xref: /linux/include/linux/sched.h (revision e3c9fc78f096b83e81329b213c25fb9a376e373a)
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/mutex.h>
18 #include <linux/plist.h>
19 #include <linux/hrtimer.h>
20 #include <linux/irqflags.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/syscall_user_dispatch.h>
31 #include <linux/mm_types_task.h>
32 #include <linux/task_io_accounting.h>
33 #include <linux/posix-timers.h>
34 #include <linux/rseq.h>
35 #include <linux/seqlock.h>
36 #include <linux/kcsan.h>
37 #include <linux/rv.h>
38 #include <asm/kmap_size.h>
39 
40 /* task_struct member predeclarations (sorted alphabetically): */
41 struct audit_context;
42 struct backing_dev_info;
43 struct bio_list;
44 struct blk_plug;
45 struct bpf_local_storage;
46 struct bpf_run_ctx;
47 struct capture_control;
48 struct cfs_rq;
49 struct fs_struct;
50 struct futex_pi_state;
51 struct io_context;
52 struct io_uring_task;
53 struct mempolicy;
54 struct nameidata;
55 struct nsproxy;
56 struct perf_event_context;
57 struct pid_namespace;
58 struct pipe_inode_info;
59 struct rcu_node;
60 struct reclaim_state;
61 struct robust_list_head;
62 struct root_domain;
63 struct rq;
64 struct sched_attr;
65 struct sched_param;
66 struct seq_file;
67 struct sighand_struct;
68 struct signal_struct;
69 struct task_delay_info;
70 struct task_group;
71 
72 /*
73  * Task state bitmask. NOTE! These bits are also
74  * encoded in fs/proc/array.c: get_task_state().
75  *
76  * We have two separate sets of flags: task->state
77  * is about runnability, while task->exit_state are
78  * about the task exiting. Confusing, but this way
79  * modifying one set can't modify the other one by
80  * mistake.
81  */
82 
83 /* Used in tsk->state: */
84 #define TASK_RUNNING			0x0000
85 #define TASK_INTERRUPTIBLE		0x0001
86 #define TASK_UNINTERRUPTIBLE		0x0002
87 #define __TASK_STOPPED			0x0004
88 #define __TASK_TRACED			0x0008
89 /* Used in tsk->exit_state: */
90 #define EXIT_DEAD			0x0010
91 #define EXIT_ZOMBIE			0x0020
92 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
93 /* Used in tsk->state again: */
94 #define TASK_PARKED			0x0040
95 #define TASK_DEAD			0x0080
96 #define TASK_WAKEKILL			0x0100
97 #define TASK_WAKING			0x0200
98 #define TASK_NOLOAD			0x0400
99 #define TASK_NEW			0x0800
100 /* RT specific auxilliary flag to mark RT lock waiters */
101 #define TASK_RTLOCK_WAIT		0x1000
102 #define TASK_STATE_MAX			0x2000
103 
104 /* Convenience macros for the sake of set_current_state: */
105 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
106 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
107 #define TASK_TRACED			__TASK_TRACED
108 
109 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
110 
111 /* Convenience macros for the sake of wake_up(): */
112 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
113 
114 /* get_task_state(): */
115 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
116 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
117 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
118 					 TASK_PARKED)
119 
120 #define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
121 
122 #define task_is_traced(task)		((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
123 #define task_is_stopped(task)		((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
124 #define task_is_stopped_or_traced(task)	((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
125 
126 /*
127  * Special states are those that do not use the normal wait-loop pattern. See
128  * the comment with set_special_state().
129  */
130 #define is_special_task_state(state)				\
131 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
132 
133 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
134 # define debug_normal_state_change(state_value)				\
135 	do {								\
136 		WARN_ON_ONCE(is_special_task_state(state_value));	\
137 		current->task_state_change = _THIS_IP_;			\
138 	} while (0)
139 
140 # define debug_special_state_change(state_value)			\
141 	do {								\
142 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
143 		current->task_state_change = _THIS_IP_;			\
144 	} while (0)
145 
146 # define debug_rtlock_wait_set_state()					\
147 	do {								 \
148 		current->saved_state_change = current->task_state_change;\
149 		current->task_state_change = _THIS_IP_;			 \
150 	} while (0)
151 
152 # define debug_rtlock_wait_restore_state()				\
153 	do {								 \
154 		current->task_state_change = current->saved_state_change;\
155 	} while (0)
156 
157 #else
158 # define debug_normal_state_change(cond)	do { } while (0)
159 # define debug_special_state_change(cond)	do { } while (0)
160 # define debug_rtlock_wait_set_state()		do { } while (0)
161 # define debug_rtlock_wait_restore_state()	do { } while (0)
162 #endif
163 
164 /*
165  * set_current_state() includes a barrier so that the write of current->state
166  * is correctly serialised wrt the caller's subsequent test of whether to
167  * actually sleep:
168  *
169  *   for (;;) {
170  *	set_current_state(TASK_UNINTERRUPTIBLE);
171  *	if (CONDITION)
172  *	   break;
173  *
174  *	schedule();
175  *   }
176  *   __set_current_state(TASK_RUNNING);
177  *
178  * If the caller does not need such serialisation (because, for instance, the
179  * CONDITION test and condition change and wakeup are under the same lock) then
180  * use __set_current_state().
181  *
182  * The above is typically ordered against the wakeup, which does:
183  *
184  *   CONDITION = 1;
185  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
186  *
187  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
188  * accessing p->state.
189  *
190  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
191  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
192  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
193  *
194  * However, with slightly different timing the wakeup TASK_RUNNING store can
195  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
196  * a problem either because that will result in one extra go around the loop
197  * and our @cond test will save the day.
198  *
199  * Also see the comments of try_to_wake_up().
200  */
201 #define __set_current_state(state_value)				\
202 	do {								\
203 		debug_normal_state_change((state_value));		\
204 		WRITE_ONCE(current->__state, (state_value));		\
205 	} while (0)
206 
207 #define set_current_state(state_value)					\
208 	do {								\
209 		debug_normal_state_change((state_value));		\
210 		smp_store_mb(current->__state, (state_value));		\
211 	} while (0)
212 
213 /*
214  * set_special_state() should be used for those states when the blocking task
215  * can not use the regular condition based wait-loop. In that case we must
216  * serialize against wakeups such that any possible in-flight TASK_RUNNING
217  * stores will not collide with our state change.
218  */
219 #define set_special_state(state_value)					\
220 	do {								\
221 		unsigned long flags; /* may shadow */			\
222 									\
223 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
224 		debug_special_state_change((state_value));		\
225 		WRITE_ONCE(current->__state, (state_value));		\
226 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
227 	} while (0)
228 
229 /*
230  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
231  *
232  * RT's spin/rwlock substitutions are state preserving. The state of the
233  * task when blocking on the lock is saved in task_struct::saved_state and
234  * restored after the lock has been acquired.  These operations are
235  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
236  * lock related wakeups while the task is blocked on the lock are
237  * redirected to operate on task_struct::saved_state to ensure that these
238  * are not dropped. On restore task_struct::saved_state is set to
239  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
240  *
241  * The lock operation looks like this:
242  *
243  *	current_save_and_set_rtlock_wait_state();
244  *	for (;;) {
245  *		if (try_lock())
246  *			break;
247  *		raw_spin_unlock_irq(&lock->wait_lock);
248  *		schedule_rtlock();
249  *		raw_spin_lock_irq(&lock->wait_lock);
250  *		set_current_state(TASK_RTLOCK_WAIT);
251  *	}
252  *	current_restore_rtlock_saved_state();
253  */
254 #define current_save_and_set_rtlock_wait_state()			\
255 	do {								\
256 		lockdep_assert_irqs_disabled();				\
257 		raw_spin_lock(&current->pi_lock);			\
258 		current->saved_state = current->__state;		\
259 		debug_rtlock_wait_set_state();				\
260 		WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);		\
261 		raw_spin_unlock(&current->pi_lock);			\
262 	} while (0);
263 
264 #define current_restore_rtlock_saved_state()				\
265 	do {								\
266 		lockdep_assert_irqs_disabled();				\
267 		raw_spin_lock(&current->pi_lock);			\
268 		debug_rtlock_wait_restore_state();			\
269 		WRITE_ONCE(current->__state, current->saved_state);	\
270 		current->saved_state = TASK_RUNNING;			\
271 		raw_spin_unlock(&current->pi_lock);			\
272 	} while (0);
273 
274 #define get_current_state()	READ_ONCE(current->__state)
275 
276 /*
277  * Define the task command name length as enum, then it can be visible to
278  * BPF programs.
279  */
280 enum {
281 	TASK_COMM_LEN = 16,
282 };
283 
284 extern void scheduler_tick(void);
285 
286 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
287 
288 extern long schedule_timeout(long timeout);
289 extern long schedule_timeout_interruptible(long timeout);
290 extern long schedule_timeout_killable(long timeout);
291 extern long schedule_timeout_uninterruptible(long timeout);
292 extern long schedule_timeout_idle(long timeout);
293 asmlinkage void schedule(void);
294 extern void schedule_preempt_disabled(void);
295 asmlinkage void preempt_schedule_irq(void);
296 #ifdef CONFIG_PREEMPT_RT
297  extern void schedule_rtlock(void);
298 #endif
299 
300 extern int __must_check io_schedule_prepare(void);
301 extern void io_schedule_finish(int token);
302 extern long io_schedule_timeout(long timeout);
303 extern void io_schedule(void);
304 
305 /**
306  * struct prev_cputime - snapshot of system and user cputime
307  * @utime: time spent in user mode
308  * @stime: time spent in system mode
309  * @lock: protects the above two fields
310  *
311  * Stores previous user/system time values such that we can guarantee
312  * monotonicity.
