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>
38 #include <asm/kmap_size.h>
40 /* task_struct member predeclarations (sorted alphabetically): */
42 struct backing_dev_info;
45 struct bpf_local_storage;
47 struct capture_control;
50 struct futex_pi_state;
56 struct perf_event_context;
58 struct pipe_inode_info;
61 struct robust_list_head;
67 struct sighand_struct;
69 struct task_delay_info;
73 * Task state bitmask. NOTE! These bits are also
74 * encoded in fs/proc/array.c: get_task_state().
76 * We have two separate sets of flags: task->state
77 * is about runnability, while task->exit_state are
78 * about the task exiting. Confusing, but this way
79 * modifying one set can't modify the other one by
83 /* Used in tsk->state: */
84 #define TASK_RUNNING 0x00000000
85 #define TASK_INTERRUPTIBLE 0x00000001
86 #define TASK_UNINTERRUPTIBLE 0x00000002
87 #define __TASK_STOPPED 0x00000004
88 #define __TASK_TRACED 0x00000008
89 /* Used in tsk->exit_state: */
90 #define EXIT_DEAD 0x00000010
91 #define EXIT_ZOMBIE 0x00000020
92 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
93 /* Used in tsk->state again: */
94 #define TASK_PARKED 0x00000040
95 #define TASK_DEAD 0x00000080
96 #define TASK_WAKEKILL 0x00000100
97 #define TASK_WAKING 0x00000200
98 #define TASK_NOLOAD 0x00000400
99 #define TASK_NEW 0x00000800
100 /* RT specific auxilliary flag to mark RT lock waiters */
101 #define TASK_RTLOCK_WAIT 0x00001000
102 #define TASK_STATE_MAX 0x00002000
104 #define TASK_ANY (TASK_STATE_MAX-1)
106 /* Convenience macros for the sake of set_current_state: */
107 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
108 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
109 #define TASK_TRACED __TASK_TRACED
111 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
113 /* Convenience macros for the sake of wake_up(): */
114 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
116 /* get_task_state(): */
117 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
118 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
119 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
122 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
124 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
125 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
126 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
129 * Special states are those that do not use the normal wait-loop pattern. See
130 * the comment with set_special_state().
132 #define is_special_task_state(state) \
133 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
135 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
136 # define debug_normal_state_change(state_value) \
138 WARN_ON_ONCE(is_special_task_state(state_value)); \
139 current->task_state_change = _THIS_IP_; \
142 # define debug_special_state_change(state_value) \
144 WARN_ON_ONCE(!is_special_task_state(state_value)); \
145 current->task_state_change = _THIS_IP_; \
148 # define debug_rtlock_wait_set_state() \
150 current->saved_state_change = current->task_state_change;\
151 current->task_state_change = _THIS_IP_; \
154 # define debug_rtlock_wait_restore_state() \
156 current->task_state_change = current->saved_state_change;\
160 # define debug_normal_state_change(cond) do { } while (0)
161 # define debug_special_state_change(cond) do { } while (0)
162 # define debug_rtlock_wait_set_state() do { } while (0)
163 # define debug_rtlock_wait_restore_state() do { } while (0)
167 * set_current_state() includes a barrier so that the write of current->state
168 * is correctly serialised wrt the caller's subsequent test of whether to
172 * set_current_state(TASK_UNINTERRUPTIBLE);
178 * __set_current_state(TASK_RUNNING);
180 * If the caller does not need such serialisation (because, for instance, the
181 * CONDITION test and condition change and wakeup are under the same lock) then
182 * use __set_current_state().
184 * The above is typically ordered against the wakeup, which does:
187 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
189 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
190 * accessing p->state.
192 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
193 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
194 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
196 * However, with slightly different timing the wakeup TASK_RUNNING store can
197 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
198 * a problem either because that will result in one extra go around the loop
199 * and our @cond test will save the day.
201 * Also see the comments of try_to_wake_up().
203 #define __set_current_state(state_value) \
205 debug_normal_state_change((state_value)); \
206 WRITE_ONCE(current->__state, (state_value)); \
209 #define set_current_state(state_value) \
211 debug_normal_state_change((state_value)); \
212 smp_store_mb(current->__state, (state_value)); \
216 * set_special_state() should be used for those states when the blocking task
217 * can not use the regular condition based wait-loop. In that case we must
218 * serialize against wakeups such that any possible in-flight TASK_RUNNING
219 * stores will not collide with our state change.
221 #define set_special_state(state_value) \
223 unsigned long flags; /* may shadow */ \
225 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
226 debug_special_state_change((state_value)); \
227 WRITE_ONCE(current->__state, (state_value)); \
228 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
232 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
234 * RT's spin/rwlock substitutions are state preserving. The state of the
235 * task when blocking on the lock is saved in task_struct::saved_state and
236 * restored after the lock has been acquired. These operations are
237 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
238 * lock related wakeups while the task is blocked on the lock are
239 * redirected to operate on task_struct::saved_state to ensure that these
240 * are not dropped. On restore task_struct::saved_state is set to
241 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
243 * The lock operation looks like this:
245 * current_save_and_set_rtlock_wait_state();
249 * raw_spin_unlock_irq(&lock->wait_lock);
251 * raw_spin_lock_irq(&lock->wait_lock);
252 * set_current_state(TASK_RTLOCK_WAIT);
254 * current_restore_rtlock_saved_state();
256 #define current_save_and_set_rtlock_wait_state() \
258 lockdep_assert_irqs_disabled(); \
259 raw_spin_lock(¤t->pi_lock); \
260 current->saved_state = current->__state; \
261 debug_rtlock_wait_set_state(); \
262 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
263 raw_spin_unlock(¤t->pi_lock); \
266 #define current_restore_rtlock_saved_state() \
268 lockdep_assert_irqs_disabled(); \
269 raw_spin_lock(¤t->pi_lock); \
270 debug_rtlock_wait_restore_state(); \
271 WRITE_ONCE(current->__state, current->saved_state); \
272 current->saved_state = TASK_RUNNING; \
273 raw_spin_unlock(¤t->pi_lock); \
276 #define get_current_state() READ_ONCE(current->__state)
279 * Define the task command name length as enum, then it can be visible to
286 extern void scheduler_tick(void);
288 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
290 extern long schedule_timeout(long timeout);
291 extern long schedule_timeout_interruptible(long timeout);
292 extern long schedule_timeout_killable(long timeout);
293 extern long schedule_timeout_uninterruptible(long timeout);
294 extern long schedule_timeout_idle(long timeout);
295 asmlinkage void schedule(void);
296 extern void schedule_preempt_disabled(void);
297 asmlinkage void preempt_schedule_irq(void);
298 #ifdef CONFIG_PREEMPT_RT
299 extern void schedule_rtlock(void);
302 extern int __must_check io_schedule_prepare(void);
303 extern void io_schedule_finish(int token);
304 extern long io_schedule_timeout(long timeout);
305 extern void io_schedule(void);
308 * struct prev_cputime - snapshot of system and user cputime
309 * @utime: time spent in user mode
310 * @stime: time spent in system mode
311 * @lock: protects the above two fields
313 * Stores previous user/system time values such that we can guarantee
316 struct prev_cputime {
317 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
325 /* Task is sleeping or running in a CPU with VTIME inactive: */
329 /* Task runs in kernelspace in a CPU with VTIME active: */
331 /* Task runs in userspace in a CPU with VTIME active: */
333 /* Task runs as guests in a CPU with VTIME active: */
339 unsigned long long starttime;
340 enum vtime_state state;
348 * Utilization clamp constraints.
