4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
8 #include <linux/mutex.h>
9 #include <linux/plist.h>
10 #include <linux/mm_types_task.h>
11 #include <asm/ptrace.h>
13 #include <linux/sem.h>
14 #include <linux/shm.h>
15 #include <linux/signal.h>
16 #include <linux/signal_types.h>
17 #include <linux/pid.h>
18 #include <linux/seccomp.h>
19 #include <linux/rculist.h>
20 #include <linux/rtmutex.h>
22 #include <linux/resource.h>
23 #include <linux/hrtimer.h>
24 #include <linux/kcov.h>
25 #include <linux/task_io_accounting.h>
26 #include <linux/latencytop.h>
27 #include <linux/cred.h>
28 #include <linux/gfp.h>
29 #include <linux/topology.h>
30 #include <linux/magic.h>
31 #include <linux/cgroup-defs.h>
33 #include <asm/current.h>
35 /* task_struct member predeclarations: */
38 struct backing_dev_info;
44 struct futex_pi_state;
49 struct perf_event_context;
51 struct pipe_inode_info;
54 struct robust_list_head;
58 struct sighand_struct;
60 struct task_delay_info;
66 * Task state bitmask. NOTE! These bits are also
67 * encoded in fs/proc/array.c: get_task_state().
69 * We have two separate sets of flags: task->state
70 * is about runnability, while task->exit_state are
71 * about the task exiting. Confusing, but this way
72 * modifying one set can't modify the other one by
75 #define TASK_RUNNING 0
76 #define TASK_INTERRUPTIBLE 1
77 #define TASK_UNINTERRUPTIBLE 2
78 #define __TASK_STOPPED 4
79 #define __TASK_TRACED 8
80 /* in tsk->exit_state */
82 #define EXIT_ZOMBIE 32
83 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
84 /* in tsk->state again */
86 #define TASK_WAKEKILL 128
87 #define TASK_WAKING 256
88 #define TASK_PARKED 512
89 #define TASK_NOLOAD 1024
91 #define TASK_STATE_MAX 4096
93 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
95 /* Convenience macros for the sake of set_current_state */
96 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
97 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
98 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
100 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
102 /* Convenience macros for the sake of wake_up */
103 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
104 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
106 /* get_task_state() */
107 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
108 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
109 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
111 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
112 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
113 #define task_is_stopped_or_traced(task) \
114 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
115 #define task_contributes_to_load(task) \
116 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
117 (task->flags & PF_FROZEN) == 0 && \
118 (task->state & TASK_NOLOAD) == 0)
120 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
122 #define __set_current_state(state_value) \
124 current->task_state_change = _THIS_IP_; \
125 current->state = (state_value); \
127 #define set_current_state(state_value) \
129 current->task_state_change = _THIS_IP_; \
130 smp_store_mb(current->state, (state_value)); \
135 * set_current_state() includes a barrier so that the write of current->state
136 * is correctly serialised wrt the caller's subsequent test of whether to
140 * set_current_state(TASK_UNINTERRUPTIBLE);
146 * __set_current_state(TASK_RUNNING);
148 * If the caller does not need such serialisation (because, for instance, the
149 * condition test and condition change and wakeup are under the same lock) then
150 * use __set_current_state().
152 * The above is typically ordered against the wakeup, which does:
154 * need_sleep = false;
155 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
157 * Where wake_up_state() (and all other wakeup primitives) imply enough
158 * barriers to order the store of the variable against wakeup.
160 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
161 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
162 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
164 * This is obviously fine, since they both store the exact same value.
166 * Also see the comments of try_to_wake_up().
168 #define __set_current_state(state_value) \
169 do { current->state = (state_value); } while (0)
170 #define set_current_state(state_value) \
171 smp_store_mb(current->state, (state_value))
175 /* Task command name length */
176 #define TASK_COMM_LEN 16
178 extern cpumask_var_t cpu_isolated_map;
180 extern int runqueue_is_locked(int cpu);
182 extern void scheduler_tick(void);
184 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
185 extern signed long schedule_timeout(signed long timeout);
186 extern signed long schedule_timeout_interruptible(signed long timeout);
187 extern signed long schedule_timeout_killable(signed long timeout);
188 extern signed long schedule_timeout_uninterruptible(signed long timeout);
189 extern signed long schedule_timeout_idle(signed long timeout);
190 asmlinkage void schedule(void);
191 extern void schedule_preempt_disabled(void);
193 extern int __must_check io_schedule_prepare(void);
194 extern void io_schedule_finish(int token);
195 extern long io_schedule_timeout(long timeout);
196 extern void io_schedule(void);
199 * struct prev_cputime - snaphsot of system and user cputime
200 * @utime: time spent in user mode
201 * @stime: time spent in system mode
202 * @lock: protects the above two fields
204 * Stores previous user/system time values such that we can guarantee
207 struct prev_cputime {
208 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
216 * struct task_cputime - collected CPU time counts
217 * @utime: time spent in user mode, in nanoseconds
218 * @stime: time spent in kernel mode, in nanoseconds
219 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
221 * This structure groups together three kinds of CPU time that are tracked for
222 * threads and thread groups. Most things considering CPU time want to group
223 * these counts together and treat all three of them in parallel.
