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>
13 #include <asm/processor.h>
14 #include <linux/thread_info.h>
15 #include <linux/preempt.h>
16 #include <linux/cpumask_types.h>
18 #include <linux/cache.h>
19 #include <linux/irqflags_types.h>
20 #include <linux/smp_types.h>
21 #include <linux/pid_types.h>
22 #include <linux/sem_types.h>
23 #include <linux/shm.h>
24 #include <linux/kmsan_types.h>
25 #include <linux/mutex_types.h>
26 #include <linux/plist_types.h>
27 #include <linux/hrtimer_types.h>
28 #include <linux/timer_types.h>
29 #include <linux/seccomp_types.h>
30 #include <linux/nodemask_types.h>
31 #include <linux/refcount_types.h>
32 #include <linux/resource.h>
33 #include <linux/latencytop.h>
34 #include <linux/sched/prio.h>
35 #include <linux/sched/types.h>
36 #include <linux/signal_types.h>
37 #include <linux/syscall_user_dispatch_types.h>
38 #include <linux/mm_types_task.h>
39 #include <linux/netdevice_xmit.h>
40 #include <linux/task_io_accounting.h>
41 #include <linux/posix-timers_types.h>
42 #include <linux/restart_block.h>
43 #include <uapi/linux/rseq.h>
44 #include <linux/seqlock_types.h>
45 #include <linux/kcsan.h>
47 #include <linux/livepatch_sched.h>
48 #include <linux/uidgid_types.h>
49 #include <asm/kmap_size.h>
51 /* task_struct member predeclarations (sorted alphabetically): */
55 struct bpf_local_storage;
57 struct bpf_net_context;
58 struct capture_control;
61 struct futex_pi_state;
67 struct perf_event_context;
69 struct pipe_inode_info;
72 struct robust_list_head;
76 struct sched_dl_entity;
78 struct sighand_struct;
80 struct task_delay_info;
86 * Task state bitmask. NOTE! These bits are also
87 * encoded in fs/proc/array.c: get_task_state().
89 * We have two separate sets of flags: task->__state
90 * is about runnability, while task->exit_state are
91 * about the task exiting. Confusing, but this way
92 * modifying one set can't modify the other one by
96 /* Used in tsk->__state: */
97 #define TASK_RUNNING 0x00000000
98 #define TASK_INTERRUPTIBLE 0x00000001
99 #define TASK_UNINTERRUPTIBLE 0x00000002
100 #define __TASK_STOPPED 0x00000004
101 #define __TASK_TRACED 0x00000008
102 /* Used in tsk->exit_state: */
103 #define EXIT_DEAD 0x00000010
104 #define EXIT_ZOMBIE 0x00000020
105 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
106 /* Used in tsk->__state again: */
107 #define TASK_PARKED 0x00000040
108 #define TASK_DEAD 0x00000080
109 #define TASK_WAKEKILL 0x00000100
110 #define TASK_WAKING 0x00000200
111 #define TASK_NOLOAD 0x00000400
112 #define TASK_NEW 0x00000800
113 #define TASK_RTLOCK_WAIT 0x00001000
114 #define TASK_FREEZABLE 0x00002000
115 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
116 #define TASK_FROZEN 0x00008000
117 #define TASK_STATE_MAX 0x00010000
119 #define TASK_ANY (TASK_STATE_MAX-1)
122 * DO NOT ADD ANY NEW USERS !
124 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
126 /* Convenience macros for the sake of set_current_state: */
127 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
128 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
129 #define TASK_TRACED __TASK_TRACED
131 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
133 /* Convenience macros for the sake of wake_up(): */
134 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
136 /* get_task_state(): */
137 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
138 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
139 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
142 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
144 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
145 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
146 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
149 * Special states are those that do not use the normal wait-loop pattern. See
150 * the comment with set_special_state().
152 #define is_special_task_state(state) \
153 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | \
154 TASK_DEAD | TASK_FROZEN))
156 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
157 # define debug_normal_state_change(state_value) \
159 WARN_ON_ONCE(is_special_task_state(state_value)); \
160 current->task_state_change = _THIS_IP_; \
163 # define debug_special_state_change(state_value) \
165 WARN_ON_ONCE(!is_special_task_state(state_value)); \
166 current->task_state_change = _THIS_IP_; \
169 # define debug_rtlock_wait_set_state() \
171 current->saved_state_change = current->task_state_change;\
172 current->task_state_change = _THIS_IP_; \
175 # define debug_rtlock_wait_restore_state() \
177 current->task_state_change = current->saved_state_change;\
181 # define debug_normal_state_change(cond) do { } while (0)
182 # define debug_special_state_change(cond) do { } while (0)
183 # define debug_rtlock_wait_set_state() do { } while (0)
184 # define debug_rtlock_wait_restore_state() do { } while (0)
188 * set_current_state() includes a barrier so that the write of current->__state
189 * is correctly serialised wrt the caller's subsequent test of whether to
193 * set_current_state(TASK_UNINTERRUPTIBLE);
199 * __set_current_state(TASK_RUNNING);
201 * If the caller does not need such serialisation (because, for instance, the
202 * CONDITION test and condition change and wakeup are under the same lock) then
203 * use __set_current_state().
205 * The above is typically ordered against the wakeup, which does:
208 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
210 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
211 * accessing p->__state.
213 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
214 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
215 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
217 * However, with slightly different timing the wakeup TASK_RUNNING store can
218 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
219 * a problem either because that will result in one extra go around the loop
220 * and our @cond test will save the day.
222 * Also see the comments of try_to_wake_up().
224 #define __set_current_state(state_value) \
226 debug_normal_state_change((state_value)); \
227 WRITE_ONCE(current->__state, (state_value)); \
230 #define set_current_state(state_value) \
232 debug_normal_state_change((state_value)); \
233 smp_store_mb(current->__state, (state_value)); \
237 * set_special_state() should be used for those states when the blocking task
238 * can not use the regular condition based wait-loop. In that case we must
239 * serialize against wakeups such that any possible in-flight TASK_RUNNING
240 * stores will not collide with our state change.
