xref: /linux/include/linux/sched.h (revision 2169e6daa1ffa6e9869fcc56ff7df23c9287f1ec)
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/signal_types.h>
29 #include <linux/mm_types_task.h>
30 #include <linux/task_io_accounting.h>
31 #include <linux/rseq.h>
32 
33 /* task_struct member predeclarations (sorted alphabetically): */
34 struct audit_context;
35 struct backing_dev_info;
36 struct bio_list;
37 struct blk_plug;
38 struct capture_control;
39 struct cfs_rq;
40 struct fs_struct;
41 struct futex_pi_state;
42 struct io_context;
43 struct mempolicy;
44 struct nameidata;
45 struct nsproxy;
46 struct perf_event_context;
47 struct pid_namespace;
48 struct pipe_inode_info;
49 struct rcu_node;
50 struct reclaim_state;
51 struct robust_list_head;
52 struct root_domain;
53 struct rq;
54 struct sched_attr;
55 struct sched_param;
56 struct seq_file;
57 struct sighand_struct;
58 struct signal_struct;
59 struct task_delay_info;
60 struct task_group;
61 
62 /*
63  * Task state bitmask. NOTE! These bits are also
64  * encoded in fs/proc/array.c: get_task_state().
65  *
66  * We have two separate sets of flags: task->state
67  * is about runnability, while task->exit_state are
68  * about the task exiting. Confusing, but this way
69  * modifying one set can't modify the other one by
70  * mistake.
71  */
72 
73 /* Used in tsk->state: */
74 #define TASK_RUNNING			0x0000
75 #define TASK_INTERRUPTIBLE		0x0001
76 #define TASK_UNINTERRUPTIBLE		0x0002
77 #define __TASK_STOPPED			0x0004
78 #define __TASK_TRACED			0x0008
79 /* Used in tsk->exit_state: */
80 #define EXIT_DEAD			0x0010
81 #define EXIT_ZOMBIE			0x0020
82 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
83 /* Used in tsk->state again: */
84 #define TASK_PARKED			0x0040
85 #define TASK_DEAD			0x0080
86 #define TASK_WAKEKILL			0x0100
87 #define TASK_WAKING			0x0200
88 #define TASK_NOLOAD			0x0400
89 #define TASK_NEW			0x0800
90 #define TASK_STATE_MAX			0x1000
91 
92 /* Convenience macros for the sake of set_current_state: */
93 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
94 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
95 #define TASK_TRACED			(TASK_WAKEKILL | __TASK_TRACED)
96 
97 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
98 
99 /* Convenience macros for the sake of wake_up(): */
100 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
101 
102 /* get_task_state(): */
103 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
104 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
105 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
106 					 TASK_PARKED)
107 
108 #define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
109 
110 #define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
111 
112 #define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
113 
114 #define task_contributes_to_load(task)	((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
115 					 (task->flags & PF_FROZEN) == 0 && \
116 					 (task->state & TASK_NOLOAD) == 0)
117 
118 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
119 
120 /*
121  * Special states are those that do not use the normal wait-loop pattern. See
122  * the comment with set_special_state().
123  */
124 #define is_special_task_state(state)				\
125 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
126 
127 #define __set_current_state(state_value)			\
128 	do {							\
129 		WARN_ON_ONCE(is_special_task_state(state_value));\
130 		current->task_state_change = _THIS_IP_;		\
131 		current->state = (state_value);			\
132 	} while (0)
133 
134 #define set_current_state(state_value)				\
135 	do {							\
136 		WARN_ON_ONCE(is_special_task_state(state_value));\
137 		current->task_state_change = _THIS_IP_;		\
138 		smp_store_mb(current->state, (state_value));	\
139 	} while (0)
140 
141 #define set_special_state(state_value)					\
142 	do {								\
143 		unsigned long flags; /* may shadow */			\
144 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
145 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
146 		current->task_state_change = _THIS_IP_;			\
147 		current->state = (state_value);				\
148 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
149 	} while (0)
150 #else
151 /*
152  * set_current_state() includes a barrier so that the write of current->state
153  * is correctly serialised wrt the caller's subsequent test of whether to
154  * actually sleep:
155  *
156  *   for (;;) {
157  *	set_current_state(TASK_UNINTERRUPTIBLE);
158  *	if (!need_sleep)
159  *		break;
160  *
161  *	schedule();
162  *   }
163  *   __set_current_state(TASK_RUNNING);
164  *
165  * If the caller does not need such serialisation (because, for instance, the
166  * condition test and condition change and wakeup are under the same lock) then
167  * use __set_current_state().
168  *
169  * The above is typically ordered against the wakeup, which does:
170  *
171  *   need_sleep = false;
172  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
173  *
174  * where wake_up_state() executes a full memory barrier before accessing the
175  * task state.
176  *
177  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
178  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
179  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
180  *
181  * However, with slightly different timing the wakeup TASK_RUNNING store can
182  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
183  * a problem either because that will result in one extra go around the loop
184  * and our @cond test will save the day.
185  *
186  * Also see the comments of try_to_wake_up().
187  */
188 #define __set_current_state(state_value)				\
189 	current->state = (state_value)
190 
191 #define set_current_state(state_value)					\
192 	smp_store_mb(current->state, (state_value))
193 
194 /*
195  * set_special_state() should be used for those states when the blocking task
196  * can not use the regular condition based wait-loop. In that case we must
197  * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
198  * will not collide with our state change.
199  */
200 #define set_special_state(state_value)					\
201 	do {								\
202 		unsigned long flags; /* may shadow */			\
203 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
204 		current->state = (state_value);				\
205 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
206 	} while (0)
207 
208 #endif
209 
210 /* Task command name length: */
211 #define TASK_COMM_LEN			16
212 
213 extern void scheduler_tick(void);
214 
215 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
216 
217 extern long schedule_timeout(long timeout);
218 extern long schedule_timeout_interruptible(long timeout);
219 extern long schedule_timeout_killable(long timeout);
220 extern long schedule_timeout_uninterruptible(long timeout);
221 extern long schedule_timeout_idle(long timeout);
222 asmlinkage void schedule(void);
223 extern void schedule_preempt_disabled(void);
224 
225 extern int __must_check io_schedule_prepare(void);
226 extern void io_schedule_finish(int token);
227 extern long io_schedule_timeout(long timeout);
228 extern void io_schedule(void);
229 
230 /**
231  * struct prev_cputime - snapshot of system and user cputime
232  * @utime: time spent in user mode
233  * @stime: time spent in system mode
234  * @lock: protects the above two fields
235  *
236  * Stores previous user/system time values such that we can guarantee
237  * monotonicity.
238  */
239 struct prev_cputime {
240 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
241 	u64				utime;
242 	u64				stime;
243 	raw_spinlock_t			lock;
244 #endif
245 };
246 
247 /**
248  * struct task_cputime - collected CPU time counts
249  * @utime:		time spent in user mode, in nanoseconds
250  * @stime:		time spent in kernel mode, in nanoseconds
251  * @sum_exec_runtime:	total time spent on the CPU, in nanoseconds
252  *
253  * This structure groups together three kinds of CPU time that are tracked for
254  * threads and thread groups.  Most things considering CPU time want to group
255  * these counts together and treat all three of them in parallel.
