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.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/task_io_accounting.h>
40 #include <linux/posix-timers_types.h>
41 #include <linux/restart_block.h>
42 #include <uapi/linux/rseq.h>
43 #include <linux/seqlock_types.h>
44 #include <linux/kcsan.h>
46 #include <linux/livepatch_sched.h>
47 #include <linux/uidgid_types.h>
48 #include <asm/kmap_size.h>
50 /* task_struct member predeclarations (sorted alphabetically): */
54 struct bpf_local_storage;
56 struct capture_control;
59 struct futex_pi_state;
65 struct perf_event_context;
67 struct pipe_inode_info;
70 struct robust_list_head;
74 struct sched_dl_entity;
76 struct sighand_struct;
78 struct task_delay_info;
84 * Task state bitmask. NOTE! These bits are also
85 * encoded in fs/proc/array.c: get_task_state().
87 * We have two separate sets of flags: task->__state
88 * is about runnability, while task->exit_state are
89 * about the task exiting. Confusing, but this way
90 * modifying one set can't modify the other one by
94 /* Used in tsk->__state: */
95 #define TASK_RUNNING 0x00000000
96 #define TASK_INTERRUPTIBLE 0x00000001
97 #define TASK_UNINTERRUPTIBLE 0x00000002
98 #define __TASK_STOPPED 0x00000004
99 #define __TASK_TRACED 0x00000008
100 /* Used in tsk->exit_state: */
101 #define EXIT_DEAD 0x00000010
102 #define EXIT_ZOMBIE 0x00000020
103 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
104 /* Used in tsk->__state again: */
105 #define TASK_PARKED 0x00000040
106 #define TASK_DEAD 0x00000080
107 #define TASK_WAKEKILL 0x00000100
108 #define TASK_WAKING 0x00000200
109 #define TASK_NOLOAD 0x00000400
110 #define TASK_NEW 0x00000800
111 #define TASK_RTLOCK_WAIT 0x00001000
112 #define TASK_FREEZABLE 0x00002000
113 #define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
114 #define TASK_FROZEN 0x00008000
115 #define TASK_STATE_MAX 0x00010000
117 #define TASK_ANY (TASK_STATE_MAX-1)
120 * DO NOT ADD ANY NEW USERS !
122 #define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
124 /* Convenience macros for the sake of set_current_state: */
125 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
126 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
127 #define TASK_TRACED __TASK_TRACED
129 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
131 /* Convenience macros for the sake of wake_up(): */
132 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
134 /* get_task_state(): */
135 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
136 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
137 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
140 #define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
142 #define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
143 #define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
144 #define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
147 * Special states are those that do not use the normal wait-loop pattern. See
148 * the comment with set_special_state().
150 #define is_special_task_state(state) \
151 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
153 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
154 # define debug_normal_state_change(state_value) \
156 WARN_ON_ONCE(is_special_task_state(state_value)); \
157 current->task_state_change = _THIS_IP_; \
160 # define debug_special_state_change(state_value) \
162 WARN_ON_ONCE(!is_special_task_state(state_value)); \
163 current->task_state_change = _THIS_IP_; \
166 # define debug_rtlock_wait_set_state() \
168 current->saved_state_change = current->task_state_change;\
169 current->task_state_change = _THIS_IP_; \
172 # define debug_rtlock_wait_restore_state() \
174 current->task_state_change = current->saved_state_change;\
178 # define debug_normal_state_change(cond) do { } while (0)
179 # define debug_special_state_change(cond) do { } while (0)
180 # define debug_rtlock_wait_set_state() do { } while (0)
181 # define debug_rtlock_wait_restore_state() do { } while (0)
185 * set_current_state() includes a barrier so that the write of current->__state
186 * is correctly serialised wrt the caller's subsequent test of whether to
190 * set_current_state(TASK_UNINTERRUPTIBLE);
196 * __set_current_state(TASK_RUNNING);
198 * If the caller does not need such serialisation (because, for instance, the
199 * CONDITION test and condition change and wakeup are under the same lock) then
200 * use __set_current_state().
202 * The above is typically ordered against the wakeup, which does:
205 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
207 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
208 * accessing p->__state.
210 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
211 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
212 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
214 * However, with slightly different timing the wakeup TASK_RUNNING store can
215 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
216 * a problem either because that will result in one extra go around the loop
217 * and our @cond test will save the day.
219 * Also see the comments of try_to_wake_up().
221 #define __set_current_state(state_value) \
223 debug_normal_state_change((state_value)); \
224 WRITE_ONCE(current->__state, (state_value)); \
227 #define set_current_state(state_value) \
229 debug_normal_state_change((state_value)); \
230 smp_store_mb(current->__state, (state_value)); \
234 * set_special_state() should be used for those states when the blocking task
235 * can not use the regular condition based wait-loop. In that case we must
236 * serialize against wakeups such that any possible in-flight TASK_RUNNING
237 * stores will not collide with our state change.
239 #define set_special_state(state_value) \
241 unsigned long flags; /* may shadow */ \
243 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
244 debug_special_state_change((state_value)); \
245 WRITE_ONCE(current->__state, (state_value)); \
246 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
250 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
252 * RT's spin/rwlock substitutions are state preserving. The state of the
253 * task when blocking on the lock is saved in task_struct::saved_state and
254 * restored after the lock has been acquired. These operations are
255 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
256 * lock related wakeups while the task is blocked on the lock are
257 * redirected to operate on task_struct::saved_state to ensure that these
258 * are not dropped. On restore task_struct::saved_state is set to
259 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
261 * The lock operation looks like this:
263 * current_save_and_set_rtlock_wait_state();
267 * raw_spin_unlock_irq(&lock->wait_lock);
269 * raw_spin_lock_irq(&lock->wait_lock);
270 * set_current_state(TASK_RTLOCK_WAIT);
272 * current_restore_rtlock_saved_state();
274 #define current_save_and_set_rtlock_wait_state() \
276 lockdep_assert_irqs_disabled(); \
277 raw_spin_lock(¤t->pi_lock); \
278 current->saved_state = current->__state; \
279 debug_rtlock_wait_set_state(); \
280 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
281 raw_spin_unlock(¤t->pi_lock); \
284 #define current_restore_rtlock_saved_state() \
286 lockdep_assert_irqs_disabled(); \
287 raw_spin_lock(¤t->pi_lock); \
288 debug_rtlock_wait_restore_state(); \
289 WRITE_ONCE(current->__state, current->saved_state); \
290 current->saved_state = TASK_RUNNING; \
291 raw_spin_unlock(¤t->pi_lock); \
294 #define get_current_state() READ_ONCE(current->__state)
297 * Define the task command name length as enum, then it can be visible to
304 extern void sched_tick(void);
306 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
308 extern long schedule_timeout(long timeout);
309 extern long schedule_timeout_interruptible(long timeout);
310 extern long schedule_timeout_killable(long timeout);
311 extern long schedule_timeout_uninterruptible(long timeout);
312 extern long schedule_timeout_idle(long timeout);
313 asmlinkage void schedule(void);
314 extern void schedule_preempt_disabled(void);
315 asmlinkage void preempt_schedule_irq(void);
316 #ifdef CONFIG_PREEMPT_RT
317 extern void schedule_rtlock(void);
320 extern int __must_check io_schedule_prepare(void);
321 extern void io_schedule_finish(int token);
322 extern long io_schedule_timeout(long timeout);
323 extern void io_schedule(void);
326 * struct prev_cputime - snapshot of system and user cputime
327 * @utime: time spent in user mode
328 * @stime: time spent in system mode
329 * @lock: protects the above two fields
331 * Stores previous user/system time values such that we can guarantee
334 struct prev_cputime {
335 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
343 /* Task is sleeping or running in a CPU with VTIME inactive: */
347 /* Task runs in kernelspace in a CPU with VTIME active: */
349 /* Task runs in userspace in a CPU with VTIME active: */
351 /* Task runs as guests in a CPU with VTIME active: */
357 unsigned long long starttime;
358 enum vtime_state state;
366 * Utilization clamp constraints.
