xref: /linux/include/linux/sched.h (revision 1110ce6a1e34fe1fdc1bfe4ad52405f327d5083b)
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 	/* Max time spent waiting on a runqueue: */
402 	unsigned long long		max_run_delay;
403 
404 	/* Min time spent waiting on a runqueue: */
405 	unsigned long long		min_run_delay;
406 
407 	/* Timestamps: */
408 
409 	/* When did we last run on a CPU? */
410 	unsigned long long		last_arrival;
411 
412 	/* When were we last queued to run? */
413 	unsigned long long		last_queued;
414 
415 #endif /* CONFIG_SCHED_INFO */
416 };
417 
418 /*
419  * Integer metrics need fixed point arithmetic, e.g., sched/fair
420  * has a few: load, load_avg, util_avg, freq, and capacity.
421  *
422  * We define a basic fixed point arithmetic range, and then formalize
423  * all these metrics based on that basic range.
424  */
425 # define SCHED_FIXEDPOINT_SHIFT		10
426 # define SCHED_FIXEDPOINT_SCALE		(1L << SCHED_FIXEDPOINT_SHIFT)
427 
428 /* Increase resolution of cpu_capacity calculations */
429 # define SCHED_CAPACITY_SHIFT		SCHED_FIXEDPOINT_SHIFT
430 # define SCHED_CAPACITY_SCALE		(1L << SCHED_CAPACITY_SHIFT)
431 
432 struct load_weight {
433 	unsigned long			weight;
434 	u32				inv_weight;
435 };
436 
437 /*
438  * The load/runnable/util_avg accumulates an infinite geometric series
439  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
440  *
441  * [load_avg definition]
442  *
443  *   load_avg = runnable% * scale_load_down(load)
444  *
445  * [runnable_avg definition]
446  *
447  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
448  *
449  * [util_avg definition]
450  *
451  *   util_avg = running% * SCHED_CAPACITY_SCALE
452  *
453  * where runnable% is the time ratio that a sched_entity is runnable and
454  * running% the time ratio that a sched_entity is running.
455  *
456  * For cfs_rq, they are the aggregated values of all runnable and blocked
457  * sched_entities.
458  *
459  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
460  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
461  * for computing those signals (see update_rq_clock_pelt())
462  *
463  * N.B., the above ratios (runnable% and running%) themselves are in the
464  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
465  * to as large a range as necessary. This is for example reflected by
466  * util_avg's SCHED_CAPACITY_SCALE.
467  *
468  * [Overflow issue]
469  *
470  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
471  * with the highest load (=88761), always runnable on a single cfs_rq,
472  * and should not overflow as the number already hits PID_MAX_LIMIT.
473  *
474  * For all other cases (including 32-bit kernels), struct load_weight's
475  * weight will overflow first before we do, because:
476  *
477  *    Max(load_avg) <= Max(load.weight)
478  *
479  * Then it is the load_weight's responsibility to consider overflow
480  * issues.
481  */
482 struct sched_avg {
483 	u64				last_update_time;
484 	u64				load_sum;
485 	u64				runnable_sum;
486 	u32				util_sum;
487 	u32				period_contrib;
488 	unsigned long			load_avg;
489 	unsigned long			runnable_avg;
490 	unsigned long			util_avg;
491 	unsigned int			util_est;
492 } ____cacheline_aligned;
493 
494 /*
495  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
496  * updates. When a task is dequeued, its util_est should not be updated if its
497  * util_avg has not been updated in the meantime.
498  * This information is mapped into the MSB bit of util_est at dequeue time.
499  * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
500  * it is safe to use MSB.
501  */
502 #define UTIL_EST_WEIGHT_SHIFT		2
503 #define UTIL_AVG_UNCHANGED		0x80000000
504 
505 struct sched_statistics {
506 #ifdef CONFIG_SCHEDSTATS
507 	u64				wait_start;
508 	u64				wait_max;
509 	u64				wait_count;
510 	u64				wait_sum;
511 	u64				iowait_count;
512 	u64				iowait_sum;
513 
514 	u64				sleep_start;
515 	u64				sleep_max;
516 	s64				sum_sleep_runtime;
517 
518 	u64				block_start;
519 	u64				block_max;
520 	s64				sum_block_runtime;
521 
522 	s64				exec_max;
523 	u64				slice_max;
524 
525 	u64				nr_migrations_cold;
526 	u64				nr_failed_migrations_affine;
527 	u64				nr_failed_migrations_running;
528 	u64				nr_failed_migrations_hot;
529 	u64				nr_forced_migrations;
530 
531 	u64				nr_wakeups;
532 	u64				nr_wakeups_sync;
533 	u64				nr_wakeups_migrate;
534 	u64				nr_wakeups_local;
535 	u64				nr_wakeups_remote;
536 	u64				nr_wakeups_affine;
537 	u64				nr_wakeups_affine_attempts;
538 	u64				nr_wakeups_passive;
539 	u64				nr_wakeups_idle;
540 
541 #ifdef CONFIG_SCHED_CORE
542 	u64				core_forceidle_sum;
543 #endif
544 #endif /* CONFIG_SCHEDSTATS */
545 } ____cacheline_aligned;
546 
547 struct sched_entity {
548 	/* For load-balancing: */
549 	struct load_weight		load;
550 	struct rb_node			run_node;
551 	u64				deadline;
552 	u64				min_vruntime;
553 	u64				min_slice;
554 
555 	struct list_head		group_node;
556 	unsigned char			on_rq;
557 	unsigned char			sched_delayed;
558 	unsigned char			rel_deadline;
559 	unsigned char			custom_slice;
560 					/* hole */
561 
562 	u64				exec_start;
563 	u64				sum_exec_runtime;
564 	u64				prev_sum_exec_runtime;
565 	u64				vruntime;
566 	s64				vlag;
567 	u64				slice;
568 
569 	u64				nr_migrations;
570 
571 #ifdef CONFIG_FAIR_GROUP_SCHED
572 	int				depth;
573 	struct sched_entity		*parent;
574 	/* rq on which this entity is (to be) queued: */
575 	struct cfs_rq			*cfs_rq;
576 	/* rq "owned" by this entity/group: */
577 	struct cfs_rq			*my_q;
578 	/* cached value of my_q->h_nr_running */
579 	unsigned long			runnable_weight;
580 #endif
581 
582 #ifdef CONFIG_SMP
583 	/*
584 	 * Per entity load average tracking.
585 	 *
586 	 * Put into separate cache line so it does not
587 	 * collide with read-mostly values above.
588 	 */
589 	struct sched_avg		avg;
590 #endif
591 };
592 
593 struct sched_rt_entity {
594 	struct list_head		run_list;
595 	unsigned long			timeout;
596 	unsigned long			watchdog_stamp;
597 	unsigned int			time_slice;
598 	unsigned short			on_rq;
599 	unsigned short			on_list;
600 
601 	struct sched_rt_entity		*back;
602 #ifdef CONFIG_RT_GROUP_SCHED
603 	struct sched_rt_entity		*parent;
604 	/* rq on which this entity is (to be) queued: */
605 	struct rt_rq			*rt_rq;
606 	/* rq "owned" by this entity/group: */
607 	struct rt_rq			*my_q;
608 #endif
609 } __randomize_layout;
610 
611 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
612 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
613 
614 struct sched_dl_entity {
615 	struct rb_node			rb_node;
616 
617 	/*
618 	 * Original scheduling parameters. Copied here from sched_attr
619 	 * during sched_setattr(), they will remain the same until
620 	 * the next sched_setattr().
621 	 */
622 	u64				dl_runtime;	/* Maximum runtime for each instance	*/
623 	u64				dl_deadline;	/* Relative deadline of each instance	*/
624 	u64				dl_period;	/* Separation of two instances (period) */
625 	u64				dl_bw;		/* dl_runtime / dl_period		*/
626 	u64				dl_density;	/* dl_runtime / dl_deadline		*/
627 
628 	/*
629 	 * Actual scheduling parameters. Initialized with the values above,
630 	 * they are continuously updated during task execution. Note that
631 	 * the remaining runtime could be < 0 in case we are in overrun.
632 	 */
633 	s64				runtime;	/* Remaining runtime for this instance	*/
634 	u64				deadline;	/* Absolute deadline for this instance	*/
635 	unsigned int			flags;		/* Specifying the scheduler behaviour	*/
636 
637 	/*
638 	 * Some bool flags:
639 	 *
640 	 * @dl_throttled tells if we exhausted the runtime. If so, the
641 	 * task has to wait for a replenishment to be performed at the
642 	 * next firing of dl_timer.
