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.h>
17
18 #include <linux/cache.h>
19 #include <linux/irqflags_types.h>
20 #include <linux/smp_types.h>
21 #include <linux/pid_types.h>
22 #include <linux/sem_types.h>
23 #include <linux/shm.h>
24 #include <linux/kmsan_types.h>
25 #include <linux/mutex_types.h>
26 #include <linux/plist_types.h>
27 #include <linux/hrtimer_types.h>
28 #include <linux/timer_types.h>
29 #include <linux/seccomp_types.h>
30 #include <linux/nodemask_types.h>
31 #include <linux/refcount_types.h>
32 #include <linux/resource.h>
33 #include <linux/latencytop.h>
34 #include <linux/sched/prio.h>
35 #include <linux/sched/types.h>
36 #include <linux/signal_types.h>
37 #include <linux/syscall_user_dispatch_types.h>
38 #include <linux/mm_types_task.h>
39 #include <linux/task_io_accounting.h>
40 #include <linux/posix-timers_types.h>
41 #include <linux/restart_block.h>
42 #include <uapi/linux/rseq.h>
43 #include <linux/seqlock_types.h>
44 #include <linux/kcsan.h>
45 #include <linux/rv.h>
46 #include <linux/livepatch_sched.h>
47 #include <linux/uidgid_types.h>
48 #include <asm/kmap_size.h>
49
50 /* task_struct member predeclarations (sorted alphabetically): */
51 struct audit_context;
52 struct bio_list;
53 struct blk_plug;
54 struct bpf_local_storage;
55 struct bpf_run_ctx;
56 struct capture_control;
57 struct cfs_rq;
58 struct fs_struct;
59 struct futex_pi_state;
60 struct io_context;
61 struct io_uring_task;
62 struct mempolicy;
63 struct nameidata;
64 struct nsproxy;
65 struct perf_event_context;
66 struct pid_namespace;
67 struct pipe_inode_info;
68 struct rcu_node;
69 struct reclaim_state;
70 struct robust_list_head;
71 struct root_domain;
72 struct rq;
73 struct sched_attr;
74 struct sched_dl_entity;
75 struct seq_file;
76 struct sighand_struct;
77 struct signal_struct;
78 struct task_delay_info;
79 struct task_group;
80 struct task_struct;
81 struct user_event_mm;
82
83 /*
84 * Task state bitmask. NOTE! These bits are also
85 * encoded in fs/proc/array.c: get_task_state().
86 *
87 * We have two separate sets of flags: task->__state
88 * is about runnability, while task->exit_state are
89 * about the task exiting. Confusing, but this way
90 * modifying one set can't modify the other one by
91 * mistake.
92 */
93
94 /* Used in tsk->__state: */
95 #define TASK_RUNNING 0x00000000
96 #define TASK_INTERRUPTIBLE 0x00000001
97 #define TASK_UNINTERRUPTIBLE 0x00000002
98 #define __TASK_STOPPED 0x00000004
99 #define __TASK_TRACED 0x00000008
100 /* Used in tsk->exit_state: */
101 #define EXIT_DEAD 0x00000010
102 #define EXIT_ZOMBIE 0x00000020
103 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
104 /* Used in tsk->__state again: */
105 #define TASK_PARKED 0x00000040
106 #define TASK_DEAD 0x00000080
107 #define TASK_WAKEKILL 0x00000100
108 #define TASK_WAKING 0x00000200
109 #define TASK_NOLOAD 0x00000400
110 #define TASK_NEW 0x00000800
111 #define TASK_RTLOCK_WAIT 0x00001000
112 #define TASK_FREEZABLE 0x00002000
113 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
114 #define TASK_FROZEN 0x00008000
115 #define TASK_STATE_MAX 0x00010000
116
117 #define TASK_ANY (TASK_STATE_MAX-1)
118
119 /*
120 * DO NOT ADD ANY NEW USERS !
121 */
122 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
123
124 /* Convenience macros for the sake of set_current_state: */
125 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
126 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
127 #define TASK_TRACED __TASK_TRACED
128
129 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
130
131 /* Convenience macros for the sake of wake_up(): */
132 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
133
134 /* get_task_state(): */
135 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
136 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
137 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
138 TASK_PARKED)
139
140 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
141
142 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
143 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
144 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
145
146 /*
147 * Special states are those that do not use the normal wait-loop pattern. See
148 * the comment with set_special_state().
149 */
150 #define is_special_task_state(state) \
151 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
152
153 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
154 # define debug_normal_state_change(state_value) \
155 do { \
156 WARN_ON_ONCE(is_special_task_state(state_value)); \
157 current->task_state_change = _THIS_IP_; \
158 } while (0)
159
160 # define debug_special_state_change(state_value) \
161 do { \
162 WARN_ON_ONCE(!is_special_task_state(state_value)); \
163 current->task_state_change = _THIS_IP_; \
164 } while (0)
165
166 # define debug_rtlock_wait_set_state() \
167 do { \
168 current->saved_state_change = current->task_state_change;\
169 current->task_state_change = _THIS_IP_; \
170 } while (0)
171
172 # define debug_rtlock_wait_restore_state() \
173 do { \
174 current->task_state_change = current->saved_state_change;\
175 } while (0)
176
177 #else
178 # define debug_normal_state_change(cond) do { } while (0)
179 # define debug_special_state_change(cond) do { } while (0)
180 # define debug_rtlock_wait_set_state() do { } while (0)
181 # define debug_rtlock_wait_restore_state() do { } while (0)
182 #endif
183
184 /*
185 * set_current_state() includes a barrier so that the write of current->__state
186 * is correctly serialised wrt the caller's subsequent test of whether to
187 * actually sleep:
188 *
189 * for (;;) {
190 * set_current_state(TASK_UNINTERRUPTIBLE);
191 * if (CONDITION)
192 * break;
193 *
194 * schedule();
195 * }
196 * __set_current_state(TASK_RUNNING);
197 *
198 * If the caller does not need such serialisation (because, for instance, the
199 * CONDITION test and condition change and wakeup are under the same lock) then
200 * use __set_current_state().
201 *
202 * The above is typically ordered against the wakeup, which does:
203 *
204 * CONDITION = 1;
205 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
206 *
207 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
208 * accessing p->__state.
209 *
210 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
211 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
212 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
213 *
214 * However, with slightly different timing the wakeup TASK_RUNNING store can
215 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
216 * a problem either because that will result in one extra go around the loop
217 * and our @cond test will save the day.
218 *
219 * Also see the comments of try_to_wake_up().
220 */
221 #define __set_current_state(state_value) \
222 do { \
223 debug_normal_state_change((state_value)); \
224 WRITE_ONCE(current->__state, (state_value)); \
225 } while (0)
226
227 #define set_current_state(state_value) \
228 do { \
229 debug_normal_state_change((state_value)); \
230 smp_store_mb(current->__state, (state_value)); \
231 } while (0)
232
233 /*
234 * set_special_state() should be used for those states when the blocking task
235 * can not use the regular condition based wait-loop. In that case we must
236 * serialize against wakeups such that any possible in-flight TASK_RUNNING
237 * stores will not collide with our state change.
238 */
239 #define set_special_state(state_value) \
240 do { \
241 unsigned long flags; /* may shadow */ \
242 \
243 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
244 debug_special_state_change((state_value)); \
245 WRITE_ONCE(current->__state, (state_value)); \
246 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
247 } while (0)
248
249 /*
250 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
251 *
252 * RT's spin/rwlock substitutions are state preserving. The state of the
253 * task when blocking on the lock is saved in task_struct::saved_state and
254 * restored after the lock has been acquired. These operations are
255 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
256 * lock related wakeups while the task is blocked on the lock are
257 * redirected to operate on task_struct::saved_state to ensure that these
258 * are not dropped. On restore task_struct::saved_state is set to
259 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
260 *
261 * The lock operation looks like this:
262 *
263 * current_save_and_set_rtlock_wait_state();
264 * for (;;) {
265 * if (try_lock())
266 * break;
267 * raw_spin_unlock_irq(&lock->wait_lock);
268 * schedule_rtlock();
269 * raw_spin_lock_irq(&lock->wait_lock);
270 * set_current_state(TASK_RTLOCK_WAIT);
271 * }
272 * current_restore_rtlock_saved_state();
273 */
274 #define current_save_and_set_rtlock_wait_state() \
275 do { \
276 lockdep_assert_irqs_disabled(); \
277 raw_spin_lock(¤t->pi_lock); \
278 current->saved_state = current->__state; \
279 debug_rtlock_wait_set_state(); \
280 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
281 raw_spin_unlock(¤t->pi_lock); \
282 } while (0);
283
284 #define current_restore_rtlock_saved_state() \
285 do { \
286 lockdep_assert_irqs_disabled(); \
287 raw_spin_lock(¤t->pi_lock); \
288 debug_rtlock_wait_restore_state(); \
289 WRITE_ONCE(current->__state, current->saved_state); \
290 current->saved_state = TASK_RUNNING; \
291 raw_spin_unlock(¤t->pi_lock); \
292 } while (0);
293
294 #define get_current_state() READ_ONCE(current->__state)
295
296 /*
297 * Define the task command name length as enum, then it can be visible to
298 * BPF programs.
