4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
8 #include <linux/capability.h>
9 #include <linux/mutex.h>
10 #include <linux/plist.h>
11 #include <linux/mm_types_task.h>
12 #include <asm/ptrace.h>
14 #include <linux/sem.h>
15 #include <linux/shm.h>
16 #include <linux/signal.h>
17 #include <linux/signal_types.h>
18 #include <linux/pid.h>
19 #include <linux/seccomp.h>
20 #include <linux/rculist.h>
21 #include <linux/rtmutex.h>
23 #include <linux/resource.h>
24 #include <linux/hrtimer.h>
25 #include <linux/kcov.h>
26 #include <linux/task_io_accounting.h>
27 #include <linux/latencytop.h>
28 #include <linux/cred.h>
29 #include <linux/gfp.h>
30 #include <linux/topology.h>
31 #include <linux/magic.h>
32 #include <linux/cgroup-defs.h>
34 #include <asm/current.h>
36 /* task_struct member predeclarations: */
39 struct backing_dev_info;
45 struct futex_pi_state;
50 struct perf_event_context;
52 struct pipe_inode_info;
55 struct robust_list_head;
59 struct sighand_struct;
61 struct task_delay_info;
67 * Task state bitmask. NOTE! These bits are also
68 * encoded in fs/proc/array.c: get_task_state().
70 * We have two separate sets of flags: task->state
71 * is about runnability, while task->exit_state are
72 * about the task exiting. Confusing, but this way
73 * modifying one set can't modify the other one by
76 #define TASK_RUNNING 0
77 #define TASK_INTERRUPTIBLE 1
78 #define TASK_UNINTERRUPTIBLE 2
79 #define __TASK_STOPPED 4
80 #define __TASK_TRACED 8
81 /* in tsk->exit_state */
83 #define EXIT_ZOMBIE 32
84 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
85 /* in tsk->state again */
87 #define TASK_WAKEKILL 128
88 #define TASK_WAKING 256
89 #define TASK_PARKED 512
90 #define TASK_NOLOAD 1024
92 #define TASK_STATE_MAX 4096
94 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
96 /* Convenience macros for the sake of set_current_state */
97 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
98 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
99 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
101 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
103 /* Convenience macros for the sake of wake_up */
104 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
105 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
107 /* get_task_state() */
108 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
109 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
110 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
112 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
113 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
114 #define task_is_stopped_or_traced(task) \
115 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
116 #define task_contributes_to_load(task) \
117 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
118 (task->flags & PF_FROZEN) == 0 && \
119 (task->state & TASK_NOLOAD) == 0)
121 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
123 #define __set_current_state(state_value) \
125 current->task_state_change = _THIS_IP_; \
126 current->state = (state_value); \
128 #define set_current_state(state_value) \
130 current->task_state_change = _THIS_IP_; \
131 smp_store_mb(current->state, (state_value)); \
136 * set_current_state() includes a barrier so that the write of current->state
137 * is correctly serialised wrt the caller's subsequent test of whether to
141 * set_current_state(TASK_UNINTERRUPTIBLE);
147 * __set_current_state(TASK_RUNNING);
149 * If the caller does not need such serialisation (because, for instance, the
150 * condition test and condition change and wakeup are under the same lock) then
151 * use __set_current_state().
153 * The above is typically ordered against the wakeup, which does:
155 * need_sleep = false;
156 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
158 * Where wake_up_state() (and all other wakeup primitives) imply enough
159 * barriers to order the store of the variable against wakeup.
161 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
162 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
163 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
165 * This is obviously fine, since they both store the exact same value.
167 * Also see the comments of try_to_wake_up().
169 #define __set_current_state(state_value) \
170 do { current->state = (state_value); } while (0)
171 #define set_current_state(state_value) \
172 smp_store_mb(current->state, (state_value))
176 /* Task command name length */
177 #define TASK_COMM_LEN 16
179 extern void sched_init(void);
180 extern void sched_init_smp(void);
182 extern cpumask_var_t cpu_isolated_map;
184 extern int runqueue_is_locked(int cpu);
186 extern void cpu_init (void);
187 extern void trap_init(void);
188 extern void update_process_times(int user);
189 extern void scheduler_tick(void);
191 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
192 extern signed long schedule_timeout(signed long timeout);
193 extern signed long schedule_timeout_interruptible(signed long timeout);
194 extern signed long schedule_timeout_killable(signed long timeout);
195 extern signed long schedule_timeout_uninterruptible(signed long timeout);
196 extern signed long schedule_timeout_idle(signed long timeout);
197 asmlinkage void schedule(void);
198 extern void schedule_preempt_disabled(void);
200 extern int __must_check io_schedule_prepare(void);
201 extern void io_schedule_finish(int token);
202 extern long io_schedule_timeout(long timeout);
203 extern void io_schedule(void);
206 * struct prev_cputime - snaphsot of system and user cputime
207 * @utime: time spent in user mode
208 * @stime: time spent in system mode
209 * @lock: protects the above two fields
211 * Stores previous user/system time values such that we can guarantee
214 struct prev_cputime {
215 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
223 * struct task_cputime - collected CPU time counts
224 * @utime: time spent in user mode, in nanoseconds
225 * @stime: time spent in kernel mode, in nanoseconds
226 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
228 * This structure groups together three kinds of CPU time that are tracked for
229 * threads and thread groups. Most things considering CPU time want to group
230 * these counts together and treat all three of them in parallel.
