1 /* SPDX-License-Identifier: GPL-2.0 */
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
10 #include <uapi/linux/sched.h>
12 #include <asm/current.h>
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/mutex.h>
18 #include <linux/plist.h>
19 #include <linux/hrtimer.h>
20 #include <linux/irqflags.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/refcount.h>
25 #include <linux/resource.h>
26 #include <linux/latencytop.h>
27 #include <linux/sched/prio.h>
28 #include <linux/sched/types.h>
29 #include <linux/signal_types.h>
30 #include <linux/mm_types_task.h>
31 #include <linux/task_io_accounting.h>
32 #include <linux/posix-timers.h>
33 #include <linux/rseq.h>
34 #include <linux/seqlock.h>
35 #include <linux/kcsan.h>
37 /* task_struct member predeclarations (sorted alphabetically): */
39 struct backing_dev_info;
42 struct capture_control;
45 struct futex_pi_state;
50 struct perf_event_context;
52 struct pipe_inode_info;
55 struct robust_list_head;
61 struct sighand_struct;
63 struct task_delay_info;
68 * Task state bitmask. NOTE! These bits are also
69 * encoded in fs/proc/array.c: get_task_state().
71 * We have two separate sets of flags: task->state
72 * is about runnability, while task->exit_state are
73 * about the task exiting. Confusing, but this way
74 * modifying one set can't modify the other one by
78 /* Used in tsk->state: */
79 #define TASK_RUNNING 0x0000
80 #define TASK_INTERRUPTIBLE 0x0001
81 #define TASK_UNINTERRUPTIBLE 0x0002
82 #define __TASK_STOPPED 0x0004
83 #define __TASK_TRACED 0x0008
84 /* Used in tsk->exit_state: */
85 #define EXIT_DEAD 0x0010
86 #define EXIT_ZOMBIE 0x0020
87 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
88 /* Used in tsk->state again: */
89 #define TASK_PARKED 0x0040
90 #define TASK_DEAD 0x0080
91 #define TASK_WAKEKILL 0x0100
92 #define TASK_WAKING 0x0200
93 #define TASK_NOLOAD 0x0400
94 #define TASK_NEW 0x0800
95 #define TASK_STATE_MAX 0x1000
97 /* Convenience macros for the sake of set_current_state: */
98 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
99 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
100 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
102 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
104 /* Convenience macros for the sake of wake_up(): */
105 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
107 /* get_task_state(): */
108 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
109 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
110 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
113 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
115 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
117 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
119 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
122 * Special states are those that do not use the normal wait-loop pattern. See
123 * the comment with set_special_state().
125 #define is_special_task_state(state) \
126 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
128 #define __set_current_state(state_value) \
130 WARN_ON_ONCE(is_special_task_state(state_value));\
131 current->task_state_change = _THIS_IP_; \
132 current->state = (state_value); \
135 #define set_current_state(state_value) \
137 WARN_ON_ONCE(is_special_task_state(state_value));\
138 current->task_state_change = _THIS_IP_; \
139 smp_store_mb(current->state, (state_value)); \
142 #define set_special_state(state_value) \
144 unsigned long flags; /* may shadow */ \
145 WARN_ON_ONCE(!is_special_task_state(state_value)); \
146 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
147 current->task_state_change = _THIS_IP_; \
148 current->state = (state_value); \
149 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
153 * set_current_state() includes a barrier so that the write of current->state
154 * is correctly serialised wrt the caller's subsequent test of whether to
158 * set_current_state(TASK_UNINTERRUPTIBLE);
164 * __set_current_state(TASK_RUNNING);
166 * If the caller does not need such serialisation (because, for instance, the
167 * CONDITION test and condition change and wakeup are under the same lock) then
168 * use __set_current_state().
170 * The above is typically ordered against the wakeup, which does:
173 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
175 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
176 * accessing p->state.
178 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
179 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
180 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
182 * However, with slightly different timing the wakeup TASK_RUNNING store can
183 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
184 * a problem either because that will result in one extra go around the loop
185 * and our @cond test will save the day.
187 * Also see the comments of try_to_wake_up().
189 #define __set_current_state(state_value) \
190 current->state = (state_value)
192 #define set_current_state(state_value) \
193 smp_store_mb(current->state, (state_value))
196 * set_special_state() should be used for those states when the blocking task
197 * can not use the regular condition based wait-loop. In that case we must
198 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
199 * will not collide with our state change.
201 #define set_special_state(state_value) \
203 unsigned long flags; /* may shadow */ \
204 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
205 current->state = (state_value); \
206 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
211 /* Task command name length: */
212 #define TASK_COMM_LEN 16
214 extern void scheduler_tick(void);
216 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
218 extern long schedule_timeout(long timeout);
219 extern long schedule_timeout_interruptible(long timeout);
220 extern long schedule_timeout_killable(long timeout);
221 extern long schedule_timeout_uninterruptible(long timeout);
222 extern long schedule_timeout_idle(long timeout);
223 asmlinkage void schedule(void);
224 extern void schedule_preempt_disabled(void);
225 asmlinkage void preempt_schedule_irq(void);
227 extern int __must_check io_schedule_prepare(void);
228 extern void io_schedule_finish(int token);
229 extern long io_schedule_timeout(long timeout);
230 extern void io_schedule(void);
233 * struct prev_cputime - snapshot of system and user cputime
234 * @utime: time spent in user mode
235 * @stime: time spent in system mode
236 * @lock: protects the above two fields
238 * Stores previous user/system time values such that we can guarantee
241 struct prev_cputime {
242 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
250 /* Task is sleeping or running in a CPU with VTIME inactive: */
254 /* Task runs in kernelspace in a CPU with VTIME active: */
256 /* Task runs in userspace in a CPU with VTIME active: */
258 /* Task runs as guests in a CPU with VTIME active: */
264 unsigned long long starttime;
265 enum vtime_state state;
273 * Utilization clamp constraints.
