xref: /linux/include/linux/sched.h (revision e3b9f1e81de2083f359bacd2a94bf1c024f2ede0)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4 
5 /*
6  * Define 'struct task_struct' and provide the main scheduler
7  * APIs (schedule(), wakeup variants, etc.)
8  */
9 
10 #include <uapi/linux/sched.h>
11 
12 #include <asm/current.h>
13 
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/resource.h>
25 #include <linux/latencytop.h>
26 #include <linux/sched/prio.h>
27 #include <linux/signal_types.h>
28 #include <linux/mm_types_task.h>
29 #include <linux/task_io_accounting.h>
30 
31 /* task_struct member predeclarations (sorted alphabetically): */
32 struct audit_context;
33 struct backing_dev_info;
34 struct bio_list;
35 struct blk_plug;
36 struct cfs_rq;
37 struct fs_struct;
38 struct futex_pi_state;
39 struct io_context;
40 struct mempolicy;
41 struct nameidata;
42 struct nsproxy;
43 struct perf_event_context;
44 struct pid_namespace;
45 struct pipe_inode_info;
46 struct rcu_node;
47 struct reclaim_state;
48 struct robust_list_head;
49 struct sched_attr;
50 struct sched_param;
51 struct seq_file;
52 struct sighand_struct;
53 struct signal_struct;
54 struct task_delay_info;
55 struct task_group;
56 
57 /*
58  * Task state bitmask. NOTE! These bits are also
59  * encoded in fs/proc/array.c: get_task_state().
60  *
61  * We have two separate sets of flags: task->state
62  * is about runnability, while task->exit_state are
63  * about the task exiting. Confusing, but this way
64  * modifying one set can't modify the other one by
65  * mistake.
66  */
67 
68 /* Used in tsk->state: */
69 #define TASK_RUNNING			0x0000
70 #define TASK_INTERRUPTIBLE		0x0001
71 #define TASK_UNINTERRUPTIBLE		0x0002
72 #define __TASK_STOPPED			0x0004
73 #define __TASK_TRACED			0x0008
74 /* Used in tsk->exit_state: */
75 #define EXIT_DEAD			0x0010
76 #define EXIT_ZOMBIE			0x0020
77 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
78 /* Used in tsk->state again: */
79 #define TASK_PARKED			0x0040
80 #define TASK_DEAD			0x0080
81 #define TASK_WAKEKILL			0x0100
82 #define TASK_WAKING			0x0200
83 #define TASK_NOLOAD			0x0400
84 #define TASK_NEW			0x0800
85 #define TASK_STATE_MAX			0x1000
86 
87 /* Convenience macros for the sake of set_current_state: */
88 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
89 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
90 #define TASK_TRACED			(TASK_WAKEKILL | __TASK_TRACED)
91 
92 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
93 
94 /* Convenience macros for the sake of wake_up(): */
95 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
96 #define TASK_ALL			(TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
97 
98 /* get_task_state(): */
99 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
100 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
101 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
102 					 TASK_PARKED)
103 
104 #define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
105 
106 #define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
107 
108 #define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
109 
110 #define task_contributes_to_load(task)	((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 					 (task->flags & PF_FROZEN) == 0 && \
112 					 (task->state & TASK_NOLOAD) == 0)
113 
114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
115 
116 #define __set_current_state(state_value)			\
117 	do {							\
118 		current->task_state_change = _THIS_IP_;		\
119 		current->state = (state_value);			\
120 	} while (0)
121 #define set_current_state(state_value)				\
122 	do {							\
123 		current->task_state_change = _THIS_IP_;		\
124 		smp_store_mb(current->state, (state_value));	\
125 	} while (0)
126 
127 #else
128 /*
129  * set_current_state() includes a barrier so that the write of current->state
130  * is correctly serialised wrt the caller's subsequent test of whether to
131  * actually sleep:
132  *
133  *   for (;;) {
134  *	set_current_state(TASK_UNINTERRUPTIBLE);
135  *	if (!need_sleep)
136  *		break;
137  *
138  *	schedule();
139  *   }
140  *   __set_current_state(TASK_RUNNING);
141  *
142  * If the caller does not need such serialisation (because, for instance, the
143  * condition test and condition change and wakeup are under the same lock) then
144  * use __set_current_state().
145  *
146  * The above is typically ordered against the wakeup, which does:
147  *
148  *	need_sleep = false;
149  *	wake_up_state(p, TASK_UNINTERRUPTIBLE);
150  *
151  * Where wake_up_state() (and all other wakeup primitives) imply enough
152  * barriers to order the store of the variable against wakeup.
153  *
154  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
155  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
156  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
157  *
158  * This is obviously fine, since they both store the exact same value.
159  *
160  * Also see the comments of try_to_wake_up().
161  */
162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0)
163 #define set_current_state(state_value)	 smp_store_mb(current->state, (state_value))
164 #endif
165 
166 /* Task command name length: */
167 #define TASK_COMM_LEN			16
168 
169 extern void scheduler_tick(void);
170 
171 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
172 
173 extern long schedule_timeout(long timeout);
174 extern long schedule_timeout_interruptible(long timeout);
175 extern long schedule_timeout_killable(long timeout);
176 extern long schedule_timeout_uninterruptible(long timeout);
177 extern long schedule_timeout_idle(long timeout);
178 asmlinkage void schedule(void);
179 extern void schedule_preempt_disabled(void);
180 
181 extern int __must_check io_schedule_prepare(void);
182 extern void io_schedule_finish(int token);
183 extern long io_schedule_timeout(long timeout);
184 extern void io_schedule(void);
185 
186 /**
187  * struct prev_cputime - snapshot of system and user cputime
188  * @utime: time spent in user mode
189  * @stime: time spent in system mode
190  * @lock: protects the above two fields
191  *
192  * Stores previous user/system time values such that we can guarantee
193  * monotonicity.
194  */
195 struct prev_cputime {
196 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
197 	u64				utime;
198 	u64				stime;
199 	raw_spinlock_t			lock;
200 #endif
201 };
202 
203 /**
204  * struct task_cputime - collected CPU time counts
205  * @utime:		time spent in user mode, in nanoseconds
206  * @stime:		time spent in kernel mode, in nanoseconds
207  * @sum_exec_runtime:	total time spent on the CPU, in nanoseconds
208  *
209  * This structure groups together three kinds of CPU time that are tracked for
210  * threads and thread groups.  Most things considering CPU time want to group
211  * these counts together and treat all three of them in parallel.
