linux/include/linux/sched.h
<<
>>
Prefs
   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 <asm/kmap_size.h>
  38
  39/* task_struct member predeclarations (sorted alphabetically): */
  40struct audit_context;
  41struct backing_dev_info;
  42struct bio_list;
  43struct blk_plug;
  44struct bpf_local_storage;
  45struct capture_control;
  46struct cfs_rq;
  47struct fs_struct;
  48struct futex_pi_state;
  49struct io_context;
  50struct io_uring_task;
  51struct mempolicy;
  52struct nameidata;
  53struct nsproxy;
  54struct perf_event_context;
  55struct pid_namespace;
  56struct pipe_inode_info;
  57struct rcu_node;
  58struct reclaim_state;
  59struct robust_list_head;
  60struct root_domain;
  61struct rq;
  62struct sched_attr;
  63struct sched_param;
  64struct seq_file;
  65struct sighand_struct;
  66struct signal_struct;
  67struct task_delay_info;
  68struct task_group;
  69
  70/*
  71 * Task state bitmask. NOTE! These bits are also
  72 * encoded in fs/proc/array.c: get_task_state().
  73 *
  74 * We have two separate sets of flags: task->state
  75 * is about runnability, while task->exit_state are
  76 * about the task exiting. Confusing, but this way
  77 * modifying one set can't modify the other one by
  78 * mistake.
  79 */
  80
  81/* Used in tsk->state: */
  82#define TASK_RUNNING                    0x0000
  83#define TASK_INTERRUPTIBLE              0x0001
  84#define TASK_UNINTERRUPTIBLE            0x0002
  85#define __TASK_STOPPED                  0x0004
  86#define __TASK_TRACED                   0x0008
  87/* Used in tsk->exit_state: */
  88#define EXIT_DEAD                       0x0010
  89#define EXIT_ZOMBIE                     0x0020
  90#define EXIT_TRACE                      (EXIT_ZOMBIE | EXIT_DEAD)
  91/* Used in tsk->state again: */
  92#define TASK_PARKED                     0x0040
  93#define TASK_DEAD                       0x0080
  94#define TASK_WAKEKILL                   0x0100
  95#define TASK_WAKING                     0x0200
  96#define TASK_NOLOAD                     0x0400
  97#define TASK_NEW                        0x0800
  98#define TASK_STATE_MAX                  0x1000
  99
 100/* Convenience macros for the sake of set_current_state: */
 101#define TASK_KILLABLE                   (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
 102#define TASK_STOPPED                    (TASK_WAKEKILL | __TASK_STOPPED)
 103#define TASK_TRACED                     (TASK_WAKEKILL | __TASK_TRACED)
 104
 105#define TASK_IDLE                       (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
 106
 107/* Convenience macros for the sake of wake_up(): */
 108#define TASK_NORMAL                     (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
 109
 110/* get_task_state(): */
 111#define TASK_REPORT                     (TASK_RUNNING | TASK_INTERRUPTIBLE | \
 112                                         TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
 113                                         __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
 114                                         TASK_PARKED)
 115
 116#define task_is_running(task)           (READ_ONCE((task)->__state) == TASK_RUNNING)
 117
 118#define task_is_traced(task)            ((READ_ONCE(task->__state) & __TASK_TRACED) != 0)
 119
 120#define task_is_stopped(task)           ((READ_ONCE(task->__state) & __TASK_STOPPED) != 0)
 121
 122#define task_is_stopped_or_traced(task) ((READ_ONCE(task->__state) & (__TASK_STOPPED | __TASK_TRACED)) != 0)
 123
 124#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
 125
 126/*
 127 * Special states are those that do not use the normal wait-loop pattern. See
 128 * the comment with set_special_state().
 129 */
 130#define is_special_task_state(state)                            \
 131        ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
 132
 133#define __set_current_state(state_value)                        \
 134        do {                                                    \
 135                WARN_ON_ONCE(is_special_task_state(state_value));\
 136                current->task_state_change = _THIS_IP_;         \
 137                WRITE_ONCE(current->__state, (state_value));    \
 138        } while (0)
 139
 140#define set_current_state(state_value)                          \
 141        do {                                                    \
 142                WARN_ON_ONCE(is_special_task_state(state_value));\
 143                current->task_state_change = _THIS_IP_;         \
 144                smp_store_mb(current->__state, (state_value));  \
 145        } while (0)
 146
 147#define set_special_state(state_value)                                  \
 148        do {                                                            \
 149                unsigned long flags; /* may shadow */                   \
 150                WARN_ON_ONCE(!is_special_task_state(state_value));      \
 151                raw_spin_lock_irqsave(&current->pi_lock, flags);        \
 152                current->task_state_change = _THIS_IP_;                 \
 153                WRITE_ONCE(current->__state, (state_value));            \
 154                raw_spin_unlock_irqrestore(&current->pi_lock, flags);   \
 155        } while (0)
 156#else
 157/*
 158 * set_current_state() includes a barrier so that the write of current->state
 159 * is correctly serialised wrt the caller's subsequent test of whether to
 160 * actually sleep:
 161 *
 162 *   for (;;) {
 163 *      set_current_state(TASK_UNINTERRUPTIBLE);
 164 *      if (CONDITION)
 165 *         break;
 166 *
 167 *      schedule();
 168 *   }
 169 *   __set_current_state(TASK_RUNNING);
 170 *
 171 * If the caller does not need such serialisation (because, for instance, the
 172 * CONDITION test and condition change and wakeup are under the same lock) then
 173 * use __set_current_state().
 174 *
 175 * The above is typically ordered against the wakeup, which does:
 176 *
 177 *   CONDITION = 1;
 178 *   wake_up_state(p, TASK_UNINTERRUPTIBLE);
 179 *
 180 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
 181 * accessing p->state.
 182 *
 183 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
 184 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
 185 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
 186 *
 187 * However, with slightly different timing the wakeup TASK_RUNNING store can
 188 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
 189 * a problem either because that will result in one extra go around the loop
 190 * and our @cond test will save the day.
 191 *
 192 * Also see the comments of try_to_wake_up().
 193 */
 194#define __set_current_state(state_value)                                \
 195        WRITE_ONCE(current->__state, (state_value))
 196
 197#define set_current_state(state_value)                                  \
 198        smp_store_mb(current->__state, (state_value))
 199
 200/*
 201 * set_special_state() should be used for those states when the blocking task
 202 * can not use the regular condition based wait-loop. In that case we must
 203 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
 204 * will not collide with our state change.
 205 */
 206#define set_special_state(state_value)                                  \
 207        do {                                                            \
 208                unsigned long flags; /* may shadow */                   \
 209                raw_spin_lock_irqsave(&current->pi_lock, flags);        \
 210                WRITE_ONCE(current->__state, (state_value));            \
 211                raw_spin_unlock_irqrestore(&current->pi_lock, flags);   \
 212        } while (0)
 213
 214#endif
 215
 216#define get_current_state()     READ_ONCE(current->__state)
 217
 218/* Task command name length: */
 219#define TASK_COMM_LEN                   16
 220
 221extern void scheduler_tick(void);
 222
 223#define MAX_SCHEDULE_TIMEOUT            LONG_MAX
 224
 225extern long schedule_timeout(long timeout);
 226extern long schedule_timeout_interruptible(long timeout);
 227extern long schedule_timeout_killable(long timeout);
 228extern long schedule_timeout_uninterruptible(long timeout);
 229extern long schedule_timeout_idle(long timeout);
 230asmlinkage void schedule(void);
 231extern void schedule_preempt_disabled(void);
 232asmlinkage void preempt_schedule_irq(void);
 233
 234extern int __must_check io_schedule_prepare(void);
 235extern void io_schedule_finish(int token);
 236extern long io_schedule_timeout(long timeout);
 237extern void io_schedule(void);
 238
 239/**
 240 * struct prev_cputime - snapshot of system and user cputime
 241 * @utime: time spent in user mode
 242 * @stime: time spent in system mode
 243 * @lock: protects the above two fields
 244 *
 245 * Stores previous user/system time values such that we can guarantee
 246 * monotonicity.
 247 */
 248struct prev_cputime {
 249#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
 250        u64                             utime;
 251        u64                             stime;
 252        raw_spinlock_t                  lock;
 253#endif
 254};
 255
 256enum vtime_state {
 257        /* Task is sleeping or running in a CPU with VTIME inactive: */
 258        VTIME_INACTIVE = 0,
 259        /* Task is idle */
 260        VTIME_IDLE,
 261        /* Task runs in kernelspace in a CPU with VTIME active: */
 262        VTIME_SYS,
 263        /* Task runs in userspace in a CPU with VTIME active: */
 264        VTIME_USER,
 265        /* Task runs as guests in a CPU with VTIME active: */
 266        VTIME_GUEST,
 267};
 268
 269struct vtime {
 270        seqcount_t              seqcount;
 271        unsigned long long      starttime;
 272        enum vtime_state        state;
 273        unsigned int            cpu;
 274        u64                     utime;
 275        u64                     stime;
 276        u64                     gtime;
 277};
 278
 279/*
 280 * Utilization clamp constraints.
