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;
46 struct capture_control;
49 struct futex_pi_state;
55 struct perf_event_context;
57 struct pipe_inode_info;
60 struct robust_list_head;
66 struct sighand_struct;
68 struct task_delay_info;
72 * Task state bitmask. NOTE! These bits are also
73 * encoded in fs/proc/array.c: get_task_state().
75 * We have two separate sets of flags: task->state
76 * is about runnability, while task->exit_state are
77 * about the task exiting. Confusing, but this way
78 * modifying one set can't modify the other one by
82 /* Used in tsk->state: */
83 #define TASK_RUNNING 0x0000
84 #define TASK_INTERRUPTIBLE 0x0001
85 #define TASK_UNINTERRUPTIBLE 0x0002
86 #define __TASK_STOPPED 0x0004
87 #define __TASK_TRACED 0x0008
88 /* Used in tsk->exit_state: */
89 #define EXIT_DEAD 0x0010
90 #define EXIT_ZOMBIE 0x0020
91 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
92 /* Used in tsk->state again: */
93 #define TASK_PARKED 0x0040
94 #define TASK_DEAD 0x0080
95 #define TASK_WAKEKILL 0x0100
96 #define TASK_WAKING 0x0200
97 #define TASK_NOLOAD 0x0400
98 #define TASK_NEW 0x0800
99 /* RT specific auxilliary flag to mark RT lock waiters */
100 #define TASK_RTLOCK_WAIT 0x1000
101 #define TASK_STATE_MAX 0x2000
103 /* Convenience macros for the sake of set_current_state: */
104 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
105 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
106 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
108 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
110 /* Convenience macros for the sake of wake_up(): */
111 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
113 /* get_task_state(): */
114 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
115 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
116 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
119 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
121 #define task_is_traced(task) ((READ_ONCE(task->__state) & __TASK_TRACED) != 0)
123 #define task_is_stopped(task) ((READ_ONCE(task->__state) & __TASK_STOPPED) != 0)
125 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->__state) & (__TASK_STOPPED | __TASK_TRACED)) != 0)
128 * Special states are those that do not use the normal wait-loop pattern. See
129 * the comment with set_special_state().
131 #define is_special_task_state(state) \
132 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
134 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
135 # define debug_normal_state_change(state_value) \
137 WARN_ON_ONCE(is_special_task_state(state_value)); \
138 current->task_state_change = _THIS_IP_; \
141 # define debug_special_state_change(state_value) \
143 WARN_ON_ONCE(!is_special_task_state(state_value)); \
144 current->task_state_change = _THIS_IP_; \
147 # define debug_rtlock_wait_set_state() \
149 current->saved_state_change = current->task_state_change;\
150 current->task_state_change = _THIS_IP_; \
153 # define debug_rtlock_wait_restore_state() \
155 current->task_state_change = current->saved_state_change;\
159 # define debug_normal_state_change(cond) do { } while (0)
160 # define debug_special_state_change(cond) do { } while (0)
161 # define debug_rtlock_wait_set_state() do { } while (0)
162 # define debug_rtlock_wait_restore_state() do { } while (0)
166 * set_current_state() includes a barrier so that the write of current->state
167 * is correctly serialised wrt the caller's subsequent test of whether to
171 * set_current_state(TASK_UNINTERRUPTIBLE);
177 * __set_current_state(TASK_RUNNING);
179 * If the caller does not need such serialisation (because, for instance, the
180 * CONDITION test and condition change and wakeup are under the same lock) then
181 * use __set_current_state().
183 * The above is typically ordered against the wakeup, which does:
186 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
188 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
189 * accessing p->state.
191 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
192 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
193 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
195 * However, with slightly different timing the wakeup TASK_RUNNING store can
196 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
197 * a problem either because that will result in one extra go around the loop
198 * and our @cond test will save the day.
200 * Also see the comments of try_to_wake_up().
202 #define __set_current_state(state_value) \
204 debug_normal_state_change((state_value)); \
205 WRITE_ONCE(current->__state, (state_value)); \
208 #define set_current_state(state_value) \
210 debug_normal_state_change((state_value)); \
211 smp_store_mb(current->__state, (state_value)); \
215 * set_special_state() should be used for those states when the blocking task
216 * can not use the regular condition based wait-loop. In that case we must
217 * serialize against wakeups such that any possible in-flight TASK_RUNNING
218 * stores will not collide with our state change.
220 #define set_special_state(state_value) \
222 unsigned long flags; /* may shadow */ \
224 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
225 debug_special_state_change((state_value)); \
226 WRITE_ONCE(current->__state, (state_value)); \
227 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
231 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
233 * RT's spin/rwlock substitutions are state preserving. The state of the
234 * task when blocking on the lock is saved in task_struct::saved_state and
235 * restored after the lock has been acquired. These operations are
236 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
237 * lock related wakeups while the task is blocked on the lock are
238 * redirected to operate on task_struct::saved_state to ensure that these
239 * are not dropped. On restore task_struct::saved_state is set to
240 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
242 * The lock operation looks like this:
244 * current_save_and_set_rtlock_wait_state();
248 * raw_spin_unlock_irq(&lock->wait_lock);
250 * raw_spin_lock_irq(&lock->wait_lock);
251 * set_current_state(TASK_RTLOCK_WAIT);
253 * current_restore_rtlock_saved_state();
255 #define current_save_and_set_rtlock_wait_state() \
257 lockdep_assert_irqs_disabled(); \
258 raw_spin_lock(¤t->pi_lock); \
259 current->saved_state = current->__state; \
260 debug_rtlock_wait_set_state(); \
261 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
262 raw_spin_unlock(¤t->pi_lock); \
265 #define current_restore_rtlock_saved_state() \
267 lockdep_assert_irqs_disabled(); \
268 raw_spin_lock(¤t->pi_lock); \
269 debug_rtlock_wait_restore_state(); \
270 WRITE_ONCE(current->__state, current->saved_state); \
271 current->saved_state = TASK_RUNNING; \
272 raw_spin_unlock(¤t->pi_lock); \
275 #define get_current_state() READ_ONCE(current->__state)
277 /* Task command name length: */
278 #define TASK_COMM_LEN 16
280 extern void scheduler_tick(void);
282 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
284 extern long schedule_timeout(long timeout);
285 extern long schedule_timeout_interruptible(long timeout);
286 extern long schedule_timeout_killable(long timeout);
287 extern long schedule_timeout_uninterruptible(long timeout);
288 extern long schedule_timeout_idle(long timeout);
289 asmlinkage void schedule(void);
290 extern void schedule_preempt_disabled(void);
291 asmlinkage void preempt_schedule_irq(void);
292 #ifdef CONFIG_PREEMPT_RT
293 extern void schedule_rtlock(void);
296 extern int __must_check io_schedule_prepare(void);
297 extern void io_schedule_finish(int token);
298 extern long io_schedule_timeout(long timeout);
299 extern void io_schedule(void);
302 * struct prev_cputime - snapshot of system and user cputime
303 * @utime: time spent in user mode
304 * @stime: time spent in system mode
305 * @lock: protects the above two fields
307 * Stores previous user/system time values such that we can guarantee
310 struct prev_cputime {
311 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
319 /* Task is sleeping or running in a CPU with VTIME inactive: */
323 /* Task runs in kernelspace in a CPU with VTIME active: */
325 /* Task runs in userspace in a CPU with VTIME active: */
327 /* Task runs as guests in a CPU with VTIME active: */
333 unsigned long long starttime;
334 enum vtime_state state;
342 * Utilization clamp constraints.
