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/syscall_user_dispatch.h>
31 #include <linux/mm_types_task.h>
32 #include <linux/task_io_accounting.h>
33 #include <linux/posix-timers.h>
34 #include <linux/rseq.h>
35 #include <linux/seqlock.h>
36 #include <linux/kcsan.h>
37 #include <asm/kmap_size.h>
39 /* task_struct member predeclarations (sorted alphabetically): */
41 struct backing_dev_info;
44 struct bpf_local_storage;
45 struct capture_control;
48 struct futex_pi_state;
54 struct perf_event_context;
56 struct pipe_inode_info;
59 struct robust_list_head;
65 struct sighand_struct;
67 struct task_delay_info;
71 * Task state bitmask. NOTE! These bits are also
72 * encoded in fs/proc/array.c: get_task_state().
74 * We have two separate sets of flags: task->state
75 * is about runnability, while task->exit_state are
76 * about the task exiting. Confusing, but this way
77 * modifying one set can't modify the other one by
81 /* Used in tsk->state: */
82 #define TASK_RUNNING 0x0000
83 #define TASK_INTERRUPTIBLE 0x0001
84 #define TASK_UNINTERRUPTIBLE 0x0002
85 #define __TASK_STOPPED 0x0004
86 #define __TASK_TRACED 0x0008
87 /* Used in tsk->exit_state: */
88 #define EXIT_DEAD 0x0010
89 #define EXIT_ZOMBIE 0x0020
90 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
91 /* Used in tsk->state again: */
92 #define TASK_PARKED 0x0040
93 #define TASK_DEAD 0x0080
94 #define TASK_WAKEKILL 0x0100
95 #define TASK_WAKING 0x0200
96 #define TASK_NOLOAD 0x0400
97 #define TASK_NEW 0x0800
98 /* RT specific auxilliary flag to mark RT lock waiters */
99 #define TASK_RTLOCK_WAIT 0x1000
100 #define TASK_STATE_MAX 0x2000
102 /* Convenience macros for the sake of set_current_state: */
103 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
104 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
105 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
107 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
109 /* Convenience macros for the sake of wake_up(): */
110 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
112 /* get_task_state(): */
113 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
114 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
115 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
118 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
120 #define task_is_traced(task) ((READ_ONCE(task->__state) & __TASK_TRACED) != 0)
122 #define task_is_stopped(task) ((READ_ONCE(task->__state) & __TASK_STOPPED) != 0)
124 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->__state) & (__TASK_STOPPED | __TASK_TRACED)) != 0)
127 * Special states are those that do not use the normal wait-loop pattern. See
128 * the comment with set_special_state().
130 #define is_special_task_state(state) \
131 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
133 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
134 # define debug_normal_state_change(state_value) \
136 WARN_ON_ONCE(is_special_task_state(state_value)); \
137 current->task_state_change = _THIS_IP_; \
140 # define debug_special_state_change(state_value) \
142 WARN_ON_ONCE(!is_special_task_state(state_value)); \
143 current->task_state_change = _THIS_IP_; \
146 # define debug_rtlock_wait_set_state() \
148 current->saved_state_change = current->task_state_change;\
149 current->task_state_change = _THIS_IP_; \
152 # define debug_rtlock_wait_restore_state() \
154 current->task_state_change = current->saved_state_change;\
158 # define debug_normal_state_change(cond) do { } while (0)
159 # define debug_special_state_change(cond) do { } while (0)
160 # define debug_rtlock_wait_set_state() do { } while (0)
161 # define debug_rtlock_wait_restore_state() do { } while (0)
165 * set_current_state() includes a barrier so that the write of current->state
166 * is correctly serialised wrt the caller's subsequent test of whether to
170 * set_current_state(TASK_UNINTERRUPTIBLE);
176 * __set_current_state(TASK_RUNNING);
178 * If the caller does not need such serialisation (because, for instance, the
179 * CONDITION test and condition change and wakeup are under the same lock) then
180 * use __set_current_state().
182 * The above is typically ordered against the wakeup, which does:
185 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
187 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
188 * accessing p->state.
190 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
191 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
192 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
194 * However, with slightly different timing the wakeup TASK_RUNNING store can
195 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
196 * a problem either because that will result in one extra go around the loop
197 * and our @cond test will save the day.
199 * Also see the comments of try_to_wake_up().
201 #define __set_current_state(state_value) \
203 debug_normal_state_change((state_value)); \
204 WRITE_ONCE(current->__state, (state_value)); \
207 #define set_current_state(state_value) \
209 debug_normal_state_change((state_value)); \
210 smp_store_mb(current->__state, (state_value)); \
214 * set_special_state() should be used for those states when the blocking task
215 * can not use the regular condition based wait-loop. In that case we must
216 * serialize against wakeups such that any possible in-flight TASK_RUNNING
217 * stores will not collide with our state change.
219 #define set_special_state(state_value) \
221 unsigned long flags; /* may shadow */ \
223 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
224 debug_special_state_change((state_value)); \
225 WRITE_ONCE(current->__state, (state_value)); \
226 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
230 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
232 * RT's spin/rwlock substitutions are state preserving. The state of the
233 * task when blocking on the lock is saved in task_struct::saved_state and
234 * restored after the lock has been acquired. These operations are
235 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
236 * lock related wakeups while the task is blocked on the lock are
237 * redirected to operate on task_struct::saved_state to ensure that these
238 * are not dropped. On restore task_struct::saved_state is set to
239 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
241 * The lock operation looks like this:
243 * current_save_and_set_rtlock_wait_state();
247 * raw_spin_unlock_irq(&lock->wait_lock);
249 * raw_spin_lock_irq(&lock->wait_lock);
250 * set_current_state(TASK_RTLOCK_WAIT);
252 * current_restore_rtlock_saved_state();
254 #define current_save_and_set_rtlock_wait_state() \
256 lockdep_assert_irqs_disabled(); \
257 raw_spin_lock(¤t->pi_lock); \
258 current->saved_state = current->__state; \
259 debug_rtlock_wait_set_state(); \
260 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
261 raw_spin_unlock(¤t->pi_lock); \
264 #define current_restore_rtlock_saved_state() \
266 lockdep_assert_irqs_disabled(); \
267 raw_spin_lock(¤t->pi_lock); \
268 debug_rtlock_wait_restore_state(); \
269 WRITE_ONCE(current->__state, current->saved_state); \
270 current->saved_state = TASK_RUNNING; \
271 raw_spin_unlock(¤t->pi_lock); \
274 #define get_current_state() READ_ONCE(current->__state)
276 /* Task command name length: */
277 #define TASK_COMM_LEN 16
279 extern void scheduler_tick(void);
281 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
283 extern long schedule_timeout(long timeout);
284 extern long schedule_timeout_interruptible(long timeout);
285 extern long schedule_timeout_killable(long timeout);
286 extern long schedule_timeout_uninterruptible(long timeout);
287 extern long schedule_timeout_idle(long timeout);
288 asmlinkage void schedule(void);
289 extern void schedule_preempt_disabled(void);
290 asmlinkage void preempt_schedule_irq(void);
291 #ifdef CONFIG_PREEMPT_RT
292 extern void schedule_rtlock(void);
295 extern int __must_check io_schedule_prepare(void);
296 extern void io_schedule_finish(int token);
297 extern long io_schedule_timeout(long timeout);
298 extern void io_schedule(void);
301 * struct prev_cputime - snapshot of system and user cputime
302 * @utime: time spent in user mode
303 * @stime: time spent in system mode
304 * @lock: protects the above two fields
306 * Stores previous user/system time values such that we can guarantee
309 struct prev_cputime {
310 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
318 /* Task is sleeping or running in a CPU with VTIME inactive: */
322 /* Task runs in kernelspace in a CPU with VTIME active: */
324 /* Task runs in userspace in a CPU with VTIME active: */
326 /* Task runs as guests in a CPU with VTIME active: */
332 unsigned long long starttime;
333 enum vtime_state state;
341 * Utilization clamp constraints.
