xref: /linux/include/linux/sched.h (revision 83439a0f1ce6a592f95e41338320b5f01b98a356)
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			0x00000000
85 #define TASK_INTERRUPTIBLE		0x00000001
86 #define TASK_UNINTERRUPTIBLE		0x00000002
87 #define __TASK_STOPPED			0x00000004
88 #define __TASK_TRACED			0x00000008
89 /* Used in tsk->exit_state: */
90 #define EXIT_DEAD			0x00000010
91 #define EXIT_ZOMBIE			0x00000020
92 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
93 /* Used in tsk->state again: */
94 #define TASK_PARKED			0x00000040
95 #define TASK_DEAD			0x00000080
96 #define TASK_WAKEKILL			0x00000100
97 #define TASK_WAKING			0x00000200
98 #define TASK_NOLOAD			0x00000400
99 #define TASK_NEW			0x00000800
100 #define TASK_RTLOCK_WAIT		0x00001000
101 #define TASK_FREEZABLE			0x00002000
102 #define __TASK_FREEZABLE_UNSAFE	       (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
103 #define TASK_FROZEN			0x00008000
104 #define TASK_STATE_MAX			0x00010000
105 
106 #define TASK_ANY			(TASK_STATE_MAX-1)
107 
108 /*
109  * DO NOT ADD ANY NEW USERS !
110  */
111 #define TASK_FREEZABLE_UNSAFE		(TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
112 
113 /* Convenience macros for the sake of set_current_state: */
114 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
115 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
116 #define TASK_TRACED			__TASK_TRACED
117 
118 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
119 
120 /* Convenience macros for the sake of wake_up(): */
121 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
122 
123 /* get_task_state(): */
124 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
125 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
126 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
127 					 TASK_PARKED)
128 
129 #define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
130 
131 #define task_is_traced(task)		((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
132 #define task_is_stopped(task)		((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
133 #define task_is_stopped_or_traced(task)	((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
134 
135 /*
136  * Special states are those that do not use the normal wait-loop pattern. See
137  * the comment with set_special_state().
138  */
139 #define is_special_task_state(state)				\
140 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
141 
142 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
143 # define debug_normal_state_change(state_value)				\
144 	do {								\
145 		WARN_ON_ONCE(is_special_task_state(state_value));	\
146 		current->task_state_change = _THIS_IP_;			\
147 	} while (0)
148 
149 # define debug_special_state_change(state_value)			\
150 	do {								\
151 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
152 		current->task_state_change = _THIS_IP_;			\
153 	} while (0)
154 
155 # define debug_rtlock_wait_set_state()					\
156 	do {								 \
157 		current->saved_state_change = current->task_state_change;\
158 		current->task_state_change = _THIS_IP_;			 \
159 	} while (0)
160 
161 # define debug_rtlock_wait_restore_state()				\
162 	do {								 \
163 		current->task_state_change = current->saved_state_change;\
164 	} while (0)
165 
166 #else
167 # define debug_normal_state_change(cond)	do { } while (0)
168 # define debug_special_state_change(cond)	do { } while (0)
169 # define debug_rtlock_wait_set_state()		do { } while (0)
170 # define debug_rtlock_wait_restore_state()	do { } while (0)
171 #endif
172 
173 /*
174  * set_current_state() includes a barrier so that the write of current->state
175  * is correctly serialised wrt the caller's subsequent test of whether to
176  * actually sleep:
177  *
178  *   for (;;) {
179  *	set_current_state(TASK_UNINTERRUPTIBLE);
180  *	if (CONDITION)
181  *	   break;
182  *
183  *	schedule();
184  *   }
185  *   __set_current_state(TASK_RUNNING);
186  *
187  * If the caller does not need such serialisation (because, for instance, the
188  * CONDITION test and condition change and wakeup are under the same lock) then
189  * use __set_current_state().
190  *
191  * The above is typically ordered against the wakeup, which does:
192  *
193  *   CONDITION = 1;
194  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
195  *
196  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
197  * accessing p->state.
198  *
199  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
200  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
201  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
202  *
203  * However, with slightly different timing the wakeup TASK_RUNNING store can
204  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
205  * a problem either because that will result in one extra go around the loop
206  * and our @cond test will save the day.
207  *
208  * Also see the comments of try_to_wake_up().
209  */
210 #define __set_current_state(state_value)				\
211 	do {								\
212 		debug_normal_state_change((state_value));		\
213 		WRITE_ONCE(current->__state, (state_value));		\
214 	} while (0)
215 
216 #define set_current_state(state_value)					\
217 	do {								\
218 		debug_normal_state_change((state_value));		\
219 		smp_store_mb(current->__state, (state_value));		\
220 	} while (0)
221 
222 /*
223  * set_special_state() should be used for those states when the blocking task
224  * can not use the regular condition based wait-loop. In that case we must
225  * serialize against wakeups such that any possible in-flight TASK_RUNNING
226  * stores will not collide with our state change.
227  */
228 #define set_special_state(state_value)					\
229 	do {								\
230 		unsigned long flags; /* may shadow */			\
231 									\
232 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
233 		debug_special_state_change((state_value));		\
234 		WRITE_ONCE(current->__state, (state_value));		\
235 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
236 	} while (0)
237 
238 /*
239  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
240  *
241  * RT's spin/rwlock substitutions are state preserving. The state of the
242  * task when blocking on the lock is saved in task_struct::saved_state and
243  * restored after the lock has been acquired.  These operations are
244  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
245  * lock related wakeups while the task is blocked on the lock are
246  * redirected to operate on task_struct::saved_state to ensure that these
247  * are not dropped. On restore task_struct::saved_state is set to
248  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
249  *
250  * The lock operation looks like this:
251  *
252  *	current_save_and_set_rtlock_wait_state();
253  *	for (;;) {
254  *		if (try_lock())
255  *			break;
256  *		raw_spin_unlock_irq(&lock->wait_lock);
257  *		schedule_rtlock();
258  *		raw_spin_lock_irq(&lock->wait_lock);
259  *		set_current_state(TASK_RTLOCK_WAIT);
260  *	}
261  *	current_restore_rtlock_saved_state();
262  */
263 #define current_save_and_set_rtlock_wait_state()			\
264 	do {								\
265 		lockdep_assert_irqs_disabled();				\
266 		raw_spin_lock(&current->pi_lock);			\
267 		current->saved_state = current->__state;		\
268 		debug_rtlock_wait_set_state();				\
269 		WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);		\
270 		raw_spin_unlock(&current->pi_lock);			\
271 	} while (0);
272 
273 #define current_restore_rtlock_saved_state()				\
274 	do {								\
275 		lockdep_assert_irqs_disabled();				\
276 		raw_spin_lock(&current->pi_lock);			\
277 		debug_rtlock_wait_restore_state();			\
278 		WRITE_ONCE(current->__state, current->saved_state);	\
279 		current->saved_state = TASK_RUNNING;			\
280 		raw_spin_unlock(&current->pi_lock);			\
281 	} while (0);
282 
283 #define get_current_state()	READ_ONCE(current->__state)
284 
285 /*
286  * Define the task command name length as enum, then it can be visible to
287  * BPF programs.
288  */
289 enum {
290 	TASK_COMM_LEN = 16,
291 };
292 
293 extern void scheduler_tick(void);
294 
295 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
296 
297 extern long schedule_timeout(long timeout);
298 extern long schedule_timeout_interruptible(long timeout);
299 extern long schedule_timeout_killable(long timeout);
300 extern long schedule_timeout_uninterruptible(long timeout);
301 extern long schedule_timeout_idle(long timeout);
302 asmlinkage void schedule(void);
303 extern void schedule_preempt_disabled(void);
304 asmlinkage void preempt_schedule_irq(void);
305 #ifdef CONFIG_PREEMPT_RT
306  extern void schedule_rtlock(void);
307 #endif
308 
309 extern int __must_check io_schedule_prepare(void);
310 extern void io_schedule_finish(int token);
311 extern long io_schedule_timeout(long timeout);
312 extern void io_schedule(void);
313 
314 /**
315  * struct prev_cputime - snapshot of system and user cputime
316  * @utime: time spent in user mode
317  * @stime: time spent in system mode
318  * @lock: protects the above two fields
319  *
320  * Stores previous user/system time values such that we can guarantee
321  * monotonicity.
322  */
323 struct prev_cputime {
324 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
325 	u64				utime;
326 	u64				stime;
327 	raw_spinlock_t			lock;
328 #endif
329 };
330 
331 enum vtime_state {
332 	/* Task is sleeping or running in a CPU with VTIME inactive: */
333 	VTIME_INACTIVE = 0,
334 	/* Task is idle */
335 	VTIME_IDLE,
336 	/* Task runs in kernelspace in a CPU with VTIME active: */
337 	VTIME_SYS,
338 	/* Task runs in userspace in a CPU with VTIME active: */
339 	VTIME_USER,
340 	/* Task runs as guests in a CPU with VTIME active: */
341 	VTIME_GUEST,
342 };
343 
344 struct vtime {
345 	seqcount_t		seqcount;
346 	unsigned long long	starttime;
347 	enum vtime_state	state;
348 	unsigned int		cpu;
349 	u64			utime;
350 	u64			stime;
351 	u64			gtime;
352 };
353 
354 /*
355  * Utilization clamp constraints.
