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/refcount.h>
25#include <linux/resource.h>
26#include <linux/latencytop.h>
27#include <linux/sched/prio.h>
28#include <linux/signal_types.h>
29#include <linux/psi_types.h>
30#include <linux/mm_types_task.h>
31#include <linux/task_io_accounting.h>
32#include <linux/rseq.h>
33
34/* task_struct member predeclarations (sorted alphabetically): */
35struct audit_context;
36struct backing_dev_info;
37struct bio_list;
38struct blk_plug;
39struct cfs_rq;
40struct fs_struct;
41struct futex_pi_state;
42struct io_context;
43struct mempolicy;
44struct nameidata;
45struct nsproxy;
46struct perf_event_context;
47struct pid_namespace;
48struct pipe_inode_info;
49struct rcu_node;
50struct reclaim_state;
51struct capture_control;
52struct robust_list_head;
53struct sched_attr;
54struct sched_param;
55struct seq_file;
56struct sighand_struct;
57struct signal_struct;
58struct task_delay_info;
59struct task_group;
60
61/*
62 * Task state bitmask. NOTE! These bits are also
63 * encoded in fs/proc/array.c: get_task_state().
64 *
65 * We have two separate sets of flags: task->state
66 * is about runnability, while task->exit_state are
67 * about the task exiting. Confusing, but this way
68 * modifying one set can't modify the other one by
69 * mistake.
70 */
71
72/* Used in tsk->state: */
73#define TASK_RUNNING 0x0000
74#define TASK_INTERRUPTIBLE 0x0001
75#define TASK_UNINTERRUPTIBLE 0x0002
76#define __TASK_STOPPED 0x0004
77#define __TASK_TRACED 0x0008
78/* Used in tsk->exit_state: */
79#define EXIT_DEAD 0x0010
80#define EXIT_ZOMBIE 0x0020
81#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
82/* Used in tsk->state again: */
83#define TASK_PARKED 0x0040
84#define TASK_DEAD 0x0080
85#define TASK_WAKEKILL 0x0100
86#define TASK_WAKING 0x0200
87#define TASK_NOLOAD 0x0400
88#define TASK_NEW 0x0800
89#define TASK_STATE_MAX 0x1000
90
91/* Convenience macros for the sake of set_current_state: */
92#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
93#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
94#define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
95
96#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
97
98/* Convenience macros for the sake of wake_up(): */
99#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
100
101/* get_task_state(): */
102#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
103 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
104 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
105 TASK_PARKED)
106
107#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
108
109#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
110
111#define task_is_stopped_or_traced(task) ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
112
113#define task_contributes_to_load(task) ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
114 (task->flags & PF_FROZEN) == 0 && \
115 (task->state & TASK_NOLOAD) == 0)
116
117#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
118
119/*
120 * Special states are those that do not use the normal wait-loop pattern. See
121 * the comment with set_special_state().
122 */
123#define is_special_task_state(state) \
124 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
125
126#define __set_current_state(state_value) \
127 do { \
128 WARN_ON_ONCE(is_special_task_state(state_value));\
129 current->task_state_change = _THIS_IP_; \
130 current->state = (state_value); \
131 } while (0)
132
133#define set_current_state(state_value) \
134 do { \
135 WARN_ON_ONCE(is_special_task_state(state_value));\
136 current->task_state_change = _THIS_IP_; \
137 smp_store_mb(current->state, (state_value)); \
138 } while (0)
139
140#define set_special_state(state_value) \
141 do { \
142 unsigned long flags; /* may shadow */ \
143 WARN_ON_ONCE(!is_special_task_state(state_value)); \
144 raw_spin_lock_irqsave(&current->pi_lock, flags); \
145 current->task_state_change = _THIS_IP_; \
146 current->state = (state_value); \
147 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
148 } while (0)
149#else
150/*
151 * set_current_state() includes a barrier so that the write of current->state
152 * is correctly serialised wrt the caller's subsequent test of whether to
153 * actually sleep:
154 *
155 * for (;;) {
156 * set_current_state(TASK_UNINTERRUPTIBLE);
157 * if (!need_sleep)
158 * break;
159 *
160 * schedule();
161 * }
162 * __set_current_state(TASK_RUNNING);
163 *
164 * If the caller does not need such serialisation (because, for instance, the
165 * condition test and condition change and wakeup are under the same lock) then
166 * use __set_current_state().
167 *
168 * The above is typically ordered against the wakeup, which does:
169 *
170 * need_sleep = false;
171 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
172 *
173 * where wake_up_state() executes a full memory barrier before accessing the
174 * task state.
175 *
176 * Wakeup will do: if (@state & p->state) p->state = TASK_RUNNING, that is,
177 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
178 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
179 *
180 * However, with slightly different timing the wakeup TASK_RUNNING store can
181 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
182 * a problem either because that will result in one extra go around the loop
183 * and our @cond test will save the day.
184 *
185 * Also see the comments of try_to_wake_up().
186 */
187#define __set_current_state(state_value) \
188 current->state = (state_value)
189
190#define set_current_state(state_value) \
191 smp_store_mb(current->state, (state_value))
192
193/*
194 * set_special_state() should be used for those states when the blocking task
195 * can not use the regular condition based wait-loop. In that case we must
196 * serialize against wakeups such that any possible in-flight TASK_RUNNING stores
197 * will not collide with our state change.
198 */
199#define set_special_state(state_value) \
200 do { \
201 unsigned long flags; /* may shadow */ \
202 raw_spin_lock_irqsave(&current->pi_lock, flags); \
203 current->state = (state_value); \
204 raw_spin_unlock_irqrestore(&current->pi_lock, flags); \
205 } while (0)
206
207#endif
208
209/* Task command name length: */
210#define TASK_COMM_LEN 16
211
212extern void scheduler_tick(void);
213
214#define MAX_SCHEDULE_TIMEOUT LONG_MAX
215
216extern long schedule_timeout(long timeout);
217extern long schedule_timeout_interruptible(long timeout);
218extern long schedule_timeout_killable(long timeout);
219extern long schedule_timeout_uninterruptible(long timeout);
220extern long schedule_timeout_idle(long timeout);
221asmlinkage void schedule(void);
222extern void schedule_preempt_disabled(void);
223
224extern int __must_check io_schedule_prepare(void);
225extern void io_schedule_finish(int token);
226extern long io_schedule_timeout(long timeout);
227extern void io_schedule(void);
228
229/**
230 * struct prev_cputime - snapshot of system and user cputime
231 * @utime: time spent in user mode
232 * @stime: time spent in system mode
233 * @lock: protects the above two fields
234 *
235 * Stores previous user/system time values such that we can guarantee
236 * monotonicity.
237 */
238struct prev_cputime {
239#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
240 u64 utime;
241 u64 stime;
242 raw_spinlock_t lock;
243#endif
244};
245
246/**
247 * struct task_cputime - collected CPU time counts
248 * @utime: time spent in user mode, in nanoseconds
249 * @stime: time spent in kernel mode, in nanoseconds
250 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
251 *
252 * This structure groups together three kinds of CPU time that are tracked for
253 * threads and thread groups. Most things considering CPU time want to group
254 * these counts together and treat all three of them in parallel.
