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 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
120 * Special states are those that do not use the normal wait-loop pattern. See
121 * the comment with set_special_state().
123 #define is_special_task_state(state) \
124 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
126 #define __set_current_state(state_value) \
128 WARN_ON_ONCE(is_special_task_state(state_value));\
129 current->task_state_change = _THIS_IP_; \
130 current->state = (state_value); \
133 #define set_current_state(state_value) \
135 WARN_ON_ONCE(is_special_task_state(state_value));\
136 current->task_state_change = _THIS_IP_; \
137 smp_store_mb(current->state, (state_value)); \
140 #define set_special_state(state_value) \
142 unsigned long flags; /* may shadow */ \
143 WARN_ON_ONCE(!is_special_task_state(state_value)); \
144 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
145 current->task_state_change = _THIS_IP_; \
146 current->state = (state_value); \
147 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
151 * set_current_state() includes a barrier so that the write of current->state
152 * is correctly serialised wrt the caller's subsequent test of whether to
156 * set_current_state(TASK_UNINTERRUPTIBLE);
162 * __set_current_state(TASK_RUNNING);
164 * If the caller does not need such serialisation (because, for instance, the
165 * condition test and condition change and wakeup are under the same lock) then
166 * use __set_current_state().
168 * The above is typically ordered against the wakeup, which does:
170 * need_sleep = false;
171 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
173 * where wake_up_state() executes a full memory barrier before accessing the
176 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
177 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
178 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
180 * However, with slightly different timing the wakeup TASK_RUNNING store can
181 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
182 * a problem either because that will result in one extra go around the loop
183 * and our @cond test will save the day.
185 * Also see the comments of try_to_wake_up().
187 #define __set_current_state(state_value) \
188 current->state = (state_value)
190 #define set_current_state(state_value) \
191 smp_store_mb(current->state, (state_value))
194 * set_special_state() should be used for those states when the blocking task
195 * can not use the regular condition based wait-loop. In that case we must
196 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
197 * will not collide with our state change.
199 #define set_special_state(state_value) \
201 unsigned long flags; /* may shadow */ \
202 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
203 current->state = (state_value); \
204 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
209 /* Task command name length: */
210 #define TASK_COMM_LEN 16
212 extern void scheduler_tick(void);
214 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
216 extern long schedule_timeout(long timeout);
217 extern long schedule_timeout_interruptible(long timeout);
218 extern long schedule_timeout_killable(long timeout);
219 extern long schedule_timeout_uninterruptible(long timeout);
220 extern long schedule_timeout_idle(long timeout);
221 asmlinkage void schedule(void);
222 extern void schedule_preempt_disabled(void);
223 asmlinkage void preempt_schedule_irq(void);
225 extern int __must_check io_schedule_prepare(void);
226 extern void io_schedule_finish(int token);
227 extern long io_schedule_timeout(long timeout);
228 extern void io_schedule(void);
231 * struct prev_cputime - snapshot of system and user cputime
232 * @utime: time spent in user mode
233 * @stime: time spent in system mode
234 * @lock: protects the above two fields
236 * Stores previous user/system time values such that we can guarantee
239 struct prev_cputime {
240 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
248 /* Task is sleeping or running in a CPU with VTIME inactive: */
252 /* Task runs in kernelspace in a CPU with VTIME active: */
254 /* Task runs in userspace in a CPU with VTIME active: */
256 /* Task runs as guests in a CPU with VTIME active: */
262 unsigned long long starttime;
263 enum vtime_state state;
271 * Utilization clamp constraints.
272 * @UCLAMP_MIN: Minimum utilization
273 * @UCLAMP_MAX: Maximum utilization
274 * @UCLAMP_CNT: Utilization clamp constraints count
283 extern struct root_domain def_root_domain;
284 extern struct mutex sched_domains_mutex;
288 #ifdef CONFIG_SCHED_INFO
289 /* Cumulative counters: */
291 /* # of times we have run on this CPU: */
292 unsigned long pcount;
294 /* Time spent waiting on a runqueue: */
295 unsigned long long run_delay;
299 /* When did we last run on a CPU? */
300 unsigned long long last_arrival;
302 /* When were we last queued to run? */
303 unsigned long long last_queued;
305 #endif /* CONFIG_SCHED_INFO */
309 * Integer metrics need fixed point arithmetic, e.g., sched/fair
310 * has a few: load, load_avg, util_avg, freq, and capacity.
312 * We define a basic fixed point arithmetic range, and then formalize
313 * all these metrics based on that basic range.
315 # define SCHED_FIXEDPOINT_SHIFT 10
316 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
318 /* Increase resolution of cpu_capacity calculations */
319 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
320 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
323 unsigned long weight;
328 * struct util_est - Estimation utilization of FAIR tasks
329 * @enqueued: instantaneous estimated utilization of a task/cpu
330 * @ewma: the Exponential Weighted Moving Average (EWMA)
331 * utilization of a task
333 * Support data structure to track an Exponential Weighted Moving Average
334 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
335 * average each time a task completes an activation. Sample's weight is chosen
336 * so that the EWMA will be relatively insensitive to transient changes to the
339 * The enqueued attribute has a slightly different meaning for tasks and cpus:
340 * - task: the task's util_avg at last task dequeue time
341 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
342 * Thus, the util_est.enqueued of a task represents the contribution on the
343 * estimated utilization of the CPU where that task is currently enqueued.
345 * Only for tasks we track a moving average of the past instantaneous
346 * estimated utilization. This allows to absorb sporadic drops in utilization
347 * of an otherwise almost periodic task.
