Loading...
1#ifndef _LINUX_SCHED_H
2#define _LINUX_SCHED_H
3
4#include <uapi/linux/sched.h>
5
6#include <linux/sched/prio.h>
7
8
9struct sched_param {
10 int sched_priority;
11};
12
13#include <asm/param.h> /* for HZ */
14
15#include <linux/capability.h>
16#include <linux/threads.h>
17#include <linux/kernel.h>
18#include <linux/types.h>
19#include <linux/timex.h>
20#include <linux/jiffies.h>
21#include <linux/plist.h>
22#include <linux/rbtree.h>
23#include <linux/thread_info.h>
24#include <linux/cpumask.h>
25#include <linux/errno.h>
26#include <linux/nodemask.h>
27#include <linux/mm_types.h>
28#include <linux/preempt_mask.h>
29
30#include <asm/page.h>
31#include <asm/ptrace.h>
32#include <linux/cputime.h>
33
34#include <linux/smp.h>
35#include <linux/sem.h>
36#include <linux/signal.h>
37#include <linux/compiler.h>
38#include <linux/completion.h>
39#include <linux/pid.h>
40#include <linux/percpu.h>
41#include <linux/topology.h>
42#include <linux/proportions.h>
43#include <linux/seccomp.h>
44#include <linux/rcupdate.h>
45#include <linux/rculist.h>
46#include <linux/rtmutex.h>
47
48#include <linux/time.h>
49#include <linux/param.h>
50#include <linux/resource.h>
51#include <linux/timer.h>
52#include <linux/hrtimer.h>
53#include <linux/task_io_accounting.h>
54#include <linux/latencytop.h>
55#include <linux/cred.h>
56#include <linux/llist.h>
57#include <linux/uidgid.h>
58#include <linux/gfp.h>
59
60#include <asm/processor.h>
61
62#define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
63
64/*
65 * Extended scheduling parameters data structure.
66 *
67 * This is needed because the original struct sched_param can not be
68 * altered without introducing ABI issues with legacy applications
69 * (e.g., in sched_getparam()).
70 *
71 * However, the possibility of specifying more than just a priority for
72 * the tasks may be useful for a wide variety of application fields, e.g.,
73 * multimedia, streaming, automation and control, and many others.
74 *
75 * This variant (sched_attr) is meant at describing a so-called
76 * sporadic time-constrained task. In such model a task is specified by:
77 * - the activation period or minimum instance inter-arrival time;
78 * - the maximum (or average, depending on the actual scheduling
79 * discipline) computation time of all instances, a.k.a. runtime;
80 * - the deadline (relative to the actual activation time) of each
81 * instance.
82 * Very briefly, a periodic (sporadic) task asks for the execution of
83 * some specific computation --which is typically called an instance--
84 * (at most) every period. Moreover, each instance typically lasts no more
85 * than the runtime and must be completed by time instant t equal to
86 * the instance activation time + the deadline.
87 *
88 * This is reflected by the actual fields of the sched_attr structure:
89 *
90 * @size size of the structure, for fwd/bwd compat.
91 *
92 * @sched_policy task's scheduling policy
93 * @sched_flags for customizing the scheduler behaviour
94 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
95 * @sched_priority task's static priority (SCHED_FIFO/RR)
96 * @sched_deadline representative of the task's deadline
97 * @sched_runtime representative of the task's runtime
98 * @sched_period representative of the task's period
99 *
100 * Given this task model, there are a multiplicity of scheduling algorithms
101 * and policies, that can be used to ensure all the tasks will make their
102 * timing constraints.
103 *
104 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
105 * only user of this new interface. More information about the algorithm
106 * available in the scheduling class file or in Documentation/.
