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/livepatch_sched.h>
48 #include <linux/uidgid_types.h>
49 #include <linux/tracepoint-defs.h>
50 #include <asm/kmap_size.h>
51
52 /* task_struct member predeclarations (sorted alphabetically): */
53 struct audit_context;
54 struct bio_list;
55 struct blk_plug;
56 struct bpf_local_storage;
57 struct bpf_run_ctx;
58 struct bpf_net_context;
59 struct capture_control;
60 struct cfs_rq;
61 struct fs_struct;
62 struct futex_pi_state;
63 struct io_context;
64 struct io_uring_task;
65 struct mempolicy;
66 struct nameidata;
67 struct nsproxy;
68 struct perf_event_context;
69 struct perf_ctx_data;
70 struct pid_namespace;
71 struct pipe_inode_info;
72 struct rcu_node;
73 struct reclaim_state;
74 struct robust_list_head;
75 struct root_domain;
76 struct rq;
77 struct sched_attr;
78 struct sched_dl_entity;
79 struct seq_file;
80 struct sighand_struct;
81 struct signal_struct;
82 struct task_delay_info;
83 struct task_group;
84 struct task_struct;
85 struct user_event_mm;
86
87 #include <linux/sched/ext.h>
88
89 /*
90 * Task state bitmask. NOTE! These bits are also
91 * encoded in fs/proc/array.c: get_task_state().
92 *
93 * We have two separate sets of flags: task->__state
94 * is about runnability, while task->exit_state are
95 * about the task exiting. Confusing, but this way
96 * modifying one set can't modify the other one by
97 * mistake.
98 */
99
100 /* Used in tsk->__state: */
101 #define TASK_RUNNING 0x00000000
102 #define TASK_INTERRUPTIBLE 0x00000001
103 #define TASK_UNINTERRUPTIBLE 0x00000002
104 #define __TASK_STOPPED 0x00000004
105 #define __TASK_TRACED 0x00000008
106 /* Used in tsk->exit_state: */
107 #define EXIT_DEAD 0x00000010
108 #define EXIT_ZOMBIE 0x00000020
109 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
110 /* Used in tsk->__state again: */
111 #define TASK_PARKED 0x00000040
112 #define TASK_DEAD 0x00000080
113 #define TASK_WAKEKILL 0x00000100
114 #define TASK_WAKING 0x00000200
115 #define TASK_NOLOAD 0x00000400
116 #define TASK_NEW 0x00000800
117 #define TASK_RTLOCK_WAIT 0x00001000
118 #define TASK_FREEZABLE 0x00002000
119 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
120 #define TASK_FROZEN 0x00008000
121 #define TASK_STATE_MAX 0x00010000
122
123 #define TASK_ANY (TASK_STATE_MAX-1)
124
125 /*
126 * DO NOT ADD ANY NEW USERS !
127 */
128 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
129
130 /* Convenience macros for the sake of set_current_state: */
131 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
132 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
133 #define TASK_TRACED __TASK_TRACED
134
135 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
136
137 /* Convenience macros for the sake of wake_up(): */
138 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
139
140 /* get_task_state(): */
141 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
142 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
143 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
144 TASK_PARKED)
145
146 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
147
148 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
149 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
150 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
151
152 /*
153 * Special states are those that do not use the normal wait-loop pattern. See
154 * the comment with set_special_state().
155 */
156 #define is_special_task_state(state) \
157 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \
158 TASK_DEAD | TASK_FROZEN))
159
160 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
161 # define debug_normal_state_change(state_value) \
162 do { \
163 WARN_ON_ONCE(is_special_task_state(state_value)); \
164 current->task_state_change = _THIS_IP_; \
165 } while (0)
166
167 # define debug_special_state_change(state_value) \
168 do { \
169 WARN_ON_ONCE(!is_special_task_state(state_value)); \
170 current->task_state_change = _THIS_IP_; \
171 } while (0)
172
173 # define debug_rtlock_wait_set_state() \
174 do { \
175 current->saved_state_change = current->task_state_change;\
176 current->task_state_change = _THIS_IP_; \
177 } while (0)
178
179 # define debug_rtlock_wait_restore_state() \
180 do { \
181 current->task_state_change = current->saved_state_change;\
182 } while (0)
183
184 #else
185 # define debug_normal_state_change(cond) do { } while (0)
186 # define debug_special_state_change(cond) do { } while (0)
187 # define debug_rtlock_wait_set_state() do { } while (0)
188 # define debug_rtlock_wait_restore_state() do { } while (0)
189 #endif
190
191 #define trace_set_current_state(state_value) \
192 do { \
193 if (tracepoint_enabled(sched_set_state_tp)) \
194 __trace_set_current_state(state_value); \
195 } while (0)
196
197 /*
198 * set_current_state() includes a barrier so that the write of current->__state
199 * is correctly serialised wrt the caller's subsequent test of whether to
200 * actually sleep:
201 *
202 * for (;;) {
203 * set_current_state(TASK_UNINTERRUPTIBLE);
204 * if (CONDITION)
205 * break;
206 *
207 * schedule();
208 * }
209 * __set_current_state(TASK_RUNNING);
210 *
211 * If the caller does not need such serialisation (because, for instance, the
212 * CONDITION test and condition change and wakeup are under the same lock) then
213 * use __set_current_state().
214 *
215 * The above is typically ordered against the wakeup, which does:
216 *
217 * CONDITION = 1;
218 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
219 *
220 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
221 * accessing p->__state.
222 *
223 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
224 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
225 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
226 *
227 * However, with slightly different timing the wakeup TASK_RUNNING store can
228 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
229 * a problem either because that will result in one extra go around the loop
230 * and our @cond test will save the day.
231 *
232 * Also see the comments of try_to_wake_up().
233 */
234 #define __set_current_state(state_value) \
235 do { \
236 debug_normal_state_change((state_value)); \
237 trace_set_current_state(state_value); \
238 WRITE_ONCE(current->__state, (state_value)); \
239 } while (0)
240
241 #define set_current_state(state_value) \
242 do { \
243 debug_normal_state_change((state_value)); \
244 trace_set_current_state(state_value); \
245 smp_store_mb(current->__state, (state_value)); \
246 } while (0)
247
248 /*
249 * set_special_state() should be used for those states when the blocking task
250 * can not use the regular condition based wait-loop. In that case we must
251 * serialize against wakeups such that any possible in-flight TASK_RUNNING
252 * stores will not collide with our state change.
253 */
254 #define set_special_state(state_value) \
255 do { \
256 unsigned long flags; /* may shadow */ \
257 \
258 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
259 debug_special_state_change((state_value)); \
260 trace_set_current_state(state_value); \
261 WRITE_ONCE(current->__state, (state_value)); \
262 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
263 } while (0)
264
265 /*
266 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
267 *
268 * RT's spin/rwlock substitutions are state preserving. The state of the
269 * task when blocking on the lock is saved in task_struct::saved_state and
270 * restored after the lock has been acquired. These operations are
271 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
272 * lock related wakeups while the task is blocked on the lock are
273 * redirected to operate on task_struct::saved_state to ensure that these
274 * are not dropped. On restore task_struct::saved_state is set to
275 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
276 *
277 * The lock operation looks like this:
278 *
279 * current_save_and_set_rtlock_wait_state();
280 * for (;;) {
281 * if (try_lock())
282 * break;
283 * raw_spin_unlock_irq(&lock->wait_lock);
284 * schedule_rtlock();
285 * raw_spin_lock_irq(&lock->wait_lock);
286 * set_current_state(TASK_RTLOCK_WAIT);
287 * }
288 * current_restore_rtlock_saved_state();
289 */
290 #define current_save_and_set_rtlock_wait_state() \
291 do { \
292 lockdep_assert_irqs_disabled(); \
293 raw_spin_lock(¤t->pi_lock); \
294 current->saved_state = current->__state; \
295 debug_rtlock_wait_set_state(); \
296 trace_set_current_state(TASK_RTLOCK_WAIT); \
297 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
298 raw_spin_unlock(¤t->pi_lock); \
299 } while (0);
300
301 #define current_restore_rtlock_saved_state() \
302 do { \
303 lockdep_assert_irqs_disabled(); \
304 raw_spin_lock(¤t->pi_lock); \
305 debug_rtlock_wait_restore_state(); \
306 trace_set_current_state(current->saved_state); \
307 WRITE_ONCE(current->__state, current->saved_state); \
308 current->saved_state = TASK_RUNNING; \
309 raw_spin_unlock(¤t->pi_lock); \
310 } while (0);
311
312 #define get_current_state() READ_ONCE(current->__state)
313
314 /*
315 * Define the task command name length as enum, then it can be visible to
316 * BPF programs.