299 */
300 enum {
301 TASK_COMM_LEN = 16,
302 };
303
304 extern void sched_tick(void);
305
306 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
307
308 extern long schedule_timeout(long timeout);
309 extern long schedule_timeout_interruptible(long timeout);
310 extern long schedule_timeout_killable(long timeout);
311 extern long schedule_timeout_uninterruptible(long timeout);
312 extern long schedule_timeout_idle(long timeout);
313 asmlinkage void schedule(void);
314 extern void schedule_preempt_disabled(void);
315 asmlinkage void preempt_schedule_irq(void);
316 #ifdef CONFIG_PREEMPT_RT
317 extern void schedule_rtlock(void);
318 #endif
319
320 extern int __must_check io_schedule_prepare(void);
321 extern void io_schedule_finish(int token);
322 extern long io_schedule_timeout(long timeout);
323 extern void io_schedule(void);
324
325 /**
326 * struct prev_cputime - snapshot of system and user cputime
327 * @utime: time spent in user mode
328 * @stime: time spent in system mode
329 * @lock: protects the above two fields
330 *
331 * Stores previous user/system time values such that we can guarantee
332 * monotonicity.
333 */
334 struct prev_cputime {
335 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
336 u64 utime;
337 u64 stime;
338 raw_spinlock_t lock;
339 #endif
340 };
341
342 enum vtime_state {
343 /* Task is sleeping or running in a CPU with VTIME inactive: */
344 VTIME_INACTIVE = 0,
345 /* Task is idle */
346 VTIME_IDLE,
347 /* Task runs in kernelspace in a CPU with VTIME active: */
348 VTIME_SYS,
349 /* Task runs in userspace in a CPU with VTIME active: */
350 VTIME_USER,
351 /* Task runs as guests in a CPU with VTIME active: */
352 VTIME_GUEST,
353 };
354
355 struct vtime {
356 seqcount_t seqcount;
357 unsigned long long starttime;
358 enum vtime_state state;
359 unsigned int cpu;
360 u64 utime;
361 u64 stime;
362 u64 gtime;
363 };
364
365 /*
366 * Utilization clamp constraints.
367 * @UCLAMP_MIN: Minimum utilization
368 * @UCLAMP_MAX: Maximum utilization
369 * @UCLAMP_CNT: Utilization clamp constraints count
370 */
371 enum uclamp_id {
372 UCLAMP_MIN = 0,
373 UCLAMP_MAX,
374 UCLAMP_CNT
375 };
376
377 #ifdef CONFIG_SMP
378 extern struct root_domain def_root_domain;
379 extern struct mutex sched_domains_mutex;
380 #endif
381
382 struct sched_param {
383 int sched_priority;
384 };
385
386 struct sched_info {
387 #ifdef CONFIG_SCHED_INFO
388 /* Cumulative counters: */
389
390 /* # of times we have run on this CPU: */
391 unsigned long pcount;
392
393 /* Time spent waiting on a runqueue: */
394 unsigned long long run_delay;
395
396 /* Timestamps: */
397
398 /* When did we last run on a CPU? */
399 unsigned long long last_arrival;
400
401 /* When were we last queued to run? */
402 unsigned long long last_queued;
403
404 #endif /* CONFIG_SCHED_INFO */
405 };
406
407 /*
408 * Integer metrics need fixed point arithmetic, e.g., sched/fair
409 * has a few: load, load_avg, util_avg, freq, and capacity.
410 *
411 * We define a basic fixed point arithmetic range, and then formalize
412 * all these metrics based on that basic range.
413 */
414 # define SCHED_FIXEDPOINT_SHIFT 10
415 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
416
417 /* Increase resolution of cpu_capacity calculations */
418 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
419 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
420
421 struct load_weight {
422 unsigned long weight;
423 u32 inv_weight;
424 };
425
426 /*
427 * The load/runnable/util_avg accumulates an infinite geometric series
428 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
429 *
430 * [load_avg definition]
431 *
432 * load_avg = runnable% * scale_load_down(load)
433 *
434 * [runnable_avg definition]
435 *
436 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
437 *
438 * [util_avg definition]
439 *
440 * util_avg = running% * SCHED_CAPACITY_SCALE
441 *
442 * where runnable% is the time ratio that a sched_entity is runnable and
443 * running% the time ratio that a sched_entity is running.
444 *
445 * For cfs_rq, they are the aggregated values of all runnable and blocked
446 * sched_entities.
447 *
448 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
449 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
450 * for computing those signals (see update_rq_clock_pelt())
451 *
452 * N.B., the above ratios (runnable% and running%) themselves are in the
453 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
454 * to as large a range as necessary. This is for example reflected by
455 * util_avg's SCHED_CAPACITY_SCALE.
456 *
457 * [Overflow issue]
458 *
459 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
460 * with the highest load (=88761), always runnable on a single cfs_rq,
461 * and should not overflow as the number already hits PID_MAX_LIMIT.
462 *
463 * For all other cases (including 32-bit kernels), struct load_weight's
464 * weight will overflow first before we do, because:
465 *
466 * Max(load_avg) <= Max(load.weight)
467 *
468 * Then it is the load_weight's responsibility to consider overflow
469 * issues.
470 */
471 struct sched_avg {
472 u64 last_update_time;
473 u64 load_sum;
474 u64 runnable_sum;
475 u32 util_sum;
476 u32 period_contrib;
477 unsigned long load_avg;
478 unsigned long runnable_avg;
479 unsigned long util_avg;
480 unsigned int util_est;
481 } ____cacheline_aligned;
482
483 /*
484 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
485 * updates. When a task is dequeued, its util_est should not be updated if its
486 * util_avg has not been updated in the meantime.
487 * This information is mapped into the MSB bit of util_est at dequeue time.
488 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
489 * it is safe to use MSB.
490 */
491 #define UTIL_EST_WEIGHT_SHIFT 2
492 #define UTIL_AVG_UNCHANGED 0x80000000
493
494 struct sched_statistics {
495 #ifdef CONFIG_SCHEDSTATS
496 u64 wait_start;
497 u64 wait_max;
498 u64 wait_count;
499 u64 wait_sum;
500 u64 iowait_count;
501 u64 iowait_sum;
502
503 u64 sleep_start;
504 u64 sleep_max;
505 s64 sum_sleep_runtime;
506
507 u64 block_start;
508 u64 block_max;
509 s64 sum_block_runtime;
510
511 s64 exec_max;
512 u64 slice_max;
513
514 u64 nr_migrations_cold;
515 u64 nr_failed_migrations_affine;
516 u64 nr_failed_migrations_running;
517 u64 nr_failed_migrations_hot;
518 u64 nr_forced_migrations;
519
520 u64 nr_wakeups;
521 u64 nr_wakeups_sync;
522 u64 nr_wakeups_migrate;
523 u64 nr_wakeups_local;
524 u64 nr_wakeups_remote;
525 u64 nr_wakeups_affine;
526 u64 nr_wakeups_affine_attempts;
527 u64 nr_wakeups_passive;
528 u64 nr_wakeups_idle;
529
530 #ifdef CONFIG_SCHED_CORE
531 u64 core_forceidle_sum;
532 #endif
533 #endif /* CONFIG_SCHEDSTATS */
534 } ____cacheline_aligned;
535
536 struct sched_entity {
537 /* For load-balancing: */
538 struct load_weight load;
539 struct rb_node run_node;
540 u64 deadline;
541 u64 min_vruntime;
542
543 struct list_head group_node;
544 unsigned int on_rq;
545
546 u64 exec_start;
547 u64 sum_exec_runtime;
548 u64 prev_sum_exec_runtime;
549 u64 vruntime;
550 s64 vlag;
551 u64 slice;
552
553 u64 nr_migrations;
554
555 #ifdef CONFIG_FAIR_GROUP_SCHED
556 int depth;
557 struct sched_entity *parent;
558 /* rq on which this entity is (to be) queued: */
559 struct cfs_rq *cfs_rq;
560 /* rq "owned" by this entity/group: */
561 struct cfs_rq *my_q;
562 /* cached value of my_q->h_nr_running */
563 unsigned long runnable_weight;
564 #endif
565
566 #ifdef CONFIG_SMP
567 /*
568 * Per entity load average tracking.
