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/spinlock.h>
38#include <linux/syscall_user_dispatch_types.h>
39#include <linux/mm_types_task.h>
40#include <linux/netdevice_xmit.h>
41#include <linux/task_io_accounting.h>
42#include <linux/posix-timers_types.h>
43#include <linux/restart_block.h>
44#include <uapi/linux/rseq.h>
45#include <linux/seqlock_types.h>
46#include <linux/kcsan.h>
47#include <linux/rv.h>
48#include <linux/uidgid_types.h>
49#include <linux/tracepoint-defs.h>
50#include <linux/unwind_deferred_types.h>
51#include <asm/kmap_size.h>
52#ifndef COMPILE_OFFSETS
53#include <generated/rq-offsets.h>
54#endif
55
56/* task_struct member predeclarations (sorted alphabetically): */
57struct audit_context;
58struct bio_list;
59struct blk_plug;
60struct bpf_local_storage;
61struct bpf_run_ctx;
62struct bpf_net_context;
63struct capture_control;
64struct cfs_rq;
65struct fs_struct;
66struct futex_pi_state;
67struct io_context;
68struct io_uring_task;
69struct mempolicy;
70struct nameidata;
71struct nsproxy;
72struct perf_event_context;
73struct perf_ctx_data;
74struct pid_namespace;
75struct pipe_inode_info;
76struct rcu_node;
77struct reclaim_state;
78struct robust_list_head;
79struct root_domain;
80struct rq;
81struct sched_attr;
82struct sched_dl_entity;
83struct seq_file;
84struct sighand_struct;
85struct signal_struct;
86struct task_delay_info;
87struct task_group;
88struct task_struct;
89struct user_event_mm;
90
91#include <linux/sched/ext.h>
92
93/*
94 * Task state bitmask. NOTE! These bits are also
95 * encoded in fs/proc/array.c: get_task_state().
96 *
97 * We have two separate sets of flags: task->__state
98 * is about runnability, while task->exit_state are
99 * about the task exiting. Confusing, but this way
100 * modifying one set can't modify the other one by
101 * mistake.
102 */
103
104/* Used in tsk->__state: */
105#define TASK_RUNNING 0x00000000
106#define TASK_INTERRUPTIBLE 0x00000001
107#define TASK_UNINTERRUPTIBLE 0x00000002
108#define __TASK_STOPPED 0x00000004
109#define __TASK_TRACED 0x00000008
110/* Used in tsk->exit_state: */
111#define EXIT_DEAD 0x00000010
112#define EXIT_ZOMBIE 0x00000020
113#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
114/* Used in tsk->__state again: */
115#define TASK_PARKED 0x00000040
116#define TASK_DEAD 0x00000080
117#define TASK_WAKEKILL 0x00000100
118#define TASK_WAKING 0x00000200
119#define TASK_NOLOAD 0x00000400
120#define TASK_NEW 0x00000800
121#define TASK_RTLOCK_WAIT 0x00001000
122#define TASK_FREEZABLE 0x00002000
123#define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
124#define TASK_FROZEN 0x00008000
125#define TASK_STATE_MAX 0x00010000
126
127#define TASK_ANY (TASK_STATE_MAX-1)
128
129/*
130 * DO NOT ADD ANY NEW USERS !
131 */
132#define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
133
134/* Convenience macros for the sake of set_current_state: */
135#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
136#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
137#define TASK_TRACED __TASK_TRACED
138
139#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
140
141/* Convenience macros for the sake of wake_up(): */
142#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
143
144/* get_task_state(): */
145#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
146 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
147 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
148 TASK_PARKED)
149
150#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
151
152#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
153#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
154#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
155
156/*
157 * Special states are those that do not use the normal wait-loop pattern. See
158 * the comment with set_special_state().
159 */
160#define is_special_task_state(state) \
161 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \
162 TASK_DEAD | TASK_FROZEN))
163
164#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
165# define debug_normal_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_special_state_change(state_value) \
172 do { \
173 WARN_ON_ONCE(!is_special_task_state(state_value)); \
174 current->task_state_change = _THIS_IP_; \
175 } while (0)
176
177# define debug_rtlock_wait_set_state() \
178 do { \
179 current->saved_state_change = current->task_state_change;\
180 current->task_state_change = _THIS_IP_; \
181 } while (0)
182
183# define debug_rtlock_wait_restore_state() \
184 do { \
185 current->task_state_change = current->saved_state_change;\
186 } while (0)
187
188#else
189# define debug_normal_state_change(cond) do { } while (0)
190# define debug_special_state_change(cond) do { } while (0)
191# define debug_rtlock_wait_set_state() do { } while (0)
192# define debug_rtlock_wait_restore_state() do { } while (0)
193#endif
194
195#define trace_set_current_state(state_value) \
196 do { \
197 if (tracepoint_enabled(sched_set_state_tp)) \
198 __trace_set_current_state(state_value); \
199 } while (0)
200
201/*
202 * set_current_state() includes a barrier so that the write of current->__state
203 * is correctly serialised wrt the caller's subsequent test of whether to
204 * actually sleep:
205 *
206 * for (;;) {
207 * set_current_state(TASK_UNINTERRUPTIBLE);
208 * if (CONDITION)
209 * break;
210 *
211 * schedule();
212 * }
213 * __set_current_state(TASK_RUNNING);
214 *
215 * If the caller does not need such serialisation (because, for instance, the
216 * CONDITION test and condition change and wakeup are under the same lock) then
217 * use __set_current_state().
218 *
219 * The above is typically ordered against the wakeup, which does:
220 *
221 * CONDITION = 1;
222 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
223 *
224 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
225 * accessing p->__state.
226 *
227 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
228 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
229 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
230 *
231 * However, with slightly different timing the wakeup TASK_RUNNING store can
232 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
233 * a problem either because that will result in one extra go around the loop
234 * and our @cond test will save the day.
235 *
236 * Also see the comments of try_to_wake_up().
237 */
238#define __set_current_state(state_value) \
239 do { \
240 debug_normal_state_change((state_value)); \
241 trace_set_current_state(state_value); \
242 WRITE_ONCE(current->__state, (state_value)); \
243 } while (0)
244
245#define set_current_state(state_value) \
246 do { \
247 debug_normal_state_change((state_value)); \
248 trace_set_current_state(state_value); \
249 smp_store_mb(current->__state, (state_value)); \
250 } while (0)
251
252/*
253 * set_special_state() should be used for those states when the blocking task
254 * can not use the regular condition based wait-loop. In that case we must
255 * serialize against wakeups such that any possible in-flight TASK_RUNNING
256 * stores will not collide with our state change.
257 */
258#define set_special_state(state_value) \
259 do { \
260 unsigned long flags; /* may shadow */ \
261 \
262 raw_spin_lock_irqsave(&current->pi_lock, flags); \
263 debug_special_state_change((state_value)); \
264 trace_set_current_state(state_value); \
265 WRITE_ONCE(current->__state, (state_value)); \
266 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
267 } while (0)
268
269/*
270 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
271 *
272 * RT's spin/rwlock substitutions are state preserving. The state of the
273 * task when blocking on the lock is saved in task_struct::saved_state and
274 * restored after the lock has been acquired. These operations are
275 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
276 * lock related wakeups while the task is blocked on the lock are
277 * redirected to operate on task_struct::saved_state to ensure that these
278 * are not dropped. On restore task_struct::saved_state is set to
279 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
280 *
281 * The lock operation looks like this:
282 *
283 * current_save_and_set_rtlock_wait_state();
284 * for (;;) {
285 * if (try_lock())
286 * break;
287 * raw_spin_unlock_irq(&lock->wait_lock);
288 * schedule_rtlock();
289 * raw_spin_lock_irq(&lock->wait_lock);
290 * set_current_state(TASK_RTLOCK_WAIT);
291 * }
292 * current_restore_rtlock_saved_state();
293 */
294#define current_save_and_set_rtlock_wait_state() \
295 do { \
296 lockdep_assert_irqs_disabled(); \
297 raw_spin_lock(&current->pi_lock); \
298 current->saved_state = current->__state; \
299 debug_rtlock_wait_set_state(); \
300 trace_set_current_state(TASK_RTLOCK_WAIT); \
301 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
302 raw_spin_unlock(&current->pi_lock); \
303 } while (0);
304
305#define current_restore_rtlock_saved_state() \
306 do { \
307 lockdep_assert_irqs_disabled(); \
308 raw_spin_lock(&current->pi_lock); \
309 debug_rtlock_wait_restore_state(); \
310 trace_set_current_state(current->saved_state); \
311 WRITE_ONCE(current->__state, current->saved_state); \
312 current->saved_state = TASK_RUNNING; \
313 raw_spin_unlock(&current->pi_lock); \
314 } while (0);
315
316#define get_current_state() READ_ONCE(current->__state)
317
318/*
319 * Define the task command name length as enum, then it can be visible to
320 * BPF programs.