367 * @UCLAMP_MIN: Minimum utilization
368 * @UCLAMP_MAX: Maximum utilization
369 * @UCLAMP_CNT: Utilization clamp constraints count
378 extern struct root_domain def_root_domain;
379 extern struct mutex sched_domains_mutex;
387 #ifdef CONFIG_SCHED_INFO
388 /* Cumulative counters: */
390 /* # of times we have run on this CPU: */
391 unsigned long pcount;
393 /* Time spent waiting on a runqueue: */
394 unsigned long long run_delay;
398 /* When did we last run on a CPU? */
399 unsigned long long last_arrival;
401 /* When were we last queued to run? */
402 unsigned long long last_queued;
404 #endif /* CONFIG_SCHED_INFO */
408 * Integer metrics need fixed point arithmetic, e.g., sched/fair
409 * has a few: load, load_avg, util_avg, freq, and capacity.
411 * We define a basic fixed point arithmetic range, and then formalize
412 * all these metrics based on that basic range.
414 # define SCHED_FIXEDPOINT_SHIFT 10
415 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
417 /* Increase resolution of cpu_capacity calculations */
418 # define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
419 # define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
422 unsigned long weight;
427 * The load/runnable/util_avg accumulates an infinite geometric series
428 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
430 * [load_avg definition]
432 * load_avg = runnable% * scale_load_down(load)
434 * [runnable_avg definition]
436 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
438 * [util_avg definition]
440 * util_avg = running% * SCHED_CAPACITY_SCALE
442 * where runnable% is the time ratio that a sched_entity is runnable and
443 * running% the time ratio that a sched_entity is running.
445 * For cfs_rq, they are the aggregated values of all runnable and blocked
448 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
449 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
450 * for computing those signals (see update_rq_clock_pelt())
452 * N.B., the above ratios (runnable% and running%) themselves are in the
453 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
454 * to as large a range as necessary. This is for example reflected by
455 * util_avg's SCHED_CAPACITY_SCALE.
459 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
460 * with the highest load (=88761), always runnable on a single cfs_rq,
461 * and should not overflow as the number already hits PID_MAX_LIMIT.
463 * For all other cases (including 32-bit kernels), struct load_weight's
464 * weight will overflow first before we do, because:
466 * Max(load_avg) <= Max(load.weight)
468 * Then it is the load_weight's responsibility to consider overflow
472 u64 last_update_time;
477 unsigned long load_avg;
478 unsigned long runnable_avg;
479 unsigned long util_avg;
480 unsigned int util_est;
481 } ____cacheline_aligned;
484 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
485 * updates. When a task is dequeued, its util_est should not be updated if its
486 * util_avg has not been updated in the meantime.
487 * This information is mapped into the MSB bit of util_est at dequeue time.
488 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
489 * it is safe to use MSB.
491 #define UTIL_EST_WEIGHT_SHIFT 2
492 #define UTIL_AVG_UNCHANGED 0x80000000
494 struct sched_statistics {
495 #ifdef CONFIG_SCHEDSTATS
505 s64 sum_sleep_runtime;
509 s64 sum_block_runtime;
514 u64 nr_migrations_cold;
515 u64 nr_failed_migrations_affine;
516 u64 nr_failed_migrations_running;
517 u64 nr_failed_migrations_hot;
518 u64 nr_forced_migrations;
522 u64 nr_wakeups_migrate;
523 u64 nr_wakeups_local;
524 u64 nr_wakeups_remote;
525 u64 nr_wakeups_affine;
526 u64 nr_wakeups_affine_attempts;
527 u64 nr_wakeups_passive;
530 #ifdef CONFIG_SCHED_CORE
531 u64 core_forceidle_sum;
533 #endif /* CONFIG_SCHEDSTATS */
534 } ____cacheline_aligned;
536 struct sched_entity {
537 /* For load-balancing: */
538 struct load_weight load;
539 struct rb_node run_node;
543 struct list_head group_node;
547 u64 sum_exec_runtime;
548 u64 prev_sum_exec_runtime;
555 #ifdef CONFIG_FAIR_GROUP_SCHED
557 struct sched_entity *parent;
558 /* rq on which this entity is (to be) queued: */
559 struct cfs_rq *cfs_rq;
560 /* rq "owned" by this entity/group: */
562 /* cached value of my_q->h_nr_running */
563 unsigned long runnable_weight;
568 * Per entity load average tracking.
570 * Put into separate cache line so it does not
571 * collide with read-mostly values above.
573 struct sched_avg avg;
577 struct sched_rt_entity {
578 struct list_head run_list;
579 unsigned long timeout;
580 unsigned long watchdog_stamp;
581 unsigned int time_slice;
582 unsigned short on_rq;
583 unsigned short on_list;
585 struct sched_rt_entity *back;
586 #ifdef CONFIG_RT_GROUP_SCHED
587 struct sched_rt_entity *parent;
588 /* rq on which this entity is (to be) queued: */
590 /* rq "owned" by this entity/group: */
593 } __randomize_layout;
595 typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
596 typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
598 struct sched_dl_entity {
599 struct rb_node rb_node;
602 * Original scheduling parameters. Copied here from sched_attr
603 * during sched_setattr(), they will remain the same until
604 * the next sched_setattr().
