Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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 <linux/rseq_types.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#include <linux/time64.h>
53#ifndef COMPILE_OFFSETS
54#include <generated/rq-offsets.h>
55#endif
56
57/* task_struct member predeclarations (sorted alphabetically): */
58struct audit_context;
59struct bio_list;
60struct blk_plug;
61struct bpf_local_storage;
62struct bpf_run_ctx;
63struct bpf_net_context;
64struct capture_control;
65struct cfs_rq;
66struct fs_struct;
67struct futex_pi_state;
68struct io_context;
69struct io_uring_task;
70struct mempolicy;
71struct nameidata;
72struct nsproxy;
73struct perf_event_context;
74struct perf_ctx_data;
75struct pid_namespace;
76struct pipe_inode_info;
77struct rcu_node;
78struct reclaim_state;
79struct robust_list_head;
80struct root_domain;
81struct rq;
82struct sched_attr;
83struct sched_dl_entity;
84struct seq_file;
85struct sighand_struct;
86struct signal_struct;
87struct task_delay_info;
88struct task_group;
89struct task_struct;
90struct timespec64;
91struct user_event_mm;
92
93#include <linux/sched/ext.h>
94
95/*
96 * Task state bitmask. NOTE! These bits are also
97 * encoded in fs/proc/array.c: get_task_state().
98 *
99 * We have two separate sets of flags: task->__state
100 * is about runnability, while task->exit_state are
101 * about the task exiting. Confusing, but this way
102 * modifying one set can't modify the other one by
103 * mistake.
104 */
105
106/* Used in tsk->__state: */
107#define TASK_RUNNING 0x00000000
108#define TASK_INTERRUPTIBLE 0x00000001
109#define TASK_UNINTERRUPTIBLE 0x00000002
110#define __TASK_STOPPED 0x00000004
111#define __TASK_TRACED 0x00000008
112/* Used in tsk->exit_state: */
113#define EXIT_DEAD 0x00000010
114#define EXIT_ZOMBIE 0x00000020
115#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
116/* Used in tsk->__state again: */
117#define TASK_PARKED 0x00000040
118#define TASK_DEAD 0x00000080
119#define TASK_WAKEKILL 0x00000100
120#define TASK_WAKING 0x00000200
121#define TASK_NOLOAD 0x00000400
122#define TASK_NEW 0x00000800
123#define TASK_RTLOCK_WAIT 0x00001000
124#define TASK_FREEZABLE 0x00002000
125#define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
126#define TASK_FROZEN 0x00008000
127#define TASK_STATE_MAX 0x00010000
128
129#define TASK_ANY (TASK_STATE_MAX-1)
130
131/*
132 * DO NOT ADD ANY NEW USERS !
133 */
134#define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
135
136/* Convenience macros for the sake of set_current_state: */
137#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
138#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
139#define TASK_TRACED __TASK_TRACED
140
141#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
142
143/* Convenience macros for the sake of wake_up(): */
144#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
145
146/* get_task_state(): */
147#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
148 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
149 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
150 TASK_PARKED)
151
152#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
153
154#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
155#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
156#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
157
158/*
159 * Special states are those that do not use the normal wait-loop pattern. See
160 * the comment with set_special_state().
161 */
162#define is_special_task_state(state) \
163 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \
164 TASK_DEAD | TASK_FROZEN))
165
166#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
167# define debug_normal_state_change(state_value) \
168 do { \
169 WARN_ON_ONCE(is_special_task_state(state_value)); \
170 current->task_state_change = _THIS_IP_; \
171 } while (0)
172
173# define debug_special_state_change(state_value) \
174 do { \
175 WARN_ON_ONCE(!is_special_task_state(state_value)); \
176 current->task_state_change = _THIS_IP_; \
177 } while (0)
178
179# define debug_rtlock_wait_set_state() \
180 do { \
181 current->saved_state_change = current->task_state_change;\
182 current->task_state_change = _THIS_IP_; \
183 } while (0)
184
185# define debug_rtlock_wait_restore_state() \
186 do { \
187 current->task_state_change = current->saved_state_change;\
188 } while (0)
189
190#else
191# define debug_normal_state_change(cond) do { } while (0)
192# define debug_special_state_change(cond) do { } while (0)
193# define debug_rtlock_wait_set_state() do { } while (0)
194# define debug_rtlock_wait_restore_state() do { } while (0)
195#endif
196
197#define trace_set_current_state(state_value) \
198 do { \
199 if (tracepoint_enabled(sched_set_state_tp)) \
200 __trace_set_current_state(state_value); \
201 } while (0)
202
203/*
204 * set_current_state() includes a barrier so that the write of current->__state
205 * is correctly serialised wrt the caller's subsequent test of whether to
206 * actually sleep:
207 *
208 * for (;;) {
209 * set_current_state(TASK_UNINTERRUPTIBLE);
210 * if (CONDITION)
211 * break;
212 *
213 * schedule();
214 * }
215 * __set_current_state(TASK_RUNNING);
216 *
217 * If the caller does not need such serialisation (because, for instance, the
218 * CONDITION test and condition change and wakeup are under the same lock) then
219 * use __set_current_state().
220 *
221 * The above is typically ordered against the wakeup, which does:
222 *
223 * CONDITION = 1;
224 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
225 *
226 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
227 * accessing p->__state.
228 *
229 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
230 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
231 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
232 *
233 * However, with slightly different timing the wakeup TASK_RUNNING store can
234 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
235 * a problem either because that will result in one extra go around the loop
236 * and our @cond test will save the day.
237 *
238 * Also see the comments of try_to_wake_up().
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 WRITE_ONCE(current->__state, (state_value)); \
245 } while (0)
246
247#define set_current_state(state_value) \
248 do { \
249 debug_normal_state_change((state_value)); \
250 trace_set_current_state(state_value); \
251 smp_store_mb(current->__state, (state_value)); \
252 } while (0)
253
254/*
255 * set_special_state() should be used for those states when the blocking task
256 * can not use the regular condition based wait-loop. In that case we must
257 * serialize against wakeups such that any possible in-flight TASK_RUNNING
258 * stores will not collide with our state change.
259 */
260#define set_special_state(state_value) \
261 do { \
262 unsigned long flags; /* may shadow */ \
263 \
264 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
265 debug_special_state_change((state_value)); \
266 trace_set_current_state(state_value); \
267 WRITE_ONCE(current->__state, (state_value)); \
268 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
269 } while (0)
270
271/*
272 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
273 *
274 * RT's spin/rwlock substitutions are state preserving. The state of the
275 * task when blocking on the lock is saved in task_struct::saved_state and
276 * restored after the lock has been acquired. These operations are
277 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
278 * lock related wakeups while the task is blocked on the lock are
279 * redirected to operate on task_struct::saved_state to ensure that these
280 * are not dropped. On restore task_struct::saved_state is set to
281 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
282 *
283 * The lock operation looks like this:
284 *
285 * current_save_and_set_rtlock_wait_state();
286 * for (;;) {
287 * if (try_lock())
288 * break;
289 * raw_spin_unlock_irq(&lock->wait_lock);
290 * schedule_rtlock();
291 * raw_spin_lock_irq(&lock->wait_lock);
292 * set_current_state(TASK_RTLOCK_WAIT);
293 * }
294 * current_restore_rtlock_saved_state();
295 */
296#define current_save_and_set_rtlock_wait_state() \
297 do { \
298 lockdep_assert_irqs_disabled(); \
299 raw_spin_lock(¤t->pi_lock); \
300 current->saved_state = current->__state; \
301 debug_rtlock_wait_set_state(); \
302 trace_set_current_state(TASK_RTLOCK_WAIT); \
303 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
304 raw_spin_unlock(¤t->pi_lock); \
305 } while (0);
306
307#define current_restore_rtlock_saved_state() \
308 do { \
309 lockdep_assert_irqs_disabled(); \
310 raw_spin_lock(¤t->pi_lock); \
311 debug_rtlock_wait_restore_state(); \
312 trace_set_current_state(current->saved_state); \
313 WRITE_ONCE(current->__state, current->saved_state); \
314 current->saved_state = TASK_RUNNING; \
315 raw_spin_unlock(¤t->pi_lock); \
316 } while (0);
317
318#define get_current_state() READ_ONCE(current->__state)
319
320/*
321 * Define the task command name length as enum, then it can be visible to
322 * BPF programs.
