xref: /linux/include/linux/sched.h (revision 93431e0607e58a3c997a134adc0fad4fdc147dab)
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 asmlinkage void preempt_schedule_irq(void);
227 
228 extern int __must_check io_schedule_prepare(void);
229 extern void io_schedule_finish(int token);
230 extern long io_schedule_timeout(long timeout);
231 extern void io_schedule(void);
232 
233 /**
234  * struct prev_cputime - snapshot of system and user cputime
235  * @utime: time spent in user mode
236  * @stime: time spent in system mode
237  * @lock: protects the above two fields
238  *
239  * Stores previous user/system time values such that we can guarantee
240  * monotonicity.
241  */
242 struct prev_cputime {
243 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
244 	u64				utime;
245 	u64				stime;
246 	raw_spinlock_t			lock;
247 #endif
248 };
249 
250 enum vtime_state {
251 	/* Task is sleeping or running in a CPU with VTIME inactive: */
252 	VTIME_INACTIVE = 0,
253 	/* Task is idle */
254 	VTIME_IDLE,
255 	/* Task runs in kernelspace in a CPU with VTIME active: */
256 	VTIME_SYS,
257 	/* Task runs in userspace in a CPU with VTIME active: */
258 	VTIME_USER,
259 	/* Task runs as guests in a CPU with VTIME active: */
260 	VTIME_GUEST,
261 };
262 
263 struct vtime {
264 	seqcount_t		seqcount;
265 	unsigned long long	starttime;
266 	enum vtime_state	state;
267 	unsigned int		cpu;
268 	u64			utime;
269 	u64			stime;
270 	u64			gtime;
271 };
272 
273 /*
274  * Utilization clamp constraints.
275  * @UCLAMP_MIN:	Minimum utilization
276  * @UCLAMP_MAX:	Maximum utilization
277  * @UCLAMP_CNT:	Utilization clamp constraints count
278  */
279 enum uclamp_id {
280 	UCLAMP_MIN = 0,
281 	UCLAMP_MAX,
282 	UCLAMP_CNT
283 };
284 
285 #ifdef CONFIG_SMP
286 extern struct root_domain def_root_domain;
287 extern struct mutex sched_domains_mutex;
288 #endif
289 
290 struct sched_info {
291 #ifdef CONFIG_SCHED_INFO
292 	/* Cumulative counters: */
293 
294 	/* # of times we have run on this CPU: */
295 	unsigned long			pcount;
296 
297 	/* Time spent waiting on a runqueue: */
298 	unsigned long long		run_delay;
299 
300 	/* Timestamps: */
301 
302 	/* When did we last run on a CPU? */
303 	unsigned long long		last_arrival;
304 
305 	/* When were we last queued to run? */
306 	unsigned long long		last_queued;
307 
308 #endif /* CONFIG_SCHED_INFO */
309 };
310 
311 /*
312  * Integer metrics need fixed point arithmetic, e.g., sched/fair
313  * has a few: load, load_avg, util_avg, freq, and capacity.
314  *
315  * We define a basic fixed point arithmetic range, and then formalize
316  * all these metrics based on that basic range.
317  */
318 # define SCHED_FIXEDPOINT_SHIFT		10
319 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
320 
321 /* Increase resolution of cpu_capacity calculations */
322 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
323 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
324 
325 struct load_weight {
326 	unsigned long			weight;
327 	u32				inv_weight;
328 };
329 
330 /**
331  * struct util_est - Estimation utilization of FAIR tasks
332  * @enqueued: instantaneous estimated utilization of a task/cpu
333  * @ewma:     the Exponential Weighted Moving Average (EWMA)
334  *            utilization of a task
335  *
336  * Support data structure to track an Exponential Weighted Moving Average
337  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
338  * average each time a task completes an activation. Sample's weight is chosen
339  * so that the EWMA will be relatively insensitive to transient changes to the
340  * task's workload.
341  *
342  * The enqueued attribute has a slightly different meaning for tasks and cpus:
343  * - task:   the task's util_avg at last task dequeue time
344  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
345  * Thus, the util_est.enqueued of a task represents the contribution on the
346  * estimated utilization of the CPU where that task is currently enqueued.
347  *
348  * Only for tasks we track a moving average of the past instantaneous
349  * estimated utilization. This allows to absorb sporadic drops in utilization
350  * of an otherwise almost periodic task.
351  */
352 struct util_est {
353 	unsigned int			enqueued;
354 	unsigned int			ewma;
355 #define UTIL_EST_WEIGHT_SHIFT		2
356 } __attribute__((__aligned__(sizeof(u64))));
357 
358 /*
359  * The load/runnable/util_avg accumulates an infinite geometric series
360  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
361  *
362  * [load_avg definition]
363  *
364  *   load_avg = runnable% * scale_load_down(load)
365  *
366  * [runnable_avg definition]
367  *
368  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
369  *
370  * [util_avg definition]
371  *
372  *   util_avg = running% * SCHED_CAPACITY_SCALE
373  *
374  * where runnable% is the time ratio that a sched_entity is runnable and
375  * running% the time ratio that a sched_entity is running.
376  *
377  * For cfs_rq, they are the aggregated values of all runnable and blocked
378  * sched_entities.
379  *
380  * The load/runnable/util_avg doesn't direcly factor frequency scaling and CPU
381  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
382  * for computing those signals (see update_rq_clock_pelt())
383  *
384  * N.B., the above ratios (runnable% and running%) themselves are in the
385  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
386  * to as large a range as necessary. This is for example reflected by
387  * util_avg's SCHED_CAPACITY_SCALE.
388  *
389  * [Overflow issue]
390  *
391  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
392  * with the highest load (=88761), always runnable on a single cfs_rq,
393  * and should not overflow as the number already hits PID_MAX_LIMIT.
394  *
395  * For all other cases (including 32-bit kernels), struct load_weight's
396  * weight will overflow first before we do, because:
397  *
398  *    Max(load_avg) <= Max(load.weight)
399  *
400  * Then it is the load_weight's responsibility to consider overflow
401  * issues.
