xref: /linux/include/linux/sched.h (revision e9f0878c4b2004ac19581274c1ae4c61ae3ca70e)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4 
5 /*
6  * Define 'struct task_struct' and provide the main scheduler
7  * APIs (schedule(), wakeup variants, etc.)
8  */
9 
10 #include <uapi/linux/sched.h>
11 
12 #include <asm/current.h>
13 
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/resource.h>
25 #include <linux/latencytop.h>
26 #include <linux/sched/prio.h>
27 #include <linux/signal_types.h>
28 #include <linux/mm_types_task.h>
29 #include <linux/task_io_accounting.h>
30 #include <linux/rseq.h>
31 
32 /* task_struct member predeclarations (sorted alphabetically): */
33 struct audit_context;
34 struct backing_dev_info;
35 struct bio_list;
36 struct blk_plug;
37 struct cfs_rq;
38 struct fs_struct;
39 struct futex_pi_state;
40 struct io_context;
41 struct mempolicy;
42 struct nameidata;
43 struct nsproxy;
44 struct perf_event_context;
45 struct pid_namespace;
46 struct pipe_inode_info;
47 struct rcu_node;
48 struct reclaim_state;
49 struct robust_list_head;
50 struct sched_attr;
51 struct sched_param;
52 struct seq_file;
53 struct sighand_struct;
54 struct signal_struct;
55 struct task_delay_info;
56 struct task_group;
57 
58 /*
59  * Task state bitmask. NOTE! These bits are also
60  * encoded in fs/proc/array.c: get_task_state().
61  *
62  * We have two separate sets of flags: task->state
63  * is about runnability, while task->exit_state are
64  * about the task exiting. Confusing, but this way
65  * modifying one set can't modify the other one by
66  * mistake.
67  */
68 
69 /* Used in tsk->state: */
70 #define TASK_RUNNING			0x0000
71 #define TASK_INTERRUPTIBLE		0x0001
72 #define TASK_UNINTERRUPTIBLE		0x0002
73 #define __TASK_STOPPED			0x0004
74 #define __TASK_TRACED			0x0008
75 /* Used in tsk->exit_state: */
76 #define EXIT_DEAD			0x0010
77 #define EXIT_ZOMBIE			0x0020
78 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
79 /* Used in tsk->state again: */
80 #define TASK_PARKED			0x0040
81 #define TASK_DEAD			0x0080
82 #define TASK_WAKEKILL			0x0100
83 #define TASK_WAKING			0x0200
84 #define TASK_NOLOAD			0x0400
85 #define TASK_NEW			0x0800
86 #define TASK_STATE_MAX			0x1000
87 
88 /* Convenience macros for the sake of set_current_state: */
89 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
90 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
91 #define TASK_TRACED			(TASK_WAKEKILL | __TASK_TRACED)
92 
93 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
94 
95 /* Convenience macros for the sake of wake_up(): */
96 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
97 
98 /* get_task_state(): */
99 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
100 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
101 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
102 					 TASK_PARKED)
103 
104 #define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
105 
106 #define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
107 
108 #define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
109 
110 #define task_contributes_to_load(task)	((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 					 (task->flags & PF_FROZEN) == 0 && \
112 					 (task->state & TASK_NOLOAD) == 0)
113 
114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
115 
116 /*
117  * Special states are those that do not use the normal wait-loop pattern. See
118  * the comment with set_special_state().
119  */
120 #define is_special_task_state(state)				\
121 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
122 
123 #define __set_current_state(state_value)			\
124 	do {							\
125 		WARN_ON_ONCE(is_special_task_state(state_value));\
126 		current->task_state_change = _THIS_IP_;		\
127 		current->state = (state_value);			\
128 	} while (0)
129 
130 #define set_current_state(state_value)				\
131 	do {							\
132 		WARN_ON_ONCE(is_special_task_state(state_value));\
133 		current->task_state_change = _THIS_IP_;		\
134 		smp_store_mb(current->state, (state_value));	\
135 	} while (0)
136 
137 #define set_special_state(state_value)					\
138 	do {								\
139 		unsigned long flags; /* may shadow */			\
140 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
141 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
142 		current->task_state_change = _THIS_IP_;			\
143 		current->state = (state_value);				\
144 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
145 	} while (0)
146 #else
147 /*
148  * set_current_state() includes a barrier so that the write of current->state
149  * is correctly serialised wrt the caller's subsequent test of whether to
150  * actually sleep:
151  *
152  *   for (;;) {
153  *	set_current_state(TASK_UNINTERRUPTIBLE);
154  *	if (!need_sleep)
155  *		break;
156  *
157  *	schedule();
158  *   }
159  *   __set_current_state(TASK_RUNNING);
160  *
161  * If the caller does not need such serialisation (because, for instance, the
162  * condition test and condition change and wakeup are under the same lock) then
163  * use __set_current_state().
164  *
165  * The above is typically ordered against the wakeup, which does:
166  *
167  *   need_sleep = false;
168  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
169  *
170  * where wake_up_state() executes a full memory barrier before accessing the
171  * task state.
172  *
173  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
174  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
175  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
176  *
177  * However, with slightly different timing the wakeup TASK_RUNNING store can
178  * also collide with the TASK_UNINTERRUPTIBLE store. Loosing that store is not
179  * a problem either because that will result in one extra go around the loop
180  * and our @cond test will save the day.
181  *
182  * Also see the comments of try_to_wake_up().
183  */
184 #define __set_current_state(state_value)				\
185 	current->state = (state_value)
186 
187 #define set_current_state(state_value)					\
188 	smp_store_mb(current->state, (state_value))
189 
190 /*
191  * set_special_state() should be used for those states when the blocking task
192  * can not use the regular condition based wait-loop. In that case we must
193  * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
194  * will not collide with our state change.
195  */
196 #define set_special_state(state_value)					\
197 	do {								\
198 		unsigned long flags; /* may shadow */			\
199 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
200 		current->state = (state_value);				\
201 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
202 	} while (0)
203 
204 #endif
205 
206 /* Task command name length: */
207 #define TASK_COMM_LEN			16
208 
209 extern void scheduler_tick(void);
210 
211 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
212 
213 extern long schedule_timeout(long timeout);
214 extern long schedule_timeout_interruptible(long timeout);
215 extern long schedule_timeout_killable(long timeout);
216 extern long schedule_timeout_uninterruptible(long timeout);
217 extern long schedule_timeout_idle(long timeout);
218 asmlinkage void schedule(void);
219 extern void schedule_preempt_disabled(void);
220 
221 extern int __must_check io_schedule_prepare(void);
222 extern void io_schedule_finish(int token);
223 extern long io_schedule_timeout(long timeout);
224 extern void io_schedule(void);
225 
226 /**
227  * struct prev_cputime - snapshot of system and user cputime
228  * @utime: time spent in user mode
229  * @stime: time spent in system mode
230  * @lock: protects the above two fields
231  *
232  * Stores previous user/system time values such that we can guarantee
233  * monotonicity.
234  */
235 struct prev_cputime {
236 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
237 	u64				utime;
238 	u64				stime;
239 	raw_spinlock_t			lock;
240 #endif
241 };
242 
243 /**
244  * struct task_cputime - collected CPU time counts
245  * @utime:		time spent in user mode, in nanoseconds
246  * @stime:		time spent in kernel mode, in nanoseconds
247  * @sum_exec_runtime:	total time spent on the CPU, in nanoseconds
248  *
249  * This structure groups together three kinds of CPU time that are tracked for
250  * threads and thread groups.  Most things considering CPU time want to group
251  * these counts together and treat all three of them in parallel.
