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/signal_types.h>
29 #include <linux/mm_types_task.h>
30 #include <linux/task_io_accounting.h>
31 #include <linux/rseq.h>
33 /* task_struct member predeclarations (sorted alphabetically): */
35 struct backing_dev_info;
38 struct capture_control;
41 struct futex_pi_state;
46 struct perf_event_context;
48 struct pipe_inode_info;
51 struct robust_list_head;
57 struct sighand_struct;
59 struct task_delay_info;
63 * Task state bitmask. NOTE! These bits are also
64 * encoded in fs/proc/array.c: get_task_state().
66 * We have two separate sets of flags: task->state
67 * is about runnability, while task->exit_state are
68 * about the task exiting. Confusing, but this way
69 * modifying one set can't modify the other one by
73 /* Used in tsk->state: */
74 #define TASK_RUNNING 0x0000
75 #define TASK_INTERRUPTIBLE 0x0001
76 #define TASK_UNINTERRUPTIBLE 0x0002
77 #define __TASK_STOPPED 0x0004
78 #define __TASK_TRACED 0x0008
79 /* Used in tsk->exit_state: */
80 #define EXIT_DEAD 0x0010
81 #define EXIT_ZOMBIE 0x0020
82 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
83 /* Used in tsk->state again: */
84 #define TASK_PARKED 0x0040
85 #define TASK_DEAD 0x0080
86 #define TASK_WAKEKILL 0x0100
87 #define TASK_WAKING 0x0200
88 #define TASK_NOLOAD 0x0400
89 #define TASK_NEW 0x0800
90 #define TASK_STATE_MAX 0x1000
92 /* Convenience macros for the sake of set_current_state: */
93 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
94 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
95 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
97 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
99 /* Convenience macros for the sake of wake_up(): */
100 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
102 /* get_task_state(): */
103 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
104 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
105 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
108 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
110 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
112 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
114 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
115 (task->flags & PF_FROZEN) == 0 && \
116 (task->state & TASK_NOLOAD) == 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);
225 extern int __must_check io_schedule_prepare(void);
226 extern void io_schedule_finish(int token);
227 extern long io_schedule_timeout(long timeout);
228 extern void io_schedule(void);
231 * struct prev_cputime - snapshot of system and user cputime
232 * @utime: time spent in user mode
233 * @stime: time spent in system mode
234 * @lock: protects the above two fields
236 * Stores previous user/system time values such that we can guarantee
239 struct prev_cputime {
240 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
248 * struct task_cputime - collected CPU time counts
249 * @utime: time spent in user mode, in nanoseconds
250 * @stime: time spent in kernel mode, in nanoseconds
251 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
253 * This structure groups together three kinds of CPU time that are tracked for
254 * threads and thread groups. Most things considering CPU time want to group
255 * these counts together and treat all three of them in parallel.
257 struct task_cputime {
260 unsigned long long sum_exec_runtime;
263 /* Alternate field names when used on cache expirations: */
264 #define virt_exp utime
265 #define prof_exp stime
266 #define sched_exp sum_exec_runtime
269 /* Task is sleeping or running in a CPU with VTIME inactive: */
271 /* Task runs in userspace in a CPU with VTIME active: */
273 /* Task runs in kernelspace in a CPU with VTIME active: */
279 unsigned long long starttime;
280 enum vtime_state state;
287 * Utilization clamp constraints.
288 * @UCLAMP_MIN: Minimum utilization
289 * @UCLAMP_MAX: Maximum utilization
290 * @UCLAMP_CNT: Utilization clamp constraints count
299 extern struct root_domain def_root_domain;
300 extern struct mutex sched_domains_mutex;
304 #ifdef CONFIG_SCHED_INFO
305 /* Cumulative counters: */
307 /* # of times we have run on this CPU: */
308 unsigned long pcount;
310 /* Time spent waiting on a runqueue: */
311 unsigned long long run_delay;
315 /* When did we last run on a CPU? */
316 unsigned long long last_arrival;
318 /* When were we last queued to run? */
319 unsigned long long last_queued;
321 #endif /* CONFIG_SCHED_INFO */
325 * Integer metrics need fixed point arithmetic, e.g., sched/fair
326 * has a few: load, load_avg, util_avg, freq, and capacity.
328 * We define a basic fixed point arithmetic range, and then formalize
329 * all these metrics based on that basic range.
331 # define SCHED_FIXEDPOINT_SHIFT 10
332 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
334 /* Increase resolution of cpu_capacity calculations */
335 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
336 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
339 unsigned long weight;
344 * struct util_est - Estimation utilization of FAIR tasks
345 * @enqueued: instantaneous estimated utilization of a task/cpu
346 * @ewma: the Exponential Weighted Moving Average (EWMA)
347 * utilization of a task
349 * Support data structure to track an Exponential Weighted Moving Average
350 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
351 * average each time a task completes an activation. Sample's weight is chosen
352 * so that the EWMA will be relatively insensitive to transient changes to the
355 * The enqueued attribute has a slightly different meaning for tasks and cpus:
356 * - task: the task's util_avg at last task dequeue time
357 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
358 * Thus, the util_est.enqueued of a task represents the contribution on the
359 * estimated utilization of the CPU where that task is currently enqueued.
361 * Only for tasks we track a moving average of the past instantaneous
362 * estimated utilization. This allows to absorb sporadic drops in utilization
363 * of an otherwise almost periodic task.
