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 #ifdef CONFIG_SCHED_INFO
228 /* Cumulative counters: */
230 /* # of times we have run on this CPU: */
231 unsigned long pcount
;
233 /* Time spent waiting on a runqueue: */
234 unsigned long long run_delay
;
238 /* When did we last run on a CPU? */
239 unsigned long long last_arrival
;
241 /* When were we last queued to run? */
242 unsigned long long last_queued
;
244 #endif /* CONFIG_SCHED_INFO */
248 * Integer metrics need fixed point arithmetic, e.g., sched/fair
249 * has a few: load, load_avg, util_avg, freq, and capacity.
251 * We define a basic fixed point arithmetic range, and then formalize
252 * all these metrics based on that basic range.
254 # define SCHED_FIXEDPOINT_SHIFT 10
255 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
258 unsigned long weight
;
263 * The load_avg/util_avg accumulates an infinite geometric series
264 * (see __update_load_avg() in kernel/sched/fair.c).
266 * [load_avg definition]
268 * load_avg = runnable% * scale_load_down(load)
270 * where runnable% is the time ratio that a sched_entity is runnable.
271 * For cfs_rq, it is the aggregated load_avg of all runnable and
272 * blocked sched_entities.
274 * load_avg may also take frequency scaling into account:
276 * load_avg = runnable% * scale_load_down(load) * freq%
278 * where freq% is the CPU frequency normalized to the highest frequency.
280 * [util_avg definition]
282 * util_avg = running% * SCHED_CAPACITY_SCALE
284 * where running% is the time ratio that a sched_entity is running on
285 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
286 * and blocked sched_entities.
288 * util_avg may also factor frequency scaling and CPU capacity scaling:
290 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
292 * where freq% is the same as above, and capacity% is the CPU capacity
293 * normalized to the greatest capacity (due to uarch differences, etc).
295 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
296 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
297 * we therefore scale them to as large a range as necessary. This is for
298 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
302 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
303 * with the highest load (=88761), always runnable on a single cfs_rq,
304 * and should not overflow as the number already hits PID_MAX_LIMIT.
306 * For all other cases (including 32-bit kernels), struct load_weight's
307 * weight will overflow first before we do, because:
309 * Max(load_avg) <= Max(load.weight)
311 * Then it is the load_weight's responsibility to consider overflow
315 u64 last_update_time
;
319 unsigned long load_avg
;
320 unsigned long util_avg
;
323 struct sched_statistics
{
324 #ifdef CONFIG_SCHEDSTATS
334 s64 sum_sleep_runtime
;
341 u64 nr_migrations_cold
;
342 u64 nr_failed_migrations_affine
;
343 u64 nr_failed_migrations_running
;
344 u64 nr_failed_migrations_hot
;
345 u64 nr_forced_migrations
;
349 u64 nr_wakeups_migrate
;
350 u64 nr_wakeups_local
;
351 u64 nr_wakeups_remote
;
352 u64 nr_wakeups_affine
;
353 u64 nr_wakeups_affine_attempts
;
354 u64 nr_wakeups_passive
;
359 struct sched_entity
{
360 /* For load-balancing: */
361 struct load_weight load
;
362 struct rb_node run_node
;
363 struct list_head group_node
;
367 u64 sum_exec_runtime
;
369 u64 prev_sum_exec_runtime
;
373 struct sched_statistics statistics
;
375 #ifdef CONFIG_FAIR_GROUP_SCHED
377 struct sched_entity
*parent
;
378 /* rq on which this entity is (to be) queued: */
379 struct cfs_rq
*cfs_rq
;
380 /* rq "owned" by this entity/group: */
386 * Per entity load average tracking.
388 * Put into separate cache line so it does not
389 * collide with read-mostly values above.
391 struct sched_avg avg ____cacheline_aligned_in_smp
;
395 struct sched_rt_entity
{
396 struct list_head run_list
;
397 unsigned long timeout
;
398 unsigned long watchdog_stamp
;
399 unsigned int time_slice
;
400 unsigned short on_rq
;
401 unsigned short on_list
;
403 struct sched_rt_entity
*back
;
404 #ifdef CONFIG_RT_GROUP_SCHED
405 struct sched_rt_entity
*parent
;
406 /* rq on which this entity is (to be) queued: */
408 /* rq "owned" by this entity/group: */
413 struct sched_dl_entity
{
414 struct rb_node rb_node
;
417 * Original scheduling parameters. Copied here from sched_attr
418 * during sched_setattr(), they will remain the same until
419 * the next sched_setattr().
