1 /* SPDX-License-Identifier: GPL-2.0 */
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
10 #include <uapi/linux/sched.h>
12 #include <asm/current.h>
14 #include <linux/pid.h>
15 #include <linux/sem.h>
16 #include <linux/shm.h>
17 #include <linux/kcov.h>
18 #include <linux/mutex.h>
19 #include <linux/plist.h>
20 #include <linux/hrtimer.h>
21 #include <linux/seccomp.h>
22 #include <linux/nodemask.h>
23 #include <linux/rcupdate.h>
24 #include <linux/resource.h>
25 #include <linux/latencytop.h>
26 #include <linux/sched/prio.h>
27 #include <linux/signal_types.h>
28 #include <linux/mm_types_task.h>
29 #include <linux/task_io_accounting.h>
31 /* task_struct member predeclarations (sorted alphabetically): */
33 struct backing_dev_info
;
38 struct futex_pi_state
;
43 struct perf_event_context
;
45 struct pipe_inode_info
;
48 struct robust_list_head
;
52 struct sighand_struct
;
54 struct task_delay_info
;
58 * Task state bitmask. NOTE! These bits are also
59 * encoded in fs/proc/array.c: get_task_state().
61 * We have two separate sets of flags: task->state
62 * is about runnability, while task->exit_state are
63 * about the task exiting. Confusing, but this way
64 * modifying one set can't modify the other one by
68 /* Used in tsk->state: */
69 #define TASK_RUNNING 0x0000
70 #define TASK_INTERRUPTIBLE 0x0001
71 #define TASK_UNINTERRUPTIBLE 0x0002
72 #define __TASK_STOPPED 0x0004
73 #define __TASK_TRACED 0x0008
74 /* Used in tsk->exit_state: */
75 #define EXIT_DEAD 0x0010
76 #define EXIT_ZOMBIE 0x0020
77 #define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
78 /* Used in tsk->state again: */
79 #define TASK_PARKED 0x0040
80 #define TASK_DEAD 0x0080
81 #define TASK_WAKEKILL 0x0100
82 #define TASK_WAKING 0x0200
83 #define TASK_NOLOAD 0x0400
84 #define TASK_NEW 0x0800
85 #define TASK_STATE_MAX 0x1000
87 /* Convenience macros for the sake of set_current_state: */
88 #define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
89 #define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
90 #define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
92 #define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
94 /* Convenience macros for the sake of wake_up(): */
95 #define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
96 #define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
98 /* get_task_state(): */
99 #define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
100 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
101 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
104 #define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
106 #define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
108 #define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
110 #define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 (task->flags & PF_FROZEN) == 0 && \
112 (task->state & TASK_NOLOAD) == 0)
114 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
116 #define __set_current_state(state_value) \
118 current->task_state_change = _THIS_IP_; \
119 current->state = (state_value); \
121 #define set_current_state(state_value) \
123 current->task_state_change = _THIS_IP_; \
124 smp_store_mb(current->state, (state_value)); \
129 * set_current_state() includes a barrier so that the write of current->state
130 * is correctly serialised wrt the caller's subsequent test of whether to
134 * set_current_state(TASK_UNINTERRUPTIBLE);
140 * __set_current_state(TASK_RUNNING);
142 * If the caller does not need such serialisation (because, for instance, the
143 * condition test and condition change and wakeup are under the same lock) then
144 * use __set_current_state().
146 * The above is typically ordered against the wakeup, which does:
148 * need_sleep = false;
149 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
151 * Where wake_up_state() (and all other wakeup primitives) imply enough
152 * barriers to order the store of the variable against wakeup.
154 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
155 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
156 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
158 * This is obviously fine, since they both store the exact same value.
160 * Also see the comments of try_to_wake_up().
162 #define __set_current_state(state_value) do { current->state = (state_value); } while (0)
163 #define set_current_state(state_value) smp_store_mb(current->state, (state_value))
166 /* Task command name length: */
167 #define TASK_COMM_LEN 16
178 extern cpumask_var_t cpu_isolated_map
;
180 extern void scheduler_tick(void);
182 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
184 extern long schedule_timeout(long timeout
);
185 extern long schedule_timeout_interruptible(long timeout
);
186 extern long schedule_timeout_killable(long timeout
);
187 extern long schedule_timeout_uninterruptible(long timeout
);
188 extern long schedule_timeout_idle(long timeout
);
189 asmlinkage
void schedule(void);
190 extern void schedule_preempt_disabled(void);
192 extern int __must_check
io_schedule_prepare(void);
193 extern void io_schedule_finish(int token
);
194 extern long io_schedule_timeout(long timeout
);
195 extern void io_schedule(void);
198 * struct prev_cputime - snapshot of system and user cputime
199 * @utime: time spent in user mode
200 * @stime: time spent in system mode
201 * @lock: protects the above two fields
203 * Stores previous user/system time values such that we can guarantee
206 struct prev_cputime
{
207 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
215 * struct task_cputime - collected CPU time counts
216 * @utime: time spent in user mode, in nanoseconds
217 * @stime: time spent in kernel mode, in nanoseconds
218 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
220 * This structure groups together three kinds of CPU time that are tracked for
221 * threads and thread groups. Most things considering CPU time want to group
222 * these counts together and treat all three of them in parallel.