212  */
213 struct task_cputime {
214 	u64				utime;
215 	u64				stime;
216 	unsigned long long		sum_exec_runtime;
217 };
218 
219 /* Alternate field names when used on cache expirations: */
220 #define virt_exp			utime
221 #define prof_exp			stime
222 #define sched_exp			sum_exec_runtime
223 
224 enum vtime_state {
225 	/* Task is sleeping or running in a CPU with VTIME inactive: */
226 	VTIME_INACTIVE = 0,
227 	/* Task runs in userspace in a CPU with VTIME active: */
228 	VTIME_USER,
229 	/* Task runs in kernelspace in a CPU with VTIME active: */
230 	VTIME_SYS,
231 };
232 
233 struct vtime {
234 	seqcount_t		seqcount;
235 	unsigned long long	starttime;
236 	enum vtime_state	state;
237 	u64			utime;
238 	u64			stime;
239 	u64			gtime;
240 };
241 
242 struct sched_info {
243 #ifdef CONFIG_SCHED_INFO
244 	/* Cumulative counters: */
245 
246 	/* # of times we have run on this CPU: */
247 	unsigned long			pcount;
248 
249 	/* Time spent waiting on a runqueue: */
250 	unsigned long long		run_delay;
251 
252 	/* Timestamps: */
253 
254 	/* When did we last run on a CPU? */
255 	unsigned long long		last_arrival;
256 
257 	/* When were we last queued to run? */
258 	unsigned long long		last_queued;
259 
260 #endif /* CONFIG_SCHED_INFO */
261 };
262 
263 /*
264  * Integer metrics need fixed point arithmetic, e.g., sched/fair
265  * has a few: load, load_avg, util_avg, freq, and capacity.
266  *
267  * We define a basic fixed point arithmetic range, and then formalize
268  * all these metrics based on that basic range.
269  */
270 # define SCHED_FIXEDPOINT_SHIFT		10
271 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
272 
273 struct load_weight {
274 	unsigned long			weight;
275 	u32				inv_weight;
276 };
277 
278 /*
279  * The load_avg/util_avg accumulates an infinite geometric series
280  * (see __update_load_avg() in kernel/sched/fair.c).
281  *
282  * [load_avg definition]
283  *
284  *   load_avg = runnable% * scale_load_down(load)
285  *
286  * where runnable% is the time ratio that a sched_entity is runnable.
287  * For cfs_rq, it is the aggregated load_avg of all runnable and
288  * blocked sched_entities.
289  *
290  * load_avg may also take frequency scaling into account:
291  *
292  *   load_avg = runnable% * scale_load_down(load) * freq%
293  *
294  * where freq% is the CPU frequency normalized to the highest frequency.
295  *
296  * [util_avg definition]
297  *
298  *   util_avg = running% * SCHED_CAPACITY_SCALE
299  *
300  * where running% is the time ratio that a sched_entity is running on
301  * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
302  * and blocked sched_entities.
303  *
304  * util_avg may also factor frequency scaling and CPU capacity scaling:
305  *
306  *   util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
307  *
308  * where freq% is the same as above, and capacity% is the CPU capacity
309  * normalized to the greatest capacity (due to uarch differences, etc).
310  *
311  * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
312  * themselves are in the range of [0, 1]. To do fixed point arithmetics,
313  * we therefore scale them to as large a range as necessary. This is for
314  * example reflected by util_avg's SCHED_CAPACITY_SCALE.
315  *
316  * [Overflow issue]
317  *
318  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
319  * with the highest load (=88761), always runnable on a single cfs_rq,
320  * and should not overflow as the number already hits PID_MAX_LIMIT.
321  *
322  * For all other cases (including 32-bit kernels), struct load_weight's
323  * weight will overflow first before we do, because:
324  *
325  *    Max(load_avg) <= Max(load.weight)
326  *
327  * Then it is the load_weight's responsibility to consider overflow
328  * issues.
329  */
330 struct sched_avg {
331 	u64				last_update_time;
332 	u64				load_sum;
333 	u64				runnable_load_sum;
334 	u32				util_sum;
335 	u32				period_contrib;
336 	unsigned long			load_avg;
337 	unsigned long			runnable_load_avg;
338 	unsigned long			util_avg;
339 };
340 
341 struct sched_statistics {
342 #ifdef CONFIG_SCHEDSTATS
343 	u64				wait_start;
344 	u64				wait_max;
345 	u64				wait_count;
346 	u64				wait_sum;
347 	u64				iowait_count;
348 	u64				iowait_sum;
349 
350 	u64				sleep_start;
351 	u64				sleep_max;
352 	s64				sum_sleep_runtime;
353 
354 	u64				block_start;
355 	u64				block_max;
356 	u64				exec_max;
357 	u64				slice_max;
358 
359 	u64				nr_migrations_cold;
360 	u64				nr_failed_migrations_affine;
361 	u64				nr_failed_migrations_running;
362 	u64				nr_failed_migrations_hot;
363 	u64				nr_forced_migrations;
364 
365 	u64				nr_wakeups;
366 	u64				nr_wakeups_sync;
367 	u64				nr_wakeups_migrate;
368 	u64				nr_wakeups_local;
369 	u64				nr_wakeups_remote;
370 	u64				nr_wakeups_affine;
371 	u64				nr_wakeups_affine_attempts;
372 	u64				nr_wakeups_passive;
373 	u64				nr_wakeups_idle;
374 #endif
375 };
376 
377 struct sched_entity {
378 	/* For load-balancing: */
379 	struct load_weight		load;
380 	unsigned long			runnable_weight;
381 	struct rb_node			run_node;
382 	struct list_head		group_node;
383 	unsigned int			on_rq;
384 
385 	u64				exec_start;
386 	u64				sum_exec_runtime;
387 	u64				vruntime;
388 	u64				prev_sum_exec_runtime;
389 
390 	u64				nr_migrations;
391 
392 	struct sched_statistics		statistics;
393 
394 #ifdef CONFIG_FAIR_GROUP_SCHED
395 	int				depth;
396 	struct sched_entity		*parent;
397 	/* rq on which this entity is (to be) queued: */
398 	struct cfs_rq			*cfs_rq;
399 	/* rq "owned" by this entity/group: */
400 	struct cfs_rq			*my_q;
401 #endif
402 
403 #ifdef CONFIG_SMP
404 	/*
405 	 * Per entity load average tracking.
406 	 *
407 	 * Put into separate cache line so it does not
408 	 * collide with read-mostly values above.
409 	 */
410 	struct sched_avg		avg ____cacheline_aligned_in_smp;
411 #endif
412 };
413 
414 struct sched_rt_entity {
415 	struct list_head		run_list;
416 	unsigned long			timeout;
417 	unsigned long			watchdog_stamp;
418 	unsigned int			time_slice;
419 	unsigned short			on_rq;
420 	unsigned short			on_list;
421 
422 	struct sched_rt_entity		*back;
423 #ifdef CONFIG_RT_GROUP_SCHED
424 	struct sched_rt_entity		*parent;
425 	/* rq on which this entity is (to be) queued: */
426 	struct rt_rq			*rt_rq;
427 	/* rq "owned" by this entity/group: */
428 	struct rt_rq			*my_q;
429 #endif
430 } __randomize_layout;
431 
432 struct sched_dl_entity {
433 	struct rb_node			rb_node;
434 
435 	/*
436 	 * Original scheduling parameters. Copied here from sched_attr
437 	 * during sched_setattr(), they will remain the same until
438 	 * the next sched_setattr().
