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 - snapshot 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 /* Task is sleeping or running in a CPU with VTIME inactive: */
229 /* Task runs in userspace in a CPU with VTIME active: */
231 /* Task runs in kernelspace in a CPU with VTIME active: */
237 unsigned long long starttime;
238 enum vtime_state state;
242 #ifdef CONFIG_SCHED_INFO
243 /* Cumulative counters: */
245 /* # of times we have run on this CPU: */
246 unsigned long pcount;
248 /* Time spent waiting on a runqueue: */
249 unsigned long long run_delay;
253 /* When did we last run on a CPU? */
254 unsigned long long last_arrival;
256 /* When were we last queued to run? */
257 unsigned long long last_queued;
259 #endif /* CONFIG_SCHED_INFO */
263 * Integer metrics need fixed point arithmetic, e.g., sched/fair
264 * has a few: load, load_avg, util_avg, freq, and capacity.
266 * We define a basic fixed point arithmetic range, and then formalize
267 * all these metrics based on that basic range.
269 # define SCHED_FIXEDPOINT_SHIFT 10
270 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
273 unsigned long weight;
278 * The load_avg/util_avg accumulates an infinite geometric series
279 * (see __update_load_avg() in kernel/sched/fair.c).
281 * [load_avg definition]
283 * load_avg = runnable% * scale_load_down(load)
285 * where runnable% is the time ratio that a sched_entity is runnable.
286 * For cfs_rq, it is the aggregated load_avg of all runnable and
287 * blocked sched_entities.
289 * load_avg may also take frequency scaling into account:
291 * load_avg = runnable% * scale_load_down(load) * freq%
293 * where freq% is the CPU frequency normalized to the highest frequency.
295 * [util_avg definition]
297 * util_avg = running% * SCHED_CAPACITY_SCALE
299 * where running% is the time ratio that a sched_entity is running on
300 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
301 * and blocked sched_entities.
303 * util_avg may also factor frequency scaling and CPU capacity scaling:
305 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
307 * where freq% is the same as above, and capacity% is the CPU capacity
308 * normalized to the greatest capacity (due to uarch differences, etc).
310 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
311 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
312 * we therefore scale them to as large a range as necessary. This is for
313 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
317 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
318 * with the highest load (=88761), always runnable on a single cfs_rq,
319 * and should not overflow as the number already hits PID_MAX_LIMIT.
321 * For all other cases (including 32-bit kernels), struct load_weight's
322 * weight will overflow first before we do, because:
324 * Max(load_avg) <= Max(load.weight)
326 * Then it is the load_weight's responsibility to consider overflow
330 u64 last_update_time;
334 unsigned long load_avg;
335 unsigned long util_avg;
338 struct sched_statistics {
339 #ifdef CONFIG_SCHEDSTATS
349 s64 sum_sleep_runtime;
356 u64 nr_migrations_cold;
357 u64 nr_failed_migrations_affine;
358 u64 nr_failed_migrations_running;
359 u64 nr_failed_migrations_hot;
360 u64 nr_forced_migrations;
364 u64 nr_wakeups_migrate;
365 u64 nr_wakeups_local;
366 u64 nr_wakeups_remote;
367 u64 nr_wakeups_affine;
368 u64 nr_wakeups_affine_attempts;
369 u64 nr_wakeups_passive;
374 struct sched_entity {
375 /* For load-balancing: */
376 struct load_weight load;
377 struct rb_node run_node;
378 struct list_head group_node;
382 u64 sum_exec_runtime;
384 u64 prev_sum_exec_runtime;
388 struct sched_statistics statistics;
390 #ifdef CONFIG_FAIR_GROUP_SCHED
392 struct sched_entity *parent;
393 /* rq on which this entity is (to be) queued: */
394 struct cfs_rq *cfs_rq;
395 /* rq "owned" by this entity/group: */
401 * Per entity load average tracking.
403 * Put into separate cache line so it does not
404 * collide with read-mostly values above.
