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/kcsan.h>
37 /* task_struct member predeclarations (sorted alphabetically): */
39 struct backing_dev_info;
42 struct capture_control;
45 struct futex_pi_state;
50 struct perf_event_context;
52 struct pipe_inode_info;
55 struct robust_list_head;
61 struct sighand_struct;
63 struct task_delay_info;
67 * Task state bitmask. NOTE! These bits are also
68 * encoded in fs/proc/array.c: get_task_state().
70 * We have two separate sets of flags: task->state
71 * is about runnability, while task->exit_state are
72 * about the task exiting. Confusing, but this way
73 * modifying one set can't modify the other one by
77 /* Used in tsk->state: */
78 #define TASK_RUNNING 0x0000
79 #define TASK_INTERRUPTIBLE 0x0001
80 #define TASK_UNINTERRUPTIBLE 0x0002
81 #define __TASK_STOPPED 0x0004
82 #define __TASK_TRACED 0x0008
83 /* Used in tsk->exit_state: */
84 #define EXIT_DEAD 0x0010
85 #define EXIT_ZOMBIE 0x0020
86 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
87 /* Used in tsk->state again: */
88 #define TASK_PARKED 0x0040
89 #define TASK_DEAD 0x0080
90 #define TASK_WAKEKILL 0x0100
91 #define TASK_WAKING 0x0200
92 #define TASK_NOLOAD 0x0400
93 #define TASK_NEW 0x0800
94 #define TASK_STATE_MAX 0x1000
96 /* Convenience macros for the sake of set_current_state: */
97 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
98 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
99 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
101 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
103 /* Convenience macros for the sake of wake_up(): */
104 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
106 /* get_task_state(): */
107 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
108 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
109 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
112 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
114 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
116 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
118 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
121 * Special states are those that do not use the normal wait-loop pattern. See
122 * the comment with set_special_state().
124 #define is_special_task_state(state) \
125 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
127 #define __set_current_state(state_value) \
129 WARN_ON_ONCE(is_special_task_state(state_value));\
130 current->task_state_change = _THIS_IP_; \
131 current->state = (state_value); \
134 #define set_current_state(state_value) \
136 WARN_ON_ONCE(is_special_task_state(state_value));\
137 current->task_state_change = _THIS_IP_; \
138 smp_store_mb(current->state, (state_value)); \
141 #define set_special_state(state_value) \
143 unsigned long flags; /* may shadow */ \
144 WARN_ON_ONCE(!is_special_task_state(state_value)); \
145 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
146 current->task_state_change = _THIS_IP_; \
147 current->state = (state_value); \
148 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
152 * set_current_state() includes a barrier so that the write of current->state
153 * is correctly serialised wrt the caller's subsequent test of whether to
157 * set_current_state(TASK_UNINTERRUPTIBLE);
163 * __set_current_state(TASK_RUNNING);
165 * If the caller does not need such serialisation (because, for instance, the
166 * condition test and condition change and wakeup are under the same lock) then
167 * use __set_current_state().
169 * The above is typically ordered against the wakeup, which does:
171 * need_sleep = false;
172 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
174 * where wake_up_state() executes a full memory barrier before accessing the
177 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
178 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
179 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
181 * However, with slightly different timing the wakeup TASK_RUNNING store can
182 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
183 * a problem either because that will result in one extra go around the loop
184 * and our @cond test will save the day.
186 * Also see the comments of try_to_wake_up().
188 #define __set_current_state(state_value) \
189 current->state = (state_value)
191 #define set_current_state(state_value) \
192 smp_store_mb(current->state, (state_value))
195 * set_special_state() should be used for those states when the blocking task
196 * can not use the regular condition based wait-loop. In that case we must
197 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
198 * will not collide with our state change.
200 #define set_special_state(state_value) \
202 unsigned long flags; /* may shadow */ \
203 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
204 current->state = (state_value); \
205 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
210 /* Task command name length: */
211 #define TASK_COMM_LEN 16
213 extern void scheduler_tick(void);
215 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
217 extern long schedule_timeout(long timeout);
218 extern long schedule_timeout_interruptible(long timeout);
219 extern long schedule_timeout_killable(long timeout);
220 extern long schedule_timeout_uninterruptible(long timeout);
221 extern long schedule_timeout_idle(long timeout);
222 asmlinkage void schedule(void);
223 extern void schedule_preempt_disabled(void);
224 asmlinkage void preempt_schedule_irq(void);
226 extern int __must_check io_schedule_prepare(void);
227 extern void io_schedule_finish(int token);
228 extern long io_schedule_timeout(long timeout);
229 extern void io_schedule(void);
232 * struct prev_cputime - snapshot of system and user cputime
233 * @utime: time spent in user mode
234 * @stime: time spent in system mode
235 * @lock: protects the above two fields
237 * Stores previous user/system time values such that we can guarantee
240 struct prev_cputime {
241 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
249 /* Task is sleeping or running in a CPU with VTIME inactive: */
253 /* Task runs in kernelspace in a CPU with VTIME active: */
255 /* Task runs in userspace in a CPU with VTIME active: */
257 /* Task runs as guests in a CPU with VTIME active: */
263 unsigned long long starttime;
264 enum vtime_state state;
272 * Utilization clamp constraints.
273 * @UCLAMP_MIN: Minimum utilization
274 * @UCLAMP_MAX: Maximum utilization
275 * @UCLAMP_CNT: Utilization clamp constraints count
284 extern struct root_domain def_root_domain;
285 extern struct mutex sched_domains_mutex;
289 #ifdef CONFIG_SCHED_INFO
290 /* Cumulative counters: */
292 /* # of times we have run on this CPU: */
293 unsigned long pcount;
295 /* Time spent waiting on a runqueue: */
296 unsigned long long run_delay;
300 /* When did we last run on a CPU? */
301 unsigned long long last_arrival;
303 /* When were we last queued to run? */
304 unsigned long long last_queued;
306 #endif /* CONFIG_SCHED_INFO */
310 * Integer metrics need fixed point arithmetic, e.g., sched/fair
311 * has a few: load, load_avg, util_avg, freq, and capacity.
