4 #include <uapi/linux/sched.h>
6 #include <linux/sched/prio.h>
8 #include <linux/capability.h>
9 #include <linux/mutex.h>
10 #include <linux/plist.h>
11 #include <linux/mm_types.h>
12 #include <asm/ptrace.h>
14 #include <linux/sem.h>
15 #include <linux/shm.h>
16 #include <linux/signal.h>
17 #include <linux/signal_types.h>
18 #include <linux/pid.h>
19 #include <linux/seccomp.h>
20 #include <linux/rculist.h>
21 #include <linux/rtmutex.h>
23 #include <linux/resource.h>
24 #include <linux/hrtimer.h>
25 #include <linux/kcov.h>
26 #include <linux/task_io_accounting.h>
27 #include <linux/latencytop.h>
28 #include <linux/cred.h>
29 #include <linux/gfp.h>
30 #include <linux/topology.h>
31 #include <linux/magic.h>
32 #include <linux/cgroup-defs.h>
34 #include <asm/current.h>
36 /* task_struct member predeclarations: */
39 struct backing_dev_info;
45 struct futex_pi_state;
50 struct perf_event_context;
52 struct pipe_inode_info;
55 struct robust_list_head;
59 struct sighand_struct;
61 struct task_delay_info;
67 * Task state bitmask. NOTE! These bits are also
68 * encoded in fs/proc/array.c: get_task_state().
70 * We have two separate sets of flags: task->state
71 * is about runnability, while task->exit_state are
72 * about the task exiting. Confusing, but this way
73 * modifying one set can't modify the other one by
76 #define TASK_RUNNING 0
77 #define TASK_INTERRUPTIBLE 1
78 #define TASK_UNINTERRUPTIBLE 2
79 #define __TASK_STOPPED 4
80 #define __TASK_TRACED 8
81 /* in tsk->exit_state */
83 #define EXIT_ZOMBIE 32
84 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
85 /* in tsk->state again */
87 #define TASK_WAKEKILL 128
88 #define TASK_WAKING 256
89 #define TASK_PARKED 512
90 #define TASK_NOLOAD 1024
92 #define TASK_STATE_MAX 4096
94 #define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
96 /* Convenience macros for the sake of set_current_state */
97 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
98 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
99 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
101 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
103 /* Convenience macros for the sake of wake_up */
104 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
105 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
107 /* get_task_state() */
108 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
109 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
110 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
112 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
113 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
114 #define task_is_stopped_or_traced(task) \
115 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
116 #define task_contributes_to_load(task) \
117 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
118 (task->flags & PF_FROZEN) == 0 && \
119 (task->state & TASK_NOLOAD) == 0)
121 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
123 #define __set_current_state(state_value) \
125 current->task_state_change = _THIS_IP_; \
126 current->state = (state_value); \
128 #define set_current_state(state_value) \
130 current->task_state_change = _THIS_IP_; \
131 smp_store_mb(current->state, (state_value)); \
136 * set_current_state() includes a barrier so that the write of current->state
137 * is correctly serialised wrt the caller's subsequent test of whether to
141 * set_current_state(TASK_UNINTERRUPTIBLE);
147 * __set_current_state(TASK_RUNNING);
149 * If the caller does not need such serialisation (because, for instance, the
150 * condition test and condition change and wakeup are under the same lock) then
151 * use __set_current_state().
153 * The above is typically ordered against the wakeup, which does:
155 * need_sleep = false;
156 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
158 * Where wake_up_state() (and all other wakeup primitives) imply enough
159 * barriers to order the store of the variable against wakeup.
161 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
162 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
163 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
165 * This is obviously fine, since they both store the exact same value.
167 * Also see the comments of try_to_wake_up().
169 #define __set_current_state(state_value) \
170 do { current->state = (state_value); } while (0)
171 #define set_current_state(state_value) \
172 smp_store_mb(current->state, (state_value))
176 /* Task command name length */
177 #define TASK_COMM_LEN 16
179 extern void sched_init(void);
180 extern void sched_init_smp(void);
182 extern cpumask_var_t cpu_isolated_map;
184 extern int runqueue_is_locked(int cpu);
186 extern void cpu_init (void);
187 extern void trap_init(void);
188 extern void update_process_times(int user);
189 extern void scheduler_tick(void);
191 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
192 extern signed long schedule_timeout(signed long timeout);
193 extern signed long schedule_timeout_interruptible(signed long timeout);
194 extern signed long schedule_timeout_killable(signed long timeout);
195 extern signed long schedule_timeout_uninterruptible(signed long timeout);
196 extern signed long schedule_timeout_idle(signed long timeout);
197 asmlinkage void schedule(void);
198 extern void schedule_preempt_disabled(void);
200 extern int __must_check io_schedule_prepare(void);
201 extern void io_schedule_finish(int token);
202 extern long io_schedule_timeout(long timeout);
203 extern void io_schedule(void);
206 * struct prev_cputime - snaphsot of system and user cputime
207 * @utime: time spent in user mode
208 * @stime: time spent in system mode
209 * @lock: protects the above two fields
211 * Stores previous user/system time values such that we can guarantee
214 struct prev_cputime {
215 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
222 static inline void prev_cputime_init(struct prev_cputime *prev)
224 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
225 prev->utime = prev->stime = 0;
226 raw_spin_lock_init(&prev->lock);
231 * struct task_cputime - collected CPU time counts
232 * @utime: time spent in user mode, in nanoseconds
233 * @stime: time spent in kernel mode, in nanoseconds
234 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
236 * This structure groups together three kinds of CPU time that are tracked for
237 * threads and thread groups. Most things considering CPU time want to group
238 * these counts together and treat all three of them in parallel.
240 struct task_cputime {
243 unsigned long long sum_exec_runtime;
246 /* Alternate field names when used to cache expirations. */
247 #define virt_exp utime
248 #define prof_exp stime
249 #define sched_exp sum_exec_runtime
252 * This is the atomic variant of task_cputime, which can be used for
253 * storing and updating task_cputime statistics without locking.
