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>
33 #include <linux/smp.h>
34 #include <linux/sem.h>
35 #include <linux/shm.h>
36 #include <linux/signal.h>
37 #include <linux/compiler.h>
38 #include <linux/completion.h>
39 #include <linux/pid.h>
40 #include <linux/percpu.h>
41 #include <linux/topology.h>
42 #include <linux/seccomp.h>
43 #include <linux/rcupdate.h>
44 #include <linux/rculist.h>
45 #include <linux/rtmutex.h>
47 #include <linux/time.h>
48 #include <linux/param.h>
49 #include <linux/resource.h>
50 #include <linux/timer.h>
51 #include <linux/hrtimer.h>
52 #include <linux/kcov.h>
53 #include <linux/task_io_accounting.h>
54 #include <linux/latencytop.h>
55 #include <linux/cred.h>
56 #include <linux/llist.h>
57 #include <linux/uidgid.h>
58 #include <linux/gfp.h>
59 #include <linux/topology.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;
139 * These are the constant used to fake the fixed-point load-average
140 * counting. Some notes:
141 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
142 * a load-average precision of 10 bits integer + 11 bits fractional
143 * - if you want to count load-averages more often, you need more
144 * precision, or rounding will get you. With 2-second counting freq,
145 * the EXP_n values would be 1981, 2034 and 2043 if still using only
148 extern unsigned long avenrun[]; /* Load averages */
149 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
151 #define FSHIFT 11 /* nr of bits of precision */
152 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
153 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
154 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
155 #define EXP_5 2014 /* 1/exp(5sec/5min) */
156 #define EXP_15 2037 /* 1/exp(5sec/15min) */
158 #define CALC_LOAD(load,exp,n) \
160 load += n*(FIXED_1-exp); \
163 extern unsigned long total_forks;
164 extern int nr_threads;
165 DECLARE_PER_CPU(unsigned long, process_counts);
166 extern int nr_processes(void);
167 extern unsigned long nr_running(void);
168 extern bool single_task_running(void);
169 extern unsigned long nr_iowait(void);
170 extern unsigned long nr_iowait_cpu(int cpu);
171 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
173 extern void calc_global_load(unsigned long ticks);
175 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
176 extern void cpu_load_update_nohz_start(void);
177 extern void cpu_load_update_nohz_stop(void);
179 static inline void cpu_load_update_nohz_start(void) { }
180 static inline void cpu_load_update_nohz_stop(void) { }
183 extern void dump_cpu_task(int cpu);
188 #ifdef CONFIG_SCHED_DEBUG
189 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
190 extern void proc_sched_set_task(struct task_struct *p);
194 * Task state bitmask. NOTE! These bits are also
195 * encoded in fs/proc/array.c: get_task_state().
197 * We have two separate sets of flags: task->state
198 * is about runnability, while task->exit_state are
199 * about the task exiting. Confusing, but this way
200 * modifying one set can't modify the other one by
203 #define TASK_RUNNING 0
204 #define TASK_INTERRUPTIBLE 1
205 #define TASK_UNINTERRUPTIBLE 2
206 #define __TASK_STOPPED 4
207 #define __TASK_TRACED 8
208 /* in tsk->exit_state */
210 #define EXIT_ZOMBIE 32
211 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
212 /* in tsk->state again */
214 #define TASK_WAKEKILL 128
215 #define TASK_WAKING 256
216 #define TASK_PARKED 512
217 #define TASK_NOLOAD 1024
218 #define TASK_NEW 2048
219 #define TASK_STATE_MAX 4096
221 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
223 /* Convenience macros for the sake of set_current_state */
224 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
225 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
226 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
228 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
230 /* Convenience macros for the sake of wake_up */
231 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
232 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
234 /* get_task_state() */
235 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
236 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
237 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
239 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
240 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
241 #define task_is_stopped_or_traced(task) \
242 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
243 #define task_contributes_to_load(task) \
244 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
245 (task->flags & PF_FROZEN) == 0 && \
246 (task->state & TASK_NOLOAD) == 0)
248 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
250 #define __set_current_state(state_value) \
252 current->task_state_change = _THIS_IP_; \
253 current->state = (state_value); \
255 #define set_current_state(state_value) \
257 current->task_state_change = _THIS_IP_; \
258 smp_store_mb(current->state, (state_value)); \
263 * set_current_state() includes a barrier so that the write of current->state
264 * is correctly serialised wrt the caller's subsequent test of whether to
268 * set_current_state(TASK_UNINTERRUPTIBLE);
274 * __set_current_state(TASK_RUNNING);
276 * If the caller does not need such serialisation (because, for instance, the
277 * condition test and condition change and wakeup are under the same lock) then
278 * use __set_current_state().
280 * The above is typically ordered against the wakeup, which does:
282 * need_sleep = false;
283 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
285 * Where wake_up_state() (and all other wakeup primitives) imply enough
286 * barriers to order the store of the variable against wakeup.
288 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
289 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
290 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
292 * This is obviously fine, since they both store the exact same value.
294 * Also see the comments of try_to_wake_up().
296 #define __set_current_state(state_value) \
297 do { current->state = (state_value); } while (0)
298 #define set_current_state(state_value) \
299 smp_store_mb(current->state, (state_value))
303 /* Task command name length */
304 #define TASK_COMM_LEN 16
306 #include <linux/spinlock.h>
309 * This serializes "schedule()" and also protects
310 * the run-queue from deletions/modifications (but
311 * _adding_ to the beginning of the run-queue has
314 extern rwlock_t tasklist_lock;
315 extern spinlock_t mmlist_lock;
319 #ifdef CONFIG_PROVE_RCU
320 extern int lockdep_tasklist_lock_is_held(void);
321 #endif /* #ifdef CONFIG_PROVE_RCU */
323 extern void sched_init(void);
324 extern void sched_init_smp(void);
325 extern asmlinkage void schedule_tail(struct task_struct *prev);
326 extern void init_idle(struct task_struct *idle, int cpu);
327 extern void init_idle_bootup_task(struct task_struct *idle);
329 extern cpumask_var_t cpu_isolated_map;
331 extern int runqueue_is_locked(int cpu);
333 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
334 extern void nohz_balance_enter_idle(int cpu);
335 extern void set_cpu_sd_state_idle(void);
336 extern int get_nohz_timer_target(void);
338 static inline void nohz_balance_enter_idle(int cpu) { }
339 static inline void set_cpu_sd_state_idle(void) { }
343 * Only dump TASK_* tasks. (0 for all tasks)
345 extern void show_state_filter(unsigned long state_filter);
347 static inline void show_state(void)
349 show_state_filter(0);
352 extern void show_regs(struct pt_regs *);
355 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
356 * task), SP is the stack pointer of the first frame that should be shown in the back
357 * trace (or NULL if the entire call-chain of the task should be shown).