321 */
322enum {
323 TASK_COMM_LEN = 16,
324};
325
326extern void sched_tick(void);
327
328#define MAX_SCHEDULE_TIMEOUT LONG_MAX
329
330extern long schedule_timeout(long timeout);
331extern long schedule_timeout_interruptible(long timeout);
332extern long schedule_timeout_killable(long timeout);
333extern long schedule_timeout_uninterruptible(long timeout);
334extern long schedule_timeout_idle(long timeout);
335asmlinkage void schedule(void);
336extern void schedule_preempt_disabled(void);
337asmlinkage void preempt_schedule_irq(void);
338#ifdef CONFIG_PREEMPT_RT
339 extern void schedule_rtlock(void);
340#endif
341
342extern int __must_check io_schedule_prepare(void);
343extern void io_schedule_finish(int token);
344extern long io_schedule_timeout(long timeout);
345extern void io_schedule(void);
346
347/* wrapper functions to trace from this header file */
348DECLARE_TRACEPOINT(sched_set_state_tp);
349extern void __trace_set_current_state(int state_value);
350DECLARE_TRACEPOINT(sched_set_need_resched_tp);
351extern void __trace_set_need_resched(struct task_struct *curr, int tif);
352
353/**
354 * struct prev_cputime - snapshot of system and user cputime
355 * @utime: time spent in user mode
356 * @stime: time spent in system mode
357 * @lock: protects the above two fields
358 *
359 * Stores previous user/system time values such that we can guarantee
360 * monotonicity.
361 */
362struct prev_cputime {
363#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
364 u64 utime;
365 u64 stime;
366 raw_spinlock_t lock;
367#endif
368};
369
370enum vtime_state {
371 /* Task is sleeping or running in a CPU with VTIME inactive: */
372 VTIME_INACTIVE = 0,
373 /* Task is idle */
374 VTIME_IDLE,
375 /* Task runs in kernelspace in a CPU with VTIME active: */
376 VTIME_SYS,
377 /* Task runs in userspace in a CPU with VTIME active: */
378 VTIME_USER,
379 /* Task runs as guests in a CPU with VTIME active: */
380 VTIME_GUEST,
381};
382
383struct vtime {
384 seqcount_t seqcount;
385 unsigned long long starttime;
386 enum vtime_state state;
387 unsigned int cpu;
388 u64 utime;
389 u64 stime;
390 u64 gtime;
391};
392
393/*
394 * Utilization clamp constraints.
395 * @UCLAMP_MIN: Minimum utilization
396 * @UCLAMP_MAX: Maximum utilization
397 * @UCLAMP_CNT: Utilization clamp constraints count
398 */
399enum uclamp_id {
400 UCLAMP_MIN = 0,
401 UCLAMP_MAX,
402 UCLAMP_CNT
403};
404
405extern struct root_domain def_root_domain;
406extern struct mutex sched_domains_mutex;
407extern void sched_domains_mutex_lock(void);
408extern void sched_domains_mutex_unlock(void);
409
410struct sched_param {
411 int sched_priority;
412};
413
414struct sched_info {
415#ifdef CONFIG_SCHED_INFO
416 /* Cumulative counters: */
417
418 /* # of times we have run on this CPU: */
419 unsigned long pcount;
420
421 /* Time spent waiting on a runqueue: */
422 unsigned long long run_delay;
423
424 /* Max time spent waiting on a runqueue: */
425 unsigned long long max_run_delay;
426
427 /* Min time spent waiting on a runqueue: */
428 unsigned long long min_run_delay;
429
430 /* Timestamps: */
431
432 /* When did we last run on a CPU? */
433 unsigned long long last_arrival;
434
435 /* When were we last queued to run? */
436 unsigned long long last_queued;
437
438#endif /* CONFIG_SCHED_INFO */
439};
440
441/*
442 * Integer metrics need fixed point arithmetic, e.g., sched/fair
443 * has a few: load, load_avg, util_avg, freq, and capacity.
444 *
445 * We define a basic fixed point arithmetic range, and then formalize
446 * all these metrics based on that basic range.
447 */
448# define SCHED_FIXEDPOINT_SHIFT 10
449# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
450
451/* Increase resolution of cpu_capacity calculations */
452# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
453# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
454
455struct load_weight {
456 unsigned long weight;
457 u32 inv_weight;
458};
459
460/*
461 * The load/runnable/util_avg accumulates an infinite geometric series
462 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
463 *
464 * [load_avg definition]
465 *
466 * load_avg = runnable% * scale_load_down(load)
467 *
468 * [runnable_avg definition]
469 *
470 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
471 *
472 * [util_avg definition]
473 *
474 * util_avg = running% * SCHED_CAPACITY_SCALE
475 *
476 * where runnable% is the time ratio that a sched_entity is runnable and
477 * running% the time ratio that a sched_entity is running.
478 *
479 * For cfs_rq, they are the aggregated values of all runnable and blocked
480 * sched_entities.
481 *
482 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
483 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
484 * for computing those signals (see update_rq_clock_pelt())
485 *
486 * N.B., the above ratios (runnable% and running%) themselves are in the
487 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
488 * to as large a range as necessary. This is for example reflected by
489 * util_avg's SCHED_CAPACITY_SCALE.
490 *
491 * [Overflow issue]
492 *
493 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
494 * with the highest load (=88761), always runnable on a single cfs_rq,
495 * and should not overflow as the number already hits PID_MAX_LIMIT.
496 *
497 * For all other cases (including 32-bit kernels), struct load_weight's
498 * weight will overflow first before we do, because:
499 *
500 * Max(load_avg) <= Max(load.weight)
501 *
502 * Then it is the load_weight's responsibility to consider overflow
503 * issues.
504 */
505struct sched_avg {
506 u64 last_update_time;
507 u64 load_sum;
508 u64 runnable_sum;
509 u32 util_sum;
510 u32 period_contrib;
511 unsigned long load_avg;
512 unsigned long runnable_avg;
513 unsigned long util_avg;
514 unsigned int util_est;
515} ____cacheline_aligned;
516
517/*
518 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
519 * updates. When a task is dequeued, its util_est should not be updated if its
520 * util_avg has not been updated in the meantime.
521 * This information is mapped into the MSB bit of util_est at dequeue time.
522 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
523 * it is safe to use MSB.
524 */
525#define UTIL_EST_WEIGHT_SHIFT 2
526#define UTIL_AVG_UNCHANGED 0x80000000
527
528struct sched_statistics {
529#ifdef CONFIG_SCHEDSTATS
530 u64 wait_start;
531 u64 wait_max;
532 u64 wait_count;
533 u64 wait_sum;
534 u64 iowait_count;
535 u64 iowait_sum;
536
537 u64 sleep_start;
538 u64 sleep_max;
539 s64 sum_sleep_runtime;
540
541 u64 block_start;
542 u64 block_max;
543 s64 sum_block_runtime;
544
545 s64 exec_max;
546 u64 slice_max;
547
548 u64 nr_migrations_cold;
549 u64 nr_failed_migrations_affine;
550 u64 nr_failed_migrations_running;
551 u64 nr_failed_migrations_hot;
552 u64 nr_forced_migrations;
553
554 u64 nr_wakeups;
555 u64 nr_wakeups_sync;
556 u64 nr_wakeups_migrate;
557 u64 nr_wakeups_local;
558 u64 nr_wakeups_remote;
559 u64 nr_wakeups_affine;
560 u64 nr_wakeups_affine_attempts;
561 u64 nr_wakeups_passive;
562 u64 nr_wakeups_idle;
563
564#ifdef CONFIG_SCHED_CORE
565 u64 core_forceidle_sum;
566#endif
567#endif /* CONFIG_SCHEDSTATS */
568} ____cacheline_aligned;
569
570struct sched_entity {
571 /* For load-balancing: */
572 struct load_weight load;
573 struct rb_node run_node;
574 u64 deadline;
575 u64 min_vruntime;
576 u64 min_slice;
577
578 struct list_head group_node;
579 unsigned char on_rq;
580 unsigned char sched_delayed;
581 unsigned char rel_deadline;
582 unsigned char custom_slice;
583 /* hole */
584
585 u64 exec_start;
586 u64 sum_exec_runtime;
587 u64 prev_sum_exec_runtime;
588 u64 vruntime;
589 union {
590 /*
591 * When !@on_rq this field is vlag.
592 * When cfs_rq->curr == se (which implies @on_rq)
593 * this field is vprot. See protect_slice().
594 */
595 s64 vlag;
596 u64 vprot;
597 };
598 u64 slice;
599
600 u64 nr_migrations;
601
602#ifdef CONFIG_FAIR_GROUP_SCHED
603 int depth;
604 struct sched_entity *parent;
605 /* rq on which this entity is (to be) queued: */
606 struct cfs_rq *cfs_rq;
607 /* rq "owned" by this entity/group: */
608 struct cfs_rq *my_q;
609 /* cached value of my_q->h_nr_running */
610 unsigned long runnable_weight;
611#endif
612
613 /*
614 * Per entity load average tracking.
615 *
616 * Put into separate cache line so it does not
617 * collide with read-mostly values above.
618 */
619 struct sched_avg avg;
620};
621
622struct sched_rt_entity {
623 struct list_head run_list;
624 unsigned long timeout;
625 unsigned long watchdog_stamp;
626 unsigned int time_slice;
627 unsigned short on_rq;
628 unsigned short on_list;
629
630 struct sched_rt_entity *back;
631#ifdef CONFIG_RT_GROUP_SCHED
632 struct sched_rt_entity *parent;
633 /* rq on which this entity is (to be) queued: */
634 struct rt_rq *rt_rq;
635 /* rq "owned" by this entity/group: */
636 struct rt_rq *my_q;
637#endif
638} __randomize_layout;
639
640typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
641typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
642
643struct sched_dl_entity {
644 struct rb_node rb_node;
645
646 /*
647 * Original scheduling parameters. Copied here from sched_attr
648 * during sched_setattr(), they will remain the same until
649 * the next sched_setattr().