 281 * @UCLAMP_MIN: Minimum utilization
 282 * @UCLAMP_MAX: Maximum utilization
 283 * @UCLAMP_CNT: Utilization clamp constraints count
 284 */
 285enum uclamp_id {
 286        UCLAMP_MIN = 0,
 287        UCLAMP_MAX,
 288        UCLAMP_CNT
 289};
 290
 291#ifdef CONFIG_SMP
 292extern struct root_domain def_root_domain;
 293extern struct mutex sched_domains_mutex;
 294#endif
 295
 296struct sched_info {
 297#ifdef CONFIG_SCHED_INFO
 298        /* Cumulative counters: */
 299
 300        /* # of times we have run on this CPU: */
 301        unsigned long                   pcount;
 302
 303        /* Time spent waiting on a runqueue: */
 304        unsigned long long              run_delay;
 305
 306        /* Timestamps: */
 307
 308        /* When did we last run on a CPU? */
 309        unsigned long long              last_arrival;
 310
 311        /* When were we last queued to run? */
 312        unsigned long long              last_queued;
 313
 314#endif /* CONFIG_SCHED_INFO */
 315};
 316
 317/*
 318 * Integer metrics need fixed point arithmetic, e.g., sched/fair
 319 * has a few: load, load_avg, util_avg, freq, and capacity.
 320 *
 321 * We define a basic fixed point arithmetic range, and then formalize
 322 * all these metrics based on that basic range.
 323 */
 324# define SCHED_FIXEDPOINT_SHIFT         10
 325# define SCHED_FIXEDPOINT_SCALE         (1L << SCHED_FIXEDPOINT_SHIFT)
 326
 327/* Increase resolution of cpu_capacity calculations */
 328# define SCHED_CAPACITY_SHIFT           SCHED_FIXEDPOINT_SHIFT
 329# define SCHED_CAPACITY_SCALE           (1L << SCHED_CAPACITY_SHIFT)
 330
 331struct load_weight {
 332        unsigned long                   weight;
 333        u32                             inv_weight;
 334};
 335
 336/**
 337 * struct util_est - Estimation utilization of FAIR tasks
 338 * @enqueued: instantaneous estimated utilization of a task/cpu
 339 * @ewma:     the Exponential Weighted Moving Average (EWMA)
 340 *            utilization of a task
 341 *
 342 * Support data structure to track an Exponential Weighted Moving Average
 343 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
 344 * average each time a task completes an activation. Sample's weight is chosen
 345 * so that the EWMA will be relatively insensitive to transient changes to the
 346 * task's workload.
 347 *
 348 * The enqueued attribute has a slightly different meaning for tasks and cpus:
 349 * - task:   the task's util_avg at last task dequeue time
 350 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
 351 * Thus, the util_est.enqueued of a task represents the contribution on the
 352 * estimated utilization of the CPU where that task is currently enqueued.
 353 *
 354 * Only for tasks we track a moving average of the past instantaneous
 355 * estimated utilization. This allows to absorb sporadic drops in utilization
 356 * of an otherwise almost periodic task.
 357 *
 358 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
 359 * updates. When a task is dequeued, its util_est should not be updated if its
 360 * util_avg has not been updated in the meantime.
 361 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
 362 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
 363 * for a task) it is safe to use MSB.
 364 */
 365struct util_est {
 366        unsigned int                    enqueued;
 367        unsigned int                    ewma;
 368#define UTIL_EST_WEIGHT_SHIFT           2
 369#define UTIL_AVG_UNCHANGED              0x80000000
 370} __attribute__((__aligned__(sizeof(u64))));
 371
 372/*
 373 * The load/runnable/util_avg accumulates an infinite geometric series
 374 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
 375 *
 376 * [load_avg definition]
 377 *
 378 *   load_avg = runnable% * scale_load_down(load)
 379 *
 380 * [runnable_avg definition]
 381 *
 382 *   runnable_avg = runnable% * SCHED_CAPACITY_SCALE
 383 *
 384 * [util_avg definition]
 385 *
 386 *   util_avg = running% * SCHED_CAPACITY_SCALE
 387 *
 388 * where runnable% is the time ratio that a sched_entity is runnable and
 389 * running% the time ratio that a sched_entity is running.
 390 *
 391 * For cfs_rq, they are the aggregated values of all runnable and blocked
 392 * sched_entities.
 393 *
 394 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
 395 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
 396 * for computing those signals (see update_rq_clock_pelt())
 397 *
 398 * N.B., the above ratios (runnable% and running%) themselves are in the
 399 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
 400 * to as large a range as necessary. This is for example reflected by
 401 * util_avg's SCHED_CAPACITY_SCALE.
 402 *
 403 * [Overflow issue]
 404 *
 405 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
 406 * with the highest load (=88761), always runnable on a single cfs_rq,
 407 * and should not overflow as the number already hits PID_MAX_LIMIT.
 408 *
 409 * For all other cases (including 32-bit kernels), struct load_weight's
 410 * weight will overflow first before we do, because:
 411 *
 412 *    Max(load_avg) <= Max(load.weight)
 413 *
 414 * Then it is the load_weight's responsibility to consider overflow
 415 * issues.
 416 */
 417struct sched_avg {
 418        u64                             last_update_time;
 419        u64                             load_sum;
 420        u64                             runnable_sum;
 421        u32                             util_sum;
 422        u32                             period_contrib;
 423        unsigned long                   load_avg;
 424        unsigned long                   runnable_avg;
 425        unsigned long                   util_avg;
 426        struct util_est                 util_est;
 427} ____cacheline_aligned;
 428
 429struct sched_statistics {
 430#ifdef CONFIG_SCHEDSTATS
 431        u64                             wait_start;
 432        u64                             wait_max;
 433        u64                             wait_count;
 434        u64                             wait_sum;
 435        u64                             iowait_count;
 436        u64                             iowait_sum;
 437
 438        u64                             sleep_start;
 439        u64                             sleep_max;
 440        s64                             sum_sleep_runtime;
 441
 442        u64                             block_start;
 443        u64                             block_max;
 444        u64                             exec_max;
 445        u64                             slice_max;
 446
 447        u64                             nr_migrations_cold;
 448        u64                             nr_failed_migrations_affine;
 449        u64                             nr_failed_migrations_running;
 450        u64                             nr_failed_migrations_hot;
 451        u64                             nr_forced_migrations;
 452
 453        u64                             nr_wakeups;
 454        u64                             nr_wakeups_sync;
 455        u64                             nr_wakeups_migrate;
 456        u64                             nr_wakeups_local;
 457        u64                             nr_wakeups_remote;
 458        u64                             nr_wakeups_affine;
 459        u64                             nr_wakeups_affine_attempts;
 460        u64                             nr_wakeups_passive;
 461        u64                             nr_wakeups_idle;
 462#endif
 463};
 464
 465struct sched_entity {
 466        /* For load-balancing: */
 467        struct load_weight              load;
 468        struct rb_node                  run_node;
 469        struct list_head                group_node;
 470        unsigned int                    on_rq;
 471
 472        u64                             exec_start;
 473        u64                             sum_exec_runtime;
 474        u64                             vruntime;
 475        u64                             prev_sum_exec_runtime;
 476
 477        u64                             nr_migrations;
 478
 479        struct sched_statistics         statistics;
 480
 481#ifdef CONFIG_FAIR_GROUP_SCHED
 482        int                             depth;
 483        struct sched_entity             *parent;
 484        /* rq on which this entity is (to be) queued: */
 485        struct cfs_rq                   *cfs_rq;
 486        /* rq "owned" by this entity/group: */
 487        struct cfs_rq                   *my_q;
 488        /* cached value of my_q->h_nr_running */
 489        unsigned long                   runnable_weight;
 490#endif
 491
 492#ifdef CONFIG_SMP
 493        /*
 494         * Per entity load average tracking.
 495         *
 496         * Put into separate cache line so it does not
 497         * collide with read-mostly values above.
 498         */
 499        struct sched_avg                avg;
 500#endif
 501};
 502
 503struct sched_rt_entity {
 504        struct list_head                run_list;
 505        unsigned long                   timeout;
 506        unsigned long                   watchdog_stamp;
 507        unsigned int                    time_slice;
 508        unsigned short                  on_rq;
 509        unsigned short                  on_list;
 510
 511        struct sched_rt_entity          *back;
 512#ifdef CONFIG_RT_GROUP_SCHED
 513        struct sched_rt_entity          *parent;
 514        /* rq on which this entity is (to be) queued: */
 515        struct rt_rq                    *rt_rq;
 516        /* rq "owned" by this entity/group: */
 517        struct rt_rq                    *my_q;
 518#endif
 519} __randomize_layout;
 520
 521struct sched_dl_entity {
 522        struct rb_node                  rb_node;
 523
 524        /*
 525         * Original scheduling parameters. Copied here from sched_attr
 526         * during sched_setattr(), they will remain the same until
 527         * the next sched_setattr().
 528         */
 529        u64                             dl_runtime;     /* Maximum runtime for each instance    */
 530        u64                             dl_deadline;    /* Relative deadline of each instance   */
 531        u64                             dl_period;      /* Separation of two instances (period) */
 532        u64                             dl_bw;          /* dl_runtime / dl_period               */
 533        u64                             dl_density;     /* dl_runtime / dl_deadline             */
 534
 535        /*
 536         * Actual scheduling parameters. Initialized with the values above,
 537         * they are continuously updated during task execution. Note that
 538         * the remaining runtime could be < 0 in case we are in overrun.
