1#ifndef _LINUX_SCHED_H
2#define _LINUX_SCHED_H
3
4/*
5 * Define 'struct task_struct' and provide the main scheduler
6 * APIs (schedule(), wakeup variants, etc.)
7 */
8
9#include <uapi/linux/sched.h>
10
11#include <asm/current.h>
12
13#include <linux/pid.h>
14#include <linux/sem.h>
15#include <linux/shm.h>
16#include <linux/kcov.h>
17#include <linux/mutex.h>
18#include <linux/plist.h>
19#include <linux/hrtimer.h>
20#include <linux/seccomp.h>
21#include <linux/nodemask.h>
22#include <linux/rcupdate.h>
23#include <linux/resource.h>
24#include <linux/latencytop.h>
25#include <linux/sched/prio.h>
26#include <linux/signal_types.h>
27#include <linux/mm_types_task.h>
28#include <linux/task_io_accounting.h>
29
30/* task_struct member predeclarations (sorted alphabetically): */
31struct audit_context;
32struct backing_dev_info;
33struct bio_list;
34struct blk_plug;
35struct cfs_rq;
36struct fs_struct;
37struct futex_pi_state;
38struct io_context;
39struct mempolicy;
40struct nameidata;
41struct nsproxy;
42struct perf_event_context;
43struct pid_namespace;
44struct pipe_inode_info;
45struct rcu_node;
46struct reclaim_state;
47struct robust_list_head;
48struct sched_attr;
49struct sched_param;
50struct seq_file;
51struct sighand_struct;
52struct signal_struct;
53struct task_delay_info;
54struct task_group;
55
56/*
57 * Task state bitmask. NOTE! These bits are also
58 * encoded in fs/proc/array.c: get_task_state().
59 *
60 * We have two separate sets of flags: task->state
61 * is about runnability, while task->exit_state are
62 * about the task exiting. Confusing, but this way
63 * modifying one set can't modify the other one by
64 * mistake.
65 */
66
67/* Used in tsk->state: */
68#define TASK_RUNNING 0
69#define TASK_INTERRUPTIBLE 1
70#define TASK_UNINTERRUPTIBLE 2
71#define __TASK_STOPPED 4
72#define __TASK_TRACED 8
73/* Used in tsk->exit_state: */
74#define EXIT_DEAD 16
75#define EXIT_ZOMBIE 32
76#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
77/* Used in tsk->state again: */
78#define TASK_DEAD 64
79#define TASK_WAKEKILL 128
80#define TASK_WAKING 256
81#define TASK_PARKED 512
82#define TASK_NOLOAD 1024
83#define TASK_NEW 2048
84#define TASK_STATE_MAX 4096
85
86#define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWPNn"
87
88/* Convenience macros for the sake of set_current_state: */
89#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
90#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
91#define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
92
93#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
94
95/* Convenience macros for the sake of wake_up(): */
96#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
97#define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
98
99/* get_task_state(): */
100#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
101 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
102 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
103
104#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
105
106#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
107
108#define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
109
110#define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
111 (task->flags & PF_FROZEN) == 0 && \
112 (task->state & TASK_NOLOAD) == 0)
113
114#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
115
116#define __set_current_state(state_value) \
117 do { \
118 current->task_state_change = _THIS_IP_; \
119 current->state = (state_value); \
120 } while (0)
121#define set_current_state(state_value) \
122 do { \
123 current->task_state_change = _THIS_IP_; \
124 smp_store_mb(current->state, (state_value)); \
125 } while (0)
126
127#else
128/*
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
131 * actually sleep:
132 *
133 * for (;;) {
134 * set_current_state(TASK_UNINTERRUPTIBLE);
135 * if (!need_sleep)
136 * break;
137 *
138 * schedule();
139 * }
140 * __set_current_state(TASK_RUNNING);
141 *
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().
145 *
146 * The above is typically ordered against the wakeup, which does:
147 *
148 * need_sleep = false;
149 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
150 *
151 * Where wake_up_state() (and all other wakeup primitives) imply enough
152 * barriers to order the store of the variable against wakeup.
153 *
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).
157 *
158 * This is obviously fine, since they both store the exact same value.
159 *
160 * Also see the comments of try_to_wake_up().
161 */
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))
164#endif
165
166/* Task command name length: */
167#define TASK_COMM_LEN 16
168
169extern cpumask_var_t cpu_isolated_map;
170
171extern void scheduler_tick(void);
172
173#define MAX_SCHEDULE_TIMEOUT LONG_MAX
174
175extern long schedule_timeout(long timeout);
176extern long schedule_timeout_interruptible(long timeout);
177extern long schedule_timeout_killable(long timeout);
178extern long schedule_timeout_uninterruptible(long timeout);
179extern long schedule_timeout_idle(long timeout);
180asmlinkage void schedule(void);
181extern void schedule_preempt_disabled(void);
182
183extern int __must_check io_schedule_prepare(void);
184extern void io_schedule_finish(int token);
185extern long io_schedule_timeout(long timeout);
186extern void io_schedule(void);
187
188/**
189 * struct prev_cputime - snapshot of system and user cputime
190 * @utime: time spent in user mode
191 * @stime: time spent in system mode
192 * @lock: protects the above two fields
193 *
194 * Stores previous user/system time values such that we can guarantee
195 * monotonicity.
196 */
197struct prev_cputime {
198#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
199 u64 utime;
200 u64 stime;
201 raw_spinlock_t lock;
202#endif
203};
204
205/**
206 * struct task_cputime - collected CPU time counts
207 * @utime: time spent in user mode, in nanoseconds
208 * @stime: time spent in kernel mode, in nanoseconds
209 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
210 *
211 * This structure groups together three kinds of CPU time that are tracked for
212 * threads and thread groups. Most things considering CPU time want to group
213 * these counts together and treat all three of them in parallel.