224 struct task_cputime
{
227 unsigned long long sum_exec_runtime
;
230 /* Alternate field names when used on cache expirations: */
231 #define virt_exp utime
232 #define prof_exp stime
233 #define sched_exp sum_exec_runtime
236 /* Task is sleeping or running in a CPU with VTIME inactive: */
238 /* Task runs in userspace in a CPU with VTIME active: */
240 /* Task runs in kernelspace in a CPU with VTIME active: */
246 unsigned long long starttime
;
247 enum vtime_state state
;
254 #ifdef CONFIG_SCHED_INFO
255 /* Cumulative counters: */
257 /* # of times we have run on this CPU: */
258 unsigned long pcount
;
260 /* Time spent waiting on a runqueue: */
261 unsigned long long run_delay
;
265 /* When did we last run on a CPU? */
266 unsigned long long last_arrival
;
268 /* When were we last queued to run? */
269 unsigned long long last_queued
;
271 #endif /* CONFIG_SCHED_INFO */
275 * Integer metrics need fixed point arithmetic, e.g., sched/fair
276 * has a few: load, load_avg, util_avg, freq, and capacity.
278 * We define a basic fixed point arithmetic range, and then formalize
279 * all these metrics based on that basic range.
281 # define SCHED_FIXEDPOINT_SHIFT 10
282 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
285 unsigned long weight
;
290 * The load_avg/util_avg accumulates an infinite geometric series
291 * (see __update_load_avg() in kernel/sched/fair.c).
293 * [load_avg definition]
295 * load_avg = runnable% * scale_load_down(load)
297 * where runnable% is the time ratio that a sched_entity is runnable.
298 * For cfs_rq, it is the aggregated load_avg of all runnable and
299 * blocked sched_entities.
301 * load_avg may also take frequency scaling into account:
303 * load_avg = runnable% * scale_load_down(load) * freq%
305 * where freq% is the CPU frequency normalized to the highest frequency.
307 * [util_avg definition]
309 * util_avg = running% * SCHED_CAPACITY_SCALE
311 * where running% is the time ratio that a sched_entity is running on
312 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
313 * and blocked sched_entities.
315 * util_avg may also factor frequency scaling and CPU capacity scaling:
317 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
319 * where freq% is the same as above, and capacity% is the CPU capacity
320 * normalized to the greatest capacity (due to uarch differences, etc).
322 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
323 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
324 * we therefore scale them to as large a range as necessary. This is for
325 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
329 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
330 * with the highest load (=88761), always runnable on a single cfs_rq,
331 * and should not overflow as the number already hits PID_MAX_LIMIT.
333 * For all other cases (including 32-bit kernels), struct load_weight's
334 * weight will overflow first before we do, because:
336 * Max(load_avg) <= Max(load.weight)
338 * Then it is the load_weight's responsibility to consider overflow
342 u64 last_update_time
;
346 unsigned long load_avg
;
347 unsigned long util_avg
;
350 #ifdef CONFIG_SCHED_EHMP
352 #define ONTIME_MIGRATING 2
355 #define ontime_of(p) (&p->se.ontime)
358 u64 ontime_migration_time
;
361 unsigned long load_avg
;
364 struct ontime_entity
{
365 struct ontime_avg avg
;
371 struct sched_statistics
{
372 #ifdef CONFIG_SCHEDSTATS
382 s64 sum_sleep_runtime
;
389 u64 nr_migrations_cold
;
390 u64 nr_failed_migrations_affine
;
391 u64 nr_failed_migrations_running
;
392 u64 nr_failed_migrations_hot
;
393 u64 nr_forced_migrations
;
397 u64 nr_wakeups_migrate
;
398 u64 nr_wakeups_local
;
399 u64 nr_wakeups_remote
;
400 u64 nr_wakeups_affine
;
401 u64 nr_wakeups_affine_attempts
;
402 u64 nr_wakeups_passive
;
407 struct sched_entity
{
408 /* For load-balancing: */
409 struct load_weight load
;
410 struct rb_node run_node
;
411 struct list_head group_node
;
415 u64 sum_exec_runtime
;
417 u64 prev_sum_exec_runtime
;
421 struct sched_statistics statistics
;
423 #ifdef CONFIG_FAIR_GROUP_SCHED
425 struct sched_entity
*parent
;
426 /* rq on which this entity is (to be) queued: */
427 struct cfs_rq
*cfs_rq
;
428 /* rq "owned" by this entity/group: */
434 * Per entity load average tracking.
436 * Put into separate cache line so it does not
437 * collide with read-mostly values above.
439 struct sched_avg avg ____cacheline_aligned_in_smp
;
441 #ifdef CONFIG_SCHED_EHMP
442 struct ontime_entity ontime
;
446 #ifdef CONFIG_SCHED_WALT
447 #define RAVG_HIST_SIZE_MAX 5
449 /* ravg represents frequency scaled cpu-demand of tasks */
452 * 'mark_start' marks the beginning of an event (task waking up, task
453 * starting to execute, task being preempted) within a window
455 * 'sum' represents how runnable a task has been within current
456 * window. It incorporates both running time and wait time and is
459 * 'sum_history' keeps track of history of 'sum' seen over previous
460 * RAVG_HIST_SIZE windows. Windows where task was entirely sleeping are
463 * 'demand' represents maximum sum seen over previous
464 * sysctl_sched_ravg_hist_size windows. 'demand' could drive frequency
467 * 'curr_window' represents task's contribution to cpu busy time
468 * statistics (rq->curr_runnable_sum) in current window
470 * 'prev_window' represents task's contribution to cpu busy time
471 * statistics (rq->prev_runnable_sum) in previous window
475 u32 sum_history
[RAVG_HIST_SIZE_MAX
];
476 u32 curr_window
, prev_window
;
481 struct sched_rt_entity
{
482 struct list_head run_list
;
483 unsigned long timeout
;
484 unsigned long watchdog_stamp
;
485 unsigned int time_slice
;
486 unsigned short on_rq
;
487 unsigned short on_list
;
489 struct sched_rt_entity
*back
;
490 #ifdef CONFIG_RT_GROUP_SCHED
491 struct sched_rt_entity
*parent
;
492 /* rq on which this entity is (to be) queued: */
494 /* rq "owned" by this entity/group: */
497 } __randomize_layout
;
499 struct sched_dl_entity
{
500 struct rb_node rb_node
;
503 * Original scheduling parameters. Copied here from sched_attr
504 * during sched_setattr(), they will remain the same until
505 * the next sched_setattr().
