xref: /linux/include/linux/sched.h (revision b81a677400755cf24c971d1342c4c53d81744d82)
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/kmsan_types.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/irqflags.h>
22 #include <linux/seccomp.h>
23 #include <linux/nodemask.h>
24 #include <linux/rcupdate.h>
25 #include <linux/refcount.h>
26 #include <linux/resource.h>
27 #include <linux/latencytop.h>
28 #include <linux/sched/prio.h>
29 #include <linux/sched/types.h>
30 #include <linux/signal_types.h>
31 #include <linux/syscall_user_dispatch.h>
32 #include <linux/mm_types_task.h>
33 #include <linux/task_io_accounting.h>
34 #include <linux/posix-timers.h>
35 #include <linux/rseq.h>
36 #include <linux/seqlock.h>
37 #include <linux/kcsan.h>
38 #include <linux/rv.h>
39 #include <asm/kmap_size.h>
40 
41 /* task_struct member predeclarations (sorted alphabetically): */
42 struct audit_context;
43 struct backing_dev_info;
44 struct bio_list;
45 struct blk_plug;
46 struct bpf_local_storage;
47 struct bpf_run_ctx;
48 struct capture_control;
49 struct cfs_rq;
50 struct fs_struct;
51 struct futex_pi_state;
52 struct io_context;
53 struct io_uring_task;
54 struct mempolicy;
55 struct nameidata;
56 struct nsproxy;
57 struct perf_event_context;
58 struct pid_namespace;
59 struct pipe_inode_info;
60 struct rcu_node;
61 struct reclaim_state;
62 struct robust_list_head;
63 struct root_domain;
64 struct rq;
65 struct sched_attr;
66 struct sched_param;
67 struct seq_file;
68 struct sighand_struct;
69 struct signal_struct;
70 struct task_delay_info;
71 struct task_group;
72 
73 /*
74  * Task state bitmask. NOTE! These bits are also
75  * encoded in fs/proc/array.c: get_task_state().
76  *
77  * We have two separate sets of flags: task->state
78  * is about runnability, while task->exit_state are
79  * about the task exiting. Confusing, but this way
80  * modifying one set can't modify the other one by
81  * mistake.
82  */
83 
84 /* Used in tsk->state: */
85 #define TASK_RUNNING			0x00000000
86 #define TASK_INTERRUPTIBLE		0x00000001
87 #define TASK_UNINTERRUPTIBLE		0x00000002
88 #define __TASK_STOPPED			0x00000004
89 #define __TASK_TRACED			0x00000008
90 /* Used in tsk->exit_state: */
91 #define EXIT_DEAD			0x00000010
92 #define EXIT_ZOMBIE			0x00000020
93 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
94 /* Used in tsk->state again: */
95 #define TASK_PARKED			0x00000040
96 #define TASK_DEAD			0x00000080
97 #define TASK_WAKEKILL			0x00000100
98 #define TASK_WAKING			0x00000200
99 #define TASK_NOLOAD			0x00000400
100 #define TASK_NEW			0x00000800
101 #define TASK_RTLOCK_WAIT		0x00001000
102 #define TASK_FREEZABLE			0x00002000
103 #define __TASK_FREEZABLE_UNSAFE	       (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
104 #define TASK_FROZEN			0x00008000
105 #define TASK_STATE_MAX			0x00010000
106 
107 #define TASK_ANY			(TASK_STATE_MAX-1)
108 
109 /*
110  * DO NOT ADD ANY NEW USERS !
111  */
112 #define TASK_FREEZABLE_UNSAFE		(TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
113 
114 /* Convenience macros for the sake of set_current_state: */
115 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
116 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
117 #define TASK_TRACED			__TASK_TRACED
118 
119 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
120 
121 /* Convenience macros for the sake of wake_up(): */
122 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
123 
124 /* get_task_state(): */
125 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
126 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
127 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
128 					 TASK_PARKED)
129 
130 #define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
131 
132 #define task_is_traced(task)		((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
133 #define task_is_stopped(task)		((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
134 #define task_is_stopped_or_traced(task)	((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
135 
136 /*
137  * Special states are those that do not use the normal wait-loop pattern. See
138  * the comment with set_special_state().
139  */
140 #define is_special_task_state(state)				\
141 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
142 
143 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
144 # define debug_normal_state_change(state_value)				\
145 	do {								\
146 		WARN_ON_ONCE(is_special_task_state(state_value));	\
147 		current->task_state_change = _THIS_IP_;			\
148 	} while (0)
149 
150 # define debug_special_state_change(state_value)			\
151 	do {								\
152 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
153 		current->task_state_change = _THIS_IP_;			\
154 	} while (0)
155 
156 # define debug_rtlock_wait_set_state()					\
157 	do {								 \
158 		current->saved_state_change = current->task_state_change;\
159 		current->task_state_change = _THIS_IP_;			 \
160 	} while (0)
161 
162 # define debug_rtlock_wait_restore_state()				\
163 	do {								 \
164 		current->task_state_change = current->saved_state_change;\
165 	} while (0)
166 
167 #else
168 # define debug_normal_state_change(cond)	do { } while (0)
169 # define debug_special_state_change(cond)	do { } while (0)
170 # define debug_rtlock_wait_set_state()		do { } while (0)
171 # define debug_rtlock_wait_restore_state()	do { } while (0)
172 #endif
173 
174 /*
175  * set_current_state() includes a barrier so that the write of current->state
176  * is correctly serialised wrt the caller's subsequent test of whether to
177  * actually sleep:
178  *
179  *   for (;;) {
180  *	set_current_state(TASK_UNINTERRUPTIBLE);
181  *	if (CONDITION)
182  *	   break;
183  *
184  *	schedule();
185  *   }
186  *   __set_current_state(TASK_RUNNING);
187  *
188  * If the caller does not need such serialisation (because, for instance, the
189  * CONDITION test and condition change and wakeup are under the same lock) then
190  * use __set_current_state().
191  *
192  * The above is typically ordered against the wakeup, which does:
193  *
194  *   CONDITION = 1;
195  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
196  *
197  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
198  * accessing p->state.
199  *
200  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
201  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
202  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
203  *
204  * However, with slightly different timing the wakeup TASK_RUNNING store can
205  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
206  * a problem either because that will result in one extra go around the loop
207  * and our @cond test will save the day.
208  *
209  * Also see the comments of try_to_wake_up().
210  */
211 #define __set_current_state(state_value)				\
212 	do {								\
213 		debug_normal_state_change((state_value));		\
214 		WRITE_ONCE(current->__state, (state_value));		\
215 	} while (0)
216 
217 #define set_current_state(state_value)					\
218 	do {								\
219 		debug_normal_state_change((state_value));		\
220 		smp_store_mb(current->__state, (state_value));		\
221 	} while (0)
222 
223 /*
224  * set_special_state() should be used for those states when the blocking task
225  * can not use the regular condition based wait-loop. In that case we must
226  * serialize against wakeups such that any possible in-flight TASK_RUNNING
227  * stores will not collide with our state change.
228  */
229 #define set_special_state(state_value)					\
230 	do {								\
231 		unsigned long flags; /* may shadow */			\
232 									\
233 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
234 		debug_special_state_change((state_value));		\
235 		WRITE_ONCE(current->__state, (state_value));		\
236 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
237 	} while (0)
238 
239 /*
240  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
241  *
242  * RT's spin/rwlock substitutions are state preserving. The state of the
243  * task when blocking on the lock is saved in task_struct::saved_state and
244  * restored after the lock has been acquired.  These operations are
245  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
246  * lock related wakeups while the task is blocked on the lock are
247  * redirected to operate on task_struct::saved_state to ensure that these
248  * are not dropped. On restore task_struct::saved_state is set to
249  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
250  *
251  * The lock operation looks like this:
252  *
253  *	current_save_and_set_rtlock_wait_state();
254  *	for (;;) {
255  *		if (try_lock())
256  *			break;
257  *		raw_spin_unlock_irq(&lock->wait_lock);
258  *		schedule_rtlock();
259  *		raw_spin_lock_irq(&lock->wait_lock);
260  *		set_current_state(TASK_RTLOCK_WAIT);
261  *	}
262  *	current_restore_rtlock_saved_state();
263  */
264 #define current_save_and_set_rtlock_wait_state()			\
265 	do {								\
266 		lockdep_assert_irqs_disabled();				\
267 		raw_spin_lock(&current->pi_lock);			\
268 		current->saved_state = current->__state;		\
269 		debug_rtlock_wait_set_state();				\
270 		WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);		\
271 		raw_spin_unlock(&current->pi_lock);			\
272 	} while (0);
273 
274 #define current_restore_rtlock_saved_state()				\
275 	do {								\
276 		lockdep_assert_irqs_disabled();				\
277 		raw_spin_lock(&current->pi_lock);			\
278 		debug_rtlock_wait_restore_state();			\
279 		WRITE_ONCE(current->__state, current->saved_state);	\
280 		current->saved_state = TASK_RUNNING;			\
281 		raw_spin_unlock(&current->pi_lock);			\
282 	} while (0);
283 
284 #define get_current_state()	READ_ONCE(current->__state)
285 
286 /*
287  * Define the task command name length as enum, then it can be visible to
288  * BPF programs.