650 */
651 u64 dl_runtime; /* Maximum runtime for each instance */
652 u64 dl_deadline; /* Relative deadline of each instance */
653 u64 dl_period; /* Separation of two instances (period) */
654 u64 dl_bw; /* dl_runtime / dl_period */
655 u64 dl_density; /* dl_runtime / dl_deadline */
656
657 /*
658 * Actual scheduling parameters. Initialized with the values above,
659 * they are continuously updated during task execution. Note that
660 * the remaining runtime could be < 0 in case we are in overrun.
661 */
662 s64 runtime; /* Remaining runtime for this instance */
663 u64 deadline; /* Absolute deadline for this instance */
664 unsigned int flags; /* Specifying the scheduler behaviour */
665
666 /*
667 * Some bool flags:
668 *
669 * @dl_throttled tells if we exhausted the runtime. If so, the
670 * task has to wait for a replenishment to be performed at the
671 * next firing of dl_timer.
672 *
673 * @dl_yielded tells if task gave up the CPU before consuming
674 * all its available runtime during the last job.
675 *
676 * @dl_non_contending tells if the task is inactive while still
677 * contributing to the active utilization. In other words, it
678 * indicates if the inactive timer has been armed and its handler
679 * has not been executed yet. This flag is useful to avoid race
680 * conditions between the inactive timer handler and the wakeup
681 * code.
682 *
683 * @dl_overrun tells if the task asked to be informed about runtime
684 * overruns.
685 *
686 * @dl_server tells if this is a server entity.
687 *
688 * @dl_defer tells if this is a deferred or regular server. For
689 * now only defer server exists.
690 *
691 * @dl_defer_armed tells if the deferrable server is waiting
692 * for the replenishment timer to activate it.
693 *
694 * @dl_server_active tells if the dlserver is active(started).
695 * dlserver is started on first cfs enqueue on an idle runqueue
696 * and is stopped when a dequeue results in 0 cfs tasks on the
697 * runqueue. In other words, dlserver is active only when cpu's
698 * runqueue has atleast one cfs task.
699 *
700 * @dl_defer_running tells if the deferrable server is actually
701 * running, skipping the defer phase.
702 */
703 unsigned int dl_throttled : 1;
704 unsigned int dl_yielded : 1;
705 unsigned int dl_non_contending : 1;
706 unsigned int dl_overrun : 1;
707 unsigned int dl_server : 1;
708 unsigned int dl_server_active : 1;
709 unsigned int dl_defer : 1;
710 unsigned int dl_defer_armed : 1;
711 unsigned int dl_defer_running : 1;
712
713 /*
714 * Bandwidth enforcement timer. Each -deadline task has its
715 * own bandwidth to be enforced, thus we need one timer per task.
716 */
717 struct hrtimer dl_timer;
718
719 /*
720 * Inactive timer, responsible for decreasing the active utilization
721 * at the "0-lag time". When a -deadline task blocks, it contributes
722 * to GRUB's active utilization until the "0-lag time", hence a
723 * timer is needed to decrease the active utilization at the correct
724 * time.
725 */
726 struct hrtimer inactive_timer;
727
728 /*
729 * Bits for DL-server functionality. Also see the comment near
730 * dl_server_update().
731 *
732 * @rq the runqueue this server is for
733 *
734 * @server_has_tasks() returns true if @server_pick return a
735 * runnable task.
736 */
737 struct rq *rq;
738 dl_server_pick_f server_pick_task;
739
740#ifdef CONFIG_RT_MUTEXES
741 /*
742 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
743 * pi_se points to the donor, otherwise points to the dl_se it belongs
744 * to (the original one/itself).
745 */
746 struct sched_dl_entity *pi_se;
747#endif
748};
749
750#ifdef CONFIG_UCLAMP_TASK
751/* Number of utilization clamp buckets (shorter alias) */
752#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
753
754/*
755 * Utilization clamp for a scheduling entity
756 * @value: clamp value "assigned" to a se
757 * @bucket_id: bucket index corresponding to the "assigned" value
758 * @active: the se is currently refcounted in a rq's bucket
759 * @user_defined: the requested clamp value comes from user-space
760 *
761 * The bucket_id is the index of the clamp bucket matching the clamp value
762 * which is pre-computed and stored to avoid expensive integer divisions from
763 * the fast path.
764 *
765 * The active bit is set whenever a task has got an "effective" value assigned,
766 * which can be different from the clamp value "requested" from user-space.
767 * This allows to know a task is refcounted in the rq's bucket corresponding
768 * to the "effective" bucket_id.
769 *
770 * The user_defined bit is set whenever a task has got a task-specific clamp
771 * value requested from userspace, i.e. the system defaults apply to this task
772 * just as a restriction. This allows to relax default clamps when a less
773 * restrictive task-specific value has been requested, thus allowing to
774 * implement a "nice" semantic. For example, a task running with a 20%
775 * default boost can still drop its own boosting to 0%.
776 */
777struct uclamp_se {
778 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
779 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
780 unsigned int active : 1;
781 unsigned int user_defined : 1;
782};
783#endif /* CONFIG_UCLAMP_TASK */
784
785union rcu_special {
786 struct {
787 u8 blocked;
788 u8 need_qs;
789 u8 exp_hint; /* Hint for performance. */
790 u8 need_mb; /* Readers need smp_mb(). */
791 } b; /* Bits. */
792 u32 s; /* Set of bits. */
793};
794
795enum perf_event_task_context {
796 perf_invalid_context = -1,
797 perf_hw_context = 0,
798 perf_sw_context,
799 perf_nr_task_contexts,
800};
801
802/*
803 * Number of contexts where an event can trigger:
804 * task, softirq, hardirq, nmi.
805 */
806#define PERF_NR_CONTEXTS 4
807
808struct wake_q_node {
809 struct wake_q_node *next;
810};
811
812struct kmap_ctrl {
813#ifdef CONFIG_KMAP_LOCAL
814 int idx;
815 pte_t pteval[KM_MAX_IDX];
816#endif
817};
818
819struct task_struct {
820#ifdef CONFIG_THREAD_INFO_IN_TASK
821 /*
822 * For reasons of header soup (see current_thread_info()), this
823 * must be the first element of task_struct.
824 */
825 struct thread_info thread_info;
826#endif
827 unsigned int __state;
828
829 /* saved state for "spinlock sleepers" */
830 unsigned int saved_state;
831
832 /*
833 * This begins the randomizable portion of task_struct. Only
834 * scheduling-critical items should be added above here.
835 */
836 randomized_struct_fields_start
837
838 void *stack;
839 refcount_t usage;
840 /* Per task flags (PF_*), defined further below: */
841 unsigned int flags;
842 unsigned int ptrace;
843
844#ifdef CONFIG_MEM_ALLOC_PROFILING
845 struct alloc_tag *alloc_tag;
846#endif
847
848 int on_cpu;
849 struct __call_single_node wake_entry;
850 unsigned int wakee_flips;
851 unsigned long wakee_flip_decay_ts;
852 struct task_struct *last_wakee;
853
854 /*
855 * recent_used_cpu is initially set as the last CPU used by a task
856 * that wakes affine another task. Waker/wakee relationships can
857 * push tasks around a CPU where each wakeup moves to the next one.
858 * Tracking a recently used CPU allows a quick search for a recently
859 * used CPU that may be idle.
860 */
861 int recent_used_cpu;
862 int wake_cpu;
863 int on_rq;
864
865 int prio;
866 int static_prio;
867 int normal_prio;
868 unsigned int rt_priority;
869
870 struct sched_entity se;
871 struct sched_rt_entity rt;
872 struct sched_dl_entity dl;
873 struct sched_dl_entity *dl_server;
874#ifdef CONFIG_SCHED_CLASS_EXT
875 struct sched_ext_entity scx;
876#endif
877 const struct sched_class *sched_class;
878
879#ifdef CONFIG_SCHED_CORE
880 struct rb_node core_node;
881 unsigned long core_cookie;
882 unsigned int core_occupation;
883#endif
884
885#ifdef CONFIG_CGROUP_SCHED
886 struct task_group *sched_task_group;
887#ifdef CONFIG_CFS_BANDWIDTH
888 struct callback_head sched_throttle_work;
889 struct list_head throttle_node;
890 bool throttled;
891#endif
892#endif
893
894
895#ifdef CONFIG_UCLAMP_TASK
896 /*
897 * Clamp values requested for a scheduling entity.
898 * Must be updated with task_rq_lock() held.
899 */
900 struct uclamp_se uclamp_req[UCLAMP_CNT];
901 /*
902 * Effective clamp values used for a scheduling entity.
903 * Must be updated with task_rq_lock() held.
