1 /* SPDX-License-Identifier: GPL-2.0 */
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
10 #include <uapi/linux/sched.h>
12 #include <asm/current.h>
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/irqflags.h>
22 #include <linux/seccomp.h>
23 #include <linux/nodemask.h>
24 #include <linux/rcupdate.h>
25 #include <linux/refcount.h>
26 #include <linux/resource.h>
27 #include <linux/latencytop.h>
28 #include <linux/sched/prio.h>
29 #include <linux/sched/types.h>
30 #include <linux/signal_types.h>
31 #include <linux/syscall_user_dispatch.h>
32 #include <linux/mm_types_task.h>
33 #include <linux/task_io_accounting.h>
34 #include <linux/posix-timers.h>
35 #include <linux/rseq.h>
36 #include <linux/seqlock.h>
37 #include <linux/kcsan.h>
39 /* task_struct member predeclarations (sorted alphabetically): */
41 struct backing_dev_info;
44 struct capture_control;
47 struct futex_pi_state;
52 struct perf_event_context;
54 struct pipe_inode_info;
57 struct robust_list_head;
63 struct sighand_struct;
65 struct task_delay_info;
70 * Task state bitmask. NOTE! These bits are also
71 * encoded in fs/proc/array.c: get_task_state().
73 * We have two separate sets of flags: task->state
74 * is about runnability, while task->exit_state are
75 * about the task exiting. Confusing, but this way
76 * modifying one set can't modify the other one by
80 /* Used in tsk->state: */
81 #define TASK_RUNNING 0x0000
82 #define TASK_INTERRUPTIBLE 0x0001
83 #define TASK_UNINTERRUPTIBLE 0x0002
84 #define __TASK_STOPPED 0x0004
85 #define __TASK_TRACED 0x0008
86 /* Used in tsk->exit_state: */
87 #define EXIT_DEAD 0x0010
88 #define EXIT_ZOMBIE 0x0020
89 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
90 /* Used in tsk->state again: */
91 #define TASK_PARKED 0x0040
92 #define TASK_DEAD 0x0080
93 #define TASK_WAKEKILL 0x0100
94 #define TASK_WAKING 0x0200
95 #define TASK_NOLOAD 0x0400
96 #define TASK_NEW 0x0800
97 #define TASK_STATE_MAX 0x1000
99 /* Convenience macros for the sake of set_current_state: */
100 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
101 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
102 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
104 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
106 /* Convenience macros for the sake of wake_up(): */
107 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
109 /* get_task_state(): */
110 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
111 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
112 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
115 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
117 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
119 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 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:
175 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
177 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
178 * accessing p->state.
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 directly 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_yielded : 1;
557 unsigned int dl_non_contending : 1;
558 unsigned int dl_overrun : 1;
561 * Bandwidth enforcement timer. Each -deadline task has its
562 * own bandwidth to be enforced, thus we need one timer per task.
564 struct hrtimer dl_timer;
567 * Inactive timer, responsible for decreasing the active utilization
568 * at the "0-lag time". When a -deadline task blocks, it contributes
569 * to GRUB's active utilization until the "0-lag time", hence a
570 * timer is needed to decrease the active utilization at the correct
573 struct hrtimer inactive_timer;
575 #ifdef CONFIG_RT_MUTEXES
577 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
578 * pi_se points to the donor, otherwise points to the dl_se it belongs
579 * to (the original one/itself).
581 struct sched_dl_entity *pi_se;
585 #ifdef CONFIG_UCLAMP_TASK
586 /* Number of utilization clamp buckets (shorter alias) */
587 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
590 * Utilization clamp for a scheduling entity
591 * @value: clamp value "assigned" to a se
592 * @bucket_id: bucket index corresponding to the "assigned" value
593 * @active: the se is currently refcounted in a rq's bucket
594 * @user_defined: the requested clamp value comes from user-space
596 * The bucket_id is the index of the clamp bucket matching the clamp value
597 * which is pre-computed and stored to avoid expensive integer divisions from
600 * The active bit is set whenever a task has got an "effective" value assigned,
601 * which can be different from the clamp value "requested" from user-space.
602 * This allows to know a task is refcounted in the rq's bucket corresponding
603 * to the "effective" bucket_id.
605 * The user_defined bit is set whenever a task has got a task-specific clamp
606 * value requested from userspace, i.e. the system defaults apply to this task
607 * just as a restriction. This allows to relax default clamps when a less
608 * restrictive task-specific value has been requested, thus allowing to
609 * implement a "nice" semantic. For example, a task running with a 20%
610 * default boost can still drop its own boosting to 0%.
