1 /* SPDX-License-Identifier: GPL-2.0 */
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
10 #include <uapi/linux/sched.h>
12 #include <asm/current.h>
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/refcount.h>
25 #include <linux/resource.h>
26 #include <linux/latencytop.h>
27 #include <linux/sched/prio.h>
28 #include <linux/sched/types.h>
29 #include <linux/signal_types.h>
30 #include <linux/mm_types_task.h>
31 #include <linux/task_io_accounting.h>
32 #include <linux/posix-timers.h>
33 #include <linux/rseq.h>
34 #include <linux/kcsan.h>
36 /* task_struct member predeclarations (sorted alphabetically): */
38 struct backing_dev_info;
41 struct capture_control;
44 struct futex_pi_state;
49 struct perf_event_context;
51 struct pipe_inode_info;
54 struct robust_list_head;
60 struct sighand_struct;
62 struct task_delay_info;
66 * Task state bitmask. NOTE! These bits are also
67 * encoded in fs/proc/array.c: get_task_state().
69 * We have two separate sets of flags: task->state
70 * is about runnability, while task->exit_state are
71 * about the task exiting. Confusing, but this way
72 * modifying one set can't modify the other one by
76 /* Used in tsk->state: */
77 #define TASK_RUNNING 0x0000
78 #define TASK_INTERRUPTIBLE 0x0001
79 #define TASK_UNINTERRUPTIBLE 0x0002
80 #define __TASK_STOPPED 0x0004
81 #define __TASK_TRACED 0x0008
82 /* Used in tsk->exit_state: */
83 #define EXIT_DEAD 0x0010
84 #define EXIT_ZOMBIE 0x0020
85 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
86 /* Used in tsk->state again: */
87 #define TASK_PARKED 0x0040
88 #define TASK_DEAD 0x0080
89 #define TASK_WAKEKILL 0x0100
90 #define TASK_WAKING 0x0200
91 #define TASK_NOLOAD 0x0400
92 #define TASK_NEW 0x0800
93 #define TASK_STATE_MAX 0x1000
95 /* Convenience macros for the sake of set_current_state: */
96 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
97 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
98 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
100 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
102 /* Convenience macros for the sake of wake_up(): */
103 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
105 /* get_task_state(): */
106 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
107 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
108 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
111 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
113 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
115 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
117 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
118 (task->flags & PF_FROZEN) == 0 && \
119 (task->state & TASK_NOLOAD) == 0)
121 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
124 * Special states are those that do not use the normal wait-loop pattern. See
125 * the comment with set_special_state().
127 #define is_special_task_state(state) \
128 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
130 #define __set_current_state(state_value) \
132 WARN_ON_ONCE(is_special_task_state(state_value));\
133 current->task_state_change = _THIS_IP_; \
134 current->state = (state_value); \
137 #define set_current_state(state_value) \
139 WARN_ON_ONCE(is_special_task_state(state_value));\
140 current->task_state_change = _THIS_IP_; \
141 smp_store_mb(current->state, (state_value)); \
144 #define set_special_state(state_value) \
146 unsigned long flags; /* may shadow */ \
147 WARN_ON_ONCE(!is_special_task_state(state_value)); \
148 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
149 current->task_state_change = _THIS_IP_; \
150 current->state = (state_value); \
151 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
155 * set_current_state() includes a barrier so that the write of current->state
156 * is correctly serialised wrt the caller's subsequent test of whether to
160 * set_current_state(TASK_UNINTERRUPTIBLE);
166 * __set_current_state(TASK_RUNNING);
168 * If the caller does not need such serialisation (because, for instance, the
169 * condition test and condition change and wakeup are under the same lock) then
170 * use __set_current_state().
172 * The above is typically ordered against the wakeup, which does:
174 * need_sleep = false;
175 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
177 * where wake_up_state() executes a full memory barrier before accessing the
180 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
181 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
182 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
184 * However, with slightly different timing the wakeup TASK_RUNNING store can
185 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
186 * a problem either because that will result in one extra go around the loop
187 * and our @cond test will save the day.
189 * Also see the comments of try_to_wake_up().
191 #define __set_current_state(state_value) \
192 current->state = (state_value)
194 #define set_current_state(state_value) \
195 smp_store_mb(current->state, (state_value))
198 * set_special_state() should be used for those states when the blocking task
199 * can not use the regular condition based wait-loop. In that case we must
200 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
201 * will not collide with our state change.
203 #define set_special_state(state_value) \
205 unsigned long flags; /* may shadow */ \
206 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
207 current->state = (state_value); \
208 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
213 /* Task command name length: */
214 #define TASK_COMM_LEN 16
216 extern void scheduler_tick(void);
218 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
220 extern long schedule_timeout(long timeout);
221 extern long schedule_timeout_interruptible(long timeout);
222 extern long schedule_timeout_killable(long timeout);
223 extern long schedule_timeout_uninterruptible(long timeout);
224 extern long schedule_timeout_idle(long timeout);
225 asmlinkage void schedule(void);
226 extern void schedule_preempt_disabled(void);
227 asmlinkage void preempt_schedule_irq(void);
229 extern int __must_check io_schedule_prepare(void);
230 extern void io_schedule_finish(int token);
231 extern long io_schedule_timeout(long timeout);
232 extern void io_schedule(void);
235 * struct prev_cputime - snapshot of system and user cputime
236 * @utime: time spent in user mode
237 * @stime: time spent in system mode
238 * @lock: protects the above two fields
240 * Stores previous user/system time values such that we can guarantee
243 struct prev_cputime {
244 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
252 /* Task is sleeping or running in a CPU with VTIME inactive: */
256 /* Task runs in kernelspace in a CPU with VTIME active: */
258 /* Task runs in userspace in a CPU with VTIME active: */
260 /* Task runs as guests in a CPU with VTIME active: */
266 unsigned long long starttime;
267 enum vtime_state state;
275 * Utilization clamp constraints.