255 struct task_cputime_atomic {
258 atomic64_t sum_exec_runtime;
261 #define INIT_CPUTIME_ATOMIC \
262 (struct task_cputime_atomic) { \
263 .utime = ATOMIC64_INIT(0), \
264 .stime = ATOMIC64_INIT(0), \
265 .sum_exec_runtime = ATOMIC64_INIT(0), \
268 #define PREEMPT_DISABLED (PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
271 * Disable preemption until the scheduler is running -- use an unconditional
272 * value so that it also works on !PREEMPT_COUNT kernels.
274 * Reset by start_kernel()->sched_init()->init_idle()->init_idle_preempt_count().
276 #define INIT_PREEMPT_COUNT PREEMPT_OFFSET
279 * Initial preempt_count value; reflects the preempt_count schedule invariant
280 * which states that during context switches:
282 * preempt_count() == 2*PREEMPT_DISABLE_OFFSET
284 * Note: PREEMPT_DISABLE_OFFSET is 0 for !PREEMPT_COUNT kernels.
285 * Note: See finish_task_switch().
287 #define FORK_PREEMPT_COUNT (2*PREEMPT_DISABLE_OFFSET + PREEMPT_ENABLED)
290 * struct thread_group_cputimer - thread group interval timer counts
291 * @cputime_atomic: atomic thread group interval timers.
292 * @running: true when there are timers running and
293 * @cputime_atomic receives updates.
294 * @checking_timer: true when a thread in the group is in the
295 * process of checking for thread group timers.
297 * This structure contains the version of task_cputime, above, that is
298 * used for thread group CPU timer calculations.
300 struct thread_group_cputimer {
301 struct task_cputime_atomic cputime_atomic;
306 #include <linux/rwsem.h>
308 #ifdef CONFIG_SCHED_INFO
310 /* cumulative counters */
311 unsigned long pcount; /* # of times run on this cpu */
312 unsigned long long run_delay; /* time spent waiting on a runqueue */
315 unsigned long long last_arrival,/* when we last ran on a cpu */
316 last_queued; /* when we were last queued to run */
318 #endif /* CONFIG_SCHED_INFO */
320 static inline int sched_info_on(void)
322 #ifdef CONFIG_SCHEDSTATS
324 #elif defined(CONFIG_TASK_DELAY_ACCT)
325 extern int delayacct_on;
332 #ifdef CONFIG_SCHEDSTATS
333 void force_schedstat_enabled(void);
337 * Integer metrics need fixed point arithmetic, e.g., sched/fair
338 * has a few: load, load_avg, util_avg, freq, and capacity.
340 * We define a basic fixed point arithmetic range, and then formalize
341 * all these metrics based on that basic range.
343 # define SCHED_FIXEDPOINT_SHIFT 10
344 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
346 #ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
347 extern void prefetch_stack(struct task_struct *t);
349 static inline void prefetch_stack(struct task_struct *t) { }
353 unsigned long weight;
358 * The load_avg/util_avg accumulates an infinite geometric series
359 * (see __update_load_avg() in kernel/sched/fair.c).
361 * [load_avg definition]
363 * load_avg = runnable% * scale_load_down(load)
365 * where runnable% is the time ratio that a sched_entity is runnable.
366 * For cfs_rq, it is the aggregated load_avg of all runnable and
367 * blocked sched_entities.
369 * load_avg may also take frequency scaling into account:
371 * load_avg = runnable% * scale_load_down(load) * freq%
373 * where freq% is the CPU frequency normalized to the highest frequency.
375 * [util_avg definition]
377 * util_avg = running% * SCHED_CAPACITY_SCALE
379 * where running% is the time ratio that a sched_entity is running on
380 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
381 * and blocked sched_entities.
383 * util_avg may also factor frequency scaling and CPU capacity scaling:
385 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
387 * where freq% is the same as above, and capacity% is the CPU capacity
388 * normalized to the greatest capacity (due to uarch differences, etc).
390 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
391 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
392 * we therefore scale them to as large a range as necessary. This is for
393 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
397 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
398 * with the highest load (=88761), always runnable on a single cfs_rq,
399 * and should not overflow as the number already hits PID_MAX_LIMIT.
401 * For all other cases (including 32-bit kernels), struct load_weight's
402 * weight will overflow first before we do, because:
404 * Max(load_avg) <= Max(load.weight)
406 * Then it is the load_weight's responsibility to consider overflow
410 u64 last_update_time, load_sum;
411 u32 util_sum, period_contrib;
412 unsigned long load_avg, util_avg;
415 #ifdef CONFIG_SCHEDSTATS
416 struct sched_statistics {
426 s64 sum_sleep_runtime;
433 u64 nr_migrations_cold;
434 u64 nr_failed_migrations_affine;
435 u64 nr_failed_migrations_running;
436 u64 nr_failed_migrations_hot;
437 u64 nr_forced_migrations;
441 u64 nr_wakeups_migrate;
442 u64 nr_wakeups_local;
443 u64 nr_wakeups_remote;
444 u64 nr_wakeups_affine;
445 u64 nr_wakeups_affine_attempts;
446 u64 nr_wakeups_passive;
451 struct sched_entity {
452 struct load_weight load; /* for load-balancing */
453 struct rb_node run_node;
454 struct list_head group_node;
458 u64 sum_exec_runtime;
460 u64 prev_sum_exec_runtime;
464 #ifdef CONFIG_SCHEDSTATS
465 struct sched_statistics statistics;
468 #ifdef CONFIG_FAIR_GROUP_SCHED
470 struct sched_entity *parent;
471 /* rq on which this entity is (to be) queued: */
472 struct cfs_rq *cfs_rq;
473 /* rq "owned" by this entity/group: */
479 * Per entity load average tracking.