317 */
318 enum {
319 TASK_COMM_LEN = 16,
320 };
321
322 extern void sched_tick(void);
323
324 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
325
326 extern long schedule_timeout(long timeout);
327 extern long schedule_timeout_interruptible(long timeout);
328 extern long schedule_timeout_killable(long timeout);
329 extern long schedule_timeout_uninterruptible(long timeout);
330 extern long schedule_timeout_idle(long timeout);
331 asmlinkage void schedule(void);
332 extern void schedule_preempt_disabled(void);
333 asmlinkage void preempt_schedule_irq(void);
334 #ifdef CONFIG_PREEMPT_RT
335 extern void schedule_rtlock(void);
336 #endif
337
338 extern int __must_check io_schedule_prepare(void);
339 extern void io_schedule_finish(int token);
340 extern long io_schedule_timeout(long timeout);
341 extern void io_schedule(void);
342
343 /* wrapper function to trace from this header file */
344 DECLARE_TRACEPOINT(sched_set_state_tp);
345 extern void __trace_set_current_state(int state_value);
346
347 /**
348 * struct prev_cputime - snapshot of system and user cputime
349 * @utime: time spent in user mode
350 * @stime: time spent in system mode
351 * @lock: protects the above two fields
352 *
353 * Stores previous user/system time values such that we can guarantee
354 * monotonicity.
355 */
356 struct prev_cputime {
357 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
358 u64 utime;
359 u64 stime;
360 raw_spinlock_t lock;
361 #endif
362 };
363
364 enum vtime_state {
365 /* Task is sleeping or running in a CPU with VTIME inactive: */
366 VTIME_INACTIVE = 0,
367 /* Task is idle */
368 VTIME_IDLE,
369 /* Task runs in kernelspace in a CPU with VTIME active: */
370 VTIME_SYS,
371 /* Task runs in userspace in a CPU with VTIME active: */
372 VTIME_USER,
373 /* Task runs as guests in a CPU with VTIME active: */
374 VTIME_GUEST,
375 };
376
377 struct vtime {
378 seqcount_t seqcount;
379 unsigned long long starttime;
380 enum vtime_state state;
381 unsigned int cpu;
382 u64 utime;
383 u64 stime;
384 u64 gtime;
385 };
386
387 /*
388 * Utilization clamp constraints.
389 * @UCLAMP_MIN: Minimum utilization
390 * @UCLAMP_MAX: Maximum utilization
391 * @UCLAMP_CNT: Utilization clamp constraints count
392 */
393 enum uclamp_id {
394 UCLAMP_MIN = 0,
395 UCLAMP_MAX,
396 UCLAMP_CNT
397 };
398
399 #ifdef CONFIG_SMP
400 extern struct root_domain def_root_domain;
401 extern struct mutex sched_domains_mutex;
402 extern void sched_domains_mutex_lock(void);
403 extern void sched_domains_mutex_unlock(void);
404 #else
sched_domains_mutex_lock(void)405 static inline void sched_domains_mutex_lock(void) { }
sched_domains_mutex_unlock(void)406 static inline void sched_domains_mutex_unlock(void) { }
407 #endif
408
409 struct sched_param {
410 int sched_priority;
411 };
412
413 struct sched_info {
414 #ifdef CONFIG_SCHED_INFO
415 /* Cumulative counters: */
416
417 /* # of times we have run on this CPU: */
418 unsigned long pcount;
419
420 /* Time spent waiting on a runqueue: */
421 unsigned long long run_delay;
422
423 /* Max time spent waiting on a runqueue: */
424 unsigned long long max_run_delay;
425
426 /* Min time spent waiting on a runqueue: */
427 unsigned long long min_run_delay;
428
429 /* Timestamps: */
430
431 /* When did we last run on a CPU? */
432 unsigned long long last_arrival;
433
434 /* When were we last queued to run? */
435 unsigned long long last_queued;
436
437 #endif /* CONFIG_SCHED_INFO */
438 };
439
440 /*
441 * Integer metrics need fixed point arithmetic, e.g., sched/fair
442 * has a few: load, load_avg, util_avg, freq, and capacity.
443 *
444 * We define a basic fixed point arithmetic range, and then formalize
445 * all these metrics based on that basic range.
446 */
447 # define SCHED_FIXEDPOINT_SHIFT 10
448 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
449
450 /* Increase resolution of cpu_capacity calculations */
451 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
452 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
453
454 struct load_weight {
455 unsigned long weight;
456 u32 inv_weight;
457 };
458
459 /*
460 * The load/runnable/util_avg accumulates an infinite geometric series
461 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
462 *
463 * [load_avg definition]
464 *
465 * load_avg = runnable% * scale_load_down(load)
466 *
467 * [runnable_avg definition]
468 *
469 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
470 *
471 * [util_avg definition]
472 *
473 * util_avg = running% * SCHED_CAPACITY_SCALE
474 *
475 * where runnable% is the time ratio that a sched_entity is runnable and
476 * running% the time ratio that a sched_entity is running.
477 *
478 * For cfs_rq, they are the aggregated values of all runnable and blocked
479 * sched_entities.
480 *
481 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
482 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
483 * for computing those signals (see update_rq_clock_pelt())
484 *
485 * N.B., the above ratios (runnable% and running%) themselves are in the
486 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
487 * to as large a range as necessary. This is for example reflected by
488 * util_avg's SCHED_CAPACITY_SCALE.
489 *
490 * [Overflow issue]
491 *
492 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
493 * with the highest load (=88761), always runnable on a single cfs_rq,
494 * and should not overflow as the number already hits PID_MAX_LIMIT.
495 *
496 * For all other cases (including 32-bit kernels), struct load_weight's
497 * weight will overflow first before we do, because:
498 *
499 * Max(load_avg) <= Max(load.weight)
500 *
501 * Then it is the load_weight's responsibility to consider overflow
502 * issues.
503 */
504 struct sched_avg {
505 u64 last_update_time;
506 u64 load_sum;
507 u64 runnable_sum;
508 u32 util_sum;
509 u32 period_contrib;
510 unsigned long load_avg;
511 unsigned long runnable_avg;
512 unsigned long util_avg;
513 unsigned int util_est;
514 } ____cacheline_aligned;
515
516 /*
517 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
518 * updates. When a task is dequeued, its util_est should not be updated if its
519 * util_avg has not been updated in the meantime.
520 * This information is mapped into the MSB bit of util_est at dequeue time.
521 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
522 * it is safe to use MSB.
523 */
524 #define UTIL_EST_WEIGHT_SHIFT 2
525 #define UTIL_AVG_UNCHANGED 0x80000000
526
527 struct sched_statistics {
528 #ifdef CONFIG_SCHEDSTATS
529 u64 wait_start;
530 u64 wait_max;
531 u64 wait_count;
532 u64 wait_sum;
533 u64 iowait_count;
534 u64 iowait_sum;
535
536 u64 sleep_start;
537 u64 sleep_max;
538 s64 sum_sleep_runtime;
539
540 u64 block_start;
541 u64 block_max;
542 s64 sum_block_runtime;
543
544 s64 exec_max;
545 u64 slice_max;
546
547 u64 nr_migrations_cold;
548 u64 nr_failed_migrations_affine;
549 u64 nr_failed_migrations_running;
550 u64 nr_failed_migrations_hot;
551 u64 nr_forced_migrations;
552
553 u64 nr_wakeups;
554 u64 nr_wakeups_sync;
555 u64 nr_wakeups_migrate;
556 u64 nr_wakeups_local;
557 u64 nr_wakeups_remote;
558 u64 nr_wakeups_affine;
559 u64 nr_wakeups_affine_attempts;
560 u64 nr_wakeups_passive;
561 u64 nr_wakeups_idle;
562
563 #ifdef CONFIG_SCHED_CORE
564 u64 core_forceidle_sum;
565 #endif
566 #endif /* CONFIG_SCHEDSTATS */
567 } ____cacheline_aligned;
568
569 struct sched_entity {
570 /* For load-balancing: */
571 struct load_weight load;
572 struct rb_node run_node;
573 u64 deadline;
574 u64 min_vruntime;
575 u64 min_slice;
576
577 struct list_head group_node;
578 unsigned char on_rq;
579 unsigned char sched_delayed;
580 unsigned char rel_deadline;
581 unsigned char custom_slice;
582 /* hole */
583
584 u64 exec_start;
585 u64 sum_exec_runtime;
586 u64 prev_sum_exec_runtime;
587 u64 vruntime;
588 s64 vlag;
589 u64 slice;
590
591 u64 nr_migrations;
592
593 #ifdef CONFIG_FAIR_GROUP_SCHED
594 int depth;
595 struct sched_entity *parent;
596 /* rq on which this entity is (to be) queued: */
597 struct cfs_rq *cfs_rq;
598 /* rq "owned" by this entity/group: */
599 struct cfs_rq *my_q;
600 /* cached value of my_q->h_nr_running */
601 unsigned long runnable_weight;
602 #endif
603
604 #ifdef CONFIG_SMP
605 /*
606 * Per entity load average tracking.
