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/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/irqflags.h>
22 #include <linux/seccomp.h>
23 #include <linux/nodemask.h>
24 #include <linux/rcupdate.h>
25 #include <linux/refcount.h>
26 #include <linux/resource.h>
27 #include <linux/latencytop.h>
28 #include <linux/sched/prio.h>
29 #include <linux/sched/types.h>
30 #include <linux/signal_types.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 /* task_struct member predeclarations (sorted alphabetically): */
40 struct backing_dev_info;
43 struct capture_control;
46 struct futex_pi_state;
51 struct perf_event_context;
53 struct pipe_inode_info;
56 struct robust_list_head;
62 struct sighand_struct;
64 struct task_delay_info;
69 * Task state bitmask. NOTE! These bits are also
70 * encoded in fs/proc/array.c: get_task_state().
72 * We have two separate sets of flags: task->state
73 * is about runnability, while task->exit_state are
74 * about the task exiting. Confusing, but this way
75 * modifying one set can't modify the other one by
79 /* Used in tsk->state: */
80 #define TASK_RUNNING 0x0000
81 #define TASK_INTERRUPTIBLE 0x0001
82 #define TASK_UNINTERRUPTIBLE 0x0002
83 #define __TASK_STOPPED 0x0004
84 #define __TASK_TRACED 0x0008
85 /* Used in tsk->exit_state: */
86 #define EXIT_DEAD 0x0010
87 #define EXIT_ZOMBIE 0x0020
88 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
89 /* Used in tsk->state again: */
90 #define TASK_PARKED 0x0040
91 #define TASK_DEAD 0x0080
92 #define TASK_WAKEKILL 0x0100
93 #define TASK_WAKING 0x0200
94 #define TASK_NOLOAD 0x0400
95 #define TASK_NEW 0x0800
96 #define TASK_STATE_MAX 0x1000
98 /* Convenience macros for the sake of set_current_state: */
99 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
100 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
101 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
103 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
105 /* Convenience macros for the sake of wake_up(): */
106 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
108 /* get_task_state(): */
109 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
110 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
111 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
114 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
116 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
118 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
120 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
123 * Special states are those that do not use the normal wait-loop pattern. See
124 * the comment with set_special_state().
126 #define is_special_task_state(state) \
127 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
129 #define __set_current_state(state_value) \
131 WARN_ON_ONCE(is_special_task_state(state_value));\
132 current->task_state_change = _THIS_IP_; \
133 current->state = (state_value); \
136 #define set_current_state(state_value) \
138 WARN_ON_ONCE(is_special_task_state(state_value));\
139 current->task_state_change = _THIS_IP_; \
140 smp_store_mb(current->state, (state_value)); \
143 #define set_special_state(state_value) \
145 unsigned long flags; /* may shadow */ \
146 WARN_ON_ONCE(!is_special_task_state(state_value)); \
147 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
148 current->task_state_change = _THIS_IP_; \
149 current->state = (state_value); \
150 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
154 * set_current_state() includes a barrier so that the write of current->state
155 * is correctly serialised wrt the caller's subsequent test of whether to
159 * set_current_state(TASK_UNINTERRUPTIBLE);
165 * __set_current_state(TASK_RUNNING);
167 * If the caller does not need such serialisation (because, for instance, the
168 * CONDITION test and condition change and wakeup are under the same lock) then
169 * use __set_current_state().
171 * The above is typically ordered against the wakeup, which does:
174 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
176 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
177 * accessing p->state.
179 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
180 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
181 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
183 * However, with slightly different timing the wakeup TASK_RUNNING store can
184 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
185 * a problem either because that will result in one extra go around the loop
186 * and our @cond test will save the day.
188 * Also see the comments of try_to_wake_up().
190 #define __set_current_state(state_value) \
191 current->state = (state_value)
193 #define set_current_state(state_value) \
194 smp_store_mb(current->state, (state_value))
197 * set_special_state() should be used for those states when the blocking task
198 * can not use the regular condition based wait-loop. In that case we must
199 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
200 * will not collide with our state change.
202 #define set_special_state(state_value) \
204 unsigned long flags; /* may shadow */ \
205 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
206 current->state = (state_value); \
207 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
212 /* Task command name length: */
213 #define TASK_COMM_LEN 16
215 extern void scheduler_tick(void);
217 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
219 extern long schedule_timeout(long timeout);
220 extern long schedule_timeout_interruptible(long timeout);
221 extern long schedule_timeout_killable(long timeout);
222 extern long schedule_timeout_uninterruptible(long timeout);
223 extern long schedule_timeout_idle(long timeout);
224 asmlinkage void schedule(void);
225 extern void schedule_preempt_disabled(void);
226 asmlinkage void preempt_schedule_irq(void);
228 extern int __must_check io_schedule_prepare(void);
229 extern void io_schedule_finish(int token);
230 extern long io_schedule_timeout(long timeout);
231 extern void io_schedule(void);
234 * struct prev_cputime - snapshot of system and user cputime
235 * @utime: time spent in user mode
236 * @stime: time spent in system mode
237 * @lock: protects the above two fields
239 * Stores previous user/system time values such that we can guarantee
242 struct prev_cputime {
243 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
251 /* Task is sleeping or running in a CPU with VTIME inactive: */
255 /* Task runs in kernelspace in a CPU with VTIME active: */
257 /* Task runs in userspace in a CPU with VTIME active: */
259 /* Task runs as guests in a CPU with VTIME active: */
265 unsigned long long starttime;
266 enum vtime_state state;
274 * Utilization clamp constraints.
