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;
818 void *migration_pending;
820 unsigned short migration_disabled;
822 unsigned short migration_flags;
824 #ifdef CONFIG_PREEMPT_RCU
825 int rcu_read_lock_nesting;
826 union rcu_special rcu_read_unlock_special;
827 struct list_head rcu_node_entry;
828 struct rcu_node *rcu_blocked_node;
829 #endif /* #ifdef CONFIG_PREEMPT_RCU */
831 #ifdef CONFIG_TASKS_RCU
832 unsigned long rcu_tasks_nvcsw;
833 u8 rcu_tasks_holdout;
835 int rcu_tasks_idle_cpu;
836 struct list_head rcu_tasks_holdout_list;
837 #endif /* #ifdef CONFIG_TASKS_RCU */
839 #ifdef CONFIG_TASKS_TRACE_RCU
840 int trc_reader_nesting;
842 union rcu_special trc_reader_special;
843 bool trc_reader_checked;
844 struct list_head trc_holdout_list;
845 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
847 struct sched_info sched_info;
849 struct list_head tasks;
851 struct plist_node pushable_tasks;
852 struct rb_node pushable_dl_tasks;
855 struct mm_struct *mm;
856 struct mm_struct *active_mm;
858 /* Per-thread vma caching: */
859 struct vmacache vmacache;
861 #ifdef SPLIT_RSS_COUNTING
862 struct task_rss_stat rss_stat;
867 /* The signal sent when the parent dies: */
869 /* JOBCTL_*, siglock protected: */
870 unsigned long jobctl;
872 /* Used for emulating ABI behavior of previous Linux versions: */
873 unsigned int personality;
875 /* Scheduler bits, serialized by scheduler locks: */
876 unsigned sched_reset_on_fork:1;
877 unsigned sched_contributes_to_load:1;
878 unsigned sched_migrated:1;
880 unsigned sched_psi_wake_requeue:1;
883 /* Force alignment to the next boundary: */
886 /* Unserialized, strictly 'current' */
889 * This field must not be in the scheduler word above due to wakelist
890 * queueing no longer being serialized by p->on_cpu. However:
893 * schedule() if (p->on_rq && ..) // false
894 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
895 * deactivate_task() ttwu_queue_wakelist())
896 * p->on_rq = 0; p->sched_remote_wakeup = Y;
898 * guarantees all stores of 'current' are visible before
899 * ->sched_remote_wakeup gets used, so it can be in this word.
901 unsigned sched_remote_wakeup:1;
903 /* Bit to tell LSMs we're in execve(): */
904 unsigned in_execve:1;
905 unsigned in_iowait:1;
906 #ifndef TIF_RESTORE_SIGMASK
907 unsigned restore_sigmask:1;
910 unsigned in_user_fault:1;
912 #ifdef CONFIG_COMPAT_BRK
913 unsigned brk_randomized:1;
915 #ifdef CONFIG_CGROUPS
916 /* disallow userland-initiated cgroup migration */
917 unsigned no_cgroup_migration:1;
918 /* task is frozen/stopped (used by the cgroup freezer) */
921 #ifdef CONFIG_BLK_CGROUP
922 unsigned use_memdelay:1;
925 /* Stalled due to lack of memory */
926 unsigned in_memstall:1;
928 #ifdef CONFIG_PAGE_OWNER
929 /* Used by page_owner=on to detect recursion in page tracking. */
930 unsigned in_page_owner:1;
933 unsigned long atomic_flags; /* Flags requiring atomic access. */
935 struct restart_block restart_block;
940 #ifdef CONFIG_STACKPROTECTOR
941 /* Canary value for the -fstack-protector GCC feature: */
942 unsigned long stack_canary;
945 * Pointers to the (original) parent process, youngest child, younger sibling,
946 * older sibling, respectively. (p->father can be replaced with
947 * p->real_parent->pid)
950 /* Real parent process: */
951 struct task_struct __rcu *real_parent;
953 /* Recipient of SIGCHLD, wait4() reports: */
954 struct task_struct __rcu *parent;
957 * Children/sibling form the list of natural children:
959 struct list_head children;
960 struct list_head sibling;
961 struct task_struct *group_leader;
964 * 'ptraced' is the list of tasks this task is using ptrace() on.
