4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
13 #include <asm/param.h> /* for HZ */
15 #include <linux/capability.h>
16 #include <linux/threads.h>
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/timex.h>
20 #include <linux/jiffies.h>
21 #include <linux/plist.h>
22 #include <linux/rbtree.h>
23 #include <linux/thread_info.h>
24 #include <linux/cpumask.h>
25 #include <linux/errno.h>
26 #include <linux/nodemask.h>
27 #include <linux/mm_types.h>
28 #include <linux/preempt.h>
31 #include <asm/ptrace.h>
33 #include <linux/smp.h>
34 #include <linux/sem.h>
35 #include <linux/shm.h>
36 #include <linux/signal.h>
37 #include <linux/compiler.h>
38 #include <linux/completion.h>
39 #include <linux/pid.h>
40 #include <linux/percpu.h>
41 #include <linux/topology.h>
42 #include <linux/seccomp.h>
43 #include <linux/rcupdate.h>
44 #include <linux/rculist.h>
45 #include <linux/rtmutex.h>
47 #include <linux/time.h>
48 #include <linux/param.h>
49 #include <linux/resource.h>
50 #include <linux/timer.h>
51 #include <linux/hrtimer.h>
52 #include <linux/kcov.h>
53 #include <linux/task_io_accounting.h>
54 #include <linux/latencytop.h>
55 #include <linux/cred.h>
56 #include <linux/llist.h>
57 #include <linux/uidgid.h>
58 #include <linux/gfp.h>
59 #include <linux/magic.h>
60 #include <linux/cgroup-defs.h>
62 #include <asm/processor.h>
64 #define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
67 * Extended scheduling parameters data structure.
69 * This is needed because the original struct sched_param can not be
70 * altered without introducing ABI issues with legacy applications
71 * (e.g., in sched_getparam()).
73 * However, the possibility of specifying more than just a priority for
74 * the tasks may be useful for a wide variety of application fields, e.g.,
75 * multimedia, streaming, automation and control, and many others.
77 * This variant (sched_attr) is meant at describing a so-called
78 * sporadic time-constrained task. In such model a task is specified by:
79 * - the activation period or minimum instance inter-arrival time;
80 * - the maximum (or average, depending on the actual scheduling
81 * discipline) computation time of all instances, a.k.a. runtime;
82 * - the deadline (relative to the actual activation time) of each
84 * Very briefly, a periodic (sporadic) task asks for the execution of
85 * some specific computation --which is typically called an instance--
86 * (at most) every period. Moreover, each instance typically lasts no more
87 * than the runtime and must be completed by time instant t equal to
88 * the instance activation time + the deadline.
90 * This is reflected by the actual fields of the sched_attr structure:
92 * @size size of the structure, for fwd/bwd compat.
94 * @sched_policy task's scheduling policy
95 * @sched_flags for customizing the scheduler behaviour
96 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
97 * @sched_priority task's static priority (SCHED_FIFO/RR)
98 * @sched_deadline representative of the task's deadline
99 * @sched_runtime representative of the task's runtime
100 * @sched_period representative of the task's period
102 * Given this task model, there are a multiplicity of scheduling algorithms
103 * and policies, that can be used to ensure all the tasks will make their
104 * timing constraints.
106 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
107 * only user of this new interface. More information about the algorithm
108 * available in the scheduling class file or in Documentation/.
116 /* SCHED_NORMAL, SCHED_BATCH */
119 /* SCHED_FIFO, SCHED_RR */
128 struct futex_pi_state
;
129 struct robust_list_head
;
132 struct perf_event_context
;
138 * These are the constant used to fake the fixed-point load-average
139 * counting. Some notes:
140 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
141 * a load-average precision of 10 bits integer + 11 bits fractional
142 * - if you want to count load-averages more often, you need more
143 * precision, or rounding will get you. With 2-second counting freq,
144 * the EXP_n values would be 1981, 2034 and 2043 if still using only
147 extern unsigned long avenrun
[]; /* Load averages */
148 extern void get_avenrun(unsigned long *loads
, unsigned long offset
, int shift
);
150 #define FSHIFT 11 /* nr of bits of precision */
151 #define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
152 #define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
153 #define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
154 #define EXP_5 2014 /* 1/exp(5sec/5min) */
155 #define EXP_15 2037 /* 1/exp(5sec/15min) */
157 #define CALC_LOAD(load,exp,n) \
159 load += n*(FIXED_1-exp); \
162 extern unsigned long total_forks
;
163 extern int nr_threads
;
164 DECLARE_PER_CPU(unsigned long, process_counts
);
165 extern int nr_processes(void);
166 extern unsigned long nr_running(void);
167 extern bool single_task_running(void);
168 extern unsigned long nr_iowait(void);
169 extern unsigned long nr_iowait_cpu(int cpu
);
170 extern void get_iowait_load(unsigned long *nr_waiters
, unsigned long *load
);
172 extern void calc_global_load(unsigned long ticks
);
174 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
175 extern void cpu_load_update_nohz_start(void);
176 extern void cpu_load_update_nohz_stop(void);
178 static inline void cpu_load_update_nohz_start(void) { }
179 static inline void cpu_load_update_nohz_stop(void) { }
182 extern void dump_cpu_task(int cpu
);
187 #ifdef CONFIG_SCHED_DEBUG
188 extern void proc_sched_show_task(struct task_struct
*p
, struct seq_file
*m
);
189 extern void proc_sched_set_task(struct task_struct
*p
);
193 * Task state bitmask. NOTE! These bits are also
194 * encoded in fs/proc/array.c: get_task_state().
196 * We have two separate sets of flags: task->state
197 * is about runnability, while task->exit_state are
198 * about the task exiting. Confusing, but this way
199 * modifying one set can't modify the other one by
202 #define TASK_RUNNING 0
203 #define TASK_INTERRUPTIBLE 1
204 #define TASK_UNINTERRUPTIBLE 2
205 #define __TASK_STOPPED 4
206 #define __TASK_TRACED 8
207 /* in tsk->exit_state */
209 #define EXIT_ZOMBIE 32
210 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
211 /* in tsk->state again */
213 #define TASK_WAKEKILL 128
214 #define TASK_WAKING 256
215 #define TASK_PARKED 512
216 #define TASK_NOLOAD 1024
217 #define TASK_NEW 2048
218 #define TASK_STATE_MAX 4096
220 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
222 /* Convenience macros for the sake of set_current_state */
223 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
224 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
225 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
227 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
229 /* Convenience macros for the sake of wake_up */
230 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
231 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
233 /* get_task_state() */
234 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
235 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
236 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
238 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
239 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
240 #define task_is_stopped_or_traced(task) \
241 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
242 #define task_contributes_to_load(task) \
243 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
244 (task->flags & PF_FROZEN) == 0 && \
245 (task->state & TASK_NOLOAD) == 0)
247 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
249 #define __set_current_state(state_value) \
251 current->task_state_change = _THIS_IP_; \
252 current->state = (state_value); \
254 #define set_current_state(state_value) \
256 current->task_state_change = _THIS_IP_; \
257 smp_store_mb(current->state, (state_value)); \
262 * set_current_state() includes a barrier so that the write of current->state
263 * is correctly serialised wrt the caller's subsequent test of whether to
267 * set_current_state(TASK_UNINTERRUPTIBLE);
273 * __set_current_state(TASK_RUNNING);
275 * If the caller does not need such serialisation (because, for instance, the
276 * condition test and condition change and wakeup are under the same lock) then
277 * use __set_current_state().
279 * The above is typically ordered against the wakeup, which does:
281 * need_sleep = false;
282 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
284 * Where wake_up_state() (and all other wakeup primitives) imply enough
285 * barriers to order the store of the variable against wakeup.
287 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
288 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
289 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
291 * This is obviously fine, since they both store the exact same value.
293 * Also see the comments of try_to_wake_up().
295 #define __set_current_state(state_value) \
296 do { current->state = (state_value); } while (0)
297 #define set_current_state(state_value) \
298 smp_store_mb(current->state, (state_value))
302 /* Task command name length */
303 #define TASK_COMM_LEN 16
305 #include <linux/spinlock.h>
308 * This serializes "schedule()" and also protects
309 * the run-queue from deletions/modifications (but
310 * _adding_ to the beginning of the run-queue has
313 extern rwlock_t tasklist_lock
;
314 extern spinlock_t mmlist_lock
;
318 #ifdef CONFIG_PROVE_RCU
319 extern int lockdep_tasklist_lock_is_held(void);
320 #endif /* #ifdef CONFIG_PROVE_RCU */
322 extern void sched_init(void);
323 extern void sched_init_smp(void);
324 extern asmlinkage
void schedule_tail(struct task_struct
*prev
);
325 extern void init_idle(struct task_struct
*idle
, int cpu
);
326 extern void init_idle_bootup_task(struct task_struct
*idle
);
328 extern cpumask_var_t cpu_isolated_map
;
330 extern int runqueue_is_locked(int cpu
);
332 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
333 extern void nohz_balance_enter_idle(int cpu
);
334 extern void set_cpu_sd_state_idle(void);
335 extern int get_nohz_timer_target(void);
337 static inline void nohz_balance_enter_idle(int cpu
) { }
338 static inline void set_cpu_sd_state_idle(void) { }
342 * Only dump TASK_* tasks. (0 for all tasks)
344 extern void show_state_filter(unsigned long state_filter
);
346 static inline void show_state(void)
348 show_state_filter(0);
351 extern void show_regs(struct pt_regs
*);
354 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
355 * task), SP is the stack pointer of the first frame that should be shown in the back
356 * trace (or NULL if the entire call-chain of the task should be shown).
358 extern void show_stack(struct task_struct
*task
, unsigned long *sp
);
360 extern void cpu_init (void);
361 extern void trap_init(void);
362 extern void update_process_times(int user
);
363 extern void scheduler_tick(void);
364 extern int sched_cpu_starting(unsigned int cpu
);
365 extern int sched_cpu_activate(unsigned int cpu
);
366 extern int sched_cpu_deactivate(unsigned int cpu
);
368 #ifdef CONFIG_HOTPLUG_CPU
369 extern int sched_cpu_dying(unsigned int cpu
);
371 # define sched_cpu_dying NULL
374 extern void sched_show_task(struct task_struct
*p
);
376 #ifdef CONFIG_LOCKUP_DETECTOR
377 extern void touch_softlockup_watchdog_sched(void);
378 extern void touch_softlockup_watchdog(void);
379 extern void touch_softlockup_watchdog_sync(void);
380 extern void touch_all_softlockup_watchdogs(void);
381 extern int proc_dowatchdog_thresh(struct ctl_table
*table
, int write
,
383 size_t *lenp
, loff_t
*ppos
);
384 extern unsigned int softlockup_panic
;
385 extern unsigned int hardlockup_panic
;
386 void lockup_detector_init(void);
388 static inline void touch_softlockup_watchdog_sched(void)
391 static inline void touch_softlockup_watchdog(void)
394 static inline void touch_softlockup_watchdog_sync(void)
397 static inline void touch_all_softlockup_watchdogs(void)
400 static inline void lockup_detector_init(void)
405 #ifdef CONFIG_DETECT_HUNG_TASK
406 void reset_hung_task_detector(void);
408 static inline void reset_hung_task_detector(void)
413 /* Attach to any functions which should be ignored in wchan output. */
414 #define __sched __attribute__((__section__(".sched.text")))
416 /* Linker adds these: start and end of __sched functions */
417 extern char __sched_text_start
[], __sched_text_end
[];
419 /* Is this address in the __sched functions? */
420 extern int in_sched_functions(unsigned long addr
);
422 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
423 extern signed long schedule_timeout(signed long timeout
);
424 extern signed long schedule_timeout_interruptible(signed long timeout
);
425 extern signed long schedule_timeout_killable(signed long timeout
);
426 extern signed long schedule_timeout_uninterruptible(signed long timeout
);
427 extern signed long schedule_timeout_idle(signed long timeout
);
428 asmlinkage
void schedule(void);
429 extern void schedule_preempt_disabled(void);
431 extern int __must_check
io_schedule_prepare(void);
432 extern void io_schedule_finish(int token
);
433 extern long io_schedule_timeout(long timeout
);
434 extern void io_schedule(void);
436 void __noreturn
do_task_dead(void);
439 struct user_namespace
;
442 extern void arch_pick_mmap_layout(struct mm_struct
*mm
);
444 arch_get_unmapped_area(struct file
*, unsigned long, unsigned long,
445 unsigned long, unsigned long);
447 arch_get_unmapped_area_topdown(struct file
*filp
, unsigned long addr
,
448 unsigned long len
, unsigned long pgoff
,
449 unsigned long flags
);
451 static inline void arch_pick_mmap_layout(struct mm_struct
*mm
) {}
454 #define SUID_DUMP_DISABLE 0 /* No setuid dumping */
455 #define SUID_DUMP_USER 1 /* Dump as user of process */
456 #define SUID_DUMP_ROOT 2 /* Dump as root */
460 /* for SUID_DUMP_* above */
461 #define MMF_DUMPABLE_BITS 2
462 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
464 extern void set_dumpable(struct mm_struct
*mm
, int value
);
466 * This returns the actual value of the suid_dumpable flag. For things
467 * that are using this for checking for privilege transitions, it must
468 * test against SUID_DUMP_USER rather than treating it as a boolean
471 static inline int __get_dumpable(unsigned long mm_flags
)
473 return mm_flags
& MMF_DUMPABLE_MASK
;
476 static inline int get_dumpable(struct mm_struct
*mm
)
478 return __get_dumpable(mm
->flags
);
481 /* coredump filter bits */
482 #define MMF_DUMP_ANON_PRIVATE 2
483 #define MMF_DUMP_ANON_SHARED 3
484 #define MMF_DUMP_MAPPED_PRIVATE 4
485 #define MMF_DUMP_MAPPED_SHARED 5
486 #define MMF_DUMP_ELF_HEADERS 6
487 #define MMF_DUMP_HUGETLB_PRIVATE 7
488 #define MMF_DUMP_HUGETLB_SHARED 8
489 #define MMF_DUMP_DAX_PRIVATE 9
490 #define MMF_DUMP_DAX_SHARED 10
492 #define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
493 #define MMF_DUMP_FILTER_BITS 9
494 #define MMF_DUMP_FILTER_MASK \
495 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
496 #define MMF_DUMP_FILTER_DEFAULT \
497 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
498 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
500 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
501 # define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
503 # define MMF_DUMP_MASK_DEFAULT_ELF 0
505 /* leave room for more dump flags */
506 #define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
507 #define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
509 * This one-shot flag is dropped due to necessity of changing exe once again
512 //#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
514 #define MMF_HAS_UPROBES 19 /* has uprobes */
515 #define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
516 #define MMF_OOM_SKIP 21 /* mm is of no interest for the OOM killer */
517 #define MMF_UNSTABLE 22 /* mm is unstable for copy_from_user */
518 #define MMF_HUGE_ZERO_PAGE 23 /* mm has ever used the global huge zero page */
520 #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
522 struct sighand_struct
{
524 struct k_sigaction action
[_NSIG
];
526 wait_queue_head_t signalfd_wqh
;
529 struct pacct_struct
{
532 unsigned long ac_mem
;
533 u64 ac_utime
, ac_stime
;
534 unsigned long ac_minflt
, ac_majflt
;
543 * struct prev_cputime - snaphsot of system and user cputime
544 * @utime: time spent in user mode
545 * @stime: time spent in system mode
546 * @lock: protects the above two fields
548 * Stores previous user/system time values such that we can guarantee
551 struct prev_cputime
{
552 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
559 static inline void prev_cputime_init(struct prev_cputime
*prev
)
561 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
562 prev
->utime
= prev
->stime
= 0;
563 raw_spin_lock_init(&prev
->lock
);
568 * struct task_cputime - collected CPU time counts
569 * @utime: time spent in user mode, in nanoseconds
570 * @stime: time spent in kernel mode, in nanoseconds
571 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
573 * This structure groups together three kinds of CPU time that are tracked for
574 * threads and thread groups. Most things considering CPU time want to group
575 * these counts together and treat all three of them in parallel.