242 #define set_special_state(state_value) \
244 unsigned long flags; /* may shadow */ \
246 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
247 debug_special_state_change((state_value)); \
248 WRITE_ONCE(current->__state, (state_value)); \
249 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
253 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
255 * RT's spin/rwlock substitutions are state preserving. The state of the
256 * task when blocking on the lock is saved in task_struct::saved_state and
257 * restored after the lock has been acquired. These operations are
258 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
259 * lock related wakeups while the task is blocked on the lock are
260 * redirected to operate on task_struct::saved_state to ensure that these
261 * are not dropped. On restore task_struct::saved_state is set to
262 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
264 * The lock operation looks like this:
266 * current_save_and_set_rtlock_wait_state();
270 * raw_spin_unlock_irq(&lock->wait_lock);
272 * raw_spin_lock_irq(&lock->wait_lock);
273 * set_current_state(TASK_RTLOCK_WAIT);
275 * current_restore_rtlock_saved_state();
277 #define current_save_and_set_rtlock_wait_state() \
279 lockdep_assert_irqs_disabled(); \
280 raw_spin_lock(¤t->pi_lock); \
281 current->saved_state = current->__state; \
282 debug_rtlock_wait_set_state(); \
283 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
284 raw_spin_unlock(¤t->pi_lock); \
287 #define current_restore_rtlock_saved_state() \
289 lockdep_assert_irqs_disabled(); \
290 raw_spin_lock(¤t->pi_lock); \
291 debug_rtlock_wait_restore_state(); \
292 WRITE_ONCE(current->__state, current->saved_state); \
293 current->saved_state = TASK_RUNNING; \
294 raw_spin_unlock(¤t->pi_lock); \
297 #define get_current_state() READ_ONCE(current->__state)
300 * Define the task command name length as enum, then it can be visible to
307 extern void sched_tick(void);
309 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
311 extern long schedule_timeout(long timeout);
312 extern long schedule_timeout_interruptible(long timeout);
313 extern long schedule_timeout_killable(long timeout);
314 extern long schedule_timeout_uninterruptible(long timeout);
315 extern long schedule_timeout_idle(long timeout);
316 asmlinkage void schedule(void);
317 extern void schedule_preempt_disabled(void);
318 asmlinkage void preempt_schedule_irq(void);
319 #ifdef CONFIG_PREEMPT_RT
320 extern void schedule_rtlock(void);
323 extern int __must_check io_schedule_prepare(void);
324 extern void io_schedule_finish(int token);
325 extern long io_schedule_timeout(long timeout);
326 extern void io_schedule(void);
329 * struct prev_cputime - snapshot of system and user cputime
330 * @utime: time spent in user mode
331 * @stime: time spent in system mode
332 * @lock: protects the above two fields
334 * Stores previous user/system time values such that we can guarantee
337 struct prev_cputime {
338 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
346 /* Task is sleeping or running in a CPU with VTIME inactive: */
350 /* Task runs in kernelspace in a CPU with VTIME active: */
352 /* Task runs in userspace in a CPU with VTIME active: */
354 /* Task runs as guests in a CPU with VTIME active: */
360 unsigned long long starttime;
361 enum vtime_state state;
369 * Utilization clamp constraints.
370 * @UCLAMP_MIN: Minimum utilization
371 * @UCLAMP_MAX: Maximum utilization
372 * @UCLAMP_CNT: Utilization clamp constraints count
381 extern struct root_domain def_root_domain;
382 extern struct mutex sched_domains_mutex;
390 #ifdef CONFIG_SCHED_INFO
391 /* Cumulative counters: */
393 /* # of times we have run on this CPU: */
394 unsigned long pcount;
396 /* Time spent waiting on a runqueue: */
397 unsigned long long run_delay;
401 /* When did we last run on a CPU? */
402 unsigned long long last_arrival;
404 /* When were we last queued to run? */
405 unsigned long long last_queued;
407 #endif /* CONFIG_SCHED_INFO */
411 * Integer metrics need fixed point arithmetic, e.g., sched/fair
412 * has a few: load, load_avg, util_avg, freq, and capacity.
414 * We define a basic fixed point arithmetic range, and then formalize
415 * all these metrics based on that basic range.
417 # define SCHED_FIXEDPOINT_SHIFT 10
418 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
420 /* Increase resolution of cpu_capacity calculations */
421 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
422 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
425 unsigned long weight;
430 * The load/runnable/util_avg accumulates an infinite geometric series
431 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
433 * [load_avg definition]
435 * load_avg = runnable% * scale_load_down(load)
437 * [runnable_avg definition]
439 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
441 * [util_avg definition]
443 * util_avg = running% * SCHED_CAPACITY_SCALE
445 * where runnable% is the time ratio that a sched_entity is runnable and
446 * running% the time ratio that a sched_entity is running.
448 * For cfs_rq, they are the aggregated values of all runnable and blocked
451 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
452 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
453 * for computing those signals (see update_rq_clock_pelt())
455 * N.B., the above ratios (runnable% and running%) themselves are in the
456 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
457 * to as large a range as necessary. This is for example reflected by
458 * util_avg's SCHED_CAPACITY_SCALE.
462 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
463 * with the highest load (=88761), always runnable on a single cfs_rq,
464 * and should not overflow as the number already hits PID_MAX_LIMIT.
466 * For all other cases (including 32-bit kernels), struct load_weight's
467 * weight will overflow first before we do, because:
469 * Max(load_avg) <= Max(load.weight)
471 * Then it is the load_weight's responsibility to consider overflow
475 u64 last_update_time;
480 unsigned long load_avg;
481 unsigned long runnable_avg;
482 unsigned long util_avg;
483 unsigned int util_est;
484 } ____cacheline_aligned;
487 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
488 * updates. When a task is dequeued, its util_est should not be updated if its
489 * util_avg has not been updated in the meantime.
490 * This information is mapped into the MSB bit of util_est at dequeue time.
491 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
492 * it is safe to use MSB.
494 #define UTIL_EST_WEIGHT_SHIFT 2
495 #define UTIL_AVG_UNCHANGED 0x80000000
497 struct sched_statistics {
498 #ifdef CONFIG_SCHEDSTATS
508 s64 sum_sleep_runtime;
512 s64 sum_block_runtime;
517 u64 nr_migrations_cold;
518 u64 nr_failed_migrations_affine;
519 u64 nr_failed_migrations_running;
520 u64 nr_failed_migrations_hot;
521 u64 nr_forced_migrations;
525 u64 nr_wakeups_migrate;
526 u64 nr_wakeups_local;
527 u64 nr_wakeups_remote;
528 u64 nr_wakeups_affine;
529 u64 nr_wakeups_affine_attempts;
530 u64 nr_wakeups_passive;
533 #ifdef CONFIG_SCHED_CORE
534 u64 core_forceidle_sum;
536 #endif /* CONFIG_SCHEDSTATS */
537 } ____cacheline_aligned;
539 struct sched_entity {
540 /* For load-balancing: */
541 struct load_weight load;
542 struct rb_node run_node;
547 struct list_head group_node;
549 unsigned char sched_delayed;
550 unsigned char rel_deadline;
551 unsigned char custom_slice;
555 u64 sum_exec_runtime;
556 u64 prev_sum_exec_runtime;
563 #ifdef CONFIG_FAIR_GROUP_SCHED
565 struct sched_entity *parent;
566 /* rq on which this entity is (to be) queued: */
567 struct cfs_rq *cfs_rq;
568 /* rq "owned" by this entity/group: */
570 /* cached value of my_q->h_nr_running */
571 unsigned long runnable_weight;
576 * Per entity load average tracking.