349 * @UCLAMP_MIN: Minimum utilization
350 * @UCLAMP_MAX: Maximum utilization
351 * @UCLAMP_CNT: Utilization clamp constraints count
360 extern struct root_domain def_root_domain;
361 extern struct mutex sched_domains_mutex;
365 #ifdef CONFIG_SCHED_INFO
366 /* Cumulative counters: */
368 /* # of times we have run on this CPU: */
369 unsigned long pcount;
371 /* Time spent waiting on a runqueue: */
372 unsigned long long run_delay;
376 /* When did we last run on a CPU? */
377 unsigned long long last_arrival;
379 /* When were we last queued to run? */
380 unsigned long long last_queued;
382 #endif /* CONFIG_SCHED_INFO */
386 * Integer metrics need fixed point arithmetic, e.g., sched/fair
387 * has a few: load, load_avg, util_avg, freq, and capacity.
389 * We define a basic fixed point arithmetic range, and then formalize
390 * all these metrics based on that basic range.
392 # define SCHED_FIXEDPOINT_SHIFT 10
393 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
395 /* Increase resolution of cpu_capacity calculations */
396 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
397 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
400 unsigned long weight;
405 * struct util_est - Estimation utilization of FAIR tasks
406 * @enqueued: instantaneous estimated utilization of a task/cpu
407 * @ewma: the Exponential Weighted Moving Average (EWMA)
408 * utilization of a task
410 * Support data structure to track an Exponential Weighted Moving Average
411 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
412 * average each time a task completes an activation. Sample's weight is chosen
413 * so that the EWMA will be relatively insensitive to transient changes to the
416 * The enqueued attribute has a slightly different meaning for tasks and cpus:
417 * - task: the task's util_avg at last task dequeue time
418 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
419 * Thus, the util_est.enqueued of a task represents the contribution on the
420 * estimated utilization of the CPU where that task is currently enqueued.
422 * Only for tasks we track a moving average of the past instantaneous
423 * estimated utilization. This allows to absorb sporadic drops in utilization
424 * of an otherwise almost periodic task.
426 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
427 * updates. When a task is dequeued, its util_est should not be updated if its
428 * util_avg has not been updated in the meantime.
429 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
430 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
431 * for a task) it is safe to use MSB.
434 unsigned int enqueued;
436 #define UTIL_EST_WEIGHT_SHIFT 2
437 #define UTIL_AVG_UNCHANGED 0x80000000
438 } __attribute__((__aligned__(sizeof(u64))));
441 * The load/runnable/util_avg accumulates an infinite geometric series
442 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
444 * [load_avg definition]
446 * load_avg = runnable% * scale_load_down(load)
448 * [runnable_avg definition]
450 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
452 * [util_avg definition]
454 * util_avg = running% * SCHED_CAPACITY_SCALE
456 * where runnable% is the time ratio that a sched_entity is runnable and
457 * running% the time ratio that a sched_entity is running.
459 * For cfs_rq, they are the aggregated values of all runnable and blocked
462 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
463 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
464 * for computing those signals (see update_rq_clock_pelt())
466 * N.B., the above ratios (runnable% and running%) themselves are in the
467 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
468 * to as large a range as necessary. This is for example reflected by
469 * util_avg's SCHED_CAPACITY_SCALE.
473 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
474 * with the highest load (=88761), always runnable on a single cfs_rq,
475 * and should not overflow as the number already hits PID_MAX_LIMIT.
477 * For all other cases (including 32-bit kernels), struct load_weight's
478 * weight will overflow first before we do, because:
480 * Max(load_avg) <= Max(load.weight)
482 * Then it is the load_weight's responsibility to consider overflow
486 u64 last_update_time;
491 unsigned long load_avg;
492 unsigned long runnable_avg;
493 unsigned long util_avg;
494 struct util_est util_est;
495 } ____cacheline_aligned;
497 struct sched_statistics {
498 #ifdef CONFIG_SCHEDSTATS
508 s64 sum_sleep_runtime;
512 s64 sum_block_runtime;
517 u64 nr_migrations_cold;
518 u64 nr_failed_migrations_affine;
519 u64 nr_failed_migrations_running;
520 u64 nr_failed_migrations_hot;
521 u64 nr_forced_migrations;
525 u64 nr_wakeups_migrate;
526 u64 nr_wakeups_local;
527 u64 nr_wakeups_remote;
528 u64 nr_wakeups_affine;
529 u64 nr_wakeups_affine_attempts;
530 u64 nr_wakeups_passive;
533 #ifdef CONFIG_SCHED_CORE
534 u64 core_forceidle_sum;
536 #endif /* CONFIG_SCHEDSTATS */
537 } ____cacheline_aligned;
539 struct sched_entity {
540 /* For load-balancing: */
541 struct load_weight load;
542 struct rb_node run_node;
543 struct list_head group_node;
547 u64 sum_exec_runtime;
549 u64 prev_sum_exec_runtime;
553 #ifdef CONFIG_FAIR_GROUP_SCHED
555 struct sched_entity *parent;
556 /* rq on which this entity is (to be) queued: */
557 struct cfs_rq *cfs_rq;
558 /* rq "owned" by this entity/group: */
560 /* cached value of my_q->h_nr_running */
561 unsigned long runnable_weight;
566 * Per entity load average tracking.
568 * Put into separate cache line so it does not
569 * collide with read-mostly values above.
571 struct sched_avg avg;
575 struct sched_rt_entity {
576 struct list_head run_list;
577 unsigned long timeout;
578 unsigned long watchdog_stamp;
579 unsigned int time_slice;
580 unsigned short on_rq;
581 unsigned short on_list;
583 struct sched_rt_entity *back;
584 #ifdef CONFIG_RT_GROUP_SCHED
585 struct sched_rt_entity *parent;
586 /* rq on which this entity is (to be) queued: */
588 /* rq "owned" by this entity/group: */
591 } __randomize_layout;
593 struct sched_dl_entity {
594 struct rb_node rb_node;
597 * Original scheduling parameters. Copied here from sched_attr
598 * during sched_setattr(), they will remain the same until
599 * the next sched_setattr().