577 struct task_cputime
{
580 unsigned long long sum_exec_runtime
;
583 /* Alternate field names when used to cache expirations. */
584 #define virt_exp utime
585 #define prof_exp stime
586 #define sched_exp sum_exec_runtime
589 * This is the atomic variant of task_cputime, which can be used for
590 * storing and updating task_cputime statistics without locking.
592 struct task_cputime_atomic
{
595 atomic64_t sum_exec_runtime
;
598 #define INIT_CPUTIME_ATOMIC \
599 (struct task_cputime_atomic) { \
600 .utime = ATOMIC64_INIT(0), \
601 .stime = ATOMIC64_INIT(0), \
602 .sum_exec_runtime = ATOMIC64_INIT(0), \
605 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
608 * Disable preemption until the scheduler is running -- use an unconditional
609 * value so that it also works on !PREEMPT_COUNT kernels.
611 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
613 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
616 * Initial preempt_count value; reflects the preempt_count schedule invariant
617 * which states that during context switches:
619 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
621 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
622 * Note: See finish_task_switch().
624 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
627 * struct thread_group_cputimer - thread group interval timer counts
628 * @cputime_atomic: atomic thread group interval timers.
629 * @running: true when there are timers running and
630 * @cputime_atomic receives updates.
631 * @checking_timer: true when a thread in the group is in the
632 * process of checking for thread group timers.
634 * This structure contains the version of task_cputime, above, that is
635 * used for thread group CPU timer calculations.
637 struct thread_group_cputimer
{
638 struct task_cputime_atomic cputime_atomic
;
643 #include <linux/rwsem.h>
647 * NOTE! "signal_struct" does not have its own
648 * locking, because a shared signal_struct always
649 * implies a shared sighand_struct, so locking
650 * sighand_struct is always a proper superset of
651 * the locking of signal_struct.
653 struct signal_struct
{
657 struct list_head thread_head
;
659 wait_queue_head_t wait_chldexit
; /* for wait4() */
661 /* current thread group signal load-balancing target: */
662 struct task_struct
*curr_target
;
664 /* shared signal handling: */
665 struct sigpending shared_pending
;
667 /* thread group exit support */
670 * - notify group_exit_task when ->count is equal to notify_count
671 * - everyone except group_exit_task is stopped during signal delivery
672 * of fatal signals, group_exit_task processes the signal.
675 struct task_struct
*group_exit_task
;
677 /* thread group stop support, overloads group_exit_code too */
678 int group_stop_count
;
679 unsigned int flags
; /* see SIGNAL_* flags below */
682 * PR_SET_CHILD_SUBREAPER marks a process, like a service
683 * manager, to re-parent orphan (double-forking) child processes
684 * to this process instead of 'init'. The service manager is
685 * able to receive SIGCHLD signals and is able to investigate
686 * the process until it calls wait(). All children of this
687 * process will inherit a flag if they should look for a
688 * child_subreaper process at exit.
690 unsigned int is_child_subreaper
:1;
691 unsigned int has_child_subreaper
:1;
693 #ifdef CONFIG_POSIX_TIMERS
695 /* POSIX.1b Interval Timers */
697 struct list_head posix_timers
;
699 /* ITIMER_REAL timer for the process */
700 struct hrtimer real_timer
;
701 ktime_t it_real_incr
;
704 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
705 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
706 * values are defined to 0 and 1 respectively
708 struct cpu_itimer it
[2];
711 * Thread group totals for process CPU timers.
712 * See thread_group_cputimer(), et al, for details.
714 struct thread_group_cputimer cputimer
;
716 /* Earliest-expiration cache. */
717 struct task_cputime cputime_expires
;
719 struct list_head cpu_timers
[3];
723 struct pid
*leader_pid
;
725 #ifdef CONFIG_NO_HZ_FULL
726 atomic_t tick_dep_mask
;
729 struct pid
*tty_old_pgrp
;
731 /* boolean value for session group leader */
734 struct tty_struct
*tty
; /* NULL if no tty */
736 #ifdef CONFIG_SCHED_AUTOGROUP
737 struct autogroup
*autogroup
;
740 * Cumulative resource counters for dead threads in the group,
741 * and for reaped dead child processes forked by this group.
742 * Live threads maintain their own counters and add to these
743 * in __exit_signal, except for the group leader.
745 seqlock_t stats_lock
;
746 u64 utime
, stime
, cutime
, cstime
;
749 struct prev_cputime prev_cputime
;
750 unsigned long nvcsw
, nivcsw
, cnvcsw
, cnivcsw
;
751 unsigned long min_flt
, maj_flt
, cmin_flt
, cmaj_flt
;
752 unsigned long inblock
, oublock
, cinblock
, coublock
;
753 unsigned long maxrss
, cmaxrss
;
754 struct task_io_accounting ioac
;
757 * Cumulative ns of schedule CPU time fo dead threads in the
758 * group, not including a zombie group leader, (This only differs
759 * from jiffies_to_ns(utime + stime) if sched_clock uses something
760 * other than jiffies.)
762 unsigned long long sum_sched_runtime
;
765 * We don't bother to synchronize most readers of this at all,
766 * because there is no reader checking a limit that actually needs
767 * to get both rlim_cur and rlim_max atomically, and either one
768 * alone is a single word that can safely be read normally.
769 * getrlimit/setrlimit use task_lock(current->group_leader) to
770 * protect this instead of the siglock, because they really
771 * have no need to disable irqs.
773 struct rlimit rlim
[RLIM_NLIMITS
];
775 #ifdef CONFIG_BSD_PROCESS_ACCT
776 struct pacct_struct pacct
; /* per-process accounting information */
778 #ifdef CONFIG_TASKSTATS
779 struct taskstats
*stats
;
783 struct tty_audit_buf
*tty_audit_buf
;
787 * Thread is the potential origin of an oom condition; kill first on
790 bool oom_flag_origin
;
791 short oom_score_adj
; /* OOM kill score adjustment */
792 short oom_score_adj_min
; /* OOM kill score adjustment min value.
793 * Only settable by CAP_SYS_RESOURCE. */
794 struct mm_struct
*oom_mm
; /* recorded mm when the thread group got
795 * killed by the oom killer */
797 struct mutex cred_guard_mutex
; /* guard against foreign influences on
798 * credential calculations
799 * (notably. ptrace) */
803 * Bits in flags field of signal_struct.
805 #define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
806 #define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
807 #define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
808 #define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
810 * Pending notifications to parent.
812 #define SIGNAL_CLD_STOPPED 0x00000010
813 #define SIGNAL_CLD_CONTINUED 0x00000020
814 #define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
816 #define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
818 #define SIGNAL_STOP_MASK (SIGNAL_CLD_MASK | SIGNAL_STOP_STOPPED | \
819 SIGNAL_STOP_CONTINUED)
821 static inline void signal_set_stop_flags(struct signal_struct
*sig
,
824 WARN_ON(sig
->flags
& (SIGNAL_GROUP_EXIT
|SIGNAL_GROUP_COREDUMP
));
825 sig
->flags
= (sig
->flags
& ~SIGNAL_STOP_MASK
) | flags
;
828 /* If true, all threads except ->group_exit_task have pending SIGKILL */
829 static inline int signal_group_exit(const struct signal_struct
*sig
)
831 return (sig
->flags
& SIGNAL_GROUP_EXIT
) ||
832 (sig
->group_exit_task
!= NULL
);
836 * Some day this will be a full-fledged user tracking system..
839 atomic_t __count
; /* reference count */
840 atomic_t processes
; /* How many processes does this user have? */
841 atomic_t sigpending
; /* How many pending signals does this user have? */
842 #ifdef CONFIG_FANOTIFY
843 atomic_t fanotify_listeners
;
846 atomic_long_t epoll_watches
; /* The number of file descriptors currently watched */
848 #ifdef CONFIG_POSIX_MQUEUE
849 /* protected by mq_lock */
850 unsigned long mq_bytes
; /* How many bytes can be allocated to mqueue? */
852 unsigned long locked_shm
; /* How many pages of mlocked shm ? */
853 unsigned long unix_inflight
; /* How many files in flight in unix sockets */
854 atomic_long_t pipe_bufs
; /* how many pages are allocated in pipe buffers */
857 struct key
*uid_keyring
; /* UID specific keyring */
858 struct key
*session_keyring
; /* UID's default session keyring */
861 /* Hash table maintenance information */
862 struct hlist_node uidhash_node
;
865 #if defined(CONFIG_PERF_EVENTS) || defined(CONFIG_BPF_SYSCALL)
866 atomic_long_t locked_vm
;
870 extern int uids_sysfs_init(void);
872 extern struct user_struct
*find_user(kuid_t
);
874 extern struct user_struct root_user
;
875 #define INIT_USER (&root_user)
878 struct backing_dev_info
;
879 struct reclaim_state
;
881 #ifdef CONFIG_SCHED_INFO
883 /* cumulative counters */
884 unsigned long pcount
; /* # of times run on this cpu */
885 unsigned long long run_delay
; /* time spent waiting on a runqueue */
888 unsigned long long last_arrival
,/* when we last ran on a cpu */
889 last_queued
; /* when we were last queued to run */
891 #endif /* CONFIG_SCHED_INFO */
893 #ifdef CONFIG_TASK_DELAY_ACCT
894 struct task_delay_info
{
896 unsigned int flags
; /* Private per-task flags */
898 /* For each stat XXX, add following, aligned appropriately
900 * struct timespec XXX_start, XXX_end;
904 * Atomicity of updates to XXX_delay, XXX_count protected by
905 * single lock above (split into XXX_lock if contention is an issue).
909 * XXX_count is incremented on every XXX operation, the delay
910 * associated with the operation is added to XXX_delay.
911 * XXX_delay contains the accumulated delay time in nanoseconds.
913 u64 blkio_start
; /* Shared by blkio, swapin */
914 u64 blkio_delay
; /* wait for sync block io completion */
915 u64 swapin_delay
; /* wait for swapin block io completion */
916 u32 blkio_count
; /* total count of the number of sync block */
917 /* io operations performed */
918 u32 swapin_count
; /* total count of the number of swapin block */
919 /* io operations performed */
922 u64 freepages_delay
; /* wait for memory reclaim */
923 u32 freepages_count
; /* total count of memory reclaim */
925 #endif /* CONFIG_TASK_DELAY_ACCT */
927 static inline int sched_info_on(void)
929 #ifdef CONFIG_SCHEDSTATS
931 #elif defined(CONFIG_TASK_DELAY_ACCT)
932 extern int delayacct_on
;
939 #ifdef CONFIG_SCHEDSTATS
940 void force_schedstat_enabled(void);
951 * Integer metrics need fixed point arithmetic, e.g., sched/fair
952 * has a few: load, load_avg, util_avg, freq, and capacity.
954 * We define a basic fixed point arithmetic range, and then formalize
955 * all these metrics based on that basic range.
957 # define SCHED_FIXEDPOINT_SHIFT 10
958 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
961 * Increase resolution of cpu_capacity calculations
963 #define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
964 #define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
967 * Wake-queues are lists of tasks with a pending wakeup, whose
968 * callers have already marked the task as woken internally,
969 * and can thus carry on. A common use case is being able to
970 * do the wakeups once the corresponding user lock as been
973 * We hold reference to each task in the list across the wakeup,
974 * thus guaranteeing that the memory is still valid by the time
975 * the actual wakeups are performed in wake_up_q().
