4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
13 #include <asm/param.h> /* for HZ */
15 #include <linux/capability.h>
16 #include <linux/threads.h>
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/timex.h>
20 #include <linux/jiffies.h>
21 #include <linux/plist.h>
22 #include <linux/rbtree.h>
23 #include <linux/thread_info.h>
24 #include <linux/cpumask.h>
25 #include <linux/errno.h>
26 #include <linux/nodemask.h>
27 #include <linux/mm_types.h>
28 #include <linux/preempt.h>
31 #include <asm/ptrace.h>
32 #include <linux/cputime.h>
34 #include <linux/smp.h>
35 #include <linux/sem.h>
36 #include <linux/shm.h>
37 #include <linux/signal.h>
38 #include <linux/compiler.h>
39 #include <linux/completion.h>
40 #include <linux/pid.h>
41 #include <linux/percpu.h>
42 #include <linux/topology.h>
43 #include <linux/proportions.h>
44 #include <linux/seccomp.h>
45 #include <linux/rcupdate.h>
46 #include <linux/rculist.h>
47 #include <linux/rtmutex.h>
49 #include <linux/time.h>
50 #include <linux/param.h>
51 #include <linux/resource.h>
52 #include <linux/timer.h>
53 #include <linux/hrtimer.h>
54 #include <linux/kcov.h>
55 #include <linux/task_io_accounting.h>
56 #include <linux/latencytop.h>
57 #include <linux/cred.h>
58 #include <linux/llist.h>
59 #include <linux/uidgid.h>
60 #include <linux/gfp.h>
61 #include <linux/magic.h>
62 #include <linux/cgroup-defs.h>
64 #include <asm/processor.h>
66 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
69 * Extended scheduling parameters data structure.
71 * This is needed because the original struct sched_param can not be
72 * altered without introducing ABI issues with legacy applications
73 * (e.g., in sched_getparam()).
75 * However, the possibility of specifying more than just a priority for
76 * the tasks may be useful for a wide variety of application fields, e.g.,
77 * multimedia, streaming, automation and control, and many others.
79 * This variant (sched_attr) is meant at describing a so-called
80 * sporadic time-constrained task. In such model a task is specified by:
81 * - the activation period or minimum instance inter-arrival time;
82 * - the maximum (or average, depending on the actual scheduling
83 * discipline) computation time of all instances, a.k.a. runtime;
84 * - the deadline (relative to the actual activation time) of each
86 * Very briefly, a periodic (sporadic) task asks for the execution of
87 * some specific computation --which is typically called an instance--
88 * (at most) every period. Moreover, each instance typically lasts no more
89 * than the runtime and must be completed by time instant t equal to
90 * the instance activation time + the deadline.
92 * This is reflected by the actual fields of the sched_attr structure:
94 * @size size of the structure, for fwd/bwd compat.
96 * @sched_policy task's scheduling policy
97 * @sched_flags for customizing the scheduler behaviour
98 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
99 * @sched_priority task's static priority (SCHED_FIFO/RR)
100 * @sched_deadline representative of the task's deadline
101 * @sched_runtime representative of the task's runtime
102 * @sched_period representative of the task's period
104 * Given this task model, there are a multiplicity of scheduling algorithms
105 * and policies, that can be used to ensure all the tasks will make their
106 * timing constraints.
108 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
109 * only user of this new interface. More information about the algorithm
110 * available in the scheduling class file or in Documentation/.
118 /* SCHED_NORMAL, SCHED_BATCH */
121 /* SCHED_FIFO, SCHED_RR */
130 struct futex_pi_state;
131 struct robust_list_head;
134 struct perf_event_context;
139 #define VMACACHE_BITS 2
140 #define VMACACHE_SIZE (1U << VMACACHE_BITS)
141 #define VMACACHE_MASK (VMACACHE_SIZE - 1)
144 * These are the constant used to fake the fixed-point load-average
145 * counting. Some notes:
146 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
147 * a load-average precision of 10 bits integer + 11 bits fractional
148 * - if you want to count load-averages more often, you need more
149 * precision, or rounding will get you. With 2-second counting freq,
150 * the EXP_n values would be 1981, 2034 and 2043 if still using only
153 extern unsigned long avenrun[]; /* Load averages */
154 extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
156 #define FSHIFT 11 /* nr of bits of precision */
157 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
158 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
159 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
160 #define EXP_5 2014 /* 1/exp(5sec/5min) */
161 #define EXP_15 2037 /* 1/exp(5sec/15min) */
163 #define CALC_LOAD(load,exp,n) \
165 load += n*(FIXED_1-exp); \
168 extern unsigned long total_forks;
169 extern int nr_threads;
170 DECLARE_PER_CPU(unsigned long, process_counts);
171 extern int nr_processes(void);
172 extern unsigned long nr_running(void);
173 extern bool single_task_running(void);
174 extern unsigned long nr_iowait(void);
175 extern unsigned long nr_iowait_cpu(int cpu);
176 extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
178 extern void calc_global_load(unsigned long ticks);
180 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
181 extern void cpu_load_update_nohz_start(void);
182 extern void cpu_load_update_nohz_stop(void);
184 static inline void cpu_load_update_nohz_start(void) { }
185 static inline void cpu_load_update_nohz_stop(void) { }
188 extern void dump_cpu_task(int cpu);
193 #ifdef CONFIG_SCHED_DEBUG
194 extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
195 extern void proc_sched_set_task(struct task_struct *p);
199 * Task state bitmask. NOTE! These bits are also
200 * encoded in fs/proc/array.c: get_task_state().
202 * We have two separate sets of flags: task->state
203 * is about runnability, while task->exit_state are
204 * about the task exiting. Confusing, but this way
205 * modifying one set can't modify the other one by
208 #define TASK_RUNNING 0
209 #define TASK_INTERRUPTIBLE 1
210 #define TASK_UNINTERRUPTIBLE 2
211 #define __TASK_STOPPED 4
212 #define __TASK_TRACED 8
213 /* in tsk->exit_state */
215 #define EXIT_ZOMBIE 32
216 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
217 /* in tsk->state again */
219 #define TASK_WAKEKILL 128
220 #define TASK_WAKING 256
221 #define TASK_PARKED 512
222 #define TASK_NOLOAD 1024
223 #define TASK_STATE_MAX 2048
225 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPN"
227 extern char ___assert_task_state[1 - 2*!!(
228 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
230 /* Convenience macros for the sake of set_task_state */
231 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
232 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
233 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
235 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
237 /* Convenience macros for the sake of wake_up */
238 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
239 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
241 /* get_task_state() */
242 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
243 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
244 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
246 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
247 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
248 #define task_is_stopped_or_traced(task) \
249 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
250 #define task_contributes_to_load(task) \
251 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
252 (task->flags & PF_FROZEN) == 0 && \
253 (task->state & TASK_NOLOAD) == 0)
255 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
257 #define __set_task_state(tsk, state_value) \
259 (tsk)->task_state_change = _THIS_IP_; \
260 (tsk)->state = (state_value); \
262 #define set_task_state(tsk, state_value) \
264 (tsk)->task_state_change = _THIS_IP_; \
265 smp_store_mb((tsk)->state, (state_value)); \
269 * set_current_state() includes a barrier so that the write of current->state
270 * is correctly serialised wrt the caller's subsequent test of whether to
273 * set_current_state(TASK_UNINTERRUPTIBLE);
274 * if (do_i_need_to_sleep())
277 * If the caller does not need such serialisation then use __set_current_state()
279 #define __set_current_state(state_value) \
281 current->task_state_change = _THIS_IP_; \
282 current->state = (state_value); \
284 #define set_current_state(state_value) \
286 current->task_state_change = _THIS_IP_; \
287 smp_store_mb(current->state, (state_value)); \
292 #define __set_task_state(tsk, state_value) \
293 do { (tsk)->state = (state_value); } while (0)
294 #define set_task_state(tsk, state_value) \
295 smp_store_mb((tsk)->state, (state_value))
298 * set_current_state() includes a barrier so that the write of current->state
299 * is correctly serialised wrt the caller's subsequent test of whether to
302 * set_current_state(TASK_UNINTERRUPTIBLE);
303 * if (do_i_need_to_sleep())
306 * If the caller does not need such serialisation then use __set_current_state()
308 #define __set_current_state(state_value) \
309 do { current->state = (state_value); } while (0)
310 #define set_current_state(state_value) \
311 smp_store_mb(current->state, (state_value))
315 /* Task command name length */
316 #define TASK_COMM_LEN 16
318 #include <linux/spinlock.h>
321 * This serializes "schedule()" and also protects
322 * the run-queue from deletions/modifications (but
323 * _adding_ to the beginning of the run-queue has
326 extern rwlock_t tasklist_lock;
327 extern spinlock_t mmlist_lock;
331 #ifdef CONFIG_PROVE_RCU
332 extern int lockdep_tasklist_lock_is_held(void);
333 #endif /* #ifdef CONFIG_PROVE_RCU */
335 extern void sched_init(void);
336 extern void sched_init_smp(void);
337 extern asmlinkage void schedule_tail(struct task_struct *prev);
338 extern void init_idle(struct task_struct *idle, int cpu);
339 extern void init_idle_bootup_task(struct task_struct *idle);
341 extern cpumask_var_t cpu_isolated_map;
343 extern int runqueue_is_locked(int cpu);
345 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
346 extern void nohz_balance_enter_idle(int cpu);
347 extern void set_cpu_sd_state_idle(void);
348 extern int get_nohz_timer_target(void);
350 static inline void nohz_balance_enter_idle(int cpu) { }
351 static inline void set_cpu_sd_state_idle(void) { }
355 * Only dump TASK_* tasks. (0 for all tasks)
357 extern void show_state_filter(unsigned long state_filter);
359 static inline void show_state(void)
361 show_state_filter(0);
364 extern void show_regs(struct pt_regs *);
367 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
368 * task), SP is the stack pointer of the first frame that should be shown in the back
369 * trace (or NULL if the entire call-chain of the task should be shown).
