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/syscall_user_dispatch.h>
32 #include <linux/mm_types_task.h>
33 #include <linux/task_io_accounting.h>
34 #include <linux/posix-timers.h>
35 #include <linux/rseq.h>
36 #include <linux/seqlock.h>
37 #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;
46 struct capture_control;
49 struct futex_pi_state;
55 struct perf_event_context;
57 struct pipe_inode_info;
60 struct robust_list_head;
66 struct sighand_struct;
68 struct task_delay_info;
72 * Task state bitmask. NOTE! These bits are also
73 * encoded in fs/proc/array.c: get_task_state().
75 * We have two separate sets of flags: task->state
76 * is about runnability, while task->exit_state are
77 * about the task exiting. Confusing, but this way
78 * modifying one set can't modify the other one by
82 /* Used in tsk->state: */
83 #define TASK_RUNNING 0x0000
84 #define TASK_INTERRUPTIBLE 0x0001
85 #define TASK_UNINTERRUPTIBLE 0x0002
86 #define __TASK_STOPPED 0x0004
87 #define __TASK_TRACED 0x0008
88 /* Used in tsk->exit_state: */
89 #define EXIT_DEAD 0x0010
90 #define EXIT_ZOMBIE 0x0020
91 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
92 /* Used in tsk->state again: */
93 #define TASK_PARKED 0x0040
94 #define TASK_DEAD 0x0080
95 #define TASK_WAKEKILL 0x0100
96 #define TASK_WAKING 0x0200
97 #define TASK_NOLOAD 0x0400
98 #define TASK_NEW 0x0800
99 #define TASK_STATE_MAX 0x1000
101 /* Convenience macros for the sake of set_current_state: */
102 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
103 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
104 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
106 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
108 /* Convenience macros for the sake of wake_up(): */
109 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
111 /* get_task_state(): */
112 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
113 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
114 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
117 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
119 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
121 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
123 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
126 * Special states are those that do not use the normal wait-loop pattern. See
127 * the comment with set_special_state().
129 #define is_special_task_state(state) \
130 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
132 #define __set_current_state(state_value) \
134 WARN_ON_ONCE(is_special_task_state(state_value));\
135 current->task_state_change = _THIS_IP_; \
136 current->state = (state_value); \
139 #define set_current_state(state_value) \
141 WARN_ON_ONCE(is_special_task_state(state_value));\
142 current->task_state_change = _THIS_IP_; \
143 smp_store_mb(current->state, (state_value)); \
146 #define set_special_state(state_value) \
148 unsigned long flags; /* may shadow */ \
149 WARN_ON_ONCE(!is_special_task_state(state_value)); \
150 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
151 current->task_state_change = _THIS_IP_; \
152 current->state = (state_value); \
153 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
157 * set_current_state() includes a barrier so that the write of current->state
158 * is correctly serialised wrt the caller's subsequent test of whether to
162 * set_current_state(TASK_UNINTERRUPTIBLE);
168 * __set_current_state(TASK_RUNNING);
170 * If the caller does not need such serialisation (because, for instance, the
171 * CONDITION test and condition change and wakeup are under the same lock) then
172 * use __set_current_state().
174 * The above is typically ordered against the wakeup, which does:
177 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
179 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
180 * accessing p->state.
182 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
183 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
184 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
186 * However, with slightly different timing the wakeup TASK_RUNNING store can
187 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
188 * a problem either because that will result in one extra go around the loop
189 * and our @cond test will save the day.
191 * Also see the comments of try_to_wake_up().
193 #define __set_current_state(state_value) \
194 current->state = (state_value)
196 #define set_current_state(state_value) \
197 smp_store_mb(current->state, (state_value))
200 * set_special_state() should be used for those states when the blocking task
201 * can not use the regular condition based wait-loop. In that case we must
202 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
203 * will not collide with our state change.
205 #define set_special_state(state_value) \
207 unsigned long flags; /* may shadow */ \
208 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
209 current->state = (state_value); \
210 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
215 /* Task command name length: */
216 #define TASK_COMM_LEN 16
218 extern void scheduler_tick(void);
220 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
222 extern long schedule_timeout(long timeout);
223 extern long schedule_timeout_interruptible(long timeout);
224 extern long schedule_timeout_killable(long timeout);
225 extern long schedule_timeout_uninterruptible(long timeout);
226 extern long schedule_timeout_idle(long timeout);
227 asmlinkage void schedule(void);
228 extern void schedule_preempt_disabled(void);
229 asmlinkage void preempt_schedule_irq(void);
231 extern int __must_check io_schedule_prepare(void);
232 extern void io_schedule_finish(int token);
233 extern long io_schedule_timeout(long timeout);
234 extern void io_schedule(void);
237 * struct prev_cputime - snapshot of system and user cputime
238 * @utime: time spent in user mode
239 * @stime: time spent in system mode
240 * @lock: protects the above two fields
242 * Stores previous user/system time values such that we can guarantee
245 struct prev_cputime {
246 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
254 /* Task is sleeping or running in a CPU with VTIME inactive: */
258 /* Task runs in kernelspace in a CPU with VTIME active: */
260 /* Task runs in userspace in a CPU with VTIME active: */
262 /* Task runs as guests in a CPU with VTIME active: */
268 unsigned long long starttime;
269 enum vtime_state state;
277 * Utilization clamp constraints.
278 * @UCLAMP_MIN: Minimum utilization
279 * @UCLAMP_MAX: Maximum utilization
280 * @UCLAMP_CNT: Utilization clamp constraints count
289 extern struct root_domain def_root_domain;
290 extern struct mutex sched_domains_mutex;
294 #ifdef CONFIG_SCHED_INFO
295 /* Cumulative counters: */
297 /* # of times we have run on this CPU: */
298 unsigned long pcount;
300 /* Time spent waiting on a runqueue: */
301 unsigned long long run_delay;
305 /* When did we last run on a CPU? */
306 unsigned long long last_arrival;
308 /* When were we last queued to run? */
309 unsigned long long last_queued;
311 #endif /* CONFIG_SCHED_INFO */
315 * Integer metrics need fixed point arithmetic, e.g., sched/fair
316 * has a few: load, load_avg, util_avg, freq, and capacity.