350 unsigned int enqueued;
352 #define UTIL_EST_WEIGHT_SHIFT 2
353 } __attribute__((__aligned__(sizeof(u64))));
356 * The load/runnable/util_avg accumulates an infinite geometric series
357 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
359 * [load_avg definition]
361 * load_avg = runnable% * scale_load_down(load)
363 * [runnable_avg definition]
365 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
367 * [util_avg definition]
369 * util_avg = running% * SCHED_CAPACITY_SCALE
371 * where runnable% is the time ratio that a sched_entity is runnable and
372 * running% the time ratio that a sched_entity is running.
374 * For cfs_rq, they are the aggregated values of all runnable and blocked
377 * The load/runnable/util_avg doesn't direcly factor frequency scaling and CPU
378 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
379 * for computing those signals (see update_rq_clock_pelt())
381 * N.B., the above ratios (runnable% and running%) themselves are in the
382 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
383 * to as large a range as necessary. This is for example reflected by
384 * util_avg's SCHED_CAPACITY_SCALE.
388 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
389 * with the highest load (=88761), always runnable on a single cfs_rq,
390 * and should not overflow as the number already hits PID_MAX_LIMIT.
392 * For all other cases (including 32-bit kernels), struct load_weight's
393 * weight will overflow first before we do, because:
395 * Max(load_avg) <= Max(load.weight)
397 * Then it is the load_weight's responsibility to consider overflow
401 u64 last_update_time;
406 unsigned long load_avg;
407 unsigned long runnable_avg;
408 unsigned long util_avg;
409 struct util_est util_est;
410 } ____cacheline_aligned;
412 struct sched_statistics {
413 #ifdef CONFIG_SCHEDSTATS
423 s64 sum_sleep_runtime;
430 u64 nr_migrations_cold;
431 u64 nr_failed_migrations_affine;
432 u64 nr_failed_migrations_running;
433 u64 nr_failed_migrations_hot;
434 u64 nr_forced_migrations;
438 u64 nr_wakeups_migrate;
439 u64 nr_wakeups_local;
440 u64 nr_wakeups_remote;
441 u64 nr_wakeups_affine;
442 u64 nr_wakeups_affine_attempts;
443 u64 nr_wakeups_passive;
448 struct sched_entity {
449 /* For load-balancing: */
450 struct load_weight load;
451 struct rb_node run_node;
452 struct list_head group_node;
456 u64 sum_exec_runtime;
458 u64 prev_sum_exec_runtime;
462 struct sched_statistics statistics;
464 #ifdef CONFIG_FAIR_GROUP_SCHED
466 struct sched_entity *parent;
467 /* rq on which this entity is (to be) queued: */
468 struct cfs_rq *cfs_rq;
469 /* rq "owned" by this entity/group: */
471 /* cached value of my_q->h_nr_running */
472 unsigned long runnable_weight;
477 * Per entity load average tracking.
479 * Put into separate cache line so it does not
480 * collide with read-mostly values above.
482 struct sched_avg avg;
486 struct sched_rt_entity {
487 struct list_head run_list;
488 unsigned long timeout;
489 unsigned long watchdog_stamp;
490 unsigned int time_slice;
491 unsigned short on_rq;
492 unsigned short on_list;
494 struct sched_rt_entity *back;
495 #ifdef CONFIG_RT_GROUP_SCHED
496 struct sched_rt_entity *parent;
497 /* rq on which this entity is (to be) queued: */
499 /* rq "owned" by this entity/group: */
502 } __randomize_layout;
504 struct sched_dl_entity {
505 struct rb_node rb_node;
508 * Original scheduling parameters. Copied here from sched_attr
509 * during sched_setattr(), they will remain the same until
510 * the next sched_setattr().
512 u64 dl_runtime; /* Maximum runtime for each instance */
513 u64 dl_deadline; /* Relative deadline of each instance */
514 u64 dl_period; /* Separation of two instances (period) */
515 u64 dl_bw; /* dl_runtime / dl_period */
516 u64 dl_density; /* dl_runtime / dl_deadline */
519 * Actual scheduling parameters. Initialized with the values above,
520 * they are continuously updated during task execution. Note that
521 * the remaining runtime could be < 0 in case we are in overrun.
523 s64 runtime; /* Remaining runtime for this instance */
524 u64 deadline; /* Absolute deadline for this instance */
525 unsigned int flags; /* Specifying the scheduler behaviour */
530 * @dl_throttled tells if we exhausted the runtime. If so, the
531 * task has to wait for a replenishment to be performed at the
532 * next firing of dl_timer.
534 * @dl_boosted tells if we are boosted due to DI. If so we are
535 * outside bandwidth enforcement mechanism (but only until we
536 * exit the critical section);
538 * @dl_yielded tells if task gave up the CPU before consuming
539 * all its available runtime during the last job.
541 * @dl_non_contending tells if the task is inactive while still
542 * contributing to the active utilization. In other words, it
543 * indicates if the inactive timer has been armed and its handler
544 * has not been executed yet. This flag is useful to avoid race
545 * conditions between the inactive timer handler and the wakeup
548 * @dl_overrun tells if the task asked to be informed about runtime
551 unsigned int dl_throttled : 1;
552 unsigned int dl_boosted : 1;
553 unsigned int dl_yielded : 1;
554 unsigned int dl_non_contending : 1;
555 unsigned int dl_overrun : 1;
558 * Bandwidth enforcement timer. Each -deadline task has its
559 * own bandwidth to be enforced, thus we need one timer per task.
