5 * Define 'struct task_struct' and provide the main scheduler
6 * APIs (schedule(), wakeup variants, etc.)
9 #include <uapi/linux/sched.h>
11 #include <asm/current.h>
13 #include <linux/pid.h>
14 #include <linux/sem.h>
15 #include <linux/shm.h>
16 #include <linux/kcov.h>
17 #include <linux/mutex.h>
18 #include <linux/plist.h>
19 #include <linux/hrtimer.h>
20 #include <linux/seccomp.h>
21 #include <linux/nodemask.h>
22 #include <linux/rcupdate.h>
23 #include <linux/resource.h>
24 #include <linux/latencytop.h>
25 #include <linux/sched/prio.h>
26 #include <linux/signal_types.h>
27 #include <linux/mm_types_task.h>
28 #include <linux/task_io_accounting.h>
30 /* task_struct member predeclarations (sorted alphabetically): */
32 struct backing_dev_info;
37 struct futex_pi_state;
42 struct perf_event_context;
44 struct pipe_inode_info;
47 struct robust_list_head;
51 struct sighand_struct;
53 struct task_delay_info;
57 * Task state bitmask. NOTE! These bits are also
58 * encoded in fs/proc/array.c: get_task_state().
60 * We have two separate sets of flags: task->state
61 * is about runnability, while task->exit_state are
62 * about the task exiting. Confusing, but this way
63 * modifying one set can't modify the other one by
67 /* Used in tsk->state: */
68 #define TASK_RUNNING 0
69 #define TASK_INTERRUPTIBLE 1
70 #define TASK_UNINTERRUPTIBLE 2
71 #define __TASK_STOPPED 4
72 #define __TASK_TRACED 8
73 /* Used in tsk->exit_state: */
75 #define EXIT_ZOMBIE 32
76 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
77 /* Used in tsk->state again: */
79 #define TASK_WAKEKILL 128
80 #define TASK_WAKING 256
81 #define TASK_PARKED 512
82 #define TASK_NOLOAD 1024
84 #define TASK_STATE_MAX 4096
86 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
88 /* Convenience macros for the sake of set_current_state: */
89 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
90 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
91 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
93 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
95 /* Convenience macros for the sake of wake_up(): */
96 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
97 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
99 /* get_task_state(): */
100 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
101 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
102 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
104 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
108 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
110 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 (task->flags & PF_FROZEN) == 0 && \
112 (task->state & TASK_NOLOAD) == 0)
114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
116 #define __set_current_state(state_value) \
118 current->task_state_change = _THIS_IP_; \
119 current->state = (state_value); \
121 #define set_current_state(state_value) \
123 current->task_state_change = _THIS_IP_; \
124 smp_store_mb(current->state, (state_value)); \
129 * set_current_state() includes a barrier so that the write of current->state
130 * is correctly serialised wrt the caller's subsequent test of whether to
134 * set_current_state(TASK_UNINTERRUPTIBLE);
140 * __set_current_state(TASK_RUNNING);
142 * If the caller does not need such serialisation (because, for instance, the
143 * condition test and condition change and wakeup are under the same lock) then
144 * use __set_current_state().
146 * The above is typically ordered against the wakeup, which does:
148 * need_sleep = false;
149 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
151 * Where wake_up_state() (and all other wakeup primitives) imply enough
152 * barriers to order the store of the variable against wakeup.
154 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
155 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
156 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
158 * This is obviously fine, since they both store the exact same value.
160 * Also see the comments of try_to_wake_up().
162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0)
163 #define set_current_state(state_value) smp_store_mb(current->state, (state_value))
166 /* Task command name length: */
167 #define TASK_COMM_LEN 16
169 extern cpumask_var_t cpu_isolated_map;
171 extern void scheduler_tick(void);
173 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
175 extern long schedule_timeout(long timeout);
176 extern long schedule_timeout_interruptible(long timeout);
177 extern long schedule_timeout_killable(long timeout);
178 extern long schedule_timeout_uninterruptible(long timeout);
179 extern long schedule_timeout_idle(long timeout);
180 asmlinkage void schedule(void);
181 extern void schedule_preempt_disabled(void);
183 extern int __must_check io_schedule_prepare(void);
184 extern void io_schedule_finish(int token);
185 extern long io_schedule_timeout(long timeout);
186 extern void io_schedule(void);
189 * struct prev_cputime - snapshot of system and user cputime
190 * @utime: time spent in user mode
191 * @stime: time spent in system mode
192 * @lock: protects the above two fields
194 * Stores previous user/system time values such that we can guarantee
197 struct prev_cputime {
198 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
206 * struct task_cputime - collected CPU time counts
207 * @utime: time spent in user mode, in nanoseconds
208 * @stime: time spent in kernel mode, in nanoseconds
209 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
211 * This structure groups together three kinds of CPU time that are tracked for
212 * threads and thread groups. Most things considering CPU time want to group
213 * these counts together and treat all three of them in parallel.