289  */
290 enum {
291 	TASK_COMM_LEN = 16,
292 };
293 
294 extern void scheduler_tick(void);
295 
296 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
297 
298 extern long schedule_timeout(long timeout);
299 extern long schedule_timeout_interruptible(long timeout);
300 extern long schedule_timeout_killable(long timeout);
301 extern long schedule_timeout_uninterruptible(long timeout);
302 extern long schedule_timeout_idle(long timeout);
303 asmlinkage void schedule(void);
304 extern void schedule_preempt_disabled(void);
305 asmlinkage void preempt_schedule_irq(void);
306 #ifdef CONFIG_PREEMPT_RT
307  extern void schedule_rtlock(void);
308 #endif
309 
310 extern int __must_check io_schedule_prepare(void);
311 extern void io_schedule_finish(int token);
312 extern long io_schedule_timeout(long timeout);
313 extern void io_schedule(void);
314 
315 /**
316  * struct prev_cputime - snapshot of system and user cputime
317  * @utime: time spent in user mode
318  * @stime: time spent in system mode
319  * @lock: protects the above two fields
320  *
321  * Stores previous user/system time values such that we can guarantee
322  * monotonicity.
323  */
324 struct prev_cputime {
325 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
326 	u64				utime;
327 	u64				stime;
328 	raw_spinlock_t			lock;
329 #endif
330 };
331 
332 enum vtime_state {
333 	/* Task is sleeping or running in a CPU with VTIME inactive: */
334 	VTIME_INACTIVE = 0,
335 	/* Task is idle */
336 	VTIME_IDLE,
337 	/* Task runs in kernelspace in a CPU with VTIME active: */
338 	VTIME_SYS,
339 	/* Task runs in userspace in a CPU with VTIME active: */
340 	VTIME_USER,
341 	/* Task runs as guests in a CPU with VTIME active: */
342 	VTIME_GUEST,
343 };
344 
345 struct vtime {
346 	seqcount_t		seqcount;
347 	unsigned long long	starttime;
348 	enum vtime_state	state;
349 	unsigned int		cpu;
350 	u64			utime;
351 	u64			stime;
352 	u64			gtime;
353 };
354 
355 /*
356  * Utilization clamp constraints.
357  * @UCLAMP_MIN:	Minimum utilization
358  * @UCLAMP_MAX:	Maximum utilization
359  * @UCLAMP_CNT:	Utilization clamp constraints count
360  */
361 enum uclamp_id {
362 	UCLAMP_MIN = 0,
363 	UCLAMP_MAX,
364 	UCLAMP_CNT
365 };
366 
367 #ifdef CONFIG_SMP
368 extern struct root_domain def_root_domain;
369 extern struct mutex sched_domains_mutex;
370 #endif
371 
372 struct sched_info {
373 #ifdef CONFIG_SCHED_INFO
374 	/* Cumulative counters: */
375 
376 	/* # of times we have run on this CPU: */
377 	unsigned long			pcount;
378 
379 	/* Time spent waiting on a runqueue: */
380 	unsigned long long		run_delay;
381 
382 	/* Timestamps: */
383 
384 	/* When did we last run on a CPU? */
385 	unsigned long long		last_arrival;
386 
387 	/* When were we last queued to run? */
388 	unsigned long long		last_queued;
389 
390 #endif /* CONFIG_SCHED_INFO */
391 };
392 
393 /*
394  * Integer metrics need fixed point arithmetic, e.g., sched/fair
395  * has a few: load, load_avg, util_avg, freq, and capacity.
396  *
397  * We define a basic fixed point arithmetic range, and then formalize
398  * all these metrics based on that basic range.
399  */
400 # define SCHED_FIXEDPOINT_SHIFT		10
401 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
402 
403 /* Increase resolution of cpu_capacity calculations */
404 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
405 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
406 
407 struct load_weight {
408 	unsigned long			weight;
409 	u32				inv_weight;
410 };
411 
412 /**
413  * struct util_est - Estimation utilization of FAIR tasks
414  * @enqueued: instantaneous estimated utilization of a task/cpu
415  * @ewma:     the Exponential Weighted Moving Average (EWMA)
416  *            utilization of a task
417  *
418  * Support data structure to track an Exponential Weighted Moving Average
419  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
420  * average each time a task completes an activation. Sample's weight is chosen
421  * so that the EWMA will be relatively insensitive to transient changes to the
422  * task's workload.
423  *
424  * The enqueued attribute has a slightly different meaning for tasks and cpus:
425  * - task:   the task's util_avg at last task dequeue time
426  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
427  * Thus, the util_est.enqueued of a task represents the contribution on the
428  * estimated utilization of the CPU where that task is currently enqueued.
429  *
430  * Only for tasks we track a moving average of the past instantaneous
431  * estimated utilization. This allows to absorb sporadic drops in utilization
432  * of an otherwise almost periodic task.
433  *
434  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
435  * updates. When a task is dequeued, its util_est should not be updated if its
436  * util_avg has not been updated in the meantime.
437  * This information is mapped into the MSB bit of util_est.enqueued at dequeue
438  * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
439  * for a task) it is safe to use MSB.
440  */
441 struct util_est {
442 	unsigned int			enqueued;
443 	unsigned int			ewma;
444 #define UTIL_EST_WEIGHT_SHIFT		2
445 #define UTIL_AVG_UNCHANGED		0x80000000
446 } __attribute__((__aligned__(sizeof(u64))));
447 
448 /*
449  * The load/runnable/util_avg accumulates an infinite geometric series
450  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
451  *
452  * [load_avg definition]
453  *
454  *   load_avg = runnable% * scale_load_down(load)
455  *
456  * [runnable_avg definition]
457  *
458  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
459  *
460  * [util_avg definition]
461  *
462  *   util_avg = running% * SCHED_CAPACITY_SCALE
463  *
464  * where runnable% is the time ratio that a sched_entity is runnable and
465  * running% the time ratio that a sched_entity is running.
466  *
467  * For cfs_rq, they are the aggregated values of all runnable and blocked
468  * sched_entities.
469  *
470  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
471  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
472  * for computing those signals (see update_rq_clock_pelt())
473  *
474  * N.B., the above ratios (runnable% and running%) themselves are in the
475  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
476  * to as large a range as necessary. This is for example reflected by
477  * util_avg's SCHED_CAPACITY_SCALE.
478  *
479  * [Overflow issue]
480  *
481  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
482  * with the highest load (=88761), always runnable on a single cfs_rq,
483  * and should not overflow as the number already hits PID_MAX_LIMIT.
484  *
485  * For all other cases (including 32-bit kernels), struct load_weight's
486  * weight will overflow first before we do, because:
487  *
488  *    Max(load_avg) <= Max(load.weight)
489  *
490  * Then it is the load_weight's responsibility to consider overflow
491  * issues.
492  */
493 struct sched_avg {
494 	u64				last_update_time;
495 	u64				load_sum;
496 	u64				runnable_sum;
497 	u32				util_sum;
498 	u32				period_contrib;
499 	unsigned long			load_avg;
500 	unsigned long			runnable_avg;
501 	unsigned long			util_avg;
502 	struct util_est			util_est;
503 } ____cacheline_aligned;
504 
505 struct sched_statistics {
506 #ifdef CONFIG_SCHEDSTATS
507 	u64				wait_start;
508 	u64				wait_max;
509 	u64				wait_count;
510 	u64				wait_sum;
511 	u64				iowait_count;
512 	u64				iowait_sum;
513 
514 	u64				sleep_start;
515 	u64				sleep_max;
516 	s64				sum_sleep_runtime;
517 
518 	u64				block_start;
519 	u64				block_max;
520 	s64				sum_block_runtime;
521 
522 	u64				exec_max;
523 	u64				slice_max;
524 
525 	u64				nr_migrations_cold;
526 	u64				nr_failed_migrations_affine;
527 	u64				nr_failed_migrations_running;
528 	u64				nr_failed_migrations_hot;
529 	u64				nr_forced_migrations;
530 
531 	u64				nr_wakeups;
532 	u64				nr_wakeups_sync;
533 	u64				nr_wakeups_migrate;
534 	u64				nr_wakeups_local;
535 	u64				nr_wakeups_remote;
536 	u64				nr_wakeups_affine;
537 	u64				nr_wakeups_affine_attempts;
538 	u64				nr_wakeups_passive;
539 	u64				nr_wakeups_idle;
540 
541 #ifdef CONFIG_SCHED_CORE
542 	u64				core_forceidle_sum;
543 #endif
544 #endif /* CONFIG_SCHEDSTATS */
545 } ____cacheline_aligned;
546 
547 struct sched_entity {
548 	/* For load-balancing: */
549 	struct load_weight		load;
550 	struct rb_node			run_node;
551 	struct list_head		group_node;
552 	unsigned int			on_rq;
553 
554 	u64				exec_start;
555 	u64				sum_exec_runtime;
556 	u64				vruntime;
557 	u64				prev_sum_exec_runtime;
558 
559 	u64				nr_migrations;
560 
561 #ifdef CONFIG_FAIR_GROUP_SCHED
562 	int				depth;
563 	struct sched_entity		*parent;
564 	/* rq on which this entity is (to be) queued: */
565 	struct cfs_rq			*cfs_rq;
566 	/* rq "owned" by this entity/group: */
567 	struct cfs_rq			*my_q;
568 	/* cached value of my_q->h_nr_running */
569 	unsigned long			runnable_weight;
570 #endif
571 
572 #ifdef CONFIG_SMP
573 	/*
574 	 * Per entity load average tracking.
