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;
68 * Task state bitmask. NOTE! These bits are also
69 * encoded in fs/proc/array.c: get_task_state().
71 * We have two separate sets of flags: task->state
72 * is about runnability, while task->exit_state are
73 * about the task exiting. Confusing, but this way
74 * modifying one set can't modify the other one by
78 /* Used in tsk->state: */
79 #define TASK_RUNNING 0x0000
80 #define TASK_INTERRUPTIBLE 0x0001
81 #define TASK_UNINTERRUPTIBLE 0x0002
82 #define __TASK_STOPPED 0x0004
83 #define __TASK_TRACED 0x0008
84 /* Used in tsk->exit_state: */
85 #define EXIT_DEAD 0x0010
86 #define EXIT_ZOMBIE 0x0020
87 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
88 /* Used in tsk->state again: */
89 #define TASK_PARKED 0x0040
90 #define TASK_DEAD 0x0080
91 #define TASK_WAKEKILL 0x0100
92 #define TASK_WAKING 0x0200
93 #define TASK_NOLOAD 0x0400
94 #define TASK_NEW 0x0800
95 #define TASK_STATE_MAX 0x1000
97 /* Convenience macros for the sake of set_current_state: */
98 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
99 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
100 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
102 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
104 /* Convenience macros for the sake of wake_up(): */
105 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
107 /* get_task_state(): */
108 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
109 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
110 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
113 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
115 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
117 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
119 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
122 * Special states are those that do not use the normal wait-loop pattern. See
123 * the comment with set_special_state().
125 #define is_special_task_state(state) \
126 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
128 #define __set_current_state(state_value) \
130 WARN_ON_ONCE(is_special_task_state(state_value));\
131 current->task_state_change = _THIS_IP_; \
132 current->state = (state_value); \
135 #define set_current_state(state_value) \
137 WARN_ON_ONCE(is_special_task_state(state_value));\
138 current->task_state_change = _THIS_IP_; \
139 smp_store_mb(current->state, (state_value)); \
142 #define set_special_state(state_value) \
144 unsigned long flags; /* may shadow */ \
145 WARN_ON_ONCE(!is_special_task_state(state_value)); \
146 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
147 current->task_state_change = _THIS_IP_; \
148 current->state = (state_value); \
149 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
153 * set_current_state() includes a barrier so that the write of current->state
154 * is correctly serialised wrt the caller's subsequent test of whether to
158 * set_current_state(TASK_UNINTERRUPTIBLE);
164 * __set_current_state(TASK_RUNNING);
166 * If the caller does not need such serialisation (because, for instance, the
167 * CONDITION test and condition change and wakeup are under the same lock) then
168 * use __set_current_state().
170 * The above is typically ordered against the wakeup, which does:
173 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
175 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
176 * accessing p->state.
178 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
179 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
180 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
182 * However, with slightly different timing the wakeup TASK_RUNNING store can
183 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
184 * a problem either because that will result in one extra go around the loop
185 * and our @cond test will save the day.
187 * Also see the comments of try_to_wake_up().
189 #define __set_current_state(state_value) \
190 current->state = (state_value)
192 #define set_current_state(state_value) \
193 smp_store_mb(current->state, (state_value))
196 * set_special_state() should be used for those states when the blocking task
197 * can not use the regular condition based wait-loop. In that case we must
198 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
199 * will not collide with our state change.
201 #define set_special_state(state_value) \
203 unsigned long flags; /* may shadow */ \
204 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
205 current->state = (state_value); \
206 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
211 /* Task command name length: */
212 #define TASK_COMM_LEN 16
214 extern void scheduler_tick(void);
216 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
218 extern long schedule_timeout(long timeout);
219 extern long schedule_timeout_interruptible(long timeout);
220 extern long schedule_timeout_killable(long timeout);
221 extern long schedule_timeout_uninterruptible(long timeout);
222 extern long schedule_timeout_idle(long timeout);
223 asmlinkage void schedule(void);
224 extern void schedule_preempt_disabled(void);
225 asmlinkage void preempt_schedule_irq(void);
227 extern int __must_check io_schedule_prepare(void);
228 extern void io_schedule_finish(int token);
229 extern long io_schedule_timeout(long timeout);
230 extern void io_schedule(void);
233 * struct prev_cputime - snapshot of system and user cputime
234 * @utime: time spent in user mode
235 * @stime: time spent in system mode
236 * @lock: protects the above two fields
238 * Stores previous user/system time values such that we can guarantee
241 struct prev_cputime {
242 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
250 /* Task is sleeping or running in a CPU with VTIME inactive: */
254 /* Task runs in kernelspace in a CPU with VTIME active: */
256 /* Task runs in userspace in a CPU with VTIME active: */
258 /* Task runs as guests in a CPU with VTIME active: */
264 unsigned long long starttime;
265 enum vtime_state state;
273 * Utilization clamp constraints.
274 * @UCLAMP_MIN: Minimum utilization
275 * @UCLAMP_MAX: Maximum utilization
276 * @UCLAMP_CNT: Utilization clamp constraints count
285 extern struct root_domain def_root_domain;
286 extern struct mutex sched_domains_mutex;
290 #ifdef CONFIG_SCHED_INFO
291 /* Cumulative counters: */
293 /* # of times we have run on this CPU: */
294 unsigned long pcount;
296 /* Time spent waiting on a runqueue: */
297 unsigned long long run_delay;
301 /* When did we last run on a CPU? */
302 unsigned long long last_arrival;
304 /* When were we last queued to run? */
305 unsigned long long last_queued;
307 #endif /* CONFIG_SCHED_INFO */
311 * Integer metrics need fixed point arithmetic, e.g., sched/fair
312 * has a few: load, load_avg, util_avg, freq, and capacity.
