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/uidgid_types.h>
48#include <linux/tracepoint-defs.h>
49#include <asm/kmap_size.h>
50
51/* task_struct member predeclarations (sorted alphabetically): */
52struct audit_context;
53struct bio_list;
54struct blk_plug;
55struct bpf_local_storage;
56struct bpf_run_ctx;
57struct bpf_net_context;
58struct capture_control;
59struct cfs_rq;
60struct fs_struct;
61struct futex_pi_state;
62struct io_context;
63struct io_uring_task;
64struct mempolicy;
65struct nameidata;
66struct nsproxy;
67struct perf_event_context;
68struct perf_ctx_data;
69struct pid_namespace;
70struct pipe_inode_info;
71struct rcu_node;
72struct reclaim_state;
73struct robust_list_head;
74struct root_domain;
75struct rq;
76struct sched_attr;
77struct sched_dl_entity;
78struct seq_file;
79struct sighand_struct;
80struct signal_struct;
81struct task_delay_info;
82struct task_group;
83struct task_struct;
84struct user_event_mm;
85
86#include <linux/sched/ext.h>
87
88/*
89 * Task state bitmask. NOTE! These bits are also
90 * encoded in fs/proc/array.c: get_task_state().
91 *
92 * We have two separate sets of flags: task->__state
93 * is about runnability, while task->exit_state are
94 * about the task exiting. Confusing, but this way
95 * modifying one set can't modify the other one by
96 * mistake.
97 */
98
99/* Used in tsk->__state: */
100#define TASK_RUNNING 0x00000000
101#define TASK_INTERRUPTIBLE 0x00000001
102#define TASK_UNINTERRUPTIBLE 0x00000002
103#define __TASK_STOPPED 0x00000004
104#define __TASK_TRACED 0x00000008
105/* Used in tsk->exit_state: */
106#define EXIT_DEAD 0x00000010
107#define EXIT_ZOMBIE 0x00000020
108#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
109/* Used in tsk->__state again: */
110#define TASK_PARKED 0x00000040
111#define TASK_DEAD 0x00000080
112#define TASK_WAKEKILL 0x00000100
113#define TASK_WAKING 0x00000200
114#define TASK_NOLOAD 0x00000400
115#define TASK_NEW 0x00000800
116#define TASK_RTLOCK_WAIT 0x00001000
117#define TASK_FREEZABLE 0x00002000
118#define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
119#define TASK_FROZEN 0x00008000
120#define TASK_STATE_MAX 0x00010000
121
122#define TASK_ANY (TASK_STATE_MAX-1)
123
124/*
125 * DO NOT ADD ANY NEW USERS !
126 */
127#define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
128
129/* Convenience macros for the sake of set_current_state: */
130#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
131#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
132#define TASK_TRACED __TASK_TRACED
133
134#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
135
136/* Convenience macros for the sake of wake_up(): */
137#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
138
139/* get_task_state(): */
140#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
141 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
142 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
143 TASK_PARKED)
144
145#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
146
147#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
148#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
149#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
150
151/*
152 * Special states are those that do not use the normal wait-loop pattern. See
153 * the comment with set_special_state().
154 */
155#define is_special_task_state(state) \
156 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \
157 TASK_DEAD | TASK_FROZEN))
158
159#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
160# define debug_normal_state_change(state_value) \
161 do { \
162 WARN_ON_ONCE(is_special_task_state(state_value)); \
163 current->task_state_change = _THIS_IP_; \
164 } while (0)
165
166# define debug_special_state_change(state_value) \
167 do { \
168 WARN_ON_ONCE(!is_special_task_state(state_value)); \
169 current->task_state_change = _THIS_IP_; \
170 } while (0)
171
172# define debug_rtlock_wait_set_state() \
173 do { \
174 current->saved_state_change = current->task_state_change;\
175 current->task_state_change = _THIS_IP_; \
176 } while (0)
177
178# define debug_rtlock_wait_restore_state() \
179 do { \
180 current->task_state_change = current->saved_state_change;\
181 } while (0)
182
183#else
184# define debug_normal_state_change(cond) do { } while (0)
185# define debug_special_state_change(cond) do { } while (0)
186# define debug_rtlock_wait_set_state() do { } while (0)
187# define debug_rtlock_wait_restore_state() do { } while (0)
188#endif
189
190#define trace_set_current_state(state_value) \
191 do { \
192 if (tracepoint_enabled(sched_set_state_tp)) \
193 __trace_set_current_state(state_value); \
194 } while (0)
195
196/*
197 * set_current_state() includes a barrier so that the write of current->__state
198 * is correctly serialised wrt the caller's subsequent test of whether to
199 * actually sleep:
200 *
201 * for (;;) {
202 * set_current_state(TASK_UNINTERRUPTIBLE);
203 * if (CONDITION)
204 * break;
205 *
206 * schedule();
207 * }
208 * __set_current_state(TASK_RUNNING);
209 *
210 * If the caller does not need such serialisation (because, for instance, the
211 * CONDITION test and condition change and wakeup are under the same lock) then
212 * use __set_current_state().
213 *
214 * The above is typically ordered against the wakeup, which does:
215 *
216 * CONDITION = 1;
217 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
218 *
219 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
220 * accessing p->__state.
221 *
222 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
223 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
224 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
225 *
226 * However, with slightly different timing the wakeup TASK_RUNNING store can
227 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
228 * a problem either because that will result in one extra go around the loop
229 * and our @cond test will save the day.
230 *
231 * Also see the comments of try_to_wake_up().
232 */
233#define __set_current_state(state_value) \
234 do { \
235 debug_normal_state_change((state_value)); \
236 trace_set_current_state(state_value); \
237 WRITE_ONCE(current->__state, (state_value)); \
238 } while (0)
239
240#define set_current_state(state_value) \
241 do { \
242 debug_normal_state_change((state_value)); \
243 trace_set_current_state(state_value); \
244 smp_store_mb(current->__state, (state_value)); \
245 } while (0)
246
247/*
248 * set_special_state() should be used for those states when the blocking task
249 * can not use the regular condition based wait-loop. In that case we must
250 * serialize against wakeups such that any possible in-flight TASK_RUNNING
251 * stores will not collide with our state change.
252 */
253#define set_special_state(state_value) \
254 do { \
255 unsigned long flags; /* may shadow */ \
256 \
257 raw_spin_lock_irqsave(&current->pi_lock, flags); \
258 debug_special_state_change((state_value)); \
259 trace_set_current_state(state_value); \
260 WRITE_ONCE(current->__state, (state_value)); \
261 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
262 } while (0)
263
264/*
265 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
266 *
267 * RT's spin/rwlock substitutions are state preserving. The state of the
268 * task when blocking on the lock is saved in task_struct::saved_state and
269 * restored after the lock has been acquired. These operations are
270 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
271 * lock related wakeups while the task is blocked on the lock are
272 * redirected to operate on task_struct::saved_state to ensure that these
273 * are not dropped. On restore task_struct::saved_state is set to
274 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
275 *
276 * The lock operation looks like this:
277 *
278 * current_save_and_set_rtlock_wait_state();
279 * for (;;) {
280 * if (try_lock())
281 * break;
282 * raw_spin_unlock_irq(&lock->wait_lock);
283 * schedule_rtlock();
284 * raw_spin_lock_irq(&lock->wait_lock);
285 * set_current_state(TASK_RTLOCK_WAIT);
286 * }
287 * current_restore_rtlock_saved_state();
288 */
289#define current_save_and_set_rtlock_wait_state() \
290 do { \
291 lockdep_assert_irqs_disabled(); \
292 raw_spin_lock(&current->pi_lock); \
293 current->saved_state = current->__state; \
294 debug_rtlock_wait_set_state(); \
295 trace_set_current_state(TASK_RTLOCK_WAIT); \
296 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
297 raw_spin_unlock(&current->pi_lock); \
298 } while (0);
299
300#define current_restore_rtlock_saved_state() \
301 do { \
302 lockdep_assert_irqs_disabled(); \
303 raw_spin_lock(&current->pi_lock); \
304 debug_rtlock_wait_restore_state(); \
305 trace_set_current_state(current->saved_state); \
306 WRITE_ONCE(current->__state, current->saved_state); \
307 current->saved_state = TASK_RUNNING; \
308 raw_spin_unlock(&current->pi_lock); \
309 } while (0);
310
311#define get_current_state() READ_ONCE(current->__state)
312
313/*
314 * Define the task command name length as enum, then it can be visible to
315 * BPF programs.
