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
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kmsan_types.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/irqflags.h>
22 #include <linux/seccomp.h>
23 #include <linux/nodemask.h>
24 #include <linux/rcupdate.h>
25 #include <linux/refcount.h>
26 #include <linux/resource.h>
27 #include <linux/latencytop.h>
28 #include <linux/sched/prio.h>
29 #include <linux/sched/types.h>
30 #include <linux/signal_types.h>
31 #include <linux/syscall_user_dispatch.h>
32 #include <linux/mm_types_task.h>
33 #include <linux/task_io_accounting.h>
34 #include <linux/posix-timers.h>
35 #include <linux/rseq.h>
36 #include <linux/seqlock.h>
37 #include <linux/kcsan.h>
38 #include <linux/rv.h>
39 #include <linux/livepatch_sched.h>
40 #include <asm/kmap_size.h>
41
42 /* task_struct member predeclarations (sorted alphabetically): */
43 struct audit_context;
44 struct bio_list;
45 struct blk_plug;
46 struct bpf_local_storage;
47 struct bpf_run_ctx;
48 struct capture_control;
49 struct cfs_rq;
50 struct fs_struct;
51 struct futex_pi_state;
52 struct io_context;
53 struct io_uring_task;
54 struct mempolicy;
55 struct nameidata;
56 struct nsproxy;
57 struct perf_event_context;
58 struct pid_namespace;
59 struct pipe_inode_info;
60 struct rcu_node;
61 struct reclaim_state;
62 struct robust_list_head;
63 struct root_domain;
64 struct rq;
65 struct sched_attr;
66 struct sched_param;
67 struct seq_file;
68 struct sighand_struct;
69 struct signal_struct;
70 struct task_delay_info;
71 struct task_group;
72 struct user_event_mm;
73
74 /*
75 * Task state bitmask. NOTE! These bits are also
76 * encoded in fs/proc/array.c: get_task_state().
77 *
78 * We have two separate sets of flags: task->__state
79 * is about runnability, while task->exit_state are
80 * about the task exiting. Confusing, but this way
81 * modifying one set can't modify the other one by
82 * mistake.
83 */
84
85 /* Used in tsk->__state: */
86 #define TASK_RUNNING 0x00000000
87 #define TASK_INTERRUPTIBLE 0x00000001
88 #define TASK_UNINTERRUPTIBLE 0x00000002
89 #define __TASK_STOPPED 0x00000004
90 #define __TASK_TRACED 0x00000008
91 /* Used in tsk->exit_state: */
92 #define EXIT_DEAD 0x00000010
93 #define EXIT_ZOMBIE 0x00000020
94 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
95 /* Used in tsk->__state again: */
96 #define TASK_PARKED 0x00000040
97 #define TASK_DEAD 0x00000080
98 #define TASK_WAKEKILL 0x00000100
99 #define TASK_WAKING 0x00000200
100 #define TASK_NOLOAD 0x00000400
101 #define TASK_NEW 0x00000800
102 #define TASK_RTLOCK_WAIT 0x00001000
103 #define TASK_FREEZABLE 0x00002000
104 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
105 #define TASK_FROZEN 0x00008000
106 #define TASK_STATE_MAX 0x00010000
107
108 #define TASK_ANY (TASK_STATE_MAX-1)
109
110 /*
111 * DO NOT ADD ANY NEW USERS !
112 */
113 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
114
115 /* Convenience macros for the sake of set_current_state: */
116 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
117 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
118 #define TASK_TRACED __TASK_TRACED
119
120 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
121
122 /* Convenience macros for the sake of wake_up(): */
123 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
124
125 /* get_task_state(): */
126 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
127 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
128 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
129 TASK_PARKED)
130
131 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
132
133 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
134 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
135 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
136
137 /*
138 * Special states are those that do not use the normal wait-loop pattern. See
139 * the comment with set_special_state().
140 */
141 #define is_special_task_state(state) \
142 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
143
144 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
145 # define debug_normal_state_change(state_value) \
146 do { \
147 WARN_ON_ONCE(is_special_task_state(state_value)); \
148 current->task_state_change = _THIS_IP_; \
149 } while (0)
150
151 # define debug_special_state_change(state_value) \
152 do { \
153 WARN_ON_ONCE(!is_special_task_state(state_value)); \
154 current->task_state_change = _THIS_IP_; \
155 } while (0)
156
157 # define debug_rtlock_wait_set_state() \
158 do { \
159 current->saved_state_change = current->task_state_change;\
160 current->task_state_change = _THIS_IP_; \
161 } while (0)
162
163 # define debug_rtlock_wait_restore_state() \
164 do { \
165 current->task_state_change = current->saved_state_change;\
166 } while (0)
167
168 #else
169 # define debug_normal_state_change(cond) do { } while (0)
170 # define debug_special_state_change(cond) do { } while (0)
171 # define debug_rtlock_wait_set_state() do { } while (0)
172 # define debug_rtlock_wait_restore_state() do { } while (0)
173 #endif
174
175 /*
176 * set_current_state() includes a barrier so that the write of current->__state
177 * is correctly serialised wrt the caller's subsequent test of whether to
178 * actually sleep:
179 *
180 * for (;;) {
181 * set_current_state(TASK_UNINTERRUPTIBLE);
182 * if (CONDITION)
183 * break;
184 *
185 * schedule();
186 * }
187 * __set_current_state(TASK_RUNNING);
188 *
189 * If the caller does not need such serialisation (because, for instance, the
190 * CONDITION test and condition change and wakeup are under the same lock) then
191 * use __set_current_state().
192 *
193 * The above is typically ordered against the wakeup, which does:
194 *
195 * CONDITION = 1;
196 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
197 *
198 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
199 * accessing p->__state.
200 *
201 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
202 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
203 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
204 *
205 * However, with slightly different timing the wakeup TASK_RUNNING store can
206 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
207 * a problem either because that will result in one extra go around the loop
208 * and our @cond test will save the day.
209 *
210 * Also see the comments of try_to_wake_up().
211 */
212 #define __set_current_state(state_value) \
213 do { \
214 debug_normal_state_change((state_value)); \
215 WRITE_ONCE(current->__state, (state_value)); \
216 } while (0)
217
218 #define set_current_state(state_value) \
219 do { \
220 debug_normal_state_change((state_value)); \
221 smp_store_mb(current->__state, (state_value)); \
222 } while (0)
223
224 /*
225 * set_special_state() should be used for those states when the blocking task
226 * can not use the regular condition based wait-loop. In that case we must
227 * serialize against wakeups such that any possible in-flight TASK_RUNNING
228 * stores will not collide with our state change.
229 */
230 #define set_special_state(state_value) \
231 do { \
232 unsigned long flags; /* may shadow */ \
233 \
234 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
235 debug_special_state_change((state_value)); \
236 WRITE_ONCE(current->__state, (state_value)); \
237 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
238 } while (0)
239
240 /*
241 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
242 *
243 * RT's spin/rwlock substitutions are state preserving. The state of the
244 * task when blocking on the lock is saved in task_struct::saved_state and
245 * restored after the lock has been acquired. These operations are
246 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
247 * lock related wakeups while the task is blocked on the lock are
248 * redirected to operate on task_struct::saved_state to ensure that these
249 * are not dropped. On restore task_struct::saved_state is set to
250 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
251 *
252 * The lock operation looks like this:
253 *
254 * current_save_and_set_rtlock_wait_state();
255 * for (;;) {
256 * if (try_lock())
257 * break;
258 * raw_spin_unlock_irq(&lock->wait_lock);
259 * schedule_rtlock();
260 * raw_spin_lock_irq(&lock->wait_lock);
261 * set_current_state(TASK_RTLOCK_WAIT);
262 * }
263 * current_restore_rtlock_saved_state();
264 */
265 #define current_save_and_set_rtlock_wait_state() \
266 do { \
267 lockdep_assert_irqs_disabled(); \
268 raw_spin_lock(¤t->pi_lock); \
269 current->saved_state = current->__state; \
270 debug_rtlock_wait_set_state(); \
271 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
272 raw_spin_unlock(¤t->pi_lock); \
273 } while (0);
274
275 #define current_restore_rtlock_saved_state() \
276 do { \
277 lockdep_assert_irqs_disabled(); \
278 raw_spin_lock(¤t->pi_lock); \
279 debug_rtlock_wait_restore_state(); \
280 WRITE_ONCE(current->__state, current->saved_state); \
281 current->saved_state = TASK_RUNNING; \
282 raw_spin_unlock(¤t->pi_lock); \
283 } while (0);
284
285 #define get_current_state() READ_ONCE(current->__state)
286
287 /*
288 * Define the task command name length as enum, then it can be visible to
289 * BPF programs.
290 */
291 enum {
292 TASK_COMM_LEN = 16,
293 };
294
295 extern void scheduler_tick(void);
296
297 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
298
299 extern long schedule_timeout(long timeout);
300 extern long schedule_timeout_interruptible(long timeout);
301 extern long schedule_timeout_killable(long timeout);
302 extern long schedule_timeout_uninterruptible(long timeout);
303 extern long schedule_timeout_idle(long timeout);
304 asmlinkage void schedule(void);
305 extern void schedule_preempt_disabled(void);
306 asmlinkage void preempt_schedule_irq(void);
307 #ifdef CONFIG_PREEMPT_RT
308 extern void schedule_rtlock(void);
309 #endif
310
311 extern int __must_check io_schedule_prepare(void);
312 extern void io_schedule_finish(int token);
313 extern long io_schedule_timeout(long timeout);
314 extern void io_schedule(void);
315
316 /**
317 * struct prev_cputime - snapshot of system and user cputime
318 * @utime: time spent in user mode
319 * @stime: time spent in system mode
320 * @lock: protects the above two fields
321 *
322 * Stores previous user/system time values such that we can guarantee
323 * monotonicity.