481 * Put into separate cache line so it does not
482 * collide with read-mostly values above.
484 struct sched_avg avg ____cacheline_aligned_in_smp;
488 struct sched_rt_entity {
489 struct list_head run_list;
490 unsigned long timeout;
491 unsigned long watchdog_stamp;
492 unsigned int time_slice;
493 unsigned short on_rq;
494 unsigned short on_list;
496 struct sched_rt_entity *back;
497 #ifdef CONFIG_RT_GROUP_SCHED
498 struct sched_rt_entity *parent;
499 /* rq on which this entity is (to be) queued: */
501 /* rq "owned" by this entity/group: */
506 struct sched_dl_entity {
507 struct rb_node rb_node;
510 * Original scheduling parameters. Copied here from sched_attr
511 * during sched_setattr(), they will remain the same until
512 * the next sched_setattr().
514 u64 dl_runtime; /* maximum runtime for each instance */
515 u64 dl_deadline; /* relative deadline of each instance */
516 u64 dl_period; /* separation of two instances (period) */
517 u64 dl_bw; /* dl_runtime / dl_deadline */
520 * Actual scheduling parameters. Initialized with the values above,
521 * they are continously updated during task execution. Note that
522 * the remaining runtime could be < 0 in case we are in overrun.
524 s64 runtime; /* remaining runtime for this instance */
525 u64 deadline; /* absolute deadline for this instance */
526 unsigned int flags; /* specifying the scheduler behaviour */
531 * @dl_throttled tells if we exhausted the runtime. If so, the
532 * task has to wait for a replenishment to be performed at the
533 * next firing of dl_timer.
535 * @dl_boosted tells if we are boosted due to DI. If so we are
536 * outside bandwidth enforcement mechanism (but only until we
537 * exit the critical section);
539 * @dl_yielded tells if task gave up the cpu before consuming
540 * all its available runtime during the last job.
542 int dl_throttled, dl_boosted, dl_yielded;
545 * Bandwidth enforcement timer. Each -deadline task has its
546 * own bandwidth to be enforced, thus we need one timer per task.
548 struct hrtimer dl_timer;
556 u8 pad; /* Otherwise the compiler can store garbage here. */
558 u32 s; /* Set of bits. */
561 enum perf_event_task_context {
562 perf_invalid_context = -1,
565 perf_nr_task_contexts,
569 struct wake_q_node *next;
572 /* Track pages that require TLB flushes */
573 struct tlbflush_unmap_batch {
575 * Each bit set is a CPU that potentially has a TLB entry for one of
576 * the PFNs being flushed. See set_tlb_ubc_flush_pending().
578 struct cpumask cpumask;
580 /* True if any bit in cpumask is set */
584 * If true then the PTE was dirty when unmapped. The entry must be
585 * flushed before IO is initiated or a stale TLB entry potentially
586 * allows an update without redirtying the page.
592 #ifdef CONFIG_THREAD_INFO_IN_TASK
594 * For reasons of header soup (see current_thread_info()), this
595 * must be the first element of task_struct.
597 struct thread_info thread_info;
599 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
602 unsigned int flags; /* per process flags, defined below */
606 struct llist_node wake_entry;
608 #ifdef CONFIG_THREAD_INFO_IN_TASK
609 unsigned int cpu; /* current CPU */
611 unsigned int wakee_flips;
612 unsigned long wakee_flip_decay_ts;
613 struct task_struct *last_wakee;
619 int prio, static_prio, normal_prio;
620 unsigned int rt_priority;
621 const struct sched_class *sched_class;
622 struct sched_entity se;
623 struct sched_rt_entity rt;
624 #ifdef CONFIG_CGROUP_SCHED
625 struct task_group *sched_task_group;
627 struct sched_dl_entity dl;
629 #ifdef CONFIG_PREEMPT_NOTIFIERS
630 /* list of struct preempt_notifier: */
631 struct hlist_head preempt_notifiers;
634 #ifdef CONFIG_BLK_DEV_IO_TRACE
635 unsigned int btrace_seq;
640 cpumask_t cpus_allowed;
642 #ifdef CONFIG_PREEMPT_RCU
643 int rcu_read_lock_nesting;
644 union rcu_special rcu_read_unlock_special;
645 struct list_head rcu_node_entry;
646 struct rcu_node *rcu_blocked_node;
647 #endif /* #ifdef CONFIG_PREEMPT_RCU */
648 #ifdef CONFIG_TASKS_RCU
649 unsigned long rcu_tasks_nvcsw;
650 bool rcu_tasks_holdout;
651 struct list_head rcu_tasks_holdout_list;
652 int rcu_tasks_idle_cpu;
653 #endif /* #ifdef CONFIG_TASKS_RCU */
655 #ifdef CONFIG_SCHED_INFO
656 struct sched_info sched_info;
659 struct list_head tasks;
661 struct plist_node pushable_tasks;
662 struct rb_node pushable_dl_tasks;
665 struct mm_struct *mm, *active_mm;
667 /* Per-thread vma caching: */
668 struct vmacache vmacache;
670 #if defined(SPLIT_RSS_COUNTING)
671 struct task_rss_stat rss_stat;
675 int exit_code, exit_signal;
676 int pdeath_signal; /* The signal sent when the parent dies */
677 unsigned long jobctl; /* JOBCTL_*, siglock protected */
679 /* Used for emulating ABI behavior of previous Linux versions */
680 unsigned int personality;
682 /* scheduler bits, serialized by scheduler locks */
683 unsigned sched_reset_on_fork:1;
684 unsigned sched_contributes_to_load:1;
685 unsigned sched_migrated:1;
686 unsigned sched_remote_wakeup:1;
687 unsigned :0; /* force alignment to the next boundary */
689 /* unserialized, strictly 'current' */
690 unsigned in_execve:1; /* bit to tell LSMs we're in execve */
691 unsigned in_iowait:1;
692 #if !defined(TIF_RESTORE_SIGMASK)
693 unsigned restore_sigmask:1;
696 unsigned memcg_may_oom:1;
698 unsigned memcg_kmem_skip_account:1;
701 #ifdef CONFIG_COMPAT_BRK
702 unsigned brk_randomized:1;
705 unsigned long atomic_flags; /* Flags needing atomic access. */
707 struct restart_block restart_block;
712 #ifdef CONFIG_CC_STACKPROTECTOR
713 /* Canary value for the -fstack-protector gcc feature */
714 unsigned long stack_canary;
717 * pointers to (original) parent process, youngest child, younger sibling,
718 * older sibling, respectively. (p->father can be replaced with
719 * p->real_parent->pid)
721 struct task_struct __rcu *real_parent; /* real parent process */
722 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
724 * children/sibling forms the list of my natural children
726 struct list_head children; /* list of my children */
727 struct list_head sibling; /* linkage in my parent's children list */
728 struct task_struct *group_leader; /* threadgroup leader */
731 * ptraced is the list of tasks this task is using ptrace on.