406 struct sched_avg avg ____cacheline_aligned_in_smp;
410 struct sched_rt_entity {
411 struct list_head run_list;
412 unsigned long timeout;
413 unsigned long watchdog_stamp;
414 unsigned int time_slice;
415 unsigned short on_rq;
416 unsigned short on_list;
418 struct sched_rt_entity *back;
419 #ifdef CONFIG_RT_GROUP_SCHED
420 struct sched_rt_entity *parent;
421 /* rq on which this entity is (to be) queued: */
423 /* rq "owned" by this entity/group: */
428 struct sched_dl_entity {
429 struct rb_node rb_node;
432 * Original scheduling parameters. Copied here from sched_attr
433 * during sched_setattr(), they will remain the same until
434 * the next sched_setattr().
436 u64 dl_runtime; /* Maximum runtime for each instance */
437 u64 dl_deadline; /* Relative deadline of each instance */
438 u64 dl_period; /* Separation of two instances (period) */
439 u64 dl_bw; /* dl_runtime / dl_period */
440 u64 dl_density; /* dl_runtime / dl_deadline */
443 * Actual scheduling parameters. Initialized with the values above,
444 * they are continously updated during task execution. Note that
445 * the remaining runtime could be < 0 in case we are in overrun.
447 s64 runtime; /* Remaining runtime for this instance */
448 u64 deadline; /* Absolute deadline for this instance */
449 unsigned int flags; /* Specifying the scheduler behaviour */
454 * @dl_throttled tells if we exhausted the runtime. If so, the
455 * task has to wait for a replenishment to be performed at the
456 * next firing of dl_timer.
458 * @dl_boosted tells if we are boosted due to DI. If so we are
459 * outside bandwidth enforcement mechanism (but only until we
460 * exit the critical section);
462 * @dl_yielded tells if task gave up the CPU before consuming
463 * all its available runtime during the last job.
465 * @dl_non_contending tells if the task is inactive while still
466 * contributing to the active utilization. In other words, it
467 * indicates if the inactive timer has been armed and its handler
468 * has not been executed yet. This flag is useful to avoid race
469 * conditions between the inactive timer handler and the wakeup
475 int dl_non_contending;
478 * Bandwidth enforcement timer. Each -deadline task has its
479 * own bandwidth to be enforced, thus we need one timer per task.
481 struct hrtimer dl_timer;
484 * Inactive timer, responsible for decreasing the active utilization
485 * at the "0-lag time". When a -deadline task blocks, it contributes
486 * to GRUB's active utilization until the "0-lag time", hence a
487 * timer is needed to decrease the active utilization at the correct
490 struct hrtimer inactive_timer;
499 /* Otherwise the compiler can store garbage here: */
502 u32 s; /* Set of bits. */
505 enum perf_event_task_context {
506 perf_invalid_context = -1,
509 perf_nr_task_contexts,
513 struct wake_q_node *next;
517 #ifdef CONFIG_THREAD_INFO_IN_TASK
519 * For reasons of header soup (see current_thread_info()), this
520 * must be the first element of task_struct.