606 u64 dl_runtime; /* Maximum runtime for each instance */
607 u64 dl_deadline; /* Relative deadline of each instance */
608 u64 dl_period; /* Separation of two instances (period) */
609 u64 dl_bw; /* dl_runtime / dl_period */
610 u64 dl_density; /* dl_runtime / dl_deadline */
613 * Actual scheduling parameters. Initialized with the values above,
614 * they are continuously updated during task execution. Note that
615 * the remaining runtime could be < 0 in case we are in overrun.
617 s64 runtime; /* Remaining runtime for this instance */
618 u64 deadline; /* Absolute deadline for this instance */
619 unsigned int flags; /* Specifying the scheduler behaviour */
624 * @dl_throttled tells if we exhausted the runtime. If so, the
625 * task has to wait for a replenishment to be performed at the
626 * next firing of dl_timer.
628 * @dl_yielded tells if task gave up the CPU before consuming
629 * all its available runtime during the last job.
631 * @dl_non_contending tells if the task is inactive while still
632 * contributing to the active utilization. In other words, it
633 * indicates if the inactive timer has been armed and its handler
634 * has not been executed yet. This flag is useful to avoid race
635 * conditions between the inactive timer handler and the wakeup
638 * @dl_overrun tells if the task asked to be informed about runtime
641 unsigned int dl_throttled : 1;
642 unsigned int dl_yielded : 1;
643 unsigned int dl_non_contending : 1;
644 unsigned int dl_overrun : 1;
645 unsigned int dl_server : 1;
648 * Bandwidth enforcement timer. Each -deadline task has its
649 * own bandwidth to be enforced, thus we need one timer per task.
651 struct hrtimer dl_timer;
654 * Inactive timer, responsible for decreasing the active utilization
655 * at the "0-lag time". When a -deadline task blocks, it contributes
656 * to GRUB's active utilization until the "0-lag time", hence a
657 * timer is needed to decrease the active utilization at the correct
660 struct hrtimer inactive_timer;
663 * Bits for DL-server functionality. Also see the comment near
664 * dl_server_update().
666 * @rq the runqueue this server is for
668 * @server_has_tasks() returns true if @server_pick return a
672 dl_server_has_tasks_f server_has_tasks;
673 dl_server_pick_f server_pick;
675 #ifdef CONFIG_RT_MUTEXES
677 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
678 * pi_se points to the donor, otherwise points to the dl_se it belongs
679 * to (the original one/itself).
681 struct sched_dl_entity *pi_se;
685 #ifdef CONFIG_UCLAMP_TASK
686 /* Number of utilization clamp buckets (shorter alias) */
687 #define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
690 * Utilization clamp for a scheduling entity
691 * @value: clamp value "assigned" to a se
692 * @bucket_id: bucket index corresponding to the "assigned" value
693 * @active: the se is currently refcounted in a rq's bucket
694 * @user_defined: the requested clamp value comes from user-space
696 * The bucket_id is the index of the clamp bucket matching the clamp value
697 * which is pre-computed and stored to avoid expensive integer divisions from
700 * The active bit is set whenever a task has got an "effective" value assigned,
701 * which can be different from the clamp value "requested" from user-space.
702 * This allows to know a task is refcounted in the rq's bucket corresponding
703 * to the "effective" bucket_id.
705 * The user_defined bit is set whenever a task has got a task-specific clamp
706 * value requested from userspace, i.e. the system defaults apply to this task
707 * just as a restriction. This allows to relax default clamps when a less
708 * restrictive task-specific value has been requested, thus allowing to
709 * implement a "nice" semantic. For example, a task running with a 20%
710 * default boost can still drop its own boosting to 0%.
713 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
714 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
715 unsigned int active : 1;
716 unsigned int user_defined : 1;
718 #endif /* CONFIG_UCLAMP_TASK */
724 u8 exp_hint; /* Hint for performance. */
725 u8 need_mb; /* Readers need smp_mb(). */
727 u32 s; /* Set of bits. */
730 enum perf_event_task_context {
731 perf_invalid_context = -1,
734 perf_nr_task_contexts,
738 struct wake_q_node *next;
742 #ifdef CONFIG_KMAP_LOCAL
744 pte_t pteval[KM_MAX_IDX];
749 #ifdef CONFIG_THREAD_INFO_IN_TASK
751 * For reasons of header soup (see current_thread_info()), this
752 * must be the first element of task_struct.
754 struct thread_info thread_info;
756 unsigned int __state;
758 /* saved state for "spinlock sleepers" */
759 unsigned int saved_state;
762 * This begins the randomizable portion of task_struct. Only
763 * scheduling-critical items should be added above here.
765 randomized_struct_fields_start
769 /* Per task flags (PF_*), defined further below: */
775 struct __call_single_node wake_entry;
776 unsigned int wakee_flips;
777 unsigned long wakee_flip_decay_ts;
778 struct task_struct *last_wakee;
781 * recent_used_cpu is initially set as the last CPU used by a task
782 * that wakes affine another task. Waker/wakee relationships can
783 * push tasks around a CPU where each wakeup moves to the next one.
784 * Tracking a recently used CPU allows a quick search for a recently
785 * used CPU that may be idle.
795 unsigned int rt_priority;
797 struct sched_entity se;
798 struct sched_rt_entity rt;
799 struct sched_dl_entity dl;
800 struct sched_dl_entity *dl_server;
801 const struct sched_class *sched_class;
803 #ifdef CONFIG_SCHED_CORE
804 struct rb_node core_node;
805 unsigned long core_cookie;
806 unsigned int core_occupation;
809 #ifdef CONFIG_CGROUP_SCHED
810 struct task_group *sched_task_group;
813 #ifdef CONFIG_UCLAMP_TASK
815 * Clamp values requested for a scheduling entity.
816 * Must be updated with task_rq_lock() held.
818 struct uclamp_se uclamp_req[UCLAMP_CNT];
820 * Effective clamp values used for a scheduling entity.
821 * Must be updated with task_rq_lock() held.