643 	 *
644 	 * @dl_yielded tells if task gave up the CPU before consuming
645 	 * all its available runtime during the last job.
646 	 *
647 	 * @dl_non_contending tells if the task is inactive while still
648 	 * contributing to the active utilization. In other words, it
649 	 * indicates if the inactive timer has been armed and its handler
650 	 * has not been executed yet. This flag is useful to avoid race
651 	 * conditions between the inactive timer handler and the wakeup
652 	 * code.
653 	 *
654 	 * @dl_overrun tells if the task asked to be informed about runtime
655 	 * overruns.
656 	 *
657 	 * @dl_server tells if this is a server entity.
658 	 *
659 	 * @dl_defer tells if this is a deferred or regular server. For
660 	 * now only defer server exists.
661 	 *
662 	 * @dl_defer_armed tells if the deferrable server is waiting
663 	 * for the replenishment timer to activate it.
664 	 *
665 	 * @dl_server_active tells if the dlserver is active(started).
666 	 * dlserver is started on first cfs enqueue on an idle runqueue
667 	 * and is stopped when a dequeue results in 0 cfs tasks on the
668 	 * runqueue. In other words, dlserver is active only when cpu's
669 	 * runqueue has atleast one cfs task.
670 	 *
671 	 * @dl_defer_running tells if the deferrable server is actually
672 	 * running, skipping the defer phase.
673 	 */
674 	unsigned int			dl_throttled      : 1;
675 	unsigned int			dl_yielded        : 1;
676 	unsigned int			dl_non_contending : 1;
677 	unsigned int			dl_overrun	  : 1;
678 	unsigned int			dl_server         : 1;
679 	unsigned int			dl_server_active  : 1;
680 	unsigned int			dl_defer	  : 1;
681 	unsigned int			dl_defer_armed	  : 1;
682 	unsigned int			dl_defer_running  : 1;
683 
684 	/*
685 	 * Bandwidth enforcement timer. Each -deadline task has its
686 	 * own bandwidth to be enforced, thus we need one timer per task.
687 	 */
688 	struct hrtimer			dl_timer;
689 
690 	/*
691 	 * Inactive timer, responsible for decreasing the active utilization
692 	 * at the "0-lag time". When a -deadline task blocks, it contributes
693 	 * to GRUB's active utilization until the "0-lag time", hence a
694 	 * timer is needed to decrease the active utilization at the correct
695 	 * time.
696 	 */
697 	struct hrtimer			inactive_timer;
698 
699 	/*
700 	 * Bits for DL-server functionality. Also see the comment near
701 	 * dl_server_update().
702 	 *
703 	 * @rq the runqueue this server is for
704 	 *
705 	 * @server_has_tasks() returns true if @server_pick return a
706 	 * runnable task.
707 	 */
708 	struct rq			*rq;
709 	dl_server_has_tasks_f		server_has_tasks;
710 	dl_server_pick_f		server_pick_task;
711 
712 #ifdef CONFIG_RT_MUTEXES
713 	/*
714 	 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
715 	 * pi_se points to the donor, otherwise points to the dl_se it belongs
716 	 * to (the original one/itself).
717 	 */
718 	struct sched_dl_entity *pi_se;
719 #endif
720 };
721 
722 #ifdef CONFIG_UCLAMP_TASK
723 /* Number of utilization clamp buckets (shorter alias) */
724 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
725 
726 /*
727  * Utilization clamp for a scheduling entity
728  * @value:		clamp value "assigned" to a se
729  * @bucket_id:		bucket index corresponding to the "assigned" value
730  * @active:		the se is currently refcounted in a rq's bucket
731  * @user_defined:	the requested clamp value comes from user-space
732  *
733  * The bucket_id is the index of the clamp bucket matching the clamp value
734  * which is pre-computed and stored to avoid expensive integer divisions from
735  * the fast path.
736  *
737  * The active bit is set whenever a task has got an "effective" value assigned,
738  * which can be different from the clamp value "requested" from user-space.
739  * This allows to know a task is refcounted in the rq's bucket corresponding
740  * to the "effective" bucket_id.
741  *
742  * The user_defined bit is set whenever a task has got a task-specific clamp
743  * value requested from userspace, i.e. the system defaults apply to this task
744  * just as a restriction. This allows to relax default clamps when a less
745  * restrictive task-specific value has been requested, thus allowing to
746  * implement a "nice" semantic. For example, a task running with a 20%
747  * default boost can still drop its own boosting to 0%.
748  */
749 struct uclamp_se {
750 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
751 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
752 	unsigned int active		: 1;
753 	unsigned int user_defined	: 1;
754 };
755 #endif /* CONFIG_UCLAMP_TASK */
756 
757 union rcu_special {
758 	struct {
759 		u8			blocked;
760 		u8			need_qs;
761 		u8			exp_hint; /* Hint for performance. */
762 		u8			need_mb; /* Readers need smp_mb(). */
763 	} b; /* Bits. */
764 	u32 s; /* Set of bits. */
765 };
766 
767 enum perf_event_task_context {
768 	perf_invalid_context = -1,
769 	perf_hw_context = 0,
770 	perf_sw_context,
771 	perf_nr_task_contexts,
772 };
773 
774 /*
775  * Number of contexts where an event can trigger:
776  *      task, softirq, hardirq, nmi.
777  */
778 #define PERF_NR_CONTEXTS	4
779 
780 struct wake_q_node {
781 	struct wake_q_node *next;
782 };
783 
784 struct kmap_ctrl {
785 #ifdef CONFIG_KMAP_LOCAL
786 	int				idx;
787 	pte_t				pteval[KM_MAX_IDX];
788 #endif
789 };
790 
791 struct task_struct {
792 #ifdef CONFIG_THREAD_INFO_IN_TASK
793 	/*
794 	 * For reasons of header soup (see current_thread_info()), this
795 	 * must be the first element of task_struct.
796 	 */
797 	struct thread_info		thread_info;
798 #endif
799 	unsigned int			__state;
800 
801 	/* saved state for "spinlock sleepers" */
802 	unsigned int			saved_state;
803 
804 	/*
805 	 * This begins the randomizable portion of task_struct. Only
806 	 * scheduling-critical items should be added above here.
807 	 */
808 	randomized_struct_fields_start
809 
810 	void				*stack;
811 	refcount_t			usage;
812 	/* Per task flags (PF_*), defined further below: */
813 	unsigned int			flags;
814 	unsigned int			ptrace;
815 
816 #ifdef CONFIG_MEM_ALLOC_PROFILING
817 	struct alloc_tag		*alloc_tag;
818 #endif
819 
820 #ifdef CONFIG_SMP
821 	int				on_cpu;
822 	struct __call_single_node	wake_entry;
823 	unsigned int			wakee_flips;
824 	unsigned long			wakee_flip_decay_ts;
825 	struct task_struct		*last_wakee;
826 
827 	/*
828 	 * recent_used_cpu is initially set as the last CPU used by a task
829 	 * that wakes affine another task. Waker/wakee relationships can
830 	 * push tasks around a CPU where each wakeup moves to the next one.
831 	 * Tracking a recently used CPU allows a quick search for a recently
832 	 * used CPU that may be idle.
833 	 */
834 	int				recent_used_cpu;
835 	int				wake_cpu;
836 #endif
837 	int				on_rq;
838 
839 	int				prio;
840 	int				static_prio;
841 	int				normal_prio;
842 	unsigned int			rt_priority;
843 
844 	struct sched_entity		se;
845 	struct sched_rt_entity		rt;
846 	struct sched_dl_entity		dl;
847 	struct sched_dl_entity		*dl_server;
848 #ifdef CONFIG_SCHED_CLASS_EXT
849 	struct sched_ext_entity		scx;
850 #endif
851 	const struct sched_class	*sched_class;
852 
853 #ifdef CONFIG_SCHED_CORE
854 	struct rb_node			core_node;
855 	unsigned long			core_cookie;
856 	unsigned int			core_occupation;
857 #endif
858 
859 #ifdef CONFIG_CGROUP_SCHED
860 	struct task_group		*sched_task_group;
861 #endif
862 
863 
864 #ifdef CONFIG_UCLAMP_TASK
865 	/*
866 	 * Clamp values requested for a scheduling entity.
867 	 * Must be updated with task_rq_lock() held.
868 	 */
869 	struct uclamp_se		uclamp_req[UCLAMP_CNT];
870 	/*
871 	 * Effective clamp values used for a scheduling entity.
872 	 * Must be updated with task_rq_lock() held.