421 u64 dl_runtime
; /* Maximum runtime for each instance */
422 u64 dl_deadline
; /* Relative deadline of each instance */
423 u64 dl_period
; /* Separation of two instances (period) */
424 u64 dl_bw
; /* dl_runtime / dl_period */
425 u64 dl_density
; /* dl_runtime / dl_deadline */
428 * Actual scheduling parameters. Initialized with the values above,
429 * they are continously updated during task execution. Note that
430 * the remaining runtime could be < 0 in case we are in overrun.
432 s64 runtime
; /* Remaining runtime for this instance */
433 u64 deadline
; /* Absolute deadline for this instance */
434 unsigned int flags
; /* Specifying the scheduler behaviour */
439 * @dl_throttled tells if we exhausted the runtime. If so, the
440 * task has to wait for a replenishment to be performed at the
441 * next firing of dl_timer.
443 * @dl_boosted tells if we are boosted due to DI. If so we are
444 * outside bandwidth enforcement mechanism (but only until we
445 * exit the critical section);
447 * @dl_yielded tells if task gave up the CPU before consuming
448 * all its available runtime during the last job.
450 * @dl_non_contending tells if the task is inactive while still
451 * contributing to the active utilization. In other words, it
452 * indicates if the inactive timer has been armed and its handler
453 * has not been executed yet. This flag is useful to avoid race
454 * conditions between the inactive timer handler and the wakeup
460 int dl_non_contending
;
463 * Bandwidth enforcement timer. Each -deadline task has its
464 * own bandwidth to be enforced, thus we need one timer per task.
466 struct hrtimer dl_timer
;
469 * Inactive timer, responsible for decreasing the active utilization
470 * at the "0-lag time". When a -deadline task blocks, it contributes
471 * to GRUB's active utilization until the "0-lag time", hence a
472 * timer is needed to decrease the active utilization at the correct
475 struct hrtimer inactive_timer
;
484 /* Otherwise the compiler can store garbage here: */
487 u32 s
; /* Set of bits. */
490 enum perf_event_task_context
{
491 perf_invalid_context
= -1,
494 perf_nr_task_contexts
,
498 struct wake_q_node
*next
;
502 #ifdef CONFIG_THREAD_INFO_IN_TASK
504 * For reasons of header soup (see current_thread_info()), this
505 * must be the first element of task_struct.
507 struct thread_info thread_info
;
509 /* -1 unrunnable, 0 runnable, >0 stopped: */
513 /* Per task flags (PF_*), defined further below: */
518 struct llist_node wake_entry
;
520 #ifdef CONFIG_THREAD_INFO_IN_TASK
524 unsigned int wakee_flips
;
525 unsigned long wakee_flip_decay_ts
;
526 struct task_struct
*last_wakee
;
535 unsigned int rt_priority
;
537 const struct sched_class
*sched_class
;
538 struct sched_entity se
;
539 struct sched_rt_entity rt
;
540 #ifdef CONFIG_CGROUP_SCHED
541 struct task_group
*sched_task_group
;
543 struct sched_dl_entity dl
;
545 #ifdef CONFIG_PREEMPT_NOTIFIERS
546 /* List of struct preempt_notifier: */
547 struct hlist_head preempt_notifiers
;
550 #ifdef CONFIG_BLK_DEV_IO_TRACE
551 unsigned int btrace_seq
;
556 cpumask_t cpus_allowed
;
558 #ifdef CONFIG_PREEMPT_RCU
559 int rcu_read_lock_nesting
;
560 union rcu_special rcu_read_unlock_special
;
561 struct list_head rcu_node_entry
;
562 struct rcu_node
*rcu_blocked_node
;
563 #endif /* #ifdef CONFIG_PREEMPT_RCU */
565 #ifdef CONFIG_TASKS_RCU
566 unsigned long rcu_tasks_nvcsw
;
567 bool rcu_tasks_holdout
;
568 struct list_head rcu_tasks_holdout_list
;
569 int rcu_tasks_idle_cpu
;
570 #endif /* #ifdef CONFIG_TASKS_RCU */
572 struct sched_info sched_info
;