252  */
253 struct task_cputime {
254 	u64				utime;
255 	u64				stime;
256 	unsigned long long		sum_exec_runtime;
257 };
258 
259 /* Alternate field names when used on cache expirations: */
260 #define virt_exp			utime
261 #define prof_exp			stime
262 #define sched_exp			sum_exec_runtime
263 
264 enum vtime_state {
265 	/* Task is sleeping or running in a CPU with VTIME inactive: */
266 	VTIME_INACTIVE = 0,
267 	/* Task runs in userspace in a CPU with VTIME active: */
268 	VTIME_USER,
269 	/* Task runs in kernelspace in a CPU with VTIME active: */
270 	VTIME_SYS,
271 };
272 
273 struct vtime {
274 	seqcount_t		seqcount;
275 	unsigned long long	starttime;
276 	enum vtime_state	state;
277 	u64			utime;
278 	u64			stime;
279 	u64			gtime;
280 };
281 
282 struct sched_info {
283 #ifdef CONFIG_SCHED_INFO
284 	/* Cumulative counters: */
285 
286 	/* # of times we have run on this CPU: */
287 	unsigned long			pcount;
288 
289 	/* Time spent waiting on a runqueue: */
290 	unsigned long long		run_delay;
291 
292 	/* Timestamps: */
293 
294 	/* When did we last run on a CPU? */
295 	unsigned long long		last_arrival;
296 
297 	/* When were we last queued to run? */
298 	unsigned long long		last_queued;
299 
300 #endif /* CONFIG_SCHED_INFO */
301 };
302 
303 /*
304  * Integer metrics need fixed point arithmetic, e.g., sched/fair
305  * has a few: load, load_avg, util_avg, freq, and capacity.
306  *
307  * We define a basic fixed point arithmetic range, and then formalize
308  * all these metrics based on that basic range.
309  */
310 # define SCHED_FIXEDPOINT_SHIFT		10
311 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
312 
313 struct load_weight {
314 	unsigned long			weight;
315 	u32				inv_weight;
316 };
317 
318 /**
319  * struct util_est - Estimation utilization of FAIR tasks
320  * @enqueued: instantaneous estimated utilization of a task/cpu
321  * @ewma:     the Exponential Weighted Moving Average (EWMA)
322  *            utilization of a task
323  *
324  * Support data structure to track an Exponential Weighted Moving Average
325  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
326  * average each time a task completes an activation. Sample's weight is chosen
327  * so that the EWMA will be relatively insensitive to transient changes to the
328  * task's workload.
329  *
330  * The enqueued attribute has a slightly different meaning for tasks and cpus:
331  * - task:   the task's util_avg at last task dequeue time
332  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
333  * Thus, the util_est.enqueued of a task represents the contribution on the
334  * estimated utilization of the CPU where that task is currently enqueued.
335  *
336  * Only for tasks we track a moving average of the past instantaneous
337  * estimated utilization. This allows to absorb sporadic drops in utilization
338  * of an otherwise almost periodic task.
339  */
340 struct util_est {
341 	unsigned int			enqueued;
342 	unsigned int			ewma;
343 #define UTIL_EST_WEIGHT_SHIFT		2
344 } __attribute__((__aligned__(sizeof(u64))));
345 
346 /*
347  * The load_avg/util_avg accumulates an infinite geometric series
348  * (see __update_load_avg() in kernel/sched/fair.c).
349  *
350  * [load_avg definition]
351  *
352  *   load_avg = runnable% * scale_load_down(load)
353  *
354  * where runnable% is the time ratio that a sched_entity is runnable.
355  * For cfs_rq, it is the aggregated load_avg of all runnable and
356  * blocked sched_entities.
357  *
358  * load_avg may also take frequency scaling into account:
359  *
360  *   load_avg = runnable% * scale_load_down(load) * freq%
361  *
362  * where freq% is the CPU frequency normalized to the highest frequency.
363  *
364  * [util_avg definition]
365  *
366  *   util_avg = running% * SCHED_CAPACITY_SCALE
367  *
368  * where running% is the time ratio that a sched_entity is running on
369  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
370  * and blocked sched_entities.
371  *
372  * util_avg may also factor frequency scaling and CPU capacity scaling:
373  *
374  *   util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
375  *
376  * where freq% is the same as above, and capacity% is the CPU capacity
377  * normalized to the greatest capacity (due to uarch differences, etc).
378  *
379  * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
380  * themselves are in the range of [0, 1]. To do fixed point arithmetics,
381  * we therefore scale them to as large a range as necessary. This is for
382  * example reflected by util_avg's SCHED_CAPACITY_SCALE.
383  *
384  * [Overflow issue]
385  *
386  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
387  * with the highest load (=88761), always runnable on a single cfs_rq,
388  * and should not overflow as the number already hits PID_MAX_LIMIT.
389  *
390  * For all other cases (including 32-bit kernels), struct load_weight's
391  * weight will overflow first before we do, because:
392  *
393  *    Max(load_avg) <= Max(load.weight)
394  *
395  * Then it is the load_weight's responsibility to consider overflow
396  * issues.
397  */
398 struct sched_avg {
399 	u64				last_update_time;
400 	u64				load_sum;
401 	u64				runnable_load_sum;
402 	u32				util_sum;
403 	u32				period_contrib;
404 	unsigned long			load_avg;
405 	unsigned long			runnable_load_avg;
406 	unsigned long			util_avg;
407 	struct util_est			util_est;
408 } ____cacheline_aligned;
409 
410 struct sched_statistics {
411 #ifdef CONFIG_SCHEDSTATS
412 	u64				wait_start;
413 	u64				wait_max;
414 	u64				wait_count;
415 	u64				wait_sum;
416 	u64				iowait_count;
417 	u64				iowait_sum;
418 
419 	u64				sleep_start;
420 	u64				sleep_max;
421 	s64				sum_sleep_runtime;
422 
423 	u64				block_start;
424 	u64				block_max;
425 	u64				exec_max;
426 	u64				slice_max;
427 
428 	u64				nr_migrations_cold;
429 	u64				nr_failed_migrations_affine;
430 	u64				nr_failed_migrations_running;
431 	u64				nr_failed_migrations_hot;
432 	u64				nr_forced_migrations;
433 
434 	u64				nr_wakeups;
435 	u64				nr_wakeups_sync;
436 	u64				nr_wakeups_migrate;
437 	u64				nr_wakeups_local;
438 	u64				nr_wakeups_remote;
439 	u64				nr_wakeups_affine;
440 	u64				nr_wakeups_affine_attempts;
441 	u64				nr_wakeups_passive;
442 	u64				nr_wakeups_idle;
443 #endif
444 };
445 
446 struct sched_entity {
447 	/* For load-balancing: */
448 	struct load_weight		load;
449 	unsigned long			runnable_weight;
450 	struct rb_node			run_node;
451 	struct list_head		group_node;
452 	unsigned int			on_rq;
453 
454 	u64				exec_start;
455 	u64				sum_exec_runtime;
456 	u64				vruntime;
457 	u64				prev_sum_exec_runtime;
458 
459 	u64				nr_migrations;
460 
461 	struct sched_statistics		statistics;
462 
463 #ifdef CONFIG_FAIR_GROUP_SCHED
464 	int				depth;
465 	struct sched_entity		*parent;
466 	/* rq on which this entity is (to be) queued: */
467 	struct cfs_rq			*cfs_rq;
468 	/* rq "owned" by this entity/group: */
469 	struct cfs_rq			*my_q;
470 #endif
471 
472 #ifdef CONFIG_SMP
473 	/*
474 	 * Per entity load average tracking.