232 struct task_cputime {
235 unsigned long long sum_exec_runtime;
238 /* Alternate field names when used to cache expirations. */
239 #define virt_exp utime
240 #define prof_exp stime
241 #define sched_exp sum_exec_runtime
243 #include <linux/rwsem.h>
245 #ifdef CONFIG_SCHED_INFO
247 /* cumulative counters */
248 unsigned long pcount; /* # of times run on this cpu */
249 unsigned long long run_delay; /* time spent waiting on a runqueue */
252 unsigned long long last_arrival,/* when we last ran on a cpu */
253 last_queued; /* when we were last queued to run */
255 #endif /* CONFIG_SCHED_INFO */
258 * Integer metrics need fixed point arithmetic, e.g., sched/fair
259 * has a few: load, load_avg, util_avg, freq, and capacity.
261 * We define a basic fixed point arithmetic range, and then formalize
262 * all these metrics based on that basic range.
264 # define SCHED_FIXEDPOINT_SHIFT 10
265 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
267 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
268 extern void prefetch_stack(struct task_struct *t);
270 static inline void prefetch_stack(struct task_struct *t) { }
274 unsigned long weight;
279 * The load_avg/util_avg accumulates an infinite geometric series
280 * (see __update_load_avg() in kernel/sched/fair.c).
282 * [load_avg definition]
284 * load_avg = runnable% * scale_load_down(load)
286 * where runnable% is the time ratio that a sched_entity is runnable.
287 * For cfs_rq, it is the aggregated load_avg of all runnable and
288 * blocked sched_entities.
290 * load_avg may also take frequency scaling into account:
292 * load_avg = runnable% * scale_load_down(load) * freq%
294 * where freq% is the CPU frequency normalized to the highest frequency.
296 * [util_avg definition]
298 * util_avg = running% * SCHED_CAPACITY_SCALE
300 * where running% is the time ratio that a sched_entity is running on
301 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
302 * and blocked sched_entities.
304 * util_avg may also factor frequency scaling and CPU capacity scaling:
306 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
308 * where freq% is the same as above, and capacity% is the CPU capacity
309 * normalized to the greatest capacity (due to uarch differences, etc).
311 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
312 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
313 * we therefore scale them to as large a range as necessary. This is for
314 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
318 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
319 * with the highest load (=88761), always runnable on a single cfs_rq,
320 * and should not overflow as the number already hits PID_MAX_LIMIT.
322 * For all other cases (including 32-bit kernels), struct load_weight's
323 * weight will overflow first before we do, because:
325 * Max(load_avg) <= Max(load.weight)
327 * Then it is the load_weight's responsibility to consider overflow
331 u64 last_update_time, load_sum;
332 u32 util_sum, period_contrib;
333 unsigned long load_avg, util_avg;
336 #ifdef CONFIG_SCHEDSTATS
337 struct sched_statistics {
347 s64 sum_sleep_runtime;
354 u64 nr_migrations_cold;
355 u64 nr_failed_migrations_affine;
356 u64 nr_failed_migrations_running;
357 u64 nr_failed_migrations_hot;
358 u64 nr_forced_migrations;
362 u64 nr_wakeups_migrate;
363 u64 nr_wakeups_local;
364 u64 nr_wakeups_remote;
365 u64 nr_wakeups_affine;
366 u64 nr_wakeups_affine_attempts;
367 u64 nr_wakeups_passive;
372 struct sched_entity {
373 struct load_weight load; /* for load-balancing */
374 struct rb_node run_node;
375 struct list_head group_node;
379 u64 sum_exec_runtime;
381 u64 prev_sum_exec_runtime;
385 #ifdef CONFIG_SCHEDSTATS
386 struct sched_statistics statistics;
389 #ifdef CONFIG_FAIR_GROUP_SCHED
391 struct sched_entity *parent;
392 /* rq on which this entity is (to be) queued: */
393 struct cfs_rq *cfs_rq;
394 /* rq "owned" by this entity/group: */
400 * Per entity load average tracking.
402 * Put into separate cache line so it does not
403 * collide with read-mostly values above.