343 * @UCLAMP_MIN: Minimum utilization
344 * @UCLAMP_MAX: Maximum utilization
345 * @UCLAMP_CNT: Utilization clamp constraints count
354 extern struct root_domain def_root_domain;
355 extern struct mutex sched_domains_mutex;
359 #ifdef CONFIG_SCHED_INFO
360 /* Cumulative counters: */
362 /* # of times we have run on this CPU: */
363 unsigned long pcount;
365 /* Time spent waiting on a runqueue: */
366 unsigned long long run_delay;
370 /* When did we last run on a CPU? */
371 unsigned long long last_arrival;
373 /* When were we last queued to run? */
374 unsigned long long last_queued;
376 #endif /* CONFIG_SCHED_INFO */
380 * Integer metrics need fixed point arithmetic, e.g., sched/fair
381 * has a few: load, load_avg, util_avg, freq, and capacity.
383 * We define a basic fixed point arithmetic range, and then formalize
384 * all these metrics based on that basic range.
386 # define SCHED_FIXEDPOINT_SHIFT 10
387 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
389 /* Increase resolution of cpu_capacity calculations */
390 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
391 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
394 unsigned long weight;
399 * struct util_est - Estimation utilization of FAIR tasks
400 * @enqueued: instantaneous estimated utilization of a task/cpu
401 * @ewma: the Exponential Weighted Moving Average (EWMA)
402 * utilization of a task
404 * Support data structure to track an Exponential Weighted Moving Average
405 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
406 * average each time a task completes an activation. Sample's weight is chosen
407 * so that the EWMA will be relatively insensitive to transient changes to the
410 * The enqueued attribute has a slightly different meaning for tasks and cpus:
411 * - task: the task's util_avg at last task dequeue time
412 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
413 * Thus, the util_est.enqueued of a task represents the contribution on the
414 * estimated utilization of the CPU where that task is currently enqueued.
416 * Only for tasks we track a moving average of the past instantaneous
417 * estimated utilization. This allows to absorb sporadic drops in utilization
418 * of an otherwise almost periodic task.
420 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
421 * updates. When a task is dequeued, its util_est should not be updated if its
422 * util_avg has not been updated in the meantime.
423 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
424 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
425 * for a task) it is safe to use MSB.
428 unsigned int enqueued;
430 #define UTIL_EST_WEIGHT_SHIFT 2
431 #define UTIL_AVG_UNCHANGED 0x80000000
432 } __attribute__((__aligned__(sizeof(u64))));
435 * The load/runnable/util_avg accumulates an infinite geometric series
436 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
438 * [load_avg definition]
440 * load_avg = runnable% * scale_load_down(load)
442 * [runnable_avg definition]
444 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
446 * [util_avg definition]
448 * util_avg = running% * SCHED_CAPACITY_SCALE
450 * where runnable% is the time ratio that a sched_entity is runnable and
451 * running% the time ratio that a sched_entity is running.
453 * For cfs_rq, they are the aggregated values of all runnable and blocked
456 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
457 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
458 * for computing those signals (see update_rq_clock_pelt())
460 * N.B., the above ratios (runnable% and running%) themselves are in the
461 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
462 * to as large a range as necessary. This is for example reflected by
463 * util_avg's SCHED_CAPACITY_SCALE.
467 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
468 * with the highest load (=88761), always runnable on a single cfs_rq,
469 * and should not overflow as the number already hits PID_MAX_LIMIT.
471 * For all other cases (including 32-bit kernels), struct load_weight's
472 * weight will overflow first before we do, because:
474 * Max(load_avg) <= Max(load.weight)
476 * Then it is the load_weight's responsibility to consider overflow
480 u64 last_update_time;
485 unsigned long load_avg;
486 unsigned long runnable_avg;
487 unsigned long util_avg;
488 struct util_est util_est;
489 } ____cacheline_aligned;
491 struct sched_statistics {
492 #ifdef CONFIG_SCHEDSTATS
502 s64 sum_sleep_runtime;
506 s64 sum_block_runtime;
511 u64 nr_migrations_cold;
512 u64 nr_failed_migrations_affine;
513 u64 nr_failed_migrations_running;
514 u64 nr_failed_migrations_hot;
515 u64 nr_forced_migrations;
519 u64 nr_wakeups_migrate;
520 u64 nr_wakeups_local;
521 u64 nr_wakeups_remote;
522 u64 nr_wakeups_affine;
523 u64 nr_wakeups_affine_attempts;
524 u64 nr_wakeups_passive;
527 } ____cacheline_aligned;
529 struct sched_entity {
530 /* For load-balancing: */
531 struct load_weight load;
532 struct rb_node run_node;
533 struct list_head group_node;
537 u64 sum_exec_runtime;
539 u64 prev_sum_exec_runtime;
543 #ifdef CONFIG_FAIR_GROUP_SCHED
545 struct sched_entity *parent;
546 /* rq on which this entity is (to be) queued: */
547 struct cfs_rq *cfs_rq;
548 /* rq "owned" by this entity/group: */
550 /* cached value of my_q->h_nr_running */
551 unsigned long runnable_weight;
556 * Per entity load average tracking.