561 struct hrtimer dl_timer;
564 * Inactive timer, responsible for decreasing the active utilization
565 * at the "0-lag time". When a -deadline task blocks, it contributes
566 * to GRUB's active utilization until the "0-lag time", hence a
567 * timer is needed to decrease the active utilization at the correct
570 struct hrtimer inactive_timer;
573 #ifdef CONFIG_UCLAMP_TASK
574 /* Number of utilization clamp buckets (shorter alias) */
575 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
578 * Utilization clamp for a scheduling entity
579 * @value: clamp value "assigned" to a se
580 * @bucket_id: bucket index corresponding to the "assigned" value
581 * @active: the se is currently refcounted in a rq's bucket
582 * @user_defined: the requested clamp value comes from user-space
584 * The bucket_id is the index of the clamp bucket matching the clamp value
585 * which is pre-computed and stored to avoid expensive integer divisions from
588 * The active bit is set whenever a task has got an "effective" value assigned,
589 * which can be different from the clamp value "requested" from user-space.
590 * This allows to know a task is refcounted in the rq's bucket corresponding
591 * to the "effective" bucket_id.
593 * The user_defined bit is set whenever a task has got a task-specific clamp
594 * value requested from userspace, i.e. the system defaults apply to this task
595 * just as a restriction. This allows to relax default clamps when a less
596 * restrictive task-specific value has been requested, thus allowing to
597 * implement a "nice" semantic. For example, a task running with a 20%
598 * default boost can still drop its own boosting to 0%.
601 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
602 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
603 unsigned int active : 1;
604 unsigned int user_defined : 1;
606 #endif /* CONFIG_UCLAMP_TASK */
612 u8 exp_hint; /* Hint for performance. */
613 u8 need_mb; /* Readers need smp_mb(). */
615 u32 s; /* Set of bits. */
618 enum perf_event_task_context {
619 perf_invalid_context = -1,
622 perf_nr_task_contexts,
626 struct wake_q_node *next;
630 #ifdef CONFIG_THREAD_INFO_IN_TASK
632 * For reasons of header soup (see current_thread_info()), this
633 * must be the first element of task_struct.
635 struct thread_info thread_info;
637 /* -1 unrunnable, 0 runnable, >0 stopped: */
641 * This begins the randomizable portion of task_struct. Only
642 * scheduling-critical items should be added above here.
644 randomized_struct_fields_start
648 /* Per task flags (PF_*), defined further below: */
654 struct __call_single_node wake_entry;
655 #ifdef CONFIG_THREAD_INFO_IN_TASK
659 unsigned int wakee_flips;
660 unsigned long wakee_flip_decay_ts;
661 struct task_struct *last_wakee;
664 * recent_used_cpu is initially set as the last CPU used by a task
665 * that wakes affine another task. Waker/wakee relationships can
666 * push tasks around a CPU where each wakeup moves to the next one.
667 * Tracking a recently used CPU allows a quick search for a recently
668 * used CPU that may be idle.
678 unsigned int rt_priority;
680 const struct sched_class *sched_class;
681 struct sched_entity se;
682 struct sched_rt_entity rt;
683 #ifdef CONFIG_CGROUP_SCHED
684 struct task_group *sched_task_group;
686 struct sched_dl_entity dl;
688 #ifdef CONFIG_UCLAMP_TASK
689 /* Clamp values requested for a scheduling entity */
690 struct uclamp_se uclamp_req[UCLAMP_CNT];
691 /* Effective clamp values used for a scheduling entity */
692 struct uclamp_se uclamp[UCLAMP_CNT];
695 #ifdef CONFIG_PREEMPT_NOTIFIERS
696 /* List of struct preempt_notifier: */
697 struct hlist_head preempt_notifiers;
700 #ifdef CONFIG_BLK_DEV_IO_TRACE
701 unsigned int btrace_seq;
706 const cpumask_t *cpus_ptr;
709 #ifdef CONFIG_PREEMPT_RCU
710 int rcu_read_lock_nesting;
711 union rcu_special rcu_read_unlock_special;
712 struct list_head rcu_node_entry;
713 struct rcu_node *rcu_blocked_node;
714 #endif /* #ifdef CONFIG_PREEMPT_RCU */
716 #ifdef CONFIG_TASKS_RCU
717 unsigned long rcu_tasks_nvcsw;
718 u8 rcu_tasks_holdout;
720 int rcu_tasks_idle_cpu;
721 struct list_head rcu_tasks_holdout_list;
722 #endif /* #ifdef CONFIG_TASKS_RCU */
724 #ifdef CONFIG_TASKS_TRACE_RCU
725 int trc_reader_nesting;
727 union rcu_special trc_reader_special;
728 bool trc_reader_checked;
729 struct list_head trc_holdout_list;
730 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
732 struct sched_info sched_info;
734 struct list_head tasks;
736 struct plist_node pushable_tasks;
737 struct rb_node pushable_dl_tasks;
740 struct mm_struct *mm;
741 struct mm_struct *active_mm;
743 /* Per-thread vma caching: */
744 struct vmacache vmacache;
746 #ifdef SPLIT_RSS_COUNTING
747 struct task_rss_stat rss_stat;
752 /* The signal sent when the parent dies: */
754 /* JOBCTL_*, siglock protected: */
755 unsigned long jobctl;
757 /* Used for emulating ABI behavior of previous Linux versions: */
758 unsigned int personality;
760 /* Scheduler bits, serialized by scheduler locks: */
761 unsigned sched_reset_on_fork:1;