342 * @UCLAMP_MIN: Minimum utilization
343 * @UCLAMP_MAX: Maximum utilization
344 * @UCLAMP_CNT: Utilization clamp constraints count
353 extern struct root_domain def_root_domain;
354 extern struct mutex sched_domains_mutex;
358 #ifdef CONFIG_SCHED_INFO
359 /* Cumulative counters: */
361 /* # of times we have run on this CPU: */
362 unsigned long pcount;
364 /* Time spent waiting on a runqueue: */
365 unsigned long long run_delay;
369 /* When did we last run on a CPU? */
370 unsigned long long last_arrival;
372 /* When were we last queued to run? */
373 unsigned long long last_queued;
375 #endif /* CONFIG_SCHED_INFO */
379 * Integer metrics need fixed point arithmetic, e.g., sched/fair
380 * has a few: load, load_avg, util_avg, freq, and capacity.
382 * We define a basic fixed point arithmetic range, and then formalize
383 * all these metrics based on that basic range.
385 # define SCHED_FIXEDPOINT_SHIFT 10
386 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
388 /* Increase resolution of cpu_capacity calculations */
389 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
390 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
393 unsigned long weight;
398 * struct util_est - Estimation utilization of FAIR tasks
399 * @enqueued: instantaneous estimated utilization of a task/cpu
400 * @ewma: the Exponential Weighted Moving Average (EWMA)
401 * utilization of a task
403 * Support data structure to track an Exponential Weighted Moving Average
404 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
405 * average each time a task completes an activation. Sample's weight is chosen
406 * so that the EWMA will be relatively insensitive to transient changes to the
409 * The enqueued attribute has a slightly different meaning for tasks and cpus:
410 * - task: the task's util_avg at last task dequeue time
411 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
412 * Thus, the util_est.enqueued of a task represents the contribution on the
413 * estimated utilization of the CPU where that task is currently enqueued.
415 * Only for tasks we track a moving average of the past instantaneous
416 * estimated utilization. This allows to absorb sporadic drops in utilization
417 * of an otherwise almost periodic task.
419 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
420 * updates. When a task is dequeued, its util_est should not be updated if its
421 * util_avg has not been updated in the meantime.
422 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
423 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
424 * for a task) it is safe to use MSB.
427 unsigned int enqueued;
429 #define UTIL_EST_WEIGHT_SHIFT 2
430 #define UTIL_AVG_UNCHANGED 0x80000000
431 } __attribute__((__aligned__(sizeof(u64))));
434 * The load/runnable/util_avg accumulates an infinite geometric series
435 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
437 * [load_avg definition]
439 * load_avg = runnable% * scale_load_down(load)
441 * [runnable_avg definition]
443 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
445 * [util_avg definition]
447 * util_avg = running% * SCHED_CAPACITY_SCALE
449 * where runnable% is the time ratio that a sched_entity is runnable and
450 * running% the time ratio that a sched_entity is running.
452 * For cfs_rq, they are the aggregated values of all runnable and blocked
455 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
456 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
457 * for computing those signals (see update_rq_clock_pelt())
459 * N.B., the above ratios (runnable% and running%) themselves are in the
460 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
461 * to as large a range as necessary. This is for example reflected by
462 * util_avg's SCHED_CAPACITY_SCALE.
466 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
467 * with the highest load (=88761), always runnable on a single cfs_rq,
468 * and should not overflow as the number already hits PID_MAX_LIMIT.
470 * For all other cases (including 32-bit kernels), struct load_weight's
471 * weight will overflow first before we do, because:
473 * Max(load_avg) <= Max(load.weight)
475 * Then it is the load_weight's responsibility to consider overflow
479 u64 last_update_time;
484 unsigned long load_avg;
485 unsigned long runnable_avg;
486 unsigned long util_avg;
487 struct util_est util_est;
488 } ____cacheline_aligned;
490 struct sched_statistics {
491 #ifdef CONFIG_SCHEDSTATS
501 s64 sum_sleep_runtime;
508 u64 nr_migrations_cold;
509 u64 nr_failed_migrations_affine;
510 u64 nr_failed_migrations_running;
511 u64 nr_failed_migrations_hot;
512 u64 nr_forced_migrations;
516 u64 nr_wakeups_migrate;
517 u64 nr_wakeups_local;
518 u64 nr_wakeups_remote;
519 u64 nr_wakeups_affine;
520 u64 nr_wakeups_affine_attempts;
521 u64 nr_wakeups_passive;
526 struct sched_entity {
527 /* For load-balancing: */
528 struct load_weight load;
529 struct rb_node run_node;
530 struct list_head group_node;
534 u64 sum_exec_runtime;
536 u64 prev_sum_exec_runtime;
540 struct sched_statistics statistics;
542 #ifdef CONFIG_FAIR_GROUP_SCHED
544 struct sched_entity *parent;
545 /* rq on which this entity is (to be) queued: */
546 struct cfs_rq *cfs_rq;
547 /* rq "owned" by this entity/group: */
549 /* cached value of my_q->h_nr_running */
550 unsigned long runnable_weight;
555 * Per entity load average tracking.
557 * Put into separate cache line so it does not
558 * collide with read-mostly values above.