318 * We define a basic fixed point arithmetic range, and then formalize
319 * all these metrics based on that basic range.
321 # define SCHED_FIXEDPOINT_SHIFT 10
322 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
324 /* Increase resolution of cpu_capacity calculations */
325 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
326 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
329 unsigned long weight;
334 * struct util_est - Estimation utilization of FAIR tasks
335 * @enqueued: instantaneous estimated utilization of a task/cpu
336 * @ewma: the Exponential Weighted Moving Average (EWMA)
337 * utilization of a task
339 * Support data structure to track an Exponential Weighted Moving Average
340 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
341 * average each time a task completes an activation. Sample's weight is chosen
342 * so that the EWMA will be relatively insensitive to transient changes to the
345 * The enqueued attribute has a slightly different meaning for tasks and cpus:
346 * - task: the task's util_avg at last task dequeue time
347 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
348 * Thus, the util_est.enqueued of a task represents the contribution on the
349 * estimated utilization of the CPU where that task is currently enqueued.
351 * Only for tasks we track a moving average of the past instantaneous
352 * estimated utilization. This allows to absorb sporadic drops in utilization
353 * of an otherwise almost periodic task.
356 unsigned int enqueued;
358 #define UTIL_EST_WEIGHT_SHIFT 2
359 } __attribute__((__aligned__(sizeof(u64))));
362 * The load/runnable/util_avg accumulates an infinite geometric series
363 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
365 * [load_avg definition]
367 * load_avg = runnable% * scale_load_down(load)
369 * [runnable_avg definition]
371 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
373 * [util_avg definition]
375 * util_avg = running% * SCHED_CAPACITY_SCALE
377 * where runnable% is the time ratio that a sched_entity is runnable and
378 * running% the time ratio that a sched_entity is running.
380 * For cfs_rq, they are the aggregated values of all runnable and blocked
383 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
384 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
385 * for computing those signals (see update_rq_clock_pelt())
387 * N.B., the above ratios (runnable% and running%) themselves are in the
388 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
389 * to as large a range as necessary. This is for example reflected by
390 * util_avg's SCHED_CAPACITY_SCALE.
394 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
395 * with the highest load (=88761), always runnable on a single cfs_rq,
396 * and should not overflow as the number already hits PID_MAX_LIMIT.
398 * For all other cases (including 32-bit kernels), struct load_weight's
399 * weight will overflow first before we do, because:
401 * Max(load_avg) <= Max(load.weight)
403 * Then it is the load_weight's responsibility to consider overflow
407 u64 last_update_time;
412 unsigned long load_avg;
413 unsigned long runnable_avg;
414 unsigned long util_avg;
415 struct util_est util_est;
416 } ____cacheline_aligned;
418 struct sched_statistics {
419 #ifdef CONFIG_SCHEDSTATS
429 s64 sum_sleep_runtime;
436 u64 nr_migrations_cold;
437 u64 nr_failed_migrations_affine;
438 u64 nr_failed_migrations_running;
439 u64 nr_failed_migrations_hot;
440 u64 nr_forced_migrations;
444 u64 nr_wakeups_migrate;
445 u64 nr_wakeups_local;
446 u64 nr_wakeups_remote;
447 u64 nr_wakeups_affine;
448 u64 nr_wakeups_affine_attempts;
449 u64 nr_wakeups_passive;
454 struct sched_entity {
455 /* For load-balancing: */
456 struct load_weight load;
457 struct rb_node run_node;
458 struct list_head group_node;
462 u64 sum_exec_runtime;
464 u64 prev_sum_exec_runtime;
468 struct sched_statistics statistics;
470 #ifdef CONFIG_FAIR_GROUP_SCHED
472 struct sched_entity *parent;
473 /* rq on which this entity is (to be) queued: */
474 struct cfs_rq *cfs_rq;
475 /* rq "owned" by this entity/group: */
477 /* cached value of my_q->h_nr_running */
478 unsigned long runnable_weight;
483 * Per entity load average tracking.
485 * Put into separate cache line so it does not
486 * collide with read-mostly values above.
488 struct sched_avg avg;
492 struct sched_rt_entity {
493 struct list_head run_list;
494 unsigned long timeout;
495 unsigned long watchdog_stamp;
496 unsigned int time_slice;
497 unsigned short on_rq;
498 unsigned short on_list;
500 struct sched_rt_entity *back;
501 #ifdef CONFIG_RT_GROUP_SCHED
502 struct sched_rt_entity *parent;
503 /* rq on which this entity is (to be) queued: */
505 /* rq "owned" by this entity/group: */
508 } __randomize_layout;
510 struct sched_dl_entity {
511 struct rb_node rb_node;
514 * Original scheduling parameters. Copied here from sched_attr
515 * during sched_setattr(), they will remain the same until
516 * the next sched_setattr().
518 u64 dl_runtime; /* Maximum runtime for each instance */
519 u64 dl_deadline; /* Relative deadline of each instance */
520 u64 dl_period; /* Separation of two instances (period) */
521 u64 dl_bw; /* dl_runtime / dl_period */
522 u64 dl_density; /* dl_runtime / dl_deadline */
525 * Actual scheduling parameters. Initialized with the values above,
526 * they are continuously updated during task execution. Note that
527 * the remaining runtime could be < 0 in case we are in overrun.
529 s64 runtime; /* Remaining runtime for this instance */
530 u64 deadline; /* Absolute deadline for this instance */
531 unsigned int flags; /* Specifying the scheduler behaviour */
536 * @dl_throttled tells if we exhausted the runtime. If so, the
537 * task has to wait for a replenishment to be performed at the
538 * next firing of dl_timer.
540 * @dl_boosted tells if we are boosted due to DI. If so we are
541 * outside bandwidth enforcement mechanism (but only until we
542 * exit the critical section);
544 * @dl_yielded tells if task gave up the CPU before consuming
545 * all its available runtime during the last job.