439 	 */
440 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
441 	u64				dl_deadline;	/* Relative deadline of each instance	*/
442 	u64				dl_period;	/* Separation of two instances (period) */
443 	u64				dl_bw;		/* dl_runtime / dl_period		*/
444 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
445 
446 	/*
447 	 * Actual scheduling parameters. Initialized with the values above,
448 	 * they are continously updated during task execution. Note that
449 	 * the remaining runtime could be < 0 in case we are in overrun.
450 	 */
451 	s64				runtime;	/* Remaining runtime for this instance	*/
452 	u64				deadline;	/* Absolute deadline for this instance	*/
453 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
454 
455 	/*
456 	 * Some bool flags:
457 	 *
458 	 * @dl_throttled tells if we exhausted the runtime. If so, the
459 	 * task has to wait for a replenishment to be performed at the
460 	 * next firing of dl_timer.
461 	 *
462 	 * @dl_boosted tells if we are boosted due to DI. If so we are
463 	 * outside bandwidth enforcement mechanism (but only until we
464 	 * exit the critical section);
465 	 *
466 	 * @dl_yielded tells if task gave up the CPU before consuming
467 	 * all its available runtime during the last job.
468 	 *
469 	 * @dl_non_contending tells if the task is inactive while still
470 	 * contributing to the active utilization. In other words, it
471 	 * indicates if the inactive timer has been armed and its handler
472 	 * has not been executed yet. This flag is useful to avoid race
473 	 * conditions between the inactive timer handler and the wakeup
474 	 * code.
475 	 *
476 	 * @dl_overrun tells if the task asked to be informed about runtime
477 	 * overruns.
478 	 */
479 	unsigned int			dl_throttled      : 1;
480 	unsigned int			dl_boosted        : 1;
481 	unsigned int			dl_yielded        : 1;
482 	unsigned int			dl_non_contending : 1;
483 	unsigned int			dl_overrun	  : 1;
484 
485 	/*
486 	 * Bandwidth enforcement timer. Each -deadline task has its
487 	 * own bandwidth to be enforced, thus we need one timer per task.
488 	 */
489 	struct hrtimer			dl_timer;
490 
491 	/*
492 	 * Inactive timer, responsible for decreasing the active utilization
493 	 * at the "0-lag time". When a -deadline task blocks, it contributes
494 	 * to GRUB's active utilization until the "0-lag time", hence a
495 	 * timer is needed to decrease the active utilization at the correct
496 	 * time.
497 	 */
498 	struct hrtimer inactive_timer;
499 };
500 
501 union rcu_special {
502 	struct {
503 		u8			blocked;
504 		u8			need_qs;
505 		u8			exp_need_qs;
506 
507 		/* Otherwise the compiler can store garbage here: */
508 		u8			pad;
509 	} b; /* Bits. */
510 	u32 s; /* Set of bits. */
511 };
512 
513 enum perf_event_task_context {
514 	perf_invalid_context = -1,
515 	perf_hw_context = 0,
516 	perf_sw_context,
517 	perf_nr_task_contexts,
518 };
519 
520 struct wake_q_node {
521 	struct wake_q_node *next;
522 };
523 
524 struct task_struct {
525 #ifdef CONFIG_THREAD_INFO_IN_TASK
526 	/*
527 	 * For reasons of header soup (see current_thread_info()), this
528 	 * must be the first element of task_struct.
529 	 */
530 	struct thread_info		thread_info;
531 #endif
532 	/* -1 unrunnable, 0 runnable, >0 stopped: */
533 	volatile long			state;
534 
535 	/*
536 	 * This begins the randomizable portion of task_struct. Only
537 	 * scheduling-critical items should be added above here.
538 	 */
539 	randomized_struct_fields_start
540 
541 	void				*stack;
542 	atomic_t			usage;
543 	/* Per task flags (PF_*), defined further below: */
544 	unsigned int			flags;
545 	unsigned int			ptrace;
546 
547 #ifdef CONFIG_SMP
548 	struct llist_node		wake_entry;
549 	int				on_cpu;
550 #ifdef CONFIG_THREAD_INFO_IN_TASK
551 	/* Current CPU: */
552 	unsigned int			cpu;
553 #endif
554 	unsigned int			wakee_flips;
555 	unsigned long			wakee_flip_decay_ts;
556 	struct task_struct		*last_wakee;
557 
558 	/*
559 	 * recent_used_cpu is initially set as the last CPU used by a task
560 	 * that wakes affine another task. Waker/wakee relationships can
561 	 * push tasks around a CPU where each wakeup moves to the next one.
562 	 * Tracking a recently used CPU allows a quick search for a recently
563 	 * used CPU that may be idle.