313 * We define a basic fixed point arithmetic range, and then formalize
314 * all these metrics based on that basic range.
316 # define SCHED_FIXEDPOINT_SHIFT 10
317 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
319 /* Increase resolution of cpu_capacity calculations */
320 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
321 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
324 unsigned long weight;
329 * struct util_est - Estimation utilization of FAIR tasks
330 * @enqueued: instantaneous estimated utilization of a task/cpu
331 * @ewma: the Exponential Weighted Moving Average (EWMA)
332 * utilization of a task
334 * Support data structure to track an Exponential Weighted Moving Average
335 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
336 * average each time a task completes an activation. Sample's weight is chosen
337 * so that the EWMA will be relatively insensitive to transient changes to the
340 * The enqueued attribute has a slightly different meaning for tasks and cpus:
341 * - task: the task's util_avg at last task dequeue time
342 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
343 * Thus, the util_est.enqueued of a task represents the contribution on the
344 * estimated utilization of the CPU where that task is currently enqueued.
346 * Only for tasks we track a moving average of the past instantaneous
347 * estimated utilization. This allows to absorb sporadic drops in utilization
348 * of an otherwise almost periodic task.
351 unsigned int enqueued;
353 #define UTIL_EST_WEIGHT_SHIFT 2
354 } __attribute__((__aligned__(sizeof(u64))));
357 * The load/runnable/util_avg accumulates an infinite geometric series
358 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
360 * [load_avg definition]
362 * load_avg = runnable% * scale_load_down(load)
364 * [runnable_avg definition]
366 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
368 * [util_avg definition]
370 * util_avg = running% * SCHED_CAPACITY_SCALE
372 * where runnable% is the time ratio that a sched_entity is runnable and
373 * running% the time ratio that a sched_entity is running.
375 * For cfs_rq, they are the aggregated values of all runnable and blocked
378 * The load/runnable/util_avg doesn't direcly factor frequency scaling and CPU
379 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
380 * for computing those signals (see update_rq_clock_pelt())
382 * N.B., the above ratios (runnable% and running%) themselves are in the
383 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
384 * to as large a range as necessary. This is for example reflected by
385 * util_avg's SCHED_CAPACITY_SCALE.
389 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
390 * with the highest load (=88761), always runnable on a single cfs_rq,
391 * and should not overflow as the number already hits PID_MAX_LIMIT.
393 * For all other cases (including 32-bit kernels), struct load_weight's
394 * weight will overflow first before we do, because:
396 * Max(load_avg) <= Max(load.weight)
398 * Then it is the load_weight's responsibility to consider overflow
402 u64 last_update_time;
407 unsigned long load_avg;
408 unsigned long runnable_avg;
409 unsigned long util_avg;
410 struct util_est util_est;
411 } ____cacheline_aligned;
413 struct sched_statistics {
414 #ifdef CONFIG_SCHEDSTATS
424 s64 sum_sleep_runtime;
431 u64 nr_migrations_cold;
432 u64 nr_failed_migrations_affine;
433 u64 nr_failed_migrations_running;
434 u64 nr_failed_migrations_hot;
435 u64 nr_forced_migrations;
439 u64 nr_wakeups_migrate;
440 u64 nr_wakeups_local;
441 u64 nr_wakeups_remote;
442 u64 nr_wakeups_affine;
443 u64 nr_wakeups_affine_attempts;
444 u64 nr_wakeups_passive;
449 struct sched_entity {
450 /* For load-balancing: */
451 struct load_weight load;
452 struct rb_node run_node;
453 struct list_head group_node;
457 u64 sum_exec_runtime;
459 u64 prev_sum_exec_runtime;
463 struct sched_statistics statistics;
465 #ifdef CONFIG_FAIR_GROUP_SCHED
467 struct sched_entity *parent;
468 /* rq on which this entity is (to be) queued: */
469 struct cfs_rq *cfs_rq;
470 /* rq "owned" by this entity/group: */
472 /* cached value of my_q->h_nr_running */
473 unsigned long runnable_weight;
478 * Per entity load average tracking.
480 * Put into separate cache line so it does not
481 * collide with read-mostly values above.
483 struct sched_avg avg;
487 struct sched_rt_entity {
488 struct list_head run_list;
489 unsigned long timeout;
490 unsigned long watchdog_stamp;
491 unsigned int time_slice;
492 unsigned short on_rq;
493 unsigned short on_list;
495 struct sched_rt_entity *back;
496 #ifdef CONFIG_RT_GROUP_SCHED
497 struct sched_rt_entity *parent;
498 /* rq on which this entity is (to be) queued: */
500 /* rq "owned" by this entity/group: */
503 } __randomize_layout;
505 struct sched_dl_entity {
506 struct rb_node rb_node;
509 * Original scheduling parameters. Copied here from sched_attr
510 * during sched_setattr(), they will remain the same until
511 * the next sched_setattr().
513 u64 dl_runtime; /* Maximum runtime for each instance */
514 u64 dl_deadline; /* Relative deadline of each instance */
515 u64 dl_period; /* Separation of two instances (period) */
516 u64 dl_bw; /* dl_runtime / dl_period */
517 u64 dl_density; /* dl_runtime / dl_deadline */
520 * Actual scheduling parameters. Initialized with the values above,
521 * they are continuously updated during task execution. Note that
522 * the remaining runtime could be < 0 in case we are in overrun.
