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>
12 #include <linux/sem.h>
13 #include <linux/shm.h>
14 #include <linux/signal.h>
15 #include <linux/signal_types.h>
16 #include <linux/pid.h>
17 #include <linux/seccomp.h>
18 #include <linux/rculist.h>
20 #include <linux/resource.h>
21 #include <linux/hrtimer.h>
22 #include <linux/kcov.h>
23 #include <linux/task_io_accounting.h>
24 #include <linux/latencytop.h>
25 #include <linux/topology.h>
26 #include <linux/magic.h>
28 #include <asm/current.h>
30 /* task_struct member predeclarations: */
33 struct backing_dev_info;
39 struct futex_pi_state;
44 struct perf_event_context;
46 struct pipe_inode_info;
49 struct robust_list_head;
53 struct sighand_struct;
55 struct task_delay_info;
61 * Task state bitmask. NOTE! These bits are also
62 * encoded in fs/proc/array.c: get_task_state().
64 * We have two separate sets of flags: task->state
65 * is about runnability, while task->exit_state are
66 * about the task exiting. Confusing, but this way
67 * modifying one set can't modify the other one by
70 #define TASK_RUNNING 0
71 #define TASK_INTERRUPTIBLE 1
72 #define TASK_UNINTERRUPTIBLE 2
73 #define __TASK_STOPPED 4
74 #define __TASK_TRACED 8
75 /* in tsk->exit_state */
77 #define EXIT_ZOMBIE 32
78 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
79 /* in tsk->state again */
81 #define TASK_WAKEKILL 128
82 #define TASK_WAKING 256
83 #define TASK_PARKED 512
84 #define TASK_NOLOAD 1024
86 #define TASK_STATE_MAX 4096
88 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
90 /* Convenience macros for the sake of set_current_state */
91 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
92 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
93 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
95 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
97 /* Convenience macros for the sake of wake_up */
98 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
99 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
101 /* get_task_state() */
102 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
103 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
104 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
106 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
107 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
108 #define task_is_stopped_or_traced(task) \
109 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
110 #define task_contributes_to_load(task) \
111 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
112 (task->flags & PF_FROZEN) == 0 && \
113 (task->state & TASK_NOLOAD) == 0)
115 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
117 #define __set_current_state(state_value) \
119 current->task_state_change = _THIS_IP_; \
120 current->state = (state_value); \
122 #define set_current_state(state_value) \
124 current->task_state_change = _THIS_IP_; \
125 smp_store_mb(current->state, (state_value)); \
130 * set_current_state() includes a barrier so that the write of current->state
131 * is correctly serialised wrt the caller's subsequent test of whether to
135 * set_current_state(TASK_UNINTERRUPTIBLE);
141 * __set_current_state(TASK_RUNNING);
143 * If the caller does not need such serialisation (because, for instance, the
144 * condition test and condition change and wakeup are under the same lock) then
145 * use __set_current_state().
147 * The above is typically ordered against the wakeup, which does:
149 * need_sleep = false;
150 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
152 * Where wake_up_state() (and all other wakeup primitives) imply enough
153 * barriers to order the store of the variable against wakeup.
155 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
156 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
157 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
159 * This is obviously fine, since they both store the exact same value.
161 * Also see the comments of try_to_wake_up().
163 #define __set_current_state(state_value) \
164 do { current->state = (state_value); } while (0)
165 #define set_current_state(state_value) \
166 smp_store_mb(current->state, (state_value))
170 /* Task command name length */
171 #define TASK_COMM_LEN 16
173 extern cpumask_var_t cpu_isolated_map;
175 extern void scheduler_tick(void);
177 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
178 extern signed long schedule_timeout(signed long timeout);
179 extern signed long schedule_timeout_interruptible(signed long timeout);
180 extern signed long schedule_timeout_killable(signed long timeout);
181 extern signed long schedule_timeout_uninterruptible(signed long timeout);
182 extern signed long schedule_timeout_idle(signed long timeout);
183 asmlinkage void schedule(void);
184 extern void schedule_preempt_disabled(void);
186 extern int __must_check io_schedule_prepare(void);
187 extern void io_schedule_finish(int token);
188 extern long io_schedule_timeout(long timeout);
189 extern void io_schedule(void);
192 * struct prev_cputime - snaphsot of system and user cputime
193 * @utime: time spent in user mode
194 * @stime: time spent in system mode
195 * @lock: protects the above two fields
197 * Stores previous user/system time values such that we can guarantee
200 struct prev_cputime {
201 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
209 * struct task_cputime - collected CPU time counts
210 * @utime: time spent in user mode, in nanoseconds
211 * @stime: time spent in kernel mode, in nanoseconds
212 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
214 * This structure groups together three kinds of CPU time that are tracked for
215 * threads and thread groups. Most things considering CPU time want to group
216 * these counts together and treat all three of them in parallel.