 539         */
 540        s64                             runtime;        /* Remaining runtime for this instance  */
 541        u64                             deadline;       /* Absolute deadline for this instance  */
 542        unsigned int                    flags;          /* Specifying the scheduler behaviour   */
 543
 544        /*
 545         * Some bool flags:
 546         *
 547         * @dl_throttled tells if we exhausted the runtime. If so, the
 548         * task has to wait for a replenishment to be performed at the
 549         * next firing of dl_timer.
 550         *
 551         * @dl_boosted tells if we are boosted due to DI. If so we are
 552         * outside bandwidth enforcement mechanism (but only until we
 553         * exit the critical section);
 554         *
 555         * @dl_yielded tells if task gave up the CPU before consuming
 556         * all its available runtime during the last job.
 557         *
 558         * @dl_non_contending tells if the task is inactive while still
 559         * contributing to the active utilization. In other words, it
 560         * indicates if the inactive timer has been armed and its handler
 561         * has not been executed yet. This flag is useful to avoid race
 562         * conditions between the inactive timer handler and the wakeup
 563         * code.
 564         *
 565         * @dl_overrun tells if the task asked to be informed about runtime
 566         * overruns.
 567         */
 568        unsigned int                    dl_throttled      : 1;
 569        unsigned int                    dl_yielded        : 1;
 570        unsigned int                    dl_non_contending : 1;
 571        unsigned int                    dl_overrun        : 1;
 572
 573        /*
 574         * Bandwidth enforcement timer. Each -deadline task has its
 575         * own bandwidth to be enforced, thus we need one timer per task.
 576         */
 577        struct hrtimer                  dl_timer;
 578
 579        /*
 580         * Inactive timer, responsible for decreasing the active utilization
 581         * at the "0-lag time". When a -deadline task blocks, it contributes
 582         * to GRUB's active utilization until the "0-lag time", hence a
 583         * timer is needed to decrease the active utilization at the correct
 584         * time.
 585         */
 586        struct hrtimer inactive_timer;
 587
 588#ifdef CONFIG_RT_MUTEXES
 589        /*
 590         * Priority Inheritance. When a DEADLINE scheduling entity is boosted
 591         * pi_se points to the donor, otherwise points to the dl_se it belongs
 592         * to (the original one/itself).
 593         */
 594        struct sched_dl_entity *pi_se;
 595#endif
 596};
 597
 598#ifdef CONFIG_UCLAMP_TASK
 599/* Number of utilization clamp buckets (shorter alias) */
 600#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
 601
 602/*
 603 * Utilization clamp for a scheduling entity
 604 * @value:              clamp value "assigned" to a se
 605 * @bucket_id:          bucket index corresponding to the "assigned" value
 606 * @active:             the se is currently refcounted in a rq's bucket
 607 * @user_defined:       the requested clamp value comes from user-space
 608 *
 609 * The bucket_id is the index of the clamp bucket matching the clamp value
 610 * which is pre-computed and stored to avoid expensive integer divisions from
 611 * the fast path.
 612 *
 613 * The active bit is set whenever a task has got an "effective" value assigned,
 614 * which can be different from the clamp value "requested" from user-space.
 615 * This allows to know a task is refcounted in the rq's bucket corresponding
 616 * to the "effective" bucket_id.
 617 *
 618 * The user_defined bit is set whenever a task has got a task-specific clamp
 619 * value requested from userspace, i.e. the system defaults apply to this task
 620 * just as a restriction. This allows to relax default clamps when a less
 621 * restrictive task-specific value has been requested, thus allowing to
 622 * implement a "nice" semantic. For example, a task running with a 20%
 623 * default boost can still drop its own boosting to 0%.
 624 */
 625struct uclamp_se {
 626        unsigned int value              : bits_per(SCHED_CAPACITY_SCALE);
 627        unsigned int bucket_id          : bits_per(UCLAMP_BUCKETS);
 628        unsigned int active             : 1;
 629        unsigned int user_defined       : 1;
 630};
 631#endif /* CONFIG_UCLAMP_TASK */
 632
 633union rcu_special {
 634        struct {
 635                u8                      blocked;
 636                u8                      need_qs;
 637                u8                      exp_hint; /* Hint for performance. */
 638                u8                      need_mb; /* Readers need smp_mb(). */
 639        } b; /* Bits. */
 640        u32 s; /* Set of bits. */
 641};
 642
 643enum perf_event_task_context {
 644        perf_invalid_context = -1,
 645        perf_hw_context = 0,
 646        perf_sw_context,
 647        perf_nr_task_contexts,
 648};
 649
 650struct wake_q_node {
 651        struct wake_q_node *next;
 652};
 653
 654struct kmap_ctrl {
 655#ifdef CONFIG_KMAP_LOCAL
 656        int                             idx;
 657        pte_t                           pteval[KM_MAX_IDX];
 658#endif
 659};
 660
 661struct task_struct {
 662#ifdef CONFIG_THREAD_INFO_IN_TASK
 663        /*
 664         * For reasons of header soup (see current_thread_info()), this
 665         * must be the first element of task_struct.
 666         */
 667        struct thread_info              thread_info;
 668#endif
 669        unsigned int                    __state;
 670
 671        /*
 672         * This begins the randomizable portion of task_struct. Only
 673         * scheduling-critical items should be added above here.
 674         */
 675        randomized_struct_fields_start
 676
 677        void                            *stack;
 678        refcount_t                      usage;
 679        /* Per task flags (PF_*), defined further below: */
 680        unsigned int                    flags;
 681        unsigned int                    ptrace;
 682
 683#ifdef CONFIG_SMP
 684        int                             on_cpu;
 685        struct __call_single_node       wake_entry;
 686#ifdef CONFIG_THREAD_INFO_IN_TASK
 687        /* Current CPU: */
 688        unsigned int                    cpu;
 689#endif
 690        unsigned int                    wakee_flips;
 691        unsigned long                   wakee_flip_decay_ts;
 692        struct task_struct              *last_wakee;
 693
 694        /*
 695         * recent_used_cpu is initially set as the last CPU used by a task
 696         * that wakes affine another task. Waker/wakee relationships can
 697         * push tasks around a CPU where each wakeup moves to the next one.
 698         * Tracking a recently used CPU allows a quick search for a recently
 699         * used CPU that may be idle.
 700         */
 701        int                             recent_used_cpu;
 702        int                             wake_cpu;
 703#endif
 704        int                             on_rq;
 705
 706        int                             prio;
 707        int                             static_prio;
 708        int                             normal_prio;
 709        unsigned int                    rt_priority;
 710
 711        const struct sched_class        *sched_class;
 712        struct sched_entity             se;
 713        struct sched_rt_entity          rt;
 714        struct sched_dl_entity          dl;
 715
 716#ifdef CONFIG_SCHED_CORE
 717        struct rb_node                  core_node;
 718        unsigned long                   core_cookie;
 719        unsigned int                    core_occupation;
 720#endif
 721
 722#ifdef CONFIG_CGROUP_SCHED
 723        struct task_group               *sched_task_group;
 724#endif
 725
 726#ifdef CONFIG_UCLAMP_TASK
 727        /*
 728         * Clamp values requested for a scheduling entity.
 729         * Must be updated with task_rq_lock() held.
 730         */
 731        struct uclamp_se                uclamp_req[UCLAMP_CNT];
 732        /*
 733         * Effective clamp values used for a scheduling entity.
 734         * Must be updated with task_rq_lock() held.
