5 * Define 'struct task_struct' and provide the main scheduler
6 * APIs (schedule(), wakeup variants, etc.)
9 #include <uapi/linux/sched.h>
11 #include <asm/current.h>
13 #include <linux/pid.h>
14 #include <linux/sem.h>
15 #include <linux/shm.h>
16 #include <linux/kcov.h>
17 #include <linux/mutex.h>
18 #include <linux/plist.h>
19 #include <linux/hrtimer.h>
20 #include <linux/seccomp.h>
21 #include <linux/nodemask.h>
22 #include <linux/rcupdate.h>
23 #include <linux/resource.h>
24 #include <linux/latencytop.h>
25 #include <linux/sched/prio.h>
26 #include <linux/signal_types.h>
27 #include <linux/mm_types_task.h>
28 #include <linux/task_io_accounting.h>
30 /* task_struct member predeclarations (sorted alphabetically): */
32 struct backing_dev_info;
37 struct futex_pi_state;
42 struct perf_event_context;
44 struct pipe_inode_info;
47 struct robust_list_head;
51 struct sighand_struct;
53 struct task_delay_info;
57 * Task state bitmask. NOTE! These bits are also
58 * encoded in fs/proc/array.c: get_task_state().
60 * We have two separate sets of flags: task->state
61 * is about runnability, while task->exit_state are
62 * about the task exiting. Confusing, but this way
63 * modifying one set can't modify the other one by
67 /* Used in tsk->state: */
68 #define TASK_RUNNING 0
69 #define TASK_INTERRUPTIBLE 1
70 #define TASK_UNINTERRUPTIBLE 2
71 #define __TASK_STOPPED 4
72 #define __TASK_TRACED 8
73 /* Used in tsk->exit_state: */
75 #define EXIT_ZOMBIE 32
76 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
77 /* Used in tsk->state again: */
79 #define TASK_WAKEKILL 128
80 #define TASK_WAKING 256
81 #define TASK_PARKED 512
82 #define TASK_NOLOAD 1024
84 #define TASK_STATE_MAX 4096
86 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
88 /* Convenience macros for the sake of set_current_state: */
89 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
90 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
91 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
93 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
95 /* Convenience macros for the sake of wake_up(): */
96 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
97 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
99 /* get_task_state(): */
100 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
101 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
102 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
104 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
108 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
110 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 (task->flags & PF_FROZEN) == 0 && \
112 (task->state & TASK_NOLOAD) == 0)
114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
116 #define __set_current_state(state_value) \
118 current->task_state_change = _THIS_IP_; \
119 current->state = (state_value); \
121 #define set_current_state(state_value) \
123 current->task_state_change = _THIS_IP_; \
124 smp_store_mb(current->state, (state_value)); \
129 * set_current_state() includes a barrier so that the write of current->state
130 * is correctly serialised wrt the caller's subsequent test of whether to
134 * set_current_state(TASK_UNINTERRUPTIBLE);
140 * __set_current_state(TASK_RUNNING);
142 * If the caller does not need such serialisation (because, for instance, the
143 * condition test and condition change and wakeup are under the same lock) then
144 * use __set_current_state().
146 * The above is typically ordered against the wakeup, which does:
148 * need_sleep = false;
149 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
151 * Where wake_up_state() (and all other wakeup primitives) imply enough
152 * barriers to order the store of the variable against wakeup.
154 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
155 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
156 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
158 * This is obviously fine, since they both store the exact same value.
160 * Also see the comments of try_to_wake_up().