366 unsigned int enqueued;
368 #define UTIL_EST_WEIGHT_SHIFT 2
369 } __attribute__((__aligned__(sizeof(u64))));
372 * The load_avg/util_avg accumulates an infinite geometric series
373 * (see __update_load_avg() in kernel/sched/fair.c).
375 * [load_avg definition]
377 * load_avg = runnable% * scale_load_down(load)
379 * where runnable% is the time ratio that a sched_entity is runnable.
380 * For cfs_rq, it is the aggregated load_avg of all runnable and
381 * blocked sched_entities.
383 * [util_avg definition]
385 * util_avg = running% * SCHED_CAPACITY_SCALE
387 * where running% is the time ratio that a sched_entity is running on
388 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
389 * and blocked sched_entities.
391 * load_avg and util_avg don't direcly factor frequency scaling and CPU
392 * capacity scaling. The scaling is done through the rq_clock_pelt that
393 * is used for computing those signals (see update_rq_clock_pelt())
395 * N.B., the above ratios (runnable% and running%) themselves are in the
396 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
397 * to as large a range as necessary. This is for example reflected by
398 * util_avg's SCHED_CAPACITY_SCALE.
402 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
403 * with the highest load (=88761), always runnable on a single cfs_rq,
404 * and should not overflow as the number already hits PID_MAX_LIMIT.
406 * For all other cases (including 32-bit kernels), struct load_weight's
407 * weight will overflow first before we do, because:
409 * Max(load_avg) <= Max(load.weight)
411 * Then it is the load_weight's responsibility to consider overflow
415 u64 last_update_time;
417 u64 runnable_load_sum;
420 unsigned long load_avg;
421 unsigned long runnable_load_avg;
422 unsigned long util_avg;
423 struct util_est util_est;
424 } ____cacheline_aligned;
426 struct sched_statistics {
427 #ifdef CONFIG_SCHEDSTATS
437 s64 sum_sleep_runtime;
444 u64 nr_migrations_cold;
445 u64 nr_failed_migrations_affine;
446 u64 nr_failed_migrations_running;
447 u64 nr_failed_migrations_hot;
448 u64 nr_forced_migrations;
452 u64 nr_wakeups_migrate;
453 u64 nr_wakeups_local;
454 u64 nr_wakeups_remote;
455 u64 nr_wakeups_affine;
456 u64 nr_wakeups_affine_attempts;
457 u64 nr_wakeups_passive;
462 struct sched_entity {
463 /* For load-balancing: */
464 struct load_weight load;
465 unsigned long runnable_weight;
466 struct rb_node run_node;
467 struct list_head group_node;
471 u64 sum_exec_runtime;
473 u64 prev_sum_exec_runtime;
477 struct sched_statistics statistics;
479 #ifdef CONFIG_FAIR_GROUP_SCHED
481 struct sched_entity *parent;
482 /* rq on which this entity is (to be) queued: */
483 struct cfs_rq *cfs_rq;
484 /* rq "owned" by this entity/group: */
490 * Per entity load average tracking.
492 * Put into separate cache line so it does not
493 * collide with read-mostly values above.
495 struct sched_avg avg;
499 struct sched_rt_entity {
500 struct list_head run_list;
501 unsigned long timeout;
502 unsigned long watchdog_stamp;
503 unsigned int time_slice;
504 unsigned short on_rq;
505 unsigned short on_list;
507 struct sched_rt_entity *back;
508 #ifdef CONFIG_RT_GROUP_SCHED
509 struct sched_rt_entity *parent;
510 /* rq on which this entity is (to be) queued: */
512 /* rq "owned" by this entity/group: */
515 } __randomize_layout;
517 struct sched_dl_entity {
518 struct rb_node rb_node;
521 * Original scheduling parameters. Copied here from sched_attr
522 * during sched_setattr(), they will remain the same until
523 * the next sched_setattr().
525 u64 dl_runtime; /* Maximum runtime for each instance */
526 u64 dl_deadline; /* Relative deadline of each instance */
527 u64 dl_period; /* Separation of two instances (period) */
528 u64 dl_bw; /* dl_runtime / dl_period */
529 u64 dl_density; /* dl_runtime / dl_deadline */
532 * Actual scheduling parameters. Initialized with the values above,
533 * they are continuously updated during task execution. Note that
534 * the remaining runtime could be < 0 in case we are in overrun.
536 s64 runtime; /* Remaining runtime for this instance */
537 u64 deadline; /* Absolute deadline for this instance */
538 unsigned int flags; /* Specifying the scheduler behaviour */
543 * @dl_throttled tells if we exhausted the runtime. If so, the
544 * task has to wait for a replenishment to be performed at the
545 * next firing of dl_timer.
547 * @dl_boosted tells if we are boosted due to DI. If so we are
548 * outside bandwidth enforcement mechanism (but only until we
549 * exit the critical section);
551 * @dl_yielded tells if task gave up the CPU before consuming
552 * all its available runtime during the last job.
554 * @dl_non_contending tells if the task is inactive while still
555 * contributing to the active utilization. In other words, it
556 * indicates if the inactive timer has been armed and its handler
557 * has not been executed yet. This flag is useful to avoid race
558 * conditions between the inactive timer handler and the wakeup
561 * @dl_overrun tells if the task asked to be informed about runtime
564 unsigned int dl_throttled : 1;
565 unsigned int dl_boosted : 1;
566 unsigned int dl_yielded : 1;
567 unsigned int dl_non_contending : 1;
568 unsigned int dl_overrun : 1;
571 * Bandwidth enforcement timer. Each -deadline task has its
572 * own bandwidth to be enforced, thus we need one timer per task.