371 extern void show_stack(struct task_struct *task, unsigned long *sp);
373 extern void cpu_init (void);
374 extern void trap_init(void);
375 extern void update_process_times(int user);
376 extern void scheduler_tick(void);
377 extern int sched_cpu_starting(unsigned int cpu);
378 extern int sched_cpu_activate(unsigned int cpu);
379 extern int sched_cpu_deactivate(unsigned int cpu);
381 #ifdef CONFIG_HOTPLUG_CPU
382 extern int sched_cpu_dying(unsigned int cpu);
384 # define sched_cpu_dying NULL
387 extern void sched_show_task(struct task_struct *p);
389 #ifdef CONFIG_LOCKUP_DETECTOR
390 extern void touch_softlockup_watchdog_sched(void);
391 extern void touch_softlockup_watchdog(void);
392 extern void touch_softlockup_watchdog_sync(void);
393 extern void touch_all_softlockup_watchdogs(void);
394 extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
396 size_t *lenp, loff_t *ppos);
397 extern unsigned int softlockup_panic;
398 extern unsigned int hardlockup_panic;
399 void lockup_detector_init(void);
401 static inline void touch_softlockup_watchdog_sched(void)
404 static inline void touch_softlockup_watchdog(void)
407 static inline void touch_softlockup_watchdog_sync(void)
410 static inline void touch_all_softlockup_watchdogs(void)
413 static inline void lockup_detector_init(void)
418 #ifdef CONFIG_DETECT_HUNG_TASK
419 void reset_hung_task_detector(void);
421 static inline void reset_hung_task_detector(void)
426 /* Attach to any functions which should be ignored in wchan output. */
427 #define __sched __attribute__((__section__(".sched.text")))
429 /* Linker adds these: start and end of __sched functions */
430 extern char __sched_text_start[], __sched_text_end[];
432 /* Is this address in the __sched functions? */
433 extern int in_sched_functions(unsigned long addr);
435 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
436 extern signed long schedule_timeout(signed long timeout);
437 extern signed long schedule_timeout_interruptible(signed long timeout);
438 extern signed long schedule_timeout_killable(signed long timeout);
439 extern signed long schedule_timeout_uninterruptible(signed long timeout);
440 extern signed long schedule_timeout_idle(signed long timeout);
441 asmlinkage void schedule(void);
442 extern void schedule_preempt_disabled(void);
444 extern long io_schedule_timeout(long timeout);
446 static inline void io_schedule(void)
448 io_schedule_timeout(MAX_SCHEDULE_TIMEOUT);
452 struct user_namespace;
455 extern void arch_pick_mmap_layout(struct mm_struct *mm);
457 arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
458 unsigned long, unsigned long);
460 arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
461 unsigned long len, unsigned long pgoff,
462 unsigned long flags);
464 static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
467 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
468 #define SUID_DUMP_USER 1 /* Dump as user of process */
469 #define SUID_DUMP_ROOT 2 /* Dump as root */
473 /* for SUID_DUMP_* above */
474 #define MMF_DUMPABLE_BITS 2
475 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
477 extern void set_dumpable(struct mm_struct *mm, int value);
479 * This returns the actual value of the suid_dumpable flag. For things
480 * that are using this for checking for privilege transitions, it must
481 * test against SUID_DUMP_USER rather than treating it as a boolean
484 static inline int __get_dumpable(unsigned long mm_flags)
486 return mm_flags & MMF_DUMPABLE_MASK;
489 static inline int get_dumpable(struct mm_struct *mm)
491 return __get_dumpable(mm->flags);
494 /* coredump filter bits */
495 #define MMF_DUMP_ANON_PRIVATE 2
496 #define MMF_DUMP_ANON_SHARED 3
497 #define MMF_DUMP_MAPPED_PRIVATE 4
498 #define MMF_DUMP_MAPPED_SHARED 5
499 #define MMF_DUMP_ELF_HEADERS 6
500 #define MMF_DUMP_HUGETLB_PRIVATE 7
501 #define MMF_DUMP_HUGETLB_SHARED 8
502 #define MMF_DUMP_DAX_PRIVATE 9
503 #define MMF_DUMP_DAX_SHARED 10
505 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
506 #define MMF_DUMP_FILTER_BITS 9
507 #define MMF_DUMP_FILTER_MASK \
508 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
509 #define MMF_DUMP_FILTER_DEFAULT \
510 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
511 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
513 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
514 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
516 # define MMF_DUMP_MASK_DEFAULT_ELF 0
518 /* leave room for more dump flags */
519 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
520 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
521 #define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
523 #define MMF_HAS_UPROBES 19 /* has uprobes */
524 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
526 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
528 struct sighand_struct {
530 struct k_sigaction action[_NSIG];
532 wait_queue_head_t signalfd_wqh;
535 struct pacct_struct {
538 unsigned long ac_mem;
539 cputime_t ac_utime, ac_stime;
540 unsigned long ac_minflt, ac_majflt;
551 * struct prev_cputime - snaphsot of system and user cputime
552 * @utime: time spent in user mode
553 * @stime: time spent in system mode
554 * @lock: protects the above two fields
556 * Stores previous user/system time values such that we can guarantee
559 struct prev_cputime {
560 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
567 static inline void prev_cputime_init(struct prev_cputime *prev)
569 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
570 prev->utime = prev->stime = 0;
571 raw_spin_lock_init(&prev->lock);
576 * struct task_cputime - collected CPU time counts
577 * @utime: time spent in user mode, in &cputime_t units
578 * @stime: time spent in kernel mode, in &cputime_t units
579 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
581 * This structure groups together three kinds of CPU time that are tracked for
582 * threads and thread groups. Most things considering CPU time want to group
583 * these counts together and treat all three of them in parallel.
585 struct task_cputime {
588 unsigned long long sum_exec_runtime;
591 /* Alternate field names when used to cache expirations. */
592 #define virt_exp utime
593 #define prof_exp stime
594 #define sched_exp sum_exec_runtime
596 #define INIT_CPUTIME \
597 (struct task_cputime) { \
600 .sum_exec_runtime = 0, \
604 * This is the atomic variant of task_cputime, which can be used for
605 * storing and updating task_cputime statistics without locking.
607 struct task_cputime_atomic {
610 atomic64_t sum_exec_runtime;
613 #define INIT_CPUTIME_ATOMIC \
614 (struct task_cputime_atomic) { \
615 .utime = ATOMIC64_INIT(0), \
616 .stime = ATOMIC64_INIT(0), \
617 .sum_exec_runtime = ATOMIC64_INIT(0), \
620 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
623 * Disable preemption until the scheduler is running -- use an unconditional
624 * value so that it also works on !PREEMPT_COUNT kernels.
626 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
628 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
631 * Initial preempt_count value; reflects the preempt_count schedule invariant
632 * which states that during context switches:
634 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
636 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
637 * Note: See finish_task_switch().
639 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
642 * struct thread_group_cputimer - thread group interval timer counts
643 * @cputime_atomic: atomic thread group interval timers.
644 * @running: true when there are timers running and
645 * @cputime_atomic receives updates.
646 * @checking_timer: true when a thread in the group is in the
647 * process of checking for thread group timers.
649 * This structure contains the version of task_cputime, above, that is
650 * used for thread group CPU timer calculations.
652 struct thread_group_cputimer {
653 struct task_cputime_atomic cputime_atomic;
658 #include <linux/rwsem.h>
662 * NOTE! "signal_struct" does not have its own
663 * locking, because a shared signal_struct always
664 * implies a shared sighand_struct, so locking
665 * sighand_struct is always a proper superset of
666 * the locking of signal_struct.
668 struct signal_struct {
672 struct list_head thread_head;
674 wait_queue_head_t wait_chldexit; /* for wait4() */
676 /* current thread group signal load-balancing target: */
677 struct task_struct *curr_target;
679 /* shared signal handling: */
680 struct sigpending shared_pending;
682 /* thread group exit support */
685 * - notify group_exit_task when ->count is equal to notify_count
686 * - everyone except group_exit_task is stopped during signal delivery
687 * of fatal signals, group_exit_task processes the signal.
690 struct task_struct *group_exit_task;
692 /* thread group stop support, overloads group_exit_code too */
693 int group_stop_count;
694 unsigned int flags; /* see SIGNAL_* flags below */
697 * PR_SET_CHILD_SUBREAPER marks a process, like a service
698 * manager, to re-parent orphan (double-forking) child processes
699 * to this process instead of 'init'. The service manager is
700 * able to receive SIGCHLD signals and is able to investigate
701 * the process until it calls wait(). All children of this
702 * process will inherit a flag if they should look for a
703 * child_subreaper process at exit.
705 unsigned int is_child_subreaper:1;
706 unsigned int has_child_subreaper:1;
708 /* POSIX.1b Interval Timers */
710 struct list_head posix_timers;
712 /* ITIMER_REAL timer for the process */
713 struct hrtimer real_timer;
714 struct pid *leader_pid;
715 ktime_t it_real_incr;
718 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
719 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
720 * values are defined to 0 and 1 respectively
722 struct cpu_itimer it[2];
725 * Thread group totals for process CPU timers.
726 * See thread_group_cputimer(), et al, for details.
728 struct thread_group_cputimer cputimer;
730 /* Earliest-expiration cache. */
731 struct task_cputime cputime_expires;
733 #ifdef CONFIG_NO_HZ_FULL
734 atomic_t tick_dep_mask;
737 struct list_head cpu_timers[3];
739 struct pid *tty_old_pgrp;
741 /* boolean value for session group leader */
744 struct tty_struct *tty; /* NULL if no tty */
746 #ifdef CONFIG_SCHED_AUTOGROUP
747 struct autogroup *autogroup;
750 * Cumulative resource counters for dead threads in the group,
751 * and for reaped dead child processes forked by this group.
752 * Live threads maintain their own counters and add to these
753 * in __exit_signal, except for the group leader.
755 seqlock_t stats_lock;
756 cputime_t utime, stime, cutime, cstime;
759 struct prev_cputime prev_cputime;
760 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
761 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
762 unsigned long inblock, oublock, cinblock, coublock;
763 unsigned long maxrss, cmaxrss;
764 struct task_io_accounting ioac;
767 * Cumulative ns of schedule CPU time fo dead threads in the
768 * group, not including a zombie group leader, (This only differs
769 * from jiffies_to_ns(utime + stime) if sched_clock uses something
770 * other than jiffies.)
772 unsigned long long sum_sched_runtime;
775 * We don't bother to synchronize most readers of this at all,
776 * because there is no reader checking a limit that actually needs
777 * to get both rlim_cur and rlim_max atomically, and either one
778 * alone is a single word that can safely be read normally.
779 * getrlimit/setrlimit use task_lock(current->group_leader) to
780 * protect this instead of the siglock, because they really
781 * have no need to disable irqs.
783 struct rlimit rlim[RLIM_NLIMITS];
785 #ifdef CONFIG_BSD_PROCESS_ACCT
786 struct pacct_struct pacct; /* per-process accounting information */
788 #ifdef CONFIG_TASKSTATS
789 struct taskstats *stats;
793 struct tty_audit_buf *tty_audit_buf;
796 oom_flags_t oom_flags;
797 short oom_score_adj; /* OOM kill score adjustment */
798 short oom_score_adj_min; /* OOM kill score adjustment min value.