275 * @UCLAMP_MIN: Minimum utilization
276 * @UCLAMP_MAX: Maximum utilization
277 * @UCLAMP_CNT: Utilization clamp constraints count
286 extern struct root_domain def_root_domain;
287 extern struct mutex sched_domains_mutex;
291 #ifdef CONFIG_SCHED_INFO
292 /* Cumulative counters: */
294 /* # of times we have run on this CPU: */
295 unsigned long pcount;
297 /* Time spent waiting on a runqueue: */
298 unsigned long long run_delay;
302 /* When did we last run on a CPU? */
303 unsigned long long last_arrival;
305 /* When were we last queued to run? */
306 unsigned long long last_queued;
308 #endif /* CONFIG_SCHED_INFO */
312 * Integer metrics need fixed point arithmetic, e.g., sched/fair
313 * has a few: load, load_avg, util_avg, freq, and capacity.
315 * We define a basic fixed point arithmetic range, and then formalize
316 * all these metrics based on that basic range.
318 # define SCHED_FIXEDPOINT_SHIFT 10
319 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
321 /* Increase resolution of cpu_capacity calculations */
322 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
323 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
326 unsigned long weight;
331 * struct util_est - Estimation utilization of FAIR tasks
332 * @enqueued: instantaneous estimated utilization of a task/cpu
333 * @ewma: the Exponential Weighted Moving Average (EWMA)
334 * utilization of a task
336 * Support data structure to track an Exponential Weighted Moving Average
337 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
338 * average each time a task completes an activation. Sample's weight is chosen
339 * so that the EWMA will be relatively insensitive to transient changes to the
342 * The enqueued attribute has a slightly different meaning for tasks and cpus:
343 * - task: the task's util_avg at last task dequeue time
344 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
345 * Thus, the util_est.enqueued of a task represents the contribution on the
346 * estimated utilization of the CPU where that task is currently enqueued.
348 * Only for tasks we track a moving average of the past instantaneous
349 * estimated utilization. This allows to absorb sporadic drops in utilization
350 * of an otherwise almost periodic task.
353 unsigned int enqueued;
355 #define UTIL_EST_WEIGHT_SHIFT 2
356 } __attribute__((__aligned__(sizeof(u64))));
359 * The load/runnable/util_avg accumulates an infinite geometric series
360 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
362 * [load_avg definition]
364 * load_avg = runnable% * scale_load_down(load)
366 * [runnable_avg definition]
368 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
370 * [util_avg definition]
372 * util_avg = running% * SCHED_CAPACITY_SCALE
374 * where runnable% is the time ratio that a sched_entity is runnable and
375 * running% the time ratio that a sched_entity is running.
377 * For cfs_rq, they are the aggregated values of all runnable and blocked
380 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
381 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
382 * for computing those signals (see update_rq_clock_pelt())
384 * N.B., the above ratios (runnable% and running%) themselves are in the
385 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
386 * to as large a range as necessary. This is for example reflected by
387 * util_avg's SCHED_CAPACITY_SCALE.
391 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
392 * with the highest load (=88761), always runnable on a single cfs_rq,
393 * and should not overflow as the number already hits PID_MAX_LIMIT.
395 * For all other cases (including 32-bit kernels), struct load_weight's
396 * weight will overflow first before we do, because:
398 * Max(load_avg) <= Max(load.weight)
400 * Then it is the load_weight's responsibility to consider overflow
404 u64 last_update_time;
409 unsigned long load_avg;
410 unsigned long runnable_avg;
411 unsigned long util_avg;
412 struct util_est util_est;
413 } ____cacheline_aligned;
415 struct sched_statistics {
416 #ifdef CONFIG_SCHEDSTATS
426 s64 sum_sleep_runtime;
433 u64 nr_migrations_cold;
434 u64 nr_failed_migrations_affine;
435 u64 nr_failed_migrations_running;
436 u64 nr_failed_migrations_hot;
437 u64 nr_forced_migrations;
441 u64 nr_wakeups_migrate;
442 u64 nr_wakeups_local;
443 u64 nr_wakeups_remote;
444 u64 nr_wakeups_affine;
445 u64 nr_wakeups_affine_attempts;
446 u64 nr_wakeups_passive;
451 struct sched_entity {
452 /* For load-balancing: */
453 struct load_weight load;
454 struct rb_node run_node;
455 struct list_head group_node;
459 u64 sum_exec_runtime;
461 u64 prev_sum_exec_runtime;
465 struct sched_statistics statistics;
467 #ifdef CONFIG_FAIR_GROUP_SCHED
469 struct sched_entity *parent;
470 /* rq on which this entity is (to be) queued: */
471 struct cfs_rq *cfs_rq;
472 /* rq "owned" by this entity/group: */
474 /* cached value of my_q->h_nr_running */
475 unsigned long runnable_weight;
480 * Per entity load average tracking.
482 * Put into separate cache line so it does not
483 * collide with read-mostly values above.
485 struct sched_avg avg;
489 struct sched_rt_entity {
490 struct list_head run_list;
491 unsigned long timeout;
492 unsigned long watchdog_stamp;
493 unsigned int time_slice;
494 unsigned short on_rq;
495 unsigned short on_list;
497 struct sched_rt_entity *back;
498 #ifdef CONFIG_RT_GROUP_SCHED
499 struct sched_rt_entity *parent;
500 /* rq on which this entity is (to be) queued: */
502 /* rq "owned" by this entity/group: */
505 } __randomize_layout;
507 struct sched_dl_entity {
508 struct rb_node rb_node;
511 * Original scheduling parameters. Copied here from sched_attr
512 * during sched_setattr(), they will remain the same until
513 * the next sched_setattr().
515 u64 dl_runtime; /* Maximum runtime for each instance */
516 u64 dl_deadline; /* Relative deadline of each instance */
517 u64 dl_period; /* Separation of two instances (period) */
518 u64 dl_bw; /* dl_runtime / dl_period */
519 u64 dl_density; /* dl_runtime / dl_deadline */
522 * Actual scheduling parameters. Initialized with the values above,
523 * they are continuously updated during task execution. Note that
524 * the remaining runtime could be < 0 in case we are in overrun.