162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0)
163 #define set_current_state(state_value) smp_store_mb(current->state, (state_value))
166 /* Task command name length: */
167 #define TASK_COMM_LEN 16
169 extern cpumask_var_t cpu_isolated_map;
171 extern void scheduler_tick(void);
173 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
175 extern long schedule_timeout(long timeout);
176 extern long schedule_timeout_interruptible(long timeout);
177 extern long schedule_timeout_killable(long timeout);
178 extern long schedule_timeout_uninterruptible(long timeout);
179 extern long schedule_timeout_idle(long timeout);
180 asmlinkage void schedule(void);
181 extern void schedule_preempt_disabled(void);
183 extern int __must_check io_schedule_prepare(void);
184 extern void io_schedule_finish(int token);
185 extern long io_schedule_timeout(long timeout);
186 extern void io_schedule(void);
189 * struct prev_cputime - snaphsot of system and user cputime
190 * @utime: time spent in user mode
191 * @stime: time spent in system mode
192 * @lock: protects the above two fields
194 * Stores previous user/system time values such that we can guarantee
197 struct prev_cputime {
198 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
206 * struct task_cputime - collected CPU time counts
207 * @utime: time spent in user mode, in nanoseconds
208 * @stime: time spent in kernel mode, in nanoseconds
209 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
211 * This structure groups together three kinds of CPU time that are tracked for
212 * threads and thread groups. Most things considering CPU time want to group
213 * these counts together and treat all three of them in parallel.
215 struct task_cputime {
218 unsigned long long sum_exec_runtime;
221 /* Alternate field names when used on cache expirations: */
222 #define virt_exp utime
223 #define prof_exp stime
224 #define sched_exp sum_exec_runtime
227 #ifdef CONFIG_SCHED_INFO
228 /* Cumulative counters: */
230 /* # of times we have run on this CPU: */
231 unsigned long pcount;
233 /* Time spent waiting on a runqueue: */
234 unsigned long long run_delay;
238 /* When did we last run on a CPU? */
239 unsigned long long last_arrival;
241 /* When were we last queued to run? */
242 unsigned long long last_queued;
244 #endif /* CONFIG_SCHED_INFO */
248 * Integer metrics need fixed point arithmetic, e.g., sched/fair
249 * has a few: load, load_avg, util_avg, freq, and capacity.
251 * We define a basic fixed point arithmetic range, and then formalize
252 * all these metrics based on that basic range.
254 # define SCHED_FIXEDPOINT_SHIFT 10
255 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
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;
319 unsigned long load_avg;
320 unsigned long util_avg;
323 struct sched_statistics {
324 #ifdef CONFIG_SCHEDSTATS
334 s64 sum_sleep_runtime;
341 u64 nr_migrations_cold;
342 u64 nr_failed_migrations_affine;
343 u64 nr_failed_migrations_running;
344 u64 nr_failed_migrations_hot;
345 u64 nr_forced_migrations;
349 u64 nr_wakeups_migrate;
350 u64 nr_wakeups_local;
351 u64 nr_wakeups_remote;
352 u64 nr_wakeups_affine;
353 u64 nr_wakeups_affine_attempts;
354 u64 nr_wakeups_passive;
359 struct sched_entity {
360 /* For load-balancing: */
361 struct load_weight load;
362 struct rb_node run_node;
363 struct list_head group_node;
367 u64 sum_exec_runtime;
369 u64 prev_sum_exec_runtime;
373 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.
454 * Bandwidth enforcement timer. Each -deadline task has its
455 * own bandwidth to be enforced, thus we need one timer per task.
457 struct hrtimer dl_timer;
466 /* Otherwise the compiler can store garbage here: */
469 u32 s; /* Set of bits. */
472 enum perf_event_task_context {
473 perf_invalid_context = -1,
476 perf_nr_task_contexts,
480 struct wake_q_node *next;
484 #ifdef CONFIG_THREAD_INFO_IN_TASK
486 * For reasons of header soup (see current_thread_info()), this
487 * must be the first element of task_struct.