575 	 *
576 	 * Put into separate cache line so it does not
577 	 * collide with read-mostly values above.
578 	 */
579 	struct sched_avg		avg;
580 #endif
581 };
582 
583 struct sched_rt_entity {
584 	struct list_head		run_list;
585 	unsigned long			timeout;
586 	unsigned long			watchdog_stamp;
587 	unsigned int			time_slice;
588 	unsigned short			on_rq;
589 	unsigned short			on_list;
590 
591 	struct sched_rt_entity		*back;
592 #ifdef CONFIG_RT_GROUP_SCHED
593 	struct sched_rt_entity		*parent;
594 	/* rq on which this entity is (to be) queued: */
595 	struct rt_rq			*rt_rq;
596 	/* rq "owned" by this entity/group: */
597 	struct rt_rq			*my_q;
598 #endif
599 } __randomize_layout;
600 
601 struct sched_dl_entity {
602 	struct rb_node			rb_node;
603 
604 	/*
605 	 * Original scheduling parameters. Copied here from sched_attr
606 	 * during sched_setattr(), they will remain the same until
607 	 * the next sched_setattr().
608 	 */
609 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
610 	u64				dl_deadline;	/* Relative deadline of each instance	*/
611 	u64				dl_period;	/* Separation of two instances (period) */
612 	u64				dl_bw;		/* dl_runtime / dl_period		*/
613 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
614 
615 	/*
616 	 * Actual scheduling parameters. Initialized with the values above,
617 	 * they are continuously updated during task execution. Note that
618 	 * the remaining runtime could be < 0 in case we are in overrun.
619 	 */
620 	s64				runtime;	/* Remaining runtime for this instance	*/
621 	u64				deadline;	/* Absolute deadline for this instance	*/
622 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
623 
624 	/*
625 	 * Some bool flags:
626 	 *
627 	 * @dl_throttled tells if we exhausted the runtime. If so, the
628 	 * task has to wait for a replenishment to be performed at the
629 	 * next firing of dl_timer.
630 	 *
631 	 * @dl_yielded tells if task gave up the CPU before consuming
632 	 * all its available runtime during the last job.
633 	 *
634 	 * @dl_non_contending tells if the task is inactive while still
635 	 * contributing to the active utilization. In other words, it
636 	 * indicates if the inactive timer has been armed and its handler
637 	 * has not been executed yet. This flag is useful to avoid race
638 	 * conditions between the inactive timer handler and the wakeup
639 	 * code.
640 	 *
641 	 * @dl_overrun tells if the task asked to be informed about runtime
642 	 * overruns.
643 	 */
644 	unsigned int			dl_throttled      : 1;
645 	unsigned int			dl_yielded        : 1;
646 	unsigned int			dl_non_contending : 1;
647 	unsigned int			dl_overrun	  : 1;
648 
649 	/*
650 	 * Bandwidth enforcement timer. Each -deadline task has its
651 	 * own bandwidth to be enforced, thus we need one timer per task.
652 	 */
653 	struct hrtimer			dl_timer;
654 
655 	/*
656 	 * Inactive timer, responsible for decreasing the active utilization
657 	 * at the "0-lag time". When a -deadline task blocks, it contributes
658 	 * to GRUB's active utilization until the "0-lag time", hence a
659 	 * timer is needed to decrease the active utilization at the correct
660 	 * time.
661 	 */
662 	struct hrtimer inactive_timer;
663 
664 #ifdef CONFIG_RT_MUTEXES
665 	/*
666 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
667 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
668 	 * to (the original one/itself).
669 	 */
670 	struct sched_dl_entity *pi_se;
671 #endif
672 };
673 
674 #ifdef CONFIG_UCLAMP_TASK
675 /* Number of utilization clamp buckets (shorter alias) */
676 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
677 
678 /*
679  * Utilization clamp for a scheduling entity
680  * @value:		clamp value "assigned" to a se
681  * @bucket_id:		bucket index corresponding to the "assigned" value
682  * @active:		the se is currently refcounted in a rq's bucket
683  * @user_defined:	the requested clamp value comes from user-space
684  *
685  * The bucket_id is the index of the clamp bucket matching the clamp value
686  * which is pre-computed and stored to avoid expensive integer divisions from
687  * the fast path.
688  *
689  * The active bit is set whenever a task has got an "effective" value assigned,
690  * which can be different from the clamp value "requested" from user-space.
691  * This allows to know a task is refcounted in the rq's bucket corresponding
692  * to the "effective" bucket_id.
693  *
694  * The user_defined bit is set whenever a task has got a task-specific clamp
695  * value requested from userspace, i.e. the system defaults apply to this task
696  * just as a restriction. This allows to relax default clamps when a less
697  * restrictive task-specific value has been requested, thus allowing to
698  * implement a "nice" semantic. For example, a task running with a 20%
699  * default boost can still drop its own boosting to 0%.
700  */
701 struct uclamp_se {
702 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
703 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
704 	unsigned int active		: 1;
705 	unsigned int user_defined	: 1;
706 };
707 #endif /* CONFIG_UCLAMP_TASK */
708 
709 union rcu_special {
710 	struct {
711 		u8			blocked;
712 		u8			need_qs;
713 		u8			exp_hint; /* Hint for performance. */
714 		u8			need_mb; /* Readers need smp_mb(). */
715 	} b; /* Bits. */
716 	u32 s; /* Set of bits. */
717 };
718 
719 enum perf_event_task_context {
720 	perf_invalid_context = -1,
721 	perf_hw_context = 0,
722 	perf_sw_context,
723 	perf_nr_task_contexts,
724 };
725 
726 struct wake_q_node {
727 	struct wake_q_node *next;
728 };
729 
730 struct kmap_ctrl {
731 #ifdef CONFIG_KMAP_LOCAL
732 	int				idx;
733 	pte_t				pteval[KM_MAX_IDX];
734 #endif
735 };
736 
737 struct task_struct {
738 #ifdef CONFIG_THREAD_INFO_IN_TASK
739 	/*
740 	 * For reasons of header soup (see current_thread_info()), this
741 	 * must be the first element of task_struct.
742 	 */
743 	struct thread_info		thread_info;
744 #endif
745 	unsigned int			__state;
746 
747 #ifdef CONFIG_PREEMPT_RT
748 	/* saved state for "spinlock sleepers" */
749 	unsigned int			saved_state;
750 #endif
751 
752 	/*
753 	 * This begins the randomizable portion of task_struct. Only
754 	 * scheduling-critical items should be added above here.
755 	 */
756 	randomized_struct_fields_start
757 
758 	void				*stack;
759 	refcount_t			usage;
760 	/* Per task flags (PF_*), defined further below: */
761 	unsigned int			flags;
762 	unsigned int			ptrace;
763 
764 #ifdef CONFIG_SMP
765 	int				on_cpu;
766 	struct __call_single_node	wake_entry;
767 	unsigned int			wakee_flips;
768 	unsigned long			wakee_flip_decay_ts;
769 	struct task_struct		*last_wakee;
770 
771 	/*
772 	 * recent_used_cpu is initially set as the last CPU used by a task
773 	 * that wakes affine another task. Waker/wakee relationships can
774 	 * push tasks around a CPU where each wakeup moves to the next one.
775 	 * Tracking a recently used CPU allows a quick search for a recently
776 	 * used CPU that may be idle.
777 	 */
778 	int				recent_used_cpu;
779 	int				wake_cpu;
780 #endif
781 	int				on_rq;
782 
783 	int				prio;
784 	int				static_prio;
785 	int				normal_prio;
786 	unsigned int			rt_priority;
787 
788 	struct sched_entity		se;
789 	struct sched_rt_entity		rt;
790 	struct sched_dl_entity		dl;
791 	const struct sched_class	*sched_class;
792 
793 #ifdef CONFIG_SCHED_CORE
794 	struct rb_node			core_node;
795 	unsigned long			core_cookie;
796 	unsigned int			core_occupation;
797 #endif
798 
799 #ifdef CONFIG_CGROUP_SCHED
800 	struct task_group		*sched_task_group;
801 #endif
802 
803 #ifdef CONFIG_UCLAMP_TASK
804 	/*
805 	 * Clamp values requested for a scheduling entity.
806 	 * Must be updated with task_rq_lock() held.
807 	 */
808 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
809 	/*
810 	 * Effective clamp values used for a scheduling entity.
811 	 * Must be updated with task_rq_lock() held.