356  * @UCLAMP_MIN:	Minimum utilization
357  * @UCLAMP_MAX:	Maximum utilization
358  * @UCLAMP_CNT:	Utilization clamp constraints count
359  */
360 enum uclamp_id {
361 	UCLAMP_MIN = 0,
362 	UCLAMP_MAX,
363 	UCLAMP_CNT
364 };
365 
366 #ifdef CONFIG_SMP
367 extern struct root_domain def_root_domain;
368 extern struct mutex sched_domains_mutex;
369 #endif
370 
371 struct sched_info {
372 #ifdef CONFIG_SCHED_INFO
373 	/* Cumulative counters: */
374 
375 	/* # of times we have run on this CPU: */
376 	unsigned long			pcount;
377 
378 	/* Time spent waiting on a runqueue: */
379 	unsigned long long		run_delay;
380 
381 	/* Timestamps: */
382 
383 	/* When did we last run on a CPU? */
384 	unsigned long long		last_arrival;
385 
386 	/* When were we last queued to run? */
387 	unsigned long long		last_queued;
388 
389 #endif /* CONFIG_SCHED_INFO */
390 };
391 
392 /*
393  * Integer metrics need fixed point arithmetic, e.g., sched/fair
394  * has a few: load, load_avg, util_avg, freq, and capacity.
395  *
396  * We define a basic fixed point arithmetic range, and then formalize
397  * all these metrics based on that basic range.
398  */
399 # define SCHED_FIXEDPOINT_SHIFT		10
400 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
401 
402 /* Increase resolution of cpu_capacity calculations */
403 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
404 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
405 
406 struct load_weight {
407 	unsigned long			weight;
408 	u32				inv_weight;
409 };
410 
411 /**
412  * struct util_est - Estimation utilization of FAIR tasks
413  * @enqueued: instantaneous estimated utilization of a task/cpu
414  * @ewma:     the Exponential Weighted Moving Average (EWMA)
415  *            utilization of a task
416  *
417  * Support data structure to track an Exponential Weighted Moving Average
418  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
419  * average each time a task completes an activation. Sample's weight is chosen
420  * so that the EWMA will be relatively insensitive to transient changes to the
421  * task's workload.
422  *
423  * The enqueued attribute has a slightly different meaning for tasks and cpus:
424  * - task:   the task's util_avg at last task dequeue time
425  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
426  * Thus, the util_est.enqueued of a task represents the contribution on the
427  * estimated utilization of the CPU where that task is currently enqueued.
428  *
429  * Only for tasks we track a moving average of the past instantaneous
430  * estimated utilization. This allows to absorb sporadic drops in utilization
431  * of an otherwise almost periodic task.
432  *
433  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
434  * updates. When a task is dequeued, its util_est should not be updated if its
435  * util_avg has not been updated in the meantime.
436  * This information is mapped into the MSB bit of util_est.enqueued at dequeue
437  * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
438  * for a task) it is safe to use MSB.
439  */
440 struct util_est {
441 	unsigned int			enqueued;
442 	unsigned int			ewma;
443 #define UTIL_EST_WEIGHT_SHIFT		2
444 #define UTIL_AVG_UNCHANGED		0x80000000
445 } __attribute__((__aligned__(sizeof(u64))));
446 
447 /*
448  * The load/runnable/util_avg accumulates an infinite geometric series
449  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
450  *
451  * [load_avg definition]
452  *
453  *   load_avg = runnable% * scale_load_down(load)
454  *
455  * [runnable_avg definition]
456  *
457  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
458  *
459  * [util_avg definition]
460  *
461  *   util_avg = running% * SCHED_CAPACITY_SCALE
462  *
463  * where runnable% is the time ratio that a sched_entity is runnable and
464  * running% the time ratio that a sched_entity is running.
465  *
466  * For cfs_rq, they are the aggregated values of all runnable and blocked
467  * sched_entities.
468  *
469  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
470  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
471  * for computing those signals (see update_rq_clock_pelt())
472  *
473  * N.B., the above ratios (runnable% and running%) themselves are in the
474  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
475  * to as large a range as necessary. This is for example reflected by
476  * util_avg's SCHED_CAPACITY_SCALE.
477  *
478  * [Overflow issue]
479  *
480  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
481  * with the highest load (=88761), always runnable on a single cfs_rq,
482  * and should not overflow as the number already hits PID_MAX_LIMIT.
483  *
484  * For all other cases (including 32-bit kernels), struct load_weight's
485  * weight will overflow first before we do, because:
486  *
487  *    Max(load_avg) <= Max(load.weight)
488  *
489  * Then it is the load_weight's responsibility to consider overflow
490  * issues.
491  */
492 struct sched_avg {
493 	u64				last_update_time;
494 	u64				load_sum;
495 	u64				runnable_sum;
496 	u32				util_sum;
497 	u32				period_contrib;
498 	unsigned long			load_avg;
499 	unsigned long			runnable_avg;
500 	unsigned long			util_avg;
501 	struct util_est			util_est;
502 } ____cacheline_aligned;
503 
504 struct sched_statistics {
505 #ifdef CONFIG_SCHEDSTATS
506 	u64				wait_start;
507 	u64				wait_max;
508 	u64				wait_count;
509 	u64				wait_sum;
510 	u64				iowait_count;
511 	u64				iowait_sum;
512 
513 	u64				sleep_start;
514 	u64				sleep_max;
515 	s64				sum_sleep_runtime;
516 
517 	u64				block_start;
518 	u64				block_max;
519 	s64				sum_block_runtime;
520 
521 	u64				exec_max;
522 	u64				slice_max;
523 
524 	u64				nr_migrations_cold;
525 	u64				nr_failed_migrations_affine;
526 	u64				nr_failed_migrations_running;
527 	u64				nr_failed_migrations_hot;
528 	u64				nr_forced_migrations;
529 
530 	u64				nr_wakeups;
531 	u64				nr_wakeups_sync;
532 	u64				nr_wakeups_migrate;
533 	u64				nr_wakeups_local;
534 	u64				nr_wakeups_remote;
535 	u64				nr_wakeups_affine;
536 	u64				nr_wakeups_affine_attempts;
537 	u64				nr_wakeups_passive;
538 	u64				nr_wakeups_idle;
539 
540 #ifdef CONFIG_SCHED_CORE
541 	u64				core_forceidle_sum;
542 #endif
543 #endif /* CONFIG_SCHEDSTATS */
544 } ____cacheline_aligned;
545 
546 struct sched_entity {
547 	/* For load-balancing: */
548 	struct load_weight		load;
549 	struct rb_node			run_node;
550 	struct list_head		group_node;
551 	unsigned int			on_rq;
552 
553 	u64				exec_start;
554 	u64				sum_exec_runtime;
555 	u64				vruntime;
556 	u64				prev_sum_exec_runtime;
557 
558 	u64				nr_migrations;
559 
560 #ifdef CONFIG_FAIR_GROUP_SCHED
561 	int				depth;
562 	struct sched_entity		*parent;
563 	/* rq on which this entity is (to be) queued: */
564 	struct cfs_rq			*cfs_rq;
565 	/* rq "owned" by this entity/group: */
566 	struct cfs_rq			*my_q;
567 	/* cached value of my_q->h_nr_running */
568 	unsigned long			runnable_weight;
569 #endif
570 
571 #ifdef CONFIG_SMP
572 	/*
573 	 * Per entity load average tracking.
574 	 *
575 	 * Put into separate cache line so it does not
576 	 * collide with read-mostly values above.
577 	 */
578 	struct sched_avg		avg;
579 #endif
580 };
581 
582 struct sched_rt_entity {
583 	struct list_head		run_list;
584 	unsigned long			timeout;
585 	unsigned long			watchdog_stamp;
586 	unsigned int			time_slice;
587 	unsigned short			on_rq;
588 	unsigned short			on_list;
589 
590 	struct sched_rt_entity		*back;
591 #ifdef CONFIG_RT_GROUP_SCHED
592 	struct sched_rt_entity		*parent;
593 	/* rq on which this entity is (to be) queued: */
594 	struct rt_rq			*rt_rq;
595 	/* rq "owned" by this entity/group: */
596 	struct rt_rq			*my_q;
597 #endif
598 } __randomize_layout;
599 
600 struct sched_dl_entity {
601 	struct rb_node			rb_node;
602 
603 	/*
604 	 * Original scheduling parameters. Copied here from sched_attr
605 	 * during sched_setattr(), they will remain the same until
606 	 * the next sched_setattr().
607 	 */
608 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
609 	u64				dl_deadline;	/* Relative deadline of each instance	*/
610 	u64				dl_period;	/* Separation of two instances (period) */
611 	u64				dl_bw;		/* dl_runtime / dl_period		*/
612 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
613 
614 	/*
615 	 * Actual scheduling parameters. Initialized with the values above,
616 	 * they are continuously updated during task execution. Note that
617 	 * the remaining runtime could be < 0 in case we are in overrun.
618 	 */
619 	s64				runtime;	/* Remaining runtime for this instance	*/
620 	u64				deadline;	/* Absolute deadline for this instance	*/
621 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
622 
623 	/*
624 	 * Some bool flags:
625 	 *
626 	 * @dl_throttled tells if we exhausted the runtime. If so, the
627 	 * task has to wait for a replenishment to be performed at the
628 	 * next firing of dl_timer.
