linux/include/linux/sched.h
<<
>>
Prefs
   1#ifndef _LINUX_SCHED_H
   2#define _LINUX_SCHED_H
   3
   4#include <uapi/linux/sched.h>
   5
   6#include <linux/sched/prio.h>
   7
   8
   9struct sched_param {
  10        int sched_priority;
  11};
  12
  13#include <asm/param.h>  /* for HZ */
  14
  15#include <linux/capability.h>
  16#include <linux/threads.h>
  17#include <linux/kernel.h>
  18#include <linux/types.h>
  19#include <linux/timex.h>
  20#include <linux/jiffies.h>
  21#include <linux/plist.h>
  22#include <linux/rbtree.h>
  23#include <linux/thread_info.h>
  24#include <linux/cpumask.h>
  25#include <linux/errno.h>
  26#include <linux/nodemask.h>
  27#include <linux/mm_types.h>
  28#include <linux/preempt.h>
  29
  30#include <asm/page.h>
  31#include <asm/ptrace.h>
  32#include <linux/cputime.h>
  33
  34#include <linux/smp.h>
  35#include <linux/sem.h>
  36#include <linux/shm.h>
  37#include <linux/signal.h>
  38#include <linux/compiler.h>
  39#include <linux/completion.h>
  40#include <linux/pid.h>
  41#include <linux/percpu.h>
  42#include <linux/topology.h>
  43#include <linux/seccomp.h>
  44#include <linux/rcupdate.h>
  45#include <linux/rculist.h>
  46#include <linux/rtmutex.h>
  47
  48#include <linux/time.h>
  49#include <linux/param.h>
  50#include <linux/resource.h>
  51#include <linux/timer.h>
  52#include <linux/hrtimer.h>
  53#include <linux/kcov.h>
  54#include <linux/task_io_accounting.h>
  55#include <linux/latencytop.h>
  56#include <linux/cred.h>
  57#include <linux/llist.h>
  58#include <linux/uidgid.h>
  59#include <linux/gfp.h>
  60#include <linux/magic.h>
  61#include <linux/cgroup-defs.h>
  62
  63#include <asm/processor.h>
  64
  65#define SCHED_ATTR_SIZE_VER0    48      /* sizeof first published struct */
  66
  67/*
  68 * Extended scheduling parameters data structure.
  69 *
  70 * This is needed because the original struct sched_param can not be
  71 * altered without introducing ABI issues with legacy applications
  72 * (e.g., in sched_getparam()).
  73 *
  74 * However, the possibility of specifying more than just a priority for
  75 * the tasks may be useful for a wide variety of application fields, e.g.,
  76 * multimedia, streaming, automation and control, and many others.
  77 *
  78 * This variant (sched_attr) is meant at describing a so-called
  79 * sporadic time-constrained task. In such model a task is specified by:
  80 *  - the activation period or minimum instance inter-arrival time;
  81 *  - the maximum (or average, depending on the actual scheduling
  82 *    discipline) computation time of all instances, a.k.a. runtime;
  83 *  - the deadline (relative to the actual activation time) of each
  84 *    instance.
  85 * Very briefly, a periodic (sporadic) task asks for the execution of
  86 * some specific computation --which is typically called an instance--
  87 * (at most) every period. Moreover, each instance typically lasts no more
  88 * than the runtime and must be completed by time instant t equal to
  89 * the instance activation time + the deadline.
  90 *
  91 * This is reflected by the actual fields of the sched_attr structure:
  92 *
  93 *  @size               size of the structure, for fwd/bwd compat.
  94 *
  95 *  @sched_policy       task's scheduling policy
  96 *  @sched_flags        for customizing the scheduler behaviour
  97 *  @sched_nice         task's nice value      (SCHED_NORMAL/BATCH)
  98 *  @sched_priority     task's static priority (SCHED_FIFO/RR)
  99 *  @sched_deadline     representative of the task's deadline
 100 *  @sched_runtime      representative of the task's runtime
 101 *  @sched_period       representative of the task's period
 102 *
 103 * Given this task model, there are a multiplicity of scheduling algorithms
 104 * and policies, that can be used to ensure all the tasks will make their
 105 * timing constraints.
 106 *
 107 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
 108 * only user of this new interface. More information about the algorithm
 109 * available in the scheduling class file or in Documentation/.
 110 */
 111struct sched_attr {
 112        u32 size;
 113
 114        u32 sched_policy;
 115        u64 sched_flags;
 116
 117        /* SCHED_NORMAL, SCHED_BATCH */
 118        s32 sched_nice;
 119
 120        /* SCHED_FIFO, SCHED_RR */
 121        u32 sched_priority;
 122
 123        /* SCHED_DEADLINE */
 124        u64 sched_runtime;
 125        u64 sched_deadline;
 126        u64 sched_period;
 127};
 128
 129struct futex_pi_state;
 130struct robust_list_head;
 131struct bio_list;
 132struct fs_struct;
 133struct perf_event_context;
 134struct blk_plug;
 135struct filename;
 136struct nameidata;
 137
 138#define VMACACHE_BITS 2
 139#define VMACACHE_SIZE (1U << VMACACHE_BITS)
 140#define VMACACHE_MASK (VMACACHE_SIZE - 1)
 141
 142/*
 143 * These are the constant used to fake the fixed-point load-average
 144 * counting. Some notes:
 145 *  - 11 bit fractions expand to 22 bits by the multiplies: this gives
 146 *    a load-average precision of 10 bits integer + 11 bits fractional
 147 *  - if you want to count load-averages more often, you need more
 148 *    precision, or rounding will get you. With 2-second counting freq,
 149 *    the EXP_n values would be 1981, 2034 and 2043 if still using only
 150 *    11 bit fractions.
 151 */
 152extern unsigned long avenrun[];         /* Load averages */
 153extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
 154
 155#define FSHIFT          11              /* nr of bits of precision */
 156#define FIXED_1         (1<<FSHIFT)     /* 1.0 as fixed-point */
 157#define LOAD_FREQ       (5*HZ+1)        /* 5 sec intervals */
 158#define EXP_1           1884            /* 1/exp(5sec/1min) as fixed-point */
 159#define EXP_5           2014            /* 1/exp(5sec/5min) */
 160#define EXP_15          2037            /* 1/exp(5sec/15min) */
 161
 162#define CALC_LOAD(load,exp,n) \
 163        load *= exp; \
 164        load += n*(FIXED_1-exp); \
 165        load >>= FSHIFT;
 166
 167extern unsigned long total_forks;
 168extern int nr_threads;
 169DECLARE_PER_CPU(unsigned long, process_counts);
 170extern int nr_processes(void);
 171extern unsigned long nr_running(void);
 172extern bool single_task_running(void);
 173extern unsigned long nr_iowait(void);
 174extern unsigned long nr_iowait_cpu(int cpu);
 175extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
 176
 177extern void calc_global_load(unsigned long ticks);
 178
 179#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
 180extern void cpu_load_update_nohz_start(void);
 181extern void cpu_load_update_nohz_stop(void);
 182#else
 183static inline void cpu_load_update_nohz_start(void) { }
 184static inline void cpu_load_update_nohz_stop(void) { }
 185#endif
 186
 187extern void dump_cpu_task(int cpu);
 188
 189struct seq_file;
 190struct cfs_rq;
 191struct task_group;
 192#ifdef CONFIG_SCHED_DEBUG
 193extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
 194extern void proc_sched_set_task(struct task_struct *p);
 195#endif
 196
 197/*
 198 * Task state bitmask. NOTE! These bits are also
 199 * encoded in fs/proc/array.c: get_task_state().
 200 *
 201 * We have two separate sets of flags: task->state
 202 * is about runnability, while task->exit_state are
 203 * about the task exiting. Confusing, but this way
 204 * modifying one set can't modify the other one by
 205 * mistake.
 206 */
 207#define TASK_RUNNING            0
 208#define TASK_INTERRUPTIBLE      1
 209#define TASK_UNINTERRUPTIBLE    2
 210#define __TASK_STOPPED          4
 211#define __TASK_TRACED           8
 212/* in tsk->exit_state */
 213#define EXIT_DEAD               16
 214#define EXIT_ZOMBIE             32
 215#define EXIT_TRACE              (EXIT_ZOMBIE | EXIT_DEAD)
 216/* in tsk->state again */
 217#define TASK_DEAD               64
 218#define TASK_WAKEKILL           128
 219#define TASK_WAKING             256
 220#define TASK_PARKED             512
 221#define TASK_NOLOAD             1024
 222#define TASK_NEW                2048
 223#define TASK_STATE_MAX          4096
 224
 225#define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
 226
 227extern char ___assert_task_state[1 - 2*!!(
 228                sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
 229
 230/* Convenience macros for the sake of set_task_state */
 231#define TASK_KILLABLE           (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
 232#define TASK_STOPPED            (TASK_WAKEKILL | __TASK_STOPPED)
 233#define TASK_TRACED             (TASK_WAKEKILL | __TASK_TRACED)
 234
 235#define TASK_IDLE               (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
 236
 237/* Convenience macros for the sake of wake_up */
 238#define TASK_NORMAL             (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
 239#define TASK_ALL                (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
 240
 241/* get_task_state() */
 242#define TASK_REPORT             (TASK_RUNNING | TASK_INTERRUPTIBLE | \
 243                                 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
 244                                 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
 245
 246#define task_is_traced(task)    ((task->state & __TASK_TRACED) != 0)
 247#define task_is_stopped(task)   ((task->state & __TASK_STOPPED) != 0)
 248#define task_is_stopped_or_traced(task) \
 249                        ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
 250#define task_contributes_to_load(task)  \
 251                                ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
 252                                 (task->flags & PF_FROZEN) == 0 && \
 253                                 (task->state & TASK_NOLOAD) == 0)
 254
 255#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
 256
 257#define __set_task_state(tsk, state_value)                      \
 258        do {                                                    \
 259                (tsk)->task_state_change = _THIS_IP_;           \
 260                (tsk)->state = (state_value);                   \
 261        } while (0)
 262#define set_task_state(tsk, state_value)                        \
 263        do {                                                    \
 264                (tsk)->task_state_change = _THIS_IP_;           \
 265                smp_store_mb((tsk)->state, (state_value));              \
 266        } while (0)
 267
 268/*
 269 * set_current_state() includes a barrier so that the write of current->state
 270 * is correctly serialised wrt the caller's subsequent test of whether to
 271 * actually sleep:
 272 *
 273 *      set_current_state(TASK_UNINTERRUPTIBLE);
 274 *      if (do_i_need_to_sleep())
 275 *              schedule();
 276 *
 277 * If the caller does not need such serialisation then use __set_current_state()
 278 */
 279#define __set_current_state(state_value)                        \
 280        do {                                                    \
 281                current->task_state_change = _THIS_IP_;         \
 282                current->state = (state_value);                 \
 283        } while (0)
 284#define set_current_state(state_value)                          \
 285        do {                                                    \
 286                current->task_state_change = _THIS_IP_;         \
 287                smp_store_mb(current->state, (state_value));            \
 288        } while (0)
 289
 290#else
 291
 292#define __set_task_state(tsk, state_value)              \
 293        do { (tsk)->state = (state_value); } while (0)
 294#define set_task_state(tsk, state_value)                \
 295        smp_store_mb((tsk)->state, (state_value))
 296
 297/*
 298 * set_current_state() includes a barrier so that the write of current->state
 299 * is correctly serialised wrt the caller's subsequent test of whether to
 300 * actually sleep:
 301 *
 302 *      set_current_state(TASK_UNINTERRUPTIBLE);
 303 *      if (do_i_need_to_sleep())
 304 *              schedule();
 305 *
 306 * If the caller does not need such serialisation then use __set_current_state()
 307 */
 308#define __set_current_state(state_value)                \
 309        do { current->state = (state_value); } while (0)
 310#define set_current_state(state_value)                  \
 311        smp_store_mb(current->state, (state_value))
 312
 313#endif
 314
 315/* Task command name length */
 316#define TASK_COMM_LEN 16
 317
 318#include <linux/spinlock.h>
 319
 320/*
 321 * This serializes "schedule()" and also protects
 322 * the run-queue from deletions/modifications (but
 323 * _adding_ to the beginning of the run-queue has
 324 * a separate lock).
 325 */
 326extern rwlock_t tasklist_lock;
 327extern spinlock_t mmlist_lock;
 328
 329struct task_struct;
 330
 331#ifdef CONFIG_PROVE_RCU
 332extern int lockdep_tasklist_lock_is_held(void);
 333#endif /* #ifdef CONFIG_PROVE_RCU */
 334
 335extern void sched_init(void);
 336extern void sched_init_smp(void);
 337extern asmlinkage void schedule_tail(struct task_struct *prev);
 338extern void init_idle(struct task_struct *idle, int cpu);
 339extern void init_idle_bootup_task(struct task_struct *idle);
 340
 341extern cpumask_var_t cpu_isolated_map;
 342
 343extern int runqueue_is_locked(int cpu);
 344
 345#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
 346extern void nohz_balance_enter_idle(int cpu);
 347extern void set_cpu_sd_state_idle(void);
 348extern int get_nohz_timer_target(void);
 349#else
 350static inline void nohz_balance_enter_idle(int cpu) { }
 351static inline void set_cpu_sd_state_idle(void) { }
 352#endif
 353
 354/*
 355 * Only dump TASK_* tasks. (0 for all tasks)
 356 */
 357extern void show_state_filter(unsigned long state_filter);
 358
 359static inline void show_state(void)
 360{
 361        show_state_filter(0);
 362}
 363
 364extern void show_regs(struct pt_regs *);
 365
 366/*
 367 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
 368 * task), SP is the stack pointer of the first frame that should be shown in the back
 369 * trace (or NULL if the entire call-chain of the task should be shown).