574 struct hrtimer dl_timer;
577 * Inactive timer, responsible for decreasing the active utilization
578 * at the "0-lag time". When a -deadline task blocks, it contributes
579 * to GRUB's active utilization until the "0-lag time", hence a
580 * timer is needed to decrease the active utilization at the correct
583 struct hrtimer inactive_timer;
586 #ifdef CONFIG_UCLAMP_TASK
587 /* Number of utilization clamp buckets (shorter alias) */
588 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
591 * Utilization clamp for a scheduling entity
592 * @value: clamp value "assigned" to a se
593 * @bucket_id: bucket index corresponding to the "assigned" value
594 * @active: the se is currently refcounted in a rq's bucket
595 * @user_defined: the requested clamp value comes from user-space
597 * The bucket_id is the index of the clamp bucket matching the clamp value
598 * which is pre-computed and stored to avoid expensive integer divisions from
601 * The active bit is set whenever a task has got an "effective" value assigned,
602 * which can be different from the clamp value "requested" from user-space.
603 * This allows to know a task is refcounted in the rq's bucket corresponding
604 * to the "effective" bucket_id.
606 * The user_defined bit is set whenever a task has got a task-specific clamp
607 * value requested from userspace, i.e. the system defaults apply to this task
608 * just as a restriction. This allows to relax default clamps when a less
609 * restrictive task-specific value has been requested, thus allowing to
610 * implement a "nice" semantic. For example, a task running with a 20%
611 * default boost can still drop its own boosting to 0%.
614 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
615 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
616 unsigned int active : 1;
617 unsigned int user_defined : 1;
619 #endif /* CONFIG_UCLAMP_TASK */
625 u8 exp_hint; /* Hint for performance. */
628 u32 s; /* Set of bits. */
631 enum perf_event_task_context {
632 perf_invalid_context = -1,
635 perf_nr_task_contexts,
639 struct wake_q_node *next;
643 #ifdef CONFIG_THREAD_INFO_IN_TASK
645 * For reasons of header soup (see current_thread_info()), this
646 * must be the first element of task_struct.
648 struct thread_info thread_info;
650 /* -1 unrunnable, 0 runnable, >0 stopped: */
654 * This begins the randomizable portion of task_struct. Only
655 * scheduling-critical items should be added above here.
657 randomized_struct_fields_start
661 /* Per task flags (PF_*), defined further below: */
666 struct llist_node wake_entry;
668 #ifdef CONFIG_THREAD_INFO_IN_TASK
672 unsigned int wakee_flips;
673 unsigned long wakee_flip_decay_ts;
674 struct task_struct *last_wakee;
677 * recent_used_cpu is initially set as the last CPU used by a task
678 * that wakes affine another task. Waker/wakee relationships can
679 * push tasks around a CPU where each wakeup moves to the next one.
680 * Tracking a recently used CPU allows a quick search for a recently
681 * used CPU that may be idle.
691 unsigned int rt_priority;
693 const struct sched_class *sched_class;
694 struct sched_entity se;
695 struct sched_rt_entity rt;
696 #ifdef CONFIG_CGROUP_SCHED
697 struct task_group *sched_task_group;
699 struct sched_dl_entity dl;
701 #ifdef CONFIG_UCLAMP_TASK
702 /* Clamp values requested for a scheduling entity */
703 struct uclamp_se uclamp_req[UCLAMP_CNT];
704 /* Effective clamp values used for a scheduling entity */
705 struct uclamp_se uclamp[UCLAMP_CNT];
708 #ifdef CONFIG_PREEMPT_NOTIFIERS
709 /* List of struct preempt_notifier: */
710 struct hlist_head preempt_notifiers;
713 #ifdef CONFIG_BLK_DEV_IO_TRACE
714 unsigned int btrace_seq;
719 const cpumask_t *cpus_ptr;
722 #ifdef CONFIG_PREEMPT_RCU
723 int rcu_read_lock_nesting;
724 union rcu_special rcu_read_unlock_special;
725 struct list_head rcu_node_entry;
726 struct rcu_node *rcu_blocked_node;
727 #endif /* #ifdef CONFIG_PREEMPT_RCU */
729 #ifdef CONFIG_TASKS_RCU
730 unsigned long rcu_tasks_nvcsw;
731 u8 rcu_tasks_holdout;
733 int rcu_tasks_idle_cpu;
734 struct list_head rcu_tasks_holdout_list;
735 #endif /* #ifdef CONFIG_TASKS_RCU */
737 struct sched_info sched_info;
739 struct list_head tasks;
741 struct plist_node pushable_tasks;
742 struct rb_node pushable_dl_tasks;
745 struct mm_struct *mm;
746 struct mm_struct *active_mm;
748 /* Per-thread vma caching: */
749 struct vmacache vmacache;
751 #ifdef SPLIT_RSS_COUNTING
752 struct task_rss_stat rss_stat;
757 /* The signal sent when the parent dies: */
759 /* JOBCTL_*, siglock protected: */
760 unsigned long jobctl;
762 /* Used for emulating ABI behavior of previous Linux versions: */
763 unsigned int personality;
765 /* Scheduler bits, serialized by scheduler locks: */
766 unsigned sched_reset_on_fork:1;
767 unsigned sched_contributes_to_load:1;
768 unsigned sched_migrated:1;
769 unsigned sched_remote_wakeup:1;
