4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
13 #include <asm/param.h> /* for HZ */
15 #include <linux/capability.h>
16 #include <linux/threads.h>
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/timex.h>
20 #include <linux/jiffies.h>
21 #include <linux/plist.h>
22 #include <linux/rbtree.h>
23 #include <linux/thread_info.h>
24 #include <linux/cpumask.h>
25 #include <linux/errno.h>
26 #include <linux/nodemask.h>
27 #include <linux/mm_types.h>
28 #include <linux/preempt.h>
31 #include <asm/ptrace.h>
33 #include <linux/smp.h>
34 #include <linux/sem.h>
35 #include <linux/shm.h>
36 #include <linux/signal.h>
37 #include <linux/compiler.h>
38 #include <linux/completion.h>
39 #include <linux/pid.h>
40 #include <linux/percpu.h>
41 #include <linux/topology.h>
42 #include <linux/seccomp.h>
43 #include <linux/rcupdate.h>
44 #include <linux/rculist.h>
45 #include <linux/rtmutex.h>
47 #include <linux/time.h>
48 #include <linux/param.h>
49 #include <linux/resource.h>
50 #include <linux/timer.h>
51 #include <linux/hrtimer.h>
52 #include <linux/kcov.h>
53 #include <linux/task_io_accounting.h>
54 #include <linux/latencytop.h>
55 #include <linux/cred.h>
56 #include <linux/llist.h>
57 #include <linux/uidgid.h>
58 #include <linux/gfp.h>
59 #include <linux/magic.h>
60 #include <linux/cgroup-defs.h>
62 #include <asm/processor.h>
64 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
67 * Extended scheduling parameters data structure.
69 * This is needed because the original struct sched_param can not be
70 * altered without introducing ABI issues with legacy applications
71 * (e.g., in sched_getparam()).
73 * However, the possibility of specifying more than just a priority for
74 * the tasks may be useful for a wide variety of application fields, e.g.,
75 * multimedia, streaming, automation and control, and many others.
77 * This variant (sched_attr) is meant at describing a so-called
78 * sporadic time-constrained task. In such model a task is specified by:
79 * - the activation period or minimum instance inter-arrival time;
80 * - the maximum (or average, depending on the actual scheduling
81 * discipline) computation time of all instances, a.k.a. runtime;
82 * - the deadline (relative to the actual activation time) of each
84 * Very briefly, a periodic (sporadic) task asks for the execution of
85 * some specific computation --which is typically called an instance--
86 * (at most) every period. Moreover, each instance typically lasts no more
87 * than the runtime and must be completed by time instant t equal to
88 * the instance activation time + the deadline.
90 * This is reflected by the actual fields of the sched_attr structure:
92 * @size size of the structure, for fwd/bwd compat.
94 * @sched_policy task's scheduling policy
95 * @sched_flags for customizing the scheduler behaviour
96 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
97 * @sched_priority task's static priority (SCHED_FIFO/RR)
98 * @sched_deadline representative of the task's deadline
99 * @sched_runtime representative of the task's runtime
100 * @sched_period representative of the task's period
102 * Given this task model, there are a multiplicity of scheduling algorithms
103 * and policies, that can be used to ensure all the tasks will make their
104 * timing constraints.
106 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
107 * only user of this new interface. More information about the algorithm
108 * available in the scheduling class file or in Documentation/.
116 /* SCHED_NORMAL, SCHED_BATCH */
119 /* SCHED_FIFO, SCHED_RR */
128 struct futex_pi_state;
129 struct robust_list_head;
132 struct perf_event_context;
137 #define VMACACHE_BITS 2
138 #define VMACACHE_SIZE (1U << VMACACHE_BITS)
139 #define VMACACHE_MASK (VMACACHE_SIZE - 1)
142 * These are the constant used to fake the fixed-point load-average
143 * counting. Some notes:
144 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
145 * a load-average precision of 10 bits integer + 11 bits fractional
146 * - if you want to count load-averages more often, you need more
147 * precision, or rounding will get you. With 2-second counting freq,
148 * the EXP_n values would be 1981, 2034 and 2043 if still using only
151 extern unsigned long avenrun[]; /* Load averages */
152 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
154 #define FSHIFT 11 /* nr of bits of precision */
155 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
156 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
157 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
158 #define EXP_5 2014 /* 1/exp(5sec/5min) */
159 #define EXP_15 2037 /* 1/exp(5sec/15min) */
161 #define CALC_LOAD(load,exp,n) \
163 load += n*(FIXED_1-exp); \
166 extern unsigned long total_forks;
167 extern int nr_threads;
168 DECLARE_PER_CPU(unsigned long, process_counts);
169 extern int nr_processes(void);
170 extern unsigned long nr_running(void);
171 extern bool single_task_running(void);
172 extern unsigned long nr_iowait(void);
173 extern unsigned long nr_iowait_cpu(int cpu);
174 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
176 extern void calc_global_load(unsigned long ticks);
178 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
179 extern void cpu_load_update_nohz_start(void);
180 extern void cpu_load_update_nohz_stop(void);
182 static inline void cpu_load_update_nohz_start(void) { }
183 static inline void cpu_load_update_nohz_stop(void) { }
186 extern void dump_cpu_task(int cpu);
191 #ifdef CONFIG_SCHED_DEBUG
192 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
193 extern void proc_sched_set_task(struct task_struct *p);
197 * Task state bitmask. NOTE! These bits are also
198 * encoded in fs/proc/array.c: get_task_state().
200 * We have two separate sets of flags: task->state
201 * is about runnability, while task->exit_state are
202 * about the task exiting. Confusing, but this way
203 * modifying one set can't modify the other one by
206 #define TASK_RUNNING 0
207 #define TASK_INTERRUPTIBLE 1
208 #define TASK_UNINTERRUPTIBLE 2
209 #define __TASK_STOPPED 4
210 #define __TASK_TRACED 8
211 /* in tsk->exit_state */
213 #define EXIT_ZOMBIE 32
214 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
215 /* in tsk->state again */
217 #define TASK_WAKEKILL 128
218 #define TASK_WAKING 256
219 #define TASK_PARKED 512
220 #define TASK_NOLOAD 1024
221 #define TASK_NEW 2048
222 #define TASK_STATE_MAX 4096
224 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
226 extern char ___assert_task_state[1 - 2*!!(
227 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
229 /* Convenience macros for the sake of set_task_state */
230 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
231 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
232 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
234 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
236 /* Convenience macros for the sake of wake_up */
237 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
238 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
240 /* get_task_state() */
241 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
242 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
243 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
245 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
246 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
247 #define task_is_stopped_or_traced(task) \
248 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
249 #define task_contributes_to_load(task) \
250 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
251 (task->flags & PF_FROZEN) == 0 && \
252 (task->state & TASK_NOLOAD) == 0)
254 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
256 #define __set_task_state(tsk, state_value) \
258 (tsk)->task_state_change = _THIS_IP_; \
259 (tsk)->state = (state_value); \
261 #define set_task_state(tsk, state_value) \
263 (tsk)->task_state_change = _THIS_IP_; \
264 smp_store_mb((tsk)->state, (state_value)); \
267 #define __set_current_state(state_value) \
269 current->task_state_change = _THIS_IP_; \
270 current->state = (state_value); \
272 #define set_current_state(state_value) \
274 current->task_state_change = _THIS_IP_; \
275 smp_store_mb(current->state, (state_value)); \
281 * @tsk had better be current, or you get to keep the pieces.
283 * The only reason is that computing current can be more expensive than
284 * using a pointer that's already available.
286 * Therefore, see set_current_state().
288 #define __set_task_state(tsk, state_value) \
289 do { (tsk)->state = (state_value); } while (0)
290 #define set_task_state(tsk, state_value) \
291 smp_store_mb((tsk)->state, (state_value))
294 * set_current_state() includes a barrier so that the write of current->state
295 * is correctly serialised wrt the caller's subsequent test of whether to
299 * set_current_state(TASK_UNINTERRUPTIBLE);
305 * __set_current_state(TASK_RUNNING);
307 * If the caller does not need such serialisation (because, for instance, the
308 * condition test and condition change and wakeup are under the same lock) then
309 * use __set_current_state().
311 * The above is typically ordered against the wakeup, which does:
313 * need_sleep = false;
314 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
316 * Where wake_up_state() (and all other wakeup primitives) imply enough
317 * barriers to order the store of the variable against wakeup.
319 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
320 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
321 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
323 * This is obviously fine, since they both store the exact same value.
325 * Also see the comments of try_to_wake_up().
327 #define __set_current_state(state_value) \
328 do { current->state = (state_value); } while (0)
329 #define set_current_state(state_value) \
330 smp_store_mb(current->state, (state_value))
334 /* Task command name length */
335 #define TASK_COMM_LEN 16
337 #include <linux/spinlock.h>
340 * This serializes "schedule()" and also protects
341 * the run-queue from deletions/modifications (but
342 * _adding_ to the beginning of the run-queue has
345 extern rwlock_t tasklist_lock;
346 extern spinlock_t mmlist_lock;
350 #ifdef CONFIG_PROVE_RCU
351 extern int lockdep_tasklist_lock_is_held(void);
352 #endif /* #ifdef CONFIG_PROVE_RCU */
354 extern void sched_init(void);
355 extern void sched_init_smp(void);
356 extern asmlinkage void schedule_tail(struct task_struct *prev);
357 extern void init_idle(struct task_struct *idle, int cpu);
358 extern void init_idle_bootup_task(struct task_struct *idle);
360 extern cpumask_var_t cpu_isolated_map;
362 extern int runqueue_is_locked(int cpu);
364 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
365 extern void nohz_balance_enter_idle(int cpu);
366 extern void set_cpu_sd_state_idle(void);
367 extern int get_nohz_timer_target(void);
369 static inline void nohz_balance_enter_idle(int cpu) { }
370 static inline void set_cpu_sd_state_idle(void) { }
374 * Only dump TASK_* tasks. (0 for all tasks)
376 extern void show_state_filter(unsigned long state_filter);
378 static inline void show_state(void)
380 show_state_filter(0);
383 extern void show_regs(struct pt_regs *);
386 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
387 * task), SP is the stack pointer of the first frame that should be shown in the back
388 * trace (or NULL if the entire call-chain of the task should be shown).
