xref: /linux-6.15/include/linux/sched.h (revision d047e32b)
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 #include <asm/processor.h>
14 #include <linux/thread_info.h>
15 #include <linux/preempt.h>
16 #include <linux/cpumask_types.h>
17 
18 #include <linux/cache.h>
19 #include <linux/irqflags_types.h>
20 #include <linux/smp_types.h>
21 #include <linux/pid_types.h>
22 #include <linux/sem_types.h>
23 #include <linux/shm.h>
24 #include <linux/kmsan_types.h>
25 #include <linux/mutex_types.h>
26 #include <linux/plist_types.h>
27 #include <linux/hrtimer_types.h>
28 #include <linux/timer_types.h>
29 #include <linux/seccomp_types.h>
30 #include <linux/nodemask_types.h>
31 #include <linux/refcount_types.h>
32 #include <linux/resource.h>
33 #include <linux/latencytop.h>
34 #include <linux/sched/prio.h>
35 #include <linux/sched/types.h>
36 #include <linux/signal_types.h>
37 #include <linux/syscall_user_dispatch_types.h>
38 #include <linux/mm_types_task.h>
39 #include <linux/netdevice_xmit.h>
40 #include <linux/task_io_accounting.h>
41 #include <linux/posix-timers_types.h>
42 #include <linux/restart_block.h>
43 #include <uapi/linux/rseq.h>
44 #include <linux/seqlock_types.h>
45 #include <linux/kcsan.h>
46 #include <linux/rv.h>
47 #include <linux/livepatch_sched.h>
48 #include <linux/uidgid_types.h>
49 #include <asm/kmap_size.h>
50 
51 /* task_struct member predeclarations (sorted alphabetically): */
52 struct audit_context;
53 struct bio_list;
54 struct blk_plug;
55 struct bpf_local_storage;
56 struct bpf_run_ctx;
57 struct bpf_net_context;
58 struct capture_control;
59 struct cfs_rq;
60 struct fs_struct;
61 struct futex_pi_state;
62 struct io_context;
63 struct io_uring_task;
64 struct mempolicy;
65 struct nameidata;
66 struct nsproxy;
67 struct perf_event_context;
68 struct pid_namespace;
69 struct pipe_inode_info;
70 struct rcu_node;
71 struct reclaim_state;
72 struct robust_list_head;
73 struct root_domain;
74 struct rq;
75 struct sched_attr;
76 struct sched_dl_entity;
77 struct seq_file;
78 struct sighand_struct;
79 struct signal_struct;
80 struct task_delay_info;
81 struct task_group;
82 struct task_struct;
83 struct user_event_mm;
84 
85 #include <linux/sched/ext.h>
86 
87 /*
88  * Task state bitmask. NOTE! These bits are also
89  * encoded in fs/proc/array.c: get_task_state().
90  *
91  * We have two separate sets of flags: task->__state
92  * is about runnability, while task->exit_state are
93  * about the task exiting. Confusing, but this way
94  * modifying one set can't modify the other one by
95  * mistake.
96  */
97 
98 /* Used in tsk->__state: */
99 #define TASK_RUNNING			0x00000000
100 #define TASK_INTERRUPTIBLE		0x00000001
101 #define TASK_UNINTERRUPTIBLE		0x00000002
102 #define __TASK_STOPPED			0x00000004
103 #define __TASK_TRACED			0x00000008
104 /* Used in tsk->exit_state: */
105 #define EXIT_DEAD			0x00000010
106 #define EXIT_ZOMBIE			0x00000020
107 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
108 /* Used in tsk->__state again: */
109 #define TASK_PARKED			0x00000040
110 #define TASK_DEAD			0x00000080
111 #define TASK_WAKEKILL			0x00000100
112 #define TASK_WAKING			0x00000200
113 #define TASK_NOLOAD			0x00000400
114 #define TASK_NEW			0x00000800
115 #define TASK_RTLOCK_WAIT		0x00001000
116 #define TASK_FREEZABLE			0x00002000
117 #define __TASK_FREEZABLE_UNSAFE	       (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
118 #define TASK_FROZEN			0x00008000
119 #define TASK_STATE_MAX			0x00010000
120 
121 #define TASK_ANY			(TASK_STATE_MAX-1)
122 
123 /*
124  * DO NOT ADD ANY NEW USERS !
125  */
126 #define TASK_FREEZABLE_UNSAFE		(TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
127 
128 /* Convenience macros for the sake of set_current_state: */
129 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
130 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
131 #define TASK_TRACED			__TASK_TRACED
132 
133 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
134 
135 /* Convenience macros for the sake of wake_up(): */
136 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
137 
138 /* get_task_state(): */
139 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
140 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
141 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
142 					 TASK_PARKED)
143 
144 #define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
145 
146 #define task_is_traced(task)		((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
147 #define task_is_stopped(task)		((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
148 #define task_is_stopped_or_traced(task)	((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
149 
150 /*
151  * Special states are those that do not use the normal wait-loop pattern. See
152  * the comment with set_special_state().
153  */
154 #define is_special_task_state(state)					\
155 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED |	\
156 		    TASK_DEAD | TASK_FROZEN))
157 
158 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
159 # define debug_normal_state_change(state_value)				\
160 	do {								\
161 		WARN_ON_ONCE(is_special_task_state(state_value));	\
162 		current->task_state_change = _THIS_IP_;			\
163 	} while (0)
164 
165 # define debug_special_state_change(state_value)			\
166 	do {								\
167 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
168 		current->task_state_change = _THIS_IP_;			\
169 	} while (0)
170 
171 # define debug_rtlock_wait_set_state()					\
172 	do {								 \
173 		current->saved_state_change = current->task_state_change;\
174 		current->task_state_change = _THIS_IP_;			 \
175 	} while (0)
176 
177 # define debug_rtlock_wait_restore_state()				\
178 	do {								 \
179 		current->task_state_change = current->saved_state_change;\
180 	} while (0)
181 
182 #else
183 # define debug_normal_state_change(cond)	do { } while (0)
184 # define debug_special_state_change(cond)	do { } while (0)
185 # define debug_rtlock_wait_set_state()		do { } while (0)
186 # define debug_rtlock_wait_restore_state()	do { } while (0)
187 #endif
188 
189 /*
190  * set_current_state() includes a barrier so that the write of current->__state
191  * is correctly serialised wrt the caller's subsequent test of whether to
192  * actually sleep:
193  *
194  *   for (;;) {
195  *	set_current_state(TASK_UNINTERRUPTIBLE);
196  *	if (CONDITION)
197  *	   break;
198  *
199  *	schedule();
200  *   }
201  *   __set_current_state(TASK_RUNNING);
202  *
203  * If the caller does not need such serialisation (because, for instance, the
204  * CONDITION test and condition change and wakeup are under the same lock) then
205  * use __set_current_state().
206  *
207  * The above is typically ordered against the wakeup, which does:
208  *
209  *   CONDITION = 1;
210  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
211  *
212  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
213  * accessing p->__state.
214  *
215  * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
216  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
217  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
218  *
219  * However, with slightly different timing the wakeup TASK_RUNNING store can
220  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
221  * a problem either because that will result in one extra go around the loop
222  * and our @cond test will save the day.
223  *
224  * Also see the comments of try_to_wake_up().
225  */
226 #define __set_current_state(state_value)				\
227 	do {								\
228 		debug_normal_state_change((state_value));		\
229 		WRITE_ONCE(current->__state, (state_value));		\
230 	} while (0)
231 
232 #define set_current_state(state_value)					\
233 	do {								\
234 		debug_normal_state_change((state_value));		\
235 		smp_store_mb(current->__state, (state_value));		\
236 	} while (0)
237 
238 /*
239  * set_special_state() should be used for those states when the blocking task
240  * can not use the regular condition based wait-loop. In that case we must
241  * serialize against wakeups such that any possible in-flight TASK_RUNNING
242  * stores will not collide with our state change.
243  */
244 #define set_special_state(state_value)					\
245 	do {								\
246 		unsigned long flags; /* may shadow */			\
247 									\
248 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
249 		debug_special_state_change((state_value));		\
250 		WRITE_ONCE(current->__state, (state_value));		\
251 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
252 	} while (0)
253 
254 /*
255  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
256  *
257  * RT's spin/rwlock substitutions are state preserving. The state of the
258  * task when blocking on the lock is saved in task_struct::saved_state and
259  * restored after the lock has been acquired.  These operations are
260  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
261  * lock related wakeups while the task is blocked on the lock are
262  * redirected to operate on task_struct::saved_state to ensure that these
263  * are not dropped. On restore task_struct::saved_state is set to
264  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
265  *
266  * The lock operation looks like this:
267  *
268  *	current_save_and_set_rtlock_wait_state();
269  *	for (;;) {
270  *		if (try_lock())
271  *			break;
272  *		raw_spin_unlock_irq(&lock->wait_lock);
273  *		schedule_rtlock();
274  *		raw_spin_lock_irq(&lock->wait_lock);
275  *		set_current_state(TASK_RTLOCK_WAIT);
276  *	}
277  *	current_restore_rtlock_saved_state();
278  */
279 #define current_save_and_set_rtlock_wait_state()			\
280 	do {								\
281 		lockdep_assert_irqs_disabled();				\
282 		raw_spin_lock(&current->pi_lock);			\
283 		current->saved_state = current->__state;		\
284 		debug_rtlock_wait_set_state();				\
285 		WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);		\
286 		raw_spin_unlock(&current->pi_lock);			\
287 	} while (0);
288 
289 #define current_restore_rtlock_saved_state()				\
290 	do {								\
291 		lockdep_assert_irqs_disabled();				\
292 		raw_spin_lock(&current->pi_lock);			\
293 		debug_rtlock_wait_restore_state();			\
294 		WRITE_ONCE(current->__state, current->saved_state);	\
295 		current->saved_state = TASK_RUNNING;			\
296 		raw_spin_unlock(&current->pi_lock);			\
297 	} while (0);
298 
299 #define get_current_state()	READ_ONCE(current->__state)
300 
301 /*
302  * Define the task command name length as enum, then it can be visible to
303  * BPF programs.