256  */
257 struct task_cputime {
258 	u64				utime;
259 	u64				stime;
260 	unsigned long long		sum_exec_runtime;
261 };
262 
263 /* Alternate field names when used on cache expirations: */
264 #define virt_exp			utime
265 #define prof_exp			stime
266 #define sched_exp			sum_exec_runtime
267 
268 enum vtime_state {
269 	/* Task is sleeping or running in a CPU with VTIME inactive: */
270 	VTIME_INACTIVE = 0,
271 	/* Task runs in userspace in a CPU with VTIME active: */
272 	VTIME_USER,
273 	/* Task runs in kernelspace in a CPU with VTIME active: */
274 	VTIME_SYS,
275 };
276 
277 struct vtime {
278 	seqcount_t		seqcount;
279 	unsigned long long	starttime;
280 	enum vtime_state	state;
281 	u64			utime;
282 	u64			stime;
283 	u64			gtime;
284 };
285 
286 /*
287  * Utilization clamp constraints.
288  * @UCLAMP_MIN:	Minimum utilization
289  * @UCLAMP_MAX:	Maximum utilization
290  * @UCLAMP_CNT:	Utilization clamp constraints count
291  */
292 enum uclamp_id {
293 	UCLAMP_MIN = 0,
294 	UCLAMP_MAX,
295 	UCLAMP_CNT
296 };
297 
298 struct sched_info {
299 #ifdef CONFIG_SCHED_INFO
300 	/* Cumulative counters: */
301 
302 	/* # of times we have run on this CPU: */
303 	unsigned long			pcount;
304 
305 	/* Time spent waiting on a runqueue: */
306 	unsigned long long		run_delay;
307 
308 	/* Timestamps: */
309 
310 	/* When did we last run on a CPU? */
311 	unsigned long long		last_arrival;
312 
313 	/* When were we last queued to run? */
314 	unsigned long long		last_queued;
315 
316 #endif /* CONFIG_SCHED_INFO */
317 };
318 
319 /*
320  * Integer metrics need fixed point arithmetic, e.g., sched/fair
321  * has a few: load, load_avg, util_avg, freq, and capacity.
322  *
323  * We define a basic fixed point arithmetic range, and then formalize
324  * all these metrics based on that basic range.
325  */
326 # define SCHED_FIXEDPOINT_SHIFT		10
327 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
328 
329 /* Increase resolution of cpu_capacity calculations */
330 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
331 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
332 
333 struct load_weight {
334 	unsigned long			weight;
335 	u32				inv_weight;
336 };
337 
338 /**
339  * struct util_est - Estimation utilization of FAIR tasks
340  * @enqueued: instantaneous estimated utilization of a task/cpu
341  * @ewma:     the Exponential Weighted Moving Average (EWMA)
342  *            utilization of a task
343  *
344  * Support data structure to track an Exponential Weighted Moving Average
345  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
346  * average each time a task completes an activation. Sample's weight is chosen
347  * so that the EWMA will be relatively insensitive to transient changes to the
348  * task's workload.
349  *
350  * The enqueued attribute has a slightly different meaning for tasks and cpus:
351  * - task:   the task's util_avg at last task dequeue time
352  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
353  * Thus, the util_est.enqueued of a task represents the contribution on the
354  * estimated utilization of the CPU where that task is currently enqueued.
355  *
356  * Only for tasks we track a moving average of the past instantaneous
357  * estimated utilization. This allows to absorb sporadic drops in utilization
358  * of an otherwise almost periodic task.
359  */
360 struct util_est {
361 	unsigned int			enqueued;
362 	unsigned int			ewma;
363 #define UTIL_EST_WEIGHT_SHIFT		2
364 } __attribute__((__aligned__(sizeof(u64))));
365 
366 /*
367  * The load_avg/util_avg accumulates an infinite geometric series
368  * (see __update_load_avg() in kernel/sched/fair.c).
369  *
370  * [load_avg definition]
371  *
372  *   load_avg = runnable% * scale_load_down(load)
373  *
374  * where runnable% is the time ratio that a sched_entity is runnable.
375  * For cfs_rq, it is the aggregated load_avg of all runnable and
376  * blocked sched_entities.
377  *
378  * [util_avg definition]
379  *
380  *   util_avg = running% * SCHED_CAPACITY_SCALE
381  *
382  * where running% is the time ratio that a sched_entity is running on
383  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
384  * and blocked sched_entities.
385  *
386  * load_avg and util_avg don't direcly factor frequency scaling and CPU
387  * capacity scaling. The scaling is done through the rq_clock_pelt that
388  * is used for computing those signals (see update_rq_clock_pelt())
389  *
390  * N.B., the above ratios (runnable% and running%) themselves are in the
391  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
392  * to as large a range as necessary. This is for example reflected by
393  * util_avg's SCHED_CAPACITY_SCALE.
394  *
395  * [Overflow issue]
396  *
397  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
398  * with the highest load (=88761), always runnable on a single cfs_rq,
399  * and should not overflow as the number already hits PID_MAX_LIMIT.
400  *
401  * For all other cases (including 32-bit kernels), struct load_weight's
402  * weight will overflow first before we do, because:
403  *
404  *    Max(load_avg) <= Max(load.weight)
405  *
406  * Then it is the load_weight's responsibility to consider overflow
407  * issues.
408  */
409 struct sched_avg {
410 	u64				last_update_time;
411 	u64				load_sum;
412 	u64				runnable_load_sum;
413 	u32				util_sum;
414 	u32				period_contrib;
415 	unsigned long			load_avg;
416 	unsigned long			runnable_load_avg;
417 	unsigned long			util_avg;
418 	struct util_est			util_est;
419 } ____cacheline_aligned;
420 
421 struct sched_statistics {
422 #ifdef CONFIG_SCHEDSTATS
423 	u64				wait_start;
424 	u64				wait_max;
425 	u64				wait_count;
426 	u64				wait_sum;
427 	u64				iowait_count;
428 	u64				iowait_sum;
429 
430 	u64				sleep_start;
431 	u64				sleep_max;
432 	s64				sum_sleep_runtime;
433 
434 	u64				block_start;
435 	u64				block_max;
436 	u64				exec_max;
437 	u64				slice_max;
438 
439 	u64				nr_migrations_cold;
440 	u64				nr_failed_migrations_affine;
441 	u64				nr_failed_migrations_running;
442 	u64				nr_failed_migrations_hot;
443 	u64				nr_forced_migrations;
444 
445 	u64				nr_wakeups;
446 	u64				nr_wakeups_sync;
447 	u64				nr_wakeups_migrate;
448 	u64				nr_wakeups_local;
449 	u64				nr_wakeups_remote;
450 	u64				nr_wakeups_affine;
451 	u64				nr_wakeups_affine_attempts;
452 	u64				nr_wakeups_passive;
453 	u64				nr_wakeups_idle;
454 #endif
455 };
456 
457 struct sched_entity {
458 	/* For load-balancing: */
459 	struct load_weight		load;
460 	unsigned long			runnable_weight;
461 	struct rb_node			run_node;
462 	struct list_head		group_node;
463 	unsigned int			on_rq;
464 
465 	u64				exec_start;
466 	u64				sum_exec_runtime;
467 	u64				vruntime;
468 	u64				prev_sum_exec_runtime;
469 
470 	u64				nr_migrations;
471 
472 	struct sched_statistics		statistics;
473 
474 #ifdef CONFIG_FAIR_GROUP_SCHED
475 	int				depth;
476 	struct sched_entity		*parent;
477 	/* rq on which this entity is (to be) queued: */
478 	struct cfs_rq			*cfs_rq;
479 	/* rq "owned" by this entity/group: */
480 	struct cfs_rq			*my_q;
481 #endif
482 
483 #ifdef CONFIG_SMP
484 	/*
485 	 * Per entity load average tracking.
486 	 *
487 	 * Put into separate cache line so it does not
488 	 * collide with read-mostly values above.