214 */
215struct task_cputime {
216 u64 utime;
217 u64 stime;
218 unsigned long long sum_exec_runtime;
219};
220
221/* Alternate field names when used on cache expirations: */
222#define virt_exp utime
223#define prof_exp stime
224#define sched_exp sum_exec_runtime
225
226enum vtime_state {
227 /* Task is sleeping or running in a CPU with VTIME inactive: */
228 VTIME_INACTIVE = 0,
229 /* Task runs in userspace in a CPU with VTIME active: */
230 VTIME_USER,
231 /* Task runs in kernelspace in a CPU with VTIME active: */
232 VTIME_SYS,
233};
234
235struct vtime {
236 seqcount_t seqcount;
237 unsigned long long starttime;
238 enum vtime_state state;
239 u64 utime;
240 u64 stime;
241 u64 gtime;
242};
243
244struct sched_info {
245#ifdef CONFIG_SCHED_INFO
246 /* Cumulative counters: */
247
248 /* # of times we have run on this CPU: */
249 unsigned long pcount;
250
251 /* Time spent waiting on a runqueue: */
252 unsigned long long run_delay;
253
254 /* Timestamps: */
255
256 /* When did we last run on a CPU? */
257 unsigned long long last_arrival;
258
259 /* When were we last queued to run? */
260 unsigned long long last_queued;
261
262#endif /* CONFIG_SCHED_INFO */
263};
264
265/*
266 * Integer metrics need fixed point arithmetic, e.g., sched/fair
267 * has a few: load, load_avg, util_avg, freq, and capacity.
268 *
269 * We define a basic fixed point arithmetic range, and then formalize
270 * all these metrics based on that basic range.
271 */
272# define SCHED_FIXEDPOINT_SHIFT 10
273# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
274
275struct load_weight {
276 unsigned long weight;
277 u32 inv_weight;
278};
279
280/*
281 * The load_avg/util_avg accumulates an infinite geometric series
282 * (see __update_load_avg() in kernel/sched/fair.c).
283 *
284 * [load_avg definition]
285 *
286 * load_avg = runnable% * scale_load_down(load)
287 *
288 * where runnable% is the time ratio that a sched_entity is runnable.
289 * For cfs_rq, it is the aggregated load_avg of all runnable and
290 * blocked sched_entities.
291 *
292 * load_avg may also take frequency scaling into account:
293 *
294 * load_avg = runnable% * scale_load_down(load) * freq%
295 *
296 * where freq% is the CPU frequency normalized to the highest frequency.
297 *
298 * [util_avg definition]
299 *
300 * util_avg = running% * SCHED_CAPACITY_SCALE
301 *
302 * where running% is the time ratio that a sched_entity is running on
303 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
304 * and blocked sched_entities.
305 *
306 * util_avg may also factor frequency scaling and CPU capacity scaling:
307 *
308 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
309 *
310 * where freq% is the same as above, and capacity% is the CPU capacity
311 * normalized to the greatest capacity (due to uarch differences, etc).
312 *
313 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
314 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
315 * we therefore scale them to as large a range as necessary. This is for
316 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
317 *
318 * [Overflow issue]
319 *
320 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
321 * with the highest load (=88761), always runnable on a single cfs_rq,
322 * and should not overflow as the number already hits PID_MAX_LIMIT.
323 *
324 * For all other cases (including 32-bit kernels), struct load_weight's
325 * weight will overflow first before we do, because:
326 *
327 * Max(load_avg) <= Max(load.weight)
328 *
329 * Then it is the load_weight's responsibility to consider overflow
330 * issues.
331 */
332struct sched_avg {
333 u64 last_update_time;
334 u64 load_sum;
335 u32 util_sum;
336 u32 period_contrib;
337 unsigned long load_avg;
338 unsigned long util_avg;
339};
340
341struct sched_statistics {
342#ifdef CONFIG_SCHEDSTATS
343 u64 wait_start;
344 u64 wait_max;
345 u64 wait_count;
346 u64 wait_sum;
347 u64 iowait_count;
348 u64 iowait_sum;
349
350 u64 sleep_start;
351 u64 sleep_max;
352 s64 sum_sleep_runtime;
353
354 u64 block_start;
355 u64 block_max;
356 u64 exec_max;
357 u64 slice_max;
358
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;
364
365 u64 nr_wakeups;
366 u64 nr_wakeups_sync;
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;
373 u64 nr_wakeups_idle;
374#endif
375};
376
377struct sched_entity {
378 /* For load-balancing: */
379 struct load_weight load;
380 struct rb_node run_node;
381 struct list_head group_node;
382 unsigned int on_rq;
383
384 u64 exec_start;
385 u64 sum_exec_runtime;
386 u64 vruntime;
387 u64 prev_sum_exec_runtime;
388
389 u64 nr_migrations;
390
391 struct sched_statistics statistics;
392
393#ifdef CONFIG_FAIR_GROUP_SCHED
394 int depth;
395 struct sched_entity *parent;
396 /* rq on which this entity is (to be) queued: */
397 struct cfs_rq *cfs_rq;
398 /* rq "owned" by this entity/group: */
399 struct cfs_rq *my_q;
400#endif
401
402#ifdef CONFIG_SMP
403 /*
404 * Per entity load average tracking.
405 *
406 * Put into separate cache line so it does not
407 * collide with read-mostly values above.
408 */
409 struct sched_avg avg ____cacheline_aligned_in_smp;
410#endif
411};
412
413struct sched_rt_entity {
414 struct list_head run_list;
415 unsigned long timeout;
416 unsigned long watchdog_stamp;
417 unsigned int time_slice;
418 unsigned short on_rq;
419 unsigned short on_list;
420
421 struct sched_rt_entity *back;
422#ifdef CONFIG_RT_GROUP_SCHED
423 struct sched_rt_entity *parent;
424 /* rq on which this entity is (to be) queued: */
425 struct rt_rq *rt_rq;
426 /* rq "owned" by this entity/group: */
427 struct rt_rq *my_q;
428#endif
429} __randomize_layout;
430
431struct sched_dl_entity {
432 struct rb_node rb_node;
433
434 /*
435 * Original scheduling parameters. Copied here from sched_attr
436 * during sched_setattr(), they will remain the same until
437 * the next sched_setattr().