522 struct thread_info thread_info;
524 /* -1 unrunnable, 0 runnable, >0 stopped: */
528 /* Per task flags (PF_*), defined further below: */
533 struct llist_node wake_entry;
535 #ifdef CONFIG_THREAD_INFO_IN_TASK
539 unsigned int wakee_flips;
540 unsigned long wakee_flip_decay_ts;
541 struct task_struct *last_wakee;
550 unsigned int rt_priority;
552 const struct sched_class *sched_class;
553 struct sched_entity se;
554 struct sched_rt_entity rt;
555 #ifdef CONFIG_CGROUP_SCHED
556 struct task_group *sched_task_group;
558 struct sched_dl_entity dl;
560 #ifdef CONFIG_PREEMPT_NOTIFIERS
561 /* List of struct preempt_notifier: */
562 struct hlist_head preempt_notifiers;
565 #ifdef CONFIG_BLK_DEV_IO_TRACE
566 unsigned int btrace_seq;
571 cpumask_t cpus_allowed;
573 #ifdef CONFIG_PREEMPT_RCU
574 int rcu_read_lock_nesting;
575 union rcu_special rcu_read_unlock_special;
576 struct list_head rcu_node_entry;
577 struct rcu_node *rcu_blocked_node;
578 #endif /* #ifdef CONFIG_PREEMPT_RCU */
580 #ifdef CONFIG_TASKS_RCU
581 unsigned long rcu_tasks_nvcsw;
582 bool rcu_tasks_holdout;
583 struct list_head rcu_tasks_holdout_list;
584 int rcu_tasks_idle_cpu;
585 #endif /* #ifdef CONFIG_TASKS_RCU */
587 struct sched_info sched_info;
589 struct list_head tasks;
591 struct plist_node pushable_tasks;
592 struct rb_node pushable_dl_tasks;
595 struct mm_struct *mm;
596 struct mm_struct *active_mm;
598 /* Per-thread vma caching: */
599 struct vmacache vmacache;
601 #ifdef SPLIT_RSS_COUNTING
602 struct task_rss_stat rss_stat;
607 /* The signal sent when the parent dies: */
609 /* JOBCTL_*, siglock protected: */
610 unsigned long jobctl;
612 /* Used for emulating ABI behavior of previous Linux versions: */
613 unsigned int personality;
615 /* Scheduler bits, serialized by scheduler locks: */
616 unsigned sched_reset_on_fork:1;
617 unsigned sched_contributes_to_load:1;
618 unsigned sched_migrated:1;
619 unsigned sched_remote_wakeup:1;
620 /* Force alignment to the next boundary: */
623 /* Unserialized, strictly 'current' */
625 /* Bit to tell LSMs we're in execve(): */
626 unsigned in_execve:1;
627 unsigned in_iowait:1;
628 #ifndef TIF_RESTORE_SIGMASK
629 unsigned restore_sigmask:1;
632 unsigned memcg_may_oom:1;
634 unsigned memcg_kmem_skip_account:1;
637 #ifdef CONFIG_COMPAT_BRK
638 unsigned brk_randomized:1;
640 #ifdef CONFIG_CGROUPS
641 /* disallow userland-initiated cgroup migration */
642 unsigned no_cgroup_migration:1;
645 unsigned long atomic_flags; /* Flags requiring atomic access. */
647 struct restart_block restart_block;
652 #ifdef CONFIG_CC_STACKPROTECTOR
653 /* Canary value for the -fstack-protector GCC feature: */
654 unsigned long stack_canary;
657 * Pointers to the (original) parent process, youngest child, younger sibling,
658 * older sibling, respectively. (p->father can be replaced with
659 * p->real_parent->pid)
662 /* Real parent process: */
663 struct task_struct __rcu *real_parent;
665 /* Recipient of SIGCHLD, wait4() reports: */
666 struct task_struct __rcu *parent;
669 * Children/sibling form the list of natural children:
671 struct list_head children;
672 struct list_head sibling;
673 struct task_struct *group_leader;
676 * 'ptraced' is the list of tasks this task is using ptrace() on.
678 * This includes both natural children and PTRACE_ATTACH targets.
679 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
681 struct list_head ptraced;
682 struct list_head ptrace_entry;
684 /* PID/PID hash table linkage. */
685 struct pid_link pids[PIDTYPE_MAX];
686 struct list_head thread_group;
687 struct list_head thread_node;
689 struct completion *vfork_done;
691 /* CLONE_CHILD_SETTID: */
692 int __user *set_child_tid;
694 /* CLONE_CHILD_CLEARTID: */
695 int __user *clear_child_tid;
699 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
704 struct prev_cputime prev_cputime;
705 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
709 #ifdef CONFIG_NO_HZ_FULL
710 atomic_t tick_dep_mask;
712 /* Context switch counts: */
714 unsigned long nivcsw;
716 /* Monotonic time in nsecs: */
719 /* Boot based time in nsecs: */
722 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
723 unsigned long min_flt;
724 unsigned long maj_flt;
726 #ifdef CONFIG_POSIX_TIMERS
727 struct task_cputime cputime_expires;
728 struct list_head cpu_timers[3];
731 /* Process credentials: */
733 /* Tracer's credentials at attach: */
734 const struct cred __rcu *ptracer_cred;
736 /* Objective and real subjective task credentials (COW): */
737 const struct cred __rcu *real_cred;
739 /* Effective (overridable) subjective task credentials (COW): */
740 const struct cred __rcu *cred;
743 * executable name, excluding path.