904 */
905 struct uclamp_se uclamp[UCLAMP_CNT];
906#endif
907
908 struct sched_statistics stats;
909
910#ifdef CONFIG_PREEMPT_NOTIFIERS
911 /* List of struct preempt_notifier: */
912 struct hlist_head preempt_notifiers;
913#endif
914
915#ifdef CONFIG_BLK_DEV_IO_TRACE
916 unsigned int btrace_seq;
917#endif
918
919 unsigned int policy;
920 unsigned long max_allowed_capacity;
921 int nr_cpus_allowed;
922 const cpumask_t *cpus_ptr;
923 cpumask_t *user_cpus_ptr;
924 cpumask_t cpus_mask;
925 void *migration_pending;
926 unsigned short migration_disabled;
927 unsigned short migration_flags;
928
929#ifdef CONFIG_PREEMPT_RCU
930 int rcu_read_lock_nesting;
931 union rcu_special rcu_read_unlock_special;
932 struct list_head rcu_node_entry;
933 struct rcu_node *rcu_blocked_node;
934#endif /* #ifdef CONFIG_PREEMPT_RCU */
935
936#ifdef CONFIG_TASKS_RCU
937 unsigned long rcu_tasks_nvcsw;
938 u8 rcu_tasks_holdout;
939 u8 rcu_tasks_idx;
940 int rcu_tasks_idle_cpu;
941 struct list_head rcu_tasks_holdout_list;
942 int rcu_tasks_exit_cpu;
943 struct list_head rcu_tasks_exit_list;
944#endif /* #ifdef CONFIG_TASKS_RCU */
945
946#ifdef CONFIG_TASKS_TRACE_RCU
947 int trc_reader_nesting;
948 int trc_ipi_to_cpu;
949 union rcu_special trc_reader_special;
950 struct list_head trc_holdout_list;
951 struct list_head trc_blkd_node;
952 int trc_blkd_cpu;
953#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
954
955 struct sched_info sched_info;
956
957 struct list_head tasks;
958 struct plist_node pushable_tasks;
959 struct rb_node pushable_dl_tasks;
960
961 struct mm_struct *mm;
962 struct mm_struct *active_mm;
963 struct address_space *faults_disabled_mapping;
964
965 int exit_state;
966 int exit_code;
967 int exit_signal;
968 /* The signal sent when the parent dies: */
969 int pdeath_signal;
970 /* JOBCTL_*, siglock protected: */
971 unsigned long jobctl;
972
973 /* Used for emulating ABI behavior of previous Linux versions: */
974 unsigned int personality;
975
976 /* Scheduler bits, serialized by scheduler locks: */
977 unsigned sched_reset_on_fork:1;
978 unsigned sched_contributes_to_load:1;
979 unsigned sched_migrated:1;
980 unsigned sched_task_hot:1;
981
982 /* Force alignment to the next boundary: */
983 unsigned :0;
984
985 /* Unserialized, strictly 'current' */
986
987 /*
988 * This field must not be in the scheduler word above due to wakelist
989 * queueing no longer being serialized by p->on_cpu. However:
990 *
991 * p->XXX = X; ttwu()
992 * schedule() if (p->on_rq && ..) // false
993 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
994 * deactivate_task() ttwu_queue_wakelist())
995 * p->on_rq = 0; p->sched_remote_wakeup = Y;
996 *
997 * guarantees all stores of 'current' are visible before
998 * ->sched_remote_wakeup gets used, so it can be in this word.
999 */
1000 unsigned sched_remote_wakeup:1;
1001#ifdef CONFIG_RT_MUTEXES
1002 unsigned sched_rt_mutex:1;
1003#endif
1004
1005 /* Bit to tell TOMOYO we're in execve(): */
1006 unsigned in_execve:1;
1007 unsigned in_iowait:1;
1008#ifndef TIF_RESTORE_SIGMASK
1009 unsigned restore_sigmask:1;
1010#endif
1011#ifdef CONFIG_MEMCG_V1
1012 unsigned in_user_fault:1;
1013#endif
1014#ifdef CONFIG_LRU_GEN
1015 /* whether the LRU algorithm may apply to this access */
1016 unsigned in_lru_fault:1;
1017#endif
1018#ifdef CONFIG_COMPAT_BRK
1019 unsigned brk_randomized:1;
1020#endif
1021#ifdef CONFIG_CGROUPS
1022 /* disallow userland-initiated cgroup migration */
1023 unsigned no_cgroup_migration:1;
1024 /* task is frozen/stopped (used by the cgroup freezer) */
1025 unsigned frozen:1;
1026#endif
1027#ifdef CONFIG_BLK_CGROUP
1028 unsigned use_memdelay:1;
1029#endif
1030#ifdef CONFIG_PSI
1031 /* Stalled due to lack of memory */
1032 unsigned in_memstall:1;
1033#endif
1034#ifdef CONFIG_PAGE_OWNER
1035 /* Used by page_owner=on to detect recursion in page tracking. */
1036 unsigned in_page_owner:1;
1037#endif
1038#ifdef CONFIG_EVENTFD
1039 /* Recursion prevention for eventfd_signal() */
1040 unsigned in_eventfd:1;
1041#endif
1042#ifdef CONFIG_ARCH_HAS_CPU_PASID
1043 unsigned pasid_activated:1;
1044#endif
1045#ifdef CONFIG_X86_BUS_LOCK_DETECT
1046 unsigned reported_split_lock:1;
1047#endif
1048#ifdef CONFIG_TASK_DELAY_ACCT
1049 /* delay due to memory thrashing */
1050 unsigned in_thrashing:1;
1051#endif
1052 unsigned in_nf_duplicate:1;
1053#ifdef CONFIG_PREEMPT_RT
1054 struct netdev_xmit net_xmit;
1055#endif
1056 unsigned long atomic_flags; /* Flags requiring atomic access. */
1057
1058 struct restart_block restart_block;
1059
1060 pid_t pid;
1061 pid_t tgid;
1062
1063#ifdef CONFIG_STACKPROTECTOR
1064 /* Canary value for the -fstack-protector GCC feature: */
1065 unsigned long stack_canary;
1066#endif
1067 /*
1068 * Pointers to the (original) parent process, youngest child, younger sibling,
1069 * older sibling, respectively. (p->father can be replaced with
1070 * p->real_parent->pid)
1071 */
1072
1073 /* Real parent process: */
1074 struct task_struct __rcu *real_parent;
1075
1076 /* Recipient of SIGCHLD, wait4() reports: */
1077 struct task_struct __rcu *parent;
1078
1079 /*
1080 * Children/sibling form the list of natural children:
1081 */
1082 struct list_head children;
1083 struct list_head sibling;
1084 struct task_struct *group_leader;
1085
1086 /*
1087 * 'ptraced' is the list of tasks this task is using ptrace() on.
1088 *
1089 * This includes both natural children and PTRACE_ATTACH targets.
1090 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1091 */
1092 struct list_head ptraced;
1093 struct list_head ptrace_entry;
1094
1095 /* PID/PID hash table linkage. */
1096 struct pid *thread_pid;
1097 struct hlist_node pid_links[PIDTYPE_MAX];
1098 struct list_head thread_node;
1099
1100 struct completion *vfork_done;
1101
1102 /* CLONE_CHILD_SETTID: */
1103 int __user *set_child_tid;
1104
1105 /* CLONE_CHILD_CLEARTID: */
1106 int __user *clear_child_tid;
1107
1108 /* PF_KTHREAD | PF_IO_WORKER */
1109 void *worker_private;
1110
1111 u64 utime;
1112 u64 stime;
1113#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1114 u64 utimescaled;
1115 u64 stimescaled;
1116#endif
1117 u64 gtime;
1118 struct prev_cputime prev_cputime;
1119#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1120 struct vtime vtime;
1121#endif
1122
1123#ifdef CONFIG_NO_HZ_FULL
1124 atomic_t tick_dep_mask;
1125#endif
1126 /* Context switch counts: */
1127 unsigned long nvcsw;
1128 unsigned long nivcsw;
1129
1130 /* Monotonic time in nsecs: */
1131 u64 start_time;
1132
1133 /* Boot based time in nsecs: */
1134 u64 start_boottime;
1135
1136 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1137 unsigned long min_flt;
1138 unsigned long maj_flt;
1139
1140 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1141 struct posix_cputimers posix_cputimers;
1142
1143#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1144 struct posix_cputimers_work posix_cputimers_work;
1145#endif
1146
1147 /* Process credentials: */
1148
1149 /* Tracer's credentials at attach: */
1150 const struct cred __rcu *ptracer_cred;
1151
1152 /* Objective and real subjective task credentials (COW): */
1153 const struct cred __rcu *real_cred;
1154
1155 /* Effective (overridable) subjective task credentials (COW): */
1156 const struct cred __rcu *cred;
1157
1158#ifdef CONFIG_KEYS
1159 /* Cached requested key. */
1160 struct key *cached_requested_key;
1161#endif
1162
1163 /*
1164 * executable name, excluding path.
1165 *
1166 * - normally initialized begin_new_exec()
1167 * - set it with set_task_comm()
1168 * - strscpy_pad() to ensure it is always NUL-terminated and
1169 * zero-padded
1170 * - task_lock() to ensure the operation is atomic and the name is
1171 * fully updated.
1172 */
1173 char comm[TASK_COMM_LEN];
1174
1175 struct nameidata *nameidata;
1176
1177#ifdef CONFIG_SYSVIPC
1178 struct sysv_sem sysvsem;
1179 struct sysv_shm sysvshm;
1180#endif
1181#ifdef CONFIG_DETECT_HUNG_TASK
1182 unsigned long last_switch_count;
1183 unsigned long last_switch_time;
1184#endif
1185 /* Filesystem information: */
1186 struct fs_struct *fs;
1187
1188 /* Open file information: */
1189 struct files_struct *files;
1190
1191#ifdef CONFIG_IO_URING
1192 struct io_uring_task *io_uring;
1193#endif
1194
1195 /* Namespaces: */
1196 struct nsproxy *nsproxy;
1197
1198 /* Signal handlers: */
1199 struct signal_struct *signal;
1200 struct sighand_struct __rcu *sighand;
1201 sigset_t blocked;
1202 sigset_t real_blocked;
1203 /* Restored if set_restore_sigmask() was used: */
1204 sigset_t saved_sigmask;
1205 struct sigpending pending;
1206 unsigned long sas_ss_sp;
1207 size_t sas_ss_size;
1208 unsigned int sas_ss_flags;
1209
1210 struct callback_head *task_works;
1211
1212#ifdef CONFIG_AUDIT
1213#ifdef CONFIG_AUDITSYSCALL
1214 struct audit_context *audit_context;
1215#endif
1216 kuid_t loginuid;
1217 unsigned int sessionid;
1218#endif
1219 struct seccomp seccomp;
1220 struct syscall_user_dispatch syscall_dispatch;
1221
1222 /* Thread group tracking: */
1223 u64 parent_exec_id;
1224 u64 self_exec_id;
1225
1226 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1227 spinlock_t alloc_lock;
1228
1229 /* Protection of the PI data structures: */
1230 raw_spinlock_t pi_lock;
1231
1232 struct wake_q_node wake_q;
1233
1234#ifdef CONFIG_RT_MUTEXES
1235 /* PI waiters blocked on a rt_mutex held by this task: */
1236 struct rb_root_cached pi_waiters;
1237 /* Updated under owner's pi_lock and rq lock */
1238 struct task_struct *pi_top_task;
1239 /* Deadlock detection and priority inheritance handling: */
1240 struct rt_mutex_waiter *pi_blocked_on;
1241#endif
1242
1243 struct mutex *blocked_on; /* lock we're blocked on */
1244
1245#ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER
1246 /*
1247 * Encoded lock address causing task block (lower 2 bits = type from
1248 * <linux/hung_task.h>). Accessed via hung_task_*() helpers.