489 	 */
490 	struct sched_avg		avg;
491 #endif
492 };
493 
494 struct sched_rt_entity {
495 	struct list_head		run_list;
496 	unsigned long			timeout;
497 	unsigned long			watchdog_stamp;
498 	unsigned int			time_slice;
499 	unsigned short			on_rq;
500 	unsigned short			on_list;
501 
502 	struct sched_rt_entity		*back;
503 #ifdef CONFIG_RT_GROUP_SCHED
504 	struct sched_rt_entity		*parent;
505 	/* rq on which this entity is (to be) queued: */
506 	struct rt_rq			*rt_rq;
507 	/* rq "owned" by this entity/group: */
508 	struct rt_rq			*my_q;
509 #endif
510 } __randomize_layout;
511 
512 struct sched_dl_entity {
513 	struct rb_node			rb_node;
514 
515 	/*
516 	 * Original scheduling parameters. Copied here from sched_attr
517 	 * during sched_setattr(), they will remain the same until
518 	 * the next sched_setattr().
519 	 */
520 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
521 	u64				dl_deadline;	/* Relative deadline of each instance	*/
522 	u64				dl_period;	/* Separation of two instances (period) */
523 	u64				dl_bw;		/* dl_runtime / dl_period		*/
524 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
525 
526 	/*
527 	 * Actual scheduling parameters. Initialized with the values above,
528 	 * they are continuously updated during task execution. Note that
529 	 * the remaining runtime could be < 0 in case we are in overrun.
530 	 */
531 	s64				runtime;	/* Remaining runtime for this instance	*/
532 	u64				deadline;	/* Absolute deadline for this instance	*/
533 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
534 
535 	/*
536 	 * Some bool flags:
537 	 *
538 	 * @dl_throttled tells if we exhausted the runtime. If so, the
539 	 * task has to wait for a replenishment to be performed at the
540 	 * next firing of dl_timer.
541 	 *
542 	 * @dl_boosted tells if we are boosted due to DI. If so we are
543 	 * outside bandwidth enforcement mechanism (but only until we
544 	 * exit the critical section);
545 	 *
546 	 * @dl_yielded tells if task gave up the CPU before consuming
547 	 * all its available runtime during the last job.
548 	 *
549 	 * @dl_non_contending tells if the task is inactive while still
550 	 * contributing to the active utilization. In other words, it
551 	 * indicates if the inactive timer has been armed and its handler
552 	 * has not been executed yet. This flag is useful to avoid race
553 	 * conditions between the inactive timer handler and the wakeup
554 	 * code.
555 	 *
556 	 * @dl_overrun tells if the task asked to be informed about runtime
557 	 * overruns.
558 	 */
559 	unsigned int			dl_throttled      : 1;
560 	unsigned int			dl_boosted        : 1;
561 	unsigned int			dl_yielded        : 1;
562 	unsigned int			dl_non_contending : 1;
563 	unsigned int			dl_overrun	  : 1;
564 
565 	/*
566 	 * Bandwidth enforcement timer. Each -deadline task has its
567 	 * own bandwidth to be enforced, thus we need one timer per task.
568 	 */
569 	struct hrtimer			dl_timer;
570 
571 	/*
572 	 * Inactive timer, responsible for decreasing the active utilization
573 	 * at the "0-lag time". When a -deadline task blocks, it contributes
574 	 * to GRUB's active utilization until the "0-lag time", hence a
575 	 * timer is needed to decrease the active utilization at the correct
576 	 * time.
577 	 */
578 	struct hrtimer inactive_timer;
579 };
580 
581 #ifdef CONFIG_UCLAMP_TASK
582 /* Number of utilization clamp buckets (shorter alias) */
583 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
584 
585 /*
586  * Utilization clamp for a scheduling entity
587  * @value:		clamp value "assigned" to a se
588  * @bucket_id:		bucket index corresponding to the "assigned" value
589  * @active:		the se is currently refcounted in a rq's bucket
590  * @user_defined:	the requested clamp value comes from user-space
591  *
592  * The bucket_id is the index of the clamp bucket matching the clamp value
593  * which is pre-computed and stored to avoid expensive integer divisions from
594  * the fast path.
595  *
596  * The active bit is set whenever a task has got an "effective" value assigned,
597  * which can be different from the clamp value "requested" from user-space.
598  * This allows to know a task is refcounted in the rq's bucket corresponding
599  * to the "effective" bucket_id.
600  *
601  * The user_defined bit is set whenever a task has got a task-specific clamp
602  * value requested from userspace, i.e. the system defaults apply to this task
603  * just as a restriction. This allows to relax default clamps when a less
604  * restrictive task-specific value has been requested, thus allowing to
605  * implement a "nice" semantic. For example, a task running with a 20%
606  * default boost can still drop its own boosting to 0%.
607  */
608 struct uclamp_se {
609 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
610 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
611 	unsigned int active		: 1;
612 	unsigned int user_defined	: 1;
613 };
614 #endif /* CONFIG_UCLAMP_TASK */
615 
616 union rcu_special {
617 	struct {
618 		u8			blocked;
619 		u8			need_qs;
620 		u8			exp_hint; /* Hint for performance. */
621 		u8			deferred_qs;
622 	} b; /* Bits. */
623 	u32 s; /* Set of bits. */
624 };
625 
626 enum perf_event_task_context {
627 	perf_invalid_context = -1,
628 	perf_hw_context = 0,
629 	perf_sw_context,
630 	perf_nr_task_contexts,
631 };
632 
633 struct wake_q_node {
634 	struct wake_q_node *next;
635 };
636 
637 struct task_struct {
638 #ifdef CONFIG_THREAD_INFO_IN_TASK
639 	/*
640 	 * For reasons of header soup (see current_thread_info()), this
641 	 * must be the first element of task_struct.
642 	 */
643 	struct thread_info		thread_info;
644 #endif
645 	/* -1 unrunnable, 0 runnable, >0 stopped: */
646 	volatile long			state;
647 
648 	/*
649 	 * This begins the randomizable portion of task_struct. Only
650 	 * scheduling-critical items should be added above here.
651 	 */
652 	randomized_struct_fields_start
653 
654 	void				*stack;
655 	refcount_t			usage;
656 	/* Per task flags (PF_*), defined further below: */
657 	unsigned int			flags;
658 	unsigned int			ptrace;
659 
660 #ifdef CONFIG_SMP
661 	struct llist_node		wake_entry;
662 	int				on_cpu;
663 #ifdef CONFIG_THREAD_INFO_IN_TASK
664 	/* Current CPU: */
665 	unsigned int			cpu;
666 #endif
667 	unsigned int			wakee_flips;
668 	unsigned long			wakee_flip_decay_ts;
669 	struct task_struct		*last_wakee;
670 
671 	/*
672 	 * recent_used_cpu is initially set as the last CPU used by a task
673 	 * that wakes affine another task. Waker/wakee relationships can
674 	 * push tasks around a CPU where each wakeup moves to the next one.
675 	 * Tracking a recently used CPU allows a quick search for a recently
676 	 * used CPU that may be idle.
