1 /* SPDX-License-Identifier: GPL-2.0 */
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
10 #include <uapi/linux/sched.h>
12 #include <asm/current.h>
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/resource.h>
25 #include <linux/latencytop.h>
26 #include <linux/sched/prio.h>
27 #include <linux/signal_types.h>
28 #include <linux/mm_types_task.h>
29 #include <linux/task_io_accounting.h>
31 /* task_struct member predeclarations (sorted alphabetically): */
33 struct backing_dev_info;
38 struct futex_pi_state;
43 struct perf_event_context;
45 struct pipe_inode_info;
48 struct robust_list_head;
52 struct sighand_struct;
54 struct task_delay_info;
58 * Task state bitmask. NOTE! These bits are also
59 * encoded in fs/proc/array.c: get_task_state().
61 * We have two separate sets of flags: task->state
62 * is about runnability, while task->exit_state are
63 * about the task exiting. Confusing, but this way
64 * modifying one set can't modify the other one by
68 /* Used in tsk->state: */
69 #define TASK_RUNNING 0x0000
70 #define TASK_INTERRUPTIBLE 0x0001
71 #define TASK_UNINTERRUPTIBLE 0x0002
72 #define __TASK_STOPPED 0x0004
73 #define __TASK_TRACED 0x0008
74 /* Used in tsk->exit_state: */
75 #define EXIT_DEAD 0x0010
76 #define EXIT_ZOMBIE 0x0020
77 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
78 /* Used in tsk->state again: */
79 #define TASK_PARKED 0x0040
80 #define TASK_DEAD 0x0080
81 #define TASK_WAKEKILL 0x0100
82 #define TASK_WAKING 0x0200
83 #define TASK_NOLOAD 0x0400
84 #define TASK_NEW 0x0800
85 #define TASK_STATE_MAX 0x1000
87 /* Convenience macros for the sake of set_current_state: */
88 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
89 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
90 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
92 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
94 /* Convenience macros for the sake of wake_up(): */
95 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
96 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
98 /* get_task_state(): */
99 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
100 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
101 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
104 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
108 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
110 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 (task->flags & PF_FROZEN) == 0 && \
112 (task->state & TASK_NOLOAD) == 0)
114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
117 * Special states are those that do not use the normal wait-loop pattern. See
118 * the comment with set_special_state().
120 #define is_special_task_state(state) \
121 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_DEAD))
123 #define __set_current_state(state_value) \
125 WARN_ON_ONCE(is_special_task_state(state_value));\
126 current->task_state_change = _THIS_IP_; \
127 current->state = (state_value); \
130 #define set_current_state(state_value) \
132 WARN_ON_ONCE(is_special_task_state(state_value));\
133 current->task_state_change = _THIS_IP_; \
134 smp_store_mb(current->state, (state_value)); \
137 #define set_special_state(state_value) \
139 unsigned long flags; /* may shadow */ \
140 WARN_ON_ONCE(!is_special_task_state(state_value)); \
141 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
142 current->task_state_change = _THIS_IP_; \
143 current->state = (state_value); \
144 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
148 * set_current_state() includes a barrier so that the write of current->state
149 * is correctly serialised wrt the caller's subsequent test of whether to
153 * set_current_state(TASK_UNINTERRUPTIBLE);
159 * __set_current_state(TASK_RUNNING);
161 * If the caller does not need such serialisation (because, for instance, the
162 * condition test and condition change and wakeup are under the same lock) then
163 * use __set_current_state().
165 * The above is typically ordered against the wakeup, which does:
167 * need_sleep = false;
168 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
170 * Where wake_up_state() (and all other wakeup primitives) imply enough
171 * barriers to order the store of the variable against wakeup.
173 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
174 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
175 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
177 * However, with slightly different timing the wakeup TASK_RUNNING store can
178 * also collide with the TASK_UNINTERRUPTIBLE store. Loosing that store is not
179 * a problem either because that will result in one extra go around the loop
180 * and our @cond test will save the day.
182 * Also see the comments of try_to_wake_up().
184 #define __set_current_state(state_value) \
185 current->state = (state_value)
187 #define set_current_state(state_value) \
188 smp_store_mb(current->state, (state_value))
191 * set_special_state() should be used for those states when the blocking task
192 * can not use the regular condition based wait-loop. In that case we must
193 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
194 * will not collide with our state change.
