xref: /linux/include/linux/sched.h (revision 4c5a116ada953b86125ab7c70a57c57463a55a55)
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/irqflags.h>
22 #include <linux/seccomp.h>
23 #include <linux/nodemask.h>
24 #include <linux/rcupdate.h>
25 #include <linux/refcount.h>
26 #include <linux/resource.h>
27 #include <linux/latencytop.h>
28 #include <linux/sched/prio.h>
29 #include <linux/sched/types.h>
30 #include <linux/signal_types.h>
31 #include <linux/mm_types_task.h>
32 #include <linux/task_io_accounting.h>
33 #include <linux/posix-timers.h>
34 #include <linux/rseq.h>
35 #include <linux/kcsan.h>
36 
37 /* task_struct member predeclarations (sorted alphabetically): */
38 struct audit_context;
39 struct backing_dev_info;
40 struct bio_list;
41 struct blk_plug;
42 struct capture_control;
43 struct cfs_rq;
44 struct fs_struct;
45 struct futex_pi_state;
46 struct io_context;
47 struct mempolicy;
48 struct nameidata;
49 struct nsproxy;
50 struct perf_event_context;
51 struct pid_namespace;
52 struct pipe_inode_info;
53 struct rcu_node;
54 struct reclaim_state;
55 struct robust_list_head;
56 struct root_domain;
57 struct rq;
58 struct sched_attr;
59 struct sched_param;
60 struct seq_file;
61 struct sighand_struct;
62 struct signal_struct;
63 struct task_delay_info;
64 struct task_group;
65 
66 /*
67  * Task state bitmask. NOTE! These bits are also
68  * encoded in fs/proc/array.c: get_task_state().
69  *
70  * We have two separate sets of flags: task->state
71  * is about runnability, while task->exit_state are
72  * about the task exiting. Confusing, but this way
73  * modifying one set can't modify the other one by
74  * mistake.
75  */
76 
77 /* Used in tsk->state: */
78 #define TASK_RUNNING			0x0000
79 #define TASK_INTERRUPTIBLE		0x0001
80 #define TASK_UNINTERRUPTIBLE		0x0002
81 #define __TASK_STOPPED			0x0004
82 #define __TASK_TRACED			0x0008
83 /* Used in tsk->exit_state: */
84 #define EXIT_DEAD			0x0010
85 #define EXIT_ZOMBIE			0x0020
86 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
87 /* Used in tsk->state again: */
88 #define TASK_PARKED			0x0040
89 #define TASK_DEAD			0x0080
90 #define TASK_WAKEKILL			0x0100
91 #define TASK_WAKING			0x0200
92 #define TASK_NOLOAD			0x0400
93 #define TASK_NEW			0x0800
94 #define TASK_STATE_MAX			0x1000
95 
96 /* Convenience macros for the sake of set_current_state: */
97 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
98 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
99 #define TASK_TRACED			(TASK_WAKEKILL | __TASK_TRACED)
100 
101 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
102 
103 /* Convenience macros for the sake of wake_up(): */
104 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
105 
106 /* get_task_state(): */
107 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
108 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
109 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
110 					 TASK_PARKED)
111 
112 #define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
113 
114 #define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
115 
116 #define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
117 
118 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
119 
120 /*
121  * Special states are those that do not use the normal wait-loop pattern. See
122  * the comment with set_special_state().
123  */
124 #define is_special_task_state(state)				\
125 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
126 
127 #define __set_current_state(state_value)			\
128 	do {							\
129 		WARN_ON_ONCE(is_special_task_state(state_value));\
130 		current->task_state_change = _THIS_IP_;		\
131 		current->state = (state_value);			\
132 	} while (0)
133 
134 #define set_current_state(state_value)				\
135 	do {							\
136 		WARN_ON_ONCE(is_special_task_state(state_value));\
137 		current->task_state_change = _THIS_IP_;		\
138 		smp_store_mb(current->state, (state_value));	\
139 	} while (0)
140 
141 #define set_special_state(state_value)					\
142 	do {								\
143 		unsigned long flags; /* may shadow */			\
144 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
145 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
146 		current->task_state_change = _THIS_IP_;			\
147 		current->state = (state_value);				\
148 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
149 	} while (0)
150 #else
151 /*
152  * set_current_state() includes a barrier so that the write of current->state
153  * is correctly serialised wrt the caller's subsequent test of whether to
154  * actually sleep:
155  *
156  *   for (;;) {
157  *	set_current_state(TASK_UNINTERRUPTIBLE);
158  *	if (CONDITION)
159  *	   break;
160  *
161  *	schedule();
162  *   }
163  *   __set_current_state(TASK_RUNNING);
164  *
165  * If the caller does not need such serialisation (because, for instance, the
166  * CONDITION test and condition change and wakeup are under the same lock) then
167  * use __set_current_state().
168  *
169  * The above is typically ordered against the wakeup, which does:
170  *
171  *   CONDITION = 1;
172  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
173  *
174  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
175  * accessing p->state.
176  *
177  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
178  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
179  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
180  *
181  * However, with slightly different timing the wakeup TASK_RUNNING store can
182  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
183  * a problem either because that will result in one extra go around the loop
184  * and our @cond test will save the day.
185  *
186  * Also see the comments of try_to_wake_up().
187  */
188 #define __set_current_state(state_value)				\
189 	current->state = (state_value)
190 
191 #define set_current_state(state_value)					\
192 	smp_store_mb(current->state, (state_value))
193 
194 /*
195  * set_special_state() should be used for those states when the blocking task
196  * can not use the regular condition based wait-loop. In that case we must
197  * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
198  * will not collide with our state change.
199  */
200 #define set_special_state(state_value)					\
201 	do {								\
202 		unsigned long flags; /* may shadow */			\
203 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
204 		current->state = (state_value);				\
205 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
206 	} while (0)
207 
208 #endif
209 
210 /* Task command name length: */
211 #define TASK_COMM_LEN			16
212 
213 extern void scheduler_tick(void);
214 
215 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
216 
217 extern long schedule_timeout(long timeout);
218 extern long schedule_timeout_interruptible(long timeout);
219 extern long schedule_timeout_killable(long timeout);
220 extern long schedule_timeout_uninterruptible(long timeout);
221 extern long schedule_timeout_idle(long timeout);
222 asmlinkage void schedule(void);
223 extern void schedule_preempt_disabled(void);
224 asmlinkage void preempt_schedule_irq(void);
225 
226 extern int __must_check io_schedule_prepare(void);
227 extern void io_schedule_finish(int token);
228 extern long io_schedule_timeout(long timeout);
229 extern void io_schedule(void);
230 
231 /**
232  * struct prev_cputime - snapshot of system and user cputime
233  * @utime: time spent in user mode
234  * @stime: time spent in system mode
235  * @lock: protects the above two fields
236  *
237  * Stores previous user/system time values such that we can guarantee
238  * monotonicity.
239  */
240 struct prev_cputime {
241 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
242 	u64				utime;
243 	u64				stime;
244 	raw_spinlock_t			lock;
245 #endif
246 };
247 
248 enum vtime_state {
249 	/* Task is sleeping or running in a CPU with VTIME inactive: */
250 	VTIME_INACTIVE = 0,
251 	/* Task is idle */
252 	VTIME_IDLE,
253 	/* Task runs in kernelspace in a CPU with VTIME active: */
254 	VTIME_SYS,
255 	/* Task runs in userspace in a CPU with VTIME active: */
256 	VTIME_USER,
257 	/* Task runs as guests in a CPU with VTIME active: */
258 	VTIME_GUEST,
259 };
260 
261 struct vtime {
262 	seqcount_t		seqcount;
263 	unsigned long long	starttime;
264 	enum vtime_state	state;
265 	unsigned int		cpu;
266 	u64			utime;
267 	u64			stime;
268 	u64			gtime;
269 };
270 
271 /*
272  * Utilization clamp constraints.