873 	 */
874 	struct uclamp_se		uclamp[UCLAMP_CNT];
875 #endif
876 
877 	struct sched_statistics         stats;
878 
879 #ifdef CONFIG_PREEMPT_NOTIFIERS
880 	/* List of struct preempt_notifier: */
881 	struct hlist_head		preempt_notifiers;
882 #endif
883 
884 #ifdef CONFIG_BLK_DEV_IO_TRACE
885 	unsigned int			btrace_seq;
886 #endif
887 
888 	unsigned int			policy;
889 	unsigned long			max_allowed_capacity;
890 	int				nr_cpus_allowed;
891 	const cpumask_t			*cpus_ptr;
892 	cpumask_t			*user_cpus_ptr;
893 	cpumask_t			cpus_mask;
894 	void				*migration_pending;
895 #ifdef CONFIG_SMP
896 	unsigned short			migration_disabled;
897 #endif
898 	unsigned short			migration_flags;
899 
900 #ifdef CONFIG_PREEMPT_RCU
901 	int				rcu_read_lock_nesting;
902 	union rcu_special		rcu_read_unlock_special;
903 	struct list_head		rcu_node_entry;
904 	struct rcu_node			*rcu_blocked_node;
905 #endif /* #ifdef CONFIG_PREEMPT_RCU */
906 
907 #ifdef CONFIG_TASKS_RCU
908 	unsigned long			rcu_tasks_nvcsw;
909 	u8				rcu_tasks_holdout;
910 	u8				rcu_tasks_idx;
911 	int				rcu_tasks_idle_cpu;
912 	struct list_head		rcu_tasks_holdout_list;
913 	int				rcu_tasks_exit_cpu;
914 	struct list_head		rcu_tasks_exit_list;
915 #endif /* #ifdef CONFIG_TASKS_RCU */
916 
917 #ifdef CONFIG_TASKS_TRACE_RCU
918 	int				trc_reader_nesting;
919 	int				trc_ipi_to_cpu;
920 	union rcu_special		trc_reader_special;
921 	struct list_head		trc_holdout_list;
922 	struct list_head		trc_blkd_node;
923 	int				trc_blkd_cpu;
924 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
925 
926 	struct sched_info		sched_info;
927 
928 	struct list_head		tasks;
929 #ifdef CONFIG_SMP
930 	struct plist_node		pushable_tasks;
931 	struct rb_node			pushable_dl_tasks;
932 #endif
933 
934 	struct mm_struct		*mm;
935 	struct mm_struct		*active_mm;
936 	struct address_space		*faults_disabled_mapping;
937 
938 	int				exit_state;
939 	int				exit_code;
940 	int				exit_signal;
941 	/* The signal sent when the parent dies: */
942 	int				pdeath_signal;
943 	/* JOBCTL_*, siglock protected: */
944 	unsigned long			jobctl;
945 
946 	/* Used for emulating ABI behavior of previous Linux versions: */
947 	unsigned int			personality;
948 
949 	/* Scheduler bits, serialized by scheduler locks: */
950 	unsigned			sched_reset_on_fork:1;
951 	unsigned			sched_contributes_to_load:1;
952 	unsigned			sched_migrated:1;
953 	unsigned			sched_task_hot:1;
954 
955 	/* Force alignment to the next boundary: */
956 	unsigned			:0;
957 
958 	/* Unserialized, strictly 'current' */
959 
960 	/*
961 	 * This field must not be in the scheduler word above due to wakelist
962 	 * queueing no longer being serialized by p->on_cpu. However:
963 	 *
964 	 * p->XXX = X;			ttwu()
965 	 * schedule()			  if (p->on_rq && ..) // false
966 	 *   smp_mb__after_spinlock();	  if (smp_load_acquire(&p->on_cpu) && //true
967 	 *   deactivate_task()		      ttwu_queue_wakelist())
968 	 *     p->on_rq = 0;			p->sched_remote_wakeup = Y;
969 	 *
970 	 * guarantees all stores of 'current' are visible before
971 	 * ->sched_remote_wakeup gets used, so it can be in this word.
972 	 */
973 	unsigned			sched_remote_wakeup:1;
974 #ifdef CONFIG_RT_MUTEXES
975 	unsigned			sched_rt_mutex:1;
976 #endif
977 
978 	/* Bit to tell TOMOYO we're in execve(): */
979 	unsigned			in_execve:1;
980 	unsigned			in_iowait:1;
981 #ifndef TIF_RESTORE_SIGMASK
982 	unsigned			restore_sigmask:1;
983 #endif
984 #ifdef CONFIG_MEMCG_V1
985 	unsigned			in_user_fault:1;
986 #endif
987 #ifdef CONFIG_LRU_GEN
988 	/* whether the LRU algorithm may apply to this access */
989 	unsigned			in_lru_fault:1;
990 #endif
991 #ifdef CONFIG_COMPAT_BRK
992 	unsigned			brk_randomized:1;
993 #endif
994 #ifdef CONFIG_CGROUPS
995 	/* disallow userland-initiated cgroup migration */
996 	unsigned			no_cgroup_migration:1;
997 	/* task is frozen/stopped (used by the cgroup freezer) */
998 	unsigned			frozen:1;
999 #endif
1000 #ifdef CONFIG_BLK_CGROUP
1001 	unsigned			use_memdelay:1;
1002 #endif
1003 #ifdef CONFIG_PSI
1004 	/* Stalled due to lack of memory */
1005 	unsigned			in_memstall:1;
1006 #endif
1007 #ifdef CONFIG_PAGE_OWNER
1008 	/* Used by page_owner=on to detect recursion in page tracking. */
1009 	unsigned			in_page_owner:1;
1010 #endif
1011 #ifdef CONFIG_EVENTFD
1012 	/* Recursion prevention for eventfd_signal() */
1013 	unsigned			in_eventfd:1;
1014 #endif
1015 #ifdef CONFIG_ARCH_HAS_CPU_PASID
1016 	unsigned			pasid_activated:1;
1017 #endif
1018 #ifdef CONFIG_X86_BUS_LOCK_DETECT
1019 	unsigned			reported_split_lock:1;
1020 #endif
1021 #ifdef CONFIG_TASK_DELAY_ACCT
1022 	/* delay due to memory thrashing */
1023 	unsigned                        in_thrashing:1;
1024 #endif
1025 #ifdef CONFIG_PREEMPT_RT
1026 	struct netdev_xmit		net_xmit;
1027 #endif
1028 	unsigned long			atomic_flags; /* Flags requiring atomic access. */
1029 
1030 	struct restart_block		restart_block;
1031 
1032 	pid_t				pid;
1033 	pid_t				tgid;
1034 
1035 #ifdef CONFIG_STACKPROTECTOR
1036 	/* Canary value for the -fstack-protector GCC feature: */
1037 	unsigned long			stack_canary;
1038 #endif
1039 	/*
1040 	 * Pointers to the (original) parent process, youngest child, younger sibling,
1041 	 * older sibling, respectively.  (p->father can be replaced with
1042 	 * p->real_parent->pid)
1043 	 */
1044 
1045 	/* Real parent process: */
1046 	struct task_struct __rcu	*real_parent;
1047 
1048 	/* Recipient of SIGCHLD, wait4() reports: */
1049 	struct task_struct __rcu	*parent;
1050 
1051 	/*
1052 	 * Children/sibling form the list of natural children:
1053 	 */
1054 	struct list_head		children;
1055 	struct list_head		sibling;
1056 	struct task_struct		*group_leader;
1057 
1058 	/*
1059 	 * 'ptraced' is the list of tasks this task is using ptrace() on.
1060 	 *
1061 	 * This includes both natural children and PTRACE_ATTACH targets.
