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 #define TASK_STATE_MAX 0x1000
100 /* Convenience macros for the sake of set_current_state: */
101 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
102 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
103 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
105 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
107 /* Convenience macros for the sake of wake_up(): */
108 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
110 /* get_task_state(): */
111 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
112 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
113 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
116 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
118 #define task_is_traced(task) ((READ_ONCE(task->__state) & __TASK_TRACED) != 0)
120 #define task_is_stopped(task) ((READ_ONCE(task->__state) & __TASK_STOPPED) != 0)
122 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->__state) & (__TASK_STOPPED | __TASK_TRACED)) != 0)
124 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
127 * Special states are those that do not use the normal wait-loop pattern. See
128 * the comment with set_special_state().
130 #define is_special_task_state(state) \
131 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
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 WRITE_ONCE(current->__state, (state_value)); \
140 #define set_current_state(state_value) \
142 WARN_ON_ONCE(is_special_task_state(state_value));\
143 current->task_state_change = _THIS_IP_; \
144 smp_store_mb(current->__state, (state_value)); \
147 #define set_special_state(state_value) \
149 unsigned long flags; /* may shadow */ \
150 WARN_ON_ONCE(!is_special_task_state(state_value)); \
151 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
152 current->task_state_change = _THIS_IP_; \
153 WRITE_ONCE(current->__state, (state_value)); \
154 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
158 * set_current_state() includes a barrier so that the write of current->state
159 * is correctly serialised wrt the caller's subsequent test of whether to
163 * set_current_state(TASK_UNINTERRUPTIBLE);
169 * __set_current_state(TASK_RUNNING);
171 * If the caller does not need such serialisation (because, for instance, the
172 * CONDITION test and condition change and wakeup are under the same lock) then
173 * use __set_current_state().
175 * The above is typically ordered against the wakeup, which does:
178 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
180 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
181 * accessing p->state.
183 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
184 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
185 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
187 * However, with slightly different timing the wakeup TASK_RUNNING store can
188 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
189 * a problem either because that will result in one extra go around the loop
190 * and our @cond test will save the day.
192 * Also see the comments of try_to_wake_up().
194 #define __set_current_state(state_value) \
195 WRITE_ONCE(current->__state, (state_value))
197 #define set_current_state(state_value) \
198 smp_store_mb(current->__state, (state_value))
201 * set_special_state() should be used for those states when the blocking task
202 * can not use the regular condition based wait-loop. In that case we must
203 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
204 * will not collide with our state change.
206 #define set_special_state(state_value) \
208 unsigned long flags; /* may shadow */ \
209 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
210 WRITE_ONCE(current->__state, (state_value)); \
211 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
216 #define get_current_state() READ_ONCE(current->__state)
218 /* Task command name length: */
219 #define TASK_COMM_LEN 16
221 extern void scheduler_tick(void);
223 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
225 extern long schedule_timeout(long timeout);
226 extern long schedule_timeout_interruptible(long timeout);
227 extern long schedule_timeout_killable(long timeout);
228 extern long schedule_timeout_uninterruptible(long timeout);
229 extern long schedule_timeout_idle(long timeout);
230 asmlinkage void schedule(void);
231 extern void schedule_preempt_disabled(void);
232 asmlinkage void preempt_schedule_irq(void);
234 extern int __must_check io_schedule_prepare(void);
235 extern void io_schedule_finish(int token);
236 extern long io_schedule_timeout(long timeout);
237 extern void io_schedule(void);
240 * struct prev_cputime - snapshot of system and user cputime
241 * @utime: time spent in user mode
242 * @stime: time spent in system mode
243 * @lock: protects the above two fields
245 * Stores previous user/system time values such that we can guarantee
248 struct prev_cputime {
249 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
257 /* Task is sleeping or running in a CPU with VTIME inactive: */
261 /* Task runs in kernelspace in a CPU with VTIME active: */
263 /* Task runs in userspace in a CPU with VTIME active: */
265 /* Task runs as guests in a CPU with VTIME active: */
271 unsigned long long starttime;
272 enum vtime_state state;
280 * Utilization clamp constraints.
281 * @UCLAMP_MIN: Minimum utilization
282 * @UCLAMP_MAX: Maximum utilization
283 * @UCLAMP_CNT: Utilization clamp constraints count
292 extern struct root_domain def_root_domain;
293 extern struct mutex sched_domains_mutex;
297 #ifdef CONFIG_SCHED_INFO
298 /* Cumulative counters: */
300 /* # of times we have run on this CPU: */
301 unsigned long pcount;
303 /* Time spent waiting on a runqueue: */
304 unsigned long long run_delay;
308 /* When did we last run on a CPU? */
309 unsigned long long last_arrival;
311 /* When were we last queued to run? */
312 unsigned long long last_queued;
314 #endif /* CONFIG_SCHED_INFO */
318 * Integer metrics need fixed point arithmetic, e.g., sched/fair
319 * has a few: load, load_avg, util_avg, freq, and capacity.
321 * We define a basic fixed point arithmetic range, and then formalize
322 * all these metrics based on that basic range.
324 # define SCHED_FIXEDPOINT_SHIFT 10
325 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
327 /* Increase resolution of cpu_capacity calculations */
328 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
329 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
332 unsigned long weight;
337 * struct util_est - Estimation utilization of FAIR tasks
338 * @enqueued: instantaneous estimated utilization of a task/cpu
339 * @ewma: the Exponential Weighted Moving Average (EWMA)
340 * utilization of a task
342 * Support data structure to track an Exponential Weighted Moving Average
343 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
344 * average each time a task completes an activation. Sample's weight is chosen
345 * so that the EWMA will be relatively insensitive to transient changes to the
348 * The enqueued attribute has a slightly different meaning for tasks and cpus:
349 * - task: the task's util_avg at last task dequeue time
350 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
351 * Thus, the util_est.enqueued of a task represents the contribution on the
352 * estimated utilization of the CPU where that task is currently enqueued.
354 * Only for tasks we track a moving average of the past instantaneous
355 * estimated utilization. This allows to absorb sporadic drops in utilization
356 * of an otherwise almost periodic task.
358 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
359 * updates. When a task is dequeued, its util_est should not be updated if its
360 * util_avg has not been updated in the meantime.