196 #define set_special_state(state_value) \
198 unsigned long flags; /* may shadow */ \
199 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
200 current->state = (state_value); \
201 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
206 /* Task command name length: */
207 #define TASK_COMM_LEN 16
209 extern cpumask_var_t cpu_isolated_map;
211 extern void scheduler_tick(void);
213 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
215 extern long schedule_timeout(long timeout);
216 extern long schedule_timeout_interruptible(long timeout);
217 extern long schedule_timeout_killable(long timeout);
218 extern long schedule_timeout_uninterruptible(long timeout);
219 extern long schedule_timeout_idle(long timeout);
220 asmlinkage void schedule(void);
221 extern void schedule_preempt_disabled(void);
223 extern int __must_check io_schedule_prepare(void);
224 extern void io_schedule_finish(int token);
225 extern long io_schedule_timeout(long timeout);
226 extern void io_schedule(void);
229 * struct prev_cputime - snapshot of system and user cputime
230 * @utime: time spent in user mode
231 * @stime: time spent in system mode
232 * @lock: protects the above two fields
234 * Stores previous user/system time values such that we can guarantee
237 struct prev_cputime {
238 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
246 * struct task_cputime - collected CPU time counts
247 * @utime: time spent in user mode, in nanoseconds
248 * @stime: time spent in kernel mode, in nanoseconds
249 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
251 * This structure groups together three kinds of CPU time that are tracked for
252 * threads and thread groups. Most things considering CPU time want to group
253 * these counts together and treat all three of them in parallel.
255 struct task_cputime {
258 unsigned long long sum_exec_runtime;
261 /* Alternate field names when used on cache expirations: */
262 #define virt_exp utime
263 #define prof_exp stime
264 #define sched_exp sum_exec_runtime
267 /* Task is sleeping or running in a CPU with VTIME inactive: */
269 /* Task runs in userspace in a CPU with VTIME active: */
271 /* Task runs in kernelspace in a CPU with VTIME active: */
277 unsigned long long starttime;
278 enum vtime_state state;
285 #ifdef CONFIG_SCHED_INFO
286 /* Cumulative counters: */
288 /* # of times we have run on this CPU: */
289 unsigned long pcount;
291 /* Time spent waiting on a runqueue: */
292 unsigned long long run_delay;
296 /* When did we last run on a CPU? */
297 unsigned long long last_arrival;
299 /* When were we last queued to run? */
300 unsigned long long last_queued;
302 #endif /* CONFIG_SCHED_INFO */
306 * Integer metrics need fixed point arithmetic, e.g., sched/fair
307 * has a few: load, load_avg, util_avg, freq, and capacity.
309 * We define a basic fixed point arithmetic range, and then formalize
310 * all these metrics based on that basic range.
312 # define SCHED_FIXEDPOINT_SHIFT 10
313 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
316 unsigned long weight;
321 * The load_avg/util_avg accumulates an infinite geometric series
322 * (see __update_load_avg() in kernel/sched/fair.c).
324 * [load_avg definition]
326 * load_avg = runnable% * scale_load_down(load)
328 * where runnable% is the time ratio that a sched_entity is runnable.
329 * For cfs_rq, it is the aggregated load_avg of all runnable and
330 * blocked sched_entities.
332 * load_avg may also take frequency scaling into account:
334 * load_avg = runnable% * scale_load_down(load) * freq%
336 * where freq% is the CPU frequency normalized to the highest frequency.
338 * [util_avg definition]
340 * util_avg = running% * SCHED_CAPACITY_SCALE
342 * where running% is the time ratio that a sched_entity is running on
343 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
344 * and blocked sched_entities.
346 * util_avg may also factor frequency scaling and CPU capacity scaling:
348 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
350 * where freq% is the same as above, and capacity% is the CPU capacity
351 * normalized to the greatest capacity (due to uarch differences, etc).
353 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
354 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
355 * we therefore scale them to as large a range as necessary. This is for
356 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
360 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
361 * with the highest load (=88761), always runnable on a single cfs_rq,
362 * and should not overflow as the number already hits PID_MAX_LIMIT.
364 * For all other cases (including 32-bit kernels), struct load_weight's
365 * weight will overflow first before we do, because:
367 * Max(load_avg) <= Max(load.weight)
369 * Then it is the load_weight's responsibility to consider overflow
373 u64 last_update_time;
377 unsigned long load_avg;
378 unsigned long util_avg;
381 struct sched_statistics {
382 #ifdef CONFIG_SCHEDSTATS
392 s64 sum_sleep_runtime;
399 u64 nr_migrations_cold;
400 u64 nr_failed_migrations_affine;
401 u64 nr_failed_migrations_running;
402 u64 nr_failed_migrations_hot;
403 u64 nr_forced_migrations;
407 u64 nr_wakeups_migrate;
408 u64 nr_wakeups_local;
409 u64 nr_wakeups_remote;
410 u64 nr_wakeups_affine;
411 u64 nr_wakeups_affine_attempts;
412 u64 nr_wakeups_passive;
417 struct sched_entity {
418 /* For load-balancing: */
419 struct load_weight load;
420 struct rb_node run_node;
421 struct list_head group_node;
425 u64 sum_exec_runtime;
427 u64 prev_sum_exec_runtime;
431 struct sched_statistics statistics;
433 #ifdef CONFIG_FAIR_GROUP_SCHED
435 struct sched_entity *parent;
436 /* rq on which this entity is (to be) queued: */
437 struct cfs_rq *cfs_rq;
438 /* rq "owned" by this entity/group: */
444 * Per entity load average tracking.