276 * @UCLAMP_MIN: Minimum utilization
277 * @UCLAMP_MAX: Maximum utilization
278 * @UCLAMP_CNT: Utilization clamp constraints count
287 extern struct root_domain def_root_domain;
288 extern struct mutex sched_domains_mutex;
292 #ifdef CONFIG_SCHED_INFO
293 /* Cumulative counters: */
295 /* # of times we have run on this CPU: */
296 unsigned long pcount;
298 /* Time spent waiting on a runqueue: */
299 unsigned long long run_delay;
303 /* When did we last run on a CPU? */
304 unsigned long long last_arrival;
306 /* When were we last queued to run? */
307 unsigned long long last_queued;
309 #endif /* CONFIG_SCHED_INFO */
313 * Integer metrics need fixed point arithmetic, e.g., sched/fair
314 * has a few: load, load_avg, util_avg, freq, and capacity.
316 * We define a basic fixed point arithmetic range, and then formalize
317 * all these metrics based on that basic range.
319 # define SCHED_FIXEDPOINT_SHIFT 10
320 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
322 /* Increase resolution of cpu_capacity calculations */
323 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
324 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
327 unsigned long weight;
332 * struct util_est - Estimation utilization of FAIR tasks
333 * @enqueued: instantaneous estimated utilization of a task/cpu
334 * @ewma: the Exponential Weighted Moving Average (EWMA)
335 * utilization of a task
337 * Support data structure to track an Exponential Weighted Moving Average
338 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
339 * average each time a task completes an activation. Sample's weight is chosen
340 * so that the EWMA will be relatively insensitive to transient changes to the
343 * The enqueued attribute has a slightly different meaning for tasks and cpus:
344 * - task: the task's util_avg at last task dequeue time
345 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
346 * Thus, the util_est.enqueued of a task represents the contribution on the
347 * estimated utilization of the CPU where that task is currently enqueued.
349 * Only for tasks we track a moving average of the past instantaneous
350 * estimated utilization. This allows to absorb sporadic drops in utilization
351 * of an otherwise almost periodic task.
354 unsigned int enqueued;
356 #define UTIL_EST_WEIGHT_SHIFT 2
357 } __attribute__((__aligned__(sizeof(u64))));
360 * The load/runnable/util_avg accumulates an infinite geometric series
361 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
363 * [load_avg definition]
365 * load_avg = runnable% * scale_load_down(load)
367 * [runnable_avg definition]
369 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
371 * [util_avg definition]
373 * util_avg = running% * SCHED_CAPACITY_SCALE
375 * where runnable% is the time ratio that a sched_entity is runnable and
376 * running% the time ratio that a sched_entity is running.
378 * For cfs_rq, they are the aggregated values of all runnable and blocked
381 * The load/runnable/util_avg doesn't direcly factor frequency scaling and CPU
382 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
383 * for computing those signals (see update_rq_clock_pelt())
385 * N.B., the above ratios (runnable% and running%) themselves are in the
386 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
387 * to as large a range as necessary. This is for example reflected by
388 * util_avg's SCHED_CAPACITY_SCALE.
392 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
393 * with the highest load (=88761), always runnable on a single cfs_rq,
394 * and should not overflow as the number already hits PID_MAX_LIMIT.
396 * For all other cases (including 32-bit kernels), struct load_weight's
397 * weight will overflow first before we do, because:
399 * Max(load_avg) <= Max(load.weight)
401 * Then it is the load_weight's responsibility to consider overflow
405 u64 last_update_time;
410 unsigned long load_avg;
411 unsigned long runnable_avg;
412 unsigned long util_avg;
413 struct util_est util_est;
414 } ____cacheline_aligned;
416 struct sched_statistics {
417 #ifdef CONFIG_SCHEDSTATS
427 s64 sum_sleep_runtime;
434 u64 nr_migrations_cold;
435 u64 nr_failed_migrations_affine;
436 u64 nr_failed_migrations_running;
437 u64 nr_failed_migrations_hot;
438 u64 nr_forced_migrations;
442 u64 nr_wakeups_migrate;
443 u64 nr_wakeups_local;
444 u64 nr_wakeups_remote;
445 u64 nr_wakeups_affine;
446 u64 nr_wakeups_affine_attempts;
447 u64 nr_wakeups_passive;
452 struct sched_entity {
453 /* For load-balancing: */
454 struct load_weight load;
455 struct rb_node run_node;
456 struct list_head group_node;
460 u64 sum_exec_runtime;
462 u64 prev_sum_exec_runtime;
466 struct sched_statistics statistics;
468 #ifdef CONFIG_FAIR_GROUP_SCHED
470 struct sched_entity *parent;
471 /* rq on which this entity is (to be) queued: */
472 struct cfs_rq *cfs_rq;
473 /* rq "owned" by this entity/group: */
475 /* cached value of my_q->h_nr_running */
476 unsigned long runnable_weight;
481 * Per entity load average tracking.
483 * Put into separate cache line so it does not
484 * collide with read-mostly values above.