574 struct list_head tasks
;
576 struct plist_node pushable_tasks
;
577 struct rb_node pushable_dl_tasks
;
580 struct mm_struct
*mm
;
581 struct mm_struct
*active_mm
;
583 /* Per-thread vma caching: */
584 struct vmacache vmacache
;
586 #ifdef SPLIT_RSS_COUNTING
587 struct task_rss_stat rss_stat
;
592 /* The signal sent when the parent dies: */
594 /* JOBCTL_*, siglock protected: */
595 unsigned long jobctl
;
597 /* Used for emulating ABI behavior of previous Linux versions: */
598 unsigned int personality
;
600 /* Scheduler bits, serialized by scheduler locks: */
601 unsigned sched_reset_on_fork
:1;
602 unsigned sched_contributes_to_load
:1;
603 unsigned sched_migrated
:1;
604 unsigned sched_remote_wakeup
:1;
605 /* Force alignment to the next boundary: */
608 /* Unserialized, strictly 'current' */
610 /* Bit to tell LSMs we're in execve(): */
611 unsigned in_execve
:1;
612 unsigned in_iowait
:1;
613 #ifndef 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;
625 #ifdef CONFIG_CGROUPS
626 /* disallow userland-initiated cgroup migration */
627 unsigned no_cgroup_migration
:1;
630 unsigned long atomic_flags
; /* Flags requiring atomic access. */
632 struct restart_block restart_block
;
637 #ifdef CONFIG_CC_STACKPROTECTOR
638 /* Canary value for the -fstack-protector GCC feature: */
639 unsigned long stack_canary
;
642 * Pointers to the (original) parent process, youngest child, younger sibling,
643 * older sibling, respectively. (p->father can be replaced with
644 * p->real_parent->pid)
647 /* Real parent process: */
648 struct task_struct __rcu
*real_parent
;
650 /* Recipient of SIGCHLD, wait4() reports: */
651 struct task_struct __rcu
*parent
;
654 * Children/sibling form the list of natural children:
656 struct list_head children
;
657 struct list_head sibling
;
658 struct task_struct
*group_leader
;
661 * 'ptraced' is the list of tasks this task is using ptrace() on.
663 * This includes both natural children and PTRACE_ATTACH targets.
664 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
666 struct list_head ptraced
;
667 struct list_head ptrace_entry
;
669 /* PID/PID hash table linkage. */
670 struct pid_link pids
[PIDTYPE_MAX
];
671 struct list_head thread_group
;
672 struct list_head thread_node
;
674 struct completion
*vfork_done
;
676 /* CLONE_CHILD_SETTID: */
677 int __user
*set_child_tid
;
679 /* CLONE_CHILD_CLEARTID: */
680 int __user
*clear_child_tid
;
684 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
689 struct prev_cputime prev_cputime
;
690 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
691 seqcount_t vtime_seqcount
;
692 unsigned long long vtime_snap
;
694 /* Task is sleeping or running in a CPU with VTIME inactive: */
696 /* Task runs in userspace in a CPU with VTIME active: */
698 /* Task runs in kernelspace in a CPU with VTIME active: */
703 #ifdef CONFIG_NO_HZ_FULL
704 atomic_t tick_dep_mask
;
706 /* Context switch counts: */
708 unsigned long nivcsw
;
710 /* Monotonic time in nsecs: */
713 /* Boot based time in nsecs: */
716 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
717 unsigned long min_flt
;
718 unsigned long maj_flt
;
720 #ifdef CONFIG_POSIX_TIMERS
721 struct task_cputime cputime_expires
;
722 struct list_head cpu_timers
[3];
725 /* Process credentials: */
727 /* Tracer's credentials at attach: */
728 const struct cred __rcu
*ptracer_cred
;
730 /* Objective and real subjective task credentials (COW): */
731 const struct cred __rcu
*real_cred
;
733 /* Effective (overridable) subjective task credentials (COW): */
734 const struct cred __rcu
*cred
;
737 * executable name, excluding path.
