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/mutex.h>
18 #include <linux/plist.h>
19 #include <linux/hrtimer.h>
20 #include <linux/irqflags.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/syscall_user_dispatch.h>
31 #include <linux/mm_types_task.h>
32 #include <linux/task_io_accounting.h>
33 #include <linux/posix-timers.h>
34 #include <linux/rseq.h>
35 #include <linux/seqlock.h>
36 #include <linux/kcsan.h>
37 #include <asm/kmap_size.h>
39 /* task_struct member predeclarations (sorted alphabetically): */
41 struct backing_dev_info;
44 struct bpf_local_storage;
45 struct capture_control;
48 struct futex_pi_state;
54 struct perf_event_context;
56 struct pipe_inode_info;
59 struct robust_list_head;
65 struct sighand_struct;
67 struct task_delay_info;
71 * Task state bitmask. NOTE! These bits are also
72 * encoded in fs/proc/array.c: get_task_state().
74 * We have two separate sets of flags: task->state
75 * is about runnability, while task->exit_state are
76 * about the task exiting. Confusing, but this way
77 * modifying one set can't modify the other one by
81 /* Used in tsk->state: */
82 #define TASK_RUNNING 0x0000
83 #define TASK_INTERRUPTIBLE 0x0001
84 #define TASK_UNINTERRUPTIBLE 0x0002
85 #define __TASK_STOPPED 0x0004
86 #define __TASK_TRACED 0x0008
87 /* Used in tsk->exit_state: */
88 #define EXIT_DEAD 0x0010
89 #define EXIT_ZOMBIE 0x0020
90 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
91 /* Used in tsk->state again: */
92 #define TASK_PARKED 0x0040
93 #define TASK_DEAD 0x0080
94 #define TASK_WAKEKILL 0x0100
95 #define TASK_WAKING 0x0200
96 #define TASK_NOLOAD 0x0400
97 #define TASK_NEW 0x0800
98 /* RT specific auxilliary flag to mark RT lock waiters */
99 #define TASK_RTLOCK_WAIT 0x1000
100 #define TASK_STATE_MAX 0x2000
102 /* Convenience macros for the sake of set_current_state: */
103 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
104 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
105 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
107 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
109 /* Convenience macros for the sake of wake_up(): */
110 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
112 /* get_task_state(): */
113 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
114 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
115 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
118 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
120 #define task_is_traced(task) ((READ_ONCE(task->__state) & __TASK_TRACED) != 0)
122 #define task_is_stopped(task) ((READ_ONCE(task->__state) & __TASK_STOPPED) != 0)
124 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->__state) & (__TASK_STOPPED | __TASK_TRACED)) != 0)
126 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
129 * Special states are those that do not use the normal wait-loop pattern. See
130 * the comment with set_special_state().
132 #define is_special_task_state(state) \
133 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
135 #define __set_current_state(state_value) \
137 WARN_ON_ONCE(is_special_task_state(state_value));\
138 current->task_state_change = _THIS_IP_; \
139 WRITE_ONCE(current->__state, (state_value)); \
142 #define set_current_state(state_value) \
144 WARN_ON_ONCE(is_special_task_state(state_value));\
145 current->task_state_change = _THIS_IP_; \
146 smp_store_mb(current->__state, (state_value)); \
149 #define set_special_state(state_value) \
151 unsigned long flags; /* may shadow */ \
152 WARN_ON_ONCE(!is_special_task_state(state_value)); \
153 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
154 current->task_state_change = _THIS_IP_; \
155 WRITE_ONCE(current->__state, (state_value)); \
156 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
160 * set_current_state() includes a barrier so that the write of current->state
161 * is correctly serialised wrt the caller's subsequent test of whether to
165 * set_current_state(TASK_UNINTERRUPTIBLE);
171 * __set_current_state(TASK_RUNNING);
173 * If the caller does not need such serialisation (because, for instance, the
174 * CONDITION test and condition change and wakeup are under the same lock) then
175 * use __set_current_state().
177 * The above is typically ordered against the wakeup, which does:
180 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
182 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
183 * accessing p->state.
185 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
186 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
187 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
189 * However, with slightly different timing the wakeup TASK_RUNNING store can
190 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
191 * a problem either because that will result in one extra go around the loop
192 * and our @cond test will save the day.
194 * Also see the comments of try_to_wake_up().
196 #define __set_current_state(state_value) \
197 WRITE_ONCE(current->__state, (state_value))
199 #define set_current_state(state_value) \
200 smp_store_mb(current->__state, (state_value))
203 * set_special_state() should be used for those states when the blocking task
204 * can not use the regular condition based wait-loop. In that case we must
205 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
206 * will not collide with our state change.
208 #define set_special_state(state_value) \
210 unsigned long flags; /* may shadow */ \
211 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
212 WRITE_ONCE(current->__state, (state_value)); \
213 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
218 #define get_current_state() READ_ONCE(current->__state)
220 /* Task command name length: */
221 #define TASK_COMM_LEN 16
223 extern void scheduler_tick(void);
225 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
227 extern long schedule_timeout(long timeout);
228 extern long schedule_timeout_interruptible(long timeout);
229 extern long schedule_timeout_killable(long timeout);
230 extern long schedule_timeout_uninterruptible(long timeout);
231 extern long schedule_timeout_idle(long timeout);
232 asmlinkage void schedule(void);
233 extern void schedule_preempt_disabled(void);
234 asmlinkage void preempt_schedule_irq(void);
236 extern int __must_check io_schedule_prepare(void);
237 extern void io_schedule_finish(int token);
238 extern long io_schedule_timeout(long timeout);
239 extern void io_schedule(void);
242 * struct prev_cputime - snapshot of system and user cputime
243 * @utime: time spent in user mode
244 * @stime: time spent in system mode
245 * @lock: protects the above two fields
247 * Stores previous user/system time values such that we can guarantee
250 struct prev_cputime {
251 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
259 /* Task is sleeping or running in a CPU with VTIME inactive: */
263 /* Task runs in kernelspace in a CPU with VTIME active: */
265 /* Task runs in userspace in a CPU with VTIME active: */
267 /* Task runs as guests in a CPU with VTIME active: */
273 unsigned long long starttime;
274 enum vtime_state state;
282 * Utilization clamp constraints.
