4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
8 #include <linux/capability.h>
9 #include <linux/mutex.h>
10 #include <linux/plist.h>
11 #include <linux/mm_types_task.h>
12 #include <asm/ptrace.h>
14 #include <linux/sem.h>
15 #include <linux/shm.h>
16 #include <linux/signal.h>
17 #include <linux/signal_types.h>
18 #include <linux/pid.h>
19 #include <linux/seccomp.h>
20 #include <linux/rculist.h>
21 #include <linux/rtmutex.h>
23 #include <linux/resource.h>
24 #include <linux/hrtimer.h>
25 #include <linux/kcov.h>
26 #include <linux/task_io_accounting.h>
27 #include <linux/latencytop.h>
28 #include <linux/cred.h>
29 #include <linux/gfp.h>
30 #include <linux/topology.h>
31 #include <linux/magic.h>
32 #include <linux/cgroup-defs.h>
34 #include <asm/current.h>
36 /* task_struct member predeclarations: */
39 struct backing_dev_info;
45 struct futex_pi_state;
50 struct perf_event_context;
52 struct pipe_inode_info;
55 struct robust_list_head;
59 struct sighand_struct;
61 struct task_delay_info;
67 * Task state bitmask. NOTE! These bits are also
68 * encoded in fs/proc/array.c: get_task_state().
70 * We have two separate sets of flags: task->state
71 * is about runnability, while task->exit_state are
72 * about the task exiting. Confusing, but this way
73 * modifying one set can't modify the other one by
76 #define TASK_RUNNING 0
77 #define TASK_INTERRUPTIBLE 1
78 #define TASK_UNINTERRUPTIBLE 2
79 #define __TASK_STOPPED 4
80 #define __TASK_TRACED 8
81 /* in tsk->exit_state */
83 #define EXIT_ZOMBIE 32
84 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
85 /* in tsk->state again */
87 #define TASK_WAKEKILL 128
88 #define TASK_WAKING 256
89 #define TASK_PARKED 512
90 #define TASK_NOLOAD 1024
92 #define TASK_STATE_MAX 4096
94 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
96 /* Convenience macros for the sake of set_current_state */
97 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
98 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
99 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
101 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
103 /* Convenience macros for the sake of wake_up */
104 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
105 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
107 /* get_task_state() */
108 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
109 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
110 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
112 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
113 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
114 #define task_is_stopped_or_traced(task) \
115 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
116 #define task_contributes_to_load(task) \
117 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
118 (task->flags & PF_FROZEN) == 0 && \
119 (task->state & TASK_NOLOAD) == 0)
121 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
123 #define __set_current_state(state_value) \
125 current->task_state_change = _THIS_IP_; \
126 current->state = (state_value); \
128 #define set_current_state(state_value) \
130 current->task_state_change = _THIS_IP_; \
131 smp_store_mb(current->state, (state_value)); \
136 * set_current_state() includes a barrier so that the write of current->state
137 * is correctly serialised wrt the caller's subsequent test of whether to
141 * set_current_state(TASK_UNINTERRUPTIBLE);
147 * __set_current_state(TASK_RUNNING);
149 * If the caller does not need such serialisation (because, for instance, the
150 * condition test and condition change and wakeup are under the same lock) then
151 * use __set_current_state().
153 * The above is typically ordered against the wakeup, which does:
155 * need_sleep = false;
156 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
158 * Where wake_up_state() (and all other wakeup primitives) imply enough
159 * barriers to order the store of the variable against wakeup.
161 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
162 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
163 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
165 * This is obviously fine, since they both store the exact same value.
167 * Also see the comments of try_to_wake_up().
169 #define __set_current_state(state_value) \
170 do { current->state = (state_value); } while (0)
171 #define set_current_state(state_value) \
172 smp_store_mb(current->state, (state_value))
176 /* Task command name length */
177 #define TASK_COMM_LEN 16
179 extern cpumask_var_t cpu_isolated_map;
181 extern int runqueue_is_locked(int cpu);
183 extern void update_process_times(int user);
184 extern void scheduler_tick(void);
186 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
187 extern signed long schedule_timeout(signed long timeout);
188 extern signed long schedule_timeout_interruptible(signed long timeout);
189 extern signed long schedule_timeout_killable(signed long timeout);
190 extern signed long schedule_timeout_uninterruptible(signed long timeout);
191 extern signed long schedule_timeout_idle(signed long timeout);
192 asmlinkage void schedule(void);
193 extern void schedule_preempt_disabled(void);
195 extern int __must_check io_schedule_prepare(void);
196 extern void io_schedule_finish(int token);
197 extern long io_schedule_timeout(long timeout);
198 extern void io_schedule(void);
201 * struct prev_cputime - snaphsot of system and user cputime
202 * @utime: time spent in user mode
203 * @stime: time spent in system mode
204 * @lock: protects the above two fields
206 * Stores previous user/system time values such that we can guarantee
209 struct prev_cputime {
210 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
218 * struct task_cputime - collected CPU time counts
219 * @utime: time spent in user mode, in nanoseconds
220 * @stime: time spent in kernel mode, in nanoseconds
221 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
223 * This structure groups together three kinds of CPU time that are tracked for
224 * threads and thread groups. Most things considering CPU time want to group
225 * these counts together and treat all three of them in parallel.