 735         */
 736        struct uclamp_se                uclamp[UCLAMP_CNT];
 737#endif
 738
 739#ifdef CONFIG_PREEMPT_NOTIFIERS
 740        /* List of struct preempt_notifier: */
 741        struct hlist_head               preempt_notifiers;
 742#endif
 743
 744#ifdef CONFIG_BLK_DEV_IO_TRACE
 745        unsigned int                    btrace_seq;
 746#endif
 747
 748        unsigned int                    policy;
 749        int                             nr_cpus_allowed;
 750        const cpumask_t                 *cpus_ptr;
 751        cpumask_t                       cpus_mask;
 752        void                            *migration_pending;
 753#ifdef CONFIG_SMP
 754        unsigned short                  migration_disabled;
 755#endif
 756        unsigned short                  migration_flags;
 757
 758#ifdef CONFIG_PREEMPT_RCU
 759        int                             rcu_read_lock_nesting;
 760        union rcu_special               rcu_read_unlock_special;
 761        struct list_head                rcu_node_entry;
 762        struct rcu_node                 *rcu_blocked_node;
 763#endif /* #ifdef CONFIG_PREEMPT_RCU */
 764
 765#ifdef CONFIG_TASKS_RCU
 766        unsigned long                   rcu_tasks_nvcsw;
 767        u8                              rcu_tasks_holdout;
 768        u8                              rcu_tasks_idx;
 769        int                             rcu_tasks_idle_cpu;
 770        struct list_head                rcu_tasks_holdout_list;
 771#endif /* #ifdef CONFIG_TASKS_RCU */
 772
 773#ifdef CONFIG_TASKS_TRACE_RCU
 774        int                             trc_reader_nesting;
 775        int                             trc_ipi_to_cpu;
 776        union rcu_special               trc_reader_special;
 777        bool                            trc_reader_checked;
 778        struct list_head                trc_holdout_list;
 779#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
 780
 781        struct sched_info               sched_info;
 782
 783        struct list_head                tasks;
 784#ifdef CONFIG_SMP
 785        struct plist_node               pushable_tasks;
 786        struct rb_node                  pushable_dl_tasks;
 787#endif
 788
 789        struct mm_struct                *mm;
 790        struct mm_struct                *active_mm;
 791
 792        /* Per-thread vma caching: */
 793        struct vmacache                 vmacache;
 794
 795#ifdef SPLIT_RSS_COUNTING
 796        struct task_rss_stat            rss_stat;
 797#endif
 798        int                             exit_state;
 799        int                             exit_code;
 800        int                             exit_signal;
 801        /* The signal sent when the parent dies: */
 802        int                             pdeath_signal;
 803        /* JOBCTL_*, siglock protected: */
 804        unsigned long                   jobctl;
 805
 806        /* Used for emulating ABI behavior of previous Linux versions: */
 807        unsigned int                    personality;
 808
 809        /* Scheduler bits, serialized by scheduler locks: */
 810        unsigned                        sched_reset_on_fork:1;
 811        unsigned                        sched_contributes_to_load:1;
 812        unsigned                        sched_migrated:1;
 813#ifdef CONFIG_PSI
 814        unsigned                        sched_psi_wake_requeue:1;
 815#endif
 816
 817        /* Force alignment to the next boundary: */
 818        unsigned                        :0;
 819
 820        /* Unserialized, strictly 'current' */
 821
 822        /*
 823         * This field must not be in the scheduler word above due to wakelist
 824         * queueing no longer being serialized by p->on_cpu. However:
 825         *
 826         * p->XXX = X;                  ttwu()
 827         * schedule()                     if (p->on_rq && ..) // false
 828         *   smp_mb__after_spinlock();    if (smp_load_acquire(&p->on_cpu) && //true
 829         *   deactivate_task()                ttwu_queue_wakelist())
 830         *     p->on_rq = 0;                    p->sched_remote_wakeup = Y;
 831         *
 832         * guarantees all stores of 'current' are visible before
 833         * ->sched_remote_wakeup gets used, so it can be in this word.
 834         */
 835        unsigned                        sched_remote_wakeup:1;
 836
 837        /* Bit to tell LSMs we're in execve(): */
 838        unsigned                        in_execve:1;
 839        unsigned                        in_iowait:1;
 840#ifndef TIF_RESTORE_SIGMASK
 841        unsigned                        restore_sigmask:1;
 842#endif
 843#ifdef CONFIG_MEMCG
 844        unsigned                        in_user_fault:1;
 845#endif
 846#ifdef CONFIG_COMPAT_BRK
 847        unsigned                        brk_randomized:1;
 848#endif
 849#ifdef CONFIG_CGROUPS
 850        /* disallow userland-initiated cgroup migration */
 851        unsigned                        no_cgroup_migration:1;
 852        /* task is frozen/stopped (used by the cgroup freezer) */
 853        unsigned                        frozen:1;
 854#endif
 855#ifdef CONFIG_BLK_CGROUP
 856        unsigned                        use_memdelay:1;
 857#endif
 858#ifdef CONFIG_PSI
 859        /* Stalled due to lack of memory */
 860        unsigned                        in_memstall:1;
 861#endif
 862#ifdef CONFIG_PAGE_OWNER
 863        /* Used by page_owner=on to detect recursion in page tracking. */
 864        unsigned                        in_page_owner:1;
 865#endif
 866
 867        unsigned long                   atomic_flags; /* Flags requiring atomic access. */
 868
 869        struct restart_block            restart_block;
 870
 871        pid_t                           pid;
 872        pid_t                           tgid;
 873
 874#ifdef CONFIG_STACKPROTECTOR
 875        /* Canary value for the -fstack-protector GCC feature: */
 876        unsigned long                   stack_canary;
 877#endif
 878        /*
 879         * Pointers to the (original) parent process, youngest child, younger sibling,
 880         * older sibling, respectively.  (p->father can be replaced with
 881         * p->real_parent->pid)
 882         */
 883
 884        /* Real parent process: */
 885        struct task_struct __rcu        *real_parent;
 886
 887        /* Recipient of SIGCHLD, wait4() reports: */
 888        struct task_struct __rcu        *parent;
 889
 890        /*
 891         * Children/sibling form the list of natural children:
 892         */
 893        struct list_head                children;
 894        struct list_head                sibling;
 895        struct task_struct              *group_leader;
 896
 897        /*
 898         * 'ptraced' is the list of tasks this task is using ptrace() on.
 899         *
 900         * This includes both natural children and PTRACE_ATTACH targets.
 901         * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
 902         */
 903        struct list_head                ptraced;
 904        struct list_head                ptrace_entry;
 905
 906        /* PID/PID hash table linkage. */
 907        struct pid                      *thread_pid;
 908        struct hlist_node               pid_links[PIDTYPE_MAX];
 909        struct list_head                thread_group;
 910        struct list_head                thread_node;
 911
 912        struct completion               *vfork_done;
 913
 914        /* CLONE_CHILD_SETTID: */
 915        int __user                      *set_child_tid;
 916
 917        /* CLONE_CHILD_CLEARTID: */
 918        int __user                      *clear_child_tid;
 919
 920        /* PF_IO_WORKER */
 921        void                            *pf_io_worker;
 922
 923        u64                             utime;
 924        u64                             stime;
 925#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
 926        u64                             utimescaled;
 927        u64                             stimescaled;
 928#endif
 929        u64                             gtime;
 930        struct prev_cputime             prev_cputime;
 931#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
 932        struct vtime                    vtime;
 933#endif
 934
 935#ifdef CONFIG_NO_HZ_FULL
 936        atomic_t                        tick_dep_mask;
 937#endif
 938        /* Context switch counts: */
 939        unsigned long                   nvcsw;
 940        unsigned long                   nivcsw;
 941
 942        /* Monotonic time in nsecs: */
 943        u64                             start_time;
 944
 945        /* Boot based time in nsecs: */
 946        u64                             start_boottime;
 947
 948        /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
 949        unsigned long                   min_flt;
 950        unsigned long                   maj_flt;
 951
 952        /* Empty if CONFIG_POSIX_CPUTIMERS=n */
 953        struct posix_cputimers          posix_cputimers;
 954
 955#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
 956        struct posix_cputimers_work     posix_cputimers_work;
 957#endif
 958
 959        /* Process credentials: */
 960
 961        /* Tracer's credentials at attach: */
 962        const struct cred __rcu         *ptracer_cred;
 963
 964        /* Objective and real subjective task credentials (COW): */
 965        const struct cred __rcu         *real_cred;
 966
 967        /* Effective (overridable) subjective task credentials (COW): */
 968        const struct cred __rcu         *cred;
 969
 970#ifdef CONFIG_KEYS
 971        /* Cached requested key. */
 972        struct key                      *cached_requested_key;
 973#endif
 974
 975        /*
 976         * executable name, excluding path.
