xref: /linux/include/linux/sched.h (revision bf070bb0e6c62ba3075db0a666763ba52c677102)
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 
31 /* task_struct member predeclarations (sorted alphabetically): */
32 struct audit_context;
33 struct backing_dev_info;
34 struct bio_list;
35 struct blk_plug;
36 struct cfs_rq;
37 struct fs_struct;
38 struct futex_pi_state;
39 struct io_context;
40 struct mempolicy;
41 struct nameidata;
42 struct nsproxy;
43 struct perf_event_context;
44 struct pid_namespace;
45 struct pipe_inode_info;
46 struct rcu_node;
47 struct reclaim_state;
48 struct robust_list_head;
49 struct sched_attr;
50 struct sched_param;
51 struct seq_file;
52 struct sighand_struct;
53 struct signal_struct;
54 struct task_delay_info;
55 struct task_group;
56 
57 /*
58  * Task state bitmask. NOTE! These bits are also
59  * encoded in fs/proc/array.c: get_task_state().
60  *
61  * We have two separate sets of flags: task->state
62  * is about runnability, while task->exit_state are
63  * about the task exiting. Confusing, but this way
64  * modifying one set can't modify the other one by
65  * mistake.
66  */
67 
68 /* Used in tsk->state: */
69 #define TASK_RUNNING			0x0000
70 #define TASK_INTERRUPTIBLE		0x0001
71 #define TASK_UNINTERRUPTIBLE		0x0002
72 #define __TASK_STOPPED			0x0004
73 #define __TASK_TRACED			0x0008
74 /* Used in tsk->exit_state: */
75 #define EXIT_DEAD			0x0010
76 #define EXIT_ZOMBIE			0x0020
77 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
78 /* Used in tsk->state again: */
79 #define TASK_PARKED			0x0040
80 #define TASK_DEAD			0x0080
81 #define TASK_WAKEKILL			0x0100
82 #define TASK_WAKING			0x0200
83 #define TASK_NOLOAD			0x0400
84 #define TASK_NEW			0x0800
85 #define TASK_STATE_MAX			0x1000
86 
87 /* Convenience macros for the sake of set_current_state: */
88 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
89 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
90 #define TASK_TRACED			(TASK_WAKEKILL | __TASK_TRACED)
91 
92 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
93 
94 /* Convenience macros for the sake of wake_up(): */
95 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
96 #define TASK_ALL			(TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
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 #define __set_current_state(state_value)			\
117 	do {							\
118 		current->task_state_change = _THIS_IP_;		\
119 		current->state = (state_value);			\
120 	} while (0)
121 #define set_current_state(state_value)				\
122 	do {							\
123 		current->task_state_change = _THIS_IP_;		\
124 		smp_store_mb(current->state, (state_value));	\
125 	} while (0)
126 
127 #else
128 /*
129  * set_current_state() includes a barrier so that the write of current->state
130  * is correctly serialised wrt the caller's subsequent test of whether to
131  * actually sleep:
132  *
133  *   for (;;) {
134  *	set_current_state(TASK_UNINTERRUPTIBLE);
135  *	if (!need_sleep)
136  *		break;
137  *
138  *	schedule();
139  *   }
140  *   __set_current_state(TASK_RUNNING);
141  *
142  * If the caller does not need such serialisation (because, for instance, the
143  * condition test and condition change and wakeup are under the same lock) then
144  * use __set_current_state().
145  *
146  * The above is typically ordered against the wakeup, which does:
147  *
148  *	need_sleep = false;
149  *	wake_up_state(p, TASK_UNINTERRUPTIBLE);
150  *
151  * Where wake_up_state() (and all other wakeup primitives) imply enough
152  * barriers to order the store of the variable against wakeup.
153  *
154  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
155  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
156  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
157  *
158  * This is obviously fine, since they both store the exact same value.
159  *
160  * Also see the comments of try_to_wake_up().
161  */
162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0)
163 #define set_current_state(state_value)	 smp_store_mb(current->state, (state_value))
164 #endif
165 
166 /* Task command name length: */
167 #define TASK_COMM_LEN			16
168 
169 extern void scheduler_tick(void);
170 
171 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
172 
173 extern long schedule_timeout(long timeout);
174 extern long schedule_timeout_interruptible(long timeout);
175 extern long schedule_timeout_killable(long timeout);
176 extern long schedule_timeout_uninterruptible(long timeout);
177 extern long schedule_timeout_idle(long timeout);
178 asmlinkage void schedule(void);
179 extern void schedule_preempt_disabled(void);
180 
181 extern int __must_check io_schedule_prepare(void);
182 extern void io_schedule_finish(int token);
183 extern long io_schedule_timeout(long timeout);
184 extern void io_schedule(void);
185 
186 /**
187  * struct prev_cputime - snapshot of system and user cputime
188  * @utime: time spent in user mode
189  * @stime: time spent in system mode
190  * @lock: protects the above two fields
191  *
192  * Stores previous user/system time values such that we can guarantee
193  * monotonicity.
194  */
195 struct prev_cputime {
196 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
197 	u64				utime;
198 	u64				stime;
199 	raw_spinlock_t			lock;
200 #endif
201 };
202 
203 /**
204  * struct task_cputime - collected CPU time counts
205  * @utime:		time spent in user mode, in nanoseconds
206  * @stime:		time spent in kernel mode, in nanoseconds
207  * @sum_exec_runtime:	total time spent on the CPU, in nanoseconds
208  *
209  * This structure groups together three kinds of CPU time that are tracked for
210  * threads and thread groups.  Most things considering CPU time want to group
211  * these counts together and treat all three of them in parallel.
212  */
213 struct task_cputime {
214 	u64				utime;
215 	u64				stime;
216 	unsigned long long		sum_exec_runtime;
217 };
218 
219 /* Alternate field names when used on cache expirations: */
220 #define virt_exp			utime
221 #define prof_exp			stime
222 #define sched_exp			sum_exec_runtime
223 
224 enum vtime_state {
225 	/* Task is sleeping or running in a CPU with VTIME inactive: */
226 	VTIME_INACTIVE = 0,
227 	/* Task runs in userspace in a CPU with VTIME active: */
228 	VTIME_USER,
229 	/* Task runs in kernelspace in a CPU with VTIME active: */
230 	VTIME_SYS,
231 };
232 
233 struct vtime {
234 	seqcount_t		seqcount;
235 	unsigned long long	starttime;
236 	enum vtime_state	state;
237 	u64			utime;
238 	u64			stime;
239 	u64			gtime;
240 };
241 
242 struct sched_info {
243 #ifdef CONFIG_SCHED_INFO
244 	/* Cumulative counters: */
245 
246 	/* # of times we have run on this CPU: */
247 	unsigned long			pcount;
248 
249 	/* Time spent waiting on a runqueue: */
250 	unsigned long long		run_delay;
251 
252 	/* Timestamps: */
253 
254 	/* When did we last run on a CPU? */
255 	unsigned long long		last_arrival;
256 
257 	/* When were we last queued to run? */
258 	unsigned long long		last_queued;
259 
260 #endif /* CONFIG_SCHED_INFO */
261 };
262 
263 /*
264  * Integer metrics need fixed point arithmetic, e.g., sched/fair
265  * has a few: load, load_avg, util_avg, freq, and capacity.