507 u64 dl_runtime
; /* Maximum runtime for each instance */
508 u64 dl_deadline
; /* Relative deadline of each instance */
509 u64 dl_period
; /* Separation of two instances (period) */
510 u64 dl_bw
; /* dl_runtime / dl_period */
511 u64 dl_density
; /* dl_runtime / dl_deadline */
514 * Actual scheduling parameters. Initialized with the values above,
515 * they are continously updated during task execution. Note that
516 * the remaining runtime could be < 0 in case we are in overrun.
518 s64 runtime
; /* Remaining runtime for this instance */
519 u64 deadline
; /* Absolute deadline for this instance */
520 unsigned int flags
; /* Specifying the scheduler behaviour */
525 * @dl_throttled tells if we exhausted the runtime. If so, the
526 * task has to wait for a replenishment to be performed at the
527 * next firing of dl_timer.
529 * @dl_boosted tells if we are boosted due to DI. If so we are
530 * outside bandwidth enforcement mechanism (but only until we
531 * exit the critical section);
533 * @dl_yielded tells if task gave up the CPU before consuming
534 * all its available runtime during the last job.
536 * @dl_non_contending tells if the task is inactive while still
537 * contributing to the active utilization. In other words, it
538 * indicates if the inactive timer has been armed and its handler
539 * has not been executed yet. This flag is useful to avoid race
540 * conditions between the inactive timer handler and the wakeup
546 int dl_non_contending
;
549 * Bandwidth enforcement timer. Each -deadline task has its
550 * own bandwidth to be enforced, thus we need one timer per task.
552 struct hrtimer dl_timer
;
555 * Inactive timer, responsible for decreasing the active utilization
556 * at the "0-lag time". When a -deadline task blocks, it contributes
557 * to GRUB's active utilization until the "0-lag time", hence a
558 * timer is needed to decrease the active utilization at the correct
561 struct hrtimer inactive_timer
;
570 /* Otherwise the compiler can store garbage here: */
573 u32 s
; /* Set of bits. */
576 enum perf_event_task_context
{
577 perf_invalid_context
= -1,
580 perf_nr_task_contexts
,
584 struct wake_q_node
*next
;
588 #ifdef CONFIG_THREAD_INFO_IN_TASK
590 * For reasons of header soup (see current_thread_info()), this
591 * must be the first element of task_struct.
593 struct thread_info thread_info
;
595 /* -1 unrunnable, 0 runnable, >0 stopped: */
599 * This begins the randomizable portion of task_struct. Only
600 * scheduling-critical items should be added above here.
602 randomized_struct_fields_start
606 /* Per task flags (PF_*), defined further below: */
611 struct llist_node wake_entry
;
613 #ifdef CONFIG_THREAD_INFO_IN_TASK
617 unsigned int wakee_flips
;
618 unsigned long wakee_flip_decay_ts
;
619 struct task_struct
*last_wakee
;
628 unsigned int rt_priority
;
630 const struct sched_class
*sched_class
;
631 struct sched_entity se
;
632 struct sched_rt_entity rt
;
633 #ifdef CONFIG_SCHED_WALT
636 * 'init_load_pct' represents the initial task load assigned to children
643 #ifdef CONFIG_CGROUP_SCHED
644 struct task_group
*sched_task_group
;
646 struct sched_dl_entity dl
;
648 #ifdef CONFIG_PREEMPT_NOTIFIERS
649 /* List of struct preempt_notifier: */
650 struct hlist_head preempt_notifiers
;
653 #ifdef CONFIG_BLK_DEV_IO_TRACE
654 unsigned int btrace_seq
;
659 cpumask_t cpus_allowed
;
661 #ifdef CONFIG_PREEMPT_RCU
662 int rcu_read_lock_nesting
;
663 union rcu_special rcu_read_unlock_special
;
664 struct list_head rcu_node_entry
;
665 struct rcu_node
*rcu_blocked_node
;
666 #endif /* #ifdef CONFIG_PREEMPT_RCU */
668 #ifdef CONFIG_TASKS_RCU
669 unsigned long rcu_tasks_nvcsw
;
670 u8 rcu_tasks_holdout
;
672 int rcu_tasks_idle_cpu
;
673 struct list_head rcu_tasks_holdout_list
;
674 #endif /* #ifdef CONFIG_TASKS_RCU */
676 struct sched_info sched_info
;
678 struct list_head tasks
;
680 struct plist_node pushable_tasks
;
681 struct rb_node pushable_dl_tasks
;
684 struct mm_struct
*mm
;
685 struct mm_struct
*active_mm
;
687 /* Per-thread vma caching: */
688 struct vmacache vmacache
;
690 #ifdef SPLIT_RSS_COUNTING
691 struct task_rss_stat rss_stat
;
696 /* The signal sent when the parent dies: */
698 /* JOBCTL_*, siglock protected: */
699 unsigned long jobctl
;
701 /* Used for emulating ABI behavior of previous Linux versions: */
702 unsigned int personality
;
704 /* Scheduler bits, serialized by scheduler locks: */
705 unsigned sched_reset_on_fork
:1;
706 unsigned sched_contributes_to_load
:1;
707 unsigned sched_migrated
:1;
708 unsigned sched_remote_wakeup
:1;
709 /* Force alignment to the next boundary: */
712 /* Unserialized, strictly 'current' */
714 /* Bit to tell LSMs we're in execve(): */
715 unsigned in_execve
:1;
716 unsigned in_iowait
:1;
717 #ifndef TIF_RESTORE_SIGMASK
718 unsigned restore_sigmask
:1;
721 unsigned memcg_may_oom
:1;
723 unsigned memcg_kmem_skip_account
:1;
726 #ifdef CONFIG_COMPAT_BRK
727 unsigned brk_randomized
:1;
729 #ifdef CONFIG_CGROUPS
730 /* disallow userland-initiated cgroup migration */
731 unsigned no_cgroup_migration
:1;
734 unsigned long atomic_flags
; /* Flags requiring atomic access. */
736 struct restart_block restart_block
;
741 #ifdef CONFIG_CC_STACKPROTECTOR
742 /* Canary value for the -fstack-protector GCC feature: */
743 unsigned long stack_canary
;
746 * Pointers to the (original) parent process, youngest child, younger sibling,
747 * older sibling, respectively. (p->father can be replaced with
748 * p->real_parent->pid)
751 /* Real parent process: */
752 struct task_struct __rcu
*real_parent
;
754 /* Recipient of SIGCHLD, wait4() reports: */
755 struct task_struct __rcu
*parent
;
758 * Children/sibling form the list of natural children:
760 struct list_head children
;
761 struct list_head sibling
;
762 struct task_struct
*group_leader
;
765 * 'ptraced' is the list of tasks this task is using ptrace() on.