225 struct task_cputime {
228 unsigned long long sum_exec_runtime;
231 /* Alternate field names when used to cache expirations. */
232 #define virt_exp utime
233 #define prof_exp stime
234 #define sched_exp sum_exec_runtime
236 #include <linux/rwsem.h>
238 #ifdef CONFIG_SCHED_INFO
240 /* cumulative counters */
241 unsigned long pcount; /* # of times run on this cpu */
242 unsigned long long run_delay; /* time spent waiting on a runqueue */
245 unsigned long long last_arrival,/* when we last ran on a cpu */
246 last_queued; /* when we were last queued to run */
248 #endif /* CONFIG_SCHED_INFO */
251 * Integer metrics need fixed point arithmetic, e.g., sched/fair
252 * has a few: load, load_avg, util_avg, freq, and capacity.
254 * We define a basic fixed point arithmetic range, and then formalize
255 * all these metrics based on that basic range.
257 # define SCHED_FIXEDPOINT_SHIFT 10
258 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
260 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
261 extern void prefetch_stack(struct task_struct *t);
263 static inline void prefetch_stack(struct task_struct *t) { }
267 unsigned long weight;
272 * The load_avg/util_avg accumulates an infinite geometric series
273 * (see __update_load_avg() in kernel/sched/fair.c).
275 * [load_avg definition]
277 * load_avg = runnable% * scale_load_down(load)
279 * where runnable% is the time ratio that a sched_entity is runnable.
280 * For cfs_rq, it is the aggregated load_avg of all runnable and
281 * blocked sched_entities.
283 * load_avg may also take frequency scaling into account:
285 * load_avg = runnable% * scale_load_down(load) * freq%
287 * where freq% is the CPU frequency normalized to the highest frequency.
289 * [util_avg definition]
291 * util_avg = running% * SCHED_CAPACITY_SCALE
293 * where running% is the time ratio that a sched_entity is running on
294 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
295 * and blocked sched_entities.
297 * util_avg may also factor frequency scaling and CPU capacity scaling:
299 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
301 * where freq% is the same as above, and capacity% is the CPU capacity
302 * normalized to the greatest capacity (due to uarch differences, etc).
304 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
305 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
306 * we therefore scale them to as large a range as necessary. This is for
307 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
311 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
312 * with the highest load (=88761), always runnable on a single cfs_rq,
313 * and should not overflow as the number already hits PID_MAX_LIMIT.
315 * For all other cases (including 32-bit kernels), struct load_weight's
316 * weight will overflow first before we do, because:
318 * Max(load_avg) <= Max(load.weight)
320 * Then it is the load_weight's responsibility to consider overflow
324 u64 last_update_time, load_sum;
325 u32 util_sum, period_contrib;
326 unsigned long load_avg, util_avg;
329 #ifdef CONFIG_SCHEDSTATS
330 struct sched_statistics {
340 s64 sum_sleep_runtime;
347 u64 nr_migrations_cold;
348 u64 nr_failed_migrations_affine;
349 u64 nr_failed_migrations_running;
350 u64 nr_failed_migrations_hot;
351 u64 nr_forced_migrations;
355 u64 nr_wakeups_migrate;
356 u64 nr_wakeups_local;
357 u64 nr_wakeups_remote;
358 u64 nr_wakeups_affine;
359 u64 nr_wakeups_affine_attempts;
360 u64 nr_wakeups_passive;
365 struct sched_entity {
366 struct load_weight load; /* for load-balancing */
367 struct rb_node run_node;
368 struct list_head group_node;
372 u64 sum_exec_runtime;
374 u64 prev_sum_exec_runtime;
378 #ifdef CONFIG_SCHEDSTATS
379 struct sched_statistics statistics;
382 #ifdef CONFIG_FAIR_GROUP_SCHED
384 struct sched_entity *parent;
385 /* rq on which this entity is (to be) queued: */
386 struct cfs_rq *cfs_rq;
387 /* rq "owned" by this entity/group: */
393 * Per entity load average tracking.
395 * Put into separate cache line so it does not
396 * collide with read-mostly values above.