564 	 */
565 	int				recent_used_cpu;
566 	int				wake_cpu;
567 #endif
568 	int				on_rq;
569 
570 	int				prio;
571 	int				static_prio;
572 	int				normal_prio;
573 	unsigned int			rt_priority;
574 
575 	const struct sched_class	*sched_class;
576 	struct sched_entity		se;
577 	struct sched_rt_entity		rt;
578 #ifdef CONFIG_CGROUP_SCHED
579 	struct task_group		*sched_task_group;
580 #endif
581 	struct sched_dl_entity		dl;
582 
583 #ifdef CONFIG_PREEMPT_NOTIFIERS
584 	/* List of struct preempt_notifier: */
585 	struct hlist_head		preempt_notifiers;
586 #endif
587 
588 #ifdef CONFIG_BLK_DEV_IO_TRACE
589 	unsigned int			btrace_seq;
590 #endif
591 
592 	unsigned int			policy;
593 	int				nr_cpus_allowed;
594 	cpumask_t			cpus_allowed;
595 
596 #ifdef CONFIG_PREEMPT_RCU
597 	int				rcu_read_lock_nesting;
598 	union rcu_special		rcu_read_unlock_special;
599 	struct list_head		rcu_node_entry;
600 	struct rcu_node			*rcu_blocked_node;
601 #endif /* #ifdef CONFIG_PREEMPT_RCU */
602 
603 #ifdef CONFIG_TASKS_RCU
604 	unsigned long			rcu_tasks_nvcsw;
605 	u8				rcu_tasks_holdout;
606 	u8				rcu_tasks_idx;
607 	int				rcu_tasks_idle_cpu;
608 	struct list_head		rcu_tasks_holdout_list;
609 #endif /* #ifdef CONFIG_TASKS_RCU */
610 
611 	struct sched_info		sched_info;
612 
613 	struct list_head		tasks;
614 #ifdef CONFIG_SMP
615 	struct plist_node		pushable_tasks;
616 	struct rb_node			pushable_dl_tasks;
617 #endif
618 
619 	struct mm_struct		*mm;
620 	struct mm_struct		*active_mm;
621 
622 	/* Per-thread vma caching: */
623 	struct vmacache			vmacache;
624 
625 #ifdef SPLIT_RSS_COUNTING
626 	struct task_rss_stat		rss_stat;
627 #endif
628 	int				exit_state;
629 	int				exit_code;
630 	int				exit_signal;
631 	/* The signal sent when the parent dies: */
632 	int				pdeath_signal;
633 	/* JOBCTL_*, siglock protected: */
634 	unsigned long			jobctl;
635 
636 	/* Used for emulating ABI behavior of previous Linux versions: */
637 	unsigned int			personality;
638 
639 	/* Scheduler bits, serialized by scheduler locks: */
640 	unsigned			sched_reset_on_fork:1;
641 	unsigned			sched_contributes_to_load:1;
642 	unsigned			sched_migrated:1;
643 	unsigned			sched_remote_wakeup:1;
644 	/* Force alignment to the next boundary: */
645 	unsigned			:0;
646 
647 	/* Unserialized, strictly 'current' */
648 
649 	/* Bit to tell LSMs we're in execve(): */
650 	unsigned			in_execve:1;
651 	unsigned			in_iowait:1;
652 #ifndef TIF_RESTORE_SIGMASK
653 	unsigned			restore_sigmask:1;
654 #endif
655 #ifdef CONFIG_MEMCG
656 	unsigned			memcg_may_oom:1;
657 #ifndef CONFIG_SLOB
658 	unsigned			memcg_kmem_skip_account:1;
659 #endif
660 #endif
661 #ifdef CONFIG_COMPAT_BRK
662 	unsigned			brk_randomized:1;
663 #endif
664 #ifdef CONFIG_CGROUPS
665 	/* disallow userland-initiated cgroup migration */
666 	unsigned			no_cgroup_migration:1;
667 #endif
668 
669 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
670 
671 	struct restart_block		restart_block;
672 
673 	pid_t				pid;
674 	pid_t				tgid;
675 
676 #ifdef CONFIG_CC_STACKPROTECTOR
677 	/* Canary value for the -fstack-protector GCC feature: */
678 	unsigned long			stack_canary;
679 #endif
680 	/*
681 	 * Pointers to the (original) parent process, youngest child, younger sibling,
682 	 * older sibling, respectively.  (p->father can be replaced with
683 	 * p->real_parent->pid)
684 	 */
685 
686 	/* Real parent process: */
687 	struct task_struct __rcu	*real_parent;
688 
689 	/* Recipient of SIGCHLD, wait4() reports: */
690 	struct task_struct __rcu	*parent;
691 
692 	/*
693 	 * Children/sibling form the list of natural children:
694 	 */
695 	struct list_head		children;
696 	struct list_head		sibling;
697 	struct task_struct		*group_leader;
698 
699 	/*
700 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
701 	 *
702 	 * This includes both natural children and PTRACE_ATTACH targets.
703 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
704 	 */
705 	struct list_head		ptraced;
706 	struct list_head		ptrace_entry;
707 
708 	/* PID/PID hash table linkage. */
709 	struct pid_link			pids[PIDTYPE_MAX];
710 	struct list_head		thread_group;
711 	struct list_head		thread_node;
712 
713 	struct completion		*vfork_done;
714 
715 	/* CLONE_CHILD_SETTID: */
716 	int __user			*set_child_tid;
717 
718 	/* CLONE_CHILD_CLEARTID: */
719 	int __user			*clear_child_tid;
720 
721 	u64				utime;
722 	u64				stime;
723 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
724 	u64				utimescaled;
725 	u64				stimescaled;
726 #endif
727 	u64				gtime;
728 	struct prev_cputime		prev_cputime;
729 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
730 	struct vtime			vtime;
731 #endif
732 
733 #ifdef CONFIG_NO_HZ_FULL
734 	atomic_t			tick_dep_mask;
735 #endif
736 	/* Context switch counts: */
737 	unsigned long			nvcsw;
738 	unsigned long			nivcsw;
739 
740 	/* Monotonic time in nsecs: */
741 	u64				start_time;
742 
743 	/* Boot based time in nsecs: */
744 	u64				real_start_time;
745 
746 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
747 	unsigned long			min_flt;
748 	unsigned long			maj_flt;
749 
750 #ifdef CONFIG_POSIX_TIMERS
751 	struct task_cputime		cputime_expires;
752 	struct list_head		cpu_timers[3];
753 #endif
754 
755 	/* Process credentials: */
756 
757 	/* Tracer's credentials at attach: */
758 	const struct cred __rcu		*ptracer_cred;
759 
760 	/* Objective and real subjective task credentials (COW): */
761 	const struct cred __rcu		*real_cred;
762 
763 	/* Effective (overridable) subjective task credentials (COW): */
764 	const struct cred __rcu		*cred;
765 
766 	/*
767 	 * executable name, excluding path.
