4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
8 #include <linux/mutex.h>
9 #include <linux/plist.h>
10 #include <linux/mm_types_task.h>
11 #include <asm/ptrace.h>
13 #include <linux/sem.h>
14 #include <linux/shm.h>
15 #include <linux/signal.h>
16 #include <linux/signal_types.h>
17 #include <linux/pid.h>
18 #include <linux/seccomp.h>
19 #include <linux/rculist.h>
20 #include <linux/rtmutex.h>
22 #include <linux/resource.h>
23 #include <linux/hrtimer.h>
24 #include <linux/kcov.h>
25 #include <linux/task_io_accounting.h>
26 #include <linux/latencytop.h>
27 #include <linux/gfp.h>
28 #include <linux/topology.h>
29 #include <linux/magic.h>
31 #include <asm/current.h>
33 /* task_struct member predeclarations: */
36 struct backing_dev_info;
42 struct futex_pi_state;
47 struct perf_event_context;
49 struct pipe_inode_info;
52 struct robust_list_head;
56 struct sighand_struct;
58 struct task_delay_info;
64 * Task state bitmask. NOTE! These bits are also
65 * encoded in fs/proc/array.c: get_task_state().
67 * We have two separate sets of flags: task->state
68 * is about runnability, while task->exit_state are
69 * about the task exiting. Confusing, but this way
70 * modifying one set can't modify the other one by
73 #define TASK_RUNNING 0
74 #define TASK_INTERRUPTIBLE 1
75 #define TASK_UNINTERRUPTIBLE 2
76 #define __TASK_STOPPED 4
77 #define __TASK_TRACED 8
78 /* in tsk->exit_state */
80 #define EXIT_ZOMBIE 32
81 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
82 /* in tsk->state again */
84 #define TASK_WAKEKILL 128
85 #define TASK_WAKING 256
86 #define TASK_PARKED 512
87 #define TASK_NOLOAD 1024
89 #define TASK_STATE_MAX 4096
91 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
93 /* Convenience macros for the sake of set_current_state */
94 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
95 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
96 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
98 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
100 /* Convenience macros for the sake of wake_up */
101 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
102 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
104 /* get_task_state() */
105 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
106 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
107 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
109 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
110 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
111 #define task_is_stopped_or_traced(task) \
112 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
113 #define task_contributes_to_load(task) \
114 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
115 (task->flags & PF_FROZEN) == 0 && \
116 (task->state & TASK_NOLOAD) == 0)
118 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
120 #define __set_current_state(state_value) \
122 current->task_state_change = _THIS_IP_; \
123 current->state = (state_value); \
125 #define set_current_state(state_value) \
127 current->task_state_change = _THIS_IP_; \
128 smp_store_mb(current->state, (state_value)); \
133 * set_current_state() includes a barrier so that the write of current->state
134 * is correctly serialised wrt the caller's subsequent test of whether to
138 * set_current_state(TASK_UNINTERRUPTIBLE);
144 * __set_current_state(TASK_RUNNING);
146 * If the caller does not need such serialisation (because, for instance, the
147 * condition test and condition change and wakeup are under the same lock) then
148 * use __set_current_state().
150 * The above is typically ordered against the wakeup, which does:
152 * need_sleep = false;
153 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
155 * Where wake_up_state() (and all other wakeup primitives) imply enough
156 * barriers to order the store of the variable against wakeup.
158 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
159 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
160 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
162 * This is obviously fine, since they both store the exact same value.
164 * Also see the comments of try_to_wake_up().
166 #define __set_current_state(state_value) \
167 do { current->state = (state_value); } while (0)
168 #define set_current_state(state_value) \
169 smp_store_mb(current->state, (state_value))
173 /* Task command name length */
174 #define TASK_COMM_LEN 16
176 extern cpumask_var_t cpu_isolated_map;
178 extern int runqueue_is_locked(int cpu);
180 extern void scheduler_tick(void);
182 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
183 extern signed long schedule_timeout(signed long timeout);
184 extern signed long schedule_timeout_interruptible(signed long timeout);
185 extern signed long schedule_timeout_killable(signed long timeout);
186 extern signed long schedule_timeout_uninterruptible(signed long timeout);
187 extern signed long schedule_timeout_idle(signed long timeout);
188 asmlinkage void schedule(void);
189 extern void schedule_preempt_disabled(void);
191 extern int __must_check io_schedule_prepare(void);
192 extern void io_schedule_finish(int token);
193 extern long io_schedule_timeout(long timeout);
194 extern void io_schedule(void);
197 * struct prev_cputime - snaphsot of system and user cputime
198 * @utime: time spent in user mode
199 * @stime: time spent in system mode
200 * @lock: protects the above two fields
202 * Stores previous user/system time values such that we can guarantee
205 struct prev_cputime {
206 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
214 * struct task_cputime - collected CPU time counts
215 * @utime: time spent in user mode, in nanoseconds
216 * @stime: time spent in kernel mode, in nanoseconds
217 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
219 * This structure groups together three kinds of CPU time that are tracked for
220 * threads and thread groups. Most things considering CPU time want to group
221 * these counts together and treat all three of them in parallel.