402  */
403 struct sched_avg {
404 	u64				last_update_time;
405 	u64				load_sum;
406 	u64				runnable_sum;
407 	u32				util_sum;
408 	u32				period_contrib;
409 	unsigned long			load_avg;
410 	unsigned long			runnable_avg;
411 	unsigned long			util_avg;
412 	struct util_est			util_est;
413 } ____cacheline_aligned;
414 
415 struct sched_statistics {
416 #ifdef CONFIG_SCHEDSTATS
417 	u64				wait_start;
418 	u64				wait_max;
419 	u64				wait_count;
420 	u64				wait_sum;
421 	u64				iowait_count;
422 	u64				iowait_sum;
423 
424 	u64				sleep_start;
425 	u64				sleep_max;
426 	s64				sum_sleep_runtime;
427 
428 	u64				block_start;
429 	u64				block_max;
430 	u64				exec_max;
431 	u64				slice_max;
432 
433 	u64				nr_migrations_cold;
434 	u64				nr_failed_migrations_affine;
435 	u64				nr_failed_migrations_running;
436 	u64				nr_failed_migrations_hot;
437 	u64				nr_forced_migrations;
438 
439 	u64				nr_wakeups;
440 	u64				nr_wakeups_sync;
441 	u64				nr_wakeups_migrate;
442 	u64				nr_wakeups_local;
443 	u64				nr_wakeups_remote;
444 	u64				nr_wakeups_affine;
445 	u64				nr_wakeups_affine_attempts;
446 	u64				nr_wakeups_passive;
447 	u64				nr_wakeups_idle;
448 #endif
449 };
450 
451 struct sched_entity {
452 	/* For load-balancing: */
453 	struct load_weight		load;
454 	struct rb_node			run_node;
455 	struct list_head		group_node;
456 	unsigned int			on_rq;
457 
458 	u64				exec_start;
459 	u64				sum_exec_runtime;
460 	u64				vruntime;
461 	u64				prev_sum_exec_runtime;
462 
463 	u64				nr_migrations;
464 
465 	struct sched_statistics		statistics;
466 
467 #ifdef CONFIG_FAIR_GROUP_SCHED
468 	int				depth;
469 	struct sched_entity		*parent;
470 	/* rq on which this entity is (to be) queued: */
471 	struct cfs_rq			*cfs_rq;
472 	/* rq "owned" by this entity/group: */
473 	struct cfs_rq			*my_q;
474 	/* cached value of my_q->h_nr_running */
475 	unsigned long			runnable_weight;
476 #endif
477 
478 #ifdef CONFIG_SMP
479 	/*
480 	 * Per entity load average tracking.
481 	 *
482 	 * Put into separate cache line so it does not
483 	 * collide with read-mostly values above.
484 	 */
485 	struct sched_avg		avg;
486 #endif
487 };
488 
489 struct sched_rt_entity {
490 	struct list_head		run_list;
491 	unsigned long			timeout;
492 	unsigned long			watchdog_stamp;
493 	unsigned int			time_slice;
494 	unsigned short			on_rq;
495 	unsigned short			on_list;
496 
497 	struct sched_rt_entity		*back;
498 #ifdef CONFIG_RT_GROUP_SCHED
499 	struct sched_rt_entity		*parent;
500 	/* rq on which this entity is (to be) queued: */
501 	struct rt_rq			*rt_rq;
502 	/* rq "owned" by this entity/group: */
503 	struct rt_rq			*my_q;
504 #endif
505 } __randomize_layout;
506 
507 struct sched_dl_entity {
508 	struct rb_node			rb_node;
509 
510 	/*
511 	 * Original scheduling parameters. Copied here from sched_attr
512 	 * during sched_setattr(), they will remain the same until
513 	 * the next sched_setattr().
514 	 */
515 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
516 	u64				dl_deadline;	/* Relative deadline of each instance	*/
517 	u64				dl_period;	/* Separation of two instances (period) */
518 	u64				dl_bw;		/* dl_runtime / dl_period		*/
519 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
520 
521 	/*
522 	 * Actual scheduling parameters. Initialized with the values above,
523 	 * they are continuously updated during task execution. Note that
524 	 * the remaining runtime could be < 0 in case we are in overrun.
525 	 */
526 	s64				runtime;	/* Remaining runtime for this instance	*/
527 	u64				deadline;	/* Absolute deadline for this instance	*/
528 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
529 
530 	/*
531 	 * Some bool flags:
532 	 *
533 	 * @dl_throttled tells if we exhausted the runtime. If so, the
534 	 * task has to wait for a replenishment to be performed at the
535 	 * next firing of dl_timer.
536 	 *
537 	 * @dl_boosted tells if we are boosted due to DI. If so we are
538 	 * outside bandwidth enforcement mechanism (but only until we
539 	 * exit the critical section);
540 	 *
541 	 * @dl_yielded tells if task gave up the CPU before consuming
542 	 * all its available runtime during the last job.
543 	 *
544 	 * @dl_non_contending tells if the task is inactive while still
545 	 * contributing to the active utilization. In other words, it
546 	 * indicates if the inactive timer has been armed and its handler
547 	 * has not been executed yet. This flag is useful to avoid race
548 	 * conditions between the inactive timer handler and the wakeup
549 	 * code.
550 	 *
551 	 * @dl_overrun tells if the task asked to be informed about runtime
552 	 * overruns.
553 	 */
554 	unsigned int			dl_throttled      : 1;
555 	unsigned int			dl_boosted        : 1;
556 	unsigned int			dl_yielded        : 1;
557 	unsigned int			dl_non_contending : 1;
558 	unsigned int			dl_overrun	  : 1;
559 
560 	/*
561 	 * Bandwidth enforcement timer. Each -deadline task has its
562 	 * own bandwidth to be enforced, thus we need one timer per task.
563 	 */
564 	struct hrtimer			dl_timer;
565 
566 	/*
567 	 * Inactive timer, responsible for decreasing the active utilization
568 	 * at the "0-lag time". When a -deadline task blocks, it contributes
569 	 * to GRUB's active utilization until the "0-lag time", hence a
570 	 * timer is needed to decrease the active utilization at the correct
571 	 * time.
572 	 */
573 	struct hrtimer inactive_timer;
574 };
575 
576 #ifdef CONFIG_UCLAMP_TASK
577 /* Number of utilization clamp buckets (shorter alias) */
578 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
579 
580 /*
581  * Utilization clamp for a scheduling entity
582  * @value:		clamp value "assigned" to a se
583  * @bucket_id:		bucket index corresponding to the "assigned" value
584  * @active:		the se is currently refcounted in a rq's bucket
585  * @user_defined:	the requested clamp value comes from user-space
586  *
587  * The bucket_id is the index of the clamp bucket matching the clamp value
588  * which is pre-computed and stored to avoid expensive integer divisions from
589  * the fast path.
590  *
591  * The active bit is set whenever a task has got an "effective" value assigned,
592  * which can be different from the clamp value "requested" from user-space.
593  * This allows to know a task is refcounted in the rq's bucket corresponding
594  * to the "effective" bucket_id.
595  *
596  * The user_defined bit is set whenever a task has got a task-specific clamp
597  * value requested from userspace, i.e. the system defaults apply to this task
598  * just as a restriction. This allows to relax default clamps when a less
599  * restrictive task-specific value has been requested, thus allowing to
600  * implement a "nice" semantic. For example, a task running with a 20%
601  * default boost can still drop its own boosting to 0%.
602  */
603 struct uclamp_se {
604 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
605 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
606 	unsigned int active		: 1;
607 	unsigned int user_defined	: 1;
608 };
609 #endif /* CONFIG_UCLAMP_TASK */
610 
611 union rcu_special {
612 	struct {
613 		u8			blocked;
614 		u8			need_qs;
615 		u8			exp_hint; /* Hint for performance. */
616 		u8			deferred_qs;
617 	} b; /* Bits. */
618 	u32 s; /* Set of bits. */
619 };
620 
621 enum perf_event_task_context {
622 	perf_invalid_context = -1,
623 	perf_hw_context = 0,
624 	perf_sw_context,
625 	perf_nr_task_contexts,
626 };
627 
628 struct wake_q_node {
629 	struct wake_q_node *next;
630 };
631 
632 struct task_struct {
633 #ifdef CONFIG_THREAD_INFO_IN_TASK
634 	/*
635 	 * For reasons of header soup (see current_thread_info()), this
636 	 * must be the first element of task_struct.
