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 */
528 #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
529 #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
531 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
533 struct sighand_struct {
535 struct k_sigaction action[_NSIG];
537 wait_queue_head_t signalfd_wqh;
540 struct pacct_struct {
543 unsigned long ac_mem;
544 cputime_t ac_utime, ac_stime;
545 unsigned long ac_minflt, ac_majflt;
556 * struct prev_cputime - snaphsot of system and user cputime
557 * @utime: time spent in user mode
558 * @stime: time spent in system mode
559 * @lock: protects the above two fields
561 * Stores previous user/system time values such that we can guarantee
564 struct prev_cputime {
565 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
572 static inline void prev_cputime_init(struct prev_cputime *prev)
574 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
575 prev->utime = prev->stime = 0;
576 raw_spin_lock_init(&prev->lock);
581 * struct task_cputime - collected CPU time counts
582 * @utime: time spent in user mode, in &cputime_t units
583 * @stime: time spent in kernel mode, in &cputime_t units
584 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
586 * This structure groups together three kinds of CPU time that are tracked for
587 * threads and thread groups. Most things considering CPU time want to group
588 * these counts together and treat all three of them in parallel.
590 struct task_cputime {
593 unsigned long long sum_exec_runtime;
596 /* Alternate field names when used to cache expirations. */
597 #define virt_exp utime
598 #define prof_exp stime
599 #define sched_exp sum_exec_runtime
601 #define INIT_CPUTIME \
602 (struct task_cputime) { \
605 .sum_exec_runtime = 0, \
609 * This is the atomic variant of task_cputime, which can be used for
610 * storing and updating task_cputime statistics without locking.
612 struct task_cputime_atomic {
615 atomic64_t sum_exec_runtime;
618 #define INIT_CPUTIME_ATOMIC \
619 (struct task_cputime_atomic) { \
620 .utime = ATOMIC64_INIT(0), \
621 .stime = ATOMIC64_INIT(0), \
622 .sum_exec_runtime = ATOMIC64_INIT(0), \
625 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
628 * Disable preemption until the scheduler is running -- use an unconditional
629 * value so that it also works on !PREEMPT_COUNT kernels.
631 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
633 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
636 * Initial preempt_count value; reflects the preempt_count schedule invariant
637 * which states that during context switches:
639 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
641 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
642 * Note: See finish_task_switch().
644 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
647 * struct thread_group_cputimer - thread group interval timer counts
648 * @cputime_atomic: atomic thread group interval timers.
649 * @running: true when there are timers running and
650 * @cputime_atomic receives updates.
651 * @checking_timer: true when a thread in the group is in the
652 * process of checking for thread group timers.
654 * This structure contains the version of task_cputime, above, that is
655 * used for thread group CPU timer calculations.
657 struct thread_group_cputimer {
658 struct task_cputime_atomic cputime_atomic;
663 #include <linux/rwsem.h>
667 * NOTE! "signal_struct" does not have its own
668 * locking, because a shared signal_struct always
669 * implies a shared sighand_struct, so locking
670 * sighand_struct is always a proper superset of
671 * the locking of signal_struct.
673 struct signal_struct {
677 struct list_head thread_head;
679 wait_queue_head_t wait_chldexit; /* for wait4() */
681 /* current thread group signal load-balancing target: */
682 struct task_struct *curr_target;
684 /* shared signal handling: */
685 struct sigpending shared_pending;
687 /* thread group exit support */
690 * - notify group_exit_task when ->count is equal to notify_count
691 * - everyone except group_exit_task is stopped during signal delivery
692 * of fatal signals, group_exit_task processes the signal.
695 struct task_struct *group_exit_task;
697 /* thread group stop support, overloads group_exit_code too */
698 int group_stop_count;
699 unsigned int flags; /* see SIGNAL_* flags below */
702 * PR_SET_CHILD_SUBREAPER marks a process, like a service
703 * manager, to re-parent orphan (double-forking) child processes
704 * to this process instead of 'init'. The service manager is
705 * able to receive SIGCHLD signals and is able to investigate
706 * the process until it calls wait(). All children of this
707 * process will inherit a flag if they should look for a
708 * child_subreaper process at exit.
710 unsigned int is_child_subreaper:1;
711 unsigned int has_child_subreaper:1;
713 /* POSIX.1b Interval Timers */
715 struct list_head posix_timers;
717 /* ITIMER_REAL timer for the process */
718 struct hrtimer real_timer;
719 struct pid *leader_pid;
720 ktime_t it_real_incr;
723 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
724 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
725 * values are defined to 0 and 1 respectively
727 struct cpu_itimer it[2];
730 * Thread group totals for process CPU timers.
731 * See thread_group_cputimer(), et al, for details.
733 struct thread_group_cputimer cputimer;
735 /* Earliest-expiration cache. */
736 struct task_cputime cputime_expires;
738 #ifdef CONFIG_NO_HZ_FULL
739 atomic_t tick_dep_mask;
742 struct list_head cpu_timers[3];
744 struct pid *tty_old_pgrp;
746 /* boolean value for session group leader */
749 struct tty_struct *tty; /* NULL if no tty */
751 #ifdef CONFIG_SCHED_AUTOGROUP
752 struct autogroup *autogroup;
755 * Cumulative resource counters for dead threads in the group,
756 * and for reaped dead child processes forked by this group.
757 * Live threads maintain their own counters and add to these
758 * in __exit_signal, except for the group leader.
760 seqlock_t stats_lock;
761 cputime_t utime, stime, cutime, cstime;
764 struct prev_cputime prev_cputime;
765 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
766 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
767 unsigned long inblock, oublock, cinblock, coublock;
768 unsigned long maxrss, cmaxrss;
769 struct task_io_accounting ioac;
772 * Cumulative ns of schedule CPU time fo dead threads in the
773 * group, not including a zombie group leader, (This only differs
774 * from jiffies_to_ns(utime + stime) if sched_clock uses something
775 * other than jiffies.)
777 unsigned long long sum_sched_runtime;
780 * We don't bother to synchronize most readers of this at all,
781 * because there is no reader checking a limit that actually needs
782 * to get both rlim_cur and rlim_max atomically, and either one
783 * alone is a single word that can safely be read normally.
784 * getrlimit/setrlimit use task_lock(current->group_leader) to
785 * protect this instead of the siglock, because they really
786 * have no need to disable irqs.
788 struct rlimit rlim[RLIM_NLIMITS];
790 #ifdef CONFIG_BSD_PROCESS_ACCT
791 struct pacct_struct pacct; /* per-process accounting information */
793 #ifdef CONFIG_TASKSTATS
794 struct taskstats *stats;
798 struct tty_audit_buf *tty_audit_buf;
802 * Thread is the potential origin of an oom condition; kill first on
805 bool oom_flag_origin;
806 short oom_score_adj; /* OOM kill score adjustment */
807 short oom_score_adj_min; /* OOM kill score adjustment min value.
808 * Only settable by CAP_SYS_RESOURCE. */
809 struct mm_struct *oom_mm; /* recorded mm when the thread group got
810 * killed by the oom killer */
812 struct mutex cred_guard_mutex; /* guard against foreign influences on
813 * credential calculations
814 * (notably. ptrace) */
818 * Bits in flags field of signal_struct.
820 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
821 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
822 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
823 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
825 * Pending notifications to parent.
827 #define SIGNAL_CLD_STOPPED 0x00000010
828 #define SIGNAL_CLD_CONTINUED 0x00000020
829 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
831 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
833 /* If true, all threads except ->group_exit_task have pending SIGKILL */
834 static inline int signal_group_exit(const struct signal_struct *sig)
836 return (sig->flags & SIGNAL_GROUP_EXIT) ||
837 (sig->group_exit_task != NULL);
841 * Some day this will be a full-fledged user tracking system..
844 atomic_t __count; /* reference count */
845 atomic_t processes; /* How many processes does this user have? */
846 atomic_t sigpending; /* How many pending signals does this user have? */
847 #ifdef CONFIG_INOTIFY_USER
848 atomic_t inotify_watches; /* How many inotify watches does this user have? */
849 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
851 #ifdef CONFIG_FANOTIFY
852 atomic_t fanotify_listeners;
855 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
857 #ifdef CONFIG_POSIX_MQUEUE
858 /* protected by mq_lock */
859 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
861 unsigned long locked_shm; /* How many pages of mlocked shm ? */
862 unsigned long unix_inflight; /* How many files in flight in unix sockets */
863 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
866 struct key *uid_keyring; /* UID specific keyring */
867 struct key *session_keyring; /* UID's default session keyring */
870 /* Hash table maintenance information */
871 struct hlist_node uidhash_node;
874 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
875 atomic_long_t locked_vm;
879 extern int uids_sysfs_init(void);
881 extern struct user_struct *find_user(kuid_t);
883 extern struct user_struct root_user;
884 #define INIT_USER (&root_user)
887 struct backing_dev_info;
888 struct reclaim_state;
890 #ifdef CONFIG_SCHED_INFO
892 /* cumulative counters */
893 unsigned long pcount; /* # of times run on this cpu */
894 unsigned long long run_delay; /* time spent waiting on a runqueue */
897 unsigned long long last_arrival,/* when we last ran on a cpu */
898 last_queued; /* when we were last queued to run */
900 #endif /* CONFIG_SCHED_INFO */
902 #ifdef CONFIG_TASK_DELAY_ACCT
903 struct task_delay_info {
905 unsigned int flags; /* Private per-task flags */
907 /* For each stat XXX, add following, aligned appropriately
909 * struct timespec XXX_start, XXX_end;
913 * Atomicity of updates to XXX_delay, XXX_count protected by
914 * single lock above (split into XXX_lock if contention is an issue).