274 * @UCLAMP_MIN: Minimum utilization
275 * @UCLAMP_MAX: Maximum utilization
276 * @UCLAMP_CNT: Utilization clamp constraints count
285 extern struct root_domain def_root_domain;
286 extern struct mutex sched_domains_mutex;
290 #ifdef CONFIG_SCHED_INFO
291 /* Cumulative counters: */
293 /* # of times we have run on this CPU: */
294 unsigned long pcount;
296 /* Time spent waiting on a runqueue: */
297 unsigned long long run_delay;
301 /* When did we last run on a CPU? */
302 unsigned long long last_arrival;
304 /* When were we last queued to run? */
305 unsigned long long last_queued;
307 #endif /* CONFIG_SCHED_INFO */
311 * Integer metrics need fixed point arithmetic, e.g., sched/fair
312 * has a few: load, load_avg, util_avg, freq, and capacity.
314 * We define a basic fixed point arithmetic range, and then formalize
315 * all these metrics based on that basic range.
317 # define SCHED_FIXEDPOINT_SHIFT 10
318 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
320 /* Increase resolution of cpu_capacity calculations */
321 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
322 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
325 unsigned long weight;
330 * struct util_est - Estimation utilization of FAIR tasks
331 * @enqueued: instantaneous estimated utilization of a task/cpu
332 * @ewma: the Exponential Weighted Moving Average (EWMA)
333 * utilization of a task
335 * Support data structure to track an Exponential Weighted Moving Average
336 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
337 * average each time a task completes an activation. Sample's weight is chosen
338 * so that the EWMA will be relatively insensitive to transient changes to the
341 * The enqueued attribute has a slightly different meaning for tasks and cpus:
342 * - task: the task's util_avg at last task dequeue time
343 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
344 * Thus, the util_est.enqueued of a task represents the contribution on the
345 * estimated utilization of the CPU where that task is currently enqueued.
347 * Only for tasks we track a moving average of the past instantaneous
348 * estimated utilization. This allows to absorb sporadic drops in utilization
349 * of an otherwise almost periodic task.
351 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
352 * updates. When a task is dequeued, its util_est should not be updated if its
353 * util_avg has not been updated in the meantime.
354 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
355 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
356 * for a task) it is safe to use MSB.
359 unsigned int enqueued;
361 #define UTIL_EST_WEIGHT_SHIFT 2
362 #define UTIL_AVG_UNCHANGED 0x80000000
363 } __attribute__((__aligned__(sizeof(u64))));
366 * The load/runnable/util_avg accumulates an infinite geometric series
367 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
369 * [load_avg definition]
371 * load_avg = runnable% * scale_load_down(load)
373 * [runnable_avg definition]
375 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
377 * [util_avg definition]
379 * util_avg = running% * SCHED_CAPACITY_SCALE
381 * where runnable% is the time ratio that a sched_entity is runnable and
382 * running% the time ratio that a sched_entity is running.
384 * For cfs_rq, they are the aggregated values of all runnable and blocked
387 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
388 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
389 * for computing those signals (see update_rq_clock_pelt())
391 * N.B., the above ratios (runnable% and running%) themselves are in the
392 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
393 * to as large a range as necessary. This is for example reflected by
394 * util_avg's SCHED_CAPACITY_SCALE.
398 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
399 * with the highest load (=88761), always runnable on a single cfs_rq,
400 * and should not overflow as the number already hits PID_MAX_LIMIT.
402 * For all other cases (including 32-bit kernels), struct load_weight's
403 * weight will overflow first before we do, because:
405 * Max(load_avg) <= Max(load.weight)
407 * Then it is the load_weight's responsibility to consider overflow
411 u64 last_update_time;
416 unsigned long load_avg;
417 unsigned long runnable_avg;
418 unsigned long util_avg;
419 struct util_est util_est;
420 } ____cacheline_aligned;
422 struct sched_statistics {
423 #ifdef CONFIG_SCHEDSTATS
433 s64 sum_sleep_runtime;
440 u64 nr_migrations_cold;
441 u64 nr_failed_migrations_affine;
442 u64 nr_failed_migrations_running;
443 u64 nr_failed_migrations_hot;
444 u64 nr_forced_migrations;
448 u64 nr_wakeups_migrate;
449 u64 nr_wakeups_local;
450 u64 nr_wakeups_remote;
451 u64 nr_wakeups_affine;
452 u64 nr_wakeups_affine_attempts;
453 u64 nr_wakeups_passive;
458 struct sched_entity {
459 /* For load-balancing: */
460 struct load_weight load;
461 struct rb_node run_node;
462 struct list_head group_node;
466 u64 sum_exec_runtime;
468 u64 prev_sum_exec_runtime;
472 struct sched_statistics statistics;
474 #ifdef CONFIG_FAIR_GROUP_SCHED
476 struct sched_entity *parent;
477 /* rq on which this entity is (to be) queued: */
478 struct cfs_rq *cfs_rq;
479 /* rq "owned" by this entity/group: */
481 /* cached value of my_q->h_nr_running */
482 unsigned long runnable_weight;
487 * Per entity load average tracking.
489 * Put into separate cache line so it does not
490 * collide with read-mostly values above.
492 struct sched_avg avg;
496 struct sched_rt_entity {
497 struct list_head run_list;
498 unsigned long timeout;
499 unsigned long watchdog_stamp;
500 unsigned int time_slice;
501 unsigned short on_rq;
502 unsigned short on_list;
504 struct sched_rt_entity *back;
505 #ifdef CONFIG_RT_GROUP_SCHED
506 struct sched_rt_entity *parent;
507 /* rq on which this entity is (to be) queued: */
509 /* rq "owned" by this entity/group: */
512 } __randomize_layout;
514 struct sched_dl_entity {
515 struct rb_node rb_node;
518 * Original scheduling parameters. Copied here from sched_attr
519 * during sched_setattr(), they will remain the same until
520 * the next sched_setattr().