977 * One per task suffices, because there's never a need for a task to be
978 * in two wake queues simultaneously; it is forbidden to abandon a task
979 * in a wake queue (a call to wake_up_q() _must_ follow), so if a task is
980 * already in a wake queue, the wakeup will happen soon and the second
981 * waker can just skip it.
983 * The DEFINE_WAKE_Q macro declares and initializes the list head.
984 * wake_up_q() does NOT reinitialize the list; it's expected to be
985 * called near the end of a function. Otherwise, the list can be
986 * re-initialized for later re-use by wake_q_init().
988 * Note that this can cause spurious wakeups. schedule() callers
989 * must ensure the call is done inside a loop, confirming that the
990 * wakeup condition has in fact occurred.
993 struct wake_q_node
*next
;
997 struct wake_q_node
*first
;
998 struct wake_q_node
**lastp
;
1001 #define WAKE_Q_TAIL ((struct wake_q_node *) 0x01)
1003 #define DEFINE_WAKE_Q(name) \
1004 struct wake_q_head name = { WAKE_Q_TAIL, &name.first }
1006 static inline void wake_q_init(struct wake_q_head
*head
)
1008 head
->first
= WAKE_Q_TAIL
;
1009 head
->lastp
= &head
->first
;
1012 extern void wake_q_add(struct wake_q_head
*head
,
1013 struct task_struct
*task
);
1014 extern void wake_up_q(struct wake_q_head
*head
);
1017 * sched-domains (multiprocessor balancing) declarations:
1020 #define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
1021 #define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
1022 #define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
1023 #define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
1024 #define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
1025 #define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
1026 #define SD_ASYM_CPUCAPACITY 0x0040 /* Groups have different max cpu capacities */
1027 #define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu capacity */
1028 #define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
1029 #define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
1030 #define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
1031 #define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
1032 #define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
1033 #define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
1034 #define SD_NUMA 0x4000 /* cross-node balancing */
1036 #ifdef CONFIG_SCHED_SMT
1037 static inline int cpu_smt_flags(void)
1039 return SD_SHARE_CPUCAPACITY
| SD_SHARE_PKG_RESOURCES
;
1043 #ifdef CONFIG_SCHED_MC
1044 static inline int cpu_core_flags(void)
1046 return SD_SHARE_PKG_RESOURCES
;
1051 static inline int cpu_numa_flags(void)
1057 extern int arch_asym_cpu_priority(int cpu
);
1059 struct sched_domain_attr
{
1060 int relax_domain_level
;
1063 #define SD_ATTR_INIT (struct sched_domain_attr) { \
1064 .relax_domain_level = -1, \
1067 extern int sched_domain_level_max
;
1071 struct sched_domain_shared
{
1073 atomic_t nr_busy_cpus
;
1077 struct sched_domain
{
1078 /* These fields must be setup */
1079 struct sched_domain
*parent
; /* top domain must be null terminated */
1080 struct sched_domain
*child
; /* bottom domain must be null terminated */
1081 struct sched_group
*groups
; /* the balancing groups of the domain */
1082 unsigned long min_interval
; /* Minimum balance interval ms */
1083 unsigned long max_interval
; /* Maximum balance interval ms */
1084 unsigned int busy_factor
; /* less balancing by factor if busy */
1085 unsigned int imbalance_pct
; /* No balance until over watermark */
1086 unsigned int cache_nice_tries
; /* Leave cache hot tasks for # tries */
1087 unsigned int busy_idx
;
1088 unsigned int idle_idx
;
1089 unsigned int newidle_idx
;
1090 unsigned int wake_idx
;
1091 unsigned int forkexec_idx
;
1092 unsigned int smt_gain
;
1094 int nohz_idle
; /* NOHZ IDLE status */
1095 int flags
; /* See SD_* */
1098 /* Runtime fields. */
1099 unsigned long last_balance
; /* init to jiffies. units in jiffies */
1100 unsigned int balance_interval
; /* initialise to 1. units in ms. */
1101 unsigned int nr_balance_failed
; /* initialise to 0 */
1103 /* idle_balance() stats */
1104 u64 max_newidle_lb_cost
;
1105 unsigned long next_decay_max_lb_cost
;
1107 u64 avg_scan_cost
; /* select_idle_sibling */
1109 #ifdef CONFIG_SCHEDSTATS
1110 /* load_balance() stats */
1111 unsigned int lb_count
[CPU_MAX_IDLE_TYPES
];
1112 unsigned int lb_failed
[CPU_MAX_IDLE_TYPES
];
1113 unsigned int lb_balanced
[CPU_MAX_IDLE_TYPES
];
1114 unsigned int lb_imbalance
[CPU_MAX_IDLE_TYPES
];
1115 unsigned int lb_gained
[CPU_MAX_IDLE_TYPES
];
1116 unsigned int lb_hot_gained
[CPU_MAX_IDLE_TYPES
];
1117 unsigned int lb_nobusyg
[CPU_MAX_IDLE_TYPES
];
1118 unsigned int lb_nobusyq
[CPU_MAX_IDLE_TYPES
];
1120 /* Active load balancing */
1121 unsigned int alb_count
;
1122 unsigned int alb_failed
;
1123 unsigned int alb_pushed
;
1125 /* SD_BALANCE_EXEC stats */
1126 unsigned int sbe_count
;
1127 unsigned int sbe_balanced
;
1128 unsigned int sbe_pushed
;
1130 /* SD_BALANCE_FORK stats */
1131 unsigned int sbf_count
;
1132 unsigned int sbf_balanced
;
1133 unsigned int sbf_pushed
;
1135 /* try_to_wake_up() stats */
1136 unsigned int ttwu_wake_remote
;
1137 unsigned int ttwu_move_affine
;
1138 unsigned int ttwu_move_balance
;
1140 #ifdef CONFIG_SCHED_DEBUG
1144 void *private; /* used during construction */
1145 struct rcu_head rcu
; /* used during destruction */
1147 struct sched_domain_shared
*shared
;
1149 unsigned int span_weight
;
1151 * Span of all CPUs in this domain.
1153 * NOTE: this field is variable length. (Allocated dynamically
1154 * by attaching extra space to the end of the structure,
1155 * depending on how many CPUs the kernel has booted up with)
1157 unsigned long span
[0];
1160 static inline struct cpumask
*sched_domain_span(struct sched_domain
*sd
)
1162 return to_cpumask(sd
->span
);
1165 extern void partition_sched_domains(int ndoms_new
, cpumask_var_t doms_new
[],
1166 struct sched_domain_attr
*dattr_new
);
1168 /* Allocate an array of sched domains, for partition_sched_domains(). */
1169 cpumask_var_t
*alloc_sched_domains(unsigned int ndoms
);
1170 void free_sched_domains(cpumask_var_t doms
[], unsigned int ndoms
);
1172 bool cpus_share_cache(int this_cpu
, int that_cpu
);
1174 typedef const struct cpumask
*(*sched_domain_mask_f
)(int cpu
);
1175 typedef int (*sched_domain_flags_f
)(void);
1177 #define SDTL_OVERLAP 0x01
1180 struct sched_domain
**__percpu sd
;
1181 struct sched_domain_shared
**__percpu sds
;
1182 struct sched_group
**__percpu sg
;
1183 struct sched_group_capacity
**__percpu sgc
;
1186 struct sched_domain_topology_level
{
1187 sched_domain_mask_f mask
;
1188 sched_domain_flags_f sd_flags
;
1191 struct sd_data data
;
1192 #ifdef CONFIG_SCHED_DEBUG
1197 extern void set_sched_topology(struct sched_domain_topology_level
*tl
);
1198 extern void wake_up_if_idle(int cpu
);
1200 #ifdef CONFIG_SCHED_DEBUG
1201 # define SD_INIT_NAME(type) .name = #type
1203 # define SD_INIT_NAME(type)
1206 #else /* CONFIG_SMP */
1208 struct sched_domain_attr
;
1211 partition_sched_domains(int ndoms_new
, cpumask_var_t doms_new
[],
1212 struct sched_domain_attr
*dattr_new
)
1216 static inline bool cpus_share_cache(int this_cpu
, int that_cpu
)
1221 #endif /* !CONFIG_SMP */
1224 struct io_context
; /* See blkdev.h */
1227 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1228 extern void prefetch_stack(struct task_struct
*t
);
1230 static inline void prefetch_stack(struct task_struct
*t
) { }
1233 struct audit_context
; /* See audit.c */
1235 struct pipe_inode_info
;
1236 struct uts_namespace
;
1238 struct load_weight
{
1239 unsigned long weight
;
1244 * The load_avg/util_avg accumulates an infinite geometric series
1245 * (see __update_load_avg() in kernel/sched/fair.c).
1247 * [load_avg definition]
1249 * load_avg = runnable% * scale_load_down(load)
1251 * where runnable% is the time ratio that a sched_entity is runnable.
1252 * For cfs_rq, it is the aggregated load_avg of all runnable and
1253 * blocked sched_entities.
1255 * load_avg may also take frequency scaling into account:
1257 * load_avg = runnable% * scale_load_down(load) * freq%
1259 * where freq% is the CPU frequency normalized to the highest frequency.
1261 * [util_avg definition]
1263 * util_avg = running% * SCHED_CAPACITY_SCALE
1265 * where running% is the time ratio that a sched_entity is running on
1266 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
1267 * and blocked sched_entities.
1269 * util_avg may also factor frequency scaling and CPU capacity scaling:
1271 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
1273 * where freq% is the same as above, and capacity% is the CPU capacity
1274 * normalized to the greatest capacity (due to uarch differences, etc).
1276 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
1277 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
1278 * we therefore scale them to as large a range as necessary. This is for
1279 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
1283 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
1284 * with the highest load (=88761), always runnable on a single cfs_rq,
1285 * and should not overflow as the number already hits PID_MAX_LIMIT.
1287 * For all other cases (including 32-bit kernels), struct load_weight's
1288 * weight will overflow first before we do, because:
1290 * Max(load_avg) <= Max(load.weight)
1292 * Then it is the load_weight's responsibility to consider overflow
1296 u64 last_update_time
, load_sum
;
1297 u32 util_sum
, period_contrib
;
1298 unsigned long load_avg
, util_avg
;
1301 #ifdef CONFIG_SCHEDSTATS
1302 struct sched_statistics
{
1312 s64 sum_sleep_runtime
;
1319 u64 nr_migrations_cold
;
1320 u64 nr_failed_migrations_affine
;
1321 u64 nr_failed_migrations_running
;
1322 u64 nr_failed_migrations_hot
;
1323 u64 nr_forced_migrations
;
1326 u64 nr_wakeups_sync
;
1327 u64 nr_wakeups_migrate
;
1328 u64 nr_wakeups_local
;
1329 u64 nr_wakeups_remote
;
1330 u64 nr_wakeups_affine
;
1331 u64 nr_wakeups_affine_attempts
;
1332 u64 nr_wakeups_passive
;
1333 u64 nr_wakeups_idle
;
1337 struct sched_entity
{
1338 struct load_weight load
; /* for load-balancing */
1339 struct rb_node run_node
;
1340 struct list_head group_node
;
1344 u64 sum_exec_runtime
;
1346 u64 prev_sum_exec_runtime
;
1350 #ifdef CONFIG_SCHEDSTATS
1351 struct sched_statistics statistics
;
1354 #ifdef CONFIG_FAIR_GROUP_SCHED
1356 struct sched_entity
*parent
;
1357 /* rq on which this entity is (to be) queued: */
1358 struct cfs_rq
*cfs_rq
;
1359 /* rq "owned" by this entity/group: */
1360 struct cfs_rq
*my_q
;
1365 * Per entity load average tracking.
1367 * Put into separate cache line so it does not
1368 * collide with read-mostly values above.
1370 struct sched_avg avg ____cacheline_aligned_in_smp
;
1374 struct sched_rt_entity
{
1375 struct list_head run_list
;
1376 unsigned long timeout
;
1377 unsigned long watchdog_stamp
;
1378 unsigned int time_slice
;
1379 unsigned short on_rq
;
1380 unsigned short on_list
;
1382 struct sched_rt_entity
*back
;
1383 #ifdef CONFIG_RT_GROUP_SCHED
1384 struct sched_rt_entity
*parent
;
1385 /* rq on which this entity is (to be) queued: */
1386 struct rt_rq
*rt_rq
;
1387 /* rq "owned" by this entity/group: */
1392 struct sched_dl_entity
{
1393 struct rb_node rb_node
;
1396 * Original scheduling parameters. Copied here from sched_attr
1397 * during sched_setattr(), they will remain the same until
1398 * the next sched_setattr().
1400 u64 dl_runtime
; /* maximum runtime for each instance */
1401 u64 dl_deadline
; /* relative deadline of each instance */
1402 u64 dl_period
; /* separation of two instances (period) */
1403 u64 dl_bw
; /* dl_runtime / dl_deadline */
1406 * Actual scheduling parameters. Initialized with the values above,
1407 * they are continously updated during task execution. Note that
1408 * the remaining runtime could be < 0 in case we are in overrun.
1410 s64 runtime
; /* remaining runtime for this instance */
1411 u64 deadline
; /* absolute deadline for this instance */
1412 unsigned int flags
; /* specifying the scheduler behaviour */
1417 * @dl_throttled tells if we exhausted the runtime. If so, the
1418 * task has to wait for a replenishment to be performed at the
1419 * next firing of dl_timer.
1421 * @dl_boosted tells if we are boosted due to DI. If so we are
1422 * outside bandwidth enforcement mechanism (but only until we
1423 * exit the critical section);
1425 * @dl_yielded tells if task gave up the cpu before consuming
1426 * all its available runtime during the last job.
1428 int dl_throttled
, dl_boosted
, dl_yielded
;
1431 * Bandwidth enforcement timer. Each -deadline task has its
1432 * own bandwidth to be enforced, thus we need one timer per task.
