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/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/refcount.h>
25 #include <linux/resource.h>
26 #include <linux/latencytop.h>
27 #include <linux/sched/prio.h>
28 #include <linux/sched/types.h>
29 #include <linux/signal_types.h>
30 #include <linux/mm_types_task.h>
31 #include <linux/task_io_accounting.h>
32 #include <linux/posix-timers.h>
33 #include <linux/rseq.h>
35 /* task_struct member predeclarations (sorted alphabetically): */
37 struct backing_dev_info;
40 struct capture_control;
43 struct futex_pi_state;
48 struct perf_event_context;
50 struct pipe_inode_info;
53 struct robust_list_head;
59 struct sighand_struct;
61 struct task_delay_info;
65 * Task state bitmask. NOTE! These bits are also
66 * encoded in fs/proc/array.c: get_task_state().
68 * We have two separate sets of flags: task->state
69 * is about runnability, while task->exit_state are
70 * about the task exiting. Confusing, but this way
71 * modifying one set can't modify the other one by
75 /* Used in tsk->state: */
76 #define TASK_RUNNING 0x0000
77 #define TASK_INTERRUPTIBLE 0x0001
78 #define TASK_UNINTERRUPTIBLE 0x0002
79 #define __TASK_STOPPED 0x0004
80 #define __TASK_TRACED 0x0008
81 /* Used in tsk->exit_state: */
82 #define EXIT_DEAD 0x0010
83 #define EXIT_ZOMBIE 0x0020
84 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
85 /* Used in tsk->state again: */
86 #define TASK_PARKED 0x0040
87 #define TASK_DEAD 0x0080
88 #define TASK_WAKEKILL 0x0100
89 #define TASK_WAKING 0x0200
90 #define TASK_NOLOAD 0x0400
91 #define TASK_NEW 0x0800
92 #define TASK_STATE_MAX 0x1000
94 /* Convenience macros for the sake of set_current_state: */
95 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
96 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
97 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
99 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
101 /* Convenience macros for the sake of wake_up(): */
102 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
104 /* get_task_state(): */
105 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
106 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
107 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
110 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
112 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
114 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
116 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
117 (task->flags & PF_FROZEN) == 0 && \
118 (task->state & TASK_NOLOAD) == 0)
120 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
123 * Special states are those that do not use the normal wait-loop pattern. See
124 * the comment with set_special_state().
126 #define is_special_task_state(state) \
127 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
129 #define __set_current_state(state_value) \
131 WARN_ON_ONCE(is_special_task_state(state_value));\
132 current->task_state_change = _THIS_IP_; \
133 current->state = (state_value); \
136 #define set_current_state(state_value) \
138 WARN_ON_ONCE(is_special_task_state(state_value));\
139 current->task_state_change = _THIS_IP_; \
140 smp_store_mb(current->state, (state_value)); \
143 #define set_special_state(state_value) \
145 unsigned long flags; /* may shadow */ \
146 WARN_ON_ONCE(!is_special_task_state(state_value)); \
147 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
148 current->task_state_change = _THIS_IP_; \
149 current->state = (state_value); \
150 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
154 * set_current_state() includes a barrier so that the write of current->state
155 * is correctly serialised wrt the caller's subsequent test of whether to
159 * set_current_state(TASK_UNINTERRUPTIBLE);
165 * __set_current_state(TASK_RUNNING);
167 * If the caller does not need such serialisation (because, for instance, the
168 * condition test and condition change and wakeup are under the same lock) then
169 * use __set_current_state().
171 * The above is typically ordered against the wakeup, which does:
173 * need_sleep = false;
174 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
176 * where wake_up_state() executes a full memory barrier before accessing the
179 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
180 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
181 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
183 * However, with slightly different timing the wakeup TASK_RUNNING store can
184 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
185 * a problem either because that will result in one extra go around the loop
186 * and our @cond test will save the day.
188 * Also see the comments of try_to_wake_up().
190 #define __set_current_state(state_value) \
191 current->state = (state_value)
193 #define set_current_state(state_value) \
194 smp_store_mb(current->state, (state_value))
197 * set_special_state() should be used for those states when the blocking task
198 * can not use the regular condition based wait-loop. In that case we must
199 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
200 * will not collide with our state change.
202 #define set_special_state(state_value) \
204 unsigned long flags; /* may shadow */ \
205 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
206 current->state = (state_value); \
207 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
212 /* Task command name length: */
213 #define TASK_COMM_LEN 16
215 extern void scheduler_tick(void);
217 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
219 extern long schedule_timeout(long timeout);
220 extern long schedule_timeout_interruptible(long timeout);
221 extern long schedule_timeout_killable(long timeout);
222 extern long schedule_timeout_uninterruptible(long timeout);
223 extern long schedule_timeout_idle(long timeout);
224 asmlinkage void schedule(void);
225 extern void schedule_preempt_disabled(void);
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: */
252 /* Task runs in userspace in a CPU with VTIME active: */
254 /* Task runs in kernelspace in a CPU with VTIME active: */
260 unsigned long long starttime;
261 enum vtime_state state;
268 * Utilization clamp constraints.
269 * @UCLAMP_MIN: Minimum utilization
270 * @UCLAMP_MAX: Maximum utilization
271 * @UCLAMP_CNT: Utilization clamp constraints count
280 extern struct root_domain def_root_domain;
281 extern struct mutex sched_domains_mutex;
285 #ifdef CONFIG_SCHED_INFO
286 /* Cumulative counters: */
288 /* # of times we have run on this CPU: */
289 unsigned long pcount;
291 /* Time spent waiting on a runqueue: */
292 unsigned long long run_delay;
296 /* When did we last run on a CPU? */
297 unsigned long long last_arrival;
299 /* When were we last queued to run? */
300 unsigned long long last_queued;
302 #endif /* CONFIG_SCHED_INFO */
306 * Integer metrics need fixed point arithmetic, e.g., sched/fair
307 * has a few: load, load_avg, util_avg, freq, and capacity.