390 extern void show_stack(struct task_struct *task, unsigned long *sp);
392 extern void cpu_init (void);
393 extern void trap_init(void);
394 extern void update_process_times(int user);
395 extern void scheduler_tick(void);
396 extern int sched_cpu_starting(unsigned int cpu);
397 extern int sched_cpu_activate(unsigned int cpu);
398 extern int sched_cpu_deactivate(unsigned int cpu);
400 #ifdef CONFIG_HOTPLUG_CPU
401 extern int sched_cpu_dying(unsigned int cpu);
403 # define sched_cpu_dying NULL
406 extern void sched_show_task(struct task_struct *p);
408 #ifdef CONFIG_LOCKUP_DETECTOR
409 extern void touch_softlockup_watchdog_sched(void);
410 extern void touch_softlockup_watchdog(void);
411 extern void touch_softlockup_watchdog_sync(void);
412 extern void touch_all_softlockup_watchdogs(void);
413 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
415 size_t *lenp, loff_t *ppos);
416 extern unsigned int softlockup_panic;
417 extern unsigned int hardlockup_panic;
418 void lockup_detector_init(void);
420 static inline void touch_softlockup_watchdog_sched(void)
423 static inline void touch_softlockup_watchdog(void)
426 static inline void touch_softlockup_watchdog_sync(void)
429 static inline void touch_all_softlockup_watchdogs(void)
432 static inline void lockup_detector_init(void)
437 #ifdef CONFIG_DETECT_HUNG_TASK
438 void reset_hung_task_detector(void);
440 static inline void reset_hung_task_detector(void)
445 /* Attach to any functions which should be ignored in wchan output. */
446 #define __sched __attribute__((__section__(".sched.text")))
448 /* Linker adds these: start and end of __sched functions */
449 extern char __sched_text_start[], __sched_text_end[];
451 /* Is this address in the __sched functions? */
452 extern int in_sched_functions(unsigned long addr);
454 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
455 extern signed long schedule_timeout(signed long timeout);
456 extern signed long schedule_timeout_interruptible(signed long timeout);
457 extern signed long schedule_timeout_killable(signed long timeout);
458 extern signed long schedule_timeout_uninterruptible(signed long timeout);
459 extern signed long schedule_timeout_idle(signed long timeout);
460 asmlinkage void schedule(void);
461 extern void schedule_preempt_disabled(void);
463 extern int __must_check io_schedule_prepare(void);
464 extern void io_schedule_finish(int token);
465 extern long io_schedule_timeout(long timeout);
466 extern void io_schedule(void);
468 void __noreturn do_task_dead(void);
471 struct user_namespace;
474 extern void arch_pick_mmap_layout(struct mm_struct *mm);
476 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
477 unsigned long, unsigned long);
479 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
480 unsigned long len, unsigned long pgoff,
481 unsigned long flags);
483 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
486 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
487 #define SUID_DUMP_USER 1 /* Dump as user of process */
488 #define SUID_DUMP_ROOT 2 /* Dump as root */
492 /* for SUID_DUMP_* above */
493 #define MMF_DUMPABLE_BITS 2
494 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
496 extern void set_dumpable(struct mm_struct *mm, int value);
498 * This returns the actual value of the suid_dumpable flag. For things
499 * that are using this for checking for privilege transitions, it must
500 * test against SUID_DUMP_USER rather than treating it as a boolean
503 static inline int __get_dumpable(unsigned long mm_flags)
505 return mm_flags & MMF_DUMPABLE_MASK;
508 static inline int get_dumpable(struct mm_struct *mm)
510 return __get_dumpable(mm->flags);
513 /* coredump filter bits */
514 #define MMF_DUMP_ANON_PRIVATE 2
515 #define MMF_DUMP_ANON_SHARED 3
516 #define MMF_DUMP_MAPPED_PRIVATE 4
517 #define MMF_DUMP_MAPPED_SHARED 5
518 #define MMF_DUMP_ELF_HEADERS 6
519 #define MMF_DUMP_HUGETLB_PRIVATE 7
520 #define MMF_DUMP_HUGETLB_SHARED 8
521 #define MMF_DUMP_DAX_PRIVATE 9
522 #define MMF_DUMP_DAX_SHARED 10
524 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
525 #define MMF_DUMP_FILTER_BITS 9
526 #define MMF_DUMP_FILTER_MASK \
527 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
528 #define MMF_DUMP_FILTER_DEFAULT \
529 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
530 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
532 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
533 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
535 # define MMF_DUMP_MASK_DEFAULT_ELF 0
537 /* leave room for more dump flags */
538 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
539 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
541 * This one-shot flag is dropped due to necessity of changing exe once again
544 //#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
546 #define MMF_HAS_UPROBES 19 /* has uprobes */
547 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
548 #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
549 #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
550 #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
552 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
554 struct sighand_struct {
556 struct k_sigaction action[_NSIG];
558 wait_queue_head_t signalfd_wqh;
561 struct pacct_struct {
564 unsigned long ac_mem;
565 u64 ac_utime, ac_stime;
566 unsigned long ac_minflt, ac_majflt;
575 * struct prev_cputime - snaphsot of system and user cputime
576 * @utime: time spent in user mode
577 * @stime: time spent in system mode
578 * @lock: protects the above two fields
580 * Stores previous user/system time values such that we can guarantee
583 struct prev_cputime {
584 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
591 static inline void prev_cputime_init(struct prev_cputime *prev)
593 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
594 prev->utime = prev->stime = 0;
595 raw_spin_lock_init(&prev->lock);
600 * struct task_cputime - collected CPU time counts
601 * @utime: time spent in user mode, in nanoseconds
602 * @stime: time spent in kernel mode, in nanoseconds
603 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
605 * This structure groups together three kinds of CPU time that are tracked for
606 * threads and thread groups. Most things considering CPU time want to group
607 * these counts together and treat all three of them in parallel.
609 struct task_cputime {
612 unsigned long long sum_exec_runtime;
615 /* Alternate field names when used to cache expirations. */
616 #define virt_exp utime
617 #define prof_exp stime
618 #define sched_exp sum_exec_runtime
621 * This is the atomic variant of task_cputime, which can be used for
622 * storing and updating task_cputime statistics without locking.
624 struct task_cputime_atomic {
627 atomic64_t sum_exec_runtime;
630 #define INIT_CPUTIME_ATOMIC \
631 (struct task_cputime_atomic) { \
632 .utime = ATOMIC64_INIT(0), \
633 .stime = ATOMIC64_INIT(0), \
634 .sum_exec_runtime = ATOMIC64_INIT(0), \
637 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
640 * Disable preemption until the scheduler is running -- use an unconditional
641 * value so that it also works on !PREEMPT_COUNT kernels.
643 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
645 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
648 * Initial preempt_count value; reflects the preempt_count schedule invariant
649 * which states that during context switches:
651 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
653 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
654 * Note: See finish_task_switch().
656 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
659 * struct thread_group_cputimer - thread group interval timer counts
660 * @cputime_atomic: atomic thread group interval timers.
661 * @running: true when there are timers running and
662 * @cputime_atomic receives updates.
663 * @checking_timer: true when a thread in the group is in the
664 * process of checking for thread group timers.
666 * This structure contains the version of task_cputime, above, that is
667 * used for thread group CPU timer calculations.
669 struct thread_group_cputimer {
670 struct task_cputime_atomic cputime_atomic;
675 #include <linux/rwsem.h>
679 * NOTE! "signal_struct" does not have its own
680 * locking, because a shared signal_struct always
681 * implies a shared sighand_struct, so locking
682 * sighand_struct is always a proper superset of
683 * the locking of signal_struct.
685 struct signal_struct {
689 struct list_head thread_head;
691 wait_queue_head_t wait_chldexit; /* for wait4() */
693 /* current thread group signal load-balancing target: */
694 struct task_struct *curr_target;
696 /* shared signal handling: */
697 struct sigpending shared_pending;
699 /* thread group exit support */
702 * - notify group_exit_task when ->count is equal to notify_count
703 * - everyone except group_exit_task is stopped during signal delivery
704 * of fatal signals, group_exit_task processes the signal.
707 struct task_struct *group_exit_task;
709 /* thread group stop support, overloads group_exit_code too */
710 int group_stop_count;
711 unsigned int flags; /* see SIGNAL_* flags below */
714 * PR_SET_CHILD_SUBREAPER marks a process, like a service
715 * manager, to re-parent orphan (double-forking) child processes
716 * to this process instead of 'init'. The service manager is
717 * able to receive SIGCHLD signals and is able to investigate
718 * the process until it calls wait(). All children of this
719 * process will inherit a flag if they should look for a
720 * child_subreaper process at exit.
722 unsigned int is_child_subreaper:1;
723 unsigned int has_child_subreaper:1;
725 #ifdef CONFIG_POSIX_TIMERS
727 /* POSIX.1b Interval Timers */
729 struct list_head posix_timers;
731 /* ITIMER_REAL timer for the process */
732 struct hrtimer real_timer;
733 ktime_t it_real_incr;
736 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
737 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
738 * values are defined to 0 and 1 respectively
740 struct cpu_itimer it[2];
743 * Thread group totals for process CPU timers.
744 * See thread_group_cputimer(), et al, for details.
746 struct thread_group_cputimer cputimer;
748 /* Earliest-expiration cache. */
749 struct task_cputime cputime_expires;
751 struct list_head cpu_timers[3];
755 struct pid *leader_pid;
757 #ifdef CONFIG_NO_HZ_FULL
758 atomic_t tick_dep_mask;
761 struct pid *tty_old_pgrp;
763 /* boolean value for session group leader */
766 struct tty_struct *tty; /* NULL if no tty */
768 #ifdef CONFIG_SCHED_AUTOGROUP
769 struct autogroup *autogroup;
772 * Cumulative resource counters for dead threads in the group,
773 * and for reaped dead child processes forked by this group.
774 * Live threads maintain their own counters and add to these
775 * in __exit_signal, except for the group leader.
777 seqlock_t stats_lock;
778 u64 utime, stime, cutime, cstime;
781 struct prev_cputime prev_cputime;
782 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
783 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
784 unsigned long inblock, oublock, cinblock, coublock;
785 unsigned long maxrss, cmaxrss;
786 struct task_io_accounting ioac;
789 * Cumulative ns of schedule CPU time fo dead threads in the
790 * group, not including a zombie group leader, (This only differs
791 * from jiffies_to_ns(utime + stime) if sched_clock uses something
792 * other than jiffies.)
794 unsigned long long sum_sched_runtime;
797 * We don't bother to synchronize most readers of this at all,
798 * because there is no reader checking a limit that actually needs
799 * to get both rlim_cur and rlim_max atomically, and either one
800 * alone is a single word that can safely be read normally.
801 * getrlimit/setrlimit use task_lock(current->group_leader) to
802 * protect this instead of the siglock, because they really
803 * have no need to disable irqs.
805 struct rlimit rlim[RLIM_NLIMITS];
807 #ifdef CONFIG_BSD_PROCESS_ACCT
808 struct pacct_struct pacct; /* per-process accounting information */
810 #ifdef CONFIG_TASKSTATS
811 struct taskstats *stats;
815 struct tty_audit_buf *tty_audit_buf;
819 * Thread is the potential origin of an oom condition; kill first on
822 bool oom_flag_origin;
823 short oom_score_adj; /* OOM kill score adjustment */
824 short oom_score_adj_min; /* OOM kill score adjustment min value.
825 * Only settable by CAP_SYS_RESOURCE. */
826 struct mm_struct *oom_mm; /* recorded mm when the thread group got
827 * killed by the oom killer */
829 struct mutex cred_guard_mutex; /* guard against foreign influences on
830 * credential calculations
831 * (notably. ptrace) */
835 * Bits in flags field of signal_struct.
837 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
838 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
839 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
840 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
842 * Pending notifications to parent.
844 #define SIGNAL_CLD_STOPPED 0x00000010
845 #define SIGNAL_CLD_CONTINUED 0x00000020
846 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
848 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
850 #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
851 SIGNAL_STOP_CONTINUED)
853 static inline void signal_set_stop_flags(struct signal_struct *sig,
856 WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
857 sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
860 /* If true, all threads except ->group_exit_task have pending SIGKILL */
861 static inline int signal_group_exit(const struct signal_struct *sig)
863 return (sig->flags & SIGNAL_GROUP_EXIT) ||
864 (sig->group_exit_task != NULL);
868 * Some day this will be a full-fledged user tracking system..