966 * This includes both natural children and PTRACE_ATTACH targets.
967 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
969 struct list_head ptraced;
970 struct list_head ptrace_entry;
972 /* PID/PID hash table linkage. */
973 struct pid *thread_pid;
974 struct hlist_node pid_links[PIDTYPE_MAX];
975 struct list_head thread_group;
976 struct list_head thread_node;
978 struct completion *vfork_done;
980 /* CLONE_CHILD_SETTID: */
981 int __user *set_child_tid;
983 /* CLONE_CHILD_CLEARTID: */
984 int __user *clear_child_tid;
991 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
996 struct prev_cputime prev_cputime;
997 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1001 #ifdef CONFIG_NO_HZ_FULL
1002 atomic_t tick_dep_mask;
1004 /* Context switch counts: */
1005 unsigned long nvcsw;
1006 unsigned long nivcsw;
1008 /* Monotonic time in nsecs: */
1011 /* Boot based time in nsecs: */
1014 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1015 unsigned long min_flt;
1016 unsigned long maj_flt;
1018 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1019 struct posix_cputimers posix_cputimers;
1021 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1022 struct posix_cputimers_work posix_cputimers_work;
1025 /* Process credentials: */
1027 /* Tracer's credentials at attach: */
1028 const struct cred __rcu *ptracer_cred;
1030 /* Objective and real subjective task credentials (COW): */
1031 const struct cred __rcu *real_cred;
1033 /* Effective (overridable) subjective task credentials (COW): */
1034 const struct cred __rcu *cred;
1037 /* Cached requested key. */
1038 struct key *cached_requested_key;
1042 * executable name, excluding path.
1044 * - normally initialized setup_new_exec()
1045 * - access it with [gs]et_task_comm()
1046 * - lock it with task_lock()
1048 char comm[TASK_COMM_LEN];
1050 struct nameidata *nameidata;
1052 #ifdef CONFIG_SYSVIPC
1053 struct sysv_sem sysvsem;
1054 struct sysv_shm sysvshm;
1056 #ifdef CONFIG_DETECT_HUNG_TASK
1057 unsigned long last_switch_count;
1058 unsigned long last_switch_time;
1060 /* Filesystem information: */
1061 struct fs_struct *fs;
1063 /* Open file information: */
1064 struct files_struct *files;
1066 #ifdef CONFIG_IO_URING
1067 struct io_uring_task *io_uring;
1071 struct nsproxy *nsproxy;
1073 /* Signal handlers: */
1074 struct signal_struct *signal;
1075 struct sighand_struct __rcu *sighand;
1077 sigset_t real_blocked;
1078 /* Restored if set_restore_sigmask() was used: */
1079 sigset_t saved_sigmask;
1080 struct sigpending pending;
1081 unsigned long sas_ss_sp;
1083 unsigned int sas_ss_flags;
1085 struct callback_head *task_works;
1088 #ifdef CONFIG_AUDITSYSCALL
1089 struct audit_context *audit_context;
1092 unsigned int sessionid;
1094 struct seccomp seccomp;
1095 struct syscall_user_dispatch syscall_dispatch;
1097 /* Thread group tracking: */
1101 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1102 spinlock_t alloc_lock;
1104 /* Protection of the PI data structures: */
1105 raw_spinlock_t pi_lock;
1107 struct wake_q_node wake_q;
1109 #ifdef CONFIG_RT_MUTEXES
1110 /* PI waiters blocked on a rt_mutex held by this task: */
1111 struct rb_root_cached pi_waiters;
1112 /* Updated under owner's pi_lock and rq lock */
1113 struct task_struct *pi_top_task;
1114 /* Deadlock detection and priority inheritance handling: */
1115 struct rt_mutex_waiter *pi_blocked_on;
1118 #ifdef CONFIG_DEBUG_MUTEXES
1119 /* Mutex deadlock detection: */
1120 struct mutex_waiter *blocked_on;