762 unsigned sched_contributes_to_load:1;
763 unsigned sched_migrated:1;
764 unsigned sched_remote_wakeup:1;
766 unsigned sched_psi_wake_requeue:1;
769 /* Force alignment to the next boundary: */
772 /* Unserialized, strictly 'current' */
774 /* Bit to tell LSMs we're in execve(): */
775 unsigned in_execve:1;
776 unsigned in_iowait:1;
777 #ifndef TIF_RESTORE_SIGMASK
778 unsigned restore_sigmask:1;
781 unsigned in_user_fault:1;
783 #ifdef CONFIG_COMPAT_BRK
784 unsigned brk_randomized:1;
786 #ifdef CONFIG_CGROUPS
787 /* disallow userland-initiated cgroup migration */
788 unsigned no_cgroup_migration:1;
789 /* task is frozen/stopped (used by the cgroup freezer) */
792 #ifdef CONFIG_BLK_CGROUP
793 unsigned use_memdelay:1;
796 /* Stalled due to lack of memory */
797 unsigned in_memstall:1;
800 unsigned long atomic_flags; /* Flags requiring atomic access. */
802 struct restart_block restart_block;
807 #ifdef CONFIG_STACKPROTECTOR
808 /* Canary value for the -fstack-protector GCC feature: */
809 unsigned long stack_canary;
812 * Pointers to the (original) parent process, youngest child, younger sibling,
813 * older sibling, respectively. (p->father can be replaced with
814 * p->real_parent->pid)
817 /* Real parent process: */
818 struct task_struct __rcu *real_parent;
820 /* Recipient of SIGCHLD, wait4() reports: */
821 struct task_struct __rcu *parent;
824 * Children/sibling form the list of natural children:
826 struct list_head children;
827 struct list_head sibling;
828 struct task_struct *group_leader;
831 * 'ptraced' is the list of tasks this task is using ptrace() on.
833 * This includes both natural children and PTRACE_ATTACH targets.
834 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
836 struct list_head ptraced;
837 struct list_head ptrace_entry;
839 /* PID/PID hash table linkage. */
840 struct pid *thread_pid;
841 struct hlist_node pid_links[PIDTYPE_MAX];
842 struct list_head thread_group;
843 struct list_head thread_node;
845 struct completion *vfork_done;
847 /* CLONE_CHILD_SETTID: */
848 int __user *set_child_tid;
850 /* CLONE_CHILD_CLEARTID: */
851 int __user *clear_child_tid;
855 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
860 struct prev_cputime prev_cputime;
861 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
865 #ifdef CONFIG_NO_HZ_FULL
866 atomic_t tick_dep_mask;
868 /* Context switch counts: */
870 unsigned long nivcsw;
872 /* Monotonic time in nsecs: */
875 /* Boot based time in nsecs: */
878 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
879 unsigned long min_flt;
880 unsigned long maj_flt;
882 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
883 struct posix_cputimers posix_cputimers;
885 /* Process credentials: */
887 /* Tracer's credentials at attach: */
888 const struct cred __rcu *ptracer_cred;
890 /* Objective and real subjective task credentials (COW): */
891 const struct cred __rcu *real_cred;
893 /* Effective (overridable) subjective task credentials (COW): */
894 const struct cred __rcu *cred;
897 /* Cached requested key. */
898 struct key *cached_requested_key;
902 * executable name, excluding path.
904 * - normally initialized setup_new_exec()
905 * - access it with [gs]et_task_comm()
906 * - lock it with task_lock()
908 char comm[TASK_COMM_LEN];
910 struct nameidata *nameidata;
912 #ifdef CONFIG_SYSVIPC
913 struct sysv_sem sysvsem;
914 struct sysv_shm sysvshm;
916 #ifdef CONFIG_DETECT_HUNG_TASK
917 unsigned long last_switch_count;
918 unsigned long last_switch_time;
920 /* Filesystem information: */
921 struct fs_struct *fs;
923 /* Open file information: */
924 struct files_struct *files;
927 struct nsproxy *nsproxy;
929 /* Signal handlers: */
930 struct signal_struct *signal;
931 struct sighand_struct __rcu *sighand;
933 sigset_t real_blocked;
934 /* Restored if set_restore_sigmask() was used: */
935 sigset_t saved_sigmask;
936 struct sigpending pending;
937 unsigned long sas_ss_sp;
939 unsigned int sas_ss_flags;
941 struct callback_head *task_works;
944 #ifdef CONFIG_AUDITSYSCALL
945 struct audit_context *audit_context;
948 unsigned int sessionid;
950 struct seccomp seccomp;
952 /* Thread group tracking: */
956 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
957 spinlock_t alloc_lock;
959 /* Protection of the PI data structures: */
960 raw_spinlock_t pi_lock;
962 struct wake_q_node wake_q;
964 #ifdef CONFIG_RT_MUTEXES
965 /* PI waiters blocked on a rt_mutex held by this task: */
966 struct rb_root_cached pi_waiters;
967 /* Updated under owner's pi_lock and rq lock */
968 struct task_struct *pi_top_task;
969 /* Deadlock detection and priority inheritance handling: */
970 struct rt_mutex_waiter *pi_blocked_on;
973 #ifdef CONFIG_DEBUG_MUTEXES
974 /* Mutex deadlock detection: */
975 struct mutex_waiter *blocked_on;
978 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
982 #ifdef CONFIG_TRACE_IRQFLAGS
983 unsigned int irq_events;
984 unsigned int hardirq_threaded;
985 unsigned long hardirq_enable_ip;
986 unsigned long hardirq_disable_ip;
987 unsigned int hardirq_enable_event;
