1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4
5 /*
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
8 */
9
10 #include <uapi/linux/sched.h>
11
12 #include <asm/current.h>
13 #include <asm/processor.h>
14 #include <linux/thread_info.h>
15 #include <linux/preempt.h>
16 #include <linux/cpumask_types.h>
17
18 #include <linux/cache.h>
19 #include <linux/irqflags_types.h>
20 #include <linux/smp_types.h>
21 #include <linux/pid_types.h>
22 #include <linux/sem_types.h>
23 #include <linux/shm.h>
24 #include <linux/kmsan_types.h>
25 #include <linux/mutex_types.h>
26 #include <linux/plist_types.h>
27 #include <linux/hrtimer_types.h>
28 #include <linux/timer_types.h>
29 #include <linux/seccomp_types.h>
30 #include <linux/nodemask_types.h>
31 #include <linux/refcount_types.h>
32 #include <linux/resource.h>
33 #include <linux/latencytop.h>
34 #include <linux/sched/prio.h>
35 #include <linux/sched/types.h>
36 #include <linux/signal_types.h>
37 #include <linux/syscall_user_dispatch_types.h>
38 #include <linux/mm_types_task.h>
39 #include <linux/netdevice_xmit.h>
40 #include <linux/task_io_accounting.h>
41 #include <linux/posix-timers_types.h>
42 #include <linux/restart_block.h>
43 #include <uapi/linux/rseq.h>
44 #include <linux/seqlock_types.h>
45 #include <linux/kcsan.h>
46 #include <linux/rv.h>
47 #include <linux/uidgid_types.h>
48 #include <linux/tracepoint-defs.h>
49 #include <asm/kmap_size.h>
50
51 /* task_struct member predeclarations (sorted alphabetically): */
52 struct audit_context;
53 struct bio_list;
54 struct blk_plug;
55 struct bpf_local_storage;
56 struct bpf_run_ctx;
57 struct bpf_net_context;
58 struct capture_control;
59 struct cfs_rq;
60 struct fs_struct;
61 struct futex_pi_state;
62 struct io_context;
63 struct io_uring_task;
64 struct mempolicy;
65 struct nameidata;
66 struct nsproxy;
67 struct perf_event_context;
68 struct perf_ctx_data;
69 struct pid_namespace;
70 struct pipe_inode_info;
71 struct rcu_node;
72 struct reclaim_state;
73 struct robust_list_head;
74 struct root_domain;
75 struct rq;
76 struct sched_attr;
77 struct sched_dl_entity;
78 struct seq_file;
79 struct sighand_struct;
80 struct signal_struct;
81 struct task_delay_info;
82 struct task_group;
83 struct task_struct;
84 struct user_event_mm;
85
86 #include <linux/sched/ext.h>
87
88 /*
89 * Task state bitmask. NOTE! These bits are also
90 * encoded in fs/proc/array.c: get_task_state().
91 *
92 * We have two separate sets of flags: task->__state
93 * is about runnability, while task->exit_state are
94 * about the task exiting. Confusing, but this way
95 * modifying one set can't modify the other one by
96 * mistake.
97 */
98
99 /* Used in tsk->__state: */
100 #define TASK_RUNNING 0x00000000
101 #define TASK_INTERRUPTIBLE 0x00000001
102 #define TASK_UNINTERRUPTIBLE 0x00000002
103 #define __TASK_STOPPED 0x00000004
104 #define __TASK_TRACED 0x00000008
105 /* Used in tsk->exit_state: */
106 #define EXIT_DEAD 0x00000010
107 #define EXIT_ZOMBIE 0x00000020
108 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
109 /* Used in tsk->__state again: */
110 #define TASK_PARKED 0x00000040
111 #define TASK_DEAD 0x00000080
112 #define TASK_WAKEKILL 0x00000100
113 #define TASK_WAKING 0x00000200
114 #define TASK_NOLOAD 0x00000400
115 #define TASK_NEW 0x00000800
116 #define TASK_RTLOCK_WAIT 0x00001000
117 #define TASK_FREEZABLE 0x00002000
118 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
119 #define TASK_FROZEN 0x00008000
120 #define TASK_STATE_MAX 0x00010000
121
122 #define TASK_ANY (TASK_STATE_MAX-1)
123
124 /*
125 * DO NOT ADD ANY NEW USERS !
126 */
127 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
128
129 /* Convenience macros for the sake of set_current_state: */
130 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
131 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
132 #define TASK_TRACED __TASK_TRACED
133
134 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
135
136 /* Convenience macros for the sake of wake_up(): */
137 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
138
139 /* get_task_state(): */
140 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
141 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
142 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
143 TASK_PARKED)
144
145 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
146
147 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
148 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
149 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
150
151 /*
152 * Special states are those that do not use the normal wait-loop pattern. See
153 * the comment with set_special_state().
154 */
155 #define is_special_task_state(state) \
156 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \
157 TASK_DEAD | TASK_FROZEN))
158
159 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
160 # define debug_normal_state_change(state_value) \
161 do { \
162 WARN_ON_ONCE(is_special_task_state(state_value)); \
163 current->task_state_change = _THIS_IP_; \
164 } while (0)
165
166 # define debug_special_state_change(state_value) \
167 do { \
168 WARN_ON_ONCE(!is_special_task_state(state_value)); \
169 current->task_state_change = _THIS_IP_; \
170 } while (0)
171
172 # define debug_rtlock_wait_set_state() \
173 do { \
174 current->saved_state_change = current->task_state_change;\
175 current->task_state_change = _THIS_IP_; \
176 } while (0)
177
178 # define debug_rtlock_wait_restore_state() \
179 do { \
180 current->task_state_change = current->saved_state_change;\
181 } while (0)
182
183 #else
184 # define debug_normal_state_change(cond) do { } while (0)
185 # define debug_special_state_change(cond) do { } while (0)
186 # define debug_rtlock_wait_set_state() do { } while (0)
187 # define debug_rtlock_wait_restore_state() do { } while (0)
188 #endif
189
190 #define trace_set_current_state(state_value) \
191 do { \
192 if (tracepoint_enabled(sched_set_state_tp)) \
193 __trace_set_current_state(state_value); \
194 } while (0)
195
196 /*
197 * set_current_state() includes a barrier so that the write of current->__state
198 * is correctly serialised wrt the caller's subsequent test of whether to
199 * actually sleep:
200 *
201 * for (;;) {
202 * set_current_state(TASK_UNINTERRUPTIBLE);
203 * if (CONDITION)
204 * break;
205 *
206 * schedule();
207 * }
208 * __set_current_state(TASK_RUNNING);
209 *
210 * If the caller does not need such serialisation (because, for instance, the
211 * CONDITION test and condition change and wakeup are under the same lock) then
212 * use __set_current_state().
213 *
214 * The above is typically ordered against the wakeup, which does:
215 *
216 * CONDITION = 1;
217 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
218 *
219 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
220 * accessing p->__state.
221 *
222 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
223 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
224 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
225 *
226 * However, with slightly different timing the wakeup TASK_RUNNING store can
227 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
228 * a problem either because that will result in one extra go around the loop
229 * and our @cond test will save the day.
230 *
231 * Also see the comments of try_to_wake_up().
232 */
233 #define __set_current_state(state_value) \
234 do { \
235 debug_normal_state_change((state_value)); \
236 trace_set_current_state(state_value); \
237 WRITE_ONCE(current->__state, (state_value)); \
238 } while (0)
239
240 #define set_current_state(state_value) \
241 do { \
242 debug_normal_state_change((state_value)); \
243 trace_set_current_state(state_value); \
244 smp_store_mb(current->__state, (state_value)); \
245 } while (0)
246
247 /*
248 * set_special_state() should be used for those states when the blocking task
249 * can not use the regular condition based wait-loop. In that case we must
250 * serialize against wakeups such that any possible in-flight TASK_RUNNING
251 * stores will not collide with our state change.
252 */
253 #define set_special_state(state_value) \
254 do { \
255 unsigned long flags; /* may shadow */ \
256 \
257 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
258 debug_special_state_change((state_value)); \
259 trace_set_current_state(state_value); \
260 WRITE_ONCE(current->__state, (state_value)); \
261 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
262 } while (0)
263
264 /*
265 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
266 *
267 * RT's spin/rwlock substitutions are state preserving. The state of the
268 * task when blocking on the lock is saved in task_struct::saved_state and
269 * restored after the lock has been acquired. These operations are
270 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
271 * lock related wakeups while the task is blocked on the lock are
272 * redirected to operate on task_struct::saved_state to ensure that these
273 * are not dropped. On restore task_struct::saved_state is set to
274 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
275 *
276 * The lock operation looks like this:
277 *
278 * current_save_and_set_rtlock_wait_state();
279 * for (;;) {
280 * if (try_lock())
281 * break;
282 * raw_spin_unlock_irq(&lock->wait_lock);
283 * schedule_rtlock();
284 * raw_spin_lock_irq(&lock->wait_lock);
285 * set_current_state(TASK_RTLOCK_WAIT);
286 * }
287 * current_restore_rtlock_saved_state();
288 */
289 #define current_save_and_set_rtlock_wait_state() \
290 do { \
291 lockdep_assert_irqs_disabled(); \
292 raw_spin_lock(¤t->pi_lock); \
293 current->saved_state = current->__state; \
294 debug_rtlock_wait_set_state(); \
295 trace_set_current_state(TASK_RTLOCK_WAIT); \
296 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
297 raw_spin_unlock(¤t->pi_lock); \
298 } while (0);
299
300 #define current_restore_rtlock_saved_state() \
301 do { \
302 lockdep_assert_irqs_disabled(); \
303 raw_spin_lock(¤t->pi_lock); \
304 debug_rtlock_wait_restore_state(); \
305 trace_set_current_state(current->saved_state); \
306 WRITE_ONCE(current->__state, current->saved_state); \
307 current->saved_state = TASK_RUNNING; \
308 raw_spin_unlock(¤t->pi_lock); \
309 } while (0);
310
311 #define get_current_state() READ_ONCE(current->__state)
312
313 /*
314 * Define the task command name length as enum, then it can be visible to
315 * BPF programs.