1434 struct hrtimer dl_timer
;
1442 u8 pad
; /* Otherwise the compiler can store garbage here. */
1444 u32 s
; /* Set of bits. */
1448 enum perf_event_task_context
{
1449 perf_invalid_context
= -1,
1450 perf_hw_context
= 0,
1452 perf_nr_task_contexts
,
1455 /* Track pages that require TLB flushes */
1456 struct tlbflush_unmap_batch
{
1458 * Each bit set is a CPU that potentially has a TLB entry for one of
1459 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
1461 struct cpumask cpumask
;
1463 /* True if any bit in cpumask is set */
1464 bool flush_required
;
1467 * If true then the PTE was dirty when unmapped. The entry must be
1468 * flushed before IO is initiated or a stale TLB entry potentially
1469 * allows an update without redirtying the page.
1474 struct task_struct
{
1475 #ifdef CONFIG_THREAD_INFO_IN_TASK
1477 * For reasons of header soup (see current_thread_info()), this
1478 * must be the first element of task_struct.
1480 struct thread_info thread_info
;
1482 volatile long state
; /* -1 unrunnable, 0 runnable, >0 stopped */
1485 unsigned int flags
; /* per process flags, defined below */
1486 unsigned int ptrace
;
1489 struct llist_node wake_entry
;
1491 #ifdef CONFIG_THREAD_INFO_IN_TASK
1492 unsigned int cpu
; /* current CPU */
1494 unsigned int wakee_flips
;
1495 unsigned long wakee_flip_decay_ts
;
1496 struct task_struct
*last_wakee
;
1502 int prio
, static_prio
, normal_prio
;
1503 unsigned int rt_priority
;
1504 const struct sched_class
*sched_class
;
1505 struct sched_entity se
;
1506 struct sched_rt_entity rt
;
1507 #ifdef CONFIG_CGROUP_SCHED
1508 struct task_group
*sched_task_group
;
1510 struct sched_dl_entity dl
;
1512 #ifdef CONFIG_PREEMPT_NOTIFIERS
1513 /* list of struct preempt_notifier: */
1514 struct hlist_head preempt_notifiers
;
1517 #ifdef CONFIG_BLK_DEV_IO_TRACE
1518 unsigned int btrace_seq
;
1521 unsigned int policy
;
1522 int nr_cpus_allowed
;
1523 cpumask_t cpus_allowed
;
1525 #ifdef CONFIG_PREEMPT_RCU
1526 int rcu_read_lock_nesting
;
1527 union rcu_special rcu_read_unlock_special
;
1528 struct list_head rcu_node_entry
;
1529 struct rcu_node
*rcu_blocked_node
;
1530 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1531 #ifdef CONFIG_TASKS_RCU
1532 unsigned long rcu_tasks_nvcsw
;
1533 bool rcu_tasks_holdout
;
1534 struct list_head rcu_tasks_holdout_list
;
1535 int rcu_tasks_idle_cpu
;
1536 #endif /* #ifdef CONFIG_TASKS_RCU */
1538 #ifdef CONFIG_SCHED_INFO
1539 struct sched_info sched_info
;
1542 struct list_head tasks
;
1544 struct plist_node pushable_tasks
;
1545 struct rb_node pushable_dl_tasks
;
1548 struct mm_struct
*mm
, *active_mm
;
1550 /* Per-thread vma caching: */
1551 struct vmacache vmacache
;
1553 #if defined(SPLIT_RSS_COUNTING)
1554 struct task_rss_stat rss_stat
;
1558 int exit_code
, exit_signal
;
1559 int pdeath_signal
; /* The signal sent when the parent dies */
1560 unsigned long jobctl
; /* JOBCTL_*, siglock protected */
1562 /* Used for emulating ABI behavior of previous Linux versions */
1563 unsigned int personality
;
1565 /* scheduler bits, serialized by scheduler locks */
1566 unsigned sched_reset_on_fork
:1;
1567 unsigned sched_contributes_to_load
:1;
1568 unsigned sched_migrated
:1;
1569 unsigned sched_remote_wakeup
:1;
1570 unsigned :0; /* force alignment to the next boundary */
1572 /* unserialized, strictly 'current' */
1573 unsigned in_execve
:1; /* bit to tell LSMs we're in execve */
1574 unsigned in_iowait
:1;
1575 #if !defined(TIF_RESTORE_SIGMASK)
1576 unsigned restore_sigmask
:1;
1579 unsigned memcg_may_oom
:1;
1581 unsigned memcg_kmem_skip_account
:1;
1584 #ifdef CONFIG_COMPAT_BRK
1585 unsigned brk_randomized
:1;
1588 unsigned long atomic_flags
; /* Flags needing atomic access. */
1590 struct restart_block restart_block
;
1595 #ifdef CONFIG_CC_STACKPROTECTOR
1596 /* Canary value for the -fstack-protector gcc feature */
1597 unsigned long stack_canary
;
1600 * pointers to (original) parent process, youngest child, younger sibling,
1601 * older sibling, respectively. (p->father can be replaced with
1602 * p->real_parent->pid)
1604 struct task_struct __rcu
*real_parent
; /* real parent process */
1605 struct task_struct __rcu
*parent
; /* recipient of SIGCHLD, wait4() reports */
1607 * children/sibling forms the list of my natural children
1609 struct list_head children
; /* list of my children */
1610 struct list_head sibling
; /* linkage in my parent's children list */
1611 struct task_struct
*group_leader
; /* threadgroup leader */
1614 * ptraced is the list of tasks this task is using ptrace on.
1615 * This includes both natural children and PTRACE_ATTACH targets.
1616 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1618 struct list_head ptraced
;
1619 struct list_head ptrace_entry
;
1621 /* PID/PID hash table linkage. */
1622 struct pid_link pids
[PIDTYPE_MAX
];
1623 struct list_head thread_group
;
1624 struct list_head thread_node
;
1626 struct completion
*vfork_done
; /* for vfork() */
1627 int __user
*set_child_tid
; /* CLONE_CHILD_SETTID */
1628 int __user
*clear_child_tid
; /* CLONE_CHILD_CLEARTID */
1631 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1632 u64 utimescaled
, stimescaled
;
1635 struct prev_cputime prev_cputime
;
1636 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1637 seqcount_t vtime_seqcount
;
1638 unsigned long long vtime_snap
;
1640 /* Task is sleeping or running in a CPU with VTIME inactive */
1642 /* Task runs in userspace in a CPU with VTIME active */
1644 /* Task runs in kernelspace in a CPU with VTIME active */
1646 } vtime_snap_whence
;
1649 #ifdef CONFIG_NO_HZ_FULL
1650 atomic_t tick_dep_mask
;
1652 unsigned long nvcsw
, nivcsw
; /* context switch counts */
1653 u64 start_time
; /* monotonic time in nsec */
1654 u64 real_start_time
; /* boot based time in nsec */
1655 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1656 unsigned long min_flt
, maj_flt
;
1658 #ifdef CONFIG_POSIX_TIMERS
1659 struct task_cputime cputime_expires
;
1660 struct list_head cpu_timers
[3];
1663 /* process credentials */
1664 const struct cred __rcu
*ptracer_cred
; /* Tracer's credentials at attach */
1665 const struct cred __rcu
*real_cred
; /* objective and real subjective task
1666 * credentials (COW) */
1667 const struct cred __rcu
*cred
; /* effective (overridable) subjective task
1668 * credentials (COW) */
1669 char comm
[TASK_COMM_LEN
]; /* executable name excluding path
1670 - access with [gs]et_task_comm (which lock
1671 it with task_lock())
1672 - initialized normally by setup_new_exec */
1673 /* file system info */
1674 struct nameidata
*nameidata
;
1675 #ifdef CONFIG_SYSVIPC
1677 struct sysv_sem sysvsem
;
1678 struct sysv_shm sysvshm
;
1680 #ifdef CONFIG_DETECT_HUNG_TASK
1681 /* hung task detection */
1682 unsigned long last_switch_count
;
1684 /* filesystem information */
1685 struct fs_struct
*fs
;
1686 /* open file information */
1687 struct files_struct
*files
;
1689 struct nsproxy
*nsproxy
;
1690 /* signal handlers */
1691 struct signal_struct
*signal
;
1692 struct sighand_struct
*sighand
;
1694 sigset_t blocked
, real_blocked
;
1695 sigset_t saved_sigmask
; /* restored if set_restore_sigmask() was used */
1696 struct sigpending pending
;
1698 unsigned long sas_ss_sp
;
1700 unsigned sas_ss_flags
;
1702 struct callback_head
*task_works
;
1704 struct audit_context
*audit_context
;
1705 #ifdef CONFIG_AUDITSYSCALL
1707 unsigned int sessionid
;
1709 struct seccomp seccomp
;
1711 /* Thread group tracking */
1714 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1716 spinlock_t alloc_lock
;
1718 /* Protection of the PI data structures: */
1719 raw_spinlock_t pi_lock
;
1721 struct wake_q_node wake_q
;
1723 #ifdef CONFIG_RT_MUTEXES
1724 /* PI waiters blocked on a rt_mutex held by this task */
1725 struct rb_root pi_waiters
;
1726 struct rb_node
*pi_waiters_leftmost
;
1727 /* Deadlock detection and priority inheritance handling */
1728 struct rt_mutex_waiter
*pi_blocked_on
;
1731 #ifdef CONFIG_DEBUG_MUTEXES
1732 /* mutex deadlock detection */
1733 struct mutex_waiter
*blocked_on
;
1735 #ifdef CONFIG_TRACE_IRQFLAGS
1736 unsigned int irq_events
;
1737 unsigned long hardirq_enable_ip
;
1738 unsigned long hardirq_disable_ip
;
1739 unsigned int hardirq_enable_event
;
1740 unsigned int hardirq_disable_event
;
1741 int hardirqs_enabled
;
1742 int hardirq_context
;
1743 unsigned long softirq_disable_ip
;
1744 unsigned long softirq_enable_ip
;
1745 unsigned int softirq_disable_event
;
1746 unsigned int softirq_enable_event
;
1747 int softirqs_enabled
;
1748 int softirq_context
;
1750 #ifdef CONFIG_LOCKDEP
1751 # define MAX_LOCK_DEPTH 48UL
1754 unsigned int lockdep_recursion
;
1755 struct held_lock held_locks
[MAX_LOCK_DEPTH
];
1756 gfp_t lockdep_reclaim_gfp
;
1759 unsigned int in_ubsan
;
1762 /* journalling filesystem info */
1765 /* stacked block device info */
1766 struct bio_list
*bio_list
;
1769 /* stack plugging */
1770 struct blk_plug
*plug
;
1774 struct reclaim_state
*reclaim_state
;
1776 struct backing_dev_info
*backing_dev_info
;
1778 struct io_context
*io_context
;
1780 unsigned long ptrace_message
;
1781 siginfo_t
*last_siginfo
; /* For ptrace use. */
1782 struct task_io_accounting ioac
;
1783 #if defined(CONFIG_TASK_XACCT)
1784 u64 acct_rss_mem1
; /* accumulated rss usage */
1785 u64 acct_vm_mem1
; /* accumulated virtual memory usage */
1786 u64 acct_timexpd
; /* stime + utime since last update */
1788 #ifdef CONFIG_CPUSETS
1789 nodemask_t mems_allowed
; /* Protected by alloc_lock */
1790 seqcount_t mems_allowed_seq
; /* Seqence no to catch updates */
1791 int cpuset_mem_spread_rotor
;
1792 int cpuset_slab_spread_rotor
;
1794 #ifdef CONFIG_CGROUPS
1795 /* Control Group info protected by css_set_lock */
1796 struct css_set __rcu
*cgroups
;
1797 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1798 struct list_head cg_list
;
1800 #ifdef CONFIG_INTEL_RDT_A
1804 struct robust_list_head __user
*robust_list
;
1805 #ifdef CONFIG_COMPAT
1806 struct compat_robust_list_head __user
*compat_robust_list
;
1808 struct list_head pi_state_list
;
1809 struct futex_pi_state
*pi_state_cache
;
1811 #ifdef CONFIG_PERF_EVENTS
1812 struct perf_event_context
*perf_event_ctxp
[perf_nr_task_contexts
];
1813 struct mutex perf_event_mutex
;
1814 struct list_head perf_event_list
;
1816 #ifdef CONFIG_DEBUG_PREEMPT
1817 unsigned long preempt_disable_ip
;
1820 struct mempolicy
*mempolicy
; /* Protected by alloc_lock */
1822 short pref_node_fork
;
1824 #ifdef CONFIG_NUMA_BALANCING
1826 unsigned int numa_scan_period
;
1827 unsigned int numa_scan_period_max
;
1828 int numa_preferred_nid
;
1829 unsigned long numa_migrate_retry
;
1830 u64 node_stamp
; /* migration stamp */
1831 u64 last_task_numa_placement
;
1832 u64 last_sum_exec_runtime
;
1833 struct callback_head numa_work
;
1835 struct list_head numa_entry
;
1836 struct numa_group
*numa_group
;
1839 * numa_faults is an array split into four regions:
1840 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1841 * in this precise order.
1843 * faults_memory: Exponential decaying average of faults on a per-node
1844 * basis. Scheduling placement decisions are made based on these
1845 * counts. The values remain static for the duration of a PTE scan.
1846 * faults_cpu: Track the nodes the process was running on when a NUMA
1847 * hinting fault was incurred.
1848 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1849 * during the current scan window. When the scan completes, the counts
1850 * in faults_memory and faults_cpu decay and these values are copied.