405 struct sched_avg avg ____cacheline_aligned_in_smp;
409 struct sched_rt_entity {
410 struct list_head run_list;
411 unsigned long timeout;
412 unsigned long watchdog_stamp;
413 unsigned int time_slice;
414 unsigned short on_rq;
415 unsigned short on_list;
417 struct sched_rt_entity *back;
418 #ifdef CONFIG_RT_GROUP_SCHED
419 struct sched_rt_entity *parent;
420 /* rq on which this entity is (to be) queued: */
422 /* rq "owned" by this entity/group: */
427 struct sched_dl_entity {
428 struct rb_node rb_node;
431 * Original scheduling parameters. Copied here from sched_attr
432 * during sched_setattr(), they will remain the same until
433 * the next sched_setattr().
435 u64 dl_runtime; /* maximum runtime for each instance */
436 u64 dl_deadline; /* relative deadline of each instance */
437 u64 dl_period; /* separation of two instances (period) */
438 u64 dl_bw; /* dl_runtime / dl_deadline */
441 * Actual scheduling parameters. Initialized with the values above,
442 * they are continously updated during task execution. Note that
443 * the remaining runtime could be < 0 in case we are in overrun.
445 s64 runtime; /* remaining runtime for this instance */
446 u64 deadline; /* absolute deadline for this instance */
447 unsigned int flags; /* specifying the scheduler behaviour */
452 * @dl_throttled tells if we exhausted the runtime. If so, the
453 * task has to wait for a replenishment to be performed at the
454 * next firing of dl_timer.
456 * @dl_boosted tells if we are boosted due to DI. If so we are
457 * outside bandwidth enforcement mechanism (but only until we
458 * exit the critical section);
460 * @dl_yielded tells if task gave up the cpu before consuming
461 * all its available runtime during the last job.
463 int dl_throttled, dl_boosted, dl_yielded;
466 * Bandwidth enforcement timer. Each -deadline task has its
467 * own bandwidth to be enforced, thus we need one timer per task.
469 struct hrtimer dl_timer;
477 u8 pad; /* Otherwise the compiler can store garbage here. */
479 u32 s; /* Set of bits. */
482 enum perf_event_task_context {
483 perf_invalid_context = -1,
486 perf_nr_task_contexts,
490 struct wake_q_node *next;
493 /* Track pages that require TLB flushes */
494 struct tlbflush_unmap_batch {
496 * Each bit set is a CPU that potentially has a TLB entry for one of
497 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
499 struct cpumask cpumask;
501 /* True if any bit in cpumask is set */
505 * If true then the PTE was dirty when unmapped. The entry must be
506 * flushed before IO is initiated or a stale TLB entry potentially
507 * allows an update without redirtying the page.
513 #ifdef CONFIG_THREAD_INFO_IN_TASK
515 * For reasons of header soup (see current_thread_info()), this
516 * must be the first element of task_struct.
518 struct thread_info thread_info;
520 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
523 unsigned int flags; /* per process flags, defined below */
527 struct llist_node wake_entry;
529 #ifdef CONFIG_THREAD_INFO_IN_TASK
530 unsigned int cpu; /* current CPU */
532 unsigned int wakee_flips;
533 unsigned long wakee_flip_decay_ts;
534 struct task_struct *last_wakee;
540 int prio, static_prio, normal_prio;
541 unsigned int rt_priority;
542 const struct sched_class *sched_class;
543 struct sched_entity se;
544 struct sched_rt_entity rt;
545 #ifdef CONFIG_CGROUP_SCHED
546 struct task_group *sched_task_group;
548 struct sched_dl_entity dl;
550 #ifdef CONFIG_PREEMPT_NOTIFIERS
551 /* list of struct preempt_notifier: */
552 struct hlist_head preempt_notifiers;
555 #ifdef CONFIG_BLK_DEV_IO_TRACE
556 unsigned int btrace_seq;
561 cpumask_t cpus_allowed;
563 #ifdef CONFIG_PREEMPT_RCU
564 int rcu_read_lock_nesting;
565 union rcu_special rcu_read_unlock_special;
566 struct list_head rcu_node_entry;
567 struct rcu_node *rcu_blocked_node;
568 #endif /* #ifdef CONFIG_PREEMPT_RCU */
569 #ifdef CONFIG_TASKS_RCU
570 unsigned long rcu_tasks_nvcsw;
571 bool rcu_tasks_holdout;
572 struct list_head rcu_tasks_holdout_list;
573 int rcu_tasks_idle_cpu;
574 #endif /* #ifdef CONFIG_TASKS_RCU */
576 #ifdef CONFIG_SCHED_INFO
577 struct sched_info sched_info;
580 struct list_head tasks;
582 struct plist_node pushable_tasks;
583 struct rb_node pushable_dl_tasks;
586 struct mm_struct *mm, *active_mm;
588 /* Per-thread vma caching: */
589 struct vmacache vmacache;
591 #if defined(SPLIT_RSS_COUNTING)
592 struct task_rss_stat rss_stat;
596 int exit_code, exit_signal;
597 int pdeath_signal; /* The signal sent when the parent dies */
598 unsigned long jobctl; /* JOBCTL_*, siglock protected */
600 /* Used for emulating ABI behavior of previous Linux versions */
