4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
13 #include <asm/param.h> /* for HZ */
15 #include <linux/capability.h>
16 #include <linux/threads.h>
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/timex.h>
20 #include <linux/jiffies.h>
21 #include <linux/plist.h>
22 #include <linux/rbtree.h>
23 #include <linux/thread_info.h>
24 #include <linux/cpumask.h>
25 #include <linux/errno.h>
26 #include <linux/nodemask.h>
27 #include <linux/mm_types.h>
28 #include <linux/preempt.h>
31 #include <asm/ptrace.h>
32 #include <linux/cputime.h>
34 #include <linux/smp.h>
35 #include <linux/sem.h>
36 #include <linux/shm.h>
37 #include <linux/signal.h>
38 #include <linux/compiler.h>
39 #include <linux/completion.h>
40 #include <linux/pid.h>
41 #include <linux/percpu.h>
42 #include <linux/topology.h>
43 #include <linux/proportions.h>
44 #include <linux/seccomp.h>
45 #include <linux/rcupdate.h>
46 #include <linux/rculist.h>
47 #include <linux/rtmutex.h>
49 #include <linux/time.h>
50 #include <linux/param.h>
51 #include <linux/resource.h>
52 #include <linux/timer.h>
53 #include <linux/hrtimer.h>
54 #include <linux/kcov.h>
55 #include <linux/task_io_accounting.h>
56 #include <linux/latencytop.h>
57 #include <linux/cred.h>
58 #include <linux/llist.h>
59 #include <linux/uidgid.h>
60 #include <linux/gfp.h>
61 #include <linux/magic.h>
62 #include <linux/cgroup-defs.h>
64 #include <asm/processor.h>
66 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
69 * Extended scheduling parameters data structure.
71 * This is needed because the original struct sched_param can not be
72 * altered without introducing ABI issues with legacy applications
73 * (e.g., in sched_getparam()).
75 * However, the possibility of specifying more than just a priority for
76 * the tasks may be useful for a wide variety of application fields, e.g.,
77 * multimedia, streaming, automation and control, and many others.
79 * This variant (sched_attr) is meant at describing a so-called
80 * sporadic time-constrained task. In such model a task is specified by:
81 * - the activation period or minimum instance inter-arrival time;
82 * - the maximum (or average, depending on the actual scheduling
83 * discipline) computation time of all instances, a.k.a. runtime;
84 * - the deadline (relative to the actual activation time) of each
86 * Very briefly, a periodic (sporadic) task asks for the execution of
87 * some specific computation --which is typically called an instance--
88 * (at most) every period. Moreover, each instance typically lasts no more
89 * than the runtime and must be completed by time instant t equal to
90 * the instance activation time + the deadline.
92 * This is reflected by the actual fields of the sched_attr structure:
94 * @size size of the structure, for fwd/bwd compat.
96 * @sched_policy task's scheduling policy
97 * @sched_flags for customizing the scheduler behaviour
98 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
99 * @sched_priority task's static priority (SCHED_FIFO/RR)
100 * @sched_deadline representative of the task's deadline
101 * @sched_runtime representative of the task's runtime
102 * @sched_period representative of the task's period
104 * Given this task model, there are a multiplicity of scheduling algorithms
105 * and policies, that can be used to ensure all the tasks will make their
106 * timing constraints.
108 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
109 * only user of this new interface. More information about the algorithm
110 * available in the scheduling class file or in Documentation/.
118 /* SCHED_NORMAL, SCHED_BATCH */
121 /* SCHED_FIFO, SCHED_RR */
130 struct futex_pi_state;
131 struct robust_list_head;
134 struct perf_event_context;
139 #define VMACACHE_BITS 2
140 #define VMACACHE_SIZE (1U << VMACACHE_BITS)
141 #define VMACACHE_MASK (VMACACHE_SIZE - 1)
144 * These are the constant used to fake the fixed-point load-average
145 * counting. Some notes:
146 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
147 * a load-average precision of 10 bits integer + 11 bits fractional
148 * - if you want to count load-averages more often, you need more
149 * precision, or rounding will get you. With 2-second counting freq,
150 * the EXP_n values would be 1981, 2034 and 2043 if still using only
153 extern unsigned long avenrun[]; /* Load averages */
154 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
156 #define FSHIFT 11 /* nr of bits of precision */
157 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
158 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
159 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
160 #define EXP_5 2014 /* 1/exp(5sec/5min) */
161 #define EXP_15 2037 /* 1/exp(5sec/15min) */
163 #define CALC_LOAD(load,exp,n) \
165 load += n*(FIXED_1-exp); \
168 extern unsigned long total_forks;
169 extern int nr_threads;
170 DECLARE_PER_CPU(unsigned long, process_counts);
171 extern int nr_processes(void);
172 extern unsigned long nr_running(void);
173 extern bool single_task_running(void);
174 extern unsigned long nr_iowait(void);
175 extern unsigned long nr_iowait_cpu(int cpu);
176 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
177 #ifdef CONFIG_CPU_QUIET
178 extern u64 nr_running_integral(unsigned int cpu);
181 extern void sched_update_nr_prod(int cpu, long delta, bool inc);
182 extern void sched_get_nr_running_avg(int *avg, int *iowait_avg, int *big_avg,
183 unsigned int *max_nr,
184 unsigned int *big_max_nr);
185 extern u64 sched_get_cpu_last_busy_time(int cpu);
188 extern u32 sched_get_wake_up_idle(struct task_struct *p);
189 extern int sched_set_wake_up_idle(struct task_struct *p, int wake_up_idle);
191 static inline u32 sched_get_wake_up_idle(struct task_struct *p)
196 static inline int sched_set_wake_up_idle(struct task_struct *p,
201 #endif /* CONFIG_SMP */
203 extern void calc_global_load(unsigned long ticks);
205 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
206 extern void update_cpu_load_nohz(void);
208 static inline void update_cpu_load_nohz(void) { }
211 extern unsigned long get_parent_ip(unsigned long addr);
213 extern void dump_cpu_task(int cpu);
218 #ifdef CONFIG_SCHED_DEBUG
219 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
220 extern void proc_sched_set_task(struct task_struct *p);
224 * Task state bitmask. NOTE! These bits are also
225 * encoded in fs/proc/array.c: get_task_state().
227 * We have two separate sets of flags: task->state
228 * is about runnability, while task->exit_state are
229 * about the task exiting. Confusing, but this way
230 * modifying one set can't modify the other one by
233 #define TASK_RUNNING 0
234 #define TASK_INTERRUPTIBLE 1
235 #define TASK_UNINTERRUPTIBLE 2
236 #define __TASK_STOPPED 4
237 #define __TASK_TRACED 8
238 /* in tsk->exit_state */
240 #define EXIT_ZOMBIE 32
241 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
242 /* in tsk->state again */
244 #define TASK_WAKEKILL 128
245 #define TASK_WAKING 256
246 #define TASK_PARKED 512
247 #define TASK_NOLOAD 1024
248 #define TASK_NEW 2048
249 #define TASK_STATE_MAX 4096
251 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
253 extern char ___assert_task_state[1 - 2*!!(
254 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
256 /* Convenience macros for the sake of set_task_state */
257 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
258 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
259 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
261 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
263 /* Convenience macros for the sake of wake_up */
264 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
265 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
267 /* get_task_state() */
268 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
269 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
270 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
272 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
273 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
274 #define task_is_stopped_or_traced(task) \
275 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
276 #define task_contributes_to_load(task) \
277 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
278 (task->flags & PF_FROZEN) == 0 && \
279 (task->state & TASK_NOLOAD) == 0)
281 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
283 #define __set_task_state(tsk, state_value) \
285 (tsk)->task_state_change = _THIS_IP_; \
286 (tsk)->state = (state_value); \
288 #define set_task_state(tsk, state_value) \
290 (tsk)->task_state_change = _THIS_IP_; \
291 smp_store_mb((tsk)->state, (state_value)); \
295 * set_current_state() includes a barrier so that the write of current->state
296 * is correctly serialised wrt the caller's subsequent test of whether to
299 * set_current_state(TASK_UNINTERRUPTIBLE);
300 * if (do_i_need_to_sleep())
303 * If the caller does not need such serialisation then use __set_current_state()
305 #define __set_current_state(state_value) \
307 current->task_state_change = _THIS_IP_; \
308 current->state = (state_value); \
310 #define set_current_state(state_value) \
312 current->task_state_change = _THIS_IP_; \
313 smp_store_mb(current->state, (state_value)); \
318 #define __set_task_state(tsk, state_value) \
319 do { (tsk)->state = (state_value); } while (0)
320 #define set_task_state(tsk, state_value) \
321 smp_store_mb((tsk)->state, (state_value))
324 * set_current_state() includes a barrier so that the write of current->state
325 * is correctly serialised wrt the caller's subsequent test of whether to
328 * set_current_state(TASK_UNINTERRUPTIBLE);
329 * if (do_i_need_to_sleep())
332 * If the caller does not need such serialisation then use __set_current_state()
334 #define __set_current_state(state_value) \
335 do { current->state = (state_value); } while (0)
336 #define set_current_state(state_value) \
337 smp_store_mb(current->state, (state_value))
341 /* Task command name length */
342 #define TASK_COMM_LEN 16
344 extern const char *sched_window_reset_reasons[];
355 /* Note: this need to be in sync with migrate_type_names array */
361 #include <linux/spinlock.h>
364 * This serializes "schedule()" and also protects
365 * the run-queue from deletions/modifications (but
366 * _adding_ to the beginning of the run-queue has
369 extern rwlock_t tasklist_lock;
370 extern spinlock_t mmlist_lock;
374 #ifdef CONFIG_PROVE_RCU
375 extern int lockdep_tasklist_lock_is_held(void);
376 #endif /* #ifdef CONFIG_PROVE_RCU */
378 extern void sched_init(void);
379 extern void sched_init_smp(void);
380 extern asmlinkage void schedule_tail(struct task_struct *prev);
381 extern void init_idle(struct task_struct *idle, int cpu);
382 extern void init_idle_bootup_task(struct task_struct *idle);
384 extern cpumask_var_t cpu_isolated_map;
386 extern int runqueue_is_locked(int cpu);
388 #ifdef CONFIG_HOTPLUG_CPU
389 extern int sched_isolate_count(const cpumask_t *mask, bool include_offline);
390 extern int sched_isolate_cpu(int cpu);
391 extern int sched_unisolate_cpu(int cpu);
392 extern int sched_unisolate_cpu_unlocked(int cpu);
394 static inline int sched_isolate_count(const cpumask_t *mask,
395 bool include_offline)
397 cpumask_t count_mask;
400 cpumask_andnot(&count_mask, mask, cpu_online_mask);
404 return cpumask_weight(&count_mask);
407 static inline int sched_isolate_cpu(int cpu)
412 static inline int sched_unisolate_cpu(int cpu)
417 static inline int sched_unisolate_cpu_unlocked(int cpu)
423 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
424 extern void nohz_balance_enter_idle(int cpu);
425 extern void set_cpu_sd_state_idle(void);
426 extern int get_nohz_timer_target(void);
428 static inline void nohz_balance_enter_idle(int cpu) { }
429 static inline void set_cpu_sd_state_idle(void) { }
433 * Only dump TASK_* tasks. (0 for all tasks)
435 extern void show_state_filter(unsigned long state_filter);
437 static inline void show_state(void)
439 show_state_filter(0);
442 extern void show_regs(struct pt_regs *);
445 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
446 * task), SP is the stack pointer of the first frame that should be shown in the back
447 * trace (or NULL if the entire call-chain of the task should be shown).