560 struct sched_avg avg;
564 struct sched_rt_entity {
565 struct list_head run_list;
566 unsigned long timeout;
567 unsigned long watchdog_stamp;
568 unsigned int time_slice;
569 unsigned short on_rq;
570 unsigned short on_list;
572 struct sched_rt_entity *back;
573 #ifdef CONFIG_RT_GROUP_SCHED
574 struct sched_rt_entity *parent;
575 /* rq on which this entity is (to be) queued: */
577 /* rq "owned" by this entity/group: */
580 } __randomize_layout;
582 struct sched_dl_entity {
583 struct rb_node rb_node;
586 * Original scheduling parameters. Copied here from sched_attr
587 * during sched_setattr(), they will remain the same until
588 * the next sched_setattr().
590 u64 dl_runtime; /* Maximum runtime for each instance */
591 u64 dl_deadline; /* Relative deadline of each instance */
592 u64 dl_period; /* Separation of two instances (period) */
593 u64 dl_bw; /* dl_runtime / dl_period */
594 u64 dl_density; /* dl_runtime / dl_deadline */
597 * Actual scheduling parameters. Initialized with the values above,
598 * they are continuously updated during task execution. Note that
599 * the remaining runtime could be < 0 in case we are in overrun.
601 s64 runtime; /* Remaining runtime for this instance */
602 u64 deadline; /* Absolute deadline for this instance */
603 unsigned int flags; /* Specifying the scheduler behaviour */
608 * @dl_throttled tells if we exhausted the runtime. If so, the
609 * task has to wait for a replenishment to be performed at the
610 * next firing of dl_timer.
612 * @dl_boosted tells if we are boosted due to DI. If so we are
613 * outside bandwidth enforcement mechanism (but only until we
614 * exit the critical section);
616 * @dl_yielded tells if task gave up the CPU before consuming
617 * all its available runtime during the last job.
619 * @dl_non_contending tells if the task is inactive while still
620 * contributing to the active utilization. In other words, it
621 * indicates if the inactive timer has been armed and its handler
622 * has not been executed yet. This flag is useful to avoid race
623 * conditions between the inactive timer handler and the wakeup
626 * @dl_overrun tells if the task asked to be informed about runtime
629 unsigned int dl_throttled : 1;
630 unsigned int dl_yielded : 1;
631 unsigned int dl_non_contending : 1;
632 unsigned int dl_overrun : 1;
635 * Bandwidth enforcement timer. Each -deadline task has its
636 * own bandwidth to be enforced, thus we need one timer per task.
638 struct hrtimer dl_timer;
641 * Inactive timer, responsible for decreasing the active utilization
642 * at the "0-lag time". When a -deadline task blocks, it contributes
643 * to GRUB's active utilization until the "0-lag time", hence a
644 * timer is needed to decrease the active utilization at the correct
647 struct hrtimer inactive_timer;
649 #ifdef CONFIG_RT_MUTEXES
651 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
652 * pi_se points to the donor, otherwise points to the dl_se it belongs
653 * to (the original one/itself).
655 struct sched_dl_entity *pi_se;
659 #ifdef CONFIG_UCLAMP_TASK
660 /* Number of utilization clamp buckets (shorter alias) */
661 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
664 * Utilization clamp for a scheduling entity
665 * @value: clamp value "assigned" to a se
666 * @bucket_id: bucket index corresponding to the "assigned" value
667 * @active: the se is currently refcounted in a rq's bucket
668 * @user_defined: the requested clamp value comes from user-space
670 * The bucket_id is the index of the clamp bucket matching the clamp value
671 * which is pre-computed and stored to avoid expensive integer divisions from
674 * The active bit is set whenever a task has got an "effective" value assigned,
675 * which can be different from the clamp value "requested" from user-space.
676 * This allows to know a task is refcounted in the rq's bucket corresponding
677 * to the "effective" bucket_id.
679 * The user_defined bit is set whenever a task has got a task-specific clamp
680 * value requested from userspace, i.e. the system defaults apply to this task
681 * just as a restriction. This allows to relax default clamps when a less
682 * restrictive task-specific value has been requested, thus allowing to
683 * implement a "nice" semantic. For example, a task running with a 20%
684 * default boost can still drop its own boosting to 0%.
687 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
688 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
689 unsigned int active : 1;
690 unsigned int user_defined : 1;
692 #endif /* CONFIG_UCLAMP_TASK */
698 u8 exp_hint; /* Hint for performance. */
699 u8 need_mb; /* Readers need smp_mb(). */
701 u32 s; /* Set of bits. */
704 enum perf_event_task_context {
705 perf_invalid_context = -1,
708 perf_nr_task_contexts,
712 struct wake_q_node *next;
716 #ifdef CONFIG_KMAP_LOCAL
718 pte_t pteval[KM_MAX_IDX];
723 #ifdef CONFIG_THREAD_INFO_IN_TASK
725 * For reasons of header soup (see current_thread_info()), this
726 * must be the first element of task_struct.
728 struct thread_info thread_info;
730 unsigned int __state;
732 #ifdef CONFIG_PREEMPT_RT
733 /* saved state for "spinlock sleepers" */
734 unsigned int saved_state;
738 * This begins the randomizable portion of task_struct. Only
739 * scheduling-critical items should be added above here.
741 randomized_struct_fields_start
745 /* Per task flags (PF_*), defined further below: */
751 struct __call_single_node wake_entry;
752 #ifdef CONFIG_THREAD_INFO_IN_TASK
756 unsigned int wakee_flips;
757 unsigned long wakee_flip_decay_ts;
758 struct task_struct *last_wakee;
761 * recent_used_cpu is initially set as the last CPU used by a task
762 * that wakes affine another task. Waker/wakee relationships can
763 * push tasks around a CPU where each wakeup moves to the next one.
764 * Tracking a recently used CPU allows a quick search for a recently
765 * used CPU that may be idle.
775 unsigned int rt_priority;
777 const struct sched_class *sched_class;
778 struct sched_entity se;
779 struct sched_rt_entity rt;
780 struct sched_dl_entity dl;
782 #ifdef CONFIG_SCHED_CORE
783 struct rb_node core_node;
784 unsigned long core_cookie;
785 unsigned int core_occupation;
788 #ifdef CONFIG_CGROUP_SCHED
789 struct task_group *sched_task_group;
792 #ifdef CONFIG_UCLAMP_TASK
794 * Clamp values requested for a scheduling entity.
795 * Must be updated with task_rq_lock() held.
797 struct uclamp_se uclamp_req[UCLAMP_CNT];
799 * Effective clamp values used for a scheduling entity.
800 * Must be updated with task_rq_lock() held.