314 * We define a basic fixed point arithmetic range, and then formalize
315 * all these metrics based on that basic range.
317 # define SCHED_FIXEDPOINT_SHIFT 10
318 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
320 /* Increase resolution of cpu_capacity calculations */
321 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
322 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
325 unsigned long weight;
330 * struct util_est - Estimation utilization of FAIR tasks
331 * @enqueued: instantaneous estimated utilization of a task/cpu
332 * @ewma: the Exponential Weighted Moving Average (EWMA)
333 * utilization of a task
335 * Support data structure to track an Exponential Weighted Moving Average
336 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
337 * average each time a task completes an activation. Sample's weight is chosen
338 * so that the EWMA will be relatively insensitive to transient changes to the
341 * The enqueued attribute has a slightly different meaning for tasks and cpus:
342 * - task: the task's util_avg at last task dequeue time
343 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
344 * Thus, the util_est.enqueued of a task represents the contribution on the
345 * estimated utilization of the CPU where that task is currently enqueued.
347 * Only for tasks we track a moving average of the past instantaneous
348 * estimated utilization. This allows to absorb sporadic drops in utilization
349 * of an otherwise almost periodic task.
352 unsigned int enqueued;
354 #define UTIL_EST_WEIGHT_SHIFT 2
355 } __attribute__((__aligned__(sizeof(u64))));
358 * The load/runnable/util_avg accumulates an infinite geometric series
359 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
361 * [load_avg definition]
363 * load_avg = runnable% * scale_load_down(load)
365 * [runnable_avg definition]
367 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
369 * [util_avg definition]
371 * util_avg = running% * SCHED_CAPACITY_SCALE
373 * where runnable% is the time ratio that a sched_entity is runnable and
374 * running% the time ratio that a sched_entity is running.
376 * For cfs_rq, they are the aggregated values of all runnable and blocked
379 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
380 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
381 * for computing those signals (see update_rq_clock_pelt())
383 * N.B., the above ratios (runnable% and running%) themselves are in the
384 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
385 * to as large a range as necessary. This is for example reflected by
386 * util_avg's SCHED_CAPACITY_SCALE.
390 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
391 * with the highest load (=88761), always runnable on a single cfs_rq,
392 * and should not overflow as the number already hits PID_MAX_LIMIT.
394 * For all other cases (including 32-bit kernels), struct load_weight's
395 * weight will overflow first before we do, because:
397 * Max(load_avg) <= Max(load.weight)
399 * Then it is the load_weight's responsibility to consider overflow
403 u64 last_update_time;
408 unsigned long load_avg;
409 unsigned long runnable_avg;
410 unsigned long util_avg;
411 struct util_est util_est;
412 } ____cacheline_aligned;
414 struct sched_statistics {
415 #ifdef CONFIG_SCHEDSTATS
425 s64 sum_sleep_runtime;
432 u64 nr_migrations_cold;
433 u64 nr_failed_migrations_affine;
434 u64 nr_failed_migrations_running;
435 u64 nr_failed_migrations_hot;
436 u64 nr_forced_migrations;
440 u64 nr_wakeups_migrate;
441 u64 nr_wakeups_local;
442 u64 nr_wakeups_remote;
443 u64 nr_wakeups_affine;
444 u64 nr_wakeups_affine_attempts;
445 u64 nr_wakeups_passive;
450 struct sched_entity {
451 /* For load-balancing: */
452 struct load_weight load;
453 struct rb_node run_node;
454 struct list_head group_node;
458 u64 sum_exec_runtime;
460 u64 prev_sum_exec_runtime;
464 struct sched_statistics statistics;
466 #ifdef CONFIG_FAIR_GROUP_SCHED
468 struct sched_entity *parent;
469 /* rq on which this entity is (to be) queued: */
470 struct cfs_rq *cfs_rq;
471 /* rq "owned" by this entity/group: */
473 /* cached value of my_q->h_nr_running */
474 unsigned long runnable_weight;
479 * Per entity load average tracking.
481 * Put into separate cache line so it does not
482 * collide with read-mostly values above.
484 struct sched_avg avg;
488 struct sched_rt_entity {
489 struct list_head run_list;
490 unsigned long timeout;
491 unsigned long watchdog_stamp;
492 unsigned int time_slice;
493 unsigned short on_rq;
494 unsigned short on_list;
496 struct sched_rt_entity *back;
497 #ifdef CONFIG_RT_GROUP_SCHED
498 struct sched_rt_entity *parent;
499 /* rq on which this entity is (to be) queued: */
501 /* rq "owned" by this entity/group: */
504 } __randomize_layout;
506 struct sched_dl_entity {
507 struct rb_node rb_node;
510 * Original scheduling parameters. Copied here from sched_attr
511 * during sched_setattr(), they will remain the same until
512 * the next sched_setattr().
514 u64 dl_runtime; /* Maximum runtime for each instance */
515 u64 dl_deadline; /* Relative deadline of each instance */
516 u64 dl_period; /* Separation of two instances (period) */
517 u64 dl_bw; /* dl_runtime / dl_period */
518 u64 dl_density; /* dl_runtime / dl_deadline */
521 * Actual scheduling parameters. Initialized with the values above,
522 * they are continuously updated during task execution. Note that
523 * the remaining runtime could be < 0 in case we are in overrun.