215 struct task_cputime {
218 unsigned long long sum_exec_runtime;
221 /* Alternate field names when used on cache expirations: */
222 #define virt_exp utime
223 #define prof_exp stime
224 #define sched_exp sum_exec_runtime
227 /* Task is sleeping or running in a CPU with VTIME inactive: */
229 /* Task runs in userspace in a CPU with VTIME active: */
231 /* Task runs in kernelspace in a CPU with VTIME active: */
237 unsigned long long starttime;
238 enum vtime_state state;
245 #ifdef CONFIG_SCHED_INFO
246 /* Cumulative counters: */
248 /* # of times we have run on this CPU: */
249 unsigned long pcount;
251 /* Time spent waiting on a runqueue: */
252 unsigned long long run_delay;
256 /* When did we last run on a CPU? */
257 unsigned long long last_arrival;
259 /* When were we last queued to run? */
260 unsigned long long last_queued;
262 #endif /* CONFIG_SCHED_INFO */
266 * Integer metrics need fixed point arithmetic, e.g., sched/fair
267 * has a few: load, load_avg, util_avg, freq, and capacity.
269 * We define a basic fixed point arithmetic range, and then formalize
270 * all these metrics based on that basic range.
272 # define SCHED_FIXEDPOINT_SHIFT 10
273 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
276 unsigned long weight;
281 * The load_avg/util_avg accumulates an infinite geometric series
282 * (see __update_load_avg() in kernel/sched/fair.c).
284 * [load_avg definition]
286 * load_avg = runnable% * scale_load_down(load)
288 * where runnable% is the time ratio that a sched_entity is runnable.
289 * For cfs_rq, it is the aggregated load_avg of all runnable and
290 * blocked sched_entities.
292 * load_avg may also take frequency scaling into account:
294 * load_avg = runnable% * scale_load_down(load) * freq%
296 * where freq% is the CPU frequency normalized to the highest frequency.
298 * [util_avg definition]
300 * util_avg = running% * SCHED_CAPACITY_SCALE
302 * where running% is the time ratio that a sched_entity is running on
303 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
304 * and blocked sched_entities.
306 * util_avg may also factor frequency scaling and CPU capacity scaling:
308 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
310 * where freq% is the same as above, and capacity% is the CPU capacity
311 * normalized to the greatest capacity (due to uarch differences, etc).
313 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
314 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
315 * we therefore scale them to as large a range as necessary. This is for
316 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
320 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
321 * with the highest load (=88761), always runnable on a single cfs_rq,
322 * and should not overflow as the number already hits PID_MAX_LIMIT.
324 * For all other cases (including 32-bit kernels), struct load_weight's
325 * weight will overflow first before we do, because:
327 * Max(load_avg) <= Max(load.weight)
329 * Then it is the load_weight's responsibility to consider overflow
333 u64 last_update_time;
337 unsigned long load_avg;
338 unsigned long util_avg;
341 struct sched_statistics {
342 #ifdef CONFIG_SCHEDSTATS
352 s64 sum_sleep_runtime;
359 u64 nr_migrations_cold;
360 u64 nr_failed_migrations_affine;
361 u64 nr_failed_migrations_running;
362 u64 nr_failed_migrations_hot;
363 u64 nr_forced_migrations;
367 u64 nr_wakeups_migrate;
368 u64 nr_wakeups_local;
369 u64 nr_wakeups_remote;
370 u64 nr_wakeups_affine;
371 u64 nr_wakeups_affine_attempts;
372 u64 nr_wakeups_passive;
377 struct sched_entity {
378 /* For load-balancing: */
379 struct load_weight load;
380 struct rb_node run_node;
381 struct list_head group_node;
385 u64 sum_exec_runtime;
387 u64 prev_sum_exec_runtime;
391 struct sched_statistics statistics;
393 #ifdef CONFIG_FAIR_GROUP_SCHED
395 struct sched_entity *parent;
396 /* rq on which this entity is (to be) queued: */
397 struct cfs_rq *cfs_rq;
398 /* rq "owned" by this entity/group: */
404 * Per entity load average tracking.
406 * Put into separate cache line so it does not
407 * collide with read-mostly values above.