988 unsigned int hardirq_disable_event;
989 u64 hardirq_chain_key;
990 unsigned long softirq_disable_ip;
991 unsigned long softirq_enable_ip;
992 unsigned int softirq_disable_event;
993 unsigned int softirq_enable_event;
994 int softirqs_enabled;
999 #ifdef CONFIG_LOCKDEP
1000 # define MAX_LOCK_DEPTH 48UL
1003 unsigned int lockdep_recursion;
1004 struct held_lock held_locks[MAX_LOCK_DEPTH];
1008 unsigned int in_ubsan;
1011 /* Journalling filesystem info: */
1014 /* Stacked block device info: */
1015 struct bio_list *bio_list;
1018 /* Stack plugging: */
1019 struct blk_plug *plug;
1023 struct reclaim_state *reclaim_state;
1025 struct backing_dev_info *backing_dev_info;
1027 struct io_context *io_context;
1029 #ifdef CONFIG_COMPACTION
1030 struct capture_control *capture_control;
1033 unsigned long ptrace_message;
1034 kernel_siginfo_t *last_siginfo;
1036 struct task_io_accounting ioac;
1038 /* Pressure stall state */
1039 unsigned int psi_flags;
1041 #ifdef CONFIG_TASK_XACCT
1042 /* Accumulated RSS usage: */
1044 /* Accumulated virtual memory usage: */
1046 /* stime + utime since last update: */
1049 #ifdef CONFIG_CPUSETS
1050 /* Protected by ->alloc_lock: */
1051 nodemask_t mems_allowed;
1052 /* Seqence number to catch updates: */
1053 seqcount_t mems_allowed_seq;
1054 int cpuset_mem_spread_rotor;
1055 int cpuset_slab_spread_rotor;
1057 #ifdef CONFIG_CGROUPS
1058 /* Control Group info protected by css_set_lock: */
1059 struct css_set __rcu *cgroups;
1060 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1061 struct list_head cg_list;
1063 #ifdef CONFIG_X86_CPU_RESCTRL
1068 struct robust_list_head __user *robust_list;
1069 #ifdef CONFIG_COMPAT
1070 struct compat_robust_list_head __user *compat_robust_list;
1072 struct list_head pi_state_list;
1073 struct futex_pi_state *pi_state_cache;
1074 struct mutex futex_exit_mutex;
1075 unsigned int futex_state;
1077 #ifdef CONFIG_PERF_EVENTS
1078 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1079 struct mutex perf_event_mutex;
1080 struct list_head perf_event_list;
1082 #ifdef CONFIG_DEBUG_PREEMPT
1083 unsigned long preempt_disable_ip;
1086 /* Protected by alloc_lock: */
1087 struct mempolicy *mempolicy;
1089 short pref_node_fork;
1091 #ifdef CONFIG_NUMA_BALANCING
1093 unsigned int numa_scan_period;
1094 unsigned int numa_scan_period_max;
1095 int numa_preferred_nid;
1096 unsigned long numa_migrate_retry;
1097 /* Migration stamp: */
1099 u64 last_task_numa_placement;
1100 u64 last_sum_exec_runtime;
1101 struct callback_head numa_work;
1104 * This pointer is only modified for current in syscall and
1105 * pagefault context (and for tasks being destroyed), so it can be read
1106 * from any of the following contexts:
1107 * - RCU read-side critical section
1108 * - current->numa_group from everywhere
1109 * - task's runqueue locked, task not running
1111 struct numa_group __rcu *numa_group;
1114 * numa_faults is an array split into four regions:
1115 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1116 * in this precise order.
1118 * faults_memory: Exponential decaying average of faults on a per-node
1119 * basis. Scheduling placement decisions are made based on these
1120 * counts. The values remain static for the duration of a PTE scan.
1121 * faults_cpu: Track the nodes the process was running on when a NUMA
1122 * hinting fault was incurred.
1123 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1124 * during the current scan window. When the scan completes, the counts
1125 * in faults_memory and faults_cpu decay and these values are copied.
1127 unsigned long *numa_faults;
1128 unsigned long total_numa_faults;
1131 * numa_faults_locality tracks if faults recorded during the last
1132 * scan window were remote/local or failed to migrate. The task scan
1133 * period is adapted based on the locality of the faults with different
1134 * weights depending on whether they were shared or private faults
1136 unsigned long numa_faults_locality[3];
1138 unsigned long numa_pages_migrated;
1139 #endif /* CONFIG_NUMA_BALANCING */
1142 struct rseq __user *rseq;
1145 * RmW on rseq_event_mask must be performed atomically
1146 * with respect to preemption.
1148 unsigned long rseq_event_mask;
1151 struct tlbflush_unmap_batch tlb_ubc;
1154 refcount_t rcu_users;
1155 struct rcu_head rcu;
1158 /* Cache last used pipe for splice(): */
1159 struct pipe_inode_info *splice_pipe;
1161 struct page_frag task_frag;
1163 #ifdef CONFIG_TASK_DELAY_ACCT
1164 struct task_delay_info *delays;
1167 #ifdef CONFIG_FAULT_INJECTION
1169 unsigned int fail_nth;
1172 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1173 * balance_dirty_pages() for a dirty throttling pause:
1176 int nr_dirtied_pause;
1177 /* Start of a write-and-pause period: */
1178 unsigned long dirty_paused_when;
1180 #ifdef CONFIG_LATENCYTOP
1181 int latency_record_count;
1182 struct latency_record latency_record[LT_SAVECOUNT];
1185 * Time slack values; these are used to round up poll() and
1186 * select() etc timeout values. These are in nanoseconds.