526 s64 runtime; /* Remaining runtime for this instance */
527 u64 deadline; /* Absolute deadline for this instance */
528 unsigned int flags; /* Specifying the scheduler behaviour */
533 * @dl_throttled tells if we exhausted the runtime. If so, the
534 * task has to wait for a replenishment to be performed at the
535 * next firing of dl_timer.
537 * @dl_boosted tells if we are boosted due to DI. If so we are
538 * outside bandwidth enforcement mechanism (but only until we
539 * exit the critical section);
541 * @dl_yielded tells if task gave up the CPU before consuming
542 * all its available runtime during the last job.
544 * @dl_non_contending tells if the task is inactive while still
545 * contributing to the active utilization. In other words, it
546 * indicates if the inactive timer has been armed and its handler
547 * has not been executed yet. This flag is useful to avoid race
548 * conditions between the inactive timer handler and the wakeup
551 * @dl_overrun tells if the task asked to be informed about runtime
554 unsigned int dl_throttled : 1;
555 unsigned int dl_yielded : 1;
556 unsigned int dl_non_contending : 1;
557 unsigned int dl_overrun : 1;
560 * Bandwidth enforcement timer. Each -deadline task has its
561 * own bandwidth to be enforced, thus we need one timer per task.
563 struct hrtimer dl_timer;
566 * Inactive timer, responsible for decreasing the active utilization
567 * at the "0-lag time". When a -deadline task blocks, it contributes
568 * to GRUB's active utilization until the "0-lag time", hence a
569 * timer is needed to decrease the active utilization at the correct
572 struct hrtimer inactive_timer;
574 #ifdef CONFIG_RT_MUTEXES
576 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
577 * pi_se points to the donor, otherwise points to the dl_se it belongs
578 * to (the original one/itself).
580 struct sched_dl_entity *pi_se;
584 #ifdef CONFIG_UCLAMP_TASK
585 /* Number of utilization clamp buckets (shorter alias) */
586 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
589 * Utilization clamp for a scheduling entity
590 * @value: clamp value "assigned" to a se
591 * @bucket_id: bucket index corresponding to the "assigned" value
592 * @active: the se is currently refcounted in a rq's bucket
593 * @user_defined: the requested clamp value comes from user-space
595 * The bucket_id is the index of the clamp bucket matching the clamp value
596 * which is pre-computed and stored to avoid expensive integer divisions from
599 * The active bit is set whenever a task has got an "effective" value assigned,
600 * which can be different from the clamp value "requested" from user-space.
601 * This allows to know a task is refcounted in the rq's bucket corresponding
602 * to the "effective" bucket_id.
604 * The user_defined bit is set whenever a task has got a task-specific clamp
605 * value requested from userspace, i.e. the system defaults apply to this task
606 * just as a restriction. This allows to relax default clamps when a less
607 * restrictive task-specific value has been requested, thus allowing to
608 * implement a "nice" semantic. For example, a task running with a 20%
609 * default boost can still drop its own boosting to 0%.
612 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
613 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
614 unsigned int active : 1;
615 unsigned int user_defined : 1;
617 #endif /* CONFIG_UCLAMP_TASK */
623 u8 exp_hint; /* Hint for performance. */
624 u8 need_mb; /* Readers need smp_mb(). */
626 u32 s; /* Set of bits. */
629 enum perf_event_task_context {
630 perf_invalid_context = -1,
633 perf_nr_task_contexts,
637 struct wake_q_node *next;
641 #ifdef CONFIG_THREAD_INFO_IN_TASK
643 * For reasons of header soup (see current_thread_info()), this
644 * must be the first element of task_struct.
646 struct thread_info thread_info;
648 /* -1 unrunnable, 0 runnable, >0 stopped: */
652 * This begins the randomizable portion of task_struct. Only
653 * scheduling-critical items should be added above here.
655 randomized_struct_fields_start
659 /* Per task flags (PF_*), defined further below: */
665 struct __call_single_node wake_entry;
666 #ifdef CONFIG_THREAD_INFO_IN_TASK
670 unsigned int wakee_flips;
671 unsigned long wakee_flip_decay_ts;
672 struct task_struct *last_wakee;
675 * recent_used_cpu is initially set as the last CPU used by a task
676 * that wakes affine another task. Waker/wakee relationships can
677 * push tasks around a CPU where each wakeup moves to the next one.
678 * Tracking a recently used CPU allows a quick search for a recently
679 * used CPU that may be idle.
689 unsigned int rt_priority;
691 const struct sched_class *sched_class;
692 struct sched_entity se;
693 struct sched_rt_entity rt;
694 #ifdef CONFIG_CGROUP_SCHED
695 struct task_group *sched_task_group;
697 struct sched_dl_entity dl;
699 #ifdef CONFIG_UCLAMP_TASK
701 * Clamp values requested for a scheduling entity.
702 * Must be updated with task_rq_lock() held.
704 struct uclamp_se uclamp_req[UCLAMP_CNT];
706 * Effective clamp values used for a scheduling entity.
707 * Must be updated with task_rq_lock() held.