316 */
317enum {
318 TASK_COMM_LEN = 16,
319};
320
321extern void sched_tick(void);
322
323#define MAX_SCHEDULE_TIMEOUT LONG_MAX
324
325extern long schedule_timeout(long timeout);
326extern long schedule_timeout_interruptible(long timeout);
327extern long schedule_timeout_killable(long timeout);
328extern long schedule_timeout_uninterruptible(long timeout);
329extern long schedule_timeout_idle(long timeout);
330asmlinkage void schedule(void);
331extern void schedule_preempt_disabled(void);
332asmlinkage void preempt_schedule_irq(void);
333#ifdef CONFIG_PREEMPT_RT
334 extern void schedule_rtlock(void);
335#endif
336
337extern int __must_check io_schedule_prepare(void);
338extern void io_schedule_finish(int token);
339extern long io_schedule_timeout(long timeout);
340extern void io_schedule(void);
341
342/* wrapper function to trace from this header file */
343DECLARE_TRACEPOINT(sched_set_state_tp);
344extern void __trace_set_current_state(int state_value);
345
346/**
347 * struct prev_cputime - snapshot of system and user cputime
348 * @utime: time spent in user mode
349 * @stime: time spent in system mode
350 * @lock: protects the above two fields
351 *
352 * Stores previous user/system time values such that we can guarantee
353 * monotonicity.
354 */
355struct prev_cputime {
356#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
357 u64 utime;
358 u64 stime;
359 raw_spinlock_t lock;
360#endif
361};
362
363enum vtime_state {
364 /* Task is sleeping or running in a CPU with VTIME inactive: */
365 VTIME_INACTIVE = 0,
366 /* Task is idle */
367 VTIME_IDLE,
368 /* Task runs in kernelspace in a CPU with VTIME active: */
369 VTIME_SYS,
370 /* Task runs in userspace in a CPU with VTIME active: */
371 VTIME_USER,
372 /* Task runs as guests in a CPU with VTIME active: */
373 VTIME_GUEST,
374};
375
376struct vtime {
377 seqcount_t seqcount;
378 unsigned long long starttime;
379 enum vtime_state state;
380 unsigned int cpu;
381 u64 utime;
382 u64 stime;
383 u64 gtime;
384};
385
386/*
387 * Utilization clamp constraints.
388 * @UCLAMP_MIN: Minimum utilization
389 * @UCLAMP_MAX: Maximum utilization
390 * @UCLAMP_CNT: Utilization clamp constraints count
391 */
392enum uclamp_id {
393 UCLAMP_MIN = 0,
394 UCLAMP_MAX,
395 UCLAMP_CNT
396};
397
398#ifdef CONFIG_SMP
399extern struct root_domain def_root_domain;
400extern struct mutex sched_domains_mutex;
401extern void sched_domains_mutex_lock(void);
402extern void sched_domains_mutex_unlock(void);
403#else
404static inline void sched_domains_mutex_lock(void) { }
405static inline void sched_domains_mutex_unlock(void) { }
406#endif
407
408struct sched_param {
409 int sched_priority;
410};
411
412struct sched_info {
413#ifdef CONFIG_SCHED_INFO
414 /* Cumulative counters: */
415
416 /* # of times we have run on this CPU: */
417 unsigned long pcount;
418
419 /* Time spent waiting on a runqueue: */
420 unsigned long long run_delay;
421
422 /* Max time spent waiting on a runqueue: */
423 unsigned long long max_run_delay;
424
425 /* Min time spent waiting on a runqueue: */
426 unsigned long long min_run_delay;
427
428 /* Timestamps: */
429
430 /* When did we last run on a CPU? */
431 unsigned long long last_arrival;
432
433 /* When were we last queued to run? */
434 unsigned long long last_queued;
435
436#endif /* CONFIG_SCHED_INFO */
437};
438
439/*
440 * Integer metrics need fixed point arithmetic, e.g., sched/fair
441 * has a few: load, load_avg, util_avg, freq, and capacity.
442 *
443 * We define a basic fixed point arithmetic range, and then formalize
444 * all these metrics based on that basic range.
445 */
446# define SCHED_FIXEDPOINT_SHIFT 10
447# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
448
449/* Increase resolution of cpu_capacity calculations */
450# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
451# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
452
453struct load_weight {
454 unsigned long weight;
455 u32 inv_weight;
456};
457
458/*
459 * The load/runnable/util_avg accumulates an infinite geometric series
460 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
461 *
462 * [load_avg definition]
463 *
464 * load_avg = runnable% * scale_load_down(load)
465 *
466 * [runnable_avg definition]
467 *
468 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
469 *
470 * [util_avg definition]
471 *
472 * util_avg = running% * SCHED_CAPACITY_SCALE
473 *
474 * where runnable% is the time ratio that a sched_entity is runnable and
475 * running% the time ratio that a sched_entity is running.
476 *
477 * For cfs_rq, they are the aggregated values of all runnable and blocked
478 * sched_entities.
479 *
480 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
481 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
482 * for computing those signals (see update_rq_clock_pelt())
483 *
484 * N.B., the above ratios (runnable% and running%) themselves are in the
485 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
486 * to as large a range as necessary. This is for example reflected by
487 * util_avg's SCHED_CAPACITY_SCALE.
488 *
489 * [Overflow issue]
490 *
491 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
492 * with the highest load (=88761), always runnable on a single cfs_rq,
493 * and should not overflow as the number already hits PID_MAX_LIMIT.
494 *
495 * For all other cases (including 32-bit kernels), struct load_weight's
496 * weight will overflow first before we do, because:
497 *
498 * Max(load_avg) <= Max(load.weight)
499 *
500 * Then it is the load_weight's responsibility to consider overflow
501 * issues.
502 */
503struct sched_avg {
504 u64 last_update_time;
505 u64 load_sum;
506 u64 runnable_sum;
507 u32 util_sum;
508 u32 period_contrib;
509 unsigned long load_avg;
510 unsigned long runnable_avg;
511 unsigned long util_avg;
512 unsigned int util_est;
513} ____cacheline_aligned;
514
515/*
516 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
517 * updates. When a task is dequeued, its util_est should not be updated if its
518 * util_avg has not been updated in the meantime.
519 * This information is mapped into the MSB bit of util_est at dequeue time.
520 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
521 * it is safe to use MSB.
522 */
523#define UTIL_EST_WEIGHT_SHIFT 2
524#define UTIL_AVG_UNCHANGED 0x80000000
525
526struct sched_statistics {
527#ifdef CONFIG_SCHEDSTATS
528 u64 wait_start;
529 u64 wait_max;
530 u64 wait_count;
531 u64 wait_sum;
532 u64 iowait_count;
533 u64 iowait_sum;
534
535 u64 sleep_start;
536 u64 sleep_max;
537 s64 sum_sleep_runtime;
538
539 u64 block_start;
540 u64 block_max;
541 s64 sum_block_runtime;
542
543 s64 exec_max;
544 u64 slice_max;
545
546 u64 nr_migrations_cold;
547 u64 nr_failed_migrations_affine;
548 u64 nr_failed_migrations_running;
549 u64 nr_failed_migrations_hot;
550 u64 nr_forced_migrations;
551#ifdef CONFIG_NUMA_BALANCING
552 u64 numa_task_migrated;
553 u64 numa_task_swapped;
554#endif
555
556 u64 nr_wakeups;
557 u64 nr_wakeups_sync;
558 u64 nr_wakeups_migrate;
559 u64 nr_wakeups_local;
560 u64 nr_wakeups_remote;
561 u64 nr_wakeups_affine;
562 u64 nr_wakeups_affine_attempts;
563 u64 nr_wakeups_passive;
564 u64 nr_wakeups_idle;
565
566#ifdef CONFIG_SCHED_CORE
567 u64 core_forceidle_sum;
568#endif
569#endif /* CONFIG_SCHEDSTATS */
570} ____cacheline_aligned;
571
572struct sched_entity {
573 /* For load-balancing: */
574 struct load_weight load;
575 struct rb_node run_node;
576 u64 deadline;
577 u64 min_vruntime;
578 u64 min_slice;
579
580 struct list_head group_node;
581 unsigned char on_rq;
582 unsigned char sched_delayed;
583 unsigned char rel_deadline;
584 unsigned char custom_slice;
585 /* hole */
586
587 u64 exec_start;
588 u64 sum_exec_runtime;
589 u64 prev_sum_exec_runtime;
590 u64 vruntime;
591 s64 vlag;
592 u64 slice;
593
594 u64 nr_migrations;
595
596#ifdef CONFIG_FAIR_GROUP_SCHED
597 int depth;
598 struct sched_entity *parent;
599 /* rq on which this entity is (to be) queued: */
600 struct cfs_rq *cfs_rq;
601 /* rq "owned" by this entity/group: */
602 struct cfs_rq *my_q;
603 /* cached value of my_q->h_nr_running */
604 unsigned long runnable_weight;
605#endif
606
607#ifdef CONFIG_SMP
608 /*
609 * Per entity load average tracking.