823 struct uclamp_se uclamp[UCLAMP_CNT];
826 struct sched_statistics stats;
828 #ifdef CONFIG_PREEMPT_NOTIFIERS
829 /* List of struct preempt_notifier: */
830 struct hlist_head preempt_notifiers;
833 #ifdef CONFIG_BLK_DEV_IO_TRACE
834 unsigned int btrace_seq;
838 unsigned long max_allowed_capacity;
840 const cpumask_t *cpus_ptr;
841 cpumask_t *user_cpus_ptr;
843 void *migration_pending;
845 unsigned short migration_disabled;
847 unsigned short migration_flags;
849 #ifdef CONFIG_PREEMPT_RCU
850 int rcu_read_lock_nesting;
851 union rcu_special rcu_read_unlock_special;
852 struct list_head rcu_node_entry;
853 struct rcu_node *rcu_blocked_node;
854 #endif /* #ifdef CONFIG_PREEMPT_RCU */
856 #ifdef CONFIG_TASKS_RCU
857 unsigned long rcu_tasks_nvcsw;
858 u8 rcu_tasks_holdout;
860 int rcu_tasks_idle_cpu;
861 struct list_head rcu_tasks_holdout_list;
862 int rcu_tasks_exit_cpu;
863 struct list_head rcu_tasks_exit_list;
864 #endif /* #ifdef CONFIG_TASKS_RCU */
866 #ifdef CONFIG_TASKS_TRACE_RCU
867 int trc_reader_nesting;
869 union rcu_special trc_reader_special;
870 struct list_head trc_holdout_list;
871 struct list_head trc_blkd_node;
873 #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
875 struct sched_info sched_info;
877 struct list_head tasks;
879 struct plist_node pushable_tasks;
880 struct rb_node pushable_dl_tasks;
883 struct mm_struct *mm;
884 struct mm_struct *active_mm;
885 struct address_space *faults_disabled_mapping;
890 /* The signal sent when the parent dies: */
892 /* JOBCTL_*, siglock protected: */
893 unsigned long jobctl;
895 /* Used for emulating ABI behavior of previous Linux versions: */
896 unsigned int personality;
898 /* Scheduler bits, serialized by scheduler locks: */
899 unsigned sched_reset_on_fork:1;
900 unsigned sched_contributes_to_load:1;
901 unsigned sched_migrated:1;
903 /* Force alignment to the next boundary: */
906 /* Unserialized, strictly 'current' */
909 * This field must not be in the scheduler word above due to wakelist
910 * queueing no longer being serialized by p->on_cpu. However:
913 * schedule() if (p->on_rq && ..) // false
914 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
915 * deactivate_task() ttwu_queue_wakelist())
916 * p->on_rq = 0; p->sched_remote_wakeup = Y;
918 * guarantees all stores of 'current' are visible before
919 * ->sched_remote_wakeup gets used, so it can be in this word.
921 unsigned sched_remote_wakeup:1;
922 #ifdef CONFIG_RT_MUTEXES
923 unsigned sched_rt_mutex:1;
926 /* Bit to tell TOMOYO we're in execve(): */
927 unsigned in_execve:1;
928 unsigned in_iowait:1;
929 #ifndef TIF_RESTORE_SIGMASK
930 unsigned restore_sigmask:1;
933 unsigned in_user_fault:1;
935 #ifdef CONFIG_LRU_GEN
936 /* whether the LRU algorithm may apply to this access */
937 unsigned in_lru_fault:1;
939 #ifdef CONFIG_COMPAT_BRK
940 unsigned brk_randomized:1;
942 #ifdef CONFIG_CGROUPS
943 /* disallow userland-initiated cgroup migration */
944 unsigned no_cgroup_migration:1;
945 /* task is frozen/stopped (used by the cgroup freezer) */
948 #ifdef CONFIG_BLK_CGROUP
949 unsigned use_memdelay:1;
952 /* Stalled due to lack of memory */
953 unsigned in_memstall:1;
955 #ifdef CONFIG_PAGE_OWNER
956 /* Used by page_owner=on to detect recursion in page tracking. */
957 unsigned in_page_owner:1;
959 #ifdef CONFIG_EVENTFD
960 /* Recursion prevention for eventfd_signal() */
961 unsigned in_eventfd:1;
963 #ifdef CONFIG_ARCH_HAS_CPU_PASID
964 unsigned pasid_activated:1;
966 #ifdef CONFIG_CPU_SUP_INTEL
967 unsigned reported_split_lock:1;
969 #ifdef CONFIG_TASK_DELAY_ACCT
970 /* delay due to memory thrashing */
971 unsigned in_thrashing:1;
974 unsigned long atomic_flags; /* Flags requiring atomic access. */
976 struct restart_block restart_block;
981 #ifdef CONFIG_STACKPROTECTOR
982 /* Canary value for the -fstack-protector GCC feature: */
983 unsigned long stack_canary;
986 * Pointers to the (original) parent process, youngest child, younger sibling,
987 * older sibling, respectively. (p->father can be replaced with
988 * p->real_parent->pid)
991 /* Real parent process: */
992 struct task_struct __rcu *real_parent;
994 /* Recipient of SIGCHLD, wait4() reports: */
995 struct task_struct __rcu *parent;
998 * Children/sibling form the list of natural children:
1000 struct list_head children;
1001 struct list_head sibling;
1002 struct task_struct *group_leader;
1005 * 'ptraced' is the list of tasks this task is using ptrace() on.
1007 * This includes both natural children and PTRACE_ATTACH targets.
1008 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1010 struct list_head ptraced;
1011 struct list_head ptrace_entry;
1013 /* PID/PID hash table linkage. */
1014 struct pid *thread_pid;
1015 struct hlist_node pid_links[PIDTYPE_MAX];
1016 struct list_head thread_node;
1018 struct completion *vfork_done;
1020 /* CLONE_CHILD_SETTID: */
1021 int __user *set_child_tid;
1023 /* CLONE_CHILD_CLEARTID: */
1024 int __user *clear_child_tid;
1026 /* PF_KTHREAD | PF_IO_WORKER */
1027 void *worker_private;
1031 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1036 struct prev_cputime prev_cputime;
1037 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1041 #ifdef CONFIG_NO_HZ_FULL
1042 atomic_t tick_dep_mask;
1044 /* Context switch counts: */
1045 unsigned long nvcsw;
1046 unsigned long nivcsw;
1048 /* Monotonic time in nsecs: */
1051 /* Boot based time in nsecs: */
1054 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1055 unsigned long min_flt;
1056 unsigned long maj_flt;
1058 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1059 struct posix_cputimers posix_cputimers;
1061 #ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1062 struct posix_cputimers_work posix_cputimers_work;
1065 /* Process credentials: */
1067 /* Tracer's credentials at attach: */
1068 const struct cred __rcu *ptracer_cred;
1070 /* Objective and real subjective task credentials (COW): */
1071 const struct cred __rcu *real_cred;
1073 /* Effective (overridable) subjective task credentials (COW): */
1074 const struct cred __rcu *cred;
1077 /* Cached requested key. */
1078 struct key *cached_requested_key;
1082 * executable name, excluding path.