475 	 *
476 	 * Put into separate cache line so it does not
477 	 * collide with read-mostly values above.
478 	 */
479 	struct sched_avg		avg;
480 #endif
481 };
482 
483 struct sched_rt_entity {
484 	struct list_head		run_list;
485 	unsigned long			timeout;
486 	unsigned long			watchdog_stamp;
487 	unsigned int			time_slice;
488 	unsigned short			on_rq;
489 	unsigned short			on_list;
490 
491 	struct sched_rt_entity		*back;
492 #ifdef CONFIG_RT_GROUP_SCHED
493 	struct sched_rt_entity		*parent;
494 	/* rq on which this entity is (to be) queued: */
495 	struct rt_rq			*rt_rq;
496 	/* rq "owned" by this entity/group: */
497 	struct rt_rq			*my_q;
498 #endif
499 } __randomize_layout;
500 
501 struct sched_dl_entity {
502 	struct rb_node			rb_node;
503 
504 	/*
505 	 * Original scheduling parameters. Copied here from sched_attr
506 	 * during sched_setattr(), they will remain the same until
507 	 * the next sched_setattr().
508 	 */
509 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
510 	u64				dl_deadline;	/* Relative deadline of each instance	*/
511 	u64				dl_period;	/* Separation of two instances (period) */
512 	u64				dl_bw;		/* dl_runtime / dl_period		*/
513 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
514 
515 	/*
516 	 * Actual scheduling parameters. Initialized with the values above,
517 	 * they are continously updated during task execution. Note that
518 	 * the remaining runtime could be < 0 in case we are in overrun.
519 	 */
520 	s64				runtime;	/* Remaining runtime for this instance	*/
521 	u64				deadline;	/* Absolute deadline for this instance	*/
522 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
523 
524 	/*
525 	 * Some bool flags:
526 	 *
527 	 * @dl_throttled tells if we exhausted the runtime. If so, the
528 	 * task has to wait for a replenishment to be performed at the
529 	 * next firing of dl_timer.
530 	 *
531 	 * @dl_boosted tells if we are boosted due to DI. If so we are
532 	 * outside bandwidth enforcement mechanism (but only until we
533 	 * exit the critical section);
534 	 *
535 	 * @dl_yielded tells if task gave up the CPU before consuming
536 	 * all its available runtime during the last job.
537 	 *
538 	 * @dl_non_contending tells if the task is inactive while still
539 	 * contributing to the active utilization. In other words, it
540 	 * indicates if the inactive timer has been armed and its handler
541 	 * has not been executed yet. This flag is useful to avoid race
542 	 * conditions between the inactive timer handler and the wakeup
543 	 * code.
544 	 *
545 	 * @dl_overrun tells if the task asked to be informed about runtime
546 	 * overruns.
547 	 */
548 	unsigned int			dl_throttled      : 1;
549 	unsigned int			dl_boosted        : 1;
550 	unsigned int			dl_yielded        : 1;
551 	unsigned int			dl_non_contending : 1;
552 	unsigned int			dl_overrun	  : 1;
553 
554 	/*
555 	 * Bandwidth enforcement timer. Each -deadline task has its
556 	 * own bandwidth to be enforced, thus we need one timer per task.
557 	 */
558 	struct hrtimer			dl_timer;
559 
560 	/*
561 	 * Inactive timer, responsible for decreasing the active utilization
562 	 * at the "0-lag time". When a -deadline task blocks, it contributes
563 	 * to GRUB's active utilization until the "0-lag time", hence a
564 	 * timer is needed to decrease the active utilization at the correct
565 	 * time.
566 	 */
567 	struct hrtimer inactive_timer;
568 };
569 
570 union rcu_special {
571 	struct {
572 		u8			blocked;
573 		u8			need_qs;
574 		u8			exp_need_qs;
575 
576 		/* Otherwise the compiler can store garbage here: */
577 		u8			pad;
578 	} b; /* Bits. */
579 	u32 s; /* Set of bits. */
580 };
581 
582 enum perf_event_task_context {
583 	perf_invalid_context = -1,
584 	perf_hw_context = 0,
585 	perf_sw_context,
586 	perf_nr_task_contexts,
587 };
588 
589 struct wake_q_node {
590 	struct wake_q_node *next;
591 };
592 
593 struct task_struct {
594 #ifdef CONFIG_THREAD_INFO_IN_TASK
595 	/*
596 	 * For reasons of header soup (see current_thread_info()), this
597 	 * must be the first element of task_struct.
598 	 */
599 	struct thread_info		thread_info;
600 #endif
601 	/* -1 unrunnable, 0 runnable, >0 stopped: */
602 	volatile long			state;
603 
604 	/*
605 	 * This begins the randomizable portion of task_struct. Only
606 	 * scheduling-critical items should be added above here.
607 	 */
608 	randomized_struct_fields_start
609 
610 	void				*stack;
611 	atomic_t			usage;
612 	/* Per task flags (PF_*), defined further below: */
613 	unsigned int			flags;
614 	unsigned int			ptrace;
615 
616 #ifdef CONFIG_SMP
617 	struct llist_node		wake_entry;
618 	int				on_cpu;
619 #ifdef CONFIG_THREAD_INFO_IN_TASK
620 	/* Current CPU: */
621 	unsigned int			cpu;
622 #endif
623 	unsigned int			wakee_flips;
624 	unsigned long			wakee_flip_decay_ts;
625 	struct task_struct		*last_wakee;
626 
627 	/*
628 	 * recent_used_cpu is initially set as the last CPU used by a task
629 	 * that wakes affine another task. Waker/wakee relationships can
630 	 * push tasks around a CPU where each wakeup moves to the next one.
631 	 * Tracking a recently used CPU allows a quick search for a recently
632 	 * used CPU that may be idle.