486 struct sched_avg avg;
490 struct sched_rt_entity {
491 struct list_head run_list;
492 unsigned long timeout;
493 unsigned long watchdog_stamp;
494 unsigned int time_slice;
495 unsigned short on_rq;
496 unsigned short on_list;
498 struct sched_rt_entity *back;
499 #ifdef CONFIG_RT_GROUP_SCHED
500 struct sched_rt_entity *parent;
501 /* rq on which this entity is (to be) queued: */
503 /* rq "owned" by this entity/group: */
506 } __randomize_layout;
508 struct sched_dl_entity {
509 struct rb_node rb_node;
512 * Original scheduling parameters. Copied here from sched_attr
513 * during sched_setattr(), they will remain the same until
514 * the next sched_setattr().
516 u64 dl_runtime; /* Maximum runtime for each instance */
517 u64 dl_deadline; /* Relative deadline of each instance */
518 u64 dl_period; /* Separation of two instances (period) */
519 u64 dl_bw; /* dl_runtime / dl_period */
520 u64 dl_density; /* dl_runtime / dl_deadline */
523 * Actual scheduling parameters. Initialized with the values above,
524 * they are continuously updated during task execution. Note that
525 * the remaining runtime could be < 0 in case we are in overrun.
527 s64 runtime; /* Remaining runtime for this instance */
528 u64 deadline; /* Absolute deadline for this instance */
529 unsigned int flags; /* Specifying the scheduler behaviour */
534 * @dl_throttled tells if we exhausted the runtime. If so, the
535 * task has to wait for a replenishment to be performed at the
536 * next firing of dl_timer.
538 * @dl_boosted tells if we are boosted due to DI. If so we are
539 * outside bandwidth enforcement mechanism (but only until we
540 * exit the critical section);
542 * @dl_yielded tells if task gave up the CPU before consuming
543 * all its available runtime during the last job.
545 * @dl_non_contending tells if the task is inactive while still
546 * contributing to the active utilization. In other words, it
547 * indicates if the inactive timer has been armed and its handler
548 * has not been executed yet. This flag is useful to avoid race
549 * conditions between the inactive timer handler and the wakeup
552 * @dl_overrun tells if the task asked to be informed about runtime
555 unsigned int dl_throttled : 1;
556 unsigned int dl_boosted : 1;
557 unsigned int dl_yielded : 1;
558 unsigned int dl_non_contending : 1;
559 unsigned int dl_overrun : 1;
562 * Bandwidth enforcement timer. Each -deadline task has its
563 * own bandwidth to be enforced, thus we need one timer per task.
565 struct hrtimer dl_timer;
568 * Inactive timer, responsible for decreasing the active utilization
569 * at the "0-lag time". When a -deadline task blocks, it contributes
570 * to GRUB's active utilization until the "0-lag time", hence a
571 * timer is needed to decrease the active utilization at the correct
574 struct hrtimer inactive_timer;
577 #ifdef CONFIG_UCLAMP_TASK
578 /* Number of utilization clamp buckets (shorter alias) */
579 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
582 * Utilization clamp for a scheduling entity
583 * @value: clamp value "assigned" to a se
584 * @bucket_id: bucket index corresponding to the "assigned" value
585 * @active: the se is currently refcounted in a rq's bucket
586 * @user_defined: the requested clamp value comes from user-space
588 * The bucket_id is the index of the clamp bucket matching the clamp value
589 * which is pre-computed and stored to avoid expensive integer divisions from
592 * The active bit is set whenever a task has got an "effective" value assigned,
593 * which can be different from the clamp value "requested" from user-space.
594 * This allows to know a task is refcounted in the rq's bucket corresponding
595 * to the "effective" bucket_id.
597 * The user_defined bit is set whenever a task has got a task-specific clamp
598 * value requested from userspace, i.e. the system defaults apply to this task
599 * just as a restriction. This allows to relax default clamps when a less
600 * restrictive task-specific value has been requested, thus allowing to
601 * implement a "nice" semantic. For example, a task running with a 20%
602 * default boost can still drop its own boosting to 0%.
605 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
606 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
607 unsigned int active : 1;
608 unsigned int user_defined : 1;
610 #endif /* CONFIG_UCLAMP_TASK */
616 u8 exp_hint; /* Hint for performance. */
617 u8 need_mb; /* Readers need smp_mb(). */
619 u32 s; /* Set of bits. */
622 enum perf_event_task_context {
623 perf_invalid_context = -1,
626 perf_nr_task_contexts,
630 struct wake_q_node *next;
634 #ifdef CONFIG_THREAD_INFO_IN_TASK
636 * For reasons of header soup (see current_thread_info()), this
637 * must be the first element of task_struct.
639 struct thread_info thread_info;
641 /* -1 unrunnable, 0 runnable, >0 stopped: */
645 * This begins the randomizable portion of task_struct. Only
646 * scheduling-critical items should be added above here.
648 randomized_struct_fields_start
652 /* Per task flags (PF_*), defined further below: */
657 struct llist_node wake_entry;
658 unsigned int wake_entry_type;
660 #ifdef CONFIG_THREAD_INFO_IN_TASK
664 unsigned int wakee_flips;
665 unsigned long wakee_flip_decay_ts;
666 struct task_struct *last_wakee;
669 * recent_used_cpu is initially set as the last CPU used by a task
670 * that wakes affine another task. Waker/wakee relationships can
671 * push tasks around a CPU where each wakeup moves to the next one.
672 * Tracking a recently used CPU allows a quick search for a recently
673 * used CPU that may be idle.
