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/seccomp.h>
44 #include <linux/rcupdate.h>
45 #include <linux/rculist.h>
46 #include <linux/rtmutex.h>
48 #include <linux/time.h>
49 #include <linux/param.h>
50 #include <linux/resource.h>
51 #include <linux/timer.h>
52 #include <linux/hrtimer.h>
53 #include <linux/kcov.h>
54 #include <linux/task_io_accounting.h>
55 #include <linux/latencytop.h>
56 #include <linux/cred.h>
57 #include <linux/llist.h>
58 #include <linux/uidgid.h>
59 #include <linux/gfp.h>
60 #include <linux/magic.h>
61 #include <linux/cgroup-defs.h>
63 #include <asm/processor.h>
65 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
68 * Extended scheduling parameters data structure.
70 * This is needed because the original struct sched_param can not be
71 * altered without introducing ABI issues with legacy applications
72 * (e.g., in sched_getparam()).
74 * However, the possibility of specifying more than just a priority for
75 * the tasks may be useful for a wide variety of application fields, e.g.,
76 * multimedia, streaming, automation and control, and many others.
78 * This variant (sched_attr) is meant at describing a so-called
79 * sporadic time-constrained task. In such model a task is specified by:
80 * - the activation period or minimum instance inter-arrival time;
81 * - the maximum (or average, depending on the actual scheduling
82 * discipline) computation time of all instances, a.k.a. runtime;
83 * - the deadline (relative to the actual activation time) of each
85 * Very briefly, a periodic (sporadic) task asks for the execution of
86 * some specific computation --which is typically called an instance--
87 * (at most) every period. Moreover, each instance typically lasts no more
88 * than the runtime and must be completed by time instant t equal to
89 * the instance activation time + the deadline.
91 * This is reflected by the actual fields of the sched_attr structure:
93 * @size size of the structure, for fwd/bwd compat.
95 * @sched_policy task's scheduling policy
96 * @sched_flags for customizing the scheduler behaviour
97 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
98 * @sched_priority task's static priority (SCHED_FIFO/RR)
99 * @sched_deadline representative of the task's deadline
100 * @sched_runtime representative of the task's runtime
101 * @sched_period representative of the task's period
103 * Given this task model, there are a multiplicity of scheduling algorithms
104 * and policies, that can be used to ensure all the tasks will make their
105 * timing constraints.
107 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
108 * only user of this new interface. More information about the algorithm
109 * available in the scheduling class file or in Documentation/.
117 /* SCHED_NORMAL, SCHED_BATCH */
120 /* SCHED_FIFO, SCHED_RR */
129 struct futex_pi_state;
130 struct robust_list_head;
133 struct perf_event_context;
138 #define VMACACHE_BITS 2
139 #define VMACACHE_SIZE (1U << VMACACHE_BITS)
140 #define VMACACHE_MASK (VMACACHE_SIZE - 1)
143 * These are the constant used to fake the fixed-point load-average
144 * counting. Some notes:
145 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
146 * a load-average precision of 10 bits integer + 11 bits fractional
147 * - if you want to count load-averages more often, you need more
148 * precision, or rounding will get you. With 2-second counting freq,
149 * the EXP_n values would be 1981, 2034 and 2043 if still using only
152 extern unsigned long avenrun[]; /* Load averages */
153 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
155 #define FSHIFT 11 /* nr of bits of precision */
156 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
157 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
158 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
159 #define EXP_5 2014 /* 1/exp(5sec/5min) */
160 #define EXP_15 2037 /* 1/exp(5sec/15min) */
162 #define CALC_LOAD(load,exp,n) \
164 load += n*(FIXED_1-exp); \
167 extern unsigned long total_forks;
168 extern int nr_threads;
169 DECLARE_PER_CPU(unsigned long, process_counts);
170 extern int nr_processes(void);
171 extern unsigned long nr_running(void);
172 extern bool single_task_running(void);
173 extern unsigned long nr_iowait(void);
174 extern unsigned long nr_iowait_cpu(int cpu);
175 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
177 extern void calc_global_load(unsigned long ticks);
179 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
180 extern void cpu_load_update_nohz_start(void);
181 extern void cpu_load_update_nohz_stop(void);
183 static inline void cpu_load_update_nohz_start(void) { }
184 static inline void cpu_load_update_nohz_stop(void) { }
187 extern void dump_cpu_task(int cpu);
192 #ifdef CONFIG_SCHED_DEBUG
193 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
194 extern void proc_sched_set_task(struct task_struct *p);
198 * Task state bitmask. NOTE! These bits are also
199 * encoded in fs/proc/array.c: get_task_state().
201 * We have two separate sets of flags: task->state
202 * is about runnability, while task->exit_state are
203 * about the task exiting. Confusing, but this way
204 * modifying one set can't modify the other one by
207 #define TASK_RUNNING 0
208 #define TASK_INTERRUPTIBLE 1
209 #define TASK_UNINTERRUPTIBLE 2
210 #define __TASK_STOPPED 4
211 #define __TASK_TRACED 8
212 /* in tsk->exit_state */
214 #define EXIT_ZOMBIE 32
215 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
216 /* in tsk->state again */
218 #define TASK_WAKEKILL 128
219 #define TASK_WAKING 256
220 #define TASK_PARKED 512
221 #define TASK_NOLOAD 1024
222 #define TASK_NEW 2048
223 #define TASK_STATE_MAX 4096
225 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
227 extern char ___assert_task_state[1 - 2*!!(
228 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
230 /* Convenience macros for the sake of set_task_state */
231 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
232 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
233 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
235 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
237 /* Convenience macros for the sake of wake_up */
238 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
239 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
241 /* get_task_state() */
242 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
243 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
244 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
246 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
247 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
248 #define task_is_stopped_or_traced(task) \
249 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
250 #define task_contributes_to_load(task) \
251 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
252 (task->flags & PF_FROZEN) == 0 && \
253 (task->state & TASK_NOLOAD) == 0)
255 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
257 #define __set_task_state(tsk, state_value) \
259 (tsk)->task_state_change = _THIS_IP_; \
260 (tsk)->state = (state_value); \
262 #define set_task_state(tsk, state_value) \
264 (tsk)->task_state_change = _THIS_IP_; \
265 smp_store_mb((tsk)->state, (state_value)); \
269 * set_current_state() includes a barrier so that the write of current->state
270 * is correctly serialised wrt the caller's subsequent test of whether to
273 * set_current_state(TASK_UNINTERRUPTIBLE);
274 * if (do_i_need_to_sleep())
277 * If the caller does not need such serialisation then use __set_current_state()
279 #define __set_current_state(state_value) \
281 current->task_state_change = _THIS_IP_; \
282 current->state = (state_value); \
284 #define set_current_state(state_value) \
286 current->task_state_change = _THIS_IP_; \
287 smp_store_mb(current->state, (state_value)); \
292 #define __set_task_state(tsk, state_value) \
293 do { (tsk)->state = (state_value); } while (0)
294 #define set_task_state(tsk, state_value) \
295 smp_store_mb((tsk)->state, (state_value))
298 * set_current_state() includes a barrier so that the write of current->state
299 * is correctly serialised wrt the caller's subsequent test of whether to
302 * set_current_state(TASK_UNINTERRUPTIBLE);
303 * if (do_i_need_to_sleep())
306 * If the caller does not need such serialisation then use __set_current_state()
308 #define __set_current_state(state_value) \
309 do { current->state = (state_value); } while (0)
310 #define set_current_state(state_value) \
311 smp_store_mb(current->state, (state_value))
315 /* Task command name length */
316 #define TASK_COMM_LEN 16
318 #include <linux/spinlock.h>
321 * This serializes "schedule()" and also protects
322 * the run-queue from deletions/modifications (but
323 * _adding_ to the beginning of the run-queue has
326 extern rwlock_t tasklist_lock;
327 extern spinlock_t mmlist_lock;
331 #ifdef CONFIG_PROVE_RCU
332 extern int lockdep_tasklist_lock_is_held(void);
333 #endif /* #ifdef CONFIG_PROVE_RCU */
335 extern void sched_init(void);
336 extern void sched_init_smp(void);
337 extern asmlinkage void schedule_tail(struct task_struct *prev);
338 extern void init_idle(struct task_struct *idle, int cpu);
339 extern void init_idle_bootup_task(struct task_struct *idle);
341 extern cpumask_var_t cpu_isolated_map;
343 extern int runqueue_is_locked(int cpu);
345 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
346 extern void nohz_balance_enter_idle(int cpu);
347 extern void set_cpu_sd_state_idle(void);
348 extern int get_nohz_timer_target(void);
350 static inline void nohz_balance_enter_idle(int cpu) { }
351 static inline void set_cpu_sd_state_idle(void) { }
355 * Only dump TASK_* tasks. (0 for all tasks)
357 extern void show_state_filter(unsigned long state_filter);
359 static inline void show_state(void)
361 show_state_filter(0);
364 extern void show_regs(struct pt_regs *);
367 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
368 * task), SP is the stack pointer of the first frame that should be shown in the back
369 * trace (or NULL if the entire call-chain of the task should be shown).