601 unsigned int personality;
603 /* scheduler bits, serialized by scheduler locks */
604 unsigned sched_reset_on_fork:1;
605 unsigned sched_contributes_to_load:1;
606 unsigned sched_migrated:1;
607 unsigned sched_remote_wakeup:1;
608 unsigned :0; /* force alignment to the next boundary */
610 /* unserialized, strictly 'current' */
611 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
612 unsigned in_iowait:1;
613 #if !defined(TIF_RESTORE_SIGMASK)
614 unsigned restore_sigmask:1;
617 unsigned memcg_may_oom:1;
619 unsigned memcg_kmem_skip_account:1;
622 #ifdef CONFIG_COMPAT_BRK
623 unsigned brk_randomized:1;
626 unsigned long atomic_flags; /* Flags needing atomic access. */
628 struct restart_block restart_block;
633 #ifdef CONFIG_CC_STACKPROTECTOR
634 /* Canary value for the -fstack-protector gcc feature */
635 unsigned long stack_canary;
638 * pointers to (original) parent process, youngest child, younger sibling,
639 * older sibling, respectively. (p->father can be replaced with
640 * p->real_parent->pid)
642 struct task_struct __rcu *real_parent; /* real parent process */
643 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
645 * children/sibling forms the list of my natural children
647 struct list_head children; /* list of my children */
648 struct list_head sibling; /* linkage in my parent's children list */
649 struct task_struct *group_leader; /* threadgroup leader */
652 * ptraced is the list of tasks this task is using ptrace on.
653 * This includes both natural children and PTRACE_ATTACH targets.
654 * p->ptrace_entry is p's link on the p->parent->ptraced list.
656 struct list_head ptraced;
657 struct list_head ptrace_entry;
659 /* PID/PID hash table linkage. */
660 struct pid_link pids[PIDTYPE_MAX];
661 struct list_head thread_group;
662 struct list_head thread_node;
664 struct completion *vfork_done; /* for vfork() */
665 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
666 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
669 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
670 u64 utimescaled, stimescaled;
673 struct prev_cputime prev_cputime;
674 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
675 seqcount_t vtime_seqcount;
676 unsigned long long vtime_snap;
678 /* Task is sleeping or running in a CPU with VTIME inactive */
680 /* Task runs in userspace in a CPU with VTIME active */
682 /* Task runs in kernelspace in a CPU with VTIME active */
687 #ifdef CONFIG_NO_HZ_FULL
688 atomic_t tick_dep_mask;
690 unsigned long nvcsw, nivcsw; /* context switch counts */
691 u64 start_time; /* monotonic time in nsec */
692 u64 real_start_time; /* boot based time in nsec */
693 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
694 unsigned long min_flt, maj_flt;
696 #ifdef CONFIG_POSIX_TIMERS
697 struct task_cputime cputime_expires;
698 struct list_head cpu_timers[3];
701 /* process credentials */
702 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
703 const struct cred __rcu *real_cred; /* objective and real subjective task
704 * credentials (COW) */
705 const struct cred __rcu *cred; /* effective (overridable) subjective task
706 * credentials (COW) */
707 char comm[TASK_COMM_LEN]; /* executable name excluding path
708 - access with [gs]et_task_comm (which lock
710 - initialized normally by setup_new_exec */
711 /* file system info */
712 struct nameidata *nameidata;
713 #ifdef CONFIG_SYSVIPC
715 struct sysv_sem sysvsem;
716 struct sysv_shm sysvshm;
718 #ifdef CONFIG_DETECT_HUNG_TASK
719 /* hung task detection */
720 unsigned long last_switch_count;
722 /* filesystem information */
723 struct fs_struct *fs;
724 /* open file information */
725 struct files_struct *files;
727 struct nsproxy *nsproxy;
728 /* signal handlers */
729 struct signal_struct *signal;
730 struct sighand_struct *sighand;
732 sigset_t blocked, real_blocked;
733 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
734 struct sigpending pending;
736 unsigned long sas_ss_sp;
738 unsigned sas_ss_flags;
740 struct callback_head *task_works;
742 struct audit_context *audit_context;
743 #ifdef CONFIG_AUDITSYSCALL
745 unsigned int sessionid;
747 struct seccomp seccomp;
749 /* Thread group tracking */
752 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
754 spinlock_t alloc_lock;
756 /* Protection of the PI data structures: */
757 raw_spinlock_t pi_lock;
759 struct wake_q_node wake_q;
761 #ifdef CONFIG_RT_MUTEXES
762 /* PI waiters blocked on a rt_mutex held by this task */
763 struct rb_root pi_waiters;
764 struct rb_node *pi_waiters_leftmost;
765 /* Deadlock detection and priority inheritance handling */
