1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_SCHED_H
3#define _LINUX_SCHED_H
4
5/*
6 * Define 'struct task_struct' and provide the main scheduler
7 * APIs (schedule(), wakeup variants, etc.)
8 */
9
10#include <uapi/linux/sched.h>
11
12#include <asm/current.h>
13
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>
30
31/* task_struct member predeclarations (sorted alphabetically): */
32struct audit_context;
33struct backing_dev_info;
34struct bio_list;
35struct blk_plug;
36struct cfs_rq;
37struct fs_struct;
38struct futex_pi_state;
39struct io_context;
40struct mempolicy;
41struct nameidata;
42struct nsproxy;
43struct perf_event_context;
44struct pid_namespace;
45struct pipe_inode_info;
46struct rcu_node;
47struct reclaim_state;
48struct robust_list_head;
49struct sched_attr;
50struct sched_param;
51struct seq_file;
52struct sighand_struct;
53struct signal_struct;
54struct task_delay_info;
55struct task_group;
56
57/*
58 * Task state bitmask. NOTE! These bits are also
59 * encoded in fs/proc/array.c: get_task_state().
60 *
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
65 * mistake.
66 */
67
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
86
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)
91
92#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
93
94/* Convenience macros for the sake of wake_up(): */
95#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
96
97/* get_task_state(): */
98#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
99 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
100 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
101 TASK_PARKED)
102
103#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
104
105#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
106
107#define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
108
109#define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
110 (task->flags & PF_FROZEN) == 0 && \
111 (task->state & TASK_NOLOAD) == 0)
112
113#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
114
115/*
116 * Special states are those that do not use the normal wait-loop pattern. See
117 * the comment with set_special_state().
118 */
119#define is_special_task_state(state) \
120 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_DEAD))
121
122#define __set_current_state(state_value) \
123 do { \
124 WARN_ON_ONCE(is_special_task_state(state_value));\
125 current->task_state_change = _THIS_IP_; \
126 current->state = (state_value); \
127 } while (0)
128
129#define set_current_state(state_value) \
130 do { \
131 WARN_ON_ONCE(is_special_task_state(state_value));\
132 current->task_state_change = _THIS_IP_; \
133 smp_store_mb(current->state, (state_value)); \
134 } while (0)
135
136#define set_special_state(state_value) \
137 do { \
138 unsigned long flags; /* may shadow */ \
139 WARN_ON_ONCE(!is_special_task_state(state_value)); \
140 raw_spin_lock_irqsave(&current->pi_lock, flags); \
141 current->task_state_change = _THIS_IP_; \
142 current->state = (state_value); \
143 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
144 } while (0)
145#else
146/*
147 * set_current_state() includes a barrier so that the write of current->state
148 * is correctly serialised wrt the caller's subsequent test of whether to
149 * actually sleep:
150 *
151 * for (;;) {
152 * set_current_state(TASK_UNINTERRUPTIBLE);
153 * if (!need_sleep)
154 * break;
155 *
156 * schedule();
157 * }
158 * __set_current_state(TASK_RUNNING);
159 *
160 * If the caller does not need such serialisation (because, for instance, the
161 * condition test and condition change and wakeup are under the same lock) then
162 * use __set_current_state().
163 *
164 * The above is typically ordered against the wakeup, which does:
165 *
166 * need_sleep = false;
167 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
168 *
169 * Where wake_up_state() (and all other wakeup primitives) imply enough
170 * barriers to order the store of the variable against wakeup.
171 *
172 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
173 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
174 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
175 *
176 * However, with slightly different timing the wakeup TASK_RUNNING store can
177 * also collide with the TASK_UNINTERRUPTIBLE store. Loosing that store is not
178 * a problem either because that will result in one extra go around the loop
179 * and our @cond test will save the day.
180 *
181 * Also see the comments of try_to_wake_up().
182 */
183#define __set_current_state(state_value) \
184 current->state = (state_value)
185
186#define set_current_state(state_value) \
187 smp_store_mb(current->state, (state_value))
188
189/*
190 * set_special_state() should be used for those states when the blocking task
191 * can not use the regular condition based wait-loop. In that case we must
192 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
193 * will not collide with our state change.
194 */
195#define set_special_state(state_value) \
196 do { \
197 unsigned long flags; /* may shadow */ \
198 raw_spin_lock_irqsave(&current->pi_lock, flags); \
199 current->state = (state_value); \
200 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
201 } while (0)
202
203#endif
204
205/* Task command name length: */
206#define TASK_COMM_LEN 16
207
208extern void scheduler_tick(void);
209
210#define MAX_SCHEDULE_TIMEOUT LONG_MAX
211
212extern long schedule_timeout(long timeout);
213extern long schedule_timeout_interruptible(long timeout);
214extern long schedule_timeout_killable(long timeout);
215extern long schedule_timeout_uninterruptible(long timeout);
216extern long schedule_timeout_idle(long timeout);
217asmlinkage void schedule(void);
218extern void schedule_preempt_disabled(void);
219
220extern int __must_check io_schedule_prepare(void);
221extern void io_schedule_finish(int token);
222extern long io_schedule_timeout(long timeout);
223extern void io_schedule(void);
224
225/**
226 * struct prev_cputime - snapshot of system and user cputime
227 * @utime: time spent in user mode
228 * @stime: time spent in system mode
229 * @lock: protects the above two fields
230 *
231 * Stores previous user/system time values such that we can guarantee
232 * monotonicity.
233 */
234struct prev_cputime {
235#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
236 u64 utime;
237 u64 stime;
238 raw_spinlock_t lock;
239#endif
240};
241
242/**
243 * struct task_cputime - collected CPU time counts
244 * @utime: time spent in user mode, in nanoseconds
245 * @stime: time spent in kernel mode, in nanoseconds
246 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
247 *
248 * This structure groups together three kinds of CPU time that are tracked for
249 * threads and thread groups. Most things considering CPU time want to group
250 * these counts together and treat all three of them in parallel.
