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#include <linux/rseq.h>
31
32/* task_struct member predeclarations (sorted alphabetically): */
33struct audit_context;
34struct backing_dev_info;
35struct bio_list;
36struct blk_plug;
37struct cfs_rq;
38struct fs_struct;
39struct futex_pi_state;
40struct io_context;
41struct mempolicy;
42struct nameidata;
43struct nsproxy;
44struct perf_event_context;
45struct pid_namespace;
46struct pipe_inode_info;
47struct rcu_node;
48struct reclaim_state;
49struct robust_list_head;
50struct sched_attr;
51struct sched_param;
52struct seq_file;
53struct sighand_struct;
54struct signal_struct;
55struct task_delay_info;
56struct task_group;
57
58/*
59 * Task state bitmask. NOTE! These bits are also
60 * encoded in fs/proc/array.c: get_task_state().
61 *
62 * We have two separate sets of flags: task->state
63 * is about runnability, while task->exit_state are
64 * about the task exiting. Confusing, but this way
65 * modifying one set can't modify the other one by
66 * mistake.
67 */
68
69/* Used in tsk->state: */
70#define TASK_RUNNING 0x0000
71#define TASK_INTERRUPTIBLE 0x0001
72#define TASK_UNINTERRUPTIBLE 0x0002
73#define __TASK_STOPPED 0x0004
74#define __TASK_TRACED 0x0008
75/* Used in tsk->exit_state: */
76#define EXIT_DEAD 0x0010
77#define EXIT_ZOMBIE 0x0020
78#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
79/* Used in tsk->state again: */
80#define TASK_PARKED 0x0040
81#define TASK_DEAD 0x0080
82#define TASK_WAKEKILL 0x0100
83#define TASK_WAKING 0x0200
84#define TASK_NOLOAD 0x0400
85#define TASK_NEW 0x0800
86#define TASK_STATE_MAX 0x1000
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
98/* get_task_state(): */
99#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
100 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
101 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
102 TASK_PARKED)
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/*
117 * Special states are those that do not use the normal wait-loop pattern. See
118 * the comment with set_special_state().
119 */
120#define is_special_task_state(state) \
121 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
122
123#define __set_current_state(state_value) \
124 do { \
125 WARN_ON_ONCE(is_special_task_state(state_value));\
126 current->task_state_change = _THIS_IP_; \
127 current->state = (state_value); \
128 } while (0)
129
130#define set_current_state(state_value) \
131 do { \
132 WARN_ON_ONCE(is_special_task_state(state_value));\
133 current->task_state_change = _THIS_IP_; \
134 smp_store_mb(current->state, (state_value)); \
135 } while (0)
136
137#define set_special_state(state_value) \
138 do { \
139 unsigned long flags; /* may shadow */ \
140 WARN_ON_ONCE(!is_special_task_state(state_value)); \
141 raw_spin_lock_irqsave(&current->pi_lock, flags); \
142 current->task_state_change = _THIS_IP_; \
143 current->state = (state_value); \
144 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
145 } while (0)
146#else
147/*
148 * set_current_state() includes a barrier so that the write of current->state
149 * is correctly serialised wrt the caller's subsequent test of whether to
150 * actually sleep:
151 *
152 * for (;;) {
153 * set_current_state(TASK_UNINTERRUPTIBLE);
154 * if (!need_sleep)
155 * break;
156 *
157 * schedule();
158 * }
159 * __set_current_state(TASK_RUNNING);
160 *
161 * If the caller does not need such serialisation (because, for instance, the
162 * condition test and condition change and wakeup are under the same lock) then
163 * use __set_current_state().
164 *
165 * The above is typically ordered against the wakeup, which does:
166 *
167 * need_sleep = false;
168 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
169 *
170 * where wake_up_state() executes a full memory barrier before accessing the
171 * task state.
172 *
173 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
174 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
175 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
176 *
177 * However, with slightly different timing the wakeup TASK_RUNNING store can
178 * also collide with the TASK_UNINTERRUPTIBLE store. Loosing that store is not
179 * a problem either because that will result in one extra go around the loop
180 * and our @cond test will save the day.
181 *
182 * Also see the comments of try_to_wake_up().
183 */
184#define __set_current_state(state_value) \
185 current->state = (state_value)
186
187#define set_current_state(state_value) \
188 smp_store_mb(current->state, (state_value))
189
190/*
191 * set_special_state() should be used for those states when the blocking task
192 * can not use the regular condition based wait-loop. In that case we must
193 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
194 * will not collide with our state change.
195 */
196#define set_special_state(state_value) \
197 do { \
198 unsigned long flags; /* may shadow */ \
199 raw_spin_lock_irqsave(&current->pi_lock, flags); \
200 current->state = (state_value); \
201 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
202 } while (0)
203
204#endif
205
206/* Task command name length: */
207#define TASK_COMM_LEN 16
208
209extern void scheduler_tick(void);
210
211#define MAX_SCHEDULE_TIMEOUT LONG_MAX
212
213extern long schedule_timeout(long timeout);
214extern long schedule_timeout_interruptible(long timeout);
215extern long schedule_timeout_killable(long timeout);
216extern long schedule_timeout_uninterruptible(long timeout);
217extern long schedule_timeout_idle(long timeout);
218asmlinkage void schedule(void);
219extern void schedule_preempt_disabled(void);
220
221extern int __must_check io_schedule_prepare(void);
222extern void io_schedule_finish(int token);
223extern long io_schedule_timeout(long timeout);
224extern void io_schedule(void);
225
226/**
227 * struct prev_cputime - snapshot of system and user cputime
228 * @utime: time spent in user mode
229 * @stime: time spent in system mode
230 * @lock: protects the above two fields
231 *
232 * Stores previous user/system time values such that we can guarantee
233 * monotonicity.
234 */
235struct prev_cputime {
236#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
237 u64 utime;
238 u64 stime;
239 raw_spinlock_t lock;
240#endif
241};
242
243/**
244 * struct task_cputime - collected CPU time counts
245 * @utime: time spent in user mode, in nanoseconds
246 * @stime: time spent in kernel mode, in nanoseconds
247 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
248 *
249 * This structure groups together three kinds of CPU time that are tracked for
250 * threads and thread groups. Most things considering CPU time want to group
251 * these counts together and treat all three of them in parallel.