1084 * - normally initialized setup_new_exec()
1085 * - access it with [gs]et_task_comm()
1086 * - lock it with task_lock()
1088 char comm[TASK_COMM_LEN];
1090 struct nameidata *nameidata;
1092 #ifdef CONFIG_SYSVIPC
1093 struct sysv_sem sysvsem;
1094 struct sysv_shm sysvshm;
1096 #ifdef CONFIG_DETECT_HUNG_TASK
1097 unsigned long last_switch_count;
1098 unsigned long last_switch_time;
1100 /* Filesystem information: */
1101 struct fs_struct *fs;
1103 /* Open file information: */
1104 struct files_struct *files;
1106 #ifdef CONFIG_IO_URING
1107 struct io_uring_task *io_uring;
1111 struct nsproxy *nsproxy;
1113 /* Signal handlers: */
1114 struct signal_struct *signal;
1115 struct sighand_struct __rcu *sighand;
1117 sigset_t real_blocked;
1118 /* Restored if set_restore_sigmask() was used: */
1119 sigset_t saved_sigmask;
1120 struct sigpending pending;
1121 unsigned long sas_ss_sp;
1123 unsigned int sas_ss_flags;
1125 struct callback_head *task_works;
1128 #ifdef CONFIG_AUDITSYSCALL
1129 struct audit_context *audit_context;
1132 unsigned int sessionid;
1134 struct seccomp seccomp;
1135 struct syscall_user_dispatch syscall_dispatch;
1137 /* Thread group tracking: */
1141 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1142 spinlock_t alloc_lock;
1144 /* Protection of the PI data structures: */
1145 raw_spinlock_t pi_lock;
1147 struct wake_q_node wake_q;
1149 #ifdef CONFIG_RT_MUTEXES
1150 /* PI waiters blocked on a rt_mutex held by this task: */
1151 struct rb_root_cached pi_waiters;
1152 /* Updated under owner's pi_lock and rq lock */
1153 struct task_struct *pi_top_task;
1154 /* Deadlock detection and priority inheritance handling: */
1155 struct rt_mutex_waiter *pi_blocked_on;
1158 #ifdef CONFIG_DEBUG_MUTEXES
1159 /* Mutex deadlock detection: */
1160 struct mutex_waiter *blocked_on;
1163 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1164 int non_block_count;
1167 #ifdef CONFIG_TRACE_IRQFLAGS
1168 struct irqtrace_events irqtrace;
1169 unsigned int hardirq_threaded;
1170 u64 hardirq_chain_key;
1171 int softirqs_enabled;
1172 int softirq_context;
1175 #ifdef CONFIG_PREEMPT_RT
1176 int softirq_disable_cnt;
1179 #ifdef CONFIG_LOCKDEP
1180 # define MAX_LOCK_DEPTH 48UL
1183 unsigned int lockdep_recursion;
1184 struct held_lock held_locks[MAX_LOCK_DEPTH];
1187 #if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1188 unsigned int in_ubsan;
1191 /* Journalling filesystem info: */
1194 /* Stacked block device info: */
1195 struct bio_list *bio_list;
1197 /* Stack plugging: */
1198 struct blk_plug *plug;
1201 struct reclaim_state *reclaim_state;
1203 struct io_context *io_context;
1205 #ifdef CONFIG_COMPACTION
1206 struct capture_control *capture_control;
1209 unsigned long ptrace_message;
1210 kernel_siginfo_t *last_siginfo;
1212 struct task_io_accounting ioac;
1214 /* Pressure stall state */
1215 unsigned int psi_flags;
1217 #ifdef CONFIG_TASK_XACCT
1218 /* Accumulated RSS usage: */
1220 /* Accumulated virtual memory usage: */
1222 /* stime + utime since last update: */
1225 #ifdef CONFIG_CPUSETS
1226 /* Protected by ->alloc_lock: */
1227 nodemask_t mems_allowed;
1228 /* Sequence number to catch updates: */
1229 seqcount_spinlock_t mems_allowed_seq;
1230 int cpuset_mem_spread_rotor;
1231 int cpuset_slab_spread_rotor;
1233 #ifdef CONFIG_CGROUPS
1234 /* Control Group info protected by css_set_lock: */
1235 struct css_set __rcu *cgroups;
1236 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1237 struct list_head cg_list;
1239 #ifdef CONFIG_X86_CPU_RESCTRL
1244 struct robust_list_head __user *robust_list;
1245 #ifdef CONFIG_COMPAT
1246 struct compat_robust_list_head __user *compat_robust_list;
1248 struct list_head pi_state_list;
1249 struct futex_pi_state *pi_state_cache;
1250 struct mutex futex_exit_mutex;
1251 unsigned int futex_state;
1253 #ifdef CONFIG_PERF_EVENTS
1254 struct perf_event_context *perf_event_ctxp;
1255 struct mutex perf_event_mutex;
1256 struct list_head perf_event_list;
1258 #ifdef CONFIG_DEBUG_PREEMPT
1259 unsigned long preempt_disable_ip;
1262 /* Protected by alloc_lock: */
1263 struct mempolicy *mempolicy;
1266 short pref_node_fork;
1268 #ifdef CONFIG_NUMA_BALANCING
1270 unsigned int numa_scan_period;
1271 unsigned int numa_scan_period_max;
1272 int numa_preferred_nid;
1273 unsigned long numa_migrate_retry;
1274 /* Migration stamp: */
1276 u64 last_task_numa_placement;
1277 u64 last_sum_exec_runtime;
1278 struct callback_head numa_work;
1281 * This pointer is only modified for current in syscall and
1282 * pagefault context (and for tasks being destroyed), so it can be read
1283 * from any of the following contexts:
1284 * - RCU read-side critical section
1285 * - current->numa_group from everywhere
1286 * - task's runqueue locked, task not running
1288 struct numa_group __rcu *numa_group;
1291 * numa_faults is an array split into four regions:
1292 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1293 * in this precise order.
1295 * faults_memory: Exponential decaying average of faults on a per-node
1296 * basis. Scheduling placement decisions are made based on these
1297 * counts. The values remain static for the duration of a PTE scan.
1298 * faults_cpu: Track the nodes the process was running on when a NUMA
1299 * hinting fault was incurred.
1300 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1301 * during the current scan window. When the scan completes, the counts
1302 * in faults_memory and faults_cpu decay and these values are copied.
1304 unsigned long *numa_faults;
1305 unsigned long total_numa_faults;
1308 * numa_faults_locality tracks if faults recorded during the last
1309 * scan window were remote/local or failed to migrate. The task scan
1310 * period is adapted based on the locality of the faults with different
1311 * weights depending on whether they were shared or private faults
1313 unsigned long numa_faults_locality[3];
1315 unsigned long numa_pages_migrated;
1316 #endif /* CONFIG_NUMA_BALANCING */
1319 struct rseq __user *rseq;
1323 * RmW on rseq_event_mask must be performed atomically
1324 * with respect to preemption.
