struct seq_file;
59 struct sighand_struct;
60 struct signal_struct;
61 struct task_delay_info;
62 struct task_group;
63 
64 /*
65  * Task state bitmask. NOTE! These bits are also
66  * encoded in fs/proc/array.c: get_task_state().
67  *
68  * We have two separate sets of flags: task->state
69  * is about runnability, while task->exit_state are
70  * about the task exiting. Confusing, but this way
71  * modifying one set can't modify the other one by
72  * mistake.
73  */
74 
75 /* Used in tsk->state: */
76 #define TASK_RUNNING			0x0000
77 #define TASK_INTERRUPTIBLE		0x0001
78 #define TASK_UNINTERRUPTIBLE		0x0002
79 #define __TASK_STOPPED			0x0004
80 #define __TASK_TRACED			0x0008
81 /* Used in tsk->exit_state: */
82 #define EXIT_DEAD			0x0010
83 #define EXIT_ZOMBIE			0x0020
84 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
85 /* Used in tsk->state again: */
86 #define TASK_PARKED			0x0040
87 #define TASK_DEAD			0x0080
88 #define TASK_WAKEKILL			0x0100
89 #define TASK_WAKING			0x0200
90 #define TASK_NOLOAD			0x0400
91 #define TASK_NEW			0x0800
92 #define TASK_STATE_MAX			0x1000
93 
94 /* Convenience macros for the sake of set_current_state: */
95 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
96 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
97 #define TASK_TRACED			(TASK_WAKEKILL | __TASK_TRACED)
98 
99 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
100 
101 /* Convenience macros for the sake of wake_up(): */
102 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
103 
104 /* get_task_state(): */
105 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
106 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
107 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
108 					 TASK_PARKED)
109 
110 #define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
111 
112 #define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
113 
114 #define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
115 
116 #define task_contributes_to_load(task)	((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
117 					 (task->flags & PF_FROZEN) == 0 && \
118 					 (task->state & TASK_NOLOAD) == 0)
119 
120 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
121 
122 /*
123  * Special states are those that do not use the normal wait-loop pattern. See
124  * the comment with set_special_state().
125  */
126 #define is_special_task_state(state)				\
127 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
128 
129 #define __set_current_state(state_value)			\
130 	do {							\
131 		WARN_ON_ONCE(is_special_task_state(state_value));\
132 		current->task_state_change = _THIS_IP_;		\
133 		current->state = (state_value);			\
134 	} while (0)
135 
136 #define set_current_state(state_value)				\
137 	do {							\
138 		WARN_ON_ONCE(is_special_task_state(state_value));\
139 		current->task_state_change = _THIS_IP_;		\
140 		smp_store_mb(current->state, (state_value));	\
141 	} while (0)
142 
143 #define set_special_state(state_value)					\
144 	do {								\
145 		unsigned long flags; /* may shadow */			\
146 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
147 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
148 		current->task_state_change = _THIS_IP_;			\
149 		current->state = (state_value);				\
150 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
151 	} while (0)
152 #else
153 /*
154  * set_current_state() includes a barrier so that the write of current->state
155  * is correctly serialised wrt the caller's subsequent test of whether to
156  * actually sleep:
157  *
158  *   for (;;) {
159  *	set_current_state(TASK_UNINTERRUPTIBLE);
160  *	if (!need_sleep)
161  *		break;
162  *
163  *	schedule();
164  *   }
165  *   __set_current_state(TASK_RUNNING);
166  *
167  * If the caller does not need such serialisation (because, for instance, the
168  * condition test and condition change and wakeup are under the same lock) then
169  * use __set_current_state().
170  *
171  * The above is typically ordered against the wakeup, which does:
172  *
173  *   need_sleep = false;
174  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
175  *
176  * where wake_up_state() executes a full memory barrier before accessing the
177  * task state.
178  *
179  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
180  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
181  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
182  *
183  * However, with slightly different timing the wakeup TASK_RUNNING store can
184  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
185  * a problem either because that will result in one extra go around the loop
186  * and our @cond test will save the day.
187  *
188  * Also see the comments of try_to_wake_up().
189  */
190 #define __set_current_state(state_value)				\
191 	current->state = (state_value)
192 
193 #define set_current_state(state_value)					\
194 	smp_store_mb(current->state, (state_value))
195 
196 /*
197  * set_special_state() should be used for those states when the blocking task
198  * can not use the regular condition based wait-loop. In that case we must
199  * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
200  * will not collide with our state change.
201  */
202 #define set_special_state(state_value)					\
203 	do {								\
204 		unsigned long flags; /* may shadow */			\
205 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
206 		current->state = (state_value);				\
207 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
208 	} while (0)
209 
210 #endif
211 
212 /* Task command name length: */
213 #define TASK_COMM_LEN			16
214 
215 extern void scheduler_tick(void);
216 
217 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
218 
219 extern long schedule_timeout(long timeout);
220 extern long schedule_timeout_interruptible(long timeout);
221 extern long schedule_timeout_killable(long timeout);
222 extern long schedule_timeout_uninterruptible(long timeout);
223 extern long schedule_timeout_idle(long timeout);
224 asmlinkage void schedule(void);
225 extern void schedule_preempt_disabled(void);
226 
227 extern int __must_check io_schedule_prepare(void);
228 extern void io_schedule_finish(int token);
229 extern long io_schedule_timeout(long timeout);
230 extern void io_schedule(void);
231 
232 /**
233  * struct prev_cputime - snapshot of system and user cputime
234  * @utime: time spent in user mode
235  * @stime: time spent in system mode
236  * @lock: protects the above two fields
237  *
238  * Stores previous user/system time values such that we can guarantee
239  * monotonicity.
240  */
241 struct prev_cputime {
242 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
243 	u64				utime;
244 	u64				stime;
245 	raw_spinlock_t			lock;
246 #endif
247 };
248 
249 enum vtime_state {
250 	/* Task is sleeping or running in a CPU with VTIME inactive: */
251 	VTIME_INACTIVE = 0,
252 	/* Task runs in userspace in a CPU with VTIME active: */
253 	VTIME_USER,
254 	/* Task runs in kernelspace in a CPU with VTIME active: */
255 	VTIME_SYS,
256 };
257 
258 struct vtime {
259 	seqcount_t		seqcount;
260 	unsigned long long	starttime;
261 	enum vtime_state	state;
262 	u64			utime;
263 	u64			stime;
264 	u64			gtime;
265 };
266 
267 /*
268  * Utilization clamp constraints.
269  * @UCLAMP_MIN:	Minimum utilization
270  * @UCLAMP_MAX:	Maximum utilization
271  * @UCLAMP_CNT:	Utilization clamp constraints count
272  */
273 enum uclamp_id {
274 	UCLAMP_MIN = 0,
275 	UCLAMP_MAX,
276 	UCLAMP_CNT
277 };
278 
279 #ifdef CONFIG_SMP
280 extern struct root_domain def_root_domain;
281 extern struct mutex sched_domains_mutex;
282 #endif
283 
284 struct sched_info {
285 #ifdef CONFIG_SCHED_INFO
286 	/* Cumulative counters: */
287 
288 	/* # of times we have run on this CPU: */
289 	unsigned long			pcount;
290 
291 	/* Time spent waiting on a runqueue: */
292 	unsigned long long		run_delay;
293 
294 	/* Timestamps: */
295 
296 	/* When did we last run on a CPU? */
297 	unsigned long long		last_arrival;
298 
299 	/* When were we last queued to run? */
300 	unsigned long long		last_queued;
301 
302 #endif /* CONFIG_SCHED_INFO */
303 };
304 
305 /*
306  * Integer metrics need fixed point arithmetic, e.g., sched/fair
307  * has a few: load, load_avg, util_avg, freq, and capacity.