569 *
570 * Put into separate cache line so it does not
571 * collide with read-mostly values above.
572 */
573 struct sched_avg avg;
574 #endif
575 };
576
577 struct sched_rt_entity {
578 struct list_head run_list;
579 unsigned long timeout;
580 unsigned long watchdog_stamp;
581 unsigned int time_slice;
582 unsigned short on_rq;
583 unsigned short on_list;
584
585 struct sched_rt_entity *back;
586 #ifdef CONFIG_RT_GROUP_SCHED
587 struct sched_rt_entity *parent;
588 /* rq on which this entity is (to be) queued: */
589 struct rt_rq *rt_rq;
590 /* rq "owned" by this entity/group: */
591 struct rt_rq *my_q;
592 #endif
593 } __randomize_layout;
594
595 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
596 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
597
598 struct sched_dl_entity {
599 struct rb_node rb_node;
600
601 /*
602 * Original scheduling parameters. Copied here from sched_attr
603 * during sched_setattr(), they will remain the same until
604 * the next sched_setattr().
605 */
606 u64 dl_runtime; /* Maximum runtime for each instance */
607 u64 dl_deadline; /* Relative deadline of each instance */
608 u64 dl_period; /* Separation of two instances (period) */
609 u64 dl_bw; /* dl_runtime / dl_period */
610 u64 dl_density; /* dl_runtime / dl_deadline */
611
612 /*
613 * Actual scheduling parameters. Initialized with the values above,
614 * they are continuously updated during task execution. Note that
615 * the remaining runtime could be < 0 in case we are in overrun.
616 */
617 s64 runtime; /* Remaining runtime for this instance */
618 u64 deadline; /* Absolute deadline for this instance */
619 unsigned int flags; /* Specifying the scheduler behaviour */
620
621 /*
622 * Some bool flags:
623 *
624 * @dl_throttled tells if we exhausted the runtime. If so, the
625 * task has to wait for a replenishment to be performed at the
626 * next firing of dl_timer.
627 *
628 * @dl_yielded tells if task gave up the CPU before consuming
629 * all its available runtime during the last job.
630 *
631 * @dl_non_contending tells if the task is inactive while still
632 * contributing to the active utilization. In other words, it
633 * indicates if the inactive timer has been armed and its handler
634 * has not been executed yet. This flag is useful to avoid race
635 * conditions between the inactive timer handler and the wakeup
636 * code.
637 *
638 * @dl_overrun tells if the task asked to be informed about runtime
639 * overruns.
640 */
641 unsigned int dl_throttled : 1;
642 unsigned int dl_yielded : 1;
643 unsigned int dl_non_contending : 1;
644 unsigned int dl_overrun : 1;
645 unsigned int dl_server : 1;
646
647 /*
648 * Bandwidth enforcement timer. Each -deadline task has its
649 * own bandwidth to be enforced, thus we need one timer per task.
650 */
651 struct hrtimer dl_timer;
652
653 /*
654 * Inactive timer, responsible for decreasing the active utilization
655 * at the "0-lag time". When a -deadline task blocks, it contributes
656 * to GRUB's active utilization until the "0-lag time", hence a
657 * timer is needed to decrease the active utilization at the correct
658 * time.
659 */
660 struct hrtimer inactive_timer;
661
662 /*
663 * Bits for DL-server functionality. Also see the comment near
664 * dl_server_update().
665 *
666 * @rq the runqueue this server is for
667 *
668 * @server_has_tasks() returns true if @server_pick return a
669 * runnable task.
670 */
671 struct rq *rq;
672 dl_server_has_tasks_f server_has_tasks;
673 dl_server_pick_f server_pick;
674
675 #ifdef CONFIG_RT_MUTEXES
676 /*
677 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
678 * pi_se points to the donor, otherwise points to the dl_se it belongs
679 * to (the original one/itself).
680 */
681 struct sched_dl_entity *pi_se;
682 #endif
683 };
684
685 #ifdef CONFIG_UCLAMP_TASK
686 /* Number of utilization clamp buckets (shorter alias) */
687 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
688
689 /*
690 * Utilization clamp for a scheduling entity
691 * @value: clamp value "assigned" to a se
692 * @bucket_id: bucket index corresponding to the "assigned" value
693 * @active: the se is currently refcounted in a rq's bucket
694 * @user_defined: the requested clamp value comes from user-space
695 *
696 * The bucket_id is the index of the clamp bucket matching the clamp value
697 * which is pre-computed and stored to avoid expensive integer divisions from
698 * the fast path.
699 *
700 * The active bit is set whenever a task has got an "effective" value assigned,
701 * which can be different from the clamp value "requested" from user-space.
702 * This allows to know a task is refcounted in the rq's bucket corresponding
703 * to the "effective" bucket_id.
704 *
705 * The user_defined bit is set whenever a task has got a task-specific clamp
706 * value requested from userspace, i.e. the system defaults apply to this task
707 * just as a restriction. This allows to relax default clamps when a less
708 * restrictive task-specific value has been requested, thus allowing to
709 * implement a "nice" semantic. For example, a task running with a 20%
710 * default boost can still drop its own boosting to 0%.
711 */
712 struct uclamp_se {
713 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
714 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
715 unsigned int active : 1;
716 unsigned int user_defined : 1;
717 };
718 #endif /* CONFIG_UCLAMP_TASK */
719
720 union rcu_special {
721 struct {
722 u8 blocked;
723 u8 need_qs;
724 u8 exp_hint; /* Hint for performance. */
725 u8 need_mb; /* Readers need smp_mb(). */
726 } b; /* Bits. */
727 u32 s; /* Set of bits. */
728 };
729
730 enum perf_event_task_context {
731 perf_invalid_context = -1,
732 perf_hw_context = 0,
733 perf_sw_context,
734 perf_nr_task_contexts,
735 };
736
737 struct wake_q_node {
738 struct wake_q_node *next;
739 };
740
741 struct kmap_ctrl {
742 #ifdef CONFIG_KMAP_LOCAL
743 int idx;
744 pte_t pteval[KM_MAX_IDX];
745 #endif
746 };
747
748 struct task_struct {
749 #ifdef CONFIG_THREAD_INFO_IN_TASK
750 /*
751 * For reasons of header soup (see current_thread_info()), this
752 * must be the first element of task_struct.
753 */
754 struct thread_info thread_info;
755 #endif
756 unsigned int __state;
757
758 /* saved state for "spinlock sleepers" */
759 unsigned int saved_state;
760
761 /*
762 * This begins the randomizable portion of task_struct. Only
763 * scheduling-critical items should be added above here.
764 */
765 randomized_struct_fields_start
766
767 void *stack;
768 refcount_t usage;
769 /* Per task flags (PF_*), defined further below: */
770 unsigned int flags;
771 unsigned int ptrace;
772
773 #ifdef CONFIG_MEM_ALLOC_PROFILING
774 struct alloc_tag *alloc_tag;
775 #endif
776
777 #ifdef CONFIG_SMP
778 int on_cpu;
779 struct __call_single_node wake_entry;
780 unsigned int wakee_flips;
781 unsigned long wakee_flip_decay_ts;
782 struct task_struct *last_wakee;
783
784 /*
785 * recent_used_cpu is initially set as the last CPU used by a task
786 * that wakes affine another task. Waker/wakee relationships can
787 * push tasks around a CPU where each wakeup moves to the next one.
788 * Tracking a recently used CPU allows a quick search for a recently
789 * used CPU that may be idle.
790 */
791 int recent_used_cpu;
792 int wake_cpu;
793 #endif
794 int on_rq;
795
796 int prio;
797 int static_prio;
798 int normal_prio;
799 unsigned int rt_priority;
800
801 struct sched_entity se;
802 struct sched_rt_entity rt;
803 struct sched_dl_entity dl;
804 struct sched_dl_entity *dl_server;
805 const struct sched_class *sched_class;
806
807 #ifdef CONFIG_SCHED_CORE
808 struct rb_node core_node;
809 unsigned long core_cookie;
810 unsigned int core_occupation;
811 #endif
812
813 #ifdef CONFIG_CGROUP_SCHED
814 struct task_group *sched_task_group;
815 #endif
816
817
818 #ifdef CONFIG_UCLAMP_TASK
819 /*
820 * Clamp values requested for a scheduling entity.
821 * Must be updated with task_rq_lock() held.
822 */
823 struct uclamp_se uclamp_req[UCLAMP_CNT];
824 /*
825 * Effective clamp values used for a scheduling entity.
826 * Must be updated with task_rq_lock() held.
