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/mutex.h>
22 #include <linux/plist.h>
23 #include <linux/rbtree.h>
24 #include <linux/thread_info.h>
25 #include <linux/cpumask.h>
26 #include <linux/errno.h>
27 #include <linux/nodemask.h>
28 #include <linux/mm_types.h>
29 #include <linux/preempt.h>
32 #include <asm/ptrace.h>
34 #include <linux/smp.h>
35 #include <linux/sem.h>
36 #include <linux/shm.h>
37 #include <linux/signal.h>
38 #include <linux/compiler.h>
39 #include <linux/completion.h>
40 #include <linux/signal_types.h>
41 #include <linux/pid.h>
42 #include <linux/percpu.h>
43 #include <linux/topology.h>
44 #include <linux/seccomp.h>
45 #include <linux/rcupdate.h>
46 #include <linux/rculist.h>
47 #include <linux/rtmutex.h>
49 #include <linux/time.h>
50 #include <linux/param.h>
51 #include <linux/resource.h>
52 #include <linux/timer.h>
53 #include <linux/hrtimer.h>
54 #include <linux/kcov.h>
55 #include <linux/task_io_accounting.h>
56 #include <linux/latencytop.h>
57 #include <linux/cred.h>
58 #include <linux/llist.h>
59 #include <linux/uidgid.h>
60 #include <linux/gfp.h>
61 #include <linux/topology.h>
62 #include <linux/magic.h>
63 #include <linux/cgroup-defs.h>
65 #include <asm/processor.h>
67 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
70 * Extended scheduling parameters data structure.
72 * This is needed because the original struct sched_param can not be
73 * altered without introducing ABI issues with legacy applications
74 * (e.g., in sched_getparam()).
76 * However, the possibility of specifying more than just a priority for
77 * the tasks may be useful for a wide variety of application fields, e.g.,
78 * multimedia, streaming, automation and control, and many others.
80 * This variant (sched_attr) is meant at describing a so-called
81 * sporadic time-constrained task. In such model a task is specified by:
82 * - the activation period or minimum instance inter-arrival time;
83 * - the maximum (or average, depending on the actual scheduling
84 * discipline) computation time of all instances, a.k.a. runtime;
85 * - the deadline (relative to the actual activation time) of each
87 * Very briefly, a periodic (sporadic) task asks for the execution of
88 * some specific computation --which is typically called an instance--
89 * (at most) every period. Moreover, each instance typically lasts no more
90 * than the runtime and must be completed by time instant t equal to
91 * the instance activation time + the deadline.
93 * This is reflected by the actual fields of the sched_attr structure:
95 * @size size of the structure, for fwd/bwd compat.
97 * @sched_policy task's scheduling policy
98 * @sched_flags for customizing the scheduler behaviour
99 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
100 * @sched_priority task's static priority (SCHED_FIFO/RR)
101 * @sched_deadline representative of the task's deadline
102 * @sched_runtime representative of the task's runtime
103 * @sched_period representative of the task's period
105 * Given this task model, there are a multiplicity of scheduling algorithms
106 * and policies, that can be used to ensure all the tasks will make their
107 * timing constraints.
109 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
110 * only user of this new interface. More information about the algorithm
111 * available in the scheduling class file or in Documentation/.
119 /* SCHED_NORMAL, SCHED_BATCH */
122 /* SCHED_FIFO, SCHED_RR */
131 struct futex_pi_state;
132 struct robust_list_head;
135 struct perf_event_context;
141 * These are the constant used to fake the fixed-point load-average
142 * counting. Some notes:
143 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
144 * a load-average precision of 10 bits integer + 11 bits fractional
145 * - if you want to count load-averages more often, you need more
146 * precision, or rounding will get you. With 2-second counting freq,
147 * the EXP_n values would be 1981, 2034 and 2043 if still using only
150 extern unsigned long avenrun[]; /* Load averages */
151 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
153 #define FSHIFT 11 /* nr of bits of precision */
154 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
155 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
156 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
157 #define EXP_5 2014 /* 1/exp(5sec/5min) */
158 #define EXP_15 2037 /* 1/exp(5sec/15min) */
160 #define CALC_LOAD(load,exp,n) \
162 load += n*(FIXED_1-exp); \
165 extern unsigned long total_forks;
166 extern int nr_threads;
167 DECLARE_PER_CPU(unsigned long, process_counts);
168 extern int nr_processes(void);
169 extern unsigned long nr_running(void);
170 extern bool single_task_running(void);
171 extern unsigned long nr_iowait(void);
172 extern unsigned long nr_iowait_cpu(int cpu);
173 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
175 extern void calc_global_load(unsigned long ticks);
177 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
178 extern void cpu_load_update_nohz_start(void);
179 extern void cpu_load_update_nohz_stop(void);
181 static inline void cpu_load_update_nohz_start(void) { }
182 static inline void cpu_load_update_nohz_stop(void) { }
185 extern void dump_cpu_task(int cpu);
190 #ifdef CONFIG_SCHED_DEBUG
191 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
192 extern void proc_sched_set_task(struct task_struct *p);
196 * Task state bitmask. NOTE! These bits are also
197 * encoded in fs/proc/array.c: get_task_state().
199 * We have two separate sets of flags: task->state
200 * is about runnability, while task->exit_state are
201 * about the task exiting. Confusing, but this way
202 * modifying one set can't modify the other one by
205 #define TASK_RUNNING 0
206 #define TASK_INTERRUPTIBLE 1
207 #define TASK_UNINTERRUPTIBLE 2
208 #define __TASK_STOPPED 4
209 #define __TASK_TRACED 8
210 /* in tsk->exit_state */
212 #define EXIT_ZOMBIE 32
213 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
214 /* in tsk->state again */
216 #define TASK_WAKEKILL 128
217 #define TASK_WAKING 256
218 #define TASK_PARKED 512
219 #define TASK_NOLOAD 1024
220 #define TASK_NEW 2048
221 #define TASK_STATE_MAX 4096
223 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
225 /* Convenience macros for the sake of set_current_state */
226 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
227 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
228 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
230 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
232 /* Convenience macros for the sake of wake_up */
233 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
234 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
236 /* get_task_state() */
237 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
238 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
239 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
241 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
242 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
243 #define task_is_stopped_or_traced(task) \
244 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
245 #define task_contributes_to_load(task) \
246 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
247 (task->flags & PF_FROZEN) == 0 && \
248 (task->state & TASK_NOLOAD) == 0)
250 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
252 #define __set_current_state(state_value) \
254 current->task_state_change = _THIS_IP_; \
255 current->state = (state_value); \
257 #define set_current_state(state_value) \
259 current->task_state_change = _THIS_IP_; \
260 smp_store_mb(current->state, (state_value)); \
265 * set_current_state() includes a barrier so that the write of current->state
266 * is correctly serialised wrt the caller's subsequent test of whether to
270 * set_current_state(TASK_UNINTERRUPTIBLE);
276 * __set_current_state(TASK_RUNNING);
278 * If the caller does not need such serialisation (because, for instance, the
279 * condition test and condition change and wakeup are under the same lock) then
280 * use __set_current_state().
282 * The above is typically ordered against the wakeup, which does:
284 * need_sleep = false;
285 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
287 * Where wake_up_state() (and all other wakeup primitives) imply enough
288 * barriers to order the store of the variable against wakeup.
290 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
291 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
292 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
294 * This is obviously fine, since they both store the exact same value.
296 * Also see the comments of try_to_wake_up().
298 #define __set_current_state(state_value) \
299 do { current->state = (state_value); } while (0)
300 #define set_current_state(state_value) \
301 smp_store_mb(current->state, (state_value))
305 /* Task command name length */
306 #define TASK_COMM_LEN 16
308 #include <linux/spinlock.h>
311 * This serializes "schedule()" and also protects
312 * the run-queue from deletions/modifications (but
313 * _adding_ to the beginning of the run-queue has
316 extern rwlock_t tasklist_lock;
317 extern spinlock_t mmlist_lock;
321 #ifdef CONFIG_PROVE_RCU
322 extern int lockdep_tasklist_lock_is_held(void);
323 #endif /* #ifdef CONFIG_PROVE_RCU */
325 extern void sched_init(void);
326 extern void sched_init_smp(void);
327 extern asmlinkage void schedule_tail(struct task_struct *prev);
328 extern void init_idle(struct task_struct *idle, int cpu);
329 extern void init_idle_bootup_task(struct task_struct *idle);
331 extern cpumask_var_t cpu_isolated_map;
333 extern int runqueue_is_locked(int cpu);
335 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
336 extern void nohz_balance_enter_idle(int cpu);
337 extern void set_cpu_sd_state_idle(void);
338 extern int get_nohz_timer_target(void);
340 static inline void nohz_balance_enter_idle(int cpu) { }
341 static inline void set_cpu_sd_state_idle(void) { }
345 * Only dump TASK_* tasks. (0 for all tasks)
347 extern void show_state_filter(unsigned long state_filter);
349 static inline void show_state(void)
351 show_state_filter(0);
354 extern void show_regs(struct pt_regs *);
357 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
358 * task), SP is the stack pointer of the first frame that should be shown in the back
359 * trace (or NULL if the entire call-chain of the task should be shown).