323 */
324enum {
325 TASK_COMM_LEN = 16,
326};
327
328extern void sched_tick(void);
329
330#define MAX_SCHEDULE_TIMEOUT LONG_MAX
331
332extern long schedule_timeout(long timeout);
333extern long schedule_timeout_interruptible(long timeout);
334extern long schedule_timeout_killable(long timeout);
335extern long schedule_timeout_uninterruptible(long timeout);
336extern long schedule_timeout_idle(long timeout);
337asmlinkage void schedule(void);
338extern void schedule_preempt_disabled(void);
339asmlinkage void preempt_schedule_irq(void);
340#ifdef CONFIG_PREEMPT_RT
341 extern void schedule_rtlock(void);
342#endif
343
344extern int __must_check io_schedule_prepare(void);
345extern void io_schedule_finish(int token);
346extern long io_schedule_timeout(long timeout);
347extern void io_schedule(void);
348
349/* wrapper functions to trace from this header file */
350DECLARE_TRACEPOINT(sched_set_state_tp);
351extern void __trace_set_current_state(int state_value);
352DECLARE_TRACEPOINT(sched_set_need_resched_tp);
353extern void __trace_set_need_resched(struct task_struct *curr, int tif);
354
355/**
356 * struct prev_cputime - snapshot of system and user cputime
357 * @utime: time spent in user mode
358 * @stime: time spent in system mode
359 * @lock: protects the above two fields
360 *
361 * Stores previous user/system time values such that we can guarantee
362 * monotonicity.
363 */
364struct prev_cputime {
365#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
366 u64 utime;
367 u64 stime;
368 raw_spinlock_t lock;
369#endif
370};
371
372enum vtime_state {
373 /* Task is sleeping or running in a CPU with VTIME inactive: */
374 VTIME_INACTIVE = 0,
375 /* Task is idle */
376 VTIME_IDLE,
377 /* Task runs in kernelspace in a CPU with VTIME active: */
378 VTIME_SYS,
379 /* Task runs in userspace in a CPU with VTIME active: */
380 VTIME_USER,
381 /* Task runs as guests in a CPU with VTIME active: */
382 VTIME_GUEST,
383};
384
385struct vtime {
386 seqcount_t seqcount;
387 unsigned long long starttime;
388 enum vtime_state state;
389 unsigned int cpu;
390 u64 utime;
391 u64 stime;
392 u64 gtime;
393};
394
395/*
396 * Utilization clamp constraints.
397 * @UCLAMP_MIN: Minimum utilization
398 * @UCLAMP_MAX: Maximum utilization
399 * @UCLAMP_CNT: Utilization clamp constraints count
400 */
401enum uclamp_id {
402 UCLAMP_MIN = 0,
403 UCLAMP_MAX,
404 UCLAMP_CNT
405};
406
407extern struct root_domain def_root_domain;
408extern struct mutex sched_domains_mutex;
409extern void sched_domains_mutex_lock(void);
410extern void sched_domains_mutex_unlock(void);
411
412struct sched_param {
413 int sched_priority;
414};
415
416struct sched_info {
417#ifdef CONFIG_SCHED_INFO
418 /* Cumulative counters: */
419
420 /* # of times we have run on this CPU: */
421 unsigned long pcount;
422
423 /* Time spent waiting on a runqueue: */
424 unsigned long long run_delay;
425
426 /* Max time spent waiting on a runqueue: */
427 unsigned long long max_run_delay;
428
429 /* Min time spent waiting on a runqueue: */
430 unsigned long long min_run_delay;
431
432 /* Timestamps: */
433
434 /* When did we last run on a CPU? */
435 unsigned long long last_arrival;
436
437 /* When were we last queued to run? */
438 unsigned long long last_queued;
439
440 /* Timestamp of max time spent waiting on a runqueue: */
441 struct timespec64 max_run_delay_ts;
442
443#endif /* CONFIG_SCHED_INFO */
444};
445
446/*
447 * Integer metrics need fixed point arithmetic, e.g., sched/fair
448 * has a few: load, load_avg, util_avg, freq, and capacity.
449 *
450 * We define a basic fixed point arithmetic range, and then formalize
451 * all these metrics based on that basic range.
452 */
453# define SCHED_FIXEDPOINT_SHIFT 10
454# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
455
456/* Increase resolution of cpu_capacity calculations */
457# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
458# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
459
460struct load_weight {
461 unsigned long weight;
462 u32 inv_weight;
463};
464
465/*
466 * The load/runnable/util_avg accumulates an infinite geometric series
467 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
468 *
469 * [load_avg definition]
470 *
471 * load_avg = runnable% * scale_load_down(load)
472 *
473 * [runnable_avg definition]
474 *
475 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
476 *
477 * [util_avg definition]
478 *
479 * util_avg = running% * SCHED_CAPACITY_SCALE
480 *
481 * where runnable% is the time ratio that a sched_entity is runnable and
482 * running% the time ratio that a sched_entity is running.
483 *
484 * For cfs_rq, they are the aggregated values of all runnable and blocked
485 * sched_entities.
486 *
487 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
488 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
489 * for computing those signals (see update_rq_clock_pelt())
490 *
491 * N.B., the above ratios (runnable% and running%) themselves are in the
492 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
493 * to as large a range as necessary. This is for example reflected by
494 * util_avg's SCHED_CAPACITY_SCALE.
495 *
496 * [Overflow issue]
497 *
498 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
499 * with the highest load (=88761), always runnable on a single cfs_rq,
500 * and should not overflow as the number already hits PID_MAX_LIMIT.
501 *
502 * For all other cases (including 32-bit kernels), struct load_weight's
503 * weight will overflow first before we do, because:
504 *
505 * Max(load_avg) <= Max(load.weight)
506 *
507 * Then it is the load_weight's responsibility to consider overflow
508 * issues.
509 */
510struct sched_avg {
511 u64 last_update_time;
512 u64 load_sum;
513 u64 runnable_sum;
514 u32 util_sum;
515 u32 period_contrib;
516 unsigned long load_avg;
517 unsigned long runnable_avg;
518 unsigned long util_avg;
519 unsigned int util_est;
520} ____cacheline_aligned;
521
522/*
523 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
524 * updates. When a task is dequeued, its util_est should not be updated if its
525 * util_avg has not been updated in the meantime.
526 * This information is mapped into the MSB bit of util_est at dequeue time.
527 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
528 * it is safe to use MSB.
529 */
530#define UTIL_EST_WEIGHT_SHIFT 2
531#define UTIL_AVG_UNCHANGED 0x80000000
532
533struct sched_statistics {
534#ifdef CONFIG_SCHEDSTATS
535 u64 wait_start;
536 u64 wait_max;
537 u64 wait_count;
538 u64 wait_sum;
539 u64 iowait_count;
540 u64 iowait_sum;
541
542 u64 sleep_start;
543 u64 sleep_max;
544 s64 sum_sleep_runtime;
545
546 u64 block_start;
547 u64 block_max;
548 s64 sum_block_runtime;
549
550 s64 exec_max;
551 u64 slice_max;
552
553 u64 nr_migrations_cold;
554 u64 nr_failed_migrations_affine;
555 u64 nr_failed_migrations_running;
556 u64 nr_failed_migrations_hot;
557 u64 nr_forced_migrations;
558
559 u64 nr_wakeups;
560 u64 nr_wakeups_sync;
561 u64 nr_wakeups_migrate;
562 u64 nr_wakeups_local;
563 u64 nr_wakeups_remote;
564 u64 nr_wakeups_affine;
565 u64 nr_wakeups_affine_attempts;
566 u64 nr_wakeups_passive;
567 u64 nr_wakeups_idle;
568
569#ifdef CONFIG_SCHED_CORE
570 u64 core_forceidle_sum;
571#endif
572#endif /* CONFIG_SCHEDSTATS */
573} ____cacheline_aligned;
574
575struct sched_entity {
576 /* For load-balancing: */
577 struct load_weight load;
578 struct rb_node run_node;
579 u64 deadline;
580 u64 min_vruntime;
581 u64 min_slice;
582 u64 max_slice;
583
584 struct list_head group_node;
585 unsigned char on_rq;
586 unsigned char sched_delayed;
587 unsigned char rel_deadline;
588 unsigned char custom_slice;
589 /* hole */
590
591 u64 exec_start;
592 u64 sum_exec_runtime;
593 u64 prev_sum_exec_runtime;
594 u64 vruntime;
595 /* Approximated virtual lag: */
596 s64 vlag;
597 /* 'Protected' deadline, to give out minimum quantums: */
598 u64 vprot;
599 u64 slice;
600
601 u64 nr_migrations;
602
603#ifdef CONFIG_FAIR_GROUP_SCHED
604 int depth;
605 struct sched_entity *parent;
606 /* rq on which this entity is (to be) queued: */
607 struct cfs_rq *cfs_rq;
608 /* rq "owned" by this entity/group: */
609 struct cfs_rq *my_q;
610 /* cached value of my_q->h_nr_running */
611 unsigned long runnable_weight;
612#endif
613
614 /*
615 * Per entity load average tracking.
616 *
617 * Put into separate cache line so it does not
618 * collide with read-mostly values above.