633 	 */
634 	int				recent_used_cpu;
635 	int				wake_cpu;
636 #endif
637 	int				on_rq;
638 
639 	int				prio;
640 	int				static_prio;
641 	int				normal_prio;
642 	unsigned int			rt_priority;
643 
644 	const struct sched_class	*sched_class;
645 	struct sched_entity		se;
646 	struct sched_rt_entity		rt;
647 #ifdef CONFIG_CGROUP_SCHED
648 	struct task_group		*sched_task_group;
649 #endif
650 	struct sched_dl_entity		dl;
651 
652 #ifdef CONFIG_PREEMPT_NOTIFIERS
653 	/* List of struct preempt_notifier: */
654 	struct hlist_head		preempt_notifiers;
655 #endif
656 
657 #ifdef CONFIG_BLK_DEV_IO_TRACE
658 	unsigned int			btrace_seq;
659 #endif
660 
661 	unsigned int			policy;
662 	int				nr_cpus_allowed;
663 	cpumask_t			cpus_allowed;
664 
665 #ifdef CONFIG_PREEMPT_RCU
666 	int				rcu_read_lock_nesting;
667 	union rcu_special		rcu_read_unlock_special;
668 	struct list_head		rcu_node_entry;
669 	struct rcu_node			*rcu_blocked_node;
670 #endif /* #ifdef CONFIG_PREEMPT_RCU */
671 
672 #ifdef CONFIG_TASKS_RCU
673 	unsigned long			rcu_tasks_nvcsw;
674 	u8				rcu_tasks_holdout;
675 	u8				rcu_tasks_idx;
676 	int				rcu_tasks_idle_cpu;
677 	struct list_head		rcu_tasks_holdout_list;
678 #endif /* #ifdef CONFIG_TASKS_RCU */
679 
680 	struct sched_info		sched_info;
681 
682 	struct list_head		tasks;
683 #ifdef CONFIG_SMP
684 	struct plist_node		pushable_tasks;
685 	struct rb_node			pushable_dl_tasks;
686 #endif
687 
688 	struct mm_struct		*mm;
689 	struct mm_struct		*active_mm;
690 
691 	/* Per-thread vma caching: */
692 	struct vmacache			vmacache;
693 
694 #ifdef SPLIT_RSS_COUNTING
695 	struct task_rss_stat		rss_stat;
696 #endif
697 	int				exit_state;
698 	int				exit_code;
699 	int				exit_signal;
700 	/* The signal sent when the parent dies: */
701 	int				pdeath_signal;
702 	/* JOBCTL_*, siglock protected: */
703 	unsigned long			jobctl;
704 
705 	/* Used for emulating ABI behavior of previous Linux versions: */
706 	unsigned int			personality;
707 
708 	/* Scheduler bits, serialized by scheduler locks: */
709 	unsigned			sched_reset_on_fork:1;
710 	unsigned			sched_contributes_to_load:1;
711 	unsigned			sched_migrated:1;
712 	unsigned			sched_remote_wakeup:1;
713 	/* Force alignment to the next boundary: */
714 	unsigned			:0;
715 
716 	/* Unserialized, strictly 'current' */
717 
718 	/* Bit to tell LSMs we're in execve(): */
719 	unsigned			in_execve:1;
720 	unsigned			in_iowait:1;
721 #ifndef TIF_RESTORE_SIGMASK
722 	unsigned			restore_sigmask:1;
723 #endif
724 #ifdef CONFIG_MEMCG
725 	unsigned			in_user_fault:1;
726 #ifdef CONFIG_MEMCG_KMEM
727 	unsigned			memcg_kmem_skip_account:1;
728 #endif
729 #endif
730 #ifdef CONFIG_COMPAT_BRK
731 	unsigned			brk_randomized:1;
732 #endif
733 #ifdef CONFIG_CGROUPS
734 	/* disallow userland-initiated cgroup migration */
735 	unsigned			no_cgroup_migration:1;
736 #endif
737 #ifdef CONFIG_BLK_CGROUP
738 	/* to be used once the psi infrastructure lands upstream. */
739 	unsigned			use_memdelay:1;
740 #endif
741 
742 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
743 
744 	struct restart_block		restart_block;
745 
746 	pid_t				pid;
747 	pid_t				tgid;
748 
749 #ifdef CONFIG_STACKPROTECTOR
750 	/* Canary value for the -fstack-protector GCC feature: */
751 	unsigned long			stack_canary;
752 #endif
753 	/*
754 	 * Pointers to the (original) parent process, youngest child, younger sibling,
755 	 * older sibling, respectively.  (p->father can be replaced with
756 	 * p->real_parent->pid)
757 	 */
758 
759 	/* Real parent process: */
760 	struct task_struct __rcu	*real_parent;
761 
762 	/* Recipient of SIGCHLD, wait4() reports: */
763 	struct task_struct __rcu	*parent;
764 
765 	/*
766 	 * Children/sibling form the list of natural children:
767 	 */
768 	struct list_head		children;
769 	struct list_head		sibling;
770 	struct task_struct		*group_leader;
771 
772 	/*
773 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
774 	 *
775 	 * This includes both natural children and PTRACE_ATTACH targets.
776 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
777 	 */
778 	struct list_head		ptraced;
779 	struct list_head		ptrace_entry;
780 
781 	/* PID/PID hash table linkage. */
782 	struct pid			*thread_pid;
783 	struct hlist_node		pid_links[PIDTYPE_MAX];
784 	struct list_head		thread_group;
785 	struct list_head		thread_node;
786 
787 	struct completion		*vfork_done;
788 
789 	/* CLONE_CHILD_SETTID: */
790 	int __user			*set_child_tid;
791 
792 	/* CLONE_CHILD_CLEARTID: */
793 	int __user			*clear_child_tid;
794 
795 	u64				utime;
796 	u64				stime;
797 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
798 	u64				utimescaled;
799 	u64				stimescaled;
800 #endif
801 	u64				gtime;
802 	struct prev_cputime		prev_cputime;
803 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
804 	struct vtime			vtime;
805 #endif
806 
807 #ifdef CONFIG_NO_HZ_FULL
808 	atomic_t			tick_dep_mask;
809 #endif
810 	/* Context switch counts: */
811 	unsigned long			nvcsw;
812 	unsigned long			nivcsw;
813 
814 	/* Monotonic time in nsecs: */
815 	u64				start_time;
816 
817 	/* Boot based time in nsecs: */
818 	u64				real_start_time;
819 
820 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
821 	unsigned long			min_flt;
822 	unsigned long			maj_flt;
823 
824 #ifdef CONFIG_POSIX_TIMERS
825 	struct task_cputime		cputime_expires;
826 	struct list_head		cpu_timers[3];
827 #endif
828 
829 	/* Process credentials: */
830 
831 	/* Tracer's credentials at attach: */
832 	const struct cred __rcu		*ptracer_cred;
833 
834 	/* Objective and real subjective task credentials (COW): */
835 	const struct cred __rcu		*real_cred;
836 
837 	/* Effective (overridable) subjective task credentials (COW): */
838 	const struct cred __rcu		*cred;
839 
840 	/*
841 	 * executable name, excluding path.