283 * @UCLAMP_MIN: Minimum utilization
284 * @UCLAMP_MAX: Maximum utilization
285 * @UCLAMP_CNT: Utilization clamp constraints count
294 extern struct root_domain def_root_domain;
295 extern struct mutex sched_domains_mutex;
299 #ifdef CONFIG_SCHED_INFO
300 /* Cumulative counters: */
302 /* # of times we have run on this CPU: */
303 unsigned long pcount;
305 /* Time spent waiting on a runqueue: */
306 unsigned long long run_delay;
310 /* When did we last run on a CPU? */
311 unsigned long long last_arrival;
313 /* When were we last queued to run? */
314 unsigned long long last_queued;
316 #endif /* CONFIG_SCHED_INFO */
320 * Integer metrics need fixed point arithmetic, e.g., sched/fair
321 * has a few: load, load_avg, util_avg, freq, and capacity.
323 * We define a basic fixed point arithmetic range, and then formalize
324 * all these metrics based on that basic range.
326 # define SCHED_FIXEDPOINT_SHIFT 10
327 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
329 /* Increase resolution of cpu_capacity calculations */
330 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
331 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
334 unsigned long weight;
339 * struct util_est - Estimation utilization of FAIR tasks
340 * @enqueued: instantaneous estimated utilization of a task/cpu
341 * @ewma: the Exponential Weighted Moving Average (EWMA)
342 * utilization of a task
344 * Support data structure to track an Exponential Weighted Moving Average
345 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
346 * average each time a task completes an activation. Sample's weight is chosen
347 * so that the EWMA will be relatively insensitive to transient changes to the
350 * The enqueued attribute has a slightly different meaning for tasks and cpus:
351 * - task: the task's util_avg at last task dequeue time
352 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
353 * Thus, the util_est.enqueued of a task represents the contribution on the
354 * estimated utilization of the CPU where that task is currently enqueued.
356 * Only for tasks we track a moving average of the past instantaneous
357 * estimated utilization. This allows to absorb sporadic drops in utilization
358 * of an otherwise almost periodic task.
360 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
361 * updates. When a task is dequeued, its util_est should not be updated if its
362 * util_avg has not been updated in the meantime.
363 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
364 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
365 * for a task) it is safe to use MSB.
368 unsigned int enqueued;
370 #define UTIL_EST_WEIGHT_SHIFT 2
371 #define UTIL_AVG_UNCHANGED 0x80000000
372 } __attribute__((__aligned__(sizeof(u64))));
375 * The load/runnable/util_avg accumulates an infinite geometric series
376 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
378 * [load_avg definition]
380 * load_avg = runnable% * scale_load_down(load)
382 * [runnable_avg definition]
384 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
386 * [util_avg definition]
388 * util_avg = running% * SCHED_CAPACITY_SCALE
390 * where runnable% is the time ratio that a sched_entity is runnable and
391 * running% the time ratio that a sched_entity is running.
393 * For cfs_rq, they are the aggregated values of all runnable and blocked
396 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
397 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
398 * for computing those signals (see update_rq_clock_pelt())
400 * N.B., the above ratios (runnable% and running%) themselves are in the
401 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
402 * to as large a range as necessary. This is for example reflected by
403 * util_avg's SCHED_CAPACITY_SCALE.
407 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
408 * with the highest load (=88761), always runnable on a single cfs_rq,
409 * and should not overflow as the number already hits PID_MAX_LIMIT.
411 * For all other cases (including 32-bit kernels), struct load_weight's
412 * weight will overflow first before we do, because:
414 * Max(load_avg) <= Max(load.weight)
416 * Then it is the load_weight's responsibility to consider overflow
420 u64 last_update_time;
425 unsigned long load_avg;
426 unsigned long runnable_avg;
427 unsigned long util_avg;
428 struct util_est util_est;
429 } ____cacheline_aligned;
431 struct sched_statistics {
432 #ifdef CONFIG_SCHEDSTATS
442 s64 sum_sleep_runtime;
449 u64 nr_migrations_cold;
450 u64 nr_failed_migrations_affine;
451 u64 nr_failed_migrations_running;
452 u64 nr_failed_migrations_hot;
453 u64 nr_forced_migrations;
457 u64 nr_wakeups_migrate;
458 u64 nr_wakeups_local;
459 u64 nr_wakeups_remote;
460 u64 nr_wakeups_affine;
461 u64 nr_wakeups_affine_attempts;
462 u64 nr_wakeups_passive;
467 struct sched_entity {
468 /* For load-balancing: */
469 struct load_weight load;
470 struct rb_node run_node;
471 struct list_head group_node;
475 u64 sum_exec_runtime;
477 u64 prev_sum_exec_runtime;
481 struct sched_statistics statistics;
483 #ifdef CONFIG_FAIR_GROUP_SCHED
485 struct sched_entity *parent;
486 /* rq on which this entity is (to be) queued: */
487 struct cfs_rq *cfs_rq;
488 /* rq "owned" by this entity/group: */
490 /* cached value of my_q->h_nr_running */
491 unsigned long runnable_weight;
496 * Per entity load average tracking.
498 * Put into separate cache line so it does not
499 * collide with read-mostly values above.
501 struct sched_avg avg;
505 struct sched_rt_entity {
506 struct list_head run_list;
507 unsigned long timeout;
508 unsigned long watchdog_stamp;
509 unsigned int time_slice;
510 unsigned short on_rq;
511 unsigned short on_list;
513 struct sched_rt_entity *back;
514 #ifdef CONFIG_RT_GROUP_SCHED
515 struct sched_rt_entity *parent;
516 /* rq on which this entity is (to be) queued: */
518 /* rq "owned" by this entity/group: */
521 } __randomize_layout;
523 struct sched_dl_entity {
524 struct rb_node rb_node;
527 * Original scheduling parameters. Copied here from sched_attr
528 * during sched_setattr(), they will remain the same until
529 * the next sched_setattr().
531 u64 dl_runtime; /* Maximum runtime for each instance */
532 u64 dl_deadline; /* Relative deadline of each instance */
533 u64 dl_period; /* Separation of two instances (period) */
534 u64 dl_bw; /* dl_runtime / dl_period */
535 u64 dl_density; /* dl_runtime / dl_deadline */
538 * Actual scheduling parameters. Initialized with the values above,
539 * they are continuously updated during task execution. Note that
540 * the remaining runtime could be < 0 in case we are in overrun.