613 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
614 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
615 unsigned int active : 1;
616 unsigned int user_defined : 1;
618 #endif /* CONFIG_UCLAMP_TASK */
624 u8 exp_hint; /* Hint for performance. */
625 u8 need_mb; /* Readers need smp_mb(). */
627 u32 s; /* Set of bits. */
630 enum perf_event_task_context {
631 perf_invalid_context = -1,
634 perf_nr_task_contexts,
638 struct wake_q_node *next;
642 #ifdef CONFIG_THREAD_INFO_IN_TASK
644 * For reasons of header soup (see current_thread_info()), this
645 * must be the first element of task_struct.
647 struct thread_info thread_info;
649 /* -1 unrunnable, 0 runnable, >0 stopped: */
653 * This begins the randomizable portion of task_struct. Only
654 * scheduling-critical items should be added above here.
656 randomized_struct_fields_start
660 /* Per task flags (PF_*), defined further below: */
666 struct __call_single_node wake_entry;
667 #ifdef CONFIG_THREAD_INFO_IN_TASK
671 unsigned int wakee_flips;
672 unsigned long wakee_flip_decay_ts;
673 struct task_struct *last_wakee;
676 * recent_used_cpu is initially set as the last CPU used by a task
677 * that wakes affine another task. Waker/wakee relationships can
678 * push tasks around a CPU where each wakeup moves to the next one.
679 * Tracking a recently used CPU allows a quick search for a recently
680 * used CPU that may be idle.
690 unsigned int rt_priority;
692 const struct sched_class *sched_class;
693 struct sched_entity se;
694 struct sched_rt_entity rt;
695 #ifdef CONFIG_CGROUP_SCHED
696 struct task_group *sched_task_group;
698 struct sched_dl_entity dl;
700 #ifdef CONFIG_UCLAMP_TASK
702 * Clamp values requested for a scheduling entity.
703 * Must be updated with task_rq_lock() held.
705 struct uclamp_se uclamp_req[UCLAMP_CNT];
707 * Effective clamp values used for a scheduling entity.
708 * Must be updated with task_rq_lock() held.
710 struct uclamp_se uclamp[UCLAMP_CNT];
713 #ifdef CONFIG_PREEMPT_NOTIFIERS
714 /* List of struct preempt_notifier: */
715 struct hlist_head preempt_notifiers;
718 #ifdef CONFIG_BLK_DEV_IO_TRACE
719 unsigned int btrace_seq;
724 const cpumask_t *cpus_ptr;
727 #ifdef CONFIG_PREEMPT_RCU
728 int rcu_read_lock_nesting;
729 union rcu_special rcu_read_unlock_special;
730 struct list_head rcu_node_entry;
731 struct rcu_node *rcu_blocked_node;
732 #endif /* #ifdef CONFIG_PREEMPT_RCU */
734 #ifdef CONFIG_TASKS_RCU
735 unsigned long rcu_tasks_nvcsw;
736 u8 rcu_tasks_holdout;
738 int rcu_tasks_idle_cpu;
739 struct list_head rcu_tasks_holdout_list;
740 #endif /* #ifdef CONFIG_TASKS_RCU */
742 #ifdef CONFIG_TASKS_TRACE_RCU
743 int trc_reader_nesting;
745 union rcu_special trc_reader_special;
746 bool trc_reader_checked;
747 struct list_head trc_holdout_list;
748 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
750 struct sched_info sched_info;
752 struct list_head tasks;
754 struct plist_node pushable_tasks;
755 struct rb_node pushable_dl_tasks;
758 struct mm_struct *mm;
759 struct mm_struct *active_mm;
761 /* Per-thread vma caching: */
762 struct vmacache vmacache;
764 #ifdef SPLIT_RSS_COUNTING
765 struct task_rss_stat rss_stat;
770 /* The signal sent when the parent dies: */
772 /* JOBCTL_*, siglock protected: */
773 unsigned long jobctl;
775 /* Used for emulating ABI behavior of previous Linux versions: */
776 unsigned int personality;
778 /* Scheduler bits, serialized by scheduler locks: */
779 unsigned sched_reset_on_fork:1;
780 unsigned sched_contributes_to_load:1;
781 unsigned sched_migrated:1;
783 unsigned sched_psi_wake_requeue:1;
786 /* Force alignment to the next boundary: */
789 /* Unserialized, strictly 'current' */
792 * This field must not be in the scheduler word above due to wakelist
793 * queueing no longer being serialized by p->on_cpu. However:
796 * schedule() if (p->on_rq && ..) // false
797 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
798 * deactivate_task() ttwu_queue_wakelist())
799 * p->on_rq = 0; p->sched_remote_wakeup = Y;
801 * guarantees all stores of 'current' are visible before
802 * ->sched_remote_wakeup gets used, so it can be in this word.