637 	 */
638 	struct thread_info		thread_info;
639 #endif
640 	/* -1 unrunnable, 0 runnable, >0 stopped: */
641 	volatile long			state;
642 
643 	/*
644 	 * This begins the randomizable portion of task_struct. Only
645 	 * scheduling-critical items should be added above here.
646 	 */
647 	randomized_struct_fields_start
648 
649 	void				*stack;
650 	refcount_t			usage;
651 	/* Per task flags (PF_*), defined further below: */
652 	unsigned int			flags;
653 	unsigned int			ptrace;
654 
655 #ifdef CONFIG_SMP
656 	struct llist_node		wake_entry;
657 	int				on_cpu;
658 #ifdef CONFIG_THREAD_INFO_IN_TASK
659 	/* Current CPU: */
660 	unsigned int			cpu;
661 #endif
662 	unsigned int			wakee_flips;
663 	unsigned long			wakee_flip_decay_ts;
664 	struct task_struct		*last_wakee;
665 
666 	/*
667 	 * recent_used_cpu is initially set as the last CPU used by a task
668 	 * that wakes affine another task. Waker/wakee relationships can
669 	 * push tasks around a CPU where each wakeup moves to the next one.
670 	 * Tracking a recently used CPU allows a quick search for a recently
671 	 * used CPU that may be idle.
672 	 */
673 	int				recent_used_cpu;
674 	int				wake_cpu;
675 #endif
676 	int				on_rq;
677 
678 	int				prio;
679 	int				static_prio;
680 	int				normal_prio;
681 	unsigned int			rt_priority;
682 
683 	const struct sched_class	*sched_class;
684 	struct sched_entity		se;
685 	struct sched_rt_entity		rt;
686 #ifdef CONFIG_CGROUP_SCHED
687 	struct task_group		*sched_task_group;
688 #endif
689 	struct sched_dl_entity		dl;
690 
691 #ifdef CONFIG_UCLAMP_TASK
692 	/* Clamp values requested for a scheduling entity */
693 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
694 	/* Effective clamp values used for a scheduling entity */
695 	struct uclamp_se		uclamp[UCLAMP_CNT];
696 #endif
697 
698 #ifdef CONFIG_PREEMPT_NOTIFIERS
699 	/* List of struct preempt_notifier: */
700 	struct hlist_head		preempt_notifiers;
701 #endif
702 
703 #ifdef CONFIG_BLK_DEV_IO_TRACE
704 	unsigned int			btrace_seq;
705 #endif
706 
707 	unsigned int			policy;
708 	int				nr_cpus_allowed;
709 	const cpumask_t			*cpus_ptr;
710 	cpumask_t			cpus_mask;
711 
712 #ifdef CONFIG_PREEMPT_RCU
713 	int				rcu_read_lock_nesting;
714 	union rcu_special		rcu_read_unlock_special;
715 	struct list_head		rcu_node_entry;
716 	struct rcu_node			*rcu_blocked_node;
717 #endif /* #ifdef CONFIG_PREEMPT_RCU */
718 
719 #ifdef CONFIG_TASKS_RCU
720 	unsigned long			rcu_tasks_nvcsw;
721 	u8				rcu_tasks_holdout;
722 	u8				rcu_tasks_idx;
723 	int				rcu_tasks_idle_cpu;
724 	struct list_head		rcu_tasks_holdout_list;
725 #endif /* #ifdef CONFIG_TASKS_RCU */
726 
727 	struct sched_info		sched_info;
728 
729 	struct list_head		tasks;
730 #ifdef CONFIG_SMP
731 	struct plist_node		pushable_tasks;
732 	struct rb_node			pushable_dl_tasks;
733 #endif
734 
735 	struct mm_struct		*mm;
736 	struct mm_struct		*active_mm;
737 
738 	/* Per-thread vma caching: */
739 	struct vmacache			vmacache;
740 
741 #ifdef SPLIT_RSS_COUNTING
742 	struct task_rss_stat		rss_stat;
743 #endif
744 	int				exit_state;
745 	int				exit_code;
746 	int				exit_signal;
747 	/* The signal sent when the parent dies: */
748 	int				pdeath_signal;
749 	/* JOBCTL_*, siglock protected: */
750 	unsigned long			jobctl;
751 
752 	/* Used for emulating ABI behavior of previous Linux versions: */
753 	unsigned int			personality;
754 
755 	/* Scheduler bits, serialized by scheduler locks: */
756 	unsigned			sched_reset_on_fork:1;
757 	unsigned			sched_contributes_to_load:1;
758 	unsigned			sched_migrated:1;
759 	unsigned			sched_remote_wakeup:1;
760 #ifdef CONFIG_PSI
761 	unsigned			sched_psi_wake_requeue:1;
762 #endif
763 
764 	/* Force alignment to the next boundary: */
765 	unsigned			:0;
766 
767 	/* Unserialized, strictly 'current' */
768 
769 	/* Bit to tell LSMs we're in execve(): */
770 	unsigned			in_execve:1;
771 	unsigned			in_iowait:1;
772 #ifndef TIF_RESTORE_SIGMASK
773 	unsigned			restore_sigmask:1;
774 #endif
775 #ifdef CONFIG_MEMCG
776 	unsigned			in_user_fault:1;
777 #endif
778 #ifdef CONFIG_COMPAT_BRK
779 	unsigned			brk_randomized:1;
780 #endif
781 #ifdef CONFIG_CGROUPS
782 	/* disallow userland-initiated cgroup migration */
783 	unsigned			no_cgroup_migration:1;
784 	/* task is frozen/stopped (used by the cgroup freezer) */
785 	unsigned			frozen:1;
786 #endif
787 #ifdef CONFIG_BLK_CGROUP
788 	unsigned			use_memdelay:1;
789 #endif
790 #ifdef CONFIG_PSI
791 	/* Stalled due to lack of memory */
792 	unsigned			in_memstall:1;
793 #endif
794 
795 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
796 
797 	struct restart_block		restart_block;
798 
799 	pid_t				pid;
800 	pid_t				tgid;
801 
802 #ifdef CONFIG_STACKPROTECTOR
803 	/* Canary value for the -fstack-protector GCC feature: */
804 	unsigned long			stack_canary;
805 #endif
806 	/*
807 	 * Pointers to the (original) parent process, youngest child, younger sibling,
808 	 * older sibling, respectively.  (p->father can be replaced with
809 	 * p->real_parent->pid)
810 	 */
811 
812 	/* Real parent process: */
813 	struct task_struct __rcu	*real_parent;
814 
815 	/* Recipient of SIGCHLD, wait4() reports: */
816 	struct task_struct __rcu	*parent;
817 
818 	/*
819 	 * Children/sibling form the list of natural children:
820 	 */
821 	struct list_head		children;
822 	struct list_head		sibling;
823 	struct task_struct		*group_leader;
824 
825 	/*
826 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
827 	 *
828 	 * This includes both natural children and PTRACE_ATTACH targets.