223 struct task_cputime {
226 unsigned long long sum_exec_runtime;
229 /* Alternate field names when used to cache expirations. */
230 #define virt_exp utime
231 #define prof_exp stime
232 #define sched_exp sum_exec_runtime
234 #include <linux/rwsem.h>
236 #ifdef CONFIG_SCHED_INFO
238 /* cumulative counters */
239 unsigned long pcount; /* # of times run on this cpu */
240 unsigned long long run_delay; /* time spent waiting on a runqueue */
243 unsigned long long last_arrival,/* when we last ran on a cpu */
244 last_queued; /* when we were last queued to run */
246 #endif /* CONFIG_SCHED_INFO */
249 * Integer metrics need fixed point arithmetic, e.g., sched/fair
250 * has a few: load, load_avg, util_avg, freq, and capacity.
252 * We define a basic fixed point arithmetic range, and then formalize
253 * all these metrics based on that basic range.
255 # define SCHED_FIXEDPOINT_SHIFT 10
256 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
258 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
259 extern void prefetch_stack(struct task_struct *t);
261 static inline void prefetch_stack(struct task_struct *t) { }
265 unsigned long weight;
270 * The load_avg/util_avg accumulates an infinite geometric series
271 * (see __update_load_avg() in kernel/sched/fair.c).
273 * [load_avg definition]
275 * load_avg = runnable% * scale_load_down(load)
277 * where runnable% is the time ratio that a sched_entity is runnable.
278 * For cfs_rq, it is the aggregated load_avg of all runnable and
279 * blocked sched_entities.
281 * load_avg may also take frequency scaling into account:
283 * load_avg = runnable% * scale_load_down(load) * freq%
285 * where freq% is the CPU frequency normalized to the highest frequency.
287 * [util_avg definition]
289 * util_avg = running% * SCHED_CAPACITY_SCALE
291 * where running% is the time ratio that a sched_entity is running on
292 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
293 * and blocked sched_entities.
295 * util_avg may also factor frequency scaling and CPU capacity scaling:
297 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
299 * where freq% is the same as above, and capacity% is the CPU capacity
300 * normalized to the greatest capacity (due to uarch differences, etc).
302 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
303 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
304 * we therefore scale them to as large a range as necessary. This is for
305 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
309 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
310 * with the highest load (=88761), always runnable on a single cfs_rq,
311 * and should not overflow as the number already hits PID_MAX_LIMIT.
313 * For all other cases (including 32-bit kernels), struct load_weight's
314 * weight will overflow first before we do, because:
316 * Max(load_avg) <= Max(load.weight)
318 * Then it is the load_weight's responsibility to consider overflow
322 u64 last_update_time, load_sum;
323 u32 util_sum, period_contrib;
324 unsigned long load_avg, util_avg;
327 #ifdef CONFIG_SCHEDSTATS
328 struct sched_statistics {
338 s64 sum_sleep_runtime;
345 u64 nr_migrations_cold;
346 u64 nr_failed_migrations_affine;
347 u64 nr_failed_migrations_running;
348 u64 nr_failed_migrations_hot;
349 u64 nr_forced_migrations;
353 u64 nr_wakeups_migrate;
354 u64 nr_wakeups_local;
355 u64 nr_wakeups_remote;
356 u64 nr_wakeups_affine;
357 u64 nr_wakeups_affine_attempts;
358 u64 nr_wakeups_passive;
363 struct sched_entity {
364 struct load_weight load; /* for load-balancing */
365 struct rb_node run_node;
366 struct list_head group_node;
370 u64 sum_exec_runtime;
372 u64 prev_sum_exec_runtime;
376 #ifdef CONFIG_SCHEDSTATS
377 struct sched_statistics statistics;
380 #ifdef CONFIG_FAIR_GROUP_SCHED
382 struct sched_entity *parent;
383 /* rq on which this entity is (to be) queued: */
384 struct cfs_rq *cfs_rq;
385 /* rq "owned" by this entity/group: */
391 * Per entity load average tracking.
393 * Put into separate cache line so it does not
394 * collide with read-mostly values above.