610 *
611 * Put into separate cache line so it does not
612 * collide with read-mostly values above.
613 */
614 struct sched_avg avg;
615#endif
616};
617
618struct sched_rt_entity {
619 struct list_head run_list;
620 unsigned long timeout;
621 unsigned long watchdog_stamp;
622 unsigned int time_slice;
623 unsigned short on_rq;
624 unsigned short on_list;
625
626 struct sched_rt_entity *back;
627#ifdef CONFIG_RT_GROUP_SCHED
628 struct sched_rt_entity *parent;
629 /* rq on which this entity is (to be) queued: */
630 struct rt_rq *rt_rq;
631 /* rq "owned" by this entity/group: */
632 struct rt_rq *my_q;
633#endif
634} __randomize_layout;
635
636typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
637typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
638
639struct sched_dl_entity {
640 struct rb_node rb_node;
641
642 /*
643 * Original scheduling parameters. Copied here from sched_attr
644 * during sched_setattr(), they will remain the same until
645 * the next sched_setattr().
646 */
647 u64 dl_runtime; /* Maximum runtime for each instance */
648 u64 dl_deadline; /* Relative deadline of each instance */
649 u64 dl_period; /* Separation of two instances (period) */
650 u64 dl_bw; /* dl_runtime / dl_period */
651 u64 dl_density; /* dl_runtime / dl_deadline */
652
653 /*
654 * Actual scheduling parameters. Initialized with the values above,
655 * they are continuously updated during task execution. Note that
656 * the remaining runtime could be < 0 in case we are in overrun.
657 */
658 s64 runtime; /* Remaining runtime for this instance */
659 u64 deadline; /* Absolute deadline for this instance */
660 unsigned int flags; /* Specifying the scheduler behaviour */
661
662 /*
663 * Some bool flags:
664 *
665 * @dl_throttled tells if we exhausted the runtime. If so, the
666 * task has to wait for a replenishment to be performed at the
667 * next firing of dl_timer.
668 *
669 * @dl_yielded tells if task gave up the CPU before consuming
670 * all its available runtime during the last job.
671 *
672 * @dl_non_contending tells if the task is inactive while still
673 * contributing to the active utilization. In other words, it
674 * indicates if the inactive timer has been armed and its handler
675 * has not been executed yet. This flag is useful to avoid race
676 * conditions between the inactive timer handler and the wakeup
677 * code.
678 *
679 * @dl_overrun tells if the task asked to be informed about runtime
680 * overruns.
681 *
682 * @dl_server tells if this is a server entity.
683 *
684 * @dl_defer tells if this is a deferred or regular server. For
685 * now only defer server exists.
686 *
687 * @dl_defer_armed tells if the deferrable server is waiting
688 * for the replenishment timer to activate it.
689 *
690 * @dl_server_active tells if the dlserver is active(started).
691 * dlserver is started on first cfs enqueue on an idle runqueue
692 * and is stopped when a dequeue results in 0 cfs tasks on the
693 * runqueue. In other words, dlserver is active only when cpu's
694 * runqueue has atleast one cfs task.
695 *
696 * @dl_defer_running tells if the deferrable server is actually
697 * running, skipping the defer phase.
698 */
699 unsigned int dl_throttled : 1;
700 unsigned int dl_yielded : 1;
701 unsigned int dl_non_contending : 1;
702 unsigned int dl_overrun : 1;
703 unsigned int dl_server : 1;
704 unsigned int dl_server_active : 1;
705 unsigned int dl_defer : 1;
706 unsigned int dl_defer_armed : 1;
707 unsigned int dl_defer_running : 1;
708
709 /*
710 * Bandwidth enforcement timer. Each -deadline task has its
711 * own bandwidth to be enforced, thus we need one timer per task.
712 */
713 struct hrtimer dl_timer;
714
715 /*
716 * Inactive timer, responsible for decreasing the active utilization
717 * at the "0-lag time". When a -deadline task blocks, it contributes
718 * to GRUB's active utilization until the "0-lag time", hence a
719 * timer is needed to decrease the active utilization at the correct
720 * time.
721 */
722 struct hrtimer inactive_timer;
723
724 /*
725 * Bits for DL-server functionality. Also see the comment near
726 * dl_server_update().
727 *
728 * @rq the runqueue this server is for
729 *
730 * @server_has_tasks() returns true if @server_pick return a
731 * runnable task.
732 */
733 struct rq *rq;
734 dl_server_has_tasks_f server_has_tasks;
735 dl_server_pick_f server_pick_task;
736
737#ifdef CONFIG_RT_MUTEXES
738 /*
739 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
740 * pi_se points to the donor, otherwise points to the dl_se it belongs
741 * to (the original one/itself).
742 */
743 struct sched_dl_entity *pi_se;
744#endif
745};
746
747#ifdef CONFIG_UCLAMP_TASK
748/* Number of utilization clamp buckets (shorter alias) */
749#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
750
751/*
752 * Utilization clamp for a scheduling entity
753 * @value: clamp value "assigned" to a se
754 * @bucket_id: bucket index corresponding to the "assigned" value
755 * @active: the se is currently refcounted in a rq's bucket
756 * @user_defined: the requested clamp value comes from user-space
757 *
758 * The bucket_id is the index of the clamp bucket matching the clamp value
759 * which is pre-computed and stored to avoid expensive integer divisions from
760 * the fast path.
761 *
762 * The active bit is set whenever a task has got an "effective" value assigned,
763 * which can be different from the clamp value "requested" from user-space.
764 * This allows to know a task is refcounted in the rq's bucket corresponding
765 * to the "effective" bucket_id.
766 *
767 * The user_defined bit is set whenever a task has got a task-specific clamp
768 * value requested from userspace, i.e. the system defaults apply to this task
769 * just as a restriction. This allows to relax default clamps when a less
770 * restrictive task-specific value has been requested, thus allowing to
771 * implement a "nice" semantic. For example, a task running with a 20%
772 * default boost can still drop its own boosting to 0%.
773 */
774struct uclamp_se {
775 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
776 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
777 unsigned int active : 1;
778 unsigned int user_defined : 1;
779};
780#endif /* CONFIG_UCLAMP_TASK */
781
782union rcu_special {
783 struct {
784 u8 blocked;
785 u8 need_qs;
786 u8 exp_hint; /* Hint for performance. */
787 u8 need_mb; /* Readers need smp_mb(). */
788 } b; /* Bits. */
789 u32 s; /* Set of bits. */
790};
791
792enum perf_event_task_context {
793 perf_invalid_context = -1,
794 perf_hw_context = 0,
795 perf_sw_context,
796 perf_nr_task_contexts,
797};
798
799/*
800 * Number of contexts where an event can trigger:
801 * task, softirq, hardirq, nmi.
802 */
803#define PERF_NR_CONTEXTS 4
804
805struct wake_q_node {
806 struct wake_q_node *next;
807};
808
809struct kmap_ctrl {
810#ifdef CONFIG_KMAP_LOCAL
811 int idx;
812 pte_t pteval[KM_MAX_IDX];
813#endif
814};
815
816struct task_struct {
817#ifdef CONFIG_THREAD_INFO_IN_TASK
818 /*
819 * For reasons of header soup (see current_thread_info()), this
820 * must be the first element of task_struct.
821 */
822 struct thread_info thread_info;
823#endif
824 unsigned int __state;
825
826 /* saved state for "spinlock sleepers" */
827 unsigned int saved_state;
828
829 /*
830 * This begins the randomizable portion of task_struct. Only
831 * scheduling-critical items should be added above here.
832 */
833 randomized_struct_fields_start
834
835 void *stack;
836 refcount_t usage;
837 /* Per task flags (PF_*), defined further below: */
838 unsigned int flags;
839 unsigned int ptrace;
840
841#ifdef CONFIG_MEM_ALLOC_PROFILING
842 struct alloc_tag *alloc_tag;
843#endif
844
845#ifdef CONFIG_SMP
846 int on_cpu;
847 struct __call_single_node wake_entry;
848 unsigned int wakee_flips;
849 unsigned long wakee_flip_decay_ts;
850 struct task_struct *last_wakee;
851
852 /*
853 * recent_used_cpu is initially set as the last CPU used by a task
854 * that wakes affine another task. Waker/wakee relationships can
855 * push tasks around a CPU where each wakeup moves to the next one.
856 * Tracking a recently used CPU allows a quick search for a recently
857 * used CPU that may be idle.