359 extern void show_stack(struct task_struct *task, unsigned long *sp);
361 extern void cpu_init (void);
362 extern void trap_init(void);
363 extern void update_process_times(int user);
364 extern void scheduler_tick(void);
365 extern int sched_cpu_starting(unsigned int cpu);
366 extern int sched_cpu_activate(unsigned int cpu);
367 extern int sched_cpu_deactivate(unsigned int cpu);
369 #ifdef CONFIG_HOTPLUG_CPU
370 extern int sched_cpu_dying(unsigned int cpu);
372 # define sched_cpu_dying NULL
375 extern void sched_show_task(struct task_struct *p);
377 #ifdef CONFIG_LOCKUP_DETECTOR
378 extern void touch_softlockup_watchdog_sched(void);
379 extern void touch_softlockup_watchdog(void);
380 extern void touch_softlockup_watchdog_sync(void);
381 extern void touch_all_softlockup_watchdogs(void);
382 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
384 size_t *lenp, loff_t *ppos);
385 extern unsigned int softlockup_panic;
386 extern unsigned int hardlockup_panic;
387 void lockup_detector_init(void);
389 static inline void touch_softlockup_watchdog_sched(void)
392 static inline void touch_softlockup_watchdog(void)
395 static inline void touch_softlockup_watchdog_sync(void)
398 static inline void touch_all_softlockup_watchdogs(void)
401 static inline void lockup_detector_init(void)
406 #ifdef CONFIG_DETECT_HUNG_TASK
407 void reset_hung_task_detector(void);
409 static inline void reset_hung_task_detector(void)
414 /* Attach to any functions which should be ignored in wchan output. */
415 #define __sched __attribute__((__section__(".sched.text")))
417 /* Linker adds these: start and end of __sched functions */
418 extern char __sched_text_start[], __sched_text_end[];
420 /* Is this address in the __sched functions? */
421 extern int in_sched_functions(unsigned long addr);
423 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
424 extern signed long schedule_timeout(signed long timeout);
425 extern signed long schedule_timeout_interruptible(signed long timeout);
426 extern signed long schedule_timeout_killable(signed long timeout);
427 extern signed long schedule_timeout_uninterruptible(signed long timeout);
428 extern signed long schedule_timeout_idle(signed long timeout);
429 asmlinkage void schedule(void);
430 extern void schedule_preempt_disabled(void);
432 extern int __must_check io_schedule_prepare(void);
433 extern void io_schedule_finish(int token);
434 extern long io_schedule_timeout(long timeout);
435 extern void io_schedule(void);
437 void __noreturn do_task_dead(void);
440 struct user_namespace;
443 extern void arch_pick_mmap_layout(struct mm_struct *mm);
445 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
446 unsigned long, unsigned long);
448 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
449 unsigned long len, unsigned long pgoff,
450 unsigned long flags);
452 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
455 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
456 #define SUID_DUMP_USER 1 /* Dump as user of process */
457 #define SUID_DUMP_ROOT 2 /* Dump as root */
461 /* for SUID_DUMP_* above */
462 #define MMF_DUMPABLE_BITS 2
463 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
465 extern void set_dumpable(struct mm_struct *mm, int value);
467 * This returns the actual value of the suid_dumpable flag. For things
468 * that are using this for checking for privilege transitions, it must
469 * test against SUID_DUMP_USER rather than treating it as a boolean
472 static inline int __get_dumpable(unsigned long mm_flags)
474 return mm_flags & MMF_DUMPABLE_MASK;
477 static inline int get_dumpable(struct mm_struct *mm)
479 return __get_dumpable(mm->flags);
482 /* coredump filter bits */
483 #define MMF_DUMP_ANON_PRIVATE 2
484 #define MMF_DUMP_ANON_SHARED 3
485 #define MMF_DUMP_MAPPED_PRIVATE 4
486 #define MMF_DUMP_MAPPED_SHARED 5
487 #define MMF_DUMP_ELF_HEADERS 6
488 #define MMF_DUMP_HUGETLB_PRIVATE 7
489 #define MMF_DUMP_HUGETLB_SHARED 8
490 #define MMF_DUMP_DAX_PRIVATE 9
491 #define MMF_DUMP_DAX_SHARED 10
493 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
494 #define MMF_DUMP_FILTER_BITS 9
495 #define MMF_DUMP_FILTER_MASK \
496 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
497 #define MMF_DUMP_FILTER_DEFAULT \
498 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
499 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
501 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
502 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
504 # define MMF_DUMP_MASK_DEFAULT_ELF 0
506 /* leave room for more dump flags */
507 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
508 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
510 * This one-shot flag is dropped due to necessity of changing exe once again
513 //#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
515 #define MMF_HAS_UPROBES 19 /* has uprobes */
516 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
517 #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
518 #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
519 #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
521 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
523 struct sighand_struct {
525 struct k_sigaction action[_NSIG];
527 wait_queue_head_t signalfd_wqh;
530 struct pacct_struct {
533 unsigned long ac_mem;
534 u64 ac_utime, ac_stime;
535 unsigned long ac_minflt, ac_majflt;
544 * struct prev_cputime - snaphsot of system and user cputime
545 * @utime: time spent in user mode
546 * @stime: time spent in system mode
547 * @lock: protects the above two fields
549 * Stores previous user/system time values such that we can guarantee
552 struct prev_cputime {
553 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
560 static inline void prev_cputime_init(struct prev_cputime *prev)
562 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
563 prev->utime = prev->stime = 0;
564 raw_spin_lock_init(&prev->lock);
569 * struct task_cputime - collected CPU time counts
570 * @utime: time spent in user mode, in nanoseconds
571 * @stime: time spent in kernel mode, in nanoseconds
572 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
574 * This structure groups together three kinds of CPU time that are tracked for
575 * threads and thread groups. Most things considering CPU time want to group
576 * these counts together and treat all three of them in parallel.
578 struct task_cputime {
581 unsigned long long sum_exec_runtime;
584 /* Alternate field names when used to cache expirations. */
585 #define virt_exp utime
586 #define prof_exp stime
587 #define sched_exp sum_exec_runtime
590 * This is the atomic variant of task_cputime, which can be used for
591 * storing and updating task_cputime statistics without locking.
593 struct task_cputime_atomic {
596 atomic64_t sum_exec_runtime;
599 #define INIT_CPUTIME_ATOMIC \
600 (struct task_cputime_atomic) { \
601 .utime = ATOMIC64_INIT(0), \
602 .stime = ATOMIC64_INIT(0), \
603 .sum_exec_runtime = ATOMIC64_INIT(0), \
606 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
609 * Disable preemption until the scheduler is running -- use an unconditional
610 * value so that it also works on !PREEMPT_COUNT kernels.
612 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
614 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
617 * Initial preempt_count value; reflects the preempt_count schedule invariant
618 * which states that during context switches:
620 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
622 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
623 * Note: See finish_task_switch().
625 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
628 * struct thread_group_cputimer - thread group interval timer counts
629 * @cputime_atomic: atomic thread group interval timers.
630 * @running: true when there are timers running and
631 * @cputime_atomic receives updates.
632 * @checking_timer: true when a thread in the group is in the
633 * process of checking for thread group timers.
635 * This structure contains the version of task_cputime, above, that is
636 * used for thread group CPU timer calculations.
638 struct thread_group_cputimer {
639 struct task_cputime_atomic cputime_atomic;
644 #include <linux/rwsem.h>
648 * NOTE! "signal_struct" does not have its own
649 * locking, because a shared signal_struct always
650 * implies a shared sighand_struct, so locking
651 * sighand_struct is always a proper superset of
652 * the locking of signal_struct.
654 struct signal_struct {
658 struct list_head thread_head;
660 wait_queue_head_t wait_chldexit; /* for wait4() */
662 /* current thread group signal load-balancing target: */
663 struct task_struct *curr_target;
665 /* shared signal handling: */
666 struct sigpending shared_pending;
668 /* thread group exit support */
671 * - notify group_exit_task when ->count is equal to notify_count
672 * - everyone except group_exit_task is stopped during signal delivery
673 * of fatal signals, group_exit_task processes the signal.
676 struct task_struct *group_exit_task;
678 /* thread group stop support, overloads group_exit_code too */
679 int group_stop_count;
680 unsigned int flags; /* see SIGNAL_* flags below */
683 * PR_SET_CHILD_SUBREAPER marks a process, like a service
684 * manager, to re-parent orphan (double-forking) child processes
685 * to this process instead of 'init'. The service manager is
686 * able to receive SIGCHLD signals and is able to investigate
687 * the process until it calls wait(). All children of this
688 * process will inherit a flag if they should look for a
689 * child_subreaper process at exit.
691 unsigned int is_child_subreaper:1;
692 unsigned int has_child_subreaper:1;
694 #ifdef CONFIG_POSIX_TIMERS
696 /* POSIX.1b Interval Timers */
698 struct list_head posix_timers;
700 /* ITIMER_REAL timer for the process */
701 struct hrtimer real_timer;
702 ktime_t it_real_incr;
705 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
706 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
707 * values are defined to 0 and 1 respectively
709 struct cpu_itimer it[2];
712 * Thread group totals for process CPU timers.
713 * See thread_group_cputimer(), et al, for details.
715 struct thread_group_cputimer cputimer;
717 /* Earliest-expiration cache. */
718 struct task_cputime cputime_expires;
720 struct list_head cpu_timers[3];
724 struct pid *leader_pid;
726 #ifdef CONFIG_NO_HZ_FULL
727 atomic_t tick_dep_mask;
730 struct pid *tty_old_pgrp;
732 /* boolean value for session group leader */
735 struct tty_struct *tty; /* NULL if no tty */
737 #ifdef CONFIG_SCHED_AUTOGROUP
738 struct autogroup *autogroup;
741 * Cumulative resource counters for dead threads in the group,
742 * and for reaped dead child processes forked by this group.
743 * Live threads maintain their own counters and add to these
744 * in __exit_signal, except for the group leader.
746 seqlock_t stats_lock;
747 u64 utime, stime, cutime, cstime;
750 struct prev_cputime prev_cputime;
751 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
752 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
753 unsigned long inblock, oublock, cinblock, coublock;
754 unsigned long maxrss, cmaxrss;
755 struct task_io_accounting ioac;
758 * Cumulative ns of schedule CPU time fo dead threads in the
759 * group, not including a zombie group leader, (This only differs
760 * from jiffies_to_ns(utime + stime) if sched_clock uses something
761 * other than jiffies.)
763 unsigned long long sum_sched_runtime;
766 * We don't bother to synchronize most readers of this at all,
767 * because there is no reader checking a limit that actually needs
768 * to get both rlim_cur and rlim_max atomically, and either one
769 * alone is a single word that can safely be read normally.
770 * getrlimit/setrlimit use task_lock(current->group_leader) to
771 * protect this instead of the siglock, because they really
772 * have no need to disable irqs.
774 struct rlimit rlim[RLIM_NLIMITS];
776 #ifdef CONFIG_BSD_PROCESS_ACCT
777 struct pacct_struct pacct; /* per-process accounting information */
779 #ifdef CONFIG_TASKSTATS
780 struct taskstats *stats;
784 struct tty_audit_buf *tty_audit_buf;
788 * Thread is the potential origin of an oom condition; kill first on
791 bool oom_flag_origin;
792 short oom_score_adj; /* OOM kill score adjustment */
793 short oom_score_adj_min; /* OOM kill score adjustment min value.
794 * Only settable by CAP_SYS_RESOURCE. */
795 struct mm_struct *oom_mm; /* recorded mm when the thread group got
796 * killed by the oom killer */
798 struct mutex cred_guard_mutex; /* guard against foreign influences on
799 * credential calculations
800 * (notably. ptrace) */
804 * Bits in flags field of signal_struct.
806 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
807 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
808 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
809 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
811 * Pending notifications to parent.