361 * This information is mapped into the MSB bit of util_est.enqueued at dequeue
362 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
363 * for a task) it is safe to use MSB.
366 unsigned int enqueued;
368 #define UTIL_EST_WEIGHT_SHIFT 2
369 #define UTIL_AVG_UNCHANGED 0x80000000
370 } __attribute__((__aligned__(sizeof(u64))));
373 * The load/runnable/util_avg accumulates an infinite geometric series
374 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
376 * [load_avg definition]
378 * load_avg = runnable% * scale_load_down(load)
380 * [runnable_avg definition]
382 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
384 * [util_avg definition]
386 * util_avg = running% * SCHED_CAPACITY_SCALE
388 * where runnable% is the time ratio that a sched_entity is runnable and
389 * running% the time ratio that a sched_entity is running.
391 * For cfs_rq, they are the aggregated values of all runnable and blocked
394 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
395 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
396 * for computing those signals (see update_rq_clock_pelt())
398 * N.B., the above ratios (runnable% and running%) themselves are in the
399 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
400 * to as large a range as necessary. This is for example reflected by
401 * util_avg's SCHED_CAPACITY_SCALE.
405 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
406 * with the highest load (=88761), always runnable on a single cfs_rq,
407 * and should not overflow as the number already hits PID_MAX_LIMIT.
409 * For all other cases (including 32-bit kernels), struct load_weight's
410 * weight will overflow first before we do, because:
412 * Max(load_avg) <= Max(load.weight)
414 * Then it is the load_weight's responsibility to consider overflow
418 u64 last_update_time;
423 unsigned long load_avg;
424 unsigned long runnable_avg;
425 unsigned long util_avg;
426 struct util_est util_est;
427 } ____cacheline_aligned;
429 struct sched_statistics {
430 #ifdef CONFIG_SCHEDSTATS
440 s64 sum_sleep_runtime;
447 u64 nr_migrations_cold;
448 u64 nr_failed_migrations_affine;
449 u64 nr_failed_migrations_running;
450 u64 nr_failed_migrations_hot;
451 u64 nr_forced_migrations;
455 u64 nr_wakeups_migrate;
456 u64 nr_wakeups_local;
457 u64 nr_wakeups_remote;
458 u64 nr_wakeups_affine;
459 u64 nr_wakeups_affine_attempts;
460 u64 nr_wakeups_passive;
465 struct sched_entity {
466 /* For load-balancing: */
467 struct load_weight load;
468 struct rb_node run_node;
469 struct list_head group_node;
473 u64 sum_exec_runtime;
475 u64 prev_sum_exec_runtime;
479 struct sched_statistics statistics;
481 #ifdef CONFIG_FAIR_GROUP_SCHED
483 struct sched_entity *parent;
484 /* rq on which this entity is (to be) queued: */
485 struct cfs_rq *cfs_rq;
486 /* rq "owned" by this entity/group: */
488 /* cached value of my_q->h_nr_running */
489 unsigned long runnable_weight;
494 * Per entity load average tracking.
496 * Put into separate cache line so it does not
497 * collide with read-mostly values above.
499 struct sched_avg avg;
503 struct sched_rt_entity {
504 struct list_head run_list;
505 unsigned long timeout;
506 unsigned long watchdog_stamp;
507 unsigned int time_slice;
508 unsigned short on_rq;
509 unsigned short on_list;
511 struct sched_rt_entity *back;
512 #ifdef CONFIG_RT_GROUP_SCHED
513 struct sched_rt_entity *parent;
514 /* rq on which this entity is (to be) queued: */
516 /* rq "owned" by this entity/group: */
519 } __randomize_layout;
521 struct sched_dl_entity {
522 struct rb_node rb_node;
525 * Original scheduling parameters. Copied here from sched_attr
526 * during sched_setattr(), they will remain the same until
527 * the next sched_setattr().
529 u64 dl_runtime; /* Maximum runtime for each instance */
530 u64 dl_deadline; /* Relative deadline of each instance */
531 u64 dl_period; /* Separation of two instances (period) */
532 u64 dl_bw; /* dl_runtime / dl_period */
533 u64 dl_density; /* dl_runtime / dl_deadline */
536 * Actual scheduling parameters. Initialized with the values above,
537 * they are continuously updated during task execution. Note that
538 * the remaining runtime could be < 0 in case we are in overrun.
540 s64 runtime; /* Remaining runtime for this instance */
541 u64 deadline; /* Absolute deadline for this instance */
542 unsigned int flags; /* Specifying the scheduler behaviour */
547 * @dl_throttled tells if we exhausted the runtime. If so, the
548 * task has to wait for a replenishment to be performed at the
549 * next firing of dl_timer.
551 * @dl_boosted tells if we are boosted due to DI. If so we are
552 * outside bandwidth enforcement mechanism (but only until we
553 * exit the critical section);
555 * @dl_yielded tells if task gave up the CPU before consuming
556 * all its available runtime during the last job.
558 * @dl_non_contending tells if the task is inactive while still
559 * contributing to the active utilization. In other words, it
560 * indicates if the inactive timer has been armed and its handler
561 * has not been executed yet. This flag is useful to avoid race
562 * conditions between the inactive timer handler and the wakeup
565 * @dl_overrun tells if the task asked to be informed about runtime
568 unsigned int dl_throttled : 1;
569 unsigned int dl_yielded : 1;
570 unsigned int dl_non_contending : 1;
571 unsigned int dl_overrun : 1;
574 * Bandwidth enforcement timer. Each -deadline task has its
575 * own bandwidth to be enforced, thus we need one timer per task.
577 struct hrtimer dl_timer;
580 * Inactive timer, responsible for decreasing the active utilization
581 * at the "0-lag time". When a -deadline task blocks, it contributes
582 * to GRUB's active utilization until the "0-lag time", hence a
583 * timer is needed to decrease the active utilization at the correct
586 struct hrtimer inactive_timer;
588 #ifdef CONFIG_RT_MUTEXES
590 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
591 * pi_se points to the donor, otherwise points to the dl_se it belongs
592 * to (the original one/itself).
