xref: /linux/include/linux/sched.h (revision fca3aa16642200069eafa4ece17a60751bb891cd)
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 
97 /* get_task_state(): */
98 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
99 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
100 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
101 					 TASK_PARKED)
102 
103 #define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
104 
105 #define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
106 
107 #define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
108 
109 #define task_contributes_to_load(task)	((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
110 					 (task->flags & PF_FROZEN) == 0 && \
111 					 (task->state & TASK_NOLOAD) == 0)
112 
113 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
114 
115 #define __set_current_state(state_value)			\
116 	do {							\
117 		current->task_state_change = _THIS_IP_;		\
118 		current->state = (state_value);			\
119 	} while (0)
120 #define set_current_state(state_value)				\
121 	do {							\
122 		current->task_state_change = _THIS_IP_;		\
123 		smp_store_mb(current->state, (state_value));	\
124 	} while (0)
125 
126 #else
127 /*
128  * set_current_state() includes a barrier so that the write of current->state
129  * is correctly serialised wrt the caller's subsequent test of whether to
130  * actually sleep:
131  *
132  *   for (;;) {
133  *	set_current_state(TASK_UNINTERRUPTIBLE);
134  *	if (!need_sleep)
135  *		break;
136  *
137  *	schedule();
138  *   }
139  *   __set_current_state(TASK_RUNNING);
140  *
141  * If the caller does not need such serialisation (because, for instance, the
142  * condition test and condition change and wakeup are under the same lock) then
143  * use __set_current_state().
144  *
145  * The above is typically ordered against the wakeup, which does:
146  *
147  *	need_sleep = false;
148  *	wake_up_state(p, TASK_UNINTERRUPTIBLE);
149  *
150  * Where wake_up_state() (and all other wakeup primitives) imply enough
151  * barriers to order the store of the variable against wakeup.
152  *
153  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
154  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
155  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
156  *
157  * This is obviously fine, since they both store the exact same value.
158  *
159  * Also see the comments of try_to_wake_up().
160  */
161 #define __set_current_state(state_value) do { current->state = (state_value); } while (0)
162 #define set_current_state(state_value)	 smp_store_mb(current->state, (state_value))
163 #endif
164 
165 /* Task command name length: */
166 #define TASK_COMM_LEN			16
167 
168 extern void scheduler_tick(void);
169 
170 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
171 
172 extern long schedule_timeout(long timeout);
173 extern long schedule_timeout_interruptible(long timeout);
174 extern long schedule_timeout_killable(long timeout);
175 extern long schedule_timeout_uninterruptible(long timeout);
176 extern long schedule_timeout_idle(long timeout);
177 asmlinkage void schedule(void);
178 extern void schedule_preempt_disabled(void);
179 
180 extern int __must_check io_schedule_prepare(void);
181 extern void io_schedule_finish(int token);
182 extern long io_schedule_timeout(long timeout);
183 extern void io_schedule(void);
184 
185 /**
186  * struct prev_cputime - snapshot of system and user cputime
187  * @utime: time spent in user mode
188  * @stime: time spent in system mode
189  * @lock: protects the above two fields
190  *
191  * Stores previous user/system time values such that we can guarantee
192  * monotonicity.
193  */
194 struct prev_cputime {
195 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
196 	u64				utime;
197 	u64				stime;
198 	raw_spinlock_t			lock;
199 #endif
200 };
201 
202 /**
203  * struct task_cputime - collected CPU time counts
204  * @utime:		time spent in user mode, in nanoseconds
205  * @stime:		time spent in kernel mode, in nanoseconds
206  * @sum_exec_runtime:	total time spent on the CPU, in nanoseconds
207  *
208  * This structure groups together three kinds of CPU time that are tracked for
209  * threads and thread groups.  Most things considering CPU time want to group
210  * these counts together and treat all three of them in parallel.
211  */
212 struct task_cputime {
213 	u64				utime;
214 	u64				stime;
215 	unsigned long long		sum_exec_runtime;
216 };
217 
218 /* Alternate field names when used on cache expirations: */
219 #define virt_exp			utime
220 #define prof_exp			stime
221 #define sched_exp			sum_exec_runtime
222 
223 enum vtime_state {
224 	/* Task is sleeping or running in a CPU with VTIME inactive: */
225 	VTIME_INACTIVE = 0,
226 	/* Task runs in userspace in a CPU with VTIME active: */
227 	VTIME_USER,
228 	/* Task runs in kernelspace in a CPU with VTIME active: */
229 	VTIME_SYS,
230 };
231 
232 struct vtime {
233 	seqcount_t		seqcount;
234 	unsigned long long	starttime;
235 	enum vtime_state	state;
236 	u64			utime;
237 	u64			stime;
238 	u64			gtime;
239 };
240 
241 struct sched_info {
242 #ifdef CONFIG_SCHED_INFO
243 	/* Cumulative counters: */
244 
245 	/* # of times we have run on this CPU: */
246 	unsigned long			pcount;
247 
248 	/* Time spent waiting on a runqueue: */
249 	unsigned long long		run_delay;
250 
251 	/* Timestamps: */
252 
253 	/* When did we last run on a CPU? */
254 	unsigned long long		last_arrival;
255 
256 	/* When were we last queued to run? */
257 	unsigned long long		last_queued;
258 
259 #endif /* CONFIG_SCHED_INFO */
260 };
261 
262 /*
263  * Integer metrics need fixed point arithmetic, e.g., sched/fair
264  * has a few: load, load_avg, util_avg, freq, and capacity.
265  *
266  * We define a basic fixed point arithmetic range, and then formalize
267  * all these metrics based on that basic range.
268  */
269 # define SCHED_FIXEDPOINT_SHIFT		10
270 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
271 
272 struct load_weight {
273 	unsigned long			weight;
274 	u32				inv_weight;
275 };
276 
277 /**
278  * struct util_est - Estimation utilization of FAIR tasks
279  * @enqueued: instantaneous estimated utilization of a task/cpu
280  * @ewma:     the Exponential Weighted Moving Average (EWMA)
281  *            utilization of a task
282  *
283  * Support data structure to track an Exponential Weighted Moving Average
284  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
285  * average each time a task completes an activation. Sample's weight is chosen
286  * so that the EWMA will be relatively insensitive to transient changes to the
287  * task's workload.
288  *
289  * The enqueued attribute has a slightly different meaning for tasks and cpus:
290  * - task:   the task's util_avg at last task dequeue time
291  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
292  * Thus, the util_est.enqueued of a task represents the contribution on the
293  * estimated utilization of the CPU where that task is currently enqueued.