 370 */
 371extern void show_stack(struct task_struct *task, unsigned long *sp);
 372
 373extern void cpu_init (void);
 374extern void trap_init(void);
 375extern void update_process_times(int user);
 376extern void scheduler_tick(void);
 377extern int sched_cpu_starting(unsigned int cpu);
 378extern int sched_cpu_activate(unsigned int cpu);
 379extern int sched_cpu_deactivate(unsigned int cpu);
 380
 381#ifdef CONFIG_HOTPLUG_CPU
 382extern int sched_cpu_dying(unsigned int cpu);
 383#else
 384# define sched_cpu_dying        NULL
 385#endif
 386
 387extern void sched_show_task(struct task_struct *p);
 388
 389#ifdef CONFIG_LOCKUP_DETECTOR
 390extern void touch_softlockup_watchdog_sched(void);
 391extern void touch_softlockup_watchdog(void);
 392extern void touch_softlockup_watchdog_sync(void);
 393extern void touch_all_softlockup_watchdogs(void);
 394extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
 395                                  void __user *buffer,
 396                                  size_t *lenp, loff_t *ppos);
 397extern unsigned int  softlockup_panic;
 398extern unsigned int  hardlockup_panic;
 399void lockup_detector_init(void);
 400#else
 401static inline void touch_softlockup_watchdog_sched(void)
 402{
 403}
 404static inline void touch_softlockup_watchdog(void)
 405{
 406}
 407static inline void touch_softlockup_watchdog_sync(void)
 408{
 409}
 410static inline void touch_all_softlockup_watchdogs(void)
 411{
 412}
 413static inline void lockup_detector_init(void)
 414{
 415}
 416#endif
 417
 418#ifdef CONFIG_DETECT_HUNG_TASK
 419void reset_hung_task_detector(void);
 420#else
 421static inline void reset_hung_task_detector(void)
 422{
 423}
 424#endif
 425
 426/* Attach to any functions which should be ignored in wchan output. */
 427#define __sched         __attribute__((__section__(".sched.text")))
 428
 429/* Linker adds these: start and end of __sched functions */
 430extern char __sched_text_start[], __sched_text_end[];
 431
 432/* Is this address in the __sched functions? */
 433extern int in_sched_functions(unsigned long addr);
 434
 435#define MAX_SCHEDULE_TIMEOUT    LONG_MAX
 436extern signed long schedule_timeout(signed long timeout);
 437extern signed long schedule_timeout_interruptible(signed long timeout);
 438extern signed long schedule_timeout_killable(signed long timeout);
 439extern signed long schedule_timeout_uninterruptible(signed long timeout);
 440extern signed long schedule_timeout_idle(signed long timeout);
 441asmlinkage void schedule(void);
 442extern void schedule_preempt_disabled(void);
 443
 444extern long io_schedule_timeout(long timeout);
 445
 446static inline void io_schedule(void)
 447{
 448        io_schedule_timeout(MAX_SCHEDULE_TIMEOUT);
 449}
 450
 451void __noreturn do_task_dead(void);
 452
 453struct nsproxy;
 454struct user_namespace;
 455
 456#ifdef CONFIG_MMU
 457extern void arch_pick_mmap_layout(struct mm_struct *mm);
 458extern unsigned long
 459arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
 460                       unsigned long, unsigned long);
 461extern unsigned long
 462arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
 463                          unsigned long len, unsigned long pgoff,
 464                          unsigned long flags);
 465#else
 466static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
 467#endif
 468
 469#define SUID_DUMP_DISABLE       0       /* No setuid dumping */
 470#define SUID_DUMP_USER          1       /* Dump as user of process */
 471#define SUID_DUMP_ROOT          2       /* Dump as root */
 472
 473/* mm flags */
 474
 475/* for SUID_DUMP_* above */
 476#define MMF_DUMPABLE_BITS 2
 477#define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
 478
 479extern void set_dumpable(struct mm_struct *mm, int value);
 480/*
 481 * This returns the actual value of the suid_dumpable flag. For things
 482 * that are using this for checking for privilege transitions, it must
 483 * test against SUID_DUMP_USER rather than treating it as a boolean
 484 * value.
 485 */
 486static inline int __get_dumpable(unsigned long mm_flags)
 487{
 488        return mm_flags & MMF_DUMPABLE_MASK;
 489}
 490
 491static inline int get_dumpable(struct mm_struct *mm)
 492{
 493        return __get_dumpable(mm->flags);
 494}
 495
 496/* coredump filter bits */
 497#define MMF_DUMP_ANON_PRIVATE   2
 498#define MMF_DUMP_ANON_SHARED    3
 499#define MMF_DUMP_MAPPED_PRIVATE 4
 500#define MMF_DUMP_MAPPED_SHARED  5
 501#define MMF_DUMP_ELF_HEADERS    6
 502#define MMF_DUMP_HUGETLB_PRIVATE 7
 503#define MMF_DUMP_HUGETLB_SHARED  8
 504#define MMF_DUMP_DAX_PRIVATE    9
 505#define MMF_DUMP_DAX_SHARED     10
 506
 507#define MMF_DUMP_FILTER_SHIFT   MMF_DUMPABLE_BITS
 508#define MMF_DUMP_FILTER_BITS    9
 509#define MMF_DUMP_FILTER_MASK \
 510        (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
 511#define MMF_DUMP_FILTER_DEFAULT \
 512        ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
 513         (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
 514
 515#ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
 516# define MMF_DUMP_MASK_DEFAULT_ELF      (1 << MMF_DUMP_ELF_HEADERS)
 517#else
 518# define MMF_DUMP_MASK_DEFAULT_ELF      0
 519#endif
 520                                        /* leave room for more dump flags */
 521#define MMF_VM_MERGEABLE        16      /* KSM may merge identical pages */
 522#define MMF_VM_HUGEPAGE         17      /* set when VM_HUGEPAGE is set on vma */
 523#define MMF_EXE_FILE_CHANGED    18      /* see prctl_set_mm_exe_file() */
 524
 525#define MMF_HAS_UPROBES         19      /* has uprobes */
 526#define MMF_RECALC_UPROBES      20      /* MMF_HAS_UPROBES can be wrong */
 527#define MMF_OOM_SKIP            21      /* mm is of no interest for the OOM killer */
 528#define MMF_UNSTABLE            22      /* mm is unstable for copy_from_user */
 529#define MMF_HUGE_ZERO_PAGE      23      /* mm has ever used the global huge zero page */
 530
 531#define MMF_INIT_MASK           (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
 532
 533struct sighand_struct {
 534        atomic_t                count;
 535        struct k_sigaction      action[_NSIG];
 536        spinlock_t              siglock;
 537        wait_queue_head_t       signalfd_wqh;
 538};
 539
 540struct pacct_struct {
 541        int                     ac_flag;
 542        long                    ac_exitcode;
 543        unsigned long           ac_mem;
 544        cputime_t               ac_utime, ac_stime;
 545        unsigned long           ac_minflt, ac_majflt;
 546};
 547
 548struct cpu_itimer {
 549        cputime_t expires;
 550        cputime_t incr;
 551        u32 error;
 552        u32 incr_error;
 553};
 554
 555/**
 556 * struct prev_cputime - snaphsot of system and user cputime
 557 * @utime: time spent in user mode
 558 * @stime: time spent in system mode
 559 * @lock: protects the above two fields
 560 *
 561 * Stores previous user/system time values such that we can guarantee
 562 * monotonicity.
 563 */
 564struct prev_cputime {
 565#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
 566        cputime_t utime;
 567        cputime_t stime;
 568        raw_spinlock_t lock;
 569#endif
 570};
 571
 572static inline void prev_cputime_init(struct prev_cputime *prev)
 573{
 574#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
 575        prev->utime = prev->stime = 0;
 576        raw_spin_lock_init(&prev->lock);
 577#endif
 578}
 579
 580/**
 581 * struct task_cputime - collected CPU time counts
 582 * @utime:              time spent in user mode, in &cputime_t units
 583 * @stime:              time spent in kernel mode, in &cputime_t units
 584 * @sum_exec_runtime:   total time spent on the CPU, in nanoseconds
 585 *
 586 * This structure groups together three kinds of CPU time that are tracked for
 587 * threads and thread groups.  Most things considering CPU time want to group
 588 * these counts together and treat all three of them in parallel.
 589 */
 590struct task_cputime {
 591        cputime_t utime;
 592        cputime_t stime;
 593        unsigned long long sum_exec_runtime;
 594};
 595
 596/* Alternate field names when used to cache expirations. */
 597#define virt_exp        utime
 598#define prof_exp        stime
 599#define sched_exp       sum_exec_runtime
 600
 601#define INIT_CPUTIME    \
 602        (struct task_cputime) {                                 \
 603                .utime = 0,                                     \
 604                .stime = 0,                                     \
 605                .sum_exec_runtime = 0,                          \
 606        }
 607
 608/*
 609 * This is the atomic variant of task_cputime, which can be used for
 610 * storing and updating task_cputime statistics without locking.
 611 */
 612struct task_cputime_atomic {
 613        atomic64_t utime;
 614        atomic64_t stime;
 615        atomic64_t sum_exec_runtime;
 616};
 617
 618#define INIT_CPUTIME_ATOMIC \
 619        (struct task_cputime_atomic) {                          \
 620                .utime = ATOMIC64_INIT(0),                      \
 621                .stime = ATOMIC64_INIT(0),                      \
 622                .sum_exec_runtime = ATOMIC64_INIT(0),           \
 623        }
 624
 625#define PREEMPT_DISABLED        (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
 626
 627/*
 628 * Disable preemption until the scheduler is running -- use an unconditional
 629 * value so that it also works on !PREEMPT_COUNT kernels.
 630 *
 631 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
 632 */
 633#define INIT_PREEMPT_COUNT      PREEMPT_OFFSET
 634
 635/*
 636 * Initial preempt_count value; reflects the preempt_count schedule invariant
 637 * which states that during context switches:
 638 *
 639 *    preempt_count() == 2*PREEMPT_DISABLE_OFFSET
 640 *
 641 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
 642 * Note: See finish_task_switch().
 643 */
 644#define FORK_PREEMPT_COUNT      (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
 645
 646/**
 647 * struct thread_group_cputimer - thread group interval timer counts
 648 * @cputime_atomic:     atomic thread group interval timers.
 649 * @running:            true when there are timers running and
 650 *                      @cputime_atomic receives updates.
 651 * @checking_timer:     true when a thread in the group is in the
 652 *                      process of checking for thread group timers.
 653 *
 654 * This structure contains the version of task_cputime, above, that is
 655 * used for thread group CPU timer calculations.
 656 */
 657struct thread_group_cputimer {
 658        struct task_cputime_atomic cputime_atomic;
 659        bool running;
 660        bool checking_timer;
 661};
 662
 663#include <linux/rwsem.h>
 664struct autogroup;
 665
 666/*
 667 * NOTE! "signal_struct" does not have its own
 668 * locking, because a shared signal_struct always
 669 * implies a shared sighand_struct, so locking
 670 * sighand_struct is always a proper superset of
 671 * the locking of signal_struct.
 672 */
 673struct signal_struct {
 674        atomic_t                sigcnt;
 675        atomic_t                live;
 676        int                     nr_threads;
 677        struct list_head        thread_head;
 678
 679        wait_queue_head_t       wait_chldexit;  /* for wait4() */
 680
 681        /* current thread group signal load-balancing target: */
 682        struct task_struct      *curr_target;
 683
 684        /* shared signal handling: */
 685        struct sigpending       shared_pending;
 686
 687        /* thread group exit support */
 688        int                     group_exit_code;
 689        /* overloaded:
 690         * - notify group_exit_task when ->count is equal to notify_count
 691         * - everyone except group_exit_task is stopped during signal delivery
 692         *   of fatal signals, group_exit_task processes the signal.
 693         */
 694        int                     notify_count;
 695        struct task_struct      *group_exit_task;
 696
 697        /* thread group stop support, overloads group_exit_code too */
 698        int                     group_stop_count;
 699        unsigned int            flags; /* see SIGNAL_* flags below */
 700
 701        /*
 702         * PR_SET_CHILD_SUBREAPER marks a process, like a service
 703         * manager, to re-parent orphan (double-forking) child processes
 704         * to this process instead of 'init'. The service manager is
 705         * able to receive SIGCHLD signals and is able to investigate
 706         * the process until it calls wait(). All children of this
 707         * process will inherit a flag if they should look for a
 708         * child_subreaper process at exit.
 709         */
 710        unsigned int            is_child_subreaper:1;
 711        unsigned int            has_child_subreaper:1;
 712
 713        /* POSIX.1b Interval Timers */
 714        int                     posix_timer_id;
 715        struct list_head        posix_timers;
 716
 717        /* ITIMER_REAL timer for the process */
 718        struct hrtimer real_timer;
 719        struct pid *leader_pid;
 720        ktime_t it_real_incr;
 721
 722        /*
 723         * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
 724         * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
 725         * values are defined to 0 and 1 respectively
 726         */
 727        struct cpu_itimer it[2];
 728
 729        /*
 730         * Thread group totals for process CPU timers.
 731         * See thread_group_cputimer(), et al, for details.
 732         */
 733        struct thread_group_cputimer cputimer;
 734
 735        /* Earliest-expiration cache. */
 736        struct task_cputime cputime_expires;
 737
 738#ifdef CONFIG_NO_HZ_FULL
 739        atomic_t tick_dep_mask;
 740#endif
 741
 742        struct list_head cpu_timers[3];
 743
 744        struct pid *tty_old_pgrp;
 745
 746        /* boolean value for session group leader */
 747        int leader;
 748
 749        struct tty_struct *tty; /* NULL if no tty */
 750
 751#ifdef CONFIG_SCHED_AUTOGROUP
 752        struct autogroup *autogroup;
 753#endif
 754        /*
 755         * Cumulative resource counters for dead threads in the group,
 756         * and for reaped dead child processes forked by this group.
 757         * Live threads maintain their own counters and add to these
 758         * in __exit_signal, except for the group leader.
 759         */
 760        seqlock_t stats_lock;
 761        cputime_t utime, stime, cutime, cstime;
 762        cputime_t gtime;
 763        cputime_t cgtime;
 764        struct prev_cputime prev_cputime;
 765        unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
 766        unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
 767        unsigned long inblock, oublock, cinblock, coublock;
 768        unsigned long maxrss, cmaxrss;
 769        struct task_io_accounting ioac;
 770
 771        /*
 772         * Cumulative ns of schedule CPU time fo dead threads in the
 773         * group, not including a zombie group leader, (This only differs
 774         * from jiffies_to_ns(utime + stime) if sched_clock uses something
 775         * other than jiffies.)
 776         */
 777        unsigned long long sum_sched_runtime;
 778
 779        /*
 780         * We don't bother to synchronize most readers of this at all,
 781         * because there is no reader checking a limit that actually needs
 782         * to get both rlim_cur and rlim_max atomically, and either one
 783         * alone is a single word that can safely be read normally.
 784         * getrlimit/setrlimit use task_lock(current->group_leader) to
 785         * protect this instead of the siglock, because they really
 786         * have no need to disable irqs.
 787         */
 788        struct rlimit rlim[RLIM_NLIMITS];
 789
 790#ifdef CONFIG_BSD_PROCESS_ACCT
 791        struct pacct_struct pacct;      /* per-process accounting information */
 792#endif
 793#ifdef CONFIG_TASKSTATS
 794        struct taskstats *stats;
 795#endif
 796#ifdef CONFIG_AUDIT
 797        unsigned audit_tty;
 798        struct tty_audit_buf *tty_audit_buf;
 799#endif
 800
 801        /*
 802         * Thread is the potential origin of an oom condition; kill first on
 803         * oom
 804         */
 805        bool oom_flag_origin;
 806        short oom_score_adj;            /* OOM kill score adjustment */
 807        short oom_score_adj_min;        /* OOM kill score adjustment min value.
 808                                         * Only settable by CAP_SYS_RESOURCE. */
 809        struct mm_struct *oom_mm;       /* recorded mm when the thread group got
 810                                         * killed by the oom killer */
 811
 812        struct mutex cred_guard_mutex;  /* guard against foreign influences on
 813                                         * credential calculations
 814                                         * (notably. ptrace) */
 815};
 816
 817/*
 818 * Bits in flags field of signal_struct.
 819 */
 820#define SIGNAL_STOP_STOPPED     0x00000001 /* job control stop in effect */
 821#define SIGNAL_STOP_CONTINUED   0x00000002 /* SIGCONT since WCONTINUED reap */
 822#define SIGNAL_GROUP_EXIT       0x00000004 /* group exit in progress */
 823#define SIGNAL_GROUP_COREDUMP   0x00000008 /* coredump in progress */
 824/*
 825 * Pending notifications to parent.