308  *
309  * We define a basic fixed point arithmetic range, and then formalize
310  * all these metrics based on that basic range.
311  */
312 # define SCHED_FIXEDPOINT_SHIFT		10
313 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
314 
315 /* Increase resolution of cpu_capacity calculations */
316 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
317 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
318 
319 struct load_weight {
320 	unsigned long			weight;
321 	u32				inv_weight;
322 };
323 
324 /**
325  * struct util_est - Estimation utilization of FAIR tasks
326  * @enqueued: instantaneous estimated utilization of a task/cpu
327  * @ewma:     the Exponential Weighted Moving Average (EWMA)
328  *            utilization of a task
329  *
330  * Support data structure to track an Exponential Weighted Moving Average
331  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
332  * average each time a task completes an activation. Sample's weight is chosen
333  * so that the EWMA will be relatively insensitive to transient changes to the
334  * task's workload.
335  *
336  * The enqueued attribute has a slightly different meaning for tasks and cpus:
337  * - task:   the task's util_avg at last task dequeue time
338  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
339  * Thus, the util_est.enqueued of a task represents the contribution on the
340  * estimated utilization of the CPU where that task is currently enqueued.
341  *
342  * Only for tasks we track a moving average of the past instantaneous
343  * estimated utilization. This allows to absorb sporadic drops in utilization
344  * of an otherwise almost periodic task.
345  */
346 struct util_est {
347 	unsigned int			enqueued;
348 	unsigned int			ewma;
349 #define UTIL_EST_WEIGHT_SHIFT		2
350 } __attribute__((__aligned__(sizeof(u64))));
351 
352 /*
353  * The load_avg/util_avg accumulates an infinite geometric series
354  * (see __update_load_avg() in kernel/sched/fair.c).
355  *
356  * [load_avg definition]
357  *
358  *   load_avg = runnable% * scale_load_down(load)
359  *
360  * where runnable% is the time ratio that a sched_entity is runnable.
361  * For cfs_rq, it is the aggregated load_avg of all runnable and
362  * blocked sched_entities.
363  *
364  * [util_avg definition]
365  *
366  *   util_avg = running% * SCHED_CAPACITY_SCALE
367  *
368  * where running% is the time ratio that a sched_entity is running on
369  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
370  * and blocked sched_entities.
371  *
372  * load_avg and util_avg don't direcly factor frequency scaling and CPU
373  * capacity scaling. The scaling is done through the rq_clock_pelt that
374  * is used for computing those signals (see update_rq_clock_pelt())
375  *
376  * N.B., the above ratios (runnable% and running%) themselves are in the
377  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
378  * to as large a range as necessary. This is for example reflected by
379  * util_avg's SCHED_CAPACITY_SCALE.
380  *
381  * [Overflow issue]
382  *
383  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
384  * with the highest load (=88761), always runnable on a single cfs_rq,
385  * and should not overflow as the number already hits PID_MAX_LIMIT.
386  *
387  * For all other cases (including 32-bit kernels), struct load_weight's
388  * weight will overflow first before we do, because:
389  *
390  *    Max(load_avg) <= Max(load.weight)
391  *
392  * Then it is the load_weight's responsibility to consider overflow
393  * issues.
394  */
395 struct sched_avg {
396 	u64				last_update_time;
397 	u64				load_sum;
398 	u64				runnable_load_sum;
399 	u32				util_sum;
400 	u32				period_contrib;
401 	unsigned long			load_avg;
402 	unsigned long			runnable_load_avg;
403 	unsigned long			util_avg;
404 	struct util_est			util_est;
405 } ____cacheline_aligned;
406 
407 struct sched_statistics {
408 #ifdef CONFIG_SCHEDSTATS
409 	u64				wait_start;
410 	u64				wait_max;
411 	u64				wait_count;
412 	u64				wait_sum;
413 	u64				iowait_count;
414 	u64				iowait_sum;
415 
416 	u64				sleep_start;
417 	u64				sleep_max;
418 	s64				sum_sleep_runtime;
419 
420 	u64				block_start;
421 	u64				block_max;
422 	u64				exec_max;
423 	u64				slice_max;
424 
425 	u64				nr_migrations_cold;
426 	u64				nr_failed_migrations_affine;
427 	u64				nr_failed_migrations_running;
428 	u64				nr_failed_migrations_hot;
429 	u64				nr_forced_migrations;
430 
431 	u64				nr_wakeups;
432 	u64				nr_wakeups_sync;
433 	u64				nr_wakeups_migrate;
434 	u64				nr_wakeups_local;
435 	u64				nr_wakeups_remote;
436 	u64				nr_wakeups_affine;
437 	u64				nr_wakeups_affine_attempts;
438 	u64				nr_wakeups_passive;
439 	u64				nr_wakeups_idle;
440 #endif
441 };
442 
443 struct sched_entity {
444 	/* For load-balancing: */
445 	struct load_weight		load;
446 	unsigned long			runnable_weight;
447 	struct rb_node			run_node;
448 	struct list_head		group_node;
449 	unsigned int			on_rq;
450 
451 	u64				exec_start;
452 	u64				sum_exec_runtime;
453 	u64				vruntime;
454 	u64				prev_sum_exec_runtime;
455 
456 	u64				nr_migrations;
457 
458 	struct sched_statistics		statistics;
459 
460 #ifdef CONFIG_FAIR_GROUP_SCHED
461 	int				depth;
462 	struct sched_entity		*parent;
463 	/* rq on which this entity is (to be) queued: */
464 	struct cfs_rq			*cfs_rq;
465 	/* rq "owned" by this entity/group: */
466 	struct cfs_rq			*my_q;
467 #endif
468 
469 #ifdef CONFIG_SMP
470 	/*
471 	 * Per entity load average tracking.
472 	 *
473 	 * Put into separate cache line so it does not
474 	 * collide with read-mostly values above.
475 	 */
476 	struct sched_avg		avg;
477 #endif
478 };
479 
480 struct sched_rt_entity {
481 	struct list_head		run_list;
482 	unsigned long			timeout;
483 	unsigned long			watchdog_stamp;
484 	unsigned int			time_slice;
485 	unsigned short			on_rq;
486 	unsigned short			on_list;
487 
488 	struct sched_rt_entity		*back;
489 #ifdef CONFIG_RT_GROUP_SCHED
490 	struct sched_rt_entity		*parent;
491 	/* rq on which this entity is (to be) queued: */
492 	struct rt_rq			*rt_rq;
493 	/* rq "owned" by this entity/group: */
494 	struct rt_rq			*my_q;
495 #endif
496 } __randomize_layout;
497 
498 struct sched_dl_entity {
499 	struct rb_node			rb_node;
500 
501 	/*
502 	 * Original scheduling parameters. Copied here from sched_attr
503 	 * during sched_setattr(), they will remain the same until
504 	 * the next sched_setattr().
505 	 */
506 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
507 	u64				dl_deadline;	/* Relative deadline of each instance	*/
508 	u64				dl_period;	/* Separation of two instances (period) */
509 	u64				dl_bw;		/* dl_runtime / dl_period		*/
510 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
511 
512 	/*
513 	 * Actual scheduling parameters. Initialized with the values above,
514 	 * they are continuously updated during task execution. Note that
515 	 * the remaining runtime could be < 0 in case we are in overrun.