449 extern void show_stack(struct task_struct *task, unsigned long *sp);
451 extern void cpu_init (void);
452 extern void trap_init(void);
453 extern void update_process_times(int user);
454 extern void scheduler_tick(void);
456 extern void sched_show_task(struct task_struct *p);
458 #ifdef CONFIG_LOCKUP_DETECTOR
459 extern void touch_softlockup_watchdog_sched(void);
460 extern void touch_softlockup_watchdog(void);
461 extern void touch_softlockup_watchdog_sync(void);
462 extern void touch_all_softlockup_watchdogs(void);
463 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
465 size_t *lenp, loff_t *ppos);
466 extern unsigned int softlockup_panic;
467 extern unsigned int hardlockup_panic;
468 void lockup_detector_init(void);
469 extern void watchdog_enable(unsigned int cpu);
470 extern void watchdog_disable(unsigned int cpu);
471 extern bool watchdog_configured(unsigned int cpu);
473 static inline void touch_softlockup_watchdog_sched(void)
476 static inline void touch_softlockup_watchdog(void)
479 static inline void touch_softlockup_watchdog_sync(void)
482 static inline void touch_all_softlockup_watchdogs(void)
485 static inline void lockup_detector_init(void)
488 static inline void watchdog_enable(unsigned int cpu)
491 static inline void watchdog_disable(unsigned int cpu)
494 static inline bool watchdog_configured(unsigned int cpu)
497 * Predend the watchdog is always configured.
498 * We will be waiting for the watchdog to be enabled in core isolation
504 #ifdef CONFIG_DETECT_HUNG_TASK
505 void reset_hung_task_detector(void);
507 static inline void reset_hung_task_detector(void)
512 /* Attach to any functions which should be ignored in wchan output. */
513 #define __sched __attribute__((__section__(".sched.text")))
515 /* Linker adds these: start and end of __sched functions */
516 extern char __sched_text_start[], __sched_text_end[];
518 /* Is this address in the __sched functions? */
519 extern int in_sched_functions(unsigned long addr);
521 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
522 extern signed long schedule_timeout(signed long timeout);
523 extern signed long schedule_timeout_interruptible(signed long timeout);
524 extern signed long schedule_timeout_killable(signed long timeout);
525 extern signed long schedule_timeout_uninterruptible(signed long timeout);
526 asmlinkage void schedule(void);
527 extern void schedule_preempt_disabled(void);
529 extern long io_schedule_timeout(long timeout);
531 static inline void io_schedule(void)
533 io_schedule_timeout(MAX_SCHEDULE_TIMEOUT);
537 struct user_namespace;
540 extern void arch_pick_mmap_layout(struct mm_struct *mm);
542 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
543 unsigned long, unsigned long);
545 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
546 unsigned long len, unsigned long pgoff,
547 unsigned long flags);
549 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
552 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
553 #define SUID_DUMP_USER 1 /* Dump as user of process */
554 #define SUID_DUMP_ROOT 2 /* Dump as root */
558 /* for SUID_DUMP_* above */
559 #define MMF_DUMPABLE_BITS 2
560 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
562 extern void set_dumpable(struct mm_struct *mm, int value);
564 * This returns the actual value of the suid_dumpable flag. For things
565 * that are using this for checking for privilege transitions, it must
566 * test against SUID_DUMP_USER rather than treating it as a boolean
569 static inline int __get_dumpable(unsigned long mm_flags)
571 return mm_flags & MMF_DUMPABLE_MASK;
574 static inline int get_dumpable(struct mm_struct *mm)
576 return __get_dumpable(mm->flags);
579 /* coredump filter bits */
580 #define MMF_DUMP_ANON_PRIVATE 2
581 #define MMF_DUMP_ANON_SHARED 3
582 #define MMF_DUMP_MAPPED_PRIVATE 4
583 #define MMF_DUMP_MAPPED_SHARED 5
584 #define MMF_DUMP_ELF_HEADERS 6
585 #define MMF_DUMP_HUGETLB_PRIVATE 7
586 #define MMF_DUMP_HUGETLB_SHARED 8
587 #define MMF_DUMP_DAX_PRIVATE 9
588 #define MMF_DUMP_DAX_SHARED 10
590 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
591 #define MMF_DUMP_FILTER_BITS 9
592 #define MMF_DUMP_FILTER_MASK \
593 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
594 #define MMF_DUMP_FILTER_DEFAULT \
595 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
596 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
598 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
599 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
601 # define MMF_DUMP_MASK_DEFAULT_ELF 0
603 /* leave room for more dump flags */
604 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
605 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
606 #define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
608 #define MMF_HAS_UPROBES 19 /* has uprobes */
609 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
611 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
613 struct sighand_struct {
615 struct k_sigaction action[_NSIG];
617 wait_queue_head_t signalfd_wqh;
620 struct pacct_struct {
623 unsigned long ac_mem;
624 cputime_t ac_utime, ac_stime;
625 unsigned long ac_minflt, ac_majflt;
636 * struct prev_cputime - snaphsot of system and user cputime
637 * @utime: time spent in user mode
638 * @stime: time spent in system mode
639 * @lock: protects the above two fields
641 * Stores previous user/system time values such that we can guarantee
644 struct prev_cputime {
645 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
652 static inline void prev_cputime_init(struct prev_cputime *prev)
654 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
655 prev->utime = prev->stime = 0;
656 raw_spin_lock_init(&prev->lock);
661 * struct task_cputime - collected CPU time counts
662 * @utime: time spent in user mode, in &cputime_t units
663 * @stime: time spent in kernel mode, in &cputime_t units
664 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
666 * This structure groups together three kinds of CPU time that are tracked for
667 * threads and thread groups. Most things considering CPU time want to group
668 * these counts together and treat all three of them in parallel.
670 struct task_cputime {
673 unsigned long long sum_exec_runtime;
676 /* Alternate field names when used to cache expirations. */
677 #define virt_exp utime
678 #define prof_exp stime
679 #define sched_exp sum_exec_runtime
681 #define INIT_CPUTIME \
682 (struct task_cputime) { \
685 .sum_exec_runtime = 0, \
689 * This is the atomic variant of task_cputime, which can be used for
690 * storing and updating task_cputime statistics without locking.
692 struct task_cputime_atomic {
695 atomic64_t sum_exec_runtime;
698 #define INIT_CPUTIME_ATOMIC \
699 (struct task_cputime_atomic) { \
700 .utime = ATOMIC64_INIT(0), \
701 .stime = ATOMIC64_INIT(0), \
702 .sum_exec_runtime = ATOMIC64_INIT(0), \
705 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
708 * Disable preemption until the scheduler is running -- use an unconditional
709 * value so that it also works on !PREEMPT_COUNT kernels.
711 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
713 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
716 * Initial preempt_count value; reflects the preempt_count schedule invariant
717 * which states that during context switches:
719 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
721 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
722 * Note: See finish_task_switch().
724 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
727 * struct thread_group_cputimer - thread group interval timer counts
728 * @cputime_atomic: atomic thread group interval timers.
729 * @running: true when there are timers running and
730 * @cputime_atomic receives updates.
731 * @checking_timer: true when a thread in the group is in the
732 * process of checking for thread group timers.
734 * This structure contains the version of task_cputime, above, that is
735 * used for thread group CPU timer calculations.
737 struct thread_group_cputimer {
738 struct task_cputime_atomic cputime_atomic;
743 #include <linux/rwsem.h>
747 * NOTE! "signal_struct" does not have its own
748 * locking, because a shared signal_struct always
749 * implies a shared sighand_struct, so locking
750 * sighand_struct is always a proper superset of
751 * the locking of signal_struct.
753 struct signal_struct {
757 struct list_head thread_head;
759 wait_queue_head_t wait_chldexit; /* for wait4() */
761 /* current thread group signal load-balancing target: */
762 struct task_struct *curr_target;
764 /* shared signal handling: */
765 struct sigpending shared_pending;
767 /* thread group exit support */
770 * - notify group_exit_task when ->count is equal to notify_count
771 * - everyone except group_exit_task is stopped during signal delivery
772 * of fatal signals, group_exit_task processes the signal.
775 struct task_struct *group_exit_task;
777 /* thread group stop support, overloads group_exit_code too */
778 int group_stop_count;
779 unsigned int flags; /* see SIGNAL_* flags below */
782 * PR_SET_CHILD_SUBREAPER marks a process, like a service
783 * manager, to re-parent orphan (double-forking) child processes
784 * to this process instead of 'init'. The service manager is
785 * able to receive SIGCHLD signals and is able to investigate
786 * the process until it calls wait(). All children of this
787 * process will inherit a flag if they should look for a
788 * child_subreaper process at exit.
790 unsigned int is_child_subreaper:1;
791 unsigned int has_child_subreaper:1;
793 /* POSIX.1b Interval Timers */
795 struct list_head posix_timers;
797 /* ITIMER_REAL timer for the process */
798 struct hrtimer real_timer;
799 struct pid *leader_pid;
800 ktime_t it_real_incr;
803 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
804 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
805 * values are defined to 0 and 1 respectively
807 struct cpu_itimer it[2];
810 * Thread group totals for process CPU timers.
811 * See thread_group_cputimer(), et al, for details.
813 struct thread_group_cputimer cputimer;
815 /* Earliest-expiration cache. */
816 struct task_cputime cputime_expires;
818 struct list_head cpu_timers[3];
820 struct pid *tty_old_pgrp;
822 /* boolean value for session group leader */
825 struct tty_struct *tty; /* NULL if no tty */
827 #ifdef CONFIG_SCHED_AUTOGROUP
828 struct autogroup *autogroup;
831 * Cumulative resource counters for dead threads in the group,
832 * and for reaped dead child processes forked by this group.
833 * Live threads maintain their own counters and add to these
834 * in __exit_signal, except for the group leader.
836 seqlock_t stats_lock;
837 cputime_t utime, stime, cutime, cstime;
840 struct prev_cputime prev_cputime;
841 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
842 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
843 unsigned long inblock, oublock, cinblock, coublock;
844 unsigned long maxrss, cmaxrss;
845 struct task_io_accounting ioac;
848 * Cumulative ns of schedule CPU time fo dead threads in the
849 * group, not including a zombie group leader, (This only differs
850 * from jiffies_to_ns(utime + stime) if sched_clock uses something
851 * other than jiffies.)
853 unsigned long long sum_sched_runtime;
856 * We don't bother to synchronize most readers of this at all,
857 * because there is no reader checking a limit that actually needs
858 * to get both rlim_cur and rlim_max atomically, and either one
859 * alone is a single word that can safely be read normally.
860 * getrlimit/setrlimit use task_lock(current->group_leader) to
861 * protect this instead of the siglock, because they really
862 * have no need to disable irqs.
864 struct rlimit rlim[RLIM_NLIMITS];
866 #ifdef CONFIG_BSD_PROCESS_ACCT
867 struct pacct_struct pacct; /* per-process accounting information */
869 #ifdef CONFIG_TASKSTATS
870 struct taskstats *stats;
874 unsigned audit_tty_log_passwd;
875 struct tty_audit_buf *tty_audit_buf;
878 oom_flags_t oom_flags;
879 short oom_score_adj; /* OOM kill score adjustment */
880 short oom_score_adj_min; /* OOM kill score adjustment min value.
881 * Only settable by CAP_SYS_RESOURCE. */
883 struct mutex cred_guard_mutex; /* guard against foreign influences on
884 * credential calculations
885 * (notably. ptrace) */
889 * Bits in flags field of signal_struct.
891 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
892 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
893 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
894 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
896 * Pending notifications to parent.
898 #define SIGNAL_CLD_STOPPED 0x00000010
899 #define SIGNAL_CLD_CONTINUED 0x00000020
900 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
902 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
904 #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
905 SIGNAL_STOP_CONTINUED)
907 static inline void signal_set_stop_flags(struct signal_struct *sig,
910 WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
911 sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
914 /* If true, all threads except ->group_exit_task have pending SIGKILL */
915 static inline int signal_group_exit(const struct signal_struct *sig)
917 return (sig->flags & SIGNAL_GROUP_EXIT) ||
918 (sig->group_exit_task != NULL);
922 * Some day this will be a full-fledged user tracking system..