918 * XXX_count is incremented on every XXX operation, the delay
919 * associated with the operation is added to XXX_delay.
920 * XXX_delay contains the accumulated delay time in nanoseconds.
922 u64 blkio_start; /* Shared by blkio, swapin */
923 u64 blkio_delay; /* wait for sync block io completion */
924 u64 swapin_delay; /* wait for swapin block io completion */
925 u32 blkio_count; /* total count of the number of sync block */
926 /* io operations performed */
927 u32 swapin_count; /* total count of the number of swapin block */
928 /* io operations performed */
931 u64 freepages_delay; /* wait for memory reclaim */
932 u32 freepages_count; /* total count of memory reclaim */
934 #endif /* CONFIG_TASK_DELAY_ACCT */
936 static inline int sched_info_on(void)
938 #ifdef CONFIG_SCHEDSTATS
940 #elif defined(CONFIG_TASK_DELAY_ACCT)
941 extern int delayacct_on;
948 #ifdef CONFIG_SCHEDSTATS
949 void force_schedstat_enabled(void);
960 * Integer metrics need fixed point arithmetic, e.g., sched/fair
961 * has a few: load, load_avg, util_avg, freq, and capacity.
963 * We define a basic fixed point arithmetic range, and then formalize
964 * all these metrics based on that basic range.
966 # define SCHED_FIXEDPOINT_SHIFT 10
967 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
970 * Increase resolution of cpu_capacity calculations
972 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
973 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
976 * Wake-queues are lists of tasks with a pending wakeup, whose
977 * callers have already marked the task as woken internally,
978 * and can thus carry on. A common use case is being able to
979 * do the wakeups once the corresponding user lock as been
982 * We hold reference to each task in the list across the wakeup,
983 * thus guaranteeing that the memory is still valid by the time
984 * the actual wakeups are performed in wake_up_q().
986 * One per task suffices, because there's never a need for a task to be
987 * in two wake queues simultaneously; it is forbidden to abandon a task
988 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
989 * already in a wake queue, the wakeup will happen soon and the second
990 * waker can just skip it.
992 * The WAKE_Q macro declares and initializes the list head.
993 * wake_up_q() does NOT reinitialize the list; it's expected to be
994 * called near the end of a function, where the fact that the queue is
995 * not used again will be easy to see by inspection.
997 * Note that this can cause spurious wakeups. schedule() callers
998 * must ensure the call is done inside a loop, confirming that the
999 * wakeup condition has in fact occurred.
1001 struct wake_q_node {
1002 struct wake_q_node *next;
1005 struct wake_q_head {
1006 struct wake_q_node *first;
1007 struct wake_q_node **lastp;
1010 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1012 #define WAKE_Q(name) \
1013 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1015 extern void wake_q_add(struct wake_q_head *head,
1016 struct task_struct *task);
1017 extern void wake_up_q(struct wake_q_head *head);
1020 * sched-domains (multiprocessor balancing) declarations:
1023 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1024 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1025 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1026 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1027 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1028 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1029 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1030 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1031 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1032 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1033 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1034 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1035 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1036 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1037 #define SD_NUMA 0x4000 /* cross-node balancing */
1039 #ifdef CONFIG_SCHED_SMT
1040 static inline int cpu_smt_flags(void)
1042 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1046 #ifdef CONFIG_SCHED_MC
1047 static inline int cpu_core_flags(void)
1049 return SD_SHARE_PKG_RESOURCES;
1054 static inline int cpu_numa_flags(void)
1060 struct sched_domain_attr {
1061 int relax_domain_level;
1064 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1065 .relax_domain_level = -1, \
1068 extern int sched_domain_level_max;
1072 struct sched_domain_shared {
1074 atomic_t nr_busy_cpus;
1078 struct sched_domain {
1079 /* These fields must be setup */
1080 struct sched_domain *parent; /* top domain must be null terminated */
1081 struct sched_domain *child; /* bottom domain must be null terminated */
1082 struct sched_group *groups; /* the balancing groups of the domain */
1083 unsigned long min_interval; /* Minimum balance interval ms */
1084 unsigned long max_interval; /* Maximum balance interval ms */
1085 unsigned int busy_factor; /* less balancing by factor if busy */
1086 unsigned int imbalance_pct; /* No balance until over watermark */
1087 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1088 unsigned int busy_idx;
1089 unsigned int idle_idx;
1090 unsigned int newidle_idx;
1091 unsigned int wake_idx;
1092 unsigned int forkexec_idx;
1093 unsigned int smt_gain;
1095 int nohz_idle; /* NOHZ IDLE status */
1096 int flags; /* See SD_* */
1099 /* Runtime fields. */
1100 unsigned long last_balance; /* init to jiffies. units in jiffies */
1101 unsigned int balance_interval; /* initialise to 1. units in ms. */
1102 unsigned int nr_balance_failed; /* initialise to 0 */
1104 /* idle_balance() stats */
1105 u64 max_newidle_lb_cost;
1106 unsigned long next_decay_max_lb_cost;
1108 u64 avg_scan_cost; /* select_idle_sibling */
1110 #ifdef CONFIG_SCHEDSTATS
1111 /* load_balance() stats */
1112 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1113 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1114 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1115 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1116 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1117 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1118 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1119 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1121 /* Active load balancing */
1122 unsigned int alb_count;
1123 unsigned int alb_failed;
1124 unsigned int alb_pushed;
1126 /* SD_BALANCE_EXEC stats */
1127 unsigned int sbe_count;
1128 unsigned int sbe_balanced;
1129 unsigned int sbe_pushed;
1131 /* SD_BALANCE_FORK stats */
1132 unsigned int sbf_count;
1133 unsigned int sbf_balanced;
1134 unsigned int sbf_pushed;
1136 /* try_to_wake_up() stats */
1137 unsigned int ttwu_wake_remote;
1138 unsigned int ttwu_move_affine;
1139 unsigned int ttwu_move_balance;
1141 #ifdef CONFIG_SCHED_DEBUG
1145 void *private; /* used during construction */
1146 struct rcu_head rcu; /* used during destruction */
1148 struct sched_domain_shared *shared;
1150 unsigned int span_weight;
1152 * Span of all CPUs in this domain.
1154 * NOTE: this field is variable length. (Allocated dynamically
1155 * by attaching extra space to the end of the structure,
1156 * depending on how many CPUs the kernel has booted up with)
1158 unsigned long span[0];
1161 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1163 return to_cpumask(sd->span);
1166 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1167 struct sched_domain_attr *dattr_new);
1169 /* Allocate an array of sched domains, for partition_sched_domains(). */
1170 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1171 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1173 bool cpus_share_cache(int this_cpu, int that_cpu);
1175 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1176 typedef int (*sched_domain_flags_f)(void);
1178 #define SDTL_OVERLAP 0x01
1181 struct sched_domain **__percpu sd;
1182 struct sched_domain_shared **__percpu sds;
1183 struct sched_group **__percpu sg;
1184 struct sched_group_capacity **__percpu sgc;
1187 struct sched_domain_topology_level {
1188 sched_domain_mask_f mask;
1189 sched_domain_flags_f sd_flags;
1192 struct sd_data data;
1193 #ifdef CONFIG_SCHED_DEBUG
1198 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1199 extern void wake_up_if_idle(int cpu);
1201 #ifdef CONFIG_SCHED_DEBUG
1202 # define SD_INIT_NAME(type) .name = #type
1204 # define SD_INIT_NAME(type)
1207 #else /* CONFIG_SMP */
1209 struct sched_domain_attr;
1212 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1213 struct sched_domain_attr *dattr_new)
1217 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1222 #endif /* !CONFIG_SMP */
1225 struct io_context; /* See blkdev.h */
1228 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1229 extern void prefetch_stack(struct task_struct *t);
1231 static inline void prefetch_stack(struct task_struct *t) { }
1234 struct audit_context; /* See audit.c */
1236 struct pipe_inode_info;
1237 struct uts_namespace;
1239 struct load_weight {
1240 unsigned long weight;
1245 * The load_avg/util_avg accumulates an infinite geometric series
1246 * (see __update_load_avg() in kernel/sched/fair.c).
1248 * [load_avg definition]
1250 * load_avg = runnable% * scale_load_down(load)
1252 * where runnable% is the time ratio that a sched_entity is runnable.
1253 * For cfs_rq, it is the aggregated load_avg of all runnable and
1254 * blocked sched_entities.
1256 * load_avg may also take frequency scaling into account:
1258 * load_avg = runnable% * scale_load_down(load) * freq%
1260 * where freq% is the CPU frequency normalized to the highest frequency.
1262 * [util_avg definition]
1264 * util_avg = running% * SCHED_CAPACITY_SCALE
1266 * where running% is the time ratio that a sched_entity is running on
1267 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1268 * and blocked sched_entities.
1270 * util_avg may also factor frequency scaling and CPU capacity scaling:
1272 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1274 * where freq% is the same as above, and capacity% is the CPU capacity
1275 * normalized to the greatest capacity (due to uarch differences, etc).
1277 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1278 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1279 * we therefore scale them to as large a range as necessary. This is for
1280 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1284 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1285 * with the highest load (=88761), always runnable on a single cfs_rq,
1286 * and should not overflow as the number already hits PID_MAX_LIMIT.
