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 #include <linux/rwsem.h>
231 #ifdef CONFIG_SCHED_INFO
233 /* cumulative counters */
234 unsigned long pcount; /* # of times run on this cpu */
235 unsigned long long run_delay; /* time spent waiting on a runqueue */
238 unsigned long long last_arrival,/* when we last ran on a cpu */
239 last_queued; /* when we were last queued to run */
241 #endif /* CONFIG_SCHED_INFO */
244 * Integer metrics need fixed point arithmetic, e.g., sched/fair
245 * has a few: load, load_avg, util_avg, freq, and capacity.
247 * We define a basic fixed point arithmetic range, and then formalize
248 * all these metrics based on that basic range.
250 # define SCHED_FIXEDPOINT_SHIFT 10
251 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
253 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
254 extern void prefetch_stack(struct task_struct *t);
256 static inline void prefetch_stack(struct task_struct *t) { }
260 unsigned long weight;
265 * The load_avg/util_avg accumulates an infinite geometric series
266 * (see __update_load_avg() in kernel/sched/fair.c).
268 * [load_avg definition]
270 * load_avg = runnable% * scale_load_down(load)
272 * where runnable% is the time ratio that a sched_entity is runnable.
273 * For cfs_rq, it is the aggregated load_avg of all runnable and
274 * blocked sched_entities.
276 * load_avg may also take frequency scaling into account:
278 * load_avg = runnable% * scale_load_down(load) * freq%
280 * where freq% is the CPU frequency normalized to the highest frequency.
282 * [util_avg definition]
284 * util_avg = running% * SCHED_CAPACITY_SCALE
286 * where running% is the time ratio that a sched_entity is running on
287 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
288 * and blocked sched_entities.
290 * util_avg may also factor frequency scaling and CPU capacity scaling:
292 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
294 * where freq% is the same as above, and capacity% is the CPU capacity
295 * normalized to the greatest capacity (due to uarch differences, etc).
297 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
298 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
299 * we therefore scale them to as large a range as necessary. This is for
300 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
304 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
305 * with the highest load (=88761), always runnable on a single cfs_rq,
306 * and should not overflow as the number already hits PID_MAX_LIMIT.
308 * For all other cases (including 32-bit kernels), struct load_weight's
309 * weight will overflow first before we do, because:
311 * Max(load_avg) <= Max(load.weight)
313 * Then it is the load_weight's responsibility to consider overflow
317 u64 last_update_time, load_sum;
318 u32 util_sum, period_contrib;
319 unsigned long load_avg, util_avg;
322 #ifdef CONFIG_SCHEDSTATS
323 struct sched_statistics {
333 s64 sum_sleep_runtime;
340 u64 nr_migrations_cold;
341 u64 nr_failed_migrations_affine;
342 u64 nr_failed_migrations_running;
343 u64 nr_failed_migrations_hot;
344 u64 nr_forced_migrations;
348 u64 nr_wakeups_migrate;
349 u64 nr_wakeups_local;
350 u64 nr_wakeups_remote;
351 u64 nr_wakeups_affine;
352 u64 nr_wakeups_affine_attempts;
353 u64 nr_wakeups_passive;
358 struct sched_entity {
359 struct load_weight load; /* for load-balancing */
360 struct rb_node run_node;
361 struct list_head group_node;
365 u64 sum_exec_runtime;
367 u64 prev_sum_exec_runtime;
371 #ifdef CONFIG_SCHEDSTATS
372 struct sched_statistics statistics;
375 #ifdef CONFIG_FAIR_GROUP_SCHED
377 struct sched_entity *parent;
378 /* rq on which this entity is (to be) queued: */
379 struct cfs_rq *cfs_rq;
380 /* rq "owned" by this entity/group: */
386 * Per entity load average tracking.
388 * Put into separate cache line so it does not
389 * collide with read-mostly values above.