768 	 *
769 	 * - normally initialized setup_new_exec()
770 	 * - access it with [gs]et_task_comm()
771 	 * - lock it with task_lock()
772 	 */
773 	char				comm[TASK_COMM_LEN];
774 
775 	struct nameidata		*nameidata;
776 
777 #ifdef CONFIG_SYSVIPC
778 	struct sysv_sem			sysvsem;
779 	struct sysv_shm			sysvshm;
780 #endif
781 #ifdef CONFIG_DETECT_HUNG_TASK
782 	unsigned long			last_switch_count;
783 #endif
784 	/* Filesystem information: */
785 	struct fs_struct		*fs;
786 
787 	/* Open file information: */
788 	struct files_struct		*files;
789 
790 	/* Namespaces: */
791 	struct nsproxy			*nsproxy;
792 
793 	/* Signal handlers: */
794 	struct signal_struct		*signal;
795 	struct sighand_struct		*sighand;
796 	sigset_t			blocked;
797 	sigset_t			real_blocked;
798 	/* Restored if set_restore_sigmask() was used: */
799 	sigset_t			saved_sigmask;
800 	struct sigpending		pending;
801 	unsigned long			sas_ss_sp;
802 	size_t				sas_ss_size;
803 	unsigned int			sas_ss_flags;
804 
805 	struct callback_head		*task_works;
806 
807 	struct audit_context		*audit_context;
808 #ifdef CONFIG_AUDITSYSCALL
809 	kuid_t				loginuid;
810 	unsigned int			sessionid;
811 #endif
812 	struct seccomp			seccomp;
813 
814 	/* Thread group tracking: */
815 	u32				parent_exec_id;
816 	u32				self_exec_id;
817 
818 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
819 	spinlock_t			alloc_lock;
820 
821 	/* Protection of the PI data structures: */
822 	raw_spinlock_t			pi_lock;
823 
824 	struct wake_q_node		wake_q;
825 
826 #ifdef CONFIG_RT_MUTEXES
827 	/* PI waiters blocked on a rt_mutex held by this task: */
828 	struct rb_root_cached		pi_waiters;
829 	/* Updated under owner's pi_lock and rq lock */
830 	struct task_struct		*pi_top_task;
831 	/* Deadlock detection and priority inheritance handling: */
832 	struct rt_mutex_waiter		*pi_blocked_on;
833 #endif
834 
835 #ifdef CONFIG_DEBUG_MUTEXES
836 	/* Mutex deadlock detection: */
837 	struct mutex_waiter		*blocked_on;
838 #endif
839 
840 #ifdef CONFIG_TRACE_IRQFLAGS
841 	unsigned int			irq_events;
842 	unsigned long			hardirq_enable_ip;
843 	unsigned long			hardirq_disable_ip;
844 	unsigned int			hardirq_enable_event;
845 	unsigned int			hardirq_disable_event;
846 	int				hardirqs_enabled;
847 	int				hardirq_context;
848 	unsigned long			softirq_disable_ip;
849 	unsigned long			softirq_enable_ip;
850 	unsigned int			softirq_disable_event;
851 	unsigned int			softirq_enable_event;
852 	int				softirqs_enabled;
853 	int				softirq_context;
854 #endif
855 
856 #ifdef CONFIG_LOCKDEP
857 # define MAX_LOCK_DEPTH			48UL
858 	u64				curr_chain_key;
859 	int				lockdep_depth;
860 	unsigned int			lockdep_recursion;
861 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
862 #endif
863 
864 #ifdef CONFIG_UBSAN
865 	unsigned int			in_ubsan;
866 #endif
867 
868 	/* Journalling filesystem info: */
869 	void				*journal_info;
870 
871 	/* Stacked block device info: */
872 	struct bio_list			*bio_list;
873 
874 #ifdef CONFIG_BLOCK
875 	/* Stack plugging: */
876 	struct blk_plug			*plug;
877 #endif
878 
879 	/* VM state: */
880 	struct reclaim_state		*reclaim_state;
881 
882 	struct backing_dev_info		*backing_dev_info;
883 
884 	struct io_context		*io_context;
885 
886 	/* Ptrace state: */
887 	unsigned long			ptrace_message;
888 	siginfo_t			*last_siginfo;
889 
890 	struct task_io_accounting	ioac;
891 #ifdef CONFIG_TASK_XACCT
892 	/* Accumulated RSS usage: */
893 	u64				acct_rss_mem1;
894 	/* Accumulated virtual memory usage: */
895 	u64				acct_vm_mem1;
896 	/* stime + utime since last update: */
897 	u64				acct_timexpd;
898 #endif
899 #ifdef CONFIG_CPUSETS
900 	/* Protected by ->alloc_lock: */
901 	nodemask_t			mems_allowed;
902 	/* Seqence number to catch updates: */
903 	seqcount_t			mems_allowed_seq;
904 	int				cpuset_mem_spread_rotor;
905 	int				cpuset_slab_spread_rotor;
906 #endif
907 #ifdef CONFIG_CGROUPS
908 	/* Control Group info protected by css_set_lock: */
909 	struct css_set __rcu		*cgroups;
910 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
911 	struct list_head		cg_list;
912 #endif
913 #ifdef CONFIG_INTEL_RDT
914 	u32				closid;
915 	u32				rmid;
916 #endif
917 #ifdef CONFIG_FUTEX
918 	struct robust_list_head __user	*robust_list;
919 #ifdef CONFIG_COMPAT
920 	struct compat_robust_list_head __user *compat_robust_list;
921 #endif
922 	struct list_head		pi_state_list;
923 	struct futex_pi_state		*pi_state_cache;
924 #endif
925 #ifdef CONFIG_PERF_EVENTS
926 	struct perf_event_context	*perf_event_ctxp[perf_nr_task_contexts];
927 	struct mutex			perf_event_mutex;
928 	struct list_head		perf_event_list;
929 #endif
930 #ifdef CONFIG_DEBUG_PREEMPT
931 	unsigned long			preempt_disable_ip;
932 #endif
933 #ifdef CONFIG_NUMA
934 	/* Protected by alloc_lock: */
935 	struct mempolicy		*mempolicy;
936 	short				il_prev;
937 	short				pref_node_fork;
938 #endif
939 #ifdef CONFIG_NUMA_BALANCING
940 	int				numa_scan_seq;
941 	unsigned int			numa_scan_period;
942 	unsigned int			numa_scan_period_max;
943 	int				numa_preferred_nid;
944 	unsigned long			numa_migrate_retry;
945 	/* Migration stamp: */
946 	u64				node_stamp;
947 	u64				last_task_numa_placement;
948 	u64				last_sum_exec_runtime;
949 	struct callback_head		numa_work;
950 
951 	struct list_head		numa_entry;
952 	struct numa_group		*numa_group;
953 
954 	/*
955 	 * numa_faults is an array split into four regions:
956 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
957 	 * in this precise order.
958 	 *
959 	 * faults_memory: Exponential decaying average of faults on a per-node
960 	 * basis. Scheduling placement decisions are made based on these
961 	 * counts. The values remain static for the duration of a PTE scan.
962 	 * faults_cpu: Track the nodes the process was running on when a NUMA
963 	 * hinting fault was incurred.
964 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
965 	 * during the current scan window. When the scan completes, the counts
966 	 * in faults_memory and faults_cpu decay and these values are copied.