522 u64 dl_runtime; /* Maximum runtime for each instance */
523 u64 dl_deadline; /* Relative deadline of each instance */
524 u64 dl_period; /* Separation of two instances (period) */
525 u64 dl_bw; /* dl_runtime / dl_period */
526 u64 dl_density; /* dl_runtime / dl_deadline */
529 * Actual scheduling parameters. Initialized with the values above,
530 * they are continuously updated during task execution. Note that
531 * the remaining runtime could be < 0 in case we are in overrun.
533 s64 runtime; /* Remaining runtime for this instance */
534 u64 deadline; /* Absolute deadline for this instance */
535 unsigned int flags; /* Specifying the scheduler behaviour */
540 * @dl_throttled tells if we exhausted the runtime. If so, the
541 * task has to wait for a replenishment to be performed at the
542 * next firing of dl_timer.
544 * @dl_yielded tells if task gave up the CPU before consuming
545 * all its available runtime during the last job.
547 * @dl_non_contending tells if the task is inactive while still
548 * contributing to the active utilization. In other words, it
549 * indicates if the inactive timer has been armed and its handler
550 * has not been executed yet. This flag is useful to avoid race
551 * conditions between the inactive timer handler and the wakeup
554 * @dl_overrun tells if the task asked to be informed about runtime
557 unsigned int dl_throttled : 1;
558 unsigned int dl_yielded : 1;
559 unsigned int dl_non_contending : 1;
560 unsigned int dl_overrun : 1;
563 * Bandwidth enforcement timer. Each -deadline task has its
564 * own bandwidth to be enforced, thus we need one timer per task.
566 struct hrtimer dl_timer;
569 * Inactive timer, responsible for decreasing the active utilization
570 * at the "0-lag time". When a -deadline task blocks, it contributes
571 * to GRUB's active utilization until the "0-lag time", hence a
572 * timer is needed to decrease the active utilization at the correct
575 struct hrtimer inactive_timer;
577 #ifdef CONFIG_RT_MUTEXES
579 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
580 * pi_se points to the donor, otherwise points to the dl_se it belongs
581 * to (the original one/itself).
583 struct sched_dl_entity *pi_se;
587 #ifdef CONFIG_UCLAMP_TASK
588 /* Number of utilization clamp buckets (shorter alias) */
589 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
592 * Utilization clamp for a scheduling entity
593 * @value: clamp value "assigned" to a se
594 * @bucket_id: bucket index corresponding to the "assigned" value
595 * @active: the se is currently refcounted in a rq's bucket
596 * @user_defined: the requested clamp value comes from user-space
598 * The bucket_id is the index of the clamp bucket matching the clamp value
599 * which is pre-computed and stored to avoid expensive integer divisions from
602 * The active bit is set whenever a task has got an "effective" value assigned,
603 * which can be different from the clamp value "requested" from user-space.
604 * This allows to know a task is refcounted in the rq's bucket corresponding
605 * to the "effective" bucket_id.
607 * The user_defined bit is set whenever a task has got a task-specific clamp
608 * value requested from userspace, i.e. the system defaults apply to this task
609 * just as a restriction. This allows to relax default clamps when a less
610 * restrictive task-specific value has been requested, thus allowing to
611 * implement a "nice" semantic. For example, a task running with a 20%
612 * default boost can still drop its own boosting to 0%.
615 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
616 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
617 unsigned int active : 1;
618 unsigned int user_defined : 1;
620 #endif /* CONFIG_UCLAMP_TASK */
626 u8 exp_hint; /* Hint for performance. */
627 u8 need_mb; /* Readers need smp_mb(). */
629 u32 s; /* Set of bits. */
632 enum perf_event_task_context {
633 perf_invalid_context = -1,
636 perf_nr_task_contexts,
640 struct wake_q_node *next;
644 #ifdef CONFIG_THREAD_INFO_IN_TASK
646 * For reasons of header soup (see current_thread_info()), this
647 * must be the first element of task_struct.
649 struct thread_info thread_info;
651 /* -1 unrunnable, 0 runnable, >0 stopped: */
655 * This begins the randomizable portion of task_struct. Only
656 * scheduling-critical items should be added above here.
658 randomized_struct_fields_start
662 /* Per task flags (PF_*), defined further below: */
668 struct __call_single_node wake_entry;
669 #ifdef CONFIG_THREAD_INFO_IN_TASK
673 unsigned int wakee_flips;
674 unsigned long wakee_flip_decay_ts;
675 struct task_struct *last_wakee;
678 * recent_used_cpu is initially set as the last CPU used by a task
679 * that wakes affine another task. Waker/wakee relationships can
680 * push tasks around a CPU where each wakeup moves to the next one.
681 * Tracking a recently used CPU allows a quick search for a recently
682 * used CPU that may be idle.
692 unsigned int rt_priority;
694 const struct sched_class *sched_class;
695 struct sched_entity se;
696 struct sched_rt_entity rt;
697 #ifdef CONFIG_CGROUP_SCHED
698 struct task_group *sched_task_group;
700 struct sched_dl_entity dl;
702 #ifdef CONFIG_UCLAMP_TASK
704 * Clamp values requested for a scheduling entity.
705 * Must be updated with task_rq_lock() held.
707 struct uclamp_se uclamp_req[UCLAMP_CNT];
709 * Effective clamp values used for a scheduling entity.
710 * Must be updated with task_rq_lock() held.