767 * This includes both natural children and PTRACE_ATTACH targets.
768 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
770 struct list_head ptraced
;
771 struct list_head ptrace_entry
;
773 /* PID/PID hash table linkage. */
774 struct pid_link pids
[PIDTYPE_MAX
];
775 struct list_head thread_group
;
776 struct list_head thread_node
;
778 struct completion
*vfork_done
;
780 /* CLONE_CHILD_SETTID: */
781 int __user
*set_child_tid
;
783 /* CLONE_CHILD_CLEARTID: */
784 int __user
*clear_child_tid
;
788 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
793 struct prev_cputime prev_cputime
;
794 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
798 #ifdef CONFIG_NO_HZ_FULL
799 atomic_t tick_dep_mask
;
801 /* Context switch counts: */
803 unsigned long nivcsw
;
805 /* Monotonic time in nsecs: */
808 /* Boot based time in nsecs: */
811 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
812 unsigned long min_flt
;
813 unsigned long maj_flt
;
815 #ifdef CONFIG_POSIX_TIMERS
816 struct task_cputime cputime_expires
;
817 struct list_head cpu_timers
[3];
820 /* Process credentials: */
822 /* Tracer's credentials at attach: */
823 const struct cred __rcu
*ptracer_cred
;
825 /* Objective and real subjective task credentials (COW): */
826 const struct cred __rcu
*real_cred
;
828 /* Effective (overridable) subjective task credentials (COW): */
829 const struct cred __rcu
*cred
;
832 * executable name, excluding path.
834 * - normally initialized setup_new_exec()
835 * - access it with [gs]et_task_comm()
836 * - lock it with task_lock()
838 char comm
[TASK_COMM_LEN
];
840 struct nameidata
*nameidata
;
842 #ifdef CONFIG_SYSVIPC
843 struct sysv_sem sysvsem
;
844 struct sysv_shm sysvshm
;
846 #ifdef CONFIG_DETECT_HUNG_TASK
847 unsigned long last_switch_count
;
849 /* Filesystem information: */
850 struct fs_struct
*fs
;
852 /* Open file information: */
853 struct files_struct
*files
;
856 struct nsproxy
*nsproxy
;
858 /* Signal handlers: */
859 struct signal_struct
*signal
;
860 struct sighand_struct
*sighand
;
862 sigset_t real_blocked
;
863 /* Restored if set_restore_sigmask() was used: */
864 sigset_t saved_sigmask
;
865 struct sigpending pending
;
866 unsigned long sas_ss_sp
;
868 unsigned int sas_ss_flags
;
870 struct callback_head
*task_works
;
872 struct audit_context
*audit_context
;
873 #ifdef CONFIG_AUDITSYSCALL
875 unsigned int sessionid
;
877 struct seccomp seccomp
;
879 /* Thread group tracking: */
883 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
884 spinlock_t alloc_lock
;
886 /* Protection of the PI data structures: */
887 raw_spinlock_t pi_lock
;
889 struct wake_q_node wake_q
;
891 #ifdef CONFIG_RT_MUTEXES
892 /* PI waiters blocked on a rt_mutex held by this task: */
893 struct rb_root_cached pi_waiters
;
894 /* Updated under owner's pi_lock and rq lock */
895 struct task_struct
*pi_top_task
;
896 /* Deadlock detection and priority inheritance handling: */
897 struct rt_mutex_waiter
*pi_blocked_on
;
900 #ifdef CONFIG_DEBUG_MUTEXES
901 /* Mutex deadlock detection: */
902 struct mutex_waiter
*blocked_on
;
905 #ifdef CONFIG_TRACE_IRQFLAGS
906 unsigned int irq_events
;
907 unsigned long hardirq_enable_ip
;
908 unsigned long hardirq_disable_ip
;
909 unsigned int hardirq_enable_event
;
910 unsigned int hardirq_disable_event
;
911 int hardirqs_enabled
;
913 unsigned long softirq_disable_ip
;
914 unsigned long softirq_enable_ip
;
915 unsigned int softirq_disable_event
;
916 unsigned int softirq_enable_event
;
917 int softirqs_enabled
;
921 #ifdef CONFIG_LOCKDEP
922 # define MAX_LOCK_DEPTH 48UL
925 unsigned int lockdep_recursion
;
926 struct held_lock held_locks
[MAX_LOCK_DEPTH
];
929 #ifdef CONFIG_LOCKDEP_CROSSRELEASE
930 #define MAX_XHLOCKS_NR 64UL
931 struct hist_lock
*xhlocks
; /* Crossrelease history locks */
932 unsigned int xhlock_idx
;
933 /* For restoring at history boundaries */
934 unsigned int xhlock_idx_hist
[XHLOCK_CTX_NR
];
935 unsigned int hist_id
;
936 /* For overwrite check at each context exit */