925 atomic_t __count; /* reference count */
926 atomic_t processes; /* How many processes does this user have? */
927 atomic_t sigpending; /* How many pending signals does this user have? */
928 #ifdef CONFIG_INOTIFY_USER
929 atomic_t inotify_watches; /* How many inotify watches does this user have? */
930 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
932 #ifdef CONFIG_FANOTIFY
933 atomic_t fanotify_listeners;
936 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
938 #ifdef CONFIG_POSIX_MQUEUE
939 /* protected by mq_lock */
940 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
942 unsigned long locked_shm; /* How many pages of mlocked shm ? */
943 unsigned long unix_inflight; /* How many files in flight in unix sockets */
944 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
947 struct key *uid_keyring; /* UID specific keyring */
948 struct key *session_keyring; /* UID's default session keyring */
951 /* Hash table maintenance information */
952 struct hlist_node uidhash_node;
955 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
956 atomic_long_t locked_vm;
960 extern int uids_sysfs_init(void);
962 extern struct user_struct *find_user(kuid_t);
964 extern struct user_struct root_user;
965 #define INIT_USER (&root_user)
968 struct backing_dev_info;
969 struct reclaim_state;
971 #ifdef CONFIG_SCHED_INFO
973 /* cumulative counters */
974 unsigned long pcount; /* # of times run on this cpu */
975 unsigned long long run_delay; /* time spent waiting on a runqueue */
978 unsigned long long last_arrival,/* when we last ran on a cpu */
979 last_queued; /* when we were last queued to run */
981 #endif /* CONFIG_SCHED_INFO */
983 #ifdef CONFIG_TASK_DELAY_ACCT
984 struct task_delay_info {
986 unsigned int flags; /* Private per-task flags */
988 /* For each stat XXX, add following, aligned appropriately
990 * struct timespec XXX_start, XXX_end;
994 * Atomicity of updates to XXX_delay, XXX_count protected by
995 * single lock above (split into XXX_lock if contention is an issue).
999 * XXX_count is incremented on every XXX operation, the delay
1000 * associated with the operation is added to XXX_delay.
1001 * XXX_delay contains the accumulated delay time in nanoseconds.
1003 u64 blkio_start; /* Shared by blkio, swapin */
1004 u64 blkio_delay; /* wait for sync block io completion */
1005 u64 swapin_delay; /* wait for swapin block io completion */
1006 u32 blkio_count; /* total count of the number of sync block */
1007 /* io operations performed */
1008 u32 swapin_count; /* total count of the number of swapin block */
1009 /* io operations performed */
1011 u64 freepages_start;
1012 u64 freepages_delay; /* wait for memory reclaim */
1013 u32 freepages_count; /* total count of memory reclaim */
1015 #endif /* CONFIG_TASK_DELAY_ACCT */
1017 static inline int sched_info_on(void)
1019 #ifdef CONFIG_SCHEDSTATS
1021 #elif defined(CONFIG_TASK_DELAY_ACCT)
1022 extern int delayacct_on;
1023 return delayacct_on;
1029 enum cpu_idle_type {
1037 * Increase resolution of cpu_capacity calculations
1039 #define SCHED_CAPACITY_SHIFT 10
1040 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
1042 struct sched_capacity_reqs {
1047 unsigned long total;
1051 * Wake-queues are lists of tasks with a pending wakeup, whose
1052 * callers have already marked the task as woken internally,
1053 * and can thus carry on. A common use case is being able to
1054 * do the wakeups once the corresponding user lock as been
1057 * We hold reference to each task in the list across the wakeup,
1058 * thus guaranteeing that the memory is still valid by the time
1059 * the actual wakeups are performed in wake_up_q().
1061 * One per task suffices, because there's never a need for a task to be
1062 * in two wake queues simultaneously; it is forbidden to abandon a task
1063 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
1064 * already in a wake queue, the wakeup will happen soon and the second
1065 * waker can just skip it.
1067 * The WAKE_Q macro declares and initializes the list head.
1068 * wake_up_q() does NOT reinitialize the list; it's expected to be
1069 * called near the end of a function, where the fact that the queue is
1070 * not used again will be easy to see by inspection.
1072 * Note that this can cause spurious wakeups. schedule() callers
1073 * must ensure the call is done inside a loop, confirming that the
1074 * wakeup condition has in fact occurred.
1076 struct wake_q_node {
1077 struct wake_q_node *next;
1080 struct wake_q_head {
1081 struct wake_q_node *first;
1082 struct wake_q_node **lastp;
1086 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1088 #define WAKE_Q(name) \
1089 struct wake_q_head name = { WAKE_Q_TAIL, &name.first, 0 }
1091 extern void wake_q_add(struct wake_q_head *head,
1092 struct task_struct *task);
1093 extern void wake_up_q(struct wake_q_head *head);
1096 * sched-domains (multiprocessor balancing) declarations:
1099 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1100 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1101 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1102 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1103 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1104 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1105 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1106 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1107 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1108 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1109 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1110 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1111 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1112 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1113 #define SD_NUMA 0x4000 /* cross-node balancing */
1114 #define SD_SHARE_CAP_STATES 0x8000 /* Domain members share capacity state */
1116 #ifdef CONFIG_SCHED_SMT
1117 static inline int cpu_smt_flags(void)
1119 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1123 #ifdef CONFIG_SCHED_MC
1124 static inline int cpu_core_flags(void)
1126 return SD_SHARE_PKG_RESOURCES;
1131 static inline int cpu_numa_flags(void)
1137 struct sched_domain_attr {
1138 int relax_domain_level;
1141 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1142 .relax_domain_level = -1, \
1145 extern int sched_domain_level_max;
1147 struct capacity_state {
1148 unsigned long cap; /* compute capacity */
1149 unsigned long power; /* power consumption at this compute capacity */
1153 unsigned long power; /* power consumption in this idle state */
1156 struct sched_group_energy {
1157 unsigned int nr_idle_states; /* number of idle states */
1158 struct idle_state *idle_states; /* ptr to idle state array */
1159 unsigned int nr_cap_states; /* number of capacity states */
1160 struct capacity_state *cap_states; /* ptr to capacity state array */
1163 unsigned long capacity_curr_of(int cpu);
1168 /* select_idle_sibling() stats */
1171 u64 sis_cache_affine;
1176 /* select_energy_cpu_brute() stats */
1180 u64 secb_insuff_cap;
1181 u64 secb_no_nrg_sav;
1185 /* find_best_target() stats */
1193 /* select_task_rq_fair() stats */
1198 struct sched_domain {
1199 /* These fields must be setup */
1200 struct sched_domain *parent; /* top domain must be null terminated */
1201 struct sched_domain *child; /* bottom domain must be null terminated */
1202 struct sched_group *groups; /* the balancing groups of the domain */
1203 unsigned long min_interval; /* Minimum balance interval ms */
1204 unsigned long max_interval; /* Maximum balance interval ms */
1205 unsigned int busy_factor; /* less balancing by factor if busy */
1206 unsigned int imbalance_pct; /* No balance until over watermark */
1207 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1208 unsigned int busy_idx;
1209 unsigned int idle_idx;
1210 unsigned int newidle_idx;
1211 unsigned int wake_idx;
1212 unsigned int forkexec_idx;
1213 unsigned int smt_gain;
1215 int nohz_idle; /* NOHZ IDLE status */
1216 int flags; /* See SD_* */
1219 /* Runtime fields. */
1220 unsigned long last_balance; /* init to jiffies. units in jiffies */
1221 unsigned int balance_interval; /* initialise to 1. units in ms. */
1222 unsigned int nr_balance_failed; /* initialise to 0 */
1224 /* idle_balance() stats */
1225 u64 max_newidle_lb_cost;
1226 unsigned long next_decay_max_lb_cost;
1228 #ifdef CONFIG_SCHEDSTATS
1229 /* load_balance() stats */
1230 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1231 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1232 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1233 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1234 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1235 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1236 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1237 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1239 /* Active load balancing */
1240 unsigned int alb_count;
1241 unsigned int alb_failed;
1242 unsigned int alb_pushed;
1244 /* SD_BALANCE_EXEC stats */
1245 unsigned int sbe_count;
1246 unsigned int sbe_balanced;
1247 unsigned int sbe_pushed;
1249 /* SD_BALANCE_FORK stats */
1250 unsigned int sbf_count;
1251 unsigned int sbf_balanced;
1252 unsigned int sbf_pushed;
1254 /* try_to_wake_up() stats */
1255 unsigned int ttwu_wake_remote;
1256 unsigned int ttwu_move_affine;
1257 unsigned int ttwu_move_balance;
1259 struct eas_stats eas_stats;
1261 #ifdef CONFIG_SCHED_DEBUG
1265 void *private; /* used during construction */
1266 struct rcu_head rcu; /* used during destruction */
1269 unsigned int span_weight;
1271 * Span of all CPUs in this domain.
1273 * NOTE: this field is variable length. (Allocated dynamically
1274 * by attaching extra space to the end of the structure,
1275 * depending on how many CPUs the kernel has booted up with)
1277 unsigned long span[0];
1280 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1282 return to_cpumask(sd->span);
1285 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1286 struct sched_domain_attr *dattr_new);
1288 /* Allocate an array of sched domains, for partition_sched_domains(). */
1289 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1290 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1292 bool cpus_share_cache(int this_cpu, int that_cpu);
1294 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1295 typedef int (*sched_domain_flags_f)(void);
1297 const struct sched_group_energy * const(*sched_domain_energy_f)(int cpu);
1299 #define SDTL_OVERLAP 0x01
1302 struct sched_domain **__percpu sd;
1303 struct sched_group **__percpu sg;
1304 struct sched_group_capacity **__percpu sgc;
1307 struct sched_domain_topology_level {
1308 sched_domain_mask_f mask;
1309 sched_domain_flags_f sd_flags;
1310 sched_domain_energy_f energy;
1313 struct sd_data data;
1314 #ifdef CONFIG_SCHED_DEBUG
1319 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1320 extern void wake_up_if_idle(int cpu);
1322 #ifdef CONFIG_SCHED_DEBUG
1323 # define SD_INIT_NAME(type) .name = #type
1325 # define SD_INIT_NAME(type)
1328 #else /* CONFIG_SMP */
1330 struct sched_domain_attr;
1333 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1334 struct sched_domain_attr *dattr_new)
1338 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1343 #endif /* !CONFIG_SMP */
1346 struct io_context; /* See blkdev.h */
1349 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1350 extern void prefetch_stack(struct task_struct *t);
1352 static inline void prefetch_stack(struct task_struct *t) { }
1355 struct audit_context; /* See audit.c */
1357 struct pipe_inode_info;
1358 struct uts_namespace;
1360 struct load_weight {
1361 unsigned long weight;
1366 * The load_avg/util_avg accumulates an infinite geometric series.
1367 * 1) load_avg factors frequency scaling into the amount of time that a
1368 * sched_entity is runnable on a rq into its weight. For cfs_rq, it is the
1369 * aggregated such weights of all runnable and blocked sched_entities.
1370 * 2) util_avg factors frequency and cpu scaling into the amount of time
1371 * that a sched_entity is running on a CPU, in the range [0..SCHED_LOAD_SCALE].
1372 * For cfs_rq, it is the aggregated such times of all runnable and
1373 * blocked sched_entities.