251 */
252struct task_cputime {
253 u64 utime;
254 u64 stime;
255 unsigned long long sum_exec_runtime;
256};
257
258/* Alternate field names when used on cache expirations: */
259#define virt_exp utime
260#define prof_exp stime
261#define sched_exp sum_exec_runtime
262
263enum vtime_state {
264 /* Task is sleeping or running in a CPU with VTIME inactive: */
265 VTIME_INACTIVE = 0,
266 /* Task runs in userspace in a CPU with VTIME active: */
267 VTIME_USER,
268 /* Task runs in kernelspace in a CPU with VTIME active: */
269 VTIME_SYS,
270};
271
272struct vtime {
273 seqcount_t seqcount;
274 unsigned long long starttime;
275 enum vtime_state state;
276 u64 utime;
277 u64 stime;
278 u64 gtime;
279};
280
281struct sched_info {
282#ifdef CONFIG_SCHED_INFO
283 /* Cumulative counters: */
284
285 /* # of times we have run on this CPU: */
286 unsigned long pcount;
287
288 /* Time spent waiting on a runqueue: */
289 unsigned long long run_delay;
290
291 /* Timestamps: */
292
293 /* When did we last run on a CPU? */
294 unsigned long long last_arrival;
295
296 /* When were we last queued to run? */
297 unsigned long long last_queued;
298
299#endif /* CONFIG_SCHED_INFO */
300};
301
302/*
303 * Integer metrics need fixed point arithmetic, e.g., sched/fair
304 * has a few: load, load_avg, util_avg, freq, and capacity.
305 *
306 * We define a basic fixed point arithmetic range, and then formalize
307 * all these metrics based on that basic range.
308 */
309# define SCHED_FIXEDPOINT_SHIFT 10
310# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
311
312struct load_weight {
313 unsigned long weight;
314 u32 inv_weight;
315};
316
317/**
318 * struct util_est - Estimation utilization of FAIR tasks
319 * @enqueued: instantaneous estimated utilization of a task/cpu
320 * @ewma: the Exponential Weighted Moving Average (EWMA)
321 * utilization of a task
322 *
323 * Support data structure to track an Exponential Weighted Moving Average
324 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
325 * average each time a task completes an activation. Sample's weight is chosen
326 * so that the EWMA will be relatively insensitive to transient changes to the
327 * task's workload.
328 *
329 * The enqueued attribute has a slightly different meaning for tasks and cpus:
330 * - task: the task's util_avg at last task dequeue time
331 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
332 * Thus, the util_est.enqueued of a task represents the contribution on the
333 * estimated utilization of the CPU where that task is currently enqueued.
334 *
335 * Only for tasks we track a moving average of the past instantaneous
336 * estimated utilization. This allows to absorb sporadic drops in utilization
337 * of an otherwise almost periodic task.
338 */
339struct util_est {
340 unsigned int enqueued;
341 unsigned int ewma;
342#define UTIL_EST_WEIGHT_SHIFT 2
343} __attribute__((__aligned__(sizeof(u64))));
344
345/*
346 * The load_avg/util_avg accumulates an infinite geometric series
347 * (see __update_load_avg() in kernel/sched/fair.c).
348 *
349 * [load_avg definition]
350 *
351 * load_avg = runnable% * scale_load_down(load)
352 *
353 * where runnable% is the time ratio that a sched_entity is runnable.
354 * For cfs_rq, it is the aggregated load_avg of all runnable and
355 * blocked sched_entities.
356 *
357 * load_avg may also take frequency scaling into account:
358 *
359 * load_avg = runnable% * scale_load_down(load) * freq%
360 *
361 * where freq% is the CPU frequency normalized to the highest frequency.
362 *
363 * [util_avg definition]
364 *
365 * util_avg = running% * SCHED_CAPACITY_SCALE
366 *
367 * where running% is the time ratio that a sched_entity is running on
368 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
369 * and blocked sched_entities.
370 *
371 * util_avg may also factor frequency scaling and CPU capacity scaling:
372 *
373 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
374 *
375 * where freq% is the same as above, and capacity% is the CPU capacity
376 * normalized to the greatest capacity (due to uarch differences, etc).
377 *
378 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
379 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
380 * we therefore scale them to as large a range as necessary. This is for
381 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
382 *
383 * [Overflow issue]
384 *
385 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
386 * with the highest load (=88761), always runnable on a single cfs_rq,
387 * and should not overflow as the number already hits PID_MAX_LIMIT.
388 *
389 * For all other cases (including 32-bit kernels), struct load_weight's
390 * weight will overflow first before we do, because:
391 *
392 * Max(load_avg) <= Max(load.weight)
393 *
394 * Then it is the load_weight's responsibility to consider overflow
395 * issues.
396 */
397struct sched_avg {
398 u64 last_update_time;
399 u64 load_sum;
400 u64 runnable_load_sum;
401 u32 util_sum;
402 u32 period_contrib;
403 unsigned long load_avg;
404 unsigned long runnable_load_avg;
405 unsigned long util_avg;
406 struct util_est util_est;
407} ____cacheline_aligned;
408
409struct sched_statistics {
410#ifdef CONFIG_SCHEDSTATS
411 u64 wait_start;
412 u64 wait_max;
413 u64 wait_count;
414 u64 wait_sum;
415 u64 iowait_count;
416 u64 iowait_sum;
417
418 u64 sleep_start;
419 u64 sleep_max;
420 s64 sum_sleep_runtime;
421
422 u64 block_start;
423 u64 block_max;
424 u64 exec_max;
425 u64 slice_max;
426
427 u64 nr_migrations_cold;
428 u64 nr_failed_migrations_affine;
429 u64 nr_failed_migrations_running;
430 u64 nr_failed_migrations_hot;
431 u64 nr_forced_migrations;
432
433 u64 nr_wakeups;
434 u64 nr_wakeups_sync;
435 u64 nr_wakeups_migrate;
436 u64 nr_wakeups_local;
437 u64 nr_wakeups_remote;
438 u64 nr_wakeups_affine;
439 u64 nr_wakeups_affine_attempts;
440 u64 nr_wakeups_passive;
441 u64 nr_wakeups_idle;
442#endif
443};
444
445struct sched_entity {
446 /* For load-balancing: */
447 struct load_weight load;
448 unsigned long runnable_weight;
449 struct rb_node run_node;
450 struct list_head group_node;
451 unsigned int on_rq;
452
453 u64 exec_start;
454 u64 sum_exec_runtime;
455 u64 vruntime;
456 u64 prev_sum_exec_runtime;
457
458 u64 nr_migrations;
459
460 struct sched_statistics statistics;
461
462#ifdef CONFIG_FAIR_GROUP_SCHED
463 int depth;
464 struct sched_entity *parent;
465 /* rq on which this entity is (to be) queued: */
466 struct cfs_rq *cfs_rq;
467 /* rq "owned" by this entity/group: */
468 struct cfs_rq *my_q;
469#endif
470
471#ifdef CONFIG_SMP
472 /*
473 * Per entity load average tracking.