438 */
439 u64 dl_runtime; /* Maximum runtime for each instance */
440 u64 dl_deadline; /* Relative deadline of each instance */
441 u64 dl_period; /* Separation of two instances (period) */
442 u64 dl_bw; /* dl_runtime / dl_period */
443 u64 dl_density; /* dl_runtime / dl_deadline */
444
445 /*
446 * Actual scheduling parameters. Initialized with the values above,
447 * they are continously updated during task execution. Note that
448 * the remaining runtime could be < 0 in case we are in overrun.
449 */
450 s64 runtime; /* Remaining runtime for this instance */
451 u64 deadline; /* Absolute deadline for this instance */
452 unsigned int flags; /* Specifying the scheduler behaviour */
453
454 /*
455 * Some bool flags:
456 *
457 * @dl_throttled tells if we exhausted the runtime. If so, the
458 * task has to wait for a replenishment to be performed at the
459 * next firing of dl_timer.
460 *
461 * @dl_boosted tells if we are boosted due to DI. If so we are
462 * outside bandwidth enforcement mechanism (but only until we
463 * exit the critical section);
464 *
465 * @dl_yielded tells if task gave up the CPU before consuming
466 * all its available runtime during the last job.
467 *
468 * @dl_non_contending tells if the task is inactive while still
469 * contributing to the active utilization. In other words, it
470 * indicates if the inactive timer has been armed and its handler
471 * has not been executed yet. This flag is useful to avoid race
472 * conditions between the inactive timer handler and the wakeup
473 * code.
474 */
475 int dl_throttled;
476 int dl_boosted;
477 int dl_yielded;
478 int dl_non_contending;
479
480 /*
481 * Bandwidth enforcement timer. Each -deadline task has its
482 * own bandwidth to be enforced, thus we need one timer per task.
483 */
484 struct hrtimer dl_timer;
485
486 /*
487 * Inactive timer, responsible for decreasing the active utilization
488 * at the "0-lag time". When a -deadline task blocks, it contributes
489 * to GRUB's active utilization until the "0-lag time", hence a
490 * timer is needed to decrease the active utilization at the correct
491 * time.
492 */
493 struct hrtimer inactive_timer;
494};
495
496union rcu_special {
497 struct {
498 u8 blocked;
499 u8 need_qs;
500 u8 exp_need_qs;
501
502 /* Otherwise the compiler can store garbage here: */
503 u8 pad;
504 } b; /* Bits. */
505 u32 s; /* Set of bits. */
506};
507
508enum perf_event_task_context {
509 perf_invalid_context = -1,
510 perf_hw_context = 0,
511 perf_sw_context,
512 perf_nr_task_contexts,
513};
514
515struct wake_q_node {
516 struct wake_q_node *next;
517};
518
519struct task_struct {
520#ifdef CONFIG_THREAD_INFO_IN_TASK
521 /*
522 * For reasons of header soup (see current_thread_info()), this
523 * must be the first element of task_struct.
524 */
525 struct thread_info thread_info;
526#endif
527 /* -1 unrunnable, 0 runnable, >0 stopped: */
528 volatile long state;
529
530 /*
531 * This begins the randomizable portion of task_struct. Only
532 * scheduling-critical items should be added above here.
533 */
534 randomized_struct_fields_start
535
536 void *stack;
537 atomic_t usage;
538 /* Per task flags (PF_*), defined further below: */
539 unsigned int flags;
540 unsigned int ptrace;
541
542#ifdef CONFIG_SMP
543 struct llist_node wake_entry;
544 int on_cpu;
545#ifdef CONFIG_THREAD_INFO_IN_TASK
546 /* Current CPU: */
547 unsigned int cpu;
548#endif
549 unsigned int wakee_flips;
550 unsigned long wakee_flip_decay_ts;
551 struct task_struct *last_wakee;
552
553 int wake_cpu;
554#endif
555 int on_rq;
556
557 int prio;
558 int static_prio;
559 int normal_prio;
560 unsigned int rt_priority;
561
562 const struct sched_class *sched_class;
563 struct sched_entity se;
564 struct sched_rt_entity rt;
565#ifdef CONFIG_CGROUP_SCHED
566 struct task_group *sched_task_group;
567#endif
568 struct sched_dl_entity dl;
569
570#ifdef CONFIG_PREEMPT_NOTIFIERS
571 /* List of struct preempt_notifier: */
572 struct hlist_head preempt_notifiers;
573#endif
574
575#ifdef CONFIG_BLK_DEV_IO_TRACE
576 unsigned int btrace_seq;
577#endif
578
579 unsigned int policy;
580 int nr_cpus_allowed;
581 cpumask_t cpus_allowed;
582
583#ifdef CONFIG_PREEMPT_RCU
584 int rcu_read_lock_nesting;
585 union rcu_special rcu_read_unlock_special;
586 struct list_head rcu_node_entry;
587 struct rcu_node *rcu_blocked_node;
