xref: /linux/include/linux/sched.h (revision cd9626e9)
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 #include <asm/processor.h>
14 #include <linux/thread_info.h>
15 #include <linux/preempt.h>
16 #include <linux/cpumask_types.h>
17 
18 #include <linux/cache.h>
19 #include <linux/irqflags_types.h>
20 #include <linux/smp_types.h>
21 #include <linux/pid_types.h>
22 #include <linux/sem_types.h>
23 #include <linux/shm.h>
24 #include <linux/kmsan_types.h>
25 #include <linux/mutex_types.h>
26 #include <linux/plist_types.h>
27 #include <linux/hrtimer_types.h>
28 #include <linux/timer_types.h>
29 #include <linux/seccomp_types.h>
30 #include <linux/nodemask_types.h>
31 #include <linux/refcount_types.h>
32 #include <linux/resource.h>
33 #include <linux/latencytop.h>
34 #include <linux/sched/prio.h>
35 #include <linux/sched/types.h>
36 #include <linux/signal_types.h>
37 #include <linux/syscall_user_dispatch_types.h>
38 #include <linux/mm_types_task.h>
39 #include <linux/netdevice_xmit.h>
40 #include <linux/task_io_accounting.h>
41 #include <linux/posix-timers_types.h>
42 #include <linux/restart_block.h>
43 #include <uapi/linux/rseq.h>
44 #include <linux/seqlock_types.h>
45 #include <linux/kcsan.h>
46 #include <linux/rv.h>
47 #include <linux/livepatch_sched.h>
48 #include <linux/uidgid_types.h>
49 #include <asm/kmap_size.h>
50 
51 /* task_struct member predeclarations (sorted alphabetically): */
52 struct audit_context;
53 struct bio_list;
54 struct blk_plug;
55 struct bpf_local_storage;
56 struct bpf_run_ctx;
57 struct bpf_net_context;
58 struct capture_control;
59 struct cfs_rq;
60 struct fs_struct;
61 struct futex_pi_state;
62 struct io_context;
63 struct io_uring_task;
64 struct mempolicy;
65 struct nameidata;
66 struct nsproxy;
67 struct perf_event_context;
68 struct pid_namespace;
69 struct pipe_inode_info;
70 struct rcu_node;
71 struct reclaim_state;
72 struct robust_list_head;
73 struct root_domain;
74 struct rq;
75 struct sched_attr;
76 struct sched_dl_entity;
77 struct seq_file;
78 struct sighand_struct;
79 struct signal_struct;
80 struct task_delay_info;
81 struct task_group;
82 struct task_struct;
83 struct user_event_mm;
84 
85 #include <linux/sched/ext.h>
86 
87 /*
88  * Task state bitmask. NOTE! These bits are also
89  * encoded in fs/proc/array.c: get_task_state().
90  *
91  * We have two separate sets of flags: task->__state
92  * is about runnability, while task->exit_state are
93  * about the task exiting. Confusing, but this way
94  * modifying one set can't modify the other one by
95  * mistake.
96  */
97 
98 /* Used in tsk->__state: */
99 #define TASK_RUNNING			0x00000000
100 #define TASK_INTERRUPTIBLE		0x00000001
101 #define TASK_UNINTERRUPTIBLE		0x00000002
102 #define __TASK_STOPPED			0x00000004
103 #define __TASK_TRACED			0x00000008
104 /* Used in tsk->exit_state: */
105 #define EXIT_DEAD			0x00000010
106 #define EXIT_ZOMBIE			0x00000020
107 #define EXIT_TRACE			(EXIT_ZOMBIE | EXIT_DEAD)
108 /* Used in tsk->__state again: */
109 #define TASK_PARKED			0x00000040
110 #define TASK_DEAD			0x00000080
111 #define TASK_WAKEKILL			0x00000100
112 #define TASK_WAKING			0x00000200
113 #define TASK_NOLOAD			0x00000400
114 #define TASK_NEW			0x00000800
115 #define TASK_RTLOCK_WAIT		0x00001000
116 #define TASK_FREEZABLE			0x00002000
117 #define __TASK_FREEZABLE_UNSAFE	       (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
118 #define TASK_FROZEN			0x00008000
119 #define TASK_STATE_MAX			0x00010000
120 
121 #define TASK_ANY			(TASK_STATE_MAX-1)
122 
123 /*
124  * DO NOT ADD ANY NEW USERS !
125  */
126 #define TASK_FREEZABLE_UNSAFE		(TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
127 
128 /* Convenience macros for the sake of set_current_state: */
129 #define TASK_KILLABLE			(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
130 #define TASK_STOPPED			(TASK_WAKEKILL | __TASK_STOPPED)
131 #define TASK_TRACED			__TASK_TRACED
132 
133 #define TASK_IDLE			(TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
134 
135 /* Convenience macros for the sake of wake_up(): */
136 #define TASK_NORMAL			(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
137 
138 /* get_task_state(): */
139 #define TASK_REPORT			(TASK_RUNNING | TASK_INTERRUPTIBLE | \
140 					 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
141 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
142 					 TASK_PARKED)
143 
144 #define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
145 
146 #define task_is_traced(task)		((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
147 #define task_is_stopped(task)		((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
148 #define task_is_stopped_or_traced(task)	((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
149 
150 /*
151  * Special states are those that do not use the normal wait-loop pattern. See
152  * the comment with set_special_state().
153  */
154 #define is_special_task_state(state)					\
155 	((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED |	\
156 		    TASK_DEAD | TASK_FROZEN))
157 
158 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
159 # define debug_normal_state_change(state_value)				\
160 	do {								\
161 		WARN_ON_ONCE(is_special_task_state(state_value));	\
162 		current->task_state_change = _THIS_IP_;			\
163 	} while (0)
164 
165 # define debug_special_state_change(state_value)			\
166 	do {								\
167 		WARN_ON_ONCE(!is_special_task_state(state_value));	\
168 		current->task_state_change = _THIS_IP_;			\
169 	} while (0)
170 
171 # define debug_rtlock_wait_set_state()					\
172 	do {								 \
173 		current->saved_state_change = current->task_state_change;\
174 		current->task_state_change = _THIS_IP_;			 \
175 	} while (0)
176 
177 # define debug_rtlock_wait_restore_state()				\
178 	do {								 \
179 		current->task_state_change = current->saved_state_change;\
180 	} while (0)
181 
182 #else
183 # define debug_normal_state_change(cond)	do { } while (0)
184 # define debug_special_state_change(cond)	do { } while (0)
185 # define debug_rtlock_wait_set_state()		do { } while (0)
186 # define debug_rtlock_wait_restore_state()	do { } while (0)
187 #endif
188 
189 /*
190  * set_current_state() includes a barrier so that the write of current->__state
191  * is correctly serialised wrt the caller's subsequent test of whether to
192  * actually sleep:
193  *
194  *   for (;;) {
195  *	set_current_state(TASK_UNINTERRUPTIBLE);
196  *	if (CONDITION)
197  *	   break;
198  *
199  *	schedule();
200  *   }
201  *   __set_current_state(TASK_RUNNING);
202  *
203  * If the caller does not need such serialisation (because, for instance, the
204  * CONDITION test and condition change and wakeup are under the same lock) then
205  * use __set_current_state().
206  *
207  * The above is typically ordered against the wakeup, which does:
208  *
209  *   CONDITION = 1;
210  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
211  *
212  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
213  * accessing p->__state.
214  *
215  * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
216  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
217  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
218  *
219  * However, with slightly different timing the wakeup TASK_RUNNING store can
220  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
221  * a problem either because that will result in one extra go around the loop
222  * and our @cond test will save the day.
223  *
224  * Also see the comments of try_to_wake_up().
225  */
226 #define __set_current_state(state_value)				\
227 	do {								\
228 		debug_normal_state_change((state_value));		\
229 		WRITE_ONCE(current->__state, (state_value));		\
230 	} while (0)
231 
232 #define set_current_state(state_value)					\
233 	do {								\
234 		debug_normal_state_change((state_value));		\
235 		smp_store_mb(current->__state, (state_value));		\
236 	} while (0)
237 
238 /*
239  * set_special_state() should be used for those states when the blocking task
240  * can not use the regular condition based wait-loop. In that case we must
241  * serialize against wakeups such that any possible in-flight TASK_RUNNING
242  * stores will not collide with our state change.
