4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
8 #include <linux/mutex.h>
9 #include <linux/plist.h>
10 #include <linux/mm_types_task.h>
12 #include <linux/sem.h>
13 #include <linux/shm.h>
14 #include <linux/signal_types.h>
15 #include <linux/pid.h>
16 #include <linux/seccomp.h>
17 #include <linux/rculist.h>
19 #include <linux/resource.h>
20 #include <linux/hrtimer.h>
21 #include <linux/kcov.h>
22 #include <linux/task_io_accounting.h>
23 #include <linux/latencytop.h>
24 #include <linux/topology.h>
25 #include <linux/magic.h>
27 #include <asm/current.h>
29 /* task_struct member predeclarations: */
32 struct backing_dev_info;
38 struct futex_pi_state;
43 struct perf_event_context;
45 struct pipe_inode_info;
48 struct robust_list_head;
52 struct sighand_struct;
54 struct task_delay_info;
60 * Task state bitmask. NOTE! These bits are also
61 * encoded in fs/proc/array.c: get_task_state().
63 * We have two separate sets of flags: task->state
64 * is about runnability, while task->exit_state are
65 * about the task exiting. Confusing, but this way
66 * modifying one set can't modify the other one by
69 #define TASK_RUNNING 0
70 #define TASK_INTERRUPTIBLE 1
71 #define TASK_UNINTERRUPTIBLE 2
72 #define __TASK_STOPPED 4
73 #define __TASK_TRACED 8
74 /* in tsk->exit_state */
76 #define EXIT_ZOMBIE 32
77 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
78 /* in tsk->state again */
80 #define TASK_WAKEKILL 128
81 #define TASK_WAKING 256
82 #define TASK_PARKED 512
83 #define TASK_NOLOAD 1024
85 #define TASK_STATE_MAX 4096
87 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
89 /* Convenience macros for the sake of set_current_state */
90 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
91 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
92 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
94 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
96 /* Convenience macros for the sake of wake_up */
97 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
98 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
100 /* get_task_state() */
101 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
102 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
103 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
105 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
107 #define task_is_stopped_or_traced(task) \
108 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
109 #define task_contributes_to_load(task) \
110 ((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) \
163 do { current->state = (state_value); } while (0)
164 #define set_current_state(state_value) \
165 smp_store_mb(current->state, (state_value))
169 /* Task command name length */
170 #define TASK_COMM_LEN 16
172 extern cpumask_var_t cpu_isolated_map;
174 extern void scheduler_tick(void);
176 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
177 extern signed long schedule_timeout(signed long timeout);
178 extern signed long schedule_timeout_interruptible(signed long timeout);
179 extern signed long schedule_timeout_killable(signed long timeout);
180 extern signed long schedule_timeout_uninterruptible(signed long timeout);
181 extern signed long schedule_timeout_idle(signed long timeout);
182 asmlinkage void schedule(void);
183 extern void schedule_preempt_disabled(void);
185 extern int __must_check io_schedule_prepare(void);
186 extern void io_schedule_finish(int token);
187 extern long io_schedule_timeout(long timeout);
188 extern void io_schedule(void);
191 * struct prev_cputime - snaphsot of system and user cputime
192 * @utime: time spent in user mode
193 * @stime: time spent in system mode
194 * @lock: protects the above two fields
196 * Stores previous user/system time values such that we can guarantee
199 struct prev_cputime {
200 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
208 * struct task_cputime - collected CPU time counts
209 * @utime: time spent in user mode, in nanoseconds
210 * @stime: time spent in kernel mode, in nanoseconds
211 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
213 * This structure groups together three kinds of CPU time that are tracked for
214 * threads and thread groups. Most things considering CPU time want to group
215 * these counts together and treat all three of them in parallel.
217 struct task_cputime {
220 unsigned long long sum_exec_runtime;
223 /* Alternate field names when used to cache expirations. */
224 #define virt_exp utime
225 #define prof_exp stime
226 #define sched_exp sum_exec_runtime
228 #ifdef CONFIG_SCHED_INFO
230 /* cumulative counters */
231 unsigned long pcount; /* # of times run on this cpu */
232 unsigned long long run_delay; /* time spent waiting on a runqueue */
235 unsigned long long last_arrival,/* when we last ran on a cpu */
236 last_queued; /* when we were last queued to run */
238 #endif /* CONFIG_SCHED_INFO */
241 * Integer metrics need fixed point arithmetic, e.g., sched/fair
242 * has a few: load, load_avg, util_avg, freq, and capacity.