629 	 *
630 	 * @dl_yielded tells if task gave up the CPU before consuming
631 	 * all its available runtime during the last job.
632 	 *
633 	 * @dl_non_contending tells if the task is inactive while still
634 	 * contributing to the active utilization. In other words, it
635 	 * indicates if the inactive timer has been armed and its handler
636 	 * has not been executed yet. This flag is useful to avoid race
637 	 * conditions between the inactive timer handler and the wakeup
638 	 * code.
639 	 *
640 	 * @dl_overrun tells if the task asked to be informed about runtime
641 	 * overruns.
642 	 */
643 	unsigned int			dl_throttled      : 1;
644 	unsigned int			dl_yielded        : 1;
645 	unsigned int			dl_non_contending : 1;
646 	unsigned int			dl_overrun	  : 1;
647 
648 	/*
649 	 * Bandwidth enforcement timer. Each -deadline task has its
650 	 * own bandwidth to be enforced, thus we need one timer per task.
651 	 */
652 	struct hrtimer			dl_timer;
653 
654 	/*
655 	 * Inactive timer, responsible for decreasing the active utilization
656 	 * at the "0-lag time". When a -deadline task blocks, it contributes
657 	 * to GRUB's active utilization until the "0-lag time", hence a
658 	 * timer is needed to decrease the active utilization at the correct
659 	 * time.
660 	 */
661 	struct hrtimer inactive_timer;
662 
663 #ifdef CONFIG_RT_MUTEXES
664 	/*
665 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
666 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
667 	 * to (the original one/itself).
668 	 */
669 	struct sched_dl_entity *pi_se;
670 #endif
671 };
672 
673 #ifdef CONFIG_UCLAMP_TASK
674 /* Number of utilization clamp buckets (shorter alias) */
675 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
676 
677 /*
678  * Utilization clamp for a scheduling entity
679  * @value:		clamp value "assigned" to a se
680  * @bucket_id:		bucket index corresponding to the "assigned" value
681  * @active:		the se is currently refcounted in a rq's bucket
682  * @user_defined:	the requested clamp value comes from user-space
683  *
684  * The bucket_id is the index of the clamp bucket matching the clamp value
685  * which is pre-computed and stored to avoid expensive integer divisions from
686  * the fast path.
687  *
688  * The active bit is set whenever a task has got an "effective" value assigned,
689  * which can be different from the clamp value "requested" from user-space.
690  * This allows to know a task is refcounted in the rq's bucket corresponding
691  * to the "effective" bucket_id.
692  *
693  * The user_defined bit is set whenever a task has got a task-specific clamp
694  * value requested from userspace, i.e. the system defaults apply to this task
695  * just as a restriction. This allows to relax default clamps when a less
696  * restrictive task-specific value has been requested, thus allowing to
697  * implement a "nice" semantic. For example, a task running with a 20%
698  * default boost can still drop its own boosting to 0%.
699  */
700 struct uclamp_se {
701 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
702 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
703 	unsigned int active		: 1;
704 	unsigned int user_defined	: 1;
705 };
706 #endif /* CONFIG_UCLAMP_TASK */
707 
708 union rcu_special {
709 	struct {
710 		u8			blocked;
711 		u8			need_qs;
712 		u8			exp_hint; /* Hint for performance. */
713 		u8			need_mb; /* Readers need smp_mb(). */
714 	} b; /* Bits. */
715 	u32 s; /* Set of bits. */
716 };
717 
718 enum perf_event_task_context {
719 	perf_invalid_context = -1,
720 	perf_hw_context = 0,
721 	perf_sw_context,
722 	perf_nr_task_contexts,
723 };
724 
725 struct wake_q_node {
726 	struct wake_q_node *next;
727 };
728 
729 struct kmap_ctrl {
730 #ifdef CONFIG_KMAP_LOCAL
731 	int				idx;
732 	pte_t				pteval[KM_MAX_IDX];
733 #endif
734 };
735 
736 struct task_struct {
737 #ifdef CONFIG_THREAD_INFO_IN_TASK
738 	/*
739 	 * For reasons of header soup (see current_thread_info()), this
740 	 * must be the first element of task_struct.
741 	 */
742 	struct thread_info		thread_info;
743 #endif
744 	unsigned int			__state;
745 
746 #ifdef CONFIG_PREEMPT_RT
747 	/* saved state for "spinlock sleepers" */
748 	unsigned int			saved_state;
749 #endif
750 
751 	/*
752 	 * This begins the randomizable portion of task_struct. Only
753 	 * scheduling-critical items should be added above here.
754 	 */
755 	randomized_struct_fields_start
756 
757 	void				*stack;
758 	refcount_t			usage;
759 	/* Per task flags (PF_*), defined further below: */
760 	unsigned int			flags;
761 	unsigned int			ptrace;
762 
763 #ifdef CONFIG_SMP
764 	int				on_cpu;
765 	struct __call_single_node	wake_entry;
766 	unsigned int			wakee_flips;
767 	unsigned long			wakee_flip_decay_ts;
768 	struct task_struct		*last_wakee;
769 
770 	/*
771 	 * recent_used_cpu is initially set as the last CPU used by a task
772 	 * that wakes affine another task. Waker/wakee relationships can
773 	 * push tasks around a CPU where each wakeup moves to the next one.
774 	 * Tracking a recently used CPU allows a quick search for a recently
775 	 * used CPU that may be idle.
776 	 */
777 	int				recent_used_cpu;
778 	int				wake_cpu;
779 #endif
780 	int				on_rq;
781 
782 	int				prio;
783 	int				static_prio;
784 	int				normal_prio;
785 	unsigned int			rt_priority;
786 
787 	struct sched_entity		se;
788 	struct sched_rt_entity		rt;
789 	struct sched_dl_entity		dl;
790 	const struct sched_class	*sched_class;
791 
792 #ifdef CONFIG_SCHED_CORE
793 	struct rb_node			core_node;
794 	unsigned long			core_cookie;
795 	unsigned int			core_occupation;
796 #endif
797 
798 #ifdef CONFIG_CGROUP_SCHED
799 	struct task_group		*sched_task_group;
800 #endif
801 
802 #ifdef CONFIG_UCLAMP_TASK
803 	/*
804 	 * Clamp values requested for a scheduling entity.
805 	 * Must be updated with task_rq_lock() held.
806 	 */
807 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
808 	/*
809 	 * Effective clamp values used for a scheduling entity.
810 	 * Must be updated with task_rq_lock() held.
811 	 */
812 	struct uclamp_se		uclamp[UCLAMP_CNT];
813 #endif
814 
815 	struct sched_statistics         stats;
816 
817 #ifdef CONFIG_PREEMPT_NOTIFIERS
818 	/* List of struct preempt_notifier: */
819 	struct hlist_head		preempt_notifiers;
820 #endif
821 
822 #ifdef CONFIG_BLK_DEV_IO_TRACE
823 	unsigned int			btrace_seq;
824 #endif
825 
826 	unsigned int			policy;
827 	int				nr_cpus_allowed;
828 	const cpumask_t			*cpus_ptr;
829 	cpumask_t			*user_cpus_ptr;
830 	cpumask_t			cpus_mask;
831 	void				*migration_pending;
832 #ifdef CONFIG_SMP
833 	unsigned short			migration_disabled;
834 #endif
835 	unsigned short			migration_flags;
836 
837 #ifdef CONFIG_PREEMPT_RCU
838 	int				rcu_read_lock_nesting;
839 	union rcu_special		rcu_read_unlock_special;
840 	struct list_head		rcu_node_entry;
841 	struct rcu_node			*rcu_blocked_node;
842 #endif /* #ifdef CONFIG_PREEMPT_RCU */
843 
844 #ifdef CONFIG_TASKS_RCU
845 	unsigned long			rcu_tasks_nvcsw;
846 	u8				rcu_tasks_holdout;
847 	u8				rcu_tasks_idx;
848 	int				rcu_tasks_idle_cpu;
849 	struct list_head		rcu_tasks_holdout_list;
850 #endif /* #ifdef CONFIG_TASKS_RCU */
851 
852 #ifdef CONFIG_TASKS_TRACE_RCU
853 	int				trc_reader_nesting;
854 	int				trc_ipi_to_cpu;
855 	union rcu_special		trc_reader_special;
856 	struct list_head		trc_holdout_list;
857 	struct list_head		trc_blkd_node;
858 	int				trc_blkd_cpu;
859 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
860 
861 	struct sched_info		sched_info;
862 
863 	struct list_head		tasks;
864 #ifdef CONFIG_SMP
865 	struct plist_node		pushable_tasks;
866 	struct rb_node			pushable_dl_tasks;
867 #endif
868 
869 	struct mm_struct		*mm;
870 	struct mm_struct		*active_mm;
871 
872 	/* Per-thread vma caching: */
873 	struct vmacache			vmacache;
874 
875 #ifdef SPLIT_RSS_COUNTING
876 	struct task_rss_stat		rss_stat;
877 #endif
878 	int				exit_state;
879 	int				exit_code;
880 	int				exit_signal;
881 	/* The signal sent when the parent dies: */
882 	int				pdeath_signal;
883 	/* JOBCTL_*, siglock protected: */
884 	unsigned long			jobctl;
885 
886 	/* Used for emulating ABI behavior of previous Linux versions: */
887 	unsigned int			personality;
888 
889 	/* Scheduler bits, serialized by scheduler locks: */
890 	unsigned			sched_reset_on_fork:1;
891 	unsigned			sched_contributes_to_load:1;
892 	unsigned			sched_migrated:1;
893 #ifdef CONFIG_PSI
894 	unsigned			sched_psi_wake_requeue:1;
895 #endif
896 
897 	/* Force alignment to the next boundary: */
898 	unsigned			:0;
899 
900 	/* Unserialized, strictly 'current' */
901 
902 	/*
903 	 * This field must not be in the scheduler word above due to wakelist
904 	 * queueing no longer being serialized by p->on_cpu. However:
905 	 *
906 	 * p->XXX = X;			ttwu()
907 	 * schedule()			  if (p->on_rq && ..) // false
908 	 *   smp_mb__after_spinlock();	  if (smp_load_acquire(&p->on_cpu) && //true
909 	 *   deactivate_task()		      ttwu_queue_wakelist())
910 	 *     p->on_rq = 0;			p->sched_remote_wakeup = Y;
911 	 *
912 	 * guarantees all stores of 'current' are visible before
913 	 * ->sched_remote_wakeup gets used, so it can be in this word.