 977         *
 978         * - normally initialized setup_new_exec()
 979         * - access it with [gs]et_task_comm()
 980         * - lock it with task_lock()
 981         */
 982        char                            comm[TASK_COMM_LEN];
 983
 984        struct nameidata                *nameidata;
 985
 986#ifdef CONFIG_SYSVIPC
 987        struct sysv_sem                 sysvsem;
 988        struct sysv_shm                 sysvshm;
 989#endif
 990#ifdef CONFIG_DETECT_HUNG_TASK
 991        unsigned long                   last_switch_count;
 992        unsigned long                   last_switch_time;
 993#endif
 994        /* Filesystem information: */
 995        struct fs_struct                *fs;
 996
 997        /* Open file information: */
 998        struct files_struct             *files;
 999
1000#ifdef CONFIG_IO_URING
1001        struct io_uring_task            *io_uring;
1002#endif
1003
1004        /* Namespaces: */
1005        struct nsproxy                  *nsproxy;
1006
1007        /* Signal handlers: */
1008        struct signal_struct            *signal;
1009        struct sighand_struct __rcu             *sighand;
1010        sigset_t                        blocked;
1011        sigset_t                        real_blocked;
1012        /* Restored if set_restore_sigmask() was used: */
1013        sigset_t                        saved_sigmask;
1014        struct sigpending               pending;
1015        unsigned long                   sas_ss_sp;
1016        size_t                          sas_ss_size;
1017        unsigned int                    sas_ss_flags;
1018
1019        struct callback_head            *task_works;
1020
1021#ifdef CONFIG_AUDIT
1022#ifdef CONFIG_AUDITSYSCALL
1023        struct audit_context            *audit_context;
1024#endif
1025        kuid_t                          loginuid;
1026        unsigned int                    sessionid;
1027#endif
1028        struct seccomp                  seccomp;
1029        struct syscall_user_dispatch    syscall_dispatch;
1030
1031        /* Thread group tracking: */
1032        u64                             parent_exec_id;
1033        u64                             self_exec_id;
1034
1035        /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1036        spinlock_t                      alloc_lock;
1037
1038        /* Protection of the PI data structures: */
1039        raw_spinlock_t                  pi_lock;
1040
1041        struct wake_q_node              wake_q;
1042
1043#ifdef CONFIG_RT_MUTEXES
1044        /* PI waiters blocked on a rt_mutex held by this task: */
1045        struct rb_root_cached           pi_waiters;
1046        /* Updated under owner's pi_lock and rq lock */
1047        struct task_struct              *pi_top_task;
1048        /* Deadlock detection and priority inheritance handling: */
1049        struct rt_mutex_waiter          *pi_blocked_on;
1050#endif
1051
1052#ifdef CONFIG_DEBUG_MUTEXES
1053        /* Mutex deadlock detection: */
1054        struct mutex_waiter             *blocked_on;
1055#endif
1056
1057#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1058        int                             non_block_count;
1059#endif
1060
1061#ifdef CONFIG_TRACE_IRQFLAGS
1062        struct irqtrace_events          irqtrace;
1063        unsigned int                    hardirq_threaded;
1064        u64                             hardirq_chain_key;
1065        int                             softirqs_enabled;
1066        int                             softirq_context;
1067        int                             irq_config;
1068#endif
1069#ifdef CONFIG_PREEMPT_RT
1070        int                             softirq_disable_cnt;
1071#endif
1072
1073#ifdef CONFIG_LOCKDEP
1074# define MAX_LOCK_DEPTH                 48UL
1075        u64                             curr_chain_key;
1076        int                             lockdep_depth;
1077        unsigned int                    lockdep_recursion;
1078        struct held_lock                held_locks[MAX_LOCK_DEPTH];
1079#endif
1080
1081#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1082        unsigned int                    in_ubsan;
1083#endif
1084
1085        /* Journalling filesystem info: */
1086        void                            *journal_info;
1087
1088        /* Stacked block device info: */
1089        struct bio_list                 *bio_list;
1090
1091#ifdef CONFIG_BLOCK
1092        /* Stack plugging: */
1093        struct blk_plug                 *plug;
1094#endif
1095
1096        /* VM state: */
1097        struct reclaim_state            *reclaim_state;
1098
1099        struct backing_dev_info         *backing_dev_info;
1100
1101        struct io_context               *io_context;
1102
1103#ifdef CONFIG_COMPACTION
1104        struct capture_control          *capture_control;
1105#endif
1106        /* Ptrace state: */
1107        unsigned long                   ptrace_message;
1108        kernel_siginfo_t                *last_siginfo;
1109
1110        struct task_io_accounting       ioac;
1111#ifdef CONFIG_PSI
1112        /* Pressure stall state */
1113        unsigned int                    psi_flags;
1114#endif
1115#ifdef CONFIG_TASK_XACCT
1116        /* Accumulated RSS usage: */
1117        u64                             acct_rss_mem1;
1118        /* Accumulated virtual memory usage: */
1119        u64                             acct_vm_mem1;
1120        /* stime + utime since last update: */
1121        u64                             acct_timexpd;
1122#endif
1123#ifdef CONFIG_CPUSETS
1124        /* Protected by ->alloc_lock: */
1125        nodemask_t                      mems_allowed;
1126        /* Sequence number to catch updates: */
1127        seqcount_spinlock_t             mems_allowed_seq;
1128        int                             cpuset_mem_spread_rotor;
1129        int                             cpuset_slab_spread_rotor;
1130#endif
1131#ifdef CONFIG_CGROUPS
1132        /* Control Group info protected by css_set_lock: */
1133        struct css_set __rcu            *cgroups;
1134        /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1135        struct list_head                cg_list;
1136#endif
1137#ifdef CONFIG_X86_CPU_RESCTRL
1138        u32                             closid;
1139        u32                             rmid;
1140#endif
1141#ifdef CONFIG_FUTEX
1142        struct robust_list_head __user  *robust_list;
1143#ifdef CONFIG_COMPAT
1144        struct compat_robust_list_head __user *compat_robust_list;
1145#endif
1146        struct list_head                pi_state_list;
1147        struct futex_pi_state           *pi_state_cache;
1148        struct mutex                    futex_exit_mutex;
1149        unsigned int                    futex_state;
1150#endif
1151#ifdef CONFIG_PERF_EVENTS
1152        struct perf_event_context       *perf_event_ctxp[perf_nr_task_contexts];
1153        struct mutex                    perf_event_mutex;
1154        struct list_head                perf_event_list;
1155#endif
1156#ifdef CONFIG_DEBUG_PREEMPT
1157        unsigned long                   preempt_disable_ip;
1158#endif
1159#ifdef CONFIG_NUMA
1160        /* Protected by alloc_lock: */
1161        struct mempolicy                *mempolicy;
1162        short                           il_prev;
1163        short                           pref_node_fork;
1164#endif
1165#ifdef CONFIG_NUMA_BALANCING
1166        int                             numa_scan_seq;
1167        unsigned int                    numa_scan_period;
1168        unsigned int                    numa_scan_period_max;
1169        int                             numa_preferred_nid;
1170        unsigned long                   numa_migrate_retry;
1171        /* Migration stamp: */
1172        u64                             node_stamp;
1173        u64                             last_task_numa_placement;
1174        u64                             last_sum_exec_runtime;
1175        struct callback_head            numa_work;
1176
1177        /*
1178         * This pointer is only modified for current in syscall and
1179         * pagefault context (and for tasks being destroyed), so it can be read
1180         * from any of the following contexts:
1181         *  - RCU read-side critical section
1182         *  - current->numa_group from everywhere
1183         *  - task's runqueue locked, task not running
1184         */
1185        struct numa_group __rcu         *numa_group;
1186
1187        /*
1188         * numa_faults is an array split into four regions:
1189         * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1190         * in this precise order.
1191         *
1192         * faults_memory: Exponential decaying average of faults on a per-node
1193         * basis. Scheduling placement decisions are made based on these
1194         * counts. The values remain static for the duration of a PTE scan.
1195         * faults_cpu: Track the nodes the process was running on when a NUMA
1196         * hinting fault was incurred.
1197         * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1198         * during the current scan window. When the scan completes, the counts
1199         * in faults_memory and faults_cpu decay and these values are copied.
1200         */
1201        unsigned long                   *numa_faults;
1202        unsigned long                   total_numa_faults;
1203
1204        /*
1205         * numa_faults_locality tracks if faults recorded during the last
1206         * scan window were remote/local or failed to migrate. The task scan
1207         * period is adapted based on the locality of the faults with different
1208         * weights depending on whether they were shared or private faults
1209         */
1210        unsigned long                   numa_faults_locality[3];
1211
1212        unsigned long                   numa_pages_migrated;
1213#endif /* CONFIG_NUMA_BALANCING */
1214
1215#ifdef CONFIG_RSEQ
1216        struct rseq __user *rseq;
1217        u32 rseq_sig;
1218        /*
1219         * RmW on rseq_event_mask must be performed atomically
1220         * with respect to preemption.
1221         */
1222        unsigned long rseq_event_mask;
1223#endif
1224
1225        struct tlbflush_unmap_batch     tlb_ubc;
1226
1227        union {
1228                refcount_t              rcu_users;
1229                struct rcu_head         rcu;
1230        };
1231
1232        /* Cache last used pipe for splice(): */
1233        struct pipe_inode_info          *splice_pipe;
1234
1235        struct page_frag                task_frag;
1236
1237#ifdef CONFIG_TASK_DELAY_ACCT
1238        struct task_delay_info          *delays;
1239#endif
1240
1241#ifdef CONFIG_FAULT_INJECTION
1242        int                             make_it_fail;
1243        unsigned int                    fail_nth;
1244#endif
1245        /*
1246         * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1247         * balance_dirty_pages() for a dirty throttling pause:
1248         */
1249        int                             nr_dirtied;
1250        int                             nr_dirtied_pause;
1251        /* Start of a write-and-pause period: */
1252        unsigned long                   dirty_paused_when;
1253
1254#ifdef CONFIG_LATENCYTOP
1255        int                             latency_record_count;
1256        struct latency_record           latency_record[LT_SAVECOUNT];
1257#endif
1258        /*
1259         * Time slack values; these are used to round up poll() and
1260         * select() etc timeout values. These are in nanoseconds.