313  */
314 struct prev_cputime {
315 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
316 	u64				utime;
317 	u64				stime;
318 	raw_spinlock_t			lock;
319 #endif
320 };
321 
322 enum vtime_state {
323 	/* Task is sleeping or running in a CPU with VTIME inactive: */
324 	VTIME_INACTIVE = 0,
325 	/* Task is idle */
326 	VTIME_IDLE,
327 	/* Task runs in kernelspace in a CPU with VTIME active: */
328 	VTIME_SYS,
329 	/* Task runs in userspace in a CPU with VTIME active: */
330 	VTIME_USER,
331 	/* Task runs as guests in a CPU with VTIME active: */
332 	VTIME_GUEST,
333 };
334 
335 struct vtime {
336 	seqcount_t		seqcount;
337 	unsigned long long	starttime;
338 	enum vtime_state	state;
339 	unsigned int		cpu;
340 	u64			utime;
341 	u64			stime;
342 	u64			gtime;
343 };
344 
345 /*
346  * Utilization clamp constraints.
347  * @UCLAMP_MIN:	Minimum utilization
348  * @UCLAMP_MAX:	Maximum utilization
349  * @UCLAMP_CNT:	Utilization clamp constraints count
350  */
351 enum uclamp_id {
352 	UCLAMP_MIN = 0,
353 	UCLAMP_MAX,
354 	UCLAMP_CNT
355 };
356 
357 #ifdef CONFIG_SMP
358 extern struct root_domain def_root_domain;
359 extern struct mutex sched_domains_mutex;
360 #endif
361 
362 struct sched_info {
363 #ifdef CONFIG_SCHED_INFO
364 	/* Cumulative counters: */
365 
366 	/* # of times we have run on this CPU: */
367 	unsigned long			pcount;
368 
369 	/* Time spent waiting on a runqueue: */
370 	unsigned long long		run_delay;
371 
372 	/* Timestamps: */
373 
374 	/* When did we last run on a CPU? */
375 	unsigned long long		last_arrival;
376 
377 	/* When were we last queued to run? */
378 	unsigned long long		last_queued;
379 
380 #endif /* CONFIG_SCHED_INFO */
381 };
382 
383 /*
384  * Integer metrics need fixed point arithmetic, e.g., sched/fair
385  * has a few: load, load_avg, util_avg, freq, and capacity.
386  *
387  * We define a basic fixed point arithmetic range, and then formalize
388  * all these metrics based on that basic range.
389  */
390 # define SCHED_FIXEDPOINT_SHIFT		10
391 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
392 
393 /* Increase resolution of cpu_capacity calculations */
394 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
395 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
396 
397 struct load_weight {
398 	unsigned long			weight;
399 	u32				inv_weight;
400 };
401 
402 /**
403  * struct util_est - Estimation utilization of FAIR tasks
404  * @enqueued: instantaneous estimated utilization of a task/cpu
405  * @ewma:     the Exponential Weighted Moving Average (EWMA)
406  *            utilization of a task
407  *
408  * Support data structure to track an Exponential Weighted Moving Average
409  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
410  * average each time a task completes an activation. Sample's weight is chosen
411  * so that the EWMA will be relatively insensitive to transient changes to the
412  * task's workload.
413  *
414  * The enqueued attribute has a slightly different meaning for tasks and cpus:
415  * - task:   the task's util_avg at last task dequeue time
416  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
417  * Thus, the util_est.enqueued of a task represents the contribution on the
418  * estimated utilization of the CPU where that task is currently enqueued.
419  *
420  * Only for tasks we track a moving average of the past instantaneous
421  * estimated utilization. This allows to absorb sporadic drops in utilization
422  * of an otherwise almost periodic task.
423  *
424  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
425  * updates. When a task is dequeued, its util_est should not be updated if its
426  * util_avg has not been updated in the meantime.
427  * This information is mapped into the MSB bit of util_est.enqueued at dequeue
428  * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
429  * for a task) it is safe to use MSB.
430  */
431 struct util_est {
432 	unsigned int			enqueued;
433 	unsigned int			ewma;
434 #define UTIL_EST_WEIGHT_SHIFT		2
435 #define UTIL_AVG_UNCHANGED		0x80000000
436 } __attribute__((__aligned__(sizeof(u64))));
437 
438 /*
439  * The load/runnable/util_avg accumulates an infinite geometric series
440  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
441  *
442  * [load_avg definition]
443  *
444  *   load_avg = runnable% * scale_load_down(load)
445  *
446  * [runnable_avg definition]
447  *
448  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
449  *
450  * [util_avg definition]
451  *
452  *   util_avg = running% * SCHED_CAPACITY_SCALE
453  *
454  * where runnable% is the time ratio that a sched_entity is runnable and
455  * running% the time ratio that a sched_entity is running.
456  *
457  * For cfs_rq, they are the aggregated values of all runnable and blocked
458  * sched_entities.
459  *
460  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
461  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
462  * for computing those signals (see update_rq_clock_pelt())
463  *
464  * N.B., the above ratios (runnable% and running%) themselves are in the
465  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
466  * to as large a range as necessary. This is for example reflected by
467  * util_avg's SCHED_CAPACITY_SCALE.
468  *
469  * [Overflow issue]
470  *
471  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
472  * with the highest load (=88761), always runnable on a single cfs_rq,
473  * and should not overflow as the number already hits PID_MAX_LIMIT.
474  *
475  * For all other cases (including 32-bit kernels), struct load_weight's
476  * weight will overflow first before we do, because:
477  *
478  *    Max(load_avg) <= Max(load.weight)
479  *
480  * Then it is the load_weight's responsibility to consider overflow
481  * issues.
482  */
483 struct sched_avg {
484 	u64				last_update_time;
485 	u64				load_sum;
486 	u64				runnable_sum;
487 	u32				util_sum;
488 	u32				period_contrib;
489 	unsigned long			load_avg;
490 	unsigned long			runnable_avg;
491 	unsigned long			util_avg;
492 	struct util_est			util_est;
493 } ____cacheline_aligned;
494 
495 struct sched_statistics {
496 #ifdef CONFIG_SCHEDSTATS
497 	u64				wait_start;
498 	u64				wait_max;
499 	u64				wait_count;
500 	u64				wait_sum;
501 	u64				iowait_count;
502 	u64				iowait_sum;
503 
504 	u64				sleep_start;
505 	u64				sleep_max;
506 	s64				sum_sleep_runtime;
507 
508 	u64				block_start;
509 	u64				block_max;
510 	s64				sum_block_runtime;
511 
512 	u64				exec_max;
513 	u64				slice_max;
514 
515 	u64				nr_migrations_cold;
516 	u64				nr_failed_migrations_affine;
517 	u64				nr_failed_migrations_running;
518 	u64				nr_failed_migrations_hot;
519 	u64				nr_forced_migrations;
520 
521 	u64				nr_wakeups;
522 	u64				nr_wakeups_sync;
523 	u64				nr_wakeups_migrate;
524 	u64				nr_wakeups_local;
525 	u64				nr_wakeups_remote;
526 	u64				nr_wakeups_affine;
527 	u64				nr_wakeups_affine_attempts;
528 	u64				nr_wakeups_passive;
529 	u64				nr_wakeups_idle;
530 
531 #ifdef CONFIG_SCHED_CORE
532 	u64				core_forceidle_sum;
533 #endif
534 #endif /* CONFIG_SCHEDSTATS */
535 } ____cacheline_aligned;
536 
537 struct sched_entity {
538 	/* For load-balancing: */
539 	struct load_weight		load;
540 	struct rb_node			run_node;
541 	struct list_head		group_node;
542 	unsigned int			on_rq;
543 
544 	u64				exec_start;
545 	u64				sum_exec_runtime;
546 	u64				vruntime;
547 	u64				prev_sum_exec_runtime;
548 
549 	u64				nr_migrations;
550 
551 #ifdef CONFIG_FAIR_GROUP_SCHED
552 	int				depth;
553 	struct sched_entity		*parent;
554 	/* rq on which this entity is (to be) queued: */
555 	struct cfs_rq			*cfs_rq;
556 	/* rq "owned" by this entity/group: */
557 	struct cfs_rq			*my_q;
558 	/* cached value of my_q->h_nr_running */
559 	unsigned long			runnable_weight;
560 #endif
561 
562 #ifdef CONFIG_SMP
563 	/*
564 	 * Per entity load average tracking.
565 	 *
566 	 * Put into separate cache line so it does not
567 	 * collide with read-mostly values above.
568 	 */
569 	struct sched_avg		avg;
570 #endif
571 };
572 
573 struct sched_rt_entity {
574 	struct list_head		run_list;
575 	unsigned long			timeout;
576 	unsigned long			watchdog_stamp;
577 	unsigned int			time_slice;
578 	unsigned short			on_rq;
579 	unsigned short			on_list;
580 
581 	struct sched_rt_entity		*back;
582 #ifdef CONFIG_RT_GROUP_SCHED
583 	struct sched_rt_entity		*parent;
584 	/* rq on which this entity is (to be) queued: */
585 	struct rt_rq			*rt_rq;
586 	/* rq "owned" by this entity/group: */
587 	struct rt_rq			*my_q;
588 #endif
589 } __randomize_layout;
590 
591 struct sched_dl_entity {
592 	struct rb_node			rb_node;
593 
594 	/*
595 	 * Original scheduling parameters. Copied here from sched_attr
596 	 * during sched_setattr(), they will remain the same until
597 	 * the next sched_setattr().