967 	 */
968 	unsigned long			*numa_faults;
969 	unsigned long			total_numa_faults;
970 
971 	/*
972 	 * numa_faults_locality tracks if faults recorded during the last
973 	 * scan window were remote/local or failed to migrate. The task scan
974 	 * period is adapted based on the locality of the faults with different
975 	 * weights depending on whether they were shared or private faults
976 	 */
977 	unsigned long			numa_faults_locality[3];
978 
979 	unsigned long			numa_pages_migrated;
980 #endif /* CONFIG_NUMA_BALANCING */
981 
982 	struct tlbflush_unmap_batch	tlb_ubc;
983 
984 	struct rcu_head			rcu;
985 
986 	/* Cache last used pipe for splice(): */
987 	struct pipe_inode_info		*splice_pipe;
988 
989 	struct page_frag		task_frag;
990 
991 #ifdef CONFIG_TASK_DELAY_ACCT
992 	struct task_delay_info		*delays;
993 #endif
994 
995 #ifdef CONFIG_FAULT_INJECTION
996 	int				make_it_fail;
997 	unsigned int			fail_nth;
998 #endif
999 	/*
1000 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1001 	 * balance_dirty_pages() for a dirty throttling pause:
1002 	 */
1003 	int				nr_dirtied;
1004 	int				nr_dirtied_pause;
1005 	/* Start of a write-and-pause period: */
1006 	unsigned long			dirty_paused_when;
1007 
1008 #ifdef CONFIG_LATENCYTOP
1009 	int				latency_record_count;
1010 	struct latency_record		latency_record[LT_SAVECOUNT];
1011 #endif
1012 	/*
1013 	 * Time slack values; these are used to round up poll() and
1014 	 * select() etc timeout values. These are in nanoseconds.
1015 	 */
1016 	u64				timer_slack_ns;
1017 	u64				default_timer_slack_ns;
1018 
1019 #ifdef CONFIG_KASAN
1020 	unsigned int			kasan_depth;
1021 #endif
1022 
1023 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1024 	/* Index of current stored address in ret_stack: */
1025 	int				curr_ret_stack;
1026 
1027 	/* Stack of return addresses for return function tracing: */
1028 	struct ftrace_ret_stack		*ret_stack;
1029 
1030 	/* Timestamp for last schedule: */
1031 	unsigned long long		ftrace_timestamp;
1032 
1033 	/*
1034 	 * Number of functions that haven't been traced
1035 	 * because of depth overrun:
1036 	 */
1037 	atomic_t			trace_overrun;
1038 
1039 	/* Pause tracing: */
1040 	atomic_t			tracing_graph_pause;
1041 #endif
1042 
1043 #ifdef CONFIG_TRACING
1044 	/* State flags for use by tracers: */
1045 	unsigned long			trace;
1046 
1047 	/* Bitmask and counter of trace recursion: */
1048 	unsigned long			trace_recursion;
1049 #endif /* CONFIG_TRACING */
1050 
1051 #ifdef CONFIG_KCOV
1052 	/* Coverage collection mode enabled for this task (0 if disabled): */
1053 	enum kcov_mode			kcov_mode;
1054 
1055 	/* Size of the kcov_area: */
1056 	unsigned int			kcov_size;
1057 
1058 	/* Buffer for coverage collection: */
1059 	void				*kcov_area;
1060 
1061 	/* KCOV descriptor wired with this task or NULL: */
1062 	struct kcov			*kcov;
1063 #endif
1064 
1065 #ifdef CONFIG_MEMCG
1066 	struct mem_cgroup		*memcg_in_oom;
1067 	gfp_t				memcg_oom_gfp_mask;
1068 	int				memcg_oom_order;
1069 
1070 	/* Number of pages to reclaim on returning to userland: */
1071 	unsigned int			memcg_nr_pages_over_high;
1072 #endif
1073 
1074 #ifdef CONFIG_UPROBES
1075 	struct uprobe_task		*utask;
1076 #endif
1077 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1078 	unsigned int			sequential_io;
1079 	unsigned int			sequential_io_avg;
1080 #endif
1081 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1082 	unsigned long			task_state_change;
1083 #endif
1084 	int				pagefault_disabled;
1085 #ifdef CONFIG_MMU
1086 	struct task_struct		*oom_reaper_list;
1087 #endif
1088 #ifdef CONFIG_VMAP_STACK
1089 	struct vm_struct		*stack_vm_area;
1090 #endif
1091 #ifdef CONFIG_THREAD_INFO_IN_TASK
1092 	/* A live task holds one reference: */
1093 	atomic_t			stack_refcount;
1094 #endif
1095 #ifdef CONFIG_LIVEPATCH
1096 	int patch_state;
1097 #endif
1098 #ifdef CONFIG_SECURITY
1099 	/* Used by LSM modules for access restriction: */
1100 	void				*security;
1101 #endif
1102 
1103 	/*
1104 	 * New fields for task_struct should be added above here, so that
1105 	 * they are included in the randomized portion of task_struct.
1106 	 */
1107 	randomized_struct_fields_end
1108 
1109 	/* CPU-specific state of this task: */
1110 	struct thread_struct		thread;
1111 
1112 	/*
1113 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1114 	 * structure.  It *MUST* be at the end of 'task_struct'.
1115 	 *
1116 	 * Do not put anything below here!
1117 	 */
1118 };
1119 
1120 static inline struct pid *task_pid(struct task_struct *task)
1121 {
1122 	return task->pids[PIDTYPE_PID].pid;
1123 }
1124 
1125 static inline struct pid *task_tgid(struct task_struct *task)
1126 {
1127 	return task->group_leader->pids[PIDTYPE_PID].pid;
1128 }
1129 
1130 /*
1131  * Without tasklist or RCU lock it is not safe to dereference
1132  * the result of task_pgrp/task_session even if task == current,
1133  * we can race with another thread doing sys_setsid/sys_setpgid.
1134  */
1135 static inline struct pid *task_pgrp(struct task_struct *task)
1136 {
1137 	return task->group_leader->pids[PIDTYPE_PGID].pid;
1138 }
1139 
1140 static inline struct pid *task_session(struct task_struct *task)
1141 {
1142 	return task->group_leader->pids[PIDTYPE_SID].pid;
1143 }
1144 
1145 /*
1146  * the helpers to get the task's different pids as they are seen
1147  * from various namespaces
1148  *
1149  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1150  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1151  *                     current.
1152  * task_xid_nr_ns()  : id seen from the ns specified;
1153  *
1154  * see also pid_nr() etc in include/linux/pid.h
1155  */
1156 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1157 
1158 static inline pid_t task_pid_nr(struct task_struct *tsk)
1159 {
1160 	return tsk->pid;
1161 }
1162 
1163 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1164 {
1165 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1166 }
1167 
1168 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1169 {
1170 	return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1171 }
1172 
1173 
1174 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1175 {
1176 	return tsk->tgid;
1177 }
1178 
1179 /**
1180  * pid_alive - check that a task structure is not stale
1181  * @p: Task structure to be checked.
1182  *
1183  * Test if a process is not yet dead (at most zombie state)
1184  * If pid_alive fails, then pointers within the task structure
1185  * can be stale and must not be dereferenced.
1186  *
1187  * Return: 1 if the process is alive. 0 otherwise.