324 */
325 struct prev_cputime {
326 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
327 u64 utime;
328 u64 stime;
329 raw_spinlock_t lock;
330 #endif
331 };
332
333 enum vtime_state {
334 /* Task is sleeping or running in a CPU with VTIME inactive: */
335 VTIME_INACTIVE = 0,
336 /* Task is idle */
337 VTIME_IDLE,
338 /* Task runs in kernelspace in a CPU with VTIME active: */
339 VTIME_SYS,
340 /* Task runs in userspace in a CPU with VTIME active: */
341 VTIME_USER,
342 /* Task runs as guests in a CPU with VTIME active: */
343 VTIME_GUEST,
344 };
345
346 struct vtime {
347 seqcount_t seqcount;
348 unsigned long long starttime;
349 enum vtime_state state;
350 unsigned int cpu;
351 u64 utime;
352 u64 stime;
353 u64 gtime;
354 };
355
356 /*
357 * Utilization clamp constraints.
358 * @UCLAMP_MIN: Minimum utilization
359 * @UCLAMP_MAX: Maximum utilization
360 * @UCLAMP_CNT: Utilization clamp constraints count
361 */
362 enum uclamp_id {
363 UCLAMP_MIN = 0,
364 UCLAMP_MAX,
365 UCLAMP_CNT
366 };
367
368 #ifdef CONFIG_SMP
369 extern struct root_domain def_root_domain;
370 extern struct mutex sched_domains_mutex;
371 #endif
372
373 struct sched_info {
374 #ifdef CONFIG_SCHED_INFO
375 /* Cumulative counters: */
376
377 /* # of times we have run on this CPU: */
378 unsigned long pcount;
379
380 /* Time spent waiting on a runqueue: */
381 unsigned long long run_delay;
382
383 /* Timestamps: */
384
385 /* When did we last run on a CPU? */
386 unsigned long long last_arrival;
387
388 /* When were we last queued to run? */
389 unsigned long long last_queued;
390
391 #endif /* CONFIG_SCHED_INFO */
392 };
393
394 /*
395 * Integer metrics need fixed point arithmetic, e.g., sched/fair
396 * has a few: load, load_avg, util_avg, freq, and capacity.
397 *
398 * We define a basic fixed point arithmetic range, and then formalize
399 * all these metrics based on that basic range.
400 */
401 # define SCHED_FIXEDPOINT_SHIFT 10
402 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
403
404 /* Increase resolution of cpu_capacity calculations */
405 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
406 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
407
408 struct load_weight {
409 unsigned long weight;
410 u32 inv_weight;
411 };
412
413 /**
414 * struct util_est - Estimation utilization of FAIR tasks
415 * @enqueued: instantaneous estimated utilization of a task/cpu
416 * @ewma: the Exponential Weighted Moving Average (EWMA)
417 * utilization of a task
418 *
419 * Support data structure to track an Exponential Weighted Moving Average
420 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
421 * average each time a task completes an activation. Sample's weight is chosen
422 * so that the EWMA will be relatively insensitive to transient changes to the
423 * task's workload.
424 *
425 * The enqueued attribute has a slightly different meaning for tasks and cpus:
426 * - task: the task's util_avg at last task dequeue time
427 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
428 * Thus, the util_est.enqueued of a task represents the contribution on the
429 * estimated utilization of the CPU where that task is currently enqueued.
430 *
431 * Only for tasks we track a moving average of the past instantaneous
432 * estimated utilization. This allows to absorb sporadic drops in utilization
433 * of an otherwise almost periodic task.
434 *
435 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
436 * updates. When a task is dequeued, its util_est should not be updated if its
437 * util_avg has not been updated in the meantime.
438 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
439 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
440 * for a task) it is safe to use MSB.
441 */
442 struct util_est {
443 unsigned int enqueued;
444 unsigned int ewma;
445 #define UTIL_EST_WEIGHT_SHIFT 2
446 #define UTIL_AVG_UNCHANGED 0x80000000
447 } __attribute__((__aligned__(sizeof(u64))));
448
449 /*
450 * The load/runnable/util_avg accumulates an infinite geometric series
451 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
452 *
453 * [load_avg definition]
454 *
455 * load_avg = runnable% * scale_load_down(load)
456 *
457 * [runnable_avg definition]
458 *
459 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
460 *
461 * [util_avg definition]
462 *
463 * util_avg = running% * SCHED_CAPACITY_SCALE
464 *
465 * where runnable% is the time ratio that a sched_entity is runnable and
466 * running% the time ratio that a sched_entity is running.
467 *
468 * For cfs_rq, they are the aggregated values of all runnable and blocked
469 * sched_entities.
470 *
471 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
472 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
473 * for computing those signals (see update_rq_clock_pelt())
474 *
475 * N.B., the above ratios (runnable% and running%) themselves are in the
476 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
477 * to as large a range as necessary. This is for example reflected by
478 * util_avg's SCHED_CAPACITY_SCALE.
479 *
480 * [Overflow issue]
481 *
482 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
483 * with the highest load (=88761), always runnable on a single cfs_rq,
484 * and should not overflow as the number already hits PID_MAX_LIMIT.
485 *
486 * For all other cases (including 32-bit kernels), struct load_weight's
487 * weight will overflow first before we do, because:
488 *
489 * Max(load_avg) <= Max(load.weight)
490 *
491 * Then it is the load_weight's responsibility to consider overflow
492 * issues.
493 */
494 struct sched_avg {
495 u64 last_update_time;
496 u64 load_sum;
497 u64 runnable_sum;
498 u32 util_sum;
499 u32 period_contrib;
500 unsigned long load_avg;
501 unsigned long runnable_avg;
502 unsigned long util_avg;
503 struct util_est util_est;
504 } ____cacheline_aligned;
505
506 struct sched_statistics {
507 #ifdef CONFIG_SCHEDSTATS
508 u64 wait_start;
509 u64 wait_max;
510 u64 wait_count;
511 u64 wait_sum;
512 u64 iowait_count;
513 u64 iowait_sum;
514
515 u64 sleep_start;
516 u64 sleep_max;
517 s64 sum_sleep_runtime;
518
519 u64 block_start;
520 u64 block_max;
521 s64 sum_block_runtime;
522
523 u64 exec_max;
524 u64 slice_max;
525
526 u64 nr_migrations_cold;
527 u64 nr_failed_migrations_affine;
528 u64 nr_failed_migrations_running;
529 u64 nr_failed_migrations_hot;
530 u64 nr_forced_migrations;
531
532 u64 nr_wakeups;
533 u64 nr_wakeups_sync;
534 u64 nr_wakeups_migrate;
535 u64 nr_wakeups_local;
536 u64 nr_wakeups_remote;
537 u64 nr_wakeups_affine;
538 u64 nr_wakeups_affine_attempts;
539 u64 nr_wakeups_passive;
540 u64 nr_wakeups_idle;
541
542 #ifdef CONFIG_SCHED_CORE
543 u64 core_forceidle_sum;
544 #endif
545 #endif /* CONFIG_SCHEDSTATS */
546 } ____cacheline_aligned;
547
548 struct sched_entity {
549 /* For load-balancing: */
550 struct load_weight load;
551 struct rb_node run_node;
552 u64 deadline;
553 u64 min_deadline;
554
555 struct list_head group_node;
556 unsigned int on_rq;
557
558 u64 exec_start;
559 u64 sum_exec_runtime;
560 u64 prev_sum_exec_runtime;
561 u64 vruntime;
562 s64 vlag;
563 u64 slice;
564
565 u64 nr_migrations;
566
567 #ifdef CONFIG_FAIR_GROUP_SCHED
568 int depth;
569 struct sched_entity *parent;
570 /* rq on which this entity is (to be) queued: */
571 struct cfs_rq *cfs_rq;
572 /* rq "owned" by this entity/group: */
573 struct cfs_rq *my_q;
574 /* cached value of my_q->h_nr_running */
575 unsigned long runnable_weight;
576 #endif
577
578 #ifdef CONFIG_SMP
579 /*
580 * Per entity load average tracking.
581 *
582 * Put into separate cache line so it does not
583 * collide with read-mostly values above.
584 */
585 struct sched_avg avg;
586 #endif
587 };
588
589 struct sched_rt_entity {
590 struct list_head run_list;
591 unsigned long timeout;
592 unsigned long watchdog_stamp;
593 unsigned int time_slice;
594 unsigned short on_rq;
595 unsigned short on_list;
596
597 struct sched_rt_entity *back;
598 #ifdef CONFIG_RT_GROUP_SCHED
599 struct sched_rt_entity *parent;
600 /* rq on which this entity is (to be) queued: */
601 struct rt_rq *rt_rq;
602 /* rq "owned" by this entity/group: */
603 struct rt_rq *my_q;
604 #endif
605 } __randomize_layout;
606
607 struct sched_dl_entity {
608 struct rb_node rb_node;
609
610 /*
611 * Original scheduling parameters. Copied here from sched_attr
612 * during sched_setattr(), they will remain the same until
613 * the next sched_setattr().
614 */
615 u64 dl_runtime; /* Maximum runtime for each instance */
616 u64 dl_deadline; /* Relative deadline of each instance */
617 u64 dl_period; /* Separation of two instances (period) */
618 u64 dl_bw; /* dl_runtime / dl_period */
619 u64 dl_density; /* dl_runtime / dl_deadline */
620
621 /*
622 * Actual scheduling parameters. Initialized with the values above,
623 * they are continuously updated during task execution. Note that
624 * the remaining runtime could be < 0 in case we are in overrun.
