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
169 extern void scheduler_tick(void);
171 #define MAX_SCHEDULE_TIMEOUT LONG_MAX
173 extern long schedule_timeout(long timeout
);
174 extern long schedule_timeout_interruptible(long timeout
);
175 extern long schedule_timeout_killable(long timeout
);
176 extern long schedule_timeout_uninterruptible(long timeout
);
177 extern long schedule_timeout_idle(long timeout
);
178 asmlinkage
void schedule(void);
179 extern void schedule_preempt_disabled(void);
181 extern int __must_check
io_schedule_prepare(void);
182 extern void io_schedule_finish(int token
);
183 extern long io_schedule_timeout(long timeout
);
184 extern void io_schedule(void);
187 * struct prev_cputime - snapshot of system and user cputime
188 * @utime: time spent in user mode
189 * @stime: time spent in system mode
190 * @lock: protects the above two fields
192 * Stores previous user/system time values such that we can guarantee
195 struct prev_cputime
{
196 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
204 * struct task_cputime - collected CPU time counts
205 * @utime: time spent in user mode, in nanoseconds
206 * @stime: time spent in kernel mode, in nanoseconds
207 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
209 * This structure groups together three kinds of CPU time that are tracked for
210 * threads and thread groups. Most things considering CPU time want to group
211 * these counts together and treat all three of them in parallel.
213 struct task_cputime
{
216 unsigned long long sum_exec_runtime
;
219 /* Alternate field names when used on cache expirations: */
220 #define virt_exp utime
221 #define prof_exp stime
222 #define sched_exp sum_exec_runtime
225 /* Task is sleeping or running in a CPU with VTIME inactive: */
227 /* Task runs in userspace in a CPU with VTIME active: */
229 /* Task runs in kernelspace in a CPU with VTIME active: */
235 unsigned long long starttime
;
236 enum vtime_state state
;
243 #ifdef CONFIG_SCHED_INFO
244 /* Cumulative counters: */
246 /* # of times we have run on this CPU: */
247 unsigned long pcount
;
249 /* Time spent waiting on a runqueue: */
250 unsigned long long run_delay
;
254 /* When did we last run on a CPU? */
255 unsigned long long last_arrival
;
257 /* When were we last queued to run? */
258 unsigned long long last_queued
;
260 #endif /* CONFIG_SCHED_INFO */
264 * Integer metrics need fixed point arithmetic, e.g., sched/fair
265 * has a few: load, load_avg, util_avg, freq, and capacity.
267 * We define a basic fixed point arithmetic range, and then formalize
268 * all these metrics based on that basic range.
270 # define SCHED_FIXEDPOINT_SHIFT 10
271 # define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
274 unsigned long weight
;
279 * The load_avg/util_avg accumulates an infinite geometric series
280 * (see __update_load_avg() in kernel/sched/fair.c).
282 * [load_avg definition]
284 * load_avg = runnable% * scale_load_down(load)
286 * where runnable% is the time ratio that a sched_entity is runnable.
287 * For cfs_rq, it is the aggregated load_avg of all runnable and
288 * blocked sched_entities.
290 * load_avg may also take frequency scaling into account:
292 * load_avg = runnable% * scale_load_down(load) * freq%
294 * where freq% is the CPU frequency normalized to the highest frequency.
296 * [util_avg definition]
298 * util_avg = running% * SCHED_CAPACITY_SCALE
300 * where running% is the time ratio that a sched_entity is running on
301 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
302 * and blocked sched_entities.
304 * util_avg may also factor frequency scaling and CPU capacity scaling:
306 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
308 * where freq% is the same as above, and capacity% is the CPU capacity
309 * normalized to the greatest capacity (due to uarch differences, etc).
311 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
312 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
313 * we therefore scale them to as large a range as necessary. This is for
314 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
318 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
319 * with the highest load (=88761), always runnable on a single cfs_rq,
320 * and should not overflow as the number already hits PID_MAX_LIMIT.
322 * For all other cases (including 32-bit kernels), struct load_weight's
323 * weight will overflow first before we do, because:
325 * Max(load_avg) <= Max(load.weight)
327 * Then it is the load_weight's responsibility to consider overflow
331 u64 last_update_time
;
333 u64 runnable_load_sum
;
336 unsigned long load_avg
;
337 unsigned long runnable_load_avg
;
338 unsigned long util_avg
;
341 struct sched_statistics
{
342 #ifdef CONFIG_SCHEDSTATS
352 s64 sum_sleep_runtime
;
359 u64 nr_migrations_cold
;
360 u64 nr_failed_migrations_affine
;
361 u64 nr_failed_migrations_running
;
362 u64 nr_failed_migrations_hot
;
363 u64 nr_forced_migrations
;
367 u64 nr_wakeups_migrate
;
368 u64 nr_wakeups_local
;
369 u64 nr_wakeups_remote
;
370 u64 nr_wakeups_affine
;
371 u64 nr_wakeups_affine_attempts
;
372 u64 nr_wakeups_passive
;
377 struct sched_entity
{
378 /* For load-balancing: */
379 struct load_weight load
;
380 unsigned long runnable_weight
;
381 struct rb_node run_node
;
382 struct list_head group_node
;
386 u64 sum_exec_runtime
;
388 u64 prev_sum_exec_runtime
;
392 struct sched_statistics statistics
;
394 #ifdef CONFIG_FAIR_GROUP_SCHED
396 struct sched_entity
*parent
;
397 /* rq on which this entity is (to be) queued: */
398 struct cfs_rq
*cfs_rq
;
399 /* rq "owned" by this entity/group: */
405 * Per entity load average tracking.