732 * This includes both natural children and PTRACE_ATTACH targets.
733 * p->ptrace_entry is p's link on the p->parent->ptraced list.
735 struct list_head ptraced;
736 struct list_head ptrace_entry;
738 /* PID/PID hash table linkage. */
739 struct pid_link pids[PIDTYPE_MAX];
740 struct list_head thread_group;
741 struct list_head thread_node;
743 struct completion *vfork_done; /* for vfork() */
744 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
745 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
748 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
749 u64 utimescaled, stimescaled;
752 struct prev_cputime prev_cputime;
753 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
754 seqcount_t vtime_seqcount;
755 unsigned long long vtime_snap;
757 /* Task is sleeping or running in a CPU with VTIME inactive */
759 /* Task runs in userspace in a CPU with VTIME active */
761 /* Task runs in kernelspace in a CPU with VTIME active */
766 #ifdef CONFIG_NO_HZ_FULL
767 atomic_t tick_dep_mask;
769 unsigned long nvcsw, nivcsw; /* context switch counts */
770 u64 start_time; /* monotonic time in nsec */
771 u64 real_start_time; /* boot based time in nsec */
772 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
773 unsigned long min_flt, maj_flt;
775 #ifdef CONFIG_POSIX_TIMERS
776 struct task_cputime cputime_expires;
777 struct list_head cpu_timers[3];
780 /* process credentials */
781 const struct cred __rcu *ptracer_cred; /* Tracer's credentials at attach */
782 const struct cred __rcu *real_cred; /* objective and real subjective task
783 * credentials (COW) */
784 const struct cred __rcu *cred; /* effective (overridable) subjective task
785 * credentials (COW) */
786 char comm[TASK_COMM_LEN]; /* executable name excluding path
787 - access with [gs]et_task_comm (which lock
789 - initialized normally by setup_new_exec */
790 /* file system info */
791 struct nameidata *nameidata;
792 #ifdef CONFIG_SYSVIPC
794 struct sysv_sem sysvsem;
795 struct sysv_shm sysvshm;
797 #ifdef CONFIG_DETECT_HUNG_TASK
798 /* hung task detection */
799 unsigned long last_switch_count;
801 /* filesystem information */
802 struct fs_struct *fs;
803 /* open file information */
804 struct files_struct *files;
806 struct nsproxy *nsproxy;
807 /* signal handlers */
808 struct signal_struct *signal;
809 struct sighand_struct *sighand;
811 sigset_t blocked, real_blocked;
812 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
813 struct sigpending pending;
815 unsigned long sas_ss_sp;
817 unsigned sas_ss_flags;
819 struct callback_head *task_works;
821 struct audit_context *audit_context;
822 #ifdef CONFIG_AUDITSYSCALL
824 unsigned int sessionid;
826 struct seccomp seccomp;
828 /* Thread group tracking */
831 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
833 spinlock_t alloc_lock;
835 /* Protection of the PI data structures: */
836 raw_spinlock_t pi_lock;
838 struct wake_q_node wake_q;
840 #ifdef CONFIG_RT_MUTEXES
841 /* PI waiters blocked on a rt_mutex held by this task */
842 struct rb_root pi_waiters;
843 struct rb_node *pi_waiters_leftmost;
844 /* Deadlock detection and priority inheritance handling */
845 struct rt_mutex_waiter *pi_blocked_on;
848 #ifdef CONFIG_DEBUG_MUTEXES
849 /* mutex deadlock detection */
850 struct mutex_waiter *blocked_on;
852 #ifdef CONFIG_TRACE_IRQFLAGS
853 unsigned int irq_events;
854 unsigned long hardirq_enable_ip;
855 unsigned long hardirq_disable_ip;
856 unsigned int hardirq_enable_event;
857 unsigned int hardirq_disable_event;
858 int hardirqs_enabled;
860 unsigned long softirq_disable_ip;
861 unsigned long softirq_enable_ip;
862 unsigned int softirq_disable_event;
863 unsigned int softirq_enable_event;
864 int softirqs_enabled;
867 #ifdef CONFIG_LOCKDEP
868 # define MAX_LOCK_DEPTH 48UL
871 unsigned int lockdep_recursion;
872 struct held_lock held_locks[MAX_LOCK_DEPTH];
873 gfp_t lockdep_reclaim_gfp;
876 unsigned int in_ubsan;
879 /* journalling filesystem info */
882 /* stacked block device info */
883 struct bio_list *bio_list;
887 struct blk_plug *plug;
891 struct reclaim_state *reclaim_state;
893 struct backing_dev_info *backing_dev_info;
895 struct io_context *io_context;
897 unsigned long ptrace_message;
898 siginfo_t *last_siginfo; /* For ptrace use. */
899 struct task_io_accounting ioac;
900 #if defined(CONFIG_TASK_XACCT)
901 u64 acct_rss_mem1; /* accumulated rss usage */
902 u64 acct_vm_mem1; /* accumulated virtual memory usage */
903 u64 acct_timexpd; /* stime + utime since last update */
905 #ifdef CONFIG_CPUSETS
906 nodemask_t mems_allowed; /* Protected by alloc_lock */
907 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
908 int cpuset_mem_spread_rotor;
909 int cpuset_slab_spread_rotor;
911 #ifdef CONFIG_CGROUPS
912 /* Control Group info protected by css_set_lock */
913 struct css_set __rcu *cgroups;
914 /* cg_list protected by css_set_lock and tsk->alloc_lock */
915 struct list_head cg_list;
917 #ifdef CONFIG_INTEL_RDT_A
921 struct robust_list_head __user *robust_list;
923 struct compat_robust_list_head __user *compat_robust_list;