252 */
253struct task_cputime {
254 u64 utime;
255 u64 stime;
256 unsigned long long sum_exec_runtime;
257};
258
259/* Alternate field names when used on cache expirations: */
260#define virt_exp utime
261#define prof_exp stime
262#define sched_exp sum_exec_runtime
263
264enum vtime_state {
265 /* Task is sleeping or running in a CPU with VTIME inactive: */
266 VTIME_INACTIVE = 0,
267 /* Task runs in userspace in a CPU with VTIME active: */
268 VTIME_USER,
269 /* Task runs in kernelspace in a CPU with VTIME active: */
270 VTIME_SYS,
271};
272
273struct vtime {
274 seqcount_t seqcount;
275 unsigned long long starttime;
276 enum vtime_state state;
277 u64 utime;
278 u64 stime;
279 u64 gtime;
280};
281
282struct sched_info {
283#ifdef CONFIG_SCHED_INFO
284 /* Cumulative counters: */
285
286 /* # of times we have run on this CPU: */
287 unsigned long pcount;
288
289 /* Time spent waiting on a runqueue: */
290 unsigned long long run_delay;
291
292 /* Timestamps: */
293
294 /* When did we last run on a CPU? */
295 unsigned long long last_arrival;
296
297 /* When were we last queued to run? */
298 unsigned long long last_queued;
299
300#endif /* CONFIG_SCHED_INFO */
301};
302
303/*
304 * Integer metrics need fixed point arithmetic, e.g., sched/fair
305 * has a few: load, load_avg, util_avg, freq, and capacity.
306 *
307 * We define a basic fixed point arithmetic range, and then formalize
308 * all these metrics based on that basic range.
309 */
310# define SCHED_FIXEDPOINT_SHIFT 10
311# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
312
313struct load_weight {
314 unsigned long weight;
315 u32 inv_weight;
316};
317
318/**
319 * struct util_est - Estimation utilization of FAIR tasks
320 * @enqueued: instantaneous estimated utilization of a task/cpu
321 * @ewma: the Exponential Weighted Moving Average (EWMA)
322 * utilization of a task
323 *
324 * Support data structure to track an Exponential Weighted Moving Average
325 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
326 * average each time a task completes an activation. Sample's weight is chosen
327 * so that the EWMA will be relatively insensitive to transient changes to the
328 * task's workload.
329 *
330 * The enqueued attribute has a slightly different meaning for tasks and cpus:
331 * - task: the task's util_avg at last task dequeue time
332 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
333 * Thus, the util_est.enqueued of a task represents the contribution on the
334 * estimated utilization of the CPU where that task is currently enqueued.
335 *
336 * Only for tasks we track a moving average of the past instantaneous
337 * estimated utilization. This allows to absorb sporadic drops in utilization
338 * of an otherwise almost periodic task.
339 */
340struct util_est {
341 unsigned int enqueued;
342 unsigned int ewma;
343#define UTIL_EST_WEIGHT_SHIFT 2
344} __attribute__((__aligned__(sizeof(u64))));
345
346/*
347 * The load_avg/util_avg accumulates an infinite geometric series
348 * (see __update_load_avg() in kernel/sched/fair.c).
349 *
350 * [load_avg definition]
351 *
352 * load_avg = runnable% * scale_load_down(load)
353 *
354 * where runnable% is the time ratio that a sched_entity is runnable.
355 * For cfs_rq, it is the aggregated load_avg of all runnable and
356 * blocked sched_entities.
357 *
358 * load_avg may also take frequency scaling into account:
359 *
360 * load_avg = runnable% * scale_load_down(load) * freq%
361 *
362 * where freq% is the CPU frequency normalized to the highest frequency.
363 *
364 * [util_avg definition]
365 *
366 * util_avg = running% * SCHED_CAPACITY_SCALE
367 *
368 * where running% is the time ratio that a sched_entity is running on
369 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
370 * and blocked sched_entities.
371 *
372 * util_avg may also factor frequency scaling and CPU capacity scaling:
373 *
374 * util_avg = running% * SCHED_CAPACITY_SCALE * freq% * capacity%
375 *
376 * where freq% is the same as above, and capacity% is the CPU capacity
377 * normalized to the greatest capacity (due to uarch differences, etc).
378 *
379 * N.B., the above ratios (runnable%, running%, freq%, and capacity%)
380 * themselves are in the range of [0, 1]. To do fixed point arithmetics,
381 * we therefore scale them to as large a range as necessary. This is for
382 * example reflected by util_avg's SCHED_CAPACITY_SCALE.
383 *
384 * [Overflow issue]
385 *
386 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
387 * with the highest load (=88761), always runnable on a single cfs_rq,
388 * and should not overflow as the number already hits PID_MAX_LIMIT.
389 *
390 * For all other cases (including 32-bit kernels), struct load_weight's
391 * weight will overflow first before we do, because:
392 *
393 * Max(load_avg) <= Max(load.weight)
394 *
395 * Then it is the load_weight's responsibility to consider overflow
396 * issues.
397 */
398struct sched_avg {
399 u64 last_update_time;
400 u64 load_sum;
401 u64 runnable_load_sum;
402 u32 util_sum;
403 u32 period_contrib;
404 unsigned long load_avg;
405 unsigned long runnable_load_avg;
406 unsigned long util_avg;
407 struct util_est util_est;
408} ____cacheline_aligned;
409
410struct sched_statistics {
411#ifdef CONFIG_SCHEDSTATS
412 u64 wait_start;
413 u64 wait_max;
414 u64 wait_count;
415 u64 wait_sum;
416 u64 iowait_count;
417 u64 iowait_sum;
418
419 u64 sleep_start;
420 u64 sleep_max;
421 s64 sum_sleep_runtime;
422
423 u64 block_start;
424 u64 block_max;
425 u64 exec_max;
426 u64 slice_max;
427
428 u64 nr_migrations_cold;
429 u64 nr_failed_migrations_affine;
430 u64 nr_failed_migrations_running;
431 u64 nr_failed_migrations_hot;
432 u64 nr_forced_migrations;
433
434 u64 nr_wakeups;
435 u64 nr_wakeups_sync;
436 u64 nr_wakeups_migrate;
437 u64 nr_wakeups_local;
438 u64 nr_wakeups_remote;
439 u64 nr_wakeups_affine;
440 u64 nr_wakeups_affine_attempts;
441 u64 nr_wakeups_passive;
442 u64 nr_wakeups_idle;
443#endif
444};
445
446struct sched_entity {
447 /* For load-balancing: */
448 struct load_weight load;
449 unsigned long runnable_weight;
450 struct rb_node run_node;
451 struct list_head group_node;
452 unsigned int on_rq;
453
454 u64 exec_start;
455 u64 sum_exec_runtime;
456 u64 vruntime;
457 u64 prev_sum_exec_runtime;
458
459 u64 nr_migrations;
460
461 struct sched_statistics statistics;
462
463#ifdef CONFIG_FAIR_GROUP_SCHED
464 int depth;
465 struct sched_entity *parent;
466 /* rq on which this entity is (to be) queued: */
467 struct cfs_rq *cfs_rq;
468 /* rq "owned" by this entity/group: */
469 struct cfs_rq *my_q;
470#endif
471
472#ifdef CONFIG_SMP
473 /*
474 * Per entity load average tracking.
475 *
476 * Put into separate cache line so it does not
477 * collide with read-mostly values above.