799 * Only settable by CAP_SYS_RESOURCE. */
801 struct mutex cred_guard_mutex; /* guard against foreign influences on
802 * credential calculations
803 * (notably. ptrace) */
807 * Bits in flags field of signal_struct.
809 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
810 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
811 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
812 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
814 * Pending notifications to parent.
816 #define SIGNAL_CLD_STOPPED 0x00000010
817 #define SIGNAL_CLD_CONTINUED 0x00000020
818 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
820 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
822 /* If true, all threads except ->group_exit_task have pending SIGKILL */
823 static inline int signal_group_exit(const struct signal_struct *sig)
825 return (sig->flags & SIGNAL_GROUP_EXIT) ||
826 (sig->group_exit_task != NULL);
830 * Some day this will be a full-fledged user tracking system..
833 atomic_t __count; /* reference count */
834 atomic_t processes; /* How many processes does this user have? */
835 atomic_t sigpending; /* How many pending signals does this user have? */
836 #ifdef CONFIG_INOTIFY_USER
837 atomic_t inotify_watches; /* How many inotify watches does this user have? */
838 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
840 #ifdef CONFIG_FANOTIFY
841 atomic_t fanotify_listeners;
844 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
846 #ifdef CONFIG_POSIX_MQUEUE
847 /* protected by mq_lock */
848 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
850 unsigned long locked_shm; /* How many pages of mlocked shm ? */
851 unsigned long unix_inflight; /* How many files in flight in unix sockets */
852 atomic_long_t pipe_bufs; /* how many pages are allocated in pipe buffers */
855 struct key *uid_keyring; /* UID specific keyring */
856 struct key *session_keyring; /* UID's default session keyring */
859 /* Hash table maintenance information */
860 struct hlist_node uidhash_node;
863 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
864 atomic_long_t locked_vm;
868 extern int uids_sysfs_init(void);
870 extern struct user_struct *find_user(kuid_t);
872 extern struct user_struct root_user;
873 #define INIT_USER (&root_user)
876 struct backing_dev_info;
877 struct reclaim_state;
879 #ifdef CONFIG_SCHED_INFO
881 /* cumulative counters */
882 unsigned long pcount; /* # of times run on this cpu */
883 unsigned long long run_delay; /* time spent waiting on a runqueue */
886 unsigned long long last_arrival,/* when we last ran on a cpu */
887 last_queued; /* when we were last queued to run */
889 #endif /* CONFIG_SCHED_INFO */
891 #ifdef CONFIG_TASK_DELAY_ACCT
892 struct task_delay_info {
894 unsigned int flags; /* Private per-task flags */
896 /* For each stat XXX, add following, aligned appropriately
898 * struct timespec XXX_start, XXX_end;
902 * Atomicity of updates to XXX_delay, XXX_count protected by
903 * single lock above (split into XXX_lock if contention is an issue).
907 * XXX_count is incremented on every XXX operation, the delay
908 * associated with the operation is added to XXX_delay.
909 * XXX_delay contains the accumulated delay time in nanoseconds.
911 u64 blkio_start; /* Shared by blkio, swapin */
912 u64 blkio_delay; /* wait for sync block io completion */
913 u64 swapin_delay; /* wait for swapin block io completion */
914 u32 blkio_count; /* total count of the number of sync block */
915 /* io operations performed */
916 u32 swapin_count; /* total count of the number of swapin block */
917 /* io operations performed */
920 u64 freepages_delay; /* wait for memory reclaim */
921 u32 freepages_count; /* total count of memory reclaim */
923 #endif /* CONFIG_TASK_DELAY_ACCT */
925 static inline int sched_info_on(void)
927 #ifdef CONFIG_SCHEDSTATS
929 #elif defined(CONFIG_TASK_DELAY_ACCT)
930 extern int delayacct_on;
937 #ifdef CONFIG_SCHEDSTATS
938 void force_schedstat_enabled(void);
949 * Integer metrics need fixed point arithmetic, e.g., sched/fair
950 * has a few: load, load_avg, util_avg, freq, and capacity.
952 * We define a basic fixed point arithmetic range, and then formalize
953 * all these metrics based on that basic range.
955 # define SCHED_FIXEDPOINT_SHIFT 10
956 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
959 * Increase resolution of cpu_capacity calculations
961 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
962 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
965 * Wake-queues are lists of tasks with a pending wakeup, whose
966 * callers have already marked the task as woken internally,
967 * and can thus carry on. A common use case is being able to
968 * do the wakeups once the corresponding user lock as been
971 * We hold reference to each task in the list across the wakeup,
972 * thus guaranteeing that the memory is still valid by the time
973 * the actual wakeups are performed in wake_up_q().
975 * One per task suffices, because there's never a need for a task to be
976 * in two wake queues simultaneously; it is forbidden to abandon a task
977 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
978 * already in a wake queue, the wakeup will happen soon and the second
979 * waker can just skip it.
981 * The WAKE_Q macro declares and initializes the list head.
982 * wake_up_q() does NOT reinitialize the list; it's expected to be
983 * called near the end of a function, where the fact that the queue is
984 * not used again will be easy to see by inspection.
986 * Note that this can cause spurious wakeups. schedule() callers
987 * must ensure the call is done inside a loop, confirming that the
988 * wakeup condition has in fact occurred.
991 struct wake_q_node *next;
995 struct wake_q_node *first;
996 struct wake_q_node **lastp;
999 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1001 #define WAKE_Q(name) \
1002 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1004 extern void wake_q_add(struct wake_q_head *head,
1005 struct task_struct *task);
1006 extern void wake_up_q(struct wake_q_head *head);
1009 * sched-domains (multiprocessor balancing) declarations:
1012 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1013 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1014 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1015 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1016 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1017 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1018 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu power */
1019 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1020 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1021 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1022 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1023 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1024 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1025 #define SD_NUMA 0x4000 /* cross-node balancing */
1027 #ifdef CONFIG_SCHED_SMT
1028 static inline int cpu_smt_flags(void)
1030 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
1034 #ifdef CONFIG_SCHED_MC
1035 static inline int cpu_core_flags(void)
1037 return SD_SHARE_PKG_RESOURCES;
1042 static inline int cpu_numa_flags(void)
1048 struct sched_domain_attr {
1049 int relax_domain_level;
1052 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1053 .relax_domain_level = -1, \
1056 extern int sched_domain_level_max;
1060 struct sched_domain {
1061 /* These fields must be setup */
1062 struct sched_domain *parent; /* top domain must be null terminated */
1063 struct sched_domain *child; /* bottom domain must be null terminated */
1064 struct sched_group *groups; /* the balancing groups of the domain */
1065 unsigned long min_interval; /* Minimum balance interval ms */
1066 unsigned long max_interval; /* Maximum balance interval ms */
1067 unsigned int busy_factor; /* less balancing by factor if busy */
1068 unsigned int imbalance_pct; /* No balance until over watermark */
1069 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
1070 unsigned int busy_idx;
1071 unsigned int idle_idx;
1072 unsigned int newidle_idx;
1073 unsigned int wake_idx;
1074 unsigned int forkexec_idx;
1075 unsigned int smt_gain;
1077 int nohz_idle; /* NOHZ IDLE status */
1078 int flags; /* See SD_* */
1081 /* Runtime fields. */
1082 unsigned long last_balance; /* init to jiffies. units in jiffies */
1083 unsigned int balance_interval; /* initialise to 1. units in ms. */
1084 unsigned int nr_balance_failed; /* initialise to 0 */
1086 /* idle_balance() stats */
1087 u64 max_newidle_lb_cost;
1088 unsigned long next_decay_max_lb_cost;
1090 #ifdef CONFIG_SCHEDSTATS
1091 /* load_balance() stats */
1092 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
1093 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
1094 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
1095 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
1096 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
1097 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
1098 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
1099 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
1101 /* Active load balancing */
1102 unsigned int alb_count;
1103 unsigned int alb_failed;
1104 unsigned int alb_pushed;
1106 /* SD_BALANCE_EXEC stats */
1107 unsigned int sbe_count;
1108 unsigned int sbe_balanced;
1109 unsigned int sbe_pushed;
1111 /* SD_BALANCE_FORK stats */
1112 unsigned int sbf_count;
1113 unsigned int sbf_balanced;
1114 unsigned int sbf_pushed;
1116 /* try_to_wake_up() stats */
1117 unsigned int ttwu_wake_remote;
1118 unsigned int ttwu_move_affine;
1119 unsigned int ttwu_move_balance;
1121 #ifdef CONFIG_SCHED_DEBUG
1125 void *private; /* used during construction */
1126 struct rcu_head rcu; /* used during destruction */
1129 unsigned int span_weight;
1131 * Span of all CPUs in this domain.
1133 * NOTE: this field is variable length. (Allocated dynamically
1134 * by attaching extra space to the end of the structure,
1135 * depending on how many CPUs the kernel has booted up with)
1137 unsigned long span[0];
1140 static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
1142 return to_cpumask(sd->span);
1145 extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1146 struct sched_domain_attr *dattr_new);
1148 /* Allocate an array of sched domains, for partition_sched_domains(). */
1149 cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
1150 void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
1152 bool cpus_share_cache(int this_cpu, int that_cpu);
1154 typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1155 typedef int (*sched_domain_flags_f)(void);
1157 #define SDTL_OVERLAP 0x01
1160 struct sched_domain **__percpu sd;
1161 struct sched_group **__percpu sg;
1162 struct sched_group_capacity **__percpu sgc;
1165 struct sched_domain_topology_level {
1166 sched_domain_mask_f mask;
1167 sched_domain_flags_f sd_flags;
1170 struct sd_data data;
1171 #ifdef CONFIG_SCHED_DEBUG
1176 extern void set_sched_topology(struct sched_domain_topology_level *tl);
1177 extern void wake_up_if_idle(int cpu);
1179 #ifdef CONFIG_SCHED_DEBUG
1180 # define SD_INIT_NAME(type) .name = #type
1182 # define SD_INIT_NAME(type)
1185 #else /* CONFIG_SMP */
1187 struct sched_domain_attr;
1190 partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1191 struct sched_domain_attr *dattr_new)
1195 static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1200 #endif /* !CONFIG_SMP */
1203 struct io_context; /* See blkdev.h */
1206 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1207 extern void prefetch_stack(struct task_struct *t);
1209 static inline void prefetch_stack(struct task_struct *t) { }
1212 struct audit_context; /* See audit.c */
1214 struct pipe_inode_info;
1215 struct uts_namespace;
1217 struct load_weight {
1218 unsigned long weight;
1223 * The load_avg/util_avg accumulates an infinite geometric series
1224 * (see __update_load_avg() in kernel/sched/fair.c).
1226 * [load_avg definition]
1228 * load_avg = runnable% * scale_load_down(load)
1230 * where runnable% is the time ratio that a sched_entity is runnable.