525 s64 runtime; /* Remaining runtime for this instance */
526 u64 deadline; /* Absolute deadline for this instance */
527 unsigned int flags; /* Specifying the scheduler behaviour */
532 * @dl_throttled tells if we exhausted the runtime. If so, the
533 * task has to wait for a replenishment to be performed at the
534 * next firing of dl_timer.
536 * @dl_boosted tells if we are boosted due to DI. If so we are
537 * outside bandwidth enforcement mechanism (but only until we
538 * exit the critical section);
540 * @dl_yielded tells if task gave up the CPU before consuming
541 * all its available runtime during the last job.
543 * @dl_non_contending tells if the task is inactive while still
544 * contributing to the active utilization. In other words, it
545 * indicates if the inactive timer has been armed and its handler
546 * has not been executed yet. This flag is useful to avoid race
547 * conditions between the inactive timer handler and the wakeup
550 * @dl_overrun tells if the task asked to be informed about runtime
553 unsigned int dl_throttled : 1;
554 unsigned int dl_boosted : 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;
575 #ifdef CONFIG_UCLAMP_TASK
576 /* Number of utilization clamp buckets (shorter alias) */
577 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
580 * Utilization clamp for a scheduling entity
581 * @value: clamp value "assigned" to a se
582 * @bucket_id: bucket index corresponding to the "assigned" value
583 * @active: the se is currently refcounted in a rq's bucket
584 * @user_defined: the requested clamp value comes from user-space
586 * The bucket_id is the index of the clamp bucket matching the clamp value
587 * which is pre-computed and stored to avoid expensive integer divisions from
590 * The active bit is set whenever a task has got an "effective" value assigned,
591 * which can be different from the clamp value "requested" from user-space.
592 * This allows to know a task is refcounted in the rq's bucket corresponding
593 * to the "effective" bucket_id.
595 * The user_defined bit is set whenever a task has got a task-specific clamp
596 * value requested from userspace, i.e. the system defaults apply to this task
597 * just as a restriction. This allows to relax default clamps when a less
598 * restrictive task-specific value has been requested, thus allowing to
599 * implement a "nice" semantic. For example, a task running with a 20%
600 * default boost can still drop its own boosting to 0%.
603 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
604 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
605 unsigned int active : 1;
606 unsigned int user_defined : 1;
608 #endif /* CONFIG_UCLAMP_TASK */
614 u8 exp_hint; /* Hint for performance. */
615 u8 need_mb; /* Readers need smp_mb(). */
617 u32 s; /* Set of bits. */
620 enum perf_event_task_context {
621 perf_invalid_context = -1,
624 perf_nr_task_contexts,
628 struct wake_q_node *next;
632 #ifdef CONFIG_THREAD_INFO_IN_TASK
634 * For reasons of header soup (see current_thread_info()), this
635 * must be the first element of task_struct.
637 struct thread_info thread_info;
639 /* -1 unrunnable, 0 runnable, >0 stopped: */
643 * This begins the randomizable portion of task_struct. Only
644 * scheduling-critical items should be added above here.
646 randomized_struct_fields_start
650 /* Per task flags (PF_*), defined further below: */
656 struct __call_single_node wake_entry;
657 #ifdef CONFIG_THREAD_INFO_IN_TASK
661 unsigned int wakee_flips;
662 unsigned long wakee_flip_decay_ts;
663 struct task_struct *last_wakee;
666 * recent_used_cpu is initially set as the last CPU used by a task
667 * that wakes affine another task. Waker/wakee relationships can
668 * push tasks around a CPU where each wakeup moves to the next one.
669 * Tracking a recently used CPU allows a quick search for a recently
670 * used CPU that may be idle.
680 unsigned int rt_priority;
682 const struct sched_class *sched_class;
683 struct sched_entity se;
684 struct sched_rt_entity rt;
685 #ifdef CONFIG_CGROUP_SCHED
686 struct task_group *sched_task_group;
688 struct sched_dl_entity dl;
690 #ifdef CONFIG_UCLAMP_TASK
692 * Clamp values requested for a scheduling entity.
693 * Must be updated with task_rq_lock() held.
695 struct uclamp_se uclamp_req[UCLAMP_CNT];
697 * Effective clamp values used for a scheduling entity.
698 * Must be updated with task_rq_lock() held.