309 * We define a basic fixed point arithmetic range, and then formalize
310 * all these metrics based on that basic range.
312 # define SCHED_FIXEDPOINT_SHIFT 10
313 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
315 /* Increase resolution of cpu_capacity calculations */
316 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
317 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
320 unsigned long weight;
325 * struct util_est - Estimation utilization of FAIR tasks
326 * @enqueued: instantaneous estimated utilization of a task/cpu
327 * @ewma: the Exponential Weighted Moving Average (EWMA)
328 * utilization of a task
330 * Support data structure to track an Exponential Weighted Moving Average
331 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
332 * average each time a task completes an activation. Sample's weight is chosen
333 * so that the EWMA will be relatively insensitive to transient changes to the
336 * The enqueued attribute has a slightly different meaning for tasks and cpus:
337 * - task: the task's util_avg at last task dequeue time
338 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
339 * Thus, the util_est.enqueued of a task represents the contribution on the
340 * estimated utilization of the CPU where that task is currently enqueued.
342 * Only for tasks we track a moving average of the past instantaneous
343 * estimated utilization. This allows to absorb sporadic drops in utilization
344 * of an otherwise almost periodic task.
347 unsigned int enqueued;
349 #define UTIL_EST_WEIGHT_SHIFT 2
350 } __attribute__((__aligned__(sizeof(u64))));
353 * The load_avg/util_avg accumulates an infinite geometric series
354 * (see __update_load_avg() in kernel/sched/fair.c).
356 * [load_avg definition]
358 * load_avg = runnable% * scale_load_down(load)
360 * where runnable% is the time ratio that a sched_entity is runnable.
361 * For cfs_rq, it is the aggregated load_avg of all runnable and
362 * blocked sched_entities.
364 * [util_avg definition]
366 * util_avg = running% * SCHED_CAPACITY_SCALE
368 * where running% is the time ratio that a sched_entity is running on
369 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
370 * and blocked sched_entities.
372 * load_avg and util_avg don't direcly factor frequency scaling and CPU
373 * capacity scaling. The scaling is done through the rq_clock_pelt that
374 * is used for computing those signals (see update_rq_clock_pelt())
376 * N.B., the above ratios (runnable% and running%) themselves are in the
377 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
378 * to as large a range as necessary. This is for example reflected by
379 * util_avg's SCHED_CAPACITY_SCALE.
383 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
384 * with the highest load (=88761), always runnable on a single cfs_rq,
385 * and should not overflow as the number already hits PID_MAX_LIMIT.
387 * For all other cases (including 32-bit kernels), struct load_weight's
388 * weight will overflow first before we do, because:
390 * Max(load_avg) <= Max(load.weight)
392 * Then it is the load_weight's responsibility to consider overflow
396 u64 last_update_time;
398 u64 runnable_load_sum;
401 unsigned long load_avg;
402 unsigned long runnable_load_avg;
403 unsigned long util_avg;
404 struct util_est util_est;
405 } ____cacheline_aligned;
407 struct sched_statistics {
408 #ifdef CONFIG_SCHEDSTATS
418 s64 sum_sleep_runtime;
425 u64 nr_migrations_cold;
426 u64 nr_failed_migrations_affine;
427 u64 nr_failed_migrations_running;
428 u64 nr_failed_migrations_hot;
429 u64 nr_forced_migrations;
433 u64 nr_wakeups_migrate;
434 u64 nr_wakeups_local;
435 u64 nr_wakeups_remote;
436 u64 nr_wakeups_affine;
437 u64 nr_wakeups_affine_attempts;
438 u64 nr_wakeups_passive;
443 struct sched_entity {
444 /* For load-balancing: */
445 struct load_weight load;
446 unsigned long runnable_weight;
447 struct rb_node run_node;
448 struct list_head group_node;
452 u64 sum_exec_runtime;
454 u64 prev_sum_exec_runtime;
458 struct sched_statistics statistics;
460 #ifdef CONFIG_FAIR_GROUP_SCHED
462 struct sched_entity *parent;
463 /* rq on which this entity is (to be) queued: */
464 struct cfs_rq *cfs_rq;
465 /* rq "owned" by this entity/group: */
471 * Per entity load average tracking.
473 * Put into separate cache line so it does not
474 * collide with read-mostly values above.
476 struct sched_avg avg;
480 struct sched_rt_entity {
481 struct list_head run_list;
482 unsigned long timeout;
483 unsigned long watchdog_stamp;
484 unsigned int time_slice;
485 unsigned short on_rq;
486 unsigned short on_list;
488 struct sched_rt_entity *back;
489 #ifdef CONFIG_RT_GROUP_SCHED
490 struct sched_rt_entity *parent;
491 /* rq on which this entity is (to be) queued: */
493 /* rq "owned" by this entity/group: */
496 } __randomize_layout;
498 struct sched_dl_entity {
499 struct rb_node rb_node;
502 * Original scheduling parameters. Copied here from sched_attr
503 * during sched_setattr(), they will remain the same until
504 * the next sched_setattr().