107 */
108struct sched_attr {
109 u32 size;
110
111 u32 sched_policy;
112 u64 sched_flags;
113
114 /* SCHED_NORMAL, SCHED_BATCH */
115 s32 sched_nice;
116
117 /* SCHED_FIFO, SCHED_RR */
118 u32 sched_priority;
119
120 /* SCHED_DEADLINE */
121 u64 sched_runtime;
122 u64 sched_deadline;
123 u64 sched_period;
124};
125
126struct exec_domain;
127struct futex_pi_state;
128struct robust_list_head;
129struct bio_list;
130struct fs_struct;
131struct perf_event_context;
132struct blk_plug;
133struct filename;
134
135#define VMACACHE_BITS 2
136#define VMACACHE_SIZE (1U << VMACACHE_BITS)
137#define VMACACHE_MASK (VMACACHE_SIZE - 1)
138
139/*
140 * List of flags we want to share for kernel threads,
141 * if only because they are not used by them anyway.
142 */
143#define CLONE_KERNEL (CLONE_FS | CLONE_FILES | CLONE_SIGHAND)
144
145/*
146 * These are the constant used to fake the fixed-point load-average
147 * counting. Some notes:
148 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
149 * a load-average precision of 10 bits integer + 11 bits fractional
150 * - if you want to count load-averages more often, you need more
151 * precision, or rounding will get you. With 2-second counting freq,
152 * the EXP_n values would be 1981, 2034 and 2043 if still using only
153 * 11 bit fractions.
154 */
155extern unsigned long avenrun[]; /* Load averages */
156extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
157
158#define FSHIFT 11 /* nr of bits of precision */
159#define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
160#define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
161#define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
162#define EXP_5 2014 /* 1/exp(5sec/5min) */
163#define EXP_15 2037 /* 1/exp(5sec/15min) */
164
165#define CALC_LOAD(load,exp,n) \
166 load *= exp; \
167 load += n*(FIXED_1-exp); \
168 load >>= FSHIFT;
169
170extern unsigned long total_forks;
171extern int nr_threads;
172DECLARE_PER_CPU(unsigned long, process_counts);
173extern int nr_processes(void);
174extern unsigned long nr_running(void);
175extern unsigned long nr_iowait(void);
176extern unsigned long nr_iowait_cpu(int cpu);
177extern unsigned long this_cpu_load(void);
178
179
180extern void calc_global_load(unsigned long ticks);
181extern void update_cpu_load_nohz(void);
182
183extern unsigned long get_parent_ip(unsigned long addr);
184
185extern void dump_cpu_task(int cpu);
186
187struct seq_file;
188struct cfs_rq;
189struct task_group;
190#ifdef CONFIG_SCHED_DEBUG
191extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
192extern void proc_sched_set_task(struct task_struct *p);
193extern void
194print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq);
195#endif
196
197/*
198 * Task state bitmask. NOTE! These bits are also
199 * encoded in fs/proc/array.c: get_task_state().
200 *
201 * We have two separate sets of flags: task->state
202 * is about runnability, while task->exit_state are
203 * about the task exiting. Confusing, but this way
204 * modifying one set can't modify the other one by
205 * mistake.
206 */
207#define TASK_RUNNING 0
208#define TASK_INTERRUPTIBLE 1
209#define TASK_UNINTERRUPTIBLE 2
210#define __TASK_STOPPED 4
211#define __TASK_TRACED 8
212/* in tsk->exit_state */
213#define EXIT_DEAD 16
214#define EXIT_ZOMBIE 32
215#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
216/* in tsk->state again */
217#define TASK_DEAD 64
218#define TASK_WAKEKILL 128
219#define TASK_WAKING 256
220#define TASK_PARKED 512
221#define TASK_STATE_MAX 1024
222
223#define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWP"
224
225extern char ___assert_task_state[1 - 2*!!(
226 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
227
228/* Convenience macros for the sake of set_task_state */
229#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
230#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
231#define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
232
233/* Convenience macros for the sake of wake_up */
234#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
235#define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
236
237/* get_task_state() */
238#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
239 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
240 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
241
242#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
243#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
244#define task_is_stopped_or_traced(task) \
245 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
246#define task_contributes_to_load(task) \
247 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
248 (task->flags & PF_FROZEN) == 0)
249
250#define __set_task_state(tsk, state_value) \
251 do { (tsk)->state = (state_value); } while (0)
252#define set_task_state(tsk, state_value) \
253 set_mb((tsk)->state, (state_value))
254
255/*
256 * set_current_state() includes a barrier so that the write of current->state
257 * is correctly serialised wrt the caller's subsequent test of whether to
258 * actually sleep:
259 *
260 * set_current_state(TASK_UNINTERRUPTIBLE);
261 * if (do_i_need_to_sleep())
262 * schedule();
263 *
264 * If the caller does not need such serialisation then use __set_current_state()
265 */
266#define __set_current_state(state_value) \
267 do { current->state = (state_value); } while (0)
268#define set_current_state(state_value) \
269 set_mb(current->state, (state_value))
270
271/* Task command name length */
272#define TASK_COMM_LEN 16
273
274#include <linux/spinlock.h>
275
276/*
277 * This serializes "schedule()" and also protects
278 * the run-queue from deletions/modifications (but
279 * _adding_ to the beginning of the run-queue has
280 * a separate lock).