588#endif /* #ifdef CONFIG_PREEMPT_RCU */
589
590#ifdef CONFIG_TASKS_RCU
591 unsigned long rcu_tasks_nvcsw;
592 bool rcu_tasks_holdout;
593 struct list_head rcu_tasks_holdout_list;
594 int rcu_tasks_idle_cpu;
595#endif /* #ifdef CONFIG_TASKS_RCU */
596
597 struct sched_info sched_info;
598
599 struct list_head tasks;
600#ifdef CONFIG_SMP
601 struct plist_node pushable_tasks;
602 struct rb_node pushable_dl_tasks;
603#endif
604
605 struct mm_struct *mm;
606 struct mm_struct *active_mm;
607
608 /* Per-thread vma caching: */
609 struct vmacache vmacache;
610
611#ifdef SPLIT_RSS_COUNTING
612 struct task_rss_stat rss_stat;
613#endif
614 int exit_state;
615 int exit_code;
616 int exit_signal;
617 /* The signal sent when the parent dies: */
618 int pdeath_signal;
619 /* JOBCTL_*, siglock protected: */
620 unsigned long jobctl;
621
622 /* Used for emulating ABI behavior of previous Linux versions: */
623 unsigned int personality;
624
625 /* Scheduler bits, serialized by scheduler locks: */
626 unsigned sched_reset_on_fork:1;
627 unsigned sched_contributes_to_load:1;
628 unsigned sched_migrated:1;
629 unsigned sched_remote_wakeup:1;
630 /* Force alignment to the next boundary: */
631 unsigned :0;
632
633 /* Unserialized, strictly 'current' */
634
635 /* Bit to tell LSMs we're in execve(): */
636 unsigned in_execve:1;
637 unsigned in_iowait:1;
638#ifndef TIF_RESTORE_SIGMASK
639 unsigned restore_sigmask:1;
640#endif
641#ifdef CONFIG_MEMCG
642 unsigned memcg_may_oom:1;
643#ifndef CONFIG_SLOB
644 unsigned memcg_kmem_skip_account:1;
645#endif
646#endif
647#ifdef CONFIG_COMPAT_BRK
648 unsigned brk_randomized:1;
649#endif
650#ifdef CONFIG_CGROUPS
651 /* disallow userland-initiated cgroup migration */
652 unsigned no_cgroup_migration:1;
653#endif
654
655 unsigned long atomic_flags; /* Flags requiring atomic access. */
656
657 struct restart_block restart_block;
658
659 pid_t pid;
660 pid_t tgid;
661
662#ifdef CONFIG_CC_STACKPROTECTOR
663 /* Canary value for the -fstack-protector GCC feature: */
664 unsigned long stack_canary;
665#endif
666 /*
667 * Pointers to the (original) parent process, youngest child, younger sibling,
668 * older sibling, respectively. (p->father can be replaced with
669 * p->real_parent->pid)
670 */
671
672 /* Real parent process: */
673 struct task_struct __rcu *real_parent;
674
675 /* Recipient of SIGCHLD, wait4() reports: */
676 struct task_struct __rcu *parent;
677
678 /*
679 * Children/sibling form the list of natural children:
680 */
681 struct list_head children;
682 struct list_head sibling;
683 struct task_struct *group_leader;
684
685 /*
686 * 'ptraced' is the list of tasks this task is using ptrace() on.
687 *
688 * This includes both natural children and PTRACE_ATTACH targets.
689 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
690 */
691 struct list_head ptraced;
692 struct list_head ptrace_entry;
693
694 /* PID/PID hash table linkage. */
695 struct pid_link pids[PIDTYPE_MAX];
696 struct list_head thread_group;
697 struct list_head thread_node;
698
699 struct completion *vfork_done;
700
701 /* CLONE_CHILD_SETTID: */
702 int __user *set_child_tid;
703
704 /* CLONE_CHILD_CLEARTID: */
705 int __user *clear_child_tid;
706
707 u64 utime;
708 u64 stime;
709#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
710 u64 utimescaled;
711 u64 stimescaled;
712#endif
713 u64 gtime;
714 struct prev_cputime prev_cputime;
715#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
716 struct vtime vtime;
717#endif
718
719#ifdef CONFIG_NO_HZ_FULL
720 atomic_t tick_dep_mask;
721#endif
722 /* Context switch counts: */
723 unsigned long nvcsw;
724 unsigned long nivcsw;
725
726 /* Monotonic time in nsecs: */
727 u64 start_time;
728
729 /* Boot based time in nsecs: */
730 u64 real_start_time;
731
732 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
733 unsigned long min_flt;
734 unsigned long maj_flt;
735
736#ifdef CONFIG_POSIX_TIMERS
737 struct task_cputime cputime_expires;
738 struct list_head cpu_timers[3];
739#endif
740
741 /* Process credentials: */
742
743 /* Tracer's credentials at attach: */
744 const struct cred __rcu *ptracer_cred;
745
746 /* Objective and real subjective task credentials (COW): */
747 const struct cred __rcu *real_cred;
748
749 /* Effective (overridable) subjective task credentials (COW): */
750 const struct cred __rcu *cred;
751
752 /*
753 * executable name, excluding path.