914 	 */
915 	unsigned			sched_remote_wakeup:1;
916 
917 	/* Bit to tell LSMs we're in execve(): */
918 	unsigned			in_execve:1;
919 	unsigned			in_iowait:1;
920 #ifndef TIF_RESTORE_SIGMASK
921 	unsigned			restore_sigmask:1;
922 #endif
923 #ifdef CONFIG_MEMCG
924 	unsigned			in_user_fault:1;
925 #endif
926 #ifdef CONFIG_COMPAT_BRK
927 	unsigned			brk_randomized:1;
928 #endif
929 #ifdef CONFIG_CGROUPS
930 	/* disallow userland-initiated cgroup migration */
931 	unsigned			no_cgroup_migration:1;
932 	/* task is frozen/stopped (used by the cgroup freezer) */
933 	unsigned			frozen:1;
934 #endif
935 #ifdef CONFIG_BLK_CGROUP
936 	unsigned			use_memdelay:1;
937 #endif
938 #ifdef CONFIG_PSI
939 	/* Stalled due to lack of memory */
940 	unsigned			in_memstall:1;
941 #endif
942 #ifdef CONFIG_PAGE_OWNER
943 	/* Used by page_owner=on to detect recursion in page tracking. */
944 	unsigned			in_page_owner:1;
945 #endif
946 #ifdef CONFIG_EVENTFD
947 	/* Recursion prevention for eventfd_signal() */
948 	unsigned			in_eventfd:1;
949 #endif
950 #ifdef CONFIG_IOMMU_SVA
951 	unsigned			pasid_activated:1;
952 #endif
953 #ifdef	CONFIG_CPU_SUP_INTEL
954 	unsigned			reported_split_lock:1;
955 #endif
956 
957 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
958 
959 	struct restart_block		restart_block;
960 
961 	pid_t				pid;
962 	pid_t				tgid;
963 
964 #ifdef CONFIG_STACKPROTECTOR
965 	/* Canary value for the -fstack-protector GCC feature: */
966 	unsigned long			stack_canary;
967 #endif
968 	/*
969 	 * Pointers to the (original) parent process, youngest child, younger sibling,
970 	 * older sibling, respectively.  (p->father can be replaced with
971 	 * p->real_parent->pid)
972 	 */
973 
974 	/* Real parent process: */
975 	struct task_struct __rcu	*real_parent;
976 
977 	/* Recipient of SIGCHLD, wait4() reports: */
978 	struct task_struct __rcu	*parent;
979 
980 	/*
981 	 * Children/sibling form the list of natural children:
982 	 */
983 	struct list_head		children;
984 	struct list_head		sibling;
985 	struct task_struct		*group_leader;
986 
987 	/*
988 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
989 	 *
990 	 * This includes both natural children and PTRACE_ATTACH targets.
991 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
992 	 */
993 	struct list_head		ptraced;
994 	struct list_head		ptrace_entry;
995 
996 	/* PID/PID hash table linkage. */
997 	struct pid			*thread_pid;
998 	struct hlist_node		pid_links[PIDTYPE_MAX];
999 	struct list_head		thread_group;
1000 	struct list_head		thread_node;
1001 
1002 	struct completion		*vfork_done;
1003 
1004 	/* CLONE_CHILD_SETTID: */
1005 	int __user			*set_child_tid;
1006 
1007 	/* CLONE_CHILD_CLEARTID: */
1008 	int __user			*clear_child_tid;
1009 
1010 	/* PF_KTHREAD | PF_IO_WORKER */
1011 	void				*worker_private;
1012 
1013 	u64				utime;
1014 	u64				stime;
1015 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1016 	u64				utimescaled;
1017 	u64				stimescaled;
1018 #endif
1019 	u64				gtime;
1020 	struct prev_cputime		prev_cputime;
1021 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1022 	struct vtime			vtime;
1023 #endif
1024 
1025 #ifdef CONFIG_NO_HZ_FULL
1026 	atomic_t			tick_dep_mask;
1027 #endif
1028 	/* Context switch counts: */
1029 	unsigned long			nvcsw;
1030 	unsigned long			nivcsw;
1031 
1032 	/* Monotonic time in nsecs: */
1033 	u64				start_time;
1034 
1035 	/* Boot based time in nsecs: */
1036 	u64				start_boottime;
1037 
1038 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1039 	unsigned long			min_flt;
1040 	unsigned long			maj_flt;
1041 
1042 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1043 	struct posix_cputimers		posix_cputimers;
1044 
1045 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1046 	struct posix_cputimers_work	posix_cputimers_work;
1047 #endif
1048 
1049 	/* Process credentials: */
1050 
1051 	/* Tracer's credentials at attach: */
1052 	const struct cred __rcu		*ptracer_cred;
1053 
1054 	/* Objective and real subjective task credentials (COW): */
1055 	const struct cred __rcu		*real_cred;
1056 
1057 	/* Effective (overridable) subjective task credentials (COW): */
1058 	const struct cred __rcu		*cred;
1059 
1060 #ifdef CONFIG_KEYS
1061 	/* Cached requested key. */
1062 	struct key			*cached_requested_key;
1063 #endif
1064 
1065 	/*
1066 	 * executable name, excluding path.
1067 	 *
1068 	 * - normally initialized setup_new_exec()
1069 	 * - access it with [gs]et_task_comm()
1070 	 * - lock it with task_lock()
1071 	 */
1072 	char				comm[TASK_COMM_LEN];
1073 
1074 	struct nameidata		*nameidata;
1075 
1076 #ifdef CONFIG_SYSVIPC
1077 	struct sysv_sem			sysvsem;
1078 	struct sysv_shm			sysvshm;
1079 #endif
1080 #ifdef CONFIG_DETECT_HUNG_TASK
1081 	unsigned long			last_switch_count;
1082 	unsigned long			last_switch_time;
1083 #endif
1084 	/* Filesystem information: */
1085 	struct fs_struct		*fs;
1086 
1087 	/* Open file information: */
1088 	struct files_struct		*files;
1089 
1090 #ifdef CONFIG_IO_URING
1091 	struct io_uring_task		*io_uring;
1092 #endif
1093 
1094 	/* Namespaces: */
1095 	struct nsproxy			*nsproxy;
1096 
1097 	/* Signal handlers: */
1098 	struct signal_struct		*signal;
1099 	struct sighand_struct __rcu		*sighand;
1100 	sigset_t			blocked;
1101 	sigset_t			real_blocked;
1102 	/* Restored if set_restore_sigmask() was used: */
1103 	sigset_t			saved_sigmask;
1104 	struct sigpending		pending;
1105 	unsigned long			sas_ss_sp;
1106 	size_t				sas_ss_size;
1107 	unsigned int			sas_ss_flags;
1108 
1109 	struct callback_head		*task_works;
1110 
1111 #ifdef CONFIG_AUDIT
1112 #ifdef CONFIG_AUDITSYSCALL
1113 	struct audit_context		*audit_context;
1114 #endif
1115 	kuid_t				loginuid;
1116 	unsigned int			sessionid;
1117 #endif
1118 	struct seccomp			seccomp;
1119 	struct syscall_user_dispatch	syscall_dispatch;
1120 
1121 	/* Thread group tracking: */
1122 	u64				parent_exec_id;
1123 	u64				self_exec_id;
1124 
1125 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1126 	spinlock_t			alloc_lock;
1127 
1128 	/* Protection of the PI data structures: */
1129 	raw_spinlock_t			pi_lock;
1130 
1131 	struct wake_q_node		wake_q;
1132 
1133 #ifdef CONFIG_RT_MUTEXES
1134 	/* PI waiters blocked on a rt_mutex held by this task: */
1135 	struct rb_root_cached		pi_waiters;
1136 	/* Updated under owner's pi_lock and rq lock */
1137 	struct task_struct		*pi_top_task;
1138 	/* Deadlock detection and priority inheritance handling: */
1139 	struct rt_mutex_waiter		*pi_blocked_on;
1140 #endif
1141 
1142 #ifdef CONFIG_DEBUG_MUTEXES