594 struct sched_dl_entity *pi_se;
598 #ifdef CONFIG_UCLAMP_TASK
599 /* Number of utilization clamp buckets (shorter alias) */
600 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
603 * Utilization clamp for a scheduling entity
604 * @value: clamp value "assigned" to a se
605 * @bucket_id: bucket index corresponding to the "assigned" value
606 * @active: the se is currently refcounted in a rq's bucket
607 * @user_defined: the requested clamp value comes from user-space
609 * The bucket_id is the index of the clamp bucket matching the clamp value
610 * which is pre-computed and stored to avoid expensive integer divisions from
613 * The active bit is set whenever a task has got an "effective" value assigned,
614 * which can be different from the clamp value "requested" from user-space.
615 * This allows to know a task is refcounted in the rq's bucket corresponding
616 * to the "effective" bucket_id.
618 * The user_defined bit is set whenever a task has got a task-specific clamp
619 * value requested from userspace, i.e. the system defaults apply to this task
620 * just as a restriction. This allows to relax default clamps when a less
621 * restrictive task-specific value has been requested, thus allowing to
622 * implement a "nice" semantic. For example, a task running with a 20%
623 * default boost can still drop its own boosting to 0%.
626 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
627 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
628 unsigned int active : 1;
629 unsigned int user_defined : 1;
631 #endif /* CONFIG_UCLAMP_TASK */
637 u8 exp_hint; /* Hint for performance. */
638 u8 need_mb; /* Readers need smp_mb(). */
640 u32 s; /* Set of bits. */
643 enum perf_event_task_context {
644 perf_invalid_context = -1,
647 perf_nr_task_contexts,
651 struct wake_q_node *next;
655 #ifdef CONFIG_KMAP_LOCAL
657 pte_t pteval[KM_MAX_IDX];
662 #ifdef CONFIG_THREAD_INFO_IN_TASK
664 * For reasons of header soup (see current_thread_info()), this
665 * must be the first element of task_struct.
667 struct thread_info thread_info;
669 unsigned int __state;
672 * This begins the randomizable portion of task_struct. Only
673 * scheduling-critical items should be added above here.
675 randomized_struct_fields_start
679 /* Per task flags (PF_*), defined further below: */
685 struct __call_single_node wake_entry;
686 #ifdef CONFIG_THREAD_INFO_IN_TASK
690 unsigned int wakee_flips;
691 unsigned long wakee_flip_decay_ts;
692 struct task_struct *last_wakee;
695 * recent_used_cpu is initially set as the last CPU used by a task
696 * that wakes affine another task. Waker/wakee relationships can
697 * push tasks around a CPU where each wakeup moves to the next one.
698 * Tracking a recently used CPU allows a quick search for a recently
699 * used CPU that may be idle.
709 unsigned int rt_priority;
711 const struct sched_class *sched_class;
712 struct sched_entity se;
713 struct sched_rt_entity rt;
714 struct sched_dl_entity dl;
716 #ifdef CONFIG_SCHED_CORE
717 struct rb_node core_node;
718 unsigned long core_cookie;
719 unsigned int core_occupation;
722 #ifdef CONFIG_CGROUP_SCHED
723 struct task_group *sched_task_group;
726 #ifdef CONFIG_UCLAMP_TASK
728 * Clamp values requested for a scheduling entity.
729 * Must be updated with task_rq_lock() held.
731 struct uclamp_se uclamp_req[UCLAMP_CNT];
733 * Effective clamp values used for a scheduling entity.
734 * Must be updated with task_rq_lock() held.
736 struct uclamp_se uclamp[UCLAMP_CNT];
739 #ifdef CONFIG_PREEMPT_NOTIFIERS
740 /* List of struct preempt_notifier: */
741 struct hlist_head preempt_notifiers;
744 #ifdef CONFIG_BLK_DEV_IO_TRACE
745 unsigned int btrace_seq;
750 const cpumask_t *cpus_ptr;
751 cpumask_t *user_cpus_ptr;
753 void *migration_pending;
755 unsigned short migration_disabled;
757 unsigned short migration_flags;
759 #ifdef CONFIG_PREEMPT_RCU
760 int rcu_read_lock_nesting;
761 union rcu_special rcu_read_unlock_special;
762 struct list_head rcu_node_entry;
763 struct rcu_node *rcu_blocked_node;
764 #endif /* #ifdef CONFIG_PREEMPT_RCU */
766 #ifdef CONFIG_TASKS_RCU
767 unsigned long rcu_tasks_nvcsw;
768 u8 rcu_tasks_holdout;
770 int rcu_tasks_idle_cpu;
771 struct list_head rcu_tasks_holdout_list;
772 #endif /* #ifdef CONFIG_TASKS_RCU */
774 #ifdef CONFIG_TASKS_TRACE_RCU
775 int trc_reader_nesting;
777 union rcu_special trc_reader_special;
778 bool trc_reader_checked;
779 struct list_head trc_holdout_list;
780 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
782 struct sched_info sched_info;
784 struct list_head tasks;
786 struct plist_node pushable_tasks;
787 struct rb_node pushable_dl_tasks;
790 struct mm_struct *mm;
791 struct mm_struct *active_mm;
793 /* Per-thread vma caching: */
794 struct vmacache vmacache;
796 #ifdef SPLIT_RSS_COUNTING
797 struct task_rss_stat rss_stat;
802 /* The signal sent when the parent dies: */
804 /* JOBCTL_*, siglock protected: */
805 unsigned long jobctl;
807 /* Used for emulating ABI behavior of previous Linux versions: */
808 unsigned int personality;
810 /* Scheduler bits, serialized by scheduler locks: */
811 unsigned sched_reset_on_fork:1;
812 unsigned sched_contributes_to_load:1;
813 unsigned sched_migrated:1;
815 unsigned sched_psi_wake_requeue:1;
818 /* Force alignment to the next boundary: */
821 /* Unserialized, strictly 'current' */
824 * This field must not be in the scheduler word above due to wakelist
825 * queueing no longer being serialized by p->on_cpu. However:
828 * schedule() if (p->on_rq && ..) // false
829 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
830 * deactivate_task() ttwu_queue_wakelist())
831 * p->on_rq = 0; p->sched_remote_wakeup = Y;
833 * guarantees all stores of 'current' are visible before
834 * ->sched_remote_wakeup gets used, so it can be in this word.