827 */
828 struct uclamp_se uclamp[UCLAMP_CNT];
829 #endif
830
831 struct sched_statistics stats;
832
833 #ifdef CONFIG_PREEMPT_NOTIFIERS
834 /* List of struct preempt_notifier: */
835 struct hlist_head preempt_notifiers;
836 #endif
837
838 #ifdef CONFIG_BLK_DEV_IO_TRACE
839 unsigned int btrace_seq;
840 #endif
841
842 unsigned int policy;
843 unsigned long max_allowed_capacity;
844 int nr_cpus_allowed;
845 const cpumask_t *cpus_ptr;
846 cpumask_t *user_cpus_ptr;
847 cpumask_t cpus_mask;
848 void *migration_pending;
849 #ifdef CONFIG_SMP
850 unsigned short migration_disabled;
851 #endif
852 unsigned short migration_flags;
853
854 #ifdef CONFIG_PREEMPT_RCU
855 int rcu_read_lock_nesting;
856 union rcu_special rcu_read_unlock_special;
857 struct list_head rcu_node_entry;
858 struct rcu_node *rcu_blocked_node;
859 #endif /* #ifdef CONFIG_PREEMPT_RCU */
860
861 #ifdef CONFIG_TASKS_RCU
862 unsigned long rcu_tasks_nvcsw;
863 u8 rcu_tasks_holdout;
864 u8 rcu_tasks_idx;
865 int rcu_tasks_idle_cpu;
866 struct list_head rcu_tasks_holdout_list;
867 int rcu_tasks_exit_cpu;
868 struct list_head rcu_tasks_exit_list;
869 #endif /* #ifdef CONFIG_TASKS_RCU */
870
871 #ifdef CONFIG_TASKS_TRACE_RCU
872 int trc_reader_nesting;
873 int trc_ipi_to_cpu;
874 union rcu_special trc_reader_special;
875 struct list_head trc_holdout_list;
876 struct list_head trc_blkd_node;
877 int trc_blkd_cpu;
878 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
879
880 struct sched_info sched_info;
881
882 struct list_head tasks;
883 #ifdef CONFIG_SMP
884 struct plist_node pushable_tasks;
885 struct rb_node pushable_dl_tasks;
886 #endif
887
888 struct mm_struct *mm;
889 struct mm_struct *active_mm;
890 struct address_space *faults_disabled_mapping;
891
892 int exit_state;
893 int exit_code;
894 int exit_signal;
895 /* The signal sent when the parent dies: */
896 int pdeath_signal;
897 /* JOBCTL_*, siglock protected: */
898 unsigned long jobctl;
899
900 /* Used for emulating ABI behavior of previous Linux versions: */
901 unsigned int personality;
902
903 /* Scheduler bits, serialized by scheduler locks: */
904 unsigned sched_reset_on_fork:1;
905 unsigned sched_contributes_to_load:1;
906 unsigned sched_migrated:1;
907
908 /* Force alignment to the next boundary: */
909 unsigned :0;
910
911 /* Unserialized, strictly 'current' */
912
913 /*
914 * This field must not be in the scheduler word above due to wakelist
915 * queueing no longer being serialized by p->on_cpu. However:
916 *
917 * p->XXX = X; ttwu()
918 * schedule() if (p->on_rq && ..) // false
919 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
920 * deactivate_task() ttwu_queue_wakelist())
921 * p->on_rq = 0; p->sched_remote_wakeup = Y;
922 *
923 * guarantees all stores of 'current' are visible before
924 * ->sched_remote_wakeup gets used, so it can be in this word.
925 */
926 unsigned sched_remote_wakeup:1;
927 #ifdef CONFIG_RT_MUTEXES
928 unsigned sched_rt_mutex:1;
929 #endif
930
931 /* Bit to tell TOMOYO we're in execve(): */
932 unsigned in_execve:1;
933 unsigned in_iowait:1;
934 #ifndef TIF_RESTORE_SIGMASK
935 unsigned restore_sigmask:1;
936 #endif
937 #ifdef CONFIG_MEMCG
938 unsigned in_user_fault:1;
939 #endif
940 #ifdef CONFIG_LRU_GEN
941 /* whether the LRU algorithm may apply to this access */
942 unsigned in_lru_fault:1;
943 #endif
944 #ifdef CONFIG_COMPAT_BRK
945 unsigned brk_randomized:1;
946 #endif
947 #ifdef CONFIG_CGROUPS
948 /* disallow userland-initiated cgroup migration */
949 unsigned no_cgroup_migration:1;
950 /* task is frozen/stopped (used by the cgroup freezer) */
951 unsigned frozen:1;
952 #endif
953 #ifdef CONFIG_BLK_CGROUP
954 unsigned use_memdelay:1;
955 #endif
956 #ifdef CONFIG_PSI
957 /* Stalled due to lack of memory */
958 unsigned in_memstall:1;
959 #endif
960 #ifdef CONFIG_PAGE_OWNER
961 /* Used by page_owner=on to detect recursion in page tracking. */
962 unsigned in_page_owner:1;
963 #endif
964 #ifdef CONFIG_EVENTFD
965 /* Recursion prevention for eventfd_signal() */
966 unsigned in_eventfd:1;
967 #endif
968 #ifdef CONFIG_ARCH_HAS_CPU_PASID
969 unsigned pasid_activated:1;
970 #endif
971 #ifdef CONFIG_CPU_SUP_INTEL
972 unsigned reported_split_lock:1;
973 #endif
974 #ifdef CONFIG_TASK_DELAY_ACCT
975 /* delay due to memory thrashing */
976 unsigned in_thrashing:1;
977 #endif
978
979 unsigned long atomic_flags; /* Flags requiring atomic access. */
980
981 struct restart_block restart_block;
982
983 pid_t pid;
984 pid_t tgid;
985
986 #ifdef CONFIG_STACKPROTECTOR
987 /* Canary value for the -fstack-protector GCC feature: */
988 unsigned long stack_canary;
989 #endif
990 /*
991 * Pointers to the (original) parent process, youngest child, younger sibling,
992 * older sibling, respectively. (p->father can be replaced with
993 * p->real_parent->pid)
994 */
995
996 /* Real parent process: */
997 struct task_struct __rcu *real_parent;
998
999 /* Recipient of SIGCHLD, wait4() reports: */
1000 struct task_struct __rcu *parent;
1001
1002 /*
1003 * Children/sibling form the list of natural children:
1004 */
1005 struct list_head children;
1006 struct list_head sibling;
1007 struct task_struct *group_leader;
1008
1009 /*
1010 * 'ptraced' is the list of tasks this task is using ptrace() on.
1011 *
1012 * This includes both natural children and PTRACE_ATTACH targets.
1013 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1014 */
1015 struct list_head ptraced;
1016 struct list_head ptrace_entry;
1017
1018 /* PID/PID hash table linkage. */
1019 struct pid *thread_pid;
1020 struct hlist_node pid_links[PIDTYPE_MAX];
1021 struct list_head thread_node;
1022
1023 struct completion *vfork_done;
1024
1025 /* CLONE_CHILD_SETTID: */
1026 int __user *set_child_tid;
1027
1028 /* CLONE_CHILD_CLEARTID: */
1029 int __user *clear_child_tid;
1030
1031 /* PF_KTHREAD | PF_IO_WORKER */
1032 void *worker_private;
1033
1034 u64 utime;
1035 u64 stime;
1036 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1037 u64 utimescaled;
1038 u64 stimescaled;
1039 #endif
1040 u64 gtime;
1041 struct prev_cputime prev_cputime;
1042 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1043 struct vtime vtime;
1044 #endif
1045
1046 #ifdef CONFIG_NO_HZ_FULL
1047 atomic_t tick_dep_mask;
1048 #endif
1049 /* Context switch counts: */
1050 unsigned long nvcsw;
1051 unsigned long nivcsw;
1052
1053 /* Monotonic time in nsecs: */
1054 u64 start_time;
1055
1056 /* Boot based time in nsecs: */
1057 u64 start_boottime;
1058
1059 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1060 unsigned long min_flt;
1061 unsigned long maj_flt;
1062
1063 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1064 struct posix_cputimers posix_cputimers;
1065
1066 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1067 struct posix_cputimers_work posix_cputimers_work;
1068 #endif
1069
1070 /* Process credentials: */
1071
1072 /* Tracer's credentials at attach: */
1073 const struct cred __rcu *ptracer_cred;
1074
1075 /* Objective and real subjective task credentials (COW): */
1076 const struct cred __rcu *real_cred;
1077
1078 /* Effective (overridable) subjective task credentials (COW): */
1079 const struct cred __rcu *cred;
1080
1081 #ifdef CONFIG_KEYS
1082 /* Cached requested key. */
1083 struct key *cached_requested_key;
1084 #endif
1085
1086 /*
1087 * executable name, excluding path.
