sched: Widen TAKS_state literals
[platform/kernel/linux-starfive.git] / include / linux / sched.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SCHED_H
3 #define _LINUX_SCHED_H
4
5 /*
6  * Define 'struct task_struct' and provide the main scheduler
7  * APIs (schedule(), wakeup variants, etc.)
8  */
9
10 #include <uapi/linux/sched.h>
11
12 #include <asm/current.h>
13
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/mutex.h>
18 #include <linux/plist.h>
19 #include <linux/hrtimer.h>
20 #include <linux/irqflags.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/refcount.h>
25 #include <linux/resource.h>
26 #include <linux/latencytop.h>
27 #include <linux/sched/prio.h>
28 #include <linux/sched/types.h>
29 #include <linux/signal_types.h>
30 #include <linux/syscall_user_dispatch.h>
31 #include <linux/mm_types_task.h>
32 #include <linux/task_io_accounting.h>
33 #include <linux/posix-timers.h>
34 #include <linux/rseq.h>
35 #include <linux/seqlock.h>
36 #include <linux/kcsan.h>
37 #include <linux/rv.h>
38 #include <asm/kmap_size.h>
39
40 /* task_struct member predeclarations (sorted alphabetically): */
41 struct audit_context;
42 struct backing_dev_info;
43 struct bio_list;
44 struct blk_plug;
45 struct bpf_local_storage;
46 struct bpf_run_ctx;
47 struct capture_control;
48 struct cfs_rq;
49 struct fs_struct;
50 struct futex_pi_state;
51 struct io_context;
52 struct io_uring_task;
53 struct mempolicy;
54 struct nameidata;
55 struct nsproxy;
56 struct perf_event_context;
57 struct pid_namespace;
58 struct pipe_inode_info;
59 struct rcu_node;
60 struct reclaim_state;
61 struct robust_list_head;
62 struct root_domain;
63 struct rq;
64 struct sched_attr;
65 struct sched_param;
66 struct seq_file;
67 struct sighand_struct;
68 struct signal_struct;
69 struct task_delay_info;
70 struct task_group;
71
72 /*
73  * Task state bitmask. NOTE! These bits are also
74  * encoded in fs/proc/array.c: get_task_state().
75  *
76  * We have two separate sets of flags: task->state
77  * is about runnability, while task->exit_state are
78  * about the task exiting. Confusing, but this way
79  * modifying one set can't modify the other one by
80  * mistake.
81  */
82
83 /* Used in tsk->state: */
84 #define TASK_RUNNING                    0x00000000
85 #define TASK_INTERRUPTIBLE              0x00000001
86 #define TASK_UNINTERRUPTIBLE            0x00000002
87 #define __TASK_STOPPED                  0x00000004
88 #define __TASK_TRACED                   0x00000008
89 /* Used in tsk->exit_state: */
90 #define EXIT_DEAD                       0x00000010
91 #define EXIT_ZOMBIE                     0x00000020
92 #define EXIT_TRACE                      (EXIT_ZOMBIE | EXIT_DEAD)
93 /* Used in tsk->state again: */
94 #define TASK_PARKED                     0x00000040
95 #define TASK_DEAD                       0x00000080
96 #define TASK_WAKEKILL                   0x00000100
97 #define TASK_WAKING                     0x00000200
98 #define TASK_NOLOAD                     0x00000400
99 #define TASK_NEW                        0x00000800
100 /* RT specific auxilliary flag to mark RT lock waiters */
101 #define TASK_RTLOCK_WAIT                0x00001000
102 #define TASK_STATE_MAX                  0x00002000
103
104 #define TASK_ANY                        (TASK_STATE_MAX-1)
105
106 /* Convenience macros for the sake of set_current_state: */
107 #define TASK_KILLABLE                   (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
108 #define TASK_STOPPED                    (TASK_WAKEKILL | __TASK_STOPPED)
109 #define TASK_TRACED                     __TASK_TRACED
110
111 #define TASK_IDLE                       (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
112
113 /* Convenience macros for the sake of wake_up(): */
114 #define TASK_NORMAL                     (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
115
116 /* get_task_state(): */
117 #define TASK_REPORT                     (TASK_RUNNING | TASK_INTERRUPTIBLE | \
118                                          TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
119                                          __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
120                                          TASK_PARKED)
121
122 #define task_is_running(task)           (READ_ONCE((task)->__state) == TASK_RUNNING)
123
124 #define task_is_traced(task)            ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
125 #define task_is_stopped(task)           ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
126 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
127
128 /*
129  * Special states are those that do not use the normal wait-loop pattern. See
130  * the comment with set_special_state().
131  */
132 #define is_special_task_state(state)                            \
133         ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
134
135 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
136 # define debug_normal_state_change(state_value)                         \
137         do {                                                            \
138                 WARN_ON_ONCE(is_special_task_state(state_value));       \
139                 current->task_state_change = _THIS_IP_;                 \
140         } while (0)
141
142 # define debug_special_state_change(state_value)                        \
143         do {                                                            \
144                 WARN_ON_ONCE(!is_special_task_state(state_value));      \
145                 current->task_state_change = _THIS_IP_;                 \
146         } while (0)
147
148 # define debug_rtlock_wait_set_state()                                  \
149         do {                                                             \
150                 current->saved_state_change = current->task_state_change;\
151                 current->task_state_change = _THIS_IP_;                  \
152         } while (0)
153
154 # define debug_rtlock_wait_restore_state()                              \
155         do {                                                             \
156                 current->task_state_change = current->saved_state_change;\
157         } while (0)
158
159 #else
160 # define debug_normal_state_change(cond)        do { } while (0)
161 # define debug_special_state_change(cond)       do { } while (0)
162 # define debug_rtlock_wait_set_state()          do { } while (0)
163 # define debug_rtlock_wait_restore_state()      do { } while (0)
164 #endif
165
166 /*
167  * set_current_state() includes a barrier so that the write of current->state
168  * is correctly serialised wrt the caller's subsequent test of whether to
169  * actually sleep:
170  *
171  *   for (;;) {
172  *      set_current_state(TASK_UNINTERRUPTIBLE);
173  *      if (CONDITION)
174  *         break;
175  *
176  *      schedule();
177  *   }
178  *   __set_current_state(TASK_RUNNING);
179  *
180  * If the caller does not need such serialisation (because, for instance, the
181  * CONDITION test and condition change and wakeup are under the same lock) then
182  * use __set_current_state().
183  *
184  * The above is typically ordered against the wakeup, which does:
185  *
186  *   CONDITION = 1;
187  *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
188  *
189  * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
190  * accessing p->state.
191  *
192  * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
193  * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
194  * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
195  *
196  * However, with slightly different timing the wakeup TASK_RUNNING store can
197  * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
198  * a problem either because that will result in one extra go around the loop
199  * and our @cond test will save the day.
200  *
201  * Also see the comments of try_to_wake_up().
202  */
203 #define __set_current_state(state_value)                                \
204         do {                                                            \
205                 debug_normal_state_change((state_value));               \
206                 WRITE_ONCE(current->__state, (state_value));            \
207         } while (0)
208
209 #define set_current_state(state_value)                                  \
210         do {                                                            \
211                 debug_normal_state_change((state_value));               \
212                 smp_store_mb(current->__state, (state_value));          \
213         } while (0)
214
215 /*
216  * set_special_state() should be used for those states when the blocking task
217  * can not use the regular condition based wait-loop. In that case we must
218  * serialize against wakeups such that any possible in-flight TASK_RUNNING
219  * stores will not collide with our state change.
220  */
221 #define set_special_state(state_value)                                  \
222         do {                                                            \
223                 unsigned long flags; /* may shadow */                   \
224                                                                         \
225                 raw_spin_lock_irqsave(&current->pi_lock, flags);        \
226                 debug_special_state_change((state_value));              \
227                 WRITE_ONCE(current->__state, (state_value));            \
228                 raw_spin_unlock_irqrestore(&current->pi_lock, flags);   \
229         } while (0)
230
231 /*
232  * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
233  *
234  * RT's spin/rwlock substitutions are state preserving. The state of the
235  * task when blocking on the lock is saved in task_struct::saved_state and
236  * restored after the lock has been acquired.  These operations are
237  * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
238  * lock related wakeups while the task is blocked on the lock are
239  * redirected to operate on task_struct::saved_state to ensure that these
240  * are not dropped. On restore task_struct::saved_state is set to
241  * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
242  *
243  * The lock operation looks like this:
244  *
245  *      current_save_and_set_rtlock_wait_state();
246  *      for (;;) {
247  *              if (try_lock())
248  *                      break;
249  *              raw_spin_unlock_irq(&lock->wait_lock);
250  *              schedule_rtlock();
251  *              raw_spin_lock_irq(&lock->wait_lock);
252  *              set_current_state(TASK_RTLOCK_WAIT);
253  *      }
254  *      current_restore_rtlock_saved_state();
255  */
256 #define current_save_and_set_rtlock_wait_state()                        \
257         do {                                                            \
258                 lockdep_assert_irqs_disabled();                         \
259                 raw_spin_lock(&current->pi_lock);                       \
260                 current->saved_state = current->__state;                \
261                 debug_rtlock_wait_set_state();                          \
262                 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT);         \
263                 raw_spin_unlock(&current->pi_lock);                     \
264         } while (0);
265
266 #define current_restore_rtlock_saved_state()                            \
267         do {                                                            \
268                 lockdep_assert_irqs_disabled();                         \
269                 raw_spin_lock(&current->pi_lock);                       \
270                 debug_rtlock_wait_restore_state();                      \
271                 WRITE_ONCE(current->__state, current->saved_state);     \
272                 current->saved_state = TASK_RUNNING;                    \
273                 raw_spin_unlock(&current->pi_lock);                     \
274         } while (0);
275
276 #define get_current_state()     READ_ONCE(current->__state)
277
278 /*
279  * Define the task command name length as enum, then it can be visible to
280  * BPF programs.