1249 */
1250 unsigned long blocker;
1251#endif
1252
1253#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1254 int non_block_count;
1255#endif
1256
1257#ifdef CONFIG_TRACE_IRQFLAGS
1258 struct irqtrace_events irqtrace;
1259 unsigned int hardirq_threaded;
1260 u64 hardirq_chain_key;
1261 int softirqs_enabled;
1262 int softirq_context;
1263 int irq_config;
1264#endif
1265#ifdef CONFIG_PREEMPT_RT
1266 int softirq_disable_cnt;
1267#endif
1268
1269#ifdef CONFIG_LOCKDEP
1270# define MAX_LOCK_DEPTH 48UL
1271 u64 curr_chain_key;
1272 int lockdep_depth;
1273 unsigned int lockdep_recursion;
1274 struct held_lock held_locks[MAX_LOCK_DEPTH];
1275#endif
1276
1277#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1278 unsigned int in_ubsan;
1279#endif
1280
1281 /* Journalling filesystem info: */
1282 void *journal_info;
1283
1284 /* Stacked block device info: */
1285 struct bio_list *bio_list;
1286
1287 /* Stack plugging: */
1288 struct blk_plug *plug;
1289
1290 /* VM state: */
1291 struct reclaim_state *reclaim_state;
1292
1293 struct io_context *io_context;
1294
1295#ifdef CONFIG_COMPACTION
1296 struct capture_control *capture_control;
1297#endif
1298 /* Ptrace state: */
1299 unsigned long ptrace_message;
1300 kernel_siginfo_t *last_siginfo;
1301
1302 struct task_io_accounting ioac;
1303#ifdef CONFIG_PSI
1304 /* Pressure stall state */
1305 unsigned int psi_flags;
1306#endif
1307#ifdef CONFIG_TASK_XACCT
1308 /* Accumulated RSS usage: */
1309 u64 acct_rss_mem1;
1310 /* Accumulated virtual memory usage: */
1311 u64 acct_vm_mem1;
1312 /* stime + utime since last update: */
1313 u64 acct_timexpd;
1314#endif
1315#ifdef CONFIG_CPUSETS
1316 /* Protected by ->alloc_lock: */
1317 nodemask_t mems_allowed;
1318 /* Sequence number to catch updates: */
1319 seqcount_spinlock_t mems_allowed_seq;
1320 int cpuset_mem_spread_rotor;
1321#endif
1322#ifdef CONFIG_CGROUPS
1323 /* Control Group info protected by css_set_lock: */
1324 struct css_set __rcu *cgroups;
1325 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1326 struct list_head cg_list;
1327#endif
1328#ifdef CONFIG_X86_CPU_RESCTRL
1329 u32 closid;
1330 u32 rmid;
1331#endif
1332#ifdef CONFIG_FUTEX
1333 struct robust_list_head __user *robust_list;
1334#ifdef CONFIG_COMPAT
1335 struct compat_robust_list_head __user *compat_robust_list;
1336#endif
1337 struct list_head pi_state_list;
1338 struct futex_pi_state *pi_state_cache;
1339 struct mutex futex_exit_mutex;
1340 unsigned int futex_state;
1341#endif
1342#ifdef CONFIG_PERF_EVENTS
1343 u8 perf_recursion[PERF_NR_CONTEXTS];
1344 struct perf_event_context *perf_event_ctxp;
1345 struct mutex perf_event_mutex;
1346 struct list_head perf_event_list;
1347 struct perf_ctx_data __rcu *perf_ctx_data;
1348#endif
1349#ifdef CONFIG_DEBUG_PREEMPT
1350 unsigned long preempt_disable_ip;
1351#endif
1352#ifdef CONFIG_NUMA
1353 /* Protected by alloc_lock: */
1354 struct mempolicy *mempolicy;
1355 short il_prev;
1356 u8 il_weight;
1357 short pref_node_fork;
1358#endif
1359#ifdef CONFIG_NUMA_BALANCING
1360 int numa_scan_seq;
1361 unsigned int numa_scan_period;
1362 unsigned int numa_scan_period_max;
1363 int numa_preferred_nid;
1364 unsigned long numa_migrate_retry;
1365 /* Migration stamp: */
1366 u64 node_stamp;
1367 u64 last_task_numa_placement;
1368 u64 last_sum_exec_runtime;
1369 struct callback_head numa_work;
1370
1371 /*
1372 * This pointer is only modified for current in syscall and
1373 * pagefault context (and for tasks being destroyed), so it can be read
1374 * from any of the following contexts:
1375 * - RCU read-side critical section
1376 * - current->numa_group from everywhere
1377 * - task's runqueue locked, task not running
1378 */
1379 struct numa_group __rcu *numa_group;
1380
1381 /*
1382 * numa_faults is an array split into four regions:
1383 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1384 * in this precise order.
1385 *
1386 * faults_memory: Exponential decaying average of faults on a per-node
1387 * basis. Scheduling placement decisions are made based on these
1388 * counts. The values remain static for the duration of a PTE scan.
1389 * faults_cpu: Track the nodes the process was running on when a NUMA
1390 * hinting fault was incurred.
1391 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1392 * during the current scan window. When the scan completes, the counts
1393 * in faults_memory and faults_cpu decay and these values are copied.
1394 */
1395 unsigned long *numa_faults;
1396 unsigned long total_numa_faults;
1397
1398 /*
1399 * numa_faults_locality tracks if faults recorded during the last
1400 * scan window were remote/local or failed to migrate. The task scan
1401 * period is adapted based on the locality of the faults with different
1402 * weights depending on whether they were shared or private faults
1403 */
1404 unsigned long numa_faults_locality[3];
1405
1406 unsigned long numa_pages_migrated;
1407#endif /* CONFIG_NUMA_BALANCING */
1408
1409#ifdef CONFIG_RSEQ
1410 struct rseq __user *rseq;
1411 u32 rseq_len;
1412 u32 rseq_sig;
1413 /*
1414 * RmW on rseq_event_mask must be performed atomically
1415 * with respect to preemption.
1416 */
1417 unsigned long rseq_event_mask;
1418# ifdef CONFIG_DEBUG_RSEQ
1419 /*
1420 * This is a place holder to save a copy of the rseq fields for
1421 * validation of read-only fields. The struct rseq has a
1422 * variable-length array at the end, so it cannot be used
1423 * directly. Reserve a size large enough for the known fields.
1424 */
1425 char rseq_fields[sizeof(struct rseq)];
1426# endif
1427#endif
1428
1429#ifdef CONFIG_SCHED_MM_CID
1430 int mm_cid; /* Current cid in mm */
1431 int last_mm_cid; /* Most recent cid in mm */
1432 int migrate_from_cpu;
1433 int mm_cid_active; /* Whether cid bitmap is active */
1434 struct callback_head cid_work;
1435#endif
1436
1437 struct tlbflush_unmap_batch tlb_ubc;
1438
1439 /* Cache last used pipe for splice(): */
1440 struct pipe_inode_info *splice_pipe;
1441
1442 struct page_frag task_frag;
1443
1444#ifdef CONFIG_TASK_DELAY_ACCT
1445 struct task_delay_info *delays;
1446#endif
1447
1448#ifdef CONFIG_FAULT_INJECTION
1449 int make_it_fail;
1450 unsigned int fail_nth;
1451#endif
1452 /*
1453 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1454 * balance_dirty_pages() for a dirty throttling pause:
1455 */
1456 int nr_dirtied;
1457 int nr_dirtied_pause;
1458 /* Start of a write-and-pause period: */
1459 unsigned long dirty_paused_when;
1460
1461#ifdef CONFIG_LATENCYTOP
1462 int latency_record_count;
1463 struct latency_record latency_record[LT_SAVECOUNT];
1464#endif
1465 /*
1466 * Time slack values; these are used to round up poll() and
1467 * select() etc timeout values. These are in nanoseconds.