1231 * For cfs_rq, it is the aggregated load_avg of all runnable and
1232 * blocked sched_entities.
1234 * load_avg may also take frequency scaling into account:
1236 * load_avg = runnable% * scale_load_down(load) * freq%
1238 * where freq% is the CPU frequency normalized to the highest frequency.
1240 * [util_avg definition]
1242 * util_avg = running% * SCHED_CAPACITY_SCALE
1244 * where running% is the time ratio that a sched_entity is running on
1245 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1246 * and blocked sched_entities.
1248 * util_avg may also factor frequency scaling and CPU capacity scaling:
1250 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1252 * where freq% is the same as above, and capacity% is the CPU capacity
1253 * normalized to the greatest capacity (due to uarch differences, etc).
1255 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1256 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1257 * we therefore scale them to as large a range as necessary. This is for
1258 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1262 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1263 * with the highest load (=88761), always runnable on a single cfs_rq,
1264 * and should not overflow as the number already hits PID_MAX_LIMIT.
1266 * For all other cases (including 32-bit kernels), struct load_weight's
1267 * weight will overflow first before we do, because:
1269 * Max(load_avg) <= Max(load.weight)
1271 * Then it is the load_weight's responsibility to consider overflow
1275 u64 last_update_time, load_sum;
1276 u32 util_sum, period_contrib;
1277 unsigned long load_avg, util_avg;
1280 #ifdef CONFIG_SCHEDSTATS
1281 struct sched_statistics {
1291 s64 sum_sleep_runtime;
1298 u64 nr_migrations_cold;
1299 u64 nr_failed_migrations_affine;
1300 u64 nr_failed_migrations_running;
1301 u64 nr_failed_migrations_hot;
1302 u64 nr_forced_migrations;
1305 u64 nr_wakeups_sync;
1306 u64 nr_wakeups_migrate;
1307 u64 nr_wakeups_local;
1308 u64 nr_wakeups_remote;
1309 u64 nr_wakeups_affine;
1310 u64 nr_wakeups_affine_attempts;
1311 u64 nr_wakeups_passive;
1312 u64 nr_wakeups_idle;
1316 struct sched_entity {
1317 struct load_weight load; /* for load-balancing */
1318 struct rb_node run_node;
1319 struct list_head group_node;
1323 u64 sum_exec_runtime;
1325 u64 prev_sum_exec_runtime;
1329 #ifdef CONFIG_SCHEDSTATS
1330 struct sched_statistics statistics;
1333 #ifdef CONFIG_FAIR_GROUP_SCHED
1335 struct sched_entity *parent;
1336 /* rq on which this entity is (to be) queued: */
1337 struct cfs_rq *cfs_rq;
1338 /* rq "owned" by this entity/group: */
1339 struct cfs_rq *my_q;
1344 * Per entity load average tracking.
1346 * Put into separate cache line so it does not
1347 * collide with read-mostly values above.
1349 struct sched_avg avg ____cacheline_aligned_in_smp;
1353 struct sched_rt_entity {
1354 struct list_head run_list;
1355 unsigned long timeout;
1356 unsigned long watchdog_stamp;
1357 unsigned int time_slice;
1358 unsigned short on_rq;
1359 unsigned short on_list;
1361 struct sched_rt_entity *back;
1362 #ifdef CONFIG_RT_GROUP_SCHED
1363 struct sched_rt_entity *parent;
1364 /* rq on which this entity is (to be) queued: */
1365 struct rt_rq *rt_rq;
1366 /* rq "owned" by this entity/group: */
1371 struct sched_dl_entity {
1372 struct rb_node rb_node;
1375 * Original scheduling parameters. Copied here from sched_attr
1376 * during sched_setattr(), they will remain the same until
1377 * the next sched_setattr().
1379 u64 dl_runtime; /* maximum runtime for each instance */
1380 u64 dl_deadline; /* relative deadline of each instance */
1381 u64 dl_period; /* separation of two instances (period) */
1382 u64 dl_bw; /* dl_runtime / dl_deadline */
1385 * Actual scheduling parameters. Initialized with the values above,
1386 * they are continously updated during task execution. Note that
1387 * the remaining runtime could be < 0 in case we are in overrun.
1389 s64 runtime; /* remaining runtime for this instance */
1390 u64 deadline; /* absolute deadline for this instance */
1391 unsigned int flags; /* specifying the scheduler behaviour */
1396 * @dl_throttled tells if we exhausted the runtime. If so, the
1397 * task has to wait for a replenishment to be performed at the
1398 * next firing of dl_timer.
1400 * @dl_boosted tells if we are boosted due to DI. If so we are
1401 * outside bandwidth enforcement mechanism (but only until we
1402 * exit the critical section);
1404 * @dl_yielded tells if task gave up the cpu before consuming
1405 * all its available runtime during the last job.
1407 int dl_throttled, dl_boosted, dl_yielded;
1410 * Bandwidth enforcement timer. Each -deadline task has its
1411 * own bandwidth to be enforced, thus we need one timer per task.
1413 struct hrtimer dl_timer;
1421 u8 pad; /* Otherwise the compiler can store garbage here. */
1423 u32 s; /* Set of bits. */
1427 enum perf_event_task_context {
1428 perf_invalid_context = -1,
1429 perf_hw_context = 0,
1431 perf_nr_task_contexts,
1434 /* Track pages that require TLB flushes */
1435 struct tlbflush_unmap_batch {
1437 * Each bit set is a CPU that potentially has a TLB entry for one of
1438 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1440 struct cpumask cpumask;
1442 /* True if any bit in cpumask is set */
1443 bool flush_required;
1446 * If true then the PTE was dirty when unmapped. The entry must be
1447 * flushed before IO is initiated or a stale TLB entry potentially
1448 * allows an update without redirtying the page.
1453 struct task_struct {
1454 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1457 unsigned int flags; /* per process flags, defined below */
1458 unsigned int ptrace;
1461 struct llist_node wake_entry;
1463 unsigned int wakee_flips;
1464 unsigned long wakee_flip_decay_ts;
1465 struct task_struct *last_wakee;
1471 int prio, static_prio, normal_prio;
1472 unsigned int rt_priority;
1473 const struct sched_class *sched_class;
1474 struct sched_entity se;
1475 struct sched_rt_entity rt;
1476 #ifdef CONFIG_CGROUP_SCHED
1477 struct task_group *sched_task_group;
1479 struct sched_dl_entity dl;
1481 #ifdef CONFIG_PREEMPT_NOTIFIERS
1482 /* list of struct preempt_notifier: */
1483 struct hlist_head preempt_notifiers;
1486 #ifdef CONFIG_BLK_DEV_IO_TRACE
1487 unsigned int btrace_seq;
1490 unsigned int policy;
1491 int nr_cpus_allowed;
1492 cpumask_t cpus_allowed;
1494 #ifdef CONFIG_PREEMPT_RCU
1495 int rcu_read_lock_nesting;
1496 union rcu_special rcu_read_unlock_special;
1497 struct list_head rcu_node_entry;
1498 struct rcu_node *rcu_blocked_node;
1499 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1500 #ifdef CONFIG_TASKS_RCU
1501 unsigned long rcu_tasks_nvcsw;
1502 bool rcu_tasks_holdout;
1503 struct list_head rcu_tasks_holdout_list;
1504 int rcu_tasks_idle_cpu;
1505 #endif /* #ifdef CONFIG_TASKS_RCU */
1507 #ifdef CONFIG_SCHED_INFO
1508 struct sched_info sched_info;
1511 struct list_head tasks;
1513 struct plist_node pushable_tasks;
1514 struct rb_node pushable_dl_tasks;
1517 struct mm_struct *mm, *active_mm;
1518 /* per-thread vma caching */
1519 u32 vmacache_seqnum;
1520 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1521 #if defined(SPLIT_RSS_COUNTING)
1522 struct task_rss_stat rss_stat;
1526 int exit_code, exit_signal;
1527 int pdeath_signal; /* The signal sent when the parent dies */
1528 unsigned long jobctl; /* JOBCTL_*, siglock protected */
1530 /* Used for emulating ABI behavior of previous Linux versions */
1531 unsigned int personality;
1533 /* scheduler bits, serialized by scheduler locks */
1534 unsigned sched_reset_on_fork:1;
1535 unsigned sched_contributes_to_load:1;
1536 unsigned sched_migrated:1;
1537 unsigned :0; /* force alignment to the next boundary */
1539 /* unserialized, strictly 'current' */
1540 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
1541 unsigned in_iowait:1;
1543 unsigned memcg_may_oom:1;
1545 unsigned memcg_kmem_skip_account:1;
1548 #ifdef CONFIG_COMPAT_BRK
1549 unsigned brk_randomized:1;
1552 unsigned long atomic_flags; /* Flags needing atomic access. */
1554 struct restart_block restart_block;
1559 #ifdef CONFIG_CC_STACKPROTECTOR
1560 /* Canary value for the -fstack-protector gcc feature */
1561 unsigned long stack_canary;
1564 * pointers to (original) parent process, youngest child, younger sibling,
1565 * older sibling, respectively. (p->father can be replaced with
1566 * p->real_parent->pid)
1568 struct task_struct __rcu *real_parent; /* real parent process */
1569 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1571 * children/sibling forms the list of my natural children
1573 struct list_head children; /* list of my children */
1574 struct list_head sibling; /* linkage in my parent's children list */
1575 struct task_struct *group_leader; /* threadgroup leader */
1578 * ptraced is the list of tasks this task is using ptrace on.
1579 * This includes both natural children and PTRACE_ATTACH targets.