1261         */
1262        u64                             timer_slack_ns;
1263        u64                             default_timer_slack_ns;
1264
1265#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1266        unsigned int                    kasan_depth;
1267#endif
1268
1269#ifdef CONFIG_KCSAN
1270        struct kcsan_ctx                kcsan_ctx;
1271#ifdef CONFIG_TRACE_IRQFLAGS
1272        struct irqtrace_events          kcsan_save_irqtrace;
1273#endif
1274#endif
1275
1276#if IS_ENABLED(CONFIG_KUNIT)
1277        struct kunit                    *kunit_test;
1278#endif
1279
1280#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1281        /* Index of current stored address in ret_stack: */
1282        int                             curr_ret_stack;
1283        int                             curr_ret_depth;
1284
1285        /* Stack of return addresses for return function tracing: */
1286        struct ftrace_ret_stack         *ret_stack;
1287
1288        /* Timestamp for last schedule: */
1289        unsigned long long              ftrace_timestamp;
1290
1291        /*
1292         * Number of functions that haven't been traced
1293         * because of depth overrun:
1294         */
1295        atomic_t                        trace_overrun;
1296
1297        /* Pause tracing: */
1298        atomic_t                        tracing_graph_pause;
1299#endif
1300
1301#ifdef CONFIG_TRACING
1302        /* State flags for use by tracers: */
1303        unsigned long                   trace;
1304
1305        /* Bitmask and counter of trace recursion: */
1306        unsigned long                   trace_recursion;
1307#endif /* CONFIG_TRACING */
1308
1309#ifdef CONFIG_KCOV
1310        /* See kernel/kcov.c for more details. */
1311
1312        /* Coverage collection mode enabled for this task (0 if disabled): */
1313        unsigned int                    kcov_mode;
1314
1315        /* Size of the kcov_area: */
1316        unsigned int                    kcov_size;
1317
1318        /* Buffer for coverage collection: */
1319        void                            *kcov_area;
1320
1321        /* KCOV descriptor wired with this task or NULL: */
1322        struct kcov                     *kcov;
1323
1324        /* KCOV common handle for remote coverage collection: */
1325        u64                             kcov_handle;
1326
1327        /* KCOV sequence number: */
1328        int                             kcov_sequence;
1329
1330        /* Collect coverage from softirq context: */
1331        unsigned int                    kcov_softirq;
1332#endif
1333
1334#ifdef CONFIG_MEMCG
1335        struct mem_cgroup               *memcg_in_oom;
1336        gfp_t                           memcg_oom_gfp_mask;
1337        int                             memcg_oom_order;
1338
1339        /* Number of pages to reclaim on returning to userland: */
1340        unsigned int                    memcg_nr_pages_over_high;
1341
1342        /* Used by memcontrol for targeted memcg charge: */
1343        struct mem_cgroup               *active_memcg;
1344#endif
1345
1346#ifdef CONFIG_BLK_CGROUP
1347        struct request_queue            *throttle_queue;
1348#endif
1349
1350#ifdef CONFIG_UPROBES
1351        struct uprobe_task              *utask;
1352#endif
1353#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1354        unsigned int                    sequential_io;
1355        unsigned int                    sequential_io_avg;
1356#endif
1357        struct kmap_ctrl                kmap_ctrl;
1358#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1359        unsigned long                   task_state_change;
1360#endif
1361        int                             pagefault_disabled;
1362#ifdef CONFIG_MMU
1363        struct task_struct              *oom_reaper_list;
1364#endif
1365#ifdef CONFIG_VMAP_STACK
1366        struct vm_struct                *stack_vm_area;
1367#endif
1368#ifdef CONFIG_THREAD_INFO_IN_TASK
1369        /* A live task holds one reference: */
1370        refcount_t                      stack_refcount;
1371#endif
1372#ifdef CONFIG_LIVEPATCH
1373        int patch_state;
1374#endif
1375#ifdef CONFIG_SECURITY
1376        /* Used by LSM modules for access restriction: */
1377        void                            *security;
1378#endif
1379#ifdef CONFIG_BPF_SYSCALL
1380        /* Used by BPF task local storage */
1381        struct bpf_local_storage __rcu  *bpf_storage;
1382#endif
1383
1384#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1385        unsigned long                   lowest_stack;
1386        unsigned long                   prev_lowest_stack;
1387#endif
1388
1389#ifdef CONFIG_X86_MCE
1390        void __user                     *mce_vaddr;
1391        __u64                           mce_kflags;
1392        u64                             mce_addr;
1393        __u64                           mce_ripv : 1,
1394                                        mce_whole_page : 1,
1395                                        __mce_reserved : 62;
1396        struct callback_head            mce_kill_me;
1397#endif
1398
1399#ifdef CONFIG_KRETPROBES
1400        struct llist_head               kretprobe_instances;
1401#endif
1402
1403        /*
1404         * New fields for task_struct should be added above here, so that
1405         * they are included in the randomized portion of task_struct.
1406         */
1407        randomized_struct_fields_end
1408
1409        /* CPU-specific state of this task: */
1410        struct thread_struct            thread;
1411
1412        /*
1413         * WARNING: on x86, 'thread_struct' contains a variable-sized
1414         * structure.  It *MUST* be at the end of 'task_struct'.
1415         *
1416         * Do not put anything below here!
1417         */
1418};
1419
1420static inline struct pid *task_pid(struct task_struct *task)
1421{
1422        return task->thread_pid;
1423}
1424
1425/*
1426 * the helpers to get the task's different pids as they are seen
1427 * from various namespaces
1428 *
1429 * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
1430 * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
1431 *                     current.
1432 * task_xid_nr_ns()  : id seen from the ns specified;
1433 *
1434 * see also pid_nr() etc in include/linux/pid.h
1435 */
1436pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1437
1438static inline pid_t task_pid_nr(struct task_struct *tsk)
1439{
1440        return tsk->pid;
1441}
1442
1443static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1444{
1445        return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1446}
1447
1448static inline pid_t task_pid_vnr(struct task_struct *tsk)
1449{
1450        return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1451}
1452
1453
1454static inline pid_t task_tgid_nr(struct task_struct *tsk)
1455{
1456        return tsk->tgid;
1457}
1458
1459/**
1460 * pid_alive - check that a task structure is not stale
1461 * @p: Task structure to be checked.
1462 *
1463 * Test if a process is not yet dead (at most zombie state)
1464 * If pid_alive fails, then pointers within the task structure
1465 * can be stale and must not be dereferenced.
1466 *
1467 * Return: 1 if the process is alive. 0 otherwise.
1468 */
1469static inline int pid_alive(const struct task_struct *p)
1470{
1471        return p->thread_pid != NULL;
1472}
1473
1474static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1475{
1476        return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1477}
1478
1479static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1480{
1481        return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1482}
1483
1484
1485static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1486{
1487        return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1488}
1489
1490static inline pid_t task_session_vnr(struct task_struct *tsk)
1491{
1492        return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1493}
1494
1495static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1496{
1497        return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1498}
1499
1500static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1501{
1502        return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1503}
1504
1505static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1506{
1507        pid_t pid = 0;
1508
1509        rcu_read_lock();
1510        if (pid_alive(tsk))
1511                pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1512        rcu_read_unlock();
1513
1514        return pid;
1515}
1516
1517static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1518{
1519        return task_ppid_nr_ns(tsk, &init_pid_ns);
1520}
1521
1522/* Obsolete, do not use: */
1523static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1524{
1525        return task_pgrp_nr_ns(tsk, &init_pid_ns);
1526}
1527
1528#define TASK_REPORT_IDLE        (TASK_REPORT + 1)
1529#define TASK_REPORT_MAX         (TASK_REPORT_IDLE << 1)
1530
1531static inline unsigned int task_state_index(struct task_struct *tsk)
1532{
1533        unsigned int tsk_state = READ_ONCE(tsk->__state);
1534        unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1535
1536        BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1537
1538        if (tsk_state == TASK_IDLE)
1539                state = TASK_REPORT_IDLE;
1540
1541        return fls(state);
1542}
1543
1544static inline char task_index_to_char(unsigned int state)
1545{
1546        static const char state_char[] = "RSDTtXZPI";
1547
1548        BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1549
1550        return state_char[state];
1551}
1552
1553static inline char task_state_to_char(struct task_struct *tsk)
1554{
1555        return task_index_to_char(task_state_index(tsk));
1556}
1557
1558/**
1559 * is_global_init - check if a task structure is init. Since init
1560 * is free to have sub-threads we need to check tgid.
1561 * @tsk: Task structure to be checked.
1562 *
1563 * Check if a task structure is the first user space task the kernel created.
1564 *
1565 * Return: 1 if the task structure is init. 0 otherwise.
1566 */
1567static inline int is_global_init(struct task_struct *tsk)
1568{
1569        return task_tgid_nr(tsk) == 1;
1570}
1571
1572extern struct pid *cad_pid;
1573
1574/*
1575 * Per process flags
1576 */
1577#define PF_VCPU                 0x00000001      /* I'm a virtual CPU */
1578#define PF_IDLE                 0x00000002      /* I am an IDLE thread */
1579#define PF_EXITING              0x00000004      /* Getting shut down */
1580#define PF_IO_WORKER            0x00000010      /* Task is an IO worker */
1581#define PF_WQ_WORKER            0x00000020      /* I'm a workqueue worker */
1582#define PF_FORKNOEXEC           0x00000040      /* Forked but didn't exec */
1583#define PF_MCE_PROCESS          0x00000080      /* Process policy on mce errors */
1584#define PF_SUPERPRIV            0x00000100      /* Used super-user privileges */
1585#define PF_DUMPCORE             0x00000200      /* Dumped core */
1586#define PF_SIGNALED             0x00000400      /* Killed by a signal */
1587#define PF_MEMALLOC             0x00000800      /* Allocating memory */
1588#define PF_NPROC_EXCEEDED       0x00001000      /* set_user() noticed that RLIMIT_NPROC was exceeded */
1589#define PF_USED_MATH            0x00002000      /* If unset the fpu must be initialized before use */
1590#define PF_USED_ASYNC           0x00004000      /* Used async_schedule*(), used by module init */
1591#define PF_NOFREEZE             0x00008000      /* This thread should not be frozen */
1592#define PF_FROZEN               0x00010000      /* Frozen for system suspend */
1593#define PF_KSWAPD               0x00020000      /* I am kswapd */
1594#define PF_MEMALLOC_NOFS        0x00040000      /* All allocation requests will inherit GFP_NOFS */
1595#define PF_MEMALLOC_NOIO        0x00080000      /* All allocation requests will inherit GFP_NOIO */
1596#define PF_LOCAL_THROTTLE       0x00100000      /* Throttle writes only against the bdi I write to,
1597                                                 * I am cleaning dirty pages from some other bdi. */
1598#define PF_KTHREAD              0x00200000      /* I am a kernel thread */
1599#define PF_RANDOMIZE            0x00400000      /* Randomize virtual address space */
1600#define PF_SWAPWRITE            0x00800000      /* Allowed to write to swap */
1601#define PF_NO_SETAFFINITY       0x04000000      /* Userland is not allowed to meddle with cpus_mask */
1602#define PF_MCE_EARLY            0x08000000      /* Early kill for mce process policy */
1603#define PF_MEMALLOC_PIN         0x10000000      /* Allocation context constrained to zones which allow long term pinning. */
1604#define PF_FREEZER_SKIP         0x40000000      /* Freezer should not count it as freezable */
1605#define PF_SUSPEND_TASK         0x80000000      /* This thread called freeze_processes() and should not be frozen */
1606
1607/*
1608 * Only the _current_ task can read/write to tsk->flags, but other
1609 * tasks can access tsk->flags in readonly mode for example
1610 * with tsk_used_math (like during threaded core dumping).