446 * Put into separate cache line so it does not
447 * collide with read-mostly values above.
449 struct sched_avg avg ____cacheline_aligned_in_smp;
453 struct sched_rt_entity {
454 struct list_head run_list;
455 unsigned long timeout;
456 unsigned long watchdog_stamp;
457 unsigned int time_slice;
458 unsigned short on_rq;
459 unsigned short on_list;
461 struct sched_rt_entity *back;
462 #ifdef CONFIG_RT_GROUP_SCHED
463 struct sched_rt_entity *parent;
464 /* rq on which this entity is (to be) queued: */
466 /* rq "owned" by this entity/group: */
469 } __randomize_layout;
471 struct sched_dl_entity {
472 struct rb_node rb_node;
475 * Original scheduling parameters. Copied here from sched_attr
476 * during sched_setattr(), they will remain the same until
477 * the next sched_setattr().
479 u64 dl_runtime; /* Maximum runtime for each instance */
480 u64 dl_deadline; /* Relative deadline of each instance */
481 u64 dl_period; /* Separation of two instances (period) */
482 u64 dl_bw; /* dl_runtime / dl_period */
483 u64 dl_density; /* dl_runtime / dl_deadline */
486 * Actual scheduling parameters. Initialized with the values above,
487 * they are continously updated during task execution. Note that
488 * the remaining runtime could be < 0 in case we are in overrun.
490 s64 runtime; /* Remaining runtime for this instance */
491 u64 deadline; /* Absolute deadline for this instance */
492 unsigned int flags; /* Specifying the scheduler behaviour */
497 * @dl_throttled tells if we exhausted the runtime. If so, the
498 * task has to wait for a replenishment to be performed at the
499 * next firing of dl_timer.
501 * @dl_boosted tells if we are boosted due to DI. If so we are
502 * outside bandwidth enforcement mechanism (but only until we
503 * exit the critical section);
505 * @dl_yielded tells if task gave up the CPU before consuming
506 * all its available runtime during the last job.
508 * @dl_non_contending tells if the task is inactive while still
509 * contributing to the active utilization. In other words, it
510 * indicates if the inactive timer has been armed and its handler
511 * has not been executed yet. This flag is useful to avoid race
512 * conditions between the inactive timer handler and the wakeup
518 int dl_non_contending;
521 * Bandwidth enforcement timer. Each -deadline task has its
522 * own bandwidth to be enforced, thus we need one timer per task.
524 struct hrtimer dl_timer;
527 * Inactive timer, responsible for decreasing the active utilization
528 * at the "0-lag time". When a -deadline task blocks, it contributes
529 * to GRUB's active utilization until the "0-lag time", hence a
530 * timer is needed to decrease the active utilization at the correct
533 struct hrtimer inactive_timer;
542 /* Otherwise the compiler can store garbage here: */
545 u32 s; /* Set of bits. */
548 enum perf_event_task_context {
549 perf_invalid_context = -1,
552 perf_nr_task_contexts,
556 struct wake_q_node *next;
560 #ifdef CONFIG_THREAD_INFO_IN_TASK
562 * For reasons of header soup (see current_thread_info()), this
563 * must be the first element of task_struct.
565 struct thread_info thread_info;
567 /* -1 unrunnable, 0 runnable, >0 stopped: */
571 * This begins the randomizable portion of task_struct. Only
572 * scheduling-critical items should be added above here.
574 randomized_struct_fields_start
578 /* Per task flags (PF_*), defined further below: */
583 struct llist_node wake_entry;
585 #ifdef CONFIG_THREAD_INFO_IN_TASK
589 unsigned int wakee_flips;
590 unsigned long wakee_flip_decay_ts;
591 struct task_struct *last_wakee;
600 unsigned int rt_priority;
602 const struct sched_class *sched_class;
603 struct sched_entity se;
604 struct sched_rt_entity rt;
605 #ifdef CONFIG_CGROUP_SCHED
606 struct task_group *sched_task_group;
608 struct sched_dl_entity dl;
610 #ifdef CONFIG_PREEMPT_NOTIFIERS
611 /* List of struct preempt_notifier: */
612 struct hlist_head preempt_notifiers;
615 #ifdef CONFIG_BLK_DEV_IO_TRACE