712 struct uclamp_se uclamp[UCLAMP_CNT];
715 #ifdef CONFIG_PREEMPT_NOTIFIERS
716 /* List of struct preempt_notifier: */
717 struct hlist_head preempt_notifiers;
720 #ifdef CONFIG_BLK_DEV_IO_TRACE
721 unsigned int btrace_seq;
726 const cpumask_t *cpus_ptr;
729 #ifdef CONFIG_PREEMPT_RCU
730 int rcu_read_lock_nesting;
731 union rcu_special rcu_read_unlock_special;
732 struct list_head rcu_node_entry;
733 struct rcu_node *rcu_blocked_node;
734 #endif /* #ifdef CONFIG_PREEMPT_RCU */
736 #ifdef CONFIG_TASKS_RCU
737 unsigned long rcu_tasks_nvcsw;
738 u8 rcu_tasks_holdout;
740 int rcu_tasks_idle_cpu;
741 struct list_head rcu_tasks_holdout_list;
742 #endif /* #ifdef CONFIG_TASKS_RCU */
744 #ifdef CONFIG_TASKS_TRACE_RCU
745 int trc_reader_nesting;
747 union rcu_special trc_reader_special;
748 bool trc_reader_checked;
749 struct list_head trc_holdout_list;
750 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
752 struct sched_info sched_info;
754 struct list_head tasks;
756 struct plist_node pushable_tasks;
757 struct rb_node pushable_dl_tasks;
760 struct mm_struct *mm;
761 struct mm_struct *active_mm;
763 /* Per-thread vma caching: */
764 struct vmacache vmacache;
766 #ifdef SPLIT_RSS_COUNTING
767 struct task_rss_stat rss_stat;
772 /* The signal sent when the parent dies: */
774 /* JOBCTL_*, siglock protected: */
775 unsigned long jobctl;
777 /* Used for emulating ABI behavior of previous Linux versions: */
778 unsigned int personality;
780 /* Scheduler bits, serialized by scheduler locks: */
781 unsigned sched_reset_on_fork:1;
782 unsigned sched_contributes_to_load:1;
783 unsigned sched_migrated:1;
785 unsigned sched_psi_wake_requeue:1;
788 /* Force alignment to the next boundary: */
791 /* Unserialized, strictly 'current' */
794 * This field must not be in the scheduler word above due to wakelist
795 * queueing no longer being serialized by p->on_cpu. However:
798 * schedule() if (p->on_rq && ..) // false
799 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
800 * deactivate_task() ttwu_queue_wakelist())
801 * p->on_rq = 0; p->sched_remote_wakeup = Y;
803 * guarantees all stores of 'current' are visible before
804 * ->sched_remote_wakeup gets used, so it can be in this word.
806 unsigned sched_remote_wakeup:1;
808 /* Bit to tell LSMs we're in execve(): */
809 unsigned in_execve:1;
810 unsigned in_iowait:1;
811 #ifndef TIF_RESTORE_SIGMASK
812 unsigned restore_sigmask:1;
815 unsigned in_user_fault:1;
817 #ifdef CONFIG_COMPAT_BRK
818 unsigned brk_randomized:1;
820 #ifdef CONFIG_CGROUPS
821 /* disallow userland-initiated cgroup migration */
822 unsigned no_cgroup_migration:1;
823 /* task is frozen/stopped (used by the cgroup freezer) */
826 #ifdef CONFIG_BLK_CGROUP
827 unsigned use_memdelay:1;
830 /* Stalled due to lack of memory */
831 unsigned in_memstall:1;
834 unsigned long atomic_flags; /* Flags requiring atomic access. */
836 struct restart_block restart_block;
841 #ifdef CONFIG_STACKPROTECTOR
842 /* Canary value for the -fstack-protector GCC feature: */
843 unsigned long stack_canary;
846 * Pointers to the (original) parent process, youngest child, younger sibling,
847 * older sibling, respectively. (p->father can be replaced with
848 * p->real_parent->pid)
851 /* Real parent process: */
852 struct task_struct __rcu *real_parent;
854 /* Recipient of SIGCHLD, wait4() reports: */
855 struct task_struct __rcu *parent;
858 * Children/sibling form the list of natural children:
860 struct list_head children;
861 struct list_head sibling;
862 struct task_struct *group_leader;
865 * 'ptraced' is the list of tasks this task is using ptrace() on.
867 * This includes both natural children and PTRACE_ATTACH targets.
868 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
870 struct list_head ptraced;
871 struct list_head ptrace_entry;
873 /* PID/PID hash table linkage. */
874 struct pid *thread_pid;
875 struct hlist_node pid_links[PIDTYPE_MAX];
876 struct list_head thread_group;
877 struct list_head thread_node;
879 struct completion *vfork_done;
881 /* CLONE_CHILD_SETTID: */
882 int __user *set_child_tid;
884 /* CLONE_CHILD_CLEARTID: */
885 int __user *clear_child_tid;
892 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
897 struct prev_cputime prev_cputime;
898 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
902 #ifdef CONFIG_NO_HZ_FULL
903 atomic_t tick_dep_mask;
905 /* Context switch counts: */
907 unsigned long nivcsw;
909 /* Monotonic time in nsecs: */
912 /* Boot based time in nsecs: */
915 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
916 unsigned long min_flt;
917 unsigned long maj_flt;
919 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
920 struct posix_cputimers posix_cputimers;
922 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
923 struct posix_cputimers_work posix_cputimers_work;
926 /* Process credentials: */
928 /* Tracer's credentials at attach: */
929 const struct cred __rcu *ptracer_cred;
931 /* Objective and real subjective task credentials (COW): */
932 const struct cred __rcu *real_cred;
934 /* Effective (overridable) subjective task credentials (COW): */
935 const struct cred __rcu *cred;
938 /* Cached requested key. */
939 struct key *cached_requested_key;
943 * executable name, excluding path.