478 */
479 struct sched_avg avg;
480#endif
481};
482
483struct sched_rt_entity {
484 struct list_head run_list;
485 unsigned long timeout;
486 unsigned long watchdog_stamp;
487 unsigned int time_slice;
488 unsigned short on_rq;
489 unsigned short on_list;
490
491 struct sched_rt_entity *back;
492#ifdef CONFIG_RT_GROUP_SCHED
493 struct sched_rt_entity *parent;
494 /* rq on which this entity is (to be) queued: */
495 struct rt_rq *rt_rq;
496 /* rq "owned" by this entity/group: */
497 struct rt_rq *my_q;
498#endif
499} __randomize_layout;
500
501struct sched_dl_entity {
502 struct rb_node rb_node;
503
504 /*
505 * Original scheduling parameters. Copied here from sched_attr
506 * during sched_setattr(), they will remain the same until
507 * the next sched_setattr().
508 */
509 u64 dl_runtime; /* Maximum runtime for each instance */
510 u64 dl_deadline; /* Relative deadline of each instance */
511 u64 dl_period; /* Separation of two instances (period) */
512 u64 dl_bw; /* dl_runtime / dl_period */
513 u64 dl_density; /* dl_runtime / dl_deadline */
514
515 /*
516 * Actual scheduling parameters. Initialized with the values above,
517 * they are continously updated during task execution. Note that
518 * the remaining runtime could be < 0 in case we are in overrun.
519 */
520 s64 runtime; /* Remaining runtime for this instance */
521 u64 deadline; /* Absolute deadline for this instance */
522 unsigned int flags; /* Specifying the scheduler behaviour */
523
524 /*
525 * Some bool flags:
526 *
527 * @dl_throttled tells if we exhausted the runtime. If so, the
528 * task has to wait for a replenishment to be performed at the
529 * next firing of dl_timer.
530 *
531 * @dl_boosted tells if we are boosted due to DI. If so we are
532 * outside bandwidth enforcement mechanism (but only until we
533 * exit the critical section);
534 *
535 * @dl_yielded tells if task gave up the CPU before consuming
536 * all its available runtime during the last job.
537 *
538 * @dl_non_contending tells if the task is inactive while still
539 * contributing to the active utilization. In other words, it
540 * indicates if the inactive timer has been armed and its handler
541 * has not been executed yet. This flag is useful to avoid race
542 * conditions between the inactive timer handler and the wakeup
543 * code.
544 *
545 * @dl_overrun tells if the task asked to be informed about runtime
546 * overruns.
547 */
548 unsigned int dl_throttled : 1;
549 unsigned int dl_boosted : 1;
550 unsigned int dl_yielded : 1;
551 unsigned int dl_non_contending : 1;
552 unsigned int dl_overrun : 1;
553
554 /*
555 * Bandwidth enforcement timer. Each -deadline task has its
556 * own bandwidth to be enforced, thus we need one timer per task.
557 */
558 struct hrtimer dl_timer;
559
560 /*
561 * Inactive timer, responsible for decreasing the active utilization
562 * at the "0-lag time". When a -deadline task blocks, it contributes
563 * to GRUB's active utilization until the "0-lag time", hence a
564 * timer is needed to decrease the active utilization at the correct
565 * time.
566 */
567 struct hrtimer inactive_timer;
568};
569
570union rcu_special {
571 struct {
572 u8 blocked;
573 u8 need_qs;
574 u8 exp_need_qs;
575
576 /* Otherwise the compiler can store garbage here: */
577 u8 pad;
578 } b; /* Bits. */
579 u32 s; /* Set of bits. */
580};
581
582enum perf_event_task_context {
583 perf_invalid_context = -1,
584 perf_hw_context = 0,
585 perf_sw_context,
586 perf_nr_task_contexts,
587};
588
589struct wake_q_node {
590 struct wake_q_node *next;
591};
592
593struct task_struct {
594#ifdef CONFIG_THREAD_INFO_IN_TASK
595 /*
596 * For reasons of header soup (see current_thread_info()), this
597 * must be the first element of task_struct.
598 */
599 struct thread_info thread_info;
600#endif
601 /* -1 unrunnable, 0 runnable, >0 stopped: */
602 volatile long state;
603
604 /*
605 * This begins the randomizable portion of task_struct. Only
606 * scheduling-critical items should be added above here.
607 */
608 randomized_struct_fields_start
609
610 void *stack;
611 atomic_t usage;
612 /* Per task flags (PF_*), defined further below: */
613 unsigned int flags;
614 unsigned int ptrace;
615
616#ifdef CONFIG_SMP
617 struct llist_node wake_entry;
618 int on_cpu;
619#ifdef CONFIG_THREAD_INFO_IN_TASK
620 /* Current CPU: */
621 unsigned int cpu;
622#endif
623 unsigned int wakee_flips;
624 unsigned long wakee_flip_decay_ts;
625 struct task_struct *last_wakee;
626
627 /*
628 * recent_used_cpu is initially set as the last CPU used by a task
629 * that wakes affine another task. Waker/wakee relationships can
630 * push tasks around a CPU where each wakeup moves to the next one.
631 * Tracking a recently used CPU allows a quick search for a recently
632 * used CPU that may be idle.