281  */
282 enum {
283         TASK_COMM_LEN = 16,
284 };
285
286 extern void scheduler_tick(void);
287
288 #define MAX_SCHEDULE_TIMEOUT            LONG_MAX
289
290 extern long schedule_timeout(long timeout);
291 extern long schedule_timeout_interruptible(long timeout);
292 extern long schedule_timeout_killable(long timeout);
293 extern long schedule_timeout_uninterruptible(long timeout);
294 extern long schedule_timeout_idle(long timeout);
295 asmlinkage void schedule(void);
296 extern void schedule_preempt_disabled(void);
297 asmlinkage void preempt_schedule_irq(void);
298 #ifdef CONFIG_PREEMPT_RT
299  extern void schedule_rtlock(void);
300 #endif
301
302 extern int __must_check io_schedule_prepare(void);
303 extern void io_schedule_finish(int token);
304 extern long io_schedule_timeout(long timeout);
305 extern void io_schedule(void);
306
307 /**
308  * struct prev_cputime - snapshot of system and user cputime
309  * @utime: time spent in user mode
310  * @stime: time spent in system mode
311  * @lock: protects the above two fields
312  *
313  * Stores previous user/system time values such that we can guarantee
314  * monotonicity.
315  */
316 struct prev_cputime {
317 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
318         u64                             utime;
319         u64                             stime;
320         raw_spinlock_t                  lock;
321 #endif
322 };
323
324 enum vtime_state {
325         /* Task is sleeping or running in a CPU with VTIME inactive: */
326         VTIME_INACTIVE = 0,
327         /* Task is idle */
328         VTIME_IDLE,
329         /* Task runs in kernelspace in a CPU with VTIME active: */
330         VTIME_SYS,
331         /* Task runs in userspace in a CPU with VTIME active: */
332         VTIME_USER,
333         /* Task runs as guests in a CPU with VTIME active: */
334         VTIME_GUEST,
335 };
336
337 struct vtime {
338         seqcount_t              seqcount;
339         unsigned long long      starttime;
340         enum vtime_state        state;
341         unsigned int            cpu;
342         u64                     utime;
343         u64                     stime;
344         u64                     gtime;
345 };
346
347 /*
348  * Utilization clamp constraints.
349  * @UCLAMP_MIN: Minimum utilization
350  * @UCLAMP_MAX: Maximum utilization
351  * @UCLAMP_CNT: Utilization clamp constraints count
352  */
353 enum uclamp_id {
354         UCLAMP_MIN = 0,
355         UCLAMP_MAX,
356         UCLAMP_CNT
357 };
358
359 #ifdef CONFIG_SMP
360 extern struct root_domain def_root_domain;
361 extern struct mutex sched_domains_mutex;
362 #endif
363
364 struct sched_info {
365 #ifdef CONFIG_SCHED_INFO
366         /* Cumulative counters: */
367
368         /* # of times we have run on this CPU: */
369         unsigned long                   pcount;
370
371         /* Time spent waiting on a runqueue: */
372         unsigned long long              run_delay;
373
374         /* Timestamps: */
375
376         /* When did we last run on a CPU? */
377         unsigned long long              last_arrival;
378
379         /* When were we last queued to run? */
380         unsigned long long              last_queued;
381
382 #endif /* CONFIG_SCHED_INFO */
383 };
384
385 /*
386  * Integer metrics need fixed point arithmetic, e.g., sched/fair
387  * has a few: load, load_avg, util_avg, freq, and capacity.
388  *
389  * We define a basic fixed point arithmetic range, and then formalize
390  * all these metrics based on that basic range.
391  */
392 # define SCHED_FIXEDPOINT_SHIFT         10
393 # define SCHED_FIXEDPOINT_SCALE         (1L << SCHED_FIXEDPOINT_SHIFT)
394
395 /* Increase resolution of cpu_capacity calculations */
396 # define SCHED_CAPACITY_SHIFT           SCHED_FIXEDPOINT_SHIFT
397 # define SCHED_CAPACITY_SCALE           (1L << SCHED_CAPACITY_SHIFT)
398
399 struct load_weight {
400         unsigned long                   weight;
401         u32                             inv_weight;
402 };
403
404 /**
405  * struct util_est - Estimation utilization of FAIR tasks
406  * @enqueued: instantaneous estimated utilization of a task/cpu
407  * @ewma:     the Exponential Weighted Moving Average (EWMA)
408  *            utilization of a task
409  *
410  * Support data structure to track an Exponential Weighted Moving Average
411  * (EWMA) of a FAIR task's utilization. New samples are added to the moving
412  * average each time a task completes an activation. Sample's weight is chosen
413  * so that the EWMA will be relatively insensitive to transient changes to the
414  * task's workload.
415  *
416  * The enqueued attribute has a slightly different meaning for tasks and cpus:
417  * - task:   the task's util_avg at last task dequeue time
418  * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
419  * Thus, the util_est.enqueued of a task represents the contribution on the
420  * estimated utilization of the CPU where that task is currently enqueued.
421  *
422  * Only for tasks we track a moving average of the past instantaneous
423  * estimated utilization. This allows to absorb sporadic drops in utilization
424  * of an otherwise almost periodic task.
425  *
426  * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
427  * updates. When a task is dequeued, its util_est should not be updated if its
428  * util_avg has not been updated in the meantime.
429  * This information is mapped into the MSB bit of util_est.enqueued at dequeue
430  * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
431  * for a task) it is safe to use MSB.
432  */
433 struct util_est {
434         unsigned int                    enqueued;
435         unsigned int                    ewma;
436 #define UTIL_EST_WEIGHT_SHIFT           2
437 #define UTIL_AVG_UNCHANGED              0x80000000
438 } __attribute__((__aligned__(sizeof(u64))));
439
440 /*
441  * The load/runnable/util_avg accumulates an infinite geometric series
442  * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
443  *
444  * [load_avg definition]
445  *
446  *   load_avg = runnable% * scale_load_down(load)
447  *
448  * [runnable_avg definition]
449  *
450  *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
451  *
452  * [util_avg definition]
453  *
454  *   util_avg = running% * SCHED_CAPACITY_SCALE
455  *
456  * where runnable% is the time ratio that a sched_entity is runnable and
457  * running% the time ratio that a sched_entity is running.
458  *
459  * For cfs_rq, they are the aggregated values of all runnable and blocked
460  * sched_entities.
461  *
462  * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
463  * capacity scaling. The scaling is done through the rq_clock_pelt that is used
464  * for computing those signals (see update_rq_clock_pelt())
465  *
466  * N.B., the above ratios (runnable% and running%) themselves are in the
467  * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
468  * to as large a range as necessary. This is for example reflected by
469  * util_avg's SCHED_CAPACITY_SCALE.
470  *
471  * [Overflow issue]
472  *
473  * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
474  * with the highest load (=88761), always runnable on a single cfs_rq,
475  * and should not overflow as the number already hits PID_MAX_LIMIT.
476  *
477  * For all other cases (including 32-bit kernels), struct load_weight's
478  * weight will overflow first before we do, because:
479  *
480  *    Max(load_avg) <= Max(load.weight)
481  *
482  * Then it is the load_weight's responsibility to consider overflow
483  * issues.
484  */
485 struct sched_avg {
486         u64                             last_update_time;
487         u64                             load_sum;
488         u64                             runnable_sum;
489         u32                             util_sum;
490         u32                             period_contrib;
491         unsigned long                   load_avg;
492         unsigned long                   runnable_avg;
493         unsigned long                   util_avg;
494         struct util_est                 util_est;
495 } ____cacheline_aligned;
496
497 struct sched_statistics {
498 #ifdef CONFIG_SCHEDSTATS
499         u64                             wait_start;
500         u64                             wait_max;
501         u64                             wait_count;
502         u64                             wait_sum;
503         u64                             iowait_count;
504         u64                             iowait_sum;
505
506         u64                             sleep_start;
507         u64                             sleep_max;
508         s64                             sum_sleep_runtime;
509
510         u64                             block_start;
511         u64                             block_max;
512         s64                             sum_block_runtime;
513
514         u64                             exec_max;
515         u64                             slice_max;
516
517         u64                             nr_migrations_cold;
518         u64                             nr_failed_migrations_affine;
519         u64                             nr_failed_migrations_running;
520         u64                             nr_failed_migrations_hot;
521         u64                             nr_forced_migrations;
522
523         u64                             nr_wakeups;
524         u64                             nr_wakeups_sync;
525         u64                             nr_wakeups_migrate;
526         u64                             nr_wakeups_local;
527         u64                             nr_wakeups_remote;
528         u64                             nr_wakeups_affine;
529         u64                             nr_wakeups_affine_attempts;
530         u64                             nr_wakeups_passive;
531         u64                             nr_wakeups_idle;
532
533 #ifdef CONFIG_SCHED_CORE
534         u64                             core_forceidle_sum;
535 #endif
536 #endif /* CONFIG_SCHEDSTATS */
537 } ____cacheline_aligned;
538
539 struct sched_entity {
540         /* For load-balancing: */
541         struct load_weight              load;
542         struct rb_node                  run_node;
543         struct list_head                group_node;
544         unsigned int                    on_rq;
545
546         u64                             exec_start;
547         u64                             sum_exec_runtime;
548         u64                             vruntime;
549         u64                             prev_sum_exec_runtime;
550
551         u64                             nr_migrations;
552
553 #ifdef CONFIG_FAIR_GROUP_SCHED
554         int                             depth;
555         struct sched_entity             *parent;
556         /* rq on which this entity is (to be) queued: */
557         struct cfs_rq                   *cfs_rq;
558         /* rq "owned" by this entity/group: */
559         struct cfs_rq                   *my_q;
560         /* cached value of my_q->h_nr_running */
561         unsigned long                   runnable_weight;
562 #endif
563
564 #ifdef CONFIG_SMP
565         /*
566          * Per entity load average tracking.
567          *
568          * Put into separate cache line so it does not
569          * collide with read-mostly values above.