524 s64 runtime; /* Remaining runtime for this instance */
525 u64 deadline; /* Absolute deadline for this instance */
526 unsigned int flags; /* Specifying the scheduler behaviour */
531 * @dl_throttled tells if we exhausted the runtime. If so, the
532 * task has to wait for a replenishment to be performed at the
533 * next firing of dl_timer.
535 * @dl_boosted tells if we are boosted due to DI. If so we are
536 * outside bandwidth enforcement mechanism (but only until we
537 * exit the critical section);
539 * @dl_yielded tells if task gave up the CPU before consuming
540 * all its available runtime during the last job.
542 * @dl_non_contending tells if the task is inactive while still
543 * contributing to the active utilization. In other words, it
544 * indicates if the inactive timer has been armed and its handler
545 * has not been executed yet. This flag is useful to avoid race
546 * conditions between the inactive timer handler and the wakeup
549 * @dl_overrun tells if the task asked to be informed about runtime
552 unsigned int dl_throttled : 1;
553 unsigned int dl_boosted : 1;
554 unsigned int dl_yielded : 1;
555 unsigned int dl_non_contending : 1;
556 unsigned int dl_overrun : 1;
559 * Bandwidth enforcement timer. Each -deadline task has its
560 * own bandwidth to be enforced, thus we need one timer per task.
562 struct hrtimer dl_timer;
565 * Inactive timer, responsible for decreasing the active utilization
566 * at the "0-lag time". When a -deadline task blocks, it contributes
567 * to GRUB's active utilization until the "0-lag time", hence a
568 * timer is needed to decrease the active utilization at the correct
571 struct hrtimer inactive_timer;
574 #ifdef CONFIG_UCLAMP_TASK
575 /* Number of utilization clamp buckets (shorter alias) */
576 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
579 * Utilization clamp for a scheduling entity
580 * @value: clamp value "assigned" to a se
581 * @bucket_id: bucket index corresponding to the "assigned" value
582 * @active: the se is currently refcounted in a rq's bucket
583 * @user_defined: the requested clamp value comes from user-space
585 * The bucket_id is the index of the clamp bucket matching the clamp value
586 * which is pre-computed and stored to avoid expensive integer divisions from
589 * The active bit is set whenever a task has got an "effective" value assigned,
590 * which can be different from the clamp value "requested" from user-space.
591 * This allows to know a task is refcounted in the rq's bucket corresponding
592 * to the "effective" bucket_id.
594 * The user_defined bit is set whenever a task has got a task-specific clamp
595 * value requested from userspace, i.e. the system defaults apply to this task
596 * just as a restriction. This allows to relax default clamps when a less
597 * restrictive task-specific value has been requested, thus allowing to
598 * implement a "nice" semantic. For example, a task running with a 20%
599 * default boost can still drop its own boosting to 0%.
602 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
603 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
604 unsigned int active : 1;
605 unsigned int user_defined : 1;
607 #endif /* CONFIG_UCLAMP_TASK */
613 u8 exp_hint; /* Hint for performance. */
614 u8 need_mb; /* Readers need smp_mb(). */
616 u32 s; /* Set of bits. */
619 enum perf_event_task_context {
620 perf_invalid_context = -1,
623 perf_nr_task_contexts,
627 struct wake_q_node *next;
631 #ifdef CONFIG_THREAD_INFO_IN_TASK
633 * For reasons of header soup (see current_thread_info()), this
634 * must be the first element of task_struct.
636 struct thread_info thread_info;
638 /* -1 unrunnable, 0 runnable, >0 stopped: */
642 * This begins the randomizable portion of task_struct. Only
643 * scheduling-critical items should be added above here.
645 randomized_struct_fields_start
649 /* Per task flags (PF_*), defined further below: */
655 struct __call_single_node wake_entry;
656 #ifdef CONFIG_THREAD_INFO_IN_TASK
660 unsigned int wakee_flips;
661 unsigned long wakee_flip_decay_ts;
662 struct task_struct *last_wakee;
665 * recent_used_cpu is initially set as the last CPU used by a task
666 * that wakes affine another task. Waker/wakee relationships can
667 * push tasks around a CPU where each wakeup moves to the next one.
668 * Tracking a recently used CPU allows a quick search for a recently
669 * used CPU that may be idle.