304  */
305 enum {
306 	TASK_COMM_LEN = 16,
307 };
308 
309 extern void sched_tick(void);
310 
311 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
312 
313 extern long schedule_timeout(long timeout);
314 extern long schedule_timeout_interruptible(long timeout);
315 extern long schedule_timeout_killable(long timeout);
316 extern long schedule_timeout_uninterruptible(long timeout);
317 extern long schedule_timeout_idle(long timeout);
318 asmlinkage void schedule(void);
319 extern void schedule_preempt_disabled(void);
320 asmlinkage void preempt_schedule_irq(void);
321 #ifdef CONFIG_PREEMPT_RT
322  extern void schedule_rtlock(void);
323 #endif
324 
325 extern int __must_check io_schedule_prepare(void);
326 extern void io_schedule_finish(int token);
327 extern long io_schedule_timeout(long timeout);
328 extern void io_schedule(void);
329 
330 /**
331  * struct prev_cputime - snapshot of system and user cputime
332  * @utime: time spent in user mode
333  * @stime: time spent in system mode
334  * @lock: protects the above two fields
335  *
336  * Stores previous user/system time values such that we can guarantee
337  * monotonicity.
338  */
339 struct prev_cputime {
340 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
341 	u64				utime;
342 	u64				stime;
343 	raw_spinlock_t			lock;
344 #endif
345 };
346 
347 enum vtime_state {
348 	/* Task is sleeping or running in a CPU with VTIME inactive: */
349 	VTIME_INACTIVE = 0,
350 	/* Task is idle */
351 	VTIME_IDLE,
352 	/* Task runs in kernelspace in a CPU with VTIME active: */
353 	VTIME_SYS,
354 	/* Task runs in userspace in a CPU with VTIME active: */
355 	VTIME_USER,
356 	/* Task runs as guests in a CPU with VTIME active: */
357 	VTIME_GUEST,
358 };
359 
360 struct vtime {
361 	seqcount_t		seqcount;
362 	unsigned long long	starttime;
363 	enum vtime_state	state;
364 	unsigned int		cpu;
365 	u64			utime;
366 	u64			stime;
367 	u64			gtime;
368 };
369 
370 /*
371  * Utilization clamp constraints.
372  * @UCLAMP_MIN:	Minimum utilization
373  * @UCLAMP_MAX:	Maximum utilization
374  * @UCLAMP_CNT:	Utilization clamp constraints count
375  */
376 enum uclamp_id {
377 	UCLAMP_MIN = 0,
378 	UCLAMP_MAX,
379 	UCLAMP_CNT
380 };
381 
382 #ifdef CONFIG_SMP
383 extern struct root_domain def_root_domain;
384 extern struct mutex sched_domains_mutex;
385 extern void sched_domains_mutex_lock(void);
386 extern void sched_domains_mutex_unlock(void);
387 #else
388 static inline void sched_domains_mutex_lock(void) { }
389 static inline void sched_domains_mutex_unlock(void) { }
390 #endif
391 
392 struct sched_param {
393 	int sched_priority;
394 };
395 
396 struct sched_info {
397 #ifdef CONFIG_SCHED_INFO
398 	/* Cumulative counters: */
399 
400 	/* # of times we have run on this CPU: */
401 	unsigned long			pcount;
402 
403 	/* Time spent waiting on a runqueue: */
404 	unsigned long long		run_delay;
405 
406 	/* Max time spent waiting on a runqueue: */
407 	unsigned long long		max_run_delay;
408 
409 	/* Min time spent waiting on a runqueue: */
410 	unsigned long long		min_run_delay;
411 
412 	/* Timestamps: */
413 
414 	/* When did we last run on a CPU? */
415 	unsigned long long		last_arrival;
416 
417 	/* When were we last queued to run? */
418 	unsigned long long		last_queued;
419 
420 #endif /* CONFIG_SCHED_INFO */
421 };
422 
423 /*
424  * Integer metrics need fixed point arithmetic, e.g., sched/fair
425  * has a few: load, load_avg, util_avg, freq, and capacity.
426  *
427  * We define a basic fixed point arithmetic range, and then formalize
428  * all these metrics based on that basic range.
429  */
430 # define SCHED_FIXEDPOINT_SHIFT		10
431 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
432 
433 /* Increase resolution of cpu_capacity calculations */
434 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
435 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
436 
437 struct load_weight {
438 	unsigned long			weight;
439 	u32				inv_weight;
440 };
441 
442 /*
443  * The load/runnable/util_avg accumulates an infinite geometric series
444  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
445  *
446  * [load_avg definition]
447  *
448  *   load_avg = runnable% * scale_load_down(load)
449  *
450  * [runnable_avg definition]
451  *
452  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
453  *
454  * [util_avg definition]
455  *
456  *   util_avg = running% * SCHED_CAPACITY_SCALE
457  *
458  * where runnable% is the time ratio that a sched_entity is runnable and
459  * running% the time ratio that a sched_entity is running.
460  *
461  * For cfs_rq, they are the aggregated values of all runnable and blocked
462  * sched_entities.
463  *
464  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
465  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
466  * for computing those signals (see update_rq_clock_pelt())
467  *
468  * N.B., the above ratios (runnable% and running%) themselves are in the
469  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
470  * to as large a range as necessary. This is for example reflected by
471  * util_avg's SCHED_CAPACITY_SCALE.
472  *
473  * [Overflow issue]
474  *
475  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
476  * with the highest load (=88761), always runnable on a single cfs_rq,
477  * and should not overflow as the number already hits PID_MAX_LIMIT.
478  *
479  * For all other cases (including 32-bit kernels), struct load_weight's
480  * weight will overflow first before we do, because:
481  *
482  *    Max(load_avg) <= Max(load.weight)
483  *
484  * Then it is the load_weight's responsibility to consider overflow
485  * issues.
486  */
487 struct sched_avg {
488 	u64				last_update_time;
489 	u64				load_sum;
490 	u64				runnable_sum;
491 	u32				util_sum;
492 	u32				period_contrib;
493 	unsigned long			load_avg;
494 	unsigned long			runnable_avg;
495 	unsigned long			util_avg;
496 	unsigned int			util_est;
497 } ____cacheline_aligned;
498 
499 /*
500  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
501  * updates. When a task is dequeued, its util_est should not be updated if its
502  * util_avg has not been updated in the meantime.
503  * This information is mapped into the MSB bit of util_est at dequeue time.
504  * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
505  * it is safe to use MSB.
506  */
507 #define UTIL_EST_WEIGHT_SHIFT		2
508 #define UTIL_AVG_UNCHANGED		0x80000000
509 
510 struct sched_statistics {
511 #ifdef CONFIG_SCHEDSTATS
512 	u64				wait_start;
513 	u64				wait_max;
514 	u64				wait_count;
515 	u64				wait_sum;
516 	u64				iowait_count;
517 	u64				iowait_sum;
518 
519 	u64				sleep_start;
520 	u64				sleep_max;
521 	s64				sum_sleep_runtime;
522 
523 	u64				block_start;
524 	u64				block_max;
525 	s64				sum_block_runtime;
526 
527 	s64				exec_max;
528 	u64				slice_max;
529 
530 	u64				nr_migrations_cold;
531 	u64				nr_failed_migrations_affine;
532 	u64				nr_failed_migrations_running;
533 	u64				nr_failed_migrations_hot;
534 	u64				nr_forced_migrations;
535 
536 	u64				nr_wakeups;
537 	u64				nr_wakeups_sync;
538 	u64				nr_wakeups_migrate;
539 	u64				nr_wakeups_local;
540 	u64				nr_wakeups_remote;
541 	u64				nr_wakeups_affine;
542 	u64				nr_wakeups_affine_attempts;
543 	u64				nr_wakeups_passive;
544 	u64				nr_wakeups_idle;
545 
546 #ifdef CONFIG_SCHED_CORE
547 	u64				core_forceidle_sum;
548 #endif
549 #endif /* CONFIG_SCHEDSTATS */
550 } ____cacheline_aligned;
551 
552 struct sched_entity {
553 	/* For load-balancing: */
554 	struct load_weight		load;
555 	struct rb_node			run_node;
556 	u64				deadline;
557 	u64				min_vruntime;
558 	u64				min_slice;
559 
560 	struct list_head		group_node;
561 	unsigned char			on_rq;
562 	unsigned char			sched_delayed;
563 	unsigned char			rel_deadline;
564 	unsigned char			custom_slice;
565 					/* hole */
566 
567 	u64				exec_start;
568 	u64				sum_exec_runtime;
569 	u64				prev_sum_exec_runtime;
570 	u64				vruntime;
571 	s64				vlag;
572 	u64				slice;
573 
574 	u64				nr_migrations;
575 
576 #ifdef CONFIG_FAIR_GROUP_SCHED
577 	int				depth;
578 	struct sched_entity		*parent;
579 	/* rq on which this entity is (to be) queued: */
580 	struct cfs_rq			*cfs_rq;
581 	/* rq "owned" by this entity/group: */
582 	struct cfs_rq			*my_q;
583 	/* cached value of my_q->h_nr_running */
584 	unsigned long			runnable_weight;
585 #endif
586 
587 #ifdef CONFIG_SMP
588 	/*
589 	 * Per entity load average tracking.