 826 */
 827#define SIGNAL_CLD_STOPPED      0x00000010
 828#define SIGNAL_CLD_CONTINUED    0x00000020
 829#define SIGNAL_CLD_MASK         (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
 830
 831#define SIGNAL_UNKILLABLE       0x00000040 /* for init: ignore fatal signals */
 832
 833/* If true, all threads except ->group_exit_task have pending SIGKILL */
 834static inline int signal_group_exit(const struct signal_struct *sig)
 835{
 836        return  (sig->flags & SIGNAL_GROUP_EXIT) ||
 837                (sig->group_exit_task != NULL);
 838}
 839
 840/*
 841 * Some day this will be a full-fledged user tracking system..
 842 */
 843struct user_struct {
 844        atomic_t __count;       /* reference count */
 845        atomic_t processes;     /* How many processes does this user have? */
 846        atomic_t sigpending;    /* How many pending signals does this user have? */
 847#ifdef CONFIG_INOTIFY_USER
 848        atomic_t inotify_watches; /* How many inotify watches does this user have? */
 849        atomic_t inotify_devs;  /* How many inotify devs does this user have opened? */
 850#endif
 851#ifdef CONFIG_FANOTIFY
 852        atomic_t fanotify_listeners;
 853#endif
 854#ifdef CONFIG_EPOLL
 855        atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
 856#endif
 857#ifdef CONFIG_POSIX_MQUEUE
 858        /* protected by mq_lock */
 859        unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
 860#endif
 861        unsigned long locked_shm; /* How many pages of mlocked shm ? */
 862        unsigned long unix_inflight;    /* How many files in flight in unix sockets */
 863        atomic_long_t pipe_bufs;  /* how many pages are allocated in pipe buffers */
 864
 865#ifdef CONFIG_KEYS
 866        struct key *uid_keyring;        /* UID specific keyring */
 867        struct key *session_keyring;    /* UID's default session keyring */
 868#endif
 869
 870        /* Hash table maintenance information */
 871        struct hlist_node uidhash_node;
 872        kuid_t uid;
 873
 874#if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
 875        atomic_long_t locked_vm;
 876#endif
 877};
 878
 879extern int uids_sysfs_init(void);
 880
 881extern struct user_struct *find_user(kuid_t);
 882
 883extern struct user_struct root_user;
 884#define INIT_USER (&root_user)
 885
 886
 887struct backing_dev_info;
 888struct reclaim_state;
 889
 890#ifdef CONFIG_SCHED_INFO
 891struct sched_info {
 892        /* cumulative counters */
 893        unsigned long pcount;         /* # of times run on this cpu */
 894        unsigned long long run_delay; /* time spent waiting on a runqueue */
 895
 896        /* timestamps */
 897        unsigned long long last_arrival,/* when we last ran on a cpu */
 898                           last_queued; /* when we were last queued to run */
 899};
 900#endif /* CONFIG_SCHED_INFO */
 901
 902#ifdef CONFIG_TASK_DELAY_ACCT
 903struct task_delay_info {
 904        spinlock_t      lock;
 905        unsigned int    flags;  /* Private per-task flags */
 906
 907        /* For each stat XXX, add following, aligned appropriately
 908         *
 909         * struct timespec XXX_start, XXX_end;
 910         * u64 XXX_delay;
 911         * u32 XXX_count;
 912         *
 913         * Atomicity of updates to XXX_delay, XXX_count protected by
 914         * single lock above (split into XXX_lock if contention is an issue).
 915         */
 916
 917        /*
 918         * XXX_count is incremented on every XXX operation, the delay
 919         * associated with the operation is added to XXX_delay.
 920         * XXX_delay contains the accumulated delay time in nanoseconds.
 921         */
 922        u64 blkio_start;        /* Shared by blkio, swapin */
 923        u64 blkio_delay;        /* wait for sync block io completion */
 924        u64 swapin_delay;       /* wait for swapin block io completion */
 925        u32 blkio_count;        /* total count of the number of sync block */
 926                                /* io operations performed */
 927        u32 swapin_count;       /* total count of the number of swapin block */
 928                                /* io operations performed */
 929
 930        u64 freepages_start;
 931        u64 freepages_delay;    /* wait for memory reclaim */
 932        u32 freepages_count;    /* total count of memory reclaim */
 933};
 934#endif  /* CONFIG_TASK_DELAY_ACCT */
 935
 936static inline int sched_info_on(void)
 937{
 938#ifdef CONFIG_SCHEDSTATS
 939        return 1;
 940#elif defined(CONFIG_TASK_DELAY_ACCT)
 941        extern int delayacct_on;
 942        return delayacct_on;
 943#else
 944        return 0;
 945#endif
 946}
 947
 948#ifdef CONFIG_SCHEDSTATS
 949void force_schedstat_enabled(void);
 950#endif
 951
 952enum cpu_idle_type {
 953        CPU_IDLE,
 954        CPU_NOT_IDLE,
 955        CPU_NEWLY_IDLE,
 956        CPU_MAX_IDLE_TYPES
 957};
 958
 959/*
 960 * Integer metrics need fixed point arithmetic, e.g., sched/fair
 961 * has a few: load, load_avg, util_avg, freq, and capacity.
 962 *
 963 * We define a basic fixed point arithmetic range, and then formalize
 964 * all these metrics based on that basic range.
 965 */
 966# define SCHED_FIXEDPOINT_SHIFT 10
 967# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
 968
 969/*
 970 * Increase resolution of cpu_capacity calculations
 971 */
 972#define SCHED_CAPACITY_SHIFT    SCHED_FIXEDPOINT_SHIFT
 973#define SCHED_CAPACITY_SCALE    (1L << SCHED_CAPACITY_SHIFT)
 974
 975/*
 976 * Wake-queues are lists of tasks with a pending wakeup, whose
 977 * callers have already marked the task as woken internally,
 978 * and can thus carry on. A common use case is being able to
 979 * do the wakeups once the corresponding user lock as been
 980 * released.
 981 *
 982 * We hold reference to each task in the list across the wakeup,
 983 * thus guaranteeing that the memory is still valid by the time
 984 * the actual wakeups are performed in wake_up_q().
 985 *
 986 * One per task suffices, because there's never a need for a task to be
 987 * in two wake queues simultaneously; it is forbidden to abandon a task
 988 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
 989 * already in a wake queue, the wakeup will happen soon and the second
 990 * waker can just skip it.
 991 *
 992 * The WAKE_Q macro declares and initializes the list head.
 993 * wake_up_q() does NOT reinitialize the list; it's expected to be
 994 * called near the end of a function, where the fact that the queue is
 995 * not used again will be easy to see by inspection.
 996 *
 997 * Note that this can cause spurious wakeups. schedule() callers
 998 * must ensure the call is done inside a loop, confirming that the
 999 * wakeup condition has in fact occurred.
1000 */
1001struct wake_q_node {
1002        struct wake_q_node *next;
1003};
1004
1005struct wake_q_head {
1006        struct wake_q_node *first;
1007        struct wake_q_node **lastp;
1008};
1009
1010#define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1011
1012#define WAKE_Q(name)                                    \
1013        struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1014
1015extern void wake_q_add(struct wake_q_head *head,
1016                       struct task_struct *task);
1017extern void wake_up_q(struct wake_q_head *head);
1018
1019/*
1020 * sched-domains (multiprocessor balancing) declarations:
1021 */
1022#ifdef CONFIG_SMP
1023#define SD_LOAD_BALANCE         0x0001  /* Do load balancing on this domain. */
1024#define SD_BALANCE_NEWIDLE      0x0002  /* Balance when about to become idle */
1025#define SD_BALANCE_EXEC         0x0004  /* Balance on exec */
1026#define SD_BALANCE_FORK         0x0008  /* Balance on fork, clone */
1027#define SD_BALANCE_WAKE         0x0010  /* Balance on wakeup */
1028#define SD_WAKE_AFFINE          0x0020  /* Wake task to waking CPU */
1029#define SD_ASYM_CPUCAPACITY     0x0040  /* Groups have different max cpu capacities */
1030#define SD_SHARE_CPUCAPACITY    0x0080  /* Domain members share cpu capacity */
1031#define SD_SHARE_POWERDOMAIN    0x0100  /* Domain members share power domain */
1032#define SD_SHARE_PKG_RESOURCES  0x0200  /* Domain members share cpu pkg resources */
1033#define SD_SERIALIZE            0x0400  /* Only a single load balancing instance */
1034#define SD_ASYM_PACKING         0x0800  /* Place busy groups earlier in the domain */
1035#define SD_PREFER_SIBLING       0x1000  /* Prefer to place tasks in a sibling domain */
1036#define SD_OVERLAP              0x2000  /* sched_domains of this level overlap */
1037#define SD_NUMA                 0x4000  /* cross-node balancing */
1038
1039#ifdef CONFIG_SCHED_SMT
1040static inline int cpu_smt_flags(void)
1041{
1042        return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1043}
1044#endif
1045
1046#ifdef CONFIG_SCHED_MC
1047static inline int cpu_core_flags(void)
1048{
1049        return SD_SHARE_PKG_RESOURCES;
1050}
1051#endif
1052
1053#ifdef CONFIG_NUMA
1054static inline int cpu_numa_flags(void)
1055{
1056        return SD_NUMA;
1057}
1058#endif
1059
1060struct sched_domain_attr {
1061        int relax_domain_level;
1062};
1063
1064#define SD_ATTR_INIT    (struct sched_domain_attr) {    \
1065        .relax_domain_level = -1,                       \
1066}
1067
1068extern int sched_domain_level_max;
1069
1070struct sched_group;
1071
1072struct sched_domain_shared {
1073        atomic_t        ref;
1074        atomic_t        nr_busy_cpus;
1075        int             has_idle_cores;
1076};
1077
1078struct sched_domain {
1079        /* These fields must be setup */
1080        struct sched_domain *parent;    /* top domain must be null terminated */
1081        struct sched_domain *child;     /* bottom domain must be null terminated */
1082        struct sched_group *groups;     /* the balancing groups of the domain */
1083        unsigned long min_interval;     /* Minimum balance interval ms */
1084        unsigned long max_interval;     /* Maximum balance interval ms */
1085        unsigned int busy_factor;       /* less balancing by factor if busy */
1086        unsigned int imbalance_pct;     /* No balance until over watermark */
1087        unsigned int cache_nice_tries;  /* Leave cache hot tasks for # tries */
1088        unsigned int busy_idx;
1089        unsigned int idle_idx;
1090        unsigned int newidle_idx;
1091        unsigned int wake_idx;
1092        unsigned int forkexec_idx;
1093        unsigned int smt_gain;
1094
1095        int nohz_idle;                  /* NOHZ IDLE status */
1096        int flags;                      /* See SD_* */
1097        int level;
1098
1099        /* Runtime fields. */
1100        unsigned long last_balance;     /* init to jiffies. units in jiffies */
1101        unsigned int balance_interval;  /* initialise to 1. units in ms. */
1102        unsigned int nr_balance_failed; /* initialise to 0 */
1103
1104        /* idle_balance() stats */
1105        u64 max_newidle_lb_cost;
1106        unsigned long next_decay_max_lb_cost;
1107
1108        u64 avg_scan_cost;              /* select_idle_sibling */
1109
1110#ifdef CONFIG_SCHEDSTATS
1111        /* load_balance() stats */
1112        unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1113        unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1114        unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1115        unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1116        unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1117        unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1118        unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1119        unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1120
1121        /* Active load balancing */
1122        unsigned int alb_count;
1123        unsigned int alb_failed;
1124        unsigned int alb_pushed;
1125
1126        /* SD_BALANCE_EXEC stats */
1127        unsigned int sbe_count;
1128        unsigned int sbe_balanced;
1129        unsigned int sbe_pushed;
1130
1131        /* SD_BALANCE_FORK stats */
1132        unsigned int sbf_count;
1133        unsigned int sbf_balanced;
1134        unsigned int sbf_pushed;
1135
1136        /* try_to_wake_up() stats */
1137        unsigned int ttwu_wake_remote;
1138        unsigned int ttwu_move_affine;
1139        unsigned int ttwu_move_balance;
1140#endif
1141#ifdef CONFIG_SCHED_DEBUG
1142        char *name;
1143#endif
1144        union {
1145                void *private;          /* used during construction */
1146                struct rcu_head rcu;    /* used during destruction */
1147        };
1148        struct sched_domain_shared *shared;
1149
1150        unsigned int span_weight;
1151        /*
1152         * Span of all CPUs in this domain.
1153         *
1154         * NOTE: this field is variable length. (Allocated dynamically
1155         * by attaching extra space to the end of the structure,
1156         * depending on how many CPUs the kernel has booted up with)
1157         */
1158        unsigned long span[0];
1159};
1160
1161static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1162{
1163        return to_cpumask(sd->span);
1164}
1165
1166extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1167                                    struct sched_domain_attr *dattr_new);
1168
1169/* Allocate an array of sched domains, for partition_sched_domains(). */
1170cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1171void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1172
1173bool cpus_share_cache(int this_cpu, int that_cpu);
1174
1175typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1176typedef int (*sched_domain_flags_f)(void);
1177
1178#define SDTL_OVERLAP    0x01
1179
1180struct sd_data {
1181        struct sched_domain **__percpu sd;
1182        struct sched_domain_shared **__percpu sds;
1183        struct sched_group **__percpu sg;
1184        struct sched_group_capacity **__percpu sgc;
1185};
1186
1187struct sched_domain_topology_level {
1188        sched_domain_mask_f mask;
1189        sched_domain_flags_f sd_flags;
1190        int                 flags;
1191        int                 numa_level;
1192        struct sd_data      data;
1193#ifdef CONFIG_SCHED_DEBUG
1194        char                *name;
1195#endif
1196};
1197
1198extern void set_sched_topology(struct sched_domain_topology_level *tl);
1199extern void wake_up_if_idle(int cpu);
1200
1201#ifdef CONFIG_SCHED_DEBUG
1202# define SD_INIT_NAME(type)             .name = #type
1203#else
1204# define SD_INIT_NAME(type)
1205#endif
1206
1207#else /* CONFIG_SMP */
1208
1209struct sched_domain_attr;
1210
1211static inline void
1212partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1213                        struct sched_domain_attr *dattr_new)
1214{
1215}
1216
1217static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1218{
1219        return true;
1220}
1221
1222#endif  /* !CONFIG_SMP */
1223
1224
1225struct io_context;                      /* See blkdev.h */
1226
1227
1228#ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1229extern void prefetch_stack(struct task_struct *t);
1230#else
1231static inline void prefetch_stack(struct task_struct *t) { }
1232#endif
1233
1234struct audit_context;           /* See audit.c */
1235struct mempolicy;
1236struct pipe_inode_info;
1237struct uts_namespace;
1238
1239struct load_weight {
1240        unsigned long weight;
1241        u32 inv_weight;
1242};
1243
1244/*
1245 * The load_avg/util_avg accumulates an infinite geometric series
1246 * (see __update_load_avg() in kernel/sched/fair.c).
1247 *
1248 * [load_avg definition]
1249 *
1250 *   load_avg = runnable% * scale_load_down(load)
1251 *
1252 * where runnable% is the time ratio that a sched_entity is runnable.
1253 * For cfs_rq, it is the aggregated load_avg of all runnable and
1254 * blocked sched_entities.
1255 *
1256 * load_avg may also take frequency scaling into account:
1257 *
1258 *   load_avg = runnable% * scale_load_down(load) * freq%
1259 *
1260 * where freq% is the CPU frequency normalized to the highest frequency.