625 */
626 s64 runtime; /* Remaining runtime for this instance */
627 u64 deadline; /* Absolute deadline for this instance */
628 unsigned int flags; /* Specifying the scheduler behaviour */
629
630 /*
631 * Some bool flags:
632 *
633 * @dl_throttled tells if we exhausted the runtime. If so, the
634 * task has to wait for a replenishment to be performed at the
635 * next firing of dl_timer.
636 *
637 * @dl_yielded tells if task gave up the CPU before consuming
638 * all its available runtime during the last job.
639 *
640 * @dl_non_contending tells if the task is inactive while still
641 * contributing to the active utilization. In other words, it
642 * indicates if the inactive timer has been armed and its handler
643 * has not been executed yet. This flag is useful to avoid race
644 * conditions between the inactive timer handler and the wakeup
645 * code.
646 *
647 * @dl_overrun tells if the task asked to be informed about runtime
648 * overruns.
649 */
650 unsigned int dl_throttled : 1;
651 unsigned int dl_yielded : 1;
652 unsigned int dl_non_contending : 1;
653 unsigned int dl_overrun : 1;
654
655 /*
656 * Bandwidth enforcement timer. Each -deadline task has its
657 * own bandwidth to be enforced, thus we need one timer per task.
658 */
659 struct hrtimer dl_timer;
660
661 /*
662 * Inactive timer, responsible for decreasing the active utilization
663 * at the "0-lag time". When a -deadline task blocks, it contributes
664 * to GRUB's active utilization until the "0-lag time", hence a
665 * timer is needed to decrease the active utilization at the correct
666 * time.
667 */
668 struct hrtimer inactive_timer;
669
670 #ifdef CONFIG_RT_MUTEXES
671 /*
672 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
673 * pi_se points to the donor, otherwise points to the dl_se it belongs
674 * to (the original one/itself).
675 */
676 struct sched_dl_entity *pi_se;
677 #endif
678 };
679
680 #ifdef CONFIG_UCLAMP_TASK
681 /* Number of utilization clamp buckets (shorter alias) */
682 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
683
684 /*
685 * Utilization clamp for a scheduling entity
686 * @value: clamp value "assigned" to a se
687 * @bucket_id: bucket index corresponding to the "assigned" value
688 * @active: the se is currently refcounted in a rq's bucket
689 * @user_defined: the requested clamp value comes from user-space
690 *
691 * The bucket_id is the index of the clamp bucket matching the clamp value
692 * which is pre-computed and stored to avoid expensive integer divisions from
693 * the fast path.
694 *
695 * The active bit is set whenever a task has got an "effective" value assigned,
696 * which can be different from the clamp value "requested" from user-space.
697 * This allows to know a task is refcounted in the rq's bucket corresponding
698 * to the "effective" bucket_id.
699 *
700 * The user_defined bit is set whenever a task has got a task-specific clamp
701 * value requested from userspace, i.e. the system defaults apply to this task
702 * just as a restriction. This allows to relax default clamps when a less
703 * restrictive task-specific value has been requested, thus allowing to
704 * implement a "nice" semantic. For example, a task running with a 20%
705 * default boost can still drop its own boosting to 0%.
706 */
707 struct uclamp_se {
708 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
709 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
710 unsigned int active : 1;
711 unsigned int user_defined : 1;
712 };
713 #endif /* CONFIG_UCLAMP_TASK */
714
715 union rcu_special {
716 struct {
717 u8 blocked;
718 u8 need_qs;
719 u8 exp_hint; /* Hint for performance. */
720 u8 need_mb; /* Readers need smp_mb(). */
721 } b; /* Bits. */
722 u32 s; /* Set of bits. */
723 };
724
725 enum perf_event_task_context {
726 perf_invalid_context = -1,
727 perf_hw_context = 0,
728 perf_sw_context,
729 perf_nr_task_contexts,
730 };
731
732 struct wake_q_node {
733 struct wake_q_node *next;
734 };
735
736 struct kmap_ctrl {
737 #ifdef CONFIG_KMAP_LOCAL
738 int idx;
739 pte_t pteval[KM_MAX_IDX];
740 #endif
741 };
742
743 struct task_struct {
744 #ifdef CONFIG_THREAD_INFO_IN_TASK
745 /*
746 * For reasons of header soup (see current_thread_info()), this
747 * must be the first element of task_struct.
748 */
749 struct thread_info thread_info;
750 #endif
751 unsigned int __state;
752
753 #ifdef CONFIG_PREEMPT_RT
754 /* saved state for "spinlock sleepers" */
755 unsigned int saved_state;
756 #endif
757
758 /*
759 * This begins the randomizable portion of task_struct. Only
760 * scheduling-critical items should be added above here.
761 */
762 randomized_struct_fields_start
763
764 void *stack;
765 refcount_t usage;
766 /* Per task flags (PF_*), defined further below: */
767 unsigned int flags;
768 unsigned int ptrace;
769
770 #ifdef CONFIG_SMP
771 int on_cpu;
772 struct __call_single_node wake_entry;
773 unsigned int wakee_flips;
774 unsigned long wakee_flip_decay_ts;
775 struct task_struct *last_wakee;
776
777 /*
778 * recent_used_cpu is initially set as the last CPU used by a task
779 * that wakes affine another task. Waker/wakee relationships can
780 * push tasks around a CPU where each wakeup moves to the next one.
781 * Tracking a recently used CPU allows a quick search for a recently
782 * used CPU that may be idle.
783 */
784 int recent_used_cpu;
785 int wake_cpu;
786 #endif
787 int on_rq;
788
789 int prio;
790 int static_prio;
791 int normal_prio;
792 unsigned int rt_priority;
793
794 struct sched_entity se;
795 struct sched_rt_entity rt;
796 struct sched_dl_entity dl;
797 const struct sched_class *sched_class;
798
799 #ifdef CONFIG_SCHED_CORE
800 struct rb_node core_node;
801 unsigned long core_cookie;
802 unsigned int core_occupation;
803 #endif
804
805 #ifdef CONFIG_CGROUP_SCHED
806 struct task_group *sched_task_group;
807 #endif
808
809 #ifdef CONFIG_UCLAMP_TASK
810 /*
811 * Clamp values requested for a scheduling entity.
812 * Must be updated with task_rq_lock() held.
813 */
814 struct uclamp_se uclamp_req[UCLAMP_CNT];
815 /*
816 * Effective clamp values used for a scheduling entity.
817 * Must be updated with task_rq_lock() held.
818 */
819 struct uclamp_se uclamp[UCLAMP_CNT];
820 #endif
821
822 struct sched_statistics stats;
823
824 #ifdef CONFIG_PREEMPT_NOTIFIERS
825 /* List of struct preempt_notifier: */
826 struct hlist_head preempt_notifiers;
827 #endif
828
829 #ifdef CONFIG_BLK_DEV_IO_TRACE
830 unsigned int btrace_seq;
831 #endif
832
833 unsigned int policy;
834 int nr_cpus_allowed;
835 const cpumask_t *cpus_ptr;
836 cpumask_t *user_cpus_ptr;
837 cpumask_t cpus_mask;
838 void *migration_pending;
839 #ifdef CONFIG_SMP
840 unsigned short migration_disabled;
841 #endif
842 unsigned short migration_flags;
843
844 #ifdef CONFIG_PREEMPT_RCU
845 int rcu_read_lock_nesting;
846 union rcu_special rcu_read_unlock_special;
847 struct list_head rcu_node_entry;
848 struct rcu_node *rcu_blocked_node;
849 #endif /* #ifdef CONFIG_PREEMPT_RCU */
850
851 #ifdef CONFIG_TASKS_RCU
852 unsigned long rcu_tasks_nvcsw;
853 u8 rcu_tasks_holdout;
854 u8 rcu_tasks_idx;
855 int rcu_tasks_idle_cpu;
856 struct list_head rcu_tasks_holdout_list;
857 #endif /* #ifdef CONFIG_TASKS_RCU */
858
859 #ifdef CONFIG_TASKS_TRACE_RCU
860 int trc_reader_nesting;
861 int trc_ipi_to_cpu;
862 union rcu_special trc_reader_special;
863 struct list_head trc_holdout_list;
864 struct list_head trc_blkd_node;
865 int trc_blkd_cpu;
866 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
867
868 struct sched_info sched_info;
869
870 struct list_head tasks;
871 #ifdef CONFIG_SMP
872 struct plist_node pushable_tasks;
873 struct rb_node pushable_dl_tasks;
874 #endif
875
876 struct mm_struct *mm;
877 struct mm_struct *active_mm;
878
879 int exit_state;
880 int exit_code;
881 int exit_signal;
882 /* The signal sent when the parent dies: */
883 int pdeath_signal;
884 /* JOBCTL_*, siglock protected: */
885 unsigned long jobctl;
886
887 /* Used for emulating ABI behavior of previous Linux versions: */
888 unsigned int personality;
889
890 /* Scheduler bits, serialized by scheduler locks: */
891 unsigned sched_reset_on_fork:1;
892 unsigned sched_contributes_to_load:1;
893 unsigned sched_migrated:1;
894
895 /* Force alignment to the next boundary: */
896 unsigned :0;
897
898 /* Unserialized, strictly 'current' */
899
900 /*
901 * This field must not be in the scheduler word above due to wakelist
902 * queueing no longer being serialized by p->on_cpu. However:
903 *
904 * p->XXX = X; ttwu()
905 * schedule() if (p->on_rq && ..) // false
906 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
907 * deactivate_task() ttwu_queue_wakelist())
908 * p->on_rq = 0; p->sched_remote_wakeup = Y;
909 *
910 * guarantees all stores of 'current' are visible before
911 * ->sched_remote_wakeup gets used, so it can be in this word.