739 * - normally initialized setup_new_exec()
740 * - access it with [gs]et_task_comm()
741 * - lock it with task_lock()
743 char comm
[TASK_COMM_LEN
];
745 struct nameidata
*nameidata
;
747 #ifdef CONFIG_SYSVIPC
748 struct sysv_sem sysvsem
;
749 struct sysv_shm sysvshm
;
751 #ifdef CONFIG_DETECT_HUNG_TASK
752 unsigned long last_switch_count
;
754 /* Filesystem information: */
755 struct fs_struct
*fs
;
757 /* Open file information: */
758 struct files_struct
*files
;
761 struct nsproxy
*nsproxy
;
763 /* Signal handlers: */
764 struct signal_struct
*signal
;
765 struct sighand_struct
*sighand
;
767 sigset_t real_blocked
;
768 /* Restored if set_restore_sigmask() was used: */
769 sigset_t saved_sigmask
;
770 struct sigpending pending
;
771 unsigned long sas_ss_sp
;
773 unsigned int sas_ss_flags
;
775 struct callback_head
*task_works
;
777 struct audit_context
*audit_context
;
778 #ifdef CONFIG_AUDITSYSCALL
780 unsigned int sessionid
;
782 struct seccomp seccomp
;
784 /* Thread group tracking: */
788 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
789 spinlock_t alloc_lock
;
791 /* Protection of the PI data structures: */
792 raw_spinlock_t pi_lock
;
794 struct wake_q_node wake_q
;
796 #ifdef CONFIG_RT_MUTEXES
797 /* PI waiters blocked on a rt_mutex held by this task: */
798 struct rb_root pi_waiters
;
799 struct rb_node
*pi_waiters_leftmost
;
800 /* Updated under owner's pi_lock and rq lock */
801 struct task_struct
*pi_top_task
;
802 /* Deadlock detection and priority inheritance handling: */
803 struct rt_mutex_waiter
*pi_blocked_on
;
806 #ifdef CONFIG_DEBUG_MUTEXES
807 /* Mutex deadlock detection: */
808 struct mutex_waiter
*blocked_on
;
811 #ifdef CONFIG_TRACE_IRQFLAGS
812 unsigned int irq_events
;
813 unsigned long hardirq_enable_ip
;
814 unsigned long hardirq_disable_ip
;
815 unsigned int hardirq_enable_event
;
816 unsigned int hardirq_disable_event
;
817 int hardirqs_enabled
;
819 unsigned long softirq_disable_ip
;
820 unsigned long softirq_enable_ip
;
821 unsigned int softirq_disable_event
;
822 unsigned int softirq_enable_event
;
823 int softirqs_enabled
;
827 #ifdef CONFIG_LOCKDEP
828 # define MAX_LOCK_DEPTH 48UL
831 unsigned int lockdep_recursion
;
832 struct held_lock held_locks
[MAX_LOCK_DEPTH
];
833 gfp_t lockdep_reclaim_gfp
;
837 unsigned int in_ubsan
;
840 /* Journalling filesystem info: */
843 /* Stacked block device info: */
844 struct bio_list
*bio_list
;
847 /* Stack plugging: */
848 struct blk_plug
*plug
;
852 struct reclaim_state
*reclaim_state
;
854 struct backing_dev_info
*backing_dev_info
;
856 struct io_context
*io_context
;
859 unsigned long ptrace_message
;
860 siginfo_t
*last_siginfo
;
862 struct task_io_accounting ioac
;
863 #ifdef CONFIG_TASK_XACCT
864 /* Accumulated RSS usage: */
866 /* Accumulated virtual memory usage: */
868 /* stime + utime since last update: */
871 #ifdef CONFIG_CPUSETS
872 /* Protected by ->alloc_lock: */
873 nodemask_t mems_allowed
;
874 /* Seqence number to catch updates: */
875 seqcount_t mems_allowed_seq
;
876 int cpuset_mem_spread_rotor
;
877 int cpuset_slab_spread_rotor
;
879 #ifdef CONFIG_CGROUPS
880 /* Control Group info protected by css_set_lock: */
881 struct css_set __rcu
*cgroups
;
882 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
883 struct list_head cg_list
;
885 #ifdef CONFIG_INTEL_RDT_A
889 struct robust_list_head __user
*robust_list
;
891 struct compat_robust_list_head __user
*compat_robust_list
;
893 struct list_head pi_state_list
;
894 struct futex_pi_state
*pi_state_cache
;
896 #ifdef CONFIG_PERF_EVENTS
897 struct perf_event_context
*perf_event_ctxp
[perf_nr_task_contexts
];
898 struct mutex perf_event_mutex
;
899 struct list_head perf_event_list
;
901 #ifdef CONFIG_DEBUG_PREEMPT
902 unsigned long preempt_disable_ip
;
905 /* Protected by alloc_lock: */
906 struct mempolicy
*mempolicy
;
908 short pref_node_fork
;
910 #ifdef CONFIG_NUMA_BALANCING
912 unsigned int numa_scan_period
;
913 unsigned int numa_scan_period_max
;
914 int numa_preferred_nid
;
915 unsigned long numa_migrate_retry
;
916 /* Migration stamp: */
918 u64 last_task_numa_placement
;
919 u64 last_sum_exec_runtime
;
920 struct callback_head numa_work
;
922 struct list_head numa_entry
;
923 struct numa_group
*numa_group
;
926 * numa_faults is an array split into four regions:
927 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
928 * in this precise order.
930 * faults_memory: Exponential decaying average of faults on a per-node
931 * basis. Scheduling placement decisions are made based on these
932 * counts. The values remain static for the duration of a PTE scan.
933 * faults_cpu: Track the nodes the process was running on when a NUMA
934 * hinting fault was incurred.