945 * - normally initialized setup_new_exec()
946 * - access it with [gs]et_task_comm()
947 * - lock it with task_lock()
949 char comm[TASK_COMM_LEN];
951 struct nameidata *nameidata;
953 #ifdef CONFIG_SYSVIPC
954 struct sysv_sem sysvsem;
955 struct sysv_shm sysvshm;
957 #ifdef CONFIG_DETECT_HUNG_TASK
958 unsigned long last_switch_count;
959 unsigned long last_switch_time;
961 /* Filesystem information: */
962 struct fs_struct *fs;
964 /* Open file information: */
965 struct files_struct *files;
967 #ifdef CONFIG_IO_URING
968 struct io_uring_task *io_uring;
972 struct nsproxy *nsproxy;
974 /* Signal handlers: */
975 struct signal_struct *signal;
976 struct sighand_struct __rcu *sighand;
978 sigset_t real_blocked;
979 /* Restored if set_restore_sigmask() was used: */
980 sigset_t saved_sigmask;
981 struct sigpending pending;
982 unsigned long sas_ss_sp;
984 unsigned int sas_ss_flags;
986 struct callback_head *task_works;
989 #ifdef CONFIG_AUDITSYSCALL
990 struct audit_context *audit_context;
993 unsigned int sessionid;
995 struct seccomp seccomp;
997 /* Thread group tracking: */
1001 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1002 spinlock_t alloc_lock;
1004 /* Protection of the PI data structures: */
1005 raw_spinlock_t pi_lock;
1007 struct wake_q_node wake_q;
1009 #ifdef CONFIG_RT_MUTEXES
1010 /* PI waiters blocked on a rt_mutex held by this task: */
1011 struct rb_root_cached pi_waiters;
1012 /* Updated under owner's pi_lock and rq lock */
1013 struct task_struct *pi_top_task;
1014 /* Deadlock detection and priority inheritance handling: */
1015 struct rt_mutex_waiter *pi_blocked_on;
1018 #ifdef CONFIG_DEBUG_MUTEXES
1019 /* Mutex deadlock detection: */
1020 struct mutex_waiter *blocked_on;
1023 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1024 int non_block_count;
1027 #ifdef CONFIG_TRACE_IRQFLAGS
1028 struct irqtrace_events irqtrace;
1029 unsigned int hardirq_threaded;
1030 u64 hardirq_chain_key;
1031 int softirqs_enabled;
1032 int softirq_context;
1036 #ifdef CONFIG_LOCKDEP
1037 # define MAX_LOCK_DEPTH 48UL
1040 unsigned int lockdep_recursion;
1041 struct held_lock held_locks[MAX_LOCK_DEPTH];
1044 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1045 unsigned int in_ubsan;
1048 /* Journalling filesystem info: */
1051 /* Stacked block device info: */
1052 struct bio_list *bio_list;
1055 /* Stack plugging: */
1056 struct blk_plug *plug;
1060 struct reclaim_state *reclaim_state;
1062 struct backing_dev_info *backing_dev_info;
1064 struct io_context *io_context;
1066 #ifdef CONFIG_COMPACTION
1067 struct capture_control *capture_control;
1070 unsigned long ptrace_message;
1071 kernel_siginfo_t *last_siginfo;
1073 struct task_io_accounting ioac;
1075 /* Pressure stall state */
1076 unsigned int psi_flags;
1078 #ifdef CONFIG_TASK_XACCT
1079 /* Accumulated RSS usage: */
1081 /* Accumulated virtual memory usage: */
1083 /* stime + utime since last update: */
1086 #ifdef CONFIG_CPUSETS
1087 /* Protected by ->alloc_lock: */
1088 nodemask_t mems_allowed;
1089 /* Seqence number to catch updates: */
1090 seqcount_spinlock_t mems_allowed_seq;
1091 int cpuset_mem_spread_rotor;
1092 int cpuset_slab_spread_rotor;
1094 #ifdef CONFIG_CGROUPS
1095 /* Control Group info protected by css_set_lock: */
1096 struct css_set __rcu *cgroups;
1097 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1098 struct list_head cg_list;
1100 #ifdef CONFIG_X86_CPU_RESCTRL
1105 struct robust_list_head __user *robust_list;
1106 #ifdef CONFIG_COMPAT
1107 struct compat_robust_list_head __user *compat_robust_list;
1109 struct list_head pi_state_list;
1110 struct futex_pi_state *pi_state_cache;
1111 struct mutex futex_exit_mutex;
1112 unsigned int futex_state;
1114 #ifdef CONFIG_PERF_EVENTS
1115 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1116 struct mutex perf_event_mutex;
1117 struct list_head perf_event_list;
1119 #ifdef CONFIG_DEBUG_PREEMPT
1120 unsigned long preempt_disable_ip;
1123 /* Protected by alloc_lock: */
1124 struct mempolicy *mempolicy;
1126 short pref_node_fork;
1128 #ifdef CONFIG_NUMA_BALANCING
1130 unsigned int numa_scan_period;
1131 unsigned int numa_scan_period_max;
1132 int numa_preferred_nid;
1133 unsigned long numa_migrate_retry;
1134 /* Migration stamp: */
1136 u64 last_task_numa_placement;
1137 u64 last_sum_exec_runtime;
1138 struct callback_head numa_work;
1141 * This pointer is only modified for current in syscall and
1142 * pagefault context (and for tasks being destroyed), so it can be read
1143 * from any of the following contexts:
1144 * - RCU read-side critical section
1145 * - current->numa_group from everywhere
1146 * - task's runqueue locked, task not running
1148 struct numa_group __rcu *numa_group;
1151 * numa_faults is an array split into four regions:
1152 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1153 * in this precise order.
1155 * faults_memory: Exponential decaying average of faults on a per-node
1156 * basis. Scheduling placement decisions are made based on these
1157 * counts. The values remain static for the duration of a PTE scan.
1158 * faults_cpu: Track the nodes the process was running on when a NUMA
1159 * hinting fault was incurred.
1160 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1161 * during the current scan window. When the scan completes, the counts
1162 * in faults_memory and faults_cpu decay and these values are copied.
1164 unsigned long *numa_faults;
1165 unsigned long total_numa_faults;
1168 * numa_faults_locality tracks if faults recorded during the last
1169 * scan window were remote/local or failed to migrate. The task scan
1170 * period is adapted based on the locality of the faults with different
1171 * weights depending on whether they were shared or private faults
1173 unsigned long numa_faults_locality[3];
1175 unsigned long numa_pages_migrated;
1176 #endif /* CONFIG_NUMA_BALANCING */
1179 struct rseq __user *rseq;
1182 * RmW on rseq_event_mask must be performed atomically
1183 * with respect to preemption.