829 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
830 	 */
831 	struct list_head		ptraced;
832 	struct list_head		ptrace_entry;
833 
834 	/* PID/PID hash table linkage. */
835 	struct pid			*thread_pid;
836 	struct hlist_node		pid_links[PIDTYPE_MAX];
837 	struct list_head		thread_group;
838 	struct list_head		thread_node;
839 
840 	struct completion		*vfork_done;
841 
842 	/* CLONE_CHILD_SETTID: */
843 	int __user			*set_child_tid;
844 
845 	/* CLONE_CHILD_CLEARTID: */
846 	int __user			*clear_child_tid;
847 
848 	u64				utime;
849 	u64				stime;
850 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
851 	u64				utimescaled;
852 	u64				stimescaled;
853 #endif
854 	u64				gtime;
855 	struct prev_cputime		prev_cputime;
856 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
857 	struct vtime			vtime;
858 #endif
859 
860 #ifdef CONFIG_NO_HZ_FULL
861 	atomic_t			tick_dep_mask;
862 #endif
863 	/* Context switch counts: */
864 	unsigned long			nvcsw;
865 	unsigned long			nivcsw;
866 
867 	/* Monotonic time in nsecs: */
868 	u64				start_time;
869 
870 	/* Boot based time in nsecs: */
871 	u64				start_boottime;
872 
873 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
874 	unsigned long			min_flt;
875 	unsigned long			maj_flt;
876 
877 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
878 	struct posix_cputimers		posix_cputimers;
879 
880 	/* Process credentials: */
881 
882 	/* Tracer's credentials at attach: */
883 	const struct cred __rcu		*ptracer_cred;
884 
885 	/* Objective and real subjective task credentials (COW): */
886 	const struct cred __rcu		*real_cred;
887 
888 	/* Effective (overridable) subjective task credentials (COW): */
889 	const struct cred __rcu		*cred;
890 
891 #ifdef CONFIG_KEYS
892 	/* Cached requested key. */
893 	struct key			*cached_requested_key;
894 #endif
895 
896 	/*
897 	 * executable name, excluding path.
898 	 *
899 	 * - normally initialized setup_new_exec()
900 	 * - access it with [gs]et_task_comm()
901 	 * - lock it with task_lock()
902 	 */
903 	char				comm[TASK_COMM_LEN];
904 
905 	struct nameidata		*nameidata;
906 
907 #ifdef CONFIG_SYSVIPC
908 	struct sysv_sem			sysvsem;
909 	struct sysv_shm			sysvshm;
910 #endif
911 #ifdef CONFIG_DETECT_HUNG_TASK
912 	unsigned long			last_switch_count;
913 	unsigned long			last_switch_time;
914 #endif
915 	/* Filesystem information: */
916 	struct fs_struct		*fs;
917 
918 	/* Open file information: */
919 	struct files_struct		*files;
920 
921 	/* Namespaces: */
922 	struct nsproxy			*nsproxy;
923 
924 	/* Signal handlers: */
925 	struct signal_struct		*signal;
926 	struct sighand_struct __rcu		*sighand;
927 	sigset_t			blocked;
928 	sigset_t			real_blocked;
929 	/* Restored if set_restore_sigmask() was used: */
930 	sigset_t			saved_sigmask;
931 	struct sigpending		pending;
932 	unsigned long			sas_ss_sp;
933 	size_t				sas_ss_size;
934 	unsigned int			sas_ss_flags;
935 
936 	struct callback_head		*task_works;
937 
938 #ifdef CONFIG_AUDIT
939 #ifdef CONFIG_AUDITSYSCALL
940 	struct audit_context		*audit_context;
941 #endif
942 	kuid_t				loginuid;
943 	unsigned int			sessionid;
944 #endif
945 	struct seccomp			seccomp;
946 
947 	/* Thread group tracking: */
948 	u64				parent_exec_id;
949 	u64				self_exec_id;
950 
951 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
952 	spinlock_t			alloc_lock;
953 
954 	/* Protection of the PI data structures: */
955 	raw_spinlock_t			pi_lock;
956 
957 	struct wake_q_node		wake_q;
958 
959 #ifdef CONFIG_RT_MUTEXES
960 	/* PI waiters blocked on a rt_mutex held by this task: */
961 	struct rb_root_cached		pi_waiters;
962 	/* Updated under owner's pi_lock and rq lock */
963 	struct task_struct		*pi_top_task;
964 	/* Deadlock detection and priority inheritance handling: */
965 	struct rt_mutex_waiter		*pi_blocked_on;
966 #endif
967 
968 #ifdef CONFIG_DEBUG_MUTEXES
969 	/* Mutex deadlock detection: */
970 	struct mutex_waiter		*blocked_on;
971 #endif
972 
973 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
974 	int				non_block_count;
975 #endif
976 
977 #ifdef CONFIG_TRACE_IRQFLAGS
978 	unsigned int			irq_events;
979 	unsigned int			hardirq_threaded;
980 	unsigned long			hardirq_enable_ip;
981 	unsigned long			hardirq_disable_ip;
982 	unsigned int			hardirq_enable_event;
983 	unsigned int			hardirq_disable_event;
984 	int				hardirqs_enabled;
985 	int				hardirq_context;
986 	unsigned long			softirq_disable_ip;
987 	unsigned long			softirq_enable_ip;
988 	unsigned int			softirq_disable_event;
989 	unsigned int			softirq_enable_event;
990 	int				softirqs_enabled;
991 	int				softirq_context;
992 	int				irq_config;
993 #endif
994 
995 #ifdef CONFIG_LOCKDEP
996 # define MAX_LOCK_DEPTH			48UL
997 	u64				curr_chain_key;
998 	int				lockdep_depth;
999 	unsigned int			lockdep_recursion;
1000 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1001 #endif
1002 
1003 #ifdef CONFIG_UBSAN
1004 	unsigned int			in_ubsan;
1005 #endif
1006 
1007 	/* Journalling filesystem info: */
1008 	void				*journal_info;
1009 
1010 	/* Stacked block device info: */
1011 	struct bio_list			*bio_list;
1012 
1013 #ifdef CONFIG_BLOCK
1014 	/* Stack plugging: */
1015 	struct blk_plug			*plug;
1016 #endif
1017 
1018 	/* VM state: */
1019 	struct reclaim_state		*reclaim_state;
1020 
1021 	struct backing_dev_info		*backing_dev_info;