542 s64 runtime; /* Remaining runtime for this instance */
543 u64 deadline; /* Absolute deadline for this instance */
544 unsigned int flags; /* Specifying the scheduler behaviour */
549 * @dl_throttled tells if we exhausted the runtime. If so, the
550 * task has to wait for a replenishment to be performed at the
551 * next firing of dl_timer.
553 * @dl_boosted tells if we are boosted due to DI. If so we are
554 * outside bandwidth enforcement mechanism (but only until we
555 * exit the critical section);
557 * @dl_yielded tells if task gave up the CPU before consuming
558 * all its available runtime during the last job.
560 * @dl_non_contending tells if the task is inactive while still
561 * contributing to the active utilization. In other words, it
562 * indicates if the inactive timer has been armed and its handler
563 * has not been executed yet. This flag is useful to avoid race
564 * conditions between the inactive timer handler and the wakeup
567 * @dl_overrun tells if the task asked to be informed about runtime
570 unsigned int dl_throttled : 1;
571 unsigned int dl_yielded : 1;
572 unsigned int dl_non_contending : 1;
573 unsigned int dl_overrun : 1;
576 * Bandwidth enforcement timer. Each -deadline task has its
577 * own bandwidth to be enforced, thus we need one timer per task.
579 struct hrtimer dl_timer;
582 * Inactive timer, responsible for decreasing the active utilization
583 * at the "0-lag time". When a -deadline task blocks, it contributes
584 * to GRUB's active utilization until the "0-lag time", hence a
585 * timer is needed to decrease the active utilization at the correct
588 struct hrtimer inactive_timer;
590 #ifdef CONFIG_RT_MUTEXES
592 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
593 * pi_se points to the donor, otherwise points to the dl_se it belongs
594 * to (the original one/itself).
596 struct sched_dl_entity *pi_se;
600 #ifdef CONFIG_UCLAMP_TASK
601 /* Number of utilization clamp buckets (shorter alias) */
602 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
605 * Utilization clamp for a scheduling entity
606 * @value: clamp value "assigned" to a se
607 * @bucket_id: bucket index corresponding to the "assigned" value
608 * @active: the se is currently refcounted in a rq's bucket
609 * @user_defined: the requested clamp value comes from user-space
611 * The bucket_id is the index of the clamp bucket matching the clamp value
612 * which is pre-computed and stored to avoid expensive integer divisions from
615 * The active bit is set whenever a task has got an "effective" value assigned,
616 * which can be different from the clamp value "requested" from user-space.
617 * This allows to know a task is refcounted in the rq's bucket corresponding
618 * to the "effective" bucket_id.
620 * The user_defined bit is set whenever a task has got a task-specific clamp
621 * value requested from userspace, i.e. the system defaults apply to this task
622 * just as a restriction. This allows to relax default clamps when a less
623 * restrictive task-specific value has been requested, thus allowing to
624 * implement a "nice" semantic. For example, a task running with a 20%
625 * default boost can still drop its own boosting to 0%.
628 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
629 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
630 unsigned int active : 1;
631 unsigned int user_defined : 1;
633 #endif /* CONFIG_UCLAMP_TASK */
639 u8 exp_hint; /* Hint for performance. */
640 u8 need_mb; /* Readers need smp_mb(). */
642 u32 s; /* Set of bits. */
645 enum perf_event_task_context {
646 perf_invalid_context = -1,
649 perf_nr_task_contexts,
653 struct wake_q_node *next;
657 #ifdef CONFIG_KMAP_LOCAL
659 pte_t pteval[KM_MAX_IDX];
664 #ifdef CONFIG_THREAD_INFO_IN_TASK
666 * For reasons of header soup (see current_thread_info()), this
667 * must be the first element of task_struct.
669 struct thread_info thread_info;
671 unsigned int __state;
674 * This begins the randomizable portion of task_struct. Only
675 * scheduling-critical items should be added above here.
677 randomized_struct_fields_start
681 /* Per task flags (PF_*), defined further below: */
687 struct __call_single_node wake_entry;
688 #ifdef CONFIG_THREAD_INFO_IN_TASK
692 unsigned int wakee_flips;
693 unsigned long wakee_flip_decay_ts;
694 struct task_struct *last_wakee;
697 * recent_used_cpu is initially set as the last CPU used by a task
698 * that wakes affine another task. Waker/wakee relationships can
699 * push tasks around a CPU where each wakeup moves to the next one.
700 * Tracking a recently used CPU allows a quick search for a recently
701 * used CPU that may be idle.
711 unsigned int rt_priority;
713 const struct sched_class *sched_class;
714 struct sched_entity se;
715 struct sched_rt_entity rt;
716 struct sched_dl_entity dl;
718 #ifdef CONFIG_SCHED_CORE
719 struct rb_node core_node;
720 unsigned long core_cookie;
721 unsigned int core_occupation;
724 #ifdef CONFIG_CGROUP_SCHED
725 struct task_group *sched_task_group;
728 #ifdef CONFIG_UCLAMP_TASK
730 * Clamp values requested for a scheduling entity.
731 * Must be updated with task_rq_lock() held.
733 struct uclamp_se uclamp_req[UCLAMP_CNT];
735 * Effective clamp values used for a scheduling entity.
736 * Must be updated with task_rq_lock() held.