266  *
267  * We define a basic fixed point arithmetic range, and then formalize
268  * all these metrics based on that basic range.
269  */
270 # define SCHED_FIXEDPOINT_SHIFT		10
271 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
272 
273 struct load_weight {
274 	unsigned long			weight;
275 	u32				inv_weight;
276 };
277 
278 /*
279  * The load_avg/util_avg accumulates an infinite geometric series
280  * (see __update_load_avg() in kernel/sched/fair.c).
281  *
282  * [load_avg definition]
283  *
284  *   load_avg = runnable% * scale_load_down(load)
285  *
286  * where runnable% is the time ratio that a sched_entity is runnable.
287  * For cfs_rq, it is the aggregated load_avg of all runnable and
288  * blocked sched_entities.
289  *
290  * load_avg may also take frequency scaling into account:
291  *
292  *   load_avg = runnable% * scale_load_down(load) * freq%
293  *
294  * where freq% is the CPU frequency normalized to the highest frequency.
295  *
296  * [util_avg definition]
297  *
298  *   util_avg = running% * SCHED_CAPACITY_SCALE
299  *
300  * where running% is the time ratio that a sched_entity is running on
301  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
302  * and blocked sched_entities.
303  *
304  * util_avg may also factor frequency scaling and CPU capacity scaling:
305  *
306  *   util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
307  *
308  * where freq% is the same as above, and capacity% is the CPU capacity
309  * normalized to the greatest capacity (due to uarch differences, etc).
310  *
311  * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
312  * themselves are in the range of [0, 1]. To do fixed point arithmetics,
313  * we therefore scale them to as large a range as necessary. This is for
314  * example reflected by util_avg's SCHED_CAPACITY_SCALE.
315  *
316  * [Overflow issue]
317  *
318  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
319  * with the highest load (=88761), always runnable on a single cfs_rq,
320  * and should not overflow as the number already hits PID_MAX_LIMIT.
321  *
322  * For all other cases (including 32-bit kernels), struct load_weight's
323  * weight will overflow first before we do, because:
324  *
325  *    Max(load_avg) <= Max(load.weight)
326  *
327  * Then it is the load_weight's responsibility to consider overflow
328  * issues.
329  */
330 struct sched_avg {
331 	u64				last_update_time;
332 	u64				load_sum;
333 	u64				runnable_load_sum;
334 	u32				util_sum;
335 	u32				period_contrib;
336 	unsigned long			load_avg;
337 	unsigned long			runnable_load_avg;
338 	unsigned long			util_avg;
339 };
340 
341 struct sched_statistics {
342 #ifdef CONFIG_SCHEDSTATS
343 	u64				wait_start;
344 	u64				wait_max;
345 	u64				wait_count;
346 	u64				wait_sum;
347 	u64				iowait_count;
348 	u64				iowait_sum;
349 
350 	u64				sleep_start;
351 	u64				sleep_max;
352 	s64				sum_sleep_runtime;
353 
354 	u64				block_start;
355 	u64				block_max;
356 	u64				exec_max;
357 	u64				slice_max;
358 
359 	u64				nr_migrations_cold;
360 	u64				nr_failed_migrations_affine;
361 	u64				nr_failed_migrations_running;
362 	u64				nr_failed_migrations_hot;
363 	u64				nr_forced_migrations;
364 
365 	u64				nr_wakeups;
366 	u64				nr_wakeups_sync;
367 	u64				nr_wakeups_migrate;
368 	u64				nr_wakeups_local;
369 	u64				nr_wakeups_remote;
370 	u64				nr_wakeups_affine;
371 	u64				nr_wakeups_affine_attempts;
372 	u64				nr_wakeups_passive;
373 	u64				nr_wakeups_idle;
374 #endif
375 };
376 
377 struct sched_entity {
378 	/* For load-balancing: */
379 	struct load_weight		load;
380 	unsigned long			runnable_weight;
381 	struct rb_node			run_node;
382 	struct list_head		group_node;
383 	unsigned int			on_rq;
384 
385 	u64				exec_start;
386 	u64				sum_exec_runtime;
387 	u64				vruntime;
388 	u64				prev_sum_exec_runtime;
389 
390 	u64				nr_migrations;
391 
392 	struct sched_statistics		statistics;
393 
394 #ifdef CONFIG_FAIR_GROUP_SCHED
395 	int				depth;
396 	struct sched_entity		*parent;
397 	/* rq on which this entity is (to be) queued: */
398 	struct cfs_rq			*cfs_rq;
399 	/* rq "owned" by this entity/group: */
400 	struct cfs_rq			*my_q;
401 #endif
402 
403 #ifdef CONFIG_SMP
404 	/*
405 	 * Per entity load average tracking.
406 	 *
407 	 * Put into separate cache line so it does not
408 	 * collide with read-mostly values above.
409 	 */
410 	struct sched_avg		avg ____cacheline_aligned_in_smp;
411 #endif
412 };
413 
414 struct sched_rt_entity {
415 	struct list_head		run_list;
416 	unsigned long			timeout;
417 	unsigned long			watchdog_stamp;
418 	unsigned int			time_slice;
419 	unsigned short			on_rq;
420 	unsigned short			on_list;
421 
422 	struct sched_rt_entity		*back;
423 #ifdef CONFIG_RT_GROUP_SCHED
424 	struct sched_rt_entity		*parent;
425 	/* rq on which this entity is (to be) queued: */
426 	struct rt_rq			*rt_rq;
427 	/* rq "owned" by this entity/group: */
428 	struct rt_rq			*my_q;
429 #endif
430 } __randomize_layout;
431 
432 struct sched_dl_entity {
433 	struct rb_node			rb_node;
434 
435 	/*
436 	 * Original scheduling parameters. Copied here from sched_attr
437 	 * during sched_setattr(), they will remain the same until
438 	 * the next sched_setattr().