709 struct uclamp_se uclamp[UCLAMP_CNT];
712 #ifdef CONFIG_PREEMPT_NOTIFIERS
713 /* List of struct preempt_notifier: */
714 struct hlist_head preempt_notifiers;
717 #ifdef CONFIG_BLK_DEV_IO_TRACE
718 unsigned int btrace_seq;
723 const cpumask_t *cpus_ptr;
726 #ifdef CONFIG_PREEMPT_RCU
727 int rcu_read_lock_nesting;
728 union rcu_special rcu_read_unlock_special;
729 struct list_head rcu_node_entry;
730 struct rcu_node *rcu_blocked_node;
731 #endif /* #ifdef CONFIG_PREEMPT_RCU */
733 #ifdef CONFIG_TASKS_RCU
734 unsigned long rcu_tasks_nvcsw;
735 u8 rcu_tasks_holdout;
737 int rcu_tasks_idle_cpu;
738 struct list_head rcu_tasks_holdout_list;
739 #endif /* #ifdef CONFIG_TASKS_RCU */
741 #ifdef CONFIG_TASKS_TRACE_RCU
742 int trc_reader_nesting;
744 union rcu_special trc_reader_special;
745 bool trc_reader_checked;
746 struct list_head trc_holdout_list;
747 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
749 struct sched_info sched_info;
751 struct list_head tasks;
753 struct plist_node pushable_tasks;
754 struct rb_node pushable_dl_tasks;
757 struct mm_struct *mm;
758 struct mm_struct *active_mm;
760 /* Per-thread vma caching: */
761 struct vmacache vmacache;
763 #ifdef SPLIT_RSS_COUNTING
764 struct task_rss_stat rss_stat;
769 /* The signal sent when the parent dies: */
771 /* JOBCTL_*, siglock protected: */
772 unsigned long jobctl;
774 /* Used for emulating ABI behavior of previous Linux versions: */
775 unsigned int personality;
777 /* Scheduler bits, serialized by scheduler locks: */
778 unsigned sched_reset_on_fork:1;
779 unsigned sched_contributes_to_load:1;
780 unsigned sched_migrated:1;
782 unsigned sched_psi_wake_requeue:1;
785 /* Force alignment to the next boundary: */
788 /* Unserialized, strictly 'current' */
791 * This field must not be in the scheduler word above due to wakelist
792 * queueing no longer being serialized by p->on_cpu. However:
795 * schedule() if (p->on_rq && ..) // false
796 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
797 * deactivate_task() ttwu_queue_wakelist())
798 * p->on_rq = 0; p->sched_remote_wakeup = Y;
800 * guarantees all stores of 'current' are visible before
801 * ->sched_remote_wakeup gets used, so it can be in this word.
803 unsigned sched_remote_wakeup:1;
805 /* Bit to tell LSMs we're in execve(): */
806 unsigned in_execve:1;
807 unsigned in_iowait:1;
808 #ifndef TIF_RESTORE_SIGMASK
809 unsigned restore_sigmask:1;
812 unsigned in_user_fault:1;
814 #ifdef CONFIG_COMPAT_BRK
815 unsigned brk_randomized:1;
817 #ifdef CONFIG_CGROUPS
818 /* disallow userland-initiated cgroup migration */
819 unsigned no_cgroup_migration:1;
820 /* task is frozen/stopped (used by the cgroup freezer) */
823 #ifdef CONFIG_BLK_CGROUP
824 unsigned use_memdelay:1;
827 /* Stalled due to lack of memory */
828 unsigned in_memstall:1;
831 unsigned long atomic_flags; /* Flags requiring atomic access. */
833 struct restart_block restart_block;
838 #ifdef CONFIG_STACKPROTECTOR
839 /* Canary value for the -fstack-protector GCC feature: */
840 unsigned long stack_canary;
843 * Pointers to the (original) parent process, youngest child, younger sibling,
844 * older sibling, respectively. (p->father can be replaced with
845 * p->real_parent->pid)
848 /* Real parent process: */
849 struct task_struct __rcu *real_parent;
851 /* Recipient of SIGCHLD, wait4() reports: */
852 struct task_struct __rcu *parent;
855 * Children/sibling form the list of natural children:
857 struct list_head children;
858 struct list_head sibling;
859 struct task_struct *group_leader;
862 * 'ptraced' is the list of tasks this task is using ptrace() on.
864 * This includes both natural children and PTRACE_ATTACH targets.
865 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
867 struct list_head ptraced;
868 struct list_head ptrace_entry;
870 /* PID/PID hash table linkage. */
871 struct pid *thread_pid;
872 struct hlist_node pid_links[PIDTYPE_MAX];
873 struct list_head thread_group;
874 struct list_head thread_node;
876 struct completion *vfork_done;
878 /* CLONE_CHILD_SETTID: */
879 int __user *set_child_tid;
881 /* CLONE_CHILD_CLEARTID: */
882 int __user *clear_child_tid;
886 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
891 struct prev_cputime prev_cputime;
892 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
896 #ifdef CONFIG_NO_HZ_FULL
897 atomic_t tick_dep_mask;
899 /* Context switch counts: */
901 unsigned long nivcsw;
903 /* Monotonic time in nsecs: */
906 /* Boot based time in nsecs: */
909 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
910 unsigned long min_flt;
911 unsigned long maj_flt;
913 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
914 struct posix_cputimers posix_cputimers;
916 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
917 struct posix_cputimers_work posix_cputimers_work;
920 /* Process credentials: */
922 /* Tracer's credentials at attach: */
923 const struct cred __rcu *ptracer_cred;
925 /* Objective and real subjective task credentials (COW): */
926 const struct cred __rcu *real_cred;
928 /* Effective (overridable) subjective task credentials (COW): */
929 const struct cred __rcu *cred;
932 /* Cached requested key. */
933 struct key *cached_requested_key;
937 * executable name, excluding path.