738 struct uclamp_se uclamp[UCLAMP_CNT];
741 #ifdef CONFIG_PREEMPT_NOTIFIERS
742 /* List of struct preempt_notifier: */
743 struct hlist_head preempt_notifiers;
746 #ifdef CONFIG_BLK_DEV_IO_TRACE
747 unsigned int btrace_seq;
752 const cpumask_t *cpus_ptr;
754 void *migration_pending;
756 unsigned short migration_disabled;
758 unsigned short migration_flags;
760 #ifdef CONFIG_PREEMPT_RCU
761 int rcu_read_lock_nesting;
762 union rcu_special rcu_read_unlock_special;
763 struct list_head rcu_node_entry;
764 struct rcu_node *rcu_blocked_node;
765 #endif /* #ifdef CONFIG_PREEMPT_RCU */
767 #ifdef CONFIG_TASKS_RCU
768 unsigned long rcu_tasks_nvcsw;
769 u8 rcu_tasks_holdout;
771 int rcu_tasks_idle_cpu;
772 struct list_head rcu_tasks_holdout_list;
773 #endif /* #ifdef CONFIG_TASKS_RCU */
775 #ifdef CONFIG_TASKS_TRACE_RCU
776 int trc_reader_nesting;
778 union rcu_special trc_reader_special;
779 bool trc_reader_checked;
780 struct list_head trc_holdout_list;
781 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
783 struct sched_info sched_info;
785 struct list_head tasks;
787 struct plist_node pushable_tasks;
788 struct rb_node pushable_dl_tasks;
791 struct mm_struct *mm;
792 struct mm_struct *active_mm;
794 /* Per-thread vma caching: */
795 struct vmacache vmacache;
797 #ifdef SPLIT_RSS_COUNTING
798 struct task_rss_stat rss_stat;
803 /* The signal sent when the parent dies: */
805 /* JOBCTL_*, siglock protected: */
806 unsigned long jobctl;
808 /* Used for emulating ABI behavior of previous Linux versions: */
809 unsigned int personality;
811 /* Scheduler bits, serialized by scheduler locks: */
812 unsigned sched_reset_on_fork:1;
813 unsigned sched_contributes_to_load:1;
814 unsigned sched_migrated:1;
816 unsigned sched_psi_wake_requeue:1;
819 /* Force alignment to the next boundary: */
822 /* Unserialized, strictly 'current' */
825 * This field must not be in the scheduler word above due to wakelist
826 * queueing no longer being serialized by p->on_cpu. However:
829 * schedule() if (p->on_rq && ..) // false
830 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
831 * deactivate_task() ttwu_queue_wakelist())
832 * p->on_rq = 0; p->sched_remote_wakeup = Y;
834 * guarantees all stores of 'current' are visible before
835 * ->sched_remote_wakeup gets used, so it can be in this word.
837 unsigned sched_remote_wakeup:1;
839 /* Bit to tell LSMs we're in execve(): */
840 unsigned in_execve:1;
841 unsigned in_iowait:1;
842 #ifndef TIF_RESTORE_SIGMASK
843 unsigned restore_sigmask:1;
846 unsigned in_user_fault:1;
848 #ifdef CONFIG_COMPAT_BRK
849 unsigned brk_randomized:1;
851 #ifdef CONFIG_CGROUPS
852 /* disallow userland-initiated cgroup migration */
853 unsigned no_cgroup_migration:1;
854 /* task is frozen/stopped (used by the cgroup freezer) */
857 #ifdef CONFIG_BLK_CGROUP
858 unsigned use_memdelay:1;
861 /* Stalled due to lack of memory */
862 unsigned in_memstall:1;
864 #ifdef CONFIG_PAGE_OWNER
865 /* Used by page_owner=on to detect recursion in page tracking. */
866 unsigned in_page_owner:1;
869 unsigned long atomic_flags; /* Flags requiring atomic access. */
871 struct restart_block restart_block;
876 #ifdef CONFIG_STACKPROTECTOR
877 /* Canary value for the -fstack-protector GCC feature: */
878 unsigned long stack_canary;
881 * Pointers to the (original) parent process, youngest child, younger sibling,
882 * older sibling, respectively. (p->father can be replaced with
883 * p->real_parent->pid)
886 /* Real parent process: */
887 struct task_struct __rcu *real_parent;
889 /* Recipient of SIGCHLD, wait4() reports: */
890 struct task_struct __rcu *parent;
893 * Children/sibling form the list of natural children:
895 struct list_head children;
896 struct list_head sibling;
897 struct task_struct *group_leader;
900 * 'ptraced' is the list of tasks this task is using ptrace() on.
902 * This includes both natural children and PTRACE_ATTACH targets.
903 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
905 struct list_head ptraced;
906 struct list_head ptrace_entry;
908 /* PID/PID hash table linkage. */
909 struct pid *thread_pid;
910 struct hlist_node pid_links[PIDTYPE_MAX];
911 struct list_head thread_group;
912 struct list_head thread_node;
914 struct completion *vfork_done;
916 /* CLONE_CHILD_SETTID: */
917 int __user *set_child_tid;
919 /* CLONE_CHILD_CLEARTID: */
920 int __user *clear_child_tid;
927 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
932 struct prev_cputime prev_cputime;
933 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
937 #ifdef CONFIG_NO_HZ_FULL
938 atomic_t tick_dep_mask;
940 /* Context switch counts: */
942 unsigned long nivcsw;
944 /* Monotonic time in nsecs: */
947 /* Boot based time in nsecs: */
950 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
951 unsigned long min_flt;
952 unsigned long maj_flt;
954 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
955 struct posix_cputimers posix_cputimers;
957 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
958 struct posix_cputimers_work posix_cputimers_work;
961 /* Process credentials: */
963 /* Tracer's credentials at attach: */
964 const struct cred __rcu *ptracer_cred;
966 /* Objective and real subjective task credentials (COW): */
967 const struct cred __rcu *real_cred;
969 /* Effective (overridable) subjective task credentials (COW): */
970 const struct cred __rcu *cred;
973 /* Cached requested key. */
974 struct key *cached_requested_key;
978 * executable name, excluding path.