407 * Put into separate cache line so it does not
408 * collide with read-mostly values above.
410 struct sched_avg avg ____cacheline_aligned_in_smp
;
414 struct sched_rt_entity
{
415 struct list_head run_list
;
416 unsigned long timeout
;
417 unsigned long watchdog_stamp
;
418 unsigned int time_slice
;
419 unsigned short on_rq
;
420 unsigned short on_list
;
422 struct sched_rt_entity
*back
;
423 #ifdef CONFIG_RT_GROUP_SCHED
424 struct sched_rt_entity
*parent
;
425 /* rq on which this entity is (to be) queued: */
427 /* rq "owned" by this entity/group: */
430 } __randomize_layout
;
432 struct sched_dl_entity
{
433 struct rb_node rb_node
;
436 * Original scheduling parameters. Copied here from sched_attr
437 * during sched_setattr(), they will remain the same until
438 * the next sched_setattr().
440 u64 dl_runtime
; /* Maximum runtime for each instance */
441 u64 dl_deadline
; /* Relative deadline of each instance */
442 u64 dl_period
; /* Separation of two instances (period) */
443 u64 dl_bw
; /* dl_runtime / dl_period */
444 u64 dl_density
; /* dl_runtime / dl_deadline */
447 * Actual scheduling parameters. Initialized with the values above,
448 * they are continously updated during task execution. Note that
449 * the remaining runtime could be < 0 in case we are in overrun.
451 s64 runtime
; /* Remaining runtime for this instance */
452 u64 deadline
; /* Absolute deadline for this instance */
453 unsigned int flags
; /* Specifying the scheduler behaviour */
458 * @dl_throttled tells if we exhausted the runtime. If so, the
459 * task has to wait for a replenishment to be performed at the
460 * next firing of dl_timer.
462 * @dl_boosted tells if we are boosted due to DI. If so we are
463 * outside bandwidth enforcement mechanism (but only until we
464 * exit the critical section);
466 * @dl_yielded tells if task gave up the CPU before consuming
467 * all its available runtime during the last job.
469 * @dl_non_contending tells if the task is inactive while still
470 * contributing to the active utilization. In other words, it
471 * indicates if the inactive timer has been armed and its handler
472 * has not been executed yet. This flag is useful to avoid race
473 * conditions between the inactive timer handler and the wakeup
476 * @dl_overrun tells if the task asked to be informed about runtime
479 unsigned int dl_throttled
: 1;
480 unsigned int dl_boosted
: 1;
481 unsigned int dl_yielded
: 1;
482 unsigned int dl_non_contending
: 1;
483 unsigned int dl_overrun
: 1;
486 * Bandwidth enforcement timer. Each -deadline task has its
487 * own bandwidth to be enforced, thus we need one timer per task.
489 struct hrtimer dl_timer
;
492 * Inactive timer, responsible for decreasing the active utilization
493 * at the "0-lag time". When a -deadline task blocks, it contributes
494 * to GRUB's active utilization until the "0-lag time", hence a
495 * timer is needed to decrease the active utilization at the correct
498 struct hrtimer inactive_timer
;
507 /* Otherwise the compiler can store garbage here: */
510 u32 s
; /* Set of bits. */
513 enum perf_event_task_context
{
514 perf_invalid_context
= -1,
517 perf_nr_task_contexts
,
521 struct wake_q_node
*next
;
525 #ifdef CONFIG_THREAD_INFO_IN_TASK
527 * For reasons of header soup (see current_thread_info()), this
528 * must be the first element of task_struct.
530 struct thread_info thread_info
;
532 /* -1 unrunnable, 0 runnable, >0 stopped: */
536 * This begins the randomizable portion of task_struct. Only
537 * scheduling-critical items should be added above here.
539 randomized_struct_fields_start
543 /* Per task flags (PF_*), defined further below: */
548 struct llist_node wake_entry
;
550 #ifdef CONFIG_THREAD_INFO_IN_TASK
554 unsigned int wakee_flips
;
555 unsigned long wakee_flip_decay_ts
;
556 struct task_struct
*last_wakee
;
559 * recent_used_cpu is initially set as the last CPU used by a task
560 * that wakes affine another task. Waker/wakee relationships can
561 * push tasks around a CPU where each wakeup moves to the next one.
562 * Tracking a recently used CPU allows a quick search for a recently
563 * used CPU that may be idle.