244 * We define a basic fixed point arithmetic range, and then formalize
245 * all these metrics based on that basic range.
247 # define SCHED_FIXEDPOINT_SHIFT 10
248 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
250 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
251 extern void prefetch_stack(struct task_struct *t);
253 static inline void prefetch_stack(struct task_struct *t) { }
257 unsigned long weight;
262 * The load_avg/util_avg accumulates an infinite geometric series
263 * (see __update_load_avg() in kernel/sched/fair.c).
265 * [load_avg definition]
267 * load_avg = runnable% * scale_load_down(load)
269 * where runnable% is the time ratio that a sched_entity is runnable.
270 * For cfs_rq, it is the aggregated load_avg of all runnable and
271 * blocked sched_entities.
273 * load_avg may also take frequency scaling into account:
275 * load_avg = runnable% * scale_load_down(load) * freq%
277 * where freq% is the CPU frequency normalized to the highest frequency.
279 * [util_avg definition]
281 * util_avg = running% * SCHED_CAPACITY_SCALE
283 * where running% is the time ratio that a sched_entity is running on
284 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
285 * and blocked sched_entities.
287 * util_avg may also factor frequency scaling and CPU capacity scaling:
289 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
291 * where freq% is the same as above, and capacity% is the CPU capacity
292 * normalized to the greatest capacity (due to uarch differences, etc).
294 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
295 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
296 * we therefore scale them to as large a range as necessary. This is for
297 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
301 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
302 * with the highest load (=88761), always runnable on a single cfs_rq,
303 * and should not overflow as the number already hits PID_MAX_LIMIT.
305 * For all other cases (including 32-bit kernels), struct load_weight's
306 * weight will overflow first before we do, because:
308 * Max(load_avg) <= Max(load.weight)
310 * Then it is the load_weight's responsibility to consider overflow
314 u64 last_update_time, load_sum;
315 u32 util_sum, period_contrib;
316 unsigned long load_avg, util_avg;
319 #ifdef CONFIG_SCHEDSTATS
320 struct sched_statistics {
330 s64 sum_sleep_runtime;
337 u64 nr_migrations_cold;
338 u64 nr_failed_migrations_affine;
339 u64 nr_failed_migrations_running;
340 u64 nr_failed_migrations_hot;
341 u64 nr_forced_migrations;
345 u64 nr_wakeups_migrate;
346 u64 nr_wakeups_local;
347 u64 nr_wakeups_remote;
348 u64 nr_wakeups_affine;
349 u64 nr_wakeups_affine_attempts;
350 u64 nr_wakeups_passive;
355 struct sched_entity {
356 struct load_weight load; /* for load-balancing */
357 struct rb_node run_node;
358 struct list_head group_node;
362 u64 sum_exec_runtime;
364 u64 prev_sum_exec_runtime;
368 #ifdef CONFIG_SCHEDSTATS
369 struct sched_statistics statistics;
372 #ifdef CONFIG_FAIR_GROUP_SCHED
374 struct sched_entity *parent;
375 /* rq on which this entity is (to be) queued: */
376 struct cfs_rq *cfs_rq;
377 /* rq "owned" by this entity/group: */
383 * Per entity load average tracking.
385 * Put into separate cache line so it does not
386 * collide with read-mostly values above.
388 struct sched_avg avg ____cacheline_aligned_in_smp;
392 struct sched_rt_entity {
393 struct list_head run_list;
394 unsigned long timeout;
395 unsigned long watchdog_stamp;
396 unsigned int time_slice;
397 unsigned short on_rq;
398 unsigned short on_list;
400 struct sched_rt_entity *back;
401 #ifdef CONFIG_RT_GROUP_SCHED
402 struct sched_rt_entity *parent;
403 /* rq on which this entity is (to be) queued: */
405 /* rq "owned" by this entity/group: */
410 struct sched_dl_entity {
411 struct rb_node rb_node;
414 * Original scheduling parameters. Copied here from sched_attr
415 * during sched_setattr(), they will remain the same until
416 * the next sched_setattr().
418 u64 dl_runtime; /* maximum runtime for each instance */
419 u64 dl_deadline; /* relative deadline of each instance */
420 u64 dl_period; /* separation of two instances (period) */
421 u64 dl_bw; /* dl_runtime / dl_deadline */
424 * Actual scheduling parameters. Initialized with the values above,
425 * they are continously updated during task execution. Note that
426 * the remaining runtime could be < 0 in case we are in overrun.