570          */
571         struct sched_avg                avg;
572 #endif
573 };
574
575 struct sched_rt_entity {
576         struct list_head                run_list;
577         unsigned long                   timeout;
578         unsigned long                   watchdog_stamp;
579         unsigned int                    time_slice;
580         unsigned short                  on_rq;
581         unsigned short                  on_list;
582
583         struct sched_rt_entity          *back;
584 #ifdef CONFIG_RT_GROUP_SCHED
585         struct sched_rt_entity          *parent;
586         /* rq on which this entity is (to be) queued: */
587         struct rt_rq                    *rt_rq;
588         /* rq "owned" by this entity/group: */
589         struct rt_rq                    *my_q;
590 #endif
591 } __randomize_layout;
592
593 struct sched_dl_entity {
594         struct rb_node                  rb_node;
595
596         /*
597          * Original scheduling parameters. Copied here from sched_attr
598          * during sched_setattr(), they will remain the same until
599          * the next sched_setattr().
600          */
601         u64                             dl_runtime;     /* Maximum runtime for each instance    */
602         u64                             dl_deadline;    /* Relative deadline of each instance   */
603         u64                             dl_period;      /* Separation of two instances (period) */
604         u64                             dl_bw;          /* dl_runtime / dl_period               */
605         u64                             dl_density;     /* dl_runtime / dl_deadline             */
606
607         /*
608          * Actual scheduling parameters. Initialized with the values above,
609          * they are continuously updated during task execution. Note that
610          * the remaining runtime could be < 0 in case we are in overrun.
611          */
612         s64                             runtime;        /* Remaining runtime for this instance  */
613         u64                             deadline;       /* Absolute deadline for this instance  */
614         unsigned int                    flags;          /* Specifying the scheduler behaviour   */
615
616         /*
617          * Some bool flags:
618          *
619          * @dl_throttled tells if we exhausted the runtime. If so, the
620          * task has to wait for a replenishment to be performed at the
621          * next firing of dl_timer.
622          *
623          * @dl_yielded tells if task gave up the CPU before consuming
624          * all its available runtime during the last job.
625          *
626          * @dl_non_contending tells if the task is inactive while still
627          * contributing to the active utilization. In other words, it
628          * indicates if the inactive timer has been armed and its handler
629          * has not been executed yet. This flag is useful to avoid race
630          * conditions between the inactive timer handler and the wakeup
631          * code.
632          *
633          * @dl_overrun tells if the task asked to be informed about runtime
634          * overruns.
635          */
636         unsigned int                    dl_throttled      : 1;
637         unsigned int                    dl_yielded        : 1;
638         unsigned int                    dl_non_contending : 1;
639         unsigned int                    dl_overrun        : 1;
640
641         /*
642          * Bandwidth enforcement timer. Each -deadline task has its
643          * own bandwidth to be enforced, thus we need one timer per task.
644          */
645         struct hrtimer                  dl_timer;
646
647         /*
648          * Inactive timer, responsible for decreasing the active utilization
649          * at the "0-lag time". When a -deadline task blocks, it contributes
650          * to GRUB's active utilization until the "0-lag time", hence a
651          * timer is needed to decrease the active utilization at the correct
652          * time.
653          */
654         struct hrtimer inactive_timer;
655
656 #ifdef CONFIG_RT_MUTEXES
657         /*
658          * Priority Inheritance. When a DEADLINE scheduling entity is boosted
659          * pi_se points to the donor, otherwise points to the dl_se it belongs
660          * to (the original one/itself).
661          */
662         struct sched_dl_entity *pi_se;
663 #endif
664 };
665
666 #ifdef CONFIG_UCLAMP_TASK
667 /* Number of utilization clamp buckets (shorter alias) */
668 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
669
670 /*
671  * Utilization clamp for a scheduling entity
672  * @value:              clamp value "assigned" to a se
673  * @bucket_id:          bucket index corresponding to the "assigned" value
674  * @active:             the se is currently refcounted in a rq's bucket
675  * @user_defined:       the requested clamp value comes from user-space
676  *
677  * The bucket_id is the index of the clamp bucket matching the clamp value
678  * which is pre-computed and stored to avoid expensive integer divisions from
679  * the fast path.
680  *
681  * The active bit is set whenever a task has got an "effective" value assigned,
682  * which can be different from the clamp value "requested" from user-space.
683  * This allows to know a task is refcounted in the rq's bucket corresponding
684  * to the "effective" bucket_id.
685  *
686  * The user_defined bit is set whenever a task has got a task-specific clamp
687  * value requested from userspace, i.e. the system defaults apply to this task
688  * just as a restriction. This allows to relax default clamps when a less
689  * restrictive task-specific value has been requested, thus allowing to
690  * implement a "nice" semantic. For example, a task running with a 20%
691  * default boost can still drop its own boosting to 0%.
692  */
693 struct uclamp_se {
694         unsigned int value              : bits_per(SCHED_CAPACITY_SCALE);
695         unsigned int bucket_id          : bits_per(UCLAMP_BUCKETS);
696         unsigned int active             : 1;
697         unsigned int user_defined       : 1;
698 };
699 #endif /* CONFIG_UCLAMP_TASK */
700
701 union rcu_special {
702         struct {
703                 u8                      blocked;
704                 u8                      need_qs;
705                 u8                      exp_hint; /* Hint for performance. */
706                 u8                      need_mb; /* Readers need smp_mb(). */
707         } b; /* Bits. */
708         u32 s; /* Set of bits. */
709 };
710
711 enum perf_event_task_context {
712         perf_invalid_context = -1,
713         perf_hw_context = 0,
714         perf_sw_context,
715         perf_nr_task_contexts,
716 };
717
718 struct wake_q_node {
719         struct wake_q_node *next;
720 };
721
722 struct kmap_ctrl {
723 #ifdef CONFIG_KMAP_LOCAL
724         int                             idx;
725         pte_t                           pteval[KM_MAX_IDX];
726 #endif
727 };
728
729 struct task_struct {
730 #ifdef CONFIG_THREAD_INFO_IN_TASK
731         /*
732          * For reasons of header soup (see current_thread_info()), this
733          * must be the first element of task_struct.
734          */
735         struct thread_info              thread_info;
736 #endif
737         unsigned int                    __state;
738
739 #ifdef CONFIG_PREEMPT_RT
740         /* saved state for "spinlock sleepers" */
741         unsigned int                    saved_state;
742 #endif
743
744         /*
745          * This begins the randomizable portion of task_struct. Only
746          * scheduling-critical items should be added above here.
747          */
748         randomized_struct_fields_start
749
750         void                            *stack;
751         refcount_t                      usage;
752         /* Per task flags (PF_*), defined further below: */
753         unsigned int                    flags;
754         unsigned int                    ptrace;
755
756 #ifdef CONFIG_SMP
757         int                             on_cpu;
758         struct __call_single_node       wake_entry;
759         unsigned int                    wakee_flips;
760         unsigned long                   wakee_flip_decay_ts;
761         struct task_struct              *last_wakee;
762
763         /*
764          * recent_used_cpu is initially set as the last CPU used by a task
765          * that wakes affine another task. Waker/wakee relationships can
766          * push tasks around a CPU where each wakeup moves to the next one.
767          * Tracking a recently used CPU allows a quick search for a recently
768          * used CPU that may be idle.
769          */
770         int                             recent_used_cpu;
771         int                             wake_cpu;
772 #endif
773         int                             on_rq;
774
775         int                             prio;
776         int                             static_prio;
777         int                             normal_prio;
778         unsigned int                    rt_priority;
779
780         struct sched_entity             se;
781         struct sched_rt_entity          rt;
782         struct sched_dl_entity          dl;
783         const struct sched_class        *sched_class;
784
785 #ifdef CONFIG_SCHED_CORE
786         struct rb_node                  core_node;
787         unsigned long                   core_cookie;
788         unsigned int                    core_occupation;
789 #endif
790
791 #ifdef CONFIG_CGROUP_SCHED
792         struct task_group               *sched_task_group;
793 #endif
794
795 #ifdef CONFIG_UCLAMP_TASK
796         /*
797          * Clamp values requested for a scheduling entity.
798          * Must be updated with task_rq_lock() held.
799          */
800         struct uclamp_se                uclamp_req[UCLAMP_CNT];
801         /*
802          * Effective clamp values used for a scheduling entity.
803          * Must be updated with task_rq_lock() held.
804          */
805         struct uclamp_se                uclamp[UCLAMP_CNT];
806 #endif
807
808         struct sched_statistics         stats;
809
810 #ifdef CONFIG_PREEMPT_NOTIFIERS
811         /* List of struct preempt_notifier: */
812         struct hlist_head               preempt_notifiers;
813 #endif
814
815 #ifdef CONFIG_BLK_DEV_IO_TRACE
816         unsigned int                    btrace_seq;
817 #endif
818
819         unsigned int                    policy;
820         int                             nr_cpus_allowed;
821         const cpumask_t                 *cpus_ptr;
822         cpumask_t                       *user_cpus_ptr;
823         cpumask_t                       cpus_mask;
824         void                            *migration_pending;
825 #ifdef CONFIG_SMP
826         unsigned short                  migration_disabled;
827 #endif
828         unsigned short                  migration_flags;
829
830 #ifdef CONFIG_PREEMPT_RCU
831         int                             rcu_read_lock_nesting;
832         union rcu_special               rcu_read_unlock_special;
833         struct list_head                rcu_node_entry;
834         struct rcu_node                 *rcu_blocked_node;
835 #endif /* #ifdef CONFIG_PREEMPT_RCU */
836
837 #ifdef CONFIG_TASKS_RCU
838         unsigned long                   rcu_tasks_nvcsw;
839         u8                              rcu_tasks_holdout;
840         u8                              rcu_tasks_idx;
841         int                             rcu_tasks_idle_cpu;
842         struct list_head                rcu_tasks_holdout_list;
843 #endif /* #ifdef CONFIG_TASKS_RCU */
844
845 #ifdef CONFIG_TASKS_TRACE_RCU
846         int                             trc_reader_nesting;
847         int                             trc_ipi_to_cpu;
848         union rcu_special               trc_reader_special;
849         struct list_head                trc_holdout_list;
850         struct list_head                trc_blkd_node;
851         int                             trc_blkd_cpu;
852 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
853
854         struct sched_info               sched_info;
855
856         struct list_head                tasks;
857 #ifdef CONFIG_SMP
858         struct plist_node               pushable_tasks;
859         struct rb_node                  pushable_dl_tasks;
860 #endif
861
862         struct mm_struct                *mm;
863         struct mm_struct                *active_mm;
864
865         /* Per-thread vma caching: */
866         struct vmacache                 vmacache;
867
868 #ifdef SPLIT_RSS_COUNTING
869         struct task_rss_stat            rss_stat;
870 #endif
871         int                             exit_state;
872         int                             exit_code;
873         int                             exit_signal;
874         /* The signal sent when the parent dies: */
875         int                             pdeath_signal;
876         /* JOBCTL_*, siglock protected: */
877         unsigned long                   jobctl;
878
879         /* Used for emulating ABI behavior of previous Linux versions: */
880         unsigned int                    personality;
881
882         /* Scheduler bits, serialized by scheduler locks: */
883         unsigned                        sched_reset_on_fork:1;
884         unsigned                        sched_contributes_to_load:1;
885         unsigned                        sched_migrated:1;
886 #ifdef CONFIG_PSI
887         unsigned                        sched_psi_wake_requeue:1;
888 #endif
889
890         /* Force alignment to the next boundary: */
891         unsigned                        :0;
892
893         /* Unserialized, strictly 'current' */
894
895         /*
896          * This field must not be in the scheduler word above due to wakelist
897          * queueing no longer being serialized by p->on_cpu. However:
898          *
899          * p->XXX = X;                  ttwu()
900          * schedule()                     if (p->on_rq && ..) // false
901          *   smp_mb__after_spinlock();    if (smp_load_acquire(&p->on_cpu) && //true
902          *   deactivate_task()                ttwu_queue_wakelist())
903          *     p->on_rq = 0;                    p->sched_remote_wakeup = Y;
904          *
905          * guarantees all stores of 'current' are visible before
906          * ->sched_remote_wakeup gets used, so it can be in this word.