607 *
608 * Put into separate cache line so it does not
609 * collide with read-mostly values above.
610 */
611 struct sched_avg avg;
612 #endif
613 };
614
615 struct sched_rt_entity {
616 struct list_head run_list;
617 unsigned long timeout;
618 unsigned long watchdog_stamp;
619 unsigned int time_slice;
620 unsigned short on_rq;
621 unsigned short on_list;
622
623 struct sched_rt_entity *back;
624 #ifdef CONFIG_RT_GROUP_SCHED
625 struct sched_rt_entity *parent;
626 /* rq on which this entity is (to be) queued: */
627 struct rt_rq *rt_rq;
628 /* rq "owned" by this entity/group: */
629 struct rt_rq *my_q;
630 #endif
631 } __randomize_layout;
632
633 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
634 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
635
636 struct sched_dl_entity {
637 struct rb_node rb_node;
638
639 /*
640 * Original scheduling parameters. Copied here from sched_attr
641 * during sched_setattr(), they will remain the same until
642 * the next sched_setattr().
643 */
644 u64 dl_runtime; /* Maximum runtime for each instance */
645 u64 dl_deadline; /* Relative deadline of each instance */
646 u64 dl_period; /* Separation of two instances (period) */
647 u64 dl_bw; /* dl_runtime / dl_period */
648 u64 dl_density; /* dl_runtime / dl_deadline */
649
650 /*
651 * Actual scheduling parameters. Initialized with the values above,
652 * they are continuously updated during task execution. Note that
653 * the remaining runtime could be < 0 in case we are in overrun.
654 */
655 s64 runtime; /* Remaining runtime for this instance */
656 u64 deadline; /* Absolute deadline for this instance */
657 unsigned int flags; /* Specifying the scheduler behaviour */
658
659 /*
660 * Some bool flags:
661 *
662 * @dl_throttled tells if we exhausted the runtime. If so, the
663 * task has to wait for a replenishment to be performed at the
664 * next firing of dl_timer.
665 *
666 * @dl_yielded tells if task gave up the CPU before consuming
667 * all its available runtime during the last job.
668 *
669 * @dl_non_contending tells if the task is inactive while still
670 * contributing to the active utilization. In other words, it
671 * indicates if the inactive timer has been armed and its handler
672 * has not been executed yet. This flag is useful to avoid race
673 * conditions between the inactive timer handler and the wakeup
674 * code.
675 *
676 * @dl_overrun tells if the task asked to be informed about runtime
677 * overruns.
678 *
679 * @dl_server tells if this is a server entity.
680 *
681 * @dl_defer tells if this is a deferred or regular server. For
682 * now only defer server exists.
683 *
684 * @dl_defer_armed tells if the deferrable server is waiting
685 * for the replenishment timer to activate it.
686 *
687 * @dl_server_active tells if the dlserver is active(started).
688 * dlserver is started on first cfs enqueue on an idle runqueue
689 * and is stopped when a dequeue results in 0 cfs tasks on the
690 * runqueue. In other words, dlserver is active only when cpu's
691 * runqueue has atleast one cfs task.
692 *
693 * @dl_defer_running tells if the deferrable server is actually
694 * running, skipping the defer phase.
695 */
696 unsigned int dl_throttled : 1;
697 unsigned int dl_yielded : 1;
698 unsigned int dl_non_contending : 1;
699 unsigned int dl_overrun : 1;
700 unsigned int dl_server : 1;
701 unsigned int dl_server_active : 1;
702 unsigned int dl_defer : 1;
703 unsigned int dl_defer_armed : 1;
704 unsigned int dl_defer_running : 1;
705
706 /*
707 * Bandwidth enforcement timer. Each -deadline task has its
708 * own bandwidth to be enforced, thus we need one timer per task.
709 */
710 struct hrtimer dl_timer;
711
712 /*
713 * Inactive timer, responsible for decreasing the active utilization
714 * at the "0-lag time". When a -deadline task blocks, it contributes
715 * to GRUB's active utilization until the "0-lag time", hence a
716 * timer is needed to decrease the active utilization at the correct
717 * time.
718 */
719 struct hrtimer inactive_timer;
720
721 /*
722 * Bits for DL-server functionality. Also see the comment near
723 * dl_server_update().
724 *
725 * @rq the runqueue this server is for
726 *
727 * @server_has_tasks() returns true if @server_pick return a
728 * runnable task.
729 */
730 struct rq *rq;
731 dl_server_has_tasks_f server_has_tasks;
732 dl_server_pick_f server_pick_task;
733
734 #ifdef CONFIG_RT_MUTEXES
735 /*
736 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
737 * pi_se points to the donor, otherwise points to the dl_se it belongs
738 * to (the original one/itself).
739 */
740 struct sched_dl_entity *pi_se;
741 #endif
742 };
743
744 #ifdef CONFIG_UCLAMP_TASK
745 /* Number of utilization clamp buckets (shorter alias) */
746 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
747
748 /*
749 * Utilization clamp for a scheduling entity
750 * @value: clamp value "assigned" to a se
751 * @bucket_id: bucket index corresponding to the "assigned" value
752 * @active: the se is currently refcounted in a rq's bucket
753 * @user_defined: the requested clamp value comes from user-space
754 *
755 * The bucket_id is the index of the clamp bucket matching the clamp value
756 * which is pre-computed and stored to avoid expensive integer divisions from
757 * the fast path.
758 *
759 * The active bit is set whenever a task has got an "effective" value assigned,
760 * which can be different from the clamp value "requested" from user-space.
761 * This allows to know a task is refcounted in the rq's bucket corresponding
762 * to the "effective" bucket_id.
763 *
764 * The user_defined bit is set whenever a task has got a task-specific clamp
765 * value requested from userspace, i.e. the system defaults apply to this task
766 * just as a restriction. This allows to relax default clamps when a less
767 * restrictive task-specific value has been requested, thus allowing to
768 * implement a "nice" semantic. For example, a task running with a 20%
769 * default boost can still drop its own boosting to 0%.
770 */
771 struct uclamp_se {
772 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
773 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
774 unsigned int active : 1;
775 unsigned int user_defined : 1;
776 };
777 #endif /* CONFIG_UCLAMP_TASK */
778
779 union rcu_special {
780 struct {
781 u8 blocked;
782 u8 need_qs;
783 u8 exp_hint; /* Hint for performance. */
784 u8 need_mb; /* Readers need smp_mb(). */
785 } b; /* Bits. */
786 u32 s; /* Set of bits. */
787 };
788
789 enum perf_event_task_context {
790 perf_invalid_context = -1,
791 perf_hw_context = 0,
792 perf_sw_context,
793 perf_nr_task_contexts,
794 };
795
796 /*
797 * Number of contexts where an event can trigger:
798 * task, softirq, hardirq, nmi.
799 */
800 #define PERF_NR_CONTEXTS 4
801
802 struct wake_q_node {
803 struct wake_q_node *next;
804 };
805
806 struct kmap_ctrl {
807 #ifdef CONFIG_KMAP_LOCAL
808 int idx;
809 pte_t pteval[KM_MAX_IDX];
810 #endif
811 };
812
813 struct task_struct {
814 #ifdef CONFIG_THREAD_INFO_IN_TASK
815 /*
816 * For reasons of header soup (see current_thread_info()), this
817 * must be the first element of task_struct.