428 s64 runtime; /* remaining runtime for this instance */
429 u64 deadline; /* absolute deadline for this instance */
430 unsigned int flags; /* specifying the scheduler behaviour */
435 * @dl_throttled tells if we exhausted the runtime. If so, the
436 * task has to wait for a replenishment to be performed at the
437 * next firing of dl_timer.
439 * @dl_boosted tells if we are boosted due to DI. If so we are
440 * outside bandwidth enforcement mechanism (but only until we
441 * exit the critical section);
443 * @dl_yielded tells if task gave up the cpu before consuming
444 * all its available runtime during the last job.
446 int dl_throttled, dl_boosted, dl_yielded;
449 * Bandwidth enforcement timer. Each -deadline task has its
450 * own bandwidth to be enforced, thus we need one timer per task.
452 struct hrtimer dl_timer;
460 u8 pad; /* Otherwise the compiler can store garbage here. */
462 u32 s; /* Set of bits. */
465 enum perf_event_task_context {
466 perf_invalid_context = -1,
469 perf_nr_task_contexts,
473 struct wake_q_node *next;
477 #ifdef CONFIG_THREAD_INFO_IN_TASK
479 * For reasons of header soup (see current_thread_info()), this
480 * must be the first element of task_struct.
482 struct thread_info thread_info;
484 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
487 unsigned int flags; /* per process flags, defined below */
491 struct llist_node wake_entry;
493 #ifdef CONFIG_THREAD_INFO_IN_TASK
494 unsigned int cpu; /* current CPU */
496 unsigned int wakee_flips;
497 unsigned long wakee_flip_decay_ts;
498 struct task_struct *last_wakee;
504 int prio, static_prio, normal_prio;
505 unsigned int rt_priority;
506 const struct sched_class *sched_class;
507 struct sched_entity se;
508 struct sched_rt_entity rt;
509 #ifdef CONFIG_CGROUP_SCHED
510 struct task_group *sched_task_group;
512 struct sched_dl_entity dl;
514 #ifdef CONFIG_PREEMPT_NOTIFIERS
515 /* list of struct preempt_notifier: */
516 struct hlist_head preempt_notifiers;
519 #ifdef CONFIG_BLK_DEV_IO_TRACE
520 unsigned int btrace_seq;
525 cpumask_t cpus_allowed;
527 #ifdef CONFIG_PREEMPT_RCU
528 int rcu_read_lock_nesting;
529 union rcu_special rcu_read_unlock_special;
530 struct list_head rcu_node_entry;
531 struct rcu_node *rcu_blocked_node;
532 #endif /* #ifdef CONFIG_PREEMPT_RCU */
533 #ifdef CONFIG_TASKS_RCU
534 unsigned long rcu_tasks_nvcsw;
535 bool rcu_tasks_holdout;
536 struct list_head rcu_tasks_holdout_list;
537 int rcu_tasks_idle_cpu;
538 #endif /* #ifdef CONFIG_TASKS_RCU */
540 #ifdef CONFIG_SCHED_INFO
541 struct sched_info sched_info;
544 struct list_head tasks;
546 struct plist_node pushable_tasks;
547 struct rb_node pushable_dl_tasks;
550 struct mm_struct *mm, *active_mm;
552 /* Per-thread vma caching: */
553 struct vmacache vmacache;
555 #if defined(SPLIT_RSS_COUNTING)
556 struct task_rss_stat rss_stat;
560 int exit_code, exit_signal;
561 int pdeath_signal; /* The signal sent when the parent dies */
562 unsigned long jobctl; /* JOBCTL_*, siglock protected */
564 /* Used for emulating ABI behavior of previous Linux versions */
565 unsigned int personality;
567 /* scheduler bits, serialized by scheduler locks */
568 unsigned sched_reset_on_fork:1;
569 unsigned sched_contributes_to_load:1;
570 unsigned sched_migrated:1;
571 unsigned sched_remote_wakeup:1;
572 unsigned :0; /* force alignment to the next boundary */
574 /* unserialized, strictly 'current' */
575 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
576 unsigned in_iowait:1;
577 #if !defined(TIF_RESTORE_SIGMASK)
578 unsigned restore_sigmask:1;
581 unsigned memcg_may_oom:1;
583 unsigned memcg_kmem_skip_account:1;
586 #ifdef CONFIG_COMPAT_BRK
587 unsigned brk_randomized:1;
590 unsigned long atomic_flags; /* Flags needing atomic access. */
592 struct restart_block restart_block;
597 #ifdef CONFIG_CC_STACKPROTECTOR
598 /* Canary value for the -fstack-protector gcc feature */
599 unsigned long stack_canary;
602 * pointers to (original) parent process, youngest child, younger sibling,
603 * older sibling, respectively. (p->father can be replaced with
604 * p->real_parent->pid)
606 struct task_struct __rcu *real_parent; /* real parent process */
607 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
609 * children/sibling forms the list of my natural children
611 struct list_head children; /* list of my children */
612 struct list_head sibling; /* linkage in my parent's children list */
613 struct task_struct *group_leader; /* threadgroup leader */
616 * ptraced is the list of tasks this task is using ptrace on.