316 */
317 enum {
318 TASK_COMM_LEN = 16,
319 };
320
321 extern void sched_tick(void);
322
323 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
324
325 extern long schedule_timeout(long timeout);
326 extern long schedule_timeout_interruptible(long timeout);
327 extern long schedule_timeout_killable(long timeout);
328 extern long schedule_timeout_uninterruptible(long timeout);
329 extern long schedule_timeout_idle(long timeout);
330 asmlinkage void schedule(void);
331 extern void schedule_preempt_disabled(void);
332 asmlinkage void preempt_schedule_irq(void);
333 #ifdef CONFIG_PREEMPT_RT
334 extern void schedule_rtlock(void);
335 #endif
336
337 extern int __must_check io_schedule_prepare(void);
338 extern void io_schedule_finish(int token);
339 extern long io_schedule_timeout(long timeout);
340 extern void io_schedule(void);
341
342 /* wrapper function to trace from this header file */
343 DECLARE_TRACEPOINT(sched_set_state_tp);
344 extern void __trace_set_current_state(int state_value);
345
346 /**
347 * struct prev_cputime - snapshot of system and user cputime
348 * @utime: time spent in user mode
349 * @stime: time spent in system mode
350 * @lock: protects the above two fields
351 *
352 * Stores previous user/system time values such that we can guarantee
353 * monotonicity.
354 */
355 struct prev_cputime {
356 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
357 u64 utime;
358 u64 stime;
359 raw_spinlock_t lock;
360 #endif
361 };
362
363 enum vtime_state {
364 /* Task is sleeping or running in a CPU with VTIME inactive: */
365 VTIME_INACTIVE = 0,
366 /* Task is idle */
367 VTIME_IDLE,
368 /* Task runs in kernelspace in a CPU with VTIME active: */
369 VTIME_SYS,
370 /* Task runs in userspace in a CPU with VTIME active: */
371 VTIME_USER,
372 /* Task runs as guests in a CPU with VTIME active: */
373 VTIME_GUEST,
374 };
375
376 struct vtime {
377 seqcount_t seqcount;
378 unsigned long long starttime;
379 enum vtime_state state;
380 unsigned int cpu;
381 u64 utime;
382 u64 stime;
383 u64 gtime;
384 };
385
386 /*
387 * Utilization clamp constraints.
388 * @UCLAMP_MIN: Minimum utilization
389 * @UCLAMP_MAX: Maximum utilization
390 * @UCLAMP_CNT: Utilization clamp constraints count
391 */
392 enum uclamp_id {
393 UCLAMP_MIN = 0,
394 UCLAMP_MAX,
395 UCLAMP_CNT
396 };
397
398 #ifdef CONFIG_SMP
399 extern struct root_domain def_root_domain;
400 extern struct mutex sched_domains_mutex;
401 extern void sched_domains_mutex_lock(void);
402 extern void sched_domains_mutex_unlock(void);
403 #else
sched_domains_mutex_lock(void)404 static inline void sched_domains_mutex_lock(void) { }
sched_domains_mutex_unlock(void)405 static inline void sched_domains_mutex_unlock(void) { }
406 #endif
407
408 struct sched_param {
409 int sched_priority;
410 };
411
412 struct sched_info {
413 #ifdef CONFIG_SCHED_INFO
414 /* Cumulative counters: */
415
416 /* # of times we have run on this CPU: */
417 unsigned long pcount;
418
419 /* Time spent waiting on a runqueue: */
420 unsigned long long run_delay;
421
422 /* Max time spent waiting on a runqueue: */
423 unsigned long long max_run_delay;
424
425 /* Min time spent waiting on a runqueue: */
426 unsigned long long min_run_delay;
427
428 /* Timestamps: */
429
430 /* When did we last run on a CPU? */
431 unsigned long long last_arrival;
432
433 /* When were we last queued to run? */
434 unsigned long long last_queued;
435
436 #endif /* CONFIG_SCHED_INFO */
437 };
438
439 /*
440 * Integer metrics need fixed point arithmetic, e.g., sched/fair
441 * has a few: load, load_avg, util_avg, freq, and capacity.
442 *
443 * We define a basic fixed point arithmetic range, and then formalize
444 * all these metrics based on that basic range.
445 */
446 # define SCHED_FIXEDPOINT_SHIFT 10
447 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
448
449 /* Increase resolution of cpu_capacity calculations */
450 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
451 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
452
453 struct load_weight {
454 unsigned long weight;
455 u32 inv_weight;
456 };
457
458 /*
459 * The load/runnable/util_avg accumulates an infinite geometric series
460 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
461 *
462 * [load_avg definition]
463 *
464 * load_avg = runnable% * scale_load_down(load)
465 *
466 * [runnable_avg definition]
467 *
468 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
469 *
470 * [util_avg definition]
471 *
472 * util_avg = running% * SCHED_CAPACITY_SCALE
473 *
474 * where runnable% is the time ratio that a sched_entity is runnable and
475 * running% the time ratio that a sched_entity is running.
476 *
477 * For cfs_rq, they are the aggregated values of all runnable and blocked
478 * sched_entities.
479 *
480 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
481 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
482 * for computing those signals (see update_rq_clock_pelt())
483 *
484 * N.B., the above ratios (runnable% and running%) themselves are in the
485 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
486 * to as large a range as necessary. This is for example reflected by
487 * util_avg's SCHED_CAPACITY_SCALE.
488 *
489 * [Overflow issue]
490 *
491 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
492 * with the highest load (=88761), always runnable on a single cfs_rq,
493 * and should not overflow as the number already hits PID_MAX_LIMIT.
494 *
495 * For all other cases (including 32-bit kernels), struct load_weight's
496 * weight will overflow first before we do, because:
497 *
498 * Max(load_avg) <= Max(load.weight)
499 *
500 * Then it is the load_weight's responsibility to consider overflow
501 * issues.
502 */
503 struct sched_avg {
504 u64 last_update_time;
505 u64 load_sum;
506 u64 runnable_sum;
507 u32 util_sum;
508 u32 period_contrib;
509 unsigned long load_avg;
510 unsigned long runnable_avg;
511 unsigned long util_avg;
512 unsigned int util_est;
513 } ____cacheline_aligned;
514
515 /*
516 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
517 * updates. When a task is dequeued, its util_est should not be updated if its
518 * util_avg has not been updated in the meantime.
519 * This information is mapped into the MSB bit of util_est at dequeue time.
520 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
521 * it is safe to use MSB.
522 */
523 #define UTIL_EST_WEIGHT_SHIFT 2
524 #define UTIL_AVG_UNCHANGED 0x80000000
525
526 struct sched_statistics {
527 #ifdef CONFIG_SCHEDSTATS
528 u64 wait_start;
529 u64 wait_max;
530 u64 wait_count;
531 u64 wait_sum;
532 u64 iowait_count;
533 u64 iowait_sum;
534
535 u64 sleep_start;
536 u64 sleep_max;
537 s64 sum_sleep_runtime;
538
539 u64 block_start;
540 u64 block_max;
541 s64 sum_block_runtime;
542
543 s64 exec_max;
544 u64 slice_max;
545
546 u64 nr_migrations_cold;
547 u64 nr_failed_migrations_affine;
548 u64 nr_failed_migrations_running;
549 u64 nr_failed_migrations_hot;
550 u64 nr_forced_migrations;
551
552 u64 nr_wakeups;
553 u64 nr_wakeups_sync;
554 u64 nr_wakeups_migrate;
555 u64 nr_wakeups_local;
556 u64 nr_wakeups_remote;
557 u64 nr_wakeups_affine;
558 u64 nr_wakeups_affine_attempts;
559 u64 nr_wakeups_passive;
560 u64 nr_wakeups_idle;
561
562 #ifdef CONFIG_SCHED_CORE
563 u64 core_forceidle_sum;
564 #endif
565 #endif /* CONFIG_SCHEDSTATS */
566 } ____cacheline_aligned;
567
568 struct sched_entity {
569 /* For load-balancing: */
570 struct load_weight load;
571 struct rb_node run_node;
572 u64 deadline;
573 u64 min_vruntime;
574 u64 min_slice;
575
576 struct list_head group_node;
577 unsigned char on_rq;
578 unsigned char sched_delayed;
579 unsigned char rel_deadline;
580 unsigned char custom_slice;
581 /* hole */
582
583 u64 exec_start;
584 u64 sum_exec_runtime;
585 u64 prev_sum_exec_runtime;
586 u64 vruntime;
587 s64 vlag;
588 u64 slice;
589
590 u64 nr_migrations;
591
592 #ifdef CONFIG_FAIR_GROUP_SCHED
593 int depth;
594 struct sched_entity *parent;
595 /* rq on which this entity is (to be) queued: */
596 struct cfs_rq *cfs_rq;
597 /* rq "owned" by this entity/group: */
598 struct cfs_rq *my_q;
599 /* cached value of my_q->h_nr_running */
600 unsigned long runnable_weight;
601 #endif
602
603 #ifdef CONFIG_SMP
604 /*
605 * Per entity load average tracking.