1062 	 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1063 	 */
1064 	struct list_head		ptraced;
1065 	struct list_head		ptrace_entry;
1066 
1067 	/* PID/PID hash table linkage. */
1068 	struct pid			*thread_pid;
1069 	struct hlist_node		pid_links[PIDTYPE_MAX];
1070 	struct list_head		thread_node;
1071 
1072 	struct completion		*vfork_done;
1073 
1074 	/* CLONE_CHILD_SETTID: */
1075 	int __user			*set_child_tid;
1076 
1077 	/* CLONE_CHILD_CLEARTID: */
1078 	int __user			*clear_child_tid;
1079 
1080 	/* PF_KTHREAD | PF_IO_WORKER */
1081 	void				*worker_private;
1082 
1083 	u64				utime;
1084 	u64				stime;
1085 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1086 	u64				utimescaled;
1087 	u64				stimescaled;
1088 #endif
1089 	u64				gtime;
1090 	struct prev_cputime		prev_cputime;
1091 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1092 	struct vtime			vtime;
1093 #endif
1094 
1095 #ifdef CONFIG_NO_HZ_FULL
1096 	atomic_t			tick_dep_mask;
1097 #endif
1098 	/* Context switch counts: */
1099 	unsigned long			nvcsw;
1100 	unsigned long			nivcsw;
1101 
1102 	/* Monotonic time in nsecs: */
1103 	u64				start_time;
1104 
1105 	/* Boot based time in nsecs: */
1106 	u64				start_boottime;
1107 
1108 	/* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1109 	unsigned long			min_flt;
1110 	unsigned long			maj_flt;
1111 
1112 	/* Empty if CONFIG_POSIX_CPUTIMERS=n */
1113 	struct posix_cputimers		posix_cputimers;
1114 
1115 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1116 	struct posix_cputimers_work	posix_cputimers_work;
1117 #endif
1118 
1119 	/* Process credentials: */
1120 
1121 	/* Tracer's credentials at attach: */
1122 	const struct cred __rcu		*ptracer_cred;
1123 
1124 	/* Objective and real subjective task credentials (COW): */
1125 	const struct cred __rcu		*real_cred;
1126 
1127 	/* Effective (overridable) subjective task credentials (COW): */
1128 	const struct cred __rcu		*cred;
1129 
1130 #ifdef CONFIG_KEYS
1131 	/* Cached requested key. */
1132 	struct key			*cached_requested_key;
1133 #endif
1134 
1135 	/*
1136 	 * executable name, excluding path.
1137 	 *
1138 	 * - normally initialized begin_new_exec()
1139 	 * - set it with set_task_comm()
1140 	 *   - strscpy_pad() to ensure it is always NUL-terminated and
1141 	 *     zero-padded
1142 	 *   - task_lock() to ensure the operation is atomic and the name is
1143 	 *     fully updated.
1144 	 */
1145 	char				comm[TASK_COMM_LEN];
1146 
1147 	struct nameidata		*nameidata;
1148 
1149 #ifdef CONFIG_SYSVIPC
1150 	struct sysv_sem			sysvsem;
1151 	struct sysv_shm			sysvshm;
1152 #endif
1153 #ifdef CONFIG_DETECT_HUNG_TASK
1154 	unsigned long			last_switch_count;
1155 	unsigned long			last_switch_time;
1156 #endif
1157 	/* Filesystem information: */
1158 	struct fs_struct		*fs;
1159 
1160 	/* Open file information: */
1161 	struct files_struct		*files;
1162 
1163 #ifdef CONFIG_IO_URING
1164 	struct io_uring_task		*io_uring;
1165 #endif
1166 
1167 	/* Namespaces: */
1168 	struct nsproxy			*nsproxy;
1169 
1170 	/* Signal handlers: */
1171 	struct signal_struct		*signal;
1172 	struct sighand_struct __rcu		*sighand;
1173 	sigset_t			blocked;
1174 	sigset_t			real_blocked;
1175 	/* Restored if set_restore_sigmask() was used: */
1176 	sigset_t			saved_sigmask;
1177 	struct sigpending		pending;
1178 	unsigned long			sas_ss_sp;
1179 	size_t				sas_ss_size;
1180 	unsigned int			sas_ss_flags;
1181 
1182 	struct callback_head		*task_works;
1183 
1184 #ifdef CONFIG_AUDIT
1185 #ifdef CONFIG_AUDITSYSCALL
1186 	struct audit_context		*audit_context;
1187 #endif
1188 	kuid_t				loginuid;
1189 	unsigned int			sessionid;
1190 #endif
1191 	struct seccomp			seccomp;
1192 	struct syscall_user_dispatch	syscall_dispatch;
1193 
1194 	/* Thread group tracking: */
1195 	u64				parent_exec_id;
1196 	u64				self_exec_id;
1197 
1198 	/* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1199 	spinlock_t			alloc_lock;
1200 
1201 	/* Protection of the PI data structures: */
1202 	raw_spinlock_t			pi_lock;
1203 
1204 	struct wake_q_node		wake_q;
1205 
1206 #ifdef CONFIG_RT_MUTEXES
1207 	/* PI waiters blocked on a rt_mutex held by this task: */
1208 	struct rb_root_cached		pi_waiters;
1209 	/* Updated under owner's pi_lock and rq lock */
1210 	struct task_struct		*pi_top_task;
1211 	/* Deadlock detection and priority inheritance handling: */
1212 	struct rt_mutex_waiter		*pi_blocked_on;
1213 #endif
1214 
1215 #ifdef CONFIG_DEBUG_MUTEXES
1216 	/* Mutex deadlock detection: */
1217 	struct mutex_waiter		*blocked_on;
1218 #endif
1219 
1220 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1221 	int				non_block_count;
1222 #endif
1223 
1224 #ifdef CONFIG_TRACE_IRQFLAGS
1225 	struct irqtrace_events		irqtrace;
1226 	unsigned int			hardirq_threaded;
1227 	u64				hardirq_chain_key;
1228 	int				softirqs_enabled;
1229 	int				softirq_context;
1230 	int				irq_config;
1231 #endif
1232 #ifdef CONFIG_PREEMPT_RT
1233 	int				softirq_disable_cnt;
1234 #endif
1235 
1236 #ifdef CONFIG_LOCKDEP
1237 # define MAX_LOCK_DEPTH			48UL
1238 	u64				curr_chain_key;
1239 	int				lockdep_depth;
1240 	unsigned int			lockdep_recursion;
1241 	struct held_lock		held_locks[MAX_LOCK_DEPTH];
1242 #endif
1243 
1244 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1245 	unsigned int			in_ubsan;
1246 #endif
1247 
1248 	/* Journalling filesystem info: */
1249 	void				*journal_info;
1250 
1251 	/* Stacked block device info: */
1252 	struct bio_list			*bio_list;
1253 
1254 	/* Stack plugging: */
1255 	struct blk_plug			*plug;
1256 
1257 	/* VM state: */
1258 	struct reclaim_state		*reclaim_state;
1259 
1260 	struct io_context		*io_context;
1261 
1262 #ifdef CONFIG_COMPACTION
1263 	struct capture_control		*capture_control;
1264 #endif
1265 	/* Ptrace state: */
1266 	unsigned long			ptrace_message;
1267 	kernel_siginfo_t		*last_siginfo;
1268 
1269 	struct task_io_accounting	ioac;
1270 #ifdef CONFIG_PSI
1271 	/* Pressure stall state */
1272 	unsigned int			psi_flags;
1273 #endif
1274 #ifdef CONFIG_TASK_XACCT
1275 	/* Accumulated RSS usage: */
1276 	u64				acct_rss_mem1;
1277 	/* Accumulated virtual memory usage: */
1278 	u64				acct_vm_mem1;
1279 	/* stime + utime since last update: */
1280 	u64				acct_timexpd;
1281 #endif
1282 #ifdef CONFIG_CPUSETS
1283 	/* Protected by ->alloc_lock: */
1284 	nodemask_t			mems_allowed;
1285 	/* Sequence number to catch updates: */
1286 	seqcount_spinlock_t		mems_allowed_seq;
1287 	int				cpuset_mem_spread_rotor;
1288 #endif
1289 #ifdef CONFIG_CGROUPS
1290 	/* Control Group info protected by css_set_lock: */
1291 	struct css_set __rcu		*cgroups;
1292 	/* cg_list protected by css_set_lock and tsk->alloc_lock: */
1293 	struct list_head		cg_list;
1294 #endif
1295 #ifdef CONFIG_X86_CPU_RESCTRL
1296 	u32				closid;
1297 	u32				rmid;
1298 #endif
1299 #ifdef CONFIG_FUTEX
1300 	struct robust_list_head __user	*robust_list;
1301 #ifdef CONFIG_COMPAT
1302 	struct compat_robust_list_head __user *compat_robust_list;
1303 #endif
1304 	struct list_head		pi_state_list;
1305 	struct futex_pi_state		*pi_state_cache;
1306 	struct mutex			futex_exit_mutex;
1307 	unsigned int			futex_state;
1308 #endif
1309 #ifdef CONFIG_PERF_EVENTS
1310 	u8				perf_recursion[PERF_NR_CONTEXTS];
1311 	struct perf_event_context	*perf_event_ctxp;
1312 	struct mutex			perf_event_mutex;
1313 	struct list_head		perf_event_list;
1314 #endif
1315 #ifdef CONFIG_DEBUG_PREEMPT
1316 	unsigned long			preempt_disable_ip;
1317 #endif
1318 #ifdef CONFIG_NUMA
1319 	/* Protected by alloc_lock: */
1320 	struct mempolicy		*mempolicy;
1321 	short				il_prev;
1322 	u8				il_weight;
1323 	short				pref_node_fork;
1324 #endif
1325 #ifdef CONFIG_NUMA_BALANCING
1326 	int				numa_scan_seq;
1327 	unsigned int			numa_scan_period;
1328 	unsigned int			numa_scan_period_max;
1329 	int				numa_preferred_nid;
1330 	unsigned long			numa_migrate_retry;
1331 	/* Migration stamp: */
1332 	u64				node_stamp;
1333 	u64				last_task_numa_placement;
1334 	u64				last_sum_exec_runtime;
1335 	struct callback_head		numa_work;
1336 
1337 	/*
1338 	 * This pointer is only modified for current in syscall and
1339 	 * pagefault context (and for tasks being destroyed), so it can be read
1340 	 * from any of the following contexts:
1341 	 *  - RCU read-side critical section
1342 	 *  - current->numa_group from everywhere
1343 	 *  - task's runqueue locked, task not running
1344 	 */
1345 	struct numa_group __rcu		*numa_group;
1346 
1347 	/*
1348 	 * numa_faults is an array split into four regions:
1349 	 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1350 	 * in this precise order.