243  */
244 #define set_special_state(state_value)					\
245 	do {								\
246 		unsigned long flags; /* may shadow */			\
247 									\
248 		raw_spin_lock_irqsave(&current->pi_lock, flags);	\
249 		debug_special_state_change((state_value));		\
250 		WRITE_ONCE(current->__state, (state_value));		\
251 		raw_spin_unlock_irqrestore(&current->pi_lock, flags);	\
252 	} while (0)
253 
254 /*
255  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
256  *
257  * RT's spin/rwlock substitutions are state preserving. The state of the
258  * task when blocking on the lock is saved in task_struct::saved_state and
259  * restored after the lock has been acquired.  These operations are
260  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
261  * lock related wakeups while the task is blocked on the lock are
262  * redirected to operate on task_struct::saved_state to ensure that these
263  * are not dropped. On restore task_struct::saved_state is set to
264  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
265  *
266  * The lock operation looks like this:
267  *
268  *	current_save_and_set_rtlock_wait_state();
269  *	for (;;) {
270  *		if (try_lock())
271  *			break;
272  *		raw_spin_unlock_irq(&lock->wait_lock);
273  *		schedule_rtlock();
274  *		raw_spin_lock_irq(&lock->wait_lock);
275  *		set_current_state(TASK_RTLOCK_WAIT);
276  *	}
277  *	current_restore_rtlock_saved_state();
278  */
279 #define current_save_and_set_rtlock_wait_state()			\
280 	do {								\
281 		lockdep_assert_irqs_disabled();				\
282 		raw_spin_lock(&current->pi_lock);			\
283 		current->saved_state = current->__state;		\
284 		debug_rtlock_wait_set_state();				\
285 		WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);		\
286 		raw_spin_unlock(&current->pi_lock);			\
287 	} while (0);
288 
289 #define current_restore_rtlock_saved_state()				\
290 	do {								\
291 		lockdep_assert_irqs_disabled();				\
292 		raw_spin_lock(&current->pi_lock);			\
293 		debug_rtlock_wait_restore_state();			\
294 		WRITE_ONCE(current->__state, current->saved_state);	\
295 		current->saved_state = TASK_RUNNING;			\
296 		raw_spin_unlock(&current->pi_lock);			\
297 	} while (0);
298 
299 #define get_current_state()	READ_ONCE(current->__state)
300 
301 /*
302  * Define the task command name length as enum, then it can be visible to
303  * BPF programs.
304  */
305 enum {
306 	TASK_COMM_LEN = 16,
307 };
308 
309 extern void sched_tick(void);
310 
311 #define	MAX_SCHEDULE_TIMEOUT		LONG_MAX
312 
313 extern long schedule_timeout(long timeout);
314 extern long schedule_timeout_interruptible(long timeout);
315 extern long schedule_timeout_killable(long timeout);
316 extern long schedule_timeout_uninterruptible(long timeout);
317 extern long schedule_timeout_idle(long timeout);
318 asmlinkage void schedule(void);
319 extern void schedule_preempt_disabled(void);
320 asmlinkage void preempt_schedule_irq(void);
321 #ifdef CONFIG_PREEMPT_RT
322  extern void schedule_rtlock(void);
323 #endif
324 
325 extern int __must_check io_schedule_prepare(void);
326 extern void io_schedule_finish(int token);
327 extern long io_schedule_timeout(long timeout);
328 extern void io_schedule(void);
329 
330 /**
331  * struct prev_cputime - snapshot of system and user cputime
332  * @utime: time spent in user mode
333  * @stime: time spent in system mode
334  * @lock: protects the above two fields
335  *
336  * Stores previous user/system time values such that we can guarantee
337  * monotonicity.
338  */
339 struct prev_cputime {
340 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
341 	u64				utime;
342 	u64				stime;
343 	raw_spinlock_t			lock;
344 #endif
345 };
346 
347 enum vtime_state {
348 	/* Task is sleeping or running in a CPU with VTIME inactive: */
349 	VTIME_INACTIVE = 0,
350 	/* Task is idle */
351 	VTIME_IDLE,
352 	/* Task runs in kernelspace in a CPU with VTIME active: */
353 	VTIME_SYS,
354 	/* Task runs in userspace in a CPU with VTIME active: */
355 	VTIME_USER,
356 	/* Task runs as guests in a CPU with VTIME active: */
357 	VTIME_GUEST,
358 };
359 
360 struct vtime {
361 	seqcount_t		seqcount;
362 	unsigned long long	starttime;
363 	enum vtime_state	state;
364 	unsigned int		cpu;
365 	u64			utime;
366 	u64			stime;
367 	u64			gtime;
368 };
369 
370 /*
371  * Utilization clamp constraints.
372  * @UCLAMP_MIN:	Minimum utilization
373  * @UCLAMP_MAX:	Maximum utilization
374  * @UCLAMP_CNT:	Utilization clamp constraints count
375  */
376 enum uclamp_id {
377 	UCLAMP_MIN = 0,
378 	UCLAMP_MAX,
379 	UCLAMP_CNT
380 };
381 
382 #ifdef CONFIG_SMP
383 extern struct root_domain def_root_domain;
384 extern struct mutex sched_domains_mutex;
385 #endif
386 
387 struct sched_param {
388 	int sched_priority;
389 };
390 
391 struct sched_info {
392 #ifdef CONFIG_SCHED_INFO
393 	/* Cumulative counters: */
394 
395 	/* # of times we have run on this CPU: */
396 	unsigned long			pcount;
397 
398 	/* Time spent waiting on a runqueue: */
399 	unsigned long long		run_delay;
400 
401 	/* Timestamps: */
402 
403 	/* When did we last run on a CPU? */
404 	unsigned long long		last_arrival;
405 
406 	/* When were we last queued to run? */
407 	unsigned long long		last_queued;
408 
409 #endif /* CONFIG_SCHED_INFO */
410 };
411 
412 /*
413  * Integer metrics need fixed point arithmetic, e.g., sched/fair
414  * has a few: load, load_avg, util_avg, freq, and capacity.
415  *
416  * We define a basic fixed point arithmetic range, and then formalize
417  * all these metrics based on that basic range.
418  */
419 # define SCHED_FIXEDPOINT_SHIFT		10
420 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
421 
422 /* Increase resolution of cpu_capacity calculations */
423 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
424 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
425 
426 struct load_weight {
427 	unsigned long			weight;
428 	u32				inv_weight;
429 };
430 
431 /*
432  * The load/runnable/util_avg accumulates an infinite geometric series
433  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
434  *
435  * [load_avg definition]
436  *
437  *   load_avg = runnable% * scale_load_down(load)
438  *
439  * [runnable_avg definition]
440  *
441  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
442  *
443  * [util_avg definition]
444  *
445  *   util_avg = running% * SCHED_CAPACITY_SCALE
446  *
447  * where runnable% is the time ratio that a sched_entity is runnable and
448  * running% the time ratio that a sched_entity is running.
449  *
450  * For cfs_rq, they are the aggregated values of all runnable and blocked
451  * sched_entities.
452  *
453  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
454  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
455  * for computing those signals (see update_rq_clock_pelt())
456  *
457  * N.B., the above ratios (runnable% and running%) themselves are in the
458  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
459  * to as large a range as necessary. This is for example reflected by
460  * util_avg's SCHED_CAPACITY_SCALE.
461  *
462  * [Overflow issue]
463  *
464  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
465  * with the highest load (=88761), always runnable on a single cfs_rq,
466  * and should not overflow as the number already hits PID_MAX_LIMIT.
467  *
468  * For all other cases (including 32-bit kernels), struct load_weight's
469  * weight will overflow first before we do, because:
470  *
471  *    Max(load_avg) <= Max(load.weight)
472  *
473  * Then it is the load_weight's responsibility to consider overflow
474  * issues.
475  */
476 struct sched_avg {
477 	u64				last_update_time;
478 	u64				load_sum;
479 	u64				runnable_sum;
480 	u32				util_sum;
481 	u32				period_contrib;
482 	unsigned long			load_avg;
483 	unsigned long			runnable_avg;
484 	unsigned long			util_avg;
485 	unsigned int			util_est;
486 } ____cacheline_aligned;
487 
488 /*
489  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
490  * updates. When a task is dequeued, its util_est should not be updated if its
491  * util_avg has not been updated in the meantime.
492  * This information is mapped into the MSB bit of util_est at dequeue time.
493  * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
494  * it is safe to use MSB.
495  */
496 #define UTIL_EST_WEIGHT_SHIFT		2
497 #define UTIL_AVG_UNCHANGED		0x80000000
498 
499 struct sched_statistics {
500 #ifdef CONFIG_SCHEDSTATS
501 	u64				wait_start;
502 	u64				wait_max;
503 	u64				wait_count;
504 	u64				wait_sum;
505 	u64				iowait_count;
506 	u64				iowait_sum;
507 
508 	u64				sleep_start;
509 	u64				sleep_max;
510 	s64				sum_sleep_runtime;
511 
512 	u64				block_start;
513 	u64				block_max;
514 	s64				sum_block_runtime;
515 
516 	s64				exec_max;
517 	u64				slice_max;
518 
519 	u64				nr_migrations_cold;
520 	u64				nr_failed_migrations_affine;
521 	u64				nr_failed_migrations_running;
522 	u64				nr_failed_migrations_hot;
523 	u64				nr_forced_migrations;
524 
525 	u64				nr_wakeups;
526 	u64				nr_wakeups_sync;
527 	u64				nr_wakeups_migrate;
528 	u64				nr_wakeups_local;
529 	u64				nr_wakeups_remote;
530 	u64				nr_wakeups_affine;
531 	u64				nr_wakeups_affine_attempts;
532 	u64				nr_wakeups_passive;
533 	u64				nr_wakeups_idle;
534 
535 #ifdef CONFIG_SCHED_CORE
536 	u64				core_forceidle_sum;
537 #endif
538 #endif /* CONFIG_SCHEDSTATS */
539 } ____cacheline_aligned;
540 
541 struct sched_entity {
542 	/* For load-balancing: */
543 	struct load_weight		load;
544 	struct rb_node			run_node;
545 	u64				deadline;
546 	u64				min_vruntime;
547 	u64				min_slice;
548 
549 	struct list_head		group_node;
550 	unsigned char			on_rq;
551 	unsigned char			sched_delayed;
552 	unsigned char			rel_deadline;
553 	unsigned char			custom_slice;
554 					/* hole */
555 
556 	u64				exec_start;
557 	u64				sum_exec_runtime;
558 	u64				prev_sum_exec_runtime;
559 	u64				vruntime;
560 	s64				vlag;
561 	u64				slice;
562 
563 	u64				nr_migrations;
564 
565 #ifdef CONFIG_FAIR_GROUP_SCHED
566 	int				depth;
567 	struct sched_entity		*parent;
568 	/* rq on which this entity is (to be) queued: */
569 	struct cfs_rq			*cfs_rq;
570 	/* rq "owned" by this entity/group: */
571 	struct cfs_rq			*my_q;
572 	/* cached value of my_q->h_nr_running */
573 	unsigned long			runnable_weight;
574 #endif
575 
576 #ifdef CONFIG_SMP
577 	/*
578 	 * Per entity load average tracking.