802 struct uclamp_se uclamp[UCLAMP_CNT];
805 #ifdef CONFIG_PREEMPT_NOTIFIERS
806 /* List of struct preempt_notifier: */
807 struct hlist_head preempt_notifiers;
810 #ifdef CONFIG_BLK_DEV_IO_TRACE
811 unsigned int btrace_seq;
816 const cpumask_t *cpus_ptr;
817 cpumask_t *user_cpus_ptr;
819 void *migration_pending;
821 unsigned short migration_disabled;
823 unsigned short migration_flags;
825 #ifdef CONFIG_PREEMPT_RCU
826 int rcu_read_lock_nesting;
827 union rcu_special rcu_read_unlock_special;
828 struct list_head rcu_node_entry;
829 struct rcu_node *rcu_blocked_node;
830 #endif /* #ifdef CONFIG_PREEMPT_RCU */
832 #ifdef CONFIG_TASKS_RCU
833 unsigned long rcu_tasks_nvcsw;
834 u8 rcu_tasks_holdout;
836 int rcu_tasks_idle_cpu;
837 struct list_head rcu_tasks_holdout_list;
838 #endif /* #ifdef CONFIG_TASKS_RCU */
840 #ifdef CONFIG_TASKS_TRACE_RCU
841 int trc_reader_nesting;
843 union rcu_special trc_reader_special;
844 bool trc_reader_checked;
845 struct list_head trc_holdout_list;
846 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
848 struct sched_info sched_info;
850 struct list_head tasks;
852 struct plist_node pushable_tasks;
853 struct rb_node pushable_dl_tasks;
856 struct mm_struct *mm;
857 struct mm_struct *active_mm;
859 /* Per-thread vma caching: */
860 struct vmacache vmacache;
862 #ifdef SPLIT_RSS_COUNTING
863 struct task_rss_stat rss_stat;
868 /* The signal sent when the parent dies: */
870 /* JOBCTL_*, siglock protected: */
871 unsigned long jobctl;
873 /* Used for emulating ABI behavior of previous Linux versions: */
874 unsigned int personality;
876 /* Scheduler bits, serialized by scheduler locks: */
877 unsigned sched_reset_on_fork:1;
878 unsigned sched_contributes_to_load:1;
879 unsigned sched_migrated:1;
881 unsigned sched_psi_wake_requeue:1;
884 /* Force alignment to the next boundary: */
887 /* Unserialized, strictly 'current' */
890 * This field must not be in the scheduler word above due to wakelist
891 * queueing no longer being serialized by p->on_cpu. However:
894 * schedule() if (p->on_rq && ..) // false
895 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
896 * deactivate_task() ttwu_queue_wakelist())
897 * p->on_rq = 0; p->sched_remote_wakeup = Y;
899 * guarantees all stores of 'current' are visible before
900 * ->sched_remote_wakeup gets used, so it can be in this word.
902 unsigned sched_remote_wakeup:1;
904 /* Bit to tell LSMs we're in execve(): */
905 unsigned in_execve:1;
906 unsigned in_iowait:1;
907 #ifndef TIF_RESTORE_SIGMASK
908 unsigned restore_sigmask:1;
911 unsigned in_user_fault:1;
913 #ifdef CONFIG_COMPAT_BRK
914 unsigned brk_randomized:1;
916 #ifdef CONFIG_CGROUPS
917 /* disallow userland-initiated cgroup migration */
918 unsigned no_cgroup_migration:1;
919 /* task is frozen/stopped (used by the cgroup freezer) */
922 #ifdef CONFIG_BLK_CGROUP
923 unsigned use_memdelay:1;
926 /* Stalled due to lack of memory */
927 unsigned in_memstall:1;
929 #ifdef CONFIG_PAGE_OWNER
930 /* Used by page_owner=on to detect recursion in page tracking. */
931 unsigned in_page_owner:1;
933 #ifdef CONFIG_EVENTFD
934 /* Recursion prevention for eventfd_signal() */
935 unsigned in_eventfd_signal:1;
938 unsigned long atomic_flags; /* Flags requiring atomic access. */
940 struct restart_block restart_block;
945 #ifdef CONFIG_STACKPROTECTOR
946 /* Canary value for the -fstack-protector GCC feature: */
947 unsigned long stack_canary;
950 * Pointers to the (original) parent process, youngest child, younger sibling,
951 * older sibling, respectively. (p->father can be replaced with
952 * p->real_parent->pid)
955 /* Real parent process: */
956 struct task_struct __rcu *real_parent;
958 /* Recipient of SIGCHLD, wait4() reports: */
959 struct task_struct __rcu *parent;
962 * Children/sibling form the list of natural children:
964 struct list_head children;
965 struct list_head sibling;
966 struct task_struct *group_leader;
969 * 'ptraced' is the list of tasks this task is using ptrace() on.
971 * This includes both natural children and PTRACE_ATTACH targets.
972 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
974 struct list_head ptraced;
975 struct list_head ptrace_entry;
977 /* PID/PID hash table linkage. */
978 struct pid *thread_pid;
979 struct hlist_node pid_links[PIDTYPE_MAX];
980 struct list_head thread_group;
981 struct list_head thread_node;
983 struct completion *vfork_done;
985 /* CLONE_CHILD_SETTID: */
986 int __user *set_child_tid;
988 /* CLONE_CHILD_CLEARTID: */
989 int __user *clear_child_tid;
996 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1001 struct prev_cputime prev_cputime;
1002 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1006 #ifdef CONFIG_NO_HZ_FULL
1007 atomic_t tick_dep_mask;
1009 /* Context switch counts: */
1010 unsigned long nvcsw;
1011 unsigned long nivcsw;
1013 /* Monotonic time in nsecs: */
1016 /* Boot based time in nsecs: */
1019 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1020 unsigned long min_flt;
1021 unsigned long maj_flt;
1023 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1024 struct posix_cputimers posix_cputimers;
1026 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1027 struct posix_cputimers_work posix_cputimers_work;
1030 /* Process credentials: */
1032 /* Tracer's credentials at attach: */
1033 const struct cred __rcu *ptracer_cred;
1035 /* Objective and real subjective task credentials (COW): */
1036 const struct cred __rcu *real_cred;
1038 /* Effective (overridable) subjective task credentials (COW): */
1039 const struct cred __rcu *cred;
1042 /* Cached requested key. */
1043 struct key *cached_requested_key;
1047 * executable name, excluding path.