439 	 */
440 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
441 	u64				dl_deadline;	/* Relative deadline of each instance	*/
442 	u64				dl_period;	/* Separation of two instances (period) */
443 	u64				dl_bw;		/* dl_runtime / dl_period		*/
444 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
445 
446 	/*
447 	 * Actual scheduling parameters. Initialized with the values above,
448 	 * they are continously updated during task execution. Note that
449 	 * the remaining runtime could be < 0 in case we are in overrun.
450 	 */
451 	s64				runtime;	/* Remaining runtime for this instance	*/
452 	u64				deadline;	/* Absolute deadline for this instance	*/
453 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
454 
455 	/*
456 	 * Some bool flags:
457 	 *
458 	 * @dl_throttled tells if we exhausted the runtime. If so, the
459 	 * task has to wait for a replenishment to be performed at the
460 	 * next firing of dl_timer.
461 	 *
462 	 * @dl_boosted tells if we are boosted due to DI. If so we are
463 	 * outside bandwidth enforcement mechanism (but only until we
464 	 * exit the critical section);
465 	 *
466 	 * @dl_yielded tells if task gave up the CPU before consuming
467 	 * all its available runtime during the last job.
468 	 *
469 	 * @dl_non_contending tells if the task is inactive while still
470 	 * contributing to the active utilization. In other words, it
471 	 * indicates if the inactive timer has been armed and its handler
472 	 * has not been executed yet. This flag is useful to avoid race
473 	 * conditions between the inactive timer handler and the wakeup
474 	 * code.
475 	 */
476 	int				dl_throttled      : 1;
477 	int				dl_boosted        : 1;
478 	int				dl_yielded        : 1;
479 	int				dl_non_contending : 1;
480 
481 	/*
482 	 * Bandwidth enforcement timer. Each -deadline task has its
483 	 * own bandwidth to be enforced, thus we need one timer per task.
484 	 */
485 	struct hrtimer			dl_timer;
486 
487 	/*
488 	 * Inactive timer, responsible for decreasing the active utilization
489 	 * at the "0-lag time". When a -deadline task blocks, it contributes
490 	 * to GRUB's active utilization until the "0-lag time", hence a
491 	 * timer is needed to decrease the active utilization at the correct
492 	 * time.
493 	 */
494 	struct hrtimer inactive_timer;
495 };
496 
497 union rcu_special {
498 	struct {
499 		u8			blocked;
500 		u8			need_qs;
501 		u8			exp_need_qs;
502 
503 		/* Otherwise the compiler can store garbage here: */
504 		u8			pad;
505 	} b; /* Bits. */
506 	u32 s; /* Set of bits. */
507 };
508 
509 enum perf_event_task_context {
510 	perf_invalid_context = -1,
511 	perf_hw_context = 0,
512 	perf_sw_context,
513 	perf_nr_task_contexts,
514 };
515 
516 struct wake_q_node {
517 	struct wake_q_node *next;
518 };
519 
520 struct task_struct {
521 #ifdef CONFIG_THREAD_INFO_IN_TASK
522 	/*
523 	 * For reasons of header soup (see current_thread_info()), this
524 	 * must be the first element of task_struct.
525 	 */
526 	struct thread_info		thread_info;
527 #endif
528 	/* -1 unrunnable, 0 runnable, >0 stopped: */
529 	volatile long			state;
530 
531 	/*
532 	 * This begins the randomizable portion of task_struct. Only
533 	 * scheduling-critical items should be added above here.
534 	 */
535 	randomized_struct_fields_start
536 
537 	void				*stack;
538 	atomic_t			usage;
539 	/* Per task flags (PF_*), defined further below: */
540 	unsigned int			flags;
541 	unsigned int			ptrace;
542 
543 #ifdef CONFIG_SMP
544 	struct llist_node		wake_entry;
545 	int				on_cpu;
546 #ifdef CONFIG_THREAD_INFO_IN_TASK
547 	/* Current CPU: */
548 	unsigned int			cpu;
549 #endif
550 	unsigned int			wakee_flips;
551 	unsigned long			wakee_flip_decay_ts;
552 	struct task_struct		*last_wakee;
553 
554 	int				wake_cpu;
555 #endif
556 	int				on_rq;
557 
558 	int				prio;
559 	int				static_prio;
560 	int				normal_prio;
561 	unsigned int			rt_priority;
562 
563 	const struct sched_class	*sched_class;
564 	struct sched_entity		se;
565 	struct sched_rt_entity		rt;
566 #ifdef CONFIG_CGROUP_SCHED
567 	struct task_group		*sched_task_group;
568 #endif
569 	struct sched_dl_entity		dl;
570 
571 #ifdef CONFIG_PREEMPT_NOTIFIERS
572 	/* List of struct preempt_notifier: */
573 	struct hlist_head		preempt_notifiers;
574 #endif
575 
576 #ifdef CONFIG_BLK_DEV_IO_TRACE
577 	unsigned int			btrace_seq;
578 #endif
579 
580 	unsigned int			policy;
581 	int				nr_cpus_allowed;
582 	cpumask_t			cpus_allowed;
583 
584 #ifdef CONFIG_PREEMPT_RCU
585 	int				rcu_read_lock_nesting;
586 	union rcu_special		rcu_read_unlock_special;
587 	struct list_head		rcu_node_entry;
588 	struct rcu_node			*rcu_blocked_node;