396 struct sched_avg avg ____cacheline_aligned_in_smp;
400 struct sched_rt_entity {
401 struct list_head run_list;
402 unsigned long timeout;
403 unsigned long watchdog_stamp;
404 unsigned int time_slice;
405 unsigned short on_rq;
406 unsigned short on_list;
408 struct sched_rt_entity *back;
409 #ifdef CONFIG_RT_GROUP_SCHED
410 struct sched_rt_entity *parent;
411 /* rq on which this entity is (to be) queued: */
413 /* rq "owned" by this entity/group: */
418 struct sched_dl_entity {
419 struct rb_node rb_node;
422 * Original scheduling parameters. Copied here from sched_attr
423 * during sched_setattr(), they will remain the same until
424 * the next sched_setattr().
426 u64 dl_runtime; /* maximum runtime for each instance */
427 u64 dl_deadline; /* relative deadline of each instance */
428 u64 dl_period; /* separation of two instances (period) */
429 u64 dl_bw; /* dl_runtime / dl_deadline */
432 * Actual scheduling parameters. Initialized with the values above,
433 * they are continously updated during task execution. Note that
434 * the remaining runtime could be < 0 in case we are in overrun.
436 s64 runtime; /* remaining runtime for this instance */
437 u64 deadline; /* absolute deadline for this instance */
438 unsigned int flags; /* specifying the scheduler behaviour */
443 * @dl_throttled tells if we exhausted the runtime. If so, the
444 * task has to wait for a replenishment to be performed at the
445 * next firing of dl_timer.
447 * @dl_boosted tells if we are boosted due to DI. If so we are
448 * outside bandwidth enforcement mechanism (but only until we
449 * exit the critical section);
451 * @dl_yielded tells if task gave up the cpu before consuming
452 * all its available runtime during the last job.
454 int dl_throttled, dl_boosted, dl_yielded;
457 * Bandwidth enforcement timer. Each -deadline task has its
458 * own bandwidth to be enforced, thus we need one timer per task.
460 struct hrtimer dl_timer;
468 u8 pad; /* Otherwise the compiler can store garbage here. */
470 u32 s; /* Set of bits. */
473 enum perf_event_task_context {
474 perf_invalid_context = -1,
477 perf_nr_task_contexts,
481 struct wake_q_node *next;
485 #ifdef CONFIG_THREAD_INFO_IN_TASK
487 * For reasons of header soup (see current_thread_info()), this
488 * must be the first element of task_struct.
490 struct thread_info thread_info;
492 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
495 unsigned int flags; /* per process flags, defined below */
499 struct llist_node wake_entry;
501 #ifdef CONFIG_THREAD_INFO_IN_TASK
502 unsigned int cpu; /* current CPU */
504 unsigned int wakee_flips;
505 unsigned long wakee_flip_decay_ts;
506 struct task_struct *last_wakee;
512 int prio, static_prio, normal_prio;
513 unsigned int rt_priority;
514 const struct sched_class *sched_class;
515 struct sched_entity se;
516 struct sched_rt_entity rt;
517 #ifdef CONFIG_CGROUP_SCHED
518 struct task_group *sched_task_group;
520 struct sched_dl_entity dl;
522 #ifdef CONFIG_PREEMPT_NOTIFIERS
523 /* list of struct preempt_notifier: */
524 struct hlist_head preempt_notifiers;
527 #ifdef CONFIG_BLK_DEV_IO_TRACE
528 unsigned int btrace_seq;
533 cpumask_t cpus_allowed;
535 #ifdef CONFIG_PREEMPT_RCU
536 int rcu_read_lock_nesting;
537 union rcu_special rcu_read_unlock_special;
538 struct list_head rcu_node_entry;
539 struct rcu_node *rcu_blocked_node;
540 #endif /* #ifdef CONFIG_PREEMPT_RCU */
541 #ifdef CONFIG_TASKS_RCU
542 unsigned long rcu_tasks_nvcsw;
543 bool rcu_tasks_holdout;
544 struct list_head rcu_tasks_holdout_list;
545 int rcu_tasks_idle_cpu;
546 #endif /* #ifdef CONFIG_TASKS_RCU */
548 #ifdef CONFIG_SCHED_INFO
549 struct sched_info sched_info;
552 struct list_head tasks;
554 struct plist_node pushable_tasks;
555 struct rb_node pushable_dl_tasks;
558 struct mm_struct *mm, *active_mm;
560 /* Per-thread vma caching: */
561 struct vmacache vmacache;
563 #if defined(SPLIT_RSS_COUNTING)
564 struct task_rss_stat rss_stat;
568 int exit_code, exit_signal;
569 int pdeath_signal; /* The signal sent when the parent dies */
570 unsigned long jobctl; /* JOBCTL_*, siglock protected */
572 /* Used for emulating ABI behavior of previous Linux versions */
573 unsigned int personality;
575 /* scheduler bits, serialized by scheduler locks */
576 unsigned sched_reset_on_fork:1;
577 unsigned sched_contributes_to_load:1;
578 unsigned sched_migrated:1;
579 unsigned sched_remote_wakeup:1;
580 unsigned :0; /* force alignment to the next boundary */