1185 unsigned long rseq_event_mask;
1188 struct tlbflush_unmap_batch tlb_ubc;
1191 refcount_t rcu_users;
1192 struct rcu_head rcu;
1195 /* Cache last used pipe for splice(): */
1196 struct pipe_inode_info *splice_pipe;
1198 struct page_frag task_frag;
1200 #ifdef CONFIG_TASK_DELAY_ACCT
1201 struct task_delay_info *delays;
1204 #ifdef CONFIG_FAULT_INJECTION
1206 unsigned int fail_nth;
1209 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1210 * balance_dirty_pages() for a dirty throttling pause:
1213 int nr_dirtied_pause;
1214 /* Start of a write-and-pause period: */
1215 unsigned long dirty_paused_when;
1217 #ifdef CONFIG_LATENCYTOP
1218 int latency_record_count;
1219 struct latency_record latency_record[LT_SAVECOUNT];
1222 * Time slack values; these are used to round up poll() and
1223 * select() etc timeout values. These are in nanoseconds.
1226 u64 default_timer_slack_ns;
1229 unsigned int kasan_depth;
1233 struct kcsan_ctx kcsan_ctx;
1234 #ifdef CONFIG_TRACE_IRQFLAGS
1235 struct irqtrace_events kcsan_save_irqtrace;
1239 #if IS_ENABLED(CONFIG_KUNIT)
1240 struct kunit *kunit_test;
1243 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1244 /* Index of current stored address in ret_stack: */
1248 /* Stack of return addresses for return function tracing: */
1249 struct ftrace_ret_stack *ret_stack;
1251 /* Timestamp for last schedule: */
1252 unsigned long long ftrace_timestamp;
1255 * Number of functions that haven't been traced
1256 * because of depth overrun:
1258 atomic_t trace_overrun;
1260 /* Pause tracing: */
1261 atomic_t tracing_graph_pause;
1264 #ifdef CONFIG_TRACING
1265 /* State flags for use by tracers: */
1266 unsigned long trace;
1268 /* Bitmask and counter of trace recursion: */
1269 unsigned long trace_recursion;
1270 #endif /* CONFIG_TRACING */
1273 /* See kernel/kcov.c for more details. */
1275 /* Coverage collection mode enabled for this task (0 if disabled): */
1276 unsigned int kcov_mode;
1278 /* Size of the kcov_area: */
1279 unsigned int kcov_size;
1281 /* Buffer for coverage collection: */
1284 /* KCOV descriptor wired with this task or NULL: */
1287 /* KCOV common handle for remote coverage collection: */
1290 /* KCOV sequence number: */
1293 /* Collect coverage from softirq context: */
1294 unsigned int kcov_softirq;
1298 struct mem_cgroup *memcg_in_oom;
1299 gfp_t memcg_oom_gfp_mask;
1300 int memcg_oom_order;
1302 /* Number of pages to reclaim on returning to userland: */
1303 unsigned int memcg_nr_pages_over_high;
1305 /* Used by memcontrol for targeted memcg charge: */
1306 struct mem_cgroup *active_memcg;
1309 #ifdef CONFIG_BLK_CGROUP
1310 struct request_queue *throttle_queue;
1313 #ifdef CONFIG_UPROBES
1314 struct uprobe_task *utask;
1316 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1317 unsigned int sequential_io;
1318 unsigned int sequential_io_avg;
1320 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1321 unsigned long task_state_change;
1323 int pagefault_disabled;
1325 struct task_struct *oom_reaper_list;
1326 struct timer_list oom_reaper_timer;
1328 #ifdef CONFIG_VMAP_STACK
1329 struct vm_struct *stack_vm_area;
1331 #ifdef CONFIG_THREAD_INFO_IN_TASK
1332 /* A live task holds one reference: */
1333 refcount_t stack_refcount;
1335 #ifdef CONFIG_LIVEPATCH
1338 #ifdef CONFIG_SECURITY
1339 /* Used by LSM modules for access restriction: */
1343 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1344 unsigned long lowest_stack;
1345 unsigned long prev_lowest_stack;
1348 #ifdef CONFIG_X86_MCE
1349 void __user *mce_vaddr;
1354 __mce_reserved : 62;
1355 struct callback_head mce_kill_me;
1360 * New fields for task_struct should be added above here, so that
1361 * they are included in the randomized portion of task_struct.
1363 randomized_struct_fields_end
1365 /* CPU-specific state of this task: */
1366 struct thread_struct thread;
1369 * WARNING: on x86, 'thread_struct' contains a variable-sized
1370 * structure. It *MUST* be at the end of 'task_struct'.
1372 * Do not put anything below here!
1376 static inline struct pid *task_pid(struct task_struct *task)
1378 return task->thread_pid;
1382 * the helpers to get the task's different pids as they are seen
1383 * from various namespaces
1385 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1386 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1388 * task_xid_nr_ns() : id seen from the ns specified;
1390 * see also pid_nr() etc in include/linux/pid.h
1392 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1394 static inline pid_t task_pid_nr(struct task_struct *tsk)
1399 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1401 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1404 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1406 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1410 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1416 * pid_alive - check that a task structure is not stale
1417 * @p: Task structure to be checked.
1419 * Test if a process is not yet dead (at most zombie state)
1420 * If pid_alive fails, then pointers within the task structure
1421 * can be stale and must not be dereferenced.
1423 * Return: 1 if the process is alive. 0 otherwise.