391 struct sched_avg avg ____cacheline_aligned_in_smp;
395 struct sched_rt_entity {
396 struct list_head run_list;
397 unsigned long timeout;
398 unsigned long watchdog_stamp;
399 unsigned int time_slice;
400 unsigned short on_rq;
401 unsigned short on_list;
403 struct sched_rt_entity *back;
404 #ifdef CONFIG_RT_GROUP_SCHED
405 struct sched_rt_entity *parent;
406 /* rq on which this entity is (to be) queued: */
408 /* rq "owned" by this entity/group: */
413 struct sched_dl_entity {
414 struct rb_node rb_node;
417 * Original scheduling parameters. Copied here from sched_attr
418 * during sched_setattr(), they will remain the same until
419 * the next sched_setattr().
421 u64 dl_runtime; /* maximum runtime for each instance */
422 u64 dl_deadline; /* relative deadline of each instance */
423 u64 dl_period; /* separation of two instances (period) */
424 u64 dl_bw; /* dl_runtime / dl_deadline */
427 * Actual scheduling parameters. Initialized with the values above,
428 * they are continously updated during task execution. Note that
429 * the remaining runtime could be < 0 in case we are in overrun.
431 s64 runtime; /* remaining runtime for this instance */
432 u64 deadline; /* absolute deadline for this instance */
433 unsigned int flags; /* specifying the scheduler behaviour */
438 * @dl_throttled tells if we exhausted the runtime. If so, the
439 * task has to wait for a replenishment to be performed at the
440 * next firing of dl_timer.
442 * @dl_boosted tells if we are boosted due to DI. If so we are
443 * outside bandwidth enforcement mechanism (but only until we
444 * exit the critical section);
446 * @dl_yielded tells if task gave up the cpu before consuming
447 * all its available runtime during the last job.
449 int dl_throttled, dl_boosted, dl_yielded;
452 * Bandwidth enforcement timer. Each -deadline task has its
453 * own bandwidth to be enforced, thus we need one timer per task.
455 struct hrtimer dl_timer;
463 u8 pad; /* Otherwise the compiler can store garbage here. */
465 u32 s; /* Set of bits. */
468 enum perf_event_task_context {
469 perf_invalid_context = -1,
472 perf_nr_task_contexts,
476 struct wake_q_node *next;
480 #ifdef CONFIG_THREAD_INFO_IN_TASK
482 * For reasons of header soup (see current_thread_info()), this
483 * must be the first element of task_struct.
485 struct thread_info thread_info;
487 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
490 unsigned int flags; /* per process flags, defined below */
494 struct llist_node wake_entry;
496 #ifdef CONFIG_THREAD_INFO_IN_TASK
497 unsigned int cpu; /* current CPU */
499 unsigned int wakee_flips;
500 unsigned long wakee_flip_decay_ts;
501 struct task_struct *last_wakee;
507 int prio, static_prio, normal_prio;
508 unsigned int rt_priority;
509 const struct sched_class *sched_class;
510 struct sched_entity se;
511 struct sched_rt_entity rt;
512 #ifdef CONFIG_CGROUP_SCHED
513 struct task_group *sched_task_group;
515 struct sched_dl_entity dl;
517 #ifdef CONFIG_PREEMPT_NOTIFIERS
518 /* list of struct preempt_notifier: */
519 struct hlist_head preempt_notifiers;
522 #ifdef CONFIG_BLK_DEV_IO_TRACE
523 unsigned int btrace_seq;
528 cpumask_t cpus_allowed;
530 #ifdef CONFIG_PREEMPT_RCU
531 int rcu_read_lock_nesting;
532 union rcu_special rcu_read_unlock_special;
533 struct list_head rcu_node_entry;
534 struct rcu_node *rcu_blocked_node;
535 #endif /* #ifdef CONFIG_PREEMPT_RCU */
536 #ifdef CONFIG_TASKS_RCU
537 unsigned long rcu_tasks_nvcsw;
538 bool rcu_tasks_holdout;
539 struct list_head rcu_tasks_holdout_list;
540 int rcu_tasks_idle_cpu;
541 #endif /* #ifdef CONFIG_TASKS_RCU */
543 #ifdef CONFIG_SCHED_INFO
544 struct sched_info sched_info;
547 struct list_head tasks;
549 struct plist_node pushable_tasks;
550 struct rb_node pushable_dl_tasks;
553 struct mm_struct *mm, *active_mm;
555 /* Per-thread vma caching: */
556 struct vmacache vmacache;