547 * @dl_non_contending tells if the task is inactive while still
548 * contributing to the active utilization. In other words, it
549 * indicates if the inactive timer has been armed and its handler
550 * has not been executed yet. This flag is useful to avoid race
551 * conditions between the inactive timer handler and the wakeup
554 * @dl_overrun tells if the task asked to be informed about runtime
557 unsigned int dl_throttled : 1;
558 unsigned int dl_yielded : 1;
559 unsigned int dl_non_contending : 1;
560 unsigned int dl_overrun : 1;
563 * Bandwidth enforcement timer. Each -deadline task has its
564 * own bandwidth to be enforced, thus we need one timer per task.
566 struct hrtimer dl_timer;
569 * Inactive timer, responsible for decreasing the active utilization
570 * at the "0-lag time". When a -deadline task blocks, it contributes
571 * to GRUB's active utilization until the "0-lag time", hence a
572 * timer is needed to decrease the active utilization at the correct
575 struct hrtimer inactive_timer;
577 #ifdef CONFIG_RT_MUTEXES
579 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
580 * pi_se points to the donor, otherwise points to the dl_se it belongs
581 * to (the original one/itself).
583 struct sched_dl_entity *pi_se;
587 #ifdef CONFIG_UCLAMP_TASK
588 /* Number of utilization clamp buckets (shorter alias) */
589 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
592 * Utilization clamp for a scheduling entity
593 * @value: clamp value "assigned" to a se
594 * @bucket_id: bucket index corresponding to the "assigned" value
595 * @active: the se is currently refcounted in a rq's bucket
596 * @user_defined: the requested clamp value comes from user-space
598 * The bucket_id is the index of the clamp bucket matching the clamp value
599 * which is pre-computed and stored to avoid expensive integer divisions from
602 * The active bit is set whenever a task has got an "effective" value assigned,
603 * which can be different from the clamp value "requested" from user-space.
604 * This allows to know a task is refcounted in the rq's bucket corresponding
605 * to the "effective" bucket_id.
607 * The user_defined bit is set whenever a task has got a task-specific clamp
608 * value requested from userspace, i.e. the system defaults apply to this task
609 * just as a restriction. This allows to relax default clamps when a less
610 * restrictive task-specific value has been requested, thus allowing to
611 * implement a "nice" semantic. For example, a task running with a 20%
612 * default boost can still drop its own boosting to 0%.
615 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
616 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
617 unsigned int active : 1;
618 unsigned int user_defined : 1;
620 #endif /* CONFIG_UCLAMP_TASK */
626 u8 exp_hint; /* Hint for performance. */
627 u8 need_mb; /* Readers need smp_mb(). */
629 u32 s; /* Set of bits. */
632 enum perf_event_task_context {
633 perf_invalid_context = -1,
636 perf_nr_task_contexts,
640 struct wake_q_node *next;
644 #ifdef CONFIG_KMAP_LOCAL
646 pte_t pteval[KM_MAX_IDX];
651 #ifdef CONFIG_THREAD_INFO_IN_TASK
653 * For reasons of header soup (see current_thread_info()), this
654 * must be the first element of task_struct.
656 struct thread_info thread_info;
658 /* -1 unrunnable, 0 runnable, >0 stopped: */
662 * This begins the randomizable portion of task_struct. Only
663 * scheduling-critical items should be added above here.
665 randomized_struct_fields_start
669 /* Per task flags (PF_*), defined further below: */
675 struct __call_single_node wake_entry;
676 #ifdef CONFIG_THREAD_INFO_IN_TASK
680 unsigned int wakee_flips;
681 unsigned long wakee_flip_decay_ts;
682 struct task_struct *last_wakee;
685 * recent_used_cpu is initially set as the last CPU used by a task
686 * that wakes affine another task. Waker/wakee relationships can
687 * push tasks around a CPU where each wakeup moves to the next one.
688 * Tracking a recently used CPU allows a quick search for a recently
689 * used CPU that may be idle.
699 unsigned int rt_priority;
701 const struct sched_class *sched_class;
702 struct sched_entity se;
703 struct sched_rt_entity rt;
704 #ifdef CONFIG_CGROUP_SCHED
705 struct task_group *sched_task_group;
707 struct sched_dl_entity dl;
709 #ifdef CONFIG_UCLAMP_TASK
711 * Clamp values requested for a scheduling entity.
712 * Must be updated with task_rq_lock() held.
714 struct uclamp_se uclamp_req[UCLAMP_CNT];
716 * Effective clamp values used for a scheduling entity.
717 * Must be updated with task_rq_lock() held.
719 struct uclamp_se uclamp[UCLAMP_CNT];
722 #ifdef CONFIG_PREEMPT_NOTIFIERS
723 /* List of struct preempt_notifier: */
724 struct hlist_head preempt_notifiers;
727 #ifdef CONFIG_BLK_DEV_IO_TRACE
728 unsigned int btrace_seq;
733 const cpumask_t *cpus_ptr;
735 void *migration_pending;
737 unsigned short migration_disabled;
739 unsigned short migration_flags;
741 #ifdef CONFIG_PREEMPT_RCU
742 int rcu_read_lock_nesting;
743 union rcu_special rcu_read_unlock_special;
744 struct list_head rcu_node_entry;
745 struct rcu_node *rcu_blocked_node;
746 #endif /* #ifdef CONFIG_PREEMPT_RCU */
748 #ifdef CONFIG_TASKS_RCU
749 unsigned long rcu_tasks_nvcsw;
750 u8 rcu_tasks_holdout;
752 int rcu_tasks_idle_cpu;
753 struct list_head rcu_tasks_holdout_list;
754 #endif /* #ifdef CONFIG_TASKS_RCU */
756 #ifdef CONFIG_TASKS_TRACE_RCU
757 int trc_reader_nesting;
759 union rcu_special trc_reader_special;
760 bool trc_reader_checked;
761 struct list_head trc_holdout_list;
762 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
764 struct sched_info sched_info;
766 struct list_head tasks;
768 struct plist_node pushable_tasks;
769 struct rb_node pushable_dl_tasks;
772 struct mm_struct *mm;
773 struct mm_struct *active_mm;
775 /* Per-thread vma caching: */
776 struct vmacache vmacache;
778 #ifdef SPLIT_RSS_COUNTING
779 struct task_rss_stat rss_stat;
784 /* The signal sent when the parent dies: */
786 /* JOBCTL_*, siglock protected: */
787 unsigned long jobctl;
789 /* Used for emulating ABI behavior of previous Linux versions: */
790 unsigned int personality;
792 /* Scheduler bits, serialized by scheduler locks: */
793 unsigned sched_reset_on_fork:1;
794 unsigned sched_contributes_to_load:1;
795 unsigned sched_migrated:1;
797 unsigned sched_psi_wake_requeue:1;
800 /* Force alignment to the next boundary: */
803 /* Unserialized, strictly 'current' */
806 * This field must not be in the scheduler word above due to wakelist
807 * queueing no longer being serialized by p->on_cpu. However:
810 * schedule() if (p->on_rq && ..) // false
811 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
812 * deactivate_task() ttwu_queue_wakelist())
813 * p->on_rq = 0; p->sched_remote_wakeup = Y;
815 * guarantees all stores of 'current' are visible before
816 * ->sched_remote_wakeup gets used, so it can be in this word.