227 struct task_cputime {
230 unsigned long long sum_exec_runtime;
233 /* Alternate field names when used to cache expirations. */
234 #define virt_exp utime
235 #define prof_exp stime
236 #define sched_exp sum_exec_runtime
238 #include <linux/rwsem.h>
240 #ifdef CONFIG_SCHED_INFO
242 /* cumulative counters */
243 unsigned long pcount; /* # of times run on this cpu */
244 unsigned long long run_delay; /* time spent waiting on a runqueue */
247 unsigned long long last_arrival,/* when we last ran on a cpu */
248 last_queued; /* when we were last queued to run */
250 #endif /* CONFIG_SCHED_INFO */
253 * Integer metrics need fixed point arithmetic, e.g., sched/fair
254 * has a few: load, load_avg, util_avg, freq, and capacity.
256 * We define a basic fixed point arithmetic range, and then formalize
257 * all these metrics based on that basic range.
259 # define SCHED_FIXEDPOINT_SHIFT 10
260 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
262 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
263 extern void prefetch_stack(struct task_struct *t);
265 static inline void prefetch_stack(struct task_struct *t) { }
269 unsigned long weight;
274 * The load_avg/util_avg accumulates an infinite geometric series
275 * (see __update_load_avg() in kernel/sched/fair.c).
277 * [load_avg definition]
279 * load_avg = runnable% * scale_load_down(load)
281 * where runnable% is the time ratio that a sched_entity is runnable.
282 * For cfs_rq, it is the aggregated load_avg of all runnable and
283 * blocked sched_entities.
285 * load_avg may also take frequency scaling into account:
287 * load_avg = runnable% * scale_load_down(load) * freq%
289 * where freq% is the CPU frequency normalized to the highest frequency.
291 * [util_avg definition]
293 * util_avg = running% * SCHED_CAPACITY_SCALE
295 * where running% is the time ratio that a sched_entity is running on
296 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
297 * and blocked sched_entities.
299 * util_avg may also factor frequency scaling and CPU capacity scaling:
301 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
303 * where freq% is the same as above, and capacity% is the CPU capacity
304 * normalized to the greatest capacity (due to uarch differences, etc).
306 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
307 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
308 * we therefore scale them to as large a range as necessary. This is for
309 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
313 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
314 * with the highest load (=88761), always runnable on a single cfs_rq,
315 * and should not overflow as the number already hits PID_MAX_LIMIT.
317 * For all other cases (including 32-bit kernels), struct load_weight's
318 * weight will overflow first before we do, because:
320 * Max(load_avg) <= Max(load.weight)
322 * Then it is the load_weight's responsibility to consider overflow
326 u64 last_update_time, load_sum;
327 u32 util_sum, period_contrib;
328 unsigned long load_avg, util_avg;
331 #ifdef CONFIG_SCHEDSTATS
332 struct sched_statistics {
342 s64 sum_sleep_runtime;
349 u64 nr_migrations_cold;
350 u64 nr_failed_migrations_affine;
351 u64 nr_failed_migrations_running;
352 u64 nr_failed_migrations_hot;
353 u64 nr_forced_migrations;
357 u64 nr_wakeups_migrate;
358 u64 nr_wakeups_local;
359 u64 nr_wakeups_remote;
360 u64 nr_wakeups_affine;
361 u64 nr_wakeups_affine_attempts;
362 u64 nr_wakeups_passive;
367 struct sched_entity {
368 struct load_weight load; /* for load-balancing */
369 struct rb_node run_node;
370 struct list_head group_node;
374 u64 sum_exec_runtime;
376 u64 prev_sum_exec_runtime;
380 #ifdef CONFIG_SCHEDSTATS
381 struct sched_statistics statistics;
384 #ifdef CONFIG_FAIR_GROUP_SCHED
386 struct sched_entity *parent;
387 /* rq on which this entity is (to be) queued: */
388 struct cfs_rq *cfs_rq;
389 /* rq "owned" by this entity/group: */
395 * Per entity load average tracking.
397 * Put into separate cache line so it does not
398 * collide with read-mostly values above.