683 unsigned int rt_priority;
685 const struct sched_class *sched_class;
686 struct sched_entity se;
687 struct sched_rt_entity rt;
688 #ifdef CONFIG_CGROUP_SCHED
689 struct task_group *sched_task_group;
691 struct sched_dl_entity dl;
693 #ifdef CONFIG_UCLAMP_TASK
694 /* Clamp values requested for a scheduling entity */
695 struct uclamp_se uclamp_req[UCLAMP_CNT];
696 /* Effective clamp values used for a scheduling entity */
697 struct uclamp_se uclamp[UCLAMP_CNT];
700 #ifdef CONFIG_PREEMPT_NOTIFIERS
701 /* List of struct preempt_notifier: */
702 struct hlist_head preempt_notifiers;
705 #ifdef CONFIG_BLK_DEV_IO_TRACE
706 unsigned int btrace_seq;
711 const cpumask_t *cpus_ptr;
714 #ifdef CONFIG_PREEMPT_RCU
715 int rcu_read_lock_nesting;
716 union rcu_special rcu_read_unlock_special;
717 struct list_head rcu_node_entry;
718 struct rcu_node *rcu_blocked_node;
719 #endif /* #ifdef CONFIG_PREEMPT_RCU */
721 #ifdef CONFIG_TASKS_RCU
722 unsigned long rcu_tasks_nvcsw;
723 u8 rcu_tasks_holdout;
725 int rcu_tasks_idle_cpu;
726 struct list_head rcu_tasks_holdout_list;
727 #endif /* #ifdef CONFIG_TASKS_RCU */
729 #ifdef CONFIG_TASKS_TRACE_RCU
730 int trc_reader_nesting;
732 union rcu_special trc_reader_special;
733 bool trc_reader_checked;
734 struct list_head trc_holdout_list;
735 #endif /* #ifdef CONFIG_TASKS_TRACE_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 unsigned use_memdelay:1;
801 /* Stalled due to lack of memory */
802 unsigned in_memstall:1;
805 unsigned long atomic_flags; /* Flags requiring atomic access. */
807 struct restart_block restart_block;
812 #ifdef CONFIG_STACKPROTECTOR
813 /* Canary value for the -fstack-protector GCC feature: */
814 unsigned long stack_canary;
817 * Pointers to the (original) parent process, youngest child, younger sibling,
818 * older sibling, respectively. (p->father can be replaced with
819 * p->real_parent->pid)
822 /* Real parent process: */
823 struct task_struct __rcu *real_parent;
825 /* Recipient of SIGCHLD, wait4() reports: */
826 struct task_struct __rcu *parent;
829 * Children/sibling form the list of natural children:
831 struct list_head children;
832 struct list_head sibling;
833 struct task_struct *group_leader;
836 * 'ptraced' is the list of tasks this task is using ptrace() on.
838 * This includes both natural children and PTRACE_ATTACH targets.
839 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
841 struct list_head ptraced;
842 struct list_head ptrace_entry;
844 /* PID/PID hash table linkage. */
845 struct pid *thread_pid;
846 struct hlist_node pid_links[PIDTYPE_MAX];
847 struct list_head thread_group;
848 struct list_head thread_node;
850 struct completion *vfork_done;
852 /* CLONE_CHILD_SETTID: */
853 int __user *set_child_tid;
855 /* CLONE_CHILD_CLEARTID: */
856 int __user *clear_child_tid;
860 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
865 struct prev_cputime prev_cputime;
866 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
870 #ifdef CONFIG_NO_HZ_FULL
871 atomic_t tick_dep_mask;
873 /* Context switch counts: */
875 unsigned long nivcsw;
877 /* Monotonic time in nsecs: */
880 /* Boot based time in nsecs: */
883 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
884 unsigned long min_flt;
885 unsigned long maj_flt;
887 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
888 struct posix_cputimers posix_cputimers;
890 /* Process credentials: */
892 /* Tracer's credentials at attach: */
893 const struct cred __rcu *ptracer_cred;
895 /* Objective and real subjective task credentials (COW): */
896 const struct cred __rcu *real_cred;
898 /* Effective (overridable) subjective task credentials (COW): */
899 const struct cred __rcu *cred;
902 /* Cached requested key. */
903 struct key *cached_requested_key;
907 * executable name, excluding path.
909 * - normally initialized setup_new_exec()
910 * - access it with [gs]et_task_comm()
911 * - lock it with task_lock()
913 char comm[TASK_COMM_LEN];
915 struct nameidata *nameidata;
917 #ifdef CONFIG_SYSVIPC
918 struct sysv_sem sysvsem;
919 struct sysv_shm sysvshm;
921 #ifdef CONFIG_DETECT_HUNG_TASK
922 unsigned long last_switch_count;
923 unsigned long last_switch_time;
925 /* Filesystem information: */
926 struct fs_struct *fs;
928 /* Open file information: */
929 struct files_struct *files;
932 struct nsproxy *nsproxy;
934 /* Signal handlers: */
935 struct signal_struct *signal;
936 struct sighand_struct __rcu *sighand;
938 sigset_t real_blocked;
939 /* Restored if set_restore_sigmask() was used: */
940 sigset_t saved_sigmask;
941 struct sigpending pending;
942 unsigned long sas_ss_sp;
944 unsigned int sas_ss_flags;
946 struct callback_head *task_works;
949 #ifdef CONFIG_AUDITSYSCALL
950 struct audit_context *audit_context;
953 unsigned int sessionid;
955 struct seccomp seccomp;
957 /* Thread group tracking: */
961 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
962 spinlock_t alloc_lock;
964 /* Protection of the PI data structures: */