218 struct task_cputime {
221 unsigned long long sum_exec_runtime;
224 /* Alternate field names when used to cache expirations. */
225 #define virt_exp utime
226 #define prof_exp stime
227 #define sched_exp sum_exec_runtime
229 #ifdef CONFIG_SCHED_INFO
231 /* cumulative counters */
232 unsigned long pcount; /* # of times run on this cpu */
233 unsigned long long run_delay; /* time spent waiting on a runqueue */
236 unsigned long long last_arrival,/* when we last ran on a cpu */
237 last_queued; /* when we were last queued to run */
239 #endif /* CONFIG_SCHED_INFO */
242 * Integer metrics need fixed point arithmetic, e.g., sched/fair
243 * has a few: load, load_avg, util_avg, freq, and capacity.
245 * We define a basic fixed point arithmetic range, and then formalize
246 * all these metrics based on that basic range.
248 # define SCHED_FIXEDPOINT_SHIFT 10
249 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
251 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
252 extern void prefetch_stack(struct task_struct *t);
254 static inline void prefetch_stack(struct task_struct *t) { }
258 unsigned long weight;
263 * The load_avg/util_avg accumulates an infinite geometric series
264 * (see __update_load_avg() in kernel/sched/fair.c).
266 * [load_avg definition]
268 * load_avg = runnable% * scale_load_down(load)
270 * where runnable% is the time ratio that a sched_entity is runnable.
271 * For cfs_rq, it is the aggregated load_avg of all runnable and
272 * blocked sched_entities.
274 * load_avg may also take frequency scaling into account:
276 * load_avg = runnable% * scale_load_down(load) * freq%
278 * where freq% is the CPU frequency normalized to the highest frequency.
280 * [util_avg definition]
282 * util_avg = running% * SCHED_CAPACITY_SCALE
284 * where running% is the time ratio that a sched_entity is running on
285 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
286 * and blocked sched_entities.
288 * util_avg may also factor frequency scaling and CPU capacity scaling:
290 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
292 * where freq% is the same as above, and capacity% is the CPU capacity
293 * normalized to the greatest capacity (due to uarch differences, etc).
295 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
296 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
297 * we therefore scale them to as large a range as necessary. This is for
298 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
302 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
303 * with the highest load (=88761), always runnable on a single cfs_rq,
304 * and should not overflow as the number already hits PID_MAX_LIMIT.
306 * For all other cases (including 32-bit kernels), struct load_weight's
307 * weight will overflow first before we do, because:
309 * Max(load_avg) <= Max(load.weight)
311 * Then it is the load_weight's responsibility to consider overflow
315 u64 last_update_time, load_sum;
316 u32 util_sum, period_contrib;
317 unsigned long load_avg, util_avg;
320 #ifdef CONFIG_SCHEDSTATS
321 struct sched_statistics {
331 s64 sum_sleep_runtime;
338 u64 nr_migrations_cold;
339 u64 nr_failed_migrations_affine;
340 u64 nr_failed_migrations_running;
341 u64 nr_failed_migrations_hot;
342 u64 nr_forced_migrations;
346 u64 nr_wakeups_migrate;
347 u64 nr_wakeups_local;
348 u64 nr_wakeups_remote;
349 u64 nr_wakeups_affine;
350 u64 nr_wakeups_affine_attempts;
351 u64 nr_wakeups_passive;
356 struct sched_entity {
357 struct load_weight load; /* for load-balancing */
358 struct rb_node run_node;
359 struct list_head group_node;
363 u64 sum_exec_runtime;
365 u64 prev_sum_exec_runtime;
369 #ifdef CONFIG_SCHEDSTATS
370 struct sched_statistics statistics;
373 #ifdef CONFIG_FAIR_GROUP_SCHED
375 struct sched_entity *parent;
376 /* rq on which this entity is (to be) queued: */
377 struct cfs_rq *cfs_rq;
378 /* rq "owned" by this entity/group: */
384 * Per entity load average tracking.
386 * Put into separate cache line so it does not
387 * collide with read-mostly values above.