619 */
620 struct sched_avg avg;
621};
622
623struct sched_rt_entity {
624 struct list_head run_list;
625 unsigned long timeout;
626 unsigned long watchdog_stamp;
627 unsigned int time_slice;
628 unsigned short on_rq;
629 unsigned short on_list;
630
631 struct sched_rt_entity *back;
632#ifdef CONFIG_RT_GROUP_SCHED
633 struct sched_rt_entity *parent;
634 /* rq on which this entity is (to be) queued: */
635 struct rt_rq *rt_rq;
636 /* rq "owned" by this entity/group: */
637 struct rt_rq *my_q;
638#endif
639} __randomize_layout;
640
641struct rq_flags;
642typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *, struct rq_flags *rf);
643
644struct sched_dl_entity {
645 struct rb_node rb_node;
646
647 /*
648 * Original scheduling parameters. Copied here from sched_attr
649 * during sched_setattr(), they will remain the same until
650 * the next sched_setattr().
651 */
652 u64 dl_runtime; /* Maximum runtime for each instance */
653 u64 dl_deadline; /* Relative deadline of each instance */
654 u64 dl_period; /* Separation of two instances (period) */
655 u64 dl_bw; /* dl_runtime / dl_period */
656 u64 dl_density; /* dl_runtime / dl_deadline */
657
658 /*
659 * Actual scheduling parameters. Initialized with the values above,
660 * they are continuously updated during task execution. Note that
661 * the remaining runtime could be < 0 in case we are in overrun.
662 */
663 s64 runtime; /* Remaining runtime for this instance */
664 u64 deadline; /* Absolute deadline for this instance */
665 unsigned int flags; /* Specifying the scheduler behaviour */
666
667 /*
668 * Some bool flags:
669 *
670 * @dl_throttled tells if we exhausted the runtime. If so, the
671 * task has to wait for a replenishment to be performed at the
672 * next firing of dl_timer.
673 *
674 * @dl_yielded tells if task gave up the CPU before consuming
675 * all its available runtime during the last job.
676 *
677 * @dl_non_contending tells if the task is inactive while still
678 * contributing to the active utilization. In other words, it
679 * indicates if the inactive timer has been armed and its handler
680 * has not been executed yet. This flag is useful to avoid race
681 * conditions between the inactive timer handler and the wakeup
682 * code.
683 *
684 * @dl_overrun tells if the task asked to be informed about runtime
685 * overruns.
686 *
687 * @dl_server tells if this is a server entity.
688 *
689 * @dl_server_active tells if the dlserver is active(started).
690 * dlserver is started on first cfs enqueue on an idle runqueue
691 * and is stopped when a dequeue results in 0 cfs tasks on the
692 * runqueue. In other words, dlserver is active only when cpu's
693 * runqueue has atleast one cfs task.
694 *
695 * @dl_defer tells if this is a deferred or regular server. For
696 * now only defer server exists.
697 *
698 * @dl_defer_armed tells if the deferrable server is waiting
699 * for the replenishment timer to activate it.
700 *
701 * @dl_defer_running tells if the deferrable server is actually
702 * running, skipping the defer phase.
703 *
704 * @dl_defer_idle tracks idle state
705 */
706 unsigned int dl_throttled : 1;
707 unsigned int dl_yielded : 1;
708 unsigned int dl_non_contending : 1;
709 unsigned int dl_overrun : 1;
710 unsigned int dl_server : 1;
711 unsigned int dl_server_active : 1;
712 unsigned int dl_defer : 1;
713 unsigned int dl_defer_armed : 1;
714 unsigned int dl_defer_running : 1;
715 unsigned int dl_defer_idle : 1;
716
717 /*
718 * Bandwidth enforcement timer. Each -deadline task has its
719 * own bandwidth to be enforced, thus we need one timer per task.
720 */
721 struct hrtimer dl_timer;
722
723 /*
724 * Inactive timer, responsible for decreasing the active utilization
725 * at the "0-lag time". When a -deadline task blocks, it contributes
726 * to GRUB's active utilization until the "0-lag time", hence a
727 * timer is needed to decrease the active utilization at the correct
728 * time.
729 */
730 struct hrtimer inactive_timer;
731
732 /*
733 * Bits for DL-server functionality. Also see the comment near
734 * dl_server_update().
735 *
736 * @rq the runqueue this server is for
737 */
738 struct rq *rq;
739 dl_server_pick_f server_pick_task;
740
741#ifdef CONFIG_RT_MUTEXES
742 /*
743 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
744 * pi_se points to the donor, otherwise points to the dl_se it belongs
745 * to (the original one/itself).
746 */
747 struct sched_dl_entity *pi_se;
748#endif
749};
750
751#ifdef CONFIG_UCLAMP_TASK
752/* Number of utilization clamp buckets (shorter alias) */
753#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
754
755/*
756 * Utilization clamp for a scheduling entity
757 * @value: clamp value "assigned" to a se
758 * @bucket_id: bucket index corresponding to the "assigned" value
759 * @active: the se is currently refcounted in a rq's bucket
760 * @user_defined: the requested clamp value comes from user-space
761 *
762 * The bucket_id is the index of the clamp bucket matching the clamp value
763 * which is pre-computed and stored to avoid expensive integer divisions from
764 * the fast path.
765 *
766 * The active bit is set whenever a task has got an "effective" value assigned,
767 * which can be different from the clamp value "requested" from user-space.
768 * This allows to know a task is refcounted in the rq's bucket corresponding
769 * to the "effective" bucket_id.
770 *
771 * The user_defined bit is set whenever a task has got a task-specific clamp
772 * value requested from userspace, i.e. the system defaults apply to this task
773 * just as a restriction. This allows to relax default clamps when a less
774 * restrictive task-specific value has been requested, thus allowing to
775 * implement a "nice" semantic. For example, a task running with a 20%
776 * default boost can still drop its own boosting to 0%.
777 */
778struct uclamp_se {
779 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
780 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
781 unsigned int active : 1;
782 unsigned int user_defined : 1;
783};
784#endif /* CONFIG_UCLAMP_TASK */
785
786union rcu_special {
787 struct {
788 u8 blocked;
789 u8 need_qs;
790 u8 exp_hint; /* Hint for performance. */
791 u8 need_mb; /* Readers need smp_mb(). */
792 } b; /* Bits. */
793 u32 s; /* Set of bits. */
794};
795
796enum perf_event_task_context {
797 perf_invalid_context = -1,
798 perf_hw_context = 0,
799 perf_sw_context,
800 perf_nr_task_contexts,
801};
802
803/*
804 * Number of contexts where an event can trigger:
805 * task, softirq, hardirq, nmi.
806 */
807#define PERF_NR_CONTEXTS 4
808
809struct wake_q_node {
810 struct wake_q_node *next;
811};
812
813struct kmap_ctrl {
814#ifdef CONFIG_KMAP_LOCAL
815 int idx;
816 pte_t pteval[KM_MAX_IDX];
817#endif
818};
819
820struct task_struct {
821#ifdef CONFIG_THREAD_INFO_IN_TASK
822 /*
823 * For reasons of header soup (see current_thread_info()), this
824 * must be the first element of task_struct.
825 */
826 struct thread_info thread_info;
827#endif
828 unsigned int __state;
829
830 /* saved state for "spinlock sleepers" */
831 unsigned int saved_state;
832
833 /*
834 * This begins the randomizable portion of task_struct. Only
835 * scheduling-critical items should be added above here.
836 */
837 randomized_struct_fields_start
838
839 void *stack;
840 refcount_t usage;
841 /* Per task flags (PF_*), defined further below: */
842 unsigned int flags;
843 unsigned int ptrace;
844
845#ifdef CONFIG_MEM_ALLOC_PROFILING
846 struct alloc_tag *alloc_tag;
847#endif
848
849 int on_cpu;
850 struct __call_single_node wake_entry;
851 unsigned int wakee_flips;
852 unsigned long wakee_flip_decay_ts;
853 struct task_struct *last_wakee;
854
855 /*
856 * recent_used_cpu is initially set as the last CPU used by a task
857 * that wakes affine another task. Waker/wakee relationships can
858 * push tasks around a CPU where each wakeup moves to the next one.
859 * Tracking a recently used CPU allows a quick search for a recently
860 * used CPU that may be idle.
861 */
862 int recent_used_cpu;
863 int wake_cpu;
864 int on_rq;
865
866 int prio;
867 int static_prio;
868 int normal_prio;
869 unsigned int rt_priority;
870
871 struct sched_entity se;
872 struct sched_rt_entity rt;
873 struct sched_dl_entity dl;
874 struct sched_dl_entity *dl_server;
875#ifdef CONFIG_SCHED_CLASS_EXT
876 struct sched_ext_entity scx;
877#endif
878 const struct sched_class *sched_class;
879
880#ifdef CONFIG_SCHED_CORE
881 struct rb_node core_node;
882 unsigned long core_cookie;
883 unsigned int core_occupation;
884#endif
885
886#ifdef CONFIG_CGROUP_SCHED
887 struct task_group *sched_task_group;
888#ifdef CONFIG_CFS_BANDWIDTH
889 struct callback_head sched_throttle_work;
890 struct list_head throttle_node;
891 bool throttled;
892#endif
893#endif
894
895
896#ifdef CONFIG_UCLAMP_TASK
897 /*
898 * Clamp values requested for a scheduling entity.