813 #define SIGNAL_CLD_STOPPED 0x00000010
814 #define SIGNAL_CLD_CONTINUED 0x00000020
815 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
817 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
819 #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
820 SIGNAL_STOP_CONTINUED)
822 static inline void signal_set_stop_flags(struct signal_struct *sig,
825 WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
826 sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
829 /* If true, all threads except ->group_exit_task have pending SIGKILL */
830 static inline int signal_group_exit(const struct signal_struct *sig)
832 return (sig->flags & SIGNAL_GROUP_EXIT) ||
833 (sig->group_exit_task != NULL);
837 * Some day this will be a full-fledged user tracking system..
840 atomic_t __count; /* reference count */
841 atomic_t processes; /* How many processes does this user have? */
842 atomic_t sigpending; /* How many pending signals does this user have? */
843 #ifdef CONFIG_FANOTIFY
844 atomic_t fanotify_listeners;
847 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
849 #ifdef CONFIG_POSIX_MQUEUE
850 /* protected by mq_lock */
851 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
853 unsigned long locked_shm; /* How many pages of mlocked shm ? */
854 unsigned long unix_inflight; /* How many files in flight in unix sockets */
855 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
858 struct key *uid_keyring; /* UID specific keyring */
859 struct key *session_keyring; /* UID's default session keyring */
862 /* Hash table maintenance information */
863 struct hlist_node uidhash_node;
866 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
867 atomic_long_t locked_vm;
871 extern int uids_sysfs_init(void);
873 extern struct user_struct *find_user(kuid_t);
875 extern struct user_struct root_user;
876 #define INIT_USER (&root_user)
879 struct backing_dev_info;
880 struct reclaim_state;
882 #ifdef CONFIG_SCHED_INFO
884 /* cumulative counters */
885 unsigned long pcount; /* # of times run on this cpu */
886 unsigned long long run_delay; /* time spent waiting on a runqueue */
889 unsigned long long last_arrival,/* when we last ran on a cpu */
890 last_queued; /* when we were last queued to run */
892 #endif /* CONFIG_SCHED_INFO */
894 #ifdef CONFIG_TASK_DELAY_ACCT
895 struct task_delay_info {
897 unsigned int flags; /* Private per-task flags */
899 /* For each stat XXX, add following, aligned appropriately
901 * struct timespec XXX_start, XXX_end;
905 * Atomicity of updates to XXX_delay, XXX_count protected by
906 * single lock above (split into XXX_lock if contention is an issue).
910 * XXX_count is incremented on every XXX operation, the delay
911 * associated with the operation is added to XXX_delay.
912 * XXX_delay contains the accumulated delay time in nanoseconds.
914 u64 blkio_start; /* Shared by blkio, swapin */
915 u64 blkio_delay; /* wait for sync block io completion */
916 u64 swapin_delay; /* wait for swapin block io completion */
917 u32 blkio_count; /* total count of the number of sync block */
918 /* io operations performed */
919 u32 swapin_count; /* total count of the number of swapin block */
920 /* io operations performed */
923 u64 freepages_delay; /* wait for memory reclaim */
924 u32 freepages_count; /* total count of memory reclaim */
926 #endif /* CONFIG_TASK_DELAY_ACCT */
928 static inline int sched_info_on(void)
930 #ifdef CONFIG_SCHEDSTATS
932 #elif defined(CONFIG_TASK_DELAY_ACCT)
933 extern int delayacct_on;
940 #ifdef CONFIG_SCHEDSTATS
941 void force_schedstat_enabled(void);
952 * Integer metrics need fixed point arithmetic, e.g., sched/fair
953 * has a few: load, load_avg, util_avg, freq, and capacity.
955 * We define a basic fixed point arithmetic range, and then formalize
956 * all these metrics based on that basic range.
958 # define SCHED_FIXEDPOINT_SHIFT 10
959 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
962 * Increase resolution of cpu_capacity calculations
964 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
965 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
968 * Wake-queues are lists of tasks with a pending wakeup, whose
969 * callers have already marked the task as woken internally,
970 * and can thus carry on. A common use case is being able to
971 * do the wakeups once the corresponding user lock as been
974 * We hold reference to each task in the list across the wakeup,
975 * thus guaranteeing that the memory is still valid by the time
976 * the actual wakeups are performed in wake_up_q().
978 * One per task suffices, because there's never a need for a task to be
979 * in two wake queues simultaneously; it is forbidden to abandon a task
980 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
981 * already in a wake queue, the wakeup will happen soon and the second
982 * waker can just skip it.
984 * The DEFINE_WAKE_Q macro declares and initializes the list head.
985 * wake_up_q() does NOT reinitialize the list; it's expected to be
986 * called near the end of a function. Otherwise, the list can be
987 * re-initialized for later re-use by wake_q_init().
989 * Note that this can cause spurious wakeups. schedule() callers
990 * must ensure the call is done inside a loop, confirming that the
991 * wakeup condition has in fact occurred.
994 struct wake_q_node *next;
998 struct wake_q_node *first;
999 struct wake_q_node **lastp;
1002 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1004 #define DEFINE_WAKE_Q(name) \
1005 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1007 static inline void wake_q_init(struct wake_q_head *head)
1009 head->first = WAKE_Q_TAIL;
1010 head->lastp = &head->first;
1013 extern void wake_q_add(struct wake_q_head *head,
1014 struct task_struct *task);
1015 extern void wake_up_q(struct wake_q_head *head);
1018 * sched-domains (multiprocessor balancing) declarations:
1021 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1022 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1023 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1024 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1025 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1026 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1027 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1028 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1029 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1030 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1031 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1032 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1033 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1034 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1035 #define SD_NUMA 0x4000 /* cross-node balancing */
1037 #ifdef CONFIG_SCHED_SMT
1038 static inline int cpu_smt_flags(void)
1040 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1044 #ifdef CONFIG_SCHED_MC
1045 static inline int cpu_core_flags(void)
1047 return SD_SHARE_PKG_RESOURCES;
1052 static inline int cpu_numa_flags(void)
1058 extern int arch_asym_cpu_priority(int cpu);
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;
1551 /* Per-thread vma caching: */
1552 struct vmacache vmacache;
1554 #if defined(SPLIT_RSS_COUNTING)
1555 struct task_rss_stat rss_stat;
1559 int exit_code, exit_signal;
1560 int pdeath_signal; /* The signal sent when the parent dies */
1561 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1563 /* Used for emulating ABI behavior of previous Linux versions */
1564 unsigned int personality;
1566 /* scheduler bits, serialized by scheduler locks */
1567 unsigned sched_reset_on_fork:1;
1568 unsigned sched_contributes_to_load:1;
1569 unsigned sched_migrated:1;
1570 unsigned sched_remote_wakeup:1;
1571 unsigned :0; /* force alignment to the next boundary */
1573 /* unserialized, strictly 'current' */
1574 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1575 unsigned in_iowait:1;
1576 #if !defined(TIF_RESTORE_SIGMASK)
1577 unsigned restore_sigmask:1;
1580 unsigned memcg_may_oom:1;
1582 unsigned memcg_kmem_skip_account:1;
1585 #ifdef CONFIG_COMPAT_BRK
1586 unsigned brk_randomized:1;
1589 unsigned long atomic_flags; /* Flags needing atomic access. */
1591 struct restart_block restart_block;
1596 #ifdef CONFIG_CC_STACKPROTECTOR
1597 /* Canary value for the -fstack-protector gcc feature */
1598 unsigned long stack_canary;
1601 * pointers to (original) parent process, youngest child, younger sibling,
1602 * older sibling, respectively. (p->father can be replaced with
1603 * p->real_parent->pid)
1605 struct task_struct __rcu *real_parent; /* real parent process */
1606 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1608 * children/sibling forms the list of my natural children
1610 struct list_head children; /* list of my children */
1611 struct list_head sibling; /* linkage in my parent's children list */
1612 struct task_struct *group_leader; /* threadgroup leader */
1615 * ptraced is the list of tasks this task is using ptrace on.
1616 * This includes both natural children and PTRACE_ATTACH targets.