617 * This includes both natural children and PTRACE_ATTACH targets.
618 * p->ptrace_entry is p's link on the p->parent->ptraced list.
620 struct list_head ptraced;
621 struct list_head ptrace_entry;
623 /* PID/PID hash table linkage. */
624 struct pid_link pids[PIDTYPE_MAX];
625 struct list_head thread_group;
626 struct list_head thread_node;
628 struct completion *vfork_done; /* for vfork() */
629 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
630 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
633 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
634 u64 utimescaled, stimescaled;
637 struct prev_cputime prev_cputime;
638 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
639 seqcount_t vtime_seqcount;
640 unsigned long long vtime_snap;
642 /* Task is sleeping or running in a CPU with VTIME inactive */
644 /* Task runs in userspace in a CPU with VTIME active */
646 /* Task runs in kernelspace in a CPU with VTIME active */
651 #ifdef CONFIG_NO_HZ_FULL
652 atomic_t tick_dep_mask;
654 unsigned long nvcsw, nivcsw; /* context switch counts */
655 u64 start_time; /* monotonic time in nsec */
656 u64 real_start_time; /* boot based time in nsec */
657 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
658 unsigned long min_flt, maj_flt;
660 #ifdef CONFIG_POSIX_TIMERS
661 struct task_cputime cputime_expires;
662 struct list_head cpu_timers[3];
665 /* process credentials */
666 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
667 const struct cred __rcu *real_cred; /* objective and real subjective task
668 * credentials (COW) */
669 const struct cred __rcu *cred; /* effective (overridable) subjective task
670 * credentials (COW) */
671 char comm[TASK_COMM_LEN]; /* executable name excluding path
672 - access with [gs]et_task_comm (which lock
674 - initialized normally by setup_new_exec */
675 /* file system info */
676 struct nameidata *nameidata;
677 #ifdef CONFIG_SYSVIPC
679 struct sysv_sem sysvsem;
680 struct sysv_shm sysvshm;
682 #ifdef CONFIG_DETECT_HUNG_TASK
683 /* hung task detection */
684 unsigned long last_switch_count;
686 /* filesystem information */
687 struct fs_struct *fs;
688 /* open file information */
689 struct files_struct *files;
691 struct nsproxy *nsproxy;
692 /* signal handlers */
693 struct signal_struct *signal;
694 struct sighand_struct *sighand;
696 sigset_t blocked, real_blocked;
697 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
698 struct sigpending pending;
700 unsigned long sas_ss_sp;
702 unsigned sas_ss_flags;
704 struct callback_head *task_works;
706 struct audit_context *audit_context;
707 #ifdef CONFIG_AUDITSYSCALL
709 unsigned int sessionid;
711 struct seccomp seccomp;
713 /* Thread group tracking */
716 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
718 spinlock_t alloc_lock;
720 /* Protection of the PI data structures: */
721 raw_spinlock_t pi_lock;
723 struct wake_q_node wake_q;
725 #ifdef CONFIG_RT_MUTEXES
726 /* PI waiters blocked on a rt_mutex held by this task */
727 struct rb_root pi_waiters;
728 struct rb_node *pi_waiters_leftmost;
729 /* Deadlock detection and priority inheritance handling */
730 struct rt_mutex_waiter *pi_blocked_on;
733 #ifdef CONFIG_DEBUG_MUTEXES
734 /* mutex deadlock detection */
735 struct mutex_waiter *blocked_on;
737 #ifdef CONFIG_TRACE_IRQFLAGS
738 unsigned int irq_events;
739 unsigned long hardirq_enable_ip;
740 unsigned long hardirq_disable_ip;
741 unsigned int hardirq_enable_event;
742 unsigned int hardirq_disable_event;
743 int hardirqs_enabled;
745 unsigned long softirq_disable_ip;
746 unsigned long softirq_enable_ip;
747 unsigned int softirq_disable_event;
748 unsigned int softirq_enable_event;
749 int softirqs_enabled;
752 #ifdef CONFIG_LOCKDEP
753 # define MAX_LOCK_DEPTH 48UL
756 unsigned int lockdep_recursion;
757 struct held_lock held_locks[MAX_LOCK_DEPTH];