409 struct sched_avg avg ____cacheline_aligned_in_smp;
413 struct sched_rt_entity {
414 struct list_head run_list;
415 unsigned long timeout;
416 unsigned long watchdog_stamp;
417 unsigned int time_slice;
418 unsigned short on_rq;
419 unsigned short on_list;
421 struct sched_rt_entity *back;
422 #ifdef CONFIG_RT_GROUP_SCHED
423 struct sched_rt_entity *parent;
424 /* rq on which this entity is (to be) queued: */
426 /* rq "owned" by this entity/group: */
431 struct sched_dl_entity {
432 struct rb_node rb_node;
435 * Original scheduling parameters. Copied here from sched_attr
436 * during sched_setattr(), they will remain the same until
437 * the next sched_setattr().
439 u64 dl_runtime; /* Maximum runtime for each instance */
440 u64 dl_deadline; /* Relative deadline of each instance */
441 u64 dl_period; /* Separation of two instances (period) */
442 u64 dl_bw; /* dl_runtime / dl_period */
443 u64 dl_density; /* dl_runtime / dl_deadline */
446 * Actual scheduling parameters. Initialized with the values above,
447 * they are continously updated during task execution. Note that
448 * the remaining runtime could be < 0 in case we are in overrun.
450 s64 runtime; /* Remaining runtime for this instance */
451 u64 deadline; /* Absolute deadline for this instance */
452 unsigned int flags; /* Specifying the scheduler behaviour */
457 * @dl_throttled tells if we exhausted the runtime. If so, the
458 * task has to wait for a replenishment to be performed at the
459 * next firing of dl_timer.
461 * @dl_boosted tells if we are boosted due to DI. If so we are
462 * outside bandwidth enforcement mechanism (but only until we
463 * exit the critical section);
465 * @dl_yielded tells if task gave up the CPU before consuming
466 * all its available runtime during the last job.
468 * @dl_non_contending tells if the task is inactive while still
469 * contributing to the active utilization. In other words, it
470 * indicates if the inactive timer has been armed and its handler
471 * has not been executed yet. This flag is useful to avoid race
472 * conditions between the inactive timer handler and the wakeup
478 int dl_non_contending;
481 * Bandwidth enforcement timer. Each -deadline task has its
482 * own bandwidth to be enforced, thus we need one timer per task.
484 struct hrtimer dl_timer;
487 * Inactive timer, responsible for decreasing the active utilization
488 * at the "0-lag time". When a -deadline task blocks, it contributes
489 * to GRUB's active utilization until the "0-lag time", hence a
490 * timer is needed to decrease the active utilization at the correct
493 struct hrtimer inactive_timer;
502 /* Otherwise the compiler can store garbage here: */
505 u32 s; /* Set of bits. */
508 enum perf_event_task_context {
509 perf_invalid_context = -1,
512 perf_nr_task_contexts,
516 struct wake_q_node *next;
520 #ifdef CONFIG_THREAD_INFO_IN_TASK
522 * For reasons of header soup (see current_thread_info()), this
523 * must be the first element of task_struct.
525 struct thread_info thread_info;
527 /* -1 unrunnable, 0 runnable, >0 stopped: */
531 /* Per task flags (PF_*), defined further below: */
536 struct llist_node wake_entry;
538 #ifdef CONFIG_THREAD_INFO_IN_TASK
542 unsigned int wakee_flips;
543 unsigned long wakee_flip_decay_ts;
544 struct task_struct *last_wakee;
553 unsigned int rt_priority;
555 const struct sched_class *sched_class;
556 struct sched_entity se;
557 struct sched_rt_entity rt;
558 #ifdef CONFIG_CGROUP_SCHED
559 struct task_group *sched_task_group;
561 struct sched_dl_entity dl;
563 #ifdef CONFIG_PREEMPT_NOTIFIERS
564 /* List of struct preempt_notifier: */
565 struct hlist_head preempt_notifiers;
568 #ifdef CONFIG_BLK_DEV_IO_TRACE
569 unsigned int btrace_seq;
574 cpumask_t cpus_allowed;
576 #ifdef CONFIG_PREEMPT_RCU
577 int rcu_read_lock_nesting;