812 	 */
813 	struct uclamp_se		uclamp[UCLAMP_CNT];
814 #endif
815 
816 	struct sched_statistics         stats;
817 
818 #ifdef CONFIG_PREEMPT_NOTIFIERS
819 	/* List of struct preempt_notifier: */
820 	struct hlist_head		preempt_notifiers;
821 #endif
822 
823 #ifdef CONFIG_BLK_DEV_IO_TRACE
824 	unsigned int			btrace_seq;
825 #endif
826 
827 	unsigned int			policy;
828 	int				nr_cpus_allowed;
829 	const cpumask_t			*cpus_ptr;
830 	cpumask_t			*user_cpus_ptr;
831 	cpumask_t			cpus_mask;
832 	void				*migration_pending;
833 #ifdef CONFIG_SMP
834 	unsigned short			migration_disabled;
835 #endif
836 	unsigned short			migration_flags;
837 
838 #ifdef CONFIG_PREEMPT_RCU
839 	int				rcu_read_lock_nesting;
840 	union rcu_special		rcu_read_unlock_special;
841 	struct list_head		rcu_node_entry;
842 	struct rcu_node			*rcu_blocked_node;
843 #endif /* #ifdef CONFIG_PREEMPT_RCU */
844 
845 #ifdef CONFIG_TASKS_RCU
846 	unsigned long			rcu_tasks_nvcsw;
847 	u8				rcu_tasks_holdout;
848 	u8				rcu_tasks_idx;
849 	int				rcu_tasks_idle_cpu;
850 	struct list_head		rcu_tasks_holdout_list;
851 #endif /* #ifdef CONFIG_TASKS_RCU */
852 
853 #ifdef CONFIG_TASKS_TRACE_RCU
854 	int				trc_reader_nesting;
855 	int				trc_ipi_to_cpu;
856 	union rcu_special		trc_reader_special;
857 	struct list_head		trc_holdout_list;
858 	struct list_head		trc_blkd_node;
859 	int				trc_blkd_cpu;
860 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
861 
862 	struct sched_info		sched_info;
863 
864 	struct list_head		tasks;
865 #ifdef CONFIG_SMP
866 	struct plist_node		pushable_tasks;
867 	struct rb_node			pushable_dl_tasks;
868 #endif
869 
870 	struct mm_struct		*mm;
871 	struct mm_struct		*active_mm;
872 
873 	/* Per-thread vma caching: */
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_LRU_GEN
927 	/* whether the LRU algorithm may apply to this access */
928 	unsigned			in_lru_fault:1;
929 #endif
930 #ifdef CONFIG_COMPAT_BRK
931 	unsigned			brk_randomized:1;
932 #endif
933 #ifdef CONFIG_CGROUPS
934 	/* disallow userland-initiated cgroup migration */
935 	unsigned			no_cgroup_migration:1;
936 	/* task is frozen/stopped (used by the cgroup freezer) */
937 	unsigned			frozen:1;
938 #endif
939 #ifdef CONFIG_BLK_CGROUP
940 	unsigned			use_memdelay:1;
941 #endif
942 #ifdef CONFIG_PSI
943 	/* Stalled due to lack of memory */
944 	unsigned			in_memstall:1;
945 #endif
946 #ifdef CONFIG_PAGE_OWNER
947 	/* Used by page_owner=on to detect recursion in page tracking. */
948 	unsigned			in_page_owner:1;
949 #endif
950 #ifdef CONFIG_EVENTFD
951 	/* Recursion prevention for eventfd_signal() */
952 	unsigned			in_eventfd:1;
953 #endif
954 #ifdef CONFIG_IOMMU_SVA
955 	unsigned			pasid_activated:1;
956 #endif
957 #ifdef	CONFIG_CPU_SUP_INTEL
958 	unsigned			reported_split_lock:1;
959 #endif
960 #ifdef CONFIG_TASK_DELAY_ACCT
961 	/* delay due to memory thrashing */
962 	unsigned                        in_thrashing:1;
963 #endif
964 
965 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
966 
967 	struct restart_block		restart_block;
968 
969 	pid_t				pid;
970 	pid_t				tgid;
971 
972 #ifdef CONFIG_STACKPROTECTOR
973 	/* Canary value for the -fstack-protector GCC feature: */
974 	unsigned long			stack_canary;
975 #endif
976 	/*
977 	 * Pointers to the (original) parent process, youngest child, younger sibling,
978 	 * older sibling, respectively.  (p->father can be replaced with
979 	 * p->real_parent->pid)
980 	 */
981 
982 	/* Real parent process: */
983 	struct task_struct __rcu	*real_parent;
984 
985 	/* Recipient of SIGCHLD, wait4() reports: */
986 	struct task_struct __rcu	*parent;
987 
988 	/*
989 	 * Children/sibling form the list of natural children:
990 	 */
991 	struct list_head		children;
992 	struct list_head		sibling;
993 	struct task_struct		*group_leader;
994 
995 	/*
996 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
997 	 *
998 	 * This includes both natural children and PTRACE_ATTACH targets.
999 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1000 	 */
1001 	struct list_head		ptraced;
1002 	struct list_head		ptrace_entry;
1003 
1004 	/* PID/PID hash table linkage. */
1005 	struct pid			*thread_pid;
1006 	struct hlist_node		pid_links[PIDTYPE_MAX];
1007 	struct list_head		thread_group;
1008 	struct list_head		thread_node;
1009 
1010 	struct completion		*vfork_done;
1011 
1012 	/* CLONE_CHILD_SETTID: */
1013 	int __user			*set_child_tid;
1014 
1015 	/* CLONE_CHILD_CLEARTID: */
1016 	int __user			*clear_child_tid;
1017 
1018 	/* PF_KTHREAD | PF_IO_WORKER */
1019 	void				*worker_private;
1020 
1021 	u64				utime;
1022 	u64				stime;
1023 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1024 	u64				utimescaled;
1025 	u64				stimescaled;
1026 #endif
1027 	u64				gtime;
1028 	struct prev_cputime		prev_cputime;
1029 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1030 	struct vtime			vtime;
1031 #endif
1032 
1033 #ifdef CONFIG_NO_HZ_FULL
1034 	atomic_t			tick_dep_mask;
1035 #endif
1036 	/* Context switch counts: */
1037 	unsigned long			nvcsw;
1038 	unsigned long			nivcsw;
1039 
1040 	/* Monotonic time in nsecs: */
1041 	u64				start_time;
1042 
1043 	/* Boot based time in nsecs: */
1044 	u64				start_boottime;
1045 
1046 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1047 	unsigned long			min_flt;
1048 	unsigned long			maj_flt;
1049 
1050 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1051 	struct posix_cputimers		posix_cputimers;
1052 
1053 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1054 	struct posix_cputimers_work	posix_cputimers_work;
1055 #endif
1056 
1057 	/* Process credentials: */
1058 
1059 	/* Tracer's credentials at attach: */
1060 	const struct cred __rcu		*ptracer_cred;
1061 
1062 	/* Objective and real subjective task credentials (COW): */
1063 	const struct cred __rcu		*real_cred;
1064 
1065 	/* Effective (overridable) subjective task credentials (COW): */
1066 	const struct cred __rcu		*cred;
1067 
1068 #ifdef CONFIG_KEYS
1069 	/* Cached requested key. */
1070 	struct key			*cached_requested_key;
1071 #endif
1072 
1073 	/*
1074 	 * executable name, excluding path.
1075 	 *
1076 	 * - normally initialized setup_new_exec()
1077 	 * - access it with [gs]et_task_comm()
1078 	 * - lock it with task_lock()
1079 	 */
1080 	char				comm[TASK_COMM_LEN];
1081 
1082 	struct nameidata		*nameidata;
1083 
1084 #ifdef CONFIG_SYSVIPC
1085 	struct sysv_sem			sysvsem;
1086 	struct sysv_shm			sysvshm;
1087 #endif
1088 #ifdef CONFIG_DETECT_HUNG_TASK
1089 	unsigned long			last_switch_count;
1090 	unsigned long			last_switch_time;
1091 #endif
1092 	/* Filesystem information: */
1093 	struct fs_struct		*fs;
1094 
1095 	/* Open file information: */
1096 	struct files_struct		*files;
1097 
1098 #ifdef CONFIG_IO_URING
1099 	struct io_uring_task		*io_uring;
1100 #endif
1101 
1102 	/* Namespaces: */
1103 	struct nsproxy			*nsproxy;
1104 
1105 	/* Signal handlers: */
1106 	struct signal_struct		*signal;
1107 	struct sighand_struct __rcu		*sighand;
1108 	sigset_t			blocked;
1109 	sigset_t			real_blocked;
1110 	/* Restored if set_restore_sigmask() was used: */
1111 	sigset_t			saved_sigmask;
1112 	struct sigpending		pending;
1113 	unsigned long			sas_ss_sp;
1114 	size_t				sas_ss_size;
1115 	unsigned int			sas_ss_flags;
1116 
1117 	struct callback_head		*task_works;
1118 
1119 #ifdef CONFIG_AUDIT
1120 #ifdef CONFIG_AUDITSYSCALL
1121 	struct audit_context		*audit_context;
1122 #endif