573 unsigned int rt_priority
;
575 const struct sched_class
*sched_class
;
576 struct sched_entity se
;
577 struct sched_rt_entity rt
;
578 #ifdef CONFIG_CGROUP_SCHED
579 struct task_group
*sched_task_group
;
581 struct sched_dl_entity dl
;
583 #ifdef CONFIG_PREEMPT_NOTIFIERS
584 /* List of struct preempt_notifier: */
585 struct hlist_head preempt_notifiers
;
588 #ifdef CONFIG_BLK_DEV_IO_TRACE
589 unsigned int btrace_seq
;
594 cpumask_t cpus_allowed
;
596 #ifdef CONFIG_PREEMPT_RCU
597 int rcu_read_lock_nesting
;
598 union rcu_special rcu_read_unlock_special
;
599 struct list_head rcu_node_entry
;
600 struct rcu_node
*rcu_blocked_node
;
601 #endif /* #ifdef CONFIG_PREEMPT_RCU */
603 #ifdef CONFIG_TASKS_RCU
604 unsigned long rcu_tasks_nvcsw
;
605 u8 rcu_tasks_holdout
;
607 int rcu_tasks_idle_cpu
;
608 struct list_head rcu_tasks_holdout_list
;
609 #endif /* #ifdef CONFIG_TASKS_RCU */
611 struct sched_info sched_info
;
613 struct list_head tasks
;
615 struct plist_node pushable_tasks
;
616 struct rb_node pushable_dl_tasks
;
619 struct mm_struct
*mm
;
620 struct mm_struct
*active_mm
;
622 /* Per-thread vma caching: */
623 struct vmacache vmacache
;
625 #ifdef SPLIT_RSS_COUNTING
626 struct task_rss_stat rss_stat
;
631 /* The signal sent when the parent dies: */
633 /* JOBCTL_*, siglock protected: */
634 unsigned long jobctl
;
636 /* Used for emulating ABI behavior of previous Linux versions: */
637 unsigned int personality
;
639 /* Scheduler bits, serialized by scheduler locks: */
640 unsigned sched_reset_on_fork
:1;
641 unsigned sched_contributes_to_load
:1;
642 unsigned sched_migrated
:1;
643 unsigned sched_remote_wakeup
:1;
644 /* Force alignment to the next boundary: */
647 /* Unserialized, strictly 'current' */
649 /* Bit to tell LSMs we're in execve(): */
650 unsigned in_execve
:1;
651 unsigned in_iowait
:1;
652 #ifndef TIF_RESTORE_SIGMASK
653 unsigned restore_sigmask
:1;
656 unsigned memcg_may_oom
:1;
658 unsigned memcg_kmem_skip_account
:1;
661 #ifdef CONFIG_COMPAT_BRK
662 unsigned brk_randomized
:1;
664 #ifdef CONFIG_CGROUPS
665 /* disallow userland-initiated cgroup migration */
666 unsigned no_cgroup_migration
:1;
669 unsigned long atomic_flags
; /* Flags requiring atomic access. */
671 struct restart_block restart_block
;
676 #ifdef CONFIG_CC_STACKPROTECTOR
677 /* Canary value for the -fstack-protector GCC feature: */
678 unsigned long stack_canary
;
681 * Pointers to the (original) parent process, youngest child, younger sibling,
682 * older sibling, respectively. (p->father can be replaced with
683 * p->real_parent->pid)
686 /* Real parent process: */
687 struct task_struct __rcu
*real_parent
;
689 /* Recipient of SIGCHLD, wait4() reports: */
690 struct task_struct __rcu
*parent
;
693 * Children/sibling form the list of natural children:
695 struct list_head children
;
696 struct list_head sibling
;
697 struct task_struct
*group_leader
;
700 * 'ptraced' is the list of tasks this task is using ptrace() on.
702 * This includes both natural children and PTRACE_ATTACH targets.
703 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
705 struct list_head ptraced
;
706 struct list_head ptrace_entry
;
708 /* PID/PID hash table linkage. */
709 struct pid_link pids
[PIDTYPE_MAX
];
710 struct list_head thread_group
;
711 struct list_head thread_node
;
713 struct completion
*vfork_done
;
715 /* CLONE_CHILD_SETTID: */
716 int __user
*set_child_tid
;
718 /* CLONE_CHILD_CLEARTID: */
719 int __user
*clear_child_tid
;
723 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
728 struct prev_cputime prev_cputime
;
729 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
733 #ifdef CONFIG_NO_HZ_FULL
734 atomic_t tick_dep_mask
;
736 /* Context switch counts: */
738 unsigned long nivcsw
;
740 /* Monotonic time in nsecs: */
743 /* Boot based time in nsecs: */
746 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
747 unsigned long min_flt
;
748 unsigned long maj_flt
;
750 #ifdef CONFIG_POSIX_TIMERS
751 struct task_cputime cputime_expires
;
752 struct list_head cpu_timers
[3];
755 /* Process credentials: */
757 /* Tracer's credentials at attach: */
758 const struct cred __rcu
*ptracer_cred
;
760 /* Objective and real subjective task credentials (COW): */
761 const struct cred __rcu
*real_cred
;
763 /* Effective (overridable) subjective task credentials (COW): */
764 const struct cred __rcu
*cred
;
767 * executable name, excluding path.