818 unsigned sched_remote_wakeup:1;
820 /* Bit to tell LSMs we're in execve(): */
821 unsigned in_execve:1;
822 unsigned in_iowait:1;
823 #ifndef TIF_RESTORE_SIGMASK
824 unsigned restore_sigmask:1;
827 unsigned in_user_fault:1;
829 #ifdef CONFIG_COMPAT_BRK
830 unsigned brk_randomized:1;
832 #ifdef CONFIG_CGROUPS
833 /* disallow userland-initiated cgroup migration */
834 unsigned no_cgroup_migration:1;
835 /* task is frozen/stopped (used by the cgroup freezer) */
838 #ifdef CONFIG_BLK_CGROUP
839 unsigned use_memdelay:1;
842 /* Stalled due to lack of memory */
843 unsigned in_memstall:1;
846 unsigned long atomic_flags; /* Flags requiring atomic access. */
848 struct restart_block restart_block;
853 #ifdef CONFIG_STACKPROTECTOR
854 /* Canary value for the -fstack-protector GCC feature: */
855 unsigned long stack_canary;
858 * Pointers to the (original) parent process, youngest child, younger sibling,
859 * older sibling, respectively. (p->father can be replaced with
860 * p->real_parent->pid)
863 /* Real parent process: */
864 struct task_struct __rcu *real_parent;
866 /* Recipient of SIGCHLD, wait4() reports: */
867 struct task_struct __rcu *parent;
870 * Children/sibling form the list of natural children:
872 struct list_head children;
873 struct list_head sibling;
874 struct task_struct *group_leader;
877 * 'ptraced' is the list of tasks this task is using ptrace() on.
879 * This includes both natural children and PTRACE_ATTACH targets.
880 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
882 struct list_head ptraced;
883 struct list_head ptrace_entry;
885 /* PID/PID hash table linkage. */
886 struct pid *thread_pid;
887 struct hlist_node pid_links[PIDTYPE_MAX];
888 struct list_head thread_group;
889 struct list_head thread_node;
891 struct completion *vfork_done;
893 /* CLONE_CHILD_SETTID: */
894 int __user *set_child_tid;
896 /* CLONE_CHILD_CLEARTID: */
897 int __user *clear_child_tid;
904 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
909 struct prev_cputime prev_cputime;
910 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
914 #ifdef CONFIG_NO_HZ_FULL
915 atomic_t tick_dep_mask;
917 /* Context switch counts: */
919 unsigned long nivcsw;
921 /* Monotonic time in nsecs: */
924 /* Boot based time in nsecs: */
927 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
928 unsigned long min_flt;
929 unsigned long maj_flt;
931 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
932 struct posix_cputimers posix_cputimers;
934 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
935 struct posix_cputimers_work posix_cputimers_work;
938 /* Process credentials: */
940 /* Tracer's credentials at attach: */
941 const struct cred __rcu *ptracer_cred;
943 /* Objective and real subjective task credentials (COW): */
944 const struct cred __rcu *real_cred;
946 /* Effective (overridable) subjective task credentials (COW): */
947 const struct cred __rcu *cred;
950 /* Cached requested key. */
951 struct key *cached_requested_key;
955 * executable name, excluding path.