700 struct uclamp_se uclamp[UCLAMP_CNT];
703 #ifdef CONFIG_PREEMPT_NOTIFIERS
704 /* List of struct preempt_notifier: */
705 struct hlist_head preempt_notifiers;
708 #ifdef CONFIG_BLK_DEV_IO_TRACE
709 unsigned int btrace_seq;
714 const cpumask_t *cpus_ptr;
717 #ifdef CONFIG_PREEMPT_RCU
718 int rcu_read_lock_nesting;
719 union rcu_special rcu_read_unlock_special;
720 struct list_head rcu_node_entry;
721 struct rcu_node *rcu_blocked_node;
722 #endif /* #ifdef CONFIG_PREEMPT_RCU */
724 #ifdef CONFIG_TASKS_RCU
725 unsigned long rcu_tasks_nvcsw;
726 u8 rcu_tasks_holdout;
728 int rcu_tasks_idle_cpu;
729 struct list_head rcu_tasks_holdout_list;
730 #endif /* #ifdef CONFIG_TASKS_RCU */
732 #ifdef CONFIG_TASKS_TRACE_RCU
733 int trc_reader_nesting;
735 union rcu_special trc_reader_special;
736 bool trc_reader_checked;
737 struct list_head trc_holdout_list;
738 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
740 struct sched_info sched_info;
742 struct list_head tasks;
744 struct plist_node pushable_tasks;
745 struct rb_node pushable_dl_tasks;
748 struct mm_struct *mm;
749 struct mm_struct *active_mm;
751 /* Per-thread vma caching: */
752 struct vmacache vmacache;
754 #ifdef SPLIT_RSS_COUNTING
755 struct task_rss_stat rss_stat;
760 /* The signal sent when the parent dies: */
762 /* JOBCTL_*, siglock protected: */
763 unsigned long jobctl;
765 /* Used for emulating ABI behavior of previous Linux versions: */
766 unsigned int personality;
768 /* Scheduler bits, serialized by scheduler locks: */
769 unsigned sched_reset_on_fork:1;
770 unsigned sched_contributes_to_load:1;
771 unsigned sched_migrated:1;
772 unsigned sched_remote_wakeup:1;
774 unsigned sched_psi_wake_requeue:1;
777 /* Force alignment to the next boundary: */
780 /* Unserialized, strictly 'current' */
782 /* Bit to tell LSMs we're in execve(): */
783 unsigned in_execve:1;
784 unsigned in_iowait:1;
785 #ifndef TIF_RESTORE_SIGMASK
786 unsigned restore_sigmask:1;
789 unsigned in_user_fault:1;
791 #ifdef CONFIG_COMPAT_BRK
792 unsigned brk_randomized:1;
794 #ifdef CONFIG_CGROUPS
795 /* disallow userland-initiated cgroup migration */
796 unsigned no_cgroup_migration:1;
797 /* task is frozen/stopped (used by the cgroup freezer) */
800 #ifdef CONFIG_BLK_CGROUP
801 unsigned use_memdelay:1;
804 /* Stalled due to lack of memory */
805 unsigned in_memstall:1;
808 unsigned long atomic_flags; /* Flags requiring atomic access. */
810 struct restart_block restart_block;
815 #ifdef CONFIG_STACKPROTECTOR
816 /* Canary value for the -fstack-protector GCC feature: */
817 unsigned long stack_canary;
820 * Pointers to the (original) parent process, youngest child, younger sibling,
821 * older sibling, respectively. (p->father can be replaced with
822 * p->real_parent->pid)
825 /* Real parent process: */
826 struct task_struct __rcu *real_parent;
828 /* Recipient of SIGCHLD, wait4() reports: */
829 struct task_struct __rcu *parent;
832 * Children/sibling form the list of natural children:
834 struct list_head children;
835 struct list_head sibling;
836 struct task_struct *group_leader;
839 * 'ptraced' is the list of tasks this task is using ptrace() on.
841 * This includes both natural children and PTRACE_ATTACH targets.
842 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
844 struct list_head ptraced;
845 struct list_head ptrace_entry;
847 /* PID/PID hash table linkage. */
848 struct pid *thread_pid;
849 struct hlist_node pid_links[PIDTYPE_MAX];
850 struct list_head thread_group;
851 struct list_head thread_node;
853 struct completion *vfork_done;
855 /* CLONE_CHILD_SETTID: */
856 int __user *set_child_tid;
858 /* CLONE_CHILD_CLEARTID: */
859 int __user *clear_child_tid;
863 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
868 struct prev_cputime prev_cputime;
869 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
873 #ifdef CONFIG_NO_HZ_FULL
874 atomic_t tick_dep_mask;
876 /* Context switch counts: */
878 unsigned long nivcsw;
880 /* Monotonic time in nsecs: */
883 /* Boot based time in nsecs: */
886 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
887 unsigned long min_flt;
888 unsigned long maj_flt;
890 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
891 struct posix_cputimers posix_cputimers;
893 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
894 struct posix_cputimers_work posix_cputimers_work;
897 /* Process credentials: */
899 /* Tracer's credentials at attach: */
900 const struct cred __rcu *ptracer_cred;
902 /* Objective and real subjective task credentials (COW): */
903 const struct cred __rcu *real_cred;
905 /* Effective (overridable) subjective task credentials (COW): */
906 const struct cred __rcu *cred;
909 /* Cached requested key. */
910 struct key *cached_requested_key;
914 * executable name, excluding path.