925 struct list_head pi_state_list;
926 struct futex_pi_state *pi_state_cache;
928 #ifdef CONFIG_PERF_EVENTS
929 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
930 struct mutex perf_event_mutex;
931 struct list_head perf_event_list;
933 #ifdef CONFIG_DEBUG_PREEMPT
934 unsigned long preempt_disable_ip;
937 struct mempolicy *mempolicy; /* Protected by alloc_lock */
939 short pref_node_fork;
941 #ifdef CONFIG_NUMA_BALANCING
943 unsigned int numa_scan_period;
944 unsigned int numa_scan_period_max;
945 int numa_preferred_nid;
946 unsigned long numa_migrate_retry;
947 u64 node_stamp; /* migration stamp */
948 u64 last_task_numa_placement;
949 u64 last_sum_exec_runtime;
950 struct callback_head numa_work;
952 struct list_head numa_entry;
953 struct numa_group *numa_group;
956 * numa_faults is an array split into four regions:
957 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
958 * in this precise order.
960 * faults_memory: Exponential decaying average of faults on a per-node
961 * basis. Scheduling placement decisions are made based on these
962 * counts. The values remain static for the duration of a PTE scan.
963 * faults_cpu: Track the nodes the process was running on when a NUMA
964 * hinting fault was incurred.
965 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
966 * during the current scan window. When the scan completes, the counts
967 * in faults_memory and faults_cpu decay and these values are copied.
969 unsigned long *numa_faults;
970 unsigned long total_numa_faults;
973 * numa_faults_locality tracks if faults recorded during the last
974 * scan window were remote/local or failed to migrate. The task scan
975 * period is adapted based on the locality of the faults with different
976 * weights depending on whether they were shared or private faults
978 unsigned long numa_faults_locality[3];
980 unsigned long numa_pages_migrated;
981 #endif /* CONFIG_NUMA_BALANCING */
983 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
984 struct tlbflush_unmap_batch tlb_ubc;
990 * cache last used pipe for splice
992 struct pipe_inode_info *splice_pipe;
994 struct page_frag task_frag;
996 #ifdef CONFIG_TASK_DELAY_ACCT
997 struct task_delay_info *delays;
1000 #ifdef CONFIG_FAULT_INJECTION
1004 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1005 * balance_dirty_pages() for some dirty throttling pause
1008 int nr_dirtied_pause;
1009 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1011 #ifdef CONFIG_LATENCYTOP
1012 int latency_record_count;
1013 struct latency_record latency_record[LT_SAVECOUNT];
1016 * time slack values; these are used to round up poll() and
1017 * select() etc timeout values. These are in nanoseconds.
1020 u64 default_timer_slack_ns;
1023 unsigned int kasan_depth;
1025 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1026 /* Index of current stored address in ret_stack */
1028 /* Stack of return addresses for return function tracing */
1029 struct ftrace_ret_stack *ret_stack;
1030 /* time stamp for last schedule */
1031 unsigned long long ftrace_timestamp;
1033 * Number of functions that haven't been traced
1034 * because of depth overrun.
1036 atomic_t trace_overrun;
1037 /* Pause for the tracing */
1038 atomic_t tracing_graph_pause;
1040 #ifdef CONFIG_TRACING
1041 /* state flags for use by tracers */
1042 unsigned long trace;
1043 /* bitmask and counter of trace recursion */
1044 unsigned long trace_recursion;
1045 #endif /* CONFIG_TRACING */
1047 /* Coverage collection mode enabled for this task (0 if disabled). */
1048 enum kcov_mode kcov_mode;
1049 /* Size of the kcov_area. */
1051 /* Buffer for coverage collection. */
1053 /* kcov desciptor wired with this task or NULL. */
1057 struct mem_cgroup *memcg_in_oom;
1058 gfp_t memcg_oom_gfp_mask;
1059 int memcg_oom_order;
1061 /* number of pages to reclaim on returning to userland */
1062 unsigned int memcg_nr_pages_over_high;
1064 #ifdef CONFIG_UPROBES
1065 struct uprobe_task *utask;
1067 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1068 unsigned int sequential_io;
1069 unsigned int sequential_io_avg;
1071 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1072 unsigned long task_state_change;
1074 int pagefault_disabled;
1076 struct task_struct *oom_reaper_list;
1078 #ifdef CONFIG_VMAP_STACK
1079 struct vm_struct *stack_vm_area;
1081 #ifdef CONFIG_THREAD_INFO_IN_TASK
1082 /* A live task holds one reference. */
1083 atomic_t stack_refcount;
1085 /* CPU-specific state of this task */
1086 struct thread_struct thread;
1088 * WARNING: on x86, 'thread_struct' contains a variable-sized
1089 * structure. It *MUST* be at the end of 'task_struct'.
1091 * Do not put anything below here!
1095 static inline struct pid *task_pid(struct task_struct *task)
1097 return task->pids[PIDTYPE_PID].pid;
1100 static inline struct pid *task_tgid(struct task_struct *task)
1102 return task->group_leader->pids[PIDTYPE_PID].pid;
1106 * Without tasklist or rcu lock it is not safe to dereference
1107 * the result of task_pgrp/task_session even if task == current,
1108 * we can race with another thread doing sys_setsid/sys_setpgid.