558 #if defined(SPLIT_RSS_COUNTING)
559 struct task_rss_stat rss_stat;
563 int exit_code, exit_signal;
564 int pdeath_signal; /* The signal sent when the parent dies */
565 unsigned long jobctl; /* JOBCTL_*, siglock protected */
567 /* Used for emulating ABI behavior of previous Linux versions */
568 unsigned int personality;
570 /* scheduler bits, serialized by scheduler locks */
571 unsigned sched_reset_on_fork:1;
572 unsigned sched_contributes_to_load:1;
573 unsigned sched_migrated:1;
574 unsigned sched_remote_wakeup:1;
575 unsigned :0; /* force alignment to the next boundary */
577 /* unserialized, strictly 'current' */
578 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
579 unsigned in_iowait:1;
580 #if !defined(TIF_RESTORE_SIGMASK)
581 unsigned restore_sigmask:1;
584 unsigned memcg_may_oom:1;
586 unsigned memcg_kmem_skip_account:1;
589 #ifdef CONFIG_COMPAT_BRK
590 unsigned brk_randomized:1;
593 unsigned long atomic_flags; /* Flags needing atomic access. */
595 struct restart_block restart_block;
600 #ifdef CONFIG_CC_STACKPROTECTOR
601 /* Canary value for the -fstack-protector gcc feature */
602 unsigned long stack_canary;
605 * pointers to (original) parent process, youngest child, younger sibling,
606 * older sibling, respectively. (p->father can be replaced with
607 * p->real_parent->pid)
609 struct task_struct __rcu *real_parent; /* real parent process */
610 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
612 * children/sibling forms the list of my natural children
614 struct list_head children; /* list of my children */
615 struct list_head sibling; /* linkage in my parent's children list */
616 struct task_struct *group_leader; /* threadgroup leader */
619 * ptraced is the list of tasks this task is using ptrace on.
620 * This includes both natural children and PTRACE_ATTACH targets.
621 * p->ptrace_entry is p's link on the p->parent->ptraced list.
623 struct list_head ptraced;
624 struct list_head ptrace_entry;
626 /* PID/PID hash table linkage. */
627 struct pid_link pids[PIDTYPE_MAX];
628 struct list_head thread_group;
629 struct list_head thread_node;
631 struct completion *vfork_done; /* for vfork() */
632 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
633 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
636 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
637 u64 utimescaled, stimescaled;
640 struct prev_cputime prev_cputime;
641 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
642 seqcount_t vtime_seqcount;
643 unsigned long long vtime_snap;
645 /* Task is sleeping or running in a CPU with VTIME inactive */
647 /* Task runs in userspace in a CPU with VTIME active */
649 /* Task runs in kernelspace in a CPU with VTIME active */
654 #ifdef CONFIG_NO_HZ_FULL
655 atomic_t tick_dep_mask;
657 unsigned long nvcsw, nivcsw; /* context switch counts */
658 u64 start_time; /* monotonic time in nsec */
659 u64 real_start_time; /* boot based time in nsec */
660 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
661 unsigned long min_flt, maj_flt;
663 #ifdef CONFIG_POSIX_TIMERS
664 struct task_cputime cputime_expires;
665 struct list_head cpu_timers[3];
668 /* process credentials */
669 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
670 const struct cred __rcu *real_cred; /* objective and real subjective task
671 * credentials (COW) */
672 const struct cred __rcu *cred; /* effective (overridable) subjective task
673 * credentials (COW) */
674 char comm[TASK_COMM_LEN]; /* executable name excluding path
675 - access with [gs]et_task_comm (which lock
677 - initialized normally by setup_new_exec */
678 /* file system info */
679 struct nameidata *nameidata;
680 #ifdef CONFIG_SYSVIPC
682 struct sysv_sem sysvsem;
683 struct sysv_shm sysvshm;
685 #ifdef CONFIG_DETECT_HUNG_TASK
686 /* hung task detection */
687 unsigned long last_switch_count;