804 unsigned sched_remote_wakeup:1;
806 /* Bit to tell LSMs we're in execve(): */
807 unsigned in_execve:1;
808 unsigned in_iowait:1;
809 #ifndef TIF_RESTORE_SIGMASK
810 unsigned restore_sigmask:1;
813 unsigned in_user_fault:1;
815 #ifdef CONFIG_COMPAT_BRK
816 unsigned brk_randomized:1;
818 #ifdef CONFIG_CGROUPS
819 /* disallow userland-initiated cgroup migration */
820 unsigned no_cgroup_migration:1;
821 /* task is frozen/stopped (used by the cgroup freezer) */
824 #ifdef CONFIG_BLK_CGROUP
825 unsigned use_memdelay:1;
828 /* Stalled due to lack of memory */
829 unsigned in_memstall:1;
832 unsigned long atomic_flags; /* Flags requiring atomic access. */
834 struct restart_block restart_block;
839 #ifdef CONFIG_STACKPROTECTOR
840 /* Canary value for the -fstack-protector GCC feature: */
841 unsigned long stack_canary;
844 * Pointers to the (original) parent process, youngest child, younger sibling,
845 * older sibling, respectively. (p->father can be replaced with
846 * p->real_parent->pid)
849 /* Real parent process: */
850 struct task_struct __rcu *real_parent;
852 /* Recipient of SIGCHLD, wait4() reports: */
853 struct task_struct __rcu *parent;
856 * Children/sibling form the list of natural children:
858 struct list_head children;
859 struct list_head sibling;
860 struct task_struct *group_leader;
863 * 'ptraced' is the list of tasks this task is using ptrace() on.
865 * This includes both natural children and PTRACE_ATTACH targets.
866 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
868 struct list_head ptraced;
869 struct list_head ptrace_entry;
871 /* PID/PID hash table linkage. */
872 struct pid *thread_pid;
873 struct hlist_node pid_links[PIDTYPE_MAX];
874 struct list_head thread_group;
875 struct list_head thread_node;
877 struct completion *vfork_done;
879 /* CLONE_CHILD_SETTID: */
880 int __user *set_child_tid;
882 /* CLONE_CHILD_CLEARTID: */
883 int __user *clear_child_tid;
887 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
892 struct prev_cputime prev_cputime;
893 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
897 #ifdef CONFIG_NO_HZ_FULL
898 atomic_t tick_dep_mask;
900 /* Context switch counts: */
902 unsigned long nivcsw;
904 /* Monotonic time in nsecs: */
907 /* Boot based time in nsecs: */
910 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
911 unsigned long min_flt;
912 unsigned long maj_flt;
914 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
915 struct posix_cputimers posix_cputimers;
917 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
918 struct posix_cputimers_work posix_cputimers_work;
921 /* Process credentials: */
923 /* Tracer's credentials at attach: */
924 const struct cred __rcu *ptracer_cred;
926 /* Objective and real subjective task credentials (COW): */
927 const struct cred __rcu *real_cred;
929 /* Effective (overridable) subjective task credentials (COW): */
930 const struct cred __rcu *cred;
933 /* Cached requested key. */
934 struct key *cached_requested_key;
938 * executable name, excluding path.