589 #endif /* #ifdef CONFIG_PREEMPT_RCU */
590 
591 #ifdef CONFIG_TASKS_RCU
592 	unsigned long			rcu_tasks_nvcsw;
593 	u8				rcu_tasks_holdout;
594 	u8				rcu_tasks_idx;
595 	int				rcu_tasks_idle_cpu;
596 	struct list_head		rcu_tasks_holdout_list;
597 #endif /* #ifdef CONFIG_TASKS_RCU */
598 
599 	struct sched_info		sched_info;
600 
601 	struct list_head		tasks;
602 #ifdef CONFIG_SMP
603 	struct plist_node		pushable_tasks;
604 	struct rb_node			pushable_dl_tasks;
605 #endif
606 
607 	struct mm_struct		*mm;
608 	struct mm_struct		*active_mm;
609 
610 	/* Per-thread vma caching: */
611 	struct vmacache			vmacache;
612 
613 #ifdef SPLIT_RSS_COUNTING
614 	struct task_rss_stat		rss_stat;
615 #endif
616 	int				exit_state;
617 	int				exit_code;
618 	int				exit_signal;
619 	/* The signal sent when the parent dies: */
620 	int				pdeath_signal;
621 	/* JOBCTL_*, siglock protected: */
622 	unsigned long			jobctl;
623 
624 	/* Used for emulating ABI behavior of previous Linux versions: */
625 	unsigned int			personality;
626 
627 	/* Scheduler bits, serialized by scheduler locks: */
628 	unsigned			sched_reset_on_fork:1;
629 	unsigned			sched_contributes_to_load:1;
630 	unsigned			sched_migrated:1;
631 	unsigned			sched_remote_wakeup:1;
632 	/* Force alignment to the next boundary: */
633 	unsigned			:0;
634 
635 	/* Unserialized, strictly 'current' */
636 
637 	/* Bit to tell LSMs we're in execve(): */
638 	unsigned			in_execve:1;
639 	unsigned			in_iowait:1;
640 #ifndef TIF_RESTORE_SIGMASK
641 	unsigned			restore_sigmask:1;
642 #endif
643 #ifdef CONFIG_MEMCG
644 	unsigned			memcg_may_oom:1;
645 #ifndef CONFIG_SLOB
646 	unsigned			memcg_kmem_skip_account:1;
647 #endif
648 #endif
649 #ifdef CONFIG_COMPAT_BRK
650 	unsigned			brk_randomized:1;
651 #endif
652 #ifdef CONFIG_CGROUPS
653 	/* disallow userland-initiated cgroup migration */
654 	unsigned			no_cgroup_migration:1;
655 #endif
656 
657 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
658 
659 	struct restart_block		restart_block;
660 
661 	pid_t				pid;
662 	pid_t				tgid;
663 
664 #ifdef CONFIG_CC_STACKPROTECTOR
665 	/* Canary value for the -fstack-protector GCC feature: */
666 	unsigned long			stack_canary;
667 #endif
668 	/*
669 	 * Pointers to the (original) parent process, youngest child, younger sibling,
670 	 * older sibling, respectively.  (p->father can be replaced with
671 	 * p->real_parent->pid)
672 	 */
673 
674 	/* Real parent process: */
675 	struct task_struct __rcu	*real_parent;
676 
677 	/* Recipient of SIGCHLD, wait4() reports: */
678 	struct task_struct __rcu	*parent;
679 
680 	/*
681 	 * Children/sibling form the list of natural children:
682 	 */
683 	struct list_head		children;
684 	struct list_head		sibling;
685 	struct task_struct		*group_leader;
686 
687 	/*
688 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
689 	 *
690 	 * This includes both natural children and PTRACE_ATTACH targets.
691 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
692 	 */
693 	struct list_head		ptraced;
694 	struct list_head		ptrace_entry;
695 
696 	/* PID/PID hash table linkage. */
697 	struct pid_link			pids[PIDTYPE_MAX];
698 	struct list_head		thread_group;
699 	struct list_head		thread_node;
700 
701 	struct completion		*vfork_done;
702 
703 	/* CLONE_CHILD_SETTID: */
704 	int __user			*set_child_tid;
705 
706 	/* CLONE_CHILD_CLEARTID: */
707 	int __user			*clear_child_tid;
708 
709 	u64				utime;
710 	u64				stime;
711 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
712 	u64				utimescaled;
713 	u64				stimescaled;
714 #endif
715 	u64				gtime;
716 	struct prev_cputime		prev_cputime;
717 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
718 	struct vtime			vtime;
719 #endif
720 
721 #ifdef CONFIG_NO_HZ_FULL
722 	atomic_t			tick_dep_mask;
723 #endif
724 	/* Context switch counts: */
725 	unsigned long			nvcsw;
726 	unsigned long			nivcsw;
727 
728 	/* Monotonic time in nsecs: */
729 	u64				start_time;
730 
731 	/* Boot based time in nsecs: */
732 	u64				real_start_time;
733 
734 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
735 	unsigned long			min_flt;
736 	unsigned long			maj_flt;
737 
738 #ifdef CONFIG_POSIX_TIMERS
739 	struct task_cputime		cputime_expires;
740 	struct list_head		cpu_timers[3];
741 #endif
742 
743 	/* Process credentials: */
744 
745 	/* Tracer's credentials at attach: */
746 	const struct cred __rcu		*ptracer_cred;
747 
748 	/* Objective and real subjective task credentials (COW): */
749 	const struct cred __rcu		*real_cred;
750 
751 	/* Effective (overridable) subjective task credentials (COW): */
752 	const struct cred __rcu		*cred;
753 
754 	/*
755 	 * executable name, excluding path.