1351 	 *
1352 	 * faults_memory: Exponential decaying average of faults on a per-node
1353 	 * basis. Scheduling placement decisions are made based on these
1354 	 * counts. The values remain static for the duration of a PTE scan.
1355 	 * faults_cpu: Track the nodes the process was running on when a NUMA
1356 	 * hinting fault was incurred.
1357 	 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1358 	 * during the current scan window. When the scan completes, the counts
1359 	 * in faults_memory and faults_cpu decay and these values are copied.
1360 	 */
1361 	unsigned long			*numa_faults;
1362 	unsigned long			total_numa_faults;
1363 
1364 	/*
1365 	 * numa_faults_locality tracks if faults recorded during the last
1366 	 * scan window were remote/local or failed to migrate. The task scan
1367 	 * period is adapted based on the locality of the faults with different
1368 	 * weights depending on whether they were shared or private faults
1369 	 */
1370 	unsigned long			numa_faults_locality[3];
1371 
1372 	unsigned long			numa_pages_migrated;
1373 #endif /* CONFIG_NUMA_BALANCING */
1374 
1375 #ifdef CONFIG_RSEQ
1376 	struct rseq __user *rseq;
1377 	u32 rseq_len;
1378 	u32 rseq_sig;
1379 	/*
1380 	 * RmW on rseq_event_mask must be performed atomically
1381 	 * with respect to preemption.
1382 	 */
1383 	unsigned long rseq_event_mask;
1384 # ifdef CONFIG_DEBUG_RSEQ
1385 	/*
1386 	 * This is a place holder to save a copy of the rseq fields for
1387 	 * validation of read-only fields. The struct rseq has a
1388 	 * variable-length array at the end, so it cannot be used
1389 	 * directly. Reserve a size large enough for the known fields.
1390 	 */
1391 	char				rseq_fields[sizeof(struct rseq)];
1392 # endif
1393 #endif
1394 
1395 #ifdef CONFIG_SCHED_MM_CID
1396 	int				mm_cid;		/* Current cid in mm */
1397 	int				last_mm_cid;	/* Most recent cid in mm */
1398 	int				migrate_from_cpu;
1399 	int				mm_cid_active;	/* Whether cid bitmap is active */
1400 	struct callback_head		cid_work;
1401 #endif
1402 
1403 	struct tlbflush_unmap_batch	tlb_ubc;
1404 
1405 	/* Cache last used pipe for splice(): */
1406 	struct pipe_inode_info		*splice_pipe;
1407 
1408 	struct page_frag		task_frag;
1409 
1410 #ifdef CONFIG_TASK_DELAY_ACCT
1411 	struct task_delay_info		*delays;
1412 #endif
1413 
1414 #ifdef CONFIG_FAULT_INJECTION
1415 	int				make_it_fail;
1416 	unsigned int			fail_nth;
1417 #endif
1418 	/*
1419 	 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1420 	 * balance_dirty_pages() for a dirty throttling pause:
1421 	 */
1422 	int				nr_dirtied;
1423 	int				nr_dirtied_pause;
1424 	/* Start of a write-and-pause period: */
1425 	unsigned long			dirty_paused_when;
1426 
1427 #ifdef CONFIG_LATENCYTOP
1428 	int				latency_record_count;
1429 	struct latency_record		latency_record[LT_SAVECOUNT];
1430 #endif
1431 	/*
1432 	 * Time slack values; these are used to round up poll() and
1433 	 * select() etc timeout values. These are in nanoseconds.
1434 	 */
1435 	u64				timer_slack_ns;
1436 	u64				default_timer_slack_ns;
1437 
1438 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1439 	unsigned int			kasan_depth;
1440 #endif
1441 
1442 #ifdef CONFIG_KCSAN
1443 	struct kcsan_ctx		kcsan_ctx;
1444 #ifdef CONFIG_TRACE_IRQFLAGS
1445 	struct irqtrace_events		kcsan_save_irqtrace;
1446 #endif
1447 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1448 	int				kcsan_stack_depth;
1449 #endif
1450 #endif
1451 
1452 #ifdef CONFIG_KMSAN
1453 	struct kmsan_ctx		kmsan_ctx;
1454 #endif
1455 
1456 #if IS_ENABLED(CONFIG_KUNIT)
1457 	struct kunit			*kunit_test;
1458 #endif
1459 
1460 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1461 	/* Index of current stored address in ret_stack: */
1462 	int				curr_ret_stack;
1463 	int				curr_ret_depth;
1464 
1465 	/* Stack of return addresses for return function tracing: */
1466 	unsigned long			*ret_stack;
1467 
1468 	/* Timestamp for last schedule: */
1469 	unsigned long long		ftrace_timestamp;
1470 	unsigned long long		ftrace_sleeptime;
1471 
1472 	/*
1473 	 * Number of functions that haven't been traced
1474 	 * because of depth overrun:
1475 	 */
1476 	atomic_t			trace_overrun;
1477 
1478 	/* Pause tracing: */
1479 	atomic_t			tracing_graph_pause;
1480 #endif
1481 
1482 #ifdef CONFIG_TRACING
1483 	/* Bitmask and counter of trace recursion: */
1484 	unsigned long			trace_recursion;
1485 #endif /* CONFIG_TRACING */
1486 
1487 #ifdef CONFIG_KCOV
1488 	/* See kernel/kcov.c for more details. */
1489 
1490 	/* Coverage collection mode enabled for this task (0 if disabled): */
1491 	unsigned int			kcov_mode;
1492 
1493 	/* Size of the kcov_area: */
1494 	unsigned int			kcov_size;
1495 
1496 	/* Buffer for coverage collection: */
1497 	void				*kcov_area;
1498 
1499 	/* KCOV descriptor wired with this task or NULL: */
1500 	struct kcov			*kcov;
1501 
1502 	/* KCOV common handle for remote coverage collection: */
1503 	u64				kcov_handle;
1504 
1505 	/* KCOV sequence number: */
1506 	int				kcov_sequence;
1507 
1508 	/* Collect coverage from softirq context: */
1509 	unsigned int			kcov_softirq;
1510 #endif
1511 
1512 #ifdef CONFIG_MEMCG_V1
1513 	struct mem_cgroup		*memcg_in_oom;
1514 #endif
1515 
1516 #ifdef CONFIG_MEMCG
1517 	/* Number of pages to reclaim on returning to userland: */
1518 	unsigned int			memcg_nr_pages_over_high;
1519 
1520 	/* Used by memcontrol for targeted memcg charge: */
1521 	struct mem_cgroup		*active_memcg;
1522 
1523 	/* Cache for current->cgroups->memcg->objcg lookups: */
1524 	struct obj_cgroup		*objcg;
1525 #endif
1526 
1527 #ifdef CONFIG_BLK_CGROUP
1528 	struct gendisk			*throttle_disk;
1529 #endif
1530 
1531 #ifdef CONFIG_UPROBES
1532 	struct uprobe_task		*utask;
1533 #endif
1534 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1535 	unsigned int			sequential_io;
1536 	unsigned int			sequential_io_avg;
1537 #endif
1538 	struct kmap_ctrl		kmap_ctrl;
1539 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1540 	unsigned long			task_state_change;
1541 # ifdef CONFIG_PREEMPT_RT
1542 	unsigned long			saved_state_change;
1543 # endif
1544 #endif
1545 	struct rcu_head			rcu;
1546 	refcount_t			rcu_users;
1547 	int				pagefault_disabled;
1548 #ifdef CONFIG_MMU
1549 	struct task_struct		*oom_reaper_list;
1550 	struct timer_list		oom_reaper_timer;
1551 #endif
1552 #ifdef CONFIG_VMAP_STACK
1553 	struct vm_struct		*stack_vm_area;
1554 #endif
1555 #ifdef CONFIG_THREAD_INFO_IN_TASK
1556 	/* A live task holds one reference: */
1557 	refcount_t			stack_refcount;
1558 #endif
1559 #ifdef CONFIG_LIVEPATCH
1560 	int patch_state;
1561 #endif
1562 #ifdef CONFIG_SECURITY
1563 	/* Used by LSM modules for access restriction: */
1564 	void				*security;
1565 #endif
1566 #ifdef CONFIG_BPF_SYSCALL
1567 	/* Used by BPF task local storage */
1568 	struct bpf_local_storage __rcu	*bpf_storage;
1569 	/* Used for BPF run context */
1570 	struct bpf_run_ctx		*bpf_ctx;
1571 #endif
1572 	/* Used by BPF for per-TASK xdp storage */
1573 	struct bpf_net_context		*bpf_net_context;
1574 
1575 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1576 	unsigned long			lowest_stack;
1577 	unsigned long			prev_lowest_stack;
1578 #endif
1579 
1580 #ifdef CONFIG_X86_MCE
1581 	void __user			*mce_vaddr;
1582 	__u64				mce_kflags;
1583 	u64				mce_addr;
1584 	__u64				mce_ripv : 1,
1585 					mce_whole_page : 1,
1586 					__mce_reserved : 62;
1587 	struct callback_head		mce_kill_me;
1588 	int				mce_count;
1589 #endif
1590 
1591 #ifdef CONFIG_KRETPROBES
1592 	struct llist_head               kretprobe_instances;
1593 #endif
1594 #ifdef CONFIG_RETHOOK
1595 	struct llist_head               rethooks;
1596 #endif
1597 
1598 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1599 	/*
1600 	 * If L1D flush is supported on mm context switch
1601 	 * then we use this callback head to queue kill work
1602 	 * to kill tasks that are not running on SMT disabled
1603 	 * cores
1604 	 */
1605 	struct callback_head		l1d_flush_kill;
1606 #endif
1607 
1608 #ifdef CONFIG_RV
1609 	/*
1610 	 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1611 	 * If we find justification for more monitors, we can think
1612 	 * about adding more or developing a dynamic method. So far,
1613 	 * none of these are justified.