899 * Must be updated with task_rq_lock() held.
900 */
901 struct uclamp_se uclamp_req[UCLAMP_CNT];
902 /*
903 * Effective clamp values used for a scheduling entity.
904 * Must be updated with task_rq_lock() held.
905 */
906 struct uclamp_se uclamp[UCLAMP_CNT];
907#endif
908
909 struct sched_statistics stats;
910
911#ifdef CONFIG_PREEMPT_NOTIFIERS
912 /* List of struct preempt_notifier: */
913 struct hlist_head preempt_notifiers;
914#endif
915
916#ifdef CONFIG_BLK_DEV_IO_TRACE
917 unsigned int btrace_seq;
918#endif
919
920 unsigned int policy;
921 unsigned long max_allowed_capacity;
922 int nr_cpus_allowed;
923 const cpumask_t *cpus_ptr;
924 cpumask_t *user_cpus_ptr;
925 cpumask_t cpus_mask;
926 void *migration_pending;
927 unsigned short migration_disabled;
928 unsigned short migration_flags;
929
930#ifdef CONFIG_PREEMPT_RCU
931 int rcu_read_lock_nesting;
932 union rcu_special rcu_read_unlock_special;
933 struct list_head rcu_node_entry;
934 struct rcu_node *rcu_blocked_node;
935#endif /* #ifdef CONFIG_PREEMPT_RCU */
936
937#ifdef CONFIG_TASKS_RCU
938 unsigned long rcu_tasks_nvcsw;
939 u8 rcu_tasks_holdout;
940 u8 rcu_tasks_idx;
941 int rcu_tasks_idle_cpu;
942 struct list_head rcu_tasks_holdout_list;
943 int rcu_tasks_exit_cpu;
944 struct list_head rcu_tasks_exit_list;
945#endif /* #ifdef CONFIG_TASKS_RCU */
946
947#ifdef CONFIG_TASKS_TRACE_RCU
948 int trc_reader_nesting;
949 struct srcu_ctr __percpu *trc_reader_scp;
950#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
951
952#ifdef CONFIG_TRIVIAL_PREEMPT_RCU
953 int rcu_trivial_preempt_nesting;
954#endif /* #ifdef CONFIG_TRIVIAL_PREEMPT_RCU */
955
956 struct sched_info sched_info;
957
958 struct list_head tasks;
959 struct plist_node pushable_tasks;
960 struct rb_node pushable_dl_tasks;
961
962 struct mm_struct *mm;
963 struct mm_struct *active_mm;
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 by begin_new_exec()
1167 * - set it with set_task_comm() to ensure it is always
1168 * NUL-terminated and zero-padded
1169 */
1170 char comm[TASK_COMM_LEN];
1171
1172 struct nameidata *nameidata;
1173
1174#ifdef CONFIG_SYSVIPC
1175 struct sysv_sem sysvsem;
1176 struct sysv_shm sysvshm;
1177#endif
1178#ifdef CONFIG_DETECT_HUNG_TASK
1179 unsigned long last_switch_count;
1180 unsigned long last_switch_time;
1181#endif
1182 /* Filesystem information: */
1183 struct fs_struct *fs;
1184
1185 /* Open file information: */
1186 struct files_struct *files;
1187
1188#ifdef CONFIG_IO_URING
1189 struct io_uring_task *io_uring;
1190 struct io_restriction *io_uring_restrict;
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 struct mutex *blocked_on; /* lock we're blocked on */
1242 raw_spinlock_t blocked_lock;
1243
1244#ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER
1245 /*
1246 * Encoded lock address causing task block (lower 2 bits = type from
1247 * <linux/hung_task.h>). Accessed via hung_task_*() helpers.
1248 */
1249 unsigned long blocker;
1250#endif
1251
1252#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1253 int non_block_count;
1254#endif
1255
1256#ifdef CONFIG_TRACE_IRQFLAGS
1257 struct irqtrace_events irqtrace;
1258 unsigned int hardirq_threaded;
1259 u64 hardirq_chain_key;
1260 int softirqs_enabled;
1261 int softirq_context;
1262 int irq_config;
1263#endif
1264#ifdef CONFIG_PREEMPT_RT
1265 int softirq_disable_cnt;
1266#endif
1267
1268#ifdef CONFIG_LOCKDEP
1269# define MAX_LOCK_DEPTH 48UL
1270 u64 curr_chain_key;
1271 int lockdep_depth;
1272 unsigned int lockdep_recursion;
1273 struct held_lock held_locks[MAX_LOCK_DEPTH];
1274#endif
1275
1276#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1277 unsigned int in_ubsan;
1278#endif
1279
1280 /* Journalling filesystem info: */
1281 void *journal_info;
1282
1283 /* Stacked block device info: */
1284 struct bio_list *bio_list;
1285
1286 /* Stack plugging: */
1287 struct blk_plug *plug;
1288
1289 /* VM state: */
1290 struct reclaim_state *reclaim_state;
1291
1292 struct io_context *io_context;
1293
1294#ifdef CONFIG_COMPACTION
1295 struct capture_control *capture_control;
1296#endif
1297 /* Ptrace state: */
1298 unsigned long ptrace_message;
1299 kernel_siginfo_t *last_siginfo;
1300
1301 struct task_io_accounting ioac;
1302#ifdef CONFIG_PSI
1303 /* Pressure stall state */
1304 unsigned int psi_flags;
1305#endif
1306#ifdef CONFIG_TASK_XACCT
1307 /* Accumulated RSS usage: */
1308 u64 acct_rss_mem1;
1309 /* Accumulated virtual memory usage: */
1310 u64 acct_vm_mem1;
1311 /* stime + utime since last update: */
1312 u64 acct_timexpd;
1313#endif
1314#ifdef CONFIG_CPUSETS
1315 /* Protected by ->alloc_lock: */
1316 nodemask_t mems_allowed;
1317 /* Sequence number to catch updates: */
1318 seqcount_spinlock_t mems_allowed_seq;
1319 int cpuset_mem_spread_rotor;
1320#endif
1321#ifdef CONFIG_CGROUPS
1322 /* Control Group info protected by css_set_lock: */
1323 struct css_set __rcu *cgroups;
1324 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1325 struct list_head cg_list;
1326#ifdef CONFIG_PREEMPT_RT
1327 struct llist_node cg_dead_lnode;
1328#endif /* CONFIG_PREEMPT_RT */
1329#endif /* CONFIG_CGROUPS */
1330#ifdef CONFIG_X86_CPU_RESCTRL
1331 u32 closid;
1332 u32 rmid;
1333#endif
1334#ifdef CONFIG_FUTEX
1335 struct robust_list_head __user *robust_list;
1336#ifdef CONFIG_COMPAT
1337 struct compat_robust_list_head __user *compat_robust_list;
1338#endif
1339 struct list_head pi_state_list;
1340 struct futex_pi_state *pi_state_cache;
1341 struct mutex futex_exit_mutex;
1342 unsigned int futex_state;
1343#endif
1344#ifdef CONFIG_PERF_EVENTS
1345 u8 perf_recursion[PERF_NR_CONTEXTS];
1346 struct perf_event_context *perf_event_ctxp;
1347 struct mutex perf_event_mutex;
1348 struct list_head perf_event_list;
1349 struct perf_ctx_data __rcu *perf_ctx_data;
1350#endif
1351#ifdef CONFIG_DEBUG_PREEMPT
1352 unsigned long preempt_disable_ip;
1353#endif
1354#ifdef CONFIG_NUMA
1355 /* Protected by alloc_lock: */
1356 struct mempolicy *mempolicy;
1357 short il_prev;
1358 u8 il_weight;
1359 short pref_node_fork;
1360#endif
1361#ifdef CONFIG_NUMA_BALANCING
1362 int numa_scan_seq;
1363 unsigned int numa_scan_period;
1364 unsigned int numa_scan_period_max;
1365 int numa_preferred_nid;
1366 unsigned long numa_migrate_retry;
1367 /* Migration stamp: */
1368 u64 node_stamp;
1369 u64 last_task_numa_placement;
1370 u64 last_sum_exec_runtime;
1371 struct callback_head numa_work;
1372
1373 /*
1374 * This pointer is only modified for current in syscall and
1375 * pagefault context (and for tasks being destroyed), so it can be read
1376 * from any of the following contexts:
1377 * - RCU read-side critical section
1378 * - current->numa_group from everywhere
1379 * - task's runqueue locked, task not running
1380 */
1381 struct numa_group __rcu *numa_group;
1382
1383 /*
1384 * numa_faults is an array split into four regions:
1385 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1386 * in this precise order.
1387 *
1388 * faults_memory: Exponential decaying average of faults on a per-node
1389 * basis. Scheduling placement decisions are made based on these
1390 * counts. The values remain static for the duration of a PTE scan.
1391 * faults_cpu: Track the nodes the process was running on when a NUMA
1392 * hinting fault was incurred.