766 struct rt_mutex_waiter *pi_blocked_on;
769 #ifdef CONFIG_DEBUG_MUTEXES
770 /* mutex deadlock detection */
771 struct mutex_waiter *blocked_on;
773 #ifdef CONFIG_TRACE_IRQFLAGS
774 unsigned int irq_events;
775 unsigned long hardirq_enable_ip;
776 unsigned long hardirq_disable_ip;
777 unsigned int hardirq_enable_event;
778 unsigned int hardirq_disable_event;
779 int hardirqs_enabled;
781 unsigned long softirq_disable_ip;
782 unsigned long softirq_enable_ip;
783 unsigned int softirq_disable_event;
784 unsigned int softirq_enable_event;
785 int softirqs_enabled;
788 #ifdef CONFIG_LOCKDEP
789 # define MAX_LOCK_DEPTH 48UL
792 unsigned int lockdep_recursion;
793 struct held_lock held_locks[MAX_LOCK_DEPTH];
794 gfp_t lockdep_reclaim_gfp;
797 unsigned int in_ubsan;
800 /* journalling filesystem info */
803 /* stacked block device info */
804 struct bio_list *bio_list;
808 struct blk_plug *plug;
812 struct reclaim_state *reclaim_state;
814 struct backing_dev_info *backing_dev_info;
816 struct io_context *io_context;
818 unsigned long ptrace_message;
819 siginfo_t *last_siginfo; /* For ptrace use. */
820 struct task_io_accounting ioac;
821 #if defined(CONFIG_TASK_XACCT)
822 u64 acct_rss_mem1; /* accumulated rss usage */
823 u64 acct_vm_mem1; /* accumulated virtual memory usage */
824 u64 acct_timexpd; /* stime + utime since last update */
826 #ifdef CONFIG_CPUSETS
827 nodemask_t mems_allowed; /* Protected by alloc_lock */
828 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
829 int cpuset_mem_spread_rotor;
830 int cpuset_slab_spread_rotor;
832 #ifdef CONFIG_CGROUPS
833 /* Control Group info protected by css_set_lock */
834 struct css_set __rcu *cgroups;
835 /* cg_list protected by css_set_lock and tsk->alloc_lock */
836 struct list_head cg_list;
838 #ifdef CONFIG_INTEL_RDT_A
842 struct robust_list_head __user *robust_list;
844 struct compat_robust_list_head __user *compat_robust_list;
846 struct list_head pi_state_list;
847 struct futex_pi_state *pi_state_cache;
849 #ifdef CONFIG_PERF_EVENTS
850 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
851 struct mutex perf_event_mutex;
852 struct list_head perf_event_list;
854 #ifdef CONFIG_DEBUG_PREEMPT
855 unsigned long preempt_disable_ip;
858 struct mempolicy *mempolicy; /* Protected by alloc_lock */
860 short pref_node_fork;
862 #ifdef CONFIG_NUMA_BALANCING
864 unsigned int numa_scan_period;
865 unsigned int numa_scan_period_max;
866 int numa_preferred_nid;
867 unsigned long numa_migrate_retry;
868 u64 node_stamp; /* migration stamp */
869 u64 last_task_numa_placement;
870 u64 last_sum_exec_runtime;
871 struct callback_head numa_work;
873 struct list_head numa_entry;
874 struct numa_group *numa_group;
877 * numa_faults is an array split into four regions:
878 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
879 * in this precise order.
881 * faults_memory: Exponential decaying average of faults on a per-node
882 * basis. Scheduling placement decisions are made based on these
883 * counts. The values remain static for the duration of a PTE scan.
884 * faults_cpu: Track the nodes the process was running on when a NUMA
885 * hinting fault was incurred.
886 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
887 * during the current scan window. When the scan completes, the counts
888 * in faults_memory and faults_cpu decay and these values are copied.
890 unsigned long *numa_faults;
891 unsigned long total_numa_faults;
894 * numa_faults_locality tracks if faults recorded during the last
895 * scan window were remote/local or failed to migrate. The task scan
896 * period is adapted based on the locality of the faults with different
897 * weights depending on whether they were shared or private faults
899 unsigned long numa_faults_locality[3];
901 unsigned long numa_pages_migrated;
902 #endif /* CONFIG_NUMA_BALANCING */
904 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
905 struct tlbflush_unmap_batch tlb_ubc;
911 * cache last used pipe for splice
913 struct pipe_inode_info *splice_pipe;
915 struct page_frag task_frag;
917 #ifdef CONFIG_TASK_DELAY_ACCT
918 struct task_delay_info *delays;
921 #ifdef CONFIG_FAULT_INJECTION
925 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
926 * balance_dirty_pages() for some dirty throttling pause
929 int nr_dirtied_pause;
930 unsigned long dirty_paused_when; /* start of a write-and-pause period */
932 #ifdef CONFIG_LATENCYTOP
933 int latency_record_count;
934 struct latency_record latency_record[LT_SAVECOUNT];
937 * time slack values; these are used to round up poll() and
938 * select() etc timeout values. These are in nanoseconds.