758 gfp_t lockdep_reclaim_gfp;
761 unsigned int in_ubsan;
764 /* journalling filesystem info */
767 /* stacked block device info */
768 struct bio_list *bio_list;
772 struct blk_plug *plug;
776 struct reclaim_state *reclaim_state;
778 struct backing_dev_info *backing_dev_info;
780 struct io_context *io_context;
782 unsigned long ptrace_message;
783 siginfo_t *last_siginfo; /* For ptrace use. */
784 struct task_io_accounting ioac;
785 #if defined(CONFIG_TASK_XACCT)
786 u64 acct_rss_mem1; /* accumulated rss usage */
787 u64 acct_vm_mem1; /* accumulated virtual memory usage */
788 u64 acct_timexpd; /* stime + utime since last update */
790 #ifdef CONFIG_CPUSETS
791 nodemask_t mems_allowed; /* Protected by alloc_lock */
792 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
793 int cpuset_mem_spread_rotor;
794 int cpuset_slab_spread_rotor;
796 #ifdef CONFIG_CGROUPS
797 /* Control Group info protected by css_set_lock */
798 struct css_set __rcu *cgroups;
799 /* cg_list protected by css_set_lock and tsk->alloc_lock */
800 struct list_head cg_list;
802 #ifdef CONFIG_INTEL_RDT_A
806 struct robust_list_head __user *robust_list;
808 struct compat_robust_list_head __user *compat_robust_list;
810 struct list_head pi_state_list;
811 struct futex_pi_state *pi_state_cache;
813 #ifdef CONFIG_PERF_EVENTS
814 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
815 struct mutex perf_event_mutex;
816 struct list_head perf_event_list;
818 #ifdef CONFIG_DEBUG_PREEMPT
819 unsigned long preempt_disable_ip;
822 struct mempolicy *mempolicy; /* Protected by alloc_lock */
824 short pref_node_fork;
826 #ifdef CONFIG_NUMA_BALANCING
828 unsigned int numa_scan_period;
829 unsigned int numa_scan_period_max;
830 int numa_preferred_nid;
831 unsigned long numa_migrate_retry;
832 u64 node_stamp; /* migration stamp */
833 u64 last_task_numa_placement;
834 u64 last_sum_exec_runtime;
835 struct callback_head numa_work;
837 struct list_head numa_entry;
838 struct numa_group *numa_group;
841 * numa_faults is an array split into four regions:
842 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
843 * in this precise order.
845 * faults_memory: Exponential decaying average of faults on a per-node
846 * basis. Scheduling placement decisions are made based on these
847 * counts. The values remain static for the duration of a PTE scan.
848 * faults_cpu: Track the nodes the process was running on when a NUMA
849 * hinting fault was incurred.
850 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
851 * during the current scan window. When the scan completes, the counts
852 * in faults_memory and faults_cpu decay and these values are copied.
854 unsigned long *numa_faults;
855 unsigned long total_numa_faults;
858 * numa_faults_locality tracks if faults recorded during the last
859 * scan window were remote/local or failed to migrate. The task scan
860 * period is adapted based on the locality of the faults with different
861 * weights depending on whether they were shared or private faults
863 unsigned long numa_faults_locality[3];
865 unsigned long numa_pages_migrated;
866 #endif /* CONFIG_NUMA_BALANCING */
868 struct tlbflush_unmap_batch tlb_ubc;
873 * cache last used pipe for splice
875 struct pipe_inode_info *splice_pipe;
877 struct page_frag task_frag;
879 #ifdef CONFIG_TASK_DELAY_ACCT
880 struct task_delay_info *delays;
883 #ifdef CONFIG_FAULT_INJECTION
887 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
888 * balance_dirty_pages() for some dirty throttling pause
891 int nr_dirtied_pause;
892 unsigned long dirty_paused_when; /* start of a write-and-pause period */
894 #ifdef CONFIG_LATENCYTOP
895 int latency_record_count;
896 struct latency_record latency_record[LT_SAVECOUNT];
899 * time slack values; these are used to round up poll() and
900 * select() etc timeout values. These are in nanoseconds.