769 * - normally initialized setup_new_exec()
770 * - access it with [gs]et_task_comm()
771 * - lock it with task_lock()
773 char comm
[TASK_COMM_LEN
];
775 struct nameidata
*nameidata
;
777 #ifdef CONFIG_SYSVIPC
778 struct sysv_sem sysvsem
;
779 struct sysv_shm sysvshm
;
781 #ifdef CONFIG_DETECT_HUNG_TASK
782 unsigned long last_switch_count
;
784 /* Filesystem information: */
785 struct fs_struct
*fs
;
787 /* Open file information: */
788 struct files_struct
*files
;
791 struct nsproxy
*nsproxy
;
793 /* Signal handlers: */
794 struct signal_struct
*signal
;
795 struct sighand_struct
*sighand
;
797 sigset_t real_blocked
;
798 /* Restored if set_restore_sigmask() was used: */
799 sigset_t saved_sigmask
;
800 struct sigpending pending
;
801 unsigned long sas_ss_sp
;
803 unsigned int sas_ss_flags
;
805 struct callback_head
*task_works
;
807 struct audit_context
*audit_context
;
808 #ifdef CONFIG_AUDITSYSCALL
810 unsigned int sessionid
;
812 struct seccomp seccomp
;
814 /* Thread group tracking: */
818 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
819 spinlock_t alloc_lock
;
821 /* Protection of the PI data structures: */
822 raw_spinlock_t pi_lock
;
824 struct wake_q_node wake_q
;
826 #ifdef CONFIG_RT_MUTEXES
827 /* PI waiters blocked on a rt_mutex held by this task: */
828 struct rb_root_cached pi_waiters
;
829 /* Updated under owner's pi_lock and rq lock */
830 struct task_struct
*pi_top_task
;
831 /* Deadlock detection and priority inheritance handling: */
832 struct rt_mutex_waiter
*pi_blocked_on
;
835 #ifdef CONFIG_DEBUG_MUTEXES
836 /* Mutex deadlock detection: */
837 struct mutex_waiter
*blocked_on
;
840 #ifdef CONFIG_TRACE_IRQFLAGS
841 unsigned int irq_events
;
842 unsigned long hardirq_enable_ip
;
843 unsigned long hardirq_disable_ip
;
844 unsigned int hardirq_enable_event
;
845 unsigned int hardirq_disable_event
;
846 int hardirqs_enabled
;
848 unsigned long softirq_disable_ip
;
849 unsigned long softirq_enable_ip
;
850 unsigned int softirq_disable_event
;
851 unsigned int softirq_enable_event
;
852 int softirqs_enabled
;
856 #ifdef CONFIG_LOCKDEP
857 # define MAX_LOCK_DEPTH 48UL
860 unsigned int lockdep_recursion
;
861 struct held_lock held_locks
[MAX_LOCK_DEPTH
];
865 unsigned int in_ubsan
;
868 /* Journalling filesystem info: */
871 /* Stacked block device info: */
872 struct bio_list
*bio_list
;
875 /* Stack plugging: */
876 struct blk_plug
*plug
;
880 struct reclaim_state
*reclaim_state
;
882 struct backing_dev_info
*backing_dev_info
;
884 struct io_context
*io_context
;
887 unsigned long ptrace_message
;
888 siginfo_t
*last_siginfo
;
890 struct task_io_accounting ioac
;
891 #ifdef CONFIG_TASK_XACCT
892 /* Accumulated RSS usage: */
894 /* Accumulated virtual memory usage: */
896 /* stime + utime since last update: */
899 #ifdef CONFIG_CPUSETS
900 /* Protected by ->alloc_lock: */
901 nodemask_t mems_allowed
;
902 /* Seqence number to catch updates: */
903 seqcount_t mems_allowed_seq
;
904 int cpuset_mem_spread_rotor
;
905 int cpuset_slab_spread_rotor
;
907 #ifdef CONFIG_CGROUPS
908 /* Control Group info protected by css_set_lock: */
909 struct css_set __rcu
*cgroups
;
910 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
911 struct list_head cg_list
;
913 #ifdef CONFIG_INTEL_RDT
918 struct robust_list_head __user
*robust_list
;
920 struct compat_robust_list_head __user
*compat_robust_list
;
922 struct list_head pi_state_list
;
923 struct futex_pi_state
*pi_state_cache
;
925 #ifdef CONFIG_PERF_EVENTS
926 struct perf_event_context
*perf_event_ctxp
[perf_nr_task_contexts
];
927 struct mutex perf_event_mutex
;
928 struct list_head perf_event_list
;
930 #ifdef CONFIG_DEBUG_PREEMPT
931 unsigned long preempt_disable_ip
;
934 /* Protected by alloc_lock: */
935 struct mempolicy
*mempolicy
;
937 short pref_node_fork
;
939 #ifdef CONFIG_NUMA_BALANCING
941 unsigned int numa_scan_period
;
942 unsigned int numa_scan_period_max
;
943 int numa_preferred_nid
;
944 unsigned long numa_migrate_retry
;
945 /* Migration stamp: */
947 u64 last_task_numa_placement
;
948 u64 last_sum_exec_runtime
;
949 struct callback_head numa_work
;
951 struct list_head numa_entry
;
952 struct numa_group
*numa_group
;
955 * numa_faults is an array split into four regions:
956 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
957 * in this precise order.
959 * faults_memory: Exponential decaying average of faults on a per-node
960 * basis. Scheduling placement decisions are made based on these
961 * counts. The values remain static for the duration of a PTE scan.
962 * faults_cpu: Track the nodes the process was running on when a NUMA
963 * hinting fault was incurred.
964 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
965 * during the current scan window. When the scan completes, the counts
966 * in faults_memory and faults_cpu decay and these values are copied.