1580 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1582 struct list_head ptraced;
1583 struct list_head ptrace_entry;
1585 /* PID/PID hash table linkage. */
1586 struct pid_link pids[PIDTYPE_MAX];
1587 struct list_head thread_group;
1588 struct list_head thread_node;
1590 struct completion *vfork_done; /* for vfork() */
1591 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1592 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1594 cputime_t utime, stime, utimescaled, stimescaled;
1596 struct prev_cputime prev_cputime;
1597 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1598 seqcount_t vtime_seqcount;
1599 unsigned long long vtime_snap;
1601 /* Task is sleeping or running in a CPU with VTIME inactive */
1603 /* Task runs in userspace in a CPU with VTIME active */
1605 /* Task runs in kernelspace in a CPU with VTIME active */
1607 } vtime_snap_whence;
1610 #ifdef CONFIG_NO_HZ_FULL
1611 atomic_t tick_dep_mask;
1613 unsigned long nvcsw, nivcsw; /* context switch counts */
1614 u64 start_time; /* monotonic time in nsec */
1615 u64 real_start_time; /* boot based time in nsec */
1616 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1617 unsigned long min_flt, maj_flt;
1619 struct task_cputime cputime_expires;
1620 struct list_head cpu_timers[3];
1622 /* process credentials */
1623 const struct cred __rcu *real_cred; /* objective and real subjective task
1624 * credentials (COW) */
1625 const struct cred __rcu *cred; /* effective (overridable) subjective task
1626 * credentials (COW) */
1627 char comm[TASK_COMM_LEN]; /* executable name excluding path
1628 - access with [gs]et_task_comm (which lock
1629 it with task_lock())
1630 - initialized normally by setup_new_exec */
1631 /* file system info */
1632 struct nameidata *nameidata;
1633 #ifdef CONFIG_SYSVIPC
1635 struct sysv_sem sysvsem;
1636 struct sysv_shm sysvshm;
1638 #ifdef CONFIG_DETECT_HUNG_TASK
1639 /* hung task detection */
1640 unsigned long last_switch_count;
1642 /* filesystem information */
1643 struct fs_struct *fs;
1644 /* open file information */
1645 struct files_struct *files;
1647 struct nsproxy *nsproxy;
1648 /* signal handlers */
1649 struct signal_struct *signal;
1650 struct sighand_struct *sighand;
1652 sigset_t blocked, real_blocked;
1653 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1654 struct sigpending pending;
1656 unsigned long sas_ss_sp;
1659 struct callback_head *task_works;
1661 struct audit_context *audit_context;
1662 #ifdef CONFIG_AUDITSYSCALL
1664 unsigned int sessionid;
1666 struct seccomp seccomp;
1668 /* Thread group tracking */
1671 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1673 spinlock_t alloc_lock;
1675 /* Protection of the PI data structures: */
1676 raw_spinlock_t pi_lock;
1678 struct wake_q_node wake_q;
1680 #ifdef CONFIG_RT_MUTEXES
1681 /* PI waiters blocked on a rt_mutex held by this task */
1682 struct rb_root pi_waiters;
1683 struct rb_node *pi_waiters_leftmost;
1684 /* Deadlock detection and priority inheritance handling */
1685 struct rt_mutex_waiter *pi_blocked_on;
1688 #ifdef CONFIG_DEBUG_MUTEXES
1689 /* mutex deadlock detection */
1690 struct mutex_waiter *blocked_on;
1692 #ifdef CONFIG_TRACE_IRQFLAGS
1693 unsigned int irq_events;
1694 unsigned long hardirq_enable_ip;
1695 unsigned long hardirq_disable_ip;
1696 unsigned int hardirq_enable_event;
1697 unsigned int hardirq_disable_event;
1698 int hardirqs_enabled;
1699 int hardirq_context;
1700 unsigned long softirq_disable_ip;
1701 unsigned long softirq_enable_ip;
1702 unsigned int softirq_disable_event;
1703 unsigned int softirq_enable_event;
1704 int softirqs_enabled;
1705 int softirq_context;
1707 #ifdef CONFIG_LOCKDEP
1708 # define MAX_LOCK_DEPTH 48UL
1711 unsigned int lockdep_recursion;
1712 struct held_lock held_locks[MAX_LOCK_DEPTH];
1713 gfp_t lockdep_reclaim_gfp;
1716 unsigned int in_ubsan;
1719 /* journalling filesystem info */
1722 /* stacked block device info */
1723 struct bio_list *bio_list;
1726 /* stack plugging */
1727 struct blk_plug *plug;
1731 struct reclaim_state *reclaim_state;
1733 struct backing_dev_info *backing_dev_info;
1735 struct io_context *io_context;
1737 unsigned long ptrace_message;
1738 siginfo_t *last_siginfo; /* For ptrace use. */
1739 struct task_io_accounting ioac;
1740 #if defined(CONFIG_TASK_XACCT)
1741 u64 acct_rss_mem1; /* accumulated rss usage */
1742 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1743 cputime_t acct_timexpd; /* stime + utime since last update */
1745 #ifdef CONFIG_CPUSETS
1746 nodemask_t mems_allowed; /* Protected by alloc_lock */
1747 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1748 int cpuset_mem_spread_rotor;
1749 int cpuset_slab_spread_rotor;
1751 #ifdef CONFIG_CGROUPS
1752 /* Control Group info protected by css_set_lock */
1753 struct css_set __rcu *cgroups;
1754 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1755 struct list_head cg_list;
1758 struct robust_list_head __user *robust_list;
1759 #ifdef CONFIG_COMPAT
1760 struct compat_robust_list_head __user *compat_robust_list;
1762 struct list_head pi_state_list;
1763 struct futex_pi_state *pi_state_cache;
1765 #ifdef CONFIG_PERF_EVENTS
1766 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1767 struct mutex perf_event_mutex;
1768 struct list_head perf_event_list;
1770 #ifdef CONFIG_DEBUG_PREEMPT
1771 unsigned long preempt_disable_ip;
1774 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1776 short pref_node_fork;
1778 #ifdef CONFIG_NUMA_BALANCING
1780 unsigned int numa_scan_period;
1781 unsigned int numa_scan_period_max;
1782 int numa_preferred_nid;
1783 unsigned long numa_migrate_retry;
1784 u64 node_stamp; /* migration stamp */
1785 u64 last_task_numa_placement;
1786 u64 last_sum_exec_runtime;
1787 struct callback_head numa_work;
1789 struct list_head numa_entry;
1790 struct numa_group *numa_group;
1793 * numa_faults is an array split into four regions:
1794 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1795 * in this precise order.
1797 * faults_memory: Exponential decaying average of faults on a per-node
1798 * basis. Scheduling placement decisions are made based on these
1799 * counts. The values remain static for the duration of a PTE scan.
1800 * faults_cpu: Track the nodes the process was running on when a NUMA
1801 * hinting fault was incurred.
1802 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1803 * during the current scan window. When the scan completes, the counts
1804 * in faults_memory and faults_cpu decay and these values are copied.
1806 unsigned long *numa_faults;
1807 unsigned long total_numa_faults;
1810 * numa_faults_locality tracks if faults recorded during the last
1811 * scan window were remote/local or failed to migrate. The task scan
1812 * period is adapted based on the locality of the faults with different
1813 * weights depending on whether they were shared or private faults
1815 unsigned long numa_faults_locality[3];
1817 unsigned long numa_pages_migrated;
1818 #endif /* CONFIG_NUMA_BALANCING */
1820 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1821 struct tlbflush_unmap_batch tlb_ubc;
1824 struct rcu_head rcu;
1827 * cache last used pipe for splice
1829 struct pipe_inode_info *splice_pipe;
1831 struct page_frag task_frag;
1833 #ifdef CONFIG_TASK_DELAY_ACCT
1834 struct task_delay_info *delays;
1836 #ifdef CONFIG_FAULT_INJECTION
1840 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1841 * balance_dirty_pages() for some dirty throttling pause
1844 int nr_dirtied_pause;
1845 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1847 #ifdef CONFIG_LATENCYTOP
1848 int latency_record_count;
1849 struct latency_record latency_record[LT_SAVECOUNT];
1852 * time slack values; these are used to round up poll() and
1853 * select() etc timeout values. These are in nanoseconds.
1856 u64 default_timer_slack_ns;
1859 unsigned int kasan_depth;
1861 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1862 /* Index of current stored address in ret_stack */
1864 /* Stack of return addresses for return function tracing */
1865 struct ftrace_ret_stack *ret_stack;
1866 /* time stamp for last schedule */
1867 unsigned long long ftrace_timestamp;
1869 * Number of functions that haven't been traced
1870 * because of depth overrun.
1872 atomic_t trace_overrun;
1873 /* Pause for the tracing */
1874 atomic_t tracing_graph_pause;
1876 #ifdef CONFIG_TRACING
1877 /* state flags for use by tracers */
1878 unsigned long trace;
1879 /* bitmask and counter of trace recursion */
1880 unsigned long trace_recursion;
1881 #endif /* CONFIG_TRACING */
1883 /* Coverage collection mode enabled for this task (0 if disabled). */
1884 enum kcov_mode kcov_mode;
1885 /* Size of the kcov_area. */
1887 /* Buffer for coverage collection. */
1889 /* kcov desciptor wired with this task or NULL. */
1893 struct mem_cgroup *memcg_in_oom;
1894 gfp_t memcg_oom_gfp_mask;
1895 int memcg_oom_order;
1897 /* number of pages to reclaim on returning to userland */
1898 unsigned int memcg_nr_pages_over_high;
1900 #ifdef CONFIG_UPROBES
1901 struct uprobe_task *utask;
1903 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1904 unsigned int sequential_io;
1905 unsigned int sequential_io_avg;
1907 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1908 unsigned long task_state_change;
1910 int pagefault_disabled;
1912 struct task_struct *oom_reaper_list;
1914 /* CPU-specific state of this task */
1915 struct thread_struct thread;
1917 * WARNING: on x86, 'thread_struct' contains a variable-sized
1918 * structure. It *MUST* be at the end of 'task_struct'.
1920 * Do not put anything below here!
1924 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1925 extern int arch_task_struct_size __read_mostly;
1927 # define arch_task_struct_size (sizeof(struct task_struct))
1930 /* Future-safe accessor for struct task_struct's cpus_allowed. */
1931 #define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1933 #define TNF_MIGRATED 0x01
1934 #define TNF_NO_GROUP 0x02
1935 #define TNF_SHARED 0x04
1936 #define TNF_FAULT_LOCAL 0x08
1937 #define TNF_MIGRATE_FAIL 0x10
1939 #ifdef CONFIG_NUMA_BALANCING
1940 extern void task_numa_fault(int last_node, int node, int pages, int flags);
1941 extern pid_t task_numa_group_id(struct task_struct *p);
1942 extern void set_numabalancing_state(bool enabled);
1943 extern void task_numa_free(struct task_struct *p);
1944 extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
1945 int src_nid, int dst_cpu);
1947 static inline void task_numa_fault(int last_node, int node, int pages,
1951 static inline pid_t task_numa_group_id(struct task_struct *p)
1955 static inline void set_numabalancing_state(bool enabled)
1958 static inline void task_numa_free(struct task_struct *p)
1961 static inline bool should_numa_migrate_memory(struct task_struct *p,
1962 struct page *page, int src_nid, int dst_cpu)
1968 static inline struct pid *task_pid(struct task_struct *task)
1970 return task->pids[PIDTYPE_PID].pid;
1973 static inline struct pid *task_tgid(struct task_struct *task)
1975 return task->group_leader->pids[PIDTYPE_PID].pid;
1979 * Without tasklist or rcu lock it is not safe to dereference
1980 * the result of task_pgrp/task_session even if task == current,
1981 * we can race with another thread doing sys_setsid/sys_setpgid.