255 */
256struct task_cputime {
257 u64 utime;
258 u64 stime;
259 unsigned long long sum_exec_runtime;
260};
261
262/* Alternate field names when used on cache expirations: */
263#define virt_exp utime
264#define prof_exp stime
265#define sched_exp sum_exec_runtime
266
267enum vtime_state {
268 /* Task is sleeping or running in a CPU with VTIME inactive: */
269 VTIME_INACTIVE = 0,
270 /* Task runs in userspace in a CPU with VTIME active: */
271 VTIME_USER,
272 /* Task runs in kernelspace in a CPU with VTIME active: */
273 VTIME_SYS,
274};
275
276struct vtime {
277 seqcount_t seqcount;
278 unsigned long long starttime;
279 enum vtime_state state;
280 u64 utime;
281 u64 stime;
282 u64 gtime;
283};
284
285struct sched_info {
286#ifdef CONFIG_SCHED_INFO
287 /* Cumulative counters: */
288
289 /* # of times we have run on this CPU: */
290 unsigned long pcount;
291
292 /* Time spent waiting on a runqueue: */
293 unsigned long long run_delay;
294
295 /* Timestamps: */
296
297 /* When did we last run on a CPU? */
298 unsigned long long last_arrival;
299
300 /* When were we last queued to run? */
301 unsigned long long last_queued;
302
303#endif /* CONFIG_SCHED_INFO */
304};
305
306/*
307 * Integer metrics need fixed point arithmetic, e.g., sched/fair
308 * has a few: load, load_avg, util_avg, freq, and capacity.
309 *
310 * We define a basic fixed point arithmetic range, and then formalize
311 * all these metrics based on that basic range.
312 */
313# define SCHED_FIXEDPOINT_SHIFT 10
314# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
315
316struct load_weight {
317 unsigned long weight;
318 u32 inv_weight;
319};
320
321/**
322 * struct util_est - Estimation utilization of FAIR tasks
323 * @enqueued: instantaneous estimated utilization of a task/cpu
324 * @ewma: the Exponential Weighted Moving Average (EWMA)
325 * utilization of a task
326 *
327 * Support data structure to track an Exponential Weighted Moving Average
328 * (EWMA) of a FAIR task's utilization. New samples are added to the moving
329 * average each time a task completes an activation. Sample's weight is chosen
330 * so that the EWMA will be relatively insensitive to transient changes to the
331 * task's workload.
332 *
333 * The enqueued attribute has a slightly different meaning for tasks and cpus:
334 * - task: the task's util_avg at last task dequeue time
335 * - cfs_rq: the sum of util_est.enqueued for each RUNNABLE task on that CPU
336 * Thus, the util_est.enqueued of a task represents the contribution on the
337 * estimated utilization of the CPU where that task is currently enqueued.
338 *
339 * Only for tasks we track a moving average of the past instantaneous
340 * estimated utilization. This allows to absorb sporadic drops in utilization
341 * of an otherwise almost periodic task.
342 */
343struct util_est {
344 unsigned int enqueued;
345 unsigned int ewma;
346#define UTIL_EST_WEIGHT_SHIFT 2
347} __attribute__((__aligned__(sizeof(u64))));
348
349/*
350 * The load_avg/util_avg accumulates an infinite geometric series
351 * (see __update_load_avg() in kernel/sched/fair.c).
352 *
353 * [load_avg definition]
354 *
355 * load_avg = runnable% * scale_load_down(load)
356 *
357 * where runnable% is the time ratio that a sched_entity is runnable.
358 * For cfs_rq, it is the aggregated load_avg of all runnable and
359 * blocked sched_entities.
360 *
361 * [util_avg definition]
362 *
363 * util_avg = running% * SCHED_CAPACITY_SCALE
364 *
365 * where running% is the time ratio that a sched_entity is running on
366 * a CPU. For cfs_rq, it is the aggregated util_avg of all runnable
367 * and blocked sched_entities.
368 *
369 * load_avg and util_avg don't direcly factor frequency scaling and CPU
370 * capacity scaling. The scaling is done through the rq_clock_pelt that
371 * is used for computing those signals (see update_rq_clock_pelt())
372 *
373 * N.B., the above ratios (runnable% and running%) themselves are in the
374 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
375 * to as large a range as necessary. This is for example reflected by
376 * util_avg's SCHED_CAPACITY_SCALE.
377 *
378 * [Overflow issue]
379 *
380 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
381 * with the highest load (=88761), always runnable on a single cfs_rq,
382 * and should not overflow as the number already hits PID_MAX_LIMIT.
383 *
384 * For all other cases (including 32-bit kernels), struct load_weight's
385 * weight will overflow first before we do, because:
386 *
387 * Max(load_avg) <= Max(load.weight)
388 *
389 * Then it is the load_weight's responsibility to consider overflow
390 * issues.
391 */
392struct sched_avg {
393 u64 last_update_time;
394 u64 load_sum;
395 u64 runnable_load_sum;
396 u32 util_sum;
397 u32 period_contrib;
398 unsigned long load_avg;
399 unsigned long runnable_load_avg;
400 unsigned long util_avg;
401 struct util_est util_est;
402} ____cacheline_aligned;
403
404struct sched_statistics {
405#ifdef CONFIG_SCHEDSTATS
406 u64 wait_start;
407 u64 wait_max;
408 u64 wait_count;
409 u64 wait_sum;
410 u64 iowait_count;
411 u64 iowait_sum;
412
413 u64 sleep_start;
414 u64 sleep_max;
415 s64 sum_sleep_runtime;
416
417 u64 block_start;
418 u64 block_max;
419 u64 exec_max;
420 u64 slice_max;
421
422 u64 nr_migrations_cold;
423 u64 nr_failed_migrations_affine;
424 u64 nr_failed_migrations_running;
425 u64 nr_failed_migrations_hot;
426 u64 nr_forced_migrations;
427
428 u64 nr_wakeups;
429 u64 nr_wakeups_sync;
430 u64 nr_wakeups_migrate;
431 u64 nr_wakeups_local;
432 u64 nr_wakeups_remote;
433 u64 nr_wakeups_affine;
434 u64 nr_wakeups_affine_attempts;
435 u64 nr_wakeups_passive;
436 u64 nr_wakeups_idle;
437#endif
438};
439
440struct sched_entity {
441 /* For load-balancing: */
442 struct load_weight load;
443 unsigned long runnable_weight;
444 struct rb_node run_node;
445 struct list_head group_node;
446 unsigned int on_rq;
447
448 u64 exec_start;
449 u64 sum_exec_runtime;
450 u64 vruntime;
451 u64 prev_sum_exec_runtime;
452
453 u64 nr_migrations;
454
455 struct sched_statistics statistics;
456
457#ifdef CONFIG_FAIR_GROUP_SCHED
458 int depth;
459 struct sched_entity *parent;
460 /* rq on which this entity is (to be) queued: */
461 struct cfs_rq *cfs_rq;
462 /* rq "owned" by this entity/group: */
463 struct cfs_rq *my_q;
464#endif
465
466#ifdef CONFIG_SMP
467 /*
468 * Per entity load average tracking.
469 *
470 * Put into separate cache line so it does not
471 * collide with read-mostly values above.