965 raw_spinlock_t pi_lock;
967 struct wake_q_node wake_q;
969 #ifdef CONFIG_RT_MUTEXES
970 /* PI waiters blocked on a rt_mutex held by this task: */
971 struct rb_root_cached pi_waiters;
972 /* Updated under owner's pi_lock and rq lock */
973 struct task_struct *pi_top_task;
974 /* Deadlock detection and priority inheritance handling: */
975 struct rt_mutex_waiter *pi_blocked_on;
978 #ifdef CONFIG_DEBUG_MUTEXES
979 /* Mutex deadlock detection: */
980 struct mutex_waiter *blocked_on;
983 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
987 #ifdef CONFIG_TRACE_IRQFLAGS
988 unsigned int irq_events;
989 unsigned int hardirq_threaded;
990 unsigned long hardirq_enable_ip;
991 unsigned long hardirq_disable_ip;
992 unsigned int hardirq_enable_event;
993 unsigned int hardirq_disable_event;
994 int hardirqs_enabled;
996 u64 hardirq_chain_key;
997 unsigned long softirq_disable_ip;
998 unsigned long softirq_enable_ip;
999 unsigned int softirq_disable_event;
1000 unsigned int softirq_enable_event;
1001 int softirqs_enabled;
1002 int softirq_context;
1006 #ifdef CONFIG_LOCKDEP
1007 # define MAX_LOCK_DEPTH 48UL
1010 unsigned int lockdep_recursion;
1011 struct held_lock held_locks[MAX_LOCK_DEPTH];
1015 unsigned int in_ubsan;
1018 /* Journalling filesystem info: */
1021 /* Stacked block device info: */
1022 struct bio_list *bio_list;
1025 /* Stack plugging: */
1026 struct blk_plug *plug;
1030 struct reclaim_state *reclaim_state;
1032 struct backing_dev_info *backing_dev_info;
1034 struct io_context *io_context;
1036 #ifdef CONFIG_COMPACTION
1037 struct capture_control *capture_control;
1040 unsigned long ptrace_message;
1041 kernel_siginfo_t *last_siginfo;
1043 struct task_io_accounting ioac;
1045 /* Pressure stall state */
1046 unsigned int psi_flags;
1048 #ifdef CONFIG_TASK_XACCT
1049 /* Accumulated RSS usage: */
1051 /* Accumulated virtual memory usage: */
1053 /* stime + utime since last update: */
1056 #ifdef CONFIG_CPUSETS
1057 /* Protected by ->alloc_lock: */
1058 nodemask_t mems_allowed;
1059 /* Seqence number to catch updates: */
1060 seqcount_t mems_allowed_seq;
1061 int cpuset_mem_spread_rotor;
1062 int cpuset_slab_spread_rotor;
1064 #ifdef CONFIG_CGROUPS
1065 /* Control Group info protected by css_set_lock: */
1066 struct css_set __rcu *cgroups;
1067 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1068 struct list_head cg_list;
1070 #ifdef CONFIG_X86_CPU_RESCTRL
1075 struct robust_list_head __user *robust_list;
1076 #ifdef CONFIG_COMPAT
1077 struct compat_robust_list_head __user *compat_robust_list;
1079 struct list_head pi_state_list;
1080 struct futex_pi_state *pi_state_cache;
1081 struct mutex futex_exit_mutex;
1082 unsigned int futex_state;
1084 #ifdef CONFIG_PERF_EVENTS
1085 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1086 struct mutex perf_event_mutex;
1087 struct list_head perf_event_list;
1089 #ifdef CONFIG_DEBUG_PREEMPT
1090 unsigned long preempt_disable_ip;
1093 /* Protected by alloc_lock: */
1094 struct mempolicy *mempolicy;
1096 short pref_node_fork;
1098 #ifdef CONFIG_NUMA_BALANCING
1100 unsigned int numa_scan_period;
1101 unsigned int numa_scan_period_max;
1102 int numa_preferred_nid;
1103 unsigned long numa_migrate_retry;
1104 /* Migration stamp: */
1106 u64 last_task_numa_placement;
1107 u64 last_sum_exec_runtime;
1108 struct callback_head numa_work;
1111 * This pointer is only modified for current in syscall and
1112 * pagefault context (and for tasks being destroyed), so it can be read
1113 * from any of the following contexts:
1114 * - RCU read-side critical section
1115 * - current->numa_group from everywhere
1116 * - task's runqueue locked, task not running
1118 struct numa_group __rcu *numa_group;
1121 * numa_faults is an array split into four regions:
1122 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1123 * in this precise order.
1125 * faults_memory: Exponential decaying average of faults on a per-node
1126 * basis. Scheduling placement decisions are made based on these
1127 * counts. The values remain static for the duration of a PTE scan.
1128 * faults_cpu: Track the nodes the process was running on when a NUMA
1129 * hinting fault was incurred.
1130 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1131 * during the current scan window. When the scan completes, the counts
1132 * in faults_memory and faults_cpu decay and these values are copied.
1134 unsigned long *numa_faults;
1135 unsigned long total_numa_faults;
1138 * numa_faults_locality tracks if faults recorded during the last
1139 * scan window were remote/local or failed to migrate. The task scan
1140 * period is adapted based on the locality of the faults with different
1141 * weights depending on whether they were shared or private faults
1143 unsigned long numa_faults_locality[3];
1145 unsigned long numa_pages_migrated;
1146 #endif /* CONFIG_NUMA_BALANCING */
1149 struct rseq __user *rseq;
1152 * RmW on rseq_event_mask must be performed atomically
1153 * with respect to preemption.