689 /* filesystem information */
690 struct fs_struct *fs;
691 /* open file information */
692 struct files_struct *files;
694 struct nsproxy *nsproxy;
695 /* signal handlers */
696 struct signal_struct *signal;
697 struct sighand_struct *sighand;
699 sigset_t blocked, real_blocked;
700 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
701 struct sigpending pending;
703 unsigned long sas_ss_sp;
705 unsigned sas_ss_flags;
707 struct callback_head *task_works;
709 struct audit_context *audit_context;
710 #ifdef CONFIG_AUDITSYSCALL
712 unsigned int sessionid;
714 struct seccomp seccomp;
716 /* Thread group tracking */
719 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
721 spinlock_t alloc_lock;
723 /* Protection of the PI data structures: */
724 raw_spinlock_t pi_lock;
726 struct wake_q_node wake_q;
728 #ifdef CONFIG_RT_MUTEXES
729 /* PI waiters blocked on a rt_mutex held by this task */
730 struct rb_root pi_waiters;
731 struct rb_node *pi_waiters_leftmost;
732 /* Deadlock detection and priority inheritance handling */
733 struct rt_mutex_waiter *pi_blocked_on;
736 #ifdef CONFIG_DEBUG_MUTEXES
737 /* mutex deadlock detection */
738 struct mutex_waiter *blocked_on;
740 #ifdef CONFIG_TRACE_IRQFLAGS
741 unsigned int irq_events;
742 unsigned long hardirq_enable_ip;
743 unsigned long hardirq_disable_ip;
744 unsigned int hardirq_enable_event;
745 unsigned int hardirq_disable_event;
746 int hardirqs_enabled;
748 unsigned long softirq_disable_ip;
749 unsigned long softirq_enable_ip;
750 unsigned int softirq_disable_event;
751 unsigned int softirq_enable_event;
752 int softirqs_enabled;
755 #ifdef CONFIG_LOCKDEP
756 # define MAX_LOCK_DEPTH 48UL
759 unsigned int lockdep_recursion;
760 struct held_lock held_locks[MAX_LOCK_DEPTH];
761 gfp_t lockdep_reclaim_gfp;
764 unsigned int in_ubsan;
767 /* journalling filesystem info */
770 /* stacked block device info */
771 struct bio_list *bio_list;
775 struct blk_plug *plug;
779 struct reclaim_state *reclaim_state;
781 struct backing_dev_info *backing_dev_info;
783 struct io_context *io_context;
785 unsigned long ptrace_message;
786 siginfo_t *last_siginfo; /* For ptrace use. */
787 struct task_io_accounting ioac;
788 #if defined(CONFIG_TASK_XACCT)
789 u64 acct_rss_mem1; /* accumulated rss usage */
790 u64 acct_vm_mem1; /* accumulated virtual memory usage */
791 u64 acct_timexpd; /* stime + utime since last update */
793 #ifdef CONFIG_CPUSETS
794 nodemask_t mems_allowed; /* Protected by alloc_lock */
795 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
796 int cpuset_mem_spread_rotor;
797 int cpuset_slab_spread_rotor;
799 #ifdef CONFIG_CGROUPS
800 /* Control Group info protected by css_set_lock */
801 struct css_set __rcu *cgroups;
802 /* cg_list protected by css_set_lock and tsk->alloc_lock */
803 struct list_head cg_list;
805 #ifdef CONFIG_INTEL_RDT_A
809 struct robust_list_head __user *robust_list;
811 struct compat_robust_list_head __user *compat_robust_list;
813 struct list_head pi_state_list;
814 struct futex_pi_state *pi_state_cache;
816 #ifdef CONFIG_PERF_EVENTS
817 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
818 struct mutex perf_event_mutex;
819 struct list_head perf_event_list;
821 #ifdef CONFIG_DEBUG_PREEMPT
822 unsigned long preempt_disable_ip;
825 struct mempolicy *mempolicy; /* Protected by alloc_lock */
827 short pref_node_fork;
829 #ifdef CONFIG_NUMA_BALANCING
831 unsigned int numa_scan_period;
832 unsigned int numa_scan_period_max;
833 int numa_preferred_nid;
834 unsigned long numa_migrate_retry;
835 u64 node_stamp; /* migration stamp */
836 u64 last_task_numa_placement;
837 u64 last_sum_exec_runtime;
838 struct callback_head numa_work;
840 struct list_head numa_entry;
841 struct numa_group *numa_group;
844 * numa_faults is an array split into four regions:
845 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
846 * in this precise order.