598 	 */
599 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
600 	u64				dl_deadline;	/* Relative deadline of each instance	*/
601 	u64				dl_period;	/* Separation of two instances (period) */
602 	u64				dl_bw;		/* dl_runtime / dl_period		*/
603 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
604 
605 	/*
606 	 * Actual scheduling parameters. Initialized with the values above,
607 	 * they are continuously updated during task execution. Note that
608 	 * the remaining runtime could be < 0 in case we are in overrun.
609 	 */
610 	s64				runtime;	/* Remaining runtime for this instance	*/
611 	u64				deadline;	/* Absolute deadline for this instance	*/
612 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
613 
614 	/*
615 	 * Some bool flags:
616 	 *
617 	 * @dl_throttled tells if we exhausted the runtime. If so, the
618 	 * task has to wait for a replenishment to be performed at the
619 	 * next firing of dl_timer.
620 	 *
621 	 * @dl_yielded tells if task gave up the CPU before consuming
622 	 * all its available runtime during the last job.
623 	 *
624 	 * @dl_non_contending tells if the task is inactive while still
625 	 * contributing to the active utilization. In other words, it
626 	 * indicates if the inactive timer has been armed and its handler
627 	 * has not been executed yet. This flag is useful to avoid race
628 	 * conditions between the inactive timer handler and the wakeup
629 	 * code.
630 	 *
631 	 * @dl_overrun tells if the task asked to be informed about runtime
632 	 * overruns.
633 	 */
634 	unsigned int			dl_throttled      : 1;
635 	unsigned int			dl_yielded        : 1;
636 	unsigned int			dl_non_contending : 1;
637 	unsigned int			dl_overrun	  : 1;
638 
639 	/*
640 	 * Bandwidth enforcement timer. Each -deadline task has its
641 	 * own bandwidth to be enforced, thus we need one timer per task.
642 	 */
643 	struct hrtimer			dl_timer;
644 
645 	/*
646 	 * Inactive timer, responsible for decreasing the active utilization
647 	 * at the "0-lag time". When a -deadline task blocks, it contributes
648 	 * to GRUB's active utilization until the "0-lag time", hence a
649 	 * timer is needed to decrease the active utilization at the correct
650 	 * time.
651 	 */
652 	struct hrtimer inactive_timer;
653 
654 #ifdef CONFIG_RT_MUTEXES
655 	/*
656 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
657 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
658 	 * to (the original one/itself).
659 	 */
660 	struct sched_dl_entity *pi_se;
661 #endif
662 };
663 
664 #ifdef CONFIG_UCLAMP_TASK
665 /* Number of utilization clamp buckets (shorter alias) */
666 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
667 
668 /*
669  * Utilization clamp for a scheduling entity
670  * @value:		clamp value "assigned" to a se
671  * @bucket_id:		bucket index corresponding to the "assigned" value
672  * @active:		the se is currently refcounted in a rq's bucket
673  * @user_defined:	the requested clamp value comes from user-space
674  *
675  * The bucket_id is the index of the clamp bucket matching the clamp value
676  * which is pre-computed and stored to avoid expensive integer divisions from
677  * the fast path.
678  *
679  * The active bit is set whenever a task has got an "effective" value assigned,
680  * which can be different from the clamp value "requested" from user-space.
681  * This allows to know a task is refcounted in the rq's bucket corresponding
682  * to the "effective" bucket_id.
683  *
684  * The user_defined bit is set whenever a task has got a task-specific clamp
685  * value requested from userspace, i.e. the system defaults apply to this task
686  * just as a restriction. This allows to relax default clamps when a less
687  * restrictive task-specific value has been requested, thus allowing to
688  * implement a "nice" semantic. For example, a task running with a 20%
689  * default boost can still drop its own boosting to 0%.
690  */
691 struct uclamp_se {
692 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
693 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
694 	unsigned int active		: 1;
695 	unsigned int user_defined	: 1;
696 };
697 #endif /* CONFIG_UCLAMP_TASK */
698 
699 union rcu_special {
700 	struct {
701 		u8			blocked;
702 		u8			need_qs;
703 		u8			exp_hint; /* Hint for performance. */
704 		u8			need_mb; /* Readers need smp_mb(). */
705 	} b; /* Bits. */
706 	u32 s; /* Set of bits. */
707 };
708 
709 enum perf_event_task_context {
710 	perf_invalid_context = -1,
711 	perf_hw_context = 0,
712 	perf_sw_context,
713 	perf_nr_task_contexts,
714 };
715 
716 struct wake_q_node {
717 	struct wake_q_node *next;
718 };
719 
720 struct kmap_ctrl {
721 #ifdef CONFIG_KMAP_LOCAL
722 	int				idx;
723 	pte_t				pteval[KM_MAX_IDX];
724 #endif
725 };
726 
727 struct task_struct {
728 #ifdef CONFIG_THREAD_INFO_IN_TASK
729 	/*
730 	 * For reasons of header soup (see current_thread_info()), this
731 	 * must be the first element of task_struct.
732 	 */
733 	struct thread_info		thread_info;
734 #endif
735 	unsigned int			__state;
736 
737 #ifdef CONFIG_PREEMPT_RT
738 	/* saved state for "spinlock sleepers" */
739 	unsigned int			saved_state;
740 #endif
741 
742 	/*
743 	 * This begins the randomizable portion of task_struct. Only
744 	 * scheduling-critical items should be added above here.
745 	 */
746 	randomized_struct_fields_start
747 
748 	void				*stack;
749 	refcount_t			usage;
750 	/* Per task flags (PF_*), defined further below: */
751 	unsigned int			flags;
752 	unsigned int			ptrace;
753 
754 #ifdef CONFIG_SMP
755 	int				on_cpu;
756 	struct __call_single_node	wake_entry;
757 	unsigned int			wakee_flips;
758 	unsigned long			wakee_flip_decay_ts;
759 	struct task_struct		*last_wakee;
760 
761 	/*
762 	 * recent_used_cpu is initially set as the last CPU used by a task
763 	 * that wakes affine another task. Waker/wakee relationships can
764 	 * push tasks around a CPU where each wakeup moves to the next one.
765 	 * Tracking a recently used CPU allows a quick search for a recently
766 	 * used CPU that may be idle.
767 	 */
768 	int				recent_used_cpu;
769 	int				wake_cpu;
770 #endif
771 	int				on_rq;
772 
773 	int				prio;
774 	int				static_prio;
775 	int				normal_prio;
776 	unsigned int			rt_priority;
777 
778 	struct sched_entity		se;
779 	struct sched_rt_entity		rt;
780 	struct sched_dl_entity		dl;
781 	const struct sched_class	*sched_class;
782 
783 #ifdef CONFIG_SCHED_CORE
784 	struct rb_node			core_node;
785 	unsigned long			core_cookie;
786 	unsigned int			core_occupation;
787 #endif
788 
789 #ifdef CONFIG_CGROUP_SCHED
790 	struct task_group		*sched_task_group;
791 #endif
792 
793 #ifdef CONFIG_UCLAMP_TASK
794 	/*
795 	 * Clamp values requested for a scheduling entity.
796 	 * Must be updated with task_rq_lock() held.
797 	 */
798 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
799 	/*
800 	 * Effective clamp values used for a scheduling entity.
801 	 * Must be updated with task_rq_lock() held.