606 *
607 * Put into separate cache line so it does not
608 * collide with read-mostly values above.
609 */
610 struct sched_avg avg;
611 #endif
612 };
613
614 struct sched_rt_entity {
615 struct list_head run_list;
616 unsigned long timeout;
617 unsigned long watchdog_stamp;
618 unsigned int time_slice;
619 unsigned short on_rq;
620 unsigned short on_list;
621
622 struct sched_rt_entity *back;
623 #ifdef CONFIG_RT_GROUP_SCHED
624 struct sched_rt_entity *parent;
625 /* rq on which this entity is (to be) queued: */
626 struct rt_rq *rt_rq;
627 /* rq "owned" by this entity/group: */
628 struct rt_rq *my_q;
629 #endif
630 } __randomize_layout;
631
632 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
633 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
634
635 struct sched_dl_entity {
636 struct rb_node rb_node;
637
638 /*
639 * Original scheduling parameters. Copied here from sched_attr
640 * during sched_setattr(), they will remain the same until
641 * the next sched_setattr().
642 */
643 u64 dl_runtime; /* Maximum runtime for each instance */
644 u64 dl_deadline; /* Relative deadline of each instance */
645 u64 dl_period; /* Separation of two instances (period) */
646 u64 dl_bw; /* dl_runtime / dl_period */
647 u64 dl_density; /* dl_runtime / dl_deadline */
648
649 /*
650 * Actual scheduling parameters. Initialized with the values above,
651 * they are continuously updated during task execution. Note that
652 * the remaining runtime could be < 0 in case we are in overrun.
653 */
654 s64 runtime; /* Remaining runtime for this instance */
655 u64 deadline; /* Absolute deadline for this instance */
656 unsigned int flags; /* Specifying the scheduler behaviour */
657
658 /*
659 * Some bool flags:
660 *
661 * @dl_throttled tells if we exhausted the runtime. If so, the
662 * task has to wait for a replenishment to be performed at the
663 * next firing of dl_timer.
664 *
665 * @dl_yielded tells if task gave up the CPU before consuming
666 * all its available runtime during the last job.
667 *
668 * @dl_non_contending tells if the task is inactive while still
669 * contributing to the active utilization. In other words, it
670 * indicates if the inactive timer has been armed and its handler
671 * has not been executed yet. This flag is useful to avoid race
672 * conditions between the inactive timer handler and the wakeup
673 * code.
674 *
675 * @dl_overrun tells if the task asked to be informed about runtime
676 * overruns.
677 *
678 * @dl_server tells if this is a server entity.
679 *
680 * @dl_defer tells if this is a deferred or regular server. For
681 * now only defer server exists.
682 *
683 * @dl_defer_armed tells if the deferrable server is waiting
684 * for the replenishment timer to activate it.
685 *
686 * @dl_server_active tells if the dlserver is active(started).
687 * dlserver is started on first cfs enqueue on an idle runqueue
688 * and is stopped when a dequeue results in 0 cfs tasks on the
689 * runqueue. In other words, dlserver is active only when cpu's
690 * runqueue has atleast one cfs task.
691 *
692 * @dl_defer_running tells if the deferrable server is actually
693 * running, skipping the defer phase.
694 */
695 unsigned int dl_throttled : 1;
696 unsigned int dl_yielded : 1;
697 unsigned int dl_non_contending : 1;
698 unsigned int dl_overrun : 1;
699 unsigned int dl_server : 1;
700 unsigned int dl_server_active : 1;
701 unsigned int dl_defer : 1;
702 unsigned int dl_defer_armed : 1;
703 unsigned int dl_defer_running : 1;
704
705 /*
706 * Bandwidth enforcement timer. Each -deadline task has its
707 * own bandwidth to be enforced, thus we need one timer per task.
708 */
709 struct hrtimer dl_timer;
710
711 /*
712 * Inactive timer, responsible for decreasing the active utilization
713 * at the "0-lag time". When a -deadline task blocks, it contributes
714 * to GRUB's active utilization until the "0-lag time", hence a
715 * timer is needed to decrease the active utilization at the correct
716 * time.
717 */
718 struct hrtimer inactive_timer;
719
720 /*
721 * Bits for DL-server functionality. Also see the comment near
722 * dl_server_update().
723 *
724 * @rq the runqueue this server is for
725 *
726 * @server_has_tasks() returns true if @server_pick return a
727 * runnable task.
728 */
729 struct rq *rq;
730 dl_server_has_tasks_f server_has_tasks;
731 dl_server_pick_f server_pick_task;
732
733 #ifdef CONFIG_RT_MUTEXES
734 /*
735 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
736 * pi_se points to the donor, otherwise points to the dl_se it belongs
737 * to (the original one/itself).
738 */
739 struct sched_dl_entity *pi_se;
740 #endif
741 };
742
743 #ifdef CONFIG_UCLAMP_TASK
744 /* Number of utilization clamp buckets (shorter alias) */
745 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
746
747 /*
748 * Utilization clamp for a scheduling entity
749 * @value: clamp value "assigned" to a se
750 * @bucket_id: bucket index corresponding to the "assigned" value
751 * @active: the se is currently refcounted in a rq's bucket
752 * @user_defined: the requested clamp value comes from user-space
753 *
754 * The bucket_id is the index of the clamp bucket matching the clamp value
755 * which is pre-computed and stored to avoid expensive integer divisions from
756 * the fast path.
757 *
758 * The active bit is set whenever a task has got an "effective" value assigned,
759 * which can be different from the clamp value "requested" from user-space.
760 * This allows to know a task is refcounted in the rq's bucket corresponding
761 * to the "effective" bucket_id.
762 *
763 * The user_defined bit is set whenever a task has got a task-specific clamp
764 * value requested from userspace, i.e. the system defaults apply to this task
765 * just as a restriction. This allows to relax default clamps when a less
766 * restrictive task-specific value has been requested, thus allowing to
767 * implement a "nice" semantic. For example, a task running with a 20%
768 * default boost can still drop its own boosting to 0%.
769 */
770 struct uclamp_se {
771 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
772 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
773 unsigned int active : 1;
774 unsigned int user_defined : 1;
775 };
776 #endif /* CONFIG_UCLAMP_TASK */
777
778 union rcu_special {
779 struct {
780 u8 blocked;
781 u8 need_qs;
782 u8 exp_hint; /* Hint for performance. */
783 u8 need_mb; /* Readers need smp_mb(). */
784 } b; /* Bits. */
785 u32 s; /* Set of bits. */
786 };
787
788 enum perf_event_task_context {
789 perf_invalid_context = -1,
790 perf_hw_context = 0,
791 perf_sw_context,
792 perf_nr_task_contexts,
793 };
794
795 /*
796 * Number of contexts where an event can trigger:
797 * task, softirq, hardirq, nmi.
798 */
799 #define PERF_NR_CONTEXTS 4
800
801 struct wake_q_node {
802 struct wake_q_node *next;
803 };
804
805 struct kmap_ctrl {
806 #ifdef CONFIG_KMAP_LOCAL
807 int idx;
808 pte_t pteval[KM_MAX_IDX];
809 #endif
810 };
811
812 struct task_struct {
813 #ifdef CONFIG_THREAD_INFO_IN_TASK
814 /*
815 * For reasons of header soup (see current_thread_info()), this
816 * must be the first element of task_struct.
817 */
818 struct thread_info thread_info;
819 #endif
820 unsigned int __state;
821
822 /* saved state for "spinlock sleepers" */
823 unsigned int saved_state;
824
825 /*
826 * This begins the randomizable portion of task_struct. Only
827 * scheduling-critical items should be added above here.
828 */
829 randomized_struct_fields_start
830
831 void *stack;
832 refcount_t usage;
833 /* Per task flags (PF_*), defined further below: */
834 unsigned int flags;
835 unsigned int ptrace;
836
837 #ifdef CONFIG_MEM_ALLOC_PROFILING
838 struct alloc_tag *alloc_tag;
839 #endif
840
841 #ifdef CONFIG_SMP
842 int on_cpu;
843 struct __call_single_node wake_entry;
844 unsigned int wakee_flips;
845 unsigned long wakee_flip_decay_ts;
846 struct task_struct *last_wakee;
847
848 /*
849 * recent_used_cpu is initially set as the last CPU used by a task
850 * that wakes affine another task. Waker/wakee relationships can
851 * push tasks around a CPU where each wakeup moves to the next one.