1049 * - normally initialized setup_new_exec()
1050 * - access it with [gs]et_task_comm()
1051 * - lock it with task_lock()
1053 char comm[TASK_COMM_LEN];
1055 struct nameidata *nameidata;
1057 #ifdef CONFIG_SYSVIPC
1058 struct sysv_sem sysvsem;
1059 struct sysv_shm sysvshm;
1061 #ifdef CONFIG_DETECT_HUNG_TASK
1062 unsigned long last_switch_count;
1063 unsigned long last_switch_time;
1065 /* Filesystem information: */
1066 struct fs_struct *fs;
1068 /* Open file information: */
1069 struct files_struct *files;
1071 #ifdef CONFIG_IO_URING
1072 struct io_uring_task *io_uring;
1076 struct nsproxy *nsproxy;
1078 /* Signal handlers: */
1079 struct signal_struct *signal;
1080 struct sighand_struct __rcu *sighand;
1082 sigset_t real_blocked;
1083 /* Restored if set_restore_sigmask() was used: */
1084 sigset_t saved_sigmask;
1085 struct sigpending pending;
1086 unsigned long sas_ss_sp;
1088 unsigned int sas_ss_flags;
1090 struct callback_head *task_works;
1093 #ifdef CONFIG_AUDITSYSCALL
1094 struct audit_context *audit_context;
1097 unsigned int sessionid;
1099 struct seccomp seccomp;
1100 struct syscall_user_dispatch syscall_dispatch;
1102 /* Thread group tracking: */
1106 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1107 spinlock_t alloc_lock;
1109 /* Protection of the PI data structures: */
1110 raw_spinlock_t pi_lock;
1112 struct wake_q_node wake_q;
1114 #ifdef CONFIG_RT_MUTEXES
1115 /* PI waiters blocked on a rt_mutex held by this task: */
1116 struct rb_root_cached pi_waiters;
1117 /* Updated under owner's pi_lock and rq lock */
1118 struct task_struct *pi_top_task;
1119 /* Deadlock detection and priority inheritance handling: */
1120 struct rt_mutex_waiter *pi_blocked_on;
1123 #ifdef CONFIG_DEBUG_MUTEXES
1124 /* Mutex deadlock detection: */
1125 struct mutex_waiter *blocked_on;
1128 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1129 int non_block_count;
1132 #ifdef CONFIG_TRACE_IRQFLAGS
1133 struct irqtrace_events irqtrace;
1134 unsigned int hardirq_threaded;
1135 u64 hardirq_chain_key;
1136 int softirqs_enabled;
1137 int softirq_context;
1140 #ifdef CONFIG_PREEMPT_RT
1141 int softirq_disable_cnt;
1144 #ifdef CONFIG_LOCKDEP
1145 # define MAX_LOCK_DEPTH 48UL
1148 unsigned int lockdep_recursion;
1149 struct held_lock held_locks[MAX_LOCK_DEPTH];
1152 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1153 unsigned int in_ubsan;
1156 /* Journalling filesystem info: */
1159 /* Stacked block device info: */
1160 struct bio_list *bio_list;
1163 /* Stack plugging: */
1164 struct blk_plug *plug;
1168 struct reclaim_state *reclaim_state;
1170 struct backing_dev_info *backing_dev_info;
1172 struct io_context *io_context;
1174 #ifdef CONFIG_COMPACTION
1175 struct capture_control *capture_control;
1178 unsigned long ptrace_message;
1179 kernel_siginfo_t *last_siginfo;
1181 struct task_io_accounting ioac;
1183 /* Pressure stall state */
1184 unsigned int psi_flags;
1186 #ifdef CONFIG_TASK_XACCT
1187 /* Accumulated RSS usage: */
1189 /* Accumulated virtual memory usage: */
1191 /* stime + utime since last update: */
1194 #ifdef CONFIG_CPUSETS
1195 /* Protected by ->alloc_lock: */
1196 nodemask_t mems_allowed;
1197 /* Sequence number to catch updates: */
1198 seqcount_spinlock_t mems_allowed_seq;
1199 int cpuset_mem_spread_rotor;
1200 int cpuset_slab_spread_rotor;
1202 #ifdef CONFIG_CGROUPS
1203 /* Control Group info protected by css_set_lock: */
1204 struct css_set __rcu *cgroups;
1205 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1206 struct list_head cg_list;
1208 #ifdef CONFIG_X86_CPU_RESCTRL
1213 struct robust_list_head __user *robust_list;
1214 #ifdef CONFIG_COMPAT
1215 struct compat_robust_list_head __user *compat_robust_list;
1217 struct list_head pi_state_list;
1218 struct futex_pi_state *pi_state_cache;
1219 struct mutex futex_exit_mutex;
1220 unsigned int futex_state;
1222 #ifdef CONFIG_PERF_EVENTS
1223 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1224 struct mutex perf_event_mutex;
1225 struct list_head perf_event_list;
1227 #ifdef CONFIG_DEBUG_PREEMPT
1228 unsigned long preempt_disable_ip;
1231 /* Protected by alloc_lock: */
1232 struct mempolicy *mempolicy;
1234 short pref_node_fork;
1236 #ifdef CONFIG_NUMA_BALANCING
1238 unsigned int numa_scan_period;
1239 unsigned int numa_scan_period_max;
1240 int numa_preferred_nid;
1241 unsigned long numa_migrate_retry;
1242 /* Migration stamp: */
1244 u64 last_task_numa_placement;
1245 u64 last_sum_exec_runtime;
1246 struct callback_head numa_work;
1249 * This pointer is only modified for current in syscall and
1250 * pagefault context (and for tasks being destroyed), so it can be read
1251 * from any of the following contexts:
1252 * - RCU read-side critical section
1253 * - current->numa_group from everywhere
1254 * - task's runqueue locked, task not running
1256 struct numa_group __rcu *numa_group;
1259 * numa_faults is an array split into four regions:
1260 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1261 * in this precise order.
1263 * faults_memory: Exponential decaying average of faults on a per-node
1264 * basis. Scheduling placement decisions are made based on these
1265 * counts. The values remain static for the duration of a PTE scan.
1266 * faults_cpu: Track the nodes the process was running on when a NUMA
1267 * hinting fault was incurred.
1268 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1269 * during the current scan window. When the scan completes, the counts
1270 * in faults_memory and faults_cpu decay and these values are copied.
1272 unsigned long *numa_faults;
1273 unsigned long total_numa_faults;
1276 * numa_faults_locality tracks if faults recorded during the last
1277 * scan window were remote/local or failed to migrate. The task scan
1278 * period is adapted based on the locality of the faults with different
1279 * weights depending on whether they were shared or private faults
1281 unsigned long numa_faults_locality[3];
1283 unsigned long numa_pages_migrated;
1284 #endif /* CONFIG_NUMA_BALANCING */
1287 struct rseq __user *rseq;
1290 * RmW on rseq_event_mask must be performed atomically
1291 * with respect to preemption.