506 u64 dl_runtime; /* Maximum runtime for each instance */
507 u64 dl_deadline; /* Relative deadline of each instance */
508 u64 dl_period; /* Separation of two instances (period) */
509 u64 dl_bw; /* dl_runtime / dl_period */
510 u64 dl_density; /* dl_runtime / dl_deadline */
513 * Actual scheduling parameters. Initialized with the values above,
514 * they are continuously updated during task execution. Note that
515 * the remaining runtime could be < 0 in case we are in overrun.
517 s64 runtime; /* Remaining runtime for this instance */
518 u64 deadline; /* Absolute deadline for this instance */
519 unsigned int flags; /* Specifying the scheduler behaviour */
524 * @dl_throttled tells if we exhausted the runtime. If so, the
525 * task has to wait for a replenishment to be performed at the
526 * next firing of dl_timer.
528 * @dl_boosted tells if we are boosted due to DI. If so we are
529 * outside bandwidth enforcement mechanism (but only until we
530 * exit the critical section);
532 * @dl_yielded tells if task gave up the CPU before consuming
533 * all its available runtime during the last job.
535 * @dl_non_contending tells if the task is inactive while still
536 * contributing to the active utilization. In other words, it
537 * indicates if the inactive timer has been armed and its handler
538 * has not been executed yet. This flag is useful to avoid race
539 * conditions between the inactive timer handler and the wakeup
542 * @dl_overrun tells if the task asked to be informed about runtime
545 unsigned int dl_throttled : 1;
546 unsigned int dl_boosted : 1;
547 unsigned int dl_yielded : 1;
548 unsigned int dl_non_contending : 1;
549 unsigned int dl_overrun : 1;
552 * Bandwidth enforcement timer. Each -deadline task has its
553 * own bandwidth to be enforced, thus we need one timer per task.
555 struct hrtimer dl_timer;
558 * Inactive timer, responsible for decreasing the active utilization
559 * at the "0-lag time". When a -deadline task blocks, it contributes
560 * to GRUB's active utilization until the "0-lag time", hence a
561 * timer is needed to decrease the active utilization at the correct
564 struct hrtimer inactive_timer;
567 #ifdef CONFIG_UCLAMP_TASK
568 /* Number of utilization clamp buckets (shorter alias) */
569 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
572 * Utilization clamp for a scheduling entity
573 * @value: clamp value "assigned" to a se
574 * @bucket_id: bucket index corresponding to the "assigned" value
575 * @active: the se is currently refcounted in a rq's bucket
576 * @user_defined: the requested clamp value comes from user-space
578 * The bucket_id is the index of the clamp bucket matching the clamp value
579 * which is pre-computed and stored to avoid expensive integer divisions from
582 * The active bit is set whenever a task has got an "effective" value assigned,
583 * which can be different from the clamp value "requested" from user-space.
584 * This allows to know a task is refcounted in the rq's bucket corresponding
585 * to the "effective" bucket_id.
587 * The user_defined bit is set whenever a task has got a task-specific clamp
588 * value requested from userspace, i.e. the system defaults apply to this task
589 * just as a restriction. This allows to relax default clamps when a less
590 * restrictive task-specific value has been requested, thus allowing to
591 * implement a "nice" semantic. For example, a task running with a 20%
592 * default boost can still drop its own boosting to 0%.
595 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
596 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
597 unsigned int active : 1;
598 unsigned int user_defined : 1;
600 #endif /* CONFIG_UCLAMP_TASK */
606 u8 exp_hint; /* Hint for performance. */
609 u32 s; /* Set of bits. */
612 enum perf_event_task_context {
613 perf_invalid_context = -1,
616 perf_nr_task_contexts,
620 struct wake_q_node *next;
624 #ifdef CONFIG_THREAD_INFO_IN_TASK
626 * For reasons of header soup (see current_thread_info()), this
627 * must be the first element of task_struct.
629 struct thread_info thread_info;
631 /* -1 unrunnable, 0 runnable, >0 stopped: */
635 * This begins the randomizable portion of task_struct. Only
636 * scheduling-critical items should be added above here.
638 randomized_struct_fields_start
642 /* Per task flags (PF_*), defined further below: */
647 struct llist_node wake_entry;
649 #ifdef CONFIG_THREAD_INFO_IN_TASK
653 unsigned int wakee_flips;
654 unsigned long wakee_flip_decay_ts;
655 struct task_struct *last_wakee;
658 * recent_used_cpu is initially set as the last CPU used by a task
659 * that wakes affine another task. Waker/wakee relationships can
660 * push tasks around a CPU where each wakeup moves to the next one.
661 * Tracking a recently used CPU allows a quick search for a recently
662 * used CPU that may be idle.