848 * faults_memory: Exponential decaying average of faults on a per-node
849 * basis. Scheduling placement decisions are made based on these
850 * counts. The values remain static for the duration of a PTE scan.
851 * faults_cpu: Track the nodes the process was running on when a NUMA
852 * hinting fault was incurred.
853 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
854 * during the current scan window. When the scan completes, the counts
855 * in faults_memory and faults_cpu decay and these values are copied.
857 unsigned long *numa_faults;
858 unsigned long total_numa_faults;
861 * numa_faults_locality tracks if faults recorded during the last
862 * scan window were remote/local or failed to migrate. The task scan
863 * period is adapted based on the locality of the faults with different
864 * weights depending on whether they were shared or private faults
866 unsigned long numa_faults_locality[3];
868 unsigned long numa_pages_migrated;
869 #endif /* CONFIG_NUMA_BALANCING */
871 struct tlbflush_unmap_batch tlb_ubc;
876 * cache last used pipe for splice
878 struct pipe_inode_info *splice_pipe;
880 struct page_frag task_frag;
882 #ifdef CONFIG_TASK_DELAY_ACCT
883 struct task_delay_info *delays;
886 #ifdef CONFIG_FAULT_INJECTION
890 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
891 * balance_dirty_pages() for some dirty throttling pause
894 int nr_dirtied_pause;
895 unsigned long dirty_paused_when; /* start of a write-and-pause period */
897 #ifdef CONFIG_LATENCYTOP
898 int latency_record_count;
899 struct latency_record latency_record[LT_SAVECOUNT];
902 * time slack values; these are used to round up poll() and
903 * select() etc timeout values. These are in nanoseconds.
906 u64 default_timer_slack_ns;
909 unsigned int kasan_depth;
911 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
912 /* Index of current stored address in ret_stack */
914 /* Stack of return addresses for return function tracing */
915 struct ftrace_ret_stack *ret_stack;
916 /* time stamp for last schedule */
917 unsigned long long ftrace_timestamp;
919 * Number of functions that haven't been traced
920 * because of depth overrun.
922 atomic_t trace_overrun;
923 /* Pause for the tracing */
924 atomic_t tracing_graph_pause;
926 #ifdef CONFIG_TRACING
927 /* state flags for use by tracers */
929 /* bitmask and counter of trace recursion */
930 unsigned long trace_recursion;
931 #endif /* CONFIG_TRACING */
933 /* Coverage collection mode enabled for this task (0 if disabled). */
934 enum kcov_mode kcov_mode;
935 /* Size of the kcov_area. */
937 /* Buffer for coverage collection. */
939 /* kcov desciptor wired with this task or NULL. */
943 struct mem_cgroup *memcg_in_oom;
944 gfp_t memcg_oom_gfp_mask;
947 /* number of pages to reclaim on returning to userland */
948 unsigned int memcg_nr_pages_over_high;
950 #ifdef CONFIG_UPROBES
951 struct uprobe_task *utask;
953 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
954 unsigned int sequential_io;
955 unsigned int sequential_io_avg;
957 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
958 unsigned long task_state_change;
960 int pagefault_disabled;
962 struct task_struct *oom_reaper_list;
964 #ifdef CONFIG_VMAP_STACK
965 struct vm_struct *stack_vm_area;
967 #ifdef CONFIG_THREAD_INFO_IN_TASK
968 /* A live task holds one reference. */
969 atomic_t stack_refcount;
971 /* CPU-specific state of this task */
972 struct thread_struct thread;
974 * WARNING: on x86, 'thread_struct' contains a variable-sized
975 * structure. It *MUST* be at the end of 'task_struct'.