802 	 */
803 	struct uclamp_se		uclamp[UCLAMP_CNT];
804 #endif
805 
806 	struct sched_statistics         stats;
807 
808 #ifdef CONFIG_PREEMPT_NOTIFIERS
809 	/* List of struct preempt_notifier: */
810 	struct hlist_head		preempt_notifiers;
811 #endif
812 
813 #ifdef CONFIG_BLK_DEV_IO_TRACE
814 	unsigned int			btrace_seq;
815 #endif
816 
817 	unsigned int			policy;
818 	int				nr_cpus_allowed;
819 	const cpumask_t			*cpus_ptr;
820 	cpumask_t			*user_cpus_ptr;
821 	cpumask_t			cpus_mask;
822 	void				*migration_pending;
823 #ifdef CONFIG_SMP
824 	unsigned short			migration_disabled;
825 #endif
826 	unsigned short			migration_flags;
827 
828 #ifdef CONFIG_PREEMPT_RCU
829 	int				rcu_read_lock_nesting;
830 	union rcu_special		rcu_read_unlock_special;
831 	struct list_head		rcu_node_entry;
832 	struct rcu_node			*rcu_blocked_node;
833 #endif /* #ifdef CONFIG_PREEMPT_RCU */
834 
835 #ifdef CONFIG_TASKS_RCU
836 	unsigned long			rcu_tasks_nvcsw;
837 	u8				rcu_tasks_holdout;
838 	u8				rcu_tasks_idx;
839 	int				rcu_tasks_idle_cpu;
840 	struct list_head		rcu_tasks_holdout_list;
841 #endif /* #ifdef CONFIG_TASKS_RCU */
842 
843 #ifdef CONFIG_TASKS_TRACE_RCU
844 	int				trc_reader_nesting;
845 	int				trc_ipi_to_cpu;
846 	union rcu_special		trc_reader_special;
847 	bool				trc_reader_checked;
848 	struct list_head		trc_holdout_list;
849 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
850 
851 	struct sched_info		sched_info;
852 
853 	struct list_head		tasks;
854 #ifdef CONFIG_SMP
855 	struct plist_node		pushable_tasks;
856 	struct rb_node			pushable_dl_tasks;
857 #endif
858 
859 	struct mm_struct		*mm;
860 	struct mm_struct		*active_mm;
861 
862 	/* Per-thread vma caching: */
863 	struct vmacache			vmacache;
864 
865 #ifdef SPLIT_RSS_COUNTING
866 	struct task_rss_stat		rss_stat;
867 #endif
868 	int				exit_state;
869 	int				exit_code;
870 	int				exit_signal;
871 	/* The signal sent when the parent dies: */
872 	int				pdeath_signal;
873 	/* JOBCTL_*, siglock protected: */
874 	unsigned long			jobctl;
875 
876 	/* Used for emulating ABI behavior of previous Linux versions: */
877 	unsigned int			personality;
878 
879 	/* Scheduler bits, serialized by scheduler locks: */
880 	unsigned			sched_reset_on_fork:1;
881 	unsigned			sched_contributes_to_load:1;
882 	unsigned			sched_migrated:1;
883 #ifdef CONFIG_PSI
884 	unsigned			sched_psi_wake_requeue:1;
885 #endif
886 
887 	/* Force alignment to the next boundary: */
888 	unsigned			:0;
889 
890 	/* Unserialized, strictly 'current' */
891 
892 	/*
893 	 * This field must not be in the scheduler word above due to wakelist
894 	 * queueing no longer being serialized by p->on_cpu. However:
895 	 *
896 	 * p->XXX = X;			ttwu()
897 	 * schedule()			  if (p->on_rq && ..) // false
898 	 *   smp_mb__after_spinlock();	  if (smp_load_acquire(&p->on_cpu) && //true
899 	 *   deactivate_task()		      ttwu_queue_wakelist())
900 	 *     p->on_rq = 0;			p->sched_remote_wakeup = Y;
901 	 *
902 	 * guarantees all stores of 'current' are visible before
903 	 * ->sched_remote_wakeup gets used, so it can be in this word.
904 	 */
905 	unsigned			sched_remote_wakeup:1;
906 
907 	/* Bit to tell LSMs we're in execve(): */
908 	unsigned			in_execve:1;
909 	unsigned			in_iowait:1;
910 #ifndef TIF_RESTORE_SIGMASK
911 	unsigned			restore_sigmask:1;
912 #endif
913 #ifdef CONFIG_MEMCG
914 	unsigned			in_user_fault:1;
915 #endif
916 #ifdef CONFIG_COMPAT_BRK
917 	unsigned			brk_randomized:1;
918 #endif
919 #ifdef CONFIG_CGROUPS
920 	/* disallow userland-initiated cgroup migration */
921 	unsigned			no_cgroup_migration:1;
922 	/* task is frozen/stopped (used by the cgroup freezer) */
923 	unsigned			frozen:1;
924 #endif
925 #ifdef CONFIG_BLK_CGROUP
926 	unsigned			use_memdelay:1;
927 #endif
928 #ifdef CONFIG_PSI
929 	/* Stalled due to lack of memory */
930 	unsigned			in_memstall:1;
931 #endif
932 #ifdef CONFIG_PAGE_OWNER
933 	/* Used by page_owner=on to detect recursion in page tracking. */
934 	unsigned			in_page_owner:1;
935 #endif
936 #ifdef CONFIG_EVENTFD
937 	/* Recursion prevention for eventfd_signal() */
938 	unsigned			in_eventfd_signal:1;
939 #endif
940 #ifdef CONFIG_IOMMU_SVA
941 	unsigned			pasid_activated:1;
942 #endif
943 #ifdef	CONFIG_CPU_SUP_INTEL
944 	unsigned			reported_split_lock:1;
945 #endif
946 
947 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
948 
949 	struct restart_block		restart_block;
950 
951 	pid_t				pid;
952 	pid_t				tgid;
953 
954 #ifdef CONFIG_STACKPROTECTOR
955 	/* Canary value for the -fstack-protector GCC feature: */
956 	unsigned long			stack_canary;
957 #endif
958 	/*
959 	 * Pointers to the (original) parent process, youngest child, younger sibling,
960 	 * older sibling, respectively.  (p->father can be replaced with
961 	 * p->real_parent->pid)
962 	 */
963 
964 	/* Real parent process: */
965 	struct task_struct __rcu	*real_parent;
966 
967 	/* Recipient of SIGCHLD, wait4() reports: */
968 	struct task_struct __rcu	*parent;
969 
970 	/*
971 	 * Children/sibling form the list of natural children:
972 	 */
973 	struct list_head		children;
974 	struct list_head		sibling;
975 	struct task_struct		*group_leader;
976 
977 	/*
978 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
979 	 *
980 	 * This includes both natural children and PTRACE_ATTACH targets.
981 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
982 	 */
983 	struct list_head		ptraced;
984 	struct list_head		ptrace_entry;
985 
986 	/* PID/PID hash table linkage. */
987 	struct pid			*thread_pid;
988 	struct hlist_node		pid_links[PIDTYPE_MAX];
989 	struct list_head		thread_group;
990 	struct list_head		thread_node;
991 
992 	struct completion		*vfork_done;
993 
994 	/* CLONE_CHILD_SETTID: */
995 	int __user			*set_child_tid;
996 
997 	/* CLONE_CHILD_CLEARTID: */
998 	int __user			*clear_child_tid;
999 
1000 	/* PF_KTHREAD | PF_IO_WORKER */
1001 	void				*worker_private;
1002 
1003 	u64				utime;
1004 	u64				stime;
1005 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1006 	u64				utimescaled;
1007 	u64				stimescaled;
1008 #endif
1009 	u64				gtime;
1010 	struct prev_cputime		prev_cputime;
1011 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1012 	struct vtime			vtime;
1013 #endif
1014 
1015 #ifdef CONFIG_NO_HZ_FULL
1016 	atomic_t			tick_dep_mask;
1017 #endif
1018 	/* Context switch counts: */
1019 	unsigned long			nvcsw;
1020 	unsigned long			nivcsw;
1021 
1022 	/* Monotonic time in nsecs: */
1023 	u64				start_time;
1024 
1025 	/* Boot based time in nsecs: */
1026 	u64				start_boottime;
1027 
1028 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1029 	unsigned long			min_flt;
1030 	unsigned long			maj_flt;
1031 
1032 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1033 	struct posix_cputimers		posix_cputimers;
1034 
1035 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1036 	struct posix_cputimers_work	posix_cputimers_work;
1037 #endif
1038 
1039 	/* Process credentials: */
1040 
1041 	/* Tracer's credentials at attach: */
1042 	const struct cred __rcu		*ptracer_cred;
1043 
1044 	/* Objective and real subjective task credentials (COW): */
1045 	const struct cred __rcu		*real_cred;
1046 
1047 	/* Effective (overridable) subjective task credentials (COW): */
1048 	const struct cred __rcu		*cred;
1049 
1050 #ifdef CONFIG_KEYS
1051 	/* Cached requested key. */
1052 	struct key			*cached_requested_key;
1053 #endif
1054 
1055 	/*
1056 	 * executable name, excluding path.