273  * @UCLAMP_MIN:	Minimum utilization
274  * @UCLAMP_MAX:	Maximum utilization
275  * @UCLAMP_CNT:	Utilization clamp constraints count
276  */
277 enum uclamp_id {
278 	UCLAMP_MIN = 0,
279 	UCLAMP_MAX,
280 	UCLAMP_CNT
281 };
282 
283 #ifdef CONFIG_SMP
284 extern struct root_domain def_root_domain;
285 extern struct mutex sched_domains_mutex;
286 #endif
287 
288 struct sched_info {
289 #ifdef CONFIG_SCHED_INFO
290 	/* Cumulative counters: */
291 
292 	/* # of times we have run on this CPU: */
293 	unsigned long			pcount;
294 
295 	/* Time spent waiting on a runqueue: */
296 	unsigned long long		run_delay;
297 
298 	/* Timestamps: */
299 
300 	/* When did we last run on a CPU? */
301 	unsigned long long		last_arrival;
302 
303 	/* When were we last queued to run? */
304 	unsigned long long		last_queued;
305 
306 #endif /* CONFIG_SCHED_INFO */
307 };
308 
309 /*
310  * Integer metrics need fixed point arithmetic, e.g., sched/fair
311  * has a few: load, load_avg, util_avg, freq, and capacity.
312  *
313  * We define a basic fixed point arithmetic range, and then formalize
314  * all these metrics based on that basic range.
315  */
316 # define SCHED_FIXEDPOINT_SHIFT		10
317 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
318 
319 /* Increase resolution of cpu_capacity calculations */
320 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
321 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
322 
323 struct load_weight {
324 	unsigned long			weight;
325 	u32				inv_weight;
326 };
327 
328 /**
329  * struct util_est - Estimation utilization of FAIR tasks
330  * @enqueued: instantaneous estimated utilization of a task/cpu
331  * @ewma:     the Exponential Weighted Moving Average (EWMA)
332  *            utilization of a task
333  *
334  * Support data structure to track an Exponential Weighted Moving Average
335  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
336  * average each time a task completes an activation. Sample's weight is chosen
337  * so that the EWMA will be relatively insensitive to transient changes to the
338  * task's workload.
339  *
340  * The enqueued attribute has a slightly different meaning for tasks and cpus:
341  * - task:   the task's util_avg at last task dequeue time
342  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
343  * Thus, the util_est.enqueued of a task represents the contribution on the
344  * estimated utilization of the CPU where that task is currently enqueued.
345  *
346  * Only for tasks we track a moving average of the past instantaneous
347  * estimated utilization. This allows to absorb sporadic drops in utilization
348  * of an otherwise almost periodic task.
349  */
350 struct util_est {
351 	unsigned int			enqueued;
352 	unsigned int			ewma;
353 #define UTIL_EST_WEIGHT_SHIFT		2
354 } __attribute__((__aligned__(sizeof(u64))));
355 
356 /*
357  * The load/runnable/util_avg accumulates an infinite geometric series
358  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
359  *
360  * [load_avg definition]
361  *
362  *   load_avg = runnable% * scale_load_down(load)
363  *
364  * [runnable_avg definition]
365  *
366  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
367  *
368  * [util_avg definition]
369  *
370  *   util_avg = running% * SCHED_CAPACITY_SCALE
371  *
372  * where runnable% is the time ratio that a sched_entity is runnable and
373  * running% the time ratio that a sched_entity is running.
374  *
375  * For cfs_rq, they are the aggregated values of all runnable and blocked
376  * sched_entities.
377  *
378  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
379  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
380  * for computing those signals (see update_rq_clock_pelt())
381  *
382  * N.B., the above ratios (runnable% and running%) themselves are in the
383  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
384  * to as large a range as necessary. This is for example reflected by
385  * util_avg's SCHED_CAPACITY_SCALE.
386  *
387  * [Overflow issue]
388  *
389  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
390  * with the highest load (=88761), always runnable on a single cfs_rq,
391  * and should not overflow as the number already hits PID_MAX_LIMIT.
392  *
393  * For all other cases (including 32-bit kernels), struct load_weight's
394  * weight will overflow first before we do, because:
395  *
396  *    Max(load_avg) <= Max(load.weight)
397  *
398  * Then it is the load_weight's responsibility to consider overflow
399  * issues.
400  */
401 struct sched_avg {
402 	u64				last_update_time;
403 	u64				load_sum;
404 	u64				runnable_sum;
405 	u32				util_sum;
406 	u32				period_contrib;
407 	unsigned long			load_avg;
408 	unsigned long			runnable_avg;
409 	unsigned long			util_avg;
410 	struct util_est			util_est;
411 } ____cacheline_aligned;
412 
413 struct sched_statistics {
414 #ifdef CONFIG_SCHEDSTATS
415 	u64				wait_start;
416 	u64				wait_max;
417 	u64				wait_count;
418 	u64				wait_sum;
419 	u64				iowait_count;
420 	u64				iowait_sum;
421 
422 	u64				sleep_start;
423 	u64				sleep_max;
424 	s64				sum_sleep_runtime;
425 
426 	u64				block_start;
427 	u64				block_max;
428 	u64				exec_max;
429 	u64				slice_max;
430 
431 	u64				nr_migrations_cold;
432 	u64				nr_failed_migrations_affine;
433 	u64				nr_failed_migrations_running;
434 	u64				nr_failed_migrations_hot;
435 	u64				nr_forced_migrations;
436 
437 	u64				nr_wakeups;
438 	u64				nr_wakeups_sync;
439 	u64				nr_wakeups_migrate;
440 	u64				nr_wakeups_local;
441 	u64				nr_wakeups_remote;
442 	u64				nr_wakeups_affine;
443 	u64				nr_wakeups_affine_attempts;
444 	u64				nr_wakeups_passive;
445 	u64				nr_wakeups_idle;
446 #endif
447 };
448 
449 struct sched_entity {
450 	/* For load-balancing: */
451 	struct load_weight		load;
452 	struct rb_node			run_node;
453 	struct list_head		group_node;
454 	unsigned int			on_rq;
455 
456 	u64				exec_start;
457 	u64				sum_exec_runtime;
458 	u64				vruntime;
459 	u64				prev_sum_exec_runtime;
460 
461 	u64				nr_migrations;
462 
463 	struct sched_statistics		statistics;
464 
465 #ifdef CONFIG_FAIR_GROUP_SCHED
466 	int				depth;
467 	struct sched_entity		*parent;
468 	/* rq on which this entity is (to be) queued: */
469 	struct cfs_rq			*cfs_rq;
470 	/* rq "owned" by this entity/group: */
471 	struct cfs_rq			*my_q;
472 	/* cached value of my_q->h_nr_running */
473 	unsigned long			runnable_weight;
474 #endif
475 
476 #ifdef CONFIG_SMP
477 	/*
478 	 * Per entity load average tracking.
479 	 *
480 	 * Put into separate cache line so it does not
481 	 * collide with read-mostly values above.
482 	 */
483 	struct sched_avg		avg;
484 #endif
485 };
486 
487 struct sched_rt_entity {
488 	struct list_head		run_list;
489 	unsigned long			timeout;
490 	unsigned long			watchdog_stamp;
491 	unsigned int			time_slice;
492 	unsigned short			on_rq;
493 	unsigned short			on_list;
494 
495 	struct sched_rt_entity		*back;
496 #ifdef CONFIG_RT_GROUP_SCHED
497 	struct sched_rt_entity		*parent;
498 	/* rq on which this entity is (to be) queued: */
499 	struct rt_rq			*rt_rq;
500 	/* rq "owned" by this entity/group: */
501 	struct rt_rq			*my_q;
502 #endif
503 } __randomize_layout;
504 
505 struct sched_dl_entity {
506 	struct rb_node			rb_node;
507 
508 	/*
509 	 * Original scheduling parameters. Copied here from sched_attr
510 	 * during sched_setattr(), they will remain the same until
511 	 * the next sched_setattr().