1261 *
1262 * [util_avg definition]
1263 *
1264 *   util_avg = running% * SCHED_CAPACITY_SCALE
1265 *
1266 * where running% is the time ratio that a sched_entity is running on
1267 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1268 * and blocked sched_entities.
1269 *
1270 * util_avg may also factor frequency scaling and CPU capacity scaling:
1271 *
1272 *   util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1273 *
1274 * where freq% is the same as above, and capacity% is the CPU capacity
1275 * normalized to the greatest capacity (due to uarch differences, etc).
1276 *
1277 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1278 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1279 * we therefore scale them to as large a range as necessary. This is for
1280 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1281 *
1282 * [Overflow issue]
1283 *
1284 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1285 * with the highest load (=88761), always runnable on a single cfs_rq,
1286 * and should not overflow as the number already hits PID_MAX_LIMIT.
1287 *
1288 * For all other cases (including 32-bit kernels), struct load_weight's
1289 * weight will overflow first before we do, because:
1290 *
1291 *    Max(load_avg) <= Max(load.weight)
1292 *
1293 * Then it is the load_weight's responsibility to consider overflow
1294 * issues.
1295 */
1296struct sched_avg {
1297        u64 last_update_time, load_sum;
1298        u32 util_sum, period_contrib;
1299        unsigned long load_avg, util_avg;
1300};
1301
1302#ifdef CONFIG_SCHEDSTATS
1303struct sched_statistics {
1304        u64                     wait_start;
1305        u64                     wait_max;
1306        u64                     wait_count;
1307        u64                     wait_sum;
1308        u64                     iowait_count;
1309        u64                     iowait_sum;
1310
1311        u64                     sleep_start;
1312        u64                     sleep_max;
1313        s64                     sum_sleep_runtime;
1314
1315        u64                     block_start;
1316        u64                     block_max;
1317        u64                     exec_max;
1318        u64                     slice_max;
1319
1320        u64                     nr_migrations_cold;
1321        u64                     nr_failed_migrations_affine;
1322        u64                     nr_failed_migrations_running;
1323        u64                     nr_failed_migrations_hot;
1324        u64                     nr_forced_migrations;
1325
1326        u64                     nr_wakeups;
1327        u64                     nr_wakeups_sync;
1328        u64                     nr_wakeups_migrate;
1329        u64                     nr_wakeups_local;
1330        u64                     nr_wakeups_remote;
1331        u64                     nr_wakeups_affine;
1332        u64                     nr_wakeups_affine_attempts;
1333        u64                     nr_wakeups_passive;
1334        u64                     nr_wakeups_idle;
1335};
1336#endif
1337
1338struct sched_entity {
1339        struct load_weight      load;           /* for load-balancing */
1340        struct rb_node          run_node;
1341        struct list_head        group_node;
1342        unsigned int            on_rq;
1343
1344        u64                     exec_start;
1345        u64                     sum_exec_runtime;
1346        u64                     vruntime;
1347        u64                     prev_sum_exec_runtime;
1348
1349        u64                     nr_migrations;
1350
1351#ifdef CONFIG_SCHEDSTATS
1352        struct sched_statistics statistics;
1353#endif
1354
1355#ifdef CONFIG_FAIR_GROUP_SCHED
1356        int                     depth;
1357        struct sched_entity     *parent;
1358        /* rq on which this entity is (to be) queued: */
1359        struct cfs_rq           *cfs_rq;
1360        /* rq "owned" by this entity/group: */
1361        struct cfs_rq           *my_q;
1362#endif
1363
1364#ifdef CONFIG_SMP
1365        /*
1366         * Per entity load average tracking.
1367         *
1368         * Put into separate cache line so it does not
1369         * collide with read-mostly values above.
1370         */
1371        struct sched_avg        avg ____cacheline_aligned_in_smp;
1372#endif
1373};
1374
1375struct sched_rt_entity {
1376        struct list_head run_list;
1377        unsigned long timeout;
1378        unsigned long watchdog_stamp;
1379        unsigned int time_slice;
1380        unsigned short on_rq;
1381        unsigned short on_list;
1382
1383        struct sched_rt_entity *back;
1384#ifdef CONFIG_RT_GROUP_SCHED
1385        struct sched_rt_entity  *parent;
1386        /* rq on which this entity is (to be) queued: */
1387        struct rt_rq            *rt_rq;
1388        /* rq "owned" by this entity/group: */
1389        struct rt_rq            *my_q;
1390#endif
1391};
1392
1393struct sched_dl_entity {
1394        struct rb_node  rb_node;
1395
1396        /*
1397         * Original scheduling parameters. Copied here from sched_attr
1398         * during sched_setattr(), they will remain the same until
1399         * the next sched_setattr().
1400         */
1401        u64 dl_runtime;         /* maximum runtime for each instance    */
1402        u64 dl_deadline;        /* relative deadline of each instance   */
1403        u64 dl_period;          /* separation of two instances (period) */
1404        u64 dl_bw;              /* dl_runtime / dl_deadline             */
1405
1406        /*
1407         * Actual scheduling parameters. Initialized with the values above,
1408         * they are continously updated during task execution. Note that
1409         * the remaining runtime could be < 0 in case we are in overrun.
1410         */
1411        s64 runtime;            /* remaining runtime for this instance  */
1412        u64 deadline;           /* absolute deadline for this instance  */
1413        unsigned int flags;     /* specifying the scheduler behaviour   */
1414
1415        /*
1416         * Some bool flags:
1417         *
1418         * @dl_throttled tells if we exhausted the runtime. If so, the
1419         * task has to wait for a replenishment to be performed at the
1420         * next firing of dl_timer.
1421         *
1422         * @dl_boosted tells if we are boosted due to DI. If so we are
1423         * outside bandwidth enforcement mechanism (but only until we
1424         * exit the critical section);
1425         *
1426         * @dl_yielded tells if task gave up the cpu before consuming
1427         * all its available runtime during the last job.
1428         */
1429        int dl_throttled, dl_boosted, dl_yielded;
1430
1431        /*
1432         * Bandwidth enforcement timer. Each -deadline task has its
1433         * own bandwidth to be enforced, thus we need one timer per task.
1434         */
1435        struct hrtimer dl_timer;
1436};
1437
1438union rcu_special {
1439        struct {
1440                u8 blocked;
1441                u8 need_qs;
1442                u8 exp_need_qs;
1443                u8 pad; /* Otherwise the compiler can store garbage here. */
1444        } b; /* Bits. */
1445        u32 s; /* Set of bits. */
1446};
1447struct rcu_node;
1448
1449enum perf_event_task_context {
1450        perf_invalid_context = -1,
1451        perf_hw_context = 0,
1452        perf_sw_context,
1453        perf_nr_task_contexts,
1454};
1455
1456/* Track pages that require TLB flushes */
1457struct tlbflush_unmap_batch {
1458        /*
1459         * Each bit set is a CPU that potentially has a TLB entry for one of
1460         * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1461         */
1462        struct cpumask cpumask;
1463
1464        /* True if any bit in cpumask is set */
1465        bool flush_required;
1466
1467        /*
1468         * If true then the PTE was dirty when unmapped. The entry must be
1469         * flushed before IO is initiated or a stale TLB entry potentially
1470         * allows an update without redirtying the page.
1471         */
1472        bool writable;
1473};
1474
1475struct task_struct {
1476#ifdef CONFIG_THREAD_INFO_IN_TASK
1477        /*
1478         * For reasons of header soup (see current_thread_info()), this
1479         * must be the first element of task_struct.
1480         */
1481        struct thread_info thread_info;
1482#endif
1483        volatile long state;    /* -1 unrunnable, 0 runnable, >0 stopped */
1484        void *stack;
1485        atomic_t usage;
1486        unsigned int flags;     /* per process flags, defined below */
1487        unsigned int ptrace;
1488
1489#ifdef CONFIG_SMP
1490        struct llist_node wake_entry;
1491        int on_cpu;
1492#ifdef CONFIG_THREAD_INFO_IN_TASK
1493        unsigned int cpu;       /* current CPU */
1494#endif
1495        unsigned int wakee_flips;
1496        unsigned long wakee_flip_decay_ts;
1497        struct task_struct *last_wakee;
1498
1499        int wake_cpu;
1500#endif
1501        int on_rq;
1502
1503        int prio, static_prio, normal_prio;
1504        unsigned int rt_priority;
1505        const struct sched_class *sched_class;
1506        struct sched_entity se;
1507        struct sched_rt_entity rt;
1508#ifdef CONFIG_CGROUP_SCHED
1509        struct task_group *sched_task_group;
1510#endif
1511        struct sched_dl_entity dl;
1512
1513#ifdef CONFIG_PREEMPT_NOTIFIERS
1514        /* list of struct preempt_notifier: */
1515        struct hlist_head preempt_notifiers;
1516#endif
1517
1518#ifdef CONFIG_BLK_DEV_IO_TRACE
1519        unsigned int btrace_seq;
1520#endif
1521
1522        unsigned int policy;
1523        int nr_cpus_allowed;
1524        cpumask_t cpus_allowed;
1525
1526#ifdef CONFIG_PREEMPT_RCU
1527        int rcu_read_lock_nesting;
1528        union rcu_special rcu_read_unlock_special;
1529        struct list_head rcu_node_entry;
1530        struct rcu_node *rcu_blocked_node;
1531#endif /* #ifdef CONFIG_PREEMPT_RCU */
1532#ifdef CONFIG_TASKS_RCU
1533        unsigned long rcu_tasks_nvcsw;
1534        bool rcu_tasks_holdout;
1535        struct list_head rcu_tasks_holdout_list;
1536        int rcu_tasks_idle_cpu;
1537#endif /* #ifdef CONFIG_TASKS_RCU */
1538
1539#ifdef CONFIG_SCHED_INFO
1540        struct sched_info sched_info;
1541#endif
1542
1543        struct list_head tasks;
1544#ifdef CONFIG_SMP
1545        struct plist_node pushable_tasks;
1546        struct rb_node pushable_dl_tasks;
1547#endif
1548
1549        struct mm_struct *mm, *active_mm;
1550        /* per-thread vma caching */
1551        u32 vmacache_seqnum;
1552        struct vm_area_struct *vmacache[VMACACHE_SIZE];
1553#if defined(SPLIT_RSS_COUNTING)
1554        struct task_rss_stat    rss_stat;
1555#endif
1556/* task state */
1557        int exit_state;
1558        int exit_code, exit_signal;
1559        int pdeath_signal;  /*  The signal sent when the parent dies  */
1560        unsigned long jobctl;   /* JOBCTL_*, siglock protected */
1561
1562        /* Used for emulating ABI behavior of previous Linux versions */
1563        unsigned int personality;
1564
1565        /* scheduler bits, serialized by scheduler locks */
1566        unsigned sched_reset_on_fork:1;
1567        unsigned sched_contributes_to_load:1;
1568        unsigned sched_migrated:1;
1569        unsigned sched_remote_wakeup:1;
1570        unsigned :0; /* force alignment to the next boundary */
1571
1572        /* unserialized, strictly 'current' */
1573        unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1574        unsigned in_iowait:1;
1575#if !defined(TIF_RESTORE_SIGMASK)
1576        unsigned restore_sigmask:1;
1577#endif
1578#ifdef CONFIG_MEMCG
1579        unsigned memcg_may_oom:1;
1580#ifndef CONFIG_SLOB
1581        unsigned memcg_kmem_skip_account:1;
1582#endif
1583#endif
1584#ifdef CONFIG_COMPAT_BRK
1585        unsigned brk_randomized:1;
1586#endif
1587
1588        unsigned long atomic_flags; /* Flags needing atomic access. */
1589
1590        struct restart_block restart_block;
1591
1592        pid_t pid;
1593        pid_t tgid;
1594
1595#ifdef CONFIG_CC_STACKPROTECTOR
1596        /* Canary value for the -fstack-protector gcc feature */
1597        unsigned long stack_canary;
1598#endif
1599        /*
1600         * pointers to (original) parent process, youngest child, younger sibling,
1601         * older sibling, respectively.  (p->father can be replaced with
1602         * p->real_parent->pid)
1603         */
1604        struct task_struct __rcu *real_parent; /* real parent process */
1605        struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1606        /*
1607         * children/sibling forms the list of my natural children
1608         */
1609        struct list_head children;      /* list of my children */
1610        struct list_head sibling;       /* linkage in my parent's children list */
1611        struct task_struct *group_leader;       /* threadgroup leader */
1612
1613        /*
1614         * ptraced is the list of tasks this task is using ptrace on.
1615         * This includes both natural children and PTRACE_ATTACH targets.
1616         * p->ptrace_entry is p's link on the p->parent->ptraced list.