852 * Tracking a recently used CPU allows a quick search for a recently
853 * used CPU that may be idle.
854 */
855 int recent_used_cpu;
856 int wake_cpu;
857 #endif
858 int on_rq;
859
860 int prio;
861 int static_prio;
862 int normal_prio;
863 unsigned int rt_priority;
864
865 struct sched_entity se;
866 struct sched_rt_entity rt;
867 struct sched_dl_entity dl;
868 struct sched_dl_entity *dl_server;
869 #ifdef CONFIG_SCHED_CLASS_EXT
870 struct sched_ext_entity scx;
871 #endif
872 const struct sched_class *sched_class;
873
874 #ifdef CONFIG_SCHED_CORE
875 struct rb_node core_node;
876 unsigned long core_cookie;
877 unsigned int core_occupation;
878 #endif
879
880 #ifdef CONFIG_CGROUP_SCHED
881 struct task_group *sched_task_group;
882 #endif
883
884
885 #ifdef CONFIG_UCLAMP_TASK
886 /*
887 * Clamp values requested for a scheduling entity.
888 * Must be updated with task_rq_lock() held.
889 */
890 struct uclamp_se uclamp_req[UCLAMP_CNT];
891 /*
892 * Effective clamp values used for a scheduling entity.
893 * Must be updated with task_rq_lock() held.
894 */
895 struct uclamp_se uclamp[UCLAMP_CNT];
896 #endif
897
898 struct sched_statistics stats;
899
900 #ifdef CONFIG_PREEMPT_NOTIFIERS
901 /* List of struct preempt_notifier: */
902 struct hlist_head preempt_notifiers;
903 #endif
904
905 #ifdef CONFIG_BLK_DEV_IO_TRACE
906 unsigned int btrace_seq;
907 #endif
908
909 unsigned int policy;
910 unsigned long max_allowed_capacity;
911 int nr_cpus_allowed;
912 const cpumask_t *cpus_ptr;
913 cpumask_t *user_cpus_ptr;
914 cpumask_t cpus_mask;
915 void *migration_pending;
916 #ifdef CONFIG_SMP
917 unsigned short migration_disabled;
918 #endif
919 unsigned short migration_flags;
920
921 #ifdef CONFIG_PREEMPT_RCU
922 int rcu_read_lock_nesting;
923 union rcu_special rcu_read_unlock_special;
924 struct list_head rcu_node_entry;
925 struct rcu_node *rcu_blocked_node;
926 #endif /* #ifdef CONFIG_PREEMPT_RCU */
927
928 #ifdef CONFIG_TASKS_RCU
929 unsigned long rcu_tasks_nvcsw;
930 u8 rcu_tasks_holdout;
931 u8 rcu_tasks_idx;
932 int rcu_tasks_idle_cpu;
933 struct list_head rcu_tasks_holdout_list;
934 int rcu_tasks_exit_cpu;
935 struct list_head rcu_tasks_exit_list;
936 #endif /* #ifdef CONFIG_TASKS_RCU */
937
938 #ifdef CONFIG_TASKS_TRACE_RCU
939 int trc_reader_nesting;
940 int trc_ipi_to_cpu;
941 union rcu_special trc_reader_special;
942 struct list_head trc_holdout_list;
943 struct list_head trc_blkd_node;
944 int trc_blkd_cpu;
945 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
946
947 struct sched_info sched_info;
948
949 struct list_head tasks;
950 #ifdef CONFIG_SMP
951 struct plist_node pushable_tasks;
952 struct rb_node pushable_dl_tasks;
953 #endif
954
955 struct mm_struct *mm;
956 struct mm_struct *active_mm;
957 struct address_space *faults_disabled_mapping;
958
959 int exit_state;
960 int exit_code;
961 int exit_signal;
962 /* The signal sent when the parent dies: */
963 int pdeath_signal;
964 /* JOBCTL_*, siglock protected: */
965 unsigned long jobctl;
966
967 /* Used for emulating ABI behavior of previous Linux versions: */
968 unsigned int personality;
969
970 /* Scheduler bits, serialized by scheduler locks: */
971 unsigned sched_reset_on_fork:1;
972 unsigned sched_contributes_to_load:1;
973 unsigned sched_migrated:1;
974 unsigned sched_task_hot:1;
975
976 /* Force alignment to the next boundary: */
977 unsigned :0;
978
979 /* Unserialized, strictly 'current' */
980
981 /*
982 * This field must not be in the scheduler word above due to wakelist
983 * queueing no longer being serialized by p->on_cpu. However:
984 *
985 * p->XXX = X; ttwu()
986 * schedule() if (p->on_rq && ..) // false
987 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
988 * deactivate_task() ttwu_queue_wakelist())
989 * p->on_rq = 0; p->sched_remote_wakeup = Y;
990 *
991 * guarantees all stores of 'current' are visible before
992 * ->sched_remote_wakeup gets used, so it can be in this word.
993 */
994 unsigned sched_remote_wakeup:1;
995 #ifdef CONFIG_RT_MUTEXES
996 unsigned sched_rt_mutex:1;
997 #endif
998
999 /* Bit to tell TOMOYO we're in execve(): */
1000 unsigned in_execve:1;
1001 unsigned in_iowait:1;
1002 #ifndef TIF_RESTORE_SIGMASK
1003 unsigned restore_sigmask:1;
1004 #endif
1005 #ifdef CONFIG_MEMCG_V1
1006 unsigned in_user_fault:1;
1007 #endif
1008 #ifdef CONFIG_LRU_GEN
1009 /* whether the LRU algorithm may apply to this access */
1010 unsigned in_lru_fault:1;
1011 #endif
1012 #ifdef CONFIG_COMPAT_BRK
1013 unsigned brk_randomized:1;
1014 #endif
1015 #ifdef CONFIG_CGROUPS
1016 /* disallow userland-initiated cgroup migration */
1017 unsigned no_cgroup_migration:1;
1018 /* task is frozen/stopped (used by the cgroup freezer) */
1019 unsigned frozen:1;
1020 #endif
1021 #ifdef CONFIG_BLK_CGROUP
1022 unsigned use_memdelay:1;
1023 #endif
1024 #ifdef CONFIG_PSI
1025 /* Stalled due to lack of memory */
1026 unsigned in_memstall:1;
1027 #endif
1028 #ifdef CONFIG_PAGE_OWNER
1029 /* Used by page_owner=on to detect recursion in page tracking. */
1030 unsigned in_page_owner:1;
1031 #endif
1032 #ifdef CONFIG_EVENTFD
1033 /* Recursion prevention for eventfd_signal() */
1034 unsigned in_eventfd:1;
1035 #endif
1036 #ifdef CONFIG_ARCH_HAS_CPU_PASID
1037 unsigned pasid_activated:1;
1038 #endif
1039 #ifdef CONFIG_X86_BUS_LOCK_DETECT
1040 unsigned reported_split_lock:1;
1041 #endif
1042 #ifdef CONFIG_TASK_DELAY_ACCT
1043 /* delay due to memory thrashing */
1044 unsigned in_thrashing:1;
1045 #endif
1046 unsigned in_nf_duplicate:1;
1047 #ifdef CONFIG_PREEMPT_RT
1048 struct netdev_xmit net_xmit;
1049 #endif
1050 unsigned long atomic_flags; /* Flags requiring atomic access. */
1051
1052 struct restart_block restart_block;
1053
1054 pid_t pid;
1055 pid_t tgid;
1056
1057 #ifdef CONFIG_STACKPROTECTOR
1058 /* Canary value for the -fstack-protector GCC feature: */
1059 unsigned long stack_canary;
1060 #endif
1061 /*
1062 * Pointers to the (original) parent process, youngest child, younger sibling,
1063 * older sibling, respectively. (p->father can be replaced with
1064 * p->real_parent->pid)
1065 */
1066
1067 /* Real parent process: */
1068 struct task_struct __rcu *real_parent;
1069
1070 /* Recipient of SIGCHLD, wait4() reports: */
1071 struct task_struct __rcu *parent;
1072
1073 /*
1074 * Children/sibling form the list of natural children:
1075 */
1076 struct list_head children;
1077 struct list_head sibling;
1078 struct task_struct *group_leader;
1079
1080 /*
1081 * 'ptraced' is the list of tasks this task is using ptrace() on.
1082 *
1083 * This includes both natural children and PTRACE_ATTACH targets.