1057 	 *
1058 	 * - normally initialized setup_new_exec()
1059 	 * - access it with [gs]et_task_comm()
1060 	 * - lock it with task_lock()
1061 	 */
1062 	char				comm[TASK_COMM_LEN];
1063 
1064 	struct nameidata		*nameidata;
1065 
1066 #ifdef CONFIG_SYSVIPC
1067 	struct sysv_sem			sysvsem;
1068 	struct sysv_shm			sysvshm;
1069 #endif
1070 #ifdef CONFIG_DETECT_HUNG_TASK
1071 	unsigned long			last_switch_count;
1072 	unsigned long			last_switch_time;
1073 #endif
1074 	/* Filesystem information: */
1075 	struct fs_struct		*fs;
1076 
1077 	/* Open file information: */
1078 	struct files_struct		*files;
1079 
1080 #ifdef CONFIG_IO_URING
1081 	struct io_uring_task		*io_uring;
1082 #endif
1083 
1084 	/* Namespaces: */
1085 	struct nsproxy			*nsproxy;
1086 
1087 	/* Signal handlers: */
1088 	struct signal_struct		*signal;
1089 	struct sighand_struct __rcu		*sighand;
1090 	sigset_t			blocked;
1091 	sigset_t			real_blocked;
1092 	/* Restored if set_restore_sigmask() was used: */
1093 	sigset_t			saved_sigmask;
1094 	struct sigpending		pending;
1095 	unsigned long			sas_ss_sp;
1096 	size_t				sas_ss_size;
1097 	unsigned int			sas_ss_flags;
1098 
1099 	struct callback_head		*task_works;
1100 
1101 #ifdef CONFIG_AUDIT
1102 #ifdef CONFIG_AUDITSYSCALL
1103 	struct audit_context		*audit_context;
1104 #endif
1105 	kuid_t				loginuid;
1106 	unsigned int			sessionid;
1107 #endif
1108 	struct seccomp			seccomp;
1109 	struct syscall_user_dispatch	syscall_dispatch;
1110 
1111 	/* Thread group tracking: */
1112 	u64				parent_exec_id;
1113 	u64				self_exec_id;
1114 
1115 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1116 	spinlock_t			alloc_lock;
1117 
1118 	/* Protection of the PI data structures: */
1119 	raw_spinlock_t			pi_lock;
1120 
1121 	struct wake_q_node		wake_q;
1122 
1123 #ifdef CONFIG_RT_MUTEXES
1124 	/* PI waiters blocked on a rt_mutex held by this task: */
1125 	struct rb_root_cached		pi_waiters;
1126 	/* Updated under owner's pi_lock and rq lock */
1127 	struct task_struct		*pi_top_task;
1128 	/* Deadlock detection and priority inheritance handling: */
1129 	struct rt_mutex_waiter		*pi_blocked_on;
1130 #endif
1131 
1132 #ifdef CONFIG_DEBUG_MUTEXES
1133 	/* Mutex deadlock detection: */
1134 	struct mutex_waiter		*blocked_on;
1135 #endif
1136 
1137 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1138 	int				non_block_count;
1139 #endif
1140 
1141 #ifdef CONFIG_TRACE_IRQFLAGS
1142 	struct irqtrace_events		irqtrace;
1143 	unsigned int			hardirq_threaded;
1144 	u64				hardirq_chain_key;
1145 	int				softirqs_enabled;
1146 	int				softirq_context;
1147 	int				irq_config;
1148 #endif
1149 #ifdef CONFIG_PREEMPT_RT
1150 	int				softirq_disable_cnt;
1151 #endif
1152 
1153 #ifdef CONFIG_LOCKDEP
1154 # define MAX_LOCK_DEPTH			48UL
1155 	u64				curr_chain_key;
1156 	int				lockdep_depth;
1157 	unsigned int			lockdep_recursion;
1158 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1159 #endif
1160 
1161 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1162 	unsigned int			in_ubsan;
1163 #endif
1164 
1165 	/* Journalling filesystem info: */
1166 	void				*journal_info;
1167 
1168 	/* Stacked block device info: */
1169 	struct bio_list			*bio_list;
1170 
1171 	/* Stack plugging: */
1172 	struct blk_plug			*plug;
1173 
1174 	/* VM state: */
1175 	struct reclaim_state		*reclaim_state;
1176 
1177 	struct backing_dev_info		*backing_dev_info;
1178 
1179 	struct io_context		*io_context;
1180 
1181 #ifdef CONFIG_COMPACTION
1182 	struct capture_control		*capture_control;
1183 #endif
1184 	/* Ptrace state: */
1185 	unsigned long			ptrace_message;
1186 	kernel_siginfo_t		*last_siginfo;
1187 
1188 	struct task_io_accounting	ioac;
1189 #ifdef CONFIG_PSI
1190 	/* Pressure stall state */
1191 	unsigned int			psi_flags;
1192 #endif
1193 #ifdef CONFIG_TASK_XACCT
1194 	/* Accumulated RSS usage: */
1195 	u64				acct_rss_mem1;
1196 	/* Accumulated virtual memory usage: */
1197 	u64				acct_vm_mem1;
1198 	/* stime + utime since last update: */
1199 	u64				acct_timexpd;
1200 #endif
1201 #ifdef CONFIG_CPUSETS
1202 	/* Protected by ->alloc_lock: */
1203 	nodemask_t			mems_allowed;
1204 	/* Sequence number to catch updates: */
1205 	seqcount_spinlock_t		mems_allowed_seq;
1206 	int				cpuset_mem_spread_rotor;
1207 	int				cpuset_slab_spread_rotor;
1208 #endif
1209 #ifdef CONFIG_CGROUPS
1210 	/* Control Group info protected by css_set_lock: */
1211 	struct css_set __rcu		*cgroups;
1212 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1213 	struct list_head		cg_list;
1214 #endif
1215 #ifdef CONFIG_X86_CPU_RESCTRL
1216 	u32				closid;
1217 	u32				rmid;
1218 #endif
1219 #ifdef CONFIG_FUTEX
1220 	struct robust_list_head __user	*robust_list;
1221 #ifdef CONFIG_COMPAT
1222 	struct compat_robust_list_head __user *compat_robust_list;
1223 #endif
1224 	struct list_head		pi_state_list;
1225 	struct futex_pi_state		*pi_state_cache;
1226 	struct mutex			futex_exit_mutex;
1227 	unsigned int			futex_state;
1228 #endif
1229 #ifdef CONFIG_PERF_EVENTS
1230 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1231 	struct mutex			perf_event_mutex;
1232 	struct list_head		perf_event_list;
1233 #endif
1234 #ifdef CONFIG_DEBUG_PREEMPT
1235 	unsigned long			preempt_disable_ip;
1236 #endif
1237 #ifdef CONFIG_NUMA
1238 	/* Protected by alloc_lock: */
1239 	struct mempolicy		*mempolicy;
1240 	short				il_prev;
1241 	short				pref_node_fork;
1242 #endif
1243 #ifdef CONFIG_NUMA_BALANCING
1244 	int				numa_scan_seq;
1245 	unsigned int			numa_scan_period;
1246 	unsigned int			numa_scan_period_max;
1247 	int				numa_preferred_nid;
1248 	unsigned long			numa_migrate_retry;
1249 	/* Migration stamp: */
1250 	u64				node_stamp;
1251 	u64				last_task_numa_placement;
1252 	u64				last_sum_exec_runtime;
1253 	struct callback_head		numa_work;
1254 
1255 	/*
1256 	 * This pointer is only modified for current in syscall and
1257 	 * pagefault context (and for tasks being destroyed), so it can be read
1258 	 * from any of the following contexts:
1259 	 *  - RCU read-side critical section
1260 	 *  - current->numa_group from everywhere
1261 	 *  - task's runqueue locked, task not running
1262 	 */
1263 	struct numa_group __rcu		*numa_group;
1264 
1265 	/*
1266 	 * numa_faults is an array split into four regions:
1267 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1268 	 * in this precise order.
1269 	 *
1270 	 * faults_memory: Exponential decaying average of faults on a per-node
1271 	 * basis. Scheduling placement decisions are made based on these
1272 	 * counts. The values remain static for the duration of a PTE scan.
1273 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1274 	 * hinting fault was incurred.
1275 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1276 	 * during the current scan window. When the scan completes, the counts
1277 	 * in faults_memory and faults_cpu decay and these values are copied.
1278 	 */
1279 	unsigned long			*numa_faults;
1280 	unsigned long			total_numa_faults;
1281 
1282 	/*
1283 	 * numa_faults_locality tracks if faults recorded during the last
1284 	 * scan window were remote/local or failed to migrate. The task scan
1285 	 * period is adapted based on the locality of the faults with different
1286 	 * weights depending on whether they were shared or private faults
1287 	 */
1288 	unsigned long			numa_faults_locality[3];
1289 
1290 	unsigned long			numa_pages_migrated;
1291 #endif /* CONFIG_NUMA_BALANCING */
1292 
1293 #ifdef CONFIG_RSEQ
1294 	struct rseq __user *rseq;
1295 	u32 rseq_sig;
1296 	/*
1297 	 * RmW on rseq_event_mask must be performed atomically
1298 	 * with respect to preemption.
1299 	 */
1300 	unsigned long rseq_event_mask;
1301 #endif
1302 
1303 	struct tlbflush_unmap_batch	tlb_ubc;
1304 
1305 	union {
1306 		refcount_t		rcu_users;
1307 		struct rcu_head		rcu;
1308 	};
1309 
1310 	/* Cache last used pipe for splice(): */
1311 	struct pipe_inode_info		*splice_pipe;
1312 
1313 	struct page_frag		task_frag;
1314 
1315 #ifdef CONFIG_TASK_DELAY_ACCT
1316 	struct task_delay_info		*delays;
1317 #endif
1318 
1319 #ifdef CONFIG_FAULT_INJECTION
1320 	int				make_it_fail;
1321 	unsigned int			fail_nth;
1322 #endif
1323 	/*
1324 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1325 	 * balance_dirty_pages() for a dirty throttling pause:
1326 	 */
1327 	int				nr_dirtied;
1328 	int				nr_dirtied_pause;
1329 	/* Start of a write-and-pause period: */
1330 	unsigned long			dirty_paused_when;
1331 
1332 #ifdef CONFIG_LATENCYTOP
1333 	int				latency_record_count;
1334 	struct latency_record		latency_record[LT_SAVECOUNT];
1335 #endif
1336 	/*
1337 	 * Time slack values; these are used to round up poll() and
1338 	 * select() etc timeout values. These are in nanoseconds.