858 */
859 int recent_used_cpu;
860 int wake_cpu;
861#endif
862 int on_rq;
863
864 int prio;
865 int static_prio;
866 int normal_prio;
867 unsigned int rt_priority;
868
869 struct sched_entity se;
870 struct sched_rt_entity rt;
871 struct sched_dl_entity dl;
872 struct sched_dl_entity *dl_server;
873#ifdef CONFIG_SCHED_CLASS_EXT
874 struct sched_ext_entity scx;
875#endif
876 const struct sched_class *sched_class;
877
878#ifdef CONFIG_SCHED_CORE
879 struct rb_node core_node;
880 unsigned long core_cookie;
881 unsigned int core_occupation;
882#endif
883
884#ifdef CONFIG_CGROUP_SCHED
885 struct task_group *sched_task_group;
886#endif
887
888
889#ifdef CONFIG_UCLAMP_TASK
890 /*
891 * Clamp values requested for a scheduling entity.
892 * Must be updated with task_rq_lock() held.
893 */
894 struct uclamp_se uclamp_req[UCLAMP_CNT];
895 /*
896 * Effective clamp values used for a scheduling entity.
897 * Must be updated with task_rq_lock() held.
898 */
899 struct uclamp_se uclamp[UCLAMP_CNT];
900#endif
901
902 struct sched_statistics stats;
903
904#ifdef CONFIG_PREEMPT_NOTIFIERS
905 /* List of struct preempt_notifier: */
906 struct hlist_head preempt_notifiers;
907#endif
908
909#ifdef CONFIG_BLK_DEV_IO_TRACE
910 unsigned int btrace_seq;
911#endif
912
913 unsigned int policy;
914 unsigned long max_allowed_capacity;
915 int nr_cpus_allowed;
916 const cpumask_t *cpus_ptr;
917 cpumask_t *user_cpus_ptr;
918 cpumask_t cpus_mask;
919 void *migration_pending;
920#ifdef CONFIG_SMP
921 unsigned short migration_disabled;
922#endif
923 unsigned short migration_flags;
924
925#ifdef CONFIG_PREEMPT_RCU
926 int rcu_read_lock_nesting;
927 union rcu_special rcu_read_unlock_special;
928 struct list_head rcu_node_entry;
929 struct rcu_node *rcu_blocked_node;
930#endif /* #ifdef CONFIG_PREEMPT_RCU */
931
932#ifdef CONFIG_TASKS_RCU
933 unsigned long rcu_tasks_nvcsw;
934 u8 rcu_tasks_holdout;
935 u8 rcu_tasks_idx;
936 int rcu_tasks_idle_cpu;
937 struct list_head rcu_tasks_holdout_list;
938 int rcu_tasks_exit_cpu;
939 struct list_head rcu_tasks_exit_list;
940#endif /* #ifdef CONFIG_TASKS_RCU */
941
942#ifdef CONFIG_TASKS_TRACE_RCU
943 int trc_reader_nesting;
944 int trc_ipi_to_cpu;
945 union rcu_special trc_reader_special;
946 struct list_head trc_holdout_list;
947 struct list_head trc_blkd_node;
948 int trc_blkd_cpu;
949#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
950
951 struct sched_info sched_info;
952
953 struct list_head tasks;
954#ifdef CONFIG_SMP
955 struct plist_node pushable_tasks;
956 struct rb_node pushable_dl_tasks;
957#endif
958
959 struct mm_struct *mm;
960 struct mm_struct *active_mm;
961 struct address_space *faults_disabled_mapping;
962
963 int exit_state;
964 int exit_code;
965 int exit_signal;
966 /* The signal sent when the parent dies: */
967 int pdeath_signal;
968 /* JOBCTL_*, siglock protected: */
969 unsigned long jobctl;
970
971 /* Used for emulating ABI behavior of previous Linux versions: */
972 unsigned int personality;
973
974 /* Scheduler bits, serialized by scheduler locks: */
975 unsigned sched_reset_on_fork:1;
976 unsigned sched_contributes_to_load:1;
977 unsigned sched_migrated:1;
978 unsigned sched_task_hot:1;
979
980 /* Force alignment to the next boundary: */
981 unsigned :0;
982
983 /* Unserialized, strictly 'current' */
984
985 /*
986 * This field must not be in the scheduler word above due to wakelist
987 * queueing no longer being serialized by p->on_cpu. However:
988 *
989 * p->XXX = X; ttwu()
990 * schedule() if (p->on_rq && ..) // false
991 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
992 * deactivate_task() ttwu_queue_wakelist())
993 * p->on_rq = 0; p->sched_remote_wakeup = Y;
994 *
995 * guarantees all stores of 'current' are visible before
996 * ->sched_remote_wakeup gets used, so it can be in this word.
997 */
998 unsigned sched_remote_wakeup:1;
999#ifdef CONFIG_RT_MUTEXES
1000 unsigned sched_rt_mutex:1;
1001#endif
1002
1003 /* Bit to tell TOMOYO we're in execve(): */
1004 unsigned in_execve:1;
1005 unsigned in_iowait:1;
1006#ifndef TIF_RESTORE_SIGMASK
1007 unsigned restore_sigmask:1;
1008#endif
1009#ifdef CONFIG_MEMCG_V1
1010 unsigned in_user_fault:1;
1011#endif
1012#ifdef CONFIG_LRU_GEN
1013 /* whether the LRU algorithm may apply to this access */
1014 unsigned in_lru_fault:1;
1015#endif
1016#ifdef CONFIG_COMPAT_BRK
1017 unsigned brk_randomized:1;
1018#endif
1019#ifdef CONFIG_CGROUPS
1020 /* disallow userland-initiated cgroup migration */
1021 unsigned no_cgroup_migration:1;
1022 /* task is frozen/stopped (used by the cgroup freezer) */
1023 unsigned frozen:1;
1024#endif
1025#ifdef CONFIG_BLK_CGROUP
1026 unsigned use_memdelay:1;
1027#endif
1028#ifdef CONFIG_PSI
1029 /* Stalled due to lack of memory */
1030 unsigned in_memstall:1;
1031#endif
1032#ifdef CONFIG_PAGE_OWNER
1033 /* Used by page_owner=on to detect recursion in page tracking. */
1034 unsigned in_page_owner:1;
1035#endif
1036#ifdef CONFIG_EVENTFD
1037 /* Recursion prevention for eventfd_signal() */
1038 unsigned in_eventfd:1;
1039#endif
1040#ifdef CONFIG_ARCH_HAS_CPU_PASID
1041 unsigned pasid_activated:1;
1042#endif
1043#ifdef CONFIG_X86_BUS_LOCK_DETECT
1044 unsigned reported_split_lock:1;
1045#endif
1046#ifdef CONFIG_TASK_DELAY_ACCT
1047 /* delay due to memory thrashing */
1048 unsigned in_thrashing:1;
1049#endif
1050 unsigned in_nf_duplicate:1;
1051#ifdef CONFIG_PREEMPT_RT
1052 struct netdev_xmit net_xmit;
1053#endif
1054 unsigned long atomic_flags; /* Flags requiring atomic access. */
1055
1056 struct restart_block restart_block;
1057
1058 pid_t pid;
1059 pid_t tgid;
1060
1061#ifdef CONFIG_STACKPROTECTOR
1062 /* Canary value for the -fstack-protector GCC feature: */
1063 unsigned long stack_canary;
1064#endif
1065 /*
1066 * Pointers to the (original) parent process, youngest child, younger sibling,
1067 * older sibling, respectively. (p->father can be replaced with
1068 * p->real_parent->pid)
1069 */
1070
1071 /* Real parent process: */
1072 struct task_struct __rcu *real_parent;
1073
1074 /* Recipient of SIGCHLD, wait4() reports: */
1075 struct task_struct __rcu *parent;
1076
1077 /*
1078 * Children/sibling form the list of natural children:
1079 */
1080 struct list_head children;
1081 struct list_head sibling;
1082 struct task_struct *group_leader;
1083
1084 /*
1085 * 'ptraced' is the list of tasks this task is using ptrace() on.
1086 *
1087 * This includes both natural children and PTRACE_ATTACH targets.