980 * - normally initialized setup_new_exec()
981 * - access it with [gs]et_task_comm()
982 * - lock it with task_lock()
984 char comm[TASK_COMM_LEN];
986 struct nameidata *nameidata;
988 #ifdef CONFIG_SYSVIPC
989 struct sysv_sem sysvsem;
990 struct sysv_shm sysvshm;
992 #ifdef CONFIG_DETECT_HUNG_TASK
993 unsigned long last_switch_count;
994 unsigned long last_switch_time;
996 /* Filesystem information: */
997 struct fs_struct *fs;
999 /* Open file information: */
1000 struct files_struct *files;
1002 #ifdef CONFIG_IO_URING
1003 struct io_uring_task *io_uring;
1007 struct nsproxy *nsproxy;
1009 /* Signal handlers: */
1010 struct signal_struct *signal;
1011 struct sighand_struct __rcu *sighand;
1013 sigset_t real_blocked;
1014 /* Restored if set_restore_sigmask() was used: */
1015 sigset_t saved_sigmask;
1016 struct sigpending pending;
1017 unsigned long sas_ss_sp;
1019 unsigned int sas_ss_flags;
1021 struct callback_head *task_works;
1024 #ifdef CONFIG_AUDITSYSCALL
1025 struct audit_context *audit_context;
1028 unsigned int sessionid;
1030 struct seccomp seccomp;
1031 struct syscall_user_dispatch syscall_dispatch;
1033 /* Thread group tracking: */
1037 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1038 spinlock_t alloc_lock;
1040 /* Protection of the PI data structures: */
1041 raw_spinlock_t pi_lock;
1043 struct wake_q_node wake_q;
1045 #ifdef CONFIG_RT_MUTEXES
1046 /* PI waiters blocked on a rt_mutex held by this task: */
1047 struct rb_root_cached pi_waiters;
1048 /* Updated under owner's pi_lock and rq lock */
1049 struct task_struct *pi_top_task;
1050 /* Deadlock detection and priority inheritance handling: */
1051 struct rt_mutex_waiter *pi_blocked_on;
1054 #ifdef CONFIG_DEBUG_MUTEXES
1055 /* Mutex deadlock detection: */
1056 struct mutex_waiter *blocked_on;
1059 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1060 int non_block_count;
1063 #ifdef CONFIG_TRACE_IRQFLAGS
1064 struct irqtrace_events irqtrace;
1065 unsigned int hardirq_threaded;
1066 u64 hardirq_chain_key;
1067 int softirqs_enabled;
1068 int softirq_context;
1071 #ifdef CONFIG_PREEMPT_RT
1072 int softirq_disable_cnt;
1075 #ifdef CONFIG_LOCKDEP
1076 # define MAX_LOCK_DEPTH 48UL
1079 unsigned int lockdep_recursion;
1080 struct held_lock held_locks[MAX_LOCK_DEPTH];
1083 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1084 unsigned int in_ubsan;
1087 /* Journalling filesystem info: */
1090 /* Stacked block device info: */
1091 struct bio_list *bio_list;
1094 /* Stack plugging: */
1095 struct blk_plug *plug;
1099 struct reclaim_state *reclaim_state;
1101 struct backing_dev_info *backing_dev_info;
1103 struct io_context *io_context;
1105 #ifdef CONFIG_COMPACTION
1106 struct capture_control *capture_control;
1109 unsigned long ptrace_message;
1110 kernel_siginfo_t *last_siginfo;
1112 struct task_io_accounting ioac;
1114 /* Pressure stall state */
1115 unsigned int psi_flags;
1117 #ifdef CONFIG_TASK_XACCT
1118 /* Accumulated RSS usage: */
1120 /* Accumulated virtual memory usage: */
1122 /* stime + utime since last update: */
1125 #ifdef CONFIG_CPUSETS
1126 /* Protected by ->alloc_lock: */
1127 nodemask_t mems_allowed;
1128 /* Sequence number to catch updates: */
1129 seqcount_spinlock_t mems_allowed_seq;
1130 int cpuset_mem_spread_rotor;
1131 int cpuset_slab_spread_rotor;
1133 #ifdef CONFIG_CGROUPS
1134 /* Control Group info protected by css_set_lock: */
1135 struct css_set __rcu *cgroups;
1136 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1137 struct list_head cg_list;
1139 #ifdef CONFIG_X86_CPU_RESCTRL
1144 struct robust_list_head __user *robust_list;
1145 #ifdef CONFIG_COMPAT
1146 struct compat_robust_list_head __user *compat_robust_list;
1148 struct list_head pi_state_list;
1149 struct futex_pi_state *pi_state_cache;
1150 struct mutex futex_exit_mutex;
1151 unsigned int futex_state;
1153 #ifdef CONFIG_PERF_EVENTS
1154 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1155 struct mutex perf_event_mutex;
1156 struct list_head perf_event_list;
1158 #ifdef CONFIG_DEBUG_PREEMPT
1159 unsigned long preempt_disable_ip;
1162 /* Protected by alloc_lock: */
1163 struct mempolicy *mempolicy;
1165 short pref_node_fork;
1167 #ifdef CONFIG_NUMA_BALANCING
1169 unsigned int numa_scan_period;
1170 unsigned int numa_scan_period_max;
1171 int numa_preferred_nid;
1172 unsigned long numa_migrate_retry;
1173 /* Migration stamp: */
1175 u64 last_task_numa_placement;
1176 u64 last_sum_exec_runtime;
1177 struct callback_head numa_work;
1180 * This pointer is only modified for current in syscall and
1181 * pagefault context (and for tasks being destroyed), so it can be read
1182 * from any of the following contexts:
1183 * - RCU read-side critical section
1184 * - current->numa_group from everywhere
1185 * - task's runqueue locked, task not running
1187 struct numa_group __rcu *numa_group;
1190 * numa_faults is an array split into four regions:
1191 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1192 * in this precise order.
1194 * faults_memory: Exponential decaying average of faults on a per-node
1195 * basis. Scheduling placement decisions are made based on these
1196 * counts. The values remain static for the duration of a PTE scan.
1197 * faults_cpu: Track the nodes the process was running on when a NUMA
1198 * hinting fault was incurred.
1199 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1200 * during the current scan window. When the scan completes, the counts
1201 * in faults_memory and faults_cpu decay and these values are copied.
1203 unsigned long *numa_faults;
1204 unsigned long total_numa_faults;
1207 * numa_faults_locality tracks if faults recorded during the last
1208 * scan window were remote/local or failed to migrate. The task scan
1209 * period is adapted based on the locality of the faults with different
1210 * weights depending on whether they were shared or private faults
1212 unsigned long numa_faults_locality[3];
1214 unsigned long numa_pages_migrated;
1215 #endif /* CONFIG_NUMA_BALANCING */
1218 struct rseq __user *rseq;
1221 * RmW on rseq_event_mask must be performed atomically
1222 * with respect to preemption.
1224 unsigned long rseq_event_mask;
1227 struct tlbflush_unmap_batch tlb_ubc;
1230 refcount_t rcu_users;
1231 struct rcu_head rcu;
1234 /* Cache last used pipe for splice(): */
1235 struct pipe_inode_info *splice_pipe;
1237 struct page_frag task_frag;
1239 #ifdef CONFIG_TASK_DELAY_ACCT
1240 struct task_delay_info *delays;
1243 #ifdef CONFIG_FAULT_INJECTION
1245 unsigned int fail_nth;
1248 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1249 * balance_dirty_pages() for a dirty throttling pause:
1252 int nr_dirtied_pause;
1253 /* Start of a write-and-pause period: */
1254 unsigned long dirty_paused_when;
1256 #ifdef CONFIG_LATENCYTOP
1257 int latency_record_count;
1258 struct latency_record latency_record[LT_SAVECOUNT];
1261 * Time slack values; these are used to round up poll() and
1262 * select() etc timeout values. These are in nanoseconds.