474 *
475 * Put into separate cache line so it does not
476 * collide with read-mostly values above.
477 */
478 struct sched_avg avg;
479#endif
480};
481
482struct sched_rt_entity {
483 struct list_head run_list;
484 unsigned long timeout;
485 unsigned long watchdog_stamp;
486 unsigned int time_slice;
487 unsigned short on_rq;
488 unsigned short on_list;
489
490 struct sched_rt_entity *back;
491#ifdef CONFIG_RT_GROUP_SCHED
492 struct sched_rt_entity *parent;
493 /* rq on which this entity is (to be) queued: */
494 struct rt_rq *rt_rq;
495 /* rq "owned" by this entity/group: */
496 struct rt_rq *my_q;
497#endif
498} __randomize_layout;
499
500struct sched_dl_entity {
501 struct rb_node rb_node;
502
503 /*
504 * Original scheduling parameters. Copied here from sched_attr
505 * during sched_setattr(), they will remain the same until
506 * the next sched_setattr().
507 */
508 u64 dl_runtime; /* Maximum runtime for each instance */
509 u64 dl_deadline; /* Relative deadline of each instance */
510 u64 dl_period; /* Separation of two instances (period) */
511 u64 dl_bw; /* dl_runtime / dl_period */
512 u64 dl_density; /* dl_runtime / dl_deadline */
513
514 /*
515 * Actual scheduling parameters. Initialized with the values above,
516 * they are continously updated during task execution. Note that
517 * the remaining runtime could be < 0 in case we are in overrun.
518 */
519 s64 runtime; /* Remaining runtime for this instance */
520 u64 deadline; /* Absolute deadline for this instance */
521 unsigned int flags; /* Specifying the scheduler behaviour */
522
523 /*
524 * Some bool flags:
525 *
526 * @dl_throttled tells if we exhausted the runtime. If so, the
527 * task has to wait for a replenishment to be performed at the
528 * next firing of dl_timer.
529 *
530 * @dl_boosted tells if we are boosted due to DI. If so we are
531 * outside bandwidth enforcement mechanism (but only until we
532 * exit the critical section);
533 *
534 * @dl_yielded tells if task gave up the CPU before consuming
535 * all its available runtime during the last job.
536 *
537 * @dl_non_contending tells if the task is inactive while still
538 * contributing to the active utilization. In other words, it
539 * indicates if the inactive timer has been armed and its handler
540 * has not been executed yet. This flag is useful to avoid race
541 * conditions between the inactive timer handler and the wakeup
542 * code.
543 *
544 * @dl_overrun tells if the task asked to be informed about runtime
545 * overruns.
546 */
547 unsigned int dl_throttled : 1;
548 unsigned int dl_boosted : 1;
549 unsigned int dl_yielded : 1;
550 unsigned int dl_non_contending : 1;
551 unsigned int dl_overrun : 1;
552
553 /*
554 * Bandwidth enforcement timer. Each -deadline task has its
555 * own bandwidth to be enforced, thus we need one timer per task.
556 */
557 struct hrtimer dl_timer;
558
559 /*
560 * Inactive timer, responsible for decreasing the active utilization
561 * at the "0-lag time". When a -deadline task blocks, it contributes
562 * to GRUB's active utilization until the "0-lag time", hence a
563 * timer is needed to decrease the active utilization at the correct
564 * time.
565 */
566 struct hrtimer inactive_timer;
567};
568
569union rcu_special {
570 struct {
571 u8 blocked;
572 u8 need_qs;
573 u8 exp_need_qs;
574
575 /* Otherwise the compiler can store garbage here: */
576 u8 pad;
577 } b; /* Bits. */
578 u32 s; /* Set of bits. */
579};
580
581enum perf_event_task_context {
582 perf_invalid_context = -1,
583 perf_hw_context = 0,
584 perf_sw_context,
585 perf_nr_task_contexts,
586};
587
588struct wake_q_node {
589 struct wake_q_node *next;
590};
591
592struct task_struct {
593#ifdef CONFIG_THREAD_INFO_IN_TASK
594 /*
595 * For reasons of header soup (see current_thread_info()), this
596 * must be the first element of task_struct.
597 */
598 struct thread_info thread_info;
599#endif
600 /* -1 unrunnable, 0 runnable, >0 stopped: */
601 volatile long state;
602
603 /*
604 * This begins the randomizable portion of task_struct. Only
605 * scheduling-critical items should be added above here.
606 */
607 randomized_struct_fields_start
608
609 void *stack;
610 atomic_t usage;
611 /* Per task flags (PF_*), defined further below: */
612 unsigned int flags;
613 unsigned int ptrace;
614
615#ifdef CONFIG_SMP
616 struct llist_node wake_entry;
617 int on_cpu;
618#ifdef CONFIG_THREAD_INFO_IN_TASK
619 /* Current CPU: */
620 unsigned int cpu;
621#endif
622 unsigned int wakee_flips;
623 unsigned long wakee_flip_decay_ts;
624 struct task_struct *last_wakee;
625
626 /*
627 * recent_used_cpu is initially set as the last CPU used by a task
628 * that wakes affine another task. Waker/wakee relationships can
629 * push tasks around a CPU where each wakeup moves to the next one.
630 * Tracking a recently used CPU allows a quick search for a recently
631 * used CPU that may be idle.