400 struct sched_avg avg ____cacheline_aligned_in_smp;
404 struct sched_rt_entity {
405 struct list_head run_list;
406 unsigned long timeout;
407 unsigned long watchdog_stamp;
408 unsigned int time_slice;
409 unsigned short on_rq;
410 unsigned short on_list;
412 struct sched_rt_entity *back;
413 #ifdef CONFIG_RT_GROUP_SCHED
414 struct sched_rt_entity *parent;
415 /* rq on which this entity is (to be) queued: */
417 /* rq "owned" by this entity/group: */
422 struct sched_dl_entity {
423 struct rb_node rb_node;
426 * Original scheduling parameters. Copied here from sched_attr
427 * during sched_setattr(), they will remain the same until
428 * the next sched_setattr().
430 u64 dl_runtime; /* maximum runtime for each instance */
431 u64 dl_deadline; /* relative deadline of each instance */
432 u64 dl_period; /* separation of two instances (period) */
433 u64 dl_bw; /* dl_runtime / dl_deadline */
436 * Actual scheduling parameters. Initialized with the values above,
437 * they are continously updated during task execution. Note that
438 * the remaining runtime could be < 0 in case we are in overrun.
440 s64 runtime; /* remaining runtime for this instance */
441 u64 deadline; /* absolute deadline for this instance */
442 unsigned int flags; /* specifying the scheduler behaviour */
447 * @dl_throttled tells if we exhausted the runtime. If so, the
448 * task has to wait for a replenishment to be performed at the
449 * next firing of dl_timer.
451 * @dl_boosted tells if we are boosted due to DI. If so we are
452 * outside bandwidth enforcement mechanism (but only until we
453 * exit the critical section);
455 * @dl_yielded tells if task gave up the cpu before consuming
456 * all its available runtime during the last job.
458 int dl_throttled, dl_boosted, dl_yielded;
461 * Bandwidth enforcement timer. Each -deadline task has its
462 * own bandwidth to be enforced, thus we need one timer per task.
464 struct hrtimer dl_timer;
472 u8 pad; /* Otherwise the compiler can store garbage here. */
474 u32 s; /* Set of bits. */
477 enum perf_event_task_context {
478 perf_invalid_context = -1,
481 perf_nr_task_contexts,
485 struct wake_q_node *next;
488 /* Track pages that require TLB flushes */
489 struct tlbflush_unmap_batch {
491 * Each bit set is a CPU that potentially has a TLB entry for one of
492 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
494 struct cpumask cpumask;
496 /* True if any bit in cpumask is set */
500 * If true then the PTE was dirty when unmapped. The entry must be
501 * flushed before IO is initiated or a stale TLB entry potentially
502 * allows an update without redirtying the page.
508 #ifdef CONFIG_THREAD_INFO_IN_TASK
510 * For reasons of header soup (see current_thread_info()), this
511 * must be the first element of task_struct.
513 struct thread_info thread_info;
515 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
518 unsigned int flags; /* per process flags, defined below */
522 struct llist_node wake_entry;
524 #ifdef CONFIG_THREAD_INFO_IN_TASK
525 unsigned int cpu; /* current CPU */
527 unsigned int wakee_flips;
528 unsigned long wakee_flip_decay_ts;
529 struct task_struct *last_wakee;
535 int prio, static_prio, normal_prio;
536 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 */
564 #ifdef CONFIG_TASKS_RCU
565 unsigned long rcu_tasks_nvcsw;
566 bool rcu_tasks_holdout;
567 struct list_head rcu_tasks_holdout_list;
568 int rcu_tasks_idle_cpu;
569 #endif /* #ifdef CONFIG_TASKS_RCU */
571 #ifdef CONFIG_SCHED_INFO
572 struct sched_info sched_info;
575 struct list_head tasks;
577 struct plist_node pushable_tasks;
578 struct rb_node pushable_dl_tasks;
581 struct mm_struct *mm, *active_mm;
583 /* Per-thread vma caching: */
584 struct vmacache vmacache;
586 #if defined(SPLIT_RSS_COUNTING)
587 struct task_rss_stat rss_stat;
591 int exit_code, exit_signal;
592 int pdeath_signal; /* The signal sent when the parent dies */
593 unsigned long jobctl; /* JOBCTL_*, siglock protected */
595 /* Used for emulating ABI behavior of previous Linux versions */
596 unsigned int personality;
598 /* scheduler bits, serialized by scheduler locks */
599 unsigned sched_reset_on_fork:1;
600 unsigned sched_contributes_to_load:1;
601 unsigned sched_migrated:1;
602 unsigned sched_remote_wakeup:1;
603 unsigned :0; /* force alignment to the next boundary */
605 /* unserialized, strictly 'current' */
606 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
607 unsigned in_iowait:1;
608 #if !defined(TIF_RESTORE_SIGMASK)
609 unsigned restore_sigmask:1;
612 unsigned memcg_may_oom:1;
614 unsigned memcg_kmem_skip_account:1;
617 #ifdef CONFIG_COMPAT_BRK
618 unsigned brk_randomized:1;
621 unsigned long atomic_flags; /* Flags needing atomic access. */
623 struct restart_block restart_block;
628 #ifdef CONFIG_CC_STACKPROTECTOR
629 /* Canary value for the -fstack-protector gcc feature */
630 unsigned long stack_canary;
633 * pointers to (original) parent process, youngest child, younger sibling,
634 * older sibling, respectively. (p->father can be replaced with
635 * p->real_parent->pid)
637 struct task_struct __rcu *real_parent; /* real parent process */
638 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
640 * children/sibling forms the list of my natural children
642 struct list_head children; /* list of my children */
643 struct list_head sibling; /* linkage in my parent's children list */
644 struct task_struct *group_leader; /* threadgroup leader */
647 * ptraced is the list of tasks this task is using ptrace on.