633 */
634 int recent_used_cpu;
635 int wake_cpu;
636#endif
637 int on_rq;
638
639 int prio;
640 int static_prio;
641 int normal_prio;
642 unsigned int rt_priority;
643
644 const struct sched_class *sched_class;
645 struct sched_entity se;
646 struct sched_rt_entity rt;
647#ifdef CONFIG_CGROUP_SCHED
648 struct task_group *sched_task_group;
649#endif
650 struct sched_dl_entity dl;
651
652#ifdef CONFIG_PREEMPT_NOTIFIERS
653 /* List of struct preempt_notifier: */
654 struct hlist_head preempt_notifiers;
655#endif
656
657#ifdef CONFIG_BLK_DEV_IO_TRACE
658 unsigned int btrace_seq;
659#endif
660
661 unsigned int policy;
662 int nr_cpus_allowed;
663 cpumask_t cpus_allowed;
664
665#ifdef CONFIG_PREEMPT_RCU
666 int rcu_read_lock_nesting;
667 union rcu_special rcu_read_unlock_special;
668 struct list_head rcu_node_entry;
669 struct rcu_node *rcu_blocked_node;
670#endif /* #ifdef CONFIG_PREEMPT_RCU */
671
672#ifdef CONFIG_TASKS_RCU
673 unsigned long rcu_tasks_nvcsw;
674 u8 rcu_tasks_holdout;
675 u8 rcu_tasks_idx;
676 int rcu_tasks_idle_cpu;
677 struct list_head rcu_tasks_holdout_list;
678#endif /* #ifdef CONFIG_TASKS_RCU */
679
680 struct sched_info sched_info;
681
682 struct list_head tasks;
683#ifdef CONFIG_SMP
684 struct plist_node pushable_tasks;
685 struct rb_node pushable_dl_tasks;
686#endif
687
688 struct mm_struct *mm;
689 struct mm_struct *active_mm;
690
691 /* Per-thread vma caching: */
692 struct vmacache vmacache;
693
694#ifdef SPLIT_RSS_COUNTING
695 struct task_rss_stat rss_stat;
696#endif
697 int exit_state;
698 int exit_code;
699 int exit_signal;
700 /* The signal sent when the parent dies: */
701 int pdeath_signal;
702 /* JOBCTL_*, siglock protected: */
703 unsigned long jobctl;
704
705 /* Used for emulating ABI behavior of previous Linux versions: */
706 unsigned int personality;
707
708 /* Scheduler bits, serialized by scheduler locks: */
709 unsigned sched_reset_on_fork:1;
710 unsigned sched_contributes_to_load:1;
711 unsigned sched_migrated:1;
712 unsigned sched_remote_wakeup:1;
713 /* Force alignment to the next boundary: */
714 unsigned :0;
715
716 /* Unserialized, strictly 'current' */
717
718 /* Bit to tell LSMs we're in execve(): */
719 unsigned in_execve:1;
720 unsigned in_iowait:1;
721#ifndef TIF_RESTORE_SIGMASK
722 unsigned restore_sigmask:1;
723#endif
724#ifdef CONFIG_MEMCG
725 unsigned in_user_fault:1;
726#ifdef CONFIG_MEMCG_KMEM
727 unsigned memcg_kmem_skip_account:1;
728#endif
729#endif
730#ifdef CONFIG_COMPAT_BRK
731 unsigned brk_randomized:1;
732#endif
733#ifdef CONFIG_CGROUPS
734 /* disallow userland-initiated cgroup migration */
735 unsigned no_cgroup_migration:1;
736#endif
737#ifdef CONFIG_BLK_CGROUP
738 /* to be used once the psi infrastructure lands upstream. */
739 unsigned use_memdelay:1;
740#endif
741
742 unsigned long atomic_flags; /* Flags requiring atomic access. */
743
744 struct restart_block restart_block;
745
746 pid_t pid;
747 pid_t tgid;
748
749#ifdef CONFIG_STACKPROTECTOR
750 /* Canary value for the -fstack-protector GCC feature: */
751 unsigned long stack_canary;
752#endif
753 /*
754 * Pointers to the (original) parent process, youngest child, younger sibling,
755 * older sibling, respectively. (p->father can be replaced with
756 * p->real_parent->pid)
757 */
758
759 /* Real parent process: */
760 struct task_struct __rcu *real_parent;
761
762 /* Recipient of SIGCHLD, wait4() reports: */
763 struct task_struct __rcu *parent;
764
765 /*
766 * Children/sibling form the list of natural children:
767 */
768 struct list_head children;
769 struct list_head sibling;
770 struct task_struct *group_leader;
771
772 /*
773 * 'ptraced' is the list of tasks this task is using ptrace() on.
774 *
775 * This includes both natural children and PTRACE_ATTACH targets.
776 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
777 */
778 struct list_head ptraced;
779 struct list_head ptrace_entry;
780
781 /* PID/PID hash table linkage. */
782 struct pid *thread_pid;
783 struct hlist_node pid_links[PIDTYPE_MAX];
784 struct list_head thread_group;
785 struct list_head thread_node;
786
787 struct completion *vfork_done;
788
789 /* CLONE_CHILD_SETTID: */
790 int __user *set_child_tid;
791
792 /* CLONE_CHILD_CLEARTID: */
793 int __user *clear_child_tid;
794
795 u64 utime;
796 u64 stime;
797#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
798 u64 utimescaled;
799 u64 stimescaled;
800#endif
801 u64 gtime;
802 struct prev_cputime prev_cputime;
803#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
804 struct vtime vtime;
805#endif
806
807#ifdef CONFIG_NO_HZ_FULL
808 atomic_t tick_dep_mask;
809#endif
810 /* Context switch counts: */
811 unsigned long nvcsw;
812 unsigned long nivcsw;
813
814 /* Monotonic time in nsecs: */
815 u64 start_time;
816
817 /* Boot based time in nsecs: */
818 u64 real_start_time;
819
820 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
821 unsigned long min_flt;
822 unsigned long maj_flt;
823
824#ifdef CONFIG_POSIX_TIMERS
825 struct task_cputime cputime_expires;
826 struct list_head cpu_timers[3];
827#endif
828
829 /* Process credentials: */
830
831 /* Tracer's credentials at attach: */
832 const struct cred __rcu *ptracer_cred;
833
834 /* Objective and real subjective task credentials (COW): */
835 const struct cred __rcu *real_cred;
836
837 /* Effective (overridable) subjective task credentials (COW): */
838 const struct cred __rcu *cred;
839
840 /*
841 * executable name, excluding path.