1155 unsigned long rseq_event_mask;
1158 struct tlbflush_unmap_batch tlb_ubc;
1161 refcount_t rcu_users;
1162 struct rcu_head rcu;
1165 /* Cache last used pipe for splice(): */
1166 struct pipe_inode_info *splice_pipe;
1168 struct page_frag task_frag;
1170 #ifdef CONFIG_TASK_DELAY_ACCT
1171 struct task_delay_info *delays;
1174 #ifdef CONFIG_FAULT_INJECTION
1176 unsigned int fail_nth;
1179 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1180 * balance_dirty_pages() for a dirty throttling pause:
1183 int nr_dirtied_pause;
1184 /* Start of a write-and-pause period: */
1185 unsigned long dirty_paused_when;
1187 #ifdef CONFIG_LATENCYTOP
1188 int latency_record_count;
1189 struct latency_record latency_record[LT_SAVECOUNT];
1192 * Time slack values; these are used to round up poll() and
1193 * select() etc timeout values. These are in nanoseconds.
1196 u64 default_timer_slack_ns;
1199 unsigned int kasan_depth;
1202 struct kcsan_ctx kcsan_ctx;
1205 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1206 /* Index of current stored address in ret_stack: */
1210 /* Stack of return addresses for return function tracing: */
1211 struct ftrace_ret_stack *ret_stack;
1213 /* Timestamp for last schedule: */
1214 unsigned long long ftrace_timestamp;
1217 * Number of functions that haven't been traced
1218 * because of depth overrun:
1220 atomic_t trace_overrun;
1222 /* Pause tracing: */
1223 atomic_t tracing_graph_pause;
1226 #ifdef CONFIG_TRACING
1227 /* State flags for use by tracers: */
1228 unsigned long trace;
1230 /* Bitmask and counter of trace recursion: */
1231 unsigned long trace_recursion;
1232 #endif /* CONFIG_TRACING */
1235 /* See kernel/kcov.c for more details. */
1237 /* Coverage collection mode enabled for this task (0 if disabled): */
1238 unsigned int kcov_mode;
1240 /* Size of the kcov_area: */
1241 unsigned int kcov_size;
1243 /* Buffer for coverage collection: */
1246 /* KCOV descriptor wired with this task or NULL: */
1249 /* KCOV common handle for remote coverage collection: */
1252 /* KCOV sequence number: */
1255 /* Collect coverage from softirq context: */
1256 unsigned int kcov_softirq;
1260 struct mem_cgroup *memcg_in_oom;
1261 gfp_t memcg_oom_gfp_mask;
1262 int memcg_oom_order;
1264 /* Number of pages to reclaim on returning to userland: */
1265 unsigned int memcg_nr_pages_over_high;
1267 /* Used by memcontrol for targeted memcg charge: */
1268 struct mem_cgroup *active_memcg;
1271 #ifdef CONFIG_BLK_CGROUP
1272 struct request_queue *throttle_queue;
1275 #ifdef CONFIG_UPROBES
1276 struct uprobe_task *utask;
1278 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1279 unsigned int sequential_io;
1280 unsigned int sequential_io_avg;
1282 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1283 unsigned long task_state_change;
1285 int pagefault_disabled;
1287 struct task_struct *oom_reaper_list;
1289 #ifdef CONFIG_VMAP_STACK
1290 struct vm_struct *stack_vm_area;
1292 #ifdef CONFIG_THREAD_INFO_IN_TASK
1293 /* A live task holds one reference: */
1294 refcount_t stack_refcount;
1296 #ifdef CONFIG_LIVEPATCH
1299 #ifdef CONFIG_SECURITY
1300 /* Used by LSM modules for access restriction: */
1304 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1305 unsigned long lowest_stack;
1306 unsigned long prev_lowest_stack;
1309 #ifdef CONFIG_X86_MCE
1313 __mce_reserved : 62;
1314 struct callback_head mce_kill_me;
1318 * New fields for task_struct should be added above here, so that
1319 * they are included in the randomized portion of task_struct.
1321 randomized_struct_fields_end
1323 /* CPU-specific state of this task: */
1324 struct thread_struct thread;
1327 * WARNING: on x86, 'thread_struct' contains a variable-sized
1328 * structure. It *MUST* be at the end of 'task_struct'.
1330 * Do not put anything below here!
1334 static inline struct pid *task_pid(struct task_struct *task)
1336 return task->thread_pid;
1340 * the helpers to get the task's different pids as they are seen
1341 * from various namespaces
1343 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1344 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1346 * task_xid_nr_ns() : id seen from the ns specified;
1348 * see also pid_nr() etc in include/linux/pid.h
1350 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1352 static inline pid_t task_pid_nr(struct task_struct *tsk)
1357 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1359 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1362 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1364 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1368 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1374 * pid_alive - check that a task structure is not stale
1375 * @p: Task structure to be checked.
1377 * Test if a process is not yet dead (at most zombie state)
1378 * If pid_alive fails, then pointers within the task structure
1379 * can be stale and must not be dereferenced.
1381 * Return: 1 if the process is alive. 0 otherwise.
1383 static inline int pid_alive(const struct task_struct *p)
1385 return p->thread_pid != NULL;
1388 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1390 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1393 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1395 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1399 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1401 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1404 static inline pid_t task_session_vnr(struct task_struct *tsk)
1406 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1409 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1411 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1414 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1416 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1419 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1425 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1431 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1433 return task_ppid_nr_ns(tsk, &init_pid_ns);
1436 /* Obsolete, do not use: */
1437 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1439 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1442 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1443 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1445 static inline unsigned int task_state_index(struct task_struct *tsk)
1447 unsigned int tsk_state = READ_ONCE(tsk->state);
1448 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1450 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1452 if (tsk_state == TASK_IDLE)
1453 state = TASK_REPORT_IDLE;
1458 static inline char task_index_to_char(unsigned int state)
1460 static const char state_char[] = "RSDTtXZPI";
1462 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1464 return state_char[state];
1467 static inline char task_state_to_char(struct task_struct *tsk)
1469 return task_index_to_char(task_state_index(tsk));
1473 * is_global_init - check if a task structure is init. Since init
1474 * is free to have sub-threads we need to check tgid.