672 unsigned int rt_priority;
674 const struct sched_class *sched_class;
675 struct sched_entity se;
676 struct sched_rt_entity rt;
677 #ifdef CONFIG_CGROUP_SCHED
678 struct task_group *sched_task_group;
680 struct sched_dl_entity dl;
682 #ifdef CONFIG_UCLAMP_TASK
683 /* Clamp values requested for a scheduling entity */
684 struct uclamp_se uclamp_req[UCLAMP_CNT];
685 /* Effective clamp values used for a scheduling entity */
686 struct uclamp_se uclamp[UCLAMP_CNT];
689 #ifdef CONFIG_PREEMPT_NOTIFIERS
690 /* List of struct preempt_notifier: */
691 struct hlist_head preempt_notifiers;
694 #ifdef CONFIG_BLK_DEV_IO_TRACE
695 unsigned int btrace_seq;
700 const cpumask_t *cpus_ptr;
703 #ifdef CONFIG_PREEMPT_RCU
704 int rcu_read_lock_nesting;
705 union rcu_special rcu_read_unlock_special;
706 struct list_head rcu_node_entry;
707 struct rcu_node *rcu_blocked_node;
708 #endif /* #ifdef CONFIG_PREEMPT_RCU */
710 #ifdef CONFIG_TASKS_RCU
711 unsigned long rcu_tasks_nvcsw;
712 u8 rcu_tasks_holdout;
714 int rcu_tasks_idle_cpu;
715 struct list_head rcu_tasks_holdout_list;
716 #endif /* #ifdef CONFIG_TASKS_RCU */
718 struct sched_info sched_info;
720 struct list_head tasks;
722 struct plist_node pushable_tasks;
723 struct rb_node pushable_dl_tasks;
726 struct mm_struct *mm;
727 struct mm_struct *active_mm;
729 /* Per-thread vma caching: */
730 struct vmacache vmacache;
732 #ifdef SPLIT_RSS_COUNTING
733 struct task_rss_stat rss_stat;
738 /* The signal sent when the parent dies: */
740 /* JOBCTL_*, siglock protected: */
741 unsigned long jobctl;
743 /* Used for emulating ABI behavior of previous Linux versions: */
744 unsigned int personality;
746 /* Scheduler bits, serialized by scheduler locks: */
747 unsigned sched_reset_on_fork:1;
748 unsigned sched_contributes_to_load:1;
749 unsigned sched_migrated:1;
750 unsigned sched_remote_wakeup:1;
752 unsigned sched_psi_wake_requeue:1;
755 /* Force alignment to the next boundary: */
758 /* Unserialized, strictly 'current' */
760 /* Bit to tell LSMs we're in execve(): */
761 unsigned in_execve:1;
762 unsigned in_iowait:1;
763 #ifndef TIF_RESTORE_SIGMASK
764 unsigned restore_sigmask:1;
767 unsigned in_user_fault:1;
769 #ifdef CONFIG_COMPAT_BRK
770 unsigned brk_randomized:1;
772 #ifdef CONFIG_CGROUPS
773 /* disallow userland-initiated cgroup migration */
774 unsigned no_cgroup_migration:1;
775 /* task is frozen/stopped (used by the cgroup freezer) */
778 #ifdef CONFIG_BLK_CGROUP
779 /* to be used once the psi infrastructure lands upstream. */
780 unsigned use_memdelay:1;
783 unsigned long atomic_flags; /* Flags requiring atomic access. */
785 struct restart_block restart_block;
790 #ifdef CONFIG_STACKPROTECTOR
791 /* Canary value for the -fstack-protector GCC feature: */
792 unsigned long stack_canary;
795 * Pointers to the (original) parent process, youngest child, younger sibling,
796 * older sibling, respectively. (p->father can be replaced with
797 * p->real_parent->pid)
800 /* Real parent process: */
801 struct task_struct __rcu *real_parent;
803 /* Recipient of SIGCHLD, wait4() reports: */
804 struct task_struct __rcu *parent;
807 * Children/sibling form the list of natural children:
809 struct list_head children;
810 struct list_head sibling;
811 struct task_struct *group_leader;
814 * 'ptraced' is the list of tasks this task is using ptrace() on.
816 * This includes both natural children and PTRACE_ATTACH targets.
817 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
819 struct list_head ptraced;
820 struct list_head ptrace_entry;
822 /* PID/PID hash table linkage. */
823 struct pid *thread_pid;
824 struct hlist_node pid_links[PIDTYPE_MAX];
825 struct list_head thread_group;
826 struct list_head thread_node;
828 struct completion *vfork_done;
830 /* CLONE_CHILD_SETTID: */
831 int __user *set_child_tid;
833 /* CLONE_CHILD_CLEARTID: */
834 int __user *clear_child_tid;
838 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
843 struct prev_cputime prev_cputime;
844 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
848 #ifdef CONFIG_NO_HZ_FULL
849 atomic_t tick_dep_mask;
851 /* Context switch counts: */
853 unsigned long nivcsw;
855 /* Monotonic time in nsecs: */
858 /* Boot based time in nsecs: */
861 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
862 unsigned long min_flt;
863 unsigned long maj_flt;
865 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
866 struct posix_cputimers posix_cputimers;
868 /* Process credentials: */
870 /* Tracer's credentials at attach: */
871 const struct cred __rcu *ptracer_cred;
873 /* Objective and real subjective task credentials (COW): */
874 const struct cred __rcu *real_cred;
876 /* Effective (overridable) subjective task credentials (COW): */
877 const struct cred __rcu *cred;
880 /* Cached requested key. */
881 struct key *cached_requested_key;
885 * executable name, excluding path.