818 */
819 struct thread_info thread_info;
820 #endif
821 unsigned int __state;
822
823 /* saved state for "spinlock sleepers" */
824 unsigned int saved_state;
825
826 /*
827 * This begins the randomizable portion of task_struct. Only
828 * scheduling-critical items should be added above here.
829 */
830 randomized_struct_fields_start
831
832 void *stack;
833 refcount_t usage;
834 /* Per task flags (PF_*), defined further below: */
835 unsigned int flags;
836 unsigned int ptrace;
837
838 #ifdef CONFIG_MEM_ALLOC_PROFILING
839 struct alloc_tag *alloc_tag;
840 #endif
841
842 #ifdef CONFIG_SMP
843 int on_cpu;
844 struct __call_single_node wake_entry;
845 unsigned int wakee_flips;
846 unsigned long wakee_flip_decay_ts;
847 struct task_struct *last_wakee;
848
849 /*
850 * recent_used_cpu is initially set as the last CPU used by a task
851 * that wakes affine another task. Waker/wakee relationships can
852 * push tasks around a CPU where each wakeup moves to the next one.
853 * Tracking a recently used CPU allows a quick search for a recently
854 * used CPU that may be idle.
855 */
856 int recent_used_cpu;
857 int wake_cpu;
858 #endif
859 int on_rq;
860
861 int prio;
862 int static_prio;
863 int normal_prio;
864 unsigned int rt_priority;
865
866 struct sched_entity se;
867 struct sched_rt_entity rt;
868 struct sched_dl_entity dl;
869 struct sched_dl_entity *dl_server;
870 #ifdef CONFIG_SCHED_CLASS_EXT
871 struct sched_ext_entity scx;
872 #endif
873 const struct sched_class *sched_class;
874
875 #ifdef CONFIG_SCHED_CORE
876 struct rb_node core_node;
877 unsigned long core_cookie;
878 unsigned int core_occupation;
879 #endif
880
881 #ifdef CONFIG_CGROUP_SCHED
882 struct task_group *sched_task_group;
883 #endif
884
885
886 #ifdef CONFIG_UCLAMP_TASK
887 /*
888 * Clamp values requested for a scheduling entity.
889 * Must be updated with task_rq_lock() held.
890 */
891 struct uclamp_se uclamp_req[UCLAMP_CNT];
892 /*
893 * Effective clamp values used for a scheduling entity.
894 * Must be updated with task_rq_lock() held.
895 */
896 struct uclamp_se uclamp[UCLAMP_CNT];
897 #endif
898
899 struct sched_statistics stats;
900
901 #ifdef CONFIG_PREEMPT_NOTIFIERS
902 /* List of struct preempt_notifier: */
903 struct hlist_head preempt_notifiers;
904 #endif
905
906 #ifdef CONFIG_BLK_DEV_IO_TRACE
907 unsigned int btrace_seq;
908 #endif
909
910 unsigned int policy;
911 unsigned long max_allowed_capacity;
912 int nr_cpus_allowed;
913 const cpumask_t *cpus_ptr;
914 cpumask_t *user_cpus_ptr;
915 cpumask_t cpus_mask;
916 void *migration_pending;
917 #ifdef CONFIG_SMP
918 unsigned short migration_disabled;
919 #endif
920 unsigned short migration_flags;
921
922 #ifdef CONFIG_PREEMPT_RCU
923 int rcu_read_lock_nesting;
924 union rcu_special rcu_read_unlock_special;
925 struct list_head rcu_node_entry;
926 struct rcu_node *rcu_blocked_node;
927 #endif /* #ifdef CONFIG_PREEMPT_RCU */
928
929 #ifdef CONFIG_TASKS_RCU
930 unsigned long rcu_tasks_nvcsw;
931 u8 rcu_tasks_holdout;
932 u8 rcu_tasks_idx;
933 int rcu_tasks_idle_cpu;
934 struct list_head rcu_tasks_holdout_list;
935 int rcu_tasks_exit_cpu;
936 struct list_head rcu_tasks_exit_list;
937 #endif /* #ifdef CONFIG_TASKS_RCU */
938
939 #ifdef CONFIG_TASKS_TRACE_RCU
940 int trc_reader_nesting;
941 int trc_ipi_to_cpu;
942 union rcu_special trc_reader_special;
943 struct list_head trc_holdout_list;
944 struct list_head trc_blkd_node;
945 int trc_blkd_cpu;
946 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
947
948 struct sched_info sched_info;
949
950 struct list_head tasks;
951 #ifdef CONFIG_SMP
952 struct plist_node pushable_tasks;
953 struct rb_node pushable_dl_tasks;
954 #endif
955
956 struct mm_struct *mm;
957 struct mm_struct *active_mm;
958 struct address_space *faults_disabled_mapping;
959
960 int exit_state;
961 int exit_code;
962 int exit_signal;
963 /* The signal sent when the parent dies: */
964 int pdeath_signal;
965 /* JOBCTL_*, siglock protected: */
966 unsigned long jobctl;
967
968 /* Used for emulating ABI behavior of previous Linux versions: */
969 unsigned int personality;
970
971 /* Scheduler bits, serialized by scheduler locks: */
972 unsigned sched_reset_on_fork:1;
973 unsigned sched_contributes_to_load:1;
974 unsigned sched_migrated:1;
975 unsigned sched_task_hot:1;
976
977 /* Force alignment to the next boundary: */
978 unsigned :0;
979
980 /* Unserialized, strictly 'current' */
981
982 /*
983 * This field must not be in the scheduler word above due to wakelist
984 * queueing no longer being serialized by p->on_cpu. However:
985 *
986 * p->XXX = X; ttwu()
987 * schedule() if (p->on_rq && ..) // false
988 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
989 * deactivate_task() ttwu_queue_wakelist())
990 * p->on_rq = 0; p->sched_remote_wakeup = Y;
991 *
992 * guarantees all stores of 'current' are visible before
993 * ->sched_remote_wakeup gets used, so it can be in this word.