871 atomic_t __count; /* reference count */
872 atomic_t processes; /* How many processes does this user have? */
873 atomic_t sigpending; /* How many pending signals does this user have? */
874 #ifdef CONFIG_INOTIFY_USER
875 atomic_t inotify_watches; /* How many inotify watches does this user have? */
876 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
878 #ifdef CONFIG_FANOTIFY
879 atomic_t fanotify_listeners;
882 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
884 #ifdef CONFIG_POSIX_MQUEUE
885 /* protected by mq_lock */
886 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
888 unsigned long locked_shm; /* How many pages of mlocked shm ? */
889 unsigned long unix_inflight; /* How many files in flight in unix sockets */
890 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
893 struct key *uid_keyring; /* UID specific keyring */
894 struct key *session_keyring; /* UID's default session keyring */
897 /* Hash table maintenance information */
898 struct hlist_node uidhash_node;
901 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
902 atomic_long_t locked_vm;
906 extern int uids_sysfs_init(void);
908 extern struct user_struct *find_user(kuid_t);
910 extern struct user_struct root_user;
911 #define INIT_USER (&root_user)
914 struct backing_dev_info;
915 struct reclaim_state;
917 #ifdef CONFIG_SCHED_INFO
919 /* cumulative counters */
920 unsigned long pcount; /* # of times run on this cpu */
921 unsigned long long run_delay; /* time spent waiting on a runqueue */
924 unsigned long long last_arrival,/* when we last ran on a cpu */
925 last_queued; /* when we were last queued to run */
927 #endif /* CONFIG_SCHED_INFO */
929 #ifdef CONFIG_TASK_DELAY_ACCT
930 struct task_delay_info {
932 unsigned int flags; /* Private per-task flags */
934 /* For each stat XXX, add following, aligned appropriately
936 * struct timespec XXX_start, XXX_end;
940 * Atomicity of updates to XXX_delay, XXX_count protected by
941 * single lock above (split into XXX_lock if contention is an issue).
945 * XXX_count is incremented on every XXX operation, the delay
946 * associated with the operation is added to XXX_delay.
947 * XXX_delay contains the accumulated delay time in nanoseconds.
949 u64 blkio_start; /* Shared by blkio, swapin */
950 u64 blkio_delay; /* wait for sync block io completion */
951 u64 swapin_delay; /* wait for swapin block io completion */
952 u32 blkio_count; /* total count of the number of sync block */
953 /* io operations performed */
954 u32 swapin_count; /* total count of the number of swapin block */
955 /* io operations performed */
958 u64 freepages_delay; /* wait for memory reclaim */
959 u32 freepages_count; /* total count of memory reclaim */
961 #endif /* CONFIG_TASK_DELAY_ACCT */
963 static inline int sched_info_on(void)
965 #ifdef CONFIG_SCHEDSTATS
967 #elif defined(CONFIG_TASK_DELAY_ACCT)
968 extern int delayacct_on;
975 #ifdef CONFIG_SCHEDSTATS
976 void force_schedstat_enabled(void);
987 * Integer metrics need fixed point arithmetic, e.g., sched/fair
988 * has a few: load, load_avg, util_avg, freq, and capacity.
990 * We define a basic fixed point arithmetic range, and then formalize
991 * all these metrics based on that basic range.
993 # define SCHED_FIXEDPOINT_SHIFT 10
994 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
997 * Increase resolution of cpu_capacity calculations
999 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
1000 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
1003 * Wake-queues are lists of tasks with a pending wakeup, whose
1004 * callers have already marked the task as woken internally,
1005 * and can thus carry on. A common use case is being able to
1006 * do the wakeups once the corresponding user lock as been
1009 * We hold reference to each task in the list across the wakeup,
1010 * thus guaranteeing that the memory is still valid by the time
1011 * the actual wakeups are performed in wake_up_q().
1013 * One per task suffices, because there's never a need for a task to be
1014 * in two wake queues simultaneously; it is forbidden to abandon a task
1015 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
1016 * already in a wake queue, the wakeup will happen soon and the second
1017 * waker can just skip it.
1019 * The DEFINE_WAKE_Q macro declares and initializes the list head.
1020 * wake_up_q() does NOT reinitialize the list; it's expected to be
1021 * called near the end of a function, where the fact that the queue is
1022 * not used again will be easy to see by inspection.
1024 * Note that this can cause spurious wakeups. schedule() callers
1025 * must ensure the call is done inside a loop, confirming that the
1026 * wakeup condition has in fact occurred.
1028 struct wake_q_node {
1029 struct wake_q_node *next;
1032 struct wake_q_head {
1033 struct wake_q_node *first;
1034 struct wake_q_node **lastp;
1037 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1039 #define DEFINE_WAKE_Q(name) \
1040 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1042 extern void wake_q_add(struct wake_q_head *head,
1043 struct task_struct *task);
1044 extern void wake_up_q(struct wake_q_head *head);
1047 * sched-domains (multiprocessor balancing) declarations:
1050 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1051 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1052 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1053 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1054 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1055 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1056 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1057 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1058 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1059 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1060 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1061 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1062 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1063 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1064 #define SD_NUMA 0x4000 /* cross-node balancing */
1066 #ifdef CONFIG_SCHED_SMT
1067 static inline int cpu_smt_flags(void)
1069 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1073 #ifdef CONFIG_SCHED_MC
1074 static inline int cpu_core_flags(void)
1076 return SD_SHARE_PKG_RESOURCES;
1081 static inline int cpu_numa_flags(void)
1087 extern int arch_asym_cpu_priority(int cpu);
1089 struct sched_domain_attr {
1090 int relax_domain_level;
1093 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1094 .relax_domain_level = -1, \
1097 extern int sched_domain_level_max;
1101 struct sched_domain_shared {
1103 atomic_t nr_busy_cpus;
1107 struct sched_domain {
1108 /* These fields must be setup */
1109 struct sched_domain *parent; /* top domain must be null terminated */
1110 struct sched_domain *child; /* bottom domain must be null terminated */
1111 struct sched_group *groups; /* the balancing groups of the domain */
1112 unsigned long min_interval; /* Minimum balance interval ms */
1113 unsigned long max_interval; /* Maximum balance interval ms */
1114 unsigned int busy_factor; /* less balancing by factor if busy */
1115 unsigned int imbalance_pct; /* No balance until over watermark */
1116 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1117 unsigned int busy_idx;
1118 unsigned int idle_idx;
1119 unsigned int newidle_idx;
1120 unsigned int wake_idx;
1121 unsigned int forkexec_idx;
1122 unsigned int smt_gain;
1124 int nohz_idle; /* NOHZ IDLE status */
1125 int flags; /* See SD_* */
1128 /* Runtime fields. */
1129 unsigned long last_balance; /* init to jiffies. units in jiffies */
1130 unsigned int balance_interval; /* initialise to 1. units in ms. */
1131 unsigned int nr_balance_failed; /* initialise to 0 */
1133 /* idle_balance() stats */
1134 u64 max_newidle_lb_cost;
1135 unsigned long next_decay_max_lb_cost;
1137 u64 avg_scan_cost; /* select_idle_sibling */
1139 #ifdef CONFIG_SCHEDSTATS
1140 /* load_balance() stats */
1141 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1142 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1143 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1144 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1145 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1146 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1147 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1148 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1150 /* Active load balancing */
1151 unsigned int alb_count;
1152 unsigned int alb_failed;
1153 unsigned int alb_pushed;
1155 /* SD_BALANCE_EXEC stats */
1156 unsigned int sbe_count;
1157 unsigned int sbe_balanced;
1158 unsigned int sbe_pushed;
1160 /* SD_BALANCE_FORK stats */
1161 unsigned int sbf_count;
1162 unsigned int sbf_balanced;
1163 unsigned int sbf_pushed;
1165 /* try_to_wake_up() stats */
1166 unsigned int ttwu_wake_remote;
1167 unsigned int ttwu_move_affine;
1168 unsigned int ttwu_move_balance;
1170 #ifdef CONFIG_SCHED_DEBUG
1174 void *private; /* used during construction */
1175 struct rcu_head rcu; /* used during destruction */
1177 struct sched_domain_shared *shared;
1179 unsigned int span_weight;
1181 * Span of all CPUs in this domain.
1183 * NOTE: this field is variable length. (Allocated dynamically
1184 * by attaching extra space to the end of the structure,
1185 * depending on how many CPUs the kernel has booted up with)
1187 unsigned long span[0];
1190 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1192 return to_cpumask(sd->span);
1195 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1196 struct sched_domain_attr *dattr_new);
1198 /* Allocate an array of sched domains, for partition_sched_domains(). */
1199 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1200 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1202 bool cpus_share_cache(int this_cpu, int that_cpu);
1204 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1205 typedef int (*sched_domain_flags_f)(void);
1207 #define SDTL_OVERLAP 0x01
1210 struct sched_domain **__percpu sd;
1211 struct sched_domain_shared **__percpu sds;
1212 struct sched_group **__percpu sg;
1213 struct sched_group_capacity **__percpu sgc;
1216 struct sched_domain_topology_level {
1217 sched_domain_mask_f mask;
1218 sched_domain_flags_f sd_flags;
1221 struct sd_data data;
1222 #ifdef CONFIG_SCHED_DEBUG
1227 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1228 extern void wake_up_if_idle(int cpu);
1230 #ifdef CONFIG_SCHED_DEBUG
1231 # define SD_INIT_NAME(type) .name = #type
1233 # define SD_INIT_NAME(type)
1236 #else /* CONFIG_SMP */
1238 struct sched_domain_attr;
1241 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1242 struct sched_domain_attr *dattr_new)
1246 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1251 #endif /* !CONFIG_SMP */
1254 struct io_context; /* See blkdev.h */
1257 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1258 extern void prefetch_stack(struct task_struct *t);
1260 static inline void prefetch_stack(struct task_struct *t) { }
1263 struct audit_context; /* See audit.c */
1265 struct pipe_inode_info;
1266 struct uts_namespace;
1268 struct load_weight {
1269 unsigned long weight;
1274 * The load_avg/util_avg accumulates an infinite geometric series
1275 * (see __update_load_avg() in kernel/sched/fair.c).
1277 * [load_avg definition]
1279 * load_avg = runnable% * scale_load_down(load)
1281 * where runnable% is the time ratio that a sched_entity is runnable.
1282 * For cfs_rq, it is the aggregated load_avg of all runnable and
1283 * blocked sched_entities.
1285 * load_avg may also take frequency scaling into account:
1287 * load_avg = runnable% * scale_load_down(load) * freq%
1289 * where freq% is the CPU frequency normalized to the highest frequency.
1291 * [util_avg definition]
1293 * util_avg = running% * SCHED_CAPACITY_SCALE
1295 * where running% is the time ratio that a sched_entity is running on
1296 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1297 * and blocked sched_entities.
1299 * util_avg may also factor frequency scaling and CPU capacity scaling:
1301 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1303 * where freq% is the same as above, and capacity% is the CPU capacity
1304 * normalized to the greatest capacity (due to uarch differences, etc).
1306 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1307 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1308 * we therefore scale them to as large a range as necessary. This is for
1309 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1313 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1314 * with the highest load (=88761), always runnable on a single cfs_rq,
1315 * and should not overflow as the number already hits PID_MAX_LIMIT.