371 extern void show_stack(struct task_struct *task, unsigned long *sp);
373 extern void cpu_init (void);
374 extern void trap_init(void);
375 extern void update_process_times(int user);
376 extern void scheduler_tick(void);
377 extern int sched_cpu_starting(unsigned int cpu);
378 extern int sched_cpu_activate(unsigned int cpu);
379 extern int sched_cpu_deactivate(unsigned int cpu);
381 #ifdef CONFIG_HOTPLUG_CPU
382 extern int sched_cpu_dying(unsigned int cpu);
384 # define sched_cpu_dying NULL
387 extern void sched_show_task(struct task_struct *p);
389 #ifdef CONFIG_LOCKUP_DETECTOR
390 extern void touch_softlockup_watchdog_sched(void);
391 extern void touch_softlockup_watchdog(void);
392 extern void touch_softlockup_watchdog_sync(void);
393 extern void touch_all_softlockup_watchdogs(void);
394 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
396 size_t *lenp, loff_t *ppos);
397 extern unsigned int softlockup_panic;
398 extern unsigned int hardlockup_panic;
399 void lockup_detector_init(void);
401 static inline void touch_softlockup_watchdog_sched(void)
404 static inline void touch_softlockup_watchdog(void)
407 static inline void touch_softlockup_watchdog_sync(void)
410 static inline void touch_all_softlockup_watchdogs(void)
413 static inline void lockup_detector_init(void)
418 #ifdef CONFIG_DETECT_HUNG_TASK
419 void reset_hung_task_detector(void);
421 static inline void reset_hung_task_detector(void)
426 /* Attach to any functions which should be ignored in wchan output. */
427 #define __sched __attribute__((__section__(".sched.text")))
429 /* Linker adds these: start and end of __sched functions */
430 extern char __sched_text_start[], __sched_text_end[];
432 /* Is this address in the __sched functions? */
433 extern int in_sched_functions(unsigned long addr);
435 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
436 extern signed long schedule_timeout(signed long timeout);
437 extern signed long schedule_timeout_interruptible(signed long timeout);
438 extern signed long schedule_timeout_killable(signed long timeout);
439 extern signed long schedule_timeout_uninterruptible(signed long timeout);
440 extern signed long schedule_timeout_idle(signed long timeout);
441 asmlinkage void schedule(void);
442 extern void schedule_preempt_disabled(void);
444 extern long io_schedule_timeout(long timeout);
446 static inline void io_schedule(void)
448 io_schedule_timeout(MAX_SCHEDULE_TIMEOUT);
451 void __noreturn do_task_dead(void);
454 struct user_namespace;
457 extern void arch_pick_mmap_layout(struct mm_struct *mm);
459 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
460 unsigned long, unsigned long);
462 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
463 unsigned long len, unsigned long pgoff,
464 unsigned long flags);
466 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
469 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
470 #define SUID_DUMP_USER 1 /* Dump as user of process */
471 #define SUID_DUMP_ROOT 2 /* Dump as root */
475 /* for SUID_DUMP_* above */
476 #define MMF_DUMPABLE_BITS 2
477 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
479 extern void set_dumpable(struct mm_struct *mm, int value);
481 * This returns the actual value of the suid_dumpable flag. For things
482 * that are using this for checking for privilege transitions, it must
483 * test against SUID_DUMP_USER rather than treating it as a boolean
486 static inline int __get_dumpable(unsigned long mm_flags)
488 return mm_flags & MMF_DUMPABLE_MASK;
491 static inline int get_dumpable(struct mm_struct *mm)
493 return __get_dumpable(mm->flags);
496 /* coredump filter bits */
497 #define MMF_DUMP_ANON_PRIVATE 2
498 #define MMF_DUMP_ANON_SHARED 3
499 #define MMF_DUMP_MAPPED_PRIVATE 4
500 #define MMF_DUMP_MAPPED_SHARED 5
501 #define MMF_DUMP_ELF_HEADERS 6
502 #define MMF_DUMP_HUGETLB_PRIVATE 7
503 #define MMF_DUMP_HUGETLB_SHARED 8
504 #define MMF_DUMP_DAX_PRIVATE 9
505 #define MMF_DUMP_DAX_SHARED 10
507 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
508 #define MMF_DUMP_FILTER_BITS 9
509 #define MMF_DUMP_FILTER_MASK \
510 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
511 #define MMF_DUMP_FILTER_DEFAULT \
512 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
513 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
515 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
516 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
518 # define MMF_DUMP_MASK_DEFAULT_ELF 0
520 /* leave room for more dump flags */
521 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
522 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
523 #define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
525 #define MMF_HAS_UPROBES 19 /* has uprobes */
526 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
527 #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
529 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
531 struct sighand_struct {
533 struct k_sigaction action[_NSIG];
535 wait_queue_head_t signalfd_wqh;
538 struct pacct_struct {
541 unsigned long ac_mem;
542 cputime_t ac_utime, ac_stime;
543 unsigned long ac_minflt, ac_majflt;
554 * struct prev_cputime - snaphsot of system and user cputime
555 * @utime: time spent in user mode
556 * @stime: time spent in system mode
557 * @lock: protects the above two fields
559 * Stores previous user/system time values such that we can guarantee
562 struct prev_cputime {
563 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
570 static inline void prev_cputime_init(struct prev_cputime *prev)
572 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
573 prev->utime = prev->stime = 0;
574 raw_spin_lock_init(&prev->lock);
579 * struct task_cputime - collected CPU time counts
580 * @utime: time spent in user mode, in &cputime_t units
581 * @stime: time spent in kernel mode, in &cputime_t units
582 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
584 * This structure groups together three kinds of CPU time that are tracked for
585 * threads and thread groups. Most things considering CPU time want to group
586 * these counts together and treat all three of them in parallel.
588 struct task_cputime {
591 unsigned long long sum_exec_runtime;
594 /* Alternate field names when used to cache expirations. */
595 #define virt_exp utime
596 #define prof_exp stime
597 #define sched_exp sum_exec_runtime
599 #define INIT_CPUTIME \
600 (struct task_cputime) { \
603 .sum_exec_runtime = 0, \
607 * This is the atomic variant of task_cputime, which can be used for
608 * storing and updating task_cputime statistics without locking.
610 struct task_cputime_atomic {
613 atomic64_t sum_exec_runtime;
616 #define INIT_CPUTIME_ATOMIC \
617 (struct task_cputime_atomic) { \
618 .utime = ATOMIC64_INIT(0), \
619 .stime = ATOMIC64_INIT(0), \
620 .sum_exec_runtime = ATOMIC64_INIT(0), \
623 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
626 * Disable preemption until the scheduler is running -- use an unconditional
627 * value so that it also works on !PREEMPT_COUNT kernels.
629 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
631 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
634 * Initial preempt_count value; reflects the preempt_count schedule invariant
635 * which states that during context switches:
637 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
639 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
640 * Note: See finish_task_switch().
642 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
645 * struct thread_group_cputimer - thread group interval timer counts
646 * @cputime_atomic: atomic thread group interval timers.
647 * @running: true when there are timers running and
648 * @cputime_atomic receives updates.
649 * @checking_timer: true when a thread in the group is in the
650 * process of checking for thread group timers.
652 * This structure contains the version of task_cputime, above, that is
653 * used for thread group CPU timer calculations.
655 struct thread_group_cputimer {
656 struct task_cputime_atomic cputime_atomic;
661 #include <linux/rwsem.h>
665 * NOTE! "signal_struct" does not have its own
666 * locking, because a shared signal_struct always
667 * implies a shared sighand_struct, so locking
668 * sighand_struct is always a proper superset of
669 * the locking of signal_struct.
671 struct signal_struct {
675 struct list_head thread_head;
677 wait_queue_head_t wait_chldexit; /* for wait4() */
679 /* current thread group signal load-balancing target: */
680 struct task_struct *curr_target;
682 /* shared signal handling: */
683 struct sigpending shared_pending;
685 /* thread group exit support */
688 * - notify group_exit_task when ->count is equal to notify_count
689 * - everyone except group_exit_task is stopped during signal delivery
690 * of fatal signals, group_exit_task processes the signal.
693 struct task_struct *group_exit_task;
695 /* thread group stop support, overloads group_exit_code too */
696 int group_stop_count;
697 unsigned int flags; /* see SIGNAL_* flags below */
700 * PR_SET_CHILD_SUBREAPER marks a process, like a service
701 * manager, to re-parent orphan (double-forking) child processes
702 * to this process instead of 'init'. The service manager is
703 * able to receive SIGCHLD signals and is able to investigate
704 * the process until it calls wait(). All children of this
705 * process will inherit a flag if they should look for a
706 * child_subreaper process at exit.
708 unsigned int is_child_subreaper:1;
709 unsigned int has_child_subreaper:1;
711 /* POSIX.1b Interval Timers */
713 struct list_head posix_timers;
715 /* ITIMER_REAL timer for the process */
716 struct hrtimer real_timer;
717 struct pid *leader_pid;
718 ktime_t it_real_incr;
721 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
722 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
723 * values are defined to 0 and 1 respectively
725 struct cpu_itimer it[2];
728 * Thread group totals for process CPU timers.
729 * See thread_group_cputimer(), et al, for details.
731 struct thread_group_cputimer cputimer;
733 /* Earliest-expiration cache. */
734 struct task_cputime cputime_expires;
736 #ifdef CONFIG_NO_HZ_FULL
737 atomic_t tick_dep_mask;
740 struct list_head cpu_timers[3];
742 struct pid *tty_old_pgrp;
744 /* boolean value for session group leader */
747 struct tty_struct *tty; /* NULL if no tty */
749 #ifdef CONFIG_SCHED_AUTOGROUP
750 struct autogroup *autogroup;
753 * Cumulative resource counters for dead threads in the group,
754 * and for reaped dead child processes forked by this group.
755 * Live threads maintain their own counters and add to these
756 * in __exit_signal, except for the group leader.
758 seqlock_t stats_lock;
759 cputime_t utime, stime, cutime, cstime;
762 struct prev_cputime prev_cputime;
763 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
764 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
765 unsigned long inblock, oublock, cinblock, coublock;
766 unsigned long maxrss, cmaxrss;
767 struct task_io_accounting ioac;
770 * Cumulative ns of schedule CPU time fo dead threads in the
771 * group, not including a zombie group leader, (This only differs
772 * from jiffies_to_ns(utime + stime) if sched_clock uses something
773 * other than jiffies.)
775 unsigned long long sum_sched_runtime;
778 * We don't bother to synchronize most readers of this at all,
779 * because there is no reader checking a limit that actually needs
780 * to get both rlim_cur and rlim_max atomically, and either one
781 * alone is a single word that can safely be read normally.
782 * getrlimit/setrlimit use task_lock(current->group_leader) to
783 * protect this instead of the siglock, because they really
784 * have no need to disable irqs.
786 struct rlimit rlim[RLIM_NLIMITS];
788 #ifdef CONFIG_BSD_PROCESS_ACCT
789 struct pacct_struct pacct; /* per-process accounting information */
791 #ifdef CONFIG_TASKSTATS
792 struct taskstats *stats;
796 struct tty_audit_buf *tty_audit_buf;
800 * Thread is the potential origin of an oom condition; kill first on
803 bool oom_flag_origin;
804 short oom_score_adj; /* OOM kill score adjustment */
805 short oom_score_adj_min; /* OOM kill score adjustment min value.
806 * Only settable by CAP_SYS_RESOURCE. */
807 struct mm_struct *oom_mm; /* recorded mm when the thread group got
808 * killed by the oom killer */
810 struct mutex cred_guard_mutex; /* guard against foreign influences on
811 * credential calculations
812 * (notably. ptrace) */
816 * Bits in flags field of signal_struct.
818 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
819 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
820 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
821 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
823 * Pending notifications to parent.
825 #define SIGNAL_CLD_STOPPED 0x00000010
826 #define SIGNAL_CLD_CONTINUED 0x00000020
827 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
829 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
831 /* If true, all threads except ->group_exit_task have pending SIGKILL */
832 static inline int signal_group_exit(const struct signal_struct *sig)
834 return (sig->flags & SIGNAL_GROUP_EXIT) ||
835 (sig->group_exit_task != NULL);
839 * Some day this will be a full-fledged user tracking system..