1123 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1124 int non_block_count;
1127 #ifdef CONFIG_TRACE_IRQFLAGS
1128 struct irqtrace_events irqtrace;
1129 unsigned int hardirq_threaded;
1130 u64 hardirq_chain_key;
1131 int softirqs_enabled;
1132 int softirq_context;
1135 #ifdef CONFIG_PREEMPT_RT
1136 int softirq_disable_cnt;
1139 #ifdef CONFIG_LOCKDEP
1140 # define MAX_LOCK_DEPTH 48UL
1143 unsigned int lockdep_recursion;
1144 struct held_lock held_locks[MAX_LOCK_DEPTH];
1147 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1148 unsigned int in_ubsan;
1151 /* Journalling filesystem info: */
1154 /* Stacked block device info: */
1155 struct bio_list *bio_list;
1158 /* Stack plugging: */
1159 struct blk_plug *plug;
1163 struct reclaim_state *reclaim_state;
1165 struct backing_dev_info *backing_dev_info;
1167 struct io_context *io_context;
1169 #ifdef CONFIG_COMPACTION
1170 struct capture_control *capture_control;
1173 unsigned long ptrace_message;
1174 kernel_siginfo_t *last_siginfo;
1176 struct task_io_accounting ioac;
1178 /* Pressure stall state */
1179 unsigned int psi_flags;
1181 #ifdef CONFIG_TASK_XACCT
1182 /* Accumulated RSS usage: */
1184 /* Accumulated virtual memory usage: */
1186 /* stime + utime since last update: */
1189 #ifdef CONFIG_CPUSETS
1190 /* Protected by ->alloc_lock: */
1191 nodemask_t mems_allowed;
1192 /* Sequence number to catch updates: */
1193 seqcount_spinlock_t mems_allowed_seq;
1194 int cpuset_mem_spread_rotor;
1195 int cpuset_slab_spread_rotor;
1197 #ifdef CONFIG_CGROUPS
1198 /* Control Group info protected by css_set_lock: */
1199 struct css_set __rcu *cgroups;
1200 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1201 struct list_head cg_list;
1203 #ifdef CONFIG_X86_CPU_RESCTRL
1208 struct robust_list_head __user *robust_list;
1209 #ifdef CONFIG_COMPAT
1210 struct compat_robust_list_head __user *compat_robust_list;
1212 struct list_head pi_state_list;
1213 struct futex_pi_state *pi_state_cache;
1214 struct mutex futex_exit_mutex;
1215 unsigned int futex_state;
1217 #ifdef CONFIG_PERF_EVENTS
1218 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1219 struct mutex perf_event_mutex;
1220 struct list_head perf_event_list;
1222 #ifdef CONFIG_DEBUG_PREEMPT
1223 unsigned long preempt_disable_ip;
1226 /* Protected by alloc_lock: */
1227 struct mempolicy *mempolicy;
1229 short pref_node_fork;
1231 #ifdef CONFIG_NUMA_BALANCING
1233 unsigned int numa_scan_period;
1234 unsigned int numa_scan_period_max;
1235 int numa_preferred_nid;
1236 unsigned long numa_migrate_retry;
1237 /* Migration stamp: */
1239 u64 last_task_numa_placement;
1240 u64 last_sum_exec_runtime;
1241 struct callback_head numa_work;
1244 * This pointer is only modified for current in syscall and
1245 * pagefault context (and for tasks being destroyed), so it can be read
1246 * from any of the following contexts:
1247 * - RCU read-side critical section
1248 * - current->numa_group from everywhere
1249 * - task's runqueue locked, task not running
1251 struct numa_group __rcu *numa_group;
1254 * numa_faults is an array split into four regions:
1255 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1256 * in this precise order.
1258 * faults_memory: Exponential decaying average of faults on a per-node
1259 * basis. Scheduling placement decisions are made based on these
1260 * counts. The values remain static for the duration of a PTE scan.
1261 * faults_cpu: Track the nodes the process was running on when a NUMA
1262 * hinting fault was incurred.