1022 
1023 	struct io_context		*io_context;
1024 
1025 #ifdef CONFIG_COMPACTION
1026 	struct capture_control		*capture_control;
1027 #endif
1028 	/* Ptrace state: */
1029 	unsigned long			ptrace_message;
1030 	kernel_siginfo_t		*last_siginfo;
1031 
1032 	struct task_io_accounting	ioac;
1033 #ifdef CONFIG_PSI
1034 	/* Pressure stall state */
1035 	unsigned int			psi_flags;
1036 #endif
1037 #ifdef CONFIG_TASK_XACCT
1038 	/* Accumulated RSS usage: */
1039 	u64				acct_rss_mem1;
1040 	/* Accumulated virtual memory usage: */
1041 	u64				acct_vm_mem1;
1042 	/* stime + utime since last update: */
1043 	u64				acct_timexpd;
1044 #endif
1045 #ifdef CONFIG_CPUSETS
1046 	/* Protected by ->alloc_lock: */
1047 	nodemask_t			mems_allowed;
1048 	/* Seqence number to catch updates: */
1049 	seqcount_t			mems_allowed_seq;
1050 	int				cpuset_mem_spread_rotor;
1051 	int				cpuset_slab_spread_rotor;
1052 #endif
1053 #ifdef CONFIG_CGROUPS
1054 	/* Control Group info protected by css_set_lock: */
1055 	struct css_set __rcu		*cgroups;
1056 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1057 	struct list_head		cg_list;
1058 #endif
1059 #ifdef CONFIG_X86_CPU_RESCTRL
1060 	u32				closid;
1061 	u32				rmid;
1062 #endif
1063 #ifdef CONFIG_FUTEX
1064 	struct robust_list_head __user	*robust_list;
1065 #ifdef CONFIG_COMPAT
1066 	struct compat_robust_list_head __user *compat_robust_list;
1067 #endif
1068 	struct list_head		pi_state_list;
1069 	struct futex_pi_state		*pi_state_cache;
1070 	struct mutex			futex_exit_mutex;
1071 	unsigned int			futex_state;
1072 #endif
1073 #ifdef CONFIG_PERF_EVENTS
1074 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1075 	struct mutex			perf_event_mutex;
1076 	struct list_head		perf_event_list;
1077 #endif
1078 #ifdef CONFIG_DEBUG_PREEMPT
1079 	unsigned long			preempt_disable_ip;
1080 #endif
1081 #ifdef CONFIG_NUMA
1082 	/* Protected by alloc_lock: */
1083 	struct mempolicy		*mempolicy;
1084 	short				il_prev;
1085 	short				pref_node_fork;
1086 #endif
1087 #ifdef CONFIG_NUMA_BALANCING
1088 	int				numa_scan_seq;
1089 	unsigned int			numa_scan_period;
1090 	unsigned int			numa_scan_period_max;
1091 	int				numa_preferred_nid;
1092 	unsigned long			numa_migrate_retry;
1093 	/* Migration stamp: */
1094 	u64				node_stamp;
1095 	u64				last_task_numa_placement;
1096 	u64				last_sum_exec_runtime;
1097 	struct callback_head		numa_work;
1098 
1099 	/*
1100 	 * This pointer is only modified for current in syscall and
1101 	 * pagefault context (and for tasks being destroyed), so it can be read
1102 	 * from any of the following contexts:
1103 	 *  - RCU read-side critical section
1104 	 *  - current->numa_group from everywhere
1105 	 *  - task's runqueue locked, task not running
1106 	 */
1107 	struct numa_group __rcu		*numa_group;
1108 
1109 	/*
1110 	 * numa_faults is an array split into four regions:
1111 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1112 	 * in this precise order.
1113 	 *
1114 	 * faults_memory: Exponential decaying average of faults on a per-node
1115 	 * basis. Scheduling placement decisions are made based on these
1116 	 * counts. The values remain static for the duration of a PTE scan.
1117 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1118 	 * hinting fault was incurred.
1119 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1120 	 * during the current scan window. When the scan completes, the counts
1121 	 * in faults_memory and faults_cpu decay and these values are copied.
1122 	 */
1123 	unsigned long			*numa_faults;
1124 	unsigned long			total_numa_faults;
1125 
1126 	/*
1127 	 * numa_faults_locality tracks if faults recorded during the last
1128 	 * scan window were remote/local or failed to migrate. The task scan
1129 	 * period is adapted based on the locality of the faults with different
1130 	 * weights depending on whether they were shared or private faults
1131 	 */
1132 	unsigned long			numa_faults_locality[3];
1133 
1134 	unsigned long			numa_pages_migrated;
1135 #endif /* CONFIG_NUMA_BALANCING */
1136 
1137 #ifdef CONFIG_RSEQ
1138 	struct rseq __user *rseq;
1139 	u32 rseq_sig;
1140 	/*
1141 	 * RmW on rseq_event_mask must be performed atomically
1142 	 * with respect to preemption.
1143 	 */
1144 	unsigned long rseq_event_mask;
1145 #endif
1146 
1147 	struct tlbflush_unmap_batch	tlb_ubc;
1148 
1149 	union {
1150 		refcount_t		rcu_users;
1151 		struct rcu_head		rcu;
1152 	};
1153 
1154 	/* Cache last used pipe for splice(): */
1155 	struct pipe_inode_info		*splice_pipe;
1156 
1157 	struct page_frag		task_frag;
1158 
1159 #ifdef CONFIG_TASK_DELAY_ACCT
1160 	struct task_delay_info		*delays;
1161 #endif
1162 
1163 #ifdef CONFIG_FAULT_INJECTION
1164 	int				make_it_fail;
1165 	unsigned int			fail_nth;
1166 #endif
1167 	/*
1168 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1169 	 * balance_dirty_pages() for a dirty throttling pause:
1170 	 */
1171 	int				nr_dirtied;
1172 	int				nr_dirtied_pause;
1173 	/* Start of a write-and-pause period: */
1174 	unsigned long			dirty_paused_when;
1175 
1176 #ifdef CONFIG_LATENCYTOP
1177 	int				latency_record_count;
1178 	struct latency_record		latency_record[LT_SAVECOUNT];
1179 #endif
1180 	/*
1181 	 * Time slack values; these are used to round up poll() and
1182 	 * select() etc timeout values. These are in nanoseconds.