590 	 *
591 	 * Put into separate cache line so it does not
592 	 * collide with read-mostly values above.
593 	 */
594 	struct sched_avg		avg;
595 #endif
596 };
597 
598 struct sched_rt_entity {
599 	struct list_head		run_list;
600 	unsigned long			timeout;
601 	unsigned long			watchdog_stamp;
602 	unsigned int			time_slice;
603 	unsigned short			on_rq;
604 	unsigned short			on_list;
605 
606 	struct sched_rt_entity		*back;
607 #ifdef CONFIG_RT_GROUP_SCHED
608 	struct sched_rt_entity		*parent;
609 	/* rq on which this entity is (to be) queued: */
610 	struct rt_rq			*rt_rq;
611 	/* rq "owned" by this entity/group: */
612 	struct rt_rq			*my_q;
613 #endif
614 } __randomize_layout;
615 
616 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
617 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
618 
619 struct sched_dl_entity {
620 	struct rb_node			rb_node;
621 
622 	/*
623 	 * Original scheduling parameters. Copied here from sched_attr
624 	 * during sched_setattr(), they will remain the same until
625 	 * the next sched_setattr().
626 	 */
627 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
628 	u64				dl_deadline;	/* Relative deadline of each instance	*/
629 	u64				dl_period;	/* Separation of two instances (period) */
630 	u64				dl_bw;		/* dl_runtime / dl_period		*/
631 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
632 
633 	/*
634 	 * Actual scheduling parameters. Initialized with the values above,
635 	 * they are continuously updated during task execution. Note that
636 	 * the remaining runtime could be < 0 in case we are in overrun.
637 	 */
638 	s64				runtime;	/* Remaining runtime for this instance	*/
639 	u64				deadline;	/* Absolute deadline for this instance	*/
640 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
641 
642 	/*
643 	 * Some bool flags:
644 	 *
645 	 * @dl_throttled tells if we exhausted the runtime. If so, the
646 	 * task has to wait for a replenishment to be performed at the
647 	 * next firing of dl_timer.
648 	 *
649 	 * @dl_yielded tells if task gave up the CPU before consuming
650 	 * all its available runtime during the last job.
651 	 *
652 	 * @dl_non_contending tells if the task is inactive while still
653 	 * contributing to the active utilization. In other words, it
654 	 * indicates if the inactive timer has been armed and its handler
655 	 * has not been executed yet. This flag is useful to avoid race
656 	 * conditions between the inactive timer handler and the wakeup
657 	 * code.
658 	 *
659 	 * @dl_overrun tells if the task asked to be informed about runtime
660 	 * overruns.
661 	 *
662 	 * @dl_server tells if this is a server entity.
663 	 *
664 	 * @dl_defer tells if this is a deferred or regular server. For
665 	 * now only defer server exists.
666 	 *
667 	 * @dl_defer_armed tells if the deferrable server is waiting
668 	 * for the replenishment timer to activate it.
669 	 *
670 	 * @dl_server_active tells if the dlserver is active(started).
671 	 * dlserver is started on first cfs enqueue on an idle runqueue
672 	 * and is stopped when a dequeue results in 0 cfs tasks on the
673 	 * runqueue. In other words, dlserver is active only when cpu's
674 	 * runqueue has atleast one cfs task.
675 	 *
676 	 * @dl_defer_running tells if the deferrable server is actually
677 	 * running, skipping the defer phase.
678 	 */
679 	unsigned int			dl_throttled      : 1;
680 	unsigned int			dl_yielded        : 1;
681 	unsigned int			dl_non_contending : 1;
682 	unsigned int			dl_overrun	  : 1;
683 	unsigned int			dl_server         : 1;
684 	unsigned int			dl_server_active  : 1;
685 	unsigned int			dl_defer	  : 1;
686 	unsigned int			dl_defer_armed	  : 1;
687 	unsigned int			dl_defer_running  : 1;
688 
689 	/*
690 	 * Bandwidth enforcement timer. Each -deadline task has its
691 	 * own bandwidth to be enforced, thus we need one timer per task.
692 	 */
693 	struct hrtimer			dl_timer;
694 
695 	/*
696 	 * Inactive timer, responsible for decreasing the active utilization
697 	 * at the "0-lag time". When a -deadline task blocks, it contributes
698 	 * to GRUB's active utilization until the "0-lag time", hence a
699 	 * timer is needed to decrease the active utilization at the correct
700 	 * time.
701 	 */
702 	struct hrtimer			inactive_timer;
703 
704 	/*
705 	 * Bits for DL-server functionality. Also see the comment near
706 	 * dl_server_update().
707 	 *
708 	 * @rq the runqueue this server is for
709 	 *
710 	 * @server_has_tasks() returns true if @server_pick return a
711 	 * runnable task.
712 	 */
713 	struct rq			*rq;
714 	dl_server_has_tasks_f		server_has_tasks;
715 	dl_server_pick_f		server_pick_task;
716 
717 #ifdef CONFIG_RT_MUTEXES
718 	/*
719 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
720 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
721 	 * to (the original one/itself).
722 	 */
723 	struct sched_dl_entity *pi_se;
724 #endif
725 };
726 
727 #ifdef CONFIG_UCLAMP_TASK
728 /* Number of utilization clamp buckets (shorter alias) */
729 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
730 
731 /*
732  * Utilization clamp for a scheduling entity
733  * @value:		clamp value "assigned" to a se
734  * @bucket_id:		bucket index corresponding to the "assigned" value
735  * @active:		the se is currently refcounted in a rq's bucket
736  * @user_defined:	the requested clamp value comes from user-space
737  *
738  * The bucket_id is the index of the clamp bucket matching the clamp value
739  * which is pre-computed and stored to avoid expensive integer divisions from
740  * the fast path.
741  *
742  * The active bit is set whenever a task has got an "effective" value assigned,
743  * which can be different from the clamp value "requested" from user-space.
744  * This allows to know a task is refcounted in the rq's bucket corresponding
745  * to the "effective" bucket_id.
746  *
747  * The user_defined bit is set whenever a task has got a task-specific clamp
748  * value requested from userspace, i.e. the system defaults apply to this task
749  * just as a restriction. This allows to relax default clamps when a less
750  * restrictive task-specific value has been requested, thus allowing to
751  * implement a "nice" semantic. For example, a task running with a 20%
752  * default boost can still drop its own boosting to 0%.
753  */
754 struct uclamp_se {
755 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
756 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
757 	unsigned int active		: 1;
758 	unsigned int user_defined	: 1;
759 };
760 #endif /* CONFIG_UCLAMP_TASK */
761 
762 union rcu_special {
763 	struct {
764 		u8			blocked;
765 		u8			need_qs;
766 		u8			exp_hint; /* Hint for performance. */
767 		u8			need_mb; /* Readers need smp_mb(). */
768 	} b; /* Bits. */
769 	u32 s; /* Set of bits. */
770 };
771 
772 enum perf_event_task_context {
773 	perf_invalid_context = -1,
774 	perf_hw_context = 0,
775 	perf_sw_context,
776 	perf_nr_task_contexts,
777 };
778 
779 /*
780  * Number of contexts where an event can trigger:
781  *      task, softirq, hardirq, nmi.
782  */
783 #define PERF_NR_CONTEXTS	4
784 
785 struct wake_q_node {
786 	struct wake_q_node *next;
787 };
788 
789 struct kmap_ctrl {
790 #ifdef CONFIG_KMAP_LOCAL
791 	int				idx;
792 	pte_t				pteval[KM_MAX_IDX];
793 #endif
794 };
795 
796 struct task_struct {
797 #ifdef CONFIG_THREAD_INFO_IN_TASK
798 	/*
799 	 * For reasons of header soup (see current_thread_info()), this
800 	 * must be the first element of task_struct.
801 	 */
802 	struct thread_info		thread_info;
803 #endif
804 	unsigned int			__state;
805 
806 	/* saved state for "spinlock sleepers" */
807 	unsigned int			saved_state;
808 
809 	/*
810 	 * This begins the randomizable portion of task_struct. Only
811 	 * scheduling-critical items should be added above here.
812 	 */
813 	randomized_struct_fields_start
814 
815 	void				*stack;
816 	refcount_t			usage;
817 	/* Per task flags (PF_*), defined further below: */
818 	unsigned int			flags;
819 	unsigned int			ptrace;
820 
821 #ifdef CONFIG_MEM_ALLOC_PROFILING
822 	struct alloc_tag		*alloc_tag;
823 #endif
824 
825 #ifdef CONFIG_SMP
826 	int				on_cpu;
827 	struct __call_single_node	wake_entry;
828 	unsigned int			wakee_flips;
829 	unsigned long			wakee_flip_decay_ts;
830 	struct task_struct		*last_wakee;
831 
832 	/*
833 	 * recent_used_cpu is initially set as the last CPU used by a task
834 	 * that wakes affine another task. Waker/wakee relationships can
835 	 * push tasks around a CPU where each wakeup moves to the next one.