1123 	kuid_t				loginuid;
1124 	unsigned int			sessionid;
1125 #endif
1126 	struct seccomp			seccomp;
1127 	struct syscall_user_dispatch	syscall_dispatch;
1128 
1129 	/* Thread group tracking: */
1130 	u64				parent_exec_id;
1131 	u64				self_exec_id;
1132 
1133 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1134 	spinlock_t			alloc_lock;
1135 
1136 	/* Protection of the PI data structures: */
1137 	raw_spinlock_t			pi_lock;
1138 
1139 	struct wake_q_node		wake_q;
1140 
1141 #ifdef CONFIG_RT_MUTEXES
1142 	/* PI waiters blocked on a rt_mutex held by this task: */
1143 	struct rb_root_cached		pi_waiters;
1144 	/* Updated under owner's pi_lock and rq lock */
1145 	struct task_struct		*pi_top_task;
1146 	/* Deadlock detection and priority inheritance handling: */
1147 	struct rt_mutex_waiter		*pi_blocked_on;
1148 #endif
1149 
1150 #ifdef CONFIG_DEBUG_MUTEXES
1151 	/* Mutex deadlock detection: */
1152 	struct mutex_waiter		*blocked_on;
1153 #endif
1154 
1155 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1156 	int				non_block_count;
1157 #endif
1158 
1159 #ifdef CONFIG_TRACE_IRQFLAGS
1160 	struct irqtrace_events		irqtrace;
1161 	unsigned int			hardirq_threaded;
1162 	u64				hardirq_chain_key;
1163 	int				softirqs_enabled;
1164 	int				softirq_context;
1165 	int				irq_config;
1166 #endif
1167 #ifdef CONFIG_PREEMPT_RT
1168 	int				softirq_disable_cnt;
1169 #endif
1170 
1171 #ifdef CONFIG_LOCKDEP
1172 # define MAX_LOCK_DEPTH			48UL
1173 	u64				curr_chain_key;
1174 	int				lockdep_depth;
1175 	unsigned int			lockdep_recursion;
1176 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1177 #endif
1178 
1179 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1180 	unsigned int			in_ubsan;
1181 #endif
1182 
1183 	/* Journalling filesystem info: */
1184 	void				*journal_info;
1185 
1186 	/* Stacked block device info: */
1187 	struct bio_list			*bio_list;
1188 
1189 	/* Stack plugging: */
1190 	struct blk_plug			*plug;
1191 
1192 	/* VM state: */
1193 	struct reclaim_state		*reclaim_state;
1194 
1195 	struct backing_dev_info		*backing_dev_info;
1196 
1197 	struct io_context		*io_context;
1198 
1199 #ifdef CONFIG_COMPACTION
1200 	struct capture_control		*capture_control;
1201 #endif
1202 	/* Ptrace state: */
1203 	unsigned long			ptrace_message;
1204 	kernel_siginfo_t		*last_siginfo;
1205 
1206 	struct task_io_accounting	ioac;
1207 #ifdef CONFIG_PSI
1208 	/* Pressure stall state */
1209 	unsigned int			psi_flags;
1210 #endif
1211 #ifdef CONFIG_TASK_XACCT
1212 	/* Accumulated RSS usage: */
1213 	u64				acct_rss_mem1;
1214 	/* Accumulated virtual memory usage: */
1215 	u64				acct_vm_mem1;
1216 	/* stime + utime since last update: */
1217 	u64				acct_timexpd;
1218 #endif
1219 #ifdef CONFIG_CPUSETS
1220 	/* Protected by ->alloc_lock: */
1221 	nodemask_t			mems_allowed;
1222 	/* Sequence number to catch updates: */
1223 	seqcount_spinlock_t		mems_allowed_seq;
1224 	int				cpuset_mem_spread_rotor;
1225 	int				cpuset_slab_spread_rotor;
1226 #endif
1227 #ifdef CONFIG_CGROUPS
1228 	/* Control Group info protected by css_set_lock: */
1229 	struct css_set __rcu		*cgroups;
1230 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1231 	struct list_head		cg_list;
1232 #endif
1233 #ifdef CONFIG_X86_CPU_RESCTRL
1234 	u32				closid;
1235 	u32				rmid;
1236 #endif
1237 #ifdef CONFIG_FUTEX
1238 	struct robust_list_head __user	*robust_list;
1239 #ifdef CONFIG_COMPAT
1240 	struct compat_robust_list_head __user *compat_robust_list;
1241 #endif
1242 	struct list_head		pi_state_list;
1243 	struct futex_pi_state		*pi_state_cache;
1244 	struct mutex			futex_exit_mutex;
1245 	unsigned int			futex_state;
1246 #endif
1247 #ifdef CONFIG_PERF_EVENTS
1248 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1249 	struct mutex			perf_event_mutex;
1250 	struct list_head		perf_event_list;
1251 #endif
1252 #ifdef CONFIG_DEBUG_PREEMPT
1253 	unsigned long			preempt_disable_ip;
1254 #endif
1255 #ifdef CONFIG_NUMA
1256 	/* Protected by alloc_lock: */
1257 	struct mempolicy		*mempolicy;
1258 	short				il_prev;
1259 	short				pref_node_fork;
1260 #endif
1261 #ifdef CONFIG_NUMA_BALANCING
1262 	int				numa_scan_seq;
1263 	unsigned int			numa_scan_period;
1264 	unsigned int			numa_scan_period_max;
1265 	int				numa_preferred_nid;
1266 	unsigned long			numa_migrate_retry;
1267 	/* Migration stamp: */
1268 	u64				node_stamp;
1269 	u64				last_task_numa_placement;
1270 	u64				last_sum_exec_runtime;
1271 	struct callback_head		numa_work;
1272 
1273 	/*
1274 	 * This pointer is only modified for current in syscall and
1275 	 * pagefault context (and for tasks being destroyed), so it can be read
1276 	 * from any of the following contexts:
1277 	 *  - RCU read-side critical section
1278 	 *  - current->numa_group from everywhere
1279 	 *  - task's runqueue locked, task not running
1280 	 */
1281 	struct numa_group __rcu		*numa_group;
1282 
1283 	/*
1284 	 * numa_faults is an array split into four regions:
1285 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1286 	 * in this precise order.
1287 	 *
1288 	 * faults_memory: Exponential decaying average of faults on a per-node
1289 	 * basis. Scheduling placement decisions are made based on these
1290 	 * counts. The values remain static for the duration of a PTE scan.
1291 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1292 	 * hinting fault was incurred.
1293 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1294 	 * during the current scan window. When the scan completes, the counts
1295 	 * in faults_memory and faults_cpu decay and these values are copied.
1296 	 */
1297 	unsigned long			*numa_faults;
1298 	unsigned long			total_numa_faults;
1299 
1300 	/*
1301 	 * numa_faults_locality tracks if faults recorded during the last
1302 	 * scan window were remote/local or failed to migrate. The task scan
1303 	 * period is adapted based on the locality of the faults with different
1304 	 * weights depending on whether they were shared or private faults
1305 	 */
1306 	unsigned long			numa_faults_locality[3];
1307 
1308 	unsigned long			numa_pages_migrated;
1309 #endif /* CONFIG_NUMA_BALANCING */
1310 
1311 #ifdef CONFIG_RSEQ
1312 	struct rseq __user *rseq;
1313 	u32 rseq_sig;
1314 	/*
1315 	 * RmW on rseq_event_mask must be performed atomically
1316 	 * with respect to preemption.
1317 	 */
1318 	unsigned long rseq_event_mask;
1319 #endif
1320 
1321 	struct tlbflush_unmap_batch	tlb_ubc;
1322 
1323 	union {
1324 		refcount_t		rcu_users;
1325 		struct rcu_head		rcu;
1326 	};
1327 
1328 	/* Cache last used pipe for splice(): */
1329 	struct pipe_inode_info		*splice_pipe;
1330 
1331 	struct page_frag		task_frag;
1332 
1333 #ifdef CONFIG_TASK_DELAY_ACCT
1334 	struct task_delay_info		*delays;
1335 #endif
1336 
1337 #ifdef CONFIG_FAULT_INJECTION
1338 	int				make_it_fail;
1339 	unsigned int			fail_nth;
1340 #endif
1341 	/*
1342 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1343 	 * balance_dirty_pages() for a dirty throttling pause:
1344 	 */
1345 	int				nr_dirtied;
1346 	int				nr_dirtied_pause;
1347 	/* Start of a write-and-pause period: */
1348 	unsigned long			dirty_paused_when;
1349 
1350 #ifdef CONFIG_LATENCYTOP
1351 	int				latency_record_count;
1352 	struct latency_record		latency_record[LT_SAVECOUNT];
1353 #endif
1354 	/*
1355 	 * Time slack values; these are used to round up poll() and
1356 	 * select() etc timeout values. These are in nanoseconds.