1317 * For all other cases (including 32-bit kernels), struct load_weight's
1318 * weight will overflow first before we do, because:
1320 * Max(load_avg) <= Max(load.weight)
1322 * Then it is the load_weight's responsibility to consider overflow
1326 u64 last_update_time, load_sum;
1327 u32 util_sum, period_contrib;
1328 unsigned long load_avg, util_avg;
1331 #ifdef CONFIG_SCHEDSTATS
1332 struct sched_statistics {
1342 s64 sum_sleep_runtime;
1349 u64 nr_migrations_cold;
1350 u64 nr_failed_migrations_affine;
1351 u64 nr_failed_migrations_running;
1352 u64 nr_failed_migrations_hot;
1353 u64 nr_forced_migrations;
1356 u64 nr_wakeups_sync;
1357 u64 nr_wakeups_migrate;
1358 u64 nr_wakeups_local;
1359 u64 nr_wakeups_remote;
1360 u64 nr_wakeups_affine;
1361 u64 nr_wakeups_affine_attempts;
1362 u64 nr_wakeups_passive;
1363 u64 nr_wakeups_idle;
1367 struct sched_entity {
1368 struct load_weight load; /* for load-balancing */
1369 struct rb_node run_node;
1370 struct list_head group_node;
1374 u64 sum_exec_runtime;
1376 u64 prev_sum_exec_runtime;
1380 #ifdef CONFIG_SCHEDSTATS
1381 struct sched_statistics statistics;
1384 #ifdef CONFIG_FAIR_GROUP_SCHED
1386 struct sched_entity *parent;
1387 /* rq on which this entity is (to be) queued: */
1388 struct cfs_rq *cfs_rq;
1389 /* rq "owned" by this entity/group: */
1390 struct cfs_rq *my_q;
1395 * Per entity load average tracking.
1397 * Put into separate cache line so it does not
1398 * collide with read-mostly values above.
1400 struct sched_avg avg ____cacheline_aligned_in_smp;
1404 struct sched_rt_entity {
1405 struct list_head run_list;
1406 unsigned long timeout;
1407 unsigned long watchdog_stamp;
1408 unsigned int time_slice;
1409 unsigned short on_rq;
1410 unsigned short on_list;
1412 struct sched_rt_entity *back;
1413 #ifdef CONFIG_RT_GROUP_SCHED
1414 struct sched_rt_entity *parent;
1415 /* rq on which this entity is (to be) queued: */
1416 struct rt_rq *rt_rq;
1417 /* rq "owned" by this entity/group: */
1422 struct sched_dl_entity {
1423 struct rb_node rb_node;
1426 * Original scheduling parameters. Copied here from sched_attr
1427 * during sched_setattr(), they will remain the same until
1428 * the next sched_setattr().
1430 u64 dl_runtime; /* maximum runtime for each instance */
1431 u64 dl_deadline; /* relative deadline of each instance */
1432 u64 dl_period; /* separation of two instances (period) */
1433 u64 dl_bw; /* dl_runtime / dl_deadline */
1436 * Actual scheduling parameters. Initialized with the values above,
1437 * they are continously updated during task execution. Note that
1438 * the remaining runtime could be < 0 in case we are in overrun.
1440 s64 runtime; /* remaining runtime for this instance */
1441 u64 deadline; /* absolute deadline for this instance */
1442 unsigned int flags; /* specifying the scheduler behaviour */
1447 * @dl_throttled tells if we exhausted the runtime. If so, the
1448 * task has to wait for a replenishment to be performed at the
1449 * next firing of dl_timer.
1451 * @dl_boosted tells if we are boosted due to DI. If so we are
1452 * outside bandwidth enforcement mechanism (but only until we
1453 * exit the critical section);
1455 * @dl_yielded tells if task gave up the cpu before consuming
1456 * all its available runtime during the last job.
1458 int dl_throttled, dl_boosted, dl_yielded;
1461 * Bandwidth enforcement timer. Each -deadline task has its
1462 * own bandwidth to be enforced, thus we need one timer per task.
1464 struct hrtimer dl_timer;
1472 u8 pad; /* Otherwise the compiler can store garbage here. */
1474 u32 s; /* Set of bits. */
1478 enum perf_event_task_context {
1479 perf_invalid_context = -1,
1480 perf_hw_context = 0,
1482 perf_nr_task_contexts,
1485 /* Track pages that require TLB flushes */
1486 struct tlbflush_unmap_batch {
1488 * Each bit set is a CPU that potentially has a TLB entry for one of
1489 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1491 struct cpumask cpumask;
1493 /* True if any bit in cpumask is set */
1494 bool flush_required;
1497 * If true then the PTE was dirty when unmapped. The entry must be
1498 * flushed before IO is initiated or a stale TLB entry potentially
1499 * allows an update without redirtying the page.
1504 struct task_struct {
1505 #ifdef CONFIG_THREAD_INFO_IN_TASK
1507 * For reasons of header soup (see current_thread_info()), this
1508 * must be the first element of task_struct.
1510 struct thread_info thread_info;
1512 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1515 unsigned int flags; /* per process flags, defined below */
1516 unsigned int ptrace;
1519 struct llist_node wake_entry;
1521 #ifdef CONFIG_THREAD_INFO_IN_TASK
1522 unsigned int cpu; /* current CPU */
1524 unsigned int wakee_flips;
1525 unsigned long wakee_flip_decay_ts;
1526 struct task_struct *last_wakee;
1532 int prio, static_prio, normal_prio;
1533 unsigned int rt_priority;
1534 const struct sched_class *sched_class;
1535 struct sched_entity se;
1536 struct sched_rt_entity rt;
1537 #ifdef CONFIG_CGROUP_SCHED
1538 struct task_group *sched_task_group;
1540 struct sched_dl_entity dl;
1542 #ifdef CONFIG_PREEMPT_NOTIFIERS
1543 /* list of struct preempt_notifier: */
1544 struct hlist_head preempt_notifiers;
1547 #ifdef CONFIG_BLK_DEV_IO_TRACE
1548 unsigned int btrace_seq;
1551 unsigned int policy;
1552 int nr_cpus_allowed;
1553 cpumask_t cpus_allowed;
1555 #ifdef CONFIG_PREEMPT_RCU
1556 int rcu_read_lock_nesting;
1557 union rcu_special rcu_read_unlock_special;
1558 struct list_head rcu_node_entry;
1559 struct rcu_node *rcu_blocked_node;
1560 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1561 #ifdef CONFIG_TASKS_RCU
1562 unsigned long rcu_tasks_nvcsw;
1563 bool rcu_tasks_holdout;
1564 struct list_head rcu_tasks_holdout_list;
1565 int rcu_tasks_idle_cpu;
1566 #endif /* #ifdef CONFIG_TASKS_RCU */
1568 #ifdef CONFIG_SCHED_INFO
1569 struct sched_info sched_info;
1572 struct list_head tasks;
1574 struct plist_node pushable_tasks;
1575 struct rb_node pushable_dl_tasks;
1578 struct mm_struct *mm, *active_mm;
1579 /* per-thread vma caching */
1580 u32 vmacache_seqnum;
1581 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1582 #if defined(SPLIT_RSS_COUNTING)
1583 struct task_rss_stat rss_stat;
1587 int exit_code, exit_signal;
1588 int pdeath_signal; /* The signal sent when the parent dies */
1589 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1591 /* Used for emulating ABI behavior of previous Linux versions */
1592 unsigned int personality;
1594 /* scheduler bits, serialized by scheduler locks */
1595 unsigned sched_reset_on_fork:1;
1596 unsigned sched_contributes_to_load:1;
1597 unsigned sched_migrated:1;
1598 unsigned sched_remote_wakeup:1;
1599 unsigned :0; /* force alignment to the next boundary */
1601 /* unserialized, strictly 'current' */
1602 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1603 unsigned in_iowait:1;
1604 #if !defined(TIF_RESTORE_SIGMASK)
1605 unsigned restore_sigmask:1;
1608 unsigned memcg_may_oom:1;
1610 unsigned memcg_kmem_skip_account:1;
1613 #ifdef CONFIG_COMPAT_BRK
1614 unsigned brk_randomized:1;
1617 unsigned long atomic_flags; /* Flags needing atomic access. */
1619 struct restart_block restart_block;
1624 #ifdef CONFIG_CC_STACKPROTECTOR
1625 /* Canary value for the -fstack-protector gcc feature */
1626 unsigned long stack_canary;
1629 * pointers to (original) parent process, youngest child, younger sibling,
1630 * older sibling, respectively. (p->father can be replaced with
1631 * p->real_parent->pid)
1633 struct task_struct __rcu *real_parent; /* real parent process */
1634 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1636 * children/sibling forms the list of my natural children
1638 struct list_head children; /* list of my children */
1639 struct list_head sibling; /* linkage in my parent's children list */
1640 struct task_struct *group_leader; /* threadgroup leader */
1643 * ptraced is the list of tasks this task is using ptrace on.
1644 * This includes both natural children and PTRACE_ATTACH targets.