516 	 */
517 	s64				runtime;	/* Remaining runtime for this instance	*/
518 	u64				deadline;	/* Absolute deadline for this instance	*/
519 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
520 
521 	/*
522 	 * Some bool flags:
523 	 *
524 	 * @dl_throttled tells if we exhausted the runtime. If so, the
525 	 * task has to wait for a replenishment to be performed at the
526 	 * next firing of dl_timer.
527 	 *
528 	 * @dl_boosted tells if we are boosted due to DI. If so we are
529 	 * outside bandwidth enforcement mechanism (but only until we
530 	 * exit the critical section);
531 	 *
532 	 * @dl_yielded tells if task gave up the CPU before consuming
533 	 * all its available runtime during the last job.
534 	 *
535 	 * @dl_non_contending tells if the task is inactive while still
536 	 * contributing to the active utilization. In other words, it
537 	 * indicates if the inactive timer has been armed and its handler
538 	 * has not been executed yet. This flag is useful to avoid race
539 	 * conditions between the inactive timer handler and the wakeup
540 	 * code.
541 	 *
542 	 * @dl_overrun tells if the task asked to be informed about runtime
543 	 * overruns.
544 	 */
545 	unsigned int			dl_throttled      : 1;
546 	unsigned int			dl_boosted        : 1;
547 	unsigned int			dl_yielded        : 1;
548 	unsigned int			dl_non_contending : 1;
549 	unsigned int			dl_overrun	  : 1;
550 
551 	/*
552 	 * Bandwidth enforcement timer. Each -deadline task has its
553 	 * own bandwidth to be enforced, thus we need one timer per task.
554 	 */
555 	struct hrtimer			dl_timer;
556 
557 	/*
558 	 * Inactive timer, responsible for decreasing the active utilization
559 	 * at the "0-lag time". When a -deadline task blocks, it contributes
560 	 * to GRUB's active utilization until the "0-lag time", hence a
561 	 * timer is needed to decrease the active utilization at the correct
562 	 * time.
563 	 */
564 	struct hrtimer inactive_timer;
565 };
566 
567 #ifdef CONFIG_UCLAMP_TASK
568 /* Number of utilization clamp buckets (shorter alias) */
569 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
570 
571 /*
572  * Utilization clamp for a scheduling entity
573  * @value:		clamp value "assigned" to a se
574  * @bucket_id:		bucket index corresponding to the "assigned" value
575  * @active:		the se is currently refcounted in a rq's bucket
576  * @user_defined:	the requested clamp value comes from user-space
577  *
578  * The bucket_id is the index of the clamp bucket matching the clamp value
579  * which is pre-computed and stored to avoid expensive integer divisions from
580  * the fast path.
581  *
582  * The active bit is set whenever a task has got an "effective" value assigned,
583  * which can be different from the clamp value "requested" from user-space.
584  * This allows to know a task is refcounted in the rq's bucket corresponding
585  * to the "effective" bucket_id.
586  *
587  * The user_defined bit is set whenever a task has got a task-specific clamp
588  * value requested from userspace, i.e. the system defaults apply to this task
589  * just as a restriction. This allows to relax default clamps when a less
590  * restrictive task-specific value has been requested, thus allowing to
591  * implement a "nice" semantic. For example, a task running with a 20%
592  * default boost can still drop its own boosting to 0%.
593  */
594 struct uclamp_se {
595 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
596 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
597 	unsigned int active		: 1;
598 	unsigned int user_defined	: 1;
599 };
600 #endif /* CONFIG_UCLAMP_TASK */
601 
602 union rcu_special {
603 	struct {
604 		u8			blocked;
605 		u8			need_qs;
606 		u8			exp_hint; /* Hint for performance. */
607 		u8			deferred_qs;
608 	} b; /* Bits. */
609 	u32 s; /* Set of bits. */
610 };
611 
612 enum perf_event_task_context {
613 	perf_invalid_context = -1,
614 	perf_hw_context = 0,
615 	perf_sw_context,
616 	perf_nr_task_contexts,
617 };
618 
619 struct wake_q_node {
620 	struct wake_q_node *next;
621 };
622 
623 struct task_struct {
624 #ifdef CONFIG_THREAD_INFO_IN_TASK
625 	/*
626 	 * For reasons of header soup (see current_thread_info()), this
627 	 * must be the first element of task_struct.
628 	 */
629 	struct thread_info		thread_info;
630 #endif
631 	/* -1 unrunnable, 0 runnable, >0 stopped: */
632 	volatile long			state;
633 
634 	/*
635 	 * This begins the randomizable portion of task_struct. Only
636 	 * scheduling-critical items should be added above here.
637 	 */
638 	randomized_struct_fields_start
639 
640 	void				*stack;
641 	refcount_t			usage;
642 	/* Per task flags (PF_*), defined further below: */
643 	unsigned int			flags;
644 	unsigned int			ptrace;
645 
646 #ifdef CONFIG_SMP
647 	struct llist_node		wake_entry;
648 	int				on_cpu;
649 #ifdef CONFIG_THREAD_INFO_IN_TASK
650 	/* Current CPU: */
651 	unsigned int			cpu;
652 #endif
653 	unsigned int			wakee_flips;
654 	unsigned long			wakee_flip_decay_ts;
655 	struct task_struct		*last_wakee;
656 
657 	/*
658 	 * recent_used_cpu is initially set as the last CPU used by a task
659 	 * that wakes affine another task. Waker/wakee relationships can
660 	 * push tasks around a CPU where each wakeup moves to the next one.
661 	 * Tracking a recently used CPU allows a quick search for a recently
662 	 * used CPU that may be idle.