1374 * The 64 bit load_sum can:
1375 * 1) for cfs_rq, afford 4353082796 (=2^64/47742/88761) entities with
1376 * the highest weight (=88761) always runnable, we should not overflow
1377 * 2) for entity, support any load.weight always runnable
1380 u64 last_update_time, load_sum;
1381 u32 util_sum, period_contrib;
1382 unsigned long load_avg, util_avg;
1385 #ifdef CONFIG_SCHEDSTATS
1386 struct sched_statistics {
1396 s64 sum_sleep_runtime;
1403 u64 nr_migrations_cold;
1404 u64 nr_failed_migrations_affine;
1405 u64 nr_failed_migrations_running;
1406 u64 nr_failed_migrations_hot;
1407 u64 nr_forced_migrations;
1410 u64 nr_wakeups_sync;
1411 u64 nr_wakeups_migrate;
1412 u64 nr_wakeups_local;
1413 u64 nr_wakeups_remote;
1414 u64 nr_wakeups_affine;
1415 u64 nr_wakeups_affine_attempts;
1416 u64 nr_wakeups_passive;
1417 u64 nr_wakeups_idle;
1419 /* select_idle_sibling() */
1420 u64 nr_wakeups_sis_attempts;
1421 u64 nr_wakeups_sis_idle;
1422 u64 nr_wakeups_sis_cache_affine;
1423 u64 nr_wakeups_sis_suff_cap;
1424 u64 nr_wakeups_sis_idle_cpu;
1425 u64 nr_wakeups_sis_count;
1427 /* energy_aware_wake_cpu() */
1428 u64 nr_wakeups_secb_attempts;
1429 u64 nr_wakeups_secb_sync;
1430 u64 nr_wakeups_secb_idle_bt;
1431 u64 nr_wakeups_secb_insuff_cap;
1432 u64 nr_wakeups_secb_no_nrg_sav;
1433 u64 nr_wakeups_secb_nrg_sav;
1434 u64 nr_wakeups_secb_count;
1436 /* find_best_target() */
1437 u64 nr_wakeups_fbt_attempts;
1438 u64 nr_wakeups_fbt_no_cpu;
1439 u64 nr_wakeups_fbt_no_sd;
1440 u64 nr_wakeups_fbt_pref_idle;
1441 u64 nr_wakeups_fbt_count;
1444 /* select_task_rq_fair() */
1445 u64 nr_wakeups_cas_attempts;
1446 u64 nr_wakeups_cas_count;
1450 #define RAVG_HIST_SIZE_MAX 5
1451 #define NUM_BUSY_BUCKETS 10
1453 /* ravg represents frequency scaled cpu-demand of tasks */
1456 * 'mark_start' marks the beginning of an event (task waking up, task
1457 * starting to execute, task being preempted) within a window
1459 * 'sum' represents how runnable a task has been within current
1460 * window. It incorporates both running time and wait time and is
1463 * 'sum_history' keeps track of history of 'sum' seen over previous
1464 * RAVG_HIST_SIZE windows. Windows where task was entirely sleeping are
1467 * 'demand' represents maximum sum seen over previous
1468 * sysctl_sched_ravg_hist_size windows. 'demand' could drive frequency
1471 * 'curr_window_cpu' represents task's contribution to cpu busy time on
1472 * various CPUs in the current window
1474 * 'prev_window_cpu' represents task's contribution to cpu busy time on
1475 * various CPUs in the previous window
1477 * 'curr_window' represents the sum of all entries in curr_window_cpu
1479 * 'prev_window' represents the sum of all entries in prev_window_cpu
1481 * 'pred_demand' represents task's current predicted cpu busy time
1483 * 'busy_buckets' groups historical busy time into different buckets
1484 * used for prediction
1488 u32 sum_history[RAVG_HIST_SIZE_MAX];
1489 u32 *curr_window_cpu, *prev_window_cpu;
1490 u32 curr_window, prev_window;
1491 u64 curr_burst, avg_burst, avg_sleep_time;
1494 u8 busy_buckets[NUM_BUSY_BUCKETS];
1497 struct sched_entity {
1498 struct load_weight load; /* for load-balancing */
1499 struct rb_node run_node;
1500 struct list_head group_node;
1504 u64 sum_exec_runtime;
1506 u64 prev_sum_exec_runtime;
1510 #ifdef CONFIG_SCHEDSTATS
1511 struct sched_statistics statistics;
1514 #ifdef CONFIG_FAIR_GROUP_SCHED
1516 struct sched_entity *parent;
1517 /* rq on which this entity is (to be) queued: */
1518 struct cfs_rq *cfs_rq;
1519 /* rq "owned" by this entity/group: */
1520 struct cfs_rq *my_q;
1524 /* Per entity load average tracking */
1525 struct sched_avg avg;
1529 struct sched_rt_entity {
1530 struct list_head run_list;
1531 unsigned long timeout;
1532 unsigned long watchdog_stamp;
1533 unsigned int time_slice;
1534 unsigned short on_rq;
1535 unsigned short on_list;
1537 /* Accesses for these must be guarded by rq->lock of the task's rq */
1538 bool schedtune_enqueued;
1539 struct hrtimer schedtune_timer;
1541 struct sched_rt_entity *back;
1542 #ifdef CONFIG_RT_GROUP_SCHED
1543 struct sched_rt_entity *parent;
1544 /* rq on which this entity is (to be) queued: */
1545 struct rt_rq *rt_rq;
1546 /* rq "owned" by this entity/group: */
1551 struct sched_dl_entity {
1552 struct rb_node rb_node;
1555 * Original scheduling parameters. Copied here from sched_attr
1556 * during sched_setattr(), they will remain the same until
1557 * the next sched_setattr().
1559 u64 dl_runtime; /* maximum runtime for each instance */
1560 u64 dl_deadline; /* relative deadline of each instance */
1561 u64 dl_period; /* separation of two instances (period) */
1562 u64 dl_bw; /* dl_runtime / dl_deadline */
1563 u64 dl_density; /* dl_runtime / dl_deadline */
1566 * Actual scheduling parameters. Initialized with the values above,
1567 * they are continously updated during task execution. Note that
1568 * the remaining runtime could be < 0 in case we are in overrun.
1570 s64 runtime; /* remaining runtime for this instance */
1571 u64 deadline; /* absolute deadline for this instance */
1572 unsigned int flags; /* specifying the scheduler behaviour */
1577 * @dl_throttled tells if we exhausted the runtime. If so, the
1578 * task has to wait for a replenishment to be performed at the
1579 * next firing of dl_timer.
1581 * @dl_new tells if a new instance arrived. If so we must
1582 * start executing it with full runtime and reset its absolute
1585 * @dl_boosted tells if we are boosted due to DI. If so we are
1586 * outside bandwidth enforcement mechanism (but only until we
1587 * exit the critical section);
1589 * @dl_yielded tells if task gave up the cpu before consuming
1590 * all its available runtime during the last job.
1592 int dl_throttled, dl_new, dl_boosted, dl_yielded;
1595 * Bandwidth enforcement timer. Each -deadline task has its
1596 * own bandwidth to be enforced, thus we need one timer per task.
1598 struct hrtimer dl_timer;
1606 u8 pad; /* Otherwise the compiler can store garbage here. */
1608 u32 s; /* Set of bits. */
1612 enum perf_event_task_context {
1613 perf_invalid_context = -1,
1614 perf_hw_context = 0,
1616 perf_nr_task_contexts,
1619 /* Track pages that require TLB flushes */
1620 struct tlbflush_unmap_batch {
1622 * Each bit set is a CPU that potentially has a TLB entry for one of
1623 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1625 struct cpumask cpumask;
1627 /* True if any bit in cpumask is set */
1628 bool flush_required;
1631 * If true then the PTE was dirty when unmapped. The entry must be
1632 * flushed before IO is initiated or a stale TLB entry potentially
1633 * allows an update without redirtying the page.
1638 struct task_struct {
1639 #ifdef CONFIG_THREAD_INFO_IN_TASK
1641 * For reasons of header soup (see current_thread_info()), this
1642 * must be the first element of task_struct.
1644 struct thread_info thread_info;
1646 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1649 unsigned int flags; /* per process flags, defined below */
1650 unsigned int ptrace;
1653 struct llist_node wake_entry;
1655 #ifdef CONFIG_THREAD_INFO_IN_TASK
1656 unsigned int cpu; /* current CPU */
1658 unsigned int wakee_flips;
1659 unsigned long wakee_flip_decay_ts;
1660 struct task_struct *last_wakee;
1666 int prio, static_prio, normal_prio;
1667 unsigned int rt_priority;
1668 const struct sched_class *sched_class;
1669 struct sched_entity se;
1670 struct sched_rt_entity rt;
1671 #ifdef CONFIG_SCHED_HMP
1674 * 'init_load_pct' represents the initial task load assigned to children
1679 u64 last_switch_out_ts;
1680 u64 last_cpu_selected_ts;
1681 struct related_thread_group *grp;
1682 struct list_head grp_list;
1686 #ifdef CONFIG_CGROUP_SCHED
1687 struct task_group *sched_task_group;
1689 struct sched_dl_entity dl;
1691 #ifdef CONFIG_PREEMPT_NOTIFIERS
1692 /* list of struct preempt_notifier: */
1693 struct hlist_head preempt_notifiers;
1696 #ifdef CONFIG_BLK_DEV_IO_TRACE
1697 unsigned int btrace_seq;
1700 unsigned int policy;
1701 int nr_cpus_allowed;
1702 cpumask_t cpus_allowed;
1704 #ifdef CONFIG_PREEMPT_RCU
1705 int rcu_read_lock_nesting;
1706 union rcu_special rcu_read_unlock_special;
1707 struct list_head rcu_node_entry;
1708 struct rcu_node *rcu_blocked_node;
1709 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1710 #ifdef CONFIG_TASKS_RCU
1711 unsigned long rcu_tasks_nvcsw;
1712 bool rcu_tasks_holdout;
1713 struct list_head rcu_tasks_holdout_list;
1714 int rcu_tasks_idle_cpu;
1715 #endif /* #ifdef CONFIG_TASKS_RCU */
1717 #ifdef CONFIG_SCHED_INFO
1718 struct sched_info sched_info;
1721 struct list_head tasks;
1723 struct plist_node pushable_tasks;
1724 struct rb_node pushable_dl_tasks;
1727 struct mm_struct *mm, *active_mm;
1728 /* per-thread vma caching */
1729 u64 vmacache_seqnum;
1730 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1731 #if defined(SPLIT_RSS_COUNTING)
1732 struct task_rss_stat rss_stat;
1736 int exit_code, exit_signal;
1737 int pdeath_signal; /* The signal sent when the parent dies */
1738 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1740 /* Used for emulating ABI behavior of previous Linux versions */
1741 unsigned int personality;
1743 /* scheduler bits, serialized by scheduler locks */
1744 unsigned sched_reset_on_fork:1;
1745 unsigned sched_contributes_to_load:1;
1746 unsigned sched_migrated:1;
1747 unsigned :0; /* force alignment to the next boundary */
1749 /* unserialized, strictly 'current' */
1750 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1751 unsigned in_iowait:1;
1753 unsigned memcg_may_oom:1;
1755 #ifdef CONFIG_MEMCG_KMEM
1756 unsigned memcg_kmem_skip_account:1;
1758 #ifdef CONFIG_COMPAT_BRK
1759 unsigned brk_randomized:1;
1761 #ifdef CONFIG_CGROUPS
1762 /* disallow userland-initiated cgroup migration */
1763 unsigned no_cgroup_migration:1;
1766 unsigned long atomic_flags; /* Flags needing atomic access. */
1768 struct restart_block restart_block;
1773 #ifdef CONFIG_CC_STACKPROTECTOR
1774 /* Canary value for the -fstack-protector gcc feature */
1775 unsigned long stack_canary;
1778 * pointers to (original) parent process, youngest child, younger sibling,
1779 * older sibling, respectively. (p->father can be replaced with
1780 * p->real_parent->pid)
1782 struct task_struct __rcu *real_parent; /* real parent process */
1783 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1785 * children/sibling forms the list of my natural children
1787 struct list_head children; /* list of my children */
1788 struct list_head sibling; /* linkage in my parent's children list */
1789 struct task_struct *group_leader; /* threadgroup leader */
1792 * ptraced is the list of tasks this task is using ptrace on.
1793 * This includes both natural children and PTRACE_ATTACH targets.