489 struct thread_info thread_info;
491 /* -1 unrunnable, 0 runnable, >0 stopped: */
495 /* Per task flags (PF_*), defined further below: */
500 struct llist_node wake_entry;
502 #ifdef CONFIG_THREAD_INFO_IN_TASK
506 unsigned int wakee_flips;
507 unsigned long wakee_flip_decay_ts;
508 struct task_struct *last_wakee;
517 unsigned int rt_priority;
519 const struct sched_class *sched_class;
520 struct sched_entity se;
521 struct sched_rt_entity rt;
522 #ifdef CONFIG_CGROUP_SCHED
523 struct task_group *sched_task_group;
525 struct sched_dl_entity dl;
527 #ifdef CONFIG_PREEMPT_NOTIFIERS
528 /* List of struct preempt_notifier: */
529 struct hlist_head preempt_notifiers;
532 #ifdef CONFIG_BLK_DEV_IO_TRACE
533 unsigned int btrace_seq;
538 cpumask_t cpus_allowed;
540 #ifdef CONFIG_PREEMPT_RCU
541 int rcu_read_lock_nesting;
542 union rcu_special rcu_read_unlock_special;
543 struct list_head rcu_node_entry;
544 struct rcu_node *rcu_blocked_node;
545 #endif /* #ifdef CONFIG_PREEMPT_RCU */
547 #ifdef CONFIG_TASKS_RCU
548 unsigned long rcu_tasks_nvcsw;
549 bool rcu_tasks_holdout;
550 struct list_head rcu_tasks_holdout_list;
551 int rcu_tasks_idle_cpu;
552 #endif /* #ifdef CONFIG_TASKS_RCU */
554 struct sched_info sched_info;
556 struct list_head tasks;
558 struct plist_node pushable_tasks;
559 struct rb_node pushable_dl_tasks;
562 struct mm_struct *mm;
563 struct mm_struct *active_mm;
565 /* Per-thread vma caching: */
566 struct vmacache vmacache;
568 #ifdef SPLIT_RSS_COUNTING
569 struct task_rss_stat rss_stat;
574 /* The signal sent when the parent dies: */
576 /* JOBCTL_*, siglock protected: */
577 unsigned long jobctl;
579 /* Used for emulating ABI behavior of previous Linux versions: */
580 unsigned int personality;
582 /* Scheduler bits, serialized by scheduler locks: */
583 unsigned sched_reset_on_fork:1;
584 unsigned sched_contributes_to_load:1;
585 unsigned sched_migrated:1;
586 unsigned sched_remote_wakeup:1;
587 /* Force alignment to the next boundary: */
590 /* Unserialized, strictly 'current' */
592 /* Bit to tell LSMs we're in execve(): */
593 unsigned in_execve:1;
594 unsigned in_iowait:1;
595 #ifndef TIF_RESTORE_SIGMASK
596 unsigned restore_sigmask:1;
599 unsigned memcg_may_oom:1;
601 unsigned memcg_kmem_skip_account:1;
604 #ifdef CONFIG_COMPAT_BRK
605 unsigned brk_randomized:1;
608 unsigned long atomic_flags; /* Flags requiring atomic access. */
610 struct restart_block restart_block;
615 #ifdef CONFIG_CC_STACKPROTECTOR
616 /* Canary value for the -fstack-protector GCC feature: */
617 unsigned long stack_canary;
620 * Pointers to the (original) parent process, youngest child, younger sibling,
621 * older sibling, respectively. (p->father can be replaced with
622 * p->real_parent->pid)
625 /* Real parent process: */
626 struct task_struct __rcu *real_parent;
628 /* Recipient of SIGCHLD, wait4() reports: */
629 struct task_struct __rcu *parent;
632 * Children/sibling form the list of natural children:
634 struct list_head children;
635 struct list_head sibling;
636 struct task_struct *group_leader;
639 * 'ptraced' is the list of tasks this task is using ptrace() on.
641 * This includes both natural children and PTRACE_ATTACH targets.