472 */
473 struct sched_avg avg;
474#endif
475};
476
477struct sched_rt_entity {
478 struct list_head run_list;
479 unsigned long timeout;
480 unsigned long watchdog_stamp;
481 unsigned int time_slice;
482 unsigned short on_rq;
483 unsigned short on_list;
484
485 struct sched_rt_entity *back;
486#ifdef CONFIG_RT_GROUP_SCHED
487 struct sched_rt_entity *parent;
488 /* rq on which this entity is (to be) queued: */
489 struct rt_rq *rt_rq;
490 /* rq "owned" by this entity/group: */
491 struct rt_rq *my_q;
492#endif
493} __randomize_layout;
494
495struct sched_dl_entity {
496 struct rb_node rb_node;
497
498 /*
499 * Original scheduling parameters. Copied here from sched_attr
500 * during sched_setattr(), they will remain the same until
501 * the next sched_setattr().
502 */
503 u64 dl_runtime; /* Maximum runtime for each instance */
504 u64 dl_deadline; /* Relative deadline of each instance */
505 u64 dl_period; /* Separation of two instances (period) */
506 u64 dl_bw; /* dl_runtime / dl_period */
507 u64 dl_density; /* dl_runtime / dl_deadline */
508
509 /*
510 * Actual scheduling parameters. Initialized with the values above,
511 * they are continuously updated during task execution. Note that
512 * the remaining runtime could be < 0 in case we are in overrun.
513 */
514 s64 runtime; /* Remaining runtime for this instance */
515 u64 deadline; /* Absolute deadline for this instance */
516 unsigned int flags; /* Specifying the scheduler behaviour */
517
518 /*
519 * Some bool flags:
520 *
521 * @dl_throttled tells if we exhausted the runtime. If so, the
522 * task has to wait for a replenishment to be performed at the
523 * next firing of dl_timer.
524 *
525 * @dl_boosted tells if we are boosted due to DI. If so we are
526 * outside bandwidth enforcement mechanism (but only until we
527 * exit the critical section);
528 *
529 * @dl_yielded tells if task gave up the CPU before consuming
530 * all its available runtime during the last job.
531 *
532 * @dl_non_contending tells if the task is inactive while still
533 * contributing to the active utilization. In other words, it
534 * indicates if the inactive timer has been armed and its handler
535 * has not been executed yet. This flag is useful to avoid race
536 * conditions between the inactive timer handler and the wakeup
537 * code.
538 *
539 * @dl_overrun tells if the task asked to be informed about runtime
540 * overruns.
541 */
542 unsigned int dl_throttled : 1;
543 unsigned int dl_boosted : 1;
544 unsigned int dl_yielded : 1;
545 unsigned int dl_non_contending : 1;
546 unsigned int dl_overrun : 1;
547
548 /*
549 * Bandwidth enforcement timer. Each -deadline task has its
550 * own bandwidth to be enforced, thus we need one timer per task.
551 */
552 struct hrtimer dl_timer;
553
554 /*
555 * Inactive timer, responsible for decreasing the active utilization
556 * at the "0-lag time". When a -deadline task blocks, it contributes
557 * to GRUB's active utilization until the "0-lag time", hence a
558 * timer is needed to decrease the active utilization at the correct
559 * time.
560 */
561 struct hrtimer inactive_timer;
562};
563
564union rcu_special {
565 struct {
566 u8 blocked;
567 u8 need_qs;
568 u8 exp_hint; /* Hint for performance. */
569 u8 pad; /* No garbage from compiler! */
570 } b; /* Bits. */
571 u32 s; /* Set of bits. */
572};
573
574enum perf_event_task_context {
575 perf_invalid_context = -1,
576 perf_hw_context = 0,
577 perf_sw_context,
578 perf_nr_task_contexts,
579};
580
581struct wake_q_node {
582 struct wake_q_node *next;
583};
584
585struct task_struct {
586#ifdef CONFIG_THREAD_INFO_IN_TASK
587 /*
588 * For reasons of header soup (see current_thread_info()), this
589 * must be the first element of task_struct.
590 */
591 struct thread_info thread_info;
592#endif
593 /* -1 unrunnable, 0 runnable, >0 stopped: */
594 volatile long state;
595
596 /*
597 * This begins the randomizable portion of task_struct. Only
598 * scheduling-critical items should be added above here.
599 */
600 randomized_struct_fields_start
601
602 void *stack;
603 refcount_t usage;
604 /* Per task flags (PF_*), defined further below: */
605 unsigned int flags;
606 unsigned int ptrace;
607
608#ifdef CONFIG_SMP
609 struct llist_node wake_entry;
610 int on_cpu;
611#ifdef CONFIG_THREAD_INFO_IN_TASK
612 /* Current CPU: */
613 unsigned int cpu;
614#endif
615 unsigned int wakee_flips;
616 unsigned long wakee_flip_decay_ts;
617 struct task_struct *last_wakee;
618
619 /*
620 * recent_used_cpu is initially set as the last CPU used by a task
621 * that wakes affine another task. Waker/wakee relationships can
622 * push tasks around a CPU where each wakeup moves to the next one.
623 * Tracking a recently used CPU allows a quick search for a recently
624 * used CPU that may be idle.