940 * - normally initialized setup_new_exec()
941 * - access it with [gs]et_task_comm()
942 * - lock it with task_lock()
944 char comm[TASK_COMM_LEN];
946 struct nameidata *nameidata;
948 #ifdef CONFIG_SYSVIPC
949 struct sysv_sem sysvsem;
950 struct sysv_shm sysvshm;
952 #ifdef CONFIG_DETECT_HUNG_TASK
953 unsigned long last_switch_count;
954 unsigned long last_switch_time;
956 /* Filesystem information: */
957 struct fs_struct *fs;
959 /* Open file information: */
960 struct files_struct *files;
962 #ifdef CONFIG_IO_URING
963 struct io_uring_task *io_uring;
967 struct nsproxy *nsproxy;
969 /* Signal handlers: */
970 struct signal_struct *signal;
971 struct sighand_struct __rcu *sighand;
973 sigset_t real_blocked;
974 /* Restored if set_restore_sigmask() was used: */
975 sigset_t saved_sigmask;
976 struct sigpending pending;
977 unsigned long sas_ss_sp;
979 unsigned int sas_ss_flags;
981 struct callback_head *task_works;
984 #ifdef CONFIG_AUDITSYSCALL
985 struct audit_context *audit_context;
988 unsigned int sessionid;
990 struct seccomp seccomp;
991 struct syscall_user_dispatch syscall_dispatch;
993 /* Thread group tracking: */
997 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
998 spinlock_t alloc_lock;
1000 /* Protection of the PI data structures: */
1001 raw_spinlock_t pi_lock;
1003 struct wake_q_node wake_q;
1005 #ifdef CONFIG_RT_MUTEXES
1006 /* PI waiters blocked on a rt_mutex held by this task: */
1007 struct rb_root_cached pi_waiters;
1008 /* Updated under owner's pi_lock and rq lock */
1009 struct task_struct *pi_top_task;
1010 /* Deadlock detection and priority inheritance handling: */
1011 struct rt_mutex_waiter *pi_blocked_on;
1014 #ifdef CONFIG_DEBUG_MUTEXES
1015 /* Mutex deadlock detection: */
1016 struct mutex_waiter *blocked_on;
1019 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1020 int non_block_count;
1023 #ifdef CONFIG_TRACE_IRQFLAGS
1024 struct irqtrace_events irqtrace;
1025 unsigned int hardirq_threaded;
1026 u64 hardirq_chain_key;
1027 int softirqs_enabled;
1028 int softirq_context;
1032 #ifdef CONFIG_LOCKDEP
1033 # define MAX_LOCK_DEPTH 48UL
1036 unsigned int lockdep_recursion;
1037 struct held_lock held_locks[MAX_LOCK_DEPTH];
1040 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1041 unsigned int in_ubsan;
1044 /* Journalling filesystem info: */
1047 /* Stacked block device info: */
1048 struct bio_list *bio_list;
1051 /* Stack plugging: */
1052 struct blk_plug *plug;
1056 struct reclaim_state *reclaim_state;
1058 struct backing_dev_info *backing_dev_info;
1060 struct io_context *io_context;
1062 #ifdef CONFIG_COMPACTION
1063 struct capture_control *capture_control;
1066 unsigned long ptrace_message;
1067 kernel_siginfo_t *last_siginfo;
1069 struct task_io_accounting ioac;
1071 /* Pressure stall state */
1072 unsigned int psi_flags;
1074 #ifdef CONFIG_TASK_XACCT
1075 /* Accumulated RSS usage: */
1077 /* Accumulated virtual memory usage: */
1079 /* stime + utime since last update: */
1082 #ifdef CONFIG_CPUSETS
1083 /* Protected by ->alloc_lock: */
1084 nodemask_t mems_allowed;
1085 /* Seqence number to catch updates: */
1086 seqcount_spinlock_t mems_allowed_seq;
1087 int cpuset_mem_spread_rotor;
1088 int cpuset_slab_spread_rotor;
1090 #ifdef CONFIG_CGROUPS
1091 /* Control Group info protected by css_set_lock: */
1092 struct css_set __rcu *cgroups;
1093 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1094 struct list_head cg_list;
1096 #ifdef CONFIG_X86_CPU_RESCTRL
1101 struct robust_list_head __user *robust_list;
1102 #ifdef CONFIG_COMPAT
1103 struct compat_robust_list_head __user *compat_robust_list;
1105 struct list_head pi_state_list;
1106 struct futex_pi_state *pi_state_cache;
1107 struct mutex futex_exit_mutex;
1108 unsigned int futex_state;
1110 #ifdef CONFIG_PERF_EVENTS
1111 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1112 struct mutex perf_event_mutex;
1113 struct list_head perf_event_list;
1115 #ifdef CONFIG_DEBUG_PREEMPT
1116 unsigned long preempt_disable_ip;
1119 /* Protected by alloc_lock: */
1120 struct mempolicy *mempolicy;
1122 short pref_node_fork;
1124 #ifdef CONFIG_NUMA_BALANCING
1126 unsigned int numa_scan_period;
1127 unsigned int numa_scan_period_max;
1128 int numa_preferred_nid;
1129 unsigned long numa_migrate_retry;
1130 /* Migration stamp: */
1132 u64 last_task_numa_placement;
1133 u64 last_sum_exec_runtime;
1134 struct callback_head numa_work;
1137 * This pointer is only modified for current in syscall and
1138 * pagefault context (and for tasks being destroyed), so it can be read
1139 * from any of the following contexts:
1140 * - RCU read-side critical section
1141 * - current->numa_group from everywhere
1142 * - task's runqueue locked, task not running
1144 struct numa_group __rcu *numa_group;
1147 * numa_faults is an array split into four regions:
1148 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1149 * in this precise order.
1151 * faults_memory: Exponential decaying average of faults on a per-node
1152 * basis. Scheduling placement decisions are made based on these
1153 * counts. The values remain static for the duration of a PTE scan.