1084 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1085 */
1086 struct list_head ptraced;
1087 struct list_head ptrace_entry;
1088
1089 /* PID/PID hash table linkage. */
1090 struct pid *thread_pid;
1091 struct hlist_node pid_links[PIDTYPE_MAX];
1092 struct list_head thread_node;
1093
1094 struct completion *vfork_done;
1095
1096 /* CLONE_CHILD_SETTID: */
1097 int __user *set_child_tid;
1098
1099 /* CLONE_CHILD_CLEARTID: */
1100 int __user *clear_child_tid;
1101
1102 /* PF_KTHREAD | PF_IO_WORKER */
1103 void *worker_private;
1104
1105 u64 utime;
1106 u64 stime;
1107 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1108 u64 utimescaled;
1109 u64 stimescaled;
1110 #endif
1111 u64 gtime;
1112 struct prev_cputime prev_cputime;
1113 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1114 struct vtime vtime;
1115 #endif
1116
1117 #ifdef CONFIG_NO_HZ_FULL
1118 atomic_t tick_dep_mask;
1119 #endif
1120 /* Context switch counts: */
1121 unsigned long nvcsw;
1122 unsigned long nivcsw;
1123
1124 /* Monotonic time in nsecs: */
1125 u64 start_time;
1126
1127 /* Boot based time in nsecs: */
1128 u64 start_boottime;
1129
1130 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1131 unsigned long min_flt;
1132 unsigned long maj_flt;
1133
1134 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1135 struct posix_cputimers posix_cputimers;
1136
1137 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1138 struct posix_cputimers_work posix_cputimers_work;
1139 #endif
1140
1141 /* Process credentials: */
1142
1143 /* Tracer's credentials at attach: */
1144 const struct cred __rcu *ptracer_cred;
1145
1146 /* Objective and real subjective task credentials (COW): */
1147 const struct cred __rcu *real_cred;
1148
1149 /* Effective (overridable) subjective task credentials (COW): */
1150 const struct cred __rcu *cred;
1151
1152 #ifdef CONFIG_KEYS
1153 /* Cached requested key. */
1154 struct key *cached_requested_key;
1155 #endif
1156
1157 /*
1158 * executable name, excluding path.
1159 *
1160 * - normally initialized begin_new_exec()
1161 * - set it with set_task_comm()
1162 * - strscpy_pad() to ensure it is always NUL-terminated and
1163 * zero-padded
1164 * - task_lock() to ensure the operation is atomic and the name is
1165 * fully updated.
1166 */
1167 char comm[TASK_COMM_LEN];
1168
1169 struct nameidata *nameidata;
1170
1171 #ifdef CONFIG_SYSVIPC
1172 struct sysv_sem sysvsem;
1173 struct sysv_shm sysvshm;
1174 #endif
1175 #ifdef CONFIG_DETECT_HUNG_TASK
1176 unsigned long last_switch_count;
1177 unsigned long last_switch_time;
1178 #endif
1179 /* Filesystem information: */
1180 struct fs_struct *fs;
1181
1182 /* Open file information: */
1183 struct files_struct *files;
1184
1185 #ifdef CONFIG_IO_URING
1186 struct io_uring_task *io_uring;
1187 #endif
1188
1189 /* Namespaces: */
1190 struct nsproxy *nsproxy;
1191
1192 /* Signal handlers: */
1193 struct signal_struct *signal;
1194 struct sighand_struct __rcu *sighand;
1195 sigset_t blocked;
1196 sigset_t real_blocked;
1197 /* Restored if set_restore_sigmask() was used: */
1198 sigset_t saved_sigmask;
1199 struct sigpending pending;
1200 unsigned long sas_ss_sp;
1201 size_t sas_ss_size;
1202 unsigned int sas_ss_flags;
1203
1204 struct callback_head *task_works;
1205
1206 #ifdef CONFIG_AUDIT
1207 #ifdef CONFIG_AUDITSYSCALL
1208 struct audit_context *audit_context;
1209 #endif
1210 kuid_t loginuid;
1211 unsigned int sessionid;
1212 #endif
1213 struct seccomp seccomp;
1214 struct syscall_user_dispatch syscall_dispatch;
1215
1216 /* Thread group tracking: */
1217 u64 parent_exec_id;
1218 u64 self_exec_id;
1219
1220 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1221 spinlock_t alloc_lock;
1222
1223 /* Protection of the PI data structures: */
1224 raw_spinlock_t pi_lock;
1225
1226 struct wake_q_node wake_q;
1227
1228 #ifdef CONFIG_RT_MUTEXES
1229 /* PI waiters blocked on a rt_mutex held by this task: */
1230 struct rb_root_cached pi_waiters;
1231 /* Updated under owner's pi_lock and rq lock */
1232 struct task_struct *pi_top_task;
1233 /* Deadlock detection and priority inheritance handling: */
1234 struct rt_mutex_waiter *pi_blocked_on;
1235 #endif
1236
1237 #ifdef CONFIG_DEBUG_MUTEXES
1238 /* Mutex deadlock detection: */
1239 struct mutex_waiter *blocked_on;
1240 #endif
1241
1242 #ifdef CONFIG_DETECT_HUNG_TASK_BLOCKER
1243 /*
1244 * Encoded lock address causing task block (lower 2 bits = type from
1245 * <linux/hung_task.h>). Accessed via hung_task_*() helpers.
1246 */
1247 unsigned long blocker;
1248 #endif
1249
1250 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1251 int non_block_count;
1252 #endif
1253
1254 #ifdef CONFIG_TRACE_IRQFLAGS
1255 struct irqtrace_events irqtrace;
1256 unsigned int hardirq_threaded;
1257 u64 hardirq_chain_key;
1258 int softirqs_enabled;
1259 int softirq_context;
1260 int irq_config;
1261 #endif
1262 #ifdef CONFIG_PREEMPT_RT
1263 int softirq_disable_cnt;
1264 #endif
1265
1266 #ifdef CONFIG_LOCKDEP
1267 # define MAX_LOCK_DEPTH 48UL
1268 u64 curr_chain_key;
1269 int lockdep_depth;
1270 unsigned int lockdep_recursion;
1271 struct held_lock held_locks[MAX_LOCK_DEPTH];
1272 #endif
1273
1274 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1275 unsigned int in_ubsan;
1276 #endif
1277
1278 /* Journalling filesystem info: */
1279 void *journal_info;
1280
1281 /* Stacked block device info: */
1282 struct bio_list *bio_list;
1283
1284 /* Stack plugging: */
1285 struct blk_plug *plug;
1286
1287 /* VM state: */
1288 struct reclaim_state *reclaim_state;
1289
1290 struct io_context *io_context;
1291
1292 #ifdef CONFIG_COMPACTION
1293 struct capture_control *capture_control;
1294 #endif
1295 /* Ptrace state: */
1296 unsigned long ptrace_message;
1297 kernel_siginfo_t *last_siginfo;
1298
1299 struct task_io_accounting ioac;
1300 #ifdef CONFIG_PSI
1301 /* Pressure stall state */
1302 unsigned int psi_flags;
1303 #endif
1304 #ifdef CONFIG_TASK_XACCT
1305 /* Accumulated RSS usage: */
1306 u64 acct_rss_mem1;
1307 /* Accumulated virtual memory usage: */
1308 u64 acct_vm_mem1;
1309 /* stime + utime since last update: */
1310 u64 acct_timexpd;
1311 #endif
1312 #ifdef CONFIG_CPUSETS
1313 /* Protected by ->alloc_lock: */
1314 nodemask_t mems_allowed;
1315 /* Sequence number to catch updates: */
1316 seqcount_spinlock_t mems_allowed_seq;
1317 int cpuset_mem_spread_rotor;
1318 #endif
1319 #ifdef CONFIG_CGROUPS
1320 /* Control Group info protected by css_set_lock: */
1321 struct css_set __rcu *cgroups;
1322 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1323 struct list_head cg_list;
1324 #endif
1325 #ifdef CONFIG_X86_CPU_RESCTRL
1326 u32 closid;
1327 u32 rmid;
1328 #endif
1329 #ifdef CONFIG_FUTEX
1330 struct robust_list_head __user *robust_list;
1331 #ifdef CONFIG_COMPAT
1332 struct compat_robust_list_head __user *compat_robust_list;
1333 #endif
1334 struct list_head pi_state_list;
1335 struct futex_pi_state *pi_state_cache;
1336 struct mutex futex_exit_mutex;
1337 unsigned int futex_state;
1338 #endif
1339 #ifdef CONFIG_PERF_EVENTS
1340 u8 perf_recursion[PERF_NR_CONTEXTS];
1341 struct perf_event_context *perf_event_ctxp;
1342 struct mutex perf_event_mutex;
1343 struct list_head perf_event_list;
1344 struct perf_ctx_data __rcu *perf_ctx_data;
1345 #endif
1346 #ifdef CONFIG_DEBUG_PREEMPT
1347 unsigned long preempt_disable_ip;
1348 #endif
1349 #ifdef CONFIG_NUMA
1350 /* Protected by alloc_lock: */
1351 struct mempolicy *mempolicy;
1352 short il_prev;
1353 u8 il_weight;
1354 short pref_node_fork;
1355 #endif
1356 #ifdef CONFIG_NUMA_BALANCING
1357 int numa_scan_seq;
1358 unsigned int numa_scan_period;
1359 unsigned int numa_scan_period_max;
1360 int numa_preferred_nid;
1361 unsigned long numa_migrate_retry;
1362 /* Migration stamp: */
1363 u64 node_stamp;
1364 u64 last_task_numa_placement;
1365 u64 last_sum_exec_runtime;
1366 struct callback_head numa_work;
1367
1368 /*
1369 * This pointer is only modified for current in syscall and
1370 * pagefault context (and for tasks being destroyed), so it can be read
1371 * from any of the following contexts:
1372 * - RCU read-side critical section
1373 * - current->numa_group from everywhere
1374 * - task's runqueue locked, task not running
1375 */
1376 struct numa_group __rcu *numa_group;
1377
1378 /*
1379 * numa_faults is an array split into four regions:
1380 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1381 * in this precise order.