679 unsigned int rt_priority;
681 const struct sched_class *sched_class;
682 struct sched_entity se;
683 struct sched_rt_entity rt;
684 #ifdef CONFIG_CGROUP_SCHED
685 struct task_group *sched_task_group;
687 struct sched_dl_entity dl;
689 #ifdef CONFIG_UCLAMP_TASK
690 /* Clamp values requested for a scheduling entity */
691 struct uclamp_se uclamp_req[UCLAMP_CNT];
692 /* Effective clamp values used for a scheduling entity */
693 struct uclamp_se uclamp[UCLAMP_CNT];
696 #ifdef CONFIG_PREEMPT_NOTIFIERS
697 /* List of struct preempt_notifier: */
698 struct hlist_head preempt_notifiers;
701 #ifdef CONFIG_BLK_DEV_IO_TRACE
702 unsigned int btrace_seq;
707 const cpumask_t *cpus_ptr;
710 #ifdef CONFIG_PREEMPT_RCU
711 int rcu_read_lock_nesting;
712 union rcu_special rcu_read_unlock_special;
713 struct list_head rcu_node_entry;
714 struct rcu_node *rcu_blocked_node;
715 #endif /* #ifdef CONFIG_PREEMPT_RCU */
717 #ifdef CONFIG_TASKS_RCU
718 unsigned long rcu_tasks_nvcsw;
719 u8 rcu_tasks_holdout;
721 int rcu_tasks_idle_cpu;
722 struct list_head rcu_tasks_holdout_list;
723 #endif /* #ifdef CONFIG_TASKS_RCU */
725 #ifdef CONFIG_TASKS_TRACE_RCU
726 int trc_reader_nesting;
728 union rcu_special trc_reader_special;
729 bool trc_reader_checked;
730 struct list_head trc_holdout_list;
731 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
733 struct sched_info sched_info;
735 struct list_head tasks;
737 struct plist_node pushable_tasks;
738 struct rb_node pushable_dl_tasks;
741 struct mm_struct *mm;
742 struct mm_struct *active_mm;
744 /* Per-thread vma caching: */
745 struct vmacache vmacache;
747 #ifdef SPLIT_RSS_COUNTING
748 struct task_rss_stat rss_stat;
753 /* The signal sent when the parent dies: */
755 /* JOBCTL_*, siglock protected: */
756 unsigned long jobctl;
758 /* Used for emulating ABI behavior of previous Linux versions: */
759 unsigned int personality;
761 /* Scheduler bits, serialized by scheduler locks: */
762 unsigned sched_reset_on_fork:1;
763 unsigned sched_contributes_to_load:1;
764 unsigned sched_migrated:1;
765 unsigned sched_remote_wakeup:1;
767 unsigned sched_psi_wake_requeue:1;
770 /* Force alignment to the next boundary: */
773 /* Unserialized, strictly 'current' */
775 /* Bit to tell LSMs we're in execve(): */
776 unsigned in_execve:1;
777 unsigned in_iowait:1;
778 #ifndef TIF_RESTORE_SIGMASK
779 unsigned restore_sigmask:1;
782 unsigned in_user_fault:1;
784 #ifdef CONFIG_COMPAT_BRK
785 unsigned brk_randomized:1;
787 #ifdef CONFIG_CGROUPS
788 /* disallow userland-initiated cgroup migration */
789 unsigned no_cgroup_migration:1;
790 /* task is frozen/stopped (used by the cgroup freezer) */
793 #ifdef CONFIG_BLK_CGROUP
794 unsigned use_memdelay:1;
797 /* Stalled due to lack of memory */
798 unsigned in_memstall:1;
801 unsigned long atomic_flags; /* Flags requiring atomic access. */
803 struct restart_block restart_block;
808 #ifdef CONFIG_STACKPROTECTOR
809 /* Canary value for the -fstack-protector GCC feature: */
810 unsigned long stack_canary;
813 * Pointers to the (original) parent process, youngest child, younger sibling,
814 * older sibling, respectively. (p->father can be replaced with
815 * p->real_parent->pid)
818 /* Real parent process: */
819 struct task_struct __rcu *real_parent;
821 /* Recipient of SIGCHLD, wait4() reports: */
822 struct task_struct __rcu *parent;
825 * Children/sibling form the list of natural children:
827 struct list_head children;
828 struct list_head sibling;
829 struct task_struct *group_leader;
832 * 'ptraced' is the list of tasks this task is using ptrace() on.
834 * This includes both natural children and PTRACE_ATTACH targets.
835 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
837 struct list_head ptraced;
838 struct list_head ptrace_entry;
840 /* PID/PID hash table linkage. */
841 struct pid *thread_pid;
842 struct hlist_node pid_links[PIDTYPE_MAX];
843 struct list_head thread_group;
844 struct list_head thread_node;
846 struct completion *vfork_done;
848 /* CLONE_CHILD_SETTID: */
849 int __user *set_child_tid;
851 /* CLONE_CHILD_CLEARTID: */
852 int __user *clear_child_tid;
856 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
861 struct prev_cputime prev_cputime;
862 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
866 #ifdef CONFIG_NO_HZ_FULL
867 atomic_t tick_dep_mask;
869 /* Context switch counts: */
871 unsigned long nivcsw;
873 /* Monotonic time in nsecs: */
876 /* Boot based time in nsecs: */
879 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
880 unsigned long min_flt;
881 unsigned long maj_flt;
883 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
884 struct posix_cputimers posix_cputimers;
886 /* Process credentials: */
888 /* Tracer's credentials at attach: */
889 const struct cred __rcu *ptracer_cred;
891 /* Objective and real subjective task credentials (COW): */
892 const struct cred __rcu *real_cred;
894 /* Effective (overridable) subjective task credentials (COW): */
895 const struct cred __rcu *cred;
898 /* Cached requested key. */
899 struct key *cached_requested_key;
903 * executable name, excluding path.