361 extern void show_stack(struct task_struct *task, unsigned long *sp);
363 extern void cpu_init (void);
364 extern void trap_init(void);
365 extern void update_process_times(int user);
366 extern void scheduler_tick(void);
367 extern int sched_cpu_starting(unsigned int cpu);
368 extern int sched_cpu_activate(unsigned int cpu);
369 extern int sched_cpu_deactivate(unsigned int cpu);
371 #ifdef CONFIG_HOTPLUG_CPU
372 extern int sched_cpu_dying(unsigned int cpu);
374 # define sched_cpu_dying NULL
377 extern void sched_show_task(struct task_struct *p);
379 #ifdef CONFIG_LOCKUP_DETECTOR
380 extern void touch_softlockup_watchdog_sched(void);
381 extern void touch_softlockup_watchdog(void);
382 extern void touch_softlockup_watchdog_sync(void);
383 extern void touch_all_softlockup_watchdogs(void);
384 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
386 size_t *lenp, loff_t *ppos);
387 extern unsigned int softlockup_panic;
388 extern unsigned int hardlockup_panic;
389 void lockup_detector_init(void);
391 static inline void touch_softlockup_watchdog_sched(void)
394 static inline void touch_softlockup_watchdog(void)
397 static inline void touch_softlockup_watchdog_sync(void)
400 static inline void touch_all_softlockup_watchdogs(void)
403 static inline void lockup_detector_init(void)
408 #ifdef CONFIG_DETECT_HUNG_TASK
409 void reset_hung_task_detector(void);
411 static inline void reset_hung_task_detector(void)
416 /* Attach to any functions which should be ignored in wchan output. */
417 #define __sched __attribute__((__section__(".sched.text")))
419 /* Linker adds these: start and end of __sched functions */
420 extern char __sched_text_start[], __sched_text_end[];
422 /* Is this address in the __sched functions? */
423 extern int in_sched_functions(unsigned long addr);
425 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
426 extern signed long schedule_timeout(signed long timeout);
427 extern signed long schedule_timeout_interruptible(signed long timeout);
428 extern signed long schedule_timeout_killable(signed long timeout);
429 extern signed long schedule_timeout_uninterruptible(signed long timeout);
430 extern signed long schedule_timeout_idle(signed long timeout);
431 asmlinkage void schedule(void);
432 extern void schedule_preempt_disabled(void);
434 extern int __must_check io_schedule_prepare(void);
435 extern void io_schedule_finish(int token);
436 extern long io_schedule_timeout(long timeout);
437 extern void io_schedule(void);
439 void __noreturn do_task_dead(void);
442 struct user_namespace;
445 extern void arch_pick_mmap_layout(struct mm_struct *mm);
447 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
448 unsigned long, unsigned long);
450 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
451 unsigned long len, unsigned long pgoff,
452 unsigned long flags);
454 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
457 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
458 #define SUID_DUMP_USER 1 /* Dump as user of process */
459 #define SUID_DUMP_ROOT 2 /* Dump as root */
463 /* for SUID_DUMP_* above */
464 #define MMF_DUMPABLE_BITS 2
465 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
467 extern void set_dumpable(struct mm_struct *mm, int value);
469 * This returns the actual value of the suid_dumpable flag. For things
470 * that are using this for checking for privilege transitions, it must
471 * test against SUID_DUMP_USER rather than treating it as a boolean
474 static inline int __get_dumpable(unsigned long mm_flags)
476 return mm_flags & MMF_DUMPABLE_MASK;
479 static inline int get_dumpable(struct mm_struct *mm)
481 return __get_dumpable(mm->flags);
484 /* coredump filter bits */
485 #define MMF_DUMP_ANON_PRIVATE 2
486 #define MMF_DUMP_ANON_SHARED 3
487 #define MMF_DUMP_MAPPED_PRIVATE 4
488 #define MMF_DUMP_MAPPED_SHARED 5
489 #define MMF_DUMP_ELF_HEADERS 6
490 #define MMF_DUMP_HUGETLB_PRIVATE 7
491 #define MMF_DUMP_HUGETLB_SHARED 8
492 #define MMF_DUMP_DAX_PRIVATE 9
493 #define MMF_DUMP_DAX_SHARED 10
495 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
496 #define MMF_DUMP_FILTER_BITS 9
497 #define MMF_DUMP_FILTER_MASK \
498 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
499 #define MMF_DUMP_FILTER_DEFAULT \
500 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
501 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
503 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
504 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
506 # define MMF_DUMP_MASK_DEFAULT_ELF 0
508 /* leave room for more dump flags */
509 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
510 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
512 * This one-shot flag is dropped due to necessity of changing exe once again
515 //#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
517 #define MMF_HAS_UPROBES 19 /* has uprobes */
518 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
519 #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
520 #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
521 #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
523 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
525 struct sighand_struct {
527 struct k_sigaction action[_NSIG];
529 wait_queue_head_t signalfd_wqh;
532 struct pacct_struct {
535 unsigned long ac_mem;
536 u64 ac_utime, ac_stime;
537 unsigned long ac_minflt, ac_majflt;
546 * struct prev_cputime - snaphsot of system and user cputime
547 * @utime: time spent in user mode
548 * @stime: time spent in system mode
549 * @lock: protects the above two fields
551 * Stores previous user/system time values such that we can guarantee
554 struct prev_cputime {
555 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
562 static inline void prev_cputime_init(struct prev_cputime *prev)
564 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
565 prev->utime = prev->stime = 0;
566 raw_spin_lock_init(&prev->lock);
571 * struct task_cputime - collected CPU time counts
572 * @utime: time spent in user mode, in nanoseconds
573 * @stime: time spent in kernel mode, in nanoseconds
574 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
576 * This structure groups together three kinds of CPU time that are tracked for
577 * threads and thread groups. Most things considering CPU time want to group
578 * these counts together and treat all three of them in parallel.
580 struct task_cputime {
583 unsigned long long sum_exec_runtime;
586 /* Alternate field names when used to cache expirations. */
587 #define virt_exp utime
588 #define prof_exp stime
589 #define sched_exp sum_exec_runtime
592 * This is the atomic variant of task_cputime, which can be used for
593 * storing and updating task_cputime statistics without locking.
595 struct task_cputime_atomic {
598 atomic64_t sum_exec_runtime;
601 #define INIT_CPUTIME_ATOMIC \
602 (struct task_cputime_atomic) { \
603 .utime = ATOMIC64_INIT(0), \
604 .stime = ATOMIC64_INIT(0), \
605 .sum_exec_runtime = ATOMIC64_INIT(0), \
608 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
611 * Disable preemption until the scheduler is running -- use an unconditional
612 * value so that it also works on !PREEMPT_COUNT kernels.
614 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
616 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
619 * Initial preempt_count value; reflects the preempt_count schedule invariant
620 * which states that during context switches:
622 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
624 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
625 * Note: See finish_task_switch().
627 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
630 * struct thread_group_cputimer - thread group interval timer counts
631 * @cputime_atomic: atomic thread group interval timers.
632 * @running: true when there are timers running and
633 * @cputime_atomic receives updates.
634 * @checking_timer: true when a thread in the group is in the
635 * process of checking for thread group timers.
637 * This structure contains the version of task_cputime, above, that is
638 * used for thread group CPU timer calculations.
640 struct thread_group_cputimer {
641 struct task_cputime_atomic cputime_atomic;
646 #include <linux/rwsem.h>
650 * NOTE! "signal_struct" does not have its own
651 * locking, because a shared signal_struct always
652 * implies a shared sighand_struct, so locking
653 * sighand_struct is always a proper superset of
654 * the locking of signal_struct.
656 struct signal_struct {
660 struct list_head thread_head;
662 wait_queue_head_t wait_chldexit; /* for wait4() */
664 /* current thread group signal load-balancing target: */
665 struct task_struct *curr_target;
667 /* shared signal handling: */
668 struct sigpending shared_pending;
670 /* thread group exit support */
673 * - notify group_exit_task when ->count is equal to notify_count
674 * - everyone except group_exit_task is stopped during signal delivery
675 * of fatal signals, group_exit_task processes the signal.
678 struct task_struct *group_exit_task;
680 /* thread group stop support, overloads group_exit_code too */
681 int group_stop_count;
682 unsigned int flags; /* see SIGNAL_* flags below */
685 * PR_SET_CHILD_SUBREAPER marks a process, like a service
686 * manager, to re-parent orphan (double-forking) child processes
687 * to this process instead of 'init'. The service manager is
688 * able to receive SIGCHLD signals and is able to investigate
689 * the process until it calls wait(). All children of this
690 * process will inherit a flag if they should look for a
691 * child_subreaper process at exit.
693 unsigned int is_child_subreaper:1;
694 unsigned int has_child_subreaper:1;
696 #ifdef CONFIG_POSIX_TIMERS
698 /* POSIX.1b Interval Timers */
700 struct list_head posix_timers;
702 /* ITIMER_REAL timer for the process */
703 struct hrtimer real_timer;
704 ktime_t it_real_incr;
707 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
708 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
709 * values are defined to 0 and 1 respectively
711 struct cpu_itimer it[2];
714 * Thread group totals for process CPU timers.
715 * See thread_group_cputimer(), et al, for details.
717 struct thread_group_cputimer cputimer;
719 /* Earliest-expiration cache. */
720 struct task_cputime cputime_expires;
722 struct list_head cpu_timers[3];
726 struct pid *leader_pid;
728 #ifdef CONFIG_NO_HZ_FULL
729 atomic_t tick_dep_mask;
732 struct pid *tty_old_pgrp;
734 /* boolean value for session group leader */
737 struct tty_struct *tty; /* NULL if no tty */
739 #ifdef CONFIG_SCHED_AUTOGROUP
740 struct autogroup *autogroup;
743 * Cumulative resource counters for dead threads in the group,
744 * and for reaped dead child processes forked by this group.
745 * Live threads maintain their own counters and add to these
746 * in __exit_signal, except for the group leader.
748 seqlock_t stats_lock;
749 u64 utime, stime, cutime, cstime;
752 struct prev_cputime prev_cputime;
753 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
754 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
755 unsigned long inblock, oublock, cinblock, coublock;
756 unsigned long maxrss, cmaxrss;
757 struct task_io_accounting ioac;
760 * Cumulative ns of schedule CPU time fo dead threads in the
761 * group, not including a zombie group leader, (This only differs
762 * from jiffies_to_ns(utime + stime) if sched_clock uses something
763 * other than jiffies.)
765 unsigned long long sum_sched_runtime;
768 * We don't bother to synchronize most readers of this at all,
769 * because there is no reader checking a limit that actually needs
770 * to get both rlim_cur and rlim_max atomically, and either one
771 * alone is a single word that can safely be read normally.
772 * getrlimit/setrlimit use task_lock(current->group_leader) to
773 * protect this instead of the siglock, because they really
774 * have no need to disable irqs.
776 struct rlimit rlim[RLIM_NLIMITS];
778 #ifdef CONFIG_BSD_PROCESS_ACCT
779 struct pacct_struct pacct; /* per-process accounting information */
781 #ifdef CONFIG_TASKSTATS
782 struct taskstats *stats;
786 struct tty_audit_buf *tty_audit_buf;
790 * Thread is the potential origin of an oom condition; kill first on
793 bool oom_flag_origin;
794 short oom_score_adj; /* OOM kill score adjustment */
795 short oom_score_adj_min; /* OOM kill score adjustment min value.
796 * Only settable by CAP_SYS_RESOURCE. */
797 struct mm_struct *oom_mm; /* recorded mm when the thread group got
798 * killed by the oom killer */
800 struct mutex cred_guard_mutex; /* guard against foreign influences on
801 * credential calculations
802 * (notably. ptrace) */
806 * Bits in flags field of signal_struct.
808 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
809 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
810 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
811 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
813 * Pending notifications to parent.