579 	 *
580 	 * Put into separate cache line so it does not
581 	 * collide with read-mostly values above.
582 	 */
583 	struct sched_avg		avg;
584 #endif
585 };
586 
587 struct sched_rt_entity {
588 	struct list_head		run_list;
589 	unsigned long			timeout;
590 	unsigned long			watchdog_stamp;
591 	unsigned int			time_slice;
592 	unsigned short			on_rq;
593 	unsigned short			on_list;
594 
595 	struct sched_rt_entity		*back;
596 #ifdef CONFIG_RT_GROUP_SCHED
597 	struct sched_rt_entity		*parent;
598 	/* rq on which this entity is (to be) queued: */
599 	struct rt_rq			*rt_rq;
600 	/* rq "owned" by this entity/group: */
601 	struct rt_rq			*my_q;
602 #endif
603 } __randomize_layout;
604 
605 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
606 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
607 
608 struct sched_dl_entity {
609 	struct rb_node			rb_node;
610 
611 	/*
612 	 * Original scheduling parameters. Copied here from sched_attr
613 	 * during sched_setattr(), they will remain the same until
614 	 * the next sched_setattr().
615 	 */
616 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
617 	u64				dl_deadline;	/* Relative deadline of each instance	*/
618 	u64				dl_period;	/* Separation of two instances (period) */
619 	u64				dl_bw;		/* dl_runtime / dl_period		*/
620 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
621 
622 	/*
623 	 * Actual scheduling parameters. Initialized with the values above,
624 	 * they are continuously updated during task execution. Note that
625 	 * the remaining runtime could be < 0 in case we are in overrun.
626 	 */
627 	s64				runtime;	/* Remaining runtime for this instance	*/
628 	u64				deadline;	/* Absolute deadline for this instance	*/
629 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
630 
631 	/*
632 	 * Some bool flags:
633 	 *
634 	 * @dl_throttled tells if we exhausted the runtime. If so, the
635 	 * task has to wait for a replenishment to be performed at the
636 	 * next firing of dl_timer.
637 	 *
638 	 * @dl_yielded tells if task gave up the CPU before consuming
639 	 * all its available runtime during the last job.
640 	 *
641 	 * @dl_non_contending tells if the task is inactive while still
642 	 * contributing to the active utilization. In other words, it
643 	 * indicates if the inactive timer has been armed and its handler
644 	 * has not been executed yet. This flag is useful to avoid race
645 	 * conditions between the inactive timer handler and the wakeup
646 	 * code.
647 	 *
648 	 * @dl_overrun tells if the task asked to be informed about runtime
649 	 * overruns.
650 	 *
651 	 * @dl_server tells if this is a server entity.
652 	 *
653 	 * @dl_defer tells if this is a deferred or regular server. For
654 	 * now only defer server exists.
655 	 *
656 	 * @dl_defer_armed tells if the deferrable server is waiting
657 	 * for the replenishment timer to activate it.
658 	 *
659 	 * @dl_defer_running tells if the deferrable server is actually
660 	 * running, skipping the defer phase.
661 	 */
662 	unsigned int			dl_throttled      : 1;
663 	unsigned int			dl_yielded        : 1;
664 	unsigned int			dl_non_contending : 1;
665 	unsigned int			dl_overrun	  : 1;
666 	unsigned int			dl_server         : 1;
667 	unsigned int			dl_defer	  : 1;
668 	unsigned int			dl_defer_armed	  : 1;
669 	unsigned int			dl_defer_running  : 1;
670 
671 	/*
672 	 * Bandwidth enforcement timer. Each -deadline task has its
673 	 * own bandwidth to be enforced, thus we need one timer per task.
674 	 */
675 	struct hrtimer			dl_timer;
676 
677 	/*
678 	 * Inactive timer, responsible for decreasing the active utilization
679 	 * at the "0-lag time". When a -deadline task blocks, it contributes
680 	 * to GRUB's active utilization until the "0-lag time", hence a
681 	 * timer is needed to decrease the active utilization at the correct
682 	 * time.
683 	 */
684 	struct hrtimer			inactive_timer;
685 
686 	/*
687 	 * Bits for DL-server functionality. Also see the comment near
688 	 * dl_server_update().
689 	 *
690 	 * @rq the runqueue this server is for
691 	 *
692 	 * @server_has_tasks() returns true if @server_pick return a
693 	 * runnable task.
694 	 */
695 	struct rq			*rq;
696 	dl_server_has_tasks_f		server_has_tasks;
697 	dl_server_pick_f		server_pick_task;
698 
699 #ifdef CONFIG_RT_MUTEXES
700 	/*
701 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
702 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
703 	 * to (the original one/itself).
704 	 */
705 	struct sched_dl_entity *pi_se;
706 #endif
707 };
708 
709 #ifdef CONFIG_UCLAMP_TASK
710 /* Number of utilization clamp buckets (shorter alias) */
711 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
712 
713 /*
714  * Utilization clamp for a scheduling entity
715  * @value:		clamp value "assigned" to a se
716  * @bucket_id:		bucket index corresponding to the "assigned" value
717  * @active:		the se is currently refcounted in a rq's bucket
718  * @user_defined:	the requested clamp value comes from user-space
719  *
720  * The bucket_id is the index of the clamp bucket matching the clamp value
721  * which is pre-computed and stored to avoid expensive integer divisions from
722  * the fast path.
723  *
724  * The active bit is set whenever a task has got an "effective" value assigned,
725  * which can be different from the clamp value "requested" from user-space.
726  * This allows to know a task is refcounted in the rq's bucket corresponding
727  * to the "effective" bucket_id.
728  *
729  * The user_defined bit is set whenever a task has got a task-specific clamp
730  * value requested from userspace, i.e. the system defaults apply to this task
731  * just as a restriction. This allows to relax default clamps when a less
732  * restrictive task-specific value has been requested, thus allowing to
733  * implement a "nice" semantic. For example, a task running with a 20%
734  * default boost can still drop its own boosting to 0%.
735  */
736 struct uclamp_se {
737 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
738 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
739 	unsigned int active		: 1;
740 	unsigned int user_defined	: 1;
741 };
742 #endif /* CONFIG_UCLAMP_TASK */
743 
744 union rcu_special {
745 	struct {
746 		u8			blocked;
747 		u8			need_qs;
748 		u8			exp_hint; /* Hint for performance. */
749 		u8			need_mb; /* Readers need smp_mb(). */
750 	} b; /* Bits. */
751 	u32 s; /* Set of bits. */
752 };
753 
754 enum perf_event_task_context {
755 	perf_invalid_context = -1,
756 	perf_hw_context = 0,
757 	perf_sw_context,
758 	perf_nr_task_contexts,
759 };
760 
761 /*
762  * Number of contexts where an event can trigger:
763  *      task, softirq, hardirq, nmi.
764  */
765 #define PERF_NR_CONTEXTS	4
766 
767 struct wake_q_node {
768 	struct wake_q_node *next;
769 };
770 
771 struct kmap_ctrl {
772 #ifdef CONFIG_KMAP_LOCAL
773 	int				idx;
774 	pte_t				pteval[KM_MAX_IDX];
775 #endif
776 };
777 
778 struct task_struct {
779 #ifdef CONFIG_THREAD_INFO_IN_TASK
780 	/*
781 	 * For reasons of header soup (see current_thread_info()), this
782 	 * must be the first element of task_struct.
783 	 */
784 	struct thread_info		thread_info;
785 #endif
786 	unsigned int			__state;
787 
788 	/* saved state for "spinlock sleepers" */
789 	unsigned int			saved_state;
790 
791 	/*
792 	 * This begins the randomizable portion of task_struct. Only
793 	 * scheduling-critical items should be added above here.
794 	 */
795 	randomized_struct_fields_start
796 
797 	void				*stack;
798 	refcount_t			usage;
799 	/* Per task flags (PF_*), defined further below: */
800 	unsigned int			flags;
801 	unsigned int			ptrace;
802 
803 #ifdef CONFIG_MEM_ALLOC_PROFILING
804 	struct alloc_tag		*alloc_tag;
805 #endif
806 
807 #ifdef CONFIG_SMP
808 	int				on_cpu;
809 	struct __call_single_node	wake_entry;
810 	unsigned int			wakee_flips;
811 	unsigned long			wakee_flip_decay_ts;
812 	struct task_struct		*last_wakee;
813 
814 	/*
815 	 * recent_used_cpu is initially set as the last CPU used by a task
816 	 * that wakes affine another task. Waker/wakee relationships can
817 	 * push tasks around a CPU where each wakeup moves to the next one.
818 	 * Tracking a recently used CPU allows a quick search for a recently
819 	 * used CPU that may be idle.