836 unsigned sched_remote_wakeup:1;
838 /* Bit to tell LSMs we're in execve(): */
839 unsigned in_execve:1;
840 unsigned in_iowait:1;
841 #ifndef TIF_RESTORE_SIGMASK
842 unsigned restore_sigmask:1;
845 unsigned in_user_fault:1;
847 #ifdef CONFIG_COMPAT_BRK
848 unsigned brk_randomized:1;
850 #ifdef CONFIG_CGROUPS
851 /* disallow userland-initiated cgroup migration */
852 unsigned no_cgroup_migration:1;
853 /* task is frozen/stopped (used by the cgroup freezer) */
856 #ifdef CONFIG_BLK_CGROUP
857 unsigned use_memdelay:1;
860 /* Stalled due to lack of memory */
861 unsigned in_memstall:1;
863 #ifdef CONFIG_PAGE_OWNER
864 /* Used by page_owner=on to detect recursion in page tracking. */
865 unsigned in_page_owner:1;
868 unsigned long atomic_flags; /* Flags requiring atomic access. */
870 struct restart_block restart_block;
875 #ifdef CONFIG_STACKPROTECTOR
876 /* Canary value for the -fstack-protector GCC feature: */
877 unsigned long stack_canary;
880 * Pointers to the (original) parent process, youngest child, younger sibling,
881 * older sibling, respectively. (p->father can be replaced with
882 * p->real_parent->pid)
885 /* Real parent process: */
886 struct task_struct __rcu *real_parent;
888 /* Recipient of SIGCHLD, wait4() reports: */
889 struct task_struct __rcu *parent;
892 * Children/sibling form the list of natural children:
894 struct list_head children;
895 struct list_head sibling;
896 struct task_struct *group_leader;
899 * 'ptraced' is the list of tasks this task is using ptrace() on.
901 * This includes both natural children and PTRACE_ATTACH targets.
902 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
904 struct list_head ptraced;
905 struct list_head ptrace_entry;
907 /* PID/PID hash table linkage. */
908 struct pid *thread_pid;
909 struct hlist_node pid_links[PIDTYPE_MAX];
910 struct list_head thread_group;
911 struct list_head thread_node;
913 struct completion *vfork_done;
915 /* CLONE_CHILD_SETTID: */
916 int __user *set_child_tid;
918 /* CLONE_CHILD_CLEARTID: */
919 int __user *clear_child_tid;
926 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
931 struct prev_cputime prev_cputime;
932 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
936 #ifdef CONFIG_NO_HZ_FULL
937 atomic_t tick_dep_mask;
939 /* Context switch counts: */
941 unsigned long nivcsw;
943 /* Monotonic time in nsecs: */
946 /* Boot based time in nsecs: */
949 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
950 unsigned long min_flt;
951 unsigned long maj_flt;
953 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
954 struct posix_cputimers posix_cputimers;
956 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
957 struct posix_cputimers_work posix_cputimers_work;
960 /* Process credentials: */
962 /* Tracer's credentials at attach: */
963 const struct cred __rcu *ptracer_cred;
965 /* Objective and real subjective task credentials (COW): */
966 const struct cred __rcu *real_cred;
968 /* Effective (overridable) subjective task credentials (COW): */
969 const struct cred __rcu *cred;
972 /* Cached requested key. */
973 struct key *cached_requested_key;
977 * executable name, excluding path.
979 * - normally initialized setup_new_exec()
980 * - access it with [gs]et_task_comm()
981 * - lock it with task_lock()
983 char comm[TASK_COMM_LEN];
985 struct nameidata *nameidata;
987 #ifdef CONFIG_SYSVIPC
988 struct sysv_sem sysvsem;
989 struct sysv_shm sysvshm;
991 #ifdef CONFIG_DETECT_HUNG_TASK
992 unsigned long last_switch_count;
993 unsigned long last_switch_time;
995 /* Filesystem information: */
996 struct fs_struct *fs;
998 /* Open file information: */
999 struct files_struct *files;
1001 #ifdef CONFIG_IO_URING
1002 struct io_uring_task *io_uring;
1006 struct nsproxy *nsproxy;
1008 /* Signal handlers: */
1009 struct signal_struct *signal;
1010 struct sighand_struct __rcu *sighand;
1012 sigset_t real_blocked;
1013 /* Restored if set_restore_sigmask() was used: */
1014 sigset_t saved_sigmask;
1015 struct sigpending pending;
1016 unsigned long sas_ss_sp;
1018 unsigned int sas_ss_flags;
1020 struct callback_head *task_works;
1023 #ifdef CONFIG_AUDITSYSCALL
1024 struct audit_context *audit_context;
1027 unsigned int sessionid;
1029 struct seccomp seccomp;
1030 struct syscall_user_dispatch syscall_dispatch;
1032 /* Thread group tracking: */
1036 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1037 spinlock_t alloc_lock;
1039 /* Protection of the PI data structures: */
1040 raw_spinlock_t pi_lock;
1042 struct wake_q_node wake_q;
1044 #ifdef CONFIG_RT_MUTEXES
1045 /* PI waiters blocked on a rt_mutex held by this task: */
1046 struct rb_root_cached pi_waiters;
1047 /* Updated under owner's pi_lock and rq lock */
1048 struct task_struct *pi_top_task;
1049 /* Deadlock detection and priority inheritance handling: */
1050 struct rt_mutex_waiter *pi_blocked_on;
1053 #ifdef CONFIG_DEBUG_MUTEXES
1054 /* Mutex deadlock detection: */
1055 struct mutex_waiter *blocked_on;
1058 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1059 int non_block_count;
1062 #ifdef CONFIG_TRACE_IRQFLAGS