756 	 *
757 	 * - normally initialized setup_new_exec()
758 	 * - access it with [gs]et_task_comm()
759 	 * - lock it with task_lock()
760 	 */
761 	char				comm[TASK_COMM_LEN];
762 
763 	struct nameidata		*nameidata;
764 
765 #ifdef CONFIG_SYSVIPC
766 	struct sysv_sem			sysvsem;
767 	struct sysv_shm			sysvshm;
768 #endif
769 #ifdef CONFIG_DETECT_HUNG_TASK
770 	unsigned long			last_switch_count;
771 #endif
772 	/* Filesystem information: */
773 	struct fs_struct		*fs;
774 
775 	/* Open file information: */
776 	struct files_struct		*files;
777 
778 	/* Namespaces: */
779 	struct nsproxy			*nsproxy;
780 
781 	/* Signal handlers: */
782 	struct signal_struct		*signal;
783 	struct sighand_struct		*sighand;
784 	sigset_t			blocked;
785 	sigset_t			real_blocked;
786 	/* Restored if set_restore_sigmask() was used: */
787 	sigset_t			saved_sigmask;
788 	struct sigpending		pending;
789 	unsigned long			sas_ss_sp;
790 	size_t				sas_ss_size;
791 	unsigned int			sas_ss_flags;
792 
793 	struct callback_head		*task_works;
794 
795 	struct audit_context		*audit_context;
796 #ifdef CONFIG_AUDITSYSCALL
797 	kuid_t				loginuid;
798 	unsigned int			sessionid;
799 #endif
800 	struct seccomp			seccomp;
801 
802 	/* Thread group tracking: */
803 	u32				parent_exec_id;
804 	u32				self_exec_id;
805 
806 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
807 	spinlock_t			alloc_lock;
808 
809 	/* Protection of the PI data structures: */
810 	raw_spinlock_t			pi_lock;
811 
812 	struct wake_q_node		wake_q;
813 
814 #ifdef CONFIG_RT_MUTEXES
815 	/* PI waiters blocked on a rt_mutex held by this task: */
816 	struct rb_root_cached		pi_waiters;
817 	/* Updated under owner's pi_lock and rq lock */
818 	struct task_struct		*pi_top_task;
819 	/* Deadlock detection and priority inheritance handling: */
820 	struct rt_mutex_waiter		*pi_blocked_on;
821 #endif
822 
823 #ifdef CONFIG_DEBUG_MUTEXES
824 	/* Mutex deadlock detection: */
825 	struct mutex_waiter		*blocked_on;
826 #endif
827 
828 #ifdef CONFIG_TRACE_IRQFLAGS
829 	unsigned int			irq_events;
830 	unsigned long			hardirq_enable_ip;
831 	unsigned long			hardirq_disable_ip;
832 	unsigned int			hardirq_enable_event;
833 	unsigned int			hardirq_disable_event;
834 	int				hardirqs_enabled;
835 	int				hardirq_context;
836 	unsigned long			softirq_disable_ip;
837 	unsigned long			softirq_enable_ip;
838 	unsigned int			softirq_disable_event;
839 	unsigned int			softirq_enable_event;
840 	int				softirqs_enabled;
841 	int				softirq_context;
842 #endif
843 
844 #ifdef CONFIG_LOCKDEP
845 # define MAX_LOCK_DEPTH			48UL
846 	u64				curr_chain_key;
847 	int				lockdep_depth;
848 	unsigned int			lockdep_recursion;
849 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
850 #endif
851 
852 #ifdef CONFIG_LOCKDEP_CROSSRELEASE
853 #define MAX_XHLOCKS_NR 64UL
854 	struct hist_lock *xhlocks; /* Crossrelease history locks */
855 	unsigned int xhlock_idx;
856 	/* For restoring at history boundaries */
857 	unsigned int xhlock_idx_hist[XHLOCK_CTX_NR];
858 	unsigned int hist_id;
859 	/* For overwrite check at each context exit */
860 	unsigned int hist_id_save[XHLOCK_CTX_NR];
861 #endif
862 
863 #ifdef CONFIG_UBSAN
864 	unsigned int			in_ubsan;
865 #endif
866 
867 	/* Journalling filesystem info: */
868 	void				*journal_info;
869 
870 	/* Stacked block device info: */
871 	struct bio_list			*bio_list;
872 
873 #ifdef CONFIG_BLOCK
874 	/* Stack plugging: */
875 	struct blk_plug			*plug;
876 #endif
877 
878 	/* VM state: */
879 	struct reclaim_state		*reclaim_state;
880 
881 	struct backing_dev_info		*backing_dev_info;
882 
883 	struct io_context		*io_context;
884 
885 	/* Ptrace state: */
886 	unsigned long			ptrace_message;
887 	siginfo_t			*last_siginfo;
888 
889 	struct task_io_accounting	ioac;
890 #ifdef CONFIG_TASK_XACCT
891 	/* Accumulated RSS usage: */
892 	u64				acct_rss_mem1;
893 	/* Accumulated virtual memory usage: */
894 	u64				acct_vm_mem1;
895 	/* stime + utime since last update: */
896 	u64				acct_timexpd;
897 #endif
898 #ifdef CONFIG_CPUSETS
899 	/* Protected by ->alloc_lock: */
900 	nodemask_t			mems_allowed;
901 	/* Seqence number to catch updates: */
902 	seqcount_t			mems_allowed_seq;
903 	int				cpuset_mem_spread_rotor;
904 	int				cpuset_slab_spread_rotor;
905 #endif
906 #ifdef CONFIG_CGROUPS
907 	/* Control Group info protected by css_set_lock: */
908 	struct css_set __rcu		*cgroups;
909 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
910 	struct list_head		cg_list;
911 #endif
912 #ifdef CONFIG_INTEL_RDT
913 	u32				closid;
914 	u32				rmid;
915 #endif
916 #ifdef CONFIG_FUTEX
917 	struct robust_list_head __user	*robust_list;
918 #ifdef CONFIG_COMPAT
919 	struct compat_robust_list_head __user *compat_robust_list;
920 #endif
921 	struct list_head		pi_state_list;
922 	struct futex_pi_state		*pi_state_cache;
923 #endif
924 #ifdef CONFIG_PERF_EVENTS
925 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
926 	struct mutex			perf_event_mutex;
927 	struct list_head		perf_event_list;
928 #endif
929 #ifdef CONFIG_DEBUG_PREEMPT
930 	unsigned long			preempt_disable_ip;
931 #endif
932 #ifdef CONFIG_NUMA
933 	/* Protected by alloc_lock: */
934 	struct mempolicy		*mempolicy;
935 	short				il_prev;
936 	short				pref_node_fork;
937 #endif
938 #ifdef CONFIG_NUMA_BALANCING
939 	int				numa_scan_seq;
940 	unsigned int			numa_scan_period;
941 	unsigned int			numa_scan_period_max;
942 	int				numa_preferred_nid;
943 	unsigned long			numa_migrate_retry;
944 	/* Migration stamp: */
945 	u64				node_stamp;
946 	u64				last_task_numa_placement;
947 	u64				last_sum_exec_runtime;
948 	struct callback_head		numa_work;
949 
950 	struct list_head		numa_entry;
951 	struct numa_group		*numa_group;
952 
953 	/*
954 	 * numa_faults is an array split into four regions:
955 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
956 	 * in this precise order.
957 	 *
958 	 * faults_memory: Exponential decaying average of faults on a per-node
959 	 * basis. Scheduling placement decisions are made based on these
960 	 * counts. The values remain static for the duration of a PTE scan.
961 	 * faults_cpu: Track the nodes the process was running on when a NUMA
962 	 * hinting fault was incurred.
963 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
964 	 * during the current scan window. When the scan completes, the counts
965 	 * in faults_memory and faults_cpu decay and these values are copied.