632 */
633 int recent_used_cpu;
634 int wake_cpu;
635#endif
636 int on_rq;
637
638 int prio;
639 int static_prio;
640 int normal_prio;
641 unsigned int rt_priority;
642
643 const struct sched_class *sched_class;
644 struct sched_entity se;
645 struct sched_rt_entity rt;
646#ifdef CONFIG_CGROUP_SCHED
647 struct task_group *sched_task_group;
648#endif
649 struct sched_dl_entity dl;
650
651#ifdef CONFIG_PREEMPT_NOTIFIERS
652 /* List of struct preempt_notifier: */
653 struct hlist_head preempt_notifiers;
654#endif
655
656#ifdef CONFIG_BLK_DEV_IO_TRACE
657 unsigned int btrace_seq;
658#endif
659
660 unsigned int policy;
661 int nr_cpus_allowed;
662 cpumask_t cpus_allowed;
663
664#ifdef CONFIG_PREEMPT_RCU
665 int rcu_read_lock_nesting;
666 union rcu_special rcu_read_unlock_special;
667 struct list_head rcu_node_entry;
668 struct rcu_node *rcu_blocked_node;
669#endif /* #ifdef CONFIG_PREEMPT_RCU */
670
671#ifdef CONFIG_TASKS_RCU
672 unsigned long rcu_tasks_nvcsw;
673 u8 rcu_tasks_holdout;
674 u8 rcu_tasks_idx;
675 int rcu_tasks_idle_cpu;
676 struct list_head rcu_tasks_holdout_list;
677#endif /* #ifdef CONFIG_TASKS_RCU */
678
679 struct sched_info sched_info;
680
681 struct list_head tasks;
682#ifdef CONFIG_SMP
683 struct plist_node pushable_tasks;
684 struct rb_node pushable_dl_tasks;
685#endif
686
687 struct mm_struct *mm;
688 struct mm_struct *active_mm;
689
690 /* Per-thread vma caching: */
691 struct vmacache vmacache;
692
693#ifdef SPLIT_RSS_COUNTING
694 struct task_rss_stat rss_stat;
695#endif
696 int exit_state;
697 int exit_code;
698 int exit_signal;
699 /* The signal sent when the parent dies: */
700 int pdeath_signal;
701 /* JOBCTL_*, siglock protected: */
702 unsigned long jobctl;
703
704 /* Used for emulating ABI behavior of previous Linux versions: */
705 unsigned int personality;
706
707 /* Scheduler bits, serialized by scheduler locks: */
708 unsigned sched_reset_on_fork:1;
709 unsigned sched_contributes_to_load:1;
710 unsigned sched_migrated:1;
711 unsigned sched_remote_wakeup:1;
712 /* Force alignment to the next boundary: */
713 unsigned :0;
714
715 /* Unserialized, strictly 'current' */
716
717 /* Bit to tell LSMs we're in execve(): */
718 unsigned in_execve:1;
719 unsigned in_iowait:1;
720#ifndef TIF_RESTORE_SIGMASK
721 unsigned restore_sigmask:1;
722#endif
723#ifdef CONFIG_MEMCG
724 unsigned memcg_may_oom:1;
725#ifndef CONFIG_SLOB
726 unsigned memcg_kmem_skip_account:1;
727#endif
728#endif
729#ifdef CONFIG_COMPAT_BRK
730 unsigned brk_randomized:1;
731#endif
732#ifdef CONFIG_CGROUPS
733 /* disallow userland-initiated cgroup migration */
734 unsigned no_cgroup_migration:1;
735#endif
736
737 unsigned long atomic_flags; /* Flags requiring atomic access. */
738
739 struct restart_block restart_block;
740
741 pid_t pid;
742 pid_t tgid;
743
744#ifdef CONFIG_CC_STACKPROTECTOR
745 /* Canary value for the -fstack-protector GCC feature: */
746 unsigned long stack_canary;
747#endif
748 /*
749 * Pointers to the (original) parent process, youngest child, younger sibling,
750 * older sibling, respectively. (p->father can be replaced with
751 * p->real_parent->pid)
752 */
753
754 /* Real parent process: */
755 struct task_struct __rcu *real_parent;
756
757 /* Recipient of SIGCHLD, wait4() reports: */
758 struct task_struct __rcu *parent;
759
760 /*
761 * Children/sibling form the list of natural children:
762 */
763 struct list_head children;
764 struct list_head sibling;
765 struct task_struct *group_leader;
766
767 /*
768 * 'ptraced' is the list of tasks this task is using ptrace() on.
769 *
770 * This includes both natural children and PTRACE_ATTACH targets.
771 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
772 */
773 struct list_head ptraced;
774 struct list_head ptrace_entry;
775
776 /* PID/PID hash table linkage. */
777 struct pid_link pids[PIDTYPE_MAX];
778 struct list_head thread_group;
779 struct list_head thread_node;
780
781 struct completion *vfork_done;
782
783 /* CLONE_CHILD_SETTID: */
784 int __user *set_child_tid;
785
786 /* CLONE_CHILD_CLEARTID: */
787 int __user *clear_child_tid;
788
789 u64 utime;
790 u64 stime;
791#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
792 u64 utimescaled;
793 u64 stimescaled;
794#endif
795 u64 gtime;
796 struct prev_cputime prev_cputime;
797#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
798 struct vtime vtime;
799#endif
800
801#ifdef CONFIG_NO_HZ_FULL
802 atomic_t tick_dep_mask;
803#endif
804 /* Context switch counts: */
805 unsigned long nvcsw;
806 unsigned long nivcsw;
807
808 /* Monotonic time in nsecs: */
809 u64 start_time;
810
811 /* Boot based time in nsecs: */
812 u64 real_start_time;
813
814 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
815 unsigned long min_flt;
816 unsigned long maj_flt;
817
818#ifdef CONFIG_POSIX_TIMERS
819 struct task_cputime cputime_expires;
820 struct list_head cpu_timers[3];
821#endif
822
823 /* Process credentials: */
824
825 /* Tracer's credentials at attach: */
826 const struct cred __rcu *ptracer_cred;
827
828 /* Objective and real subjective task credentials (COW): */
829 const struct cred __rcu *real_cred;
830
831 /* Effective (overridable) subjective task credentials (COW): */
832 const struct cred __rcu *cred;
833
834 /*
835 * executable name, excluding path.