648 * This includes both natural children and PTRACE_ATTACH targets.
649 * p->ptrace_entry is p's link on the p->parent->ptraced list.
651 struct list_head ptraced;
652 struct list_head ptrace_entry;
654 /* PID/PID hash table linkage. */
655 struct pid_link pids[PIDTYPE_MAX];
656 struct list_head thread_group;
657 struct list_head thread_node;
659 struct completion *vfork_done; /* for vfork() */
660 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
661 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
664 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
665 u64 utimescaled, stimescaled;
668 struct prev_cputime prev_cputime;
669 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
670 seqcount_t vtime_seqcount;
671 unsigned long long vtime_snap;
673 /* Task is sleeping or running in a CPU with VTIME inactive */
675 /* Task runs in userspace in a CPU with VTIME active */
677 /* Task runs in kernelspace in a CPU with VTIME active */
682 #ifdef CONFIG_NO_HZ_FULL
683 atomic_t tick_dep_mask;
685 unsigned long nvcsw, nivcsw; /* context switch counts */
686 u64 start_time; /* monotonic time in nsec */
687 u64 real_start_time; /* boot based time in nsec */
688 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
689 unsigned long min_flt, maj_flt;
691 #ifdef CONFIG_POSIX_TIMERS
692 struct task_cputime cputime_expires;
693 struct list_head cpu_timers[3];
696 /* process credentials */
697 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
698 const struct cred __rcu *real_cred; /* objective and real subjective task
699 * credentials (COW) */
700 const struct cred __rcu *cred; /* effective (overridable) subjective task
701 * credentials (COW) */
702 char comm[TASK_COMM_LEN]; /* executable name excluding path
703 - access with [gs]et_task_comm (which lock
705 - initialized normally by setup_new_exec */
706 /* file system info */
707 struct nameidata *nameidata;
708 #ifdef CONFIG_SYSVIPC
710 struct sysv_sem sysvsem;
711 struct sysv_shm sysvshm;
713 #ifdef CONFIG_DETECT_HUNG_TASK
714 /* hung task detection */
715 unsigned long last_switch_count;
717 /* filesystem information */
718 struct fs_struct *fs;
719 /* open file information */
720 struct files_struct *files;
722 struct nsproxy *nsproxy;
723 /* signal handlers */
724 struct signal_struct *signal;
725 struct sighand_struct *sighand;
727 sigset_t blocked, real_blocked;
728 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
729 struct sigpending pending;
731 unsigned long sas_ss_sp;
733 unsigned sas_ss_flags;
735 struct callback_head *task_works;
737 struct audit_context *audit_context;
738 #ifdef CONFIG_AUDITSYSCALL
740 unsigned int sessionid;
742 struct seccomp seccomp;
744 /* Thread group tracking */
747 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
749 spinlock_t alloc_lock;
751 /* Protection of the PI data structures: */
752 raw_spinlock_t pi_lock;
754 struct wake_q_node wake_q;
756 #ifdef CONFIG_RT_MUTEXES
757 /* PI waiters blocked on a rt_mutex held by this task */
758 struct rb_root pi_waiters;
759 struct rb_node *pi_waiters_leftmost;
760 /* Deadlock detection and priority inheritance handling */
761 struct rt_mutex_waiter *pi_blocked_on;
764 #ifdef CONFIG_DEBUG_MUTEXES
765 /* mutex deadlock detection */
766 struct mutex_waiter *blocked_on;
768 #ifdef CONFIG_TRACE_IRQFLAGS
769 unsigned int irq_events;
770 unsigned long hardirq_enable_ip;
771 unsigned long hardirq_disable_ip;
772 unsigned int hardirq_enable_event;
773 unsigned int hardirq_disable_event;
774 int hardirqs_enabled;
776 unsigned long softirq_disable_ip;
777 unsigned long softirq_enable_ip;
778 unsigned int softirq_disable_event;
779 unsigned int softirq_enable_event;
780 int softirqs_enabled;
783 #ifdef CONFIG_LOCKDEP
784 # define MAX_LOCK_DEPTH 48UL
787 unsigned int lockdep_recursion;
788 struct held_lock held_locks[MAX_LOCK_DEPTH];
789 gfp_t lockdep_reclaim_gfp;