1645 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1647 struct list_head ptraced;
1648 struct list_head ptrace_entry;
1650 /* PID/PID hash table linkage. */
1651 struct pid_link pids[PIDTYPE_MAX];
1652 struct list_head thread_group;
1653 struct list_head thread_node;
1655 struct completion *vfork_done; /* for vfork() */
1656 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1657 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1660 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1661 u64 utimescaled, stimescaled;
1664 struct prev_cputime prev_cputime;
1665 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1666 seqcount_t vtime_seqcount;
1667 unsigned long long vtime_snap;
1669 /* Task is sleeping or running in a CPU with VTIME inactive */
1671 /* Task runs in userspace in a CPU with VTIME active */
1673 /* Task runs in kernelspace in a CPU with VTIME active */
1675 } vtime_snap_whence;
1678 #ifdef CONFIG_NO_HZ_FULL
1679 atomic_t tick_dep_mask;
1681 unsigned long nvcsw, nivcsw; /* context switch counts */
1682 u64 start_time; /* monotonic time in nsec */
1683 u64 real_start_time; /* boot based time in nsec */
1684 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1685 unsigned long min_flt, maj_flt;
1687 #ifdef CONFIG_POSIX_TIMERS
1688 struct task_cputime cputime_expires;
1689 struct list_head cpu_timers[3];
1692 /* process credentials */
1693 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
1694 const struct cred __rcu *real_cred; /* objective and real subjective task
1695 * credentials (COW) */
1696 const struct cred __rcu *cred; /* effective (overridable) subjective task
1697 * credentials (COW) */
1698 char comm[TASK_COMM_LEN]; /* executable name excluding path
1699 - access with [gs]et_task_comm (which lock
1700 it with task_lock())
1701 - initialized normally by setup_new_exec */
1702 /* file system info */
1703 struct nameidata *nameidata;
1704 #ifdef CONFIG_SYSVIPC
1706 struct sysv_sem sysvsem;
1707 struct sysv_shm sysvshm;
1709 #ifdef CONFIG_DETECT_HUNG_TASK
1710 /* hung task detection */
1711 unsigned long last_switch_count;
1713 /* filesystem information */
1714 struct fs_struct *fs;
1715 /* open file information */
1716 struct files_struct *files;
1718 struct nsproxy *nsproxy;
1719 /* signal handlers */
1720 struct signal_struct *signal;
1721 struct sighand_struct *sighand;
1723 sigset_t blocked, real_blocked;
1724 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1725 struct sigpending pending;
1727 unsigned long sas_ss_sp;
1729 unsigned sas_ss_flags;
1731 struct callback_head *task_works;
1733 struct audit_context *audit_context;
1734 #ifdef CONFIG_AUDITSYSCALL
1736 unsigned int sessionid;
1738 struct seccomp seccomp;
1740 /* Thread group tracking */
1743 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1745 spinlock_t alloc_lock;
1747 /* Protection of the PI data structures: */
1748 raw_spinlock_t pi_lock;
1750 struct wake_q_node wake_q;
1752 #ifdef CONFIG_RT_MUTEXES
1753 /* PI waiters blocked on a rt_mutex held by this task */
1754 struct rb_root pi_waiters;
1755 struct rb_node *pi_waiters_leftmost;
1756 /* Deadlock detection and priority inheritance handling */
1757 struct rt_mutex_waiter *pi_blocked_on;
1760 #ifdef CONFIG_DEBUG_MUTEXES
1761 /* mutex deadlock detection */
1762 struct mutex_waiter *blocked_on;
1764 #ifdef CONFIG_TRACE_IRQFLAGS
1765 unsigned int irq_events;
1766 unsigned long hardirq_enable_ip;
1767 unsigned long hardirq_disable_ip;
1768 unsigned int hardirq_enable_event;
1769 unsigned int hardirq_disable_event;
1770 int hardirqs_enabled;
1771 int hardirq_context;
1772 unsigned long softirq_disable_ip;
1773 unsigned long softirq_enable_ip;
1774 unsigned int softirq_disable_event;
1775 unsigned int softirq_enable_event;
1776 int softirqs_enabled;
1777 int softirq_context;
1779 #ifdef CONFIG_LOCKDEP
1780 # define MAX_LOCK_DEPTH 48UL
1783 unsigned int lockdep_recursion;
1784 struct held_lock held_locks[MAX_LOCK_DEPTH];
1785 gfp_t lockdep_reclaim_gfp;
1788 unsigned int in_ubsan;
1791 /* journalling filesystem info */
1794 /* stacked block device info */
1795 struct bio_list *bio_list;
1798 /* stack plugging */
1799 struct blk_plug *plug;
1803 struct reclaim_state *reclaim_state;
1805 struct backing_dev_info *backing_dev_info;
1807 struct io_context *io_context;
1809 unsigned long ptrace_message;
1810 siginfo_t *last_siginfo; /* For ptrace use. */
1811 struct task_io_accounting ioac;
1812 #if defined(CONFIG_TASK_XACCT)
1813 u64 acct_rss_mem1; /* accumulated rss usage */
1814 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1815 u64 acct_timexpd; /* stime + utime since last update */
1817 #ifdef CONFIG_CPUSETS
1818 nodemask_t mems_allowed; /* Protected by alloc_lock */
1819 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1820 int cpuset_mem_spread_rotor;
1821 int cpuset_slab_spread_rotor;
1823 #ifdef CONFIG_CGROUPS
1824 /* Control Group info protected by css_set_lock */
1825 struct css_set __rcu *cgroups;
1826 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1827 struct list_head cg_list;
1829 #ifdef CONFIG_INTEL_RDT_A
1833 struct robust_list_head __user *robust_list;
1834 #ifdef CONFIG_COMPAT
1835 struct compat_robust_list_head __user *compat_robust_list;
1837 struct list_head pi_state_list;
1838 struct futex_pi_state *pi_state_cache;
1840 #ifdef CONFIG_PERF_EVENTS
1841 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1842 struct mutex perf_event_mutex;
1843 struct list_head perf_event_list;
1845 #ifdef CONFIG_DEBUG_PREEMPT
1846 unsigned long preempt_disable_ip;
1849 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1851 short pref_node_fork;
1853 #ifdef CONFIG_NUMA_BALANCING
1855 unsigned int numa_scan_period;
1856 unsigned int numa_scan_period_max;
1857 int numa_preferred_nid;
1858 unsigned long numa_migrate_retry;
1859 u64 node_stamp; /* migration stamp */
1860 u64 last_task_numa_placement;
1861 u64 last_sum_exec_runtime;
1862 struct callback_head numa_work;
1864 struct list_head numa_entry;
1865 struct numa_group *numa_group;
1868 * numa_faults is an array split into four regions:
1869 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1870 * in this precise order.
1872 * faults_memory: Exponential decaying average of faults on a per-node
1873 * basis. Scheduling placement decisions are made based on these
1874 * counts. The values remain static for the duration of a PTE scan.
1875 * faults_cpu: Track the nodes the process was running on when a NUMA
1876 * hinting fault was incurred.
1877 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1878 * during the current scan window. When the scan completes, the counts
1879 * in faults_memory and faults_cpu decay and these values are copied.
1881 unsigned long *numa_faults;
1882 unsigned long total_numa_faults;
1885 * numa_faults_locality tracks if faults recorded during the last
1886 * scan window were remote/local or failed to migrate. The task scan
1887 * period is adapted based on the locality of the faults with different
1888 * weights depending on whether they were shared or private faults
1890 unsigned long numa_faults_locality[3];
1892 unsigned long numa_pages_migrated;
1893 #endif /* CONFIG_NUMA_BALANCING */
1895 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1896 struct tlbflush_unmap_batch tlb_ubc;
1899 struct rcu_head rcu;
1902 * cache last used pipe for splice
1904 struct pipe_inode_info *splice_pipe;
1906 struct page_frag task_frag;
1908 #ifdef CONFIG_TASK_DELAY_ACCT
1909 struct task_delay_info *delays;
1911 #ifdef CONFIG_FAULT_INJECTION
1915 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1916 * balance_dirty_pages() for some dirty throttling pause
1919 int nr_dirtied_pause;
1920 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1922 #ifdef CONFIG_LATENCYTOP
1923 int latency_record_count;
1924 struct latency_record latency_record[LT_SAVECOUNT];
1927 * time slack values; these are used to round up poll() and
1928 * select() etc timeout values. These are in nanoseconds.
1931 u64 default_timer_slack_ns;
1934 unsigned int kasan_depth;
1936 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1937 /* Index of current stored address in ret_stack */
1939 /* Stack of return addresses for return function tracing */
1940 struct ftrace_ret_stack *ret_stack;
1941 /* time stamp for last schedule */
1942 unsigned long long ftrace_timestamp;
1944 * Number of functions that haven't been traced
1945 * because of depth overrun.
1947 atomic_t trace_overrun;
1948 /* Pause for the tracing */
1949 atomic_t tracing_graph_pause;
1951 #ifdef CONFIG_TRACING
1952 /* state flags for use by tracers */
1953 unsigned long trace;
1954 /* bitmask and counter of trace recursion */
1955 unsigned long trace_recursion;
1956 #endif /* CONFIG_TRACING */
1958 /* Coverage collection mode enabled for this task (0 if disabled). */
1959 enum kcov_mode kcov_mode;
1960 /* Size of the kcov_area. */
1962 /* Buffer for coverage collection. */
1964 /* kcov desciptor wired with this task or NULL. */
1968 struct mem_cgroup *memcg_in_oom;
1969 gfp_t memcg_oom_gfp_mask;
1970 int memcg_oom_order;
1972 /* number of pages to reclaim on returning to userland */
1973 unsigned int memcg_nr_pages_over_high;
1975 #ifdef CONFIG_UPROBES
1976 struct uprobe_task *utask;
1978 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1979 unsigned int sequential_io;
1980 unsigned int sequential_io_avg;
1982 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1983 unsigned long task_state_change;
1985 int pagefault_disabled;
1987 struct task_struct *oom_reaper_list;
1989 #ifdef CONFIG_VMAP_STACK
1990 struct vm_struct *stack_vm_area;
1992 #ifdef CONFIG_THREAD_INFO_IN_TASK
1993 /* A live task holds one reference. */
1994 atomic_t stack_refcount;
1996 /* CPU-specific state of this task */
1997 struct thread_struct thread;
1999 * WARNING: on x86, 'thread_struct' contains a variable-sized
2000 * structure. It *MUST* be at the end of 'task_struct'.
2002 * Do not put anything below here!
2006 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
2007 extern int arch_task_struct_size __read_mostly;
2009 # define arch_task_struct_size (sizeof(struct task_struct))
2012 #ifdef CONFIG_VMAP_STACK
2013 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
2015 return t->stack_vm_area;
2018 static inline struct vm_struct *task_stack_vm_area(const struct task_struct *t)
2024 /* Future-safe accessor for struct task_struct's cpus_allowed. */
2025 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
2027 static inline int tsk_nr_cpus_allowed(struct task_struct *p)
2029 return p->nr_cpus_allowed;
2032 #define TNF_MIGRATED 0x01
2033 #define TNF_NO_GROUP 0x02
2034 #define TNF_SHARED 0x04
2035 #define TNF_FAULT_LOCAL 0x08
2036 #define TNF_MIGRATE_FAIL 0x10
2038 static inline bool in_vfork(struct task_struct *tsk)
2043 * need RCU to access ->real_parent if CLONE_VM was used along with
2046 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2049 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2050 * ->real_parent is not necessarily the task doing vfork(), so in
2051 * theory we can't rely on task_lock() if we want to dereference it.
2053 * And in this case we can't trust the real_parent->mm == tsk->mm
2054 * check, it can be false negative. But we do not care, if init or
2055 * another oom-unkillable task does this it should blame itself.
2058 ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
2064 #ifdef CONFIG_NUMA_BALANCING
2065 extern void task_numa_fault(int last_node, int node, int pages, int flags);
2066 extern pid_t task_numa_group_id(struct task_struct *p);
2067 extern void set_numabalancing_state(bool enabled);
2068 extern void task_numa_free(struct task_struct *p);
2069 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
2070 int src_nid, int dst_cpu);
2072 static inline void task_numa_fault(int last_node, int node, int pages,
2076 static inline pid_t task_numa_group_id(struct task_struct *p)
2080 static inline void set_numabalancing_state(bool enabled)
2083 static inline void task_numa_free(struct task_struct *p)
2086 static inline bool should_numa_migrate_memory(struct task_struct *p,
2087 struct page *page, int src_nid, int dst_cpu)
2093 static inline struct pid *task_pid(struct task_struct *task)
2095 return task->pids[PIDTYPE_PID].pid;
2098 static inline struct pid *task_tgid(struct task_struct *task)
2100 return task->group_leader->pids[PIDTYPE_PID].pid;
2104 * Without tasklist or rcu lock it is not safe to dereference
2105 * the result of task_pgrp/task_session even if task == current,
2106 * we can race with another thread doing sys_setsid/sys_setpgid.
2108 static inline struct pid *task_pgrp(struct task_struct *task)
2110 return task->group_leader->pids[PIDTYPE_PGID].pid;
2113 static inline struct pid *task_session(struct task_struct *task)
2115 return task->group_leader->pids[PIDTYPE_SID].pid;
2118 struct pid_namespace;
2121 * the helpers to get the task's different pids as they are seen
2122 * from various namespaces
2124 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2125 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2127 * task_xid_nr_ns() : id seen from the ns specified;
2129 * set_task_vxid() : assigns a virtual id to a task;
2131 * see also pid_nr() etc in include/linux/pid.h
2133 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2134 struct pid_namespace *ns);
2136 static inline pid_t task_pid_nr(struct task_struct *tsk)
2141 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2142 struct pid_namespace *ns)
2144 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2147 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2149 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2153 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2158 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2160 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2162 return pid_vnr(task_tgid(tsk));
2166 static inline int pid_alive(const struct task_struct *p);
2167 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2173 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2179 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2181 return task_ppid_nr_ns(tsk, &init_pid_ns);
2184 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2185 struct pid_namespace *ns)
2187 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2190 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2192 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2196 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2197 struct pid_namespace *ns)
2199 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2202 static inline pid_t task_session_vnr(struct task_struct *tsk)
2204 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2207 /* obsolete, do not use */
2208 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2210 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2214 * pid_alive - check that a task structure is not stale
2215 * @p: Task structure to be checked.