745 * - normally initialized setup_new_exec()
746 * - access it with [gs]et_task_comm()
747 * - lock it with task_lock()
749 char comm[TASK_COMM_LEN];
751 struct nameidata *nameidata;
753 #ifdef CONFIG_SYSVIPC
754 struct sysv_sem sysvsem;
755 struct sysv_shm sysvshm;
757 #ifdef CONFIG_DETECT_HUNG_TASK
758 unsigned long last_switch_count;
760 /* Filesystem information: */
761 struct fs_struct *fs;
763 /* Open file information: */
764 struct files_struct *files;
767 struct nsproxy *nsproxy;
769 /* Signal handlers: */
770 struct signal_struct *signal;
771 struct sighand_struct *sighand;
773 sigset_t real_blocked;
774 /* Restored if set_restore_sigmask() was used: */
775 sigset_t saved_sigmask;
776 struct sigpending pending;
777 unsigned long sas_ss_sp;
779 unsigned int sas_ss_flags;
781 struct callback_head *task_works;
783 struct audit_context *audit_context;
784 #ifdef CONFIG_AUDITSYSCALL
786 unsigned int sessionid;
788 struct seccomp seccomp;
790 /* Thread group tracking: */
794 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
795 spinlock_t alloc_lock;
797 /* Protection of the PI data structures: */
798 raw_spinlock_t pi_lock;
800 struct wake_q_node wake_q;
802 #ifdef CONFIG_RT_MUTEXES
803 /* PI waiters blocked on a rt_mutex held by this task: */
804 struct rb_root pi_waiters;
805 struct rb_node *pi_waiters_leftmost;
806 /* Updated under owner's pi_lock and rq lock */
807 struct task_struct *pi_top_task;
808 /* Deadlock detection and priority inheritance handling: */
809 struct rt_mutex_waiter *pi_blocked_on;
812 #ifdef CONFIG_DEBUG_MUTEXES
813 /* Mutex deadlock detection: */
814 struct mutex_waiter *blocked_on;
817 #ifdef CONFIG_TRACE_IRQFLAGS
818 unsigned int irq_events;
819 unsigned long hardirq_enable_ip;
820 unsigned long hardirq_disable_ip;
821 unsigned int hardirq_enable_event;
822 unsigned int hardirq_disable_event;
823 int hardirqs_enabled;
825 unsigned long softirq_disable_ip;
826 unsigned long softirq_enable_ip;
827 unsigned int softirq_disable_event;
828 unsigned int softirq_enable_event;
829 int softirqs_enabled;
833 #ifdef CONFIG_LOCKDEP
834 # define MAX_LOCK_DEPTH 48UL
837 unsigned int lockdep_recursion;
838 struct held_lock held_locks[MAX_LOCK_DEPTH];
839 gfp_t lockdep_reclaim_gfp;
843 unsigned int in_ubsan;
846 /* Journalling filesystem info: */
849 /* Stacked block device info: */
850 struct bio_list *bio_list;
853 /* Stack plugging: */
854 struct blk_plug *plug;
858 struct reclaim_state *reclaim_state;
860 struct backing_dev_info *backing_dev_info;
862 struct io_context *io_context;
865 unsigned long ptrace_message;
866 siginfo_t *last_siginfo;
868 struct task_io_accounting ioac;
869 #ifdef CONFIG_TASK_XACCT
870 /* Accumulated RSS usage: */
872 /* Accumulated virtual memory usage: */
874 /* stime + utime since last update: */
877 #ifdef CONFIG_CPUSETS
878 /* Protected by ->alloc_lock: */
879 nodemask_t mems_allowed;
880 /* Seqence number to catch updates: */
881 seqcount_t mems_allowed_seq;
882 int cpuset_mem_spread_rotor;