1288 * For all other cases (including 32-bit kernels), struct load_weight's
1289 * weight will overflow first before we do, because:
1291 * Max(load_avg) <= Max(load.weight)
1293 * Then it is the load_weight's responsibility to consider overflow
1297 u64 last_update_time, load_sum;
1298 u32 util_sum, period_contrib;
1299 unsigned long load_avg, util_avg;
1302 #ifdef CONFIG_SCHEDSTATS
1303 struct sched_statistics {
1313 s64 sum_sleep_runtime;
1320 u64 nr_migrations_cold;
1321 u64 nr_failed_migrations_affine;
1322 u64 nr_failed_migrations_running;
1323 u64 nr_failed_migrations_hot;
1324 u64 nr_forced_migrations;
1327 u64 nr_wakeups_sync;
1328 u64 nr_wakeups_migrate;
1329 u64 nr_wakeups_local;
1330 u64 nr_wakeups_remote;
1331 u64 nr_wakeups_affine;
1332 u64 nr_wakeups_affine_attempts;
1333 u64 nr_wakeups_passive;
1334 u64 nr_wakeups_idle;
1338 struct sched_entity {
1339 struct load_weight load; /* for load-balancing */
1340 struct rb_node run_node;
1341 struct list_head group_node;
1345 u64 sum_exec_runtime;
1347 u64 prev_sum_exec_runtime;
1351 #ifdef CONFIG_SCHEDSTATS
1352 struct sched_statistics statistics;
1355 #ifdef CONFIG_FAIR_GROUP_SCHED
1357 struct sched_entity *parent;
1358 /* rq on which this entity is (to be) queued: */
1359 struct cfs_rq *cfs_rq;
1360 /* rq "owned" by this entity/group: */
1361 struct cfs_rq *my_q;
1366 * Per entity load average tracking.
1368 * Put into separate cache line so it does not
1369 * collide with read-mostly values above.
1371 struct sched_avg avg ____cacheline_aligned_in_smp;
1375 struct sched_rt_entity {
1376 struct list_head run_list;
1377 unsigned long timeout;
1378 unsigned long watchdog_stamp;
1379 unsigned int time_slice;
1380 unsigned short on_rq;
1381 unsigned short on_list;
1383 struct sched_rt_entity *back;
1384 #ifdef CONFIG_RT_GROUP_SCHED
1385 struct sched_rt_entity *parent;
1386 /* rq on which this entity is (to be) queued: */
1387 struct rt_rq *rt_rq;
1388 /* rq "owned" by this entity/group: */
1393 struct sched_dl_entity {
1394 struct rb_node rb_node;
1397 * Original scheduling parameters. Copied here from sched_attr
1398 * during sched_setattr(), they will remain the same until
1399 * the next sched_setattr().
1401 u64 dl_runtime; /* maximum runtime for each instance */
1402 u64 dl_deadline; /* relative deadline of each instance */
1403 u64 dl_period; /* separation of two instances (period) */
1404 u64 dl_bw; /* dl_runtime / dl_deadline */
1407 * Actual scheduling parameters. Initialized with the values above,
1408 * they are continously updated during task execution. Note that
1409 * the remaining runtime could be < 0 in case we are in overrun.
1411 s64 runtime; /* remaining runtime for this instance */
1412 u64 deadline; /* absolute deadline for this instance */
1413 unsigned int flags; /* specifying the scheduler behaviour */
1418 * @dl_throttled tells if we exhausted the runtime. If so, the
1419 * task has to wait for a replenishment to be performed at the
1420 * next firing of dl_timer.
1422 * @dl_boosted tells if we are boosted due to DI. If so we are
1423 * outside bandwidth enforcement mechanism (but only until we
1424 * exit the critical section);
1426 * @dl_yielded tells if task gave up the cpu before consuming
1427 * all its available runtime during the last job.
1429 int dl_throttled, dl_boosted, dl_yielded;
1432 * Bandwidth enforcement timer. Each -deadline task has its
1433 * own bandwidth to be enforced, thus we need one timer per task.
1435 struct hrtimer dl_timer;
1443 u8 pad; /* Otherwise the compiler can store garbage here. */
1445 u32 s; /* Set of bits. */
1449 enum perf_event_task_context {
1450 perf_invalid_context = -1,
1451 perf_hw_context = 0,
1453 perf_nr_task_contexts,
1456 /* Track pages that require TLB flushes */
1457 struct tlbflush_unmap_batch {
1459 * Each bit set is a CPU that potentially has a TLB entry for one of
1460 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1462 struct cpumask cpumask;
1464 /* True if any bit in cpumask is set */
1465 bool flush_required;
1468 * If true then the PTE was dirty when unmapped. The entry must be
1469 * flushed before IO is initiated or a stale TLB entry potentially
1470 * allows an update without redirtying the page.
1475 struct task_struct {
1476 #ifdef CONFIG_THREAD_INFO_IN_TASK
1478 * For reasons of header soup (see current_thread_info()), this
1479 * must be the first element of task_struct.
1481 struct thread_info thread_info;
1483 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1486 unsigned int flags; /* per process flags, defined below */
1487 unsigned int ptrace;
1490 struct llist_node wake_entry;
1492 #ifdef CONFIG_THREAD_INFO_IN_TASK
1493 unsigned int cpu; /* current CPU */
1495 unsigned int wakee_flips;
1496 unsigned long wakee_flip_decay_ts;
1497 struct task_struct *last_wakee;
1503 int prio, static_prio, normal_prio;
1504 unsigned int rt_priority;
1505 const struct sched_class *sched_class;
1506 struct sched_entity se;
1507 struct sched_rt_entity rt;
1508 #ifdef CONFIG_CGROUP_SCHED
1509 struct task_group *sched_task_group;
1511 struct sched_dl_entity dl;
1513 #ifdef CONFIG_PREEMPT_NOTIFIERS
1514 /* list of struct preempt_notifier: */
1515 struct hlist_head preempt_notifiers;
1518 #ifdef CONFIG_BLK_DEV_IO_TRACE
1519 unsigned int btrace_seq;
1522 unsigned int policy;
1523 int nr_cpus_allowed;
1524 cpumask_t cpus_allowed;
1526 #ifdef CONFIG_PREEMPT_RCU
1527 int rcu_read_lock_nesting;
1528 union rcu_special rcu_read_unlock_special;
1529 struct list_head rcu_node_entry;
1530 struct rcu_node *rcu_blocked_node;
1531 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1532 #ifdef CONFIG_TASKS_RCU
1533 unsigned long rcu_tasks_nvcsw;
1534 bool rcu_tasks_holdout;
1535 struct list_head rcu_tasks_holdout_list;
1536 int rcu_tasks_idle_cpu;
1537 #endif /* #ifdef CONFIG_TASKS_RCU */
1539 #ifdef CONFIG_SCHED_INFO
1540 struct sched_info sched_info;
1543 struct list_head tasks;
1545 struct plist_node pushable_tasks;
1546 struct rb_node pushable_dl_tasks;
1549 struct mm_struct *mm, *active_mm;
1550 /* per-thread vma caching */
1551 u32 vmacache_seqnum;
1552 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1553 #if defined(SPLIT_RSS_COUNTING)
1554 struct task_rss_stat rss_stat;
1558 int exit_code, exit_signal;
1559 int pdeath_signal; /* The signal sent when the parent dies */
1560 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1562 /* Used for emulating ABI behavior of previous Linux versions */
1563 unsigned int personality;
1565 /* scheduler bits, serialized by scheduler locks */
1566 unsigned sched_reset_on_fork:1;
1567 unsigned sched_contributes_to_load:1;
1568 unsigned sched_migrated:1;
1569 unsigned sched_remote_wakeup:1;
1570 unsigned :0; /* force alignment to the next boundary */
1572 /* unserialized, strictly 'current' */
1573 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1574 unsigned in_iowait:1;
1575 #if !defined(TIF_RESTORE_SIGMASK)
1576 unsigned restore_sigmask:1;
1579 unsigned memcg_may_oom:1;
1581 unsigned memcg_kmem_skip_account:1;
1584 #ifdef CONFIG_COMPAT_BRK
1585 unsigned brk_randomized:1;
1588 unsigned long atomic_flags; /* Flags needing atomic access. */
1590 struct restart_block restart_block;
1595 #ifdef CONFIG_CC_STACKPROTECTOR
1596 /* Canary value for the -fstack-protector gcc feature */
1597 unsigned long stack_canary;
1600 * pointers to (original) parent process, youngest child, younger sibling,
1601 * older sibling, respectively. (p->father can be replaced with
1602 * p->real_parent->pid)
1604 struct task_struct __rcu *real_parent; /* real parent process */
1605 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1607 * children/sibling forms the list of my natural children
1609 struct list_head children; /* list of my children */
1610 struct list_head sibling; /* linkage in my parent's children list */
1611 struct task_struct *group_leader; /* threadgroup leader */
1614 * ptraced is the list of tasks this task is using ptrace on.
1615 * This includes both natural children and PTRACE_ATTACH targets.