836 	 * Tracking a recently used CPU allows a quick search for a recently
837 	 * used CPU that may be idle.
838 	 */
839 	int				recent_used_cpu;
840 	int				wake_cpu;
841 #endif
842 	int				on_rq;
843 
844 	int				prio;
845 	int				static_prio;
846 	int				normal_prio;
847 	unsigned int			rt_priority;
848 
849 	struct sched_entity		se;
850 	struct sched_rt_entity		rt;
851 	struct sched_dl_entity		dl;
852 	struct sched_dl_entity		*dl_server;
853 #ifdef CONFIG_SCHED_CLASS_EXT
854 	struct sched_ext_entity		scx;
855 #endif
856 	const struct sched_class	*sched_class;
857 
858 #ifdef CONFIG_SCHED_CORE
859 	struct rb_node			core_node;
860 	unsigned long			core_cookie;
861 	unsigned int			core_occupation;
862 #endif
863 
864 #ifdef CONFIG_CGROUP_SCHED
865 	struct task_group		*sched_task_group;
866 #endif
867 
868 
869 #ifdef CONFIG_UCLAMP_TASK
870 	/*
871 	 * Clamp values requested for a scheduling entity.
872 	 * Must be updated with task_rq_lock() held.
873 	 */
874 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
875 	/*
876 	 * Effective clamp values used for a scheduling entity.
877 	 * Must be updated with task_rq_lock() held.
878 	 */
879 	struct uclamp_se		uclamp[UCLAMP_CNT];
880 #endif
881 
882 	struct sched_statistics         stats;
883 
884 #ifdef CONFIG_PREEMPT_NOTIFIERS
885 	/* List of struct preempt_notifier: */
886 	struct hlist_head		preempt_notifiers;
887 #endif
888 
889 #ifdef CONFIG_BLK_DEV_IO_TRACE
890 	unsigned int			btrace_seq;
891 #endif
892 
893 	unsigned int			policy;
894 	unsigned long			max_allowed_capacity;
895 	int				nr_cpus_allowed;
896 	const cpumask_t			*cpus_ptr;
897 	cpumask_t			*user_cpus_ptr;
898 	cpumask_t			cpus_mask;
899 	void				*migration_pending;
900 #ifdef CONFIG_SMP
901 	unsigned short			migration_disabled;
902 #endif
903 	unsigned short			migration_flags;
904 
905 #ifdef CONFIG_PREEMPT_RCU
906 	int				rcu_read_lock_nesting;
907 	union rcu_special		rcu_read_unlock_special;
908 	struct list_head		rcu_node_entry;
909 	struct rcu_node			*rcu_blocked_node;
910 #endif /* #ifdef CONFIG_PREEMPT_RCU */
911 
912 #ifdef CONFIG_TASKS_RCU
913 	unsigned long			rcu_tasks_nvcsw;
914 	u8				rcu_tasks_holdout;
915 	u8				rcu_tasks_idx;
916 	int				rcu_tasks_idle_cpu;
917 	struct list_head		rcu_tasks_holdout_list;
918 	int				rcu_tasks_exit_cpu;
919 	struct list_head		rcu_tasks_exit_list;
920 #endif /* #ifdef CONFIG_TASKS_RCU */
921 
922 #ifdef CONFIG_TASKS_TRACE_RCU
923 	int				trc_reader_nesting;
924 	int				trc_ipi_to_cpu;
925 	union rcu_special		trc_reader_special;
926 	struct list_head		trc_holdout_list;
927 	struct list_head		trc_blkd_node;
928 	int				trc_blkd_cpu;
929 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
930 
931 	struct sched_info		sched_info;
932 
933 	struct list_head		tasks;
934 #ifdef CONFIG_SMP
935 	struct plist_node		pushable_tasks;
936 	struct rb_node			pushable_dl_tasks;
937 #endif
938 
939 	struct mm_struct		*mm;
940 	struct mm_struct		*active_mm;
941 	struct address_space		*faults_disabled_mapping;
942 
943 	int				exit_state;
944 	int				exit_code;
945 	int				exit_signal;
946 	/* The signal sent when the parent dies: */
947 	int				pdeath_signal;
948 	/* JOBCTL_*, siglock protected: */
949 	unsigned long			jobctl;
950 
951 	/* Used for emulating ABI behavior of previous Linux versions: */
952 	unsigned int			personality;
953 
954 	/* Scheduler bits, serialized by scheduler locks: */
955 	unsigned			sched_reset_on_fork:1;
956 	unsigned			sched_contributes_to_load:1;
957 	unsigned			sched_migrated:1;
958 	unsigned			sched_task_hot:1;
959 
960 	/* Force alignment to the next boundary: */
961 	unsigned			:0;
962 
963 	/* Unserialized, strictly 'current' */
964 
965 	/*
966 	 * This field must not be in the scheduler word above due to wakelist
967 	 * queueing no longer being serialized by p->on_cpu. However:
968 	 *
969 	 * p->XXX = X;			ttwu()
970 	 * schedule()			  if (p->on_rq && ..) // false
971 	 *   smp_mb__after_spinlock();	  if (smp_load_acquire(&p->on_cpu) && //true
972 	 *   deactivate_task()		      ttwu_queue_wakelist())
973 	 *     p->on_rq = 0;			p->sched_remote_wakeup = Y;
974 	 *
975 	 * guarantees all stores of 'current' are visible before
976 	 * ->sched_remote_wakeup gets used, so it can be in this word.
977 	 */
978 	unsigned			sched_remote_wakeup:1;
979 #ifdef CONFIG_RT_MUTEXES
980 	unsigned			sched_rt_mutex:1;
981 #endif
982 
983 	/* Bit to tell TOMOYO we're in execve(): */
984 	unsigned			in_execve:1;
985 	unsigned			in_iowait:1;
986 #ifndef TIF_RESTORE_SIGMASK
987 	unsigned			restore_sigmask:1;
988 #endif
989 #ifdef CONFIG_MEMCG_V1
990 	unsigned			in_user_fault:1;
991 #endif
992 #ifdef CONFIG_LRU_GEN
993 	/* whether the LRU algorithm may apply to this access */
994 	unsigned			in_lru_fault:1;
995 #endif
996 #ifdef CONFIG_COMPAT_BRK
997 	unsigned			brk_randomized:1;
998 #endif
999 #ifdef CONFIG_CGROUPS
1000 	/* disallow userland-initiated cgroup migration */
1001 	unsigned			no_cgroup_migration:1;
1002 	/* task is frozen/stopped (used by the cgroup freezer) */
1003 	unsigned			frozen:1;
1004 #endif
1005 #ifdef CONFIG_BLK_CGROUP
1006 	unsigned			use_memdelay:1;
1007 #endif
1008 #ifdef CONFIG_PSI
1009 	/* Stalled due to lack of memory */
1010 	unsigned			in_memstall:1;
1011 #endif
1012 #ifdef CONFIG_PAGE_OWNER
1013 	/* Used by page_owner=on to detect recursion in page tracking. */
1014 	unsigned			in_page_owner:1;
1015 #endif
1016 #ifdef CONFIG_EVENTFD
1017 	/* Recursion prevention for eventfd_signal() */
1018 	unsigned			in_eventfd:1;
1019 #endif
1020 #ifdef CONFIG_ARCH_HAS_CPU_PASID
1021 	unsigned			pasid_activated:1;
1022 #endif
1023 #ifdef CONFIG_X86_BUS_LOCK_DETECT
1024 	unsigned			reported_split_lock:1;
1025 #endif
1026 #ifdef CONFIG_TASK_DELAY_ACCT
1027 	/* delay due to memory thrashing */
1028 	unsigned                        in_thrashing:1;
1029 #endif
1030 #ifdef CONFIG_PREEMPT_RT
1031 	struct netdev_xmit		net_xmit;
1032 #endif
1033 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
1034 
1035 	struct restart_block		restart_block;
1036 
1037 	pid_t				pid;
1038 	pid_t				tgid;
1039 
1040 #ifdef CONFIG_STACKPROTECTOR
1041 	/* Canary value for the -fstack-protector GCC feature: */
1042 	unsigned long			stack_canary;
1043 #endif
1044 	/*
1045 	 * Pointers to the (original) parent process, youngest child, younger sibling,
1046 	 * older sibling, respectively.  (p->father can be replaced with
1047 	 * p->real_parent->pid)
1048 	 */
1049 
1050 	/* Real parent process: */
1051 	struct task_struct __rcu	*real_parent;
1052 
1053 	/* Recipient of SIGCHLD, wait4() reports: */
1054 	struct task_struct __rcu	*parent;
1055 
1056 	/*
1057 	 * Children/sibling form the list of natural children:
1058 	 */
1059 	struct list_head		children;
1060 	struct list_head		sibling;
1061 	struct task_struct		*group_leader;
1062 
1063 	/*
1064 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
1065 	 *
1066 	 * This includes both natural children and PTRACE_ATTACH targets.