815 #define SIGNAL_CLD_STOPPED 0x00000010
816 #define SIGNAL_CLD_CONTINUED 0x00000020
817 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
819 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
821 #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
822 SIGNAL_STOP_CONTINUED)
824 static inline void signal_set_stop_flags(struct signal_struct *sig,
827 WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
828 sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
831 /* If true, all threads except ->group_exit_task have pending SIGKILL */
832 static inline int signal_group_exit(const struct signal_struct *sig)
834 return (sig->flags & SIGNAL_GROUP_EXIT) ||
835 (sig->group_exit_task != NULL);
839 * Some day this will be a full-fledged user tracking system..
842 atomic_t __count; /* reference count */
843 atomic_t processes; /* How many processes does this user have? */
844 atomic_t sigpending; /* How many pending signals does this user have? */
845 #ifdef CONFIG_FANOTIFY
846 atomic_t fanotify_listeners;
849 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
851 #ifdef CONFIG_POSIX_MQUEUE
852 /* protected by mq_lock */
853 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
855 unsigned long locked_shm; /* How many pages of mlocked shm ? */
856 unsigned long unix_inflight; /* How many files in flight in unix sockets */
857 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
860 struct key *uid_keyring; /* UID specific keyring */
861 struct key *session_keyring; /* UID's default session keyring */
864 /* Hash table maintenance information */
865 struct hlist_node uidhash_node;
868 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
869 atomic_long_t locked_vm;
873 extern int uids_sysfs_init(void);
875 extern struct user_struct *find_user(kuid_t);
877 extern struct user_struct root_user;
878 #define INIT_USER (&root_user)
881 struct backing_dev_info;
882 struct reclaim_state;
884 #ifdef CONFIG_SCHED_INFO
886 /* cumulative counters */
887 unsigned long pcount; /* # of times run on this cpu */
888 unsigned long long run_delay; /* time spent waiting on a runqueue */
891 unsigned long long last_arrival,/* when we last ran on a cpu */
892 last_queued; /* when we were last queued to run */
894 #endif /* CONFIG_SCHED_INFO */
896 #ifdef CONFIG_TASK_DELAY_ACCT
897 struct task_delay_info {
899 unsigned int flags; /* Private per-task flags */
901 /* For each stat XXX, add following, aligned appropriately
903 * struct timespec XXX_start, XXX_end;
907 * Atomicity of updates to XXX_delay, XXX_count protected by
908 * single lock above (split into XXX_lock if contention is an issue).
912 * XXX_count is incremented on every XXX operation, the delay
913 * associated with the operation is added to XXX_delay.
914 * XXX_delay contains the accumulated delay time in nanoseconds.
916 u64 blkio_start; /* Shared by blkio, swapin */
917 u64 blkio_delay; /* wait for sync block io completion */
918 u64 swapin_delay; /* wait for swapin block io completion */
919 u32 blkio_count; /* total count of the number of sync block */
920 /* io operations performed */
921 u32 swapin_count; /* total count of the number of swapin block */
922 /* io operations performed */
925 u64 freepages_delay; /* wait for memory reclaim */
926 u32 freepages_count; /* total count of memory reclaim */
928 #endif /* CONFIG_TASK_DELAY_ACCT */
930 static inline int sched_info_on(void)
932 #ifdef CONFIG_SCHEDSTATS
934 #elif defined(CONFIG_TASK_DELAY_ACCT)
935 extern int delayacct_on;
942 #ifdef CONFIG_SCHEDSTATS
943 void force_schedstat_enabled(void);
947 * Integer metrics need fixed point arithmetic, e.g., sched/fair
948 * has a few: load, load_avg, util_avg, freq, and capacity.
950 * We define a basic fixed point arithmetic range, and then formalize
951 * all these metrics based on that basic range.
953 # define SCHED_FIXEDPOINT_SHIFT 10
954 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
957 * Wake-queues are lists of tasks with a pending wakeup, whose
958 * callers have already marked the task as woken internally,
959 * and can thus carry on. A common use case is being able to
960 * do the wakeups once the corresponding user lock as been
963 * We hold reference to each task in the list across the wakeup,
964 * thus guaranteeing that the memory is still valid by the time
965 * the actual wakeups are performed in wake_up_q().
967 * One per task suffices, because there's never a need for a task to be
968 * in two wake queues simultaneously; it is forbidden to abandon a task
969 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
970 * already in a wake queue, the wakeup will happen soon and the second
971 * waker can just skip it.
973 * The DEFINE_WAKE_Q macro declares and initializes the list head.
974 * wake_up_q() does NOT reinitialize the list; it's expected to be
975 * called near the end of a function. Otherwise, the list can be
976 * re-initialized for later re-use by wake_q_init().
978 * Note that this can cause spurious wakeups. schedule() callers
979 * must ensure the call is done inside a loop, confirming that the
980 * wakeup condition has in fact occurred.
983 struct wake_q_node *next;
987 struct wake_q_node *first;
988 struct wake_q_node **lastp;
991 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
993 #define DEFINE_WAKE_Q(name) \
994 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
996 static inline void wake_q_init(struct wake_q_head *head)
998 head->first = WAKE_Q_TAIL;
999 head->lastp = &head->first;
1002 extern void wake_q_add(struct wake_q_head *head,
1003 struct task_struct *task);
1004 extern void wake_up_q(struct wake_q_head *head);
1006 struct io_context; /* See blkdev.h */
1009 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1010 extern void prefetch_stack(struct task_struct *t);
1012 static inline void prefetch_stack(struct task_struct *t) { }
1015 struct audit_context; /* See audit.c */
1017 struct pipe_inode_info;
1018 struct uts_namespace;
1020 struct load_weight {
1021 unsigned long weight;
1026 * The load_avg/util_avg accumulates an infinite geometric series
1027 * (see __update_load_avg() in kernel/sched/fair.c).
1029 * [load_avg definition]
1031 * load_avg = runnable% * scale_load_down(load)
1033 * where runnable% is the time ratio that a sched_entity is runnable.
1034 * For cfs_rq, it is the aggregated load_avg of all runnable and
1035 * blocked sched_entities.
1037 * load_avg may also take frequency scaling into account:
1039 * load_avg = runnable% * scale_load_down(load) * freq%
1041 * where freq% is the CPU frequency normalized to the highest frequency.
1043 * [util_avg definition]
1045 * util_avg = running% * SCHED_CAPACITY_SCALE
1047 * where running% is the time ratio that a sched_entity is running on
1048 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1049 * and blocked sched_entities.
1051 * util_avg may also factor frequency scaling and CPU capacity scaling:
1053 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1055 * where freq% is the same as above, and capacity% is the CPU capacity
1056 * normalized to the greatest capacity (due to uarch differences, etc).
1058 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1059 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1060 * we therefore scale them to as large a range as necessary. This is for
1061 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1065 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1066 * with the highest load (=88761), always runnable on a single cfs_rq,
1067 * and should not overflow as the number already hits PID_MAX_LIMIT.
1069 * For all other cases (including 32-bit kernels), struct load_weight's
1070 * weight will overflow first before we do, because:
1072 * Max(load_avg) <= Max(load.weight)
1074 * Then it is the load_weight's responsibility to consider overflow
1078 u64 last_update_time, load_sum;
1079 u32 util_sum, period_contrib;
1080 unsigned long load_avg, util_avg;
1083 #ifdef CONFIG_SCHEDSTATS
1084 struct sched_statistics {
1094 s64 sum_sleep_runtime;
1101 u64 nr_migrations_cold;
1102 u64 nr_failed_migrations_affine;
1103 u64 nr_failed_migrations_running;
1104 u64 nr_failed_migrations_hot;
1105 u64 nr_forced_migrations;
1108 u64 nr_wakeups_sync;
1109 u64 nr_wakeups_migrate;
1110 u64 nr_wakeups_local;
1111 u64 nr_wakeups_remote;
1112 u64 nr_wakeups_affine;
1113 u64 nr_wakeups_affine_attempts;
1114 u64 nr_wakeups_passive;
1115 u64 nr_wakeups_idle;
1119 struct sched_entity {
1120 struct load_weight load; /* for load-balancing */
1121 struct rb_node run_node;
1122 struct list_head group_node;
1126 u64 sum_exec_runtime;
1128 u64 prev_sum_exec_runtime;
1132 #ifdef CONFIG_SCHEDSTATS
1133 struct sched_statistics statistics;
1136 #ifdef CONFIG_FAIR_GROUP_SCHED
1138 struct sched_entity *parent;
1139 /* rq on which this entity is (to be) queued: */
1140 struct cfs_rq *cfs_rq;
1141 /* rq "owned" by this entity/group: */
1142 struct cfs_rq *my_q;
1147 * Per entity load average tracking.
1149 * Put into separate cache line so it does not
1150 * collide with read-mostly values above.
1152 struct sched_avg avg ____cacheline_aligned_in_smp;
1156 struct sched_rt_entity {
1157 struct list_head run_list;
1158 unsigned long timeout;
1159 unsigned long watchdog_stamp;
1160 unsigned int time_slice;
1161 unsigned short on_rq;
1162 unsigned short on_list;
1164 struct sched_rt_entity *back;
1165 #ifdef CONFIG_RT_GROUP_SCHED
1166 struct sched_rt_entity *parent;
1167 /* rq on which this entity is (to be) queued: */
1168 struct rt_rq *rt_rq;
1169 /* rq "owned" by this entity/group: */
1174 struct sched_dl_entity {
1175 struct rb_node rb_node;
1178 * Original scheduling parameters. Copied here from sched_attr
1179 * during sched_setattr(), they will remain the same until
1180 * the next sched_setattr().
1182 u64 dl_runtime; /* maximum runtime for each instance */
1183 u64 dl_deadline; /* relative deadline of each instance */
1184 u64 dl_period; /* separation of two instances (period) */
1185 u64 dl_bw; /* dl_runtime / dl_deadline */
1188 * Actual scheduling parameters. Initialized with the values above,
1189 * they are continously updated during task execution. Note that
1190 * the remaining runtime could be < 0 in case we are in overrun.
1192 s64 runtime; /* remaining runtime for this instance */
1193 u64 deadline; /* absolute deadline for this instance */
1194 unsigned int flags; /* specifying the scheduler behaviour */
1199 * @dl_throttled tells if we exhausted the runtime. If so, the
1200 * task has to wait for a replenishment to be performed at the
1201 * next firing of dl_timer.
1203 * @dl_boosted tells if we are boosted due to DI. If so we are
1204 * outside bandwidth enforcement mechanism (but only until we
1205 * exit the critical section);
1207 * @dl_yielded tells if task gave up the cpu before consuming
1208 * all its available runtime during the last job.
1210 int dl_throttled, dl_boosted, dl_yielded;
1213 * Bandwidth enforcement timer. Each -deadline task has its
1214 * own bandwidth to be enforced, thus we need one timer per task.