836 *
837 * - normally initialized setup_new_exec()
838 * - access it with [gs]et_task_comm()
839 * - lock it with task_lock()
840 */
841 char comm[TASK_COMM_LEN];
842
843 struct nameidata *nameidata;
844
845#ifdef CONFIG_SYSVIPC
846 struct sysv_sem sysvsem;
847 struct sysv_shm sysvshm;
848#endif
849#ifdef CONFIG_DETECT_HUNG_TASK
850 unsigned long last_switch_count;
851#endif
852 /* Filesystem information: */
853 struct fs_struct *fs;
854
855 /* Open file information: */
856 struct files_struct *files;
857
858 /* Namespaces: */
859 struct nsproxy *nsproxy;
860
861 /* Signal handlers: */
862 struct signal_struct *signal;
863 struct sighand_struct *sighand;
864 sigset_t blocked;
865 sigset_t real_blocked;
866 /* Restored if set_restore_sigmask() was used: */
867 sigset_t saved_sigmask;
868 struct sigpending pending;
869 unsigned long sas_ss_sp;
870 size_t sas_ss_size;
871 unsigned int sas_ss_flags;
872
873 struct callback_head *task_works;
874
875 struct audit_context *audit_context;
876#ifdef CONFIG_AUDITSYSCALL
877 kuid_t loginuid;
878 unsigned int sessionid;
879#endif
880 struct seccomp seccomp;
881
882 /* Thread group tracking: */
883 u32 parent_exec_id;
884 u32 self_exec_id;
885
886 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
887 spinlock_t alloc_lock;
888
889 /* Protection of the PI data structures: */
890 raw_spinlock_t pi_lock;
891
892 struct wake_q_node wake_q;
893
894#ifdef CONFIG_RT_MUTEXES
895 /* PI waiters blocked on a rt_mutex held by this task: */
896 struct rb_root_cached pi_waiters;
897 /* Updated under owner's pi_lock and rq lock */
898 struct task_struct *pi_top_task;
899 /* Deadlock detection and priority inheritance handling: */
900 struct rt_mutex_waiter *pi_blocked_on;
901#endif
902
903#ifdef CONFIG_DEBUG_MUTEXES
904 /* Mutex deadlock detection: */
905 struct mutex_waiter *blocked_on;
906#endif
907
908#ifdef CONFIG_TRACE_IRQFLAGS
909 unsigned int irq_events;
910 unsigned long hardirq_enable_ip;
911 unsigned long hardirq_disable_ip;
912 unsigned int hardirq_enable_event;
913 unsigned int hardirq_disable_event;
914 int hardirqs_enabled;
915 int hardirq_context;
916 unsigned long softirq_disable_ip;
917 unsigned long softirq_enable_ip;
918 unsigned int softirq_disable_event;
919 unsigned int softirq_enable_event;
920 int softirqs_enabled;
921 int softirq_context;
922#endif
923
924#ifdef CONFIG_LOCKDEP
925# define MAX_LOCK_DEPTH 48UL
926 u64 curr_chain_key;
927 int lockdep_depth;
928 unsigned int lockdep_recursion;
929 struct held_lock held_locks[MAX_LOCK_DEPTH];
930#endif
931
932#ifdef CONFIG_UBSAN
933 unsigned int in_ubsan;
934#endif
935
936 /* Journalling filesystem info: */
937 void *journal_info;
938
939 /* Stacked block device info: */
940 struct bio_list *bio_list;
941
942#ifdef CONFIG_BLOCK
943 /* Stack plugging: */
944 struct blk_plug *plug;
945#endif
946
947 /* VM state: */
948 struct reclaim_state *reclaim_state;
949
950 struct backing_dev_info *backing_dev_info;
951
952 struct io_context *io_context;
953
954 /* Ptrace state: */
955 unsigned long ptrace_message;
956 siginfo_t *last_siginfo;
957
958 struct task_io_accounting ioac;
959#ifdef CONFIG_TASK_XACCT
960 /* Accumulated RSS usage: */
961 u64 acct_rss_mem1;
962 /* Accumulated virtual memory usage: */
963 u64 acct_vm_mem1;
964 /* stime + utime since last update: */
965 u64 acct_timexpd;
966#endif
967#ifdef CONFIG_CPUSETS
968 /* Protected by ->alloc_lock: */
969 nodemask_t mems_allowed;
970 /* Seqence number to catch updates: */
971 seqcount_t mems_allowed_seq;
972 int cpuset_mem_spread_rotor;
973 int cpuset_slab_spread_rotor;
974#endif
975#ifdef CONFIG_CGROUPS
976 /* Control Group info protected by css_set_lock: */
977 struct css_set __rcu *cgroups;
978 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
979 struct list_head cg_list;
980#endif
981#ifdef CONFIG_INTEL_RDT
982 u32 closid;
983 u32 rmid;
984#endif
985#ifdef CONFIG_FUTEX
986 struct robust_list_head __user *robust_list;
987#ifdef CONFIG_COMPAT
988 struct compat_robust_list_head __user *compat_robust_list;
989#endif
990 struct list_head pi_state_list;
991 struct futex_pi_state *pi_state_cache;
992#endif
993#ifdef CONFIG_PERF_EVENTS
994 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
995 struct mutex perf_event_mutex;
996 struct list_head perf_event_list;
997#endif
998#ifdef CONFIG_DEBUG_PREEMPT
999 unsigned long preempt_disable_ip;
1000#endif
1001#ifdef CONFIG_NUMA
1002 /* Protected by alloc_lock: */
1003 struct mempolicy *mempolicy;
1004 short il_prev;
1005 short pref_node_fork;
1006#endif
1007#ifdef CONFIG_NUMA_BALANCING
1008 int numa_scan_seq;
1009 unsigned int numa_scan_period;
1010 unsigned int numa_scan_period_max;
1011 int numa_preferred_nid;
1012 unsigned long numa_migrate_retry;
1013 /* Migration stamp: */
1014 u64 node_stamp;
1015 u64 last_task_numa_placement;
1016 u64 last_sum_exec_runtime;
1017 struct callback_head numa_work;
1018
1019 struct list_head numa_entry;
1020 struct numa_group *numa_group;
1021
1022 /*
1023 * numa_faults is an array split into four regions:
1024 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1025 * in this precise order.
1026 *
1027 * faults_memory: Exponential decaying average of faults on a per-node
1028 * basis. Scheduling placement decisions are made based on these
1029 * counts. The values remain static for the duration of a PTE scan.
1030 * faults_cpu: Track the nodes the process was running on when a NUMA
1031 * hinting fault was incurred.
1032 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1033 * during the current scan window. When the scan completes, the counts
1034 * in faults_memory and faults_cpu decay and these values are copied.