294  *
295  * Only for tasks we track a moving average of the past instantaneous
296  * estimated utilization. This allows to absorb sporadic drops in utilization
297  * of an otherwise almost periodic task.
298  */
299 struct util_est {
300 	unsigned int			enqueued;
301 	unsigned int			ewma;
302 #define UTIL_EST_WEIGHT_SHIFT		2
303 } __attribute__((__aligned__(sizeof(u64))));
304 
305 /*
306  * The load_avg/util_avg accumulates an infinite geometric series
307  * (see __update_load_avg() in kernel/sched/fair.c).
308  *
309  * [load_avg definition]
310  *
311  *   load_avg = runnable% * scale_load_down(load)
312  *
313  * where runnable% is the time ratio that a sched_entity is runnable.
314  * For cfs_rq, it is the aggregated load_avg of all runnable and
315  * blocked sched_entities.
316  *
317  * load_avg may also take frequency scaling into account:
318  *
319  *   load_avg = runnable% * scale_load_down(load) * freq%
320  *
321  * where freq% is the CPU frequency normalized to the highest frequency.
322  *
323  * [util_avg definition]
324  *
325  *   util_avg = running% * SCHED_CAPACITY_SCALE
326  *
327  * where running% is the time ratio that a sched_entity is running on
328  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
329  * and blocked sched_entities.
330  *
331  * util_avg may also factor frequency scaling and CPU capacity scaling:
332  *
333  *   util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
334  *
335  * where freq% is the same as above, and capacity% is the CPU capacity
336  * normalized to the greatest capacity (due to uarch differences, etc).
337  *
338  * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
339  * themselves are in the range of [0, 1]. To do fixed point arithmetics,
340  * we therefore scale them to as large a range as necessary. This is for
341  * example reflected by util_avg's SCHED_CAPACITY_SCALE.
342  *
343  * [Overflow issue]
344  *
345  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
346  * with the highest load (=88761), always runnable on a single cfs_rq,
347  * and should not overflow as the number already hits PID_MAX_LIMIT.
348  *
349  * For all other cases (including 32-bit kernels), struct load_weight's
350  * weight will overflow first before we do, because:
351  *
352  *    Max(load_avg) <= Max(load.weight)
353  *
354  * Then it is the load_weight's responsibility to consider overflow
355  * issues.
356  */
357 struct sched_avg {
358 	u64				last_update_time;
359 	u64				load_sum;
360 	u64				runnable_load_sum;
361 	u32				util_sum;
362 	u32				period_contrib;
363 	unsigned long			load_avg;
364 	unsigned long			runnable_load_avg;
365 	unsigned long			util_avg;
366 	struct util_est			util_est;
367 } ____cacheline_aligned;
368 
369 struct sched_statistics {
370 #ifdef CONFIG_SCHEDSTATS
371 	u64				wait_start;
372 	u64				wait_max;
373 	u64				wait_count;
374 	u64				wait_sum;
375 	u64				iowait_count;
376 	u64				iowait_sum;
377 
378 	u64				sleep_start;
379 	u64				sleep_max;
380 	s64				sum_sleep_runtime;
381 
382 	u64				block_start;
383 	u64				block_max;
384 	u64				exec_max;
385 	u64				slice_max;
386 
387 	u64				nr_migrations_cold;
388 	u64				nr_failed_migrations_affine;
389 	u64				nr_failed_migrations_running;
390 	u64				nr_failed_migrations_hot;
391 	u64				nr_forced_migrations;
392 
393 	u64				nr_wakeups;
394 	u64				nr_wakeups_sync;
395 	u64				nr_wakeups_migrate;
396 	u64				nr_wakeups_local;
397 	u64				nr_wakeups_remote;
398 	u64				nr_wakeups_affine;
399 	u64				nr_wakeups_affine_attempts;
400 	u64				nr_wakeups_passive;
401 	u64				nr_wakeups_idle;
402 #endif
403 };
404 
405 struct sched_entity {
406 	/* For load-balancing: */
407 	struct load_weight		load;
408 	unsigned long			runnable_weight;
409 	struct rb_node			run_node;
410 	struct list_head		group_node;
411 	unsigned int			on_rq;
412 
413 	u64				exec_start;
414 	u64				sum_exec_runtime;
415 	u64				vruntime;
416 	u64				prev_sum_exec_runtime;
417 
418 	u64				nr_migrations;
419 
420 	struct sched_statistics		statistics;
421 
422 #ifdef CONFIG_FAIR_GROUP_SCHED
423 	int				depth;
424 	struct sched_entity		*parent;
425 	/* rq on which this entity is (to be) queued: */
426 	struct cfs_rq			*cfs_rq;
427 	/* rq "owned" by this entity/group: */
428 	struct cfs_rq			*my_q;
429 #endif
430 
431 #ifdef CONFIG_SMP
432 	/*
433 	 * Per entity load average tracking.
434 	 *
435 	 * Put into separate cache line so it does not
436 	 * collide with read-mostly values above.
437 	 */
438 	struct sched_avg		avg;
439 #endif
440 };
441 
442 struct sched_rt_entity {
443 	struct list_head		run_list;
444 	unsigned long			timeout;
445 	unsigned long			watchdog_stamp;
446 	unsigned int			time_slice;
447 	unsigned short			on_rq;
448 	unsigned short			on_list;
449 
450 	struct sched_rt_entity		*back;
451 #ifdef CONFIG_RT_GROUP_SCHED
452 	struct sched_rt_entity		*parent;
453 	/* rq on which this entity is (to be) queued: */
454 	struct rt_rq			*rt_rq;
455 	/* rq "owned" by this entity/group: */
456 	struct rt_rq			*my_q;
457 #endif
458 } __randomize_layout;
459 
460 struct sched_dl_entity {
461 	struct rb_node			rb_node;
462 
463 	/*
464 	 * Original scheduling parameters. Copied here from sched_attr
465 	 * during sched_setattr(), they will remain the same until
466 	 * the next sched_setattr().
467 	 */
468 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
469 	u64				dl_deadline;	/* Relative deadline of each instance	*/
470 	u64				dl_period;	/* Separation of two instances (period) */
471 	u64				dl_bw;		/* dl_runtime / dl_period		*/
472 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
473 
474 	/*
475 	 * Actual scheduling parameters. Initialized with the values above,
476 	 * they are continously updated during task execution. Note that
477 	 * the remaining runtime could be < 0 in case we are in overrun.