912 */
913 unsigned sched_remote_wakeup:1;
914
915 /* Bit to tell LSMs we're in execve(): */
916 unsigned in_execve:1;
917 unsigned in_iowait:1;
918 #ifndef TIF_RESTORE_SIGMASK
919 unsigned restore_sigmask:1;
920 #endif
921 #ifdef CONFIG_MEMCG
922 unsigned in_user_fault:1;
923 #endif
924 #ifdef CONFIG_LRU_GEN
925 /* whether the LRU algorithm may apply to this access */
926 unsigned in_lru_fault:1;
927 #endif
928 #ifdef CONFIG_COMPAT_BRK
929 unsigned brk_randomized:1;
930 #endif
931 #ifdef CONFIG_CGROUPS
932 /* disallow userland-initiated cgroup migration */
933 unsigned no_cgroup_migration:1;
934 /* task is frozen/stopped (used by the cgroup freezer) */
935 unsigned frozen:1;
936 #endif
937 #ifdef CONFIG_BLK_CGROUP
938 unsigned use_memdelay:1;
939 #endif
940 #ifdef CONFIG_PSI
941 /* Stalled due to lack of memory */
942 unsigned in_memstall:1;
943 #endif
944 #ifdef CONFIG_PAGE_OWNER
945 /* Used by page_owner=on to detect recursion in page tracking. */
946 unsigned in_page_owner:1;
947 #endif
948 #ifdef CONFIG_EVENTFD
949 /* Recursion prevention for eventfd_signal() */
950 unsigned in_eventfd:1;
951 #endif
952 #ifdef CONFIG_IOMMU_SVA
953 unsigned pasid_activated:1;
954 #endif
955 #ifdef CONFIG_CPU_SUP_INTEL
956 unsigned reported_split_lock:1;
957 #endif
958 #ifdef CONFIG_TASK_DELAY_ACCT
959 /* delay due to memory thrashing */
960 unsigned in_thrashing:1;
961 #endif
962
963 unsigned long atomic_flags; /* Flags requiring atomic access. */
964
965 struct restart_block restart_block;
966
967 pid_t pid;
968 pid_t tgid;
969
970 #ifdef CONFIG_STACKPROTECTOR
971 /* Canary value for the -fstack-protector GCC feature: */
972 unsigned long stack_canary;
973 #endif
974 /*
975 * Pointers to the (original) parent process, youngest child, younger sibling,
976 * older sibling, respectively. (p->father can be replaced with
977 * p->real_parent->pid)
978 */
979
980 /* Real parent process: */
981 struct task_struct __rcu *real_parent;
982
983 /* Recipient of SIGCHLD, wait4() reports: */
984 struct task_struct __rcu *parent;
985
986 /*
987 * Children/sibling form the list of natural children:
988 */
989 struct list_head children;
990 struct list_head sibling;
991 struct task_struct *group_leader;
992
993 /*
994 * 'ptraced' is the list of tasks this task is using ptrace() on.
995 *
996 * This includes both natural children and PTRACE_ATTACH targets.
997 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
998 */
999 struct list_head ptraced;
1000 struct list_head ptrace_entry;
1001
1002 /* PID/PID hash table linkage. */
1003 struct pid *thread_pid;
1004 struct hlist_node pid_links[PIDTYPE_MAX];
1005 struct list_head thread_group;
1006 struct list_head thread_node;
1007
1008 struct completion *vfork_done;
1009
1010 /* CLONE_CHILD_SETTID: */
1011 int __user *set_child_tid;
1012
1013 /* CLONE_CHILD_CLEARTID: */
1014 int __user *clear_child_tid;
1015
1016 /* PF_KTHREAD | PF_IO_WORKER */
1017 void *worker_private;
1018
1019 u64 utime;
1020 u64 stime;
1021 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1022 u64 utimescaled;
1023 u64 stimescaled;
1024 #endif
1025 u64 gtime;
1026 struct prev_cputime prev_cputime;
1027 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1028 struct vtime vtime;
1029 #endif
1030
1031 #ifdef CONFIG_NO_HZ_FULL
1032 atomic_t tick_dep_mask;
1033 #endif
1034 /* Context switch counts: */
1035 unsigned long nvcsw;
1036 unsigned long nivcsw;
1037
1038 /* Monotonic time in nsecs: */
1039 u64 start_time;
1040
1041 /* Boot based time in nsecs: */
1042 u64 start_boottime;
1043
1044 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1045 unsigned long min_flt;
1046 unsigned long maj_flt;
1047
1048 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1049 struct posix_cputimers posix_cputimers;
1050
1051 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1052 struct posix_cputimers_work posix_cputimers_work;
1053 #endif
1054
1055 /* Process credentials: */
1056
1057 /* Tracer's credentials at attach: */
1058 const struct cred __rcu *ptracer_cred;
1059
1060 /* Objective and real subjective task credentials (COW): */
1061 const struct cred __rcu *real_cred;
1062
1063 /* Effective (overridable) subjective task credentials (COW): */
1064 const struct cred __rcu *cred;
1065
1066 #ifdef CONFIG_KEYS
1067 /* Cached requested key. */
1068 struct key *cached_requested_key;
1069 #endif
1070
1071 /*
1072 * executable name, excluding path.
1073 *
1074 * - normally initialized setup_new_exec()
1075 * - access it with [gs]et_task_comm()
1076 * - lock it with task_lock()
1077 */
1078 char comm[TASK_COMM_LEN];
1079
1080 struct nameidata *nameidata;
1081
1082 #ifdef CONFIG_SYSVIPC
1083 struct sysv_sem sysvsem;
1084 struct sysv_shm sysvshm;
1085 #endif
1086 #ifdef CONFIG_DETECT_HUNG_TASK
1087 unsigned long last_switch_count;
1088 unsigned long last_switch_time;
1089 #endif
1090 /* Filesystem information: */
1091 struct fs_struct *fs;
1092
1093 /* Open file information: */
1094 struct files_struct *files;
1095
1096 #ifdef CONFIG_IO_URING
1097 struct io_uring_task *io_uring;
1098 #endif
1099
1100 /* Namespaces: */
1101 struct nsproxy *nsproxy;
1102
1103 /* Signal handlers: */
1104 struct signal_struct *signal;
1105 struct sighand_struct __rcu *sighand;
1106 sigset_t blocked;
1107 sigset_t real_blocked;
1108 /* Restored if set_restore_sigmask() was used: */
1109 sigset_t saved_sigmask;
1110 struct sigpending pending;
1111 unsigned long sas_ss_sp;
1112 size_t sas_ss_size;
1113 unsigned int sas_ss_flags;
1114
1115 struct callback_head *task_works;
1116
1117 #ifdef CONFIG_AUDIT
1118 #ifdef CONFIG_AUDITSYSCALL
1119 struct audit_context *audit_context;
1120 #endif
1121 kuid_t loginuid;
1122 unsigned int sessionid;
1123 #endif
1124 struct seccomp seccomp;
1125 struct syscall_user_dispatch syscall_dispatch;
1126
1127 /* Thread group tracking: */
1128 u64 parent_exec_id;
1129 u64 self_exec_id;
1130
1131 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1132 spinlock_t alloc_lock;
1133
1134 /* Protection of the PI data structures: */
1135 raw_spinlock_t pi_lock;
1136
1137 struct wake_q_node wake_q;
1138
1139 #ifdef CONFIG_RT_MUTEXES
1140 /* PI waiters blocked on a rt_mutex held by this task: */
1141 struct rb_root_cached pi_waiters;
1142 /* Updated under owner's pi_lock and rq lock */
1143 struct task_struct *pi_top_task;
1144 /* Deadlock detection and priority inheritance handling: */
1145 struct rt_mutex_waiter *pi_blocked_on;
1146 #endif
1147
1148 #ifdef CONFIG_DEBUG_MUTEXES
1149 /* Mutex deadlock detection: */
1150 struct mutex_waiter *blocked_on;
1151 #endif
1152
1153 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1154 int non_block_count;
1155 #endif
1156
1157 #ifdef CONFIG_TRACE_IRQFLAGS
1158 struct irqtrace_events irqtrace;
1159 unsigned int hardirq_threaded;
1160 u64 hardirq_chain_key;
1161 int softirqs_enabled;
1162 int softirq_context;
1163 int irq_config;
1164 #endif
1165 #ifdef CONFIG_PREEMPT_RT
1166 int softirq_disable_cnt;
1167 #endif
1168
1169 #ifdef CONFIG_LOCKDEP
1170 # define MAX_LOCK_DEPTH 48UL
1171 u64 curr_chain_key;
1172 int lockdep_depth;
1173 unsigned int lockdep_recursion;
1174 struct held_lock held_locks[MAX_LOCK_DEPTH];
1175 #endif
1176
1177 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1178 unsigned int in_ubsan;
1179 #endif
1180
1181 /* Journalling filesystem info: */
1182 void *journal_info;
1183
1184 /* Stacked block device info: */
1185 struct bio_list *bio_list;
1186
1187 /* Stack plugging: */
1188 struct blk_plug *plug;
1189
1190 /* VM state: */
1191 struct reclaim_state *reclaim_state;
1192
1193 struct io_context *io_context;
1194
1195 #ifdef CONFIG_COMPACTION
1196 struct capture_control *capture_control;
1197 #endif
1198 /* Ptrace state: */
1199 unsigned long ptrace_message;
1200 kernel_siginfo_t *last_siginfo;
1201
1202 struct task_io_accounting ioac;
1203 #ifdef CONFIG_PSI
1204 /* Pressure stall state */
1205 unsigned int psi_flags;
1206 #endif
1207 #ifdef CONFIG_TASK_XACCT
1208 /* Accumulated RSS usage: */
1209 u64 acct_rss_mem1;
1210 /* Accumulated virtual memory usage: */
1211 u64 acct_vm_mem1;
1212 /* stime + utime since last update: */
1213 u64 acct_timexpd;
1214 #endif
1215 #ifdef CONFIG_CPUSETS
1216 /* Protected by ->alloc_lock: */
1217 nodemask_t mems_allowed;
1218 /* Sequence number to catch updates: */
1219 seqcount_spinlock_t mems_allowed_seq;
1220 int cpuset_mem_spread_rotor;
1221 int cpuset_slab_spread_rotor;
1222 #endif
1223 #ifdef CONFIG_CGROUPS
1224 /* Control Group info protected by css_set_lock: */
1225 struct css_set __rcu *cgroups;
1226 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1227 struct list_head cg_list;
1228 #endif
1229 #ifdef CONFIG_X86_CPU_RESCTRL
1230 u32 closid;
1231 u32 rmid;
1232 #endif
1233 #ifdef CONFIG_FUTEX
1234 struct robust_list_head __user *robust_list;
1235 #ifdef CONFIG_COMPAT