939 * - normally initialized setup_new_exec()
940 * - access it with [gs]et_task_comm()
941 * - lock it with task_lock()
943 char comm[TASK_COMM_LEN];
945 struct nameidata *nameidata;
947 #ifdef CONFIG_SYSVIPC
948 struct sysv_sem sysvsem;
949 struct sysv_shm sysvshm;
951 #ifdef CONFIG_DETECT_HUNG_TASK
952 unsigned long last_switch_count;
953 unsigned long last_switch_time;
955 /* Filesystem information: */
956 struct fs_struct *fs;
958 /* Open file information: */
959 struct files_struct *files;
961 #ifdef CONFIG_IO_URING
962 struct io_uring_task *io_uring;
966 struct nsproxy *nsproxy;
968 /* Signal handlers: */
969 struct signal_struct *signal;
970 struct sighand_struct __rcu *sighand;
972 sigset_t real_blocked;
973 /* Restored if set_restore_sigmask() was used: */
974 sigset_t saved_sigmask;
975 struct sigpending pending;
976 unsigned long sas_ss_sp;
978 unsigned int sas_ss_flags;
980 struct callback_head *task_works;
983 #ifdef CONFIG_AUDITSYSCALL
984 struct audit_context *audit_context;
987 unsigned int sessionid;
989 struct seccomp seccomp;
991 /* Thread group tracking: */
995 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
996 spinlock_t alloc_lock;
998 /* Protection of the PI data structures: */
999 raw_spinlock_t pi_lock;
1001 struct wake_q_node wake_q;
1003 #ifdef CONFIG_RT_MUTEXES
1004 /* PI waiters blocked on a rt_mutex held by this task: */
1005 struct rb_root_cached pi_waiters;
1006 /* Updated under owner's pi_lock and rq lock */
1007 struct task_struct *pi_top_task;
1008 /* Deadlock detection and priority inheritance handling: */
1009 struct rt_mutex_waiter *pi_blocked_on;
1012 #ifdef CONFIG_DEBUG_MUTEXES
1013 /* Mutex deadlock detection: */
1014 struct mutex_waiter *blocked_on;
1017 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1018 int non_block_count;
1021 #ifdef CONFIG_TRACE_IRQFLAGS
1022 struct irqtrace_events irqtrace;
1023 unsigned int hardirq_threaded;
1024 u64 hardirq_chain_key;
1025 int softirqs_enabled;
1026 int softirq_context;
1030 #ifdef CONFIG_LOCKDEP
1031 # define MAX_LOCK_DEPTH 48UL
1034 unsigned int lockdep_recursion;
1035 struct held_lock held_locks[MAX_LOCK_DEPTH];
1038 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1039 unsigned int in_ubsan;
1042 /* Journalling filesystem info: */
1045 /* Stacked block device info: */
1046 struct bio_list *bio_list;
1049 /* Stack plugging: */
1050 struct blk_plug *plug;
1054 struct reclaim_state *reclaim_state;
1056 struct backing_dev_info *backing_dev_info;
1058 struct io_context *io_context;
1060 #ifdef CONFIG_COMPACTION
1061 struct capture_control *capture_control;
1064 unsigned long ptrace_message;
1065 kernel_siginfo_t *last_siginfo;
1067 struct task_io_accounting ioac;
1069 /* Pressure stall state */
1070 unsigned int psi_flags;
1072 #ifdef CONFIG_TASK_XACCT
1073 /* Accumulated RSS usage: */
1075 /* Accumulated virtual memory usage: */
1077 /* stime + utime since last update: */
1080 #ifdef CONFIG_CPUSETS
1081 /* Protected by ->alloc_lock: */
1082 nodemask_t mems_allowed;
1083 /* Seqence number to catch updates: */
1084 seqcount_spinlock_t mems_allowed_seq;
1085 int cpuset_mem_spread_rotor;
1086 int cpuset_slab_spread_rotor;
1088 #ifdef CONFIG_CGROUPS
1089 /* Control Group info protected by css_set_lock: */
1090 struct css_set __rcu *cgroups;
1091 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1092 struct list_head cg_list;
1094 #ifdef CONFIG_X86_CPU_RESCTRL
1099 struct robust_list_head __user *robust_list;
1100 #ifdef CONFIG_COMPAT
1101 struct compat_robust_list_head __user *compat_robust_list;
1103 struct list_head pi_state_list;
1104 struct futex_pi_state *pi_state_cache;
1105 struct mutex futex_exit_mutex;
1106 unsigned int futex_state;
1108 #ifdef CONFIG_PERF_EVENTS
1109 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1110 struct mutex perf_event_mutex;
1111 struct list_head perf_event_list;
1113 #ifdef CONFIG_DEBUG_PREEMPT
1114 unsigned long preempt_disable_ip;
1117 /* Protected by alloc_lock: */
1118 struct mempolicy *mempolicy;
1120 short pref_node_fork;
1122 #ifdef CONFIG_NUMA_BALANCING
1124 unsigned int numa_scan_period;
1125 unsigned int numa_scan_period_max;
1126 int numa_preferred_nid;
1127 unsigned long numa_migrate_retry;
1128 /* Migration stamp: */
1130 u64 last_task_numa_placement;
1131 u64 last_sum_exec_runtime;
1132 struct callback_head numa_work;
1135 * This pointer is only modified for current in syscall and
1136 * pagefault context (and for tasks being destroyed), so it can be read
1137 * from any of the following contexts:
1138 * - RCU read-side critical section
1139 * - current->numa_group from everywhere
1140 * - task's runqueue locked, task not running
1142 struct numa_group __rcu *numa_group;
1145 * numa_faults is an array split into four regions:
1146 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1147 * in this precise order.
1149 * faults_memory: Exponential decaying average of faults on a per-node
1150 * basis. Scheduling placement decisions are made based on these
1151 * counts. The values remain static for the duration of a PTE scan.
1152 * faults_cpu: Track the nodes the process was running on when a NUMA
1153 * hinting fault was incurred.
1154 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1155 * during the current scan window. When the scan completes, the counts
1156 * in faults_memory and faults_cpu decay and these values are copied.
1158 unsigned long *numa_faults;
1159 unsigned long total_numa_faults;
1162 * numa_faults_locality tracks if faults recorded during the last
1163 * scan window were remote/local or failed to migrate. The task scan
1164 * period is adapted based on the locality of the faults with different
1165 * weights depending on whether they were shared or private faults
1167 unsigned long numa_faults_locality[3];
1169 unsigned long numa_pages_migrated;
1170 #endif /* CONFIG_NUMA_BALANCING */
1173 struct rseq __user *rseq;
1176 * RmW on rseq_event_mask must be performed atomically
1177 * with respect to preemption.