478 	 */
479 	s64				runtime;	/* Remaining runtime for this instance	*/
480 	u64				deadline;	/* Absolute deadline for this instance	*/
481 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
482 
483 	/*
484 	 * Some bool flags:
485 	 *
486 	 * @dl_throttled tells if we exhausted the runtime. If so, the
487 	 * task has to wait for a replenishment to be performed at the
488 	 * next firing of dl_timer.
489 	 *
490 	 * @dl_boosted tells if we are boosted due to DI. If so we are
491 	 * outside bandwidth enforcement mechanism (but only until we
492 	 * exit the critical section);
493 	 *
494 	 * @dl_yielded tells if task gave up the CPU before consuming
495 	 * all its available runtime during the last job.
496 	 *
497 	 * @dl_non_contending tells if the task is inactive while still
498 	 * contributing to the active utilization. In other words, it
499 	 * indicates if the inactive timer has been armed and its handler
500 	 * has not been executed yet. This flag is useful to avoid race
501 	 * conditions between the inactive timer handler and the wakeup
502 	 * code.
503 	 *
504 	 * @dl_overrun tells if the task asked to be informed about runtime
505 	 * overruns.
506 	 */
507 	unsigned int			dl_throttled      : 1;
508 	unsigned int			dl_boosted        : 1;
509 	unsigned int			dl_yielded        : 1;
510 	unsigned int			dl_non_contending : 1;
511 	unsigned int			dl_overrun	  : 1;
512 
513 	/*
514 	 * Bandwidth enforcement timer. Each -deadline task has its
515 	 * own bandwidth to be enforced, thus we need one timer per task.
516 	 */
517 	struct hrtimer			dl_timer;
518 
519 	/*
520 	 * Inactive timer, responsible for decreasing the active utilization
521 	 * at the "0-lag time". When a -deadline task blocks, it contributes
522 	 * to GRUB's active utilization until the "0-lag time", hence a
523 	 * timer is needed to decrease the active utilization at the correct
524 	 * time.
525 	 */
526 	struct hrtimer inactive_timer;
527 };
528 
529 union rcu_special {
530 	struct {
531 		u8			blocked;
532 		u8			need_qs;
533 		u8			exp_need_qs;
534 
535 		/* Otherwise the compiler can store garbage here: */
536 		u8			pad;
537 	} b; /* Bits. */
538 	u32 s; /* Set of bits. */
539 };
540 
541 enum perf_event_task_context {
542 	perf_invalid_context = -1,
543 	perf_hw_context = 0,
544 	perf_sw_context,
545 	perf_nr_task_contexts,
546 };
547 
548 struct wake_q_node {
549 	struct wake_q_node *next;
550 };
551 
552 struct task_struct {
553 #ifdef CONFIG_THREAD_INFO_IN_TASK
554 	/*
555 	 * For reasons of header soup (see current_thread_info()), this
556 	 * must be the first element of task_struct.
557 	 */
558 	struct thread_info		thread_info;
559 #endif
560 	/* -1 unrunnable, 0 runnable, >0 stopped: */
561 	volatile long			state;
562 
563 	/*
564 	 * This begins the randomizable portion of task_struct. Only
565 	 * scheduling-critical items should be added above here.
566 	 */
567 	randomized_struct_fields_start
568 
569 	void				*stack;
570 	atomic_t			usage;
571 	/* Per task flags (PF_*), defined further below: */
572 	unsigned int			flags;
573 	unsigned int			ptrace;
574 
575 #ifdef CONFIG_SMP
576 	struct llist_node		wake_entry;
577 	int				on_cpu;
578 #ifdef CONFIG_THREAD_INFO_IN_TASK
579 	/* Current CPU: */
580 	unsigned int			cpu;
581 #endif
582 	unsigned int			wakee_flips;
583 	unsigned long			wakee_flip_decay_ts;
584 	struct task_struct		*last_wakee;
585 
586 	/*
587 	 * recent_used_cpu is initially set as the last CPU used by a task
588 	 * that wakes affine another task. Waker/wakee relationships can
589 	 * push tasks around a CPU where each wakeup moves to the next one.
590 	 * Tracking a recently used CPU allows a quick search for a recently
591 	 * used CPU that may be idle.
592 	 */
593 	int				recent_used_cpu;
594 	int				wake_cpu;
595 #endif
596 	int				on_rq;
597 
598 	int				prio;
599 	int				static_prio;
600 	int				normal_prio;
601 	unsigned int			rt_priority;
602 
603 	const struct sched_class	*sched_class;
604 	struct sched_entity		se;
605 	struct sched_rt_entity		rt;
606 #ifdef CONFIG_CGROUP_SCHED
607 	struct task_group		*sched_task_group;
608 #endif
609 	struct sched_dl_entity		dl;
610 
611 #ifdef CONFIG_PREEMPT_NOTIFIERS
612 	/* List of struct preempt_notifier: */
613 	struct hlist_head		preempt_notifiers;
614 #endif
615 
616 #ifdef CONFIG_BLK_DEV_IO_TRACE
617 	unsigned int			btrace_seq;
618 #endif
619 
620 	unsigned int			policy;
621 	int				nr_cpus_allowed;
622 	cpumask_t			cpus_allowed;
623 
624 #ifdef CONFIG_PREEMPT_RCU
625 	int				rcu_read_lock_nesting;
626 	union rcu_special		rcu_read_unlock_special;
627 	struct list_head		rcu_node_entry;
628 	struct rcu_node			*rcu_blocked_node;