512 	 */
513 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
514 	u64				dl_deadline;	/* Relative deadline of each instance	*/
515 	u64				dl_period;	/* Separation of two instances (period) */
516 	u64				dl_bw;		/* dl_runtime / dl_period		*/
517 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
518 
519 	/*
520 	 * Actual scheduling parameters. Initialized with the values above,
521 	 * they are continuously updated during task execution. Note that
522 	 * the remaining runtime could be < 0 in case we are in overrun.
523 	 */
524 	s64				runtime;	/* Remaining runtime for this instance	*/
525 	u64				deadline;	/* Absolute deadline for this instance	*/
526 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
527 
528 	/*
529 	 * Some bool flags:
530 	 *
531 	 * @dl_throttled tells if we exhausted the runtime. If so, the
532 	 * task has to wait for a replenishment to be performed at the
533 	 * next firing of dl_timer.
534 	 *
535 	 * @dl_boosted tells if we are boosted due to DI. If so we are
536 	 * outside bandwidth enforcement mechanism (but only until we
537 	 * exit the critical section);
538 	 *
539 	 * @dl_yielded tells if task gave up the CPU before consuming
540 	 * all its available runtime during the last job.
541 	 *
542 	 * @dl_non_contending tells if the task is inactive while still
543 	 * contributing to the active utilization. In other words, it
544 	 * indicates if the inactive timer has been armed and its handler
545 	 * has not been executed yet. This flag is useful to avoid race
546 	 * conditions between the inactive timer handler and the wakeup
547 	 * code.
548 	 *
549 	 * @dl_overrun tells if the task asked to be informed about runtime
550 	 * overruns.
551 	 */
552 	unsigned int			dl_throttled      : 1;
553 	unsigned int			dl_boosted        : 1;
554 	unsigned int			dl_yielded        : 1;
555 	unsigned int			dl_non_contending : 1;
556 	unsigned int			dl_overrun	  : 1;
557 
558 	/*
559 	 * Bandwidth enforcement timer. Each -deadline task has its
560 	 * own bandwidth to be enforced, thus we need one timer per task.
561 	 */
562 	struct hrtimer			dl_timer;
563 
564 	/*
565 	 * Inactive timer, responsible for decreasing the active utilization
566 	 * at the "0-lag time". When a -deadline task blocks, it contributes
567 	 * to GRUB's active utilization until the "0-lag time", hence a
568 	 * timer is needed to decrease the active utilization at the correct
569 	 * time.
570 	 */
571 	struct hrtimer inactive_timer;
572 };
573 
574 #ifdef CONFIG_UCLAMP_TASK
575 /* Number of utilization clamp buckets (shorter alias) */
576 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
577 
578 /*
579  * Utilization clamp for a scheduling entity
580  * @value:		clamp value "assigned" to a se
581  * @bucket_id:		bucket index corresponding to the "assigned" value
582  * @active:		the se is currently refcounted in a rq's bucket
583  * @user_defined:	the requested clamp value comes from user-space
584  *
585  * The bucket_id is the index of the clamp bucket matching the clamp value
586  * which is pre-computed and stored to avoid expensive integer divisions from
587  * the fast path.
588  *
589  * The active bit is set whenever a task has got an "effective" value assigned,
590  * which can be different from the clamp value "requested" from user-space.
591  * This allows to know a task is refcounted in the rq's bucket corresponding
592  * to the "effective" bucket_id.
593  *
594  * The user_defined bit is set whenever a task has got a task-specific clamp
595  * value requested from userspace, i.e. the system defaults apply to this task
596  * just as a restriction. This allows to relax default clamps when a less
597  * restrictive task-specific value has been requested, thus allowing to
598  * implement a "nice" semantic. For example, a task running with a 20%
599  * default boost can still drop its own boosting to 0%.
600  */
601 struct uclamp_se {
602 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
603 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
604 	unsigned int active		: 1;
605 	unsigned int user_defined	: 1;
606 };
607 #endif /* CONFIG_UCLAMP_TASK */
608 
609 union rcu_special {
610 	struct {
611 		u8			blocked;
612 		u8			need_qs;
613 		u8			exp_hint; /* Hint for performance. */
614 		u8			need_mb; /* Readers need smp_mb(). */
615 	} b; /* Bits. */
616 	u32 s; /* Set of bits. */
617 };
618 
619 enum perf_event_task_context {
620 	perf_invalid_context = -1,
621 	perf_hw_context = 0,
622 	perf_sw_context,
623 	perf_nr_task_contexts,
624 };
625 
626 struct wake_q_node {
627 	struct wake_q_node *next;
628 };
629 
630 struct task_struct {
631 #ifdef CONFIG_THREAD_INFO_IN_TASK
632 	/*
633 	 * For reasons of header soup (see current_thread_info()), this
634 	 * must be the first element of task_struct.
635 	 */
636 	struct thread_info		thread_info;
637 #endif
638 	/* -1 unrunnable, 0 runnable, >0 stopped: */
639 	volatile long			state;
640 
641 	/*
642 	 * This begins the randomizable portion of task_struct. Only
643 	 * scheduling-critical items should be added above here.
644 	 */
645 	randomized_struct_fields_start
646 
647 	void				*stack;
648 	refcount_t			usage;
649 	/* Per task flags (PF_*), defined further below: */
650 	unsigned int			flags;
651 	unsigned int			ptrace;
652 
653 #ifdef CONFIG_SMP
654 	int				on_cpu;
655 	struct __call_single_node	wake_entry;
656 #ifdef CONFIG_THREAD_INFO_IN_TASK
657 	/* Current CPU: */
658 	unsigned int			cpu;
659 #endif
660 	unsigned int			wakee_flips;
661 	unsigned long			wakee_flip_decay_ts;
662 	struct task_struct		*last_wakee;
663 
664 	/*
665 	 * recent_used_cpu is initially set as the last CPU used by a task
666 	 * that wakes affine another task. Waker/wakee relationships can
667 	 * push tasks around a CPU where each wakeup moves to the next one.
668 	 * Tracking a recently used CPU allows a quick search for a recently
669 	 * used CPU that may be idle.
670 	 */
671 	int				recent_used_cpu;
672 	int				wake_cpu;
673 #endif
674 	int				on_rq;
675 
676 	int				prio;
677 	int				static_prio;
678 	int				normal_prio;
679 	unsigned int			rt_priority;
680 
681 	const struct sched_class	*sched_class;
682 	struct sched_entity		se;
683 	struct sched_rt_entity		rt;
684 #ifdef CONFIG_CGROUP_SCHED
685 	struct task_group		*sched_task_group;
686 #endif
687 	struct sched_dl_entity		dl;
688 
689 #ifdef CONFIG_UCLAMP_TASK
690 	/*
691 	 * Clamp values requested for a scheduling entity.
692 	 * Must be updated with task_rq_lock() held.
693 	 */
694 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
695 	/*
696 	 * Effective clamp values used for a scheduling entity.
697 	 * Must be updated with task_rq_lock() held.