642 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
644 struct list_head ptraced;
645 struct list_head ptrace_entry;
647 /* PID/PID hash table linkage. */
648 struct pid_link pids[PIDTYPE_MAX];
649 struct list_head thread_group;
650 struct list_head thread_node;
652 struct completion *vfork_done;
654 /* CLONE_CHILD_SETTID: */
655 int __user *set_child_tid;
657 /* CLONE_CHILD_CLEARTID: */
658 int __user *clear_child_tid;
662 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
667 struct prev_cputime prev_cputime;
668 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
669 seqcount_t vtime_seqcount;
670 unsigned long long vtime_snap;
672 /* Task is sleeping or running in a CPU with VTIME inactive: */
674 /* Task runs in userspace in a CPU with VTIME active: */
676 /* Task runs in kernelspace in a CPU with VTIME active: */
681 #ifdef CONFIG_NO_HZ_FULL
682 atomic_t tick_dep_mask;
684 /* Context switch counts: */
686 unsigned long nivcsw;
688 /* Monotonic time in nsecs: */
691 /* Boot based time in nsecs: */
694 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
695 unsigned long min_flt;
696 unsigned long maj_flt;
698 #ifdef CONFIG_POSIX_TIMERS
699 struct task_cputime cputime_expires;
700 struct list_head cpu_timers[3];
703 /* Process credentials: */
705 /* Tracer's credentials at attach: */
706 const struct cred __rcu *ptracer_cred;
708 /* Objective and real subjective task credentials (COW): */
709 const struct cred __rcu *real_cred;
711 /* Effective (overridable) subjective task credentials (COW): */
712 const struct cred __rcu *cred;
715 * executable name, excluding path.
717 * - normally initialized setup_new_exec()
718 * - access it with [gs]et_task_comm()
719 * - lock it with task_lock()
721 char comm[TASK_COMM_LEN];
723 struct nameidata *nameidata;
725 #ifdef CONFIG_SYSVIPC
726 struct sysv_sem sysvsem;
727 struct sysv_shm sysvshm;
729 #ifdef CONFIG_DETECT_HUNG_TASK
730 unsigned long last_switch_count;
732 /* Filesystem information: */
733 struct fs_struct *fs;
735 /* Open file information: */
736 struct files_struct *files;
739 struct nsproxy *nsproxy;
741 /* Signal handlers: */
742 struct signal_struct *signal;
743 struct sighand_struct *sighand;
745 sigset_t real_blocked;
746 /* Restored if set_restore_sigmask() was used: */
747 sigset_t saved_sigmask;
748 struct sigpending pending;
749 unsigned long sas_ss_sp;
751 unsigned int sas_ss_flags;
753 struct callback_head *task_works;
755 struct audit_context *audit_context;
756 #ifdef CONFIG_AUDITSYSCALL
758 unsigned int sessionid;
760 struct seccomp seccomp;
762 /* Thread group tracking: */
766 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
767 spinlock_t alloc_lock;
769 /* Protection of the PI data structures: */
770 raw_spinlock_t pi_lock;
772 struct wake_q_node wake_q;
774 #ifdef CONFIG_RT_MUTEXES
775 /* PI waiters blocked on a rt_mutex held by this task: */
776 struct rb_root pi_waiters;
777 struct rb_node *pi_waiters_leftmost;
778 /* Deadlock detection and priority inheritance handling: */
779 struct rt_mutex_waiter *pi_blocked_on;
782 #ifdef CONFIG_DEBUG_MUTEXES
783 /* Mutex deadlock detection: */
784 struct mutex_waiter *blocked_on;
787 #ifdef CONFIG_TRACE_IRQFLAGS
788 unsigned int irq_events;
789 unsigned long hardirq_enable_ip;
790 unsigned long hardirq_disable_ip;
791 unsigned int hardirq_enable_event;
792 unsigned int hardirq_disable_event;
793 int hardirqs_enabled;
795 unsigned long softirq_disable_ip;
796 unsigned long softirq_enable_ip;
797 unsigned int softirq_disable_event;
798 unsigned int softirq_enable_event;
799 int softirqs_enabled;
803 #ifdef CONFIG_LOCKDEP
804 # define MAX_LOCK_DEPTH 48UL
807 unsigned int lockdep_recursion;
808 struct held_lock held_locks[MAX_LOCK_DEPTH];
809 gfp_t lockdep_reclaim_gfp;