771 unsigned sched_psi_wake_requeue:1;
774 /* Force alignment to the next boundary: */
777 /* Unserialized, strictly 'current' */
779 /* Bit to tell LSMs we're in execve(): */
780 unsigned in_execve:1;
781 unsigned in_iowait:1;
782 #ifndef TIF_RESTORE_SIGMASK
783 unsigned restore_sigmask:1;
786 unsigned in_user_fault:1;
788 #ifdef CONFIG_COMPAT_BRK
789 unsigned brk_randomized:1;
791 #ifdef CONFIG_CGROUPS
792 /* disallow userland-initiated cgroup migration */
793 unsigned no_cgroup_migration:1;
794 /* task is frozen/stopped (used by the cgroup freezer) */
797 #ifdef CONFIG_BLK_CGROUP
798 /* to be used once the psi infrastructure lands upstream. */
799 unsigned use_memdelay:1;
802 unsigned long atomic_flags; /* Flags requiring atomic access. */
804 struct restart_block restart_block;
809 #ifdef CONFIG_STACKPROTECTOR
810 /* Canary value for the -fstack-protector GCC feature: */
811 unsigned long stack_canary;
814 * Pointers to the (original) parent process, youngest child, younger sibling,
815 * older sibling, respectively. (p->father can be replaced with
816 * p->real_parent->pid)
819 /* Real parent process: */
820 struct task_struct __rcu *real_parent;
822 /* Recipient of SIGCHLD, wait4() reports: */
823 struct task_struct __rcu *parent;
826 * Children/sibling form the list of natural children:
828 struct list_head children;
829 struct list_head sibling;
830 struct task_struct *group_leader;
833 * 'ptraced' is the list of tasks this task is using ptrace() on.
835 * This includes both natural children and PTRACE_ATTACH targets.
836 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
838 struct list_head ptraced;
839 struct list_head ptrace_entry;
841 /* PID/PID hash table linkage. */
842 struct pid *thread_pid;
843 struct hlist_node pid_links[PIDTYPE_MAX];
844 struct list_head thread_group;
845 struct list_head thread_node;
847 struct completion *vfork_done;
849 /* CLONE_CHILD_SETTID: */
850 int __user *set_child_tid;
852 /* CLONE_CHILD_CLEARTID: */
853 int __user *clear_child_tid;
857 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
862 struct prev_cputime prev_cputime;
863 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
867 #ifdef CONFIG_NO_HZ_FULL
868 atomic_t tick_dep_mask;
870 /* Context switch counts: */
872 unsigned long nivcsw;
874 /* Monotonic time in nsecs: */
877 /* Boot based time in nsecs: */
880 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
881 unsigned long min_flt;
882 unsigned long maj_flt;
884 #ifdef CONFIG_POSIX_TIMERS
885 struct task_cputime cputime_expires;
886 struct list_head cpu_timers[3];
889 /* Process credentials: */
891 /* Tracer's credentials at attach: */
892 const struct cred __rcu *ptracer_cred;
894 /* Objective and real subjective task credentials (COW): */
895 const struct cred __rcu *real_cred;
897 /* Effective (overridable) subjective task credentials (COW): */
898 const struct cred __rcu *cred;
901 /* Cached requested key. */
902 struct key *cached_requested_key;
906 * executable name, excluding path.
908 * - normally initialized setup_new_exec()
909 * - access it with [gs]et_task_comm()
910 * - lock it with task_lock()
912 char comm[TASK_COMM_LEN];
914 struct nameidata *nameidata;
916 #ifdef CONFIG_SYSVIPC
917 struct sysv_sem sysvsem;
918 struct sysv_shm sysvshm;
920 #ifdef CONFIG_DETECT_HUNG_TASK
921 unsigned long last_switch_count;
922 unsigned long last_switch_time;
924 /* Filesystem information: */
925 struct fs_struct *fs;
927 /* Open file information: */
928 struct files_struct *files;
931 struct nsproxy *nsproxy;
933 /* Signal handlers: */
934 struct signal_struct *signal;
935 struct sighand_struct *sighand;
937 sigset_t real_blocked;
938 /* Restored if set_restore_sigmask() was used: */
939 sigset_t saved_sigmask;
940 struct sigpending pending;
941 unsigned long sas_ss_sp;
943 unsigned int sas_ss_flags;
945 struct callback_head *task_works;
948 #ifdef CONFIG_AUDITSYSCALL
949 struct audit_context *audit_context;
952 unsigned int sessionid;
954 struct seccomp seccomp;
956 /* Thread group tracking: */
960 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
961 spinlock_t alloc_lock;
963 /* Protection of the PI data structures: */
964 raw_spinlock_t pi_lock;
966 struct wake_q_node wake_q;
968 #ifdef CONFIG_RT_MUTEXES
969 /* PI waiters blocked on a rt_mutex held by this task: */
970 struct rb_root_cached pi_waiters;
971 /* Updated under owner's pi_lock and rq lock */
972 struct task_struct *pi_top_task;
973 /* Deadlock detection and priority inheritance handling: */
974 struct rt_mutex_waiter *pi_blocked_on;
977 #ifdef CONFIG_DEBUG_MUTEXES
978 /* Mutex deadlock detection: */
979 struct mutex_waiter *blocked_on;
982 #ifdef CONFIG_TRACE_IRQFLAGS
983 unsigned int irq_events;
984 unsigned long hardirq_enable_ip;
985 unsigned long hardirq_disable_ip;
986 unsigned int hardirq_enable_event;
987 unsigned int hardirq_disable_event;