698 	 */
699 	struct uclamp_se		uclamp[UCLAMP_CNT];
700 #endif
701 
702 #ifdef CONFIG_PREEMPT_NOTIFIERS
703 	/* List of struct preempt_notifier: */
704 	struct hlist_head		preempt_notifiers;
705 #endif
706 
707 #ifdef CONFIG_BLK_DEV_IO_TRACE
708 	unsigned int			btrace_seq;
709 #endif
710 
711 	unsigned int			policy;
712 	int				nr_cpus_allowed;
713 	const cpumask_t			*cpus_ptr;
714 	cpumask_t			cpus_mask;
715 
716 #ifdef CONFIG_PREEMPT_RCU
717 	int				rcu_read_lock_nesting;
718 	union rcu_special		rcu_read_unlock_special;
719 	struct list_head		rcu_node_entry;
720 	struct rcu_node			*rcu_blocked_node;
721 #endif /* #ifdef CONFIG_PREEMPT_RCU */
722 
723 #ifdef CONFIG_TASKS_RCU
724 	unsigned long			rcu_tasks_nvcsw;
725 	u8				rcu_tasks_holdout;
726 	u8				rcu_tasks_idx;
727 	int				rcu_tasks_idle_cpu;
728 	struct list_head		rcu_tasks_holdout_list;
729 #endif /* #ifdef CONFIG_TASKS_RCU */
730 
731 #ifdef CONFIG_TASKS_TRACE_RCU
732 	int				trc_reader_nesting;
733 	int				trc_ipi_to_cpu;
734 	union rcu_special		trc_reader_special;
735 	bool				trc_reader_checked;
736 	struct list_head		trc_holdout_list;
737 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
738 
739 	struct sched_info		sched_info;
740 
741 	struct list_head		tasks;
742 #ifdef CONFIG_SMP
743 	struct plist_node		pushable_tasks;
744 	struct rb_node			pushable_dl_tasks;
745 #endif
746 
747 	struct mm_struct		*mm;
748 	struct mm_struct		*active_mm;
749 
750 	/* Per-thread vma caching: */
751 	struct vmacache			vmacache;
752 
753 #ifdef SPLIT_RSS_COUNTING
754 	struct task_rss_stat		rss_stat;
755 #endif
756 	int				exit_state;
757 	int				exit_code;
758 	int				exit_signal;
759 	/* The signal sent when the parent dies: */
760 	int				pdeath_signal;
761 	/* JOBCTL_*, siglock protected: */
762 	unsigned long			jobctl;
763 
764 	/* Used for emulating ABI behavior of previous Linux versions: */
765 	unsigned int			personality;
766 
767 	/* Scheduler bits, serialized by scheduler locks: */
768 	unsigned			sched_reset_on_fork:1;
769 	unsigned			sched_contributes_to_load:1;
770 	unsigned			sched_migrated:1;
771 	unsigned			sched_remote_wakeup:1;
772 #ifdef CONFIG_PSI
773 	unsigned			sched_psi_wake_requeue:1;
774 #endif
775 
776 	/* Force alignment to the next boundary: */
777 	unsigned			:0;
778 
779 	/* Unserialized, strictly 'current' */
780 
781 	/* Bit to tell LSMs we're in execve(): */
782 	unsigned			in_execve:1;
783 	unsigned			in_iowait:1;
784 #ifndef TIF_RESTORE_SIGMASK
785 	unsigned			restore_sigmask:1;
786 #endif
787 #ifdef CONFIG_MEMCG
788 	unsigned			in_user_fault:1;
789 #endif
790 #ifdef CONFIG_COMPAT_BRK
791 	unsigned			brk_randomized:1;
792 #endif
793 #ifdef CONFIG_CGROUPS
794 	/* disallow userland-initiated cgroup migration */
795 	unsigned			no_cgroup_migration:1;
796 	/* task is frozen/stopped (used by the cgroup freezer) */
797 	unsigned			frozen:1;
798 #endif
799 #ifdef CONFIG_BLK_CGROUP
800 	unsigned			use_memdelay:1;
801 #endif
802 #ifdef CONFIG_PSI
803 	/* Stalled due to lack of memory */
804 	unsigned			in_memstall:1;
805 #endif
806 
807 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
808 
809 	struct restart_block		restart_block;
810 
811 	pid_t				pid;
812 	pid_t				tgid;
813 
814 #ifdef CONFIG_STACKPROTECTOR
815 	/* Canary value for the -fstack-protector GCC feature: */
816 	unsigned long			stack_canary;
817 #endif
818 	/*
819 	 * Pointers to the (original) parent process, youngest child, younger sibling,
820 	 * older sibling, respectively.  (p->father can be replaced with
821 	 * p->real_parent->pid)
822 	 */
823 
824 	/* Real parent process: */
825 	struct task_struct __rcu	*real_parent;
826 
827 	/* Recipient of SIGCHLD, wait4() reports: */
828 	struct task_struct __rcu	*parent;
829 
830 	/*
831 	 * Children/sibling form the list of natural children:
832 	 */
833 	struct list_head		children;
834 	struct list_head		sibling;
835 	struct task_struct		*group_leader;
836 
837 	/*
838 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
839 	 *
840 	 * This includes both natural children and PTRACE_ATTACH targets.
841 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
842 	 */
843 	struct list_head		ptraced;
844 	struct list_head		ptrace_entry;
845 
846 	/* PID/PID hash table linkage. */
847 	struct pid			*thread_pid;
848 	struct hlist_node		pid_links[PIDTYPE_MAX];
849 	struct list_head		thread_group;
850 	struct list_head		thread_node;
851 
852 	struct completion		*vfork_done;
853 
854 	/* CLONE_CHILD_SETTID: */
855 	int __user			*set_child_tid;
856 
857 	/* CLONE_CHILD_CLEARTID: */
858 	int __user			*clear_child_tid;
859 
860 	u64				utime;
861 	u64				stime;
862 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
863 	u64				utimescaled;
864 	u64				stimescaled;
865 #endif
866 	u64				gtime;
867 	struct prev_cputime		prev_cputime;
868 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
869 	struct vtime			vtime;
870 #endif
871 
872 #ifdef CONFIG_NO_HZ_FULL
873 	atomic_t			tick_dep_mask;
874 #endif
875 	/* Context switch counts: */
876 	unsigned long			nvcsw;
877 	unsigned long			nivcsw;
878 
879 	/* Monotonic time in nsecs: */
880 	u64				start_time;
881 
882 	/* Boot based time in nsecs: */
883 	u64				start_boottime;
884 
885 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
886 	unsigned long			min_flt;
887 	unsigned long			maj_flt;
888 
889 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
890 	struct posix_cputimers		posix_cputimers;
891 
892 	/* Process credentials: */
893 
894 	/* Tracer's credentials at attach: */
895 	const struct cred __rcu		*ptracer_cred;
896 
897 	/* Objective and real subjective task credentials (COW): */
898 	const struct cred __rcu		*real_cred;
899 
900 	/* Effective (overridable) subjective task credentials (COW): */
901 	const struct cred __rcu		*cred;
902 
903 #ifdef CONFIG_KEYS
904 	/* Cached requested key. */
905 	struct key			*cached_requested_key;
906 #endif
907 
908 	/*
909 	 * executable name, excluding path.