578 * Put into separate cache line so it does not
579 * collide with read-mostly values above.
581 struct sched_avg avg;
585 struct sched_rt_entity {
586 struct list_head run_list;
587 unsigned long timeout;
588 unsigned long watchdog_stamp;
589 unsigned int time_slice;
590 unsigned short on_rq;
591 unsigned short on_list;
593 struct sched_rt_entity *back;
594 #ifdef CONFIG_RT_GROUP_SCHED
595 struct sched_rt_entity *parent;
596 /* rq on which this entity is (to be) queued: */
598 /* rq "owned" by this entity/group: */
601 } __randomize_layout;
603 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
604 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
606 struct sched_dl_entity {
607 struct rb_node rb_node;
610 * Original scheduling parameters. Copied here from sched_attr
611 * during sched_setattr(), they will remain the same until
612 * the next sched_setattr().
614 u64 dl_runtime; /* Maximum runtime for each instance */
615 u64 dl_deadline; /* Relative deadline of each instance */
616 u64 dl_period; /* Separation of two instances (period) */
617 u64 dl_bw; /* dl_runtime / dl_period */
618 u64 dl_density; /* dl_runtime / dl_deadline */
621 * Actual scheduling parameters. Initialized with the values above,
622 * they are continuously updated during task execution. Note that
623 * the remaining runtime could be < 0 in case we are in overrun.
625 s64 runtime; /* Remaining runtime for this instance */
626 u64 deadline; /* Absolute deadline for this instance */
627 unsigned int flags; /* Specifying the scheduler behaviour */
632 * @dl_throttled tells if we exhausted the runtime. If so, the
633 * task has to wait for a replenishment to be performed at the
634 * next firing of dl_timer.
636 * @dl_yielded tells if task gave up the CPU before consuming
637 * all its available runtime during the last job.
639 * @dl_non_contending tells if the task is inactive while still
640 * contributing to the active utilization. In other words, it
641 * indicates if the inactive timer has been armed and its handler
642 * has not been executed yet. This flag is useful to avoid race
643 * conditions between the inactive timer handler and the wakeup
646 * @dl_overrun tells if the task asked to be informed about runtime
649 * @dl_server tells if this is a server entity.
651 * @dl_defer tells if this is a deferred or regular server. For
652 * now only defer server exists.
654 * @dl_defer_armed tells if the deferrable server is waiting
655 * for the replenishment timer to activate it.
657 * @dl_defer_running tells if the deferrable server is actually
658 * running, skipping the defer phase.
660 unsigned int dl_throttled : 1;
661 unsigned int dl_yielded : 1;
662 unsigned int dl_non_contending : 1;
663 unsigned int dl_overrun : 1;
664 unsigned int dl_server : 1;
665 unsigned int dl_defer : 1;
666 unsigned int dl_defer_armed : 1;
667 unsigned int dl_defer_running : 1;
670 * Bandwidth enforcement timer. Each -deadline task has its
671 * own bandwidth to be enforced, thus we need one timer per task.
673 struct hrtimer dl_timer;
676 * Inactive timer, responsible for decreasing the active utilization
677 * at the "0-lag time". When a -deadline task blocks, it contributes
678 * to GRUB's active utilization until the "0-lag time", hence a
679 * timer is needed to decrease the active utilization at the correct
682 struct hrtimer inactive_timer;
685 * Bits for DL-server functionality. Also see the comment near
686 * dl_server_update().
688 * @rq the runqueue this server is for
690 * @server_has_tasks() returns true if @server_pick return a
694 dl_server_has_tasks_f server_has_tasks;
695 dl_server_pick_f server_pick_task;
697 #ifdef CONFIG_RT_MUTEXES
699 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
700 * pi_se points to the donor, otherwise points to the dl_se it belongs
701 * to (the original one/itself).
703 struct sched_dl_entity *pi_se;
707 #ifdef CONFIG_UCLAMP_TASK
708 /* Number of utilization clamp buckets (shorter alias) */
709 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
712 * Utilization clamp for a scheduling entity
713 * @value: clamp value "assigned" to a se
714 * @bucket_id: bucket index corresponding to the "assigned" value
715 * @active: the se is currently refcounted in a rq's bucket
716 * @user_defined: the requested clamp value comes from user-space
718 * The bucket_id is the index of the clamp bucket matching the clamp value
719 * which is pre-computed and stored to avoid expensive integer divisions from
722 * The active bit is set whenever a task has got an "effective" value assigned,
723 * which can be different from the clamp value "requested" from user-space.
724 * This allows to know a task is refcounted in the rq's bucket corresponding
725 * to the "effective" bucket_id.
727 * The user_defined bit is set whenever a task has got a task-specific clamp
728 * value requested from userspace, i.e. the system defaults apply to this task
729 * just as a restriction. This allows to relax default clamps when a less
730 * restrictive task-specific value has been requested, thus allowing to
731 * implement a "nice" semantic. For example, a task running with a 20%
732 * default boost can still drop its own boosting to 0%.
735 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
736 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
737 unsigned int active : 1;
738 unsigned int user_defined : 1;
740 #endif /* CONFIG_UCLAMP_TASK */
746 u8 exp_hint; /* Hint for performance. */
747 u8 need_mb; /* Readers need smp_mb(). */
749 u32 s; /* Set of bits. */
752 enum perf_event_task_context {
753 perf_invalid_context = -1,
756 perf_nr_task_contexts,
760 * Number of contexts where an event can trigger:
761 * task, softirq, hardirq, nmi.
763 #define PERF_NR_CONTEXTS 4
766 struct wake_q_node *next;
770 #ifdef CONFIG_KMAP_LOCAL
772 pte_t pteval[KM_MAX_IDX];
777 #ifdef CONFIG_THREAD_INFO_IN_TASK
779 * For reasons of header soup (see current_thread_info()), this
780 * must be the first element of task_struct.
782 struct thread_info thread_info;
784 unsigned int __state;
786 /* saved state for "spinlock sleepers" */
787 unsigned int saved_state;
790 * This begins the randomizable portion of task_struct. Only
791 * scheduling-critical items should be added above here.
793 randomized_struct_fields_start
797 /* Per task flags (PF_*), defined further below: */
801 #ifdef CONFIG_MEM_ALLOC_PROFILING
802 struct alloc_tag *alloc_tag;
807 struct __call_single_node wake_entry;
808 unsigned int wakee_flips;
809 unsigned long wakee_flip_decay_ts;
810 struct task_struct *last_wakee;
813 * recent_used_cpu is initially set as the last CPU used by a task
814 * that wakes affine another task. Waker/wakee relationships can
815 * push tasks around a CPU where each wakeup moves to the next one.
816 * Tracking a recently used CPU allows a quick search for a recently
817 * used CPU that may be idle.