1063 struct irqtrace_events irqtrace;
1064 unsigned int hardirq_threaded;
1065 u64 hardirq_chain_key;
1066 int softirqs_enabled;
1067 int softirq_context;
1070 #ifdef CONFIG_PREEMPT_RT
1071 int softirq_disable_cnt;
1074 #ifdef CONFIG_LOCKDEP
1075 # define MAX_LOCK_DEPTH 48UL
1078 unsigned int lockdep_recursion;
1079 struct held_lock held_locks[MAX_LOCK_DEPTH];
1082 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1083 unsigned int in_ubsan;
1086 /* Journalling filesystem info: */
1089 /* Stacked block device info: */
1090 struct bio_list *bio_list;
1093 /* Stack plugging: */
1094 struct blk_plug *plug;
1098 struct reclaim_state *reclaim_state;
1100 struct backing_dev_info *backing_dev_info;
1102 struct io_context *io_context;
1104 #ifdef CONFIG_COMPACTION
1105 struct capture_control *capture_control;
1108 unsigned long ptrace_message;
1109 kernel_siginfo_t *last_siginfo;
1111 struct task_io_accounting ioac;
1113 /* Pressure stall state */
1114 unsigned int psi_flags;
1116 #ifdef CONFIG_TASK_XACCT
1117 /* Accumulated RSS usage: */
1119 /* Accumulated virtual memory usage: */
1121 /* stime + utime since last update: */
1124 #ifdef CONFIG_CPUSETS
1125 /* Protected by ->alloc_lock: */
1126 nodemask_t mems_allowed;
1127 /* Sequence number to catch updates: */
1128 seqcount_spinlock_t mems_allowed_seq;
1129 int cpuset_mem_spread_rotor;
1130 int cpuset_slab_spread_rotor;
1132 #ifdef CONFIG_CGROUPS
1133 /* Control Group info protected by css_set_lock: */
1134 struct css_set __rcu *cgroups;
1135 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1136 struct list_head cg_list;
1138 #ifdef CONFIG_X86_CPU_RESCTRL
1143 struct robust_list_head __user *robust_list;
1144 #ifdef CONFIG_COMPAT
1145 struct compat_robust_list_head __user *compat_robust_list;
1147 struct list_head pi_state_list;
1148 struct futex_pi_state *pi_state_cache;
1149 struct mutex futex_exit_mutex;
1150 unsigned int futex_state;
1152 #ifdef CONFIG_PERF_EVENTS
1153 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1154 struct mutex perf_event_mutex;
1155 struct list_head perf_event_list;
1157 #ifdef CONFIG_DEBUG_PREEMPT
1158 unsigned long preempt_disable_ip;
1161 /* Protected by alloc_lock: */
1162 struct mempolicy *mempolicy;
1164 short pref_node_fork;
1166 #ifdef CONFIG_NUMA_BALANCING
1168 unsigned int numa_scan_period;
1169 unsigned int numa_scan_period_max;
1170 int numa_preferred_nid;
1171 unsigned long numa_migrate_retry;
1172 /* Migration stamp: */
1174 u64 last_task_numa_placement;
1175 u64 last_sum_exec_runtime;
1176 struct callback_head numa_work;
1179 * This pointer is only modified for current in syscall and
1180 * pagefault context (and for tasks being destroyed), so it can be read
1181 * from any of the following contexts:
1182 * - RCU read-side critical section
1183 * - current->numa_group from everywhere
1184 * - task's runqueue locked, task not running
1186 struct numa_group __rcu *numa_group;
1189 * numa_faults is an array split into four regions:
1190 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1191 * in this precise order.
1193 * faults_memory: Exponential decaying average of faults on a per-node
1194 * basis. Scheduling placement decisions are made based on these
1195 * counts. The values remain static for the duration of a PTE scan.
1196 * faults_cpu: Track the nodes the process was running on when a NUMA
1197 * hinting fault was incurred.
1198 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1199 * during the current scan window. When the scan completes, the counts
1200 * in faults_memory and faults_cpu decay and these values are copied.
1202 unsigned long *numa_faults;
1203 unsigned long total_numa_faults;
1206 * numa_faults_locality tracks if faults recorded during the last
1207 * scan window were remote/local or failed to migrate. The task scan
1208 * period is adapted based on the locality of the faults with different
1209 * weights depending on whether they were shared or private faults
1211 unsigned long numa_faults_locality[3];
1213 unsigned long numa_pages_migrated;
1214 #endif /* CONFIG_NUMA_BALANCING */
1217 struct rseq __user *rseq;
1220 * RmW on rseq_event_mask must be performed atomically
1221 * with respect to preemption.
1223 unsigned long rseq_event_mask;
1226 struct tlbflush_unmap_batch tlb_ubc;
1229 refcount_t rcu_users;
1230 struct rcu_head rcu;
1233 /* Cache last used pipe for splice(): */
1234 struct pipe_inode_info *splice_pipe;
1236 struct page_frag task_frag;
1238 #ifdef CONFIG_TASK_DELAY_ACCT
1239 struct task_delay_info *delays;
1242 #ifdef CONFIG_FAULT_INJECTION
1244 unsigned int fail_nth;
1247 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1248 * balance_dirty_pages() for a dirty throttling pause:
1251 int nr_dirtied_pause;
1252 /* Start of a write-and-pause period: */
1253 unsigned long dirty_paused_when;
1255 #ifdef CONFIG_LATENCYTOP
1256 int latency_record_count;
1257 struct latency_record latency_record[LT_SAVECOUNT];
1260 * Time slack values; these are used to round up poll() and
1261 * select() etc timeout values. These are in nanoseconds.