1154 * faults_cpu: Track the nodes the process was running on when a NUMA
1155 * hinting fault was incurred.
1156 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1157 * during the current scan window. When the scan completes, the counts
1158 * in faults_memory and faults_cpu decay and these values are copied.
1160 unsigned long *numa_faults;
1161 unsigned long total_numa_faults;
1164 * numa_faults_locality tracks if faults recorded during the last
1165 * scan window were remote/local or failed to migrate. The task scan
1166 * period is adapted based on the locality of the faults with different
1167 * weights depending on whether they were shared or private faults
1169 unsigned long numa_faults_locality[3];
1171 unsigned long numa_pages_migrated;
1172 #endif /* CONFIG_NUMA_BALANCING */
1175 struct rseq __user *rseq;
1178 * RmW on rseq_event_mask must be performed atomically
1179 * with respect to preemption.
1181 unsigned long rseq_event_mask;
1184 struct tlbflush_unmap_batch tlb_ubc;
1187 refcount_t rcu_users;
1188 struct rcu_head rcu;
1191 /* Cache last used pipe for splice(): */
1192 struct pipe_inode_info *splice_pipe;
1194 struct page_frag task_frag;
1196 #ifdef CONFIG_TASK_DELAY_ACCT
1197 struct task_delay_info *delays;
1200 #ifdef CONFIG_FAULT_INJECTION
1202 unsigned int fail_nth;
1205 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1206 * balance_dirty_pages() for a dirty throttling pause:
1209 int nr_dirtied_pause;
1210 /* Start of a write-and-pause period: */
1211 unsigned long dirty_paused_when;
1213 #ifdef CONFIG_LATENCYTOP
1214 int latency_record_count;
1215 struct latency_record latency_record[LT_SAVECOUNT];
1218 * Time slack values; these are used to round up poll() and
1219 * select() etc timeout values. These are in nanoseconds.
1222 u64 default_timer_slack_ns;
1225 unsigned int kasan_depth;
1229 struct kcsan_ctx kcsan_ctx;
1230 #ifdef CONFIG_TRACE_IRQFLAGS
1231 struct irqtrace_events kcsan_save_irqtrace;
1235 #if IS_ENABLED(CONFIG_KUNIT)
1236 struct kunit *kunit_test;
1239 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1240 /* Index of current stored address in ret_stack: */
1244 /* Stack of return addresses for return function tracing: */
1245 struct ftrace_ret_stack *ret_stack;
1247 /* Timestamp for last schedule: */
1248 unsigned long long ftrace_timestamp;
1251 * Number of functions that haven't been traced
1252 * because of depth overrun:
1254 atomic_t trace_overrun;
1256 /* Pause tracing: */
1257 atomic_t tracing_graph_pause;
1260 #ifdef CONFIG_TRACING
1261 /* State flags for use by tracers: */
1262 unsigned long trace;
1264 /* Bitmask and counter of trace recursion: */
1265 unsigned long trace_recursion;
1266 #endif /* CONFIG_TRACING */
1269 /* See kernel/kcov.c for more details. */
1271 /* Coverage collection mode enabled for this task (0 if disabled): */
1272 unsigned int kcov_mode;
1274 /* Size of the kcov_area: */
1275 unsigned int kcov_size;
1277 /* Buffer for coverage collection: */
1280 /* KCOV descriptor wired with this task or NULL: */
1283 /* KCOV common handle for remote coverage collection: */
1286 /* KCOV sequence number: */
1289 /* Collect coverage from softirq context: */
1290 unsigned int kcov_softirq;
1294 struct mem_cgroup *memcg_in_oom;
1295 gfp_t memcg_oom_gfp_mask;
1296 int memcg_oom_order;
1298 /* Number of pages to reclaim on returning to userland: */
1299 unsigned int memcg_nr_pages_over_high;
1301 /* Used by memcontrol for targeted memcg charge: */
1302 struct mem_cgroup *active_memcg;
1305 #ifdef CONFIG_BLK_CGROUP
1306 struct request_queue *throttle_queue;
1309 #ifdef CONFIG_UPROBES
1310 struct uprobe_task *utask;
1312 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1313 unsigned int sequential_io;
1314 unsigned int sequential_io_avg;
1316 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1317 unsigned long task_state_change;
1319 int pagefault_disabled;
1321 struct task_struct *oom_reaper_list;
1323 #ifdef CONFIG_VMAP_STACK
1324 struct vm_struct *stack_vm_area;
1326 #ifdef CONFIG_THREAD_INFO_IN_TASK
1327 /* A live task holds one reference: */
1328 refcount_t stack_refcount;
1330 #ifdef CONFIG_LIVEPATCH
1333 #ifdef CONFIG_SECURITY
1334 /* Used by LSM modules for access restriction: */
1338 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1339 unsigned long lowest_stack;
1340 unsigned long prev_lowest_stack;
1343 #ifdef CONFIG_X86_MCE
1344 void __user *mce_vaddr;
1349 __mce_reserved : 62;
1350 struct callback_head mce_kill_me;
1353 #ifdef CONFIG_KRETPROBES
1354 struct llist_head kretprobe_instances;
1358 * New fields for task_struct should be added above here, so that
1359 * they are included in the randomized portion of task_struct.