578 union rcu_special rcu_read_unlock_special;
579 struct list_head rcu_node_entry;
580 struct rcu_node *rcu_blocked_node;
581 #endif /* #ifdef CONFIG_PREEMPT_RCU */
583 #ifdef CONFIG_TASKS_RCU
584 unsigned long rcu_tasks_nvcsw;
585 bool rcu_tasks_holdout;
586 struct list_head rcu_tasks_holdout_list;
587 int rcu_tasks_idle_cpu;
588 #endif /* #ifdef CONFIG_TASKS_RCU */
590 struct sched_info sched_info;
592 struct list_head tasks;
594 struct plist_node pushable_tasks;
595 struct rb_node pushable_dl_tasks;
598 struct mm_struct *mm;
599 struct mm_struct *active_mm;
601 /* Per-thread vma caching: */
602 struct vmacache vmacache;
604 #ifdef SPLIT_RSS_COUNTING
605 struct task_rss_stat rss_stat;
610 /* The signal sent when the parent dies: */
612 /* JOBCTL_*, siglock protected: */
613 unsigned long jobctl;
615 /* Used for emulating ABI behavior of previous Linux versions: */
616 unsigned int personality;
618 /* Scheduler bits, serialized by scheduler locks: */
619 unsigned sched_reset_on_fork:1;
620 unsigned sched_contributes_to_load:1;
621 unsigned sched_migrated:1;
622 unsigned sched_remote_wakeup:1;
623 /* Force alignment to the next boundary: */
626 /* Unserialized, strictly 'current' */
628 /* Bit to tell LSMs we're in execve(): */
629 unsigned in_execve:1;
630 unsigned in_iowait:1;
631 #ifndef TIF_RESTORE_SIGMASK
632 unsigned restore_sigmask:1;
635 unsigned memcg_may_oom:1;
637 unsigned memcg_kmem_skip_account:1;
640 #ifdef CONFIG_COMPAT_BRK
641 unsigned brk_randomized:1;
643 #ifdef CONFIG_CGROUPS
644 /* disallow userland-initiated cgroup migration */
645 unsigned no_cgroup_migration:1;
648 unsigned long atomic_flags; /* Flags requiring atomic access. */
650 struct restart_block restart_block;
655 #ifdef CONFIG_CC_STACKPROTECTOR
656 /* Canary value for the -fstack-protector GCC feature: */
657 unsigned long stack_canary;
660 * Pointers to the (original) parent process, youngest child, younger sibling,
661 * older sibling, respectively. (p->father can be replaced with
662 * p->real_parent->pid)
665 /* Real parent process: */
666 struct task_struct __rcu *real_parent;
668 /* Recipient of SIGCHLD, wait4() reports: */
669 struct task_struct __rcu *parent;
672 * Children/sibling form the list of natural children:
674 struct list_head children;
675 struct list_head sibling;
676 struct task_struct *group_leader;
679 * 'ptraced' is the list of tasks this task is using ptrace() on.
681 * This includes both natural children and PTRACE_ATTACH targets.
682 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
684 struct list_head ptraced;
685 struct list_head ptrace_entry;
687 /* PID/PID hash table linkage. */
688 struct pid_link pids[PIDTYPE_MAX];
689 struct list_head thread_group;
690 struct list_head thread_node;
692 struct completion *vfork_done;
694 /* CLONE_CHILD_SETTID: */
695 int __user *set_child_tid;
697 /* CLONE_CHILD_CLEARTID: */
698 int __user *clear_child_tid;
702 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
707 struct prev_cputime prev_cputime;
708 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
712 #ifdef CONFIG_NO_HZ_FULL
713 atomic_t tick_dep_mask;
715 /* Context switch counts: */
717 unsigned long nivcsw;
719 /* Monotonic time in nsecs: */
722 /* Boot based time in nsecs: */
725 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
726 unsigned long min_flt;
727 unsigned long maj_flt;
729 #ifdef CONFIG_POSIX_TIMERS
730 struct task_cputime cputime_expires;
731 struct list_head cpu_timers[3];
734 /* Process credentials: */
736 /* Tracer's credentials at attach: */
737 const struct cred __rcu *ptracer_cred;
739 /* Objective and real subjective task credentials (COW): */
740 const struct cred __rcu *real_cred;
742 /* Effective (overridable) subjective task credentials (COW): */
743 const struct cred __rcu *cred;
746 * executable name, excluding path.