792 unsigned int in_ubsan;
795 /* journalling filesystem info */
798 /* stacked block device info */
799 struct bio_list *bio_list;
803 struct blk_plug *plug;
807 struct reclaim_state *reclaim_state;
809 struct backing_dev_info *backing_dev_info;
811 struct io_context *io_context;
813 unsigned long ptrace_message;
814 siginfo_t *last_siginfo; /* For ptrace use. */
815 struct task_io_accounting ioac;
816 #if defined(CONFIG_TASK_XACCT)
817 u64 acct_rss_mem1; /* accumulated rss usage */
818 u64 acct_vm_mem1; /* accumulated virtual memory usage */
819 u64 acct_timexpd; /* stime + utime since last update */
821 #ifdef CONFIG_CPUSETS
822 nodemask_t mems_allowed; /* Protected by alloc_lock */
823 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
824 int cpuset_mem_spread_rotor;
825 int cpuset_slab_spread_rotor;
827 #ifdef CONFIG_CGROUPS
828 /* Control Group info protected by css_set_lock */
829 struct css_set __rcu *cgroups;
830 /* cg_list protected by css_set_lock and tsk->alloc_lock */
831 struct list_head cg_list;
833 #ifdef CONFIG_INTEL_RDT_A
837 struct robust_list_head __user *robust_list;
839 struct compat_robust_list_head __user *compat_robust_list;
841 struct list_head pi_state_list;
842 struct futex_pi_state *pi_state_cache;
844 #ifdef CONFIG_PERF_EVENTS
845 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
846 struct mutex perf_event_mutex;
847 struct list_head perf_event_list;
849 #ifdef CONFIG_DEBUG_PREEMPT
850 unsigned long preempt_disable_ip;
853 struct mempolicy *mempolicy; /* Protected by alloc_lock */
855 short pref_node_fork;
857 #ifdef CONFIG_NUMA_BALANCING
859 unsigned int numa_scan_period;
860 unsigned int numa_scan_period_max;
861 int numa_preferred_nid;
862 unsigned long numa_migrate_retry;
863 u64 node_stamp; /* migration stamp */
864 u64 last_task_numa_placement;
865 u64 last_sum_exec_runtime;
866 struct callback_head numa_work;
868 struct list_head numa_entry;
869 struct numa_group *numa_group;
872 * numa_faults is an array split into four regions:
873 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
874 * in this precise order.
876 * faults_memory: Exponential decaying average of faults on a per-node
877 * basis. Scheduling placement decisions are made based on these
878 * counts. The values remain static for the duration of a PTE scan.
879 * faults_cpu: Track the nodes the process was running on when a NUMA
880 * hinting fault was incurred.
881 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
882 * during the current scan window. When the scan completes, the counts
883 * in faults_memory and faults_cpu decay and these values are copied.
885 unsigned long *numa_faults;
886 unsigned long total_numa_faults;
889 * numa_faults_locality tracks if faults recorded during the last
890 * scan window were remote/local or failed to migrate. The task scan
891 * period is adapted based on the locality of the faults with different
892 * weights depending on whether they were shared or private faults
894 unsigned long numa_faults_locality[3];
896 unsigned long numa_pages_migrated;
897 #endif /* CONFIG_NUMA_BALANCING */
899 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
900 struct tlbflush_unmap_batch tlb_ubc;
906 * cache last used pipe for splice
908 struct pipe_inode_info *splice_pipe;
910 struct page_frag task_frag;
912 #ifdef CONFIG_TASK_DELAY_ACCT
913 struct task_delay_info *delays;
916 #ifdef CONFIG_FAULT_INJECTION
920 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
921 * balance_dirty_pages() for some dirty throttling pause
924 int nr_dirtied_pause;
925 unsigned long dirty_paused_when; /* start of a write-and-pause period */
927 #ifdef CONFIG_LATENCYTOP
928 int latency_record_count;
929 struct latency_record latency_record[LT_SAVECOUNT];
932 * time slack values; these are used to round up poll() and
933 * select() etc timeout values. These are in nanoseconds.