677 	 */
678 	int				recent_used_cpu;
679 	int				wake_cpu;
680 #endif
681 	int				on_rq;
682 
683 	int				prio;
684 	int				static_prio;
685 	int				normal_prio;
686 	unsigned int			rt_priority;
687 
688 	const struct sched_class	*sched_class;
689 	struct sched_entity		se;
690 	struct sched_rt_entity		rt;
691 #ifdef CONFIG_CGROUP_SCHED
692 	struct task_group		*sched_task_group;
693 #endif
694 	struct sched_dl_entity		dl;
695 
696 #ifdef CONFIG_UCLAMP_TASK
697 	/* Clamp values requested for a scheduling entity */
698 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
699 	/* Effective clamp values used for a scheduling entity */
700 	struct uclamp_se		uclamp[UCLAMP_CNT];
701 #endif
702 
703 #ifdef CONFIG_PREEMPT_NOTIFIERS
704 	/* List of struct preempt_notifier: */
705 	struct hlist_head		preempt_notifiers;
706 #endif
707 
708 #ifdef CONFIG_BLK_DEV_IO_TRACE
709 	unsigned int			btrace_seq;
710 #endif
711 
712 	unsigned int			policy;
713 	int				nr_cpus_allowed;
714 	const cpumask_t			*cpus_ptr;
715 	cpumask_t			cpus_mask;
716 
717 #ifdef CONFIG_PREEMPT_RCU
718 	int				rcu_read_lock_nesting;
719 	union rcu_special		rcu_read_unlock_special;
720 	struct list_head		rcu_node_entry;
721 	struct rcu_node			*rcu_blocked_node;
722 #endif /* #ifdef CONFIG_PREEMPT_RCU */
723 
724 #ifdef CONFIG_TASKS_RCU
725 	unsigned long			rcu_tasks_nvcsw;
726 	u8				rcu_tasks_holdout;
727 	u8				rcu_tasks_idx;
728 	int				rcu_tasks_idle_cpu;
729 	struct list_head		rcu_tasks_holdout_list;
730 #endif /* #ifdef CONFIG_TASKS_RCU */
731 
732 	struct sched_info		sched_info;
733 
734 	struct list_head		tasks;
735 #ifdef CONFIG_SMP
736 	struct plist_node		pushable_tasks;
737 	struct rb_node			pushable_dl_tasks;
738 #endif
739 
740 	struct mm_struct		*mm;
741 	struct mm_struct		*active_mm;
742 
743 	/* Per-thread vma caching: */
744 	struct vmacache			vmacache;
745 
746 #ifdef SPLIT_RSS_COUNTING
747 	struct task_rss_stat		rss_stat;
748 #endif
749 	int				exit_state;
750 	int				exit_code;
751 	int				exit_signal;
752 	/* The signal sent when the parent dies: */
753 	int				pdeath_signal;
754 	/* JOBCTL_*, siglock protected: */
755 	unsigned long			jobctl;
756 
757 	/* Used for emulating ABI behavior of previous Linux versions: */
758 	unsigned int			personality;
759 
760 	/* Scheduler bits, serialized by scheduler locks: */
761 	unsigned			sched_reset_on_fork:1;
762 	unsigned			sched_contributes_to_load:1;
763 	unsigned			sched_migrated:1;
764 	unsigned			sched_remote_wakeup:1;
765 #ifdef CONFIG_PSI
766 	unsigned			sched_psi_wake_requeue:1;
767 #endif
768 
769 	/* Force alignment to the next boundary: */
770 	unsigned			:0;
771 
772 	/* Unserialized, strictly 'current' */
773 
774 	/* Bit to tell LSMs we're in execve(): */
775 	unsigned			in_execve:1;
776 	unsigned			in_iowait:1;
777 #ifndef TIF_RESTORE_SIGMASK
778 	unsigned			restore_sigmask:1;
779 #endif
780 #ifdef CONFIG_MEMCG
781 	unsigned			in_user_fault:1;
782 #endif
783 #ifdef CONFIG_COMPAT_BRK
784 	unsigned			brk_randomized:1;
785 #endif
786 #ifdef CONFIG_CGROUPS
787 	/* disallow userland-initiated cgroup migration */
788 	unsigned			no_cgroup_migration:1;
789 	/* task is frozen/stopped (used by the cgroup freezer) */
790 	unsigned			frozen:1;
791 #endif
792 #ifdef CONFIG_BLK_CGROUP
793 	/* to be used once the psi infrastructure lands upstream. */
794 	unsigned			use_memdelay:1;
795 #endif
796 
797 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
798 
799 	struct restart_block		restart_block;
800 
801 	pid_t				pid;
802 	pid_t				tgid;
803 
804 #ifdef CONFIG_STACKPROTECTOR
805 	/* Canary value for the -fstack-protector GCC feature: */
806 	unsigned long			stack_canary;
807 #endif
808 	/*
809 	 * Pointers to the (original) parent process, youngest child, younger sibling,
810 	 * older sibling, respectively.  (p->father can be replaced with
811 	 * p->real_parent->pid)
812 	 */
813 
814 	/* Real parent process: */
815 	struct task_struct __rcu	*real_parent;
816 
817 	/* Recipient of SIGCHLD, wait4() reports: */
818 	struct task_struct __rcu	*parent;
819 
820 	/*
821 	 * Children/sibling form the list of natural children:
822 	 */
823 	struct list_head		children;
824 	struct list_head		sibling;
825 	struct task_struct		*group_leader;
826 
827 	/*
828 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
829 	 *
830 	 * This includes both natural children and PTRACE_ATTACH targets.
831 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
832 	 */
833 	struct list_head		ptraced;
834 	struct list_head		ptrace_entry;
835 
836 	/* PID/PID hash table linkage. */
837 	struct pid			*thread_pid;
838 	struct hlist_node		pid_links[PIDTYPE_MAX];
839 	struct list_head		thread_group;
840 	struct list_head		thread_node;
841 
842 	struct completion		*vfork_done;
843 
844 	/* CLONE_CHILD_SETTID: */
845 	int __user			*set_child_tid;
846 
847 	/* CLONE_CHILD_CLEARTID: */
848 	int __user			*clear_child_tid;
849 
850 	u64				utime;
851 	u64				stime;
852 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
853 	u64				utimescaled;
854 	u64				stimescaled;
855 #endif
856 	u64				gtime;
857 	struct prev_cputime		prev_cputime;
858 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
859 	struct vtime			vtime;
860 #endif
861 
862 #ifdef CONFIG_NO_HZ_FULL
863 	atomic_t			tick_dep_mask;
864 #endif
865 	/* Context switch counts: */
866 	unsigned long			nvcsw;
867 	unsigned long			nivcsw;
868 
869 	/* Monotonic time in nsecs: */
870 	u64				start_time;
871 
872 	/* Boot based time in nsecs: */
873 	u64				real_start_time;
874 
875 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
876 	unsigned long			min_flt;
877 	unsigned long			maj_flt;
878 
879 #ifdef CONFIG_POSIX_TIMERS
880 	struct task_cputime		cputime_expires;
881 	struct list_head		cpu_timers[3];
882 #endif
883 
884 	/* Process credentials: */
885 
886 	/* Tracer's credentials at attach: */
887 	const struct cred __rcu		*ptracer_cred;
888 
889 	/* Objective and real subjective task credentials (COW): */
890 	const struct cred __rcu		*real_cred;
891 
892 	/* Effective (overridable) subjective task credentials (COW): */
893 	const struct cred __rcu		*cred;
894 
895 #ifdef CONFIG_KEYS
896 	/* Cached requested key. */
897 	struct key			*cached_requested_key;
898 #endif
899 
900 	/*
901 	 * executable name, excluding path.