558 * Put into separate cache line so it does not
559 * collide with read-mostly values above.
561 struct sched_avg avg;
565 struct sched_rt_entity {
566 struct list_head run_list;
567 unsigned long timeout;
568 unsigned long watchdog_stamp;
569 unsigned int time_slice;
570 unsigned short on_rq;
571 unsigned short on_list;
573 struct sched_rt_entity *back;
574 #ifdef CONFIG_RT_GROUP_SCHED
575 struct sched_rt_entity *parent;
576 /* rq on which this entity is (to be) queued: */
578 /* rq "owned" by this entity/group: */
581 } __randomize_layout;
583 struct sched_dl_entity {
584 struct rb_node rb_node;
587 * Original scheduling parameters. Copied here from sched_attr
588 * during sched_setattr(), they will remain the same until
589 * the next sched_setattr().
591 u64 dl_runtime; /* Maximum runtime for each instance */
592 u64 dl_deadline; /* Relative deadline of each instance */
593 u64 dl_period; /* Separation of two instances (period) */
594 u64 dl_bw; /* dl_runtime / dl_period */
595 u64 dl_density; /* dl_runtime / dl_deadline */
598 * Actual scheduling parameters. Initialized with the values above,
599 * they are continuously updated during task execution. Note that
600 * the remaining runtime could be < 0 in case we are in overrun.
602 s64 runtime; /* Remaining runtime for this instance */
603 u64 deadline; /* Absolute deadline for this instance */
604 unsigned int flags; /* Specifying the scheduler behaviour */
609 * @dl_throttled tells if we exhausted the runtime. If so, the
610 * task has to wait for a replenishment to be performed at the
611 * next firing of dl_timer.
613 * @dl_boosted tells if we are boosted due to DI. If so we are
614 * outside bandwidth enforcement mechanism (but only until we
615 * exit the critical section);
617 * @dl_yielded tells if task gave up the CPU before consuming
618 * all its available runtime during the last job.
620 * @dl_non_contending tells if the task is inactive while still
621 * contributing to the active utilization. In other words, it
622 * indicates if the inactive timer has been armed and its handler
623 * has not been executed yet. This flag is useful to avoid race
624 * conditions between the inactive timer handler and the wakeup
627 * @dl_overrun tells if the task asked to be informed about runtime
630 unsigned int dl_throttled : 1;
631 unsigned int dl_yielded : 1;
632 unsigned int dl_non_contending : 1;
633 unsigned int dl_overrun : 1;
636 * Bandwidth enforcement timer. Each -deadline task has its
637 * own bandwidth to be enforced, thus we need one timer per task.
639 struct hrtimer dl_timer;
642 * Inactive timer, responsible for decreasing the active utilization
643 * at the "0-lag time". When a -deadline task blocks, it contributes
644 * to GRUB's active utilization until the "0-lag time", hence a
645 * timer is needed to decrease the active utilization at the correct
648 struct hrtimer inactive_timer;
650 #ifdef CONFIG_RT_MUTEXES
652 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
653 * pi_se points to the donor, otherwise points to the dl_se it belongs
654 * to (the original one/itself).
656 struct sched_dl_entity *pi_se;
660 #ifdef CONFIG_UCLAMP_TASK
661 /* Number of utilization clamp buckets (shorter alias) */
662 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
665 * Utilization clamp for a scheduling entity
666 * @value: clamp value "assigned" to a se
667 * @bucket_id: bucket index corresponding to the "assigned" value
668 * @active: the se is currently refcounted in a rq's bucket
669 * @user_defined: the requested clamp value comes from user-space
671 * The bucket_id is the index of the clamp bucket matching the clamp value
672 * which is pre-computed and stored to avoid expensive integer divisions from
675 * The active bit is set whenever a task has got an "effective" value assigned,
676 * which can be different from the clamp value "requested" from user-space.
677 * This allows to know a task is refcounted in the rq's bucket corresponding
678 * to the "effective" bucket_id.
680 * The user_defined bit is set whenever a task has got a task-specific clamp
681 * value requested from userspace, i.e. the system defaults apply to this task
682 * just as a restriction. This allows to relax default clamps when a less
683 * restrictive task-specific value has been requested, thus allowing to
684 * implement a "nice" semantic. For example, a task running with a 20%
685 * default boost can still drop its own boosting to 0%.
688 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
689 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
690 unsigned int active : 1;
691 unsigned int user_defined : 1;
693 #endif /* CONFIG_UCLAMP_TASK */
699 u8 exp_hint; /* Hint for performance. */
700 u8 need_mb; /* Readers need smp_mb(). */
702 u32 s; /* Set of bits. */
705 enum perf_event_task_context {
706 perf_invalid_context = -1,
709 perf_nr_task_contexts,
713 struct wake_q_node *next;
717 #ifdef CONFIG_KMAP_LOCAL
719 pte_t pteval[KM_MAX_IDX];
724 #ifdef CONFIG_THREAD_INFO_IN_TASK
726 * For reasons of header soup (see current_thread_info()), this
727 * must be the first element of task_struct.
729 struct thread_info thread_info;
731 unsigned int __state;
733 #ifdef CONFIG_PREEMPT_RT
734 /* saved state for "spinlock sleepers" */
735 unsigned int saved_state;
739 * This begins the randomizable portion of task_struct. Only
740 * scheduling-critical items should be added above here.
742 randomized_struct_fields_start
746 /* Per task flags (PF_*), defined further below: */
752 struct __call_single_node wake_entry;
753 unsigned int wakee_flips;
754 unsigned long wakee_flip_decay_ts;
755 struct task_struct *last_wakee;
758 * recent_used_cpu is initially set as the last CPU used by a task
759 * that wakes affine another task. Waker/wakee relationships can
760 * push tasks around a CPU where each wakeup moves to the next one.
761 * Tracking a recently used CPU allows a quick search for a recently
762 * used CPU that may be idle.