910 	 *
911 	 * - normally initialized setup_new_exec()
912 	 * - access it with [gs]et_task_comm()
913 	 * - lock it with task_lock()
914 	 */
915 	char				comm[TASK_COMM_LEN];
916 
917 	struct nameidata		*nameidata;
918 
919 #ifdef CONFIG_SYSVIPC
920 	struct sysv_sem			sysvsem;
921 	struct sysv_shm			sysvshm;
922 #endif
923 #ifdef CONFIG_DETECT_HUNG_TASK
924 	unsigned long			last_switch_count;
925 	unsigned long			last_switch_time;
926 #endif
927 	/* Filesystem information: */
928 	struct fs_struct		*fs;
929 
930 	/* Open file information: */
931 	struct files_struct		*files;
932 
933 	/* Namespaces: */
934 	struct nsproxy			*nsproxy;
935 
936 	/* Signal handlers: */
937 	struct signal_struct		*signal;
938 	struct sighand_struct __rcu		*sighand;
939 	sigset_t			blocked;
940 	sigset_t			real_blocked;
941 	/* Restored if set_restore_sigmask() was used: */
942 	sigset_t			saved_sigmask;
943 	struct sigpending		pending;
944 	unsigned long			sas_ss_sp;
945 	size_t				sas_ss_size;
946 	unsigned int			sas_ss_flags;
947 
948 	struct callback_head		*task_works;
949 
950 #ifdef CONFIG_AUDIT
951 #ifdef CONFIG_AUDITSYSCALL
952 	struct audit_context		*audit_context;
953 #endif
954 	kuid_t				loginuid;
955 	unsigned int			sessionid;
956 #endif
957 	struct seccomp			seccomp;
958 
959 	/* Thread group tracking: */
960 	u64				parent_exec_id;
961 	u64				self_exec_id;
962 
963 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
964 	spinlock_t			alloc_lock;
965 
966 	/* Protection of the PI data structures: */
967 	raw_spinlock_t			pi_lock;
968 
969 	struct wake_q_node		wake_q;
970 
971 #ifdef CONFIG_RT_MUTEXES
972 	/* PI waiters blocked on a rt_mutex held by this task: */
973 	struct rb_root_cached		pi_waiters;
974 	/* Updated under owner's pi_lock and rq lock */
975 	struct task_struct		*pi_top_task;
976 	/* Deadlock detection and priority inheritance handling: */
977 	struct rt_mutex_waiter		*pi_blocked_on;
978 #endif
979 
980 #ifdef CONFIG_DEBUG_MUTEXES
981 	/* Mutex deadlock detection: */
982 	struct mutex_waiter		*blocked_on;
983 #endif
984 
985 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
986 	int				non_block_count;
987 #endif
988 
989 #ifdef CONFIG_TRACE_IRQFLAGS
990 	struct irqtrace_events		irqtrace;
991 	unsigned int			hardirq_threaded;
992 	u64				hardirq_chain_key;
993 	int				softirqs_enabled;
994 	int				softirq_context;
995 	int				irq_config;
996 #endif
997 
998 #ifdef CONFIG_LOCKDEP
999 # define MAX_LOCK_DEPTH			48UL
1000 	u64				curr_chain_key;
1001 	int				lockdep_depth;
1002 	unsigned int			lockdep_recursion;
1003 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1004 #endif
1005 
1006 #ifdef CONFIG_UBSAN
1007 	unsigned int			in_ubsan;
1008 #endif
1009 
1010 	/* Journalling filesystem info: */
1011 	void				*journal_info;
1012 
1013 	/* Stacked block device info: */
1014 	struct bio_list			*bio_list;
1015 
1016 #ifdef CONFIG_BLOCK
1017 	/* Stack plugging: */
1018 	struct blk_plug			*plug;
1019 #endif
1020 
1021 	/* VM state: */
1022 	struct reclaim_state		*reclaim_state;
1023 
1024 	struct backing_dev_info		*backing_dev_info;
1025 
1026 	struct io_context		*io_context;
1027 
1028 #ifdef CONFIG_COMPACTION
1029 	struct capture_control		*capture_control;
1030 #endif
1031 	/* Ptrace state: */
1032 	unsigned long			ptrace_message;
1033 	kernel_siginfo_t		*last_siginfo;
1034 
1035 	struct task_io_accounting	ioac;
1036 #ifdef CONFIG_PSI
1037 	/* Pressure stall state */
1038 	unsigned int			psi_flags;
1039 #endif
1040 #ifdef CONFIG_TASK_XACCT
1041 	/* Accumulated RSS usage: */
1042 	u64				acct_rss_mem1;
1043 	/* Accumulated virtual memory usage: */
1044 	u64				acct_vm_mem1;
1045 	/* stime + utime since last update: */
1046 	u64				acct_timexpd;
1047 #endif
1048 #ifdef CONFIG_CPUSETS
1049 	/* Protected by ->alloc_lock: */
1050 	nodemask_t			mems_allowed;
1051 	/* Seqence number to catch updates: */
1052 	seqcount_t			mems_allowed_seq;
1053 	int				cpuset_mem_spread_rotor;
1054 	int				cpuset_slab_spread_rotor;
1055 #endif
1056 #ifdef CONFIG_CGROUPS
1057 	/* Control Group info protected by css_set_lock: */
1058 	struct css_set __rcu		*cgroups;
1059 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1060 	struct list_head		cg_list;
1061 #endif
1062 #ifdef CONFIG_X86_CPU_RESCTRL
1063 	u32				closid;
1064 	u32				rmid;
1065 #endif
1066 #ifdef CONFIG_FUTEX
1067 	struct robust_list_head __user	*robust_list;
1068 #ifdef CONFIG_COMPAT
1069 	struct compat_robust_list_head __user *compat_robust_list;
1070 #endif
1071 	struct list_head		pi_state_list;
1072 	struct futex_pi_state		*pi_state_cache;
1073 	struct mutex			futex_exit_mutex;
1074 	unsigned int			futex_state;
1075 #endif
1076 #ifdef CONFIG_PERF_EVENTS
1077 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
1078 	struct mutex			perf_event_mutex;
1079 	struct list_head		perf_event_list;
1080 #endif
1081 #ifdef CONFIG_DEBUG_PREEMPT
1082 	unsigned long			preempt_disable_ip;
1083 #endif
1084 #ifdef CONFIG_NUMA
1085 	/* Protected by alloc_lock: */
1086 	struct mempolicy		*mempolicy;
1087 	short				il_prev;
1088 	short				pref_node_fork;
1089 #endif
1090 #ifdef CONFIG_NUMA_BALANCING
1091 	int				numa_scan_seq;
1092 	unsigned int			numa_scan_period;
1093 	unsigned int			numa_scan_period_max;
1094 	int				numa_preferred_nid;
1095 	unsigned long			numa_migrate_retry;
1096 	/* Migration stamp: */
1097 	u64				node_stamp;
1098 	u64				last_task_numa_placement;
1099 	u64				last_sum_exec_runtime;
1100 	struct callback_head		numa_work;
1101 
1102 	/*
1103 	 * This pointer is only modified for current in syscall and
1104 	 * pagefault context (and for tasks being destroyed), so it can be read
1105 	 * from any of the following contexts:
1106 	 *  - RCU read-side critical section
1107 	 *  - current->numa_group from everywhere
1108 	 *  - task's runqueue locked, task not running
1109 	 */
1110 	struct numa_group __rcu		*numa_group;
1111 
1112 	/*
1113 	 * numa_faults is an array split into four regions:
1114 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1115 	 * in this precise order.
1116 	 *
1117 	 * faults_memory: Exponential decaying average of faults on a per-node
1118 	 * basis. Scheduling placement decisions are made based on these
1119 	 * counts. The values remain static for the duration of a PTE scan.
1120 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1121 	 * hinting fault was incurred.
1122 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1123 	 * during the current scan window. When the scan completes, the counts
1124 	 * in faults_memory and faults_cpu decay and these values are copied.
1125 	 */
1126 	unsigned long			*numa_faults;
1127 	unsigned long			total_numa_faults;
1128 
1129 	/*
1130 	 * numa_faults_locality tracks if faults recorded during the last
1131 	 * scan window were remote/local or failed to migrate. The task scan
1132 	 * period is adapted based on the locality of the faults with different
1133 	 * weights depending on whether they were shared or private faults
1134 	 */
1135 	unsigned long			numa_faults_locality[3];
1136 
1137 	unsigned long			numa_pages_migrated;
1138 #endif /* CONFIG_NUMA_BALANCING */
1139 
1140 #ifdef CONFIG_RSEQ
1141 	struct rseq __user *rseq;
1142 	u32 rseq_sig;
1143 	/*
1144 	 * RmW on rseq_event_mask must be performed atomically
1145 	 * with respect to preemption.
1146 	 */
1147 	unsigned long rseq_event_mask;
1148 #endif
1149 
1150 	struct tlbflush_unmap_batch	tlb_ubc;
1151 
1152 	union {
1153 		refcount_t		rcu_users;
1154 		struct rcu_head		rcu;
1155 	};
1156 
1157 	/* Cache last used pipe for splice(): */
1158 	struct pipe_inode_info		*splice_pipe;
1159 
1160 	struct page_frag		task_frag;
1161 
1162 #ifdef CONFIG_TASK_DELAY_ACCT
1163 	struct task_delay_info		*delays;
1164 #endif
1165 
1166 #ifdef CONFIG_FAULT_INJECTION
1167 	int				make_it_fail;
1168 	unsigned int			fail_nth;
1169 #endif
1170 	/*
1171 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1172 	 * balance_dirty_pages() for a dirty throttling pause:
1173 	 */
1174 	int				nr_dirtied;
1175 	int				nr_dirtied_pause;
1176 	/* Start of a write-and-pause period: */
1177 	unsigned long			dirty_paused_when;
1178 
1179 #ifdef CONFIG_LATENCYTOP
1180 	int				latency_record_count;
1181 	struct latency_record		latency_record[LT_SAVECOUNT];
1182 #endif
1183 	/*
1184 	 * Time slack values; these are used to round up poll() and
1185 	 * select() etc timeout values. These are in nanoseconds.