842 *
843 * - normally initialized setup_new_exec()
844 * - access it with [gs]et_task_comm()
845 * - lock it with task_lock()
846 */
847 char comm[TASK_COMM_LEN];
848
849 struct nameidata *nameidata;
850
851#ifdef CONFIG_SYSVIPC
852 struct sysv_sem sysvsem;
853 struct sysv_shm sysvshm;
854#endif
855#ifdef CONFIG_DETECT_HUNG_TASK
856 unsigned long last_switch_count;
857#endif
858 /* Filesystem information: */
859 struct fs_struct *fs;
860
861 /* Open file information: */
862 struct files_struct *files;
863
864 /* Namespaces: */
865 struct nsproxy *nsproxy;
866
867 /* Signal handlers: */
868 struct signal_struct *signal;
869 struct sighand_struct *sighand;
870 sigset_t blocked;
871 sigset_t real_blocked;
872 /* Restored if set_restore_sigmask() was used: */
873 sigset_t saved_sigmask;
874 struct sigpending pending;
875 unsigned long sas_ss_sp;
876 size_t sas_ss_size;
877 unsigned int sas_ss_flags;
878
879 struct callback_head *task_works;
880
881 struct audit_context *audit_context;
882#ifdef CONFIG_AUDITSYSCALL
883 kuid_t loginuid;
884 unsigned int sessionid;
885#endif
886 struct seccomp seccomp;
887
888 /* Thread group tracking: */
889 u32 parent_exec_id;
890 u32 self_exec_id;
891
892 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
893 spinlock_t alloc_lock;
894
895 /* Protection of the PI data structures: */
896 raw_spinlock_t pi_lock;
897
898 struct wake_q_node wake_q;
899
900#ifdef CONFIG_RT_MUTEXES
901 /* PI waiters blocked on a rt_mutex held by this task: */
902 struct rb_root_cached pi_waiters;
903 /* Updated under owner's pi_lock and rq lock */
904 struct task_struct *pi_top_task;
905 /* Deadlock detection and priority inheritance handling: */
906 struct rt_mutex_waiter *pi_blocked_on;
907#endif
908
909#ifdef CONFIG_DEBUG_MUTEXES
910 /* Mutex deadlock detection: */
911 struct mutex_waiter *blocked_on;
912#endif
913
914#ifdef CONFIG_TRACE_IRQFLAGS
915 unsigned int irq_events;
916 unsigned long hardirq_enable_ip;
917 unsigned long hardirq_disable_ip;
918 unsigned int hardirq_enable_event;
919 unsigned int hardirq_disable_event;
920 int hardirqs_enabled;
921 int hardirq_context;
922 unsigned long softirq_disable_ip;
923 unsigned long softirq_enable_ip;
924 unsigned int softirq_disable_event;
925 unsigned int softirq_enable_event;
926 int softirqs_enabled;
927 int softirq_context;
928#endif
929
930#ifdef CONFIG_LOCKDEP
931# define MAX_LOCK_DEPTH 48UL
932 u64 curr_chain_key;
933 int lockdep_depth;
934 unsigned int lockdep_recursion;
935 struct held_lock held_locks[MAX_LOCK_DEPTH];
936#endif
937
938#ifdef CONFIG_UBSAN
939 unsigned int in_ubsan;
940#endif
941
942 /* Journalling filesystem info: */
943 void *journal_info;
944
945 /* Stacked block device info: */
946 struct bio_list *bio_list;
947
948#ifdef CONFIG_BLOCK
949 /* Stack plugging: */
950 struct blk_plug *plug;
951#endif
952
953 /* VM state: */
954 struct reclaim_state *reclaim_state;
955
956 struct backing_dev_info *backing_dev_info;
957
958 struct io_context *io_context;
959
960 /* Ptrace state: */
961 unsigned long ptrace_message;
962 siginfo_t *last_siginfo;
963
964 struct task_io_accounting ioac;
965#ifdef CONFIG_TASK_XACCT
966 /* Accumulated RSS usage: */
967 u64 acct_rss_mem1;
968 /* Accumulated virtual memory usage: */
969 u64 acct_vm_mem1;
970 /* stime + utime since last update: */
971 u64 acct_timexpd;
972#endif
973#ifdef CONFIG_CPUSETS
974 /* Protected by ->alloc_lock: */
975 nodemask_t mems_allowed;
976 /* Seqence number to catch updates: */
977 seqcount_t mems_allowed_seq;
978 int cpuset_mem_spread_rotor;
979 int cpuset_slab_spread_rotor;
980#endif
981#ifdef CONFIG_CGROUPS
982 /* Control Group info protected by css_set_lock: */
983 struct css_set __rcu *cgroups;
984 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
985 struct list_head cg_list;
986#endif
987#ifdef CONFIG_INTEL_RDT
988 u32 closid;
989 u32 rmid;
990#endif
991#ifdef CONFIG_FUTEX
992 struct robust_list_head __user *robust_list;
993#ifdef CONFIG_COMPAT
994 struct compat_robust_list_head __user *compat_robust_list;
995#endif
996 struct list_head pi_state_list;
997 struct futex_pi_state *pi_state_cache;
998#endif
999#ifdef CONFIG_PERF_EVENTS
1000 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1001 struct mutex perf_event_mutex;
1002 struct list_head perf_event_list;
1003#endif
1004#ifdef CONFIG_DEBUG_PREEMPT
1005 unsigned long preempt_disable_ip;
1006#endif
1007#ifdef CONFIG_NUMA
1008 /* Protected by alloc_lock: */
1009 struct mempolicy *mempolicy;
1010 short il_prev;
1011 short pref_node_fork;
1012#endif
1013#ifdef CONFIG_NUMA_BALANCING
1014 int numa_scan_seq;
1015 unsigned int numa_scan_period;
1016 unsigned int numa_scan_period_max;
1017 int numa_preferred_nid;
1018 unsigned long numa_migrate_retry;
1019 /* Migration stamp: */
1020 u64 node_stamp;
1021 u64 last_task_numa_placement;
1022 u64 last_sum_exec_runtime;
1023 struct callback_head numa_work;
1024
1025 struct numa_group *numa_group;
1026
1027 /*
1028 * numa_faults is an array split into four regions:
1029 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1030 * in this precise order.
1031 *
1032 * faults_memory: Exponential decaying average of faults on a per-node
1033 * basis. Scheduling placement decisions are made based on these
1034 * counts. The values remain static for the duration of a PTE scan.
1035 * faults_cpu: Track the nodes the process was running on when a NUMA
1036 * hinting fault was incurred.
1037 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1038 * during the current scan window. When the scan completes, the counts
1039 * in faults_memory and faults_cpu decay and these values are copied.
1040 */
1041 unsigned long *numa_faults;
1042 unsigned long total_numa_faults;
1043
1044 /*
1045 * numa_faults_locality tracks if faults recorded during the last
1046 * scan window were remote/local or failed to migrate. The task scan
1047 * period is adapted based on the locality of the faults with different
1048 * weights depending on whether they were shared or private faults
1049 */
1050 unsigned long numa_faults_locality[3];
1051
1052 unsigned long numa_pages_migrated;
1053#endif /* CONFIG_NUMA_BALANCING */
1054
1055#ifdef CONFIG_RSEQ
1056 struct rseq __user *rseq;
1057 u32 rseq_len;
1058 u32 rseq_sig;
1059 /*
1060 * RmW on rseq_event_mask must be performed atomically
1061 * with respect to preemption.
1062 */
1063 unsigned long rseq_event_mask;
1064#endif
1065
1066 struct tlbflush_unmap_batch tlb_ubc;
1067
1068 struct rcu_head rcu;
1069
1070 /* Cache last used pipe for splice(): */
1071 struct pipe_inode_info *splice_pipe;
1072
1073 struct page_frag task_frag;
1074
1075#ifdef CONFIG_TASK_DELAY_ACCT
1076 struct task_delay_info *delays;
1077#endif
1078
1079#ifdef CONFIG_FAULT_INJECTION
1080 int make_it_fail;
1081 unsigned int fail_nth;
1082#endif
1083 /*
1084 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1085 * balance_dirty_pages() for a dirty throttling pause:
1086 */
1087 int nr_dirtied;
1088 int nr_dirtied_pause;
1089 /* Start of a write-and-pause period: */
1090 unsigned long dirty_paused_when;
1091
1092#ifdef CONFIG_LATENCYTOP
1093 int latency_record_count;
1094 struct latency_record latency_record[LT_SAVECOUNT];
1095#endif
1096 /*
1097 * Time slack values; these are used to round up poll() and
1098 * select() etc timeout values. These are in nanoseconds.