663 	 */
664 	int				recent_used_cpu;
665 	int				wake_cpu;
666 #endif
667 	int				on_rq;
668 
669 	int				prio;
670 	int				static_prio;
671 	int				normal_prio;
672 	unsigned int			rt_priority;
673 
674 	const struct sched_class	*sched_class;
675 	struct sched_entity		se;
676 	struct sched_rt_entity		rt;
677 #ifdef CONFIG_CGROUP_SCHED
678 	struct task_group		*sched_task_group;
679 #endif
680 	struct sched_dl_entity		dl;
681 
682 #ifdef CONFIG_UCLAMP_TASK
683 	/* Clamp values requested for a scheduling entity */
684 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
685 	/* Effective clamp values used for a scheduling entity */
686 	struct uclamp_se		uclamp[UCLAMP_CNT];
687 #endif
688 
689 #ifdef CONFIG_PREEMPT_NOTIFIERS
690 	/* List of struct preempt_notifier: */
691 	struct hlist_head		preempt_notifiers;
692 #endif
693 
694 #ifdef CONFIG_BLK_DEV_IO_TRACE
695 	unsigned int			btrace_seq;
696 #endif
697 
698 	unsigned int			policy;
699 	int				nr_cpus_allowed;
700 	const cpumask_t			*cpus_ptr;
701 	cpumask_t			cpus_mask;
702 
703 #ifdef CONFIG_PREEMPT_RCU
704 	int				rcu_read_lock_nesting;
705 	union rcu_special		rcu_read_unlock_special;
706 	struct list_head		rcu_node_entry;
707 	struct rcu_node			*rcu_blocked_node;
708 #endif /* #ifdef CONFIG_PREEMPT_RCU */
709 
710 #ifdef CONFIG_TASKS_RCU
711 	unsigned long			rcu_tasks_nvcsw;
712 	u8				rcu_tasks_holdout;
713 	u8				rcu_tasks_idx;
714 	int				rcu_tasks_idle_cpu;
715 	struct list_head		rcu_tasks_holdout_list;
716 #endif /* #ifdef CONFIG_TASKS_RCU */
717 
718 	struct sched_info		sched_info;
719 
720 	struct list_head		tasks;
721 #ifdef CONFIG_SMP
722 	struct plist_node		pushable_tasks;
723 	struct rb_node			pushable_dl_tasks;
724 #endif
725 
726 	struct mm_struct		*mm;
727 	struct mm_struct		*active_mm;
728 
729 	/* Per-thread vma caching: */
730 	struct vmacache			vmacache;
731 
732 #ifdef SPLIT_RSS_COUNTING
733 	struct task_rss_stat		rss_stat;
734 #endif
735 	int				exit_state;
736 	int				exit_code;
737 	int				exit_signal;
738 	/* The signal sent when the parent dies: */
739 	int				pdeath_signal;
740 	/* JOBCTL_*, siglock protected: */
741 	unsigned long			jobctl;
742 
743 	/* Used for emulating ABI behavior of previous Linux versions: */
744 	unsigned int			personality;
745 
746 	/* Scheduler bits, serialized by scheduler locks: */
747 	unsigned			sched_reset_on_fork:1;
748 	unsigned			sched_contributes_to_load:1;
749 	unsigned			sched_migrated:1;
750 	unsigned			sched_remote_wakeup:1;
751 #ifdef CONFIG_PSI
752 	unsigned			sched_psi_wake_requeue:1;
753 #endif
754 
755 	/* Force alignment to the next boundary: */
756 	unsigned			:0;
757 
758 	/* Unserialized, strictly 'current' */
759 
760 	/* Bit to tell LSMs we're in execve(): */
761 	unsigned			in_execve:1;
762 	unsigned			in_iowait:1;
763 #ifndef TIF_RESTORE_SIGMASK
764 	unsigned			restore_sigmask:1;
765 #endif
766 #ifdef CONFIG_MEMCG
767 	unsigned			in_user_fault:1;
768 #endif
769 #ifdef CONFIG_COMPAT_BRK
770 	unsigned			brk_randomized:1;
771 #endif
772 #ifdef CONFIG_CGROUPS
773 	/* disallow userland-initiated cgroup migration */
774 	unsigned			no_cgroup_migration:1;
775 	/* task is frozen/stopped (used by the cgroup freezer) */
776 	unsigned			frozen:1;
777 #endif
778 #ifdef CONFIG_BLK_CGROUP
779 	/* to be used once the psi infrastructure lands upstream. */
780 	unsigned			use_memdelay:1;
781 #endif
782 
783 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
784 
785 	struct restart_block		restart_block;
786 
787 	pid_t				pid;
788 	pid_t				tgid;
789 
790 #ifdef CONFIG_STACKPROTECTOR
791 	/* Canary value for the -fstack-protector GCC feature: */
792 	unsigned long			stack_canary;
793 #endif
794 	/*
795 	 * Pointers to the (original) parent process, youngest child, younger sibling,
796 	 * older sibling, respectively.  (p->father can be replaced with
797 	 * p->real_parent->pid)
798 	 */
799 
800 	/* Real parent process: */
801 	struct task_struct __rcu	*real_parent;
802 
803 	/* Recipient of SIGCHLD, wait4() reports: */
804 	struct task_struct __rcu	*parent;
805 
806 	/*
807 	 * Children/sibling form the list of natural children:
808 	 */
809 	struct list_head		children;
810 	struct list_head		sibling;
811 	struct task_struct		*group_leader;
812 
813 	/*
814 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
815 	 *
816 	 * This includes both natural children and PTRACE_ATTACH targets.
817 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
818 	 */
819 	struct list_head		ptraced;
820 	struct list_head		ptrace_entry;
821 
822 	/* PID/PID hash table linkage. */
823 	struct pid			*thread_pid;
824 	struct hlist_node		pid_links[PIDTYPE_MAX];
825 	struct list_head		thread_group;
826 	struct list_head		thread_node;
827 
828 	struct completion		*vfork_done;
829 
830 	/* CLONE_CHILD_SETTID: */
831 	int __user			*set_child_tid;
832 
833 	/* CLONE_CHILD_CLEARTID: */
834 	int __user			*clear_child_tid;
835 
836 	u64				utime;
837 	u64				stime;
838 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
839 	u64				utimescaled;
840 	u64				stimescaled;
841 #endif
842 	u64				gtime;
843 	struct prev_cputime		prev_cputime;
844 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
845 	struct vtime			vtime;
846 #endif
847 
848 #ifdef CONFIG_NO_HZ_FULL
849 	atomic_t			tick_dep_mask;
850 #endif
851 	/* Context switch counts: */
852 	unsigned long			nvcsw;
853 	unsigned long			nivcsw;
854 
855 	/* Monotonic time in nsecs: */
856 	u64				start_time;
857 
858 	/* Boot based time in nsecs: */
859 	u64				real_start_time;
860 
861 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
862 	unsigned long			min_flt;
863 	unsigned long			maj_flt;
864 
865 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
866 	struct posix_cputimers		posix_cputimers;
867 
868 	/* Process credentials: */
869 
870 	/* Tracer's credentials at attach: */
871 	const struct cred __rcu		*ptracer_cred;
872 
873 	/* Objective and real subjective task credentials (COW): */
874 	const struct cred __rcu		*real_cred;
875 
876 	/* Effective (overridable) subjective task credentials (COW): */
877 	const struct cred __rcu		*cred;
878 
879 #ifdef CONFIG_KEYS
880 	/* Cached requested key. */
881 	struct key			*cached_requested_key;
882 #endif
883 
884 	/*
885 	 * executable name, excluding path.