905 * - normally initialized setup_new_exec()
906 * - access it with [gs]et_task_comm()
907 * - lock it with task_lock()
909 char comm[TASK_COMM_LEN];
911 struct nameidata *nameidata;
913 #ifdef CONFIG_SYSVIPC
914 struct sysv_sem sysvsem;
915 struct sysv_shm sysvshm;
917 #ifdef CONFIG_DETECT_HUNG_TASK
918 unsigned long last_switch_count;
919 unsigned long last_switch_time;
921 /* Filesystem information: */
922 struct fs_struct *fs;
924 /* Open file information: */
925 struct files_struct *files;
928 struct nsproxy *nsproxy;
930 /* Signal handlers: */
931 struct signal_struct *signal;
932 struct sighand_struct __rcu *sighand;
934 sigset_t real_blocked;
935 /* Restored if set_restore_sigmask() was used: */
936 sigset_t saved_sigmask;
937 struct sigpending pending;
938 unsigned long sas_ss_sp;
940 unsigned int sas_ss_flags;
942 struct callback_head *task_works;
945 #ifdef CONFIG_AUDITSYSCALL
946 struct audit_context *audit_context;
949 unsigned int sessionid;
951 struct seccomp seccomp;
953 /* Thread group tracking: */
957 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
958 spinlock_t alloc_lock;
960 /* Protection of the PI data structures: */
961 raw_spinlock_t pi_lock;
963 struct wake_q_node wake_q;
965 #ifdef CONFIG_RT_MUTEXES
966 /* PI waiters blocked on a rt_mutex held by this task: */
967 struct rb_root_cached pi_waiters;
968 /* Updated under owner's pi_lock and rq lock */
969 struct task_struct *pi_top_task;
970 /* Deadlock detection and priority inheritance handling: */
971 struct rt_mutex_waiter *pi_blocked_on;
974 #ifdef CONFIG_DEBUG_MUTEXES
975 /* Mutex deadlock detection: */
976 struct mutex_waiter *blocked_on;
979 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
983 #ifdef CONFIG_TRACE_IRQFLAGS
984 struct irqtrace_events irqtrace;
985 unsigned int hardirq_threaded;
986 u64 hardirq_chain_key;
987 int softirqs_enabled;
992 #ifdef CONFIG_LOCKDEP
993 # define MAX_LOCK_DEPTH 48UL
996 unsigned int lockdep_recursion;
997 struct held_lock held_locks[MAX_LOCK_DEPTH];
1001 unsigned int in_ubsan;
1004 /* Journalling filesystem info: */
1007 /* Stacked block device info: */
1008 struct bio_list *bio_list;
1011 /* Stack plugging: */
1012 struct blk_plug *plug;
1016 struct reclaim_state *reclaim_state;
1018 struct backing_dev_info *backing_dev_info;
1020 struct io_context *io_context;
1022 #ifdef CONFIG_COMPACTION
1023 struct capture_control *capture_control;
1026 unsigned long ptrace_message;
1027 kernel_siginfo_t *last_siginfo;
1029 struct task_io_accounting ioac;
1031 /* Pressure stall state */
1032 unsigned int psi_flags;
1034 #ifdef CONFIG_TASK_XACCT
1035 /* Accumulated RSS usage: */
1037 /* Accumulated virtual memory usage: */
1039 /* stime + utime since last update: */
1042 #ifdef CONFIG_CPUSETS
1043 /* Protected by ->alloc_lock: */
1044 nodemask_t mems_allowed;
1045 /* Seqence number to catch updates: */
1046 seqcount_t mems_allowed_seq;
1047 int cpuset_mem_spread_rotor;
1048 int cpuset_slab_spread_rotor;
1050 #ifdef CONFIG_CGROUPS
1051 /* Control Group info protected by css_set_lock: */
1052 struct css_set __rcu *cgroups;
1053 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1054 struct list_head cg_list;
1056 #ifdef CONFIG_X86_CPU_RESCTRL
1061 struct robust_list_head __user *robust_list;
1062 #ifdef CONFIG_COMPAT
1063 struct compat_robust_list_head __user *compat_robust_list;
1065 struct list_head pi_state_list;
1066 struct futex_pi_state *pi_state_cache;
1067 struct mutex futex_exit_mutex;
1068 unsigned int futex_state;
1070 #ifdef CONFIG_PERF_EVENTS
1071 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1072 struct mutex perf_event_mutex;
1073 struct list_head perf_event_list;
1075 #ifdef CONFIG_DEBUG_PREEMPT
1076 unsigned long preempt_disable_ip;
1079 /* Protected by alloc_lock: */
1080 struct mempolicy *mempolicy;
1082 short pref_node_fork;
1084 #ifdef CONFIG_NUMA_BALANCING
1086 unsigned int numa_scan_period;
1087 unsigned int numa_scan_period_max;
1088 int numa_preferred_nid;
1089 unsigned long numa_migrate_retry;
1090 /* Migration stamp: */
1092 u64 last_task_numa_placement;
1093 u64 last_sum_exec_runtime;
1094 struct callback_head numa_work;
1097 * This pointer is only modified for current in syscall and
1098 * pagefault context (and for tasks being destroyed), so it can be read
1099 * from any of the following contexts:
1100 * - RCU read-side critical section
1101 * - current->numa_group from everywhere
1102 * - task's runqueue locked, task not running
1104 struct numa_group __rcu *numa_group;
1107 * numa_faults is an array split into four regions:
1108 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1109 * in this precise order.
1111 * faults_memory: Exponential decaying average of faults on a per-node
1112 * basis. Scheduling placement decisions are made based on these
1113 * counts. The values remain static for the duration of a PTE scan.
1114 * faults_cpu: Track the nodes the process was running on when a NUMA
1115 * hinting fault was incurred.
1116 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1117 * during the current scan window. When the scan completes, the counts
1118 * in faults_memory and faults_cpu decay and these values are copied.
1120 unsigned long *numa_faults;
1121 unsigned long total_numa_faults;
1124 * numa_faults_locality tracks if faults recorded during the last
1125 * scan window were remote/local or failed to migrate. The task scan
1126 * period is adapted based on the locality of the faults with different
1127 * weights depending on whether they were shared or private faults
1129 unsigned long numa_faults_locality[3];
1131 unsigned long numa_pages_migrated;
1132 #endif /* CONFIG_NUMA_BALANCING */
1135 struct rseq __user *rseq;
1138 * RmW on rseq_event_mask must be performed atomically
1139 * with respect to preemption.