625 */
626 int recent_used_cpu;
627 int wake_cpu;
628#endif
629 int on_rq;
630
631 int prio;
632 int static_prio;
633 int normal_prio;
634 unsigned int rt_priority;
635
636 const struct sched_class *sched_class;
637 struct sched_entity se;
638 struct sched_rt_entity rt;
639#ifdef CONFIG_CGROUP_SCHED
640 struct task_group *sched_task_group;
641#endif
642 struct sched_dl_entity dl;
643
644#ifdef CONFIG_PREEMPT_NOTIFIERS
645 /* List of struct preempt_notifier: */
646 struct hlist_head preempt_notifiers;
647#endif
648
649#ifdef CONFIG_BLK_DEV_IO_TRACE
650 unsigned int btrace_seq;
651#endif
652
653 unsigned int policy;
654 int nr_cpus_allowed;
655 cpumask_t cpus_allowed;
656
657#ifdef CONFIG_PREEMPT_RCU
658 int rcu_read_lock_nesting;
659 union rcu_special rcu_read_unlock_special;
660 struct list_head rcu_node_entry;
661 struct rcu_node *rcu_blocked_node;
662#endif /* #ifdef CONFIG_PREEMPT_RCU */
663
664#ifdef CONFIG_TASKS_RCU
665 unsigned long rcu_tasks_nvcsw;
666 u8 rcu_tasks_holdout;
667 u8 rcu_tasks_idx;
668 int rcu_tasks_idle_cpu;
669 struct list_head rcu_tasks_holdout_list;
670#endif /* #ifdef CONFIG_TASKS_RCU */
671
672 struct sched_info sched_info;
673
674 struct list_head tasks;
675#ifdef CONFIG_SMP
676 struct plist_node pushable_tasks;
677 struct rb_node pushable_dl_tasks;
678#endif
679
680 struct mm_struct *mm;
681 struct mm_struct *active_mm;
682
683 /* Per-thread vma caching: */
684 struct vmacache vmacache;
685
686#ifdef SPLIT_RSS_COUNTING
687 struct task_rss_stat rss_stat;
688#endif
689 int exit_state;
690 int exit_code;
691 int exit_signal;
692 /* The signal sent when the parent dies: */
693 int pdeath_signal;
694 /* JOBCTL_*, siglock protected: */
695 unsigned long jobctl;
696
697 /* Used for emulating ABI behavior of previous Linux versions: */
698 unsigned int personality;
699
700 /* Scheduler bits, serialized by scheduler locks: */
701 unsigned sched_reset_on_fork:1;
702 unsigned sched_contributes_to_load:1;
703 unsigned sched_migrated:1;
704 unsigned sched_remote_wakeup:1;
705#ifdef CONFIG_PSI
706 unsigned sched_psi_wake_requeue:1;
707#endif
708
709 /* Force alignment to the next boundary: */
710 unsigned :0;
711
712 /* Unserialized, strictly 'current' */
713
714 /* Bit to tell LSMs we're in execve(): */
715 unsigned in_execve:1;
716 unsigned in_iowait:1;
717#ifndef TIF_RESTORE_SIGMASK
718 unsigned restore_sigmask:1;
719#endif
720#ifdef CONFIG_MEMCG
721 unsigned in_user_fault:1;
722#endif
723#ifdef CONFIG_COMPAT_BRK
724 unsigned brk_randomized:1;
725#endif
726#ifdef CONFIG_CGROUPS
727 /* disallow userland-initiated cgroup migration */
728 unsigned no_cgroup_migration:1;
729#endif
730#ifdef CONFIG_BLK_CGROUP
731 /* to be used once the psi infrastructure lands upstream. */
732 unsigned use_memdelay:1;
733#endif
734
735 unsigned long atomic_flags; /* Flags requiring atomic access. */
736
737 struct restart_block restart_block;
738
739 pid_t pid;
740 pid_t tgid;
741
742#ifdef CONFIG_STACKPROTECTOR
743 /* Canary value for the -fstack-protector GCC feature: */
744 unsigned long stack_canary;
745#endif
746 /*
747 * Pointers to the (original) parent process, youngest child, younger sibling,
748 * older sibling, respectively. (p->father can be replaced with
749 * p->real_parent->pid)
750 */
751
752 /* Real parent process: */
753 struct task_struct __rcu *real_parent;
754
755 /* Recipient of SIGCHLD, wait4() reports: */
756 struct task_struct __rcu *parent;
757
758 /*
759 * Children/sibling form the list of natural children:
760 */
761 struct list_head children;
762 struct list_head sibling;
763 struct task_struct *group_leader;
764
765 /*
766 * 'ptraced' is the list of tasks this task is using ptrace() on.
767 *
768 * This includes both natural children and PTRACE_ATTACH targets.
769 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
770 */
771 struct list_head ptraced;
772 struct list_head ptrace_entry;
773
774 /* PID/PID hash table linkage. */
775 struct pid *thread_pid;
776 struct hlist_node pid_links[PIDTYPE_MAX];
777 struct list_head thread_group;
778 struct list_head thread_node;
779
780 struct completion *vfork_done;
781
782 /* CLONE_CHILD_SETTID: */
783 int __user *set_child_tid;
784
785 /* CLONE_CHILD_CLEARTID: */
786 int __user *clear_child_tid;
787
788 u64 utime;
789 u64 stime;
790#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
791 u64 utimescaled;
792 u64 stimescaled;
793#endif
794 u64 gtime;
795 struct prev_cputime prev_cputime;
796#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
797 struct vtime vtime;
798#endif
799
800#ifdef CONFIG_NO_HZ_FULL
801 atomic_t tick_dep_mask;
802#endif
803 /* Context switch counts: */
804 unsigned long nvcsw;
805 unsigned long nivcsw;
806
807 /* Monotonic time in nsecs: */
808 u64 start_time;
809
810 /* Boot based time in nsecs: */
811 u64 real_start_time;
812
813 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
814 unsigned long min_flt;
815 unsigned long maj_flt;
816
817#ifdef CONFIG_POSIX_TIMERS
818 struct task_cputime cputime_expires;
819 struct list_head cpu_timers[3];
820#endif
821
822 /* Process credentials: */
823
824 /* Tracer's credentials at attach: */
825 const struct cred __rcu *ptracer_cred;
826
827 /* Objective and real subjective task credentials (COW): */
828 const struct cred __rcu *real_cred;
829
830 /* Effective (overridable) subjective task credentials (COW): */
831 const struct cred __rcu *cred;
832
833 /*
834 * executable name, excluding path.
835 *
836 * - normally initialized setup_new_exec()
837 * - access it with [gs]et_task_comm()
838 * - lock it with task_lock()
839 */
840 char comm[TASK_COMM_LEN];
841
842 struct nameidata *nameidata;
843
844#ifdef CONFIG_SYSVIPC
845 struct sysv_sem sysvsem;
846 struct sysv_shm sysvshm;
847#endif
848#ifdef CONFIG_DETECT_HUNG_TASK
849 unsigned long last_switch_count;
850 unsigned long last_switch_time;
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#ifdef CONFIG_AUDIT
876#ifdef CONFIG_AUDITSYSCALL
877 struct audit_context *audit_context;
878#endif
879 kuid_t loginuid;
880 unsigned int sessionid;
881#endif
882 struct seccomp seccomp;
883
884 /* Thread group tracking: */
885 u32 parent_exec_id;
886 u32 self_exec_id;
887
888 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
889 spinlock_t alloc_lock;