1614 	 */
1615 	union rv_task_monitor		rv[RV_PER_TASK_MONITORS];
1616 #endif
1617 
1618 #ifdef CONFIG_USER_EVENTS
1619 	struct user_event_mm		*user_event_mm;
1620 #endif
1621 
1622 	/*
1623 	 * New fields for task_struct should be added above here, so that
1624 	 * they are included in the randomized portion of task_struct.
1625 	 */
1626 	randomized_struct_fields_end
1627 
1628 	/* CPU-specific state of this task: */
1629 	struct thread_struct		thread;
1630 
1631 	/*
1632 	 * WARNING: on x86, 'thread_struct' contains a variable-sized
1633 	 * structure.  It *MUST* be at the end of 'task_struct'.
1634 	 *
1635 	 * Do not put anything below here!
1636 	 */
1637 };
1638 
1639 #define TASK_REPORT_IDLE	(TASK_REPORT + 1)
1640 #define TASK_REPORT_MAX		(TASK_REPORT_IDLE << 1)
1641 
__task_state_index(unsigned int tsk_state,unsigned int tsk_exit_state)1642 static inline unsigned int __task_state_index(unsigned int tsk_state,
1643 					      unsigned int tsk_exit_state)
1644 {
1645 	unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1646 
1647 	BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1648 
1649 	if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1650 		state = TASK_REPORT_IDLE;
1651 
1652 	/*
1653 	 * We're lying here, but rather than expose a completely new task state
1654 	 * to userspace, we can make this appear as if the task has gone through
1655 	 * a regular rt_mutex_lock() call.
1656 	 * Report frozen tasks as uninterruptible.
1657 	 */
1658 	if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
1659 		state = TASK_UNINTERRUPTIBLE;
1660 
1661 	return fls(state);
1662 }
1663 
task_state_index(struct task_struct * tsk)1664 static inline unsigned int task_state_index(struct task_struct *tsk)
1665 {
1666 	return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1667 }
1668 
task_index_to_char(unsigned int state)1669 static inline char task_index_to_char(unsigned int state)
1670 {
1671 	static const char state_char[] = "RSDTtXZPI";
1672 
1673 	BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1674 
1675 	return state_char[state];
1676 }
1677 
task_state_to_char(struct task_struct * tsk)1678 static inline char task_state_to_char(struct task_struct *tsk)
1679 {
1680 	return task_index_to_char(task_state_index(tsk));
1681 }
1682 
1683 extern struct pid *cad_pid;
1684 
1685 /*
1686  * Per process flags
1687  */
1688 #define PF_VCPU			0x00000001	/* I'm a virtual CPU */
1689 #define PF_IDLE			0x00000002	/* I am an IDLE thread */
1690 #define PF_EXITING		0x00000004	/* Getting shut down */
1691 #define PF_POSTCOREDUMP		0x00000008	/* Coredumps should ignore this task */
1692 #define PF_IO_WORKER		0x00000010	/* Task is an IO worker */
1693 #define PF_WQ_WORKER		0x00000020	/* I'm a workqueue worker */
1694 #define PF_FORKNOEXEC		0x00000040	/* Forked but didn't exec */
1695 #define PF_MCE_PROCESS		0x00000080      /* Process policy on mce errors */
1696 #define PF_SUPERPRIV		0x00000100	/* Used super-user privileges */
1697 #define PF_DUMPCORE		0x00000200	/* Dumped core */
1698 #define PF_SIGNALED		0x00000400	/* Killed by a signal */
1699 #define PF_MEMALLOC		0x00000800	/* Allocating memory to free memory. See memalloc_noreclaim_save() */
1700 #define PF_NPROC_EXCEEDED	0x00001000	/* set_user() noticed that RLIMIT_NPROC was exceeded */
1701 #define PF_USED_MATH		0x00002000	/* If unset the fpu must be initialized before use */
1702 #define PF_USER_WORKER		0x00004000	/* Kernel thread cloned from userspace thread */
1703 #define PF_NOFREEZE		0x00008000	/* This thread should not be frozen */
1704 #define PF_KCOMPACTD		0x00010000	/* I am kcompactd */
1705 #define PF_KSWAPD		0x00020000	/* I am kswapd */
1706 #define PF_MEMALLOC_NOFS	0x00040000	/* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1707 #define PF_MEMALLOC_NOIO	0x00080000	/* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1708 #define PF_LOCAL_THROTTLE	0x00100000	/* Throttle writes only against the bdi I write to,
1709 						 * I am cleaning dirty pages from some other bdi. */
1710 #define PF_KTHREAD		0x00200000	/* I am a kernel thread */
1711 #define PF_RANDOMIZE		0x00400000	/* Randomize virtual address space */
1712 #define PF__HOLE__00800000	0x00800000
1713 #define PF__HOLE__01000000	0x01000000
1714 #define PF__HOLE__02000000	0x02000000
1715 #define PF_NO_SETAFFINITY	0x04000000	/* Userland is not allowed to meddle with cpus_mask */
1716 #define PF_MCE_EARLY		0x08000000      /* Early kill for mce process policy */
1717 #define PF_MEMALLOC_PIN		0x10000000	/* Allocations constrained to zones which allow long term pinning.
1718 						 * See memalloc_pin_save() */
1719 #define PF_BLOCK_TS		0x20000000	/* plug has ts that needs updating */
1720 #define PF__HOLE__40000000	0x40000000
1721 #define PF_SUSPEND_TASK		0x80000000      /* This thread called freeze_processes() and should not be frozen */
1722 
1723 /*
1724  * Only the _current_ task can read/write to tsk->flags, but other
1725  * tasks can access tsk->flags in readonly mode for example
1726  * with tsk_used_math (like during threaded core dumping).
1727  * There is however an exception to this rule during ptrace
1728  * or during fork: the ptracer task is allowed to write to the
1729  * child->flags of its traced child (same goes for fork, the parent
1730  * can write to the child->flags), because we're guaranteed the
1731  * child is not running and in turn not changing child->flags
1732  * at the same time the parent does it.