398 struct sched_avg avg ____cacheline_aligned_in_smp;
402 struct sched_rt_entity {
403 struct list_head run_list;
404 unsigned long timeout;
405 unsigned long watchdog_stamp;
406 unsigned int time_slice;
407 unsigned short on_rq;
408 unsigned short on_list;
410 struct sched_rt_entity *back;
411 #ifdef CONFIG_RT_GROUP_SCHED
412 struct sched_rt_entity *parent;
413 /* rq on which this entity is (to be) queued: */
415 /* rq "owned" by this entity/group: */
420 struct sched_dl_entity {
421 struct rb_node rb_node;
424 * Original scheduling parameters. Copied here from sched_attr
425 * during sched_setattr(), they will remain the same until
426 * the next sched_setattr().
428 u64 dl_runtime; /* maximum runtime for each instance */
429 u64 dl_deadline; /* relative deadline of each instance */
430 u64 dl_period; /* separation of two instances (period) */
431 u64 dl_bw; /* dl_runtime / dl_deadline */
434 * Actual scheduling parameters. Initialized with the values above,
435 * they are continously updated during task execution. Note that
436 * the remaining runtime could be < 0 in case we are in overrun.
438 s64 runtime; /* remaining runtime for this instance */
439 u64 deadline; /* absolute deadline for this instance */
440 unsigned int flags; /* specifying the scheduler behaviour */
445 * @dl_throttled tells if we exhausted the runtime. If so, the
446 * task has to wait for a replenishment to be performed at the
447 * next firing of dl_timer.
449 * @dl_boosted tells if we are boosted due to DI. If so we are
450 * outside bandwidth enforcement mechanism (but only until we
451 * exit the critical section);
453 * @dl_yielded tells if task gave up the cpu before consuming
454 * all its available runtime during the last job.
456 int dl_throttled, dl_boosted, dl_yielded;
459 * Bandwidth enforcement timer. Each -deadline task has its
460 * own bandwidth to be enforced, thus we need one timer per task.
462 struct hrtimer dl_timer;
470 u8 pad; /* Otherwise the compiler can store garbage here. */
472 u32 s; /* Set of bits. */
475 enum perf_event_task_context {
476 perf_invalid_context = -1,
479 perf_nr_task_contexts,
483 struct wake_q_node *next;
487 #ifdef CONFIG_THREAD_INFO_IN_TASK
489 * For reasons of header soup (see current_thread_info()), this
490 * must be the first element of task_struct.
492 struct thread_info thread_info;
494 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
497 unsigned int flags; /* per process flags, defined below */
501 struct llist_node wake_entry;
503 #ifdef CONFIG_THREAD_INFO_IN_TASK
504 unsigned int cpu; /* current CPU */
506 unsigned int wakee_flips;
507 unsigned long wakee_flip_decay_ts;
508 struct task_struct *last_wakee;
514 int prio, static_prio, normal_prio;
515 unsigned int rt_priority;
516 const struct sched_class *sched_class;
517 struct sched_entity se;
518 struct sched_rt_entity rt;
519 #ifdef CONFIG_CGROUP_SCHED
520 struct task_group *sched_task_group;
522 struct sched_dl_entity dl;
524 #ifdef CONFIG_PREEMPT_NOTIFIERS
525 /* list of struct preempt_notifier: */
526 struct hlist_head preempt_notifiers;
529 #ifdef CONFIG_BLK_DEV_IO_TRACE
530 unsigned int btrace_seq;
535 cpumask_t cpus_allowed;
537 #ifdef CONFIG_PREEMPT_RCU
538 int rcu_read_lock_nesting;
539 union rcu_special rcu_read_unlock_special;
540 struct list_head rcu_node_entry;
541 struct rcu_node *rcu_blocked_node;
542 #endif /* #ifdef CONFIG_PREEMPT_RCU */
543 #ifdef CONFIG_TASKS_RCU
544 unsigned long rcu_tasks_nvcsw;
545 bool rcu_tasks_holdout;
546 struct list_head rcu_tasks_holdout_list;
547 int rcu_tasks_idle_cpu;
548 #endif /* #ifdef CONFIG_TASKS_RCU */
550 #ifdef CONFIG_SCHED_INFO
551 struct sched_info sched_info;
554 struct list_head tasks;
556 struct plist_node pushable_tasks;
557 struct rb_node pushable_dl_tasks;
560 struct mm_struct *mm, *active_mm;
562 /* Per-thread vma caching: */
563 struct vmacache vmacache;
565 #if defined(SPLIT_RSS_COUNTING)
566 struct task_rss_stat rss_stat;
570 int exit_code, exit_signal;
571 int pdeath_signal; /* The signal sent when the parent dies */
572 unsigned long jobctl; /* JOBCTL_*, siglock protected */
574 /* Used for emulating ABI behavior of previous Linux versions */
575 unsigned int personality;
577 /* scheduler bits, serialized by scheduler locks */
578 unsigned sched_reset_on_fork:1;
579 unsigned sched_contributes_to_load:1;
580 unsigned sched_migrated:1;
581 unsigned sched_remote_wakeup:1;
582 unsigned :0; /* force alignment to the next boundary */