1326 unsigned long rseq_event_mask;
1329 #ifdef CONFIG_SCHED_MM_CID
1330 int mm_cid; /* Current cid in mm */
1331 int last_mm_cid; /* Most recent cid in mm */
1332 int migrate_from_cpu;
1333 int mm_cid_active; /* Whether cid bitmap is active */
1334 struct callback_head cid_work;
1337 struct tlbflush_unmap_batch tlb_ubc;
1339 /* Cache last used pipe for splice(): */
1340 struct pipe_inode_info *splice_pipe;
1342 struct page_frag task_frag;
1344 #ifdef CONFIG_TASK_DELAY_ACCT
1345 struct task_delay_info *delays;
1348 #ifdef CONFIG_FAULT_INJECTION
1350 unsigned int fail_nth;
1353 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1354 * balance_dirty_pages() for a dirty throttling pause:
1357 int nr_dirtied_pause;
1358 /* Start of a write-and-pause period: */
1359 unsigned long dirty_paused_when;
1361 #ifdef CONFIG_LATENCYTOP
1362 int latency_record_count;
1363 struct latency_record latency_record[LT_SAVECOUNT];
1366 * Time slack values; these are used to round up poll() and
1367 * select() etc timeout values. These are in nanoseconds.
1370 u64 default_timer_slack_ns;
1372 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1373 unsigned int kasan_depth;
1377 struct kcsan_ctx kcsan_ctx;
1378 #ifdef CONFIG_TRACE_IRQFLAGS
1379 struct irqtrace_events kcsan_save_irqtrace;
1381 #ifdef CONFIG_KCSAN_WEAK_MEMORY
1382 int kcsan_stack_depth;
1387 struct kmsan_ctx kmsan_ctx;
1390 #if IS_ENABLED(CONFIG_KUNIT)
1391 struct kunit *kunit_test;
1394 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1395 /* Index of current stored address in ret_stack: */
1399 /* Stack of return addresses for return function tracing: */
1400 struct ftrace_ret_stack *ret_stack;
1402 /* Timestamp for last schedule: */
1403 unsigned long long ftrace_timestamp;
1406 * Number of functions that haven't been traced
1407 * because of depth overrun:
1409 atomic_t trace_overrun;
1411 /* Pause tracing: */
1412 atomic_t tracing_graph_pause;
1415 #ifdef CONFIG_TRACING
1416 /* Bitmask and counter of trace recursion: */
1417 unsigned long trace_recursion;
1418 #endif /* CONFIG_TRACING */
1421 /* See kernel/kcov.c for more details. */
1423 /* Coverage collection mode enabled for this task (0 if disabled): */
1424 unsigned int kcov_mode;
1426 /* Size of the kcov_area: */
1427 unsigned int kcov_size;
1429 /* Buffer for coverage collection: */
1432 /* KCOV descriptor wired with this task or NULL: */
1435 /* KCOV common handle for remote coverage collection: */
1438 /* KCOV sequence number: */
1441 /* Collect coverage from softirq context: */
1442 unsigned int kcov_softirq;
1446 struct mem_cgroup *memcg_in_oom;
1447 gfp_t memcg_oom_gfp_mask;
1448 int memcg_oom_order;
1450 /* Number of pages to reclaim on returning to userland: */
1451 unsigned int memcg_nr_pages_over_high;
1453 /* Used by memcontrol for targeted memcg charge: */
1454 struct mem_cgroup *active_memcg;
1457 #ifdef CONFIG_MEMCG_KMEM
1458 struct obj_cgroup *objcg;
1461 #ifdef CONFIG_BLK_CGROUP
1462 struct gendisk *throttle_disk;
1465 #ifdef CONFIG_UPROBES
1466 struct uprobe_task *utask;
1468 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1469 unsigned int sequential_io;
1470 unsigned int sequential_io_avg;
1472 struct kmap_ctrl kmap_ctrl;
1473 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1474 unsigned long task_state_change;
1475 # ifdef CONFIG_PREEMPT_RT
1476 unsigned long saved_state_change;
1479 struct rcu_head rcu;
1480 refcount_t rcu_users;
1481 int pagefault_disabled;
1483 struct task_struct *oom_reaper_list;
1484 struct timer_list oom_reaper_timer;
1486 #ifdef CONFIG_VMAP_STACK
1487 struct vm_struct *stack_vm_area;
1489 #ifdef CONFIG_THREAD_INFO_IN_TASK
1490 /* A live task holds one reference: */
1491 refcount_t stack_refcount;
1493 #ifdef CONFIG_LIVEPATCH
1496 #ifdef CONFIG_SECURITY
1497 /* Used by LSM modules for access restriction: */
1500 #ifdef CONFIG_BPF_SYSCALL
1501 /* Used by BPF task local storage */
1502 struct bpf_local_storage __rcu *bpf_storage;
1503 /* Used for BPF run context */
1504 struct bpf_run_ctx *bpf_ctx;
1507 #ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1508 unsigned long lowest_stack;
1509 unsigned long prev_lowest_stack;
1512 #ifdef CONFIG_X86_MCE
1513 void __user *mce_vaddr;
1518 __mce_reserved : 62;
1519 struct callback_head mce_kill_me;
1523 #ifdef CONFIG_KRETPROBES
1524 struct llist_head kretprobe_instances;
1526 #ifdef CONFIG_RETHOOK
1527 struct llist_head rethooks;
1530 #ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1532 * If L1D flush is supported on mm context switch
1533 * then we use this callback head to queue kill work
1534 * to kill tasks that are not running on SMT disabled
1537 struct callback_head l1d_flush_kill;
1542 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1543 * If we find justification for more monitors, we can think
1544 * about adding more or developing a dynamic method. So far,
1545 * none of these are justified.
1547 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1550 #ifdef CONFIG_USER_EVENTS
1551 struct user_event_mm *user_event_mm;
1555 * New fields for task_struct should be added above here, so that
1556 * they are included in the randomized portion of task_struct.
1558 randomized_struct_fields_end
1560 /* CPU-specific state of this task: */
1561 struct thread_struct thread;
1564 * WARNING: on x86, 'thread_struct' contains a variable-sized
1565 * structure. It *MUST* be at the end of 'task_struct'.
1567 * Do not put anything below here!
1571 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1572 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1574 static inline unsigned int __task_state_index(unsigned int tsk_state,
1575 unsigned int tsk_exit_state)
1577 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1579 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1581 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1582 state = TASK_REPORT_IDLE;
1585 * We're lying here, but rather than expose a completely new task state
1586 * to userspace, we can make this appear as if the task has gone through
1587 * a regular rt_mutex_lock() call.