1611 * There is however an exception to this rule during ptrace
1612 * or during fork: the ptracer task is allowed to write to the
1613 * child->flags of its traced child (same goes for fork, the parent
1614 * can write to the child->flags), because we're guaranteed the
1615 * child is not running and in turn not changing child->flags
1616 * at the same time the parent does it.
1617 */
1618#define clear_stopped_child_used_math(child)    do { (child)->flags &= ~PF_USED_MATH; } while (0)
1619#define set_stopped_child_used_math(child)      do { (child)->flags |= PF_USED_MATH; } while (0)
1620#define clear_used_math()                       clear_stopped_child_used_math(current)
1621#define set_used_math()                         set_stopped_child_used_math(current)
1622
1623#define conditional_stopped_child_used_math(condition, child) \
1624        do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1625
1626#define conditional_used_math(condition)        conditional_stopped_child_used_math(condition, current)
1627
1628#define copy_to_stopped_child_used_math(child) \
1629        do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1630
1631/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1632#define tsk_used_math(p)                        ((p)->flags & PF_USED_MATH)
1633#define used_math()                             tsk_used_math(current)
1634
1635static inline bool is_percpu_thread(void)
1636{
1637#ifdef CONFIG_SMP
1638        return (current->flags & PF_NO_SETAFFINITY) &&
1639                (current->nr_cpus_allowed  == 1);
1640#else
1641        return true;
1642#endif
1643}
1644
1645/* Per-process atomic flags. */
1646#define PFA_NO_NEW_PRIVS                0       /* May not gain new privileges. */
1647#define PFA_SPREAD_PAGE                 1       /* Spread page cache over cpuset */
1648#define PFA_SPREAD_SLAB                 2       /* Spread some slab caches over cpuset */
1649#define PFA_SPEC_SSB_DISABLE            3       /* Speculative Store Bypass disabled */
1650#define PFA_SPEC_SSB_FORCE_DISABLE      4       /* Speculative Store Bypass force disabled*/
1651#define PFA_SPEC_IB_DISABLE             5       /* Indirect branch speculation restricted */
1652#define PFA_SPEC_IB_FORCE_DISABLE       6       /* Indirect branch speculation permanently restricted */
1653#define PFA_SPEC_SSB_NOEXEC             7       /* Speculative Store Bypass clear on execve() */
1654
1655#define TASK_PFA_TEST(name, func)                                       \
1656        static inline bool task_##func(struct task_struct *p)           \
1657        { return test_bit(PFA_##name, &p->atomic_flags); }
1658
1659#define TASK_PFA_SET(name, func)                                        \
1660        static inline void task_set_##func(struct task_struct *p)       \
1661        { set_bit(PFA_##name, &p->atomic_flags); }
1662
1663#define TASK_PFA_CLEAR(name, func)                                      \
1664        static inline void task_clear_##func(struct task_struct *p)     \
1665        { clear_bit(PFA_##name, &p->atomic_flags); }
1666
1667TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1668TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1669
1670TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1671TASK_PFA_SET(SPREAD_PAGE, spread_page)
1672TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1673
1674TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1675TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1676TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1677
1678TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1679TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1680TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1681
1682TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1683TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1684TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1685
1686TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1687TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1688
1689TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1690TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1691TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1692
1693TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1694TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1695
1696static inline void
1697current_restore_flags(unsigned long orig_flags, unsigned long flags)
1698{
1699        current->flags &= ~flags;
1700        current->flags |= orig_flags & flags;
1701}
1702
1703extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1704extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1705#ifdef CONFIG_SMP
1706extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1707extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1708#else
1709static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1710{
1711}
1712static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1713{
1714        if (!cpumask_test_cpu(0, new_mask))
1715                return -EINVAL;
1716        return 0;
1717}
1718#endif
1719
1720extern int yield_to(struct task_struct *p, bool preempt);
1721extern void set_user_nice(struct task_struct *p, long nice);
1722extern int task_prio(const struct task_struct *p);
1723
1724/**
1725 * task_nice - return the nice value of a given task.
1726 * @p: the task in question.
1727 *
1728 * Return: The nice value [ -20 ... 0 ... 19 ].
1729 */
1730static inline int task_nice(const struct task_struct *p)
1731{
1732        return PRIO_TO_NICE((p)->static_prio);
1733}
1734
1735extern int can_nice(const struct task_struct *p, const int nice);
1736extern int task_curr(const struct task_struct *p);
1737extern int idle_cpu(int cpu);
1738extern int available_idle_cpu(int cpu);
1739extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1740extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1741extern void sched_set_fifo(struct task_struct *p);
1742extern void sched_set_fifo_low(struct task_struct *p);
1743extern void sched_set_normal(struct task_struct *p, int nice);
1744extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1745extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1746extern struct task_struct *idle_task(int cpu);
1747
1748/**
1749 * is_idle_task - is the specified task an idle task?
1750 * @p: the task in question.
1751 *
1752 * Return: 1 if @p is an idle task. 0 otherwise.
1753 */
1754static __always_inline bool is_idle_task(const struct task_struct *p)
1755{
1756        return !!(p->flags & PF_IDLE);
1757}
1758
1759extern struct task_struct *curr_task(int cpu);
1760extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1761
1762void yield(void);
1763
1764union thread_union {
1765#ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1766        struct task_struct task;
1767#endif
1768#ifndef CONFIG_THREAD_INFO_IN_TASK
1769        struct thread_info thread_info;
1770#endif
1771        unsigned long stack[THREAD_SIZE/sizeof(long)];
1772};
1773
1774#ifndef CONFIG_THREAD_INFO_IN_TASK
1775extern struct thread_info init_thread_info;
1776#endif
1777
1778extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1779
1780#ifdef CONFIG_THREAD_INFO_IN_TASK
1781static inline struct thread_info *task_thread_info(struct task_struct *task)
1782{
1783        return &task->thread_info;
1784}
1785#elif !defined(__HAVE_THREAD_FUNCTIONS)
1786# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1787#endif
1788
1789/*
1790 * find a task by one of its numerical ids
1791 *
1792 * find_task_by_pid_ns():
1793 *      finds a task by its pid in the specified namespace
1794 * find_task_by_vpid():
1795 *      finds a task by its virtual pid
1796 *
1797 * see also find_vpid() etc in include/linux/pid.h
1798 */
1799
1800extern struct task_struct *find_task_by_vpid(pid_t nr);
1801extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1802
1803/*
1804 * find a task by its virtual pid and get the task struct
1805 */
1806extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1807
1808extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1809extern int wake_up_process(struct task_struct *tsk);
1810extern void wake_up_new_task(struct task_struct *tsk);
1811
1812#ifdef CONFIG_SMP
1813extern void kick_process(struct task_struct *tsk);
1814#else
1815static inline void kick_process(struct task_struct *tsk) { }
1816#endif
1817
1818extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1819
1820static inline void set_task_comm(struct task_struct *tsk, const char *from)
1821{
1822        __set_task_comm(tsk, from, false);
1823}
1824
1825extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1826#define get_task_comm(buf, tsk) ({                      \
1827        BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN);     \
1828        __get_task_comm(buf, sizeof(buf), tsk);         \
1829})
1830
1831#ifdef CONFIG_SMP
1832static __always_inline void scheduler_ipi(void)
1833{
1834        /*
1835         * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1836         * TIF_NEED_RESCHED remotely (for the first time) will also send
1837         * this IPI.
1838         */
1839        preempt_fold_need_resched();
1840}
1841extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1842#else
1843static inline void scheduler_ipi(void) { }
1844static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1845{
1846        return 1;
1847}
1848#endif
1849
1850/*
1851 * Set thread flags in other task's structures.