1088 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1089 */
1090 struct list_head ptraced;
1091 struct list_head ptrace_entry;
1092
1093 /* PID/PID hash table linkage. */
1094 struct pid *thread_pid;
1095 struct hlist_node pid_links[PIDTYPE_MAX];
1096 struct list_head thread_node;
1097
1098 struct completion *vfork_done;
1099
1100 /* CLONE_CHILD_SETTID: */
1101 int __user *set_child_tid;
1102
1103 /* CLONE_CHILD_CLEARTID: */
1104 int __user *clear_child_tid;
1105
1106 /* PF_KTHREAD | PF_IO_WORKER */
1107 void *worker_private;
1108
1109 u64 utime;
1110 u64 stime;
1111#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1112 u64 utimescaled;
1113 u64 stimescaled;
1114#endif
1115 u64 gtime;
1116 struct prev_cputime prev_cputime;
1117#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1118 struct vtime vtime;
1119#endif
1120
1121#ifdef CONFIG_NO_HZ_FULL
1122 atomic_t tick_dep_mask;
1123#endif
1124 /* Context switch counts: */
1125 unsigned long nvcsw;
1126 unsigned long nivcsw;
1127
1128 /* Monotonic time in nsecs: */
1129 u64 start_time;
1130
1131 /* Boot based time in nsecs: */
1132 u64 start_boottime;
1133
1134 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1135 unsigned long min_flt;
1136 unsigned long maj_flt;
1137
1138 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1139 struct posix_cputimers posix_cputimers;
1140
1141#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1142 struct posix_cputimers_work posix_cputimers_work;
1143#endif
1144
1145 /* Process credentials: */
1146
1147 /* Tracer's credentials at attach: */
1148 const struct cred __rcu *ptracer_cred;
1149
1150 /* Objective and real subjective task credentials (COW): */
1151 const struct cred __rcu *real_cred;
1152
1153 /* Effective (overridable) subjective task credentials (COW): */
1154 const struct cred __rcu *cred;
1155
1156#ifdef CONFIG_KEYS
1157 /* Cached requested key. */
1158 struct key *cached_requested_key;
1159#endif
1160
1161 /*
1162 * executable name, excluding path.
1163 *
1164 * - normally initialized begin_new_exec()
1165 * - set it with set_task_comm()
1166 * - strscpy_pad() to ensure it is always NUL-terminated and
1167 * zero-padded
1168 * - task_lock() to ensure the operation is atomic and the name is
1169 * fully updated.
1170 */
1171 char comm[TASK_COMM_LEN];
1172
1173 struct nameidata *nameidata;
1174
1175#ifdef CONFIG_SYSVIPC
1176 struct sysv_sem sysvsem;
1177 struct sysv_shm sysvshm;
1178#endif
1179#ifdef CONFIG_DETECT_HUNG_TASK
1180 unsigned long last_switch_count;
1181 unsigned long last_switch_time;
1182#endif
1183 /* Filesystem information: */
1184 struct fs_struct *fs;
1185
1186 /* Open file information: */
1187 struct files_struct *files;
1188
1189#ifdef CONFIG_IO_URING
1190 struct io_uring_task *io_uring;
1191#endif
1192
1193 /* Namespaces: */
1194 struct nsproxy *nsproxy;
1195
1196 /* Signal handlers: */
1197 struct signal_struct *signal;
1198 struct sighand_struct __rcu *sighand;
1199 sigset_t blocked;
1200 sigset_t real_blocked;
1201 /* Restored if set_restore_sigmask() was used: */
1202 sigset_t saved_sigmask;
1203 struct sigpending pending;
1204 unsigned long sas_ss_sp;
1205 size_t sas_ss_size;
1206 unsigned int sas_ss_flags;
1207
1208 struct callback_head *task_works;
1209
1210#ifdef CONFIG_AUDIT
1211#ifdef CONFIG_AUDITSYSCALL
1212 struct audit_context *audit_context;
1213#endif
1214 kuid_t loginuid;
1215 unsigned int sessionid;
1216#endif
1217 struct seccomp seccomp;
1218 struct syscall_user_dispatch syscall_dispatch;
1219
1220 /* Thread group tracking: */
1221 u64 parent_exec_id;
1222 u64 self_exec_id;
1223
1224 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1225 spinlock_t alloc_lock;
1226
1227 /* Protection of the PI data structures: */
1228 raw_spinlock_t pi_lock;
1229
1230 struct wake_q_node wake_q;
1231
1232#ifdef CONFIG_RT_MUTEXES
1233 /* PI waiters blocked on a rt_mutex held by this task: */
1234 struct rb_root_cached pi_waiters;
1235 /* Updated under owner's pi_lock and rq lock */
1236 struct task_struct *pi_top_task;
1237 /* Deadlock detection and priority inheritance handling: */
1238 struct rt_mutex_waiter *pi_blocked_on;
1239#endif
1240
1241#ifdef CONFIG_DEBUG_MUTEXES
1242 /* Mutex deadlock detection: */
1243 struct mutex_waiter *blocked_on;
1244#endif
1245
1246#ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER
1247 /*
1248 * Encoded lock address causing task block (lower 2 bits = type from
1249 * <linux/hung_task.h>). Accessed via hung_task_*() helpers.
1250 */
1251 unsigned long blocker;
1252#endif
1253
1254#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1255 int non_block_count;
1256#endif
1257
1258#ifdef CONFIG_TRACE_IRQFLAGS
1259 struct irqtrace_events irqtrace;
1260 unsigned int hardirq_threaded;
1261 u64 hardirq_chain_key;
1262 int softirqs_enabled;
1263 int softirq_context;
1264 int irq_config;
1265#endif
1266#ifdef CONFIG_PREEMPT_RT
1267 int softirq_disable_cnt;
1268#endif
1269
1270#ifdef CONFIG_LOCKDEP
1271# define MAX_LOCK_DEPTH 48UL
1272 u64 curr_chain_key;
1273 int lockdep_depth;
1274 unsigned int lockdep_recursion;
1275 struct held_lock held_locks[MAX_LOCK_DEPTH];
1276#endif
1277
1278#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1279 unsigned int in_ubsan;
1280#endif
1281
1282 /* Journalling filesystem info: */
1283 void *journal_info;
1284
1285 /* Stacked block device info: */
1286 struct bio_list *bio_list;
1287
1288 /* Stack plugging: */
1289 struct blk_plug *plug;
1290
1291 /* VM state: */
1292 struct reclaim_state *reclaim_state;
1293
1294 struct io_context *io_context;
1295
1296#ifdef CONFIG_COMPACTION
1297 struct capture_control *capture_control;
1298#endif
1299 /* Ptrace state: */
1300 unsigned long ptrace_message;
1301 kernel_siginfo_t *last_siginfo;
1302
1303 struct task_io_accounting ioac;
1304#ifdef CONFIG_PSI
1305 /* Pressure stall state */
1306 unsigned int psi_flags;
1307#endif
1308#ifdef CONFIG_TASK_XACCT
1309 /* Accumulated RSS usage: */
1310 u64 acct_rss_mem1;
1311 /* Accumulated virtual memory usage: */
1312 u64 acct_vm_mem1;
1313 /* stime + utime since last update: */
1314 u64 acct_timexpd;
1315#endif
1316#ifdef CONFIG_CPUSETS
1317 /* Protected by ->alloc_lock: */
1318 nodemask_t mems_allowed;
1319 /* Sequence number to catch updates: */
1320 seqcount_spinlock_t mems_allowed_seq;
1321 int cpuset_mem_spread_rotor;
1322#endif
1323#ifdef CONFIG_CGROUPS
1324 /* Control Group info protected by css_set_lock: */
1325 struct css_set __rcu *cgroups;
1326 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1327 struct list_head cg_list;
1328#endif
1329#ifdef CONFIG_X86_CPU_RESCTRL
1330 u32 closid;
1331 u32 rmid;
1332#endif
1333#ifdef CONFIG_FUTEX
1334 struct robust_list_head __user *robust_list;
1335#ifdef CONFIG_COMPAT
1336 struct compat_robust_list_head __user *compat_robust_list;
1337#endif
1338 struct list_head pi_state_list;
1339 struct futex_pi_state *pi_state_cache;
1340 struct mutex futex_exit_mutex;
1341 unsigned int futex_state;
1342#endif
1343#ifdef CONFIG_PERF_EVENTS
1344 u8 perf_recursion[PERF_NR_CONTEXTS];
1345 struct perf_event_context *perf_event_ctxp;
1346 struct mutex perf_event_mutex;
1347 struct list_head perf_event_list;
1348 struct perf_ctx_data __rcu *perf_ctx_data;
1349#endif
1350#ifdef CONFIG_DEBUG_PREEMPT
1351 unsigned long preempt_disable_ip;
1352#endif
1353#ifdef CONFIG_NUMA
1354 /* Protected by alloc_lock: */
1355 struct mempolicy *mempolicy;
1356 short il_prev;
1357 u8 il_weight;
1358 short pref_node_fork;
1359#endif
1360#ifdef CONFIG_NUMA_BALANCING
1361 int numa_scan_seq;
1362 unsigned int numa_scan_period;
1363 unsigned int numa_scan_period_max;
1364 int numa_preferred_nid;
1365 unsigned long numa_migrate_retry;
1366 /* Migration stamp: */
1367 u64 node_stamp;
1368 u64 last_task_numa_placement;
1369 u64 last_sum_exec_runtime;
1370 struct callback_head numa_work;
1371
1372 /*
1373 * This pointer is only modified for current in syscall and
1374 * pagefault context (and for tasks being destroyed), so it can be read
1375 * from any of the following contexts:
1376 * - RCU read-side critical section
1377 * - current->numa_group from everywhere
1378 * - task's runqueue locked, task not running
1379 */
1380 struct numa_group __rcu *numa_group;
1381
1382 /*
1383 * numa_faults is an array split into four regions:
1384 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1385 * in this precise order.