935 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
936 * during the current scan window. When the scan completes, the counts
937 * in faults_memory and faults_cpu decay and these values are copied.
939 unsigned long *numa_faults
;
940 unsigned long total_numa_faults
;
943 * numa_faults_locality tracks if faults recorded during the last
944 * scan window were remote/local or failed to migrate. The task scan
945 * period is adapted based on the locality of the faults with different
946 * weights depending on whether they were shared or private faults
948 unsigned long numa_faults_locality
[3];
950 unsigned long numa_pages_migrated
;
951 #endif /* CONFIG_NUMA_BALANCING */
953 struct tlbflush_unmap_batch tlb_ubc
;
957 /* Cache last used pipe for splice(): */
958 struct pipe_inode_info
*splice_pipe
;
960 struct page_frag task_frag
;
962 #ifdef CONFIG_TASK_DELAY_ACCT
963 struct task_delay_info
*delays
;
966 #ifdef CONFIG_FAULT_INJECTION
970 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
971 * balance_dirty_pages() for a dirty throttling pause:
974 int nr_dirtied_pause
;
975 /* Start of a write-and-pause period: */
976 unsigned long dirty_paused_when
;
978 #ifdef CONFIG_LATENCYTOP
979 int latency_record_count
;
980 struct latency_record latency_record
[LT_SAVECOUNT
];
983 * Time slack values; these are used to round up poll() and
984 * select() etc timeout values. These are in nanoseconds.
987 u64 default_timer_slack_ns
;
990 unsigned int kasan_depth
;
993 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
994 /* Index of current stored address in ret_stack: */
997 /* Stack of return addresses for return function tracing: */
998 struct ftrace_ret_stack
*ret_stack
;
1000 /* Timestamp for last schedule: */
1001 unsigned long long ftrace_timestamp
;
1004 * Number of functions that haven't been traced
1005 * because of depth overrun:
1007 atomic_t trace_overrun
;
1009 /* Pause tracing: */
1010 atomic_t tracing_graph_pause
;
1013 #ifdef CONFIG_TRACING
1014 /* State flags for use by tracers: */
1015 unsigned long trace
;
1017 /* Bitmask and counter of trace recursion: */
1018 unsigned long trace_recursion
;
1019 #endif /* CONFIG_TRACING */
1022 /* Coverage collection mode enabled for this task (0 if disabled): */
1023 enum kcov_mode kcov_mode
;
1025 /* Size of the kcov_area: */
1026 unsigned int kcov_size
;
1028 /* Buffer for coverage collection: */
1031 /* KCOV descriptor wired with this task or NULL: */
1036 struct mem_cgroup
*memcg_in_oom
;
1037 gfp_t memcg_oom_gfp_mask
;
1038 int memcg_oom_order
;
1040 /* Number of pages to reclaim on returning to userland: */
1041 unsigned int memcg_nr_pages_over_high
;
1044 #ifdef CONFIG_UPROBES
1045 struct uprobe_task
*utask
;
1047 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1048 unsigned int sequential_io
;
1049 unsigned int sequential_io_avg
;
1051 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1052 unsigned long task_state_change
;
1054 int pagefault_disabled
;
1056 struct task_struct
*oom_reaper_list
;
1058 #ifdef CONFIG_VMAP_STACK
1059 struct vm_struct
*stack_vm_area
;
1061 #ifdef CONFIG_THREAD_INFO_IN_TASK
1062 /* A live task holds one reference: */
1063 atomic_t stack_refcount
;
1065 #ifdef CONFIG_LIVEPATCH
1068 #ifdef CONFIG_SECURITY
1069 /* Used by LSM modules for access restriction: */
1072 /* CPU-specific state of this task: */
1073 struct thread_struct thread
;
1076 * WARNING: on x86, 'thread_struct' contains a variable-sized
1077 * structure. It *MUST* be at the end of 'task_struct'.
1079 * Do not put anything below here!
1083 static inline struct pid
*task_pid(struct task_struct
*task
)
1085 return task
->pids
[PIDTYPE_PID
].pid
;
1088 static inline struct pid
*task_tgid(struct task_struct
*task
)
1090 return task
->group_leader
->pids
[PIDTYPE_PID
].pid
;
1094 * Without tasklist or RCU lock it is not safe to dereference
1095 * the result of task_pgrp/task_session even if task == current,
1096 * we can race with another thread doing sys_setsid/sys_setpgid.