389 struct sched_avg avg ____cacheline_aligned_in_smp;
393 struct sched_rt_entity {
394 struct list_head run_list;
395 unsigned long timeout;
396 unsigned long watchdog_stamp;
397 unsigned int time_slice;
398 unsigned short on_rq;
399 unsigned short on_list;
401 struct sched_rt_entity *back;
402 #ifdef CONFIG_RT_GROUP_SCHED
403 struct sched_rt_entity *parent;
404 /* rq on which this entity is (to be) queued: */
406 /* rq "owned" by this entity/group: */
411 struct sched_dl_entity {
412 struct rb_node rb_node;
415 * Original scheduling parameters. Copied here from sched_attr
416 * during sched_setattr(), they will remain the same until
417 * the next sched_setattr().
419 u64 dl_runtime; /* maximum runtime for each instance */
420 u64 dl_deadline; /* relative deadline of each instance */
421 u64 dl_period; /* separation of two instances (period) */
422 u64 dl_bw; /* dl_runtime / dl_deadline */
425 * Actual scheduling parameters. Initialized with the values above,
426 * they are continously updated during task execution. Note that
427 * the remaining runtime could be < 0 in case we are in overrun.
429 s64 runtime; /* remaining runtime for this instance */
430 u64 deadline; /* absolute deadline for this instance */
431 unsigned int flags; /* specifying the scheduler behaviour */
436 * @dl_throttled tells if we exhausted the runtime. If so, the
437 * task has to wait for a replenishment to be performed at the
438 * next firing of dl_timer.
440 * @dl_boosted tells if we are boosted due to DI. If so we are
441 * outside bandwidth enforcement mechanism (but only until we
442 * exit the critical section);
444 * @dl_yielded tells if task gave up the cpu before consuming
445 * all its available runtime during the last job.
447 int dl_throttled, dl_boosted, dl_yielded;
450 * Bandwidth enforcement timer. Each -deadline task has its
451 * own bandwidth to be enforced, thus we need one timer per task.
453 struct hrtimer dl_timer;
461 u8 pad; /* Otherwise the compiler can store garbage here. */
463 u32 s; /* Set of bits. */
466 enum perf_event_task_context {
467 perf_invalid_context = -1,
470 perf_nr_task_contexts,
474 struct wake_q_node *next;
478 #ifdef CONFIG_THREAD_INFO_IN_TASK
480 * For reasons of header soup (see current_thread_info()), this
481 * must be the first element of task_struct.
483 struct thread_info thread_info;
485 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
488 unsigned int flags; /* per process flags, defined below */
492 struct llist_node wake_entry;
494 #ifdef CONFIG_THREAD_INFO_IN_TASK
495 unsigned int cpu; /* current CPU */
497 unsigned int wakee_flips;
498 unsigned long wakee_flip_decay_ts;
499 struct task_struct *last_wakee;
505 int prio, static_prio, normal_prio;
506 unsigned int rt_priority;
507 const struct sched_class *sched_class;
508 struct sched_entity se;
509 struct sched_rt_entity rt;
510 #ifdef CONFIG_CGROUP_SCHED
511 struct task_group *sched_task_group;
513 struct sched_dl_entity dl;
515 #ifdef CONFIG_PREEMPT_NOTIFIERS
516 /* list of struct preempt_notifier: */
517 struct hlist_head preempt_notifiers;
520 #ifdef CONFIG_BLK_DEV_IO_TRACE
521 unsigned int btrace_seq;
526 cpumask_t cpus_allowed;
528 #ifdef CONFIG_PREEMPT_RCU
529 int rcu_read_lock_nesting;
530 union rcu_special rcu_read_unlock_special;
531 struct list_head rcu_node_entry;
532 struct rcu_node *rcu_blocked_node;
533 #endif /* #ifdef CONFIG_PREEMPT_RCU */
534 #ifdef CONFIG_TASKS_RCU
535 unsigned long rcu_tasks_nvcsw;
536 bool rcu_tasks_holdout;
537 struct list_head rcu_tasks_holdout_list;
538 int rcu_tasks_idle_cpu;
539 #endif /* #ifdef CONFIG_TASKS_RCU */
541 #ifdef CONFIG_SCHED_INFO
542 struct sched_info sched_info;
545 struct list_head tasks;
547 struct plist_node pushable_tasks;
548 struct rb_node pushable_dl_tasks;
551 struct mm_struct *mm, *active_mm;
553 /* Per-thread vma caching: */
554 struct vmacache vmacache;
556 #if defined(SPLIT_RSS_COUNTING)
557 struct task_rss_stat rss_stat;
561 int exit_code, exit_signal;
562 int pdeath_signal; /* The signal sent when the parent dies */
563 unsigned long jobctl; /* JOBCTL_*, siglock protected */
565 /* Used for emulating ABI behavior of previous Linux versions */
566 unsigned int personality;
568 /* scheduler bits, serialized by scheduler locks */
569 unsigned sched_reset_on_fork:1;
570 unsigned sched_contributes_to_load:1;
571 unsigned sched_migrated:1;
572 unsigned sched_remote_wakeup:1;
573 unsigned :0; /* force alignment to the next boundary */