1339 	 */
1340 	u64				timer_slack_ns;
1341 	u64				default_timer_slack_ns;
1342 
1343 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1344 	unsigned int			kasan_depth;
1345 #endif
1346 
1347 #ifdef CONFIG_KCSAN
1348 	struct kcsan_ctx		kcsan_ctx;
1349 #ifdef CONFIG_TRACE_IRQFLAGS
1350 	struct irqtrace_events		kcsan_save_irqtrace;
1351 #endif
1352 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1353 	int				kcsan_stack_depth;
1354 #endif
1355 #endif
1356 
1357 #if IS_ENABLED(CONFIG_KUNIT)
1358 	struct kunit			*kunit_test;
1359 #endif
1360 
1361 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1362 	/* Index of current stored address in ret_stack: */
1363 	int				curr_ret_stack;
1364 	int				curr_ret_depth;
1365 
1366 	/* Stack of return addresses for return function tracing: */
1367 	struct ftrace_ret_stack		*ret_stack;
1368 
1369 	/* Timestamp for last schedule: */
1370 	unsigned long long		ftrace_timestamp;
1371 
1372 	/*
1373 	 * Number of functions that haven't been traced
1374 	 * because of depth overrun:
1375 	 */
1376 	atomic_t			trace_overrun;
1377 
1378 	/* Pause tracing: */
1379 	atomic_t			tracing_graph_pause;
1380 #endif
1381 
1382 #ifdef CONFIG_TRACING
1383 	/* State flags for use by tracers: */
1384 	unsigned long			trace;
1385 
1386 	/* Bitmask and counter of trace recursion: */
1387 	unsigned long			trace_recursion;
1388 #endif /* CONFIG_TRACING */
1389 
1390 #ifdef CONFIG_KCOV
1391 	/* See kernel/kcov.c for more details. */
1392 
1393 	/* Coverage collection mode enabled for this task (0 if disabled): */
1394 	unsigned int			kcov_mode;
1395 
1396 	/* Size of the kcov_area: */
1397 	unsigned int			kcov_size;
1398 
1399 	/* Buffer for coverage collection: */
1400 	void				*kcov_area;
1401 
1402 	/* KCOV descriptor wired with this task or NULL: */
1403 	struct kcov			*kcov;
1404 
1405 	/* KCOV common handle for remote coverage collection: */
1406 	u64				kcov_handle;
1407 
1408 	/* KCOV sequence number: */
1409 	int				kcov_sequence;
1410 
1411 	/* Collect coverage from softirq context: */
1412 	unsigned int			kcov_softirq;
1413 #endif
1414 
1415 #ifdef CONFIG_MEMCG
1416 	struct mem_cgroup		*memcg_in_oom;
1417 	gfp_t				memcg_oom_gfp_mask;
1418 	int				memcg_oom_order;
1419 
1420 	/* Number of pages to reclaim on returning to userland: */
1421 	unsigned int			memcg_nr_pages_over_high;
1422 
1423 	/* Used by memcontrol for targeted memcg charge: */
1424 	struct mem_cgroup		*active_memcg;
1425 #endif
1426 
1427 #ifdef CONFIG_BLK_CGROUP
1428 	struct request_queue		*throttle_queue;
1429 #endif
1430 
1431 #ifdef CONFIG_UPROBES
1432 	struct uprobe_task		*utask;
1433 #endif
1434 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1435 	unsigned int			sequential_io;
1436 	unsigned int			sequential_io_avg;
1437 #endif
1438 	struct kmap_ctrl		kmap_ctrl;
1439 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1440 	unsigned long			task_state_change;
1441 # ifdef CONFIG_PREEMPT_RT
1442 	unsigned long			saved_state_change;
1443 # endif
1444 #endif
1445 	int				pagefault_disabled;
1446 #ifdef CONFIG_MMU
1447 	struct task_struct		*oom_reaper_list;
1448 	struct timer_list		oom_reaper_timer;
1449 #endif
1450 #ifdef CONFIG_VMAP_STACK
1451 	struct vm_struct		*stack_vm_area;
1452 #endif
1453 #ifdef CONFIG_THREAD_INFO_IN_TASK
1454 	/* A live task holds one reference: */
1455 	refcount_t			stack_refcount;
1456 #endif
1457 #ifdef CONFIG_LIVEPATCH
1458 	int patch_state;
1459 #endif
1460 #ifdef CONFIG_SECURITY
1461 	/* Used by LSM modules for access restriction: */
1462 	void				*security;
1463 #endif
1464 #ifdef CONFIG_BPF_SYSCALL
1465 	/* Used by BPF task local storage */
1466 	struct bpf_local_storage __rcu	*bpf_storage;
1467 	/* Used for BPF run context */
1468 	struct bpf_run_ctx		*bpf_ctx;
1469 #endif
1470 
1471 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1472 	unsigned long			lowest_stack;
1473 	unsigned long			prev_lowest_stack;
1474 #endif
1475 
1476 #ifdef CONFIG_X86_MCE
1477 	void __user			*mce_vaddr;
1478 	__u64				mce_kflags;
1479 	u64				mce_addr;
1480 	__u64				mce_ripv : 1,
1481 					mce_whole_page : 1,
1482 					__mce_reserved : 62;
1483 	struct callback_head		mce_kill_me;
1484 	int				mce_count;
1485 #endif
1486 
1487 #ifdef CONFIG_KRETPROBES
1488 	struct llist_head               kretprobe_instances;
1489 #endif
1490 #ifdef CONFIG_RETHOOK
1491 	struct llist_head               rethooks;
1492 #endif
1493 
1494 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1495 	/*
1496 	 * If L1D flush is supported on mm context switch
1497 	 * then we use this callback head to queue kill work
1498 	 * to kill tasks that are not running on SMT disabled
1499 	 * cores
1500 	 */
1501 	struct callback_head		l1d_flush_kill;
1502 #endif
1503 
1504 #ifdef CONFIG_RV
1505 	/*
1506 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1507 	 * If we find justification for more monitors, we can think
1508 	 * about adding more or developing a dynamic method. So far,
1509 	 * none of these are justified.
1510 	 */
1511 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1512 #endif
1513 
1514 	/*
1515 	 * New fields for task_struct should be added above here, so that
1516 	 * they are included in the randomized portion of task_struct.
1517 	 */
1518 	randomized_struct_fields_end
1519 
1520 	/* CPU-specific state of this task: */
1521 	struct thread_struct		thread;
1522 
1523 	/*
1524 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1525 	 * structure.  It *MUST* be at the end of 'task_struct'.
1526 	 *
1527 	 * Do not put anything below here!
1528 	 */
1529 };
1530 
1531 static inline struct pid *task_pid(struct task_struct *task)
1532 {
1533 	return task->thread_pid;
1534 }
1535 
1536 /*
1537  * the helpers to get the task's different pids as they are seen
1538  * from various namespaces
1539  *
1540  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1541  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1542  *                     current.
1543  * task_xid_nr_ns()  : id seen from the ns specified;
1544  *
1545  * see also pid_nr() etc in include/linux/pid.h
1546  */
1547 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1548 
1549 static inline pid_t task_pid_nr(struct task_struct *tsk)
1550 {
1551 	return tsk->pid;
1552 }
1553 
1554 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1555 {
1556 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1557 }
1558 
1559 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1560 {
1561 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1562 }
1563 
1564 
1565 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1566 {
1567 	return tsk->tgid;
1568 }
1569 
1570 /**
1571  * pid_alive - check that a task structure is not stale
1572  * @p: Task structure to be checked.
1573  *
1574  * Test if a process is not yet dead (at most zombie state)
1575  * If pid_alive fails, then pointers within the task structure
1576  * can be stale and must not be dereferenced.
1577  *
1578  * Return: 1 if the process is alive. 0 otherwise.
1579  */
1580 static inline int pid_alive(const struct task_struct *p)
1581 {
1582 	return p->thread_pid != NULL;
1583 }
1584 
1585 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1586 {
1587 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1588 }
1589 
1590 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1591 {
1592 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1593 }
1594 
1595 
1596 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1597 {
1598 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1599 }
1600 
1601 static inline pid_t task_session_vnr(struct task_struct *tsk)
1602 {
1603 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1604 }
1605 
1606 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1607 {
1608 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1609 }
1610 
1611 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1612 {
1613 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1614 }
1615 
1616 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1617 {
1618 	pid_t pid = 0;
1619 
1620 	rcu_read_lock();
1621 	if (pid_alive(tsk))
1622 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1623 	rcu_read_unlock();
1624 
1625 	return pid;
1626 }
1627 
1628 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1629 {
1630 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1631 }
1632 
1633 /* Obsolete, do not use: */
1634 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1635 {
1636 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1637 }
1638 
1639 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1640 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1641 
1642 static inline unsigned int __task_state_index(unsigned int tsk_state,
1643 					      unsigned int tsk_exit_state)
1644 {
1645 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1646 
1647 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1648 
1649 	if (tsk_state == TASK_IDLE)
1650 		state = TASK_REPORT_IDLE;
1651 
1652 	/*
1653 	 * We're lying here, but rather than expose a completely new task state
1654 	 * to userspace, we can make this appear as if the task has gone through
1655 	 * a regular rt_mutex_lock() call.
1656 	 */
1657 	if (tsk_state == TASK_RTLOCK_WAIT)
1658 		state = TASK_UNINTERRUPTIBLE;
1659 
1660 	return fls(state);
1661 }
1662 
1663 static inline unsigned int task_state_index(struct task_struct *tsk)
1664 {
1665 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1666 }
1667 
1668 static inline char task_index_to_char(unsigned int state)
1669 {
1670 	static const char state_char[] = "RSDTtXZPI";
1671 
1672 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1673 
1674 	return state_char[state];
1675 }
1676 
1677 static inline char task_state_to_char(struct task_struct *tsk)
1678 {
1679 	return task_index_to_char(task_state_index(tsk));
1680 }
1681 
1682 /**
1683  * is_global_init - check if a task structure is init. Since init
1684  * is free to have sub-threads we need to check tgid.
1685  * @tsk: Task structure to be checked.
1686  *
1687  * Check if a task structure is the first user space task the kernel created.
1688  *
1689  * Return: 1 if the task structure is init. 0 otherwise.