886 	 *
887 	 * - normally initialized setup_new_exec()
888 	 * - access it with [gs]et_task_comm()
889 	 * - lock it with task_lock()
890 	 */
891 	char				comm[TASK_COMM_LEN];
892 
893 	struct nameidata		*nameidata;
894 
895 #ifdef CONFIG_SYSVIPC
896 	struct sysv_sem			sysvsem;
897 	struct sysv_shm			sysvshm;
898 #endif
899 #ifdef CONFIG_DETECT_HUNG_TASK
900 	unsigned long			last_switch_count;
901 	unsigned long			last_switch_time;
902 #endif
903 	/* Filesystem information: */
904 	struct fs_struct		*fs;
905 
906 	/* Open file information: */
907 	struct files_struct		*files;
908 
909 	/* Namespaces: */
910 	struct nsproxy			*nsproxy;
911 
912 	/* Signal handlers: */
913 	struct signal_struct		*signal;
914 	struct sighand_struct		*sighand;
915 	sigset_t			blocked;
916 	sigset_t			real_blocked;
917 	/* Restored if set_restore_sigmask() was used: */
918 	sigset_t			saved_sigmask;
919 	struct sigpending		pending;
920 	unsigned long			sas_ss_sp;
921 	size_t				sas_ss_size;
922 	unsigned int			sas_ss_flags;
923 
924 	struct callback_head		*task_works;
925 
926 #ifdef CONFIG_AUDIT
927 #ifdef CONFIG_AUDITSYSCALL
928 	struct audit_context		*audit_context;
929 #endif
930 	kuid_t				loginuid;
931 	unsigned int			sessionid;
932 #endif
933 	struct seccomp			seccomp;
934 
935 	/* Thread group tracking: */
936 	u32				parent_exec_id;
937 	u32				self_exec_id;
938 
939 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
940 	spinlock_t			alloc_lock;
941 
942 	/* Protection of the PI data structures: */
943 	raw_spinlock_t			pi_lock;
944 
945 	struct wake_q_node		wake_q;
946 
947 #ifdef CONFIG_RT_MUTEXES
948 	/* PI waiters blocked on a rt_mutex held by this task: */
949 	struct rb_root_cached		pi_waiters;
950 	/* Updated under owner's pi_lock and rq lock */
951 	struct task_struct		*pi_top_task;
952 	/* Deadlock detection and priority inheritance handling: */
953 	struct rt_mutex_waiter		*pi_blocked_on;
954 #endif
955 
956 #ifdef CONFIG_DEBUG_MUTEXES
957 	/* Mutex deadlock detection: */
958 	struct mutex_waiter		*blocked_on;
959 #endif
960 
961 #ifdef CONFIG_TRACE_IRQFLAGS
962 	unsigned int			irq_events;
963 	unsigned long			hardirq_enable_ip;
964 	unsigned long			hardirq_disable_ip;
965 	unsigned int			hardirq_enable_event;
966 	unsigned int			hardirq_disable_event;
967 	int				hardirqs_enabled;
968 	int				hardirq_context;
969 	unsigned long			softirq_disable_ip;
970 	unsigned long			softirq_enable_ip;
971 	unsigned int			softirq_disable_event;
972 	unsigned int			softirq_enable_event;
973 	int				softirqs_enabled;
974 	int				softirq_context;
975 #endif
976 
977 #ifdef CONFIG_LOCKDEP
978 # define MAX_LOCK_DEPTH			48UL
979 	u64				curr_chain_key;
980 	int				lockdep_depth;
981 	unsigned int			lockdep_recursion;
982 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
983 #endif
984 
985 #ifdef CONFIG_UBSAN
986 	unsigned int			in_ubsan;
987 #endif
988 
989 	/* Journalling filesystem info: */
990 	void				*journal_info;
991 
992 	/* Stacked block device info: */
993 	struct bio_list			*bio_list;
994 
995 #ifdef CONFIG_BLOCK
996 	/* Stack plugging: */
997 	struct blk_plug			*plug;
998 #endif
999 
1000 	/* VM state: */
1001 	struct reclaim_state		*reclaim_state;
1002 
1003 	struct backing_dev_info		*backing_dev_info;
1004 
1005 	struct io_context		*io_context;
1006 
1007 #ifdef CONFIG_COMPACTION
1008 	struct capture_control		*capture_control;
1009 #endif
1010 	/* Ptrace state: */
1011 	unsigned long			ptrace_message;
1012 	kernel_siginfo_t		*last_siginfo;
1013 
1014 	struct task_io_accounting	ioac;
1015 #ifdef CONFIG_PSI
1016 	/* Pressure stall state */
1017 	unsigned int			psi_flags;
1018 #endif
1019 #ifdef CONFIG_TASK_XACCT
1020 	/* Accumulated RSS usage: */
1021 	u64				acct_rss_mem1;
1022 	/* Accumulated virtual memory usage: */
1023 	u64				acct_vm_mem1;
1024 	/* stime + utime since last update: */
1025 	u64				acct_timexpd;
1026 #endif
1027 #ifdef CONFIG_CPUSETS
1028 	/* Protected by ->alloc_lock: */
1029 	nodemask_t			mems_allowed;
1030 	/* Seqence number to catch updates: */
1031 	seqcount_t			mems_allowed_seq;
1032 	int				cpuset_mem_spread_rotor;
1033 	int				cpuset_slab_spread_rotor;
1034 #endif
1035 #ifdef CONFIG_CGROUPS
1036 	/* Control Group info protected by css_set_lock: */
1037 	struct css_set __rcu		*cgroups;
1038 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1039 	struct list_head		cg_list;
1040 #endif
1041 #ifdef CONFIG_X86_CPU_RESCTRL
1042 	u32				closid;
1043 	u32				rmid;
1044 #endif
1045 #ifdef CONFIG_FUTEX
1046 	struct robust_list_head __user	*robust_list;
1047 #ifdef CONFIG_COMPAT
1048 	struct compat_robust_list_head __user *compat_robust_list;
1049 #endif
1050 	struct list_head		pi_state_list;
1051 	struct futex_pi_state		*pi_state_cache;
1052 #endif
1053 #ifdef CONFIG_PERF_EVENTS
1054 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1055 	struct mutex			perf_event_mutex;
1056 	struct list_head		perf_event_list;
1057 #endif
1058 #ifdef CONFIG_DEBUG_PREEMPT
1059 	unsigned long			preempt_disable_ip;
1060 #endif
1061 #ifdef CONFIG_NUMA
1062 	/* Protected by alloc_lock: */
1063 	struct mempolicy		*mempolicy;
1064 	short				il_prev;
1065 	short				pref_node_fork;
1066 #endif
1067 #ifdef CONFIG_NUMA_BALANCING
1068 	int				numa_scan_seq;
1069 	unsigned int			numa_scan_period;
1070 	unsigned int			numa_scan_period_max;
1071 	int				numa_preferred_nid;
1072 	unsigned long			numa_migrate_retry;
1073 	/* Migration stamp: */
1074 	u64				node_stamp;
1075 	u64				last_task_numa_placement;
1076 	u64				last_sum_exec_runtime;
1077 	struct callback_head		numa_work;
1078 
1079 	/*
1080 	 * This pointer is only modified for current in syscall and
1081 	 * pagefault context (and for tasks being destroyed), so it can be read
1082 	 * from any of the following contexts:
1083 	 *  - RCU read-side critical section
1084 	 *  - current->numa_group from everywhere
1085 	 *  - task's runqueue locked, task not running
1086 	 */
1087 	struct numa_group __rcu		*numa_group;
1088 
1089 	/*
1090 	 * numa_faults is an array split into four regions:
1091 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1092 	 * in this precise order.
1093 	 *
1094 	 * faults_memory: Exponential decaying average of faults on a per-node
1095 	 * basis. Scheduling placement decisions are made based on these
1096 	 * counts. The values remain static for the duration of a PTE scan.
1097 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1098 	 * hinting fault was incurred.
1099 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1100 	 * during the current scan window. When the scan completes, the counts
1101 	 * in faults_memory and faults_cpu decay and these values are copied.
1102 	 */
1103 	unsigned long			*numa_faults;
1104 	unsigned long			total_numa_faults;
1105 
1106 	/*
1107 	 * numa_faults_locality tracks if faults recorded during the last
1108 	 * scan window were remote/local or failed to migrate. The task scan
1109 	 * period is adapted based on the locality of the faults with different
1110 	 * weights depending on whether they were shared or private faults
1111 	 */
1112 	unsigned long			numa_faults_locality[3];
1113 
1114 	unsigned long			numa_pages_migrated;
1115 #endif /* CONFIG_NUMA_BALANCING */
1116 
1117 #ifdef CONFIG_RSEQ
1118 	struct rseq __user *rseq;
1119 	u32 rseq_sig;
1120 	/*
1121 	 * RmW on rseq_event_mask must be performed atomically
1122 	 * with respect to preemption.