820 	 */
821 	int				recent_used_cpu;
822 	int				wake_cpu;
823 #endif
824 	int				on_rq;
825 
826 	int				prio;
827 	int				static_prio;
828 	int				normal_prio;
829 	unsigned int			rt_priority;
830 
831 	struct sched_entity		se;
832 	struct sched_rt_entity		rt;
833 	struct sched_dl_entity		dl;
834 	struct sched_dl_entity		*dl_server;
835 #ifdef CONFIG_SCHED_CLASS_EXT
836 	struct sched_ext_entity		scx;
837 #endif
838 	const struct sched_class	*sched_class;
839 
840 #ifdef CONFIG_SCHED_CORE
841 	struct rb_node			core_node;
842 	unsigned long			core_cookie;
843 	unsigned int			core_occupation;
844 #endif
845 
846 #ifdef CONFIG_CGROUP_SCHED
847 	struct task_group		*sched_task_group;
848 #endif
849 
850 
851 #ifdef CONFIG_UCLAMP_TASK
852 	/*
853 	 * Clamp values requested for a scheduling entity.
854 	 * Must be updated with task_rq_lock() held.
855 	 */
856 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
857 	/*
858 	 * Effective clamp values used for a scheduling entity.
859 	 * Must be updated with task_rq_lock() held.
860 	 */
861 	struct uclamp_se		uclamp[UCLAMP_CNT];
862 #endif
863 
864 	struct sched_statistics         stats;
865 
866 #ifdef CONFIG_PREEMPT_NOTIFIERS
867 	/* List of struct preempt_notifier: */
868 	struct hlist_head		preempt_notifiers;
869 #endif
870 
871 #ifdef CONFIG_BLK_DEV_IO_TRACE
872 	unsigned int			btrace_seq;
873 #endif
874 
875 	unsigned int			policy;
876 	unsigned long			max_allowed_capacity;
877 	int				nr_cpus_allowed;
878 	const cpumask_t			*cpus_ptr;
879 	cpumask_t			*user_cpus_ptr;
880 	cpumask_t			cpus_mask;
881 	void				*migration_pending;
882 #ifdef CONFIG_SMP
883 	unsigned short			migration_disabled;
884 #endif
885 	unsigned short			migration_flags;
886 
887 #ifdef CONFIG_PREEMPT_RCU
888 	int				rcu_read_lock_nesting;
889 	union rcu_special		rcu_read_unlock_special;
890 	struct list_head		rcu_node_entry;
891 	struct rcu_node			*rcu_blocked_node;
892 #endif /* #ifdef CONFIG_PREEMPT_RCU */
893 
894 #ifdef CONFIG_TASKS_RCU
895 	unsigned long			rcu_tasks_nvcsw;
896 	u8				rcu_tasks_holdout;
897 	u8				rcu_tasks_idx;
898 	int				rcu_tasks_idle_cpu;
899 	struct list_head		rcu_tasks_holdout_list;
900 	int				rcu_tasks_exit_cpu;
901 	struct list_head		rcu_tasks_exit_list;
902 #endif /* #ifdef CONFIG_TASKS_RCU */
903 
904 #ifdef CONFIG_TASKS_TRACE_RCU
905 	int				trc_reader_nesting;
906 	int				trc_ipi_to_cpu;
907 	union rcu_special		trc_reader_special;
908 	struct list_head		trc_holdout_list;
909 	struct list_head		trc_blkd_node;
910 	int				trc_blkd_cpu;
911 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
912 
913 	struct sched_info		sched_info;
914 
915 	struct list_head		tasks;
916 #ifdef CONFIG_SMP
917 	struct plist_node		pushable_tasks;
918 	struct rb_node			pushable_dl_tasks;
919 #endif
920 
921 	struct mm_struct		*mm;
922 	struct mm_struct		*active_mm;
923 	struct address_space		*faults_disabled_mapping;
924 
925 	int				exit_state;
926 	int				exit_code;
927 	int				exit_signal;
928 	/* The signal sent when the parent dies: */
929 	int				pdeath_signal;
930 	/* JOBCTL_*, siglock protected: */
931 	unsigned long			jobctl;
932 
933 	/* Used for emulating ABI behavior of previous Linux versions: */
934 	unsigned int			personality;
935 
936 	/* Scheduler bits, serialized by scheduler locks: */
937 	unsigned			sched_reset_on_fork:1;
938 	unsigned			sched_contributes_to_load:1;
939 	unsigned			sched_migrated:1;
940 
941 	/* Force alignment to the next boundary: */
942 	unsigned			:0;
943 
944 	/* Unserialized, strictly 'current' */
945 
946 	/*
947 	 * This field must not be in the scheduler word above due to wakelist
948 	 * queueing no longer being serialized by p->on_cpu. However:
949 	 *
950 	 * p->XXX = X;			ttwu()
951 	 * schedule()			  if (p->on_rq && ..) // false
952 	 *   smp_mb__after_spinlock();	  if (smp_load_acquire(&p->on_cpu) && //true
953 	 *   deactivate_task()		      ttwu_queue_wakelist())
954 	 *     p->on_rq = 0;			p->sched_remote_wakeup = Y;
955 	 *
956 	 * guarantees all stores of 'current' are visible before
957 	 * ->sched_remote_wakeup gets used, so it can be in this word.
958 	 */
959 	unsigned			sched_remote_wakeup:1;
960 #ifdef CONFIG_RT_MUTEXES
961 	unsigned			sched_rt_mutex:1;
962 #endif
963 
964 	/* Bit to tell TOMOYO we're in execve(): */
965 	unsigned			in_execve:1;
966 	unsigned			in_iowait:1;
967 #ifndef TIF_RESTORE_SIGMASK
968 	unsigned			restore_sigmask:1;
969 #endif
970 #ifdef CONFIG_MEMCG_V1
971 	unsigned			in_user_fault:1;
972 #endif
973 #ifdef CONFIG_LRU_GEN
974 	/* whether the LRU algorithm may apply to this access */
975 	unsigned			in_lru_fault:1;
976 #endif
977 #ifdef CONFIG_COMPAT_BRK
978 	unsigned			brk_randomized:1;
979 #endif
980 #ifdef CONFIG_CGROUPS
981 	/* disallow userland-initiated cgroup migration */
982 	unsigned			no_cgroup_migration:1;
983 	/* task is frozen/stopped (used by the cgroup freezer) */
984 	unsigned			frozen:1;
985 #endif
986 #ifdef CONFIG_BLK_CGROUP
987 	unsigned			use_memdelay:1;
988 #endif
989 #ifdef CONFIG_PSI
990 	/* Stalled due to lack of memory */
991 	unsigned			in_memstall:1;
992 #endif
993 #ifdef CONFIG_PAGE_OWNER
994 	/* Used by page_owner=on to detect recursion in page tracking. */
995 	unsigned			in_page_owner:1;
996 #endif
997 #ifdef CONFIG_EVENTFD
998 	/* Recursion prevention for eventfd_signal() */
999 	unsigned			in_eventfd:1;
1000 #endif
1001 #ifdef CONFIG_ARCH_HAS_CPU_PASID
1002 	unsigned			pasid_activated:1;
1003 #endif
1004 #ifdef	CONFIG_CPU_SUP_INTEL
1005 	unsigned			reported_split_lock:1;
1006 #endif
1007 #ifdef CONFIG_TASK_DELAY_ACCT
1008 	/* delay due to memory thrashing */
1009 	unsigned                        in_thrashing:1;
1010 #endif
1011 #ifdef CONFIG_PREEMPT_RT
1012 	struct netdev_xmit		net_xmit;
1013 #endif
1014 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
1015 
1016 	struct restart_block		restart_block;
1017 
1018 	pid_t				pid;
1019 	pid_t				tgid;
1020 
1021 #ifdef CONFIG_STACKPROTECTOR
1022 	/* Canary value for the -fstack-protector GCC feature: */
1023 	unsigned long			stack_canary;
1024 #endif
1025 	/*
1026 	 * Pointers to the (original) parent process, youngest child, younger sibling,
1027 	 * older sibling, respectively.  (p->father can be replaced with
1028 	 * p->real_parent->pid)
1029 	 */
1030 
1031 	/* Real parent process: */
1032 	struct task_struct __rcu	*real_parent;
1033 
1034 	/* Recipient of SIGCHLD, wait4() reports: */
1035 	struct task_struct __rcu	*parent;
1036 
1037 	/*
1038 	 * Children/sibling form the list of natural children:
1039 	 */
1040 	struct list_head		children;
1041 	struct list_head		sibling;
1042 	struct task_struct		*group_leader;
1043 
1044 	/*
1045 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
1046 	 *
1047 	 * This includes both natural children and PTRACE_ATTACH targets.