1393 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1394 * during the current scan window. When the scan completes, the counts
1395 * in faults_memory and faults_cpu decay and these values are copied.
1396 */
1397 unsigned long *numa_faults;
1398 unsigned long total_numa_faults;
1399
1400 /*
1401 * numa_faults_locality tracks if faults recorded during the last
1402 * scan window were remote/local or failed to migrate. The task scan
1403 * period is adapted based on the locality of the faults with different
1404 * weights depending on whether they were shared or private faults
1405 */
1406 unsigned long numa_faults_locality[3];
1407
1408 unsigned long numa_pages_migrated;
1409#endif /* CONFIG_NUMA_BALANCING */
1410
1411 struct rseq_data rseq;
1412 struct sched_mm_cid mm_cid;
1413
1414 struct tlbflush_unmap_batch tlb_ubc;
1415
1416 /* Cache last used pipe for splice(): */
1417 struct pipe_inode_info *splice_pipe;
1418
1419 struct page_frag task_frag;
1420
1421#ifdef CONFIG_ARCH_HAS_LAZY_MMU_MODE
1422 struct lazy_mmu_state lazy_mmu_state;
1423#endif
1424
1425#ifdef CONFIG_TASK_DELAY_ACCT
1426 struct task_delay_info *delays;
1427#endif
1428
1429#ifdef CONFIG_FAULT_INJECTION
1430 int make_it_fail;
1431 unsigned int fail_nth;
1432#endif
1433 /*
1434 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1435 * balance_dirty_pages() for a dirty throttling pause:
1436 */
1437 int nr_dirtied;
1438 int nr_dirtied_pause;
1439 /* Start of a write-and-pause period: */
1440 unsigned long dirty_paused_when;
1441
1442#ifdef CONFIG_LATENCYTOP
1443 int latency_record_count;
1444 struct latency_record latency_record[LT_SAVECOUNT];
1445#endif
1446 /*
1447 * Time slack values; these are used to round up poll() and
1448 * select() etc timeout values. These are in nanoseconds.
1449 */
1450 u64 timer_slack_ns;
1451 u64 default_timer_slack_ns;
1452
1453#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1454 unsigned int kasan_depth;
1455#endif
1456
1457#ifdef CONFIG_KCSAN
1458 struct kcsan_ctx kcsan_ctx;
1459#ifdef CONFIG_TRACE_IRQFLAGS
1460 struct irqtrace_events kcsan_save_irqtrace;
1461#endif
1462#ifdef CONFIG_KCSAN_WEAK_MEMORY
1463 int kcsan_stack_depth;
1464#endif
1465#endif
1466
1467#ifdef CONFIG_KMSAN
1468 struct kmsan_ctx kmsan_ctx;
1469#endif
1470
1471#if IS_ENABLED(CONFIG_KUNIT)
1472 struct kunit *kunit_test;
1473#endif
1474
1475#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1476 /* Index of current stored address in ret_stack: */
1477 int curr_ret_stack;
1478 int curr_ret_depth;
1479
1480 /* Stack of return addresses for return function tracing: */
1481 unsigned long *ret_stack;
1482
1483 /* Timestamp for last schedule: */
1484 unsigned long long ftrace_timestamp;
1485 unsigned long long ftrace_sleeptime;
1486
1487 /*
1488 * Number of functions that haven't been traced
1489 * because of depth overrun:
1490 */
1491 atomic_t trace_overrun;
1492
1493 /* Pause tracing: */
1494 atomic_t tracing_graph_pause;
1495#endif
1496
1497#ifdef CONFIG_TRACING
1498 /* Bitmask and counter of trace recursion: */
1499 unsigned long trace_recursion;
1500#endif /* CONFIG_TRACING */
1501
1502#ifdef CONFIG_KCOV
1503 /* See kernel/kcov.c for more details. */
1504
1505 /* Coverage collection mode enabled for this task (0 if disabled): */
1506 unsigned int kcov_mode;
1507
1508 /* Size of the kcov_area: */
1509 unsigned int kcov_size;
1510
1511 /* Buffer for coverage collection: */
1512 void *kcov_area;
1513
1514 /* KCOV descriptor wired with this task or NULL: */
1515 struct kcov *kcov;
1516
1517 /* KCOV common handle for remote coverage collection: */
1518 u64 kcov_handle;
1519
1520 /* KCOV sequence number: */
1521 int kcov_sequence;
1522
1523 /* Collect coverage from softirq context: */
1524 unsigned int kcov_softirq;
1525#endif
1526
1527#ifdef CONFIG_MEMCG_V1
1528 struct mem_cgroup *memcg_in_oom;
1529#endif
1530
1531#ifdef CONFIG_MEMCG
1532 /* Number of pages to reclaim on returning to userland: */
1533 unsigned int memcg_nr_pages_over_high;
1534
1535 /* Used by memcontrol for targeted memcg charge: */
1536 struct mem_cgroup *active_memcg;
1537
1538 /* Cache for current->cgroups->memcg->nodeinfo[nid]->objcg lookups: */
1539 struct obj_cgroup *objcg;
1540#endif
1541
1542#ifdef CONFIG_BLK_CGROUP
1543 struct gendisk *throttle_disk;
1544#endif
1545
1546#ifdef CONFIG_UPROBES
1547 struct uprobe_task *utask;
1548#endif
1549#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1550 unsigned int sequential_io;
1551 unsigned int sequential_io_avg;
1552#endif
1553 struct kmap_ctrl kmap_ctrl;
1554#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1555 unsigned long task_state_change;
1556# ifdef CONFIG_PREEMPT_RT
1557 unsigned long saved_state_change;
1558# endif
1559#endif
1560 struct rcu_head rcu;
1561 refcount_t rcu_users;
1562 int pagefault_disabled;
1563#ifdef CONFIG_MMU
1564 struct task_struct *oom_reaper_list;
1565 struct timer_list oom_reaper_timer;
1566#endif
1567#ifdef CONFIG_VMAP_STACK
1568 struct vm_struct *stack_vm_area;
1569#endif
1570#ifdef CONFIG_THREAD_INFO_IN_TASK
1571 /* A live task holds one reference: */
1572 refcount_t stack_refcount;
1573#endif
1574#ifdef CONFIG_LIVEPATCH
1575 int patch_state;
1576#endif
1577#ifdef CONFIG_SECURITY
1578 /* Used by LSM modules for access restriction: */
1579 void *security;
1580#endif
1581#ifdef CONFIG_BPF_SYSCALL
1582 /* Used by BPF task local storage */
1583 struct bpf_local_storage __rcu *bpf_storage;
1584 /* Used for BPF run context */
1585 struct bpf_run_ctx *bpf_ctx;
1586#endif
1587 /* Used by BPF for per-TASK xdp storage */
1588 struct bpf_net_context *bpf_net_context;
1589
1590#ifdef CONFIG_KSTACK_ERASE
1591 unsigned long lowest_stack;
1592#endif
1593#ifdef CONFIG_KSTACK_ERASE_METRICS
1594 unsigned long prev_lowest_stack;
1595#endif
1596
1597#ifdef CONFIG_X86_MCE
1598 void __user *mce_vaddr;
1599 __u64 mce_kflags;
1600 u64 mce_addr;
1601 __u64 mce_ripv : 1,
1602 mce_whole_page : 1,
1603 __mce_reserved : 62;
1604 struct callback_head mce_kill_me;
1605 int mce_count;
1606#endif
1607
1608#ifdef CONFIG_KRETPROBES
1609 struct llist_head kretprobe_instances;
1610#endif
1611#ifdef CONFIG_RETHOOK
1612 struct llist_head rethooks;
1613#endif
1614
1615#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1616 /*
1617 * If L1D flush is supported on mm context switch
1618 * then we use this callback head to queue kill work
1619 * to kill tasks that are not running on SMT disabled
1620 * cores
1621 */
1622 struct callback_head l1d_flush_kill;
1623#endif
1624
1625#ifdef CONFIG_RV
1626 /*
1627 * Per-task RV monitor, fixed in CONFIG_RV_PER_TASK_MONITORS.
1628 * If memory becomes a concern, we can think about a dynamic method.
1629 */
1630 union rv_task_monitor rv[CONFIG_RV_PER_TASK_MONITORS];
1631#endif
1632
1633#ifdef CONFIG_USER_EVENTS
1634 struct user_event_mm *user_event_mm;
1635#endif
1636
1637#ifdef CONFIG_UNWIND_USER
1638 struct unwind_task_info unwind_info;
1639#endif
1640
1641 /* CPU-specific state of this task: */
1642 struct thread_struct thread;
1643
1644 /*
1645 * New fields for task_struct should be added above here, so that
1646 * they are included in the randomized portion of task_struct.