1189 u64 default_timer_slack_ns;
1192 unsigned int kasan_depth;
1195 struct kcsan_ctx kcsan_ctx;
1198 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1199 /* Index of current stored address in ret_stack: */
1203 /* Stack of return addresses for return function tracing: */
1204 struct ftrace_ret_stack *ret_stack;
1206 /* Timestamp for last schedule: */
1207 unsigned long long ftrace_timestamp;
1210 * Number of functions that haven't been traced
1211 * because of depth overrun:
1213 atomic_t trace_overrun;
1215 /* Pause tracing: */
1216 atomic_t tracing_graph_pause;
1219 #ifdef CONFIG_TRACING
1220 /* State flags for use by tracers: */
1221 unsigned long trace;
1223 /* Bitmask and counter of trace recursion: */
1224 unsigned long trace_recursion;
1225 #endif /* CONFIG_TRACING */
1228 /* See kernel/kcov.c for more details. */
1230 /* Coverage collection mode enabled for this task (0 if disabled): */
1231 unsigned int kcov_mode;
1233 /* Size of the kcov_area: */
1234 unsigned int kcov_size;
1236 /* Buffer for coverage collection: */
1239 /* KCOV descriptor wired with this task or NULL: */
1242 /* KCOV common handle for remote coverage collection: */
1245 /* KCOV sequence number: */
1248 /* Collect coverage from softirq context: */
1249 unsigned int kcov_softirq;
1253 struct mem_cgroup *memcg_in_oom;
1254 gfp_t memcg_oom_gfp_mask;
1255 int memcg_oom_order;
1257 /* Number of pages to reclaim on returning to userland: */
1258 unsigned int memcg_nr_pages_over_high;
1260 /* Used by memcontrol for targeted memcg charge: */
1261 struct mem_cgroup *active_memcg;
1264 #ifdef CONFIG_BLK_CGROUP
1265 struct request_queue *throttle_queue;
1268 #ifdef CONFIG_UPROBES
1269 struct uprobe_task *utask;
1271 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1272 unsigned int sequential_io;
1273 unsigned int sequential_io_avg;
1275 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1276 unsigned long task_state_change;
1278 int pagefault_disabled;
1280 struct task_struct *oom_reaper_list;
1282 #ifdef CONFIG_VMAP_STACK
1283 struct vm_struct *stack_vm_area;
1285 #ifdef CONFIG_THREAD_INFO_IN_TASK
1286 /* A live task holds one reference: */
1287 refcount_t stack_refcount;
1289 #ifdef CONFIG_LIVEPATCH
1292 #ifdef CONFIG_SECURITY
1293 /* Used by LSM modules for access restriction: */
1297 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1298 unsigned long lowest_stack;
1299 unsigned long prev_lowest_stack;
1302 #ifdef CONFIG_X86_MCE
1306 __mce_reserved : 62;
1307 struct callback_head mce_kill_me;
1311 * New fields for task_struct should be added above here, so that
1312 * they are included in the randomized portion of task_struct.
1314 randomized_struct_fields_end
1316 /* CPU-specific state of this task: */
1317 struct thread_struct thread;
1320 * WARNING: on x86, 'thread_struct' contains a variable-sized
1321 * structure. It *MUST* be at the end of 'task_struct'.
1323 * Do not put anything below here!
1327 static inline struct pid *task_pid(struct task_struct *task)
1329 return task->thread_pid;
1333 * the helpers to get the task's different pids as they are seen
1334 * from various namespaces
1336 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1337 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1339 * task_xid_nr_ns() : id seen from the ns specified;
1341 * see also pid_nr() etc in include/linux/pid.h
1343 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1345 static inline pid_t task_pid_nr(struct task_struct *tsk)
1350 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1352 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1355 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1357 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1361 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1367 * pid_alive - check that a task structure is not stale
1368 * @p: Task structure to be checked.
1370 * Test if a process is not yet dead (at most zombie state)
1371 * If pid_alive fails, then pointers within the task structure
1372 * can be stale and must not be dereferenced.
1374 * Return: 1 if the process is alive. 0 otherwise.
1376 static inline int pid_alive(const struct task_struct *p)
1378 return p->thread_pid != NULL;
1381 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1383 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1386 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1388 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1392 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1394 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1397 static inline pid_t task_session_vnr(struct task_struct *tsk)
1399 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1402 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1404 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1407 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1409 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1412 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1418 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1424 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1426 return task_ppid_nr_ns(tsk, &init_pid_ns);
1429 /* Obsolete, do not use: */
1430 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1432 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1435 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1436 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1438 static inline unsigned int task_state_index(struct task_struct *tsk)
1440 unsigned int tsk_state = READ_ONCE(tsk->state);
1441 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1443 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1445 if (tsk_state == TASK_IDLE)
1446 state = TASK_REPORT_IDLE;
1451 static inline char task_index_to_char(unsigned int state)
1453 static const char state_char[] = "RSDTtXZPI";
1455 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1457 return state_char[state];
1460 static inline char task_state_to_char(struct task_struct *tsk)
1462 return task_index_to_char(task_state_index(tsk));
1466 * is_global_init - check if a task structure is init. Since init
1467 * is free to have sub-threads we need to check tgid.
1468 * @tsk: Task structure to be checked.
1470 * Check if a task structure is the first user space task the kernel created.