1088 *
1089 * - normally initialized setup_new_exec()
1090 * - access it with [gs]et_task_comm()
1091 * - lock it with task_lock()
1092 */
1093 char comm[TASK_COMM_LEN];
1094
1095 struct nameidata *nameidata;
1096
1097 #ifdef CONFIG_SYSVIPC
1098 struct sysv_sem sysvsem;
1099 struct sysv_shm sysvshm;
1100 #endif
1101 #ifdef CONFIG_DETECT_HUNG_TASK
1102 unsigned long last_switch_count;
1103 unsigned long last_switch_time;
1104 #endif
1105 /* Filesystem information: */
1106 struct fs_struct *fs;
1107
1108 /* Open file information: */
1109 struct files_struct *files;
1110
1111 #ifdef CONFIG_IO_URING
1112 struct io_uring_task *io_uring;
1113 #endif
1114
1115 /* Namespaces: */
1116 struct nsproxy *nsproxy;
1117
1118 /* Signal handlers: */
1119 struct signal_struct *signal;
1120 struct sighand_struct __rcu *sighand;
1121 sigset_t blocked;
1122 sigset_t real_blocked;
1123 /* Restored if set_restore_sigmask() was used: */
1124 sigset_t saved_sigmask;
1125 struct sigpending pending;
1126 unsigned long sas_ss_sp;
1127 size_t sas_ss_size;
1128 unsigned int sas_ss_flags;
1129
1130 struct callback_head *task_works;
1131
1132 #ifdef CONFIG_AUDIT
1133 #ifdef CONFIG_AUDITSYSCALL
1134 struct audit_context *audit_context;
1135 #endif
1136 kuid_t loginuid;
1137 unsigned int sessionid;
1138 #endif
1139 struct seccomp seccomp;
1140 struct syscall_user_dispatch syscall_dispatch;
1141
1142 /* Thread group tracking: */
1143 u64 parent_exec_id;
1144 u64 self_exec_id;
1145
1146 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1147 spinlock_t alloc_lock;
1148
1149 /* Protection of the PI data structures: */
1150 raw_spinlock_t pi_lock;
1151
1152 struct wake_q_node wake_q;
1153
1154 #ifdef CONFIG_RT_MUTEXES
1155 /* PI waiters blocked on a rt_mutex held by this task: */
1156 struct rb_root_cached pi_waiters;
1157 /* Updated under owner's pi_lock and rq lock */
1158 struct task_struct *pi_top_task;
1159 /* Deadlock detection and priority inheritance handling: */
1160 struct rt_mutex_waiter *pi_blocked_on;
1161 #endif
1162
1163 #ifdef CONFIG_DEBUG_MUTEXES
1164 /* Mutex deadlock detection: */
1165 struct mutex_waiter *blocked_on;
1166 #endif
1167
1168 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1169 int non_block_count;
1170 #endif
1171
1172 #ifdef CONFIG_TRACE_IRQFLAGS
1173 struct irqtrace_events irqtrace;
1174 unsigned int hardirq_threaded;
1175 u64 hardirq_chain_key;
1176 int softirqs_enabled;
1177 int softirq_context;
1178 int irq_config;
1179 #endif
1180 #ifdef CONFIG_PREEMPT_RT
1181 int softirq_disable_cnt;
1182 #endif
1183
1184 #ifdef CONFIG_LOCKDEP
1185 # define MAX_LOCK_DEPTH 48UL
1186 u64 curr_chain_key;
1187 int lockdep_depth;
1188 unsigned int lockdep_recursion;
1189 struct held_lock held_locks[MAX_LOCK_DEPTH];
1190 #endif
1191
1192 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1193 unsigned int in_ubsan;
1194 #endif
1195
1196 /* Journalling filesystem info: */
1197 void *journal_info;
1198
1199 /* Stacked block device info: */
1200 struct bio_list *bio_list;
1201
1202 /* Stack plugging: */
1203 struct blk_plug *plug;
1204
1205 /* VM state: */
1206 struct reclaim_state *reclaim_state;
1207
1208 struct io_context *io_context;
1209
1210 #ifdef CONFIG_COMPACTION
1211 struct capture_control *capture_control;
1212 #endif
1213 /* Ptrace state: */
1214 unsigned long ptrace_message;
1215 kernel_siginfo_t *last_siginfo;
1216
1217 struct task_io_accounting ioac;
1218 #ifdef CONFIG_PSI
1219 /* Pressure stall state */
1220 unsigned int psi_flags;
1221 #endif
1222 #ifdef CONFIG_TASK_XACCT
1223 /* Accumulated RSS usage: */
1224 u64 acct_rss_mem1;
1225 /* Accumulated virtual memory usage: */
1226 u64 acct_vm_mem1;
1227 /* stime + utime since last update: */
1228 u64 acct_timexpd;
1229 #endif
1230 #ifdef CONFIG_CPUSETS
1231 /* Protected by ->alloc_lock: */
1232 nodemask_t mems_allowed;
1233 /* Sequence number to catch updates: */
1234 seqcount_spinlock_t mems_allowed_seq;
1235 int cpuset_mem_spread_rotor;
1236 int cpuset_slab_spread_rotor;
1237 #endif
1238 #ifdef CONFIG_CGROUPS
1239 /* Control Group info protected by css_set_lock: */
1240 struct css_set __rcu *cgroups;
1241 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1242 struct list_head cg_list;
1243 #endif
1244 #ifdef CONFIG_X86_CPU_RESCTRL
1245 u32 closid;
1246 u32 rmid;
1247 #endif
1248 #ifdef CONFIG_FUTEX
1249 struct robust_list_head __user *robust_list;
1250 #ifdef CONFIG_COMPAT
1251 struct compat_robust_list_head __user *compat_robust_list;
1252 #endif
1253 struct list_head pi_state_list;
1254 struct futex_pi_state *pi_state_cache;
1255 struct mutex futex_exit_mutex;
1256 unsigned int futex_state;
1257 #endif
1258 #ifdef CONFIG_PERF_EVENTS
1259 struct perf_event_context *perf_event_ctxp;
1260 struct mutex perf_event_mutex;
1261 struct list_head perf_event_list;
1262 #endif
1263 #ifdef CONFIG_DEBUG_PREEMPT
1264 unsigned long preempt_disable_ip;
1265 #endif
1266 #ifdef CONFIG_NUMA
1267 /* Protected by alloc_lock: */
1268 struct mempolicy *mempolicy;
1269 short il_prev;
1270 u8 il_weight;
1271 short pref_node_fork;
1272 #endif
1273 #ifdef CONFIG_NUMA_BALANCING
1274 int numa_scan_seq;
1275 unsigned int numa_scan_period;
1276 unsigned int numa_scan_period_max;
1277 int numa_preferred_nid;
1278 unsigned long numa_migrate_retry;
1279 /* Migration stamp: */
1280 u64 node_stamp;
1281 u64 last_task_numa_placement;
1282 u64 last_sum_exec_runtime;
1283 struct callback_head numa_work;
1284
1285 /*
1286 * This pointer is only modified for current in syscall and
1287 * pagefault context (and for tasks being destroyed), so it can be read
1288 * from any of the following contexts:
1289 * - RCU read-side critical section
1290 * - current->numa_group from everywhere
1291 * - task's runqueue locked, task not running
1292 */
1293 struct numa_group __rcu *numa_group;
1294
1295 /*
1296 * numa_faults is an array split into four regions:
1297 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1298 * in this precise order.
1299 *
1300 * faults_memory: Exponential decaying average of faults on a per-node
1301 * basis. Scheduling placement decisions are made based on these
1302 * counts. The values remain static for the duration of a PTE scan.
1303 * faults_cpu: Track the nodes the process was running on when a NUMA
1304 * hinting fault was incurred.
1305 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1306 * during the current scan window. When the scan completes, the counts
1307 * in faults_memory and faults_cpu decay and these values are copied.
1308 */
1309 unsigned long *numa_faults;
1310 unsigned long total_numa_faults;
1311
1312 /*
1313 * numa_faults_locality tracks if faults recorded during the last
1314 * scan window were remote/local or failed to migrate. The task scan
1315 * period is adapted based on the locality of the faults with different
1316 * weights depending on whether they were shared or private faults
1317 */
1318 unsigned long numa_faults_locality[3];
1319
1320 unsigned long numa_pages_migrated;
1321 #endif /* CONFIG_NUMA_BALANCING */
1322
1323 #ifdef CONFIG_RSEQ
1324 struct rseq __user *rseq;
1325 u32 rseq_len;
1326 u32 rseq_sig;
1327 /*
1328 * RmW on rseq_event_mask must be performed atomically
1329 * with respect to preemption.