748 * - normally initialized setup_new_exec()
749 * - access it with [gs]et_task_comm()
750 * - lock it with task_lock()
752 char comm[TASK_COMM_LEN];
754 struct nameidata *nameidata;
756 #ifdef CONFIG_SYSVIPC
757 struct sysv_sem sysvsem;
758 struct sysv_shm sysvshm;
760 #ifdef CONFIG_DETECT_HUNG_TASK
761 unsigned long last_switch_count;
763 /* Filesystem information: */
764 struct fs_struct *fs;
766 /* Open file information: */
767 struct files_struct *files;
770 struct nsproxy *nsproxy;
772 /* Signal handlers: */
773 struct signal_struct *signal;
774 struct sighand_struct *sighand;
776 sigset_t real_blocked;
777 /* Restored if set_restore_sigmask() was used: */
778 sigset_t saved_sigmask;
779 struct sigpending pending;
780 unsigned long sas_ss_sp;
782 unsigned int sas_ss_flags;
784 struct callback_head *task_works;
786 struct audit_context *audit_context;
787 #ifdef CONFIG_AUDITSYSCALL
789 unsigned int sessionid;
791 struct seccomp seccomp;
793 /* Thread group tracking: */
797 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
798 spinlock_t alloc_lock;
800 /* Protection of the PI data structures: */
801 raw_spinlock_t pi_lock;
803 struct wake_q_node wake_q;
805 #ifdef CONFIG_RT_MUTEXES
806 /* PI waiters blocked on a rt_mutex held by this task: */
807 struct rb_root pi_waiters;
808 struct rb_node *pi_waiters_leftmost;
809 /* Updated under owner's pi_lock and rq lock */
810 struct task_struct *pi_top_task;
811 /* Deadlock detection and priority inheritance handling: */
812 struct rt_mutex_waiter *pi_blocked_on;
815 #ifdef CONFIG_DEBUG_MUTEXES
816 /* Mutex deadlock detection: */
817 struct mutex_waiter *blocked_on;
820 #ifdef CONFIG_TRACE_IRQFLAGS
821 unsigned int irq_events;
822 unsigned long hardirq_enable_ip;
823 unsigned long hardirq_disable_ip;
824 unsigned int hardirq_enable_event;
825 unsigned int hardirq_disable_event;
826 int hardirqs_enabled;
828 unsigned long softirq_disable_ip;
829 unsigned long softirq_enable_ip;
830 unsigned int softirq_disable_event;
831 unsigned int softirq_enable_event;
832 int softirqs_enabled;
836 #ifdef CONFIG_LOCKDEP
837 # define MAX_LOCK_DEPTH 48UL
840 unsigned int lockdep_recursion;
841 struct held_lock held_locks[MAX_LOCK_DEPTH];
842 gfp_t lockdep_reclaim_gfp;
846 unsigned int in_ubsan;
849 /* Journalling filesystem info: */
852 /* Stacked block device info: */
853 struct bio_list *bio_list;
856 /* Stack plugging: */
857 struct blk_plug *plug;
861 struct reclaim_state *reclaim_state;
863 struct backing_dev_info *backing_dev_info;
865 struct io_context *io_context;
868 unsigned long ptrace_message;
869 siginfo_t *last_siginfo;
871 struct task_io_accounting ioac;
872 #ifdef CONFIG_TASK_XACCT
873 /* Accumulated RSS usage: */
875 /* Accumulated virtual memory usage: */
877 /* stime + utime since last update: */
880 #ifdef CONFIG_CPUSETS
881 /* Protected by ->alloc_lock: */
882 nodemask_t mems_allowed;
883 /* Seqence number to catch updates: */
884 seqcount_t mems_allowed_seq;
885 int cpuset_mem_spread_rotor;
886 int cpuset_slab_spread_rotor;
888 #ifdef CONFIG_CGROUPS
889 /* Control Group info protected by css_set_lock: */
890 struct css_set __rcu *cgroups;
891 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
892 struct list_head cg_list;
894 #ifdef CONFIG_INTEL_RDT_A
898 struct robust_list_head __user *robust_list;
900 struct compat_robust_list_head __user *compat_robust_list;
902 struct list_head pi_state_list;
903 struct futex_pi_state *pi_state_cache;
905 #ifdef CONFIG_PERF_EVENTS
906 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
907 struct mutex perf_event_mutex;
908 struct list_head perf_event_list;
910 #ifdef CONFIG_DEBUG_PREEMPT
911 unsigned long preempt_disable_ip;
914 /* Protected by alloc_lock: */
915 struct mempolicy *mempolicy;
917 short pref_node_fork;
919 #ifdef CONFIG_NUMA_BALANCING
921 unsigned int numa_scan_period;
922 unsigned int numa_scan_period_max;
923 int numa_preferred_nid;
924 unsigned long numa_migrate_retry;
925 /* Migration stamp: */
927 u64 last_task_numa_placement;
928 u64 last_sum_exec_runtime;
929 struct callback_head numa_work;
931 struct list_head numa_entry;
932 struct numa_group *numa_group;
935 * numa_faults is an array split into four regions:
936 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
937 * in this precise order.
939 * faults_memory: Exponential decaying average of faults on a per-node
940 * basis. Scheduling placement decisions are made based on these
941 * counts. The values remain static for the duration of a PTE scan.
942 * faults_cpu: Track the nodes the process was running on when a NUMA
943 * hinting fault was incurred.
944 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
945 * during the current scan window. When the scan completes, the counts
946 * in faults_memory and faults_cpu decay and these values are copied.