1179 unsigned long rseq_event_mask;
1182 struct tlbflush_unmap_batch tlb_ubc;
1185 refcount_t rcu_users;
1186 struct rcu_head rcu;
1189 /* Cache last used pipe for splice(): */
1190 struct pipe_inode_info *splice_pipe;
1192 struct page_frag task_frag;
1194 #ifdef CONFIG_TASK_DELAY_ACCT
1195 struct task_delay_info *delays;
1198 #ifdef CONFIG_FAULT_INJECTION
1200 unsigned int fail_nth;
1203 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1204 * balance_dirty_pages() for a dirty throttling pause:
1207 int nr_dirtied_pause;
1208 /* Start of a write-and-pause period: */
1209 unsigned long dirty_paused_when;
1211 #ifdef CONFIG_LATENCYTOP
1212 int latency_record_count;
1213 struct latency_record latency_record[LT_SAVECOUNT];
1216 * Time slack values; these are used to round up poll() and
1217 * select() etc timeout values. These are in nanoseconds.
1220 u64 default_timer_slack_ns;
1223 unsigned int kasan_depth;
1227 struct kcsan_ctx kcsan_ctx;
1228 #ifdef CONFIG_TRACE_IRQFLAGS
1229 struct irqtrace_events kcsan_save_irqtrace;
1233 #if IS_ENABLED(CONFIG_KUNIT)
1234 struct kunit *kunit_test;
1237 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1238 /* Index of current stored address in ret_stack: */
1242 /* Stack of return addresses for return function tracing: */
1243 struct ftrace_ret_stack *ret_stack;
1245 /* Timestamp for last schedule: */
1246 unsigned long long ftrace_timestamp;
1249 * Number of functions that haven't been traced
1250 * because of depth overrun:
1252 atomic_t trace_overrun;
1254 /* Pause tracing: */
1255 atomic_t tracing_graph_pause;
1258 #ifdef CONFIG_TRACING
1259 /* State flags for use by tracers: */
1260 unsigned long trace;
1262 /* Bitmask and counter of trace recursion: */
1263 unsigned long trace_recursion;
1264 #endif /* CONFIG_TRACING */
1267 /* See kernel/kcov.c for more details. */
1269 /* Coverage collection mode enabled for this task (0 if disabled): */
1270 unsigned int kcov_mode;
1272 /* Size of the kcov_area: */
1273 unsigned int kcov_size;
1275 /* Buffer for coverage collection: */
1278 /* KCOV descriptor wired with this task or NULL: */
1281 /* KCOV common handle for remote coverage collection: */
1284 /* KCOV sequence number: */
1287 /* Collect coverage from softirq context: */
1288 unsigned int kcov_softirq;
1292 struct mem_cgroup *memcg_in_oom;
1293 gfp_t memcg_oom_gfp_mask;
1294 int memcg_oom_order;
1296 /* Number of pages to reclaim on returning to userland: */
1297 unsigned int memcg_nr_pages_over_high;
1299 /* Used by memcontrol for targeted memcg charge: */
1300 struct mem_cgroup *active_memcg;
1303 #ifdef CONFIG_BLK_CGROUP
1304 struct request_queue *throttle_queue;
1307 #ifdef CONFIG_UPROBES
1308 struct uprobe_task *utask;
1310 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1311 unsigned int sequential_io;
1312 unsigned int sequential_io_avg;
1314 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1315 unsigned long task_state_change;
1317 int pagefault_disabled;
1319 struct task_struct *oom_reaper_list;
1321 #ifdef CONFIG_VMAP_STACK
1322 struct vm_struct *stack_vm_area;
1324 #ifdef CONFIG_THREAD_INFO_IN_TASK
1325 /* A live task holds one reference: */
1326 refcount_t stack_refcount;
1328 #ifdef CONFIG_LIVEPATCH
1331 #ifdef CONFIG_SECURITY
1332 /* Used by LSM modules for access restriction: */
1336 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1337 unsigned long lowest_stack;
1338 unsigned long prev_lowest_stack;
1341 #ifdef CONFIG_X86_MCE
1342 void __user *mce_vaddr;
1347 __mce_reserved : 62;
1348 struct callback_head mce_kill_me;
1352 * New fields for task_struct should be added above here, so that
1353 * they are included in the randomized portion of task_struct.
1355 randomized_struct_fields_end
1357 /* CPU-specific state of this task: */
1358 struct thread_struct thread;
1361 * WARNING: on x86, 'thread_struct' contains a variable-sized
1362 * structure. It *MUST* be at the end of 'task_struct'.
1364 * Do not put anything below here!
1368 static inline struct pid *task_pid(struct task_struct *task)
1370 return task->thread_pid;
1374 * the helpers to get the task's different pids as they are seen
1375 * from various namespaces
1377 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1378 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1380 * task_xid_nr_ns() : id seen from the ns specified;
1382 * see also pid_nr() etc in include/linux/pid.h
1384 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1386 static inline pid_t task_pid_nr(struct task_struct *tsk)
1391 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1393 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1396 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1398 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1402 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1408 * pid_alive - check that a task structure is not stale
1409 * @p: Task structure to be checked.
1411 * Test if a process is not yet dead (at most zombie state)
1412 * If pid_alive fails, then pointers within the task structure
1413 * can be stale and must not be dereferenced.
1415 * Return: 1 if the process is alive. 0 otherwise.