1293 unsigned long rseq_event_mask;
1296 struct tlbflush_unmap_batch tlb_ubc;
1299 refcount_t rcu_users;
1300 struct rcu_head rcu;
1303 /* Cache last used pipe for splice(): */
1304 struct pipe_inode_info *splice_pipe;
1306 struct page_frag task_frag;
1308 #ifdef CONFIG_TASK_DELAY_ACCT
1309 struct task_delay_info *delays;
1312 #ifdef CONFIG_FAULT_INJECTION
1314 unsigned int fail_nth;
1317 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1318 * balance_dirty_pages() for a dirty throttling pause:
1321 int nr_dirtied_pause;
1322 /* Start of a write-and-pause period: */
1323 unsigned long dirty_paused_when;
1325 #ifdef CONFIG_LATENCYTOP
1326 int latency_record_count;
1327 struct latency_record latency_record[LT_SAVECOUNT];
1330 * Time slack values; these are used to round up poll() and
1331 * select() etc timeout values. These are in nanoseconds.
1334 u64 default_timer_slack_ns;
1336 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1337 unsigned int kasan_depth;
1341 struct kcsan_ctx kcsan_ctx;
1342 #ifdef CONFIG_TRACE_IRQFLAGS
1343 struct irqtrace_events kcsan_save_irqtrace;
1347 #if IS_ENABLED(CONFIG_KUNIT)
1348 struct kunit *kunit_test;
1351 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1352 /* Index of current stored address in ret_stack: */
1356 /* Stack of return addresses for return function tracing: */
1357 struct ftrace_ret_stack *ret_stack;
1359 /* Timestamp for last schedule: */
1360 unsigned long long ftrace_timestamp;
1363 * Number of functions that haven't been traced
1364 * because of depth overrun:
1366 atomic_t trace_overrun;
1368 /* Pause tracing: */
1369 atomic_t tracing_graph_pause;
1372 #ifdef CONFIG_TRACING
1373 /* State flags for use by tracers: */
1374 unsigned long trace;
1376 /* Bitmask and counter of trace recursion: */
1377 unsigned long trace_recursion;
1378 #endif /* CONFIG_TRACING */
1381 /* See kernel/kcov.c for more details. */
1383 /* Coverage collection mode enabled for this task (0 if disabled): */
1384 unsigned int kcov_mode;
1386 /* Size of the kcov_area: */
1387 unsigned int kcov_size;
1389 /* Buffer for coverage collection: */
1392 /* KCOV descriptor wired with this task or NULL: */
1395 /* KCOV common handle for remote coverage collection: */
1398 /* KCOV sequence number: */
1401 /* Collect coverage from softirq context: */
1402 unsigned int kcov_softirq;
1406 struct mem_cgroup *memcg_in_oom;
1407 gfp_t memcg_oom_gfp_mask;
1408 int memcg_oom_order;
1410 /* Number of pages to reclaim on returning to userland: */
1411 unsigned int memcg_nr_pages_over_high;
1413 /* Used by memcontrol for targeted memcg charge: */
1414 struct mem_cgroup *active_memcg;
1417 #ifdef CONFIG_BLK_CGROUP
1418 struct request_queue *throttle_queue;
1421 #ifdef CONFIG_UPROBES
1422 struct uprobe_task *utask;
1424 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1425 unsigned int sequential_io;
1426 unsigned int sequential_io_avg;
1428 struct kmap_ctrl kmap_ctrl;
1429 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1430 unsigned long task_state_change;
1431 # ifdef CONFIG_PREEMPT_RT
1432 unsigned long saved_state_change;
1435 int pagefault_disabled;
1437 struct task_struct *oom_reaper_list;
1439 #ifdef CONFIG_VMAP_STACK
1440 struct vm_struct *stack_vm_area;
1442 #ifdef CONFIG_THREAD_INFO_IN_TASK
1443 /* A live task holds one reference: */
1444 refcount_t stack_refcount;
1446 #ifdef CONFIG_LIVEPATCH
1449 #ifdef CONFIG_SECURITY
1450 /* Used by LSM modules for access restriction: */
1453 #ifdef CONFIG_BPF_SYSCALL
1454 /* Used by BPF task local storage */
1455 struct bpf_local_storage __rcu *bpf_storage;
1458 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1459 unsigned long lowest_stack;
1460 unsigned long prev_lowest_stack;
1463 #ifdef CONFIG_X86_MCE
1464 void __user *mce_vaddr;
1469 __mce_reserved : 62;
1470 struct callback_head mce_kill_me;
1473 #ifdef CONFIG_KRETPROBES
1474 struct llist_head kretprobe_instances;
1478 * New fields for task_struct should be added above here, so that
1479 * they are included in the randomized portion of task_struct.
1481 randomized_struct_fields_end
1483 /* CPU-specific state of this task: */
1484 struct thread_struct thread;
1487 * WARNING: on x86, 'thread_struct' contains a variable-sized
1488 * structure. It *MUST* be at the end of 'task_struct'.
1490 * Do not put anything below here!
1494 static inline struct pid *task_pid(struct task_struct *task)
1496 return task->thread_pid;
1500 * the helpers to get the task's different pids as they are seen
1501 * from various namespaces
1503 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1504 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1506 * task_xid_nr_ns() : id seen from the ns specified;
1508 * see also pid_nr() etc in include/linux/pid.h
1510 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1512 static inline pid_t task_pid_nr(struct task_struct *tsk)
1517 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1519 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1522 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1524 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1528 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1534 * pid_alive - check that a task structure is not stale
1535 * @p: Task structure to be checked.
1537 * Test if a process is not yet dead (at most zombie state)
1538 * If pid_alive fails, then pointers within the task structure
1539 * can be stale and must not be dereferenced.
1541 * Return: 1 if the process is alive. 0 otherwise.
1543 static inline int pid_alive(const struct task_struct *p)
1545 return p->thread_pid != NULL;
1548 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1550 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1553 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1555 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1559 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1561 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1564 static inline pid_t task_session_vnr(struct task_struct *tsk)
1566 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1569 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1571 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1574 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1576 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1579 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1585 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1591 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1593 return task_ppid_nr_ns(tsk, &init_pid_ns);
1596 /* Obsolete, do not use: */
1597 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1599 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1602 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1603 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1605 static inline unsigned int task_state_index(struct task_struct *tsk)
1607 unsigned int tsk_state = READ_ONCE(tsk->__state);
1608 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1610 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1612 if (tsk_state == TASK_IDLE)
1613 state = TASK_REPORT_IDLE;
1618 static inline char task_index_to_char(unsigned int state)
1620 static const char state_char[] = "RSDTtXZPI";
1622 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1624 return state_char[state];
1627 static inline char task_state_to_char(struct task_struct *tsk)
1629 return task_index_to_char(task_state_index(tsk));
1633 * is_global_init - check if a task structure is init. Since init
1634 * is free to have sub-threads we need to check tgid.