616 unsigned int btrace_seq;
621 cpumask_t cpus_allowed;
623 #ifdef CONFIG_PREEMPT_RCU
624 int rcu_read_lock_nesting;
625 union rcu_special rcu_read_unlock_special;
626 struct list_head rcu_node_entry;
627 struct rcu_node *rcu_blocked_node;
628 #endif /* #ifdef CONFIG_PREEMPT_RCU */
630 #ifdef CONFIG_TASKS_RCU
631 unsigned long rcu_tasks_nvcsw;
632 u8 rcu_tasks_holdout;
634 int rcu_tasks_idle_cpu;
635 struct list_head rcu_tasks_holdout_list;
636 #endif /* #ifdef CONFIG_TASKS_RCU */
638 struct sched_info sched_info;
640 struct list_head tasks;
642 struct plist_node pushable_tasks;
643 struct rb_node pushable_dl_tasks;
646 struct mm_struct *mm;
647 struct mm_struct *active_mm;
649 /* Per-thread vma caching: */
650 struct vmacache vmacache;
652 #ifdef SPLIT_RSS_COUNTING
653 struct task_rss_stat rss_stat;
658 /* The signal sent when the parent dies: */
660 /* JOBCTL_*, siglock protected: */
661 unsigned long jobctl;
663 /* Used for emulating ABI behavior of previous Linux versions: */
664 unsigned int personality;
666 /* Scheduler bits, serialized by scheduler locks: */
667 unsigned sched_reset_on_fork:1;
668 unsigned sched_contributes_to_load:1;
669 unsigned sched_migrated:1;
670 unsigned sched_remote_wakeup:1;
671 /* Force alignment to the next boundary: */
674 /* Unserialized, strictly 'current' */
676 /* Bit to tell LSMs we're in execve(): */
677 unsigned in_execve:1;
678 unsigned in_iowait:1;
679 #ifndef TIF_RESTORE_SIGMASK
680 unsigned restore_sigmask:1;
683 unsigned memcg_may_oom:1;
685 unsigned memcg_kmem_skip_account:1;
688 #ifdef CONFIG_COMPAT_BRK
689 unsigned brk_randomized:1;
691 #ifdef CONFIG_CGROUPS
692 /* disallow userland-initiated cgroup migration */
693 unsigned no_cgroup_migration:1;
696 unsigned long atomic_flags; /* Flags requiring atomic access. */
698 struct restart_block restart_block;
703 #ifdef CONFIG_CC_STACKPROTECTOR
704 /* Canary value for the -fstack-protector GCC feature: */
705 unsigned long stack_canary;
708 * Pointers to the (original) parent process, youngest child, younger sibling,
709 * older sibling, respectively. (p->father can be replaced with
710 * p->real_parent->pid)
713 /* Real parent process: */
714 struct task_struct __rcu *real_parent;
716 /* Recipient of SIGCHLD, wait4() reports: */
717 struct task_struct __rcu *parent;
720 * Children/sibling form the list of natural children:
722 struct list_head children;
723 struct list_head sibling;
724 struct task_struct *group_leader;
727 * 'ptraced' is the list of tasks this task is using ptrace() on.
729 * This includes both natural children and PTRACE_ATTACH targets.
730 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
732 struct list_head ptraced;
733 struct list_head ptrace_entry;
735 /* PID/PID hash table linkage. */
736 struct pid_link pids[PIDTYPE_MAX];
737 struct list_head thread_group;
738 struct list_head thread_node;
740 struct completion *vfork_done;
742 /* CLONE_CHILD_SETTID: */
743 int __user *set_child_tid;
745 /* CLONE_CHILD_CLEARTID: */
746 int __user *clear_child_tid;
750 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
755 struct prev_cputime prev_cputime;
756 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
760 #ifdef CONFIG_NO_HZ_FULL
761 atomic_t tick_dep_mask;
763 /* Context switch counts: */
765 unsigned long nivcsw;
767 /* Monotonic time in nsecs: */
770 /* Boot based time in nsecs: */
773 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
774 unsigned long min_flt;
775 unsigned long maj_flt;
777 #ifdef CONFIG_POSIX_TIMERS
778 struct task_cputime cputime_expires;
779 struct list_head cpu_timers[3];
782 /* Process credentials: */
784 /* Tracer's credentials at attach: */
785 const struct cred __rcu *ptracer_cred;
787 /* Objective and real subjective task credentials (COW): */
788 const struct cred __rcu *real_cred;
790 /* Effective (overridable) subjective task credentials (COW): */
791 const struct cred __rcu *cred;
794 * executable name, excluding path.