601 u64 dl_runtime; /* Maximum runtime for each instance */
602 u64 dl_deadline; /* Relative deadline of each instance */
603 u64 dl_period; /* Separation of two instances (period) */
604 u64 dl_bw; /* dl_runtime / dl_period */
605 u64 dl_density; /* dl_runtime / dl_deadline */
608 * Actual scheduling parameters. Initialized with the values above,
609 * they are continuously updated during task execution. Note that
610 * the remaining runtime could be < 0 in case we are in overrun.
612 s64 runtime; /* Remaining runtime for this instance */
613 u64 deadline; /* Absolute deadline for this instance */
614 unsigned int flags; /* Specifying the scheduler behaviour */
619 * @dl_throttled tells if we exhausted the runtime. If so, the
620 * task has to wait for a replenishment to be performed at the
621 * next firing of dl_timer.
623 * @dl_yielded tells if task gave up the CPU before consuming
624 * all its available runtime during the last job.
626 * @dl_non_contending tells if the task is inactive while still
627 * contributing to the active utilization. In other words, it
628 * indicates if the inactive timer has been armed and its handler
629 * has not been executed yet. This flag is useful to avoid race
630 * conditions between the inactive timer handler and the wakeup
633 * @dl_overrun tells if the task asked to be informed about runtime
636 unsigned int dl_throttled : 1;
637 unsigned int dl_yielded : 1;
638 unsigned int dl_non_contending : 1;
639 unsigned int dl_overrun : 1;
642 * Bandwidth enforcement timer. Each -deadline task has its
643 * own bandwidth to be enforced, thus we need one timer per task.
645 struct hrtimer dl_timer;
648 * Inactive timer, responsible for decreasing the active utilization
649 * at the "0-lag time". When a -deadline task blocks, it contributes
650 * to GRUB's active utilization until the "0-lag time", hence a
651 * timer is needed to decrease the active utilization at the correct
654 struct hrtimer inactive_timer;
656 #ifdef CONFIG_RT_MUTEXES
658 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
659 * pi_se points to the donor, otherwise points to the dl_se it belongs
660 * to (the original one/itself).
662 struct sched_dl_entity *pi_se;
666 #ifdef CONFIG_UCLAMP_TASK
667 /* Number of utilization clamp buckets (shorter alias) */
668 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
671 * Utilization clamp for a scheduling entity
672 * @value: clamp value "assigned" to a se
673 * @bucket_id: bucket index corresponding to the "assigned" value
674 * @active: the se is currently refcounted in a rq's bucket
675 * @user_defined: the requested clamp value comes from user-space
677 * The bucket_id is the index of the clamp bucket matching the clamp value
678 * which is pre-computed and stored to avoid expensive integer divisions from
681 * The active bit is set whenever a task has got an "effective" value assigned,
682 * which can be different from the clamp value "requested" from user-space.
683 * This allows to know a task is refcounted in the rq's bucket corresponding
684 * to the "effective" bucket_id.
686 * The user_defined bit is set whenever a task has got a task-specific clamp
687 * value requested from userspace, i.e. the system defaults apply to this task
688 * just as a restriction. This allows to relax default clamps when a less
689 * restrictive task-specific value has been requested, thus allowing to
690 * implement a "nice" semantic. For example, a task running with a 20%
691 * default boost can still drop its own boosting to 0%.
694 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
695 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
696 unsigned int active : 1;
697 unsigned int user_defined : 1;
699 #endif /* CONFIG_UCLAMP_TASK */
705 u8 exp_hint; /* Hint for performance. */
706 u8 need_mb; /* Readers need smp_mb(). */
708 u32 s; /* Set of bits. */
711 enum perf_event_task_context {
712 perf_invalid_context = -1,
715 perf_nr_task_contexts,
719 struct wake_q_node *next;
723 #ifdef CONFIG_KMAP_LOCAL
725 pte_t pteval[KM_MAX_IDX];
730 #ifdef CONFIG_THREAD_INFO_IN_TASK
732 * For reasons of header soup (see current_thread_info()), this
733 * must be the first element of task_struct.
735 struct thread_info thread_info;
737 unsigned int __state;
739 #ifdef CONFIG_PREEMPT_RT
740 /* saved state for "spinlock sleepers" */
741 unsigned int saved_state;
745 * This begins the randomizable portion of task_struct. Only
746 * scheduling-critical items should be added above here.
748 randomized_struct_fields_start
752 /* Per task flags (PF_*), defined further below: */
758 struct __call_single_node wake_entry;
759 unsigned int wakee_flips;
760 unsigned long wakee_flip_decay_ts;
761 struct task_struct *last_wakee;
764 * recent_used_cpu is initially set as the last CPU used by a task
765 * that wakes affine another task. Waker/wakee relationships can
766 * push tasks around a CPU where each wakeup moves to the next one.
767 * Tracking a recently used CPU allows a quick search for a recently
768 * used CPU that may be idle.
778 unsigned int rt_priority;
780 struct sched_entity se;
781 struct sched_rt_entity rt;
782 struct sched_dl_entity dl;
783 const struct sched_class *sched_class;
785 #ifdef CONFIG_SCHED_CORE
786 struct rb_node core_node;
787 unsigned long core_cookie;
788 unsigned int core_occupation;
791 #ifdef CONFIG_CGROUP_SCHED
792 struct task_group *sched_task_group;
795 #ifdef CONFIG_UCLAMP_TASK
797 * Clamp values requested for a scheduling entity.
798 * Must be updated with task_rq_lock() held.
800 struct uclamp_se uclamp_req[UCLAMP_CNT];
802 * Effective clamp values used for a scheduling entity.
803 * Must be updated with task_rq_lock() held.