902 	 *
903 	 * - normally initialized setup_new_exec()
904 	 * - access it with [gs]et_task_comm()
905 	 * - lock it with task_lock()
906 	 */
907 	char				comm[TASK_COMM_LEN];
908 
909 	struct nameidata		*nameidata;
910 
911 #ifdef CONFIG_SYSVIPC
912 	struct sysv_sem			sysvsem;
913 	struct sysv_shm			sysvshm;
914 #endif
915 #ifdef CONFIG_DETECT_HUNG_TASK
916 	unsigned long			last_switch_count;
917 	unsigned long			last_switch_time;
918 #endif
919 	/* Filesystem information: */
920 	struct fs_struct		*fs;
921 
922 	/* Open file information: */
923 	struct files_struct		*files;
924 
925 	/* Namespaces: */
926 	struct nsproxy			*nsproxy;
927 
928 	/* Signal handlers: */
929 	struct signal_struct		*signal;
930 	struct sighand_struct		*sighand;
931 	sigset_t			blocked;
932 	sigset_t			real_blocked;
933 	/* Restored if set_restore_sigmask() was used: */
934 	sigset_t			saved_sigmask;
935 	struct sigpending		pending;
936 	unsigned long			sas_ss_sp;
937 	size_t				sas_ss_size;
938 	unsigned int			sas_ss_flags;
939 
940 	struct callback_head		*task_works;
941 
942 #ifdef CONFIG_AUDIT
943 #ifdef CONFIG_AUDITSYSCALL
944 	struct audit_context		*audit_context;
945 #endif
946 	kuid_t				loginuid;
947 	unsigned int			sessionid;
948 #endif
949 	struct seccomp			seccomp;
950 
951 	/* Thread group tracking: */
952 	u32				parent_exec_id;
953 	u32				self_exec_id;
954 
955 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
956 	spinlock_t			alloc_lock;
957 
958 	/* Protection of the PI data structures: */
959 	raw_spinlock_t			pi_lock;
960 
961 	struct wake_q_node		wake_q;
962 
963 #ifdef CONFIG_RT_MUTEXES
964 	/* PI waiters blocked on a rt_mutex held by this task: */
965 	struct rb_root_cached		pi_waiters;
966 	/* Updated under owner's pi_lock and rq lock */
967 	struct task_struct		*pi_top_task;
968 	/* Deadlock detection and priority inheritance handling: */
969 	struct rt_mutex_waiter		*pi_blocked_on;
970 #endif
971 
972 #ifdef CONFIG_DEBUG_MUTEXES
973 	/* Mutex deadlock detection: */
974 	struct mutex_waiter		*blocked_on;
975 #endif
976 
977 #ifdef CONFIG_TRACE_IRQFLAGS
978 	unsigned int			irq_events;
979 	unsigned long			hardirq_enable_ip;
980 	unsigned long			hardirq_disable_ip;
981 	unsigned int			hardirq_enable_event;
982 	unsigned int			hardirq_disable_event;
983 	int				hardirqs_enabled;
984 	int				hardirq_context;
985 	unsigned long			softirq_disable_ip;
986 	unsigned long			softirq_enable_ip;
987 	unsigned int			softirq_disable_event;
988 	unsigned int			softirq_enable_event;
989 	int				softirqs_enabled;
990 	int				softirq_context;
991 #endif
992 
993 #ifdef CONFIG_LOCKDEP
994 # define MAX_LOCK_DEPTH			48UL
995 	u64				curr_chain_key;
996 	int				lockdep_depth;
997 	unsigned int			lockdep_recursion;
998 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
999 #endif
1000 
1001 #ifdef CONFIG_UBSAN
1002 	unsigned int			in_ubsan;
1003 #endif
1004 
1005 	/* Journalling filesystem info: */
1006 	void				*journal_info;
1007 
1008 	/* Stacked block device info: */
1009 	struct bio_list			*bio_list;
1010 
1011 #ifdef CONFIG_BLOCK
1012 	/* Stack plugging: */
1013 	struct blk_plug			*plug;
1014 #endif
1015 
1016 	/* VM state: */
1017 	struct reclaim_state		*reclaim_state;
1018 
1019 	struct backing_dev_info		*backing_dev_info;
1020 
1021 	struct io_context		*io_context;
1022 
1023 #ifdef CONFIG_COMPACTION
1024 	struct capture_control		*capture_control;
1025 #endif
1026 	/* Ptrace state: */
1027 	unsigned long			ptrace_message;
1028 	kernel_siginfo_t		*last_siginfo;
1029 
1030 	struct task_io_accounting	ioac;
1031 #ifdef CONFIG_PSI
1032 	/* Pressure stall state */
1033 	unsigned int			psi_flags;
1034 #endif
1035 #ifdef CONFIG_TASK_XACCT
1036 	/* Accumulated RSS usage: */
1037 	u64				acct_rss_mem1;
1038 	/* Accumulated virtual memory usage: */
1039 	u64				acct_vm_mem1;
1040 	/* stime + utime since last update: */
1041 	u64				acct_timexpd;
1042 #endif
1043 #ifdef CONFIG_CPUSETS
1044 	/* Protected by ->alloc_lock: */
1045 	nodemask_t			mems_allowed;
1046 	/* Seqence number to catch updates: */
1047 	seqcount_t			mems_allowed_seq;
1048 	int				cpuset_mem_spread_rotor;
1049 	int				cpuset_slab_spread_rotor;
1050 #endif
1051 #ifdef CONFIG_CGROUPS
1052 	/* Control Group info protected by css_set_lock: */
1053 	struct css_set __rcu		*cgroups;
1054 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1055 	struct list_head		cg_list;
1056 #endif
1057 #ifdef CONFIG_X86_CPU_RESCTRL
1058 	u32				closid;
1059 	u32				rmid;
1060 #endif
1061 #ifdef CONFIG_FUTEX
1062 	struct robust_list_head __user	*robust_list;
1063 #ifdef CONFIG_COMPAT
1064 	struct compat_robust_list_head __user *compat_robust_list;
1065 #endif
1066 	struct list_head		pi_state_list;
1067 	struct futex_pi_state		*pi_state_cache;
1068 #endif
1069 #ifdef CONFIG_PERF_EVENTS
1070 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1071 	struct mutex			perf_event_mutex;
1072 	struct list_head		perf_event_list;
1073 #endif
1074 #ifdef CONFIG_DEBUG_PREEMPT
1075 	unsigned long			preempt_disable_ip;
1076 #endif
1077 #ifdef CONFIG_NUMA
1078 	/* Protected by alloc_lock: */
1079 	struct mempolicy		*mempolicy;
1080 	short				il_prev;
1081 	short				pref_node_fork;
1082 #endif
1083 #ifdef CONFIG_NUMA_BALANCING
1084 	int				numa_scan_seq;
1085 	unsigned int			numa_scan_period;
1086 	unsigned int			numa_scan_period_max;
1087 	int				numa_preferred_nid;
1088 	unsigned long			numa_migrate_retry;
1089 	/* Migration stamp: */
1090 	u64				node_stamp;
1091 	u64				last_task_numa_placement;
1092 	u64				last_sum_exec_runtime;
1093 	struct callback_head		numa_work;
1094 
1095 	struct numa_group		*numa_group;
1096 
1097 	/*
1098 	 * numa_faults is an array split into four regions:
1099 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1100 	 * in this precise order.
1101 	 *
1102 	 * faults_memory: Exponential decaying average of faults on a per-node
1103 	 * basis. Scheduling placement decisions are made based on these
1104 	 * counts. The values remain static for the duration of a PTE scan.
1105 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1106 	 * hinting fault was incurred.
1107 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1108 	 * during the current scan window. When the scan completes, the counts
1109 	 * in faults_memory and faults_cpu decay and these values are copied.
1110 	 */
1111 	unsigned long			*numa_faults;
1112 	unsigned long			total_numa_faults;
1113 
1114 	/*
1115 	 * numa_faults_locality tracks if faults recorded during the last
1116 	 * scan window were remote/local or failed to migrate. The task scan
1117 	 * period is adapted based on the locality of the faults with different
1118 	 * weights depending on whether they were shared or private faults
1119 	 */
1120 	unsigned long			numa_faults_locality[3];
1121 
1122 	unsigned long			numa_pages_migrated;
1123 #endif /* CONFIG_NUMA_BALANCING */
1124 
1125 #ifdef CONFIG_RSEQ
1126 	struct rseq __user *rseq;
1127 	u32 rseq_sig;
1128 	/*
1129 	 * RmW on rseq_event_mask must be performed atomically
1130 	 * with respect to preemption.