1589 if (tsk_state & TASK_RTLOCK_WAIT)
1590 state = TASK_UNINTERRUPTIBLE;
1595 static inline unsigned int task_state_index(struct task_struct *tsk)
1597 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1600 static inline char task_index_to_char(unsigned int state)
1602 static const char state_char[] = "RSDTtXZPI";
1604 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1606 return state_char[state];
1609 static inline char task_state_to_char(struct task_struct *tsk)
1611 return task_index_to_char(task_state_index(tsk));
1614 extern struct pid *cad_pid;
1619 #define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1620 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1621 #define PF_EXITING 0x00000004 /* Getting shut down */
1622 #define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1623 #define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1624 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1625 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1626 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1627 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1628 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1629 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1630 #define PF_MEMALLOC 0x00000800 /* Allocating memory to free memory. See memalloc_noreclaim_save() */
1631 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1632 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1633 #define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1634 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1635 #define PF__HOLE__00010000 0x00010000
1636 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1637 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocations inherit GFP_NOFS. See memalloc_nfs_save() */
1638 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocations inherit GFP_NOIO. See memalloc_noio_save() */
1639 #define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1640 * I am cleaning dirty pages from some other bdi. */
1641 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1642 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1643 #define PF_MEMALLOC_NORECLAIM 0x00800000 /* All allocation requests will clear __GFP_DIRECT_RECLAIM */
1644 #define PF_MEMALLOC_NOWARN 0x01000000 /* All allocation requests will inherit __GFP_NOWARN */
1645 #define PF__HOLE__02000000 0x02000000
1646 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1647 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1648 #define PF_MEMALLOC_PIN 0x10000000 /* Allocations constrained to zones which allow long term pinning.
1649 * See memalloc_pin_save() */
1650 #define PF_BLOCK_TS 0x20000000 /* plug has ts that needs updating */
1651 #define PF__HOLE__40000000 0x40000000
1652 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1655 * Only the _current_ task can read/write to tsk->flags, but other
1656 * tasks can access tsk->flags in readonly mode for example
1657 * with tsk_used_math (like during threaded core dumping).
1658 * There is however an exception to this rule during ptrace
1659 * or during fork: the ptracer task is allowed to write to the
1660 * child->flags of its traced child (same goes for fork, the parent
1661 * can write to the child->flags), because we're guaranteed the
1662 * child is not running and in turn not changing child->flags
1663 * at the same time the parent does it.
1665 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1666 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1667 #define clear_used_math() clear_stopped_child_used_math(current)
1668 #define set_used_math() set_stopped_child_used_math(current)
1670 #define conditional_stopped_child_used_math(condition, child) \
1671 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1673 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1675 #define copy_to_stopped_child_used_math(child) \
1676 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1678 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1679 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1680 #define used_math() tsk_used_math(current)
1682 static __always_inline bool is_percpu_thread(void)
1685 return (current->flags & PF_NO_SETAFFINITY) &&
1686 (current->nr_cpus_allowed == 1);
1692 /* Per-process atomic flags. */
1693 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1694 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1695 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1696 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1697 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1698 #define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1699 #define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1700 #define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1702 #define TASK_PFA_TEST(name, func) \
1703 static inline bool task_##func(struct task_struct *p) \
1704 { return test_bit(PFA_##name, &p->atomic_flags); }
1706 #define TASK_PFA_SET(name, func) \
1707 static inline void task_set_##func(struct task_struct *p) \
1708 { set_bit(PFA_##name, &p->atomic_flags); }
1710 #define TASK_PFA_CLEAR(name, func) \
1711 static inline void task_clear_##func(struct task_struct *p) \
1712 { clear_bit(PFA_##name, &p->atomic_flags); }
1714 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1715 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1717 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1718 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1719 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1721 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1722 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1723 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1725 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1726 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1727 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1729 TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1730 TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1731 TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1733 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1734 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1736 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1737 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1738 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1740 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1741 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1744 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1746 current->flags &= ~flags;
1747 current->flags |= orig_flags & flags;
1750 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1751 extern int task_can_attach(struct task_struct *p);
1752 extern int dl_bw_alloc(int cpu, u64 dl_bw);
1753 extern void dl_bw_free(int cpu, u64 dl_bw);
1756 /* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1757 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1760 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1762 * @new_mask: CPU affinity mask
1764 * Return: zero if successful, or a negative error code
1766 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1767 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1768 extern void release_user_cpus_ptr(struct task_struct *p);
1769 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1770 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1771 extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1773 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1776 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1778 if (!cpumask_test_cpu(0, new_mask))
1782 static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1784 if (src->user_cpus_ptr)
1788 static inline void release_user_cpus_ptr(struct task_struct *p)
1790 WARN_ON(p->user_cpus_ptr);
1793 static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1799 extern int yield_to(struct task_struct *p, bool preempt);
1800 extern void set_user_nice(struct task_struct *p, long nice);
1801 extern int task_prio(const struct task_struct *p);
1804 * task_nice - return the nice value of a given task.
1805 * @p: the task in question.
1807 * Return: The nice value [ -20 ... 0 ... 19 ].
1809 static inline int task_nice(const struct task_struct *p)
1811 return PRIO_TO_NICE((p)->static_prio);
1814 extern int can_nice(const struct task_struct *p, const int nice);
1815 extern int task_curr(const struct task_struct *p);
1816 extern int idle_cpu(int cpu);
1817 extern int available_idle_cpu(int cpu);
1818 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1819 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1820 extern void sched_set_fifo(struct task_struct *p);
1821 extern void sched_set_fifo_low(struct task_struct *p);
1822 extern void sched_set_normal(struct task_struct *p, int nice);
1823 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1824 extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1825 extern struct task_struct *idle_task(int cpu);
1828 * is_idle_task - is the specified task an idle task?
1829 * @p: the task in question.
1831 * Return: 1 if @p is an idle task. 0 otherwise.