772 unsigned int rt_priority;
774 struct sched_entity se;
775 struct sched_rt_entity rt;
776 struct sched_dl_entity dl;
777 const struct sched_class *sched_class;
779 #ifdef CONFIG_SCHED_CORE
780 struct rb_node core_node;
781 unsigned long core_cookie;
782 unsigned int core_occupation;
785 #ifdef CONFIG_CGROUP_SCHED
786 struct task_group *sched_task_group;
789 #ifdef CONFIG_UCLAMP_TASK
791 * Clamp values requested for a scheduling entity.
792 * Must be updated with task_rq_lock() held.
794 struct uclamp_se uclamp_req[UCLAMP_CNT];
796 * Effective clamp values used for a scheduling entity.
797 * Must be updated with task_rq_lock() held.
799 struct uclamp_se uclamp[UCLAMP_CNT];
802 struct sched_statistics stats;
804 #ifdef CONFIG_PREEMPT_NOTIFIERS
805 /* List of struct preempt_notifier: */
806 struct hlist_head preempt_notifiers;
809 #ifdef CONFIG_BLK_DEV_IO_TRACE
810 unsigned int btrace_seq;
815 const cpumask_t *cpus_ptr;
816 cpumask_t *user_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;
932 #ifdef CONFIG_EVENTFD
933 /* Recursion prevention for eventfd_signal() */
934 unsigned in_eventfd_signal:1;
937 unsigned long atomic_flags; /* Flags requiring atomic access. */
939 struct restart_block restart_block;
944 #ifdef CONFIG_STACKPROTECTOR
945 /* Canary value for the -fstack-protector GCC feature: */
946 unsigned long stack_canary;
949 * Pointers to the (original) parent process, youngest child, younger sibling,
950 * older sibling, respectively. (p->father can be replaced with
951 * p->real_parent->pid)
954 /* Real parent process: */
955 struct task_struct __rcu *real_parent;
957 /* Recipient of SIGCHLD, wait4() reports: */
958 struct task_struct __rcu *parent;
961 * Children/sibling form the list of natural children:
963 struct list_head children;
964 struct list_head sibling;
965 struct task_struct *group_leader;
968 * 'ptraced' is the list of tasks this task is using ptrace() on.
970 * This includes both natural children and PTRACE_ATTACH targets.
971 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
973 struct list_head ptraced;
974 struct list_head ptrace_entry;
976 /* PID/PID hash table linkage. */
977 struct pid *thread_pid;
978 struct hlist_node pid_links[PIDTYPE_MAX];
979 struct list_head thread_group;
980 struct list_head thread_node;
982 struct completion *vfork_done;
984 /* CLONE_CHILD_SETTID: */
985 int __user *set_child_tid;
987 /* CLONE_CHILD_CLEARTID: */
988 int __user *clear_child_tid;
995 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1000 struct prev_cputime prev_cputime;
1001 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1005 #ifdef CONFIG_NO_HZ_FULL
1006 atomic_t tick_dep_mask;
1008 /* Context switch counts: */
1009 unsigned long nvcsw;
1010 unsigned long nivcsw;
1012 /* Monotonic time in nsecs: */
1015 /* Boot based time in nsecs: */
1018 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1019 unsigned long min_flt;
1020 unsigned long maj_flt;
1022 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1023 struct posix_cputimers posix_cputimers;
1025 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1026 struct posix_cputimers_work posix_cputimers_work;
1029 /* Process credentials: */
1031 /* Tracer's credentials at attach: */
1032 const struct cred __rcu *ptracer_cred;
1034 /* Objective and real subjective task credentials (COW): */
1035 const struct cred __rcu *real_cred;
1037 /* Effective (overridable) subjective task credentials (COW): */
1038 const struct cred __rcu *cred;
1041 /* Cached requested key. */
1042 struct key *cached_requested_key;
1046 * executable name, excluding path.
1048 * - normally initialized setup_new_exec()
1049 * - access it with [gs]et_task_comm()
1050 * - lock it with task_lock()
1052 char comm[TASK_COMM_LEN];
1054 struct nameidata *nameidata;
1056 #ifdef CONFIG_SYSVIPC
1057 struct sysv_sem sysvsem;
1058 struct sysv_shm sysvshm;
1060 #ifdef CONFIG_DETECT_HUNG_TASK
1061 unsigned long last_switch_count;
1062 unsigned long last_switch_time;
1064 /* Filesystem information: */
1065 struct fs_struct *fs;
1067 /* Open file information: */
1068 struct files_struct *files;
1070 #ifdef CONFIG_IO_URING
1071 struct io_uring_task *io_uring;
1075 struct nsproxy *nsproxy;
1077 /* Signal handlers: */
1078 struct signal_struct *signal;
1079 struct sighand_struct __rcu *sighand;
1081 sigset_t real_blocked;
1082 /* Restored if set_restore_sigmask() was used: */
1083 sigset_t saved_sigmask;
1084 struct sigpending pending;
1085 unsigned long sas_ss_sp;
1087 unsigned int sas_ss_flags;
1089 struct callback_head *task_works;
1092 #ifdef CONFIG_AUDITSYSCALL
1093 struct audit_context *audit_context;
1096 unsigned int sessionid;
1098 struct seccomp seccomp;
1099 struct syscall_user_dispatch syscall_dispatch;
1101 /* Thread group tracking: */
1105 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1106 spinlock_t alloc_lock;
1108 /* Protection of the PI data structures: */
1109 raw_spinlock_t pi_lock;
1111 struct wake_q_node wake_q;
1113 #ifdef CONFIG_RT_MUTEXES
1114 /* PI waiters blocked on a rt_mutex held by this task: */
1115 struct rb_root_cached pi_waiters;
1116 /* Updated under owner's pi_lock and rq lock */
1117 struct task_struct *pi_top_task;
1118 /* Deadlock detection and priority inheritance handling: */
1119 struct rt_mutex_waiter *pi_blocked_on;