842 	 *
843 	 * - normally initialized setup_new_exec()
844 	 * - access it with [gs]et_task_comm()
845 	 * - lock it with task_lock()
846 	 */
847 	char				comm[TASK_COMM_LEN];
848 
849 	struct nameidata		*nameidata;
850 
851 #ifdef CONFIG_SYSVIPC
852 	struct sysv_sem			sysvsem;
853 	struct sysv_shm			sysvshm;
854 #endif
855 #ifdef CONFIG_DETECT_HUNG_TASK
856 	unsigned long			last_switch_count;
857 	unsigned long			last_switch_time;
858 #endif
859 	/* Filesystem information: */
860 	struct fs_struct		*fs;
861 
862 	/* Open file information: */
863 	struct files_struct		*files;
864 
865 	/* Namespaces: */
866 	struct nsproxy			*nsproxy;
867 
868 	/* Signal handlers: */
869 	struct signal_struct		*signal;
870 	struct sighand_struct		*sighand;
871 	sigset_t			blocked;
872 	sigset_t			real_blocked;
873 	/* Restored if set_restore_sigmask() was used: */
874 	sigset_t			saved_sigmask;
875 	struct sigpending		pending;
876 	unsigned long			sas_ss_sp;
877 	size_t				sas_ss_size;
878 	unsigned int			sas_ss_flags;
879 
880 	struct callback_head		*task_works;
881 
882 	struct audit_context		*audit_context;
883 #ifdef CONFIG_AUDITSYSCALL
884 	kuid_t				loginuid;
885 	unsigned int			sessionid;
886 #endif
887 	struct seccomp			seccomp;
888 
889 	/* Thread group tracking: */
890 	u32				parent_exec_id;
891 	u32				self_exec_id;
892 
893 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
894 	spinlock_t			alloc_lock;
895 
896 	/* Protection of the PI data structures: */
897 	raw_spinlock_t			pi_lock;
898 
899 	struct wake_q_node		wake_q;
900 
901 #ifdef CONFIG_RT_MUTEXES
902 	/* PI waiters blocked on a rt_mutex held by this task: */
903 	struct rb_root_cached		pi_waiters;
904 	/* Updated under owner's pi_lock and rq lock */
905 	struct task_struct		*pi_top_task;
906 	/* Deadlock detection and priority inheritance handling: */
907 	struct rt_mutex_waiter		*pi_blocked_on;
908 #endif
909 
910 #ifdef CONFIG_DEBUG_MUTEXES
911 	/* Mutex deadlock detection: */
912 	struct mutex_waiter		*blocked_on;
913 #endif
914 
915 #ifdef CONFIG_TRACE_IRQFLAGS
916 	unsigned int			irq_events;
917 	unsigned long			hardirq_enable_ip;
918 	unsigned long			hardirq_disable_ip;
919 	unsigned int			hardirq_enable_event;
920 	unsigned int			hardirq_disable_event;
921 	int				hardirqs_enabled;
922 	int				hardirq_context;
923 	unsigned long			softirq_disable_ip;
924 	unsigned long			softirq_enable_ip;
925 	unsigned int			softirq_disable_event;
926 	unsigned int			softirq_enable_event;
927 	int				softirqs_enabled;
928 	int				softirq_context;
929 #endif
930 
931 #ifdef CONFIG_LOCKDEP
932 # define MAX_LOCK_DEPTH			48UL
933 	u64				curr_chain_key;
934 	int				lockdep_depth;
935 	unsigned int			lockdep_recursion;
936 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
937 #endif
938 
939 #ifdef CONFIG_UBSAN
940 	unsigned int			in_ubsan;
941 #endif
942 
943 	/* Journalling filesystem info: */
944 	void				*journal_info;
945 
946 	/* Stacked block device info: */
947 	struct bio_list			*bio_list;
948 
949 #ifdef CONFIG_BLOCK
950 	/* Stack plugging: */
951 	struct blk_plug			*plug;
952 #endif
953 
954 	/* VM state: */
955 	struct reclaim_state		*reclaim_state;
956 
957 	struct backing_dev_info		*backing_dev_info;
958 
959 	struct io_context		*io_context;
960 
961 	/* Ptrace state: */
962 	unsigned long			ptrace_message;
963 	siginfo_t			*last_siginfo;
964 
965 	struct task_io_accounting	ioac;
966 #ifdef CONFIG_TASK_XACCT
967 	/* Accumulated RSS usage: */
968 	u64				acct_rss_mem1;
969 	/* Accumulated virtual memory usage: */
970 	u64				acct_vm_mem1;
971 	/* stime + utime since last update: */
972 	u64				acct_timexpd;
973 #endif
974 #ifdef CONFIG_CPUSETS
975 	/* Protected by ->alloc_lock: */
976 	nodemask_t			mems_allowed;
977 	/* Seqence number to catch updates: */
978 	seqcount_t			mems_allowed_seq;
979 	int				cpuset_mem_spread_rotor;
980 	int				cpuset_slab_spread_rotor;
981 #endif
982 #ifdef CONFIG_CGROUPS
983 	/* Control Group info protected by css_set_lock: */
984 	struct css_set __rcu		*cgroups;
985 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
986 	struct list_head		cg_list;
987 #endif
988 #ifdef CONFIG_INTEL_RDT
989 	u32				closid;
990 	u32				rmid;
991 #endif
992 #ifdef CONFIG_FUTEX
993 	struct robust_list_head __user	*robust_list;
994 #ifdef CONFIG_COMPAT
995 	struct compat_robust_list_head __user *compat_robust_list;
996 #endif
997 	struct list_head		pi_state_list;
998 	struct futex_pi_state		*pi_state_cache;
999 #endif
1000 #ifdef CONFIG_PERF_EVENTS
1001 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1002 	struct mutex			perf_event_mutex;
1003 	struct list_head		perf_event_list;
1004 #endif
1005 #ifdef CONFIG_DEBUG_PREEMPT
1006 	unsigned long			preempt_disable_ip;
1007 #endif
1008 #ifdef CONFIG_NUMA
1009 	/* Protected by alloc_lock: */
1010 	struct mempolicy		*mempolicy;
1011 	short				il_prev;
1012 	short				pref_node_fork;
1013 #endif
1014 #ifdef CONFIG_NUMA_BALANCING
1015 	int				numa_scan_seq;
1016 	unsigned int			numa_scan_period;
1017 	unsigned int			numa_scan_period_max;
1018 	int				numa_preferred_nid;
1019 	unsigned long			numa_migrate_retry;
1020 	/* Migration stamp: */
1021 	u64				node_stamp;
1022 	u64				last_task_numa_placement;
1023 	u64				last_sum_exec_runtime;
1024 	struct callback_head		numa_work;
1025 
1026 	struct numa_group		*numa_group;
1027 
1028 	/*
1029 	 * numa_faults is an array split into four regions:
1030 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1031 	 * in this precise order.
1032 	 *
1033 	 * faults_memory: Exponential decaying average of faults on a per-node
1034 	 * basis. Scheduling placement decisions are made based on these
1035 	 * counts. The values remain static for the duration of a PTE scan.
1036 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1037 	 * hinting fault was incurred.
1038 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1039 	 * during the current scan window. When the scan completes, the counts
1040 	 * in faults_memory and faults_cpu decay and these values are copied.
1041 	 */
1042 	unsigned long			*numa_faults;
1043 	unsigned long			total_numa_faults;
1044 
1045 	/*
1046 	 * numa_faults_locality tracks if faults recorded during the last
1047 	 * scan window were remote/local or failed to migrate. The task scan
1048 	 * period is adapted based on the locality of the faults with different
1049 	 * weights depending on whether they were shared or private faults
1050 	 */
1051 	unsigned long			numa_faults_locality[3];
1052 
1053 	unsigned long			numa_pages_migrated;
1054 #endif /* CONFIG_NUMA_BALANCING */
1055 
1056 #ifdef CONFIG_RSEQ
1057 	struct rseq __user *rseq;
1058 	u32 rseq_len;
1059 	u32 rseq_sig;
1060 	/*
1061 	 * RmW on rseq_event_mask must be performed atomically
1062 	 * with respect to preemption.