1263 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1264 * during the current scan window. When the scan completes, the counts
1265 * in faults_memory and faults_cpu decay and these values are copied.
1267 unsigned long *numa_faults;
1268 unsigned long total_numa_faults;
1271 * numa_faults_locality tracks if faults recorded during the last
1272 * scan window were remote/local or failed to migrate. The task scan
1273 * period is adapted based on the locality of the faults with different
1274 * weights depending on whether they were shared or private faults
1276 unsigned long numa_faults_locality[3];
1278 unsigned long numa_pages_migrated;
1279 #endif /* CONFIG_NUMA_BALANCING */
1282 struct rseq __user *rseq;
1285 * RmW on rseq_event_mask must be performed atomically
1286 * with respect to preemption.
1288 unsigned long rseq_event_mask;
1291 struct tlbflush_unmap_batch tlb_ubc;
1294 refcount_t rcu_users;
1295 struct rcu_head rcu;
1298 /* Cache last used pipe for splice(): */
1299 struct pipe_inode_info *splice_pipe;
1301 struct page_frag task_frag;
1303 #ifdef CONFIG_TASK_DELAY_ACCT
1304 struct task_delay_info *delays;
1307 #ifdef CONFIG_FAULT_INJECTION
1309 unsigned int fail_nth;
1312 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1313 * balance_dirty_pages() for a dirty throttling pause:
1316 int nr_dirtied_pause;
1317 /* Start of a write-and-pause period: */
1318 unsigned long dirty_paused_when;
1320 #ifdef CONFIG_LATENCYTOP
1321 int latency_record_count;
1322 struct latency_record latency_record[LT_SAVECOUNT];
1325 * Time slack values; these are used to round up poll() and
1326 * select() etc timeout values. These are in nanoseconds.
1329 u64 default_timer_slack_ns;
1331 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1332 unsigned int kasan_depth;
1336 struct kcsan_ctx kcsan_ctx;
1337 #ifdef CONFIG_TRACE_IRQFLAGS
1338 struct irqtrace_events kcsan_save_irqtrace;
1342 #if IS_ENABLED(CONFIG_KUNIT)
1343 struct kunit *kunit_test;
1346 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1347 /* Index of current stored address in ret_stack: */
1351 /* Stack of return addresses for return function tracing: */
1352 struct ftrace_ret_stack *ret_stack;
1354 /* Timestamp for last schedule: */
1355 unsigned long long ftrace_timestamp;
1358 * Number of functions that haven't been traced
1359 * because of depth overrun:
1361 atomic_t trace_overrun;
1363 /* Pause tracing: */
1364 atomic_t tracing_graph_pause;
1367 #ifdef CONFIG_TRACING
1368 /* State flags for use by tracers: */
1369 unsigned long trace;
1371 /* Bitmask and counter of trace recursion: */
1372 unsigned long trace_recursion;
1373 #endif /* CONFIG_TRACING */
1376 /* See kernel/kcov.c for more details. */
1378 /* Coverage collection mode enabled for this task (0 if disabled): */
1379 unsigned int kcov_mode;
1381 /* Size of the kcov_area: */
1382 unsigned int kcov_size;
1384 /* Buffer for coverage collection: */
1387 /* KCOV descriptor wired with this task or NULL: */
1390 /* KCOV common handle for remote coverage collection: */
1393 /* KCOV sequence number: */
1396 /* Collect coverage from softirq context: */
1397 unsigned int kcov_softirq;
1401 struct mem_cgroup *memcg_in_oom;
1402 gfp_t memcg_oom_gfp_mask;
1403 int memcg_oom_order;
1405 /* Number of pages to reclaim on returning to userland: */
1406 unsigned int memcg_nr_pages_over_high;
1408 /* Used by memcontrol for targeted memcg charge: */
1409 struct mem_cgroup *active_memcg;
1412 #ifdef CONFIG_BLK_CGROUP
1413 struct request_queue *throttle_queue;
1416 #ifdef CONFIG_UPROBES
1417 struct uprobe_task *utask;
1419 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1420 unsigned int sequential_io;
1421 unsigned int sequential_io_avg;
1423 struct kmap_ctrl kmap_ctrl;
1424 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1425 unsigned long task_state_change;
1426 # ifdef CONFIG_PREEMPT_RT
1427 unsigned long saved_state_change;
1430 int pagefault_disabled;
1432 struct task_struct *oom_reaper_list;
1434 #ifdef CONFIG_VMAP_STACK
1435 struct vm_struct *stack_vm_area;
1437 #ifdef CONFIG_THREAD_INFO_IN_TASK
1438 /* A live task holds one reference: */
1439 refcount_t stack_refcount;
1441 #ifdef CONFIG_LIVEPATCH
1444 #ifdef CONFIG_SECURITY
1445 /* Used by LSM modules for access restriction: */