1216 struct hrtimer dl_timer;
1224 u8 pad; /* Otherwise the compiler can store garbage here. */
1226 u32 s; /* Set of bits. */
1230 enum perf_event_task_context {
1231 perf_invalid_context = -1,
1232 perf_hw_context = 0,
1234 perf_nr_task_contexts,
1237 /* Track pages that require TLB flushes */
1238 struct tlbflush_unmap_batch {
1240 * Each bit set is a CPU that potentially has a TLB entry for one of
1241 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1243 struct cpumask cpumask;
1245 /* True if any bit in cpumask is set */
1246 bool flush_required;
1249 * If true then the PTE was dirty when unmapped. The entry must be
1250 * flushed before IO is initiated or a stale TLB entry potentially
1251 * allows an update without redirtying the page.
1256 struct task_struct {
1257 #ifdef CONFIG_THREAD_INFO_IN_TASK
1259 * For reasons of header soup (see current_thread_info()), this
1260 * must be the first element of task_struct.
1262 struct thread_info thread_info;
1264 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1267 unsigned int flags; /* per process flags, defined below */
1268 unsigned int ptrace;
1271 struct llist_node wake_entry;
1273 #ifdef CONFIG_THREAD_INFO_IN_TASK
1274 unsigned int cpu; /* current CPU */
1276 unsigned int wakee_flips;
1277 unsigned long wakee_flip_decay_ts;
1278 struct task_struct *last_wakee;
1284 int prio, static_prio, normal_prio;
1285 unsigned int rt_priority;
1286 const struct sched_class *sched_class;
1287 struct sched_entity se;
1288 struct sched_rt_entity rt;
1289 #ifdef CONFIG_CGROUP_SCHED
1290 struct task_group *sched_task_group;
1292 struct sched_dl_entity dl;
1294 #ifdef CONFIG_PREEMPT_NOTIFIERS
1295 /* list of struct preempt_notifier: */
1296 struct hlist_head preempt_notifiers;
1299 #ifdef CONFIG_BLK_DEV_IO_TRACE
1300 unsigned int btrace_seq;
1303 unsigned int policy;
1304 int nr_cpus_allowed;
1305 cpumask_t cpus_allowed;
1307 #ifdef CONFIG_PREEMPT_RCU
1308 int rcu_read_lock_nesting;
1309 union rcu_special rcu_read_unlock_special;
1310 struct list_head rcu_node_entry;
1311 struct rcu_node *rcu_blocked_node;
1312 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1313 #ifdef CONFIG_TASKS_RCU
1314 unsigned long rcu_tasks_nvcsw;
1315 bool rcu_tasks_holdout;
1316 struct list_head rcu_tasks_holdout_list;
1317 int rcu_tasks_idle_cpu;
1318 #endif /* #ifdef CONFIG_TASKS_RCU */
1320 #ifdef CONFIG_SCHED_INFO
1321 struct sched_info sched_info;
1324 struct list_head tasks;
1326 struct plist_node pushable_tasks;
1327 struct rb_node pushable_dl_tasks;
1330 struct mm_struct *mm, *active_mm;
1332 /* Per-thread vma caching: */
1333 struct vmacache vmacache;
1335 #if defined(SPLIT_RSS_COUNTING)
1336 struct task_rss_stat rss_stat;
1340 int exit_code, exit_signal;
1341 int pdeath_signal; /* The signal sent when the parent dies */
1342 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1344 /* Used for emulating ABI behavior of previous Linux versions */
1345 unsigned int personality;
1347 /* scheduler bits, serialized by scheduler locks */
1348 unsigned sched_reset_on_fork:1;
1349 unsigned sched_contributes_to_load:1;
1350 unsigned sched_migrated:1;
1351 unsigned sched_remote_wakeup:1;
1352 unsigned :0; /* force alignment to the next boundary */
1354 /* unserialized, strictly 'current' */
1355 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1356 unsigned in_iowait:1;
1357 #if !defined(TIF_RESTORE_SIGMASK)
1358 unsigned restore_sigmask:1;
1361 unsigned memcg_may_oom:1;
1363 unsigned memcg_kmem_skip_account:1;
1366 #ifdef CONFIG_COMPAT_BRK
1367 unsigned brk_randomized:1;
1370 unsigned long atomic_flags; /* Flags needing atomic access. */
1372 struct restart_block restart_block;
1377 #ifdef CONFIG_CC_STACKPROTECTOR
1378 /* Canary value for the -fstack-protector gcc feature */
1379 unsigned long stack_canary;
1382 * pointers to (original) parent process, youngest child, younger sibling,
1383 * older sibling, respectively. (p->father can be replaced with
1384 * p->real_parent->pid)
1386 struct task_struct __rcu *real_parent; /* real parent process */
1387 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1389 * children/sibling forms the list of my natural children
1391 struct list_head children; /* list of my children */
1392 struct list_head sibling; /* linkage in my parent's children list */
1393 struct task_struct *group_leader; /* threadgroup leader */
1396 * ptraced is the list of tasks this task is using ptrace on.
1397 * This includes both natural children and PTRACE_ATTACH targets.
1398 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1400 struct list_head ptraced;
1401 struct list_head ptrace_entry;
1403 /* PID/PID hash table linkage. */
1404 struct pid_link pids[PIDTYPE_MAX];
1405 struct list_head thread_group;
1406 struct list_head thread_node;
1408 struct completion *vfork_done; /* for vfork() */
1409 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1410 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1413 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1414 u64 utimescaled, stimescaled;
1417 struct prev_cputime prev_cputime;
1418 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1419 seqcount_t vtime_seqcount;
1420 unsigned long long vtime_snap;
1422 /* Task is sleeping or running in a CPU with VTIME inactive */
1424 /* Task runs in userspace in a CPU with VTIME active */
1426 /* Task runs in kernelspace in a CPU with VTIME active */
1428 } vtime_snap_whence;
1431 #ifdef CONFIG_NO_HZ_FULL
1432 atomic_t tick_dep_mask;
1434 unsigned long nvcsw, nivcsw; /* context switch counts */
1435 u64 start_time; /* monotonic time in nsec */
1436 u64 real_start_time; /* boot based time in nsec */
1437 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1438 unsigned long min_flt, maj_flt;
1440 #ifdef CONFIG_POSIX_TIMERS
1441 struct task_cputime cputime_expires;
1442 struct list_head cpu_timers[3];
1445 /* process credentials */
1446 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1447 const struct cred __rcu *real_cred; /* objective and real subjective task
1448 * credentials (COW) */
1449 const struct cred __rcu *cred; /* effective (overridable) subjective task
1450 * credentials (COW) */
1451 char comm[TASK_COMM_LEN]; /* executable name excluding path
1452 - access with [gs]et_task_comm (which lock
1453 it with task_lock())
1454 - initialized normally by setup_new_exec */
1455 /* file system info */
1456 struct nameidata *nameidata;
1457 #ifdef CONFIG_SYSVIPC
1459 struct sysv_sem sysvsem;
1460 struct sysv_shm sysvshm;
1462 #ifdef CONFIG_DETECT_HUNG_TASK
1463 /* hung task detection */
1464 unsigned long last_switch_count;
1466 /* filesystem information */
1467 struct fs_struct *fs;
1468 /* open file information */
1469 struct files_struct *files;
1471 struct nsproxy *nsproxy;
1472 /* signal handlers */
1473 struct signal_struct *signal;
1474 struct sighand_struct *sighand;
1476 sigset_t blocked, real_blocked;
1477 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1478 struct sigpending pending;
1480 unsigned long sas_ss_sp;
1482 unsigned sas_ss_flags;
1484 struct callback_head *task_works;
1486 struct audit_context *audit_context;
1487 #ifdef CONFIG_AUDITSYSCALL
1489 unsigned int sessionid;
1491 struct seccomp seccomp;
1493 /* Thread group tracking */
1496 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1498 spinlock_t alloc_lock;
1500 /* Protection of the PI data structures: */
1501 raw_spinlock_t pi_lock;
1503 struct wake_q_node wake_q;
1505 #ifdef CONFIG_RT_MUTEXES
1506 /* PI waiters blocked on a rt_mutex held by this task */
1507 struct rb_root pi_waiters;
1508 struct rb_node *pi_waiters_leftmost;
1509 /* Deadlock detection and priority inheritance handling */
1510 struct rt_mutex_waiter *pi_blocked_on;
1513 #ifdef CONFIG_DEBUG_MUTEXES
1514 /* mutex deadlock detection */
1515 struct mutex_waiter *blocked_on;
1517 #ifdef CONFIG_TRACE_IRQFLAGS
1518 unsigned int irq_events;
1519 unsigned long hardirq_enable_ip;
1520 unsigned long hardirq_disable_ip;
1521 unsigned int hardirq_enable_event;
1522 unsigned int hardirq_disable_event;
1523 int hardirqs_enabled;
1524 int hardirq_context;
1525 unsigned long softirq_disable_ip;
1526 unsigned long softirq_enable_ip;
1527 unsigned int softirq_disable_event;
1528 unsigned int softirq_enable_event;
1529 int softirqs_enabled;
1530 int softirq_context;
1532 #ifdef CONFIG_LOCKDEP
1533 # define MAX_LOCK_DEPTH 48UL
1536 unsigned int lockdep_recursion;
1537 struct held_lock held_locks[MAX_LOCK_DEPTH];
1538 gfp_t lockdep_reclaim_gfp;
1541 unsigned int in_ubsan;
1544 /* journalling filesystem info */
1547 /* stacked block device info */
1548 struct bio_list *bio_list;
1551 /* stack plugging */
1552 struct blk_plug *plug;
1556 struct reclaim_state *reclaim_state;
1558 struct backing_dev_info *backing_dev_info;
1560 struct io_context *io_context;
1562 unsigned long ptrace_message;
1563 siginfo_t *last_siginfo; /* For ptrace use. */
1564 struct task_io_accounting ioac;
1565 #if defined(CONFIG_TASK_XACCT)
1566 u64 acct_rss_mem1; /* accumulated rss usage */
1567 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1568 u64 acct_timexpd; /* stime + utime since last update */
1570 #ifdef CONFIG_CPUSETS
1571 nodemask_t mems_allowed; /* Protected by alloc_lock */
1572 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1573 int cpuset_mem_spread_rotor;
1574 int cpuset_slab_spread_rotor;
1576 #ifdef CONFIG_CGROUPS
1577 /* Control Group info protected by css_set_lock */
1578 struct css_set __rcu *cgroups;
1579 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1580 struct list_head cg_list;
1582 #ifdef CONFIG_INTEL_RDT_A
1586 struct robust_list_head __user *robust_list;
1587 #ifdef CONFIG_COMPAT
1588 struct compat_robust_list_head __user *compat_robust_list;
1590 struct list_head pi_state_list;
1591 struct futex_pi_state *pi_state_cache;
1593 #ifdef CONFIG_PERF_EVENTS
1594 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1595 struct mutex perf_event_mutex;
1596 struct list_head perf_event_list;
1598 #ifdef CONFIG_DEBUG_PREEMPT
1599 unsigned long preempt_disable_ip;
1602 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1604 short pref_node_fork;
1606 #ifdef CONFIG_NUMA_BALANCING
1608 unsigned int numa_scan_period;
1609 unsigned int numa_scan_period_max;
1610 int numa_preferred_nid;
1611 unsigned long numa_migrate_retry;
1612 u64 node_stamp; /* migration stamp */
1613 u64 last_task_numa_placement;
1614 u64 last_sum_exec_runtime;
1615 struct callback_head numa_work;
1617 struct list_head numa_entry;
1618 struct numa_group *numa_group;
1621 * numa_faults is an array split into four regions:
1622 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1623 * in this precise order.