1063 	 */
1064 	unsigned long rseq_event_mask;
1065 #endif
1066 
1067 	struct tlbflush_unmap_batch	tlb_ubc;
1068 
1069 	struct rcu_head			rcu;
1070 
1071 	/* Cache last used pipe for splice(): */
1072 	struct pipe_inode_info		*splice_pipe;
1073 
1074 	struct page_frag		task_frag;
1075 
1076 #ifdef CONFIG_TASK_DELAY_ACCT
1077 	struct task_delay_info		*delays;
1078 #endif
1079 
1080 #ifdef CONFIG_FAULT_INJECTION
1081 	int				make_it_fail;
1082 	unsigned int			fail_nth;
1083 #endif
1084 	/*
1085 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1086 	 * balance_dirty_pages() for a dirty throttling pause:
1087 	 */
1088 	int				nr_dirtied;
1089 	int				nr_dirtied_pause;
1090 	/* Start of a write-and-pause period: */
1091 	unsigned long			dirty_paused_when;
1092 
1093 #ifdef CONFIG_LATENCYTOP
1094 	int				latency_record_count;
1095 	struct latency_record		latency_record[LT_SAVECOUNT];
1096 #endif
1097 	/*
1098 	 * Time slack values; these are used to round up poll() and
1099 	 * select() etc timeout values. These are in nanoseconds.
1100 	 */
1101 	u64				timer_slack_ns;
1102 	u64				default_timer_slack_ns;
1103 
1104 #ifdef CONFIG_KASAN
1105 	unsigned int			kasan_depth;
1106 #endif
1107 
1108 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1109 	/* Index of current stored address in ret_stack: */
1110 	int				curr_ret_stack;
1111 
1112 	/* Stack of return addresses for return function tracing: */
1113 	struct ftrace_ret_stack		*ret_stack;
1114 
1115 	/* Timestamp for last schedule: */
1116 	unsigned long long		ftrace_timestamp;
1117 
1118 	/*
1119 	 * Number of functions that haven't been traced
1120 	 * because of depth overrun:
1121 	 */
1122 	atomic_t			trace_overrun;
1123 
1124 	/* Pause tracing: */
1125 	atomic_t			tracing_graph_pause;
1126 #endif
1127 
1128 #ifdef CONFIG_TRACING
1129 	/* State flags for use by tracers: */
1130 	unsigned long			trace;
1131 
1132 	/* Bitmask and counter of trace recursion: */
1133 	unsigned long			trace_recursion;
1134 #endif /* CONFIG_TRACING */
1135 
1136 #ifdef CONFIG_KCOV
1137 	/* Coverage collection mode enabled for this task (0 if disabled): */
1138 	unsigned int			kcov_mode;
1139 
1140 	/* Size of the kcov_area: */
1141 	unsigned int			kcov_size;
1142 
1143 	/* Buffer for coverage collection: */
1144 	void				*kcov_area;
1145 
1146 	/* KCOV descriptor wired with this task or NULL: */
1147 	struct kcov			*kcov;
1148 #endif
1149 
1150 #ifdef CONFIG_MEMCG
1151 	struct mem_cgroup		*memcg_in_oom;
1152 	gfp_t				memcg_oom_gfp_mask;
1153 	int				memcg_oom_order;
1154 
1155 	/* Number of pages to reclaim on returning to userland: */
1156 	unsigned int			memcg_nr_pages_over_high;
1157 
1158 	/* Used by memcontrol for targeted memcg charge: */
1159 	struct mem_cgroup		*active_memcg;
1160 #endif
1161 
1162 #ifdef CONFIG_BLK_CGROUP
1163 	struct request_queue		*throttle_queue;
1164 #endif
1165 
1166 #ifdef CONFIG_UPROBES
1167 	struct uprobe_task		*utask;
1168 #endif
1169 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1170 	unsigned int			sequential_io;
1171 	unsigned int			sequential_io_avg;
1172 #endif
1173 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1174 	unsigned long			task_state_change;
1175 #endif
1176 	int				pagefault_disabled;
1177 #ifdef CONFIG_MMU
1178 	struct task_struct		*oom_reaper_list;
1179 #endif
1180 #ifdef CONFIG_VMAP_STACK
1181 	struct vm_struct		*stack_vm_area;
1182 #endif
1183 #ifdef CONFIG_THREAD_INFO_IN_TASK
1184 	/* A live task holds one reference: */
1185 	atomic_t			stack_refcount;
1186 #endif
1187 #ifdef CONFIG_LIVEPATCH
1188 	int patch_state;
1189 #endif
1190 #ifdef CONFIG_SECURITY
1191 	/* Used by LSM modules for access restriction: */
1192 	void				*security;
1193 #endif
1194 
1195 	/*
1196 	 * New fields for task_struct should be added above here, so that
1197 	 * they are included in the randomized portion of task_struct.
1198 	 */
1199 	randomized_struct_fields_end
1200 
1201 	/* CPU-specific state of this task: */
1202 	struct thread_struct		thread;
1203 
1204 	/*
1205 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1206 	 * structure.  It *MUST* be at the end of 'task_struct'.
1207 	 *
1208 	 * Do not put anything below here!
1209 	 */
1210 };
1211 
1212 static inline struct pid *task_pid(struct task_struct *task)
1213 {
1214 	return task->thread_pid;
1215 }
1216 
1217 /*
1218  * the helpers to get the task's different pids as they are seen
1219  * from various namespaces
1220  *
1221  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1222  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1223  *                     current.
1224  * task_xid_nr_ns()  : id seen from the ns specified;
1225  *
1226  * see also pid_nr() etc in include/linux/pid.h
1227  */
1228 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1229 
1230 static inline pid_t task_pid_nr(struct task_struct *tsk)
1231 {
1232 	return tsk->pid;
1233 }
1234 
1235 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1236 {
1237 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1238 }
1239 
1240 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1241 {
1242 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1243 }
1244 
1245 
1246 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1247 {
1248 	return tsk->tgid;
1249 }
1250 
1251 /**
1252  * pid_alive - check that a task structure is not stale
1253  * @p: Task structure to be checked.
1254  *
1255  * Test if a process is not yet dead (at most zombie state)
1256  * If pid_alive fails, then pointers within the task structure
1257  * can be stale and must not be dereferenced.
1258  *
1259  * Return: 1 if the process is alive. 0 otherwise.
1260  */
1261 static inline int pid_alive(const struct task_struct *p)
1262 {
1263 	return p->thread_pid != NULL;
1264 }
1265 
1266 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1267 {
1268 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1269 }
1270 
1271 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1272 {
1273 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1274 }
1275 
1276 
1277 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1278 {
1279 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1280 }
1281 
1282 static inline pid_t task_session_vnr(struct task_struct *tsk)
1283 {
1284 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1285 }
1286 
1287 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1288 {
1289 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1290 }
1291 
1292 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1293 {
1294 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1295 }
1296 
1297 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1298 {
1299 	pid_t pid = 0;
1300 
1301 	rcu_read_lock();
1302 	if (pid_alive(tsk))
1303 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1304 	rcu_read_unlock();
1305 
1306 	return pid;
1307 }
1308 
1309 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1310 {
1311 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1312 }
1313 
1314 /* Obsolete, do not use: */
1315 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1316 {
1317 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1318 }
1319 
1320 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1321 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1322 
1323 static inline unsigned int task_state_index(struct task_struct *tsk)
1324 {
1325 	unsigned int tsk_state = READ_ONCE(tsk->state);
1326 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1327 
1328 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1329 
1330 	if (tsk_state == TASK_IDLE)
1331 		state = TASK_REPORT_IDLE;
1332 
1333 	return fls(state);
1334 }
1335 
1336 static inline char task_index_to_char(unsigned int state)
1337 {
1338 	static const char state_char[] = "RSDTtXZPI";
1339 
1340 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1341 
1342 	return state_char[state];
1343 }
1344 
1345 static inline char task_state_to_char(struct task_struct *tsk)
1346 {
1347 	return task_index_to_char(task_state_index(tsk));
1348 }
1349 
1350 /**
1351  * is_global_init - check if a task structure is init. Since init
1352  * is free to have sub-threads we need to check tgid.