968 unsigned long *numa_faults
;
969 unsigned long total_numa_faults
;
972 * numa_faults_locality tracks if faults recorded during the last
973 * scan window were remote/local or failed to migrate. The task scan
974 * period is adapted based on the locality of the faults with different
975 * weights depending on whether they were shared or private faults
977 unsigned long numa_faults_locality
[3];
979 unsigned long numa_pages_migrated
;
980 #endif /* CONFIG_NUMA_BALANCING */
982 struct tlbflush_unmap_batch tlb_ubc
;
986 /* Cache last used pipe for splice(): */
987 struct pipe_inode_info
*splice_pipe
;
989 struct page_frag task_frag
;
991 #ifdef CONFIG_TASK_DELAY_ACCT
992 struct task_delay_info
*delays
;
995 #ifdef CONFIG_FAULT_INJECTION
997 unsigned int fail_nth
;
1000 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1001 * balance_dirty_pages() for a dirty throttling pause:
1004 int nr_dirtied_pause
;
1005 /* Start of a write-and-pause period: */
1006 unsigned long dirty_paused_when
;
1008 #ifdef CONFIG_LATENCYTOP
1009 int latency_record_count
;
1010 struct latency_record latency_record
[LT_SAVECOUNT
];
1013 * Time slack values; these are used to round up poll() and
1014 * select() etc timeout values. These are in nanoseconds.
1017 u64 default_timer_slack_ns
;
1020 unsigned int kasan_depth
;
1023 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1024 /* Index of current stored address in ret_stack: */
1027 /* Stack of return addresses for return function tracing: */
1028 struct ftrace_ret_stack
*ret_stack
;
1030 /* Timestamp for last schedule: */
1031 unsigned long long ftrace_timestamp
;
1034 * Number of functions that haven't been traced
1035 * because of depth overrun:
1037 atomic_t trace_overrun
;
1039 /* Pause tracing: */
1040 atomic_t tracing_graph_pause
;
1043 #ifdef CONFIG_TRACING
1044 /* State flags for use by tracers: */
1045 unsigned long trace
;
1047 /* Bitmask and counter of trace recursion: */
1048 unsigned long trace_recursion
;
1049 #endif /* CONFIG_TRACING */
1052 /* Coverage collection mode enabled for this task (0 if disabled): */
1053 enum kcov_mode kcov_mode
;
1055 /* Size of the kcov_area: */
1056 unsigned int kcov_size
;
1058 /* Buffer for coverage collection: */
1061 /* KCOV descriptor wired with this task or NULL: */
1066 struct mem_cgroup
*memcg_in_oom
;
1067 gfp_t memcg_oom_gfp_mask
;
1068 int memcg_oom_order
;
1070 /* Number of pages to reclaim on returning to userland: */
1071 unsigned int memcg_nr_pages_over_high
;
1074 #ifdef CONFIG_UPROBES
1075 struct uprobe_task
*utask
;
1077 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1078 unsigned int sequential_io
;
1079 unsigned int sequential_io_avg
;
1081 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1082 unsigned long task_state_change
;
1084 int pagefault_disabled
;
1086 struct task_struct
*oom_reaper_list
;
1088 #ifdef CONFIG_VMAP_STACK
1089 struct vm_struct
*stack_vm_area
;
1091 #ifdef CONFIG_THREAD_INFO_IN_TASK
1092 /* A live task holds one reference: */
1093 atomic_t stack_refcount
;
1095 #ifdef CONFIG_LIVEPATCH
1098 #ifdef CONFIG_SECURITY
1099 /* Used by LSM modules for access restriction: */
1104 * New fields for task_struct should be added above here, so that
1105 * they are included in the randomized portion of task_struct.
1107 randomized_struct_fields_end
1109 /* CPU-specific state of this task: */
1110 struct thread_struct thread
;
1113 * WARNING: on x86, 'thread_struct' contains a variable-sized
1114 * structure. It *MUST* be at the end of 'task_struct'.
1116 * Do not put anything below here!
1120 static inline struct pid
*task_pid(struct task_struct
*task
)
1122 return task
->pids
[PIDTYPE_PID
].pid
;
1125 static inline struct pid
*task_tgid(struct task_struct
*task
)
1127 return task
->group_leader
->pids
[PIDTYPE_PID
].pid
;
1131 * Without tasklist or RCU lock it is not safe to dereference
1132 * the result of task_pgrp/task_session even if task == current,
1133 * we can race with another thread doing sys_setsid/sys_setpgid.
1135 static inline struct pid
*task_pgrp(struct task_struct
*task
)
1137 return task
->group_leader
->pids
[PIDTYPE_PGID
].pid
;
1140 static inline struct pid
*task_session(struct task_struct
*task
)
1142 return task
->group_leader
->pids
[PIDTYPE_SID
].pid
;
1146 * the helpers to get the task's different pids as they are seen
1147 * from various namespaces
1149 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1150 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1152 * task_xid_nr_ns() : id seen from the ns specified;
1154 * see also pid_nr() etc in include/linux/pid.h
1156 pid_t
__task_pid_nr_ns(struct task_struct
*task
, enum pid_type type
, struct pid_namespace
*ns
);
1158 static inline pid_t
task_pid_nr(struct task_struct
*tsk
)
1163 static inline pid_t
task_pid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1165 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, ns
);
1168 static inline pid_t
task_pid_vnr(struct task_struct
*tsk
)
1170 return __task_pid_nr_ns(tsk
, PIDTYPE_PID
, NULL
);
1174 static inline pid_t
task_tgid_nr(struct task_struct
*tsk
)
1180 * pid_alive - check that a task structure is not stale
1181 * @p: Task structure to be checked.