1330 */
1331 unsigned long rseq_event_mask;
1332 #endif
1333
1334 #ifdef CONFIG_SCHED_MM_CID
1335 int mm_cid; /* Current cid in mm */
1336 int last_mm_cid; /* Most recent cid in mm */
1337 int migrate_from_cpu;
1338 int mm_cid_active; /* Whether cid bitmap is active */
1339 struct callback_head cid_work;
1340 #endif
1341
1342 struct tlbflush_unmap_batch tlb_ubc;
1343
1344 /* Cache last used pipe for splice(): */
1345 struct pipe_inode_info *splice_pipe;
1346
1347 struct page_frag task_frag;
1348
1349 #ifdef CONFIG_TASK_DELAY_ACCT
1350 struct task_delay_info *delays;
1351 #endif
1352
1353 #ifdef CONFIG_FAULT_INJECTION
1354 int make_it_fail;
1355 unsigned int fail_nth;
1356 #endif
1357 /*
1358 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1359 * balance_dirty_pages() for a dirty throttling pause:
1360 */
1361 int nr_dirtied;
1362 int nr_dirtied_pause;
1363 /* Start of a write-and-pause period: */
1364 unsigned long dirty_paused_when;
1365
1366 #ifdef CONFIG_LATENCYTOP
1367 int latency_record_count;
1368 struct latency_record latency_record[LT_SAVECOUNT];
1369 #endif
1370 /*
1371 * Time slack values; these are used to round up poll() and
1372 * select() etc timeout values. These are in nanoseconds.
1373 */
1374 u64 timer_slack_ns;
1375 u64 default_timer_slack_ns;
1376
1377 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1378 unsigned int kasan_depth;
1379 #endif
1380
1381 #ifdef CONFIG_KCSAN
1382 struct kcsan_ctx kcsan_ctx;
1383 #ifdef CONFIG_TRACE_IRQFLAGS
1384 struct irqtrace_events kcsan_save_irqtrace;
1385 #endif
1386 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1387 int kcsan_stack_depth;
1388 #endif
1389 #endif
1390
1391 #ifdef CONFIG_KMSAN
1392 struct kmsan_ctx kmsan_ctx;
1393 #endif
1394
1395 #if IS_ENABLED(CONFIG_KUNIT)
1396 struct kunit *kunit_test;
1397 #endif
1398
1399 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1400 /* Index of current stored address in ret_stack: */
1401 int curr_ret_stack;
1402 int curr_ret_depth;
1403
1404 /* Stack of return addresses for return function tracing: */
1405 struct ftrace_ret_stack *ret_stack;
1406
1407 /* Timestamp for last schedule: */
1408 unsigned long long ftrace_timestamp;
1409
1410 /*
1411 * Number of functions that haven't been traced
1412 * because of depth overrun:
1413 */
1414 atomic_t trace_overrun;
1415
1416 /* Pause tracing: */
1417 atomic_t tracing_graph_pause;
1418 #endif
1419
1420 #ifdef CONFIG_TRACING
1421 /* Bitmask and counter of trace recursion: */
1422 unsigned long trace_recursion;
1423 #endif /* CONFIG_TRACING */
1424
1425 #ifdef CONFIG_KCOV
1426 /* See kernel/kcov.c for more details. */
1427
1428 /* Coverage collection mode enabled for this task (0 if disabled): */
1429 unsigned int kcov_mode;
1430
1431 /* Size of the kcov_area: */
1432 unsigned int kcov_size;
1433
1434 /* Buffer for coverage collection: */
1435 void *kcov_area;
1436
1437 /* KCOV descriptor wired with this task or NULL: */
1438 struct kcov *kcov;
1439
1440 /* KCOV common handle for remote coverage collection: */
1441 u64 kcov_handle;
1442
1443 /* KCOV sequence number: */
1444 int kcov_sequence;
1445
1446 /* Collect coverage from softirq context: */
1447 unsigned int kcov_softirq;
1448 #endif
1449
1450 #ifdef CONFIG_MEMCG
1451 struct mem_cgroup *memcg_in_oom;
1452
1453 /* Number of pages to reclaim on returning to userland: */
1454 unsigned int memcg_nr_pages_over_high;
1455
1456 /* Used by memcontrol for targeted memcg charge: */
1457 struct mem_cgroup *active_memcg;
1458 #endif
1459
1460 #ifdef CONFIG_MEMCG_KMEM
1461 struct obj_cgroup *objcg;
1462 #endif
1463
1464 #ifdef CONFIG_BLK_CGROUP
1465 struct gendisk *throttle_disk;
1466 #endif
1467
1468 #ifdef CONFIG_UPROBES
1469 struct uprobe_task *utask;
1470 #endif
1471 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1472 unsigned int sequential_io;
1473 unsigned int sequential_io_avg;
1474 #endif
1475 struct kmap_ctrl kmap_ctrl;
1476 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1477 unsigned long task_state_change;
1478 # ifdef CONFIG_PREEMPT_RT
1479 unsigned long saved_state_change;
1480 # endif
1481 #endif
1482 struct rcu_head rcu;
1483 refcount_t rcu_users;
1484 int pagefault_disabled;
1485 #ifdef CONFIG_MMU
1486 struct task_struct *oom_reaper_list;
1487 struct timer_list oom_reaper_timer;
1488 #endif
1489 #ifdef CONFIG_VMAP_STACK
1490 struct vm_struct *stack_vm_area;
1491 #endif
1492 #ifdef CONFIG_THREAD_INFO_IN_TASK
1493 /* A live task holds one reference: */
1494 refcount_t stack_refcount;
1495 #endif
1496 #ifdef CONFIG_LIVEPATCH
1497 int patch_state;
1498 #endif
1499 #ifdef CONFIG_SECURITY
1500 /* Used by LSM modules for access restriction: */
1501 void *security;
1502 #endif
1503 #ifdef CONFIG_BPF_SYSCALL
1504 /* Used by BPF task local storage */
1505 struct bpf_local_storage __rcu *bpf_storage;
1506 /* Used for BPF run context */
1507 struct bpf_run_ctx *bpf_ctx;
1508 #endif
1509
1510 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1511 unsigned long lowest_stack;
1512 unsigned long prev_lowest_stack;
1513 #endif
1514
1515 #ifdef CONFIG_X86_MCE
1516 void __user *mce_vaddr;
1517 __u64 mce_kflags;
1518 u64 mce_addr;
1519 __u64 mce_ripv : 1,
1520 mce_whole_page : 1,
1521 __mce_reserved : 62;
1522 struct callback_head mce_kill_me;
1523 int mce_count;
1524 #endif
1525
1526 #ifdef CONFIG_KRETPROBES
1527 struct llist_head kretprobe_instances;
1528 #endif
1529 #ifdef CONFIG_RETHOOK
1530 struct llist_head rethooks;
1531 #endif
1532
1533 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1534 /*
1535 * If L1D flush is supported on mm context switch
1536 * then we use this callback head to queue kill work
1537 * to kill tasks that are not running on SMT disabled
1538 * cores
1539 */
1540 struct callback_head l1d_flush_kill;
1541 #endif
1542
1543 #ifdef CONFIG_RV
1544 /*
1545 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1546 * If we find justification for more monitors, we can think
1547 * about adding more or developing a dynamic method. So far,
1548 * none of these are justified.
1549 */
1550 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1551 #endif
1552
1553 #ifdef CONFIG_USER_EVENTS
1554 struct user_event_mm *user_event_mm;
1555 #endif
1556
1557 /*
1558 * New fields for task_struct should be added above here, so that
1559 * they are included in the randomized portion of task_struct.
1560 */
1561 randomized_struct_fields_end
1562
1563 /* CPU-specific state of this task: */
1564 struct thread_struct thread;
1565
1566 /*
1567 * WARNING: on x86, 'thread_struct' contains a variable-sized
1568 * structure. It *MUST* be at the end of 'task_struct'.
1569 *
1570 * Do not put anything below here!
1571 */
1572 };
1573
1574 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1575 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1576
__task_state_index(unsigned int tsk_state,unsigned int tsk_exit_state)1577 static inline unsigned int __task_state_index(unsigned int tsk_state,
1578 unsigned int tsk_exit_state)
1579 {
1580 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1581
1582 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1583
1584 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1585 state = TASK_REPORT_IDLE;
1586
1587 /*
1588 * We're lying here, but rather than expose a completely new task state
1589 * to userspace, we can make this appear as if the task has gone through
1590 * a regular rt_mutex_lock() call.