629 #endif /* #ifdef CONFIG_PREEMPT_RCU */
630 
631 #ifdef CONFIG_TASKS_RCU
632 	unsigned long			rcu_tasks_nvcsw;
633 	u8				rcu_tasks_holdout;
634 	u8				rcu_tasks_idx;
635 	int				rcu_tasks_idle_cpu;
636 	struct list_head		rcu_tasks_holdout_list;
637 #endif /* #ifdef CONFIG_TASKS_RCU */
638 
639 	struct sched_info		sched_info;
640 
641 	struct list_head		tasks;
642 #ifdef CONFIG_SMP
643 	struct plist_node		pushable_tasks;
644 	struct rb_node			pushable_dl_tasks;
645 #endif
646 
647 	struct mm_struct		*mm;
648 	struct mm_struct		*active_mm;
649 
650 	/* Per-thread vma caching: */
651 	struct vmacache			vmacache;
652 
653 #ifdef SPLIT_RSS_COUNTING
654 	struct task_rss_stat		rss_stat;
655 #endif
656 	int				exit_state;
657 	int				exit_code;
658 	int				exit_signal;
659 	/* The signal sent when the parent dies: */
660 	int				pdeath_signal;
661 	/* JOBCTL_*, siglock protected: */
662 	unsigned long			jobctl;
663 
664 	/* Used for emulating ABI behavior of previous Linux versions: */
665 	unsigned int			personality;
666 
667 	/* Scheduler bits, serialized by scheduler locks: */
668 	unsigned			sched_reset_on_fork:1;
669 	unsigned			sched_contributes_to_load:1;
670 	unsigned			sched_migrated:1;
671 	unsigned			sched_remote_wakeup:1;
672 	/* Force alignment to the next boundary: */
673 	unsigned			:0;
674 
675 	/* Unserialized, strictly 'current' */
676 
677 	/* Bit to tell LSMs we're in execve(): */
678 	unsigned			in_execve:1;
679 	unsigned			in_iowait:1;
680 #ifndef TIF_RESTORE_SIGMASK
681 	unsigned			restore_sigmask:1;
682 #endif
683 #ifdef CONFIG_MEMCG
684 	unsigned			memcg_may_oom:1;
685 #ifndef CONFIG_SLOB
686 	unsigned			memcg_kmem_skip_account:1;
687 #endif
688 #endif
689 #ifdef CONFIG_COMPAT_BRK
690 	unsigned			brk_randomized:1;
691 #endif
692 #ifdef CONFIG_CGROUPS
693 	/* disallow userland-initiated cgroup migration */
694 	unsigned			no_cgroup_migration:1;
695 #endif
696 
697 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
698 
699 	struct restart_block		restart_block;
700 
701 	pid_t				pid;
702 	pid_t				tgid;
703 
704 #ifdef CONFIG_CC_STACKPROTECTOR
705 	/* Canary value for the -fstack-protector GCC feature: */
706 	unsigned long			stack_canary;
707 #endif
708 	/*
709 	 * Pointers to the (original) parent process, youngest child, younger sibling,
710 	 * older sibling, respectively.  (p->father can be replaced with
711 	 * p->real_parent->pid)
712 	 */
713 
714 	/* Real parent process: */
715 	struct task_struct __rcu	*real_parent;
716 
717 	/* Recipient of SIGCHLD, wait4() reports: */
718 	struct task_struct __rcu	*parent;
719 
720 	/*
721 	 * Children/sibling form the list of natural children:
722 	 */
723 	struct list_head		children;
724 	struct list_head		sibling;
725 	struct task_struct		*group_leader;
726 
727 	/*
728 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
729 	 *
730 	 * This includes both natural children and PTRACE_ATTACH targets.
731 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
732 	 */
733 	struct list_head		ptraced;
734 	struct list_head		ptrace_entry;
735 
736 	/* PID/PID hash table linkage. */
737 	struct pid_link			pids[PIDTYPE_MAX];
738 	struct list_head		thread_group;
739 	struct list_head		thread_node;
740 
741 	struct completion		*vfork_done;
742 
743 	/* CLONE_CHILD_SETTID: */
744 	int __user			*set_child_tid;
745 
746 	/* CLONE_CHILD_CLEARTID: */
747 	int __user			*clear_child_tid;
748 
749 	u64				utime;
750 	u64				stime;
751 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
752 	u64				utimescaled;
753 	u64				stimescaled;
754 #endif
755 	u64				gtime;
756 	struct prev_cputime		prev_cputime;
757 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
758 	struct vtime			vtime;
759 #endif
760 
761 #ifdef CONFIG_NO_HZ_FULL
762 	atomic_t			tick_dep_mask;
763 #endif
764 	/* Context switch counts: */
765 	unsigned long			nvcsw;
766 	unsigned long			nivcsw;
767 
768 	/* Monotonic time in nsecs: */
769 	u64				start_time;
770 
771 	/* Boot based time in nsecs: */
772 	u64				real_start_time;
773 
774 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
775 	unsigned long			min_flt;
776 	unsigned long			maj_flt;
777 
778 #ifdef CONFIG_POSIX_TIMERS
779 	struct task_cputime		cputime_expires;
780 	struct list_head		cpu_timers[3];
781 #endif
782 
783 	/* Process credentials: */
784 
785 	/* Tracer's credentials at attach: */
786 	const struct cred __rcu		*ptracer_cred;
787 
788 	/* Objective and real subjective task credentials (COW): */
789 	const struct cred __rcu		*real_cred;
790 
791 	/* Effective (overridable) subjective task credentials (COW): */
792 	const struct cred __rcu		*cred;
793 
794 	/*
795 	 * executable name, excluding path.