813 unsigned int in_ubsan;
816 /* Journalling filesystem info: */
819 /* Stacked block device info: */
820 struct bio_list *bio_list;
823 /* Stack plugging: */
824 struct blk_plug *plug;
828 struct reclaim_state *reclaim_state;
830 struct backing_dev_info *backing_dev_info;
832 struct io_context *io_context;
835 unsigned long ptrace_message;
836 siginfo_t *last_siginfo;
838 struct task_io_accounting ioac;
839 #ifdef CONFIG_TASK_XACCT
840 /* Accumulated RSS usage: */
842 /* Accumulated virtual memory usage: */
844 /* stime + utime since last update: */
847 #ifdef CONFIG_CPUSETS
848 /* Protected by ->alloc_lock: */
849 nodemask_t mems_allowed;
850 /* Seqence number to catch updates: */
851 seqcount_t mems_allowed_seq;
852 int cpuset_mem_spread_rotor;
853 int cpuset_slab_spread_rotor;
855 #ifdef CONFIG_CGROUPS
856 /* Control Group info protected by css_set_lock: */
857 struct css_set __rcu *cgroups;
858 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
859 struct list_head cg_list;
861 #ifdef CONFIG_INTEL_RDT_A
865 struct robust_list_head __user *robust_list;
867 struct compat_robust_list_head __user *compat_robust_list;
869 struct list_head pi_state_list;
870 struct futex_pi_state *pi_state_cache;
872 #ifdef CONFIG_PERF_EVENTS
873 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
874 struct mutex perf_event_mutex;
875 struct list_head perf_event_list;
877 #ifdef CONFIG_DEBUG_PREEMPT
878 unsigned long preempt_disable_ip;
881 /* Protected by alloc_lock: */
882 struct mempolicy *mempolicy;
884 short pref_node_fork;
886 #ifdef CONFIG_NUMA_BALANCING
888 unsigned int numa_scan_period;
889 unsigned int numa_scan_period_max;
890 int numa_preferred_nid;
891 unsigned long numa_migrate_retry;
892 /* Migration stamp: */
894 u64 last_task_numa_placement;
895 u64 last_sum_exec_runtime;
896 struct callback_head numa_work;
898 struct list_head numa_entry;
899 struct numa_group *numa_group;
902 * numa_faults is an array split into four regions:
903 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
904 * in this precise order.
906 * faults_memory: Exponential decaying average of faults on a per-node
907 * basis. Scheduling placement decisions are made based on these
908 * counts. The values remain static for the duration of a PTE scan.
909 * faults_cpu: Track the nodes the process was running on when a NUMA
910 * hinting fault was incurred.
911 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
912 * during the current scan window. When the scan completes, the counts
913 * in faults_memory and faults_cpu decay and these values are copied.
915 unsigned long *numa_faults;
916 unsigned long total_numa_faults;
919 * numa_faults_locality tracks if faults recorded during the last
920 * scan window were remote/local or failed to migrate. The task scan
921 * period is adapted based on the locality of the faults with different
922 * weights depending on whether they were shared or private faults
924 unsigned long numa_faults_locality[3];
926 unsigned long numa_pages_migrated;
927 #endif /* CONFIG_NUMA_BALANCING */
929 struct tlbflush_unmap_batch tlb_ubc;
933 /* Cache last used pipe for splice(): */
934 struct pipe_inode_info *splice_pipe;
936 struct page_frag task_frag;
938 #ifdef CONFIG_TASK_DELAY_ACCT
939 struct task_delay_info *delays;
942 #ifdef CONFIG_FAULT_INJECTION
946 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
947 * balance_dirty_pages() for a dirty throttling pause:
950 int nr_dirtied_pause;
951 /* Start of a write-and-pause period: */
952 unsigned long dirty_paused_when;
954 #ifdef CONFIG_LATENCYTOP
955 int latency_record_count;
956 struct latency_record latency_record[LT_SAVECOUNT];
959 * Time slack values; these are used to round up poll() and
960 * select() etc timeout values. These are in nanoseconds.