1617         */
1618        struct list_head ptraced;
1619        struct list_head ptrace_entry;
1620
1621        /* PID/PID hash table linkage. */
1622        struct pid_link pids[PIDTYPE_MAX];
1623        struct list_head thread_group;
1624        struct list_head thread_node;
1625
1626        struct completion *vfork_done;          /* for vfork() */
1627        int __user *set_child_tid;              /* CLONE_CHILD_SETTID */
1628        int __user *clear_child_tid;            /* CLONE_CHILD_CLEARTID */
1629
1630        cputime_t utime, stime, utimescaled, stimescaled;
1631        cputime_t gtime;
1632        struct prev_cputime prev_cputime;
1633#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1634        seqcount_t vtime_seqcount;
1635        unsigned long long vtime_snap;
1636        enum {
1637                /* Task is sleeping or running in a CPU with VTIME inactive */
1638                VTIME_INACTIVE = 0,
1639                /* Task runs in userspace in a CPU with VTIME active */
1640                VTIME_USER,
1641                /* Task runs in kernelspace in a CPU with VTIME active */
1642                VTIME_SYS,
1643        } vtime_snap_whence;
1644#endif
1645
1646#ifdef CONFIG_NO_HZ_FULL
1647        atomic_t tick_dep_mask;
1648#endif
1649        unsigned long nvcsw, nivcsw; /* context switch counts */
1650        u64 start_time;         /* monotonic time in nsec */
1651        u64 real_start_time;    /* boot based time in nsec */
1652/* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1653        unsigned long min_flt, maj_flt;
1654
1655        struct task_cputime cputime_expires;
1656        struct list_head cpu_timers[3];
1657
1658/* process credentials */
1659        const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1660        const struct cred __rcu *real_cred; /* objective and real subjective task
1661                                         * credentials (COW) */
1662        const struct cred __rcu *cred;  /* effective (overridable) subjective task
1663                                         * credentials (COW) */
1664        char comm[TASK_COMM_LEN]; /* executable name excluding path
1665                                     - access with [gs]et_task_comm (which lock
1666                                       it with task_lock())
1667                                     - initialized normally by setup_new_exec */
1668/* file system info */
1669        struct nameidata *nameidata;
1670#ifdef CONFIG_SYSVIPC
1671/* ipc stuff */
1672        struct sysv_sem sysvsem;
1673        struct sysv_shm sysvshm;
1674#endif
1675#ifdef CONFIG_DETECT_HUNG_TASK
1676/* hung task detection */
1677        unsigned long last_switch_count;
1678#endif
1679/* filesystem information */
1680        struct fs_struct *fs;
1681/* open file information */
1682        struct files_struct *files;
1683/* namespaces */
1684        struct nsproxy *nsproxy;
1685/* signal handlers */
1686        struct signal_struct *signal;
1687        struct sighand_struct *sighand;
1688
1689        sigset_t blocked, real_blocked;
1690        sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1691        struct sigpending pending;
1692
1693        unsigned long sas_ss_sp;
1694        size_t sas_ss_size;
1695        unsigned sas_ss_flags;
1696
1697        struct callback_head *task_works;
1698
1699        struct audit_context *audit_context;
1700#ifdef CONFIG_AUDITSYSCALL
1701        kuid_t loginuid;
1702        unsigned int sessionid;
1703#endif
1704        struct seccomp seccomp;
1705
1706/* Thread group tracking */
1707        u32 parent_exec_id;
1708        u32 self_exec_id;
1709/* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1710 * mempolicy */
1711        spinlock_t alloc_lock;
1712
1713        /* Protection of the PI data structures: */
1714        raw_spinlock_t pi_lock;
1715
1716        struct wake_q_node wake_q;
1717
1718#ifdef CONFIG_RT_MUTEXES
1719        /* PI waiters blocked on a rt_mutex held by this task */
1720        struct rb_root pi_waiters;
1721        struct rb_node *pi_waiters_leftmost;
1722        /* Deadlock detection and priority inheritance handling */
1723        struct rt_mutex_waiter *pi_blocked_on;
1724#endif
1725
1726#ifdef CONFIG_DEBUG_MUTEXES
1727        /* mutex deadlock detection */
1728        struct mutex_waiter *blocked_on;
1729#endif
1730#ifdef CONFIG_TRACE_IRQFLAGS
1731        unsigned int irq_events;
1732        unsigned long hardirq_enable_ip;
1733        unsigned long hardirq_disable_ip;
1734        unsigned int hardirq_enable_event;
1735        unsigned int hardirq_disable_event;
1736        int hardirqs_enabled;
1737        int hardirq_context;
1738        unsigned long softirq_disable_ip;
1739        unsigned long softirq_enable_ip;
1740        unsigned int softirq_disable_event;
1741        unsigned int softirq_enable_event;
1742        int softirqs_enabled;
1743        int softirq_context;
1744#endif
1745#ifdef CONFIG_LOCKDEP
1746# define MAX_LOCK_DEPTH 48UL
1747        u64 curr_chain_key;
1748        int lockdep_depth;
1749        unsigned int lockdep_recursion;
1750        struct held_lock held_locks[MAX_LOCK_DEPTH];
1751        gfp_t lockdep_reclaim_gfp;
1752#endif
1753#ifdef CONFIG_UBSAN
1754        unsigned int in_ubsan;
1755#endif
1756
1757/* journalling filesystem info */
1758        void *journal_info;
1759
1760/* stacked block device info */
1761        struct bio_list *bio_list;
1762
1763#ifdef CONFIG_BLOCK
1764/* stack plugging */
1765        struct blk_plug *plug;
1766#endif
1767
1768/* VM state */
1769        struct reclaim_state *reclaim_state;
1770
1771        struct backing_dev_info *backing_dev_info;
1772
1773        struct io_context *io_context;
1774
1775        unsigned long ptrace_message;
1776        siginfo_t *last_siginfo; /* For ptrace use.  */
1777        struct task_io_accounting ioac;
1778#if defined(CONFIG_TASK_XACCT)
1779        u64 acct_rss_mem1;      /* accumulated rss usage */
1780        u64 acct_vm_mem1;       /* accumulated virtual memory usage */
1781        cputime_t acct_timexpd; /* stime + utime since last update */
1782#endif
1783#ifdef CONFIG_CPUSETS
1784        nodemask_t mems_allowed;        /* Protected by alloc_lock */
1785        seqcount_t mems_allowed_seq;    /* Seqence no to catch updates */
1786        int cpuset_mem_spread_rotor;
1787        int cpuset_slab_spread_rotor;
1788#endif
1789#ifdef CONFIG_CGROUPS
1790        /* Control Group info protected by css_set_lock */
1791        struct css_set __rcu *cgroups;
1792        /* cg_list protected by css_set_lock and tsk->alloc_lock */
1793        struct list_head cg_list;
1794#endif
1795#ifdef CONFIG_FUTEX
1796        struct robust_list_head __user *robust_list;
1797#ifdef CONFIG_COMPAT
1798        struct compat_robust_list_head __user *compat_robust_list;
1799#endif
1800        struct list_head pi_state_list;
1801        struct futex_pi_state *pi_state_cache;
1802#endif
1803#ifdef CONFIG_PERF_EVENTS
1804        struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1805        struct mutex perf_event_mutex;
1806        struct list_head perf_event_list;
1807#endif
1808#ifdef CONFIG_DEBUG_PREEMPT
1809        unsigned long preempt_disable_ip;
1810#endif
1811#ifdef CONFIG_NUMA
1812        struct mempolicy *mempolicy;    /* Protected by alloc_lock */
1813        short il_next;
1814        short pref_node_fork;
1815#endif
1816#ifdef CONFIG_NUMA_BALANCING
1817        int numa_scan_seq;
1818        unsigned int numa_scan_period;
1819        unsigned int numa_scan_period_max;
1820        int numa_preferred_nid;
1821        unsigned long numa_migrate_retry;
1822        u64 node_stamp;                 /* migration stamp  */
1823        u64 last_task_numa_placement;
1824        u64 last_sum_exec_runtime;
1825        struct callback_head numa_work;
1826
1827        struct list_head numa_entry;
1828        struct numa_group *numa_group;
1829
1830        /*
1831         * numa_faults is an array split into four regions:
1832         * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1833         * in this precise order.
1834         *
1835         * faults_memory: Exponential decaying average of faults on a per-node
1836         * basis. Scheduling placement decisions are made based on these
1837         * counts. The values remain static for the duration of a PTE scan.
1838         * faults_cpu: Track the nodes the process was running on when a NUMA
1839         * hinting fault was incurred.
1840         * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1841         * during the current scan window. When the scan completes, the counts
1842         * in faults_memory and faults_cpu decay and these values are copied.
1843         */
1844        unsigned long *numa_faults;
1845        unsigned long total_numa_faults;
1846
1847        /*
1848         * numa_faults_locality tracks if faults recorded during the last
1849         * scan window were remote/local or failed to migrate. The task scan
1850         * period is adapted based on the locality of the faults with different
1851         * weights depending on whether they were shared or private faults
1852         */
1853        unsigned long numa_faults_locality[3];
1854
1855        unsigned long numa_pages_migrated;
1856#endif /* CONFIG_NUMA_BALANCING */
1857
1858#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1859        struct tlbflush_unmap_batch tlb_ubc;
1860#endif
1861
1862        struct rcu_head rcu;
1863
1864        /*
1865         * cache last used pipe for splice
1866         */
1867        struct pipe_inode_info *splice_pipe;
1868
1869        struct page_frag task_frag;
1870
1871#ifdef  CONFIG_TASK_DELAY_ACCT
1872        struct task_delay_info *delays;
1873#endif
1874#ifdef CONFIG_FAULT_INJECTION
1875        int make_it_fail;
1876#endif
1877        /*
1878         * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1879         * balance_dirty_pages() for some dirty throttling pause
1880         */
1881        int nr_dirtied;
1882        int nr_dirtied_pause;
1883        unsigned long dirty_paused_when; /* start of a write-and-pause period */
1884
1885#ifdef CONFIG_LATENCYTOP
1886        int latency_record_count;
1887        struct latency_record latency_record[LT_SAVECOUNT];
1888#endif
1889        /*
1890         * time slack values; these are used to round up poll() and
1891         * select() etc timeout values. These are in nanoseconds.
1892         */
1893        u64 timer_slack_ns;
1894        u64 default_timer_slack_ns;
1895
1896#ifdef CONFIG_KASAN
1897        unsigned int kasan_depth;
1898#endif
1899#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1900        /* Index of current stored address in ret_stack */
1901        int curr_ret_stack;
1902        /* Stack of return addresses for return function tracing */
1903        struct ftrace_ret_stack *ret_stack;
1904        /* time stamp for last schedule */
1905        unsigned long long ftrace_timestamp;
1906        /*
1907         * Number of functions that haven't been traced
1908         * because of depth overrun.
1909         */
1910        atomic_t trace_overrun;
1911        /* Pause for the tracing */
1912        atomic_t tracing_graph_pause;
1913#endif
1914#ifdef CONFIG_TRACING
1915        /* state flags for use by tracers */
1916        unsigned long trace;
1917        /* bitmask and counter of trace recursion */
1918        unsigned long trace_recursion;
1919#endif /* CONFIG_TRACING */
1920#ifdef CONFIG_KCOV
1921        /* Coverage collection mode enabled for this task (0 if disabled). */
1922        enum kcov_mode kcov_mode;
1923        /* Size of the kcov_area. */
1924        unsigned        kcov_size;
1925        /* Buffer for coverage collection. */
1926        void            *kcov_area;
1927        /* kcov desciptor wired with this task or NULL. */
1928        struct kcov     *kcov;
1929#endif
1930#ifdef CONFIG_MEMCG
1931        struct mem_cgroup *memcg_in_oom;
1932        gfp_t memcg_oom_gfp_mask;
1933        int memcg_oom_order;
1934
1935        /* number of pages to reclaim on returning to userland */
1936        unsigned int memcg_nr_pages_over_high;
1937#endif
1938#ifdef CONFIG_UPROBES
1939        struct uprobe_task *utask;
1940#endif
1941#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1942        unsigned int    sequential_io;
1943        unsigned int    sequential_io_avg;
1944#endif
1945#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1946        unsigned long   task_state_change;
1947#endif
1948        int pagefault_disabled;
1949#ifdef CONFIG_MMU
1950        struct task_struct *oom_reaper_list;
1951#endif
1952#ifdef CONFIG_VMAP_STACK
1953        struct vm_struct *stack_vm_area;
1954#endif
1955#ifdef CONFIG_THREAD_INFO_IN_TASK
1956        /* A live task holds one reference. */
1957        atomic_t stack_refcount;
1958#endif
1959/* CPU-specific state of this task */
1960        struct thread_struct thread;
1961/*
1962 * WARNING: on x86, 'thread_struct' contains a variable-sized
1963 * structure.  It *MUST* be at the end of 'task_struct'.
1964 *
1965 * Do not put anything below here!
1966 */
1967};
1968
1969#ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1970extern int arch_task_struct_size __read_mostly;
1971#else
1972# define arch_task_struct_size (sizeof(struct task_struct))
1973#endif
1974
1975#ifdef CONFIG_VMAP_STACK
1976static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1977{
1978        return t->stack_vm_area;
1979}
1980#else
1981static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1982{
1983        return NULL;
1984}
1985#endif
1986
1987/* Future-safe accessor for struct task_struct's cpus_allowed. */
1988#define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1989
1990static inline int tsk_nr_cpus_allowed(struct task_struct *p)
1991{
1992        return p->nr_cpus_allowed;
1993}
1994
1995#define TNF_MIGRATED    0x01
1996#define TNF_NO_GROUP    0x02
1997#define TNF_SHARED      0x04
1998#define TNF_FAULT_LOCAL 0x08
1999#define TNF_MIGRATE_FAIL 0x10
2000
2001static inline bool in_vfork(struct task_struct *tsk)
2002{
2003        bool ret;
2004
2005        /*
2006         * need RCU to access ->real_parent if CLONE_VM was used along with
2007         * CLONE_PARENT.
2008         *
2009         * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2010         * imply CLONE_VM
2011         *
2012         * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2013         * ->real_parent is not necessarily the task doing vfork(), so in
2014         * theory we can't rely on task_lock() if we want to dereference it.
2015         *
2016         * And in this case we can't trust the real_parent->mm == tsk->mm
2017         * check, it can be false negative. But we do not care, if init or
2018         * another oom-unkillable task does this it should blame itself.
2019         */
2020        rcu_read_lock();
2021        ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2022        rcu_read_unlock();
2023
2024        return ret;
2025}
2026
2027#ifdef CONFIG_NUMA_BALANCING
2028extern void task_numa_fault(int last_node, int node, int pages, int flags);
2029extern pid_t task_numa_group_id(struct task_struct *p);
2030extern void set_numabalancing_state(bool enabled);
2031extern void task_numa_free(struct task_struct *p);
2032extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2033                                        int src_nid, int dst_cpu);
2034#else
2035static inline void task_numa_fault(int last_node, int node, int pages,
2036                                   int flags)
2037{
2038}
2039static inline pid_t task_numa_group_id(struct task_struct *p)
2040{
2041        return 0;
2042}
2043static inline void set_numabalancing_state(bool enabled)
2044{
2045}
2046static inline void task_numa_free(struct task_struct *p)
2047{
2048}
2049static inline bool should_numa_migrate_memory(struct task_struct *p,
2050                                struct page *page, int src_nid, int dst_cpu)
2051{
2052        return true;
2053}
2054#endif
2055
2056static inline struct pid *task_pid(struct task_struct *task)
2057{
2058        return task->pids[PIDTYPE_PID].pid;
2059}
2060
2061static inline struct pid *task_tgid(struct task_struct *task)
2062{
2063        return task->group_leader->pids[PIDTYPE_PID].pid;
2064}
2065
2066/*
2067 * Without tasklist or rcu lock it is not safe to dereference
2068 * the result of task_pgrp/task_session even if task == current,
2069 * we can race with another thread doing sys_setsid/sys_setpgid.
2070 */
2071static inline struct pid *task_pgrp(struct task_struct *task)
2072{
2073        return task->group_leader->pids[PIDTYPE_PGID].pid;
2074}
2075
2076static inline struct pid *task_session(struct task_struct *task)
2077{
2078        return task->group_leader->pids[PIDTYPE_SID].pid;
2079}
2080
2081struct pid_namespace;
2082
2083/*
2084 * the helpers to get the task's different pids as they are seen
2085 * from various namespaces
2086 *
2087 * task_xid_nr()     : global id, i.e. the id seen from the init namespace;
2088 * task_xid_vnr()    : virtual id, i.e. the id seen from the pid namespace of
2089 *                     current.
2090 * task_xid_nr_ns()  : id seen from the ns specified;
2091 *
2092 * set_task_vxid()   : assigns a virtual id to a task;
2093 *
2094 * see also pid_nr() etc in include/linux/pid.h
2095 */
2096pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2097                        struct pid_namespace *ns);
2098
2099static inline pid_t task_pid_nr(struct task_struct *tsk)
2100{
2101        return tsk->pid;
2102}
2103
2104static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2105                                        struct pid_namespace *ns)
2106{
2107        return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2108}
2109
2110static inline pid_t task_pid_vnr(struct task_struct *tsk)
2111{
2112        return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2113}
2114
2115
2116static inline pid_t task_tgid_nr(struct task_struct *tsk)
2117{
2118        return tsk->tgid;
2119}
2120
2121pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2122
2123static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2124{
2125        return pid_vnr(task_tgid(tsk));
2126}
2127
2128
2129static inline int pid_alive(const struct task_struct *p);
2130static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2131{
2132        pid_t pid = 0;
2133
2134        rcu_read_lock();
2135        if (pid_alive(tsk))
2136                pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2137        rcu_read_unlock();
2138
2139        return pid;
2140}
2141
2142static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2143{
2144        return task_ppid_nr_ns(tsk, &init_pid_ns);
2145}
2146
2147static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2148                                        struct pid_namespace *ns)
2149{
2150        return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2151}
2152
2153static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2154{
2155        return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2156}
2157
2158
2159static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2160                                        struct pid_namespace *ns)
2161{
2162        return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2163}
2164
2165static inline pid_t task_session_vnr(struct task_struct *tsk)
2166{
2167        return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2168}
2169
2170/* obsolete, do not use */
2171static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2172{
2173        return task_pgrp_nr_ns(tsk, &init_pid_ns);
2174}
2175
2176/**
2177 * pid_alive - check that a task structure is not stale
2178 * @p: Task structure to be checked.