1236 struct compat_robust_list_head __user *compat_robust_list;
1237 #endif
1238 struct list_head pi_state_list;
1239 struct futex_pi_state *pi_state_cache;
1240 struct mutex futex_exit_mutex;
1241 unsigned int futex_state;
1242 #endif
1243 #ifdef CONFIG_PERF_EVENTS
1244 struct perf_event_context *perf_event_ctxp;
1245 struct mutex perf_event_mutex;
1246 struct list_head perf_event_list;
1247 #endif
1248 #ifdef CONFIG_DEBUG_PREEMPT
1249 unsigned long preempt_disable_ip;
1250 #endif
1251 #ifdef CONFIG_NUMA
1252 /* Protected by alloc_lock: */
1253 struct mempolicy *mempolicy;
1254 short il_prev;
1255 short pref_node_fork;
1256 #endif
1257 #ifdef CONFIG_NUMA_BALANCING
1258 int numa_scan_seq;
1259 unsigned int numa_scan_period;
1260 unsigned int numa_scan_period_max;
1261 int numa_preferred_nid;
1262 unsigned long numa_migrate_retry;
1263 /* Migration stamp: */
1264 u64 node_stamp;
1265 u64 last_task_numa_placement;
1266 u64 last_sum_exec_runtime;
1267 struct callback_head numa_work;
1268
1269 /*
1270 * This pointer is only modified for current in syscall and
1271 * pagefault context (and for tasks being destroyed), so it can be read
1272 * from any of the following contexts:
1273 * - RCU read-side critical section
1274 * - current->numa_group from everywhere
1275 * - task's runqueue locked, task not running
1276 */
1277 struct numa_group __rcu *numa_group;
1278
1279 /*
1280 * numa_faults is an array split into four regions:
1281 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1282 * in this precise order.
1283 *
1284 * faults_memory: Exponential decaying average of faults on a per-node
1285 * basis. Scheduling placement decisions are made based on these
1286 * counts. The values remain static for the duration of a PTE scan.
1287 * faults_cpu: Track the nodes the process was running on when a NUMA
1288 * hinting fault was incurred.
1289 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1290 * during the current scan window. When the scan completes, the counts
1291 * in faults_memory and faults_cpu decay and these values are copied.
1292 */
1293 unsigned long *numa_faults;
1294 unsigned long total_numa_faults;
1295
1296 /*
1297 * numa_faults_locality tracks if faults recorded during the last
1298 * scan window were remote/local or failed to migrate. The task scan
1299 * period is adapted based on the locality of the faults with different
1300 * weights depending on whether they were shared or private faults
1301 */
1302 unsigned long numa_faults_locality[3];
1303
1304 unsigned long numa_pages_migrated;
1305 #endif /* CONFIG_NUMA_BALANCING */
1306
1307 #ifdef CONFIG_RSEQ
1308 struct rseq __user *rseq;
1309 u32 rseq_len;
1310 u32 rseq_sig;
1311 /*
1312 * RmW on rseq_event_mask must be performed atomically
1313 * with respect to preemption.
1314 */
1315 unsigned long rseq_event_mask;
1316 #endif
1317
1318 #ifdef CONFIG_SCHED_MM_CID
1319 int mm_cid; /* Current cid in mm */
1320 int last_mm_cid; /* Most recent cid in mm */
1321 int migrate_from_cpu;
1322 int mm_cid_active; /* Whether cid bitmap is active */
1323 struct callback_head cid_work;
1324 #endif
1325
1326 struct tlbflush_unmap_batch tlb_ubc;
1327
1328 /* Cache last used pipe for splice(): */
1329 struct pipe_inode_info *splice_pipe;
1330
1331 struct page_frag task_frag;
1332
1333 #ifdef CONFIG_TASK_DELAY_ACCT
1334 struct task_delay_info *delays;
1335 #endif
1336
1337 #ifdef CONFIG_FAULT_INJECTION
1338 int make_it_fail;
1339 unsigned int fail_nth;
1340 #endif
1341 /*
1342 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1343 * balance_dirty_pages() for a dirty throttling pause:
1344 */
1345 int nr_dirtied;
1346 int nr_dirtied_pause;
1347 /* Start of a write-and-pause period: */
1348 unsigned long dirty_paused_when;
1349
1350 #ifdef CONFIG_LATENCYTOP
1351 int latency_record_count;
1352 struct latency_record latency_record[LT_SAVECOUNT];
1353 #endif
1354 /*
1355 * Time slack values; these are used to round up poll() and
1356 * select() etc timeout values. These are in nanoseconds.
1357 */
1358 u64 timer_slack_ns;
1359 u64 default_timer_slack_ns;
1360
1361 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1362 unsigned int kasan_depth;
1363 #endif
1364
1365 #ifdef CONFIG_KCSAN
1366 struct kcsan_ctx kcsan_ctx;
1367 #ifdef CONFIG_TRACE_IRQFLAGS
1368 struct irqtrace_events kcsan_save_irqtrace;
1369 #endif
1370 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1371 int kcsan_stack_depth;
1372 #endif
1373 #endif
1374
1375 #ifdef CONFIG_KMSAN
1376 struct kmsan_ctx kmsan_ctx;
1377 #endif
1378
1379 #if IS_ENABLED(CONFIG_KUNIT)
1380 struct kunit *kunit_test;
1381 #endif
1382
1383 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1384 /* Index of current stored address in ret_stack: */
1385 int curr_ret_stack;
1386 int curr_ret_depth;
1387
1388 /* Stack of return addresses for return function tracing: */
1389 struct ftrace_ret_stack *ret_stack;
1390
1391 /* Timestamp for last schedule: */
1392 unsigned long long ftrace_timestamp;
1393
1394 /*
1395 * Number of functions that haven't been traced
1396 * because of depth overrun:
1397 */
1398 atomic_t trace_overrun;
1399
1400 /* Pause tracing: */
1401 atomic_t tracing_graph_pause;
1402 #endif
1403
1404 #ifdef CONFIG_TRACING
1405 /* Bitmask and counter of trace recursion: */
1406 unsigned long trace_recursion;
1407 #endif /* CONFIG_TRACING */
1408
1409 #ifdef CONFIG_KCOV
1410 /* See kernel/kcov.c for more details. */
1411
1412 /* Coverage collection mode enabled for this task (0 if disabled): */
1413 unsigned int kcov_mode;
1414
1415 /* Size of the kcov_area: */
1416 unsigned int kcov_size;
1417
1418 /* Buffer for coverage collection: */
1419 void *kcov_area;
1420
1421 /* KCOV descriptor wired with this task or NULL: */
1422 struct kcov *kcov;
1423
1424 /* KCOV common handle for remote coverage collection: */
1425 u64 kcov_handle;
1426
1427 /* KCOV sequence number: */
1428 int kcov_sequence;
1429
1430 /* Collect coverage from softirq context: */
1431 unsigned int kcov_softirq;
1432 #endif
1433
1434 #ifdef CONFIG_MEMCG
1435 struct mem_cgroup *memcg_in_oom;
1436 gfp_t memcg_oom_gfp_mask;
1437 int memcg_oom_order;
1438
1439 /* Number of pages to reclaim on returning to userland: */
1440 unsigned int memcg_nr_pages_over_high;
1441
1442 /* Used by memcontrol for targeted memcg charge: */
1443 struct mem_cgroup *active_memcg;
1444 #endif
1445
1446 #ifdef CONFIG_BLK_CGROUP
1447 struct gendisk *throttle_disk;
1448 #endif
1449
1450 #ifdef CONFIG_UPROBES
1451 struct uprobe_task *utask;
1452 #endif
1453 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1454 unsigned int sequential_io;
1455 unsigned int sequential_io_avg;
1456 #endif
1457 struct kmap_ctrl kmap_ctrl;
1458 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1459 unsigned long task_state_change;
1460 # ifdef CONFIG_PREEMPT_RT
1461 unsigned long saved_state_change;
1462 # endif
1463 #endif
1464 struct rcu_head rcu;
1465 refcount_t rcu_users;
1466 int pagefault_disabled;
1467 #ifdef CONFIG_MMU
1468 struct task_struct *oom_reaper_list;
1469 struct timer_list oom_reaper_timer;
1470 #endif
1471 #ifdef CONFIG_VMAP_STACK
1472 struct vm_struct *stack_vm_area;
1473 #endif
1474 #ifdef CONFIG_THREAD_INFO_IN_TASK
1475 /* A live task holds one reference: */
1476 refcount_t stack_refcount;
1477 #endif
1478 #ifdef CONFIG_LIVEPATCH
1479 int patch_state;
1480 #endif
1481 #ifdef CONFIG_SECURITY
1482 /* Used by LSM modules for access restriction: */
1483 void *security;
1484 #endif
1485 #ifdef CONFIG_BPF_SYSCALL
1486 /* Used by BPF task local storage */
1487 struct bpf_local_storage __rcu *bpf_storage;
1488 /* Used for BPF run context */
1489 struct bpf_run_ctx *bpf_ctx;
1490 #endif
1491
1492 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1493 unsigned long lowest_stack;
1494 unsigned long prev_lowest_stack;
1495 #endif
1496
1497 #ifdef CONFIG_X86_MCE
1498 void __user *mce_vaddr;
1499 __u64 mce_kflags;
1500 u64 mce_addr;
1501 __u64 mce_ripv : 1,
1502 mce_whole_page : 1,
1503 __mce_reserved : 62;
1504 struct callback_head mce_kill_me;
1505 int mce_count;
1506 #endif
1507
1508 #ifdef CONFIG_KRETPROBES
1509 struct llist_head kretprobe_instances;
1510 #endif
1511 #ifdef CONFIG_RETHOOK
1512 struct llist_head rethooks;
1513 #endif
1514
1515 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1516 /*
1517 * If L1D flush is supported on mm context switch
1518 * then we use this callback head to queue kill work
1519 * to kill tasks that are not running on SMT disabled
1520 * cores
1521 */
1522 struct callback_head l1d_flush_kill;
1523 #endif
1524
1525 #ifdef CONFIG_RV
1526 /*
1527 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1528 * If we find justification for more monitors, we can think
1529 * about adding more or developing a dynamic method. So far,
1530 * none of these are justified.