1035 */
1036 unsigned long *numa_faults;
1037 unsigned long total_numa_faults;
1038
1039 /*
1040 * numa_faults_locality tracks if faults recorded during the last
1041 * scan window were remote/local or failed to migrate. The task scan
1042 * period is adapted based on the locality of the faults with different
1043 * weights depending on whether they were shared or private faults
1044 */
1045 unsigned long numa_faults_locality[3];
1046
1047 unsigned long numa_pages_migrated;
1048#endif /* CONFIG_NUMA_BALANCING */
1049
1050 struct tlbflush_unmap_batch tlb_ubc;
1051
1052 struct rcu_head rcu;
1053
1054 /* Cache last used pipe for splice(): */
1055 struct pipe_inode_info *splice_pipe;
1056
1057 struct page_frag task_frag;
1058
1059#ifdef CONFIG_TASK_DELAY_ACCT
1060 struct task_delay_info *delays;
1061#endif
1062
1063#ifdef CONFIG_FAULT_INJECTION
1064 int make_it_fail;
1065 unsigned int fail_nth;
1066#endif
1067 /*
1068 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1069 * balance_dirty_pages() for a dirty throttling pause:
1070 */
1071 int nr_dirtied;
1072 int nr_dirtied_pause;
1073 /* Start of a write-and-pause period: */
1074 unsigned long dirty_paused_when;
1075
1076#ifdef CONFIG_LATENCYTOP
1077 int latency_record_count;
1078 struct latency_record latency_record[LT_SAVECOUNT];
1079#endif
1080 /*
1081 * Time slack values; these are used to round up poll() and
1082 * select() etc timeout values. These are in nanoseconds.
1083 */
1084 u64 timer_slack_ns;
1085 u64 default_timer_slack_ns;
1086
1087#ifdef CONFIG_KASAN
1088 unsigned int kasan_depth;
1089#endif
1090
1091#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1092 /* Index of current stored address in ret_stack: */
1093 int curr_ret_stack;
1094
1095 /* Stack of return addresses for return function tracing: */
1096 struct ftrace_ret_stack *ret_stack;
1097
1098 /* Timestamp for last schedule: */
1099 unsigned long long ftrace_timestamp;
1100
1101 /*
1102 * Number of functions that haven't been traced
1103 * because of depth overrun:
1104 */
1105 atomic_t trace_overrun;
1106
1107 /* Pause tracing: */
1108 atomic_t tracing_graph_pause;
1109#endif
1110
1111#ifdef CONFIG_TRACING
1112 /* State flags for use by tracers: */
1113 unsigned long trace;
1114
1115 /* Bitmask and counter of trace recursion: */
1116 unsigned long trace_recursion;
1117#endif /* CONFIG_TRACING */
1118
1119#ifdef CONFIG_KCOV
1120 /* Coverage collection mode enabled for this task (0 if disabled): */
1121 enum kcov_mode kcov_mode;
1122
1123 /* Size of the kcov_area: */
1124 unsigned int kcov_size;
1125
1126 /* Buffer for coverage collection: */
1127 void *kcov_area;
1128
1129 /* KCOV descriptor wired with this task or NULL: */
1130 struct kcov *kcov;
1131#endif
1132
1133#ifdef CONFIG_MEMCG
1134 struct mem_cgroup *memcg_in_oom;
1135 gfp_t memcg_oom_gfp_mask;
1136 int memcg_oom_order;
1137
1138 /* Number of pages to reclaim on returning to userland: */
1139 unsigned int memcg_nr_pages_over_high;
1140#endif
1141
1142#ifdef CONFIG_UPROBES
1143 struct uprobe_task *utask;
1144#endif
1145#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1146 unsigned int sequential_io;
1147 unsigned int sequential_io_avg;
1148#endif
1149#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1150 unsigned long task_state_change;
1151#endif
1152 int pagefault_disabled;
1153#ifdef CONFIG_MMU
1154 struct task_struct *oom_reaper_list;
1155#endif
1156#ifdef CONFIG_VMAP_STACK
1157 struct vm_struct *stack_vm_area;
1158#endif
1159#ifdef CONFIG_THREAD_INFO_IN_TASK
1160 /* A live task holds one reference: */
1161 atomic_t stack_refcount;
1162#endif
1163#ifdef CONFIG_LIVEPATCH
1164 int patch_state;
1165#endif
1166#ifdef CONFIG_SECURITY
1167 /* Used by LSM modules for access restriction: */
1168 void *security;
1169#endif
1170
1171 /*
1172 * New fields for task_struct should be added above here, so that
1173 * they are included in the randomized portion of task_struct.
1174 */
1175 randomized_struct_fields_end
1176
1177 /* CPU-specific state of this task: */
1178 struct thread_struct thread;
1179
1180 /*
1181 * WARNING: on x86, 'thread_struct' contains a variable-sized
1182 * structure. It *MUST* be at the end of 'task_struct'.
1183 *
1184 * Do not put anything below here!
1185 */
1186};
1187
1188static inline struct pid *task_pid(struct task_struct *task)
1189{
1190 return task->pids[PIDTYPE_PID].pid;
1191}
1192
1193static inline struct pid *task_tgid(struct task_struct *task)
1194{
1195 return task->group_leader->pids[PIDTYPE_PID].pid;
1196}
1197
1198/*
1199 * Without tasklist or RCU lock it is not safe to dereference
1200 * the result of task_pgrp/task_session even if task == current,
1201 * we can race with another thread doing sys_setsid/sys_setpgid.
1202 */
1203static inline struct pid *task_pgrp(struct task_struct *task)
1204{
1205 return task->group_leader->pids[PIDTYPE_PGID].pid;
1206}
1207
1208static inline struct pid *task_session(struct task_struct *task)
1209{
1210 return task->group_leader->pids[PIDTYPE_SID].pid;
1211}
1212
1213/*
1214 * the helpers to get the task's different pids as they are seen
1215 * from various namespaces
1216 *
1217 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1218 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1219 * current.
1220 * task_xid_nr_ns() : id seen from the ns specified;
1221 *
1222 * see also pid_nr() etc in include/linux/pid.h
1223 */
1224pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1225
1226static inline pid_t task_pid_nr(struct task_struct *tsk)
1227{
1228 return tsk->pid;
1229}
1230
1231static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1232{
1233 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1234}
1235
1236static inline pid_t task_pid_vnr(struct task_struct *tsk)
1237{
1238 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1239}
1240
1241
1242static inline pid_t task_tgid_nr(struct task_struct *tsk)
1243{
1244 return tsk->tgid;
1245}
1246
1247/**
1248 * pid_alive - check that a task structure is not stale
1249 * @p: Task structure to be checked.
1250 *
1251 * Test if a process is not yet dead (at most zombie state)
1252 * If pid_alive fails, then pointers within the task structure
1253 * can be stale and must not be dereferenced.
1254 *
1255 * Return: 1 if the process is alive. 0 otherwise.