937 unsigned int hist_id_save
[XHLOCK_CTX_NR
];
941 unsigned int in_ubsan
;
944 /* Journalling filesystem info: */
947 /* Stacked block device info: */
948 struct bio_list
*bio_list
;
951 /* Stack plugging: */
952 struct blk_plug
*plug
;
956 struct reclaim_state
*reclaim_state
;
958 struct backing_dev_info
*backing_dev_info
;
960 struct io_context
*io_context
;
963 unsigned long ptrace_message
;
964 siginfo_t
*last_siginfo
;
966 struct task_io_accounting ioac
;
967 #ifdef CONFIG_TASK_XACCT
968 /* Accumulated RSS usage: */
970 /* Accumulated virtual memory usage: */
972 /* stime + utime since last update: */
975 #ifdef CONFIG_CPUSETS
976 /* Protected by ->alloc_lock: */
977 nodemask_t mems_allowed
;
978 /* Seqence number to catch updates: */
979 seqcount_t mems_allowed_seq
;
980 int cpuset_mem_spread_rotor
;
981 int cpuset_slab_spread_rotor
;
983 #ifdef CONFIG_CGROUPS
984 /* Control Group info protected by css_set_lock: */
985 struct css_set __rcu
*cgroups
;
986 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
987 struct list_head cg_list
;
989 #ifdef CONFIG_INTEL_RDT
994 struct robust_list_head __user
*robust_list
;
996 struct compat_robust_list_head __user
*compat_robust_list
;
998 struct list_head pi_state_list
;
999 struct futex_pi_state
*pi_state_cache
;
1001 #ifdef CONFIG_PERF_EVENTS
1002 struct perf_event_context
*perf_event_ctxp
[perf_nr_task_contexts
];
1003 struct mutex perf_event_mutex
;
1004 struct list_head perf_event_list
;
1006 #ifdef CONFIG_DEBUG_PREEMPT
1007 unsigned long preempt_disable_ip
;
1010 /* Protected by alloc_lock: */
1011 struct mempolicy
*mempolicy
;
1013 short pref_node_fork
;
1015 #ifdef CONFIG_NUMA_BALANCING
1017 unsigned int numa_scan_period
;
1018 unsigned int numa_scan_period_max
;
1019 int numa_preferred_nid
;
1020 unsigned long numa_migrate_retry
;
1021 /* Migration stamp: */
1023 u64 last_task_numa_placement
;
1024 u64 last_sum_exec_runtime
;
1025 struct callback_head numa_work
;
1027 struct list_head numa_entry
;
1028 struct numa_group
*numa_group
;
1031 * numa_faults is an array split into four regions:
1032 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1033 * in this precise order.
1035 * faults_memory: Exponential decaying average of faults on a per-node
1036 * basis. Scheduling placement decisions are made based on these
1037 * counts. The values remain static for the duration of a PTE scan.
1038 * faults_cpu: Track the nodes the process was running on when a NUMA
1039 * hinting fault was incurred.
1040 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1041 * during the current scan window. When the scan completes, the counts
1042 * in faults_memory and faults_cpu decay and these values are copied.
1044 unsigned long *numa_faults
;
1045 unsigned long total_numa_faults
;
1048 * numa_faults_locality tracks if faults recorded during the last
1049 * scan window were remote/local or failed to migrate. The task scan
1050 * period is adapted based on the locality of the faults with different
1051 * weights depending on whether they were shared or private faults
1053 unsigned long numa_faults_locality
[3];
1055 unsigned long numa_pages_migrated
;
1056 #endif /* CONFIG_NUMA_BALANCING */
1058 struct tlbflush_unmap_batch tlb_ubc
;
1060 struct rcu_head rcu
;
1062 /* Cache last used pipe for splice(): */
1063 struct pipe_inode_info
*splice_pipe
;
1065 struct page_frag task_frag
;
1067 #ifdef CONFIG_TASK_DELAY_ACCT
1068 struct task_delay_info
*delays
;
1071 #ifdef CONFIG_FAULT_INJECTION
1073 unsigned int fail_nth
;
1076 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1077 * balance_dirty_pages() for a dirty throttling pause:
1080 int nr_dirtied_pause
;
1081 /* Start of a write-and-pause period: */
1082 unsigned long dirty_paused_when
;
1084 #ifdef CONFIG_LATENCYTOP
1085 int latency_record_count
;
1086 struct latency_record latency_record
[LT_SAVECOUNT
];
1089 * Time slack values; these are used to round up poll() and
1090 * select() etc timeout values. These are in nanoseconds.