883 int cpuset_slab_spread_rotor;
885 #ifdef CONFIG_CGROUPS
886 /* Control Group info protected by css_set_lock: */
887 struct css_set __rcu *cgroups;
888 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
889 struct list_head cg_list;
891 #ifdef CONFIG_INTEL_RDT_A
895 struct robust_list_head __user *robust_list;
897 struct compat_robust_list_head __user *compat_robust_list;
899 struct list_head pi_state_list;
900 struct futex_pi_state *pi_state_cache;
902 #ifdef CONFIG_PERF_EVENTS
903 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
904 struct mutex perf_event_mutex;
905 struct list_head perf_event_list;
907 #ifdef CONFIG_DEBUG_PREEMPT
908 unsigned long preempt_disable_ip;
911 /* Protected by alloc_lock: */
912 struct mempolicy *mempolicy;
914 short pref_node_fork;
916 #ifdef CONFIG_NUMA_BALANCING
918 unsigned int numa_scan_period;
919 unsigned int numa_scan_period_max;
920 int numa_preferred_nid;
921 unsigned long numa_migrate_retry;
922 /* Migration stamp: */
924 u64 last_task_numa_placement;
925 u64 last_sum_exec_runtime;
926 struct callback_head numa_work;
928 struct list_head numa_entry;
929 struct numa_group *numa_group;
932 * numa_faults is an array split into four regions:
933 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
934 * in this precise order.
936 * faults_memory: Exponential decaying average of faults on a per-node
937 * basis. Scheduling placement decisions are made based on these
938 * counts. The values remain static for the duration of a PTE scan.
939 * faults_cpu: Track the nodes the process was running on when a NUMA
940 * hinting fault was incurred.
941 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
942 * during the current scan window. When the scan completes, the counts
943 * in faults_memory and faults_cpu decay and these values are copied.
945 unsigned long *numa_faults;
946 unsigned long total_numa_faults;
949 * numa_faults_locality tracks if faults recorded during the last
950 * scan window were remote/local or failed to migrate. The task scan
951 * period is adapted based on the locality of the faults with different
952 * weights depending on whether they were shared or private faults
954 unsigned long numa_faults_locality[3];
956 unsigned long numa_pages_migrated;
957 #endif /* CONFIG_NUMA_BALANCING */
959 struct tlbflush_unmap_batch tlb_ubc;
963 /* Cache last used pipe for splice(): */
964 struct pipe_inode_info *splice_pipe;
966 struct page_frag task_frag;
968 #ifdef CONFIG_TASK_DELAY_ACCT
969 struct task_delay_info *delays;
972 #ifdef CONFIG_FAULT_INJECTION
976 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
977 * balance_dirty_pages() for a dirty throttling pause:
980 int nr_dirtied_pause;
981 /* Start of a write-and-pause period: */
982 unsigned long dirty_paused_when;
984 #ifdef CONFIG_LATENCYTOP
985 int latency_record_count;
986 struct latency_record latency_record[LT_SAVECOUNT];
989 * Time slack values; these are used to round up poll() and
990 * select() etc timeout values. These are in nanoseconds.