582 /* unserialized, strictly 'current' */
583 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
584 unsigned in_iowait:1;
585 #if !defined(TIF_RESTORE_SIGMASK)
586 unsigned restore_sigmask:1;
589 unsigned memcg_may_oom:1;
591 unsigned memcg_kmem_skip_account:1;
594 #ifdef CONFIG_COMPAT_BRK
595 unsigned brk_randomized:1;
598 unsigned long atomic_flags; /* Flags needing atomic access. */
600 struct restart_block restart_block;
605 #ifdef CONFIG_CC_STACKPROTECTOR
606 /* Canary value for the -fstack-protector gcc feature */
607 unsigned long stack_canary;
610 * pointers to (original) parent process, youngest child, younger sibling,
611 * older sibling, respectively. (p->father can be replaced with
612 * p->real_parent->pid)
614 struct task_struct __rcu *real_parent; /* real parent process */
615 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
617 * children/sibling forms the list of my natural children
619 struct list_head children; /* list of my children */
620 struct list_head sibling; /* linkage in my parent's children list */
621 struct task_struct *group_leader; /* threadgroup leader */
624 * ptraced is the list of tasks this task is using ptrace on.
625 * This includes both natural children and PTRACE_ATTACH targets.
626 * p->ptrace_entry is p's link on the p->parent->ptraced list.
628 struct list_head ptraced;
629 struct list_head ptrace_entry;
631 /* PID/PID hash table linkage. */
632 struct pid_link pids[PIDTYPE_MAX];
633 struct list_head thread_group;
634 struct list_head thread_node;
636 struct completion *vfork_done; /* for vfork() */
637 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
638 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
641 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
642 u64 utimescaled, stimescaled;
645 struct prev_cputime prev_cputime;
646 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
647 seqcount_t vtime_seqcount;
648 unsigned long long vtime_snap;
650 /* Task is sleeping or running in a CPU with VTIME inactive */
652 /* Task runs in userspace in a CPU with VTIME active */
654 /* Task runs in kernelspace in a CPU with VTIME active */
659 #ifdef CONFIG_NO_HZ_FULL
660 atomic_t tick_dep_mask;
662 unsigned long nvcsw, nivcsw; /* context switch counts */
663 u64 start_time; /* monotonic time in nsec */
664 u64 real_start_time; /* boot based time in nsec */
665 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
666 unsigned long min_flt, maj_flt;
668 #ifdef CONFIG_POSIX_TIMERS
669 struct task_cputime cputime_expires;
670 struct list_head cpu_timers[3];
673 /* process credentials */
674 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
675 const struct cred __rcu *real_cred; /* objective and real subjective task
676 * credentials (COW) */
677 const struct cred __rcu *cred; /* effective (overridable) subjective task
678 * credentials (COW) */
679 char comm[TASK_COMM_LEN]; /* executable name excluding path
680 - access with [gs]et_task_comm (which lock
682 - initialized normally by setup_new_exec */
683 /* file system info */
684 struct nameidata *nameidata;
685 #ifdef CONFIG_SYSVIPC
687 struct sysv_sem sysvsem;
688 struct sysv_shm sysvshm;
690 #ifdef CONFIG_DETECT_HUNG_TASK
691 /* hung task detection */
692 unsigned long last_switch_count;
694 /* filesystem information */
695 struct fs_struct *fs;
696 /* open file information */
697 struct files_struct *files;
699 struct nsproxy *nsproxy;
700 /* signal handlers */
701 struct signal_struct *signal;
702 struct sighand_struct *sighand;
704 sigset_t blocked, real_blocked;
705 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
706 struct sigpending pending;
708 unsigned long sas_ss_sp;
710 unsigned sas_ss_flags;
712 struct callback_head *task_works;
714 struct audit_context *audit_context;
715 #ifdef CONFIG_AUDITSYSCALL
717 unsigned int sessionid;
719 struct seccomp seccomp;
721 /* Thread group tracking */
724 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
726 spinlock_t alloc_lock;
728 /* Protection of the PI data structures: */
729 raw_spinlock_t pi_lock;
731 struct wake_q_node wake_q;
733 #ifdef CONFIG_RT_MUTEXES
734 /* PI waiters blocked on a rt_mutex held by this task */
735 struct rb_root pi_waiters;
736 struct rb_node *pi_waiters_leftmost;
737 /* Deadlock detection and priority inheritance handling */
738 struct rt_mutex_waiter *pi_blocked_on;
741 #ifdef CONFIG_DEBUG_MUTEXES
742 /* mutex deadlock detection */