842 atomic_t __count; /* reference count */
843 atomic_t processes; /* How many processes does this user have? */
844 atomic_t sigpending; /* How many pending signals does this user have? */
845 #ifdef CONFIG_INOTIFY_USER
846 atomic_t inotify_watches; /* How many inotify watches does this user have? */
847 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
849 #ifdef CONFIG_FANOTIFY
850 atomic_t fanotify_listeners;
853 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
855 #ifdef CONFIG_POSIX_MQUEUE
856 /* protected by mq_lock */
857 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
859 unsigned long locked_shm; /* How many pages of mlocked shm ? */
860 unsigned long unix_inflight; /* How many files in flight in unix sockets */
861 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
864 struct key *uid_keyring; /* UID specific keyring */
865 struct key *session_keyring; /* UID's default session keyring */
868 /* Hash table maintenance information */
869 struct hlist_node uidhash_node;
872 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
873 atomic_long_t locked_vm;
877 extern int uids_sysfs_init(void);
879 extern struct user_struct *find_user(kuid_t);
881 extern struct user_struct root_user;
882 #define INIT_USER (&root_user)
885 struct backing_dev_info;
886 struct reclaim_state;
888 #ifdef CONFIG_SCHED_INFO
890 /* cumulative counters */
891 unsigned long pcount; /* # of times run on this cpu */
892 unsigned long long run_delay; /* time spent waiting on a runqueue */
895 unsigned long long last_arrival,/* when we last ran on a cpu */
896 last_queued; /* when we were last queued to run */
898 #endif /* CONFIG_SCHED_INFO */
900 #ifdef CONFIG_TASK_DELAY_ACCT
901 struct task_delay_info {
903 unsigned int flags; /* Private per-task flags */
905 /* For each stat XXX, add following, aligned appropriately
907 * struct timespec XXX_start, XXX_end;
911 * Atomicity of updates to XXX_delay, XXX_count protected by
912 * single lock above (split into XXX_lock if contention is an issue).
916 * XXX_count is incremented on every XXX operation, the delay
917 * associated with the operation is added to XXX_delay.
918 * XXX_delay contains the accumulated delay time in nanoseconds.
920 u64 blkio_start; /* Shared by blkio, swapin */
921 u64 blkio_delay; /* wait for sync block io completion */
922 u64 swapin_delay; /* wait for swapin block io completion */
923 u32 blkio_count; /* total count of the number of sync block */
924 /* io operations performed */
925 u32 swapin_count; /* total count of the number of swapin block */
926 /* io operations performed */
929 u64 freepages_delay; /* wait for memory reclaim */
930 u32 freepages_count; /* total count of memory reclaim */
932 #endif /* CONFIG_TASK_DELAY_ACCT */
934 static inline int sched_info_on(void)
936 #ifdef CONFIG_SCHEDSTATS
938 #elif defined(CONFIG_TASK_DELAY_ACCT)
939 extern int delayacct_on;
946 #ifdef CONFIG_SCHEDSTATS
947 void force_schedstat_enabled(void);
958 * Integer metrics need fixed point arithmetic, e.g., sched/fair
959 * has a few: load, load_avg, util_avg, freq, and capacity.
961 * We define a basic fixed point arithmetic range, and then formalize
962 * all these metrics based on that basic range.
964 # define SCHED_FIXEDPOINT_SHIFT 10
965 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
968 * Increase resolution of cpu_capacity calculations
970 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
971 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
974 * Wake-queues are lists of tasks with a pending wakeup, whose
975 * callers have already marked the task as woken internally,
976 * and can thus carry on. A common use case is being able to
977 * do the wakeups once the corresponding user lock as been
980 * We hold reference to each task in the list across the wakeup,
981 * thus guaranteeing that the memory is still valid by the time
982 * the actual wakeups are performed in wake_up_q().
984 * One per task suffices, because there's never a need for a task to be
985 * in two wake queues simultaneously; it is forbidden to abandon a task
986 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
987 * already in a wake queue, the wakeup will happen soon and the second
988 * waker can just skip it.
990 * The WAKE_Q macro declares and initializes the list head.
991 * wake_up_q() does NOT reinitialize the list; it's expected to be
992 * called near the end of a function, where the fact that the queue is
993 * not used again will be easy to see by inspection.
995 * Note that this can cause spurious wakeups. schedule() callers
996 * must ensure the call is done inside a loop, confirming that the
997 * wakeup condition has in fact occurred.
1000 struct wake_q_node *next;
1003 struct wake_q_head {
1004 struct wake_q_node *first;
1005 struct wake_q_node **lastp;
1008 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1010 #define WAKE_Q(name) \
1011 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1013 extern void wake_q_add(struct wake_q_head *head,
1014 struct task_struct *task);
1015 extern void wake_up_q(struct wake_q_head *head);
1018 * sched-domains (multiprocessor balancing) declarations:
1021 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1022 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1023 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1024 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1025 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1026 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1027 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1028 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1029 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1030 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1031 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1032 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1033 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1034 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1035 #define SD_NUMA 0x4000 /* cross-node balancing */
1037 #ifdef CONFIG_SCHED_SMT
1038 static inline int cpu_smt_flags(void)
1040 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1044 #ifdef CONFIG_SCHED_MC
1045 static inline int cpu_core_flags(void)
1047 return SD_SHARE_PKG_RESOURCES;
1052 static inline int cpu_numa_flags(void)
1058 struct sched_domain_attr {
1059 int relax_domain_level;
1062 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1063 .relax_domain_level = -1, \
1066 extern int sched_domain_level_max;
1070 struct sched_domain_shared {
1072 atomic_t nr_busy_cpus;
1076 struct sched_domain {
1077 /* These fields must be setup */
1078 struct sched_domain *parent; /* top domain must be null terminated */
1079 struct sched_domain *child; /* bottom domain must be null terminated */
1080 struct sched_group *groups; /* the balancing groups of the domain */
1081 unsigned long min_interval; /* Minimum balance interval ms */
1082 unsigned long max_interval; /* Maximum balance interval ms */
1083 unsigned int busy_factor; /* less balancing by factor if busy */
1084 unsigned int imbalance_pct; /* No balance until over watermark */
1085 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1086 unsigned int busy_idx;
1087 unsigned int idle_idx;
1088 unsigned int newidle_idx;
1089 unsigned int wake_idx;
1090 unsigned int forkexec_idx;
1091 unsigned int smt_gain;
1093 int nohz_idle; /* NOHZ IDLE status */
1094 int flags; /* See SD_* */
1097 /* Runtime fields. */
1098 unsigned long last_balance; /* init to jiffies. units in jiffies */
1099 unsigned int balance_interval; /* initialise to 1. units in ms. */
1100 unsigned int nr_balance_failed; /* initialise to 0 */
1102 /* idle_balance() stats */
1103 u64 max_newidle_lb_cost;
1104 unsigned long next_decay_max_lb_cost;
1106 u64 avg_scan_cost; /* select_idle_sibling */
1108 #ifdef CONFIG_SCHEDSTATS
1109 /* load_balance() stats */
1110 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1111 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1112 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1113 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1114 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1115 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1116 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1117 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1119 /* Active load balancing */
1120 unsigned int alb_count;
1121 unsigned int alb_failed;
1122 unsigned int alb_pushed;
1124 /* SD_BALANCE_EXEC stats */
1125 unsigned int sbe_count;
1126 unsigned int sbe_balanced;
1127 unsigned int sbe_pushed;
1129 /* SD_BALANCE_FORK stats */
1130 unsigned int sbf_count;
1131 unsigned int sbf_balanced;
1132 unsigned int sbf_pushed;
1134 /* try_to_wake_up() stats */
1135 unsigned int ttwu_wake_remote;
1136 unsigned int ttwu_move_affine;
1137 unsigned int ttwu_move_balance;
1139 #ifdef CONFIG_SCHED_DEBUG
1143 void *private; /* used during construction */
1144 struct rcu_head rcu; /* used during destruction */
1146 struct sched_domain_shared *shared;
1148 unsigned int span_weight;
1150 * Span of all CPUs in this domain.
1152 * NOTE: this field is variable length. (Allocated dynamically
1153 * by attaching extra space to the end of the structure,
1154 * depending on how many CPUs the kernel has booted up with)
1156 unsigned long span[0];
1159 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1161 return to_cpumask(sd->span);
1164 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1165 struct sched_domain_attr *dattr_new);
1167 /* Allocate an array of sched domains, for partition_sched_domains(). */
1168 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1169 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1171 bool cpus_share_cache(int this_cpu, int that_cpu);
1173 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1174 typedef int (*sched_domain_flags_f)(void);
1176 #define SDTL_OVERLAP 0x01
1179 struct sched_domain **__percpu sd;
1180 struct sched_domain_shared **__percpu sds;
1181 struct sched_group **__percpu sg;
1182 struct sched_group_capacity **__percpu sgc;
1185 struct sched_domain_topology_level {
1186 sched_domain_mask_f mask;
1187 sched_domain_flags_f sd_flags;
1190 struct sd_data data;
1191 #ifdef CONFIG_SCHED_DEBUG
1196 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1197 extern void wake_up_if_idle(int cpu);
1199 #ifdef CONFIG_SCHED_DEBUG
1200 # define SD_INIT_NAME(type) .name = #type
1202 # define SD_INIT_NAME(type)
1205 #else /* CONFIG_SMP */
1207 struct sched_domain_attr;
1210 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1211 struct sched_domain_attr *dattr_new)
1215 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1220 #endif /* !CONFIG_SMP */
1223 struct io_context; /* See blkdev.h */
1226 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1227 extern void prefetch_stack(struct task_struct *t);
1229 static inline void prefetch_stack(struct task_struct *t) { }
1232 struct audit_context; /* See audit.c */
1234 struct pipe_inode_info;
1235 struct uts_namespace;
1237 struct load_weight {
1238 unsigned long weight;
1243 * The load_avg/util_avg accumulates an infinite geometric series
1244 * (see __update_load_avg() in kernel/sched/fair.c).
1246 * [load_avg definition]
1248 * load_avg = runnable% * scale_load_down(load)
1250 * where runnable% is the time ratio that a sched_entity is runnable.
1251 * For cfs_rq, it is the aggregated load_avg of all runnable and
1252 * blocked sched_entities.
1254 * load_avg may also take frequency scaling into account:
1256 * load_avg = runnable% * scale_load_down(load) * freq%
1258 * where freq% is the CPU frequency normalized to the highest frequency.
1260 * [util_avg definition]
1262 * util_avg = running% * SCHED_CAPACITY_SCALE
1264 * where running% is the time ratio that a sched_entity is running on
1265 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1266 * and blocked sched_entities.
1268 * util_avg may also factor frequency scaling and CPU capacity scaling:
1270 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1272 * where freq% is the same as above, and capacity% is the CPU capacity
1273 * normalized to the greatest capacity (due to uarch differences, etc).
1275 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1276 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1277 * we therefore scale them to as large a range as necessary. This is for
1278 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1282 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1283 * with the highest load (=88761), always runnable on a single cfs_rq,
1284 * and should not overflow as the number already hits PID_MAX_LIMIT.