1531 */
1532 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1533 #endif
1534
1535 #ifdef CONFIG_USER_EVENTS
1536 struct user_event_mm *user_event_mm;
1537 #endif
1538
1539 /*
1540 * New fields for task_struct should be added above here, so that
1541 * they are included in the randomized portion of task_struct.
1542 */
1543 randomized_struct_fields_end
1544
1545 /* CPU-specific state of this task: */
1546 struct thread_struct thread;
1547
1548 /*
1549 * WARNING: on x86, 'thread_struct' contains a variable-sized
1550 * structure. It *MUST* be at the end of 'task_struct'.
1551 *
1552 * Do not put anything below here!
1553 */
1554 };
1555
task_pid(struct task_struct * task)1556 static inline struct pid *task_pid(struct task_struct *task)
1557 {
1558 return task->thread_pid;
1559 }
1560
1561 /*
1562 * the helpers to get the task's different pids as they are seen
1563 * from various namespaces
1564 *
1565 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1566 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1567 * current.
1568 * task_xid_nr_ns() : id seen from the ns specified;
1569 *
1570 * see also pid_nr() etc in include/linux/pid.h
1571 */
1572 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1573
task_pid_nr(struct task_struct * tsk)1574 static inline pid_t task_pid_nr(struct task_struct *tsk)
1575 {
1576 return tsk->pid;
1577 }
1578
task_pid_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1579 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1580 {
1581 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1582 }
1583
task_pid_vnr(struct task_struct * tsk)1584 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1585 {
1586 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1587 }
1588
1589
task_tgid_nr(struct task_struct * tsk)1590 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1591 {
1592 return tsk->tgid;
1593 }
1594
1595 /**
1596 * pid_alive - check that a task structure is not stale
1597 * @p: Task structure to be checked.
1598 *
1599 * Test if a process is not yet dead (at most zombie state)
1600 * If pid_alive fails, then pointers within the task structure
1601 * can be stale and must not be dereferenced.
1602 *
1603 * Return: 1 if the process is alive. 0 otherwise.
1604 */
pid_alive(const struct task_struct * p)1605 static inline int pid_alive(const struct task_struct *p)
1606 {
1607 return p->thread_pid != NULL;
1608 }
1609
task_pgrp_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1610 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1611 {
1612 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1613 }
1614
task_pgrp_vnr(struct task_struct * tsk)1615 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1616 {
1617 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1618 }
1619
1620
task_session_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1621 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1622 {
1623 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1624 }
1625
task_session_vnr(struct task_struct * tsk)1626 static inline pid_t task_session_vnr(struct task_struct *tsk)
1627 {
1628 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1629 }
1630
task_tgid_nr_ns(struct task_struct * tsk,struct pid_namespace * ns)1631 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1632 {
1633 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1634 }
1635
task_tgid_vnr(struct task_struct * tsk)1636 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1637 {
1638 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1639 }
1640
task_ppid_nr_ns(const struct task_struct * tsk,struct pid_namespace * ns)1641 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1642 {
1643 pid_t pid = 0;
1644
1645 rcu_read_lock();
1646 if (pid_alive(tsk))
1647 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1648 rcu_read_unlock();
1649
1650 return pid;
1651 }
1652
task_ppid_nr(const struct task_struct * tsk)1653 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1654 {
1655 return task_ppid_nr_ns(tsk, &init_pid_ns);
1656 }
1657
1658 /* Obsolete, do not use: */
task_pgrp_nr(struct task_struct * tsk)1659 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1660 {
1661 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1662 }
1663
1664 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1665 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1666
__task_state_index(unsigned int tsk_state,unsigned int tsk_exit_state)1667 static 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 */
1682 if (tsk_state & TASK_RTLOCK_WAIT)
1683 state = TASK_UNINTERRUPTIBLE;
1684
1685 return fls(state);
1686 }
1687
task_state_index(struct task_struct * tsk)1688 static inline unsigned int task_state_index(struct task_struct *tsk)
1689 {
1690 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1691 }
1692
task_index_to_char(unsigned int state)1693 static inline char task_index_to_char(unsigned int state)
1694 {
1695 static const char state_char[] = "RSDTtXZPI";
1696
1697 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1698
1699 return state_char[state];
1700 }
1701
task_state_to_char(struct task_struct * tsk)1702 static inline char task_state_to_char(struct task_struct *tsk)
1703 {
1704 return task_index_to_char(task_state_index(tsk));
1705 }
1706
1707 /**
1708 * is_global_init - check if a task structure is init. Since init
1709 * is free to have sub-threads we need to check tgid.
1710 * @tsk: Task structure to be checked.
1711 *
1712 * Check if a task structure is the first user space task the kernel created.
1713 *
1714 * Return: 1 if the task structure is init. 0 otherwise.
1715 */
is_global_init(struct task_struct * tsk)1716 static inline int is_global_init(struct task_struct *tsk)
1717 {
1718 return task_tgid_nr(tsk) == 1;
1719 }
1720
1721 extern struct pid *cad_pid;
1722
1723 /*
1724 * Per process flags
1725 */
1726 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1727 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1728 #define PF_EXITING 0x00000004 /* Getting shut down */
1729 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1730 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1731 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1732 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1733 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1734 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1735 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1736 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1737 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1738 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1739 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1740 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1741 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1742 #define PF__HOLE__00010000 0x00010000
1743 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1744 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1745 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1746 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1747 * I am cleaning dirty pages from some other bdi. */
1748 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1749 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1750 #define PF__HOLE__00800000 0x00800000
1751 #define PF__HOLE__01000000 0x01000000
1752 #define PF__HOLE__02000000 0x02000000
1753 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1754 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1755 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1756 #define PF__HOLE__20000000 0x20000000
1757 #define PF__HOLE__40000000 0x40000000
1758 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1759
1760 /*
1761 * Only the _current_ task can read/write to tsk->flags, but other
1762 * tasks can access tsk->flags in readonly mode for example
1763 * with tsk_used_math (like during threaded core dumping).
1764 * There is however an exception to this rule during ptrace
1765 * or during fork: the ptracer task is allowed to write to the
1766 * child->flags of its traced child (same goes for fork, the parent
1767 * can write to the child->flags), because we're guaranteed the
1768 * child is not running and in turn not changing child->flags
1769 * at the same time the parent does it.