1265 u64 default_timer_slack_ns;
1267 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1268 unsigned int kasan_depth;
1272 struct kcsan_ctx kcsan_ctx;
1273 #ifdef CONFIG_TRACE_IRQFLAGS
1274 struct irqtrace_events kcsan_save_irqtrace;
1278 #if IS_ENABLED(CONFIG_KUNIT)
1279 struct kunit *kunit_test;
1282 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1283 /* Index of current stored address in ret_stack: */
1287 /* Stack of return addresses for return function tracing: */
1288 struct ftrace_ret_stack *ret_stack;
1290 /* Timestamp for last schedule: */
1291 unsigned long long ftrace_timestamp;
1294 * Number of functions that haven't been traced
1295 * because of depth overrun:
1297 atomic_t trace_overrun;
1299 /* Pause tracing: */
1300 atomic_t tracing_graph_pause;
1303 #ifdef CONFIG_TRACING
1304 /* State flags for use by tracers: */
1305 unsigned long trace;
1307 /* Bitmask and counter of trace recursion: */
1308 unsigned long trace_recursion;
1309 #endif /* CONFIG_TRACING */
1312 /* See kernel/kcov.c for more details. */
1314 /* Coverage collection mode enabled for this task (0 if disabled): */
1315 unsigned int kcov_mode;
1317 /* Size of the kcov_area: */
1318 unsigned int kcov_size;
1320 /* Buffer for coverage collection: */
1323 /* KCOV descriptor wired with this task or NULL: */
1326 /* KCOV common handle for remote coverage collection: */
1329 /* KCOV sequence number: */
1332 /* Collect coverage from softirq context: */
1333 unsigned int kcov_softirq;
1337 struct mem_cgroup *memcg_in_oom;
1338 gfp_t memcg_oom_gfp_mask;
1339 int memcg_oom_order;
1341 /* Number of pages to reclaim on returning to userland: */
1342 unsigned int memcg_nr_pages_over_high;
1344 /* Used by memcontrol for targeted memcg charge: */
1345 struct mem_cgroup *active_memcg;
1348 #ifdef CONFIG_BLK_CGROUP
1349 struct request_queue *throttle_queue;
1352 #ifdef CONFIG_UPROBES
1353 struct uprobe_task *utask;
1355 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1356 unsigned int sequential_io;
1357 unsigned int sequential_io_avg;
1359 struct kmap_ctrl kmap_ctrl;
1360 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1361 unsigned long task_state_change;
1363 int pagefault_disabled;
1365 struct task_struct *oom_reaper_list;
1367 #ifdef CONFIG_VMAP_STACK
1368 struct vm_struct *stack_vm_area;
1370 #ifdef CONFIG_THREAD_INFO_IN_TASK
1371 /* A live task holds one reference: */
1372 refcount_t stack_refcount;
1374 #ifdef CONFIG_LIVEPATCH
1377 #ifdef CONFIG_SECURITY
1378 /* Used by LSM modules for access restriction: */
1381 #ifdef CONFIG_BPF_SYSCALL
1382 /* Used by BPF task local storage */
1383 struct bpf_local_storage __rcu *bpf_storage;
1386 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1387 unsigned long lowest_stack;
1388 unsigned long prev_lowest_stack;
1391 #ifdef CONFIG_X86_MCE
1392 void __user *mce_vaddr;
1397 __mce_reserved : 62;
1398 struct callback_head mce_kill_me;
1401 #ifdef CONFIG_KRETPROBES
1402 struct llist_head kretprobe_instances;
1406 * New fields for task_struct should be added above here, so that
1407 * they are included in the randomized portion of task_struct.
1409 randomized_struct_fields_end
1411 /* CPU-specific state of this task: */
1412 struct thread_struct thread;
1415 * WARNING: on x86, 'thread_struct' contains a variable-sized
1416 * structure. It *MUST* be at the end of 'task_struct'.
1418 * Do not put anything below here!
1422 static inline struct pid *task_pid(struct task_struct *task)
1424 return task->thread_pid;
1428 * the helpers to get the task's different pids as they are seen
1429 * from various namespaces
1431 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1432 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1434 * task_xid_nr_ns() : id seen from the ns specified;
1436 * see also pid_nr() etc in include/linux/pid.h
1438 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1440 static inline pid_t task_pid_nr(struct task_struct *tsk)
1445 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1447 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1450 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1452 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1456 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1462 * pid_alive - check that a task structure is not stale
1463 * @p: Task structure to be checked.
1465 * Test if a process is not yet dead (at most zombie state)
1466 * If pid_alive fails, then pointers within the task structure
1467 * can be stale and must not be dereferenced.
1469 * Return: 1 if the process is alive. 0 otherwise.
1471 static inline int pid_alive(const struct task_struct *p)
1473 return p->thread_pid != NULL;
1476 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1478 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1481 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1483 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1487 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1489 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1492 static inline pid_t task_session_vnr(struct task_struct *tsk)
1494 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1497 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1499 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1502 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1504 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1507 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1513 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1519 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1521 return task_ppid_nr_ns(tsk, &init_pid_ns);
1524 /* Obsolete, do not use: */
1525 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1527 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1530 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1531 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1533 static inline unsigned int task_state_index(struct task_struct *tsk)
1535 unsigned int tsk_state = READ_ONCE(tsk->__state);
1536 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1538 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1540 if (tsk_state == TASK_IDLE)
1541 state = TASK_REPORT_IDLE;
1546 static inline char task_index_to_char(unsigned int state)
1548 static const char state_char[] = "RSDTtXZPI";
1550 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1552 return state_char[state];
1555 static inline char task_state_to_char(struct task_struct *tsk)
1557 return task_index_to_char(task_state_index(tsk));
1561 * is_global_init - check if a task structure is init. Since init
1562 * is free to have sub-threads we need to check tgid.
1563 * @tsk: Task structure to be checked.
1565 * Check if a task structure is the first user space task the kernel created.
1567 * Return: 1 if the task structure is init. 0 otherwise.