584 /* unserialized, strictly 'current' */
585 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
586 unsigned in_iowait:1;
587 #if !defined(TIF_RESTORE_SIGMASK)
588 unsigned restore_sigmask:1;
591 unsigned memcg_may_oom:1;
593 unsigned memcg_kmem_skip_account:1;
596 #ifdef CONFIG_COMPAT_BRK
597 unsigned brk_randomized:1;
600 unsigned long atomic_flags; /* Flags needing atomic access. */
602 struct restart_block restart_block;
607 #ifdef CONFIG_CC_STACKPROTECTOR
608 /* Canary value for the -fstack-protector gcc feature */
609 unsigned long stack_canary;
612 * pointers to (original) parent process, youngest child, younger sibling,
613 * older sibling, respectively. (p->father can be replaced with
614 * p->real_parent->pid)
616 struct task_struct __rcu *real_parent; /* real parent process */
617 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
619 * children/sibling forms the list of my natural children
621 struct list_head children; /* list of my children */
622 struct list_head sibling; /* linkage in my parent's children list */
623 struct task_struct *group_leader; /* threadgroup leader */
626 * ptraced is the list of tasks this task is using ptrace on.
627 * This includes both natural children and PTRACE_ATTACH targets.
628 * p->ptrace_entry is p's link on the p->parent->ptraced list.
630 struct list_head ptraced;
631 struct list_head ptrace_entry;
633 /* PID/PID hash table linkage. */
634 struct pid_link pids[PIDTYPE_MAX];
635 struct list_head thread_group;
636 struct list_head thread_node;
638 struct completion *vfork_done; /* for vfork() */
639 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
640 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
643 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
644 u64 utimescaled, stimescaled;
647 struct prev_cputime prev_cputime;
648 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
649 seqcount_t vtime_seqcount;
650 unsigned long long vtime_snap;
652 /* Task is sleeping or running in a CPU with VTIME inactive */
654 /* Task runs in userspace in a CPU with VTIME active */
656 /* Task runs in kernelspace in a CPU with VTIME active */
661 #ifdef CONFIG_NO_HZ_FULL
662 atomic_t tick_dep_mask;
664 unsigned long nvcsw, nivcsw; /* context switch counts */
665 u64 start_time; /* monotonic time in nsec */
666 u64 real_start_time; /* boot based time in nsec */
667 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
668 unsigned long min_flt, maj_flt;
670 #ifdef CONFIG_POSIX_TIMERS
671 struct task_cputime cputime_expires;
672 struct list_head cpu_timers[3];
675 /* process credentials */
676 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
677 const struct cred __rcu *real_cred; /* objective and real subjective task
678 * credentials (COW) */
679 const struct cred __rcu *cred; /* effective (overridable) subjective task
680 * credentials (COW) */
681 char comm[TASK_COMM_LEN]; /* executable name excluding path
682 - access with [gs]et_task_comm (which lock
684 - initialized normally by setup_new_exec */
685 /* file system info */
686 struct nameidata *nameidata;
687 #ifdef CONFIG_SYSVIPC
689 struct sysv_sem sysvsem;
690 struct sysv_shm sysvshm;
692 #ifdef CONFIG_DETECT_HUNG_TASK
693 /* hung task detection */
694 unsigned long last_switch_count;
696 /* filesystem information */
697 struct fs_struct *fs;
698 /* open file information */
699 struct files_struct *files;
701 struct nsproxy *nsproxy;
702 /* signal handlers */
703 struct signal_struct *signal;
704 struct sighand_struct *sighand;
706 sigset_t blocked, real_blocked;
707 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
708 struct sigpending pending;
710 unsigned long sas_ss_sp;
712 unsigned sas_ss_flags;
714 struct callback_head *task_works;
716 struct audit_context *audit_context;
717 #ifdef CONFIG_AUDITSYSCALL
719 unsigned int sessionid;
721 struct seccomp seccomp;
723 /* Thread group tracking */
726 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
728 spinlock_t alloc_lock;
730 /* Protection of the PI data structures: */
731 raw_spinlock_t pi_lock;
733 struct wake_q_node wake_q;
735 #ifdef CONFIG_RT_MUTEXES
736 /* PI waiters blocked on a rt_mutex held by this task */
737 struct rb_root pi_waiters;
738 struct rb_node *pi_waiters_leftmost;
739 /* Deadlock detection and priority inheritance handling */
740 struct rt_mutex_waiter *pi_blocked_on;
743 #ifdef CONFIG_DEBUG_MUTEXES
744 /* mutex deadlock detection */