805 struct uclamp_se uclamp[UCLAMP_CNT];
808 struct sched_statistics stats;
810 #ifdef CONFIG_PREEMPT_NOTIFIERS
811 /* List of struct preempt_notifier: */
812 struct hlist_head preempt_notifiers;
815 #ifdef CONFIG_BLK_DEV_IO_TRACE
816 unsigned int btrace_seq;
821 const cpumask_t *cpus_ptr;
822 cpumask_t *user_cpus_ptr;
824 void *migration_pending;
826 unsigned short migration_disabled;
828 unsigned short migration_flags;
830 #ifdef CONFIG_PREEMPT_RCU
831 int rcu_read_lock_nesting;
832 union rcu_special rcu_read_unlock_special;
833 struct list_head rcu_node_entry;
834 struct rcu_node *rcu_blocked_node;
835 #endif /* #ifdef CONFIG_PREEMPT_RCU */
837 #ifdef CONFIG_TASKS_RCU
838 unsigned long rcu_tasks_nvcsw;
839 u8 rcu_tasks_holdout;
841 int rcu_tasks_idle_cpu;
842 struct list_head rcu_tasks_holdout_list;
843 #endif /* #ifdef CONFIG_TASKS_RCU */
845 #ifdef CONFIG_TASKS_TRACE_RCU
846 int trc_reader_nesting;
848 union rcu_special trc_reader_special;
849 struct list_head trc_holdout_list;
850 struct list_head trc_blkd_node;
852 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
854 struct sched_info sched_info;
856 struct list_head tasks;
858 struct plist_node pushable_tasks;
859 struct rb_node pushable_dl_tasks;
862 struct mm_struct *mm;
863 struct mm_struct *active_mm;
865 /* Per-thread vma caching: */
866 struct vmacache vmacache;
868 #ifdef SPLIT_RSS_COUNTING
869 struct task_rss_stat rss_stat;
874 /* The signal sent when the parent dies: */
876 /* JOBCTL_*, siglock protected: */
877 unsigned long jobctl;
879 /* Used for emulating ABI behavior of previous Linux versions: */
880 unsigned int personality;
882 /* Scheduler bits, serialized by scheduler locks: */
883 unsigned sched_reset_on_fork:1;
884 unsigned sched_contributes_to_load:1;
885 unsigned sched_migrated:1;
887 unsigned sched_psi_wake_requeue:1;
890 /* Force alignment to the next boundary: */
893 /* Unserialized, strictly 'current' */
896 * This field must not be in the scheduler word above due to wakelist
897 * queueing no longer being serialized by p->on_cpu. However:
900 * schedule() if (p->on_rq && ..) // false
901 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
902 * deactivate_task() ttwu_queue_wakelist())
903 * p->on_rq = 0; p->sched_remote_wakeup = Y;
905 * guarantees all stores of 'current' are visible before
906 * ->sched_remote_wakeup gets used, so it can be in this word.
908 unsigned sched_remote_wakeup:1;
910 /* Bit to tell LSMs we're in execve(): */
911 unsigned in_execve:1;
912 unsigned in_iowait:1;
913 #ifndef TIF_RESTORE_SIGMASK
914 unsigned restore_sigmask:1;
917 unsigned in_user_fault:1;
919 #ifdef CONFIG_COMPAT_BRK
920 unsigned brk_randomized:1;
922 #ifdef CONFIG_CGROUPS
923 /* disallow userland-initiated cgroup migration */
924 unsigned no_cgroup_migration:1;
925 /* task is frozen/stopped (used by the cgroup freezer) */
928 #ifdef CONFIG_BLK_CGROUP
929 unsigned use_memdelay:1;
932 /* Stalled due to lack of memory */
933 unsigned in_memstall:1;
935 #ifdef CONFIG_PAGE_OWNER
936 /* Used by page_owner=on to detect recursion in page tracking. */
937 unsigned in_page_owner:1;
939 #ifdef CONFIG_EVENTFD
940 /* Recursion prevention for eventfd_signal() */
941 unsigned in_eventfd_signal:1;
943 #ifdef CONFIG_IOMMU_SVA
944 unsigned pasid_activated:1;
946 #ifdef CONFIG_CPU_SUP_INTEL
947 unsigned reported_split_lock:1;
950 unsigned long atomic_flags; /* Flags requiring atomic access. */
952 struct restart_block restart_block;
957 #ifdef CONFIG_STACKPROTECTOR
958 /* Canary value for the -fstack-protector GCC feature: */
959 unsigned long stack_canary;
962 * Pointers to the (original) parent process, youngest child, younger sibling,
963 * older sibling, respectively. (p->father can be replaced with
964 * p->real_parent->pid)
967 /* Real parent process: */
968 struct task_struct __rcu *real_parent;
970 /* Recipient of SIGCHLD, wait4() reports: */
971 struct task_struct __rcu *parent;
974 * Children/sibling form the list of natural children:
976 struct list_head children;
977 struct list_head sibling;
978 struct task_struct *group_leader;
981 * 'ptraced' is the list of tasks this task is using ptrace() on.
983 * This includes both natural children and PTRACE_ATTACH targets.
984 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
986 struct list_head ptraced;
987 struct list_head ptrace_entry;
989 /* PID/PID hash table linkage. */
990 struct pid *thread_pid;
991 struct hlist_node pid_links[PIDTYPE_MAX];
992 struct list_head thread_group;
993 struct list_head thread_node;
995 struct completion *vfork_done;
997 /* CLONE_CHILD_SETTID: */
998 int __user *set_child_tid;
1000 /* CLONE_CHILD_CLEARTID: */
1001 int __user *clear_child_tid;
1003 /* PF_KTHREAD | PF_IO_WORKER */
1004 void *worker_private;
1008 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1013 struct prev_cputime prev_cputime;
1014 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1018 #ifdef CONFIG_NO_HZ_FULL
1019 atomic_t tick_dep_mask;
1021 /* Context switch counts: */
1022 unsigned long nvcsw;
1023 unsigned long nivcsw;
1025 /* Monotonic time in nsecs: */
1028 /* Boot based time in nsecs: */
1031 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1032 unsigned long min_flt;
1033 unsigned long maj_flt;
1035 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1036 struct posix_cputimers posix_cputimers;
1038 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1039 struct posix_cputimers_work posix_cputimers_work;
1042 /* Process credentials: */
1044 /* Tracer's credentials at attach: */
1045 const struct cred __rcu *ptracer_cred;
1047 /* Objective and real subjective task credentials (COW): */
1048 const struct cred __rcu *real_cred;
1050 /* Effective (overridable) subjective task credentials (COW): */
1051 const struct cred __rcu *cred;
1054 /* Cached requested key. */
1055 struct key *cached_requested_key;
1059 * executable name, excluding path.