941 u64 default_timer_slack_ns;
944 unsigned int kasan_depth;
946 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
947 /* Index of current stored address in ret_stack */
949 /* Stack of return addresses for return function tracing */
950 struct ftrace_ret_stack *ret_stack;
951 /* time stamp for last schedule */
952 unsigned long long ftrace_timestamp;
954 * Number of functions that haven't been traced
955 * because of depth overrun.
957 atomic_t trace_overrun;
958 /* Pause for the tracing */
959 atomic_t tracing_graph_pause;
961 #ifdef CONFIG_TRACING
962 /* state flags for use by tracers */
964 /* bitmask and counter of trace recursion */
965 unsigned long trace_recursion;
966 #endif /* CONFIG_TRACING */
968 /* Coverage collection mode enabled for this task (0 if disabled). */
969 enum kcov_mode kcov_mode;
970 /* Size of the kcov_area. */
972 /* Buffer for coverage collection. */
974 /* kcov desciptor wired with this task or NULL. */
978 struct mem_cgroup *memcg_in_oom;
979 gfp_t memcg_oom_gfp_mask;
982 /* number of pages to reclaim on returning to userland */
983 unsigned int memcg_nr_pages_over_high;
985 #ifdef CONFIG_UPROBES
986 struct uprobe_task *utask;
988 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
989 unsigned int sequential_io;
990 unsigned int sequential_io_avg;
992 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
993 unsigned long task_state_change;
995 int pagefault_disabled;
997 struct task_struct *oom_reaper_list;
999 #ifdef CONFIG_VMAP_STACK
1000 struct vm_struct *stack_vm_area;
1002 #ifdef CONFIG_THREAD_INFO_IN_TASK
1003 /* A live task holds one reference. */
1004 atomic_t stack_refcount;
1006 /* CPU-specific state of this task */
1007 struct thread_struct thread;
1009 * WARNING: on x86, 'thread_struct' contains a variable-sized
1010 * structure. It *MUST* be at the end of 'task_struct'.
1012 * Do not put anything below here!
1016 static inline struct pid *task_pid(struct task_struct *task)
1018 return task->pids[PIDTYPE_PID].pid;
1021 static inline struct pid *task_tgid(struct task_struct *task)
1023 return task->group_leader->pids[PIDTYPE_PID].pid;
1027 * Without tasklist or rcu lock it is not safe to dereference
1028 * the result of task_pgrp/task_session even if task == current,
1029 * we can race with another thread doing sys_setsid/sys_setpgid.
1031 static inline struct pid *task_pgrp(struct task_struct *task)
1033 return task->group_leader->pids[PIDTYPE_PGID].pid;
1036 static inline struct pid *task_session(struct task_struct *task)
1038 return task->group_leader->pids[PIDTYPE_SID].pid;
1042 * the helpers to get the task's different pids as they are seen
1043 * from various namespaces
1045 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1046 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1048 * task_xid_nr_ns() : id seen from the ns specified;
1050 * set_task_vxid() : assigns a virtual id to a task;
1052 * see also pid_nr() etc in include/linux/pid.h
1054 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1055 struct pid_namespace *ns);
1057 static inline pid_t task_pid_nr(struct task_struct *tsk)
1062 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1063 struct pid_namespace *ns)
1065 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1068 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1070 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1074 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1079 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1081 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1083 return pid_vnr(task_tgid(tsk));
1087 static inline int pid_alive(const struct task_struct *p);
1088 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1094 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1100 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1102 return task_ppid_nr_ns(tsk, &init_pid_ns);
1105 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1106 struct pid_namespace *ns)
1108 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1111 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1113 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1117 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1118 struct pid_namespace *ns)
1120 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1123 static inline pid_t task_session_vnr(struct task_struct *tsk)
1125 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1128 /* obsolete, do not use */
1129 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1131 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1135 * pid_alive - check that a task structure is not stale
1136 * @p: Task structure to be checked.
1138 * Test if a process is not yet dead (at most zombie state)
1139 * If pid_alive fails, then pointers within the task structure
1140 * can be stale and must not be dereferenced.
1142 * Return: 1 if the process is alive. 0 otherwise.
1144 static inline int pid_alive(const struct task_struct *p)
1146 return p->pids[PIDTYPE_PID].pid != NULL;
1150 * is_global_init - check if a task structure is init. Since init
1151 * is free to have sub-threads we need to check tgid.
1152 * @tsk: Task structure to be checked.