1833 static __always_inline bool is_idle_task(const struct task_struct *p)
1835 return !!(p->flags & PF_IDLE);
1838 extern struct task_struct *curr_task(int cpu);
1839 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1843 union thread_union {
1844 struct task_struct task;
1845 #ifndef CONFIG_THREAD_INFO_IN_TASK
1846 struct thread_info thread_info;
1848 unsigned long stack[THREAD_SIZE/sizeof(long)];
1851 #ifndef CONFIG_THREAD_INFO_IN_TASK
1852 extern struct thread_info init_thread_info;
1855 extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1857 #ifdef CONFIG_THREAD_INFO_IN_TASK
1858 # define task_thread_info(task) (&(task)->thread_info)
1859 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1860 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1864 * find a task by one of its numerical ids
1866 * find_task_by_pid_ns():
1867 * finds a task by its pid in the specified namespace
1868 * find_task_by_vpid():
1869 * finds a task by its virtual pid
1871 * see also find_vpid() etc in include/linux/pid.h
1874 extern struct task_struct *find_task_by_vpid(pid_t nr);
1875 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1878 * find a task by its virtual pid and get the task struct
1880 extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1882 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1883 extern int wake_up_process(struct task_struct *tsk);
1884 extern void wake_up_new_task(struct task_struct *tsk);
1887 extern void kick_process(struct task_struct *tsk);
1889 static inline void kick_process(struct task_struct *tsk) { }
1892 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1894 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1896 __set_task_comm(tsk, from, false);
1899 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1900 #define get_task_comm(buf, tsk) ({ \
1901 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1902 __get_task_comm(buf, sizeof(buf), tsk); \
1906 static __always_inline void scheduler_ipi(void)
1909 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1910 * TIF_NEED_RESCHED remotely (for the first time) will also send
1913 preempt_fold_need_resched();
1916 static inline void scheduler_ipi(void) { }
1919 extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1922 * Set thread flags in other task's structures.
1923 * See asm/thread_info.h for TIF_xxxx flags available:
1925 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1927 set_ti_thread_flag(task_thread_info(tsk), flag);
1930 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1932 clear_ti_thread_flag(task_thread_info(tsk), flag);
1935 static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1938 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1941 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1943 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1946 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1948 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1951 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1953 return test_ti_thread_flag(task_thread_info(tsk), flag);
1956 static inline void set_tsk_need_resched(struct task_struct *tsk)
1958 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1961 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1963 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1966 static inline int test_tsk_need_resched(struct task_struct *tsk)
1968 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1972 * cond_resched() and cond_resched_lock(): latency reduction via
1973 * explicit rescheduling in places that are safe. The return
1974 * value indicates whether a reschedule was done in fact.
1975 * cond_resched_lock() will drop the spinlock before scheduling,
1977 #if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1978 extern int __cond_resched(void);
1980 #if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
1982 void sched_dynamic_klp_enable(void);
1983 void sched_dynamic_klp_disable(void);
1985 DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1987 static __always_inline int _cond_resched(void)
1989 return static_call_mod(cond_resched)();
1992 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
1994 extern int dynamic_cond_resched(void);
1996 static __always_inline int _cond_resched(void)
1998 return dynamic_cond_resched();
2001 #else /* !CONFIG_PREEMPTION */
2003 static inline int _cond_resched(void)
2005 klp_sched_try_switch();
2006 return __cond_resched();
2009 #endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2011 #else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2013 static inline int _cond_resched(void)
2015 klp_sched_try_switch();
2019 #endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2021 #define cond_resched() ({ \
2022 __might_resched(__FILE__, __LINE__, 0); \
2026 extern int __cond_resched_lock(spinlock_t *lock);
2027 extern int __cond_resched_rwlock_read(rwlock_t *lock);
2028 extern int __cond_resched_rwlock_write(rwlock_t *lock);
2030 #define MIGHT_RESCHED_RCU_SHIFT 8
2031 #define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2033 #ifndef CONFIG_PREEMPT_RT
2035 * Non RT kernels have an elevated preempt count due to the held lock,
2036 * but are not allowed to be inside a RCU read side critical section
2038 # define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2041 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2042 * cond_resched*lock() has to take that into account because it checks for
2043 * preempt_count() and rcu_preempt_depth().
2045 # define PREEMPT_LOCK_RESCHED_OFFSETS \
2046 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2049 #define cond_resched_lock(lock) ({ \
2050 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2051 __cond_resched_lock(lock); \
2054 #define cond_resched_rwlock_read(lock) ({ \
2055 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2056 __cond_resched_rwlock_read(lock); \
2059 #define cond_resched_rwlock_write(lock) ({ \
2060 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2061 __cond_resched_rwlock_write(lock); \
2064 #ifdef CONFIG_PREEMPT_DYNAMIC
2066 extern bool preempt_model_none(void);
2067 extern bool preempt_model_voluntary(void);
2068 extern bool preempt_model_full(void);
2072 static inline bool preempt_model_none(void)
2074 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2076 static inline bool preempt_model_voluntary(void)
2078 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2080 static inline bool preempt_model_full(void)
2082 return IS_ENABLED(CONFIG_PREEMPT);
2087 static inline bool preempt_model_rt(void)
2089 return IS_ENABLED(CONFIG_PREEMPT_RT);
2093 * Does the preemption model allow non-cooperative preemption?
2095 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2096 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2097 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2098 * PREEMPT_NONE model.
2100 static inline bool preempt_model_preemptible(void)
2102 return preempt_model_full() || preempt_model_rt();
2105 static __always_inline bool need_resched(void)
2107 return unlikely(tif_need_resched());
2111 * Wrappers for p->thread_info->cpu access. No-op on UP.
2115 static inline unsigned int task_cpu(const struct task_struct *p)
2117 return READ_ONCE(task_thread_info(p)->cpu);
2120 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2124 static inline unsigned int task_cpu(const struct task_struct *p)
2129 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2133 #endif /* CONFIG_SMP */
2135 extern bool sched_task_on_rq(struct task_struct *p);
2136 extern unsigned long get_wchan(struct task_struct *p);
2137 extern struct task_struct *cpu_curr_snapshot(int cpu);
2139 #include <linux/spinlock.h>
2142 * In order to reduce various lock holder preemption latencies provide an
2143 * interface to see if a vCPU is currently running or not.
2145 * This allows us to terminate optimistic spin loops and block, analogous to
2146 * the native optimistic spin heuristic of testing if the lock owner task is
2149 #ifndef vcpu_is_preempted
2150 static inline bool vcpu_is_preempted(int cpu)
2156 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2157 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2159 #ifndef TASK_SIZE_OF
2160 #define TASK_SIZE_OF(tsk) TASK_SIZE
2164 static inline bool owner_on_cpu(struct task_struct *owner)
2167 * As lock holder preemption issue, we both skip spinning if
2168 * task is not on cpu or its cpu is preempted
2170 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2173 /* Returns effective CPU energy utilization, as seen by the scheduler */
2174 unsigned long sched_cpu_util(int cpu);
2175 #endif /* CONFIG_SMP */
2177 #ifdef CONFIG_SCHED_CORE
2178 extern void sched_core_free(struct task_struct *tsk);
2179 extern void sched_core_fork(struct task_struct *p);
2180 extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2181 unsigned long uaddr);
2182 extern int sched_core_idle_cpu(int cpu);
2184 static inline void sched_core_free(struct task_struct *tsk) { }
2185 static inline void sched_core_fork(struct task_struct *p) { }
2186 static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2189 extern void sched_set_stop_task(int cpu, struct task_struct *stop);