993 u64 default_timer_slack_ns;
996 unsigned int kasan_depth;
999 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1000 /* Index of current stored address in ret_stack: */
1003 /* Stack of return addresses for return function tracing: */
1004 struct ftrace_ret_stack *ret_stack;
1006 /* Timestamp for last schedule: */
1007 unsigned long long ftrace_timestamp;
1010 * Number of functions that haven't been traced
1011 * because of depth overrun:
1013 atomic_t trace_overrun;
1015 /* Pause tracing: */
1016 atomic_t tracing_graph_pause;
1019 #ifdef CONFIG_TRACING
1020 /* State flags for use by tracers: */
1021 unsigned long trace;
1023 /* Bitmask and counter of trace recursion: */
1024 unsigned long trace_recursion;
1025 #endif /* CONFIG_TRACING */
1028 /* Coverage collection mode enabled for this task (0 if disabled): */
1029 enum kcov_mode kcov_mode;
1031 /* Size of the kcov_area: */
1032 unsigned int kcov_size;
1034 /* Buffer for coverage collection: */
1037 /* KCOV descriptor wired with this task or NULL: */
1042 struct mem_cgroup *memcg_in_oom;
1043 gfp_t memcg_oom_gfp_mask;
1044 int memcg_oom_order;
1046 /* Number of pages to reclaim on returning to userland: */
1047 unsigned int memcg_nr_pages_over_high;
1050 #ifdef CONFIG_UPROBES
1051 struct uprobe_task *utask;
1053 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1054 unsigned int sequential_io;
1055 unsigned int sequential_io_avg;
1057 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1058 unsigned long task_state_change;
1060 int pagefault_disabled;
1062 struct task_struct *oom_reaper_list;
1064 #ifdef CONFIG_VMAP_STACK
1065 struct vm_struct *stack_vm_area;
1067 #ifdef CONFIG_THREAD_INFO_IN_TASK
1068 /* A live task holds one reference: */
1069 atomic_t stack_refcount;
1071 #ifdef CONFIG_LIVEPATCH
1074 #ifdef CONFIG_SECURITY
1075 /* Used by LSM modules for access restriction: */
1078 /* CPU-specific state of this task: */
1079 struct thread_struct thread;
1082 * WARNING: on x86, 'thread_struct' contains a variable-sized
1083 * structure. It *MUST* be at the end of 'task_struct'.
1085 * Do not put anything below here!
1089 static inline struct pid *task_pid(struct task_struct *task)
1091 return task->pids[PIDTYPE_PID].pid;
1094 static inline struct pid *task_tgid(struct task_struct *task)
1096 return task->group_leader->pids[PIDTYPE_PID].pid;
1100 * Without tasklist or RCU lock it is not safe to dereference
1101 * the result of task_pgrp/task_session even if task == current,
1102 * we can race with another thread doing sys_setsid/sys_setpgid.
1104 static inline struct pid *task_pgrp(struct task_struct *task)
1106 return task->group_leader->pids[PIDTYPE_PGID].pid;
1109 static inline struct pid *task_session(struct task_struct *task)
1111 return task->group_leader->pids[PIDTYPE_SID].pid;
1115 * the helpers to get the task's different pids as they are seen
1116 * from various namespaces
1118 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1119 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1121 * task_xid_nr_ns() : id seen from the ns specified;
1123 * see also pid_nr() etc in include/linux/pid.h
1125 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1127 static inline pid_t task_pid_nr(struct task_struct *tsk)
1132 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1134 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1137 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1139 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1143 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1148 extern pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1150 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1152 return pid_vnr(task_tgid(tsk));
1156 * pid_alive - check that a task structure is not stale
1157 * @p: Task structure to be checked.
1159 * Test if a process is not yet dead (at most zombie state)
1160 * If pid_alive fails, then pointers within the task structure
1161 * can be stale and must not be dereferenced.
1163 * Return: 1 if the process is alive. 0 otherwise.
1165 static inline int pid_alive(const struct task_struct *p)
1167 return p->pids[PIDTYPE_PID].pid != NULL;
1170 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1176 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1182 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1184 return task_ppid_nr_ns(tsk, &init_pid_ns);
1187 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1189 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1192 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1194 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1198 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1200 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1203 static inline pid_t task_session_vnr(struct task_struct *tsk)
1205 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1208 /* Obsolete, do not use: */
1209 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1211 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1215 * is_global_init - check if a task structure is init. Since init
1216 * is free to have sub-threads we need to check tgid.
1217 * @tsk: Task structure to be checked.