575 /* unserialized, strictly 'current' */
576 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
577 unsigned in_iowait:1;
578 #if !defined(TIF_RESTORE_SIGMASK)
579 unsigned restore_sigmask:1;
582 unsigned memcg_may_oom:1;
584 unsigned memcg_kmem_skip_account:1;
587 #ifdef CONFIG_COMPAT_BRK
588 unsigned brk_randomized:1;
591 unsigned long atomic_flags; /* Flags needing atomic access. */
593 struct restart_block restart_block;
598 #ifdef CONFIG_CC_STACKPROTECTOR
599 /* Canary value for the -fstack-protector gcc feature */
600 unsigned long stack_canary;
603 * pointers to (original) parent process, youngest child, younger sibling,
604 * older sibling, respectively. (p->father can be replaced with
605 * p->real_parent->pid)
607 struct task_struct __rcu *real_parent; /* real parent process */
608 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
610 * children/sibling forms the list of my natural children
612 struct list_head children; /* list of my children */
613 struct list_head sibling; /* linkage in my parent's children list */
614 struct task_struct *group_leader; /* threadgroup leader */
617 * ptraced is the list of tasks this task is using ptrace on.
618 * This includes both natural children and PTRACE_ATTACH targets.
619 * p->ptrace_entry is p's link on the p->parent->ptraced list.
621 struct list_head ptraced;
622 struct list_head ptrace_entry;
624 /* PID/PID hash table linkage. */
625 struct pid_link pids[PIDTYPE_MAX];
626 struct list_head thread_group;
627 struct list_head thread_node;
629 struct completion *vfork_done; /* for vfork() */
630 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
631 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
634 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
635 u64 utimescaled, stimescaled;
638 struct prev_cputime prev_cputime;
639 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
640 seqcount_t vtime_seqcount;
641 unsigned long long vtime_snap;
643 /* Task is sleeping or running in a CPU with VTIME inactive */
645 /* Task runs in userspace in a CPU with VTIME active */
647 /* Task runs in kernelspace in a CPU with VTIME active */
652 #ifdef CONFIG_NO_HZ_FULL
653 atomic_t tick_dep_mask;
655 unsigned long nvcsw, nivcsw; /* context switch counts */
656 u64 start_time; /* monotonic time in nsec */
657 u64 real_start_time; /* boot based time in nsec */
658 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
659 unsigned long min_flt, maj_flt;
661 #ifdef CONFIG_POSIX_TIMERS
662 struct task_cputime cputime_expires;
663 struct list_head cpu_timers[3];
666 /* process credentials */
667 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
668 const struct cred __rcu *real_cred; /* objective and real subjective task
669 * credentials (COW) */
670 const struct cred __rcu *cred; /* effective (overridable) subjective task
671 * credentials (COW) */
672 char comm[TASK_COMM_LEN]; /* executable name excluding path
673 - access with [gs]et_task_comm (which lock
675 - initialized normally by setup_new_exec */
676 /* file system info */
677 struct nameidata *nameidata;
678 #ifdef CONFIG_SYSVIPC
680 struct sysv_sem sysvsem;
681 struct sysv_shm sysvshm;
683 #ifdef CONFIG_DETECT_HUNG_TASK
684 /* hung task detection */
685 unsigned long last_switch_count;
687 /* filesystem information */
688 struct fs_struct *fs;
689 /* open file information */
690 struct files_struct *files;
692 struct nsproxy *nsproxy;
693 /* signal handlers */
694 struct signal_struct *signal;
695 struct sighand_struct *sighand;
697 sigset_t blocked, real_blocked;
698 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
699 struct sigpending pending;
701 unsigned long sas_ss_sp;
703 unsigned sas_ss_flags;
705 struct callback_head *task_works;
707 struct audit_context *audit_context;
708 #ifdef CONFIG_AUDITSYSCALL
710 unsigned int sessionid;
712 struct seccomp seccomp;
714 /* Thread group tracking */
717 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
719 spinlock_t alloc_lock;
721 /* Protection of the PI data structures: */
722 raw_spinlock_t pi_lock;
724 struct wake_q_node wake_q;
726 #ifdef CONFIG_RT_MUTEXES
727 /* PI waiters blocked on a rt_mutex held by this task */
728 struct rb_root pi_waiters;
729 struct rb_node *pi_waiters_leftmost;
730 /* Deadlock detection and priority inheritance handling */
731 struct rt_mutex_waiter *pi_blocked_on;
734 #ifdef CONFIG_DEBUG_MUTEXES
735 /* mutex deadlock detection */