743 struct mutex_waiter *blocked_on;
745 #ifdef CONFIG_TRACE_IRQFLAGS
746 unsigned int irq_events;
747 unsigned long hardirq_enable_ip;
748 unsigned long hardirq_disable_ip;
749 unsigned int hardirq_enable_event;
750 unsigned int hardirq_disable_event;
751 int hardirqs_enabled;
753 unsigned long softirq_disable_ip;
754 unsigned long softirq_enable_ip;
755 unsigned int softirq_disable_event;
756 unsigned int softirq_enable_event;
757 int softirqs_enabled;
760 #ifdef CONFIG_LOCKDEP
761 # define MAX_LOCK_DEPTH 48UL
764 unsigned int lockdep_recursion;
765 struct held_lock held_locks[MAX_LOCK_DEPTH];
766 gfp_t lockdep_reclaim_gfp;
769 unsigned int in_ubsan;
772 /* journalling filesystem info */
775 /* stacked block device info */
776 struct bio_list *bio_list;
780 struct blk_plug *plug;
784 struct reclaim_state *reclaim_state;
786 struct backing_dev_info *backing_dev_info;
788 struct io_context *io_context;
790 unsigned long ptrace_message;
791 siginfo_t *last_siginfo; /* For ptrace use. */
792 struct task_io_accounting ioac;
793 #if defined(CONFIG_TASK_XACCT)
794 u64 acct_rss_mem1; /* accumulated rss usage */
795 u64 acct_vm_mem1; /* accumulated virtual memory usage */
796 u64 acct_timexpd; /* stime + utime since last update */
798 #ifdef CONFIG_CPUSETS
799 nodemask_t mems_allowed; /* Protected by alloc_lock */
800 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
801 int cpuset_mem_spread_rotor;
802 int cpuset_slab_spread_rotor;
804 #ifdef CONFIG_CGROUPS
805 /* Control Group info protected by css_set_lock */
806 struct css_set __rcu *cgroups;
807 /* cg_list protected by css_set_lock and tsk->alloc_lock */
808 struct list_head cg_list;
810 #ifdef CONFIG_INTEL_RDT_A
814 struct robust_list_head __user *robust_list;
816 struct compat_robust_list_head __user *compat_robust_list;
818 struct list_head pi_state_list;
819 struct futex_pi_state *pi_state_cache;
821 #ifdef CONFIG_PERF_EVENTS
822 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
823 struct mutex perf_event_mutex;
824 struct list_head perf_event_list;
826 #ifdef CONFIG_DEBUG_PREEMPT
827 unsigned long preempt_disable_ip;
830 struct mempolicy *mempolicy; /* Protected by alloc_lock */
832 short pref_node_fork;
834 #ifdef CONFIG_NUMA_BALANCING
836 unsigned int numa_scan_period;
837 unsigned int numa_scan_period_max;
838 int numa_preferred_nid;
839 unsigned long numa_migrate_retry;
840 u64 node_stamp; /* migration stamp */
841 u64 last_task_numa_placement;
842 u64 last_sum_exec_runtime;
843 struct callback_head numa_work;
845 struct list_head numa_entry;
846 struct numa_group *numa_group;
849 * numa_faults is an array split into four regions:
850 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
851 * in this precise order.
853 * faults_memory: Exponential decaying average of faults on a per-node
854 * basis. Scheduling placement decisions are made based on these
855 * counts. The values remain static for the duration of a PTE scan.
856 * faults_cpu: Track the nodes the process was running on when a NUMA
857 * hinting fault was incurred.
858 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
859 * during the current scan window. When the scan completes, the counts
860 * in faults_memory and faults_cpu decay and these values are copied.
862 unsigned long *numa_faults;
863 unsigned long total_numa_faults;
866 * numa_faults_locality tracks if faults recorded during the last
867 * scan window were remote/local or failed to migrate. The task scan
868 * period is adapted based on the locality of the faults with different
869 * weights depending on whether they were shared or private faults
871 unsigned long numa_faults_locality[3];
873 unsigned long numa_pages_migrated;
874 #endif /* CONFIG_NUMA_BALANCING */
876 struct tlbflush_unmap_batch tlb_ubc;
881 * cache last used pipe for splice
883 struct pipe_inode_info *splice_pipe;
885 struct page_frag task_frag;
887 #ifdef CONFIG_TASK_DELAY_ACCT
888 struct task_delay_info *delays;
891 #ifdef CONFIG_FAULT_INJECTION
895 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
896 * balance_dirty_pages() for some dirty throttling pause
899 int nr_dirtied_pause;
900 unsigned long dirty_paused_when; /* start of a write-and-pause period */
902 #ifdef CONFIG_LATENCYTOP
903 int latency_record_count;
904 struct latency_record latency_record[LT_SAVECOUNT];
907 * time slack values; these are used to round up poll() and
908 * select() etc timeout values. These are in nanoseconds.