1617 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1619 struct list_head ptraced;
1620 struct list_head ptrace_entry;
1622 /* PID/PID hash table linkage. */
1623 struct pid_link pids[PIDTYPE_MAX];
1624 struct list_head thread_group;
1625 struct list_head thread_node;
1627 struct completion *vfork_done; /* for vfork() */
1628 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1629 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1632 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1633 u64 utimescaled, stimescaled;
1636 struct prev_cputime prev_cputime;
1637 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1638 seqcount_t vtime_seqcount;
1639 unsigned long long vtime_snap;
1641 /* Task is sleeping or running in a CPU with VTIME inactive */
1643 /* Task runs in userspace in a CPU with VTIME active */
1645 /* Task runs in kernelspace in a CPU with VTIME active */
1647 } vtime_snap_whence;
1650 #ifdef CONFIG_NO_HZ_FULL
1651 atomic_t tick_dep_mask;
1653 unsigned long nvcsw, nivcsw; /* context switch counts */
1654 u64 start_time; /* monotonic time in nsec */
1655 u64 real_start_time; /* boot based time in nsec */
1656 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1657 unsigned long min_flt, maj_flt;
1659 #ifdef CONFIG_POSIX_TIMERS
1660 struct task_cputime cputime_expires;
1661 struct list_head cpu_timers[3];
1664 /* process credentials */
1665 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1666 const struct cred __rcu *real_cred; /* objective and real subjective task
1667 * credentials (COW) */
1668 const struct cred __rcu *cred; /* effective (overridable) subjective task
1669 * credentials (COW) */
1670 char comm[TASK_COMM_LEN]; /* executable name excluding path
1671 - access with [gs]et_task_comm (which lock
1672 it with task_lock())
1673 - initialized normally by setup_new_exec */
1674 /* file system info */
1675 struct nameidata *nameidata;
1676 #ifdef CONFIG_SYSVIPC
1678 struct sysv_sem sysvsem;
1679 struct sysv_shm sysvshm;
1681 #ifdef CONFIG_DETECT_HUNG_TASK
1682 /* hung task detection */
1683 unsigned long last_switch_count;
1685 /* filesystem information */
1686 struct fs_struct *fs;
1687 /* open file information */
1688 struct files_struct *files;
1690 struct nsproxy *nsproxy;
1691 /* signal handlers */
1692 struct signal_struct *signal;
1693 struct sighand_struct *sighand;
1695 sigset_t blocked, real_blocked;
1696 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1697 struct sigpending pending;
1699 unsigned long sas_ss_sp;
1701 unsigned sas_ss_flags;
1703 struct callback_head *task_works;
1705 struct audit_context *audit_context;
1706 #ifdef CONFIG_AUDITSYSCALL
1708 unsigned int sessionid;
1710 struct seccomp seccomp;
1712 /* Thread group tracking */
1715 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1717 spinlock_t alloc_lock;
1719 /* Protection of the PI data structures: */
1720 raw_spinlock_t pi_lock;
1722 struct wake_q_node wake_q;
1724 #ifdef CONFIG_RT_MUTEXES
1725 /* PI waiters blocked on a rt_mutex held by this task */
1726 struct rb_root pi_waiters;
1727 struct rb_node *pi_waiters_leftmost;
1728 /* Deadlock detection and priority inheritance handling */
1729 struct rt_mutex_waiter *pi_blocked_on;
1732 #ifdef CONFIG_DEBUG_MUTEXES
1733 /* mutex deadlock detection */
1734 struct mutex_waiter *blocked_on;
1736 #ifdef CONFIG_TRACE_IRQFLAGS
1737 unsigned int irq_events;
1738 unsigned long hardirq_enable_ip;
1739 unsigned long hardirq_disable_ip;
1740 unsigned int hardirq_enable_event;
1741 unsigned int hardirq_disable_event;
1742 int hardirqs_enabled;
1743 int hardirq_context;
1744 unsigned long softirq_disable_ip;
1745 unsigned long softirq_enable_ip;
1746 unsigned int softirq_disable_event;
1747 unsigned int softirq_enable_event;
1748 int softirqs_enabled;
1749 int softirq_context;
1751 #ifdef CONFIG_LOCKDEP
1752 # define MAX_LOCK_DEPTH 48UL
1755 unsigned int lockdep_recursion;
1756 struct held_lock held_locks[MAX_LOCK_DEPTH];
1757 gfp_t lockdep_reclaim_gfp;
1760 unsigned int in_ubsan;
1763 /* journalling filesystem info */
1766 /* stacked block device info */
1767 struct bio_list *bio_list;
1770 /* stack plugging */
1771 struct blk_plug *plug;
1775 struct reclaim_state *reclaim_state;
1777 struct backing_dev_info *backing_dev_info;
1779 struct io_context *io_context;
1781 unsigned long ptrace_message;
1782 siginfo_t *last_siginfo; /* For ptrace use. */
1783 struct task_io_accounting ioac;
1784 #if defined(CONFIG_TASK_XACCT)
1785 u64 acct_rss_mem1; /* accumulated rss usage */
1786 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1787 u64 acct_timexpd; /* stime + utime since last update */
1789 #ifdef CONFIG_CPUSETS
1790 nodemask_t mems_allowed; /* Protected by alloc_lock */
1791 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1792 int cpuset_mem_spread_rotor;
1793 int cpuset_slab_spread_rotor;
1795 #ifdef CONFIG_CGROUPS
1796 /* Control Group info protected by css_set_lock */
1797 struct css_set __rcu *cgroups;
1798 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1799 struct list_head cg_list;
1801 #ifdef CONFIG_INTEL_RDT_A
1805 struct robust_list_head __user *robust_list;
1806 #ifdef CONFIG_COMPAT
1807 struct compat_robust_list_head __user *compat_robust_list;
1809 struct list_head pi_state_list;
1810 struct futex_pi_state *pi_state_cache;
1812 #ifdef CONFIG_PERF_EVENTS
1813 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1814 struct mutex perf_event_mutex;
1815 struct list_head perf_event_list;
1817 #ifdef CONFIG_DEBUG_PREEMPT
1818 unsigned long preempt_disable_ip;
1821 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1823 short pref_node_fork;
1825 #ifdef CONFIG_NUMA_BALANCING
1827 unsigned int numa_scan_period;
1828 unsigned int numa_scan_period_max;
1829 int numa_preferred_nid;
1830 unsigned long numa_migrate_retry;
1831 u64 node_stamp; /* migration stamp */
1832 u64 last_task_numa_placement;
1833 u64 last_sum_exec_runtime;
1834 struct callback_head numa_work;
1836 struct list_head numa_entry;
1837 struct numa_group *numa_group;
1840 * numa_faults is an array split into four regions:
1841 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1842 * in this precise order.
1844 * faults_memory: Exponential decaying average of faults on a per-node
1845 * basis. Scheduling placement decisions are made based on these
1846 * counts. The values remain static for the duration of a PTE scan.
1847 * faults_cpu: Track the nodes the process was running on when a NUMA
1848 * hinting fault was incurred.
1849 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1850 * during the current scan window. When the scan completes, the counts
1851 * in faults_memory and faults_cpu decay and these values are copied.
1853 unsigned long *numa_faults;
1854 unsigned long total_numa_faults;
1857 * numa_faults_locality tracks if faults recorded during the last
1858 * scan window were remote/local or failed to migrate. The task scan
1859 * period is adapted based on the locality of the faults with different
1860 * weights depending on whether they were shared or private faults
1862 unsigned long numa_faults_locality[3];
1864 unsigned long numa_pages_migrated;
1865 #endif /* CONFIG_NUMA_BALANCING */
1867 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1868 struct tlbflush_unmap_batch tlb_ubc;
1871 struct rcu_head rcu;
1874 * cache last used pipe for splice
1876 struct pipe_inode_info *splice_pipe;
1878 struct page_frag task_frag;
1880 #ifdef CONFIG_TASK_DELAY_ACCT
1881 struct task_delay_info *delays;
1883 #ifdef CONFIG_FAULT_INJECTION
1887 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1888 * balance_dirty_pages() for some dirty throttling pause
1891 int nr_dirtied_pause;
1892 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1894 #ifdef CONFIG_LATENCYTOP
1895 int latency_record_count;
1896 struct latency_record latency_record[LT_SAVECOUNT];
1899 * time slack values; these are used to round up poll() and
1900 * select() etc timeout values. These are in nanoseconds.
1903 u64 default_timer_slack_ns;
1906 unsigned int kasan_depth;
1908 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1909 /* Index of current stored address in ret_stack */
1911 /* Stack of return addresses for return function tracing */
1912 struct ftrace_ret_stack *ret_stack;
1913 /* time stamp for last schedule */
1914 unsigned long long ftrace_timestamp;
1916 * Number of functions that haven't been traced
1917 * because of depth overrun.
1919 atomic_t trace_overrun;
1920 /* Pause for the tracing */
1921 atomic_t tracing_graph_pause;
1923 #ifdef CONFIG_TRACING
1924 /* state flags for use by tracers */
1925 unsigned long trace;
1926 /* bitmask and counter of trace recursion */
1927 unsigned long trace_recursion;
1928 #endif /* CONFIG_TRACING */
1930 /* Coverage collection mode enabled for this task (0 if disabled). */
1931 enum kcov_mode kcov_mode;
1932 /* Size of the kcov_area. */
1934 /* Buffer for coverage collection. */
1936 /* kcov desciptor wired with this task or NULL. */
1940 struct mem_cgroup *memcg_in_oom;
1941 gfp_t memcg_oom_gfp_mask;
1942 int memcg_oom_order;
1944 /* number of pages to reclaim on returning to userland */
1945 unsigned int memcg_nr_pages_over_high;
1947 #ifdef CONFIG_UPROBES
1948 struct uprobe_task *utask;
1950 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1951 unsigned int sequential_io;
1952 unsigned int sequential_io_avg;
1954 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1955 unsigned long task_state_change;
1957 int pagefault_disabled;
1959 struct task_struct *oom_reaper_list;
1961 #ifdef CONFIG_VMAP_STACK
1962 struct vm_struct *stack_vm_area;
1964 #ifdef CONFIG_THREAD_INFO_IN_TASK
1965 /* A live task holds one reference. */
1966 atomic_t stack_refcount;
1968 /* CPU-specific state of this task */
1969 struct thread_struct thread;
1971 * WARNING: on x86, 'thread_struct' contains a variable-sized
1972 * structure. It *MUST* be at the end of 'task_struct'.
1974 * Do not put anything below here!