1123 	 */
1124 	unsigned long rseq_event_mask;
1125 #endif
1126 
1127 	struct tlbflush_unmap_batch	tlb_ubc;
1128 
1129 	struct rcu_head			rcu;
1130 
1131 	/* Cache last used pipe for splice(): */
1132 	struct pipe_inode_info		*splice_pipe;
1133 
1134 	struct page_frag		task_frag;
1135 
1136 #ifdef CONFIG_TASK_DELAY_ACCT
1137 	struct task_delay_info		*delays;
1138 #endif
1139 
1140 #ifdef CONFIG_FAULT_INJECTION
1141 	int				make_it_fail;
1142 	unsigned int			fail_nth;
1143 #endif
1144 	/*
1145 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1146 	 * balance_dirty_pages() for a dirty throttling pause:
1147 	 */
1148 	int				nr_dirtied;
1149 	int				nr_dirtied_pause;
1150 	/* Start of a write-and-pause period: */
1151 	unsigned long			dirty_paused_when;
1152 
1153 #ifdef CONFIG_LATENCYTOP
1154 	int				latency_record_count;
1155 	struct latency_record		latency_record[LT_SAVECOUNT];
1156 #endif
1157 	/*
1158 	 * Time slack values; these are used to round up poll() and
1159 	 * select() etc timeout values. These are in nanoseconds.
1160 	 */
1161 	u64				timer_slack_ns;
1162 	u64				default_timer_slack_ns;
1163 
1164 #ifdef CONFIG_KASAN
1165 	unsigned int			kasan_depth;
1166 #endif
1167 
1168 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1169 	/* Index of current stored address in ret_stack: */
1170 	int				curr_ret_stack;
1171 	int				curr_ret_depth;
1172 
1173 	/* Stack of return addresses for return function tracing: */
1174 	struct ftrace_ret_stack		*ret_stack;
1175 
1176 	/* Timestamp for last schedule: */
1177 	unsigned long long		ftrace_timestamp;
1178 
1179 	/*
1180 	 * Number of functions that haven't been traced
1181 	 * because of depth overrun:
1182 	 */
1183 	atomic_t			trace_overrun;
1184 
1185 	/* Pause tracing: */
1186 	atomic_t			tracing_graph_pause;
1187 #endif
1188 
1189 #ifdef CONFIG_TRACING
1190 	/* State flags for use by tracers: */
1191 	unsigned long			trace;
1192 
1193 	/* Bitmask and counter of trace recursion: */
1194 	unsigned long			trace_recursion;
1195 #endif /* CONFIG_TRACING */
1196 
1197 #ifdef CONFIG_KCOV
1198 	/* Coverage collection mode enabled for this task (0 if disabled): */
1199 	unsigned int			kcov_mode;
1200 
1201 	/* Size of the kcov_area: */
1202 	unsigned int			kcov_size;
1203 
1204 	/* Buffer for coverage collection: */
1205 	void				*kcov_area;
1206 
1207 	/* KCOV descriptor wired with this task or NULL: */
1208 	struct kcov			*kcov;
1209 #endif
1210 
1211 #ifdef CONFIG_MEMCG
1212 	struct mem_cgroup		*memcg_in_oom;
1213 	gfp_t				memcg_oom_gfp_mask;
1214 	int				memcg_oom_order;
1215 
1216 	/* Number of pages to reclaim on returning to userland: */
1217 	unsigned int			memcg_nr_pages_over_high;
1218 
1219 	/* Used by memcontrol for targeted memcg charge: */
1220 	struct mem_cgroup		*active_memcg;
1221 #endif
1222 
1223 #ifdef CONFIG_BLK_CGROUP
1224 	struct request_queue		*throttle_queue;
1225 #endif
1226 
1227 #ifdef CONFIG_UPROBES
1228 	struct uprobe_task		*utask;
1229 #endif
1230 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1231 	unsigned int			sequential_io;
1232 	unsigned int			sequential_io_avg;
1233 #endif
1234 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1235 	unsigned long			task_state_change;
1236 #endif
1237 	int				pagefault_disabled;
1238 #ifdef CONFIG_MMU
1239 	struct task_struct		*oom_reaper_list;
1240 #endif
1241 #ifdef CONFIG_VMAP_STACK
1242 	struct vm_struct		*stack_vm_area;
1243 #endif
1244 #ifdef CONFIG_THREAD_INFO_IN_TASK
1245 	/* A live task holds one reference: */
1246 	refcount_t			stack_refcount;
1247 #endif
1248 #ifdef CONFIG_LIVEPATCH
1249 	int patch_state;
1250 #endif
1251 #ifdef CONFIG_SECURITY
1252 	/* Used by LSM modules for access restriction: */
1253 	void				*security;
1254 #endif
1255 
1256 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1257 	unsigned long			lowest_stack;
1258 	unsigned long			prev_lowest_stack;
1259 #endif
1260 
1261 	/*
1262 	 * New fields for task_struct should be added above here, so that
1263 	 * they are included in the randomized portion of task_struct.
1264 	 */
1265 	randomized_struct_fields_end
1266 
1267 	/* CPU-specific state of this task: */
1268 	struct thread_struct		thread;
1269 
1270 	/*
1271 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1272 	 * structure.  It *MUST* be at the end of 'task_struct'.
1273 	 *
1274 	 * Do not put anything below here!
1275 	 */
1276 };
1277 
1278 static inline struct pid *task_pid(struct task_struct *task)
1279 {
1280 	return task->thread_pid;
1281 }
1282 
1283 /*
1284  * the helpers to get the task's different pids as they are seen
1285  * from various namespaces
1286  *
1287  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1288  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1289  *                     current.
1290  * task_xid_nr_ns()  : id seen from the ns specified;
1291  *
1292  * see also pid_nr() etc in include/linux/pid.h
1293  */
1294 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1295 
1296 static inline pid_t task_pid_nr(struct task_struct *tsk)
1297 {
1298 	return tsk->pid;
1299 }
1300 
1301 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1302 {
1303 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1304 }
1305 
1306 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1307 {
1308 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1309 }
1310 
1311 
1312 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1313 {
1314 	return tsk->tgid;
1315 }
1316 
1317 /**
1318  * pid_alive - check that a task structure is not stale
1319  * @p: Task structure to be checked.
1320  *
1321  * Test if a process is not yet dead (at most zombie state)
1322  * If pid_alive fails, then pointers within the task structure
1323  * can be stale and must not be dereferenced.
1324  *
1325  * Return: 1 if the process is alive. 0 otherwise.
1326  */
1327 static inline int pid_alive(const struct task_struct *p)
1328 {
1329 	return p->thread_pid != NULL;
1330 }
1331 
1332 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1333 {
1334 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1335 }
1336 
1337 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1338 {
1339 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1340 }
1341 
1342 
1343 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1344 {
1345 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1346 }
1347 
1348 static inline pid_t task_session_vnr(struct task_struct *tsk)
1349 {
1350 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1351 }
1352 
1353 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1354 {
1355 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1356 }
1357 
1358 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1359 {
1360 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1361 }
1362 
1363 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1364 {
1365 	pid_t pid = 0;
1366 
1367 	rcu_read_lock();
1368 	if (pid_alive(tsk))
1369 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1370 	rcu_read_unlock();
1371 
1372 	return pid;
1373 }
1374 
1375 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1376 {
1377 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1378 }
1379 
1380 /* Obsolete, do not use: */
1381 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1382 {
1383 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1384 }
1385 
1386 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1387 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1388 
1389 static inline unsigned int task_state_index(struct task_struct *tsk)
1390 {
1391 	unsigned int tsk_state = READ_ONCE(tsk->state);
1392 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1393 
1394 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1395 
1396 	if (tsk_state == TASK_IDLE)
1397 		state = TASK_REPORT_IDLE;
1398 
1399 	return fls(state);
1400 }
1401 
1402 static inline char task_index_to_char(unsigned int state)
1403 {
1404 	static const char state_char[] = "RSDTtXZPI";
1405 
1406 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1407 
1408 	return state_char[state];
1409 }
1410 
1411 static inline char task_state_to_char(struct task_struct *tsk)
1412 {
1413 	return task_index_to_char(task_state_index(tsk));
1414 }
1415 
1416 /**
1417  * is_global_init - check if a task structure is init. Since init
1418  * is free to have sub-threads we need to check tgid.