1382 *
1383 * faults_memory: Exponential decaying average of faults on a per-node
1384 * basis. Scheduling placement decisions are made based on these
1385 * counts. The values remain static for the duration of a PTE scan.
1386 * faults_cpu: Track the nodes the process was running on when a NUMA
1387 * hinting fault was incurred.
1388 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1389 * during the current scan window. When the scan completes, the counts
1390 * in faults_memory and faults_cpu decay and these values are copied.
1391 */
1392 unsigned long *numa_faults;
1393 unsigned long total_numa_faults;
1394
1395 /*
1396 * numa_faults_locality tracks if faults recorded during the last
1397 * scan window were remote/local or failed to migrate. The task scan
1398 * period is adapted based on the locality of the faults with different
1399 * weights depending on whether they were shared or private faults
1400 */
1401 unsigned long numa_faults_locality[3];
1402
1403 unsigned long numa_pages_migrated;
1404 #endif /* CONFIG_NUMA_BALANCING */
1405
1406 #ifdef CONFIG_RSEQ
1407 struct rseq __user *rseq;
1408 u32 rseq_len;
1409 u32 rseq_sig;
1410 /*
1411 * RmW on rseq_event_mask must be performed atomically
1412 * with respect to preemption.
1413 */
1414 unsigned long rseq_event_mask;
1415 # ifdef CONFIG_DEBUG_RSEQ
1416 /*
1417 * This is a place holder to save a copy of the rseq fields for
1418 * validation of read-only fields. The struct rseq has a
1419 * variable-length array at the end, so it cannot be used
1420 * directly. Reserve a size large enough for the known fields.
1421 */
1422 char rseq_fields[sizeof(struct rseq)];
1423 # endif
1424 #endif
1425
1426 #ifdef CONFIG_SCHED_MM_CID
1427 int mm_cid; /* Current cid in mm */
1428 int last_mm_cid; /* Most recent cid in mm */
1429 int migrate_from_cpu;
1430 int mm_cid_active; /* Whether cid bitmap is active */
1431 struct callback_head cid_work;
1432 #endif
1433
1434 struct tlbflush_unmap_batch tlb_ubc;
1435
1436 /* Cache last used pipe for splice(): */
1437 struct pipe_inode_info *splice_pipe;
1438
1439 struct page_frag task_frag;
1440
1441 #ifdef CONFIG_TASK_DELAY_ACCT
1442 struct task_delay_info *delays;
1443 #endif
1444
1445 #ifdef CONFIG_FAULT_INJECTION
1446 int make_it_fail;
1447 unsigned int fail_nth;
1448 #endif
1449 /*
1450 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1451 * balance_dirty_pages() for a dirty throttling pause:
1452 */
1453 int nr_dirtied;
1454 int nr_dirtied_pause;
1455 /* Start of a write-and-pause period: */
1456 unsigned long dirty_paused_when;
1457
1458 #ifdef CONFIG_LATENCYTOP
1459 int latency_record_count;
1460 struct latency_record latency_record[LT_SAVECOUNT];
1461 #endif
1462 /*
1463 * Time slack values; these are used to round up poll() and
1464 * select() etc timeout values. These are in nanoseconds.
1465 */
1466 u64 timer_slack_ns;
1467 u64 default_timer_slack_ns;
1468
1469 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1470 unsigned int kasan_depth;
1471 #endif
1472
1473 #ifdef CONFIG_KCSAN
1474 struct kcsan_ctx kcsan_ctx;
1475 #ifdef CONFIG_TRACE_IRQFLAGS
1476 struct irqtrace_events kcsan_save_irqtrace;
1477 #endif
1478 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1479 int kcsan_stack_depth;
1480 #endif
1481 #endif
1482
1483 #ifdef CONFIG_KMSAN
1484 struct kmsan_ctx kmsan_ctx;
1485 #endif
1486
1487 #if IS_ENABLED(CONFIG_KUNIT)
1488 struct kunit *kunit_test;
1489 #endif
1490
1491 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1492 /* Index of current stored address in ret_stack: */
1493 int curr_ret_stack;
1494 int curr_ret_depth;
1495
1496 /* Stack of return addresses for return function tracing: */
1497 unsigned long *ret_stack;
1498
1499 /* Timestamp for last schedule: */
1500 unsigned long long ftrace_timestamp;
1501 unsigned long long ftrace_sleeptime;
1502
1503 /*
1504 * Number of functions that haven't been traced
1505 * because of depth overrun:
1506 */
1507 atomic_t trace_overrun;
1508
1509 /* Pause tracing: */
1510 atomic_t tracing_graph_pause;
1511 #endif
1512
1513 #ifdef CONFIG_TRACING
1514 /* Bitmask and counter of trace recursion: */
1515 unsigned long trace_recursion;
1516 #endif /* CONFIG_TRACING */
1517
1518 #ifdef CONFIG_KCOV
1519 /* See kernel/kcov.c for more details. */
1520
1521 /* Coverage collection mode enabled for this task (0 if disabled): */
1522 unsigned int kcov_mode;
1523
1524 /* Size of the kcov_area: */
1525 unsigned int kcov_size;
1526
1527 /* Buffer for coverage collection: */
1528 void *kcov_area;
1529
1530 /* KCOV descriptor wired with this task or NULL: */
1531 struct kcov *kcov;
1532
1533 /* KCOV common handle for remote coverage collection: */
1534 u64 kcov_handle;
1535
1536 /* KCOV sequence number: */
1537 int kcov_sequence;
1538
1539 /* Collect coverage from softirq context: */
1540 unsigned int kcov_softirq;
1541 #endif
1542
1543 #ifdef CONFIG_MEMCG_V1
1544 struct mem_cgroup *memcg_in_oom;
1545 #endif
1546
1547 #ifdef CONFIG_MEMCG
1548 /* Number of pages to reclaim on returning to userland: */
1549 unsigned int memcg_nr_pages_over_high;
1550
1551 /* Used by memcontrol for targeted memcg charge: */
1552 struct mem_cgroup *active_memcg;
1553
1554 /* Cache for current->cgroups->memcg->objcg lookups: */
1555 struct obj_cgroup *objcg;
1556 #endif
1557
1558 #ifdef CONFIG_BLK_CGROUP
1559 struct gendisk *throttle_disk;
1560 #endif
1561
1562 #ifdef CONFIG_UPROBES
1563 struct uprobe_task *utask;
1564 #endif
1565 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1566 unsigned int sequential_io;
1567 unsigned int sequential_io_avg;
1568 #endif
1569 struct kmap_ctrl kmap_ctrl;
1570 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1571 unsigned long task_state_change;
1572 # ifdef CONFIG_PREEMPT_RT
1573 unsigned long saved_state_change;
1574 # endif
1575 #endif
1576 struct rcu_head rcu;
1577 refcount_t rcu_users;
1578 int pagefault_disabled;
1579 #ifdef CONFIG_MMU
1580 struct task_struct *oom_reaper_list;
1581 struct timer_list oom_reaper_timer;
1582 #endif
1583 #ifdef CONFIG_VMAP_STACK
1584 struct vm_struct *stack_vm_area;
1585 #endif
1586 #ifdef CONFIG_THREAD_INFO_IN_TASK
1587 /* A live task holds one reference: */
1588 refcount_t stack_refcount;
1589 #endif
1590 #ifdef CONFIG_LIVEPATCH
1591 int patch_state;
1592 #endif
1593 #ifdef CONFIG_SECURITY
1594 /* Used by LSM modules for access restriction: */
1595 void *security;
1596 #endif
1597 #ifdef CONFIG_BPF_SYSCALL
1598 /* Used by BPF task local storage */
1599 struct bpf_local_storage __rcu *bpf_storage;
1600 /* Used for BPF run context */
1601 struct bpf_run_ctx *bpf_ctx;
1602 #endif
1603 /* Used by BPF for per-TASK xdp storage */
1604 struct bpf_net_context *bpf_net_context;
1605
1606 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1607 unsigned long lowest_stack;
1608 unsigned long prev_lowest_stack;
1609 #endif
1610
1611 #ifdef CONFIG_X86_MCE
1612 void __user *mce_vaddr;
1613 __u64 mce_kflags;
1614 u64 mce_addr;
1615 __u64 mce_ripv : 1,
1616 mce_whole_page : 1,
1617 __mce_reserved : 62;
1618 struct callback_head mce_kill_me;
1619 int mce_count;
1620 #endif
1621
1622 #ifdef CONFIG_KRETPROBES
1623 struct llist_head kretprobe_instances;
1624 #endif
1625 #ifdef CONFIG_RETHOOK
1626 struct llist_head rethooks;
1627 #endif
1628
1629 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1630 /*
1631 * If L1D flush is supported on mm context switch
1632 * then we use this callback head to queue kill work
1633 * to kill tasks that are not running on SMT disabled
1634 * cores
1635 */
1636 struct callback_head l1d_flush_kill;
1637 #endif
1638
1639 #ifdef CONFIG_RV
1640 /*
1641 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1642 * If we find justification for more monitors, we can think
1643 * about adding more or developing a dynamic method. So far,
1644 * none of these are justified.