1048 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1049 	 */
1050 	struct list_head		ptraced;
1051 	struct list_head		ptrace_entry;
1052 
1053 	/* PID/PID hash table linkage. */
1054 	struct pid			*thread_pid;
1055 	struct hlist_node		pid_links[PIDTYPE_MAX];
1056 	struct list_head		thread_node;
1057 
1058 	struct completion		*vfork_done;
1059 
1060 	/* CLONE_CHILD_SETTID: */
1061 	int __user			*set_child_tid;
1062 
1063 	/* CLONE_CHILD_CLEARTID: */
1064 	int __user			*clear_child_tid;
1065 
1066 	/* PF_KTHREAD | PF_IO_WORKER */
1067 	void				*worker_private;
1068 
1069 	u64				utime;
1070 	u64				stime;
1071 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1072 	u64				utimescaled;
1073 	u64				stimescaled;
1074 #endif
1075 	u64				gtime;
1076 	struct prev_cputime		prev_cputime;
1077 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1078 	struct vtime			vtime;
1079 #endif
1080 
1081 #ifdef CONFIG_NO_HZ_FULL
1082 	atomic_t			tick_dep_mask;
1083 #endif
1084 	/* Context switch counts: */
1085 	unsigned long			nvcsw;
1086 	unsigned long			nivcsw;
1087 
1088 	/* Monotonic time in nsecs: */
1089 	u64				start_time;
1090 
1091 	/* Boot based time in nsecs: */
1092 	u64				start_boottime;
1093 
1094 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1095 	unsigned long			min_flt;
1096 	unsigned long			maj_flt;
1097 
1098 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1099 	struct posix_cputimers		posix_cputimers;
1100 
1101 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1102 	struct posix_cputimers_work	posix_cputimers_work;
1103 #endif
1104 
1105 	/* Process credentials: */
1106 
1107 	/* Tracer's credentials at attach: */
1108 	const struct cred __rcu		*ptracer_cred;
1109 
1110 	/* Objective and real subjective task credentials (COW): */
1111 	const struct cred __rcu		*real_cred;
1112 
1113 	/* Effective (overridable) subjective task credentials (COW): */
1114 	const struct cred __rcu		*cred;
1115 
1116 #ifdef CONFIG_KEYS
1117 	/* Cached requested key. */
1118 	struct key			*cached_requested_key;
1119 #endif
1120 
1121 	/*
1122 	 * executable name, excluding path.
1123 	 *
1124 	 * - normally initialized setup_new_exec()
1125 	 * - access it with [gs]et_task_comm()
1126 	 * - lock it with task_lock()
1127 	 */
1128 	char				comm[TASK_COMM_LEN];
1129 
1130 	struct nameidata		*nameidata;
1131 
1132 #ifdef CONFIG_SYSVIPC
1133 	struct sysv_sem			sysvsem;
1134 	struct sysv_shm			sysvshm;
1135 #endif
1136 #ifdef CONFIG_DETECT_HUNG_TASK
1137 	unsigned long			last_switch_count;
1138 	unsigned long			last_switch_time;
1139 #endif
1140 	/* Filesystem information: */
1141 	struct fs_struct		*fs;
1142 
1143 	/* Open file information: */
1144 	struct files_struct		*files;
1145 
1146 #ifdef CONFIG_IO_URING
1147 	struct io_uring_task		*io_uring;
1148 #endif
1149 
1150 	/* Namespaces: */
1151 	struct nsproxy			*nsproxy;
1152 
1153 	/* Signal handlers: */
1154 	struct signal_struct		*signal;
1155 	struct sighand_struct __rcu		*sighand;
1156 	sigset_t			blocked;
1157 	sigset_t			real_blocked;
1158 	/* Restored if set_restore_sigmask() was used: */
1159 	sigset_t			saved_sigmask;
1160 	struct sigpending		pending;
1161 	unsigned long			sas_ss_sp;
1162 	size_t				sas_ss_size;
1163 	unsigned int			sas_ss_flags;
1164 
1165 	struct callback_head		*task_works;
1166 
1167 #ifdef CONFIG_AUDIT
1168 #ifdef CONFIG_AUDITSYSCALL
1169 	struct audit_context		*audit_context;
1170 #endif
1171 	kuid_t				loginuid;
1172 	unsigned int			sessionid;
1173 #endif
1174 	struct seccomp			seccomp;
1175 	struct syscall_user_dispatch	syscall_dispatch;
1176 
1177 	/* Thread group tracking: */
1178 	u64				parent_exec_id;
1179 	u64				self_exec_id;
1180 
1181 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1182 	spinlock_t			alloc_lock;
1183 
1184 	/* Protection of the PI data structures: */
1185 	raw_spinlock_t			pi_lock;
1186 
1187 	struct wake_q_node		wake_q;
1188 
1189 #ifdef CONFIG_RT_MUTEXES
1190 	/* PI waiters blocked on a rt_mutex held by this task: */
1191 	struct rb_root_cached		pi_waiters;
1192 	/* Updated under owner's pi_lock and rq lock */
1193 	struct task_struct		*pi_top_task;
1194 	/* Deadlock detection and priority inheritance handling: */
1195 	struct rt_mutex_waiter		*pi_blocked_on;
1196 #endif
1197 
1198 #ifdef CONFIG_DEBUG_MUTEXES
1199 	/* Mutex deadlock detection: */
1200 	struct mutex_waiter		*blocked_on;
1201 #endif
1202 
1203 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1204 	int				non_block_count;
1205 #endif
1206 
1207 #ifdef CONFIG_TRACE_IRQFLAGS
1208 	struct irqtrace_events		irqtrace;
1209 	unsigned int			hardirq_threaded;
1210 	u64				hardirq_chain_key;
1211 	int				softirqs_enabled;
1212 	int				softirq_context;
1213 	int				irq_config;
1214 #endif
1215 #ifdef CONFIG_PREEMPT_RT
1216 	int				softirq_disable_cnt;
1217 #endif
1218 
1219 #ifdef CONFIG_LOCKDEP
1220 # define MAX_LOCK_DEPTH			48UL
1221 	u64				curr_chain_key;
1222 	int				lockdep_depth;
1223 	unsigned int			lockdep_recursion;
1224 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1225 #endif
1226 
1227 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1228 	unsigned int			in_ubsan;
1229 #endif
1230 
1231 	/* Journalling filesystem info: */
1232 	void				*journal_info;
1233 
1234 	/* Stacked block device info: */
1235 	struct bio_list			*bio_list;
1236 
1237 	/* Stack plugging: */
1238 	struct blk_plug			*plug;
1239 
1240 	/* VM state: */
1241 	struct reclaim_state		*reclaim_state;
1242 
1243 	struct io_context		*io_context;
1244 
1245 #ifdef CONFIG_COMPACTION
1246 	struct capture_control		*capture_control;
1247 #endif
1248 	/* Ptrace state: */
1249 	unsigned long			ptrace_message;
1250 	kernel_siginfo_t		*last_siginfo;
1251 
1252 	struct task_io_accounting	ioac;
1253 #ifdef CONFIG_PSI
1254 	/* Pressure stall state */
1255 	unsigned int			psi_flags;
1256 #endif
1257 #ifdef CONFIG_TASK_XACCT
1258 	/* Accumulated RSS usage: */
1259 	u64				acct_rss_mem1;
1260 	/* Accumulated virtual memory usage: */
1261 	u64				acct_vm_mem1;
1262 	/* stime + utime since last update: */
1263 	u64				acct_timexpd;
1264 #endif
1265 #ifdef CONFIG_CPUSETS
1266 	/* Protected by ->alloc_lock: */
1267 	nodemask_t			mems_allowed;
1268 	/* Sequence number to catch updates: */
1269 	seqcount_spinlock_t		mems_allowed_seq;
1270 	int				cpuset_mem_spread_rotor;
1271 #endif
1272 #ifdef CONFIG_CGROUPS
1273 	/* Control Group info protected by css_set_lock: */
1274 	struct css_set __rcu		*cgroups;
1275 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1276 	struct list_head		cg_list;
1277 #endif
1278 #ifdef CONFIG_X86_CPU_RESCTRL
1279 	u32				closid;
1280 	u32				rmid;
1281 #endif
1282 #ifdef CONFIG_FUTEX
1283 	struct robust_list_head __user	*robust_list;
1284 #ifdef CONFIG_COMPAT
1285 	struct compat_robust_list_head __user *compat_robust_list;
1286 #endif
1287 	struct list_head		pi_state_list;
1288 	struct futex_pi_state		*pi_state_cache;
1289 	struct mutex			futex_exit_mutex;
1290 	unsigned int			futex_state;
1291 #endif
1292 #ifdef CONFIG_PERF_EVENTS
1293 	u8				perf_recursion[PERF_NR_CONTEXTS];
1294 	struct perf_event_context	*perf_event_ctxp;
1295 	struct mutex			perf_event_mutex;
1296 	struct list_head		perf_event_list;
1297 #endif
1298 #ifdef CONFIG_DEBUG_PREEMPT
1299 	unsigned long			preempt_disable_ip;
1300 #endif
1301 #ifdef CONFIG_NUMA
1302 	/* Protected by alloc_lock: */
1303 	struct mempolicy		*mempolicy;
1304 	short				il_prev;
1305 	u8				il_weight;
1306 	short				pref_node_fork;
1307 #endif
1308 #ifdef CONFIG_NUMA_BALANCING
1309 	int				numa_scan_seq;
1310 	unsigned int			numa_scan_period;
1311 	unsigned int			numa_scan_period_max;
1312 	int				numa_preferred_nid;
1313 	unsigned long			numa_migrate_retry;
1314 	/* Migration stamp: */
1315 	u64				node_stamp;
1316 	u64				last_task_numa_placement;
1317 	u64				last_sum_exec_runtime;
1318 	struct callback_head		numa_work;
1319 
1320 	/*
1321 	 * This pointer is only modified for current in syscall and
1322 	 * pagefault context (and for tasks being destroyed), so it can be read
1323 	 * from any of the following contexts:
1324 	 *  - RCU read-side critical section
1325 	 *  - current->numa_group from everywhere
1326 	 *  - task's runqueue locked, task not running
1327 	 */
1328 	struct numa_group __rcu		*numa_group;
1329 
1330 	/*
1331 	 * numa_faults is an array split into four regions:
1332 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1333 	 * in this precise order.