1978 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1979 extern int arch_task_struct_size __read_mostly;
1981 # define arch_task_struct_size (sizeof(struct task_struct))
1984 #ifdef CONFIG_VMAP_STACK
1985 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1987 return t->stack_vm_area;
1990 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1996 #define TNF_MIGRATED 0x01
1997 #define TNF_NO_GROUP 0x02
1998 #define TNF_SHARED 0x04
1999 #define TNF_FAULT_LOCAL 0x08
2000 #define TNF_MIGRATE_FAIL 0x10
2002 static inline bool in_vfork(struct task_struct *tsk)
2007 * need RCU to access ->real_parent if CLONE_VM was used along with
2010 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2013 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2014 * ->real_parent is not necessarily the task doing vfork(), so in
2015 * theory we can't rely on task_lock() if we want to dereference it.
2017 * And in this case we can't trust the real_parent->mm == tsk->mm
2018 * check, it can be false negative. But we do not care, if init or
2019 * another oom-unkillable task does this it should blame itself.
2022 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2028 #ifdef CONFIG_NUMA_BALANCING
2029 extern void task_numa_fault(int last_node, int node, int pages, int flags);
2030 extern pid_t task_numa_group_id(struct task_struct *p);
2031 extern void set_numabalancing_state(bool enabled);
2032 extern void task_numa_free(struct task_struct *p);
2033 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2034 int src_nid, int dst_cpu);
2036 static inline void task_numa_fault(int last_node, int node, int pages,
2040 static inline pid_t task_numa_group_id(struct task_struct *p)
2044 static inline void set_numabalancing_state(bool enabled)
2047 static inline void task_numa_free(struct task_struct *p)
2050 static inline bool should_numa_migrate_memory(struct task_struct *p,
2051 struct page *page, int src_nid, int dst_cpu)
2057 static inline struct pid *task_pid(struct task_struct *task)
2059 return task->pids[PIDTYPE_PID].pid;
2062 static inline struct pid *task_tgid(struct task_struct *task)
2064 return task->group_leader->pids[PIDTYPE_PID].pid;
2068 * Without tasklist or rcu lock it is not safe to dereference
2069 * the result of task_pgrp/task_session even if task == current,
2070 * we can race with another thread doing sys_setsid/sys_setpgid.
2072 static inline struct pid *task_pgrp(struct task_struct *task)
2074 return task->group_leader->pids[PIDTYPE_PGID].pid;
2077 static inline struct pid *task_session(struct task_struct *task)
2079 return task->group_leader->pids[PIDTYPE_SID].pid;
2082 struct pid_namespace;
2085 * the helpers to get the task's different pids as they are seen
2086 * from various namespaces
2088 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2089 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2091 * task_xid_nr_ns() : id seen from the ns specified;
2093 * set_task_vxid() : assigns a virtual id to a task;
2095 * see also pid_nr() etc in include/linux/pid.h
2097 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2098 struct pid_namespace *ns);
2100 static inline pid_t task_pid_nr(struct task_struct *tsk)
2105 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2106 struct pid_namespace *ns)
2108 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2111 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2113 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2117 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2122 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2124 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2126 return pid_vnr(task_tgid(tsk));
2130 static inline int pid_alive(const struct task_struct *p);
2131 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2137 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2143 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2145 return task_ppid_nr_ns(tsk, &init_pid_ns);
2148 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2149 struct pid_namespace *ns)
2151 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2154 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2156 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2160 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2161 struct pid_namespace *ns)
2163 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2166 static inline pid_t task_session_vnr(struct task_struct *tsk)
2168 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2171 /* obsolete, do not use */
2172 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2174 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2178 * pid_alive - check that a task structure is not stale
2179 * @p: Task structure to be checked.
2181 * Test if a process is not yet dead (at most zombie state)
2182 * If pid_alive fails, then pointers within the task structure
2183 * can be stale and must not be dereferenced.
2185 * Return: 1 if the process is alive. 0 otherwise.
2187 static inline int pid_alive(const struct task_struct *p)
2189 return p->pids[PIDTYPE_PID].pid != NULL;
2193 * is_global_init - check if a task structure is init. Since init
2194 * is free to have sub-threads we need to check tgid.
2195 * @tsk: Task structure to be checked.
2197 * Check if a task structure is the first user space task the kernel created.
2199 * Return: 1 if the task structure is init. 0 otherwise.
2201 static inline int is_global_init(struct task_struct *tsk)
2203 return task_tgid_nr(tsk) == 1;
2206 extern struct pid *cad_pid;
2208 extern void free_task(struct task_struct *tsk);
2209 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2211 extern void __put_task_struct(struct task_struct *t);
2213 static inline void put_task_struct(struct task_struct *t)
2215 if (atomic_dec_and_test(&t->usage))
2216 __put_task_struct(t);
2219 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2220 struct task_struct *try_get_task_struct(struct task_struct **ptask);
2222 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2223 extern void task_cputime(struct task_struct *t,
2224 u64 *utime, u64 *stime);
2225 extern u64 task_gtime(struct task_struct *t);
2227 static inline void task_cputime(struct task_struct *t,
2228 u64 *utime, u64 *stime)
2234 static inline u64 task_gtime(struct task_struct *t)
2240 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
2241 static inline void task_cputime_scaled(struct task_struct *t,
2245 *utimescaled = t->utimescaled;
2246 *stimescaled = t->stimescaled;
2249 static inline void task_cputime_scaled(struct task_struct *t,
2253 task_cputime(t, utimescaled, stimescaled);
2257 extern void task_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2258 extern void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2263 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
2264 #define PF_EXITING 0x00000004 /* getting shut down */
2265 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2266 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2267 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2268 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2269 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2270 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2271 #define PF_DUMPCORE 0x00000200 /* dumped core */
2272 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2273 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2274 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2275 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2276 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2277 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2278 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2279 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2280 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2281 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2282 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2283 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2284 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2285 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2286 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2287 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2288 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2289 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2290 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2293 * Only the _current_ task can read/write to tsk->flags, but other
2294 * tasks can access tsk->flags in readonly mode for example
2295 * with tsk_used_math (like during threaded core dumping).
2296 * There is however an exception to this rule during ptrace
2297 * or during fork: the ptracer task is allowed to write to the
2298 * child->flags of its traced child (same goes for fork, the parent
2299 * can write to the child->flags), because we're guaranteed the
2300 * child is not running and in turn not changing child->flags
2301 * at the same time the parent does it.
2303 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2304 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2305 #define clear_used_math() clear_stopped_child_used_math(current)
2306 #define set_used_math() set_stopped_child_used_math(current)
2307 #define conditional_stopped_child_used_math(condition, child) \
2308 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2309 #define conditional_used_math(condition) \
2310 conditional_stopped_child_used_math(condition, current)
2311 #define copy_to_stopped_child_used_math(child) \
2312 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2313 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2314 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2315 #define used_math() tsk_used_math(current)
2317 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2318 * __GFP_FS is also cleared as it implies __GFP_IO.
2320 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2322 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2323 flags &= ~(__GFP_IO | __GFP_FS);
2327 static inline unsigned int memalloc_noio_save(void)
2329 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2330 current->flags |= PF_MEMALLOC_NOIO;
2334 static inline void memalloc_noio_restore(unsigned int flags)
2336 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2339 /* Per-process atomic flags. */
2340 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2341 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2342 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2343 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2346 #define TASK_PFA_TEST(name, func) \
2347 static inline bool task_##func(struct task_struct *p) \
2348 { return test_bit(PFA_##name, &p->atomic_flags); }
2349 #define TASK_PFA_SET(name, func) \
2350 static inline void task_set_##func(struct task_struct *p) \
2351 { set_bit(PFA_##name, &p->atomic_flags); }
2352 #define TASK_PFA_CLEAR(name, func) \
2353 static inline void task_clear_##func(struct task_struct *p) \
2354 { clear_bit(PFA_##name, &p->atomic_flags); }
2356 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2357 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2359 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2360 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2361 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2363 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2364 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2365 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2367 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2368 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2371 * task->jobctl flags
2373 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2375 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2376 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2377 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2378 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2379 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2380 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2381 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2383 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2384 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2385 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2386 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2387 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2388 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2389 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2391 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2392 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2394 extern bool task_set_jobctl_pending(struct task_struct *task,
2395 unsigned long mask);
2396 extern void task_clear_jobctl_trapping(struct task_struct *task);
2397 extern void task_clear_jobctl_pending(struct task_struct *task,
2398 unsigned long mask);
2400 static inline void rcu_copy_process(struct task_struct *p)
2402 #ifdef CONFIG_PREEMPT_RCU
2403 p->rcu_read_lock_nesting = 0;
2404 p->rcu_read_unlock_special.s = 0;
2405 p->rcu_blocked_node = NULL;
2406 INIT_LIST_HEAD(&p->rcu_node_entry);
2407 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2408 #ifdef CONFIG_TASKS_RCU
2409 p->rcu_tasks_holdout = false;
2410 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2411 p->rcu_tasks_idle_cpu = -1;
2412 #endif /* #ifdef CONFIG_TASKS_RCU */
2415 static inline void tsk_restore_flags(struct task_struct *task,
2416 unsigned long orig_flags, unsigned long flags)
2418 task->flags &= ~flags;
2419 task->flags |= orig_flags & flags;
2422 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2423 const struct cpumask *trial);
2424 extern int task_can_attach(struct task_struct *p,
2425 const struct cpumask *cs_cpus_allowed);
2427 extern void do_set_cpus_allowed(struct task_struct *p,
2428 const struct cpumask *new_mask);
2430 extern int set_cpus_allowed_ptr(struct task_struct *p,
2431 const struct cpumask *new_mask);
2433 static inline void do_set_cpus_allowed(struct task_struct *p,
2434 const struct cpumask *new_mask)
2437 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2438 const struct cpumask *new_mask)
2440 if (!cpumask_test_cpu(0, new_mask))
2446 #ifdef CONFIG_NO_HZ_COMMON
2447 void calc_load_enter_idle(void);
2448 void calc_load_exit_idle(void);
2450 static inline void calc_load_enter_idle(void) { }
2451 static inline void calc_load_exit_idle(void) { }
2452 #endif /* CONFIG_NO_HZ_COMMON */
2454 #ifndef cpu_relax_yield
2455 #define cpu_relax_yield() cpu_relax()
2459 * Do not use outside of architecture code which knows its limitations.