936 u64 default_timer_slack_ns;
939 unsigned int kasan_depth;
941 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
942 /* Index of current stored address in ret_stack */
944 /* Stack of return addresses for return function tracing */
945 struct ftrace_ret_stack *ret_stack;
946 /* time stamp for last schedule */
947 unsigned long long ftrace_timestamp;
949 * Number of functions that haven't been traced
950 * because of depth overrun.
952 atomic_t trace_overrun;
953 /* Pause for the tracing */
954 atomic_t tracing_graph_pause;
956 #ifdef CONFIG_TRACING
957 /* state flags for use by tracers */
959 /* bitmask and counter of trace recursion */
960 unsigned long trace_recursion;
961 #endif /* CONFIG_TRACING */
963 /* Coverage collection mode enabled for this task (0 if disabled). */
964 enum kcov_mode kcov_mode;
965 /* Size of the kcov_area. */
967 /* Buffer for coverage collection. */
969 /* kcov desciptor wired with this task or NULL. */
973 struct mem_cgroup *memcg_in_oom;
974 gfp_t memcg_oom_gfp_mask;
977 /* number of pages to reclaim on returning to userland */
978 unsigned int memcg_nr_pages_over_high;
980 #ifdef CONFIG_UPROBES
981 struct uprobe_task *utask;
983 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
984 unsigned int sequential_io;
985 unsigned int sequential_io_avg;
987 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
988 unsigned long task_state_change;
990 int pagefault_disabled;
992 struct task_struct *oom_reaper_list;
994 #ifdef CONFIG_VMAP_STACK
995 struct vm_struct *stack_vm_area;
997 #ifdef CONFIG_THREAD_INFO_IN_TASK
998 /* A live task holds one reference. */
999 atomic_t stack_refcount;
1001 /* CPU-specific state of this task */
1002 struct thread_struct thread;
1004 * WARNING: on x86, 'thread_struct' contains a variable-sized
1005 * structure. It *MUST* be at the end of 'task_struct'.
1007 * Do not put anything below here!
1011 static inline struct pid *task_pid(struct task_struct *task)
1013 return task->pids[PIDTYPE_PID].pid;
1016 static inline struct pid *task_tgid(struct task_struct *task)
1018 return task->group_leader->pids[PIDTYPE_PID].pid;
1022 * Without tasklist or rcu lock it is not safe to dereference
1023 * the result of task_pgrp/task_session even if task == current,
1024 * we can race with another thread doing sys_setsid/sys_setpgid.
1026 static inline struct pid *task_pgrp(struct task_struct *task)
1028 return task->group_leader->pids[PIDTYPE_PGID].pid;
1031 static inline struct pid *task_session(struct task_struct *task)
1033 return task->group_leader->pids[PIDTYPE_SID].pid;
1037 * the helpers to get the task's different pids as they are seen
1038 * from various namespaces
1040 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1041 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1043 * task_xid_nr_ns() : id seen from the ns specified;
1045 * set_task_vxid() : assigns a virtual id to a task;
1047 * see also pid_nr() etc in include/linux/pid.h
1049 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1050 struct pid_namespace *ns);
1052 static inline pid_t task_pid_nr(struct task_struct *tsk)
1057 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1058 struct pid_namespace *ns)
1060 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1063 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1065 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1069 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1074 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1076 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1078 return pid_vnr(task_tgid(tsk));
1082 static inline int pid_alive(const struct task_struct *p);
1083 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1089 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1095 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1097 return task_ppid_nr_ns(tsk, &init_pid_ns);
1100 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1101 struct pid_namespace *ns)
1103 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1106 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1108 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1112 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1113 struct pid_namespace *ns)
1115 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1118 static inline pid_t task_session_vnr(struct task_struct *tsk)
1120 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1123 /* obsolete, do not use */
1124 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1126 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1130 * pid_alive - check that a task structure is not stale
1131 * @p: Task structure to be checked.
1133 * Test if a process is not yet dead (at most zombie state)
1134 * If pid_alive fails, then pointers within the task structure
1135 * can be stale and must not be dereferenced.
1137 * Return: 1 if the process is alive. 0 otherwise.
1139 static inline int pid_alive(const struct task_struct *p)
1141 return p->pids[PIDTYPE_PID].pid != NULL;
1145 * is_global_init - check if a task structure is init. Since init
1146 * is free to have sub-threads we need to check tgid.
1147 * @tsk: Task structure to be checked.
1149 * Check if a task structure is the first user space task the kernel created.
1151 * Return: 1 if the task structure is init. 0 otherwise.