988 int hardirqs_enabled;
990 unsigned long softirq_disable_ip;
991 unsigned long softirq_enable_ip;
992 unsigned int softirq_disable_event;
993 unsigned int softirq_enable_event;
994 int softirqs_enabled;
998 #ifdef CONFIG_LOCKDEP
999 # define MAX_LOCK_DEPTH 48UL
1002 unsigned int lockdep_recursion;
1003 struct held_lock held_locks[MAX_LOCK_DEPTH];
1007 unsigned int in_ubsan;
1010 /* Journalling filesystem info: */
1013 /* Stacked block device info: */
1014 struct bio_list *bio_list;
1017 /* Stack plugging: */
1018 struct blk_plug *plug;
1022 struct reclaim_state *reclaim_state;
1024 struct backing_dev_info *backing_dev_info;
1026 struct io_context *io_context;
1028 #ifdef CONFIG_COMPACTION
1029 struct capture_control *capture_control;
1032 unsigned long ptrace_message;
1033 kernel_siginfo_t *last_siginfo;
1035 struct task_io_accounting ioac;
1037 /* Pressure stall state */
1038 unsigned int psi_flags;
1040 #ifdef CONFIG_TASK_XACCT
1041 /* Accumulated RSS usage: */
1043 /* Accumulated virtual memory usage: */
1045 /* stime + utime since last update: */
1048 #ifdef CONFIG_CPUSETS
1049 /* Protected by ->alloc_lock: */
1050 nodemask_t mems_allowed;
1051 /* Seqence number to catch updates: */
1052 seqcount_t mems_allowed_seq;
1053 int cpuset_mem_spread_rotor;
1054 int cpuset_slab_spread_rotor;
1056 #ifdef CONFIG_CGROUPS
1057 /* Control Group info protected by css_set_lock: */
1058 struct css_set __rcu *cgroups;
1059 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1060 struct list_head cg_list;
1062 #ifdef CONFIG_X86_CPU_RESCTRL
1067 struct robust_list_head __user *robust_list;
1068 #ifdef CONFIG_COMPAT
1069 struct compat_robust_list_head __user *compat_robust_list;
1071 struct list_head pi_state_list;
1072 struct futex_pi_state *pi_state_cache;
1074 #ifdef CONFIG_PERF_EVENTS
1075 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1076 struct mutex perf_event_mutex;
1077 struct list_head perf_event_list;
1079 #ifdef CONFIG_DEBUG_PREEMPT
1080 unsigned long preempt_disable_ip;
1083 /* Protected by alloc_lock: */
1084 struct mempolicy *mempolicy;
1086 short pref_node_fork;
1088 #ifdef CONFIG_NUMA_BALANCING
1090 unsigned int numa_scan_period;
1091 unsigned int numa_scan_period_max;
1092 int numa_preferred_nid;
1093 unsigned long numa_migrate_retry;
1094 /* Migration stamp: */
1096 u64 last_task_numa_placement;
1097 u64 last_sum_exec_runtime;
1098 struct callback_head numa_work;
1101 * This pointer is only modified for current in syscall and
1102 * pagefault context (and for tasks being destroyed), so it can be read
1103 * from any of the following contexts:
1104 * - RCU read-side critical section
1105 * - current->numa_group from everywhere
1106 * - task's runqueue locked, task not running
1108 struct numa_group __rcu *numa_group;
1111 * numa_faults is an array split into four regions:
1112 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1113 * in this precise order.
1115 * faults_memory: Exponential decaying average of faults on a per-node
1116 * basis. Scheduling placement decisions are made based on these
1117 * counts. The values remain static for the duration of a PTE scan.
1118 * faults_cpu: Track the nodes the process was running on when a NUMA
1119 * hinting fault was incurred.
1120 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1121 * during the current scan window. When the scan completes, the counts
1122 * in faults_memory and faults_cpu decay and these values are copied.
1124 unsigned long *numa_faults;
1125 unsigned long total_numa_faults;
1128 * numa_faults_locality tracks if faults recorded during the last
1129 * scan window were remote/local or failed to migrate. The task scan
1130 * period is adapted based on the locality of the faults with different
1131 * weights depending on whether they were shared or private faults
1133 unsigned long numa_faults_locality[3];
1135 unsigned long numa_pages_migrated;
1136 #endif /* CONFIG_NUMA_BALANCING */
1139 struct rseq __user *rseq;
1142 * RmW on rseq_event_mask must be performed atomically
1143 * with respect to preemption.
1145 unsigned long rseq_event_mask;
1148 struct tlbflush_unmap_batch tlb_ubc;
1150 struct rcu_head rcu;
1152 /* Cache last used pipe for splice(): */
1153 struct pipe_inode_info *splice_pipe;
1155 struct page_frag task_frag;
1157 #ifdef CONFIG_TASK_DELAY_ACCT
1158 struct task_delay_info *delays;
1161 #ifdef CONFIG_FAULT_INJECTION
1163 unsigned int fail_nth;
1166 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1167 * balance_dirty_pages() for a dirty throttling pause:
1170 int nr_dirtied_pause;
1171 /* Start of a write-and-pause period: */
1172 unsigned long dirty_paused_when;
1174 #ifdef CONFIG_LATENCYTOP
1175 int latency_record_count;
1176 struct latency_record latency_record[LT_SAVECOUNT];
1179 * Time slack values; these are used to round up poll() and
1180 * select() etc timeout values. These are in nanoseconds.