736 struct mutex_waiter *blocked_on;
738 #ifdef CONFIG_TRACE_IRQFLAGS
739 unsigned int irq_events;
740 unsigned long hardirq_enable_ip;
741 unsigned long hardirq_disable_ip;
742 unsigned int hardirq_enable_event;
743 unsigned int hardirq_disable_event;
744 int hardirqs_enabled;
746 unsigned long softirq_disable_ip;
747 unsigned long softirq_enable_ip;
748 unsigned int softirq_disable_event;
749 unsigned int softirq_enable_event;
750 int softirqs_enabled;
753 #ifdef CONFIG_LOCKDEP
754 # define MAX_LOCK_DEPTH 48UL
757 unsigned int lockdep_recursion;
758 struct held_lock held_locks[MAX_LOCK_DEPTH];
759 gfp_t lockdep_reclaim_gfp;
762 unsigned int in_ubsan;
765 /* journalling filesystem info */
768 /* stacked block device info */
769 struct bio_list *bio_list;
773 struct blk_plug *plug;
777 struct reclaim_state *reclaim_state;
779 struct backing_dev_info *backing_dev_info;
781 struct io_context *io_context;
783 unsigned long ptrace_message;
784 siginfo_t *last_siginfo; /* For ptrace use. */
785 struct task_io_accounting ioac;
786 #if defined(CONFIG_TASK_XACCT)
787 u64 acct_rss_mem1; /* accumulated rss usage */
788 u64 acct_vm_mem1; /* accumulated virtual memory usage */
789 u64 acct_timexpd; /* stime + utime since last update */
791 #ifdef CONFIG_CPUSETS
792 nodemask_t mems_allowed; /* Protected by alloc_lock */
793 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
794 int cpuset_mem_spread_rotor;
795 int cpuset_slab_spread_rotor;
797 #ifdef CONFIG_CGROUPS
798 /* Control Group info protected by css_set_lock */
799 struct css_set __rcu *cgroups;
800 /* cg_list protected by css_set_lock and tsk->alloc_lock */
801 struct list_head cg_list;
803 #ifdef CONFIG_INTEL_RDT_A
807 struct robust_list_head __user *robust_list;
809 struct compat_robust_list_head __user *compat_robust_list;
811 struct list_head pi_state_list;
812 struct futex_pi_state *pi_state_cache;
814 #ifdef CONFIG_PERF_EVENTS
815 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
816 struct mutex perf_event_mutex;
817 struct list_head perf_event_list;
819 #ifdef CONFIG_DEBUG_PREEMPT
820 unsigned long preempt_disable_ip;
823 struct mempolicy *mempolicy; /* Protected by alloc_lock */
825 short pref_node_fork;
827 #ifdef CONFIG_NUMA_BALANCING
829 unsigned int numa_scan_period;
830 unsigned int numa_scan_period_max;
831 int numa_preferred_nid;
832 unsigned long numa_migrate_retry;
833 u64 node_stamp; /* migration stamp */
834 u64 last_task_numa_placement;
835 u64 last_sum_exec_runtime;
836 struct callback_head numa_work;
838 struct list_head numa_entry;
839 struct numa_group *numa_group;
842 * numa_faults is an array split into four regions:
843 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
844 * in this precise order.
846 * faults_memory: Exponential decaying average of faults on a per-node
847 * basis. Scheduling placement decisions are made based on these
848 * counts. The values remain static for the duration of a PTE scan.
849 * faults_cpu: Track the nodes the process was running on when a NUMA
850 * hinting fault was incurred.
851 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
852 * during the current scan window. When the scan completes, the counts
853 * in faults_memory and faults_cpu decay and these values are copied.
855 unsigned long *numa_faults;
856 unsigned long total_numa_faults;
859 * numa_faults_locality tracks if faults recorded during the last
860 * scan window were remote/local or failed to migrate. The task scan
861 * period is adapted based on the locality of the faults with different
862 * weights depending on whether they were shared or private faults
864 unsigned long numa_faults_locality[3];
866 unsigned long numa_pages_migrated;
867 #endif /* CONFIG_NUMA_BALANCING */
869 struct tlbflush_unmap_batch tlb_ubc;
874 * cache last used pipe for splice
876 struct pipe_inode_info *splice_pipe;
878 struct page_frag task_frag;
880 #ifdef CONFIG_TASK_DELAY_ACCT
881 struct task_delay_info *delays;
884 #ifdef CONFIG_FAULT_INJECTION
888 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
889 * balance_dirty_pages() for some dirty throttling pause
892 int nr_dirtied_pause;
893 unsigned long dirty_paused_when; /* start of a write-and-pause period */
895 #ifdef CONFIG_LATENCYTOP
896 int latency_record_count;
897 struct latency_record latency_record[LT_SAVECOUNT];
900 * time slack values; these are used to round up poll() and
901 * select() etc timeout values. These are in nanoseconds.