2179 *
2180 * Test if a process is not yet dead (at most zombie state)
2181 * If pid_alive fails, then pointers within the task structure
2182 * can be stale and must not be dereferenced.
2183 *
2184 * Return: 1 if the process is alive. 0 otherwise.
2185 */
2186static inline int pid_alive(const struct task_struct *p)
2187{
2188        return p->pids[PIDTYPE_PID].pid != NULL;
2189}
2190
2191/**
2192 * is_global_init - check if a task structure is init. Since init
2193 * is free to have sub-threads we need to check tgid.
2194 * @tsk: Task structure to be checked.
2195 *
2196 * Check if a task structure is the first user space task the kernel created.
2197 *
2198 * Return: 1 if the task structure is init. 0 otherwise.
2199 */
2200static inline int is_global_init(struct task_struct *tsk)
2201{
2202        return task_tgid_nr(tsk) == 1;
2203}
2204
2205extern struct pid *cad_pid;
2206
2207extern void free_task(struct task_struct *tsk);
2208#define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2209
2210extern void __put_task_struct(struct task_struct *t);
2211
2212static inline void put_task_struct(struct task_struct *t)
2213{
2214        if (atomic_dec_and_test(&t->usage))
2215                __put_task_struct(t);
2216}
2217
2218struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2219struct task_struct *try_get_task_struct(struct task_struct **ptask);
2220
2221#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2222extern void task_cputime(struct task_struct *t,
2223                         cputime_t *utime, cputime_t *stime);
2224extern void task_cputime_scaled(struct task_struct *t,
2225                                cputime_t *utimescaled, cputime_t *stimescaled);
2226extern cputime_t task_gtime(struct task_struct *t);
2227#else
2228static inline void task_cputime(struct task_struct *t,
2229                                cputime_t *utime, cputime_t *stime)
2230{
2231        if (utime)
2232                *utime = t->utime;
2233        if (stime)
2234                *stime = t->stime;
2235}
2236
2237static inline void task_cputime_scaled(struct task_struct *t,
2238                                       cputime_t *utimescaled,
2239                                       cputime_t *stimescaled)
2240{
2241        if (utimescaled)
2242                *utimescaled = t->utimescaled;
2243        if (stimescaled)
2244                *stimescaled = t->stimescaled;
2245}
2246
2247static inline cputime_t task_gtime(struct task_struct *t)
2248{
2249        return t->gtime;
2250}
2251#endif
2252extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2253extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2254
2255/*
2256 * Per process flags
2257 */
2258#define PF_EXITING      0x00000004      /* getting shut down */
2259#define PF_EXITPIDONE   0x00000008      /* pi exit done on shut down */
2260#define PF_VCPU         0x00000010      /* I'm a virtual CPU */
2261#define PF_WQ_WORKER    0x00000020      /* I'm a workqueue worker */
2262#define PF_FORKNOEXEC   0x00000040      /* forked but didn't exec */
2263#define PF_MCE_PROCESS  0x00000080      /* process policy on mce errors */
2264#define PF_SUPERPRIV    0x00000100      /* used super-user privileges */
2265#define PF_DUMPCORE     0x00000200      /* dumped core */
2266#define PF_SIGNALED     0x00000400      /* killed by a signal */
2267#define PF_MEMALLOC     0x00000800      /* Allocating memory */
2268#define PF_NPROC_EXCEEDED 0x00001000    /* set_user noticed that RLIMIT_NPROC was exceeded */
2269#define PF_USED_MATH    0x00002000      /* if unset the fpu must be initialized before use */
2270#define PF_USED_ASYNC   0x00004000      /* used async_schedule*(), used by module init */
2271#define PF_NOFREEZE     0x00008000      /* this thread should not be frozen */
2272#define PF_FROZEN       0x00010000      /* frozen for system suspend */
2273#define PF_FSTRANS      0x00020000      /* inside a filesystem transaction */
2274#define PF_KSWAPD       0x00040000      /* I am kswapd */
2275#define PF_MEMALLOC_NOIO 0x00080000     /* Allocating memory without IO involved */
2276#define PF_LESS_THROTTLE 0x00100000     /* Throttle me less: I clean memory */
2277#define PF_KTHREAD      0x00200000      /* I am a kernel thread */
2278#define PF_RANDOMIZE    0x00400000      /* randomize virtual address space */
2279#define PF_SWAPWRITE    0x00800000      /* Allowed to write to swap */
2280#define PF_NO_SETAFFINITY 0x04000000    /* Userland is not allowed to meddle with cpus_allowed */
2281#define PF_MCE_EARLY    0x08000000      /* Early kill for mce process policy */
2282#define PF_MUTEX_TESTER 0x20000000      /* Thread belongs to the rt mutex tester */
2283#define PF_FREEZER_SKIP 0x40000000      /* Freezer should not count it as freezable */
2284#define PF_SUSPEND_TASK 0x80000000      /* this thread called freeze_processes and should not be frozen */
2285
2286/*
2287 * Only the _current_ task can read/write to tsk->flags, but other
2288 * tasks can access tsk->flags in readonly mode for example
2289 * with tsk_used_math (like during threaded core dumping).
2290 * There is however an exception to this rule during ptrace
2291 * or during fork: the ptracer task is allowed to write to the
2292 * child->flags of its traced child (same goes for fork, the parent
2293 * can write to the child->flags), because we're guaranteed the
2294 * child is not running and in turn not changing child->flags
2295 * at the same time the parent does it.
2296 */
2297#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2298#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2299#define clear_used_math() clear_stopped_child_used_math(current)
2300#define set_used_math() set_stopped_child_used_math(current)
2301#define conditional_stopped_child_used_math(condition, child) \
2302        do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2303#define conditional_used_math(condition) \
2304        conditional_stopped_child_used_math(condition, current)
2305#define copy_to_stopped_child_used_math(child) \
2306        do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2307/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2308#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2309#define used_math() tsk_used_math(current)
2310
2311/* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2312 * __GFP_FS is also cleared as it implies __GFP_IO.
2313 */
2314static inline gfp_t memalloc_noio_flags(gfp_t flags)
2315{
2316        if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2317                flags &= ~(__GFP_IO | __GFP_FS);
2318        return flags;
2319}
2320
2321static inline unsigned int memalloc_noio_save(void)
2322{
2323        unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2324        current->flags |= PF_MEMALLOC_NOIO;
2325        return flags;
2326}
2327
2328static inline void memalloc_noio_restore(unsigned int flags)
2329{
2330        current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2331}
2332
2333/* Per-process atomic flags. */
2334#define PFA_NO_NEW_PRIVS 0      /* May not gain new privileges. */
2335#define PFA_SPREAD_PAGE  1      /* Spread page cache over cpuset */
2336#define PFA_SPREAD_SLAB  2      /* Spread some slab caches over cpuset */
2337#define PFA_LMK_WAITING  3      /* Lowmemorykiller is waiting */
2338
2339
2340#define TASK_PFA_TEST(name, func)                                       \
2341        static inline bool task_##func(struct task_struct *p)           \
2342        { return test_bit(PFA_##name, &p->atomic_flags); }
2343#define TASK_PFA_SET(name, func)                                        \
2344        static inline void task_set_##func(struct task_struct *p)       \
2345        { set_bit(PFA_##name, &p->atomic_flags); }
2346#define TASK_PFA_CLEAR(name, func)                                      \
2347        static inline void task_clear_##func(struct task_struct *p)     \
2348        { clear_bit(PFA_##name, &p->atomic_flags); }
2349
2350TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2351TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2352
2353TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2354TASK_PFA_SET(SPREAD_PAGE, spread_page)
2355TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2356
2357TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2358TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2359TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2360
2361TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2362TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2363
2364/*
2365 * task->jobctl flags
2366 */
2367#define JOBCTL_STOP_SIGMASK     0xffff  /* signr of the last group stop */
2368
2369#define JOBCTL_STOP_DEQUEUED_BIT 16     /* stop signal dequeued */
2370#define JOBCTL_STOP_PENDING_BIT 17      /* task should stop for group stop */
2371#define JOBCTL_STOP_CONSUME_BIT 18      /* consume group stop count */
2372#define JOBCTL_TRAP_STOP_BIT    19      /* trap for STOP */
2373#define JOBCTL_TRAP_NOTIFY_BIT  20      /* trap for NOTIFY */
2374#define JOBCTL_TRAPPING_BIT     21      /* switching to TRACED */
2375#define JOBCTL_LISTENING_BIT    22      /* ptracer is listening for events */
2376
2377#define JOBCTL_STOP_DEQUEUED    (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2378#define JOBCTL_STOP_PENDING     (1UL << JOBCTL_STOP_PENDING_BIT)
2379#define JOBCTL_STOP_CONSUME     (1UL << JOBCTL_STOP_CONSUME_BIT)
2380#define JOBCTL_TRAP_STOP        (1UL << JOBCTL_TRAP_STOP_BIT)
2381#define JOBCTL_TRAP_NOTIFY      (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2382#define JOBCTL_TRAPPING         (1UL << JOBCTL_TRAPPING_BIT)
2383#define JOBCTL_LISTENING        (1UL << JOBCTL_LISTENING_BIT)
2384
2385#define JOBCTL_TRAP_MASK        (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2386#define JOBCTL_PENDING_MASK     (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2387
2388extern bool task_set_jobctl_pending(struct task_struct *task,
2389                                    unsigned long mask);
2390extern void task_clear_jobctl_trapping(struct task_struct *task);
2391extern void task_clear_jobctl_pending(struct task_struct *task,
2392                                      unsigned long mask);
2393
2394static inline void rcu_copy_process(struct task_struct *p)
2395{
2396#ifdef CONFIG_PREEMPT_RCU
2397        p->rcu_read_lock_nesting = 0;
2398        p->rcu_read_unlock_special.s = 0;
2399        p->rcu_blocked_node = NULL;
2400        INIT_LIST_HEAD(&p->rcu_node_entry);
2401#endif /* #ifdef CONFIG_PREEMPT_RCU */
2402#ifdef CONFIG_TASKS_RCU
2403        p->rcu_tasks_holdout = false;
2404        INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2405        p->rcu_tasks_idle_cpu = -1;
2406#endif /* #ifdef CONFIG_TASKS_RCU */
2407}
2408
2409static inline void tsk_restore_flags(struct task_struct *task,
2410                                unsigned long orig_flags, unsigned long flags)
2411{
2412        task->flags &= ~flags;
2413        task->flags |= orig_flags & flags;
2414}
2415
2416extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2417                                     const struct cpumask *trial);
2418extern int task_can_attach(struct task_struct *p,
2419                           const struct cpumask *cs_cpus_allowed);
2420#ifdef CONFIG_SMP
2421extern void do_set_cpus_allowed(struct task_struct *p,
2422                               const struct cpumask *new_mask);
2423
2424extern int set_cpus_allowed_ptr(struct task_struct *p,
2425                                const struct cpumask *new_mask);
2426#else
2427static inline void do_set_cpus_allowed(struct task_struct *p,
2428                                      const struct cpumask *new_mask)
2429{
2430}
2431static inline int set_cpus_allowed_ptr(struct task_struct *p,
2432                                       const struct cpumask *new_mask)
2433{
2434        if (!cpumask_test_cpu(0, new_mask))
2435                return -EINVAL;
2436        return 0;
2437}
2438#endif
2439
2440#ifdef CONFIG_NO_HZ_COMMON
2441void calc_load_enter_idle(void);
2442void calc_load_exit_idle(void);
2443#else
2444static inline void calc_load_enter_idle(void) { }
2445static inline void calc_load_exit_idle(void) { }
2446#endif /* CONFIG_NO_HZ_COMMON */
2447
2448/*
2449 * Do not use outside of architecture code which knows its limitations.
2450 *
2451 * sched_clock() has no promise of monotonicity or bounded drift between
2452 * CPUs, use (which you should not) requires disabling IRQs.
2453 *
2454 * Please use one of the three interfaces below.
2455 */
2456extern unsigned long long notrace sched_clock(void);
2457/*
2458 * See the comment in kernel/sched/clock.c
2459 */
2460extern u64 running_clock(void);
2461extern u64 sched_clock_cpu(int cpu);
2462
2463
2464extern void sched_clock_init(void);
2465
2466#ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2467static inline void sched_clock_tick(void)
2468{
2469}
2470
2471static inline void sched_clock_idle_sleep_event(void)
2472{
2473}
2474
2475static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2476{
2477}
2478
2479static inline u64 cpu_clock(int cpu)
2480{
2481        return sched_clock();
2482}
2483
2484static inline u64 local_clock(void)
2485{
2486        return sched_clock();
2487}
2488#else
2489/*
2490 * Architectures can set this to 1 if they have specified
2491 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2492 * but then during bootup it turns out that sched_clock()
2493 * is reliable after all:
2494 */
2495extern int sched_clock_stable(void);
2496extern void set_sched_clock_stable(void);
2497extern void clear_sched_clock_stable(void);
2498
2499extern void sched_clock_tick(void);
2500extern void sched_clock_idle_sleep_event(void);
2501extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2502
2503/*
2504 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2505 * time source that is monotonic per cpu argument and has bounded drift
2506 * between cpus.
2507 *
2508 * ######################### BIG FAT WARNING ##########################
2509 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2510 * # go backwards !!                                                  #
2511 * ####################################################################
2512 */
2513static inline u64 cpu_clock(int cpu)
2514{
2515        return sched_clock_cpu(cpu);
2516}
2517
2518static inline u64 local_clock(void)
2519{
2520        return sched_clock_cpu(raw_smp_processor_id());
2521}
2522#endif
2523
2524#ifdef CONFIG_IRQ_TIME_ACCOUNTING
2525/*
2526 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2527 * The reason for this explicit opt-in is not to have perf penalty with
2528 * slow sched_clocks.
2529 */
2530extern void enable_sched_clock_irqtime(void);
2531extern void disable_sched_clock_irqtime(void);
2532#else
2533static inline void enable_sched_clock_irqtime(void) {}
2534static inline void disable_sched_clock_irqtime(void) {}
2535#endif
2536
2537extern unsigned long long
2538task_sched_runtime(struct task_struct *task);
2539
2540/* sched_exec is called by processes performing an exec */
2541#ifdef CONFIG_SMP
2542extern void sched_exec(void);
2543#else
2544#define sched_exec()   {}
2545#endif
2546
2547extern void sched_clock_idle_sleep_event(void);
2548extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2549
2550#ifdef CONFIG_HOTPLUG_CPU
2551extern void idle_task_exit(void);
2552#else
2553static inline void idle_task_exit(void) {}
2554#endif
2555
2556#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2557extern void wake_up_nohz_cpu(int cpu);
2558#else
2559static inline void wake_up_nohz_cpu(int cpu) { }
2560#endif
2561
2562#ifdef CONFIG_NO_HZ_FULL
2563extern u64 scheduler_tick_max_deferment(void);
2564#endif
2565
2566#ifdef CONFIG_SCHED_AUTOGROUP
2567extern void sched_autogroup_create_attach(struct task_struct *p);
2568extern void sched_autogroup_detach(struct task_struct *p);
2569extern void sched_autogroup_fork(struct signal_struct *sig);
2570extern void sched_autogroup_exit(struct signal_struct *sig);
2571extern void sched_autogroup_exit_task(struct task_struct *p);
2572#ifdef CONFIG_PROC_FS
2573extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2574extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2575#endif
2576#else
2577static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2578static inline void sched_autogroup_detach(struct task_struct *p) { }
2579static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2580static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2581static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2582#endif
2583
2584extern int yield_to(struct task_struct *p, bool preempt);
2585extern void set_user_nice(struct task_struct *p, long nice);
2586extern int task_prio(const struct task_struct *p);
2587/**
2588 * task_nice - return the nice value of a given task.