966 	 */
967 	unsigned long			*numa_faults;
968 	unsigned long			total_numa_faults;
969 
970 	/*
971 	 * numa_faults_locality tracks if faults recorded during the last
972 	 * scan window were remote/local or failed to migrate. The task scan
973 	 * period is adapted based on the locality of the faults with different
974 	 * weights depending on whether they were shared or private faults
975 	 */
976 	unsigned long			numa_faults_locality[3];
977 
978 	unsigned long			numa_pages_migrated;
979 #endif /* CONFIG_NUMA_BALANCING */
980 
981 	struct tlbflush_unmap_batch	tlb_ubc;
982 
983 	struct rcu_head			rcu;
984 
985 	/* Cache last used pipe for splice(): */
986 	struct pipe_inode_info		*splice_pipe;
987 
988 	struct page_frag		task_frag;
989 
990 #ifdef CONFIG_TASK_DELAY_ACCT
991 	struct task_delay_info		*delays;
992 #endif
993 
994 #ifdef CONFIG_FAULT_INJECTION
995 	int				make_it_fail;
996 	unsigned int			fail_nth;
997 #endif
998 	/*
999 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1000 	 * balance_dirty_pages() for a dirty throttling pause:
1001 	 */
1002 	int				nr_dirtied;
1003 	int				nr_dirtied_pause;
1004 	/* Start of a write-and-pause period: */
1005 	unsigned long			dirty_paused_when;
1006 
1007 #ifdef CONFIG_LATENCYTOP
1008 	int				latency_record_count;
1009 	struct latency_record		latency_record[LT_SAVECOUNT];
1010 #endif
1011 	/*
1012 	 * Time slack values; these are used to round up poll() and
1013 	 * select() etc timeout values. These are in nanoseconds.
1014 	 */
1015 	u64				timer_slack_ns;
1016 	u64				default_timer_slack_ns;
1017 
1018 #ifdef CONFIG_KASAN
1019 	unsigned int			kasan_depth;
1020 #endif
1021 
1022 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1023 	/* Index of current stored address in ret_stack: */
1024 	int				curr_ret_stack;
1025 
1026 	/* Stack of return addresses for return function tracing: */
1027 	struct ftrace_ret_stack		*ret_stack;
1028 
1029 	/* Timestamp for last schedule: */
1030 	unsigned long long		ftrace_timestamp;
1031 
1032 	/*
1033 	 * Number of functions that haven't been traced
1034 	 * because of depth overrun:
1035 	 */
1036 	atomic_t			trace_overrun;
1037 
1038 	/* Pause tracing: */
1039 	atomic_t			tracing_graph_pause;
1040 #endif
1041 
1042 #ifdef CONFIG_TRACING
1043 	/* State flags for use by tracers: */
1044 	unsigned long			trace;
1045 
1046 	/* Bitmask and counter of trace recursion: */
1047 	unsigned long			trace_recursion;
1048 #endif /* CONFIG_TRACING */
1049 
1050 #ifdef CONFIG_KCOV
1051 	/* Coverage collection mode enabled for this task (0 if disabled): */
1052 	enum kcov_mode			kcov_mode;
1053 
1054 	/* Size of the kcov_area: */
1055 	unsigned int			kcov_size;
1056 
1057 	/* Buffer for coverage collection: */
1058 	void				*kcov_area;
1059 
1060 	/* KCOV descriptor wired with this task or NULL: */
1061 	struct kcov			*kcov;
1062 #endif
1063 
1064 #ifdef CONFIG_MEMCG
1065 	struct mem_cgroup		*memcg_in_oom;
1066 	gfp_t				memcg_oom_gfp_mask;
1067 	int				memcg_oom_order;
1068 
1069 	/* Number of pages to reclaim on returning to userland: */
1070 	unsigned int			memcg_nr_pages_over_high;
1071 #endif
1072 
1073 #ifdef CONFIG_UPROBES
1074 	struct uprobe_task		*utask;
1075 #endif
1076 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1077 	unsigned int			sequential_io;
1078 	unsigned int			sequential_io_avg;
1079 #endif
1080 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1081 	unsigned long			task_state_change;
1082 #endif
1083 	int				pagefault_disabled;
1084 #ifdef CONFIG_MMU
1085 	struct task_struct		*oom_reaper_list;
1086 #endif
1087 #ifdef CONFIG_VMAP_STACK
1088 	struct vm_struct		*stack_vm_area;
1089 #endif
1090 #ifdef CONFIG_THREAD_INFO_IN_TASK
1091 	/* A live task holds one reference: */
1092 	atomic_t			stack_refcount;
1093 #endif
1094 #ifdef CONFIG_LIVEPATCH
1095 	int patch_state;
1096 #endif
1097 #ifdef CONFIG_SECURITY
1098 	/* Used by LSM modules for access restriction: */
1099 	void				*security;
1100 #endif
1101 
1102 	/*
1103 	 * New fields for task_struct should be added above here, so that
1104 	 * they are included in the randomized portion of task_struct.
1105 	 */
1106 	randomized_struct_fields_end
1107 
1108 	/* CPU-specific state of this task: */
1109 	struct thread_struct		thread;
1110 
1111 	/*
1112 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1113 	 * structure.  It *MUST* be at the end of 'task_struct'.
1114 	 *
1115 	 * Do not put anything below here!
1116 	 */
1117 };
1118 
1119 static inline struct pid *task_pid(struct task_struct *task)
1120 {
1121 	return task->pids[PIDTYPE_PID].pid;
1122 }
1123 
1124 static inline struct pid *task_tgid(struct task_struct *task)
1125 {
1126 	return task->group_leader->pids[PIDTYPE_PID].pid;
1127 }
1128 
1129 /*
1130  * Without tasklist or RCU lock it is not safe to dereference
1131  * the result of task_pgrp/task_session even if task == current,
1132  * we can race with another thread doing sys_setsid/sys_setpgid.
1133  */
1134 static inline struct pid *task_pgrp(struct task_struct *task)
1135 {
1136 	return task->group_leader->pids[PIDTYPE_PGID].pid;
1137 }
1138 
1139 static inline struct pid *task_session(struct task_struct *task)
1140 {
1141 	return task->group_leader->pids[PIDTYPE_SID].pid;
1142 }
1143 
1144 /*
1145  * the helpers to get the task's different pids as they are seen
1146  * from various namespaces
1147  *
1148  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1149  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1150  *                     current.
1151  * task_xid_nr_ns()  : id seen from the ns specified;
1152  *
1153  * see also pid_nr() etc in include/linux/pid.h
1154  */
1155 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1156 
1157 static inline pid_t task_pid_nr(struct task_struct *tsk)
1158 {
1159 	return tsk->pid;
1160 }
1161 
1162 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1163 {
1164 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1165 }
1166 
1167 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1168 {
1169 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1170 }
1171 
1172 
1173 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1174 {
1175 	return tsk->tgid;
1176 }
1177 
1178 /**
1179  * pid_alive - check that a task structure is not stale
1180  * @p: Task structure to be checked.