1419  * @tsk: Task structure to be checked.
1420  *
1421  * Check if a task structure is the first user space task the kernel created.
1422  *
1423  * Return: 1 if the task structure is init. 0 otherwise.
1424  */
1425 static inline int is_global_init(struct task_struct *tsk)
1426 {
1427 	return task_tgid_nr(tsk) == 1;
1428 }
1429 
1430 extern struct pid *cad_pid;
1431 
1432 /*
1433  * Per process flags
1434  */
1435 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1436 #define PF_EXITING		0x00000004	/* Getting shut down */
1437 #define PF_EXITPIDONE		0x00000008	/* PI exit done on shut down */
1438 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1439 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1440 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1441 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1442 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1443 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1444 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1445 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1446 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1447 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1448 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1449 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1450 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1451 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1452 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1453 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1454 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1455 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1456 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1457 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1458 #define PF_MEMSTALL		0x01000000	/* Stalled due to lack of memory */
1459 #define PF_UMH			0x02000000	/* I'm an Usermodehelper process */
1460 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1461 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1462 #define PF_MEMALLOC_NOCMA	0x10000000	/* All allocation request will have _GFP_MOVABLE cleared */
1463 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1464 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1465 
1466 /*
1467  * Only the _current_ task can read/write to tsk->flags, but other
1468  * tasks can access tsk->flags in readonly mode for example
1469  * with tsk_used_math (like during threaded core dumping).
1470  * There is however an exception to this rule during ptrace
1471  * or during fork: the ptracer task is allowed to write to the
1472  * child->flags of its traced child (same goes for fork, the parent
1473  * can write to the child->flags), because we're guaranteed the
1474  * child is not running and in turn not changing child->flags
1475  * at the same time the parent does it.
1476  */
1477 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1478 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1479 #define clear_used_math()			clear_stopped_child_used_math(current)
1480 #define set_used_math()				set_stopped_child_used_math(current)
1481 
1482 #define conditional_stopped_child_used_math(condition, child) \
1483 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1484 
1485 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1486 
1487 #define copy_to_stopped_child_used_math(child) \
1488 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1489 
1490 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1491 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1492 #define used_math()				tsk_used_math(current)
1493 
1494 static inline bool is_percpu_thread(void)
1495 {
1496 #ifdef CONFIG_SMP
1497 	return (current->flags & PF_NO_SETAFFINITY) &&
1498 		(current->nr_cpus_allowed  == 1);
1499 #else
1500 	return true;
1501 #endif
1502 }
1503 
1504 /* Per-process atomic flags. */
1505 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1506 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1507 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1508 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1509 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1510 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1511 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1512 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1513 
1514 #define TASK_PFA_TEST(name, func)					\
1515 	static inline bool task_##func(struct task_struct *p)		\
1516 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1517 
1518 #define TASK_PFA_SET(name, func)					\
1519 	static inline void task_set_##func(struct task_struct *p)	\
1520 	{ set_bit(PFA_##name, &p->atomic_flags); }
1521 
1522 #define TASK_PFA_CLEAR(name, func)					\
1523 	static inline void task_clear_##func(struct task_struct *p)	\
1524 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1525 
1526 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1527 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1528 
1529 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1530 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1531 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1532 
1533 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1534 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1535 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1536 
1537 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1538 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1539 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1540 
1541 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1542 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1543 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1544 
1545 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1546 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1547 
1548 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1549 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1550 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1551 
1552 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1553 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1554 
1555 static inline void
1556 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1557 {
1558 	current->flags &= ~flags;
1559 	current->flags |= orig_flags & flags;
1560 }
1561 
1562 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1563 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1564 #ifdef CONFIG_SMP
1565 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1566 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1567 #else
1568 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1569 {
1570 }
1571 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1572 {
1573 	if (!cpumask_test_cpu(0, new_mask))
1574 		return -EINVAL;
1575 	return 0;
1576 }
1577 #endif
1578 
1579 extern int yield_to(struct task_struct *p, bool preempt);
1580 extern void set_user_nice(struct task_struct *p, long nice);
1581 extern int task_prio(const struct task_struct *p);
1582 
1583 /**
1584  * task_nice - return the nice value of a given task.
1585  * @p: the task in question.
1586  *
1587  * Return: The nice value [ -20 ... 0 ... 19 ].
1588  */
1589 static inline int task_nice(const struct task_struct *p)
1590 {
1591 	return PRIO_TO_NICE((p)->static_prio);
1592 }
1593 
1594 extern int can_nice(const struct task_struct *p, const int nice);
1595 extern int task_curr(const struct task_struct *p);
1596 extern int idle_cpu(int cpu);
1597 extern int available_idle_cpu(int cpu);
1598 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1599 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1600 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1601 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1602 extern struct task_struct *idle_task(int cpu);
1603 
1604 /**
1605  * is_idle_task - is the specified task an idle task?
1606  * @p: the task in question.
1607  *
1608  * Return: 1 if @p is an idle task. 0 otherwise.
1609  */
1610 static inline bool is_idle_task(const struct task_struct *p)
1611 {
1612 	return !!(p->flags & PF_IDLE);
1613 }
1614 
1615 extern struct task_struct *curr_task(int cpu);
1616 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1617 
1618 void yield(void);
1619 
1620 union thread_union {
1621 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1622 	struct task_struct task;
1623 #endif
1624 #ifndef CONFIG_THREAD_INFO_IN_TASK
1625 	struct thread_info thread_info;
1626 #endif
1627 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1628 };
1629 
1630 #ifndef CONFIG_THREAD_INFO_IN_TASK
1631 extern struct thread_info init_thread_info;
1632 #endif
1633 
1634 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1635 
1636 #ifdef CONFIG_THREAD_INFO_IN_TASK
1637 static inline struct thread_info *task_thread_info(struct task_struct *task)
1638 {
1639 	return &task->thread_info;
1640 }
1641 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1642 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1643 #endif
1644 
1645 /*
1646  * find a task by one of its numerical ids
1647  *
1648  * find_task_by_pid_ns():
1649  *      finds a task by its pid in the specified namespace
1650  * find_task_by_vpid():
1651  *      finds a task by its virtual pid
1652  *
1653  * see also find_vpid() etc in include/linux/pid.h
1654  */
1655 
1656 extern struct task_struct *find_task_by_vpid(pid_t nr);
1657 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1658 
1659 /*
1660  * find a task by its virtual pid and get the task struct
1661  */
1662 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1663 
1664 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1665 extern int wake_up_process(struct task_struct *tsk);
1666 extern void wake_up_new_task(struct task_struct *tsk);
1667 
1668 #ifdef CONFIG_SMP
1669 extern void kick_process(struct task_struct *tsk);
1670 #else
1671 static inline void kick_process(struct task_struct *tsk) { }
1672 #endif
1673 
1674 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1675 
1676 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1677 {
1678 	__set_task_comm(tsk, from, false);
1679 }
1680 
1681 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1682 #define get_task_comm(buf, tsk) ({			\
1683 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1684 	__get_task_comm(buf, sizeof(buf), tsk);		\
1685 })
1686 
1687 #ifdef CONFIG_SMP
1688 void scheduler_ipi(void);
1689 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1690 #else
1691 static inline void scheduler_ipi(void) { }
1692 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1693 {
1694 	return 1;
1695 }
1696 #endif
1697 
1698 /*
1699  * Set thread flags in other task's structures.