796 * - normally initialized setup_new_exec()
797 * - access it with [gs]et_task_comm()
798 * - lock it with task_lock()
800 char comm[TASK_COMM_LEN];
802 struct nameidata *nameidata;
804 #ifdef CONFIG_SYSVIPC
805 struct sysv_sem sysvsem;
806 struct sysv_shm sysvshm;
808 #ifdef CONFIG_DETECT_HUNG_TASK
809 unsigned long last_switch_count;
811 /* Filesystem information: */
812 struct fs_struct *fs;
814 /* Open file information: */
815 struct files_struct *files;
818 struct nsproxy *nsproxy;
820 /* Signal handlers: */
821 struct signal_struct *signal;
822 struct sighand_struct *sighand;
824 sigset_t real_blocked;
825 /* Restored if set_restore_sigmask() was used: */
826 sigset_t saved_sigmask;
827 struct sigpending pending;
828 unsigned long sas_ss_sp;
830 unsigned int sas_ss_flags;
832 struct callback_head *task_works;
834 struct audit_context *audit_context;
835 #ifdef CONFIG_AUDITSYSCALL
837 unsigned int sessionid;
839 struct seccomp seccomp;
841 /* Thread group tracking: */
845 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
846 spinlock_t alloc_lock;
848 /* Protection of the PI data structures: */
849 raw_spinlock_t pi_lock;
851 struct wake_q_node wake_q;
853 #ifdef CONFIG_RT_MUTEXES
854 /* PI waiters blocked on a rt_mutex held by this task: */
855 struct rb_root_cached pi_waiters;
856 /* Updated under owner's pi_lock and rq lock */
857 struct task_struct *pi_top_task;
858 /* Deadlock detection and priority inheritance handling: */
859 struct rt_mutex_waiter *pi_blocked_on;
862 #ifdef CONFIG_DEBUG_MUTEXES
863 /* Mutex deadlock detection: */
864 struct mutex_waiter *blocked_on;
867 #ifdef CONFIG_TRACE_IRQFLAGS
868 unsigned int irq_events;
869 unsigned long hardirq_enable_ip;
870 unsigned long hardirq_disable_ip;
871 unsigned int hardirq_enable_event;
872 unsigned int hardirq_disable_event;
873 int hardirqs_enabled;
875 unsigned long softirq_disable_ip;
876 unsigned long softirq_enable_ip;
877 unsigned int softirq_disable_event;
878 unsigned int softirq_enable_event;
879 int softirqs_enabled;
883 #ifdef CONFIG_LOCKDEP
884 # define MAX_LOCK_DEPTH 48UL
887 unsigned int lockdep_recursion;
888 struct held_lock held_locks[MAX_LOCK_DEPTH];
891 #ifdef CONFIG_LOCKDEP_CROSSRELEASE
892 #define MAX_XHLOCKS_NR 64UL
893 struct hist_lock *xhlocks; /* Crossrelease history locks */
894 unsigned int xhlock_idx;
895 /* For restoring at history boundaries */
896 unsigned int xhlock_idx_hist[XHLOCK_CTX_NR];
897 unsigned int hist_id;
898 /* For overwrite check at each context exit */
899 unsigned int hist_id_save[XHLOCK_CTX_NR];
903 unsigned int in_ubsan;
906 /* Journalling filesystem info: */
909 /* Stacked block device info: */
910 struct bio_list *bio_list;
913 /* Stack plugging: */
914 struct blk_plug *plug;
918 struct reclaim_state *reclaim_state;
920 struct backing_dev_info *backing_dev_info;
922 struct io_context *io_context;
925 unsigned long ptrace_message;
926 siginfo_t *last_siginfo;
928 struct task_io_accounting ioac;
929 #ifdef CONFIG_TASK_XACCT
930 /* Accumulated RSS usage: */
932 /* Accumulated virtual memory usage: */
934 /* stime + utime since last update: */
937 #ifdef CONFIG_CPUSETS
938 /* Protected by ->alloc_lock: */
939 nodemask_t mems_allowed;
940 /* Seqence number to catch updates: */
941 seqcount_t mems_allowed_seq;
942 int cpuset_mem_spread_rotor;
943 int cpuset_slab_spread_rotor;
945 #ifdef CONFIG_CGROUPS
946 /* Control Group info protected by css_set_lock: */
947 struct css_set __rcu *cgroups;
948 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
949 struct list_head cg_list;
951 #ifdef CONFIG_INTEL_RDT
956 struct robust_list_head __user *robust_list;
958 struct compat_robust_list_head __user *compat_robust_list;
960 struct list_head pi_state_list;
961 struct futex_pi_state *pi_state_cache;
963 #ifdef CONFIG_PERF_EVENTS
964 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
965 struct mutex perf_event_mutex;
966 struct list_head perf_event_list;
968 #ifdef CONFIG_DEBUG_PREEMPT
969 unsigned long preempt_disable_ip;
972 /* Protected by alloc_lock: */
973 struct mempolicy *mempolicy;
975 short pref_node_fork;
977 #ifdef CONFIG_NUMA_BALANCING
979 unsigned int numa_scan_period;
980 unsigned int numa_scan_period_max;
981 int numa_preferred_nid;
982 unsigned long numa_migrate_retry;
983 /* Migration stamp: */
985 u64 last_task_numa_placement;
986 u64 last_sum_exec_runtime;
987 struct callback_head numa_work;
989 struct list_head numa_entry;
990 struct numa_group *numa_group;
993 * numa_faults is an array split into four regions:
994 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
995 * in this precise order.
997 * faults_memory: Exponential decaying average of faults on a per-node
998 * basis. Scheduling placement decisions are made based on these
999 * counts. The values remain static for the duration of a PTE scan.
1000 * faults_cpu: Track the nodes the process was running on when a NUMA
1001 * hinting fault was incurred.
1002 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1003 * during the current scan window. When the scan completes, the counts
1004 * in faults_memory and faults_cpu decay and these values are copied.