796 	 *
797 	 * - normally initialized setup_new_exec()
798 	 * - access it with [gs]et_task_comm()
799 	 * - lock it with task_lock()
800 	 */
801 	char				comm[TASK_COMM_LEN];
802 
803 	struct nameidata		*nameidata;
804 
805 #ifdef CONFIG_SYSVIPC
806 	struct sysv_sem			sysvsem;
807 	struct sysv_shm			sysvshm;
808 #endif
809 #ifdef CONFIG_DETECT_HUNG_TASK
810 	unsigned long			last_switch_count;
811 #endif
812 	/* Filesystem information: */
813 	struct fs_struct		*fs;
814 
815 	/* Open file information: */
816 	struct files_struct		*files;
817 
818 	/* Namespaces: */
819 	struct nsproxy			*nsproxy;
820 
821 	/* Signal handlers: */
822 	struct signal_struct		*signal;
823 	struct sighand_struct		*sighand;
824 	sigset_t			blocked;
825 	sigset_t			real_blocked;
826 	/* Restored if set_restore_sigmask() was used: */
827 	sigset_t			saved_sigmask;
828 	struct sigpending		pending;
829 	unsigned long			sas_ss_sp;
830 	size_t				sas_ss_size;
831 	unsigned int			sas_ss_flags;
832 
833 	struct callback_head		*task_works;
834 
835 	struct audit_context		*audit_context;
836 #ifdef CONFIG_AUDITSYSCALL
837 	kuid_t				loginuid;
838 	unsigned int			sessionid;
839 #endif
840 	struct seccomp			seccomp;
841 
842 	/* Thread group tracking: */
843 	u32				parent_exec_id;
844 	u32				self_exec_id;
845 
846 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
847 	spinlock_t			alloc_lock;
848 
849 	/* Protection of the PI data structures: */
850 	raw_spinlock_t			pi_lock;
851 
852 	struct wake_q_node		wake_q;
853 
854 #ifdef CONFIG_RT_MUTEXES
855 	/* PI waiters blocked on a rt_mutex held by this task: */
856 	struct rb_root_cached		pi_waiters;
857 	/* Updated under owner's pi_lock and rq lock */
858 	struct task_struct		*pi_top_task;
859 	/* Deadlock detection and priority inheritance handling: */
860 	struct rt_mutex_waiter		*pi_blocked_on;
861 #endif
862 
863 #ifdef CONFIG_DEBUG_MUTEXES
864 	/* Mutex deadlock detection: */
865 	struct mutex_waiter		*blocked_on;
866 #endif
867 
868 #ifdef CONFIG_TRACE_IRQFLAGS
869 	unsigned int			irq_events;
870 	unsigned long			hardirq_enable_ip;
871 	unsigned long			hardirq_disable_ip;
872 	unsigned int			hardirq_enable_event;
873 	unsigned int			hardirq_disable_event;
874 	int				hardirqs_enabled;
875 	int				hardirq_context;
876 	unsigned long			softirq_disable_ip;
877 	unsigned long			softirq_enable_ip;
878 	unsigned int			softirq_disable_event;
879 	unsigned int			softirq_enable_event;
880 	int				softirqs_enabled;
881 	int				softirq_context;
882 #endif
883 
884 #ifdef CONFIG_LOCKDEP
885 # define MAX_LOCK_DEPTH			48UL
886 	u64				curr_chain_key;
887 	int				lockdep_depth;
888 	unsigned int			lockdep_recursion;
889 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
890 #endif
891 
892 #ifdef CONFIG_UBSAN
893 	unsigned int			in_ubsan;
894 #endif
895 
896 	/* Journalling filesystem info: */
897 	void				*journal_info;
898 
899 	/* Stacked block device info: */
900 	struct bio_list			*bio_list;
901 
902 #ifdef CONFIG_BLOCK
903 	/* Stack plugging: */
904 	struct blk_plug			*plug;
905 #endif
906 
907 	/* VM state: */
908 	struct reclaim_state		*reclaim_state;
909 
910 	struct backing_dev_info		*backing_dev_info;
911 
912 	struct io_context		*io_context;
913 
914 	/* Ptrace state: */
915 	unsigned long			ptrace_message;
916 	siginfo_t			*last_siginfo;
917 
918 	struct task_io_accounting	ioac;
919 #ifdef CONFIG_TASK_XACCT
920 	/* Accumulated RSS usage: */
921 	u64				acct_rss_mem1;
922 	/* Accumulated virtual memory usage: */
923 	u64				acct_vm_mem1;
924 	/* stime + utime since last update: */
925 	u64				acct_timexpd;
926 #endif
927 #ifdef CONFIG_CPUSETS
928 	/* Protected by ->alloc_lock: */
929 	nodemask_t			mems_allowed;
930 	/* Seqence number to catch updates: */
931 	seqcount_t			mems_allowed_seq;
932 	int				cpuset_mem_spread_rotor;
933 	int				cpuset_slab_spread_rotor;
934 #endif
935 #ifdef CONFIG_CGROUPS
936 	/* Control Group info protected by css_set_lock: */
937 	struct css_set __rcu		*cgroups;
938 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
939 	struct list_head		cg_list;
940 #endif
941 #ifdef CONFIG_INTEL_RDT
942 	u32				closid;
943 	u32				rmid;
944 #endif
945 #ifdef CONFIG_FUTEX
946 	struct robust_list_head __user	*robust_list;
947 #ifdef CONFIG_COMPAT
948 	struct compat_robust_list_head __user *compat_robust_list;
949 #endif
950 	struct list_head		pi_state_list;
951 	struct futex_pi_state		*pi_state_cache;
952 #endif
953 #ifdef CONFIG_PERF_EVENTS
954 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
955 	struct mutex			perf_event_mutex;
956 	struct list_head		perf_event_list;
957 #endif
958 #ifdef CONFIG_DEBUG_PREEMPT
959 	unsigned long			preempt_disable_ip;
960 #endif
961 #ifdef CONFIG_NUMA
962 	/* Protected by alloc_lock: */
963 	struct mempolicy		*mempolicy;
964 	short				il_prev;
965 	short				pref_node_fork;
966 #endif
967 #ifdef CONFIG_NUMA_BALANCING
968 	int				numa_scan_seq;
969 	unsigned int			numa_scan_period;
970 	unsigned int			numa_scan_period_max;
971 	int				numa_preferred_nid;
972 	unsigned long			numa_migrate_retry;
973 	/* Migration stamp: */
974 	u64				node_stamp;
975 	u64				last_task_numa_placement;
976 	u64				last_sum_exec_runtime;
977 	struct callback_head		numa_work;
978 
979 	struct list_head		numa_entry;
980 	struct numa_group		*numa_group;
981 
982 	/*
983 	 * numa_faults is an array split into four regions:
984 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
985 	 * in this precise order.
986 	 *
987 	 * faults_memory: Exponential decaying average of faults on a per-node
988 	 * basis. Scheduling placement decisions are made based on these
989 	 * counts. The values remain static for the duration of a PTE scan.
990 	 * faults_cpu: Track the nodes the process was running on when a NUMA
991 	 * hinting fault was incurred.
992 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
993 	 * during the current scan window. When the scan completes, the counts
994 	 * in faults_memory and faults_cpu decay and these values are copied.