1067 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1068 	 */
1069 	struct list_head		ptraced;
1070 	struct list_head		ptrace_entry;
1071 
1072 	/* PID/PID hash table linkage. */
1073 	struct pid			*thread_pid;
1074 	struct hlist_node		pid_links[PIDTYPE_MAX];
1075 	struct list_head		thread_node;
1076 
1077 	struct completion		*vfork_done;
1078 
1079 	/* CLONE_CHILD_SETTID: */
1080 	int __user			*set_child_tid;
1081 
1082 	/* CLONE_CHILD_CLEARTID: */
1083 	int __user			*clear_child_tid;
1084 
1085 	/* PF_KTHREAD | PF_IO_WORKER */
1086 	void				*worker_private;
1087 
1088 	u64				utime;
1089 	u64				stime;
1090 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1091 	u64				utimescaled;
1092 	u64				stimescaled;
1093 #endif
1094 	u64				gtime;
1095 	struct prev_cputime		prev_cputime;
1096 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1097 	struct vtime			vtime;
1098 #endif
1099 
1100 #ifdef CONFIG_NO_HZ_FULL
1101 	atomic_t			tick_dep_mask;
1102 #endif
1103 	/* Context switch counts: */
1104 	unsigned long			nvcsw;
1105 	unsigned long			nivcsw;
1106 
1107 	/* Monotonic time in nsecs: */
1108 	u64				start_time;
1109 
1110 	/* Boot based time in nsecs: */
1111 	u64				start_boottime;
1112 
1113 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1114 	unsigned long			min_flt;
1115 	unsigned long			maj_flt;
1116 
1117 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1118 	struct posix_cputimers		posix_cputimers;
1119 
1120 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1121 	struct posix_cputimers_work	posix_cputimers_work;
1122 #endif
1123 
1124 	/* Process credentials: */
1125 
1126 	/* Tracer's credentials at attach: */
1127 	const struct cred __rcu		*ptracer_cred;
1128 
1129 	/* Objective and real subjective task credentials (COW): */
1130 	const struct cred __rcu		*real_cred;
1131 
1132 	/* Effective (overridable) subjective task credentials (COW): */
1133 	const struct cred __rcu		*cred;
1134 
1135 #ifdef CONFIG_KEYS
1136 	/* Cached requested key. */
1137 	struct key			*cached_requested_key;
1138 #endif
1139 
1140 	/*
1141 	 * executable name, excluding path.
1142 	 *
1143 	 * - normally initialized begin_new_exec()
1144 	 * - set it with set_task_comm()
1145 	 *   - strscpy_pad() to ensure it is always NUL-terminated and
1146 	 *     zero-padded
1147 	 *   - task_lock() to ensure the operation is atomic and the name is
1148 	 *     fully updated.
1149 	 */
1150 	char				comm[TASK_COMM_LEN];
1151 
1152 	struct nameidata		*nameidata;
1153 
1154 #ifdef CONFIG_SYSVIPC
1155 	struct sysv_sem			sysvsem;
1156 	struct sysv_shm			sysvshm;
1157 #endif
1158 #ifdef CONFIG_DETECT_HUNG_TASK
1159 	unsigned long			last_switch_count;
1160 	unsigned long			last_switch_time;
1161 #endif
1162 	/* Filesystem information: */
1163 	struct fs_struct		*fs;
1164 
1165 	/* Open file information: */
1166 	struct files_struct		*files;
1167 
1168 #ifdef CONFIG_IO_URING
1169 	struct io_uring_task		*io_uring;
1170 #endif
1171 
1172 	/* Namespaces: */
1173 	struct nsproxy			*nsproxy;
1174 
1175 	/* Signal handlers: */
1176 	struct signal_struct		*signal;
1177 	struct sighand_struct __rcu		*sighand;
1178 	sigset_t			blocked;
1179 	sigset_t			real_blocked;
1180 	/* Restored if set_restore_sigmask() was used: */
1181 	sigset_t			saved_sigmask;
1182 	struct sigpending		pending;
1183 	unsigned long			sas_ss_sp;
1184 	size_t				sas_ss_size;
1185 	unsigned int			sas_ss_flags;
1186 
1187 	struct callback_head		*task_works;
1188 
1189 #ifdef CONFIG_AUDIT
1190 #ifdef CONFIG_AUDITSYSCALL
1191 	struct audit_context		*audit_context;
1192 #endif
1193 	kuid_t				loginuid;
1194 	unsigned int			sessionid;
1195 #endif
1196 	struct seccomp			seccomp;
1197 	struct syscall_user_dispatch	syscall_dispatch;
1198 
1199 	/* Thread group tracking: */
1200 	u64				parent_exec_id;
1201 	u64				self_exec_id;
1202 
1203 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1204 	spinlock_t			alloc_lock;
1205 
1206 	/* Protection of the PI data structures: */
1207 	raw_spinlock_t			pi_lock;
1208 
1209 	struct wake_q_node		wake_q;
1210 
1211 #ifdef CONFIG_RT_MUTEXES
1212 	/* PI waiters blocked on a rt_mutex held by this task: */
1213 	struct rb_root_cached		pi_waiters;
1214 	/* Updated under owner's pi_lock and rq lock */
1215 	struct task_struct		*pi_top_task;
1216 	/* Deadlock detection and priority inheritance handling: */
1217 	struct rt_mutex_waiter		*pi_blocked_on;
1218 #endif
1219 
1220 #ifdef CONFIG_DEBUG_MUTEXES
1221 	/* Mutex deadlock detection: */
1222 	struct mutex_waiter		*blocked_on;
1223 #endif
1224 
1225 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1226 	int				non_block_count;
1227 #endif
1228 
1229 #ifdef CONFIG_TRACE_IRQFLAGS
1230 	struct irqtrace_events		irqtrace;
1231 	unsigned int			hardirq_threaded;
1232 	u64				hardirq_chain_key;
1233 	int				softirqs_enabled;
1234 	int				softirq_context;
1235 	int				irq_config;
1236 #endif
1237 #ifdef CONFIG_PREEMPT_RT
1238 	int				softirq_disable_cnt;
1239 #endif
1240 
1241 #ifdef CONFIG_LOCKDEP
1242 # define MAX_LOCK_DEPTH			48UL
1243 	u64				curr_chain_key;
1244 	int				lockdep_depth;
1245 	unsigned int			lockdep_recursion;
1246 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1247 #endif
1248 
1249 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1250 	unsigned int			in_ubsan;
1251 #endif
1252 
1253 	/* Journalling filesystem info: */
1254 	void				*journal_info;
1255 
1256 	/* Stacked block device info: */
1257 	struct bio_list			*bio_list;
1258 
1259 	/* Stack plugging: */
1260 	struct blk_plug			*plug;
1261 
1262 	/* VM state: */
1263 	struct reclaim_state		*reclaim_state;
1264 
1265 	struct io_context		*io_context;
1266 
1267 #ifdef CONFIG_COMPACTION
1268 	struct capture_control		*capture_control;
1269 #endif
1270 	/* Ptrace state: */
1271 	unsigned long			ptrace_message;
1272 	kernel_siginfo_t		*last_siginfo;
1273 
1274 	struct task_io_accounting	ioac;
1275 #ifdef CONFIG_PSI
1276 	/* Pressure stall state */
1277 	unsigned int			psi_flags;
1278 #endif
1279 #ifdef CONFIG_TASK_XACCT
1280 	/* Accumulated RSS usage: */
1281 	u64				acct_rss_mem1;
1282 	/* Accumulated virtual memory usage: */
1283 	u64				acct_vm_mem1;
1284 	/* stime + utime since last update: */
1285 	u64				acct_timexpd;
1286 #endif
1287 #ifdef CONFIG_CPUSETS
1288 	/* Protected by ->alloc_lock: */
1289 	nodemask_t			mems_allowed;
1290 	/* Sequence number to catch updates: */
1291 	seqcount_spinlock_t		mems_allowed_seq;
1292 	int				cpuset_mem_spread_rotor;
1293 #endif
1294 #ifdef CONFIG_CGROUPS
1295 	/* Control Group info protected by css_set_lock: */
1296 	struct css_set __rcu		*cgroups;
1297 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1298 	struct list_head		cg_list;
1299 #endif
1300 #ifdef CONFIG_X86_CPU_RESCTRL
1301 	u32				closid;
1302 	u32				rmid;
1303 #endif
1304 #ifdef CONFIG_FUTEX
1305 	struct robust_list_head __user	*robust_list;
1306 #ifdef CONFIG_COMPAT
1307 	struct compat_robust_list_head __user *compat_robust_list;
1308 #endif
1309 	struct list_head		pi_state_list;
1310 	struct futex_pi_state		*pi_state_cache;
1311 	struct mutex			futex_exit_mutex;
1312 	unsigned int			futex_state;
1313 #endif
1314 #ifdef CONFIG_PERF_EVENTS
1315 	u8				perf_recursion[PERF_NR_CONTEXTS];
1316 	struct perf_event_context	*perf_event_ctxp;
1317 	struct mutex			perf_event_mutex;
1318 	struct list_head		perf_event_list;
1319 #endif
1320 #ifdef CONFIG_DEBUG_PREEMPT
1321 	unsigned long			preempt_disable_ip;
1322 #endif
1323 #ifdef CONFIG_NUMA
1324 	/* Protected by alloc_lock: */
1325 	struct mempolicy		*mempolicy;
1326 	short				il_prev;
1327 	u8				il_weight;
1328 	short				pref_node_fork;
1329 #endif
1330 #ifdef CONFIG_NUMA_BALANCING
1331 	int				numa_scan_seq;
1332 	unsigned int			numa_scan_period;
1333 	unsigned int			numa_scan_period_max;
1334 	int				numa_preferred_nid;
1335 	unsigned long			numa_migrate_retry;
1336 	/* Migration stamp: */
1337 	u64				node_stamp;
1338 	u64				last_task_numa_placement;
1339 	u64				last_sum_exec_runtime;
1340 	struct callback_head		numa_work;
1341 
1342 	/*
1343 	 * This pointer is only modified for current in syscall and
1344 	 * pagefault context (and for tasks being destroyed), so it can be read
1345 	 * from any of the following contexts:
1346 	 *  - RCU read-side critical section
1347 	 *  - current->numa_group from everywhere
1348 	 *  - task's runqueue locked, task not running
1349 	 */
1350 	struct numa_group __rcu		*numa_group;
1351 
1352 	/*
1353 	 * numa_faults is an array split into four regions:
1354 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1355 	 * in this precise order.