1852 unsigned long *numa_faults
;
1853 unsigned long total_numa_faults
;
1856 * numa_faults_locality tracks if faults recorded during the last
1857 * scan window were remote/local or failed to migrate. The task scan
1858 * period is adapted based on the locality of the faults with different
1859 * weights depending on whether they were shared or private faults
1861 unsigned long numa_faults_locality
[3];
1863 unsigned long numa_pages_migrated
;
1864 #endif /* CONFIG_NUMA_BALANCING */
1866 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1867 struct tlbflush_unmap_batch tlb_ubc
;
1870 struct rcu_head rcu
;
1873 * cache last used pipe for splice
1875 struct pipe_inode_info
*splice_pipe
;
1877 struct page_frag task_frag
;
1879 #ifdef CONFIG_TASK_DELAY_ACCT
1880 struct task_delay_info
*delays
;
1882 #ifdef CONFIG_FAULT_INJECTION
1886 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1887 * balance_dirty_pages() for some dirty throttling pause
1890 int nr_dirtied_pause
;
1891 unsigned long dirty_paused_when
; /* start of a write-and-pause period */
1893 #ifdef CONFIG_LATENCYTOP
1894 int latency_record_count
;
1895 struct latency_record latency_record
[LT_SAVECOUNT
];
1898 * time slack values; these are used to round up poll() and
1899 * select() etc timeout values. These are in nanoseconds.
1902 u64 default_timer_slack_ns
;
1905 unsigned int kasan_depth
;
1907 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1908 /* Index of current stored address in ret_stack */
1910 /* Stack of return addresses for return function tracing */
1911 struct ftrace_ret_stack
*ret_stack
;
1912 /* time stamp for last schedule */
1913 unsigned long long ftrace_timestamp
;
1915 * Number of functions that haven't been traced
1916 * because of depth overrun.
1918 atomic_t trace_overrun
;
1919 /* Pause for the tracing */
1920 atomic_t tracing_graph_pause
;
1922 #ifdef CONFIG_TRACING
1923 /* state flags for use by tracers */
1924 unsigned long trace
;
1925 /* bitmask and counter of trace recursion */
1926 unsigned long trace_recursion
;
1927 #endif /* CONFIG_TRACING */
1929 /* Coverage collection mode enabled for this task (0 if disabled). */
1930 enum kcov_mode kcov_mode
;
1931 /* Size of the kcov_area. */
1933 /* Buffer for coverage collection. */
1935 /* kcov desciptor wired with this task or NULL. */
1939 struct mem_cgroup
*memcg_in_oom
;
1940 gfp_t memcg_oom_gfp_mask
;
1941 int memcg_oom_order
;
1943 /* number of pages to reclaim on returning to userland */
1944 unsigned int memcg_nr_pages_over_high
;
1946 #ifdef CONFIG_UPROBES
1947 struct uprobe_task
*utask
;
1949 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1950 unsigned int sequential_io
;
1951 unsigned int sequential_io_avg
;
1953 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1954 unsigned long task_state_change
;
1956 int pagefault_disabled
;
1958 struct task_struct
*oom_reaper_list
;
1960 #ifdef CONFIG_VMAP_STACK
1961 struct vm_struct
*stack_vm_area
;
1963 #ifdef CONFIG_THREAD_INFO_IN_TASK
1964 /* A live task holds one reference. */
1965 atomic_t stack_refcount
;
1967 /* CPU-specific state of this task */
1968 struct thread_struct thread
;
1970 * WARNING: on x86, 'thread_struct' contains a variable-sized
1971 * structure. It *MUST* be at the end of 'task_struct'.
1973 * Do not put anything below here!
1977 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
1978 extern int arch_task_struct_size __read_mostly
;
1980 # define arch_task_struct_size (sizeof(struct task_struct))
1983 #ifdef CONFIG_VMAP_STACK
1984 static inline struct vm_struct
*task_stack_vm_area(const struct task_struct
*t
)
1986 return t
->stack_vm_area
;
1989 static inline struct vm_struct
*task_stack_vm_area(const struct task_struct
*t
)
1995 #define TNF_MIGRATED 0x01
1996 #define TNF_NO_GROUP 0x02
1997 #define TNF_SHARED 0x04
1998 #define TNF_FAULT_LOCAL 0x08
1999 #define TNF_MIGRATE_FAIL 0x10
2001 static inline bool in_vfork(struct task_struct
*tsk
)
2006 * need RCU to access ->real_parent if CLONE_VM was used along with
2009 * We check real_parent->mm == tsk->mm because CLONE_VFORK does not
2012 * CLONE_VFORK can be used with CLONE_PARENT/CLONE_THREAD and thus
2013 * ->real_parent is not necessarily the task doing vfork(), so in
2014 * theory we can't rely on task_lock() if we want to dereference it.
2016 * And in this case we can't trust the real_parent->mm == tsk->mm
2017 * check, it can be false negative. But we do not care, if init or
2018 * another oom-unkillable task does this it should blame itself.
2021 ret
= tsk
->vfork_done
&& tsk
->real_parent
->mm
== tsk
->mm
;
2027 #ifdef CONFIG_NUMA_BALANCING
2028 extern void task_numa_fault(int last_node
, int node
, int pages
, int flags
);
2029 extern pid_t
task_numa_group_id(struct task_struct
*p
);
2030 extern void set_numabalancing_state(bool enabled
);
2031 extern void task_numa_free(struct task_struct
*p
);
2032 extern bool should_numa_migrate_memory(struct task_struct
*p
, struct page
*page
,
2033 int src_nid
, int dst_cpu
);
2035 static inline void task_numa_fault(int last_node
, int node
, int pages
,
2039 static inline pid_t
task_numa_group_id(struct task_struct
*p
)
2043 static inline void set_numabalancing_state(bool enabled
)
2046 static inline void task_numa_free(struct task_struct
*p
)
2049 static inline bool should_numa_migrate_memory(struct task_struct
*p
,
2050 struct page
*page
, int src_nid
, int dst_cpu
)
2056 static inline struct pid
*task_pid(struct task_struct
*task
)
2058 return task
->pids
[PIDTYPE_PID
].pid
;
2061 static inline struct pid
*task_tgid(struct task_struct
*task
)
2063 return task
->group_leader
->pids
[PIDTYPE_PID
].pid
;
2067 * Without tasklist or rcu lock it is not safe to dereference
2068 * the result of task_pgrp/task_session even if task == current,
2069 * we can race with another thread doing sys_setsid/sys_setpgid.
2071 static inline struct pid
*task_pgrp(struct task_struct
*task
)
2073 return task
->group_leader
->pids
[PIDTYPE_PGID
].pid
;
2076 static inline struct pid
*task_session(struct task_struct
*task
)
2078 return task
->group_leader
->pids
[PIDTYPE_SID
].pid
;
2081 struct pid_namespace
;
2084 * the helpers to get the task's different pids as they are seen
2085 * from various namespaces
2087 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
2088 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
2090 * task_xid_nr_ns() : id seen from the ns specified;
2092 * set_task_vxid() : assigns a virtual id to a task;
2094 * see also pid_nr() etc in include/linux/pid.h
2096 pid_t
__task_pid_nr_ns(struct task_struct
*task
, enum pid_type type
,
2097 struct pid_namespace
*ns
);
2099 static inline pid_t
task_pid_nr(struct task_struct
*tsk
)
2104 static inline pid_t
task_pid_nr_ns(struct task_struct
*tsk
,
2105 struct pid_namespace
*ns
)
2107 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, ns
);
2110 static inline pid_t
task_pid_vnr(struct task_struct
*tsk
)
2112 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, NULL
);
2116 static inline pid_t
task_tgid_nr(struct task_struct
*tsk
)
2121 pid_t
task_tgid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
);
2123 static inline pid_t
task_tgid_vnr(struct task_struct
*tsk
)
2125 return pid_vnr(task_tgid(tsk
));
2129 static inline int pid_alive(const struct task_struct
*p
);
2130 static inline pid_t
task_ppid_nr_ns(const struct task_struct
*tsk
, struct pid_namespace
*ns
)
2136 pid
= task_tgid_nr_ns(rcu_dereference(tsk
->real_parent
), ns
);
2142 static inline pid_t
task_ppid_nr(const struct task_struct
*tsk
)
2144 return task_ppid_nr_ns(tsk
, &init_pid_ns
);
2147 static inline pid_t
task_pgrp_nr_ns(struct task_struct
*tsk
,
2148 struct pid_namespace
*ns
)
2150 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, ns
);
2153 static inline pid_t
task_pgrp_vnr(struct task_struct
*tsk
)
2155 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, NULL
);
2159 static inline pid_t
task_session_nr_ns(struct task_struct
*tsk
,
2160 struct pid_namespace
*ns
)
2162 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, ns
);
2165 static inline pid_t
task_session_vnr(struct task_struct
*tsk
)
2167 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, NULL
);
2170 /* obsolete, do not use */
2171 static inline pid_t
task_pgrp_nr(struct task_struct
*tsk
)
2173 return task_pgrp_nr_ns(tsk
, &init_pid_ns
);
2177 * pid_alive - check that a task structure is not stale
2178 * @p: Task structure to be checked.
2180 * Test if a process is not yet dead (at most zombie state)
2181 * If pid_alive fails, then pointers within the task structure
2182 * can be stale and must not be dereferenced.
2184 * Return: 1 if the process is alive. 0 otherwise.
2186 static inline int pid_alive(const struct task_struct
*p
)
2188 return p
->pids
[PIDTYPE_PID
].pid
!= NULL
;
2192 * is_global_init - check if a task structure is init. Since init
2193 * is free to have sub-threads we need to check tgid.
2194 * @tsk: Task structure to be checked.
2196 * Check if a task structure is the first user space task the kernel created.
2198 * Return: 1 if the task structure is init. 0 otherwise.
2200 static inline int is_global_init(struct task_struct
*tsk
)
2202 return task_tgid_nr(tsk
) == 1;
2205 extern struct pid
*cad_pid
;
2207 extern void free_task(struct task_struct
*tsk
);
2208 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
2210 extern void __put_task_struct(struct task_struct
*t
);
2212 static inline void put_task_struct(struct task_struct
*t
)
2214 if (atomic_dec_and_test(&t
->usage
))
2215 __put_task_struct(t
);
2218 struct task_struct
*task_rcu_dereference(struct task_struct
**ptask
);
2219 struct task_struct
*try_get_task_struct(struct task_struct
**ptask
);
2221 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
2222 extern void task_cputime(struct task_struct
*t
,
2223 u64
*utime
, u64
*stime
);
2224 extern u64
task_gtime(struct task_struct
*t
);
2226 static inline void task_cputime(struct task_struct
*t
,
2227 u64
*utime
, u64
*stime
)
2233 static inline u64
task_gtime(struct task_struct
*t
)
2239 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
2240 static inline void task_cputime_scaled(struct task_struct
*t
,
2244 *utimescaled
= t
->utimescaled
;
2245 *stimescaled
= t
->stimescaled
;
2248 static inline void task_cputime_scaled(struct task_struct
*t
,
2252 task_cputime(t
, utimescaled
, stimescaled
);
2256 extern void task_cputime_adjusted(struct task_struct
*p
, u64
*ut
, u64
*st
);
2257 extern void thread_group_cputime_adjusted(struct task_struct
*p
, u64
*ut
, u64
*st
);
2262 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
2263 #define PF_EXITING 0x00000004 /* getting shut down */
2264 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
2265 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
2266 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
2267 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
2268 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
2269 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
2270 #define PF_DUMPCORE 0x00000200 /* dumped core */
2271 #define PF_SIGNALED 0x00000400 /* killed by a signal */
2272 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
2273 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
2274 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
2275 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
2276 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
2277 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
2278 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
2279 #define PF_KSWAPD 0x00040000 /* I am kswapd */
2280 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
2281 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
2282 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
2283 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
2284 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
2285 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
2286 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
2287 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
2288 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
2289 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
2292 * Only the _current_ task can read/write to tsk->flags, but other
2293 * tasks can access tsk->flags in readonly mode for example
2294 * with tsk_used_math (like during threaded core dumping).
2295 * There is however an exception to this rule during ptrace
2296 * or during fork: the ptracer task is allowed to write to the
2297 * child->flags of its traced child (same goes for fork, the parent
2298 * can write to the child->flags), because we're guaranteed the
2299 * child is not running and in turn not changing child->flags
2300 * at the same time the parent does it.
2302 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
2303 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
2304 #define clear_used_math() clear_stopped_child_used_math(current)
2305 #define set_used_math() set_stopped_child_used_math(current)
2306 #define conditional_stopped_child_used_math(condition, child) \
2307 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
2308 #define conditional_used_math(condition) \
2309 conditional_stopped_child_used_math(condition, current)
2310 #define copy_to_stopped_child_used_math(child) \
2311 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
2312 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
2313 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
2314 #define used_math() tsk_used_math(current)
2316 /* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags
2317 * __GFP_FS is also cleared as it implies __GFP_IO.