1616 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1618 struct list_head ptraced;
1619 struct list_head ptrace_entry;
1621 /* PID/PID hash table linkage. */
1622 struct pid_link pids[PIDTYPE_MAX];
1623 struct list_head thread_group;
1624 struct list_head thread_node;
1626 struct completion *vfork_done; /* for vfork() */
1627 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1628 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1630 cputime_t utime, stime, utimescaled, stimescaled;
1632 struct prev_cputime prev_cputime;
1633 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1634 seqcount_t vtime_seqcount;
1635 unsigned long long vtime_snap;
1637 /* Task is sleeping or running in a CPU with VTIME inactive */
1639 /* Task runs in userspace in a CPU with VTIME active */
1641 /* Task runs in kernelspace in a CPU with VTIME active */
1643 } vtime_snap_whence;
1646 #ifdef CONFIG_NO_HZ_FULL
1647 atomic_t tick_dep_mask;
1649 unsigned long nvcsw, nivcsw; /* context switch counts */
1650 u64 start_time; /* monotonic time in nsec */
1651 u64 real_start_time; /* boot based time in nsec */
1652 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1653 unsigned long min_flt, maj_flt;
1655 struct task_cputime cputime_expires;
1656 struct list_head cpu_timers[3];
1658 /* process credentials */
1659 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1660 const struct cred __rcu *real_cred; /* objective and real subjective task
1661 * credentials (COW) */
1662 const struct cred __rcu *cred; /* effective (overridable) subjective task
1663 * credentials (COW) */
1664 char comm[TASK_COMM_LEN]; /* executable name excluding path
1665 - access with [gs]et_task_comm (which lock
1666 it with task_lock())
1667 - initialized normally by setup_new_exec */
1668 /* file system info */
1669 struct nameidata *nameidata;
1670 #ifdef CONFIG_SYSVIPC
1672 struct sysv_sem sysvsem;
1673 struct sysv_shm sysvshm;
1675 #ifdef CONFIG_DETECT_HUNG_TASK
1676 /* hung task detection */
1677 unsigned long last_switch_count;
1679 /* filesystem information */
1680 struct fs_struct *fs;
1681 /* open file information */
1682 struct files_struct *files;
1684 struct nsproxy *nsproxy;
1685 /* signal handlers */
1686 struct signal_struct *signal;
1687 struct sighand_struct *sighand;
1689 sigset_t blocked, real_blocked;
1690 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1691 struct sigpending pending;
1693 unsigned long sas_ss_sp;
1695 unsigned sas_ss_flags;
1697 struct callback_head *task_works;
1699 struct audit_context *audit_context;
1700 #ifdef CONFIG_AUDITSYSCALL
1702 unsigned int sessionid;
1704 struct seccomp seccomp;
1706 /* Thread group tracking */
1709 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1711 spinlock_t alloc_lock;
1713 /* Protection of the PI data structures: */
1714 raw_spinlock_t pi_lock;
1716 struct wake_q_node wake_q;
1718 #ifdef CONFIG_RT_MUTEXES
1719 /* PI waiters blocked on a rt_mutex held by this task */
1720 struct rb_root pi_waiters;
1721 struct rb_node *pi_waiters_leftmost;
1722 /* Deadlock detection and priority inheritance handling */
1723 struct rt_mutex_waiter *pi_blocked_on;
1726 #ifdef CONFIG_DEBUG_MUTEXES
1727 /* mutex deadlock detection */
1728 struct mutex_waiter *blocked_on;
1730 #ifdef CONFIG_TRACE_IRQFLAGS
1731 unsigned int irq_events;
1732 unsigned long hardirq_enable_ip;
1733 unsigned long hardirq_disable_ip;
1734 unsigned int hardirq_enable_event;
1735 unsigned int hardirq_disable_event;
1736 int hardirqs_enabled;
1737 int hardirq_context;
1738 unsigned long softirq_disable_ip;
1739 unsigned long softirq_enable_ip;
1740 unsigned int softirq_disable_event;
1741 unsigned int softirq_enable_event;
1742 int softirqs_enabled;
1743 int softirq_context;
1745 #ifdef CONFIG_LOCKDEP
1746 # define MAX_LOCK_DEPTH 48UL
1749 unsigned int lockdep_recursion;
1750 struct held_lock held_locks[MAX_LOCK_DEPTH];
1751 gfp_t lockdep_reclaim_gfp;
1754 unsigned int in_ubsan;
1757 /* journalling filesystem info */
1760 /* stacked block device info */
1761 struct bio_list *bio_list;
1764 /* stack plugging */
1765 struct blk_plug *plug;
1769 struct reclaim_state *reclaim_state;
1771 struct backing_dev_info *backing_dev_info;
1773 struct io_context *io_context;
1775 unsigned long ptrace_message;
1776 siginfo_t *last_siginfo; /* For ptrace use. */
1777 struct task_io_accounting ioac;
1778 #if defined(CONFIG_TASK_XACCT)
1779 u64 acct_rss_mem1; /* accumulated rss usage */
1780 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1781 cputime_t acct_timexpd; /* stime + utime since last update */
1783 #ifdef CONFIG_CPUSETS
1784 nodemask_t mems_allowed; /* Protected by alloc_lock */
1785 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1786 int cpuset_mem_spread_rotor;
1787 int cpuset_slab_spread_rotor;
1789 #ifdef CONFIG_CGROUPS
1790 /* Control Group info protected by css_set_lock */
1791 struct css_set __rcu *cgroups;
1792 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1793 struct list_head cg_list;
1796 struct robust_list_head __user *robust_list;
1797 #ifdef CONFIG_COMPAT
1798 struct compat_robust_list_head __user *compat_robust_list;
1800 struct list_head pi_state_list;
1801 struct futex_pi_state *pi_state_cache;
1803 #ifdef CONFIG_PERF_EVENTS
1804 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1805 struct mutex perf_event_mutex;
1806 struct list_head perf_event_list;
1808 #ifdef CONFIG_DEBUG_PREEMPT
1809 unsigned long preempt_disable_ip;
1812 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1814 short pref_node_fork;
1816 #ifdef CONFIG_NUMA_BALANCING
1818 unsigned int numa_scan_period;
1819 unsigned int numa_scan_period_max;
1820 int numa_preferred_nid;
1821 unsigned long numa_migrate_retry;
1822 u64 node_stamp; /* migration stamp */
1823 u64 last_task_numa_placement;
1824 u64 last_sum_exec_runtime;
1825 struct callback_head numa_work;
1827 struct list_head numa_entry;
1828 struct numa_group *numa_group;
1831 * numa_faults is an array split into four regions:
1832 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1833 * in this precise order.
1835 * faults_memory: Exponential decaying average of faults on a per-node
1836 * basis. Scheduling placement decisions are made based on these
1837 * counts. The values remain static for the duration of a PTE scan.
1838 * faults_cpu: Track the nodes the process was running on when a NUMA
1839 * hinting fault was incurred.
1840 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1841 * during the current scan window. When the scan completes, the counts
1842 * in faults_memory and faults_cpu decay and these values are copied.
1844 unsigned long *numa_faults;
1845 unsigned long total_numa_faults;
1848 * numa_faults_locality tracks if faults recorded during the last
1849 * scan window were remote/local or failed to migrate. The task scan
1850 * period is adapted based on the locality of the faults with different
1851 * weights depending on whether they were shared or private faults
1853 unsigned long numa_faults_locality[3];
1855 unsigned long numa_pages_migrated;
1856 #endif /* CONFIG_NUMA_BALANCING */
1858 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1859 struct tlbflush_unmap_batch tlb_ubc;
1862 struct rcu_head rcu;
1865 * cache last used pipe for splice
1867 struct pipe_inode_info *splice_pipe;
1869 struct page_frag task_frag;
1871 #ifdef CONFIG_TASK_DELAY_ACCT
1872 struct task_delay_info *delays;
1874 #ifdef CONFIG_FAULT_INJECTION
1878 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1879 * balance_dirty_pages() for some dirty throttling pause
1882 int nr_dirtied_pause;
1883 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1885 #ifdef CONFIG_LATENCYTOP
1886 int latency_record_count;
1887 struct latency_record latency_record[LT_SAVECOUNT];
1890 * time slack values; these are used to round up poll() and
1891 * select() etc timeout values. These are in nanoseconds.
1894 u64 default_timer_slack_ns;
1897 unsigned int kasan_depth;
1899 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1900 /* Index of current stored address in ret_stack */
1902 /* Stack of return addresses for return function tracing */
1903 struct ftrace_ret_stack *ret_stack;
1904 /* time stamp for last schedule */
1905 unsigned long long ftrace_timestamp;
1907 * Number of functions that haven't been traced
1908 * because of depth overrun.
1910 atomic_t trace_overrun;
1911 /* Pause for the tracing */
1912 atomic_t tracing_graph_pause;
1914 #ifdef CONFIG_TRACING
1915 /* state flags for use by tracers */
1916 unsigned long trace;
1917 /* bitmask and counter of trace recursion */
1918 unsigned long trace_recursion;
1919 #endif /* CONFIG_TRACING */
1921 /* Coverage collection mode enabled for this task (0 if disabled). */
1922 enum kcov_mode kcov_mode;
1923 /* Size of the kcov_area. */
1925 /* Buffer for coverage collection. */
1927 /* kcov desciptor wired with this task or NULL. */
1931 struct mem_cgroup *memcg_in_oom;
1932 gfp_t memcg_oom_gfp_mask;
1933 int memcg_oom_order;
1935 /* number of pages to reclaim on returning to userland */
1936 unsigned int memcg_nr_pages_over_high;
1938 #ifdef CONFIG_UPROBES
1939 struct uprobe_task *utask;
1941 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1942 unsigned int sequential_io;
1943 unsigned int sequential_io_avg;
1945 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1946 unsigned long task_state_change;
1948 int pagefault_disabled;
1950 struct task_struct *oom_reaper_list;
1952 #ifdef CONFIG_VMAP_STACK
1953 struct vm_struct *stack_vm_area;
1955 #ifdef CONFIG_THREAD_INFO_IN_TASK
1956 /* A live task holds one reference. */
1957 atomic_t stack_refcount;
1959 /* CPU-specific state of this task */
1960 struct thread_struct thread;
1962 * WARNING: on x86, 'thread_struct' contains a variable-sized
1963 * structure. It *MUST* be at the end of 'task_struct'.