1733  */
1734 #define clear_stopped_child_used_math(child)	do { (child)->flags &= ~PF_USED_MATH; } while (0)
1735 #define set_stopped_child_used_math(child)	do { (child)->flags |= PF_USED_MATH; } while (0)
1736 #define clear_used_math()			clear_stopped_child_used_math(current)
1737 #define set_used_math()				set_stopped_child_used_math(current)
1738 
1739 #define conditional_stopped_child_used_math(condition, child) \
1740 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1741 
1742 #define conditional_used_math(condition)	conditional_stopped_child_used_math(condition, current)
1743 
1744 #define copy_to_stopped_child_used_math(child) \
1745 	do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1746 
1747 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1748 #define tsk_used_math(p)			((p)->flags & PF_USED_MATH)
1749 #define used_math()				tsk_used_math(current)
1750 
is_percpu_thread(void)1751 static __always_inline bool is_percpu_thread(void)
1752 {
1753 #ifdef CONFIG_SMP
1754 	return (current->flags & PF_NO_SETAFFINITY) &&
1755 		(current->nr_cpus_allowed  == 1);
1756 #else
1757 	return true;
1758 #endif
1759 }
1760 
1761 /* Per-process atomic flags. */
1762 #define PFA_NO_NEW_PRIVS		0	/* May not gain new privileges. */
1763 #define PFA_SPREAD_PAGE			1	/* Spread page cache over cpuset */
1764 #define PFA_SPREAD_SLAB			2	/* Spread some slab caches over cpuset */
1765 #define PFA_SPEC_SSB_DISABLE		3	/* Speculative Store Bypass disabled */
1766 #define PFA_SPEC_SSB_FORCE_DISABLE	4	/* Speculative Store Bypass force disabled*/
1767 #define PFA_SPEC_IB_DISABLE		5	/* Indirect branch speculation restricted */
1768 #define PFA_SPEC_IB_FORCE_DISABLE	6	/* Indirect branch speculation permanently restricted */
1769 #define PFA_SPEC_SSB_NOEXEC		7	/* Speculative Store Bypass clear on execve() */
1770 
1771 #define TASK_PFA_TEST(name, func)					\
1772 	static inline bool task_##func(struct task_struct *p)		\
1773 	{ return test_bit(PFA_##name, &p->atomic_flags); }
1774 
1775 #define TASK_PFA_SET(name, func)					\
1776 	static inline void task_set_##func(struct task_struct *p)	\
1777 	{ set_bit(PFA_##name, &p->atomic_flags); }
1778 
1779 #define TASK_PFA_CLEAR(name, func)					\
1780 	static inline void task_clear_##func(struct task_struct *p)	\
1781 	{ clear_bit(PFA_##name, &p->atomic_flags); }
1782 
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1783 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1784 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1785 
1786 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1787 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1788 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1789 
1790 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1791 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1792 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1793 
1794 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1795 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1796 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1797 
1798 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1799 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1800 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1801 
1802 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1803 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1804 
1805 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1806 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1807 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1808 
1809 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1810 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1811 
1812 static inline void
1813 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1814 {
1815 	current->flags &= ~flags;
1816 	current->flags |= orig_flags & flags;
1817 }
1818 
1819 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1820 extern int task_can_attach(struct task_struct *p);
1821 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1822 extern void dl_bw_free(int cpu, u64 dl_bw);
1823 #ifdef CONFIG_SMP
1824 
1825 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1826 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1827 
1828 /**
1829  * set_cpus_allowed_ptr - set CPU affinity mask of a task
1830  * @p: the task
1831  * @new_mask: CPU affinity mask
1832  *
1833  * Return: zero if successful, or a negative error code
1834  */
1835 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1836 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1837 extern void release_user_cpus_ptr(struct task_struct *p);
1838 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1839 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1840 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1841 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1842 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1843 {
1844 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1845 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1846 {
1847 	/* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1848 	if ((*cpumask_bits(new_mask) & 1) == 0)
1849 		return -EINVAL;
1850 	return 0;
1851 }
dup_user_cpus_ptr(struct task_struct * dst,struct task_struct * src,int node)1852 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1853 {
1854 	if (src->user_cpus_ptr)
1855 		return -EINVAL;
1856 	return 0;
1857 }
release_user_cpus_ptr(struct task_struct * p)1858 static inline void release_user_cpus_ptr(struct task_struct *p)
1859 {
1860 	WARN_ON(p->user_cpus_ptr);
1861 }
1862 
dl_task_check_affinity(struct task_struct * p,const struct cpumask * mask)1863 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1864 {
1865 	return 0;
1866 }
1867 #endif
1868 
1869 extern int yield_to(struct task_struct *p, bool preempt);
1870 extern void set_user_nice(struct task_struct *p, long nice);
1871 extern int task_prio(const struct task_struct *p);
1872 
1873 /**
1874  * task_nice - return the nice value of a given task.
1875  * @p: the task in question.
1876  *
1877  * Return: The nice value [ -20 ... 0 ... 19 ].
1878  */
task_nice(const struct task_struct * p)1879 static inline int task_nice(const struct task_struct *p)
1880 {
1881 	return PRIO_TO_NICE((p)->static_prio);
1882 }
1883 
1884 extern int can_nice(const struct task_struct *p, const int nice);
1885 extern int task_curr(const struct task_struct *p);
1886 extern int idle_cpu(int cpu);
1887 extern int available_idle_cpu(int cpu);
1888 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1889 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1890 extern void sched_set_fifo(struct task_struct *p);
1891 extern void sched_set_fifo_low(struct task_struct *p);
1892 extern void sched_set_normal(struct task_struct *p, int nice);
1893 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1894 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1895 extern struct task_struct *idle_task(int cpu);
1896 
1897 /**
1898  * is_idle_task - is the specified task an idle task?
1899  * @p: the task in question.
1900  *
1901  * Return: 1 if @p is an idle task. 0 otherwise.
1902  */
is_idle_task(const struct task_struct * p)1903 static __always_inline bool is_idle_task(const struct task_struct *p)
1904 {
1905 	return !!(p->flags & PF_IDLE);
1906 }
1907 
1908 extern struct task_struct *curr_task(int cpu);
1909 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1910 
1911 void yield(void);
1912 
1913 union thread_union {
1914 	struct task_struct task;
1915 #ifndef CONFIG_THREAD_INFO_IN_TASK
1916 	struct thread_info thread_info;
1917 #endif
1918 	unsigned long stack[THREAD_SIZE/sizeof(long)];
1919 };
1920 
1921 #ifndef CONFIG_THREAD_INFO_IN_TASK
1922 extern struct thread_info init_thread_info;
1923 #endif
1924 
1925 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1926 
1927 #ifdef CONFIG_THREAD_INFO_IN_TASK
1928 # define task_thread_info(task)	(&(task)->thread_info)
1929 #else
1930 # define task_thread_info(task)	((struct thread_info *)(task)->stack)
1931 #endif
1932 
1933 /*
1934  * find a task by one of its numerical ids
1935  *
1936  * find_task_by_pid_ns():
1937  *      finds a task by its pid in the specified namespace
1938  * find_task_by_vpid():
1939  *      finds a task by its virtual pid
1940  *
1941  * see also find_vpid() etc in include/linux/pid.h
1942  */
1943 
1944 extern struct task_struct *find_task_by_vpid(pid_t nr);
1945 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1946 
1947 /*
1948  * find a task by its virtual pid and get the task struct
1949  */
1950 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1951 
1952 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1953 extern int wake_up_process(struct task_struct *tsk);
1954 extern void wake_up_new_task(struct task_struct *tsk);
1955 
1956 #ifdef CONFIG_SMP
1957 extern void kick_process(struct task_struct *tsk);
1958 #else
kick_process(struct task_struct * tsk)1959 static inline void kick_process(struct task_struct *tsk) { }
1960 #endif
1961 
1962 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1963 #define set_task_comm(tsk, from) ({			\
1964 	BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN);	\
1965 	__set_task_comm(tsk, from, false);		\
1966 })
1967 
1968 /*
1969  * - Why not use task_lock()?
1970  *   User space can randomly change their names anyway, so locking for readers
1971  *   doesn't make sense. For writers, locking is probably necessary, as a race
1972  *   condition could lead to long-term mixed results.
1973  *   The strscpy_pad() in __set_task_comm() can ensure that the task comm is
1974  *   always NUL-terminated and zero-padded. Therefore the race condition between
1975  *   reader and writer is not an issue.
1976  *
1977  * - BUILD_BUG_ON() can help prevent the buf from being truncated.
1978  *   Since the callers don't perform any return value checks, this safeguard is
1979  *   necessary.
1980  */
1981 #define get_task_comm(buf, tsk) ({			\
1982 	BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN);	\
1983 	strscpy_pad(buf, (tsk)->comm);			\
1984 	buf;						\
1985 })
1986 
1987 #ifdef CONFIG_SMP
scheduler_ipi(void)1988 static __always_inline void scheduler_ipi(void)
1989 {
1990 	/*
1991 	 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1992 	 * TIF_NEED_RESCHED remotely (for the first time) will also send
1993 	 * this IPI.