1386 *
1387 * faults_memory: Exponential decaying average of faults on a per-node
1388 * basis. Scheduling placement decisions are made based on these
1389 * counts. The values remain static for the duration of a PTE scan.
1390 * faults_cpu: Track the nodes the process was running on when a NUMA
1391 * hinting fault was incurred.
1392 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1393 * during the current scan window. When the scan completes, the counts
1394 * in faults_memory and faults_cpu decay and these values are copied.
1395 */
1396 unsigned long *numa_faults;
1397 unsigned long total_numa_faults;
1398
1399 /*
1400 * numa_faults_locality tracks if faults recorded during the last
1401 * scan window were remote/local or failed to migrate. The task scan
1402 * period is adapted based on the locality of the faults with different
1403 * weights depending on whether they were shared or private faults
1404 */
1405 unsigned long numa_faults_locality[3];
1406
1407 unsigned long numa_pages_migrated;
1408#endif /* CONFIG_NUMA_BALANCING */
1409
1410#ifdef CONFIG_RSEQ
1411 struct rseq __user *rseq;
1412 u32 rseq_len;
1413 u32 rseq_sig;
1414 /*
1415 * RmW on rseq_event_mask must be performed atomically
1416 * with respect to preemption.
1417 */
1418 unsigned long rseq_event_mask;
1419# ifdef CONFIG_DEBUG_RSEQ
1420 /*
1421 * This is a place holder to save a copy of the rseq fields for
1422 * validation of read-only fields. The struct rseq has a
1423 * variable-length array at the end, so it cannot be used
1424 * directly. Reserve a size large enough for the known fields.
1425 */
1426 char rseq_fields[sizeof(struct rseq)];
1427# endif
1428#endif
1429
1430#ifdef CONFIG_SCHED_MM_CID
1431 int mm_cid; /* Current cid in mm */
1432 int last_mm_cid; /* Most recent cid in mm */
1433 int migrate_from_cpu;
1434 int mm_cid_active; /* Whether cid bitmap is active */
1435 struct callback_head cid_work;
1436#endif
1437
1438 struct tlbflush_unmap_batch tlb_ubc;
1439
1440 /* Cache last used pipe for splice(): */
1441 struct pipe_inode_info *splice_pipe;
1442
1443 struct page_frag task_frag;
1444
1445#ifdef CONFIG_TASK_DELAY_ACCT
1446 struct task_delay_info *delays;
1447#endif
1448
1449#ifdef CONFIG_FAULT_INJECTION
1450 int make_it_fail;
1451 unsigned int fail_nth;
1452#endif
1453 /*
1454 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1455 * balance_dirty_pages() for a dirty throttling pause:
1456 */
1457 int nr_dirtied;
1458 int nr_dirtied_pause;
1459 /* Start of a write-and-pause period: */
1460 unsigned long dirty_paused_when;
1461
1462#ifdef CONFIG_LATENCYTOP
1463 int latency_record_count;
1464 struct latency_record latency_record[LT_SAVECOUNT];
1465#endif
1466 /*
1467 * Time slack values; these are used to round up poll() and
1468 * select() etc timeout values. These are in nanoseconds.
1469 */
1470 u64 timer_slack_ns;
1471 u64 default_timer_slack_ns;
1472
1473#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1474 unsigned int kasan_depth;
1475#endif
1476
1477#ifdef CONFIG_KCSAN
1478 struct kcsan_ctx kcsan_ctx;
1479#ifdef CONFIG_TRACE_IRQFLAGS
1480 struct irqtrace_events kcsan_save_irqtrace;
1481#endif
1482#ifdef CONFIG_KCSAN_WEAK_MEMORY
1483 int kcsan_stack_depth;
1484#endif
1485#endif
1486
1487#ifdef CONFIG_KMSAN
1488 struct kmsan_ctx kmsan_ctx;
1489#endif
1490
1491#if IS_ENABLED(CONFIG_KUNIT)
1492 struct kunit *kunit_test;
1493#endif
1494
1495#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1496 /* Index of current stored address in ret_stack: */
1497 int curr_ret_stack;
1498 int curr_ret_depth;
1499
1500 /* Stack of return addresses for return function tracing: */
1501 unsigned long *ret_stack;
1502
1503 /* Timestamp for last schedule: */
1504 unsigned long long ftrace_timestamp;
1505 unsigned long long ftrace_sleeptime;
1506
1507 /*
1508 * Number of functions that haven't been traced
1509 * because of depth overrun:
1510 */
1511 atomic_t trace_overrun;
1512
1513 /* Pause tracing: */
1514 atomic_t tracing_graph_pause;
1515#endif
1516
1517#ifdef CONFIG_TRACING
1518 /* Bitmask and counter of trace recursion: */
1519 unsigned long trace_recursion;
1520#endif /* CONFIG_TRACING */
1521
1522#ifdef CONFIG_KCOV
1523 /* See kernel/kcov.c for more details. */
1524
1525 /* Coverage collection mode enabled for this task (0 if disabled): */
1526 unsigned int kcov_mode;
1527
1528 /* Size of the kcov_area: */
1529 unsigned int kcov_size;
1530
1531 /* Buffer for coverage collection: */
1532 void *kcov_area;
1533
1534 /* KCOV descriptor wired with this task or NULL: */
1535 struct kcov *kcov;
1536
1537 /* KCOV common handle for remote coverage collection: */
1538 u64 kcov_handle;
1539
1540 /* KCOV sequence number: */
1541 int kcov_sequence;
1542
1543 /* Collect coverage from softirq context: */
1544 unsigned int kcov_softirq;
1545#endif
1546
1547#ifdef CONFIG_MEMCG_V1
1548 struct mem_cgroup *memcg_in_oom;
1549#endif
1550
1551#ifdef CONFIG_MEMCG
1552 /* Number of pages to reclaim on returning to userland: */
1553 unsigned int memcg_nr_pages_over_high;
1554
1555 /* Used by memcontrol for targeted memcg charge: */
1556 struct mem_cgroup *active_memcg;
1557
1558 /* Cache for current->cgroups->memcg->objcg lookups: */
1559 struct obj_cgroup *objcg;
1560#endif
1561
1562#ifdef CONFIG_BLK_CGROUP
1563 struct gendisk *throttle_disk;
1564#endif
1565
1566#ifdef CONFIG_UPROBES
1567 struct uprobe_task *utask;
1568#endif
1569#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1570 unsigned int sequential_io;
1571 unsigned int sequential_io_avg;
1572#endif
1573 struct kmap_ctrl kmap_ctrl;
1574#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1575 unsigned long task_state_change;
1576# ifdef CONFIG_PREEMPT_RT
1577 unsigned long saved_state_change;
1578# endif
1579#endif
1580 struct rcu_head rcu;
1581 refcount_t rcu_users;
1582 int pagefault_disabled;
1583#ifdef CONFIG_MMU
1584 struct task_struct *oom_reaper_list;
1585 struct timer_list oom_reaper_timer;
1586#endif
1587#ifdef CONFIG_VMAP_STACK
1588 struct vm_struct *stack_vm_area;
1589#endif
1590#ifdef CONFIG_THREAD_INFO_IN_TASK
1591 /* A live task holds one reference: */
1592 refcount_t stack_refcount;
1593#endif
1594#ifdef CONFIG_LIVEPATCH
1595 int patch_state;
1596#endif
1597#ifdef CONFIG_SECURITY
1598 /* Used by LSM modules for access restriction: */
1599 void *security;
1600#endif
1601#ifdef CONFIG_BPF_SYSCALL
1602 /* Used by BPF task local storage */
1603 struct bpf_local_storage __rcu *bpf_storage;
1604 /* Used for BPF run context */
1605 struct bpf_run_ctx *bpf_ctx;
1606#endif
1607 /* Used by BPF for per-TASK xdp storage */
1608 struct bpf_net_context *bpf_net_context;
1609
1610#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1611 unsigned long lowest_stack;
1612 unsigned long prev_lowest_stack;
1613#endif
1614
1615#ifdef CONFIG_X86_MCE
1616 void __user *mce_vaddr;
1617 __u64 mce_kflags;
1618 u64 mce_addr;
1619 __u64 mce_ripv : 1,
1620 mce_whole_page : 1,
1621 __mce_reserved : 62;
1622 struct callback_head mce_kill_me;
1623 int mce_count;
1624#endif
1625
1626#ifdef CONFIG_KRETPROBES
1627 struct llist_head kretprobe_instances;
1628#endif
1629#ifdef CONFIG_RETHOOK
1630 struct llist_head rethooks;
1631#endif
1632
1633#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1634 /*
1635 * If L1D flush is supported on mm context switch
1636 * then we use this callback head to queue kill work
1637 * to kill tasks that are not running on SMT disabled
1638 * cores
1639 */
1640 struct callback_head l1d_flush_kill;
1641#endif
1642
1643#ifdef CONFIG_RV
1644 /*
1645 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1646 * If we find justification for more monitors, we can think
1647 * about adding more or developing a dynamic method. So far,
1648 * none of these are justified.