1475 * @tsk: Task structure to be checked.
1477 * Check if a task structure is the first user space task the kernel created.
1479 * Return: 1 if the task structure is init. 0 otherwise.
1481 static inline int is_global_init(struct task_struct *tsk)
1483 return task_tgid_nr(tsk) == 1;
1486 extern struct pid *cad_pid;
1491 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1492 #define PF_EXITING 0x00000004 /* Getting shut down */
1493 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1494 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1495 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1496 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1497 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1498 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1499 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1500 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1501 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1502 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1503 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1504 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1505 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1506 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1507 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1508 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1509 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1510 * I am cleaning dirty pages from some other bdi. */
1511 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1512 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1513 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1514 #define PF_UMH 0x02000000 /* I'm an Usermodehelper process */
1515 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1516 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1517 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1518 #define PF_IO_WORKER 0x20000000 /* Task is an IO worker */
1519 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1520 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1523 * Only the _current_ task can read/write to tsk->flags, but other
1524 * tasks can access tsk->flags in readonly mode for example
1525 * with tsk_used_math (like during threaded core dumping).
1526 * There is however an exception to this rule during ptrace
1527 * or during fork: the ptracer task is allowed to write to the
1528 * child->flags of its traced child (same goes for fork, the parent
1529 * can write to the child->flags), because we're guaranteed the
1530 * child is not running and in turn not changing child->flags
1531 * at the same time the parent does it.
1533 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1534 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1535 #define clear_used_math() clear_stopped_child_used_math(current)
1536 #define set_used_math() set_stopped_child_used_math(current)
1538 #define conditional_stopped_child_used_math(condition, child) \
1539 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1541 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1543 #define copy_to_stopped_child_used_math(child) \
1544 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1546 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1547 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1548 #define used_math() tsk_used_math(current)
1550 static inline bool is_percpu_thread(void)
1553 return (current->flags & PF_NO_SETAFFINITY) &&
1554 (current->nr_cpus_allowed == 1);
1560 /* Per-process atomic flags. */
1561 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1562 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1563 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1564 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1565 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1566 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1567 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1568 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1570 #define TASK_PFA_TEST(name, func) \
1571 static inline bool task_##func(struct task_struct *p) \
1572 { return test_bit(PFA_##name, &p->atomic_flags); }
1574 #define TASK_PFA_SET(name, func) \
1575 static inline void task_set_##func(struct task_struct *p) \
1576 { set_bit(PFA_##name, &p->atomic_flags); }
1578 #define TASK_PFA_CLEAR(name, func) \
1579 static inline void task_clear_##func(struct task_struct *p) \
1580 { clear_bit(PFA_##name, &p->atomic_flags); }
1582 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1583 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1585 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1586 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1587 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1589 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1590 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1591 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1593 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1594 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1595 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1597 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1598 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1599 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1601 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1602 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1604 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1605 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1606 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1608 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1609 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1612 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1614 current->flags &= ~flags;
1615 current->flags |= orig_flags & flags;
1618 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1619 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1621 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1622 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1624 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1627 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1629 if (!cpumask_test_cpu(0, new_mask))
1635 extern int yield_to(struct task_struct *p, bool preempt);
1636 extern void set_user_nice(struct task_struct *p, long nice);
1637 extern int task_prio(const struct task_struct *p);
1640 * task_nice - return the nice value of a given task.
1641 * @p: the task in question.
1643 * Return: The nice value [ -20 ... 0 ... 19 ].
1645 static inline int task_nice(const struct task_struct *p)
1647 return PRIO_TO_NICE((p)->static_prio);
1650 extern int can_nice(const struct task_struct *p, const int nice);
1651 extern int task_curr(const struct task_struct *p);
1652 extern int idle_cpu(int cpu);
1653 extern int available_idle_cpu(int cpu);
1654 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1655 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1656 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1657 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1658 extern struct task_struct *idle_task(int cpu);
1661 * is_idle_task - is the specified task an idle task?
1662 * @p: the task in question.
1664 * Return: 1 if @p is an idle task. 0 otherwise.
1666 static inline bool is_idle_task(const struct task_struct *p)
1668 return !!(p->flags & PF_IDLE);
1671 extern struct task_struct *curr_task(int cpu);
1672 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1676 union thread_union {
1677 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1678 struct task_struct task;
1680 #ifndef CONFIG_THREAD_INFO_IN_TASK
1681 struct thread_info thread_info;
1683 unsigned long stack[THREAD_SIZE/sizeof(long)];
1686 #ifndef CONFIG_THREAD_INFO_IN_TASK
1687 extern struct thread_info init_thread_info;
1690 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1692 #ifdef CONFIG_THREAD_INFO_IN_TASK
1693 static inline struct thread_info *task_thread_info(struct task_struct *task)
1695 return &task->thread_info;
1697 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1698 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1702 * find a task by one of its numerical ids
1704 * find_task_by_pid_ns():
1705 * finds a task by its pid in the specified namespace
1706 * find_task_by_vpid():
1707 * finds a task by its virtual pid
1709 * see also find_vpid() etc in include/linux/pid.h
1712 extern struct task_struct *find_task_by_vpid(pid_t nr);
1713 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1716 * find a task by its virtual pid and get the task struct
1718 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1720 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1721 extern int wake_up_process(struct task_struct *tsk);
1722 extern void wake_up_new_task(struct task_struct *tsk);
1725 extern void kick_process(struct task_struct *tsk);
1727 static inline void kick_process(struct task_struct *tsk) { }
1730 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1732 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1734 __set_task_comm(tsk, from, false);
1737 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1738 #define get_task_comm(buf, tsk) ({ \
1739 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1740 __get_task_comm(buf, sizeof(buf), tsk); \
1744 static __always_inline void scheduler_ipi(void)
1747 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1748 * TIF_NEED_RESCHED remotely (for the first time) will also send
1751 preempt_fold_need_resched();
1753 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1755 static inline void scheduler_ipi(void) { }
1756 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1763 * Set thread flags in other task's structures.