1468 */
1469 u64 timer_slack_ns;
1470 u64 default_timer_slack_ns;
1471
1472#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1473 unsigned int kasan_depth;
1474#endif
1475
1476#ifdef CONFIG_KCSAN
1477 struct kcsan_ctx kcsan_ctx;
1478#ifdef CONFIG_TRACE_IRQFLAGS
1479 struct irqtrace_events kcsan_save_irqtrace;
1480#endif
1481#ifdef CONFIG_KCSAN_WEAK_MEMORY
1482 int kcsan_stack_depth;
1483#endif
1484#endif
1485
1486#ifdef CONFIG_KMSAN
1487 struct kmsan_ctx kmsan_ctx;
1488#endif
1489
1490#if IS_ENABLED(CONFIG_KUNIT)
1491 struct kunit *kunit_test;
1492#endif
1493
1494#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1495 /* Index of current stored address in ret_stack: */
1496 int curr_ret_stack;
1497 int curr_ret_depth;
1498
1499 /* Stack of return addresses for return function tracing: */
1500 unsigned long *ret_stack;
1501
1502 /* Timestamp for last schedule: */
1503 unsigned long long ftrace_timestamp;
1504 unsigned long long ftrace_sleeptime;
1505
1506 /*
1507 * Number of functions that haven't been traced
1508 * because of depth overrun:
1509 */
1510 atomic_t trace_overrun;
1511
1512 /* Pause tracing: */
1513 atomic_t tracing_graph_pause;
1514#endif
1515
1516#ifdef CONFIG_TRACING
1517 /* Bitmask and counter of trace recursion: */
1518 unsigned long trace_recursion;
1519#endif /* CONFIG_TRACING */
1520
1521#ifdef CONFIG_KCOV
1522 /* See kernel/kcov.c for more details. */
1523
1524 /* Coverage collection mode enabled for this task (0 if disabled): */
1525 unsigned int kcov_mode;
1526
1527 /* Size of the kcov_area: */
1528 unsigned int kcov_size;
1529
1530 /* Buffer for coverage collection: */
1531 void *kcov_area;
1532
1533 /* KCOV descriptor wired with this task or NULL: */
1534 struct kcov *kcov;
1535
1536 /* KCOV common handle for remote coverage collection: */
1537 u64 kcov_handle;
1538
1539 /* KCOV sequence number: */
1540 int kcov_sequence;
1541
1542 /* Collect coverage from softirq context: */
1543 unsigned int kcov_softirq;
1544#endif
1545
1546#ifdef CONFIG_MEMCG_V1
1547 struct mem_cgroup *memcg_in_oom;
1548#endif
1549
1550#ifdef CONFIG_MEMCG
1551 /* Number of pages to reclaim on returning to userland: */
1552 unsigned int memcg_nr_pages_over_high;
1553
1554 /* Used by memcontrol for targeted memcg charge: */
1555 struct mem_cgroup *active_memcg;
1556
1557 /* Cache for current->cgroups->memcg->objcg lookups: */
1558 struct obj_cgroup *objcg;
1559#endif
1560
1561#ifdef CONFIG_BLK_CGROUP
1562 struct gendisk *throttle_disk;
1563#endif
1564
1565#ifdef CONFIG_UPROBES
1566 struct uprobe_task *utask;
1567#endif
1568#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1569 unsigned int sequential_io;
1570 unsigned int sequential_io_avg;
1571#endif
1572 struct kmap_ctrl kmap_ctrl;
1573#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1574 unsigned long task_state_change;
1575# ifdef CONFIG_PREEMPT_RT
1576 unsigned long saved_state_change;
1577# endif
1578#endif
1579 struct rcu_head rcu;
1580 refcount_t rcu_users;
1581 int pagefault_disabled;
1582#ifdef CONFIG_MMU
1583 struct task_struct *oom_reaper_list;
1584 struct timer_list oom_reaper_timer;
1585#endif
1586#ifdef CONFIG_VMAP_STACK
1587 struct vm_struct *stack_vm_area;
1588#endif
1589#ifdef CONFIG_THREAD_INFO_IN_TASK
1590 /* A live task holds one reference: */
1591 refcount_t stack_refcount;
1592#endif
1593#ifdef CONFIG_LIVEPATCH
1594 int patch_state;
1595#endif
1596#ifdef CONFIG_SECURITY
1597 /* Used by LSM modules for access restriction: */
1598 void *security;
1599#endif
1600#ifdef CONFIG_BPF_SYSCALL
1601 /* Used by BPF task local storage */
1602 struct bpf_local_storage __rcu *bpf_storage;
1603 /* Used for BPF run context */
1604 struct bpf_run_ctx *bpf_ctx;
1605#endif
1606 /* Used by BPF for per-TASK xdp storage */
1607 struct bpf_net_context *bpf_net_context;
1608
1609#ifdef CONFIG_KSTACK_ERASE
1610 unsigned long lowest_stack;
1611#endif
1612#ifdef CONFIG_KSTACK_ERASE_METRICS
1613 unsigned long prev_lowest_stack;
1614#endif
1615
1616#ifdef CONFIG_X86_MCE
1617 void __user *mce_vaddr;
1618 __u64 mce_kflags;
1619 u64 mce_addr;
1620 __u64 mce_ripv : 1,
1621 mce_whole_page : 1,
1622 __mce_reserved : 62;
1623 struct callback_head mce_kill_me;
1624 int mce_count;
1625#endif
1626
1627#ifdef CONFIG_KRETPROBES
1628 struct llist_head kretprobe_instances;
1629#endif
1630#ifdef CONFIG_RETHOOK
1631 struct llist_head rethooks;
1632#endif
1633
1634#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1635 /*
1636 * If L1D flush is supported on mm context switch
1637 * then we use this callback head to queue kill work
1638 * to kill tasks that are not running on SMT disabled
1639 * cores
1640 */
1641 struct callback_head l1d_flush_kill;
1642#endif
1643
1644#ifdef CONFIG_RV
1645 /*
1646 * Per-task RV monitor, fixed in CONFIG_RV_PER_TASK_MONITORS.
1647 * If memory becomes a concern, we can think about a dynamic method.
1648 */
1649 union rv_task_monitor rv[CONFIG_RV_PER_TASK_MONITORS];
1650#endif
1651
1652#ifdef CONFIG_USER_EVENTS
1653 struct user_event_mm *user_event_mm;
1654#endif
1655
1656#ifdef CONFIG_UNWIND_USER
1657 struct unwind_task_info unwind_info;
1658#endif
1659
1660 /* CPU-specific state of this task: */
1661 struct thread_struct thread;
1662
1663 /*
1664 * New fields for task_struct should be added above here, so that
1665 * they are included in the randomized portion of task_struct.
1666 */
1667 randomized_struct_fields_end
1668} __attribute__ ((aligned (64)));
1669
1670#ifdef CONFIG_SCHED_PROXY_EXEC
1671DECLARE_STATIC_KEY_TRUE(__sched_proxy_exec);
1672static inline bool sched_proxy_exec(void)
1673{
1674 return static_branch_likely(&__sched_proxy_exec);
1675}
1676#else
1677static inline bool sched_proxy_exec(void)
1678{
1679 return false;
1680}
1681#endif
1682
1683#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1684#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1685
1686static inline unsigned int __task_state_index(unsigned int tsk_state,
1687 unsigned int tsk_exit_state)
1688{
1689 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1690
1691 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1692
1693 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1694 state = TASK_REPORT_IDLE;
1695
1696 /*
1697 * We're lying here, but rather than expose a completely new task state
1698 * to userspace, we can make this appear as if the task has gone through
1699 * a regular rt_mutex_lock() call.
1700 * Report frozen tasks as uninterruptible.
1701 */
1702 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
1703 state = TASK_UNINTERRUPTIBLE;
1704
1705 return fls(x: state);
1706}
1707
1708static inline unsigned int task_state_index(struct task_struct *tsk)
1709{
1710 return __task_state_index(READ_ONCE(tsk->__state), tsk_exit_state: tsk->exit_state);
1711}
1712
1713static inline char task_index_to_char(unsigned int state)
1714{
1715 static const char state_char[] = "RSDTtXZPI";
1716
1717 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1718
1719 return state_char[state];
1720}
1721
1722static inline char task_state_to_char(struct task_struct *tsk)
1723{
1724 return task_index_to_char(state: task_state_index(tsk));
1725}
1726
1727extern struct pid *cad_pid;
1728
1729/*
1730 * Per process flags
1731 */
1732#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1733#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1734#define PF_EXITING 0x00000004 /* Getting shut down */
1735#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1736#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1737#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1738#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1739#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1740#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1741#define PF_DUMPCORE 0x00000200 /* Dumped core */
1742#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1743#define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1744#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1745#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1746#define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1747#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1748#define PF_KCOMPACTD 0x00010000 /* I am kcompactd */
1749#define PF_KSWAPD 0x00020000 /* I am kswapd */
1750#define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1751#define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1752#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1753 * I am cleaning dirty pages from some other bdi. */
1754#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1755#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1756#define PF__HOLE__00800000 0x00800000
1757#define PF__HOLE__01000000 0x01000000
1758#define PF__HOLE__02000000 0x02000000
1759#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1760#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1761#define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1762 * See memalloc_pin_save() */
1763#define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1764#define PF__HOLE__40000000 0x40000000
1765#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1766
1767/*
1768 * Only the _current_ task can read/write to tsk->flags, but other
1769 * tasks can access tsk->flags in readonly mode for example
1770 * with tsk_used_math (like during threaded core dumping).
1771 * There is however an exception to this rule during ptrace
1772 * or during fork: the ptracer task is allowed to write to the
1773 * child->flags of its traced child (same goes for fork, the parent
1774 * can write to the child->flags), because we're guaranteed the
1775 * child is not running and in turn not changing child->flags
1776 * at the same time the parent does it.