994 */
995 unsigned sched_remote_wakeup:1;
996 #ifdef CONFIG_RT_MUTEXES
997 unsigned sched_rt_mutex:1;
998 #endif
999
1000 /* Bit to tell TOMOYO we're in execve(): */
1001 unsigned in_execve:1;
1002 unsigned in_iowait:1;
1003 #ifndef TIF_RESTORE_SIGMASK
1004 unsigned restore_sigmask:1;
1005 #endif
1006 #ifdef CONFIG_MEMCG_V1
1007 unsigned in_user_fault:1;
1008 #endif
1009 #ifdef CONFIG_LRU_GEN
1010 /* whether the LRU algorithm may apply to this access */
1011 unsigned in_lru_fault:1;
1012 #endif
1013 #ifdef CONFIG_COMPAT_BRK
1014 unsigned brk_randomized:1;
1015 #endif
1016 #ifdef CONFIG_CGROUPS
1017 /* disallow userland-initiated cgroup migration */
1018 unsigned no_cgroup_migration:1;
1019 /* task is frozen/stopped (used by the cgroup freezer) */
1020 unsigned frozen:1;
1021 #endif
1022 #ifdef CONFIG_BLK_CGROUP
1023 unsigned use_memdelay:1;
1024 #endif
1025 #ifdef CONFIG_PSI
1026 /* Stalled due to lack of memory */
1027 unsigned in_memstall:1;
1028 #endif
1029 #ifdef CONFIG_PAGE_OWNER
1030 /* Used by page_owner=on to detect recursion in page tracking. */
1031 unsigned in_page_owner:1;
1032 #endif
1033 #ifdef CONFIG_EVENTFD
1034 /* Recursion prevention for eventfd_signal() */
1035 unsigned in_eventfd:1;
1036 #endif
1037 #ifdef CONFIG_ARCH_HAS_CPU_PASID
1038 unsigned pasid_activated:1;
1039 #endif
1040 #ifdef CONFIG_X86_BUS_LOCK_DETECT
1041 unsigned reported_split_lock:1;
1042 #endif
1043 #ifdef CONFIG_TASK_DELAY_ACCT
1044 /* delay due to memory thrashing */
1045 unsigned in_thrashing:1;
1046 #endif
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_DEBUG_ATOMIC_SLEEP
1243 int non_block_count;
1244 #endif
1245
1246 #ifdef CONFIG_TRACE_IRQFLAGS
1247 struct irqtrace_events irqtrace;
1248 unsigned int hardirq_threaded;
1249 u64 hardirq_chain_key;
1250 int softirqs_enabled;
1251 int softirq_context;
1252 int irq_config;
1253 #endif
1254 #ifdef CONFIG_PREEMPT_RT
1255 int softirq_disable_cnt;
1256 #endif
1257
1258 #ifdef CONFIG_LOCKDEP
1259 # define MAX_LOCK_DEPTH 48UL
1260 u64 curr_chain_key;
1261 int lockdep_depth;
1262 unsigned int lockdep_recursion;
1263 struct held_lock held_locks[MAX_LOCK_DEPTH];
1264 #endif
1265
1266 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1267 unsigned int in_ubsan;
1268 #endif
1269
1270 /* Journalling filesystem info: */
1271 void *journal_info;
1272
1273 /* Stacked block device info: */
1274 struct bio_list *bio_list;
1275
1276 /* Stack plugging: */
1277 struct blk_plug *plug;
1278
1279 /* VM state: */
1280 struct reclaim_state *reclaim_state;
1281
1282 struct io_context *io_context;
1283
1284 #ifdef CONFIG_COMPACTION
1285 struct capture_control *capture_control;
1286 #endif
1287 /* Ptrace state: */
1288 unsigned long ptrace_message;
1289 kernel_siginfo_t *last_siginfo;
1290
1291 struct task_io_accounting ioac;
1292 #ifdef CONFIG_PSI
1293 /* Pressure stall state */
1294 unsigned int psi_flags;
1295 #endif
1296 #ifdef CONFIG_TASK_XACCT
1297 /* Accumulated RSS usage: */
1298 u64 acct_rss_mem1;
1299 /* Accumulated virtual memory usage: */
1300 u64 acct_vm_mem1;
1301 /* stime + utime since last update: */
1302 u64 acct_timexpd;
1303 #endif
1304 #ifdef CONFIG_CPUSETS
1305 /* Protected by ->alloc_lock: */
1306 nodemask_t mems_allowed;
1307 /* Sequence number to catch updates: */
1308 seqcount_spinlock_t mems_allowed_seq;
1309 int cpuset_mem_spread_rotor;
1310 #endif
1311 #ifdef CONFIG_CGROUPS
1312 /* Control Group info protected by css_set_lock: */
1313 struct css_set __rcu *cgroups;
1314 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1315 struct list_head cg_list;
1316 #endif
1317 #ifdef CONFIG_X86_CPU_RESCTRL
1318 u32 closid;
1319 u32 rmid;
1320 #endif
1321 #ifdef CONFIG_FUTEX
1322 struct robust_list_head __user *robust_list;
1323 #ifdef CONFIG_COMPAT
1324 struct compat_robust_list_head __user *compat_robust_list;
1325 #endif
1326 struct list_head pi_state_list;
1327 struct futex_pi_state *pi_state_cache;
1328 struct mutex futex_exit_mutex;
1329 unsigned int futex_state;
1330 #endif
1331 #ifdef CONFIG_PERF_EVENTS
1332 u8 perf_recursion[PERF_NR_CONTEXTS];
1333 struct perf_event_context *perf_event_ctxp;
1334 struct mutex perf_event_mutex;
1335 struct list_head perf_event_list;
1336 struct perf_ctx_data __rcu *perf_ctx_data;
1337 #endif
1338 #ifdef CONFIG_DEBUG_PREEMPT
1339 unsigned long preempt_disable_ip;
1340 #endif
1341 #ifdef CONFIG_NUMA
1342 /* Protected by alloc_lock: */
1343 struct mempolicy *mempolicy;
1344 short il_prev;
1345 u8 il_weight;
1346 short pref_node_fork;
1347 #endif
1348 #ifdef CONFIG_NUMA_BALANCING
1349 int numa_scan_seq;
1350 unsigned int numa_scan_period;
1351 unsigned int numa_scan_period_max;
1352 int numa_preferred_nid;
1353 unsigned long numa_migrate_retry;
1354 /* Migration stamp: */
1355 u64 node_stamp;
1356 u64 last_task_numa_placement;
1357 u64 last_sum_exec_runtime;
1358 struct callback_head numa_work;
1359
1360 /*
1361 * This pointer is only modified for current in syscall and
1362 * pagefault context (and for tasks being destroyed), so it can be read
1363 * from any of the following contexts:
1364 * - RCU read-side critical section
1365 * - current->numa_group from everywhere
1366 * - task's runqueue locked, task not running
1367 */
1368 struct numa_group __rcu *numa_group;
1369
1370 /*
1371 * numa_faults is an array split into four regions:
1372 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1373 * in this precise order.
1374 *
1375 * faults_memory: Exponential decaying average of faults on a per-node
1376 * basis. Scheduling placement decisions are made based on these
1377 * counts. The values remain static for the duration of a PTE scan.
1378 * faults_cpu: Track the nodes the process was running on when a NUMA
1379 * hinting fault was incurred.
1380 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1381 * during the current scan window. When the scan completes, the counts
1382 * in faults_memory and faults_cpu decay and these values are copied.
1383 */
1384 unsigned long *numa_faults;
1385 unsigned long total_numa_faults;
1386
1387 /*
1388 * numa_faults_locality tracks if faults recorded during the last
1389 * scan window were remote/local or failed to migrate. The task scan
1390 * period is adapted based on the locality of the faults with different
1391 * weights depending on whether they were shared or private faults
1392 */
1393 unsigned long numa_faults_locality[3];
1394
1395 unsigned long numa_pages_migrated;
1396 #endif /* CONFIG_NUMA_BALANCING */
1397
1398 #ifdef CONFIG_RSEQ
1399 struct rseq __user *rseq;
1400 u32 rseq_len;
1401 u32 rseq_sig;
1402 /*
1403 * RmW on rseq_event_mask must be performed atomically
1404 * with respect to preemption.
1405 */
1406 unsigned long rseq_event_mask;
1407 # ifdef CONFIG_DEBUG_RSEQ
1408 /*
1409 * This is a place holder to save a copy of the rseq fields for
1410 * validation of read-only fields. The struct rseq has a
1411 * variable-length array at the end, so it cannot be used
1412 * directly. Reserve a size large enough for the known fields.
1413 */
1414 char rseq_fields[sizeof(struct rseq)];
1415 # endif
1416 #endif
1417
1418 #ifdef CONFIG_SCHED_MM_CID
1419 int mm_cid; /* Current cid in mm */
1420 int last_mm_cid; /* Most recent cid in mm */
1421 int migrate_from_cpu;
1422 int mm_cid_active; /* Whether cid bitmap is active */
1423 struct callback_head cid_work;
1424 #endif
1425
1426 struct tlbflush_unmap_batch tlb_ubc;
1427
1428 /* Cache last used pipe for splice(): */
1429 struct pipe_inode_info *splice_pipe;
1430
1431 struct page_frag task_frag;
1432
1433 #ifdef CONFIG_TASK_DELAY_ACCT
1434 struct task_delay_info *delays;
1435 #endif
1436
1437 #ifdef CONFIG_FAULT_INJECTION
1438 int make_it_fail;
1439 unsigned int fail_nth;
1440 #endif
1441 /*
1442 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1443 * balance_dirty_pages() for a dirty throttling pause:
1444 */
1445 int nr_dirtied;
1446 int nr_dirtied_pause;
1447 /* Start of a write-and-pause period: */
1448 unsigned long dirty_paused_when;
1449
1450 #ifdef CONFIG_LATENCYTOP
1451 int latency_record_count;
1452 struct latency_record latency_record[LT_SAVECOUNT];
1453 #endif
1454 /*
1455 * Time slack values; these are used to round up poll() and
1456 * select() etc timeout values. These are in nanoseconds.