887 * - normally initialized setup_new_exec()
888 * - access it with [gs]et_task_comm()
889 * - lock it with task_lock()
891 char comm[TASK_COMM_LEN];
893 struct nameidata *nameidata;
895 #ifdef CONFIG_SYSVIPC
896 struct sysv_sem sysvsem;
897 struct sysv_shm sysvshm;
899 #ifdef CONFIG_DETECT_HUNG_TASK
900 unsigned long last_switch_count;
901 unsigned long last_switch_time;
903 /* Filesystem information: */
904 struct fs_struct *fs;
906 /* Open file information: */
907 struct files_struct *files;
910 struct nsproxy *nsproxy;
912 /* Signal handlers: */
913 struct signal_struct *signal;
914 struct sighand_struct *sighand;
916 sigset_t real_blocked;
917 /* Restored if set_restore_sigmask() was used: */
918 sigset_t saved_sigmask;
919 struct sigpending pending;
920 unsigned long sas_ss_sp;
922 unsigned int sas_ss_flags;
924 struct callback_head *task_works;
927 #ifdef CONFIG_AUDITSYSCALL
928 struct audit_context *audit_context;
931 unsigned int sessionid;
933 struct seccomp seccomp;
935 /* Thread group tracking: */
939 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
940 spinlock_t alloc_lock;
942 /* Protection of the PI data structures: */
943 raw_spinlock_t pi_lock;
945 struct wake_q_node wake_q;
947 #ifdef CONFIG_RT_MUTEXES
948 /* PI waiters blocked on a rt_mutex held by this task: */
949 struct rb_root_cached pi_waiters;
950 /* Updated under owner's pi_lock and rq lock */
951 struct task_struct *pi_top_task;
952 /* Deadlock detection and priority inheritance handling: */
953 struct rt_mutex_waiter *pi_blocked_on;
956 #ifdef CONFIG_DEBUG_MUTEXES
957 /* Mutex deadlock detection: */
958 struct mutex_waiter *blocked_on;
961 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
965 #ifdef CONFIG_TRACE_IRQFLAGS
966 unsigned int irq_events;
967 unsigned long hardirq_enable_ip;
968 unsigned long hardirq_disable_ip;
969 unsigned int hardirq_enable_event;
970 unsigned int hardirq_disable_event;
971 int hardirqs_enabled;
973 unsigned long softirq_disable_ip;
974 unsigned long softirq_enable_ip;
975 unsigned int softirq_disable_event;
976 unsigned int softirq_enable_event;
977 int softirqs_enabled;
981 #ifdef CONFIG_LOCKDEP
982 # define MAX_LOCK_DEPTH 48UL
985 unsigned int lockdep_recursion;
986 struct held_lock held_locks[MAX_LOCK_DEPTH];
990 unsigned int in_ubsan;
993 /* Journalling filesystem info: */
996 /* Stacked block device info: */
997 struct bio_list *bio_list;
1000 /* Stack plugging: */
1001 struct blk_plug *plug;
1005 struct reclaim_state *reclaim_state;
1007 struct backing_dev_info *backing_dev_info;
1009 struct io_context *io_context;
1011 #ifdef CONFIG_COMPACTION
1012 struct capture_control *capture_control;
1015 unsigned long ptrace_message;
1016 kernel_siginfo_t *last_siginfo;
1018 struct task_io_accounting ioac;
1020 /* Pressure stall state */
1021 unsigned int psi_flags;
1023 #ifdef CONFIG_TASK_XACCT
1024 /* Accumulated RSS usage: */
1026 /* Accumulated virtual memory usage: */
1028 /* stime + utime since last update: */
1031 #ifdef CONFIG_CPUSETS
1032 /* Protected by ->alloc_lock: */
1033 nodemask_t mems_allowed;
1034 /* Seqence number to catch updates: */
1035 seqcount_t mems_allowed_seq;
1036 int cpuset_mem_spread_rotor;
1037 int cpuset_slab_spread_rotor;
1039 #ifdef CONFIG_CGROUPS
1040 /* Control Group info protected by css_set_lock: */
1041 struct css_set __rcu *cgroups;
1042 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1043 struct list_head cg_list;
1045 #ifdef CONFIG_X86_CPU_RESCTRL
1050 struct robust_list_head __user *robust_list;
1051 #ifdef CONFIG_COMPAT
1052 struct compat_robust_list_head __user *compat_robust_list;
1054 struct list_head pi_state_list;
1055 struct futex_pi_state *pi_state_cache;
1057 #ifdef CONFIG_PERF_EVENTS
1058 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1059 struct mutex perf_event_mutex;
1060 struct list_head perf_event_list;
1062 #ifdef CONFIG_DEBUG_PREEMPT
1063 unsigned long preempt_disable_ip;
1066 /* Protected by alloc_lock: */
1067 struct mempolicy *mempolicy;
1069 short pref_node_fork;
1071 #ifdef CONFIG_NUMA_BALANCING
1073 unsigned int numa_scan_period;
1074 unsigned int numa_scan_period_max;
1075 int numa_preferred_nid;
1076 unsigned long numa_migrate_retry;
1077 /* Migration stamp: */
1079 u64 last_task_numa_placement;
1080 u64 last_sum_exec_runtime;
1081 struct callback_head numa_work;
1084 * This pointer is only modified for current in syscall and
1085 * pagefault context (and for tasks being destroyed), so it can be read
1086 * from any of the following contexts:
1087 * - RCU read-side critical section
1088 * - current->numa_group from everywhere
1089 * - task's runqueue locked, task not running
1091 struct numa_group __rcu *numa_group;
1094 * numa_faults is an array split into four regions:
1095 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1096 * in this precise order.
1098 * faults_memory: Exponential decaying average of faults on a per-node
1099 * basis. Scheduling placement decisions are made based on these
1100 * counts. The values remain static for the duration of a PTE scan.
1101 * faults_cpu: Track the nodes the process was running on when a NUMA
1102 * hinting fault was incurred.
1103 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1104 * during the current scan window. When the scan completes, the counts
1105 * in faults_memory and faults_cpu decay and these values are copied.
1107 unsigned long *numa_faults;
1108 unsigned long total_numa_faults;
1111 * numa_faults_locality tracks if faults recorded during the last
1112 * scan window were remote/local or failed to migrate. The task scan
1113 * period is adapted based on the locality of the faults with different
1114 * weights depending on whether they were shared or private faults
1116 unsigned long numa_faults_locality[3];
1118 unsigned long numa_pages_migrated;
1119 #endif /* CONFIG_NUMA_BALANCING */
1122 struct rseq __user *rseq;
1125 * RmW on rseq_event_mask must be performed atomically
1126 * with respect to preemption.