1183 * Test if a process is not yet dead (at most zombie state)
1184 * If pid_alive fails, then pointers within the task structure
1185 * can be stale and must not be dereferenced.
1187 * Return: 1 if the process is alive. 0 otherwise.
1189 static inline int pid_alive(const struct task_struct
*p
)
1191 return p
->pids
[PIDTYPE_PID
].pid
!= NULL
;
1194 static inline pid_t
task_pgrp_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1196 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, ns
);
1199 static inline pid_t
task_pgrp_vnr(struct task_struct
*tsk
)
1201 return __task_pid_nr_ns(tsk
, PIDTYPE_PGID
, NULL
);
1205 static inline pid_t
task_session_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1207 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, ns
);
1210 static inline pid_t
task_session_vnr(struct task_struct
*tsk
)
1212 return __task_pid_nr_ns(tsk
, PIDTYPE_SID
, NULL
);
1215 static inline pid_t
task_tgid_nr_ns(struct task_struct
*tsk
, struct pid_namespace
*ns
)
1217 return __task_pid_nr_ns(tsk
, __PIDTYPE_TGID
, ns
);
1220 static inline pid_t
task_tgid_vnr(struct task_struct
*tsk
)
1222 return __task_pid_nr_ns(tsk
, __PIDTYPE_TGID
, NULL
);
1225 static inline pid_t
task_ppid_nr_ns(const struct task_struct
*tsk
, struct pid_namespace
*ns
)
1231 pid
= task_tgid_nr_ns(rcu_dereference(tsk
->real_parent
), ns
);
1237 static inline pid_t
task_ppid_nr(const struct task_struct
*tsk
)
1239 return task_ppid_nr_ns(tsk
, &init_pid_ns
);
1242 /* Obsolete, do not use: */
1243 static inline pid_t
task_pgrp_nr(struct task_struct
*tsk
)
1245 return task_pgrp_nr_ns(tsk
, &init_pid_ns
);
1248 #define TASK_REPORT_IDLE (TASK_REPORT + 1)
1249 #define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1251 static inline unsigned int task_state_index(struct task_struct
*tsk
)
1253 unsigned int tsk_state
= READ_ONCE(tsk
->state
);
1254 unsigned int state
= (tsk_state
| tsk
->exit_state
) & TASK_REPORT
;
1256 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX
);
1258 if (tsk_state
== TASK_IDLE
)
1259 state
= TASK_REPORT_IDLE
;
1264 static inline char task_index_to_char(unsigned int state
)
1266 static const char state_char
[] = "RSDTtXZPI";
1268 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX
) != sizeof(state_char
) - 1);
1270 return state_char
[state
];
1273 static inline char task_state_to_char(struct task_struct
*tsk
)
1275 return task_index_to_char(task_state_index(tsk
));
1279 * is_global_init - check if a task structure is init. Since init
1280 * is free to have sub-threads we need to check tgid.
1281 * @tsk: Task structure to be checked.
1283 * Check if a task structure is the first user space task the kernel created.
1285 * Return: 1 if the task structure is init. 0 otherwise.
1287 static inline int is_global_init(struct task_struct
*tsk
)
1289 return task_tgid_nr(tsk
) == 1;
1292 extern struct pid
*cad_pid
;
1297 #define PF_IDLE 0x00000002 /* I am an IDLE thread */
1298 #define PF_EXITING 0x00000004 /* Getting shut down */
1299 #define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1300 #define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1301 #define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1302 #define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1303 #define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1304 #define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1305 #define PF_DUMPCORE 0x00000200 /* Dumped core */
1306 #define PF_SIGNALED 0x00000400 /* Killed by a signal */
1307 #define PF_MEMALLOC 0x00000800 /* Allocating memory */
1308 #define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1309 #define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1310 #define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1311 #define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1312 #define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1313 #define PF_KSWAPD 0x00020000 /* I am kswapd */
1314 #define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1315 #define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1316 #define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1317 #define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1318 #define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1319 #define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1320 #define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1321 #define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1322 #define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1323 #define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1324 #define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1327 * Only the _current_ task can read/write to tsk->flags, but other
1328 * tasks can access tsk->flags in readonly mode for example
1329 * with tsk_used_math (like during threaded core dumping).
1330 * There is however an exception to this rule during ptrace
1331 * or during fork: the ptracer task is allowed to write to the
1332 * child->flags of its traced child (same goes for fork, the parent
1333 * can write to the child->flags), because we're guaranteed the
1334 * child is not running and in turn not changing child->flags
1335 * at the same time the parent does it.