1457 */
1458 u64 timer_slack_ns;
1459 u64 default_timer_slack_ns;
1460
1461 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1462 unsigned int kasan_depth;
1463 #endif
1464
1465 #ifdef CONFIG_KCSAN
1466 struct kcsan_ctx kcsan_ctx;
1467 #ifdef CONFIG_TRACE_IRQFLAGS
1468 struct irqtrace_events kcsan_save_irqtrace;
1469 #endif
1470 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1471 int kcsan_stack_depth;
1472 #endif
1473 #endif
1474
1475 #ifdef CONFIG_KMSAN
1476 struct kmsan_ctx kmsan_ctx;
1477 #endif
1478
1479 #if IS_ENABLED(CONFIG_KUNIT)
1480 struct kunit *kunit_test;
1481 #endif
1482
1483 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1484 /* Index of current stored address in ret_stack: */
1485 int curr_ret_stack;
1486 int curr_ret_depth;
1487
1488 /* Stack of return addresses for return function tracing: */
1489 unsigned long *ret_stack;
1490
1491 /* Timestamp for last schedule: */
1492 unsigned long long ftrace_timestamp;
1493 unsigned long long ftrace_sleeptime;
1494
1495 /*
1496 * Number of functions that haven't been traced
1497 * because of depth overrun:
1498 */
1499 atomic_t trace_overrun;
1500
1501 /* Pause tracing: */
1502 atomic_t tracing_graph_pause;
1503 #endif
1504
1505 #ifdef CONFIG_TRACING
1506 /* Bitmask and counter of trace recursion: */
1507 unsigned long trace_recursion;
1508 #endif /* CONFIG_TRACING */
1509
1510 #ifdef CONFIG_KCOV
1511 /* See kernel/kcov.c for more details. */
1512
1513 /* Coverage collection mode enabled for this task (0 if disabled): */
1514 unsigned int kcov_mode;
1515
1516 /* Size of the kcov_area: */
1517 unsigned int kcov_size;
1518
1519 /* Buffer for coverage collection: */
1520 void *kcov_area;
1521
1522 /* KCOV descriptor wired with this task or NULL: */
1523 struct kcov *kcov;
1524
1525 /* KCOV common handle for remote coverage collection: */
1526 u64 kcov_handle;
1527
1528 /* KCOV sequence number: */
1529 int kcov_sequence;
1530
1531 /* Collect coverage from softirq context: */
1532 unsigned int kcov_softirq;
1533 #endif
1534
1535 #ifdef CONFIG_MEMCG_V1
1536 struct mem_cgroup *memcg_in_oom;
1537 #endif
1538
1539 #ifdef CONFIG_MEMCG
1540 /* Number of pages to reclaim on returning to userland: */
1541 unsigned int memcg_nr_pages_over_high;
1542
1543 /* Used by memcontrol for targeted memcg charge: */
1544 struct mem_cgroup *active_memcg;
1545
1546 /* Cache for current->cgroups->memcg->objcg lookups: */
1547 struct obj_cgroup *objcg;
1548 #endif
1549
1550 #ifdef CONFIG_BLK_CGROUP
1551 struct gendisk *throttle_disk;
1552 #endif
1553
1554 #ifdef CONFIG_UPROBES
1555 struct uprobe_task *utask;
1556 #endif
1557 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1558 unsigned int sequential_io;
1559 unsigned int sequential_io_avg;
1560 #endif
1561 struct kmap_ctrl kmap_ctrl;
1562 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1563 unsigned long task_state_change;
1564 # ifdef CONFIG_PREEMPT_RT
1565 unsigned long saved_state_change;
1566 # endif
1567 #endif
1568 struct rcu_head rcu;
1569 refcount_t rcu_users;
1570 int pagefault_disabled;
1571 #ifdef CONFIG_MMU
1572 struct task_struct *oom_reaper_list;
1573 struct timer_list oom_reaper_timer;
1574 #endif
1575 #ifdef CONFIG_VMAP_STACK
1576 struct vm_struct *stack_vm_area;
1577 #endif
1578 #ifdef CONFIG_THREAD_INFO_IN_TASK
1579 /* A live task holds one reference: */
1580 refcount_t stack_refcount;
1581 #endif
1582 #ifdef CONFIG_LIVEPATCH
1583 int patch_state;
1584 #endif
1585 #ifdef CONFIG_SECURITY
1586 /* Used by LSM modules for access restriction: */
1587 void *security;
1588 #endif
1589 #ifdef CONFIG_BPF_SYSCALL
1590 /* Used by BPF task local storage */
1591 struct bpf_local_storage __rcu *bpf_storage;
1592 /* Used for BPF run context */
1593 struct bpf_run_ctx *bpf_ctx;
1594 #endif
1595 /* Used by BPF for per-TASK xdp storage */
1596 struct bpf_net_context *bpf_net_context;
1597
1598 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1599 unsigned long lowest_stack;
1600 unsigned long prev_lowest_stack;
1601 #endif
1602
1603 #ifdef CONFIG_X86_MCE
1604 void __user *mce_vaddr;
1605 __u64 mce_kflags;
1606 u64 mce_addr;
1607 __u64 mce_ripv : 1,
1608 mce_whole_page : 1,
1609 __mce_reserved : 62;
1610 struct callback_head mce_kill_me;
1611 int mce_count;
1612 #endif
1613
1614 #ifdef CONFIG_KRETPROBES
1615 struct llist_head kretprobe_instances;
1616 #endif
1617 #ifdef CONFIG_RETHOOK
1618 struct llist_head rethooks;
1619 #endif
1620
1621 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1622 /*
1623 * If L1D flush is supported on mm context switch
1624 * then we use this callback head to queue kill work
1625 * to kill tasks that are not running on SMT disabled
1626 * cores
1627 */
1628 struct callback_head l1d_flush_kill;
1629 #endif
1630
1631 #ifdef CONFIG_RV
1632 /*
1633 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1634 * If we find justification for more monitors, we can think
1635 * about adding more or developing a dynamic method. So far,
1636 * none of these are justified.
1637 */
1638 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1639 #endif
1640
1641 #ifdef CONFIG_USER_EVENTS
1642 struct user_event_mm *user_event_mm;
1643 #endif
1644
1645 /*
1646 * New fields for task_struct should be added above here, so that
1647 * they are included in the randomized portion of task_struct.
1648 */
1649 randomized_struct_fields_end
1650
1651 /* CPU-specific state of this task: */
1652 struct thread_struct thread;
1653
1654 /*
1655 * WARNING: on x86, 'thread_struct' contains a variable-sized
1656 * structure. It *MUST* be at the end of 'task_struct'.
1657 *
1658 * Do not put anything below here!
1659 */
1660 };
1661
1662 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1663 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1664
__task_state_index(unsigned int tsk_state,unsigned int tsk_exit_state)1665 static inline unsigned int __task_state_index(unsigned int tsk_state,
1666 unsigned int tsk_exit_state)
1667 {
1668 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1669
1670 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1671
1672 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1673 state = TASK_REPORT_IDLE;
1674
1675 /*
1676 * We're lying here, but rather than expose a completely new task state
1677 * to userspace, we can make this appear as if the task has gone through
1678 * a regular rt_mutex_lock() call.
1679 * Report frozen tasks as uninterruptible.
1680 */
1681 if ((tsk_state & TASK_RTLOCK_WAIT) || (tsk_state & TASK_FROZEN))
1682 state = TASK_UNINTERRUPTIBLE;
1683
1684 return fls(state);
1685 }
1686
task_state_index(struct task_struct * tsk)1687 static inline unsigned int task_state_index(struct task_struct *tsk)
1688 {
1689 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1690 }
1691
task_index_to_char(unsigned int state)1692 static inline char task_index_to_char(unsigned int state)
1693 {
1694 static const char state_char[] = "RSDTtXZPI";
1695
1696 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1697
1698 return state_char[state];
1699 }
1700
task_state_to_char(struct task_struct * tsk)1701 static inline char task_state_to_char(struct task_struct *tsk)
1702 {
1703 return task_index_to_char(task_state_index(tsk));
1704 }
1705
1706 extern struct pid *cad_pid;
1707
1708 /*
1709 * Per process flags
1710 */
1711 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1712 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1713 #define PF_EXITING 0x00000004 /* Getting shut down */
1714 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1715 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1716 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1717 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1718 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1719 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1720 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1721 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1722 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1723 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1724 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1725 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1726 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1727 #define PF_KCOMPACTD 0x00010000 /* I am kcompactd */
1728 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1729 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1730 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1731 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1732 * I am cleaning dirty pages from some other bdi. */
1733 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1734 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1735 #define PF__HOLE__00800000 0x00800000
1736 #define PF__HOLE__01000000 0x01000000
1737 #define PF__HOLE__02000000 0x02000000
1738 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1739 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1740 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1741 * See memalloc_pin_save() */
1742 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1743 #define PF__HOLE__40000000 0x40000000
1744 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1745
1746 /*
1747 * Only the _current_ task can read/write to tsk->flags, but other
1748 * tasks can access tsk->flags in readonly mode for example
1749 * with tsk_used_math (like during threaded core dumping).