1357 	 */
1358 	u64				timer_slack_ns;
1359 	u64				default_timer_slack_ns;
1360 
1361 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1362 	unsigned int			kasan_depth;
1363 #endif
1364 
1365 #ifdef CONFIG_KCSAN
1366 	struct kcsan_ctx		kcsan_ctx;
1367 #ifdef CONFIG_TRACE_IRQFLAGS
1368 	struct irqtrace_events		kcsan_save_irqtrace;
1369 #endif
1370 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1371 	int				kcsan_stack_depth;
1372 #endif
1373 #endif
1374 
1375 #ifdef CONFIG_KMSAN
1376 	struct kmsan_ctx		kmsan_ctx;
1377 #endif
1378 
1379 #if IS_ENABLED(CONFIG_KUNIT)
1380 	struct kunit			*kunit_test;
1381 #endif
1382 
1383 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1384 	/* Index of current stored address in ret_stack: */
1385 	int				curr_ret_stack;
1386 	int				curr_ret_depth;
1387 
1388 	/* Stack of return addresses for return function tracing: */
1389 	struct ftrace_ret_stack		*ret_stack;
1390 
1391 	/* Timestamp for last schedule: */
1392 	unsigned long long		ftrace_timestamp;
1393 
1394 	/*
1395 	 * Number of functions that haven't been traced
1396 	 * because of depth overrun:
1397 	 */
1398 	atomic_t			trace_overrun;
1399 
1400 	/* Pause tracing: */
1401 	atomic_t			tracing_graph_pause;
1402 #endif
1403 
1404 #ifdef CONFIG_TRACING
1405 	/* Bitmask and counter of trace recursion: */
1406 	unsigned long			trace_recursion;
1407 #endif /* CONFIG_TRACING */
1408 
1409 #ifdef CONFIG_KCOV
1410 	/* See kernel/kcov.c for more details. */
1411 
1412 	/* Coverage collection mode enabled for this task (0 if disabled): */
1413 	unsigned int			kcov_mode;
1414 
1415 	/* Size of the kcov_area: */
1416 	unsigned int			kcov_size;
1417 
1418 	/* Buffer for coverage collection: */
1419 	void				*kcov_area;
1420 
1421 	/* KCOV descriptor wired with this task or NULL: */
1422 	struct kcov			*kcov;
1423 
1424 	/* KCOV common handle for remote coverage collection: */
1425 	u64				kcov_handle;
1426 
1427 	/* KCOV sequence number: */
1428 	int				kcov_sequence;
1429 
1430 	/* Collect coverage from softirq context: */
1431 	unsigned int			kcov_softirq;
1432 #endif
1433 
1434 #ifdef CONFIG_MEMCG
1435 	struct mem_cgroup		*memcg_in_oom;
1436 	gfp_t				memcg_oom_gfp_mask;
1437 	int				memcg_oom_order;
1438 
1439 	/* Number of pages to reclaim on returning to userland: */
1440 	unsigned int			memcg_nr_pages_over_high;
1441 
1442 	/* Used by memcontrol for targeted memcg charge: */
1443 	struct mem_cgroup		*active_memcg;
1444 #endif
1445 
1446 #ifdef CONFIG_BLK_CGROUP
1447 	struct request_queue		*throttle_queue;
1448 #endif
1449 
1450 #ifdef CONFIG_UPROBES
1451 	struct uprobe_task		*utask;
1452 #endif
1453 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1454 	unsigned int			sequential_io;
1455 	unsigned int			sequential_io_avg;
1456 #endif
1457 	struct kmap_ctrl		kmap_ctrl;
1458 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1459 	unsigned long			task_state_change;
1460 # ifdef CONFIG_PREEMPT_RT
1461 	unsigned long			saved_state_change;
1462 # endif
1463 #endif
1464 	int				pagefault_disabled;
1465 #ifdef CONFIG_MMU
1466 	struct task_struct		*oom_reaper_list;
1467 	struct timer_list		oom_reaper_timer;
1468 #endif
1469 #ifdef CONFIG_VMAP_STACK
1470 	struct vm_struct		*stack_vm_area;
1471 #endif
1472 #ifdef CONFIG_THREAD_INFO_IN_TASK
1473 	/* A live task holds one reference: */
1474 	refcount_t			stack_refcount;
1475 #endif
1476 #ifdef CONFIG_LIVEPATCH
1477 	int patch_state;
1478 #endif
1479 #ifdef CONFIG_SECURITY
1480 	/* Used by LSM modules for access restriction: */
1481 	void				*security;
1482 #endif
1483 #ifdef CONFIG_BPF_SYSCALL
1484 	/* Used by BPF task local storage */
1485 	struct bpf_local_storage __rcu	*bpf_storage;
1486 	/* Used for BPF run context */
1487 	struct bpf_run_ctx		*bpf_ctx;
1488 #endif
1489 
1490 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1491 	unsigned long			lowest_stack;
1492 	unsigned long			prev_lowest_stack;
1493 #endif
1494 
1495 #ifdef CONFIG_X86_MCE
1496 	void __user			*mce_vaddr;
1497 	__u64				mce_kflags;
1498 	u64				mce_addr;
1499 	__u64				mce_ripv : 1,
1500 					mce_whole_page : 1,
1501 					__mce_reserved : 62;
1502 	struct callback_head		mce_kill_me;
1503 	int				mce_count;
1504 #endif
1505 
1506 #ifdef CONFIG_KRETPROBES
1507 	struct llist_head               kretprobe_instances;
1508 #endif
1509 #ifdef CONFIG_RETHOOK
1510 	struct llist_head               rethooks;
1511 #endif
1512 
1513 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1514 	/*
1515 	 * If L1D flush is supported on mm context switch
1516 	 * then we use this callback head to queue kill work
1517 	 * to kill tasks that are not running on SMT disabled
1518 	 * cores
1519 	 */
1520 	struct callback_head		l1d_flush_kill;
1521 #endif
1522 
1523 #ifdef CONFIG_RV
1524 	/*
1525 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1526 	 * If we find justification for more monitors, we can think
1527 	 * about adding more or developing a dynamic method. So far,
1528 	 * none of these are justified.
1529 	 */
1530 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1531 #endif
1532 
1533 	/*
1534 	 * New fields for task_struct should be added above here, so that
1535 	 * they are included in the randomized portion of task_struct.
1536 	 */
1537 	randomized_struct_fields_end
1538 
1539 	/* CPU-specific state of this task: */
1540 	struct thread_struct		thread;
1541 
1542 	/*
1543 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1544 	 * structure.  It *MUST* be at the end of 'task_struct'.
1545 	 *
1546 	 * Do not put anything below here!
1547 	 */
1548 };
1549 
1550 static inline struct pid *task_pid(struct task_struct *task)
1551 {
1552 	return task->thread_pid;
1553 }
1554 
1555 /*
1556  * the helpers to get the task's different pids as they are seen
1557  * from various namespaces
1558  *
1559  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1560  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1561  *                     current.
1562  * task_xid_nr_ns()  : id seen from the ns specified;
1563  *
1564  * see also pid_nr() etc in include/linux/pid.h
1565  */
1566 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1567 
1568 static inline pid_t task_pid_nr(struct task_struct *tsk)
1569 {
1570 	return tsk->pid;
1571 }
1572 
1573 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1574 {
1575 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1576 }
1577 
1578 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1579 {
1580 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1581 }
1582 
1583 
1584 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1585 {
1586 	return tsk->tgid;
1587 }
1588 
1589 /**
1590  * pid_alive - check that a task structure is not stale
1591  * @p: Task structure to be checked.
1592  *
1593  * Test if a process is not yet dead (at most zombie state)
1594  * If pid_alive fails, then pointers within the task structure
1595  * can be stale and must not be dereferenced.
1596  *
1597  * Return: 1 if the process is alive. 0 otherwise.
1598  */
1599 static inline int pid_alive(const struct task_struct *p)
1600 {
1601 	return p->thread_pid != NULL;
1602 }
1603 
1604 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1605 {
1606 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1607 }
1608 
1609 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1610 {
1611 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1612 }
1613 
1614 
1615 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1616 {
1617 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1618 }
1619 
1620 static inline pid_t task_session_vnr(struct task_struct *tsk)
1621 {
1622 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1623 }
1624 
1625 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1626 {
1627 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1628 }
1629 
1630 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1631 {
1632 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1633 }
1634 
1635 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1636 {
1637 	pid_t pid = 0;
1638 
1639 	rcu_read_lock();
1640 	if (pid_alive(tsk))
1641 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1642 	rcu_read_unlock();
1643 
1644 	return pid;
1645 }
1646 
1647 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1648 {
1649 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1650 }
1651 
1652 /* Obsolete, do not use: */
1653 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1654 {
1655 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1656 }
1657 
1658 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1659 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1660 
1661 static inline unsigned int __task_state_index(unsigned int tsk_state,
1662 					      unsigned int tsk_exit_state)
1663 {
1664 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1665 
1666 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1667 
1668 	if (tsk_state == TASK_IDLE)
1669 		state = TASK_REPORT_IDLE;
1670 
1671 	/*
1672 	 * We're lying here, but rather than expose a completely new task state
1673 	 * to userspace, we can make this appear as if the task has gone through
1674 	 * a regular rt_mutex_lock() call.
1675 	 */
1676 	if (tsk_state == TASK_RTLOCK_WAIT)
1677 		state = TASK_UNINTERRUPTIBLE;
1678 
1679 	return fls(state);
1680 }
1681 
1682 static inline unsigned int task_state_index(struct task_struct *tsk)
1683 {
1684 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1685 }
1686 
1687 static inline char task_index_to_char(unsigned int state)
1688 {
1689 	static const char state_char[] = "RSDTtXZPI";
1690 
1691 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1692 
1693 	return state_char[state];
1694 }
1695 
1696 static inline char task_state_to_char(struct task_struct *tsk)
1697 {
1698 	return task_index_to_char(task_state_index(tsk));
1699 }
1700 
1701 /**
1702  * is_global_init - check if a task structure is init. Since init
1703  * is free to have sub-threads we need to check tgid.
1704  * @tsk: Task structure to be checked.
1705  *
1706  * Check if a task structure is the first user space task the kernel created.
1707  *
1708  * Return: 1 if the task structure is init. 0 otherwise.
1709  */
1710 static inline int is_global_init(struct task_struct *tsk)
1711 {
1712 	return task_tgid_nr(tsk) == 1;
1713 }
1714 
1715 extern struct pid *cad_pid;
1716 
1717 /*
1718  * Per process flags
1719  */
1720 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1721 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1722 #define PF_EXITING		0x00000004	/* Getting shut down */
1723 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1724 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1725 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1726 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1727 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1728 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1729 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1730 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1731 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1732 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1733 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1734 #define PF__HOLE__00004000	0x00004000
1735 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1736 #define PF__HOLE__00010000	0x00010000
1737 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1738 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1739 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1740 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1741 						 * I am cleaning dirty pages from some other bdi. */
1742 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1743 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1744 #define PF__HOLE__00800000	0x00800000
1745 #define PF__HOLE__01000000	0x01000000
1746 #define PF__HOLE__02000000	0x02000000
1747 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1748 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1749 #define PF_MEMALLOC_PIN		0x10000000	/* Allocation context constrained to zones which allow long term pinning. */
1750 #define PF__HOLE__20000000	0x20000000
1751 #define PF__HOLE__40000000	0x40000000
1752 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1753 
1754 /*
1755  * Only the _current_ task can read/write to tsk->flags, but other
1756  * tasks can access tsk->flags in readonly mode for example
1757  * with tsk_used_math (like during threaded core dumping).