1448 #ifdef CONFIG_BPF_SYSCALL
1449 /* Used by BPF task local storage */
1450 struct bpf_local_storage __rcu *bpf_storage;
1453 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1454 unsigned long lowest_stack;
1455 unsigned long prev_lowest_stack;
1458 #ifdef CONFIG_X86_MCE
1459 void __user *mce_vaddr;
1464 __mce_reserved : 62;
1465 struct callback_head mce_kill_me;
1468 #ifdef CONFIG_KRETPROBES
1469 struct llist_head kretprobe_instances;
1473 * New fields for task_struct should be added above here, so that
1474 * they are included in the randomized portion of task_struct.
1476 randomized_struct_fields_end
1478 /* CPU-specific state of this task: */
1479 struct thread_struct thread;
1482 * WARNING: on x86, 'thread_struct' contains a variable-sized
1483 * structure. It *MUST* be at the end of 'task_struct'.
1485 * Do not put anything below here!
1489 static inline struct pid *task_pid(struct task_struct *task)
1491 return task->thread_pid;
1495 * the helpers to get the task's different pids as they are seen
1496 * from various namespaces
1498 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1499 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1501 * task_xid_nr_ns() : id seen from the ns specified;
1503 * see also pid_nr() etc in include/linux/pid.h
1505 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1507 static inline pid_t task_pid_nr(struct task_struct *tsk)
1512 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1514 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1517 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1519 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1523 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1529 * pid_alive - check that a task structure is not stale
1530 * @p: Task structure to be checked.
1532 * Test if a process is not yet dead (at most zombie state)
1533 * If pid_alive fails, then pointers within the task structure
1534 * can be stale and must not be dereferenced.
1536 * Return: 1 if the process is alive. 0 otherwise.
1538 static inline int pid_alive(const struct task_struct *p)
1540 return p->thread_pid != NULL;
1543 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1545 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1548 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1550 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1554 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1556 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1559 static inline pid_t task_session_vnr(struct task_struct *tsk)
1561 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1564 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1566 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1569 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1571 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1574 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1580 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1586 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1588 return task_ppid_nr_ns(tsk, &init_pid_ns);
1591 /* Obsolete, do not use: */
1592 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1594 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1597 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1598 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1600 static inline unsigned int task_state_index(struct task_struct *tsk)
1602 unsigned int tsk_state = READ_ONCE(tsk->__state);
1603 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1605 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1607 if (tsk_state == TASK_IDLE)
1608 state = TASK_REPORT_IDLE;
1613 static inline char task_index_to_char(unsigned int state)
1615 static const char state_char[] = "RSDTtXZPI";
1617 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1619 return state_char[state];
1622 static inline char task_state_to_char(struct task_struct *tsk)
1624 return task_index_to_char(task_state_index(tsk));
1628 * is_global_init - check if a task structure is init. Since init
1629 * is free to have sub-threads we need to check tgid.
1630 * @tsk: Task structure to be checked.
1632 * Check if a task structure is the first user space task the kernel created.
1634 * Return: 1 if the task structure is init. 0 otherwise.