1750 * There is however an exception to this rule during ptrace
1751 * or during fork: the ptracer task is allowed to write to the
1752 * child->flags of its traced child (same goes for fork, the parent
1753 * can write to the child->flags), because we're guaranteed the
1754 * child is not running and in turn not changing child->flags
1755 * at the same time the parent does it.
1756 */
1757 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1758 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1759 #define clear_used_math() clear_stopped_child_used_math(current)
1760 #define set_used_math() set_stopped_child_used_math(current)
1761
1762 #define conditional_stopped_child_used_math(condition, child) \
1763 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1764
1765 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1766
1767 #define copy_to_stopped_child_used_math(child) \
1768 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1769
1770 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1771 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1772 #define used_math() tsk_used_math(current)
1773
is_percpu_thread(void)1774 static __always_inline bool is_percpu_thread(void)
1775 {
1776 #ifdef CONFIG_SMP
1777 return (current->flags & PF_NO_SETAFFINITY) &&
1778 (current->nr_cpus_allowed == 1);
1779 #else
1780 return true;
1781 #endif
1782 }
1783
1784 /* Per-process atomic flags. */
1785 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1786 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1787 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1788 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1789 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1790 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1791 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1792 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1793
1794 #define TASK_PFA_TEST(name, func) \
1795 static inline bool task_##func(struct task_struct *p) \
1796 { return test_bit(PFA_##name, &p->atomic_flags); }
1797
1798 #define TASK_PFA_SET(name, func) \
1799 static inline void task_set_##func(struct task_struct *p) \
1800 { set_bit(PFA_##name, &p->atomic_flags); }
1801
1802 #define TASK_PFA_CLEAR(name, func) \
1803 static inline void task_clear_##func(struct task_struct *p) \
1804 { clear_bit(PFA_##name, &p->atomic_flags); }
1805
TASK_PFA_TEST(NO_NEW_PRIVS,no_new_privs)1806 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1807 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1808
1809 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1810 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1811 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1812
1813 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1814 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1815 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1816
1817 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1818 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1819 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1820
1821 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1822 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1823 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1824
1825 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1826 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1827
1828 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1829 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1830 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1831
1832 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1833 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1834
1835 static inline void
1836 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1837 {
1838 current->flags &= ~flags;
1839 current->flags |= orig_flags & flags;
1840 }
1841
1842 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1843 extern int task_can_attach(struct task_struct *p);
1844 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1845 extern void dl_bw_free(int cpu, u64 dl_bw);
1846 #ifdef CONFIG_SMP
1847
1848 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1849 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1850
1851 /**
1852 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1853 * @p: the task
1854 * @new_mask: CPU affinity mask
1855 *
1856 * Return: zero if successful, or a negative error code
1857 */
1858 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1859 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1860 extern void release_user_cpus_ptr(struct task_struct *p);
1861 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1862 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1863 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1864 #else
do_set_cpus_allowed(struct task_struct * p,const struct cpumask * new_mask)1865 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1866 {
1867 }
set_cpus_allowed_ptr(struct task_struct * p,const struct cpumask * new_mask)1868 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1869 {
1870 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1871 if ((*cpumask_bits(new_mask) & 1) == 0)
1872 return -EINVAL;
1873 return 0;
1874 }
dup_user_cpus_ptr(struct task_struct * dst,struct task_struct * src,int node)1875 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1876 {
1877 if (src->user_cpus_ptr)
1878 return -EINVAL;
1879 return 0;
1880 }
release_user_cpus_ptr(struct task_struct * p)1881 static inline void release_user_cpus_ptr(struct task_struct *p)
1882 {
1883 WARN_ON(p->user_cpus_ptr);
1884 }
1885
dl_task_check_affinity(struct task_struct * p,const struct cpumask * mask)1886 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1887 {
1888 return 0;
1889 }
1890 #endif
1891
1892 extern int yield_to(struct task_struct *p, bool preempt);
1893 extern void set_user_nice(struct task_struct *p, long nice);
1894 extern int task_prio(const struct task_struct *p);
1895
1896 /**
1897 * task_nice - return the nice value of a given task.
1898 * @p: the task in question.
1899 *
1900 * Return: The nice value [ -20 ... 0 ... 19 ].
1901 */
task_nice(const struct task_struct * p)1902 static inline int task_nice(const struct task_struct *p)
1903 {
1904 return PRIO_TO_NICE((p)->static_prio);
1905 }
1906
1907 extern int can_nice(const struct task_struct *p, const int nice);
1908 extern int task_curr(const struct task_struct *p);
1909 extern int idle_cpu(int cpu);
1910 extern int available_idle_cpu(int cpu);
1911 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1912 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1913 extern void sched_set_fifo(struct task_struct *p);
1914 extern void sched_set_fifo_low(struct task_struct *p);
1915 extern void sched_set_normal(struct task_struct *p, int nice);
1916 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1917 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1918 extern struct task_struct *idle_task(int cpu);
1919
1920 /**
1921 * is_idle_task - is the specified task an idle task?
1922 * @p: the task in question.
1923 *
1924 * Return: 1 if @p is an idle task. 0 otherwise.
1925 */
is_idle_task(const struct task_struct * p)1926 static __always_inline bool is_idle_task(const struct task_struct *p)
1927 {
1928 return !!(p->flags & PF_IDLE);
1929 }
1930
1931 extern struct task_struct *curr_task(int cpu);
1932 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1933
1934 void yield(void);
1935
1936 union thread_union {
1937 struct task_struct task;
1938 #ifndef CONFIG_THREAD_INFO_IN_TASK
1939 struct thread_info thread_info;
1940 #endif
1941 unsigned long stack[THREAD_SIZE/sizeof(long)];
1942 };
1943
1944 #ifndef CONFIG_THREAD_INFO_IN_TASK
1945 extern struct thread_info init_thread_info;
1946 #endif
1947
1948 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1949
1950 #ifdef CONFIG_THREAD_INFO_IN_TASK
1951 # define task_thread_info(task) (&(task)->thread_info)
1952 #else
1953 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1954 #endif
1955
1956 /*
1957 * find a task by one of its numerical ids
1958 *
1959 * find_task_by_pid_ns():
1960 * finds a task by its pid in the specified namespace
1961 * find_task_by_vpid():
1962 * finds a task by its virtual pid
1963 *
1964 * see also find_vpid() etc in include/linux/pid.h
1965 */
1966
1967 extern struct task_struct *find_task_by_vpid(pid_t nr);
1968 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1969
1970 /*
1971 * find a task by its virtual pid and get the task struct
1972 */
1973 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1974
1975 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1976 extern int wake_up_process(struct task_struct *tsk);
1977 extern void wake_up_new_task(struct task_struct *tsk);
1978
1979 #ifdef CONFIG_SMP
1980 extern void kick_process(struct task_struct *tsk);
1981 #else
kick_process(struct task_struct * tsk)1982 static inline void kick_process(struct task_struct *tsk) { }
1983 #endif
1984
1985 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1986 #define set_task_comm(tsk, from) ({ \
1987 BUILD_BUG_ON(sizeof(from) != TASK_COMM_LEN); \
1988 __set_task_comm(tsk, from, false); \
1989 })
1990
1991 /*
1992 * - Why not use task_lock()?
1993 * User space can randomly change their names anyway, so locking for readers
1994 * doesn't make sense. For writers, locking is probably necessary, as a race
1995 * condition could lead to long-term mixed results.
1996 * The strscpy_pad() in __set_task_comm() can ensure that the task comm is
1997 * always NUL-terminated and zero-padded. Therefore the race condition between
1998 * reader and writer is not an issue.
1999 *
2000 * - BUILD_BUG_ON() can help prevent the buf from being truncated.
2001 * Since the callers don't perform any return value checks, this safeguard is
2002 * necessary.
2003 */
2004 #define get_task_comm(buf, tsk) ({ \
2005 BUILD_BUG_ON(sizeof(buf) < TASK_COMM_LEN); \
2006 strscpy_pad(buf, (tsk)->comm); \
2007 buf; \
2008 })
2009
2010 #ifdef CONFIG_SMP
scheduler_ipi(void)2011 static __always_inline void scheduler_ipi(void)
2012 {
2013 /*
2014 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
2015 * TIF_NEED_RESCHED remotely (for the first time) will also send
2016 * this IPI.