1143 	/* Mutex deadlock detection: */
1144 	struct mutex_waiter		*blocked_on;
1145 #endif
1146 
1147 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1148 	int				non_block_count;
1149 #endif
1150 
1151 #ifdef CONFIG_TRACE_IRQFLAGS
1152 	struct irqtrace_events		irqtrace;
1153 	unsigned int			hardirq_threaded;
1154 	u64				hardirq_chain_key;
1155 	int				softirqs_enabled;
1156 	int				softirq_context;
1157 	int				irq_config;
1158 #endif
1159 #ifdef CONFIG_PREEMPT_RT
1160 	int				softirq_disable_cnt;
1161 #endif
1162 
1163 #ifdef CONFIG_LOCKDEP
1164 # define MAX_LOCK_DEPTH			48UL
1165 	u64				curr_chain_key;
1166 	int				lockdep_depth;
1167 	unsigned int			lockdep_recursion;
1168 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1169 #endif
1170 
1171 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1172 	unsigned int			in_ubsan;
1173 #endif
1174 
1175 	/* Journalling filesystem info: */
1176 	void				*journal_info;
1177 
1178 	/* Stacked block device info: */
1179 	struct bio_list			*bio_list;
1180 
1181 	/* Stack plugging: */
1182 	struct blk_plug			*plug;
1183 
1184 	/* VM state: */
1185 	struct reclaim_state		*reclaim_state;
1186 
1187 	struct backing_dev_info		*backing_dev_info;
1188 
1189 	struct io_context		*io_context;
1190 
1191 #ifdef CONFIG_COMPACTION
1192 	struct capture_control		*capture_control;
1193 #endif
1194 	/* Ptrace state: */
1195 	unsigned long			ptrace_message;
1196 	kernel_siginfo_t		*last_siginfo;
1197 
1198 	struct task_io_accounting	ioac;
1199 #ifdef CONFIG_PSI
1200 	/* Pressure stall state */
1201 	unsigned int			psi_flags;
1202 #endif
1203 #ifdef CONFIG_TASK_XACCT
1204 	/* Accumulated RSS usage: */
1205 	u64				acct_rss_mem1;
1206 	/* Accumulated virtual memory usage: */
1207 	u64				acct_vm_mem1;
1208 	/* stime + utime since last update: */
1209 	u64				acct_timexpd;
1210 #endif
1211 #ifdef CONFIG_CPUSETS
1212 	/* Protected by ->alloc_lock: */
1213 	nodemask_t			mems_allowed;
1214 	/* Sequence number to catch updates: */
1215 	seqcount_spinlock_t		mems_allowed_seq;
1216 	int				cpuset_mem_spread_rotor;
1217 	int				cpuset_slab_spread_rotor;
1218 #endif
1219 #ifdef CONFIG_CGROUPS
1220 	/* Control Group info protected by css_set_lock: */
1221 	struct css_set __rcu		*cgroups;
1222 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1223 	struct list_head		cg_list;
1224 #endif
1225 #ifdef CONFIG_X86_CPU_RESCTRL
1226 	u32				closid;
1227 	u32				rmid;
1228 #endif
1229 #ifdef CONFIG_FUTEX
1230 	struct robust_list_head __user	*robust_list;
1231 #ifdef CONFIG_COMPAT
1232 	struct compat_robust_list_head __user *compat_robust_list;
1233 #endif
1234 	struct list_head		pi_state_list;
1235 	struct futex_pi_state		*pi_state_cache;
1236 	struct mutex			futex_exit_mutex;
1237 	unsigned int			futex_state;
1238 #endif
1239 #ifdef CONFIG_PERF_EVENTS
1240 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1241 	struct mutex			perf_event_mutex;
1242 	struct list_head		perf_event_list;
1243 #endif
1244 #ifdef CONFIG_DEBUG_PREEMPT
1245 	unsigned long			preempt_disable_ip;
1246 #endif
1247 #ifdef CONFIG_NUMA
1248 	/* Protected by alloc_lock: */
1249 	struct mempolicy		*mempolicy;
1250 	short				il_prev;
1251 	short				pref_node_fork;
1252 #endif
1253 #ifdef CONFIG_NUMA_BALANCING
1254 	int				numa_scan_seq;
1255 	unsigned int			numa_scan_period;
1256 	unsigned int			numa_scan_period_max;
1257 	int				numa_preferred_nid;
1258 	unsigned long			numa_migrate_retry;
1259 	/* Migration stamp: */
1260 	u64				node_stamp;
1261 	u64				last_task_numa_placement;
1262 	u64				last_sum_exec_runtime;
1263 	struct callback_head		numa_work;
1264 
1265 	/*
1266 	 * This pointer is only modified for current in syscall and
1267 	 * pagefault context (and for tasks being destroyed), so it can be read
1268 	 * from any of the following contexts:
1269 	 *  - RCU read-side critical section
1270 	 *  - current->numa_group from everywhere
1271 	 *  - task's runqueue locked, task not running
1272 	 */
1273 	struct numa_group __rcu		*numa_group;
1274 
1275 	/*
1276 	 * numa_faults is an array split into four regions:
1277 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1278 	 * in this precise order.
1279 	 *
1280 	 * faults_memory: Exponential decaying average of faults on a per-node
1281 	 * basis. Scheduling placement decisions are made based on these
1282 	 * counts. The values remain static for the duration of a PTE scan.
1283 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1284 	 * hinting fault was incurred.
1285 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1286 	 * during the current scan window. When the scan completes, the counts
1287 	 * in faults_memory and faults_cpu decay and these values are copied.
1288 	 */
1289 	unsigned long			*numa_faults;
1290 	unsigned long			total_numa_faults;
1291 
1292 	/*
1293 	 * numa_faults_locality tracks if faults recorded during the last
1294 	 * scan window were remote/local or failed to migrate. The task scan
1295 	 * period is adapted based on the locality of the faults with different
1296 	 * weights depending on whether they were shared or private faults
1297 	 */
1298 	unsigned long			numa_faults_locality[3];
1299 
1300 	unsigned long			numa_pages_migrated;
1301 #endif /* CONFIG_NUMA_BALANCING */
1302 
1303 #ifdef CONFIG_RSEQ
1304 	struct rseq __user *rseq;
1305 	u32 rseq_sig;
1306 	/*
1307 	 * RmW on rseq_event_mask must be performed atomically
1308 	 * with respect to preemption.
1309 	 */
1310 	unsigned long rseq_event_mask;
1311 #endif
1312 
1313 	struct tlbflush_unmap_batch	tlb_ubc;
1314 
1315 	union {
1316 		refcount_t		rcu_users;
1317 		struct rcu_head		rcu;
1318 	};
1319 
1320 	/* Cache last used pipe for splice(): */
1321 	struct pipe_inode_info		*splice_pipe;
1322 
1323 	struct page_frag		task_frag;
1324 
1325 #ifdef CONFIG_TASK_DELAY_ACCT
1326 	struct task_delay_info		*delays;
1327 #endif
1328 
1329 #ifdef CONFIG_FAULT_INJECTION
1330 	int				make_it_fail;
1331 	unsigned int			fail_nth;
1332 #endif
1333 	/*
1334 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1335 	 * balance_dirty_pages() for a dirty throttling pause:
1336 	 */
1337 	int				nr_dirtied;
1338 	int				nr_dirtied_pause;
1339 	/* Start of a write-and-pause period: */
1340 	unsigned long			dirty_paused_when;
1341 
1342 #ifdef CONFIG_LATENCYTOP
1343 	int				latency_record_count;
1344 	struct latency_record		latency_record[LT_SAVECOUNT];
1345 #endif
1346 	/*
1347 	 * Time slack values; these are used to round up poll() and
1348 	 * select() etc timeout values. These are in nanoseconds.