1061 * - normally initialized setup_new_exec()
1062 * - access it with [gs]et_task_comm()
1063 * - lock it with task_lock()
1065 char comm[TASK_COMM_LEN];
1067 struct nameidata *nameidata;
1069 #ifdef CONFIG_SYSVIPC
1070 struct sysv_sem sysvsem;
1071 struct sysv_shm sysvshm;
1073 #ifdef CONFIG_DETECT_HUNG_TASK
1074 unsigned long last_switch_count;
1075 unsigned long last_switch_time;
1077 /* Filesystem information: */
1078 struct fs_struct *fs;
1080 /* Open file information: */
1081 struct files_struct *files;
1083 #ifdef CONFIG_IO_URING
1084 struct io_uring_task *io_uring;
1088 struct nsproxy *nsproxy;
1090 /* Signal handlers: */
1091 struct signal_struct *signal;
1092 struct sighand_struct __rcu *sighand;
1094 sigset_t real_blocked;
1095 /* Restored if set_restore_sigmask() was used: */
1096 sigset_t saved_sigmask;
1097 struct sigpending pending;
1098 unsigned long sas_ss_sp;
1100 unsigned int sas_ss_flags;
1102 struct callback_head *task_works;
1105 #ifdef CONFIG_AUDITSYSCALL
1106 struct audit_context *audit_context;
1109 unsigned int sessionid;
1111 struct seccomp seccomp;
1112 struct syscall_user_dispatch syscall_dispatch;
1114 /* Thread group tracking: */
1118 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1119 spinlock_t alloc_lock;
1121 /* Protection of the PI data structures: */
1122 raw_spinlock_t pi_lock;
1124 struct wake_q_node wake_q;
1126 #ifdef CONFIG_RT_MUTEXES
1127 /* PI waiters blocked on a rt_mutex held by this task: */
1128 struct rb_root_cached pi_waiters;
1129 /* Updated under owner's pi_lock and rq lock */
1130 struct task_struct *pi_top_task;
1131 /* Deadlock detection and priority inheritance handling: */
1132 struct rt_mutex_waiter *pi_blocked_on;
1135 #ifdef CONFIG_DEBUG_MUTEXES
1136 /* Mutex deadlock detection: */
1137 struct mutex_waiter *blocked_on;
1140 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1141 int non_block_count;
1144 #ifdef CONFIG_TRACE_IRQFLAGS
1145 struct irqtrace_events irqtrace;
1146 unsigned int hardirq_threaded;
1147 u64 hardirq_chain_key;
1148 int softirqs_enabled;
1149 int softirq_context;
1152 #ifdef CONFIG_PREEMPT_RT
1153 int softirq_disable_cnt;
1156 #ifdef CONFIG_LOCKDEP
1157 # define MAX_LOCK_DEPTH 48UL
1160 unsigned int lockdep_recursion;
1161 struct held_lock held_locks[MAX_LOCK_DEPTH];
1164 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1165 unsigned int in_ubsan;
1168 /* Journalling filesystem info: */
1171 /* Stacked block device info: */
1172 struct bio_list *bio_list;
1174 /* Stack plugging: */
1175 struct blk_plug *plug;
1178 struct reclaim_state *reclaim_state;
1180 struct backing_dev_info *backing_dev_info;
1182 struct io_context *io_context;
1184 #ifdef CONFIG_COMPACTION
1185 struct capture_control *capture_control;
1188 unsigned long ptrace_message;
1189 kernel_siginfo_t *last_siginfo;
1191 struct task_io_accounting ioac;
1193 /* Pressure stall state */
1194 unsigned int psi_flags;
1196 #ifdef CONFIG_TASK_XACCT
1197 /* Accumulated RSS usage: */
1199 /* Accumulated virtual memory usage: */
1201 /* stime + utime since last update: */
1204 #ifdef CONFIG_CPUSETS
1205 /* Protected by ->alloc_lock: */
1206 nodemask_t mems_allowed;
1207 /* Sequence number to catch updates: */
1208 seqcount_spinlock_t mems_allowed_seq;
1209 int cpuset_mem_spread_rotor;
1210 int cpuset_slab_spread_rotor;
1212 #ifdef CONFIG_CGROUPS
1213 /* Control Group info protected by css_set_lock: */
1214 struct css_set __rcu *cgroups;
1215 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1216 struct list_head cg_list;
1218 #ifdef CONFIG_X86_CPU_RESCTRL
1223 struct robust_list_head __user *robust_list;
1224 #ifdef CONFIG_COMPAT
1225 struct compat_robust_list_head __user *compat_robust_list;
1227 struct list_head pi_state_list;
1228 struct futex_pi_state *pi_state_cache;
1229 struct mutex futex_exit_mutex;
1230 unsigned int futex_state;
1232 #ifdef CONFIG_PERF_EVENTS
1233 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1234 struct mutex perf_event_mutex;
1235 struct list_head perf_event_list;
1237 #ifdef CONFIG_DEBUG_PREEMPT
1238 unsigned long preempt_disable_ip;
1241 /* Protected by alloc_lock: */
1242 struct mempolicy *mempolicy;
1244 short pref_node_fork;
1246 #ifdef CONFIG_NUMA_BALANCING
1248 unsigned int numa_scan_period;
1249 unsigned int numa_scan_period_max;
1250 int numa_preferred_nid;
1251 unsigned long numa_migrate_retry;
1252 /* Migration stamp: */
1254 u64 last_task_numa_placement;
1255 u64 last_sum_exec_runtime;
1256 struct callback_head numa_work;
1259 * This pointer is only modified for current in syscall and
1260 * pagefault context (and for tasks being destroyed), so it can be read
1261 * from any of the following contexts:
1262 * - RCU read-side critical section
1263 * - current->numa_group from everywhere
1264 * - task's runqueue locked, task not running
1266 struct numa_group __rcu *numa_group;
1269 * numa_faults is an array split into four regions:
1270 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1271 * in this precise order.
1273 * faults_memory: Exponential decaying average of faults on a per-node
1274 * basis. Scheduling placement decisions are made based on these
1275 * counts. The values remain static for the duration of a PTE scan.
1276 * faults_cpu: Track the nodes the process was running on when a NUMA
1277 * hinting fault was incurred.
1278 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1279 * during the current scan window. When the scan completes, the counts
1280 * in faults_memory and faults_cpu decay and these values are copied.
1282 unsigned long *numa_faults;
1283 unsigned long total_numa_faults;
1286 * numa_faults_locality tracks if faults recorded during the last
1287 * scan window were remote/local or failed to migrate. The task scan
1288 * period is adapted based on the locality of the faults with different
1289 * weights depending on whether they were shared or private faults
1291 unsigned long numa_faults_locality[3];
1293 unsigned long numa_pages_migrated;
1294 #endif /* CONFIG_NUMA_BALANCING */
1297 struct rseq __user *rseq;
1300 * RmW on rseq_event_mask must be performed atomically
1301 * with respect to preemption.
1303 unsigned long rseq_event_mask;
1306 struct tlbflush_unmap_batch tlb_ubc;
1309 refcount_t rcu_users;
1310 struct rcu_head rcu;
1313 /* Cache last used pipe for splice(): */
1314 struct pipe_inode_info *splice_pipe;
1316 struct page_frag task_frag;
1318 #ifdef CONFIG_TASK_DELAY_ACCT
1319 struct task_delay_info *delays;
1322 #ifdef CONFIG_FAULT_INJECTION
1324 unsigned int fail_nth;
1327 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1328 * balance_dirty_pages() for a dirty throttling pause:
1331 int nr_dirtied_pause;
1332 /* Start of a write-and-pause period: */
1333 unsigned long dirty_paused_when;
1335 #ifdef CONFIG_LATENCYTOP
1336 int latency_record_count;
1337 struct latency_record latency_record[LT_SAVECOUNT];
1340 * Time slack values; these are used to round up poll() and
1341 * select() etc timeout values. These are in nanoseconds.