1361 randomized_struct_fields_end
1363 /* CPU-specific state of this task: */
1364 struct thread_struct thread;
1367 * WARNING: on x86, 'thread_struct' contains a variable-sized
1368 * structure. It *MUST* be at the end of 'task_struct'.
1370 * Do not put anything below here!
1374 static inline struct pid *task_pid(struct task_struct *task)
1376 return task->thread_pid;
1380 * the helpers to get the task's different pids as they are seen
1381 * from various namespaces
1383 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1384 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1386 * task_xid_nr_ns() : id seen from the ns specified;
1388 * see also pid_nr() etc in include/linux/pid.h
1390 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1392 static inline pid_t task_pid_nr(struct task_struct *tsk)
1397 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1399 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1402 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1404 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1408 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1414 * pid_alive - check that a task structure is not stale
1415 * @p: Task structure to be checked.
1417 * Test if a process is not yet dead (at most zombie state)
1418 * If pid_alive fails, then pointers within the task structure
1419 * can be stale and must not be dereferenced.
1421 * Return: 1 if the process is alive. 0 otherwise.
1423 static inline int pid_alive(const struct task_struct *p)
1425 return p->thread_pid != NULL;
1428 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1430 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1433 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1435 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1439 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1441 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1444 static inline pid_t task_session_vnr(struct task_struct *tsk)
1446 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1449 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1451 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1454 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1456 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1459 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1465 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1471 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1473 return task_ppid_nr_ns(tsk, &init_pid_ns);
1476 /* Obsolete, do not use: */
1477 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1479 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1482 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1483 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1485 static inline unsigned int task_state_index(struct task_struct *tsk)
1487 unsigned int tsk_state = READ_ONCE(tsk->state);
1488 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1490 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1492 if (tsk_state == TASK_IDLE)
1493 state = TASK_REPORT_IDLE;
1498 static inline char task_index_to_char(unsigned int state)
1500 static const char state_char[] = "RSDTtXZPI";
1502 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1504 return state_char[state];
1507 static inline char task_state_to_char(struct task_struct *tsk)
1509 return task_index_to_char(task_state_index(tsk));
1513 * is_global_init - check if a task structure is init. Since init
1514 * is free to have sub-threads we need to check tgid.
1515 * @tsk: Task structure to be checked.
1517 * Check if a task structure is the first user space task the kernel created.
1519 * Return: 1 if the task structure is init. 0 otherwise.
1521 static inline int is_global_init(struct task_struct *tsk)
1523 return task_tgid_nr(tsk) == 1;
1526 extern struct pid *cad_pid;
1531 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1532 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1533 #define PF_EXITING 0x00000004 /* Getting shut down */
1534 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1535 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1536 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1537 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1538 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1539 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1540 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1541 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1542 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1543 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1544 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1545 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1546 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1547 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1548 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1549 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1550 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1551 * I am cleaning dirty pages from some other bdi. */
1552 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1553 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1554 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1555 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1556 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1557 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1558 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1559 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1562 * Only the _current_ task can read/write to tsk->flags, but other
1563 * tasks can access tsk->flags in readonly mode for example
1564 * with tsk_used_math (like during threaded core dumping).
1565 * There is however an exception to this rule during ptrace
1566 * or during fork: the ptracer task is allowed to write to the
1567 * child->flags of its traced child (same goes for fork, the parent
1568 * can write to the child->flags), because we're guaranteed the
1569 * child is not running and in turn not changing child->flags
1570 * at the same time the parent does it.