1356 	 *
1357 	 * faults_memory: Exponential decaying average of faults on a per-node
1358 	 * basis. Scheduling placement decisions are made based on these
1359 	 * counts. The values remain static for the duration of a PTE scan.
1360 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1361 	 * hinting fault was incurred.
1362 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1363 	 * during the current scan window. When the scan completes, the counts
1364 	 * in faults_memory and faults_cpu decay and these values are copied.
1365 	 */
1366 	unsigned long			*numa_faults;
1367 	unsigned long			total_numa_faults;
1368 
1369 	/*
1370 	 * numa_faults_locality tracks if faults recorded during the last
1371 	 * scan window were remote/local or failed to migrate. The task scan
1372 	 * period is adapted based on the locality of the faults with different
1373 	 * weights depending on whether they were shared or private faults
1374 	 */
1375 	unsigned long			numa_faults_locality[3];
1376 
1377 	unsigned long			numa_pages_migrated;
1378 #endif /* CONFIG_NUMA_BALANCING */
1379 
1380 #ifdef CONFIG_RSEQ
1381 	struct rseq __user *rseq;
1382 	u32 rseq_len;
1383 	u32 rseq_sig;
1384 	/*
1385 	 * RmW on rseq_event_mask must be performed atomically
1386 	 * with respect to preemption.
1387 	 */
1388 	unsigned long rseq_event_mask;
1389 # ifdef CONFIG_DEBUG_RSEQ
1390 	/*
1391 	 * This is a place holder to save a copy of the rseq fields for
1392 	 * validation of read-only fields. The struct rseq has a
1393 	 * variable-length array at the end, so it cannot be used
1394 	 * directly. Reserve a size large enough for the known fields.
1395 	 */
1396 	char				rseq_fields[sizeof(struct rseq)];
1397 # endif
1398 #endif
1399 
1400 #ifdef CONFIG_SCHED_MM_CID
1401 	int				mm_cid;		/* Current cid in mm */
1402 	int				last_mm_cid;	/* Most recent cid in mm */
1403 	int				migrate_from_cpu;
1404 	int				mm_cid_active;	/* Whether cid bitmap is active */
1405 	struct callback_head		cid_work;
1406 #endif
1407 
1408 	struct tlbflush_unmap_batch	tlb_ubc;
1409 
1410 	/* Cache last used pipe for splice(): */
1411 	struct pipe_inode_info		*splice_pipe;
1412 
1413 	struct page_frag		task_frag;
1414 
1415 #ifdef CONFIG_TASK_DELAY_ACCT
1416 	struct task_delay_info		*delays;
1417 #endif
1418 
1419 #ifdef CONFIG_FAULT_INJECTION
1420 	int				make_it_fail;
1421 	unsigned int			fail_nth;
1422 #endif
1423 	/*
1424 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1425 	 * balance_dirty_pages() for a dirty throttling pause:
1426 	 */
1427 	int				nr_dirtied;
1428 	int				nr_dirtied_pause;
1429 	/* Start of a write-and-pause period: */
1430 	unsigned long			dirty_paused_when;
1431 
1432 #ifdef CONFIG_LATENCYTOP
1433 	int				latency_record_count;
1434 	struct latency_record		latency_record[LT_SAVECOUNT];
1435 #endif
1436 	/*
1437 	 * Time slack values; these are used to round up poll() and
1438 	 * select() etc timeout values. These are in nanoseconds.
1439 	 */
1440 	u64				timer_slack_ns;
1441 	u64				default_timer_slack_ns;
1442 
1443 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1444 	unsigned int			kasan_depth;
1445 #endif
1446 
1447 #ifdef CONFIG_KCSAN
1448 	struct kcsan_ctx		kcsan_ctx;
1449 #ifdef CONFIG_TRACE_IRQFLAGS
1450 	struct irqtrace_events		kcsan_save_irqtrace;
1451 #endif
1452 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1453 	int				kcsan_stack_depth;
1454 #endif
1455 #endif
1456 
1457 #ifdef CONFIG_KMSAN
1458 	struct kmsan_ctx		kmsan_ctx;
1459 #endif
1460 
1461 #if IS_ENABLED(CONFIG_KUNIT)
1462 	struct kunit			*kunit_test;
1463 #endif
1464 
1465 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1466 	/* Index of current stored address in ret_stack: */
1467 	int				curr_ret_stack;
1468 	int				curr_ret_depth;
1469 
1470 	/* Stack of return addresses for return function tracing: */
1471 	unsigned long			*ret_stack;
1472 
1473 	/* Timestamp for last schedule: */
1474 	unsigned long long		ftrace_timestamp;
1475 	unsigned long long		ftrace_sleeptime;
1476 
1477 	/*
1478 	 * Number of functions that haven't been traced
1479 	 * because of depth overrun:
1480 	 */
1481 	atomic_t			trace_overrun;
1482 
1483 	/* Pause tracing: */
1484 	atomic_t			tracing_graph_pause;
1485 #endif
1486 
1487 #ifdef CONFIG_TRACING
1488 	/* Bitmask and counter of trace recursion: */
1489 	unsigned long			trace_recursion;
1490 #endif /* CONFIG_TRACING */
1491 
1492 #ifdef CONFIG_KCOV
1493 	/* See kernel/kcov.c for more details. */
1494 
1495 	/* Coverage collection mode enabled for this task (0 if disabled): */
1496 	unsigned int			kcov_mode;
1497 
1498 	/* Size of the kcov_area: */
1499 	unsigned int			kcov_size;
1500 
1501 	/* Buffer for coverage collection: */
1502 	void				*kcov_area;
1503 
1504 	/* KCOV descriptor wired with this task or NULL: */
1505 	struct kcov			*kcov;
1506 
1507 	/* KCOV common handle for remote coverage collection: */
1508 	u64				kcov_handle;
1509 
1510 	/* KCOV sequence number: */
1511 	int				kcov_sequence;
1512 
1513 	/* Collect coverage from softirq context: */
1514 	unsigned int			kcov_softirq;
1515 #endif
1516 
1517 #ifdef CONFIG_MEMCG_V1
1518 	struct mem_cgroup		*memcg_in_oom;
1519 #endif
1520 
1521 #ifdef CONFIG_MEMCG
1522 	/* Number of pages to reclaim on returning to userland: */
1523 	unsigned int			memcg_nr_pages_over_high;
1524 
1525 	/* Used by memcontrol for targeted memcg charge: */
1526 	struct mem_cgroup		*active_memcg;
1527 
1528 	/* Cache for current->cgroups->memcg->objcg lookups: */
1529 	struct obj_cgroup		*objcg;
1530 #endif
1531 
1532 #ifdef CONFIG_BLK_CGROUP
1533 	struct gendisk			*throttle_disk;
1534 #endif
1535 
1536 #ifdef CONFIG_UPROBES
1537 	struct uprobe_task		*utask;
1538 #endif
1539 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1540 	unsigned int			sequential_io;
1541 	unsigned int			sequential_io_avg;
1542 #endif
1543 	struct kmap_ctrl		kmap_ctrl;
1544 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1545 	unsigned long			task_state_change;
1546 # ifdef CONFIG_PREEMPT_RT
1547 	unsigned long			saved_state_change;
1548 # endif
1549 #endif
1550 	struct rcu_head			rcu;
1551 	refcount_t			rcu_users;
1552 	int				pagefault_disabled;
1553 #ifdef CONFIG_MMU
1554 	struct task_struct		*oom_reaper_list;
1555 	struct timer_list		oom_reaper_timer;
1556 #endif
1557 #ifdef CONFIG_VMAP_STACK
1558 	struct vm_struct		*stack_vm_area;
1559 #endif
1560 #ifdef CONFIG_THREAD_INFO_IN_TASK
1561 	/* A live task holds one reference: */
1562 	refcount_t			stack_refcount;
1563 #endif
1564 #ifdef CONFIG_LIVEPATCH
1565 	int patch_state;
1566 #endif
1567 #ifdef CONFIG_SECURITY
1568 	/* Used by LSM modules for access restriction: */
1569 	void				*security;
1570 #endif
1571 #ifdef CONFIG_BPF_SYSCALL
1572 	/* Used by BPF task local storage */
1573 	struct bpf_local_storage __rcu	*bpf_storage;
1574 	/* Used for BPF run context */
1575 	struct bpf_run_ctx		*bpf_ctx;
1576 #endif
1577 	/* Used by BPF for per-TASK xdp storage */
1578 	struct bpf_net_context		*bpf_net_context;
1579 
1580 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1581 	unsigned long			lowest_stack;
1582 	unsigned long			prev_lowest_stack;
1583 #endif
1584 
1585 #ifdef CONFIG_X86_MCE
1586 	void __user			*mce_vaddr;
1587 	__u64				mce_kflags;
1588 	u64				mce_addr;
1589 	__u64				mce_ripv : 1,
1590 					mce_whole_page : 1,
1591 					__mce_reserved : 62;
1592 	struct callback_head		mce_kill_me;
1593 	int				mce_count;
1594 #endif
1595 
1596 #ifdef CONFIG_KRETPROBES
1597 	struct llist_head               kretprobe_instances;
1598 #endif
1599 #ifdef CONFIG_RETHOOK
1600 	struct llist_head               rethooks;
1601 #endif
1602 
1603 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1604 	/*
1605 	 * If L1D flush is supported on mm context switch
1606 	 * then we use this callback head to queue kill work
1607 	 * to kill tasks that are not running on SMT disabled
1608 	 * cores
1609 	 */
1610 	struct callback_head		l1d_flush_kill;
1611 #endif
1612 
1613 #ifdef CONFIG_RV
1614 	/*
1615 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1616 	 * If we find justification for more monitors, we can think
1617 	 * about adding more or developing a dynamic method. So far,
1618 	 * none of these are justified.