1983 static inline struct pid *task_pgrp(struct task_struct *task)
1985 return task->group_leader->pids[PIDTYPE_PGID].pid;
1988 static inline struct pid *task_session(struct task_struct *task)
1990 return task->group_leader->pids[PIDTYPE_SID].pid;
1993 struct pid_namespace;
1996 * the helpers to get the task's different pids as they are seen
1997 * from various namespaces
1999 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2000 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2002 * task_xid_nr_ns() : id seen from the ns specified;
2004 * set_task_vxid() : assigns a virtual id to a task;
2006 * see also pid_nr() etc in include/linux/pid.h
2008 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
2009 struct pid_namespace *ns);
2011 static inline pid_t task_pid_nr(struct task_struct *tsk)
2016 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
2017 struct pid_namespace *ns)
2019 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
2022 static inline pid_t task_pid_vnr(struct task_struct *tsk)
2024 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
2028 static inline pid_t task_tgid_nr(struct task_struct *tsk)
2033 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
2035 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
2037 return pid_vnr(task_tgid(tsk));
2041 static inline int pid_alive(const struct task_struct *p);
2042 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
2048 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
2054 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
2056 return task_ppid_nr_ns(tsk, &init_pid_ns);
2059 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
2060 struct pid_namespace *ns)
2062 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
2065 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
2067 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
2071 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
2072 struct pid_namespace *ns)
2074 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
2077 static inline pid_t task_session_vnr(struct task_struct *tsk)
2079 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
2082 /* obsolete, do not use */
2083 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
2085 return task_pgrp_nr_ns(tsk, &init_pid_ns);
2089 * pid_alive - check that a task structure is not stale
2090 * @p: Task structure to be checked.
2092 * Test if a process is not yet dead (at most zombie state)
2093 * If pid_alive fails, then pointers within the task structure
2094 * can be stale and must not be dereferenced.
2096 * Return: 1 if the process is alive. 0 otherwise.
2098 static inline int pid_alive(const struct task_struct *p)
2100 return p->pids[PIDTYPE_PID].pid != NULL;
2104 * is_global_init - check if a task structure is init. Since init
2105 * is free to have sub-threads we need to check tgid.
2106 * @tsk: Task structure to be checked.
2108 * Check if a task structure is the first user space task the kernel created.
2110 * Return: 1 if the task structure is init. 0 otherwise.
2112 static inline int is_global_init(struct task_struct *tsk)
2114 return task_tgid_nr(tsk) == 1;
2117 extern struct pid *cad_pid;
2119 extern void free_task(struct task_struct *tsk);
2120 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2122 extern void __put_task_struct(struct task_struct *t);
2124 static inline void put_task_struct(struct task_struct *t)
2126 if (atomic_dec_and_test(&t->usage))
2127 __put_task_struct(t);
2130 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2131 extern void task_cputime(struct task_struct *t,
2132 cputime_t *utime, cputime_t *stime);
2133 extern void task_cputime_scaled(struct task_struct *t,
2134 cputime_t *utimescaled, cputime_t *stimescaled);
2135 extern cputime_t task_gtime(struct task_struct *t);
2137 static inline void task_cputime(struct task_struct *t,
2138 cputime_t *utime, cputime_t *stime)
2146 static inline void task_cputime_scaled(struct task_struct *t,
2147 cputime_t *utimescaled,
2148 cputime_t *stimescaled)
2151 *utimescaled = t->utimescaled;
2153 *stimescaled = t->stimescaled;
2156 static inline cputime_t task_gtime(struct task_struct *t)
2161 extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2162 extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
2167 #define PF_EXITING 0x00000004 /* getting shut down */
2168 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2169 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2170 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2171 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2172 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2173 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2174 #define PF_DUMPCORE 0x00000200 /* dumped core */
2175 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2176 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2177 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2178 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2179 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2180 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2181 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2182 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2183 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2184 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2185 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2186 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2187 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2188 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2189 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2190 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2191 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2192 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2193 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2196 * Only the _current_ task can read/write to tsk->flags, but other
2197 * tasks can access tsk->flags in readonly mode for example
2198 * with tsk_used_math (like during threaded core dumping).
2199 * There is however an exception to this rule during ptrace
2200 * or during fork: the ptracer task is allowed to write to the
2201 * child->flags of its traced child (same goes for fork, the parent
2202 * can write to the child->flags), because we're guaranteed the
2203 * child is not running and in turn not changing child->flags
2204 * at the same time the parent does it.
2206 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2207 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2208 #define clear_used_math() clear_stopped_child_used_math(current)
2209 #define set_used_math() set_stopped_child_used_math(current)
2210 #define conditional_stopped_child_used_math(condition, child) \
2211 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2212 #define conditional_used_math(condition) \
2213 conditional_stopped_child_used_math(condition, current)
2214 #define copy_to_stopped_child_used_math(child) \
2215 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2216 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2217 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2218 #define used_math() tsk_used_math(current)
2220 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2221 * __GFP_FS is also cleared as it implies __GFP_IO.
2223 static inline gfp_t memalloc_noio_flags(gfp_t flags)
2225 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
2226 flags &= ~(__GFP_IO | __GFP_FS);
2230 static inline unsigned int memalloc_noio_save(void)
2232 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
2233 current->flags |= PF_MEMALLOC_NOIO;
2237 static inline void memalloc_noio_restore(unsigned int flags)
2239 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
2242 /* Per-process atomic flags. */
2243 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2244 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2245 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2248 #define TASK_PFA_TEST(name, func) \
2249 static inline bool task_##func(struct task_struct *p) \
2250 { return test_bit(PFA_##name, &p->atomic_flags); }
2251 #define TASK_PFA_SET(name, func) \
2252 static inline void task_set_##func(struct task_struct *p) \
2253 { set_bit(PFA_##name, &p->atomic_flags); }
2254 #define TASK_PFA_CLEAR(name, func) \
2255 static inline void task_clear_##func(struct task_struct *p) \
2256 { clear_bit(PFA_##name, &p->atomic_flags); }
2258 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
2259 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
2261 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
2262 TASK_PFA_SET(SPREAD_PAGE, spread_page)
2263 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
2265 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
2266 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
2267 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
2270 * task->jobctl flags
2272 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2274 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2275 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2276 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2277 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2278 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2279 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2280 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2282 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2283 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2284 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2285 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2286 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2287 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2288 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2290 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2291 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2293 extern bool task_set_jobctl_pending(struct task_struct *task,
2294 unsigned long mask);
2295 extern void task_clear_jobctl_trapping(struct task_struct *task);
2296 extern void task_clear_jobctl_pending(struct task_struct *task,
2297 unsigned long mask);
2299 static inline void rcu_copy_process(struct task_struct *p)
2301 #ifdef CONFIG_PREEMPT_RCU
2302 p->rcu_read_lock_nesting = 0;
2303 p->rcu_read_unlock_special.s = 0;
2304 p->rcu_blocked_node = NULL;
2305 INIT_LIST_HEAD(&p->rcu_node_entry);
2306 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2307 #ifdef CONFIG_TASKS_RCU
2308 p->rcu_tasks_holdout = false;
2309 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
2310 p->rcu_tasks_idle_cpu = -1;
2311 #endif /* #ifdef CONFIG_TASKS_RCU */
2314 static inline void tsk_restore_flags(struct task_struct *task,
2315 unsigned long orig_flags, unsigned long flags)
2317 task->flags &= ~flags;
2318 task->flags |= orig_flags & flags;
2321 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
2322 const struct cpumask *trial);
2323 extern int task_can_attach(struct task_struct *p,
2324 const struct cpumask *cs_cpus_allowed);
2326 extern void do_set_cpus_allowed(struct task_struct *p,
2327 const struct cpumask *new_mask);
2329 extern int set_cpus_allowed_ptr(struct task_struct *p,
2330 const struct cpumask *new_mask);
2332 static inline void do_set_cpus_allowed(struct task_struct *p,
2333 const struct cpumask *new_mask)
2336 static inline int set_cpus_allowed_ptr(struct task_struct *p,
2337 const struct cpumask *new_mask)
2339 if (!cpumask_test_cpu(0, new_mask))
2345 #ifdef CONFIG_NO_HZ_COMMON
2346 void calc_load_enter_idle(void);
2347 void calc_load_exit_idle(void);
2349 static inline void calc_load_enter_idle(void) { }
2350 static inline void calc_load_exit_idle(void) { }
2351 #endif /* CONFIG_NO_HZ_COMMON */
2354 * Do not use outside of architecture code which knows its limitations.
2356 * sched_clock() has no promise of monotonicity or bounded drift between
2357 * CPUs, use (which you should not) requires disabling IRQs.
2359 * Please use one of the three interfaces below.
2361 extern unsigned long long notrace sched_clock(void);
2363 * See the comment in kernel/sched/clock.c
2365 extern u64 running_clock(void);
2366 extern u64 sched_clock_cpu(int cpu);
2369 extern void sched_clock_init(void);
2371 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2372 static inline void sched_clock_tick(void)
2376 static inline void sched_clock_idle_sleep_event(void)
2380 static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2384 static inline u64 cpu_clock(int cpu)
2386 return sched_clock();
2389 static inline u64 local_clock(void)
2391 return sched_clock();
2395 * Architectures can set this to 1 if they have specified
2396 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2397 * but then during bootup it turns out that sched_clock()
2398 * is reliable after all:
2400 extern int sched_clock_stable(void);
2401 extern void set_sched_clock_stable(void);
2402 extern void clear_sched_clock_stable(void);
2404 extern void sched_clock_tick(void);
2405 extern void sched_clock_idle_sleep_event(void);
2406 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2409 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2410 * time source that is monotonic per cpu argument and has bounded drift
2413 * ######################### BIG FAT WARNING ##########################
2414 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2415 * # go backwards !! #
2416 * ####################################################################
2418 static inline u64 cpu_clock(int cpu)
2420 return sched_clock_cpu(cpu);
2423 static inline u64 local_clock(void)
2425 return sched_clock_cpu(raw_smp_processor_id());
2429 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2431 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2432 * The reason for this explicit opt-in is not to have perf penalty with
2433 * slow sched_clocks.