754 *
755 * - normally initialized setup_new_exec()
756 * - access it with [gs]et_task_comm()
757 * - lock it with task_lock()
758 */
759 char comm[TASK_COMM_LEN];
760
761 struct nameidata *nameidata;
762
763#ifdef CONFIG_SYSVIPC
764 struct sysv_sem sysvsem;
765 struct sysv_shm sysvshm;
766#endif
767#ifdef CONFIG_DETECT_HUNG_TASK
768 unsigned long last_switch_count;
769#endif
770 /* Filesystem information: */
771 struct fs_struct *fs;
772
773 /* Open file information: */
774 struct files_struct *files;
775
776 /* Namespaces: */
777 struct nsproxy *nsproxy;
778
779 /* Signal handlers: */
780 struct signal_struct *signal;
781 struct sighand_struct *sighand;
782 sigset_t blocked;
783 sigset_t real_blocked;
784 /* Restored if set_restore_sigmask() was used: */
785 sigset_t saved_sigmask;
786 struct sigpending pending;
787 unsigned long sas_ss_sp;
788 size_t sas_ss_size;
789 unsigned int sas_ss_flags;
790
791 struct callback_head *task_works;
792
793 struct audit_context *audit_context;
794#ifdef CONFIG_AUDITSYSCALL
795 kuid_t loginuid;
796 unsigned int sessionid;
797#endif
798 struct seccomp seccomp;
799
800 /* Thread group tracking: */
801 u32 parent_exec_id;
802 u32 self_exec_id;
803
804 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
805 spinlock_t alloc_lock;
806
807 /* Protection of the PI data structures: */
808 raw_spinlock_t pi_lock;
809
810 struct wake_q_node wake_q;
811
812#ifdef CONFIG_RT_MUTEXES
813 /* PI waiters blocked on a rt_mutex held by this task: */
814 struct rb_root pi_waiters;
815 struct rb_node *pi_waiters_leftmost;
816 /* Updated under owner's pi_lock and rq lock */
817 struct task_struct *pi_top_task;
818 /* Deadlock detection and priority inheritance handling: */
819 struct rt_mutex_waiter *pi_blocked_on;
820#endif
821
822#ifdef CONFIG_DEBUG_MUTEXES
823 /* Mutex deadlock detection: */
824 struct mutex_waiter *blocked_on;
825#endif
826
827#ifdef CONFIG_TRACE_IRQFLAGS
828 unsigned int irq_events;
829 unsigned long hardirq_enable_ip;
830 unsigned long hardirq_disable_ip;
831 unsigned int hardirq_enable_event;
832 unsigned int hardirq_disable_event;
833 int hardirqs_enabled;
834 int hardirq_context;
835 unsigned long softirq_disable_ip;
836 unsigned long softirq_enable_ip;
837 unsigned int softirq_disable_event;
838 unsigned int softirq_enable_event;
839 int softirqs_enabled;
840 int softirq_context;
841#endif
842
843#ifdef CONFIG_LOCKDEP
844# define MAX_LOCK_DEPTH 48UL
845 u64 curr_chain_key;
846 int lockdep_depth;
847 unsigned int lockdep_recursion;
848 struct held_lock held_locks[MAX_LOCK_DEPTH];
849 gfp_t lockdep_reclaim_gfp;
850#endif
851
852#ifdef CONFIG_UBSAN
853 unsigned int in_ubsan;
854#endif
855
856 /* Journalling filesystem info: */
857 void *journal_info;
858
859 /* Stacked block device info: */
860 struct bio_list *bio_list;
861
862#ifdef CONFIG_BLOCK
863 /* Stack plugging: */
864 struct blk_plug *plug;
865#endif
866
867 /* VM state: */
868 struct reclaim_state *reclaim_state;
869
870 struct backing_dev_info *backing_dev_info;
871
872 struct io_context *io_context;
873
874 /* Ptrace state: */
875 unsigned long ptrace_message;
876 siginfo_t *last_siginfo;
877
878 struct task_io_accounting ioac;
879#ifdef CONFIG_TASK_XACCT
880 /* Accumulated RSS usage: */
881 u64 acct_rss_mem1;
882 /* Accumulated virtual memory usage: */
883 u64 acct_vm_mem1;
884 /* stime + utime since last update: */
885 u64 acct_timexpd;
886#endif
887#ifdef CONFIG_CPUSETS
888 /* Protected by ->alloc_lock: */
889 nodemask_t mems_allowed;
890 /* Seqence number to catch updates: */
891 seqcount_t mems_allowed_seq;
892 int cpuset_mem_spread_rotor;
893 int cpuset_slab_spread_rotor;
894#endif
895#ifdef CONFIG_CGROUPS
896 /* Control Group info protected by css_set_lock: */
897 struct css_set __rcu *cgroups;
898 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
899 struct list_head cg_list;
900#endif
901#ifdef CONFIG_INTEL_RDT_A
902 int closid;
903#endif
904#ifdef CONFIG_FUTEX
905 struct robust_list_head __user *robust_list;
906#ifdef CONFIG_COMPAT
907 struct compat_robust_list_head __user *compat_robust_list;
908#endif
909 struct list_head pi_state_list;
910 struct futex_pi_state *pi_state_cache;
911#endif
912#ifdef CONFIG_PERF_EVENTS
913 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
914 struct mutex perf_event_mutex;
915 struct list_head perf_event_list;
916#endif
917#ifdef CONFIG_DEBUG_PREEMPT
918 unsigned long preempt_disable_ip;
919#endif
920#ifdef CONFIG_NUMA
921 /* Protected by alloc_lock: */
922 struct mempolicy *mempolicy;
923 short il_prev;
924 short pref_node_fork;
925#endif
926#ifdef CONFIG_NUMA_BALANCING
927 int numa_scan_seq;
928 unsigned int numa_scan_period;
929 unsigned int numa_scan_period_max;
930 int numa_preferred_nid;
931 unsigned long numa_migrate_retry;
932 /* Migration stamp: */
933 u64 node_stamp;
934 u64 last_task_numa_placement;
935 u64 last_sum_exec_runtime;
936 struct callback_head numa_work;
937
938 struct list_head numa_entry;
939 struct numa_group *numa_group;
940
941 /*
942 * numa_faults is an array split into four regions:
943 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
944 * in this precise order.
945 *
946 * faults_memory: Exponential decaying average of faults on a per-node
947 * basis. Scheduling placement decisions are made based on these
948 * counts. The values remain static for the duration of a PTE scan.
949 * faults_cpu: Track the nodes the process was running on when a NUMA
950 * hinting fault was incurred.
951 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
952 * during the current scan window. When the scan completes, the counts
953 * in faults_memory and faults_cpu decay and these values are copied.