1153 static inline int is_global_init(struct task_struct *tsk)
1155 return task_tgid_nr(tsk) == 1;
1158 extern struct pid *cad_pid;
1160 extern void free_task(struct task_struct *tsk);
1161 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
1163 extern void __put_task_struct(struct task_struct *t);
1165 static inline void put_task_struct(struct task_struct *t)
1167 if (atomic_dec_and_test(&t->usage))
1168 __put_task_struct(t);
1171 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
1172 struct task_struct *try_get_task_struct(struct task_struct **ptask);
1177 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1178 #define PF_EXITING 0x00000004 /* getting shut down */
1179 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1180 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1181 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1182 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1183 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1184 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1185 #define PF_DUMPCORE 0x00000200 /* dumped core */
1186 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1187 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1188 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1189 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1190 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1191 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1192 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1193 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1194 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1195 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1196 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1197 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1198 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1199 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1200 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1201 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1202 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1203 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1204 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1207 * Only the _current_ task can read/write to tsk->flags, but other
1208 * tasks can access tsk->flags in readonly mode for example
1209 * with tsk_used_math (like during threaded core dumping).
1210 * There is however an exception to this rule during ptrace
1211 * or during fork: the ptracer task is allowed to write to the
1212 * child->flags of its traced child (same goes for fork, the parent
1213 * can write to the child->flags), because we're guaranteed the
1214 * child is not running and in turn not changing child->flags
1215 * at the same time the parent does it.
1217 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1218 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1219 #define clear_used_math() clear_stopped_child_used_math(current)
1220 #define set_used_math() set_stopped_child_used_math(current)
1221 #define conditional_stopped_child_used_math(condition, child) \
1222 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1223 #define conditional_used_math(condition) \
1224 conditional_stopped_child_used_math(condition, current)
1225 #define copy_to_stopped_child_used_math(child) \
1226 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1227 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1228 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1229 #define used_math() tsk_used_math(current)
1231 /* Per-process atomic flags. */
1232 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1233 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1234 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1235 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1238 #define TASK_PFA_TEST(name, func) \
1239 static inline bool task_##func(struct task_struct *p) \
1240 { return test_bit(PFA_##name, &p->atomic_flags); }
1241 #define TASK_PFA_SET(name, func) \
1242 static inline void task_set_##func(struct task_struct *p) \
1243 { set_bit(PFA_##name, &p->atomic_flags); }
1244 #define TASK_PFA_CLEAR(name, func) \
1245 static inline void task_clear_##func(struct task_struct *p) \
1246 { clear_bit(PFA_##name, &p->atomic_flags); }
1248 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1249 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1251 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1252 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1253 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1255 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1256 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1257 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1259 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1260 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1262 static inline void tsk_restore_flags(struct task_struct *task,
1263 unsigned long orig_flags, unsigned long flags)
1265 task->flags &= ~flags;
1266 task->flags |= orig_flags & flags;
1269 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1270 const struct cpumask *trial);
1271 extern int task_can_attach(struct task_struct *p,
1272 const struct cpumask *cs_cpus_allowed);
1274 extern void do_set_cpus_allowed(struct task_struct *p,
1275 const struct cpumask *new_mask);
1277 extern int set_cpus_allowed_ptr(struct task_struct *p,
1278 const struct cpumask *new_mask);
1280 static inline void do_set_cpus_allowed(struct task_struct *p,
1281 const struct cpumask *new_mask)
1284 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1285 const struct cpumask *new_mask)
1287 if (!cpumask_test_cpu(0, new_mask))
1293 #ifndef cpu_relax_yield
1294 #define cpu_relax_yield() cpu_relax()
1297 /* sched_exec is called by processes performing an exec */
1299 extern void sched_exec(void);
1301 #define sched_exec() {}
1304 extern int yield_to(struct task_struct *p, bool preempt);
1305 extern void set_user_nice(struct task_struct *p, long nice);
1306 extern int task_prio(const struct task_struct *p);
1308 * task_nice - return the nice value of a given task.
1309 * @p: the task in question.
1311 * Return: The nice value [ -20 ... 0 ... 19 ].
1313 static inline int task_nice(const struct task_struct *p)
1315 return PRIO_TO_NICE((p)->static_prio);
1317 extern int can_nice(const struct task_struct *p, const int nice);
1318 extern int task_curr(const struct task_struct *p);
1319 extern int idle_cpu(int cpu);
1320 extern int sched_setscheduler(struct task_struct *, int,
1321 const struct sched_param *);
1322 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1323 const struct sched_param *);
1324 extern int sched_setattr(struct task_struct *,
1325 const struct sched_attr *);
1326 extern struct task_struct *idle_task(int cpu);
1328 * is_idle_task - is the specified task an idle task?
1329 * @p: the task in question.
1331 * Return: 1 if @p is an idle task. 0 otherwise.