1636 static inline int is_global_init(struct task_struct *tsk)
1638 return task_tgid_nr(tsk) == 1;
1641 extern struct pid *cad_pid;
1646 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1647 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1648 #define PF_EXITING 0x00000004 /* Getting shut down */
1649 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1650 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1651 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1652 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1653 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1654 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1655 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1656 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1657 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1658 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1659 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1660 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1661 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1662 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1663 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1664 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1665 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1666 * I am cleaning dirty pages from some other bdi. */
1667 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1668 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1669 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1670 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1671 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1672 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1673 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1674 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1677 * Only the _current_ task can read/write to tsk->flags, but other
1678 * tasks can access tsk->flags in readonly mode for example
1679 * with tsk_used_math (like during threaded core dumping).
1680 * There is however an exception to this rule during ptrace
1681 * or during fork: the ptracer task is allowed to write to the
1682 * child->flags of its traced child (same goes for fork, the parent
1683 * can write to the child->flags), because we're guaranteed the
1684 * child is not running and in turn not changing child->flags
1685 * at the same time the parent does it.
1687 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1688 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1689 #define clear_used_math() clear_stopped_child_used_math(current)
1690 #define set_used_math() set_stopped_child_used_math(current)
1692 #define conditional_stopped_child_used_math(condition, child) \
1693 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1695 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1697 #define copy_to_stopped_child_used_math(child) \
1698 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1700 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1701 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1702 #define used_math() tsk_used_math(current)
1704 static inline bool is_percpu_thread(void)
1707 return (current->flags & PF_NO_SETAFFINITY) &&
1708 (current->nr_cpus_allowed == 1);
1714 /* Per-process atomic flags. */
1715 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1716 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1717 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1718 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1719 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1720 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1721 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1722 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1724 #define TASK_PFA_TEST(name, func) \
1725 static inline bool task_##func(struct task_struct *p) \
1726 { return test_bit(PFA_##name, &p->atomic_flags); }
1728 #define TASK_PFA_SET(name, func) \
1729 static inline void task_set_##func(struct task_struct *p) \
1730 { set_bit(PFA_##name, &p->atomic_flags); }
1732 #define TASK_PFA_CLEAR(name, func) \
1733 static inline void task_clear_##func(struct task_struct *p) \
1734 { clear_bit(PFA_##name, &p->atomic_flags); }
1736 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1737 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1739 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1740 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1741 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1743 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1744 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1745 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1747 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1748 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1749 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1751 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1752 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1753 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1755 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1756 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1758 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1759 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1760 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1762 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1763 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1766 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1768 current->flags &= ~flags;
1769 current->flags |= orig_flags & flags;
1772 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1773 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1775 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1776 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1778 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1781 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1783 if (!cpumask_test_cpu(0, new_mask))
1789 extern int yield_to(struct task_struct *p, bool preempt);
1790 extern void set_user_nice(struct task_struct *p, long nice);
1791 extern int task_prio(const struct task_struct *p);
1794 * task_nice - return the nice value of a given task.
1795 * @p: the task in question.
1797 * Return: The nice value [ -20 ... 0 ... 19 ].
1799 static inline int task_nice(const struct task_struct *p)
1801 return PRIO_TO_NICE((p)->static_prio);
1804 extern int can_nice(const struct task_struct *p, const int nice);
1805 extern int task_curr(const struct task_struct *p);
1806 extern int idle_cpu(int cpu);
1807 extern int available_idle_cpu(int cpu);
1808 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1809 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1810 extern void sched_set_fifo(struct task_struct *p);
1811 extern void sched_set_fifo_low(struct task_struct *p);
1812 extern void sched_set_normal(struct task_struct *p, int nice);
1813 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1814 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1815 extern struct task_struct *idle_task(int cpu);
1818 * is_idle_task - is the specified task an idle task?
1819 * @p: the task in question.
1821 * Return: 1 if @p is an idle task. 0 otherwise.