1181  *
1182  * Test if a process is not yet dead (at most zombie state)
1183  * If pid_alive fails, then pointers within the task structure
1184  * can be stale and must not be dereferenced.
1185  *
1186  * Return: 1 if the process is alive. 0 otherwise.
1187  */
1188 static inline int pid_alive(const struct task_struct *p)
1189 {
1190 	return p->pids[PIDTYPE_PID].pid != NULL;
1191 }
1192 
1193 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1194 {
1195 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1196 }
1197 
1198 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1199 {
1200 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1201 }
1202 
1203 
1204 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1205 {
1206 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1207 }
1208 
1209 static inline pid_t task_session_vnr(struct task_struct *tsk)
1210 {
1211 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1212 }
1213 
1214 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1215 {
1216 	return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
1217 }
1218 
1219 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1220 {
1221 	return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
1222 }
1223 
1224 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1225 {
1226 	pid_t pid = 0;
1227 
1228 	rcu_read_lock();
1229 	if (pid_alive(tsk))
1230 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1231 	rcu_read_unlock();
1232 
1233 	return pid;
1234 }
1235 
1236 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1237 {
1238 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1239 }
1240 
1241 /* Obsolete, do not use: */
1242 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1243 {
1244 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1245 }
1246 
1247 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1248 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1249 
1250 static inline unsigned int task_state_index(struct task_struct *tsk)
1251 {
1252 	unsigned int tsk_state = READ_ONCE(tsk->state);
1253 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1254 
1255 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1256 
1257 	if (tsk_state == TASK_IDLE)
1258 		state = TASK_REPORT_IDLE;
1259 
1260 	return fls(state);
1261 }
1262 
1263 static inline char task_index_to_char(unsigned int state)
1264 {
1265 	static const char state_char[] = "RSDTtXZPI";
1266 
1267 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1268 
1269 	return state_char[state];
1270 }
1271 
1272 static inline char task_state_to_char(struct task_struct *tsk)
1273 {
1274 	return task_index_to_char(task_state_index(tsk));
1275 }
1276 
1277 /**
1278  * is_global_init - check if a task structure is init. Since init
1279  * is free to have sub-threads we need to check tgid.
1280  * @tsk: Task structure to be checked.
1281  *
1282  * Check if a task structure is the first user space task the kernel created.
1283  *
1284  * Return: 1 if the task structure is init. 0 otherwise.
1285  */
1286 static inline int is_global_init(struct task_struct *tsk)
1287 {
1288 	return task_tgid_nr(tsk) == 1;
1289 }
1290 
1291 extern struct pid *cad_pid;
1292 
1293 /*
1294  * Per process flags
1295  */
1296 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1297 #define PF_EXITING		0x00000004	/* Getting shut down */
1298 #define PF_EXITPIDONE		0x00000008	/* PI exit done on shut down */
1299 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1300 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1301 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1302 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1303 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1304 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1305 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1306 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1307 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1308 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1309 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1310 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1311 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1312 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1313 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1314 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1315 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1316 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1317 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1318 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1319 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_allowed */
1320 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1321 #define PF_MUTEX_TESTER		0x20000000	/* Thread belongs to the rt mutex tester */
1322 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1323 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1324 
1325 /*
1326  * Only the _current_ task can read/write to tsk->flags, but other
1327  * tasks can access tsk->flags in readonly mode for example
1328  * with tsk_used_math (like during threaded core dumping).
1329  * There is however an exception to this rule during ptrace
1330  * or during fork: the ptracer task is allowed to write to the
1331  * child->flags of its traced child (same goes for fork, the parent
1332  * can write to the child->flags), because we're guaranteed the
1333  * child is not running and in turn not changing child->flags
1334  * at the same time the parent does it.
1335  */
1336 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1337 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1338 #define clear_used_math()			clear_stopped_child_used_math(current)
1339 #define set_used_math()				set_stopped_child_used_math(current)
1340 
1341 #define conditional_stopped_child_used_math(condition, child) \
1342 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1343 
1344 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1345 
1346 #define copy_to_stopped_child_used_math(child) \
1347 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1348 
1349 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1350 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1351 #define used_math()				tsk_used_math(current)
1352 
1353 static inline bool is_percpu_thread(void)
1354 {
1355 #ifdef CONFIG_SMP
1356 	return (current->flags & PF_NO_SETAFFINITY) &&
1357 		(current->nr_cpus_allowed  == 1);
1358 #else
1359 	return true;
1360 #endif
1361 }
1362 
1363 /* Per-process atomic flags. */
1364 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1365 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1366 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1367 
1368 
1369 #define TASK_PFA_TEST(name, func)					\
1370 	static inline bool task_##func(struct task_struct *p)		\
1371 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1372 
1373 #define TASK_PFA_SET(name, func)					\
1374 	static inline void task_set_##func(struct task_struct *p)	\
1375 	{ set_bit(PFA_##name, &p->atomic_flags); }
1376 
1377 #define TASK_PFA_CLEAR(name, func)					\
1378 	static inline void task_clear_##func(struct task_struct *p)	\
1379 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1380 
1381 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1382 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1383 
1384 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1385 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1386 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1387 
1388 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1389 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1390 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1391 
1392 static inline void
1393 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1394 {
1395 	current->flags &= ~flags;
1396 	current->flags |= orig_flags & flags;
1397 }
1398 
1399 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1400 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1401 #ifdef CONFIG_SMP
1402 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1403 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1404 #else
1405 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1406 {
1407 }
1408 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1409 {
1410 	if (!cpumask_test_cpu(0, new_mask))
1411 		return -EINVAL;
1412 	return 0;
1413 }
1414 #endif
1415 
1416 #ifndef cpu_relax_yield
1417 #define cpu_relax_yield() cpu_relax()
1418 #endif
1419 
1420 extern int yield_to(struct task_struct *p, bool preempt);
1421 extern void set_user_nice(struct task_struct *p, long nice);
1422 extern int task_prio(const struct task_struct *p);
1423 
1424 /**
1425  * task_nice - return the nice value of a given task.
1426  * @p: the task in question.
1427  *
1428  * Return: The nice value [ -20 ... 0 ... 19 ].