1472 * Return: 1 if the task structure is init. 0 otherwise.
1474 static inline int is_global_init(struct task_struct *tsk)
1476 return task_tgid_nr(tsk) == 1;
1479 extern struct pid *cad_pid;
1484 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1485 #define PF_EXITING 0x00000004 /* Getting shut down */
1486 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1487 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1488 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1489 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1490 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1491 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1492 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1493 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1494 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1495 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1496 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1497 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1498 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1499 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1500 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1501 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1502 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1503 * I am cleaning dirty pages from some other bdi. */
1504 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1505 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1506 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1507 #define PF_UMH 0x02000000 /* I'm an Usermodehelper process */
1508 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1509 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1510 #define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1511 #define PF_IO_WORKER 0x20000000 /* Task is an IO worker */
1512 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1513 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1516 * Only the _current_ task can read/write to tsk->flags, but other
1517 * tasks can access tsk->flags in readonly mode for example
1518 * with tsk_used_math (like during threaded core dumping).
1519 * There is however an exception to this rule during ptrace
1520 * or during fork: the ptracer task is allowed to write to the
1521 * child->flags of its traced child (same goes for fork, the parent
1522 * can write to the child->flags), because we're guaranteed the
1523 * child is not running and in turn not changing child->flags
1524 * at the same time the parent does it.
1526 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1527 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1528 #define clear_used_math() clear_stopped_child_used_math(current)
1529 #define set_used_math() set_stopped_child_used_math(current)
1531 #define conditional_stopped_child_used_math(condition, child) \
1532 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1534 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1536 #define copy_to_stopped_child_used_math(child) \
1537 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1539 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1540 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1541 #define used_math() tsk_used_math(current)
1543 static inline bool is_percpu_thread(void)
1546 return (current->flags & PF_NO_SETAFFINITY) &&
1547 (current->nr_cpus_allowed == 1);
1553 /* Per-process atomic flags. */
1554 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1555 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1556 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1557 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1558 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1559 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1560 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1561 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1563 #define TASK_PFA_TEST(name, func) \
1564 static inline bool task_##func(struct task_struct *p) \
1565 { return test_bit(PFA_##name, &p->atomic_flags); }
1567 #define TASK_PFA_SET(name, func) \
1568 static inline void task_set_##func(struct task_struct *p) \
1569 { set_bit(PFA_##name, &p->atomic_flags); }
1571 #define TASK_PFA_CLEAR(name, func) \
1572 static inline void task_clear_##func(struct task_struct *p) \
1573 { clear_bit(PFA_##name, &p->atomic_flags); }
1575 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1576 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1578 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1579 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1580 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1582 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1583 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1584 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1586 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1587 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1588 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1590 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1591 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1592 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1594 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1595 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1597 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1598 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1599 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1601 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1602 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1605 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1607 current->flags &= ~flags;
1608 current->flags |= orig_flags & flags;
1611 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1612 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1614 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1615 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1617 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1620 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1622 if (!cpumask_test_cpu(0, new_mask))
1628 extern int yield_to(struct task_struct *p, bool preempt);
1629 extern void set_user_nice(struct task_struct *p, long nice);
1630 extern int task_prio(const struct task_struct *p);
1633 * task_nice - return the nice value of a given task.
1634 * @p: the task in question.
1636 * Return: The nice value [ -20 ... 0 ... 19 ].
1638 static inline int task_nice(const struct task_struct *p)
1640 return PRIO_TO_NICE((p)->static_prio);
1643 extern int can_nice(const struct task_struct *p, const int nice);
1644 extern int task_curr(const struct task_struct *p);
1645 extern int idle_cpu(int cpu);
1646 extern int available_idle_cpu(int cpu);
1647 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1648 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1649 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1650 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1651 extern struct task_struct *idle_task(int cpu);
1654 * is_idle_task - is the specified task an idle task?
1655 * @p: the task in question.
1657 * Return: 1 if @p is an idle task. 0 otherwise.
1659 static inline bool is_idle_task(const struct task_struct *p)
1661 return !!(p->flags & PF_IDLE);
1664 extern struct task_struct *curr_task(int cpu);
1665 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1669 union thread_union {
1670 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1671 struct task_struct task;
1673 #ifndef CONFIG_THREAD_INFO_IN_TASK
1674 struct thread_info thread_info;
1676 unsigned long stack[THREAD_SIZE/sizeof(long)];
1679 #ifndef CONFIG_THREAD_INFO_IN_TASK
1680 extern struct thread_info init_thread_info;
1683 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1685 #ifdef CONFIG_THREAD_INFO_IN_TASK
1686 static inline struct thread_info *task_thread_info(struct task_struct *task)
1688 return &task->thread_info;
1690 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1691 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1695 * find a task by one of its numerical ids
1697 * find_task_by_pid_ns():
1698 * finds a task by its pid in the specified namespace
1699 * find_task_by_vpid():
1700 * finds a task by its virtual pid
1702 * see also find_vpid() etc in include/linux/pid.h
1705 extern struct task_struct *find_task_by_vpid(pid_t nr);
1706 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1709 * find a task by its virtual pid and get the task struct
1711 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1713 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1714 extern int wake_up_process(struct task_struct *tsk);
1715 extern void wake_up_new_task(struct task_struct *tsk);
1718 extern void kick_process(struct task_struct *tsk);
1720 static inline void kick_process(struct task_struct *tsk) { }
1723 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1725 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1727 __set_task_comm(tsk, from, false);
1730 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1731 #define get_task_comm(buf, tsk) ({ \
1732 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1733 __get_task_comm(buf, sizeof(buf), tsk); \
1737 static __always_inline void scheduler_ipi(void)
1740 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1741 * TIF_NEED_RESCHED remotely (for the first time) will also send
1744 preempt_fold_need_resched();
1746 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1748 static inline void scheduler_ipi(void) { }
1749 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1756 * Set thread flags in other task's structures.