1286 * For all other cases (including 32-bit kernels), struct load_weight's
1287 * weight will overflow first before we do, because:
1289 * Max(load_avg) <= Max(load.weight)
1291 * Then it is the load_weight's responsibility to consider overflow
1295 u64 last_update_time, load_sum;
1296 u32 util_sum, period_contrib;
1297 unsigned long load_avg, util_avg;
1300 #ifdef CONFIG_SCHEDSTATS
1301 struct sched_statistics {
1311 s64 sum_sleep_runtime;
1318 u64 nr_migrations_cold;
1319 u64 nr_failed_migrations_affine;
1320 u64 nr_failed_migrations_running;
1321 u64 nr_failed_migrations_hot;
1322 u64 nr_forced_migrations;
1325 u64 nr_wakeups_sync;
1326 u64 nr_wakeups_migrate;
1327 u64 nr_wakeups_local;
1328 u64 nr_wakeups_remote;
1329 u64 nr_wakeups_affine;
1330 u64 nr_wakeups_affine_attempts;
1331 u64 nr_wakeups_passive;
1332 u64 nr_wakeups_idle;
1336 struct sched_entity {
1337 struct load_weight load; /* for load-balancing */
1338 struct rb_node run_node;
1339 struct list_head group_node;
1343 u64 sum_exec_runtime;
1345 u64 prev_sum_exec_runtime;
1349 #ifdef CONFIG_SCHEDSTATS
1350 struct sched_statistics statistics;
1353 #ifdef CONFIG_FAIR_GROUP_SCHED
1355 struct sched_entity *parent;
1356 /* rq on which this entity is (to be) queued: */
1357 struct cfs_rq *cfs_rq;
1358 /* rq "owned" by this entity/group: */
1359 struct cfs_rq *my_q;
1364 * Per entity load average tracking.
1366 * Put into separate cache line so it does not
1367 * collide with read-mostly values above.
1369 struct sched_avg avg ____cacheline_aligned_in_smp;
1373 struct sched_rt_entity {
1374 struct list_head run_list;
1375 unsigned long timeout;
1376 unsigned long watchdog_stamp;
1377 unsigned int time_slice;
1378 unsigned short on_rq;
1379 unsigned short on_list;
1381 struct sched_rt_entity *back;
1382 #ifdef CONFIG_RT_GROUP_SCHED
1383 struct sched_rt_entity *parent;
1384 /* rq on which this entity is (to be) queued: */
1385 struct rt_rq *rt_rq;
1386 /* rq "owned" by this entity/group: */
1391 struct sched_dl_entity {
1392 struct rb_node rb_node;
1395 * Original scheduling parameters. Copied here from sched_attr
1396 * during sched_setattr(), they will remain the same until
1397 * the next sched_setattr().
1399 u64 dl_runtime; /* maximum runtime for each instance */
1400 u64 dl_deadline; /* relative deadline of each instance */
1401 u64 dl_period; /* separation of two instances (period) */
1402 u64 dl_bw; /* dl_runtime / dl_deadline */
1405 * Actual scheduling parameters. Initialized with the values above,
1406 * they are continously updated during task execution. Note that
1407 * the remaining runtime could be < 0 in case we are in overrun.
1409 s64 runtime; /* remaining runtime for this instance */
1410 u64 deadline; /* absolute deadline for this instance */
1411 unsigned int flags; /* specifying the scheduler behaviour */
1416 * @dl_throttled tells if we exhausted the runtime. If so, the
1417 * task has to wait for a replenishment to be performed at the
1418 * next firing of dl_timer.
1420 * @dl_boosted tells if we are boosted due to DI. If so we are
1421 * outside bandwidth enforcement mechanism (but only until we
1422 * exit the critical section);
1424 * @dl_yielded tells if task gave up the cpu before consuming
1425 * all its available runtime during the last job.
1427 int dl_throttled, dl_boosted, dl_yielded;
1430 * Bandwidth enforcement timer. Each -deadline task has its
1431 * own bandwidth to be enforced, thus we need one timer per task.
1433 struct hrtimer dl_timer;
1441 u8 pad; /* Otherwise the compiler can store garbage here. */
1443 u32 s; /* Set of bits. */
1447 enum perf_event_task_context {
1448 perf_invalid_context = -1,
1449 perf_hw_context = 0,
1451 perf_nr_task_contexts,
1454 /* Track pages that require TLB flushes */
1455 struct tlbflush_unmap_batch {
1457 * Each bit set is a CPU that potentially has a TLB entry for one of
1458 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1460 struct cpumask cpumask;
1462 /* True if any bit in cpumask is set */
1463 bool flush_required;
1466 * If true then the PTE was dirty when unmapped. The entry must be
1467 * flushed before IO is initiated or a stale TLB entry potentially
1468 * allows an update without redirtying the page.
1473 struct task_struct {
1474 #ifdef CONFIG_THREAD_INFO_IN_TASK
1476 * For reasons of header soup (see current_thread_info()), this
1477 * must be the first element of task_struct.
1479 struct thread_info thread_info;
1481 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1484 unsigned int flags; /* per process flags, defined below */
1485 unsigned int ptrace;
1488 struct llist_node wake_entry;
1490 #ifdef CONFIG_THREAD_INFO_IN_TASK
1491 unsigned int cpu; /* current CPU */
1493 unsigned int wakee_flips;
1494 unsigned long wakee_flip_decay_ts;
1495 struct task_struct *last_wakee;
1501 int prio, static_prio, normal_prio;
1502 unsigned int rt_priority;
1503 const struct sched_class *sched_class;
1504 struct sched_entity se;
1505 struct sched_rt_entity rt;
1506 #ifdef CONFIG_CGROUP_SCHED
1507 struct task_group *sched_task_group;
1509 struct sched_dl_entity dl;
1511 #ifdef CONFIG_PREEMPT_NOTIFIERS
1512 /* list of struct preempt_notifier: */
1513 struct hlist_head preempt_notifiers;
1516 #ifdef CONFIG_BLK_DEV_IO_TRACE
1517 unsigned int btrace_seq;
1520 unsigned int policy;
1521 int nr_cpus_allowed;
1522 cpumask_t cpus_allowed;
1524 #ifdef CONFIG_PREEMPT_RCU
1525 int rcu_read_lock_nesting;
1526 union rcu_special rcu_read_unlock_special;
1527 struct list_head rcu_node_entry;
1528 struct rcu_node *rcu_blocked_node;
1529 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1530 #ifdef CONFIG_TASKS_RCU
1531 unsigned long rcu_tasks_nvcsw;
1532 bool rcu_tasks_holdout;
1533 struct list_head rcu_tasks_holdout_list;
1534 int rcu_tasks_idle_cpu;
1535 #endif /* #ifdef CONFIG_TASKS_RCU */
1537 #ifdef CONFIG_SCHED_INFO
1538 struct sched_info sched_info;
1541 struct list_head tasks;
1543 struct plist_node pushable_tasks;
1544 struct rb_node pushable_dl_tasks;
1547 struct mm_struct *mm, *active_mm;
1548 /* per-thread vma caching */
1549 u32 vmacache_seqnum;
1550 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1551 #if defined(SPLIT_RSS_COUNTING)
1552 struct task_rss_stat rss_stat;
1556 int exit_code, exit_signal;
1557 int pdeath_signal; /* The signal sent when the parent dies */
1558 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1560 /* Used for emulating ABI behavior of previous Linux versions */
1561 unsigned int personality;
1563 /* scheduler bits, serialized by scheduler locks */
1564 unsigned sched_reset_on_fork:1;
1565 unsigned sched_contributes_to_load:1;
1566 unsigned sched_migrated:1;
1567 unsigned sched_remote_wakeup:1;
1568 unsigned :0; /* force alignment to the next boundary */
1570 /* unserialized, strictly 'current' */
1571 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1572 unsigned in_iowait:1;
1573 #if !defined(TIF_RESTORE_SIGMASK)
1574 unsigned restore_sigmask:1;
1577 unsigned memcg_may_oom:1;
1579 unsigned memcg_kmem_skip_account:1;
1582 #ifdef CONFIG_COMPAT_BRK
1583 unsigned brk_randomized:1;
1586 unsigned long atomic_flags; /* Flags needing atomic access. */
1588 struct restart_block restart_block;
1593 #ifdef CONFIG_CC_STACKPROTECTOR
1594 /* Canary value for the -fstack-protector gcc feature */
1595 unsigned long stack_canary;
1598 * pointers to (original) parent process, youngest child, younger sibling,
1599 * older sibling, respectively. (p->father can be replaced with
1600 * p->real_parent->pid)
1602 struct task_struct __rcu *real_parent; /* real parent process */
1603 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1605 * children/sibling forms the list of my natural children
1607 struct list_head children; /* list of my children */
1608 struct list_head sibling; /* linkage in my parent's children list */
1609 struct task_struct *group_leader; /* threadgroup leader */
1612 * ptraced is the list of tasks this task is using ptrace on.
1613 * This includes both natural children and PTRACE_ATTACH targets.