1186 	 */
1187 	u64				timer_slack_ns;
1188 	u64				default_timer_slack_ns;
1189 
1190 #ifdef CONFIG_KASAN
1191 	unsigned int			kasan_depth;
1192 #endif
1193 
1194 #ifdef CONFIG_KCSAN
1195 	struct kcsan_ctx		kcsan_ctx;
1196 #ifdef CONFIG_TRACE_IRQFLAGS
1197 	struct irqtrace_events		kcsan_save_irqtrace;
1198 #endif
1199 #endif
1200 
1201 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1202 	/* Index of current stored address in ret_stack: */
1203 	int				curr_ret_stack;
1204 	int				curr_ret_depth;
1205 
1206 	/* Stack of return addresses for return function tracing: */
1207 	struct ftrace_ret_stack		*ret_stack;
1208 
1209 	/* Timestamp for last schedule: */
1210 	unsigned long long		ftrace_timestamp;
1211 
1212 	/*
1213 	 * Number of functions that haven't been traced
1214 	 * because of depth overrun:
1215 	 */
1216 	atomic_t			trace_overrun;
1217 
1218 	/* Pause tracing: */
1219 	atomic_t			tracing_graph_pause;
1220 #endif
1221 
1222 #ifdef CONFIG_TRACING
1223 	/* State flags for use by tracers: */
1224 	unsigned long			trace;
1225 
1226 	/* Bitmask and counter of trace recursion: */
1227 	unsigned long			trace_recursion;
1228 #endif /* CONFIG_TRACING */
1229 
1230 #ifdef CONFIG_KCOV
1231 	/* See kernel/kcov.c for more details. */
1232 
1233 	/* Coverage collection mode enabled for this task (0 if disabled): */
1234 	unsigned int			kcov_mode;
1235 
1236 	/* Size of the kcov_area: */
1237 	unsigned int			kcov_size;
1238 
1239 	/* Buffer for coverage collection: */
1240 	void				*kcov_area;
1241 
1242 	/* KCOV descriptor wired with this task or NULL: */
1243 	struct kcov			*kcov;
1244 
1245 	/* KCOV common handle for remote coverage collection: */
1246 	u64				kcov_handle;
1247 
1248 	/* KCOV sequence number: */
1249 	int				kcov_sequence;
1250 
1251 	/* Collect coverage from softirq context: */
1252 	unsigned int			kcov_softirq;
1253 #endif
1254 
1255 #ifdef CONFIG_MEMCG
1256 	struct mem_cgroup		*memcg_in_oom;
1257 	gfp_t				memcg_oom_gfp_mask;
1258 	int				memcg_oom_order;
1259 
1260 	/* Number of pages to reclaim on returning to userland: */
1261 	unsigned int			memcg_nr_pages_over_high;
1262 
1263 	/* Used by memcontrol for targeted memcg charge: */
1264 	struct mem_cgroup		*active_memcg;
1265 #endif
1266 
1267 #ifdef CONFIG_BLK_CGROUP
1268 	struct request_queue		*throttle_queue;
1269 #endif
1270 
1271 #ifdef CONFIG_UPROBES
1272 	struct uprobe_task		*utask;
1273 #endif
1274 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1275 	unsigned int			sequential_io;
1276 	unsigned int			sequential_io_avg;
1277 #endif
1278 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1279 	unsigned long			task_state_change;
1280 #endif
1281 	int				pagefault_disabled;
1282 #ifdef CONFIG_MMU
1283 	struct task_struct		*oom_reaper_list;
1284 #endif
1285 #ifdef CONFIG_VMAP_STACK
1286 	struct vm_struct		*stack_vm_area;
1287 #endif
1288 #ifdef CONFIG_THREAD_INFO_IN_TASK
1289 	/* A live task holds one reference: */
1290 	refcount_t			stack_refcount;
1291 #endif
1292 #ifdef CONFIG_LIVEPATCH
1293 	int patch_state;
1294 #endif
1295 #ifdef CONFIG_SECURITY
1296 	/* Used by LSM modules for access restriction: */
1297 	void				*security;
1298 #endif
1299 
1300 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1301 	unsigned long			lowest_stack;
1302 	unsigned long			prev_lowest_stack;
1303 #endif
1304 
1305 #ifdef CONFIG_X86_MCE
1306 	u64				mce_addr;
1307 	__u64				mce_ripv : 1,
1308 					mce_whole_page : 1,
1309 					__mce_reserved : 62;
1310 	struct callback_head		mce_kill_me;
1311 #endif
1312 
1313 	/*
1314 	 * New fields for task_struct should be added above here, so that
1315 	 * they are included in the randomized portion of task_struct.
1316 	 */
1317 	randomized_struct_fields_end
1318 
1319 	/* CPU-specific state of this task: */
1320 	struct thread_struct		thread;
1321 
1322 	/*
1323 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1324 	 * structure.  It *MUST* be at the end of 'task_struct'.
1325 	 *
1326 	 * Do not put anything below here!
1327 	 */
1328 };
1329 
1330 static inline struct pid *task_pid(struct task_struct *task)
1331 {
1332 	return task->thread_pid;
1333 }
1334 
1335 /*
1336  * the helpers to get the task's different pids as they are seen
1337  * from various namespaces
1338  *
1339  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1340  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1341  *                     current.
1342  * task_xid_nr_ns()  : id seen from the ns specified;
1343  *
1344  * see also pid_nr() etc in include/linux/pid.h
1345  */
1346 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1347 
1348 static inline pid_t task_pid_nr(struct task_struct *tsk)
1349 {
1350 	return tsk->pid;
1351 }
1352 
1353 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1354 {
1355 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1356 }
1357 
1358 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1359 {
1360 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1361 }
1362 
1363 
1364 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1365 {
1366 	return tsk->tgid;
1367 }
1368 
1369 /**
1370  * pid_alive - check that a task structure is not stale
1371  * @p: Task structure to be checked.
1372  *
1373  * Test if a process is not yet dead (at most zombie state)
1374  * If pid_alive fails, then pointers within the task structure
1375  * can be stale and must not be dereferenced.
1376  *
1377  * Return: 1 if the process is alive. 0 otherwise.
1378  */
1379 static inline int pid_alive(const struct task_struct *p)
1380 {
1381 	return p->thread_pid != NULL;
1382 }
1383 
1384 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1385 {
1386 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1387 }
1388 
1389 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1390 {
1391 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1392 }
1393 
1394 
1395 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1396 {
1397 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1398 }
1399 
1400 static inline pid_t task_session_vnr(struct task_struct *tsk)
1401 {
1402 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1403 }
1404 
1405 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1406 {
1407 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1408 }
1409 
1410 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1411 {
1412 	return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1413 }
1414 
1415 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1416 {
1417 	pid_t pid = 0;
1418 
1419 	rcu_read_lock();
1420 	if (pid_alive(tsk))
1421 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1422 	rcu_read_unlock();
1423 
1424 	return pid;
1425 }
1426 
1427 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1428 {
1429 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1430 }
1431 
1432 /* Obsolete, do not use: */
1433 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1434 {
1435 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1436 }
1437 
1438 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1439 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1440 
1441 static inline unsigned int task_state_index(struct task_struct *tsk)
1442 {
1443 	unsigned int tsk_state = READ_ONCE(tsk->state);
1444 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1445 
1446 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1447 
1448 	if (tsk_state == TASK_IDLE)
1449 		state = TASK_REPORT_IDLE;
1450 
1451 	return fls(state);
1452 }
1453 
1454 static inline char task_index_to_char(unsigned int state)
1455 {
1456 	static const char state_char[] = "RSDTtXZPI";
1457 
1458 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1459 
1460 	return state_char[state];
1461 }
1462 
1463 static inline char task_state_to_char(struct task_struct *tsk)
1464 {
1465 	return task_index_to_char(task_state_index(tsk));
1466 }
1467 
1468 /**
1469  * is_global_init - check if a task structure is init. Since init
1470  * is free to have sub-threads we need to check tgid.