1334 	 *
1335 	 * faults_memory: Exponential decaying average of faults on a per-node
1336 	 * basis. Scheduling placement decisions are made based on these
1337 	 * counts. The values remain static for the duration of a PTE scan.
1338 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1339 	 * hinting fault was incurred.
1340 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1341 	 * during the current scan window. When the scan completes, the counts
1342 	 * in faults_memory and faults_cpu decay and these values are copied.
1343 	 */
1344 	unsigned long			*numa_faults;
1345 	unsigned long			total_numa_faults;
1346 
1347 	/*
1348 	 * numa_faults_locality tracks if faults recorded during the last
1349 	 * scan window were remote/local or failed to migrate. The task scan
1350 	 * period is adapted based on the locality of the faults with different
1351 	 * weights depending on whether they were shared or private faults
1352 	 */
1353 	unsigned long			numa_faults_locality[3];
1354 
1355 	unsigned long			numa_pages_migrated;
1356 #endif /* CONFIG_NUMA_BALANCING */
1357 
1358 #ifdef CONFIG_RSEQ
1359 	struct rseq __user *rseq;
1360 	u32 rseq_len;
1361 	u32 rseq_sig;
1362 	/*
1363 	 * RmW on rseq_event_mask must be performed atomically
1364 	 * with respect to preemption.
1365 	 */
1366 	unsigned long rseq_event_mask;
1367 #endif
1368 
1369 #ifdef CONFIG_SCHED_MM_CID
1370 	int				mm_cid;		/* Current cid in mm */
1371 	int				last_mm_cid;	/* Most recent cid in mm */
1372 	int				migrate_from_cpu;
1373 	int				mm_cid_active;	/* Whether cid bitmap is active */
1374 	struct callback_head		cid_work;
1375 #endif
1376 
1377 	struct tlbflush_unmap_batch	tlb_ubc;
1378 
1379 	/* Cache last used pipe for splice(): */
1380 	struct pipe_inode_info		*splice_pipe;
1381 
1382 	struct page_frag		task_frag;
1383 
1384 #ifdef CONFIG_TASK_DELAY_ACCT
1385 	struct task_delay_info		*delays;
1386 #endif
1387 
1388 #ifdef CONFIG_FAULT_INJECTION
1389 	int				make_it_fail;
1390 	unsigned int			fail_nth;
1391 #endif
1392 	/*
1393 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1394 	 * balance_dirty_pages() for a dirty throttling pause:
1395 	 */
1396 	int				nr_dirtied;
1397 	int				nr_dirtied_pause;
1398 	/* Start of a write-and-pause period: */
1399 	unsigned long			dirty_paused_when;
1400 
1401 #ifdef CONFIG_LATENCYTOP
1402 	int				latency_record_count;
1403 	struct latency_record		latency_record[LT_SAVECOUNT];
1404 #endif
1405 	/*
1406 	 * Time slack values; these are used to round up poll() and
1407 	 * select() etc timeout values. These are in nanoseconds.
1408 	 */
1409 	u64				timer_slack_ns;
1410 	u64				default_timer_slack_ns;
1411 
1412 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1413 	unsigned int			kasan_depth;
1414 #endif
1415 
1416 #ifdef CONFIG_KCSAN
1417 	struct kcsan_ctx		kcsan_ctx;
1418 #ifdef CONFIG_TRACE_IRQFLAGS
1419 	struct irqtrace_events		kcsan_save_irqtrace;
1420 #endif
1421 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1422 	int				kcsan_stack_depth;
1423 #endif
1424 #endif
1425 
1426 #ifdef CONFIG_KMSAN
1427 	struct kmsan_ctx		kmsan_ctx;
1428 #endif
1429 
1430 #if IS_ENABLED(CONFIG_KUNIT)
1431 	struct kunit			*kunit_test;
1432 #endif
1433 
1434 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1435 	/* Index of current stored address in ret_stack: */
1436 	int				curr_ret_stack;
1437 	int				curr_ret_depth;
1438 
1439 	/* Stack of return addresses for return function tracing: */
1440 	unsigned long			*ret_stack;
1441 
1442 	/* Timestamp for last schedule: */
1443 	unsigned long long		ftrace_timestamp;
1444 
1445 	/*
1446 	 * Number of functions that haven't been traced
1447 	 * because of depth overrun:
1448 	 */
1449 	atomic_t			trace_overrun;
1450 
1451 	/* Pause tracing: */
1452 	atomic_t			tracing_graph_pause;
1453 #endif
1454 
1455 #ifdef CONFIG_TRACING
1456 	/* Bitmask and counter of trace recursion: */
1457 	unsigned long			trace_recursion;
1458 #endif /* CONFIG_TRACING */
1459 
1460 #ifdef CONFIG_KCOV
1461 	/* See kernel/kcov.c for more details. */
1462 
1463 	/* Coverage collection mode enabled for this task (0 if disabled): */
1464 	unsigned int			kcov_mode;
1465 
1466 	/* Size of the kcov_area: */
1467 	unsigned int			kcov_size;
1468 
1469 	/* Buffer for coverage collection: */
1470 	void				*kcov_area;
1471 
1472 	/* KCOV descriptor wired with this task or NULL: */
1473 	struct kcov			*kcov;
1474 
1475 	/* KCOV common handle for remote coverage collection: */
1476 	u64				kcov_handle;
1477 
1478 	/* KCOV sequence number: */
1479 	int				kcov_sequence;
1480 
1481 	/* Collect coverage from softirq context: */
1482 	unsigned int			kcov_softirq;
1483 #endif
1484 
1485 #ifdef CONFIG_MEMCG_V1
1486 	struct mem_cgroup		*memcg_in_oom;
1487 #endif
1488 
1489 #ifdef CONFIG_MEMCG
1490 	/* Number of pages to reclaim on returning to userland: */
1491 	unsigned int			memcg_nr_pages_over_high;
1492 
1493 	/* Used by memcontrol for targeted memcg charge: */
1494 	struct mem_cgroup		*active_memcg;
1495 
1496 	/* Cache for current->cgroups->memcg->objcg lookups: */
1497 	struct obj_cgroup		*objcg;
1498 #endif
1499 
1500 #ifdef CONFIG_BLK_CGROUP
1501 	struct gendisk			*throttle_disk;
1502 #endif
1503 
1504 #ifdef CONFIG_UPROBES
1505 	struct uprobe_task		*utask;
1506 #endif
1507 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1508 	unsigned int			sequential_io;
1509 	unsigned int			sequential_io_avg;
1510 #endif
1511 	struct kmap_ctrl		kmap_ctrl;
1512 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1513 	unsigned long			task_state_change;
1514 # ifdef CONFIG_PREEMPT_RT
1515 	unsigned long			saved_state_change;
1516 # endif
1517 #endif
1518 	struct rcu_head			rcu;
1519 	refcount_t			rcu_users;
1520 	int				pagefault_disabled;
1521 #ifdef CONFIG_MMU
1522 	struct task_struct		*oom_reaper_list;
1523 	struct timer_list		oom_reaper_timer;
1524 #endif
1525 #ifdef CONFIG_VMAP_STACK
1526 	struct vm_struct		*stack_vm_area;
1527 #endif
1528 #ifdef CONFIG_THREAD_INFO_IN_TASK
1529 	/* A live task holds one reference: */
1530 	refcount_t			stack_refcount;
1531 #endif
1532 #ifdef CONFIG_LIVEPATCH
1533 	int patch_state;
1534 #endif
1535 #ifdef CONFIG_SECURITY
1536 	/* Used by LSM modules for access restriction: */
1537 	void				*security;
1538 #endif
1539 #ifdef CONFIG_BPF_SYSCALL
1540 	/* Used by BPF task local storage */
1541 	struct bpf_local_storage __rcu	*bpf_storage;
1542 	/* Used for BPF run context */
1543 	struct bpf_run_ctx		*bpf_ctx;
1544 #endif
1545 	/* Used by BPF for per-TASK xdp storage */
1546 	struct bpf_net_context		*bpf_net_context;
1547 
1548 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1549 	unsigned long			lowest_stack;
1550 	unsigned long			prev_lowest_stack;
1551 #endif
1552 
1553 #ifdef CONFIG_X86_MCE
1554 	void __user			*mce_vaddr;
1555 	__u64				mce_kflags;
1556 	u64				mce_addr;
1557 	__u64				mce_ripv : 1,
1558 					mce_whole_page : 1,
1559 					__mce_reserved : 62;
1560 	struct callback_head		mce_kill_me;
1561 	int				mce_count;
1562 #endif
1563 
1564 #ifdef CONFIG_KRETPROBES
1565 	struct llist_head               kretprobe_instances;
1566 #endif
1567 #ifdef CONFIG_RETHOOK
1568 	struct llist_head               rethooks;
1569 #endif
1570 
1571 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1572 	/*
1573 	 * If L1D flush is supported on mm context switch
1574 	 * then we use this callback head to queue kill work
1575 	 * to kill tasks that are not running on SMT disabled
1576 	 * cores
1577 	 */
1578 	struct callback_head		l1d_flush_kill;
1579 #endif
1580 
1581 #ifdef CONFIG_RV
1582 	/*
1583 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1584 	 * If we find justification for more monitors, we can think
1585 	 * about adding more or developing a dynamic method. So far,
1586 	 * none of these are justified.