907          */
908         unsigned                        sched_remote_wakeup:1;
909
910         /* Bit to tell LSMs we're in execve(): */
911         unsigned                        in_execve:1;
912         unsigned                        in_iowait:1;
913 #ifndef TIF_RESTORE_SIGMASK
914         unsigned                        restore_sigmask:1;
915 #endif
916 #ifdef CONFIG_MEMCG
917         unsigned                        in_user_fault:1;
918 #endif
919 #ifdef CONFIG_COMPAT_BRK
920         unsigned                        brk_randomized:1;
921 #endif
922 #ifdef CONFIG_CGROUPS
923         /* disallow userland-initiated cgroup migration */
924         unsigned                        no_cgroup_migration:1;
925         /* task is frozen/stopped (used by the cgroup freezer) */
926         unsigned                        frozen:1;
927 #endif
928 #ifdef CONFIG_BLK_CGROUP
929         unsigned                        use_memdelay:1;
930 #endif
931 #ifdef CONFIG_PSI
932         /* Stalled due to lack of memory */
933         unsigned                        in_memstall:1;
934 #endif
935 #ifdef CONFIG_PAGE_OWNER
936         /* Used by page_owner=on to detect recursion in page tracking. */
937         unsigned                        in_page_owner:1;
938 #endif
939 #ifdef CONFIG_EVENTFD
940         /* Recursion prevention for eventfd_signal() */
941         unsigned                        in_eventfd_signal:1;
942 #endif
943 #ifdef CONFIG_IOMMU_SVA
944         unsigned                        pasid_activated:1;
945 #endif
946 #ifdef  CONFIG_CPU_SUP_INTEL
947         unsigned                        reported_split_lock:1;
948 #endif
949
950         unsigned long                   atomic_flags; /* Flags requiring atomic access. */
951
952         struct restart_block            restart_block;
953
954         pid_t                           pid;
955         pid_t                           tgid;
956
957 #ifdef CONFIG_STACKPROTECTOR
958         /* Canary value for the -fstack-protector GCC feature: */
959         unsigned long                   stack_canary;
960 #endif
961         /*
962          * Pointers to the (original) parent process, youngest child, younger sibling,
963          * older sibling, respectively.  (p->father can be replaced with
964          * p->real_parent->pid)
965          */
966
967         /* Real parent process: */
968         struct task_struct __rcu        *real_parent;
969
970         /* Recipient of SIGCHLD, wait4() reports: */
971         struct task_struct __rcu        *parent;
972
973         /*
974          * Children/sibling form the list of natural children:
975          */
976         struct list_head                children;
977         struct list_head                sibling;
978         struct task_struct              *group_leader;
979
980         /*
981          * 'ptraced' is the list of tasks this task is using ptrace() on.
982          *
983          * This includes both natural children and PTRACE_ATTACH targets.
984          * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
985          */
986         struct list_head                ptraced;
987         struct list_head                ptrace_entry;
988
989         /* PID/PID hash table linkage. */
990         struct pid                      *thread_pid;
991         struct hlist_node               pid_links[PIDTYPE_MAX];
992         struct list_head                thread_group;
993         struct list_head                thread_node;
994
995         struct completion               *vfork_done;
996
997         /* CLONE_CHILD_SETTID: */
998         int __user                      *set_child_tid;
999
1000         /* CLONE_CHILD_CLEARTID: */
1001         int __user                      *clear_child_tid;
1002
1003         /* PF_KTHREAD | PF_IO_WORKER */
1004         void                            *worker_private;
1005
1006         u64                             utime;
1007         u64                             stime;
1008 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1009         u64                             utimescaled;
1010         u64                             stimescaled;
1011 #endif
1012         u64                             gtime;
1013         struct prev_cputime             prev_cputime;
1014 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1015         struct vtime                    vtime;
1016 #endif
1017
1018 #ifdef CONFIG_NO_HZ_FULL
1019         atomic_t                        tick_dep_mask;
1020 #endif
1021         /* Context switch counts: */
1022         unsigned long                   nvcsw;
1023         unsigned long                   nivcsw;
1024
1025         /* Monotonic time in nsecs: */
1026         u64                             start_time;
1027
1028         /* Boot based time in nsecs: */
1029         u64                             start_boottime;
1030
1031         /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1032         unsigned long                   min_flt;
1033         unsigned long                   maj_flt;
1034
1035         /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1036         struct posix_cputimers          posix_cputimers;
1037
1038 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1039         struct posix_cputimers_work     posix_cputimers_work;
1040 #endif
1041
1042         /* Process credentials: */
1043
1044         /* Tracer's credentials at attach: */
1045         const struct cred __rcu         *ptracer_cred;
1046
1047         /* Objective and real subjective task credentials (COW): */
1048         const struct cred __rcu         *real_cred;
1049
1050         /* Effective (overridable) subjective task credentials (COW): */
1051         const struct cred __rcu         *cred;
1052
1053 #ifdef CONFIG_KEYS
1054         /* Cached requested key. */
1055         struct key                      *cached_requested_key;
1056 #endif
1057
1058         /*
1059          * executable name, excluding path.
1060          *
1061          * - normally initialized setup_new_exec()
1062          * - access it with [gs]et_task_comm()
1063          * - lock it with task_lock()
1064          */
1065         char                            comm[TASK_COMM_LEN];
1066
1067         struct nameidata                *nameidata;
1068
1069 #ifdef CONFIG_SYSVIPC
1070         struct sysv_sem                 sysvsem;
1071         struct sysv_shm                 sysvshm;
1072 #endif
1073 #ifdef CONFIG_DETECT_HUNG_TASK
1074         unsigned long                   last_switch_count;
1075         unsigned long                   last_switch_time;
1076 #endif
1077         /* Filesystem information: */
1078         struct fs_struct                *fs;
1079
1080         /* Open file information: */
1081         struct files_struct             *files;
1082
1083 #ifdef CONFIG_IO_URING
1084         struct io_uring_task            *io_uring;
1085 #endif
1086
1087         /* Namespaces: */
1088         struct nsproxy                  *nsproxy;
1089
1090         /* Signal handlers: */
1091         struct signal_struct            *signal;
1092         struct sighand_struct __rcu             *sighand;
1093         sigset_t                        blocked;
1094         sigset_t                        real_blocked;
1095         /* Restored if set_restore_sigmask() was used: */
1096         sigset_t                        saved_sigmask;
1097         struct sigpending               pending;
1098         unsigned long                   sas_ss_sp;
1099         size_t                          sas_ss_size;
1100         unsigned int                    sas_ss_flags;
1101
1102         struct callback_head            *task_works;
1103
1104 #ifdef CONFIG_AUDIT
1105 #ifdef CONFIG_AUDITSYSCALL
1106         struct audit_context            *audit_context;
1107 #endif
1108         kuid_t                          loginuid;
1109         unsigned int                    sessionid;
1110 #endif
1111         struct seccomp                  seccomp;
1112         struct syscall_user_dispatch    syscall_dispatch;
1113
1114         /* Thread group tracking: */
1115         u64                             parent_exec_id;
1116         u64                             self_exec_id;
1117
1118         /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1119         spinlock_t                      alloc_lock;
1120
1121         /* Protection of the PI data structures: */
1122         raw_spinlock_t                  pi_lock;
1123
1124         struct wake_q_node              wake_q;
1125
1126 #ifdef CONFIG_RT_MUTEXES
1127         /* PI waiters blocked on a rt_mutex held by this task: */
1128         struct rb_root_cached           pi_waiters;
1129         /* Updated under owner's pi_lock and rq lock */
1130         struct task_struct              *pi_top_task;
1131         /* Deadlock detection and priority inheritance handling: */
1132         struct rt_mutex_waiter          *pi_blocked_on;
1133 #endif
1134
1135 #ifdef CONFIG_DEBUG_MUTEXES
1136         /* Mutex deadlock detection: */
1137         struct mutex_waiter             *blocked_on;
1138 #endif
1139
1140 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1141         int                             non_block_count;
1142 #endif
1143
1144 #ifdef CONFIG_TRACE_IRQFLAGS
1145         struct irqtrace_events          irqtrace;
1146         unsigned int                    hardirq_threaded;
1147         u64                             hardirq_chain_key;
1148         int                             softirqs_enabled;
1149         int                             softirq_context;
1150         int                             irq_config;
1151 #endif
1152 #ifdef CONFIG_PREEMPT_RT
1153         int                             softirq_disable_cnt;
1154 #endif
1155
1156 #ifdef CONFIG_LOCKDEP
1157 # define MAX_LOCK_DEPTH                 48UL
1158         u64                             curr_chain_key;
1159         int                             lockdep_depth;
1160         unsigned int                    lockdep_recursion;
1161         struct held_lock                held_locks[MAX_LOCK_DEPTH];
1162 #endif
1163
1164 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1165         unsigned int                    in_ubsan;
1166 #endif
1167
1168         /* Journalling filesystem info: */
1169         void                            *journal_info;
1170
1171         /* Stacked block device info: */
1172         struct bio_list                 *bio_list;
1173
1174         /* Stack plugging: */
1175         struct blk_plug                 *plug;
1176
1177         /* VM state: */
1178         struct reclaim_state            *reclaim_state;
1179
1180         struct backing_dev_info         *backing_dev_info;
1181
1182         struct io_context               *io_context;
1183
1184 #ifdef CONFIG_COMPACTION
1185         struct capture_control          *capture_control;
1186 #endif
1187         /* Ptrace state: */
1188         unsigned long                   ptrace_message;
1189         kernel_siginfo_t                *last_siginfo;
1190
1191         struct task_io_accounting       ioac;
1192 #ifdef CONFIG_PSI
1193         /* Pressure stall state */
1194         unsigned int                    psi_flags;
1195 #endif
1196 #ifdef CONFIG_TASK_XACCT
1197         /* Accumulated RSS usage: */
1198         u64                             acct_rss_mem1;
1199         /* Accumulated virtual memory usage: */
1200         u64                             acct_vm_mem1;
1201         /* stime + utime since last update: */
1202         u64                             acct_timexpd;
1203 #endif
1204 #ifdef CONFIG_CPUSETS
1205         /* Protected by ->alloc_lock: */
1206         nodemask_t                      mems_allowed;
1207         /* Sequence number to catch updates: */
1208         seqcount_spinlock_t             mems_allowed_seq;
1209         int                             cpuset_mem_spread_rotor;
1210         int                             cpuset_slab_spread_rotor;
1211 #endif
1212 #ifdef CONFIG_CGROUPS
1213         /* Control Group info protected by css_set_lock: */
1214         struct css_set __rcu            *cgroups;