1110 static inline struct pid *task_pgrp(struct task_struct *task)
1112 return task->group_leader->pids[PIDTYPE_PGID].pid;
1115 static inline struct pid *task_session(struct task_struct *task)
1117 return task->group_leader->pids[PIDTYPE_SID].pid;
1121 * the helpers to get the task's different pids as they are seen
1122 * from various namespaces
1124 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1125 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1127 * task_xid_nr_ns() : id seen from the ns specified;
1129 * set_task_vxid() : assigns a virtual id to a task;
1131 * see also pid_nr() etc in include/linux/pid.h
1133 pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1134 struct pid_namespace *ns);
1136 static inline pid_t task_pid_nr(struct task_struct *tsk)
1141 static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1142 struct pid_namespace *ns)
1144 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1147 static inline pid_t task_pid_vnr(struct task_struct *tsk)
1149 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1153 static inline pid_t task_tgid_nr(struct task_struct *tsk)
1158 pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1160 static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1162 return pid_vnr(task_tgid(tsk));
1166 static inline int pid_alive(const struct task_struct *p);
1167 static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1173 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1179 static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1181 return task_ppid_nr_ns(tsk, &init_pid_ns);
1184 static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1185 struct pid_namespace *ns)
1187 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1190 static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1192 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1196 static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1197 struct pid_namespace *ns)
1199 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1202 static inline pid_t task_session_vnr(struct task_struct *tsk)
1204 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1207 /* obsolete, do not use */
1208 static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1210 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1214 * pid_alive - check that a task structure is not stale
1215 * @p: Task structure to be checked.
1217 * Test if a process is not yet dead (at most zombie state)
1218 * If pid_alive fails, then pointers within the task structure
1219 * can be stale and must not be dereferenced.
1221 * Return: 1 if the process is alive. 0 otherwise.
1223 static inline int pid_alive(const struct task_struct *p)
1225 return p->pids[PIDTYPE_PID].pid != NULL;
1229 * is_global_init - check if a task structure is init. Since init
1230 * is free to have sub-threads we need to check tgid.
1231 * @tsk: Task structure to be checked.
1233 * Check if a task structure is the first user space task the kernel created.
1235 * Return: 1 if the task structure is init. 0 otherwise.
1237 static inline int is_global_init(struct task_struct *tsk)
1239 return task_tgid_nr(tsk) == 1;
1242 extern struct pid *cad_pid;
1244 extern void free_task(struct task_struct *tsk);
1245 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
1247 extern void __put_task_struct(struct task_struct *t);
1249 static inline void put_task_struct(struct task_struct *t)
1251 if (atomic_dec_and_test(&t->usage))
1252 __put_task_struct(t);
1255 struct task_struct *task_rcu_dereference(struct task_struct **ptask);
1256 struct task_struct *try_get_task_struct(struct task_struct **ptask);
1258 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1259 extern void task_cputime(struct task_struct *t,
1260 u64 *utime, u64 *stime);
1261 extern u64 task_gtime(struct task_struct *t);
1263 static inline void task_cputime(struct task_struct *t,
1264 u64 *utime, u64 *stime)
1270 static inline u64 task_gtime(struct task_struct *t)
1276 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1277 static inline void task_cputime_scaled(struct task_struct *t,
1281 *utimescaled = t->utimescaled;
1282 *stimescaled = t->stimescaled;
1285 static inline void task_cputime_scaled(struct task_struct *t,
1289 task_cputime(t, utimescaled, stimescaled);
1293 extern void task_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
1294 extern void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st);
1299 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1300 #define PF_EXITING 0x00000004 /* getting shut down */
1301 #define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1302 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1303 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1304 #define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1305 #define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1306 #define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1307 #define PF_DUMPCORE 0x00000200 /* dumped core */
1308 #define PF_SIGNALED 0x00000400 /* killed by a signal */
1309 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1310 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1311 #define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1312 #define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1313 #define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1314 #define PF_FROZEN 0x00010000 /* frozen for system suspend */
1315 #define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1316 #define PF_KSWAPD 0x00040000 /* I am kswapd */
1317 #define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1318 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1319 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1320 #define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1321 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1322 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1323 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1324 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1325 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1326 #define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1329 * Only the _current_ task can read/write to tsk->flags, but other
1330 * tasks can access tsk->flags in readonly mode for example
1331 * with tsk_used_math (like during threaded core dumping).
1332 * There is however an exception to this rule during ptrace
1333 * or during fork: the ptracer task is allowed to write to the
1334 * child->flags of its traced child (same goes for fork, the parent
1335 * can write to the child->flags), because we're guaranteed the
1336 * child is not running and in turn not changing child->flags
1337 * at the same time the parent does it.