1006 unsigned long *numa_faults;
1007 unsigned long total_numa_faults;
1010 * numa_faults_locality tracks if faults recorded during the last
1011 * scan window were remote/local or failed to migrate. The task scan
1012 * period is adapted based on the locality of the faults with different
1013 * weights depending on whether they were shared or private faults
1015 unsigned long numa_faults_locality[3];
1017 unsigned long numa_pages_migrated;
1018 #endif /* CONFIG_NUMA_BALANCING */
1020 struct tlbflush_unmap_batch tlb_ubc;
1022 struct rcu_head rcu;
1024 /* Cache last used pipe for splice(): */
1025 struct pipe_inode_info *splice_pipe;
1027 struct page_frag task_frag;
1029 #ifdef CONFIG_TASK_DELAY_ACCT
1030 struct task_delay_info *delays;
1033 #ifdef CONFIG_FAULT_INJECTION
1035 unsigned int fail_nth;
1038 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1039 * balance_dirty_pages() for a dirty throttling pause:
1042 int nr_dirtied_pause;
1043 /* Start of a write-and-pause period: */
1044 unsigned long dirty_paused_when;
1046 #ifdef CONFIG_LATENCYTOP
1047 int latency_record_count;
1048 struct latency_record latency_record[LT_SAVECOUNT];
1051 * Time slack values; these are used to round up poll() and
1052 * select() etc timeout values. These are in nanoseconds.
1055 u64 default_timer_slack_ns;
1058 unsigned int kasan_depth;
1061 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1062 /* Index of current stored address in ret_stack: */
1065 /* Stack of return addresses for return function tracing: */
1066 struct ftrace_ret_stack *ret_stack;
1068 /* Timestamp for last schedule: */
1069 unsigned long long ftrace_timestamp;
1072 * Number of functions that haven't been traced
1073 * because of depth overrun:
1075 atomic_t trace_overrun;
1077 /* Pause tracing: */
1078 atomic_t tracing_graph_pause;
1081 #ifdef CONFIG_TRACING
1082 /* State flags for use by tracers: */
1083 unsigned long trace;
1085 /* Bitmask and counter of trace recursion: */
1086 unsigned long trace_recursion;
1087 #endif /* CONFIG_TRACING */
1090 /* Coverage collection mode enabled for this task (0 if disabled): */
1091 enum kcov_mode kcov_mode;
1093 /* Size of the kcov_area: */
1094 unsigned int kcov_size;
1096 /* Buffer for coverage collection: */
1099 /* KCOV descriptor wired with this task or NULL: */
1104 struct mem_cgroup *memcg_in_oom;
1105 gfp_t memcg_oom_gfp_mask;
1106 int memcg_oom_order;
1108 /* Number of pages to reclaim on returning to userland: */
1109 unsigned int memcg_nr_pages_over_high;
1112 #ifdef CONFIG_UPROBES
1113 struct uprobe_task *utask;
1115 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1116 unsigned int sequential_io;
1117 unsigned int sequential_io_avg;
1119 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1120 unsigned long task_state_change;
1122 int pagefault_disabled;
1124 struct task_struct *oom_reaper_list;
1126 #ifdef CONFIG_VMAP_STACK
1127 struct vm_struct *stack_vm_area;
1129 #ifdef CONFIG_THREAD_INFO_IN_TASK
1130 /* A live task holds one reference: */
1131 atomic_t stack_refcount;
1133 #ifdef CONFIG_LIVEPATCH
1136 #ifdef CONFIG_SECURITY
1137 /* Used by LSM modules for access restriction: */
1142 * New fields for task_struct should be added above here, so that
1143 * they are included in the randomized portion of task_struct.
1145 randomized_struct_fields_end
1147 /* CPU-specific state of this task: */
1148 struct thread_struct thread;
1151 * WARNING: on x86, 'thread_struct' contains a variable-sized
1152 * structure. It *MUST* be at the end of 'task_struct'.
1154 * Do not put anything below here!
1158 static inline struct pid *task_pid(struct task_struct *task)
1160 return task->pids[PIDTYPE_PID].pid;
1163 static inline struct pid *task_tgid(struct task_struct *task)
1165 return task->group_leader->pids[PIDTYPE_PID].pid;
1169 * Without tasklist or RCU lock it is not safe to dereference
1170 * the result of task_pgrp/task_session even if task == current,
1171 * we can race with another thread doing sys_setsid/sys_setpgid.
1173 static inline struct pid *task_pgrp(struct task_struct *task)
1175 return task->group_leader->pids[PIDTYPE_PGID].pid;
1178 static inline struct pid *task_session(struct task_struct *task)
1180 return task->group_leader->pids[PIDTYPE_SID].pid;
1184 * the helpers to get the task's different pids as they are seen
1185 * from various namespaces
1187 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1188 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1190 * task_xid_nr_ns() : id seen from the ns specified;
1192 * see also pid_nr() etc in include/linux/pid.h
1194 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1196 static inline pid_t task_pid_nr(struct task_struct *tsk)
1201 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1203 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1206 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1208 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1212 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1218 * pid_alive - check that a task structure is not stale
1219 * @p: Task structure to be checked.