1965 * Do not put anything below here!
1969 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1970 extern int arch_task_struct_size __read_mostly;
1972 # define arch_task_struct_size (sizeof(struct task_struct))
1975 #ifdef CONFIG_VMAP_STACK
1976 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1978 return t->stack_vm_area;
1981 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1987 /* Future-safe accessor for struct task_struct's cpus_allowed. */
1988 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1990 static inline int tsk_nr_cpus_allowed(struct task_struct *p)
1992 return p->nr_cpus_allowed;
1995 #define TNF_MIGRATED 0x01
1996 #define TNF_NO_GROUP 0x02
1997 #define TNF_SHARED 0x04
1998 #define TNF_FAULT_LOCAL 0x08
1999 #define TNF_MIGRATE_FAIL 0x10
2001 static inline bool in_vfork(struct task_struct *tsk)
2006 * need RCU to access ->real_parent if CLONE_VM was used along with
2009 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2012 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2013 * ->real_parent is not necessarily the task doing vfork(), so in
2014 * theory we can't rely on task_lock() if we want to dereference it.
2016 * And in this case we can't trust the real_parent->mm == tsk->mm
2017 * check, it can be false negative. But we do not care, if init or
2018 * another oom-unkillable task does this it should blame itself.
2021 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2027 #ifdef CONFIG_NUMA_BALANCING
2028 extern void task_numa_fault(int last_node, int node, int pages, int flags);
2029 extern pid_t task_numa_group_id(struct task_struct *p);
2030 extern void set_numabalancing_state(bool enabled);
2031 extern void task_numa_free(struct task_struct *p);
2032 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2033 int src_nid, int dst_cpu);
2035 static inline void task_numa_fault(int last_node, int node, int pages,
2039 static inline pid_t task_numa_group_id(struct task_struct *p)
2043 static inline void set_numabalancing_state(bool enabled)
2046 static inline void task_numa_free(struct task_struct *p)
2049 static inline bool should_numa_migrate_memory(struct task_struct *p,
2050 struct page *page, int src_nid, int dst_cpu)
2056 static inline struct pid *task_pid(struct task_struct *task)
2058 return task->pids[PIDTYPE_PID].pid;
2061 static inline struct pid *task_tgid(struct task_struct *task)
2063 return task->group_leader->pids[PIDTYPE_PID].pid;
2067 * Without tasklist or rcu lock it is not safe to dereference
2068 * the result of task_pgrp/task_session even if task == current,
2069 * we can race with another thread doing sys_setsid/sys_setpgid.
2071 static inline struct pid *task_pgrp(struct task_struct *task)
2073 return task->group_leader->pids[PIDTYPE_PGID].pid;
2076 static inline struct pid *task_session(struct task_struct *task)
2078 return task->group_leader->pids[PIDTYPE_SID].pid;
2081 struct pid_namespace;
2084 * the helpers to get the task's different pids as they are seen
2085 * from various namespaces
2087 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2088 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2090 * task_xid_nr_ns() : id seen from the ns specified;
2092 * set_task_vxid() : assigns a virtual id to a task;
2094 * see also pid_nr() etc in include/linux/pid.h
2096 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2097 struct pid_namespace *ns);
2099 static inline pid_t task_pid_nr(struct task_struct *tsk)
2104 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2105 struct pid_namespace *ns)
2107 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2110 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2112 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2116 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2121 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2123 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2125 return pid_vnr(task_tgid(tsk));
2129 static inline int pid_alive(const struct task_struct *p);
2130 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2136 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2142 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2144 return task_ppid_nr_ns(tsk, &init_pid_ns);
2147 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2148 struct pid_namespace *ns)
2150 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2153 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2155 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2159 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2160 struct pid_namespace *ns)
2162 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2165 static inline pid_t task_session_vnr(struct task_struct *tsk)
2167 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2170 /* obsolete, do not use */
2171 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2173 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2177 * pid_alive - check that a task structure is not stale
2178 * @p: Task structure to be checked.
2180 * Test if a process is not yet dead (at most zombie state)
2181 * If pid_alive fails, then pointers within the task structure
2182 * can be stale and must not be dereferenced.
2184 * Return: 1 if the process is alive. 0 otherwise.
2186 static inline int pid_alive(const struct task_struct *p)
2188 return p->pids[PIDTYPE_PID].pid != NULL;
2192 * is_global_init - check if a task structure is init. Since init
2193 * is free to have sub-threads we need to check tgid.
2194 * @tsk: Task structure to be checked.
2196 * Check if a task structure is the first user space task the kernel created.
2198 * Return: 1 if the task structure is init. 0 otherwise.
2200 static inline int is_global_init(struct task_struct *tsk)
2202 return task_tgid_nr(tsk) == 1;
2205 extern struct pid *cad_pid;
2207 extern void free_task(struct task_struct *tsk);
2208 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2210 extern void __put_task_struct(struct task_struct *t);
2212 static inline void put_task_struct(struct task_struct *t)
2214 if (atomic_dec_and_test(&t->usage))
2215 __put_task_struct(t);
2218 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2219 struct task_struct *try_get_task_struct(struct task_struct **ptask);
2221 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2222 extern void task_cputime(struct task_struct *t,
2223 cputime_t *utime, cputime_t *stime);
2224 extern void task_cputime_scaled(struct task_struct *t,
2225 cputime_t *utimescaled, cputime_t *stimescaled);
2226 extern cputime_t task_gtime(struct task_struct *t);
2228 static inline void task_cputime(struct task_struct *t,
2229 cputime_t *utime, cputime_t *stime)
2237 static inline void task_cputime_scaled(struct task_struct *t,
2238 cputime_t *utimescaled,
2239 cputime_t *stimescaled)
2242 *utimescaled = t->utimescaled;
2244 *stimescaled = t->stimescaled;
2247 static inline cputime_t task_gtime(struct task_struct *t)
2252 extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2253 extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2258 #define PF_EXITING 0x00000004 /* getting shut down */
2259 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2260 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2261 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2262 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2263 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2264 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2265 #define PF_DUMPCORE 0x00000200 /* dumped core */
2266 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2267 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2268 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2269 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2270 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2271 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2272 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2273 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2274 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2275 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2276 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2277 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2278 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2279 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2280 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2281 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2282 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2283 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2284 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2287 * Only the _current_ task can read/write to tsk->flags, but other
2288 * tasks can access tsk->flags in readonly mode for example
2289 * with tsk_used_math (like during threaded core dumping).
2290 * There is however an exception to this rule during ptrace
2291 * or during fork: the ptracer task is allowed to write to the
2292 * child->flags of its traced child (same goes for fork, the parent
2293 * can write to the child->flags), because we're guaranteed the
2294 * child is not running and in turn not changing child->flags
2295 * at the same time the parent does it.
2297 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2298 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2299 #define clear_used_math() clear_stopped_child_used_math(current)
2300 #define set_used_math() set_stopped_child_used_math(current)
2301 #define conditional_stopped_child_used_math(condition, child) \
2302 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2303 #define conditional_used_math(condition) \
2304 conditional_stopped_child_used_math(condition, current)
2305 #define copy_to_stopped_child_used_math(child) \
2306 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2307 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2308 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2309 #define used_math() tsk_used_math(current)
2311 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2312 * __GFP_FS is also cleared as it implies __GFP_IO.