1256 */
1257static inline int pid_alive(const struct task_struct *p)
1258{
1259 return p->pids[PIDTYPE_PID].pid != NULL;
1260}
1261
1262static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1263{
1264 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1265}
1266
1267static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1268{
1269 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1270}
1271
1272
1273static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1274{
1275 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1276}
1277
1278static inline pid_t task_session_vnr(struct task_struct *tsk)
1279{
1280 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1281}
1282
1283static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1284{
1285 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, ns);
1286}
1287
1288static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1289{
1290 return __task_pid_nr_ns(tsk, __PIDTYPE_TGID, NULL);
1291}
1292
1293static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1294{
1295 pid_t pid = 0;
1296
1297 rcu_read_lock();
1298 if (pid_alive(tsk))
1299 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1300 rcu_read_unlock();
1301
1302 return pid;
1303}
1304
1305static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1306{
1307 return task_ppid_nr_ns(tsk, &init_pid_ns);
1308}
1309
1310/* Obsolete, do not use: */
1311static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1312{
1313 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1314}
1315
1316#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1317#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1318
1319static inline unsigned int task_state_index(struct task_struct *tsk)
1320{
1321 unsigned int tsk_state = READ_ONCE(tsk->state);
1322 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1323
1324 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1325
1326 if (tsk_state == TASK_IDLE)
1327 state = TASK_REPORT_IDLE;
1328
1329 return fls(state);
1330}
1331
1332static inline char task_index_to_char(unsigned int state)
1333{
1334 static const char state_char[] = "RSDTtXZPI";
1335
1336 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1337
1338 return state_char[state];
1339}
1340
1341static inline char task_state_to_char(struct task_struct *tsk)
1342{
1343 return task_index_to_char(task_state_index(tsk));
1344}
1345
1346/**
1347 * is_global_init - check if a task structure is init. Since init
1348 * is free to have sub-threads we need to check tgid.
1349 * @tsk: Task structure to be checked.
1350 *
1351 * Check if a task structure is the first user space task the kernel created.
1352 *
1353 * Return: 1 if the task structure is init. 0 otherwise.
1354 */
1355static inline int is_global_init(struct task_struct *tsk)
1356{
1357 return task_tgid_nr(tsk) == 1;
1358}
1359
1360extern struct pid *cad_pid;
1361
1362/*
1363 * Per process flags
1364 */
1365#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1366#define PF_EXITING 0x00000004 /* Getting shut down */
1367#define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1368#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1369#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1370#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1371#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1372#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1373#define PF_DUMPCORE 0x00000200 /* Dumped core */
1374#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1375#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1376#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1377#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1378#define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1379#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1380#define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1381#define PF_KSWAPD 0x00020000 /* I am kswapd */
1382#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1383#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1384#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1385#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1386#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1387#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1388#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1389#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1390#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1391#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1392#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1393
1394/*
1395 * Only the _current_ task can read/write to tsk->flags, but other
1396 * tasks can access tsk->flags in readonly mode for example
1397 * with tsk_used_math (like during threaded core dumping).
1398 * There is however an exception to this rule during ptrace
1399 * or during fork: the ptracer task is allowed to write to the
1400 * child->flags of its traced child (same goes for fork, the parent
1401 * can write to the child->flags), because we're guaranteed the
1402 * child is not running and in turn not changing child->flags
1403 * at the same time the parent does it.
1404 */
1405#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1406#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1407#define clear_used_math() clear_stopped_child_used_math(current)
1408#define set_used_math() set_stopped_child_used_math(current)
1409
1410#define conditional_stopped_child_used_math(condition, child) \
1411 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1412
1413#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1414
1415#define copy_to_stopped_child_used_math(child) \
1416 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1417
1418/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1419#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1420#define used_math() tsk_used_math(current)
1421
1422static inline bool is_percpu_thread(void)
1423{
1424#ifdef CONFIG_SMP
1425 return (current->flags & PF_NO_SETAFFINITY) &&
1426 (current->nr_cpus_allowed == 1);
1427#else
1428 return true;
1429#endif
1430}
1431
1432/* Per-process atomic flags. */
1433#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1434#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1435#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1436#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1437#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1438
1439#define TASK_PFA_TEST(name, func) \
1440 static inline bool task_##func(struct task_struct *p) \
1441 { return test_bit(PFA_##name, &p->atomic_flags); }
1442
1443#define TASK_PFA_SET(name, func) \
1444 static inline void task_set_##func(struct task_struct *p) \
1445 { set_bit(PFA_##name, &p->atomic_flags); }
1446
1447#define TASK_PFA_CLEAR(name, func) \
1448 static inline void task_clear_##func(struct task_struct *p) \
1449 { clear_bit(PFA_##name, &p->atomic_flags); }
1450
1451TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1452TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1453
1454TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1455TASK_PFA_SET(SPREAD_PAGE, spread_page)
1456TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1457
1458TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1459TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1460TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1461
1462TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1463TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1464TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1465
1466TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1467TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1468
1469static inline void
1470current_restore_flags(unsigned long orig_flags, unsigned long flags)
1471{
1472 current->flags &= ~flags;
1473 current->flags |= orig_flags & flags;
1474}
1475
1476extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1477extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1478#ifdef CONFIG_SMP
1479extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1480extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1481#else
1482static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1483{
1484}
1485static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1486{
1487 if (!cpumask_test_cpu(0, new_mask))
1488 return -EINVAL;
1489 return 0;
1490}
1491#endif
1492
1493#ifndef cpu_relax_yield
1494#define cpu_relax_yield() cpu_relax()
1495#endif
1496
1497extern int yield_to(struct task_struct *p, bool preempt);
1498extern void set_user_nice(struct task_struct *p, long nice);
1499extern int task_prio(const struct task_struct *p);
1500
1501/**
1502 * task_nice - return the nice value of a given task.
1503 * @p: the task in question.
1504 *
1505 * Return: The nice value [ -20 ... 0 ... 19 ].
1506 */
1507static inline int task_nice(const struct task_struct *p)
1508{
1509 return PRIO_TO_NICE((p)->static_prio);
1510}
1511
1512extern int can_nice(const struct task_struct *p, const int nice);
1513extern int task_curr(const struct task_struct *p);
1514extern int idle_cpu(int cpu);
1515extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1516extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1517extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1518extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1519extern struct task_struct *idle_task(int cpu);
1520
1521/**
1522 * is_idle_task - is the specified task an idle task?
1523 * @p: the task in question.