948 unsigned long *numa_faults;
949 unsigned long total_numa_faults;
952 * numa_faults_locality tracks if faults recorded during the last
953 * scan window were remote/local or failed to migrate. The task scan
954 * period is adapted based on the locality of the faults with different
955 * weights depending on whether they were shared or private faults
957 unsigned long numa_faults_locality[3];
959 unsigned long numa_pages_migrated;
960 #endif /* CONFIG_NUMA_BALANCING */
962 struct tlbflush_unmap_batch tlb_ubc;
966 /* Cache last used pipe for splice(): */
967 struct pipe_inode_info *splice_pipe;
969 struct page_frag task_frag;
971 #ifdef CONFIG_TASK_DELAY_ACCT
972 struct task_delay_info *delays;
975 #ifdef CONFIG_FAULT_INJECTION
979 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
980 * balance_dirty_pages() for a dirty throttling pause:
983 int nr_dirtied_pause;
984 /* Start of a write-and-pause period: */
985 unsigned long dirty_paused_when;
987 #ifdef CONFIG_LATENCYTOP
988 int latency_record_count;
989 struct latency_record latency_record[LT_SAVECOUNT];
992 * Time slack values; these are used to round up poll() and
993 * select() etc timeout values. These are in nanoseconds.
996 u64 default_timer_slack_ns;
999 unsigned int kasan_depth;
1002 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1003 /* Index of current stored address in ret_stack: */
1006 /* Stack of return addresses for return function tracing: */
1007 struct ftrace_ret_stack *ret_stack;
1009 /* Timestamp for last schedule: */
1010 unsigned long long ftrace_timestamp;
1013 * Number of functions that haven't been traced
1014 * because of depth overrun:
1016 atomic_t trace_overrun;
1018 /* Pause tracing: */
1019 atomic_t tracing_graph_pause;
1022 #ifdef CONFIG_TRACING
1023 /* State flags for use by tracers: */
1024 unsigned long trace;
1026 /* Bitmask and counter of trace recursion: */
1027 unsigned long trace_recursion;
1028 #endif /* CONFIG_TRACING */
1031 /* Coverage collection mode enabled for this task (0 if disabled): */
1032 enum kcov_mode kcov_mode;
1034 /* Size of the kcov_area: */
1035 unsigned int kcov_size;
1037 /* Buffer for coverage collection: */
1040 /* KCOV descriptor wired with this task or NULL: */
1045 struct mem_cgroup *memcg_in_oom;
1046 gfp_t memcg_oom_gfp_mask;
1047 int memcg_oom_order;
1049 /* Number of pages to reclaim on returning to userland: */
1050 unsigned int memcg_nr_pages_over_high;
1053 #ifdef CONFIG_UPROBES
1054 struct uprobe_task *utask;
1056 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1057 unsigned int sequential_io;
1058 unsigned int sequential_io_avg;
1060 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1061 unsigned long task_state_change;
1063 int pagefault_disabled;
1065 struct task_struct *oom_reaper_list;
1067 #ifdef CONFIG_VMAP_STACK
1068 struct vm_struct *stack_vm_area;
1070 #ifdef CONFIG_THREAD_INFO_IN_TASK
1071 /* A live task holds one reference: */
1072 atomic_t stack_refcount;
1074 #ifdef CONFIG_LIVEPATCH
1077 #ifdef CONFIG_SECURITY
1078 /* Used by LSM modules for access restriction: */
1081 /* CPU-specific state of this task: */
1082 struct thread_struct thread;
1085 * WARNING: on x86, 'thread_struct' contains a variable-sized
1086 * structure. It *MUST* be at the end of 'task_struct'.
1088 * Do not put anything below here!
1092 static inline struct pid *task_pid(struct task_struct *task)
1094 return task->pids[PIDTYPE_PID].pid;
1097 static inline struct pid *task_tgid(struct task_struct *task)
1099 return task->group_leader->pids[PIDTYPE_PID].pid;
1103 * Without tasklist or RCU lock it is not safe to dereference
1104 * the result of task_pgrp/task_session even if task == current,
1105 * we can race with another thread doing sys_setsid/sys_setpgid.
1107 static inline struct pid *task_pgrp(struct task_struct *task)
1109 return task->group_leader->pids[PIDTYPE_PGID].pid;
1112 static inline struct pid *task_session(struct task_struct *task)
1114 return task->group_leader->pids[PIDTYPE_SID].pid;
1118 * the helpers to get the task's different pids as they are seen
1119 * from various namespaces
1121 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1122 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1124 * task_xid_nr_ns() : id seen from the ns specified;
1126 * see also pid_nr() etc in include/linux/pid.h
1128 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1130 static inline pid_t task_pid_nr(struct task_struct *tsk)
1135 static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1137 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1140 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1142 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1146 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1151 extern pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1153 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1155 return pid_vnr(task_tgid(tsk));
1159 * pid_alive - check that a task structure is not stale
1160 * @p: Task structure to be checked.