745 struct mutex_waiter *blocked_on;
747 #ifdef CONFIG_TRACE_IRQFLAGS
748 unsigned int irq_events;
749 unsigned long hardirq_enable_ip;
750 unsigned long hardirq_disable_ip;
751 unsigned int hardirq_enable_event;
752 unsigned int hardirq_disable_event;
753 int hardirqs_enabled;
755 unsigned long softirq_disable_ip;
756 unsigned long softirq_enable_ip;
757 unsigned int softirq_disable_event;
758 unsigned int softirq_enable_event;
759 int softirqs_enabled;
762 #ifdef CONFIG_LOCKDEP
763 # define MAX_LOCK_DEPTH 48UL
766 unsigned int lockdep_recursion;
767 struct held_lock held_locks[MAX_LOCK_DEPTH];
768 gfp_t lockdep_reclaim_gfp;
771 unsigned int in_ubsan;
774 /* journalling filesystem info */
777 /* stacked block device info */
778 struct bio_list *bio_list;
782 struct blk_plug *plug;
786 struct reclaim_state *reclaim_state;
788 struct backing_dev_info *backing_dev_info;
790 struct io_context *io_context;
792 unsigned long ptrace_message;
793 siginfo_t *last_siginfo; /* For ptrace use. */
794 struct task_io_accounting ioac;
795 #if defined(CONFIG_TASK_XACCT)
796 u64 acct_rss_mem1; /* accumulated rss usage */
797 u64 acct_vm_mem1; /* accumulated virtual memory usage */
798 u64 acct_timexpd; /* stime + utime since last update */
800 #ifdef CONFIG_CPUSETS
801 nodemask_t mems_allowed; /* Protected by alloc_lock */
802 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
803 int cpuset_mem_spread_rotor;
804 int cpuset_slab_spread_rotor;
806 #ifdef CONFIG_CGROUPS
807 /* Control Group info protected by css_set_lock */
808 struct css_set __rcu *cgroups;
809 /* cg_list protected by css_set_lock and tsk->alloc_lock */
810 struct list_head cg_list;
812 #ifdef CONFIG_INTEL_RDT_A
816 struct robust_list_head __user *robust_list;
818 struct compat_robust_list_head __user *compat_robust_list;
820 struct list_head pi_state_list;
821 struct futex_pi_state *pi_state_cache;
823 #ifdef CONFIG_PERF_EVENTS
824 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
825 struct mutex perf_event_mutex;
826 struct list_head perf_event_list;
828 #ifdef CONFIG_DEBUG_PREEMPT
829 unsigned long preempt_disable_ip;
832 struct mempolicy *mempolicy; /* Protected by alloc_lock */
834 short pref_node_fork;
836 #ifdef CONFIG_NUMA_BALANCING
838 unsigned int numa_scan_period;
839 unsigned int numa_scan_period_max;
840 int numa_preferred_nid;
841 unsigned long numa_migrate_retry;
842 u64 node_stamp; /* migration stamp */
843 u64 last_task_numa_placement;
844 u64 last_sum_exec_runtime;
845 struct callback_head numa_work;
847 struct list_head numa_entry;
848 struct numa_group *numa_group;
851 * numa_faults is an array split into four regions:
852 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
853 * in this precise order.
855 * faults_memory: Exponential decaying average of faults on a per-node
856 * basis. Scheduling placement decisions are made based on these
857 * counts. The values remain static for the duration of a PTE scan.
858 * faults_cpu: Track the nodes the process was running on when a NUMA
859 * hinting fault was incurred.
860 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
861 * during the current scan window. When the scan completes, the counts
862 * in faults_memory and faults_cpu decay and these values are copied.
864 unsigned long *numa_faults;
865 unsigned long total_numa_faults;
868 * numa_faults_locality tracks if faults recorded during the last
869 * scan window were remote/local or failed to migrate. The task scan
870 * period is adapted based on the locality of the faults with different
871 * weights depending on whether they were shared or private faults
873 unsigned long numa_faults_locality[3];
875 unsigned long numa_pages_migrated;
876 #endif /* CONFIG_NUMA_BALANCING */
878 struct tlbflush_unmap_batch tlb_ubc;
883 * cache last used pipe for splice
885 struct pipe_inode_info *splice_pipe;
887 struct page_frag task_frag;
889 #ifdef CONFIG_TASK_DELAY_ACCT
890 struct task_delay_info *delays;
893 #ifdef CONFIG_FAULT_INJECTION
897 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
898 * balance_dirty_pages() for some dirty throttling pause
901 int nr_dirtied_pause;
902 unsigned long dirty_paused_when; /* start of a write-and-pause period */
904 #ifdef CONFIG_LATENCYTOP
905 int latency_record_count;
906 struct latency_record latency_record[LT_SAVECOUNT];
909 * time slack values; these are used to round up poll() and
910 * select() etc timeout values. These are in nanoseconds.