1635 * @tsk: Task structure to be checked.
1637 * Check if a task structure is the first user space task the kernel created.
1639 * Return: 1 if the task structure is init. 0 otherwise.
1641 static inline int is_global_init(struct task_struct *tsk)
1643 return task_tgid_nr(tsk) == 1;
1646 extern struct pid *cad_pid;
1651 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1652 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1653 #define PF_EXITING 0x00000004 /* Getting shut down */
1654 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1655 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1656 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1657 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1658 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1659 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1660 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1661 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1662 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1663 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1664 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1665 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1666 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1667 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1668 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1669 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1670 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1671 * I am cleaning dirty pages from some other bdi. */
1672 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1673 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1674 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1675 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1676 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1677 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1678 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1679 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1682 * Only the _current_ task can read/write to tsk->flags, but other
1683 * tasks can access tsk->flags in readonly mode for example
1684 * with tsk_used_math (like during threaded core dumping).
1685 * There is however an exception to this rule during ptrace
1686 * or during fork: the ptracer task is allowed to write to the
1687 * child->flags of its traced child (same goes for fork, the parent
1688 * can write to the child->flags), because we're guaranteed the
1689 * child is not running and in turn not changing child->flags
1690 * at the same time the parent does it.
1692 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1693 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1694 #define clear_used_math() clear_stopped_child_used_math(current)
1695 #define set_used_math() set_stopped_child_used_math(current)
1697 #define conditional_stopped_child_used_math(condition, child) \
1698 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1700 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1702 #define copy_to_stopped_child_used_math(child) \
1703 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1705 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1706 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1707 #define used_math() tsk_used_math(current)
1709 static inline bool is_percpu_thread(void)
1712 return (current->flags & PF_NO_SETAFFINITY) &&
1713 (current->nr_cpus_allowed == 1);
1719 /* Per-process atomic flags. */
1720 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1721 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1722 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1723 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1724 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1725 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1726 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1727 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1729 #define TASK_PFA_TEST(name, func) \
1730 static inline bool task_##func(struct task_struct *p) \
1731 { return test_bit(PFA_##name, &p->atomic_flags); }
1733 #define TASK_PFA_SET(name, func) \
1734 static inline void task_set_##func(struct task_struct *p) \
1735 { set_bit(PFA_##name, &p->atomic_flags); }
1737 #define TASK_PFA_CLEAR(name, func) \
1738 static inline void task_clear_##func(struct task_struct *p) \
1739 { clear_bit(PFA_##name, &p->atomic_flags); }
1741 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1742 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1744 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1745 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1746 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1748 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1749 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1750 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1752 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1753 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1754 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1756 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1757 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1758 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1760 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1761 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1763 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1764 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1765 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1767 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1768 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1771 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1773 current->flags &= ~flags;
1774 current->flags |= orig_flags & flags;
1777 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1778 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1780 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1781 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1782 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1783 extern void release_user_cpus_ptr(struct task_struct *p);
1784 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1785 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1786 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1788 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1791 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1793 if (!cpumask_test_cpu(0, new_mask))
1797 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1799 if (src->user_cpus_ptr)
1803 static inline void release_user_cpus_ptr(struct task_struct *p)
1805 WARN_ON(p->user_cpus_ptr);
1808 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1814 extern int yield_to(struct task_struct *p, bool preempt);
1815 extern void set_user_nice(struct task_struct *p, long nice);
1816 extern int task_prio(const struct task_struct *p);
1819 * task_nice - return the nice value of a given task.
1820 * @p: the task in question.
1822 * Return: The nice value [ -20 ... 0 ... 19 ].
1824 static inline int task_nice(const struct task_struct *p)
1826 return PRIO_TO_NICE((p)->static_prio);
1829 extern int can_nice(const struct task_struct *p, const int nice);
1830 extern int task_curr(const struct task_struct *p);
1831 extern int idle_cpu(int cpu);
1832 extern int available_idle_cpu(int cpu);
1833 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1834 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1835 extern void sched_set_fifo(struct task_struct *p);
1836 extern void sched_set_fifo_low(struct task_struct *p);
1837 extern void sched_set_normal(struct task_struct *p, int nice);
1838 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1839 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1840 extern struct task_struct *idle_task(int cpu);
1843 * is_idle_task - is the specified task an idle task?
1844 * @p: the task in question.
1846 * Return: 1 if @p is an idle task. 0 otherwise.
1848 static __always_inline bool is_idle_task(const struct task_struct *p)
1850 return !!(p->flags & PF_IDLE);
1853 extern struct task_struct *curr_task(int cpu);
1854 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1858 union thread_union {
1859 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1860 struct task_struct task;
1862 #ifndef CONFIG_THREAD_INFO_IN_TASK
1863 struct thread_info thread_info;
1865 unsigned long stack[THREAD_SIZE/sizeof(long)];
1868 #ifndef CONFIG_THREAD_INFO_IN_TASK
1869 extern struct thread_info init_thread_info;
1872 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1874 #ifdef CONFIG_THREAD_INFO_IN_TASK
1875 static inline struct thread_info *task_thread_info(struct task_struct *task)
1877 return &task->thread_info;
1879 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1880 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1884 * find a task by one of its numerical ids
1886 * find_task_by_pid_ns():
1887 * finds a task by its pid in the specified namespace
1888 * find_task_by_vpid():
1889 * finds a task by its virtual pid
1891 * see also find_vpid() etc in include/linux/pid.h
1894 extern struct task_struct *find_task_by_vpid(pid_t nr);
1895 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1898 * find a task by its virtual pid and get the task struct
1900 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1902 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1903 extern int wake_up_process(struct task_struct *tsk);
1904 extern void wake_up_new_task(struct task_struct *tsk);
1907 extern void kick_process(struct task_struct *tsk);
1909 static inline void kick_process(struct task_struct *tsk) { }
1912 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1914 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1916 __set_task_comm(tsk, from, false);
1919 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1920 #define get_task_comm(buf, tsk) ({ \
1921 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1922 __get_task_comm(buf, sizeof(buf), tsk); \
1926 static __always_inline void scheduler_ipi(void)
1929 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1930 * TIF_NEED_RESCHED remotely (for the first time) will also send
1933 preempt_fold_need_resched();
1935 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1937 static inline void scheduler_ipi(void) { }
1938 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1945 * Set thread flags in other task's structures.