1333 static inline bool is_idle_task(const struct task_struct *p)
1335 return !!(p->flags & PF_IDLE);
1337 extern struct task_struct *curr_task(int cpu);
1338 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1342 union thread_union {
1343 #ifndef CONFIG_THREAD_INFO_IN_TASK
1344 struct thread_info thread_info;
1346 unsigned long stack[THREAD_SIZE/sizeof(long)];
1349 #ifdef CONFIG_THREAD_INFO_IN_TASK
1350 static inline struct thread_info *task_thread_info(struct task_struct *task)
1352 return &task->thread_info;
1354 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1355 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1358 extern struct pid_namespace init_pid_ns;
1361 * find a task by one of its numerical ids
1363 * find_task_by_pid_ns():
1364 * finds a task by its pid in the specified namespace
1365 * find_task_by_vpid():
1366 * finds a task by its virtual pid
1368 * see also find_vpid() etc in include/linux/pid.h
1371 extern struct task_struct *find_task_by_vpid(pid_t nr);
1372 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1373 struct pid_namespace *ns);
1375 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1376 extern int wake_up_process(struct task_struct *tsk);
1377 extern void wake_up_new_task(struct task_struct *tsk);
1379 extern void kick_process(struct task_struct *tsk);
1381 static inline void kick_process(struct task_struct *tsk) { }
1384 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1385 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1387 __set_task_comm(tsk, from, false);
1389 extern char *get_task_comm(char *to, struct task_struct *tsk);
1392 void scheduler_ipi(void);
1393 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1395 static inline void scheduler_ipi(void) { }
1396 static inline unsigned long wait_task_inactive(struct task_struct *p,
1403 /* set thread flags in other task's structures
1404 * - see asm/thread_info.h for TIF_xxxx flags available
1406 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1408 set_ti_thread_flag(task_thread_info(tsk), flag);
1411 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1413 clear_ti_thread_flag(task_thread_info(tsk), flag);
1416 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1418 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1421 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1423 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1426 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1428 return test_ti_thread_flag(task_thread_info(tsk), flag);
1431 static inline void set_tsk_need_resched(struct task_struct *tsk)
1433 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1436 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1438 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1441 static inline int test_tsk_need_resched(struct task_struct *tsk)
1443 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1447 * cond_resched() and cond_resched_lock(): latency reduction via
1448 * explicit rescheduling in places that are safe. The return
1449 * value indicates whether a reschedule was done in fact.
1450 * cond_resched_lock() will drop the spinlock before scheduling,
1451 * cond_resched_softirq() will enable bhs before scheduling.
1453 #ifndef CONFIG_PREEMPT
1454 extern int _cond_resched(void);
1456 static inline int _cond_resched(void) { return 0; }
1459 #define cond_resched() ({ \
1460 ___might_sleep(__FILE__, __LINE__, 0); \
1464 extern int __cond_resched_lock(spinlock_t *lock);
1466 #define cond_resched_lock(lock) ({ \
1467 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1468 __cond_resched_lock(lock); \
1471 extern int __cond_resched_softirq(void);
1473 #define cond_resched_softirq() ({ \
1474 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1475 __cond_resched_softirq(); \
1478 static inline void cond_resched_rcu(void)
1480 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1488 * Does a critical section need to be broken due to another
1489 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1490 * but a general need for low latency)
1492 static inline int spin_needbreak(spinlock_t *lock)
1494 #ifdef CONFIG_PREEMPT
1495 return spin_is_contended(lock);
1501 static __always_inline bool need_resched(void)
1503 return unlikely(tif_need_resched());
1507 * Wrappers for p->thread_info->cpu access. No-op on UP.
1511 static inline unsigned int task_cpu(const struct task_struct *p)
1513 #ifdef CONFIG_THREAD_INFO_IN_TASK
1516 return task_thread_info(p)->cpu;
1520 static inline int task_node(const struct task_struct *p)
1522 return cpu_to_node(task_cpu(p));
1525 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1529 static inline unsigned int task_cpu(const struct task_struct *p)
1534 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1538 #endif /* CONFIG_SMP */
1541 * In order to reduce various lock holder preemption latencies provide an
1542 * interface to see if a vCPU is currently running or not.
1544 * This allows us to terminate optimistic spin loops and block, analogous to
1545 * the native optimistic spin heuristic of testing if the lock owner task is
1548 #ifndef vcpu_is_preempted
1549 # define vcpu_is_preempted(cpu) false
1552 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1553 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1555 extern int task_can_switch_user(struct user_struct *up,
1556 struct task_struct *tsk);
1558 #ifndef TASK_SIZE_OF
1559 #define TASK_SIZE_OF(tsk) TASK_SIZE