911 u64 default_timer_slack_ns;
914 unsigned int kasan_depth;
916 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
917 /* Index of current stored address in ret_stack */
919 /* Stack of return addresses for return function tracing */
920 struct ftrace_ret_stack *ret_stack;
921 /* time stamp for last schedule */
922 unsigned long long ftrace_timestamp;
924 * Number of functions that haven't been traced
925 * because of depth overrun.
927 atomic_t trace_overrun;
928 /* Pause for the tracing */
929 atomic_t tracing_graph_pause;
931 #ifdef CONFIG_TRACING
932 /* state flags for use by tracers */
934 /* bitmask and counter of trace recursion */
935 unsigned long trace_recursion;
936 #endif /* CONFIG_TRACING */
938 /* Coverage collection mode enabled for this task (0 if disabled). */
939 enum kcov_mode kcov_mode;
940 /* Size of the kcov_area. */
942 /* Buffer for coverage collection. */
944 /* kcov desciptor wired with this task or NULL. */
948 struct mem_cgroup *memcg_in_oom;
949 gfp_t memcg_oom_gfp_mask;
952 /* number of pages to reclaim on returning to userland */
953 unsigned int memcg_nr_pages_over_high;
955 #ifdef CONFIG_UPROBES
956 struct uprobe_task *utask;
958 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
959 unsigned int sequential_io;
960 unsigned int sequential_io_avg;
962 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
963 unsigned long task_state_change;
965 int pagefault_disabled;
967 struct task_struct *oom_reaper_list;
969 #ifdef CONFIG_VMAP_STACK
970 struct vm_struct *stack_vm_area;
972 #ifdef CONFIG_THREAD_INFO_IN_TASK
973 /* A live task holds one reference. */
974 atomic_t stack_refcount;
976 /* CPU-specific state of this task */
977 struct thread_struct thread;
979 * WARNING: on x86, 'thread_struct' contains a variable-sized
980 * structure. It *MUST* be at the end of 'task_struct'.
982 * Do not put anything below here!
986 static inline struct pid *task_pid(struct task_struct *task)
988 return task->pids[PIDTYPE_PID].pid;
991 static inline struct pid *task_tgid(struct task_struct *task)
993 return task->group_leader->pids[PIDTYPE_PID].pid;
997 * Without tasklist or rcu lock it is not safe to dereference
998 * the result of task_pgrp/task_session even if task == current,
999 * we can race with another thread doing sys_setsid/sys_setpgid.
1001 static inline struct pid *task_pgrp(struct task_struct *task)
1003 return task->group_leader->pids[PIDTYPE_PGID].pid;
1006 static inline struct pid *task_session(struct task_struct *task)
1008 return task->group_leader->pids[PIDTYPE_SID].pid;
1012 * the helpers to get the task's different pids as they are seen
1013 * from various namespaces
1015 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1016 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1018 * task_xid_nr_ns() : id seen from the ns specified;
1020 * set_task_vxid() : assigns a virtual id to a task;
1022 * see also pid_nr() etc in include/linux/pid.h
1024 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1025 struct pid_namespace *ns);
1027 static inline pid_t task_pid_nr(struct task_struct *tsk)
1032 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1033 struct pid_namespace *ns)
1035 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1038 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1040 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1044 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1049 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1051 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1053 return pid_vnr(task_tgid(tsk));
1057 static inline int pid_alive(const struct task_struct *p);
1058 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1064 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1070 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1072 return task_ppid_nr_ns(tsk, &init_pid_ns);
1075 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1076 struct pid_namespace *ns)
1078 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1081 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1083 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1087 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1088 struct pid_namespace *ns)
1090 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1093 static inline pid_t task_session_vnr(struct task_struct *tsk)
1095 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1098 /* obsolete, do not use */
1099 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1101 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1105 * pid_alive - check that a task structure is not stale
1106 * @p: Task structure to be checked.
1108 * Test if a process is not yet dead (at most zombie state)
1109 * If pid_alive fails, then pointers within the task structure
1110 * can be stale and must not be dereferenced.
1112 * Return: 1 if the process is alive. 0 otherwise.