1128 unsigned long rseq_event_mask;
1131 struct tlbflush_unmap_batch tlb_ubc;
1134 refcount_t rcu_users;
1135 struct rcu_head rcu;
1138 /* Cache last used pipe for splice(): */
1139 struct pipe_inode_info *splice_pipe;
1141 struct page_frag task_frag;
1143 #ifdef CONFIG_TASK_DELAY_ACCT
1144 struct task_delay_info *delays;
1147 #ifdef CONFIG_FAULT_INJECTION
1149 unsigned int fail_nth;
1152 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1153 * balance_dirty_pages() for a dirty throttling pause:
1156 int nr_dirtied_pause;
1157 /* Start of a write-and-pause period: */
1158 unsigned long dirty_paused_when;
1160 #ifdef CONFIG_LATENCYTOP
1161 int latency_record_count;
1162 struct latency_record latency_record[LT_SAVECOUNT];
1165 * Time slack values; these are used to round up poll() and
1166 * select() etc timeout values. These are in nanoseconds.
1169 u64 default_timer_slack_ns;
1172 unsigned int kasan_depth;
1175 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1176 /* Index of current stored address in ret_stack: */
1180 /* Stack of return addresses for return function tracing: */
1181 struct ftrace_ret_stack *ret_stack;
1183 /* Timestamp for last schedule: */
1184 unsigned long long ftrace_timestamp;
1187 * Number of functions that haven't been traced
1188 * because of depth overrun:
1190 atomic_t trace_overrun;
1192 /* Pause tracing: */
1193 atomic_t tracing_graph_pause;
1196 #ifdef CONFIG_TRACING
1197 /* State flags for use by tracers: */
1198 unsigned long trace;
1200 /* Bitmask and counter of trace recursion: */
1201 unsigned long trace_recursion;
1202 #endif /* CONFIG_TRACING */
1205 /* Coverage collection mode enabled for this task (0 if disabled): */
1206 unsigned int kcov_mode;
1208 /* Size of the kcov_area: */
1209 unsigned int kcov_size;
1211 /* Buffer for coverage collection: */
1214 /* KCOV descriptor wired with this task or NULL: */
1219 struct mem_cgroup *memcg_in_oom;
1220 gfp_t memcg_oom_gfp_mask;
1221 int memcg_oom_order;
1223 /* Number of pages to reclaim on returning to userland: */
1224 unsigned int memcg_nr_pages_over_high;
1226 /* Used by memcontrol for targeted memcg charge: */
1227 struct mem_cgroup *active_memcg;
1230 #ifdef CONFIG_BLK_CGROUP
1231 struct request_queue *throttle_queue;
1234 #ifdef CONFIG_UPROBES
1235 struct uprobe_task *utask;
1237 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1238 unsigned int sequential_io;
1239 unsigned int sequential_io_avg;
1241 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1242 unsigned long task_state_change;
1244 int pagefault_disabled;
1246 struct task_struct *oom_reaper_list;
1248 #ifdef CONFIG_VMAP_STACK
1249 struct vm_struct *stack_vm_area;
1251 #ifdef CONFIG_THREAD_INFO_IN_TASK
1252 /* A live task holds one reference: */
1253 refcount_t stack_refcount;
1255 #ifdef CONFIG_LIVEPATCH
1258 #ifdef CONFIG_SECURITY
1259 /* Used by LSM modules for access restriction: */
1263 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1264 unsigned long lowest_stack;
1265 unsigned long prev_lowest_stack;
1269 * New fields for task_struct should be added above here, so that
1270 * they are included in the randomized portion of task_struct.
1272 randomized_struct_fields_end
1274 /* CPU-specific state of this task: */
1275 struct thread_struct thread;
1278 * WARNING: on x86, 'thread_struct' contains a variable-sized
1279 * structure. It *MUST* be at the end of 'task_struct'.
1281 * Do not put anything below here!
1285 static inline struct pid *task_pid(struct task_struct *task)
1287 return task->thread_pid;
1291 * the helpers to get the task's different pids as they are seen
1292 * from various namespaces
1294 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1295 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1297 * task_xid_nr_ns() : id seen from the ns specified;
1299 * see also pid_nr() etc in include/linux/pid.h
1301 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1303 static inline pid_t task_pid_nr(struct task_struct *tsk)
1308 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1310 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1313 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1315 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1319 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1325 * pid_alive - check that a task structure is not stale
1326 * @p: Task structure to be checked.
1328 * Test if a process is not yet dead (at most zombie state)
1329 * If pid_alive fails, then pointers within the task structure
1330 * can be stale and must not be dereferenced.
1332 * Return: 1 if the process is alive. 0 otherwise.
1334 static inline int pid_alive(const struct task_struct *p)
1336 return p->thread_pid != NULL;
1339 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1341 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1344 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1346 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1350 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1352 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1355 static inline pid_t task_session_vnr(struct task_struct *tsk)
1357 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1360 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1362 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1365 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1367 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1370 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1376 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1382 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1384 return task_ppid_nr_ns(tsk, &init_pid_ns);
1387 /* Obsolete, do not use: */
1388 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1390 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1393 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1394 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1396 static inline unsigned int task_state_index(struct task_struct *tsk)
1398 unsigned int tsk_state = READ_ONCE(tsk->state);
1399 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1401 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1403 if (tsk_state == TASK_IDLE)
1404 state = TASK_REPORT_IDLE;
1409 static inline char task_index_to_char(unsigned int state)
1411 static const char state_char[] = "RSDTtXZPI";
1413 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1415 return state_char[state];
1418 static inline char task_state_to_char(struct task_struct *tsk)
1420 return task_index_to_char(task_state_index(tsk));
1424 * is_global_init - check if a task structure is init. Since init
1425 * is free to have sub-threads we need to check tgid.