1215         /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1216         struct list_head                cg_list;
1217 #endif
1218 #ifdef CONFIG_X86_CPU_RESCTRL
1219         u32                             closid;
1220         u32                             rmid;
1221 #endif
1222 #ifdef CONFIG_FUTEX
1223         struct robust_list_head __user  *robust_list;
1224 #ifdef CONFIG_COMPAT
1225         struct compat_robust_list_head __user *compat_robust_list;
1226 #endif
1227         struct list_head                pi_state_list;
1228         struct futex_pi_state           *pi_state_cache;
1229         struct mutex                    futex_exit_mutex;
1230         unsigned int                    futex_state;
1231 #endif
1232 #ifdef CONFIG_PERF_EVENTS
1233         struct perf_event_context       *perf_event_ctxp[perf_nr_task_contexts];
1234         struct mutex                    perf_event_mutex;
1235         struct list_head                perf_event_list;
1236 #endif
1237 #ifdef CONFIG_DEBUG_PREEMPT
1238         unsigned long                   preempt_disable_ip;
1239 #endif
1240 #ifdef CONFIG_NUMA
1241         /* Protected by alloc_lock: */
1242         struct mempolicy                *mempolicy;
1243         short                           il_prev;
1244         short                           pref_node_fork;
1245 #endif
1246 #ifdef CONFIG_NUMA_BALANCING
1247         int                             numa_scan_seq;
1248         unsigned int                    numa_scan_period;
1249         unsigned int                    numa_scan_period_max;
1250         int                             numa_preferred_nid;
1251         unsigned long                   numa_migrate_retry;
1252         /* Migration stamp: */
1253         u64                             node_stamp;
1254         u64                             last_task_numa_placement;
1255         u64                             last_sum_exec_runtime;
1256         struct callback_head            numa_work;
1257
1258         /*
1259          * This pointer is only modified for current in syscall and
1260          * pagefault context (and for tasks being destroyed), so it can be read
1261          * from any of the following contexts:
1262          *  - RCU read-side critical section
1263          *  - current->numa_group from everywhere
1264          *  - task's runqueue locked, task not running
1265          */
1266         struct numa_group __rcu         *numa_group;
1267
1268         /*
1269          * numa_faults is an array split into four regions:
1270          * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1271          * in this precise order.
1272          *
1273          * faults_memory: Exponential decaying average of faults on a per-node
1274          * basis. Scheduling placement decisions are made based on these
1275          * counts. The values remain static for the duration of a PTE scan.
1276          * faults_cpu: Track the nodes the process was running on when a NUMA
1277          * hinting fault was incurred.
1278          * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1279          * during the current scan window. When the scan completes, the counts
1280          * in faults_memory and faults_cpu decay and these values are copied.
1281          */
1282         unsigned long                   *numa_faults;
1283         unsigned long                   total_numa_faults;
1284
1285         /*
1286          * numa_faults_locality tracks if faults recorded during the last
1287          * scan window were remote/local or failed to migrate. The task scan
1288          * period is adapted based on the locality of the faults with different
1289          * weights depending on whether they were shared or private faults
1290          */
1291         unsigned long                   numa_faults_locality[3];
1292
1293         unsigned long                   numa_pages_migrated;
1294 #endif /* CONFIG_NUMA_BALANCING */
1295
1296 #ifdef CONFIG_RSEQ
1297         struct rseq __user *rseq;
1298         u32 rseq_sig;
1299         /*
1300          * RmW on rseq_event_mask must be performed atomically
1301          * with respect to preemption.
1302          */
1303         unsigned long rseq_event_mask;
1304 #endif
1305
1306         struct tlbflush_unmap_batch     tlb_ubc;
1307
1308         union {
1309                 refcount_t              rcu_users;
1310                 struct rcu_head         rcu;
1311         };
1312
1313         /* Cache last used pipe for splice(): */
1314         struct pipe_inode_info          *splice_pipe;
1315
1316         struct page_frag                task_frag;
1317
1318 #ifdef CONFIG_TASK_DELAY_ACCT
1319         struct task_delay_info          *delays;
1320 #endif
1321
1322 #ifdef CONFIG_FAULT_INJECTION
1323         int                             make_it_fail;
1324         unsigned int                    fail_nth;
1325 #endif
1326         /*
1327          * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1328          * balance_dirty_pages() for a dirty throttling pause:
1329          */
1330         int                             nr_dirtied;
1331         int                             nr_dirtied_pause;
1332         /* Start of a write-and-pause period: */
1333         unsigned long                   dirty_paused_when;
1334
1335 #ifdef CONFIG_LATENCYTOP
1336         int                             latency_record_count;
1337         struct latency_record           latency_record[LT_SAVECOUNT];
1338 #endif
1339         /*
1340          * Time slack values; these are used to round up poll() and
1341          * select() etc timeout values. These are in nanoseconds.
1342          */
1343         u64                             timer_slack_ns;
1344         u64                             default_timer_slack_ns;
1345
1346 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1347         unsigned int                    kasan_depth;
1348 #endif
1349
1350 #ifdef CONFIG_KCSAN
1351         struct kcsan_ctx                kcsan_ctx;
1352 #ifdef CONFIG_TRACE_IRQFLAGS
1353         struct irqtrace_events          kcsan_save_irqtrace;
1354 #endif
1355 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1356         int                             kcsan_stack_depth;
1357 #endif
1358 #endif
1359
1360 #if IS_ENABLED(CONFIG_KUNIT)
1361         struct kunit                    *kunit_test;
1362 #endif
1363
1364 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1365         /* Index of current stored address in ret_stack: */
1366         int                             curr_ret_stack;
1367         int                             curr_ret_depth;
1368
1369         /* Stack of return addresses for return function tracing: */
1370         struct ftrace_ret_stack         *ret_stack;
1371
1372         /* Timestamp for last schedule: */
1373         unsigned long long              ftrace_timestamp;
1374
1375         /*
1376          * Number of functions that haven't been traced
1377          * because of depth overrun:
1378          */
1379         atomic_t                        trace_overrun;
1380
1381         /* Pause tracing: */
1382         atomic_t                        tracing_graph_pause;
1383 #endif
1384
1385 #ifdef CONFIG_TRACING
1386         /* State flags for use by tracers: */
1387         unsigned long                   trace;
1388
1389         /* Bitmask and counter of trace recursion: */
1390         unsigned long                   trace_recursion;
1391 #endif /* CONFIG_TRACING */
1392
1393 #ifdef CONFIG_KCOV
1394         /* See kernel/kcov.c for more details. */
1395
1396         /* Coverage collection mode enabled for this task (0 if disabled): */
1397         unsigned int                    kcov_mode;
1398
1399         /* Size of the kcov_area: */
1400         unsigned int                    kcov_size;
1401
1402         /* Buffer for coverage collection: */
1403         void                            *kcov_area;
1404
1405         /* KCOV descriptor wired with this task or NULL: */
1406         struct kcov                     *kcov;
1407
1408         /* KCOV common handle for remote coverage collection: */
1409         u64                             kcov_handle;
1410
1411         /* KCOV sequence number: */
1412         int                             kcov_sequence;
1413
1414         /* Collect coverage from softirq context: */
1415         unsigned int                    kcov_softirq;
1416 #endif
1417
1418 #ifdef CONFIG_MEMCG
1419         struct mem_cgroup               *memcg_in_oom;
1420         gfp_t                           memcg_oom_gfp_mask;
1421         int                             memcg_oom_order;
1422
1423         /* Number of pages to reclaim on returning to userland: */
1424         unsigned int                    memcg_nr_pages_over_high;
1425
1426         /* Used by memcontrol for targeted memcg charge: */
1427         struct mem_cgroup               *active_memcg;
1428 #endif
1429
1430 #ifdef CONFIG_BLK_CGROUP
1431         struct request_queue            *throttle_queue;
1432 #endif
1433
1434 #ifdef CONFIG_UPROBES
1435         struct uprobe_task              *utask;
1436 #endif
1437 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1438         unsigned int                    sequential_io;
1439         unsigned int                    sequential_io_avg;
1440 #endif
1441         struct kmap_ctrl                kmap_ctrl;
1442 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1443         unsigned long                   task_state_change;
1444 # ifdef CONFIG_PREEMPT_RT
1445         unsigned long                   saved_state_change;
1446 # endif
1447 #endif
1448         int                             pagefault_disabled;
1449 #ifdef CONFIG_MMU
1450         struct task_struct              *oom_reaper_list;
1451         struct timer_list               oom_reaper_timer;
1452 #endif
1453 #ifdef CONFIG_VMAP_STACK
1454         struct vm_struct                *stack_vm_area;
1455 #endif
1456 #ifdef CONFIG_THREAD_INFO_IN_TASK
1457         /* A live task holds one reference: */
1458         refcount_t                      stack_refcount;
1459 #endif
1460 #ifdef CONFIG_LIVEPATCH
1461         int patch_state;
1462 #endif
1463 #ifdef CONFIG_SECURITY
1464         /* Used by LSM modules for access restriction: */
1465         void                            *security;
1466 #endif
1467 #ifdef CONFIG_BPF_SYSCALL
1468         /* Used by BPF task local storage */
1469         struct bpf_local_storage __rcu  *bpf_storage;
1470         /* Used for BPF run context */
1471         struct bpf_run_ctx              *bpf_ctx;
1472 #endif
1473
1474 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1475         unsigned long                   lowest_stack;
1476         unsigned long                   prev_lowest_stack;
1477 #endif
1478
1479 #ifdef CONFIG_X86_MCE
1480         void __user                     *mce_vaddr;
1481         __u64                           mce_kflags;
1482         u64                             mce_addr;
1483         __u64                           mce_ripv : 1,
1484                                         mce_whole_page : 1,
1485                                         __mce_reserved : 62;
1486         struct callback_head            mce_kill_me;
1487         int                             mce_count;
1488 #endif
1489
1490 #ifdef CONFIG_KRETPROBES
1491         struct llist_head               kretprobe_instances;
1492 #endif
1493 #ifdef CONFIG_RETHOOK
1494         struct llist_head               rethooks;
1495 #endif
1496
1497 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1498         /*
1499          * If L1D flush is supported on mm context switch
1500          * then we use this callback head to queue kill work
1501          * to kill tasks that are not running on SMT disabled
1502          * cores
1503          */
1504         struct callback_head            l1d_flush_kill;
1505 #endif
1506
1507 #ifdef CONFIG_RV
1508         /*
1509          * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1510          * If we find justification for more monitors, we can think
1511          * about adding more or developing a dynamic method. So far,
1512          * none of these are justified.