827 unsigned int rt_priority;
829 struct sched_entity se;
830 struct sched_rt_entity rt;
831 struct sched_dl_entity dl;
832 struct sched_dl_entity *dl_server;
833 const struct sched_class *sched_class;
835 #ifdef CONFIG_SCHED_CORE
836 struct rb_node core_node;
837 unsigned long core_cookie;
838 unsigned int core_occupation;
841 #ifdef CONFIG_CGROUP_SCHED
842 struct task_group *sched_task_group;
846 #ifdef CONFIG_UCLAMP_TASK
848 * Clamp values requested for a scheduling entity.
849 * Must be updated with task_rq_lock() held.
851 struct uclamp_se uclamp_req[UCLAMP_CNT];
853 * Effective clamp values used for a scheduling entity.
854 * Must be updated with task_rq_lock() held.
856 struct uclamp_se uclamp[UCLAMP_CNT];
859 struct sched_statistics stats;
861 #ifdef CONFIG_PREEMPT_NOTIFIERS
862 /* List of struct preempt_notifier: */
863 struct hlist_head preempt_notifiers;
866 #ifdef CONFIG_BLK_DEV_IO_TRACE
867 unsigned int btrace_seq;
871 unsigned long max_allowed_capacity;
873 const cpumask_t *cpus_ptr;
874 cpumask_t *user_cpus_ptr;
876 void *migration_pending;
878 unsigned short migration_disabled;
880 unsigned short migration_flags;
882 #ifdef CONFIG_PREEMPT_RCU
883 int rcu_read_lock_nesting;
884 union rcu_special rcu_read_unlock_special;
885 struct list_head rcu_node_entry;
886 struct rcu_node *rcu_blocked_node;
887 #endif /* #ifdef CONFIG_PREEMPT_RCU */
889 #ifdef CONFIG_TASKS_RCU
890 unsigned long rcu_tasks_nvcsw;
891 u8 rcu_tasks_holdout;
893 int rcu_tasks_idle_cpu;
894 struct list_head rcu_tasks_holdout_list;
895 int rcu_tasks_exit_cpu;
896 struct list_head rcu_tasks_exit_list;
897 #endif /* #ifdef CONFIG_TASKS_RCU */
899 #ifdef CONFIG_TASKS_TRACE_RCU
900 int trc_reader_nesting;
902 union rcu_special trc_reader_special;
903 struct list_head trc_holdout_list;
904 struct list_head trc_blkd_node;
906 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
908 struct sched_info sched_info;
910 struct list_head tasks;
912 struct plist_node pushable_tasks;
913 struct rb_node pushable_dl_tasks;
916 struct mm_struct *mm;
917 struct mm_struct *active_mm;
918 struct address_space *faults_disabled_mapping;
923 /* The signal sent when the parent dies: */
925 /* JOBCTL_*, siglock protected: */
926 unsigned long jobctl;
928 /* Used for emulating ABI behavior of previous Linux versions: */
929 unsigned int personality;
931 /* Scheduler bits, serialized by scheduler locks: */
932 unsigned sched_reset_on_fork:1;
933 unsigned sched_contributes_to_load:1;
934 unsigned sched_migrated:1;
936 /* Force alignment to the next boundary: */
939 /* Unserialized, strictly 'current' */
942 * This field must not be in the scheduler word above due to wakelist
943 * queueing no longer being serialized by p->on_cpu. However:
946 * schedule() if (p->on_rq && ..) // false
947 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
948 * deactivate_task() ttwu_queue_wakelist())
949 * p->on_rq = 0; p->sched_remote_wakeup = Y;
951 * guarantees all stores of 'current' are visible before
952 * ->sched_remote_wakeup gets used, so it can be in this word.
954 unsigned sched_remote_wakeup:1;
955 #ifdef CONFIG_RT_MUTEXES
956 unsigned sched_rt_mutex:1;
959 /* Bit to tell TOMOYO we're in execve(): */
960 unsigned in_execve:1;
961 unsigned in_iowait:1;
962 #ifndef TIF_RESTORE_SIGMASK
963 unsigned restore_sigmask:1;
965 #ifdef CONFIG_MEMCG_V1
966 unsigned in_user_fault:1;
968 #ifdef CONFIG_LRU_GEN
969 /* whether the LRU algorithm may apply to this access */
970 unsigned in_lru_fault:1;
972 #ifdef CONFIG_COMPAT_BRK
973 unsigned brk_randomized:1;
975 #ifdef CONFIG_CGROUPS
976 /* disallow userland-initiated cgroup migration */
977 unsigned no_cgroup_migration:1;
978 /* task is frozen/stopped (used by the cgroup freezer) */
981 #ifdef CONFIG_BLK_CGROUP
982 unsigned use_memdelay:1;
985 /* Stalled due to lack of memory */
986 unsigned in_memstall:1;
988 #ifdef CONFIG_PAGE_OWNER
989 /* Used by page_owner=on to detect recursion in page tracking. */
990 unsigned in_page_owner:1;
992 #ifdef CONFIG_EVENTFD
993 /* Recursion prevention for eventfd_signal() */
994 unsigned in_eventfd:1;
996 #ifdef CONFIG_ARCH_HAS_CPU_PASID
997 unsigned pasid_activated:1;
999 #ifdef CONFIG_CPU_SUP_INTEL
1000 unsigned reported_split_lock:1;
1002 #ifdef CONFIG_TASK_DELAY_ACCT
1003 /* delay due to memory thrashing */
1004 unsigned in_thrashing:1;
1006 #ifdef CONFIG_PREEMPT_RT
1007 struct netdev_xmit net_xmit;
1009 unsigned long atomic_flags; /* Flags requiring atomic access. */
1011 struct restart_block restart_block;
1016 #ifdef CONFIG_STACKPROTECTOR
1017 /* Canary value for the -fstack-protector GCC feature: */
1018 unsigned long stack_canary;
1021 * Pointers to the (original) parent process, youngest child, younger sibling,
1022 * older sibling, respectively. (p->father can be replaced with
1023 * p->real_parent->pid)
1026 /* Real parent process: */
1027 struct task_struct __rcu *real_parent;
1029 /* Recipient of SIGCHLD, wait4() reports: */
1030 struct task_struct __rcu *parent;
1033 * Children/sibling form the list of natural children:
1035 struct list_head children;
1036 struct list_head sibling;
1037 struct task_struct *group_leader;
1040 * 'ptraced' is the list of tasks this task is using ptrace() on.
1042 * This includes both natural children and PTRACE_ATTACH targets.