890
891 /* Protection of the PI data structures: */
892 raw_spinlock_t pi_lock;
893
894 struct wake_q_node wake_q;
895
896#ifdef CONFIG_RT_MUTEXES
897 /* PI waiters blocked on a rt_mutex held by this task: */
898 struct rb_root_cached pi_waiters;
899 /* Updated under owner's pi_lock and rq lock */
900 struct task_struct *pi_top_task;
901 /* Deadlock detection and priority inheritance handling: */
902 struct rt_mutex_waiter *pi_blocked_on;
903#endif
904
905#ifdef CONFIG_DEBUG_MUTEXES
906 /* Mutex deadlock detection: */
907 struct mutex_waiter *blocked_on;
908#endif
909
910#ifdef CONFIG_TRACE_IRQFLAGS
911 unsigned int irq_events;
912 unsigned long hardirq_enable_ip;
913 unsigned long hardirq_disable_ip;
914 unsigned int hardirq_enable_event;
915 unsigned int hardirq_disable_event;
916 int hardirqs_enabled;
917 int hardirq_context;
918 unsigned long softirq_disable_ip;
919 unsigned long softirq_enable_ip;
920 unsigned int softirq_disable_event;
921 unsigned int softirq_enable_event;
922 int softirqs_enabled;
923 int softirq_context;
924#endif
925
926#ifdef CONFIG_LOCKDEP
927# define MAX_LOCK_DEPTH 48UL
928 u64 curr_chain_key;
929 int lockdep_depth;
930 unsigned int lockdep_recursion;
931 struct held_lock held_locks[MAX_LOCK_DEPTH];
932#endif
933
934#ifdef CONFIG_UBSAN
935 unsigned int in_ubsan;
936#endif
937
938 /* Journalling filesystem info: */
939 void *journal_info;
940
941 /* Stacked block device info: */
942 struct bio_list *bio_list;
943
944#ifdef CONFIG_BLOCK
945 /* Stack plugging: */
946 struct blk_plug *plug;
947#endif
948
949 /* VM state: */
950 struct reclaim_state *reclaim_state;
951
952 struct backing_dev_info *backing_dev_info;
953
954 struct io_context *io_context;
955
956#ifdef CONFIG_COMPACTION
957 struct capture_control *capture_control;
958#endif
959 /* Ptrace state: */
960 unsigned long ptrace_message;
961 kernel_siginfo_t *last_siginfo;
962
963 struct task_io_accounting ioac;
964#ifdef CONFIG_PSI
965 /* Pressure stall state */
966 unsigned int psi_flags;
967#endif
968#ifdef CONFIG_TASK_XACCT
969 /* Accumulated RSS usage: */
970 u64 acct_rss_mem1;
971 /* Accumulated virtual memory usage: */
972 u64 acct_vm_mem1;
973 /* stime + utime since last update: */
974 u64 acct_timexpd;
975#endif
976#ifdef CONFIG_CPUSETS
977 /* Protected by ->alloc_lock: */
978 nodemask_t mems_allowed;
979 /* Seqence number to catch updates: */
980 seqcount_t mems_allowed_seq;
981 int cpuset_mem_spread_rotor;
982 int cpuset_slab_spread_rotor;
983#endif
984#ifdef CONFIG_CGROUPS
985 /* Control Group info protected by css_set_lock: */
986 struct css_set __rcu *cgroups;
987 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
988 struct list_head cg_list;
989#endif
990#ifdef CONFIG_X86_CPU_RESCTRL
991 u32 closid;
992 u32 rmid;
993#endif
994#ifdef CONFIG_FUTEX
995 struct robust_list_head __user *robust_list;
996#ifdef CONFIG_COMPAT
997 struct compat_robust_list_head __user *compat_robust_list;
998#endif
999 struct list_head pi_state_list;
1000 struct futex_pi_state *pi_state_cache;
1001#endif
1002#ifdef CONFIG_PERF_EVENTS
1003 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1004 struct mutex perf_event_mutex;
1005 struct list_head perf_event_list;
1006#endif
1007#ifdef CONFIG_DEBUG_PREEMPT
1008 unsigned long preempt_disable_ip;
1009#endif
1010#ifdef CONFIG_NUMA
1011 /* Protected by alloc_lock: */
1012 struct mempolicy *mempolicy;
1013 short il_prev;
1014 short pref_node_fork;
1015#endif
1016#ifdef CONFIG_NUMA_BALANCING
1017 int numa_scan_seq;
1018 unsigned int numa_scan_period;
1019 unsigned int numa_scan_period_max;
1020 int numa_preferred_nid;
1021 unsigned long numa_migrate_retry;
1022 /* Migration stamp: */
1023 u64 node_stamp;
1024 u64 last_task_numa_placement;
1025 u64 last_sum_exec_runtime;
1026 struct callback_head numa_work;
1027
1028 struct numa_group *numa_group;
1029
1030 /*
1031 * numa_faults is an array split into four regions:
1032 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1033 * in this precise order.
1034 *
1035 * faults_memory: Exponential decaying average of faults on a per-node
1036 * basis. Scheduling placement decisions are made based on these
1037 * counts. The values remain static for the duration of a PTE scan.
1038 * faults_cpu: Track the nodes the process was running on when a NUMA
1039 * hinting fault was incurred.
1040 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1041 * during the current scan window. When the scan completes, the counts
1042 * in faults_memory and faults_cpu decay and these values are copied.
1043 */
1044 unsigned long *numa_faults;
1045 unsigned long total_numa_faults;
1046
1047 /*
1048 * numa_faults_locality tracks if faults recorded during the last
1049 * scan window were remote/local or failed to migrate. The task scan
1050 * period is adapted based on the locality of the faults with different
1051 * weights depending on whether they were shared or private faults
1052 */
1053 unsigned long numa_faults_locality[3];
1054
1055 unsigned long numa_pages_migrated;
1056#endif /* CONFIG_NUMA_BALANCING */
1057
1058#ifdef CONFIG_RSEQ
1059 struct rseq __user *rseq;
1060 u32 rseq_len;
1061 u32 rseq_sig;
1062 /*
1063 * RmW on rseq_event_mask must be performed atomically
1064 * with respect to preemption.
1065 */
1066 unsigned long rseq_event_mask;
1067#endif
1068
1069 struct tlbflush_unmap_batch tlb_ubc;
1070
1071 struct rcu_head rcu;
1072
1073 /* Cache last used pipe for splice(): */
1074 struct pipe_inode_info *splice_pipe;
1075
1076 struct page_frag task_frag;
1077
1078#ifdef CONFIG_TASK_DELAY_ACCT
1079 struct task_delay_info *delays;
1080#endif
1081
1082#ifdef CONFIG_FAULT_INJECTION
1083 int make_it_fail;
1084 unsigned int fail_nth;
1085#endif
1086 /*
1087 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1088 * balance_dirty_pages() for a dirty throttling pause:
1089 */
1090 int nr_dirtied;
1091 int nr_dirtied_pause;
1092 /* Start of a write-and-pause period: */
1093 unsigned long dirty_paused_when;
1094
1095#ifdef CONFIG_LATENCYTOP
1096 int latency_record_count;
1097 struct latency_record latency_record[LT_SAVECOUNT];
1098#endif
1099 /*
1100 * Time slack values; these are used to round up poll() and
1101 * select() etc timeout values. These are in nanoseconds.