1524 *
1525 * Return: 1 if @p is an idle task. 0 otherwise.
1526 */
1527static inline bool is_idle_task(const struct task_struct *p)
1528{
1529 return !!(p->flags & PF_IDLE);
1530}
1531
1532extern struct task_struct *curr_task(int cpu);
1533extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1534
1535void yield(void);
1536
1537union thread_union {
1538#ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1539 struct task_struct task;
1540#endif
1541#ifndef CONFIG_THREAD_INFO_IN_TASK
1542 struct thread_info thread_info;
1543#endif
1544 unsigned long stack[THREAD_SIZE/sizeof(long)];
1545};
1546
1547#ifndef CONFIG_THREAD_INFO_IN_TASK
1548extern struct thread_info init_thread_info;
1549#endif
1550
1551extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1552
1553#ifdef CONFIG_THREAD_INFO_IN_TASK
1554static inline struct thread_info *task_thread_info(struct task_struct *task)
1555{
1556 return &task->thread_info;
1557}
1558#elif !defined(__HAVE_THREAD_FUNCTIONS)
1559# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1560#endif
1561
1562/*
1563 * find a task by one of its numerical ids
1564 *
1565 * find_task_by_pid_ns():
1566 * finds a task by its pid in the specified namespace
1567 * find_task_by_vpid():
1568 * finds a task by its virtual pid
1569 *
1570 * see also find_vpid() etc in include/linux/pid.h
1571 */
1572
1573extern struct task_struct *find_task_by_vpid(pid_t nr);
1574extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1575
1576/*
1577 * find a task by its virtual pid and get the task struct
1578 */
1579extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1580
1581extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1582extern int wake_up_process(struct task_struct *tsk);
1583extern void wake_up_new_task(struct task_struct *tsk);
1584
1585#ifdef CONFIG_SMP
1586extern void kick_process(struct task_struct *tsk);
1587#else
1588static inline void kick_process(struct task_struct *tsk) { }
1589#endif
1590
1591extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1592
1593static inline void set_task_comm(struct task_struct *tsk, const char *from)
1594{
1595 __set_task_comm(tsk, from, false);
1596}
1597
1598extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1599#define get_task_comm(buf, tsk) ({ \
1600 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1601 __get_task_comm(buf, sizeof(buf), tsk); \
1602})
1603
1604#ifdef CONFIG_SMP
1605void scheduler_ipi(void);
1606extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1607#else
1608static inline void scheduler_ipi(void) { }
1609static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1610{
1611 return 1;
1612}
1613#endif
1614
1615/*
1616 * Set thread flags in other task's structures.
1617 * See asm/thread_info.h for TIF_xxxx flags available:
1618 */
1619static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1620{
1621 set_ti_thread_flag(task_thread_info(tsk), flag);
1622}
1623
1624static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1625{
1626 clear_ti_thread_flag(task_thread_info(tsk), flag);
1627}
1628
1629static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1630{
1631 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1632}
1633
1634static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1635{
1636 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1637}
1638
1639static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1640{
1641 return test_ti_thread_flag(task_thread_info(tsk), flag);
1642}
1643
1644static inline void set_tsk_need_resched(struct task_struct *tsk)
1645{
1646 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1647}
1648
1649static inline void clear_tsk_need_resched(struct task_struct *tsk)
1650{
1651 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1652}
1653
1654static inline int test_tsk_need_resched(struct task_struct *tsk)
1655{
1656 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1657}
1658
1659/*
1660 * cond_resched() and cond_resched_lock(): latency reduction via
1661 * explicit rescheduling in places that are safe. The return
1662 * value indicates whether a reschedule was done in fact.
1663 * cond_resched_lock() will drop the spinlock before scheduling,
1664 * cond_resched_softirq() will enable bhs before scheduling.
1665 */
1666#ifndef CONFIG_PREEMPT
1667extern int _cond_resched(void);
1668#else
1669static inline int _cond_resched(void) { return 0; }
1670#endif
1671
1672#define cond_resched() ({ \
1673 ___might_sleep(__FILE__, __LINE__, 0); \
1674 _cond_resched(); \
1675})
1676
1677extern int __cond_resched_lock(spinlock_t *lock);
1678
1679#define cond_resched_lock(lock) ({ \
1680 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1681 __cond_resched_lock(lock); \
1682})
1683
1684extern int __cond_resched_softirq(void);
1685
1686#define cond_resched_softirq() ({ \
1687 ___might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
1688 __cond_resched_softirq(); \
1689})
1690
1691static inline void cond_resched_rcu(void)
1692{
1693#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1694 rcu_read_unlock();
1695 cond_resched();
1696 rcu_read_lock();
1697#endif
1698}
1699
1700/*
1701 * Does a critical section need to be broken due to another
1702 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1703 * but a general need for low latency)
1704 */
1705static inline int spin_needbreak(spinlock_t *lock)
1706{
1707#ifdef CONFIG_PREEMPT
1708 return spin_is_contended(lock);
1709#else
1710 return 0;
1711#endif
1712}
1713
1714static __always_inline bool need_resched(void)
1715{
1716 return unlikely(tif_need_resched());
1717}
1718
1719/*
1720 * Wrappers for p->thread_info->cpu access. No-op on UP.
1721 */
1722#ifdef CONFIG_SMP
1723
1724static inline unsigned int task_cpu(const struct task_struct *p)
1725{
1726#ifdef CONFIG_THREAD_INFO_IN_TASK
1727 return p->cpu;
1728#else
1729 return task_thread_info(p)->cpu;
1730#endif
1731}
1732
1733extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1734
1735#else
1736
1737static inline unsigned int task_cpu(const struct task_struct *p)
1738{
1739 return 0;
1740}
1741
1742static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1743{
1744}
1745
1746#endif /* CONFIG_SMP */
1747
1748/*
1749 * In order to reduce various lock holder preemption latencies provide an
1750 * interface to see if a vCPU is currently running or not.
1751 *
1752 * This allows us to terminate optimistic spin loops and block, analogous to
1753 * the native optimistic spin heuristic of testing if the lock owner task is
1754 * running or not.
1755 */
1756#ifndef vcpu_is_preempted
1757# define vcpu_is_preempted(cpu) false
1758#endif
1759
1760extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1761extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1762
1763#ifndef TASK_SIZE_OF
1764#define TASK_SIZE_OF(tsk) TASK_SIZE
1765#endif
1766
1767#endif
1768