1794 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1796 struct list_head ptraced;
1797 struct list_head ptrace_entry;
1799 /* PID/PID hash table linkage. */
1800 struct pid_link pids[PIDTYPE_MAX];
1801 struct list_head thread_group;
1802 struct list_head thread_node;
1804 struct completion *vfork_done; /* for vfork() */
1805 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1806 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1808 cputime_t utime, stime, utimescaled, stimescaled;
1810 #ifdef CONFIG_CPU_FREQ_TIMES
1812 unsigned int max_state;
1814 struct prev_cputime prev_cputime;
1815 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1816 seqlock_t vtime_seqlock;
1817 unsigned long long vtime_snap;
1822 } vtime_snap_whence;
1824 unsigned long nvcsw, nivcsw; /* context switch counts */
1825 u64 start_time; /* monotonic time in nsec */
1826 u64 real_start_time; /* boot based time in nsec */
1827 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1828 unsigned long min_flt, maj_flt;
1830 struct task_cputime cputime_expires;
1831 struct list_head cpu_timers[3];
1833 /* process credentials */
1834 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1835 const struct cred __rcu *real_cred; /* objective and real subjective task
1836 * credentials (COW) */
1837 const struct cred __rcu *cred; /* effective (overridable) subjective task
1838 * credentials (COW) */
1839 char comm[TASK_COMM_LEN]; /* executable name excluding path
1840 - access with [gs]et_task_comm (which lock
1841 it with task_lock())
1842 - initialized normally by setup_new_exec */
1843 /* file system info */
1844 struct nameidata *nameidata;
1845 #ifdef CONFIG_SYSVIPC
1847 struct sysv_sem sysvsem;
1848 struct sysv_shm sysvshm;
1850 #ifdef CONFIG_DETECT_HUNG_TASK
1851 /* hung task detection */
1852 unsigned long last_switch_count;
1854 /* filesystem information */
1855 struct fs_struct *fs;
1856 /* open file information */
1857 struct files_struct *files;
1859 struct nsproxy *nsproxy;
1860 /* signal handlers */
1861 struct signal_struct *signal;
1862 struct sighand_struct *sighand;
1864 sigset_t blocked, real_blocked;
1865 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1866 struct sigpending pending;
1868 unsigned long sas_ss_sp;
1871 struct callback_head *task_works;
1873 struct audit_context *audit_context;
1874 #ifdef CONFIG_AUDITSYSCALL
1876 unsigned int sessionid;
1878 struct seccomp seccomp;
1880 /* Thread group tracking */
1883 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1885 spinlock_t alloc_lock;
1887 /* Protection of the PI data structures: */
1888 raw_spinlock_t pi_lock;
1890 struct wake_q_node wake_q;
1892 #ifdef CONFIG_RT_MUTEXES
1893 /* PI waiters blocked on a rt_mutex held by this task */
1894 struct rb_root pi_waiters;
1895 struct rb_node *pi_waiters_leftmost;
1896 /* Deadlock detection and priority inheritance handling */
1897 struct rt_mutex_waiter *pi_blocked_on;
1900 #ifdef CONFIG_DEBUG_MUTEXES
1901 /* mutex deadlock detection */
1902 struct mutex_waiter *blocked_on;
1904 #ifdef CONFIG_TRACE_IRQFLAGS
1905 unsigned int irq_events;
1906 unsigned long hardirq_enable_ip;
1907 unsigned long hardirq_disable_ip;
1908 unsigned int hardirq_enable_event;
1909 unsigned int hardirq_disable_event;
1910 int hardirqs_enabled;
1911 int hardirq_context;
1912 unsigned long softirq_disable_ip;
1913 unsigned long softirq_enable_ip;
1914 unsigned int softirq_disable_event;
1915 unsigned int softirq_enable_event;
1916 int softirqs_enabled;
1917 int softirq_context;
1919 #ifdef CONFIG_LOCKDEP
1920 # define MAX_LOCK_DEPTH 48UL
1923 unsigned int lockdep_recursion;
1924 struct held_lock held_locks[MAX_LOCK_DEPTH];
1925 gfp_t lockdep_reclaim_gfp;
1928 unsigned int in_ubsan;
1931 /* journalling filesystem info */
1934 /* stacked block device info */
1935 struct bio_list *bio_list;
1938 /* stack plugging */
1939 struct blk_plug *plug;
1943 struct reclaim_state *reclaim_state;
1945 struct backing_dev_info *backing_dev_info;
1947 struct io_context *io_context;
1949 unsigned long ptrace_message;
1950 siginfo_t *last_siginfo; /* For ptrace use. */
1951 struct task_io_accounting ioac;
1952 #if defined(CONFIG_TASK_XACCT)
1953 u64 acct_rss_mem1; /* accumulated rss usage */
1954 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1955 cputime_t acct_timexpd; /* stime + utime since last update */
1957 #ifdef CONFIG_CPUSETS
1958 nodemask_t mems_allowed; /* Protected by alloc_lock */
1959 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1960 int cpuset_mem_spread_rotor;
1961 int cpuset_slab_spread_rotor;
1963 #ifdef CONFIG_CGROUPS
1964 /* Control Group info protected by css_set_lock */
1965 struct css_set __rcu *cgroups;
1966 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1967 struct list_head cg_list;
1970 struct robust_list_head __user *robust_list;
1971 #ifdef CONFIG_COMPAT
1972 struct compat_robust_list_head __user *compat_robust_list;
1974 struct list_head pi_state_list;
1975 struct futex_pi_state *pi_state_cache;
1977 #ifdef CONFIG_PERF_EVENTS
1978 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1979 struct mutex perf_event_mutex;
1980 struct list_head perf_event_list;
1982 #ifdef CONFIG_DEBUG_PREEMPT
1983 unsigned long preempt_disable_ip;
1986 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1988 short pref_node_fork;
1990 #ifdef CONFIG_NUMA_BALANCING
1992 unsigned int numa_scan_period;
1993 unsigned int numa_scan_period_max;
1994 int numa_preferred_nid;
1995 unsigned long numa_migrate_retry;
1996 u64 node_stamp; /* migration stamp */
1997 u64 last_task_numa_placement;
1998 u64 last_sum_exec_runtime;
1999 struct callback_head numa_work;
2001 struct list_head numa_entry;
2002 struct numa_group *numa_group;
2005 * numa_faults is an array split into four regions:
2006 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
2007 * in this precise order.
2009 * faults_memory: Exponential decaying average of faults on a per-node
2010 * basis. Scheduling placement decisions are made based on these
2011 * counts. The values remain static for the duration of a PTE scan.
2012 * faults_cpu: Track the nodes the process was running on when a NUMA
2013 * hinting fault was incurred.
2014 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
2015 * during the current scan window. When the scan completes, the counts
2016 * in faults_memory and faults_cpu decay and these values are copied.
2018 unsigned long *numa_faults;
2019 unsigned long total_numa_faults;
2022 * numa_faults_locality tracks if faults recorded during the last
2023 * scan window were remote/local or failed to migrate. The task scan
2024 * period is adapted based on the locality of the faults with different
2025 * weights depending on whether they were shared or private faults
2027 unsigned long numa_faults_locality[3];
2029 unsigned long numa_pages_migrated;
2030 #endif /* CONFIG_NUMA_BALANCING */
2032 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
2033 struct tlbflush_unmap_batch tlb_ubc;
2036 struct rcu_head rcu;
2039 * cache last used pipe for splice
2041 struct pipe_inode_info *splice_pipe;
2043 struct page_frag task_frag;
2045 #ifdef CONFIG_TASK_DELAY_ACCT
2046 struct task_delay_info *delays;
2048 #ifdef CONFIG_FAULT_INJECTION
2052 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
2053 * balance_dirty_pages() for some dirty throttling pause
2056 int nr_dirtied_pause;
2057 unsigned long dirty_paused_when; /* start of a write-and-pause period */
2059 #ifdef CONFIG_LATENCYTOP
2060 int latency_record_count;
2061 struct latency_record latency_record[LT_SAVECOUNT];
2064 * time slack values; these are used to round up poll() and
2065 * select() etc timeout values. These are in nanoseconds.
2068 u64 default_timer_slack_ns;
2071 unsigned int kasan_depth;
2073 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
2074 /* Index of current stored address in ret_stack */
2076 /* Stack of return addresses for return function tracing */
2077 struct ftrace_ret_stack *ret_stack;
2078 /* time stamp for last schedule */
2079 unsigned long long ftrace_timestamp;
2081 * Number of functions that haven't been traced
2082 * because of depth overrun.
2084 atomic_t trace_overrun;
2085 /* Pause for the tracing */
2086 atomic_t tracing_graph_pause;
2088 #ifdef CONFIG_TRACING
2089 /* state flags for use by tracers */
2090 unsigned long trace;
2091 /* bitmask and counter of trace recursion */
2092 unsigned long trace_recursion;
2093 #endif /* CONFIG_TRACING */
2095 /* Coverage collection mode enabled for this task (0 if disabled). */
2096 enum kcov_mode kcov_mode;
2097 /* Size of the kcov_area. */
2099 /* Buffer for coverage collection. */
2101 /* kcov desciptor wired with this task or NULL. */
2105 struct mem_cgroup *memcg_in_oom;
2106 gfp_t memcg_oom_gfp_mask;
2107 int memcg_oom_order;
2109 /* number of pages to reclaim on returning to userland */
2110 unsigned int memcg_nr_pages_over_high;
2112 #ifdef CONFIG_UPROBES
2113 struct uprobe_task *utask;
2115 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
2116 unsigned int sequential_io;
2117 unsigned int sequential_io_avg;
2119 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
2120 unsigned long task_state_change;
2122 int pagefault_disabled;
2123 /* CPU-specific state of this task */
2124 struct thread_struct thread;
2126 * WARNING: on x86, 'thread_struct' contains a variable-sized
2127 * structure. It *MUST* be at the end of 'task_struct'.
2129 * Do not put anything below here!
2133 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
2134 extern int arch_task_struct_size __read_mostly;
2136 # define arch_task_struct_size (sizeof(struct task_struct))
2139 /* Future-safe accessor for struct task_struct's cpus_allowed. */
2140 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
2142 #define TNF_MIGRATED 0x01
2143 #define TNF_NO_GROUP 0x02
2144 #define TNF_SHARED 0x04
2145 #define TNF_FAULT_LOCAL 0x08
2146 #define TNF_MIGRATE_FAIL 0x10
2148 #ifdef CONFIG_NUMA_BALANCING
2149 extern void task_numa_fault(int last_node, int node, int pages, int flags);
2150 extern pid_t task_numa_group_id(struct task_struct *p);
2151 extern void set_numabalancing_state(bool enabled);
2152 extern void task_numa_free(struct task_struct *p, bool final);
2153 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2154 int src_nid, int dst_cpu);
2156 static inline void task_numa_fault(int last_node, int node, int pages,
2160 static inline pid_t task_numa_group_id(struct task_struct *p)
2164 static inline void set_numabalancing_state(bool enabled)
2167 static inline void task_numa_free(struct task_struct *p, bool final)
2170 static inline bool should_numa_migrate_memory(struct task_struct *p,
2171 struct page *page, int src_nid, int dst_cpu)
2177 static inline struct pid *task_pid(struct task_struct *task)
2179 return task->pids[PIDTYPE_PID].pid;
2182 static inline struct pid *task_tgid(struct task_struct *task)
2184 return task->group_leader->pids[PIDTYPE_PID].pid;
2188 * Without tasklist or rcu lock it is not safe to dereference
2189 * the result of task_pgrp/task_session even if task == current,
2190 * we can race with another thread doing sys_setsid/sys_setpgid.
2192 static inline struct pid *task_pgrp(struct task_struct *task)
2194 return task->group_leader->pids[PIDTYPE_PGID].pid;
2197 static inline struct pid *task_session(struct task_struct *task)
2199 return task->group_leader->pids[PIDTYPE_SID].pid;
2202 struct pid_namespace;
2205 * the helpers to get the task's different pids as they are seen
2206 * from various namespaces
2208 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2209 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2211 * task_xid_nr_ns() : id seen from the ns specified;
2213 * set_task_vxid() : assigns a virtual id to a task;
2215 * see also pid_nr() etc in include/linux/pid.h
2217 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2218 struct pid_namespace *ns);
2220 static inline pid_t task_pid_nr(struct task_struct *tsk)
2225 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2226 struct pid_namespace *ns)
2228 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2231 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2233 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2237 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2242 static inline int pid_alive(const struct task_struct *p);
2243 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2245 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2246 struct pid_namespace *ns)
2248 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2251 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2253 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2257 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2258 struct pid_namespace *ns)
2260 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2263 static inline pid_t task_session_vnr(struct task_struct *tsk)
2265 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2268 static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
2270 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
2273 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2275 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
2278 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2284 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2290 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2292 return task_ppid_nr_ns(tsk, &init_pid_ns);
2295 /* obsolete, do not use */
2296 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2298 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2302 * pid_alive - check that a task structure is not stale
2303 * @p: Task structure to be checked.