2314 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2316 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2317 flags &= ~(__GFP_IO | __GFP_FS);
2321 static inline unsigned int memalloc_noio_save(void)
2323 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2324 current->flags |= PF_MEMALLOC_NOIO;
2328 static inline void memalloc_noio_restore(unsigned int flags)
2330 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2333 /* Per-process atomic flags. */
2334 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2335 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2336 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2337 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2340 #define TASK_PFA_TEST(name, func) \
2341 static inline bool task_##func(struct task_struct *p) \
2342 { return test_bit(PFA_##name, &p->atomic_flags); }
2343 #define TASK_PFA_SET(name, func) \
2344 static inline void task_set_##func(struct task_struct *p) \
2345 { set_bit(PFA_##name, &p->atomic_flags); }
2346 #define TASK_PFA_CLEAR(name, func) \
2347 static inline void task_clear_##func(struct task_struct *p) \
2348 { clear_bit(PFA_##name, &p->atomic_flags); }
2350 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2351 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2353 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2354 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2355 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2357 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2358 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2359 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2361 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2362 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2365 * task->jobctl flags
2367 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2369 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2370 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2371 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2372 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2373 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2374 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2375 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2377 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2378 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2379 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2380 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2381 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2382 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2383 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2385 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2386 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2388 extern bool task_set_jobctl_pending(struct task_struct *task,
2389 unsigned long mask);
2390 extern void task_clear_jobctl_trapping(struct task_struct *task);
2391 extern void task_clear_jobctl_pending(struct task_struct *task,
2392 unsigned long mask);
2394 static inline void rcu_copy_process(struct task_struct *p)
2396 #ifdef CONFIG_PREEMPT_RCU
2397 p->rcu_read_lock_nesting = 0;
2398 p->rcu_read_unlock_special.s = 0;
2399 p->rcu_blocked_node = NULL;
2400 INIT_LIST_HEAD(&p->rcu_node_entry);
2401 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2402 #ifdef CONFIG_TASKS_RCU
2403 p->rcu_tasks_holdout = false;
2404 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2405 p->rcu_tasks_idle_cpu = -1;
2406 #endif /* #ifdef CONFIG_TASKS_RCU */
2409 static inline void tsk_restore_flags(struct task_struct *task,
2410 unsigned long orig_flags, unsigned long flags)
2412 task->flags &= ~flags;
2413 task->flags |= orig_flags & flags;
2416 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2417 const struct cpumask *trial);
2418 extern int task_can_attach(struct task_struct *p,
2419 const struct cpumask *cs_cpus_allowed);
2421 extern void do_set_cpus_allowed(struct task_struct *p,
2422 const struct cpumask *new_mask);
2424 extern int set_cpus_allowed_ptr(struct task_struct *p,
2425 const struct cpumask *new_mask);
2427 static inline void do_set_cpus_allowed(struct task_struct *p,
2428 const struct cpumask *new_mask)
2431 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2432 const struct cpumask *new_mask)
2434 if (!cpumask_test_cpu(0, new_mask))
2440 #ifdef CONFIG_NO_HZ_COMMON
2441 void calc_load_enter_idle(void);
2442 void calc_load_exit_idle(void);
2444 static inline void calc_load_enter_idle(void) { }
2445 static inline void calc_load_exit_idle(void) { }
2446 #endif /* CONFIG_NO_HZ_COMMON */
2449 * Do not use outside of architecture code which knows its limitations.
2451 * sched_clock() has no promise of monotonicity or bounded drift between
2452 * CPUs, use (which you should not) requires disabling IRQs.
2454 * Please use one of the three interfaces below.
2456 extern unsigned long long notrace sched_clock(void);
2458 * See the comment in kernel/sched/clock.c
2460 extern u64 running_clock(void);
2461 extern u64 sched_clock_cpu(int cpu);
2464 extern void sched_clock_init(void);
2466 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2467 static inline void sched_clock_tick(void)
2471 static inline void sched_clock_idle_sleep_event(void)
2475 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2479 static inline u64 cpu_clock(int cpu)
2481 return sched_clock();
2484 static inline u64 local_clock(void)
2486 return sched_clock();
2490 * Architectures can set this to 1 if they have specified
2491 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2492 * but then during bootup it turns out that sched_clock()
2493 * is reliable after all:
2495 extern int sched_clock_stable(void);
2496 extern void set_sched_clock_stable(void);
2497 extern void clear_sched_clock_stable(void);
2499 extern void sched_clock_tick(void);
2500 extern void sched_clock_idle_sleep_event(void);
2501 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2504 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2505 * time source that is monotonic per cpu argument and has bounded drift
2508 * ######################### BIG FAT WARNING ##########################
2509 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2510 * # go backwards !! #
2511 * ####################################################################
2513 static inline u64 cpu_clock(int cpu)
2515 return sched_clock_cpu(cpu);
2518 static inline u64 local_clock(void)
2520 return sched_clock_cpu(raw_smp_processor_id());
2524 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2526 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2527 * The reason for this explicit opt-in is not to have perf penalty with
2528 * slow sched_clocks.
2530 extern void enable_sched_clock_irqtime(void);
2531 extern void disable_sched_clock_irqtime(void);
2533 static inline void enable_sched_clock_irqtime(void) {}
2534 static inline void disable_sched_clock_irqtime(void) {}
2537 extern unsigned long long
2538 task_sched_runtime(struct task_struct *task);
2540 /* sched_exec is called by processes performing an exec */
2542 extern void sched_exec(void);
2544 #define sched_exec() {}
2547 extern void sched_clock_idle_sleep_event(void);
2548 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2550 #ifdef CONFIG_HOTPLUG_CPU
2551 extern void idle_task_exit(void);
2553 static inline void idle_task_exit(void) {}
2556 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2557 extern void wake_up_nohz_cpu(int cpu);
2559 static inline void wake_up_nohz_cpu(int cpu) { }
2562 #ifdef CONFIG_NO_HZ_FULL
2563 extern u64 scheduler_tick_max_deferment(void);
2566 #ifdef CONFIG_SCHED_AUTOGROUP
2567 extern void sched_autogroup_create_attach(struct task_struct *p);
2568 extern void sched_autogroup_detach(struct task_struct *p);
2569 extern void sched_autogroup_fork(struct signal_struct *sig);
2570 extern void sched_autogroup_exit(struct signal_struct *sig);
2571 extern void sched_autogroup_exit_task(struct task_struct *p);
2572 #ifdef CONFIG_PROC_FS
2573 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2574 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2577 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2578 static inline void sched_autogroup_detach(struct task_struct *p) { }
2579 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2580 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2581 static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2584 extern int yield_to(struct task_struct *p, bool preempt);
2585 extern void set_user_nice(struct task_struct *p, long nice);
2586 extern int task_prio(const struct task_struct *p);
2588 * task_nice - return the nice value of a given task.
2589 * @p: the task in question.
2591 * Return: The nice value [ -20 ... 0 ... 19 ].
2593 static inline int task_nice(const struct task_struct *p)
2595 return PRIO_TO_NICE((p)->static_prio);
2597 extern int can_nice(const struct task_struct *p, const int nice);
2598 extern int task_curr(const struct task_struct *p);
2599 extern int idle_cpu(int cpu);
2600 extern int sched_setscheduler(struct task_struct *, int,
2601 const struct sched_param *);
2602 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2603 const struct sched_param *);
2604 extern int sched_setattr(struct task_struct *,
2605 const struct sched_attr *);
2606 extern struct task_struct *idle_task(int cpu);
2608 * is_idle_task - is the specified task an idle task?
2609 * @p: the task in question.
2611 * Return: 1 if @p is an idle task. 0 otherwise.
2613 static inline bool is_idle_task(const struct task_struct *p)
2617 extern struct task_struct *curr_task(int cpu);
2618 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2622 union thread_union {
2623 #ifndef CONFIG_THREAD_INFO_IN_TASK
2624 struct thread_info thread_info;
2626 unsigned long stack[THREAD_SIZE/sizeof(long)];
2629 #ifndef __HAVE_ARCH_KSTACK_END
2630 static inline int kstack_end(void *addr)
2632 /* Reliable end of stack detection:
2633 * Some APM bios versions misalign the stack
2635 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2639 extern union thread_union init_thread_union;
2640 extern struct task_struct init_task;
2642 extern struct mm_struct init_mm;
2644 extern struct pid_namespace init_pid_ns;
2647 * find a task by one of its numerical ids
2649 * find_task_by_pid_ns():
2650 * finds a task by its pid in the specified namespace
2651 * find_task_by_vpid():
2652 * finds a task by its virtual pid
2654 * see also find_vpid() etc in include/linux/pid.h
2657 extern struct task_struct *find_task_by_vpid(pid_t nr);
2658 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2659 struct pid_namespace *ns);
2661 /* per-UID process charging. */
2662 extern struct user_struct * alloc_uid(kuid_t);
2663 static inline struct user_struct *get_uid(struct user_struct *u)
2665 atomic_inc(&u->__count);
2668 extern void free_uid(struct user_struct *);
2670 #include <asm/current.h>
2672 extern void xtime_update(unsigned long ticks);
2674 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2675 extern int wake_up_process(struct task_struct *tsk);
2676 extern void wake_up_new_task(struct task_struct *tsk);
2678 extern void kick_process(struct task_struct *tsk);
2680 static inline void kick_process(struct task_struct *tsk) { }
2682 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2683 extern void sched_dead(struct task_struct *p);
2685 extern void proc_caches_init(void);
2686 extern void flush_signals(struct task_struct *);
2687 extern void ignore_signals(struct task_struct *);
2688 extern void flush_signal_handlers(struct task_struct *, int force_default);
2689 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2691 static inline int kernel_dequeue_signal(siginfo_t *info)
2693 struct task_struct *tsk = current;
2697 spin_lock_irq(&tsk->sighand->siglock);
2698 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2699 spin_unlock_irq(&tsk->sighand->siglock);
2704 static inline void kernel_signal_stop(void)
2706 spin_lock_irq(¤t->sighand->siglock);
2707 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2708 __set_current_state(TASK_STOPPED);
2709 spin_unlock_irq(¤t->sighand->siglock);
2714 extern void release_task(struct task_struct * p);
2715 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2716 extern int force_sigsegv(int, struct task_struct *);
2717 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2718 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2719 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2720 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2721 const struct cred *, u32);
2722 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2723 extern int kill_pid(struct pid *pid, int sig, int priv);
2724 extern int kill_proc_info(int, struct siginfo *, pid_t);
2725 extern __must_check bool do_notify_parent(struct task_struct *, int);
2726 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2727 extern void force_sig(int, struct task_struct *);
2728 extern int send_sig(int, struct task_struct *, int);
2729 extern int zap_other_threads(struct task_struct *p);
2730 extern struct sigqueue *sigqueue_alloc(void);
2731 extern void sigqueue_free(struct sigqueue *);
2732 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2733 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2735 #ifdef TIF_RESTORE_SIGMASK
2737 * Legacy restore_sigmask accessors. These are inefficient on
2738 * SMP architectures because they require atomic operations.