2017 */
2018 preempt_fold_need_resched();
2019 }
2020 #else
scheduler_ipi(void)2021 static inline void scheduler_ipi(void) { }
2022 #endif
2023
2024 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
2025
2026 /*
2027 * Set thread flags in other task's structures.
2028 * See asm/thread_info.h for TIF_xxxx flags available:
2029 */
set_tsk_thread_flag(struct task_struct * tsk,int flag)2030 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2031 {
2032 set_ti_thread_flag(task_thread_info(tsk), flag);
2033 }
2034
clear_tsk_thread_flag(struct task_struct * tsk,int flag)2035 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2036 {
2037 clear_ti_thread_flag(task_thread_info(tsk), flag);
2038 }
2039
update_tsk_thread_flag(struct task_struct * tsk,int flag,bool value)2040 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2041 bool value)
2042 {
2043 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2044 }
2045
test_and_set_tsk_thread_flag(struct task_struct * tsk,int flag)2046 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2047 {
2048 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2049 }
2050
test_and_clear_tsk_thread_flag(struct task_struct * tsk,int flag)2051 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2052 {
2053 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2054 }
2055
test_tsk_thread_flag(struct task_struct * tsk,int flag)2056 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2057 {
2058 return test_ti_thread_flag(task_thread_info(tsk), flag);
2059 }
2060
set_tsk_need_resched(struct task_struct * tsk)2061 static inline void set_tsk_need_resched(struct task_struct *tsk)
2062 {
2063 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2064 }
2065
clear_tsk_need_resched(struct task_struct * tsk)2066 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2067 {
2068 atomic_long_andnot(_TIF_NEED_RESCHED | _TIF_NEED_RESCHED_LAZY,
2069 (atomic_long_t *)&task_thread_info(tsk)->flags);
2070 }
2071
test_tsk_need_resched(struct task_struct * tsk)2072 static inline int test_tsk_need_resched(struct task_struct *tsk)
2073 {
2074 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2075 }
2076
2077 /*
2078 * cond_resched() and cond_resched_lock(): latency reduction via
2079 * explicit rescheduling in places that are safe. The return
2080 * value indicates whether a reschedule was done in fact.
2081 * cond_resched_lock() will drop the spinlock before scheduling,
2082 */
2083 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2084 extern int __cond_resched(void);
2085
2086 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2087
2088 void sched_dynamic_klp_enable(void);
2089 void sched_dynamic_klp_disable(void);
2090
2091 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2092
_cond_resched(void)2093 static __always_inline int _cond_resched(void)
2094 {
2095 return static_call_mod(cond_resched)();
2096 }
2097
2098 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2099
2100 extern int dynamic_cond_resched(void);
2101
_cond_resched(void)2102 static __always_inline int _cond_resched(void)
2103 {
2104 return dynamic_cond_resched();
2105 }
2106
2107 #else /* !CONFIG_PREEMPTION */
2108
_cond_resched(void)2109 static inline int _cond_resched(void)
2110 {
2111 klp_sched_try_switch();
2112 return __cond_resched();
2113 }
2114
2115 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2116
2117 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2118
_cond_resched(void)2119 static inline int _cond_resched(void)
2120 {
2121 klp_sched_try_switch();
2122 return 0;
2123 }
2124
2125 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2126
2127 #define cond_resched() ({ \
2128 __might_resched(__FILE__, __LINE__, 0); \
2129 _cond_resched(); \
2130 })
2131
2132 extern int __cond_resched_lock(spinlock_t *lock);
2133 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2134 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2135
2136 #define MIGHT_RESCHED_RCU_SHIFT 8
2137 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2138
2139 #ifndef CONFIG_PREEMPT_RT
2140 /*
2141 * Non RT kernels have an elevated preempt count due to the held lock,
2142 * but are not allowed to be inside a RCU read side critical section
2143 */
2144 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2145 #else
2146 /*
2147 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2148 * cond_resched*lock() has to take that into account because it checks for
2149 * preempt_count() and rcu_preempt_depth().
2150 */
2151 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2152 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2153 #endif
2154
2155 #define cond_resched_lock(lock) ({ \
2156 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2157 __cond_resched_lock(lock); \
2158 })
2159
2160 #define cond_resched_rwlock_read(lock) ({ \
2161 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2162 __cond_resched_rwlock_read(lock); \
2163 })
2164
2165 #define cond_resched_rwlock_write(lock) ({ \
2166 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2167 __cond_resched_rwlock_write(lock); \
2168 })
2169
need_resched(void)2170 static __always_inline bool need_resched(void)
2171 {
2172 return unlikely(tif_need_resched());
2173 }
2174
2175 /*
2176 * Wrappers for p->thread_info->cpu access. No-op on UP.
2177 */
2178 #ifdef CONFIG_SMP
2179
task_cpu(const struct task_struct * p)2180 static inline unsigned int task_cpu(const struct task_struct *p)
2181 {
2182 return READ_ONCE(task_thread_info(p)->cpu);
2183 }
2184
2185 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2186
2187 #else
2188
task_cpu(const struct task_struct * p)2189 static inline unsigned int task_cpu(const struct task_struct *p)
2190 {
2191 return 0;
2192 }
2193
set_task_cpu(struct task_struct * p,unsigned int cpu)2194 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2195 {
2196 }
2197
2198 #endif /* CONFIG_SMP */
2199
task_is_runnable(struct task_struct * p)2200 static inline bool task_is_runnable(struct task_struct *p)
2201 {
2202 return p->on_rq && !p->se.sched_delayed;
2203 }
2204
2205 extern bool sched_task_on_rq(struct task_struct *p);
2206 extern unsigned long get_wchan(struct task_struct *p);
2207 extern struct task_struct *cpu_curr_snapshot(int cpu);
2208
2209 #include <linux/spinlock.h>
2210
2211 /*
2212 * In order to reduce various lock holder preemption latencies provide an
2213 * interface to see if a vCPU is currently running or not.
2214 *
2215 * This allows us to terminate optimistic spin loops and block, analogous to
2216 * the native optimistic spin heuristic of testing if the lock owner task is
2217 * running or not.
2218 */
2219 #ifndef vcpu_is_preempted
vcpu_is_preempted(int cpu)2220 static inline bool vcpu_is_preempted(int cpu)
2221 {
2222 return false;
2223 }
2224 #endif
2225
2226 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2227 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2228
2229 #ifndef TASK_SIZE_OF
2230 #define TASK_SIZE_OF(tsk) TASK_SIZE
2231 #endif
2232
2233 #ifdef CONFIG_SMP
owner_on_cpu(struct task_struct * owner)2234 static inline bool owner_on_cpu(struct task_struct *owner)
2235 {
2236 /*
2237 * As lock holder preemption issue, we both skip spinning if
2238 * task is not on cpu or its cpu is preempted
2239 */
2240 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2241 }
2242
2243 /* Returns effective CPU energy utilization, as seen by the scheduler */
2244 unsigned long sched_cpu_util(int cpu);
2245 #endif /* CONFIG_SMP */
2246
2247 #ifdef CONFIG_SCHED_CORE
2248 extern void sched_core_free(struct task_struct *tsk);
2249 extern void sched_core_fork(struct task_struct *p);
2250 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2251 unsigned long uaddr);
2252 extern int sched_core_idle_cpu(int cpu);
2253 #else
sched_core_free(struct task_struct * tsk)2254 static inline void sched_core_free(struct task_struct *tsk) { }
sched_core_fork(struct task_struct * p)2255 static inline void sched_core_fork(struct task_struct *p) { }
sched_core_idle_cpu(int cpu)2256 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2257 #endif
2258
2259 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2260
2261 #ifdef CONFIG_MEM_ALLOC_PROFILING
alloc_tag_save(struct alloc_tag * tag)2262 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2263 {
2264 swap(current->alloc_tag, tag);
2265 return tag;
2266 }
2267
alloc_tag_restore(struct alloc_tag * tag,struct alloc_tag * old)2268 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2269 {
2270 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2271 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2272 #endif
2273 current->alloc_tag = old;
2274 }
2275 #else
2276 #define alloc_tag_save(_tag) NULL
2277 #define alloc_tag_restore(_tag, _old) do {} while (0)
2278 #endif
2279
2280 #endif
2281