1353  * @tsk: Task structure to be checked.
1354  *
1355  * Check if a task structure is the first user space task the kernel created.
1356  *
1357  * Return: 1 if the task structure is init. 0 otherwise.
1358  */
1359 static inline int is_global_init(struct task_struct *tsk)
1360 {
1361 	return task_tgid_nr(tsk) == 1;
1362 }
1363 
1364 extern struct pid *cad_pid;
1365 
1366 /*
1367  * Per process flags
1368  */
1369 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1370 #define PF_EXITING		0x00000004	/* Getting shut down */
1371 #define PF_EXITPIDONE		0x00000008	/* PI exit done on shut down */
1372 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1373 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1374 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1375 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1376 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1377 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1378 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1379 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1380 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1381 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1382 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1383 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1384 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1385 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1386 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1387 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1388 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1389 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1390 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1391 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1392 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_allowed */
1393 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1394 #define PF_MUTEX_TESTER		0x20000000	/* Thread belongs to the rt mutex tester */
1395 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1396 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1397 
1398 /*
1399  * Only the _current_ task can read/write to tsk->flags, but other
1400  * tasks can access tsk->flags in readonly mode for example
1401  * with tsk_used_math (like during threaded core dumping).
1402  * There is however an exception to this rule during ptrace
1403  * or during fork: the ptracer task is allowed to write to the
1404  * child->flags of its traced child (same goes for fork, the parent
1405  * can write to the child->flags), because we're guaranteed the
1406  * child is not running and in turn not changing child->flags
1407  * at the same time the parent does it.
1408  */
1409 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1410 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1411 #define clear_used_math()			clear_stopped_child_used_math(current)
1412 #define set_used_math()				set_stopped_child_used_math(current)
1413 
1414 #define conditional_stopped_child_used_math(condition, child) \
1415 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1416 
1417 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1418 
1419 #define copy_to_stopped_child_used_math(child) \
1420 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1421 
1422 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1423 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1424 #define used_math()				tsk_used_math(current)
1425 
1426 static inline bool is_percpu_thread(void)
1427 {
1428 #ifdef CONFIG_SMP
1429 	return (current->flags & PF_NO_SETAFFINITY) &&
1430 		(current->nr_cpus_allowed  == 1);
1431 #else
1432 	return true;
1433 #endif
1434 }
1435 
1436 /* Per-process atomic flags. */
1437 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1438 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1439 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1440 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1441 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1442 
1443 #define TASK_PFA_TEST(name, func)					\
1444 	static inline bool task_##func(struct task_struct *p)		\
1445 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1446 
1447 #define TASK_PFA_SET(name, func)					\
1448 	static inline void task_set_##func(struct task_struct *p)	\
1449 	{ set_bit(PFA_##name, &p->atomic_flags); }
1450 
1451 #define TASK_PFA_CLEAR(name, func)					\
1452 	static inline void task_clear_##func(struct task_struct *p)	\
1453 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1454 
1455 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1456 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1457 
1458 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1459 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1460 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1461 
1462 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1463 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1464 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1465 
1466 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1467 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1468 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1469 
1470 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1471 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1472 
1473 static inline void
1474 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1475 {
1476 	current->flags &= ~flags;
1477 	current->flags |= orig_flags & flags;
1478 }
1479 
1480 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1481 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1482 #ifdef CONFIG_SMP
1483 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1484 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1485 #else
1486 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1487 {
1488 }
1489 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1490 {
1491 	if (!cpumask_test_cpu(0, new_mask))
1492 		return -EINVAL;
1493 	return 0;
1494 }
1495 #endif
1496 
1497 #ifndef cpu_relax_yield
1498 #define cpu_relax_yield() cpu_relax()
1499 #endif
1500 
1501 extern int yield_to(struct task_struct *p, bool preempt);
1502 extern void set_user_nice(struct task_struct *p, long nice);
1503 extern int task_prio(const struct task_struct *p);
1504 
1505 /**
1506  * task_nice - return the nice value of a given task.
1507  * @p: the task in question.
1508  *
1509  * Return: The nice value [ -20 ... 0 ... 19 ].
1510  */
1511 static inline int task_nice(const struct task_struct *p)
1512 {
1513 	return PRIO_TO_NICE((p)->static_prio);
1514 }
1515 
1516 extern int can_nice(const struct task_struct *p, const int nice);
1517 extern int task_curr(const struct task_struct *p);
1518 extern int idle_cpu(int cpu);
1519 extern int available_idle_cpu(int cpu);
1520 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1521 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1522 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1523 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1524 extern struct task_struct *idle_task(int cpu);
1525 
1526 /**
1527  * is_idle_task - is the specified task an idle task?
1528  * @p: the task in question.
1529  *
1530  * Return: 1 if @p is an idle task. 0 otherwise.
1531  */
1532 static inline bool is_idle_task(const struct task_struct *p)
1533 {
1534 	return !!(p->flags & PF_IDLE);
1535 }
1536 
1537 extern struct task_struct *curr_task(int cpu);
1538 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1539 
1540 void yield(void);
1541 
1542 union thread_union {
1543 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1544 	struct task_struct task;
1545 #endif
1546 #ifndef CONFIG_THREAD_INFO_IN_TASK
1547 	struct thread_info thread_info;
1548 #endif
1549 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1550 };
1551 
1552 #ifndef CONFIG_THREAD_INFO_IN_TASK
1553 extern struct thread_info init_thread_info;
1554 #endif
1555 
1556 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1557 
1558 #ifdef CONFIG_THREAD_INFO_IN_TASK
1559 static inline struct thread_info *task_thread_info(struct task_struct *task)
1560 {
1561 	return &task->thread_info;
1562 }
1563 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1564 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1565 #endif
1566 
1567 /*
1568  * find a task by one of its numerical ids
1569  *
1570  * find_task_by_pid_ns():
1571  *      finds a task by its pid in the specified namespace
1572  * find_task_by_vpid():
1573  *      finds a task by its virtual pid
1574  *
1575  * see also find_vpid() etc in include/linux/pid.h
1576  */
1577 
1578 extern struct task_struct *find_task_by_vpid(pid_t nr);
1579 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1580 
1581 /*
1582  * find a task by its virtual pid and get the task struct
1583  */
1584 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1585 
1586 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1587 extern int wake_up_process(struct task_struct *tsk);
1588 extern void wake_up_new_task(struct task_struct *tsk);
1589 
1590 #ifdef CONFIG_SMP
1591 extern void kick_process(struct task_struct *tsk);
1592 #else
1593 static inline void kick_process(struct task_struct *tsk) { }
1594 #endif
1595 
1596 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1597 
1598 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1599 {
1600 	__set_task_comm(tsk, from, false);
1601 }
1602 
1603 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1604 #define get_task_comm(buf, tsk) ({			\
1605 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1606 	__get_task_comm(buf, sizeof(buf), tsk);		\
1607 })
1608 
1609 #ifdef CONFIG_SMP
1610 void scheduler_ipi(void);
1611 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1612 #else
1613 static inline void scheduler_ipi(void) { }
1614 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1615 {
1616 	return 1;
1617 }
1618 #endif
1619 
1620 /*
1621  * Set thread flags in other task's structures.