1690  */
1691 static inline int is_global_init(struct task_struct *tsk)
1692 {
1693 	return task_tgid_nr(tsk) == 1;
1694 }
1695 
1696 extern struct pid *cad_pid;
1697 
1698 /*
1699  * Per process flags
1700  */
1701 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1702 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1703 #define PF_EXITING		0x00000004	/* Getting shut down */
1704 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1705 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1706 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1707 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1708 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1709 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1710 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1711 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1712 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1713 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1714 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1715 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1716 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1717 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1718 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1719 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1720 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1721 						 * I am cleaning dirty pages from some other bdi. */
1722 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1723 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1724 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1725 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1726 #define PF_MEMALLOC_PIN		0x10000000	/* Allocation context constrained to zones which allow long term pinning. */
1727 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1728 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1729 
1730 /*
1731  * Only the _current_ task can read/write to tsk->flags, but other
1732  * tasks can access tsk->flags in readonly mode for example
1733  * with tsk_used_math (like during threaded core dumping).
1734  * There is however an exception to this rule during ptrace
1735  * or during fork: the ptracer task is allowed to write to the
1736  * child->flags of its traced child (same goes for fork, the parent
1737  * can write to the child->flags), because we're guaranteed the
1738  * child is not running and in turn not changing child->flags
1739  * at the same time the parent does it.
1740  */
1741 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1742 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1743 #define clear_used_math()			clear_stopped_child_used_math(current)
1744 #define set_used_math()				set_stopped_child_used_math(current)
1745 
1746 #define conditional_stopped_child_used_math(condition, child) \
1747 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1748 
1749 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1750 
1751 #define copy_to_stopped_child_used_math(child) \
1752 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1753 
1754 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1755 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1756 #define used_math()				tsk_used_math(current)
1757 
1758 static __always_inline bool is_percpu_thread(void)
1759 {
1760 #ifdef CONFIG_SMP
1761 	return (current->flags & PF_NO_SETAFFINITY) &&
1762 		(current->nr_cpus_allowed  == 1);
1763 #else
1764 	return true;
1765 #endif
1766 }
1767 
1768 /* Per-process atomic flags. */
1769 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1770 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1771 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1772 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1773 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1774 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1775 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1776 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1777 
1778 #define TASK_PFA_TEST(name, func)					\
1779 	static inline bool task_##func(struct task_struct *p)		\
1780 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1781 
1782 #define TASK_PFA_SET(name, func)					\
1783 	static inline void task_set_##func(struct task_struct *p)	\
1784 	{ set_bit(PFA_##name, &p->atomic_flags); }
1785 
1786 #define TASK_PFA_CLEAR(name, func)					\
1787 	static inline void task_clear_##func(struct task_struct *p)	\
1788 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1789 
1790 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1791 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1792 
1793 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1794 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1795 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1796 
1797 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1798 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1799 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1800 
1801 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1802 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1803 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1804 
1805 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1806 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1807 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1808 
1809 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1810 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1811 
1812 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1813 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1814 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1815 
1816 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1817 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1818 
1819 static inline void
1820 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1821 {
1822 	current->flags &= ~flags;
1823 	current->flags |= orig_flags & flags;
1824 }
1825 
1826 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1827 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1828 #ifdef CONFIG_SMP
1829 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1830 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1831 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1832 extern void release_user_cpus_ptr(struct task_struct *p);
1833 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1834 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1835 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1836 #else
1837 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1838 {
1839 }
1840 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1841 {
1842 	if (!cpumask_test_cpu(0, new_mask))
1843 		return -EINVAL;
1844 	return 0;
1845 }
1846 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1847 {
1848 	if (src->user_cpus_ptr)
1849 		return -EINVAL;
1850 	return 0;
1851 }
1852 static inline void release_user_cpus_ptr(struct task_struct *p)
1853 {
1854 	WARN_ON(p->user_cpus_ptr);
1855 }
1856 
1857 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1858 {
1859 	return 0;
1860 }
1861 #endif
1862 
1863 extern int yield_to(struct task_struct *p, bool preempt);
1864 extern void set_user_nice(struct task_struct *p, long nice);
1865 extern int task_prio(const struct task_struct *p);
1866 
1867 /**
1868  * task_nice - return the nice value of a given task.
1869  * @p: the task in question.
1870  *
1871  * Return: The nice value [ -20 ... 0 ... 19 ].
1872  */
1873 static inline int task_nice(const struct task_struct *p)
1874 {
1875 	return PRIO_TO_NICE((p)->static_prio);
1876 }
1877 
1878 extern int can_nice(const struct task_struct *p, const int nice);
1879 extern int task_curr(const struct task_struct *p);
1880 extern int idle_cpu(int cpu);
1881 extern int available_idle_cpu(int cpu);
1882 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1883 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1884 extern void sched_set_fifo(struct task_struct *p);
1885 extern void sched_set_fifo_low(struct task_struct *p);
1886 extern void sched_set_normal(struct task_struct *p, int nice);
1887 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1888 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1889 extern struct task_struct *idle_task(int cpu);
1890 
1891 /**
1892  * is_idle_task - is the specified task an idle task?
1893  * @p: the task in question.
1894  *
1895  * Return: 1 if @p is an idle task. 0 otherwise.
1896  */
1897 static __always_inline bool is_idle_task(const struct task_struct *p)
1898 {
1899 	return !!(p->flags & PF_IDLE);
1900 }
1901 
1902 extern struct task_struct *curr_task(int cpu);
1903 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1904 
1905 void yield(void);
1906 
1907 union thread_union {
1908 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1909 	struct task_struct task;
1910 #endif
1911 #ifndef CONFIG_THREAD_INFO_IN_TASK
1912 	struct thread_info thread_info;
1913 #endif
1914 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1915 };
1916 
1917 #ifndef CONFIG_THREAD_INFO_IN_TASK
1918 extern struct thread_info init_thread_info;
1919 #endif
1920 
1921 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1922 
1923 #ifdef CONFIG_THREAD_INFO_IN_TASK
1924 # define task_thread_info(task)	(&(task)->thread_info)
1925 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1926 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1927 #endif
1928 
1929 /*
1930  * find a task by one of its numerical ids
1931  *
1932  * find_task_by_pid_ns():
1933  *      finds a task by its pid in the specified namespace
1934  * find_task_by_vpid():
1935  *      finds a task by its virtual pid
1936  *
1937  * see also find_vpid() etc in include/linux/pid.h
1938  */
1939 
1940 extern struct task_struct *find_task_by_vpid(pid_t nr);
1941 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1942 
1943 /*
1944  * find a task by its virtual pid and get the task struct
1945  */
1946 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1947 
1948 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1949 extern int wake_up_process(struct task_struct *tsk);
1950 extern void wake_up_new_task(struct task_struct *tsk);
1951 
1952 #ifdef CONFIG_SMP
1953 extern void kick_process(struct task_struct *tsk);
1954 #else
1955 static inline void kick_process(struct task_struct *tsk) { }
1956 #endif
1957 
1958 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1959 
1960 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1961 {
1962 	__set_task_comm(tsk, from, false);
1963 }
1964 
1965 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1966 #define get_task_comm(buf, tsk) ({			\
1967 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1968 	__get_task_comm(buf, sizeof(buf), tsk);		\
1969 })
1970 
1971 #ifdef CONFIG_SMP
1972 static __always_inline void scheduler_ipi(void)
1973 {
1974 	/*
1975 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1976 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1977 	 * this IPI.
1978 	 */
1979 	preempt_fold_need_resched();
1980 }
1981 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1982 #else
1983 static inline void scheduler_ipi(void) { }
1984 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1985 {
1986 	return 1;
1987 }
1988 #endif
1989 
1990 /*
1991  * Set thread flags in other task's structures.
1992  * See asm/thread_info.h for TIF_xxxx flags available:
1993  */
1994 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1995 {
1996 	set_ti_thread_flag(task_thread_info(tsk), flag);
1997 }
1998 
1999 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2000 {
2001 	clear_ti_thread_flag(task_thread_info(tsk), flag);
2002 }
2003 
2004 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2005 					  bool value)
2006 {
2007 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
2008 }
2009 
2010 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2011 {
2012 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2013 }
2014 
2015 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2016 {
2017 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2018 }
2019 
2020 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2021 {
2022 	return test_ti_thread_flag(task_thread_info(tsk), flag);
2023 }
2024 
2025 static inline void set_tsk_need_resched(struct task_struct *tsk)
2026 {
2027 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2028 }
2029 
2030 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2031 {
2032 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2033 }
2034 
2035 static inline int test_tsk_need_resched(struct task_struct *tsk)
2036 {
2037 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2038 }
2039 
2040 /*
2041  * cond_resched() and cond_resched_lock(): latency reduction via
2042  * explicit rescheduling in places that are safe. The return
2043  * value indicates whether a reschedule was done in fact.