1162 * Test if a process is not yet dead (at most zombie state)
1163 * If pid_alive fails, then pointers within the task structure
1164 * can be stale and must not be dereferenced.
1166 * Return: 1 if the process is alive. 0 otherwise.
1168 static inline int pid_alive(const struct task_struct *p)
1170 return p->pids[PIDTYPE_PID].pid != NULL;
1173 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1179 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1185 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1187 return task_ppid_nr_ns(tsk, &init_pid_ns);
1190 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1192 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1195 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1197 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1201 static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1203 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1206 static inline pid_t task_session_vnr(struct task_struct *tsk)
1208 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1211 /* Obsolete, do not use: */
1212 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1214 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1218 * is_global_init - check if a task structure is init. Since init
1219 * is free to have sub-threads we need to check tgid.
1220 * @tsk: Task structure to be checked.
1222 * Check if a task structure is the first user space task the kernel created.
1224 * Return: 1 if the task structure is init. 0 otherwise.
1226 static inline int is_global_init(struct task_struct *tsk)
1228 return task_tgid_nr(tsk) == 1;
1231 extern struct pid *cad_pid;
1236 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1237 #define PF_EXITING 0x00000004 /* Getting shut down */
1238 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1239 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1240 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1241 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1242 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1243 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1244 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1245 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1246 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1247 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1248 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1249 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1250 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1251 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1252 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1253 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1254 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1255 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1256 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1257 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1258 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1259 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1260 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1261 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1262 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1263 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1266 * Only the _current_ task can read/write to tsk->flags, but other
1267 * tasks can access tsk->flags in readonly mode for example
1268 * with tsk_used_math (like during threaded core dumping).
1269 * There is however an exception to this rule during ptrace
1270 * or during fork: the ptracer task is allowed to write to the
1271 * child->flags of its traced child (same goes for fork, the parent
1272 * can write to the child->flags), because we're guaranteed the
1273 * child is not running and in turn not changing child->flags
1274 * at the same time the parent does it.
1276 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1277 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1278 #define clear_used_math() clear_stopped_child_used_math(current)
1279 #define set_used_math() set_stopped_child_used_math(current)
1281 #define conditional_stopped_child_used_math(condition, child) \
1282 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1284 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1286 #define copy_to_stopped_child_used_math(child) \
1287 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1289 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1290 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1291 #define used_math() tsk_used_math(current)
1293 static inline bool is_percpu_thread(void)
1296 return (current->flags & PF_NO_SETAFFINITY) &&
1297 (current->nr_cpus_allowed == 1);
1303 /* Per-process atomic flags. */
1304 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1305 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1306 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1309 #define TASK_PFA_TEST(name, func) \
1310 static inline bool task_##func(struct task_struct *p) \
1311 { return test_bit(PFA_##name, &p->atomic_flags); }
1313 #define TASK_PFA_SET(name, func) \
1314 static inline void task_set_##func(struct task_struct *p) \
1315 { set_bit(PFA_##name, &p->atomic_flags); }
1317 #define TASK_PFA_CLEAR(name, func) \
1318 static inline void task_clear_##func(struct task_struct *p) \
1319 { clear_bit(PFA_##name, &p->atomic_flags); }
1321 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1322 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1324 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1325 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1326 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1328 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1329 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1330 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1333 current_restore_flags(unsigned long orig_flags, unsigned long flags)
1335 current->flags &= ~flags;
1336 current->flags |= orig_flags & flags;
1339 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1340 extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1342 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1343 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1345 static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1348 static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1350 if (!cpumask_test_cpu(0, new_mask))
1356 #ifndef cpu_relax_yield
1357 #define cpu_relax_yield() cpu_relax()
1360 extern int yield_to(struct task_struct *p, bool preempt);
1361 extern void set_user_nice(struct task_struct *p, long nice);
1362 extern int task_prio(const struct task_struct *p);
1365 * task_nice - return the nice value of a given task.
1366 * @p: the task in question.
1368 * Return: The nice value [ -20 ... 0 ... 19 ].
1370 static inline int task_nice(const struct task_struct *p)
1372 return PRIO_TO_NICE((p)->static_prio);
1375 extern int can_nice(const struct task_struct *p, const int nice);
1376 extern int task_curr(const struct task_struct *p);
1377 extern int idle_cpu(int cpu);
1378 extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1379 extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1380 extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1381 extern struct task_struct *idle_task(int cpu);
1384 * is_idle_task - is the specified task an idle task?