2589 * @p: the task in question.
2590 *
2591 * Return: The nice value [ -20 ... 0 ... 19 ].
2592 */
2593static inline int task_nice(const struct task_struct *p)
2594{
2595        return PRIO_TO_NICE((p)->static_prio);
2596}
2597extern int can_nice(const struct task_struct *p, const int nice);
2598extern int task_curr(const struct task_struct *p);
2599extern int idle_cpu(int cpu);
2600extern int sched_setscheduler(struct task_struct *, int,
2601                              const struct sched_param *);
2602extern int sched_setscheduler_nocheck(struct task_struct *, int,
2603                                      const struct sched_param *);
2604extern int sched_setattr(struct task_struct *,
2605                         const struct sched_attr *);
2606extern struct task_struct *idle_task(int cpu);
2607/**
2608 * is_idle_task - is the specified task an idle task?
2609 * @p: the task in question.
2610 *
2611 * Return: 1 if @p is an idle task. 0 otherwise.
2612 */
2613static inline bool is_idle_task(const struct task_struct *p)
2614{
2615        return p->pid == 0;
2616}
2617extern struct task_struct *curr_task(int cpu);
2618extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2619
2620void yield(void);
2621
2622union thread_union {
2623#ifndef CONFIG_THREAD_INFO_IN_TASK
2624        struct thread_info thread_info;
2625#endif
2626        unsigned long stack[THREAD_SIZE/sizeof(long)];
2627};
2628
2629#ifndef __HAVE_ARCH_KSTACK_END
2630static inline int kstack_end(void *addr)
2631{
2632        /* Reliable end of stack detection:
2633         * Some APM bios versions misalign the stack
2634         */
2635        return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2636}
2637#endif
2638
2639extern union thread_union init_thread_union;
2640extern struct task_struct init_task;
2641
2642extern struct   mm_struct init_mm;
2643
2644extern struct pid_namespace init_pid_ns;
2645
2646/*
2647 * find a task by one of its numerical ids
2648 *
2649 * find_task_by_pid_ns():
2650 *      finds a task by its pid in the specified namespace
2651 * find_task_by_vpid():
2652 *      finds a task by its virtual pid
2653 *
2654 * see also find_vpid() etc in include/linux/pid.h
2655 */
2656
2657extern struct task_struct *find_task_by_vpid(pid_t nr);
2658extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2659                struct pid_namespace *ns);
2660
2661/* per-UID process charging. */
2662extern struct user_struct * alloc_uid(kuid_t);
2663static inline struct user_struct *get_uid(struct user_struct *u)
2664{
2665        atomic_inc(&u->__count);
2666        return u;
2667}
2668extern void free_uid(struct user_struct *);
2669
2670#include <asm/current.h>
2671
2672extern void xtime_update(unsigned long ticks);
2673
2674extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2675extern int wake_up_process(struct task_struct *tsk);
2676extern void wake_up_new_task(struct task_struct *tsk);
2677#ifdef CONFIG_SMP
2678 extern void kick_process(struct task_struct *tsk);
2679#else
2680 static inline void kick_process(struct task_struct *tsk) { }
2681#endif
2682extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2683extern void sched_dead(struct task_struct *p);
2684
2685extern void proc_caches_init(void);
2686extern void flush_signals(struct task_struct *);
2687extern void ignore_signals(struct task_struct *);
2688extern void flush_signal_handlers(struct task_struct *, int force_default);
2689extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2690
2691static inline int kernel_dequeue_signal(siginfo_t *info)
2692{
2693        struct task_struct *tsk = current;
2694        siginfo_t __info;
2695        int ret;
2696
2697        spin_lock_irq(&tsk->sighand->siglock);
2698        ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2699        spin_unlock_irq(&tsk->sighand->siglock);
2700
2701        return ret;
2702}
2703
2704static inline void kernel_signal_stop(void)
2705{
2706        spin_lock_irq(&current->sighand->siglock);
2707        if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2708                __set_current_state(TASK_STOPPED);
2709        spin_unlock_irq(&current->sighand->siglock);
2710
2711        schedule();
2712}
2713
2714extern void release_task(struct task_struct * p);
2715extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2716extern int force_sigsegv(int, struct task_struct *);
2717extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2718extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2719extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2720extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2721                                const struct cred *, u32);
2722extern int kill_pgrp(struct pid *pid, int sig, int priv);
2723extern int kill_pid(struct pid *pid, int sig, int priv);
2724extern int kill_proc_info(int, struct siginfo *, pid_t);
2725extern __must_check bool do_notify_parent(struct task_struct *, int);
2726extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2727extern void force_sig(int, struct task_struct *);
2728extern int send_sig(int, struct task_struct *, int);
2729extern int zap_other_threads(struct task_struct *p);
2730extern struct sigqueue *sigqueue_alloc(void);
2731extern void sigqueue_free(struct sigqueue *);
2732extern int send_sigqueue(struct sigqueue *,  struct task_struct *, int group);
2733extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2734
2735#ifdef TIF_RESTORE_SIGMASK
2736/*
2737 * Legacy restore_sigmask accessors.  These are inefficient on
2738 * SMP architectures because they require atomic operations.
2739 */
2740
2741/**
2742 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2743 *
2744 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2745 * will run before returning to user mode, to process the flag.  For
2746 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2747 * it.  TIF_RESTORE_SIGMASK need not be in the set of bits that the
2748 * arch code will notice on return to user mode, in case those bits
2749 * are scarce.  We set TIF_SIGPENDING here to ensure that the arch
2750 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2751 */
2752static inline void set_restore_sigmask(void)
2753{
2754        set_thread_flag(TIF_RESTORE_SIGMASK);
2755        WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2756}
2757static inline void clear_restore_sigmask(void)
2758{
2759        clear_thread_flag(TIF_RESTORE_SIGMASK);
2760}
2761static inline bool test_restore_sigmask(void)
2762{
2763        return test_thread_flag(TIF_RESTORE_SIGMASK);
2764}
2765static inline bool test_and_clear_restore_sigmask(void)
2766{
2767        return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2768}
2769
2770#else   /* TIF_RESTORE_SIGMASK */
2771
2772/* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2773static inline void set_restore_sigmask(void)
2774{
2775        current->restore_sigmask = true;
2776        WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2777}
2778static inline void clear_restore_sigmask(void)
2779{
2780        current->restore_sigmask = false;
2781}
2782static inline bool test_restore_sigmask(void)
2783{
2784        return current->restore_sigmask;
2785}
2786static inline bool test_and_clear_restore_sigmask(void)
2787{
2788        if (!current->restore_sigmask)
2789                return false;
2790        current->restore_sigmask = false;
2791        return true;
2792}
2793#endif
2794
2795static inline void restore_saved_sigmask(void)
2796{
2797        if (test_and_clear_restore_sigmask())
2798                __set_current_blocked(&current->saved_sigmask);
2799}
2800
2801static inline sigset_t *sigmask_to_save(void)
2802{
2803        sigset_t *res = &current->blocked;
2804        if (unlikely(test_restore_sigmask()))
2805                res = &current->saved_sigmask;
2806        return res;
2807}
2808
2809static inline int kill_cad_pid(int sig, int priv)
2810{
2811        return kill_pid(cad_pid, sig, priv);
2812}
2813
2814/* These can be the second arg to send_sig_info/send_group_sig_info.  */
2815#define SEND_SIG_NOINFO ((struct siginfo *) 0)
2816#define SEND_SIG_PRIV   ((struct siginfo *) 1)
2817#define SEND_SIG_FORCED ((struct siginfo *) 2)
2818
2819/*
2820 * True if we are on the alternate signal stack.
2821 */
2822static inline int on_sig_stack(unsigned long sp)
2823{
2824        /*
2825         * If the signal stack is SS_AUTODISARM then, by construction, we
2826         * can't be on the signal stack unless user code deliberately set
2827         * SS_AUTODISARM when we were already on it.
2828         *
2829         * This improves reliability: if user state gets corrupted such that
2830         * the stack pointer points very close to the end of the signal stack,
2831         * then this check will enable the signal to be handled anyway.
2832         */
2833        if (current->sas_ss_flags & SS_AUTODISARM)
2834                return 0;
2835
2836#ifdef CONFIG_STACK_GROWSUP
2837        return sp >= current->sas_ss_sp &&
2838                sp - current->sas_ss_sp < current->sas_ss_size;
2839#else
2840        return sp > current->sas_ss_sp &&
2841                sp - current->sas_ss_sp <= current->sas_ss_size;
2842#endif
2843}
2844
2845static inline int sas_ss_flags(unsigned long sp)
2846{
2847        if (!current->sas_ss_size)
2848                return SS_DISABLE;
2849
2850        return on_sig_stack(sp) ? SS_ONSTACK : 0;
2851}
2852
2853static inline void sas_ss_reset(struct task_struct *p)
2854{
2855        p->sas_ss_sp = 0;
2856        p->sas_ss_size = 0;
2857        p->sas_ss_flags = SS_DISABLE;
2858}
2859
2860static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2861{
2862        if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2863#ifdef CONFIG_STACK_GROWSUP
2864                return current->sas_ss_sp;
2865#else
2866                return current->sas_ss_sp + current->sas_ss_size;
2867#endif
2868        return sp;
2869}
2870
2871/*
2872 * Routines for handling mm_structs
2873 */
2874extern struct mm_struct * mm_alloc(void);
2875
2876/* mmdrop drops the mm and the page tables */
2877extern void __mmdrop(struct mm_struct *);
2878static inline void mmdrop(struct mm_struct *mm)
2879{
2880        if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2881                __mmdrop(mm);
2882}
2883
2884static inline void mmdrop_async_fn(struct work_struct *work)
2885{
2886        struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2887        __mmdrop(mm);
2888}
2889
2890static inline void mmdrop_async(struct mm_struct *mm)
2891{
2892        if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2893                INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2894                schedule_work(&mm->async_put_work);
2895        }
2896}
2897
2898static inline bool mmget_not_zero(struct mm_struct *mm)
2899{
2900        return atomic_inc_not_zero(&mm->mm_users);
2901}
2902
2903/* mmput gets rid of the mappings and all user-space */
2904extern void mmput(struct mm_struct *);
2905#ifdef CONFIG_MMU
2906/* same as above but performs the slow path from the async context. Can
2907 * be called from the atomic context as well
2908 */
2909extern void mmput_async(struct mm_struct *);
2910#endif
2911
2912/* Grab a reference to a task's mm, if it is not already going away */
2913extern struct mm_struct *get_task_mm(struct task_struct *task);
2914/*
2915 * Grab a reference to a task's mm, if it is not already going away
2916 * and ptrace_may_access with the mode parameter passed to it
2917 * succeeds.
2918 */
2919extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2920/* Remove the current tasks stale references to the old mm_struct */
2921extern void mm_release(struct task_struct *, struct mm_struct *);
2922
2923#ifdef CONFIG_HAVE_COPY_THREAD_TLS
2924extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2925                        struct task_struct *, unsigned long);
2926#else
2927extern int copy_thread(unsigned long, unsigned long, unsigned long,
2928                        struct task_struct *);
2929
2930/* Architectures that haven't opted into copy_thread_tls get the tls argument
2931 * via pt_regs, so ignore the tls argument passed via C. */
2932static inline int copy_thread_tls(
2933                unsigned long clone_flags, unsigned long sp, unsigned long arg,
2934                struct task_struct *p, unsigned long tls)
2935{
2936        return copy_thread(clone_flags, sp, arg, p);
2937}
2938#endif
2939extern void flush_thread(void);
2940
2941#ifdef CONFIG_HAVE_EXIT_THREAD
2942extern void exit_thread(struct task_struct *tsk);
2943#else
2944static inline void exit_thread(struct task_struct *tsk)
2945{
2946}
2947#endif
2948
2949extern void exit_files(struct task_struct *);
2950extern void __cleanup_sighand(struct sighand_struct *);
2951
2952extern void exit_itimers(struct signal_struct *);
2953extern void flush_itimer_signals(void);
2954
2955extern void do_group_exit(int);
2956
2957extern int do_execve(struct filename *,
2958                     const char __user * const __user *,
2959                     const char __user * const __user *);
2960extern int do_execveat(int, struct filename *,
2961                       const char __user * const __user *,
2962                       const char __user * const __user *,
2963                       int);
2964extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
2965extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2966struct task_struct *fork_idle(int);
2967extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2968
2969extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2970static inline void set_task_comm(struct task_struct *tsk, const char *from)
2971{
2972        __set_task_comm(tsk, from, false);
2973}
2974extern char *get_task_comm(char *to, struct task_struct *tsk);
2975
2976#ifdef CONFIG_SMP
2977void scheduler_ipi(void);
2978extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2979#else
2980static inline void scheduler_ipi(void) { }
2981static inline unsigned long wait_task_inactive(struct task_struct *p,
2982                                               long match_state)
2983{
2984        return 1;
2985}
2986#endif
2987
2988#define tasklist_empty() \
2989        list_empty(&init_task.tasks)
2990
2991#define next_task(p) \
2992        list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2993
2994#define for_each_process(p) \
2995        for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2996
2997extern bool current_is_single_threaded(void);
2998
2999/*
3000 * Careful: do_each_thread/while_each_thread is a double loop so
3001 *          'break' will not work as expected - use goto instead.
3002 */
3003#define do_each_thread(g, t) \
3004        for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3005
3006#define while_each_thread(g, t) \
3007        while ((t = next_thread(t)) != g)
3008
3009#define __for_each_thread(signal, t)    \
3010        list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3011
3012#define for_each_thread(p, t)           \
3013        __for_each_thread((p)->signal, t)
3014
3015/* Careful: this is a double loop, 'break' won't work as expected. */
3016#define for_each_process_thread(p, t)   \
3017        for_each_process(p) for_each_thread(p, t)
3018
3019static inline int get_nr_threads(struct task_struct *tsk)
3020{
3021        return tsk->signal->nr_threads;
3022}
3023
3024static inline bool thread_group_leader(struct task_struct *p)
3025{
3026        return p->exit_signal >= 0;
3027}
3028
3029/* Do to the insanities of de_thread it is possible for a process
3030 * to have the pid of the thread group leader without actually being
3031 * the thread group leader.  For iteration through the pids in proc
3032 * all we care about is that we have a task with the appropriate
3033 * pid, we don't actually care if we have the right task.