904 u64 default_timer_slack_ns;
907 unsigned int kasan_depth;
909 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
910 /* Index of current stored address in ret_stack */
912 /* Stack of return addresses for return function tracing */
913 struct ftrace_ret_stack *ret_stack;
914 /* time stamp for last schedule */
915 unsigned long long ftrace_timestamp;
917 * Number of functions that haven't been traced
918 * because of depth overrun.
920 atomic_t trace_overrun;
921 /* Pause for the tracing */
922 atomic_t tracing_graph_pause;
924 #ifdef CONFIG_TRACING
925 /* state flags for use by tracers */
927 /* bitmask and counter of trace recursion */
928 unsigned long trace_recursion;
929 #endif /* CONFIG_TRACING */
931 /* Coverage collection mode enabled for this task (0 if disabled). */
932 enum kcov_mode kcov_mode;
933 /* Size of the kcov_area. */
935 /* Buffer for coverage collection. */
937 /* kcov desciptor wired with this task or NULL. */
941 struct mem_cgroup *memcg_in_oom;
942 gfp_t memcg_oom_gfp_mask;
945 /* number of pages to reclaim on returning to userland */
946 unsigned int memcg_nr_pages_over_high;
948 #ifdef CONFIG_UPROBES
949 struct uprobe_task *utask;
951 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
952 unsigned int sequential_io;
953 unsigned int sequential_io_avg;
955 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
956 unsigned long task_state_change;
958 int pagefault_disabled;
960 struct task_struct *oom_reaper_list;
962 #ifdef CONFIG_VMAP_STACK
963 struct vm_struct *stack_vm_area;
965 #ifdef CONFIG_THREAD_INFO_IN_TASK
966 /* A live task holds one reference. */
967 atomic_t stack_refcount;
969 /* CPU-specific state of this task */
970 struct thread_struct thread;
972 * WARNING: on x86, 'thread_struct' contains a variable-sized
973 * structure. It *MUST* be at the end of 'task_struct'.
975 * Do not put anything below here!
979 static inline struct pid *task_pid(struct task_struct *task)
981 return task->pids[PIDTYPE_PID].pid;
984 static inline struct pid *task_tgid(struct task_struct *task)
986 return task->group_leader->pids[PIDTYPE_PID].pid;
990 * Without tasklist or rcu lock it is not safe to dereference
991 * the result of task_pgrp/task_session even if task == current,
992 * we can race with another thread doing sys_setsid/sys_setpgid.
994 static inline struct pid *task_pgrp(struct task_struct *task)
996 return task->group_leader->pids[PIDTYPE_PGID].pid;
999 static inline struct pid *task_session(struct task_struct *task)
1001 return task->group_leader->pids[PIDTYPE_SID].pid;
1005 * the helpers to get the task's different pids as they are seen
1006 * from various namespaces
1008 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1009 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1011 * task_xid_nr_ns() : id seen from the ns specified;
1013 * set_task_vxid() : assigns a virtual id to a task;
1015 * see also pid_nr() etc in include/linux/pid.h
1017 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1018 struct pid_namespace *ns);
1020 static inline pid_t task_pid_nr(struct task_struct *tsk)
1025 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1026 struct pid_namespace *ns)
1028 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1031 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1033 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1037 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1042 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1044 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1046 return pid_vnr(task_tgid(tsk));
1050 static inline int pid_alive(const struct task_struct *p);
1051 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1057 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1063 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1065 return task_ppid_nr_ns(tsk, &init_pid_ns);
1068 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1069 struct pid_namespace *ns)
1071 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1074 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1076 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1080 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1081 struct pid_namespace *ns)
1083 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1086 static inline pid_t task_session_vnr(struct task_struct *tsk)
1088 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1091 /* obsolete, do not use */
1092 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1094 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1098 * pid_alive - check that a task structure is not stale
1099 * @p: Task structure to be checked.
1101 * Test if a process is not yet dead (at most zombie state)
1102 * If pid_alive fails, then pointers within the task structure
1103 * can be stale and must not be dereferenced.
1105 * Return: 1 if the process is alive. 0 otherwise.