1614 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1616 struct list_head ptraced;
1617 struct list_head ptrace_entry;
1619 /* PID/PID hash table linkage. */
1620 struct pid_link pids[PIDTYPE_MAX];
1621 struct list_head thread_group;
1622 struct list_head thread_node;
1624 struct completion *vfork_done; /* for vfork() */
1625 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1626 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1628 cputime_t utime, stime, utimescaled, stimescaled;
1630 struct prev_cputime prev_cputime;
1631 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1632 seqcount_t vtime_seqcount;
1633 unsigned long long vtime_snap;
1635 /* Task is sleeping or running in a CPU with VTIME inactive */
1637 /* Task runs in userspace in a CPU with VTIME active */
1639 /* Task runs in kernelspace in a CPU with VTIME active */
1641 } vtime_snap_whence;
1644 #ifdef CONFIG_NO_HZ_FULL
1645 atomic_t tick_dep_mask;
1647 unsigned long nvcsw, nivcsw; /* context switch counts */
1648 u64 start_time; /* monotonic time in nsec */
1649 u64 real_start_time; /* boot based time in nsec */
1650 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1651 unsigned long min_flt, maj_flt;
1653 struct task_cputime cputime_expires;
1654 struct list_head cpu_timers[3];
1656 /* process credentials */
1657 const struct cred __rcu *real_cred; /* objective and real subjective task
1658 * credentials (COW) */
1659 const struct cred __rcu *cred; /* effective (overridable) subjective task
1660 * credentials (COW) */
1661 char comm[TASK_COMM_LEN]; /* executable name excluding path
1662 - access with [gs]et_task_comm (which lock
1663 it with task_lock())
1664 - initialized normally by setup_new_exec */
1665 /* file system info */
1666 struct nameidata *nameidata;
1667 #ifdef CONFIG_SYSVIPC
1669 struct sysv_sem sysvsem;
1670 struct sysv_shm sysvshm;
1672 #ifdef CONFIG_DETECT_HUNG_TASK
1673 /* hung task detection */
1674 unsigned long last_switch_count;
1676 /* filesystem information */
1677 struct fs_struct *fs;
1678 /* open file information */
1679 struct files_struct *files;
1681 struct nsproxy *nsproxy;
1682 /* signal handlers */
1683 struct signal_struct *signal;
1684 struct sighand_struct *sighand;
1686 sigset_t blocked, real_blocked;
1687 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1688 struct sigpending pending;
1690 unsigned long sas_ss_sp;
1692 unsigned sas_ss_flags;
1694 struct callback_head *task_works;
1696 struct audit_context *audit_context;
1697 #ifdef CONFIG_AUDITSYSCALL
1699 unsigned int sessionid;
1701 struct seccomp seccomp;
1703 /* Thread group tracking */
1706 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1708 spinlock_t alloc_lock;
1710 /* Protection of the PI data structures: */
1711 raw_spinlock_t pi_lock;
1713 struct wake_q_node wake_q;
1715 #ifdef CONFIG_RT_MUTEXES
1716 /* PI waiters blocked on a rt_mutex held by this task */
1717 struct rb_root pi_waiters;
1718 struct rb_node *pi_waiters_leftmost;
1719 /* Deadlock detection and priority inheritance handling */
1720 struct rt_mutex_waiter *pi_blocked_on;
1723 #ifdef CONFIG_DEBUG_MUTEXES
1724 /* mutex deadlock detection */
1725 struct mutex_waiter *blocked_on;
1727 #ifdef CONFIG_TRACE_IRQFLAGS
1728 unsigned int irq_events;
1729 unsigned long hardirq_enable_ip;
1730 unsigned long hardirq_disable_ip;
1731 unsigned int hardirq_enable_event;
1732 unsigned int hardirq_disable_event;
1733 int hardirqs_enabled;
1734 int hardirq_context;
1735 unsigned long softirq_disable_ip;
1736 unsigned long softirq_enable_ip;
1737 unsigned int softirq_disable_event;
1738 unsigned int softirq_enable_event;
1739 int softirqs_enabled;
1740 int softirq_context;
1742 #ifdef CONFIG_LOCKDEP
1743 # define MAX_LOCK_DEPTH 48UL
1746 unsigned int lockdep_recursion;
1747 struct held_lock held_locks[MAX_LOCK_DEPTH];
1748 gfp_t lockdep_reclaim_gfp;
1751 unsigned int in_ubsan;
1754 /* journalling filesystem info */
1757 /* stacked block device info */
1758 struct bio_list *bio_list;
1761 /* stack plugging */
1762 struct blk_plug *plug;
1766 struct reclaim_state *reclaim_state;
1768 struct backing_dev_info *backing_dev_info;
1770 struct io_context *io_context;
1772 unsigned long ptrace_message;
1773 siginfo_t *last_siginfo; /* For ptrace use. */
1774 struct task_io_accounting ioac;
1775 #if defined(CONFIG_TASK_XACCT)
1776 u64 acct_rss_mem1; /* accumulated rss usage */
1777 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1778 cputime_t acct_timexpd; /* stime + utime since last update */
1780 #ifdef CONFIG_CPUSETS
1781 nodemask_t mems_allowed; /* Protected by alloc_lock */
1782 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1783 int cpuset_mem_spread_rotor;
1784 int cpuset_slab_spread_rotor;
1786 #ifdef CONFIG_CGROUPS
1787 /* Control Group info protected by css_set_lock */
1788 struct css_set __rcu *cgroups;
1789 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1790 struct list_head cg_list;
1793 struct robust_list_head __user *robust_list;
1794 #ifdef CONFIG_COMPAT
1795 struct compat_robust_list_head __user *compat_robust_list;
1797 struct list_head pi_state_list;
1798 struct futex_pi_state *pi_state_cache;
1800 #ifdef CONFIG_PERF_EVENTS
1801 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1802 struct mutex perf_event_mutex;
1803 struct list_head perf_event_list;
1805 #ifdef CONFIG_DEBUG_PREEMPT
1806 unsigned long preempt_disable_ip;
1809 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1811 short pref_node_fork;
1813 #ifdef CONFIG_NUMA_BALANCING
1815 unsigned int numa_scan_period;
1816 unsigned int numa_scan_period_max;
1817 int numa_preferred_nid;
1818 unsigned long numa_migrate_retry;
1819 u64 node_stamp; /* migration stamp */
1820 u64 last_task_numa_placement;
1821 u64 last_sum_exec_runtime;
1822 struct callback_head numa_work;
1824 struct list_head numa_entry;
1825 struct numa_group *numa_group;
1828 * numa_faults is an array split into four regions:
1829 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1830 * in this precise order.
1832 * faults_memory: Exponential decaying average of faults on a per-node
1833 * basis. Scheduling placement decisions are made based on these
1834 * counts. The values remain static for the duration of a PTE scan.
1835 * faults_cpu: Track the nodes the process was running on when a NUMA
1836 * hinting fault was incurred.
1837 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1838 * during the current scan window. When the scan completes, the counts
1839 * in faults_memory and faults_cpu decay and these values are copied.
1841 unsigned long *numa_faults;
1842 unsigned long total_numa_faults;
1845 * numa_faults_locality tracks if faults recorded during the last
1846 * scan window were remote/local or failed to migrate. The task scan
1847 * period is adapted based on the locality of the faults with different
1848 * weights depending on whether they were shared or private faults
1850 unsigned long numa_faults_locality[3];
1852 unsigned long numa_pages_migrated;
1853 #endif /* CONFIG_NUMA_BALANCING */
1855 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1856 struct tlbflush_unmap_batch tlb_ubc;
1859 struct rcu_head rcu;
1862 * cache last used pipe for splice
1864 struct pipe_inode_info *splice_pipe;
1866 struct page_frag task_frag;
1868 #ifdef CONFIG_TASK_DELAY_ACCT
1869 struct task_delay_info *delays;
1871 #ifdef CONFIG_FAULT_INJECTION
1875 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1876 * balance_dirty_pages() for some dirty throttling pause
1879 int nr_dirtied_pause;
1880 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1882 #ifdef CONFIG_LATENCYTOP
1883 int latency_record_count;
1884 struct latency_record latency_record[LT_SAVECOUNT];
1887 * time slack values; these are used to round up poll() and
1888 * select() etc timeout values. These are in nanoseconds.
1891 u64 default_timer_slack_ns;
1894 unsigned int kasan_depth;
1896 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1897 /* Index of current stored address in ret_stack */
1899 /* Stack of return addresses for return function tracing */
1900 struct ftrace_ret_stack *ret_stack;
1901 /* time stamp for last schedule */
1902 unsigned long long ftrace_timestamp;
1904 * Number of functions that haven't been traced
1905 * because of depth overrun.
1907 atomic_t trace_overrun;
1908 /* Pause for the tracing */
1909 atomic_t tracing_graph_pause;
1911 #ifdef CONFIG_TRACING
1912 /* state flags for use by tracers */
1913 unsigned long trace;
1914 /* bitmask and counter of trace recursion */
1915 unsigned long trace_recursion;
1916 #endif /* CONFIG_TRACING */
1918 /* Coverage collection mode enabled for this task (0 if disabled). */
1919 enum kcov_mode kcov_mode;
1920 /* Size of the kcov_area. */
1922 /* Buffer for coverage collection. */
1924 /* kcov desciptor wired with this task or NULL. */
1928 struct mem_cgroup *memcg_in_oom;
1929 gfp_t memcg_oom_gfp_mask;
1930 int memcg_oom_order;
1932 /* number of pages to reclaim on returning to userland */
1933 unsigned int memcg_nr_pages_over_high;
1935 #ifdef CONFIG_UPROBES
1936 struct uprobe_task *utask;
1938 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1939 unsigned int sequential_io;
1940 unsigned int sequential_io_avg;
1942 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1943 unsigned long task_state_change;
1945 int pagefault_disabled;
1947 struct task_struct *oom_reaper_list;
1949 #ifdef CONFIG_VMAP_STACK
1950 struct vm_struct *stack_vm_area;
1952 #ifdef CONFIG_THREAD_INFO_IN_TASK
1953 /* A live task holds one reference. */
1954 atomic_t stack_refcount;
1956 /* CPU-specific state of this task */
1957 struct thread_struct thread;
1959 * WARNING: on x86, 'thread_struct' contains a variable-sized
1960 * structure. It *MUST* be at the end of 'task_struct'.
1962 * Do not put anything below here!
1966 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1967 extern int arch_task_struct_size __read_mostly;
1969 # define arch_task_struct_size (sizeof(struct task_struct))
1972 #ifdef CONFIG_VMAP_STACK
1973 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1975 return t->stack_vm_area;
1978 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1984 /* Future-safe accessor for struct task_struct's cpus_allowed. */
1985 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1987 static inline int tsk_nr_cpus_allowed(struct task_struct *p)
1989 return p->nr_cpus_allowed;
1992 #define TNF_MIGRATED 0x01
1993 #define TNF_NO_GROUP 0x02
1994 #define TNF_SHARED 0x04
1995 #define TNF_FAULT_LOCAL 0x08
1996 #define TNF_MIGRATE_FAIL 0x10
1998 static inline bool in_vfork(struct task_struct *tsk)
2003 * need RCU to access ->real_parent if CLONE_VM was used along with
2006 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2009 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2010 * ->real_parent is not necessarily the task doing vfork(), so in
2011 * theory we can't rely on task_lock() if we want to dereference it.
2013 * And in this case we can't trust the real_parent->mm == tsk->mm
2014 * check, it can be false negative. But we do not care, if init or
2015 * another oom-unkillable task does this it should blame itself.
2018 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2024 #ifdef CONFIG_NUMA_BALANCING
2025 extern void task_numa_fault(int last_node, int node, int pages, int flags);
2026 extern pid_t task_numa_group_id(struct task_struct *p);
2027 extern void set_numabalancing_state(bool enabled);
2028 extern void task_numa_free(struct task_struct *p);
2029 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2030 int src_nid, int dst_cpu);
2032 static inline void task_numa_fault(int last_node, int node, int pages,
2036 static inline pid_t task_numa_group_id(struct task_struct *p)
2040 static inline void set_numabalancing_state(bool enabled)
2043 static inline void task_numa_free(struct task_struct *p)
2046 static inline bool should_numa_migrate_memory(struct task_struct *p,
2047 struct page *page, int src_nid, int dst_cpu)
2053 static inline struct pid *task_pid(struct task_struct *task)
2055 return task->pids[PIDTYPE_PID].pid;
2058 static inline struct pid *task_tgid(struct task_struct *task)
2060 return task->group_leader->pids[PIDTYPE_PID].pid;
2064 * Without tasklist or rcu lock it is not safe to dereference
2065 * the result of task_pgrp/task_session even if task == current,
2066 * we can race with another thread doing sys_setsid/sys_setpgid.
2068 static inline struct pid *task_pgrp(struct task_struct *task)
2070 return task->group_leader->pids[PIDTYPE_PGID].pid;
2073 static inline struct pid *task_session(struct task_struct *task)
2075 return task->group_leader->pids[PIDTYPE_SID].pid;
2078 struct pid_namespace;
2081 * the helpers to get the task's different pids as they are seen
2082 * from various namespaces
2084 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2085 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2087 * task_xid_nr_ns() : id seen from the ns specified;
2089 * set_task_vxid() : assigns a virtual id to a task;
2091 * see also pid_nr() etc in include/linux/pid.h
2093 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2094 struct pid_namespace *ns);
2096 static inline pid_t task_pid_nr(struct task_struct *tsk)
2101 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2102 struct pid_namespace *ns)
2104 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2107 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2109 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2113 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2118 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2120 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2122 return pid_vnr(task_tgid(tsk));
2126 static inline int pid_alive(const struct task_struct *p);
2127 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2133 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2139 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2141 return task_ppid_nr_ns(tsk, &init_pid_ns);
2144 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2145 struct pid_namespace *ns)
2147 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2150 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2152 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2156 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2157 struct pid_namespace *ns)
2159 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2162 static inline pid_t task_session_vnr(struct task_struct *tsk)
2164 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2167 /* obsolete, do not use */
2168 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2170 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2174 * pid_alive - check that a task structure is not stale
2175 * @p: Task structure to be checked.