1645 */
1646 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1647 #endif
1648
1649 #ifdef CONFIG_USER_EVENTS
1650 struct user_event_mm *user_event_mm;
1651 #endif
1652
1653 /* CPU-specific state of this task: */
1654 struct thread_struct thread;
1655
1656 /*
1657 * New fields for task_struct should be added above here, so that
1658 * they are included in the randomized portion of task_struct.
1659 */
1660 randomized_struct_fields_end
1661 } __attribute__ ((aligned (64)));
1662
1663 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1664 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1665
__task_state_index(unsigned int tsk_state,unsigned int tsk_exit_state)1666 static inline unsigned int __task_state_index(unsigned int tsk_state,
1667 unsigned int tsk_exit_state)
1668 {
1669 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1670
1671 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1672
1673 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1674 state = TASK_REPORT_IDLE;
1675
1676 /*
1677 * We're lying here, but rather than expose a completely new task state
1678 * to userspace, we can make this appear as if the task has gone through
1679 * a regular rt_mutex_lock() call.
1680 * Report frozen tasks as uninterruptible.
1681 */
1682 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
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(TASK_REPORT_MAX * 2 != 1 << (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 extern struct pid *cad_pid;
1708
1709 /*
1710 * Per process flags
1711 */
1712 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1713 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1714 #define PF_EXITING 0x00000004 /* Getting shut down */
1715 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1716 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1717 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1718 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1719 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1720 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1721 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1722 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1723 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1724 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1725 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1726 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1727 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1728 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */
1729 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1730 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1731 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1732 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1733 * I am cleaning dirty pages from some other bdi. */
1734 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1735 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1736 #define PF__HOLE__00800000 0x00800000
1737 #define PF__HOLE__01000000 0x01000000
1738 #define PF__HOLE__02000000 0x02000000
1739 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1740 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1741 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1742 * See memalloc_pin_save() */
1743 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1744 #define PF__HOLE__40000000 0x40000000
1745 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1746
1747 /*
1748 * Only the _current_ task can read/write to tsk->flags, but other
1749 * tasks can access tsk->flags in readonly mode for example
1750 * with tsk_used_math (like during threaded core dumping).
1751 * There is however an exception to this rule during ptrace
1752 * or during fork: the ptracer task is allowed to write to the
1753 * child->flags of its traced child (same goes for fork, the parent
1754 * can write to the child->flags), because we're guaranteed the
1755 * child is not running and in turn not changing child->flags
1756 * at the same time the parent does it.
1757 */
1758 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1759 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1760 #define clear_used_math() clear_stopped_child_used_math(current)
1761 #define set_used_math() set_stopped_child_used_math(current)
1762
1763 #define conditional_stopped_child_used_math(condition, child) \
1764 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1765
1766 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1767
1768 #define copy_to_stopped_child_used_math(child) \
1769 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1770
1771 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1772 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1773 #define used_math() tsk_used_math(current)
1774
is_percpu_thread(void)1775 static __always_inline bool is_percpu_thread(void)
1776 {
1777 #ifdef CONFIG_SMP
1778 return (current->flags & PF_NO_SETAFFINITY) &&
1779 (current->nr_cpus_allowed == 1);
1780 #else
1781 return true;
1782 #endif
1783 }
1784
1785 /* Per-process atomic flags. */
1786 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1787 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1788 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1789 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1790 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1791 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1792 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1793 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1794
1795 #define TASK_PFA_TEST(name, func) \
1796 static inline bool task_##func(struct task_struct *p) \
1797 { return test_bit(PFA_##name, &p->atomic_flags); }
1798
1799 #define TASK_PFA_SET(name, func) \
1800 static inline void task_set_##func(struct task_struct *p) \
1801 { set_bit(PFA_##name, &p->atomic_flags); }
1802
1803 #define TASK_PFA_CLEAR(name, func) \
1804 static inline void task_clear_##func(struct task_struct *p) \
1805 { clear_bit(PFA_##name, &p->atomic_flags); }
1806
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1807 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1808 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1809
1810 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1811 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1812 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1813
1814 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1815 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1816 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1817
1818 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1819 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1820 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1821
1822 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1823 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1824 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1825
1826 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1827 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1828
1829 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1830 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1831 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1832
1833 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1834 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1835
1836 static inline void
1837 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1838 {
1839 current->flags &= ~flags;
1840 current->flags |= orig_flags & flags;
1841 }
1842
1843 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1844 extern int task_can_attach(struct task_struct *p);
1845 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1846 extern void dl_bw_free(int cpu, u64 dl_bw);
1847 #ifdef CONFIG_SMP
1848
1849 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1850 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1851
1852 /**
1853 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1854 * @p: the task
1855 * @new_mask: CPU affinity mask
1856 *
1857 * Return: zero if successful, or a negative error code
1858 */
1859 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1860 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1861 extern void release_user_cpus_ptr(struct task_struct *p);
1862 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1863 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1864 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1865 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1866 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1867 {
1868 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1869 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1870 {
1871 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1872 if ((*cpumask_bits(new_mask) & 1) == 0)
1873 return -EINVAL;
1874 return 0;
1875 }
dup_user_cpus_ptr(struct task_struct * dst,struct task_struct * src,int node)1876 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1877 {
1878 if (src->user_cpus_ptr)
1879 return -EINVAL;
1880 return 0;
1881 }
release_user_cpus_ptr(struct task_struct * p)1882 static inline void release_user_cpus_ptr(struct task_struct *p)
1883 {
1884 WARN_ON(p->user_cpus_ptr);
1885 }
1886
dl_task_check_affinity(struct task_struct * p,const struct cpumask * mask)1887 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1888 {
1889 return 0;
1890 }
1891 #endif
1892
1893 extern int yield_to(struct task_struct *p, bool preempt);
1894 extern void set_user_nice(struct task_struct *p, long nice);
1895 extern int task_prio(const struct task_struct *p);
1896
1897 /**
1898 * task_nice - return the nice value of a given task.
1899 * @p: the task in question.
1900 *
1901 * Return: The nice value [ -20 ... 0 ... 19 ].
1902 */
task_nice(const struct task_struct * p)1903 static inline int task_nice(const struct task_struct *p)
1904 {
1905 return PRIO_TO_NICE((p)->static_prio);
1906 }
1907
1908 extern int can_nice(const struct task_struct *p, const int nice);
1909 extern int task_curr(const struct task_struct *p);
1910 extern int idle_cpu(int cpu);
1911 extern int available_idle_cpu(int cpu);
1912 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1913 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1914 extern void sched_set_fifo(struct task_struct *p);
1915 extern void sched_set_fifo_low(struct task_struct *p);
1916 extern void sched_set_normal(struct task_struct *p, int nice);
1917 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1918 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1919 extern struct task_struct *idle_task(int cpu);
1920
1921 /**
1922 * is_idle_task - is the specified task an idle task?
1923 * @p: the task in question.
1924 *
1925 * Return: 1 if @p is an idle task. 0 otherwise.
1926 */
is_idle_task(const struct task_struct * p)1927 static __always_inline bool is_idle_task(const struct task_struct *p)
1928 {
1929 return !!(p->flags & PF_IDLE);
1930 }
1931
1932 extern struct task_struct *curr_task(int cpu);
1933 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1934
1935 void yield(void);
1936
1937 union thread_union {
1938 struct task_struct task;
1939 #ifndef CONFIG_THREAD_INFO_IN_TASK
1940 struct thread_info thread_info;
1941 #endif
1942 unsigned long stack[THREAD_SIZE/sizeof(long)];
1943 };
1944
1945 #ifndef CONFIG_THREAD_INFO_IN_TASK
1946 extern struct thread_info init_thread_info;
1947 #endif
1948
1949 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1950
1951 #ifdef CONFIG_THREAD_INFO_IN_TASK
1952 # define task_thread_info(task) (&(task)->thread_info)
1953 #else
1954 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1955 #endif
1956
1957 /*
1958 * find a task by one of its numerical ids
1959 *
1960 * find_task_by_pid_ns():
1961 * finds a task by its pid in the specified namespace
1962 * find_task_by_vpid():
1963 * finds a task by its virtual pid
1964 *
1965 * see also find_vpid() etc in include/linux/pid.h
1966 */
1967
1968 extern struct task_struct *find_task_by_vpid(pid_t nr);
1969 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1970
1971 /*
1972 * find a task by its virtual pid and get the task struct
1973 */
1974 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1975
1976 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1977 extern int wake_up_process(struct task_struct *tsk);
1978 extern void wake_up_new_task(struct task_struct *tsk);
1979
1980 #ifdef CONFIG_SMP
1981 extern void kick_process(struct task_struct *tsk);
1982 #else
kick_process(struct task_struct * tsk)1983 static inline void kick_process(struct task_struct *tsk) { }
1984 #endif
1985
1986 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1987 #define set_task_comm(tsk, from) ({ \
1988 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \
1989 __set_task_comm(tsk, from, false); \
1990 })
1991
1992 /*
1993 * - Why not use task_lock()?