995 	 */
996 	unsigned long			*numa_faults;
997 	unsigned long			total_numa_faults;
998 
999 	/*
1000 	 * numa_faults_locality tracks if faults recorded during the last
1001 	 * scan window were remote/local or failed to migrate. The task scan
1002 	 * period is adapted based on the locality of the faults with different
1003 	 * weights depending on whether they were shared or private faults
1004 	 */
1005 	unsigned long			numa_faults_locality[3];
1006 
1007 	unsigned long			numa_pages_migrated;
1008 #endif /* CONFIG_NUMA_BALANCING */
1009 
1010 	struct tlbflush_unmap_batch	tlb_ubc;
1011 
1012 	struct rcu_head			rcu;
1013 
1014 	/* Cache last used pipe for splice(): */
1015 	struct pipe_inode_info		*splice_pipe;
1016 
1017 	struct page_frag		task_frag;
1018 
1019 #ifdef CONFIG_TASK_DELAY_ACCT
1020 	struct task_delay_info		*delays;
1021 #endif
1022 
1023 #ifdef CONFIG_FAULT_INJECTION
1024 	int				make_it_fail;
1025 	unsigned int			fail_nth;
1026 #endif
1027 	/*
1028 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1029 	 * balance_dirty_pages() for a dirty throttling pause:
1030 	 */
1031 	int				nr_dirtied;
1032 	int				nr_dirtied_pause;
1033 	/* Start of a write-and-pause period: */
1034 	unsigned long			dirty_paused_when;
1035 
1036 #ifdef CONFIG_LATENCYTOP
1037 	int				latency_record_count;
1038 	struct latency_record		latency_record[LT_SAVECOUNT];
1039 #endif
1040 	/*
1041 	 * Time slack values; these are used to round up poll() and
1042 	 * select() etc timeout values. These are in nanoseconds.
1043 	 */
1044 	u64				timer_slack_ns;
1045 	u64				default_timer_slack_ns;
1046 
1047 #ifdef CONFIG_KASAN
1048 	unsigned int			kasan_depth;
1049 #endif
1050 
1051 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1052 	/* Index of current stored address in ret_stack: */
1053 	int				curr_ret_stack;
1054 
1055 	/* Stack of return addresses for return function tracing: */
1056 	struct ftrace_ret_stack		*ret_stack;
1057 
1058 	/* Timestamp for last schedule: */
1059 	unsigned long long		ftrace_timestamp;
1060 
1061 	/*
1062 	 * Number of functions that haven't been traced
1063 	 * because of depth overrun:
1064 	 */
1065 	atomic_t			trace_overrun;
1066 
1067 	/* Pause tracing: */
1068 	atomic_t			tracing_graph_pause;
1069 #endif
1070 
1071 #ifdef CONFIG_TRACING
1072 	/* State flags for use by tracers: */
1073 	unsigned long			trace;
1074 
1075 	/* Bitmask and counter of trace recursion: */
1076 	unsigned long			trace_recursion;
1077 #endif /* CONFIG_TRACING */
1078 
1079 #ifdef CONFIG_KCOV
1080 	/* Coverage collection mode enabled for this task (0 if disabled): */
1081 	enum kcov_mode			kcov_mode;
1082 
1083 	/* Size of the kcov_area: */
1084 	unsigned int			kcov_size;
1085 
1086 	/* Buffer for coverage collection: */
1087 	void				*kcov_area;
1088 
1089 	/* KCOV descriptor wired with this task or NULL: */
1090 	struct kcov			*kcov;
1091 #endif
1092 
1093 #ifdef CONFIG_MEMCG
1094 	struct mem_cgroup		*memcg_in_oom;
1095 	gfp_t				memcg_oom_gfp_mask;
1096 	int				memcg_oom_order;
1097 
1098 	/* Number of pages to reclaim on returning to userland: */
1099 	unsigned int			memcg_nr_pages_over_high;
1100 #endif
1101 
1102 #ifdef CONFIG_UPROBES
1103 	struct uprobe_task		*utask;
1104 #endif
1105 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1106 	unsigned int			sequential_io;
1107 	unsigned int			sequential_io_avg;
1108 #endif
1109 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1110 	unsigned long			task_state_change;
1111 #endif
1112 	int				pagefault_disabled;
1113 #ifdef CONFIG_MMU
1114 	struct task_struct		*oom_reaper_list;
1115 #endif
1116 #ifdef CONFIG_VMAP_STACK
1117 	struct vm_struct		*stack_vm_area;
1118 #endif
1119 #ifdef CONFIG_THREAD_INFO_IN_TASK
1120 	/* A live task holds one reference: */
1121 	atomic_t			stack_refcount;
1122 #endif
1123 #ifdef CONFIG_LIVEPATCH
1124 	int patch_state;
1125 #endif
1126 #ifdef CONFIG_SECURITY
1127 	/* Used by LSM modules for access restriction: */
1128 	void				*security;
1129 #endif
1130 
1131 	/*
1132 	 * New fields for task_struct should be added above here, so that
1133 	 * they are included in the randomized portion of task_struct.
1134 	 */
1135 	randomized_struct_fields_end
1136 
1137 	/* CPU-specific state of this task: */
1138 	struct thread_struct		thread;
1139 
1140 	/*
1141 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1142 	 * structure.  It *MUST* be at the end of 'task_struct'.
1143 	 *
1144 	 * Do not put anything below here!
1145 	 */
1146 };
1147 
1148 static inline struct pid *task_pid(struct task_struct *task)
1149 {
1150 	return task->pids[PIDTYPE_PID].pid;
1151 }
1152 
1153 static inline struct pid *task_tgid(struct task_struct *task)
1154 {
1155 	return task->group_leader->pids[PIDTYPE_PID].pid;
1156 }
1157 
1158 /*
1159  * Without tasklist or RCU lock it is not safe to dereference
1160  * the result of task_pgrp/task_session even if task == current,
1161  * we can race with another thread doing sys_setsid/sys_setpgid.
1162  */
1163 static inline struct pid *task_pgrp(struct task_struct *task)
1164 {
1165 	return task->group_leader->pids[PIDTYPE_PGID].pid;
1166 }
1167 
1168 static inline struct pid *task_session(struct task_struct *task)
1169 {
1170 	return task->group_leader->pids[PIDTYPE_SID].pid;
1171 }
1172 
1173 /*
1174  * the helpers to get the task's different pids as they are seen
1175  * from various namespaces
1176  *
1177  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1178  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1179  *                     current.
1180  * task_xid_nr_ns()  : id seen from the ns specified;
1181  *
1182  * see also pid_nr() etc in include/linux/pid.h
1183  */
1184 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1185 
1186 static inline pid_t task_pid_nr(struct task_struct *tsk)
1187 {
1188 	return tsk->pid;
1189 }
1190 
1191 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1192 {
1193 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1194 }
1195 
1196 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1197 {
1198 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1199 }
1200 
1201 
1202 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1203 {
1204 	return tsk->tgid;
1205 }
1206 
1207 /**
1208  * pid_alive - check that a task structure is not stale
1209  * @p: Task structure to be checked.
1210  *
1211  * Test if a process is not yet dead (at most zombie state)
1212  * If pid_alive fails, then pointers within the task structure
1213  * can be stale and must not be dereferenced.