2177 * Test if a process is not yet dead (at most zombie state)
2178 * If pid_alive fails, then pointers within the task structure
2179 * can be stale and must not be dereferenced.
2181 * Return: 1 if the process is alive. 0 otherwise.
2183 static inline int pid_alive(const struct task_struct *p)
2185 return p->pids[PIDTYPE_PID].pid != NULL;
2189 * is_global_init - check if a task structure is init. Since init
2190 * is free to have sub-threads we need to check tgid.
2191 * @tsk: Task structure to be checked.
2193 * Check if a task structure is the first user space task the kernel created.
2195 * Return: 1 if the task structure is init. 0 otherwise.
2197 static inline int is_global_init(struct task_struct *tsk)
2199 return task_tgid_nr(tsk) == 1;
2202 extern struct pid *cad_pid;
2204 extern void free_task(struct task_struct *tsk);
2205 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2207 extern void __put_task_struct(struct task_struct *t);
2209 static inline void put_task_struct(struct task_struct *t)
2211 if (atomic_dec_and_test(&t->usage))
2212 __put_task_struct(t);
2215 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2216 struct task_struct *try_get_task_struct(struct task_struct **ptask);
2218 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2219 extern void task_cputime(struct task_struct *t,
2220 cputime_t *utime, cputime_t *stime);
2221 extern void task_cputime_scaled(struct task_struct *t,
2222 cputime_t *utimescaled, cputime_t *stimescaled);
2223 extern cputime_t task_gtime(struct task_struct *t);
2225 static inline void task_cputime(struct task_struct *t,
2226 cputime_t *utime, cputime_t *stime)
2234 static inline void task_cputime_scaled(struct task_struct *t,
2235 cputime_t *utimescaled,
2236 cputime_t *stimescaled)
2239 *utimescaled = t->utimescaled;
2241 *stimescaled = t->stimescaled;
2244 static inline cputime_t task_gtime(struct task_struct *t)
2249 extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2250 extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2255 #define PF_EXITING 0x00000004 /* getting shut down */
2256 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2257 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2258 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2259 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2260 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2261 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2262 #define PF_DUMPCORE 0x00000200 /* dumped core */
2263 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2264 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2265 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2266 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2267 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2268 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2269 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2270 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2271 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2272 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2273 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2274 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2275 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2276 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2277 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2278 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2279 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2280 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2281 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2284 * Only the _current_ task can read/write to tsk->flags, but other
2285 * tasks can access tsk->flags in readonly mode for example
2286 * with tsk_used_math (like during threaded core dumping).
2287 * There is however an exception to this rule during ptrace
2288 * or during fork: the ptracer task is allowed to write to the
2289 * child->flags of its traced child (same goes for fork, the parent
2290 * can write to the child->flags), because we're guaranteed the
2291 * child is not running and in turn not changing child->flags
2292 * at the same time the parent does it.
2294 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2295 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2296 #define clear_used_math() clear_stopped_child_used_math(current)
2297 #define set_used_math() set_stopped_child_used_math(current)
2298 #define conditional_stopped_child_used_math(condition, child) \
2299 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2300 #define conditional_used_math(condition) \
2301 conditional_stopped_child_used_math(condition, current)
2302 #define copy_to_stopped_child_used_math(child) \
2303 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2304 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2305 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2306 #define used_math() tsk_used_math(current)
2308 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2309 * __GFP_FS is also cleared as it implies __GFP_IO.
2311 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2313 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2314 flags &= ~(__GFP_IO | __GFP_FS);
2318 static inline unsigned int memalloc_noio_save(void)
2320 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2321 current->flags |= PF_MEMALLOC_NOIO;
2325 static inline void memalloc_noio_restore(unsigned int flags)
2327 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2330 /* Per-process atomic flags. */
2331 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2332 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2333 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2334 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2337 #define TASK_PFA_TEST(name, func) \
2338 static inline bool task_##func(struct task_struct *p) \
2339 { return test_bit(PFA_##name, &p->atomic_flags); }
2340 #define TASK_PFA_SET(name, func) \
2341 static inline void task_set_##func(struct task_struct *p) \
2342 { set_bit(PFA_##name, &p->atomic_flags); }
2343 #define TASK_PFA_CLEAR(name, func) \
2344 static inline void task_clear_##func(struct task_struct *p) \
2345 { clear_bit(PFA_##name, &p->atomic_flags); }
2347 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2348 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2350 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2351 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2352 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2354 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2355 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2356 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2358 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2359 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2362 * task->jobctl flags
2364 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2366 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2367 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2368 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2369 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2370 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2371 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2372 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2374 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2375 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2376 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2377 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2378 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2379 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2380 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2382 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2383 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2385 extern bool task_set_jobctl_pending(struct task_struct *task,
2386 unsigned long mask);
2387 extern void task_clear_jobctl_trapping(struct task_struct *task);
2388 extern void task_clear_jobctl_pending(struct task_struct *task,
2389 unsigned long mask);
2391 static inline void rcu_copy_process(struct task_struct *p)
2393 #ifdef CONFIG_PREEMPT_RCU
2394 p->rcu_read_lock_nesting = 0;
2395 p->rcu_read_unlock_special.s = 0;
2396 p->rcu_blocked_node = NULL;
2397 INIT_LIST_HEAD(&p->rcu_node_entry);
2398 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2399 #ifdef CONFIG_TASKS_RCU
2400 p->rcu_tasks_holdout = false;
2401 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2402 p->rcu_tasks_idle_cpu = -1;
2403 #endif /* #ifdef CONFIG_TASKS_RCU */
2406 static inline void tsk_restore_flags(struct task_struct *task,
2407 unsigned long orig_flags, unsigned long flags)
2409 task->flags &= ~flags;
2410 task->flags |= orig_flags & flags;
2413 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2414 const struct cpumask *trial);
2415 extern int task_can_attach(struct task_struct *p,
2416 const struct cpumask *cs_cpus_allowed);
2418 extern void do_set_cpus_allowed(struct task_struct *p,
2419 const struct cpumask *new_mask);
2421 extern int set_cpus_allowed_ptr(struct task_struct *p,
2422 const struct cpumask *new_mask);
2424 static inline void do_set_cpus_allowed(struct task_struct *p,
2425 const struct cpumask *new_mask)
2428 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2429 const struct cpumask *new_mask)
2431 if (!cpumask_test_cpu(0, new_mask))
2437 #ifdef CONFIG_NO_HZ_COMMON
2438 void calc_load_enter_idle(void);
2439 void calc_load_exit_idle(void);
2441 static inline void calc_load_enter_idle(void) { }
2442 static inline void calc_load_exit_idle(void) { }
2443 #endif /* CONFIG_NO_HZ_COMMON */
2446 * Do not use outside of architecture code which knows its limitations.
2448 * sched_clock() has no promise of monotonicity or bounded drift between
2449 * CPUs, use (which you should not) requires disabling IRQs.
2451 * Please use one of the three interfaces below.
2453 extern unsigned long long notrace sched_clock(void);
2455 * See the comment in kernel/sched/clock.c
2457 extern u64 running_clock(void);
2458 extern u64 sched_clock_cpu(int cpu);
2461 extern void sched_clock_init(void);
2463 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2464 static inline void sched_clock_tick(void)
2468 static inline void sched_clock_idle_sleep_event(void)
2472 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2476 static inline u64 cpu_clock(int cpu)
2478 return sched_clock();
2481 static inline u64 local_clock(void)
2483 return sched_clock();
2487 * Architectures can set this to 1 if they have specified
2488 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2489 * but then during bootup it turns out that sched_clock()
2490 * is reliable after all:
2492 extern int sched_clock_stable(void);
2493 extern void set_sched_clock_stable(void);
2494 extern void clear_sched_clock_stable(void);
2496 extern void sched_clock_tick(void);
2497 extern void sched_clock_idle_sleep_event(void);
2498 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2501 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2502 * time source that is monotonic per cpu argument and has bounded drift
2505 * ######################### BIG FAT WARNING ##########################
2506 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2507 * # go backwards !! #
2508 * ####################################################################
2510 static inline u64 cpu_clock(int cpu)
2512 return sched_clock_cpu(cpu);
2515 static inline u64 local_clock(void)
2517 return sched_clock_cpu(raw_smp_processor_id());
2521 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2523 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2524 * The reason for this explicit opt-in is not to have perf penalty with
2525 * slow sched_clocks.
2527 extern void enable_sched_clock_irqtime(void);
2528 extern void disable_sched_clock_irqtime(void);
2530 static inline void enable_sched_clock_irqtime(void) {}
2531 static inline void disable_sched_clock_irqtime(void) {}
2534 extern unsigned long long
2535 task_sched_runtime(struct task_struct *task);
2537 /* sched_exec is called by processes performing an exec */
2539 extern void sched_exec(void);
2541 #define sched_exec() {}
2544 extern void sched_clock_idle_sleep_event(void);
2545 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2547 #ifdef CONFIG_HOTPLUG_CPU
2548 extern void idle_task_exit(void);
2550 static inline void idle_task_exit(void) {}
2553 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2554 extern void wake_up_nohz_cpu(int cpu);
2556 static inline void wake_up_nohz_cpu(int cpu) { }
2559 #ifdef CONFIG_NO_HZ_FULL
2560 extern u64 scheduler_tick_max_deferment(void);
2563 #ifdef CONFIG_SCHED_AUTOGROUP
2564 extern void sched_autogroup_create_attach(struct task_struct *p);
2565 extern void sched_autogroup_detach(struct task_struct *p);
2566 extern void sched_autogroup_fork(struct signal_struct *sig);
2567 extern void sched_autogroup_exit(struct signal_struct *sig);
2568 #ifdef CONFIG_PROC_FS
2569 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2570 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2573 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2574 static inline void sched_autogroup_detach(struct task_struct *p) { }
2575 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2576 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2579 extern int yield_to(struct task_struct *p, bool preempt);
2580 extern void set_user_nice(struct task_struct *p, long nice);
2581 extern int task_prio(const struct task_struct *p);
2583 * task_nice - return the nice value of a given task.
2584 * @p: the task in question.
2586 * Return: The nice value [ -20 ... 0 ... 19 ].
2588 static inline int task_nice(const struct task_struct *p)
2590 return PRIO_TO_NICE((p)->static_prio);
2592 extern int can_nice(const struct task_struct *p, const int nice);
2593 extern int task_curr(const struct task_struct *p);
2594 extern int idle_cpu(int cpu);
2595 extern int sched_setscheduler(struct task_struct *, int,
2596 const struct sched_param *);
2597 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2598 const struct sched_param *);
2599 extern int sched_setattr(struct task_struct *,
2600 const struct sched_attr *);
2601 extern struct task_struct *idle_task(int cpu);
2603 * is_idle_task - is the specified task an idle task?
2604 * @p: the task in question.
2606 * Return: 1 if @p is an idle task. 0 otherwise.