2217 * Test if a process is not yet dead (at most zombie state)
2218 * If pid_alive fails, then pointers within the task structure
2219 * can be stale and must not be dereferenced.
2221 * Return: 1 if the process is alive. 0 otherwise.
2223 static inline int pid_alive(const struct task_struct *p)
2225 return p->pids[PIDTYPE_PID].pid != NULL;
2229 * is_global_init - check if a task structure is init. Since init
2230 * is free to have sub-threads we need to check tgid.
2231 * @tsk: Task structure to be checked.
2233 * Check if a task structure is the first user space task the kernel created.
2235 * Return: 1 if the task structure is init. 0 otherwise.
2237 static inline int is_global_init(struct task_struct *tsk)
2239 return task_tgid_nr(tsk) == 1;
2242 extern struct pid *cad_pid;
2244 extern void free_task(struct task_struct *tsk);
2245 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2247 extern void __put_task_struct(struct task_struct *t);
2249 static inline void put_task_struct(struct task_struct *t)
2251 if (atomic_dec_and_test(&t->usage))
2252 __put_task_struct(t);
2255 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
2256 struct task_struct *try_get_task_struct(struct task_struct **ptask);
2258 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2259 extern void task_cputime(struct task_struct *t,
2260 u64 *utime, u64 *stime);
2261 extern u64 task_gtime(struct task_struct *t);
2263 static inline void task_cputime(struct task_struct *t,
2264 u64 *utime, u64 *stime)
2270 static inline u64 task_gtime(struct task_struct *t)
2276 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
2277 static inline void task_cputime_scaled(struct task_struct *t,
2281 *utimescaled = t->utimescaled;
2282 *stimescaled = t->stimescaled;
2285 static inline void task_cputime_scaled(struct task_struct *t,
2289 task_cputime(t, utimescaled, stimescaled);
2293 extern void task_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2294 extern void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
2299 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
2300 #define PF_EXITING 0x00000004 /* getting shut down */
2301 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2302 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2303 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2304 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2305 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2306 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2307 #define PF_DUMPCORE 0x00000200 /* dumped core */
2308 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2309 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2310 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2311 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2312 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2313 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2314 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2315 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2316 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2317 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2318 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2319 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2320 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2321 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2322 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2323 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2324 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2325 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2326 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2329 * Only the _current_ task can read/write to tsk->flags, but other
2330 * tasks can access tsk->flags in readonly mode for example
2331 * with tsk_used_math (like during threaded core dumping).
2332 * There is however an exception to this rule during ptrace
2333 * or during fork: the ptracer task is allowed to write to the
2334 * child->flags of its traced child (same goes for fork, the parent
2335 * can write to the child->flags), because we're guaranteed the
2336 * child is not running and in turn not changing child->flags
2337 * at the same time the parent does it.
2339 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2340 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2341 #define clear_used_math() clear_stopped_child_used_math(current)
2342 #define set_used_math() set_stopped_child_used_math(current)
2343 #define conditional_stopped_child_used_math(condition, child) \
2344 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2345 #define conditional_used_math(condition) \
2346 conditional_stopped_child_used_math(condition, current)
2347 #define copy_to_stopped_child_used_math(child) \
2348 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2349 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2350 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2351 #define used_math() tsk_used_math(current)
2353 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2354 * __GFP_FS is also cleared as it implies __GFP_IO.
2356 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2358 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2359 flags &= ~(__GFP_IO | __GFP_FS);
2363 static inline unsigned int memalloc_noio_save(void)
2365 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2366 current->flags |= PF_MEMALLOC_NOIO;
2370 static inline void memalloc_noio_restore(unsigned int flags)
2372 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2375 /* Per-process atomic flags. */
2376 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2377 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2378 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2379 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2382 #define TASK_PFA_TEST(name, func) \
2383 static inline bool task_##func(struct task_struct *p) \
2384 { return test_bit(PFA_##name, &p->atomic_flags); }
2385 #define TASK_PFA_SET(name, func) \
2386 static inline void task_set_##func(struct task_struct *p) \
2387 { set_bit(PFA_##name, &p->atomic_flags); }
2388 #define TASK_PFA_CLEAR(name, func) \
2389 static inline void task_clear_##func(struct task_struct *p) \
2390 { clear_bit(PFA_##name, &p->atomic_flags); }
2392 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2393 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2395 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2396 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2397 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2399 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2400 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2401 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2403 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
2404 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
2407 * task->jobctl flags
2409 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2411 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2412 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2413 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2414 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2415 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2416 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2417 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2419 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2420 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2421 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2422 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2423 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2424 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2425 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2427 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2428 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2430 extern bool task_set_jobctl_pending(struct task_struct *task,
2431 unsigned long mask);
2432 extern void task_clear_jobctl_trapping(struct task_struct *task);
2433 extern void task_clear_jobctl_pending(struct task_struct *task,
2434 unsigned long mask);
2436 static inline void rcu_copy_process(struct task_struct *p)
2438 #ifdef CONFIG_PREEMPT_RCU
2439 p->rcu_read_lock_nesting = 0;
2440 p->rcu_read_unlock_special.s = 0;
2441 p->rcu_blocked_node = NULL;
2442 INIT_LIST_HEAD(&p->rcu_node_entry);
2443 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2444 #ifdef CONFIG_TASKS_RCU
2445 p->rcu_tasks_holdout = false;
2446 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2447 p->rcu_tasks_idle_cpu = -1;
2448 #endif /* #ifdef CONFIG_TASKS_RCU */
2451 static inline void tsk_restore_flags(struct task_struct *task,
2452 unsigned long orig_flags, unsigned long flags)
2454 task->flags &= ~flags;
2455 task->flags |= orig_flags & flags;
2458 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2459 const struct cpumask *trial);
2460 extern int task_can_attach(struct task_struct *p,
2461 const struct cpumask *cs_cpus_allowed);
2463 extern void do_set_cpus_allowed(struct task_struct *p,
2464 const struct cpumask *new_mask);
2466 extern int set_cpus_allowed_ptr(struct task_struct *p,
2467 const struct cpumask *new_mask);
2469 static inline void do_set_cpus_allowed(struct task_struct *p,
2470 const struct cpumask *new_mask)
2473 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2474 const struct cpumask *new_mask)
2476 if (!cpumask_test_cpu(0, new_mask))
2482 #ifdef CONFIG_NO_HZ_COMMON
2483 void calc_load_enter_idle(void);
2484 void calc_load_exit_idle(void);
2486 static inline void calc_load_enter_idle(void) { }
2487 static inline void calc_load_exit_idle(void) { }
2488 #endif /* CONFIG_NO_HZ_COMMON */
2490 #ifndef cpu_relax_yield
2491 #define cpu_relax_yield() cpu_relax()
2495 * Do not use outside of architecture code which knows its limitations.
2497 * sched_clock() has no promise of monotonicity or bounded drift between
2498 * CPUs, use (which you should not) requires disabling IRQs.
2500 * Please use one of the three interfaces below.
2502 extern unsigned long long notrace sched_clock(void);
2504 * See the comment in kernel/sched/clock.c
2506 extern u64 running_clock(void);
2507 extern u64 sched_clock_cpu(int cpu);
2510 extern void sched_clock_init(void);
2512 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2513 static inline void sched_clock_init_late(void)
2517 static inline void sched_clock_tick(void)
2521 static inline void clear_sched_clock_stable(void)
2525 static inline void sched_clock_idle_sleep_event(void)
2529 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2533 static inline u64 cpu_clock(int cpu)
2535 return sched_clock();
2538 static inline u64 local_clock(void)
2540 return sched_clock();
2543 extern void sched_clock_init_late(void);
2545 * Architectures can set this to 1 if they have specified
2546 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2547 * but then during bootup it turns out that sched_clock()
2548 * is reliable after all:
2550 extern int sched_clock_stable(void);
2551 extern void clear_sched_clock_stable(void);
2553 extern void sched_clock_tick(void);
2554 extern void sched_clock_idle_sleep_event(void);
2555 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2558 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2559 * time source that is monotonic per cpu argument and has bounded drift
2562 * ######################### BIG FAT WARNING ##########################
2563 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2564 * # go backwards !! #
2565 * ####################################################################
2567 static inline u64 cpu_clock(int cpu)
2569 return sched_clock_cpu(cpu);
2572 static inline u64 local_clock(void)
2574 return sched_clock_cpu(raw_smp_processor_id());
2578 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2580 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2581 * The reason for this explicit opt-in is not to have perf penalty with
2582 * slow sched_clocks.
2584 extern void enable_sched_clock_irqtime(void);
2585 extern void disable_sched_clock_irqtime(void);
2587 static inline void enable_sched_clock_irqtime(void) {}
2588 static inline void disable_sched_clock_irqtime(void) {}
2591 extern unsigned long long
2592 task_sched_runtime(struct task_struct *task);
2594 /* sched_exec is called by processes performing an exec */
2596 extern void sched_exec(void);
2598 #define sched_exec() {}
2601 extern void sched_clock_idle_sleep_event(void);
2602 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2604 #ifdef CONFIG_HOTPLUG_CPU
2605 extern void idle_task_exit(void);
2607 static inline void idle_task_exit(void) {}
2610 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2611 extern void wake_up_nohz_cpu(int cpu);
2613 static inline void wake_up_nohz_cpu(int cpu) { }
2616 #ifdef CONFIG_NO_HZ_FULL
2617 extern u64 scheduler_tick_max_deferment(void);
2620 #ifdef CONFIG_SCHED_AUTOGROUP
2621 extern void sched_autogroup_create_attach(struct task_struct *p);
2622 extern void sched_autogroup_detach(struct task_struct *p);
2623 extern void sched_autogroup_fork(struct signal_struct *sig);
2624 extern void sched_autogroup_exit(struct signal_struct *sig);
2625 extern void sched_autogroup_exit_task(struct task_struct *p);
2626 #ifdef CONFIG_PROC_FS
2627 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2628 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2631 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2632 static inline void sched_autogroup_detach(struct task_struct *p) { }
2633 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2634 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2635 static inline void sched_autogroup_exit_task(struct task_struct *p) { }
2638 extern int yield_to(struct task_struct *p, bool preempt);
2639 extern void set_user_nice(struct task_struct *p, long nice);
2640 extern int task_prio(const struct task_struct *p);
2642 * task_nice - return the nice value of a given task.
2643 * @p: the task in question.
2645 * Return: The nice value [ -20 ... 0 ... 19 ].
2647 static inline int task_nice(const struct task_struct *p)
2649 return PRIO_TO_NICE((p)->static_prio);
2651 extern int can_nice(const struct task_struct *p, const int nice);
2652 extern int task_curr(const struct task_struct *p);
2653 extern int idle_cpu(int cpu);
2654 extern int sched_setscheduler(struct task_struct *, int,
2655 const struct sched_param *);
2656 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2657 const struct sched_param *);
2658 extern int sched_setattr(struct task_struct *,
2659 const struct sched_attr *);
2660 extern struct task_struct *idle_task(int cpu);
2662 * is_idle_task - is the specified task an idle task?
2663 * @p: the task in question.
2665 * Return: 1 if @p is an idle task. 0 otherwise.