2305 * Test if a process is not yet dead (at most zombie state)
2306 * If pid_alive fails, then pointers within the task structure
2307 * can be stale and must not be dereferenced.
2309 * Return: 1 if the process is alive. 0 otherwise.
2311 static inline int pid_alive(const struct task_struct *p)
2313 return p->pids[PIDTYPE_PID].pid != NULL;
2317 * is_global_init - check if a task structure is init. Since init
2318 * is free to have sub-threads we need to check tgid.
2319 * @tsk: Task structure to be checked.
2321 * Check if a task structure is the first user space task the kernel created.
2323 * Return: 1 if the task structure is init. 0 otherwise.
2325 static inline int is_global_init(struct task_struct *tsk)
2327 return task_tgid_nr(tsk) == 1;
2330 extern struct pid *cad_pid;
2332 extern void free_task(struct task_struct *tsk);
2333 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2335 extern void __put_task_struct(struct task_struct *t);
2337 static inline void put_task_struct(struct task_struct *t)
2339 if (atomic_dec_and_test(&t->usage))
2340 __put_task_struct(t);
2343 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2344 extern void task_cputime(struct task_struct *t,
2345 cputime_t *utime, cputime_t *stime);
2346 extern void task_cputime_scaled(struct task_struct *t,
2347 cputime_t *utimescaled, cputime_t *stimescaled);
2348 extern cputime_t task_gtime(struct task_struct *t);
2350 static inline void task_cputime(struct task_struct *t,
2351 cputime_t *utime, cputime_t *stime)
2359 static inline void task_cputime_scaled(struct task_struct *t,
2360 cputime_t *utimescaled,
2361 cputime_t *stimescaled)
2364 *utimescaled = t->utimescaled;
2366 *stimescaled = t->stimescaled;
2369 static inline cputime_t task_gtime(struct task_struct *t)
2374 extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2375 extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2380 #define PF_WAKE_UP_IDLE 0x00000002 /* try to wake up on an idle CPU */
2381 #define PF_EXITING 0x00000004 /* getting shut down */
2382 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2383 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2384 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2385 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2386 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2387 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2388 #define PF_DUMPCORE 0x00000200 /* dumped core */
2389 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2390 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2391 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2392 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2393 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2394 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2395 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2396 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2397 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2398 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2399 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2400 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2401 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2402 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2403 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2404 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2405 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2406 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2407 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2410 * Only the _current_ task can read/write to tsk->flags, but other
2411 * tasks can access tsk->flags in readonly mode for example
2412 * with tsk_used_math (like during threaded core dumping).
2413 * There is however an exception to this rule during ptrace
2414 * or during fork: the ptracer task is allowed to write to the
2415 * child->flags of its traced child (same goes for fork, the parent
2416 * can write to the child->flags), because we're guaranteed the
2417 * child is not running and in turn not changing child->flags
2418 * at the same time the parent does it.
2420 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2421 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2422 #define clear_used_math() clear_stopped_child_used_math(current)
2423 #define set_used_math() set_stopped_child_used_math(current)
2424 #define conditional_stopped_child_used_math(condition, child) \
2425 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2426 #define conditional_used_math(condition) \
2427 conditional_stopped_child_used_math(condition, current)
2428 #define copy_to_stopped_child_used_math(child) \
2429 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2430 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2431 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2432 #define used_math() tsk_used_math(current)
2434 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2435 * __GFP_FS is also cleared as it implies __GFP_IO.
2437 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2439 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2440 flags &= ~(__GFP_IO | __GFP_FS);
2444 static inline unsigned int memalloc_noio_save(void)
2446 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2447 current->flags |= PF_MEMALLOC_NOIO;
2451 static inline void memalloc_noio_restore(unsigned int flags)
2453 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2456 /* Per-process atomic flags. */
2457 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2458 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2459 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2460 #define PFA_SPEC_SSB_DISABLE 4 /* Speculative Store Bypass disabled */
2461 #define PFA_SPEC_SSB_FORCE_DISABLE 5 /* Speculative Store Bypass force disabled*/
2462 #define PFA_SPEC_IB_DISABLE 6 /* Indirect branch speculation restricted */
2463 #define PFA_SPEC_IB_FORCE_DISABLE 7 /* Indirect branch speculation permanently restricted */
2466 #define TASK_PFA_TEST(name, func) \
2467 static inline bool task_##func(struct task_struct *p) \
2468 { return test_bit(PFA_##name, &p->atomic_flags); }
2469 #define TASK_PFA_SET(name, func) \
2470 static inline void task_set_##func(struct task_struct *p) \
2471 { set_bit(PFA_##name, &p->atomic_flags); }
2472 #define TASK_PFA_CLEAR(name, func) \
2473 static inline void task_clear_##func(struct task_struct *p) \
2474 { clear_bit(PFA_##name, &p->atomic_flags); }
2476 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2477 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2479 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2480 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2481 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2483 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2484 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2485 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2487 TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
2488 TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
2489 TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
2491 TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
2492 TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
2494 TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
2495 TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
2496 TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
2498 TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
2499 TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
2502 * task->jobctl flags
2504 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2506 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2507 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2508 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2509 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2510 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2511 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2512 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2514 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2515 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2516 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2517 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2518 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2519 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2520 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2522 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2523 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2525 extern bool task_set_jobctl_pending(struct task_struct *task,
2526 unsigned long mask);
2527 extern void task_clear_jobctl_trapping(struct task_struct *task);
2528 extern void task_clear_jobctl_pending(struct task_struct *task,
2529 unsigned long mask);
2531 static inline void rcu_copy_process(struct task_struct *p)
2533 #ifdef CONFIG_PREEMPT_RCU
2534 p->rcu_read_lock_nesting = 0;
2535 p->rcu_read_unlock_special.s = 0;
2536 p->rcu_blocked_node = NULL;
2537 INIT_LIST_HEAD(&p->rcu_node_entry);
2538 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2539 #ifdef CONFIG_TASKS_RCU
2540 p->rcu_tasks_holdout = false;
2541 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2542 p->rcu_tasks_idle_cpu = -1;
2543 #endif /* #ifdef CONFIG_TASKS_RCU */
2546 static inline void tsk_restore_flags(struct task_struct *task,
2547 unsigned long orig_flags, unsigned long flags)
2549 task->flags &= ~flags;
2550 task->flags |= orig_flags & flags;
2553 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2554 const struct cpumask *trial);
2555 extern int task_can_attach(struct task_struct *p,
2556 const struct cpumask *cs_cpus_allowed);
2558 extern void do_set_cpus_allowed(struct task_struct *p,
2559 const struct cpumask *new_mask);
2561 extern int set_cpus_allowed_ptr(struct task_struct *p,
2562 const struct cpumask *new_mask);
2563 extern bool cpupri_check_rt(void);
2565 static inline void do_set_cpus_allowed(struct task_struct *p,
2566 const struct cpumask *new_mask)
2569 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2570 const struct cpumask *new_mask)
2572 if (!cpumask_test_cpu(0, new_mask))
2576 static inline bool cpupri_check_rt(void)
2583 unsigned long prev_load;
2584 unsigned long new_task_load;
2585 unsigned long predicted_load;
2588 struct cpu_cycle_counter_cb {
2589 u64 (*get_cpu_cycle_counter)(int cpu);
2592 #define MAX_NUM_CGROUP_COLOC_ID 20
2594 #ifdef CONFIG_SCHED_HMP
2595 extern void free_task_load_ptrs(struct task_struct *p);
2596 extern int sched_set_window(u64 window_start, unsigned int window_size);
2597 extern unsigned long sched_get_busy(int cpu);
2598 extern void sched_get_cpus_busy(struct sched_load *busy,
2599 const struct cpumask *query_cpus);
2600 extern void sched_set_io_is_busy(int val);
2601 extern int sched_set_boost(int enable);
2602 extern int sched_set_init_task_load(struct task_struct *p, int init_load_pct);
2603 extern u32 sched_get_init_task_load(struct task_struct *p);
2604 extern int sched_set_static_cpu_pwr_cost(int cpu, unsigned int cost);
2605 extern unsigned int sched_get_static_cpu_pwr_cost(int cpu);
2606 extern int sched_set_static_cluster_pwr_cost(int cpu, unsigned int cost);
2607 extern unsigned int sched_get_static_cluster_pwr_cost(int cpu);
2608 extern int sched_set_cluster_wake_idle(int cpu, unsigned int wake_idle);
2609 extern unsigned int sched_get_cluster_wake_idle(int cpu);
2610 extern int sched_update_freq_max_load(const cpumask_t *cpumask);
2611 extern void sched_update_cpu_freq_min_max(const cpumask_t *cpus,
2612 u32 fmin, u32 fmax);
2613 extern void sched_set_cpu_cstate(int cpu, int cstate,
2614 int wakeup_energy, int wakeup_latency);
2615 extern void sched_set_cluster_dstate(const cpumask_t *cluster_cpus, int dstate,
2616 int wakeup_energy, int wakeup_latency);
2617 extern int register_cpu_cycle_counter_cb(struct cpu_cycle_counter_cb *cb);
2618 extern u64 sched_ktime_clock(void);
2619 extern int sched_set_group_id(struct task_struct *p, unsigned int group_id);
2620 extern unsigned int sched_get_group_id(struct task_struct *p);
2622 #else /* CONFIG_SCHED_HMP */
2623 static inline void free_task_load_ptrs(struct task_struct *p) { }
2625 static inline u64 sched_ktime_clock(void)
2631 register_cpu_cycle_counter_cb(struct cpu_cycle_counter_cb *cb)
2636 static inline int sched_set_window(u64 window_start, unsigned int window_size)
2640 static inline unsigned long sched_get_busy(int cpu)
2644 static inline void sched_get_cpus_busy(struct sched_load *busy,
2645 const struct cpumask *query_cpus) {};
2647 static inline void sched_set_io_is_busy(int val) {};
2649 static inline int sched_set_boost(int enable)
2654 static inline int sched_update_freq_max_load(const cpumask_t *cpumask)
2659 static inline void sched_update_cpu_freq_min_max(const cpumask_t *cpus,
2660 u32 fmin, u32 fmax) { }
2663 sched_set_cpu_cstate(int cpu, int cstate, int wakeup_energy, int wakeup_latency)
2667 static inline void sched_set_cluster_dstate(const cpumask_t *cluster_cpus,
2668 int dstate, int wakeup_energy, int wakeup_latency)
2671 #endif /* CONFIG_SCHED_HMP */
2673 #ifdef CONFIG_NO_HZ_COMMON
2674 void calc_load_enter_idle(void);
2675 void calc_load_exit_idle(void);
2677 static inline void calc_load_enter_idle(void) { }
2678 static inline void calc_load_exit_idle(void) { }
2679 #endif /* CONFIG_NO_HZ_COMMON */
2681 static inline void set_wake_up_idle(bool enabled)
2684 current->flags |= PF_WAKE_UP_IDLE;
2686 current->flags &= ~PF_WAKE_UP_IDLE;
2690 * Do not use outside of architecture code which knows its limitations.
2692 * sched_clock() has no promise of monotonicity or bounded drift between
2693 * CPUs, use (which you should not) requires disabling IRQs.
2695 * Please use one of the three interfaces below.