1569 static inline int is_global_init(struct task_struct *tsk)
1571 return task_tgid_nr(tsk) == 1;
1574 extern struct pid *cad_pid;
1579 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1580 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1581 #define PF_EXITING 0x00000004 /* Getting shut down */
1582 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1583 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1584 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1585 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1586 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1587 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1588 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1589 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1590 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1591 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1592 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1593 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1594 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1595 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1596 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1597 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1598 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1599 * I am cleaning dirty pages from some other bdi. */
1600 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1601 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1602 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1603 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1604 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1605 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1606 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1607 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1610 * Only the _current_ task can read/write to tsk->flags, but other
1611 * tasks can access tsk->flags in readonly mode for example
1612 * with tsk_used_math (like during threaded core dumping).
1613 * There is however an exception to this rule during ptrace
1614 * or during fork: the ptracer task is allowed to write to the
1615 * child->flags of its traced child (same goes for fork, the parent
1616 * can write to the child->flags), because we're guaranteed the
1617 * child is not running and in turn not changing child->flags
1618 * at the same time the parent does it.
1620 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1621 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1622 #define clear_used_math() clear_stopped_child_used_math(current)
1623 #define set_used_math() set_stopped_child_used_math(current)
1625 #define conditional_stopped_child_used_math(condition, child) \
1626 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1628 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1630 #define copy_to_stopped_child_used_math(child) \
1631 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1633 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1634 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1635 #define used_math() tsk_used_math(current)
1637 static inline bool is_percpu_thread(void)
1640 return (current->flags & PF_NO_SETAFFINITY) &&
1641 (current->nr_cpus_allowed == 1);
1647 /* Per-process atomic flags. */
1648 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1649 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1650 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1651 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1652 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1653 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1654 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1655 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1657 #define TASK_PFA_TEST(name, func) \
1658 static inline bool task_##func(struct task_struct *p) \
1659 { return test_bit(PFA_##name, &p->atomic_flags); }
1661 #define TASK_PFA_SET(name, func) \
1662 static inline void task_set_##func(struct task_struct *p) \
1663 { set_bit(PFA_##name, &p->atomic_flags); }
1665 #define TASK_PFA_CLEAR(name, func) \
1666 static inline void task_clear_##func(struct task_struct *p) \
1667 { clear_bit(PFA_##name, &p->atomic_flags); }
1669 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1670 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1672 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1673 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1674 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1676 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1677 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1678 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1680 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1681 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1682 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1684 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1685 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1686 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1688 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1689 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1691 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1692 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1693 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1695 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1696 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1699 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1701 current->flags &= ~flags;
1702 current->flags |= orig_flags & flags;
1705 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1706 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1708 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1709 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1711 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1714 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1716 if (!cpumask_test_cpu(0, new_mask))
1722 extern int yield_to(struct task_struct *p, bool preempt);
1723 extern void set_user_nice(struct task_struct *p, long nice);
1724 extern int task_prio(const struct task_struct *p);
1727 * task_nice - return the nice value of a given task.
1728 * @p: the task in question.
1730 * Return: The nice value [ -20 ... 0 ... 19 ].
1732 static inline int task_nice(const struct task_struct *p)
1734 return PRIO_TO_NICE((p)->static_prio);
1737 extern int can_nice(const struct task_struct *p, const int nice);
1738 extern int task_curr(const struct task_struct *p);
1739 extern int idle_cpu(int cpu);
1740 extern int available_idle_cpu(int cpu);
1741 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1742 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1743 extern void sched_set_fifo(struct task_struct *p);
1744 extern void sched_set_fifo_low(struct task_struct *p);
1745 extern void sched_set_normal(struct task_struct *p, int nice);
1746 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1747 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1748 extern struct task_struct *idle_task(int cpu);
1751 * is_idle_task - is the specified task an idle task?
1752 * @p: the task in question.
1754 * Return: 1 if @p is an idle task. 0 otherwise.
1756 static __always_inline bool is_idle_task(const struct task_struct *p)
1758 return !!(p->flags & PF_IDLE);
1761 extern struct task_struct *curr_task(int cpu);
1762 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1766 union thread_union {
1767 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1768 struct task_struct task;
1770 #ifndef CONFIG_THREAD_INFO_IN_TASK
1771 struct thread_info thread_info;
1773 unsigned long stack[THREAD_SIZE/sizeof(long)];
1776 #ifndef CONFIG_THREAD_INFO_IN_TASK
1777 extern struct thread_info init_thread_info;
1780 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1782 #ifdef CONFIG_THREAD_INFO_IN_TASK
1783 static inline struct thread_info *task_thread_info(struct task_struct *task)
1785 return &task->thread_info;
1787 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1788 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1792 * find a task by one of its numerical ids
1794 * find_task_by_pid_ns():
1795 * finds a task by its pid in the specified namespace
1796 * find_task_by_vpid():
1797 * finds a task by its virtual pid
1799 * see also find_vpid() etc in include/linux/pid.h
1802 extern struct task_struct *find_task_by_vpid(pid_t nr);
1803 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1806 * find a task by its virtual pid and get the task struct
1808 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1810 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1811 extern int wake_up_process(struct task_struct *tsk);
1812 extern void wake_up_new_task(struct task_struct *tsk);
1815 extern void kick_process(struct task_struct *tsk);
1817 static inline void kick_process(struct task_struct *tsk) { }
1820 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1822 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1824 __set_task_comm(tsk, from, false);
1827 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1828 #define get_task_comm(buf, tsk) ({ \
1829 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1830 __get_task_comm(buf, sizeof(buf), tsk); \
1834 static __always_inline void scheduler_ipi(void)
1837 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1838 * TIF_NEED_RESCHED remotely (for the first time) will also send
1841 preempt_fold_need_resched();
1843 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1845 static inline void scheduler_ipi(void) { }
1846 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1853 * Set thread flags in other task's structures.
1854 * See asm/thread_info.h for TIF_xxxx flags available:
1856 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1858 set_ti_thread_flag(task_thread_info(tsk), flag);
1861 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1863 clear_ti_thread_flag(task_thread_info(tsk), flag);
1866 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1869 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1872 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1874 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1877 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1879 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1882 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1884 return test_ti_thread_flag(task_thread_info(tsk), flag);
1887 static inline void set_tsk_need_resched(struct task_struct *tsk)
1889 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1892 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1894 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1897 static inline int test_tsk_need_resched(struct task_struct *tsk)
1899 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1903 * cond_resched() and cond_resched_lock(): latency reduction via
1904 * explicit rescheduling in places that are safe. The return
1905 * value indicates whether a reschedule was done in fact.