1102 */
1103 u64 timer_slack_ns;
1104 u64 default_timer_slack_ns;
1105
1106#ifdef CONFIG_KASAN
1107 unsigned int kasan_depth;
1108#endif
1109
1110#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1111 /* Index of current stored address in ret_stack: */
1112 int curr_ret_stack;
1113 int curr_ret_depth;
1114
1115 /* Stack of return addresses for return function tracing: */
1116 struct ftrace_ret_stack *ret_stack;
1117
1118 /* Timestamp for last schedule: */
1119 unsigned long long ftrace_timestamp;
1120
1121 /*
1122 * Number of functions that haven't been traced
1123 * because of depth overrun:
1124 */
1125 atomic_t trace_overrun;
1126
1127 /* Pause tracing: */
1128 atomic_t tracing_graph_pause;
1129#endif
1130
1131#ifdef CONFIG_TRACING
1132 /* State flags for use by tracers: */
1133 unsigned long trace;
1134
1135 /* Bitmask and counter of trace recursion: */
1136 unsigned long trace_recursion;
1137#endif /* CONFIG_TRACING */
1138
1139#ifdef CONFIG_KCOV
1140 /* Coverage collection mode enabled for this task (0 if disabled): */
1141 unsigned int kcov_mode;
1142
1143 /* Size of the kcov_area: */
1144 unsigned int kcov_size;
1145
1146 /* Buffer for coverage collection: */
1147 void *kcov_area;
1148
1149 /* KCOV descriptor wired with this task or NULL: */
1150 struct kcov *kcov;
1151#endif
1152
1153#ifdef CONFIG_MEMCG
1154 struct mem_cgroup *memcg_in_oom;
1155 gfp_t memcg_oom_gfp_mask;
1156 int memcg_oom_order;
1157
1158 /* Number of pages to reclaim on returning to userland: */
1159 unsigned int memcg_nr_pages_over_high;
1160
1161 /* Used by memcontrol for targeted memcg charge: */
1162 struct mem_cgroup *active_memcg;
1163#endif
1164
1165#ifdef CONFIG_BLK_CGROUP
1166 struct request_queue *throttle_queue;
1167#endif
1168
1169#ifdef CONFIG_UPROBES
1170 struct uprobe_task *utask;
1171#endif
1172#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1173 unsigned int sequential_io;
1174 unsigned int sequential_io_avg;
1175#endif
1176#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1177 unsigned long task_state_change;
1178#endif
1179 int pagefault_disabled;
1180#ifdef CONFIG_MMU
1181 struct task_struct *oom_reaper_list;
1182#endif
1183#ifdef CONFIG_VMAP_STACK
1184 struct vm_struct *stack_vm_area;
1185#endif
1186#ifdef CONFIG_THREAD_INFO_IN_TASK
1187 /* A live task holds one reference: */
1188 refcount_t stack_refcount;
1189#endif
1190#ifdef CONFIG_LIVEPATCH
1191 int patch_state;
1192#endif
1193#ifdef CONFIG_SECURITY
1194 /* Used by LSM modules for access restriction: */
1195 void *security;
1196#endif
1197
1198#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1199 unsigned long lowest_stack;
1200 unsigned long prev_lowest_stack;
1201#endif
1202
1203 /*
1204 * New fields for task_struct should be added above here, so that
1205 * they are included in the randomized portion of task_struct.
1206 */
1207 randomized_struct_fields_end
1208
1209 /* CPU-specific state of this task: */
1210 struct thread_struct thread;
1211
1212 /*
1213 * WARNING: on x86, 'thread_struct' contains a variable-sized
1214 * structure. It *MUST* be at the end of 'task_struct'.
1215 *
1216 * Do not put anything below here!
1217 */
1218};
1219
1220static inline struct pid *task_pid(struct task_struct *task)
1221{
1222 return task->thread_pid;
1223}
1224
1225/*
1226 * the helpers to get the task's different pids as they are seen
1227 * from various namespaces
1228 *
1229 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1230 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1231 * current.
1232 * task_xid_nr_ns() : id seen from the ns specified;
1233 *
1234 * see also pid_nr() etc in include/linux/pid.h
1235 */
1236pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, struct pid_namespace *ns);
1237
1238static inline pid_t task_pid_nr(struct task_struct *tsk)
1239{
1240 return tsk->pid;
1241}
1242
1243static inline pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1244{
1245 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1246}
1247
1248static inline pid_t task_pid_vnr(struct task_struct *tsk)
1249{
1250 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1251}
1252
1253
1254static inline pid_t task_tgid_nr(struct task_struct *tsk)
1255{
1256 return tsk->tgid;
1257}
1258
1259/**
1260 * pid_alive - check that a task structure is not stale
1261 * @p: Task structure to be checked.
1262 *
1263 * Test if a process is not yet dead (at most zombie state)
1264 * If pid_alive fails, then pointers within the task structure
1265 * can be stale and must not be dereferenced.
1266 *
1267 * Return: 1 if the process is alive. 0 otherwise.
1268 */
1269static inline int pid_alive(const struct task_struct *p)
1270{
1271 return p->thread_pid != NULL;
1272}
1273
1274static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1275{
1276 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1277}
1278
1279static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1280{
1281 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1282}
1283
1284
1285static inline pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1286{
1287 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1288}
1289
1290static inline pid_t task_session_vnr(struct task_struct *tsk)
1291{
1292 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1293}
1294
1295static inline pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
1296{
1297 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, ns);
1298}
1299
1300static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1301{
1302 return __task_pid_nr_ns(tsk, PIDTYPE_TGID, NULL);
1303}
1304
1305static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1306{
1307 pid_t pid = 0;
1308
1309 rcu_read_lock();
1310 if (pid_alive(tsk))
1311 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1312 rcu_read_unlock();
1313
1314 return pid;
1315}
1316
1317static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1318{
1319 return task_ppid_nr_ns(tsk, &init_pid_ns);
1320}
1321
1322/* Obsolete, do not use: */
1323static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1324{
1325 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1326}
1327
1328#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1329#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1330
1331static inline unsigned int task_state_index(struct task_struct *tsk)
1332{
1333 unsigned int tsk_state = READ_ONCE(tsk->state);
1334 unsigned int state = (tsk_state | tsk->exit_state) & TASK_REPORT;
1335
1336 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1337
1338 if (tsk_state == TASK_IDLE)
1339 state = TASK_REPORT_IDLE;
1340
1341 return fls(state);
1342}
1343
1344static inline char task_index_to_char(unsigned int state)
1345{
1346 static const char state_char[] = "RSDTtXZPI";
1347
1348 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1349
1350 return state_char[state];
1351}
1352
1353static inline char task_state_to_char(struct task_struct *tsk)
1354{
1355 return task_index_to_char(task_state_index(tsk));
1356}
1357
1358/**
1359 * is_global_init - check if a task structure is init. Since init
1360 * is free to have sub-threads we need to check tgid.
1361 * @tsk: Task structure to be checked.
1362 *
1363 * Check if a task structure is the first user space task the kernel created.
1364 *
1365 * Return: 1 if the task structure is init. 0 otherwise.
1366 */
1367static inline int is_global_init(struct task_struct *tsk)
1368{
1369 return task_tgid_nr(tsk) == 1;
1370}
1371
1372extern struct pid *cad_pid;
1373
1374/*
1375 * Per process flags
1376 */
1377#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1378#define PF_EXITING 0x00000004 /* Getting shut down */
1379#define PF_EXITPIDONE 0x00000008 /* PI exit done on shut down */
1380#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1381#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1382#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1383#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1384#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1385#define PF_DUMPCORE 0x00000200 /* Dumped core */
1386#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1387#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1388#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1389#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1390#define PF_USED_ASYNC 0x00004000 /* Used async_schedule*(), used by module init */
1391#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1392#define PF_FROZEN 0x00010000 /* Frozen for system suspend */
1393#define PF_KSWAPD 0x00020000 /* I am kswapd */
1394#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1395#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1396#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1397#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1398#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1399#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1400#define PF_MEMSTALL 0x01000000 /* Stalled due to lack of memory */
1401#define PF_UMH 0x02000000 /* I'm an Usermodehelper process */
1402#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1403#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1404#define PF_MEMALLOC_NOCMA 0x10000000 /* All allocation request will have _GFP_MOVABLE cleared */
1405#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1406#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1407
1408/*
1409 * Only the _current_ task can read/write to tsk->flags, but other
1410 * tasks can access tsk->flags in readonly mode for example
1411 * with tsk_used_math (like during threaded core dumping).