1649 */
1650 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1651#endif
1652
1653#ifdef CONFIG_USER_EVENTS
1654 struct user_event_mm *user_event_mm;
1655#endif
1656
1657 /* CPU-specific state of this task: */
1658 struct thread_struct thread;
1659
1660 /*
1661 * New fields for task_struct should be added above here, so that
1662 * they are included in the randomized portion of task_struct.
1663 */
1664 randomized_struct_fields_end
1665} __attribute__ ((aligned (64)));
1666
1667#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1668#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1669
1670static inline unsigned int __task_state_index(unsigned int tsk_state,
1671 unsigned int tsk_exit_state)
1672{
1673 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1674
1675 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1676
1677 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1678 state = TASK_REPORT_IDLE;
1679
1680 /*
1681 * We're lying here, but rather than expose a completely new task state
1682 * to userspace, we can make this appear as if the task has gone through
1683 * a regular rt_mutex_lock() call.
1684 * Report frozen tasks as uninterruptible.
1685 */
1686 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
1687 state = TASK_UNINTERRUPTIBLE;
1688
1689 return fls(x: state);
1690}
1691
1692static inline unsigned int task_state_index(struct task_struct *tsk)
1693{
1694 return __task_state_index(READ_ONCE(tsk->__state), tsk_exit_state: tsk->exit_state);
1695}
1696
1697static inline char task_index_to_char(unsigned int state)
1698{
1699 static const char state_char[] = "RSDTtXZPI";
1700
1701 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1702
1703 return state_char[state];
1704}
1705
1706static inline char task_state_to_char(struct task_struct *tsk)
1707{
1708 return task_index_to_char(state: task_state_index(tsk));
1709}
1710
1711extern struct pid *cad_pid;
1712
1713/*
1714 * Per process flags
1715 */
1716#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1717#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1718#define PF_EXITING 0x00000004 /* Getting shut down */
1719#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1720#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1721#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1722#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1723#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1724#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1725#define PF_DUMPCORE 0x00000200 /* Dumped core */
1726#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1727#define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1728#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1729#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1730#define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1731#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1732#define PF_KCOMPACTD 0x00010000 /* I am kcompactd */
1733#define PF_KSWAPD 0x00020000 /* I am kswapd */
1734#define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1735#define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1736#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1737 * I am cleaning dirty pages from some other bdi. */
1738#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1739#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1740#define PF__HOLE__00800000 0x00800000
1741#define PF__HOLE__01000000 0x01000000
1742#define PF__HOLE__02000000 0x02000000
1743#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1744#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1745#define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1746 * See memalloc_pin_save() */
1747#define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1748#define PF__HOLE__40000000 0x40000000
1749#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1750
1751/*
1752 * Only the _current_ task can read/write to tsk->flags, but other
1753 * tasks can access tsk->flags in readonly mode for example
1754 * with tsk_used_math (like during threaded core dumping).
1755 * There is however an exception to this rule during ptrace
1756 * or during fork: the ptracer task is allowed to write to the
1757 * child->flags of its traced child (same goes for fork, the parent
1758 * can write to the child->flags), because we're guaranteed the
1759 * child is not running and in turn not changing child->flags
1760 * at the same time the parent does it.
1761 */
1762#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1763#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1764#define clear_used_math() clear_stopped_child_used_math(current)
1765#define set_used_math() set_stopped_child_used_math(current)
1766
1767#define conditional_stopped_child_used_math(condition, child) \
1768 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1769
1770#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1771
1772#define copy_to_stopped_child_used_math(child) \
1773 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1774
1775/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1776#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1777#define used_math() tsk_used_math(current)
1778
1779static __always_inline bool is_percpu_thread(void)
1780{
1781#ifdef CONFIG_SMP
1782 return (current->flags & PF_NO_SETAFFINITY) &&
1783 (current->nr_cpus_allowed == 1);
1784#else
1785 return true;
1786#endif
1787}
1788
1789/* Per-process atomic flags. */
1790#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1791#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1792#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1793#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1794#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1795#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1796#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1797#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1798
1799#define TASK_PFA_TEST(name, func) \
1800 static inline bool task_##func(struct task_struct *p) \
1801 { return test_bit(PFA_##name, &p->atomic_flags); }
1802
1803#define TASK_PFA_SET(name, func) \
1804 static inline void task_set_##func(struct task_struct *p) \
1805 { set_bit(PFA_##name, &p->atomic_flags); }
1806
1807#define TASK_PFA_CLEAR(name, func) \
1808 static inline void task_clear_##func(struct task_struct *p) \
1809 { clear_bit(PFA_##name, &p->atomic_flags); }
1810
1811TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1812TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1813
1814TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1815TASK_PFA_SET(SPREAD_PAGE, spread_page)
1816TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1817
1818TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1819TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1820TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1821
1822TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1823TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1824TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1825
1826TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1827TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1828TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1829
1830TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1831TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1832
1833TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1834TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1835TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1836
1837TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1838TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1839
1840static inline void
1841current_restore_flags(unsigned long orig_flags, unsigned long flags)
1842{
1843 current->flags &= ~flags;
1844 current->flags |= orig_flags & flags;
1845}
1846
1847extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1848extern int task_can_attach(struct task_struct *p);
1849extern int dl_bw_alloc(int cpu, u64 dl_bw);
1850extern void dl_bw_free(int cpu, u64 dl_bw);
1851#ifdef CONFIG_SMP
1852
1853/* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1854extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1855
1856/**
1857 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1858 * @p: the task
1859 * @new_mask: CPU affinity mask
1860 *
1861 * Return: zero if successful, or a negative error code
1862 */
1863extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1864extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1865extern void release_user_cpus_ptr(struct task_struct *p);
1866extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1867extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1868extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1869#else
1870static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1871{
1872}
1873static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1874{
1875 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1876 if ((*cpumask_bits(new_mask) & 1) == 0)
1877 return -EINVAL;
1878 return 0;
1879}
1880static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1881{
1882 if (src->user_cpus_ptr)
1883 return -EINVAL;
1884 return 0;
1885}
1886static inline void release_user_cpus_ptr(struct task_struct *p)
1887{
1888 WARN_ON(p->user_cpus_ptr);
1889}
1890
1891static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1892{
1893 return 0;
1894}
1895#endif
1896
1897extern int yield_to(struct task_struct *p, bool preempt);
1898extern void set_user_nice(struct task_struct *p, long nice);
1899extern int task_prio(const struct task_struct *p);
1900
1901/**
1902 * task_nice - return the nice value of a given task.
1903 * @p: the task in question.
1904 *
1905 * Return: The nice value [ -20 ... 0 ... 19 ].
1906 */
1907static inline int task_nice(const struct task_struct *p)
1908{
1909 return PRIO_TO_NICE((p)->static_prio);
1910}
1911
1912extern int can_nice(const struct task_struct *p, const int nice);
1913extern int task_curr(const struct task_struct *p);
1914extern int idle_cpu(int cpu);
1915extern int available_idle_cpu(int cpu);
1916extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1917extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1918extern void sched_set_fifo(struct task_struct *p);
1919extern void sched_set_fifo_low(struct task_struct *p);
1920extern void sched_set_normal(struct task_struct *p, int nice);
1921extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1922extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1923extern struct task_struct *idle_task(int cpu);
1924
1925/**
1926 * is_idle_task - is the specified task an idle task?
1927 * @p: the task in question.
1928 *
1929 * Return: 1 if @p is an idle task. 0 otherwise.