1188  */
1189 static inline int pid_alive(const struct task_struct *p)
1190 {
1191 	return p->pids[PIDTYPE_PID].pid != NULL;
1192 }
1193 
1194 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1195 {
1196 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1197 }
1198 
1199 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1200 {
1201 	return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1202 }
1203 
1204 
1205 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1206 {
1207 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1208 }
1209 
1210 static inline pid_t task_session_vnr(struct task_struct *tsk)
1211 {
1212 	return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1213 }
1214 
1215 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1216 {
1217 	return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
1218 }
1219 
1220 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1221 {
1222 	return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
1223 }
1224 
1225 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1226 {
1227 	pid_t pid = 0;
1228 
1229 	rcu_read_lock();
1230 	if (pid_alive(tsk))
1231 		pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1232 	rcu_read_unlock();
1233 
1234 	return pid;
1235 }
1236 
1237 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1238 {
1239 	return task_ppid_nr_ns(tsk, &init_pid_ns);
1240 }
1241 
1242 /* Obsolete, do not use: */
1243 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1244 {
1245 	return task_pgrp_nr_ns(tsk, &init_pid_ns);
1246 }
1247 
1248 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1249 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1250 
1251 static inline unsigned int task_state_index(struct task_struct *tsk)
1252 {
1253 	unsigned int tsk_state = READ_ONCE(tsk->state);
1254 	unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1255 
1256 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1257 
1258 	if (tsk_state == TASK_IDLE)
1259 		state = TASK_REPORT_IDLE;
1260 
1261 	return fls(state);
1262 }
1263 
1264 static inline char task_index_to_char(unsigned int state)
1265 {
1266 	static const char state_char[] = "RSDTtXZPI";
1267 
1268 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1269 
1270 	return state_char[state];
1271 }
1272 
1273 static inline char task_state_to_char(struct task_struct *tsk)
1274 {
1275 	return task_index_to_char(task_state_index(tsk));
1276 }
1277 
1278 /**
1279  * is_global_init - check if a task structure is init. Since init
1280  * is free to have sub-threads we need to check tgid.
1281  * @tsk: Task structure to be checked.
1282  *
1283  * Check if a task structure is the first user space task the kernel created.
1284  *
1285  * Return: 1 if the task structure is init. 0 otherwise.
1286  */
1287 static inline int is_global_init(struct task_struct *tsk)
1288 {
1289 	return task_tgid_nr(tsk) == 1;
1290 }
1291 
1292 extern struct pid *cad_pid;
1293 
1294 /*
1295  * Per process flags
1296  */
1297 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1298 #define PF_EXITING		0x00000004	/* Getting shut down */
1299 #define PF_EXITPIDONE		0x00000008	/* PI exit done on shut down */
1300 #define PF_VCPU			0x00000010	/* I'm a virtual CPU */
1301 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1302 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1303 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1304 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1305 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1306 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1307 #define PF_MEMALLOC		0x00000800	/* Allocating memory */
1308 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1309 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1310 #define PF_USED_ASYNC		0x00004000	/* Used async_schedule*(), used by module init */
1311 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1312 #define PF_FROZEN		0x00010000	/* Frozen for system suspend */
1313 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1314 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocation requests will inherit GFP_NOFS */
1315 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocation requests will inherit GFP_NOIO */
1316 #define PF_LESS_THROTTLE	0x00100000	/* Throttle me less: I clean memory */
1317 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1318 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1319 #define PF_SWAPWRITE		0x00800000	/* Allowed to write to swap */
1320 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_allowed */
1321 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1322 #define PF_MUTEX_TESTER		0x20000000	/* Thread belongs to the rt mutex tester */
1323 #define PF_FREEZER_SKIP		0x40000000	/* Freezer should not count it as freezable */
1324 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1325 
1326 /*
1327  * Only the _current_ task can read/write to tsk->flags, but other
1328  * tasks can access tsk->flags in readonly mode for example
1329  * with tsk_used_math (like during threaded core dumping).
1330  * There is however an exception to this rule during ptrace
1331  * or during fork: the ptracer task is allowed to write to the
1332  * child->flags of its traced child (same goes for fork, the parent
1333  * can write to the child->flags), because we're guaranteed the
1334  * child is not running and in turn not changing child->flags
1335  * at the same time the parent does it.
1336  */
1337 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1338 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1339 #define clear_used_math()			clear_stopped_child_used_math(current)
1340 #define set_used_math()				set_stopped_child_used_math(current)
1341 
1342 #define conditional_stopped_child_used_math(condition, child) \
1343 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1344 
1345 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1346 
1347 #define copy_to_stopped_child_used_math(child) \
1348 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1349 
1350 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1351 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1352 #define used_math()				tsk_used_math(current)
1353 
1354 static inline bool is_percpu_thread(void)
1355 {
1356 #ifdef CONFIG_SMP
1357 	return (current->flags & PF_NO_SETAFFINITY) &&
1358 		(current->nr_cpus_allowed  == 1);
1359 #else
1360 	return true;
1361 #endif
1362 }
1363 
1364 /* Per-process atomic flags. */
1365 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1366 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1367 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1368 
1369 
1370 #define TASK_PFA_TEST(name, func)					\
1371 	static inline bool task_##func(struct task_struct *p)		\
1372 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1373 
1374 #define TASK_PFA_SET(name, func)					\
1375 	static inline void task_set_##func(struct task_struct *p)	\
1376 	{ set_bit(PFA_##name, &p->atomic_flags); }
1377 
1378 #define TASK_PFA_CLEAR(name, func)					\
1379 	static inline void task_clear_##func(struct task_struct *p)	\
1380 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1381 
1382 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1383 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1384 
1385 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1386 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1387 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1388 
1389 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1390 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1391 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1392 
1393 static inline void
1394 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1395 {
1396 	current->flags &= ~flags;
1397 	current->flags |= orig_flags & flags;
1398 }
1399 
1400 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1401 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1402 #ifdef CONFIG_SMP
1403 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1404 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1405 #else
1406 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1407 {
1408 }
1409 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1410 {
1411 	if (!cpumask_test_cpu(0, new_mask))
1412 		return -EINVAL;
1413 	return 0;
1414 }
1415 #endif
1416 
1417 #ifndef cpu_relax_yield
1418 #define cpu_relax_yield() cpu_relax()
1419 #endif
1420 
1421 extern int yield_to(struct task_struct *p, bool preempt);
1422 extern void set_user_nice(struct task_struct *p, long nice);
1423 extern int task_prio(const struct task_struct *p);
1424 
1425 /**
1426  * task_nice - return the nice value of a given task.
1427  * @p: the task in question.
1428  *
1429  * Return: The nice value [ -20 ... 0 ... 19 ].
1430  */
1431 static inline int task_nice(const struct task_struct *p)
1432 {
1433 	return PRIO_TO_NICE((p)->static_prio);
1434 }
1435 
1436 extern int can_nice(const struct task_struct *p, const int nice);
1437 extern int task_curr(const struct task_struct *p);
1438 extern int idle_cpu(int cpu);
1439 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1440 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1441 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1442 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1443 extern struct task_struct *idle_task(int cpu);
1444 
1445 /**
1446  * is_idle_task - is the specified task an idle task?