1625 * faults_memory: Exponential decaying average of faults on a per-node
1626 * basis. Scheduling placement decisions are made based on these
1627 * counts. The values remain static for the duration of a PTE scan.
1628 * faults_cpu: Track the nodes the process was running on when a NUMA
1629 * hinting fault was incurred.
1630 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1631 * during the current scan window. When the scan completes, the counts
1632 * in faults_memory and faults_cpu decay and these values are copied.
1634 unsigned long *numa_faults;
1635 unsigned long total_numa_faults;
1638 * numa_faults_locality tracks if faults recorded during the last
1639 * scan window were remote/local or failed to migrate. The task scan
1640 * period is adapted based on the locality of the faults with different
1641 * weights depending on whether they were shared or private faults
1643 unsigned long numa_faults_locality[3];
1645 unsigned long numa_pages_migrated;
1646 #endif /* CONFIG_NUMA_BALANCING */
1648 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1649 struct tlbflush_unmap_batch tlb_ubc;
1652 struct rcu_head rcu;
1655 * cache last used pipe for splice
1657 struct pipe_inode_info *splice_pipe;
1659 struct page_frag task_frag;
1661 #ifdef CONFIG_TASK_DELAY_ACCT
1662 struct task_delay_info *delays;
1664 #ifdef CONFIG_FAULT_INJECTION
1668 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1669 * balance_dirty_pages() for some dirty throttling pause
1672 int nr_dirtied_pause;
1673 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1675 #ifdef CONFIG_LATENCYTOP
1676 int latency_record_count;
1677 struct latency_record latency_record[LT_SAVECOUNT];
1680 * time slack values; these are used to round up poll() and
1681 * select() etc timeout values. These are in nanoseconds.
1684 u64 default_timer_slack_ns;
1687 unsigned int kasan_depth;
1689 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1690 /* Index of current stored address in ret_stack */
1692 /* Stack of return addresses for return function tracing */
1693 struct ftrace_ret_stack *ret_stack;
1694 /* time stamp for last schedule */
1695 unsigned long long ftrace_timestamp;
1697 * Number of functions that haven't been traced
1698 * because of depth overrun.
1700 atomic_t trace_overrun;
1701 /* Pause for the tracing */
1702 atomic_t tracing_graph_pause;
1704 #ifdef CONFIG_TRACING
1705 /* state flags for use by tracers */
1706 unsigned long trace;
1707 /* bitmask and counter of trace recursion */
1708 unsigned long trace_recursion;
1709 #endif /* CONFIG_TRACING */
1711 /* Coverage collection mode enabled for this task (0 if disabled). */
1712 enum kcov_mode kcov_mode;
1713 /* Size of the kcov_area. */
1715 /* Buffer for coverage collection. */
1717 /* kcov desciptor wired with this task or NULL. */
1721 struct mem_cgroup *memcg_in_oom;
1722 gfp_t memcg_oom_gfp_mask;
1723 int memcg_oom_order;
1725 /* number of pages to reclaim on returning to userland */
1726 unsigned int memcg_nr_pages_over_high;
1728 #ifdef CONFIG_UPROBES
1729 struct uprobe_task *utask;
1731 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1732 unsigned int sequential_io;
1733 unsigned int sequential_io_avg;
1735 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1736 unsigned long task_state_change;
1738 int pagefault_disabled;
1740 struct task_struct *oom_reaper_list;
1742 #ifdef CONFIG_VMAP_STACK
1743 struct vm_struct *stack_vm_area;
1745 #ifdef CONFIG_THREAD_INFO_IN_TASK
1746 /* A live task holds one reference. */
1747 atomic_t stack_refcount;
1749 /* CPU-specific state of this task */
1750 struct thread_struct thread;
1752 * WARNING: on x86, 'thread_struct' contains a variable-sized
1753 * structure. It *MUST* be at the end of 'task_struct'.
1755 * Do not put anything below here!
1759 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1760 extern int arch_task_struct_size __read_mostly;
1762 # define arch_task_struct_size (sizeof(struct task_struct))
1765 #ifdef CONFIG_VMAP_STACK
1766 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1768 return t->stack_vm_area;
1771 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
1777 #define TNF_MIGRATED 0x01
1778 #define TNF_NO_GROUP 0x02
1779 #define TNF_SHARED 0x04
1780 #define TNF_FAULT_LOCAL 0x08
1781 #define TNF_MIGRATE_FAIL 0x10
1783 static inline bool in_vfork(struct task_struct *tsk)
1788 * need RCU to access ->real_parent if CLONE_VM was used along with
1791 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
1794 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
1795 * ->real_parent is not necessarily the task doing vfork(), so in
1796 * theory we can't rely on task_lock() if we want to dereference it.
1798 * And in this case we can't trust the real_parent->mm == tsk->mm
1799 * check, it can be false negative. But we do not care, if init or
1800 * another oom-unkillable task does this it should blame itself.
1803 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
1809 #ifdef CONFIG_NUMA_BALANCING
1810 extern void task_numa_fault(int last_node, int node, int pages, int flags);
1811 extern pid_t task_numa_group_id(struct task_struct *p);
1812 extern void set_numabalancing_state(bool enabled);
1813 extern void task_numa_free(struct task_struct *p);
1814 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
1815 int src_nid, int dst_cpu);
1817 static inline void task_numa_fault(int last_node, int node, int pages,
1821 static inline pid_t task_numa_group_id(struct task_struct *p)
1825 static inline void set_numabalancing_state(bool enabled)
1828 static inline void task_numa_free(struct task_struct *p)
1831 static inline bool should_numa_migrate_memory(struct task_struct *p,
1832 struct page *page, int src_nid, int dst_cpu)
1838 static inline struct pid *task_pid(struct task_struct *task)
1840 return task->pids[PIDTYPE_PID].pid;
1843 static inline struct pid *task_tgid(struct task_struct *task)
1845 return task->group_leader->pids[PIDTYPE_PID].pid;
1849 * Without tasklist or rcu lock it is not safe to dereference
1850 * the result of task_pgrp/task_session even if task == current,
1851 * we can race with another thread doing sys_setsid/sys_setpgid.
1853 static inline struct pid *task_pgrp(struct task_struct *task)
1855 return task->group_leader->pids[PIDTYPE_PGID].pid;
1858 static inline struct pid *task_session(struct task_struct *task)
1860 return task->group_leader->pids[PIDTYPE_SID].pid;
1863 struct pid_namespace;
1866 * the helpers to get the task's different pids as they are seen
1867 * from various namespaces
1869 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1870 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1872 * task_xid_nr_ns() : id seen from the ns specified;
1874 * set_task_vxid() : assigns a virtual id to a task;
1876 * see also pid_nr() etc in include/linux/pid.h
1878 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1879 struct pid_namespace *ns);
1881 static inline pid_t task_pid_nr(struct task_struct *tsk)
1886 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1887 struct pid_namespace *ns)
1889 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1892 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1894 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1898 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1903 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1905 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1907 return pid_vnr(task_tgid(tsk));
1911 static inline int pid_alive(const struct task_struct *p);
1912 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1918 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1924 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1926 return task_ppid_nr_ns(tsk, &init_pid_ns);
1929 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1930 struct pid_namespace *ns)
1932 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1935 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1937 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1941 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1942 struct pid_namespace *ns)
1944 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1947 static inline pid_t task_session_vnr(struct task_struct *tsk)
1949 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1952 /* obsolete, do not use */
1953 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1955 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1959 * pid_alive - check that a task structure is not stale
1960 * @p: Task structure to be checked.
1962 * Test if a process is not yet dead (at most zombie state)
1963 * If pid_alive fails, then pointers within the task structure
1964 * can be stale and must not be dereferenced.
1966 * Return: 1 if the process is alive. 0 otherwise.
1968 static inline int pid_alive(const struct task_struct *p)
1970 return p->pids[PIDTYPE_PID].pid != NULL;
1974 * is_global_init - check if a task structure is init. Since init
1975 * is free to have sub-threads we need to check tgid.
1976 * @tsk: Task structure to be checked.
1978 * Check if a task structure is the first user space task the kernel created.
1980 * Return: 1 if the task structure is init. 0 otherwise.
1982 static inline int is_global_init(struct task_struct *tsk)
1984 return task_tgid_nr(tsk) == 1;
1987 extern struct pid *cad_pid;
1989 extern void free_task(struct task_struct *tsk);
1990 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
1992 extern void __put_task_struct(struct task_struct *t);
1994 static inline void put_task_struct(struct task_struct *t)
1996 if (atomic_dec_and_test(&t->usage))
1997 __put_task_struct(t);
2000 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2001 struct task_struct *try_get_task_struct(struct task_struct **ptask);
2003 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2004 extern void task_cputime(struct task_struct *t,
2005 u64 *utime, u64 *stime);
2006 extern u64 task_gtime(struct task_struct *t);
2008 static inline void task_cputime(struct task_struct *t,
2009 u64 *utime, u64 *stime)
2015 static inline u64 task_gtime(struct task_struct *t)
2021 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
2022 static inline void task_cputime_scaled(struct task_struct *t,
2026 *utimescaled = t->utimescaled;
2027 *stimescaled = t->stimescaled;
2030 static inline void task_cputime_scaled(struct task_struct *t,
2034 task_cputime(t, utimescaled, stimescaled);
2038 extern void task_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2039 extern void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2044 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
2045 #define PF_EXITING 0x00000004 /* getting shut down */
2046 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2047 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2048 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2049 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2050 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2051 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2052 #define PF_DUMPCORE 0x00000200 /* dumped core */
2053 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2054 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2055 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2056 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2057 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2058 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2059 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2060 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2061 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2062 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2063 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2064 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2065 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2066 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2067 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2068 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2069 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2070 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2071 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2074 * Only the _current_ task can read/write to tsk->flags, but other
2075 * tasks can access tsk->flags in readonly mode for example
2076 * with tsk_used_math (like during threaded core dumping).