1349 	 */
1350 	u64				timer_slack_ns;
1351 	u64				default_timer_slack_ns;
1352 
1353 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1354 	unsigned int			kasan_depth;
1355 #endif
1356 
1357 #ifdef CONFIG_KCSAN
1358 	struct kcsan_ctx		kcsan_ctx;
1359 #ifdef CONFIG_TRACE_IRQFLAGS
1360 	struct irqtrace_events		kcsan_save_irqtrace;
1361 #endif
1362 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1363 	int				kcsan_stack_depth;
1364 #endif
1365 #endif
1366 
1367 #if IS_ENABLED(CONFIG_KUNIT)
1368 	struct kunit			*kunit_test;
1369 #endif
1370 
1371 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1372 	/* Index of current stored address in ret_stack: */
1373 	int				curr_ret_stack;
1374 	int				curr_ret_depth;
1375 
1376 	/* Stack of return addresses for return function tracing: */
1377 	struct ftrace_ret_stack		*ret_stack;
1378 
1379 	/* Timestamp for last schedule: */
1380 	unsigned long long		ftrace_timestamp;
1381 
1382 	/*
1383 	 * Number of functions that haven't been traced
1384 	 * because of depth overrun:
1385 	 */
1386 	atomic_t			trace_overrun;
1387 
1388 	/* Pause tracing: */
1389 	atomic_t			tracing_graph_pause;
1390 #endif
1391 
1392 #ifdef CONFIG_TRACING
1393 	/* Bitmask and counter of trace recursion: */
1394 	unsigned long			trace_recursion;
1395 #endif /* CONFIG_TRACING */
1396 
1397 #ifdef CONFIG_KCOV
1398 	/* See kernel/kcov.c for more details. */
1399 
1400 	/* Coverage collection mode enabled for this task (0 if disabled): */
1401 	unsigned int			kcov_mode;
1402 
1403 	/* Size of the kcov_area: */
1404 	unsigned int			kcov_size;
1405 
1406 	/* Buffer for coverage collection: */
1407 	void				*kcov_area;
1408 
1409 	/* KCOV descriptor wired with this task or NULL: */
1410 	struct kcov			*kcov;
1411 
1412 	/* KCOV common handle for remote coverage collection: */
1413 	u64				kcov_handle;
1414 
1415 	/* KCOV sequence number: */
1416 	int				kcov_sequence;
1417 
1418 	/* Collect coverage from softirq context: */
1419 	unsigned int			kcov_softirq;
1420 #endif
1421 
1422 #ifdef CONFIG_MEMCG
1423 	struct mem_cgroup		*memcg_in_oom;
1424 	gfp_t				memcg_oom_gfp_mask;
1425 	int				memcg_oom_order;
1426 
1427 	/* Number of pages to reclaim on returning to userland: */
1428 	unsigned int			memcg_nr_pages_over_high;
1429 
1430 	/* Used by memcontrol for targeted memcg charge: */
1431 	struct mem_cgroup		*active_memcg;
1432 #endif
1433 
1434 #ifdef CONFIG_BLK_CGROUP
1435 	struct request_queue		*throttle_queue;
1436 #endif
1437 
1438 #ifdef CONFIG_UPROBES
1439 	struct uprobe_task		*utask;
1440 #endif
1441 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1442 	unsigned int			sequential_io;
1443 	unsigned int			sequential_io_avg;
1444 #endif
1445 	struct kmap_ctrl		kmap_ctrl;
1446 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1447 	unsigned long			task_state_change;
1448 # ifdef CONFIG_PREEMPT_RT
1449 	unsigned long			saved_state_change;
1450 # endif
1451 #endif
1452 	int				pagefault_disabled;
1453 #ifdef CONFIG_MMU
1454 	struct task_struct		*oom_reaper_list;
1455 	struct timer_list		oom_reaper_timer;
1456 #endif
1457 #ifdef CONFIG_VMAP_STACK
1458 	struct vm_struct		*stack_vm_area;
1459 #endif
1460 #ifdef CONFIG_THREAD_INFO_IN_TASK
1461 	/* A live task holds one reference: */
1462 	refcount_t			stack_refcount;
1463 #endif
1464 #ifdef CONFIG_LIVEPATCH
1465 	int patch_state;
1466 #endif
1467 #ifdef CONFIG_SECURITY
1468 	/* Used by LSM modules for access restriction: */
1469 	void				*security;
1470 #endif
1471 #ifdef CONFIG_BPF_SYSCALL
1472 	/* Used by BPF task local storage */
1473 	struct bpf_local_storage __rcu	*bpf_storage;
1474 	/* Used for BPF run context */
1475 	struct bpf_run_ctx		*bpf_ctx;
1476 #endif
1477 
1478 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1479 	unsigned long			lowest_stack;
1480 	unsigned long			prev_lowest_stack;
1481 #endif
1482 
1483 #ifdef CONFIG_X86_MCE
1484 	void __user			*mce_vaddr;
1485 	__u64				mce_kflags;
1486 	u64				mce_addr;
1487 	__u64				mce_ripv : 1,
1488 					mce_whole_page : 1,
1489 					__mce_reserved : 62;
1490 	struct callback_head		mce_kill_me;
1491 	int				mce_count;
1492 #endif
1493 
1494 #ifdef CONFIG_KRETPROBES
1495 	struct llist_head               kretprobe_instances;
1496 #endif
1497 #ifdef CONFIG_RETHOOK
1498 	struct llist_head               rethooks;
1499 #endif
1500 
1501 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1502 	/*
1503 	 * If L1D flush is supported on mm context switch
1504 	 * then we use this callback head to queue kill work
1505 	 * to kill tasks that are not running on SMT disabled
1506 	 * cores
1507 	 */
1508 	struct callback_head		l1d_flush_kill;
1509 #endif
1510 
1511 #ifdef CONFIG_RV
1512 	/*
1513 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1514 	 * If we find justification for more monitors, we can think
1515 	 * about adding more or developing a dynamic method. So far,
1516 	 * none of these are justified.
1517 	 */
1518 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1519 #endif
1520 
1521 	/*
1522 	 * New fields for task_struct should be added above here, so that
1523 	 * they are included in the randomized portion of task_struct.
1524 	 */
1525 	randomized_struct_fields_end
1526 
1527 	/* CPU-specific state of this task: */
1528 	struct thread_struct		thread;
1529 
1530 	/*
1531 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1532 	 * structure.  It *MUST* be at the end of 'task_struct'.
1533 	 *
1534 	 * Do not put anything below here!
1535 	 */
1536 };
1537 
1538 static inline struct pid *task_pid(struct task_struct *task)
1539 {
1540 	return task->thread_pid;
1541 }
1542 
1543 /*
1544  * the helpers to get the task's different pids as they are seen
1545  * from various namespaces
1546  *
1547  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1548  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1549  *                     current.
1550  * task_xid_nr_ns()  : id seen from the ns specified;
1551  *
1552  * see also pid_nr() etc in include/linux/pid.h
1553  */
1554 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1555 
1556 static inline pid_t task_pid_nr(struct task_struct *tsk)
1557 {
1558 	return tsk->pid;
1559 }
1560 
1561 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1562 {
1563 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1564 }
1565 
1566 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1567 {
1568 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1569 }
1570 
1571 
1572 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1573 {
1574 	return tsk->tgid;
1575 }
1576 
1577 /**
1578  * pid_alive - check that a task structure is not stale
1579  * @p: Task structure to be checked.
1580  *
1581  * Test if a process is not yet dead (at most zombie state)
1582  * If pid_alive fails, then pointers within the task structure
1583  * can be stale and must not be dereferenced.
1584  *
1585  * Return: 1 if the process is alive. 0 otherwise.
1586  */
1587 static inline int pid_alive(const struct task_struct *p)
1588 {
1589 	return p->thread_pid != NULL;
1590 }
1591 
1592 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1593 {
1594 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1595 }
1596 
1597 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1598 {
1599 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1600 }
1601 
1602 
1603 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1604 {
1605 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1606 }
1607 
1608 static inline pid_t task_session_vnr(struct task_struct *tsk)
1609 {
1610 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1611 }
1612 
1613 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1614 {
1615 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1616 }
1617 
1618 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1619 {
1620 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1621 }
1622 
1623 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1624 {
1625 	pid_t pid = 0;
1626 
1627 	rcu_read_lock();
1628 	if (pid_alive(tsk))
1629 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1630 	rcu_read_unlock();
1631 
1632 	return pid;
1633 }
1634 
1635 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1636 {
1637 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1638 }
1639 
1640 /* Obsolete, do not use: */
1641 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1642 {
1643 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1644 }
1645 
1646 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1647 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1648 
1649 static inline unsigned int __task_state_index(unsigned int tsk_state,
1650 					      unsigned int tsk_exit_state)
1651 {
1652 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1653 
1654 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1655 
1656 	if (tsk_state == TASK_IDLE)
1657 		state = TASK_REPORT_IDLE;
1658 
1659 	/*
1660 	 * We're lying here, but rather than expose a completely new task state
1661 	 * to userspace, we can make this appear as if the task has gone through
1662 	 * a regular rt_mutex_lock() call.
1663 	 */
1664 	if (tsk_state == TASK_RTLOCK_WAIT)
1665 		state = TASK_UNINTERRUPTIBLE;
1666 
1667 	return fls(state);
1668 }
1669 
1670 static inline unsigned int task_state_index(struct task_struct *tsk)
1671 {
1672 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1673 }
1674 
1675 static inline char task_index_to_char(unsigned int state)
1676 {
1677 	static const char state_char[] = "RSDTtXZPI";
1678 
1679 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1680 
1681 	return state_char[state];
1682 }
1683 
1684 static inline char task_state_to_char(struct task_struct *tsk)
1685 {
1686 	return task_index_to_char(task_state_index(tsk));
1687 }
1688 
1689 /**
1690  * is_global_init - check if a task structure is init. Since init
1691  * is free to have sub-threads we need to check tgid.
1692  * @tsk: Task structure to be checked.
1693  *
1694  * Check if a task structure is the first user space task the kernel created.
1695  *
1696  * Return: 1 if the task structure is init. 0 otherwise.
1697  */
1698 static inline int is_global_init(struct task_struct *tsk)
1699 {
1700 	return task_tgid_nr(tsk) == 1;
1701 }
1702 
1703 extern struct pid *cad_pid;
1704 
1705 /*
1706  * Per process flags
1707  */
1708 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1709 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1710 #define PF_EXITING		0x00000004	/* Getting shut down */
1711 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1712 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1713 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1714 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1715 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1716 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1717 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1718 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1719 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1720 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1721 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1722 #define PF__HOLE__00004000	0x00004000
1723 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1724 #define PF__HOLE__00010000	0x00010000
1725 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1726 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1727 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1728 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1729 						 * I am cleaning dirty pages from some other bdi. */
1730 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1731 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1732 #define PF__HOLE__00800000	0x00800000
1733 #define PF__HOLE__01000000	0x01000000
1734 #define PF__HOLE__02000000	0x02000000
1735 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1736 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1737 #define PF_MEMALLOC_PIN		0x10000000	/* Allocation context constrained to zones which allow long term pinning. */
1738 #define PF__HOLE__20000000	0x20000000
1739 #define PF__HOLE__40000000	0x40000000
1740 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1741 
1742 /*
1743  * Only the _current_ task can read/write to tsk->flags, but other
1744  * tasks can access tsk->flags in readonly mode for example
1745  * with tsk_used_math (like during threaded core dumping).