1093 u64 default_timer_slack_ns
;
1096 unsigned int kasan_depth
;
1099 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1100 /* Index of current stored address in ret_stack: */
1103 /* Stack of return addresses for return function tracing: */
1104 struct ftrace_ret_stack
*ret_stack
;
1106 /* Timestamp for last schedule: */
1107 unsigned long long ftrace_timestamp
;
1110 * Number of functions that haven't been traced
1111 * because of depth overrun:
1113 atomic_t trace_overrun
;
1115 /* Pause tracing: */
1116 atomic_t tracing_graph_pause
;
1119 #ifdef CONFIG_TRACING
1120 /* State flags for use by tracers: */
1121 unsigned long trace
;
1123 /* Bitmask and counter of trace recursion: */
1124 unsigned long trace_recursion
;
1125 #endif /* CONFIG_TRACING */
1128 /* Coverage collection mode enabled for this task (0 if disabled): */
1129 enum kcov_mode kcov_mode
;
1131 /* Size of the kcov_area: */
1132 unsigned int kcov_size
;
1134 /* Buffer for coverage collection: */
1137 /* KCOV descriptor wired with this task or NULL: */
1142 struct mem_cgroup
*memcg_in_oom
;
1143 gfp_t memcg_oom_gfp_mask
;
1144 int memcg_oom_order
;
1146 /* Number of pages to reclaim on returning to userland: */
1147 unsigned int memcg_nr_pages_over_high
;
1150 #ifdef CONFIG_UPROBES
1151 struct uprobe_task
*utask
;
1153 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1154 unsigned int sequential_io
;
1155 unsigned int sequential_io_avg
;
1157 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1158 unsigned long task_state_change
;
1160 int pagefault_disabled
;
1162 struct task_struct
*oom_reaper_list
;
1164 #ifdef CONFIG_VMAP_STACK
1165 struct vm_struct
*stack_vm_area
;
1167 #ifdef CONFIG_THREAD_INFO_IN_TASK
1168 /* A live task holds one reference: */
1169 atomic_t stack_refcount
;
1171 #ifdef CONFIG_LIVEPATCH
1174 #ifdef CONFIG_SECURITY
1175 /* Used by LSM modules for access restriction: */
1180 * New fields for task_struct should be added above here, so that
1181 * they are included in the randomized portion of task_struct.
1183 randomized_struct_fields_end
1185 /* CPU-specific state of this task: */
1186 struct thread_struct thread
;
1189 * WARNING: on x86, 'thread_struct' contains a variable-sized
1190 * structure. It *MUST* be at the end of 'task_struct'.
1192 * Do not put anything below here!
1196 static inline struct pid
*task_pid(struct task_struct
*task
)
1198 return task
->pids
[PIDTYPE_PID
].pid
;
1201 static inline struct pid
*task_tgid(struct task_struct
*task
)
1203 return task
->group_leader
->pids
[PIDTYPE_PID
].pid
;
1207 * Without tasklist or RCU lock it is not safe to dereference
1208 * the result of task_pgrp/task_session even if task == current,
1209 * we can race with another thread doing sys_setsid/sys_setpgid.
1211 static inline struct pid
*task_pgrp(struct task_struct
*task
)
1213 return task
->group_leader
->pids
[PIDTYPE_PGID
].pid
;
1216 static inline struct pid
*task_session(struct task_struct
*task
)
1218 return task
->group_leader
->pids
[PIDTYPE_SID
].pid
;
1222 * the helpers to get the task's different pids as they are seen
1223 * from various namespaces
1225 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1226 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1228 * task_xid_nr_ns() : id seen from the ns specified;
1230 * see also pid_nr() etc in include/linux/pid.h
1232 pid_t
__task_pid_nr_ns(struct task_struct
*task
, enum pid_type type
, struct pid_namespace
*ns
);
1234 static inline pid_t
task_pid_nr(struct task_struct
*tsk
)
1239 static inline pid_t
task_pid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1241 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, ns
);
1244 static inline pid_t
task_pid_vnr(struct task_struct
*tsk
)
1246 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, NULL
);
1250 static inline pid_t
task_tgid_nr(struct task_struct
*tsk
)
1256 * pid_alive - check that a task structure is not stale
1257 * @p: Task structure to be checked.
1259 * Test if a process is not yet dead (at most zombie state)
1260 * If pid_alive fails, then pointers within the task structure
1261 * can be stale and must not be dereferenced.
1263 * Return: 1 if the process is alive. 0 otherwise.
1265 static inline int pid_alive(const struct task_struct
*p
)
1267 return p
->pids
[PIDTYPE_PID
].pid
!= NULL
;
1270 static inline pid_t
task_pgrp_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1272 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, ns
);
1275 static inline pid_t
task_pgrp_vnr(struct task_struct
*tsk
)
1277 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, NULL
);
1281 static inline pid_t
task_session_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1283 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, ns
);
1286 static inline pid_t
task_session_vnr(struct task_struct
*tsk
)
1288 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, NULL
);
1291 static inline pid_t
task_tgid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1293 return __task_pid_nr_ns(tsk
, __PIDTYPE_TGID
, ns
);
1296 static inline pid_t
task_tgid_vnr(struct task_struct
*tsk
)
1298 return __task_pid_nr_ns(tsk
, __PIDTYPE_TGID
, NULL
);
1301 static inline pid_t
task_ppid_nr_ns(const struct task_struct
*tsk
, struct pid_namespace
*ns
)
1307 pid
= task_tgid_nr_ns(rcu_dereference(tsk
->real_parent
), ns
);
1313 static inline pid_t
task_ppid_nr(const struct task_struct
*tsk
)
1315 return task_ppid_nr_ns(tsk
, &init_pid_ns
);
1318 /* Obsolete, do not use: */
1319 static inline pid_t
task_pgrp_nr(struct task_struct
*tsk
)
1321 return task_pgrp_nr_ns(tsk
, &init_pid_ns
);
1324 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1325 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1327 static inline unsigned int __get_task_state(struct task_struct
*tsk
)
1329 unsigned int tsk_state
= READ_ONCE(tsk
->state
);
1330 unsigned int state
= (tsk_state
| tsk
->exit_state
) & TASK_REPORT
;
1332 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX
);
1334 if (tsk_state
== TASK_IDLE
)
1335 state
= TASK_REPORT_IDLE
;
1340 static inline char __task_state_to_char(unsigned int state
)
1342 static const char state_char
[] = "RSDTtXZPI";
1344 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX
) != sizeof(state_char
) - 1);
1346 return state_char
[state
];
1349 static inline char task_state_to_char(struct task_struct
*tsk
)
1351 return __task_state_to_char(__get_task_state(tsk
));
1355 * is_global_init - check if a task structure is init. Since init
1356 * is free to have sub-threads we need to check tgid.