2697 extern unsigned long long notrace sched_clock(void);
2699 * See the comment in kernel/sched/clock.c
2701 extern u64 cpu_clock(int cpu);
2702 extern u64 local_clock(void);
2703 extern u64 running_clock(void);
2704 extern u64 sched_clock_cpu(int cpu);
2706 extern void sched_clock_init(void);
2707 extern int sched_clock_initialized(void);
2709 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2710 static inline void sched_clock_tick(void)
2714 static inline void sched_clock_idle_sleep_event(void)
2718 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2723 * Architectures can set this to 1 if they have specified
2724 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2725 * but then during bootup it turns out that sched_clock()
2726 * is reliable after all:
2728 extern int sched_clock_stable(void);
2729 extern void set_sched_clock_stable(void);
2730 extern void clear_sched_clock_stable(void);
2732 extern void sched_clock_tick(void);
2733 extern void sched_clock_idle_sleep_event(void);
2734 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2737 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2739 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2740 * The reason for this explicit opt-in is not to have perf penalty with
2741 * slow sched_clocks.
2743 extern void enable_sched_clock_irqtime(void);
2744 extern void disable_sched_clock_irqtime(void);
2746 static inline void enable_sched_clock_irqtime(void) {}
2747 static inline void disable_sched_clock_irqtime(void) {}
2750 extern unsigned long long
2751 task_sched_runtime(struct task_struct *task);
2753 /* sched_exec is called by processes performing an exec */
2754 #if defined(CONFIG_SMP)
2755 extern void sched_exec(void);
2757 #define sched_exec() {}
2760 extern void sched_clock_idle_sleep_event(void);
2761 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2763 #ifdef CONFIG_HOTPLUG_CPU
2764 extern void idle_task_exit(void);
2766 static inline void idle_task_exit(void) {}
2769 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2770 extern void wake_up_nohz_cpu(int cpu);
2772 static inline void wake_up_nohz_cpu(int cpu) { }
2775 #ifdef CONFIG_NO_HZ_FULL
2776 extern bool sched_can_stop_tick(void);
2777 extern u64 scheduler_tick_max_deferment(void);
2779 static inline bool sched_can_stop_tick(void) { return false; }
2782 #ifdef CONFIG_SCHED_AUTOGROUP
2783 extern void sched_autogroup_create_attach(struct task_struct *p);
2784 extern void sched_autogroup_detach(struct task_struct *p);
2785 extern void sched_autogroup_fork(struct signal_struct *sig);
2786 extern void sched_autogroup_exit(struct signal_struct *sig);
2787 #ifdef CONFIG_PROC_FS
2788 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2789 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2792 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2793 static inline void sched_autogroup_detach(struct task_struct *p) { }
2794 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2795 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2798 extern int yield_to(struct task_struct *p, bool preempt);
2799 extern void set_user_nice(struct task_struct *p, long nice);
2800 extern int task_prio(const struct task_struct *p);
2802 * task_nice - return the nice value of a given task.
2803 * @p: the task in question.
2805 * Return: The nice value [ -20 ... 0 ... 19 ].
2807 static inline int task_nice(const struct task_struct *p)
2809 return PRIO_TO_NICE((p)->static_prio);
2811 extern int can_nice(const struct task_struct *p, const int nice);
2812 extern int task_curr(const struct task_struct *p);
2813 extern int idle_cpu(int cpu);
2814 extern int sched_setscheduler(struct task_struct *, int,
2815 const struct sched_param *);
2816 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2817 const struct sched_param *);
2818 extern int sched_setattr(struct task_struct *,
2819 const struct sched_attr *);
2820 extern struct task_struct *idle_task(int cpu);
2822 * is_idle_task - is the specified task an idle task?
2823 * @p: the task in question.
2825 * Return: 1 if @p is an idle task. 0 otherwise.
2827 static inline bool is_idle_task(const struct task_struct *p)
2831 extern struct task_struct *curr_task(int cpu);
2832 extern void set_curr_task(int cpu, struct task_struct *p);
2836 union thread_union {
2837 #ifndef CONFIG_THREAD_INFO_IN_TASK
2838 struct thread_info thread_info;
2840 unsigned long stack[THREAD_SIZE/sizeof(long)];
2843 #ifndef __HAVE_ARCH_KSTACK_END
2844 static inline int kstack_end(void *addr)
2846 /* Reliable end of stack detection:
2847 * Some APM bios versions misalign the stack
2849 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2853 extern union thread_union init_thread_union;
2854 extern struct task_struct init_task;
2856 extern struct mm_struct init_mm;
2858 extern struct pid_namespace init_pid_ns;
2861 * find a task by one of its numerical ids
2863 * find_task_by_pid_ns():
2864 * finds a task by its pid in the specified namespace
2865 * find_task_by_vpid():
2866 * finds a task by its virtual pid
2868 * see also find_vpid() etc in include/linux/pid.h
2871 extern struct task_struct *find_task_by_vpid(pid_t nr);
2872 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2873 struct pid_namespace *ns);
2875 /* per-UID process charging. */
2876 extern struct user_struct * alloc_uid(kuid_t);
2877 static inline struct user_struct *get_uid(struct user_struct *u)
2879 atomic_inc(&u->__count);
2882 extern void free_uid(struct user_struct *);
2884 #include <asm/current.h>
2886 extern void xtime_update(unsigned long ticks);
2888 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2889 extern int wake_up_process(struct task_struct *tsk);
2890 extern int wake_up_process_no_notif(struct task_struct *tsk);
2891 extern void wake_up_new_task(struct task_struct *tsk);
2893 extern void kick_process(struct task_struct *tsk);
2895 static inline void kick_process(struct task_struct *tsk) { }
2897 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2898 extern void sched_dead(struct task_struct *p);
2899 #ifdef CONFIG_SCHED_HMP
2900 extern void sched_exit(struct task_struct *p);
2902 static inline void sched_exit(struct task_struct *p) { }
2905 extern void proc_caches_init(void);
2906 extern void flush_signals(struct task_struct *);
2907 extern void ignore_signals(struct task_struct *);
2908 extern void flush_signal_handlers(struct task_struct *, int force_default);
2909 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2911 static inline int kernel_dequeue_signal(siginfo_t *info)
2913 struct task_struct *tsk = current;
2917 spin_lock_irq(&tsk->sighand->siglock);
2918 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2919 spin_unlock_irq(&tsk->sighand->siglock);
2924 static inline void kernel_signal_stop(void)
2926 spin_lock_irq(¤t->sighand->siglock);
2927 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2928 __set_current_state(TASK_STOPPED);
2929 spin_unlock_irq(¤t->sighand->siglock);
2934 extern void release_task(struct task_struct * p);
2935 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2936 extern int force_sigsegv(int, struct task_struct *);
2937 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2938 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2939 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2940 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2941 const struct cred *, u32);
2942 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2943 extern int kill_pid(struct pid *pid, int sig, int priv);
2944 extern int kill_proc_info(int, struct siginfo *, pid_t);
2945 extern __must_check bool do_notify_parent(struct task_struct *, int);
2946 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2947 extern void force_sig(int, struct task_struct *);
2948 extern int send_sig(int, struct task_struct *, int);
2949 extern int zap_other_threads(struct task_struct *p);
2950 extern struct sigqueue *sigqueue_alloc(void);
2951 extern void sigqueue_free(struct sigqueue *);
2952 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2953 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2955 static inline void restore_saved_sigmask(void)
2957 if (test_and_clear_restore_sigmask())
2958 __set_current_blocked(¤t->saved_sigmask);
2961 static inline sigset_t *sigmask_to_save(void)
2963 sigset_t *res = ¤t->blocked;
2964 if (unlikely(test_restore_sigmask()))
2965 res = ¤t->saved_sigmask;
2969 static inline int kill_cad_pid(int sig, int priv)
2971 return kill_pid(cad_pid, sig, priv);
2974 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2975 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2976 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2977 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2980 * True if we are on the alternate signal stack.
2982 static inline int on_sig_stack(unsigned long sp)
2984 #ifdef CONFIG_STACK_GROWSUP
2985 return sp >= current->sas_ss_sp &&
2986 sp - current->sas_ss_sp < current->sas_ss_size;
2988 return sp > current->sas_ss_sp &&
2989 sp - current->sas_ss_sp <= current->sas_ss_size;
2993 static inline int sas_ss_flags(unsigned long sp)
2995 if (!current->sas_ss_size)
2998 return on_sig_stack(sp) ? SS_ONSTACK : 0;
3001 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
3003 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
3004 #ifdef CONFIG_STACK_GROWSUP
3005 return current->sas_ss_sp;
3007 return current->sas_ss_sp + current->sas_ss_size;
3013 * Routines for handling mm_structs
3015 extern struct mm_struct * mm_alloc(void);
3017 /* mmdrop drops the mm and the page tables */
3018 extern void __mmdrop(struct mm_struct *);
3019 static inline void mmdrop(struct mm_struct *mm)
3021 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
3025 static inline bool mmget_not_zero(struct mm_struct *mm)
3027 return atomic_inc_not_zero(&mm->mm_users);
3030 /* mmput gets rid of the mappings and all user-space */
3031 extern int mmput(struct mm_struct *);
3032 /* same as above but performs the slow path from the async kontext. Can
3033 * be called from the atomic context as well
3035 extern void mmput_async(struct mm_struct *);
3037 /* Grab a reference to a task's mm, if it is not already going away */
3038 extern struct mm_struct *get_task_mm(struct task_struct *task);
3040 * Grab a reference to a task's mm, if it is not already going away
3041 * and ptrace_may_access with the mode parameter passed to it
3044 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
3045 /* Remove the current tasks stale references to the old mm_struct */
3046 extern void mm_release(struct task_struct *, struct mm_struct *);
3048 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
3049 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
3050 struct task_struct *, unsigned long);
3052 extern int copy_thread(unsigned long, unsigned long, unsigned long,
3053 struct task_struct *);
3055 /* Architectures that haven't opted into copy_thread_tls get the tls argument
3056 * via pt_regs, so ignore the tls argument passed via C. */
3057 static inline int copy_thread_tls(
3058 unsigned long clone_flags, unsigned long sp, unsigned long arg,
3059 struct task_struct *p, unsigned long tls)
3061 return copy_thread(clone_flags, sp, arg, p);
3064 extern void flush_thread(void);
3066 #ifdef CONFIG_HAVE_EXIT_THREAD
3067 extern void exit_thread(struct task_struct *tsk);
3069 static inline void exit_thread(struct task_struct *tsk)
3074 extern void exit_files(struct task_struct *);
3075 extern void __cleanup_sighand(struct sighand_struct *);
3077 extern void exit_itimers(struct signal_struct *);
3078 extern void flush_itimer_signals(void);
3080 extern void do_group_exit(int);
3082 extern int do_execve(struct filename *,
3083 const char __user * const __user *,
3084 const char __user * const __user *);
3085 extern int do_execveat(int, struct filename *,
3086 const char __user * const __user *,
3087 const char __user * const __user *,
3089 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
3090 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
3091 struct task_struct *fork_idle(int);
3092 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
3094 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
3095 static inline void set_task_comm(struct task_struct *tsk, const char *from)
3097 __set_task_comm(tsk, from, false);
3100 extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
3101 #define get_task_comm(buf, tsk) ({ \
3102 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
3103 __get_task_comm(buf, sizeof(buf), tsk); \
3107 void scheduler_ipi(void);
3108 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
3110 static inline void scheduler_ipi(void) { }
3111 static inline unsigned long wait_task_inactive(struct task_struct *p,
3118 #define tasklist_empty() \
3119 list_empty(&init_task.tasks)
3121 #define next_task(p) \
3122 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
3124 #define for_each_process(p) \
3125 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
3127 extern bool current_is_single_threaded(void);
3130 * Careful: do_each_thread/while_each_thread is a double loop so
3131 * 'break' will not work as expected - use goto instead.