903 u64 default_timer_slack_ns;
906 unsigned int kasan_depth;
908 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
909 /* Index of current stored address in ret_stack */
911 /* Stack of return addresses for return function tracing */
912 struct ftrace_ret_stack *ret_stack;
913 /* time stamp for last schedule */
914 unsigned long long ftrace_timestamp;
916 * Number of functions that haven't been traced
917 * because of depth overrun.
919 atomic_t trace_overrun;
920 /* Pause for the tracing */
921 atomic_t tracing_graph_pause;
923 #ifdef CONFIG_TRACING
924 /* state flags for use by tracers */
926 /* bitmask and counter of trace recursion */
927 unsigned long trace_recursion;
928 #endif /* CONFIG_TRACING */
930 /* Coverage collection mode enabled for this task (0 if disabled). */
931 enum kcov_mode kcov_mode;
932 /* Size of the kcov_area. */
934 /* Buffer for coverage collection. */
936 /* kcov desciptor wired with this task or NULL. */
940 struct mem_cgroup *memcg_in_oom;
941 gfp_t memcg_oom_gfp_mask;
944 /* number of pages to reclaim on returning to userland */
945 unsigned int memcg_nr_pages_over_high;
947 #ifdef CONFIG_UPROBES
948 struct uprobe_task *utask;
950 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
951 unsigned int sequential_io;
952 unsigned int sequential_io_avg;
954 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
955 unsigned long task_state_change;
957 int pagefault_disabled;
959 struct task_struct *oom_reaper_list;
961 #ifdef CONFIG_VMAP_STACK
962 struct vm_struct *stack_vm_area;
964 #ifdef CONFIG_THREAD_INFO_IN_TASK
965 /* A live task holds one reference. */
966 atomic_t stack_refcount;
968 /* CPU-specific state of this task */
969 struct thread_struct thread;
971 * WARNING: on x86, 'thread_struct' contains a variable-sized
972 * structure. It *MUST* be at the end of 'task_struct'.
974 * Do not put anything below here!
978 static inline struct pid *task_pid(struct task_struct *task)
980 return task->pids[PIDTYPE_PID].pid;
983 static inline struct pid *task_tgid(struct task_struct *task)
985 return task->group_leader->pids[PIDTYPE_PID].pid;
989 * Without tasklist or rcu lock it is not safe to dereference
990 * the result of task_pgrp/task_session even if task == current,
991 * we can race with another thread doing sys_setsid/sys_setpgid.
993 static inline struct pid *task_pgrp(struct task_struct *task)
995 return task->group_leader->pids[PIDTYPE_PGID].pid;
998 static inline struct pid *task_session(struct task_struct *task)
1000 return task->group_leader->pids[PIDTYPE_SID].pid;
1004 * the helpers to get the task's different pids as they are seen
1005 * from various namespaces
1007 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1008 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1010 * task_xid_nr_ns() : id seen from the ns specified;
1012 * set_task_vxid() : assigns a virtual id to a task;
1014 * see also pid_nr() etc in include/linux/pid.h
1016 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1017 struct pid_namespace *ns);
1019 static inline pid_t task_pid_nr(struct task_struct *tsk)
1024 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1025 struct pid_namespace *ns)
1027 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1030 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1032 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1036 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1041 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1043 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1045 return pid_vnr(task_tgid(tsk));
1049 static inline int pid_alive(const struct task_struct *p);
1050 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1056 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1062 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1064 return task_ppid_nr_ns(tsk, &init_pid_ns);
1067 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1068 struct pid_namespace *ns)
1070 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1073 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1075 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1079 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1080 struct pid_namespace *ns)
1082 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1085 static inline pid_t task_session_vnr(struct task_struct *tsk)
1087 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1090 /* obsolete, do not use */
1091 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1093 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1097 * pid_alive - check that a task structure is not stale
1098 * @p: Task structure to be checked.
1100 * Test if a process is not yet dead (at most zombie state)
1101 * If pid_alive fails, then pointers within the task structure
1102 * can be stale and must not be dereferenced.
1104 * Return: 1 if the process is alive. 0 otherwise.
1106 static inline int pid_alive(const struct task_struct *p)
1108 return p->pids[PIDTYPE_PID].pid != NULL;
1112 * is_global_init - check if a task structure is init. Since init
1113 * is free to have sub-threads we need to check tgid.
1114 * @tsk: Task structure to be checked.
1116 * Check if a task structure is the first user space task the kernel created.