1758  * There is however an exception to this rule during ptrace
1759  * or during fork: the ptracer task is allowed to write to the
1760  * child->flags of its traced child (same goes for fork, the parent
1761  * can write to the child->flags), because we're guaranteed the
1762  * child is not running and in turn not changing child->flags
1763  * at the same time the parent does it.
1764  */
1765 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1766 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1767 #define clear_used_math()			clear_stopped_child_used_math(current)
1768 #define set_used_math()				set_stopped_child_used_math(current)
1769 
1770 #define conditional_stopped_child_used_math(condition, child) \
1771 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1772 
1773 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1774 
1775 #define copy_to_stopped_child_used_math(child) \
1776 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1777 
1778 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1779 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1780 #define used_math()				tsk_used_math(current)
1781 
1782 static __always_inline bool is_percpu_thread(void)
1783 {
1784 #ifdef CONFIG_SMP
1785 	return (current->flags & PF_NO_SETAFFINITY) &&
1786 		(current->nr_cpus_allowed  == 1);
1787 #else
1788 	return true;
1789 #endif
1790 }
1791 
1792 /* Per-process atomic flags. */
1793 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1794 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1795 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1796 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1797 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1798 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1799 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1800 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1801 
1802 #define TASK_PFA_TEST(name, func)					\
1803 	static inline bool task_##func(struct task_struct *p)		\
1804 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1805 
1806 #define TASK_PFA_SET(name, func)					\
1807 	static inline void task_set_##func(struct task_struct *p)	\
1808 	{ set_bit(PFA_##name, &p->atomic_flags); }
1809 
1810 #define TASK_PFA_CLEAR(name, func)					\
1811 	static inline void task_clear_##func(struct task_struct *p)	\
1812 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1813 
1814 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1815 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1816 
1817 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1818 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1819 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1820 
1821 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1822 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1823 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1824 
1825 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1826 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1827 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1828 
1829 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1830 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1831 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1832 
1833 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1834 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1835 
1836 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1837 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1838 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1839 
1840 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1841 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1842 
1843 static inline void
1844 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1845 {
1846 	current->flags &= ~flags;
1847 	current->flags |= orig_flags & flags;
1848 }
1849 
1850 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1851 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effective_cpus);
1852 #ifdef CONFIG_SMP
1853 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1854 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1855 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1856 extern void release_user_cpus_ptr(struct task_struct *p);
1857 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1858 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1859 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1860 #else
1861 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1862 {
1863 }
1864 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1865 {
1866 	if (!cpumask_test_cpu(0, new_mask))
1867 		return -EINVAL;
1868 	return 0;
1869 }
1870 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1871 {
1872 	if (src->user_cpus_ptr)
1873 		return -EINVAL;
1874 	return 0;
1875 }
1876 static inline void release_user_cpus_ptr(struct task_struct *p)
1877 {
1878 	WARN_ON(p->user_cpus_ptr);
1879 }
1880 
1881 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1882 {
1883 	return 0;
1884 }
1885 #endif
1886 
1887 extern int yield_to(struct task_struct *p, bool preempt);
1888 extern void set_user_nice(struct task_struct *p, long nice);
1889 extern int task_prio(const struct task_struct *p);
1890 
1891 /**
1892  * task_nice - return the nice value of a given task.
1893  * @p: the task in question.
1894  *
1895  * Return: The nice value [ -20 ... 0 ... 19 ].
1896  */
1897 static inline int task_nice(const struct task_struct *p)
1898 {
1899 	return PRIO_TO_NICE((p)->static_prio);
1900 }
1901 
1902 extern int can_nice(const struct task_struct *p, const int nice);
1903 extern int task_curr(const struct task_struct *p);
1904 extern int idle_cpu(int cpu);
1905 extern int available_idle_cpu(int cpu);
1906 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1907 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1908 extern void sched_set_fifo(struct task_struct *p);
1909 extern void sched_set_fifo_low(struct task_struct *p);
1910 extern void sched_set_normal(struct task_struct *p, int nice);
1911 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1912 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1913 extern struct task_struct *idle_task(int cpu);
1914 
1915 /**
1916  * is_idle_task - is the specified task an idle task?
1917  * @p: the task in question.
1918  *
1919  * Return: 1 if @p is an idle task. 0 otherwise.
1920  */
1921 static __always_inline bool is_idle_task(const struct task_struct *p)
1922 {
1923 	return !!(p->flags & PF_IDLE);
1924 }
1925 
1926 extern struct task_struct *curr_task(int cpu);
1927 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1928 
1929 void yield(void);
1930 
1931 union thread_union {
1932 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1933 	struct task_struct task;
1934 #endif
1935 #ifndef CONFIG_THREAD_INFO_IN_TASK
1936 	struct thread_info thread_info;
1937 #endif
1938 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1939 };
1940 
1941 #ifndef CONFIG_THREAD_INFO_IN_TASK
1942 extern struct thread_info init_thread_info;
1943 #endif
1944 
1945 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1946 
1947 #ifdef CONFIG_THREAD_INFO_IN_TASK
1948 # define task_thread_info(task)	(&(task)->thread_info)
1949 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1950 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1951 #endif
1952 
1953 /*
1954  * find a task by one of its numerical ids
1955  *
1956  * find_task_by_pid_ns():
1957  *      finds a task by its pid in the specified namespace
1958  * find_task_by_vpid():
1959  *      finds a task by its virtual pid
1960  *
1961  * see also find_vpid() etc in include/linux/pid.h
1962  */
1963 
1964 extern struct task_struct *find_task_by_vpid(pid_t nr);
1965 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1966 
1967 /*
1968  * find a task by its virtual pid and get the task struct
1969  */
1970 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1971 
1972 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1973 extern int wake_up_process(struct task_struct *tsk);
1974 extern void wake_up_new_task(struct task_struct *tsk);
1975 
1976 #ifdef CONFIG_SMP
1977 extern void kick_process(struct task_struct *tsk);
1978 #else
1979 static inline void kick_process(struct task_struct *tsk) { }
1980 #endif
1981 
1982 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1983 
1984 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1985 {
1986 	__set_task_comm(tsk, from, false);
1987 }
1988 
1989 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1990 #define get_task_comm(buf, tsk) ({			\
1991 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1992 	__get_task_comm(buf, sizeof(buf), tsk);		\
1993 })
1994 
1995 #ifdef CONFIG_SMP
1996 static __always_inline void scheduler_ipi(void)
1997 {
1998 	/*
1999 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2000 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
2001 	 * this IPI.
2002 	 */
2003 	preempt_fold_need_resched();
2004 }
2005 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2006 #else
2007 static inline void scheduler_ipi(void) { }
2008 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
2009 {
2010 	return 1;
2011 }
2012 #endif
2013 
2014 /*
2015  * Set thread flags in other task's structures.
2016  * See asm/thread_info.h for TIF_xxxx flags available:
2017  */
2018 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2019 {
2020 	set_ti_thread_flag(task_thread_info(tsk), flag);
2021 }
2022 
2023 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2024 {
2025 	clear_ti_thread_flag(task_thread_info(tsk), flag);
2026 }
2027 
2028 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2029 					  bool value)
2030 {
2031 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
2032 }
2033 
2034 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2035 {
2036 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2037 }
2038 
2039 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2040 {
2041 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2042 }
2043 
2044 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2045 {
2046 	return test_ti_thread_flag(task_thread_info(tsk), flag);
2047 }
2048 
2049 static inline void set_tsk_need_resched(struct task_struct *tsk)
2050 {
2051 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2052 }
2053 
2054 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2055 {
2056 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2057 }
2058 
2059 static inline int test_tsk_need_resched(struct task_struct *tsk)
2060 {
2061 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2062 }
2063 
2064 /*
2065  * cond_resched() and cond_resched_lock(): latency reduction via
2066  * explicit rescheduling in places that are safe. The return
2067  * value indicates whether a reschedule was done in fact.
2068  * cond_resched_lock() will drop the spinlock before scheduling,
2069  */
2070 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2071 extern int __cond_resched(void);
2072 
2073 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2074 
2075 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2076 
2077 static __always_inline int _cond_resched(void)
2078 {
2079 	return static_call_mod(cond_resched)();
2080 }
2081 
2082 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2083 extern int dynamic_cond_resched(void);
2084 
2085 static __always_inline int _cond_resched(void)
2086 {
2087 	return dynamic_cond_resched();
2088 }
2089 
2090 #else
2091 
2092 static inline int _cond_resched(void)
2093 {
2094 	return __cond_resched();
2095 }
2096 
2097 #endif /* CONFIG_PREEMPT_DYNAMIC */
2098 
2099 #else
2100 
2101 static inline int _cond_resched(void) { return 0; }
2102 
2103 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2104 
2105 #define cond_resched() ({			\
2106 	__might_resched(__FILE__, __LINE__, 0);	\
2107 	_cond_resched();			\
2108 })
2109 
2110 extern int __cond_resched_lock(spinlock_t *lock);
2111 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2112 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2113 
2114 #define MIGHT_RESCHED_RCU_SHIFT		8
2115 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2116 
2117 #ifndef CONFIG_PREEMPT_RT
2118 /*
2119  * Non RT kernels have an elevated preempt count due to the held lock,
2120  * but are not allowed to be inside a RCU read side critical section
2121  */
2122 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2123 #else
2124 /*
2125  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2126  * cond_resched*lock() has to take that into account because it checks for
2127  * preempt_count() and rcu_preempt_depth().