1417 static inline int pid_alive(const struct task_struct *p)
1419 return p->thread_pid != NULL;
1422 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1424 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1427 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1429 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1433 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1435 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1438 static inline pid_t task_session_vnr(struct task_struct *tsk)
1440 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1443 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1445 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1448 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1450 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1453 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1459 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1465 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1467 return task_ppid_nr_ns(tsk, &init_pid_ns);
1470 /* Obsolete, do not use: */
1471 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1473 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1476 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1477 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1479 static inline unsigned int task_state_index(struct task_struct *tsk)
1481 unsigned int tsk_state = READ_ONCE(tsk->state);
1482 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1484 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1486 if (tsk_state == TASK_IDLE)
1487 state = TASK_REPORT_IDLE;
1492 static inline char task_index_to_char(unsigned int state)
1494 static const char state_char[] = "RSDTtXZPI";
1496 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1498 return state_char[state];
1501 static inline char task_state_to_char(struct task_struct *tsk)
1503 return task_index_to_char(task_state_index(tsk));
1507 * is_global_init - check if a task structure is init. Since init
1508 * is free to have sub-threads we need to check tgid.
1509 * @tsk: Task structure to be checked.
1511 * Check if a task structure is the first user space task the kernel created.
1513 * Return: 1 if the task structure is init. 0 otherwise.
1515 static inline int is_global_init(struct task_struct *tsk)
1517 return task_tgid_nr(tsk) == 1;
1520 extern struct pid *cad_pid;
1525 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1526 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1527 #define PF_EXITING 0x00000004 /* Getting shut down */
1528 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1529 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1530 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1531 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1532 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1533 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1534 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1535 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1536 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1537 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1538 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1539 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1540 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1541 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1542 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1543 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1544 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1545 * I am cleaning dirty pages from some other bdi. */
1546 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1547 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1548 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1549 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1550 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1551 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1552 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1553 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1556 * Only the _current_ task can read/write to tsk->flags, but other
1557 * tasks can access tsk->flags in readonly mode for example
1558 * with tsk_used_math (like during threaded core dumping).
1559 * There is however an exception to this rule during ptrace
1560 * or during fork: the ptracer task is allowed to write to the
1561 * child->flags of its traced child (same goes for fork, the parent
1562 * can write to the child->flags), because we're guaranteed the
1563 * child is not running and in turn not changing child->flags
1564 * at the same time the parent does it.
1566 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1567 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1568 #define clear_used_math() clear_stopped_child_used_math(current)
1569 #define set_used_math() set_stopped_child_used_math(current)
1571 #define conditional_stopped_child_used_math(condition, child) \
1572 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1574 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1576 #define copy_to_stopped_child_used_math(child) \
1577 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1579 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1580 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1581 #define used_math() tsk_used_math(current)
1583 static inline bool is_percpu_thread(void)
1586 return (current->flags & PF_NO_SETAFFINITY) &&
1587 (current->nr_cpus_allowed == 1);
1593 /* Per-process atomic flags. */
1594 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1595 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1596 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1597 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1598 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1599 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1600 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1601 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1603 #define TASK_PFA_TEST(name, func) \
1604 static inline bool task_##func(struct task_struct *p) \
1605 { return test_bit(PFA_##name, &p->atomic_flags); }
1607 #define TASK_PFA_SET(name, func) \
1608 static inline void task_set_##func(struct task_struct *p) \
1609 { set_bit(PFA_##name, &p->atomic_flags); }
1611 #define TASK_PFA_CLEAR(name, func) \
1612 static inline void task_clear_##func(struct task_struct *p) \
1613 { clear_bit(PFA_##name, &p->atomic_flags); }
1615 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1616 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1618 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1619 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1620 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1622 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1623 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1624 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1626 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1627 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1628 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1630 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1631 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1632 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1634 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1635 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1637 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1638 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1639 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1641 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1642 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1645 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1647 current->flags &= ~flags;
1648 current->flags |= orig_flags & flags;
1651 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1652 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1654 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1655 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1657 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1660 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1662 if (!cpumask_test_cpu(0, new_mask))
1668 extern int yield_to(struct task_struct *p, bool preempt);
1669 extern void set_user_nice(struct task_struct *p, long nice);
1670 extern int task_prio(const struct task_struct *p);
1673 * task_nice - return the nice value of a given task.
1674 * @p: the task in question.
1676 * Return: The nice value [ -20 ... 0 ... 19 ].
1678 static inline int task_nice(const struct task_struct *p)
1680 return PRIO_TO_NICE((p)->static_prio);
1683 extern int can_nice(const struct task_struct *p, const int nice);
1684 extern int task_curr(const struct task_struct *p);
1685 extern int idle_cpu(int cpu);
1686 extern int available_idle_cpu(int cpu);
1687 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1688 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1689 extern void sched_set_fifo(struct task_struct *p);
1690 extern void sched_set_fifo_low(struct task_struct *p);
1691 extern void sched_set_normal(struct task_struct *p, int nice);
1692 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1693 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1694 extern struct task_struct *idle_task(int cpu);
1697 * is_idle_task - is the specified task an idle task?
1698 * @p: the task in question.
1700 * Return: 1 if @p is an idle task. 0 otherwise.
1702 static __always_inline bool is_idle_task(const struct task_struct *p)
1704 return !!(p->flags & PF_IDLE);
1707 extern struct task_struct *curr_task(int cpu);
1708 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1712 union thread_union {
1713 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1714 struct task_struct task;
1716 #ifndef CONFIG_THREAD_INFO_IN_TASK
1717 struct thread_info thread_info;
1719 unsigned long stack[THREAD_SIZE/sizeof(long)];
1722 #ifndef CONFIG_THREAD_INFO_IN_TASK
1723 extern struct thread_info init_thread_info;
1726 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1728 #ifdef CONFIG_THREAD_INFO_IN_TASK
1729 static inline struct thread_info *task_thread_info(struct task_struct *task)
1731 return &task->thread_info;
1733 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1734 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1738 * find a task by one of its numerical ids
1740 * find_task_by_pid_ns():
1741 * finds a task by its pid in the specified namespace
1742 * find_task_by_vpid():
1743 * finds a task by its virtual pid
1745 * see also find_vpid() etc in include/linux/pid.h
1748 extern struct task_struct *find_task_by_vpid(pid_t nr);
1749 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1752 * find a task by its virtual pid and get the task struct
1754 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1756 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1757 extern int wake_up_process(struct task_struct *tsk);
1758 extern void wake_up_new_task(struct task_struct *tsk);
1761 extern void kick_process(struct task_struct *tsk);
1763 static inline void kick_process(struct task_struct *tsk) { }
1766 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1768 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1770 __set_task_comm(tsk, from, false);
1773 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1774 #define get_task_comm(buf, tsk) ({ \
1775 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1776 __get_task_comm(buf, sizeof(buf), tsk); \
1780 static __always_inline void scheduler_ipi(void)
1783 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1784 * TIF_NEED_RESCHED remotely (for the first time) will also send
1787 preempt_fold_need_resched();
1789 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1791 static inline void scheduler_ipi(void) { }
1792 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1799 * Set thread flags in other task's structures.