916 * - normally initialized setup_new_exec()
917 * - access it with [gs]et_task_comm()
918 * - lock it with task_lock()
920 char comm[TASK_COMM_LEN];
922 struct nameidata *nameidata;
924 #ifdef CONFIG_SYSVIPC
925 struct sysv_sem sysvsem;
926 struct sysv_shm sysvshm;
928 #ifdef CONFIG_DETECT_HUNG_TASK
929 unsigned long last_switch_count;
930 unsigned long last_switch_time;
932 /* Filesystem information: */
933 struct fs_struct *fs;
935 /* Open file information: */
936 struct files_struct *files;
939 struct nsproxy *nsproxy;
941 /* Signal handlers: */
942 struct signal_struct *signal;
943 struct sighand_struct __rcu *sighand;
945 sigset_t real_blocked;
946 /* Restored if set_restore_sigmask() was used: */
947 sigset_t saved_sigmask;
948 struct sigpending pending;
949 unsigned long sas_ss_sp;
951 unsigned int sas_ss_flags;
953 struct callback_head *task_works;
956 #ifdef CONFIG_AUDITSYSCALL
957 struct audit_context *audit_context;
960 unsigned int sessionid;
962 struct seccomp seccomp;
964 /* Thread group tracking: */
968 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
969 spinlock_t alloc_lock;
971 /* Protection of the PI data structures: */
972 raw_spinlock_t pi_lock;
974 struct wake_q_node wake_q;
976 #ifdef CONFIG_RT_MUTEXES
977 /* PI waiters blocked on a rt_mutex held by this task: */
978 struct rb_root_cached pi_waiters;
979 /* Updated under owner's pi_lock and rq lock */
980 struct task_struct *pi_top_task;
981 /* Deadlock detection and priority inheritance handling: */
982 struct rt_mutex_waiter *pi_blocked_on;
985 #ifdef CONFIG_DEBUG_MUTEXES
986 /* Mutex deadlock detection: */
987 struct mutex_waiter *blocked_on;
990 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
994 #ifdef CONFIG_TRACE_IRQFLAGS
995 struct irqtrace_events irqtrace;
996 unsigned int hardirq_threaded;
997 u64 hardirq_chain_key;
998 int softirqs_enabled;
1003 #ifdef CONFIG_LOCKDEP
1004 # define MAX_LOCK_DEPTH 48UL
1007 unsigned int lockdep_recursion;
1008 struct held_lock held_locks[MAX_LOCK_DEPTH];
1012 unsigned int in_ubsan;
1015 /* Journalling filesystem info: */
1018 /* Stacked block device info: */
1019 struct bio_list *bio_list;
1022 /* Stack plugging: */
1023 struct blk_plug *plug;
1027 struct reclaim_state *reclaim_state;
1029 struct backing_dev_info *backing_dev_info;
1031 struct io_context *io_context;
1033 #ifdef CONFIG_COMPACTION
1034 struct capture_control *capture_control;
1037 unsigned long ptrace_message;
1038 kernel_siginfo_t *last_siginfo;
1040 struct task_io_accounting ioac;
1042 /* Pressure stall state */
1043 unsigned int psi_flags;
1045 #ifdef CONFIG_TASK_XACCT
1046 /* Accumulated RSS usage: */
1048 /* Accumulated virtual memory usage: */
1050 /* stime + utime since last update: */
1053 #ifdef CONFIG_CPUSETS
1054 /* Protected by ->alloc_lock: */
1055 nodemask_t mems_allowed;
1056 /* Seqence number to catch updates: */
1057 seqcount_spinlock_t mems_allowed_seq;
1058 int cpuset_mem_spread_rotor;
1059 int cpuset_slab_spread_rotor;
1061 #ifdef CONFIG_CGROUPS
1062 /* Control Group info protected by css_set_lock: */
1063 struct css_set __rcu *cgroups;
1064 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1065 struct list_head cg_list;
1067 #ifdef CONFIG_X86_CPU_RESCTRL
1072 struct robust_list_head __user *robust_list;
1073 #ifdef CONFIG_COMPAT
1074 struct compat_robust_list_head __user *compat_robust_list;
1076 struct list_head pi_state_list;
1077 struct futex_pi_state *pi_state_cache;
1078 struct mutex futex_exit_mutex;
1079 unsigned int futex_state;
1081 #ifdef CONFIG_PERF_EVENTS
1082 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1083 struct mutex perf_event_mutex;
1084 struct list_head perf_event_list;
1086 #ifdef CONFIG_DEBUG_PREEMPT
1087 unsigned long preempt_disable_ip;
1090 /* Protected by alloc_lock: */
1091 struct mempolicy *mempolicy;
1093 short pref_node_fork;
1095 #ifdef CONFIG_NUMA_BALANCING
1097 unsigned int numa_scan_period;
1098 unsigned int numa_scan_period_max;
1099 int numa_preferred_nid;
1100 unsigned long numa_migrate_retry;
1101 /* Migration stamp: */
1103 u64 last_task_numa_placement;
1104 u64 last_sum_exec_runtime;
1105 struct callback_head numa_work;
1108 * This pointer is only modified for current in syscall and
1109 * pagefault context (and for tasks being destroyed), so it can be read
1110 * from any of the following contexts:
1111 * - RCU read-side critical section
1112 * - current->numa_group from everywhere
1113 * - task's runqueue locked, task not running
1115 struct numa_group __rcu *numa_group;
1118 * numa_faults is an array split into four regions:
1119 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1120 * in this precise order.
1122 * faults_memory: Exponential decaying average of faults on a per-node
1123 * basis. Scheduling placement decisions are made based on these
1124 * counts. The values remain static for the duration of a PTE scan.
1125 * faults_cpu: Track the nodes the process was running on when a NUMA
1126 * hinting fault was incurred.
1127 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1128 * during the current scan window. When the scan completes, the counts
1129 * in faults_memory and faults_cpu decay and these values are copied.
1131 unsigned long *numa_faults;
1132 unsigned long total_numa_faults;
1135 * numa_faults_locality tracks if faults recorded during the last
1136 * scan window were remote/local or failed to migrate. The task scan
1137 * period is adapted based on the locality of the faults with different
1138 * weights depending on whether they were shared or private faults
1140 unsigned long numa_faults_locality[3];
1142 unsigned long numa_pages_migrated;
1143 #endif /* CONFIG_NUMA_BALANCING */
1146 struct rseq __user *rseq;
1149 * RmW on rseq_event_mask must be performed atomically
1150 * with respect to preemption.
1152 unsigned long rseq_event_mask;
1155 struct tlbflush_unmap_batch tlb_ubc;
1158 refcount_t rcu_users;
1159 struct rcu_head rcu;
1162 /* Cache last used pipe for splice(): */
1163 struct pipe_inode_info *splice_pipe;
1165 struct page_frag task_frag;
1167 #ifdef CONFIG_TASK_DELAY_ACCT
1168 struct task_delay_info *delays;
1171 #ifdef CONFIG_FAULT_INJECTION
1173 unsigned int fail_nth;
1176 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1177 * balance_dirty_pages() for a dirty throttling pause:
1180 int nr_dirtied_pause;
1181 /* Start of a write-and-pause period: */
1182 unsigned long dirty_paused_when;
1184 #ifdef CONFIG_LATENCYTOP
1185 int latency_record_count;
1186 struct latency_record latency_record[LT_SAVECOUNT];
1189 * Time slack values; these are used to round up poll() and
1190 * select() etc timeout values. These are in nanoseconds.