1183 u64 default_timer_slack_ns;
1186 unsigned int kasan_depth;
1189 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1190 /* Index of current stored address in ret_stack: */
1194 /* Stack of return addresses for return function tracing: */
1195 struct ftrace_ret_stack *ret_stack;
1197 /* Timestamp for last schedule: */
1198 unsigned long long ftrace_timestamp;
1201 * Number of functions that haven't been traced
1202 * because of depth overrun:
1204 atomic_t trace_overrun;
1206 /* Pause tracing: */
1207 atomic_t tracing_graph_pause;
1210 #ifdef CONFIG_TRACING
1211 /* State flags for use by tracers: */
1212 unsigned long trace;
1214 /* Bitmask and counter of trace recursion: */
1215 unsigned long trace_recursion;
1216 #endif /* CONFIG_TRACING */
1219 /* Coverage collection mode enabled for this task (0 if disabled): */
1220 unsigned int kcov_mode;
1222 /* Size of the kcov_area: */
1223 unsigned int kcov_size;
1225 /* Buffer for coverage collection: */
1228 /* KCOV descriptor wired with this task or NULL: */
1233 struct mem_cgroup *memcg_in_oom;
1234 gfp_t memcg_oom_gfp_mask;
1235 int memcg_oom_order;
1237 /* Number of pages to reclaim on returning to userland: */
1238 unsigned int memcg_nr_pages_over_high;
1240 /* Used by memcontrol for targeted memcg charge: */
1241 struct mem_cgroup *active_memcg;
1244 #ifdef CONFIG_BLK_CGROUP
1245 struct request_queue *throttle_queue;
1248 #ifdef CONFIG_UPROBES
1249 struct uprobe_task *utask;
1251 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1252 unsigned int sequential_io;
1253 unsigned int sequential_io_avg;
1255 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1256 unsigned long task_state_change;
1258 int pagefault_disabled;
1260 struct task_struct *oom_reaper_list;
1262 #ifdef CONFIG_VMAP_STACK
1263 struct vm_struct *stack_vm_area;
1265 #ifdef CONFIG_THREAD_INFO_IN_TASK
1266 /* A live task holds one reference: */
1267 refcount_t stack_refcount;
1269 #ifdef CONFIG_LIVEPATCH
1272 #ifdef CONFIG_SECURITY
1273 /* Used by LSM modules for access restriction: */
1277 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1278 unsigned long lowest_stack;
1279 unsigned long prev_lowest_stack;
1283 * New fields for task_struct should be added above here, so that
1284 * they are included in the randomized portion of task_struct.
1286 randomized_struct_fields_end
1288 /* CPU-specific state of this task: */
1289 struct thread_struct thread;
1292 * WARNING: on x86, 'thread_struct' contains a variable-sized
1293 * structure. It *MUST* be at the end of 'task_struct'.
1295 * Do not put anything below here!
1299 static inline struct pid *task_pid(struct task_struct *task)
1301 return task->thread_pid;
1305 * the helpers to get the task's different pids as they are seen
1306 * from various namespaces
1308 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1309 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1311 * task_xid_nr_ns() : id seen from the ns specified;
1313 * see also pid_nr() etc in include/linux/pid.h
1315 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1317 static inline pid_t task_pid_nr(struct task_struct *tsk)
1322 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1324 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1327 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1329 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1333 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1339 * pid_alive - check that a task structure is not stale
1340 * @p: Task structure to be checked.
1342 * Test if a process is not yet dead (at most zombie state)
1343 * If pid_alive fails, then pointers within the task structure
1344 * can be stale and must not be dereferenced.
1346 * Return: 1 if the process is alive. 0 otherwise.
1348 static inline int pid_alive(const struct task_struct *p)
1350 return p->thread_pid != NULL;
1353 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1355 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1358 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1360 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1364 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1366 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1369 static inline pid_t task_session_vnr(struct task_struct *tsk)
1371 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1374 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1376 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1379 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1381 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1384 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1390 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1396 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1398 return task_ppid_nr_ns(tsk, &init_pid_ns);
1401 /* Obsolete, do not use: */
1402 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1404 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1407 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1408 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1410 static inline unsigned int task_state_index(struct task_struct *tsk)
1412 unsigned int tsk_state = READ_ONCE(tsk->state);
1413 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1415 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1417 if (tsk_state == TASK_IDLE)
1418 state = TASK_REPORT_IDLE;
1423 static inline char task_index_to_char(unsigned int state)
1425 static const char state_char[] = "RSDTtXZPI";
1427 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1429 return state_char[state];
1432 static inline char task_state_to_char(struct task_struct *tsk)
1434 return task_index_to_char(task_state_index(tsk));
1438 * is_global_init - check if a task structure is init. Since init
1439 * is free to have sub-threads we need to check tgid.
1440 * @tsk: Task structure to be checked.
1442 * Check if a task structure is the first user space task the kernel created.
1444 * Return: 1 if the task structure is init. 0 otherwise.