1098 static inline struct pid
*task_pgrp(struct task_struct
*task
)
1100 return task
->group_leader
->pids
[PIDTYPE_PGID
].pid
;
1103 static inline struct pid
*task_session(struct task_struct
*task
)
1105 return task
->group_leader
->pids
[PIDTYPE_SID
].pid
;
1109 * the helpers to get the task's different pids as they are seen
1110 * from various namespaces
1112 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1113 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1115 * task_xid_nr_ns() : id seen from the ns specified;
1117 * see also pid_nr() etc in include/linux/pid.h
1119 pid_t
__task_pid_nr_ns(struct task_struct
*task
, enum pid_type type
, struct pid_namespace
*ns
);
1121 static inline pid_t
task_pid_nr(struct task_struct
*tsk
)
1126 static inline pid_t
task_pid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1128 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, ns
);
1131 static inline pid_t
task_pid_vnr(struct task_struct
*tsk
)
1133 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, NULL
);
1137 static inline pid_t
task_tgid_nr(struct task_struct
*tsk
)
1142 extern pid_t
task_tgid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
);
1144 static inline pid_t
task_tgid_vnr(struct task_struct
*tsk
)
1146 return pid_vnr(task_tgid(tsk
));
1150 * pid_alive - check that a task structure is not stale
1151 * @p: Task structure to be checked.
1153 * Test if a process is not yet dead (at most zombie state)
1154 * If pid_alive fails, then pointers within the task structure
1155 * can be stale and must not be dereferenced.
1157 * Return: 1 if the process is alive. 0 otherwise.
1159 static inline int pid_alive(const struct task_struct
*p
)
1161 return p
->pids
[PIDTYPE_PID
].pid
!= NULL
;
1164 static inline pid_t
task_ppid_nr_ns(const struct task_struct
*tsk
, struct pid_namespace
*ns
)
1170 pid
= task_tgid_nr_ns(rcu_dereference(tsk
->real_parent
), ns
);
1176 static inline pid_t
task_ppid_nr(const struct task_struct
*tsk
)
1178 return task_ppid_nr_ns(tsk
, &init_pid_ns
);
1181 static inline pid_t
task_pgrp_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1183 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, ns
);
1186 static inline pid_t
task_pgrp_vnr(struct task_struct
*tsk
)
1188 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, NULL
);
1192 static inline pid_t
task_session_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1194 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, ns
);
1197 static inline pid_t
task_session_vnr(struct task_struct
*tsk
)
1199 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, NULL
);
1202 /* Obsolete, do not use: */
1203 static inline pid_t
task_pgrp_nr(struct task_struct
*tsk
)
1205 return task_pgrp_nr_ns(tsk
, &init_pid_ns
);
1209 * is_global_init - check if a task structure is init. Since init
1210 * is free to have sub-threads we need to check tgid.
1211 * @tsk: Task structure to be checked.
1213 * Check if a task structure is the first user space task the kernel created.
1215 * Return: 1 if the task structure is init. 0 otherwise.
1217 static inline int is_global_init(struct task_struct
*tsk
)
1219 return task_tgid_nr(tsk
) == 1;
1222 extern struct pid
*cad_pid
;
1227 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1228 #define PF_EXITING 0x00000004 /* Getting shut down */
1229 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1230 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1231 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1232 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1233 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1234 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1235 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1236 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1237 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1238 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1239 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1240 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1241 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1242 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1243 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1244 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1245 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1246 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1247 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1248 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1249 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1250 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1251 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1252 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1253 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1254 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1257 * Only the _current_ task can read/write to tsk->flags, but other
1258 * tasks can access tsk->flags in readonly mode for example
1259 * with tsk_used_math (like during threaded core dumping).
1260 * There is however an exception to this rule during ptrace
1261 * or during fork: the ptracer task is allowed to write to the
1262 * child->flags of its traced child (same goes for fork, the parent
1263 * can write to the child->flags), because we're guaranteed the
1264 * child is not running and in turn not changing child->flags
1265 * at the same time the parent does it.