1426 * @tsk: Task structure to be checked.
1428 * Check if a task structure is the first user space task the kernel created.
1430 * Return: 1 if the task structure is init. 0 otherwise.
1432 static inline int is_global_init(struct task_struct *tsk)
1434 return task_tgid_nr(tsk) == 1;
1437 extern struct pid *cad_pid;
1442 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1443 #define PF_EXITING 0x00000004 /* Getting shut down */
1444 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1445 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1446 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1447 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1448 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1449 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1450 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1451 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1452 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1453 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1454 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1455 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1456 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1457 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1458 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1459 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1460 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1461 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1462 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1463 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1464 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1465 #define PF_MEMSTALL 0x01000000 /* Stalled due to lack of memory */
1466 #define PF_UMH 0x02000000 /* I'm an Usermodehelper process */
1467 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1468 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1469 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1470 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1471 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1474 * Only the _current_ task can read/write to tsk->flags, but other
1475 * tasks can access tsk->flags in readonly mode for example
1476 * with tsk_used_math (like during threaded core dumping).
1477 * There is however an exception to this rule during ptrace
1478 * or during fork: the ptracer task is allowed to write to the
1479 * child->flags of its traced child (same goes for fork, the parent
1480 * can write to the child->flags), because we're guaranteed the
1481 * child is not running and in turn not changing child->flags
1482 * at the same time the parent does it.
1484 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1485 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1486 #define clear_used_math() clear_stopped_child_used_math(current)
1487 #define set_used_math() set_stopped_child_used_math(current)
1489 #define conditional_stopped_child_used_math(condition, child) \
1490 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1492 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1494 #define copy_to_stopped_child_used_math(child) \
1495 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1497 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1498 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1499 #define used_math() tsk_used_math(current)
1501 static inline bool is_percpu_thread(void)
1504 return (current->flags & PF_NO_SETAFFINITY) &&
1505 (current->nr_cpus_allowed == 1);
1511 /* Per-process atomic flags. */
1512 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1513 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1514 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1515 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1516 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1517 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1518 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1519 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1521 #define TASK_PFA_TEST(name, func) \
1522 static inline bool task_##func(struct task_struct *p) \
1523 { return test_bit(PFA_##name, &p->atomic_flags); }
1525 #define TASK_PFA_SET(name, func) \
1526 static inline void task_set_##func(struct task_struct *p) \
1527 { set_bit(PFA_##name, &p->atomic_flags); }
1529 #define TASK_PFA_CLEAR(name, func) \
1530 static inline void task_clear_##func(struct task_struct *p) \
1531 { clear_bit(PFA_##name, &p->atomic_flags); }
1533 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1534 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1536 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1537 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1538 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1540 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1541 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1542 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1544 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1545 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1546 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1548 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1549 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1550 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1552 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1553 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1555 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1556 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1557 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1559 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1560 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1563 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1565 current->flags &= ~flags;
1566 current->flags |= orig_flags & flags;
1569 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1570 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1572 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1573 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1575 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1578 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1580 if (!cpumask_test_cpu(0, new_mask))
1586 extern int yield_to(struct task_struct *p, bool preempt);
1587 extern void set_user_nice(struct task_struct *p, long nice);
1588 extern int task_prio(const struct task_struct *p);
1591 * task_nice - return the nice value of a given task.
1592 * @p: the task in question.
1594 * Return: The nice value [ -20 ... 0 ... 19 ].
1596 static inline int task_nice(const struct task_struct *p)
1598 return PRIO_TO_NICE((p)->static_prio);
1601 extern int can_nice(const struct task_struct *p, const int nice);
1602 extern int task_curr(const struct task_struct *p);
1603 extern int idle_cpu(int cpu);
1604 extern int available_idle_cpu(int cpu);
1605 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1606 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1607 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1608 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1609 extern struct task_struct *idle_task(int cpu);
1612 * is_idle_task - is the specified task an idle task?
1613 * @p: the task in question.
1615 * Return: 1 if @p is an idle task. 0 otherwise.
1617 static inline bool is_idle_task(const struct task_struct *p)
1619 return !!(p->flags & PF_IDLE);
1622 extern struct task_struct *curr_task(int cpu);
1623 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1627 union thread_union {
1628 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1629 struct task_struct task;
1631 #ifndef CONFIG_THREAD_INFO_IN_TASK
1632 struct thread_info thread_info;
1634 unsigned long stack[THREAD_SIZE/sizeof(long)];
1637 #ifndef CONFIG_THREAD_INFO_IN_TASK
1638 extern struct thread_info init_thread_info;
1641 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1643 #ifdef CONFIG_THREAD_INFO_IN_TASK
1644 static inline struct thread_info *task_thread_info(struct task_struct *task)
1646 return &task->thread_info;
1648 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1649 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1653 * find a task by one of its numerical ids
1655 * find_task_by_pid_ns():
1656 * finds a task by its pid in the specified namespace
1657 * find_task_by_vpid():
1658 * finds a task by its virtual pid
1660 * see also find_vpid() etc in include/linux/pid.h
1663 extern struct task_struct *find_task_by_vpid(pid_t nr);
1664 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1667 * find a task by its virtual pid and get the task struct
1669 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1671 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1672 extern int wake_up_process(struct task_struct *tsk);
1673 extern void wake_up_new_task(struct task_struct *tsk);
1676 extern void kick_process(struct task_struct *tsk);
1678 static inline void kick_process(struct task_struct *tsk) { }
1681 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1683 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1685 __set_task_comm(tsk, from, false);
1688 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1689 #define get_task_comm(buf, tsk) ({ \
1690 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1691 __get_task_comm(buf, sizeof(buf), tsk); \
1695 void scheduler_ipi(void);
1696 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1698 static inline void scheduler_ipi(void) { }
1699 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1706 * Set thread flags in other task's structures.