977 * Do not put anything below here!
981 static inline struct pid *task_pid(struct task_struct *task)
983 return task->pids[PIDTYPE_PID].pid;
986 static inline struct pid *task_tgid(struct task_struct *task)
988 return task->group_leader->pids[PIDTYPE_PID].pid;
992 * Without tasklist or rcu lock it is not safe to dereference
993 * the result of task_pgrp/task_session even if task == current,
994 * we can race with another thread doing sys_setsid/sys_setpgid.
996 static inline struct pid *task_pgrp(struct task_struct *task)
998 return task->group_leader->pids[PIDTYPE_PGID].pid;
1001 static inline struct pid *task_session(struct task_struct *task)
1003 return task->group_leader->pids[PIDTYPE_SID].pid;
1007 * the helpers to get the task's different pids as they are seen
1008 * from various namespaces
1010 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1011 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1013 * task_xid_nr_ns() : id seen from the ns specified;
1015 * set_task_vxid() : assigns a virtual id to a task;
1017 * see also pid_nr() etc in include/linux/pid.h
1019 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1020 struct pid_namespace *ns);
1022 static inline pid_t task_pid_nr(struct task_struct *tsk)
1027 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1028 struct pid_namespace *ns)
1030 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1033 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1035 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1039 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1044 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1046 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1048 return pid_vnr(task_tgid(tsk));
1052 static inline int pid_alive(const struct task_struct *p);
1053 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1059 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1065 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1067 return task_ppid_nr_ns(tsk, &init_pid_ns);
1070 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1071 struct pid_namespace *ns)
1073 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1076 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1078 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1082 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1083 struct pid_namespace *ns)
1085 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1088 static inline pid_t task_session_vnr(struct task_struct *tsk)
1090 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1093 /* obsolete, do not use */
1094 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1096 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1100 * pid_alive - check that a task structure is not stale
1101 * @p: Task structure to be checked.
1103 * Test if a process is not yet dead (at most zombie state)
1104 * If pid_alive fails, then pointers within the task structure
1105 * can be stale and must not be dereferenced.
1107 * Return: 1 if the process is alive. 0 otherwise.
1109 static inline int pid_alive(const struct task_struct *p)
1111 return p->pids[PIDTYPE_PID].pid != NULL;
1115 * is_global_init - check if a task structure is init. Since init
1116 * is free to have sub-threads we need to check tgid.
1117 * @tsk: Task structure to be checked.
1119 * Check if a task structure is the first user space task the kernel created.
1121 * Return: 1 if the task structure is init. 0 otherwise.
1123 static inline int is_global_init(struct task_struct *tsk)
1125 return task_tgid_nr(tsk) == 1;
1128 extern struct pid *cad_pid;
1133 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1134 #define PF_EXITING 0x00000004 /* getting shut down */
1135 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1136 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1137 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1138 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1139 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1140 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1141 #define PF_DUMPCORE 0x00000200 /* dumped core */
1142 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1143 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1144 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1145 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1146 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1147 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1148 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1149 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1150 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1151 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1152 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1153 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1154 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1155 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1156 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1157 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1158 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1159 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1160 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1163 * Only the _current_ task can read/write to tsk->flags, but other
1164 * tasks can access tsk->flags in readonly mode for example
1165 * with tsk_used_math (like during threaded core dumping).
1166 * There is however an exception to this rule during ptrace
1167 * or during fork: the ptracer task is allowed to write to the
1168 * child->flags of its traced child (same goes for fork, the parent
1169 * can write to the child->flags), because we're guaranteed the
1170 * child is not running and in turn not changing child->flags
1171 * at the same time the parent does it.