1446 static inline int is_global_init(struct task_struct *tsk)
1448 return task_tgid_nr(tsk) == 1;
1451 extern struct pid *cad_pid;
1456 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1457 #define PF_EXITING 0x00000004 /* Getting shut down */
1458 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1459 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1460 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1461 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1462 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1463 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1464 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1465 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1466 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1467 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1468 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1469 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1470 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1471 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1472 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1473 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1474 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1475 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1476 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1477 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1478 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1479 #define PF_MEMSTALL 0x01000000 /* Stalled due to lack of memory */
1480 #define PF_UMH 0x02000000 /* I'm an Usermodehelper process */
1481 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1482 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1483 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1484 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1485 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1488 * Only the _current_ task can read/write to tsk->flags, but other
1489 * tasks can access tsk->flags in readonly mode for example
1490 * with tsk_used_math (like during threaded core dumping).
1491 * There is however an exception to this rule during ptrace
1492 * or during fork: the ptracer task is allowed to write to the
1493 * child->flags of its traced child (same goes for fork, the parent
1494 * can write to the child->flags), because we're guaranteed the
1495 * child is not running and in turn not changing child->flags
1496 * at the same time the parent does it.
1498 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1499 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1500 #define clear_used_math() clear_stopped_child_used_math(current)
1501 #define set_used_math() set_stopped_child_used_math(current)
1503 #define conditional_stopped_child_used_math(condition, child) \
1504 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1506 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1508 #define copy_to_stopped_child_used_math(child) \
1509 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1511 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1512 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1513 #define used_math() tsk_used_math(current)
1515 static inline bool is_percpu_thread(void)
1518 return (current->flags & PF_NO_SETAFFINITY) &&
1519 (current->nr_cpus_allowed == 1);
1525 /* Per-process atomic flags. */
1526 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1527 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1528 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1529 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1530 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1531 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1532 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1533 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1535 #define TASK_PFA_TEST(name, func) \
1536 static inline bool task_##func(struct task_struct *p) \
1537 { return test_bit(PFA_##name, &p->atomic_flags); }
1539 #define TASK_PFA_SET(name, func) \
1540 static inline void task_set_##func(struct task_struct *p) \
1541 { set_bit(PFA_##name, &p->atomic_flags); }
1543 #define TASK_PFA_CLEAR(name, func) \
1544 static inline void task_clear_##func(struct task_struct *p) \
1545 { clear_bit(PFA_##name, &p->atomic_flags); }
1547 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1548 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1550 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1551 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1552 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1554 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1555 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1556 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1558 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1559 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1560 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1562 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1563 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1564 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1566 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1567 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1569 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1570 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1571 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1573 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1574 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1577 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1579 current->flags &= ~flags;
1580 current->flags |= orig_flags & flags;
1583 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1584 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1586 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1587 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1589 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1592 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1594 if (!cpumask_test_cpu(0, new_mask))
1600 extern int yield_to(struct task_struct *p, bool preempt);
1601 extern void set_user_nice(struct task_struct *p, long nice);
1602 extern int task_prio(const struct task_struct *p);
1605 * task_nice - return the nice value of a given task.
1606 * @p: the task in question.
1608 * Return: The nice value [ -20 ... 0 ... 19 ].
1610 static inline int task_nice(const struct task_struct *p)
1612 return PRIO_TO_NICE((p)->static_prio);
1615 extern int can_nice(const struct task_struct *p, const int nice);
1616 extern int task_curr(const struct task_struct *p);
1617 extern int idle_cpu(int cpu);
1618 extern int available_idle_cpu(int cpu);
1619 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1620 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1621 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1622 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1623 extern struct task_struct *idle_task(int cpu);
1626 * is_idle_task - is the specified task an idle task?
1627 * @p: the task in question.
1629 * Return: 1 if @p is an idle task. 0 otherwise.
1631 static inline bool is_idle_task(const struct task_struct *p)
1633 return !!(p->flags & PF_IDLE);
1636 extern struct task_struct *curr_task(int cpu);
1637 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1641 union thread_union {
1642 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1643 struct task_struct task;
1645 #ifndef CONFIG_THREAD_INFO_IN_TASK
1646 struct thread_info thread_info;
1648 unsigned long stack[THREAD_SIZE/sizeof(long)];
1651 #ifndef CONFIG_THREAD_INFO_IN_TASK
1652 extern struct thread_info init_thread_info;
1655 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1657 #ifdef CONFIG_THREAD_INFO_IN_TASK
1658 static inline struct thread_info *task_thread_info(struct task_struct *task)
1660 return &task->thread_info;
1662 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1663 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1667 * find a task by one of its numerical ids
1669 * find_task_by_pid_ns():
1670 * finds a task by its pid in the specified namespace
1671 * find_task_by_vpid():
1672 * finds a task by its virtual pid
1674 * see also find_vpid() etc in include/linux/pid.h
1677 extern struct task_struct *find_task_by_vpid(pid_t nr);
1678 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1681 * find a task by its virtual pid and get the task struct
1683 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1685 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1686 extern int wake_up_process(struct task_struct *tsk);
1687 extern void wake_up_new_task(struct task_struct *tsk);
1690 extern void kick_process(struct task_struct *tsk);
1692 static inline void kick_process(struct task_struct *tsk) { }
1695 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1697 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1699 __set_task_comm(tsk, from, false);
1702 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1703 #define get_task_comm(buf, tsk) ({ \
1704 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1705 __get_task_comm(buf, sizeof(buf), tsk); \
1709 void scheduler_ipi(void);
1710 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1712 static inline void scheduler_ipi(void) { }
1713 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1720 * Set thread flags in other task's structures.