2667 static inline bool is_idle_task(const struct task_struct *p)
2669 return !!(p->flags & PF_IDLE);
2671 extern struct task_struct *curr_task(int cpu);
2672 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
2676 union thread_union {
2677 #ifndef CONFIG_THREAD_INFO_IN_TASK
2678 struct thread_info thread_info;
2680 unsigned long stack[THREAD_SIZE/sizeof(long)];
2683 #ifndef __HAVE_ARCH_KSTACK_END
2684 static inline int kstack_end(void *addr)
2686 /* Reliable end of stack detection:
2687 * Some APM bios versions misalign the stack
2689 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2693 extern union thread_union init_thread_union;
2694 extern struct task_struct init_task;
2696 extern struct mm_struct init_mm;
2698 extern struct pid_namespace init_pid_ns;
2701 * find a task by one of its numerical ids
2703 * find_task_by_pid_ns():
2704 * finds a task by its pid in the specified namespace
2705 * find_task_by_vpid():
2706 * finds a task by its virtual pid
2708 * see also find_vpid() etc in include/linux/pid.h
2711 extern struct task_struct *find_task_by_vpid(pid_t nr);
2712 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2713 struct pid_namespace *ns);
2715 /* per-UID process charging. */
2716 extern struct user_struct * alloc_uid(kuid_t);
2717 static inline struct user_struct *get_uid(struct user_struct *u)
2719 atomic_inc(&u->__count);
2722 extern void free_uid(struct user_struct *);
2724 #include <asm/current.h>
2726 extern void xtime_update(unsigned long ticks);
2728 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2729 extern int wake_up_process(struct task_struct *tsk);
2730 extern void wake_up_new_task(struct task_struct *tsk);
2732 extern void kick_process(struct task_struct *tsk);
2734 static inline void kick_process(struct task_struct *tsk) { }
2736 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2737 extern void sched_dead(struct task_struct *p);
2739 extern void proc_caches_init(void);
2740 extern void flush_signals(struct task_struct *);
2741 extern void ignore_signals(struct task_struct *);
2742 extern void flush_signal_handlers(struct task_struct *, int force_default);
2743 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2745 static inline int kernel_dequeue_signal(siginfo_t *info)
2747 struct task_struct *tsk = current;
2751 spin_lock_irq(&tsk->sighand->siglock);
2752 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2753 spin_unlock_irq(&tsk->sighand->siglock);
2758 static inline void kernel_signal_stop(void)
2760 spin_lock_irq(¤t->sighand->siglock);
2761 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2762 __set_current_state(TASK_STOPPED);
2763 spin_unlock_irq(¤t->sighand->siglock);
2768 extern void release_task(struct task_struct * p);
2769 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2770 extern int force_sigsegv(int, struct task_struct *);
2771 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2772 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2773 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2774 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2775 const struct cred *, u32);
2776 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2777 extern int kill_pid(struct pid *pid, int sig, int priv);
2778 extern int kill_proc_info(int, struct siginfo *, pid_t);
2779 extern __must_check bool do_notify_parent(struct task_struct *, int);
2780 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2781 extern void force_sig(int, struct task_struct *);
2782 extern int send_sig(int, struct task_struct *, int);
2783 extern int zap_other_threads(struct task_struct *p);
2784 extern struct sigqueue *sigqueue_alloc(void);
2785 extern void sigqueue_free(struct sigqueue *);
2786 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2787 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2789 #ifdef TIF_RESTORE_SIGMASK
2791 * Legacy restore_sigmask accessors. These are inefficient on
2792 * SMP architectures because they require atomic operations.
2796 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2798 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2799 * will run before returning to user mode, to process the flag. For
2800 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2801 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2802 * arch code will notice on return to user mode, in case those bits
2803 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2804 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2806 static inline void set_restore_sigmask(void)
2808 set_thread_flag(TIF_RESTORE_SIGMASK);
2809 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2811 static inline void clear_restore_sigmask(void)
2813 clear_thread_flag(TIF_RESTORE_SIGMASK);
2815 static inline bool test_restore_sigmask(void)
2817 return test_thread_flag(TIF_RESTORE_SIGMASK);
2819 static inline bool test_and_clear_restore_sigmask(void)
2821 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK);
2824 #else /* TIF_RESTORE_SIGMASK */
2826 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2827 static inline void set_restore_sigmask(void)
2829 current->restore_sigmask = true;
2830 WARN_ON(!test_thread_flag(TIF_SIGPENDING));
2832 static inline void clear_restore_sigmask(void)
2834 current->restore_sigmask = false;
2836 static inline bool test_restore_sigmask(void)
2838 return current->restore_sigmask;
2840 static inline bool test_and_clear_restore_sigmask(void)
2842 if (!current->restore_sigmask)
2844 current->restore_sigmask = false;
2849 static inline void restore_saved_sigmask(void)
2851 if (test_and_clear_restore_sigmask())
2852 __set_current_blocked(¤t->saved_sigmask);
2855 static inline sigset_t *sigmask_to_save(void)
2857 sigset_t *res = ¤t->blocked;
2858 if (unlikely(test_restore_sigmask()))
2859 res = ¤t->saved_sigmask;
2863 static inline int kill_cad_pid(int sig, int priv)
2865 return kill_pid(cad_pid, sig, priv);
2868 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2869 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2870 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2871 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2874 * True if we are on the alternate signal stack.
2876 static inline int on_sig_stack(unsigned long sp)
2879 * If the signal stack is SS_AUTODISARM then, by construction, we
2880 * can't be on the signal stack unless user code deliberately set
2881 * SS_AUTODISARM when we were already on it.
2883 * This improves reliability: if user state gets corrupted such that
2884 * the stack pointer points very close to the end of the signal stack,
2885 * then this check will enable the signal to be handled anyway.
2887 if (current->sas_ss_flags & SS_AUTODISARM)
2890 #ifdef CONFIG_STACK_GROWSUP
2891 return sp >= current->sas_ss_sp &&
2892 sp - current->sas_ss_sp < current->sas_ss_size;
2894 return sp > current->sas_ss_sp &&
2895 sp - current->sas_ss_sp <= current->sas_ss_size;
2899 static inline int sas_ss_flags(unsigned long sp)
2901 if (!current->sas_ss_size)
2904 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2907 static inline void sas_ss_reset(struct task_struct *p)
2911 p->sas_ss_flags = SS_DISABLE;
2914 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2916 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2917 #ifdef CONFIG_STACK_GROWSUP
2918 return current->sas_ss_sp;
2920 return current->sas_ss_sp + current->sas_ss_size;
2926 * Routines for handling mm_structs
2928 extern struct mm_struct * mm_alloc(void);
2930 /* mmdrop drops the mm and the page tables */
2931 extern void __mmdrop(struct mm_struct *);
2932 static inline void mmdrop(struct mm_struct *mm)
2934 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2938 static inline void mmdrop_async_fn(struct work_struct *work)
2940 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
2944 static inline void mmdrop_async(struct mm_struct *mm)
2946 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
2947 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
2948 schedule_work(&mm->async_put_work);
2952 static inline bool mmget_not_zero(struct mm_struct *mm)
2954 return atomic_inc_not_zero(&mm->mm_users);
2957 /* mmput gets rid of the mappings and all user-space */
2958 extern void mmput(struct mm_struct *);
2960 /* same as above but performs the slow path from the async context. Can
2961 * be called from the atomic context as well
2963 extern void mmput_async(struct mm_struct *);
2966 /* Grab a reference to a task's mm, if it is not already going away */
2967 extern struct mm_struct *get_task_mm(struct task_struct *task);
2969 * Grab a reference to a task's mm, if it is not already going away
2970 * and ptrace_may_access with the mode parameter passed to it
2973 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2974 /* Remove the current tasks stale references to the old mm_struct */
2975 extern void mm_release(struct task_struct *, struct mm_struct *);
2977 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2978 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2979 struct task_struct *, unsigned long);
2981 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2982 struct task_struct *);
2984 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2985 * via pt_regs, so ignore the tls argument passed via C. */
2986 static inline int copy_thread_tls(
2987 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2988 struct task_struct *p, unsigned long tls)
2990 return copy_thread(clone_flags, sp, arg, p);
2993 extern void flush_thread(void);
2995 #ifdef CONFIG_HAVE_EXIT_THREAD
2996 extern void exit_thread(struct task_struct *tsk);
2998 static inline void exit_thread(struct task_struct *tsk)
3003 extern void exit_files(struct task_struct *);
3004 extern void __cleanup_sighand(struct sighand_struct *);
3006 extern void exit_itimers(struct signal_struct *);
3007 extern void flush_itimer_signals(void);
3009 extern void do_group_exit(int);
3011 extern int do_execve(struct filename *,
3012 const char __user * const __user *,
3013 const char __user * const __user *);
3014 extern int do_execveat(int, struct filename *,
3015 const char __user * const __user *,
3016 const char __user * const __user *,
3018 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
3019 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
3020 struct task_struct *fork_idle(int);
3021 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
3023 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
3024 static inline void set_task_comm(struct task_struct *tsk, const char *from)
3026 __set_task_comm(tsk, from, false);
3028 extern char *get_task_comm(char *to, struct task_struct *tsk);
3031 void scheduler_ipi(void);
3032 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
3034 static inline void scheduler_ipi(void) { }
3035 static inline unsigned long wait_task_inactive(struct task_struct *p,
3042 #define tasklist_empty() \
3043 list_empty(&init_task.tasks)
3045 #define next_task(p) \
3046 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
3048 #define for_each_process(p) \
3049 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
3051 extern bool current_is_single_threaded(void);
3054 * Careful: do_each_thread/while_each_thread is a double loop so
3055 * 'break' will not work as expected - use goto instead.
3057 #define do_each_thread(g, t) \
3058 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3060 #define while_each_thread(g, t) \
3061 while ((t = next_thread(t)) != g)
3063 #define __for_each_thread(signal, t) \
3064 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3066 #define for_each_thread(p, t) \
3067 __for_each_thread((p)->signal, t)
3069 /* Careful: this is a double loop, 'break' won't work as expected. */
3070 #define for_each_process_thread(p, t) \
3071 for_each_process(p) for_each_thread(p, t)
3073 static inline int get_nr_threads(struct task_struct *tsk)
3075 return tsk->signal->nr_threads;
3078 static inline bool thread_group_leader(struct task_struct *p)
3080 return p->exit_signal >= 0;
3083 /* Do to the insanities of de_thread it is possible for a process
3084 * to have the pid of the thread group leader without actually being
3085 * the thread group leader. For iteration through the pids in proc
3086 * all we care about is that we have a task with the appropriate
3087 * pid, we don't actually care if we have the right task.
3089 static inline bool has_group_leader_pid(struct task_struct *p)
3091 return task_pid(p) == p->signal->leader_pid;
3095 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
3097 return p1->signal == p2->signal;
3100 static inline struct task_struct *next_thread(const struct task_struct *p)
3102 return list_entry_rcu(p->thread_group.next,
3103 struct task_struct, thread_group);
3106 static inline int thread_group_empty(struct task_struct *p)
3108 return list_empty(&p->thread_group);
3111 #define delay_group_leader(p) \
3112 (thread_group_leader(p) && !thread_group_empty(p))
3115 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3116 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3117 * pins the final release of task.io_context. Also protects ->cpuset and
3118 * ->cgroup.subsys[]. And ->vfork_done.
3120 * Nests both inside and outside of read_lock(&tasklist_lock).
3121 * It must not be nested with write_lock_irq(&tasklist_lock),
3122 * neither inside nor outside.