1385 * @p: the task in question.
1387 * Return: 1 if @p is an idle task. 0 otherwise.
1389 static inline bool is_idle_task(const struct task_struct *p)
1391 return !!(p->flags & PF_IDLE);
1394 extern struct task_struct *curr_task(int cpu);
1395 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1399 union thread_union {
1400 #ifndef CONFIG_THREAD_INFO_IN_TASK
1401 struct thread_info thread_info;
1403 unsigned long stack[THREAD_SIZE/sizeof(long)];
1406 #ifdef CONFIG_THREAD_INFO_IN_TASK
1407 static inline struct thread_info *task_thread_info(struct task_struct *task)
1409 return &task->thread_info;
1411 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1412 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1416 * find a task by one of its numerical ids
1418 * find_task_by_pid_ns():
1419 * finds a task by its pid in the specified namespace
1420 * find_task_by_vpid():
1421 * finds a task by its virtual pid
1423 * see also find_vpid() etc in include/linux/pid.h
1426 extern struct task_struct *find_task_by_vpid(pid_t nr);
1427 extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1429 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1430 extern int wake_up_process(struct task_struct *tsk);
1431 extern void wake_up_new_task(struct task_struct *tsk);
1434 extern void kick_process(struct task_struct *tsk);
1436 static inline void kick_process(struct task_struct *tsk) { }
1439 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1441 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1443 __set_task_comm(tsk, from, false);
1446 extern char *get_task_comm(char *to, struct task_struct *tsk);
1449 void scheduler_ipi(void);
1450 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1452 static inline void scheduler_ipi(void) { }
1453 static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1460 * Set thread flags in other task's structures.
1461 * See asm/thread_info.h for TIF_xxxx flags available:
1463 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1465 set_ti_thread_flag(task_thread_info(tsk), flag);
1468 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1470 clear_ti_thread_flag(task_thread_info(tsk), flag);
1473 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1475 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1478 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1480 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1483 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1485 return test_ti_thread_flag(task_thread_info(tsk), flag);
1488 static inline void set_tsk_need_resched(struct task_struct *tsk)
1490 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1493 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1495 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1498 static inline int test_tsk_need_resched(struct task_struct *tsk)
1500 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1504 * cond_resched() and cond_resched_lock(): latency reduction via
1505 * explicit rescheduling in places that are safe. The return
1506 * value indicates whether a reschedule was done in fact.
1507 * cond_resched_lock() will drop the spinlock before scheduling,
1508 * cond_resched_softirq() will enable bhs before scheduling.
1510 #ifndef CONFIG_PREEMPT
1511 extern int _cond_resched(void);
1513 static inline int _cond_resched(void) { return 0; }
1516 #define cond_resched() ({ \
1517 ___might_sleep(__FILE__, __LINE__, 0); \
1521 extern int __cond_resched_lock(spinlock_t *lock);
1523 #define cond_resched_lock(lock) ({ \
1524 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1525 __cond_resched_lock(lock); \
1528 extern int __cond_resched_softirq(void);
1530 #define cond_resched_softirq() ({ \
1531 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1532 __cond_resched_softirq(); \
1535 static inline void cond_resched_rcu(void)
1537 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1545 * Does a critical section need to be broken due to another
1546 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1547 * but a general need for low latency)
1549 static inline int spin_needbreak(spinlock_t *lock)
1551 #ifdef CONFIG_PREEMPT
1552 return spin_is_contended(lock);
1558 static __always_inline bool need_resched(void)
1560 return unlikely(tif_need_resched());
1564 * Wrappers for p->thread_info->cpu access. No-op on UP.
1568 static inline unsigned int task_cpu(const struct task_struct *p)
1570 #ifdef CONFIG_THREAD_INFO_IN_TASK
1573 return task_thread_info(p)->cpu;
1577 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1581 static inline unsigned int task_cpu(const struct task_struct *p)
1586 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1590 #endif /* CONFIG_SMP */
1593 * In order to reduce various lock holder preemption latencies provide an
1594 * interface to see if a vCPU is currently running or not.
1596 * This allows us to terminate optimistic spin loops and block, analogous to
1597 * the native optimistic spin heuristic of testing if the lock owner task is
1600 #ifndef vcpu_is_preempted
1601 # define vcpu_is_preempted(cpu) false
1604 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1605 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1607 #ifndef TASK_SIZE_OF
1608 #define TASK_SIZE_OF(tsk) TASK_SIZE