1043 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1045 struct list_head ptraced;
1046 struct list_head ptrace_entry;
1048 /* PID/PID hash table linkage. */
1049 struct pid *thread_pid;
1050 struct hlist_node pid_links[PIDTYPE_MAX];
1051 struct list_head thread_node;
1053 struct completion *vfork_done;
1055 /* CLONE_CHILD_SETTID: */
1056 int __user *set_child_tid;
1058 /* CLONE_CHILD_CLEARTID: */
1059 int __user *clear_child_tid;
1061 /* PF_KTHREAD | PF_IO_WORKER */
1062 void *worker_private;
1066 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1071 struct prev_cputime prev_cputime;
1072 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1076 #ifdef CONFIG_NO_HZ_FULL
1077 atomic_t tick_dep_mask;
1079 /* Context switch counts: */
1080 unsigned long nvcsw;
1081 unsigned long nivcsw;
1083 /* Monotonic time in nsecs: */
1086 /* Boot based time in nsecs: */
1089 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1090 unsigned long min_flt;
1091 unsigned long maj_flt;
1093 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1094 struct posix_cputimers posix_cputimers;
1096 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1097 struct posix_cputimers_work posix_cputimers_work;
1100 /* Process credentials: */
1102 /* Tracer's credentials at attach: */
1103 const struct cred __rcu *ptracer_cred;
1105 /* Objective and real subjective task credentials (COW): */
1106 const struct cred __rcu *real_cred;
1108 /* Effective (overridable) subjective task credentials (COW): */
1109 const struct cred __rcu *cred;
1112 /* Cached requested key. */
1113 struct key *cached_requested_key;
1117 * executable name, excluding path.
1119 * - normally initialized setup_new_exec()
1120 * - access it with [gs]et_task_comm()
1121 * - lock it with task_lock()
1123 char comm[TASK_COMM_LEN];
1125 struct nameidata *nameidata;
1127 #ifdef CONFIG_SYSVIPC
1128 struct sysv_sem sysvsem;
1129 struct sysv_shm sysvshm;
1131 #ifdef CONFIG_DETECT_HUNG_TASK
1132 unsigned long last_switch_count;
1133 unsigned long last_switch_time;
1135 /* Filesystem information: */
1136 struct fs_struct *fs;
1138 /* Open file information: */
1139 struct files_struct *files;
1141 #ifdef CONFIG_IO_URING
1142 struct io_uring_task *io_uring;
1146 struct nsproxy *nsproxy;
1148 /* Signal handlers: */
1149 struct signal_struct *signal;
1150 struct sighand_struct __rcu *sighand;
1152 sigset_t real_blocked;
1153 /* Restored if set_restore_sigmask() was used: */
1154 sigset_t saved_sigmask;
1155 struct sigpending pending;
1156 unsigned long sas_ss_sp;
1158 unsigned int sas_ss_flags;
1160 struct callback_head *task_works;
1163 #ifdef CONFIG_AUDITSYSCALL
1164 struct audit_context *audit_context;
1167 unsigned int sessionid;
1169 struct seccomp seccomp;
1170 struct syscall_user_dispatch syscall_dispatch;
1172 /* Thread group tracking: */
1176 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1177 spinlock_t alloc_lock;
1179 /* Protection of the PI data structures: */
1180 raw_spinlock_t pi_lock;
1182 struct wake_q_node wake_q;
1184 #ifdef CONFIG_RT_MUTEXES
1185 /* PI waiters blocked on a rt_mutex held by this task: */
1186 struct rb_root_cached pi_waiters;
1187 /* Updated under owner's pi_lock and rq lock */
1188 struct task_struct *pi_top_task;
1189 /* Deadlock detection and priority inheritance handling: */
1190 struct rt_mutex_waiter *pi_blocked_on;
1193 #ifdef CONFIG_DEBUG_MUTEXES
1194 /* Mutex deadlock detection: */
1195 struct mutex_waiter *blocked_on;
1198 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1199 int non_block_count;
1202 #ifdef CONFIG_TRACE_IRQFLAGS
1203 struct irqtrace_events irqtrace;
1204 unsigned int hardirq_threaded;
1205 u64 hardirq_chain_key;
1206 int softirqs_enabled;
1207 int softirq_context;
1210 #ifdef CONFIG_PREEMPT_RT
1211 int softirq_disable_cnt;
1214 #ifdef CONFIG_LOCKDEP
1215 # define MAX_LOCK_DEPTH 48UL
1218 unsigned int lockdep_recursion;
1219 struct held_lock held_locks[MAX_LOCK_DEPTH];
1222 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1223 unsigned int in_ubsan;
1226 /* Journalling filesystem info: */
1229 /* Stacked block device info: */
1230 struct bio_list *bio_list;
1232 /* Stack plugging: */
1233 struct blk_plug *plug;
1236 struct reclaim_state *reclaim_state;
1238 struct io_context *io_context;
1240 #ifdef CONFIG_COMPACTION
1241 struct capture_control *capture_control;
1244 unsigned long ptrace_message;
1245 kernel_siginfo_t *last_siginfo;
1247 struct task_io_accounting ioac;
1249 /* Pressure stall state */
1250 unsigned int psi_flags;
1252 #ifdef CONFIG_TASK_XACCT
1253 /* Accumulated RSS usage: */
1255 /* Accumulated virtual memory usage: */
1257 /* stime + utime since last update: */
1260 #ifdef CONFIG_CPUSETS
1261 /* Protected by ->alloc_lock: */
1262 nodemask_t mems_allowed;
1263 /* Sequence number to catch updates: */
1264 seqcount_spinlock_t mems_allowed_seq;
1265 int cpuset_mem_spread_rotor;
1267 #ifdef CONFIG_CGROUPS
1268 /* Control Group info protected by css_set_lock: */
1269 struct css_set __rcu *cgroups;
1270 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1271 struct list_head cg_list;
1273 #ifdef CONFIG_X86_CPU_RESCTRL
1278 struct robust_list_head __user *robust_list;
1279 #ifdef CONFIG_COMPAT
1280 struct compat_robust_list_head __user *compat_robust_list;
1282 struct list_head pi_state_list;
1283 struct futex_pi_state *pi_state_cache;
1284 struct mutex futex_exit_mutex;
1285 unsigned int futex_state;
1287 #ifdef CONFIG_PERF_EVENTS
1288 u8 perf_recursion[PERF_NR_CONTEXTS];
1289 struct perf_event_context *perf_event_ctxp;
1290 struct mutex perf_event_mutex;
1291 struct list_head perf_event_list;
1293 #ifdef CONFIG_DEBUG_PREEMPT
1294 unsigned long preempt_disable_ip;
1297 /* Protected by alloc_lock: */
1298 struct mempolicy *mempolicy;
1301 short pref_node_fork;
1303 #ifdef CONFIG_NUMA_BALANCING
1305 unsigned int numa_scan_period;
1306 unsigned int numa_scan_period_max;
1307 int numa_preferred_nid;
1308 unsigned long numa_migrate_retry;
1309 /* Migration stamp: */
1311 u64 last_task_numa_placement;
1312 u64 last_sum_exec_runtime;
1313 struct callback_head numa_work;
1316 * This pointer is only modified for current in syscall and
1317 * pagefault context (and for tasks being destroyed), so it can be read
1318 * from any of the following contexts:
1319 * - RCU read-side critical section
1320 * - current->numa_group from everywhere
1321 * - task's runqueue locked, task not running
1323 struct numa_group __rcu *numa_group;
1326 * numa_faults is an array split into four regions:
1327 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1328 * in this precise order.