1173 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1174 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1175 #define clear_used_math() clear_stopped_child_used_math(current)
1176 #define set_used_math() set_stopped_child_used_math(current)
1177 #define conditional_stopped_child_used_math(condition, child) \
1178 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1179 #define conditional_used_math(condition) \
1180 conditional_stopped_child_used_math(condition, current)
1181 #define copy_to_stopped_child_used_math(child) \
1182 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1183 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1184 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1185 #define used_math() tsk_used_math(current)
1187 /* Per-process atomic flags. */
1188 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1189 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1190 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1191 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1194 #define TASK_PFA_TEST(name, func) \
1195 static inline bool task_##func(struct task_struct *p) \
1196 { return test_bit(PFA_##name, &p->atomic_flags); }
1197 #define TASK_PFA_SET(name, func) \
1198 static inline void task_set_##func(struct task_struct *p) \
1199 { set_bit(PFA_##name, &p->atomic_flags); }
1200 #define TASK_PFA_CLEAR(name, func) \
1201 static inline void task_clear_##func(struct task_struct *p) \
1202 { clear_bit(PFA_##name, &p->atomic_flags); }
1204 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1205 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1207 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1208 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1209 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1211 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1212 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1213 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1215 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1216 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1218 static inline void tsk_restore_flags(struct task_struct *task,
1219 unsigned long orig_flags, unsigned long flags)
1221 task->flags &= ~flags;
1222 task->flags |= orig_flags & flags;
1225 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1226 const struct cpumask *trial);
1227 extern int task_can_attach(struct task_struct *p,
1228 const struct cpumask *cs_cpus_allowed);
1230 extern void do_set_cpus_allowed(struct task_struct *p,
1231 const struct cpumask *new_mask);
1233 extern int set_cpus_allowed_ptr(struct task_struct *p,
1234 const struct cpumask *new_mask);
1236 static inline void do_set_cpus_allowed(struct task_struct *p,
1237 const struct cpumask *new_mask)
1240 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1241 const struct cpumask *new_mask)
1243 if (!cpumask_test_cpu(0, new_mask))
1249 #ifndef cpu_relax_yield
1250 #define cpu_relax_yield() cpu_relax()
1253 extern int yield_to(struct task_struct *p, bool preempt);
1254 extern void set_user_nice(struct task_struct *p, long nice);
1255 extern int task_prio(const struct task_struct *p);
1257 * task_nice - return the nice value of a given task.
1258 * @p: the task in question.
1260 * Return: The nice value [ -20 ... 0 ... 19 ].
1262 static inline int task_nice(const struct task_struct *p)
1264 return PRIO_TO_NICE((p)->static_prio);
1266 extern int can_nice(const struct task_struct *p, const int nice);
1267 extern int task_curr(const struct task_struct *p);
1268 extern int idle_cpu(int cpu);
1269 extern int sched_setscheduler(struct task_struct *, int,
1270 const struct sched_param *);
1271 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1272 const struct sched_param *);
1273 extern int sched_setattr(struct task_struct *,
1274 const struct sched_attr *);
1275 extern struct task_struct *idle_task(int cpu);
1277 * is_idle_task - is the specified task an idle task?
1278 * @p: the task in question.
1280 * Return: 1 if @p is an idle task. 0 otherwise.