3034 */
3035static inline bool has_group_leader_pid(struct task_struct *p)
3036{
3037        return task_pid(p) == p->signal->leader_pid;
3038}
3039
3040static inline
3041bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3042{
3043        return p1->signal == p2->signal;
3044}
3045
3046static inline struct task_struct *next_thread(const struct task_struct *p)
3047{
3048        return list_entry_rcu(p->thread_group.next,
3049                              struct task_struct, thread_group);
3050}
3051
3052static inline int thread_group_empty(struct task_struct *p)
3053{
3054        return list_empty(&p->thread_group);
3055}
3056
3057#define delay_group_leader(p) \
3058                (thread_group_leader(p) && !thread_group_empty(p))
3059
3060/*
3061 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3062 * subscriptions and synchronises with wait4().  Also used in procfs.  Also
3063 * pins the final release of task.io_context.  Also protects ->cpuset and
3064 * ->cgroup.subsys[]. And ->vfork_done.
3065 *
3066 * Nests both inside and outside of read_lock(&tasklist_lock).
3067 * It must not be nested with write_lock_irq(&tasklist_lock),
3068 * neither inside nor outside.
3069 */
3070static inline void task_lock(struct task_struct *p)
3071{
3072        spin_lock(&p->alloc_lock);
3073}
3074
3075static inline void task_unlock(struct task_struct *p)
3076{
3077        spin_unlock(&p->alloc_lock);
3078}
3079
3080extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3081                                                        unsigned long *flags);
3082
3083static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3084                                                       unsigned long *flags)
3085{
3086        struct sighand_struct *ret;
3087
3088        ret = __lock_task_sighand(tsk, flags);
3089        (void)__cond_lock(&tsk->sighand->siglock, ret);
3090        return ret;
3091}
3092
3093static inline void unlock_task_sighand(struct task_struct *tsk,
3094                                                unsigned long *flags)
3095{
3096        spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3097}
3098
3099/**
3100 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3101 * @tsk: task causing the changes
3102 *
3103 * All operations which modify a threadgroup - a new thread joining the
3104 * group, death of a member thread (the assertion of PF_EXITING) and
3105 * exec(2) dethreading the process and replacing the leader - are wrapped
3106 * by threadgroup_change_{begin|end}().  This is to provide a place which
3107 * subsystems needing threadgroup stability can hook into for
3108 * synchronization.
3109 */
3110static inline void threadgroup_change_begin(struct task_struct *tsk)
3111{
3112        might_sleep();
3113        cgroup_threadgroup_change_begin(tsk);
3114}
3115
3116/**
3117 * threadgroup_change_end - mark the end of changes to a threadgroup
3118 * @tsk: task causing the changes
3119 *
3120 * See threadgroup_change_begin().
3121 */
3122static inline void threadgroup_change_end(struct task_struct *tsk)
3123{
3124        cgroup_threadgroup_change_end(tsk);
3125}
3126
3127#ifdef CONFIG_THREAD_INFO_IN_TASK
3128
3129static inline struct thread_info *task_thread_info(struct task_struct *task)
3130{
3131        return &task->thread_info;
3132}
3133
3134/*
3135 * When accessing the stack of a non-current task that might exit, use
3136 * try_get_task_stack() instead.  task_stack_page will return a pointer
3137 * that could get freed out from under you.
3138 */
3139static inline void *task_stack_page(const struct task_struct *task)
3140{
3141        return task->stack;
3142}
3143
3144#define setup_thread_stack(new,old)     do { } while(0)
3145
3146static inline unsigned long *end_of_stack(const struct task_struct *task)
3147{
3148        return task->stack;
3149}
3150
3151#elif !defined(__HAVE_THREAD_FUNCTIONS)
3152
3153#define task_thread_info(task)  ((struct thread_info *)(task)->stack)
3154#define task_stack_page(task)   ((void *)(task)->stack)
3155
3156static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3157{
3158        *task_thread_info(p) = *task_thread_info(org);
3159        task_thread_info(p)->task = p;
3160}
3161
3162/*
3163 * Return the address of the last usable long on the stack.
3164 *
3165 * When the stack grows down, this is just above the thread
3166 * info struct. Going any lower will corrupt the threadinfo.
3167 *
3168 * When the stack grows up, this is the highest address.
3169 * Beyond that position, we corrupt data on the next page.
3170 */
3171static inline unsigned long *end_of_stack(struct task_struct *p)
3172{
3173#ifdef CONFIG_STACK_GROWSUP
3174        return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3175#else
3176        return (unsigned long *)(task_thread_info(p) + 1);
3177#endif
3178}
3179
3180#endif
3181
3182#ifdef CONFIG_THREAD_INFO_IN_TASK
3183static inline void *try_get_task_stack(struct task_struct *tsk)
3184{
3185        return atomic_inc_not_zero(&tsk->stack_refcount) ?
3186                task_stack_page(tsk) : NULL;
3187}
3188
3189extern void put_task_stack(struct task_struct *tsk);
3190#else
3191static inline void *try_get_task_stack(struct task_struct *tsk)
3192{
3193        return task_stack_page(tsk);
3194}
3195
3196static inline void put_task_stack(struct task_struct *tsk) {}
3197#endif
3198
3199#define task_stack_end_corrupted(task) \
3200                (*(end_of_stack(task)) != STACK_END_MAGIC)
3201
3202static inline int object_is_on_stack(void *obj)
3203{
3204        void *stack = task_stack_page(current);
3205
3206        return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3207}
3208
3209extern void thread_stack_cache_init(void);
3210
3211#ifdef CONFIG_DEBUG_STACK_USAGE
3212static inline unsigned long stack_not_used(struct task_struct *p)
3213{
3214        unsigned long *n = end_of_stack(p);
3215
3216        do {    /* Skip over canary */
3217# ifdef CONFIG_STACK_GROWSUP
3218                n--;
3219# else
3220                n++;
3221# endif
3222        } while (!*n);
3223
3224# ifdef CONFIG_STACK_GROWSUP
3225        return (unsigned long)end_of_stack(p) - (unsigned long)n;
3226# else
3227        return (unsigned long)n - (unsigned long)end_of_stack(p);
3228# endif
3229}
3230#endif
3231extern void set_task_stack_end_magic(struct task_struct *tsk);
3232
3233/* set thread flags in other task's structures
3234 * - see asm/thread_info.h for TIF_xxxx flags available
3235 */
3236static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3237{
3238        set_ti_thread_flag(task_thread_info(tsk), flag);
3239}
3240
3241static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3242{
3243        clear_ti_thread_flag(task_thread_info(tsk), flag);
3244}
3245
3246static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3247{
3248        return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3249}
3250
3251static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3252{
3253        return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3254}
3255
3256static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3257{
3258        return test_ti_thread_flag(task_thread_info(tsk), flag);
3259}
3260
3261static inline void set_tsk_need_resched(struct task_struct *tsk)
3262{
3263        set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3264}
3265
3266static inline void clear_tsk_need_resched(struct task_struct *tsk)
3267{
3268        clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3269}
3270
3271static inline int test_tsk_need_resched(struct task_struct *tsk)
3272{
3273        return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3274}
3275
3276static inline int restart_syscall(void)
3277{
3278        set_tsk_thread_flag(current, TIF_SIGPENDING);
3279        return -ERESTARTNOINTR;
3280}
3281
3282static inline int signal_pending(struct task_struct *p)
3283{
3284        return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3285}
3286
3287static inline int __fatal_signal_pending(struct task_struct *p)
3288{
3289        return unlikely(sigismember(&p->pending.signal, SIGKILL));
3290}
3291
3292static inline int fatal_signal_pending(struct task_struct *p)
3293{
3294        return signal_pending(p) && __fatal_signal_pending(p);
3295}
3296
3297static inline int signal_pending_state(long state, struct task_struct *p)
3298{
3299        if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3300                return 0;
3301        if (!signal_pending(p))
3302                return 0;
3303
3304        return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3305}
3306
3307/*
3308 * cond_resched() and cond_resched_lock(): latency reduction via
3309 * explicit rescheduling in places that are safe. The return
3310 * value indicates whether a reschedule was done in fact.
3311 * cond_resched_lock() will drop the spinlock before scheduling,
3312 * cond_resched_softirq() will enable bhs before scheduling.
3313 */
3314#ifndef CONFIG_PREEMPT
3315extern int _cond_resched(void);
3316#else
3317static inline int _cond_resched(void) { return 0; }
3318#endif
3319
3320#define cond_resched() ({                       \
3321        ___might_sleep(__FILE__, __LINE__, 0);  \
3322        _cond_resched();                        \
3323})
3324
3325extern int __cond_resched_lock(spinlock_t *lock);
3326
3327#define cond_resched_lock(lock) ({                              \
3328        ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3329        __cond_resched_lock(lock);                              \
3330})
3331
3332extern int __cond_resched_softirq(void);
3333
3334#define cond_resched_softirq() ({                                       \
3335        ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET);     \
3336        __cond_resched_softirq();                                       \
3337})
3338
3339static inline void cond_resched_rcu(void)
3340{
3341#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3342        rcu_read_unlock();
3343        cond_resched();
3344        rcu_read_lock();
3345#endif
3346}
3347
3348static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3349{
3350#ifdef CONFIG_DEBUG_PREEMPT
3351        return p->preempt_disable_ip;
3352#else
3353        return 0;
3354#endif
3355}
3356
3357/*
3358 * Does a critical section need to be broken due to another
3359 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3360 * but a general need for low latency)
3361 */
3362static inline int spin_needbreak(spinlock_t *lock)
3363{
3364#ifdef CONFIG_PREEMPT
3365        return spin_is_contended(lock);
3366#else
3367        return 0;
3368#endif
3369}
3370
3371/*
3372 * Idle thread specific functions to determine the need_resched
3373 * polling state.
3374 */
3375#ifdef TIF_POLLING_NRFLAG
3376static inline int tsk_is_polling(struct task_struct *p)
3377{
3378        return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3379}
3380
3381static inline void __current_set_polling(void)
3382{
3383        set_thread_flag(TIF_POLLING_NRFLAG);
3384}
3385
3386static inline bool __must_check current_set_polling_and_test(void)
3387{
3388        __current_set_polling();
3389
3390        /*
3391         * Polling state must be visible before we test NEED_RESCHED,
3392         * paired by resched_curr()
3393         */
3394        smp_mb__after_atomic();
3395
3396        return unlikely(tif_need_resched());
3397}
3398
3399static inline void __current_clr_polling(void)
3400{
3401        clear_thread_flag(TIF_POLLING_NRFLAG);
3402}
3403
3404static inline bool __must_check current_clr_polling_and_test(void)
3405{
3406        __current_clr_polling();
3407
3408        /*
3409         * Polling state must be visible before we test NEED_RESCHED,
3410         * paired by resched_curr()
3411         */
3412        smp_mb__after_atomic();
3413
3414        return unlikely(tif_need_resched());
3415}
3416
3417#else
3418static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3419static inline void __current_set_polling(void) { }
3420static inline void __current_clr_polling(void) { }
3421
3422static inline bool __must_check current_set_polling_and_test(void)
3423{
3424        return unlikely(tif_need_resched());
3425}
3426static inline bool __must_check current_clr_polling_and_test(void)
3427{
3428        return unlikely(tif_need_resched());
3429}
3430#endif
3431
3432static inline void current_clr_polling(void)
3433{
3434        __current_clr_polling();
3435
3436        /*
3437         * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3438         * Once the bit is cleared, we'll get IPIs with every new
3439         * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3440         * fold.
3441         */
3442        smp_mb(); /* paired with resched_curr() */
3443
3444        preempt_fold_need_resched();
3445}
3446
3447static __always_inline bool need_resched(void)
3448{
3449        return unlikely(tif_need_resched());
3450}
3451
3452/*
3453 * Thread group CPU time accounting.
3454 */
3455void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3456void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3457
3458/*
3459 * Reevaluate whether the task has signals pending delivery.
3460 * Wake the task if so.
3461 * This is required every time the blocked sigset_t changes.
3462 * callers must hold sighand->siglock.
3463 */
3464extern void recalc_sigpending_and_wake(struct task_struct *t);
3465extern void recalc_sigpending(void);
3466
3467extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3468
3469static inline void signal_wake_up(struct task_struct *t, bool resume)
3470{
3471        signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3472}
3473static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3474{
3475        signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3476}
3477
3478/*
3479 * Wrappers for p->thread_info->cpu access. No-op on UP.
3480 */
3481#ifdef CONFIG_SMP
3482
3483static inline unsigned int task_cpu(const struct task_struct *p)
3484{
3485#ifdef CONFIG_THREAD_INFO_IN_TASK
3486        return p->cpu;
3487#else
3488        return task_thread_info(p)->cpu;
3489#endif
3490}
3491
3492static inline int task_node(const struct task_struct *p)
3493{
3494        return cpu_to_node(task_cpu(p));
3495}
3496
3497extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3498
3499#else
3500
3501static inline unsigned int task_cpu(const struct task_struct *p)
3502{
3503        return 0;
3504}
3505
3506static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3507{
3508}
3509
3510#endif /* CONFIG_SMP */
3511
3512extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3513extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3514
3515#ifdef CONFIG_CGROUP_SCHED
3516extern struct task_group root_task_group;
3517#endif /* CONFIG_CGROUP_SCHED */
3518
3519extern int task_can_switch_user(struct user_struct *up,
3520                                        struct task_struct *tsk);
3521
3522#ifdef CONFIG_TASK_XACCT
3523static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3524{
3525        tsk->ioac.rchar += amt;
3526}
3527
3528static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3529{
3530        tsk->ioac.wchar += amt;
3531}
3532
3533static inline void inc_syscr(struct task_struct *tsk)
3534{
3535        tsk->ioac.syscr++;
3536}
3537
3538static inline void inc_syscw(struct task_struct *tsk)
3539{
3540        tsk->ioac.syscw++;
3541}
3542#else
3543static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3544{
3545}
3546
3547static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3548{
3549}
3550
3551static inline void inc_syscr(struct task_struct *tsk)
3552{
3553}
3554
3555static inline void inc_syscw(struct task_struct *tsk)
3556{
3557}
3558#endif
3559
3560#ifndef TASK_SIZE_OF
3561#define TASK_SIZE_OF(tsk)       TASK_SIZE
3562#endif
3563
3564#ifdef CONFIG_MEMCG
3565extern void mm_update_next_owner(struct mm_struct *mm);
3566#else
3567static inline void mm_update_next_owner(struct mm_struct *mm)
3568{
3569}
3570#endif /* CONFIG_MEMCG */
3571
3572static inline unsigned long task_rlimit(const struct task_struct *tsk,
3573                unsigned int limit)
3574{
3575        return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3576}
3577
3578static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3579                unsigned int limit)
3580{
3581        return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3582}
3583
3584static inline unsigned long rlimit(unsigned int limit)
3585{
3586        return task_rlimit(current, limit);
3587}
3588
3589static inline unsigned long rlimit_max(unsigned int limit)
3590{
3591        return task_rlimit_max(current, limit);
3592}
3593
3594#define SCHED_CPUFREQ_RT        (1U << 0)
3595#define SCHED_CPUFREQ_DL        (1U << 1)
3596#define SCHED_CPUFREQ_IOWAIT    (1U << 2)
3597
3598#define SCHED_CPUFREQ_RT_DL     (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3599
3600#ifdef CONFIG_CPU_FREQ
3601struct update_util_data {
3602       void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3603};
3604
3605void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3606                       void (*func)(struct update_util_data *data, u64 time,
3607                                    unsigned int flags));
3608void cpufreq_remove_update_util_hook(int cpu);
3609#endif /* CONFIG_CPU_FREQ */
3610
3611#endif
3612
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.