2077 * There is however an exception to this rule during ptrace
2078 * or during fork: the ptracer task is allowed to write to the
2079 * child->flags of its traced child (same goes for fork, the parent
2080 * can write to the child->flags), because we're guaranteed the
2081 * child is not running and in turn not changing child->flags
2082 * at the same time the parent does it.
2084 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2085 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2086 #define clear_used_math() clear_stopped_child_used_math(current)
2087 #define set_used_math() set_stopped_child_used_math(current)
2088 #define conditional_stopped_child_used_math(condition, child) \
2089 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2090 #define conditional_used_math(condition) \
2091 conditional_stopped_child_used_math(condition, current)
2092 #define copy_to_stopped_child_used_math(child) \
2093 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2094 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2095 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2096 #define used_math() tsk_used_math(current)
2098 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2099 * __GFP_FS is also cleared as it implies __GFP_IO.
2101 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2103 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2104 flags &= ~(__GFP_IO | __GFP_FS);
2108 static inline unsigned int memalloc_noio_save(void)
2110 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2111 current->flags |= PF_MEMALLOC_NOIO;
2115 static inline void memalloc_noio_restore(unsigned int flags)
2117 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2120 /* Per-process atomic flags. */
2121 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2122 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2123 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2124 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2127 #define TASK_PFA_TEST(name, func) \
2128 static inline bool task_##func(struct task_struct *p) \
2129 { return test_bit(PFA_##name, &p->atomic_flags); }
2130 #define TASK_PFA_SET(name, func) \
2131 static inline void task_set_##func(struct task_struct *p) \
2132 { set_bit(PFA_##name, &p->atomic_flags); }
2133 #define TASK_PFA_CLEAR(name, func) \
2134 static inline void task_clear_##func(struct task_struct *p) \
2135 { clear_bit(PFA_##name, &p->atomic_flags); }
2137 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2138 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2140 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2141 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2142 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2144 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2145 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2146 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2148 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2149 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2152 * task->jobctl flags
2154 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2156 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2157 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2158 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2159 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2160 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2161 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2162 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2164 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2165 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2166 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2167 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2168 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2169 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2170 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2172 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2173 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2175 extern bool task_set_jobctl_pending(struct task_struct *task,
2176 unsigned long mask);
2177 extern void task_clear_jobctl_trapping(struct task_struct *task);
2178 extern void task_clear_jobctl_pending(struct task_struct *task,
2179 unsigned long mask);
2181 static inline void rcu_copy_process(struct task_struct *p)
2183 #ifdef CONFIG_PREEMPT_RCU
2184 p->rcu_read_lock_nesting = 0;
2185 p->rcu_read_unlock_special.s = 0;
2186 p->rcu_blocked_node = NULL;
2187 INIT_LIST_HEAD(&p->rcu_node_entry);
2188 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2189 #ifdef CONFIG_TASKS_RCU
2190 p->rcu_tasks_holdout = false;
2191 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2192 p->rcu_tasks_idle_cpu = -1;
2193 #endif /* #ifdef CONFIG_TASKS_RCU */
2196 static inline void tsk_restore_flags(struct task_struct *task,
2197 unsigned long orig_flags, unsigned long flags)
2199 task->flags &= ~flags;
2200 task->flags |= orig_flags & flags;
2203 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2204 const struct cpumask *trial);
2205 extern int task_can_attach(struct task_struct *p,
2206 const struct cpumask *cs_cpus_allowed);
2208 extern void do_set_cpus_allowed(struct task_struct *p,
2209 const struct cpumask *new_mask);
2211 extern int set_cpus_allowed_ptr(struct task_struct *p,
2212 const struct cpumask *new_mask);
2214 static inline void do_set_cpus_allowed(struct task_struct *p,
2215 const struct cpumask *new_mask)
2218 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2219 const struct cpumask *new_mask)
2221 if (!cpumask_test_cpu(0, new_mask))
2227 #ifdef CONFIG_NO_HZ_COMMON
2228 void calc_load_enter_idle(void);
2229 void calc_load_exit_idle(void);
2231 static inline void calc_load_enter_idle(void) { }
2232 static inline void calc_load_exit_idle(void) { }
2233 #endif /* CONFIG_NO_HZ_COMMON */
2235 #ifndef cpu_relax_yield
2236 #define cpu_relax_yield() cpu_relax()
2240 * Do not use outside of architecture code which knows its limitations.
2242 * sched_clock() has no promise of monotonicity or bounded drift between
2243 * CPUs, use (which you should not) requires disabling IRQs.
2245 * Please use one of the three interfaces below.
2247 extern unsigned long long notrace sched_clock(void);
2249 * See the comment in kernel/sched/clock.c
2251 extern u64 running_clock(void);
2252 extern u64 sched_clock_cpu(int cpu);
2255 extern void sched_clock_init(void);
2257 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2258 static inline void sched_clock_init_late(void)
2262 static inline void sched_clock_tick(void)
2266 static inline void clear_sched_clock_stable(void)
2270 static inline void sched_clock_idle_sleep_event(void)
2274 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2278 static inline u64 cpu_clock(int cpu)
2280 return sched_clock();
2283 static inline u64 local_clock(void)
2285 return sched_clock();
2288 extern void sched_clock_init_late(void);
2290 * Architectures can set this to 1 if they have specified
2291 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2292 * but then during bootup it turns out that sched_clock()
2293 * is reliable after all:
2295 extern int sched_clock_stable(void);
2296 extern void clear_sched_clock_stable(void);
2298 extern void sched_clock_tick(void);
2299 extern void sched_clock_idle_sleep_event(void);
2300 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2303 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2304 * time source that is monotonic per cpu argument and has bounded drift
2307 * ######################### BIG FAT WARNING ##########################
2308 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2309 * # go backwards !! #
2310 * ####################################################################
2312 static inline u64 cpu_clock(int cpu)
2314 return sched_clock_cpu(cpu);
2317 static inline u64 local_clock(void)
2319 return sched_clock_cpu(raw_smp_processor_id());
2323 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2325 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2326 * The reason for this explicit opt-in is not to have perf penalty with
2327 * slow sched_clocks.
2329 extern void enable_sched_clock_irqtime(void);
2330 extern void disable_sched_clock_irqtime(void);
2332 static inline void enable_sched_clock_irqtime(void) {}
2333 static inline void disable_sched_clock_irqtime(void) {}
2336 extern unsigned long long
2337 task_sched_runtime(struct task_struct *task);
2339 /* sched_exec is called by processes performing an exec */
2341 extern void sched_exec(void);
2343 #define sched_exec() {}
2346 extern void sched_clock_idle_sleep_event(void);
2347 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2349 #ifdef CONFIG_HOTPLUG_CPU
2350 extern void idle_task_exit(void);
2352 static inline void idle_task_exit(void) {}
2355 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2356 extern void wake_up_nohz_cpu(int cpu);
2358 static inline void wake_up_nohz_cpu(int cpu) { }
2361 #ifdef CONFIG_NO_HZ_FULL
2362 extern u64 scheduler_tick_max_deferment(void);
2365 #ifdef CONFIG_SCHED_AUTOGROUP
2366 extern void sched_autogroup_create_attach(struct task_struct *p);
2367 extern void sched_autogroup_detach(struct task_struct *p);
2368 extern void sched_autogroup_fork(struct signal_struct *sig);
2369 extern void sched_autogroup_exit(struct signal_struct *sig);
2370 extern void sched_autogroup_exit_task(struct task_struct *p);
2371 #ifdef CONFIG_PROC_FS
2372 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2373 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2376 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2377 static inline void sched_autogroup_detach(struct task_struct *p) { }
2378 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2379 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2380 static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2383 extern int yield_to(struct task_struct *p, bool preempt);
2384 extern void set_user_nice(struct task_struct *p, long nice);
2385 extern int task_prio(const struct task_struct *p);
2387 * task_nice - return the nice value of a given task.
2388 * @p: the task in question.
2390 * Return: The nice value [ -20 ... 0 ... 19 ].
2392 static inline int task_nice(const struct task_struct *p)
2394 return PRIO_TO_NICE((p)->static_prio);
2396 extern int can_nice(const struct task_struct *p, const int nice);
2397 extern int task_curr(const struct task_struct *p);
2398 extern int idle_cpu(int cpu);
2399 extern int sched_setscheduler(struct task_struct *, int,
2400 const struct sched_param *);
2401 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2402 const struct sched_param *);
2403 extern int sched_setattr(struct task_struct *,
2404 const struct sched_attr *);
2405 extern struct task_struct *idle_task(int cpu);
2407 * is_idle_task - is the specified task an idle task?
2408 * @p: the task in question.
2410 * Return: 1 if @p is an idle task. 0 otherwise.
2412 static inline bool is_idle_task(const struct task_struct *p)
2414 return !!(p->flags & PF_IDLE);
2416 extern struct task_struct *curr_task(int cpu);
2417 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2421 union thread_union {
2422 #ifndef CONFIG_THREAD_INFO_IN_TASK
2423 struct thread_info thread_info;
2425 unsigned long stack[THREAD_SIZE/sizeof(long)];
2428 #ifndef __HAVE_ARCH_KSTACK_END
2429 static inline int kstack_end(void *addr)
2431 /* Reliable end of stack detection:
2432 * Some APM bios versions misalign the stack
2434 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2438 extern union thread_union init_thread_union;
2439 extern struct task_struct init_task;
2441 extern struct mm_struct init_mm;
2443 extern struct pid_namespace init_pid_ns;
2446 * find a task by one of its numerical ids
2448 * find_task_by_pid_ns():
2449 * finds a task by its pid in the specified namespace
2450 * find_task_by_vpid():
2451 * finds a task by its virtual pid
2453 * see also find_vpid() etc in include/linux/pid.h
2456 extern struct task_struct *find_task_by_vpid(pid_t nr);
2457 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2458 struct pid_namespace *ns);
2460 /* per-UID process charging. */
2461 extern struct user_struct * alloc_uid(kuid_t);
2462 static inline struct user_struct *get_uid(struct user_struct *u)
2464 atomic_inc(&u->__count);
2467 extern void free_uid(struct user_struct *);
2469 #include <asm/current.h>
2471 extern void xtime_update(unsigned long ticks);
2473 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2474 extern int wake_up_process(struct task_struct *tsk);
2475 extern void wake_up_new_task(struct task_struct *tsk);
2477 extern void kick_process(struct task_struct *tsk);
2479 static inline void kick_process(struct task_struct *tsk) { }
2481 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2482 extern void sched_dead(struct task_struct *p);
2484 extern void proc_caches_init(void);
2485 extern void flush_signals(struct task_struct *);
2486 extern void ignore_signals(struct task_struct *);
2487 extern void flush_signal_handlers(struct task_struct *, int force_default);
2488 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2490 static inline int kernel_dequeue_signal(siginfo_t *info)
2492 struct task_struct *tsk = current;
2496 spin_lock_irq(&tsk->sighand->siglock);
2497 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2498 spin_unlock_irq(&tsk->sighand->siglock);
2503 static inline void kernel_signal_stop(void)
2505 spin_lock_irq(¤t->sighand->siglock);
2506 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2507 __set_current_state(TASK_STOPPED);
2508 spin_unlock_irq(¤t->sighand->siglock);
2513 extern void release_task(struct task_struct * p);
2514 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2515 extern int force_sigsegv(int, struct task_struct *);
2516 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2517 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2518 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2519 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2520 const struct cred *, u32);
2521 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2522 extern int kill_pid(struct pid *pid, int sig, int priv);
2523 extern int kill_proc_info(int, struct siginfo *, pid_t);
2524 extern __must_check bool do_notify_parent(struct task_struct *, int);
2525 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2526 extern void force_sig(int, struct task_struct *);
2527 extern int send_sig(int, struct task_struct *, int);
2528 extern int zap_other_threads(struct task_struct *p);
2529 extern struct sigqueue *sigqueue_alloc(void);
2530 extern void sigqueue_free(struct sigqueue *);
2531 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2532 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2534 #ifdef TIF_RESTORE_SIGMASK
2536 * Legacy restore_sigmask accessors. These are inefficient on
2537 * SMP architectures because they require atomic operations.