1264 u64 default_timer_slack_ns;
1266 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1267 unsigned int kasan_depth;
1271 struct kcsan_ctx kcsan_ctx;
1272 #ifdef CONFIG_TRACE_IRQFLAGS
1273 struct irqtrace_events kcsan_save_irqtrace;
1277 #if IS_ENABLED(CONFIG_KUNIT)
1278 struct kunit *kunit_test;
1281 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1282 /* Index of current stored address in ret_stack: */
1286 /* Stack of return addresses for return function tracing: */
1287 struct ftrace_ret_stack *ret_stack;
1289 /* Timestamp for last schedule: */
1290 unsigned long long ftrace_timestamp;
1293 * Number of functions that haven't been traced
1294 * because of depth overrun:
1296 atomic_t trace_overrun;
1298 /* Pause tracing: */
1299 atomic_t tracing_graph_pause;
1302 #ifdef CONFIG_TRACING
1303 /* State flags for use by tracers: */
1304 unsigned long trace;
1306 /* Bitmask and counter of trace recursion: */
1307 unsigned long trace_recursion;
1308 #endif /* CONFIG_TRACING */
1311 /* See kernel/kcov.c for more details. */
1313 /* Coverage collection mode enabled for this task (0 if disabled): */
1314 unsigned int kcov_mode;
1316 /* Size of the kcov_area: */
1317 unsigned int kcov_size;
1319 /* Buffer for coverage collection: */
1322 /* KCOV descriptor wired with this task or NULL: */
1325 /* KCOV common handle for remote coverage collection: */
1328 /* KCOV sequence number: */
1331 /* Collect coverage from softirq context: */
1332 unsigned int kcov_softirq;
1336 struct mem_cgroup *memcg_in_oom;
1337 gfp_t memcg_oom_gfp_mask;
1338 int memcg_oom_order;
1340 /* Number of pages to reclaim on returning to userland: */
1341 unsigned int memcg_nr_pages_over_high;
1343 /* Used by memcontrol for targeted memcg charge: */
1344 struct mem_cgroup *active_memcg;
1347 #ifdef CONFIG_BLK_CGROUP
1348 struct request_queue *throttle_queue;
1351 #ifdef CONFIG_UPROBES
1352 struct uprobe_task *utask;
1354 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1355 unsigned int sequential_io;
1356 unsigned int sequential_io_avg;
1358 struct kmap_ctrl kmap_ctrl;
1359 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1360 unsigned long task_state_change;
1362 int pagefault_disabled;
1364 struct task_struct *oom_reaper_list;
1366 #ifdef CONFIG_VMAP_STACK
1367 struct vm_struct *stack_vm_area;
1369 #ifdef CONFIG_THREAD_INFO_IN_TASK
1370 /* A live task holds one reference: */
1371 refcount_t stack_refcount;
1373 #ifdef CONFIG_LIVEPATCH
1376 #ifdef CONFIG_SECURITY
1377 /* Used by LSM modules for access restriction: */
1380 #ifdef CONFIG_BPF_SYSCALL
1381 /* Used by BPF task local storage */
1382 struct bpf_local_storage __rcu *bpf_storage;
1385 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1386 unsigned long lowest_stack;
1387 unsigned long prev_lowest_stack;
1390 #ifdef CONFIG_X86_MCE
1391 void __user *mce_vaddr;
1396 __mce_reserved : 62;
1397 struct callback_head mce_kill_me;
1400 #ifdef CONFIG_KRETPROBES
1401 struct llist_head kretprobe_instances;
1405 * New fields for task_struct should be added above here, so that
1406 * they are included in the randomized portion of task_struct.
1408 randomized_struct_fields_end
1410 /* CPU-specific state of this task: */
1411 struct thread_struct thread;
1414 * WARNING: on x86, 'thread_struct' contains a variable-sized
1415 * structure. It *MUST* be at the end of 'task_struct'.
1417 * Do not put anything below here!
1421 static inline struct pid *task_pid(struct task_struct *task)
1423 return task->thread_pid;
1427 * the helpers to get the task's different pids as they are seen
1428 * from various namespaces
1430 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1431 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1433 * task_xid_nr_ns() : id seen from the ns specified;
1435 * see also pid_nr() etc in include/linux/pid.h
1437 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1439 static inline pid_t task_pid_nr(struct task_struct *tsk)
1444 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1446 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1449 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1451 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1455 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1461 * pid_alive - check that a task structure is not stale
1462 * @p: Task structure to be checked.
1464 * Test if a process is not yet dead (at most zombie state)
1465 * If pid_alive fails, then pointers within the task structure
1466 * can be stale and must not be dereferenced.
1468 * Return: 1 if the process is alive. 0 otherwise.
1470 static inline int pid_alive(const struct task_struct *p)
1472 return p->thread_pid != NULL;
1475 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1477 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1480 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1482 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1486 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1488 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1491 static inline pid_t task_session_vnr(struct task_struct *tsk)
1493 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1496 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1498 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1501 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1503 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1506 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1512 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1518 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1520 return task_ppid_nr_ns(tsk, &init_pid_ns);
1523 /* Obsolete, do not use: */
1524 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1526 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1529 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1530 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1532 static inline unsigned int task_state_index(struct task_struct *tsk)
1534 unsigned int tsk_state = READ_ONCE(tsk->__state);
1535 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1537 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1539 if (tsk_state == TASK_IDLE)
1540 state = TASK_REPORT_IDLE;
1545 static inline char task_index_to_char(unsigned int state)
1547 static const char state_char[] = "RSDTtXZPI";
1549 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1551 return state_char[state];
1554 static inline char task_state_to_char(struct task_struct *tsk)
1556 return task_index_to_char(task_state_index(tsk));
1560 * is_global_init - check if a task structure is init. Since init
1561 * is free to have sub-threads we need to check tgid.
1562 * @tsk: Task structure to be checked.
1564 * Check if a task structure is the first user space task the kernel created.
1566 * Return: 1 if the task structure is init. 0 otherwise.
1568 static inline int is_global_init(struct task_struct *tsk)
1570 return task_tgid_nr(tsk) == 1;
1573 extern struct pid *cad_pid;
1578 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1579 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1580 #define PF_EXITING 0x00000004 /* Getting shut down */
1581 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1582 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1583 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1584 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1585 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1586 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1587 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1588 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1589 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1590 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1591 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1592 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1593 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1594 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1595 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1596 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1597 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1598 * I am cleaning dirty pages from some other bdi. */
1599 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1600 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1601 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1602 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1603 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1604 #define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1605 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1606 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1609 * Only the _current_ task can read/write to tsk->flags, but other
1610 * tasks can access tsk->flags in readonly mode for example
1611 * with tsk_used_math (like during threaded core dumping).
1612 * There is however an exception to this rule during ptrace
1613 * or during fork: the ptracer task is allowed to write to the
1614 * child->flags of its traced child (same goes for fork, the parent
1615 * can write to the child->flags), because we're guaranteed the
1616 * child is not running and in turn not changing child->flags
1617 * at the same time the parent does it.