1777 */
1778#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1779#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1780#define clear_used_math() clear_stopped_child_used_math(current)
1781#define set_used_math() set_stopped_child_used_math(current)
1782
1783#define conditional_stopped_child_used_math(condition, child) \
1784 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1785
1786#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1787
1788#define copy_to_stopped_child_used_math(child) \
1789 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1790
1791/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1792#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1793#define used_math() tsk_used_math(current)
1794
1795static __always_inline bool is_percpu_thread(void)
1796{
1797 return (current->flags & PF_NO_SETAFFINITY) &&
1798 (current->nr_cpus_allowed == 1);
1799}
1800
1801/* Per-process atomic flags. */
1802#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1803#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1804#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1805#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1806#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1807#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1808#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1809#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1810
1811#define TASK_PFA_TEST(name, func) \
1812 static inline bool task_##func(struct task_struct *p) \
1813 { return test_bit(PFA_##name, &p->atomic_flags); }
1814
1815#define TASK_PFA_SET(name, func) \
1816 static inline void task_set_##func(struct task_struct *p) \
1817 { set_bit(PFA_##name, &p->atomic_flags); }
1818
1819#define TASK_PFA_CLEAR(name, func) \
1820 static inline void task_clear_##func(struct task_struct *p) \
1821 { clear_bit(PFA_##name, &p->atomic_flags); }
1822
1823TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1824TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1825
1826TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1827TASK_PFA_SET(SPREAD_PAGE, spread_page)
1828TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1829
1830TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1831TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1832TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1833
1834TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1835TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1836TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1837
1838TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1839TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1840TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1841
1842TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1843TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1844
1845TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1846TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1847TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1848
1849TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1850TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1851
1852static inline void
1853current_restore_flags(unsigned long orig_flags, unsigned long flags)
1854{
1855 current->flags &= ~flags;
1856 current->flags |= orig_flags & flags;
1857}
1858
1859extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1860extern int task_can_attach(struct task_struct *p);
1861extern int dl_bw_alloc(int cpu, u64 dl_bw);
1862extern void dl_bw_free(int cpu, u64 dl_bw);
1863
1864/* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1865extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1866
1867/**
1868 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1869 * @p: the task
1870 * @new_mask: CPU affinity mask
1871 *
1872 * Return: zero if successful, or a negative error code
1873 */
1874extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1875extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1876extern void release_user_cpus_ptr(struct task_struct *p);
1877extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1878extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1879extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1880
1881extern int yield_to(struct task_struct *p, bool preempt);
1882extern void set_user_nice(struct task_struct *p, long nice);
1883extern int task_prio(const struct task_struct *p);
1884
1885/**
1886 * task_nice - return the nice value of a given task.
1887 * @p: the task in question.
1888 *
1889 * Return: The nice value [ -20 ... 0 ... 19 ].
1890 */
1891static inline int task_nice(const struct task_struct *p)
1892{
1893 return PRIO_TO_NICE((p)->static_prio);
1894}
1895
1896extern int can_nice(const struct task_struct *p, const int nice);
1897extern int task_curr(const struct task_struct *p);
1898extern int idle_cpu(int cpu);
1899extern int available_idle_cpu(int cpu);
1900extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1901extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1902extern void sched_set_fifo(struct task_struct *p);
1903extern void sched_set_fifo_low(struct task_struct *p);
1904extern void sched_set_normal(struct task_struct *p, int nice);
1905extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1906extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1907extern struct task_struct *idle_task(int cpu);
1908
1909/**
1910 * is_idle_task - is the specified task an idle task?
1911 * @p: the task in question.
1912 *
1913 * Return: 1 if @p is an idle task. 0 otherwise.
1914 */
1915static __always_inline bool is_idle_task(const struct task_struct *p)
1916{
1917 return !!(p->flags & PF_IDLE);
1918}
1919
1920extern struct task_struct *curr_task(int cpu);
1921extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1922
1923void yield(void);
1924
1925union thread_union {
1926 struct task_struct task;
1927#ifndef CONFIG_THREAD_INFO_IN_TASK
1928 struct thread_info thread_info;
1929#endif
1930 unsigned long stack[THREAD_SIZE/sizeof(long)];
1931};
1932
1933#ifndef CONFIG_THREAD_INFO_IN_TASK
1934extern struct thread_info init_thread_info;
1935#endif
1936
1937extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1938
1939#ifdef CONFIG_THREAD_INFO_IN_TASK
1940# define task_thread_info(task) (&(task)->thread_info)
1941#else
1942# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1943#endif
1944
1945/*
1946 * find a task by one of its numerical ids
1947 *
1948 * find_task_by_pid_ns():
1949 * finds a task by its pid in the specified namespace
1950 * find_task_by_vpid():
1951 * finds a task by its virtual pid
1952 *
1953 * see also find_vpid() etc in include/linux/pid.h
1954 */
1955
1956extern struct task_struct *find_task_by_vpid(pid_t nr);
1957extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1958
1959/*
1960 * find a task by its virtual pid and get the task struct
1961 */
1962extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1963
1964extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1965extern int wake_up_process(struct task_struct *tsk);
1966extern void wake_up_new_task(struct task_struct *tsk);
1967
1968extern void kick_process(struct task_struct *tsk);
1969
1970extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1971#define set_task_comm(tsk, from) ({ \
1972 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \
1973 __set_task_comm(tsk, from, false); \
1974})
1975
1976/*
1977 * - Why not use task_lock()?
1978 * User space can randomly change their names anyway, so locking for readers
1979 * doesn't make sense. For writers, locking is probably necessary, as a race
1980 * condition could lead to long-term mixed results.
1981 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is
1982 * always NUL-terminated and zero-padded. Therefore the race condition between
1983 * reader and writer is not an issue.
1984 *
1985 * - BUILD_BUG_ON() can help prevent the buf from being truncated.
1986 * Since the callers don't perform any return value checks, this safeguard is
1987 * necessary.
1988 */
1989#define get_task_comm(buf, tsk) ({ \
1990 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \
1991 strscpy_pad(buf, (tsk)->comm); \
1992 buf; \
1993})
1994
1995static __always_inline void scheduler_ipi(void)
1996{
1997 /*
1998 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1999 * TIF_NEED_RESCHED remotely (for the first time) will also send
2000 * this IPI.
2001 */
2002 preempt_fold_need_resched();
2003}
2004
2005extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2006
2007/*
2008 * Set thread flags in other task's structures.
2009 * See asm/thread_info.h for TIF_xxxx flags available:
2010 */
2011static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2012{
2013 set_ti_thread_flag(task_thread_info(tsk), flag);
2014}
2015
2016static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2017{
2018 clear_ti_thread_flag(task_thread_info(tsk), flag);
2019}
2020
2021static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2022 bool value)
2023{
2024 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2025}
2026
2027static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2028{
2029 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2030}
2031
2032static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2033{
2034 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2035}
2036
2037static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2038{
2039 return test_ti_thread_flag(task_thread_info(tsk), flag);
2040}
2041
2042static inline void set_tsk_need_resched(struct task_struct *tsk)
2043{
2044 if (tracepoint_enabled(sched_set_need_resched_tp) &&
2045 !test_tsk_thread_flag(tsk, TIF_NEED_RESCHED))
2046 __trace_set_need_resched(curr: tsk, TIF_NEED_RESCHED);
2047 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2048}
2049
2050static inline void clear_tsk_need_resched(struct task_struct *tsk)
2051{
2052 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2053 v: (atomic_long_t *)&task_thread_info(tsk)->flags);
2054}
2055
2056static inline int test_tsk_need_resched(struct task_struct *tsk)
2057{
2058 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2059}
2060
2061/*
2062 * cond_resched() and cond_resched_lock(): latency reduction via
2063 * explicit rescheduling in places that are safe. The return
2064 * value indicates whether a reschedule was done in fact.
2065 * cond_resched_lock() will drop the spinlock before scheduling,
2066 */
2067#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2068extern int __cond_resched(void);
2069
2070#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2071
2072DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2073
2074static __always_inline int _cond_resched(void)
2075{
2076 return static_call_mod(cond_resched)();
2077}
2078
2079#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2080
2081extern int dynamic_cond_resched(void);
2082
2083static __always_inline int _cond_resched(void)
2084{
2085 return dynamic_cond_resched();
2086}
2087
2088#else /* !CONFIG_PREEMPTION */
2089
2090static inline int _cond_resched(void)
2091{
2092 return __cond_resched();
2093}
2094
2095#endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2096
2097#else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2098
2099static inline int _cond_resched(void)
2100{
2101 return 0;
2102}
2103
2104#endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2105
2106#define cond_resched() ({ \
2107 __might_resched(__FILE__, __LINE__, 0); \
2108 _cond_resched(); \
2109})
2110
2111extern int __cond_resched_lock(spinlock_t *lock);
2112extern int __cond_resched_rwlock_read(rwlock_t *lock);
2113extern int __cond_resched_rwlock_write(rwlock_t *lock);
2114
2115#define MIGHT_RESCHED_RCU_SHIFT 8
2116#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2117
2118#ifndef CONFIG_PREEMPT_RT
2119/*
2120 * Non RT kernels have an elevated preempt count due to the held lock,
2121 * but are not allowed to be inside a RCU read side critical section
2122 */
2123# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2124#else
2125/*
2126 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2127 * cond_resched*lock() has to take that into account because it checks for
2128 * preempt_count() and rcu_preempt_depth().