1357 * @tsk: Task structure to be checked.
1359 * Check if a task structure is the first user space task the kernel created.
1361 * Return: 1 if the task structure is init. 0 otherwise.
1363 static inline int is_global_init(struct task_struct
*tsk
)
1365 return task_tgid_nr(tsk
) == 1;
1368 extern struct pid
*cad_pid
;
1373 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1374 #define PF_EXITING 0x00000004 /* Getting shut down */
1375 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1376 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1377 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1378 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1379 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1380 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1381 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1382 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1383 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1384 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1385 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1386 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1387 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1388 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1389 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1390 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1391 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1392 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1393 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1394 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1395 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1396 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1397 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1398 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1399 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1400 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1403 * Only the _current_ task can read/write to tsk->flags, but other
1404 * tasks can access tsk->flags in readonly mode for example
1405 * with tsk_used_math (like during threaded core dumping).
1406 * There is however an exception to this rule during ptrace
1407 * or during fork: the ptracer task is allowed to write to the
1408 * child->flags of its traced child (same goes for fork, the parent
1409 * can write to the child->flags), because we're guaranteed the
1410 * child is not running and in turn not changing child->flags
1411 * at the same time the parent does it.
1413 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1414 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1415 #define clear_used_math() clear_stopped_child_used_math(current)
1416 #define set_used_math() set_stopped_child_used_math(current)
1418 #define conditional_stopped_child_used_math(condition, child) \
1419 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1421 #define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1423 #define copy_to_stopped_child_used_math(child) \
1424 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1426 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1427 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1428 #define used_math() tsk_used_math(current)
1430 static inline bool is_percpu_thread(void)
1433 return (current
->flags
& PF_NO_SETAFFINITY
) &&
1434 (current
->nr_cpus_allowed
== 1);
1440 /* Per-process atomic flags. */
1441 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1442 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1443 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1446 #define TASK_PFA_TEST(name, func) \
1447 static inline bool task_##func(struct task_struct *p) \
1448 { return test_bit(PFA_##name, &p->atomic_flags); }
1450 #define TASK_PFA_SET(name, func) \
1451 static inline void task_set_##func(struct task_struct *p) \
1452 { set_bit(PFA_##name, &p->atomic_flags); }
1454 #define TASK_PFA_CLEAR(name, func) \
1455 static inline void task_clear_##func(struct task_struct *p) \
1456 { clear_bit(PFA_##name, &p->atomic_flags); }
1458 TASK_PFA_TEST(NO_NEW_PRIVS
, no_new_privs
)
1459 TASK_PFA_SET(NO_NEW_PRIVS
, no_new_privs
)
1461 TASK_PFA_TEST(SPREAD_PAGE
, spread_page
)
1462 TASK_PFA_SET(SPREAD_PAGE
, spread_page
)
1463 TASK_PFA_CLEAR(SPREAD_PAGE
, spread_page
)
1465 TASK_PFA_TEST(SPREAD_SLAB
, spread_slab
)
1466 TASK_PFA_SET(SPREAD_SLAB
, spread_slab
)
1467 TASK_PFA_CLEAR(SPREAD_SLAB
, spread_slab
)
1470 current_restore_flags(unsigned long orig_flags
, unsigned long flags
)
1472 current
->flags
&= ~flags
;
1473 current
->flags
|= orig_flags
& flags
;
1476 extern int cpuset_cpumask_can_shrink(const struct cpumask
*cur
, const struct cpumask
*trial
);
1477 extern int task_can_attach(struct task_struct
*p
, const struct cpumask
*cs_cpus_allowed
);
1479 extern void do_set_cpus_allowed(struct task_struct
*p
, const struct cpumask
*new_mask
);
1480 extern int set_cpus_allowed_ptr(struct task_struct
*p
, const struct cpumask
*new_mask
);
1482 static inline void do_set_cpus_allowed(struct task_struct
*p
, const struct cpumask
*new_mask
)
1485 static inline int set_cpus_allowed_ptr(struct task_struct
*p
, const struct cpumask
*new_mask
)
1487 if (!cpumask_test_cpu(0, new_mask
))
1493 #ifndef cpu_relax_yield
1494 #define cpu_relax_yield() cpu_relax()
1497 extern int yield_to(struct task_struct
*p
, bool preempt
);
1498 extern void set_user_nice(struct task_struct
*p
, long nice
);
1499 extern int task_prio(const struct task_struct
*p
);
1502 * task_nice - return the nice value of a given task.
1503 * @p: the task in question.
1505 * Return: The nice value [ -20 ... 0 ... 19 ].
1507 static inline int task_nice(const struct task_struct
*p
)
1509 return PRIO_TO_NICE((p
)->static_prio
);
1512 extern int can_nice(const struct task_struct
*p
, const int nice
);
1513 extern int task_curr(const struct task_struct
*p
);
1514 extern int idle_cpu(int cpu
);
1515 extern int sched_setscheduler(struct task_struct
*, int, const struct sched_param
*);
1516 extern int sched_setscheduler_nocheck(struct task_struct
*, int, const struct sched_param
*);
1517 extern int sched_setattr(struct task_struct
*, const struct sched_attr
*);
1518 extern struct task_struct
*idle_task(int cpu
);
1521 * is_idle_task - is the specified task an idle task?
1522 * @p: the task in question.
1524 * Return: 1 if @p is an idle task. 0 otherwise.