1330 * faults_memory: Exponential decaying average of faults on a per-node
1331 * basis. Scheduling placement decisions are made based on these
1332 * counts. The values remain static for the duration of a PTE scan.
1333 * faults_cpu: Track the nodes the process was running on when a NUMA
1334 * hinting fault was incurred.
1335 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1336 * during the current scan window. When the scan completes, the counts
1337 * in faults_memory and faults_cpu decay and these values are copied.
1339 unsigned long *numa_faults;
1340 unsigned long total_numa_faults;
1343 * numa_faults_locality tracks if faults recorded during the last
1344 * scan window were remote/local or failed to migrate. The task scan
1345 * period is adapted based on the locality of the faults with different
1346 * weights depending on whether they were shared or private faults
1348 unsigned long numa_faults_locality[3];
1350 unsigned long numa_pages_migrated;
1351 #endif /* CONFIG_NUMA_BALANCING */
1354 struct rseq __user *rseq;
1358 * RmW on rseq_event_mask must be performed atomically
1359 * with respect to preemption.
1361 unsigned long rseq_event_mask;
1364 #ifdef CONFIG_SCHED_MM_CID
1365 int mm_cid; /* Current cid in mm */
1366 int last_mm_cid; /* Most recent cid in mm */
1367 int migrate_from_cpu;
1368 int mm_cid_active; /* Whether cid bitmap is active */
1369 struct callback_head cid_work;
1372 struct tlbflush_unmap_batch tlb_ubc;
1374 /* Cache last used pipe for splice(): */
1375 struct pipe_inode_info *splice_pipe;
1377 struct page_frag task_frag;
1379 #ifdef CONFIG_TASK_DELAY_ACCT
1380 struct task_delay_info *delays;
1383 #ifdef CONFIG_FAULT_INJECTION
1385 unsigned int fail_nth;
1388 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1389 * balance_dirty_pages() for a dirty throttling pause:
1392 int nr_dirtied_pause;
1393 /* Start of a write-and-pause period: */
1394 unsigned long dirty_paused_when;
1396 #ifdef CONFIG_LATENCYTOP
1397 int latency_record_count;
1398 struct latency_record latency_record[LT_SAVECOUNT];
1401 * Time slack values; these are used to round up poll() and
1402 * select() etc timeout values. These are in nanoseconds.
1405 u64 default_timer_slack_ns;
1407 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1408 unsigned int kasan_depth;
1412 struct kcsan_ctx kcsan_ctx;
1413 #ifdef CONFIG_TRACE_IRQFLAGS
1414 struct irqtrace_events kcsan_save_irqtrace;
1416 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1417 int kcsan_stack_depth;
1422 struct kmsan_ctx kmsan_ctx;
1425 #if IS_ENABLED(CONFIG_KUNIT)
1426 struct kunit *kunit_test;
1429 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1430 /* Index of current stored address in ret_stack: */
1434 /* Stack of return addresses for return function tracing: */
1435 unsigned long *ret_stack;
1437 /* Timestamp for last schedule: */
1438 unsigned long long ftrace_timestamp;
1441 * Number of functions that haven't been traced
1442 * because of depth overrun:
1444 atomic_t trace_overrun;
1446 /* Pause tracing: */
1447 atomic_t tracing_graph_pause;
1450 #ifdef CONFIG_TRACING
1451 /* Bitmask and counter of trace recursion: */
1452 unsigned long trace_recursion;
1453 #endif /* CONFIG_TRACING */
1456 /* See kernel/kcov.c for more details. */
1458 /* Coverage collection mode enabled for this task (0 if disabled): */
1459 unsigned int kcov_mode;
1461 /* Size of the kcov_area: */
1462 unsigned int kcov_size;
1464 /* Buffer for coverage collection: */
1467 /* KCOV descriptor wired with this task or NULL: */
1470 /* KCOV common handle for remote coverage collection: */
1473 /* KCOV sequence number: */
1476 /* Collect coverage from softirq context: */
1477 unsigned int kcov_softirq;
1480 #ifdef CONFIG_MEMCG_V1
1481 struct mem_cgroup *memcg_in_oom;
1485 /* Number of pages to reclaim on returning to userland: */
1486 unsigned int memcg_nr_pages_over_high;
1488 /* Used by memcontrol for targeted memcg charge: */
1489 struct mem_cgroup *active_memcg;
1491 /* Cache for current->cgroups->memcg->objcg lookups: */
1492 struct obj_cgroup *objcg;
1495 #ifdef CONFIG_BLK_CGROUP
1496 struct gendisk *throttle_disk;
1499 #ifdef CONFIG_UPROBES
1500 struct uprobe_task *utask;
1502 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1503 unsigned int sequential_io;
1504 unsigned int sequential_io_avg;
1506 struct kmap_ctrl kmap_ctrl;
1507 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1508 unsigned long task_state_change;
1509 # ifdef CONFIG_PREEMPT_RT
1510 unsigned long saved_state_change;
1513 struct rcu_head rcu;
1514 refcount_t rcu_users;
1515 int pagefault_disabled;
1517 struct task_struct *oom_reaper_list;
1518 struct timer_list oom_reaper_timer;
1520 #ifdef CONFIG_VMAP_STACK
1521 struct vm_struct *stack_vm_area;
1523 #ifdef CONFIG_THREAD_INFO_IN_TASK
1524 /* A live task holds one reference: */
1525 refcount_t stack_refcount;
1527 #ifdef CONFIG_LIVEPATCH
1530 #ifdef CONFIG_SECURITY
1531 /* Used by LSM modules for access restriction: */
1534 #ifdef CONFIG_BPF_SYSCALL
1535 /* Used by BPF task local storage */
1536 struct bpf_local_storage __rcu *bpf_storage;
1537 /* Used for BPF run context */
1538 struct bpf_run_ctx *bpf_ctx;
1540 /* Used by BPF for per-TASK xdp storage */
1541 struct bpf_net_context *bpf_net_context;
1543 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1544 unsigned long lowest_stack;
1545 unsigned long prev_lowest_stack;
1548 #ifdef CONFIG_X86_MCE
1549 void __user *mce_vaddr;
1554 __mce_reserved : 62;
1555 struct callback_head mce_kill_me;
1559 #ifdef CONFIG_KRETPROBES
1560 struct llist_head kretprobe_instances;
1562 #ifdef CONFIG_RETHOOK
1563 struct llist_head rethooks;
1566 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1568 * If L1D flush is supported on mm context switch
1569 * then we use this callback head to queue kill work
1570 * to kill tasks that are not running on SMT disabled
1573 struct callback_head l1d_flush_kill;
1578 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1579 * If we find justification for more monitors, we can think
1580 * about adding more or developing a dynamic method. So far,
1581 * none of these are justified.