1141 unsigned long rseq_event_mask;
1144 struct tlbflush_unmap_batch tlb_ubc;
1147 refcount_t rcu_users;
1148 struct rcu_head rcu;
1151 /* Cache last used pipe for splice(): */
1152 struct pipe_inode_info *splice_pipe;
1154 struct page_frag task_frag;
1156 #ifdef CONFIG_TASK_DELAY_ACCT
1157 struct task_delay_info *delays;
1160 #ifdef CONFIG_FAULT_INJECTION
1162 unsigned int fail_nth;
1165 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1166 * balance_dirty_pages() for a dirty throttling pause:
1169 int nr_dirtied_pause;
1170 /* Start of a write-and-pause period: */
1171 unsigned long dirty_paused_when;
1173 #ifdef CONFIG_LATENCYTOP
1174 int latency_record_count;
1175 struct latency_record latency_record[LT_SAVECOUNT];
1178 * Time slack values; these are used to round up poll() and
1179 * select() etc timeout values. These are in nanoseconds.
1182 u64 default_timer_slack_ns;
1185 unsigned int kasan_depth;
1189 struct kcsan_ctx kcsan_ctx;
1190 #ifdef CONFIG_TRACE_IRQFLAGS
1191 struct irqtrace_events kcsan_save_irqtrace;
1195 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1196 /* Index of current stored address in ret_stack: */
1200 /* Stack of return addresses for return function tracing: */
1201 struct ftrace_ret_stack *ret_stack;
1203 /* Timestamp for last schedule: */
1204 unsigned long long ftrace_timestamp;
1207 * Number of functions that haven't been traced
1208 * because of depth overrun:
1210 atomic_t trace_overrun;
1212 /* Pause tracing: */
1213 atomic_t tracing_graph_pause;
1216 #ifdef CONFIG_TRACING
1217 /* State flags for use by tracers: */
1218 unsigned long trace;
1220 /* Bitmask and counter of trace recursion: */
1221 unsigned long trace_recursion;
1222 #endif /* CONFIG_TRACING */
1225 /* See kernel/kcov.c for more details. */
1227 /* Coverage collection mode enabled for this task (0 if disabled): */
1228 unsigned int kcov_mode;
1230 /* Size of the kcov_area: */
1231 unsigned int kcov_size;
1233 /* Buffer for coverage collection: */
1236 /* KCOV descriptor wired with this task or NULL: */
1239 /* KCOV common handle for remote coverage collection: */
1242 /* KCOV sequence number: */
1245 /* Collect coverage from softirq context: */
1246 unsigned int kcov_softirq;
1250 struct mem_cgroup *memcg_in_oom;
1251 gfp_t memcg_oom_gfp_mask;
1252 int memcg_oom_order;
1254 /* Number of pages to reclaim on returning to userland: */
1255 unsigned int memcg_nr_pages_over_high;
1257 /* Used by memcontrol for targeted memcg charge: */
1258 struct mem_cgroup *active_memcg;
1261 #ifdef CONFIG_BLK_CGROUP
1262 struct request_queue *throttle_queue;
1265 #ifdef CONFIG_UPROBES
1266 struct uprobe_task *utask;
1268 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1269 unsigned int sequential_io;
1270 unsigned int sequential_io_avg;
1272 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1273 unsigned long task_state_change;
1275 int pagefault_disabled;
1277 struct task_struct *oom_reaper_list;
1279 #ifdef CONFIG_VMAP_STACK
1280 struct vm_struct *stack_vm_area;
1282 #ifdef CONFIG_THREAD_INFO_IN_TASK
1283 /* A live task holds one reference: */
1284 refcount_t stack_refcount;
1286 #ifdef CONFIG_LIVEPATCH
1289 #ifdef CONFIG_SECURITY
1290 /* Used by LSM modules for access restriction: */
1294 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1295 unsigned long lowest_stack;
1296 unsigned long prev_lowest_stack;
1299 #ifdef CONFIG_X86_MCE
1303 __mce_reserved : 62;
1304 struct callback_head mce_kill_me;
1308 * New fields for task_struct should be added above here, so that
1309 * they are included in the randomized portion of task_struct.
1311 randomized_struct_fields_end
1313 /* CPU-specific state of this task: */
1314 struct thread_struct thread;
1317 * WARNING: on x86, 'thread_struct' contains a variable-sized
1318 * structure. It *MUST* be at the end of 'task_struct'.
1320 * Do not put anything below here!
1324 static inline struct pid *task_pid(struct task_struct *task)
1326 return task->thread_pid;
1330 * the helpers to get the task's different pids as they are seen
1331 * from various namespaces
1333 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1334 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1336 * task_xid_nr_ns() : id seen from the ns specified;
1338 * see also pid_nr() etc in include/linux/pid.h
1340 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1342 static inline pid_t task_pid_nr(struct task_struct *tsk)
1347 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1349 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1352 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1354 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1358 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1364 * pid_alive - check that a task structure is not stale
1365 * @p: Task structure to be checked.
1367 * Test if a process is not yet dead (at most zombie state)
1368 * If pid_alive fails, then pointers within the task structure
1369 * can be stale and must not be dereferenced.
1371 * Return: 1 if the process is alive. 0 otherwise.
1373 static inline int pid_alive(const struct task_struct *p)
1375 return p->thread_pid != NULL;
1378 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1380 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1383 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1385 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1389 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1391 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1394 static inline pid_t task_session_vnr(struct task_struct *tsk)
1396 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1399 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1401 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1404 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1406 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1409 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1415 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1421 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1423 return task_ppid_nr_ns(tsk, &init_pid_ns);
1426 /* Obsolete, do not use: */
1427 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1429 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1432 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1433 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1435 static inline unsigned int task_state_index(struct task_struct *tsk)
1437 unsigned int tsk_state = READ_ONCE(tsk->state);
1438 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1440 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1442 if (tsk_state == TASK_IDLE)
1443 state = TASK_REPORT_IDLE;
1448 static inline char task_index_to_char(unsigned int state)
1450 static const char state_char[] = "RSDTtXZPI";
1452 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1454 return state_char[state];
1457 static inline char task_state_to_char(struct task_struct *tsk)
1459 return task_index_to_char(task_state_index(tsk));
1463 * is_global_init - check if a task structure is init. Since init
1464 * is free to have sub-threads we need to check tgid.