1183 	 */
1184 	u64				timer_slack_ns;
1185 	u64				default_timer_slack_ns;
1186 
1187 #ifdef CONFIG_KASAN
1188 	unsigned int			kasan_depth;
1189 #endif
1190 
1191 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1192 	/* Index of current stored address in ret_stack: */
1193 	int				curr_ret_stack;
1194 	int				curr_ret_depth;
1195 
1196 	/* Stack of return addresses for return function tracing: */
1197 	struct ftrace_ret_stack		*ret_stack;
1198 
1199 	/* Timestamp for last schedule: */
1200 	unsigned long long		ftrace_timestamp;
1201 
1202 	/*
1203 	 * Number of functions that haven't been traced
1204 	 * because of depth overrun:
1205 	 */
1206 	atomic_t			trace_overrun;
1207 
1208 	/* Pause tracing: */
1209 	atomic_t			tracing_graph_pause;
1210 #endif
1211 
1212 #ifdef CONFIG_TRACING
1213 	/* State flags for use by tracers: */
1214 	unsigned long			trace;
1215 
1216 	/* Bitmask and counter of trace recursion: */
1217 	unsigned long			trace_recursion;
1218 #endif /* CONFIG_TRACING */
1219 
1220 #ifdef CONFIG_KCOV
1221 	/* See kernel/kcov.c for more details. */
1222 
1223 	/* Coverage collection mode enabled for this task (0 if disabled): */
1224 	unsigned int			kcov_mode;
1225 
1226 	/* Size of the kcov_area: */
1227 	unsigned int			kcov_size;
1228 
1229 	/* Buffer for coverage collection: */
1230 	void				*kcov_area;
1231 
1232 	/* KCOV descriptor wired with this task or NULL: */
1233 	struct kcov			*kcov;
1234 
1235 	/* KCOV common handle for remote coverage collection: */
1236 	u64				kcov_handle;
1237 
1238 	/* KCOV sequence number: */
1239 	int				kcov_sequence;
1240 #endif
1241 
1242 #ifdef CONFIG_MEMCG
1243 	struct mem_cgroup		*memcg_in_oom;
1244 	gfp_t				memcg_oom_gfp_mask;
1245 	int				memcg_oom_order;
1246 
1247 	/* Number of pages to reclaim on returning to userland: */
1248 	unsigned int			memcg_nr_pages_over_high;
1249 
1250 	/* Used by memcontrol for targeted memcg charge: */
1251 	struct mem_cgroup		*active_memcg;
1252 #endif
1253 
1254 #ifdef CONFIG_BLK_CGROUP
1255 	struct request_queue		*throttle_queue;
1256 #endif
1257 
1258 #ifdef CONFIG_UPROBES
1259 	struct uprobe_task		*utask;
1260 #endif
1261 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1262 	unsigned int			sequential_io;
1263 	unsigned int			sequential_io_avg;
1264 #endif
1265 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1266 	unsigned long			task_state_change;
1267 #endif
1268 	int				pagefault_disabled;
1269 #ifdef CONFIG_MMU
1270 	struct task_struct		*oom_reaper_list;
1271 #endif
1272 #ifdef CONFIG_VMAP_STACK
1273 	struct vm_struct		*stack_vm_area;
1274 #endif
1275 #ifdef CONFIG_THREAD_INFO_IN_TASK
1276 	/* A live task holds one reference: */
1277 	refcount_t			stack_refcount;
1278 #endif
1279 #ifdef CONFIG_LIVEPATCH
1280 	int patch_state;
1281 #endif
1282 #ifdef CONFIG_SECURITY
1283 	/* Used by LSM modules for access restriction: */
1284 	void				*security;
1285 #endif
1286 
1287 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1288 	unsigned long			lowest_stack;
1289 	unsigned long			prev_lowest_stack;
1290 #endif
1291 
1292 	/*
1293 	 * New fields for task_struct should be added above here, so that
1294 	 * they are included in the randomized portion of task_struct.
1295 	 */
1296 	randomized_struct_fields_end
1297 
1298 	/* CPU-specific state of this task: */
1299 	struct thread_struct		thread;
1300 
1301 	/*
1302 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1303 	 * structure.  It *MUST* be at the end of 'task_struct'.
1304 	 *
1305 	 * Do not put anything below here!
1306 	 */
1307 };
1308 
1309 static inline struct pid *task_pid(struct task_struct *task)
1310 {
1311 	return task->thread_pid;
1312 }
1313 
1314 /*
1315  * the helpers to get the task's different pids as they are seen
1316  * from various namespaces
1317  *
1318  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1319  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1320  *                     current.
1321  * task_xid_nr_ns()  : id seen from the ns specified;
1322  *
1323  * see also pid_nr() etc in include/linux/pid.h
1324  */
1325 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1326 
1327 static inline pid_t task_pid_nr(struct task_struct *tsk)
1328 {
1329 	return tsk->pid;
1330 }
1331 
1332 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1333 {
1334 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1335 }
1336 
1337 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1338 {
1339 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1340 }
1341 
1342 
1343 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1344 {
1345 	return tsk->tgid;
1346 }
1347 
1348 /**
1349  * pid_alive - check that a task structure is not stale
1350  * @p: Task structure to be checked.
1351  *
1352  * Test if a process is not yet dead (at most zombie state)
1353  * If pid_alive fails, then pointers within the task structure
1354  * can be stale and must not be dereferenced.
1355  *
1356  * Return: 1 if the process is alive. 0 otherwise.
1357  */
1358 static inline int pid_alive(const struct task_struct *p)
1359 {
1360 	return p->thread_pid != NULL;
1361 }
1362 
1363 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1364 {
1365 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1366 }
1367 
1368 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1369 {
1370 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1371 }
1372 
1373 
1374 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1375 {
1376 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1377 }
1378 
1379 static inline pid_t task_session_vnr(struct task_struct *tsk)
1380 {
1381 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1382 }
1383 
1384 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1385 {
1386 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1387 }
1388 
1389 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1390 {
1391 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1392 }
1393 
1394 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1395 {
1396 	pid_t pid = 0;
1397 
1398 	rcu_read_lock();
1399 	if (pid_alive(tsk))
1400 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1401 	rcu_read_unlock();
1402 
1403 	return pid;
1404 }
1405 
1406 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1407 {
1408 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1409 }
1410 
1411 /* Obsolete, do not use: */
1412 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1413 {
1414 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1415 }
1416 
1417 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1418 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1419 
1420 static inline unsigned int task_state_index(struct task_struct *tsk)
1421 {
1422 	unsigned int tsk_state = READ_ONCE(tsk->state);
1423 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1424 
1425 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1426 
1427 	if (tsk_state == TASK_IDLE)
1428 		state = TASK_REPORT_IDLE;
1429 
1430 	return fls(state);
1431 }
1432 
1433 static inline char task_index_to_char(unsigned int state)
1434 {
1435 	static const char state_char[] = "RSDTtXZPI";
1436 
1437 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1438 
1439 	return state_char[state];
1440 }
1441 
1442 static inline char task_state_to_char(struct task_struct *tsk)
1443 {
1444 	return task_index_to_char(task_state_index(tsk));
1445 }
1446 
1447 /**
1448  * is_global_init - check if a task structure is init. Since init
1449  * is free to have sub-threads we need to check tgid.
1450  * @tsk: Task structure to be checked.
1451  *
1452  * Check if a task structure is the first user space task the kernel created.
1453  *
1454  * Return: 1 if the task structure is init. 0 otherwise.