1214  *
1215  * Return: 1 if the process is alive. 0 otherwise.
1216  */
1217 static inline int pid_alive(const struct task_struct *p)
1218 {
1219 	return p->pids[PIDTYPE_PID].pid != NULL;
1220 }
1221 
1222 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1223 {
1224 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1225 }
1226 
1227 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1228 {
1229 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1230 }
1231 
1232 
1233 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1234 {
1235 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1236 }
1237 
1238 static inline pid_t task_session_vnr(struct task_struct *tsk)
1239 {
1240 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1241 }
1242 
1243 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1244 {
1245 	return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
1246 }
1247 
1248 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1249 {
1250 	return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
1251 }
1252 
1253 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1254 {
1255 	pid_t pid = 0;
1256 
1257 	rcu_read_lock();
1258 	if (pid_alive(tsk))
1259 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1260 	rcu_read_unlock();
1261 
1262 	return pid;
1263 }
1264 
1265 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1266 {
1267 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1268 }
1269 
1270 /* Obsolete, do not use: */
1271 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1272 {
1273 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1274 }
1275 
1276 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1277 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1278 
1279 static inline unsigned int task_state_index(struct task_struct *tsk)
1280 {
1281 	unsigned int tsk_state = READ_ONCE(tsk->state);
1282 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1283 
1284 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1285 
1286 	if (tsk_state == TASK_IDLE)
1287 		state = TASK_REPORT_IDLE;
1288 
1289 	return fls(state);
1290 }
1291 
1292 static inline char task_index_to_char(unsigned int state)
1293 {
1294 	static const char state_char[] = "RSDTtXZPI";
1295 
1296 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1297 
1298 	return state_char[state];
1299 }
1300 
1301 static inline char task_state_to_char(struct task_struct *tsk)
1302 {
1303 	return task_index_to_char(task_state_index(tsk));
1304 }
1305 
1306 /**
1307  * is_global_init - check if a task structure is init. Since init
1308  * is free to have sub-threads we need to check tgid.
1309  * @tsk: Task structure to be checked.
1310  *
1311  * Check if a task structure is the first user space task the kernel created.
1312  *
1313  * Return: 1 if the task structure is init. 0 otherwise.
1314  */
1315 static inline int is_global_init(struct task_struct *tsk)
1316 {
1317 	return task_tgid_nr(tsk) == 1;
1318 }
1319 
1320 extern struct pid *cad_pid;
1321 
1322 /*
1323  * Per process flags
1324  */
1325 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1326 #define PF_EXITING		0x00000004	/* Getting shut down */
1327 #define PF_EXITPIDONE		0x00000008	/* PI exit done on shut down */
1328 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1329 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1330 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1331 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1332 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1333 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1334 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1335 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1336 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1337 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1338 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1339 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1340 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1341 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1342 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1343 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1344 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1345 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1346 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1347 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1348 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_allowed */
1349 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1350 #define PF_MUTEX_TESTER		0x20000000	/* Thread belongs to the rt mutex tester */
1351 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1352 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1353 
1354 /*
1355  * Only the _current_ task can read/write to tsk->flags, but other
1356  * tasks can access tsk->flags in readonly mode for example
1357  * with tsk_used_math (like during threaded core dumping).
1358  * There is however an exception to this rule during ptrace
1359  * or during fork: the ptracer task is allowed to write to the
1360  * child->flags of its traced child (same goes for fork, the parent
1361  * can write to the child->flags), because we're guaranteed the
1362  * child is not running and in turn not changing child->flags
1363  * at the same time the parent does it.
1364  */
1365 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1366 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1367 #define clear_used_math()			clear_stopped_child_used_math(current)
1368 #define set_used_math()				set_stopped_child_used_math(current)
1369 
1370 #define conditional_stopped_child_used_math(condition, child) \
1371 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1372 
1373 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1374 
1375 #define copy_to_stopped_child_used_math(child) \
1376 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1377 
1378 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1379 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1380 #define used_math()				tsk_used_math(current)
1381 
1382 static inline bool is_percpu_thread(void)
1383 {
1384 #ifdef CONFIG_SMP
1385 	return (current->flags & PF_NO_SETAFFINITY) &&
1386 		(current->nr_cpus_allowed  == 1);
1387 #else
1388 	return true;
1389 #endif
1390 }
1391 
1392 /* Per-process atomic flags. */
1393 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1394 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1395 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1396 
1397 
1398 #define TASK_PFA_TEST(name, func)					\
1399 	static inline bool task_##func(struct task_struct *p)		\
1400 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1401 
1402 #define TASK_PFA_SET(name, func)					\
1403 	static inline void task_set_##func(struct task_struct *p)	\
1404 	{ set_bit(PFA_##name, &p->atomic_flags); }
1405 
1406 #define TASK_PFA_CLEAR(name, func)					\
1407 	static inline void task_clear_##func(struct task_struct *p)	\
1408 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1409 
1410 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1411 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1412 
1413 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1414 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1415 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1416 
1417 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1418 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1419 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1420 
1421 static inline void
1422 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1423 {
1424 	current->flags &= ~flags;
1425 	current->flags |= orig_flags & flags;
1426 }
1427 
1428 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1429 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1430 #ifdef CONFIG_SMP
1431 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1432 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1433 #else
1434 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1435 {
1436 }
1437 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1438 {
1439 	if (!cpumask_test_cpu(0, new_mask))
1440 		return -EINVAL;
1441 	return 0;
1442 }
1443 #endif
1444 
1445 #ifndef cpu_relax_yield
1446 #define cpu_relax_yield() cpu_relax()
1447 #endif
1448 
1449 extern int yield_to(struct task_struct *p, bool preempt);
1450 extern void set_user_nice(struct task_struct *p, long nice);
1451 extern int task_prio(const struct task_struct *p);
1452 
1453 /**
1454  * task_nice - return the nice value of a given task.
1455  * @p: the task in question.
1456  *
1457  * Return: The nice value [ -20 ... 0 ... 19 ].
1458  */
1459 static inline int task_nice(const struct task_struct *p)
1460 {
1461 	return PRIO_TO_NICE((p)->static_prio);
1462 }
1463 
1464 extern int can_nice(const struct task_struct *p, const int nice);
1465 extern int task_curr(const struct task_struct *p);
1466 extern int idle_cpu(int cpu);
1467 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1468 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1469 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1470 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1471 extern struct task_struct *idle_task(int cpu);
1472 
1473 /**
1474  * is_idle_task - is the specified task an idle task?