1647 */
1648 randomized_struct_fields_end
1649} __attribute__ ((aligned (64)));
1650
1651#ifdef CONFIG_SCHED_PROXY_EXEC
1652DECLARE_STATIC_KEY_TRUE(__sched_proxy_exec);
1653static inline bool sched_proxy_exec(void)
1654{
1655 return static_branch_likely(&__sched_proxy_exec);
1656}
1657#else
1658static inline bool sched_proxy_exec(void)
1659{
1660 return false;
1661}
1662#endif
1663
1664#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1665#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1666
1667static inline unsigned int __task_state_index(unsigned int tsk_state,
1668 unsigned int tsk_exit_state)
1669{
1670 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1671
1672 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1673
1674 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1675 state = TASK_REPORT_IDLE;
1676
1677 /*
1678 * We're lying here, but rather than expose a completely new task state
1679 * to userspace, we can make this appear as if the task has gone through
1680 * a regular rt_mutex_lock() call.
1681 * Report frozen tasks as uninterruptible.
1682 */
1683 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
1684 state = TASK_UNINTERRUPTIBLE;
1685
1686 return fls(state);
1687}
1688
1689static inline unsigned int task_state_index(struct task_struct *tsk)
1690{
1691 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1692}
1693
1694static inline char task_index_to_char(unsigned int state)
1695{
1696 static const char state_char[] = "RSDTtXZPI";
1697
1698 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1699
1700 return state_char[state];
1701}
1702
1703static inline char task_state_to_char(struct task_struct *tsk)
1704{
1705 return task_index_to_char(task_state_index(tsk));
1706}
1707
1708#ifdef CONFIG_ARCH_HAS_LAZY_MMU_MODE
1709/**
1710 * __task_lazy_mmu_mode_active() - Test the lazy MMU mode state for a task.
1711 * @tsk: The task to check.
1712 *
1713 * Test whether @tsk has its lazy MMU mode state set to active (i.e. enabled
1714 * and not paused).
1715 *
1716 * This function only considers the state saved in task_struct; to test whether
1717 * current actually is in lazy MMU mode, is_lazy_mmu_mode_active() should be
1718 * used instead.
1719 *
1720 * This function is intended for architectures that implement the lazy MMU
1721 * mode; it must not be called from generic code.
1722 */
1723static inline bool __task_lazy_mmu_mode_active(struct task_struct *tsk)
1724{
1725 struct lazy_mmu_state *state = &tsk->lazy_mmu_state;
1726
1727 return state->enable_count > 0 && state->pause_count == 0;
1728}
1729
1730/**
1731 * is_lazy_mmu_mode_active() - Test whether we are currently in lazy MMU mode.
1732 *
1733 * Test whether the current context is in lazy MMU mode. This is true if both:
1734 * 1. We are not in interrupt context
1735 * 2. Lazy MMU mode is active for the current task
1736 *
1737 * This function is intended for architectures that implement the lazy MMU
1738 * mode; it must not be called from generic code.
1739 */
1740static inline bool is_lazy_mmu_mode_active(void)
1741{
1742 if (in_interrupt())
1743 return false;
1744
1745 return __task_lazy_mmu_mode_active(current);
1746}
1747#endif
1748
1749extern struct pid *cad_pid;
1750
1751/*
1752 * Per process flags
1753 */
1754#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1755#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1756#define PF_EXITING 0x00000004 /* Getting shut down */
1757#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1758#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1759#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1760#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1761#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1762#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1763#define PF_DUMPCORE 0x00000200 /* Dumped core */
1764#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1765#define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1766#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1767#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1768#define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1769#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1770#define PF_KCOMPACTD 0x00010000 /* I am kcompactd */
1771#define PF_KSWAPD 0x00020000 /* I am kswapd */
1772#define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1773#define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1774#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1775 * I am cleaning dirty pages from some other bdi. */
1776#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1777#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1778#define PF__HOLE__00800000 0x00800000
1779#define PF__HOLE__01000000 0x01000000
1780#define PF__HOLE__02000000 0x02000000
1781#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1782#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1783#define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1784 * See memalloc_pin_save() */
1785#define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1786#define PF__HOLE__40000000 0x40000000
1787#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1788
1789/*
1790 * Only the _current_ task can read/write to tsk->flags, but other
1791 * tasks can access tsk->flags in readonly mode for example
1792 * with tsk_used_math (like during threaded core dumping).
1793 * There is however an exception to this rule during ptrace
1794 * or during fork: the ptracer task is allowed to write to the
1795 * child->flags of its traced child (same goes for fork, the parent
1796 * can write to the child->flags), because we're guaranteed the
1797 * child is not running and in turn not changing child->flags
1798 * at the same time the parent does it.
1799 */
1800#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1801#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1802#define clear_used_math() clear_stopped_child_used_math(current)
1803#define set_used_math() set_stopped_child_used_math(current)
1804
1805#define conditional_stopped_child_used_math(condition, child) \
1806 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1807
1808#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1809
1810#define copy_to_stopped_child_used_math(child) \
1811 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1812
1813/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1814#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1815#define used_math() tsk_used_math(current)
1816
1817static __always_inline bool is_percpu_thread(void)
1818{
1819 return (current->flags & PF_NO_SETAFFINITY) &&
1820 (current->nr_cpus_allowed == 1);
1821}
1822
1823static __always_inline bool is_user_task(struct task_struct *task)
1824{
1825 return task->mm && !(task->flags & (PF_KTHREAD | PF_USER_WORKER));
1826}
1827
1828/* Per-process atomic flags. */
1829#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1830#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1831#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1832#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1833#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1834#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1835#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1836#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1837
1838#define TASK_PFA_TEST(name, func) \
1839 static inline bool task_##func(struct task_struct *p) \
1840 { return test_bit(PFA_##name, &p->atomic_flags); }
1841
1842#define TASK_PFA_SET(name, func) \
1843 static inline void task_set_##func(struct task_struct *p) \
1844 { set_bit(PFA_##name, &p->atomic_flags); }
1845
1846#define TASK_PFA_CLEAR(name, func) \
1847 static inline void task_clear_##func(struct task_struct *p) \
1848 { clear_bit(PFA_##name, &p->atomic_flags); }
1849
1850TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1851TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1852
1853TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1854TASK_PFA_SET(SPREAD_PAGE, spread_page)
1855TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1856
1857TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1858TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1859TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1860
1861TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1862TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1863TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1864
1865TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1866TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1867TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1868
1869TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1870TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1871
1872TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1873TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1874TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1875
1876TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1877TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1878
1879static inline void
1880current_restore_flags(unsigned long orig_flags, unsigned long flags)
1881{
1882 current->flags &= ~flags;
1883 current->flags |= orig_flags & flags;
1884}
1885
1886extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1887extern int task_can_attach(struct task_struct *p);
1888extern int dl_bw_alloc(int cpu, u64 dl_bw);
1889extern void dl_bw_free(int cpu, u64 dl_bw);
1890
1891/* set_cpus_allowed_force() - consider using set_cpus_allowed_ptr() instead */
1892extern void set_cpus_allowed_force(struct task_struct *p, const struct cpumask *new_mask);
1893
1894/**
1895 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1896 * @p: the task
1897 * @new_mask: CPU affinity mask
1898 *
1899 * Return: zero if successful, or a negative error code
1900 */
1901extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1902extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1903extern void release_user_cpus_ptr(struct task_struct *p);
1904extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1905extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1906extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1907
1908extern int yield_to(struct task_struct *p, bool preempt);
1909extern void set_user_nice(struct task_struct *p, long nice);
1910extern int task_prio(const struct task_struct *p);
1911
1912/**
1913 * task_nice - return the nice value of a given task.
1914 * @p: the task in question.
1915 *
1916 * Return: The nice value [ -20 ... 0 ... 19 ].
1917 */
1918static inline int task_nice(const struct task_struct *p)
1919{
1920 return PRIO_TO_NICE((p)->static_prio);
1921}
1922
1923extern int can_nice(const struct task_struct *p, const int nice);
1924extern int task_curr(const struct task_struct *p);
1925extern int idle_cpu(int cpu);
1926extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1927extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1928extern void sched_set_fifo(struct task_struct *p);
1929extern void sched_set_fifo_low(struct task_struct *p);
1930extern void sched_set_fifo_secondary(struct task_struct *p);
1931extern void sched_set_normal(struct task_struct *p, int nice);
1932extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1933extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1934extern struct task_struct *idle_task(int cpu);
1935
1936/**
1937 * is_idle_task - is the specified task an idle task?
1938 * @p: the task in question.
1939 *
1940 * Return: 1 if @p is an idle task. 0 otherwise.