1513          */
1514         union rv_task_monitor           rv[RV_PER_TASK_MONITORS];
1515 #endif
1516
1517         /*
1518          * New fields for task_struct should be added above here, so that
1519          * they are included in the randomized portion of task_struct.
1520          */
1521         randomized_struct_fields_end
1522
1523         /* CPU-specific state of this task: */
1524         struct thread_struct            thread;
1525
1526         /*
1527          * WARNING: on x86, 'thread_struct' contains a variable-sized
1528          * structure.  It *MUST* be at the end of 'task_struct'.
1529          *
1530          * Do not put anything below here!
1531          */
1532 };
1533
1534 static inline struct pid *task_pid(struct task_struct *task)
1535 {
1536         return task->thread_pid;
1537 }
1538
1539 /*
1540  * the helpers to get the task's different pids as they are seen
1541  * from various namespaces
1542  *
1543  * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1544  * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1545  *                     current.
1546  * task_xid_nr_ns()  : id seen from the ns specified;
1547  *
1548  * see also pid_nr() etc in include/linux/pid.h
1549  */
1550 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1551
1552 static inline pid_t task_pid_nr(struct task_struct *tsk)
1553 {
1554         return tsk->pid;
1555 }
1556
1557 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1558 {
1559         return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1560 }
1561
1562 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1563 {
1564         return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1565 }
1566
1567
1568 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1569 {
1570         return tsk->tgid;
1571 }
1572
1573 /**
1574  * pid_alive - check that a task structure is not stale
1575  * @p: Task structure to be checked.
1576  *
1577  * Test if a process is not yet dead (at most zombie state)
1578  * If pid_alive fails, then pointers within the task structure
1579  * can be stale and must not be dereferenced.
1580  *
1581  * Return: 1 if the process is alive. 0 otherwise.
1582  */
1583 static inline int pid_alive(const struct task_struct *p)
1584 {
1585         return p->thread_pid != NULL;
1586 }
1587
1588 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1589 {
1590         return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1591 }
1592
1593 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1594 {
1595         return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1596 }
1597
1598
1599 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1600 {
1601         return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1602 }
1603
1604 static inline pid_t task_session_vnr(struct task_struct *tsk)
1605 {
1606         return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1607 }
1608
1609 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1610 {
1611         return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1612 }
1613
1614 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1615 {
1616         return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1617 }
1618
1619 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1620 {
1621         pid_t pid = 0;
1622
1623         rcu_read_lock();
1624         if (pid_alive(tsk))
1625                 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1626         rcu_read_unlock();
1627
1628         return pid;
1629 }
1630
1631 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1632 {
1633         return task_ppid_nr_ns(tsk, &init_pid_ns);
1634 }
1635
1636 /* Obsolete, do not use: */
1637 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1638 {
1639         return task_pgrp_nr_ns(tsk, &init_pid_ns);
1640 }
1641
1642 #define TASK_REPORT_IDLE        (TASK_REPORT + 1)
1643 #define TASK_REPORT_MAX         (TASK_REPORT_IDLE << 1)
1644
1645 static inline unsigned int __task_state_index(unsigned int tsk_state,
1646                                               unsigned int tsk_exit_state)
1647 {
1648         unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1649
1650         BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1651
1652         if (tsk_state == TASK_IDLE)
1653                 state = TASK_REPORT_IDLE;
1654
1655         /*
1656          * We're lying here, but rather than expose a completely new task state
1657          * to userspace, we can make this appear as if the task has gone through
1658          * a regular rt_mutex_lock() call.
1659          */
1660         if (tsk_state == TASK_RTLOCK_WAIT)
1661                 state = TASK_UNINTERRUPTIBLE;
1662
1663         return fls(state);
1664 }
1665
1666 static inline unsigned int task_state_index(struct task_struct *tsk)
1667 {
1668         return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1669 }
1670
1671 static inline char task_index_to_char(unsigned int state)
1672 {
1673         static const char state_char[] = "RSDTtXZPI";
1674
1675         BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1676
1677         return state_char[state];
1678 }
1679
1680 static inline char task_state_to_char(struct task_struct *tsk)
1681 {
1682         return task_index_to_char(task_state_index(tsk));
1683 }
1684
1685 /**
1686  * is_global_init - check if a task structure is init. Since init
1687  * is free to have sub-threads we need to check tgid.
1688  * @tsk: Task structure to be checked.
1689  *
1690  * Check if a task structure is the first user space task the kernel created.
1691  *
1692  * Return: 1 if the task structure is init. 0 otherwise.
1693  */
1694 static inline int is_global_init(struct task_struct *tsk)
1695 {
1696         return task_tgid_nr(tsk) == 1;
1697 }
1698
1699 extern struct pid *cad_pid;
1700
1701 /*
1702  * Per process flags
1703  */
1704 #define PF_VCPU                 0x00000001      /* I'm a virtual CPU */
1705 #define PF_IDLE                 0x00000002      /* I am an IDLE thread */
1706 #define PF_EXITING              0x00000004      /* Getting shut down */
1707 #define PF_POSTCOREDUMP         0x00000008      /* Coredumps should ignore this task */
1708 #define PF_IO_WORKER            0x00000010      /* Task is an IO worker */
1709 #define PF_WQ_WORKER            0x00000020      /* I'm a workqueue worker */
1710 #define PF_FORKNOEXEC           0x00000040      /* Forked but didn't exec */
1711 #define PF_MCE_PROCESS          0x00000080      /* Process policy on mce errors */
1712 #define PF_SUPERPRIV            0x00000100      /* Used super-user privileges */
1713 #define PF_DUMPCORE             0x00000200      /* Dumped core */
1714 #define PF_SIGNALED             0x00000400      /* Killed by a signal */
1715 #define PF_MEMALLOC             0x00000800      /* Allocating memory */
1716 #define PF_NPROC_EXCEEDED       0x00001000      /* set_user() noticed that RLIMIT_NPROC was exceeded */
1717 #define PF_USED_MATH            0x00002000      /* If unset the fpu must be initialized before use */
1718 #define PF_NOFREEZE             0x00008000      /* This thread should not be frozen */
1719 #define PF_FROZEN               0x00010000      /* Frozen for system suspend */
1720 #define PF_KSWAPD               0x00020000      /* I am kswapd */
1721 #define PF_MEMALLOC_NOFS        0x00040000      /* All allocation requests will inherit GFP_NOFS */
1722 #define PF_MEMALLOC_NOIO        0x00080000      /* All allocation requests will inherit GFP_NOIO */
1723 #define PF_LOCAL_THROTTLE       0x00100000      /* Throttle writes only against the bdi I write to,
1724                                                  * I am cleaning dirty pages from some other bdi. */
1725 #define PF_KTHREAD              0x00200000      /* I am a kernel thread */
1726 #define PF_RANDOMIZE            0x00400000      /* Randomize virtual address space */
1727 #define PF_NO_SETAFFINITY       0x04000000      /* Userland is not allowed to meddle with cpus_mask */
1728 #define PF_MCE_EARLY            0x08000000      /* Early kill for mce process policy */
1729 #define PF_MEMALLOC_PIN         0x10000000      /* Allocation context constrained to zones which allow long term pinning. */
1730 #define PF_FREEZER_SKIP         0x40000000      /* Freezer should not count it as freezable */
1731 #define PF_SUSPEND_TASK         0x80000000      /* This thread called freeze_processes() and should not be frozen */
1732
1733 /*
1734  * Only the _current_ task can read/write to tsk->flags, but other
1735  * tasks can access tsk->flags in readonly mode for example
1736  * with tsk_used_math (like during threaded core dumping).