1193 u64 default_timer_slack_ns;
1196 unsigned int kasan_depth;
1200 struct kcsan_ctx kcsan_ctx;
1201 #ifdef CONFIG_TRACE_IRQFLAGS
1202 struct irqtrace_events kcsan_save_irqtrace;
1206 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1207 /* Index of current stored address in ret_stack: */
1211 /* Stack of return addresses for return function tracing: */
1212 struct ftrace_ret_stack *ret_stack;
1214 /* Timestamp for last schedule: */
1215 unsigned long long ftrace_timestamp;
1218 * Number of functions that haven't been traced
1219 * because of depth overrun:
1221 atomic_t trace_overrun;
1223 /* Pause tracing: */
1224 atomic_t tracing_graph_pause;
1227 #ifdef CONFIG_TRACING
1228 /* State flags for use by tracers: */
1229 unsigned long trace;
1231 /* Bitmask and counter of trace recursion: */
1232 unsigned long trace_recursion;
1233 #endif /* CONFIG_TRACING */
1236 /* See kernel/kcov.c for more details. */
1238 /* Coverage collection mode enabled for this task (0 if disabled): */
1239 unsigned int kcov_mode;
1241 /* Size of the kcov_area: */
1242 unsigned int kcov_size;
1244 /* Buffer for coverage collection: */
1247 /* KCOV descriptor wired with this task or NULL: */
1250 /* KCOV common handle for remote coverage collection: */
1253 /* KCOV sequence number: */
1256 /* Collect coverage from softirq context: */
1257 unsigned int kcov_softirq;
1261 struct mem_cgroup *memcg_in_oom;
1262 gfp_t memcg_oom_gfp_mask;
1263 int memcg_oom_order;
1265 /* Number of pages to reclaim on returning to userland: */
1266 unsigned int memcg_nr_pages_over_high;
1268 /* Used by memcontrol for targeted memcg charge: */
1269 struct mem_cgroup *active_memcg;
1272 #ifdef CONFIG_BLK_CGROUP
1273 struct request_queue *throttle_queue;
1276 #ifdef CONFIG_UPROBES
1277 struct uprobe_task *utask;
1279 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1280 unsigned int sequential_io;
1281 unsigned int sequential_io_avg;
1283 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1284 unsigned long task_state_change;
1286 int pagefault_disabled;
1288 struct task_struct *oom_reaper_list;
1290 #ifdef CONFIG_VMAP_STACK
1291 struct vm_struct *stack_vm_area;
1293 #ifdef CONFIG_THREAD_INFO_IN_TASK
1294 /* A live task holds one reference: */
1295 refcount_t stack_refcount;
1297 #ifdef CONFIG_LIVEPATCH
1300 #ifdef CONFIG_SECURITY
1301 /* Used by LSM modules for access restriction: */
1305 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1306 unsigned long lowest_stack;
1307 unsigned long prev_lowest_stack;
1310 #ifdef CONFIG_X86_MCE
1314 __mce_reserved : 62;
1315 struct callback_head mce_kill_me;
1319 * New fields for task_struct should be added above here, so that
1320 * they are included in the randomized portion of task_struct.
1322 randomized_struct_fields_end
1324 /* CPU-specific state of this task: */
1325 struct thread_struct thread;
1328 * WARNING: on x86, 'thread_struct' contains a variable-sized
1329 * structure. It *MUST* be at the end of 'task_struct'.
1331 * Do not put anything below here!
1335 static inline struct pid *task_pid(struct task_struct *task)
1337 return task->thread_pid;
1341 * the helpers to get the task's different pids as they are seen
1342 * from various namespaces
1344 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1345 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1347 * task_xid_nr_ns() : id seen from the ns specified;
1349 * see also pid_nr() etc in include/linux/pid.h
1351 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1353 static inline pid_t task_pid_nr(struct task_struct *tsk)
1358 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1360 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1363 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1365 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1369 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1375 * pid_alive - check that a task structure is not stale
1376 * @p: Task structure to be checked.
1378 * Test if a process is not yet dead (at most zombie state)
1379 * If pid_alive fails, then pointers within the task structure
1380 * can be stale and must not be dereferenced.
1382 * Return: 1 if the process is alive. 0 otherwise.
1384 static inline int pid_alive(const struct task_struct *p)
1386 return p->thread_pid != NULL;
1389 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1391 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1394 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1396 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1400 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1402 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1405 static inline pid_t task_session_vnr(struct task_struct *tsk)
1407 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1410 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1412 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1415 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1417 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1420 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1426 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1432 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1434 return task_ppid_nr_ns(tsk, &init_pid_ns);
1437 /* Obsolete, do not use: */
1438 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1440 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1443 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1444 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1446 static inline unsigned int task_state_index(struct task_struct *tsk)
1448 unsigned int tsk_state = READ_ONCE(tsk->state);
1449 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1451 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1453 if (tsk_state == TASK_IDLE)
1454 state = TASK_REPORT_IDLE;
1459 static inline char task_index_to_char(unsigned int state)
1461 static const char state_char[] = "RSDTtXZPI";
1463 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1465 return state_char[state];
1468 static inline char task_state_to_char(struct task_struct *tsk)
1470 return task_index_to_char(task_state_index(tsk));
1474 * is_global_init - check if a task structure is init. Since init
1475 * is free to have sub-threads we need to check tgid.