1572 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1573 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1574 #define clear_used_math() clear_stopped_child_used_math(current)
1575 #define set_used_math() set_stopped_child_used_math(current)
1577 #define conditional_stopped_child_used_math(condition, child) \
1578 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1580 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1582 #define copy_to_stopped_child_used_math(child) \
1583 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1585 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1586 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1587 #define used_math() tsk_used_math(current)
1589 static inline bool is_percpu_thread(void)
1592 return (current->flags & PF_NO_SETAFFINITY) &&
1593 (current->nr_cpus_allowed == 1);
1599 /* Per-process atomic flags. */
1600 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1601 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1602 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1603 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1604 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1605 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1606 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1607 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1609 #define TASK_PFA_TEST(name, func) \
1610 static inline bool task_##func(struct task_struct *p) \
1611 { return test_bit(PFA_##name, &p->atomic_flags); }
1613 #define TASK_PFA_SET(name, func) \
1614 static inline void task_set_##func(struct task_struct *p) \
1615 { set_bit(PFA_##name, &p->atomic_flags); }
1617 #define TASK_PFA_CLEAR(name, func) \
1618 static inline void task_clear_##func(struct task_struct *p) \
1619 { clear_bit(PFA_##name, &p->atomic_flags); }
1621 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1622 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1624 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1625 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1626 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1628 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1629 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1630 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1632 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1633 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1634 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1636 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1637 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1638 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1640 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1641 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1643 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1644 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1645 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1647 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1648 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1651 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1653 current->flags &= ~flags;
1654 current->flags |= orig_flags & flags;
1657 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1658 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1660 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1661 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1663 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1666 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1668 if (!cpumask_test_cpu(0, new_mask))
1674 extern int yield_to(struct task_struct *p, bool preempt);
1675 extern void set_user_nice(struct task_struct *p, long nice);
1676 extern int task_prio(const struct task_struct *p);
1679 * task_nice - return the nice value of a given task.
1680 * @p: the task in question.
1682 * Return: The nice value [ -20 ... 0 ... 19 ].
1684 static inline int task_nice(const struct task_struct *p)
1686 return PRIO_TO_NICE((p)->static_prio);
1689 extern int can_nice(const struct task_struct *p, const int nice);
1690 extern int task_curr(const struct task_struct *p);
1691 extern int idle_cpu(int cpu);
1692 extern int available_idle_cpu(int cpu);
1693 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1694 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1695 extern void sched_set_fifo(struct task_struct *p);
1696 extern void sched_set_fifo_low(struct task_struct *p);
1697 extern void sched_set_normal(struct task_struct *p, int nice);
1698 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1699 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1700 extern struct task_struct *idle_task(int cpu);
1703 * is_idle_task - is the specified task an idle task?
1704 * @p: the task in question.
1706 * Return: 1 if @p is an idle task. 0 otherwise.
1708 static __always_inline bool is_idle_task(const struct task_struct *p)
1710 return !!(p->flags & PF_IDLE);
1713 extern struct task_struct *curr_task(int cpu);
1714 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1718 union thread_union {
1719 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1720 struct task_struct task;
1722 #ifndef CONFIG_THREAD_INFO_IN_TASK
1723 struct thread_info thread_info;
1725 unsigned long stack[THREAD_SIZE/sizeof(long)];
1728 #ifndef CONFIG_THREAD_INFO_IN_TASK
1729 extern struct thread_info init_thread_info;
1732 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1734 #ifdef CONFIG_THREAD_INFO_IN_TASK
1735 static inline struct thread_info *task_thread_info(struct task_struct *task)
1737 return &task->thread_info;
1739 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1740 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1744 * find a task by one of its numerical ids
1746 * find_task_by_pid_ns():
1747 * finds a task by its pid in the specified namespace
1748 * find_task_by_vpid():
1749 * finds a task by its virtual pid
1751 * see also find_vpid() etc in include/linux/pid.h
1754 extern struct task_struct *find_task_by_vpid(pid_t nr);
1755 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1758 * find a task by its virtual pid and get the task struct
1760 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1762 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1763 extern int wake_up_process(struct task_struct *tsk);
1764 extern void wake_up_new_task(struct task_struct *tsk);
1767 extern void kick_process(struct task_struct *tsk);
1769 static inline void kick_process(struct task_struct *tsk) { }
1772 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1774 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1776 __set_task_comm(tsk, from, false);
1779 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1780 #define get_task_comm(buf, tsk) ({ \
1781 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1782 __get_task_comm(buf, sizeof(buf), tsk); \
1786 static __always_inline void scheduler_ipi(void)
1789 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1790 * TIF_NEED_RESCHED remotely (for the first time) will also send
1793 preempt_fold_need_resched();
1795 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1797 static inline void scheduler_ipi(void) { }
1798 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1805 * Set thread flags in other task's structures.