1475  * @p: the task in question.
1476  *
1477  * Return: 1 if @p is an idle task. 0 otherwise.
1478  */
1479 static inline bool is_idle_task(const struct task_struct *p)
1480 {
1481 	return !!(p->flags & PF_IDLE);
1482 }
1483 
1484 extern struct task_struct *curr_task(int cpu);
1485 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1486 
1487 void yield(void);
1488 
1489 union thread_union {
1490 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1491 	struct task_struct task;
1492 #endif
1493 #ifndef CONFIG_THREAD_INFO_IN_TASK
1494 	struct thread_info thread_info;
1495 #endif
1496 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1497 };
1498 
1499 #ifndef CONFIG_THREAD_INFO_IN_TASK
1500 extern struct thread_info init_thread_info;
1501 #endif
1502 
1503 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1504 
1505 #ifdef CONFIG_THREAD_INFO_IN_TASK
1506 static inline struct thread_info *task_thread_info(struct task_struct *task)
1507 {
1508 	return &task->thread_info;
1509 }
1510 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1511 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1512 #endif
1513 
1514 /*
1515  * find a task by one of its numerical ids
1516  *
1517  * find_task_by_pid_ns():
1518  *      finds a task by its pid in the specified namespace
1519  * find_task_by_vpid():
1520  *      finds a task by its virtual pid
1521  *
1522  * see also find_vpid() etc in include/linux/pid.h
1523  */
1524 
1525 extern struct task_struct *find_task_by_vpid(pid_t nr);
1526 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1527 
1528 /*
1529  * find a task by its virtual pid and get the task struct
1530  */
1531 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1532 
1533 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1534 extern int wake_up_process(struct task_struct *tsk);
1535 extern void wake_up_new_task(struct task_struct *tsk);
1536 
1537 #ifdef CONFIG_SMP
1538 extern void kick_process(struct task_struct *tsk);
1539 #else
1540 static inline void kick_process(struct task_struct *tsk) { }
1541 #endif
1542 
1543 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1544 
1545 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1546 {
1547 	__set_task_comm(tsk, from, false);
1548 }
1549 
1550 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1551 #define get_task_comm(buf, tsk) ({			\
1552 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1553 	__get_task_comm(buf, sizeof(buf), tsk);		\
1554 })
1555 
1556 #ifdef CONFIG_SMP
1557 void scheduler_ipi(void);
1558 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1559 #else
1560 static inline void scheduler_ipi(void) { }
1561 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1562 {
1563 	return 1;
1564 }
1565 #endif
1566 
1567 /*
1568  * Set thread flags in other task's structures.
1569  * See asm/thread_info.h for TIF_xxxx flags available:
1570  */
1571 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1572 {
1573 	set_ti_thread_flag(task_thread_info(tsk), flag);
1574 }
1575 
1576 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1577 {
1578 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1579 }
1580 
1581 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1582 {
1583 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1584 }
1585 
1586 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1587 {
1588 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1589 }
1590 
1591 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1592 {
1593 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1594 }
1595 
1596 static inline void set_tsk_need_resched(struct task_struct *tsk)
1597 {
1598 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1599 }
1600 
1601 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1602 {
1603 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1604 }
1605 
1606 static inline int test_tsk_need_resched(struct task_struct *tsk)
1607 {
1608 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1609 }
1610 
1611 /*
1612  * cond_resched() and cond_resched_lock(): latency reduction via
1613  * explicit rescheduling in places that are safe. The return
1614  * value indicates whether a reschedule was done in fact.
1615  * cond_resched_lock() will drop the spinlock before scheduling,
1616  * cond_resched_softirq() will enable bhs before scheduling.
1617  */
1618 #ifndef CONFIG_PREEMPT
1619 extern int _cond_resched(void);
1620 #else
1621 static inline int _cond_resched(void) { return 0; }
1622 #endif
1623 
1624 #define cond_resched() ({			\
1625 	___might_sleep(__FILE__, __LINE__, 0);	\
1626 	_cond_resched();			\
1627 })
1628 
1629 extern int __cond_resched_lock(spinlock_t *lock);
1630 
1631 #define cond_resched_lock(lock) ({				\
1632 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1633 	__cond_resched_lock(lock);				\
1634 })
1635 
1636 extern int __cond_resched_softirq(void);
1637 
1638 #define cond_resched_softirq() ({					\
1639 	___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET);	\
1640 	__cond_resched_softirq();					\
1641 })
1642 
1643 static inline void cond_resched_rcu(void)
1644 {
1645 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1646 	rcu_read_unlock();
1647 	cond_resched();
1648 	rcu_read_lock();
1649 #endif
1650 }
1651 
1652 /*
1653  * Does a critical section need to be broken due to another
1654  * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1655  * but a general need for low latency)
1656  */
1657 static inline int spin_needbreak(spinlock_t *lock)
1658 {
1659 #ifdef CONFIG_PREEMPT
1660 	return spin_is_contended(lock);
1661 #else
1662 	return 0;
1663 #endif
1664 }
1665 
1666 static __always_inline bool need_resched(void)
1667 {
1668 	return unlikely(tif_need_resched());
1669 }
1670 
1671 /*
1672  * Wrappers for p->thread_info->cpu access. No-op on UP.
1673  */
1674 #ifdef CONFIG_SMP
1675 
1676 static inline unsigned int task_cpu(const struct task_struct *p)
1677 {
1678 #ifdef CONFIG_THREAD_INFO_IN_TASK
1679 	return p->cpu;
1680 #else
1681 	return task_thread_info(p)->cpu;
1682 #endif
1683 }
1684 
1685 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1686 
1687 #else
1688 
1689 static inline unsigned int task_cpu(const struct task_struct *p)
1690 {
1691 	return 0;
1692 }
1693 
1694 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1695 {
1696 }
1697 
1698 #endif /* CONFIG_SMP */
1699 
1700 /*
1701  * In order to reduce various lock holder preemption latencies provide an
1702  * interface to see if a vCPU is currently running or not.
1703  *
1704  * This allows us to terminate optimistic spin loops and block, analogous to
1705  * the native optimistic spin heuristic of testing if the lock owner task is
1706  * running or not.
1707  */
1708 #ifndef vcpu_is_preempted
1709 # define vcpu_is_preempted(cpu)	false
1710 #endif
1711 
1712 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1713 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1714 
1715 #ifndef TASK_SIZE_OF
1716 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1717 #endif
1718 
1719 #endif
1720