1591 */
1592 if (tsk_state & TASK_RTLOCK_WAIT)
1593 state = TASK_UNINTERRUPTIBLE;
1594
1595 return fls(state);
1596 }
1597
task_state_index(struct task_struct * tsk)1598 static inline unsigned int task_state_index(struct task_struct *tsk)
1599 {
1600 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1601 }
1602
task_index_to_char(unsigned int state)1603 static inline char task_index_to_char(unsigned int state)
1604 {
1605 static const char state_char[] = "RSDTtXZPI";
1606
1607 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1608
1609 return state_char[state];
1610 }
1611
task_state_to_char(struct task_struct * tsk)1612 static inline char task_state_to_char(struct task_struct *tsk)
1613 {
1614 return task_index_to_char(task_state_index(tsk));
1615 }
1616
1617 extern struct pid *cad_pid;
1618
1619 /*
1620 * Per process flags
1621 */
1622 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1623 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1624 #define PF_EXITING 0x00000004 /* Getting shut down */
1625 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1626 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1627 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1628 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1629 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1630 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1631 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1632 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1633 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1634 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1635 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1636 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1637 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1638 #define PF__HOLE__00010000 0x00010000
1639 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1640 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1641 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1642 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1643 * I am cleaning dirty pages from some other bdi. */
1644 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1645 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1646 #define PF_MEMALLOC_NORECLAIM 0x00800000 /* All allocation requests will clear __GFP_DIRECT_RECLAIM */
1647 #define PF_MEMALLOC_NOWARN 0x01000000 /* All allocation requests will inherit __GFP_NOWARN */
1648 #define PF__HOLE__02000000 0x02000000
1649 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1650 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1651 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1652 * See memalloc_pin_save() */
1653 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1654 #define PF__HOLE__40000000 0x40000000
1655 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1656
1657 /*
1658 * Only the _current_ task can read/write to tsk->flags, but other
1659 * tasks can access tsk->flags in readonly mode for example
1660 * with tsk_used_math (like during threaded core dumping).
1661 * There is however an exception to this rule during ptrace
1662 * or during fork: the ptracer task is allowed to write to the
1663 * child->flags of its traced child (same goes for fork, the parent
1664 * can write to the child->flags), because we're guaranteed the
1665 * child is not running and in turn not changing child->flags
1666 * at the same time the parent does it.
1667 */
1668 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1669 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1670 #define clear_used_math() clear_stopped_child_used_math(current)
1671 #define set_used_math() set_stopped_child_used_math(current)
1672
1673 #define conditional_stopped_child_used_math(condition, child) \
1674 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1675
1676 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1677
1678 #define copy_to_stopped_child_used_math(child) \
1679 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1680
1681 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1682 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1683 #define used_math() tsk_used_math(current)
1684
is_percpu_thread(void)1685 static __always_inline bool is_percpu_thread(void)
1686 {
1687 #ifdef CONFIG_SMP
1688 return (current->flags & PF_NO_SETAFFINITY) &&
1689 (current->nr_cpus_allowed == 1);
1690 #else
1691 return true;
1692 #endif
1693 }
1694
1695 /* Per-process atomic flags. */
1696 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1697 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1698 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1699 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1700 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1701 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1702 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1703 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1704
1705 #define TASK_PFA_TEST(name, func) \
1706 static inline bool task_##func(struct task_struct *p) \
1707 { return test_bit(PFA_##name, &p->atomic_flags); }
1708
1709 #define TASK_PFA_SET(name, func) \
1710 static inline void task_set_##func(struct task_struct *p) \
1711 { set_bit(PFA_##name, &p->atomic_flags); }
1712
1713 #define TASK_PFA_CLEAR(name, func) \
1714 static inline void task_clear_##func(struct task_struct *p) \
1715 { clear_bit(PFA_##name, &p->atomic_flags); }
1716
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1717 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1718 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1719
1720 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1721 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1722 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1723
1724 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1725 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1726 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1727
1728 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1729 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1730 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1731
1732 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1733 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1734 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1735
1736 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1737 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1738
1739 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1740 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1741 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1742
1743 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1744 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1745
1746 static inline void
1747 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1748 {
1749 current->flags &= ~flags;
1750 current->flags |= orig_flags & flags;
1751 }
1752
1753 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1754 extern int task_can_attach(struct task_struct *p);
1755 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1756 extern void dl_bw_free(int cpu, u64 dl_bw);
1757 #ifdef CONFIG_SMP
1758
1759 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1760 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1761
1762 /**
1763 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1764 * @p: the task
1765 * @new_mask: CPU affinity mask
1766 *
1767 * Return: zero if successful, or a negative error code
1768 */
1769 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1770 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1771 extern void release_user_cpus_ptr(struct task_struct *p);
1772 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1773 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1774 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1775 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1776 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1777 {
1778 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1779 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1780 {
1781 if (!cpumask_test_cpu(0, new_mask))
1782 return -EINVAL;
1783 return 0;
1784 }
dup_user_cpus_ptr(struct task_struct * dst,struct task_struct * src,int node)1785 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1786 {
1787 if (src->user_cpus_ptr)
1788 return -EINVAL;
1789 return 0;
1790 }
release_user_cpus_ptr(struct task_struct * p)1791 static inline void release_user_cpus_ptr(struct task_struct *p)
1792 {
1793 WARN_ON(p->user_cpus_ptr);
1794 }
1795
dl_task_check_affinity(struct task_struct * p,const struct cpumask * mask)1796 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1797 {
1798 return 0;
1799 }
1800 #endif
1801
1802 extern int yield_to(struct task_struct *p, bool preempt);
1803 extern void set_user_nice(struct task_struct *p, long nice);
1804 extern int task_prio(const struct task_struct *p);
1805
1806 /**
1807 * task_nice - return the nice value of a given task.
1808 * @p: the task in question.
1809 *
1810 * Return: The nice value [ -20 ... 0 ... 19 ].
1811 */
task_nice(const struct task_struct * p)1812 static inline int task_nice(const struct task_struct *p)
1813 {
1814 return PRIO_TO_NICE((p)->static_prio);
1815 }
1816
1817 extern int can_nice(const struct task_struct *p, const int nice);
1818 extern int task_curr(const struct task_struct *p);
1819 extern int idle_cpu(int cpu);
1820 extern int available_idle_cpu(int cpu);
1821 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1822 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1823 extern void sched_set_fifo(struct task_struct *p);
1824 extern void sched_set_fifo_low(struct task_struct *p);
1825 extern void sched_set_normal(struct task_struct *p, int nice);
1826 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1827 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1828 extern struct task_struct *idle_task(int cpu);
1829
1830 /**
1831 * is_idle_task - is the specified task an idle task?
1832 * @p: the task in question.
1833 *
1834 * Return: 1 if @p is an idle task. 0 otherwise.
1835 */
is_idle_task(const struct task_struct * p)1836 static __always_inline bool is_idle_task(const struct task_struct *p)
1837 {
1838 return !!(p->flags & PF_IDLE);
1839 }
1840
1841 extern struct task_struct *curr_task(int cpu);
1842 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1843
1844 void yield(void);
1845
1846 union thread_union {
1847 struct task_struct task;
1848 #ifndef CONFIG_THREAD_INFO_IN_TASK
1849 struct thread_info thread_info;
1850 #endif
1851 unsigned long stack[THREAD_SIZE/sizeof(long)];
1852 };
1853
1854 #ifndef CONFIG_THREAD_INFO_IN_TASK
1855 extern struct thread_info init_thread_info;
1856 #endif
1857
1858 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1859
1860 #ifdef CONFIG_THREAD_INFO_IN_TASK
1861 # define task_thread_info(task) (&(task)->thread_info)
1862 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1863 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1864 #endif
1865
1866 /*
1867 * find a task by one of its numerical ids
1868 *
1869 * find_task_by_pid_ns():
1870 * finds a task by its pid in the specified namespace
1871 * find_task_by_vpid():
1872 * finds a task by its virtual pid
1873 *
1874 * see also find_vpid() etc in include/linux/pid.h
1875 */
1876
1877 extern struct task_struct *find_task_by_vpid(pid_t nr);
1878 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1879
1880 /*
1881 * find a task by its virtual pid and get the task struct
1882 */
1883 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1884
1885 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1886 extern int wake_up_process(struct task_struct *tsk);
1887 extern void wake_up_new_task(struct task_struct *tsk);
1888
1889 #ifdef CONFIG_SMP
1890 extern void kick_process(struct task_struct *tsk);
1891 #else
kick_process(struct task_struct * tsk)1892 static inline void kick_process(struct task_struct *tsk) { }
1893 #endif
1894
1895 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1896
set_task_comm(struct task_struct * tsk,const char * from)1897 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1898 {
1899 __set_task_comm(tsk, from, false);
1900 }
1901
1902 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1903 #define get_task_comm(buf, tsk) ({ \
1904 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1905 __get_task_comm(buf, sizeof(buf), tsk); \
1906 })
1907
1908 #ifdef CONFIG_SMP
scheduler_ipi(void)1909 static __always_inline void scheduler_ipi(void)
1910 {
1911 /*
1912 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1913 * TIF_NEED_RESCHED remotely (for the first time) will also send
1914 * this IPI.