2461 * sched_clock() has no promise of monotonicity or bounded drift between
2462 * CPUs, use (which you should not) requires disabling IRQs.
2464 * Please use one of the three interfaces below.
2466 extern unsigned long long notrace sched_clock(void);
2468 * See the comment in kernel/sched/clock.c
2470 extern u64 running_clock(void);
2471 extern u64 sched_clock_cpu(int cpu);
2474 extern void sched_clock_init(void);
2476 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2477 static inline void sched_clock_init_late(void)
2481 static inline void sched_clock_tick(void)
2485 static inline void clear_sched_clock_stable(void)
2489 static inline void sched_clock_idle_sleep_event(void)
2493 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2497 static inline u64 cpu_clock(int cpu)
2499 return sched_clock();
2502 static inline u64 local_clock(void)
2504 return sched_clock();
2507 extern void sched_clock_init_late(void);
2509 * Architectures can set this to 1 if they have specified
2510 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2511 * but then during bootup it turns out that sched_clock()
2512 * is reliable after all:
2514 extern int sched_clock_stable(void);
2515 extern void clear_sched_clock_stable(void);
2517 extern void sched_clock_tick(void);
2518 extern void sched_clock_idle_sleep_event(void);
2519 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2522 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2523 * time source that is monotonic per cpu argument and has bounded drift
2526 * ######################### BIG FAT WARNING ##########################
2527 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2528 * # go backwards !! #
2529 * ####################################################################
2531 static inline u64 cpu_clock(int cpu)
2533 return sched_clock_cpu(cpu);
2536 static inline u64 local_clock(void)
2538 return sched_clock_cpu(raw_smp_processor_id());
2542 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2544 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2545 * The reason for this explicit opt-in is not to have perf penalty with
2546 * slow sched_clocks.
2548 extern void enable_sched_clock_irqtime(void);
2549 extern void disable_sched_clock_irqtime(void);
2551 static inline void enable_sched_clock_irqtime(void) {}
2552 static inline void disable_sched_clock_irqtime(void) {}
2555 extern unsigned long long
2556 task_sched_runtime(struct task_struct *task);
2558 /* sched_exec is called by processes performing an exec */
2560 extern void sched_exec(void);
2562 #define sched_exec() {}
2565 extern void sched_clock_idle_sleep_event(void);
2566 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2568 #ifdef CONFIG_HOTPLUG_CPU
2569 extern void idle_task_exit(void);
2571 static inline void idle_task_exit(void) {}
2574 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2575 extern void wake_up_nohz_cpu(int cpu);
2577 static inline void wake_up_nohz_cpu(int cpu) { }
2580 #ifdef CONFIG_NO_HZ_FULL
2581 extern u64 scheduler_tick_max_deferment(void);
2584 #ifdef CONFIG_SCHED_AUTOGROUP
2585 extern void sched_autogroup_create_attach(struct task_struct *p);
2586 extern void sched_autogroup_detach(struct task_struct *p);
2587 extern void sched_autogroup_fork(struct signal_struct *sig);
2588 extern void sched_autogroup_exit(struct signal_struct *sig);
2589 extern void sched_autogroup_exit_task(struct task_struct *p);
2590 #ifdef CONFIG_PROC_FS
2591 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2592 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2595 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2596 static inline void sched_autogroup_detach(struct task_struct *p) { }
2597 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2598 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2599 static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2602 extern int yield_to(struct task_struct *p, bool preempt);
2603 extern void set_user_nice(struct task_struct *p, long nice);
2604 extern int task_prio(const struct task_struct *p);
2606 * task_nice - return the nice value of a given task.
2607 * @p: the task in question.
2609 * Return: The nice value [ -20 ... 0 ... 19 ].
2611 static inline int task_nice(const struct task_struct *p)
2613 return PRIO_TO_NICE((p)->static_prio);
2615 extern int can_nice(const struct task_struct *p, const int nice);
2616 extern int task_curr(const struct task_struct *p);
2617 extern int idle_cpu(int cpu);
2618 extern int sched_setscheduler(struct task_struct *, int,
2619 const struct sched_param *);
2620 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2621 const struct sched_param *);
2622 extern int sched_setattr(struct task_struct *,
2623 const struct sched_attr *);
2624 extern struct task_struct *idle_task(int cpu);
2626 * is_idle_task - is the specified task an idle task?
2627 * @p: the task in question.
2629 * Return: 1 if @p is an idle task. 0 otherwise.
2631 static inline bool is_idle_task(const struct task_struct *p)
2633 return !!(p->flags & PF_IDLE);
2635 extern struct task_struct *curr_task(int cpu);
2636 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2640 union thread_union {
2641 #ifndef CONFIG_THREAD_INFO_IN_TASK
2642 struct thread_info thread_info;
2644 unsigned long stack[THREAD_SIZE/sizeof(long)];
2647 #ifndef __HAVE_ARCH_KSTACK_END
2648 static inline int kstack_end(void *addr)
2650 /* Reliable end of stack detection:
2651 * Some APM bios versions misalign the stack
2653 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2657 extern union thread_union init_thread_union;
2658 extern struct task_struct init_task;
2660 extern struct mm_struct init_mm;
2662 extern struct pid_namespace init_pid_ns;
2665 * find a task by one of its numerical ids
2667 * find_task_by_pid_ns():
2668 * finds a task by its pid in the specified namespace
2669 * find_task_by_vpid():
2670 * finds a task by its virtual pid
2672 * see also find_vpid() etc in include/linux/pid.h
2675 extern struct task_struct *find_task_by_vpid(pid_t nr);
2676 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2677 struct pid_namespace *ns);
2679 /* per-UID process charging. */
2680 extern struct user_struct * alloc_uid(kuid_t);
2681 static inline struct user_struct *get_uid(struct user_struct *u)
2683 atomic_inc(&u->__count);
2686 extern void free_uid(struct user_struct *);
2688 #include <asm/current.h>
2690 extern void xtime_update(unsigned long ticks);
2692 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2693 extern int wake_up_process(struct task_struct *tsk);
2694 extern void wake_up_new_task(struct task_struct *tsk);
2696 extern void kick_process(struct task_struct *tsk);
2698 static inline void kick_process(struct task_struct *tsk) { }
2700 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2701 extern void sched_dead(struct task_struct *p);
2703 extern void proc_caches_init(void);
2704 extern void flush_signals(struct task_struct *);
2705 extern void ignore_signals(struct task_struct *);
2706 extern void flush_signal_handlers(struct task_struct *, int force_default);
2707 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2709 static inline int kernel_dequeue_signal(siginfo_t *info)
2711 struct task_struct *tsk = current;
2715 spin_lock_irq(&tsk->sighand->siglock);
2716 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2717 spin_unlock_irq(&tsk->sighand->siglock);
2722 static inline void kernel_signal_stop(void)
2724 spin_lock_irq(¤t->sighand->siglock);
2725 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2726 __set_current_state(TASK_STOPPED);
2727 spin_unlock_irq(¤t->sighand->siglock);
2732 extern void release_task(struct task_struct * p);
2733 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2734 extern int force_sigsegv(int, struct task_struct *);
2735 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2736 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2737 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2738 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2739 const struct cred *, u32);
2740 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2741 extern int kill_pid(struct pid *pid, int sig, int priv);
2742 extern int kill_proc_info(int, struct siginfo *, pid_t);
2743 extern __must_check bool do_notify_parent(struct task_struct *, int);
2744 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2745 extern void force_sig(int, struct task_struct *);
2746 extern int send_sig(int, struct task_struct *, int);
2747 extern int zap_other_threads(struct task_struct *p);
2748 extern struct sigqueue *sigqueue_alloc(void);
2749 extern void sigqueue_free(struct sigqueue *);
2750 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2751 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2753 #ifdef TIF_RESTORE_SIGMASK
2755 * Legacy restore_sigmask accessors. These are inefficient on
2756 * SMP architectures because they require atomic operations.