1337 #define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1338 #define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1339 #define clear_used_math() clear_stopped_child_used_math(current)
1340 #define set_used_math() set_stopped_child_used_math(current)
1342 #define conditional_stopped_child_used_math(condition, child) \
1343 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1345 #define conditional_used_math(condition) 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)
1350 /* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1351 #define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1352 #define used_math() tsk_used_math(current)
1354 static inline bool is_percpu_thread(void)
1357 return (current
->flags
& PF_NO_SETAFFINITY
) &&
1358 (current
->nr_cpus_allowed
== 1);
1364 /* Per-process atomic flags. */
1365 #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1366 #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1367 #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1370 #define TASK_PFA_TEST(name, func) \
1371 static inline bool task_##func(struct task_struct *p) \
1372 { return test_bit(PFA_##name, &p->atomic_flags); }
1374 #define TASK_PFA_SET(name, func) \
1375 static inline void task_set_##func(struct task_struct *p) \
1376 { set_bit(PFA_##name, &p->atomic_flags); }
1378 #define TASK_PFA_CLEAR(name, func) \
1379 static inline void task_clear_##func(struct task_struct *p) \
1380 { clear_bit(PFA_##name, &p->atomic_flags); }
1382 TASK_PFA_TEST(NO_NEW_PRIVS
, no_new_privs
)
1383 TASK_PFA_SET(NO_NEW_PRIVS
, no_new_privs
)
1385 TASK_PFA_TEST(SPREAD_PAGE
, spread_page
)
1386 TASK_PFA_SET(SPREAD_PAGE
, spread_page
)
1387 TASK_PFA_CLEAR(SPREAD_PAGE
, spread_page
)
1389 TASK_PFA_TEST(SPREAD_SLAB
, spread_slab
)
1390 TASK_PFA_SET(SPREAD_SLAB
, spread_slab
)
1391 TASK_PFA_CLEAR(SPREAD_SLAB
, spread_slab
)
1394 current_restore_flags(unsigned long orig_flags
, unsigned long flags
)
1396 current
->flags
&= ~flags
;
1397 current
->flags
|= orig_flags
& flags
;
1400 extern int cpuset_cpumask_can_shrink(const struct cpumask
*cur
, const struct cpumask
*trial
);
1401 extern int task_can_attach(struct task_struct
*p
, const struct cpumask
*cs_cpus_allowed
);
1403 extern void do_set_cpus_allowed(struct task_struct
*p
, const struct cpumask
*new_mask
);
1404 extern int set_cpus_allowed_ptr(struct task_struct
*p
, const struct cpumask
*new_mask
);
1406 static inline void do_set_cpus_allowed(struct task_struct
*p
, const struct cpumask
*new_mask
)
1409 static inline int set_cpus_allowed_ptr(struct task_struct
*p
, const struct cpumask
*new_mask
)
1411 if (!cpumask_test_cpu(0, new_mask
))
1417 #ifndef cpu_relax_yield
1418 #define cpu_relax_yield() cpu_relax()
1421 extern int yield_to(struct task_struct
*p
, bool preempt
);
1422 extern void set_user_nice(struct task_struct
*p
, long nice
);
1423 extern int task_prio(const struct task_struct
*p
);
1426 * task_nice - return the nice value of a given task.
1427 * @p: the task in question.
1429 * Return: The nice value [ -20 ... 0 ... 19 ].
1431 static inline int task_nice(const struct task_struct
*p
)
1433 return PRIO_TO_NICE((p
)->static_prio
);
1436 extern int can_nice(const struct task_struct
*p
, const int nice
);
1437 extern int task_curr(const struct task_struct
*p
);
1438 extern int idle_cpu(int cpu
);
1439 extern int sched_setscheduler(struct task_struct
*, int, const struct sched_param
*);
1440 extern int sched_setscheduler_nocheck(struct task_struct
*, int, const struct sched_param
*);
1441 extern int sched_setattr(struct task_struct
*, const struct sched_attr
*);
1442 extern int sched_setattr_nocheck(struct task_struct
*, const struct sched_attr
*);
1443 extern struct task_struct
*idle_task(int cpu
);
1446 * is_idle_task - is the specified task an idle task?
1447 * @p: the task in question.
1449 * Return: 1 if @p is an idle task. 0 otherwise.