1344 u64 default_timer_slack_ns;
1346 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1347 unsigned int kasan_depth;
1351 struct kcsan_ctx kcsan_ctx;
1352 #ifdef CONFIG_TRACE_IRQFLAGS
1353 struct irqtrace_events kcsan_save_irqtrace;
1355 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1356 int kcsan_stack_depth;
1360 #if IS_ENABLED(CONFIG_KUNIT)
1361 struct kunit *kunit_test;
1364 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1365 /* Index of current stored address in ret_stack: */
1369 /* Stack of return addresses for return function tracing: */
1370 struct ftrace_ret_stack *ret_stack;
1372 /* Timestamp for last schedule: */
1373 unsigned long long ftrace_timestamp;
1376 * Number of functions that haven't been traced
1377 * because of depth overrun:
1379 atomic_t trace_overrun;
1381 /* Pause tracing: */
1382 atomic_t tracing_graph_pause;
1385 #ifdef CONFIG_TRACING
1386 /* State flags for use by tracers: */
1387 unsigned long trace;
1389 /* Bitmask and counter of trace recursion: */
1390 unsigned long trace_recursion;
1391 #endif /* CONFIG_TRACING */
1394 /* See kernel/kcov.c for more details. */
1396 /* Coverage collection mode enabled for this task (0 if disabled): */
1397 unsigned int kcov_mode;
1399 /* Size of the kcov_area: */
1400 unsigned int kcov_size;
1402 /* Buffer for coverage collection: */
1405 /* KCOV descriptor wired with this task or NULL: */
1408 /* KCOV common handle for remote coverage collection: */
1411 /* KCOV sequence number: */
1414 /* Collect coverage from softirq context: */
1415 unsigned int kcov_softirq;
1419 struct mem_cgroup *memcg_in_oom;
1420 gfp_t memcg_oom_gfp_mask;
1421 int memcg_oom_order;
1423 /* Number of pages to reclaim on returning to userland: */
1424 unsigned int memcg_nr_pages_over_high;
1426 /* Used by memcontrol for targeted memcg charge: */
1427 struct mem_cgroup *active_memcg;
1430 #ifdef CONFIG_BLK_CGROUP
1431 struct request_queue *throttle_queue;
1434 #ifdef CONFIG_UPROBES
1435 struct uprobe_task *utask;
1437 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1438 unsigned int sequential_io;
1439 unsigned int sequential_io_avg;
1441 struct kmap_ctrl kmap_ctrl;
1442 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1443 unsigned long task_state_change;
1444 # ifdef CONFIG_PREEMPT_RT
1445 unsigned long saved_state_change;
1448 int pagefault_disabled;
1450 struct task_struct *oom_reaper_list;
1451 struct timer_list oom_reaper_timer;
1453 #ifdef CONFIG_VMAP_STACK
1454 struct vm_struct *stack_vm_area;
1456 #ifdef CONFIG_THREAD_INFO_IN_TASK
1457 /* A live task holds one reference: */
1458 refcount_t stack_refcount;
1460 #ifdef CONFIG_LIVEPATCH
1463 #ifdef CONFIG_SECURITY
1464 /* Used by LSM modules for access restriction: */
1467 #ifdef CONFIG_BPF_SYSCALL
1468 /* Used by BPF task local storage */
1469 struct bpf_local_storage __rcu *bpf_storage;
1470 /* Used for BPF run context */
1471 struct bpf_run_ctx *bpf_ctx;
1474 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1475 unsigned long lowest_stack;
1476 unsigned long prev_lowest_stack;
1479 #ifdef CONFIG_X86_MCE
1480 void __user *mce_vaddr;
1485 __mce_reserved : 62;
1486 struct callback_head mce_kill_me;
1490 #ifdef CONFIG_KRETPROBES
1491 struct llist_head kretprobe_instances;
1493 #ifdef CONFIG_RETHOOK
1494 struct llist_head rethooks;
1497 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1499 * If L1D flush is supported on mm context switch
1500 * then we use this callback head to queue kill work
1501 * to kill tasks that are not running on SMT disabled
1504 struct callback_head l1d_flush_kill;
1509 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1510 * If we find justification for more monitors, we can think
1511 * about adding more or developing a dynamic method. So far,
1512 * none of these are justified.
1514 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1518 * New fields for task_struct should be added above here, so that
1519 * they are included in the randomized portion of task_struct.
1521 randomized_struct_fields_end
1523 /* CPU-specific state of this task: */
1524 struct thread_struct thread;
1527 * WARNING: on x86, 'thread_struct' contains a variable-sized
1528 * structure. It *MUST* be at the end of 'task_struct'.
1530 * Do not put anything below here!
1534 static inline struct pid *task_pid(struct task_struct *task)
1536 return task->thread_pid;
1540 * the helpers to get the task's different pids as they are seen
1541 * from various namespaces
1543 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1544 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1546 * task_xid_nr_ns() : id seen from the ns specified;
1548 * see also pid_nr() etc in include/linux/pid.h
1550 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1552 static inline pid_t task_pid_nr(struct task_struct *tsk)
1557 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1559 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1562 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1564 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1568 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1574 * pid_alive - check that a task structure is not stale
1575 * @p: Task structure to be checked.
1577 * Test if a process is not yet dead (at most zombie state)
1578 * If pid_alive fails, then pointers within the task structure
1579 * can be stale and must not be dereferenced.
1581 * Return: 1 if the process is alive. 0 otherwise.
1583 static inline int pid_alive(const struct task_struct *p)
1585 return p->thread_pid != NULL;
1588 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1590 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1593 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1595 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1599 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1601 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1604 static inline pid_t task_session_vnr(struct task_struct *tsk)
1606 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1609 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1611 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1614 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1616 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1619 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1625 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1631 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1633 return task_ppid_nr_ns(tsk, &init_pid_ns);
1636 /* Obsolete, do not use: */
1637 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1639 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1642 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1643 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1645 static inline unsigned int __task_state_index(unsigned int tsk_state,
1646 unsigned int tsk_exit_state)
1648 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1650 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1652 if (tsk_state == TASK_IDLE)
1653 state = TASK_REPORT_IDLE;
1656 * We're lying here, but rather than expose a completely new task state
1657 * to userspace, we can make this appear as if the task has gone through
1658 * a regular rt_mutex_lock() call.
1660 if (tsk_state == TASK_RTLOCK_WAIT)
1661 state = TASK_UNINTERRUPTIBLE;
1666 static inline unsigned int task_state_index(struct task_struct *tsk)
1668 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1671 static inline char task_index_to_char(unsigned int state)
1673 static const char state_char[] = "RSDTtXZPI";
1675 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1677 return state_char[state];
1680 static inline char task_state_to_char(struct task_struct *tsk)
1682 return task_index_to_char(task_state_index(tsk));
1686 * is_global_init - check if a task structure is init. Since init
1687 * is free to have sub-threads we need to check tgid.
1688 * @tsk: Task structure to be checked.
1690 * Check if a task structure is the first user space task the kernel created.
1692 * Return: 1 if the task structure is init. 0 otherwise.