2128  */
2129 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2130 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2131 #endif
2132 
2133 #define cond_resched_lock(lock) ({						\
2134 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2135 	__cond_resched_lock(lock);						\
2136 })
2137 
2138 #define cond_resched_rwlock_read(lock) ({					\
2139 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2140 	__cond_resched_rwlock_read(lock);					\
2141 })
2142 
2143 #define cond_resched_rwlock_write(lock) ({					\
2144 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2145 	__cond_resched_rwlock_write(lock);					\
2146 })
2147 
2148 static inline void cond_resched_rcu(void)
2149 {
2150 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2151 	rcu_read_unlock();
2152 	cond_resched();
2153 	rcu_read_lock();
2154 #endif
2155 }
2156 
2157 #ifdef CONFIG_PREEMPT_DYNAMIC
2158 
2159 extern bool preempt_model_none(void);
2160 extern bool preempt_model_voluntary(void);
2161 extern bool preempt_model_full(void);
2162 
2163 #else
2164 
2165 static inline bool preempt_model_none(void)
2166 {
2167 	return IS_ENABLED(CONFIG_PREEMPT_NONE);
2168 }
2169 static inline bool preempt_model_voluntary(void)
2170 {
2171 	return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2172 }
2173 static inline bool preempt_model_full(void)
2174 {
2175 	return IS_ENABLED(CONFIG_PREEMPT);
2176 }
2177 
2178 #endif
2179 
2180 static inline bool preempt_model_rt(void)
2181 {
2182 	return IS_ENABLED(CONFIG_PREEMPT_RT);
2183 }
2184 
2185 /*
2186  * Does the preemption model allow non-cooperative preemption?
2187  *
2188  * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2189  * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2190  * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2191  * PREEMPT_NONE model.
2192  */
2193 static inline bool preempt_model_preemptible(void)
2194 {
2195 	return preempt_model_full() || preempt_model_rt();
2196 }
2197 
2198 /*
2199  * Does a critical section need to be broken due to another
2200  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2201  * but a general need for low latency)
2202  */
2203 static inline int spin_needbreak(spinlock_t *lock)
2204 {
2205 #ifdef CONFIG_PREEMPTION
2206 	return spin_is_contended(lock);
2207 #else
2208 	return 0;
2209 #endif
2210 }
2211 
2212 /*
2213  * Check if a rwlock is contended.
2214  * Returns non-zero if there is another task waiting on the rwlock.
2215  * Returns zero if the lock is not contended or the system / underlying
2216  * rwlock implementation does not support contention detection.
2217  * Technically does not depend on CONFIG_PREEMPTION, but a general need
2218  * for low latency.
2219  */
2220 static inline int rwlock_needbreak(rwlock_t *lock)
2221 {
2222 #ifdef CONFIG_PREEMPTION
2223 	return rwlock_is_contended(lock);
2224 #else
2225 	return 0;
2226 #endif
2227 }
2228 
2229 static __always_inline bool need_resched(void)
2230 {
2231 	return unlikely(tif_need_resched());
2232 }
2233 
2234 /*
2235  * Wrappers for p->thread_info->cpu access. No-op on UP.
2236  */
2237 #ifdef CONFIG_SMP
2238 
2239 static inline unsigned int task_cpu(const struct task_struct *p)
2240 {
2241 	return READ_ONCE(task_thread_info(p)->cpu);
2242 }
2243 
2244 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2245 
2246 #else
2247 
2248 static inline unsigned int task_cpu(const struct task_struct *p)
2249 {
2250 	return 0;
2251 }
2252 
2253 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2254 {
2255 }
2256 
2257 #endif /* CONFIG_SMP */
2258 
2259 extern bool sched_task_on_rq(struct task_struct *p);
2260 extern unsigned long get_wchan(struct task_struct *p);
2261 extern struct task_struct *cpu_curr_snapshot(int cpu);
2262 
2263 /*
2264  * In order to reduce various lock holder preemption latencies provide an
2265  * interface to see if a vCPU is currently running or not.
2266  *
2267  * This allows us to terminate optimistic spin loops and block, analogous to
2268  * the native optimistic spin heuristic of testing if the lock owner task is
2269  * running or not.
2270  */
2271 #ifndef vcpu_is_preempted
2272 static inline bool vcpu_is_preempted(int cpu)
2273 {
2274 	return false;
2275 }
2276 #endif
2277 
2278 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2279 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2280 
2281 #ifndef TASK_SIZE_OF
2282 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2283 #endif
2284 
2285 #ifdef CONFIG_SMP
2286 static inline bool owner_on_cpu(struct task_struct *owner)
2287 {
2288 	/*
2289 	 * As lock holder preemption issue, we both skip spinning if
2290 	 * task is not on cpu or its cpu is preempted
2291 	 */
2292 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2293 }
2294 
2295 /* Returns effective CPU energy utilization, as seen by the scheduler */
2296 unsigned long sched_cpu_util(int cpu);
2297 #endif /* CONFIG_SMP */
2298 
2299 #ifdef CONFIG_RSEQ
2300 
2301 /*
2302  * Map the event mask on the user-space ABI enum rseq_cs_flags
2303  * for direct mask checks.
2304  */
2305 enum rseq_event_mask_bits {
2306 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2307 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2308 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2309 };
2310 
2311 enum rseq_event_mask {
2312 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
2313 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
2314 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
2315 };
2316 
2317 static inline void rseq_set_notify_resume(struct task_struct *t)
2318 {
2319 	if (t->rseq)
2320 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2321 }
2322 
2323 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2324 
2325 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2326 					     struct pt_regs *regs)
2327 {
2328 	if (current->rseq)
2329 		__rseq_handle_notify_resume(ksig, regs);
2330 }
2331 
2332 static inline void rseq_signal_deliver(struct ksignal *ksig,
2333 				       struct pt_regs *regs)
2334 {
2335 	preempt_disable();
2336 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2337 	preempt_enable();
2338 	rseq_handle_notify_resume(ksig, regs);
2339 }
2340 
2341 /* rseq_preempt() requires preemption to be disabled. */
2342 static inline void rseq_preempt(struct task_struct *t)
2343 {
2344 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2345 	rseq_set_notify_resume(t);
2346 }
2347 
2348 /* rseq_migrate() requires preemption to be disabled. */
2349 static inline void rseq_migrate(struct task_struct *t)
2350 {
2351 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2352 	rseq_set_notify_resume(t);
2353 }
2354 
2355 /*
2356  * If parent process has a registered restartable sequences area, the
2357  * child inherits. Unregister rseq for a clone with CLONE_VM set.
2358  */
2359 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2360 {
2361 	if (clone_flags & CLONE_VM) {
2362 		t->rseq = NULL;
2363 		t->rseq_sig = 0;
2364 		t->rseq_event_mask = 0;
2365 	} else {
2366 		t->rseq = current->rseq;
2367 		t->rseq_sig = current->rseq_sig;
2368 		t->rseq_event_mask = current->rseq_event_mask;
2369 	}
2370 }
2371 
2372 static inline void rseq_execve(struct task_struct *t)
2373 {
2374 	t->rseq = NULL;
2375 	t->rseq_sig = 0;
2376 	t->rseq_event_mask = 0;
2377 }
2378 
2379 #else
2380 
2381 static inline void rseq_set_notify_resume(struct task_struct *t)
2382 {
2383 }
2384 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2385 					     struct pt_regs *regs)
2386 {
2387 }
2388 static inline void rseq_signal_deliver(struct ksignal *ksig,
2389 				       struct pt_regs *regs)
2390 {
2391 }
2392 static inline void rseq_preempt(struct task_struct *t)
2393 {
2394 }
2395 static inline void rseq_migrate(struct task_struct *t)
2396 {
2397 }
2398 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2399 {
2400 }
2401 static inline void rseq_execve(struct task_struct *t)
2402 {
2403 }
2404 
2405 #endif
2406 
2407 #ifdef CONFIG_DEBUG_RSEQ
2408 
2409 void rseq_syscall(struct pt_regs *regs);
2410 
2411 #else
2412 
2413 static inline void rseq_syscall(struct pt_regs *regs)
2414 {
2415 }
2416 
2417 #endif
2418 
2419 #ifdef CONFIG_SCHED_CORE
2420 extern void sched_core_free(struct task_struct *tsk);
2421 extern void sched_core_fork(struct task_struct *p);
2422 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2423 				unsigned long uaddr);
2424 #else
2425 static inline void sched_core_free(struct task_struct *tsk) { }
2426 static inline void sched_core_fork(struct task_struct *p) { }
2427 #endif
2428 
2429 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2430 
2431 #endif
2432