1221 * Test if a process is not yet dead (at most zombie state)
1222 * If pid_alive fails, then pointers within the task structure
1223 * can be stale and must not be dereferenced.
1225 * Return: 1 if the process is alive. 0 otherwise.
1227 static inline int pid_alive(const struct task_struct *p)
1229 return p->pids[PIDTYPE_PID].pid != NULL;
1232 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1234 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1237 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1239 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1243 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1245 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1248 static inline pid_t task_session_vnr(struct task_struct *tsk)
1250 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1253 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1255 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
1258 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1260 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
1263 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1269 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1275 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1277 return task_ppid_nr_ns(tsk, &init_pid_ns);
1280 /* Obsolete, do not use: */
1281 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1283 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1286 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1287 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1289 static inline unsigned int __get_task_state(struct task_struct *tsk)
1291 unsigned int tsk_state = READ_ONCE(tsk->state);
1292 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1294 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1296 if (tsk_state == TASK_IDLE)
1297 state = TASK_REPORT_IDLE;
1302 static inline char __task_state_to_char(unsigned int state)
1304 static const char state_char[] = "RSDTtXZPI";
1306 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1308 return state_char[state];
1311 static inline char task_state_to_char(struct task_struct *tsk)
1313 return __task_state_to_char(__get_task_state(tsk));
1317 * is_global_init - check if a task structure is init. Since init
1318 * is free to have sub-threads we need to check tgid.
1319 * @tsk: Task structure to be checked.
1321 * Check if a task structure is the first user space task the kernel created.
1323 * Return: 1 if the task structure is init. 0 otherwise.
1325 static inline int is_global_init(struct task_struct *tsk)
1327 return task_tgid_nr(tsk) == 1;
1330 extern struct pid *cad_pid;
1335 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1336 #define PF_EXITING 0x00000004 /* Getting shut down */
1337 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1338 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1339 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1340 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1341 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1342 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1343 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1344 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1345 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1346 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1347 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1348 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1349 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1350 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1351 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1352 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1353 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1354 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1355 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1356 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1357 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1358 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1359 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1360 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1361 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1362 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1365 * Only the _current_ task can read/write to tsk->flags, but other
1366 * tasks can access tsk->flags in readonly mode for example
1367 * with tsk_used_math (like during threaded core dumping).
1368 * There is however an exception to this rule during ptrace
1369 * or during fork: the ptracer task is allowed to write to the
1370 * child->flags of its traced child (same goes for fork, the parent
1371 * can write to the child->flags), because we're guaranteed the
1372 * child is not running and in turn not changing child->flags
1373 * at the same time the parent does it.
1375 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1376 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1377 #define clear_used_math() clear_stopped_child_used_math(current)
1378 #define set_used_math() set_stopped_child_used_math(current)
1380 #define conditional_stopped_child_used_math(condition, child) \
1381 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1383 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1385 #define copy_to_stopped_child_used_math(child) \
1386 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1388 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1389 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1390 #define used_math() tsk_used_math(current)
1392 static inline bool is_percpu_thread(void)
1395 return (current->flags & PF_NO_SETAFFINITY) &&
1396 (current->nr_cpus_allowed == 1);
1402 /* Per-process atomic flags. */
1403 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1404 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1405 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1406 #define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1407 #define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1409 #define TASK_PFA_TEST(name, func) \
1410 static inline bool task_##func(struct task_struct *p) \
1411 { return test_bit(PFA_##name, &p->atomic_flags); }
1413 #define TASK_PFA_SET(name, func) \
1414 static inline void task_set_##func(struct task_struct *p) \
1415 { set_bit(PFA_##name, &p->atomic_flags); }
1417 #define TASK_PFA_CLEAR(name, func) \
1418 static inline void task_clear_##func(struct task_struct *p) \
1419 { clear_bit(PFA_##name, &p->atomic_flags); }
1421 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1422 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1424 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1425 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1426 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1428 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1429 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1430 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1432 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1433 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1434 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1436 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1437 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1440 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1442 current->flags &= ~flags;
1443 current->flags |= orig_flags & flags;
1446 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1447 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1449 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1450 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1452 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1455 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1457 if (!cpumask_test_cpu(0, new_mask))
1463 #ifndef cpu_relax_yield
1464 #define cpu_relax_yield() cpu_relax()
1467 extern int yield_to(struct task_struct *p, bool preempt);
1468 extern void set_user_nice(struct task_struct *p, long nice);
1469 extern int task_prio(const struct task_struct *p);
1472 * task_nice - return the nice value of a given task.
1473 * @p: the task in question.
1475 * Return: The nice value [ -20 ... 0 ... 19 ].