963 u64 default_timer_slack_ns;
966 unsigned int kasan_depth;
969 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
970 /* Index of current stored address in ret_stack: */
973 /* Stack of return addresses for return function tracing: */
974 struct ftrace_ret_stack *ret_stack;
976 /* Timestamp for last schedule: */
977 unsigned long long ftrace_timestamp;
980 * Number of functions that haven't been traced
981 * because of depth overrun:
983 atomic_t trace_overrun;
986 atomic_t tracing_graph_pause;
989 #ifdef CONFIG_TRACING
990 /* State flags for use by tracers: */
993 /* Bitmask and counter of trace recursion: */
994 unsigned long trace_recursion;
995 #endif /* CONFIG_TRACING */
998 /* Coverage collection mode enabled for this task (0 if disabled): */
999 enum kcov_mode kcov_mode;
1001 /* Size of the kcov_area: */
1002 unsigned int kcov_size;
1004 /* Buffer for coverage collection: */
1007 /* KCOV descriptor wired with this task or NULL: */
1012 struct mem_cgroup *memcg_in_oom;
1013 gfp_t memcg_oom_gfp_mask;
1014 int memcg_oom_order;
1016 /* Number of pages to reclaim on returning to userland: */
1017 unsigned int memcg_nr_pages_over_high;
1020 #ifdef CONFIG_UPROBES
1021 struct uprobe_task *utask;
1023 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1024 unsigned int sequential_io;
1025 unsigned int sequential_io_avg;
1027 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1028 unsigned long task_state_change;
1030 int pagefault_disabled;
1032 struct task_struct *oom_reaper_list;
1034 #ifdef CONFIG_VMAP_STACK
1035 struct vm_struct *stack_vm_area;
1037 #ifdef CONFIG_THREAD_INFO_IN_TASK
1038 /* A live task holds one reference: */
1039 atomic_t stack_refcount;
1041 /* CPU-specific state of this task: */
1042 struct thread_struct thread;
1045 * WARNING: on x86, 'thread_struct' contains a variable-sized
1046 * structure. It *MUST* be at the end of 'task_struct'.
1048 * Do not put anything below here!
1052 static inline struct pid *task_pid(struct task_struct *task)
1054 return task->pids[PIDTYPE_PID].pid;
1057 static inline struct pid *task_tgid(struct task_struct *task)
1059 return task->group_leader->pids[PIDTYPE_PID].pid;
1063 * Without tasklist or RCU lock it is not safe to dereference
1064 * the result of task_pgrp/task_session even if task == current,
1065 * we can race with another thread doing sys_setsid/sys_setpgid.
1067 static inline struct pid *task_pgrp(struct task_struct *task)
1069 return task->group_leader->pids[PIDTYPE_PGID].pid;
1072 static inline struct pid *task_session(struct task_struct *task)
1074 return task->group_leader->pids[PIDTYPE_SID].pid;
1078 * the helpers to get the task's different pids as they are seen
1079 * from various namespaces
1081 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1082 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1084 * task_xid_nr_ns() : id seen from the ns specified;
1086 * set_task_vxid() : assigns a virtual id to a task;
1088 * see also pid_nr() etc in include/linux/pid.h
1090 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1092 static inline pid_t task_pid_nr(struct task_struct *tsk)
1097 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1099 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1102 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1104 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1108 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1113 extern pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1115 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1117 return pid_vnr(task_tgid(tsk));
1121 * pid_alive - check that a task structure is not stale
1122 * @p: Task structure to be checked.
1124 * Test if a process is not yet dead (at most zombie state)
1125 * If pid_alive fails, then pointers within the task structure
1126 * can be stale and must not be dereferenced.
1128 * Return: 1 if the process is alive. 0 otherwise.
1130 static inline int pid_alive(const struct task_struct *p)
1132 return p->pids[PIDTYPE_PID].pid != NULL;
1135 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1141 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1147 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1149 return task_ppid_nr_ns(tsk, &init_pid_ns);
1152 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1154 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1157 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1159 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1163 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1165 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1168 static inline pid_t task_session_vnr(struct task_struct *tsk)
1170 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1173 /* Obsolete, do not use: */
1174 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1176 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1180 * is_global_init - check if a task structure is init. Since init
1181 * is free to have sub-threads we need to check tgid.