1915 */
1916 preempt_fold_need_resched();
1917 }
1918 #else
scheduler_ipi(void)1919 static inline void scheduler_ipi(void) { }
1920 #endif
1921
1922 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1923
1924 /*
1925 * Set thread flags in other task's structures.
1926 * See asm/thread_info.h for TIF_xxxx flags available:
1927 */
set_tsk_thread_flag(struct task_struct * tsk,int flag)1928 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1929 {
1930 set_ti_thread_flag(task_thread_info(tsk), flag);
1931 }
1932
clear_tsk_thread_flag(struct task_struct * tsk,int flag)1933 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1934 {
1935 clear_ti_thread_flag(task_thread_info(tsk), flag);
1936 }
1937
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)1938 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1939 bool value)
1940 {
1941 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1942 }
1943
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)1944 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1945 {
1946 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1947 }
1948
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)1949 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1950 {
1951 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1952 }
1953
test_tsk_thread_flag(struct task_struct * tsk,int flag)1954 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1955 {
1956 return test_ti_thread_flag(task_thread_info(tsk), flag);
1957 }
1958
set_tsk_need_resched(struct task_struct * tsk)1959 static inline void set_tsk_need_resched(struct task_struct *tsk)
1960 {
1961 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1962 }
1963
clear_tsk_need_resched(struct task_struct * tsk)1964 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1965 {
1966 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1967 }
1968
test_tsk_need_resched(struct task_struct * tsk)1969 static inline int test_tsk_need_resched(struct task_struct *tsk)
1970 {
1971 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1972 }
1973
1974 /*
1975 * cond_resched() and cond_resched_lock(): latency reduction via
1976 * explicit rescheduling in places that are safe. The return
1977 * value indicates whether a reschedule was done in fact.
1978 * cond_resched_lock() will drop the spinlock before scheduling,
1979 */
1980 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1981 extern int __cond_resched(void);
1982
1983 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
1984
1985 void sched_dynamic_klp_enable(void);
1986 void sched_dynamic_klp_disable(void);
1987
1988 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1989
_cond_resched(void)1990 static __always_inline int _cond_resched(void)
1991 {
1992 return static_call_mod(cond_resched)();
1993 }
1994
1995 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
1996
1997 extern int dynamic_cond_resched(void);
1998
_cond_resched(void)1999 static __always_inline int _cond_resched(void)
2000 {
2001 return dynamic_cond_resched();
2002 }
2003
2004 #else /* !CONFIG_PREEMPTION */
2005
_cond_resched(void)2006 static inline int _cond_resched(void)
2007 {
2008 klp_sched_try_switch();
2009 return __cond_resched();
2010 }
2011
2012 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2013
2014 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2015
_cond_resched(void)2016 static inline int _cond_resched(void)
2017 {
2018 klp_sched_try_switch();
2019 return 0;
2020 }
2021
2022 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2023
2024 #define cond_resched() ({ \
2025 __might_resched(__FILE__, __LINE__, 0); \
2026 _cond_resched(); \
2027 })
2028
2029 extern int __cond_resched_lock(spinlock_t *lock);
2030 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2031 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2032
2033 #define MIGHT_RESCHED_RCU_SHIFT 8
2034 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2035
2036 #ifndef CONFIG_PREEMPT_RT
2037 /*
2038 * Non RT kernels have an elevated preempt count due to the held lock,
2039 * but are not allowed to be inside a RCU read side critical section
2040 */
2041 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2042 #else
2043 /*
2044 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2045 * cond_resched*lock() has to take that into account because it checks for
2046 * preempt_count() and rcu_preempt_depth().
2047 */
2048 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2049 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2050 #endif
2051
2052 #define cond_resched_lock(lock) ({ \
2053 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2054 __cond_resched_lock(lock); \
2055 })
2056
2057 #define cond_resched_rwlock_read(lock) ({ \
2058 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2059 __cond_resched_rwlock_read(lock); \
2060 })
2061
2062 #define cond_resched_rwlock_write(lock) ({ \
2063 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2064 __cond_resched_rwlock_write(lock); \
2065 })
2066
2067 #ifdef CONFIG_PREEMPT_DYNAMIC
2068
2069 extern bool preempt_model_none(void);
2070 extern bool preempt_model_voluntary(void);
2071 extern bool preempt_model_full(void);
2072
2073 #else
2074
preempt_model_none(void)2075 static inline bool preempt_model_none(void)
2076 {
2077 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2078 }
preempt_model_voluntary(void)2079 static inline bool preempt_model_voluntary(void)
2080 {
2081 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2082 }
preempt_model_full(void)2083 static inline bool preempt_model_full(void)
2084 {
2085 return IS_ENABLED(CONFIG_PREEMPT);
2086 }
2087
2088 #endif
2089
preempt_model_rt(void)2090 static inline bool preempt_model_rt(void)
2091 {
2092 return IS_ENABLED(CONFIG_PREEMPT_RT);
2093 }
2094
2095 /*
2096 * Does the preemption model allow non-cooperative preemption?
2097 *
2098 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2099 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2100 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2101 * PREEMPT_NONE model.
2102 */
preempt_model_preemptible(void)2103 static inline bool preempt_model_preemptible(void)
2104 {
2105 return preempt_model_full() || preempt_model_rt();
2106 }
2107
need_resched(void)2108 static __always_inline bool need_resched(void)
2109 {
2110 return unlikely(tif_need_resched());
2111 }
2112
2113 /*
2114 * Wrappers for p->thread_info->cpu access. No-op on UP.
2115 */
2116 #ifdef CONFIG_SMP
2117
task_cpu(const struct task_struct * p)2118 static inline unsigned int task_cpu(const struct task_struct *p)
2119 {
2120 return READ_ONCE(task_thread_info(p)->cpu);
2121 }
2122
2123 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2124
2125 #else
2126
task_cpu(const struct task_struct * p)2127 static inline unsigned int task_cpu(const struct task_struct *p)
2128 {
2129 return 0;
2130 }
2131
set_task_cpu(struct task_struct * p,unsigned int cpu)2132 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2133 {
2134 }
2135
2136 #endif /* CONFIG_SMP */
2137
2138 extern bool sched_task_on_rq(struct task_struct *p);
2139 extern unsigned long get_wchan(struct task_struct *p);
2140 extern struct task_struct *cpu_curr_snapshot(int cpu);
2141
2142 #include <linux/spinlock.h>
2143
2144 /*
2145 * In order to reduce various lock holder preemption latencies provide an
2146 * interface to see if a vCPU is currently running or not.
2147 *
2148 * This allows us to terminate optimistic spin loops and block, analogous to
2149 * the native optimistic spin heuristic of testing if the lock owner task is
2150 * running or not.
2151 */
2152 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2153 static inline bool vcpu_is_preempted(int cpu)
2154 {
2155 return false;
2156 }
2157 #endif
2158
2159 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2160 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2161
2162 #ifndef TASK_SIZE_OF
2163 #define TASK_SIZE_OF(tsk) TASK_SIZE
2164 #endif
2165
2166 #ifdef CONFIG_SMP
owner_on_cpu(struct task_struct * owner)2167 static inline bool owner_on_cpu(struct task_struct *owner)
2168 {
2169 /*
2170 * As lock holder preemption issue, we both skip spinning if
2171 * task is not on cpu or its cpu is preempted
2172 */
2173 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2174 }
2175
2176 /* Returns effective CPU energy utilization, as seen by the scheduler */
2177 unsigned long sched_cpu_util(int cpu);
2178 #endif /* CONFIG_SMP */
2179
2180 #ifdef CONFIG_SCHED_CORE
2181 extern void sched_core_free(struct task_struct *tsk);
2182 extern void sched_core_fork(struct task_struct *p);
2183 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2184 unsigned long uaddr);
2185 extern int sched_core_idle_cpu(int cpu);
2186 #else
sched_core_free(struct task_struct * tsk)2187 static inline void sched_core_free(struct task_struct *tsk) { }
sched_core_fork(struct task_struct * p)2188 static inline void sched_core_fork(struct task_struct *p) { }
sched_core_idle_cpu(int cpu)2189 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2190 #endif
2191
2192 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2193
2194 #ifdef CONFIG_MEM_ALLOC_PROFILING
alloc_tag_save(struct alloc_tag * tag)2195 static inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2196 {
2197 swap(current->alloc_tag, tag);
2198 return tag;
2199 }
2200
alloc_tag_restore(struct alloc_tag * tag,struct alloc_tag * old)2201 static inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2202 {
2203 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2204 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2205 #endif
2206 current->alloc_tag = old;
2207 }
2208 #else
2209 #define alloc_tag_save(_tag) NULL
2210 #define alloc_tag_restore(_tag, _old) do {} while (0)
2211 #endif
2212
2213 #endif
2214