1941 */
1942static __always_inline bool is_idle_task(const struct task_struct *p)
1943{
1944 return !!(p->flags & PF_IDLE);
1945}
1946
1947extern struct task_struct *curr_task(int cpu);
1948extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1949
1950void yield(void);
1951
1952union thread_union {
1953 struct task_struct task;
1954#ifndef CONFIG_THREAD_INFO_IN_TASK
1955 struct thread_info thread_info;
1956#endif
1957 unsigned long stack[THREAD_SIZE/sizeof(long)];
1958};
1959
1960#ifndef CONFIG_THREAD_INFO_IN_TASK
1961extern struct thread_info init_thread_info;
1962#endif
1963
1964extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1965
1966#ifdef CONFIG_THREAD_INFO_IN_TASK
1967# define task_thread_info(task) (&(task)->thread_info)
1968#else
1969# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1970#endif
1971
1972/*
1973 * find a task by one of its numerical ids
1974 *
1975 * find_task_by_pid_ns():
1976 * finds a task by its pid in the specified namespace
1977 * find_task_by_vpid():
1978 * finds a task by its virtual pid
1979 *
1980 * see also find_vpid() etc in include/linux/pid.h
1981 */
1982
1983extern struct task_struct *find_task_by_vpid(pid_t nr);
1984extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1985
1986/*
1987 * find a task by its virtual pid and get the task struct
1988 */
1989extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1990
1991extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1992extern int wake_up_process(struct task_struct *tsk);
1993extern void wake_up_new_task(struct task_struct *tsk);
1994
1995extern void kick_process(struct task_struct *tsk);
1996
1997extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1998#define set_task_comm(tsk, from) ({ \
1999 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \
2000 __set_task_comm(tsk, from, false); \
2001})
2002
2003/*
2004 * - Why not use task_lock()?
2005 * User space can randomly change their names anyway, so locking for readers
2006 * doesn't make sense. For writers, locking is probably necessary, as a race
2007 * condition could lead to long-term mixed results.
2008 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is
2009 * always NUL-terminated and zero-padded. Therefore the race condition between
2010 * reader and writer is not an issue.
2011 *
2012 * - BUILD_BUG_ON() can help prevent the buf from being truncated.
2013 * Since the callers don't perform any return value checks, this safeguard is
2014 * necessary.
2015 */
2016#define get_task_comm(buf, tsk) ({ \
2017 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \
2018 strscpy_pad(buf, (tsk)->comm); \
2019 buf; \
2020})
2021
2022static __always_inline void scheduler_ipi(void)
2023{
2024 /*
2025 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2026 * TIF_NEED_RESCHED remotely (for the first time) will also send
2027 * this IPI.
2028 */
2029 preempt_fold_need_resched();
2030}
2031
2032extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2033
2034/*
2035 * Set thread flags in other task's structures.
2036 * See asm/thread_info.h for TIF_xxxx flags available:
2037 */
2038static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2039{
2040 set_ti_thread_flag(task_thread_info(tsk), flag);
2041}
2042
2043static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2044{
2045 clear_ti_thread_flag(task_thread_info(tsk), flag);
2046}
2047
2048static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2049 bool value)
2050{
2051 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2052}
2053
2054static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2055{
2056 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2057}
2058
2059static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2060{
2061 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2062}
2063
2064static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2065{
2066 return test_ti_thread_flag(task_thread_info(tsk), flag);
2067}
2068
2069static inline void set_tsk_need_resched(struct task_struct *tsk)
2070{
2071 if (tracepoint_enabled(sched_set_need_resched_tp) &&
2072 !test_tsk_thread_flag(tsk, TIF_NEED_RESCHED))
2073 __trace_set_need_resched(tsk, TIF_NEED_RESCHED);
2074 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2075}
2076
2077static inline void clear_tsk_need_resched(struct task_struct *tsk)
2078{
2079 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2080 (atomic_long_t *)&task_thread_info(tsk)->flags);
2081}
2082
2083static inline int test_tsk_need_resched(struct task_struct *tsk)
2084{
2085 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2086}
2087
2088static inline void set_need_resched_current(void)
2089{
2090 lockdep_assert_irqs_disabled();
2091 set_tsk_need_resched(current);
2092 set_preempt_need_resched();
2093}
2094
2095/*
2096 * cond_resched() and cond_resched_lock(): latency reduction via
2097 * explicit rescheduling in places that are safe. The return
2098 * value indicates whether a reschedule was done in fact.
2099 * cond_resched_lock() will drop the spinlock before scheduling,
2100 */
2101#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2102extern int __cond_resched(void);
2103
2104#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2105
2106DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2107
2108static __always_inline int _cond_resched(void)
2109{
2110 return static_call_mod(cond_resched)();
2111}
2112
2113#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2114
2115extern int dynamic_cond_resched(void);
2116
2117static __always_inline int _cond_resched(void)
2118{
2119 return dynamic_cond_resched();
2120}
2121
2122#else /* !CONFIG_PREEMPTION */
2123
2124static inline int _cond_resched(void)
2125{
2126 return __cond_resched();
2127}
2128
2129#endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2130
2131#else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2132
2133static inline int _cond_resched(void)
2134{
2135 return 0;
2136}
2137
2138#endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2139
2140#define cond_resched() ({ \
2141 __might_resched(__FILE__, __LINE__, 0); \
2142 _cond_resched(); \
2143})
2144
2145extern int __cond_resched_lock(spinlock_t *lock) __must_hold(lock);
2146extern int __cond_resched_rwlock_read(rwlock_t *lock) __must_hold_shared(lock);
2147extern int __cond_resched_rwlock_write(rwlock_t *lock) __must_hold(lock);
2148
2149#define MIGHT_RESCHED_RCU_SHIFT 8
2150#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2151
2152#ifndef CONFIG_PREEMPT_RT
2153/*
2154 * Non RT kernels have an elevated preempt count due to the held lock,
2155 * but are not allowed to be inside a RCU read side critical section
2156 */
2157# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2158#else
2159/*
2160 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2161 * cond_resched*lock() has to take that into account because it checks for
2162 * preempt_count() and rcu_preempt_depth().
2163 */
2164# define PREEMPT_LOCK_RESCHED_OFFSETS \
2165 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2166#endif
2167
2168#define cond_resched_lock(lock) ({ \
2169 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2170 __cond_resched_lock(lock); \
2171})
2172
2173#define cond_resched_rwlock_read(lock) ({ \
2174 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2175 __cond_resched_rwlock_read(lock); \
2176})
2177
2178#define cond_resched_rwlock_write(lock) ({ \
2179 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2180 __cond_resched_rwlock_write(lock); \
2181})
2182
2183#ifndef CONFIG_PREEMPT_RT
2184
2185/*
2186 * With proxy exec, if a task has been proxy-migrated, it may be a donor
2187 * on a cpu that it can't actually run on. Thus we need a special state
2188 * to denote that the task is being woken, but that it needs to be
2189 * evaluated for return-migration before it is run. So if the task is
2190 * blocked_on PROXY_WAKING, return migrate it before running it.
2191 */
2192#define PROXY_WAKING ((struct mutex *)(-1L))
2193
2194static inline struct mutex *__get_task_blocked_on(struct task_struct *p)
2195{
2196 lockdep_assert_held_once(&p->blocked_lock);
2197 return p->blocked_on == PROXY_WAKING ? NULL : p->blocked_on;
2198}
2199
2200static inline void __set_task_blocked_on(struct task_struct *p, struct mutex *m)
2201{
2202 WARN_ON_ONCE(!m);
2203 /* The task should only be setting itself as blocked */
2204 WARN_ON_ONCE(p != current);
2205 /* Currently we serialize blocked_on under the task::blocked_lock */
2206 lockdep_assert_held_once(&p->blocked_lock);
2207 /*
2208 * Check ensure we don't overwrite existing mutex value
2209 * with a different mutex. Note, setting it to the same
2210 * lock repeatedly is ok.
2211 */
2212 WARN_ON_ONCE(p->blocked_on && p->blocked_on != m);
2213 p->blocked_on = m;
2214}
2215
2216static inline void __clear_task_blocked_on(struct task_struct *p, struct mutex *m)
2217{
2218 /* Currently we serialize blocked_on under the task::blocked_lock */
2219 lockdep_assert_held_once(&p->blocked_lock);
2220 /*
2221 * There may be cases where we re-clear already cleared
2222 * blocked_on relationships, but make sure we are not
2223 * clearing the relationship with a different lock.
2224 */
2225 WARN_ON_ONCE(m && p->blocked_on && p->blocked_on != m && p->blocked_on != PROXY_WAKING);
2226 p->blocked_on = NULL;
2227}
2228
2229static inline void clear_task_blocked_on(struct task_struct *p, struct mutex *m)
2230{
2231 guard(raw_spinlock_irqsave)(&p->blocked_lock);
2232 __clear_task_blocked_on(p, m);
2233}
2234
2235static inline void __set_task_blocked_on_waking(struct task_struct *p, struct mutex *m)
2236{
2237 /* Currently we serialize blocked_on under the task::blocked_lock */
2238 lockdep_assert_held_once(&p->blocked_lock);
2239
2240 if (!sched_proxy_exec()) {
2241 __clear_task_blocked_on(p, m);
2242 return;
2243 }
2244
2245 /* Don't set PROXY_WAKING if blocked_on was already cleared */
2246 if (!p->blocked_on)
2247 return;
2248 /*
2249 * There may be cases where we set PROXY_WAKING on tasks that were
2250 * already set to waking, but make sure we are not changing
2251 * the relationship with a different lock.