1737  * There is however an exception to this rule during ptrace
1738  * or during fork: the ptracer task is allowed to write to the
1739  * child->flags of its traced child (same goes for fork, the parent
1740  * can write to the child->flags), because we're guaranteed the
1741  * child is not running and in turn not changing child->flags
1742  * at the same time the parent does it.
1743  */
1744 #define clear_stopped_child_used_math(child)    do { (child)->flags &= ~PF_USED_MATH; } while (0)
1745 #define set_stopped_child_used_math(child)      do { (child)->flags |= PF_USED_MATH; } while (0)
1746 #define clear_used_math()                       clear_stopped_child_used_math(current)
1747 #define set_used_math()                         set_stopped_child_used_math(current)
1748
1749 #define conditional_stopped_child_used_math(condition, child) \
1750         do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1751
1752 #define conditional_used_math(condition)        conditional_stopped_child_used_math(condition, current)
1753
1754 #define copy_to_stopped_child_used_math(child) \
1755         do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1756
1757 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1758 #define tsk_used_math(p)                        ((p)->flags & PF_USED_MATH)
1759 #define used_math()                             tsk_used_math(current)
1760
1761 static __always_inline bool is_percpu_thread(void)
1762 {
1763 #ifdef CONFIG_SMP
1764         return (current->flags & PF_NO_SETAFFINITY) &&
1765                 (current->nr_cpus_allowed  == 1);
1766 #else
1767         return true;
1768 #endif
1769 }
1770
1771 /* Per-process atomic flags. */
1772 #define PFA_NO_NEW_PRIVS                0       /* May not gain new privileges. */
1773 #define PFA_SPREAD_PAGE                 1       /* Spread page cache over cpuset */
1774 #define PFA_SPREAD_SLAB                 2       /* Spread some slab caches over cpuset */
1775 #define PFA_SPEC_SSB_DISABLE            3       /* Speculative Store Bypass disabled */
1776 #define PFA_SPEC_SSB_FORCE_DISABLE      4       /* Speculative Store Bypass force disabled*/
1777 #define PFA_SPEC_IB_DISABLE             5       /* Indirect branch speculation restricted */
1778 #define PFA_SPEC_IB_FORCE_DISABLE       6       /* Indirect branch speculation permanently restricted */
1779 #define PFA_SPEC_SSB_NOEXEC             7       /* Speculative Store Bypass clear on execve() */
1780
1781 #define TASK_PFA_TEST(name, func)                                       \
1782         static inline bool task_##func(struct task_struct *p)           \
1783         { return test_bit(PFA_##name, &p->atomic_flags); }
1784
1785 #define TASK_PFA_SET(name, func)                                        \
1786         static inline void task_set_##func(struct task_struct *p)       \
1787         { set_bit(PFA_##name, &p->atomic_flags); }
1788
1789 #define TASK_PFA_CLEAR(name, func)                                      \
1790         static inline void task_clear_##func(struct task_struct *p)     \
1791         { clear_bit(PFA_##name, &p->atomic_flags); }
1792
1793 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1794 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1795
1796 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1797 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1798 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1799
1800 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1801 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1802 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1803
1804 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1805 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1806 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1807
1808 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1809 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1810 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1811
1812 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1813 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1814
1815 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1816 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1817 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1818
1819 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1820 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1821
1822 static inline void
1823 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1824 {
1825         current->flags &= ~flags;
1826         current->flags |= orig_flags & flags;
1827 }
1828
1829 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1830 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effective_cpus);
1831 #ifdef CONFIG_SMP
1832 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1833 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1834 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1835 extern void release_user_cpus_ptr(struct task_struct *p);
1836 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1837 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1838 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1839 #else
1840 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1841 {
1842 }
1843 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1844 {
1845         if (!cpumask_test_cpu(0, new_mask))
1846                 return -EINVAL;
1847         return 0;
1848 }
1849 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1850 {
1851         if (src->user_cpus_ptr)
1852                 return -EINVAL;
1853         return 0;
1854 }
1855 static inline void release_user_cpus_ptr(struct task_struct *p)
1856 {
1857         WARN_ON(p->user_cpus_ptr);
1858 }
1859
1860 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1861 {
1862         return 0;
1863 }
1864 #endif
1865
1866 extern int yield_to(struct task_struct *p, bool preempt);
1867 extern void set_user_nice(struct task_struct *p, long nice);
1868 extern int task_prio(const struct task_struct *p);
1869
1870 /**
1871  * task_nice - return the nice value of a given task.
1872  * @p: the task in question.
1873  *
1874  * Return: The nice value [ -20 ... 0 ... 19 ].
1875  */
1876 static inline int task_nice(const struct task_struct *p)
1877 {
1878         return PRIO_TO_NICE((p)->static_prio);
1879 }
1880
1881 extern int can_nice(const struct task_struct *p, const int nice);
1882 extern int task_curr(const struct task_struct *p);
1883 extern int idle_cpu(int cpu);
1884 extern int available_idle_cpu(int cpu);
1885 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1886 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1887 extern void sched_set_fifo(struct task_struct *p);
1888 extern void sched_set_fifo_low(struct task_struct *p);
1889 extern void sched_set_normal(struct task_struct *p, int nice);
1890 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1891 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1892 extern struct task_struct *idle_task(int cpu);
1893
1894 /**
1895  * is_idle_task - is the specified task an idle task?
1896  * @p: the task in question.
1897  *
1898  * Return: 1 if @p is an idle task. 0 otherwise.
1899  */
1900 static __always_inline bool is_idle_task(const struct task_struct *p)
1901 {
1902         return !!(p->flags & PF_IDLE);
1903 }
1904
1905 extern struct task_struct *curr_task(int cpu);
1906 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1907
1908 void yield(void);
1909
1910 union thread_union {
1911 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1912         struct task_struct task;
1913 #endif
1914 #ifndef CONFIG_THREAD_INFO_IN_TASK
1915         struct thread_info thread_info;
1916 #endif
1917         unsigned long stack[THREAD_SIZE/sizeof(long)];
1918 };
1919
1920 #ifndef CONFIG_THREAD_INFO_IN_TASK
1921 extern struct thread_info init_thread_info;
1922 #endif
1923
1924 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1925
1926 #ifdef CONFIG_THREAD_INFO_IN_TASK
1927 # define task_thread_info(task) (&(task)->thread_info)
1928 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1929 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1930 #endif
1931
1932 /*
1933  * find a task by one of its numerical ids
1934  *
1935  * find_task_by_pid_ns():
1936  *      finds a task by its pid in the specified namespace
1937  * find_task_by_vpid():
1938  *      finds a task by its virtual pid
1939  *
1940  * see also find_vpid() etc in include/linux/pid.h
1941  */
1942
1943 extern struct task_struct *find_task_by_vpid(pid_t nr);
1944 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1945
1946 /*
1947  * find a task by its virtual pid and get the task struct
1948  */
1949 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1950
1951 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1952 extern int wake_up_process(struct task_struct *tsk);
1953 extern void wake_up_new_task(struct task_struct *tsk);
1954
1955 #ifdef CONFIG_SMP
1956 extern void kick_process(struct task_struct *tsk);
1957 #else
1958 static inline void kick_process(struct task_struct *tsk) { }
1959 #endif
1960
1961 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1962
1963 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1964 {
1965         __set_task_comm(tsk, from, false);
1966 }
1967
1968 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1969 #define get_task_comm(buf, tsk) ({                      \
1970         BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);     \
1971         __get_task_comm(buf, sizeof(buf), tsk);         \
1972 })
1973
1974 #ifdef CONFIG_SMP
1975 static __always_inline void scheduler_ipi(void)
1976 {
1977         /*
1978          * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1979          * TIF_NEED_RESCHED remotely (for the first time) will also send
1980          * this IPI.
1981          */
1982         preempt_fold_need_resched();
1983 }
1984 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1985 #else
1986 static inline void scheduler_ipi(void) { }
1987 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1988 {
1989         return 1;
1990 }
1991 #endif
1992
1993 /*
1994  * Set thread flags in other task's structures.
1995  * See asm/thread_info.h for TIF_xxxx flags available:
1996  */
1997 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1998 {
1999         set_ti_thread_flag(task_thread_info(tsk), flag);
2000 }
2001
2002 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2003 {
2004         clear_ti_thread_flag(task_thread_info(tsk), flag);
2005 }
2006
2007 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2008                                           bool value)
2009 {
2010         update_ti_thread_flag(task_thread_info(tsk), flag, value);
2011 }
2012
2013 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2014 {
2015         return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2016 }
2017
2018 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2019 {
2020         return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2021 }
2022
2023 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2024 {
2025         return test_ti_thread_flag(task_thread_info(tsk), flag);
2026 }
2027
2028 static inline void set_tsk_need_resched(struct task_struct *tsk)
2029 {
2030         set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2031 }
2032
2033 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2034 {
2035         clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2036 }
2037
2038 static inline int test_tsk_need_resched(struct task_struct *tsk)
2039 {
2040         return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2041 }
2042
2043 /*
2044  * cond_resched() and cond_resched_lock(): latency reduction via
2045  * explicit rescheduling in places that are safe. The return
2046  * value indicates whether a reschedule was done in fact.