1099 */
1100 u64 timer_slack_ns;
1101 u64 default_timer_slack_ns;
1102
1103#ifdef CONFIG_KASAN
1104 unsigned int kasan_depth;
1105#endif
1106
1107#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1108 /* Index of current stored address in ret_stack: */
1109 int curr_ret_stack;
1110
1111 /* Stack of return addresses for return function tracing: */
1112 struct ftrace_ret_stack *ret_stack;
1113
1114 /* Timestamp for last schedule: */
1115 unsigned long long ftrace_timestamp;
1116
1117 /*
1118 * Number of functions that haven't been traced
1119 * because of depth overrun:
1120 */
1121 atomic_t trace_overrun;
1122
1123 /* Pause tracing: */
1124 atomic_t tracing_graph_pause;
1125#endif
1126
1127#ifdef CONFIG_TRACING
1128 /* State flags for use by tracers: */
1129 unsigned long trace;
1130
1131 /* Bitmask and counter of trace recursion: */
1132 unsigned long trace_recursion;
1133#endif /* CONFIG_TRACING */
1134
1135#ifdef CONFIG_KCOV
1136 /* Coverage collection mode enabled for this task (0 if disabled): */
1137 unsigned int kcov_mode;
1138
1139 /* Size of the kcov_area: */
1140 unsigned int kcov_size;
1141
1142 /* Buffer for coverage collection: */
1143 void *kcov_area;
1144
1145 /* KCOV descriptor wired with this task or NULL: */
1146 struct kcov *kcov;
1147#endif
1148
1149#ifdef CONFIG_MEMCG
1150 struct mem_cgroup *memcg_in_oom;
1151 gfp_t memcg_oom_gfp_mask;
1152 int memcg_oom_order;
1153
1154 /* Number of pages to reclaim on returning to userland: */
1155 unsigned int memcg_nr_pages_over_high;
1156
1157 /* Used by memcontrol for targeted memcg charge: */
1158 struct mem_cgroup *active_memcg;
1159#endif
1160
1161#ifdef CONFIG_BLK_CGROUP
1162 struct request_queue *throttle_queue;
1163#endif
1164
1165#ifdef CONFIG_UPROBES
1166 struct uprobe_task *utask;
1167#endif
1168#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1169 unsigned int sequential_io;
1170 unsigned int sequential_io_avg;
1171#endif
1172#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1173 unsigned long task_state_change;
1174#endif
1175 int pagefault_disabled;
1176#ifdef CONFIG_MMU
1177 struct task_struct *oom_reaper_list;
1178#endif
1179#ifdef CONFIG_VMAP_STACK
1180 struct vm_struct *stack_vm_area;
1181#endif
1182#ifdef CONFIG_THREAD_INFO_IN_TASK
1183 /* A live task holds one reference: */
1184 atomic_t stack_refcount;
1185#endif
1186#ifdef CONFIG_LIVEPATCH
1187 int patch_state;
1188#endif
1189#ifdef CONFIG_SECURITY
1190 /* Used by LSM modules for access restriction: */
1191 void *security;
1192#endif
1193
1194 /*
1195 * New fields for task_struct should be added above here, so that
1196 * they are included in the randomized portion of task_struct.
1197 */
1198 randomized_struct_fields_end
1199
1200 /* CPU-specific state of this task: */
1201 struct thread_struct thread;
1202
1203 /*
1204 * WARNING: on x86, 'thread_struct' contains a variable-sized
1205 * structure. It *MUST* be at the end of 'task_struct'.
1206 *
1207 * Do not put anything below here!
1208 */
1209};
1210
1211static inline struct pid *task_pid(struct task_struct *task)
1212{
1213 return task->thread_pid;
1214}
1215
1216/*
1217 * the helpers to get the task's different pids as they are seen
1218 * from various namespaces
1219 *
1220 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1221 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1222 * current.
1223 * task_xid_nr_ns() : id seen from the ns specified;
1224 *
1225 * see also pid_nr() etc in include/linux/pid.h
1226 */
1227pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1228
1229static inline pid_t task_pid_nr(struct task_struct *tsk)
1230{
1231 return tsk->pid;
1232}
1233
1234static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1235{
1236 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1237}
1238
1239static inline pid_t task_pid_vnr(struct task_struct *tsk)
1240{
1241 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1242}
1243
1244
1245static inline pid_t task_tgid_nr(struct task_struct *tsk)
1246{
1247 return tsk->tgid;
1248}
1249
1250/**
1251 * pid_alive - check that a task structure is not stale
1252 * @p: Task structure to be checked.
1253 *
1254 * Test if a process is not yet dead (at most zombie state)
1255 * If pid_alive fails, then pointers within the task structure
1256 * can be stale and must not be dereferenced.
1257 *
1258 * Return: 1 if the process is alive. 0 otherwise.
1259 */
1260static inline int pid_alive(const struct task_struct *p)
1261{
1262 return p->thread_pid != NULL;
1263}
1264
1265static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1266{
1267 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1268}
1269
1270static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1271{
1272 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1273}
1274
1275
1276static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1277{
1278 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1279}
1280
1281static inline pid_t task_session_vnr(struct task_struct *tsk)
1282{
1283 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1284}
1285
1286static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1287{
1288 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1289}
1290
1291static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1292{
1293 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1294}
1295
1296static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1297{
1298 pid_t pid = 0;
1299
1300 rcu_read_lock();
1301 if (pid_alive(tsk))
1302 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1303 rcu_read_unlock();
1304
1305 return pid;
1306}
1307
1308static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1309{
1310 return task_ppid_nr_ns(tsk, &init_pid_ns);
1311}
1312
1313/* Obsolete, do not use: */
1314static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1315{
1316 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1317}
1318
1319#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1320#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1321
1322static inline unsigned int task_state_index(struct task_struct *tsk)
1323{
1324 unsigned int tsk_state = READ_ONCE(tsk->state);
1325 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1326
1327 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1328
1329 if (tsk_state == TASK_IDLE)
1330 state = TASK_REPORT_IDLE;
1331
1332 return fls(state);
1333}
1334
1335static inline char task_index_to_char(unsigned int state)
1336{
1337 static const char state_char[] = "RSDTtXZPI";
1338
1339 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1340
1341 return state_char[state];
1342}
1343
1344static inline char task_state_to_char(struct task_struct *tsk)
1345{
1346 return task_index_to_char(task_state_index(tsk));
1347}
1348
1349/**
1350 * is_global_init - check if a task structure is init. Since init
1351 * is free to have sub-threads we need to check tgid.
1352 * @tsk: Task structure to be checked.
1353 *
1354 * Check if a task structure is the first user space task the kernel created.
1355 *
1356 * Return: 1 if the task structure is init. 0 otherwise.