3133 #define do_each_thread(g, t) \
3134 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3136 #define while_each_thread(g, t) \
3137 while ((t = next_thread(t)) != g)
3139 #define __for_each_thread(signal, t) \
3140 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3142 #define for_each_thread(p, t) \
3143 __for_each_thread((p)->signal, t)
3145 /* Careful: this is a double loop, 'break' won't work as expected. */
3146 #define for_each_process_thread(p, t) \
3147 for_each_process(p) for_each_thread(p, t)
3149 static inline int get_nr_threads(struct task_struct *tsk)
3151 return tsk->signal->nr_threads;
3154 static inline bool thread_group_leader(struct task_struct *p)
3156 return p->exit_signal >= 0;
3159 /* Do to the insanities of de_thread it is possible for a process
3160 * to have the pid of the thread group leader without actually being
3161 * the thread group leader. For iteration through the pids in proc
3162 * all we care about is that we have a task with the appropriate
3163 * pid, we don't actually care if we have the right task.
3165 static inline bool has_group_leader_pid(struct task_struct *p)
3167 return task_pid(p) == p->signal->leader_pid;
3171 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3173 return p1->signal == p2->signal;
3176 static inline struct task_struct *next_thread(const struct task_struct *p)
3178 return list_entry_rcu(p->thread_group.next,
3179 struct task_struct, thread_group);
3182 static inline int thread_group_empty(struct task_struct *p)
3184 return list_empty(&p->thread_group);
3187 #define delay_group_leader(p) \
3188 (thread_group_leader(p) && !thread_group_empty(p))
3191 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3192 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3193 * pins the final release of task.io_context. Also protects ->cpuset and
3194 * ->cgroup.subsys[]. And ->vfork_done.
3196 * Nests both inside and outside of read_lock(&tasklist_lock).
3197 * It must not be nested with write_lock_irq(&tasklist_lock),
3198 * neither inside nor outside.
3200 static inline void task_lock(struct task_struct *p)
3202 spin_lock(&p->alloc_lock);
3205 static inline void task_unlock(struct task_struct *p)
3207 spin_unlock(&p->alloc_lock);
3210 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3211 unsigned long *flags);
3213 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3214 unsigned long *flags)
3216 struct sighand_struct *ret;
3218 ret = __lock_task_sighand(tsk, flags);
3219 (void)__cond_lock(&tsk->sighand->siglock, ret);
3223 static inline void unlock_task_sighand(struct task_struct *tsk,
3224 unsigned long *flags)
3226 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3230 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3231 * @tsk: task causing the changes
3233 * All operations which modify a threadgroup - a new thread joining the
3234 * group, death of a member thread (the assertion of PF_EXITING) and
3235 * exec(2) dethreading the process and replacing the leader - are wrapped
3236 * by threadgroup_change_{begin|end}(). This is to provide a place which
3237 * subsystems needing threadgroup stability can hook into for
3240 static inline void threadgroup_change_begin(struct task_struct *tsk)
3243 cgroup_threadgroup_change_begin(tsk);
3247 * threadgroup_change_end - mark the end of changes to a threadgroup
3248 * @tsk: task causing the changes
3250 * See threadgroup_change_begin().
3252 static inline void threadgroup_change_end(struct task_struct *tsk)
3254 cgroup_threadgroup_change_end(tsk);
3257 #ifdef CONFIG_THREAD_INFO_IN_TASK
3259 static inline struct thread_info *task_thread_info(struct task_struct *task)
3261 return &task->thread_info;
3265 * When accessing the stack of a non-current task that might exit, use
3266 * try_get_task_stack() instead. task_stack_page will return a pointer
3267 * that could get freed out from under you.
3269 static inline void *task_stack_page(const struct task_struct *task)
3274 #define setup_thread_stack(new,old) do { } while(0)
3276 static inline unsigned long *end_of_stack(const struct task_struct *task)
3281 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3283 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3284 #define task_stack_page(task) ((void *)(task)->stack)
3286 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3288 *task_thread_info(p) = *task_thread_info(org);
3289 task_thread_info(p)->task = p;
3293 * Return the address of the last usable long on the stack.
3295 * When the stack grows down, this is just above the thread
3296 * info struct. Going any lower will corrupt the threadinfo.
3298 * When the stack grows up, this is the highest address.
3299 * Beyond that position, we corrupt data on the next page.
3301 static inline unsigned long *end_of_stack(struct task_struct *p)
3303 #ifdef CONFIG_STACK_GROWSUP
3304 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3306 return (unsigned long *)(task_thread_info(p) + 1);
3312 static inline void *try_get_task_stack(struct task_struct *tsk)
3314 return task_stack_page(tsk);
3317 static inline void put_task_stack(struct task_struct *tsk) {}
3319 #define task_stack_end_corrupted(task) \
3320 (*(end_of_stack(task)) != STACK_END_MAGIC)
3322 static inline int object_is_on_stack(void *obj)
3324 void *stack = task_stack_page(current);
3326 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3329 extern void thread_stack_cache_init(void);
3331 #ifdef CONFIG_DEBUG_STACK_USAGE
3332 static inline unsigned long stack_not_used(struct task_struct *p)
3334 unsigned long *n = end_of_stack(p);
3336 do { /* Skip over canary */
3340 return (unsigned long)n - (unsigned long)end_of_stack(p);
3343 extern void set_task_stack_end_magic(struct task_struct *tsk);
3345 /* set thread flags in other task's structures
3346 * - see asm/thread_info.h for TIF_xxxx flags available
3348 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3350 set_ti_thread_flag(task_thread_info(tsk), flag);
3353 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3355 clear_ti_thread_flag(task_thread_info(tsk), flag);
3358 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3360 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3363 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3365 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3368 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3370 return test_ti_thread_flag(task_thread_info(tsk), flag);
3373 static inline void set_tsk_need_resched(struct task_struct *tsk)
3375 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3378 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3380 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3383 static inline int test_tsk_need_resched(struct task_struct *tsk)
3385 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3388 static inline int restart_syscall(void)
3390 set_tsk_thread_flag(current, TIF_SIGPENDING);
3391 return -ERESTARTNOINTR;
3394 static inline int signal_pending(struct task_struct *p)
3396 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3399 static inline int __fatal_signal_pending(struct task_struct *p)
3401 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3404 static inline int fatal_signal_pending(struct task_struct *p)
3406 return signal_pending(p) && __fatal_signal_pending(p);
3409 static inline int signal_pending_state(long state, struct task_struct *p)
3411 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3413 if (!signal_pending(p))
3416 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3420 * cond_resched() and cond_resched_lock(): latency reduction via
3421 * explicit rescheduling in places that are safe. The return
3422 * value indicates whether a reschedule was done in fact.
3423 * cond_resched_lock() will drop the spinlock before scheduling,
3424 * cond_resched_softirq() will enable bhs before scheduling.
3426 extern int _cond_resched(void);
3428 #define cond_resched() ({ \
3429 ___might_sleep(__FILE__, __LINE__, 0); \
3433 extern int __cond_resched_lock(spinlock_t *lock);
3435 #define cond_resched_lock(lock) ({ \
3436 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3437 __cond_resched_lock(lock); \
3440 extern int __cond_resched_softirq(void);
3442 #define cond_resched_softirq() ({ \
3443 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3444 __cond_resched_softirq(); \
3447 static inline void cond_resched_rcu(void)
3449 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3456 static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3458 #ifdef CONFIG_DEBUG_PREEMPT
3459 return p->preempt_disable_ip;
3466 * Does a critical section need to be broken due to another
3467 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3468 * but a general need for low latency)
3470 static inline int spin_needbreak(spinlock_t *lock)
3472 #ifdef CONFIG_PREEMPT
3473 return spin_is_contended(lock);
3480 * Idle thread specific functions to determine the need_resched
3483 #ifdef TIF_POLLING_NRFLAG
3484 static inline int tsk_is_polling(struct task_struct *p)
3486 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3489 static inline void __current_set_polling(void)
3491 set_thread_flag(TIF_POLLING_NRFLAG);
3494 static inline bool __must_check current_set_polling_and_test(void)
3496 __current_set_polling();
3499 * Polling state must be visible before we test NEED_RESCHED,
3500 * paired by resched_curr()
3502 smp_mb__after_atomic();
3504 return unlikely(tif_need_resched());
3507 static inline void __current_clr_polling(void)
3509 clear_thread_flag(TIF_POLLING_NRFLAG);
3512 static inline bool __must_check current_clr_polling_and_test(void)
3514 __current_clr_polling();
3517 * Polling state must be visible before we test NEED_RESCHED,
3518 * paired by resched_curr()
3520 smp_mb__after_atomic();
3522 return unlikely(tif_need_resched());
3526 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3527 static inline void __current_set_polling(void) { }
3528 static inline void __current_clr_polling(void) { }
3530 static inline bool __must_check current_set_polling_and_test(void)
3532 return unlikely(tif_need_resched());
3534 static inline bool __must_check current_clr_polling_and_test(void)
3536 return unlikely(tif_need_resched());
3540 static inline void current_clr_polling(void)
3542 __current_clr_polling();
3545 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3546 * Once the bit is cleared, we'll get IPIs with every new
3547 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3550 smp_mb(); /* paired with resched_curr() */
3552 preempt_fold_need_resched();
3555 static __always_inline bool need_resched(void)
3557 return unlikely(tif_need_resched());
3561 * Thread group CPU time accounting.
3563 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3564 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3567 * Reevaluate whether the task has signals pending delivery.
3568 * Wake the task if so.
3569 * This is required every time the blocked sigset_t changes.
3570 * callers must hold sighand->siglock.
3572 extern void recalc_sigpending_and_wake(struct task_struct *t);
3573 extern void recalc_sigpending(void);
3575 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3577 static inline void signal_wake_up(struct task_struct *t, bool resume)
3579 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3581 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3583 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3587 * Wrappers for p->thread_info->cpu access. No-op on UP.
3591 static inline unsigned int task_cpu(const struct task_struct *p)
3593 #ifdef CONFIG_THREAD_INFO_IN_TASK
3596 return task_thread_info(p)->cpu;
3600 static inline int task_node(const struct task_struct *p)
3602 return cpu_to_node(task_cpu(p));
3605 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3609 static inline unsigned int task_cpu(const struct task_struct *p)
3614 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3618 #endif /* CONFIG_SMP */
3620 extern struct atomic_notifier_head migration_notifier_head;
3621 struct migration_notify_data {
3627 extern struct atomic_notifier_head load_alert_notifier_head;
3629 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3630 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3632 #ifdef CONFIG_CGROUP_SCHED
3633 extern struct task_group root_task_group;
3634 #endif /* CONFIG_CGROUP_SCHED */
3636 extern int task_can_switch_user(struct user_struct *up,
3637 struct task_struct *tsk);
3639 #ifdef CONFIG_TASK_XACCT
3640 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3642 tsk->ioac.rchar += amt;
3645 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3647 tsk->ioac.wchar += amt;
3650 static inline void inc_syscr(struct task_struct *tsk)
3655 static inline void inc_syscw(struct task_struct *tsk)
3660 static inline void inc_syscfs(struct task_struct *tsk)
3665 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3669 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3673 static inline void inc_syscr(struct task_struct *tsk)
3677 static inline void inc_syscw(struct task_struct *tsk)
3680 static inline void inc_syscfs(struct task_struct *tsk)
3685 #ifndef TASK_SIZE_OF
3686 #define TASK_SIZE_OF(tsk) TASK_SIZE
3690 extern void mm_update_next_owner(struct mm_struct *mm);
3692 static inline void mm_update_next_owner(struct mm_struct *mm)
3695 #endif /* CONFIG_MEMCG */
3697 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3700 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3703 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3706 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3709 static inline unsigned long rlimit(unsigned int limit)
3711 return task_rlimit(current, limit);
3714 static inline unsigned long rlimit_max(unsigned int limit)
3716 return task_rlimit_max(current, limit);
3719 #define SCHED_CPUFREQ_RT (1U << 0)
3720 #define SCHED_CPUFREQ_DL (1U << 1)
3721 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3723 #ifdef CONFIG_CPU_FREQ
3724 struct update_util_data {
3725 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3728 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3729 void (*func)(struct update_util_data *data, u64 time,
3730 unsigned int flags));
3731 void cpufreq_remove_update_util_hook(int cpu);
3732 #endif /* CONFIG_CPU_FREQ */