1619 	 */
1620 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1621 #endif
1622 
1623 #ifdef CONFIG_USER_EVENTS
1624 	struct user_event_mm		*user_event_mm;
1625 #endif
1626 
1627 	/*
1628 	 * New fields for task_struct should be added above here, so that
1629 	 * they are included in the randomized portion of task_struct.
1630 	 */
1631 	randomized_struct_fields_end
1632 
1633 	/* CPU-specific state of this task: */
1634 	struct thread_struct		thread;
1635 
1636 	/*
1637 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1638 	 * structure.  It *MUST* be at the end of 'task_struct'.
1639 	 *
1640 	 * Do not put anything below here!
1641 	 */
1642 };
1643 
1644 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1645 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1646 
1647 static inline unsigned int __task_state_index(unsigned int tsk_state,
1648 					      unsigned int tsk_exit_state)
1649 {
1650 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1651 
1652 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1653 
1654 	if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1655 		state = TASK_REPORT_IDLE;
1656 
1657 	/*
1658 	 * We're lying here, but rather than expose a completely new task state
1659 	 * to userspace, we can make this appear as if the task has gone through
1660 	 * a regular rt_mutex_lock() call.
1661 	 * Report frozen tasks as uninterruptible.
1662 	 */
1663 	if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
1664 		state = TASK_UNINTERRUPTIBLE;
1665 
1666 	return fls(state);
1667 }
1668 
1669 static inline unsigned int task_state_index(struct task_struct *tsk)
1670 {
1671 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1672 }
1673 
1674 static inline char task_index_to_char(unsigned int state)
1675 {
1676 	static const char state_char[] = "RSDTtXZPI";
1677 
1678 	BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1679 
1680 	return state_char[state];
1681 }
1682 
1683 static inline char task_state_to_char(struct task_struct *tsk)
1684 {
1685 	return task_index_to_char(task_state_index(tsk));
1686 }
1687 
1688 extern struct pid *cad_pid;
1689 
1690 /*
1691  * Per process flags
1692  */
1693 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1694 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1695 #define PF_EXITING		0x00000004	/* Getting shut down */
1696 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1697 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1698 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1699 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1700 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1701 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1702 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1703 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1704 #define PF_MEMALLOC		0x00000800	/* Allocating memory to free memory. See memalloc_noreclaim_save() */
1705 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1706 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1707 #define PF_USER_WORKER		0x00004000	/* Kernel thread cloned from userspace thread */
1708 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1709 #define PF__HOLE__00010000	0x00010000
1710 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1711 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1712 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1713 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1714 						 * I am cleaning dirty pages from some other bdi. */
1715 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1716 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1717 #define PF__HOLE__00800000	0x00800000
1718 #define PF__HOLE__01000000	0x01000000
1719 #define PF__HOLE__02000000	0x02000000
1720 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1721 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1722 #define PF_MEMALLOC_PIN		0x10000000	/* Allocations constrained to zones which allow long term pinning.
1723 						 * See memalloc_pin_save() */
1724 #define PF_BLOCK_TS		0x20000000	/* plug has ts that needs updating */
1725 #define PF__HOLE__40000000	0x40000000
1726 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1727 
1728 /*
1729  * Only the _current_ task can read/write to tsk->flags, but other
1730  * tasks can access tsk->flags in readonly mode for example
1731  * with tsk_used_math (like during threaded core dumping).
1732  * There is however an exception to this rule during ptrace
1733  * or during fork: the ptracer task is allowed to write to the
1734  * child->flags of its traced child (same goes for fork, the parent
1735  * can write to the child->flags), because we're guaranteed the
1736  * child is not running and in turn not changing child->flags
1737  * at the same time the parent does it.
1738  */
1739 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1740 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1741 #define clear_used_math()			clear_stopped_child_used_math(current)
1742 #define set_used_math()				set_stopped_child_used_math(current)
1743 
1744 #define conditional_stopped_child_used_math(condition, child) \
1745 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1746 
1747 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1748 
1749 #define copy_to_stopped_child_used_math(child) \
1750 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1751 
1752 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1753 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1754 #define used_math()				tsk_used_math(current)
1755 
1756 static __always_inline bool is_percpu_thread(void)
1757 {
1758 #ifdef CONFIG_SMP
1759 	return (current->flags & PF_NO_SETAFFINITY) &&
1760 		(current->nr_cpus_allowed  == 1);
1761 #else
1762 	return true;
1763 #endif
1764 }
1765 
1766 /* Per-process atomic flags. */
1767 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1768 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1769 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1770 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1771 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1772 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1773 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1774 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1775 
1776 #define TASK_PFA_TEST(name, func)					\
1777 	static inline bool task_##func(struct task_struct *p)		\
1778 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1779 
1780 #define TASK_PFA_SET(name, func)					\
1781 	static inline void task_set_##func(struct task_struct *p)	\
1782 	{ set_bit(PFA_##name, &p->atomic_flags); }
1783 
1784 #define TASK_PFA_CLEAR(name, func)					\
1785 	static inline void task_clear_##func(struct task_struct *p)	\
1786 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1787 
1788 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1789 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1790 
1791 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1792 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1793 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1794 
1795 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1796 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1797 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1798 
1799 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1800 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1801 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1802 
1803 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1804 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1805 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1806 
1807 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1808 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1809 
1810 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1811 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1812 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1813 
1814 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1815 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1816 
1817 static inline void
1818 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1819 {
1820 	current->flags &= ~flags;
1821 	current->flags |= orig_flags & flags;
1822 }
1823 
1824 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1825 extern int task_can_attach(struct task_struct *p);
1826 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1827 extern void dl_bw_free(int cpu, u64 dl_bw);
1828 #ifdef CONFIG_SMP
1829 
1830 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1831 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1832 
1833 /**
1834  * set_cpus_allowed_ptr - set CPU affinity mask of a task
1835  * @p: the task
1836  * @new_mask: CPU affinity mask
1837  *
1838  * Return: zero if successful, or a negative error code
1839  */
1840 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1841 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1842 extern void release_user_cpus_ptr(struct task_struct *p);
1843 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1844 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1845 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1846 #else
1847 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1848 {
1849 }
1850 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1851 {
1852 	/* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1853 	if ((*cpumask_bits(new_mask) & 1) == 0)
1854 		return -EINVAL;
1855 	return 0;
1856 }
1857 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1858 {
1859 	if (src->user_cpus_ptr)
1860 		return -EINVAL;
1861 	return 0;
1862 }
1863 static inline void release_user_cpus_ptr(struct task_struct *p)
1864 {
1865 	WARN_ON(p->user_cpus_ptr);
1866 }
1867 
1868 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1869 {
1870 	return 0;
1871 }
1872 #endif
1873 
1874 extern int yield_to(struct task_struct *p, bool preempt);
1875 extern void set_user_nice(struct task_struct *p, long nice);
1876 extern int task_prio(const struct task_struct *p);
1877 
1878 /**
1879  * task_nice - return the nice value of a given task.
1880  * @p: the task in question.
1881  *
1882  * Return: The nice value [ -20 ... 0 ... 19 ].
1883  */
1884 static inline int task_nice(const struct task_struct *p)
1885 {
1886 	return PRIO_TO_NICE((p)->static_prio);
1887 }
1888 
1889 extern int can_nice(const struct task_struct *p, const int nice);
1890 extern int task_curr(const struct task_struct *p);
1891 extern int idle_cpu(int cpu);
1892 extern int available_idle_cpu(int cpu);
1893 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1894 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1895 extern void sched_set_fifo(struct task_struct *p);
1896 extern void sched_set_fifo_low(struct task_struct *p);
1897 extern void sched_set_normal(struct task_struct *p, int nice);
1898 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1899 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1900 extern struct task_struct *idle_task(int cpu);
1901 
1902 /**
1903  * is_idle_task - is the specified task an idle task?
1904  * @p: the task in question.
1905  *
1906  * Return: 1 if @p is an idle task. 0 otherwise.