1122 #ifdef CONFIG_DEBUG_MUTEXES
1123 /* Mutex deadlock detection: */
1124 struct mutex_waiter *blocked_on;
1127 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1128 int non_block_count;
1131 #ifdef CONFIG_TRACE_IRQFLAGS
1132 struct irqtrace_events irqtrace;
1133 unsigned int hardirq_threaded;
1134 u64 hardirq_chain_key;
1135 int softirqs_enabled;
1136 int softirq_context;
1139 #ifdef CONFIG_PREEMPT_RT
1140 int softirq_disable_cnt;
1143 #ifdef CONFIG_LOCKDEP
1144 # define MAX_LOCK_DEPTH 48UL
1147 unsigned int lockdep_recursion;
1148 struct held_lock held_locks[MAX_LOCK_DEPTH];
1151 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1152 unsigned int in_ubsan;
1155 /* Journalling filesystem info: */
1158 /* Stacked block device info: */
1159 struct bio_list *bio_list;
1161 /* Stack plugging: */
1162 struct blk_plug *plug;
1165 struct reclaim_state *reclaim_state;
1167 struct backing_dev_info *backing_dev_info;
1169 struct io_context *io_context;
1171 #ifdef CONFIG_COMPACTION
1172 struct capture_control *capture_control;
1175 unsigned long ptrace_message;
1176 kernel_siginfo_t *last_siginfo;
1178 struct task_io_accounting ioac;
1180 /* Pressure stall state */
1181 unsigned int psi_flags;
1183 #ifdef CONFIG_TASK_XACCT
1184 /* Accumulated RSS usage: */
1186 /* Accumulated virtual memory usage: */
1188 /* stime + utime since last update: */
1191 #ifdef CONFIG_CPUSETS
1192 /* Protected by ->alloc_lock: */
1193 nodemask_t mems_allowed;
1194 /* Sequence number to catch updates: */
1195 seqcount_spinlock_t mems_allowed_seq;
1196 int cpuset_mem_spread_rotor;
1197 int cpuset_slab_spread_rotor;
1199 #ifdef CONFIG_CGROUPS
1200 /* Control Group info protected by css_set_lock: */
1201 struct css_set __rcu *cgroups;
1202 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1203 struct list_head cg_list;
1205 #ifdef CONFIG_X86_CPU_RESCTRL
1210 struct robust_list_head __user *robust_list;
1211 #ifdef CONFIG_COMPAT
1212 struct compat_robust_list_head __user *compat_robust_list;
1214 struct list_head pi_state_list;
1215 struct futex_pi_state *pi_state_cache;
1216 struct mutex futex_exit_mutex;
1217 unsigned int futex_state;
1219 #ifdef CONFIG_PERF_EVENTS
1220 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1221 struct mutex perf_event_mutex;
1222 struct list_head perf_event_list;
1224 #ifdef CONFIG_DEBUG_PREEMPT
1225 unsigned long preempt_disable_ip;
1228 /* Protected by alloc_lock: */
1229 struct mempolicy *mempolicy;
1231 short pref_node_fork;
1233 #ifdef CONFIG_NUMA_BALANCING
1235 unsigned int numa_scan_period;
1236 unsigned int numa_scan_period_max;
1237 int numa_preferred_nid;
1238 unsigned long numa_migrate_retry;
1239 /* Migration stamp: */
1241 u64 last_task_numa_placement;
1242 u64 last_sum_exec_runtime;
1243 struct callback_head numa_work;
1246 * This pointer is only modified for current in syscall and
1247 * pagefault context (and for tasks being destroyed), so it can be read
1248 * from any of the following contexts:
1249 * - RCU read-side critical section
1250 * - current->numa_group from everywhere
1251 * - task's runqueue locked, task not running
1253 struct numa_group __rcu *numa_group;
1256 * numa_faults is an array split into four regions:
1257 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1258 * in this precise order.
1260 * faults_memory: Exponential decaying average of faults on a per-node
1261 * basis. Scheduling placement decisions are made based on these
1262 * counts. The values remain static for the duration of a PTE scan.
1263 * faults_cpu: Track the nodes the process was running on when a NUMA
1264 * hinting fault was incurred.
1265 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1266 * during the current scan window. When the scan completes, the counts
1267 * in faults_memory and faults_cpu decay and these values are copied.
1269 unsigned long *numa_faults;
1270 unsigned long total_numa_faults;
1273 * numa_faults_locality tracks if faults recorded during the last
1274 * scan window were remote/local or failed to migrate. The task scan
1275 * period is adapted based on the locality of the faults with different
1276 * weights depending on whether they were shared or private faults
1278 unsigned long numa_faults_locality[3];
1280 unsigned long numa_pages_migrated;
1281 #endif /* CONFIG_NUMA_BALANCING */
1284 struct rseq __user *rseq;
1287 * RmW on rseq_event_mask must be performed atomically
1288 * with respect to preemption.
1290 unsigned long rseq_event_mask;
1293 struct tlbflush_unmap_batch tlb_ubc;
1296 refcount_t rcu_users;
1297 struct rcu_head rcu;
1300 /* Cache last used pipe for splice(): */
1301 struct pipe_inode_info *splice_pipe;
1303 struct page_frag task_frag;
1305 #ifdef CONFIG_TASK_DELAY_ACCT
1306 struct task_delay_info *delays;
1309 #ifdef CONFIG_FAULT_INJECTION
1311 unsigned int fail_nth;
1314 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1315 * balance_dirty_pages() for a dirty throttling pause:
1318 int nr_dirtied_pause;
1319 /* Start of a write-and-pause period: */
1320 unsigned long dirty_paused_when;
1322 #ifdef CONFIG_LATENCYTOP
1323 int latency_record_count;
1324 struct latency_record latency_record[LT_SAVECOUNT];
1327 * Time slack values; these are used to round up poll() and
1328 * select() etc timeout values. These are in nanoseconds.