1282 static inline bool is_idle_task(const struct task_struct *p)
1284 return !!(p->flags & PF_IDLE);
1286 extern struct task_struct *curr_task(int cpu);
1287 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1291 union thread_union {
1292 #ifndef CONFIG_THREAD_INFO_IN_TASK
1293 struct thread_info thread_info;
1295 unsigned long stack[THREAD_SIZE/sizeof(long)];
1298 #ifdef CONFIG_THREAD_INFO_IN_TASK
1299 static inline struct thread_info *task_thread_info(struct task_struct *task)
1301 return &task->thread_info;
1303 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1304 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1307 extern struct pid_namespace init_pid_ns;
1310 * find a task by one of its numerical ids
1312 * find_task_by_pid_ns():
1313 * finds a task by its pid in the specified namespace
1314 * find_task_by_vpid():
1315 * finds a task by its virtual pid
1317 * see also find_vpid() etc in include/linux/pid.h
1320 extern struct task_struct *find_task_by_vpid(pid_t nr);
1321 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1322 struct pid_namespace *ns);
1324 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1325 extern int wake_up_process(struct task_struct *tsk);
1326 extern void wake_up_new_task(struct task_struct *tsk);
1328 extern void kick_process(struct task_struct *tsk);
1330 static inline void kick_process(struct task_struct *tsk) { }
1333 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1334 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1336 __set_task_comm(tsk, from, false);
1338 extern char *get_task_comm(char *to, struct task_struct *tsk);
1341 void scheduler_ipi(void);
1342 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1344 static inline void scheduler_ipi(void) { }
1345 static inline unsigned long wait_task_inactive(struct task_struct *p,
1352 /* set thread flags in other task's structures
1353 * - see asm/thread_info.h for TIF_xxxx flags available
1355 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1357 set_ti_thread_flag(task_thread_info(tsk), flag);
1360 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1362 clear_ti_thread_flag(task_thread_info(tsk), flag);
1365 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1367 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1370 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1372 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1375 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1377 return test_ti_thread_flag(task_thread_info(tsk), flag);
1380 static inline void set_tsk_need_resched(struct task_struct *tsk)
1382 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1385 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1387 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1390 static inline int test_tsk_need_resched(struct task_struct *tsk)
1392 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1396 * cond_resched() and cond_resched_lock(): latency reduction via
1397 * explicit rescheduling in places that are safe. The return
1398 * value indicates whether a reschedule was done in fact.
1399 * cond_resched_lock() will drop the spinlock before scheduling,
1400 * cond_resched_softirq() will enable bhs before scheduling.
1402 #ifndef CONFIG_PREEMPT
1403 extern int _cond_resched(void);
1405 static inline int _cond_resched(void) { return 0; }
1408 #define cond_resched() ({ \
1409 ___might_sleep(__FILE__, __LINE__, 0); \
1413 extern int __cond_resched_lock(spinlock_t *lock);
1415 #define cond_resched_lock(lock) ({ \
1416 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1417 __cond_resched_lock(lock); \
1420 extern int __cond_resched_softirq(void);
1422 #define cond_resched_softirq() ({ \
1423 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1424 __cond_resched_softirq(); \
1427 static inline void cond_resched_rcu(void)
1429 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1437 * Does a critical section need to be broken due to another
1438 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1439 * but a general need for low latency)
1441 static inline int spin_needbreak(spinlock_t *lock)
1443 #ifdef CONFIG_PREEMPT
1444 return spin_is_contended(lock);
1450 static __always_inline bool need_resched(void)
1452 return unlikely(tif_need_resched());
1456 * Wrappers for p->thread_info->cpu access. No-op on UP.
1460 static inline unsigned int task_cpu(const struct task_struct *p)
1462 #ifdef CONFIG_THREAD_INFO_IN_TASK
1465 return task_thread_info(p)->cpu;
1469 static inline int task_node(const struct task_struct *p)
1471 return cpu_to_node(task_cpu(p));
1474 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1478 static inline unsigned int task_cpu(const struct task_struct *p)
1483 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1487 #endif /* CONFIG_SMP */
1490 * In order to reduce various lock holder preemption latencies provide an
1491 * interface to see if a vCPU is currently running or not.
1493 * This allows us to terminate optimistic spin loops and block, analogous to
1494 * the native optimistic spin heuristic of testing if the lock owner task is
1497 #ifndef vcpu_is_preempted
1498 # define vcpu_is_preempted(cpu) false
1501 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1502 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1504 #ifndef TASK_SIZE_OF
1505 #define TASK_SIZE_OF(tsk) TASK_SIZE