957 * - normally initialized setup_new_exec()
958 * - access it with [gs]et_task_comm()
959 * - lock it with task_lock()
961 char comm[TASK_COMM_LEN];
963 struct nameidata *nameidata;
965 #ifdef CONFIG_SYSVIPC
966 struct sysv_sem sysvsem;
967 struct sysv_shm sysvshm;
969 #ifdef CONFIG_DETECT_HUNG_TASK
970 unsigned long last_switch_count;
971 unsigned long last_switch_time;
973 /* Filesystem information: */
974 struct fs_struct *fs;
976 /* Open file information: */
977 struct files_struct *files;
979 #ifdef CONFIG_IO_URING
980 struct io_uring_task *io_uring;
984 struct nsproxy *nsproxy;
986 /* Signal handlers: */
987 struct signal_struct *signal;
988 struct sighand_struct __rcu *sighand;
990 sigset_t real_blocked;
991 /* Restored if set_restore_sigmask() was used: */
992 sigset_t saved_sigmask;
993 struct sigpending pending;
994 unsigned long sas_ss_sp;
996 unsigned int sas_ss_flags;
998 struct callback_head *task_works;
1001 #ifdef CONFIG_AUDITSYSCALL
1002 struct audit_context *audit_context;
1005 unsigned int sessionid;
1007 struct seccomp seccomp;
1008 struct syscall_user_dispatch syscall_dispatch;
1010 /* Thread group tracking: */
1014 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1015 spinlock_t alloc_lock;
1017 /* Protection of the PI data structures: */
1018 raw_spinlock_t pi_lock;
1020 struct wake_q_node wake_q;
1022 #ifdef CONFIG_RT_MUTEXES
1023 /* PI waiters blocked on a rt_mutex held by this task: */
1024 struct rb_root_cached pi_waiters;
1025 /* Updated under owner's pi_lock and rq lock */
1026 struct task_struct *pi_top_task;
1027 /* Deadlock detection and priority inheritance handling: */
1028 struct rt_mutex_waiter *pi_blocked_on;
1031 #ifdef CONFIG_DEBUG_MUTEXES
1032 /* Mutex deadlock detection: */
1033 struct mutex_waiter *blocked_on;
1036 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1037 int non_block_count;
1040 #ifdef CONFIG_TRACE_IRQFLAGS
1041 struct irqtrace_events irqtrace;
1042 unsigned int hardirq_threaded;
1043 u64 hardirq_chain_key;
1044 int softirqs_enabled;
1045 int softirq_context;
1049 #ifdef CONFIG_LOCKDEP
1050 # define MAX_LOCK_DEPTH 48UL
1053 unsigned int lockdep_recursion;
1054 struct held_lock held_locks[MAX_LOCK_DEPTH];
1057 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1058 unsigned int in_ubsan;
1061 /* Journalling filesystem info: */
1064 /* Stacked block device info: */
1065 struct bio_list *bio_list;
1068 /* Stack plugging: */
1069 struct blk_plug *plug;
1073 struct reclaim_state *reclaim_state;
1075 struct backing_dev_info *backing_dev_info;
1077 struct io_context *io_context;
1079 #ifdef CONFIG_COMPACTION
1080 struct capture_control *capture_control;
1083 unsigned long ptrace_message;
1084 kernel_siginfo_t *last_siginfo;
1086 struct task_io_accounting ioac;
1088 /* Pressure stall state */
1089 unsigned int psi_flags;
1091 #ifdef CONFIG_TASK_XACCT
1092 /* Accumulated RSS usage: */
1094 /* Accumulated virtual memory usage: */
1096 /* stime + utime since last update: */
1099 #ifdef CONFIG_CPUSETS
1100 /* Protected by ->alloc_lock: */
1101 nodemask_t mems_allowed;
1102 /* Seqence number to catch updates: */
1103 seqcount_spinlock_t mems_allowed_seq;
1104 int cpuset_mem_spread_rotor;
1105 int cpuset_slab_spread_rotor;
1107 #ifdef CONFIG_CGROUPS
1108 /* Control Group info protected by css_set_lock: */
1109 struct css_set __rcu *cgroups;
1110 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1111 struct list_head cg_list;
1113 #ifdef CONFIG_X86_CPU_RESCTRL
1118 struct robust_list_head __user *robust_list;
1119 #ifdef CONFIG_COMPAT
1120 struct compat_robust_list_head __user *compat_robust_list;
1122 struct list_head pi_state_list;
1123 struct futex_pi_state *pi_state_cache;
1124 struct mutex futex_exit_mutex;
1125 unsigned int futex_state;
1127 #ifdef CONFIG_PERF_EVENTS
1128 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1129 struct mutex perf_event_mutex;
1130 struct list_head perf_event_list;
1132 #ifdef CONFIG_DEBUG_PREEMPT
1133 unsigned long preempt_disable_ip;
1136 /* Protected by alloc_lock: */
1137 struct mempolicy *mempolicy;
1139 short pref_node_fork;
1141 #ifdef CONFIG_NUMA_BALANCING
1143 unsigned int numa_scan_period;
1144 unsigned int numa_scan_period_max;
1145 int numa_preferred_nid;
1146 unsigned long numa_migrate_retry;
1147 /* Migration stamp: */
1149 u64 last_task_numa_placement;
1150 u64 last_sum_exec_runtime;
1151 struct callback_head numa_work;
1154 * This pointer is only modified for current in syscall and
1155 * pagefault context (and for tasks being destroyed), so it can be read
1156 * from any of the following contexts:
1157 * - RCU read-side critical section
1158 * - current->numa_group from everywhere
1159 * - task's runqueue locked, task not running
1161 struct numa_group __rcu *numa_group;
1164 * numa_faults is an array split into four regions:
1165 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1166 * in this precise order.
1168 * faults_memory: Exponential decaying average of faults on a per-node
1169 * basis. Scheduling placement decisions are made based on these
1170 * counts. The values remain static for the duration of a PTE scan.
1171 * faults_cpu: Track the nodes the process was running on when a NUMA
1172 * hinting fault was incurred.
1173 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1174 * during the current scan window. When the scan completes, the counts
1175 * in faults_memory and faults_cpu decay and these values are copied.
1177 unsigned long *numa_faults;
1178 unsigned long total_numa_faults;
1181 * numa_faults_locality tracks if faults recorded during the last
1182 * scan window were remote/local or failed to migrate. The task scan
1183 * period is adapted based on the locality of the faults with different
1184 * weights depending on whether they were shared or private faults
1186 unsigned long numa_faults_locality[3];
1188 unsigned long numa_pages_migrated;
1189 #endif /* CONFIG_NUMA_BALANCING */
1192 struct rseq __user *rseq;
1195 * RmW on rseq_event_mask must be performed atomically
1196 * with respect to preemption.
1198 unsigned long rseq_event_mask;
1201 struct tlbflush_unmap_batch tlb_ubc;
1204 refcount_t rcu_users;
1205 struct rcu_head rcu;
1208 /* Cache last used pipe for splice(): */
1209 struct pipe_inode_info *splice_pipe;
1211 struct page_frag task_frag;
1213 #ifdef CONFIG_TASK_DELAY_ACCT
1214 struct task_delay_info *delays;
1217 #ifdef CONFIG_FAULT_INJECTION
1219 unsigned int fail_nth;
1222 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1223 * balance_dirty_pages() for a dirty throttling pause:
1226 int nr_dirtied_pause;
1227 /* Start of a write-and-pause period: */
1228 unsigned long dirty_paused_when;
1230 #ifdef CONFIG_LATENCYTOP
1231 int latency_record_count;
1232 struct latency_record latency_record[LT_SAVECOUNT];
1235 * Time slack values; these are used to round up poll() and
1236 * select() etc timeout values. These are in nanoseconds.