1526 static inline bool is_idle_task(const struct task_struct
*p
)
1528 return !!(p
->flags
& PF_IDLE
);
1531 extern struct task_struct
*curr_task(int cpu
);
1532 extern void ia64_set_curr_task(int cpu
, struct task_struct
*p
);
1536 union thread_union
{
1537 #ifndef CONFIG_THREAD_INFO_IN_TASK
1538 struct thread_info thread_info
;
1540 unsigned long stack
[THREAD_SIZE
/sizeof(long)];
1543 #ifdef CONFIG_THREAD_INFO_IN_TASK
1544 static inline struct thread_info
*task_thread_info(struct task_struct
*task
)
1546 return &task
->thread_info
;
1548 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1549 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1553 * find a task by one of its numerical ids
1555 * find_task_by_pid_ns():
1556 * finds a task by its pid in the specified namespace
1557 * find_task_by_vpid():
1558 * finds a task by its virtual pid
1560 * see also find_vpid() etc in include/linux/pid.h
1563 extern struct task_struct
*find_task_by_vpid(pid_t nr
);
1564 extern struct task_struct
*find_task_by_pid_ns(pid_t nr
, struct pid_namespace
*ns
);
1566 extern int wake_up_state(struct task_struct
*tsk
, unsigned int state
);
1567 extern int wake_up_process(struct task_struct
*tsk
);
1568 extern void wake_up_new_task(struct task_struct
*tsk
);
1571 extern void kick_process(struct task_struct
*tsk
);
1573 static inline void kick_process(struct task_struct
*tsk
) { }
1576 extern void __set_task_comm(struct task_struct
*tsk
, const char *from
, bool exec
);
1578 static inline void set_task_comm(struct task_struct
*tsk
, const char *from
)
1580 __set_task_comm(tsk
, from
, false);
1583 extern char *__get_task_comm(char *to
, size_t len
, struct task_struct
*tsk
);
1584 #define get_task_comm(buf, tsk) ({ \
1585 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1586 __get_task_comm(buf, sizeof(buf), tsk); \
1590 void scheduler_ipi(void);
1591 extern unsigned long wait_task_inactive(struct task_struct
*, long match_state
);
1593 static inline void scheduler_ipi(void) { }
1594 static inline unsigned long wait_task_inactive(struct task_struct
*p
, long match_state
)
1601 * Set thread flags in other task's structures.
1602 * See asm/thread_info.h for TIF_xxxx flags available:
1604 static inline void set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1606 set_ti_thread_flag(task_thread_info(tsk
), flag
);
1609 static inline void clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1611 clear_ti_thread_flag(task_thread_info(tsk
), flag
);
1614 static inline int test_and_set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1616 return test_and_set_ti_thread_flag(task_thread_info(tsk
), flag
);
1619 static inline int test_and_clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1621 return test_and_clear_ti_thread_flag(task_thread_info(tsk
), flag
);
1624 static inline int test_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1626 return test_ti_thread_flag(task_thread_info(tsk
), flag
);
1629 static inline void set_tsk_need_resched(struct task_struct
*tsk
)
1631 set_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
1634 static inline void clear_tsk_need_resched(struct task_struct
*tsk
)
1636 clear_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
1639 static inline int test_tsk_need_resched(struct task_struct
*tsk
)
1641 return unlikely(test_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
));
1645 * cond_resched() and cond_resched_lock(): latency reduction via
1646 * explicit rescheduling in places that are safe. The return
1647 * value indicates whether a reschedule was done in fact.
1648 * cond_resched_lock() will drop the spinlock before scheduling,
1649 * cond_resched_softirq() will enable bhs before scheduling.
1651 #ifndef CONFIG_PREEMPT
1652 extern int _cond_resched(void);
1654 static inline int _cond_resched(void) { return 0; }
1657 #define cond_resched() ({ \
1658 ___might_sleep(__FILE__, __LINE__, 0); \
1662 extern int __cond_resched_lock(spinlock_t
*lock
);
1664 #define cond_resched_lock(lock) ({ \
1665 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1666 __cond_resched_lock(lock); \
1669 extern int __cond_resched_softirq(void);
1671 #define cond_resched_softirq() ({ \
1672 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1673 __cond_resched_softirq(); \
1676 static inline void cond_resched_rcu(void)
1678 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1686 * Does a critical section need to be broken due to another
1687 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1688 * but a general need for low latency)
1690 static inline int spin_needbreak(spinlock_t
*lock
)
1692 #ifdef CONFIG_PREEMPT
1693 return spin_is_contended(lock
);
1699 static __always_inline
bool need_resched(void)
1701 return unlikely(tif_need_resched());
1705 * Wrappers for p->thread_info->cpu access. No-op on UP.
1709 static inline unsigned int task_cpu(const struct task_struct
*p
)
1711 #ifdef CONFIG_THREAD_INFO_IN_TASK
1714 return task_thread_info(p
)->cpu
;
1718 extern void set_task_cpu(struct task_struct
*p
, unsigned int cpu
);
1722 static inline unsigned int task_cpu(const struct task_struct
*p
)
1727 static inline void set_task_cpu(struct task_struct
*p
, unsigned int cpu
)
1731 #endif /* CONFIG_SMP */
1734 * In order to reduce various lock holder preemption latencies provide an
1735 * interface to see if a vCPU is currently running or not.
1737 * This allows us to terminate optimistic spin loops and block, analogous to
1738 * the native optimistic spin heuristic of testing if the lock owner task is
1741 #ifndef vcpu_is_preempted
1742 # define vcpu_is_preempted(cpu) false
1745 extern long sched_setaffinity(pid_t pid
, const struct cpumask
*new_mask
);
1746 extern long sched_getaffinity(pid_t pid
, struct cpumask
*mask
);
1748 #ifndef TASK_SIZE_OF
1749 #define TASK_SIZE_OF(tsk) TASK_SIZE