1994 	 */
1995 	preempt_fold_need_resched();
1996 }
1997 #else
scheduler_ipi(void)1998 static inline void scheduler_ipi(void) { }
1999 #endif
2000 
2001 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2002 
2003 /*
2004  * Set thread flags in other task's structures.
2005  * See asm/thread_info.h for TIF_xxxx flags available:
2006  */
set_tsk_thread_flag(struct task_struct * tsk,int flag)2007 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2008 {
2009 	set_ti_thread_flag(task_thread_info(tsk), flag);
2010 }
2011 
clear_tsk_thread_flag(struct task_struct * tsk,int flag)2012 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2013 {
2014 	clear_ti_thread_flag(task_thread_info(tsk), flag);
2015 }
2016 
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)2017 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2018 					  bool value)
2019 {
2020 	update_ti_thread_flag(task_thread_info(tsk), flag, value);
2021 }
2022 
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)2023 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2024 {
2025 	return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2026 }
2027 
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)2028 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2029 {
2030 	return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2031 }
2032 
test_tsk_thread_flag(struct task_struct * tsk,int flag)2033 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2034 {
2035 	return test_ti_thread_flag(task_thread_info(tsk), flag);
2036 }
2037 
set_tsk_need_resched(struct task_struct * tsk)2038 static inline void set_tsk_need_resched(struct task_struct *tsk)
2039 {
2040 	set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2041 }
2042 
clear_tsk_need_resched(struct task_struct * tsk)2043 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2044 {
2045 	atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2046 			   (atomic_long_t *)&task_thread_info(tsk)->flags);
2047 }
2048 
test_tsk_need_resched(struct task_struct * tsk)2049 static inline int test_tsk_need_resched(struct task_struct *tsk)
2050 {
2051 	return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2052 }
2053 
2054 /*
2055  * cond_resched() and cond_resched_lock(): latency reduction via
2056  * explicit rescheduling in places that are safe. The return
2057  * value indicates whether a reschedule was done in fact.
2058  * cond_resched_lock() will drop the spinlock before scheduling,
2059  */
2060 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2061 extern int __cond_resched(void);
2062 
2063 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2064 
2065 void sched_dynamic_klp_enable(void);
2066 void sched_dynamic_klp_disable(void);
2067 
2068 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2069 
_cond_resched(void)2070 static __always_inline int _cond_resched(void)
2071 {
2072 	return static_call_mod(cond_resched)();
2073 }
2074 
2075 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2076 
2077 extern int dynamic_cond_resched(void);
2078 
_cond_resched(void)2079 static __always_inline int _cond_resched(void)
2080 {
2081 	return dynamic_cond_resched();
2082 }
2083 
2084 #else /* !CONFIG_PREEMPTION */
2085 
_cond_resched(void)2086 static inline int _cond_resched(void)
2087 {
2088 	klp_sched_try_switch();
2089 	return __cond_resched();
2090 }
2091 
2092 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2093 
2094 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2095 
_cond_resched(void)2096 static inline int _cond_resched(void)
2097 {
2098 	klp_sched_try_switch();
2099 	return 0;
2100 }
2101 
2102 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2103 
2104 #define cond_resched() ({			\
2105 	__might_resched(__FILE__, __LINE__, 0);	\
2106 	_cond_resched();			\
2107 })
2108 
2109 extern int __cond_resched_lock(spinlock_t *lock);
2110 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2111 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2112 
2113 #define MIGHT_RESCHED_RCU_SHIFT		8
2114 #define MIGHT_RESCHED_PREEMPT_MASK	((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2115 
2116 #ifndef CONFIG_PREEMPT_RT
2117 /*
2118  * Non RT kernels have an elevated preempt count due to the held lock,
2119  * but are not allowed to be inside a RCU read side critical section
2120  */
2121 # define PREEMPT_LOCK_RESCHED_OFFSETS	PREEMPT_LOCK_OFFSET
2122 #else
2123 /*
2124  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2125  * cond_resched*lock() has to take that into account because it checks for
2126  * preempt_count() and rcu_preempt_depth().
2127  */
2128 # define PREEMPT_LOCK_RESCHED_OFFSETS	\
2129 	(PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2130 #endif
2131 
2132 #define cond_resched_lock(lock) ({						\
2133 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2134 	__cond_resched_lock(lock);						\
2135 })
2136 
2137 #define cond_resched_rwlock_read(lock) ({					\
2138 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2139 	__cond_resched_rwlock_read(lock);					\
2140 })
2141 
2142 #define cond_resched_rwlock_write(lock) ({					\
2143 	__might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);	\
2144 	__cond_resched_rwlock_write(lock);					\
2145 })
2146 
need_resched(void)2147 static __always_inline bool need_resched(void)
2148 {
2149 	return unlikely(tif_need_resched());
2150 }
2151 
2152 /*
2153  * Wrappers for p->thread_info->cpu access. No-op on UP.
2154  */
2155 #ifdef CONFIG_SMP
2156 
task_cpu(const struct task_struct * p)2157 static inline unsigned int task_cpu(const struct task_struct *p)
2158 {
2159 	return READ_ONCE(task_thread_info(p)->cpu);
2160 }
2161 
2162 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2163 
2164 #else
2165 
task_cpu(const struct task_struct * p)2166 static inline unsigned int task_cpu(const struct task_struct *p)
2167 {
2168 	return 0;
2169 }
2170 
set_task_cpu(struct task_struct * p,unsigned int cpu)2171 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2172 {
2173 }
2174 
2175 #endif /* CONFIG_SMP */
2176 
task_is_runnable(struct task_struct * p)2177 static inline bool task_is_runnable(struct task_struct *p)
2178 {
2179 	return p->on_rq && !p->se.sched_delayed;
2180 }
2181 
2182 extern bool sched_task_on_rq(struct task_struct *p);
2183 extern unsigned long get_wchan(struct task_struct *p);
2184 extern struct task_struct *cpu_curr_snapshot(int cpu);
2185 
2186 #include <linux/spinlock.h>
2187 
2188 /*
2189  * In order to reduce various lock holder preemption latencies provide an
2190  * interface to see if a vCPU is currently running or not.
2191  *
2192  * This allows us to terminate optimistic spin loops and block, analogous to
2193  * the native optimistic spin heuristic of testing if the lock owner task is
2194  * running or not.
2195  */
2196 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2197 static inline bool vcpu_is_preempted(int cpu)
2198 {
2199 	return false;
2200 }
2201 #endif
2202 
2203 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2204 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2205 
2206 #ifndef TASK_SIZE_OF
2207 #define TASK_SIZE_OF(tsk)	TASK_SIZE
2208 #endif
2209 
2210 #ifdef CONFIG_SMP
owner_on_cpu(struct task_struct * owner)2211 static inline bool owner_on_cpu(struct task_struct *owner)
2212 {
2213 	/*
2214 	 * As lock holder preemption issue, we both skip spinning if
2215 	 * task is not on cpu or its cpu is preempted
2216 	 */
2217 	return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2218 }
2219 
2220 /* Returns effective CPU energy utilization, as seen by the scheduler */
2221 unsigned long sched_cpu_util(int cpu);
2222 #endif /* CONFIG_SMP */
2223 
2224 #ifdef CONFIG_SCHED_CORE
2225 extern void sched_core_free(struct task_struct *tsk);
2226 extern void sched_core_fork(struct task_struct *p);
2227 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2228 				unsigned long uaddr);
2229 extern int sched_core_idle_cpu(int cpu);
2230 #else
sched_core_free(struct task_struct * tsk)2231 static inline void sched_core_free(struct task_struct *tsk) { }
sched_core_fork(struct task_struct * p)2232 static inline void sched_core_fork(struct task_struct *p) { }
sched_core_idle_cpu(int cpu)2233 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2234 #endif
2235 
2236 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2237 
2238 #ifdef CONFIG_MEM_ALLOC_PROFILING
alloc_tag_save(struct alloc_tag * tag)2239 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2240 {
2241 	swap(current->alloc_tag, tag);
2242 	return tag;
2243 }
2244 
alloc_tag_restore(struct alloc_tag * tag,struct alloc_tag * old)2245 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2246 {
2247 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2248 	WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2249 #endif
2250 	current->alloc_tag = old;
2251 }
2252 #else
2253 #define alloc_tag_save(_tag)			NULL
2254 #define alloc_tag_restore(_tag, _old)		do {} while (0)
2255 #endif
2256 
2257 #endif
2258