2435 extern void enable_sched_clock_irqtime(void);
2436 extern void disable_sched_clock_irqtime(void);
2438 static inline void enable_sched_clock_irqtime(void) {}
2439 static inline void disable_sched_clock_irqtime(void) {}
2442 extern unsigned long long
2443 task_sched_runtime(struct task_struct *task);
2445 /* sched_exec is called by processes performing an exec */
2447 extern void sched_exec(void);
2449 #define sched_exec() {}
2452 extern void sched_clock_idle_sleep_event(void);
2453 extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2455 #ifdef CONFIG_HOTPLUG_CPU
2456 extern void idle_task_exit(void);
2458 static inline void idle_task_exit(void) {}
2461 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2462 extern void wake_up_nohz_cpu(int cpu);
2464 static inline void wake_up_nohz_cpu(int cpu) { }
2467 #ifdef CONFIG_NO_HZ_FULL
2468 extern u64 scheduler_tick_max_deferment(void);
2471 #ifdef CONFIG_SCHED_AUTOGROUP
2472 extern void sched_autogroup_create_attach(struct task_struct *p);
2473 extern void sched_autogroup_detach(struct task_struct *p);
2474 extern void sched_autogroup_fork(struct signal_struct *sig);
2475 extern void sched_autogroup_exit(struct signal_struct *sig);
2476 #ifdef CONFIG_PROC_FS
2477 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2478 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2481 static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2482 static inline void sched_autogroup_detach(struct task_struct *p) { }
2483 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2484 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2487 extern int yield_to(struct task_struct *p, bool preempt);
2488 extern void set_user_nice(struct task_struct *p, long nice);
2489 extern int task_prio(const struct task_struct *p);
2491 * task_nice - return the nice value of a given task.
2492 * @p: the task in question.
2494 * Return: The nice value [ -20 ... 0 ... 19 ].
2496 static inline int task_nice(const struct task_struct *p)
2498 return PRIO_TO_NICE((p)->static_prio);
2500 extern int can_nice(const struct task_struct *p, const int nice);
2501 extern int task_curr(const struct task_struct *p);
2502 extern int idle_cpu(int cpu);
2503 extern int sched_setscheduler(struct task_struct *, int,
2504 const struct sched_param *);
2505 extern int sched_setscheduler_nocheck(struct task_struct *, int,
2506 const struct sched_param *);
2507 extern int sched_setattr(struct task_struct *,
2508 const struct sched_attr *);
2509 extern struct task_struct *idle_task(int cpu);
2511 * is_idle_task - is the specified task an idle task?
2512 * @p: the task in question.
2514 * Return: 1 if @p is an idle task. 0 otherwise.
2516 static inline bool is_idle_task(const struct task_struct *p)
2520 extern struct task_struct *curr_task(int cpu);
2521 extern void set_curr_task(int cpu, struct task_struct *p);
2525 union thread_union {
2526 struct thread_info thread_info;
2527 unsigned long stack[THREAD_SIZE/sizeof(long)];
2530 #ifndef __HAVE_ARCH_KSTACK_END
2531 static inline int kstack_end(void *addr)
2533 /* Reliable end of stack detection:
2534 * Some APM bios versions misalign the stack
2536 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2540 extern union thread_union init_thread_union;
2541 extern struct task_struct init_task;
2543 extern struct mm_struct init_mm;
2545 extern struct pid_namespace init_pid_ns;
2548 * find a task by one of its numerical ids
2550 * find_task_by_pid_ns():
2551 * finds a task by its pid in the specified namespace
2552 * find_task_by_vpid():
2553 * finds a task by its virtual pid
2555 * see also find_vpid() etc in include/linux/pid.h
2558 extern struct task_struct *find_task_by_vpid(pid_t nr);
2559 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2560 struct pid_namespace *ns);
2562 /* per-UID process charging. */
2563 extern struct user_struct * alloc_uid(kuid_t);
2564 static inline struct user_struct *get_uid(struct user_struct *u)
2566 atomic_inc(&u->__count);
2569 extern void free_uid(struct user_struct *);
2571 #include <asm/current.h>
2573 extern void xtime_update(unsigned long ticks);
2575 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2576 extern int wake_up_process(struct task_struct *tsk);
2577 extern void wake_up_new_task(struct task_struct *tsk);
2579 extern void kick_process(struct task_struct *tsk);
2581 static inline void kick_process(struct task_struct *tsk) { }
2583 extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2584 extern void sched_dead(struct task_struct *p);
2586 extern void proc_caches_init(void);
2587 extern void flush_signals(struct task_struct *);
2588 extern void ignore_signals(struct task_struct *);
2589 extern void flush_signal_handlers(struct task_struct *, int force_default);
2590 extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2592 static inline int kernel_dequeue_signal(siginfo_t *info)
2594 struct task_struct *tsk = current;
2598 spin_lock_irq(&tsk->sighand->siglock);
2599 ret = dequeue_signal(tsk, &tsk->blocked, info ?: &__info);
2600 spin_unlock_irq(&tsk->sighand->siglock);
2605 static inline void kernel_signal_stop(void)
2607 spin_lock_irq(¤t->sighand->siglock);
2608 if (current->jobctl & JOBCTL_STOP_DEQUEUED)
2609 __set_current_state(TASK_STOPPED);
2610 spin_unlock_irq(¤t->sighand->siglock);
2615 extern void release_task(struct task_struct * p);
2616 extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2617 extern int force_sigsegv(int, struct task_struct *);
2618 extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2619 extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2620 extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2621 extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2622 const struct cred *, u32);
2623 extern int kill_pgrp(struct pid *pid, int sig, int priv);
2624 extern int kill_pid(struct pid *pid, int sig, int priv);
2625 extern int kill_proc_info(int, struct siginfo *, pid_t);
2626 extern __must_check bool do_notify_parent(struct task_struct *, int);
2627 extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2628 extern void force_sig(int, struct task_struct *);
2629 extern int send_sig(int, struct task_struct *, int);
2630 extern int zap_other_threads(struct task_struct *p);
2631 extern struct sigqueue *sigqueue_alloc(void);
2632 extern void sigqueue_free(struct sigqueue *);
2633 extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2634 extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2636 static inline void restore_saved_sigmask(void)
2638 if (test_and_clear_restore_sigmask())
2639 __set_current_blocked(¤t->saved_sigmask);
2642 static inline sigset_t *sigmask_to_save(void)
2644 sigset_t *res = ¤t->blocked;
2645 if (unlikely(test_restore_sigmask()))
2646 res = ¤t->saved_sigmask;
2650 static inline int kill_cad_pid(int sig, int priv)
2652 return kill_pid(cad_pid, sig, priv);
2655 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2656 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2657 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2658 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2661 * True if we are on the alternate signal stack.
2663 static inline int on_sig_stack(unsigned long sp)
2665 #ifdef CONFIG_STACK_GROWSUP
2666 return sp >= current->sas_ss_sp &&
2667 sp - current->sas_ss_sp < current->sas_ss_size;
2669 return sp > current->sas_ss_sp &&
2670 sp - current->sas_ss_sp <= current->sas_ss_size;
2674 static inline int sas_ss_flags(unsigned long sp)
2676 if (!current->sas_ss_size)
2679 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2682 static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2684 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2685 #ifdef CONFIG_STACK_GROWSUP
2686 return current->sas_ss_sp;
2688 return current->sas_ss_sp + current->sas_ss_size;
2694 * Routines for handling mm_structs
2696 extern struct mm_struct * mm_alloc(void);
2698 /* mmdrop drops the mm and the page tables */
2699 extern void __mmdrop(struct mm_struct *);
2700 static inline void mmdrop(struct mm_struct * mm)
2702 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2706 /* mmput gets rid of the mappings and all user-space */
2707 extern void mmput(struct mm_struct *);
2708 /* Grab a reference to a task's mm, if it is not already going away */
2709 extern struct mm_struct *get_task_mm(struct task_struct *task);
2711 * Grab a reference to a task's mm, if it is not already going away
2712 * and ptrace_may_access with the mode parameter passed to it
2715 extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2716 /* Remove the current tasks stale references to the old mm_struct */
2717 extern void mm_release(struct task_struct *, struct mm_struct *);
2719 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2720 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2721 struct task_struct *, unsigned long);
2723 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2724 struct task_struct *);
2726 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2727 * via pt_regs, so ignore the tls argument passed via C. */
2728 static inline int copy_thread_tls(
2729 unsigned long clone_flags, unsigned long sp, unsigned long arg,
2730 struct task_struct *p, unsigned long tls)
2732 return copy_thread(clone_flags, sp, arg, p);
2735 extern void flush_thread(void);
2736 extern void exit_thread(void);
2738 extern void exit_files(struct task_struct *);
2739 extern void __cleanup_sighand(struct sighand_struct *);
2741 extern void exit_itimers(struct signal_struct *);
2742 extern void flush_itimer_signals(void);
2744 extern void do_group_exit(int);
2746 extern int do_execve(struct filename *,
2747 const char __user * const __user *,
2748 const char __user * const __user *);
2749 extern int do_execveat(int, struct filename *,
2750 const char __user * const __user *,
2751 const char __user * const __user *,
2753 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *, unsigned long);
2754 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2755 struct task_struct *fork_idle(int);
2756 extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2758 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2759 static inline void set_task_comm(struct task_struct *tsk, const char *from)
2761 __set_task_comm(tsk, from, false);
2763 extern char *get_task_comm(char *to, struct task_struct *tsk);
2766 void scheduler_ipi(void);
2767 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2769 static inline void scheduler_ipi(void) { }
2770 static inline unsigned long wait_task_inactive(struct task_struct *p,
2777 #define tasklist_empty() \
2778 list_empty(&init_task.tasks)
2780 #define next_task(p) \
2781 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2783 #define for_each_process(p) \
2784 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2786 extern bool current_is_single_threaded(void);
2789 * Careful: do_each_thread/while_each_thread is a double loop so
2790 * 'break' will not work as expected - use goto instead.
2792 #define do_each_thread(g, t) \
2793 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
2795 #define while_each_thread(g, t) \
2796 while ((t = next_thread(t)) != g)
2798 #define __for_each_thread(signal, t) \
2799 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
2801 #define for_each_thread(p, t) \
2802 __for_each_thread((p)->signal, t)
2804 /* Careful: this is a double loop, 'break' won't work as expected. */
2805 #define for_each_process_thread(p, t) \
2806 for_each_process(p) for_each_thread(p, t)
2808 static inline int get_nr_threads(struct task_struct *tsk)
2810 return tsk->signal->nr_threads;
2813 static inline bool thread_group_leader(struct task_struct *p)
2815 return p->exit_signal >= 0;
2818 /* Do to the insanities of de_thread it is possible for a process
2819 * to have the pid of the thread group leader without actually being
2820 * the thread group leader. For iteration through the pids in proc
2821 * all we care about is that we have a task with the appropriate
2822 * pid, we don't actually care if we have the right task.