1619 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1620 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1621 #define clear_used_math() clear_stopped_child_used_math(current)
1622 #define set_used_math() set_stopped_child_used_math(current)
1624 #define conditional_stopped_child_used_math(condition, child) \
1625 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1627 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1629 #define copy_to_stopped_child_used_math(child) \
1630 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1632 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1633 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1634 #define used_math() tsk_used_math(current)
1636 static inline bool is_percpu_thread(void)
1639 return (current->flags & PF_NO_SETAFFINITY) &&
1640 (current->nr_cpus_allowed == 1);
1646 /* Per-process atomic flags. */
1647 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1648 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1649 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1650 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1651 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1652 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1653 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1654 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1656 #define TASK_PFA_TEST(name, func) \
1657 static inline bool task_##func(struct task_struct *p) \
1658 { return test_bit(PFA_##name, &p->atomic_flags); }
1660 #define TASK_PFA_SET(name, func) \
1661 static inline void task_set_##func(struct task_struct *p) \
1662 { set_bit(PFA_##name, &p->atomic_flags); }
1664 #define TASK_PFA_CLEAR(name, func) \
1665 static inline void task_clear_##func(struct task_struct *p) \
1666 { clear_bit(PFA_##name, &p->atomic_flags); }
1668 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1669 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1671 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1672 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1673 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1675 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1676 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1677 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1679 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1680 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1681 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1683 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1684 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1685 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1687 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1688 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1690 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1691 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1692 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1694 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1695 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1698 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1700 current->flags &= ~flags;
1701 current->flags |= orig_flags & flags;
1704 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1705 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1707 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1708 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1709 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1710 extern void release_user_cpus_ptr(struct task_struct *p);
1711 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1712 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1713 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1715 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1718 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1720 if (!cpumask_test_cpu(0, new_mask))
1724 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1726 if (src->user_cpus_ptr)
1730 static inline void release_user_cpus_ptr(struct task_struct *p)
1732 WARN_ON(p->user_cpus_ptr);
1735 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1741 extern int yield_to(struct task_struct *p, bool preempt);
1742 extern void set_user_nice(struct task_struct *p, long nice);
1743 extern int task_prio(const struct task_struct *p);
1746 * task_nice - return the nice value of a given task.
1747 * @p: the task in question.
1749 * Return: The nice value [ -20 ... 0 ... 19 ].
1751 static inline int task_nice(const struct task_struct *p)
1753 return PRIO_TO_NICE((p)->static_prio);
1756 extern int can_nice(const struct task_struct *p, const int nice);
1757 extern int task_curr(const struct task_struct *p);
1758 extern int idle_cpu(int cpu);
1759 extern int available_idle_cpu(int cpu);
1760 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1761 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1762 extern void sched_set_fifo(struct task_struct *p);
1763 extern void sched_set_fifo_low(struct task_struct *p);
1764 extern void sched_set_normal(struct task_struct *p, int nice);
1765 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1766 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1767 extern struct task_struct *idle_task(int cpu);
1770 * is_idle_task - is the specified task an idle task?
1771 * @p: the task in question.
1773 * Return: 1 if @p is an idle task. 0 otherwise.
1775 static __always_inline bool is_idle_task(const struct task_struct *p)
1777 return !!(p->flags & PF_IDLE);
1780 extern struct task_struct *curr_task(int cpu);
1781 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1785 union thread_union {
1786 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1787 struct task_struct task;
1789 #ifndef CONFIG_THREAD_INFO_IN_TASK
1790 struct thread_info thread_info;
1792 unsigned long stack[THREAD_SIZE/sizeof(long)];
1795 #ifndef CONFIG_THREAD_INFO_IN_TASK
1796 extern struct thread_info init_thread_info;
1799 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1801 #ifdef CONFIG_THREAD_INFO_IN_TASK
1802 static inline struct thread_info *task_thread_info(struct task_struct *task)
1804 return &task->thread_info;
1806 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1807 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1811 * find a task by one of its numerical ids
1813 * find_task_by_pid_ns():
1814 * finds a task by its pid in the specified namespace
1815 * find_task_by_vpid():
1816 * finds a task by its virtual pid
1818 * see also find_vpid() etc in include/linux/pid.h
1821 extern struct task_struct *find_task_by_vpid(pid_t nr);
1822 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1825 * find a task by its virtual pid and get the task struct
1827 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1829 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1830 extern int wake_up_process(struct task_struct *tsk);
1831 extern void wake_up_new_task(struct task_struct *tsk);
1834 extern void kick_process(struct task_struct *tsk);
1836 static inline void kick_process(struct task_struct *tsk) { }
1839 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1841 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1843 __set_task_comm(tsk, from, false);
1846 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1847 #define get_task_comm(buf, tsk) ({ \
1848 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1849 __get_task_comm(buf, sizeof(buf), tsk); \
1853 static __always_inline void scheduler_ipi(void)
1856 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1857 * TIF_NEED_RESCHED remotely (for the first time) will also send
1860 preempt_fold_need_resched();
1862 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1864 static inline void scheduler_ipi(void) { }
1865 static inline unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state)
1872 * Set thread flags in other task's structures.
1873 * See asm/thread_info.h for TIF_xxxx flags available:
1875 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1877 set_ti_thread_flag(task_thread_info(tsk), flag);
1880 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1882 clear_ti_thread_flag(task_thread_info(tsk), flag);
1885 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1888 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1891 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1893 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1896 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1898 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1901 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1903 return test_ti_thread_flag(task_thread_info(tsk), flag);
1906 static inline void set_tsk_need_resched(struct task_struct *tsk)
1908 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1911 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1913 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1916 static inline int test_tsk_need_resched(struct task_struct *tsk)
1918 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1922 * cond_resched() and cond_resched_lock(): latency reduction via
1923 * explicit rescheduling in places that are safe. The return
1924 * value indicates whether a reschedule was done in fact.