2252 */
2253 WARN_ON_ONCE(m && p->blocked_on != m && p->blocked_on != PROXY_WAKING);
2254 p->blocked_on = PROXY_WAKING;
2255}
2256
2257static inline void set_task_blocked_on_waking(struct task_struct *p, struct mutex *m)
2258{
2259 guard(raw_spinlock_irqsave)(&p->blocked_lock);
2260 __set_task_blocked_on_waking(p, m);
2261}
2262
2263#else
2264static inline void __clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m)
2265{
2266}
2267
2268static inline void clear_task_blocked_on(struct task_struct *p, struct rt_mutex *m)
2269{
2270}
2271
2272static inline void __set_task_blocked_on_waking(struct task_struct *p, struct rt_mutex *m)
2273{
2274}
2275
2276static inline void set_task_blocked_on_waking(struct task_struct *p, struct rt_mutex *m)
2277{
2278}
2279#endif /* !CONFIG_PREEMPT_RT */
2280
2281static __always_inline bool need_resched(void)
2282{
2283 return unlikely(tif_need_resched());
2284}
2285
2286/*
2287 * Wrappers for p->thread_info->cpu access. No-op on UP.
2288 */
2289#ifdef CONFIG_SMP
2290
2291static inline unsigned int task_cpu(const struct task_struct *p)
2292{
2293 return READ_ONCE(task_thread_info(p)->cpu);
2294}
2295
2296extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2297
2298#else
2299
2300static inline unsigned int task_cpu(const struct task_struct *p)
2301{
2302 return 0;
2303}
2304
2305static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2306{
2307}
2308
2309#endif /* CONFIG_SMP */
2310
2311static inline bool task_is_runnable(struct task_struct *p)
2312{
2313 return p->on_rq && !p->se.sched_delayed;
2314}
2315
2316extern bool sched_task_on_rq(struct task_struct *p);
2317extern unsigned long get_wchan(struct task_struct *p);
2318extern struct task_struct *cpu_curr_snapshot(int cpu);
2319
2320/*
2321 * In order to reduce various lock holder preemption latencies provide an
2322 * interface to see if a vCPU is currently running or not.
2323 *
2324 * This allows us to terminate optimistic spin loops and block, analogous to
2325 * the native optimistic spin heuristic of testing if the lock owner task is
2326 * running or not.
2327 */
2328#ifndef vcpu_is_preempted
2329static inline bool vcpu_is_preempted(int cpu)
2330{
2331 return false;
2332}
2333#endif
2334
2335extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2336extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2337
2338#ifndef TASK_SIZE_OF
2339#define TASK_SIZE_OF(tsk) TASK_SIZE
2340#endif
2341
2342static inline bool owner_on_cpu(struct task_struct *owner)
2343{
2344 /*
2345 * As lock holder preemption issue, we both skip spinning if
2346 * task is not on cpu or its cpu is preempted
2347 */
2348 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2349}
2350
2351/* Returns effective CPU energy utilization, as seen by the scheduler */
2352unsigned long sched_cpu_util(int cpu);
2353
2354#ifdef CONFIG_SCHED_CORE
2355extern void sched_core_free(struct task_struct *tsk);
2356extern void sched_core_fork(struct task_struct *p);
2357extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2358 unsigned long uaddr);
2359extern int sched_core_idle_cpu(int cpu);
2360#else
2361static inline void sched_core_free(struct task_struct *tsk) { }
2362static inline void sched_core_fork(struct task_struct *p) { }
2363static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2364#endif
2365
2366extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2367
2368#ifdef CONFIG_MEM_ALLOC_PROFILING
2369static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2370{
2371 swap(current->alloc_tag, tag);
2372 return tag;
2373}
2374
2375static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2376{
2377#ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2378 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2379#endif
2380 current->alloc_tag = old;
2381}
2382#else
2383#define alloc_tag_save(_tag) NULL
2384#define alloc_tag_restore(_tag, _old) do {} while (0)
2385#endif
2386
2387/* Avoids recursive inclusion hell */
2388#ifdef CONFIG_SCHED_MM_CID
2389void sched_mm_cid_before_execve(struct task_struct *t);
2390void sched_mm_cid_after_execve(struct task_struct *t);
2391void sched_mm_cid_exit(struct task_struct *t);
2392static __always_inline int task_mm_cid(struct task_struct *t)
2393{
2394 return t->mm_cid.cid & ~(MM_CID_ONCPU | MM_CID_TRANSIT);
2395}
2396#else
2397static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2398static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2399static inline void sched_mm_cid_exit(struct task_struct *t) { }
2400static __always_inline int task_mm_cid(struct task_struct *t)
2401{
2402 /*
2403 * Use the processor id as a fall-back when the mm cid feature is
2404 * disabled. This provides functional per-cpu data structure accesses
2405 * in user-space, althrough it won't provide the memory usage benefits.
2406 */
2407 return task_cpu(t);
2408}
2409#endif
2410
2411#ifndef MODULE
2412#ifndef COMPILE_OFFSETS
2413
2414extern void ___migrate_enable(void);
2415
2416struct rq;
2417DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
2418
2419/*
2420 * The "struct rq" is not available here, so we can't access the
2421 * "runqueues" with this_cpu_ptr(), as the compilation will fail in
2422 * this_cpu_ptr() -> raw_cpu_ptr() -> __verify_pcpu_ptr():
2423 * typeof((ptr) + 0)
2424 *
2425 * So use arch_raw_cpu_ptr()/PERCPU_PTR() directly here.
2426 */
2427#ifdef CONFIG_SMP
2428#define this_rq_raw() arch_raw_cpu_ptr(&runqueues)
2429#else
2430#define this_rq_raw() PERCPU_PTR(&runqueues)
2431#endif
2432#define this_rq_pinned() (*(unsigned int *)((void *)this_rq_raw() + RQ_nr_pinned))
2433
2434static inline void __migrate_enable(void)
2435{
2436 struct task_struct *p = current;
2437
2438#ifdef CONFIG_DEBUG_PREEMPT
2439 /*
2440 * Check both overflow from migrate_disable() and superfluous
2441 * migrate_enable().
2442 */
2443 if (WARN_ON_ONCE((s16)p->migration_disabled <= 0))
2444 return;
2445#endif
2446
2447 if (p->migration_disabled > 1) {
2448 p->migration_disabled--;
2449 return;
2450 }
2451
2452 /*
2453 * Ensure stop_task runs either before or after this, and that
2454 * __set_cpus_allowed_ptr(SCA_MIGRATE_ENABLE) doesn't schedule().
2455 */
2456 guard(preempt)();
2457 if (unlikely(p->cpus_ptr != &p->cpus_mask))
2458 ___migrate_enable();
2459 /*
2460 * Mustn't clear migration_disabled() until cpus_ptr points back at the
2461 * regular cpus_mask, otherwise things that race (eg.
2462 * select_fallback_rq) get confused.
2463 */
2464 barrier();
2465 p->migration_disabled = 0;
2466 this_rq_pinned()--;
2467}
2468
2469static inline void __migrate_disable(void)
2470{
2471 struct task_struct *p = current;
2472
2473 if (p->migration_disabled) {
2474#ifdef CONFIG_DEBUG_PREEMPT
2475 /*
2476 *Warn about overflow half-way through the range.
2477 */
2478 WARN_ON_ONCE((s16)p->migration_disabled < 0);
2479#endif
2480 p->migration_disabled++;
2481 return;
2482 }
2483
2484 guard(preempt)();
2485 this_rq_pinned()++;
2486 p->migration_disabled = 1;
2487}
2488#else /* !COMPILE_OFFSETS */
2489static inline void __migrate_disable(void) { }
2490static inline void __migrate_enable(void) { }
2491#endif /* !COMPILE_OFFSETS */
2492
2493/*
2494 * So that it is possible to not export the runqueues variable, define and
2495 * export migrate_enable/migrate_disable in kernel/sched/core.c too, and use
2496 * them for the modules. The macro "INSTANTIATE_EXPORTED_MIGRATE_DISABLE" will
2497 * be defined in kernel/sched/core.c.
2498 */
2499#ifndef INSTANTIATE_EXPORTED_MIGRATE_DISABLE
2500static __always_inline void migrate_disable(void)
2501{
2502 __migrate_disable();
2503}
2504
2505static __always_inline void migrate_enable(void)
2506{
2507 __migrate_enable();
2508}
2509#else /* INSTANTIATE_EXPORTED_MIGRATE_DISABLE */
2510extern void migrate_disable(void);
2511extern void migrate_enable(void);
2512#endif /* INSTANTIATE_EXPORTED_MIGRATE_DISABLE */
2513
2514#else /* MODULE */
2515extern void migrate_disable(void);
2516extern void migrate_enable(void);
2517#endif /* MODULE */
2518
2519DEFINE_LOCK_GUARD_0(migrate, migrate_disable(), migrate_enable())
2520
2521#endif