1583 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1586 #ifdef CONFIG_USER_EVENTS
1587 struct user_event_mm *user_event_mm;
1591 * New fields for task_struct should be added above here, so that
1592 * they are included in the randomized portion of task_struct.
1594 randomized_struct_fields_end
1596 /* CPU-specific state of this task: */
1597 struct thread_struct thread;
1600 * WARNING: on x86, 'thread_struct' contains a variable-sized
1601 * structure. It *MUST* be at the end of 'task_struct'.
1603 * Do not put anything below here!
1607 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1608 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1610 static inline unsigned int __task_state_index(unsigned int tsk_state,
1611 unsigned int tsk_exit_state)
1613 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1615 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1617 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1618 state = TASK_REPORT_IDLE;
1621 * We're lying here, but rather than expose a completely new task state
1622 * to userspace, we can make this appear as if the task has gone through
1623 * a regular rt_mutex_lock() call.
1625 if (tsk_state & TASK_RTLOCK_WAIT)
1626 state = TASK_UNINTERRUPTIBLE;
1631 static inline unsigned int task_state_index(struct task_struct *tsk)
1633 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1636 static inline char task_index_to_char(unsigned int state)
1638 static const char state_char[] = "RSDTtXZPI";
1640 BUILD_BUG_ON(TASK_REPORT_MAX * 2 != 1 << (sizeof(state_char) - 1));
1642 return state_char[state];
1645 static inline char task_state_to_char(struct task_struct *tsk)
1647 return task_index_to_char(task_state_index(tsk));
1650 extern struct pid *cad_pid;
1655 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1656 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1657 #define PF_EXITING 0x00000004 /* Getting shut down */
1658 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1659 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1660 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1661 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1662 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1663 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1664 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1665 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1666 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1667 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1668 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1669 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1670 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1671 #define PF__HOLE__00010000 0x00010000
1672 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1673 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1674 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1675 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1676 * I am cleaning dirty pages from some other bdi. */
1677 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1678 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1679 #define PF_MEMALLOC_NORECLAIM 0x00800000 /* All allocation requests will clear __GFP_DIRECT_RECLAIM */
1680 #define PF_MEMALLOC_NOWARN 0x01000000 /* All allocation requests will inherit __GFP_NOWARN */
1681 #define PF__HOLE__02000000 0x02000000
1682 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1683 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1684 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1685 * See memalloc_pin_save() */
1686 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1687 #define PF__HOLE__40000000 0x40000000
1688 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1691 * Only the _current_ task can read/write to tsk->flags, but other
1692 * tasks can access tsk->flags in readonly mode for example
1693 * with tsk_used_math (like during threaded core dumping).
1694 * There is however an exception to this rule during ptrace
1695 * or during fork: the ptracer task is allowed to write to the
1696 * child->flags of its traced child (same goes for fork, the parent
1697 * can write to the child->flags), because we're guaranteed the
1698 * child is not running and in turn not changing child->flags
1699 * at the same time the parent does it.
1701 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1702 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1703 #define clear_used_math() clear_stopped_child_used_math(current)
1704 #define set_used_math() set_stopped_child_used_math(current)
1706 #define conditional_stopped_child_used_math(condition, child) \
1707 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1709 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1711 #define copy_to_stopped_child_used_math(child) \
1712 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1714 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1715 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1716 #define used_math() tsk_used_math(current)
1718 static __always_inline bool is_percpu_thread(void)
1721 return (current->flags & PF_NO_SETAFFINITY) &&
1722 (current->nr_cpus_allowed == 1);
1728 /* Per-process atomic flags. */
1729 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1730 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1731 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1732 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1733 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1734 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1735 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1736 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1738 #define TASK_PFA_TEST(name, func) \
1739 static inline bool task_##func(struct task_struct *p) \
1740 { return test_bit(PFA_##name, &p->atomic_flags); }
1742 #define TASK_PFA_SET(name, func) \
1743 static inline void task_set_##func(struct task_struct *p) \
1744 { set_bit(PFA_##name, &p->atomic_flags); }
1746 #define TASK_PFA_CLEAR(name, func) \
1747 static inline void task_clear_##func(struct task_struct *p) \
1748 { clear_bit(PFA_##name, &p->atomic_flags); }
1750 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1751 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1753 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1754 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1755 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1757 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1758 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1759 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1761 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1762 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1763 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1765 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1766 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1767 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1769 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1770 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1772 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1773 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1774 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1776 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1777 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1780 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1782 current->flags &= ~flags;
1783 current->flags |= orig_flags & flags;
1786 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1787 extern int task_can_attach(struct task_struct *p);
1788 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1789 extern void dl_bw_free(int cpu, u64 dl_bw);
1792 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1793 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1796 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1798 * @new_mask: CPU affinity mask
1800 * Return: zero if successful, or a negative error code
1802 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1803 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1804 extern void release_user_cpus_ptr(struct task_struct *p);
1805 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1806 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1807 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1809 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1812 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1814 /* Opencoded cpumask_test_cpu(0, new_mask) to avoid dependency on cpumask.h */
1815 if ((*cpumask_bits(new_mask) & 1) == 0)
1819 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1821 if (src->user_cpus_ptr)
1825 static inline void release_user_cpus_ptr(struct task_struct *p)
1827 WARN_ON(p->user_cpus_ptr);
1830 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1836 extern int yield_to(struct task_struct *p, bool preempt);
1837 extern void set_user_nice(struct task_struct *p, long nice);
1838 extern int task_prio(const struct task_struct *p);
1841 * task_nice - return the nice value of a given task.
1842 * @p: the task in question.
1844 * Return: The nice value [ -20 ... 0 ... 19 ].
1846 static inline int task_nice(const struct task_struct *p)
1848 return PRIO_TO_NICE((p)->static_prio);
1851 extern int can_nice(const struct task_struct *p, const int nice);
1852 extern int task_curr(const struct task_struct *p);
1853 extern int idle_cpu(int cpu);
1854 extern int available_idle_cpu(int cpu);
1855 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1856 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1857 extern void sched_set_fifo(struct task_struct *p);
1858 extern void sched_set_fifo_low(struct task_struct *p);
1859 extern void sched_set_normal(struct task_struct *p, int nice);
1860 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1861 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1862 extern struct task_struct *idle_task(int cpu);
1865 * is_idle_task - is the specified task an idle task?
1866 * @p: the task in question.
1868 * Return: 1 if @p is an idle task. 0 otherwise.