1267 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1268 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1269 #define clear_used_math() clear_stopped_child_used_math(current)
1270 #define set_used_math() set_stopped_child_used_math(current)
1272 #define conditional_stopped_child_used_math(condition, child) \
1273 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1275 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1277 #define copy_to_stopped_child_used_math(child) \
1278 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1280 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1281 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1282 #define used_math() tsk_used_math(current)
1284 static inline bool is_percpu_thread(void)
1287 return (current
->flags
& PF_NO_SETAFFINITY
) &&
1288 (current
->nr_cpus_allowed
== 1);
1294 /* Per-process atomic flags. */
1295 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1296 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1297 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1300 #define TASK_PFA_TEST(name, func) \
1301 static inline bool task_##func(struct task_struct *p) \
1302 { return test_bit(PFA_##name, &p->atomic_flags); }
1304 #define TASK_PFA_SET(name, func) \
1305 static inline void task_set_##func(struct task_struct *p) \
1306 { set_bit(PFA_##name, &p->atomic_flags); }
1308 #define TASK_PFA_CLEAR(name, func) \
1309 static inline void task_clear_##func(struct task_struct *p) \
1310 { clear_bit(PFA_##name, &p->atomic_flags); }
1312 TASK_PFA_TEST(NO_NEW_PRIVS
, no_new_privs
)
1313 TASK_PFA_SET(NO_NEW_PRIVS
, no_new_privs
)
1315 TASK_PFA_TEST(SPREAD_PAGE
, spread_page
)
1316 TASK_PFA_SET(SPREAD_PAGE
, spread_page
)
1317 TASK_PFA_CLEAR(SPREAD_PAGE
, spread_page
)
1319 TASK_PFA_TEST(SPREAD_SLAB
, spread_slab
)
1320 TASK_PFA_SET(SPREAD_SLAB
, spread_slab
)
1321 TASK_PFA_CLEAR(SPREAD_SLAB
, spread_slab
)
1324 current_restore_flags(unsigned long orig_flags
, unsigned long flags
)
1326 current
->flags
&= ~flags
;
1327 current
->flags
|= orig_flags
& flags
;
1330 extern int cpuset_cpumask_can_shrink(const struct cpumask
*cur
, const struct cpumask
*trial
);
1331 extern int task_can_attach(struct task_struct
*p
, const struct cpumask
*cs_cpus_allowed
);
1333 extern void do_set_cpus_allowed(struct task_struct
*p
, const struct cpumask
*new_mask
);
1334 extern int set_cpus_allowed_ptr(struct task_struct
*p
, const struct cpumask
*new_mask
);
1336 static inline void do_set_cpus_allowed(struct task_struct
*p
, const struct cpumask
*new_mask
)
1339 static inline int set_cpus_allowed_ptr(struct task_struct
*p
, const struct cpumask
*new_mask
)
1341 if (!cpumask_test_cpu(0, new_mask
))
1347 #ifndef cpu_relax_yield
1348 #define cpu_relax_yield() cpu_relax()
1351 extern int yield_to(struct task_struct
*p
, bool preempt
);
1352 extern void set_user_nice(struct task_struct
*p
, long nice
);
1353 extern int task_prio(const struct task_struct
*p
);
1356 * task_nice - return the nice value of a given task.
1357 * @p: the task in question.
1359 * Return: The nice value [ -20 ... 0 ... 19 ].
1361 static inline int task_nice(const struct task_struct
*p
)
1363 return PRIO_TO_NICE((p
)->static_prio
);
1366 extern int can_nice(const struct task_struct
*p
, const int nice
);
1367 extern int task_curr(const struct task_struct
*p
);
1368 extern int idle_cpu(int cpu
);
1369 extern int sched_setscheduler(struct task_struct
*, int, const struct sched_param
*);
1370 extern int sched_setscheduler_nocheck(struct task_struct
*, int, const struct sched_param
*);
1371 extern int sched_setattr(struct task_struct
*, const struct sched_attr
*);
1372 extern struct task_struct
*idle_task(int cpu
);
1375 * is_idle_task - is the specified task an idle task?
1376 * @p: the task in question.
1378 * Return: 1 if @p is an idle task. 0 otherwise.
1380 static inline bool is_idle_task(const struct task_struct
*p
)
1382 return !!(p
->flags
& PF_IDLE
);
1385 extern struct task_struct
*curr_task(int cpu
);
1386 extern void ia64_set_curr_task(int cpu
, struct task_struct
*p
);
1390 union thread_union
{
1391 #ifndef CONFIG_THREAD_INFO_IN_TASK
1392 struct thread_info thread_info
;
1394 unsigned long stack
[THREAD_SIZE
/sizeof(long)];
1397 #ifdef CONFIG_THREAD_INFO_IN_TASK
1398 static inline struct thread_info
*task_thread_info(struct task_struct
*task
)
1400 return &task
->thread_info
;
1402 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1403 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1407 * find a task by one of its numerical ids
1409 * find_task_by_pid_ns():
1410 * finds a task by its pid in the specified namespace
1411 * find_task_by_vpid():
1412 * finds a task by its virtual pid
1414 * see also find_vpid() etc in include/linux/pid.h
1417 extern struct task_struct
*find_task_by_vpid(pid_t nr
);
1418 extern struct task_struct
*find_task_by_pid_ns(pid_t nr
, struct pid_namespace
*ns
);
1420 extern int wake_up_state(struct task_struct
*tsk
, unsigned int state
);
1421 extern int wake_up_process(struct task_struct
*tsk
);
1422 extern void wake_up_new_task(struct task_struct
*tsk
);
1425 extern void kick_process(struct task_struct
*tsk
);
1427 static inline void kick_process(struct task_struct
*tsk
) { }
1430 extern void __set_task_comm(struct task_struct
*tsk
, const char *from
, bool exec
);
1432 static inline void set_task_comm(struct task_struct
*tsk
, const char *from
)
1434 __set_task_comm(tsk
, from
, false);
1437 extern char *get_task_comm(char *to
, struct task_struct
*tsk
);
1440 void scheduler_ipi(void);
1441 extern unsigned long wait_task_inactive(struct task_struct
*, long match_state
);
1443 static inline void scheduler_ipi(void) { }
1444 static inline unsigned long wait_task_inactive(struct task_struct
*p
, long match_state
)
1451 * Set thread flags in other task's structures.