1770 */
1771 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1772 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1773 #define clear_used_math() clear_stopped_child_used_math(current)
1774 #define set_used_math() set_stopped_child_used_math(current)
1775
1776 #define conditional_stopped_child_used_math(condition, child) \
1777 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1778
1779 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1780
1781 #define copy_to_stopped_child_used_math(child) \
1782 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1783
1784 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1785 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1786 #define used_math() tsk_used_math(current)
1787
is_percpu_thread(void)1788 static __always_inline bool is_percpu_thread(void)
1789 {
1790 #ifdef CONFIG_SMP
1791 return (current->flags & PF_NO_SETAFFINITY) &&
1792 (current->nr_cpus_allowed == 1);
1793 #else
1794 return true;
1795 #endif
1796 }
1797
1798 /* Per-process atomic flags. */
1799 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1800 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1801 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1802 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1803 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1804 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1805 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1806 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1807
1808 #define TASK_PFA_TEST(name, func) \
1809 static inline bool task_##func(struct task_struct *p) \
1810 { return test_bit(PFA_##name, &p->atomic_flags); }
1811
1812 #define TASK_PFA_SET(name, func) \
1813 static inline void task_set_##func(struct task_struct *p) \
1814 { set_bit(PFA_##name, &p->atomic_flags); }
1815
1816 #define TASK_PFA_CLEAR(name, func) \
1817 static inline void task_clear_##func(struct task_struct *p) \
1818 { clear_bit(PFA_##name, &p->atomic_flags); }
1819
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1820 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1821 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1822
1823 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1824 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1825 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1826
1827 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1828 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1829 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1830
1831 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1832 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1833 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1834
1835 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1836 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1837 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1838
1839 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1840 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1841
1842 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1843 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1844 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1845
1846 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1847 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1848
1849 static inline void
1850 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1851 {
1852 current->flags &= ~flags;
1853 current->flags |= orig_flags & flags;
1854 }
1855
1856 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1857 extern int task_can_attach(struct task_struct *p);
1858 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1859 extern void dl_bw_free(int cpu, u64 dl_bw);
1860 #ifdef CONFIG_SMP
1861
1862 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1863 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1864
1865 /**
1866 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1867 * @p: the task
1868 * @new_mask: CPU affinity mask
1869 *
1870 * Return: zero if successful, or a negative error code
1871 */
1872 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1873 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1874 extern void release_user_cpus_ptr(struct task_struct *p);
1875 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1876 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1877 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1878 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1879 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1880 {
1881 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1882 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1883 {
1884 if (!cpumask_test_cpu(0, new_mask))
1885 return -EINVAL;
1886 return 0;
1887 }
dup_user_cpus_ptr(struct task_struct * dst,struct task_struct * src,int node)1888 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1889 {
1890 if (src->user_cpus_ptr)
1891 return -EINVAL;
1892 return 0;
1893 }
release_user_cpus_ptr(struct task_struct * p)1894 static inline void release_user_cpus_ptr(struct task_struct *p)
1895 {
1896 WARN_ON(p->user_cpus_ptr);
1897 }
1898
dl_task_check_affinity(struct task_struct * p,const struct cpumask * mask)1899 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1900 {
1901 return 0;
1902 }
1903 #endif
1904
1905 extern int yield_to(struct task_struct *p, bool preempt);
1906 extern void set_user_nice(struct task_struct *p, long nice);
1907 extern int task_prio(const struct task_struct *p);
1908
1909 /**
1910 * task_nice - return the nice value of a given task.
1911 * @p: the task in question.
1912 *
1913 * Return: The nice value [ -20 ... 0 ... 19 ].
1914 */
task_nice(const struct task_struct * p)1915 static inline int task_nice(const struct task_struct *p)
1916 {
1917 return PRIO_TO_NICE((p)->static_prio);
1918 }
1919
1920 extern int can_nice(const struct task_struct *p, const int nice);
1921 extern int task_curr(const struct task_struct *p);
1922 extern int idle_cpu(int cpu);
1923 extern int available_idle_cpu(int cpu);
1924 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1925 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1926 extern void sched_set_fifo(struct task_struct *p);
1927 extern void sched_set_fifo_low(struct task_struct *p);
1928 extern void sched_set_normal(struct task_struct *p, int nice);
1929 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1930 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1931 extern struct task_struct *idle_task(int cpu);
1932
1933 /**
1934 * is_idle_task - is the specified task an idle task?
1935 * @p: the task in question.
1936 *
1937 * Return: 1 if @p is an idle task. 0 otherwise.
1938 */
is_idle_task(const struct task_struct * p)1939 static __always_inline bool is_idle_task(const struct task_struct *p)
1940 {
1941 return !!(p->flags & PF_IDLE);
1942 }
1943
1944 extern struct task_struct *curr_task(int cpu);
1945 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1946
1947 void yield(void);
1948
1949 union thread_union {
1950 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1951 struct task_struct task;
1952 #endif
1953 #ifndef CONFIG_THREAD_INFO_IN_TASK
1954 struct thread_info thread_info;
1955 #endif
1956 unsigned long stack[THREAD_SIZE/sizeof(long)];
1957 };
1958
1959 #ifndef CONFIG_THREAD_INFO_IN_TASK
1960 extern struct thread_info init_thread_info;
1961 #endif
1962
1963 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1964
1965 #ifdef CONFIG_THREAD_INFO_IN_TASK
1966 # define task_thread_info(task) (&(task)->thread_info)
1967 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1968 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1969 #endif
1970
1971 /*
1972 * find a task by one of its numerical ids
1973 *
1974 * find_task_by_pid_ns():
1975 * finds a task by its pid in the specified namespace
1976 * find_task_by_vpid():
1977 * finds a task by its virtual pid
1978 *
1979 * see also find_vpid() etc in include/linux/pid.h
1980 */
1981
1982 extern struct task_struct *find_task_by_vpid(pid_t nr);
1983 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1984
1985 /*
1986 * find a task by its virtual pid and get the task struct
1987 */
1988 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1989
1990 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1991 extern int wake_up_process(struct task_struct *tsk);
1992 extern void wake_up_new_task(struct task_struct *tsk);
1993
1994 #ifdef CONFIG_SMP
1995 extern void kick_process(struct task_struct *tsk);
1996 #else
kick_process(struct task_struct * tsk)1997 static inline void kick_process(struct task_struct *tsk) { }
1998 #endif
1999
2000 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2001
set_task_comm(struct task_struct * tsk,const char * from)2002 static inline void set_task_comm(struct task_struct *tsk, const char *from)
2003 {
2004 __set_task_comm(tsk, from, false);
2005 }
2006
2007 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
2008 #define get_task_comm(buf, tsk) ({ \
2009 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
2010 __get_task_comm(buf, sizeof(buf), tsk); \
2011 })
2012
2013 #ifdef CONFIG_SMP
scheduler_ipi(void)2014 static __always_inline void scheduler_ipi(void)
2015 {
2016 /*
2017 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2018 * TIF_NEED_RESCHED remotely (for the first time) will also send
2019 * this IPI.
2020 */
2021 preempt_fold_need_resched();
2022 }
2023 #else
scheduler_ipi(void)2024 static inline void scheduler_ipi(void) { }
2025 #endif
2026
2027 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2028
2029 /*
2030 * Set thread flags in other task's structures.
2031 * See asm/thread_info.h for TIF_xxxx flags available:
2032 */
set_tsk_thread_flag(struct task_struct * tsk,int flag)2033 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2034 {
2035 set_ti_thread_flag(task_thread_info(tsk), flag);
2036 }
2037
clear_tsk_thread_flag(struct task_struct * tsk,int flag)2038 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2039 {
2040 clear_ti_thread_flag(task_thread_info(tsk), flag);
2041 }
2042
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)2043 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2044 bool value)
2045 {
2046 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2047 }
2048
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)2049 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2050 {
2051 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2052 }
2053
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)2054 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2055 {
2056 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2057 }
2058
test_tsk_thread_flag(struct task_struct * tsk,int flag)2059 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2060 {
2061 return test_ti_thread_flag(task_thread_info(tsk), flag);
2062 }
2063
set_tsk_need_resched(struct task_struct * tsk)2064 static inline void set_tsk_need_resched(struct task_struct *tsk)
2065 {
2066 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2067 }
2068
clear_tsk_need_resched(struct task_struct * tsk)2069 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2070 {
2071 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2072 }
2073
test_tsk_need_resched(struct task_struct * tsk)2074 static inline int test_tsk_need_resched(struct task_struct *tsk)
2075 {
2076 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2077 }
2078
2079 /*
2080 * cond_resched() and cond_resched_lock(): latency reduction via
2081 * explicit rescheduling in places that are safe. The return
2082 * value indicates whether a reschedule was done in fact.
2083 * cond_resched_lock() will drop the spinlock before scheduling,
2084 */
2085 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2086 extern int __cond_resched(void);
2087
2088 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2089
2090 void sched_dynamic_klp_enable(void);
2091 void sched_dynamic_klp_disable(void);
2092
2093 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2094
_cond_resched(void)2095 static __always_inline int _cond_resched(void)
2096 {
2097 return static_call_mod(cond_resched)();
2098 }
2099
2100 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2101
2102 extern int dynamic_cond_resched(void);
2103
_cond_resched(void)2104 static __always_inline int _cond_resched(void)
2105 {
2106 return dynamic_cond_resched();
2107 }
2108
2109 #else /* !CONFIG_PREEMPTION */
2110
_cond_resched(void)2111 static inline int _cond_resched(void)
2112 {
2113 klp_sched_try_switch();
2114 return __cond_resched();
2115 }
2116
2117 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2118
2119 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2120
_cond_resched(void)2121 static inline int _cond_resched(void)
2122 {
2123 klp_sched_try_switch();
2124 return 0;
2125 }
2126
2127 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2128
2129 #define cond_resched() ({ \
2130 __might_resched(__FILE__, __LINE__, 0); \
2131 _cond_resched(); \
2132 })
2133
2134 extern int __cond_resched_lock(spinlock_t *lock);
2135 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2136 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2137
2138 #define MIGHT_RESCHED_RCU_SHIFT 8
2139 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2140
2141 #ifndef CONFIG_PREEMPT_RT
2142 /*
2143 * Non RT kernels have an elevated preempt count due to the held lock,
2144 * but are not allowed to be inside a RCU read side critical section
2145 */
2146 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2147 #else
2148 /*
2149 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2150 * cond_resched*lock() has to take that into account because it checks for
2151 * preempt_count() and rcu_preempt_depth().