1721 * See asm/thread_info.h for TIF_xxxx flags available:
1723 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1725 set_ti_thread_flag(task_thread_info(tsk), flag);
1728 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1730 clear_ti_thread_flag(task_thread_info(tsk), flag);
1733 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1736 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1739 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1741 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1744 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1746 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1749 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1751 return test_ti_thread_flag(task_thread_info(tsk), flag);
1754 static inline void set_tsk_need_resched(struct task_struct *tsk)
1756 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1759 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1761 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1764 static inline int test_tsk_need_resched(struct task_struct *tsk)
1766 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1770 * cond_resched() and cond_resched_lock(): latency reduction via
1771 * explicit rescheduling in places that are safe. The return
1772 * value indicates whether a reschedule was done in fact.
1773 * cond_resched_lock() will drop the spinlock before scheduling,
1775 #ifndef CONFIG_PREEMPTION
1776 extern int _cond_resched(void);
1778 static inline int _cond_resched(void) { return 0; }
1781 #define cond_resched() ({ \
1782 ___might_sleep(__FILE__, __LINE__, 0); \
1786 extern int __cond_resched_lock(spinlock_t *lock);
1788 #define cond_resched_lock(lock) ({ \
1789 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1790 __cond_resched_lock(lock); \
1793 static inline void cond_resched_rcu(void)
1795 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1803 * Does a critical section need to be broken due to another
1804 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1805 * but a general need for low latency)
1807 static inline int spin_needbreak(spinlock_t *lock)
1809 #ifdef CONFIG_PREEMPTION
1810 return spin_is_contended(lock);
1816 static __always_inline bool need_resched(void)
1818 return unlikely(tif_need_resched());
1822 * Wrappers for p->thread_info->cpu access. No-op on UP.
1826 static inline unsigned int task_cpu(const struct task_struct *p)
1828 #ifdef CONFIG_THREAD_INFO_IN_TASK
1829 return READ_ONCE(p->cpu);
1831 return READ_ONCE(task_thread_info(p)->cpu);
1835 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1839 static inline unsigned int task_cpu(const struct task_struct *p)
1844 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1848 #endif /* CONFIG_SMP */
1851 * In order to reduce various lock holder preemption latencies provide an
1852 * interface to see if a vCPU is currently running or not.
1854 * This allows us to terminate optimistic spin loops and block, analogous to
1855 * the native optimistic spin heuristic of testing if the lock owner task is
1858 #ifndef vcpu_is_preempted
1859 # define vcpu_is_preempted(cpu) false
1862 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1863 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1865 #ifndef TASK_SIZE_OF
1866 #define TASK_SIZE_OF(tsk) TASK_SIZE
1872 * Map the event mask on the user-space ABI enum rseq_cs_flags
1873 * for direct mask checks.
1875 enum rseq_event_mask_bits {
1876 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1877 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1878 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1881 enum rseq_event_mask {
1882 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1883 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1884 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1887 static inline void rseq_set_notify_resume(struct task_struct *t)
1890 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1893 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1895 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1896 struct pt_regs *regs)
1899 __rseq_handle_notify_resume(ksig, regs);
1902 static inline void rseq_signal_deliver(struct ksignal *ksig,
1903 struct pt_regs *regs)
1906 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1908 rseq_handle_notify_resume(ksig, regs);
1911 /* rseq_preempt() requires preemption to be disabled. */
1912 static inline void rseq_preempt(struct task_struct *t)
1914 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1915 rseq_set_notify_resume(t);
1918 /* rseq_migrate() requires preemption to be disabled. */
1919 static inline void rseq_migrate(struct task_struct *t)
1921 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1922 rseq_set_notify_resume(t);
1926 * If parent process has a registered restartable sequences area, the
1927 * child inherits. Only applies when forking a process, not a thread.
1929 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1931 if (clone_flags & CLONE_THREAD) {
1934 t->rseq_event_mask = 0;
1936 t->rseq = current->rseq;
1937 t->rseq_sig = current->rseq_sig;
1938 t->rseq_event_mask = current->rseq_event_mask;
1942 static inline void rseq_execve(struct task_struct *t)
1946 t->rseq_event_mask = 0;
1951 static inline void rseq_set_notify_resume(struct task_struct *t)
1954 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1955 struct pt_regs *regs)
1958 static inline void rseq_signal_deliver(struct ksignal *ksig,
1959 struct pt_regs *regs)
1962 static inline void rseq_preempt(struct task_struct *t)
1965 static inline void rseq_migrate(struct task_struct *t)
1968 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1971 static inline void rseq_execve(struct task_struct *t)
1977 void __exit_umh(struct task_struct *tsk);
1979 static inline void exit_umh(struct task_struct *tsk)
1981 if (unlikely(tsk->flags & PF_UMH))
1985 #ifdef CONFIG_DEBUG_RSEQ
1987 void rseq_syscall(struct pt_regs *regs);
1991 static inline void rseq_syscall(struct pt_regs *regs)
1997 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
1998 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
1999 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2001 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2002 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2003 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2005 int sched_trace_rq_cpu(struct rq *rq);
2007 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);