1412 * There is however an exception to this rule during ptrace
1413 * or during fork: the ptracer task is allowed to write to the
1414 * child->flags of its traced child (same goes for fork, the parent
1415 * can write to the child->flags), because we're guaranteed the
1416 * child is not running and in turn not changing child->flags
1417 * at the same time the parent does it.
1418 */
1419#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1420#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1421#define clear_used_math() clear_stopped_child_used_math(current)
1422#define set_used_math() set_stopped_child_used_math(current)
1423
1424#define conditional_stopped_child_used_math(condition, child) \
1425 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1426
1427#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1428
1429#define copy_to_stopped_child_used_math(child) \
1430 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1431
1432/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1433#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1434#define used_math() tsk_used_math(current)
1435
1436static inline bool is_percpu_thread(void)
1437{
1438#ifdef CONFIG_SMP
1439 return (current->flags & PF_NO_SETAFFINITY) &&
1440 (current->nr_cpus_allowed == 1);
1441#else
1442 return true;
1443#endif
1444}
1445
1446/* Per-process atomic flags. */
1447#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1448#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1449#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1450#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1451#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1452#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1453#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1454#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1455
1456#define TASK_PFA_TEST(name, func) \
1457 static inline bool task_##func(struct task_struct *p) \
1458 { return test_bit(PFA_##name, &p->atomic_flags); }
1459
1460#define TASK_PFA_SET(name, func) \
1461 static inline void task_set_##func(struct task_struct *p) \
1462 { set_bit(PFA_##name, &p->atomic_flags); }
1463
1464#define TASK_PFA_CLEAR(name, func) \
1465 static inline void task_clear_##func(struct task_struct *p) \
1466 { clear_bit(PFA_##name, &p->atomic_flags); }
1467
1468TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1469TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1470
1471TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1472TASK_PFA_SET(SPREAD_PAGE, spread_page)
1473TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1474
1475TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1476TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1477TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1478
1479TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1480TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1481TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1482
1483TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1484TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1485TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1486
1487TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1488TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1489
1490TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1491TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1492TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1493
1494TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1495TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1496
1497static inline void
1498current_restore_flags(unsigned long orig_flags, unsigned long flags)
1499{
1500 current->flags &= ~flags;
1501 current->flags |= orig_flags & flags;
1502}
1503
1504extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1505extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_cpus_allowed);
1506#ifdef CONFIG_SMP
1507extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1508extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1509#else
1510static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1511{
1512}
1513static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1514{
1515 if (!cpumask_test_cpu(0, new_mask))
1516 return -EINVAL;
1517 return 0;
1518}
1519#endif
1520
1521#ifndef cpu_relax_yield
1522#define cpu_relax_yield() cpu_relax()
1523#endif
1524
1525extern int yield_to(struct task_struct *p, bool preempt);
1526extern void set_user_nice(struct task_struct *p, long nice);
1527extern int task_prio(const struct task_struct *p);
1528
1529/**
1530 * task_nice - return the nice value of a given task.
1531 * @p: the task in question.
1532 *
1533 * Return: The nice value [ -20 ... 0 ... 19 ].
1534 */
1535static inline int task_nice(const struct task_struct *p)
1536{
1537 return PRIO_TO_NICE((p)->static_prio);
1538}
1539
1540extern int can_nice(const struct task_struct *p, const int nice);
1541extern int task_curr(const struct task_struct *p);
1542extern int idle_cpu(int cpu);
1543extern int available_idle_cpu(int cpu);
1544extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1545extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1546extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1547extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1548extern struct task_struct *idle_task(int cpu);
1549
1550/**
1551 * is_idle_task - is the specified task an idle task?
1552 * @p: the task in question.
1553 *
1554 * Return: 1 if @p is an idle task. 0 otherwise.
1555 */
1556static inline bool is_idle_task(const struct task_struct *p)
1557{
1558 return !!(p->flags & PF_IDLE);
1559}
1560
1561extern struct task_struct *curr_task(int cpu);
1562extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1563
1564void yield(void);
1565
1566union thread_union {
1567#ifndef CONFIG_ARCH_TASK_STRUCT_ON_STACK
1568 struct task_struct task;
1569#endif
1570#ifndef CONFIG_THREAD_INFO_IN_TASK
1571 struct thread_info thread_info;
1572#endif
1573 unsigned long stack[THREAD_SIZE/sizeof(long)];
1574};
1575
1576#ifndef CONFIG_THREAD_INFO_IN_TASK
1577extern struct thread_info init_thread_info;
1578#endif
1579
1580extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1581
1582#ifdef CONFIG_THREAD_INFO_IN_TASK
1583static inline struct thread_info *task_thread_info(struct task_struct *task)
1584{
1585 return &task->thread_info;
1586}
1587#elif !defined(__HAVE_THREAD_FUNCTIONS)
1588# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1589#endif
1590
1591/*
1592 * find a task by one of its numerical ids
1593 *
1594 * find_task_by_pid_ns():
1595 * finds a task by its pid in the specified namespace
1596 * find_task_by_vpid():
1597 * finds a task by its virtual pid
1598 *
1599 * see also find_vpid() etc in include/linux/pid.h
1600 */
1601
1602extern struct task_struct *find_task_by_vpid(pid_t nr);
1603extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1604
1605/*
1606 * find a task by its virtual pid and get the task struct
1607 */
1608extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1609
1610extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1611extern int wake_up_process(struct task_struct *tsk);
1612extern void wake_up_new_task(struct task_struct *tsk);
1613
1614#ifdef CONFIG_SMP
1615extern void kick_process(struct task_struct *tsk);
1616#else
1617static inline void kick_process(struct task_struct *tsk) { }
1618#endif
1619
1620extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1621
1622static inline void set_task_comm(struct task_struct *tsk, const char *from)
1623{
1624 __set_task_comm(tsk, from, false);
1625}
1626
1627extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1628#define get_task_comm(buf, tsk) ({ \
1629 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1630 __get_task_comm(buf, sizeof(buf), tsk); \
1631})
1632
1633#ifdef CONFIG_SMP
1634void scheduler_ipi(void);
1635extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
1636#else
1637static inline void scheduler_ipi(void) { }
1638static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state)
1639{
1640 return 1;
1641}
1642#endif
1643
1644/*
1645 * Set thread flags in other task's structures.
1646 * See asm/thread_info.h for TIF_xxxx flags available:
1647 */
1648static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1649{
1650 set_ti_thread_flag(task_thread_info(tsk), flag);
1651}
1652
1653static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1654{
1655 clear_ti_thread_flag(task_thread_info(tsk), flag);
1656}
1657
1658static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1659 bool value)
1660{
1661 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1662}
1663
1664static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1665{
1666 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1667}
1668
1669static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1670{
1671 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1672}
1673
1674static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1675{
1676 return test_ti_thread_flag(task_thread_info(tsk), flag);
1677}
1678
1679static inline void set_tsk_need_resched(struct task_struct *tsk)
1680{
1681 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1682}
1683
1684static inline void clear_tsk_need_resched(struct task_struct *tsk)
1685{
1686 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1687}
1688
1689static inline int test_tsk_need_resched(struct task_struct *tsk)
1690{
1691 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1692}
1693
1694/*
1695 * cond_resched() and cond_resched_lock(): latency reduction via
1696 * explicit rescheduling in places that are safe. The return
1697 * value indicates whether a reschedule was done in fact.