1239 u64 default_timer_slack_ns;
1241 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1242 unsigned int kasan_depth;
1246 struct kcsan_ctx kcsan_ctx;
1247 #ifdef CONFIG_TRACE_IRQFLAGS
1248 struct irqtrace_events kcsan_save_irqtrace;
1252 #if IS_ENABLED(CONFIG_KUNIT)
1253 struct kunit *kunit_test;
1256 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1257 /* Index of current stored address in ret_stack: */
1261 /* Stack of return addresses for return function tracing: */
1262 struct ftrace_ret_stack *ret_stack;
1264 /* Timestamp for last schedule: */
1265 unsigned long long ftrace_timestamp;
1268 * Number of functions that haven't been traced
1269 * because of depth overrun:
1271 atomic_t trace_overrun;
1273 /* Pause tracing: */
1274 atomic_t tracing_graph_pause;
1277 #ifdef CONFIG_TRACING
1278 /* State flags for use by tracers: */
1279 unsigned long trace;
1281 /* Bitmask and counter of trace recursion: */
1282 unsigned long trace_recursion;
1283 #endif /* CONFIG_TRACING */
1286 /* See kernel/kcov.c for more details. */
1288 /* Coverage collection mode enabled for this task (0 if disabled): */
1289 unsigned int kcov_mode;
1291 /* Size of the kcov_area: */
1292 unsigned int kcov_size;
1294 /* Buffer for coverage collection: */
1297 /* KCOV descriptor wired with this task or NULL: */
1300 /* KCOV common handle for remote coverage collection: */
1303 /* KCOV sequence number: */
1306 /* Collect coverage from softirq context: */
1307 unsigned int kcov_softirq;
1311 struct mem_cgroup *memcg_in_oom;
1312 gfp_t memcg_oom_gfp_mask;
1313 int memcg_oom_order;
1315 /* Number of pages to reclaim on returning to userland: */
1316 unsigned int memcg_nr_pages_over_high;
1318 /* Used by memcontrol for targeted memcg charge: */
1319 struct mem_cgroup *active_memcg;
1322 #ifdef CONFIG_BLK_CGROUP
1323 struct request_queue *throttle_queue;
1326 #ifdef CONFIG_UPROBES
1327 struct uprobe_task *utask;
1329 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1330 unsigned int sequential_io;
1331 unsigned int sequential_io_avg;
1333 struct kmap_ctrl kmap_ctrl;
1334 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1335 unsigned long task_state_change;
1337 int pagefault_disabled;
1339 struct task_struct *oom_reaper_list;
1341 #ifdef CONFIG_VMAP_STACK
1342 struct vm_struct *stack_vm_area;
1344 #ifdef CONFIG_THREAD_INFO_IN_TASK
1345 /* A live task holds one reference: */
1346 refcount_t stack_refcount;
1348 #ifdef CONFIG_LIVEPATCH
1351 #ifdef CONFIG_SECURITY
1352 /* Used by LSM modules for access restriction: */
1355 #ifdef CONFIG_BPF_SYSCALL
1356 /* Used by BPF task local storage */
1357 struct bpf_local_storage __rcu *bpf_storage;
1360 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1361 unsigned long lowest_stack;
1362 unsigned long prev_lowest_stack;
1365 #ifdef CONFIG_X86_MCE
1366 void __user *mce_vaddr;
1371 __mce_reserved : 62;
1372 struct callback_head mce_kill_me;
1375 #ifdef CONFIG_KRETPROBES
1376 struct llist_head kretprobe_instances;
1380 * New fields for task_struct should be added above here, so that
1381 * they are included in the randomized portion of task_struct.
1383 randomized_struct_fields_end
1385 /* CPU-specific state of this task: */
1386 struct thread_struct thread;
1389 * WARNING: on x86, 'thread_struct' contains a variable-sized
1390 * structure. It *MUST* be at the end of 'task_struct'.
1392 * Do not put anything below here!
1396 static inline struct pid *task_pid(struct task_struct *task)
1398 return task->thread_pid;
1402 * the helpers to get the task's different pids as they are seen
1403 * from various namespaces
1405 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1406 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1408 * task_xid_nr_ns() : id seen from the ns specified;
1410 * see also pid_nr() etc in include/linux/pid.h
1412 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1414 static inline pid_t task_pid_nr(struct task_struct *tsk)
1419 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1421 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1424 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1426 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1430 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1436 * pid_alive - check that a task structure is not stale
1437 * @p: Task structure to be checked.
1439 * Test if a process is not yet dead (at most zombie state)
1440 * If pid_alive fails, then pointers within the task structure
1441 * can be stale and must not be dereferenced.
1443 * Return: 1 if the process is alive. 0 otherwise.
1445 static inline int pid_alive(const struct task_struct *p)
1447 return p->thread_pid != NULL;
1450 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1452 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1455 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1457 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1461 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1463 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1466 static inline pid_t task_session_vnr(struct task_struct *tsk)
1468 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1471 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1473 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1476 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1478 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1481 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1487 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1493 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1495 return task_ppid_nr_ns(tsk, &init_pid_ns);
1498 /* Obsolete, do not use: */
1499 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1501 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1504 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1505 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1507 static inline unsigned int task_state_index(struct task_struct *tsk)
1509 unsigned int tsk_state = READ_ONCE(tsk->state);
1510 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1512 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1514 if (tsk_state == TASK_IDLE)
1515 state = TASK_REPORT_IDLE;
1520 static inline char task_index_to_char(unsigned int state)
1522 static const char state_char[] = "RSDTtXZPI";
1524 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1526 return state_char[state];
1529 static inline char task_state_to_char(struct task_struct *tsk)
1531 return task_index_to_char(task_state_index(tsk));
1535 * is_global_init - check if a task structure is init. Since init
1536 * is free to have sub-threads we need to check tgid.
1537 * @tsk: Task structure to be checked.
1539 * Check if a task structure is the first user space task the kernel created.
1541 * Return: 1 if the task structure is init. 0 otherwise.