913 u64 default_timer_slack_ns;
916 unsigned int kasan_depth;
918 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
919 /* Index of current stored address in ret_stack */
921 /* Stack of return addresses for return function tracing */
922 struct ftrace_ret_stack *ret_stack;
923 /* time stamp for last schedule */
924 unsigned long long ftrace_timestamp;
926 * Number of functions that haven't been traced
927 * because of depth overrun.
929 atomic_t trace_overrun;
930 /* Pause for the tracing */
931 atomic_t tracing_graph_pause;
933 #ifdef CONFIG_TRACING
934 /* state flags for use by tracers */
936 /* bitmask and counter of trace recursion */
937 unsigned long trace_recursion;
938 #endif /* CONFIG_TRACING */
940 /* Coverage collection mode enabled for this task (0 if disabled). */
941 enum kcov_mode kcov_mode;
942 /* Size of the kcov_area. */
944 /* Buffer for coverage collection. */
946 /* kcov desciptor wired with this task or NULL. */
950 struct mem_cgroup *memcg_in_oom;
951 gfp_t memcg_oom_gfp_mask;
954 /* number of pages to reclaim on returning to userland */
955 unsigned int memcg_nr_pages_over_high;
957 #ifdef CONFIG_UPROBES
958 struct uprobe_task *utask;
960 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
961 unsigned int sequential_io;
962 unsigned int sequential_io_avg;
964 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
965 unsigned long task_state_change;
967 int pagefault_disabled;
969 struct task_struct *oom_reaper_list;
971 #ifdef CONFIG_VMAP_STACK
972 struct vm_struct *stack_vm_area;
974 #ifdef CONFIG_THREAD_INFO_IN_TASK
975 /* A live task holds one reference. */
976 atomic_t stack_refcount;
978 /* CPU-specific state of this task */
979 struct thread_struct thread;
981 * WARNING: on x86, 'thread_struct' contains a variable-sized
982 * structure. It *MUST* be at the end of 'task_struct'.
984 * Do not put anything below here!
988 static inline struct pid *task_pid(struct task_struct *task)
990 return task->pids[PIDTYPE_PID].pid;
993 static inline struct pid *task_tgid(struct task_struct *task)
995 return task->group_leader->pids[PIDTYPE_PID].pid;
999 * Without tasklist or rcu lock it is not safe to dereference
1000 * the result of task_pgrp/task_session even if task == current,
1001 * we can race with another thread doing sys_setsid/sys_setpgid.
1003 static inline struct pid *task_pgrp(struct task_struct *task)
1005 return task->group_leader->pids[PIDTYPE_PGID].pid;
1008 static inline struct pid *task_session(struct task_struct *task)
1010 return task->group_leader->pids[PIDTYPE_SID].pid;
1014 * the helpers to get the task's different pids as they are seen
1015 * from various namespaces
1017 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1018 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1020 * task_xid_nr_ns() : id seen from the ns specified;
1022 * set_task_vxid() : assigns a virtual id to a task;
1024 * see also pid_nr() etc in include/linux/pid.h
1026 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1027 struct pid_namespace *ns);
1029 static inline pid_t task_pid_nr(struct task_struct *tsk)
1034 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1035 struct pid_namespace *ns)
1037 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1040 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1042 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1046 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1051 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1053 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1055 return pid_vnr(task_tgid(tsk));
1059 static inline int pid_alive(const struct task_struct *p);
1060 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1066 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1072 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1074 return task_ppid_nr_ns(tsk, &init_pid_ns);
1077 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1078 struct pid_namespace *ns)
1080 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1083 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1085 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1089 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1090 struct pid_namespace *ns)
1092 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1095 static inline pid_t task_session_vnr(struct task_struct *tsk)
1097 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1100 /* obsolete, do not use */
1101 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1103 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1107 * pid_alive - check that a task structure is not stale
1108 * @p: Task structure to be checked.
1110 * Test if a process is not yet dead (at most zombie state)
1111 * If pid_alive fails, then pointers within the task structure
1112 * can be stale and must not be dereferenced.
1114 * Return: 1 if the process is alive. 0 otherwise.