2608 static inline bool is_idle_task(const struct task_struct *p)
2612 extern struct task_struct *curr_task(int cpu);
2613 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2617 union thread_union {
2618 #ifndef CONFIG_THREAD_INFO_IN_TASK
2619 struct thread_info thread_info;
2621 unsigned long stack[THREAD_SIZE/sizeof(long)];
2624 #ifndef __HAVE_ARCH_KSTACK_END
2625 static inline int kstack_end(void *addr)
2627 /* Reliable end of stack detection:
2628 * Some APM bios versions misalign the stack
2630 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2634 extern union thread_union init_thread_union;
2635 extern struct task_struct init_task;
2637 extern struct mm_struct init_mm;
2639 extern struct pid_namespace init_pid_ns;
2642 * find a task by one of its numerical ids
2644 * find_task_by_pid_ns():
2645 * finds a task by its pid in the specified namespace
2646 * find_task_by_vpid():
2647 * finds a task by its virtual pid
2649 * see also find_vpid() etc in include/linux/pid.h
2652 extern struct task_struct *find_task_by_vpid(pid_t nr);
2653 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2654 struct pid_namespace *ns);
2656 /* per-UID process charging. */
2657 extern struct user_struct * alloc_uid(kuid_t);
2658 static inline struct user_struct *get_uid(struct user_struct *u)
2660 atomic_inc(&u->__count);
2663 extern void free_uid(struct user_struct *);
2665 #include <asm/current.h>
2667 extern void xtime_update(unsigned long ticks);
2669 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2670 extern int wake_up_process(struct task_struct *tsk);
2671 extern void wake_up_new_task(struct task_struct *tsk);
2673 extern void kick_process(struct task_struct *tsk);
2675 static inline void kick_process(struct task_struct *tsk) { }
2677 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2678 extern void sched_dead(struct task_struct *p);
2680 extern void proc_caches_init(void);
2681 extern void flush_signals(struct task_struct *);
2682 extern void ignore_signals(struct task_struct *);
2683 extern void flush_signal_handlers(struct task_struct *, int force_default);
2684 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2686 static inline int kernel_dequeue_signal(siginfo_t *info)
2688 struct task_struct *tsk = current;
2692 spin_lock_irq(&tsk->sighand->siglock);
2693 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2694 spin_unlock_irq(&tsk->sighand->siglock);
2699 static inline void kernel_signal_stop(void)
2701 spin_lock_irq(¤t->sighand->siglock);
2702 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2703 __set_current_state(TASK_STOPPED);
2704 spin_unlock_irq(¤t->sighand->siglock);
2709 extern void release_task(struct task_struct * p);
2710 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2711 extern int force_sigsegv(int, struct task_struct *);
2712 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2713 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2714 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2715 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2716 const struct cred *, u32);
2717 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2718 extern int kill_pid(struct pid *pid, int sig, int priv);
2719 extern int kill_proc_info(int, struct siginfo *, pid_t);
2720 extern __must_check bool do_notify_parent(struct task_struct *, int);
2721 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2722 extern void force_sig(int, struct task_struct *);
2723 extern int send_sig(int, struct task_struct *, int);
2724 extern int zap_other_threads(struct task_struct *p);
2725 extern struct sigqueue *sigqueue_alloc(void);
2726 extern void sigqueue_free(struct sigqueue *);
2727 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2728 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2730 #ifdef TIF_RESTORE_SIGMASK
2732 * Legacy restore_sigmask accessors. These are inefficient on
2733 * SMP architectures because they require atomic operations.
2737 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2739 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2740 * will run before returning to user mode, to process the flag. For
2741 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2742 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2743 * arch code will notice on return to user mode, in case those bits
2744 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2745 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2747 static inline void set_restore_sigmask(void)
2749 set_thread_flag(TIF_RESTORE_SIGMASK);
2750 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2752 static inline void clear_restore_sigmask(void)
2754 clear_thread_flag(TIF_RESTORE_SIGMASK);
2756 static inline bool test_restore_sigmask(void)
2758 return test_thread_flag(TIF_RESTORE_SIGMASK);
2760 static inline bool test_and_clear_restore_sigmask(void)
2762 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2765 #else /* TIF_RESTORE_SIGMASK */
2767 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2768 static inline void set_restore_sigmask(void)
2770 current->restore_sigmask = true;
2771 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2773 static inline void clear_restore_sigmask(void)
2775 current->restore_sigmask = false;
2777 static inline bool test_restore_sigmask(void)
2779 return current->restore_sigmask;
2781 static inline bool test_and_clear_restore_sigmask(void)
2783 if (!current->restore_sigmask)
2785 current->restore_sigmask = false;
2790 static inline void restore_saved_sigmask(void)
2792 if (test_and_clear_restore_sigmask())
2793 __set_current_blocked(¤t->saved_sigmask);
2796 static inline sigset_t *sigmask_to_save(void)
2798 sigset_t *res = ¤t->blocked;
2799 if (unlikely(test_restore_sigmask()))
2800 res = ¤t->saved_sigmask;
2804 static inline int kill_cad_pid(int sig, int priv)
2806 return kill_pid(cad_pid, sig, priv);
2809 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2810 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2811 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2812 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2815 * True if we are on the alternate signal stack.
2817 static inline int on_sig_stack(unsigned long sp)
2820 * If the signal stack is SS_AUTODISARM then, by construction, we
2821 * can't be on the signal stack unless user code deliberately set
2822 * SS_AUTODISARM when we were already on it.
2824 * This improves reliability: if user state gets corrupted such that
2825 * the stack pointer points very close to the end of the signal stack,
2826 * then this check will enable the signal to be handled anyway.
2828 if (current->sas_ss_flags & SS_AUTODISARM)
2831 #ifdef CONFIG_STACK_GROWSUP
2832 return sp >= current->sas_ss_sp &&
2833 sp - current->sas_ss_sp < current->sas_ss_size;
2835 return sp > current->sas_ss_sp &&
2836 sp - current->sas_ss_sp <= current->sas_ss_size;
2840 static inline int sas_ss_flags(unsigned long sp)
2842 if (!current->sas_ss_size)
2845 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2848 static inline void sas_ss_reset(struct task_struct *p)
2852 p->sas_ss_flags = SS_DISABLE;
2855 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2857 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2858 #ifdef CONFIG_STACK_GROWSUP
2859 return current->sas_ss_sp;
2861 return current->sas_ss_sp + current->sas_ss_size;
2867 * Routines for handling mm_structs
2869 extern struct mm_struct * mm_alloc(void);
2871 /* mmdrop drops the mm and the page tables */
2872 extern void __mmdrop(struct mm_struct *);
2873 static inline void mmdrop(struct mm_struct *mm)
2875 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2879 static inline void mmdrop_async_fn(struct work_struct *work)
2881 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2885 static inline void mmdrop_async(struct mm_struct *mm)
2887 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2888 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2889 schedule_work(&mm->async_put_work);
2893 static inline bool mmget_not_zero(struct mm_struct *mm)
2895 return atomic_inc_not_zero(&mm->mm_users);
2898 /* mmput gets rid of the mappings and all user-space */
2899 extern void mmput(struct mm_struct *);
2901 /* same as above but performs the slow path from the async context. Can
2902 * be called from the atomic context as well
2904 extern void mmput_async(struct mm_struct *);
2907 /* Grab a reference to a task's mm, if it is not already going away */
2908 extern struct mm_struct *get_task_mm(struct task_struct *task);
2910 * Grab a reference to a task's mm, if it is not already going away
2911 * and ptrace_may_access with the mode parameter passed to it
2914 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2915 /* Remove the current tasks stale references to the old mm_struct */
2916 extern void mm_release(struct task_struct *, struct mm_struct *);
2918 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2919 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2920 struct task_struct *, unsigned long);
2922 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2923 struct task_struct *);
2925 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2926 * via pt_regs, so ignore the tls argument passed via C. */
2927 static inline int copy_thread_tls(
2928 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2929 struct task_struct *p, unsigned long tls)
2931 return copy_thread(clone_flags, sp, arg, p);
2934 extern void flush_thread(void);
2936 #ifdef CONFIG_HAVE_EXIT_THREAD
2937 extern void exit_thread(struct task_struct *tsk);
2939 static inline void exit_thread(struct task_struct *tsk)
2944 extern void exit_files(struct task_struct *);
2945 extern void __cleanup_sighand(struct sighand_struct *);
2947 extern void exit_itimers(struct signal_struct *);
2948 extern void flush_itimer_signals(void);
2950 extern void do_group_exit(int);
2952 extern int do_execve(struct filename *,
2953 const char __user * const __user *,
2954 const char __user * const __user *);
2955 extern int do_execveat(int, struct filename *,
2956 const char __user * const __user *,
2957 const char __user * const __user *,
2959 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
2960 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2961 struct task_struct *fork_idle(int);
2962 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2964 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2965 static inline void set_task_comm(struct task_struct *tsk, const char *from)
2967 __set_task_comm(tsk, from, false);
2969 extern char *get_task_comm(char *to, struct task_struct *tsk);
2972 void scheduler_ipi(void);
2973 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2975 static inline void scheduler_ipi(void) { }
2976 static inline unsigned long wait_task_inactive(struct task_struct *p,
2983 #define tasklist_empty() \
2984 list_empty(&init_task.tasks)
2986 #define next_task(p) \
2987 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2989 #define for_each_process(p) \
2990 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2992 extern bool current_is_single_threaded(void);
2995 * Careful: do_each_thread/while_each_thread is a double loop so
2996 * 'break' will not work as expected - use goto instead.
2998 #define do_each_thread(g, t) \
2999 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3001 #define while_each_thread(g, t) \
3002 while ((t = next_thread(t)) != g)
3004 #define __for_each_thread(signal, t) \
3005 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3007 #define for_each_thread(p, t) \
3008 __for_each_thread((p)->signal, t)
3010 /* Careful: this is a double loop, 'break' won't work as expected. */
3011 #define for_each_process_thread(p, t) \
3012 for_each_process(p) for_each_thread(p, t)
3014 static inline int get_nr_threads(struct task_struct *tsk)
3016 return tsk->signal->nr_threads;
3019 static inline bool thread_group_leader(struct task_struct *p)
3021 return p->exit_signal >= 0;
3024 /* Do to the insanities of de_thread it is possible for a process
3025 * to have the pid of the thread group leader without actually being
3026 * the thread group leader. For iteration through the pids in proc
3027 * all we care about is that we have a task with the appropriate
3028 * pid, we don't actually care if we have the right task.
3030 static inline bool has_group_leader_pid(struct task_struct *p)
3032 return task_pid(p) == p->signal->leader_pid;
3036 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3038 return p1->signal == p2->signal;
3041 static inline struct task_struct *next_thread(const struct task_struct *p)
3043 return list_entry_rcu(p->thread_group.next,
3044 struct task_struct, thread_group);
3047 static inline int thread_group_empty(struct task_struct *p)
3049 return list_empty(&p->thread_group);
3052 #define delay_group_leader(p) \
3053 (thread_group_leader(p) && !thread_group_empty(p))
3056 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3057 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3058 * pins the final release of task.io_context. Also protects ->cpuset and
3059 * ->cgroup.subsys[]. And ->vfork_done.