1451 static inline bool is_idle_task(const struct task_struct
*p
)
1453 return !!(p
->flags
& PF_IDLE
);
1456 extern struct task_struct
*curr_task(int cpu
);
1457 extern void ia64_set_curr_task(int cpu
, struct task_struct
*p
);
1461 union thread_union
{
1462 #ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1463 struct task_struct task
;
1465 #ifndef CONFIG_THREAD_INFO_IN_TASK
1466 struct thread_info thread_info
;
1468 unsigned long stack
[THREAD_SIZE
/sizeof(long)];
1471 #ifndef CONFIG_THREAD_INFO_IN_TASK
1472 extern struct thread_info init_thread_info
;
1475 extern unsigned long init_stack
[THREAD_SIZE
/ sizeof(unsigned long)];
1477 #ifdef CONFIG_THREAD_INFO_IN_TASK
1478 static inline struct thread_info
*task_thread_info(struct task_struct
*task
)
1480 return &task
->thread_info
;
1482 #elif !defined(__HAVE_THREAD_FUNCTIONS)
1483 # define task_thread_info(task) ((struct thread_info *)(task)->stack)
1487 * find a task by one of its numerical ids
1489 * find_task_by_pid_ns():
1490 * finds a task by its pid in the specified namespace
1491 * find_task_by_vpid():
1492 * finds a task by its virtual pid
1494 * see also find_vpid() etc in include/linux/pid.h
1497 extern struct task_struct
*find_task_by_vpid(pid_t nr
);
1498 extern struct task_struct
*find_task_by_pid_ns(pid_t nr
, struct pid_namespace
*ns
);
1501 * find a task by its virtual pid and get the task struct
1503 extern struct task_struct
*find_get_task_by_vpid(pid_t nr
);
1505 extern int wake_up_state(struct task_struct
*tsk
, unsigned int state
);
1506 extern int wake_up_process(struct task_struct
*tsk
);
1507 extern void wake_up_new_task(struct task_struct
*tsk
);
1510 extern void kick_process(struct task_struct
*tsk
);
1512 static inline void kick_process(struct task_struct
*tsk
) { }
1515 extern void __set_task_comm(struct task_struct
*tsk
, const char *from
, bool exec
);
1517 static inline void set_task_comm(struct task_struct
*tsk
, const char *from
)
1519 __set_task_comm(tsk
, from
, false);
1522 extern char *__get_task_comm(char *to
, size_t len
, struct task_struct
*tsk
);
1523 #define get_task_comm(buf, tsk) ({ \
1524 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1525 __get_task_comm(buf, sizeof(buf), tsk); \
1529 void scheduler_ipi(void);
1530 extern unsigned long wait_task_inactive(struct task_struct
*, long match_state
);
1532 static inline void scheduler_ipi(void) { }
1533 static inline unsigned long wait_task_inactive(struct task_struct
*p
, long match_state
)
1540 * Set thread flags in other task's structures.
1541 * See asm/thread_info.h for TIF_xxxx flags available:
1543 static inline void set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1545 set_ti_thread_flag(task_thread_info(tsk
), flag
);
1548 static inline void clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1550 clear_ti_thread_flag(task_thread_info(tsk
), flag
);
1553 static inline int test_and_set_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1555 return test_and_set_ti_thread_flag(task_thread_info(tsk
), flag
);
1558 static inline int test_and_clear_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1560 return test_and_clear_ti_thread_flag(task_thread_info(tsk
), flag
);
1563 static inline int test_tsk_thread_flag(struct task_struct
*tsk
, int flag
)
1565 return test_ti_thread_flag(task_thread_info(tsk
), flag
);
1568 static inline void set_tsk_need_resched(struct task_struct
*tsk
)
1570 set_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
1573 static inline void clear_tsk_need_resched(struct task_struct
*tsk
)
1575 clear_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
);
1578 static inline int test_tsk_need_resched(struct task_struct
*tsk
)
1580 return unlikely(test_tsk_thread_flag(tsk
,TIF_NEED_RESCHED
));
1584 * cond_resched() and cond_resched_lock(): latency reduction via
1585 * explicit rescheduling in places that are safe. The return
1586 * value indicates whether a reschedule was done in fact.
1587 * cond_resched_lock() will drop the spinlock before scheduling,
1588 * cond_resched_softirq() will enable bhs before scheduling.
1590 #ifndef CONFIG_PREEMPT
1591 extern int _cond_resched(void);
1593 static inline int _cond_resched(void) { return 0; }
1596 #define cond_resched() ({ \
1597 ___might_sleep(__FILE__, __LINE__, 0); \
1601 extern int __cond_resched_lock(spinlock_t
*lock
);
1603 #define cond_resched_lock(lock) ({ \
1604 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1605 __cond_resched_lock(lock); \
1608 extern int __cond_resched_softirq(void);
1610 #define cond_resched_softirq() ({ \
1611 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1612 __cond_resched_softirq(); \
1615 static inline void cond_resched_rcu(void)
1617 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1625 * Does a critical section need to be broken due to another
1626 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1627 * but a general need for low latency)
1629 static inline int spin_needbreak(spinlock_t
*lock
)
1631 #ifdef CONFIG_PREEMPT
1632 return spin_is_contended(lock
);
1638 static __always_inline
bool need_resched(void)
1640 return unlikely(tif_need_resched());
1644 * Wrappers for p->thread_info->cpu access. No-op on UP.
1648 static inline unsigned int task_cpu(const struct task_struct
*p
)
1650 #ifdef CONFIG_THREAD_INFO_IN_TASK
1653 return task_thread_info(p
)->cpu
;
1657 extern void set_task_cpu(struct task_struct
*p
, unsigned int cpu
);
1661 static inline unsigned int task_cpu(const struct task_struct
*p
)
1666 static inline void set_task_cpu(struct task_struct
*p
, unsigned int cpu
)
1670 #endif /* CONFIG_SMP */
1673 * In order to reduce various lock holder preemption latencies provide an
1674 * interface to see if a vCPU is currently running or not.
1676 * This allows us to terminate optimistic spin loops and block, analogous to
1677 * the native optimistic spin heuristic of testing if the lock owner task is
1680 #ifndef vcpu_is_preempted
1681 # define vcpu_is_preempted(cpu) false
1684 extern long sched_setaffinity(pid_t pid
, const struct cpumask
*new_mask
);
1685 extern long sched_getaffinity(pid_t pid
, struct cpumask
*mask
);
1687 #ifndef TASK_SIZE_OF
1688 #define TASK_SIZE_OF(tsk) TASK_SIZE