1906 * cond_resched_lock() will drop the spinlock before scheduling,
1908 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1909 extern int __cond_resched(void);
1911 #ifdef CONFIG_PREEMPT_DYNAMIC
1913 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1915 static __always_inline int _cond_resched(void)
1917 return static_call_mod(cond_resched)();
1922 static inline int _cond_resched(void)
1924 return __cond_resched();
1927 #endif /* CONFIG_PREEMPT_DYNAMIC */
1931 static inline int _cond_resched(void) { return 0; }
1933 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
1935 #define cond_resched() ({ \
1936 ___might_sleep(__FILE__, __LINE__, 0); \
1940 extern int __cond_resched_lock(spinlock_t *lock);
1941 extern int __cond_resched_rwlock_read(rwlock_t *lock);
1942 extern int __cond_resched_rwlock_write(rwlock_t *lock);
1944 #define cond_resched_lock(lock) ({ \
1945 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1946 __cond_resched_lock(lock); \
1949 #define cond_resched_rwlock_read(lock) ({ \
1950 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1951 __cond_resched_rwlock_read(lock); \
1954 #define cond_resched_rwlock_write(lock) ({ \
1955 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1956 __cond_resched_rwlock_write(lock); \
1959 static inline void cond_resched_rcu(void)
1961 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1969 * Does a critical section need to be broken due to another
1970 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1971 * but a general need for low latency)
1973 static inline int spin_needbreak(spinlock_t *lock)
1975 #ifdef CONFIG_PREEMPTION
1976 return spin_is_contended(lock);
1983 * Check if a rwlock is contended.
1984 * Returns non-zero if there is another task waiting on the rwlock.
1985 * Returns zero if the lock is not contended or the system / underlying
1986 * rwlock implementation does not support contention detection.
1987 * Technically does not depend on CONFIG_PREEMPTION, but a general need
1990 static inline int rwlock_needbreak(rwlock_t *lock)
1992 #ifdef CONFIG_PREEMPTION
1993 return rwlock_is_contended(lock);
1999 static __always_inline bool need_resched(void)
2001 return unlikely(tif_need_resched());
2005 * Wrappers for p->thread_info->cpu access. No-op on UP.
2009 static inline unsigned int task_cpu(const struct task_struct *p)
2011 #ifdef CONFIG_THREAD_INFO_IN_TASK
2012 return READ_ONCE(p->cpu);
2014 return READ_ONCE(task_thread_info(p)->cpu);
2018 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2022 static inline unsigned int task_cpu(const struct task_struct *p)
2027 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2031 #endif /* CONFIG_SMP */
2033 extern bool sched_task_on_rq(struct task_struct *p);
2036 * In order to reduce various lock holder preemption latencies provide an
2037 * interface to see if a vCPU is currently running or not.
2039 * This allows us to terminate optimistic spin loops and block, analogous to
2040 * the native optimistic spin heuristic of testing if the lock owner task is
2043 #ifndef vcpu_is_preempted
2044 static inline bool vcpu_is_preempted(int cpu)
2050 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2051 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2053 #ifndef TASK_SIZE_OF
2054 #define TASK_SIZE_OF(tsk) TASK_SIZE
2058 /* Returns effective CPU energy utilization, as seen by the scheduler */
2059 unsigned long sched_cpu_util(int cpu, unsigned long max);
2060 #endif /* CONFIG_SMP */
2065 * Map the event mask on the user-space ABI enum rseq_cs_flags
2066 * for direct mask checks.
2068 enum rseq_event_mask_bits {
2069 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2070 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2071 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2074 enum rseq_event_mask {
2075 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2076 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2077 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2080 static inline void rseq_set_notify_resume(struct task_struct *t)
2083 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2086 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2088 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2089 struct pt_regs *regs)
2092 __rseq_handle_notify_resume(ksig, regs);
2095 static inline void rseq_signal_deliver(struct ksignal *ksig,
2096 struct pt_regs *regs)
2099 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2101 rseq_handle_notify_resume(ksig, regs);
2104 /* rseq_preempt() requires preemption to be disabled. */
2105 static inline void rseq_preempt(struct task_struct *t)
2107 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2108 rseq_set_notify_resume(t);
2111 /* rseq_migrate() requires preemption to be disabled. */
2112 static inline void rseq_migrate(struct task_struct *t)
2114 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2115 rseq_set_notify_resume(t);
2119 * If parent process has a registered restartable sequences area, the
2120 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2122 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2124 if (clone_flags & CLONE_VM) {
2127 t->rseq_event_mask = 0;
2129 t->rseq = current->rseq;
2130 t->rseq_sig = current->rseq_sig;
2131 t->rseq_event_mask = current->rseq_event_mask;
2135 static inline void rseq_execve(struct task_struct *t)
2139 t->rseq_event_mask = 0;
2144 static inline void rseq_set_notify_resume(struct task_struct *t)
2147 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2148 struct pt_regs *regs)
2151 static inline void rseq_signal_deliver(struct ksignal *ksig,
2152 struct pt_regs *regs)
2155 static inline void rseq_preempt(struct task_struct *t)
2158 static inline void rseq_migrate(struct task_struct *t)
2161 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2164 static inline void rseq_execve(struct task_struct *t)
2170 #ifdef CONFIG_DEBUG_RSEQ
2172 void rseq_syscall(struct pt_regs *regs);
2176 static inline void rseq_syscall(struct pt_regs *regs)
2182 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2183 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2184 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2186 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2187 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2188 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2190 int sched_trace_rq_cpu(struct rq *rq);
2191 int sched_trace_rq_cpu_capacity(struct rq *rq);
2192 int sched_trace_rq_nr_running(struct rq *rq);
2194 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2196 #ifdef CONFIG_SCHED_CORE
2197 extern void sched_core_free(struct task_struct *tsk);
2198 extern void sched_core_fork(struct task_struct *p);
2199 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2200 unsigned long uaddr);
2202 static inline void sched_core_free(struct task_struct *tsk) { }
2203 static inline void sched_core_fork(struct task_struct *p) { }