1357 */
1358static inline int is_global_init(struct task_struct *tsk)
1359{
1360 return task_tgid_nr(tsk) == 1;
1361}
1362
1363extern struct pid *cad_pid;
1364
1365/*
1366 * Per process flags
1367 */
1368#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1369#define PF_EXITING 0x00000004 /* Getting shut down */
1370#define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1371#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1372#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1373#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1374#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1375#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1376#define PF_DUMPCORE 0x00000200 /* Dumped core */
1377#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1378#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1379#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1380#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1381#define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1382#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1383#define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1384#define PF_KSWAPD 0x00020000 /* I am kswapd */
1385#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1386#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1387#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1388#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1389#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1390#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1391#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1392#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1393#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1394#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1395#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1396
1397/*
1398 * Only the _current_ task can read/write to tsk->flags, but other
1399 * tasks can access tsk->flags in readonly mode for example
1400 * with tsk_used_math (like during threaded core dumping).
1401 * There is however an exception to this rule during ptrace
1402 * or during fork: the ptracer task is allowed to write to the
1403 * child->flags of its traced child (same goes for fork, the parent
1404 * can write to the child->flags), because we're guaranteed the
1405 * child is not running and in turn not changing child->flags
1406 * at the same time the parent does it.
1407 */
1408#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1409#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1410#define clear_used_math() clear_stopped_child_used_math(current)
1411#define set_used_math() set_stopped_child_used_math(current)
1412
1413#define conditional_stopped_child_used_math(condition, child) \
1414 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1415
1416#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1417
1418#define copy_to_stopped_child_used_math(child) \
1419 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1420
1421/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1422#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1423#define used_math() tsk_used_math(current)
1424
1425static inline bool is_percpu_thread(void)
1426{
1427#ifdef CONFIG_SMP
1428 return (current->flags & PF_NO_SETAFFINITY) &&
1429 (current->nr_cpus_allowed == 1);
1430#else
1431 return true;
1432#endif
1433}
1434
1435/* Per-process atomic flags. */
1436#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1437#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1438#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1439#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1440#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1441
1442#define TASK_PFA_TEST(name, func) \
1443 static inline bool task_##func(struct task_struct *p) \
1444 { return test_bit(PFA_##name, &p->atomic_flags); }
1445
1446#define TASK_PFA_SET(name, func) \
1447 static inline void task_set_##func(struct task_struct *p) \
1448 { set_bit(PFA_##name, &p->atomic_flags); }
1449
1450#define TASK_PFA_CLEAR(name, func) \
1451 static inline void task_clear_##func(struct task_struct *p) \
1452 { clear_bit(PFA_##name, &p->atomic_flags); }
1453
1454TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1455TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1456
1457TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1458TASK_PFA_SET(SPREAD_PAGE, spread_page)
1459TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1460
1461TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1462TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1463TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1464
1465TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1466TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1467TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1468
1469TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1470TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1471
1472static inline void
1473current_restore_flags(unsigned long orig_flags, unsigned long flags)
1474{
1475 current->flags &= ~flags;
1476 current->flags |= orig_flags & flags;
1477}
1478
1479extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1480extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1481#ifdef CONFIG_SMP
1482extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1483extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1484#else
1485static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1486{
1487}
1488static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1489{
1490 if (!cpumask_test_cpu(0, new_mask))
1491 return -EINVAL;
1492 return 0;
1493}
1494#endif
1495
1496#ifndef cpu_relax_yield
1497#define cpu_relax_yield() cpu_relax()
1498#endif
1499
1500extern int yield_to(struct task_struct *p, bool preempt);
1501extern void set_user_nice(struct task_struct *p, long nice);
1502extern int task_prio(const struct task_struct *p);
1503
1504/**
1505 * task_nice - return the nice value of a given task.
1506 * @p: the task in question.
1507 *
1508 * Return: The nice value [ -20 ... 0 ... 19 ].
1509 */
1510static inline int task_nice(const struct task_struct *p)
1511{
1512 return PRIO_TO_NICE((p)->static_prio);
1513}
1514
1515extern int can_nice(const struct task_struct *p, const int nice);
1516extern int task_curr(const struct task_struct *p);
1517extern int idle_cpu(int cpu);
1518extern int available_idle_cpu(int cpu);
1519extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1520extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1521extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1522extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1523extern struct task_struct *idle_task(int cpu);
1524
1525/**
1526 * is_idle_task - is the specified task an idle task?
1527 * @p: the task in question.
1528 *
1529 * Return: 1 if @p is an idle task. 0 otherwise.
1530 */
1531static inline bool is_idle_task(const struct task_struct *p)
1532{
1533 return !!(p->flags & PF_IDLE);
1534}
1535
1536extern struct task_struct *curr_task(int cpu);
1537extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1538
1539void yield(void);
1540
1541union thread_union {
1542#ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1543 struct task_struct task;
1544#endif
1545#ifndef CONFIG_THREAD_INFO_IN_TASK
1546 struct thread_info thread_info;
1547#endif
1548 unsigned long stack[THREAD_SIZE/sizeof(long)];
1549};
1550
1551#ifndef CONFIG_THREAD_INFO_IN_TASK
1552extern struct thread_info init_thread_info;
1553#endif
1554
1555extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1556
1557#ifdef CONFIG_THREAD_INFO_IN_TASK
1558static inline struct thread_info *task_thread_info(struct task_struct *task)
1559{
1560 return &task->thread_info;
1561}
1562#elif !defined(__HAVE_THREAD_FUNCTIONS)
1563# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1564#endif
1565
1566/*
1567 * find a task by one of its numerical ids
1568 *
1569 * find_task_by_pid_ns():
1570 * finds a task by its pid in the specified namespace
1571 * find_task_by_vpid():
1572 * finds a task by its virtual pid
1573 *
1574 * see also find_vpid() etc in include/linux/pid.h
1575 */
1576
1577extern struct task_struct *find_task_by_vpid(pid_t nr);
1578extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1579
1580/*
1581 * find a task by its virtual pid and get the task struct
1582 */
1583extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1584
1585extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1586extern int wake_up_process(struct task_struct *tsk);
1587extern void wake_up_new_task(struct task_struct *tsk);
1588
1589#ifdef CONFIG_SMP
1590extern void kick_process(struct task_struct *tsk);
1591#else
1592static inline void kick_process(struct task_struct *tsk) { }
1593#endif
1594
1595extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1596
1597static inline void set_task_comm(struct task_struct *tsk, const char *from)
1598{
1599 __set_task_comm(tsk, from, false);
1600}
1601
1602extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1603#define get_task_comm(buf, tsk) ({ \
1604 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1605 __get_task_comm(buf, sizeof(buf), tsk); \
1606})
1607
1608#ifdef CONFIG_SMP
1609void scheduler_ipi(void);
1610extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1611#else
1612static inline void scheduler_ipi(void) { }
1613static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1614{
1615 return 1;
1616}
1617#endif
1618
1619/*
1620 * Set thread flags in other task's structures.