1587 	 */
1588 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1589 #endif
1590 
1591 #ifdef CONFIG_USER_EVENTS
1592 	struct user_event_mm		*user_event_mm;
1593 #endif
1594 
1595 	/*
1596 	 * New fields for task_struct should be added above here, so that
1597 	 * they are included in the randomized portion of task_struct.
1598 	 */
1599 	randomized_struct_fields_end
1600 
1601 	/* CPU-specific state of this task: */
1602 	struct thread_struct		thread;
1603 
1604 	/*
1605 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1606 	 * structure.  It *MUST* be at the end of 'task_struct'.
1607 	 *
1608 	 * Do not put anything below here!
1609 	 */
1610 };
1611 
1612 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1613 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1614 
__task_state_index(unsigned int tsk_state,unsigned int tsk_exit_state)1615 static inline unsigned int __task_state_index(unsigned int tsk_state,
1616 					      unsigned int tsk_exit_state)
1617 {
1618 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1619 
1620 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1621 
1622 	if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1623 		state = TASK_REPORT_IDLE;
1624 
1625 	/*
1626 	 * We're lying here, but rather than expose a completely new task state
1627 	 * to userspace, we can make this appear as if the task has gone through
1628 	 * a regular rt_mutex_lock() call.
1629 	 */
1630 	if (tsk_state & TASK_RTLOCK_WAIT)
1631 		state = TASK_UNINTERRUPTIBLE;
1632 
1633 	return fls(state);
1634 }
1635 
task_state_index(struct task_struct * tsk)1636 static inline unsigned int task_state_index(struct task_struct *tsk)
1637 {
1638 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1639 }
1640 
task_index_to_char(unsigned int state)1641 static inline char task_index_to_char(unsigned int state)
1642 {
1643 	static const char state_char[] = "RSDTtXZPI";
1644 
1645 	BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1646 
1647 	return state_char[state];
1648 }
1649 
task_state_to_char(struct task_struct * tsk)1650 static inline char task_state_to_char(struct task_struct *tsk)
1651 {
1652 	return task_index_to_char(task_state_index(tsk));
1653 }
1654 
1655 extern struct pid *cad_pid;
1656 
1657 /*
1658  * Per process flags
1659  */
1660 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1661 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1662 #define PF_EXITING		0x00000004	/* Getting shut down */
1663 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1664 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1665 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1666 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1667 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1668 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1669 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1670 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1671 #define PF_MEMALLOC		0x00000800	/* Allocating memory to free memory. See memalloc_noreclaim_save() */
1672 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1673 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1674 #define PF_USER_WORKER		0x00004000	/* Kernel thread cloned from userspace thread */
1675 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1676 #define PF__HOLE__00010000	0x00010000
1677 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1678 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1679 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1680 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1681 						 * I am cleaning dirty pages from some other bdi. */
1682 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1683 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1684 #define PF__HOLE__00800000	0x00800000
1685 #define PF__HOLE__01000000	0x01000000
1686 #define PF__HOLE__02000000	0x02000000
1687 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1688 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1689 #define PF_MEMALLOC_PIN		0x10000000	/* Allocations constrained to zones which allow long term pinning.
1690 						 * See memalloc_pin_save() */
1691 #define PF_BLOCK_TS		0x20000000	/* plug has ts that needs updating */
1692 #define PF__HOLE__40000000	0x40000000
1693 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1694 
1695 /*
1696  * Only the _current_ task can read/write to tsk->flags, but other
1697  * tasks can access tsk->flags in readonly mode for example
1698  * with tsk_used_math (like during threaded core dumping).
1699  * There is however an exception to this rule during ptrace
1700  * or during fork: the ptracer task is allowed to write to the
1701  * child->flags of its traced child (same goes for fork, the parent
1702  * can write to the child->flags), because we're guaranteed the
1703  * child is not running and in turn not changing child->flags
1704  * at the same time the parent does it.
1705  */
1706 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1707 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1708 #define clear_used_math()			clear_stopped_child_used_math(current)
1709 #define set_used_math()				set_stopped_child_used_math(current)
1710 
1711 #define conditional_stopped_child_used_math(condition, child) \
1712 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1713 
1714 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1715 
1716 #define copy_to_stopped_child_used_math(child) \
1717 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1718 
1719 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1720 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1721 #define used_math()				tsk_used_math(current)
1722 
is_percpu_thread(void)1723 static __always_inline bool is_percpu_thread(void)
1724 {
1725 #ifdef CONFIG_SMP
1726 	return (current->flags & PF_NO_SETAFFINITY) &&
1727 		(current->nr_cpus_allowed  == 1);
1728 #else
1729 	return true;
1730 #endif
1731 }
1732 
1733 /* Per-process atomic flags. */
1734 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1735 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1736 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1737 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1738 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1739 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1740 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1741 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1742 
1743 #define TASK_PFA_TEST(name, func)					\
1744 	static inline bool task_##func(struct task_struct *p)		\
1745 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1746 
1747 #define TASK_PFA_SET(name, func)					\
1748 	static inline void task_set_##func(struct task_struct *p)	\
1749 	{ set_bit(PFA_##name, &p->atomic_flags); }
1750 
1751 #define TASK_PFA_CLEAR(name, func)					\
1752 	static inline void task_clear_##func(struct task_struct *p)	\
1753 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1754 
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1755 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1756 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1757 
1758 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1759 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1760 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1761 
1762 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1763 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1764 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1765 
1766 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1767 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1768 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1769 
1770 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1771 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1772 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1773 
1774 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1775 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1776 
1777 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1778 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1779 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1780 
1781 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1782 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1783 
1784 static inline void
1785 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1786 {
1787 	current->flags &= ~flags;
1788 	current->flags |= orig_flags & flags;
1789 }
1790 
1791 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1792 extern int task_can_attach(struct task_struct *p);
1793 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1794 extern void dl_bw_free(int cpu, u64 dl_bw);
1795 #ifdef CONFIG_SMP
1796 
1797 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1798 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1799 
1800 /**
1801  * set_cpus_allowed_ptr - set CPU affinity mask of a task
1802  * @p: the task
1803  * @new_mask: CPU affinity mask
1804  *
1805  * Return: zero if successful, or a negative error code
1806  */
1807 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1808 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1809 extern void release_user_cpus_ptr(struct task_struct *p);
1810 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1811 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1812 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1813 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1814 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1815 {
1816 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1817 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1818 {
1819 	/* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1820 	if ((*cpumask_bits(new_mask) & 1) == 0)
1821 		return -EINVAL;
1822 	return 0;
1823 }
dup_user_cpus_ptr(struct task_struct * dst,struct task_struct * src,int node)1824 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1825 {
1826 	if (src->user_cpus_ptr)
1827 		return -EINVAL;
1828 	return 0;
1829 }
release_user_cpus_ptr(struct task_struct * p)1830 static inline void release_user_cpus_ptr(struct task_struct *p)
1831 {
1832 	WARN_ON(p->user_cpus_ptr);
1833 }
1834 
dl_task_check_affinity(struct task_struct * p,const struct cpumask * mask)1835 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1836 {
1837 	return 0;
1838 }
1839 #endif
1840 
1841 extern int yield_to(struct task_struct *p, bool preempt);
1842 extern void set_user_nice(struct task_struct *p, long nice);
1843 extern int task_prio(const struct task_struct *p);
1844 
1845 /**
1846  * task_nice - return the nice value of a given task.
1847  * @p: the task in question.
1848  *
1849  * Return: The nice value [ -20 ... 0 ... 19 ].
1850  */
task_nice(const struct task_struct * p)1851 static inline int task_nice(const struct task_struct *p)
1852 {
1853 	return PRIO_TO_NICE((p)->static_prio);
1854 }
1855 
1856 extern int can_nice(const struct task_struct *p, const int nice);
1857 extern int task_curr(const struct task_struct *p);
1858 extern int idle_cpu(int cpu);
1859 extern int available_idle_cpu(int cpu);
1860 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1861 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1862 extern void sched_set_fifo(struct task_struct *p);
1863 extern void sched_set_fifo_low(struct task_struct *p);
1864 extern void sched_set_normal(struct task_struct *p, int nice);
1865 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1866 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1867 extern struct task_struct *idle_task(int cpu);
1868 
1869 /**
1870  * is_idle_task - is the specified task an idle task?
1871  * @p: the task in question.
1872  *
1873  * Return: 1 if @p is an idle task. 0 otherwise.