1694 static inline int is_global_init(struct task_struct *tsk)
1696 return task_tgid_nr(tsk) == 1;
1699 extern struct pid *cad_pid;
1704 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1705 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1706 #define PF_EXITING 0x00000004 /* Getting shut down */
1707 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1708 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1709 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1710 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1711 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1712 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1713 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1714 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1715 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1716 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1717 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1718 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1719 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1720 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1721 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1722 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1723 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1724 * I am cleaning dirty pages from some other bdi. */
1725 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1726 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1727 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1728 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1729 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1730 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1731 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1734 * Only the _current_ task can read/write to tsk->flags, but other
1735 * tasks can access tsk->flags in readonly mode for example
1736 * with tsk_used_math (like during threaded core dumping).
1737 * There is however an exception to this rule during ptrace
1738 * or during fork: the ptracer task is allowed to write to the
1739 * child->flags of its traced child (same goes for fork, the parent
1740 * can write to the child->flags), because we're guaranteed the
1741 * child is not running and in turn not changing child->flags
1742 * at the same time the parent does it.
1744 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1745 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1746 #define clear_used_math() clear_stopped_child_used_math(current)
1747 #define set_used_math() set_stopped_child_used_math(current)
1749 #define conditional_stopped_child_used_math(condition, child) \
1750 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1752 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1754 #define copy_to_stopped_child_used_math(child) \
1755 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1757 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1758 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1759 #define used_math() tsk_used_math(current)
1761 static __always_inline bool is_percpu_thread(void)
1764 return (current->flags & PF_NO_SETAFFINITY) &&
1765 (current->nr_cpus_allowed == 1);
1771 /* Per-process atomic flags. */
1772 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1773 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1774 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1775 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1776 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1777 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1778 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1779 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1781 #define TASK_PFA_TEST(name, func) \
1782 static inline bool task_##func(struct task_struct *p) \
1783 { return test_bit(PFA_##name, &p->atomic_flags); }
1785 #define TASK_PFA_SET(name, func) \
1786 static inline void task_set_##func(struct task_struct *p) \
1787 { set_bit(PFA_##name, &p->atomic_flags); }
1789 #define TASK_PFA_CLEAR(name, func) \
1790 static inline void task_clear_##func(struct task_struct *p) \
1791 { clear_bit(PFA_##name, &p->atomic_flags); }
1793 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1794 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1796 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1797 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1798 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1800 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1801 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1802 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1804 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1805 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1806 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1808 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1809 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1810 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1812 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1813 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1815 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1816 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1817 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1819 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1820 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1823 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1825 current->flags &= ~flags;
1826 current->flags |= orig_flags & flags;
1829 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1830 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effective_cpus);
1832 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1833 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1834 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1835 extern void release_user_cpus_ptr(struct task_struct *p);
1836 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1837 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1838 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1840 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1843 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1845 if (!cpumask_test_cpu(0, new_mask))
1849 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1851 if (src->user_cpus_ptr)
1855 static inline void release_user_cpus_ptr(struct task_struct *p)
1857 WARN_ON(p->user_cpus_ptr);
1860 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1866 extern int yield_to(struct task_struct *p, bool preempt);
1867 extern void set_user_nice(struct task_struct *p, long nice);
1868 extern int task_prio(const struct task_struct *p);
1871 * task_nice - return the nice value of a given task.
1872 * @p: the task in question.
1874 * Return: The nice value [ -20 ... 0 ... 19 ].
1876 static inline int task_nice(const struct task_struct *p)
1878 return PRIO_TO_NICE((p)->static_prio);
1881 extern int can_nice(const struct task_struct *p, const int nice);
1882 extern int task_curr(const struct task_struct *p);
1883 extern int idle_cpu(int cpu);
1884 extern int available_idle_cpu(int cpu);
1885 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1886 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1887 extern void sched_set_fifo(struct task_struct *p);
1888 extern void sched_set_fifo_low(struct task_struct *p);
1889 extern void sched_set_normal(struct task_struct *p, int nice);
1890 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1891 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1892 extern struct task_struct *idle_task(int cpu);
1895 * is_idle_task - is the specified task an idle task?
1896 * @p: the task in question.
1898 * Return: 1 if @p is an idle task. 0 otherwise.
1900 static __always_inline bool is_idle_task(const struct task_struct *p)
1902 return !!(p->flags & PF_IDLE);
1905 extern struct task_struct *curr_task(int cpu);
1906 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1910 union thread_union {
1911 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1912 struct task_struct task;
1914 #ifndef CONFIG_THREAD_INFO_IN_TASK
1915 struct thread_info thread_info;
1917 unsigned long stack[THREAD_SIZE/sizeof(long)];
1920 #ifndef CONFIG_THREAD_INFO_IN_TASK
1921 extern struct thread_info init_thread_info;
1924 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1926 #ifdef CONFIG_THREAD_INFO_IN_TASK
1927 # define task_thread_info(task) (&(task)->thread_info)
1928 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1929 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1933 * find a task by one of its numerical ids
1935 * find_task_by_pid_ns():
1936 * finds a task by its pid in the specified namespace
1937 * find_task_by_vpid():
1938 * finds a task by its virtual pid
1940 * see also find_vpid() etc in include/linux/pid.h
1943 extern struct task_struct *find_task_by_vpid(pid_t nr);
1944 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1947 * find a task by its virtual pid and get the task struct
1949 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1951 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1952 extern int wake_up_process(struct task_struct *tsk);
1953 extern void wake_up_new_task(struct task_struct *tsk);
1956 extern void kick_process(struct task_struct *tsk);
1958 static inline void kick_process(struct task_struct *tsk) { }
1961 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1963 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1965 __set_task_comm(tsk, from, false);
1968 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1969 #define get_task_comm(buf, tsk) ({ \
1970 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1971 __get_task_comm(buf, sizeof(buf), tsk); \
1975 static __always_inline void scheduler_ipi(void)
1978 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1979 * TIF_NEED_RESCHED remotely (for the first time) will also send
1982 preempt_fold_need_resched();
1984 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1986 static inline void scheduler_ipi(void) { }
1987 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1994 * Set thread flags in other task's structures.
1995 * See asm/thread_info.h for TIF_xxxx flags available:
1997 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1999 set_ti_thread_flag(task_thread_info(tsk), flag);
2002 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2004 clear_ti_thread_flag(task_thread_info(tsk), flag);
2007 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
2010 update_ti_thread_flag(task_thread_info(tsk), flag, value);
2013 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2015 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2018 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2020 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2023 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2025 return test_ti_thread_flag(task_thread_info(tsk), flag);
2028 static inline void set_tsk_need_resched(struct task_struct *tsk)
2030 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2033 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2035 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2038 static inline int test_tsk_need_resched(struct task_struct *tsk)
2040 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2044 * cond_resched() and cond_resched_lock(): latency reduction via
2045 * explicit rescheduling in places that are safe. The return
2046 * value indicates whether a reschedule was done in fact.