2129 */
2130# define PREEMPT_LOCK_RESCHED_OFFSETS \
2131 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2132#endif
2133
2134#define cond_resched_lock(lock) ({ \
2135 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2136 __cond_resched_lock(lock); \
2137})
2138
2139#define cond_resched_rwlock_read(lock) ({ \
2140 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2141 __cond_resched_rwlock_read(lock); \
2142})
2143
2144#define cond_resched_rwlock_write(lock) ({ \
2145 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2146 __cond_resched_rwlock_write(lock); \
2147})
2148
2149#ifndef CONFIG_PREEMPT_RT
2150static inline struct mutex *__get_task_blocked_on(struct task_struct *p)
2151{
2152 struct mutex *m = p->blocked_on;
2153
2154 if (m)
2155 lockdep_assert_held_once(&m->wait_lock);
2156 return m;
2157}
2158
2159static inline void __set_task_blocked_on(struct task_struct *p, struct mutex *m)
2160{
2161 struct mutex *blocked_on = READ_ONCE(p->blocked_on);
2162
2163 WARN_ON_ONCE(!m);
2164 /* The task should only be setting itself as blocked */
2165 WARN_ON_ONCE(p != current);
2166 /* Currently we serialize blocked_on under the mutex::wait_lock */
2167 lockdep_assert_held_once(&m->wait_lock);
2168 /*
2169 * Check ensure we don't overwrite existing mutex value
2170 * with a different mutex. Note, setting it to the same
2171 * lock repeatedly is ok.
2172 */
2173 WARN_ON_ONCE(blocked_on && blocked_on != m);
2174 WRITE_ONCE(p->blocked_on, m);
2175}
2176
2177static inline void set_task_blocked_on(struct task_struct *p, struct mutex *m)
2178{
2179 guard(raw_spinlock_irqsave)(l: &m->wait_lock);
2180 __set_task_blocked_on(p, m);
2181}
2182
2183static inline void __clear_task_blocked_on(struct task_struct *p, struct mutex *m)
2184{
2185 if (m) {
2186 struct mutex *blocked_on = READ_ONCE(p->blocked_on);
2187
2188 /* Currently we serialize blocked_on under the mutex::wait_lock */
2189 lockdep_assert_held_once(&m->wait_lock);
2190 /*
2191 * There may be cases where we re-clear already cleared
2192 * blocked_on relationships, but make sure we are not
2193 * clearing the relationship with a different lock.
2194 */
2195 WARN_ON_ONCE(blocked_on && blocked_on != m);
2196 }
2197 WRITE_ONCE(p->blocked_on, NULL);
2198}
2199
2200static inline void clear_task_blocked_on(struct task_struct *p, struct mutex *m)
2201{
2202 guard(raw_spinlock_irqsave)(l: &m->wait_lock);
2203 __clear_task_blocked_on(p, m);
2204}
2205#else
2206static inline void __clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m)
2207{
2208}
2209
2210static inline void clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m)
2211{
2212}
2213#endif /* !CONFIG_PREEMPT_RT */
2214
2215static __always_inline bool need_resched(void)
2216{
2217 return unlikely(tif_need_resched());
2218}
2219
2220/*
2221 * Wrappers for p->thread_info->cpu access. No-op on UP.
2222 */
2223#ifdef CONFIG_SMP
2224
2225static inline unsigned int task_cpu(const struct task_struct *p)
2226{
2227 return READ_ONCE(task_thread_info(p)->cpu);
2228}
2229
2230extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2231
2232#else
2233
2234static inline unsigned int task_cpu(const struct task_struct *p)
2235{
2236 return 0;
2237}
2238
2239static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2240{
2241}
2242
2243#endif /* CONFIG_SMP */
2244
2245static inline bool task_is_runnable(struct task_struct *p)
2246{
2247 return p->on_rq && !p->se.sched_delayed;
2248}
2249
2250extern bool sched_task_on_rq(struct task_struct *p);
2251extern unsigned long get_wchan(struct task_struct *p);
2252extern struct task_struct *cpu_curr_snapshot(int cpu);
2253
2254/*
2255 * In order to reduce various lock holder preemption latencies provide an
2256 * interface to see if a vCPU is currently running or not.
2257 *
2258 * This allows us to terminate optimistic spin loops and block, analogous to
2259 * the native optimistic spin heuristic of testing if the lock owner task is
2260 * running or not.
2261 */
2262#ifndef vcpu_is_preempted
2263static inline bool vcpu_is_preempted(int cpu)
2264{
2265 return false;
2266}
2267#endif
2268
2269extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2270extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2271
2272#ifndef TASK_SIZE_OF
2273#define TASK_SIZE_OF(tsk) TASK_SIZE
2274#endif
2275
2276static inline bool owner_on_cpu(struct task_struct *owner)
2277{
2278 /*
2279 * As lock holder preemption issue, we both skip spinning if
2280 * task is not on cpu or its cpu is preempted
2281 */
2282 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(cpu: task_cpu(p: owner));
2283}
2284
2285/* Returns effective CPU energy utilization, as seen by the scheduler */
2286unsigned long sched_cpu_util(int cpu);
2287
2288#ifdef CONFIG_SCHED_CORE
2289extern void sched_core_free(struct task_struct *tsk);
2290extern void sched_core_fork(struct task_struct *p);
2291extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2292 unsigned long uaddr);
2293extern int sched_core_idle_cpu(int cpu);
2294#else
2295static inline void sched_core_free(struct task_struct *tsk) { }
2296static inline void sched_core_fork(struct task_struct *p) { }
2297static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2298#endif
2299
2300extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2301
2302#ifdef CONFIG_MEM_ALLOC_PROFILING
2303static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2304{
2305 swap(current->alloc_tag, tag);
2306 return tag;
2307}
2308
2309static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2310{
2311#ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2312 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2313#endif
2314 current->alloc_tag = old;
2315}
2316#else
2317#define alloc_tag_save(_tag) NULL
2318#define alloc_tag_restore(_tag, _old) do {} while (0)
2319#endif
2320
2321#ifndef MODULE
2322#ifndef COMPILE_OFFSETS
2323
2324extern void ___migrate_enable(void);
2325
2326struct rq;
2327DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
2328
2329/*
2330 * The "struct rq" is not available here, so we can't access the
2331 * "runqueues" with this_cpu_ptr(), as the compilation will fail in
2332 * this_cpu_ptr() -> raw_cpu_ptr() -> __verify_pcpu_ptr():
2333 * typeof((ptr) + 0)
2334 *
2335 * So use arch_raw_cpu_ptr()/PERCPU_PTR() directly here.
2336 */
2337#ifdef CONFIG_SMP
2338#define this_rq_raw() arch_raw_cpu_ptr(&runqueues)
2339#else
2340#define this_rq_raw() PERCPU_PTR(&runqueues)
2341#endif
2342#define this_rq_pinned() (*(unsigned int *)((void *)this_rq_raw() + RQ_nr_pinned))
2343
2344static inline void __migrate_enable(void)
2345{
2346 struct task_struct *p = current;
2347
2348#ifdef CONFIG_DEBUG_PREEMPT
2349 /*
2350 * Check both overflow from migrate_disable() and superfluous
2351 * migrate_enable().
2352 */
2353 if (WARN_ON_ONCE((s16)p->migration_disabled <= 0))
2354 return;
2355#endif
2356
2357 if (p->migration_disabled > 1) {
2358 p->migration_disabled--;
2359 return;
2360 }
2361
2362 /*
2363 * Ensure stop_task runs either before or after this, and that
2364 * __set_cpus_allowed_ptr(SCA_MIGRATE_ENABLE) doesn't schedule().
2365 */
2366 guard(preempt)();
2367 if (unlikely(p->cpus_ptr != &p->cpus_mask))
2368 ___migrate_enable();
2369 /*
2370 * Mustn't clear migration_disabled() until cpus_ptr points back at the
2371 * regular cpus_mask, otherwise things that race (eg.
2372 * select_fallback_rq) get confused.
2373 */
2374 barrier();
2375 p->migration_disabled = 0;
2376 this_rq_pinned()--;
2377}
2378
2379static inline void __migrate_disable(void)
2380{
2381 struct task_struct *p = current;
2382
2383 if (p->migration_disabled) {
2384#ifdef CONFIG_DEBUG_PREEMPT
2385 /*
2386 *Warn about overflow half-way through the range.
2387 */
2388 WARN_ON_ONCE((s16)p->migration_disabled < 0);
2389#endif
2390 p->migration_disabled++;
2391 return;
2392 }
2393
2394 guard(preempt)();
2395 this_rq_pinned()++;
2396 p->migration_disabled = 1;
2397}
2398#else /* !COMPILE_OFFSETS */
2399static inline void __migrate_disable(void) { }
2400static inline void __migrate_enable(void) { }
2401#endif /* !COMPILE_OFFSETS */
2402
2403/*
2404 * So that it is possible to not export the runqueues variable, define and
2405 * export migrate_enable/migrate_disable in kernel/sched/core.c too, and use
2406 * them for the modules. The macro "INSTANTIATE_EXPORTED_MIGRATE_DISABLE" will
2407 * be defined in kernel/sched/core.c.
2408 */
2409#ifndef INSTANTIATE_EXPORTED_MIGRATE_DISABLE
2410static inline void migrate_disable(void)
2411{
2412 __migrate_disable();
2413}
2414
2415static inline void migrate_enable(void)
2416{
2417 __migrate_enable();
2418}
2419#else /* INSTANTIATE_EXPORTED_MIGRATE_DISABLE */
2420extern void migrate_disable(void);
2421extern void migrate_enable(void);
2422#endif /* INSTANTIATE_EXPORTED_MIGRATE_DISABLE */
2423
2424#else /* MODULE */
2425extern void migrate_disable(void);
2426extern void migrate_enable(void);
2427#endif /* MODULE */
2428
2429DEFINE_LOCK_GUARD_0(migrate, migrate_disable(), migrate_enable())
2430
2431#endif
2432