1994 * User space can randomly change their names anyway, so locking for readers
1995 * doesn't make sense. For writers, locking is probably necessary, as a race
1996 * condition could lead to long-term mixed results.
1997 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is
1998 * always NUL-terminated and zero-padded. Therefore the race condition between
1999 * reader and writer is not an issue.
2000 *
2001 * - BUILD_BUG_ON() can help prevent the buf from being truncated.
2002 * Since the callers don't perform any return value checks, this safeguard is
2003 * necessary.
2004 */
2005 #define get_task_comm(buf, tsk) ({ \
2006 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \
2007 strscpy_pad(buf, (tsk)->comm); \
2008 buf; \
2009 })
2010
2011 #ifdef CONFIG_SMP
scheduler_ipi(void)2012 static __always_inline void scheduler_ipi(void)
2013 {
2014 /*
2015 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2016 * TIF_NEED_RESCHED remotely (for the first time) will also send
2017 * this IPI.
2018 */
2019 preempt_fold_need_resched();
2020 }
2021 #else
scheduler_ipi(void)2022 static inline void scheduler_ipi(void) { }
2023 #endif
2024
2025 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2026
2027 /*
2028 * Set thread flags in other task's structures.
2029 * See asm/thread_info.h for TIF_xxxx flags available:
2030 */
set_tsk_thread_flag(struct task_struct * tsk,int flag)2031 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2032 {
2033 set_ti_thread_flag(task_thread_info(tsk), flag);
2034 }
2035
clear_tsk_thread_flag(struct task_struct * tsk,int flag)2036 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2037 {
2038 clear_ti_thread_flag(task_thread_info(tsk), flag);
2039 }
2040
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)2041 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2042 bool value)
2043 {
2044 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2045 }
2046
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)2047 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2048 {
2049 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2050 }
2051
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)2052 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2053 {
2054 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2055 }
2056
test_tsk_thread_flag(struct task_struct * tsk,int flag)2057 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2058 {
2059 return test_ti_thread_flag(task_thread_info(tsk), flag);
2060 }
2061
set_tsk_need_resched(struct task_struct * tsk)2062 static inline void set_tsk_need_resched(struct task_struct *tsk)
2063 {
2064 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2065 }
2066
clear_tsk_need_resched(struct task_struct * tsk)2067 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2068 {
2069 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2070 (atomic_long_t *)&task_thread_info(tsk)->flags);
2071 }
2072
test_tsk_need_resched(struct task_struct * tsk)2073 static inline int test_tsk_need_resched(struct task_struct *tsk)
2074 {
2075 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2076 }
2077
2078 /*
2079 * cond_resched() and cond_resched_lock(): latency reduction via
2080 * explicit rescheduling in places that are safe. The return
2081 * value indicates whether a reschedule was done in fact.
2082 * cond_resched_lock() will drop the spinlock before scheduling,
2083 */
2084 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2085 extern int __cond_resched(void);
2086
2087 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2088
2089 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2090
_cond_resched(void)2091 static __always_inline int _cond_resched(void)
2092 {
2093 return static_call_mod(cond_resched)();
2094 }
2095
2096 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2097
2098 extern int dynamic_cond_resched(void);
2099
_cond_resched(void)2100 static __always_inline int _cond_resched(void)
2101 {
2102 return dynamic_cond_resched();
2103 }
2104
2105 #else /* !CONFIG_PREEMPTION */
2106
_cond_resched(void)2107 static inline int _cond_resched(void)
2108 {
2109 return __cond_resched();
2110 }
2111
2112 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2113
2114 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2115
_cond_resched(void)2116 static inline int _cond_resched(void)
2117 {
2118 return 0;
2119 }
2120
2121 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2122
2123 #define cond_resched() ({ \
2124 __might_resched(__FILE__, __LINE__, 0); \
2125 _cond_resched(); \
2126 })
2127
2128 extern int __cond_resched_lock(spinlock_t *lock);
2129 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2130 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2131
2132 #define MIGHT_RESCHED_RCU_SHIFT 8
2133 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2134
2135 #ifndef CONFIG_PREEMPT_RT
2136 /*
2137 * Non RT kernels have an elevated preempt count due to the held lock,
2138 * but are not allowed to be inside a RCU read side critical section
2139 */
2140 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2141 #else
2142 /*
2143 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2144 * cond_resched*lock() has to take that into account because it checks for
2145 * preempt_count() and rcu_preempt_depth().
2146 */
2147 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2148 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2149 #endif
2150
2151 #define cond_resched_lock(lock) ({ \
2152 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2153 __cond_resched_lock(lock); \
2154 })
2155
2156 #define cond_resched_rwlock_read(lock) ({ \
2157 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2158 __cond_resched_rwlock_read(lock); \
2159 })
2160
2161 #define cond_resched_rwlock_write(lock) ({ \
2162 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2163 __cond_resched_rwlock_write(lock); \
2164 })
2165
need_resched(void)2166 static __always_inline bool need_resched(void)
2167 {
2168 return unlikely(tif_need_resched());
2169 }
2170
2171 /*
2172 * Wrappers for p->thread_info->cpu access. No-op on UP.
2173 */
2174 #ifdef CONFIG_SMP
2175
task_cpu(const struct task_struct * p)2176 static inline unsigned int task_cpu(const struct task_struct *p)
2177 {
2178 return READ_ONCE(task_thread_info(p)->cpu);
2179 }
2180
2181 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2182
2183 #else
2184
task_cpu(const struct task_struct * p)2185 static inline unsigned int task_cpu(const struct task_struct *p)
2186 {
2187 return 0;
2188 }
2189
set_task_cpu(struct task_struct * p,unsigned int cpu)2190 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2191 {
2192 }
2193
2194 #endif /* CONFIG_SMP */
2195
task_is_runnable(struct task_struct * p)2196 static inline bool task_is_runnable(struct task_struct *p)
2197 {
2198 return p->on_rq && !p->se.sched_delayed;
2199 }
2200
2201 extern bool sched_task_on_rq(struct task_struct *p);
2202 extern unsigned long get_wchan(struct task_struct *p);
2203 extern struct task_struct *cpu_curr_snapshot(int cpu);
2204
2205 #include <linux/spinlock.h>
2206
2207 /*
2208 * In order to reduce various lock holder preemption latencies provide an
2209 * interface to see if a vCPU is currently running or not.
2210 *
2211 * This allows us to terminate optimistic spin loops and block, analogous to
2212 * the native optimistic spin heuristic of testing if the lock owner task is
2213 * running or not.
2214 */
2215 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2216 static inline bool vcpu_is_preempted(int cpu)
2217 {
2218 return false;
2219 }
2220 #endif
2221
2222 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2223 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2224
2225 #ifndef TASK_SIZE_OF
2226 #define TASK_SIZE_OF(tsk) TASK_SIZE
2227 #endif
2228
2229 #ifdef CONFIG_SMP
owner_on_cpu(struct task_struct * owner)2230 static inline bool owner_on_cpu(struct task_struct *owner)
2231 {
2232 /*
2233 * As lock holder preemption issue, we both skip spinning if
2234 * task is not on cpu or its cpu is preempted
2235 */
2236 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2237 }
2238
2239 /* Returns effective CPU energy utilization, as seen by the scheduler */
2240 unsigned long sched_cpu_util(int cpu);
2241 #endif /* CONFIG_SMP */
2242
2243 #ifdef CONFIG_SCHED_CORE
2244 extern void sched_core_free(struct task_struct *tsk);
2245 extern void sched_core_fork(struct task_struct *p);
2246 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2247 unsigned long uaddr);
2248 extern int sched_core_idle_cpu(int cpu);
2249 #else
sched_core_free(struct task_struct * tsk)2250 static inline void sched_core_free(struct task_struct *tsk) { }
sched_core_fork(struct task_struct * p)2251 static inline void sched_core_fork(struct task_struct *p) { }
sched_core_idle_cpu(int cpu)2252 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2253 #endif
2254
2255 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2256
2257 #ifdef CONFIG_MEM_ALLOC_PROFILING
alloc_tag_save(struct alloc_tag * tag)2258 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2259 {
2260 swap(current->alloc_tag, tag);
2261 return tag;
2262 }
2263
alloc_tag_restore(struct alloc_tag * tag,struct alloc_tag * old)2264 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2265 {
2266 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2267 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2268 #endif
2269 current->alloc_tag = old;
2270 }
2271 #else
2272 #define alloc_tag_save(_tag) NULL
2273 #define alloc_tag_restore(_tag, _old) do {} while (0)
2274 #endif
2275
2276 #endif
2277