1764 * See asm/thread_info.h for TIF_xxxx flags available:
1766 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1768 set_ti_thread_flag(task_thread_info(tsk), flag);
1771 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1773 clear_ti_thread_flag(task_thread_info(tsk), flag);
1776 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1779 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1782 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1784 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1787 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1789 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1792 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1794 return test_ti_thread_flag(task_thread_info(tsk), flag);
1797 static inline void set_tsk_need_resched(struct task_struct *tsk)
1799 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1802 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1804 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1807 static inline int test_tsk_need_resched(struct task_struct *tsk)
1809 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1813 * cond_resched() and cond_resched_lock(): latency reduction via
1814 * explicit rescheduling in places that are safe. The return
1815 * value indicates whether a reschedule was done in fact.
1816 * cond_resched_lock() will drop the spinlock before scheduling,
1818 #ifndef CONFIG_PREEMPTION
1819 extern int _cond_resched(void);
1821 static inline int _cond_resched(void) { return 0; }
1824 #define cond_resched() ({ \
1825 ___might_sleep(__FILE__, __LINE__, 0); \
1829 extern int __cond_resched_lock(spinlock_t *lock);
1831 #define cond_resched_lock(lock) ({ \
1832 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1833 __cond_resched_lock(lock); \
1836 static inline void cond_resched_rcu(void)
1838 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1846 * Does a critical section need to be broken due to another
1847 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1848 * but a general need for low latency)
1850 static inline int spin_needbreak(spinlock_t *lock)
1852 #ifdef CONFIG_PREEMPTION
1853 return spin_is_contended(lock);
1859 static __always_inline bool need_resched(void)
1861 return unlikely(tif_need_resched());
1865 * Wrappers for p->thread_info->cpu access. No-op on UP.
1869 static inline unsigned int task_cpu(const struct task_struct *p)
1871 #ifdef CONFIG_THREAD_INFO_IN_TASK
1872 return READ_ONCE(p->cpu);
1874 return READ_ONCE(task_thread_info(p)->cpu);
1878 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1882 static inline unsigned int task_cpu(const struct task_struct *p)
1887 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1891 #endif /* CONFIG_SMP */
1894 * In order to reduce various lock holder preemption latencies provide an
1895 * interface to see if a vCPU is currently running or not.
1897 * This allows us to terminate optimistic spin loops and block, analogous to
1898 * the native optimistic spin heuristic of testing if the lock owner task is
1901 #ifndef vcpu_is_preempted
1902 static inline bool vcpu_is_preempted(int cpu)
1908 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1909 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1911 #ifndef TASK_SIZE_OF
1912 #define TASK_SIZE_OF(tsk) TASK_SIZE
1918 * Map the event mask on the user-space ABI enum rseq_cs_flags
1919 * for direct mask checks.
1921 enum rseq_event_mask_bits {
1922 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1923 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1924 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1927 enum rseq_event_mask {
1928 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1929 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1930 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1933 static inline void rseq_set_notify_resume(struct task_struct *t)
1936 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1939 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1941 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1942 struct pt_regs *regs)
1945 __rseq_handle_notify_resume(ksig, regs);
1948 static inline void rseq_signal_deliver(struct ksignal *ksig,
1949 struct pt_regs *regs)
1952 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1954 rseq_handle_notify_resume(ksig, regs);
1957 /* rseq_preempt() requires preemption to be disabled. */
1958 static inline void rseq_preempt(struct task_struct *t)
1960 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1961 rseq_set_notify_resume(t);
1964 /* rseq_migrate() requires preemption to be disabled. */
1965 static inline void rseq_migrate(struct task_struct *t)
1967 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1968 rseq_set_notify_resume(t);
1972 * If parent process has a registered restartable sequences area, the
1973 * child inherits. Unregister rseq for a clone with CLONE_VM set.
1975 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1977 if (clone_flags & CLONE_VM) {
1980 t->rseq_event_mask = 0;
1982 t->rseq = current->rseq;
1983 t->rseq_sig = current->rseq_sig;
1984 t->rseq_event_mask = current->rseq_event_mask;
1988 static inline void rseq_execve(struct task_struct *t)
1992 t->rseq_event_mask = 0;
1997 static inline void rseq_set_notify_resume(struct task_struct *t)
2000 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2001 struct pt_regs *regs)
2004 static inline void rseq_signal_deliver(struct ksignal *ksig,
2005 struct pt_regs *regs)
2008 static inline void rseq_preempt(struct task_struct *t)
2011 static inline void rseq_migrate(struct task_struct *t)
2014 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2017 static inline void rseq_execve(struct task_struct *t)
2023 void __exit_umh(struct task_struct *tsk);
2025 static inline void exit_umh(struct task_struct *tsk)
2027 if (unlikely(tsk->flags & PF_UMH))
2031 #ifdef CONFIG_DEBUG_RSEQ
2033 void rseq_syscall(struct pt_regs *regs);
2037 static inline void rseq_syscall(struct pt_regs *regs)
2043 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2044 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2045 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2047 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2048 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2049 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2051 int sched_trace_rq_cpu(struct rq *rq);
2053 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);