2760 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2762 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2763 * will run before returning to user mode, to process the flag. For
2764 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2765 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2766 * arch code will notice on return to user mode, in case those bits
2767 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2768 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2770 static inline void set_restore_sigmask(void)
2772 set_thread_flag(TIF_RESTORE_SIGMASK);
2773 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2775 static inline void clear_restore_sigmask(void)
2777 clear_thread_flag(TIF_RESTORE_SIGMASK);
2779 static inline bool test_restore_sigmask(void)
2781 return test_thread_flag(TIF_RESTORE_SIGMASK);
2783 static inline bool test_and_clear_restore_sigmask(void)
2785 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2788 #else /* TIF_RESTORE_SIGMASK */
2790 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2791 static inline void set_restore_sigmask(void)
2793 current->restore_sigmask = true;
2794 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2796 static inline void clear_restore_sigmask(void)
2798 current->restore_sigmask = false;
2800 static inline bool test_restore_sigmask(void)
2802 return current->restore_sigmask;
2804 static inline bool test_and_clear_restore_sigmask(void)
2806 if (!current->restore_sigmask)
2808 current->restore_sigmask = false;
2813 static inline void restore_saved_sigmask(void)
2815 if (test_and_clear_restore_sigmask())
2816 __set_current_blocked(¤t->saved_sigmask);
2819 static inline sigset_t *sigmask_to_save(void)
2821 sigset_t *res = ¤t->blocked;
2822 if (unlikely(test_restore_sigmask()))
2823 res = ¤t->saved_sigmask;
2827 static inline int kill_cad_pid(int sig, int priv)
2829 return kill_pid(cad_pid, sig, priv);
2832 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2833 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2834 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2835 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2838 * True if we are on the alternate signal stack.
2840 static inline int on_sig_stack(unsigned long sp)
2843 * If the signal stack is SS_AUTODISARM then, by construction, we
2844 * can't be on the signal stack unless user code deliberately set
2845 * SS_AUTODISARM when we were already on it.
2847 * This improves reliability: if user state gets corrupted such that
2848 * the stack pointer points very close to the end of the signal stack,
2849 * then this check will enable the signal to be handled anyway.
2851 if (current->sas_ss_flags & SS_AUTODISARM)
2854 #ifdef CONFIG_STACK_GROWSUP
2855 return sp >= current->sas_ss_sp &&
2856 sp - current->sas_ss_sp < current->sas_ss_size;
2858 return sp > current->sas_ss_sp &&
2859 sp - current->sas_ss_sp <= current->sas_ss_size;
2863 static inline int sas_ss_flags(unsigned long sp)
2865 if (!current->sas_ss_size)
2868 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2871 static inline void sas_ss_reset(struct task_struct *p)
2875 p->sas_ss_flags = SS_DISABLE;
2878 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2880 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2881 #ifdef CONFIG_STACK_GROWSUP
2882 return current->sas_ss_sp;
2884 return current->sas_ss_sp + current->sas_ss_size;
2890 * Routines for handling mm_structs
2892 extern struct mm_struct * mm_alloc(void);
2895 * mmgrab() - Pin a &struct mm_struct.
2896 * @mm: The &struct mm_struct to pin.
2898 * Make sure that @mm will not get freed even after the owning task
2899 * exits. This doesn't guarantee that the associated address space
2900 * will still exist later on and mmget_not_zero() has to be used before
2903 * This is a preferred way to to pin @mm for a longer/unbounded amount
2906 * Use mmdrop() to release the reference acquired by mmgrab().
2908 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2909 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2911 static inline void mmgrab(struct mm_struct *mm)
2913 atomic_inc(&mm->mm_count);
2916 /* mmdrop drops the mm and the page tables */
2917 extern void __mmdrop(struct mm_struct *);
2918 static inline void mmdrop(struct mm_struct *mm)
2920 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2924 static inline void mmdrop_async_fn(struct work_struct *work)
2926 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2930 static inline void mmdrop_async(struct mm_struct *mm)
2932 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2933 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2934 schedule_work(&mm->async_put_work);
2939 * mmget() - Pin the address space associated with a &struct mm_struct.
2940 * @mm: The address space to pin.
2942 * Make sure that the address space of the given &struct mm_struct doesn't
2943 * go away. This does not protect against parts of the address space being
2944 * modified or freed, however.
2946 * Never use this function to pin this address space for an
2947 * unbounded/indefinite amount of time.
2949 * Use mmput() to release the reference acquired by mmget().
2951 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2952 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2954 static inline void mmget(struct mm_struct *mm)
2956 atomic_inc(&mm->mm_users);
2959 static inline bool mmget_not_zero(struct mm_struct *mm)
2961 return atomic_inc_not_zero(&mm->mm_users);
2964 /* mmput gets rid of the mappings and all user-space */
2965 extern void mmput(struct mm_struct *);
2967 /* same as above but performs the slow path from the async context. Can
2968 * be called from the atomic context as well
2970 extern void mmput_async(struct mm_struct *);
2973 /* Grab a reference to a task's mm, if it is not already going away */
2974 extern struct mm_struct *get_task_mm(struct task_struct *task);
2976 * Grab a reference to a task's mm, if it is not already going away
2977 * and ptrace_may_access with the mode parameter passed to it
2980 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2981 /* Remove the current tasks stale references to the old mm_struct */
2982 extern void mm_release(struct task_struct *, struct mm_struct *);
2984 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2985 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2986 struct task_struct *, unsigned long);
2988 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2989 struct task_struct *);
2991 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2992 * via pt_regs, so ignore the tls argument passed via C. */
2993 static inline int copy_thread_tls(
2994 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2995 struct task_struct *p, unsigned long tls)
2997 return copy_thread(clone_flags, sp, arg, p);
3000 extern void flush_thread(void);
3002 #ifdef CONFIG_HAVE_EXIT_THREAD
3003 extern void exit_thread(struct task_struct *tsk);
3005 static inline void exit_thread(struct task_struct *tsk)
3010 extern void exit_files(struct task_struct *);
3011 extern void __cleanup_sighand(struct sighand_struct *);
3013 extern void exit_itimers(struct signal_struct *);
3014 extern void flush_itimer_signals(void);
3016 extern void do_group_exit(int);
3018 extern int do_execve(struct filename *,
3019 const char __user * const __user *,
3020 const char __user * const __user *);
3021 extern int do_execveat(int, struct filename *,
3022 const char __user * const __user *,
3023 const char __user * const __user *,
3025 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
3026 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
3027 struct task_struct *fork_idle(int);
3028 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
3030 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
3031 static inline void set_task_comm(struct task_struct *tsk, const char *from)
3033 __set_task_comm(tsk, from, false);
3035 extern char *get_task_comm(char *to, struct task_struct *tsk);
3038 void scheduler_ipi(void);
3039 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
3041 static inline void scheduler_ipi(void) { }
3042 static inline unsigned long wait_task_inactive(struct task_struct *p,
3049 #define tasklist_empty() \
3050 list_empty(&init_task.tasks)
3052 #define next_task(p) \
3053 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
3055 #define for_each_process(p) \
3056 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
3058 extern bool current_is_single_threaded(void);
3061 * Careful: do_each_thread/while_each_thread is a double loop so
3062 * 'break' will not work as expected - use goto instead.
3064 #define do_each_thread(g, t) \
3065 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3067 #define while_each_thread(g, t) \
3068 while ((t = next_thread(t)) != g)
3070 #define __for_each_thread(signal, t) \
3071 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3073 #define for_each_thread(p, t) \
3074 __for_each_thread((p)->signal, t)
3076 /* Careful: this is a double loop, 'break' won't work as expected. */
3077 #define for_each_process_thread(p, t) \
3078 for_each_process(p) for_each_thread(p, t)
3080 typedef int (*proc_visitor)(struct task_struct *p, void *data);
3081 void walk_process_tree(struct task_struct *top, proc_visitor, void *);
3083 static inline int get_nr_threads(struct task_struct *tsk)
3085 return tsk->signal->nr_threads;
3088 static inline bool thread_group_leader(struct task_struct *p)
3090 return p->exit_signal >= 0;
3093 /* Do to the insanities of de_thread it is possible for a process
3094 * to have the pid of the thread group leader without actually being
3095 * the thread group leader. For iteration through the pids in proc
3096 * all we care about is that we have a task with the appropriate
3097 * pid, we don't actually care if we have the right task.
3099 static inline bool has_group_leader_pid(struct task_struct *p)
3101 return task_pid(p) == p->signal->leader_pid;
3105 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3107 return p1->signal == p2->signal;
3110 static inline struct task_struct *next_thread(const struct task_struct *p)
3112 return list_entry_rcu(p->thread_group.next,
3113 struct task_struct, thread_group);
3116 static inline int thread_group_empty(struct task_struct *p)
3118 return list_empty(&p->thread_group);
3121 #define delay_group_leader(p) \
3122 (thread_group_leader(p) && !thread_group_empty(p))
3125 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3126 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3127 * pins the final release of task.io_context. Also protects ->cpuset and
3128 * ->cgroup.subsys[]. And ->vfork_done.
3130 * Nests both inside and outside of read_lock(&tasklist_lock).
3131 * It must not be nested with write_lock_irq(&tasklist_lock),
3132 * neither inside nor outside.
3134 static inline void task_lock(struct task_struct *p)
3136 spin_lock(&p->alloc_lock);
3139 static inline void task_unlock(struct task_struct *p)
3141 spin_unlock(&p->alloc_lock);
3144 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3145 unsigned long *flags);
3147 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3148 unsigned long *flags)
3150 struct sighand_struct *ret;
3152 ret = __lock_task_sighand(tsk, flags);
3153 (void)__cond_lock(&tsk->sighand->siglock, ret);
3157 static inline void unlock_task_sighand(struct task_struct *tsk,
3158 unsigned long *flags)
3160 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3163 #ifdef CONFIG_THREAD_INFO_IN_TASK
3165 static inline struct thread_info *task_thread_info(struct task_struct *task)
3167 return &task->thread_info;
3171 * When accessing the stack of a non-current task that might exit, use
3172 * try_get_task_stack() instead. task_stack_page will return a pointer
3173 * that could get freed out from under you.