1118 * Return: 1 if the task structure is init. 0 otherwise.
1120 static inline int is_global_init(struct task_struct *tsk)
1122 return task_tgid_nr(tsk) == 1;
1125 extern struct pid *cad_pid;
1130 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1131 #define PF_EXITING 0x00000004 /* getting shut down */
1132 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1133 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1134 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1135 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1136 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1137 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1138 #define PF_DUMPCORE 0x00000200 /* dumped core */
1139 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1140 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1141 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1142 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1143 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1144 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1145 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1146 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1147 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1148 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1149 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1150 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1151 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1152 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1153 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1154 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1155 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1156 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1157 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1160 * Only the _current_ task can read/write to tsk->flags, but other
1161 * tasks can access tsk->flags in readonly mode for example
1162 * with tsk_used_math (like during threaded core dumping).
1163 * There is however an exception to this rule during ptrace
1164 * or during fork: the ptracer task is allowed to write to the
1165 * child->flags of its traced child (same goes for fork, the parent
1166 * can write to the child->flags), because we're guaranteed the
1167 * child is not running and in turn not changing child->flags
1168 * at the same time the parent does it.
1170 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1171 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1172 #define clear_used_math() clear_stopped_child_used_math(current)
1173 #define set_used_math() set_stopped_child_used_math(current)
1174 #define conditional_stopped_child_used_math(condition, child) \
1175 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1176 #define conditional_used_math(condition) \
1177 conditional_stopped_child_used_math(condition, current)
1178 #define copy_to_stopped_child_used_math(child) \
1179 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1180 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1181 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1182 #define used_math() tsk_used_math(current)
1184 /* Per-process atomic flags. */
1185 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1186 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1187 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1188 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1191 #define TASK_PFA_TEST(name, func) \
1192 static inline bool task_##func(struct task_struct *p) \
1193 { return test_bit(PFA_##name, &p->atomic_flags); }
1194 #define TASK_PFA_SET(name, func) \
1195 static inline void task_set_##func(struct task_struct *p) \
1196 { set_bit(PFA_##name, &p->atomic_flags); }
1197 #define TASK_PFA_CLEAR(name, func) \
1198 static inline void task_clear_##func(struct task_struct *p) \
1199 { clear_bit(PFA_##name, &p->atomic_flags); }
1201 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1202 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1204 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1205 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1206 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1208 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1209 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1210 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1212 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1213 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1215 static inline void tsk_restore_flags(struct task_struct *task,
1216 unsigned long orig_flags, unsigned long flags)
1218 task->flags &= ~flags;
1219 task->flags |= orig_flags & flags;
1222 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1223 const struct cpumask *trial);
1224 extern int task_can_attach(struct task_struct *p,
1225 const struct cpumask *cs_cpus_allowed);
1227 extern void do_set_cpus_allowed(struct task_struct *p,
1228 const struct cpumask *new_mask);
1230 extern int set_cpus_allowed_ptr(struct task_struct *p,
1231 const struct cpumask *new_mask);
1233 static inline void do_set_cpus_allowed(struct task_struct *p,
1234 const struct cpumask *new_mask)
1237 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1238 const struct cpumask *new_mask)
1240 if (!cpumask_test_cpu(0, new_mask))
1246 #ifndef cpu_relax_yield
1247 #define cpu_relax_yield() cpu_relax()
1250 extern int yield_to(struct task_struct *p, bool preempt);
1251 extern void set_user_nice(struct task_struct *p, long nice);
1252 extern int task_prio(const struct task_struct *p);
1254 * task_nice - return the nice value of a given task.
1255 * @p: the task in question.
1257 * Return: The nice value [ -20 ... 0 ... 19 ].
1259 static inline int task_nice(const struct task_struct *p)
1261 return PRIO_TO_NICE((p)->static_prio);
1263 extern int can_nice(const struct task_struct *p, const int nice);
1264 extern int task_curr(const struct task_struct *p);
1265 extern int idle_cpu(int cpu);
1266 extern int sched_setscheduler(struct task_struct *, int,
1267 const struct sched_param *);
1268 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1269 const struct sched_param *);
1270 extern int sched_setattr(struct task_struct *,
1271 const struct sched_attr *);
1272 extern struct task_struct *idle_task(int cpu);
1274 * is_idle_task - is the specified task an idle task?
1275 * @p: the task in question.
1277 * Return: 1 if @p is an idle task. 0 otherwise.