2044  * cond_resched_lock() will drop the spinlock before scheduling,
2045  */
2046 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2047 extern int __cond_resched(void);
2048 
2049 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2050 
2051 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2052 
2053 static __always_inline int _cond_resched(void)
2054 {
2055 	return static_call_mod(cond_resched)();
2056 }
2057 
2058 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2059 extern int dynamic_cond_resched(void);
2060 
2061 static __always_inline int _cond_resched(void)
2062 {
2063 	return dynamic_cond_resched();
2064 }
2065 
2066 #else
2067 
2068 static inline int _cond_resched(void)
2069 {
2070 	return __cond_resched();
2071 }
2072 
2073 #endif /* CONFIG_PREEMPT_DYNAMIC */
2074 
2075 #else
2076 
2077 static inline int _cond_resched(void) { return 0; }
2078 
2079 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2080 
2081 #define cond_resched() ({			\
2082 	__might_resched(__FILE__, __LINE__, 0);	\
2083 	_cond_resched();			\
2084 })
2085 
2086 extern int __cond_resched_lock(spinlock_t *lock);
2087 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2088 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2089 
2090 #define MIGHT_RESCHED_RCU_SHIFT		8
2091 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2092 
2093 #ifndef CONFIG_PREEMPT_RT
2094 /*
2095  * Non RT kernels have an elevated preempt count due to the held lock,
2096  * but are not allowed to be inside a RCU read side critical section
2097  */
2098 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2099 #else
2100 /*
2101  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2102  * cond_resched*lock() has to take that into account because it checks for
2103  * preempt_count() and rcu_preempt_depth().
2104  */
2105 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2106 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2107 #endif
2108 
2109 #define cond_resched_lock(lock) ({						\
2110 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2111 	__cond_resched_lock(lock);						\
2112 })
2113 
2114 #define cond_resched_rwlock_read(lock) ({					\
2115 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2116 	__cond_resched_rwlock_read(lock);					\
2117 })
2118 
2119 #define cond_resched_rwlock_write(lock) ({					\
2120 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2121 	__cond_resched_rwlock_write(lock);					\
2122 })
2123 
2124 static inline void cond_resched_rcu(void)
2125 {
2126 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2127 	rcu_read_unlock();
2128 	cond_resched();
2129 	rcu_read_lock();
2130 #endif
2131 }
2132 
2133 #ifdef CONFIG_PREEMPT_DYNAMIC
2134 
2135 extern bool preempt_model_none(void);
2136 extern bool preempt_model_voluntary(void);
2137 extern bool preempt_model_full(void);
2138 
2139 #else
2140 
2141 static inline bool preempt_model_none(void)
2142 {
2143 	return IS_ENABLED(CONFIG_PREEMPT_NONE);
2144 }
2145 static inline bool preempt_model_voluntary(void)
2146 {
2147 	return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2148 }
2149 static inline bool preempt_model_full(void)
2150 {
2151 	return IS_ENABLED(CONFIG_PREEMPT);
2152 }
2153 
2154 #endif
2155 
2156 static inline bool preempt_model_rt(void)
2157 {
2158 	return IS_ENABLED(CONFIG_PREEMPT_RT);
2159 }
2160 
2161 /*
2162  * Does the preemption model allow non-cooperative preemption?
2163  *
2164  * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2165  * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2166  * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2167  * PREEMPT_NONE model.
2168  */
2169 static inline bool preempt_model_preemptible(void)
2170 {
2171 	return preempt_model_full() || preempt_model_rt();
2172 }
2173 
2174 /*
2175  * Does a critical section need to be broken due to another
2176  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2177  * but a general need for low latency)
2178  */
2179 static inline int spin_needbreak(spinlock_t *lock)
2180 {
2181 #ifdef CONFIG_PREEMPTION
2182 	return spin_is_contended(lock);
2183 #else
2184 	return 0;
2185 #endif
2186 }
2187 
2188 /*
2189  * Check if a rwlock is contended.
2190  * Returns non-zero if there is another task waiting on the rwlock.
2191  * Returns zero if the lock is not contended or the system / underlying
2192  * rwlock implementation does not support contention detection.
2193  * Technically does not depend on CONFIG_PREEMPTION, but a general need
2194  * for low latency.
2195  */
2196 static inline int rwlock_needbreak(rwlock_t *lock)
2197 {
2198 #ifdef CONFIG_PREEMPTION
2199 	return rwlock_is_contended(lock);
2200 #else
2201 	return 0;
2202 #endif
2203 }
2204 
2205 static __always_inline bool need_resched(void)
2206 {
2207 	return unlikely(tif_need_resched());
2208 }
2209 
2210 /*
2211  * Wrappers for p->thread_info->cpu access. No-op on UP.
2212  */
2213 #ifdef CONFIG_SMP
2214 
2215 static inline unsigned int task_cpu(const struct task_struct *p)
2216 {
2217 	return READ_ONCE(task_thread_info(p)->cpu);
2218 }
2219 
2220 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2221 
2222 #else
2223 
2224 static inline unsigned int task_cpu(const struct task_struct *p)
2225 {
2226 	return 0;
2227 }
2228 
2229 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2230 {
2231 }
2232 
2233 #endif /* CONFIG_SMP */
2234 
2235 extern bool sched_task_on_rq(struct task_struct *p);
2236 extern unsigned long get_wchan(struct task_struct *p);
2237 
2238 /*
2239  * In order to reduce various lock holder preemption latencies provide an
2240  * interface to see if a vCPU is currently running or not.
2241  *
2242  * This allows us to terminate optimistic spin loops and block, analogous to
2243  * the native optimistic spin heuristic of testing if the lock owner task is
2244  * running or not.
2245  */
2246 #ifndef vcpu_is_preempted
2247 static inline bool vcpu_is_preempted(int cpu)
2248 {
2249 	return false;
2250 }
2251 #endif
2252 
2253 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2254 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2255 
2256 #ifndef TASK_SIZE_OF
2257 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2258 #endif
2259 
2260 #ifdef CONFIG_SMP
2261 static inline bool owner_on_cpu(struct task_struct *owner)
2262 {
2263 	/*
2264 	 * As lock holder preemption issue, we both skip spinning if
2265 	 * task is not on cpu or its cpu is preempted
2266 	 */
2267 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2268 }
2269 
2270 /* Returns effective CPU energy utilization, as seen by the scheduler */
2271 unsigned long sched_cpu_util(int cpu, unsigned long max);
2272 #endif /* CONFIG_SMP */
2273 
2274 #ifdef CONFIG_RSEQ
2275 
2276 /*
2277  * Map the event mask on the user-space ABI enum rseq_cs_flags
2278  * for direct mask checks.
2279  */
2280 enum rseq_event_mask_bits {
2281 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2282 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2283 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2284 };
2285 
2286 enum rseq_event_mask {
2287 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
2288 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
2289 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
2290 };
2291 
2292 static inline void rseq_set_notify_resume(struct task_struct *t)
2293 {
2294 	if (t->rseq)
2295 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2296 }
2297 
2298 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2299 
2300 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2301 					     struct pt_regs *regs)
2302 {
2303 	if (current->rseq)
2304 		__rseq_handle_notify_resume(ksig, regs);
2305 }
2306 
2307 static inline void rseq_signal_deliver(struct ksignal *ksig,
2308 				       struct pt_regs *regs)
2309 {
2310 	preempt_disable();
2311 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2312 	preempt_enable();
2313 	rseq_handle_notify_resume(ksig, regs);
2314 }
2315 
2316 /* rseq_preempt() requires preemption to be disabled. */
2317 static inline void rseq_preempt(struct task_struct *t)
2318 {
2319 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2320 	rseq_set_notify_resume(t);
2321 }
2322 
2323 /* rseq_migrate() requires preemption to be disabled. */
2324 static inline void rseq_migrate(struct task_struct *t)
2325 {
2326 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2327 	rseq_set_notify_resume(t);
2328 }
2329 
2330 /*
2331  * If parent process has a registered restartable sequences area, the
2332  * child inherits. Unregister rseq for a clone with CLONE_VM set.
2333  */
2334 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2335 {
2336 	if (clone_flags & CLONE_VM) {
2337 		t->rseq = NULL;
2338 		t->rseq_sig = 0;
2339 		t->rseq_event_mask = 0;
2340 	} else {
2341 		t->rseq = current->rseq;
2342 		t->rseq_sig = current->rseq_sig;
2343 		t->rseq_event_mask = current->rseq_event_mask;
2344 	}
2345 }
2346 
2347 static inline void rseq_execve(struct task_struct *t)
2348 {
2349 	t->rseq = NULL;
2350 	t->rseq_sig = 0;
2351 	t->rseq_event_mask = 0;
2352 }
2353 
2354 #else
2355 
2356 static inline void rseq_set_notify_resume(struct task_struct *t)
2357 {
2358 }
2359 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2360 					     struct pt_regs *regs)
2361 {
2362 }
2363 static inline void rseq_signal_deliver(struct ksignal *ksig,
2364 				       struct pt_regs *regs)
2365 {
2366 }
2367 static inline void rseq_preempt(struct task_struct *t)
2368 {
2369 }
2370 static inline void rseq_migrate(struct task_struct *t)
2371 {
2372 }
2373 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2374 {
2375 }
2376 static inline void rseq_execve(struct task_struct *t)
2377 {
2378 }
2379 
2380 #endif
2381 
2382 #ifdef CONFIG_DEBUG_RSEQ
2383 
2384 void rseq_syscall(struct pt_regs *regs);
2385 
2386 #else
2387 
2388 static inline void rseq_syscall(struct pt_regs *regs)
2389 {
2390 }
2391 
2392 #endif
2393 
2394 #ifdef CONFIG_SCHED_CORE
2395 extern void sched_core_free(struct task_struct *tsk);
2396 extern void sched_core_fork(struct task_struct *p);
2397 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2398 				unsigned long uaddr);
2399 #else
2400 static inline void sched_core_free(struct task_struct *tsk) { }
2401 static inline void sched_core_fork(struct task_struct *p) { }
2402 #endif
2403 
2404 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2405 
2406 #endif
2407