1746  * There is however an exception to this rule during ptrace
1747  * or during fork: the ptracer task is allowed to write to the
1748  * child->flags of its traced child (same goes for fork, the parent
1749  * can write to the child->flags), because we're guaranteed the
1750  * child is not running and in turn not changing child->flags
1751  * at the same time the parent does it.
1752  */
1753 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1754 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1755 #define clear_used_math()			clear_stopped_child_used_math(current)
1756 #define set_used_math()				set_stopped_child_used_math(current)
1757 
1758 #define conditional_stopped_child_used_math(condition, child) \
1759 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1760 
1761 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1762 
1763 #define copy_to_stopped_child_used_math(child) \
1764 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1765 
1766 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1767 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1768 #define used_math()				tsk_used_math(current)
1769 
1770 static __always_inline bool is_percpu_thread(void)
1771 {
1772 #ifdef CONFIG_SMP
1773 	return (current->flags & PF_NO_SETAFFINITY) &&
1774 		(current->nr_cpus_allowed  == 1);
1775 #else
1776 	return true;
1777 #endif
1778 }
1779 
1780 /* Per-process atomic flags. */
1781 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1782 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1783 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1784 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1785 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1786 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1787 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1788 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1789 
1790 #define TASK_PFA_TEST(name, func)					\
1791 	static inline bool task_##func(struct task_struct *p)		\
1792 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1793 
1794 #define TASK_PFA_SET(name, func)					\
1795 	static inline void task_set_##func(struct task_struct *p)	\
1796 	{ set_bit(PFA_##name, &p->atomic_flags); }
1797 
1798 #define TASK_PFA_CLEAR(name, func)					\
1799 	static inline void task_clear_##func(struct task_struct *p)	\
1800 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1801 
1802 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1803 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1804 
1805 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1806 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1807 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1808 
1809 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1810 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1811 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1812 
1813 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1814 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1815 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1816 
1817 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1818 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1819 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1820 
1821 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1822 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1823 
1824 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1825 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1826 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1827 
1828 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1829 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1830 
1831 static inline void
1832 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1833 {
1834 	current->flags &= ~flags;
1835 	current->flags |= orig_flags & flags;
1836 }
1837 
1838 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1839 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effective_cpus);
1840 #ifdef CONFIG_SMP
1841 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1842 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1843 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1844 extern void release_user_cpus_ptr(struct task_struct *p);
1845 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1846 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1847 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1848 #else
1849 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1850 {
1851 }
1852 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1853 {
1854 	if (!cpumask_test_cpu(0, new_mask))
1855 		return -EINVAL;
1856 	return 0;
1857 }
1858 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1859 {
1860 	if (src->user_cpus_ptr)
1861 		return -EINVAL;
1862 	return 0;
1863 }
1864 static inline void release_user_cpus_ptr(struct task_struct *p)
1865 {
1866 	WARN_ON(p->user_cpus_ptr);
1867 }
1868 
1869 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1870 {
1871 	return 0;
1872 }
1873 #endif
1874 
1875 extern int yield_to(struct task_struct *p, bool preempt);
1876 extern void set_user_nice(struct task_struct *p, long nice);
1877 extern int task_prio(const struct task_struct *p);
1878 
1879 /**
1880  * task_nice - return the nice value of a given task.
1881  * @p: the task in question.
1882  *
1883  * Return: The nice value [ -20 ... 0 ... 19 ].
1884  */
1885 static inline int task_nice(const struct task_struct *p)
1886 {
1887 	return PRIO_TO_NICE((p)->static_prio);
1888 }
1889 
1890 extern int can_nice(const struct task_struct *p, const int nice);
1891 extern int task_curr(const struct task_struct *p);
1892 extern int idle_cpu(int cpu);
1893 extern int available_idle_cpu(int cpu);
1894 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1895 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1896 extern void sched_set_fifo(struct task_struct *p);
1897 extern void sched_set_fifo_low(struct task_struct *p);
1898 extern void sched_set_normal(struct task_struct *p, int nice);
1899 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1900 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1901 extern struct task_struct *idle_task(int cpu);
1902 
1903 /**
1904  * is_idle_task - is the specified task an idle task?
1905  * @p: the task in question.
1906  *
1907  * Return: 1 if @p is an idle task. 0 otherwise.
1908  */
1909 static __always_inline bool is_idle_task(const struct task_struct *p)
1910 {
1911 	return !!(p->flags & PF_IDLE);
1912 }
1913 
1914 extern struct task_struct *curr_task(int cpu);
1915 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1916 
1917 void yield(void);
1918 
1919 union thread_union {
1920 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1921 	struct task_struct task;
1922 #endif
1923 #ifndef CONFIG_THREAD_INFO_IN_TASK
1924 	struct thread_info thread_info;
1925 #endif
1926 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1927 };
1928 
1929 #ifndef CONFIG_THREAD_INFO_IN_TASK
1930 extern struct thread_info init_thread_info;
1931 #endif
1932 
1933 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1934 
1935 #ifdef CONFIG_THREAD_INFO_IN_TASK
1936 # define task_thread_info(task)	(&(task)->thread_info)
1937 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1938 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1939 #endif
1940 
1941 /*
1942  * find a task by one of its numerical ids
1943  *
1944  * find_task_by_pid_ns():
1945  *      finds a task by its pid in the specified namespace
1946  * find_task_by_vpid():
1947  *      finds a task by its virtual pid
1948  *
1949  * see also find_vpid() etc in include/linux/pid.h
1950  */
1951 
1952 extern struct task_struct *find_task_by_vpid(pid_t nr);
1953 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1954 
1955 /*
1956  * find a task by its virtual pid and get the task struct
1957  */
1958 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1959 
1960 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1961 extern int wake_up_process(struct task_struct *tsk);
1962 extern void wake_up_new_task(struct task_struct *tsk);
1963 
1964 #ifdef CONFIG_SMP
1965 extern void kick_process(struct task_struct *tsk);
1966 #else
1967 static inline void kick_process(struct task_struct *tsk) { }
1968 #endif
1969 
1970 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1971 
1972 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1973 {
1974 	__set_task_comm(tsk, from, false);
1975 }
1976 
1977 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1978 #define get_task_comm(buf, tsk) ({			\
1979 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1980 	__get_task_comm(buf, sizeof(buf), tsk);		\
1981 })
1982 
1983 #ifdef CONFIG_SMP
1984 static __always_inline void scheduler_ipi(void)
1985 {
1986 	/*
1987 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1988 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1989 	 * this IPI.
1990 	 */
1991 	preempt_fold_need_resched();
1992 }
1993 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1994 #else
1995 static inline void scheduler_ipi(void) { }
1996 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1997 {
1998 	return 1;
1999 }
2000 #endif
2001 
2002 /*
2003  * Set thread flags in other task's structures.
2004  * See asm/thread_info.h for TIF_xxxx flags available:
2005  */
2006 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2007 {
2008 	set_ti_thread_flag(task_thread_info(tsk), flag);
2009 }
2010 
2011 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2012 {
2013 	clear_ti_thread_flag(task_thread_info(tsk), flag);
2014 }
2015 
2016 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2017 					  bool value)
2018 {
2019 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
2020 }
2021 
2022 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2023 {
2024 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2025 }
2026 
2027 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2028 {
2029 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2030 }
2031 
2032 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2033 {
2034 	return test_ti_thread_flag(task_thread_info(tsk), flag);
2035 }
2036 
2037 static inline void set_tsk_need_resched(struct task_struct *tsk)
2038 {
2039 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2040 }
2041 
2042 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2043 {
2044 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2045 }
2046 
2047 static inline int test_tsk_need_resched(struct task_struct *tsk)
2048 {
2049 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2050 }
2051 
2052 /*
2053  * cond_resched() and cond_resched_lock(): latency reduction via
2054  * explicit rescheduling in places that are safe. The return
2055  * value indicates whether a reschedule was done in fact.
2056  * cond_resched_lock() will drop the spinlock before scheduling,
2057  */
2058 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2059 extern int __cond_resched(void);
2060 
2061 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2062 
2063 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2064 
2065 static __always_inline int _cond_resched(void)
2066 {
2067 	return static_call_mod(cond_resched)();
2068 }
2069 
2070 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2071 extern int dynamic_cond_resched(void);
2072 
2073 static __always_inline int _cond_resched(void)
2074 {
2075 	return dynamic_cond_resched();
2076 }
2077 
2078 #else
2079 
2080 static inline int _cond_resched(void)
2081 {
2082 	return __cond_resched();
2083 }
2084 
2085 #endif /* CONFIG_PREEMPT_DYNAMIC */
2086 
2087 #else
2088 
2089 static inline int _cond_resched(void) { return 0; }
2090 
2091 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2092 
2093 #define cond_resched() ({			\
2094 	__might_resched(__FILE__, __LINE__, 0);	\
2095 	_cond_resched();			\
2096 })
2097 
2098 extern int __cond_resched_lock(spinlock_t *lock);
2099 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2100 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2101 
2102 #define MIGHT_RESCHED_RCU_SHIFT		8
2103 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2104 
2105 #ifndef CONFIG_PREEMPT_RT
2106 /*
2107  * Non RT kernels have an elevated preempt count due to the held lock,
2108  * but are not allowed to be inside a RCU read side critical section
2109  */
2110 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2111 #else
2112 /*
2113  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2114  * cond_resched*lock() has to take that into account because it checks for
2115  * preempt_count() and rcu_preempt_depth().