2541 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2543 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2544 * will run before returning to user mode, to process the flag. For
2545 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2546 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2547 * arch code will notice on return to user mode, in case those bits
2548 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2549 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2551 static inline void set_restore_sigmask(void)
2553 set_thread_flag(TIF_RESTORE_SIGMASK);
2554 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2556 static inline void clear_restore_sigmask(void)
2558 clear_thread_flag(TIF_RESTORE_SIGMASK);
2560 static inline bool test_restore_sigmask(void)
2562 return test_thread_flag(TIF_RESTORE_SIGMASK);
2564 static inline bool test_and_clear_restore_sigmask(void)
2566 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2569 #else /* TIF_RESTORE_SIGMASK */
2571 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2572 static inline void set_restore_sigmask(void)
2574 current->restore_sigmask = true;
2575 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2577 static inline void clear_restore_sigmask(void)
2579 current->restore_sigmask = false;
2581 static inline bool test_restore_sigmask(void)
2583 return current->restore_sigmask;
2585 static inline bool test_and_clear_restore_sigmask(void)
2587 if (!current->restore_sigmask)
2589 current->restore_sigmask = false;
2594 static inline void restore_saved_sigmask(void)
2596 if (test_and_clear_restore_sigmask())
2597 __set_current_blocked(¤t->saved_sigmask);
2600 static inline sigset_t *sigmask_to_save(void)
2602 sigset_t *res = ¤t->blocked;
2603 if (unlikely(test_restore_sigmask()))
2604 res = ¤t->saved_sigmask;
2608 static inline int kill_cad_pid(int sig, int priv)
2610 return kill_pid(cad_pid, sig, priv);
2613 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2614 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2615 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2616 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2619 * True if we are on the alternate signal stack.
2621 static inline int on_sig_stack(unsigned long sp)
2624 * If the signal stack is SS_AUTODISARM then, by construction, we
2625 * can't be on the signal stack unless user code deliberately set
2626 * SS_AUTODISARM when we were already on it.
2628 * This improves reliability: if user state gets corrupted such that
2629 * the stack pointer points very close to the end of the signal stack,
2630 * then this check will enable the signal to be handled anyway.
2632 if (current->sas_ss_flags & SS_AUTODISARM)
2635 #ifdef CONFIG_STACK_GROWSUP
2636 return sp >= current->sas_ss_sp &&
2637 sp - current->sas_ss_sp < current->sas_ss_size;
2639 return sp > current->sas_ss_sp &&
2640 sp - current->sas_ss_sp <= current->sas_ss_size;
2644 static inline int sas_ss_flags(unsigned long sp)
2646 if (!current->sas_ss_size)
2649 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2652 static inline void sas_ss_reset(struct task_struct *p)
2656 p->sas_ss_flags = SS_DISABLE;
2659 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2661 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2662 #ifdef CONFIG_STACK_GROWSUP
2663 return current->sas_ss_sp;
2665 return current->sas_ss_sp + current->sas_ss_size;
2671 * Routines for handling mm_structs
2673 extern struct mm_struct * mm_alloc(void);
2676 * mmgrab() - Pin a &struct mm_struct.
2677 * @mm: The &struct mm_struct to pin.
2679 * Make sure that @mm will not get freed even after the owning task
2680 * exits. This doesn't guarantee that the associated address space
2681 * will still exist later on and mmget_not_zero() has to be used before
2684 * This is a preferred way to to pin @mm for a longer/unbounded amount
2687 * Use mmdrop() to release the reference acquired by mmgrab().
2689 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2690 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2692 static inline void mmgrab(struct mm_struct *mm)
2694 atomic_inc(&mm->mm_count);
2697 /* mmdrop drops the mm and the page tables */
2698 extern void __mmdrop(struct mm_struct *);
2699 static inline void mmdrop(struct mm_struct *mm)
2701 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2705 static inline void mmdrop_async_fn(struct work_struct *work)
2707 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2711 static inline void mmdrop_async(struct mm_struct *mm)
2713 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2714 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2715 schedule_work(&mm->async_put_work);
2720 * mmget() - Pin the address space associated with a &struct mm_struct.
2721 * @mm: The address space to pin.
2723 * Make sure that the address space of the given &struct mm_struct doesn't
2724 * go away. This does not protect against parts of the address space being
2725 * modified or freed, however.
2727 * Never use this function to pin this address space for an
2728 * unbounded/indefinite amount of time.
2730 * Use mmput() to release the reference acquired by mmget().
2732 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2733 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2735 static inline void mmget(struct mm_struct *mm)
2737 atomic_inc(&mm->mm_users);
2740 static inline bool mmget_not_zero(struct mm_struct *mm)
2742 return atomic_inc_not_zero(&mm->mm_users);
2745 /* mmput gets rid of the mappings and all user-space */
2746 extern void mmput(struct mm_struct *);
2748 /* same as above but performs the slow path from the async context. Can
2749 * be called from the atomic context as well
2751 extern void mmput_async(struct mm_struct *);
2754 /* Grab a reference to a task's mm, if it is not already going away */
2755 extern struct mm_struct *get_task_mm(struct task_struct *task);
2757 * Grab a reference to a task's mm, if it is not already going away
2758 * and ptrace_may_access with the mode parameter passed to it
2761 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2762 /* Remove the current tasks stale references to the old mm_struct */
2763 extern void mm_release(struct task_struct *, struct mm_struct *);
2765 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2766 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2767 struct task_struct *, unsigned long);
2769 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2770 struct task_struct *);
2772 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2773 * via pt_regs, so ignore the tls argument passed via C. */
2774 static inline int copy_thread_tls(
2775 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2776 struct task_struct *p, unsigned long tls)
2778 return copy_thread(clone_flags, sp, arg, p);
2781 extern void flush_thread(void);
2783 #ifdef CONFIG_HAVE_EXIT_THREAD
2784 extern void exit_thread(struct task_struct *tsk);
2786 static inline void exit_thread(struct task_struct *tsk)
2791 extern void exit_files(struct task_struct *);
2792 extern void __cleanup_sighand(struct sighand_struct *);
2794 extern void exit_itimers(struct signal_struct *);
2795 extern void flush_itimer_signals(void);
2797 extern void do_group_exit(int);
2799 extern int do_execve(struct filename *,
2800 const char __user * const __user *,
2801 const char __user * const __user *);
2802 extern int do_execveat(int, struct filename *,
2803 const char __user * const __user *,
2804 const char __user * const __user *,
2806 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
2807 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2808 struct task_struct *fork_idle(int);
2809 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2811 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2812 static inline void set_task_comm(struct task_struct *tsk, const char *from)
2814 __set_task_comm(tsk, from, false);
2816 extern char *get_task_comm(char *to, struct task_struct *tsk);
2819 void scheduler_ipi(void);
2820 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2822 static inline void scheduler_ipi(void) { }
2823 static inline unsigned long wait_task_inactive(struct task_struct *p,
2830 #define tasklist_empty() \
2831 list_empty(&init_task.tasks)
2833 #define next_task(p) \
2834 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2836 #define for_each_process(p) \
2837 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2839 extern bool current_is_single_threaded(void);
2842 * Careful: do_each_thread/while_each_thread is a double loop so
2843 * 'break' will not work as expected - use goto instead.
2845 #define do_each_thread(g, t) \
2846 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
2848 #define while_each_thread(g, t) \
2849 while ((t = next_thread(t)) != g)
2851 #define __for_each_thread(signal, t) \
2852 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
2854 #define for_each_thread(p, t) \
2855 __for_each_thread((p)->signal, t)
2857 /* Careful: this is a double loop, 'break' won't work as expected. */
2858 #define for_each_process_thread(p, t) \
2859 for_each_process(p) for_each_thread(p, t)
2861 typedef int (*proc_visitor)(struct task_struct *p, void *data);
2862 void walk_process_tree(struct task_struct *top, proc_visitor, void *);
2864 static inline int get_nr_threads(struct task_struct *tsk)
2866 return tsk->signal->nr_threads;
2869 static inline bool thread_group_leader(struct task_struct *p)
2871 return p->exit_signal >= 0;
2874 /* Do to the insanities of de_thread it is possible for a process
2875 * to have the pid of the thread group leader without actually being
2876 * the thread group leader. For iteration through the pids in proc
2877 * all we care about is that we have a task with the appropriate
2878 * pid, we don't actually care if we have the right task.
2880 static inline bool has_group_leader_pid(struct task_struct *p)
2882 return task_pid(p) == p->signal->leader_pid;
2886 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
2888 return p1->signal == p2->signal;
2891 static inline struct task_struct *next_thread(const struct task_struct *p)
2893 return list_entry_rcu(p->thread_group.next,
2894 struct task_struct, thread_group);
2897 static inline int thread_group_empty(struct task_struct *p)
2899 return list_empty(&p->thread_group);
2902 #define delay_group_leader(p) \
2903 (thread_group_leader(p) && !thread_group_empty(p))
2906 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
2907 * subscriptions and synchronises with wait4(). Also used in procfs. Also
2908 * pins the final release of task.io_context. Also protects ->cpuset and
2909 * ->cgroup.subsys[]. And ->vfork_done.