1339 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1340 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1341 #define clear_used_math() clear_stopped_child_used_math(current)
1342 #define set_used_math() set_stopped_child_used_math(current)
1343 #define conditional_stopped_child_used_math(condition, child) \
1344 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1345 #define conditional_used_math(condition) \
1346 conditional_stopped_child_used_math(condition, current)
1347 #define copy_to_stopped_child_used_math(child) \
1348 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1349 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1350 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1351 #define used_math() tsk_used_math(current)
1353 /* Per-process atomic flags. */
1354 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1355 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1356 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1357 #define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */
1360 #define TASK_PFA_TEST(name, func) \
1361 static inline bool task_##func(struct task_struct *p) \
1362 { return test_bit(PFA_##name, &p->atomic_flags); }
1363 #define TASK_PFA_SET(name, func) \
1364 static inline void task_set_##func(struct task_struct *p) \
1365 { set_bit(PFA_##name, &p->atomic_flags); }
1366 #define TASK_PFA_CLEAR(name, func) \
1367 static inline void task_clear_##func(struct task_struct *p) \
1368 { clear_bit(PFA_##name, &p->atomic_flags); }
1370 TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1371 TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1373 TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1374 TASK_PFA_SET(SPREAD_PAGE, spread_page)
1375 TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1377 TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1378 TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1379 TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1381 TASK_PFA_TEST(LMK_WAITING, lmk_waiting)
1382 TASK_PFA_SET(LMK_WAITING, lmk_waiting)
1384 static inline void tsk_restore_flags(struct task_struct *task,
1385 unsigned long orig_flags, unsigned long flags)
1387 task->flags &= ~flags;
1388 task->flags |= orig_flags & flags;
1391 extern int cpuset_cpumask_can_shrink(const struct cpumask *cur,
1392 const struct cpumask *trial);
1393 extern int task_can_attach(struct task_struct *p,
1394 const struct cpumask *cs_cpus_allowed);
1396 extern void do_set_cpus_allowed(struct task_struct *p,
1397 const struct cpumask *new_mask);
1399 extern int set_cpus_allowed_ptr(struct task_struct *p,
1400 const struct cpumask *new_mask);
1402 static inline void do_set_cpus_allowed(struct task_struct *p,
1403 const struct cpumask *new_mask)
1406 static inline int set_cpus_allowed_ptr(struct task_struct *p,
1407 const struct cpumask *new_mask)
1409 if (!cpumask_test_cpu(0, new_mask))
1415 #ifndef cpu_relax_yield
1416 #define cpu_relax_yield() cpu_relax()
1419 extern unsigned long long
1420 task_sched_runtime(struct task_struct *task);
1422 /* sched_exec is called by processes performing an exec */
1424 extern void sched_exec(void);
1426 #define sched_exec() {}
1429 extern int yield_to(struct task_struct *p, bool preempt);
1430 extern void set_user_nice(struct task_struct *p, long nice);
1431 extern int task_prio(const struct task_struct *p);
1433 * task_nice - return the nice value of a given task.
1434 * @p: the task in question.
1436 * Return: The nice value [ -20 ... 0 ... 19 ].
1438 static inline int task_nice(const struct task_struct *p)
1440 return PRIO_TO_NICE((p)->static_prio);
1442 extern int can_nice(const struct task_struct *p, const int nice);
1443 extern int task_curr(const struct task_struct *p);
1444 extern int idle_cpu(int cpu);
1445 extern int sched_setscheduler(struct task_struct *, int,
1446 const struct sched_param *);
1447 extern int sched_setscheduler_nocheck(struct task_struct *, int,
1448 const struct sched_param *);
1449 extern int sched_setattr(struct task_struct *,
1450 const struct sched_attr *);
1451 extern struct task_struct *idle_task(int cpu);
1453 * is_idle_task - is the specified task an idle task?
1454 * @p: the task in question.
1456 * Return: 1 if @p is an idle task. 0 otherwise.
1458 static inline bool is_idle_task(const struct task_struct *p)
1460 return !!(p->flags & PF_IDLE);
1462 extern struct task_struct *curr_task(int cpu);
1463 extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1467 union thread_union {
1468 #ifndef CONFIG_THREAD_INFO_IN_TASK
1469 struct thread_info thread_info;
1471 unsigned long stack[THREAD_SIZE/sizeof(long)];
1474 #ifndef __HAVE_ARCH_KSTACK_END
1475 static inline int kstack_end(void *addr)
1477 /* Reliable end of stack detection:
1478 * Some APM bios versions misalign the stack
1480 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
1484 extern union thread_union init_thread_union;
1485 extern struct task_struct init_task;
1487 extern struct pid_namespace init_pid_ns;
1490 * find a task by one of its numerical ids
1492 * find_task_by_pid_ns():
1493 * finds a task by its pid in the specified namespace
1494 * find_task_by_vpid():
1495 * finds a task by its virtual pid
1497 * see also find_vpid() etc in include/linux/pid.h
1500 extern struct task_struct *find_task_by_vpid(pid_t nr);
1501 extern struct task_struct *find_task_by_pid_ns(pid_t nr,
1502 struct pid_namespace *ns);
1504 extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1505 extern int wake_up_process(struct task_struct *tsk);
1506 extern void wake_up_new_task(struct task_struct *tsk);
1508 extern void kick_process(struct task_struct *tsk);
1510 static inline void kick_process(struct task_struct *tsk) { }
1513 extern void exit_files(struct task_struct *);
1515 extern void exit_itimers(struct signal_struct *);
1517 extern int do_execve(struct filename *,
1518 const char __user * const __user *,
1519 const char __user * const __user *);
1520 extern int do_execveat(int, struct filename *,
1521 const char __user * const __user *,
1522 const char __user * const __user *,
1525 extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1526 static inline void set_task_comm(struct task_struct *tsk, const char *from)
1528 __set_task_comm(tsk, from, false);
1530 extern char *get_task_comm(char *to, struct task_struct *tsk);
1533 void scheduler_ipi(void);
1534 extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1536 static inline void scheduler_ipi(void) { }
1537 static inline unsigned long wait_task_inactive(struct task_struct *p,
1545 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
1546 * subscriptions and synchronises with wait4(). Also used in procfs. Also
1547 * pins the final release of task.io_context. Also protects ->cpuset and
1548 * ->cgroup.subsys[]. And ->vfork_done.
1550 * Nests both inside and outside of read_lock(&tasklist_lock).
1551 * It must not be nested with write_lock_irq(&tasklist_lock),
1552 * neither inside nor outside.
1554 static inline void task_lock(struct task_struct *p)
1556 spin_lock(&p->alloc_lock);
1559 static inline void task_unlock(struct task_struct *p)
1561 spin_unlock(&p->alloc_lock);
1564 #ifdef CONFIG_THREAD_INFO_IN_TASK
1566 static inline struct thread_info *task_thread_info(struct task_struct *task)
1568 return &task->thread_info;
1572 * When accessing the stack of a non-current task that might exit, use
1573 * try_get_task_stack() instead. task_stack_page will return a pointer
1574 * that could get freed out from under you.