1114 static inline int pid_alive(const struct task_struct *p)
1116 return p->pids[PIDTYPE_PID].pid != NULL;
1120 * is_global_init - check if a task structure is init. Since init
1121 * is free to have sub-threads we need to check tgid.
1122 * @tsk: Task structure to be checked.
1124 * Check if a task structure is the first user space task the kernel created.
1126 * Return: 1 if the task structure is init. 0 otherwise.
1128 static inline int is_global_init(struct task_struct *tsk)
1130 return task_tgid_nr(tsk) == 1;
1133 extern struct pid *cad_pid;
1138 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1139 #define PF_EXITING 0x00000004 /* getting shut down */
1140 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1141 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1142 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1143 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1144 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1145 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1146 #define PF_DUMPCORE 0x00000200 /* dumped core */
1147 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1148 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1149 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1150 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1151 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1152 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1153 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1154 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1155 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1156 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1157 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1158 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1159 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1160 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1161 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1162 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1163 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1164 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1165 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1168 * Only the _current_ task can read/write to tsk->flags, but other
1169 * tasks can access tsk->flags in readonly mode for example
1170 * with tsk_used_math (like during threaded core dumping).
1171 * There is however an exception to this rule during ptrace
1172 * or during fork: the ptracer task is allowed to write to the
1173 * child->flags of its traced child (same goes for fork, the parent
1174 * can write to the child->flags), because we're guaranteed the
1175 * child is not running and in turn not changing child->flags
1176 * at the same time the parent does it.
1178 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1179 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1180 #define clear_used_math() clear_stopped_child_used_math(current)
1181 #define set_used_math() set_stopped_child_used_math(current)
1182 #define conditional_stopped_child_used_math(condition, child) \
1183 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1184 #define conditional_used_math(condition) \
1185 conditional_stopped_child_used_math(condition, current)
1186 #define copy_to_stopped_child_used_math(child) \
1187 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1188 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1189 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1190 #define used_math() tsk_used_math(current)
1192 /* Per-process atomic flags. */
1193 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1194 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1195 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1196 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1199 #define TASK_PFA_TEST(name, func) \
1200 static inline bool task_##func(struct task_struct *p) \
1201 { return test_bit(PFA_##name, &p->atomic_flags); }
1202 #define TASK_PFA_SET(name, func) \
1203 static inline void task_set_##func(struct task_struct *p) \
1204 { set_bit(PFA_##name, &p->atomic_flags); }
1205 #define TASK_PFA_CLEAR(name, func) \
1206 static inline void task_clear_##func(struct task_struct *p) \
1207 { clear_bit(PFA_##name, &p->atomic_flags); }
1209 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1210 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1212 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1213 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1214 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1216 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1217 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1218 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1220 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1221 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1223 static inline void tsk_restore_flags(struct task_struct *task,
1224 unsigned long orig_flags, unsigned long flags)
1226 task->flags &= ~flags;
1227 task->flags |= orig_flags & flags;
1230 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1231 const struct cpumask *trial);
1232 extern int task_can_attach(struct task_struct *p,
1233 const struct cpumask *cs_cpus_allowed);
1235 extern void do_set_cpus_allowed(struct task_struct *p,
1236 const struct cpumask *new_mask);
1238 extern int set_cpus_allowed_ptr(struct task_struct *p,
1239 const struct cpumask *new_mask);
1241 static inline void do_set_cpus_allowed(struct task_struct *p,
1242 const struct cpumask *new_mask)
1245 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1246 const struct cpumask *new_mask)
1248 if (!cpumask_test_cpu(0, new_mask))
1254 #ifndef cpu_relax_yield
1255 #define cpu_relax_yield() cpu_relax()
1258 extern int yield_to(struct task_struct *p, bool preempt);
1259 extern void set_user_nice(struct task_struct *p, long nice);
1260 extern int task_prio(const struct task_struct *p);
1262 * task_nice - return the nice value of a given task.
1263 * @p: the task in question.
1265 * Return: The nice value [ -20 ... 0 ... 19 ].
1267 static inline int task_nice(const struct task_struct *p)
1269 return PRIO_TO_NICE((p)->static_prio);
1271 extern int can_nice(const struct task_struct *p, const int nice);
1272 extern int task_curr(const struct task_struct *p);
1273 extern int idle_cpu(int cpu);
1274 extern int sched_setscheduler(struct task_struct *, int,
1275 const struct sched_param *);
1276 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1277 const struct sched_param *);
1278 extern int sched_setattr(struct task_struct *,
1279 const struct sched_attr *);
1280 extern struct task_struct *idle_task(int cpu);
1282 * is_idle_task - is the specified task an idle task?
1283 * @p: the task in question.