1219 * Check if a task structure is the first user space task the kernel created.
1221 * Return: 1 if the task structure is init. 0 otherwise.
1223 static inline int is_global_init(struct task_struct *tsk)
1225 return task_tgid_nr(tsk) == 1;
1228 extern struct pid *cad_pid;
1233 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1234 #define PF_EXITING 0x00000004 /* Getting shut down */
1235 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1236 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1237 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1238 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1239 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1240 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1241 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1242 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1243 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1244 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1245 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1246 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1247 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1248 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1249 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1250 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1251 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1252 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1253 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1254 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1255 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1256 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1257 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1258 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1259 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1260 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1263 * Only the _current_ task can read/write to tsk->flags, but other
1264 * tasks can access tsk->flags in readonly mode for example
1265 * with tsk_used_math (like during threaded core dumping).
1266 * There is however an exception to this rule during ptrace
1267 * or during fork: the ptracer task is allowed to write to the
1268 * child->flags of its traced child (same goes for fork, the parent
1269 * can write to the child->flags), because we're guaranteed the
1270 * child is not running and in turn not changing child->flags
1271 * at the same time the parent does it.
1273 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1274 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1275 #define clear_used_math() clear_stopped_child_used_math(current)
1276 #define set_used_math() set_stopped_child_used_math(current)
1278 #define conditional_stopped_child_used_math(condition, child) \
1279 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1281 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1283 #define copy_to_stopped_child_used_math(child) \
1284 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1286 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1287 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1288 #define used_math() tsk_used_math(current)
1290 static inline bool is_percpu_thread(void)
1293 return (current->flags & PF_NO_SETAFFINITY) &&
1294 (current->nr_cpus_allowed == 1);
1300 /* Per-process atomic flags. */
1301 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1302 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1303 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1306 #define TASK_PFA_TEST(name, func) \
1307 static inline bool task_##func(struct task_struct *p) \
1308 { return test_bit(PFA_##name, &p->atomic_flags); }
1310 #define TASK_PFA_SET(name, func) \
1311 static inline void task_set_##func(struct task_struct *p) \
1312 { set_bit(PFA_##name, &p->atomic_flags); }
1314 #define TASK_PFA_CLEAR(name, func) \
1315 static inline void task_clear_##func(struct task_struct *p) \
1316 { clear_bit(PFA_##name, &p->atomic_flags); }
1318 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1319 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1321 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1322 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1323 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1325 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1326 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1327 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1330 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1332 current->flags &= ~flags;
1333 current->flags |= orig_flags & flags;
1336 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1337 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1339 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1340 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1342 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1345 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1347 if (!cpumask_test_cpu(0, new_mask))
1353 #ifndef cpu_relax_yield
1354 #define cpu_relax_yield() cpu_relax()
1357 extern int yield_to(struct task_struct *p, bool preempt);
1358 extern void set_user_nice(struct task_struct *p, long nice);
1359 extern int task_prio(const struct task_struct *p);
1362 * task_nice - return the nice value of a given task.
1363 * @p: the task in question.
1365 * Return: The nice value [ -20 ... 0 ... 19 ].
1367 static inline int task_nice(const struct task_struct *p)
1369 return PRIO_TO_NICE((p)->static_prio);
1372 extern int can_nice(const struct task_struct *p, const int nice);
1373 extern int task_curr(const struct task_struct *p);
1374 extern int idle_cpu(int cpu);
1375 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1376 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1377 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1378 extern struct task_struct *idle_task(int cpu);
1381 * is_idle_task - is the specified task an idle task?
1382 * @p: the task in question.
1384 * Return: 1 if @p is an idle task. 0 otherwise.