1465 * @tsk: Task structure to be checked.
1467 * Check if a task structure is the first user space task the kernel created.
1469 * Return: 1 if the task structure is init. 0 otherwise.
1471 static inline int is_global_init(struct task_struct *tsk)
1473 return task_tgid_nr(tsk) == 1;
1476 extern struct pid *cad_pid;
1481 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1482 #define PF_EXITING 0x00000004 /* Getting shut down */
1483 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1484 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1485 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1486 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1487 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1488 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1489 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1490 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1491 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1492 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1493 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1494 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1495 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1496 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1497 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1498 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1499 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1500 * I am cleaning dirty pages from some other bdi. */
1501 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1502 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1503 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1504 #define PF_UMH 0x02000000 /* I'm an Usermodehelper process */
1505 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1506 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1507 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1508 #define PF_IO_WORKER 0x20000000 /* Task is an IO worker */
1509 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1510 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1513 * Only the _current_ task can read/write to tsk->flags, but other
1514 * tasks can access tsk->flags in readonly mode for example
1515 * with tsk_used_math (like during threaded core dumping).
1516 * There is however an exception to this rule during ptrace
1517 * or during fork: the ptracer task is allowed to write to the
1518 * child->flags of its traced child (same goes for fork, the parent
1519 * can write to the child->flags), because we're guaranteed the
1520 * child is not running and in turn not changing child->flags
1521 * at the same time the parent does it.
1523 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1524 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1525 #define clear_used_math() clear_stopped_child_used_math(current)
1526 #define set_used_math() set_stopped_child_used_math(current)
1528 #define conditional_stopped_child_used_math(condition, child) \
1529 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1531 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1533 #define copy_to_stopped_child_used_math(child) \
1534 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1536 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1537 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1538 #define used_math() tsk_used_math(current)
1540 static inline bool is_percpu_thread(void)
1543 return (current->flags & PF_NO_SETAFFINITY) &&
1544 (current->nr_cpus_allowed == 1);
1550 /* Per-process atomic flags. */
1551 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1552 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1553 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1554 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1555 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1556 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1557 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1558 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1560 #define TASK_PFA_TEST(name, func) \
1561 static inline bool task_##func(struct task_struct *p) \
1562 { return test_bit(PFA_##name, &p->atomic_flags); }
1564 #define TASK_PFA_SET(name, func) \
1565 static inline void task_set_##func(struct task_struct *p) \
1566 { set_bit(PFA_##name, &p->atomic_flags); }
1568 #define TASK_PFA_CLEAR(name, func) \
1569 static inline void task_clear_##func(struct task_struct *p) \
1570 { clear_bit(PFA_##name, &p->atomic_flags); }
1572 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1573 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1575 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1576 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1577 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1579 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1580 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1581 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1583 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1584 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1585 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1587 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1588 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1589 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1591 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1592 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1594 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1595 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1596 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1598 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1599 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1602 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1604 current->flags &= ~flags;
1605 current->flags |= orig_flags & flags;
1608 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1609 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1611 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1612 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1614 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1617 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1619 if (!cpumask_test_cpu(0, new_mask))
1625 extern int yield_to(struct task_struct *p, bool preempt);
1626 extern void set_user_nice(struct task_struct *p, long nice);
1627 extern int task_prio(const struct task_struct *p);
1630 * task_nice - return the nice value of a given task.
1631 * @p: the task in question.
1633 * Return: The nice value [ -20 ... 0 ... 19 ].
1635 static inline int task_nice(const struct task_struct *p)
1637 return PRIO_TO_NICE((p)->static_prio);
1640 extern int can_nice(const struct task_struct *p, const int nice);
1641 extern int task_curr(const struct task_struct *p);
1642 extern int idle_cpu(int cpu);
1643 extern int available_idle_cpu(int cpu);
1644 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1645 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1646 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1647 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1648 extern struct task_struct *idle_task(int cpu);
1651 * is_idle_task - is the specified task an idle task?
1652 * @p: the task in question.
1654 * Return: 1 if @p is an idle task. 0 otherwise.
1656 static inline bool is_idle_task(const struct task_struct *p)
1658 return !!(p->flags & PF_IDLE);
1661 extern struct task_struct *curr_task(int cpu);
1662 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1666 union thread_union {
1667 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1668 struct task_struct task;
1670 #ifndef CONFIG_THREAD_INFO_IN_TASK
1671 struct thread_info thread_info;
1673 unsigned long stack[THREAD_SIZE/sizeof(long)];
1676 #ifndef CONFIG_THREAD_INFO_IN_TASK
1677 extern struct thread_info init_thread_info;
1680 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1682 #ifdef CONFIG_THREAD_INFO_IN_TASK
1683 static inline struct thread_info *task_thread_info(struct task_struct *task)
1685 return &task->thread_info;
1687 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1688 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1692 * find a task by one of its numerical ids
1694 * find_task_by_pid_ns():
1695 * finds a task by its pid in the specified namespace
1696 * find_task_by_vpid():
1697 * finds a task by its virtual pid
1699 * see also find_vpid() etc in include/linux/pid.h
1702 extern struct task_struct *find_task_by_vpid(pid_t nr);
1703 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1706 * find a task by its virtual pid and get the task struct
1708 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1710 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1711 extern int wake_up_process(struct task_struct *tsk);
1712 extern void wake_up_new_task(struct task_struct *tsk);
1715 extern void kick_process(struct task_struct *tsk);
1717 static inline void kick_process(struct task_struct *tsk) { }
1720 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1722 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1724 __set_task_comm(tsk, from, false);
1727 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1728 #define get_task_comm(buf, tsk) ({ \
1729 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1730 __get_task_comm(buf, sizeof(buf), tsk); \
1734 static __always_inline void scheduler_ipi(void)
1737 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1738 * TIF_NEED_RESCHED remotely (for the first time) will also send
1741 preempt_fold_need_resched();
1743 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1745 static inline void scheduler_ipi(void) { }
1746 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1753 * Set thread flags in other task's structures.