1946 * See asm/thread_info.h for TIF_xxxx flags available:
1948 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1950 set_ti_thread_flag(task_thread_info(tsk), flag);
1953 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1955 clear_ti_thread_flag(task_thread_info(tsk), flag);
1958 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1961 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1964 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1966 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1969 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1971 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1974 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1976 return test_ti_thread_flag(task_thread_info(tsk), flag);
1979 static inline void set_tsk_need_resched(struct task_struct *tsk)
1981 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1984 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1986 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1989 static inline int test_tsk_need_resched(struct task_struct *tsk)
1991 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1995 * cond_resched() and cond_resched_lock(): latency reduction via
1996 * explicit rescheduling in places that are safe. The return
1997 * value indicates whether a reschedule was done in fact.
1998 * cond_resched_lock() will drop the spinlock before scheduling,
2000 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2001 extern int __cond_resched(void);
2003 #ifdef CONFIG_PREEMPT_DYNAMIC
2005 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2007 static __always_inline int _cond_resched(void)
2009 return static_call_mod(cond_resched)();
2014 static inline int _cond_resched(void)
2016 return __cond_resched();
2019 #endif /* CONFIG_PREEMPT_DYNAMIC */
2023 static inline int _cond_resched(void) { return 0; }
2025 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2027 #define cond_resched() ({ \
2028 ___might_sleep(__FILE__, __LINE__, 0); \
2032 extern int __cond_resched_lock(spinlock_t *lock);
2033 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2034 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2036 #define cond_resched_lock(lock) ({ \
2037 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
2038 __cond_resched_lock(lock); \
2041 #define cond_resched_rwlock_read(lock) ({ \
2042 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2043 __cond_resched_rwlock_read(lock); \
2046 #define cond_resched_rwlock_write(lock) ({ \
2047 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2048 __cond_resched_rwlock_write(lock); \
2051 static inline void cond_resched_rcu(void)
2053 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2061 * Does a critical section need to be broken due to another
2062 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2063 * but a general need for low latency)
2065 static inline int spin_needbreak(spinlock_t *lock)
2067 #ifdef CONFIG_PREEMPTION
2068 return spin_is_contended(lock);
2075 * Check if a rwlock is contended.
2076 * Returns non-zero if there is another task waiting on the rwlock.
2077 * Returns zero if the lock is not contended or the system / underlying
2078 * rwlock implementation does not support contention detection.
2079 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2082 static inline int rwlock_needbreak(rwlock_t *lock)
2084 #ifdef CONFIG_PREEMPTION
2085 return rwlock_is_contended(lock);
2091 static __always_inline bool need_resched(void)
2093 return unlikely(tif_need_resched());
2097 * Wrappers for p->thread_info->cpu access. No-op on UP.
2101 static inline unsigned int task_cpu(const struct task_struct *p)
2103 #ifdef CONFIG_THREAD_INFO_IN_TASK
2104 return READ_ONCE(p->cpu);
2106 return READ_ONCE(task_thread_info(p)->cpu);
2110 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2114 static inline unsigned int task_cpu(const struct task_struct *p)
2119 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2123 #endif /* CONFIG_SMP */
2125 extern bool sched_task_on_rq(struct task_struct *p);
2128 * In order to reduce various lock holder preemption latencies provide an
2129 * interface to see if a vCPU is currently running or not.
2131 * This allows us to terminate optimistic spin loops and block, analogous to
2132 * the native optimistic spin heuristic of testing if the lock owner task is
2135 #ifndef vcpu_is_preempted
2136 static inline bool vcpu_is_preempted(int cpu)
2142 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2143 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2145 #ifndef TASK_SIZE_OF
2146 #define TASK_SIZE_OF(tsk) TASK_SIZE
2150 /* Returns effective CPU energy utilization, as seen by the scheduler */
2151 unsigned long sched_cpu_util(int cpu, unsigned long max);
2152 #endif /* CONFIG_SMP */
2157 * Map the event mask on the user-space ABI enum rseq_cs_flags
2158 * for direct mask checks.
2160 enum rseq_event_mask_bits {
2161 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2162 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2163 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2166 enum rseq_event_mask {
2167 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2168 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2169 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2172 static inline void rseq_set_notify_resume(struct task_struct *t)
2175 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2178 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2180 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2181 struct pt_regs *regs)
2184 __rseq_handle_notify_resume(ksig, regs);
2187 static inline void rseq_signal_deliver(struct ksignal *ksig,
2188 struct pt_regs *regs)
2191 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2193 rseq_handle_notify_resume(ksig, regs);
2196 /* rseq_preempt() requires preemption to be disabled. */
2197 static inline void rseq_preempt(struct task_struct *t)
2199 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2200 rseq_set_notify_resume(t);
2203 /* rseq_migrate() requires preemption to be disabled. */
2204 static inline void rseq_migrate(struct task_struct *t)
2206 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2207 rseq_set_notify_resume(t);
2211 * If parent process has a registered restartable sequences area, the
2212 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2214 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2216 if (clone_flags & CLONE_VM) {
2219 t->rseq_event_mask = 0;
2221 t->rseq = current->rseq;
2222 t->rseq_sig = current->rseq_sig;
2223 t->rseq_event_mask = current->rseq_event_mask;
2227 static inline void rseq_execve(struct task_struct *t)
2231 t->rseq_event_mask = 0;
2236 static inline void rseq_set_notify_resume(struct task_struct *t)
2239 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2240 struct pt_regs *regs)
2243 static inline void rseq_signal_deliver(struct ksignal *ksig,
2244 struct pt_regs *regs)
2247 static inline void rseq_preempt(struct task_struct *t)
2250 static inline void rseq_migrate(struct task_struct *t)
2253 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2256 static inline void rseq_execve(struct task_struct *t)
2262 #ifdef CONFIG_DEBUG_RSEQ
2264 void rseq_syscall(struct pt_regs *regs);
2268 static inline void rseq_syscall(struct pt_regs *regs)
2274 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2275 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2276 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2278 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2279 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2280 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2282 int sched_trace_rq_cpu(struct rq *rq);
2283 int sched_trace_rq_cpu_capacity(struct rq *rq);
2284 int sched_trace_rq_nr_running(struct rq *rq);
2286 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2288 #ifdef CONFIG_SCHED_CORE
2289 extern void sched_core_free(struct task_struct *tsk);
2290 extern void sched_core_fork(struct task_struct *p);
2291 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2292 unsigned long uaddr);
2294 static inline void sched_core_free(struct task_struct *tsk) { }
2295 static inline void sched_core_fork(struct task_struct *p) { }