1452 * See asm/thread_info.h for TIF_xxxx flags available:
1454 static inline void set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1456 set_ti_thread_flag(task_thread_info(tsk
), flag
);
1459 static inline void clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1461 clear_ti_thread_flag(task_thread_info(tsk
), flag
);
1464 static inline int test_and_set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1466 return test_and_set_ti_thread_flag(task_thread_info(tsk
), flag
);
1469 static inline int test_and_clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1471 return test_and_clear_ti_thread_flag(task_thread_info(tsk
), flag
);
1474 static inline int test_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1476 return test_ti_thread_flag(task_thread_info(tsk
), flag
);
1479 static inline void set_tsk_need_resched(struct task_struct
*tsk
)
1481 set_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
1484 static inline void clear_tsk_need_resched(struct task_struct
*tsk
)
1486 clear_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
1489 static inline int test_tsk_need_resched(struct task_struct
*tsk
)
1491 return unlikely(test_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
));
1495 * cond_resched() and cond_resched_lock(): latency reduction via
1496 * explicit rescheduling in places that are safe. The return
1497 * value indicates whether a reschedule was done in fact.
1498 * cond_resched_lock() will drop the spinlock before scheduling,
1499 * cond_resched_softirq() will enable bhs before scheduling.
1501 #ifndef CONFIG_PREEMPT
1502 extern int _cond_resched(void);
1504 static inline int _cond_resched(void) { return 0; }
1507 #define cond_resched() ({ \
1508 ___might_sleep(__FILE__, __LINE__, 0); \
1512 extern int __cond_resched_lock(spinlock_t
*lock
);
1514 #define cond_resched_lock(lock) ({ \
1515 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1516 __cond_resched_lock(lock); \
1519 extern int __cond_resched_softirq(void);
1521 #define cond_resched_softirq() ({ \
1522 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1523 __cond_resched_softirq(); \
1526 static inline void cond_resched_rcu(void)
1528 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1536 * Does a critical section need to be broken due to another
1537 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1538 * but a general need for low latency)
1540 static inline int spin_needbreak(spinlock_t
*lock
)
1542 #ifdef CONFIG_PREEMPT
1543 return spin_is_contended(lock
);
1549 static __always_inline
bool need_resched(void)
1551 return unlikely(tif_need_resched());
1555 * Wrappers for p->thread_info->cpu access. No-op on UP.
1559 static inline unsigned int task_cpu(const struct task_struct
*p
)
1561 #ifdef CONFIG_THREAD_INFO_IN_TASK
1564 return task_thread_info(p
)->cpu
;
1568 extern void set_task_cpu(struct task_struct
*p
, unsigned int cpu
);
1572 static inline unsigned int task_cpu(const struct task_struct
*p
)
1577 static inline void set_task_cpu(struct task_struct
*p
, unsigned int cpu
)
1581 #endif /* CONFIG_SMP */
1584 * In order to reduce various lock holder preemption latencies provide an
1585 * interface to see if a vCPU is currently running or not.
1587 * This allows us to terminate optimistic spin loops and block, analogous to
1588 * the native optimistic spin heuristic of testing if the lock owner task is
1591 #ifndef vcpu_is_preempted
1592 # define vcpu_is_preempted(cpu) false
1595 extern long sched_setaffinity(pid_t pid
, const struct cpumask
*new_mask
);
1596 extern long sched_getaffinity(pid_t pid
, struct cpumask
*mask
);
1598 #ifndef TASK_SIZE_OF
1599 #define TASK_SIZE_OF(tsk) TASK_SIZE