2824 static inline bool has_group_leader_pid(struct task_struct *p)
2826 return task_pid(p) == p->signal->leader_pid;
2830 bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
2832 return p1->signal == p2->signal;
2835 static inline struct task_struct *next_thread(const struct task_struct *p)
2837 return list_entry_rcu(p->thread_group.next,
2838 struct task_struct, thread_group);
2841 static inline int thread_group_empty(struct task_struct *p)
2843 return list_empty(&p->thread_group);
2846 #define delay_group_leader(p) \
2847 (thread_group_leader(p) && !thread_group_empty(p))
2850 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
2851 * subscriptions and synchronises with wait4(). Also used in procfs. Also
2852 * pins the final release of task.io_context. Also protects ->cpuset and
2853 * ->cgroup.subsys[]. And ->vfork_done.
2855 * Nests both inside and outside of read_lock(&tasklist_lock).
2856 * It must not be nested with write_lock_irq(&tasklist_lock),
2857 * neither inside nor outside.
2859 static inline void task_lock(struct task_struct *p)
2861 spin_lock(&p->alloc_lock);
2864 static inline void task_unlock(struct task_struct *p)
2866 spin_unlock(&p->alloc_lock);
2869 extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
2870 unsigned long *flags);
2872 static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
2873 unsigned long *flags)
2875 struct sighand_struct *ret;
2877 ret = __lock_task_sighand(tsk, flags);
2878 (void)__cond_lock(&tsk->sighand->siglock, ret);
2882 static inline void unlock_task_sighand(struct task_struct *tsk,
2883 unsigned long *flags)
2885 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
2889 * threadgroup_change_begin - mark the beginning of changes to a threadgroup
2890 * @tsk: task causing the changes
2892 * All operations which modify a threadgroup - a new thread joining the
2893 * group, death of a member thread (the assertion of PF_EXITING) and
2894 * exec(2) dethreading the process and replacing the leader - are wrapped
2895 * by threadgroup_change_{begin|end}(). This is to provide a place which
2896 * subsystems needing threadgroup stability can hook into for
2899 static inline void threadgroup_change_begin(struct task_struct *tsk)
2902 cgroup_threadgroup_change_begin(tsk);
2906 * threadgroup_change_end - mark the end of changes to a threadgroup
2907 * @tsk: task causing the changes
2909 * See threadgroup_change_begin().
2911 static inline void threadgroup_change_end(struct task_struct *tsk)
2913 cgroup_threadgroup_change_end(tsk);
2916 #ifndef __HAVE_THREAD_FUNCTIONS
2918 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
2919 #define task_stack_page(task) ((task)->stack)
2921 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
2923 *task_thread_info(p) = *task_thread_info(org);
2924 task_thread_info(p)->task = p;
2928 * Return the address of the last usable long on the stack.
2930 * When the stack grows down, this is just above the thread
2931 * info struct. Going any lower will corrupt the threadinfo.
2933 * When the stack grows up, this is the highest address.
2934 * Beyond that position, we corrupt data on the next page.
2936 static inline unsigned long *end_of_stack(struct task_struct *p)
2938 #ifdef CONFIG_STACK_GROWSUP
2939 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
2941 return (unsigned long *)(task_thread_info(p) + 1);
2946 #define task_stack_end_corrupted(task) \
2947 (*(end_of_stack(task)) != STACK_END_MAGIC)
2949 static inline int object_is_on_stack(void *obj)
2951 void *stack = task_stack_page(current);
2953 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
2956 extern void thread_info_cache_init(void);
2958 #ifdef CONFIG_DEBUG_STACK_USAGE
2959 static inline unsigned long stack_not_used(struct task_struct *p)
2961 unsigned long *n = end_of_stack(p);
2963 do { /* Skip over canary */
2964 # ifdef CONFIG_STACK_GROWSUP
2971 # ifdef CONFIG_STACK_GROWSUP
2972 return (unsigned long)end_of_stack(p) - (unsigned long)n;
2974 return (unsigned long)n - (unsigned long)end_of_stack(p);
2978 extern void set_task_stack_end_magic(struct task_struct *tsk);
2980 /* set thread flags in other task's structures
2981 * - see asm/thread_info.h for TIF_xxxx flags available
2983 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2985 set_ti_thread_flag(task_thread_info(tsk), flag);
2988 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2990 clear_ti_thread_flag(task_thread_info(tsk), flag);
2993 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2995 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2998 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
3000 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
3003 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
3005 return test_ti_thread_flag(task_thread_info(tsk), flag);
3008 static inline void set_tsk_need_resched(struct task_struct *tsk)
3010 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3013 static inline void clear_tsk_need_resched(struct task_struct *tsk)
3015 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
3018 static inline int test_tsk_need_resched(struct task_struct *tsk)
3020 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
3023 static inline int restart_syscall(void)
3025 set_tsk_thread_flag(current, TIF_SIGPENDING);
3026 return -ERESTARTNOINTR;
3029 static inline int signal_pending(struct task_struct *p)
3031 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
3034 static inline int __fatal_signal_pending(struct task_struct *p)
3036 return unlikely(sigismember(&p->pending.signal, SIGKILL));
3039 static inline int fatal_signal_pending(struct task_struct *p)
3041 return signal_pending(p) && __fatal_signal_pending(p);
3044 static inline int signal_pending_state(long state, struct task_struct *p)
3046 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
3048 if (!signal_pending(p))
3051 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
3055 * cond_resched() and cond_resched_lock(): latency reduction via
3056 * explicit rescheduling in places that are safe. The return
3057 * value indicates whether a reschedule was done in fact.
3058 * cond_resched_lock() will drop the spinlock before scheduling,
3059 * cond_resched_softirq() will enable bhs before scheduling.
3061 extern int _cond_resched(void);
3063 #define cond_resched() ({ \
3064 ___might_sleep(__FILE__, __LINE__, 0); \
3068 extern int __cond_resched_lock(spinlock_t *lock);
3070 #define cond_resched_lock(lock) ({ \
3071 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3072 __cond_resched_lock(lock); \
3075 extern int __cond_resched_softirq(void);
3077 #define cond_resched_softirq() ({ \
3078 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3079 __cond_resched_softirq(); \
3082 static inline void cond_resched_rcu(void)
3084 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3092 * Does a critical section need to be broken due to another
3093 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3094 * but a general need for low latency)
3096 static inline int spin_needbreak(spinlock_t *lock)
3098 #ifdef CONFIG_PREEMPT
3099 return spin_is_contended(lock);
3106 * Idle thread specific functions to determine the need_resched
3109 #ifdef TIF_POLLING_NRFLAG
3110 static inline int tsk_is_polling(struct task_struct *p)
3112 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
3115 static inline void __current_set_polling(void)
3117 set_thread_flag(TIF_POLLING_NRFLAG);
3120 static inline bool __must_check current_set_polling_and_test(void)
3122 __current_set_polling();
3125 * Polling state must be visible before we test NEED_RESCHED,
3126 * paired by resched_curr()
3128 smp_mb__after_atomic();
3130 return unlikely(tif_need_resched());
3133 static inline void __current_clr_polling(void)
3135 clear_thread_flag(TIF_POLLING_NRFLAG);
3138 static inline bool __must_check current_clr_polling_and_test(void)
3140 __current_clr_polling();
3143 * Polling state must be visible before we test NEED_RESCHED,
3144 * paired by resched_curr()
3146 smp_mb__after_atomic();
3148 return unlikely(tif_need_resched());
3152 static inline int tsk_is_polling(struct task_struct *p) { return 0; }
3153 static inline void __current_set_polling(void) { }
3154 static inline void __current_clr_polling(void) { }
3156 static inline bool __must_check current_set_polling_and_test(void)
3158 return unlikely(tif_need_resched());
3160 static inline bool __must_check current_clr_polling_and_test(void)
3162 return unlikely(tif_need_resched());
3166 static inline void current_clr_polling(void)
3168 __current_clr_polling();
3171 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3172 * Once the bit is cleared, we'll get IPIs with every new
3173 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3176 smp_mb(); /* paired with resched_curr() */
3178 preempt_fold_need_resched();
3181 static __always_inline bool need_resched(void)
3183 return unlikely(tif_need_resched());
3187 * Thread group CPU time accounting.
3189 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
3190 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
3193 * Reevaluate whether the task has signals pending delivery.
3194 * Wake the task if so.
3195 * This is required every time the blocked sigset_t changes.
3196 * callers must hold sighand->siglock.
3198 extern void recalc_sigpending_and_wake(struct task_struct *t);
3199 extern void recalc_sigpending(void);
3201 extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
3203 static inline void signal_wake_up(struct task_struct *t, bool resume)
3205 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
3207 static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
3209 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
3213 * Wrappers for p->thread_info->cpu access. No-op on UP.
3217 static inline unsigned int task_cpu(const struct task_struct *p)
3219 return task_thread_info(p)->cpu;
3222 static inline int task_node(const struct task_struct *p)
3224 return cpu_to_node(task_cpu(p));
3227 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
3231 static inline unsigned int task_cpu(const struct task_struct *p)
3236 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
3240 #endif /* CONFIG_SMP */
3242 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
3243 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
3245 #ifdef CONFIG_CGROUP_SCHED
3246 extern struct task_group root_task_group;
3247 #endif /* CONFIG_CGROUP_SCHED */
3249 extern int task_can_switch_user(struct user_struct *up,
3250 struct task_struct *tsk);
3252 #ifdef CONFIG_TASK_XACCT
3253 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3255 tsk->ioac.rchar += amt;
3258 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3260 tsk->ioac.wchar += amt;
3263 static inline void inc_syscr(struct task_struct *tsk)
3268 static inline void inc_syscw(struct task_struct *tsk)
3273 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
3277 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
3281 static inline void inc_syscr(struct task_struct *tsk)
3285 static inline void inc_syscw(struct task_struct *tsk)
3290 #ifndef TASK_SIZE_OF
3291 #define TASK_SIZE_OF(tsk) TASK_SIZE
3295 extern void mm_update_next_owner(struct mm_struct *mm);
3297 static inline void mm_update_next_owner(struct mm_struct *mm)
3300 #endif /* CONFIG_MEMCG */
3302 static inline unsigned long task_rlimit(const struct task_struct *tsk,
3305 return READ_ONCE(tsk->signal->rlim[limit].rlim_cur);
3308 static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3311 return READ_ONCE(tsk->signal->rlim[limit].rlim_max);
3314 static inline unsigned long rlimit(unsigned int limit)
3316 return task_rlimit(current, limit);
3319 static inline unsigned long rlimit_max(unsigned int limit)
3321 return task_rlimit_max(current, limit);
3324 #ifdef CONFIG_CPU_FREQ
3325 struct update_util_data {
3326 void (*func)(struct update_util_data *data,
3327 u64 time, unsigned long util, unsigned long max);
3330 void cpufreq_set_update_util_data(int cpu, struct update_util_data *data);
3331 #endif /* CONFIG_CPU_FREQ */