1107 static inline int pid_alive(const struct task_struct *p)
1109 return p->pids[PIDTYPE_PID].pid != NULL;
1113 * is_global_init - check if a task structure is init. Since init
1114 * is free to have sub-threads we need to check tgid.
1115 * @tsk: Task structure to be checked.
1117 * Check if a task structure is the first user space task the kernel created.
1119 * Return: 1 if the task structure is init. 0 otherwise.
1121 static inline int is_global_init(struct task_struct *tsk)
1123 return task_tgid_nr(tsk) == 1;
1126 extern struct pid *cad_pid;
1131 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1132 #define PF_EXITING 0x00000004 /* getting shut down */
1133 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1134 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1135 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1136 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1137 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1138 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1139 #define PF_DUMPCORE 0x00000200 /* dumped core */
1140 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1141 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1142 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1143 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1144 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1145 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1146 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1147 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1148 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1149 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1150 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1151 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1152 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1153 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1154 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1155 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1156 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1157 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1158 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1161 * Only the _current_ task can read/write to tsk->flags, but other
1162 * tasks can access tsk->flags in readonly mode for example
1163 * with tsk_used_math (like during threaded core dumping).
1164 * There is however an exception to this rule during ptrace
1165 * or during fork: the ptracer task is allowed to write to the
1166 * child->flags of its traced child (same goes for fork, the parent
1167 * can write to the child->flags), because we're guaranteed the
1168 * child is not running and in turn not changing child->flags
1169 * at the same time the parent does it.
1171 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1172 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1173 #define clear_used_math() clear_stopped_child_used_math(current)
1174 #define set_used_math() set_stopped_child_used_math(current)
1175 #define conditional_stopped_child_used_math(condition, child) \
1176 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1177 #define conditional_used_math(condition) \
1178 conditional_stopped_child_used_math(condition, current)
1179 #define copy_to_stopped_child_used_math(child) \
1180 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1181 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1182 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1183 #define used_math() tsk_used_math(current)
1185 /* Per-process atomic flags. */
1186 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1187 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1188 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1189 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1192 #define TASK_PFA_TEST(name, func) \
1193 static inline bool task_##func(struct task_struct *p) \
1194 { return test_bit(PFA_##name, &p->atomic_flags); }
1195 #define TASK_PFA_SET(name, func) \
1196 static inline void task_set_##func(struct task_struct *p) \
1197 { set_bit(PFA_##name, &p->atomic_flags); }
1198 #define TASK_PFA_CLEAR(name, func) \
1199 static inline void task_clear_##func(struct task_struct *p) \
1200 { clear_bit(PFA_##name, &p->atomic_flags); }
1202 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1203 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1205 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1206 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1207 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1209 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1210 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1211 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1213 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1214 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1216 static inline void tsk_restore_flags(struct task_struct *task,
1217 unsigned long orig_flags, unsigned long flags)
1219 task->flags &= ~flags;
1220 task->flags |= orig_flags & flags;
1223 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1224 const struct cpumask *trial);
1225 extern int task_can_attach(struct task_struct *p,
1226 const struct cpumask *cs_cpus_allowed);
1228 extern void do_set_cpus_allowed(struct task_struct *p,
1229 const struct cpumask *new_mask);
1231 extern int set_cpus_allowed_ptr(struct task_struct *p,
1232 const struct cpumask *new_mask);
1234 static inline void do_set_cpus_allowed(struct task_struct *p,
1235 const struct cpumask *new_mask)
1238 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1239 const struct cpumask *new_mask)
1241 if (!cpumask_test_cpu(0, new_mask))
1247 #ifndef cpu_relax_yield
1248 #define cpu_relax_yield() cpu_relax()
1251 extern int yield_to(struct task_struct *p, bool preempt);
1252 extern void set_user_nice(struct task_struct *p, long nice);
1253 extern int task_prio(const struct task_struct *p);
1255 * task_nice - return the nice value of a given task.
1256 * @p: the task in question.
1258 * Return: The nice value [ -20 ... 0 ... 19 ].
1260 static inline int task_nice(const struct task_struct *p)
1262 return PRIO_TO_NICE((p)->static_prio);
1264 extern int can_nice(const struct task_struct *p, const int nice);
1265 extern int task_curr(const struct task_struct *p);
1266 extern int idle_cpu(int cpu);
1267 extern int sched_setscheduler(struct task_struct *, int,
1268 const struct sched_param *);
1269 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1270 const struct sched_param *);
1271 extern int sched_setattr(struct task_struct *,
1272 const struct sched_attr *);
1273 extern struct task_struct *idle_task(int cpu);
1275 * is_idle_task - is the specified task an idle task?
1276 * @p: the task in question.
1278 * Return: 1 if @p is an idle task. 0 otherwise.