954 */
955 unsigned long *numa_faults;
956 unsigned long total_numa_faults;
957
958 /*
959 * numa_faults_locality tracks if faults recorded during the last
960 * scan window were remote/local or failed to migrate. The task scan
961 * period is adapted based on the locality of the faults with different
962 * weights depending on whether they were shared or private faults
963 */
964 unsigned long numa_faults_locality[3];
965
966 unsigned long numa_pages_migrated;
967#endif /* CONFIG_NUMA_BALANCING */
968
969 struct tlbflush_unmap_batch tlb_ubc;
970
971 struct rcu_head rcu;
972
973 /* Cache last used pipe for splice(): */
974 struct pipe_inode_info *splice_pipe;
975
976 struct page_frag task_frag;
977
978#ifdef CONFIG_TASK_DELAY_ACCT
979 struct task_delay_info *delays;
980#endif
981
982#ifdef CONFIG_FAULT_INJECTION
983 int make_it_fail;
984 unsigned int fail_nth;
985#endif
986 /*
987 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
988 * balance_dirty_pages() for a dirty throttling pause:
989 */
990 int nr_dirtied;
991 int nr_dirtied_pause;
992 /* Start of a write-and-pause period: */
993 unsigned long dirty_paused_when;
994
995#ifdef CONFIG_LATENCYTOP
996 int latency_record_count;
997 struct latency_record latency_record[LT_SAVECOUNT];
998#endif
999 /*
1000 * Time slack values; these are used to round up poll() and
1001 * select() etc timeout values. These are in nanoseconds.
1002 */
1003 u64 timer_slack_ns;
1004 u64 default_timer_slack_ns;
1005
1006#ifdef CONFIG_KASAN
1007 unsigned int kasan_depth;
1008#endif
1009
1010#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1011 /* Index of current stored address in ret_stack: */
1012 int curr_ret_stack;
1013
1014 /* Stack of return addresses for return function tracing: */
1015 struct ftrace_ret_stack *ret_stack;
1016
1017 /* Timestamp for last schedule: */
1018 unsigned long long ftrace_timestamp;
1019
1020 /*
1021 * Number of functions that haven't been traced
1022 * because of depth overrun:
1023 */
1024 atomic_t trace_overrun;
1025
1026 /* Pause tracing: */
1027 atomic_t tracing_graph_pause;
1028#endif
1029
1030#ifdef CONFIG_TRACING
1031 /* State flags for use by tracers: */
1032 unsigned long trace;
1033
1034 /* Bitmask and counter of trace recursion: */
1035 unsigned long trace_recursion;
1036#endif /* CONFIG_TRACING */
1037
1038#ifdef CONFIG_KCOV
1039 /* Coverage collection mode enabled for this task (0 if disabled): */
1040 enum kcov_mode kcov_mode;
1041
1042 /* Size of the kcov_area: */
1043 unsigned int kcov_size;
1044
1045 /* Buffer for coverage collection: */
1046 void *kcov_area;
1047
1048 /* KCOV descriptor wired with this task or NULL: */
1049 struct kcov *kcov;
1050#endif
1051
1052#ifdef CONFIG_MEMCG
1053 struct mem_cgroup *memcg_in_oom;
1054 gfp_t memcg_oom_gfp_mask;
1055 int memcg_oom_order;
1056
1057 /* Number of pages to reclaim on returning to userland: */
1058 unsigned int memcg_nr_pages_over_high;
1059#endif
1060
1061#ifdef CONFIG_UPROBES
1062 struct uprobe_task *utask;
1063#endif
1064#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1065 unsigned int sequential_io;
1066 unsigned int sequential_io_avg;
1067#endif
1068#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1069 unsigned long task_state_change;
1070#endif
1071 int pagefault_disabled;
1072#ifdef CONFIG_MMU
1073 struct task_struct *oom_reaper_list;
1074#endif
1075#ifdef CONFIG_VMAP_STACK
1076 struct vm_struct *stack_vm_area;
1077#endif
1078#ifdef CONFIG_THREAD_INFO_IN_TASK
1079 /* A live task holds one reference: */
1080 atomic_t stack_refcount;
1081#endif
1082#ifdef CONFIG_LIVEPATCH
1083 int patch_state;
1084#endif
1085#ifdef CONFIG_SECURITY
1086 /* Used by LSM modules for access restriction: */
1087 void *security;
1088#endif
1089
1090 /*
1091 * New fields for task_struct should be added above here, so that
1092 * they are included in the randomized portion of task_struct.
1093 */
1094 randomized_struct_fields_end
1095
1096 /* CPU-specific state of this task: */
1097 struct thread_struct thread;
1098
1099 /*
1100 * WARNING: on x86, 'thread_struct' contains a variable-sized
1101 * structure. It *MUST* be at the end of 'task_struct'.
1102 *
1103 * Do not put anything below here!
1104 */
1105};
1106
1107static inline struct pid *task_pid(struct task_struct *task)
1108{
1109 return task->pids[PIDTYPE_PID].pid;
1110}
1111
1112static inline struct pid *task_tgid(struct task_struct *task)
1113{
1114 return task->group_leader->pids[PIDTYPE_PID].pid;
1115}
1116
1117/*
1118 * Without tasklist or RCU lock it is not safe to dereference
1119 * the result of task_pgrp/task_session even if task == current,
1120 * we can race with another thread doing sys_setsid/sys_setpgid.
1121 */
1122static inline struct pid *task_pgrp(struct task_struct *task)
1123{
1124 return task->group_leader->pids[PIDTYPE_PGID].pid;
1125}
1126
1127static inline struct pid *task_session(struct task_struct *task)
1128{
1129 return task->group_leader->pids[PIDTYPE_SID].pid;
1130}
1131
1132/*
1133 * the helpers to get the task's different pids as they are seen
1134 * from various namespaces
1135 *
1136 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1137 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1138 * current.
1139 * task_xid_nr_ns() : id seen from the ns specified;
1140 *
1141 * see also pid_nr() etc in include/linux/pid.h
1142 */
1143pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1144
1145static inline pid_t task_pid_nr(struct task_struct *tsk)
1146{
1147 return tsk->pid;
1148}
1149
1150static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1151{
1152 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1153}
1154
1155static inline pid_t task_pid_vnr(struct task_struct *tsk)
1156{
1157 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1158}
1159
1160
1161static inline pid_t task_tgid_nr(struct task_struct *tsk)
1162{
1163 return tsk->tgid;
1164}
1165
1166/**
1167 * pid_alive - check that a task structure is not stale
1168 * @p: Task structure to be checked.