1182 * @tsk: Task structure to be checked.
1184 * Check if a task structure is the first user space task the kernel created.
1186 * Return: 1 if the task structure is init. 0 otherwise.
1188 static inline int is_global_init(struct task_struct *tsk)
1190 return task_tgid_nr(tsk) == 1;
1193 extern struct pid *cad_pid;
1198 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1199 #define PF_EXITING 0x00000004 /* Getting shut down */
1200 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1201 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1202 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1203 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1204 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1205 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1206 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1207 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1208 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1209 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1210 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1211 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1212 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1213 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1214 #define PF_FSTRANS 0x00020000 /* Inside a filesystem transaction */
1215 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1216 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1217 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1218 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1219 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1220 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1221 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1222 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1223 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1224 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1225 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1228 * Only the _current_ task can read/write to tsk->flags, but other
1229 * tasks can access tsk->flags in readonly mode for example
1230 * with tsk_used_math (like during threaded core dumping).
1231 * There is however an exception to this rule during ptrace
1232 * or during fork: the ptracer task is allowed to write to the
1233 * child->flags of its traced child (same goes for fork, the parent
1234 * can write to the child->flags), because we're guaranteed the
1235 * child is not running and in turn not changing child->flags
1236 * at the same time the parent does it.
1238 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1239 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1240 #define clear_used_math() clear_stopped_child_used_math(current)
1241 #define set_used_math() set_stopped_child_used_math(current)
1243 #define conditional_stopped_child_used_math(condition, child) \
1244 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1246 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1248 #define copy_to_stopped_child_used_math(child) \
1249 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1251 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1252 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1253 #define used_math() tsk_used_math(current)
1255 /* Per-process atomic flags. */
1256 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1257 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1258 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1259 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1262 #define TASK_PFA_TEST(name, func) \
1263 static inline bool task_##func(struct task_struct *p) \
1264 { return test_bit(PFA_##name, &p->atomic_flags); }
1266 #define TASK_PFA_SET(name, func) \
1267 static inline void task_set_##func(struct task_struct *p) \
1268 { set_bit(PFA_##name, &p->atomic_flags); }
1270 #define TASK_PFA_CLEAR(name, func) \
1271 static inline void task_clear_##func(struct task_struct *p) \
1272 { clear_bit(PFA_##name, &p->atomic_flags); }
1274 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1275 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1277 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1278 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1279 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1281 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1282 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1283 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1285 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1286 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1289 tsk_restore_flags(struct task_struct *task, unsigned long orig_flags, unsigned long flags)
1291 task->flags &= ~flags;
1292 task->flags |= orig_flags & flags;
1295 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1296 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1298 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1299 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1301 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1304 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1306 if (!cpumask_test_cpu(0, new_mask))
1312 #ifndef cpu_relax_yield
1313 #define cpu_relax_yield() cpu_relax()
1316 extern int yield_to(struct task_struct *p, bool preempt);
1317 extern void set_user_nice(struct task_struct *p, long nice);
1318 extern int task_prio(const struct task_struct *p);
1321 * task_nice - return the nice value of a given task.
1322 * @p: the task in question.
1324 * Return: The nice value [ -20 ... 0 ... 19 ].
1326 static inline int task_nice(const struct task_struct *p)
1328 return PRIO_TO_NICE((p)->static_prio);
1331 extern int can_nice(const struct task_struct *p, const int nice);
1332 extern int task_curr(const struct task_struct *p);
1333 extern int idle_cpu(int cpu);
1334 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1335 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1336 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1337 extern struct task_struct *idle_task(int cpu);
1340 * is_idle_task - is the specified task an idle task?
1341 * @p: the task in question.
1343 * Return: 1 if @p is an idle task. 0 otherwise.