1131 	 */
1132 	unsigned long rseq_event_mask;
1133 #endif
1134 
1135 	struct tlbflush_unmap_batch	tlb_ubc;
1136 
1137 	struct rcu_head			rcu;
1138 
1139 	/* Cache last used pipe for splice(): */
1140 	struct pipe_inode_info		*splice_pipe;
1141 
1142 	struct page_frag		task_frag;
1143 
1144 #ifdef CONFIG_TASK_DELAY_ACCT
1145 	struct task_delay_info		*delays;
1146 #endif
1147 
1148 #ifdef CONFIG_FAULT_INJECTION
1149 	int				make_it_fail;
1150 	unsigned int			fail_nth;
1151 #endif
1152 	/*
1153 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1154 	 * balance_dirty_pages() for a dirty throttling pause:
1155 	 */
1156 	int				nr_dirtied;
1157 	int				nr_dirtied_pause;
1158 	/* Start of a write-and-pause period: */
1159 	unsigned long			dirty_paused_when;
1160 
1161 #ifdef CONFIG_LATENCYTOP
1162 	int				latency_record_count;
1163 	struct latency_record		latency_record[LT_SAVECOUNT];
1164 #endif
1165 	/*
1166 	 * Time slack values; these are used to round up poll() and
1167 	 * select() etc timeout values. These are in nanoseconds.
1168 	 */
1169 	u64				timer_slack_ns;
1170 	u64				default_timer_slack_ns;
1171 
1172 #ifdef CONFIG_KASAN
1173 	unsigned int			kasan_depth;
1174 #endif
1175 
1176 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1177 	/* Index of current stored address in ret_stack: */
1178 	int				curr_ret_stack;
1179 	int				curr_ret_depth;
1180 
1181 	/* Stack of return addresses for return function tracing: */
1182 	struct ftrace_ret_stack		*ret_stack;
1183 
1184 	/* Timestamp for last schedule: */
1185 	unsigned long long		ftrace_timestamp;
1186 
1187 	/*
1188 	 * Number of functions that haven't been traced
1189 	 * because of depth overrun:
1190 	 */
1191 	atomic_t			trace_overrun;
1192 
1193 	/* Pause tracing: */
1194 	atomic_t			tracing_graph_pause;
1195 #endif
1196 
1197 #ifdef CONFIG_TRACING
1198 	/* State flags for use by tracers: */
1199 	unsigned long			trace;
1200 
1201 	/* Bitmask and counter of trace recursion: */
1202 	unsigned long			trace_recursion;
1203 #endif /* CONFIG_TRACING */
1204 
1205 #ifdef CONFIG_KCOV
1206 	/* Coverage collection mode enabled for this task (0 if disabled): */
1207 	unsigned int			kcov_mode;
1208 
1209 	/* Size of the kcov_area: */
1210 	unsigned int			kcov_size;
1211 
1212 	/* Buffer for coverage collection: */
1213 	void				*kcov_area;
1214 
1215 	/* KCOV descriptor wired with this task or NULL: */
1216 	struct kcov			*kcov;
1217 #endif
1218 
1219 #ifdef CONFIG_MEMCG
1220 	struct mem_cgroup		*memcg_in_oom;
1221 	gfp_t				memcg_oom_gfp_mask;
1222 	int				memcg_oom_order;
1223 
1224 	/* Number of pages to reclaim on returning to userland: */
1225 	unsigned int			memcg_nr_pages_over_high;
1226 
1227 	/* Used by memcontrol for targeted memcg charge: */
1228 	struct mem_cgroup		*active_memcg;
1229 #endif
1230 
1231 #ifdef CONFIG_BLK_CGROUP
1232 	struct request_queue		*throttle_queue;
1233 #endif
1234 
1235 #ifdef CONFIG_UPROBES
1236 	struct uprobe_task		*utask;
1237 #endif
1238 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1239 	unsigned int			sequential_io;
1240 	unsigned int			sequential_io_avg;
1241 #endif
1242 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1243 	unsigned long			task_state_change;
1244 #endif
1245 	int				pagefault_disabled;
1246 #ifdef CONFIG_MMU
1247 	struct task_struct		*oom_reaper_list;
1248 #endif
1249 #ifdef CONFIG_VMAP_STACK
1250 	struct vm_struct		*stack_vm_area;
1251 #endif
1252 #ifdef CONFIG_THREAD_INFO_IN_TASK
1253 	/* A live task holds one reference: */
1254 	refcount_t			stack_refcount;
1255 #endif
1256 #ifdef CONFIG_LIVEPATCH
1257 	int patch_state;
1258 #endif
1259 #ifdef CONFIG_SECURITY
1260 	/* Used by LSM modules for access restriction: */
1261 	void				*security;
1262 #endif
1263 
1264 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1265 	unsigned long			lowest_stack;
1266 	unsigned long			prev_lowest_stack;
1267 #endif
1268 
1269 	/*
1270 	 * New fields for task_struct should be added above here, so that
1271 	 * they are included in the randomized portion of task_struct.
1272 	 */
1273 	randomized_struct_fields_end
1274 
1275 	/* CPU-specific state of this task: */
1276 	struct thread_struct		thread;
1277 
1278 	/*
1279 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1280 	 * structure.  It *MUST* be at the end of 'task_struct'.
1281 	 *
1282 	 * Do not put anything below here!
1283 	 */
1284 };
1285 
1286 static inline struct pid *task_pid(struct task_struct *task)
1287 {
1288 	return task->thread_pid;
1289 }
1290 
1291 /*
1292  * the helpers to get the task's different pids as they are seen
1293  * from various namespaces
1294  *
1295  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1296  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1297  *                     current.
1298  * task_xid_nr_ns()  : id seen from the ns specified;
1299  *
1300  * see also pid_nr() etc in include/linux/pid.h
1301  */
1302 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1303 
1304 static inline pid_t task_pid_nr(struct task_struct *tsk)
1305 {
1306 	return tsk->pid;
1307 }
1308 
1309 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1310 {
1311 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1312 }
1313 
1314 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1315 {
1316 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1317 }
1318 
1319 
1320 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1321 {
1322 	return tsk->tgid;
1323 }
1324 
1325 /**
1326  * pid_alive - check that a task structure is not stale
1327  * @p: Task structure to be checked.
1328  *
1329  * Test if a process is not yet dead (at most zombie state)
1330  * If pid_alive fails, then pointers within the task structure
1331  * can be stale and must not be dereferenced.
1332  *
1333  * Return: 1 if the process is alive. 0 otherwise.
1334  */
1335 static inline int pid_alive(const struct task_struct *p)
1336 {
1337 	return p->thread_pid != NULL;
1338 }
1339 
1340 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1341 {
1342 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1343 }
1344 
1345 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1346 {
1347 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1348 }
1349 
1350 
1351 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1352 {
1353 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1354 }
1355 
1356 static inline pid_t task_session_vnr(struct task_struct *tsk)
1357 {
1358 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1359 }
1360 
1361 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1362 {
1363 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1364 }
1365 
1366 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1367 {
1368 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1369 }
1370 
1371 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1372 {
1373 	pid_t pid = 0;
1374 
1375 	rcu_read_lock();
1376 	if (pid_alive(tsk))
1377 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1378 	rcu_read_unlock();
1379 
1380 	return pid;
1381 }
1382 
1383 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1384 {
1385 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1386 }
1387 
1388 /* Obsolete, do not use: */
1389 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1390 {
1391 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1392 }
1393 
1394 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1395 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1396 
1397 static inline unsigned int task_state_index(struct task_struct *tsk)
1398 {
1399 	unsigned int tsk_state = READ_ONCE(tsk->state);
1400 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1401 
1402 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1403 
1404 	if (tsk_state == TASK_IDLE)
1405 		state = TASK_REPORT_IDLE;
1406 
1407 	return fls(state);
1408 }
1409 
1410 static inline char task_index_to_char(unsigned int state)
1411 {
1412 	static const char state_char[] = "RSDTtXZPI";
1413 
1414 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1415 
1416 	return state_char[state];
1417 }
1418 
1419 static inline char task_state_to_char(struct task_struct *tsk)
1420 {
1421 	return task_index_to_char(task_state_index(tsk));
1422 }
1423 
1424 /**
1425  * is_global_init - check if a task structure is init. Since init
1426  * is free to have sub-threads we need to check tgid.