1280 static inline bool is_idle_task(const struct task_struct *p)
1282 return !!(p->flags & PF_IDLE);
1284 extern struct task_struct *curr_task(int cpu);
1285 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1289 union thread_union {
1290 #ifndef CONFIG_THREAD_INFO_IN_TASK
1291 struct thread_info thread_info;
1293 unsigned long stack[THREAD_SIZE/sizeof(long)];
1296 #ifdef CONFIG_THREAD_INFO_IN_TASK
1297 static inline struct thread_info *task_thread_info(struct task_struct *task)
1299 return &task->thread_info;
1301 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1302 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1305 extern struct pid_namespace init_pid_ns;
1308 * find a task by one of its numerical ids
1310 * find_task_by_pid_ns():
1311 * finds a task by its pid in the specified namespace
1312 * find_task_by_vpid():
1313 * finds a task by its virtual pid
1315 * see also find_vpid() etc in include/linux/pid.h
1318 extern struct task_struct *find_task_by_vpid(pid_t nr);
1319 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1320 struct pid_namespace *ns);
1322 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1323 extern int wake_up_process(struct task_struct *tsk);
1324 extern void wake_up_new_task(struct task_struct *tsk);
1326 extern void kick_process(struct task_struct *tsk);
1328 static inline void kick_process(struct task_struct *tsk) { }
1331 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1332 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1334 __set_task_comm(tsk, from, false);
1336 extern char *get_task_comm(char *to, struct task_struct *tsk);
1339 void scheduler_ipi(void);
1340 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1342 static inline void scheduler_ipi(void) { }
1343 static inline unsigned long wait_task_inactive(struct task_struct *p,
1350 /* set thread flags in other task's structures
1351 * - see asm/thread_info.h for TIF_xxxx flags available
1353 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1355 set_ti_thread_flag(task_thread_info(tsk), flag);
1358 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1360 clear_ti_thread_flag(task_thread_info(tsk), flag);
1363 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1365 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1368 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1370 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1373 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1375 return test_ti_thread_flag(task_thread_info(tsk), flag);
1378 static inline void set_tsk_need_resched(struct task_struct *tsk)
1380 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1383 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1385 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1388 static inline int test_tsk_need_resched(struct task_struct *tsk)
1390 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1394 * cond_resched() and cond_resched_lock(): latency reduction via
1395 * explicit rescheduling in places that are safe. The return
1396 * value indicates whether a reschedule was done in fact.
1397 * cond_resched_lock() will drop the spinlock before scheduling,
1398 * cond_resched_softirq() will enable bhs before scheduling.
1400 #ifndef CONFIG_PREEMPT
1401 extern int _cond_resched(void);
1403 static inline int _cond_resched(void) { return 0; }
1406 #define cond_resched() ({ \
1407 ___might_sleep(__FILE__, __LINE__, 0); \
1411 extern int __cond_resched_lock(spinlock_t *lock);
1413 #define cond_resched_lock(lock) ({ \
1414 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1415 __cond_resched_lock(lock); \
1418 extern int __cond_resched_softirq(void);
1420 #define cond_resched_softirq() ({ \
1421 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1422 __cond_resched_softirq(); \
1425 static inline void cond_resched_rcu(void)
1427 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1435 * Does a critical section need to be broken due to another
1436 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1437 * but a general need for low latency)
1439 static inline int spin_needbreak(spinlock_t *lock)
1441 #ifdef CONFIG_PREEMPT
1442 return spin_is_contended(lock);
1448 static __always_inline bool need_resched(void)
1450 return unlikely(tif_need_resched());
1454 * Wrappers for p->thread_info->cpu access. No-op on UP.
1458 static inline unsigned int task_cpu(const struct task_struct *p)
1460 #ifdef CONFIG_THREAD_INFO_IN_TASK
1463 return task_thread_info(p)->cpu;
1467 static inline int task_node(const struct task_struct *p)
1469 return cpu_to_node(task_cpu(p));
1472 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1476 static inline unsigned int task_cpu(const struct task_struct *p)
1481 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1485 #endif /* CONFIG_SMP */
1488 * In order to reduce various lock holder preemption latencies provide an
1489 * interface to see if a vCPU is currently running or not.
1491 * This allows us to terminate optimistic spin loops and block, analogous to
1492 * the native optimistic spin heuristic of testing if the lock owner task is
1495 #ifndef vcpu_is_preempted
1496 # define vcpu_is_preempted(cpu) false
1499 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1500 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1502 #ifndef TASK_SIZE_OF
1503 #define TASK_SIZE_OF(tsk) TASK_SIZE