3061 * Nests both inside and outside of read_lock(&tasklist_lock).
3062 * It must not be nested with write_lock_irq(&tasklist_lock),
3063 * neither inside nor outside.
3065 static inline void task_lock(struct task_struct *p)
3067 spin_lock(&p->alloc_lock);
3070 static inline void task_unlock(struct task_struct *p)
3072 spin_unlock(&p->alloc_lock);
3075 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3076 unsigned long *flags);
3078 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3079 unsigned long *flags)
3081 struct sighand_struct *ret;
3083 ret = __lock_task_sighand(tsk, flags);
3084 (void)__cond_lock(&tsk->sighand->siglock, ret);
3088 static inline void unlock_task_sighand(struct task_struct *tsk,
3089 unsigned long *flags)
3091 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3095 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3096 * @tsk: task causing the changes
3098 * All operations which modify a threadgroup - a new thread joining the
3099 * group, death of a member thread (the assertion of PF_EXITING) and
3100 * exec(2) dethreading the process and replacing the leader - are wrapped
3101 * by threadgroup_change_{begin|end}(). This is to provide a place which
3102 * subsystems needing threadgroup stability can hook into for
3105 static inline void threadgroup_change_begin(struct task_struct *tsk)
3108 cgroup_threadgroup_change_begin(tsk);
3112 * threadgroup_change_end - mark the end of changes to a threadgroup
3113 * @tsk: task causing the changes
3115 * See threadgroup_change_begin().
3117 static inline void threadgroup_change_end(struct task_struct *tsk)
3119 cgroup_threadgroup_change_end(tsk);
3122 #ifdef CONFIG_THREAD_INFO_IN_TASK
3124 static inline struct thread_info *task_thread_info(struct task_struct *task)
3126 return &task->thread_info;
3130 * When accessing the stack of a non-current task that might exit, use
3131 * try_get_task_stack() instead. task_stack_page will return a pointer
3132 * that could get freed out from under you.
3134 static inline void *task_stack_page(const struct task_struct *task)
3139 #define setup_thread_stack(new,old) do { } while(0)
3141 static inline unsigned long *end_of_stack(const struct task_struct *task)
3146 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3148 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3149 #define task_stack_page(task) ((void *)(task)->stack)
3151 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3153 *task_thread_info(p) = *task_thread_info(org);
3154 task_thread_info(p)->task = p;
3158 * Return the address of the last usable long on the stack.
3160 * When the stack grows down, this is just above the thread
3161 * info struct. Going any lower will corrupt the threadinfo.
3163 * When the stack grows up, this is the highest address.
3164 * Beyond that position, we corrupt data on the next page.
3166 static inline unsigned long *end_of_stack(struct task_struct *p)
3168 #ifdef CONFIG_STACK_GROWSUP
3169 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3171 return (unsigned long *)(task_thread_info(p) + 1);
3177 #ifdef CONFIG_THREAD_INFO_IN_TASK
3178 static inline void *try_get_task_stack(struct task_struct *tsk)
3180 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3181 task_stack_page(tsk) : NULL;
3184 extern void put_task_stack(struct task_struct *tsk);
3186 static inline void *try_get_task_stack(struct task_struct *tsk)
3188 return task_stack_page(tsk);
3191 static inline void put_task_stack(struct task_struct *tsk) {}
3194 #define task_stack_end_corrupted(task) \
3195 (*(end_of_stack(task)) != STACK_END_MAGIC)
3197 static inline int object_is_on_stack(void *obj)
3199 void *stack = task_stack_page(current);
3201 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3204 extern void thread_stack_cache_init(void);
3206 #ifdef CONFIG_DEBUG_STACK_USAGE
3207 static inline unsigned long stack_not_used(struct task_struct *p)
3209 unsigned long *n = end_of_stack(p);
3211 do { /* Skip over canary */
3212 # ifdef CONFIG_STACK_GROWSUP
3219 # ifdef CONFIG_STACK_GROWSUP
3220 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3222 return (unsigned long)n - (unsigned long)end_of_stack(p);
3226 extern void set_task_stack_end_magic(struct task_struct *tsk);
3228 /* set thread flags in other task's structures
3229 * - see asm/thread_info.h for TIF_xxxx flags available
3231 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3233 set_ti_thread_flag(task_thread_info(tsk), flag);
3236 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3238 clear_ti_thread_flag(task_thread_info(tsk), flag);
3241 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3243 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3246 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3248 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3251 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3253 return test_ti_thread_flag(task_thread_info(tsk), flag);
3256 static inline void set_tsk_need_resched(struct task_struct *tsk)
3258 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3261 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3263 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3266 static inline int test_tsk_need_resched(struct task_struct *tsk)
3268 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3271 static inline int restart_syscall(void)
3273 set_tsk_thread_flag(current, TIF_SIGPENDING);
3274 return -ERESTARTNOINTR;
3277 static inline int signal_pending(struct task_struct *p)
3279 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3282 static inline int __fatal_signal_pending(struct task_struct *p)
3284 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3287 static inline int fatal_signal_pending(struct task_struct *p)
3289 return signal_pending(p) && __fatal_signal_pending(p);
3292 static inline int signal_pending_state(long state, struct task_struct *p)
3294 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3296 if (!signal_pending(p))
3299 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3303 * cond_resched() and cond_resched_lock(): latency reduction via
3304 * explicit rescheduling in places that are safe. The return
3305 * value indicates whether a reschedule was done in fact.
3306 * cond_resched_lock() will drop the spinlock before scheduling,
3307 * cond_resched_softirq() will enable bhs before scheduling.
3309 #ifndef CONFIG_PREEMPT
3310 extern int _cond_resched(void);
3312 static inline int _cond_resched(void) { return 0; }
3315 #define cond_resched() ({ \
3316 ___might_sleep(__FILE__, __LINE__, 0); \
3320 extern int __cond_resched_lock(spinlock_t *lock);
3322 #define cond_resched_lock(lock) ({ \
3323 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3324 __cond_resched_lock(lock); \
3327 extern int __cond_resched_softirq(void);
3329 #define cond_resched_softirq() ({ \
3330 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3331 __cond_resched_softirq(); \
3334 static inline void cond_resched_rcu(void)
3336 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3343 static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3345 #ifdef CONFIG_DEBUG_PREEMPT
3346 return p->preempt_disable_ip;
3353 * Does a critical section need to be broken due to another
3354 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3355 * but a general need for low latency)
3357 static inline int spin_needbreak(spinlock_t *lock)
3359 #ifdef CONFIG_PREEMPT
3360 return spin_is_contended(lock);
3367 * Idle thread specific functions to determine the need_resched
3370 #ifdef TIF_POLLING_NRFLAG
3371 static inline int tsk_is_polling(struct task_struct *p)
3373 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3376 static inline void __current_set_polling(void)
3378 set_thread_flag(TIF_POLLING_NRFLAG);
3381 static inline bool __must_check current_set_polling_and_test(void)
3383 __current_set_polling();
3386 * Polling state must be visible before we test NEED_RESCHED,
3387 * paired by resched_curr()
3389 smp_mb__after_atomic();
3391 return unlikely(tif_need_resched());
3394 static inline void __current_clr_polling(void)
3396 clear_thread_flag(TIF_POLLING_NRFLAG);
3399 static inline bool __must_check current_clr_polling_and_test(void)
3401 __current_clr_polling();
3404 * Polling state must be visible before we test NEED_RESCHED,
3405 * paired by resched_curr()
3407 smp_mb__after_atomic();
3409 return unlikely(tif_need_resched());
3413 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3414 static inline void __current_set_polling(void) { }
3415 static inline void __current_clr_polling(void) { }
3417 static inline bool __must_check current_set_polling_and_test(void)
3419 return unlikely(tif_need_resched());
3421 static inline bool __must_check current_clr_polling_and_test(void)
3423 return unlikely(tif_need_resched());
3427 static inline void current_clr_polling(void)
3429 __current_clr_polling();
3432 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3433 * Once the bit is cleared, we'll get IPIs with every new
3434 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3437 smp_mb(); /* paired with resched_curr() */
3439 preempt_fold_need_resched();
3442 static __always_inline bool need_resched(void)
3444 return unlikely(tif_need_resched());
3448 * Thread group CPU time accounting.
3450 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3451 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3454 * Reevaluate whether the task has signals pending delivery.
3455 * Wake the task if so.
3456 * This is required every time the blocked sigset_t changes.
3457 * callers must hold sighand->siglock.
3459 extern void recalc_sigpending_and_wake(struct task_struct *t);
3460 extern void recalc_sigpending(void);
3462 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3464 static inline void signal_wake_up(struct task_struct *t, bool resume)
3466 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3468 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3470 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3474 * Wrappers for p->thread_info->cpu access. No-op on UP.
3478 static inline unsigned int task_cpu(const struct task_struct *p)
3480 #ifdef CONFIG_THREAD_INFO_IN_TASK
3483 return task_thread_info(p)->cpu;
3487 static inline int task_node(const struct task_struct *p)
3489 return cpu_to_node(task_cpu(p));
3492 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3496 static inline unsigned int task_cpu(const struct task_struct *p)
3501 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3505 #endif /* CONFIG_SMP */
3507 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3508 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3510 #ifdef CONFIG_CGROUP_SCHED
3511 extern struct task_group root_task_group;
3512 #endif /* CONFIG_CGROUP_SCHED */
3514 extern int task_can_switch_user(struct user_struct *up,
3515 struct task_struct *tsk);
3517 #ifdef CONFIG_TASK_XACCT
3518 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3520 tsk->ioac.rchar += amt;
3523 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3525 tsk->ioac.wchar += amt;
3528 static inline void inc_syscr(struct task_struct *tsk)
3533 static inline void inc_syscw(struct task_struct *tsk)
3538 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3542 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3546 static inline void inc_syscr(struct task_struct *tsk)
3550 static inline void inc_syscw(struct task_struct *tsk)
3555 #ifndef TASK_SIZE_OF
3556 #define TASK_SIZE_OF(tsk) TASK_SIZE
3560 extern void mm_update_next_owner(struct mm_struct *mm);
3562 static inline void mm_update_next_owner(struct mm_struct *mm)
3565 #endif /* CONFIG_MEMCG */
3567 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3570 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3573 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3576 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3579 static inline unsigned long rlimit(unsigned int limit)
3581 return task_rlimit(current, limit);
3584 static inline unsigned long rlimit_max(unsigned int limit)
3586 return task_rlimit_max(current, limit);
3589 #define SCHED_CPUFREQ_RT (1U << 0)
3590 #define SCHED_CPUFREQ_DL (1U << 1)
3591 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3593 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3595 #ifdef CONFIG_CPU_FREQ
3596 struct update_util_data {
3597 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3600 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3601 void (*func)(struct update_util_data *data, u64 time,
3602 unsigned int flags));
3603 void cpufreq_remove_update_util_hook(int cpu);
3604 #endif /* CONFIG_CPU_FREQ */