2047  * cond_resched_lock() will drop the spinlock before scheduling,
2048  */
2049 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2050 extern int __cond_resched(void);
2051
2052 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2053
2054 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2055
2056 static __always_inline int _cond_resched(void)
2057 {
2058         return static_call_mod(cond_resched)();
2059 }
2060
2061 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2062 extern int dynamic_cond_resched(void);
2063
2064 static __always_inline int _cond_resched(void)
2065 {
2066         return dynamic_cond_resched();
2067 }
2068
2069 #else
2070
2071 static inline int _cond_resched(void)
2072 {
2073         return __cond_resched();
2074 }
2075
2076 #endif /* CONFIG_PREEMPT_DYNAMIC */
2077
2078 #else
2079
2080 static inline int _cond_resched(void) { return 0; }
2081
2082 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2083
2084 #define cond_resched() ({                       \
2085         __might_resched(__FILE__, __LINE__, 0); \
2086         _cond_resched();                        \
2087 })
2088
2089 extern int __cond_resched_lock(spinlock_t *lock);
2090 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2091 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2092
2093 #define MIGHT_RESCHED_RCU_SHIFT         8
2094 #define MIGHT_RESCHED_PREEMPT_MASK      ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2095
2096 #ifndef CONFIG_PREEMPT_RT
2097 /*
2098  * Non RT kernels have an elevated preempt count due to the held lock,
2099  * but are not allowed to be inside a RCU read side critical section
2100  */
2101 # define PREEMPT_LOCK_RESCHED_OFFSETS   PREEMPT_LOCK_OFFSET
2102 #else
2103 /*
2104  * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2105  * cond_resched*lock() has to take that into account because it checks for
2106  * preempt_count() and rcu_preempt_depth().
2107  */
2108 # define PREEMPT_LOCK_RESCHED_OFFSETS   \
2109         (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2110 #endif
2111
2112 #define cond_resched_lock(lock) ({                                              \
2113         __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);      \
2114         __cond_resched_lock(lock);                                              \
2115 })
2116
2117 #define cond_resched_rwlock_read(lock) ({                                       \
2118         __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);      \
2119         __cond_resched_rwlock_read(lock);                                       \
2120 })
2121
2122 #define cond_resched_rwlock_write(lock) ({                                      \
2123         __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS);      \
2124         __cond_resched_rwlock_write(lock);                                      \
2125 })
2126
2127 static inline void cond_resched_rcu(void)
2128 {
2129 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2130         rcu_read_unlock();
2131         cond_resched();
2132         rcu_read_lock();
2133 #endif
2134 }
2135
2136 #ifdef CONFIG_PREEMPT_DYNAMIC
2137
2138 extern bool preempt_model_none(void);
2139 extern bool preempt_model_voluntary(void);
2140 extern bool preempt_model_full(void);
2141
2142 #else
2143
2144 static inline bool preempt_model_none(void)
2145 {
2146         return IS_ENABLED(CONFIG_PREEMPT_NONE);
2147 }
2148 static inline bool preempt_model_voluntary(void)
2149 {
2150         return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2151 }
2152 static inline bool preempt_model_full(void)
2153 {
2154         return IS_ENABLED(CONFIG_PREEMPT);
2155 }
2156
2157 #endif
2158
2159 static inline bool preempt_model_rt(void)
2160 {
2161         return IS_ENABLED(CONFIG_PREEMPT_RT);
2162 }
2163
2164 /*
2165  * Does the preemption model allow non-cooperative preemption?
2166  *
2167  * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2168  * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2169  * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2170  * PREEMPT_NONE model.
2171  */
2172 static inline bool preempt_model_preemptible(void)
2173 {
2174         return preempt_model_full() || preempt_model_rt();
2175 }
2176
2177 /*
2178  * Does a critical section need to be broken due to another
2179  * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2180  * but a general need for low latency)
2181  */
2182 static inline int spin_needbreak(spinlock_t *lock)
2183 {
2184 #ifdef CONFIG_PREEMPTION
2185         return spin_is_contended(lock);
2186 #else
2187         return 0;
2188 #endif
2189 }
2190
2191 /*
2192  * Check if a rwlock is contended.
2193  * Returns non-zero if there is another task waiting on the rwlock.
2194  * Returns zero if the lock is not contended or the system / underlying
2195  * rwlock implementation does not support contention detection.
2196  * Technically does not depend on CONFIG_PREEMPTION, but a general need
2197  * for low latency.
2198  */
2199 static inline int rwlock_needbreak(rwlock_t *lock)
2200 {
2201 #ifdef CONFIG_PREEMPTION
2202         return rwlock_is_contended(lock);
2203 #else
2204         return 0;
2205 #endif
2206 }
2207
2208 static __always_inline bool need_resched(void)
2209 {
2210         return unlikely(tif_need_resched());
2211 }
2212
2213 /*
2214  * Wrappers for p->thread_info->cpu access. No-op on UP.
2215  */
2216 #ifdef CONFIG_SMP
2217
2218 static inline unsigned int task_cpu(const struct task_struct *p)
2219 {
2220         return READ_ONCE(task_thread_info(p)->cpu);
2221 }
2222
2223 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2224
2225 #else
2226
2227 static inline unsigned int task_cpu(const struct task_struct *p)
2228 {
2229         return 0;
2230 }
2231
2232 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2233 {
2234 }
2235
2236 #endif /* CONFIG_SMP */
2237
2238 extern bool sched_task_on_rq(struct task_struct *p);
2239 extern unsigned long get_wchan(struct task_struct *p);
2240 extern struct task_struct *cpu_curr_snapshot(int cpu);
2241
2242 /*
2243  * In order to reduce various lock holder preemption latencies provide an
2244  * interface to see if a vCPU is currently running or not.
2245  *
2246  * This allows us to terminate optimistic spin loops and block, analogous to
2247  * the native optimistic spin heuristic of testing if the lock owner task is
2248  * running or not.
2249  */
2250 #ifndef vcpu_is_preempted
2251 static inline bool vcpu_is_preempted(int cpu)
2252 {
2253         return false;
2254 }
2255 #endif
2256
2257 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2258 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2259
2260 #ifndef TASK_SIZE_OF
2261 #define TASK_SIZE_OF(tsk)       TASK_SIZE
2262 #endif
2263
2264 #ifdef CONFIG_SMP
2265 static inline bool owner_on_cpu(struct task_struct *owner)
2266 {
2267         /*
2268          * As lock holder preemption issue, we both skip spinning if
2269          * task is not on cpu or its cpu is preempted
2270          */
2271         return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2272 }
2273
2274 /* Returns effective CPU energy utilization, as seen by the scheduler */
2275 unsigned long sched_cpu_util(int cpu);
2276 #endif /* CONFIG_SMP */
2277
2278 #ifdef CONFIG_RSEQ
2279
2280 /*
2281  * Map the event mask on the user-space ABI enum rseq_cs_flags
2282  * for direct mask checks.
2283  */
2284 enum rseq_event_mask_bits {
2285         RSEQ_EVENT_PREEMPT_BIT  = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2286         RSEQ_EVENT_SIGNAL_BIT   = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2287         RSEQ_EVENT_MIGRATE_BIT  = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2288 };
2289
2290 enum rseq_event_mask {
2291         RSEQ_EVENT_PREEMPT      = (1U << RSEQ_EVENT_PREEMPT_BIT),
2292         RSEQ_EVENT_SIGNAL       = (1U << RSEQ_EVENT_SIGNAL_BIT),
2293         RSEQ_EVENT_MIGRATE      = (1U << RSEQ_EVENT_MIGRATE_BIT),
2294 };
2295
2296 static inline void rseq_set_notify_resume(struct task_struct *t)
2297 {
2298         if (t->rseq)
2299                 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2300 }
2301
2302 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2303
2304 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2305                                              struct pt_regs *regs)
2306 {
2307         if (current->rseq)
2308                 __rseq_handle_notify_resume(ksig, regs);
2309 }
2310
2311 static inline void rseq_signal_deliver(struct ksignal *ksig,
2312                                        struct pt_regs *regs)
2313 {
2314         preempt_disable();
2315         __set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2316         preempt_enable();
2317         rseq_handle_notify_resume(ksig, regs);
2318 }
2319
2320 /* rseq_preempt() requires preemption to be disabled. */
2321 static inline void rseq_preempt(struct task_struct *t)
2322 {
2323         __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2324         rseq_set_notify_resume(t);
2325 }
2326
2327 /* rseq_migrate() requires preemption to be disabled. */
2328 static inline void rseq_migrate(struct task_struct *t)
2329 {
2330         __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2331         rseq_set_notify_resume(t);
2332 }
2333
2334 /*
2335  * If parent process has a registered restartable sequences area, the
2336  * child inherits. Unregister rseq for a clone with CLONE_VM set.
2337  */
2338 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2339 {
2340         if (clone_flags & CLONE_VM) {
2341                 t->rseq = NULL;
2342                 t->rseq_sig = 0;
2343                 t->rseq_event_mask = 0;
2344         } else {
2345                 t->rseq = current->rseq;
2346                 t->rseq_sig = current->rseq_sig;
2347                 t->rseq_event_mask = current->rseq_event_mask;
2348         }
2349 }
2350
2351 static inline void rseq_execve(struct task_struct *t)
2352 {
2353         t->rseq = NULL;
2354         t->rseq_sig = 0;
2355         t->rseq_event_mask = 0;
2356 }
2357
2358 #else
2359
2360 static inline void rseq_set_notify_resume(struct task_struct *t)
2361 {
2362 }
2363 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2364                                              struct pt_regs *regs)
2365 {
2366 }
2367 static inline void rseq_signal_deliver(struct ksignal *ksig,
2368                                        struct pt_regs *regs)
2369 {
2370 }
2371 static inline void rseq_preempt(struct task_struct *t)
2372 {
2373 }
2374 static inline void rseq_migrate(struct task_struct *t)
2375 {
2376 }
2377 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2378 {
2379 }
2380 static inline void rseq_execve(struct task_struct *t)
2381 {
2382 }
2383
2384 #endif
2385
2386 #ifdef CONFIG_DEBUG_RSEQ
2387
2388 void rseq_syscall(struct pt_regs *regs);
2389
2390 #else
2391
2392 static inline void rseq_syscall(struct pt_regs *regs)
2393 {
2394 }
2395
2396 #endif
2397
2398 #ifdef CONFIG_SCHED_CORE
2399 extern void sched_core_free(struct task_struct *tsk);
2400 extern void sched_core_fork(struct task_struct *p);
2401 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2402                                 unsigned long uaddr);
2403 #else
2404 static inline void sched_core_free(struct task_struct *tsk) { }
2405 static inline void sched_core_fork(struct task_struct *p) { }
2406 #endif
2407
2408 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2409
2410 #endif