1757 * See asm/thread_info.h for TIF_xxxx flags available:
1759 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1761 set_ti_thread_flag(task_thread_info(tsk), flag);
1764 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1766 clear_ti_thread_flag(task_thread_info(tsk), flag);
1769 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1772 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1775 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1777 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1780 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1782 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1785 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1787 return test_ti_thread_flag(task_thread_info(tsk), flag);
1790 static inline void set_tsk_need_resched(struct task_struct *tsk)
1792 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1795 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1797 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1800 static inline int test_tsk_need_resched(struct task_struct *tsk)
1802 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1806 * cond_resched() and cond_resched_lock(): latency reduction via
1807 * explicit rescheduling in places that are safe. The return
1808 * value indicates whether a reschedule was done in fact.
1809 * cond_resched_lock() will drop the spinlock before scheduling,
1811 #ifndef CONFIG_PREEMPTION
1812 extern int _cond_resched(void);
1814 static inline int _cond_resched(void) { return 0; }
1817 #define cond_resched() ({ \
1818 ___might_sleep(__FILE__, __LINE__, 0); \
1822 extern int __cond_resched_lock(spinlock_t *lock);
1824 #define cond_resched_lock(lock) ({ \
1825 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1826 __cond_resched_lock(lock); \
1829 static inline void cond_resched_rcu(void)
1831 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1839 * Does a critical section need to be broken due to another
1840 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1841 * but a general need for low latency)
1843 static inline int spin_needbreak(spinlock_t *lock)
1845 #ifdef CONFIG_PREEMPTION
1846 return spin_is_contended(lock);
1852 static __always_inline bool need_resched(void)
1854 return unlikely(tif_need_resched());
1858 * Wrappers for p->thread_info->cpu access. No-op on UP.
1862 static inline unsigned int task_cpu(const struct task_struct *p)
1864 #ifdef CONFIG_THREAD_INFO_IN_TASK
1865 return READ_ONCE(p->cpu);
1867 return READ_ONCE(task_thread_info(p)->cpu);
1871 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1875 static inline unsigned int task_cpu(const struct task_struct *p)
1880 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1884 #endif /* CONFIG_SMP */
1887 * In order to reduce various lock holder preemption latencies provide an
1888 * interface to see if a vCPU is currently running or not.
1890 * This allows us to terminate optimistic spin loops and block, analogous to
1891 * the native optimistic spin heuristic of testing if the lock owner task is
1894 #ifndef vcpu_is_preempted
1895 static inline bool vcpu_is_preempted(int cpu)
1901 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1902 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1904 #ifndef TASK_SIZE_OF
1905 #define TASK_SIZE_OF(tsk) TASK_SIZE
1911 * Map the event mask on the user-space ABI enum rseq_cs_flags
1912 * for direct mask checks.
1914 enum rseq_event_mask_bits {
1915 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1916 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1917 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1920 enum rseq_event_mask {
1921 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1922 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1923 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1926 static inline void rseq_set_notify_resume(struct task_struct *t)
1929 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1932 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1934 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1935 struct pt_regs *regs)
1938 __rseq_handle_notify_resume(ksig, regs);
1941 static inline void rseq_signal_deliver(struct ksignal *ksig,
1942 struct pt_regs *regs)
1945 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
1947 rseq_handle_notify_resume(ksig, regs);
1950 /* rseq_preempt() requires preemption to be disabled. */
1951 static inline void rseq_preempt(struct task_struct *t)
1953 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1954 rseq_set_notify_resume(t);
1957 /* rseq_migrate() requires preemption to be disabled. */
1958 static inline void rseq_migrate(struct task_struct *t)
1960 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1961 rseq_set_notify_resume(t);
1965 * If parent process has a registered restartable sequences area, the
1966 * child inherits. Unregister rseq for a clone with CLONE_VM set.
1968 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1970 if (clone_flags & CLONE_VM) {
1973 t->rseq_event_mask = 0;
1975 t->rseq = current->rseq;
1976 t->rseq_sig = current->rseq_sig;
1977 t->rseq_event_mask = current->rseq_event_mask;
1981 static inline void rseq_execve(struct task_struct *t)
1985 t->rseq_event_mask = 0;
1990 static inline void rseq_set_notify_resume(struct task_struct *t)
1993 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1994 struct pt_regs *regs)
1997 static inline void rseq_signal_deliver(struct ksignal *ksig,
1998 struct pt_regs *regs)
2001 static inline void rseq_preempt(struct task_struct *t)
2004 static inline void rseq_migrate(struct task_struct *t)
2007 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2010 static inline void rseq_execve(struct task_struct *t)
2016 void __exit_umh(struct task_struct *tsk);
2018 static inline void exit_umh(struct task_struct *tsk)
2020 if (unlikely(tsk->flags & PF_UMH))
2024 #ifdef CONFIG_DEBUG_RSEQ
2026 void rseq_syscall(struct pt_regs *regs);
2030 static inline void rseq_syscall(struct pt_regs *regs)
2036 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2037 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2038 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2040 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2041 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2042 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2044 int sched_trace_rq_cpu(struct rq *rq);
2046 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);