2152 */
2153 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2154 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2155 #endif
2156
2157 #define cond_resched_lock(lock) ({ \
2158 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2159 __cond_resched_lock(lock); \
2160 })
2161
2162 #define cond_resched_rwlock_read(lock) ({ \
2163 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2164 __cond_resched_rwlock_read(lock); \
2165 })
2166
2167 #define cond_resched_rwlock_write(lock) ({ \
2168 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2169 __cond_resched_rwlock_write(lock); \
2170 })
2171
cond_resched_rcu(void)2172 static inline void cond_resched_rcu(void)
2173 {
2174 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2175 rcu_read_unlock();
2176 cond_resched();
2177 rcu_read_lock();
2178 #endif
2179 }
2180
2181 #ifdef CONFIG_PREEMPT_DYNAMIC
2182
2183 extern bool preempt_model_none(void);
2184 extern bool preempt_model_voluntary(void);
2185 extern bool preempt_model_full(void);
2186
2187 #else
2188
preempt_model_none(void)2189 static inline bool preempt_model_none(void)
2190 {
2191 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2192 }
preempt_model_voluntary(void)2193 static inline bool preempt_model_voluntary(void)
2194 {
2195 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2196 }
preempt_model_full(void)2197 static inline bool preempt_model_full(void)
2198 {
2199 return IS_ENABLED(CONFIG_PREEMPT);
2200 }
2201
2202 #endif
2203
preempt_model_rt(void)2204 static inline bool preempt_model_rt(void)
2205 {
2206 return IS_ENABLED(CONFIG_PREEMPT_RT);
2207 }
2208
2209 /*
2210 * Does the preemption model allow non-cooperative preemption?
2211 *
2212 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2213 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2214 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2215 * PREEMPT_NONE model.
2216 */
preempt_model_preemptible(void)2217 static inline bool preempt_model_preemptible(void)
2218 {
2219 return preempt_model_full() || preempt_model_rt();
2220 }
2221
2222 /*
2223 * Does a critical section need to be broken due to another
2224 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2225 * but a general need for low latency)
2226 */
spin_needbreak(spinlock_t * lock)2227 static inline int spin_needbreak(spinlock_t *lock)
2228 {
2229 #ifdef CONFIG_PREEMPTION
2230 return spin_is_contended(lock);
2231 #else
2232 return 0;
2233 #endif
2234 }
2235
2236 /*
2237 * Check if a rwlock is contended.
2238 * Returns non-zero if there is another task waiting on the rwlock.
2239 * Returns zero if the lock is not contended or the system / underlying
2240 * rwlock implementation does not support contention detection.
2241 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2242 * for low latency.
2243 */
rwlock_needbreak(rwlock_t * lock)2244 static inline int rwlock_needbreak(rwlock_t *lock)
2245 {
2246 #ifdef CONFIG_PREEMPTION
2247 return rwlock_is_contended(lock);
2248 #else
2249 return 0;
2250 #endif
2251 }
2252
need_resched(void)2253 static __always_inline bool need_resched(void)
2254 {
2255 return unlikely(tif_need_resched());
2256 }
2257
2258 /*
2259 * Wrappers for p->thread_info->cpu access. No-op on UP.
2260 */
2261 #ifdef CONFIG_SMP
2262
task_cpu(const struct task_struct * p)2263 static inline unsigned int task_cpu(const struct task_struct *p)
2264 {
2265 return READ_ONCE(task_thread_info(p)->cpu);
2266 }
2267
2268 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2269
2270 #else
2271
task_cpu(const struct task_struct * p)2272 static inline unsigned int task_cpu(const struct task_struct *p)
2273 {
2274 return 0;
2275 }
2276
set_task_cpu(struct task_struct * p,unsigned int cpu)2277 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2278 {
2279 }
2280
2281 #endif /* CONFIG_SMP */
2282
2283 extern bool sched_task_on_rq(struct task_struct *p);
2284 extern unsigned long get_wchan(struct task_struct *p);
2285 extern struct task_struct *cpu_curr_snapshot(int cpu);
2286
2287 /*
2288 * In order to reduce various lock holder preemption latencies provide an
2289 * interface to see if a vCPU is currently running or not.
2290 *
2291 * This allows us to terminate optimistic spin loops and block, analogous to
2292 * the native optimistic spin heuristic of testing if the lock owner task is
2293 * running or not.
2294 */
2295 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2296 static inline bool vcpu_is_preempted(int cpu)
2297 {
2298 return false;
2299 }
2300 #endif
2301
2302 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2303 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2304
2305 #ifndef TASK_SIZE_OF
2306 #define TASK_SIZE_OF(tsk) TASK_SIZE
2307 #endif
2308
2309 #ifdef CONFIG_SMP
owner_on_cpu(struct task_struct * owner)2310 static inline bool owner_on_cpu(struct task_struct *owner)
2311 {
2312 /*
2313 * As lock holder preemption issue, we both skip spinning if
2314 * task is not on cpu or its cpu is preempted
2315 */
2316 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2317 }
2318
2319 /* Returns effective CPU energy utilization, as seen by the scheduler */
2320 unsigned long sched_cpu_util(int cpu);
2321 #endif /* CONFIG_SMP */
2322
2323 #ifdef CONFIG_RSEQ
2324
2325 /*
2326 * Map the event mask on the user-space ABI enum rseq_cs_flags
2327 * for direct mask checks.
2328 */
2329 enum rseq_event_mask_bits {
2330 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2331 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2332 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2333 };
2334
2335 enum rseq_event_mask {
2336 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2337 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2338 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2339 };
2340
rseq_set_notify_resume(struct task_struct * t)2341 static inline void rseq_set_notify_resume(struct task_struct *t)
2342 {
2343 if (t->rseq)
2344 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2345 }
2346
2347 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2348
rseq_handle_notify_resume(struct ksignal * ksig,struct pt_regs * regs)2349 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2350 struct pt_regs *regs)
2351 {
2352 if (current->rseq)
2353 __rseq_handle_notify_resume(ksig, regs);
2354 }
2355
rseq_signal_deliver(struct ksignal * ksig,struct pt_regs * regs)2356 static inline void rseq_signal_deliver(struct ksignal *ksig,
2357 struct pt_regs *regs)
2358 {
2359 preempt_disable();
2360 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2361 preempt_enable();
2362 rseq_handle_notify_resume(ksig, regs);
2363 }
2364
2365 /* rseq_preempt() requires preemption to be disabled. */
rseq_preempt(struct task_struct * t)2366 static inline void rseq_preempt(struct task_struct *t)
2367 {
2368 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2369 rseq_set_notify_resume(t);
2370 }
2371
2372 /* rseq_migrate() requires preemption to be disabled. */
rseq_migrate(struct task_struct * t)2373 static inline void rseq_migrate(struct task_struct *t)
2374 {
2375 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2376 rseq_set_notify_resume(t);
2377 }
2378
2379 /*
2380 * If parent process has a registered restartable sequences area, the
2381 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2382 */
rseq_fork(struct task_struct * t,unsigned long clone_flags)2383 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2384 {
2385 if (clone_flags & CLONE_VM) {
2386 t->rseq = NULL;
2387 t->rseq_len = 0;
2388 t->rseq_sig = 0;
2389 t->rseq_event_mask = 0;
2390 } else {
2391 t->rseq = current->rseq;
2392 t->rseq_len = current->rseq_len;
2393 t->rseq_sig = current->rseq_sig;
2394 t->rseq_event_mask = current->rseq_event_mask;
2395 }
2396 }
2397
rseq_execve(struct task_struct * t)2398 static inline void rseq_execve(struct task_struct *t)
2399 {
2400 t->rseq = NULL;
2401 t->rseq_len = 0;
2402 t->rseq_sig = 0;
2403 t->rseq_event_mask = 0;
2404 }
2405
2406 #else
2407
rseq_set_notify_resume(struct task_struct * t)2408 static inline void rseq_set_notify_resume(struct task_struct *t)
2409 {
2410 }
rseq_handle_notify_resume(struct ksignal * ksig,struct pt_regs * regs)2411 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2412 struct pt_regs *regs)
2413 {
2414 }
rseq_signal_deliver(struct ksignal * ksig,struct pt_regs * regs)2415 static inline void rseq_signal_deliver(struct ksignal *ksig,
2416 struct pt_regs *regs)
2417 {
2418 }
rseq_preempt(struct task_struct * t)2419 static inline void rseq_preempt(struct task_struct *t)
2420 {
2421 }
rseq_migrate(struct task_struct * t)2422 static inline void rseq_migrate(struct task_struct *t)
2423 {
2424 }
rseq_fork(struct task_struct * t,unsigned long clone_flags)2425 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2426 {
2427 }
rseq_execve(struct task_struct * t)2428 static inline void rseq_execve(struct task_struct *t)
2429 {
2430 }
2431
2432 #endif
2433
2434 #ifdef CONFIG_DEBUG_RSEQ
2435
2436 void rseq_syscall(struct pt_regs *regs);
2437
2438 #else
2439
rseq_syscall(struct pt_regs * regs)2440 static inline void rseq_syscall(struct pt_regs *regs)
2441 {
2442 }
2443
2444 #endif
2445
2446 #ifdef CONFIG_SCHED_CORE
2447 extern void sched_core_free(struct task_struct *tsk);
2448 extern void sched_core_fork(struct task_struct *p);
2449 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2450 unsigned long uaddr);
2451 extern int sched_core_idle_cpu(int cpu);
2452 #else
sched_core_free(struct task_struct * tsk)2453 static inline void sched_core_free(struct task_struct *tsk) { }
sched_core_fork(struct task_struct * p)2454 static inline void sched_core_fork(struct task_struct *p) { }
sched_core_idle_cpu(int cpu)2455 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2456 #endif
2457
2458 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2459
2460 #endif
2461