1852 * See asm/thread_info.h for TIF_xxxx flags available:
1853 */
1854static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1855{
1856        set_ti_thread_flag(task_thread_info(tsk), flag);
1857}
1858
1859static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1860{
1861        clear_ti_thread_flag(task_thread_info(tsk), flag);
1862}
1863
1864static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1865                                          bool value)
1866{
1867        update_ti_thread_flag(task_thread_info(tsk), flag, value);
1868}
1869
1870static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1871{
1872        return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1873}
1874
1875static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1876{
1877        return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1878}
1879
1880static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1881{
1882        return test_ti_thread_flag(task_thread_info(tsk), flag);
1883}
1884
1885static inline void set_tsk_need_resched(struct task_struct *tsk)
1886{
1887        set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1888}
1889
1890static inline void clear_tsk_need_resched(struct task_struct *tsk)
1891{
1892        clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1893}
1894
1895static inline int test_tsk_need_resched(struct task_struct *tsk)
1896{
1897        return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1898}
1899
1900/*
1901 * cond_resched() and cond_resched_lock(): latency reduction via
1902 * explicit rescheduling in places that are safe. The return
1903 * value indicates whether a reschedule was done in fact.
1904 * cond_resched_lock() will drop the spinlock before scheduling,
1905 */
1906#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1907extern int __cond_resched(void);
1908
1909#ifdef CONFIG_PREEMPT_DYNAMIC
1910
1911DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1912
1913static __always_inline int _cond_resched(void)
1914{
1915        return static_call_mod(cond_resched)();
1916}
1917
1918#else
1919
1920static inline int _cond_resched(void)
1921{
1922        return __cond_resched();
1923}
1924
1925#endif /* CONFIG_PREEMPT_DYNAMIC */
1926
1927#else
1928
1929static inline int _cond_resched(void) { return 0; }
1930
1931#endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
1932
1933#define cond_resched() ({                       \
1934        ___might_sleep(__FILE__, __LINE__, 0);  \
1935        _cond_resched();                        \
1936})
1937
1938extern int __cond_resched_lock(spinlock_t *lock);
1939extern int __cond_resched_rwlock_read(rwlock_t *lock);
1940extern int __cond_resched_rwlock_write(rwlock_t *lock);
1941
1942#define cond_resched_lock(lock) ({                              \
1943        ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1944        __cond_resched_lock(lock);                              \
1945})
1946
1947#define cond_resched_rwlock_read(lock) ({                       \
1948        __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1949        __cond_resched_rwlock_read(lock);                       \
1950})
1951
1952#define cond_resched_rwlock_write(lock) ({                      \
1953        __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1954        __cond_resched_rwlock_write(lock);                      \
1955})
1956
1957static inline void cond_resched_rcu(void)
1958{
1959#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1960        rcu_read_unlock();
1961        cond_resched();
1962        rcu_read_lock();
1963#endif
1964}
1965
1966/*
1967 * Does a critical section need to be broken due to another
1968 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1969 * but a general need for low latency)
1970 */
1971static inline int spin_needbreak(spinlock_t *lock)
1972{
1973#ifdef CONFIG_PREEMPTION
1974        return spin_is_contended(lock);
1975#else
1976        return 0;
1977#endif
1978}
1979
1980/*
1981 * Check if a rwlock is contended.
1982 * Returns non-zero if there is another task waiting on the rwlock.
1983 * Returns zero if the lock is not contended or the system / underlying
1984 * rwlock implementation does not support contention detection.
1985 * Technically does not depend on CONFIG_PREEMPTION, but a general need
1986 * for low latency.
1987 */
1988static inline int rwlock_needbreak(rwlock_t *lock)
1989{
1990#ifdef CONFIG_PREEMPTION
1991        return rwlock_is_contended(lock);
1992#else
1993        return 0;
1994#endif
1995}
1996
1997static __always_inline bool need_resched(void)
1998{
1999        return unlikely(tif_need_resched());
2000}
2001
2002/*
2003 * Wrappers for p->thread_info->cpu access. No-op on UP.
2004 */
2005#ifdef CONFIG_SMP
2006
2007static inline unsigned int task_cpu(const struct task_struct *p)
2008{
2009#ifdef CONFIG_THREAD_INFO_IN_TASK
2010        return READ_ONCE(p->cpu);
2011#else
2012        return READ_ONCE(task_thread_info(p)->cpu);
2013#endif
2014}
2015
2016extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2017
2018#else
2019
2020static inline unsigned int task_cpu(const struct task_struct *p)
2021{
2022        return 0;
2023}
2024
2025static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2026{
2027}
2028
2029#endif /* CONFIG_SMP */
2030
2031extern bool sched_task_on_rq(struct task_struct *p);
2032
2033/*
2034 * In order to reduce various lock holder preemption latencies provide an
2035 * interface to see if a vCPU is currently running or not.
2036 *
2037 * This allows us to terminate optimistic spin loops and block, analogous to
2038 * the native optimistic spin heuristic of testing if the lock owner task is
2039 * running or not.
2040 */
2041#ifndef vcpu_is_preempted
2042static inline bool vcpu_is_preempted(int cpu)
2043{
2044        return false;
2045}
2046#endif
2047
2048extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2049extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2050
2051#ifndef TASK_SIZE_OF
2052#define TASK_SIZE_OF(tsk)       TASK_SIZE
2053#endif
2054
2055#ifdef CONFIG_SMP
2056/* Returns effective CPU energy utilization, as seen by the scheduler */
2057unsigned long sched_cpu_util(int cpu, unsigned long max);
2058#endif /* CONFIG_SMP */
2059
2060#ifdef CONFIG_RSEQ
2061
2062/*
2063 * Map the event mask on the user-space ABI enum rseq_cs_flags
2064 * for direct mask checks.
2065 */
2066enum rseq_event_mask_bits {
2067        RSEQ_EVENT_PREEMPT_BIT  = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2068        RSEQ_EVENT_SIGNAL_BIT   = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2069        RSEQ_EVENT_MIGRATE_BIT  = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2070};
2071
2072enum rseq_event_mask {
2073        RSEQ_EVENT_PREEMPT      = (1U << RSEQ_EVENT_PREEMPT_BIT),
2074        RSEQ_EVENT_SIGNAL       = (1U << RSEQ_EVENT_SIGNAL_BIT),
2075        RSEQ_EVENT_MIGRATE      = (1U << RSEQ_EVENT_MIGRATE_BIT),
2076};
2077
2078static inline void rseq_set_notify_resume(struct task_struct *t)
2079{
2080        if (t->rseq)
2081                set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2082}
2083
2084void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2085
2086static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2087                                             struct pt_regs *regs)
2088{
2089        if (current->rseq)
2090                __rseq_handle_notify_resume(ksig, regs);
2091}
2092
2093static inline void rseq_signal_deliver(struct ksignal *ksig,
2094                                       struct pt_regs *regs)
2095{
2096        preempt_disable();
2097        __set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
2098        preempt_enable();
2099        rseq_handle_notify_resume(ksig, regs);
2100}
2101
2102/* rseq_preempt() requires preemption to be disabled. */
2103static inline void rseq_preempt(struct task_struct *t)
2104{
2105        __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2106        rseq_set_notify_resume(t);
2107}
2108
2109/* rseq_migrate() requires preemption to be disabled. */
2110static inline void rseq_migrate(struct task_struct *t)
2111{
2112        __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2113        rseq_set_notify_resume(t);
2114}
2115
2116/*
2117 * If parent process has a registered restartable sequences area, the
2118 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2119 */
2120static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2121{
2122        if (clone_flags & CLONE_VM) {
2123                t->rseq = NULL;
2124                t->rseq_sig = 0;
2125                t->rseq_event_mask = 0;
2126        } else {
2127                t->rseq = current->rseq;
2128                t->rseq_sig = current->rseq_sig;
2129                t->rseq_event_mask = current->rseq_event_mask;
2130        }
2131}
2132
2133static inline void rseq_execve(struct task_struct *t)
2134{
2135        t->rseq = NULL;
2136        t->rseq_sig = 0;
2137        t->rseq_event_mask = 0;
2138}
2139
2140#else
2141
2142static inline void rseq_set_notify_resume(struct task_struct *t)
2143{
2144}
2145static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2146                                             struct pt_regs *regs)
2147{
2148}
2149static inline void rseq_signal_deliver(struct ksignal *ksig,
2150                                       struct pt_regs *regs)
2151{
2152}
2153static inline void rseq_preempt(struct task_struct *t)
2154{
2155}
2156static inline void rseq_migrate(struct task_struct *t)
2157{
2158}
2159static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2160{
2161}
2162static inline void rseq_execve(struct task_struct *t)
2163{
2164}
2165
2166#endif
2167
2168#ifdef CONFIG_DEBUG_RSEQ
2169
2170void rseq_syscall(struct pt_regs *regs);
2171
2172#else
2173
2174static inline void rseq_syscall(struct pt_regs *regs)
2175{
2176}
2177
2178#endif
2179
2180const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2181char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2182int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2183
2184const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2185const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2186const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2187
2188int sched_trace_rq_cpu(struct rq *rq);
2189int sched_trace_rq_cpu_capacity(struct rq *rq);
2190int sched_trace_rq_nr_running(struct rq *rq);
2191
2192const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2193
2194#ifdef CONFIG_SCHED_CORE
2195extern void sched_core_free(struct task_struct *tsk);
2196extern void sched_core_fork(struct task_struct *p);
2197extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2198                                unsigned long uaddr);
2199#else
2200static inline void sched_core_free(struct task_struct *tsk) { }
2201static inline void sched_core_fork(struct task_struct *p) { }
2202#endif
2203
2204#endif
2205
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.