1870 static __always_inline bool is_idle_task(const struct task_struct *p)
1872 return !!(p->flags & PF_IDLE);
1875 extern struct task_struct *curr_task(int cpu);
1876 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1880 union thread_union {
1881 struct task_struct task;
1882 #ifndef CONFIG_THREAD_INFO_IN_TASK
1883 struct thread_info thread_info;
1885 unsigned long stack[THREAD_SIZE/sizeof(long)];
1888 #ifndef CONFIG_THREAD_INFO_IN_TASK
1889 extern struct thread_info init_thread_info;
1892 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1894 #ifdef CONFIG_THREAD_INFO_IN_TASK
1895 # define task_thread_info(task) (&(task)->thread_info)
1896 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1897 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1901 * find a task by one of its numerical ids
1903 * find_task_by_pid_ns():
1904 * finds a task by its pid in the specified namespace
1905 * find_task_by_vpid():
1906 * finds a task by its virtual pid
1908 * see also find_vpid() etc in include/linux/pid.h
1911 extern struct task_struct *find_task_by_vpid(pid_t nr);
1912 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1915 * find a task by its virtual pid and get the task struct
1917 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1919 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1920 extern int wake_up_process(struct task_struct *tsk);
1921 extern void wake_up_new_task(struct task_struct *tsk);
1924 extern void kick_process(struct task_struct *tsk);
1926 static inline void kick_process(struct task_struct *tsk) { }
1929 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1931 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1933 __set_task_comm(tsk, from, false);
1936 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1937 #define get_task_comm(buf, tsk) ({ \
1938 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1939 __get_task_comm(buf, sizeof(buf), tsk); \
1943 static __always_inline void scheduler_ipi(void)
1946 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1947 * TIF_NEED_RESCHED remotely (for the first time) will also send
1950 preempt_fold_need_resched();
1953 static inline void scheduler_ipi(void) { }
1956 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1959 * Set thread flags in other task's structures.
1960 * See asm/thread_info.h for TIF_xxxx flags available:
1962 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1964 set_ti_thread_flag(task_thread_info(tsk), flag);
1967 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1969 clear_ti_thread_flag(task_thread_info(tsk), flag);
1972 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1975 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1978 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1980 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1983 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1985 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1988 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1990 return test_ti_thread_flag(task_thread_info(tsk), flag);
1993 static inline void set_tsk_need_resched(struct task_struct *tsk)
1995 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1998 static inline void clear_tsk_need_resched(struct task_struct *tsk)
2000 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2003 static inline int test_tsk_need_resched(struct task_struct *tsk)
2005 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2009 * cond_resched() and cond_resched_lock(): latency reduction via
2010 * explicit rescheduling in places that are safe. The return
2011 * value indicates whether a reschedule was done in fact.
2012 * cond_resched_lock() will drop the spinlock before scheduling,
2014 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
2015 extern int __cond_resched(void);
2017 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
2019 void sched_dynamic_klp_enable(void);
2020 void sched_dynamic_klp_disable(void);
2022 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
2024 static __always_inline int _cond_resched(void)
2026 return static_call_mod(cond_resched)();
2029 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
2031 extern int dynamic_cond_resched(void);
2033 static __always_inline int _cond_resched(void)
2035 return dynamic_cond_resched();
2038 #else /* !CONFIG_PREEMPTION */
2040 static inline int _cond_resched(void)
2042 klp_sched_try_switch();
2043 return __cond_resched();
2046 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2048 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2050 static inline int _cond_resched(void)
2052 klp_sched_try_switch();
2056 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2058 #define cond_resched() ({ \
2059 __might_resched(__FILE__, __LINE__, 0); \
2063 extern int __cond_resched_lock(spinlock_t *lock);
2064 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2065 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2067 #define MIGHT_RESCHED_RCU_SHIFT 8
2068 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2070 #ifndef CONFIG_PREEMPT_RT
2072 * Non RT kernels have an elevated preempt count due to the held lock,
2073 * but are not allowed to be inside a RCU read side critical section
2075 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2078 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2079 * cond_resched*lock() has to take that into account because it checks for
2080 * preempt_count() and rcu_preempt_depth().
2082 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2083 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2086 #define cond_resched_lock(lock) ({ \
2087 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2088 __cond_resched_lock(lock); \
2091 #define cond_resched_rwlock_read(lock) ({ \
2092 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2093 __cond_resched_rwlock_read(lock); \
2096 #define cond_resched_rwlock_write(lock) ({ \
2097 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2098 __cond_resched_rwlock_write(lock); \
2101 static __always_inline bool need_resched(void)
2103 return unlikely(tif_need_resched());
2107 * Wrappers for p->thread_info->cpu access. No-op on UP.
2111 static inline unsigned int task_cpu(const struct task_struct *p)
2113 return READ_ONCE(task_thread_info(p)->cpu);
2116 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2120 static inline unsigned int task_cpu(const struct task_struct *p)
2125 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2129 #endif /* CONFIG_SMP */
2131 extern bool sched_task_on_rq(struct task_struct *p);
2132 extern unsigned long get_wchan(struct task_struct *p);
2133 extern struct task_struct *cpu_curr_snapshot(int cpu);
2135 #include <linux/spinlock.h>
2138 * In order to reduce various lock holder preemption latencies provide an
2139 * interface to see if a vCPU is currently running or not.
2141 * This allows us to terminate optimistic spin loops and block, analogous to
2142 * the native optimistic spin heuristic of testing if the lock owner task is
2145 #ifndef vcpu_is_preempted
2146 static inline bool vcpu_is_preempted(int cpu)
2152 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2153 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2155 #ifndef TASK_SIZE_OF
2156 #define TASK_SIZE_OF(tsk) TASK_SIZE
2160 static inline bool owner_on_cpu(struct task_struct *owner)
2163 * As lock holder preemption issue, we both skip spinning if
2164 * task is not on cpu or its cpu is preempted
2166 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2169 /* Returns effective CPU energy utilization, as seen by the scheduler */
2170 unsigned long sched_cpu_util(int cpu);
2171 #endif /* CONFIG_SMP */
2173 #ifdef CONFIG_SCHED_CORE
2174 extern void sched_core_free(struct task_struct *tsk);
2175 extern void sched_core_fork(struct task_struct *p);
2176 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2177 unsigned long uaddr);
2178 extern int sched_core_idle_cpu(int cpu);
2180 static inline void sched_core_free(struct task_struct *tsk) { }
2181 static inline void sched_core_fork(struct task_struct *p) { }
2182 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2185 extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2187 #ifdef CONFIG_MEM_ALLOC_PROFILING
2188 static __always_inline struct alloc_tag *alloc_tag_save(struct alloc_tag *tag)
2190 swap(current->alloc_tag, tag);
2194 static __always_inline void alloc_tag_restore(struct alloc_tag *tag, struct alloc_tag *old)
2196 #ifdef CONFIG_MEM_ALLOC_PROFILING_DEBUG
2197 WARN(current->alloc_tag != tag, "current->alloc_tag was changed:\n");
2199 current->alloc_tag = old;
2202 #define alloc_tag_save(_tag) NULL
2203 #define alloc_tag_restore(_tag, _old) do {} while (0)