1823 static __always_inline bool is_idle_task(const struct task_struct *p)
1825 return !!(p->flags & PF_IDLE);
1828 extern struct task_struct *curr_task(int cpu);
1829 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1833 union thread_union {
1834 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1835 struct task_struct task;
1837 #ifndef CONFIG_THREAD_INFO_IN_TASK
1838 struct thread_info thread_info;
1840 unsigned long stack[THREAD_SIZE/sizeof(long)];
1843 #ifndef CONFIG_THREAD_INFO_IN_TASK
1844 extern struct thread_info init_thread_info;
1847 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1849 #ifdef CONFIG_THREAD_INFO_IN_TASK
1850 static inline struct thread_info *task_thread_info(struct task_struct *task)
1852 return &task->thread_info;
1854 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1855 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1859 * find a task by one of its numerical ids
1861 * find_task_by_pid_ns():
1862 * finds a task by its pid in the specified namespace
1863 * find_task_by_vpid():
1864 * finds a task by its virtual pid
1866 * see also find_vpid() etc in include/linux/pid.h
1869 extern struct task_struct *find_task_by_vpid(pid_t nr);
1870 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1873 * find a task by its virtual pid and get the task struct
1875 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1877 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1878 extern int wake_up_process(struct task_struct *tsk);
1879 extern void wake_up_new_task(struct task_struct *tsk);
1882 extern void kick_process(struct task_struct *tsk);
1884 static inline void kick_process(struct task_struct *tsk) { }
1887 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1889 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1891 __set_task_comm(tsk, from, false);
1894 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1895 #define get_task_comm(buf, tsk) ({ \
1896 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1897 __get_task_comm(buf, sizeof(buf), tsk); \
1901 static __always_inline void scheduler_ipi(void)
1904 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1905 * TIF_NEED_RESCHED remotely (for the first time) will also send
1908 preempt_fold_need_resched();
1910 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1912 static inline void scheduler_ipi(void) { }
1913 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1920 * Set thread flags in other task's structures.
1921 * See asm/thread_info.h for TIF_xxxx flags available:
1923 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1925 set_ti_thread_flag(task_thread_info(tsk), flag);
1928 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1930 clear_ti_thread_flag(task_thread_info(tsk), flag);
1933 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1936 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1939 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1941 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1944 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1946 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1949 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1951 return test_ti_thread_flag(task_thread_info(tsk), flag);
1954 static inline void set_tsk_need_resched(struct task_struct *tsk)
1956 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1959 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1961 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1964 static inline int test_tsk_need_resched(struct task_struct *tsk)
1966 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1970 * cond_resched() and cond_resched_lock(): latency reduction via
1971 * explicit rescheduling in places that are safe. The return
1972 * value indicates whether a reschedule was done in fact.
1973 * cond_resched_lock() will drop the spinlock before scheduling,
1975 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1976 extern int __cond_resched(void);
1978 #ifdef CONFIG_PREEMPT_DYNAMIC
1980 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1982 static __always_inline int _cond_resched(void)
1984 return static_call_mod(cond_resched)();
1989 static inline int _cond_resched(void)
1991 return __cond_resched();
1994 #endif /* CONFIG_PREEMPT_DYNAMIC */
1998 static inline int _cond_resched(void) { return 0; }
2000 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2002 #define cond_resched() ({ \
2003 ___might_sleep(__FILE__, __LINE__, 0); \
2007 extern int __cond_resched_lock(spinlock_t *lock);
2008 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2009 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2011 #define cond_resched_lock(lock) ({ \
2012 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
2013 __cond_resched_lock(lock); \
2016 #define cond_resched_rwlock_read(lock) ({ \
2017 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2018 __cond_resched_rwlock_read(lock); \
2021 #define cond_resched_rwlock_write(lock) ({ \
2022 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2023 __cond_resched_rwlock_write(lock); \
2026 static inline void cond_resched_rcu(void)
2028 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2036 * Does a critical section need to be broken due to another
2037 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2038 * but a general need for low latency)
2040 static inline int spin_needbreak(spinlock_t *lock)
2042 #ifdef CONFIG_PREEMPTION
2043 return spin_is_contended(lock);
2050 * Check if a rwlock is contended.
2051 * Returns non-zero if there is another task waiting on the rwlock.
2052 * Returns zero if the lock is not contended or the system / underlying
2053 * rwlock implementation does not support contention detection.