1285 * Return: 1 if @p is an idle task. 0 otherwise.
1287 static inline bool is_idle_task(const struct task_struct *p)
1289 return !!(p->flags & PF_IDLE);
1291 extern struct task_struct *curr_task(int cpu);
1292 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1296 union thread_union {
1297 #ifndef CONFIG_THREAD_INFO_IN_TASK
1298 struct thread_info thread_info;
1300 unsigned long stack[THREAD_SIZE/sizeof(long)];
1303 #ifdef CONFIG_THREAD_INFO_IN_TASK
1304 static inline struct thread_info *task_thread_info(struct task_struct *task)
1306 return &task->thread_info;
1308 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1309 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1312 extern struct pid_namespace init_pid_ns;
1315 * find a task by one of its numerical ids
1317 * find_task_by_pid_ns():
1318 * finds a task by its pid in the specified namespace
1319 * find_task_by_vpid():
1320 * finds a task by its virtual pid
1322 * see also find_vpid() etc in include/linux/pid.h
1325 extern struct task_struct *find_task_by_vpid(pid_t nr);
1326 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1327 struct pid_namespace *ns);
1329 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1330 extern int wake_up_process(struct task_struct *tsk);
1331 extern void wake_up_new_task(struct task_struct *tsk);
1333 extern void kick_process(struct task_struct *tsk);
1335 static inline void kick_process(struct task_struct *tsk) { }
1338 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1339 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1341 __set_task_comm(tsk, from, false);
1343 extern char *get_task_comm(char *to, struct task_struct *tsk);
1346 void scheduler_ipi(void);
1347 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1349 static inline void scheduler_ipi(void) { }
1350 static inline unsigned long wait_task_inactive(struct task_struct *p,
1357 /* set thread flags in other task's structures
1358 * - see asm/thread_info.h for TIF_xxxx flags available
1360 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1362 set_ti_thread_flag(task_thread_info(tsk), flag);
1365 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1367 clear_ti_thread_flag(task_thread_info(tsk), flag);
1370 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1372 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1375 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1377 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1380 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1382 return test_ti_thread_flag(task_thread_info(tsk), flag);
1385 static inline void set_tsk_need_resched(struct task_struct *tsk)
1387 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1390 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1392 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1395 static inline int test_tsk_need_resched(struct task_struct *tsk)
1397 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1401 * cond_resched() and cond_resched_lock(): latency reduction via
1402 * explicit rescheduling in places that are safe. The return
1403 * value indicates whether a reschedule was done in fact.
1404 * cond_resched_lock() will drop the spinlock before scheduling,
1405 * cond_resched_softirq() will enable bhs before scheduling.
1407 #ifndef CONFIG_PREEMPT
1408 extern int _cond_resched(void);
1410 static inline int _cond_resched(void) { return 0; }
1413 #define cond_resched() ({ \
1414 ___might_sleep(__FILE__, __LINE__, 0); \
1418 extern int __cond_resched_lock(spinlock_t *lock);
1420 #define cond_resched_lock(lock) ({ \
1421 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1422 __cond_resched_lock(lock); \
1425 extern int __cond_resched_softirq(void);
1427 #define cond_resched_softirq() ({ \
1428 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1429 __cond_resched_softirq(); \
1432 static inline void cond_resched_rcu(void)
1434 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1442 * Does a critical section need to be broken due to another
1443 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1444 * but a general need for low latency)
1446 static inline int spin_needbreak(spinlock_t *lock)
1448 #ifdef CONFIG_PREEMPT
1449 return spin_is_contended(lock);
1455 static __always_inline bool need_resched(void)
1457 return unlikely(tif_need_resched());
1461 * Wrappers for p->thread_info->cpu access. No-op on UP.
1465 static inline unsigned int task_cpu(const struct task_struct *p)
1467 #ifdef CONFIG_THREAD_INFO_IN_TASK
1470 return task_thread_info(p)->cpu;
1474 static inline int task_node(const struct task_struct *p)
1476 return cpu_to_node(task_cpu(p));
1479 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1483 static inline unsigned int task_cpu(const struct task_struct *p)
1488 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1492 #endif /* CONFIG_SMP */
1495 * In order to reduce various lock holder preemption latencies provide an
1496 * interface to see if a vCPU is currently running or not.
1498 * This allows us to terminate optimistic spin loops and block, analogous to
1499 * the native optimistic spin heuristic of testing if the lock owner task is
1502 #ifndef vcpu_is_preempted
1503 # define vcpu_is_preempted(cpu) false
1506 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1507 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1509 #ifndef TASK_SIZE_OF
1510 #define TASK_SIZE_OF(tsk) TASK_SIZE