1907  */
1908 static __always_inline bool is_idle_task(const struct task_struct *p)
1909 {
1910 	return !!(p->flags & PF_IDLE);
1911 }
1912 
1913 extern struct task_struct *curr_task(int cpu);
1914 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1915 
1916 void yield(void);
1917 
1918 union thread_union {
1919 	struct task_struct task;
1920 #ifndef CONFIG_THREAD_INFO_IN_TASK
1921 	struct thread_info thread_info;
1922 #endif
1923 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1924 };
1925 
1926 #ifndef CONFIG_THREAD_INFO_IN_TASK
1927 extern struct thread_info init_thread_info;
1928 #endif
1929 
1930 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1931 
1932 #ifdef CONFIG_THREAD_INFO_IN_TASK
1933 # define task_thread_info(task)	(&(task)->thread_info)
1934 #else
1935 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1936 #endif
1937 
1938 /*
1939  * find a task by one of its numerical ids
1940  *
1941  * find_task_by_pid_ns():
1942  *      finds a task by its pid in the specified namespace
1943  * find_task_by_vpid():
1944  *      finds a task by its virtual pid
1945  *
1946  * see also find_vpid() etc in include/linux/pid.h
1947  */
1948 
1949 extern struct task_struct *find_task_by_vpid(pid_t nr);
1950 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1951 
1952 /*
1953  * find a task by its virtual pid and get the task struct
1954  */
1955 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1956 
1957 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1958 extern int wake_up_process(struct task_struct *tsk);
1959 extern void wake_up_new_task(struct task_struct *tsk);
1960 
1961 #ifdef CONFIG_SMP
1962 extern void kick_process(struct task_struct *tsk);
1963 #else
1964 static inline void kick_process(struct task_struct *tsk) { }
1965 #endif
1966 
1967 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1968 #define set_task_comm(tsk, from) ({			\
1969 	BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN);	\
1970 	__set_task_comm(tsk, from, false);		\
1971 })
1972 
1973 /*
1974  * - Why not use task_lock()?
1975  *   User space can randomly change their names anyway, so locking for readers
1976  *   doesn't make sense. For writers, locking is probably necessary, as a race
1977  *   condition could lead to long-term mixed results.
1978  *   The strscpy_pad() in __set_task_comm() can ensure that the task comm is
1979  *   always NUL-terminated and zero-padded. Therefore the race condition between
1980  *   reader and writer is not an issue.
1981  *
1982  * - BUILD_BUG_ON() can help prevent the buf from being truncated.
1983  *   Since the callers don't perform any return value checks, this safeguard is
1984  *   necessary.
1985  */
1986 #define get_task_comm(buf, tsk) ({			\
1987 	BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN);	\
1988 	strscpy_pad(buf, (tsk)->comm);			\
1989 	buf;						\
1990 })
1991 
1992 #ifdef CONFIG_SMP
1993 static __always_inline void scheduler_ipi(void)
1994 {
1995 	/*
1996 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1997 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1998 	 * this IPI.
1999 	 */
2000 	preempt_fold_need_resched();
2001 }
2002 #else
2003 static inline void scheduler_ipi(void) { }
2004 #endif
2005 
2006 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2007 
2008 /*
2009  * Set thread flags in other task's structures.
2010  * See asm/thread_info.h for TIF_xxxx flags available:
2011  */
2012 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2013 {
2014 	set_ti_thread_flag(task_thread_info(tsk), flag);
2015 }
2016 
2017 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2018 {
2019 	clear_ti_thread_flag(task_thread_info(tsk), flag);
2020 }
2021 
2022 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2023 					  bool value)
2024 {
2025 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
2026 }
2027 
2028 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2029 {
2030 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2031 }
2032 
2033 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2034 {
2035 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2036 }
2037 
2038 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2039 {
2040 	return test_ti_thread_flag(task_thread_info(tsk), flag);
2041 }
2042 
2043 static inline void set_tsk_need_resched(struct task_struct *tsk)
2044 {
2045 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2046 }
2047 
2048 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2049 {
2050 	atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2051 			   (atomic_long_t *)&task_thread_info(tsk)->flags);
2052 }
2053 
2054 static inline int test_tsk_need_resched(struct task_struct *tsk)
2055 {
2056 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2057 }
2058 
2059 /*
2060  * cond_resched() and cond_resched_lock(): latency reduction via
2061  * explicit rescheduling in places that are safe. The return
2062  * value indicates whether a reschedule was done in fact.
2063  * cond_resched_lock() will drop the spinlock before scheduling,
2064  */
2065 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2066 extern int __cond_resched(void);
2067 
2068 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2069 
2070 void sched_dynamic_klp_enable(void);
2071 void sched_dynamic_klp_disable(void);
2072 
2073 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2074 
2075 static __always_inline int _cond_resched(void)
2076 {
2077 	return static_call_mod(cond_resched)();
2078 }
2079 
2080 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2081 
2082 extern int dynamic_cond_resched(void);
2083 
2084 static __always_inline int _cond_resched(void)
2085 {
2086 	return dynamic_cond_resched();
2087 }
2088 
2089 #else /* !CONFIG_PREEMPTION */
2090 
2091 static inline int _cond_resched(void)
2092 {
2093 	klp_sched_try_switch();
2094 	return __cond_resched();
2095 }
2096 
2097 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2098 
2099 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2100 
2101 static inline int _cond_resched(void)
2102 {
2103 	klp_sched_try_switch();
2104 	return 0;
2105 }
2106 
2107 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2108 
2109 #define cond_resched() ({			\
2110 	__might_resched(__FILE__, __LINE__, 0);	\
2111 	_cond_resched();			\
2112 })
2113 
2114 extern int __cond_resched_lock(spinlock_t *lock);
2115 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2116 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2117 
2118 #define MIGHT_RESCHED_RCU_SHIFT		8
2119 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2120 
2121 #ifndef CONFIG_PREEMPT_RT
2122 /*
2123  * Non RT kernels have an elevated preempt count due to the held lock,
2124  * but are not allowed to be inside a RCU read side critical section
2125  */
2126 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2127 #else
2128 /*
2129  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2130  * cond_resched*lock() has to take that into account because it checks for
2131  * preempt_count() and rcu_preempt_depth().
2132  */
2133 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2134 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2135 #endif
2136 
2137 #define cond_resched_lock(lock) ({						\
2138 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2139 	__cond_resched_lock(lock);						\
2140 })
2141 
2142 #define cond_resched_rwlock_read(lock) ({					\
2143 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2144 	__cond_resched_rwlock_read(lock);					\
2145 })
2146 
2147 #define cond_resched_rwlock_write(lock) ({					\
2148 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2149 	__cond_resched_rwlock_write(lock);					\
2150 })
2151 
2152 static __always_inline bool need_resched(void)
2153 {
2154 	return unlikely(tif_need_resched());
2155 }
2156 
2157 /*
2158  * Wrappers for p->thread_info->cpu access. No-op on UP.
2159  */
2160 #ifdef CONFIG_SMP
2161 
2162 static inline unsigned int task_cpu(const struct task_struct *p)
2163 {
2164 	return READ_ONCE(task_thread_info(p)->cpu);
2165 }
2166 
2167 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2168 
2169 #else
2170 
2171 static inline unsigned int task_cpu(const struct task_struct *p)
2172 {
2173 	return 0;
2174 }
2175 
2176 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2177 {
2178 }
2179 
2180 #endif /* CONFIG_SMP */
2181 
2182 static inline bool task_is_runnable(struct task_struct *p)
2183 {
2184 	return p->on_rq && !p->se.sched_delayed;
2185 }
2186 
2187 extern bool sched_task_on_rq(struct task_struct *p);
2188 extern unsigned long get_wchan(struct task_struct *p);
2189 extern struct task_struct *cpu_curr_snapshot(int cpu);
2190 
2191 #include <linux/spinlock.h>
2192 
2193 /*
2194  * In order to reduce various lock holder preemption latencies provide an
2195  * interface to see if a vCPU is currently running or not.
2196  *
2197  * This allows us to terminate optimistic spin loops and block, analogous to
2198  * the native optimistic spin heuristic of testing if the lock owner task is
2199  * running or not.
2200  */
2201 #ifndef vcpu_is_preempted
2202 static inline bool vcpu_is_preempted(int cpu)
2203 {
2204 	return false;
2205 }
2206 #endif
2207 
2208 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2209 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2210 
2211 #ifndef TASK_SIZE_OF
2212 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2213 #endif
2214 
2215 #ifdef CONFIG_SMP
2216 static inline bool owner_on_cpu(struct task_struct *owner)
2217 {
2218 	/*
2219 	 * As lock holder preemption issue, we both skip spinning if
2220 	 * task is not on cpu or its cpu is preempted
2221 	 */
2222 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2223 }
2224 
2225 /* Returns effective CPU energy utilization, as seen by the scheduler */
2226 unsigned long sched_cpu_util(int cpu);
2227 #endif /* CONFIG_SMP */
2228 
2229 #ifdef CONFIG_SCHED_CORE
2230 extern void sched_core_free(struct task_struct *tsk);
2231 extern void sched_core_fork(struct task_struct *p);
2232 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2233 				unsigned long uaddr);
2234 extern int sched_core_idle_cpu(int cpu);
2235 #else
2236 static inline void sched_core_free(struct task_struct *tsk) { }
2237 static inline void sched_core_fork(struct task_struct *p) { }
2238 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2239 #endif
2240 
2241 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2242 
2243 #ifdef CONFIG_MEM_ALLOC_PROFILING
2244 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2245 {
2246 	swap(current->alloc_tag, tag);
2247 	return tag;
2248 }
2249 
2250 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2251 {
2252 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2253 	WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2254 #endif
2255 	current->alloc_tag = old;
2256 }
2257 #else
2258 #define alloc_tag_save(_tag)			NULL
2259 #define alloc_tag_restore(_tag, _old)		do {} while (0)
2260 #endif
2261 
2262 #endif
2263