2742 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2744 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2745 * will run before returning to user mode, to process the flag. For
2746 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2747 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2748 * arch code will notice on return to user mode, in case those bits
2749 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2750 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2752 static inline void set_restore_sigmask(void)
2754 set_thread_flag(TIF_RESTORE_SIGMASK);
2755 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2757 static inline void clear_restore_sigmask(void)
2759 clear_thread_flag(TIF_RESTORE_SIGMASK);
2761 static inline bool test_restore_sigmask(void)
2763 return test_thread_flag(TIF_RESTORE_SIGMASK);
2765 static inline bool test_and_clear_restore_sigmask(void)
2767 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2770 #else /* TIF_RESTORE_SIGMASK */
2772 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2773 static inline void set_restore_sigmask(void)
2775 current->restore_sigmask = true;
2776 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2778 static inline void clear_restore_sigmask(void)
2780 current->restore_sigmask = false;
2782 static inline bool test_restore_sigmask(void)
2784 return current->restore_sigmask;
2786 static inline bool test_and_clear_restore_sigmask(void)
2788 if (!current->restore_sigmask)
2790 current->restore_sigmask = false;
2795 static inline void restore_saved_sigmask(void)
2797 if (test_and_clear_restore_sigmask())
2798 __set_current_blocked(¤t->saved_sigmask);
2801 static inline sigset_t *sigmask_to_save(void)
2803 sigset_t *res = ¤t->blocked;
2804 if (unlikely(test_restore_sigmask()))
2805 res = ¤t->saved_sigmask;
2809 static inline int kill_cad_pid(int sig, int priv)
2811 return kill_pid(cad_pid, sig, priv);
2814 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2815 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2816 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2817 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2820 * True if we are on the alternate signal stack.
2822 static inline int on_sig_stack(unsigned long sp)
2825 * If the signal stack is SS_AUTODISARM then, by construction, we
2826 * can't be on the signal stack unless user code deliberately set
2827 * SS_AUTODISARM when we were already on it.
2829 * This improves reliability: if user state gets corrupted such that
2830 * the stack pointer points very close to the end of the signal stack,
2831 * then this check will enable the signal to be handled anyway.
2833 if (current->sas_ss_flags & SS_AUTODISARM)
2836 #ifdef CONFIG_STACK_GROWSUP
2837 return sp >= current->sas_ss_sp &&
2838 sp - current->sas_ss_sp < current->sas_ss_size;
2840 return sp > current->sas_ss_sp &&
2841 sp - current->sas_ss_sp <= current->sas_ss_size;
2845 static inline int sas_ss_flags(unsigned long sp)
2847 if (!current->sas_ss_size)
2850 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2853 static inline void sas_ss_reset(struct task_struct *p)
2857 p->sas_ss_flags = SS_DISABLE;
2860 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2862 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2863 #ifdef CONFIG_STACK_GROWSUP
2864 return current->sas_ss_sp;
2866 return current->sas_ss_sp + current->sas_ss_size;
2872 * Routines for handling mm_structs
2874 extern struct mm_struct * mm_alloc(void);
2876 /* mmdrop drops the mm and the page tables */
2877 extern void __mmdrop(struct mm_struct *);
2878 static inline void mmdrop(struct mm_struct *mm)
2880 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2884 static inline void mmdrop_async_fn(struct work_struct *work)
2886 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2890 static inline void mmdrop_async(struct mm_struct *mm)
2892 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2893 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2894 schedule_work(&mm->async_put_work);
2898 static inline bool mmget_not_zero(struct mm_struct *mm)
2900 return atomic_inc_not_zero(&mm->mm_users);
2903 /* mmput gets rid of the mappings and all user-space */
2904 extern void mmput(struct mm_struct *);
2906 /* same as above but performs the slow path from the async context. Can
2907 * be called from the atomic context as well
2909 extern void mmput_async(struct mm_struct *);
2912 /* Grab a reference to a task's mm, if it is not already going away */
2913 extern struct mm_struct *get_task_mm(struct task_struct *task);
2915 * Grab a reference to a task's mm, if it is not already going away
2916 * and ptrace_may_access with the mode parameter passed to it
2919 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2920 /* Remove the current tasks stale references to the old mm_struct */
2921 extern void mm_release(struct task_struct *, struct mm_struct *);
2923 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2924 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2925 struct task_struct *, unsigned long);
2927 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2928 struct task_struct *);
2930 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2931 * via pt_regs, so ignore the tls argument passed via C. */
2932 static inline int copy_thread_tls(
2933 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2934 struct task_struct *p, unsigned long tls)
2936 return copy_thread(clone_flags, sp, arg, p);
2939 extern void flush_thread(void);
2941 #ifdef CONFIG_HAVE_EXIT_THREAD
2942 extern void exit_thread(struct task_struct *tsk);
2944 static inline void exit_thread(struct task_struct *tsk)
2949 extern void exit_files(struct task_struct *);
2950 extern void __cleanup_sighand(struct sighand_struct *);
2952 extern void exit_itimers(struct signal_struct *);
2953 extern void flush_itimer_signals(void);
2955 extern void do_group_exit(int);
2957 extern int do_execve(struct filename *,
2958 const char __user * const __user *,
2959 const char __user * const __user *);
2960 extern int do_execveat(int, struct filename *,
2961 const char __user * const __user *,
2962 const char __user * const __user *,
2964 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
2965 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2966 struct task_struct *fork_idle(int);
2967 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2969 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2970 static inline void set_task_comm(struct task_struct *tsk, const char *from)
2972 __set_task_comm(tsk, from, false);
2974 extern char *get_task_comm(char *to, struct task_struct *tsk);
2977 void scheduler_ipi(void);
2978 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2980 static inline void scheduler_ipi(void) { }
2981 static inline unsigned long wait_task_inactive(struct task_struct *p,
2988 #define tasklist_empty() \
2989 list_empty(&init_task.tasks)
2991 #define next_task(p) \
2992 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2994 #define for_each_process(p) \
2995 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2997 extern bool current_is_single_threaded(void);
3000 * Careful: do_each_thread/while_each_thread is a double loop so
3001 * 'break' will not work as expected - use goto instead.
3003 #define do_each_thread(g, t) \
3004 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3006 #define while_each_thread(g, t) \
3007 while ((t = next_thread(t)) != g)
3009 #define __for_each_thread(signal, t) \
3010 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3012 #define for_each_thread(p, t) \
3013 __for_each_thread((p)->signal, t)
3015 /* Careful: this is a double loop, 'break' won't work as expected. */
3016 #define for_each_process_thread(p, t) \
3017 for_each_process(p) for_each_thread(p, t)
3019 static inline int get_nr_threads(struct task_struct *tsk)
3021 return tsk->signal->nr_threads;
3024 static inline bool thread_group_leader(struct task_struct *p)
3026 return p->exit_signal >= 0;
3029 /* Do to the insanities of de_thread it is possible for a process
3030 * to have the pid of the thread group leader without actually being
3031 * the thread group leader. For iteration through the pids in proc
3032 * all we care about is that we have a task with the appropriate
3033 * pid, we don't actually care if we have the right task.
3035 static inline bool has_group_leader_pid(struct task_struct *p)
3037 return task_pid(p) == p->signal->leader_pid;
3041 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3043 return p1->signal == p2->signal;
3046 static inline struct task_struct *next_thread(const struct task_struct *p)
3048 return list_entry_rcu(p->thread_group.next,
3049 struct task_struct, thread_group);
3052 static inline int thread_group_empty(struct task_struct *p)
3054 return list_empty(&p->thread_group);
3057 #define delay_group_leader(p) \
3058 (thread_group_leader(p) && !thread_group_empty(p))
3061 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3062 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3063 * pins the final release of task.io_context. Also protects ->cpuset and
3064 * ->cgroup.subsys[]. And ->vfork_done.
3066 * Nests both inside and outside of read_lock(&tasklist_lock).
3067 * It must not be nested with write_lock_irq(&tasklist_lock),
3068 * neither inside nor outside.
3070 static inline void task_lock(struct task_struct *p)
3072 spin_lock(&p->alloc_lock);
3075 static inline void task_unlock(struct task_struct *p)
3077 spin_unlock(&p->alloc_lock);
3080 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3081 unsigned long *flags);
3083 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3084 unsigned long *flags)
3086 struct sighand_struct *ret;
3088 ret = __lock_task_sighand(tsk, flags);
3089 (void)__cond_lock(&tsk->sighand->siglock, ret);
3093 static inline void unlock_task_sighand(struct task_struct *tsk,
3094 unsigned long *flags)
3096 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3100 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3101 * @tsk: task causing the changes
3103 * All operations which modify a threadgroup - a new thread joining the
3104 * group, death of a member thread (the assertion of PF_EXITING) and
3105 * exec(2) dethreading the process and replacing the leader - are wrapped
3106 * by threadgroup_change_{begin|end}(). This is to provide a place which
3107 * subsystems needing threadgroup stability can hook into for
3110 static inline void threadgroup_change_begin(struct task_struct *tsk)
3113 cgroup_threadgroup_change_begin(tsk);
3117 * threadgroup_change_end - mark the end of changes to a threadgroup
3118 * @tsk: task causing the changes
3120 * See threadgroup_change_begin().