1622  * See asm/thread_info.h for TIF_xxxx flags available:
1623  */
1624 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1625 {
1626 	set_ti_thread_flag(task_thread_info(tsk), flag);
1627 }
1628 
1629 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1630 {
1631 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1632 }
1633 
1634 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1635 					  bool value)
1636 {
1637 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1638 }
1639 
1640 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1641 {
1642 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1643 }
1644 
1645 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1646 {
1647 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1648 }
1649 
1650 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1651 {
1652 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1653 }
1654 
1655 static inline void set_tsk_need_resched(struct task_struct *tsk)
1656 {
1657 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1658 }
1659 
1660 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1661 {
1662 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1663 }
1664 
1665 static inline int test_tsk_need_resched(struct task_struct *tsk)
1666 {
1667 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1668 }
1669 
1670 /*
1671  * cond_resched() and cond_resched_lock(): latency reduction via
1672  * explicit rescheduling in places that are safe. The return
1673  * value indicates whether a reschedule was done in fact.
1674  * cond_resched_lock() will drop the spinlock before scheduling,
1675  */
1676 #ifndef CONFIG_PREEMPT
1677 extern int _cond_resched(void);
1678 #else
1679 static inline int _cond_resched(void) { return 0; }
1680 #endif
1681 
1682 #define cond_resched() ({			\
1683 	___might_sleep(__FILE__, __LINE__, 0);	\
1684 	_cond_resched();			\
1685 })
1686 
1687 extern int __cond_resched_lock(spinlock_t *lock);
1688 
1689 #define cond_resched_lock(lock) ({				\
1690 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1691 	__cond_resched_lock(lock);				\
1692 })
1693 
1694 static inline void cond_resched_rcu(void)
1695 {
1696 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1697 	rcu_read_unlock();
1698 	cond_resched();
1699 	rcu_read_lock();
1700 #endif
1701 }
1702 
1703 /*
1704  * Does a critical section need to be broken due to another
1705  * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1706  * but a general need for low latency)
1707  */
1708 static inline int spin_needbreak(spinlock_t *lock)
1709 {
1710 #ifdef CONFIG_PREEMPT
1711 	return spin_is_contended(lock);
1712 #else
1713 	return 0;
1714 #endif
1715 }
1716 
1717 static __always_inline bool need_resched(void)
1718 {
1719 	return unlikely(tif_need_resched());
1720 }
1721 
1722 /*
1723  * Wrappers for p->thread_info->cpu access. No-op on UP.
1724  */
1725 #ifdef CONFIG_SMP
1726 
1727 static inline unsigned int task_cpu(const struct task_struct *p)
1728 {
1729 #ifdef CONFIG_THREAD_INFO_IN_TASK
1730 	return p->cpu;
1731 #else
1732 	return task_thread_info(p)->cpu;
1733 #endif
1734 }
1735 
1736 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1737 
1738 #else
1739 
1740 static inline unsigned int task_cpu(const struct task_struct *p)
1741 {
1742 	return 0;
1743 }
1744 
1745 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1746 {
1747 }
1748 
1749 #endif /* CONFIG_SMP */
1750 
1751 /*
1752  * In order to reduce various lock holder preemption latencies provide an
1753  * interface to see if a vCPU is currently running or not.
1754  *
1755  * This allows us to terminate optimistic spin loops and block, analogous to
1756  * the native optimistic spin heuristic of testing if the lock owner task is
1757  * running or not.
1758  */
1759 #ifndef vcpu_is_preempted
1760 # define vcpu_is_preempted(cpu)	false
1761 #endif
1762 
1763 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1764 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1765 
1766 #ifndef TASK_SIZE_OF
1767 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1768 #endif
1769 
1770 #ifdef CONFIG_RSEQ
1771 
1772 /*
1773  * Map the event mask on the user-space ABI enum rseq_cs_flags
1774  * for direct mask checks.
1775  */
1776 enum rseq_event_mask_bits {
1777 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1778 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1779 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1780 };
1781 
1782 enum rseq_event_mask {
1783 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
1784 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
1785 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
1786 };
1787 
1788 static inline void rseq_set_notify_resume(struct task_struct *t)
1789 {
1790 	if (t->rseq)
1791 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1792 }
1793 
1794 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1795 
1796 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1797 					     struct pt_regs *regs)
1798 {
1799 	if (current->rseq)
1800 		__rseq_handle_notify_resume(ksig, regs);
1801 }
1802 
1803 static inline void rseq_signal_deliver(struct ksignal *ksig,
1804 				       struct pt_regs *regs)
1805 {
1806 	preempt_disable();
1807 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1808 	preempt_enable();
1809 	rseq_handle_notify_resume(ksig, regs);
1810 }
1811 
1812 /* rseq_preempt() requires preemption to be disabled. */
1813 static inline void rseq_preempt(struct task_struct *t)
1814 {
1815 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1816 	rseq_set_notify_resume(t);
1817 }
1818 
1819 /* rseq_migrate() requires preemption to be disabled. */
1820 static inline void rseq_migrate(struct task_struct *t)
1821 {
1822 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1823 	rseq_set_notify_resume(t);
1824 }
1825 
1826 /*
1827  * If parent process has a registered restartable sequences area, the
1828  * child inherits. Only applies when forking a process, not a thread.
1829  */
1830 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1831 {
1832 	if (clone_flags & CLONE_THREAD) {
1833 		t->rseq = NULL;
1834 		t->rseq_len = 0;
1835 		t->rseq_sig = 0;
1836 		t->rseq_event_mask = 0;
1837 	} else {
1838 		t->rseq = current->rseq;
1839 		t->rseq_len = current->rseq_len;
1840 		t->rseq_sig = current->rseq_sig;
1841 		t->rseq_event_mask = current->rseq_event_mask;
1842 	}
1843 }
1844 
1845 static inline void rseq_execve(struct task_struct *t)
1846 {
1847 	t->rseq = NULL;
1848 	t->rseq_len = 0;
1849 	t->rseq_sig = 0;
1850 	t->rseq_event_mask = 0;
1851 }
1852 
1853 #else
1854 
1855 static inline void rseq_set_notify_resume(struct task_struct *t)
1856 {
1857 }
1858 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1859 					     struct pt_regs *regs)
1860 {
1861 }
1862 static inline void rseq_signal_deliver(struct ksignal *ksig,
1863 				       struct pt_regs *regs)
1864 {
1865 }
1866 static inline void rseq_preempt(struct task_struct *t)
1867 {
1868 }
1869 static inline void rseq_migrate(struct task_struct *t)
1870 {
1871 }
1872 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1873 {
1874 }
1875 static inline void rseq_execve(struct task_struct *t)
1876 {
1877 }
1878 
1879 #endif
1880 
1881 #ifdef CONFIG_DEBUG_RSEQ
1882 
1883 void rseq_syscall(struct pt_regs *regs);
1884 
1885 #else
1886 
1887 static inline void rseq_syscall(struct pt_regs *regs)
1888 {
1889 }
1890 
1891 #endif
1892 
1893 #endif
1894