1455  */
1456 static inline int is_global_init(struct task_struct *tsk)
1457 {
1458 	return task_tgid_nr(tsk) == 1;
1459 }
1460 
1461 extern struct pid *cad_pid;
1462 
1463 /*
1464  * Per process flags
1465  */
1466 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1467 #define PF_EXITING		0x00000004	/* Getting shut down */
1468 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1469 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1470 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1471 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1472 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1473 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1474 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1475 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1476 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1477 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1478 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1479 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1480 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1481 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1482 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1483 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1484 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1485 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1486 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1487 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1488 #define PF_UMH			0x02000000	/* I'm an Usermodehelper process */
1489 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1490 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1491 #define PF_MEMALLOC_NOCMA	0x10000000	/* All allocation request will have _GFP_MOVABLE cleared */
1492 #define PF_IO_WORKER		0x20000000	/* Task is an IO worker */
1493 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1494 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1495 
1496 /*
1497  * Only the _current_ task can read/write to tsk->flags, but other
1498  * tasks can access tsk->flags in readonly mode for example
1499  * with tsk_used_math (like during threaded core dumping).
1500  * There is however an exception to this rule during ptrace
1501  * or during fork: the ptracer task is allowed to write to the
1502  * child->flags of its traced child (same goes for fork, the parent
1503  * can write to the child->flags), because we're guaranteed the
1504  * child is not running and in turn not changing child->flags
1505  * at the same time the parent does it.
1506  */
1507 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1508 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1509 #define clear_used_math()			clear_stopped_child_used_math(current)
1510 #define set_used_math()				set_stopped_child_used_math(current)
1511 
1512 #define conditional_stopped_child_used_math(condition, child) \
1513 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1514 
1515 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1516 
1517 #define copy_to_stopped_child_used_math(child) \
1518 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1519 
1520 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1521 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1522 #define used_math()				tsk_used_math(current)
1523 
1524 static inline bool is_percpu_thread(void)
1525 {
1526 #ifdef CONFIG_SMP
1527 	return (current->flags & PF_NO_SETAFFINITY) &&
1528 		(current->nr_cpus_allowed  == 1);
1529 #else
1530 	return true;
1531 #endif
1532 }
1533 
1534 /* Per-process atomic flags. */
1535 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1536 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1537 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1538 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1539 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1540 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1541 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1542 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1543 
1544 #define TASK_PFA_TEST(name, func)					\
1545 	static inline bool task_##func(struct task_struct *p)		\
1546 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1547 
1548 #define TASK_PFA_SET(name, func)					\
1549 	static inline void task_set_##func(struct task_struct *p)	\
1550 	{ set_bit(PFA_##name, &p->atomic_flags); }
1551 
1552 #define TASK_PFA_CLEAR(name, func)					\
1553 	static inline void task_clear_##func(struct task_struct *p)	\
1554 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1555 
1556 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1557 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1558 
1559 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1560 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1561 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1562 
1563 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1564 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1565 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1566 
1567 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1568 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1569 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1570 
1571 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1572 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1573 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1574 
1575 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1576 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1577 
1578 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1579 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1580 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1581 
1582 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1583 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1584 
1585 static inline void
1586 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1587 {
1588 	current->flags &= ~flags;
1589 	current->flags |= orig_flags & flags;
1590 }
1591 
1592 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1593 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1594 #ifdef CONFIG_SMP
1595 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1596 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1597 #else
1598 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1599 {
1600 }
1601 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1602 {
1603 	if (!cpumask_test_cpu(0, new_mask))
1604 		return -EINVAL;
1605 	return 0;
1606 }
1607 #endif
1608 
1609 extern int yield_to(struct task_struct *p, bool preempt);
1610 extern void set_user_nice(struct task_struct *p, long nice);
1611 extern int task_prio(const struct task_struct *p);
1612 
1613 /**
1614  * task_nice - return the nice value of a given task.
1615  * @p: the task in question.
1616  *
1617  * Return: The nice value [ -20 ... 0 ... 19 ].
1618  */
1619 static inline int task_nice(const struct task_struct *p)
1620 {
1621 	return PRIO_TO_NICE((p)->static_prio);
1622 }
1623 
1624 extern int can_nice(const struct task_struct *p, const int nice);
1625 extern int task_curr(const struct task_struct *p);
1626 extern int idle_cpu(int cpu);
1627 extern int available_idle_cpu(int cpu);
1628 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1629 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1630 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1631 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1632 extern struct task_struct *idle_task(int cpu);
1633 
1634 /**
1635  * is_idle_task - is the specified task an idle task?
1636  * @p: the task in question.
1637  *
1638  * Return: 1 if @p is an idle task. 0 otherwise.
1639  */
1640 static inline bool is_idle_task(const struct task_struct *p)
1641 {
1642 	return !!(p->flags & PF_IDLE);
1643 }
1644 
1645 extern struct task_struct *curr_task(int cpu);
1646 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1647 
1648 void yield(void);
1649 
1650 union thread_union {
1651 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1652 	struct task_struct task;
1653 #endif
1654 #ifndef CONFIG_THREAD_INFO_IN_TASK
1655 	struct thread_info thread_info;
1656 #endif
1657 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1658 };
1659 
1660 #ifndef CONFIG_THREAD_INFO_IN_TASK
1661 extern struct thread_info init_thread_info;
1662 #endif
1663 
1664 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1665 
1666 #ifdef CONFIG_THREAD_INFO_IN_TASK
1667 static inline struct thread_info *task_thread_info(struct task_struct *task)
1668 {
1669 	return &task->thread_info;
1670 }
1671 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1672 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1673 #endif
1674 
1675 /*
1676  * find a task by one of its numerical ids
1677  *
1678  * find_task_by_pid_ns():
1679  *      finds a task by its pid in the specified namespace
1680  * find_task_by_vpid():
1681  *      finds a task by its virtual pid
1682  *
1683  * see also find_vpid() etc in include/linux/pid.h
1684  */
1685 
1686 extern struct task_struct *find_task_by_vpid(pid_t nr);
1687 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1688 
1689 /*
1690  * find a task by its virtual pid and get the task struct
1691  */
1692 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1693 
1694 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1695 extern int wake_up_process(struct task_struct *tsk);
1696 extern void wake_up_new_task(struct task_struct *tsk);
1697 
1698 #ifdef CONFIG_SMP
1699 extern void kick_process(struct task_struct *tsk);
1700 #else
1701 static inline void kick_process(struct task_struct *tsk) { }
1702 #endif
1703 
1704 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1705 
1706 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1707 {
1708 	__set_task_comm(tsk, from, false);
1709 }
1710 
1711 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1712 #define get_task_comm(buf, tsk) ({			\
1713 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1714 	__get_task_comm(buf, sizeof(buf), tsk);		\
1715 })
1716 
1717 #ifdef CONFIG_SMP
1718 void scheduler_ipi(void);
1719 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1720 #else
1721 static inline void scheduler_ipi(void) { }
1722 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1723 {
1724 	return 1;
1725 }
1726 #endif
1727 
1728 /*
1729  * Set thread flags in other task's structures.