2319 static inline gfp_t
memalloc_noio_flags(gfp_t flags
)
2321 if (unlikely(current
->flags
& PF_MEMALLOC_NOIO
))
2322 flags
&= ~(__GFP_IO
| __GFP_FS
);
2326 static inline unsigned int memalloc_noio_save(void)
2328 unsigned int flags
= current
->flags
& PF_MEMALLOC_NOIO
;
2329 current
->flags
|= PF_MEMALLOC_NOIO
;
2333 static inline void memalloc_noio_restore(unsigned int flags
)
2335 current
->flags
= (current
->flags
& ~PF_MEMALLOC_NOIO
) | flags
;
2338 /* Per-process atomic flags. */
2339 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
2340 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
2341 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
2342 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
2345 #define TASK_PFA_TEST(name, func) \
2346 static inline bool task_##func(struct task_struct *p) \
2347 { return test_bit(PFA_##name, &p->atomic_flags); }
2348 #define TASK_PFA_SET(name, func) \
2349 static inline void task_set_##func(struct task_struct *p) \
2350 { set_bit(PFA_##name, &p->atomic_flags); }
2351 #define TASK_PFA_CLEAR(name, func) \
2352 static inline void task_clear_##func(struct task_struct *p) \
2353 { clear_bit(PFA_##name, &p->atomic_flags); }
2355 TASK_PFA_TEST(NO_NEW_PRIVS
, no_new_privs
)
2356 TASK_PFA_SET(NO_NEW_PRIVS
, no_new_privs
)
2358 TASK_PFA_TEST(SPREAD_PAGE
, spread_page
)
2359 TASK_PFA_SET(SPREAD_PAGE
, spread_page
)
2360 TASK_PFA_CLEAR(SPREAD_PAGE
, spread_page
)
2362 TASK_PFA_TEST(SPREAD_SLAB
, spread_slab
)
2363 TASK_PFA_SET(SPREAD_SLAB
, spread_slab
)
2364 TASK_PFA_CLEAR(SPREAD_SLAB
, spread_slab
)
2366 TASK_PFA_TEST(LMK_WAITING
, lmk_waiting
)
2367 TASK_PFA_SET(LMK_WAITING
, lmk_waiting
)
2370 * task->jobctl flags
2372 #define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
2374 #define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
2375 #define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
2376 #define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
2377 #define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
2378 #define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
2379 #define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
2380 #define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
2382 #define JOBCTL_STOP_DEQUEUED (1UL << JOBCTL_STOP_DEQUEUED_BIT)
2383 #define JOBCTL_STOP_PENDING (1UL << JOBCTL_STOP_PENDING_BIT)
2384 #define JOBCTL_STOP_CONSUME (1UL << JOBCTL_STOP_CONSUME_BIT)
2385 #define JOBCTL_TRAP_STOP (1UL << JOBCTL_TRAP_STOP_BIT)
2386 #define JOBCTL_TRAP_NOTIFY (1UL << JOBCTL_TRAP_NOTIFY_BIT)
2387 #define JOBCTL_TRAPPING (1UL << JOBCTL_TRAPPING_BIT)
2388 #define JOBCTL_LISTENING (1UL << JOBCTL_LISTENING_BIT)
2390 #define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2391 #define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2393 extern bool task_set_jobctl_pending(struct task_struct
*task
,
2394 unsigned long mask
);
2395 extern void task_clear_jobctl_trapping(struct task_struct
*task
);
2396 extern void task_clear_jobctl_pending(struct task_struct
*task
,
2397 unsigned long mask
);
2399 static inline void rcu_copy_process(struct task_struct
*p
)
2401 #ifdef CONFIG_PREEMPT_RCU
2402 p
->rcu_read_lock_nesting
= 0;
2403 p
->rcu_read_unlock_special
.s
= 0;
2404 p
->rcu_blocked_node
= NULL
;
2405 INIT_LIST_HEAD(&p
->rcu_node_entry
);
2406 #endif /* #ifdef CONFIG_PREEMPT_RCU */
2407 #ifdef CONFIG_TASKS_RCU
2408 p
->rcu_tasks_holdout
= false;
2409 INIT_LIST_HEAD(&p
->rcu_tasks_holdout_list
);
2410 p
->rcu_tasks_idle_cpu
= -1;
2411 #endif /* #ifdef CONFIG_TASKS_RCU */
2414 static inline void tsk_restore_flags(struct task_struct
*task
,
2415 unsigned long orig_flags
, unsigned long flags
)
2417 task
->flags
&= ~flags
;
2418 task
->flags
|= orig_flags
& flags
;
2421 extern int cpuset_cpumask_can_shrink(const struct cpumask
*cur
,
2422 const struct cpumask
*trial
);
2423 extern int task_can_attach(struct task_struct
*p
,
2424 const struct cpumask
*cs_cpus_allowed
);
2426 extern void do_set_cpus_allowed(struct task_struct
*p
,
2427 const struct cpumask
*new_mask
);
2429 extern int set_cpus_allowed_ptr(struct task_struct
*p
,
2430 const struct cpumask
*new_mask
);
2432 static inline void do_set_cpus_allowed(struct task_struct
*p
,
2433 const struct cpumask
*new_mask
)
2436 static inline int set_cpus_allowed_ptr(struct task_struct
*p
,
2437 const struct cpumask
*new_mask
)
2439 if (!cpumask_test_cpu(0, new_mask
))
2445 #ifdef CONFIG_NO_HZ_COMMON
2446 void calc_load_enter_idle(void);
2447 void calc_load_exit_idle(void);
2449 static inline void calc_load_enter_idle(void) { }
2450 static inline void calc_load_exit_idle(void) { }
2451 #endif /* CONFIG_NO_HZ_COMMON */
2453 #ifndef cpu_relax_yield
2454 #define cpu_relax_yield() cpu_relax()
2458 * Do not use outside of architecture code which knows its limitations.
2460 * sched_clock() has no promise of monotonicity or bounded drift between
2461 * CPUs, use (which you should not) requires disabling IRQs.
2463 * Please use one of the three interfaces below.
2465 extern unsigned long long notrace
sched_clock(void);
2467 * See the comment in kernel/sched/clock.c
2469 extern u64
running_clock(void);
2470 extern u64
sched_clock_cpu(int cpu
);
2473 extern void sched_clock_init(void);
2475 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2476 static inline void sched_clock_init_late(void)
2480 static inline void sched_clock_tick(void)
2484 static inline void clear_sched_clock_stable(void)
2488 static inline void sched_clock_idle_sleep_event(void)
2492 static inline void sched_clock_idle_wakeup_event(u64 delta_ns
)
2496 static inline u64
cpu_clock(int cpu
)
2498 return sched_clock();
2501 static inline u64
local_clock(void)
2503 return sched_clock();
2506 extern void sched_clock_init_late(void);
2508 * Architectures can set this to 1 if they have specified
2509 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2510 * but then during bootup it turns out that sched_clock()
2511 * is reliable after all:
2513 extern int sched_clock_stable(void);
2514 extern void clear_sched_clock_stable(void);
2516 extern void sched_clock_tick(void);
2517 extern void sched_clock_idle_sleep_event(void);
2518 extern void sched_clock_idle_wakeup_event(u64 delta_ns
);
2521 * As outlined in clock.c, provides a fast, high resolution, nanosecond
2522 * time source that is monotonic per cpu argument and has bounded drift
2525 * ######################### BIG FAT WARNING ##########################
2526 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
2527 * # go backwards !! #
2528 * ####################################################################
2530 static inline u64
cpu_clock(int cpu
)
2532 return sched_clock_cpu(cpu
);
2535 static inline u64
local_clock(void)
2537 return sched_clock_cpu(raw_smp_processor_id());
2541 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
2543 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2544 * The reason for this explicit opt-in is not to have perf penalty with
2545 * slow sched_clocks.
2547 extern void enable_sched_clock_irqtime(void);
2548 extern void disable_sched_clock_irqtime(void);
2550 static inline void enable_sched_clock_irqtime(void) {}
2551 static inline void disable_sched_clock_irqtime(void) {}
2554 extern unsigned long long
2555 task_sched_runtime(struct task_struct
*task
);
2557 /* sched_exec is called by processes performing an exec */
2559 extern void sched_exec(void);
2561 #define sched_exec() {}
2564 extern void sched_clock_idle_sleep_event(void);
2565 extern void sched_clock_idle_wakeup_event(u64 delta_ns
);
2567 #ifdef CONFIG_HOTPLUG_CPU
2568 extern void idle_task_exit(void);
2570 static inline void idle_task_exit(void) {}
2573 #if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2574 extern void wake_up_nohz_cpu(int cpu
);
2576 static inline void wake_up_nohz_cpu(int cpu
) { }
2579 #ifdef CONFIG_NO_HZ_FULL
2580 extern u64
scheduler_tick_max_deferment(void);
2583 #ifdef CONFIG_SCHED_AUTOGROUP
2584 extern void sched_autogroup_create_attach(struct task_struct
*p
);
2585 extern void sched_autogroup_detach(struct task_struct
*p
);
2586 extern void sched_autogroup_fork(struct signal_struct
*sig
);
2587 extern void sched_autogroup_exit(struct signal_struct
*sig
);
2588 extern void sched_autogroup_exit_task(struct task_struct
*p
);
2589 #ifdef CONFIG_PROC_FS
2590 extern void proc_sched_autogroup_show_task(struct task_struct
*p
, struct seq_file
*m
);
2591 extern int proc_sched_autogroup_set_nice(struct task_struct
*p
, int nice
);
2594 static inline void sched_autogroup_create_attach(struct task_struct
*p
) { }
2595 static inline void sched_autogroup_detach(struct task_struct
*p
) { }
2596 static inline void sched_autogroup_fork(struct signal_struct
*sig
) { }
2597 static inline void sched_autogroup_exit(struct signal_struct
*sig
) { }
2598 static inline void sched_autogroup_exit_task(struct task_struct
*p
) { }
2601 extern int yield_to(struct task_struct
*p
, bool preempt
);
2602 extern void set_user_nice(struct task_struct
*p
, long nice
);
2603 extern int task_prio(const struct task_struct
*p
);
2605 * task_nice - return the nice value of a given task.
2606 * @p: the task in question.
2608 * Return: The nice value [ -20 ... 0 ... 19 ].
2610 static inline int task_nice(const struct task_struct
*p
)
2612 return PRIO_TO_NICE((p
)->static_prio
);
2614 extern int can_nice(const struct task_struct
*p
, const int nice
);
2615 extern int task_curr(const struct task_struct
*p
);
2616 extern int idle_cpu(int cpu
);
2617 extern int sched_setscheduler(struct task_struct
*, int,
2618 const struct sched_param
*);
2619 extern int sched_setscheduler_nocheck(struct task_struct
*, int,
2620 const struct sched_param
*);
2621 extern int sched_setattr(struct task_struct
*,
2622 const struct sched_attr
*);
2623 extern struct task_struct
*idle_task(int cpu
);
2625 * is_idle_task - is the specified task an idle task?
2626 * @p: the task in question.
2628 * Return: 1 if @p is an idle task. 0 otherwise.
2630 static inline bool is_idle_task(const struct task_struct
*p
)
2632 return !!(p
->flags
& PF_IDLE
);
2634 extern struct task_struct
*curr_task(int cpu
);
2635 extern void ia64_set_curr_task(int cpu
, struct task_struct
*p
);
2639 union thread_union
{
2640 #ifndef CONFIG_THREAD_INFO_IN_TASK
2641 struct thread_info thread_info
;
2643 unsigned long stack
[THREAD_SIZE
/sizeof(long)];
2646 #ifndef __HAVE_ARCH_KSTACK_END
2647 static inline int kstack_end(void *addr
)
2649 /* Reliable end of stack detection:
2650 * Some APM bios versions misalign the stack
2652 return !(((unsigned long)addr
+sizeof(void*)-1) & (THREAD_SIZE
-sizeof(void*)));
2656 extern union thread_union init_thread_union
;
2657 extern struct task_struct init_task
;
2659 extern struct mm_struct init_mm
;
2661 extern struct pid_namespace init_pid_ns
;
2664 * find a task by one of its numerical ids
2666 * find_task_by_pid_ns():
2667 * finds a task by its pid in the specified namespace
2668 * find_task_by_vpid():
2669 * finds a task by its virtual pid
2671 * see also find_vpid() etc in include/linux/pid.h
2674 extern struct task_struct
*find_task_by_vpid(pid_t nr
);
2675 extern struct task_struct
*find_task_by_pid_ns(pid_t nr
,
2676 struct pid_namespace
*ns
);
2678 /* per-UID process charging. */
2679 extern struct user_struct
* alloc_uid(kuid_t
);
2680 static inline struct user_struct
*get_uid(struct user_struct
*u
)
2682 atomic_inc(&u
->__count
);
2685 extern void free_uid(struct user_struct
*);
2687 #include <asm/current.h>
2689 extern void xtime_update(unsigned long ticks
);
2691 extern int wake_up_state(struct task_struct
*tsk
, unsigned int state
);
2692 extern int wake_up_process(struct task_struct
*tsk
);
2693 extern void wake_up_new_task(struct task_struct
*tsk
);
2695 extern void kick_process(struct task_struct
*tsk
);
2697 static inline void kick_process(struct task_struct
*tsk
) { }
2699 extern int sched_fork(unsigned long clone_flags
, struct task_struct
*p
);
2700 extern void sched_dead(struct task_struct
*p
);
2702 extern void proc_caches_init(void);
2703 extern void flush_signals(struct task_struct
*);
2704 extern void ignore_signals(struct task_struct
*);
2705 extern void flush_signal_handlers(struct task_struct
*, int force_default
);
2706 extern int dequeue_signal(struct task_struct
*tsk
, sigset_t
*mask
, siginfo_t
*info
);
2708 static inline int kernel_dequeue_signal(siginfo_t
*info
)
2710 struct task_struct
*tsk
= current
;
2714 spin_lock_irq(&tsk
->sighand
->siglock
);
2715 ret
= dequeue_signal(tsk
, &tsk
->blocked
, info
?: &__info
);
2716 spin_unlock_irq(&tsk
->sighand
->siglock
);
2721 static inline void kernel_signal_stop(void)
2723 spin_lock_irq(¤t
->sighand
->siglock
);
2724 if (current
->jobctl
& JOBCTL_STOP_DEQUEUED
)
2725 __set_current_state(TASK_STOPPED
);
2726 spin_unlock_irq(¤t
->sighand
->siglock
);
2731 extern void release_task(struct task_struct
* p
);
2732 extern int send_sig_info(int, struct siginfo
*, struct task_struct
*);
2733 extern int force_sigsegv(int, struct task_struct
*);
2734 extern int force_sig_info(int, struct siginfo
*, struct task_struct
*);
2735 extern int __kill_pgrp_info(int sig
, struct siginfo
*info
, struct pid
*pgrp
);
2736 extern int kill_pid_info(int sig
, struct siginfo
*info
, struct pid
*pid
);
2737 extern int kill_pid_info_as_cred(int, struct siginfo
*, struct pid
*,
2738 const struct cred
*, u32
);
2739 extern int kill_pgrp(struct pid
*pid
, int sig
, int priv
);
2740 extern int kill_pid(struct pid
*pid
, int sig
, int priv
);
2741 extern int kill_proc_info(int, struct siginfo
*, pid_t
);
2742 extern __must_check
bool do_notify_parent(struct task_struct
*, int);
2743 extern void __wake_up_parent(struct task_struct
*p
, struct task_struct
*parent
);
2744 extern void force_sig(int, struct task_struct
*);
2745 extern int send_sig(int, struct task_struct
*, int);
2746 extern int zap_other_threads(struct task_struct
*p
);
2747 extern struct sigqueue
*sigqueue_alloc(void);
2748 extern void sigqueue_free(struct sigqueue
*);
2749 extern int send_sigqueue(struct sigqueue
*, struct task_struct
*, int group
);
2750 extern int do_sigaction(int, struct k_sigaction
*, struct k_sigaction
*);
2752 #ifdef TIF_RESTORE_SIGMASK
2754 * Legacy restore_sigmask accessors. These are inefficient on
2755 * SMP architectures because they require atomic operations.