1471  * @tsk: Task structure to be checked.
1472  *
1473  * Check if a task structure is the first user space task the kernel created.
1474  *
1475  * Return: 1 if the task structure is init. 0 otherwise.
1476  */
1477 static inline int is_global_init(struct task_struct *tsk)
1478 {
1479 	return task_tgid_nr(tsk) == 1;
1480 }
1481 
1482 extern struct pid *cad_pid;
1483 
1484 /*
1485  * Per process flags
1486  */
1487 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1488 #define PF_EXITING		0x00000004	/* Getting shut down */
1489 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1490 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1491 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1492 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1493 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1494 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1495 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1496 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1497 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1498 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1499 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1500 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1501 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1502 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1503 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1504 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1505 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1506 						 * I am cleaning dirty pages from some other bdi. */
1507 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1508 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1509 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1510 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1511 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1512 #define PF_MEMALLOC_NOCMA	0x10000000	/* All allocation request will have _GFP_MOVABLE cleared */
1513 #define PF_IO_WORKER		0x20000000	/* Task is an IO worker */
1514 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1515 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1516 
1517 /*
1518  * Only the _current_ task can read/write to tsk->flags, but other
1519  * tasks can access tsk->flags in readonly mode for example
1520  * with tsk_used_math (like during threaded core dumping).
1521  * There is however an exception to this rule during ptrace
1522  * or during fork: the ptracer task is allowed to write to the
1523  * child->flags of its traced child (same goes for fork, the parent
1524  * can write to the child->flags), because we're guaranteed the
1525  * child is not running and in turn not changing child->flags
1526  * at the same time the parent does it.
1527  */
1528 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1529 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1530 #define clear_used_math()			clear_stopped_child_used_math(current)
1531 #define set_used_math()				set_stopped_child_used_math(current)
1532 
1533 #define conditional_stopped_child_used_math(condition, child) \
1534 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1535 
1536 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1537 
1538 #define copy_to_stopped_child_used_math(child) \
1539 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1540 
1541 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1542 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1543 #define used_math()				tsk_used_math(current)
1544 
1545 static inline bool is_percpu_thread(void)
1546 {
1547 #ifdef CONFIG_SMP
1548 	return (current->flags & PF_NO_SETAFFINITY) &&
1549 		(current->nr_cpus_allowed  == 1);
1550 #else
1551 	return true;
1552 #endif
1553 }
1554 
1555 /* Per-process atomic flags. */
1556 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1557 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1558 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1559 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1560 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1561 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1562 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1563 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1564 
1565 #define TASK_PFA_TEST(name, func)					\
1566 	static inline bool task_##func(struct task_struct *p)		\
1567 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1568 
1569 #define TASK_PFA_SET(name, func)					\
1570 	static inline void task_set_##func(struct task_struct *p)	\
1571 	{ set_bit(PFA_##name, &p->atomic_flags); }
1572 
1573 #define TASK_PFA_CLEAR(name, func)					\
1574 	static inline void task_clear_##func(struct task_struct *p)	\
1575 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1576 
1577 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1578 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1579 
1580 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1581 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1582 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1583 
1584 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1585 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1586 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1587 
1588 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1589 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1590 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1591 
1592 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1593 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1594 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1595 
1596 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1597 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1598 
1599 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1600 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1601 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1602 
1603 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1604 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1605 
1606 static inline void
1607 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1608 {
1609 	current->flags &= ~flags;
1610 	current->flags |= orig_flags & flags;
1611 }
1612 
1613 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1614 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1615 #ifdef CONFIG_SMP
1616 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1617 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1618 #else
1619 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1620 {
1621 }
1622 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1623 {
1624 	if (!cpumask_test_cpu(0, new_mask))
1625 		return -EINVAL;
1626 	return 0;
1627 }
1628 #endif
1629 
1630 extern int yield_to(struct task_struct *p, bool preempt);
1631 extern void set_user_nice(struct task_struct *p, long nice);
1632 extern int task_prio(const struct task_struct *p);
1633 
1634 /**
1635  * task_nice - return the nice value of a given task.
1636  * @p: the task in question.
1637  *
1638  * Return: The nice value [ -20 ... 0 ... 19 ].
1639  */
1640 static inline int task_nice(const struct task_struct *p)
1641 {
1642 	return PRIO_TO_NICE((p)->static_prio);
1643 }
1644 
1645 extern int can_nice(const struct task_struct *p, const int nice);
1646 extern int task_curr(const struct task_struct *p);
1647 extern int idle_cpu(int cpu);
1648 extern int available_idle_cpu(int cpu);
1649 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1650 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1651 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1652 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1653 extern struct task_struct *idle_task(int cpu);
1654 
1655 /**
1656  * is_idle_task - is the specified task an idle task?
1657  * @p: the task in question.
1658  *
1659  * Return: 1 if @p is an idle task. 0 otherwise.
1660  */
1661 static inline bool is_idle_task(const struct task_struct *p)
1662 {
1663 	return !!(p->flags & PF_IDLE);
1664 }
1665 
1666 extern struct task_struct *curr_task(int cpu);
1667 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1668 
1669 void yield(void);
1670 
1671 union thread_union {
1672 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1673 	struct task_struct task;
1674 #endif
1675 #ifndef CONFIG_THREAD_INFO_IN_TASK
1676 	struct thread_info thread_info;
1677 #endif
1678 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1679 };
1680 
1681 #ifndef CONFIG_THREAD_INFO_IN_TASK
1682 extern struct thread_info init_thread_info;
1683 #endif
1684 
1685 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1686 
1687 #ifdef CONFIG_THREAD_INFO_IN_TASK
1688 static inline struct thread_info *task_thread_info(struct task_struct *task)
1689 {
1690 	return &task->thread_info;
1691 }
1692 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1693 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1694 #endif
1695 
1696 /*
1697  * find a task by one of its numerical ids
1698  *
1699  * find_task_by_pid_ns():
1700  *      finds a task by its pid in the specified namespace
1701  * find_task_by_vpid():
1702  *      finds a task by its virtual pid
1703  *
1704  * see also find_vpid() etc in include/linux/pid.h
1705  */
1706 
1707 extern struct task_struct *find_task_by_vpid(pid_t nr);
1708 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1709 
1710 /*
1711  * find a task by its virtual pid and get the task struct
1712  */
1713 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1714 
1715 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1716 extern int wake_up_process(struct task_struct *tsk);
1717 extern void wake_up_new_task(struct task_struct *tsk);
1718 
1719 #ifdef CONFIG_SMP
1720 extern void kick_process(struct task_struct *tsk);
1721 #else
1722 static inline void kick_process(struct task_struct *tsk) { }
1723 #endif
1724 
1725 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1726 
1727 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1728 {
1729 	__set_task_comm(tsk, from, false);
1730 }
1731 
1732 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1733 #define get_task_comm(buf, tsk) ({			\
1734 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1735 	__get_task_comm(buf, sizeof(buf), tsk);		\
1736 })
1737 
1738 #ifdef CONFIG_SMP
1739 static __always_inline void scheduler_ipi(void)
1740 {
1741 	/*
1742 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1743 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1744 	 * this IPI.
