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