1386 static inline bool is_idle_task(const struct task_struct *p)
1388 return !!(p->flags & PF_IDLE);
1391 extern struct task_struct *curr_task(int cpu);
1392 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1396 union thread_union {
1397 #ifndef CONFIG_THREAD_INFO_IN_TASK
1398 struct thread_info thread_info;
1400 unsigned long stack[THREAD_SIZE/sizeof(long)];
1403 #ifdef CONFIG_THREAD_INFO_IN_TASK
1404 static inline struct thread_info *task_thread_info(struct task_struct *task)
1406 return &task->thread_info;
1408 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1409 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1413 * find a task by one of its numerical ids
1415 * find_task_by_pid_ns():
1416 * finds a task by its pid in the specified namespace
1417 * find_task_by_vpid():
1418 * finds a task by its virtual pid
1420 * see also find_vpid() etc in include/linux/pid.h
1423 extern struct task_struct *find_task_by_vpid(pid_t nr);
1424 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1426 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1427 extern int wake_up_process(struct task_struct *tsk);
1428 extern void wake_up_new_task(struct task_struct *tsk);
1431 extern void kick_process(struct task_struct *tsk);
1433 static inline void kick_process(struct task_struct *tsk) { }
1436 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1438 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1440 __set_task_comm(tsk, from, false);
1443 extern char *get_task_comm(char *to, struct task_struct *tsk);
1446 void scheduler_ipi(void);
1447 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1449 static inline void scheduler_ipi(void) { }
1450 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1457 * Set thread flags in other task's structures.
1458 * See asm/thread_info.h for TIF_xxxx flags available:
1460 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1462 set_ti_thread_flag(task_thread_info(tsk), flag);
1465 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1467 clear_ti_thread_flag(task_thread_info(tsk), flag);
1470 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1472 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1475 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1477 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1480 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1482 return test_ti_thread_flag(task_thread_info(tsk), flag);
1485 static inline void set_tsk_need_resched(struct task_struct *tsk)
1487 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1490 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1492 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1495 static inline int test_tsk_need_resched(struct task_struct *tsk)
1497 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1501 * cond_resched() and cond_resched_lock(): latency reduction via
1502 * explicit rescheduling in places that are safe. The return
1503 * value indicates whether a reschedule was done in fact.
1504 * cond_resched_lock() will drop the spinlock before scheduling,
1505 * cond_resched_softirq() will enable bhs before scheduling.
1507 #ifndef CONFIG_PREEMPT
1508 extern int _cond_resched(void);
1510 static inline int _cond_resched(void) { return 0; }
1513 #define cond_resched() ({ \
1514 ___might_sleep(__FILE__, __LINE__, 0); \
1518 extern int __cond_resched_lock(spinlock_t *lock);
1520 #define cond_resched_lock(lock) ({ \
1521 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1522 __cond_resched_lock(lock); \
1525 extern int __cond_resched_softirq(void);
1527 #define cond_resched_softirq() ({ \
1528 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1529 __cond_resched_softirq(); \
1532 static inline void cond_resched_rcu(void)
1534 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1542 * Does a critical section need to be broken due to another
1543 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1544 * but a general need for low latency)
1546 static inline int spin_needbreak(spinlock_t *lock)
1548 #ifdef CONFIG_PREEMPT
1549 return spin_is_contended(lock);
1555 static __always_inline bool need_resched(void)
1557 return unlikely(tif_need_resched());
1561 * Wrappers for p->thread_info->cpu access. No-op on UP.
1565 static inline unsigned int task_cpu(const struct task_struct *p)
1567 #ifdef CONFIG_THREAD_INFO_IN_TASK
1570 return task_thread_info(p)->cpu;
1574 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1578 static inline unsigned int task_cpu(const struct task_struct *p)
1583 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1587 #endif /* CONFIG_SMP */
1590 * In order to reduce various lock holder preemption latencies provide an
1591 * interface to see if a vCPU is currently running or not.
1593 * This allows us to terminate optimistic spin loops and block, analogous to
1594 * the native optimistic spin heuristic of testing if the lock owner task is
1597 #ifndef vcpu_is_preempted
1598 # define vcpu_is_preempted(cpu) false
1601 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1602 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1604 #ifndef TASK_SIZE_OF
1605 #define TASK_SIZE_OF(tsk) TASK_SIZE