1730  * See asm/thread_info.h for TIF_xxxx flags available:
1731  */
1732 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1733 {
1734 	set_ti_thread_flag(task_thread_info(tsk), flag);
1735 }
1736 
1737 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1738 {
1739 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1740 }
1741 
1742 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1743 					  bool value)
1744 {
1745 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1746 }
1747 
1748 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1749 {
1750 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1751 }
1752 
1753 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1754 {
1755 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1756 }
1757 
1758 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1759 {
1760 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1761 }
1762 
1763 static inline void set_tsk_need_resched(struct task_struct *tsk)
1764 {
1765 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1766 }
1767 
1768 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1769 {
1770 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1771 }
1772 
1773 static inline int test_tsk_need_resched(struct task_struct *tsk)
1774 {
1775 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1776 }
1777 
1778 /*
1779  * cond_resched() and cond_resched_lock(): latency reduction via
1780  * explicit rescheduling in places that are safe. The return
1781  * value indicates whether a reschedule was done in fact.
1782  * cond_resched_lock() will drop the spinlock before scheduling,
1783  */
1784 #ifndef CONFIG_PREEMPTION
1785 extern int _cond_resched(void);
1786 #else
1787 static inline int _cond_resched(void) { return 0; }
1788 #endif
1789 
1790 #define cond_resched() ({			\
1791 	___might_sleep(__FILE__, __LINE__, 0);	\
1792 	_cond_resched();			\
1793 })
1794 
1795 extern int __cond_resched_lock(spinlock_t *lock);
1796 
1797 #define cond_resched_lock(lock) ({				\
1798 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1799 	__cond_resched_lock(lock);				\
1800 })
1801 
1802 static inline void cond_resched_rcu(void)
1803 {
1804 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1805 	rcu_read_unlock();
1806 	cond_resched();
1807 	rcu_read_lock();
1808 #endif
1809 }
1810 
1811 /*
1812  * Does a critical section need to be broken due to another
1813  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1814  * but a general need for low latency)
1815  */
1816 static inline int spin_needbreak(spinlock_t *lock)
1817 {
1818 #ifdef CONFIG_PREEMPTION
1819 	return spin_is_contended(lock);
1820 #else
1821 	return 0;
1822 #endif
1823 }
1824 
1825 static __always_inline bool need_resched(void)
1826 {
1827 	return unlikely(tif_need_resched());
1828 }
1829 
1830 /*
1831  * Wrappers for p->thread_info->cpu access. No-op on UP.
1832  */
1833 #ifdef CONFIG_SMP
1834 
1835 static inline unsigned int task_cpu(const struct task_struct *p)
1836 {
1837 #ifdef CONFIG_THREAD_INFO_IN_TASK
1838 	return READ_ONCE(p->cpu);
1839 #else
1840 	return READ_ONCE(task_thread_info(p)->cpu);
1841 #endif
1842 }
1843 
1844 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1845 
1846 #else
1847 
1848 static inline unsigned int task_cpu(const struct task_struct *p)
1849 {
1850 	return 0;
1851 }
1852 
1853 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1854 {
1855 }
1856 
1857 #endif /* CONFIG_SMP */
1858 
1859 /*
1860  * In order to reduce various lock holder preemption latencies provide an
1861  * interface to see if a vCPU is currently running or not.
1862  *
1863  * This allows us to terminate optimistic spin loops and block, analogous to
1864  * the native optimistic spin heuristic of testing if the lock owner task is
1865  * running or not.
1866  */
1867 #ifndef vcpu_is_preempted
1868 static inline bool vcpu_is_preempted(int cpu)
1869 {
1870 	return false;
1871 }
1872 #endif
1873 
1874 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1875 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1876 
1877 #ifndef TASK_SIZE_OF
1878 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1879 #endif
1880 
1881 #ifdef CONFIG_RSEQ
1882 
1883 /*
1884  * Map the event mask on the user-space ABI enum rseq_cs_flags
1885  * for direct mask checks.
1886  */
1887 enum rseq_event_mask_bits {
1888 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1889 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1890 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1891 };
1892 
1893 enum rseq_event_mask {
1894 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
1895 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
1896 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
1897 };
1898 
1899 static inline void rseq_set_notify_resume(struct task_struct *t)
1900 {
1901 	if (t->rseq)
1902 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1903 }
1904 
1905 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1906 
1907 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1908 					     struct pt_regs *regs)
1909 {
1910 	if (current->rseq)
1911 		__rseq_handle_notify_resume(ksig, regs);
1912 }
1913 
1914 static inline void rseq_signal_deliver(struct ksignal *ksig,
1915 				       struct pt_regs *regs)
1916 {
1917 	preempt_disable();
1918 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1919 	preempt_enable();
1920 	rseq_handle_notify_resume(ksig, regs);
1921 }
1922 
1923 /* rseq_preempt() requires preemption to be disabled. */
1924 static inline void rseq_preempt(struct task_struct *t)
1925 {
1926 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1927 	rseq_set_notify_resume(t);
1928 }
1929 
1930 /* rseq_migrate() requires preemption to be disabled. */
1931 static inline void rseq_migrate(struct task_struct *t)
1932 {
1933 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1934 	rseq_set_notify_resume(t);
1935 }
1936 
1937 /*
1938  * If parent process has a registered restartable sequences area, the
1939  * child inherits. Unregister rseq for a clone with CLONE_VM set.
1940  */
1941 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1942 {
1943 	if (clone_flags & CLONE_VM) {
1944 		t->rseq = NULL;
1945 		t->rseq_sig = 0;
1946 		t->rseq_event_mask = 0;
1947 	} else {
1948 		t->rseq = current->rseq;
1949 		t->rseq_sig = current->rseq_sig;
1950 		t->rseq_event_mask = current->rseq_event_mask;
1951 	}
1952 }
1953 
1954 static inline void rseq_execve(struct task_struct *t)
1955 {
1956 	t->rseq = NULL;
1957 	t->rseq_sig = 0;
1958 	t->rseq_event_mask = 0;
1959 }
1960 
1961 #else
1962 
1963 static inline void rseq_set_notify_resume(struct task_struct *t)
1964 {
1965 }
1966 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1967 					     struct pt_regs *regs)
1968 {
1969 }
1970 static inline void rseq_signal_deliver(struct ksignal *ksig,
1971 				       struct pt_regs *regs)
1972 {
1973 }
1974 static inline void rseq_preempt(struct task_struct *t)
1975 {
1976 }
1977 static inline void rseq_migrate(struct task_struct *t)
1978 {
1979 }
1980 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1981 {
1982 }
1983 static inline void rseq_execve(struct task_struct *t)
1984 {
1985 }
1986 
1987 #endif
1988 
1989 void __exit_umh(struct task_struct *tsk);
1990 
1991 static inline void exit_umh(struct task_struct *tsk)
1992 {
1993 	if (unlikely(tsk->flags & PF_UMH))
1994 		__exit_umh(tsk);
1995 }
1996 
1997 #ifdef CONFIG_DEBUG_RSEQ
1998 
1999 void rseq_syscall(struct pt_regs *regs);
2000 
2001 #else
2002 
2003 static inline void rseq_syscall(struct pt_regs *regs)
2004 {
2005 }
2006 
2007 #endif
2008 
2009 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2010 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2011 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2012 
2013 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2014 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2015 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2016 
2017 int sched_trace_rq_cpu(struct rq *rq);
2018 
2019 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2020 
2021 #endif
2022