3175 static inline void *task_stack_page(const struct task_struct *task)
3180 #define setup_thread_stack(new,old) do { } while(0)
3182 static inline unsigned long *end_of_stack(const struct task_struct *task)
3187 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3189 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3190 #define task_stack_page(task) ((void *)(task)->stack)
3192 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3194 *task_thread_info(p) = *task_thread_info(org);
3195 task_thread_info(p)->task = p;
3199 * Return the address of the last usable long on the stack.
3201 * When the stack grows down, this is just above the thread
3202 * info struct. Going any lower will corrupt the threadinfo.
3204 * When the stack grows up, this is the highest address.
3205 * Beyond that position, we corrupt data on the next page.
3207 static inline unsigned long *end_of_stack(struct task_struct *p)
3209 #ifdef CONFIG_STACK_GROWSUP
3210 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3212 return (unsigned long *)(task_thread_info(p) + 1);
3218 #ifdef CONFIG_THREAD_INFO_IN_TASK
3219 static inline void *try_get_task_stack(struct task_struct *tsk)
3221 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3222 task_stack_page(tsk) : NULL;
3225 extern void put_task_stack(struct task_struct *tsk);
3227 static inline void *try_get_task_stack(struct task_struct *tsk)
3229 return task_stack_page(tsk);
3232 static inline void put_task_stack(struct task_struct *tsk) {}
3235 #define task_stack_end_corrupted(task) \
3236 (*(end_of_stack(task)) != STACK_END_MAGIC)
3238 static inline int object_is_on_stack(void *obj)
3240 void *stack = task_stack_page(current);
3242 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3245 extern void thread_stack_cache_init(void);
3247 #ifdef CONFIG_DEBUG_STACK_USAGE
3248 static inline unsigned long stack_not_used(struct task_struct *p)
3250 unsigned long *n = end_of_stack(p);
3252 do { /* Skip over canary */
3253 # ifdef CONFIG_STACK_GROWSUP
3260 # ifdef CONFIG_STACK_GROWSUP
3261 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3263 return (unsigned long)n - (unsigned long)end_of_stack(p);
3267 extern void set_task_stack_end_magic(struct task_struct *tsk);
3269 /* set thread flags in other task's structures
3270 * - see asm/thread_info.h for TIF_xxxx flags available
3272 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3274 set_ti_thread_flag(task_thread_info(tsk), flag);
3277 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3279 clear_ti_thread_flag(task_thread_info(tsk), flag);
3282 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3284 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3287 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3289 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3292 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3294 return test_ti_thread_flag(task_thread_info(tsk), flag);
3297 static inline void set_tsk_need_resched(struct task_struct *tsk)
3299 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3302 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3304 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3307 static inline int test_tsk_need_resched(struct task_struct *tsk)
3309 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3312 static inline int restart_syscall(void)
3314 set_tsk_thread_flag(current, TIF_SIGPENDING);
3315 return -ERESTARTNOINTR;
3318 static inline int signal_pending(struct task_struct *p)
3320 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3323 static inline int __fatal_signal_pending(struct task_struct *p)
3325 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3328 static inline int fatal_signal_pending(struct task_struct *p)
3330 return signal_pending(p) && __fatal_signal_pending(p);
3333 static inline int signal_pending_state(long state, struct task_struct *p)
3335 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3337 if (!signal_pending(p))
3340 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3344 * cond_resched() and cond_resched_lock(): latency reduction via
3345 * explicit rescheduling in places that are safe. The return
3346 * value indicates whether a reschedule was done in fact.
3347 * cond_resched_lock() will drop the spinlock before scheduling,
3348 * cond_resched_softirq() will enable bhs before scheduling.
3350 #ifndef CONFIG_PREEMPT
3351 extern int _cond_resched(void);
3353 static inline int _cond_resched(void) { return 0; }
3356 #define cond_resched() ({ \
3357 ___might_sleep(__FILE__, __LINE__, 0); \
3361 extern int __cond_resched_lock(spinlock_t *lock);
3363 #define cond_resched_lock(lock) ({ \
3364 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3365 __cond_resched_lock(lock); \
3368 extern int __cond_resched_softirq(void);
3370 #define cond_resched_softirq() ({ \
3371 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3372 __cond_resched_softirq(); \
3375 static inline void cond_resched_rcu(void)
3377 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3385 * Does a critical section need to be broken due to another
3386 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3387 * but a general need for low latency)
3389 static inline int spin_needbreak(spinlock_t *lock)
3391 #ifdef CONFIG_PREEMPT
3392 return spin_is_contended(lock);
3399 * Idle thread specific functions to determine the need_resched
3402 #ifdef TIF_POLLING_NRFLAG
3403 static inline int tsk_is_polling(struct task_struct *p)
3405 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3408 static inline void __current_set_polling(void)
3410 set_thread_flag(TIF_POLLING_NRFLAG);
3413 static inline bool __must_check current_set_polling_and_test(void)
3415 __current_set_polling();
3418 * Polling state must be visible before we test NEED_RESCHED,
3419 * paired by resched_curr()
3421 smp_mb__after_atomic();
3423 return unlikely(tif_need_resched());
3426 static inline void __current_clr_polling(void)
3428 clear_thread_flag(TIF_POLLING_NRFLAG);
3431 static inline bool __must_check current_clr_polling_and_test(void)
3433 __current_clr_polling();
3436 * Polling state must be visible before we test NEED_RESCHED,
3437 * paired by resched_curr()
3439 smp_mb__after_atomic();
3441 return unlikely(tif_need_resched());
3445 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3446 static inline void __current_set_polling(void) { }
3447 static inline void __current_clr_polling(void) { }
3449 static inline bool __must_check current_set_polling_and_test(void)
3451 return unlikely(tif_need_resched());
3453 static inline bool __must_check current_clr_polling_and_test(void)
3455 return unlikely(tif_need_resched());
3459 static inline void current_clr_polling(void)
3461 __current_clr_polling();
3464 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3465 * Once the bit is cleared, we'll get IPIs with every new
3466 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3469 smp_mb(); /* paired with resched_curr() */
3471 preempt_fold_need_resched();
3474 static __always_inline bool need_resched(void)
3476 return unlikely(tif_need_resched());
3480 * Thread group CPU time accounting.
3482 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3483 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3486 * Reevaluate whether the task has signals pending delivery.
3487 * Wake the task if so.
3488 * This is required every time the blocked sigset_t changes.
3489 * callers must hold sighand->siglock.
3491 extern void recalc_sigpending_and_wake(struct task_struct *t);
3492 extern void recalc_sigpending(void);
3494 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3496 static inline void signal_wake_up(struct task_struct *t, bool resume)
3498 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3500 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3502 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3506 * Wrappers for p->thread_info->cpu access. No-op on UP.
3510 static inline unsigned int task_cpu(const struct task_struct *p)
3512 #ifdef CONFIG_THREAD_INFO_IN_TASK
3515 return task_thread_info(p)->cpu;
3519 static inline int task_node(const struct task_struct *p)
3521 return cpu_to_node(task_cpu(p));
3524 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3528 static inline unsigned int task_cpu(const struct task_struct *p)
3533 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3537 #endif /* CONFIG_SMP */
3540 * In order to reduce various lock holder preemption latencies provide an
3541 * interface to see if a vCPU is currently running or not.
3543 * This allows us to terminate optimistic spin loops and block, analogous to
3544 * the native optimistic spin heuristic of testing if the lock owner task is
3547 #ifndef vcpu_is_preempted
3548 # define vcpu_is_preempted(cpu) false
3551 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3552 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3554 #ifdef CONFIG_CGROUP_SCHED
3555 extern struct task_group root_task_group;
3556 #endif /* CONFIG_CGROUP_SCHED */
3558 extern int task_can_switch_user(struct user_struct *up,
3559 struct task_struct *tsk);
3561 #ifdef CONFIG_TASK_XACCT
3562 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3564 tsk->ioac.rchar += amt;
3567 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3569 tsk->ioac.wchar += amt;
3572 static inline void inc_syscr(struct task_struct *tsk)
3577 static inline void inc_syscw(struct task_struct *tsk)
3582 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3586 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3590 static inline void inc_syscr(struct task_struct *tsk)
3594 static inline void inc_syscw(struct task_struct *tsk)
3599 #ifndef TASK_SIZE_OF
3600 #define TASK_SIZE_OF(tsk) TASK_SIZE
3604 extern void mm_update_next_owner(struct mm_struct *mm);
3606 static inline void mm_update_next_owner(struct mm_struct *mm)
3609 #endif /* CONFIG_MEMCG */
3611 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3614 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3617 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3620 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3623 static inline unsigned long rlimit(unsigned int limit)
3625 return task_rlimit(current, limit);
3628 static inline unsigned long rlimit_max(unsigned int limit)
3630 return task_rlimit_max(current, limit);
3633 #define SCHED_CPUFREQ_RT (1U << 0)
3634 #define SCHED_CPUFREQ_DL (1U << 1)
3635 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3637 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3639 #ifdef CONFIG_CPU_FREQ
3640 struct update_util_data {
3641 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3644 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3645 void (*func)(struct update_util_data *data, u64 time,
3646 unsigned int flags));
3647 void cpufreq_remove_update_util_hook(int cpu);
3648 #endif /* CONFIG_CPU_FREQ */