1331 u64 default_timer_slack_ns;
1333 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1334 unsigned int kasan_depth;
1338 struct kcsan_ctx kcsan_ctx;
1339 #ifdef CONFIG_TRACE_IRQFLAGS
1340 struct irqtrace_events kcsan_save_irqtrace;
1344 #if IS_ENABLED(CONFIG_KUNIT)
1345 struct kunit *kunit_test;
1348 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1349 /* Index of current stored address in ret_stack: */
1353 /* Stack of return addresses for return function tracing: */
1354 struct ftrace_ret_stack *ret_stack;
1356 /* Timestamp for last schedule: */
1357 unsigned long long ftrace_timestamp;
1360 * Number of functions that haven't been traced
1361 * because of depth overrun:
1363 atomic_t trace_overrun;
1365 /* Pause tracing: */
1366 atomic_t tracing_graph_pause;
1369 #ifdef CONFIG_TRACING
1370 /* State flags for use by tracers: */
1371 unsigned long trace;
1373 /* Bitmask and counter of trace recursion: */
1374 unsigned long trace_recursion;
1375 #endif /* CONFIG_TRACING */
1378 /* See kernel/kcov.c for more details. */
1380 /* Coverage collection mode enabled for this task (0 if disabled): */
1381 unsigned int kcov_mode;
1383 /* Size of the kcov_area: */
1384 unsigned int kcov_size;
1386 /* Buffer for coverage collection: */
1389 /* KCOV descriptor wired with this task or NULL: */
1392 /* KCOV common handle for remote coverage collection: */
1395 /* KCOV sequence number: */
1398 /* Collect coverage from softirq context: */
1399 unsigned int kcov_softirq;
1403 struct mem_cgroup *memcg_in_oom;
1404 gfp_t memcg_oom_gfp_mask;
1405 int memcg_oom_order;
1407 /* Number of pages to reclaim on returning to userland: */
1408 unsigned int memcg_nr_pages_over_high;
1410 /* Used by memcontrol for targeted memcg charge: */
1411 struct mem_cgroup *active_memcg;
1414 #ifdef CONFIG_BLK_CGROUP
1415 struct request_queue *throttle_queue;
1418 #ifdef CONFIG_UPROBES
1419 struct uprobe_task *utask;
1421 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1422 unsigned int sequential_io;
1423 unsigned int sequential_io_avg;
1425 struct kmap_ctrl kmap_ctrl;
1426 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1427 unsigned long task_state_change;
1428 # ifdef CONFIG_PREEMPT_RT
1429 unsigned long saved_state_change;
1432 int pagefault_disabled;
1434 struct task_struct *oom_reaper_list;
1436 #ifdef CONFIG_VMAP_STACK
1437 struct vm_struct *stack_vm_area;
1439 #ifdef CONFIG_THREAD_INFO_IN_TASK
1440 /* A live task holds one reference: */
1441 refcount_t stack_refcount;
1443 #ifdef CONFIG_LIVEPATCH
1446 #ifdef CONFIG_SECURITY
1447 /* Used by LSM modules for access restriction: */
1450 #ifdef CONFIG_BPF_SYSCALL
1451 /* Used by BPF task local storage */
1452 struct bpf_local_storage __rcu *bpf_storage;
1453 /* Used for BPF run context */
1454 struct bpf_run_ctx *bpf_ctx;
1457 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1458 unsigned long lowest_stack;
1459 unsigned long prev_lowest_stack;
1462 #ifdef CONFIG_X86_MCE
1463 void __user *mce_vaddr;
1468 __mce_reserved : 62;
1469 struct callback_head mce_kill_me;
1473 #ifdef CONFIG_KRETPROBES
1474 struct llist_head kretprobe_instances;
1477 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1479 * If L1D flush is supported on mm context switch
1480 * then we use this callback head to queue kill work
1481 * to kill tasks that are not running on SMT disabled
1484 struct callback_head l1d_flush_kill;
1488 * New fields for task_struct should be added above here, so that
1489 * they are included in the randomized portion of task_struct.
1491 randomized_struct_fields_end
1493 /* CPU-specific state of this task: */
1494 struct thread_struct thread;
1497 * WARNING: on x86, 'thread_struct' contains a variable-sized
1498 * structure. It *MUST* be at the end of 'task_struct'.
1500 * Do not put anything below here!
1504 static inline struct pid *task_pid(struct task_struct *task)
1506 return task->thread_pid;
1510 * the helpers to get the task's different pids as they are seen
1511 * from various namespaces
1513 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1514 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1516 * task_xid_nr_ns() : id seen from the ns specified;
1518 * see also pid_nr() etc in include/linux/pid.h
1520 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1522 static inline pid_t task_pid_nr(struct task_struct *tsk)
1527 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1529 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1532 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1534 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1538 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1544 * pid_alive - check that a task structure is not stale
1545 * @p: Task structure to be checked.
1547 * Test if a process is not yet dead (at most zombie state)
1548 * If pid_alive fails, then pointers within the task structure
1549 * can be stale and must not be dereferenced.
1551 * Return: 1 if the process is alive. 0 otherwise.
1553 static inline int pid_alive(const struct task_struct *p)
1555 return p->thread_pid != NULL;
1558 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1560 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1563 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1565 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1569 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1571 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1574 static inline pid_t task_session_vnr(struct task_struct *tsk)
1576 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1579 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1581 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1584 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1586 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1589 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1595 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1601 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1603 return task_ppid_nr_ns(tsk, &init_pid_ns);
1606 /* Obsolete, do not use: */
1607 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1609 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1612 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1613 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1615 static inline unsigned int task_state_index(struct task_struct *tsk)
1617 unsigned int tsk_state = READ_ONCE(tsk->__state);
1618 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1620 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1622 if (tsk_state == TASK_IDLE)
1623 state = TASK_REPORT_IDLE;
1628 static inline char task_index_to_char(unsigned int state)
1630 static const char state_char[] = "RSDTtXZPI";
1632 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1634 return state_char[state];
1637 static inline char task_state_to_char(struct task_struct *tsk)
1639 return task_index_to_char(task_state_index(tsk));
1643 * is_global_init - check if a task structure is init. Since init
1644 * is free to have sub-threads we need to check tgid.
1645 * @tsk: Task structure to be checked.
1647 * Check if a task structure is the first user space task the kernel created.
1649 * Return: 1 if the task structure is init. 0 otherwise.
1651 static inline int is_global_init(struct task_struct *tsk)
1653 return task_tgid_nr(tsk) == 1;
1656 extern struct pid *cad_pid;
1661 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1662 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1663 #define PF_EXITING 0x00000004 /* Getting shut down */
1664 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1665 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1666 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1667 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1668 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1669 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1670 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1671 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1672 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1673 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1674 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1675 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1676 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1677 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1678 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1679 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1680 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1681 * I am cleaning dirty pages from some other bdi. */
1682 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1683 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1684 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1685 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1686 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1687 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1688 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1689 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1692 * Only the _current_ task can read/write to tsk->flags, but other
1693 * tasks can access tsk->flags in readonly mode for example
1694 * with tsk_used_math (like during threaded core dumping).