1707 * See asm/thread_info.h for TIF_xxxx flags available:
1709 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1711 set_ti_thread_flag(task_thread_info(tsk), flag);
1714 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1716 clear_ti_thread_flag(task_thread_info(tsk), flag);
1719 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1722 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1725 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1727 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1730 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1732 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1735 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1737 return test_ti_thread_flag(task_thread_info(tsk), flag);
1740 static inline void set_tsk_need_resched(struct task_struct *tsk)
1742 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1745 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1747 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1750 static inline int test_tsk_need_resched(struct task_struct *tsk)
1752 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1756 * cond_resched() and cond_resched_lock(): latency reduction via
1757 * explicit rescheduling in places that are safe. The return
1758 * value indicates whether a reschedule was done in fact.
1759 * cond_resched_lock() will drop the spinlock before scheduling,
1761 #ifndef CONFIG_PREEMPTION
1762 extern int _cond_resched(void);
1764 static inline int _cond_resched(void) { return 0; }
1767 #define cond_resched() ({ \
1768 ___might_sleep(__FILE__, __LINE__, 0); \
1772 extern int __cond_resched_lock(spinlock_t *lock);
1774 #define cond_resched_lock(lock) ({ \
1775 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1776 __cond_resched_lock(lock); \
1779 static inline void cond_resched_rcu(void)
1781 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1789 * Does a critical section need to be broken due to another
1790 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1791 * but a general need for low latency)
1793 static inline int spin_needbreak(spinlock_t *lock)
1795 #ifdef CONFIG_PREEMPTION
1796 return spin_is_contended(lock);
1802 static __always_inline bool need_resched(void)
1804 return unlikely(tif_need_resched());
1808 * Wrappers for p->thread_info->cpu access. No-op on UP.
1812 static inline unsigned int task_cpu(const struct task_struct *p)
1814 #ifdef CONFIG_THREAD_INFO_IN_TASK
1815 return READ_ONCE(p->cpu);
1817 return READ_ONCE(task_thread_info(p)->cpu);
1821 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1825 static inline unsigned int task_cpu(const struct task_struct *p)
1830 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1834 #endif /* CONFIG_SMP */
1837 * In order to reduce various lock holder preemption latencies provide an
1838 * interface to see if a vCPU is currently running or not.
1840 * This allows us to terminate optimistic spin loops and block, analogous to
1841 * the native optimistic spin heuristic of testing if the lock owner task is
1844 #ifndef vcpu_is_preempted
1845 static inline bool vcpu_is_preempted(int cpu)
1851 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1852 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1854 #ifndef TASK_SIZE_OF
1855 #define TASK_SIZE_OF(tsk) TASK_SIZE
1861 * Map the event mask on the user-space ABI enum rseq_cs_flags
1862 * for direct mask checks.
1864 enum rseq_event_mask_bits {
1865 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1866 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1867 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1870 enum rseq_event_mask {
1871 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1872 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1873 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1876 static inline void rseq_set_notify_resume(struct task_struct *t)
1879 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1882 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1884 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1885 struct pt_regs *regs)
1888 __rseq_handle_notify_resume(ksig, regs);
1891 static inline void rseq_signal_deliver(struct ksignal *ksig,
1892 struct pt_regs *regs)
1895 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1897 rseq_handle_notify_resume(ksig, regs);
1900 /* rseq_preempt() requires preemption to be disabled. */
1901 static inline void rseq_preempt(struct task_struct *t)
1903 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1904 rseq_set_notify_resume(t);
1907 /* rseq_migrate() requires preemption to be disabled. */
1908 static inline void rseq_migrate(struct task_struct *t)
1910 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1911 rseq_set_notify_resume(t);
1915 * If parent process has a registered restartable sequences area, the
1916 * child inherits. Only applies when forking a process, not a thread.
1918 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1920 if (clone_flags & CLONE_THREAD) {
1923 t->rseq_event_mask = 0;
1925 t->rseq = current->rseq;
1926 t->rseq_sig = current->rseq_sig;
1927 t->rseq_event_mask = current->rseq_event_mask;
1931 static inline void rseq_execve(struct task_struct *t)
1935 t->rseq_event_mask = 0;
1940 static inline void rseq_set_notify_resume(struct task_struct *t)
1943 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1944 struct pt_regs *regs)
1947 static inline void rseq_signal_deliver(struct ksignal *ksig,
1948 struct pt_regs *regs)
1951 static inline void rseq_preempt(struct task_struct *t)
1954 static inline void rseq_migrate(struct task_struct *t)
1957 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1960 static inline void rseq_execve(struct task_struct *t)
1966 void __exit_umh(struct task_struct *tsk);
1968 static inline void exit_umh(struct task_struct *tsk)
1970 if (unlikely(tsk->flags & PF_UMH))
1974 #ifdef CONFIG_DEBUG_RSEQ
1976 void rseq_syscall(struct pt_regs *regs);
1980 static inline void rseq_syscall(struct pt_regs *regs)
1986 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
1987 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
1988 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
1990 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
1991 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
1992 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
1994 int sched_trace_rq_cpu(struct rq *rq);
1996 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);