2054 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2057 static inline int rwlock_needbreak(rwlock_t *lock)
2059 #ifdef CONFIG_PREEMPTION
2060 return rwlock_is_contended(lock);
2066 static __always_inline bool need_resched(void)
2068 return unlikely(tif_need_resched());
2072 * Wrappers for p->thread_info->cpu access. No-op on UP.
2076 static inline unsigned int task_cpu(const struct task_struct *p)
2078 #ifdef CONFIG_THREAD_INFO_IN_TASK
2079 return READ_ONCE(p->cpu);
2081 return READ_ONCE(task_thread_info(p)->cpu);
2085 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2089 static inline unsigned int task_cpu(const struct task_struct *p)
2094 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2098 #endif /* CONFIG_SMP */
2100 extern bool sched_task_on_rq(struct task_struct *p);
2103 * In order to reduce various lock holder preemption latencies provide an
2104 * interface to see if a vCPU is currently running or not.
2106 * This allows us to terminate optimistic spin loops and block, analogous to
2107 * the native optimistic spin heuristic of testing if the lock owner task is
2110 #ifndef vcpu_is_preempted
2111 static inline bool vcpu_is_preempted(int cpu)
2117 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2118 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2120 #ifndef TASK_SIZE_OF
2121 #define TASK_SIZE_OF(tsk) TASK_SIZE
2125 /* Returns effective CPU energy utilization, as seen by the scheduler */
2126 unsigned long sched_cpu_util(int cpu, unsigned long max);
2127 #endif /* CONFIG_SMP */
2132 * Map the event mask on the user-space ABI enum rseq_cs_flags
2133 * for direct mask checks.
2135 enum rseq_event_mask_bits {
2136 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2137 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2138 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2141 enum rseq_event_mask {
2142 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2143 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2144 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2147 static inline void rseq_set_notify_resume(struct task_struct *t)
2150 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2153 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2155 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2156 struct pt_regs *regs)
2159 __rseq_handle_notify_resume(ksig, regs);
2162 static inline void rseq_signal_deliver(struct ksignal *ksig,
2163 struct pt_regs *regs)
2166 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2168 rseq_handle_notify_resume(ksig, regs);
2171 /* rseq_preempt() requires preemption to be disabled. */
2172 static inline void rseq_preempt(struct task_struct *t)
2174 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2175 rseq_set_notify_resume(t);
2178 /* rseq_migrate() requires preemption to be disabled. */
2179 static inline void rseq_migrate(struct task_struct *t)
2181 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2182 rseq_set_notify_resume(t);
2186 * If parent process has a registered restartable sequences area, the
2187 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2189 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2191 if (clone_flags & CLONE_VM) {
2194 t->rseq_event_mask = 0;
2196 t->rseq = current->rseq;
2197 t->rseq_sig = current->rseq_sig;
2198 t->rseq_event_mask = current->rseq_event_mask;
2202 static inline void rseq_execve(struct task_struct *t)
2206 t->rseq_event_mask = 0;
2211 static inline void rseq_set_notify_resume(struct task_struct *t)
2214 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2215 struct pt_regs *regs)
2218 static inline void rseq_signal_deliver(struct ksignal *ksig,
2219 struct pt_regs *regs)
2222 static inline void rseq_preempt(struct task_struct *t)
2225 static inline void rseq_migrate(struct task_struct *t)
2228 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2231 static inline void rseq_execve(struct task_struct *t)
2237 #ifdef CONFIG_DEBUG_RSEQ
2239 void rseq_syscall(struct pt_regs *regs);
2243 static inline void rseq_syscall(struct pt_regs *regs)
2249 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2250 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2251 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2253 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2254 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2255 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2257 int sched_trace_rq_cpu(struct rq *rq);
2258 int sched_trace_rq_cpu_capacity(struct rq *rq);
2259 int sched_trace_rq_nr_running(struct rq *rq);
2261 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2263 #ifdef CONFIG_SCHED_CORE
2264 extern void sched_core_free(struct task_struct *tsk);
2265 extern void sched_core_fork(struct task_struct *p);
2266 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2267 unsigned long uaddr);
2269 static inline void sched_core_free(struct task_struct *tsk) { }
2270 static inline void sched_core_fork(struct task_struct *p) { }