2759 * set_restore_sigmask() - make sure saved_sigmask processing gets done
2761 * This sets TIF_RESTORE_SIGMASK and ensures that the arch signal code
2762 * will run before returning to user mode, to process the flag. For
2763 * all callers, TIF_SIGPENDING is already set or it's no harm to set
2764 * it. TIF_RESTORE_SIGMASK need not be in the set of bits that the
2765 * arch code will notice on return to user mode, in case those bits
2766 * are scarce. We set TIF_SIGPENDING here to ensure that the arch
2767 * signal code always gets run when TIF_RESTORE_SIGMASK is set.
2769 static inline void set_restore_sigmask(void)
2771 set_thread_flag(TIF_RESTORE_SIGMASK
);
2772 WARN_ON(!test_thread_flag(TIF_SIGPENDING
));
2774 static inline void clear_restore_sigmask(void)
2776 clear_thread_flag(TIF_RESTORE_SIGMASK
);
2778 static inline bool test_restore_sigmask(void)
2780 return test_thread_flag(TIF_RESTORE_SIGMASK
);
2782 static inline bool test_and_clear_restore_sigmask(void)
2784 return test_and_clear_thread_flag(TIF_RESTORE_SIGMASK
);
2787 #else /* TIF_RESTORE_SIGMASK */
2789 /* Higher-quality implementation, used if TIF_RESTORE_SIGMASK doesn't exist. */
2790 static inline void set_restore_sigmask(void)
2792 current
->restore_sigmask
= true;
2793 WARN_ON(!test_thread_flag(TIF_SIGPENDING
));
2795 static inline void clear_restore_sigmask(void)
2797 current
->restore_sigmask
= false;
2799 static inline bool test_restore_sigmask(void)
2801 return current
->restore_sigmask
;
2803 static inline bool test_and_clear_restore_sigmask(void)
2805 if (!current
->restore_sigmask
)
2807 current
->restore_sigmask
= false;
2812 static inline void restore_saved_sigmask(void)
2814 if (test_and_clear_restore_sigmask())
2815 __set_current_blocked(¤t
->saved_sigmask
);
2818 static inline sigset_t
*sigmask_to_save(void)
2820 sigset_t
*res
= ¤t
->blocked
;
2821 if (unlikely(test_restore_sigmask()))
2822 res
= ¤t
->saved_sigmask
;
2826 static inline int kill_cad_pid(int sig
, int priv
)
2828 return kill_pid(cad_pid
, sig
, priv
);
2831 /* These can be the second arg to send_sig_info/send_group_sig_info. */
2832 #define SEND_SIG_NOINFO ((struct siginfo *) 0)
2833 #define SEND_SIG_PRIV ((struct siginfo *) 1)
2834 #define SEND_SIG_FORCED ((struct siginfo *) 2)
2837 * True if we are on the alternate signal stack.
2839 static inline int on_sig_stack(unsigned long sp
)
2842 * If the signal stack is SS_AUTODISARM then, by construction, we
2843 * can't be on the signal stack unless user code deliberately set
2844 * SS_AUTODISARM when we were already on it.
2846 * This improves reliability: if user state gets corrupted such that
2847 * the stack pointer points very close to the end of the signal stack,
2848 * then this check will enable the signal to be handled anyway.
2850 if (current
->sas_ss_flags
& SS_AUTODISARM
)
2853 #ifdef CONFIG_STACK_GROWSUP
2854 return sp
>= current
->sas_ss_sp
&&
2855 sp
- current
->sas_ss_sp
< current
->sas_ss_size
;
2857 return sp
> current
->sas_ss_sp
&&
2858 sp
- current
->sas_ss_sp
<= current
->sas_ss_size
;
2862 static inline int sas_ss_flags(unsigned long sp
)
2864 if (!current
->sas_ss_size
)
2867 return on_sig_stack(sp
) ? SS_ONSTACK
: 0;
2870 static inline void sas_ss_reset(struct task_struct
*p
)
2874 p
->sas_ss_flags
= SS_DISABLE
;
2877 static inline unsigned long sigsp(unsigned long sp
, struct ksignal
*ksig
)
2879 if (unlikely((ksig
->ka
.sa
.sa_flags
& SA_ONSTACK
)) && ! sas_ss_flags(sp
))
2880 #ifdef CONFIG_STACK_GROWSUP
2881 return current
->sas_ss_sp
;
2883 return current
->sas_ss_sp
+ current
->sas_ss_size
;
2889 * Routines for handling mm_structs
2891 extern struct mm_struct
* mm_alloc(void);
2894 * mmgrab() - Pin a &struct mm_struct.
2895 * @mm: The &struct mm_struct to pin.
2897 * Make sure that @mm will not get freed even after the owning task
2898 * exits. This doesn't guarantee that the associated address space
2899 * will still exist later on and mmget_not_zero() has to be used before
2902 * This is a preferred way to to pin @mm for a longer/unbounded amount
2905 * Use mmdrop() to release the reference acquired by mmgrab().
2907 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2908 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2910 static inline void mmgrab(struct mm_struct
*mm
)
2912 atomic_inc(&mm
->mm_count
);
2915 /* mmdrop drops the mm and the page tables */
2916 extern void __mmdrop(struct mm_struct
*);
2917 static inline void mmdrop(struct mm_struct
*mm
)
2919 if (unlikely(atomic_dec_and_test(&mm
->mm_count
)))
2923 static inline void mmdrop_async_fn(struct work_struct
*work
)
2925 struct mm_struct
*mm
= container_of(work
, struct mm_struct
, async_put_work
);
2929 static inline void mmdrop_async(struct mm_struct
*mm
)
2931 if (unlikely(atomic_dec_and_test(&mm
->mm_count
))) {
2932 INIT_WORK(&mm
->async_put_work
, mmdrop_async_fn
);
2933 schedule_work(&mm
->async_put_work
);
2938 * mmget() - Pin the address space associated with a &struct mm_struct.
2939 * @mm: The address space to pin.
2941 * Make sure that the address space of the given &struct mm_struct doesn't
2942 * go away. This does not protect against parts of the address space being
2943 * modified or freed, however.
2945 * Never use this function to pin this address space for an
2946 * unbounded/indefinite amount of time.
2948 * Use mmput() to release the reference acquired by mmget().
2950 * See also <Documentation/vm/active_mm.txt> for an in-depth explanation
2951 * of &mm_struct.mm_count vs &mm_struct.mm_users.
2953 static inline void mmget(struct mm_struct
*mm
)
2955 atomic_inc(&mm
->mm_users
);
2958 static inline bool mmget_not_zero(struct mm_struct
*mm
)
2960 return atomic_inc_not_zero(&mm
->mm_users
);
2963 /* mmput gets rid of the mappings and all user-space */
2964 extern void mmput(struct mm_struct
*);
2966 /* same as above but performs the slow path from the async context. Can
2967 * be called from the atomic context as well
2969 extern void mmput_async(struct mm_struct
*);
2972 /* Grab a reference to a task's mm, if it is not already going away */
2973 extern struct mm_struct
*get_task_mm(struct task_struct
*task
);
2975 * Grab a reference to a task's mm, if it is not already going away
2976 * and ptrace_may_access with the mode parameter passed to it
2979 extern struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
);
2980 /* Remove the current tasks stale references to the old mm_struct */
2981 extern void mm_release(struct task_struct
*, struct mm_struct
*);
2983 #ifdef CONFIG_HAVE_COPY_THREAD_TLS
2984 extern int copy_thread_tls(unsigned long, unsigned long, unsigned long,
2985 struct task_struct
*, unsigned long);
2987 extern int copy_thread(unsigned long, unsigned long, unsigned long,
2988 struct task_struct
*);
2990 /* Architectures that haven't opted into copy_thread_tls get the tls argument
2991 * via pt_regs, so ignore the tls argument passed via C. */
2992 static inline int copy_thread_tls(
2993 unsigned long clone_flags
, unsigned long sp
, unsigned long arg
,
2994 struct task_struct
*p
, unsigned long tls
)
2996 return copy_thread(clone_flags
, sp
, arg
, p
);
2999 extern void flush_thread(void);
3001 #ifdef CONFIG_HAVE_EXIT_THREAD
3002 extern void exit_thread(struct task_struct
*tsk
);
3004 static inline void exit_thread(struct task_struct
*tsk
)
3009 extern void exit_files(struct task_struct
*);
3010 extern void __cleanup_sighand(struct sighand_struct
*);
3012 extern void exit_itimers(struct signal_struct
*);
3013 extern void flush_itimer_signals(void);
3015 extern void do_group_exit(int);
3017 extern int do_execve(struct filename
*,
3018 const char __user
* const __user
*,
3019 const char __user
* const __user
*);
3020 extern int do_execveat(int, struct filename
*,
3021 const char __user
* const __user
*,
3022 const char __user
* const __user
*,
3024 extern long _do_fork(unsigned long, unsigned long, unsigned long, int __user
*, int __user
*, unsigned long);
3025 extern long do_fork(unsigned long, unsigned long, unsigned long, int __user
*, int __user
*);
3026 struct task_struct
*fork_idle(int);
3027 extern pid_t
kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
);
3029 extern void __set_task_comm(struct task_struct
*tsk
, const char *from
, bool exec
);
3030 static inline void set_task_comm(struct task_struct
*tsk
, const char *from
)
3032 __set_task_comm(tsk
, from
, false);
3034 extern char *get_task_comm(char *to
, struct task_struct
*tsk
);
3037 void scheduler_ipi(void);
3038 extern unsigned long wait_task_inactive(struct task_struct
*, long match_state
);
3040 static inline void scheduler_ipi(void) { }
3041 static inline unsigned long wait_task_inactive(struct task_struct
*p
,
3048 #define tasklist_empty() \
3049 list_empty(&init_task.tasks)
3051 #define next_task(p) \
3052 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
3054 #define for_each_process(p) \
3055 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
3057 extern bool current_is_single_threaded(void);
3060 * Careful: do_each_thread/while_each_thread is a double loop so
3061 * 'break' will not work as expected - use goto instead.
3063 #define do_each_thread(g, t) \
3064 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
3066 #define while_each_thread(g, t) \
3067 while ((t = next_thread(t)) != g)
3069 #define __for_each_thread(signal, t) \
3070 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
3072 #define for_each_thread(p, t) \
3073 __for_each_thread((p)->signal, t)
3075 /* Careful: this is a double loop, 'break' won't work as expected. */
3076 #define for_each_process_thread(p, t) \
3077 for_each_process(p) for_each_thread(p, t)
3079 typedef int (*proc_visitor
)(struct task_struct
*p
, void *data
);
3080 void walk_process_tree(struct task_struct
*top
, proc_visitor
, void *);
3082 static inline int get_nr_threads(struct task_struct
*tsk
)
3084 return tsk
->signal
->nr_threads
;
3087 static inline bool thread_group_leader(struct task_struct
*p
)
3089 return p
->exit_signal
>= 0;
3092 /* Do to the insanities of de_thread it is possible for a process
3093 * to have the pid of the thread group leader without actually being
3094 * the thread group leader. For iteration through the pids in proc
3095 * all we care about is that we have a task with the appropriate
3096 * pid, we don't actually care if we have the right task.
3098 static inline bool has_group_leader_pid(struct task_struct
*p
)
3100 return task_pid(p
) == p
->signal
->leader_pid
;
3104 bool same_thread_group(struct task_struct
*p1
, struct task_struct
*p2
)
3106 return p1
->signal
== p2
->signal
;
3109 static inline struct task_struct
*next_thread(const struct task_struct
*p
)
3111 return list_entry_rcu(p
->thread_group
.next
,
3112 struct task_struct
, thread_group
);
3115 static inline int thread_group_empty(struct task_struct
*p
)
3117 return list_empty(&p
->thread_group
);
3120 #define delay_group_leader(p) \
3121 (thread_group_leader(p) && !thread_group_empty(p))
3124 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
3125 * subscriptions and synchronises with wait4(). Also used in procfs. Also
3126 * pins the final release of task.io_context. Also protects ->cpuset and
3127 * ->cgroup.subsys[]. And ->vfork_done.
3129 * Nests both inside and outside of read_lock(&tasklist_lock).
3130 * It must not be nested with write_lock_irq(&tasklist_lock),
3131 * neither inside nor outside.