1476 * @tsk: Task structure to be checked.
1478 * Check if a task structure is the first user space task the kernel created.
1480 * Return: 1 if the task structure is init. 0 otherwise.
1482 static inline int is_global_init(struct task_struct *tsk)
1484 return task_tgid_nr(tsk) == 1;
1487 extern struct pid *cad_pid;
1492 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1493 #define PF_EXITING 0x00000004 /* Getting shut down */
1494 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1495 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1496 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1497 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1498 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1499 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1500 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1501 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1502 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1503 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1504 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1505 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1506 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1507 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1508 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1509 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1510 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1511 * I am cleaning dirty pages from some other bdi. */
1512 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1513 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1514 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1515 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1516 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1517 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1518 #define PF_IO_WORKER 0x20000000 /* Task is an IO worker */
1519 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1520 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1523 * Only the _current_ task can read/write to tsk->flags, but other
1524 * tasks can access tsk->flags in readonly mode for example
1525 * with tsk_used_math (like during threaded core dumping).
1526 * There is however an exception to this rule during ptrace
1527 * or during fork: the ptracer task is allowed to write to the
1528 * child->flags of its traced child (same goes for fork, the parent
1529 * can write to the child->flags), because we're guaranteed the
1530 * child is not running and in turn not changing child->flags
1531 * at the same time the parent does it.
1533 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1534 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1535 #define clear_used_math() clear_stopped_child_used_math(current)
1536 #define set_used_math() set_stopped_child_used_math(current)
1538 #define conditional_stopped_child_used_math(condition, child) \
1539 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1541 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1543 #define copy_to_stopped_child_used_math(child) \
1544 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1546 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1547 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1548 #define used_math() tsk_used_math(current)
1550 static inline bool is_percpu_thread(void)
1553 return (current->flags & PF_NO_SETAFFINITY) &&
1554 (current->nr_cpus_allowed == 1);
1560 /* Per-process atomic flags. */
1561 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1562 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1563 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1564 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1565 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1566 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1567 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1568 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1570 #define TASK_PFA_TEST(name, func) \
1571 static inline bool task_##func(struct task_struct *p) \
1572 { return test_bit(PFA_##name, &p->atomic_flags); }
1574 #define TASK_PFA_SET(name, func) \
1575 static inline void task_set_##func(struct task_struct *p) \
1576 { set_bit(PFA_##name, &p->atomic_flags); }
1578 #define TASK_PFA_CLEAR(name, func) \
1579 static inline void task_clear_##func(struct task_struct *p) \
1580 { clear_bit(PFA_##name, &p->atomic_flags); }
1582 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1583 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1585 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1586 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1587 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1589 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1590 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1591 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1593 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1594 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1595 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1597 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1598 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1599 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1601 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1602 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1604 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1605 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1606 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1608 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1609 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1612 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1614 current->flags &= ~flags;
1615 current->flags |= orig_flags & flags;
1618 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1619 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1621 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1622 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1624 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1627 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1629 if (!cpumask_test_cpu(0, new_mask))
1635 extern int yield_to(struct task_struct *p, bool preempt);
1636 extern void set_user_nice(struct task_struct *p, long nice);
1637 extern int task_prio(const struct task_struct *p);
1640 * task_nice - return the nice value of a given task.
1641 * @p: the task in question.
1643 * Return: The nice value [ -20 ... 0 ... 19 ].
1645 static inline int task_nice(const struct task_struct *p)
1647 return PRIO_TO_NICE((p)->static_prio);
1650 extern int can_nice(const struct task_struct *p, const int nice);
1651 extern int task_curr(const struct task_struct *p);
1652 extern int idle_cpu(int cpu);
1653 extern int available_idle_cpu(int cpu);
1654 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1655 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1656 extern void sched_set_fifo(struct task_struct *p);
1657 extern void sched_set_fifo_low(struct task_struct *p);
1658 extern void sched_set_normal(struct task_struct *p, int nice);
1659 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1660 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1661 extern struct task_struct *idle_task(int cpu);
1664 * is_idle_task - is the specified task an idle task?
1665 * @p: the task in question.
1667 * Return: 1 if @p is an idle task. 0 otherwise.
1669 static __always_inline bool is_idle_task(const struct task_struct *p)
1671 return !!(p->flags & PF_IDLE);
1674 extern struct task_struct *curr_task(int cpu);
1675 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1679 union thread_union {
1680 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1681 struct task_struct task;
1683 #ifndef CONFIG_THREAD_INFO_IN_TASK
1684 struct thread_info thread_info;
1686 unsigned long stack[THREAD_SIZE/sizeof(long)];
1689 #ifndef CONFIG_THREAD_INFO_IN_TASK
1690 extern struct thread_info init_thread_info;
1693 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1695 #ifdef CONFIG_THREAD_INFO_IN_TASK
1696 static inline struct thread_info *task_thread_info(struct task_struct *task)
1698 return &task->thread_info;
1700 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1701 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1705 * find a task by one of its numerical ids
1707 * find_task_by_pid_ns():
1708 * finds a task by its pid in the specified namespace
1709 * find_task_by_vpid():
1710 * finds a task by its virtual pid
1712 * see also find_vpid() etc in include/linux/pid.h
1715 extern struct task_struct *find_task_by_vpid(pid_t nr);
1716 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1719 * find a task by its virtual pid and get the task struct
1721 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1723 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1724 extern int wake_up_process(struct task_struct *tsk);
1725 extern void wake_up_new_task(struct task_struct *tsk);
1728 extern void kick_process(struct task_struct *tsk);
1730 static inline void kick_process(struct task_struct *tsk) { }
1733 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1735 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1737 __set_task_comm(tsk, from, false);
1740 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1741 #define get_task_comm(buf, tsk) ({ \
1742 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1743 __get_task_comm(buf, sizeof(buf), tsk); \
1747 static __always_inline void scheduler_ipi(void)
1750 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1751 * TIF_NEED_RESCHED remotely (for the first time) will also send
1754 preempt_fold_need_resched();
1756 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1758 static inline void scheduler_ipi(void) { }
1759 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1766 * Set thread flags in other task's structures.