1429  */
1430 static inline int task_nice(const struct task_struct *p)
1431 {
1432 	return PRIO_TO_NICE((p)->static_prio);
1433 }
1434 
1435 extern int can_nice(const struct task_struct *p, const int nice);
1436 extern int task_curr(const struct task_struct *p);
1437 extern int idle_cpu(int cpu);
1438 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1439 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1440 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1441 extern struct task_struct *idle_task(int cpu);
1442 
1443 /**
1444  * is_idle_task - is the specified task an idle task?
1445  * @p: the task in question.
1446  *
1447  * Return: 1 if @p is an idle task. 0 otherwise.
1448  */
1449 static inline bool is_idle_task(const struct task_struct *p)
1450 {
1451 	return !!(p->flags & PF_IDLE);
1452 }
1453 
1454 extern struct task_struct *curr_task(int cpu);
1455 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1456 
1457 void yield(void);
1458 
1459 union thread_union {
1460 #ifndef CONFIG_THREAD_INFO_IN_TASK
1461 	struct thread_info thread_info;
1462 #endif
1463 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1464 };
1465 
1466 #ifdef CONFIG_THREAD_INFO_IN_TASK
1467 static inline struct thread_info *task_thread_info(struct task_struct *task)
1468 {
1469 	return &task->thread_info;
1470 }
1471 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1472 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1473 #endif
1474 
1475 /*
1476  * find a task by one of its numerical ids
1477  *
1478  * find_task_by_pid_ns():
1479  *      finds a task by its pid in the specified namespace
1480  * find_task_by_vpid():
1481  *      finds a task by its virtual pid
1482  *
1483  * see also find_vpid() etc in include/linux/pid.h
1484  */
1485 
1486 extern struct task_struct *find_task_by_vpid(pid_t nr);
1487 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1488 
1489 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1490 extern int wake_up_process(struct task_struct *tsk);
1491 extern void wake_up_new_task(struct task_struct *tsk);
1492 
1493 #ifdef CONFIG_SMP
1494 extern void kick_process(struct task_struct *tsk);
1495 #else
1496 static inline void kick_process(struct task_struct *tsk) { }
1497 #endif
1498 
1499 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1500 
1501 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1502 {
1503 	__set_task_comm(tsk, from, false);
1504 }
1505 
1506 extern char *get_task_comm(char *to, struct task_struct *tsk);
1507 
1508 #ifdef CONFIG_SMP
1509 void scheduler_ipi(void);
1510 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1511 #else
1512 static inline void scheduler_ipi(void) { }
1513 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1514 {
1515 	return 1;
1516 }
1517 #endif
1518 
1519 /*
1520  * Set thread flags in other task's structures.
1521  * See asm/thread_info.h for TIF_xxxx flags available:
1522  */
1523 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1524 {
1525 	set_ti_thread_flag(task_thread_info(tsk), flag);
1526 }
1527 
1528 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1529 {
1530 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1531 }
1532 
1533 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1534 {
1535 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1536 }
1537 
1538 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1539 {
1540 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1541 }
1542 
1543 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1544 {
1545 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1546 }
1547 
1548 static inline void set_tsk_need_resched(struct task_struct *tsk)
1549 {
1550 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1551 }
1552 
1553 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1554 {
1555 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1556 }
1557 
1558 static inline int test_tsk_need_resched(struct task_struct *tsk)
1559 {
1560 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1561 }
1562 
1563 /*
1564  * cond_resched() and cond_resched_lock(): latency reduction via
1565  * explicit rescheduling in places that are safe. The return
1566  * value indicates whether a reschedule was done in fact.
1567  * cond_resched_lock() will drop the spinlock before scheduling,
1568  * cond_resched_softirq() will enable bhs before scheduling.
1569  */
1570 #ifndef CONFIG_PREEMPT
1571 extern int _cond_resched(void);
1572 #else
1573 static inline int _cond_resched(void) { return 0; }
1574 #endif
1575 
1576 #define cond_resched() ({			\
1577 	___might_sleep(__FILE__, __LINE__, 0);	\
1578 	_cond_resched();			\
1579 })
1580 
1581 extern int __cond_resched_lock(spinlock_t *lock);
1582 
1583 #define cond_resched_lock(lock) ({				\
1584 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1585 	__cond_resched_lock(lock);				\
1586 })
1587 
1588 extern int __cond_resched_softirq(void);
1589 
1590 #define cond_resched_softirq() ({					\
1591 	___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET);	\
1592 	__cond_resched_softirq();					\
1593 })
1594 
1595 static inline void cond_resched_rcu(void)
1596 {
1597 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1598 	rcu_read_unlock();
1599 	cond_resched();
1600 	rcu_read_lock();
1601 #endif
1602 }
1603 
1604 /*
1605  * Does a critical section need to be broken due to another
1606  * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1607  * but a general need for low latency)
1608  */
1609 static inline int spin_needbreak(spinlock_t *lock)
1610 {
1611 #ifdef CONFIG_PREEMPT
1612 	return spin_is_contended(lock);
1613 #else
1614 	return 0;
1615 #endif
1616 }
1617 
1618 static __always_inline bool need_resched(void)
1619 {
1620 	return unlikely(tif_need_resched());
1621 }
1622 
1623 /*
1624  * Wrappers for p->thread_info->cpu access. No-op on UP.
1625  */
1626 #ifdef CONFIG_SMP
1627 
1628 static inline unsigned int task_cpu(const struct task_struct *p)
1629 {
1630 #ifdef CONFIG_THREAD_INFO_IN_TASK
1631 	return p->cpu;
1632 #else
1633 	return task_thread_info(p)->cpu;
1634 #endif
1635 }
1636 
1637 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1638 
1639 #else
1640 
1641 static inline unsigned int task_cpu(const struct task_struct *p)
1642 {
1643 	return 0;
1644 }
1645 
1646 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1647 {
1648 }
1649 
1650 #endif /* CONFIG_SMP */
1651 
1652 /*
1653  * In order to reduce various lock holder preemption latencies provide an
1654  * interface to see if a vCPU is currently running or not.
1655  *
1656  * This allows us to terminate optimistic spin loops and block, analogous to
1657  * the native optimistic spin heuristic of testing if the lock owner task is
1658  * running or not.
1659  */
1660 #ifndef vcpu_is_preempted
1661 # define vcpu_is_preempted(cpu)	false
1662 #endif
1663 
1664 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1665 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1666 
1667 #ifndef TASK_SIZE_OF
1668 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1669 #endif
1670 
1671 #endif
1672