1169 *
1170 * Test if a process is not yet dead (at most zombie state)
1171 * If pid_alive fails, then pointers within the task structure
1172 * can be stale and must not be dereferenced.
1173 *
1174 * Return: 1 if the process is alive. 0 otherwise.
1175 */
1176static inline int pid_alive(const struct task_struct *p)
1177{
1178 return p->pids[PIDTYPE_PID].pid != NULL;
1179}
1180
1181static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1182{
1183 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1184}
1185
1186static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1187{
1188 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1189}
1190
1191
1192static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1193{
1194 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1195}
1196
1197static inline pid_t task_session_vnr(struct task_struct *tsk)
1198{
1199 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1200}
1201
1202static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1203{
1204 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
1205}
1206
1207static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1208{
1209 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
1210}
1211
1212static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1213{
1214 pid_t pid = 0;
1215
1216 rcu_read_lock();
1217 if (pid_alive(tsk))
1218 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1219 rcu_read_unlock();
1220
1221 return pid;
1222}
1223
1224static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1225{
1226 return task_ppid_nr_ns(tsk, &init_pid_ns);
1227}
1228
1229/* Obsolete, do not use: */
1230static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1231{
1232 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1233}
1234
1235/**
1236 * is_global_init - check if a task structure is init. Since init
1237 * is free to have sub-threads we need to check tgid.
1238 * @tsk: Task structure to be checked.
1239 *
1240 * Check if a task structure is the first user space task the kernel created.
1241 *
1242 * Return: 1 if the task structure is init. 0 otherwise.
1243 */
1244static inline int is_global_init(struct task_struct *tsk)
1245{
1246 return task_tgid_nr(tsk) == 1;
1247}
1248
1249extern struct pid *cad_pid;
1250
1251/*
1252 * Per process flags
1253 */
1254#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1255#define PF_EXITING 0x00000004 /* Getting shut down */
1256#define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1257#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1258#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1259#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1260#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1261#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1262#define PF_DUMPCORE 0x00000200 /* Dumped core */
1263#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1264#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1265#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1266#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1267#define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1268#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1269#define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1270#define PF_KSWAPD 0x00020000 /* I am kswapd */
1271#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1272#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1273#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1274#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1275#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1276#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1277#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1278#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1279#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1280#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1281#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1282
1283/*
1284 * Only the _current_ task can read/write to tsk->flags, but other
1285 * tasks can access tsk->flags in readonly mode for example
1286 * with tsk_used_math (like during threaded core dumping).
1287 * There is however an exception to this rule during ptrace
1288 * or during fork: the ptracer task is allowed to write to the
1289 * child->flags of its traced child (same goes for fork, the parent
1290 * can write to the child->flags), because we're guaranteed the
1291 * child is not running and in turn not changing child->flags
1292 * at the same time the parent does it.
1293 */
1294#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1295#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1296#define clear_used_math() clear_stopped_child_used_math(current)
1297#define set_used_math() set_stopped_child_used_math(current)
1298
1299#define conditional_stopped_child_used_math(condition, child) \
1300 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1301
1302#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1303
1304#define copy_to_stopped_child_used_math(child) \
1305 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1306
1307/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1308#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1309#define used_math() tsk_used_math(current)
1310
1311static inline bool is_percpu_thread(void)
1312{
1313#ifdef CONFIG_SMP
1314 return (current->flags & PF_NO_SETAFFINITY) &&
1315 (current->nr_cpus_allowed == 1);
1316#else
1317 return true;
1318#endif
1319}
1320
1321/* Per-process atomic flags. */
1322#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1323#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1324#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1325
1326
1327#define TASK_PFA_TEST(name, func) \
1328 static inline bool task_##func(struct task_struct *p) \
1329 { return test_bit(PFA_##name, &p->atomic_flags); }
1330
1331#define TASK_PFA_SET(name, func) \
1332 static inline void task_set_##func(struct task_struct *p) \
1333 { set_bit(PFA_##name, &p->atomic_flags); }
1334
1335#define TASK_PFA_CLEAR(name, func) \
1336 static inline void task_clear_##func(struct task_struct *p) \
1337 { clear_bit(PFA_##name, &p->atomic_flags); }
1338
1339TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1340TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1341
1342TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1343TASK_PFA_SET(SPREAD_PAGE, spread_page)
1344TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1345
1346TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1347TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1348TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1349
1350static inline void
1351current_restore_flags(unsigned long orig_flags, unsigned long flags)
1352{
1353 current->flags &= ~flags;
1354 current->flags |= orig_flags & flags;
1355}
1356
1357extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1358extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1359#ifdef CONFIG_SMP
1360extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1361extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1362#else
1363static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1364{
1365}
1366static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1367{
1368 if (!cpumask_test_cpu(0, new_mask))
1369 return -EINVAL;
1370 return 0;
1371}
1372#endif
1373
1374#ifndef cpu_relax_yield
1375#define cpu_relax_yield() cpu_relax()
1376#endif
1377
1378extern int yield_to(struct task_struct *p, bool preempt);
1379extern void set_user_nice(struct task_struct *p, long nice);
1380extern int task_prio(const struct task_struct *p);
1381
1382/**
1383 * task_nice - return the nice value of a given task.
1384 * @p: the task in question.
1385 *
1386 * Return: The nice value [ -20 ... 0 ... 19 ].
1387 */
1388static inline int task_nice(const struct task_struct *p)
1389{
1390 return PRIO_TO_NICE((p)->static_prio);
1391}
1392
1393extern int can_nice(const struct task_struct *p, const int nice);
1394extern int task_curr(const struct task_struct *p);
1395extern int idle_cpu(int cpu);
1396extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1397extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1398extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1399extern struct task_struct *idle_task(int cpu);
1400
1401/**
1402 * is_idle_task - is the specified task an idle task?
1403 * @p: the task in question.
1404 *
1405 * Return: 1 if @p is an idle task. 0 otherwise.