1695 * There is however an exception to this rule during ptrace
1696 * or during fork: the ptracer task is allowed to write to the
1697 * child->flags of its traced child (same goes for fork, the parent
1698 * can write to the child->flags), because we're guaranteed the
1699 * child is not running and in turn not changing child->flags
1700 * at the same time the parent does it.
1702 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1703 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1704 #define clear_used_math() clear_stopped_child_used_math(current)
1705 #define set_used_math() set_stopped_child_used_math(current)
1707 #define conditional_stopped_child_used_math(condition, child) \
1708 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1710 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1712 #define copy_to_stopped_child_used_math(child) \
1713 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1715 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1716 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1717 #define used_math() tsk_used_math(current)
1719 static __always_inline bool is_percpu_thread(void)
1722 return (current->flags & PF_NO_SETAFFINITY) &&
1723 (current->nr_cpus_allowed == 1);
1729 /* Per-process atomic flags. */
1730 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1731 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1732 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1733 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1734 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1735 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1736 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1737 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1739 #define TASK_PFA_TEST(name, func) \
1740 static inline bool task_##func(struct task_struct *p) \
1741 { return test_bit(PFA_##name, &p->atomic_flags); }
1743 #define TASK_PFA_SET(name, func) \
1744 static inline void task_set_##func(struct task_struct *p) \
1745 { set_bit(PFA_##name, &p->atomic_flags); }
1747 #define TASK_PFA_CLEAR(name, func) \
1748 static inline void task_clear_##func(struct task_struct *p) \
1749 { clear_bit(PFA_##name, &p->atomic_flags); }
1751 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1752 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1754 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1755 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1756 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1758 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1759 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1760 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1762 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1763 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1764 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1766 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1767 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1768 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1770 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1771 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1773 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1774 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1775 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1777 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1778 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1781 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1783 current->flags &= ~flags;
1784 current->flags |= orig_flags & flags;
1787 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1788 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1790 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1791 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1792 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1793 extern void release_user_cpus_ptr(struct task_struct *p);
1794 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1795 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1796 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1798 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1801 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1803 if (!cpumask_test_cpu(0, new_mask))
1807 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1809 if (src->user_cpus_ptr)
1813 static inline void release_user_cpus_ptr(struct task_struct *p)
1815 WARN_ON(p->user_cpus_ptr);
1818 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1824 extern int yield_to(struct task_struct *p, bool preempt);
1825 extern void set_user_nice(struct task_struct *p, long nice);
1826 extern int task_prio(const struct task_struct *p);
1829 * task_nice - return the nice value of a given task.
1830 * @p: the task in question.
1832 * Return: The nice value [ -20 ... 0 ... 19 ].
1834 static inline int task_nice(const struct task_struct *p)
1836 return PRIO_TO_NICE((p)->static_prio);
1839 extern int can_nice(const struct task_struct *p, const int nice);
1840 extern int task_curr(const struct task_struct *p);
1841 extern int idle_cpu(int cpu);
1842 extern int available_idle_cpu(int cpu);
1843 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1844 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1845 extern void sched_set_fifo(struct task_struct *p);
1846 extern void sched_set_fifo_low(struct task_struct *p);
1847 extern void sched_set_normal(struct task_struct *p, int nice);
1848 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1849 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1850 extern struct task_struct *idle_task(int cpu);
1853 * is_idle_task - is the specified task an idle task?
1854 * @p: the task in question.
1856 * Return: 1 if @p is an idle task. 0 otherwise.
1858 static __always_inline bool is_idle_task(const struct task_struct *p)
1860 return !!(p->flags & PF_IDLE);
1863 extern struct task_struct *curr_task(int cpu);
1864 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1868 union thread_union {
1869 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1870 struct task_struct task;
1872 #ifndef CONFIG_THREAD_INFO_IN_TASK
1873 struct thread_info thread_info;
1875 unsigned long stack[THREAD_SIZE/sizeof(long)];
1878 #ifndef CONFIG_THREAD_INFO_IN_TASK
1879 extern struct thread_info init_thread_info;
1882 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1884 #ifdef CONFIG_THREAD_INFO_IN_TASK
1885 # define task_thread_info(task) (&(task)->thread_info)
1886 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1887 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1891 * find a task by one of its numerical ids
1893 * find_task_by_pid_ns():
1894 * finds a task by its pid in the specified namespace
1895 * find_task_by_vpid():
1896 * finds a task by its virtual pid
1898 * see also find_vpid() etc in include/linux/pid.h
1901 extern struct task_struct *find_task_by_vpid(pid_t nr);
1902 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1905 * find a task by its virtual pid and get the task struct
1907 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1909 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1910 extern int wake_up_process(struct task_struct *tsk);
1911 extern void wake_up_new_task(struct task_struct *tsk);
1914 extern void kick_process(struct task_struct *tsk);
1916 static inline void kick_process(struct task_struct *tsk) { }
1919 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1921 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1923 __set_task_comm(tsk, from, false);
1926 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1927 #define get_task_comm(buf, tsk) ({ \
1928 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1929 __get_task_comm(buf, sizeof(buf), tsk); \
1933 static __always_inline void scheduler_ipi(void)
1936 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1937 * TIF_NEED_RESCHED remotely (for the first time) will also send
1940 preempt_fold_need_resched();
1942 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1944 static inline void scheduler_ipi(void) { }
1945 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1952 * Set thread flags in other task's structures.
1953 * See asm/thread_info.h for TIF_xxxx flags available:
1955 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1957 set_ti_thread_flag(task_thread_info(tsk), flag);
1960 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1962 clear_ti_thread_flag(task_thread_info(tsk), flag);
1965 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1968 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1971 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1973 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1976 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1978 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1981 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1983 return test_ti_thread_flag(task_thread_info(tsk), flag);
1986 static inline void set_tsk_need_resched(struct task_struct *tsk)
1988 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1991 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1993 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1996 static inline int test_tsk_need_resched(struct task_struct *tsk)
1998 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2002 * cond_resched() and cond_resched_lock(): latency reduction via
2003 * explicit rescheduling in places that are safe. The return
2004 * value indicates whether a reschedule was done in fact.