2911 * Nests both inside and outside of read_lock(&tasklist_lock).
2912 * It must not be nested with write_lock_irq(&tasklist_lock),
2913 * neither inside nor outside.
2915 static inline void task_lock(struct task_struct *p)
2917 spin_lock(&p->alloc_lock);
2920 static inline void task_unlock(struct task_struct *p)
2922 spin_unlock(&p->alloc_lock);
2925 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
2926 unsigned long *flags);
2928 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
2929 unsigned long *flags)
2931 struct sighand_struct *ret;
2933 ret = __lock_task_sighand(tsk, flags);
2934 (void)__cond_lock(&tsk->sighand->siglock, ret);
2938 static inline void unlock_task_sighand(struct task_struct *tsk,
2939 unsigned long *flags)
2941 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
2944 #ifdef CONFIG_THREAD_INFO_IN_TASK
2946 static inline struct thread_info *task_thread_info(struct task_struct *task)
2948 return &task->thread_info;
2952 * When accessing the stack of a non-current task that might exit, use
2953 * try_get_task_stack() instead. task_stack_page will return a pointer
2954 * that could get freed out from under you.
2956 static inline void *task_stack_page(const struct task_struct *task)
2961 #define setup_thread_stack(new,old) do { } while(0)
2963 static inline unsigned long *end_of_stack(const struct task_struct *task)
2968 #elif !defined(__HAVE_THREAD_FUNCTIONS)
2970 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
2971 #define task_stack_page(task) ((void *)(task)->stack)
2973 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
2975 *task_thread_info(p) = *task_thread_info(org);
2976 task_thread_info(p)->task = p;
2980 * Return the address of the last usable long on the stack.
2982 * When the stack grows down, this is just above the thread
2983 * info struct. Going any lower will corrupt the threadinfo.
2985 * When the stack grows up, this is the highest address.
2986 * Beyond that position, we corrupt data on the next page.
2988 static inline unsigned long *end_of_stack(struct task_struct *p)
2990 #ifdef CONFIG_STACK_GROWSUP
2991 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
2993 return (unsigned long *)(task_thread_info(p) + 1);
2999 #ifdef CONFIG_THREAD_INFO_IN_TASK
3000 static inline void *try_get_task_stack(struct task_struct *tsk)
3002 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3003 task_stack_page(tsk) : NULL;
3006 extern void put_task_stack(struct task_struct *tsk);
3008 static inline void *try_get_task_stack(struct task_struct *tsk)
3010 return task_stack_page(tsk);
3013 static inline void put_task_stack(struct task_struct *tsk) {}
3016 #define task_stack_end_corrupted(task) \
3017 (*(end_of_stack(task)) != STACK_END_MAGIC)
3019 static inline int object_is_on_stack(void *obj)
3021 void *stack = task_stack_page(current);
3023 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3026 extern void thread_stack_cache_init(void);
3028 #ifdef CONFIG_DEBUG_STACK_USAGE
3029 static inline unsigned long stack_not_used(struct task_struct *p)
3031 unsigned long *n = end_of_stack(p);
3033 do { /* Skip over canary */
3034 # ifdef CONFIG_STACK_GROWSUP
3041 # ifdef CONFIG_STACK_GROWSUP
3042 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3044 return (unsigned long)n - (unsigned long)end_of_stack(p);
3048 extern void set_task_stack_end_magic(struct task_struct *tsk);
3050 /* set thread flags in other task's structures
3051 * - see asm/thread_info.h for TIF_xxxx flags available
3053 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3055 set_ti_thread_flag(task_thread_info(tsk), flag);
3058 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3060 clear_ti_thread_flag(task_thread_info(tsk), flag);
3063 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3065 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3068 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3070 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3073 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3075 return test_ti_thread_flag(task_thread_info(tsk), flag);
3078 static inline void set_tsk_need_resched(struct task_struct *tsk)
3080 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3083 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3085 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3088 static inline int test_tsk_need_resched(struct task_struct *tsk)
3090 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3093 static inline int restart_syscall(void)
3095 set_tsk_thread_flag(current, TIF_SIGPENDING);
3096 return -ERESTARTNOINTR;
3099 static inline int signal_pending(struct task_struct *p)
3101 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3104 static inline int __fatal_signal_pending(struct task_struct *p)
3106 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3109 static inline int fatal_signal_pending(struct task_struct *p)
3111 return signal_pending(p) && __fatal_signal_pending(p);
3114 static inline int signal_pending_state(long state, struct task_struct *p)
3116 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3118 if (!signal_pending(p))
3121 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3125 * cond_resched() and cond_resched_lock(): latency reduction via
3126 * explicit rescheduling in places that are safe. The return
3127 * value indicates whether a reschedule was done in fact.
3128 * cond_resched_lock() will drop the spinlock before scheduling,
3129 * cond_resched_softirq() will enable bhs before scheduling.
3131 #ifndef CONFIG_PREEMPT
3132 extern int _cond_resched(void);
3134 static inline int _cond_resched(void) { return 0; }
3137 #define cond_resched() ({ \
3138 ___might_sleep(__FILE__, __LINE__, 0); \
3142 extern int __cond_resched_lock(spinlock_t *lock);
3144 #define cond_resched_lock(lock) ({ \
3145 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3146 __cond_resched_lock(lock); \
3149 extern int __cond_resched_softirq(void);
3151 #define cond_resched_softirq() ({ \
3152 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3153 __cond_resched_softirq(); \
3156 static inline void cond_resched_rcu(void)
3158 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3166 * Does a critical section need to be broken due to another
3167 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3168 * but a general need for low latency)
3170 static inline int spin_needbreak(spinlock_t *lock)
3172 #ifdef CONFIG_PREEMPT
3173 return spin_is_contended(lock);
3180 * Idle thread specific functions to determine the need_resched
3183 #ifdef TIF_POLLING_NRFLAG
3184 static inline int tsk_is_polling(struct task_struct *p)
3186 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3189 static inline void __current_set_polling(void)
3191 set_thread_flag(TIF_POLLING_NRFLAG);
3194 static inline bool __must_check current_set_polling_and_test(void)
3196 __current_set_polling();
3199 * Polling state must be visible before we test NEED_RESCHED,
3200 * paired by resched_curr()
3202 smp_mb__after_atomic();
3204 return unlikely(tif_need_resched());
3207 static inline void __current_clr_polling(void)
3209 clear_thread_flag(TIF_POLLING_NRFLAG);
3212 static inline bool __must_check current_clr_polling_and_test(void)
3214 __current_clr_polling();
3217 * Polling state must be visible before we test NEED_RESCHED,
3218 * paired by resched_curr()
3220 smp_mb__after_atomic();
3222 return unlikely(tif_need_resched());
3226 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3227 static inline void __current_set_polling(void) { }
3228 static inline void __current_clr_polling(void) { }
3230 static inline bool __must_check current_set_polling_and_test(void)
3232 return unlikely(tif_need_resched());
3234 static inline bool __must_check current_clr_polling_and_test(void)
3236 return unlikely(tif_need_resched());
3240 static inline void current_clr_polling(void)
3242 __current_clr_polling();
3245 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3246 * Once the bit is cleared, we'll get IPIs with every new
3247 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3250 smp_mb(); /* paired with resched_curr() */
3252 preempt_fold_need_resched();
3255 static __always_inline bool need_resched(void)
3257 return unlikely(tif_need_resched());
3261 * Thread group CPU time accounting.
3263 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3264 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3267 * Reevaluate whether the task has signals pending delivery.
3268 * Wake the task if so.
3269 * This is required every time the blocked sigset_t changes.
3270 * callers must hold sighand->siglock.
3272 extern void recalc_sigpending_and_wake(struct task_struct *t);
3273 extern void recalc_sigpending(void);
3275 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3277 static inline void signal_wake_up(struct task_struct *t, bool resume)
3279 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3281 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3283 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3287 * Wrappers for p->thread_info->cpu access. No-op on UP.
3291 static inline unsigned int task_cpu(const struct task_struct *p)
3293 #ifdef CONFIG_THREAD_INFO_IN_TASK
3296 return task_thread_info(p)->cpu;
3300 static inline int task_node(const struct task_struct *p)
3302 return cpu_to_node(task_cpu(p));
3305 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3309 static inline unsigned int task_cpu(const struct task_struct *p)
3314 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3318 #endif /* CONFIG_SMP */
3321 * In order to reduce various lock holder preemption latencies provide an
3322 * interface to see if a vCPU is currently running or not.
3324 * This allows us to terminate optimistic spin loops and block, analogous to
3325 * the native optimistic spin heuristic of testing if the lock owner task is
3328 #ifndef vcpu_is_preempted
3329 # define vcpu_is_preempted(cpu) false
3332 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3333 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3335 #ifdef CONFIG_CGROUP_SCHED
3336 extern struct task_group root_task_group;
3337 #endif /* CONFIG_CGROUP_SCHED */
3339 extern int task_can_switch_user(struct user_struct *up,
3340 struct task_struct *tsk);
3342 #ifdef CONFIG_TASK_XACCT
3343 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3345 tsk->ioac.rchar += amt;
3348 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3350 tsk->ioac.wchar += amt;
3353 static inline void inc_syscr(struct task_struct *tsk)
3358 static inline void inc_syscw(struct task_struct *tsk)
3363 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3367 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3371 static inline void inc_syscr(struct task_struct *tsk)
3375 static inline void inc_syscw(struct task_struct *tsk)
3380 #ifndef TASK_SIZE_OF
3381 #define TASK_SIZE_OF(tsk) TASK_SIZE
3385 extern void mm_update_next_owner(struct mm_struct *mm);
3387 static inline void mm_update_next_owner(struct mm_struct *mm)
3390 #endif /* CONFIG_MEMCG */
3392 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3395 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3398 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3401 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3404 static inline unsigned long rlimit(unsigned int limit)
3406 return task_rlimit(current, limit);
3409 static inline unsigned long rlimit_max(unsigned int limit)
3411 return task_rlimit_max(current, limit);
3414 #define SCHED_CPUFREQ_RT (1U << 0)
3415 #define SCHED_CPUFREQ_DL (1U << 1)
3416 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3418 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3420 #ifdef CONFIG_CPU_FREQ
3421 struct update_util_data {
3422 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3425 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3426 void (*func)(struct update_util_data *data, u64 time,
3427 unsigned int flags));
3428 void cpufreq_remove_update_util_hook(int cpu);
3429 #endif /* CONFIG_CPU_FREQ */