1543 static inline int is_global_init(struct task_struct *tsk)
1545 return task_tgid_nr(tsk) == 1;
1548 extern struct pid *cad_pid;
1553 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1554 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1555 #define PF_EXITING 0x00000004 /* Getting shut down */
1556 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1557 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1558 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1559 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1560 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1561 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1562 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1563 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1564 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1565 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1566 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1567 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1568 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1569 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1570 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1571 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1572 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1573 * I am cleaning dirty pages from some other bdi. */
1574 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1575 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1576 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1577 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1578 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1579 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1580 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1581 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1584 * Only the _current_ task can read/write to tsk->flags, but other
1585 * tasks can access tsk->flags in readonly mode for example
1586 * with tsk_used_math (like during threaded core dumping).
1587 * There is however an exception to this rule during ptrace
1588 * or during fork: the ptracer task is allowed to write to the
1589 * child->flags of its traced child (same goes for fork, the parent
1590 * can write to the child->flags), because we're guaranteed the
1591 * child is not running and in turn not changing child->flags
1592 * at the same time the parent does it.
1594 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1595 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1596 #define clear_used_math() clear_stopped_child_used_math(current)
1597 #define set_used_math() set_stopped_child_used_math(current)
1599 #define conditional_stopped_child_used_math(condition, child) \
1600 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1602 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1604 #define copy_to_stopped_child_used_math(child) \
1605 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1607 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1608 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1609 #define used_math() tsk_used_math(current)
1611 static inline bool is_percpu_thread(void)
1614 return (current->flags & PF_NO_SETAFFINITY) &&
1615 (current->nr_cpus_allowed == 1);
1621 /* Per-process atomic flags. */
1622 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1623 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1624 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1625 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1626 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1627 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1628 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1629 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1631 #define TASK_PFA_TEST(name, func) \
1632 static inline bool task_##func(struct task_struct *p) \
1633 { return test_bit(PFA_##name, &p->atomic_flags); }
1635 #define TASK_PFA_SET(name, func) \
1636 static inline void task_set_##func(struct task_struct *p) \
1637 { set_bit(PFA_##name, &p->atomic_flags); }
1639 #define TASK_PFA_CLEAR(name, func) \
1640 static inline void task_clear_##func(struct task_struct *p) \
1641 { clear_bit(PFA_##name, &p->atomic_flags); }
1643 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1644 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1646 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1647 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1648 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1650 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1651 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1652 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1654 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1655 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1656 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1658 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1659 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1660 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1662 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1663 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1665 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1666 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1667 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1669 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1670 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1673 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1675 current->flags &= ~flags;
1676 current->flags |= orig_flags & flags;
1679 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1680 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1682 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1683 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1685 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1688 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1690 if (!cpumask_test_cpu(0, new_mask))
1696 extern int yield_to(struct task_struct *p, bool preempt);
1697 extern void set_user_nice(struct task_struct *p, long nice);
1698 extern int task_prio(const struct task_struct *p);
1701 * task_nice - return the nice value of a given task.
1702 * @p: the task in question.
1704 * Return: The nice value [ -20 ... 0 ... 19 ].
1706 static inline int task_nice(const struct task_struct *p)
1708 return PRIO_TO_NICE((p)->static_prio);
1711 extern int can_nice(const struct task_struct *p, const int nice);
1712 extern int task_curr(const struct task_struct *p);
1713 extern int idle_cpu(int cpu);
1714 extern int available_idle_cpu(int cpu);
1715 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1716 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1717 extern void sched_set_fifo(struct task_struct *p);
1718 extern void sched_set_fifo_low(struct task_struct *p);
1719 extern void sched_set_normal(struct task_struct *p, int nice);
1720 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1721 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1722 extern struct task_struct *idle_task(int cpu);
1725 * is_idle_task - is the specified task an idle task?
1726 * @p: the task in question.
1728 * Return: 1 if @p is an idle task. 0 otherwise.
1730 static __always_inline bool is_idle_task(const struct task_struct *p)
1732 return !!(p->flags & PF_IDLE);
1735 extern struct task_struct *curr_task(int cpu);
1736 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1740 union thread_union {
1741 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1742 struct task_struct task;
1744 #ifndef CONFIG_THREAD_INFO_IN_TASK
1745 struct thread_info thread_info;
1747 unsigned long stack[THREAD_SIZE/sizeof(long)];
1750 #ifndef CONFIG_THREAD_INFO_IN_TASK
1751 extern struct thread_info init_thread_info;
1754 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1756 #ifdef CONFIG_THREAD_INFO_IN_TASK
1757 static inline struct thread_info *task_thread_info(struct task_struct *task)
1759 return &task->thread_info;
1761 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1762 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1766 * find a task by one of its numerical ids
1768 * find_task_by_pid_ns():
1769 * finds a task by its pid in the specified namespace
1770 * find_task_by_vpid():
1771 * finds a task by its virtual pid
1773 * see also find_vpid() etc in include/linux/pid.h
1776 extern struct task_struct *find_task_by_vpid(pid_t nr);
1777 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1780 * find a task by its virtual pid and get the task struct
1782 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1784 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1785 extern int wake_up_process(struct task_struct *tsk);
1786 extern void wake_up_new_task(struct task_struct *tsk);
1789 extern void kick_process(struct task_struct *tsk);
1791 static inline void kick_process(struct task_struct *tsk) { }
1794 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1796 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1798 __set_task_comm(tsk, from, false);
1801 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1802 #define get_task_comm(buf, tsk) ({ \
1803 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1804 __get_task_comm(buf, sizeof(buf), tsk); \
1808 static __always_inline void scheduler_ipi(void)
1811 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1812 * TIF_NEED_RESCHED remotely (for the first time) will also send
1815 preempt_fold_need_resched();
1817 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1819 static inline void scheduler_ipi(void) { }
1820 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1827 * Set thread flags in other task's structures.