1425 static inline int pid_alive(const struct task_struct *p)
1427 return p->thread_pid != NULL;
1430 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1432 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1435 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1437 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1441 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1443 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1446 static inline pid_t task_session_vnr(struct task_struct *tsk)
1448 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1451 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1453 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1456 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1458 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1461 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1467 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1473 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1475 return task_ppid_nr_ns(tsk, &init_pid_ns);
1478 /* Obsolete, do not use: */
1479 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1481 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1484 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1485 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1487 static inline unsigned int task_state_index(struct task_struct *tsk)
1489 unsigned int tsk_state = READ_ONCE(tsk->state);
1490 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1492 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1494 if (tsk_state == TASK_IDLE)
1495 state = TASK_REPORT_IDLE;
1500 static inline char task_index_to_char(unsigned int state)
1502 static const char state_char[] = "RSDTtXZPI";
1504 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1506 return state_char[state];
1509 static inline char task_state_to_char(struct task_struct *tsk)
1511 return task_index_to_char(task_state_index(tsk));
1515 * is_global_init - check if a task structure is init. Since init
1516 * is free to have sub-threads we need to check tgid.
1517 * @tsk: Task structure to be checked.
1519 * Check if a task structure is the first user space task the kernel created.
1521 * Return: 1 if the task structure is init. 0 otherwise.
1523 static inline int is_global_init(struct task_struct *tsk)
1525 return task_tgid_nr(tsk) == 1;
1528 extern struct pid *cad_pid;
1533 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1534 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1535 #define PF_EXITING 0x00000004 /* Getting shut down */
1536 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1537 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1538 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1539 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1540 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1541 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1542 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1543 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1544 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1545 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1546 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1547 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1548 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1549 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1550 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1551 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1552 * I am cleaning dirty pages from some other bdi. */
1553 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1554 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1555 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1556 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1557 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1558 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1559 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1560 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1563 * Only the _current_ task can read/write to tsk->flags, but other
1564 * tasks can access tsk->flags in readonly mode for example
1565 * with tsk_used_math (like during threaded core dumping).
1566 * There is however an exception to this rule during ptrace
1567 * or during fork: the ptracer task is allowed to write to the
1568 * child->flags of its traced child (same goes for fork, the parent
1569 * can write to the child->flags), because we're guaranteed the
1570 * child is not running and in turn not changing child->flags
1571 * at the same time the parent does it.
1573 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1574 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1575 #define clear_used_math() clear_stopped_child_used_math(current)
1576 #define set_used_math() set_stopped_child_used_math(current)
1578 #define conditional_stopped_child_used_math(condition, child) \
1579 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1581 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1583 #define copy_to_stopped_child_used_math(child) \
1584 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1586 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1587 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1588 #define used_math() tsk_used_math(current)
1590 static __always_inline bool is_percpu_thread(void)
1593 return (current->flags & PF_NO_SETAFFINITY) &&
1594 (current->nr_cpus_allowed == 1);
1600 /* Per-process atomic flags. */
1601 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1602 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1603 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1604 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1605 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1606 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1607 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1608 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1610 #define TASK_PFA_TEST(name, func) \
1611 static inline bool task_##func(struct task_struct *p) \
1612 { return test_bit(PFA_##name, &p->atomic_flags); }
1614 #define TASK_PFA_SET(name, func) \
1615 static inline void task_set_##func(struct task_struct *p) \
1616 { set_bit(PFA_##name, &p->atomic_flags); }
1618 #define TASK_PFA_CLEAR(name, func) \
1619 static inline void task_clear_##func(struct task_struct *p) \
1620 { clear_bit(PFA_##name, &p->atomic_flags); }
1622 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1623 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1625 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1626 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1627 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1629 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1630 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1631 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1633 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1634 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1635 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1637 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1638 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1639 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1641 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1642 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1644 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1645 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1646 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1648 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1649 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1652 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1654 current->flags &= ~flags;
1655 current->flags |= orig_flags & flags;
1658 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1659 extern int task_can_attach(struct task_struct *p);
1660 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1661 extern void dl_bw_free(int cpu, u64 dl_bw);
1663 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1664 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1666 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1669 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1671 if (!cpumask_test_cpu(0, new_mask))
1677 extern int yield_to(struct task_struct *p, bool preempt);
1678 extern void set_user_nice(struct task_struct *p, long nice);
1679 extern int task_prio(const struct task_struct *p);
1682 * task_nice - return the nice value of a given task.
1683 * @p: the task in question.
1685 * Return: The nice value [ -20 ... 0 ... 19 ].
1687 static inline int task_nice(const struct task_struct *p)
1689 return PRIO_TO_NICE((p)->static_prio);
1692 extern int can_nice(const struct task_struct *p, const int nice);
1693 extern int task_curr(const struct task_struct *p);
1694 extern int idle_cpu(int cpu);
1695 extern int available_idle_cpu(int cpu);
1696 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1697 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1698 extern void sched_set_fifo(struct task_struct *p);
1699 extern void sched_set_fifo_low(struct task_struct *p);
1700 extern void sched_set_normal(struct task_struct *p, int nice);
1701 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1702 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1703 extern struct task_struct *idle_task(int cpu);
1706 * is_idle_task - is the specified task an idle task?
1707 * @p: the task in question.
1709 * Return: 1 if @p is an idle task. 0 otherwise.
1711 static __always_inline bool is_idle_task(const struct task_struct *p)
1713 return !!(p->flags & PF_IDLE);
1716 extern struct task_struct *curr_task(int cpu);
1717 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1721 union thread_union {
1722 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1723 struct task_struct task;
1725 #ifndef CONFIG_THREAD_INFO_IN_TASK
1726 struct thread_info thread_info;
1728 unsigned long stack[THREAD_SIZE/sizeof(long)];
1731 #ifndef CONFIG_THREAD_INFO_IN_TASK
1732 extern struct thread_info init_thread_info;
1735 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1737 #ifdef CONFIG_THREAD_INFO_IN_TASK
1738 static inline struct thread_info *task_thread_info(struct task_struct *task)
1740 return &task->thread_info;
1742 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1743 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1747 * find a task by one of its numerical ids
1749 * find_task_by_pid_ns():
1750 * finds a task by its pid in the specified namespace
1751 * find_task_by_vpid():
1752 * finds a task by its virtual pid
1754 * see also find_vpid() etc in include/linux/pid.h
1757 extern struct task_struct *find_task_by_vpid(pid_t nr);
1758 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1761 * find a task by its virtual pid and get the task struct
1763 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1765 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1766 extern int wake_up_process(struct task_struct *tsk);
1767 extern void wake_up_new_task(struct task_struct *tsk);
1770 extern void kick_process(struct task_struct *tsk);
1772 static inline void kick_process(struct task_struct *tsk) { }
1775 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1777 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1779 __set_task_comm(tsk, from, false);
1782 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1783 #define get_task_comm(buf, tsk) ({ \
1784 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1785 __get_task_comm(buf, sizeof(buf), tsk); \
1789 static __always_inline void scheduler_ipi(void)
1792 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1793 * TIF_NEED_RESCHED remotely (for the first time) will also send
1796 preempt_fold_need_resched();
1798 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1800 static inline void scheduler_ipi(void) { }
1801 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1808 * Set thread flags in other task's structures.