3133 static inline void task_lock(struct task_struct
*p
)
3135 spin_lock(&p
->alloc_lock
);
3138 static inline void task_unlock(struct task_struct
*p
)
3140 spin_unlock(&p
->alloc_lock
);
3143 extern struct sighand_struct
*__lock_task_sighand(struct task_struct
*tsk
,
3144 unsigned long *flags
);
3146 static inline struct sighand_struct
*lock_task_sighand(struct task_struct
*tsk
,
3147 unsigned long *flags
)
3149 struct sighand_struct
*ret
;
3151 ret
= __lock_task_sighand(tsk
, flags
);
3152 (void)__cond_lock(&tsk
->sighand
->siglock
, ret
);
3156 static inline void unlock_task_sighand(struct task_struct
*tsk
,
3157 unsigned long *flags
)
3159 spin_unlock_irqrestore(&tsk
->sighand
->siglock
, *flags
);
3162 #ifdef CONFIG_THREAD_INFO_IN_TASK
3164 static inline struct thread_info
*task_thread_info(struct task_struct
*task
)
3166 return &task
->thread_info
;
3170 * When accessing the stack of a non-current task that might exit, use
3171 * try_get_task_stack() instead. task_stack_page will return a pointer
3172 * that could get freed out from under you.
3174 static inline void *task_stack_page(const struct task_struct
*task
)
3179 #define setup_thread_stack(new,old) do { } while(0)
3181 static inline unsigned long *end_of_stack(const struct task_struct
*task
)
3186 #elif !defined(__HAVE_THREAD_FUNCTIONS)
3188 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
3189 #define task_stack_page(task) ((void *)(task)->stack)
3191 static inline void setup_thread_stack(struct task_struct
*p
, struct task_struct
*org
)
3193 *task_thread_info(p
) = *task_thread_info(org
);
3194 task_thread_info(p
)->task
= p
;
3198 * Return the address of the last usable long on the stack.
3200 * When the stack grows down, this is just above the thread
3201 * info struct. Going any lower will corrupt the threadinfo.
3203 * When the stack grows up, this is the highest address.
3204 * Beyond that position, we corrupt data on the next page.
3206 static inline unsigned long *end_of_stack(struct task_struct
*p
)
3208 #ifdef CONFIG_STACK_GROWSUP
3209 return (unsigned long *)((unsigned long)task_thread_info(p
) + THREAD_SIZE
) - 1;
3211 return (unsigned long *)(task_thread_info(p
) + 1);
3217 #ifdef CONFIG_THREAD_INFO_IN_TASK
3218 static inline void *try_get_task_stack(struct task_struct
*tsk
)
3220 return atomic_inc_not_zero(&tsk
->stack_refcount
) ?
3221 task_stack_page(tsk
) : NULL
;
3224 extern void put_task_stack(struct task_struct
*tsk
);
3226 static inline void *try_get_task_stack(struct task_struct
*tsk
)
3228 return task_stack_page(tsk
);
3231 static inline void put_task_stack(struct task_struct
*tsk
) {}
3234 #define task_stack_end_corrupted(task) \
3235 (*(end_of_stack(task)) != STACK_END_MAGIC)
3237 static inline int object_is_on_stack(void *obj
)
3239 void *stack
= task_stack_page(current
);
3241 return (obj
>= stack
) && (obj
< (stack
+ THREAD_SIZE
));
3244 extern void thread_stack_cache_init(void);
3246 #ifdef CONFIG_DEBUG_STACK_USAGE
3247 static inline unsigned long stack_not_used(struct task_struct
*p
)
3249 unsigned long *n
= end_of_stack(p
);
3251 do { /* Skip over canary */
3252 # ifdef CONFIG_STACK_GROWSUP
3259 # ifdef CONFIG_STACK_GROWSUP
3260 return (unsigned long)end_of_stack(p
) - (unsigned long)n
;
3262 return (unsigned long)n
- (unsigned long)end_of_stack(p
);
3266 extern void set_task_stack_end_magic(struct task_struct
*tsk
);
3268 /* set thread flags in other task's structures
3269 * - see asm/thread_info.h for TIF_xxxx flags available
3271 static inline void set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
3273 set_ti_thread_flag(task_thread_info(tsk
), flag
);
3276 static inline void clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
3278 clear_ti_thread_flag(task_thread_info(tsk
), flag
);
3281 static inline int test_and_set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
3283 return test_and_set_ti_thread_flag(task_thread_info(tsk
), flag
);
3286 static inline int test_and_clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
3288 return test_and_clear_ti_thread_flag(task_thread_info(tsk
), flag
);
3291 static inline int test_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
3293 return test_ti_thread_flag(task_thread_info(tsk
), flag
);
3296 static inline void set_tsk_need_resched(struct task_struct
*tsk
)
3298 set_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
3301 static inline void clear_tsk_need_resched(struct task_struct
*tsk
)
3303 clear_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
3306 static inline int test_tsk_need_resched(struct task_struct
*tsk
)
3308 return unlikely(test_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
));
3311 static inline int restart_syscall(void)
3313 set_tsk_thread_flag(current
, TIF_SIGPENDING
);
3314 return -ERESTARTNOINTR
;
3317 static inline int signal_pending(struct task_struct
*p
)
3319 return unlikely(test_tsk_thread_flag(p
,TIF_SIGPENDING
));
3322 static inline int __fatal_signal_pending(struct task_struct
*p
)
3324 return unlikely(sigismember(&p
->pending
.signal
, SIGKILL
));
3327 static inline int fatal_signal_pending(struct task_struct
*p
)
3329 return signal_pending(p
) && __fatal_signal_pending(p
);
3332 static inline int signal_pending_state(long state
, struct task_struct
*p
)
3334 if (!(state
& (TASK_INTERRUPTIBLE
| TASK_WAKEKILL
)))
3336 if (!signal_pending(p
))
3339 return (state
& TASK_INTERRUPTIBLE
) || __fatal_signal_pending(p
);
3343 * cond_resched() and cond_resched_lock(): latency reduction via
3344 * explicit rescheduling in places that are safe. The return
3345 * value indicates whether a reschedule was done in fact.
3346 * cond_resched_lock() will drop the spinlock before scheduling,
3347 * cond_resched_softirq() will enable bhs before scheduling.
3349 #ifndef CONFIG_PREEMPT
3350 extern int _cond_resched(void);
3352 static inline int _cond_resched(void) { return 0; }
3355 #define cond_resched() ({ \
3356 ___might_sleep(__FILE__, __LINE__, 0); \
3360 extern int __cond_resched_lock(spinlock_t
*lock
);
3362 #define cond_resched_lock(lock) ({ \
3363 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
3364 __cond_resched_lock(lock); \
3367 extern int __cond_resched_softirq(void);
3369 #define cond_resched_softirq() ({ \
3370 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
3371 __cond_resched_softirq(); \
3374 static inline void cond_resched_rcu(void)
3376 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
3384 * Does a critical section need to be broken due to another
3385 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
3386 * but a general need for low latency)
3388 static inline int spin_needbreak(spinlock_t
*lock
)
3390 #ifdef CONFIG_PREEMPT
3391 return spin_is_contended(lock
);
3398 * Idle thread specific functions to determine the need_resched
3401 #ifdef TIF_POLLING_NRFLAG
3402 static inline int tsk_is_polling(struct task_struct
*p
)
3404 return test_tsk_thread_flag(p
, TIF_POLLING_NRFLAG
);
3407 static inline void __current_set_polling(void)
3409 set_thread_flag(TIF_POLLING_NRFLAG
);
3412 static inline bool __must_check
current_set_polling_and_test(void)
3414 __current_set_polling();
3417 * Polling state must be visible before we test NEED_RESCHED,
3418 * paired by resched_curr()
3420 smp_mb__after_atomic();
3422 return unlikely(tif_need_resched());
3425 static inline void __current_clr_polling(void)
3427 clear_thread_flag(TIF_POLLING_NRFLAG
);
3430 static inline bool __must_check
current_clr_polling_and_test(void)
3432 __current_clr_polling();
3435 * Polling state must be visible before we test NEED_RESCHED,
3436 * paired by resched_curr()
3438 smp_mb__after_atomic();
3440 return unlikely(tif_need_resched());
3444 static inline int tsk_is_polling(struct task_struct
*p
) { return 0; }
3445 static inline void __current_set_polling(void) { }
3446 static inline void __current_clr_polling(void) { }
3448 static inline bool __must_check
current_set_polling_and_test(void)
3450 return unlikely(tif_need_resched());
3452 static inline bool __must_check
current_clr_polling_and_test(void)
3454 return unlikely(tif_need_resched());
3458 static inline void current_clr_polling(void)
3460 __current_clr_polling();
3463 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
3464 * Once the bit is cleared, we'll get IPIs with every new
3465 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
3468 smp_mb(); /* paired with resched_curr() */
3470 preempt_fold_need_resched();
3473 static __always_inline
bool need_resched(void)
3475 return unlikely(tif_need_resched());
3479 * Thread group CPU time accounting.
3481 void thread_group_cputime(struct task_struct
*tsk
, struct task_cputime
*times
);
3482 void thread_group_cputimer(struct task_struct
*tsk
, struct task_cputime
*times
);
3485 * Reevaluate whether the task has signals pending delivery.
3486 * Wake the task if so.
3487 * This is required every time the blocked sigset_t changes.
3488 * callers must hold sighand->siglock.
3490 extern void recalc_sigpending_and_wake(struct task_struct
*t
);
3491 extern void recalc_sigpending(void);
3493 extern void signal_wake_up_state(struct task_struct
*t
, unsigned int state
);
3495 static inline void signal_wake_up(struct task_struct
*t
, bool resume
)
3497 signal_wake_up_state(t
, resume
? TASK_WAKEKILL
: 0);
3499 static inline void ptrace_signal_wake_up(struct task_struct
*t
, bool resume
)
3501 signal_wake_up_state(t
, resume
? __TASK_TRACED
: 0);
3505 * Wrappers for p->thread_info->cpu access. No-op on UP.
3509 static inline unsigned int task_cpu(const struct task_struct
*p
)
3511 #ifdef CONFIG_THREAD_INFO_IN_TASK
3514 return task_thread_info(p
)->cpu
;
3518 static inline int task_node(const struct task_struct
*p
)
3520 return cpu_to_node(task_cpu(p
));
3523 extern void set_task_cpu(struct task_struct
*p
, unsigned int cpu
);
3527 static inline unsigned int task_cpu(const struct task_struct
*p
)
3532 static inline void set_task_cpu(struct task_struct
*p
, unsigned int cpu
)
3536 #endif /* CONFIG_SMP */
3539 * In order to reduce various lock holder preemption latencies provide an
3540 * interface to see if a vCPU is currently running or not.
3542 * This allows us to terminate optimistic spin loops and block, analogous to
3543 * the native optimistic spin heuristic of testing if the lock owner task is
3546 #ifndef vcpu_is_preempted
3547 # define vcpu_is_preempted(cpu) false
3550 extern long sched_setaffinity(pid_t pid
, const struct cpumask
*new_mask
);
3551 extern long sched_getaffinity(pid_t pid
, struct cpumask
*mask
);
3553 #ifdef CONFIG_CGROUP_SCHED
3554 extern struct task_group root_task_group
;
3555 #endif /* CONFIG_CGROUP_SCHED */
3557 extern int task_can_switch_user(struct user_struct
*up
,
3558 struct task_struct
*tsk
);
3560 #ifdef CONFIG_TASK_XACCT
3561 static inline void add_rchar(struct task_struct
*tsk
, ssize_t amt
)
3563 tsk
->ioac
.rchar
+= amt
;
3566 static inline void add_wchar(struct task_struct
*tsk
, ssize_t amt
)
3568 tsk
->ioac
.wchar
+= amt
;
3571 static inline void inc_syscr(struct task_struct
*tsk
)
3576 static inline void inc_syscw(struct task_struct
*tsk
)
3581 static inline void add_rchar(struct task_struct
*tsk
, ssize_t amt
)
3585 static inline void add_wchar(struct task_struct
*tsk
, ssize_t amt
)
3589 static inline void inc_syscr(struct task_struct
*tsk
)
3593 static inline void inc_syscw(struct task_struct
*tsk
)
3598 #ifndef TASK_SIZE_OF
3599 #define TASK_SIZE_OF(tsk) TASK_SIZE
3603 extern void mm_update_next_owner(struct mm_struct
*mm
);
3605 static inline void mm_update_next_owner(struct mm_struct
*mm
)
3608 #endif /* CONFIG_MEMCG */
3610 static inline unsigned long task_rlimit(const struct task_struct
*tsk
,
3613 return READ_ONCE(tsk
->signal
->rlim
[limit
].rlim_cur
);
3616 static inline unsigned long task_rlimit_max(const struct task_struct
*tsk
,
3619 return READ_ONCE(tsk
->signal
->rlim
[limit
].rlim_max
);
3622 static inline unsigned long rlimit(unsigned int limit
)
3624 return task_rlimit(current
, limit
);
3627 static inline unsigned long rlimit_max(unsigned int limit
)
3629 return task_rlimit_max(current
, limit
);
3632 #define SCHED_CPUFREQ_RT (1U << 0)
3633 #define SCHED_CPUFREQ_DL (1U << 1)
3634 #define SCHED_CPUFREQ_IOWAIT (1U << 2)
3636 #define SCHED_CPUFREQ_RT_DL (SCHED_CPUFREQ_RT | SCHED_CPUFREQ_DL)
3638 #ifdef CONFIG_CPU_FREQ
3639 struct update_util_data
{
3640 void (*func
)(struct update_util_data
*data
, u64 time
, unsigned int flags
);
3643 void cpufreq_add_update_util_hook(int cpu
, struct update_util_data
*data
,
3644 void (*func
)(struct update_util_data
*data
, u64 time
,
3645 unsigned int flags
));
3646 void cpufreq_remove_update_util_hook(int cpu
);
3647 #endif /* CONFIG_CPU_FREQ */