1427  * @tsk: Task structure to be checked.
1428  *
1429  * Check if a task structure is the first user space task the kernel created.
1430  *
1431  * Return: 1 if the task structure is init. 0 otherwise.
1432  */
1433 static inline int is_global_init(struct task_struct *tsk)
1434 {
1435 	return task_tgid_nr(tsk) == 1;
1436 }
1437 
1438 extern struct pid *cad_pid;
1439 
1440 /*
1441  * Per process flags
1442  */
1443 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1444 #define PF_EXITING		0x00000004	/* Getting shut down */
1445 #define PF_EXITPIDONE		0x00000008	/* PI exit done on shut down */
1446 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1447 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1448 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1449 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1450 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1451 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1452 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1453 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1454 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1455 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1456 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1457 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1458 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1459 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1460 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1461 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1462 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1463 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1464 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1465 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1466 #define PF_MEMSTALL		0x01000000	/* Stalled due to lack of memory */
1467 #define PF_UMH			0x02000000	/* I'm an Usermodehelper process */
1468 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1469 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1470 #define PF_MEMALLOC_NOCMA	0x10000000	/* All allocation request will have _GFP_MOVABLE cleared */
1471 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1472 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1473 
1474 /*
1475  * Only the _current_ task can read/write to tsk->flags, but other
1476  * tasks can access tsk->flags in readonly mode for example
1477  * with tsk_used_math (like during threaded core dumping).
1478  * There is however an exception to this rule during ptrace
1479  * or during fork: the ptracer task is allowed to write to the
1480  * child->flags of its traced child (same goes for fork, the parent
1481  * can write to the child->flags), because we're guaranteed the
1482  * child is not running and in turn not changing child->flags
1483  * at the same time the parent does it.
1484  */
1485 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1486 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1487 #define clear_used_math()			clear_stopped_child_used_math(current)
1488 #define set_used_math()				set_stopped_child_used_math(current)
1489 
1490 #define conditional_stopped_child_used_math(condition, child) \
1491 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1492 
1493 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1494 
1495 #define copy_to_stopped_child_used_math(child) \
1496 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1497 
1498 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1499 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1500 #define used_math()				tsk_used_math(current)
1501 
1502 static inline bool is_percpu_thread(void)
1503 {
1504 #ifdef CONFIG_SMP
1505 	return (current->flags & PF_NO_SETAFFINITY) &&
1506 		(current->nr_cpus_allowed  == 1);
1507 #else
1508 	return true;
1509 #endif
1510 }
1511 
1512 /* Per-process atomic flags. */
1513 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1514 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1515 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1516 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1517 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1518 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1519 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1520 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1521 
1522 #define TASK_PFA_TEST(name, func)					\
1523 	static inline bool task_##func(struct task_struct *p)		\
1524 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1525 
1526 #define TASK_PFA_SET(name, func)					\
1527 	static inline void task_set_##func(struct task_struct *p)	\
1528 	{ set_bit(PFA_##name, &p->atomic_flags); }
1529 
1530 #define TASK_PFA_CLEAR(name, func)					\
1531 	static inline void task_clear_##func(struct task_struct *p)	\
1532 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1533 
1534 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1535 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1536 
1537 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1538 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1539 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1540 
1541 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1542 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1543 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1544 
1545 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1546 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1547 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1548 
1549 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1550 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1551 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1552 
1553 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1554 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1555 
1556 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1557 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1558 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1559 
1560 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1561 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1562 
1563 static inline void
1564 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1565 {
1566 	current->flags &= ~flags;
1567 	current->flags |= orig_flags & flags;
1568 }
1569 
1570 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1571 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1572 #ifdef CONFIG_SMP
1573 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1574 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1575 #else
1576 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1577 {
1578 }
1579 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1580 {
1581 	if (!cpumask_test_cpu(0, new_mask))
1582 		return -EINVAL;
1583 	return 0;
1584 }
1585 #endif
1586 
1587 extern int yield_to(struct task_struct *p, bool preempt);
1588 extern void set_user_nice(struct task_struct *p, long nice);
1589 extern int task_prio(const struct task_struct *p);
1590 
1591 /**
1592  * task_nice - return the nice value of a given task.
1593  * @p: the task in question.
1594  *
1595  * Return: The nice value [ -20 ... 0 ... 19 ].
1596  */
1597 static inline int task_nice(const struct task_struct *p)
1598 {
1599 	return PRIO_TO_NICE((p)->static_prio);
1600 }
1601 
1602 extern int can_nice(const struct task_struct *p, const int nice);
1603 extern int task_curr(const struct task_struct *p);
1604 extern int idle_cpu(int cpu);
1605 extern int available_idle_cpu(int cpu);
1606 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1607 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1608 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1609 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1610 extern struct task_struct *idle_task(int cpu);
1611 
1612 /**
1613  * is_idle_task - is the specified task an idle task?
1614  * @p: the task in question.
1615  *
1616  * Return: 1 if @p is an idle task. 0 otherwise.
1617  */
1618 static inline bool is_idle_task(const struct task_struct *p)
1619 {
1620 	return !!(p->flags & PF_IDLE);
1621 }
1622 
1623 extern struct task_struct *curr_task(int cpu);
1624 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1625 
1626 void yield(void);
1627 
1628 union thread_union {
1629 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1630 	struct task_struct task;
1631 #endif
1632 #ifndef CONFIG_THREAD_INFO_IN_TASK
1633 	struct thread_info thread_info;
1634 #endif
1635 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1636 };
1637 
1638 #ifndef CONFIG_THREAD_INFO_IN_TASK
1639 extern struct thread_info init_thread_info;
1640 #endif
1641 
1642 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1643 
1644 #ifdef CONFIG_THREAD_INFO_IN_TASK
1645 static inline struct thread_info *task_thread_info(struct task_struct *task)
1646 {
1647 	return &task->thread_info;
1648 }
1649 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1650 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1651 #endif
1652 
1653 /*
1654  * find a task by one of its numerical ids
1655  *
1656  * find_task_by_pid_ns():
1657  *      finds a task by its pid in the specified namespace
1658  * find_task_by_vpid():
1659  *      finds a task by its virtual pid
1660  *
1661  * see also find_vpid() etc in include/linux/pid.h
1662  */
1663 
1664 extern struct task_struct *find_task_by_vpid(pid_t nr);
1665 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1666 
1667 /*
1668  * find a task by its virtual pid and get the task struct
1669  */
1670 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1671 
1672 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1673 extern int wake_up_process(struct task_struct *tsk);
1674 extern void wake_up_new_task(struct task_struct *tsk);
1675 
1676 #ifdef CONFIG_SMP
1677 extern void kick_process(struct task_struct *tsk);
1678 #else
1679 static inline void kick_process(struct task_struct *tsk) { }
1680 #endif
1681 
1682 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1683 
1684 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1685 {
1686 	__set_task_comm(tsk, from, false);
1687 }
1688 
1689 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1690 #define get_task_comm(buf, tsk) ({			\
1691 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1692 	__get_task_comm(buf, sizeof(buf), tsk);		\
1693 })
1694 
1695 #ifdef CONFIG_SMP
1696 void scheduler_ipi(void);
1697 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1698 #else
1699 static inline void scheduler_ipi(void) { }
1700 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1701 {
1702 	return 1;
1703 }
1704 #endif
1705 
1706 /*
1707  * Set thread flags in other task's structures.