1116 static inline int pid_alive(const struct task_struct *p)
1118 return p->pids[PIDTYPE_PID].pid != NULL;
1122 * is_global_init - check if a task structure is init. Since init
1123 * is free to have sub-threads we need to check tgid.
1124 * @tsk: Task structure to be checked.
1126 * Check if a task structure is the first user space task the kernel created.
1128 * Return: 1 if the task structure is init. 0 otherwise.
1130 static inline int is_global_init(struct task_struct *tsk)
1132 return task_tgid_nr(tsk) == 1;
1135 extern struct pid *cad_pid;
1140 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1141 #define PF_EXITING 0x00000004 /* getting shut down */
1142 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1143 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1144 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1145 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1146 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1147 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1148 #define PF_DUMPCORE 0x00000200 /* dumped core */
1149 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1150 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1151 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1152 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1153 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1154 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1155 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1156 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1157 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1158 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1159 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1160 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1161 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1162 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1163 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1164 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1165 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1166 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1167 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1170 * Only the _current_ task can read/write to tsk->flags, but other
1171 * tasks can access tsk->flags in readonly mode for example
1172 * with tsk_used_math (like during threaded core dumping).
1173 * There is however an exception to this rule during ptrace
1174 * or during fork: the ptracer task is allowed to write to the
1175 * child->flags of its traced child (same goes for fork, the parent
1176 * can write to the child->flags), because we're guaranteed the
1177 * child is not running and in turn not changing child->flags
1178 * at the same time the parent does it.
1180 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1181 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1182 #define clear_used_math() clear_stopped_child_used_math(current)
1183 #define set_used_math() set_stopped_child_used_math(current)
1184 #define conditional_stopped_child_used_math(condition, child) \
1185 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1186 #define conditional_used_math(condition) \
1187 conditional_stopped_child_used_math(condition, current)
1188 #define copy_to_stopped_child_used_math(child) \
1189 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1190 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1191 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1192 #define used_math() tsk_used_math(current)
1194 /* Per-process atomic flags. */
1195 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1196 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1197 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1198 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1201 #define TASK_PFA_TEST(name, func) \
1202 static inline bool task_##func(struct task_struct *p) \
1203 { return test_bit(PFA_##name, &p->atomic_flags); }
1204 #define TASK_PFA_SET(name, func) \
1205 static inline void task_set_##func(struct task_struct *p) \
1206 { set_bit(PFA_##name, &p->atomic_flags); }
1207 #define TASK_PFA_CLEAR(name, func) \
1208 static inline void task_clear_##func(struct task_struct *p) \
1209 { clear_bit(PFA_##name, &p->atomic_flags); }
1211 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1212 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1214 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1215 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1216 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1218 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1219 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1220 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1222 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1223 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1225 static inline void tsk_restore_flags(struct task_struct *task,
1226 unsigned long orig_flags, unsigned long flags)
1228 task->flags &= ~flags;
1229 task->flags |= orig_flags & flags;
1232 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1233 const struct cpumask *trial);
1234 extern int task_can_attach(struct task_struct *p,
1235 const struct cpumask *cs_cpus_allowed);
1237 extern void do_set_cpus_allowed(struct task_struct *p,
1238 const struct cpumask *new_mask);
1240 extern int set_cpus_allowed_ptr(struct task_struct *p,
1241 const struct cpumask *new_mask);
1243 static inline void do_set_cpus_allowed(struct task_struct *p,
1244 const struct cpumask *new_mask)
1247 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1248 const struct cpumask *new_mask)
1250 if (!cpumask_test_cpu(0, new_mask))
1256 #ifndef cpu_relax_yield
1257 #define cpu_relax_yield() cpu_relax()
1260 extern int yield_to(struct task_struct *p, bool preempt);
1261 extern void set_user_nice(struct task_struct *p, long nice);
1262 extern int task_prio(const struct task_struct *p);
1264 * task_nice - return the nice value of a given task.
1265 * @p: the task in question.
1267 * Return: The nice value [ -20 ... 0 ... 19 ].
1269 static inline int task_nice(const struct task_struct *p)
1271 return PRIO_TO_NICE((p)->static_prio);
1273 extern int can_nice(const struct task_struct *p, const int nice);
1274 extern int task_curr(const struct task_struct *p);
1275 extern int idle_cpu(int cpu);
1276 extern int sched_setscheduler(struct task_struct *, int,
1277 const struct sched_param *);
1278 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1279 const struct sched_param *);
1280 extern int sched_setattr(struct task_struct *,
1281 const struct sched_attr *);
1282 extern struct task_struct *idle_task(int cpu);
1284 * is_idle_task - is the specified task an idle task?
1285 * @p: the task in question.