2116  */
2117 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2118 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2119 #endif
2120 
2121 #define cond_resched_lock(lock) ({						\
2122 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2123 	__cond_resched_lock(lock);						\
2124 })
2125 
2126 #define cond_resched_rwlock_read(lock) ({					\
2127 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2128 	__cond_resched_rwlock_read(lock);					\
2129 })
2130 
2131 #define cond_resched_rwlock_write(lock) ({					\
2132 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2133 	__cond_resched_rwlock_write(lock);					\
2134 })
2135 
2136 static inline void cond_resched_rcu(void)
2137 {
2138 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2139 	rcu_read_unlock();
2140 	cond_resched();
2141 	rcu_read_lock();
2142 #endif
2143 }
2144 
2145 #ifdef CONFIG_PREEMPT_DYNAMIC
2146 
2147 extern bool preempt_model_none(void);
2148 extern bool preempt_model_voluntary(void);
2149 extern bool preempt_model_full(void);
2150 
2151 #else
2152 
2153 static inline bool preempt_model_none(void)
2154 {
2155 	return IS_ENABLED(CONFIG_PREEMPT_NONE);
2156 }
2157 static inline bool preempt_model_voluntary(void)
2158 {
2159 	return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2160 }
2161 static inline bool preempt_model_full(void)
2162 {
2163 	return IS_ENABLED(CONFIG_PREEMPT);
2164 }
2165 
2166 #endif
2167 
2168 static inline bool preempt_model_rt(void)
2169 {
2170 	return IS_ENABLED(CONFIG_PREEMPT_RT);
2171 }
2172 
2173 /*
2174  * Does the preemption model allow non-cooperative preemption?
2175  *
2176  * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2177  * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2178  * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2179  * PREEMPT_NONE model.
2180  */
2181 static inline bool preempt_model_preemptible(void)
2182 {
2183 	return preempt_model_full() || preempt_model_rt();
2184 }
2185 
2186 /*
2187  * Does a critical section need to be broken due to another
2188  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2189  * but a general need for low latency)
2190  */
2191 static inline int spin_needbreak(spinlock_t *lock)
2192 {
2193 #ifdef CONFIG_PREEMPTION
2194 	return spin_is_contended(lock);
2195 #else
2196 	return 0;
2197 #endif
2198 }
2199 
2200 /*
2201  * Check if a rwlock is contended.
2202  * Returns non-zero if there is another task waiting on the rwlock.
2203  * Returns zero if the lock is not contended or the system / underlying
2204  * rwlock implementation does not support contention detection.
2205  * Technically does not depend on CONFIG_PREEMPTION, but a general need
2206  * for low latency.
2207  */
2208 static inline int rwlock_needbreak(rwlock_t *lock)
2209 {
2210 #ifdef CONFIG_PREEMPTION
2211 	return rwlock_is_contended(lock);
2212 #else
2213 	return 0;
2214 #endif
2215 }
2216 
2217 static __always_inline bool need_resched(void)
2218 {
2219 	return unlikely(tif_need_resched());
2220 }
2221 
2222 /*
2223  * Wrappers for p->thread_info->cpu access. No-op on UP.
2224  */
2225 #ifdef CONFIG_SMP
2226 
2227 static inline unsigned int task_cpu(const struct task_struct *p)
2228 {
2229 	return READ_ONCE(task_thread_info(p)->cpu);
2230 }
2231 
2232 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2233 
2234 #else
2235 
2236 static inline unsigned int task_cpu(const struct task_struct *p)
2237 {
2238 	return 0;
2239 }
2240 
2241 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2242 {
2243 }
2244 
2245 #endif /* CONFIG_SMP */
2246 
2247 extern bool sched_task_on_rq(struct task_struct *p);
2248 extern unsigned long get_wchan(struct task_struct *p);
2249 extern struct task_struct *cpu_curr_snapshot(int cpu);
2250 
2251 /*
2252  * In order to reduce various lock holder preemption latencies provide an
2253  * interface to see if a vCPU is currently running or not.
2254  *
2255  * This allows us to terminate optimistic spin loops and block, analogous to
2256  * the native optimistic spin heuristic of testing if the lock owner task is
2257  * running or not.
2258  */
2259 #ifndef vcpu_is_preempted
2260 static inline bool vcpu_is_preempted(int cpu)
2261 {
2262 	return false;
2263 }
2264 #endif
2265 
2266 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2267 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2268 
2269 #ifndef TASK_SIZE_OF
2270 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2271 #endif
2272 
2273 #ifdef CONFIG_SMP
2274 static inline bool owner_on_cpu(struct task_struct *owner)
2275 {
2276 	/*
2277 	 * As lock holder preemption issue, we both skip spinning if
2278 	 * task is not on cpu or its cpu is preempted
2279 	 */
2280 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2281 }
2282 
2283 /* Returns effective CPU energy utilization, as seen by the scheduler */
2284 unsigned long sched_cpu_util(int cpu);
2285 #endif /* CONFIG_SMP */
2286 
2287 #ifdef CONFIG_RSEQ
2288 
2289 /*
2290  * Map the event mask on the user-space ABI enum rseq_cs_flags
2291  * for direct mask checks.
2292  */
2293 enum rseq_event_mask_bits {
2294 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2295 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2296 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2297 };
2298 
2299 enum rseq_event_mask {
2300 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
2301 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
2302 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
2303 };
2304 
2305 static inline void rseq_set_notify_resume(struct task_struct *t)
2306 {
2307 	if (t->rseq)
2308 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2309 }
2310 
2311 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2312 
2313 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2314 					     struct pt_regs *regs)
2315 {
2316 	if (current->rseq)
2317 		__rseq_handle_notify_resume(ksig, regs);
2318 }
2319 
2320 static inline void rseq_signal_deliver(struct ksignal *ksig,
2321 				       struct pt_regs *regs)
2322 {
2323 	preempt_disable();
2324 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2325 	preempt_enable();
2326 	rseq_handle_notify_resume(ksig, regs);
2327 }
2328 
2329 /* rseq_preempt() requires preemption to be disabled. */
2330 static inline void rseq_preempt(struct task_struct *t)
2331 {
2332 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2333 	rseq_set_notify_resume(t);
2334 }
2335 
2336 /* rseq_migrate() requires preemption to be disabled. */
2337 static inline void rseq_migrate(struct task_struct *t)
2338 {
2339 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2340 	rseq_set_notify_resume(t);
2341 }
2342 
2343 /*
2344  * If parent process has a registered restartable sequences area, the
2345  * child inherits. Unregister rseq for a clone with CLONE_VM set.
2346  */
2347 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2348 {
2349 	if (clone_flags & CLONE_VM) {
2350 		t->rseq = NULL;
2351 		t->rseq_sig = 0;
2352 		t->rseq_event_mask = 0;
2353 	} else {
2354 		t->rseq = current->rseq;
2355 		t->rseq_sig = current->rseq_sig;
2356 		t->rseq_event_mask = current->rseq_event_mask;
2357 	}
2358 }
2359 
2360 static inline void rseq_execve(struct task_struct *t)
2361 {
2362 	t->rseq = NULL;
2363 	t->rseq_sig = 0;
2364 	t->rseq_event_mask = 0;
2365 }
2366 
2367 #else
2368 
2369 static inline void rseq_set_notify_resume(struct task_struct *t)
2370 {
2371 }
2372 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2373 					     struct pt_regs *regs)
2374 {
2375 }
2376 static inline void rseq_signal_deliver(struct ksignal *ksig,
2377 				       struct pt_regs *regs)
2378 {
2379 }
2380 static inline void rseq_preempt(struct task_struct *t)
2381 {
2382 }
2383 static inline void rseq_migrate(struct task_struct *t)
2384 {
2385 }
2386 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2387 {
2388 }
2389 static inline void rseq_execve(struct task_struct *t)
2390 {
2391 }
2392 
2393 #endif
2394 
2395 #ifdef CONFIG_DEBUG_RSEQ
2396 
2397 void rseq_syscall(struct pt_regs *regs);
2398 
2399 #else
2400 
2401 static inline void rseq_syscall(struct pt_regs *regs)
2402 {
2403 }
2404 
2405 #endif
2406 
2407 #ifdef CONFIG_SCHED_CORE
2408 extern void sched_core_free(struct task_struct *tsk);
2409 extern void sched_core_fork(struct task_struct *p);
2410 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2411 				unsigned long uaddr);
2412 #else
2413 static inline void sched_core_free(struct task_struct *tsk) { }
2414 static inline void sched_core_fork(struct task_struct *p) { }
2415 #endif
2416 
2417 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2418 
2419 #endif
2420