1621 * See asm/thread_info.h for TIF_xxxx flags available:
1622 */
1623static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1624{
1625 set_ti_thread_flag(task_thread_info(tsk), flag);
1626}
1627
1628static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1629{
1630 clear_ti_thread_flag(task_thread_info(tsk), flag);
1631}
1632
1633static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1634 bool value)
1635{
1636 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1637}
1638
1639static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1640{
1641 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1642}
1643
1644static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1645{
1646 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1647}
1648
1649static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1650{
1651 return test_ti_thread_flag(task_thread_info(tsk), flag);
1652}
1653
1654static inline void set_tsk_need_resched(struct task_struct *tsk)
1655{
1656 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1657}
1658
1659static inline void clear_tsk_need_resched(struct task_struct *tsk)
1660{
1661 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1662}
1663
1664static inline int test_tsk_need_resched(struct task_struct *tsk)
1665{
1666 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1667}
1668
1669/*
1670 * cond_resched() and cond_resched_lock(): latency reduction via
1671 * explicit rescheduling in places that are safe. The return
1672 * value indicates whether a reschedule was done in fact.
1673 * cond_resched_lock() will drop the spinlock before scheduling,
1674 */
1675#ifndef CONFIG_PREEMPT
1676extern int _cond_resched(void);
1677#else
1678static inline int _cond_resched(void) { return 0; }
1679#endif
1680
1681#define cond_resched() ({ \
1682 ___might_sleep(__FILE__, __LINE__, 0); \
1683 _cond_resched(); \
1684})
1685
1686extern int __cond_resched_lock(spinlock_t *lock);
1687
1688#define cond_resched_lock(lock) ({ \
1689 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1690 __cond_resched_lock(lock); \
1691})
1692
1693static inline void cond_resched_rcu(void)
1694{
1695#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1696 rcu_read_unlock();
1697 cond_resched();
1698 rcu_read_lock();
1699#endif
1700}
1701
1702/*
1703 * Does a critical section need to be broken due to another
1704 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1705 * but a general need for low latency)
1706 */
1707static inline int spin_needbreak(spinlock_t *lock)
1708{
1709#ifdef CONFIG_PREEMPT
1710 return spin_is_contended(lock);
1711#else
1712 return 0;
1713#endif
1714}
1715
1716static __always_inline bool need_resched(void)
1717{
1718 return unlikely(tif_need_resched());
1719}
1720
1721/*
1722 * Wrappers for p->thread_info->cpu access. No-op on UP.
1723 */
1724#ifdef CONFIG_SMP
1725
1726static inline unsigned int task_cpu(const struct task_struct *p)
1727{
1728#ifdef CONFIG_THREAD_INFO_IN_TASK
1729 return p->cpu;
1730#else
1731 return task_thread_info(p)->cpu;
1732#endif
1733}
1734
1735extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1736
1737#else
1738
1739static inline unsigned int task_cpu(const struct task_struct *p)
1740{
1741 return 0;
1742}
1743
1744static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1745{
1746}
1747
1748#endif /* CONFIG_SMP */
1749
1750/*
1751 * In order to reduce various lock holder preemption latencies provide an
1752 * interface to see if a vCPU is currently running or not.
1753 *
1754 * This allows us to terminate optimistic spin loops and block, analogous to
1755 * the native optimistic spin heuristic of testing if the lock owner task is
1756 * running or not.
1757 */
1758#ifndef vcpu_is_preempted
1759# define vcpu_is_preempted(cpu) false
1760#endif
1761
1762extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1763extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1764
1765#ifndef TASK_SIZE_OF
1766#define TASK_SIZE_OF(tsk) TASK_SIZE
1767#endif
1768
1769#ifdef CONFIG_RSEQ
1770
1771/*
1772 * Map the event mask on the user-space ABI enum rseq_cs_flags
1773 * for direct mask checks.
1774 */
1775enum rseq_event_mask_bits {
1776 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1777 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1778 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1779};
1780
1781enum rseq_event_mask {
1782 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1783 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1784 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1785};
1786
1787static inline void rseq_set_notify_resume(struct task_struct *t)
1788{
1789 if (t->rseq)
1790 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1791}
1792
1793void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1794
1795static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1796 struct pt_regs *regs)
1797{
1798 if (current->rseq)
1799 __rseq_handle_notify_resume(ksig, regs);
1800}
1801
1802static inline void rseq_signal_deliver(struct ksignal *ksig,
1803 struct pt_regs *regs)
1804{
1805 preempt_disable();
1806 __set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1807 preempt_enable();
1808 rseq_handle_notify_resume(ksig, regs);
1809}
1810
1811/* rseq_preempt() requires preemption to be disabled. */
1812static inline void rseq_preempt(struct task_struct *t)
1813{
1814 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1815 rseq_set_notify_resume(t);
1816}
1817
1818/* rseq_migrate() requires preemption to be disabled. */
1819static inline void rseq_migrate(struct task_struct *t)
1820{
1821 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1822 rseq_set_notify_resume(t);
1823}
1824
1825/*
1826 * If parent process has a registered restartable sequences area, the
1827 * child inherits. Only applies when forking a process, not a thread.
1828 */
1829static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1830{
1831 if (clone_flags & CLONE_THREAD) {
1832 t->rseq = NULL;
1833 t->rseq_len = 0;
1834 t->rseq_sig = 0;
1835 t->rseq_event_mask = 0;
1836 } else {
1837 t->rseq = current->rseq;
1838 t->rseq_len = current->rseq_len;
1839 t->rseq_sig = current->rseq_sig;
1840 t->rseq_event_mask = current->rseq_event_mask;
1841 }
1842}
1843
1844static inline void rseq_execve(struct task_struct *t)
1845{
1846 t->rseq = NULL;
1847 t->rseq_len = 0;
1848 t->rseq_sig = 0;
1849 t->rseq_event_mask = 0;
1850}
1851
1852#else
1853
1854static inline void rseq_set_notify_resume(struct task_struct *t)
1855{
1856}
1857static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1858 struct pt_regs *regs)
1859{
1860}
1861static inline void rseq_signal_deliver(struct ksignal *ksig,
1862 struct pt_regs *regs)
1863{
1864}
1865static inline void rseq_preempt(struct task_struct *t)
1866{
1867}
1868static inline void rseq_migrate(struct task_struct *t)
1869{
1870}
1871static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1872{
1873}
1874static inline void rseq_execve(struct task_struct *t)
1875{
1876}
1877
1878#endif
1879
1880#ifdef CONFIG_DEBUG_RSEQ
1881
1882void rseq_syscall(struct pt_regs *regs);
1883
1884#else
1885
1886static inline void rseq_syscall(struct pt_regs *regs)
1887{
1888}
1889
1890#endif
1891
1892#endif
1893