1576 static inline void *task_stack_page(const struct task_struct *task)
1581 #define setup_thread_stack(new,old) do { } while(0)
1583 static inline unsigned long *end_of_stack(const struct task_struct *task)
1588 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1590 #define task_thread_info(task) ((struct thread_info *)(task)->stack)
1591 #define task_stack_page(task) ((void *)(task)->stack)
1593 static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
1595 *task_thread_info(p) = *task_thread_info(org);
1596 task_thread_info(p)->task = p;
1600 * Return the address of the last usable long on the stack.
1602 * When the stack grows down, this is just above the thread
1603 * info struct. Going any lower will corrupt the threadinfo.
1605 * When the stack grows up, this is the highest address.
1606 * Beyond that position, we corrupt data on the next page.
1608 static inline unsigned long *end_of_stack(struct task_struct *p)
1610 #ifdef CONFIG_STACK_GROWSUP
1611 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
1613 return (unsigned long *)(task_thread_info(p) + 1);
1619 #ifdef CONFIG_THREAD_INFO_IN_TASK
1620 static inline void *try_get_task_stack(struct task_struct *tsk)
1622 return atomic_inc_not_zero(&tsk->stack_refcount) ?
1623 task_stack_page(tsk) : NULL;
1626 extern void put_task_stack(struct task_struct *tsk);
1628 static inline void *try_get_task_stack(struct task_struct *tsk)
1630 return task_stack_page(tsk);
1633 static inline void put_task_stack(struct task_struct *tsk) {}
1636 #define task_stack_end_corrupted(task) \
1637 (*(end_of_stack(task)) != STACK_END_MAGIC)
1639 static inline int object_is_on_stack(void *obj)
1641 void *stack = task_stack_page(current);
1643 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
1646 extern void thread_stack_cache_init(void);
1648 #ifdef CONFIG_DEBUG_STACK_USAGE
1649 static inline unsigned long stack_not_used(struct task_struct *p)
1651 unsigned long *n = end_of_stack(p);
1653 do { /* Skip over canary */
1654 # ifdef CONFIG_STACK_GROWSUP
1661 # ifdef CONFIG_STACK_GROWSUP
1662 return (unsigned long)end_of_stack(p) - (unsigned long)n;
1664 return (unsigned long)n - (unsigned long)end_of_stack(p);
1668 extern void set_task_stack_end_magic(struct task_struct *tsk);
1670 /* set thread flags in other task's structures
1671 * - see asm/thread_info.h for TIF_xxxx flags available
1673 static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1675 set_ti_thread_flag(task_thread_info(tsk), flag);
1678 static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1680 clear_ti_thread_flag(task_thread_info(tsk), flag);
1683 static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1685 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1688 static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1690 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1693 static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1695 return test_ti_thread_flag(task_thread_info(tsk), flag);
1698 static inline void set_tsk_need_resched(struct task_struct *tsk)
1700 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1703 static inline void clear_tsk_need_resched(struct task_struct *tsk)
1705 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1708 static inline int test_tsk_need_resched(struct task_struct *tsk)
1710 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1714 * cond_resched() and cond_resched_lock(): latency reduction via
1715 * explicit rescheduling in places that are safe. The return
1716 * value indicates whether a reschedule was done in fact.
1717 * cond_resched_lock() will drop the spinlock before scheduling,
1718 * cond_resched_softirq() will enable bhs before scheduling.
1720 #ifndef CONFIG_PREEMPT
1721 extern int _cond_resched(void);
1723 static inline int _cond_resched(void) { return 0; }
1726 #define cond_resched() ({ \
1727 ___might_sleep(__FILE__, __LINE__, 0); \
1731 extern int __cond_resched_lock(spinlock_t *lock);
1733 #define cond_resched_lock(lock) ({ \
1734 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1735 __cond_resched_lock(lock); \
1738 extern int __cond_resched_softirq(void);
1740 #define cond_resched_softirq() ({ \
1741 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1742 __cond_resched_softirq(); \
1745 static inline void cond_resched_rcu(void)
1747 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1755 * Does a critical section need to be broken due to another
1756 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1757 * but a general need for low latency)
1759 static inline int spin_needbreak(spinlock_t *lock)
1761 #ifdef CONFIG_PREEMPT
1762 return spin_is_contended(lock);
1768 static __always_inline bool need_resched(void)
1770 return unlikely(tif_need_resched());
1774 * Thread group CPU time accounting.
1776 void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
1777 void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
1780 * Wrappers for p->thread_info->cpu access. No-op on UP.
1784 static inline unsigned int task_cpu(const struct task_struct *p)
1786 #ifdef CONFIG_THREAD_INFO_IN_TASK
1789 return task_thread_info(p)->cpu;
1793 static inline int task_node(const struct task_struct *p)
1795 return cpu_to_node(task_cpu(p));
1798 extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1802 static inline unsigned int task_cpu(const struct task_struct *p)
1807 static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1811 #endif /* CONFIG_SMP */
1814 * In order to reduce various lock holder preemption latencies provide an
1815 * interface to see if a vCPU is currently running or not.
1817 * This allows us to terminate optimistic spin loops and block, analogous to
1818 * the native optimistic spin heuristic of testing if the lock owner task is
1821 #ifndef vcpu_is_preempted
1822 # define vcpu_is_preempted(cpu) false
1825 extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1826 extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1828 #ifdef CONFIG_CGROUP_SCHED
1829 extern struct task_group root_task_group;
1830 #endif /* CONFIG_CGROUP_SCHED */
1832 extern int task_can_switch_user(struct user_struct *up,
1833 struct task_struct *tsk);
1835 #ifndef TASK_SIZE_OF
1836 #define TASK_SIZE_OF(tsk) TASK_SIZE