1767 * See asm/thread_info.h for TIF_xxxx flags available:
1769 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1771 set_ti_thread_flag(task_thread_info(tsk), flag);
1774 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1776 clear_ti_thread_flag(task_thread_info(tsk), flag);
1779 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1782 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1785 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1787 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1790 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1792 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1795 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1797 return test_ti_thread_flag(task_thread_info(tsk), flag);
1800 static inline void set_tsk_need_resched(struct task_struct *tsk)
1802 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1805 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1807 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1810 static inline int test_tsk_need_resched(struct task_struct *tsk)
1812 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1816 * cond_resched() and cond_resched_lock(): latency reduction via
1817 * explicit rescheduling in places that are safe. The return
1818 * value indicates whether a reschedule was done in fact.
1819 * cond_resched_lock() will drop the spinlock before scheduling,
1821 #ifndef CONFIG_PREEMPTION
1822 extern int _cond_resched(void);
1824 static inline int _cond_resched(void) { return 0; }
1827 #define cond_resched() ({ \
1828 ___might_sleep(__FILE__, __LINE__, 0); \
1832 extern int __cond_resched_lock(spinlock_t *lock);
1834 #define cond_resched_lock(lock) ({ \
1835 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1836 __cond_resched_lock(lock); \
1839 static inline void cond_resched_rcu(void)
1841 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1849 * Does a critical section need to be broken due to another
1850 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1851 * but a general need for low latency)
1853 static inline int spin_needbreak(spinlock_t *lock)
1855 #ifdef CONFIG_PREEMPTION
1856 return spin_is_contended(lock);
1862 static __always_inline bool need_resched(void)
1864 return unlikely(tif_need_resched());
1868 * Wrappers for p->thread_info->cpu access. No-op on UP.
1872 static inline unsigned int task_cpu(const struct task_struct *p)
1874 #ifdef CONFIG_THREAD_INFO_IN_TASK
1875 return READ_ONCE(p->cpu);
1877 return READ_ONCE(task_thread_info(p)->cpu);
1881 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1885 static inline unsigned int task_cpu(const struct task_struct *p)
1890 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1894 #endif /* CONFIG_SMP */
1897 * In order to reduce various lock holder preemption latencies provide an
1898 * interface to see if a vCPU is currently running or not.
1900 * This allows us to terminate optimistic spin loops and block, analogous to
1901 * the native optimistic spin heuristic of testing if the lock owner task is
1904 #ifndef vcpu_is_preempted
1905 static inline bool vcpu_is_preempted(int cpu)
1911 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1912 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1914 #ifndef TASK_SIZE_OF
1915 #define TASK_SIZE_OF(tsk) TASK_SIZE
1921 * Map the event mask on the user-space ABI enum rseq_cs_flags
1922 * for direct mask checks.
1924 enum rseq_event_mask_bits {
1925 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1926 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1927 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1930 enum rseq_event_mask {
1931 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1932 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1933 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1936 static inline void rseq_set_notify_resume(struct task_struct *t)
1939 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1942 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1944 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1945 struct pt_regs *regs)
1948 __rseq_handle_notify_resume(ksig, regs);
1951 static inline void rseq_signal_deliver(struct ksignal *ksig,
1952 struct pt_regs *regs)
1955 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1957 rseq_handle_notify_resume(ksig, regs);
1960 /* rseq_preempt() requires preemption to be disabled. */
1961 static inline void rseq_preempt(struct task_struct *t)
1963 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1964 rseq_set_notify_resume(t);
1967 /* rseq_migrate() requires preemption to be disabled. */
1968 static inline void rseq_migrate(struct task_struct *t)
1970 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1971 rseq_set_notify_resume(t);
1975 * If parent process has a registered restartable sequences area, the
1976 * child inherits. Unregister rseq for a clone with CLONE_VM set.
1978 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1980 if (clone_flags & CLONE_VM) {
1983 t->rseq_event_mask = 0;
1985 t->rseq = current->rseq;
1986 t->rseq_sig = current->rseq_sig;
1987 t->rseq_event_mask = current->rseq_event_mask;
1991 static inline void rseq_execve(struct task_struct *t)
1995 t->rseq_event_mask = 0;
2000 static inline void rseq_set_notify_resume(struct task_struct *t)
2003 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2004 struct pt_regs *regs)
2007 static inline void rseq_signal_deliver(struct ksignal *ksig,
2008 struct pt_regs *regs)
2011 static inline void rseq_preempt(struct task_struct *t)
2014 static inline void rseq_migrate(struct task_struct *t)
2017 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2020 static inline void rseq_execve(struct task_struct *t)
2026 #ifdef CONFIG_DEBUG_RSEQ
2028 void rseq_syscall(struct pt_regs *regs);
2032 static inline void rseq_syscall(struct pt_regs *regs)
2038 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2039 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2040 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2042 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2043 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2044 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2046 int sched_trace_rq_cpu(struct rq *rq);
2047 int sched_trace_rq_nr_running(struct rq *rq);
2049 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);