3124 static inline void task_lock(struct task_struct *p)
3126 spin_lock(&p->alloc_lock);
3129 static inline void task_unlock(struct task_struct *p)
3131 spin_unlock(&p->alloc_lock);
3134 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
3135 unsigned long *flags);
3137 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
3138 unsigned long *flags)
3140 struct sighand_struct *ret;
3142 ret = __lock_task_sighand(tsk, flags);
3143 (void)__cond_lock(&tsk->sighand->siglock, ret);
3147 static inline void unlock_task_sighand(struct task_struct *tsk,
3148 unsigned long *flags)
3150 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
3154 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
3155 * @tsk: task causing the changes
3157 * All operations which modify a threadgroup - a new thread joining the
3158 * group, death of a member thread (the assertion of PF_EXITING) and
3159 * exec(2) dethreading the process and replacing the leader - are wrapped
3160 * by threadgroup_change_{begin|end}(). This is to provide a place which
3161 * subsystems needing threadgroup stability can hook into for
3164 static inline void threadgroup_change_begin(struct task_struct *tsk)
3167 cgroup_threadgroup_change_begin(tsk);
3171 * threadgroup_change_end - mark the end of changes to a threadgroup
3172 * @tsk: task causing the changes
3174 * See threadgroup_change_begin().
3176 static inline void threadgroup_change_end(struct task_struct *tsk)
3178 cgroup_threadgroup_change_end(tsk);
3181 #ifdef CONFIG_THREAD_INFO_IN_TASK
3183 static inline struct thread_info *task_thread_info(struct task_struct *task)
3185 return &task->thread_info;
3189 * When accessing the stack of a non-current task that might exit, use
3190 * try_get_task_stack() instead. task_stack_page will return a pointer
3191 * that could get freed out from under you.
3193 static inline void *task_stack_page(const struct task_struct *task)
3198 #define setup_thread_stack(new,old) do { } while(0)
3200 static inline unsigned long *end_of_stack(const struct task_struct *task)
3205 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3207 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3208 #define task_stack_page(task) ((void *)(task)->stack)
3210 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
3212 *task_thread_info(p) = *task_thread_info(org);
3213 task_thread_info(p)->task = p;
3217 * Return the address of the last usable long on the stack.
3219 * When the stack grows down, this is just above the thread
3220 * info struct. Going any lower will corrupt the threadinfo.
3222 * When the stack grows up, this is the highest address.
3223 * Beyond that position, we corrupt data on the next page.
3225 static inline unsigned long *end_of_stack(struct task_struct *p)
3227 #ifdef CONFIG_STACK_GROWSUP
3228 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
3230 return (unsigned long *)(task_thread_info(p) + 1);
3236 #ifdef CONFIG_THREAD_INFO_IN_TASK
3237 static inline void *try_get_task_stack(struct task_struct *tsk)
3239 return atomic_inc_not_zero(&tsk->stack_refcount) ?
3240 task_stack_page(tsk) : NULL;
3243 extern void put_task_stack(struct task_struct *tsk);
3245 static inline void *try_get_task_stack(struct task_struct *tsk)
3247 return task_stack_page(tsk);
3250 static inline void put_task_stack(struct task_struct *tsk) {}
3253 #define task_stack_end_corrupted(task) \
3254 (*(end_of_stack(task)) != STACK_END_MAGIC)
3256 static inline int object_is_on_stack(void *obj)
3258 void *stack = task_stack_page(current);
3260 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
3263 extern void thread_stack_cache_init(void);
3265 #ifdef CONFIG_DEBUG_STACK_USAGE
3266 static inline unsigned long stack_not_used(struct task_struct *p)
3268 unsigned long *n = end_of_stack(p);
3270 do { /* Skip over canary */
3271 # ifdef CONFIG_STACK_GROWSUP
3278 # ifdef CONFIG_STACK_GROWSUP
3279 return (unsigned long)end_of_stack(p) - (unsigned long)n;
3281 return (unsigned long)n - (unsigned long)end_of_stack(p);
3285 extern void set_task_stack_end_magic(struct task_struct *tsk);
3287 /* set thread flags in other task's structures
3288 * - see asm/thread_info.h for TIF_xxxx flags available
3290 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
3292 set_ti_thread_flag(task_thread_info(tsk), flag);
3295 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3297 clear_ti_thread_flag(task_thread_info(tsk), flag);
3300 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
3302 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
3305 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3307 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3310 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3312 return test_ti_thread_flag(task_thread_info(tsk), flag);
3315 static inline void set_tsk_need_resched(struct task_struct *tsk)
3317 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3320 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3322 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3325 static inline int test_tsk_need_resched(struct task_struct *tsk)
3327 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3330 static inline int restart_syscall(void)
3332 set_tsk_thread_flag(current, TIF_SIGPENDING);
3333 return -ERESTARTNOINTR;
3336 static inline int signal_pending(struct task_struct *p)
3338 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3341 static inline int __fatal_signal_pending(struct task_struct *p)
3343 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3346 static inline int fatal_signal_pending(struct task_struct *p)
3348 return signal_pending(p) && __fatal_signal_pending(p);
3351 static inline int signal_pending_state(long state, struct task_struct *p)
3353 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3355 if (!signal_pending(p))
3358 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3362 * cond_resched() and cond_resched_lock(): latency reduction via
3363 * explicit rescheduling in places that are safe. The return
3364 * value indicates whether a reschedule was done in fact.
3365 * cond_resched_lock() will drop the spinlock before scheduling,
3366 * cond_resched_softirq() will enable bhs before scheduling.
3368 #ifndef CONFIG_PREEMPT
3369 extern int _cond_resched(void);
3371 static inline int _cond_resched(void) { return 0; }
3374 #define cond_resched() ({ \
3375 ___might_sleep(__FILE__, __LINE__, 0); \
3379 extern int __cond_resched_lock(spinlock_t *lock);
3381 #define cond_resched_lock(lock) ({ \
3382 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3383 __cond_resched_lock(lock); \
3386 extern int __cond_resched_softirq(void);
3388 #define cond_resched_softirq() ({ \
3389 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3390 __cond_resched_softirq(); \
3393 static inline void cond_resched_rcu(void)
3395 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3402 static inline unsigned long get_preempt_disable_ip(struct task_struct *p)
3404 #ifdef CONFIG_DEBUG_PREEMPT
3405 return p->preempt_disable_ip;
3412 * Does a critical section need to be broken due to another
3413 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3414 * but a general need for low latency)
3416 static inline int spin_needbreak(spinlock_t *lock)
3418 #ifdef CONFIG_PREEMPT
3419 return spin_is_contended(lock);
3426 * Idle thread specific functions to determine the need_resched
3429 #ifdef TIF_POLLING_NRFLAG
3430 static inline int tsk_is_polling(struct task_struct *p)
3432 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3435 static inline void __current_set_polling(void)
3437 set_thread_flag(TIF_POLLING_NRFLAG);
3440 static inline bool __must_check current_set_polling_and_test(void)
3442 __current_set_polling();
3445 * Polling state must be visible before we test NEED_RESCHED,
3446 * paired by resched_curr()
3448 smp_mb__after_atomic();
3450 return unlikely(tif_need_resched());
3453 static inline void __current_clr_polling(void)
3455 clear_thread_flag(TIF_POLLING_NRFLAG);
3458 static inline bool __must_check current_clr_polling_and_test(void)
3460 __current_clr_polling();
3463 * Polling state must be visible before we test NEED_RESCHED,
3464 * paired by resched_curr()
3466 smp_mb__after_atomic();
3468 return unlikely(tif_need_resched());
3472 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3473 static inline void __current_set_polling(void) { }
3474 static inline void __current_clr_polling(void) { }
3476 static inline bool __must_check current_set_polling_and_test(void)
3478 return unlikely(tif_need_resched());
3480 static inline bool __must_check current_clr_polling_and_test(void)
3482 return unlikely(tif_need_resched());
3486 static inline void current_clr_polling(void)
3488 __current_clr_polling();
3491 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3492 * Once the bit is cleared, we'll get IPIs with every new
3493 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3496 smp_mb(); /* paired with resched_curr() */
3498 preempt_fold_need_resched();
3501 static __always_inline bool need_resched(void)
3503 return unlikely(tif_need_resched());
3507 * Thread group CPU time accounting.
3509 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3510 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3513 * Reevaluate whether the task has signals pending delivery.
3514 * Wake the task if so.
3515 * This is required every time the blocked sigset_t changes.
3516 * callers must hold sighand->siglock.
3518 extern void recalc_sigpending_and_wake(struct task_struct *t);
3519 extern void recalc_sigpending(void);
3521 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3523 static inline void signal_wake_up(struct task_struct *t, bool resume)
3525 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3527 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3529 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3533 * Wrappers for p->thread_info->cpu access. No-op on UP.
3537 static inline unsigned int task_cpu(const struct task_struct *p)
3539 #ifdef CONFIG_THREAD_INFO_IN_TASK
3542 return task_thread_info(p)->cpu;
3546 static inline int task_node(const struct task_struct *p)
3548 return cpu_to_node(task_cpu(p));
3551 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3555 static inline unsigned int task_cpu(const struct task_struct *p)
3560 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3564 #endif /* CONFIG_SMP */
3567 * In order to reduce various lock holder preemption latencies provide an
3568 * interface to see if a vCPU is currently running or not.
3570 * This allows us to terminate optimistic spin loops and block, analogous to
3571 * the native optimistic spin heuristic of testing if the lock owner task is
3574 #ifndef vcpu_is_preempted
3575 # define vcpu_is_preempted(cpu) false
3578 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3579 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3581 #ifdef CONFIG_CGROUP_SCHED
3582 extern struct task_group root_task_group;
3583 #endif /* CONFIG_CGROUP_SCHED */
3585 extern int task_can_switch_user(struct user_struct *up,
3586 struct task_struct *tsk);
3588 #ifdef CONFIG_TASK_XACCT
3589 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3591 tsk->ioac.rchar += amt;
3594 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3596 tsk->ioac.wchar += amt;
3599 static inline void inc_syscr(struct task_struct *tsk)
3604 static inline void inc_syscw(struct task_struct *tsk)
3609 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3613 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3617 static inline void inc_syscr(struct task_struct *tsk)
3621 static inline void inc_syscw(struct task_struct *tsk)
3626 #ifndef TASK_SIZE_OF
3627 #define TASK_SIZE_OF(tsk) TASK_SIZE
3631 extern void mm_update_next_owner(struct mm_struct *mm);
3633 static inline void mm_update_next_owner(struct mm_struct *mm)
3636 #endif /* CONFIG_MEMCG */
3638 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3641 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3644 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3647 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3650 static inline unsigned long rlimit(unsigned int limit)
3652 return task_rlimit(current, limit);
3655 static inline unsigned long rlimit_max(unsigned int limit)
3657 return task_rlimit_max(current, limit);
3660 #define SCHED_CPUFREQ_RT (1U << 0)
3661 #define SCHED_CPUFREQ_DL (1U << 1)
3662 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3664 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3666 #ifdef CONFIG_CPU_FREQ
3667 struct update_util_data {
3668 void (*func)(struct update_util_data *data, u64 time, unsigned int flags);
3671 void cpufreq_add_update_util_hook(int cpu, struct update_util_data *data,
3672 void (*func)(struct update_util_data *data, u64 time,
3673 unsigned int flags));
3674 void cpufreq_remove_update_util_hook(int cpu);
3675 #endif /* CONFIG_CPU_FREQ */