1930 */
1931static __always_inline bool is_idle_task(const struct task_struct *p)
1932{
1933 return !!(p->flags & PF_IDLE);
1934}
1935
1936extern struct task_struct *curr_task(int cpu);
1937extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1938
1939void yield(void);
1940
1941union thread_union {
1942 struct task_struct task;
1943#ifndef CONFIG_THREAD_INFO_IN_TASK
1944 struct thread_info thread_info;
1945#endif
1946 unsigned long stack[THREAD_SIZE/sizeof(long)];
1947};
1948
1949#ifndef CONFIG_THREAD_INFO_IN_TASK
1950extern struct thread_info init_thread_info;
1951#endif
1952
1953extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1954
1955#ifdef CONFIG_THREAD_INFO_IN_TASK
1956# define task_thread_info(task) (&(task)->thread_info)
1957#else
1958# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1959#endif
1960
1961/*
1962 * find a task by one of its numerical ids
1963 *
1964 * find_task_by_pid_ns():
1965 * finds a task by its pid in the specified namespace
1966 * find_task_by_vpid():
1967 * finds a task by its virtual pid
1968 *
1969 * see also find_vpid() etc in include/linux/pid.h
1970 */
1971
1972extern struct task_struct *find_task_by_vpid(pid_t nr);
1973extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1974
1975/*
1976 * find a task by its virtual pid and get the task struct
1977 */
1978extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1979
1980extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1981extern int wake_up_process(struct task_struct *tsk);
1982extern void wake_up_new_task(struct task_struct *tsk);
1983
1984#ifdef CONFIG_SMP
1985extern void kick_process(struct task_struct *tsk);
1986#else
1987static inline void kick_process(struct task_struct *tsk) { }
1988#endif
1989
1990extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1991#define set_task_comm(tsk, from) ({ \
1992 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \
1993 __set_task_comm(tsk, from, false); \
1994})
1995
1996/*
1997 * - Why not use task_lock()?
1998 * User space can randomly change their names anyway, so locking for readers
1999 * doesn't make sense. For writers, locking is probably necessary, as a race
2000 * condition could lead to long-term mixed results.
2001 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is
2002 * always NUL-terminated and zero-padded. Therefore the race condition between
2003 * reader and writer is not an issue.
2004 *
2005 * - BUILD_BUG_ON() can help prevent the buf from being truncated.
2006 * Since the callers don't perform any return value checks, this safeguard is
2007 * necessary.
2008 */
2009#define get_task_comm(buf, tsk) ({ \
2010 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \
2011 strscpy_pad(buf, (tsk)->comm); \
2012 buf; \
2013})
2014
2015#ifdef CONFIG_SMP
2016static __always_inline void scheduler_ipi(void)
2017{
2018 /*
2019 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2020 * TIF_NEED_RESCHED remotely (for the first time) will also send
2021 * this IPI.
2022 */
2023 preempt_fold_need_resched();
2024}
2025#else
2026static inline void scheduler_ipi(void) { }
2027#endif
2028
2029extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2030
2031/*
2032 * Set thread flags in other task's structures.
2033 * See asm/thread_info.h for TIF_xxxx flags available:
2034 */
2035static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2036{
2037 set_ti_thread_flag(task_thread_info(tsk), flag);
2038}
2039
2040static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2041{
2042 clear_ti_thread_flag(task_thread_info(tsk), flag);
2043}
2044
2045static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2046 bool value)
2047{
2048 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2049}
2050
2051static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2052{
2053 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2054}
2055
2056static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2057{
2058 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2059}
2060
2061static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2062{
2063 return test_ti_thread_flag(task_thread_info(tsk), flag);
2064}
2065
2066static inline void set_tsk_need_resched(struct task_struct *tsk)
2067{
2068 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2069}
2070
2071static inline void clear_tsk_need_resched(struct task_struct *tsk)
2072{
2073 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2074 v: (atomic_long_t *)&task_thread_info(tsk)->flags);
2075}
2076
2077static inline int test_tsk_need_resched(struct task_struct *tsk)
2078{
2079 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2080}
2081
2082/*
2083 * cond_resched() and cond_resched_lock(): latency reduction via
2084 * explicit rescheduling in places that are safe. The return
2085 * value indicates whether a reschedule was done in fact.
2086 * cond_resched_lock() will drop the spinlock before scheduling,
2087 */
2088#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2089extern int __cond_resched(void);
2090
2091#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2092
2093DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2094
2095static __always_inline int _cond_resched(void)
2096{
2097 return static_call_mod(cond_resched)();
2098}
2099
2100#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2101
2102extern int dynamic_cond_resched(void);
2103
2104static __always_inline int _cond_resched(void)
2105{
2106 return dynamic_cond_resched();
2107}
2108
2109#else /* !CONFIG_PREEMPTION */
2110
2111static inline int _cond_resched(void)
2112{
2113 return __cond_resched();
2114}
2115
2116#endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2117
2118#else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2119
2120static inline int _cond_resched(void)
2121{
2122 return 0;
2123}
2124
2125#endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2126
2127#define cond_resched() ({ \
2128 __might_resched(__FILE__, __LINE__, 0); \
2129 _cond_resched(); \
2130})
2131
2132extern int __cond_resched_lock(spinlock_t *lock);
2133extern int __cond_resched_rwlock_read(rwlock_t *lock);
2134extern int __cond_resched_rwlock_write(rwlock_t *lock);
2135
2136#define MIGHT_RESCHED_RCU_SHIFT 8
2137#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2138
2139#ifndef CONFIG_PREEMPT_RT
2140/*
2141 * Non RT kernels have an elevated preempt count due to the held lock,
2142 * but are not allowed to be inside a RCU read side critical section
2143 */
2144# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2145#else
2146/*
2147 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2148 * cond_resched*lock() has to take that into account because it checks for
2149 * preempt_count() and rcu_preempt_depth().
2150 */
2151# define PREEMPT_LOCK_RESCHED_OFFSETS \
2152 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2153#endif
2154
2155#define cond_resched_lock(lock) ({ \
2156 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2157 __cond_resched_lock(lock); \
2158})
2159
2160#define cond_resched_rwlock_read(lock) ({ \
2161 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2162 __cond_resched_rwlock_read(lock); \
2163})
2164
2165#define cond_resched_rwlock_write(lock) ({ \
2166 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2167 __cond_resched_rwlock_write(lock); \
2168})
2169
2170static __always_inline bool need_resched(void)
2171{
2172 return unlikely(tif_need_resched());
2173}
2174
2175/*
2176 * Wrappers for p->thread_info->cpu access. No-op on UP.
2177 */
2178#ifdef CONFIG_SMP
2179
2180static inline unsigned int task_cpu(const struct task_struct *p)
2181{
2182 return READ_ONCE(task_thread_info(p)->cpu);
2183}
2184
2185extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2186
2187#else
2188
2189static inline unsigned int task_cpu(const struct task_struct *p)
2190{
2191 return 0;
2192}
2193
2194static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2195{
2196}
2197
2198#endif /* CONFIG_SMP */
2199
2200static inline bool task_is_runnable(struct task_struct *p)
2201{
2202 return p->on_rq && !p->se.sched_delayed;
2203}
2204
2205extern bool sched_task_on_rq(struct task_struct *p);
2206extern unsigned long get_wchan(struct task_struct *p);
2207extern struct task_struct *cpu_curr_snapshot(int cpu);
2208
2209#include <linux/spinlock.h>
2210
2211/*
2212 * In order to reduce various lock holder preemption latencies provide an
2213 * interface to see if a vCPU is currently running or not.
2214 *
2215 * This allows us to terminate optimistic spin loops and block, analogous to
2216 * the native optimistic spin heuristic of testing if the lock owner task is
2217 * running or not.
2218 */
2219#ifndef vcpu_is_preempted
2220static inline bool vcpu_is_preempted(int cpu)
2221{
2222 return false;
2223}
2224#endif
2225
2226extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2227extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2228
2229#ifndef TASK_SIZE_OF
2230#define TASK_SIZE_OF(tsk) TASK_SIZE
2231#endif
2232
2233#ifdef CONFIG_SMP
2234static inline bool owner_on_cpu(struct task_struct *owner)
2235{
2236 /*
2237 * As lock holder preemption issue, we both skip spinning if
2238 * task is not on cpu or its cpu is preempted
2239 */
2240 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(cpu: task_cpu(p: owner));
2241}
2242
2243/* Returns effective CPU energy utilization, as seen by the scheduler */
2244unsigned long sched_cpu_util(int cpu);
2245#endif /* CONFIG_SMP */
2246
2247#ifdef CONFIG_SCHED_CORE
2248extern void sched_core_free(struct task_struct *tsk);
2249extern void sched_core_fork(struct task_struct *p);
2250extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2251 unsigned long uaddr);
2252extern int sched_core_idle_cpu(int cpu);
2253#else
2254static inline void sched_core_free(struct task_struct *tsk) { }
2255static inline void sched_core_fork(struct task_struct *p) { }
2256static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2257#endif
2258
2259extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2260
2261#ifdef CONFIG_MEM_ALLOC_PROFILING
2262static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2263{
2264 swap(current->alloc_tag, tag);
2265 return tag;
2266}
2267
2268static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2269{
2270#ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2271 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2272#endif
2273 current->alloc_tag = old;
2274}
2275#else
2276#define alloc_tag_save(_tag) NULL
2277#define alloc_tag_restore(_tag, _old) do {} while (0)
2278#endif
2279
2280#endif
2281

source code of linux/include/linux/sched.h