1447  * @p: the task in question.
1448  *
1449  * Return: 1 if @p is an idle task. 0 otherwise.
1450  */
1451 static inline bool is_idle_task(const struct task_struct *p)
1452 {
1453 	return !!(p->flags & PF_IDLE);
1454 }
1455 
1456 extern struct task_struct *curr_task(int cpu);
1457 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1458 
1459 void yield(void);
1460 
1461 union thread_union {
1462 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1463 	struct task_struct task;
1464 #endif
1465 #ifndef CONFIG_THREAD_INFO_IN_TASK
1466 	struct thread_info thread_info;
1467 #endif
1468 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1469 };
1470 
1471 #ifndef CONFIG_THREAD_INFO_IN_TASK
1472 extern struct thread_info init_thread_info;
1473 #endif
1474 
1475 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1476 
1477 #ifdef CONFIG_THREAD_INFO_IN_TASK
1478 static inline struct thread_info *task_thread_info(struct task_struct *task)
1479 {
1480 	return &task->thread_info;
1481 }
1482 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1483 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1484 #endif
1485 
1486 /*
1487  * find a task by one of its numerical ids
1488  *
1489  * find_task_by_pid_ns():
1490  *      finds a task by its pid in the specified namespace
1491  * find_task_by_vpid():
1492  *      finds a task by its virtual pid
1493  *
1494  * see also find_vpid() etc in include/linux/pid.h
1495  */
1496 
1497 extern struct task_struct *find_task_by_vpid(pid_t nr);
1498 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1499 
1500 /*
1501  * find a task by its virtual pid and get the task struct
1502  */
1503 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1504 
1505 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1506 extern int wake_up_process(struct task_struct *tsk);
1507 extern void wake_up_new_task(struct task_struct *tsk);
1508 
1509 #ifdef CONFIG_SMP
1510 extern void kick_process(struct task_struct *tsk);
1511 #else
1512 static inline void kick_process(struct task_struct *tsk) { }
1513 #endif
1514 
1515 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1516 
1517 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1518 {
1519 	__set_task_comm(tsk, from, false);
1520 }
1521 
1522 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1523 #define get_task_comm(buf, tsk) ({			\
1524 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1525 	__get_task_comm(buf, sizeof(buf), tsk);		\
1526 })
1527 
1528 #ifdef CONFIG_SMP
1529 void scheduler_ipi(void);
1530 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1531 #else
1532 static inline void scheduler_ipi(void) { }
1533 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1534 {
1535 	return 1;
1536 }
1537 #endif
1538 
1539 /*
1540  * Set thread flags in other task's structures.
1541  * See asm/thread_info.h for TIF_xxxx flags available:
1542  */
1543 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1544 {
1545 	set_ti_thread_flag(task_thread_info(tsk), flag);
1546 }
1547 
1548 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1549 {
1550 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1551 }
1552 
1553 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1554 {
1555 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1556 }
1557 
1558 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1559 {
1560 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1561 }
1562 
1563 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1564 {
1565 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1566 }
1567 
1568 static inline void set_tsk_need_resched(struct task_struct *tsk)
1569 {
1570 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1571 }
1572 
1573 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1574 {
1575 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1576 }
1577 
1578 static inline int test_tsk_need_resched(struct task_struct *tsk)
1579 {
1580 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1581 }
1582 
1583 /*
1584  * cond_resched() and cond_resched_lock(): latency reduction via
1585  * explicit rescheduling in places that are safe. The return
1586  * value indicates whether a reschedule was done in fact.
1587  * cond_resched_lock() will drop the spinlock before scheduling,
1588  * cond_resched_softirq() will enable bhs before scheduling.
1589  */
1590 #ifndef CONFIG_PREEMPT
1591 extern int _cond_resched(void);
1592 #else
1593 static inline int _cond_resched(void) { return 0; }
1594 #endif
1595 
1596 #define cond_resched() ({			\
1597 	___might_sleep(__FILE__, __LINE__, 0);	\
1598 	_cond_resched();			\
1599 })
1600 
1601 extern int __cond_resched_lock(spinlock_t *lock);
1602 
1603 #define cond_resched_lock(lock) ({				\
1604 	___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1605 	__cond_resched_lock(lock);				\
1606 })
1607 
1608 extern int __cond_resched_softirq(void);
1609 
1610 #define cond_resched_softirq() ({					\
1611 	___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET);	\
1612 	__cond_resched_softirq();					\
1613 })
1614 
1615 static inline void cond_resched_rcu(void)
1616 {
1617 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1618 	rcu_read_unlock();
1619 	cond_resched();
1620 	rcu_read_lock();
1621 #endif
1622 }
1623 
1624 /*
1625  * Does a critical section need to be broken due to another
1626  * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1627  * but a general need for low latency)
1628  */
1629 static inline int spin_needbreak(spinlock_t *lock)
1630 {
1631 #ifdef CONFIG_PREEMPT
1632 	return spin_is_contended(lock);
1633 #else
1634 	return 0;
1635 #endif
1636 }
1637 
1638 static __always_inline bool need_resched(void)
1639 {
1640 	return unlikely(tif_need_resched());
1641 }
1642 
1643 /*
1644  * Wrappers for p->thread_info->cpu access. No-op on UP.
1645  */
1646 #ifdef CONFIG_SMP
1647 
1648 static inline unsigned int task_cpu(const struct task_struct *p)
1649 {
1650 #ifdef CONFIG_THREAD_INFO_IN_TASK
1651 	return p->cpu;
1652 #else
1653 	return task_thread_info(p)->cpu;
1654 #endif
1655 }
1656 
1657 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1658 
1659 #else
1660 
1661 static inline unsigned int task_cpu(const struct task_struct *p)
1662 {
1663 	return 0;
1664 }
1665 
1666 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1667 {
1668 }
1669 
1670 #endif /* CONFIG_SMP */
1671 
1672 /*
1673  * In order to reduce various lock holder preemption latencies provide an
1674  * interface to see if a vCPU is currently running or not.
1675  *
1676  * This allows us to terminate optimistic spin loops and block, analogous to
1677  * the native optimistic spin heuristic of testing if the lock owner task is
1678  * running or not.
1679  */
1680 #ifndef vcpu_is_preempted
1681 # define vcpu_is_preempted(cpu)	false
1682 #endif
1683 
1684 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1685 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1686 
1687 #ifndef TASK_SIZE_OF
1688 #define TASK_SIZE_OF(tsk)	TASK_SIZE
1689 #endif
1690 
1691 #endif
1692