1925 * cond_resched_lock() will drop the spinlock before scheduling,
1927 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1928 extern int __cond_resched(void);
1930 #ifdef CONFIG_PREEMPT_DYNAMIC
1932 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1934 static __always_inline int _cond_resched(void)
1936 return static_call_mod(cond_resched)();
1941 static inline int _cond_resched(void)
1943 return __cond_resched();
1946 #endif /* CONFIG_PREEMPT_DYNAMIC */
1950 static inline int _cond_resched(void) { return 0; }
1952 #endif /* !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC) */
1954 #define cond_resched() ({ \
1955 ___might_sleep(__FILE__, __LINE__, 0); \
1959 extern int __cond_resched_lock(spinlock_t *lock);
1960 extern int __cond_resched_rwlock_read(rwlock_t *lock);
1961 extern int __cond_resched_rwlock_write(rwlock_t *lock);
1963 #define cond_resched_lock(lock) ({ \
1964 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1965 __cond_resched_lock(lock); \
1968 #define cond_resched_rwlock_read(lock) ({ \
1969 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1970 __cond_resched_rwlock_read(lock); \
1973 #define cond_resched_rwlock_write(lock) ({ \
1974 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
1975 __cond_resched_rwlock_write(lock); \
1978 static inline void cond_resched_rcu(void)
1980 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1988 * Does a critical section need to be broken due to another
1989 * task waiting?: (technically does not depend on CONFIG_PREEMPTION,
1990 * but a general need for low latency)
1992 static inline int spin_needbreak(spinlock_t *lock)
1994 #ifdef CONFIG_PREEMPTION
1995 return spin_is_contended(lock);
2002 * Check if a rwlock is contended.
2003 * Returns non-zero if there is another task waiting on the rwlock.
2004 * Returns zero if the lock is not contended or the system / underlying
2005 * rwlock implementation does not support contention detection.
2006 * Technically does not depend on CONFIG_PREEMPTION, but a general need
2009 static inline int rwlock_needbreak(rwlock_t *lock)
2011 #ifdef CONFIG_PREEMPTION
2012 return rwlock_is_contended(lock);
2018 static __always_inline bool need_resched(void)
2020 return unlikely(tif_need_resched());
2024 * Wrappers for p->thread_info->cpu access. No-op on UP.
2028 static inline unsigned int task_cpu(const struct task_struct *p)
2030 #ifdef CONFIG_THREAD_INFO_IN_TASK
2031 return READ_ONCE(p->cpu);
2033 return READ_ONCE(task_thread_info(p)->cpu);
2037 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2041 static inline unsigned int task_cpu(const struct task_struct *p)
2046 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2050 #endif /* CONFIG_SMP */
2052 extern bool sched_task_on_rq(struct task_struct *p);
2055 * In order to reduce various lock holder preemption latencies provide an
2056 * interface to see if a vCPU is currently running or not.
2058 * This allows us to terminate optimistic spin loops and block, analogous to
2059 * the native optimistic spin heuristic of testing if the lock owner task is
2062 #ifndef vcpu_is_preempted
2063 static inline bool vcpu_is_preempted(int cpu)
2069 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2070 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2072 #ifndef TASK_SIZE_OF
2073 #define TASK_SIZE_OF(tsk) TASK_SIZE
2077 /* Returns effective CPU energy utilization, as seen by the scheduler */
2078 unsigned long sched_cpu_util(int cpu, unsigned long max);
2079 #endif /* CONFIG_SMP */
2084 * Map the event mask on the user-space ABI enum rseq_cs_flags
2085 * for direct mask checks.
2087 enum rseq_event_mask_bits {
2088 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
2089 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
2090 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
2093 enum rseq_event_mask {
2094 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
2095 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
2096 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
2099 static inline void rseq_set_notify_resume(struct task_struct *t)
2102 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
2105 void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
2107 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2108 struct pt_regs *regs)
2111 __rseq_handle_notify_resume(ksig, regs);
2114 static inline void rseq_signal_deliver(struct ksignal *ksig,
2115 struct pt_regs *regs)
2118 __set_bit(RSEQ_EVENT_SIGNAL_BIT, ¤t->rseq_event_mask);
2120 rseq_handle_notify_resume(ksig, regs);
2123 /* rseq_preempt() requires preemption to be disabled. */
2124 static inline void rseq_preempt(struct task_struct *t)
2126 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
2127 rseq_set_notify_resume(t);
2130 /* rseq_migrate() requires preemption to be disabled. */
2131 static inline void rseq_migrate(struct task_struct *t)
2133 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
2134 rseq_set_notify_resume(t);
2138 * If parent process has a registered restartable sequences area, the
2139 * child inherits. Unregister rseq for a clone with CLONE_VM set.
2141 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2143 if (clone_flags & CLONE_VM) {
2146 t->rseq_event_mask = 0;
2148 t->rseq = current->rseq;
2149 t->rseq_sig = current->rseq_sig;
2150 t->rseq_event_mask = current->rseq_event_mask;
2154 static inline void rseq_execve(struct task_struct *t)
2158 t->rseq_event_mask = 0;
2163 static inline void rseq_set_notify_resume(struct task_struct *t)
2166 static inline void rseq_handle_notify_resume(struct ksignal *ksig,
2167 struct pt_regs *regs)
2170 static inline void rseq_signal_deliver(struct ksignal *ksig,
2171 struct pt_regs *regs)
2174 static inline void rseq_preempt(struct task_struct *t)
2177 static inline void rseq_migrate(struct task_struct *t)
2180 static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
2183 static inline void rseq_execve(struct task_struct *t)
2189 #ifdef CONFIG_DEBUG_RSEQ
2191 void rseq_syscall(struct pt_regs *regs);
2195 static inline void rseq_syscall(struct pt_regs *regs)
2201 const struct sched_avg *sched_trace_cfs_rq_avg(struct cfs_rq *cfs_rq);
2202 char *sched_trace_cfs_rq_path(struct cfs_rq *cfs_rq, char *str, int len);
2203 int sched_trace_cfs_rq_cpu(struct cfs_rq *cfs_rq);
2205 const struct sched_avg *sched_trace_rq_avg_rt(struct rq *rq);
2206 const struct sched_avg *sched_trace_rq_avg_dl(struct rq *rq);
2207 const struct sched_avg *sched_trace_rq_avg_irq(struct rq *rq);
2209 int sched_trace_rq_cpu(struct rq *rq);
2210 int sched_trace_rq_cpu_capacity(struct rq *rq);
2211 int sched_trace_rq_nr_running(struct rq *rq);
2213 const struct cpumask *sched_trace_rd_span(struct root_domain *rd);
2215 #ifdef CONFIG_SCHED_CORE
2216 extern void sched_core_free(struct task_struct *tsk);
2217 extern void sched_core_fork(struct task_struct *p);
2218 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2219 unsigned long uaddr);
2221 static inline void sched_core_free(struct task_struct *tsk) { }
2222 static inline void sched_core_fork(struct task_struct *p) { }