1287 * Return: 1 if @p is an idle task. 0 otherwise.
1289 static inline bool is_idle_task(const struct task_struct *p)
1291 return !!(p->flags & PF_IDLE);
1293 extern struct task_struct *curr_task(int cpu);
1294 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1298 union thread_union {
1299 #ifndef CONFIG_THREAD_INFO_IN_TASK
1300 struct thread_info thread_info;
1302 unsigned long stack[THREAD_SIZE/sizeof(long)];
1305 #ifdef CONFIG_THREAD_INFO_IN_TASK
1306 static inline struct thread_info *task_thread_info(struct task_struct *task)
1308 return &task->thread_info;
1310 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1311 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1314 extern struct pid_namespace init_pid_ns;
1317 * find a task by one of its numerical ids
1319 * find_task_by_pid_ns():
1320 * finds a task by its pid in the specified namespace
1321 * find_task_by_vpid():
1322 * finds a task by its virtual pid
1324 * see also find_vpid() etc in include/linux/pid.h
1327 extern struct task_struct *find_task_by_vpid(pid_t nr);
1328 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1329 struct pid_namespace *ns);
1331 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1332 extern int wake_up_process(struct task_struct *tsk);
1333 extern void wake_up_new_task(struct task_struct *tsk);
1335 extern void kick_process(struct task_struct *tsk);
1337 static inline void kick_process(struct task_struct *tsk) { }
1340 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1341 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1343 __set_task_comm(tsk, from, false);
1345 extern char *get_task_comm(char *to, struct task_struct *tsk);
1348 void scheduler_ipi(void);
1349 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1351 static inline void scheduler_ipi(void) { }
1352 static inline unsigned long wait_task_inactive(struct task_struct *p,
1359 /* set thread flags in other task's structures
1360 * - see asm/thread_info.h for TIF_xxxx flags available
1362 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1364 set_ti_thread_flag(task_thread_info(tsk), flag);
1367 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1369 clear_ti_thread_flag(task_thread_info(tsk), flag);
1372 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1374 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1377 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1379 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1382 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1384 return test_ti_thread_flag(task_thread_info(tsk), flag);
1387 static inline void set_tsk_need_resched(struct task_struct *tsk)
1389 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1392 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1394 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1397 static inline int test_tsk_need_resched(struct task_struct *tsk)
1399 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1403 * cond_resched() and cond_resched_lock(): latency reduction via
1404 * explicit rescheduling in places that are safe. The return
1405 * value indicates whether a reschedule was done in fact.
1406 * cond_resched_lock() will drop the spinlock before scheduling,
1407 * cond_resched_softirq() will enable bhs before scheduling.
1409 #ifndef CONFIG_PREEMPT
1410 extern int _cond_resched(void);
1412 static inline int _cond_resched(void) { return 0; }
1415 #define cond_resched() ({ \
1416 ___might_sleep(__FILE__, __LINE__, 0); \
1420 extern int __cond_resched_lock(spinlock_t *lock);
1422 #define cond_resched_lock(lock) ({ \
1423 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1424 __cond_resched_lock(lock); \
1427 extern int __cond_resched_softirq(void);
1429 #define cond_resched_softirq() ({ \
1430 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1431 __cond_resched_softirq(); \
1434 static inline void cond_resched_rcu(void)
1436 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1444 * Does a critical section need to be broken due to another
1445 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1446 * but a general need for low latency)
1448 static inline int spin_needbreak(spinlock_t *lock)
1450 #ifdef CONFIG_PREEMPT
1451 return spin_is_contended(lock);
1457 static __always_inline bool need_resched(void)
1459 return unlikely(tif_need_resched());
1463 * Wrappers for p->thread_info->cpu access. No-op on UP.
1467 static inline unsigned int task_cpu(const struct task_struct *p)
1469 #ifdef CONFIG_THREAD_INFO_IN_TASK
1472 return task_thread_info(p)->cpu;
1476 static inline int task_node(const struct task_struct *p)
1478 return cpu_to_node(task_cpu(p));
1481 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1485 static inline unsigned int task_cpu(const struct task_struct *p)
1490 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1494 #endif /* CONFIG_SMP */
1497 * In order to reduce various lock holder preemption latencies provide an
1498 * interface to see if a vCPU is currently running or not.
1500 * This allows us to terminate optimistic spin loops and block, analogous to
1501 * the native optimistic spin heuristic of testing if the lock owner task is
1504 #ifndef vcpu_is_preempted
1505 # define vcpu_is_preempted(cpu) false
1508 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1509 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1511 #ifndef TASK_SIZE_OF
1512 #define TASK_SIZE_OF(tsk) TASK_SIZE