2005 * cond_resched_lock() will drop the spinlock before scheduling,
2007 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2008 extern int __cond_resched(void);
2010 #ifdef CONFIG_PREEMPT_DYNAMIC
2012 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2014 static __always_inline int _cond_resched(void)
2016 return static_call_mod(cond_resched)();
2021 static inline int _cond_resched(void)
2023 return __cond_resched();
2026 #endif /* CONFIG_PREEMPT_DYNAMIC */
2030 static inline int _cond_resched(void) { return 0; }
2032 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2034 #define cond_resched() ({ \
2035 __might_resched(__FILE__, __LINE__, 0); \
2039 extern int __cond_resched_lock(spinlock_t *lock);
2040 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2041 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2043 #define MIGHT_RESCHED_RCU_SHIFT 8
2044 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2046 #ifndef CONFIG_PREEMPT_RT
2048 * Non RT kernels have an elevated preempt count due to the held lock,
2049 * but are not allowed to be inside a RCU read side critical section
2051 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2054 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2055 * cond_resched*lock() has to take that into account because it checks for
2056 * preempt_count() and rcu_preempt_depth().
2058 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2059 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2062 #define cond_resched_lock(lock) ({ \
2063 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2064 __cond_resched_lock(lock); \
2067 #define cond_resched_rwlock_read(lock) ({ \
2068 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2069 __cond_resched_rwlock_read(lock); \
2072 #define cond_resched_rwlock_write(lock) ({ \
2073 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2074 __cond_resched_rwlock_write(lock); \
2077 static inline void cond_resched_rcu(void)
2079 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2087 * Does a critical section need to be broken due to another
2088 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2089 * but a general need for low latency)
2091 static inline int spin_needbreak(spinlock_t *lock)
2093 #ifdef CONFIG_PREEMPTION
2094 return spin_is_contended(lock);
2101 * Check if a rwlock is contended.
2102 * Returns non-zero if there is another task waiting on the rwlock.
2103 * Returns zero if the lock is not contended or the system / underlying
2104 * rwlock implementation does not support contention detection.
2105 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2108 static inline int rwlock_needbreak(rwlock_t *lock)
2110 #ifdef CONFIG_PREEMPTION
2111 return rwlock_is_contended(lock);
2117 static __always_inline bool need_resched(void)
2119 return unlikely(tif_need_resched());
2123 * Wrappers for p->thread_info->cpu access. No-op on UP.
2127 static inline unsigned int task_cpu(const struct task_struct *p)
2129 return READ_ONCE(task_thread_info(p)->cpu);
2132 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2136 static inline unsigned int task_cpu(const struct task_struct *p)
2141 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2145 #endif /* CONFIG_SMP */
2147 extern bool sched_task_on_rq(struct task_struct *p);
2148 extern unsigned long get_wchan(struct task_struct *p);
2151 * In order to reduce various lock holder preemption latencies provide an
2152 * interface to see if a vCPU is currently running or not.
2154 * This allows us to terminate optimistic spin loops and block, analogous to
2155 * the native optimistic spin heuristic of testing if the lock owner task is
2158 #ifndef vcpu_is_preempted
2159 static inline bool vcpu_is_preempted(int cpu)
2165 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2166 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2168 #ifndef TASK_SIZE_OF
2169 #define TASK_SIZE_OF(tsk) TASK_SIZE
2173 /* Returns effective CPU energy utilization, as seen by the scheduler */
2174 unsigned long sched_cpu_util(int cpu, unsigned long max);
2175 #endif /* CONFIG_SMP */
2180 * Map the event mask on the user-space ABI enum rseq_cs_flags
2181 * for direct mask checks.
2183 enum rseq_event_mask_bits {
2184 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2185 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2186 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2189 enum rseq_event_mask {
2190 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2191 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2192 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2195 static inline void rseq_set_notify_resume(struct task_struct *t)
2198 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2201 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2203 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2204 struct pt_regs *regs)
2207 __rseq_handle_notify_resume(ksig, regs);
2210 static inline void rseq_signal_deliver(struct ksignal *ksig,
2211 struct pt_regs *regs)
2214 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2216 rseq_handle_notify_resume(ksig, regs);
2219 /* rseq_preempt() requires preemption to be disabled. */
2220 static inline void rseq_preempt(struct task_struct *t)
2222 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2223 rseq_set_notify_resume(t);
2226 /* rseq_migrate() requires preemption to be disabled. */
2227 static inline void rseq_migrate(struct task_struct *t)
2229 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2230 rseq_set_notify_resume(t);
2234 * If parent process has a registered restartable sequences area, the
2235 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2237 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2239 if (clone_flags & CLONE_VM) {
2242 t->rseq_event_mask = 0;
2244 t->rseq = current->rseq;
2245 t->rseq_sig = current->rseq_sig;
2246 t->rseq_event_mask = current->rseq_event_mask;
2250 static inline void rseq_execve(struct task_struct *t)
2254 t->rseq_event_mask = 0;
2259 static inline void rseq_set_notify_resume(struct task_struct *t)
2262 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2263 struct pt_regs *regs)
2266 static inline void rseq_signal_deliver(struct ksignal *ksig,
2267 struct pt_regs *regs)
2270 static inline void rseq_preempt(struct task_struct *t)
2273 static inline void rseq_migrate(struct task_struct *t)
2276 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2279 static inline void rseq_execve(struct task_struct *t)
2285 #ifdef CONFIG_DEBUG_RSEQ
2287 void rseq_syscall(struct pt_regs *regs);
2291 static inline void rseq_syscall(struct pt_regs *regs)
2297 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2298 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2299 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2301 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2302 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2303 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2305 int sched_trace_rq_cpu(struct rq *rq);
2306 int sched_trace_rq_cpu_capacity(struct rq *rq);
2307 int sched_trace_rq_nr_running(struct rq *rq);
2309 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2311 #ifdef CONFIG_SCHED_CORE
2312 extern void sched_core_free(struct task_struct *tsk);
2313 extern void sched_core_fork(struct task_struct *p);
2314 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2315 unsigned long uaddr);
2317 static inline void sched_core_free(struct task_struct *tsk) { }
2318 static inline void sched_core_fork(struct task_struct *p) { }