1154 * Check if a task structure is the first user space task the kernel created.
1156 * Return: 1 if the task structure is init. 0 otherwise.
1158 static inline int is_global_init(struct task_struct *tsk)
1160 return task_tgid_nr(tsk) == 1;
1163 extern struct pid *cad_pid;
1165 extern void free_task(struct task_struct *tsk);
1166 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
1168 extern void __put_task_struct(struct task_struct *t);
1170 static inline void put_task_struct(struct task_struct *t)
1172 if (atomic_dec_and_test(&t->usage))
1173 __put_task_struct(t);
1176 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
1177 struct task_struct *try_get_task_struct(struct task_struct **ptask);
1182 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1183 #define PF_EXITING 0x00000004 /* getting shut down */
1184 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1185 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1186 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1187 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1188 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1189 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1190 #define PF_DUMPCORE 0x00000200 /* dumped core */
1191 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1192 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1193 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1194 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1195 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1196 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1197 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1198 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1199 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1200 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1201 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1202 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1203 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1204 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1205 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1206 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1207 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1208 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1209 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1212 * Only the _current_ task can read/write to tsk->flags, but other
1213 * tasks can access tsk->flags in readonly mode for example
1214 * with tsk_used_math (like during threaded core dumping).
1215 * There is however an exception to this rule during ptrace
1216 * or during fork: the ptracer task is allowed to write to the
1217 * child->flags of its traced child (same goes for fork, the parent
1218 * can write to the child->flags), because we're guaranteed the
1219 * child is not running and in turn not changing child->flags
1220 * at the same time the parent does it.
1222 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1223 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1224 #define clear_used_math() clear_stopped_child_used_math(current)
1225 #define set_used_math() set_stopped_child_used_math(current)
1226 #define conditional_stopped_child_used_math(condition, child) \
1227 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1228 #define conditional_used_math(condition) \
1229 conditional_stopped_child_used_math(condition, current)
1230 #define copy_to_stopped_child_used_math(child) \
1231 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1232 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1233 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1234 #define used_math() tsk_used_math(current)
1236 /* Per-process atomic flags. */
1237 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1238 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1239 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1240 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1243 #define TASK_PFA_TEST(name, func) \
1244 static inline bool task_##func(struct task_struct *p) \
1245 { return test_bit(PFA_##name, &p->atomic_flags); }
1246 #define TASK_PFA_SET(name, func) \
1247 static inline void task_set_##func(struct task_struct *p) \
1248 { set_bit(PFA_##name, &p->atomic_flags); }
1249 #define TASK_PFA_CLEAR(name, func) \
1250 static inline void task_clear_##func(struct task_struct *p) \
1251 { clear_bit(PFA_##name, &p->atomic_flags); }
1253 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1254 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1256 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1257 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1258 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1260 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1261 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1262 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1264 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1265 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1267 static inline void tsk_restore_flags(struct task_struct *task,
1268 unsigned long orig_flags, unsigned long flags)
1270 task->flags &= ~flags;
1271 task->flags |= orig_flags & flags;
1274 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1275 const struct cpumask *trial);
1276 extern int task_can_attach(struct task_struct *p,
1277 const struct cpumask *cs_cpus_allowed);
1279 extern void do_set_cpus_allowed(struct task_struct *p,
1280 const struct cpumask *new_mask);
1282 extern int set_cpus_allowed_ptr(struct task_struct *p,
1283 const struct cpumask *new_mask);
1285 static inline void do_set_cpus_allowed(struct task_struct *p,
1286 const struct cpumask *new_mask)
1289 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1290 const struct cpumask *new_mask)
1292 if (!cpumask_test_cpu(0, new_mask))
1298 #ifndef cpu_relax_yield
1299 #define cpu_relax_yield() cpu_relax()
1302 /* sched_exec is called by processes performing an exec */
1304 extern void sched_exec(void);
1306 #define sched_exec() {}
1309 extern int yield_to(struct task_struct *p, bool preempt);
1310 extern void set_user_nice(struct task_struct *p, long nice);
1311 extern int task_prio(const struct task_struct *p);
1313 * task_nice - return the nice value of a given task.
1314 * @p: the task in question.
1316 * Return: The nice value [ -20 ... 0 ... 19 ].
1318 static inline int task_nice(const struct task_struct *p)
1320 return PRIO_TO_NICE((p)->static_prio);
1322 extern int can_nice(const struct task_struct *p, const int nice);
1323 extern int task_curr(const struct task_struct *p);
1324 extern int idle_cpu(int cpu);
1325 extern int sched_setscheduler(struct task_struct *, int,
1326 const struct sched_param *);
1327 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1328 const struct sched_param *);
1329 extern int sched_setattr(struct task_struct *,
1330 const struct sched_attr *);
1331 extern struct task_struct *idle_task(int cpu);
1333 * is_idle_task - is the specified task an idle task?