1754 * See asm/thread_info.h for TIF_xxxx flags available:
1756 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1758 set_ti_thread_flag(task_thread_info(tsk), flag);
1761 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1763 clear_ti_thread_flag(task_thread_info(tsk), flag);
1766 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1769 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1772 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1774 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1777 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1779 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1782 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1784 return test_ti_thread_flag(task_thread_info(tsk), flag);
1787 static inline void set_tsk_need_resched(struct task_struct *tsk)
1789 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1792 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1794 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1797 static inline int test_tsk_need_resched(struct task_struct *tsk)
1799 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1803 * cond_resched() and cond_resched_lock(): latency reduction via
1804 * explicit rescheduling in places that are safe. The return
1805 * value indicates whether a reschedule was done in fact.
1806 * cond_resched_lock() will drop the spinlock before scheduling,
1808 #ifndef CONFIG_PREEMPTION
1809 extern int _cond_resched(void);
1811 static inline int _cond_resched(void) { return 0; }
1814 #define cond_resched() ({ \
1815 ___might_sleep(__FILE__, __LINE__, 0); \
1819 extern int __cond_resched_lock(spinlock_t *lock);
1821 #define cond_resched_lock(lock) ({ \
1822 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1823 __cond_resched_lock(lock); \
1826 static inline void cond_resched_rcu(void)
1828 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1836 * Does a critical section need to be broken due to another
1837 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1838 * but a general need for low latency)
1840 static inline int spin_needbreak(spinlock_t *lock)
1842 #ifdef CONFIG_PREEMPTION
1843 return spin_is_contended(lock);
1849 static __always_inline bool need_resched(void)
1851 return unlikely(tif_need_resched());
1855 * Wrappers for p->thread_info->cpu access. No-op on UP.
1859 static inline unsigned int task_cpu(const struct task_struct *p)
1861 #ifdef CONFIG_THREAD_INFO_IN_TASK
1862 return READ_ONCE(p->cpu);
1864 return READ_ONCE(task_thread_info(p)->cpu);
1868 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1872 static inline unsigned int task_cpu(const struct task_struct *p)
1877 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1881 #endif /* CONFIG_SMP */
1884 * In order to reduce various lock holder preemption latencies provide an
1885 * interface to see if a vCPU is currently running or not.
1887 * This allows us to terminate optimistic spin loops and block, analogous to
1888 * the native optimistic spin heuristic of testing if the lock owner task is
1891 #ifndef vcpu_is_preempted
1892 static inline bool vcpu_is_preempted(int cpu)
1898 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1899 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1901 #ifndef TASK_SIZE_OF
1902 #define TASK_SIZE_OF(tsk) TASK_SIZE
1908 * Map the event mask on the user-space ABI enum rseq_cs_flags
1909 * for direct mask checks.
1911 enum rseq_event_mask_bits {
1912 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1913 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1914 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1917 enum rseq_event_mask {
1918 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1919 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1920 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1923 static inline void rseq_set_notify_resume(struct task_struct *t)
1926 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1929 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1931 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1932 struct pt_regs *regs)
1935 __rseq_handle_notify_resume(ksig, regs);
1938 static inline void rseq_signal_deliver(struct ksignal *ksig,
1939 struct pt_regs *regs)
1942 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1944 rseq_handle_notify_resume(ksig, regs);
1947 /* rseq_preempt() requires preemption to be disabled. */
1948 static inline void rseq_preempt(struct task_struct *t)
1950 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1951 rseq_set_notify_resume(t);
1954 /* rseq_migrate() requires preemption to be disabled. */
1955 static inline void rseq_migrate(struct task_struct *t)
1957 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1958 rseq_set_notify_resume(t);
1962 * If parent process has a registered restartable sequences area, the
1963 * child inherits. Unregister rseq for a clone with CLONE_VM set.
1965 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1967 if (clone_flags & CLONE_VM) {
1970 t->rseq_event_mask = 0;
1972 t->rseq = current->rseq;
1973 t->rseq_sig = current->rseq_sig;
1974 t->rseq_event_mask = current->rseq_event_mask;
1978 static inline void rseq_execve(struct task_struct *t)
1982 t->rseq_event_mask = 0;
1987 static inline void rseq_set_notify_resume(struct task_struct *t)
1990 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1991 struct pt_regs *regs)
1994 static inline void rseq_signal_deliver(struct ksignal *ksig,
1995 struct pt_regs *regs)
1998 static inline void rseq_preempt(struct task_struct *t)
2001 static inline void rseq_migrate(struct task_struct *t)
2004 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2007 static inline void rseq_execve(struct task_struct *t)
2013 void __exit_umh(struct task_struct *tsk);
2015 static inline void exit_umh(struct task_struct *tsk)
2017 if (unlikely(tsk->flags & PF_UMH))
2021 #ifdef CONFIG_DEBUG_RSEQ
2023 void rseq_syscall(struct pt_regs *regs);
2027 static inline void rseq_syscall(struct pt_regs *regs)
2033 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2034 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2035 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2037 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2038 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2039 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2041 int sched_trace_rq_cpu(struct rq *rq);
2043 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);