1745 	 */
1746 	preempt_fold_need_resched();
1747 }
1748 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1749 #else
1750 static inline void scheduler_ipi(void) { }
1751 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1752 {
1753 	return 1;
1754 }
1755 #endif
1756 
1757 /*
1758  * Set thread flags in other task's structures.
1759  * See asm/thread_info.h for TIF_xxxx flags available:
1760  */
1761 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1762 {
1763 	set_ti_thread_flag(task_thread_info(tsk), flag);
1764 }
1765 
1766 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1767 {
1768 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1769 }
1770 
1771 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1772 					  bool value)
1773 {
1774 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1775 }
1776 
1777 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1778 {
1779 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1780 }
1781 
1782 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1783 {
1784 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1785 }
1786 
1787 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1788 {
1789 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1790 }
1791 
1792 static inline void set_tsk_need_resched(struct task_struct *tsk)
1793 {
1794 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1795 }
1796 
1797 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1798 {
1799 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1800 }
1801 
1802 static inline int test_tsk_need_resched(struct task_struct *tsk)
1803 {
1804 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1805 }
1806 
1807 /*
1808  * cond_resched() and cond_resched_lock(): latency reduction via
1809  * explicit rescheduling in places that are safe. The return
1810  * value indicates whether a reschedule was done in fact.
1811  * cond_resched_lock() will drop the spinlock before scheduling,
1812  */
1813 #ifndef CONFIG_PREEMPTION
1814 extern int _cond_resched(void);
1815 #else
1816 static inline int _cond_resched(void) { return 0; }
1817 #endif
1818 
1819 #define cond_resched() ({			\
1820 	___might_sleep(__FILE__, __LINE__, 0);	\
1821 	_cond_resched();			\
1822 })
1823 
1824 extern int __cond_resched_lock(spinlock_t *lock);
1825 
1826 #define cond_resched_lock(lock) ({				\
1827 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1828 	__cond_resched_lock(lock);				\
1829 })
1830 
1831 static inline void cond_resched_rcu(void)
1832 {
1833 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1834 	rcu_read_unlock();
1835 	cond_resched();
1836 	rcu_read_lock();
1837 #endif
1838 }
1839 
1840 /*
1841  * Does a critical section need to be broken due to another
1842  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1843  * but a general need for low latency)
1844  */
1845 static inline int spin_needbreak(spinlock_t *lock)
1846 {
1847 #ifdef CONFIG_PREEMPTION
1848 	return spin_is_contended(lock);
1849 #else
1850 	return 0;
1851 #endif
1852 }
1853 
1854 static __always_inline bool need_resched(void)
1855 {
1856 	return unlikely(tif_need_resched());
1857 }
1858 
1859 /*
1860  * Wrappers for p->thread_info->cpu access. No-op on UP.
1861  */
1862 #ifdef CONFIG_SMP
1863 
1864 static inline unsigned int task_cpu(const struct task_struct *p)
1865 {
1866 #ifdef CONFIG_THREAD_INFO_IN_TASK
1867 	return READ_ONCE(p->cpu);
1868 #else
1869 	return READ_ONCE(task_thread_info(p)->cpu);
1870 #endif
1871 }
1872 
1873 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1874 
1875 #else
1876 
1877 static inline unsigned int task_cpu(const struct task_struct *p)
1878 {
1879 	return 0;
1880 }
1881 
1882 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1883 {
1884 }
1885 
1886 #endif /* CONFIG_SMP */
1887 
1888 /*
1889  * In order to reduce various lock holder preemption latencies provide an
1890  * interface to see if a vCPU is currently running or not.
1891  *
1892  * This allows us to terminate optimistic spin loops and block, analogous to
1893  * the native optimistic spin heuristic of testing if the lock owner task is
1894  * running or not.
1895  */
1896 #ifndef vcpu_is_preempted
1897 static inline bool vcpu_is_preempted(int cpu)
1898 {
1899 	return false;
1900 }
1901 #endif
1902 
1903 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1904 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1905 
1906 #ifndef TASK_SIZE_OF
1907 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1908 #endif
1909 
1910 #ifdef CONFIG_RSEQ
1911 
1912 /*
1913  * Map the event mask on the user-space ABI enum rseq_cs_flags
1914  * for direct mask checks.
1915  */
1916 enum rseq_event_mask_bits {
1917 	RSEQ_EVENT_PREEMPT_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1918 	RSEQ_EVENT_SIGNAL_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1919 	RSEQ_EVENT_MIGRATE_BIT	= RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1920 };
1921 
1922 enum rseq_event_mask {
1923 	RSEQ_EVENT_PREEMPT	= (1U << RSEQ_EVENT_PREEMPT_BIT),
1924 	RSEQ_EVENT_SIGNAL	= (1U << RSEQ_EVENT_SIGNAL_BIT),
1925 	RSEQ_EVENT_MIGRATE	= (1U << RSEQ_EVENT_MIGRATE_BIT),
1926 };
1927 
1928 static inline void rseq_set_notify_resume(struct task_struct *t)
1929 {
1930 	if (t->rseq)
1931 		set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1932 }
1933 
1934 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1935 
1936 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1937 					     struct pt_regs *regs)
1938 {
1939 	if (current->rseq)
1940 		__rseq_handle_notify_resume(ksig, regs);
1941 }
1942 
1943 static inline void rseq_signal_deliver(struct ksignal *ksig,
1944 				       struct pt_regs *regs)
1945 {
1946 	preempt_disable();
1947 	__set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1948 	preempt_enable();
1949 	rseq_handle_notify_resume(ksig, regs);
1950 }
1951 
1952 /* rseq_preempt() requires preemption to be disabled. */
1953 static inline void rseq_preempt(struct task_struct *t)
1954 {
1955 	__set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1956 	rseq_set_notify_resume(t);
1957 }
1958 
1959 /* rseq_migrate() requires preemption to be disabled. */
1960 static inline void rseq_migrate(struct task_struct *t)
1961 {
1962 	__set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1963 	rseq_set_notify_resume(t);
1964 }
1965 
1966 /*
1967  * If parent process has a registered restartable sequences area, the
1968  * child inherits. Unregister rseq for a clone with CLONE_VM set.
1969  */
1970 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1971 {
1972 	if (clone_flags & CLONE_VM) {
1973 		t->rseq = NULL;
1974 		t->rseq_sig = 0;
1975 		t->rseq_event_mask = 0;
1976 	} else {
1977 		t->rseq = current->rseq;
1978 		t->rseq_sig = current->rseq_sig;
1979 		t->rseq_event_mask = current->rseq_event_mask;
1980 	}
1981 }
1982 
1983 static inline void rseq_execve(struct task_struct *t)
1984 {
1985 	t->rseq = NULL;
1986 	t->rseq_sig = 0;
1987 	t->rseq_event_mask = 0;
1988 }
1989 
1990 #else
1991 
1992 static inline void rseq_set_notify_resume(struct task_struct *t)
1993 {
1994 }
1995 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1996 					     struct pt_regs *regs)
1997 {
1998 }
1999 static inline void rseq_signal_deliver(struct ksignal *ksig,
2000 				       struct pt_regs *regs)
2001 {
2002 }
2003 static inline void rseq_preempt(struct task_struct *t)
2004 {
2005 }
2006 static inline void rseq_migrate(struct task_struct *t)
2007 {
2008 }
2009 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2010 {
2011 }
2012 static inline void rseq_execve(struct task_struct *t)
2013 {
2014 }
2015 
2016 #endif
2017 
2018 #ifdef CONFIG_DEBUG_RSEQ
2019 
2020 void rseq_syscall(struct pt_regs *regs);
2021 
2022 #else
2023 
2024 static inline void rseq_syscall(struct pt_regs *regs)
2025 {
2026 }
2027 
2028 #endif
2029 
2030 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2031 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2032 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2033 
2034 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2035 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2036 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2037 
2038 int sched_trace_rq_cpu(struct rq *rq);
2039 int sched_trace_rq_nr_running(struct rq *rq);
2040 
2041 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2042 
2043 #endif
2044