1698 * cond_resched_lock() will drop the spinlock before scheduling,
1699 */
1700#ifndef CONFIG_PREEMPT
1701extern int _cond_resched(void);
1702#else
1703static inline int _cond_resched(void) { return 0; }
1704#endif
1705
1706#define cond_resched() ({ \
1707 ___might_sleep(__FILE__, __LINE__, 0); \
1708 _cond_resched(); \
1709})
1710
1711extern int __cond_resched_lock(spinlock_t *lock);
1712
1713#define cond_resched_lock(lock) ({ \
1714 ___might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET);\
1715 __cond_resched_lock(lock); \
1716})
1717
1718static inline void cond_resched_rcu(void)
1719{
1720#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
1721 rcu_read_unlock();
1722 cond_resched();
1723 rcu_read_lock();
1724#endif
1725}
1726
1727/*
1728 * Does a critical section need to be broken due to another
1729 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
1730 * but a general need for low latency)
1731 */
1732static inline int spin_needbreak(spinlock_t *lock)
1733{
1734#ifdef CONFIG_PREEMPT
1735 return spin_is_contended(lock);
1736#else
1737 return 0;
1738#endif
1739}
1740
1741static __always_inline bool need_resched(void)
1742{
1743 return unlikely(tif_need_resched());
1744}
1745
1746/*
1747 * Wrappers for p->thread_info->cpu access. No-op on UP.
1748 */
1749#ifdef CONFIG_SMP
1750
1751static inline unsigned int task_cpu(const struct task_struct *p)
1752{
1753#ifdef CONFIG_THREAD_INFO_IN_TASK
1754 return READ_ONCE(p->cpu);
1755#else
1756 return READ_ONCE(task_thread_info(p)->cpu);
1757#endif
1758}
1759
1760extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
1761
1762#else
1763
1764static inline unsigned int task_cpu(const struct task_struct *p)
1765{
1766 return 0;
1767}
1768
1769static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
1770{
1771}
1772
1773#endif /* CONFIG_SMP */
1774
1775/*
1776 * In order to reduce various lock holder preemption latencies provide an
1777 * interface to see if a vCPU is currently running or not.
1778 *
1779 * This allows us to terminate optimistic spin loops and block, analogous to
1780 * the native optimistic spin heuristic of testing if the lock owner task is
1781 * running or not.
1782 */
1783#ifndef vcpu_is_preempted
1784# define vcpu_is_preempted(cpu) false
1785#endif
1786
1787extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
1788extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
1789
1790#ifndef TASK_SIZE_OF
1791#define TASK_SIZE_OF(tsk) TASK_SIZE
1792#endif
1793
1794#ifdef CONFIG_RSEQ
1795
1796/*
1797 * Map the event mask on the user-space ABI enum rseq_cs_flags
1798 * for direct mask checks.
1799 */
1800enum rseq_event_mask_bits {
1801 RSEQ_EVENT_PREEMPT_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT,
1802 RSEQ_EVENT_SIGNAL_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT,
1803 RSEQ_EVENT_MIGRATE_BIT = RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT,
1804};
1805
1806enum rseq_event_mask {
1807 RSEQ_EVENT_PREEMPT = (1U << RSEQ_EVENT_PREEMPT_BIT),
1808 RSEQ_EVENT_SIGNAL = (1U << RSEQ_EVENT_SIGNAL_BIT),
1809 RSEQ_EVENT_MIGRATE = (1U << RSEQ_EVENT_MIGRATE_BIT),
1810};
1811
1812static inline void rseq_set_notify_resume(struct task_struct *t)
1813{
1814 if (t->rseq)
1815 set_tsk_thread_flag(t, TIF_NOTIFY_RESUME);
1816}
1817
1818void __rseq_handle_notify_resume(struct ksignal *sig, struct pt_regs *regs);
1819
1820static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1821 struct pt_regs *regs)
1822{
1823 if (current->rseq)
1824 __rseq_handle_notify_resume(ksig, regs);
1825}
1826
1827static inline void rseq_signal_deliver(struct ksignal *ksig,
1828 struct pt_regs *regs)
1829{
1830 preempt_disable();
1831 __set_bit(RSEQ_EVENT_SIGNAL_BIT, &current->rseq_event_mask);
1832 preempt_enable();
1833 rseq_handle_notify_resume(ksig, regs);
1834}
1835
1836/* rseq_preempt() requires preemption to be disabled. */
1837static inline void rseq_preempt(struct task_struct *t)
1838{
1839 __set_bit(RSEQ_EVENT_PREEMPT_BIT, &t->rseq_event_mask);
1840 rseq_set_notify_resume(t);
1841}
1842
1843/* rseq_migrate() requires preemption to be disabled. */
1844static inline void rseq_migrate(struct task_struct *t)
1845{
1846 __set_bit(RSEQ_EVENT_MIGRATE_BIT, &t->rseq_event_mask);
1847 rseq_set_notify_resume(t);
1848}
1849
1850/*
1851 * If parent process has a registered restartable sequences area, the
1852 * child inherits. Only applies when forking a process, not a thread.
1853 */
1854static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1855{
1856 if (clone_flags & CLONE_THREAD) {
1857 t->rseq = NULL;
1858 t->rseq_len = 0;
1859 t->rseq_sig = 0;
1860 t->rseq_event_mask = 0;
1861 } else {
1862 t->rseq = current->rseq;
1863 t->rseq_len = current->rseq_len;
1864 t->rseq_sig = current->rseq_sig;
1865 t->rseq_event_mask = current->rseq_event_mask;
1866 }
1867}
1868
1869static inline void rseq_execve(struct task_struct *t)
1870{
1871 t->rseq = NULL;
1872 t->rseq_len = 0;
1873 t->rseq_sig = 0;
1874 t->rseq_event_mask = 0;
1875}
1876
1877#else
1878
1879static inline void rseq_set_notify_resume(struct task_struct *t)
1880{
1881}
1882static inline void rseq_handle_notify_resume(struct ksignal *ksig,
1883 struct pt_regs *regs)
1884{
1885}
1886static inline void rseq_signal_deliver(struct ksignal *ksig,
1887 struct pt_regs *regs)
1888{
1889}
1890static inline void rseq_preempt(struct task_struct *t)
1891{
1892}
1893static inline void rseq_migrate(struct task_struct *t)
1894{
1895}
1896static inline void rseq_fork(struct task_struct *t, unsigned long clone_flags)
1897{
1898}
1899static inline void rseq_execve(struct task_struct *t)
1900{
1901}
1902
1903#endif
1904
1905void __exit_umh(struct task_struct *tsk);
1906
1907static inline void exit_umh(struct task_struct *tsk)
1908{
1909 if (unlikely(tsk->flags & PF_UMH))
1910 __exit_umh(tsk);
1911}
1912
1913#ifdef CONFIG_DEBUG_RSEQ
1914
1915void rseq_syscall(struct pt_regs *regs);
1916
1917#else
1918
1919static inline void rseq_syscall(struct pt_regs *regs)
1920{
1921}
1922
1923#endif
1924
1925#endif
1926