1345 static inline bool is_idle_task(const struct task_struct *p)
1347 return !!(p->flags & PF_IDLE);
1350 extern struct task_struct *curr_task(int cpu);
1351 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1355 union thread_union {
1356 #ifndef CONFIG_THREAD_INFO_IN_TASK
1357 struct thread_info thread_info;
1359 unsigned long stack[THREAD_SIZE/sizeof(long)];
1362 #ifdef CONFIG_THREAD_INFO_IN_TASK
1363 static inline struct thread_info *task_thread_info(struct task_struct *task)
1365 return &task->thread_info;
1367 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1368 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1372 * find a task by one of its numerical ids
1374 * find_task_by_pid_ns():
1375 * finds a task by its pid in the specified namespace
1376 * find_task_by_vpid():
1377 * finds a task by its virtual pid
1379 * see also find_vpid() etc in include/linux/pid.h
1382 extern struct task_struct *find_task_by_vpid(pid_t nr);
1383 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1385 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1386 extern int wake_up_process(struct task_struct *tsk);
1387 extern void wake_up_new_task(struct task_struct *tsk);
1390 extern void kick_process(struct task_struct *tsk);
1392 static inline void kick_process(struct task_struct *tsk) { }
1395 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1397 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1399 __set_task_comm(tsk, from, false);
1402 extern char *get_task_comm(char *to, struct task_struct *tsk);
1405 void scheduler_ipi(void);
1406 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1408 static inline void scheduler_ipi(void) { }
1409 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1416 * Set thread flags in other task's structures.
1417 * See asm/thread_info.h for TIF_xxxx flags available:
1419 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1421 set_ti_thread_flag(task_thread_info(tsk), flag);
1424 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1426 clear_ti_thread_flag(task_thread_info(tsk), flag);
1429 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1431 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1434 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1436 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1439 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1441 return test_ti_thread_flag(task_thread_info(tsk), flag);
1444 static inline void set_tsk_need_resched(struct task_struct *tsk)
1446 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1449 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1451 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1454 static inline int test_tsk_need_resched(struct task_struct *tsk)
1456 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1460 * cond_resched() and cond_resched_lock(): latency reduction via
1461 * explicit rescheduling in places that are safe. The return
1462 * value indicates whether a reschedule was done in fact.
1463 * cond_resched_lock() will drop the spinlock before scheduling,
1464 * cond_resched_softirq() will enable bhs before scheduling.
1466 #ifndef CONFIG_PREEMPT
1467 extern int _cond_resched(void);
1469 static inline int _cond_resched(void) { return 0; }
1472 #define cond_resched() ({ \
1473 ___might_sleep(__FILE__, __LINE__, 0); \
1477 extern int __cond_resched_lock(spinlock_t *lock);
1479 #define cond_resched_lock(lock) ({ \
1480 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1481 __cond_resched_lock(lock); \
1484 extern int __cond_resched_softirq(void);
1486 #define cond_resched_softirq() ({ \
1487 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1488 __cond_resched_softirq(); \
1491 static inline void cond_resched_rcu(void)
1493 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1501 * Does a critical section need to be broken due to another
1502 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1503 * but a general need for low latency)
1505 static inline int spin_needbreak(spinlock_t *lock)
1507 #ifdef CONFIG_PREEMPT
1508 return spin_is_contended(lock);
1514 static __always_inline bool need_resched(void)
1516 return unlikely(tif_need_resched());
1520 * Wrappers for p->thread_info->cpu access. No-op on UP.
1524 static inline unsigned int task_cpu(const struct task_struct *p)
1526 #ifdef CONFIG_THREAD_INFO_IN_TASK
1529 return task_thread_info(p)->cpu;
1533 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1537 static inline unsigned int task_cpu(const struct task_struct *p)
1542 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1546 #endif /* CONFIG_SMP */
1549 * In order to reduce various lock holder preemption latencies provide an
1550 * interface to see if a vCPU is currently running or not.
1552 * This allows us to terminate optimistic spin loops and block, analogous to
1553 * the native optimistic spin heuristic of testing if the lock owner task is
1556 #ifndef vcpu_is_preempted
1557 # define vcpu_is_preempted(cpu) false
1560 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1561 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1563 #ifndef TASK_SIZE_OF
1564 #define TASK_SIZE_OF(tsk) TASK_SIZE