1708  * See asm/thread_info.h for TIF_xxxx flags available:
1709  */
1710 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1711 {
1712 	set_ti_thread_flag(task_thread_info(tsk), flag);
1713 }
1714 
1715 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1716 {
1717 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1718 }
1719 
1720 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1721 					  bool value)
1722 {
1723 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1724 }
1725 
1726 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1727 {
1728 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1729 }
1730 
1731 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1732 {
1733 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1734 }
1735 
1736 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1737 {
1738 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1739 }
1740 
1741 static inline void set_tsk_need_resched(struct task_struct *tsk)
1742 {
1743 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1744 }
1745 
1746 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1747 {
1748 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1749 }
1750 
1751 static inline int test_tsk_need_resched(struct task_struct *tsk)
1752 {
1753 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1754 }
1755 
1756 /*
1757  * cond_resched() and cond_resched_lock(): latency reduction via
1758  * explicit rescheduling in places that are safe. The return
1759  * value indicates whether a reschedule was done in fact.
1760  * cond_resched_lock() will drop the spinlock before scheduling,
1761  */
1762 #ifndef CONFIG_PREEMPT
1763 extern int _cond_resched(void);
1764 #else
1765 static inline int _cond_resched(void) { return 0; }
1766 #endif
1767 
1768 #define cond_resched() ({			\
1769 	___might_sleep(__FILE__, __LINE__, 0);	\
1770 	_cond_resched();			\
1771 })
1772 
1773 extern int __cond_resched_lock(spinlock_t *lock);
1774 
1775 #define cond_resched_lock(lock) ({				\
1776 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1777 	__cond_resched_lock(lock);				\
1778 })
1779 
1780 static inline void cond_resched_rcu(void)
1781 {
1782 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1783 	rcu_read_unlock();
1784 	cond_resched();
1785 	rcu_read_lock();
1786 #endif
1787 }
1788 
1789 /*
1790  * Does a critical section need to be broken due to another
1791  * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1792  * but a general need for low latency)
1793  */
1794 static inline int spin_needbreak(spinlock_t *lock)
1795 {
1796 #ifdef CONFIG_PREEMPT
1797 	return spin_is_contended(lock);
1798 #else
1799 	return 0;
1800 #endif
1801 }
1802 
1803 static __always_inline bool need_resched(void)
1804 {
1805 	return unlikely(tif_need_resched());
1806 }
1807 
1808 /*
1809  * Wrappers for p->thread_info->cpu access. No-op on UP.
1810  */
1811 #ifdef CONFIG_SMP
1812 
1813 static inline unsigned int task_cpu(const struct task_struct *p)
1814 {
1815 #ifdef CONFIG_THREAD_INFO_IN_TASK
1816 	return READ_ONCE(p->cpu);
1817 #else
1818 	return READ_ONCE(task_thread_info(p)->cpu);
1819 #endif
1820 }
1821 
1822 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1823 
1824 #else
1825 
1826 static inline unsigned int task_cpu(const struct task_struct *p)
1827 {
1828 	return 0;
1829 }
1830 
1831 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1832 {
1833 }
1834 
1835 #endif /* CONFIG_SMP */
1836 
1837 /*
1838  * In order to reduce various lock holder preemption latencies provide an
1839  * interface to see if a vCPU is currently running or not.
1840  *
1841  * This allows us to terminate optimistic spin loops and block, analogous to
1842  * the native optimistic spin heuristic of testing if the lock owner task is
1843  * running or not.
1844  */
1845 #ifndef vcpu_is_preempted
1846 # define vcpu_is_preempted(cpu)	false
1847 #endif
1848 
1849 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1850 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1851 
1852 #ifndef TASK_SIZE_OF
1853 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1854 #endif
1855 
1856 #ifdef CONFIG_RSEQ
1857 
1858 /*
1859  * Map the event mask on the user-space ABI enum rseq_cs_flags
1860  * for direct mask checks.
1861  */
1862 enum rseq_event_mask_bits {
1863 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1864 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1865 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1866 };
1867 
1868 enum rseq_event_mask {
1869 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
1870 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
1871 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
1872 };
1873 
1874 static inline void rseq_set_notify_resume(struct task_struct *t)
1875 {
1876 	if (t->rseq)
1877 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1878 }
1879 
1880 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1881 
1882 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1883 					     struct pt_regs *regs)
1884 {
1885 	if (current->rseq)
1886 		__rseq_handle_notify_resume(ksig, regs);
1887 }
1888 
1889 static inline void rseq_signal_deliver(struct ksignal *ksig,
1890 				       struct pt_regs *regs)
1891 {
1892 	preempt_disable();
1893 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1894 	preempt_enable();
1895 	rseq_handle_notify_resume(ksig, regs);
1896 }
1897 
1898 /* rseq_preempt() requires preemption to be disabled. */
1899 static inline void rseq_preempt(struct task_struct *t)
1900 {
1901 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1902 	rseq_set_notify_resume(t);
1903 }
1904 
1905 /* rseq_migrate() requires preemption to be disabled. */
1906 static inline void rseq_migrate(struct task_struct *t)
1907 {
1908 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1909 	rseq_set_notify_resume(t);
1910 }
1911 
1912 /*
1913  * If parent process has a registered restartable sequences area, the
1914  * child inherits. Only applies when forking a process, not a thread.
1915  */
1916 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1917 {
1918 	if (clone_flags & CLONE_THREAD) {
1919 		t->rseq = NULL;
1920 		t->rseq_sig = 0;
1921 		t->rseq_event_mask = 0;
1922 	} else {
1923 		t->rseq = current->rseq;
1924 		t->rseq_sig = current->rseq_sig;
1925 		t->rseq_event_mask = current->rseq_event_mask;
1926 	}
1927 }
1928 
1929 static inline void rseq_execve(struct task_struct *t)
1930 {
1931 	t->rseq = NULL;
1932 	t->rseq_sig = 0;
1933 	t->rseq_event_mask = 0;
1934 }
1935 
1936 #else
1937 
1938 static inline void rseq_set_notify_resume(struct task_struct *t)
1939 {
1940 }
1941 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1942 					     struct pt_regs *regs)
1943 {
1944 }
1945 static inline void rseq_signal_deliver(struct ksignal *ksig,
1946 				       struct pt_regs *regs)
1947 {
1948 }
1949 static inline void rseq_preempt(struct task_struct *t)
1950 {
1951 }
1952 static inline void rseq_migrate(struct task_struct *t)
1953 {
1954 }
1955 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1956 {
1957 }
1958 static inline void rseq_execve(struct task_struct *t)
1959 {
1960 }
1961 
1962 #endif
1963 
1964 void __exit_umh(struct task_struct *tsk);
1965 
1966 static inline void exit_umh(struct task_struct *tsk)
1967 {
1968 	if (unlikely(tsk->flags & PF_UMH))
1969 		__exit_umh(tsk);
1970 }
1971 
1972 #ifdef CONFIG_DEBUG_RSEQ
1973 
1974 void rseq_syscall(struct pt_regs *regs);
1975 
1976 #else
1977 
1978 static inline void rseq_syscall(struct pt_regs *regs)
1979 {
1980 }
1981 
1982 #endif
1983 
1984 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
1985 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
1986 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
1987 
1988 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
1989 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
1990 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
1991 
1992 int sched_trace_rq_cpu(struct rq *rq);
1993 
1994 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
1995 
1996 #endif
1997