1279 static inline bool is_idle_task(const struct task_struct *p)
1281 return !!(p->flags & PF_IDLE);
1283 extern struct task_struct *curr_task(int cpu);
1284 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1288 union thread_union {
1289 #ifndef CONFIG_THREAD_INFO_IN_TASK
1290 struct thread_info thread_info;
1292 unsigned long stack[THREAD_SIZE/sizeof(long)];
1295 #ifdef CONFIG_THREAD_INFO_IN_TASK
1296 static inline struct thread_info *task_thread_info(struct task_struct *task)
1298 return &task->thread_info;
1300 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1301 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1305 * find a task by one of its numerical ids
1307 * find_task_by_pid_ns():
1308 * finds a task by its pid in the specified namespace
1309 * find_task_by_vpid():
1310 * finds a task by its virtual pid
1312 * see also find_vpid() etc in include/linux/pid.h
1315 extern struct task_struct *find_task_by_vpid(pid_t nr);
1316 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1317 struct pid_namespace *ns);
1319 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1320 extern int wake_up_process(struct task_struct *tsk);
1321 extern void wake_up_new_task(struct task_struct *tsk);
1323 extern void kick_process(struct task_struct *tsk);
1325 static inline void kick_process(struct task_struct *tsk) { }
1328 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1329 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1331 __set_task_comm(tsk, from, false);
1333 extern char *get_task_comm(char *to, struct task_struct *tsk);
1336 void scheduler_ipi(void);
1337 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1339 static inline void scheduler_ipi(void) { }
1340 static inline unsigned long wait_task_inactive(struct task_struct *p,
1347 /* set thread flags in other task's structures
1348 * - see asm/thread_info.h for TIF_xxxx flags available
1350 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1352 set_ti_thread_flag(task_thread_info(tsk), flag);
1355 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1357 clear_ti_thread_flag(task_thread_info(tsk), flag);
1360 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1362 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1365 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1367 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1370 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1372 return test_ti_thread_flag(task_thread_info(tsk), flag);
1375 static inline void set_tsk_need_resched(struct task_struct *tsk)
1377 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1380 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1382 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1385 static inline int test_tsk_need_resched(struct task_struct *tsk)
1387 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1391 * cond_resched() and cond_resched_lock(): latency reduction via
1392 * explicit rescheduling in places that are safe. The return
1393 * value indicates whether a reschedule was done in fact.
1394 * cond_resched_lock() will drop the spinlock before scheduling,
1395 * cond_resched_softirq() will enable bhs before scheduling.
1397 #ifndef CONFIG_PREEMPT
1398 extern int _cond_resched(void);
1400 static inline int _cond_resched(void) { return 0; }
1403 #define cond_resched() ({ \
1404 ___might_sleep(__FILE__, __LINE__, 0); \
1408 extern int __cond_resched_lock(spinlock_t *lock);
1410 #define cond_resched_lock(lock) ({ \
1411 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1412 __cond_resched_lock(lock); \
1415 extern int __cond_resched_softirq(void);
1417 #define cond_resched_softirq() ({ \
1418 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1419 __cond_resched_softirq(); \
1422 static inline void cond_resched_rcu(void)
1424 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1432 * Does a critical section need to be broken due to another
1433 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1434 * but a general need for low latency)
1436 static inline int spin_needbreak(spinlock_t *lock)
1438 #ifdef CONFIG_PREEMPT
1439 return spin_is_contended(lock);
1445 static __always_inline bool need_resched(void)
1447 return unlikely(tif_need_resched());
1451 * Wrappers for p->thread_info->cpu access. No-op on UP.
1455 static inline unsigned int task_cpu(const struct task_struct *p)
1457 #ifdef CONFIG_THREAD_INFO_IN_TASK
1460 return task_thread_info(p)->cpu;
1464 static inline int task_node(const struct task_struct *p)
1466 return cpu_to_node(task_cpu(p));
1469 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1473 static inline unsigned int task_cpu(const struct task_struct *p)
1478 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1482 #endif /* CONFIG_SMP */
1485 * In order to reduce various lock holder preemption latencies provide an
1486 * interface to see if a vCPU is currently running or not.
1488 * This allows us to terminate optimistic spin loops and block, analogous to
1489 * the native optimistic spin heuristic of testing if the lock owner task is
1492 #ifndef vcpu_is_preempted
1493 # define vcpu_is_preempted(cpu) false
1496 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1497 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1499 #ifndef TASK_SIZE_OF
1500 #define TASK_SIZE_OF(tsk) TASK_SIZE