1700  * See asm/thread_info.h for TIF_xxxx flags available:
1701  */
1702 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1703 {
1704 	set_ti_thread_flag(task_thread_info(tsk), flag);
1705 }
1706 
1707 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1708 {
1709 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1710 }
1711 
1712 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1713 					  bool value)
1714 {
1715 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1716 }
1717 
1718 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1719 {
1720 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1721 }
1722 
1723 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1724 {
1725 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1726 }
1727 
1728 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1729 {
1730 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1731 }
1732 
1733 static inline void set_tsk_need_resched(struct task_struct *tsk)
1734 {
1735 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1736 }
1737 
1738 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1739 {
1740 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1741 }
1742 
1743 static inline int test_tsk_need_resched(struct task_struct *tsk)
1744 {
1745 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1746 }
1747 
1748 /*
1749  * cond_resched() and cond_resched_lock(): latency reduction via
1750  * explicit rescheduling in places that are safe. The return
1751  * value indicates whether a reschedule was done in fact.
1752  * cond_resched_lock() will drop the spinlock before scheduling,
1753  */
1754 #ifndef CONFIG_PREEMPTION
1755 extern int _cond_resched(void);
1756 #else
1757 static inline int _cond_resched(void) { return 0; }
1758 #endif
1759 
1760 #define cond_resched() ({			\
1761 	___might_sleep(__FILE__, __LINE__, 0);	\
1762 	_cond_resched();			\
1763 })
1764 
1765 extern int __cond_resched_lock(spinlock_t *lock);
1766 
1767 #define cond_resched_lock(lock) ({				\
1768 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1769 	__cond_resched_lock(lock);				\
1770 })
1771 
1772 static inline void cond_resched_rcu(void)
1773 {
1774 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1775 	rcu_read_unlock();
1776 	cond_resched();
1777 	rcu_read_lock();
1778 #endif
1779 }
1780 
1781 /*
1782  * Does a critical section need to be broken due to another
1783  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1784  * but a general need for low latency)
1785  */
1786 static inline int spin_needbreak(spinlock_t *lock)
1787 {
1788 #ifdef CONFIG_PREEMPTION
1789 	return spin_is_contended(lock);
1790 #else
1791 	return 0;
1792 #endif
1793 }
1794 
1795 static __always_inline bool need_resched(void)
1796 {
1797 	return unlikely(tif_need_resched());
1798 }
1799 
1800 /*
1801  * Wrappers for p->thread_info->cpu access. No-op on UP.
1802  */
1803 #ifdef CONFIG_SMP
1804 
1805 static inline unsigned int task_cpu(const struct task_struct *p)
1806 {
1807 #ifdef CONFIG_THREAD_INFO_IN_TASK
1808 	return READ_ONCE(p->cpu);
1809 #else
1810 	return READ_ONCE(task_thread_info(p)->cpu);
1811 #endif
1812 }
1813 
1814 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1815 
1816 #else
1817 
1818 static inline unsigned int task_cpu(const struct task_struct *p)
1819 {
1820 	return 0;
1821 }
1822 
1823 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1824 {
1825 }
1826 
1827 #endif /* CONFIG_SMP */
1828 
1829 /*
1830  * In order to reduce various lock holder preemption latencies provide an
1831  * interface to see if a vCPU is currently running or not.
1832  *
1833  * This allows us to terminate optimistic spin loops and block, analogous to
1834  * the native optimistic spin heuristic of testing if the lock owner task is
1835  * running or not.
1836  */
1837 #ifndef vcpu_is_preempted
1838 # define vcpu_is_preempted(cpu)	false
1839 #endif
1840 
1841 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1842 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1843 
1844 #ifndef TASK_SIZE_OF
1845 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1846 #endif
1847 
1848 #ifdef CONFIG_RSEQ
1849 
1850 /*
1851  * Map the event mask on the user-space ABI enum rseq_cs_flags
1852  * for direct mask checks.
1853  */
1854 enum rseq_event_mask_bits {
1855 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1856 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1857 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1858 };
1859 
1860 enum rseq_event_mask {
1861 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
1862 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
1863 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
1864 };
1865 
1866 static inline void rseq_set_notify_resume(struct task_struct *t)
1867 {
1868 	if (t->rseq)
1869 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1870 }
1871 
1872 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1873 
1874 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1875 					     struct pt_regs *regs)
1876 {
1877 	if (current->rseq)
1878 		__rseq_handle_notify_resume(ksig, regs);
1879 }
1880 
1881 static inline void rseq_signal_deliver(struct ksignal *ksig,
1882 				       struct pt_regs *regs)
1883 {
1884 	preempt_disable();
1885 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1886 	preempt_enable();
1887 	rseq_handle_notify_resume(ksig, regs);
1888 }
1889 
1890 /* rseq_preempt() requires preemption to be disabled. */
1891 static inline void rseq_preempt(struct task_struct *t)
1892 {
1893 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1894 	rseq_set_notify_resume(t);
1895 }
1896 
1897 /* rseq_migrate() requires preemption to be disabled. */
1898 static inline void rseq_migrate(struct task_struct *t)
1899 {
1900 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1901 	rseq_set_notify_resume(t);
1902 }
1903 
1904 /*
1905  * If parent process has a registered restartable sequences area, the
1906  * child inherits. Only applies when forking a process, not a thread.
1907  */
1908 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1909 {
1910 	if (clone_flags & CLONE_THREAD) {
1911 		t->rseq = NULL;
1912 		t->rseq_sig = 0;
1913 		t->rseq_event_mask = 0;
1914 	} else {
1915 		t->rseq = current->rseq;
1916 		t->rseq_sig = current->rseq_sig;
1917 		t->rseq_event_mask = current->rseq_event_mask;
1918 	}
1919 }
1920 
1921 static inline void rseq_execve(struct task_struct *t)
1922 {
1923 	t->rseq = NULL;
1924 	t->rseq_sig = 0;
1925 	t->rseq_event_mask = 0;
1926 }
1927 
1928 #else
1929 
1930 static inline void rseq_set_notify_resume(struct task_struct *t)
1931 {
1932 }
1933 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1934 					     struct pt_regs *regs)
1935 {
1936 }
1937 static inline void rseq_signal_deliver(struct ksignal *ksig,
1938 				       struct pt_regs *regs)
1939 {
1940 }
1941 static inline void rseq_preempt(struct task_struct *t)
1942 {
1943 }
1944 static inline void rseq_migrate(struct task_struct *t)
1945 {
1946 }
1947 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1948 {
1949 }
1950 static inline void rseq_execve(struct task_struct *t)
1951 {
1952 }
1953 
1954 #endif
1955 
1956 void __exit_umh(struct task_struct *tsk);
1957 
1958 static inline void exit_umh(struct task_struct *tsk)
1959 {
1960 	if (unlikely(tsk->flags & PF_UMH))
1961 		__exit_umh(tsk);
1962 }
1963 
1964 #ifdef CONFIG_DEBUG_RSEQ
1965 
1966 void rseq_syscall(struct pt_regs *regs);
1967 
1968 #else
1969 
1970 static inline void rseq_syscall(struct pt_regs *regs)
1971 {
1972 }
1973 
1974 #endif
1975 
1976 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
1977 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
1978 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
1979 
1980 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
1981 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
1982 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
1983 
1984 int sched_trace_rq_cpu(struct rq *rq);
1985 
1986 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
1987 
1988 #endif
1989