1809 * See asm/thread_info.h for TIF_xxxx flags available:
1811 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1813 set_ti_thread_flag(task_thread_info(tsk), flag);
1816 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1818 clear_ti_thread_flag(task_thread_info(tsk), flag);
1821 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1824 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1827 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1829 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1832 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1834 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1837 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1839 return test_ti_thread_flag(task_thread_info(tsk), flag);
1842 static inline void set_tsk_need_resched(struct task_struct *tsk)
1844 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1847 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1849 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1852 static inline int test_tsk_need_resched(struct task_struct *tsk)
1854 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1858 * cond_resched() and cond_resched_lock(): latency reduction via
1859 * explicit rescheduling in places that are safe. The return
1860 * value indicates whether a reschedule was done in fact.
1861 * cond_resched_lock() will drop the spinlock before scheduling,
1863 #ifndef CONFIG_PREEMPTION
1864 extern int _cond_resched(void);
1866 static inline int _cond_resched(void) { return 0; }
1869 #define cond_resched() ({ \
1870 ___might_sleep(__FILE__, __LINE__, 0); \
1874 extern int __cond_resched_lock(spinlock_t *lock);
1876 #define cond_resched_lock(lock) ({ \
1877 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1878 __cond_resched_lock(lock); \
1881 static inline void cond_resched_rcu(void)
1883 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1891 * Does a critical section need to be broken due to another
1892 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1893 * but a general need for low latency)
1895 static inline int spin_needbreak(spinlock_t *lock)
1897 #ifdef CONFIG_PREEMPTION
1898 return spin_is_contended(lock);
1904 static __always_inline bool need_resched(void)
1906 return unlikely(tif_need_resched());
1910 * Wrappers for p->thread_info->cpu access. No-op on UP.
1914 static inline unsigned int task_cpu(const struct task_struct *p)
1916 #ifdef CONFIG_THREAD_INFO_IN_TASK
1917 return READ_ONCE(p->cpu);
1919 return READ_ONCE(task_thread_info(p)->cpu);
1923 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1927 static inline unsigned int task_cpu(const struct task_struct *p)
1932 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1936 #endif /* CONFIG_SMP */
1939 * In order to reduce various lock holder preemption latencies provide an
1940 * interface to see if a vCPU is currently running or not.
1942 * This allows us to terminate optimistic spin loops and block, analogous to
1943 * the native optimistic spin heuristic of testing if the lock owner task is
1946 #ifndef vcpu_is_preempted
1947 static inline bool vcpu_is_preempted(int cpu)
1953 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1954 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1956 #ifndef TASK_SIZE_OF
1957 #define TASK_SIZE_OF(tsk) TASK_SIZE
1963 * Map the event mask on the user-space ABI enum rseq_cs_flags
1964 * for direct mask checks.
1966 enum rseq_event_mask_bits {
1967 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1968 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1969 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1972 enum rseq_event_mask {
1973 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1974 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1975 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1978 static inline void rseq_set_notify_resume(struct task_struct *t)
1981 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1984 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1986 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1987 struct pt_regs *regs)
1990 __rseq_handle_notify_resume(ksig, regs);
1993 static inline void rseq_signal_deliver(struct ksignal *ksig,
1994 struct pt_regs *regs)
1997 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1999 rseq_handle_notify_resume(ksig, regs);
2002 /* rseq_preempt() requires preemption to be disabled. */
2003 static inline void rseq_preempt(struct task_struct *t)
2005 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2006 rseq_set_notify_resume(t);
2009 /* rseq_migrate() requires preemption to be disabled. */
2010 static inline void rseq_migrate(struct task_struct *t)
2012 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2013 rseq_set_notify_resume(t);
2017 * If parent process has a registered restartable sequences area, the
2018 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2020 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2022 if (clone_flags & CLONE_VM) {
2025 t->rseq_event_mask = 0;
2027 t->rseq = current->rseq;
2028 t->rseq_sig = current->rseq_sig;
2029 t->rseq_event_mask = current->rseq_event_mask;
2033 static inline void rseq_execve(struct task_struct *t)
2037 t->rseq_event_mask = 0;
2042 static inline void rseq_set_notify_resume(struct task_struct *t)
2045 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2046 struct pt_regs *regs)
2049 static inline void rseq_signal_deliver(struct ksignal *ksig,
2050 struct pt_regs *regs)
2053 static inline void rseq_preempt(struct task_struct *t)
2056 static inline void rseq_migrate(struct task_struct *t)
2059 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2062 static inline void rseq_execve(struct task_struct *t)
2068 #ifdef CONFIG_DEBUG_RSEQ
2070 void rseq_syscall(struct pt_regs *regs);
2074 static inline void rseq_syscall(struct pt_regs *regs)
2080 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2081 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2082 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2084 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2085 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2086 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2088 int sched_trace_rq_cpu(struct rq *rq);
2089 int sched_trace_rq_cpu_capacity(struct rq *rq);
2090 int sched_trace_rq_nr_running(struct rq *rq);
2092 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);