1406 */
1407static inline bool is_idle_task(const struct task_struct *p)
1408{
1409 return !!(p->flags & PF_IDLE);
1410}
1411
1412extern struct task_struct *curr_task(int cpu);
1413extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1414
1415void yield(void);
1416
1417union thread_union {
1418#ifndef CONFIG_THREAD_INFO_IN_TASK
1419 struct thread_info thread_info;
1420#endif
1421 unsigned long stack[THREAD_SIZE/sizeof(long)];
1422};
1423
1424#ifdef CONFIG_THREAD_INFO_IN_TASK
1425static inline struct thread_info *task_thread_info(struct task_struct *task)
1426{
1427 return &task->thread_info;
1428}
1429#elif !defined(__HAVE_THREAD_FUNCTIONS)
1430# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1431#endif
1432
1433/*
1434 * find a task by one of its numerical ids
1435 *
1436 * find_task_by_pid_ns():
1437 * finds a task by its pid in the specified namespace
1438 * find_task_by_vpid():
1439 * finds a task by its virtual pid
1440 *
1441 * see also find_vpid() etc in include/linux/pid.h
1442 */
1443
1444extern struct task_struct *find_task_by_vpid(pid_t nr);
1445extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1446
1447extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1448extern int wake_up_process(struct task_struct *tsk);
1449extern void wake_up_new_task(struct task_struct *tsk);
1450
1451#ifdef CONFIG_SMP
1452extern void kick_process(struct task_struct *tsk);
1453#else
1454static inline void kick_process(struct task_struct *tsk) { }
1455#endif
1456
1457extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1458
1459static inline void set_task_comm(struct task_struct *tsk, const char *from)
1460{
1461 __set_task_comm(tsk, from, false);
1462}
1463
1464extern char *get_task_comm(char *to, struct task_struct *tsk);
1465
1466#ifdef CONFIG_SMP
1467void scheduler_ipi(void);
1468extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1469#else
1470static inline void scheduler_ipi(void) { }
1471static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1472{
1473 return 1;
1474}
1475#endif
1476
1477/*
1478 * Set thread flags in other task's structures.
1479 * See asm/thread_info.h for TIF_xxxx flags available:
1480 */
1481static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1482{
1483 set_ti_thread_flag(task_thread_info(tsk), flag);
1484}
1485
1486static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1487{
1488 clear_ti_thread_flag(task_thread_info(tsk), flag);
1489}
1490
1491static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1492{
1493 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1494}
1495
1496static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1497{
1498 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1499}
1500
1501static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1502{
1503 return test_ti_thread_flag(task_thread_info(tsk), flag);
1504}
1505
1506static inline void set_tsk_need_resched(struct task_struct *tsk)
1507{
1508 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1509}
1510
1511static inline void clear_tsk_need_resched(struct task_struct *tsk)
1512{
1513 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1514}
1515
1516static inline int test_tsk_need_resched(struct task_struct *tsk)
1517{
1518 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1519}
1520
1521/*
1522 * cond_resched() and cond_resched_lock(): latency reduction via
1523 * explicit rescheduling in places that are safe. The return
1524 * value indicates whether a reschedule was done in fact.
1525 * cond_resched_lock() will drop the spinlock before scheduling,
1526 * cond_resched_softirq() will enable bhs before scheduling.
1527 */
1528#ifndef CONFIG_PREEMPT
1529extern int _cond_resched(void);
1530#else
1531static inline int _cond_resched(void) { return 0; }
1532#endif
1533
1534#define cond_resched() ({ \
1535 ___might_sleep(__FILE__, __LINE__, 0); \
1536 _cond_resched(); \
1537})
1538
1539extern int __cond_resched_lock(spinlock_t *lock);
1540
1541#define cond_resched_lock(lock) ({ \
1542 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1543 __cond_resched_lock(lock); \
1544})
1545
1546extern int __cond_resched_softirq(void);
1547
1548#define cond_resched_softirq() ({ \
1549 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1550 __cond_resched_softirq(); \
1551})
1552
1553static inline void cond_resched_rcu(void)
1554{
1555#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1556 rcu_read_unlock();
1557 cond_resched();
1558 rcu_read_lock();
1559#endif
1560}
1561
1562/*
1563 * Does a critical section need to be broken due to another
1564 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1565 * but a general need for low latency)
1566 */
1567static inline int spin_needbreak(spinlock_t *lock)
1568{
1569#ifdef CONFIG_PREEMPT
1570 return spin_is_contended(lock);
1571#else
1572 return 0;
1573#endif
1574}
1575
1576static __always_inline bool need_resched(void)
1577{
1578 return unlikely(tif_need_resched());
1579}
1580
1581/*
1582 * Wrappers for p->thread_info->cpu access. No-op on UP.
1583 */
1584#ifdef CONFIG_SMP
1585
1586static inline unsigned int task_cpu(const struct task_struct *p)
1587{
1588#ifdef CONFIG_THREAD_INFO_IN_TASK
1589 return p->cpu;
1590#else
1591 return task_thread_info(p)->cpu;
1592#endif
1593}
1594
1595extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1596
1597#else
1598
1599static inline unsigned int task_cpu(const struct task_struct *p)
1600{
1601 return 0;
1602}
1603
1604static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1605{
1606}
1607
1608#endif /* CONFIG_SMP */
1609
1610/*
1611 * In order to reduce various lock holder preemption latencies provide an
1612 * interface to see if a vCPU is currently running or not.
1613 *
1614 * This allows us to terminate optimistic spin loops and block, analogous to
1615 * the native optimistic spin heuristic of testing if the lock owner task is
1616 * running or not.
1617 */
1618#ifndef vcpu_is_preempted
1619# define vcpu_is_preempted(cpu) false
1620#endif
1621
1622extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1623extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1624
1625#ifndef TASK_SIZE_OF
1626#define TASK_SIZE_OF(tsk) TASK_SIZE
1627#endif
1628
1629#endif
1630