1800 * See asm/thread_info.h for TIF_xxxx flags available:
1802 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1804 set_ti_thread_flag(task_thread_info(tsk), flag);
1807 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1809 clear_ti_thread_flag(task_thread_info(tsk), flag);
1812 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1815 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1818 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1820 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1823 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1825 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1828 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1830 return test_ti_thread_flag(task_thread_info(tsk), flag);
1833 static inline void set_tsk_need_resched(struct task_struct *tsk)
1835 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1838 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1840 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1843 static inline int test_tsk_need_resched(struct task_struct *tsk)
1845 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1849 * cond_resched() and cond_resched_lock(): latency reduction via
1850 * explicit rescheduling in places that are safe. The return
1851 * value indicates whether a reschedule was done in fact.
1852 * cond_resched_lock() will drop the spinlock before scheduling,
1854 #ifndef CONFIG_PREEMPTION
1855 extern int _cond_resched(void);
1857 static inline int _cond_resched(void) { return 0; }
1860 #define cond_resched() ({ \
1861 ___might_sleep(__FILE__, __LINE__, 0); \
1865 extern int __cond_resched_lock(spinlock_t *lock);
1867 #define cond_resched_lock(lock) ({ \
1868 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1869 __cond_resched_lock(lock); \
1872 static inline void cond_resched_rcu(void)
1874 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1882 * Does a critical section need to be broken due to another
1883 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1884 * but a general need for low latency)
1886 static inline int spin_needbreak(spinlock_t *lock)
1888 #ifdef CONFIG_PREEMPTION
1889 return spin_is_contended(lock);
1895 static __always_inline bool need_resched(void)
1897 return unlikely(tif_need_resched());
1901 * Wrappers for p->thread_info->cpu access. No-op on UP.
1905 static inline unsigned int task_cpu(const struct task_struct *p)
1907 #ifdef CONFIG_THREAD_INFO_IN_TASK
1908 return READ_ONCE(p->cpu);
1910 return READ_ONCE(task_thread_info(p)->cpu);
1914 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1918 static inline unsigned int task_cpu(const struct task_struct *p)
1923 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1927 #endif /* CONFIG_SMP */
1930 * In order to reduce various lock holder preemption latencies provide an
1931 * interface to see if a vCPU is currently running or not.
1933 * This allows us to terminate optimistic spin loops and block, analogous to
1934 * the native optimistic spin heuristic of testing if the lock owner task is
1937 #ifndef vcpu_is_preempted
1938 static inline bool vcpu_is_preempted(int cpu)
1944 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1945 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1947 #ifndef TASK_SIZE_OF
1948 #define TASK_SIZE_OF(tsk) TASK_SIZE
1954 * Map the event mask on the user-space ABI enum rseq_cs_flags
1955 * for direct mask checks.
1957 enum rseq_event_mask_bits {
1958 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1959 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1960 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1963 enum rseq_event_mask {
1964 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1965 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1966 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1969 static inline void rseq_set_notify_resume(struct task_struct *t)
1972 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1975 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1977 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1978 struct pt_regs *regs)
1981 __rseq_handle_notify_resume(ksig, regs);
1984 static inline void rseq_signal_deliver(struct ksignal *ksig,
1985 struct pt_regs *regs)
1988 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1990 rseq_handle_notify_resume(ksig, regs);
1993 /* rseq_preempt() requires preemption to be disabled. */
1994 static inline void rseq_preempt(struct task_struct *t)
1996 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1997 rseq_set_notify_resume(t);
2000 /* rseq_migrate() requires preemption to be disabled. */
2001 static inline void rseq_migrate(struct task_struct *t)
2003 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2004 rseq_set_notify_resume(t);
2008 * If parent process has a registered restartable sequences area, the
2009 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2011 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2013 if (clone_flags & CLONE_VM) {
2016 t->rseq_event_mask = 0;
2018 t->rseq = current->rseq;
2019 t->rseq_sig = current->rseq_sig;
2020 t->rseq_event_mask = current->rseq_event_mask;
2024 static inline void rseq_execve(struct task_struct *t)
2028 t->rseq_event_mask = 0;
2033 static inline void rseq_set_notify_resume(struct task_struct *t)
2036 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2037 struct pt_regs *regs)
2040 static inline void rseq_signal_deliver(struct ksignal *ksig,
2041 struct pt_regs *regs)
2044 static inline void rseq_preempt(struct task_struct *t)
2047 static inline void rseq_migrate(struct task_struct *t)
2050 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2053 static inline void rseq_execve(struct task_struct *t)
2059 #ifdef CONFIG_DEBUG_RSEQ
2061 void rseq_syscall(struct pt_regs *regs);
2065 static inline void rseq_syscall(struct pt_regs *regs)
2071 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2072 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2073 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2075 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2076 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2077 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2079 int sched_trace_rq_cpu(struct rq *rq);
2080 int sched_trace_rq_cpu_capacity(struct rq *rq);
2081 int sched_trace_rq_nr_running(struct rq *rq);
2083 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);