1334 * @p: the task in question.
1336 * Return: 1 if @p is an idle task. 0 otherwise.
1338 static inline bool is_idle_task(const struct task_struct *p)
1340 return !!(p->flags & PF_IDLE);
1342 extern struct task_struct *curr_task(int cpu);
1343 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1347 union thread_union {
1348 #ifndef CONFIG_THREAD_INFO_IN_TASK
1349 struct thread_info thread_info;
1351 unsigned long stack[THREAD_SIZE/sizeof(long)];
1354 #ifdef CONFIG_THREAD_INFO_IN_TASK
1355 static inline struct thread_info *task_thread_info(struct task_struct *task)
1357 return &task->thread_info;
1359 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1360 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1363 extern struct pid_namespace init_pid_ns;
1366 * find a task by one of its numerical ids
1368 * find_task_by_pid_ns():
1369 * finds a task by its pid in the specified namespace
1370 * find_task_by_vpid():
1371 * finds a task by its virtual pid
1373 * see also find_vpid() etc in include/linux/pid.h
1376 extern struct task_struct *find_task_by_vpid(pid_t nr);
1377 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1378 struct pid_namespace *ns);
1380 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1381 extern int wake_up_process(struct task_struct *tsk);
1382 extern void wake_up_new_task(struct task_struct *tsk);
1384 extern void kick_process(struct task_struct *tsk);
1386 static inline void kick_process(struct task_struct *tsk) { }
1389 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1390 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1392 __set_task_comm(tsk, from, false);
1394 extern char *get_task_comm(char *to, struct task_struct *tsk);
1397 void scheduler_ipi(void);
1398 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1400 static inline void scheduler_ipi(void) { }
1401 static inline unsigned long wait_task_inactive(struct task_struct *p,
1408 /* set thread flags in other task's structures
1409 * - see asm/thread_info.h for TIF_xxxx flags available
1411 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1413 set_ti_thread_flag(task_thread_info(tsk), flag);
1416 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1418 clear_ti_thread_flag(task_thread_info(tsk), flag);
1421 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1423 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1426 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1428 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1431 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1433 return test_ti_thread_flag(task_thread_info(tsk), flag);
1436 static inline void set_tsk_need_resched(struct task_struct *tsk)
1438 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1441 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1443 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1446 static inline int test_tsk_need_resched(struct task_struct *tsk)
1448 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1452 * cond_resched() and cond_resched_lock(): latency reduction via
1453 * explicit rescheduling in places that are safe. The return
1454 * value indicates whether a reschedule was done in fact.
1455 * cond_resched_lock() will drop the spinlock before scheduling,
1456 * cond_resched_softirq() will enable bhs before scheduling.
1458 #ifndef CONFIG_PREEMPT
1459 extern int _cond_resched(void);
1461 static inline int _cond_resched(void) { return 0; }
1464 #define cond_resched() ({ \
1465 ___might_sleep(__FILE__, __LINE__, 0); \
1469 extern int __cond_resched_lock(spinlock_t *lock);
1471 #define cond_resched_lock(lock) ({ \
1472 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1473 __cond_resched_lock(lock); \
1476 extern int __cond_resched_softirq(void);
1478 #define cond_resched_softirq() ({ \
1479 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1480 __cond_resched_softirq(); \
1483 static inline void cond_resched_rcu(void)
1485 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1493 * Does a critical section need to be broken due to another
1494 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1495 * but a general need for low latency)
1497 static inline int spin_needbreak(spinlock_t *lock)
1499 #ifdef CONFIG_PREEMPT
1500 return spin_is_contended(lock);
1506 static __always_inline bool need_resched(void)
1508 return unlikely(tif_need_resched());
1512 * Wrappers for p->thread_info->cpu access. No-op on UP.
1516 static inline unsigned int task_cpu(const struct task_struct *p)
1518 #ifdef CONFIG_THREAD_INFO_IN_TASK
1521 return task_thread_info(p)->cpu;
1525 static inline int task_node(const struct task_struct *p)
1527 return cpu_to_node(task_cpu(p));
1530 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1534 static inline unsigned int task_cpu(const struct task_struct *p)
1539 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1543 #endif /* CONFIG_SMP */
1546 * In order to reduce various lock holder preemption latencies provide an
1547 * interface to see if a vCPU is currently running or not.
1549 * This allows us to terminate optimistic spin loops and block, analogous to
1550 * the native optimistic spin heuristic of testing if the lock owner task is
1553 #ifndef vcpu_is_preempted
1554 # define vcpu_is_preempted(cpu) false
1557 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1558 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1560 extern int task_can_switch_user(struct user_struct *up,
1561 struct task_struct *tsk);
1563 #ifndef TASK_SIZE_OF
1564 #define TASK_SIZE_OF(tsk) TASK_SIZE