1874  */
is_idle_task(const struct task_struct * p)1875 static __always_inline bool is_idle_task(const struct task_struct *p)
1876 {
1877 	return !!(p->flags & PF_IDLE);
1878 }
1879 
1880 extern struct task_struct *curr_task(int cpu);
1881 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1882 
1883 void yield(void);
1884 
1885 union thread_union {
1886 	struct task_struct task;
1887 #ifndef CONFIG_THREAD_INFO_IN_TASK
1888 	struct thread_info thread_info;
1889 #endif
1890 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1891 };
1892 
1893 #ifndef CONFIG_THREAD_INFO_IN_TASK
1894 extern struct thread_info init_thread_info;
1895 #endif
1896 
1897 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1898 
1899 #ifdef CONFIG_THREAD_INFO_IN_TASK
1900 # define task_thread_info(task)	(&(task)->thread_info)
1901 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1902 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1903 #endif
1904 
1905 /*
1906  * find a task by one of its numerical ids
1907  *
1908  * find_task_by_pid_ns():
1909  *      finds a task by its pid in the specified namespace
1910  * find_task_by_vpid():
1911  *      finds a task by its virtual pid
1912  *
1913  * see also find_vpid() etc in include/linux/pid.h
1914  */
1915 
1916 extern struct task_struct *find_task_by_vpid(pid_t nr);
1917 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1918 
1919 /*
1920  * find a task by its virtual pid and get the task struct
1921  */
1922 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1923 
1924 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1925 extern int wake_up_process(struct task_struct *tsk);
1926 extern void wake_up_new_task(struct task_struct *tsk);
1927 
1928 #ifdef CONFIG_SMP
1929 extern void kick_process(struct task_struct *tsk);
1930 #else
kick_process(struct task_struct * tsk)1931 static inline void kick_process(struct task_struct *tsk) { }
1932 #endif
1933 
1934 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1935 
set_task_comm(struct task_struct * tsk,const char * from)1936 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1937 {
1938 	__set_task_comm(tsk, from, false);
1939 }
1940 
1941 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1942 #define get_task_comm(buf, tsk) ({			\
1943 	BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);	\
1944 	__get_task_comm(buf, sizeof(buf), tsk);		\
1945 })
1946 
1947 #ifdef CONFIG_SMP
scheduler_ipi(void)1948 static __always_inline void scheduler_ipi(void)
1949 {
1950 	/*
1951 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1952 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1953 	 * this IPI.
1954 	 */
1955 	preempt_fold_need_resched();
1956 }
1957 #else
scheduler_ipi(void)1958 static inline void scheduler_ipi(void) { }
1959 #endif
1960 
1961 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1962 
1963 /*
1964  * Set thread flags in other task's structures.
1965  * See asm/thread_info.h for TIF_xxxx flags available:
1966  */
set_tsk_thread_flag(struct task_struct * tsk,int flag)1967 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1968 {
1969 	set_ti_thread_flag(task_thread_info(tsk), flag);
1970 }
1971 
clear_tsk_thread_flag(struct task_struct * tsk,int flag)1972 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1973 {
1974 	clear_ti_thread_flag(task_thread_info(tsk), flag);
1975 }
1976 
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)1977 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1978 					  bool value)
1979 {
1980 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
1981 }
1982 
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)1983 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1984 {
1985 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1986 }
1987 
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)1988 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1989 {
1990 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1991 }
1992 
test_tsk_thread_flag(struct task_struct * tsk,int flag)1993 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1994 {
1995 	return test_ti_thread_flag(task_thread_info(tsk), flag);
1996 }
1997 
set_tsk_need_resched(struct task_struct * tsk)1998 static inline void set_tsk_need_resched(struct task_struct *tsk)
1999 {
2000 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2001 }
2002 
clear_tsk_need_resched(struct task_struct * tsk)2003 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2004 {
2005 	clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2006 }
2007 
test_tsk_need_resched(struct task_struct * tsk)2008 static inline int test_tsk_need_resched(struct task_struct *tsk)
2009 {
2010 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2011 }
2012 
2013 /*
2014  * cond_resched() and cond_resched_lock(): latency reduction via
2015  * explicit rescheduling in places that are safe. The return
2016  * value indicates whether a reschedule was done in fact.
2017  * cond_resched_lock() will drop the spinlock before scheduling,
2018  */
2019 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2020 extern int __cond_resched(void);
2021 
2022 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2023 
2024 void sched_dynamic_klp_enable(void);
2025 void sched_dynamic_klp_disable(void);
2026 
2027 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2028 
_cond_resched(void)2029 static __always_inline int _cond_resched(void)
2030 {
2031 	return static_call_mod(cond_resched)();
2032 }
2033 
2034 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2035 
2036 extern int dynamic_cond_resched(void);
2037 
_cond_resched(void)2038 static __always_inline int _cond_resched(void)
2039 {
2040 	return dynamic_cond_resched();
2041 }
2042 
2043 #else /* !CONFIG_PREEMPTION */
2044 
_cond_resched(void)2045 static inline int _cond_resched(void)
2046 {
2047 	klp_sched_try_switch();
2048 	return __cond_resched();
2049 }
2050 
2051 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2052 
2053 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2054 
_cond_resched(void)2055 static inline int _cond_resched(void)
2056 {
2057 	klp_sched_try_switch();
2058 	return 0;
2059 }
2060 
2061 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2062 
2063 #define cond_resched() ({			\
2064 	__might_resched(__FILE__, __LINE__, 0);	\
2065 	_cond_resched();			\
2066 })
2067 
2068 extern int __cond_resched_lock(spinlock_t *lock);
2069 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2070 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2071 
2072 #define MIGHT_RESCHED_RCU_SHIFT		8
2073 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2074 
2075 #ifndef CONFIG_PREEMPT_RT
2076 /*
2077  * Non RT kernels have an elevated preempt count due to the held lock,
2078  * but are not allowed to be inside a RCU read side critical section
2079  */
2080 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2081 #else
2082 /*
2083  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2084  * cond_resched*lock() has to take that into account because it checks for
2085  * preempt_count() and rcu_preempt_depth().
2086  */
2087 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2088 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2089 #endif
2090 
2091 #define cond_resched_lock(lock) ({						\
2092 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2093 	__cond_resched_lock(lock);						\
2094 })
2095 
2096 #define cond_resched_rwlock_read(lock) ({					\
2097 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2098 	__cond_resched_rwlock_read(lock);					\
2099 })
2100 
2101 #define cond_resched_rwlock_write(lock) ({					\
2102 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2103 	__cond_resched_rwlock_write(lock);					\
2104 })
2105 
need_resched(void)2106 static __always_inline bool need_resched(void)
2107 {
2108 	return unlikely(tif_need_resched());
2109 }
2110 
2111 /*
2112  * Wrappers for p->thread_info->cpu access. No-op on UP.
2113  */
2114 #ifdef CONFIG_SMP
2115 
task_cpu(const struct task_struct * p)2116 static inline unsigned int task_cpu(const struct task_struct *p)
2117 {
2118 	return READ_ONCE(task_thread_info(p)->cpu);
2119 }
2120 
2121 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2122 
2123 #else
2124 
task_cpu(const struct task_struct * p)2125 static inline unsigned int task_cpu(const struct task_struct *p)
2126 {
2127 	return 0;
2128 }
2129 
set_task_cpu(struct task_struct * p,unsigned int cpu)2130 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2131 {
2132 }
2133 
2134 #endif /* CONFIG_SMP */
2135 
task_is_runnable(struct task_struct * p)2136 static inline bool task_is_runnable(struct task_struct *p)
2137 {
2138 	return p->on_rq && !p->se.sched_delayed;
2139 }
2140 
2141 extern bool sched_task_on_rq(struct task_struct *p);
2142 extern unsigned long get_wchan(struct task_struct *p);
2143 extern struct task_struct *cpu_curr_snapshot(int cpu);
2144 
2145 #include <linux/spinlock.h>
2146 
2147 /*
2148  * In order to reduce various lock holder preemption latencies provide an
2149  * interface to see if a vCPU is currently running or not.
2150  *
2151  * This allows us to terminate optimistic spin loops and block, analogous to
2152  * the native optimistic spin heuristic of testing if the lock owner task is
2153  * running or not.
2154  */
2155 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2156 static inline bool vcpu_is_preempted(int cpu)
2157 {
2158 	return false;
2159 }
2160 #endif
2161 
2162 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2163 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2164 
2165 #ifndef TASK_SIZE_OF
2166 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2167 #endif
2168 
2169 #ifdef CONFIG_SMP
owner_on_cpu(struct task_struct * owner)2170 static inline bool owner_on_cpu(struct task_struct *owner)
2171 {
2172 	/*
2173 	 * As lock holder preemption issue, we both skip spinning if
2174 	 * task is not on cpu or its cpu is preempted
2175 	 */
2176 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2177 }
2178 
2179 /* Returns effective CPU energy utilization, as seen by the scheduler */
2180 unsigned long sched_cpu_util(int cpu);
2181 #endif /* CONFIG_SMP */
2182 
2183 #ifdef CONFIG_SCHED_CORE
2184 extern void sched_core_free(struct task_struct *tsk);
2185 extern void sched_core_fork(struct task_struct *p);
2186 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2187 				unsigned long uaddr);
2188 extern int sched_core_idle_cpu(int cpu);
2189 #else
sched_core_free(struct task_struct * tsk)2190 static inline void sched_core_free(struct task_struct *tsk) { }
sched_core_fork(struct task_struct * p)2191 static inline void sched_core_fork(struct task_struct *p) { }
sched_core_idle_cpu(int cpu)2192 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2193 #endif
2194 
2195 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2196 
2197 #ifdef CONFIG_MEM_ALLOC_PROFILING
alloc_tag_save(struct alloc_tag * tag)2198 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2199 {
2200 	swap(current->alloc_tag, tag);
2201 	return tag;
2202 }
2203 
alloc_tag_restore(struct alloc_tag * tag,struct alloc_tag * old)2204 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2205 {
2206 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2207 	WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2208 #endif
2209 	current->alloc_tag = old;
2210 }
2211 #else
2212 #define alloc_tag_save(_tag)			NULL
2213 #define alloc_tag_restore(_tag, _old)		do {} while (0)
2214 #endif
2215 
2216 #endif
2217