2047 * cond_resched_lock() will drop the spinlock before scheduling,
2049 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2050 extern int __cond_resched(void);
2052 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2054 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2056 static __always_inline int _cond_resched(void)
2058 return static_call_mod(cond_resched)();
2061 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2062 extern int dynamic_cond_resched(void);
2064 static __always_inline int _cond_resched(void)
2066 return dynamic_cond_resched();
2071 static inline int _cond_resched(void)
2073 return __cond_resched();
2076 #endif /* CONFIG_PREEMPT_DYNAMIC */
2080 static inline int _cond_resched(void) { return 0; }
2082 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
2084 #define cond_resched() ({ \
2085 __might_resched(__FILE__, __LINE__, 0); \
2089 extern int __cond_resched_lock(spinlock_t *lock);
2090 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2091 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2093 #define MIGHT_RESCHED_RCU_SHIFT 8
2094 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2096 #ifndef CONFIG_PREEMPT_RT
2098 * Non RT kernels have an elevated preempt count due to the held lock,
2099 * but are not allowed to be inside a RCU read side critical section
2101 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2104 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2105 * cond_resched*lock() has to take that into account because it checks for
2106 * preempt_count() and rcu_preempt_depth().
2108 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2109 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2112 #define cond_resched_lock(lock) ({ \
2113 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2114 __cond_resched_lock(lock); \
2117 #define cond_resched_rwlock_read(lock) ({ \
2118 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2119 __cond_resched_rwlock_read(lock); \
2122 #define cond_resched_rwlock_write(lock) ({ \
2123 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2124 __cond_resched_rwlock_write(lock); \
2127 static inline void cond_resched_rcu(void)
2129 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2136 #ifdef CONFIG_PREEMPT_DYNAMIC
2138 extern bool preempt_model_none(void);
2139 extern bool preempt_model_voluntary(void);
2140 extern bool preempt_model_full(void);
2144 static inline bool preempt_model_none(void)
2146 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2148 static inline bool preempt_model_voluntary(void)
2150 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2152 static inline bool preempt_model_full(void)
2154 return IS_ENABLED(CONFIG_PREEMPT);
2159 static inline bool preempt_model_rt(void)
2161 return IS_ENABLED(CONFIG_PREEMPT_RT);
2165 * Does the preemption model allow non-cooperative preemption?
2167 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2168 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2169 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2170 * PREEMPT_NONE model.
2172 static inline bool preempt_model_preemptible(void)
2174 return preempt_model_full() || preempt_model_rt();
2178 * Does a critical section need to be broken due to another
2179 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
2180 * but a general need for low latency)
2182 static inline int spin_needbreak(spinlock_t *lock)
2184 #ifdef CONFIG_PREEMPTION
2185 return spin_is_contended(lock);
2192 * Check if a rwlock is contended.
2193 * Returns non-zero if there is another task waiting on the rwlock.
2194 * Returns zero if the lock is not contended or the system / underlying
2195 * rwlock implementation does not support contention detection.
2196 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2199 static inline int rwlock_needbreak(rwlock_t *lock)
2201 #ifdef CONFIG_PREEMPTION
2202 return rwlock_is_contended(lock);
2208 static __always_inline bool need_resched(void)
2210 return unlikely(tif_need_resched());
2214 * Wrappers for p->thread_info->cpu access. No-op on UP.
2218 static inline unsigned int task_cpu(const struct task_struct *p)
2220 return READ_ONCE(task_thread_info(p)->cpu);
2223 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2227 static inline unsigned int task_cpu(const struct task_struct *p)
2232 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2236 #endif /* CONFIG_SMP */
2238 extern bool sched_task_on_rq(struct task_struct *p);
2239 extern unsigned long get_wchan(struct task_struct *p);
2240 extern struct task_struct *cpu_curr_snapshot(int cpu);
2243 * In order to reduce various lock holder preemption latencies provide an
2244 * interface to see if a vCPU is currently running or not.
2246 * This allows us to terminate optimistic spin loops and block, analogous to
2247 * the native optimistic spin heuristic of testing if the lock owner task is
2250 #ifndef vcpu_is_preempted
2251 static inline bool vcpu_is_preempted(int cpu)
2257 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2258 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2260 #ifndef TASK_SIZE_OF
2261 #define TASK_SIZE_OF(tsk) TASK_SIZE
2265 static inline bool owner_on_cpu(struct task_struct *owner)
2268 * As lock holder preemption issue, we both skip spinning if
2269 * task is not on cpu or its cpu is preempted
2271 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2274 /* Returns effective CPU energy utilization, as seen by the scheduler */
2275 unsigned long sched_cpu_util(int cpu);
2276 #endif /* CONFIG_SMP */
2281 * Map the event mask on the user-space ABI enum rseq_cs_flags
2282 * for direct mask checks.
2284 enum rseq_event_mask_bits {
2285 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2286 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2287 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2290 enum rseq_event_mask {
2291 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2292 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2293 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2296 static inline void rseq_set_notify_resume(struct task_struct *t)
2299 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2302 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2304 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2305 struct pt_regs *regs)
2308 __rseq_handle_notify_resume(ksig, regs);
2311 static inline void rseq_signal_deliver(struct ksignal *ksig,
2312 struct pt_regs *regs)
2315 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2317 rseq_handle_notify_resume(ksig, regs);
2320 /* rseq_preempt() requires preemption to be disabled. */
2321 static inline void rseq_preempt(struct task_struct *t)
2323 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2324 rseq_set_notify_resume(t);
2327 /* rseq_migrate() requires preemption to be disabled. */
2328 static inline void rseq_migrate(struct task_struct *t)
2330 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2331 rseq_set_notify_resume(t);
2335 * If parent process has a registered restartable sequences area, the
2336 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2338 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2340 if (clone_flags & CLONE_VM) {
2343 t->rseq_event_mask = 0;
2345 t->rseq = current->rseq;
2346 t->rseq_sig = current->rseq_sig;
2347 t->rseq_event_mask = current->rseq_event_mask;
2351 static inline void rseq_execve(struct task_struct *t)
2355 t->rseq_event_mask = 0;
2360 static inline void rseq_set_notify_resume(struct task_struct *t)
2363 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2364 struct pt_regs *regs)
2367 static inline void rseq_signal_deliver(struct ksignal *ksig,
2368 struct pt_regs *regs)
2371 static inline void rseq_preempt(struct task_struct *t)
2374 static inline void rseq_migrate(struct task_struct *t)
2377 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2380 static inline void rseq_execve(struct task_struct *t)
2386 #ifdef CONFIG_DEBUG_RSEQ
2388 void rseq_syscall(struct pt_regs *regs);
2392 static inline void rseq_syscall(struct pt_regs *regs)
2398 #ifdef CONFIG_SCHED_CORE
2399 extern void sched_core_free(struct task_struct *tsk);
2400 extern void sched_core_fork(struct task_struct *p);
2401 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2402 unsigned long uaddr);
2404 static inline void sched_core_free(struct task_struct *tsk) { }
2405 static inline void sched_core_fork(struct task_struct *p) { }
2408 extern void sched_set_stop_task(int cpu, struct task_struct *stop);