1477 static inline int task_nice(const struct task_struct *p)
1479 return PRIO_TO_NICE((p)->static_prio);
1482 extern int can_nice(const struct task_struct *p, const int nice);
1483 extern int task_curr(const struct task_struct *p);
1484 extern int idle_cpu(int cpu);
1485 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1486 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1487 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1488 extern struct task_struct *idle_task(int cpu);
1491 * is_idle_task - is the specified task an idle task?
1492 * @p: the task in question.
1494 * Return: 1 if @p is an idle task. 0 otherwise.
1496 static inline bool is_idle_task(const struct task_struct *p)
1498 return !!(p->flags & PF_IDLE);
1501 extern struct task_struct *curr_task(int cpu);
1502 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1506 union thread_union {
1507 #ifndef CONFIG_THREAD_INFO_IN_TASK
1508 struct thread_info thread_info;
1510 unsigned long stack[THREAD_SIZE/sizeof(long)];
1513 #ifdef CONFIG_THREAD_INFO_IN_TASK
1514 static inline struct thread_info *task_thread_info(struct task_struct *task)
1516 return &task->thread_info;
1518 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1519 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1523 * find a task by one of its numerical ids
1525 * find_task_by_pid_ns():
1526 * finds a task by its pid in the specified namespace
1527 * find_task_by_vpid():
1528 * finds a task by its virtual pid
1530 * see also find_vpid() etc in include/linux/pid.h
1533 extern struct task_struct *find_task_by_vpid(pid_t nr);
1534 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1536 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1537 extern int wake_up_process(struct task_struct *tsk);
1538 extern void wake_up_new_task(struct task_struct *tsk);
1541 extern void kick_process(struct task_struct *tsk);
1543 static inline void kick_process(struct task_struct *tsk) { }
1546 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1548 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1550 __set_task_comm(tsk, from, false);
1553 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1554 #define get_task_comm(buf, tsk) ({ \
1555 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1556 __get_task_comm(buf, sizeof(buf), tsk); \
1560 void scheduler_ipi(void);
1561 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1563 static inline void scheduler_ipi(void) { }
1564 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1571 * Set thread flags in other task's structures.
1572 * See asm/thread_info.h for TIF_xxxx flags available:
1574 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1576 set_ti_thread_flag(task_thread_info(tsk), flag);
1579 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1581 clear_ti_thread_flag(task_thread_info(tsk), flag);
1584 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1586 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1589 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1591 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1594 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1596 return test_ti_thread_flag(task_thread_info(tsk), flag);
1599 static inline void set_tsk_need_resched(struct task_struct *tsk)
1601 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1604 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1606 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1609 static inline int test_tsk_need_resched(struct task_struct *tsk)
1611 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1615 * cond_resched() and cond_resched_lock(): latency reduction via
1616 * explicit rescheduling in places that are safe. The return
1617 * value indicates whether a reschedule was done in fact.
1618 * cond_resched_lock() will drop the spinlock before scheduling,
1619 * cond_resched_softirq() will enable bhs before scheduling.
1621 #ifndef CONFIG_PREEMPT
1622 extern int _cond_resched(void);
1624 static inline int _cond_resched(void) { return 0; }
1627 #define cond_resched() ({ \
1628 ___might_sleep(__FILE__, __LINE__, 0); \
1632 extern int __cond_resched_lock(spinlock_t *lock);
1634 #define cond_resched_lock(lock) ({ \
1635 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1636 __cond_resched_lock(lock); \
1639 extern int __cond_resched_softirq(void);
1641 #define cond_resched_softirq() ({ \
1642 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1643 __cond_resched_softirq(); \
1646 static inline void cond_resched_rcu(void)
1648 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1656 * Does a critical section need to be broken due to another
1657 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1658 * but a general need for low latency)
1660 static inline int spin_needbreak(spinlock_t *lock)
1662 #ifdef CONFIG_PREEMPT
1663 return spin_is_contended(lock);
1669 static __always_inline bool need_resched(void)
1671 return unlikely(tif_need_resched());
1675 * Wrappers for p->thread_info->cpu access. No-op on UP.
1679 static inline unsigned int task_cpu(const struct task_struct *p)
1681 #ifdef CONFIG_THREAD_INFO_IN_TASK
1684 return task_thread_info(p)->cpu;
1688 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1692 static inline unsigned int task_cpu(const struct task_struct *p)
1697 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1701 #endif /* CONFIG_SMP */
1704 * In order to reduce various lock holder preemption latencies provide an
1705 * interface to see if a vCPU is currently running or not.
1707 * This allows us to terminate optimistic spin loops and block, analogous to
1708 * the native optimistic spin heuristic of testing if the lock owner task is
1711 #ifndef vcpu_is_preempted
1712 # define vcpu_is_preempted(cpu) false
1715 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1716 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1718 #ifndef TASK_SIZE_OF
1719 #define TASK_SIZE_OF(tsk) TASK_SIZE