1806 * See asm/thread_info.h for TIF_xxxx flags available:
1808 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1810 set_ti_thread_flag(task_thread_info(tsk), flag);
1813 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1815 clear_ti_thread_flag(task_thread_info(tsk), flag);
1818 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1821 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1824 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1826 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1829 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1831 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1834 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1836 return test_ti_thread_flag(task_thread_info(tsk), flag);
1839 static inline void set_tsk_need_resched(struct task_struct *tsk)
1841 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1844 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1846 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1849 static inline int test_tsk_need_resched(struct task_struct *tsk)
1851 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1855 * cond_resched() and cond_resched_lock(): latency reduction via
1856 * explicit rescheduling in places that are safe. The return
1857 * value indicates whether a reschedule was done in fact.
1858 * cond_resched_lock() will drop the spinlock before scheduling,
1860 #ifndef CONFIG_PREEMPTION
1861 extern int _cond_resched(void);
1863 static inline int _cond_resched(void) { return 0; }
1866 #define cond_resched() ({ \
1867 ___might_sleep(__FILE__, __LINE__, 0); \
1871 extern int __cond_resched_lock(spinlock_t *lock);
1873 #define cond_resched_lock(lock) ({ \
1874 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1875 __cond_resched_lock(lock); \
1878 static inline void cond_resched_rcu(void)
1880 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1888 * Does a critical section need to be broken due to another
1889 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1890 * but a general need for low latency)
1892 static inline int spin_needbreak(spinlock_t *lock)
1894 #ifdef CONFIG_PREEMPTION
1895 return spin_is_contended(lock);
1901 static __always_inline bool need_resched(void)
1903 return unlikely(tif_need_resched());
1907 * Wrappers for p->thread_info->cpu access. No-op on UP.
1911 static inline unsigned int task_cpu(const struct task_struct *p)
1913 #ifdef CONFIG_THREAD_INFO_IN_TASK
1914 return READ_ONCE(p->cpu);
1916 return READ_ONCE(task_thread_info(p)->cpu);
1920 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1924 static inline unsigned int task_cpu(const struct task_struct *p)
1929 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1933 #endif /* CONFIG_SMP */
1936 * In order to reduce various lock holder preemption latencies provide an
1937 * interface to see if a vCPU is currently running or not.
1939 * This allows us to terminate optimistic spin loops and block, analogous to
1940 * the native optimistic spin heuristic of testing if the lock owner task is
1943 #ifndef vcpu_is_preempted
1944 static inline bool vcpu_is_preempted(int cpu)
1950 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1951 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1953 #ifndef TASK_SIZE_OF
1954 #define TASK_SIZE_OF(tsk) TASK_SIZE
1960 * Map the event mask on the user-space ABI enum rseq_cs_flags
1961 * for direct mask checks.
1963 enum rseq_event_mask_bits {
1964 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1965 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1966 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1969 enum rseq_event_mask {
1970 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1971 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1972 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1975 static inline void rseq_set_notify_resume(struct task_struct *t)
1978 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1981 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1983 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1984 struct pt_regs *regs)
1987 __rseq_handle_notify_resume(ksig, regs);
1990 static inline void rseq_signal_deliver(struct ksignal *ksig,
1991 struct pt_regs *regs)
1994 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1996 rseq_handle_notify_resume(ksig, regs);
1999 /* rseq_preempt() requires preemption to be disabled. */
2000 static inline void rseq_preempt(struct task_struct *t)
2002 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2003 rseq_set_notify_resume(t);
2006 /* rseq_migrate() requires preemption to be disabled. */
2007 static inline void rseq_migrate(struct task_struct *t)
2009 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2010 rseq_set_notify_resume(t);
2014 * If parent process has a registered restartable sequences area, the
2015 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2017 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2019 if (clone_flags & CLONE_VM) {
2022 t->rseq_event_mask = 0;
2024 t->rseq = current->rseq;
2025 t->rseq_sig = current->rseq_sig;
2026 t->rseq_event_mask = current->rseq_event_mask;
2030 static inline void rseq_execve(struct task_struct *t)
2034 t->rseq_event_mask = 0;
2039 static inline void rseq_set_notify_resume(struct task_struct *t)
2042 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2043 struct pt_regs *regs)
2046 static inline void rseq_signal_deliver(struct ksignal *ksig,
2047 struct pt_regs *regs)
2050 static inline void rseq_preempt(struct task_struct *t)
2053 static inline void rseq_migrate(struct task_struct *t)
2056 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2059 static inline void rseq_execve(struct task_struct *t)
2065 #ifdef CONFIG_DEBUG_RSEQ
2067 void rseq_syscall(struct pt_regs *regs);
2071 static inline void rseq_syscall(struct pt_regs *regs)
2077 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2078 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2079 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2081 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2082 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2083 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2085 int sched_trace_rq_cpu(struct rq *rq);
2086 int sched_trace_rq_cpu_capacity(struct rq *rq);
2087 int sched_trace_rq_nr_running(struct rq *rq);
2089 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);