1828 * See asm/thread_info.h for TIF_xxxx flags available:
1830 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1832 set_ti_thread_flag(task_thread_info(tsk), flag);
1835 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1837 clear_ti_thread_flag(task_thread_info(tsk), flag);
1840 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1843 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1846 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1848 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1851 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1853 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1856 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1858 return test_ti_thread_flag(task_thread_info(tsk), flag);
1861 static inline void set_tsk_need_resched(struct task_struct *tsk)
1863 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1866 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1868 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1871 static inline int test_tsk_need_resched(struct task_struct *tsk)
1873 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1877 * cond_resched() and cond_resched_lock(): latency reduction via
1878 * explicit rescheduling in places that are safe. The return
1879 * value indicates whether a reschedule was done in fact.
1880 * cond_resched_lock() will drop the spinlock before scheduling,
1882 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1883 extern int __cond_resched(void);
1885 #ifdef CONFIG_PREEMPT_DYNAMIC
1887 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1889 static __always_inline int _cond_resched(void)
1891 return static_call_mod(cond_resched)();
1896 static inline int _cond_resched(void)
1898 return __cond_resched();
1901 #endif /* CONFIG_PREEMPT_DYNAMIC */
1905 static inline int _cond_resched(void) { return 0; }
1907 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
1909 #define cond_resched() ({ \
1910 ___might_sleep(__FILE__, __LINE__, 0); \
1914 extern int __cond_resched_lock(spinlock_t *lock);
1915 extern int __cond_resched_rwlock_read(rwlock_t *lock);
1916 extern int __cond_resched_rwlock_write(rwlock_t *lock);
1918 #define cond_resched_lock(lock) ({ \
1919 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1920 __cond_resched_lock(lock); \
1923 #define cond_resched_rwlock_read(lock) ({ \
1924 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1925 __cond_resched_rwlock_read(lock); \
1928 #define cond_resched_rwlock_write(lock) ({ \
1929 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1930 __cond_resched_rwlock_write(lock); \
1933 static inline void cond_resched_rcu(void)
1935 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1943 * Does a critical section need to be broken due to another
1944 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1945 * but a general need for low latency)
1947 static inline int spin_needbreak(spinlock_t *lock)
1949 #ifdef CONFIG_PREEMPTION
1950 return spin_is_contended(lock);
1957 * Check if a rwlock is contended.
1958 * Returns non-zero if there is another task waiting on the rwlock.
1959 * Returns zero if the lock is not contended or the system / underlying
1960 * rwlock implementation does not support contention detection.
1961 * Technically does not depend on CONFIG_PREEMPTION, but a general need
1964 static inline int rwlock_needbreak(rwlock_t *lock)
1966 #ifdef CONFIG_PREEMPTION
1967 return rwlock_is_contended(lock);
1973 static __always_inline bool need_resched(void)
1975 return unlikely(tif_need_resched());
1979 * Wrappers for p->thread_info->cpu access. No-op on UP.
1983 static inline unsigned int task_cpu(const struct task_struct *p)
1985 #ifdef CONFIG_THREAD_INFO_IN_TASK
1986 return READ_ONCE(p->cpu);
1988 return READ_ONCE(task_thread_info(p)->cpu);
1992 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1996 static inline unsigned int task_cpu(const struct task_struct *p)
2001 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2005 #endif /* CONFIG_SMP */
2008 * In order to reduce various lock holder preemption latencies provide an
2009 * interface to see if a vCPU is currently running or not.
2011 * This allows us to terminate optimistic spin loops and block, analogous to
2012 * the native optimistic spin heuristic of testing if the lock owner task is
2015 #ifndef vcpu_is_preempted
2016 static inline bool vcpu_is_preempted(int cpu)
2022 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2023 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2025 #ifndef TASK_SIZE_OF
2026 #define TASK_SIZE_OF(tsk) TASK_SIZE
2030 /* Returns effective CPU energy utilization, as seen by the scheduler */
2031 unsigned long sched_cpu_util(int cpu, unsigned long max);
2032 #endif /* CONFIG_SMP */
2037 * Map the event mask on the user-space ABI enum rseq_cs_flags
2038 * for direct mask checks.
2040 enum rseq_event_mask_bits {
2041 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2042 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2043 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2046 enum rseq_event_mask {
2047 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2048 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2049 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2052 static inline void rseq_set_notify_resume(struct task_struct *t)
2055 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2058 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2060 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2061 struct pt_regs *regs)
2064 __rseq_handle_notify_resume(ksig, regs);
2067 static inline void rseq_signal_deliver(struct ksignal *ksig,
2068 struct pt_regs *regs)
2071 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2073 rseq_handle_notify_resume(ksig, regs);
2076 /* rseq_preempt() requires preemption to be disabled. */
2077 static inline void rseq_preempt(struct task_struct *t)
2079 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2080 rseq_set_notify_resume(t);
2083 /* rseq_migrate() requires preemption to be disabled. */
2084 static inline void rseq_migrate(struct task_struct *t)
2086 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2087 rseq_set_notify_resume(t);
2091 * If parent process has a registered restartable sequences area, the
2092 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2094 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2096 if (clone_flags & CLONE_VM) {
2099 t->rseq_event_mask = 0;
2101 t->rseq = current->rseq;
2102 t->rseq_sig = current->rseq_sig;
2103 t->rseq_event_mask = current->rseq_event_mask;
2107 static inline void rseq_execve(struct task_struct *t)
2111 t->rseq_event_mask = 0;
2116 static inline void rseq_set_notify_resume(struct task_struct *t)
2119 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2120 struct pt_regs *regs)
2123 static inline void rseq_signal_deliver(struct ksignal *ksig,
2124 struct pt_regs *regs)
2127 static inline void rseq_preempt(struct task_struct *t)
2130 static inline void rseq_migrate(struct task_struct *t)
2133 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2136 static inline void rseq_execve(struct task_struct *t)
2142 #ifdef CONFIG_DEBUG_RSEQ
2144 void rseq_syscall(struct pt_regs *regs);
2148 static inline void rseq_syscall(struct pt_regs *regs)
2154 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2155 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2156 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2158 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2159 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2160 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2162 int sched_trace_rq_cpu(struct rq *rq);
2163 int sched_trace_rq_cpu_capacity(struct rq *rq);
2164 int sched_trace_rq_nr_running(struct rq *rq);
2166 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);