3122 static inline void threadgroup_change_end(struct task_struct *tsk)
3124 cgroup_threadgroup_change_end(tsk);
3127 #ifdef CONFIG_THREAD_INFO_IN_TASK
3129 static inline struct thread_info *task_thread_info(struct task_struct *task)
3131 return &task->thread_info;
3135 * When accessing the stack of a non-current task that might exit, use
3136 * try_get_task_stack() instead. task_stack_page will return a pointer
3137 * that could get freed out from under you.
3139 static inline void *task_stack_page(const struct task_struct *task)
3144 #define setup_thread_stack(new,old) do { } while(0)
3146 static inline unsigned long *end_of_stack(const struct task_struct *task)
3151 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3153 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3154 #define task_stack_page(task) ((void *)(task)->stack)
3156 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3158 *task_thread_info(p) = *task_thread_info(org);
3159 task_thread_info(p)->task = p;
3163 * Return the address of the last usable long on the stack.
3165 * When the stack grows down, this is just above the thread
3166 * info struct. Going any lower will corrupt the threadinfo.
3168 * When the stack grows up, this is the highest address.
3169 * Beyond that position, we corrupt data on the next page.
3171 static inline unsigned long *end_of_stack(struct task_struct *p)
3173 #ifdef CONFIG_STACK_GROWSUP
3174 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3176 return (unsigned long *)(task_thread_info(p) + 1);
3182 #ifdef CONFIG_THREAD_INFO_IN_TASK
3183 static inline void *try_get_task_stack(struct task_struct *tsk)
3185 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3186 task_stack_page(tsk) : NULL;
3189 extern void put_task_stack(struct task_struct *tsk);
3191 static inline void *try_get_task_stack(struct task_struct *tsk)
3193 return task_stack_page(tsk);
3196 static inline void put_task_stack(struct task_struct *tsk) {}
3199 #define task_stack_end_corrupted(task) \
3200 (*(end_of_stack(task)) != STACK_END_MAGIC)
3202 static inline int object_is_on_stack(void *obj)
3204 void *stack = task_stack_page(current);
3206 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3209 extern void thread_stack_cache_init(void);
3211 #ifdef CONFIG_DEBUG_STACK_USAGE
3212 static inline unsigned long stack_not_used(struct task_struct *p)
3214 unsigned long *n = end_of_stack(p);
3216 do { /* Skip over canary */
3217 # ifdef CONFIG_STACK_GROWSUP
3224 # ifdef CONFIG_STACK_GROWSUP
3225 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3227 return (unsigned long)n - (unsigned long)end_of_stack(p);
3231 extern void set_task_stack_end_magic(struct task_struct *tsk);
3233 /* set thread flags in other task's structures
3234 * - see asm/thread_info.h for TIF_xxxx flags available
3236 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3238 set_ti_thread_flag(task_thread_info(tsk), flag);
3241 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3243 clear_ti_thread_flag(task_thread_info(tsk), flag);
3246 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3248 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3251 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3253 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3256 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3258 return test_ti_thread_flag(task_thread_info(tsk), flag);
3261 static inline void set_tsk_need_resched(struct task_struct *tsk)
3263 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3266 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3268 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3271 static inline int test_tsk_need_resched(struct task_struct *tsk)
3273 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3276 static inline int restart_syscall(void)
3278 set_tsk_thread_flag(current, TIF_SIGPENDING);
3279 return -ERESTARTNOINTR;
3282 static inline int signal_pending(struct task_struct *p)
3284 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3287 static inline int __fatal_signal_pending(struct task_struct *p)
3289 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3292 static inline int fatal_signal_pending(struct task_struct *p)
3294 return signal_pending(p) && __fatal_signal_pending(p);
3297 static inline int signal_pending_state(long state, struct task_struct *p)
3299 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3301 if (!signal_pending(p))
3304 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3308 * cond_resched() and cond_resched_lock(): latency reduction via
3309 * explicit rescheduling in places that are safe. The return
3310 * value indicates whether a reschedule was done in fact.
3311 * cond_resched_lock() will drop the spinlock before scheduling,
3312 * cond_resched_softirq() will enable bhs before scheduling.
3314 #ifndef CONFIG_PREEMPT
3315 extern int _cond_resched(void);
3317 static inline int _cond_resched(void) { return 0; }
3320 #define cond_resched() ({ \
3321 ___might_sleep(__FILE__, __LINE__, 0); \
3325 extern int __cond_resched_lock(spinlock_t *lock);
3327 #define cond_resched_lock(lock) ({ \
3328 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3329 __cond_resched_lock(lock); \
3332 extern int __cond_resched_softirq(void);
3334 #define cond_resched_softirq() ({ \
3335 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3336 __cond_resched_softirq(); \
3339 static inline void cond_resched_rcu(void)
3341 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3348 static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3350 #ifdef CONFIG_DEBUG_PREEMPT
3351 return p->preempt_disable_ip;
3358 * Does a critical section need to be broken due to another
3359 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3360 * but a general need for low latency)
3362 static inline int spin_needbreak(spinlock_t *lock)
3364 #ifdef CONFIG_PREEMPT
3365 return spin_is_contended(lock);
3372 * Idle thread specific functions to determine the need_resched
3375 #ifdef TIF_POLLING_NRFLAG
3376 static inline int tsk_is_polling(struct task_struct *p)
3378 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3381 static inline void __current_set_polling(void)
3383 set_thread_flag(TIF_POLLING_NRFLAG);
3386 static inline bool __must_check current_set_polling_and_test(void)
3388 __current_set_polling();
3391 * Polling state must be visible before we test NEED_RESCHED,
3392 * paired by resched_curr()
3394 smp_mb__after_atomic();
3396 return unlikely(tif_need_resched());
3399 static inline void __current_clr_polling(void)
3401 clear_thread_flag(TIF_POLLING_NRFLAG);
3404 static inline bool __must_check current_clr_polling_and_test(void)
3406 __current_clr_polling();
3409 * Polling state must be visible before we test NEED_RESCHED,
3410 * paired by resched_curr()
3412 smp_mb__after_atomic();
3414 return unlikely(tif_need_resched());
3418 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3419 static inline void __current_set_polling(void) { }
3420 static inline void __current_clr_polling(void) { }
3422 static inline bool __must_check current_set_polling_and_test(void)
3424 return unlikely(tif_need_resched());
3426 static inline bool __must_check current_clr_polling_and_test(void)
3428 return unlikely(tif_need_resched());
3432 static inline void current_clr_polling(void)
3434 __current_clr_polling();
3437 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3438 * Once the bit is cleared, we'll get IPIs with every new
3439 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3442 smp_mb(); /* paired with resched_curr() */
3444 preempt_fold_need_resched();
3447 static __always_inline bool need_resched(void)
3449 return unlikely(tif_need_resched());
3453 * Thread group CPU time accounting.
3455 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3456 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3459 * Reevaluate whether the task has signals pending delivery.
3460 * Wake the task if so.
3461 * This is required every time the blocked sigset_t changes.
3462 * callers must hold sighand->siglock.
3464 extern void recalc_sigpending_and_wake(struct task_struct *t);
3465 extern void recalc_sigpending(void);
3467 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3469 static inline void signal_wake_up(struct task_struct *t, bool resume)
3471 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3473 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3475 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3479 * Wrappers for p->thread_info->cpu access. No-op on UP.
3483 static inline unsigned int task_cpu(const struct task_struct *p)
3485 #ifdef CONFIG_THREAD_INFO_IN_TASK
3488 return task_thread_info(p)->cpu;
3492 static inline int task_node(const struct task_struct *p)
3494 return cpu_to_node(task_cpu(p));
3497 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3501 static inline unsigned int task_cpu(const struct task_struct *p)
3506 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3510 #endif /* CONFIG_SMP */
3512 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3513 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3515 #ifdef CONFIG_CGROUP_SCHED
3516 extern struct task_group root_task_group;
3517 #endif /* CONFIG_CGROUP_SCHED */
3519 extern int task_can_switch_user(struct user_struct *up,
3520 struct task_struct *tsk);
3522 #ifdef CONFIG_TASK_XACCT
3523 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3525 tsk->ioac.rchar += amt;
3528 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3530 tsk->ioac.wchar += amt;
3533 static inline void inc_syscr(struct task_struct *tsk)
3538 static inline void inc_syscw(struct task_struct *tsk)
3543 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3547 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3551 static inline void inc_syscr(struct task_struct *tsk)
3555 static inline void inc_syscw(struct task_struct *tsk)
3560 #ifndef TASK_SIZE_OF
3561 #define TASK_SIZE_OF(tsk) TASK_SIZE
3565 extern void mm_update_next_owner(struct mm_struct *mm);
3567 static inline void mm_update_next_owner(struct mm_struct *mm)
3570 #endif /* CONFIG_MEMCG */
3572 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3575 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3578 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3581 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3584 static inline unsigned long rlimit(unsigned int limit)
3586 return task_rlimit(current, limit);
3589 static inline unsigned long rlimit_max(unsigned int limit)
3591 return task_rlimit_max(current, limit);
3594 #define SCHED_CPUFREQ_RT (1U << 0)
3595 #define SCHED_CPUFREQ_DL (1U << 1)
3596 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3598 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3600 #ifdef CONFIG_CPU_FREQ
3601 struct update_util_data {
3602 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3605 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3606 void (*func)(struct update_util_data *data, u64 time,
3607 unsigned int flags));
3608 void cpufreq_remove_update_util_hook(int cpu);
3609 #endif /* CONFIG_CPU_FREQ */