281 */
282extern rwlock_t tasklist_lock;
283extern spinlock_t mmlist_lock;
284
285struct task_struct;
286
287#ifdef CONFIG_PROVE_RCU
288extern int lockdep_tasklist_lock_is_held(void);
289#endif /* #ifdef CONFIG_PROVE_RCU */
290
291extern void sched_init(void);
292extern void sched_init_smp(void);
293extern asmlinkage void schedule_tail(struct task_struct *prev);
294extern void init_idle(struct task_struct *idle, int cpu);
295extern void init_idle_bootup_task(struct task_struct *idle);
296
297extern int runqueue_is_locked(int cpu);
298
299#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
300extern void nohz_balance_enter_idle(int cpu);
301extern void set_cpu_sd_state_idle(void);
302extern int get_nohz_timer_target(int pinned);
303#else
304static inline void nohz_balance_enter_idle(int cpu) { }
305static inline void set_cpu_sd_state_idle(void) { }
306static inline int get_nohz_timer_target(int pinned)
307{
308 return smp_processor_id();
309}
310#endif
311
312/*
313 * Only dump TASK_* tasks. (0 for all tasks)
314 */
315extern void show_state_filter(unsigned long state_filter);
316
317static inline void show_state(void)
318{
319 show_state_filter(0);
320}
321
322extern void show_regs(struct pt_regs *);
323
324/*
325 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
326 * task), SP is the stack pointer of the first frame that should be shown in the back
327 * trace (or NULL if the entire call-chain of the task should be shown).
328 */
329extern void show_stack(struct task_struct *task, unsigned long *sp);
330
331void io_schedule(void);
332long io_schedule_timeout(long timeout);
333
334extern void cpu_init (void);
335extern void trap_init(void);
336extern void update_process_times(int user);
337extern void scheduler_tick(void);
338
339extern void sched_show_task(struct task_struct *p);
340
341#ifdef CONFIG_LOCKUP_DETECTOR
342extern void touch_softlockup_watchdog(void);
343extern void touch_softlockup_watchdog_sync(void);
344extern void touch_all_softlockup_watchdogs(void);
345extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
346 void __user *buffer,
347 size_t *lenp, loff_t *ppos);
348extern unsigned int softlockup_panic;
349void lockup_detector_init(void);
350#else
351static inline void touch_softlockup_watchdog(void)
352{
353}
354static inline void touch_softlockup_watchdog_sync(void)
355{
356}
357static inline void touch_all_softlockup_watchdogs(void)
358{
359}
360static inline void lockup_detector_init(void)
361{
362}
363#endif
364
365#ifdef CONFIG_DETECT_HUNG_TASK
366void reset_hung_task_detector(void);
367#else
368static inline void reset_hung_task_detector(void)
369{
370}
371#endif
372
373/* Attach to any functions which should be ignored in wchan output. */
374#define __sched __attribute__((__section__(".sched.text")))
375
376/* Linker adds these: start and end of __sched functions */
377extern char __sched_text_start[], __sched_text_end[];
378
379/* Is this address in the __sched functions? */
380extern int in_sched_functions(unsigned long addr);
381
382#define MAX_SCHEDULE_TIMEOUT LONG_MAX
383extern signed long schedule_timeout(signed long timeout);
384extern signed long schedule_timeout_interruptible(signed long timeout);
385extern signed long schedule_timeout_killable(signed long timeout);
386extern signed long schedule_timeout_uninterruptible(signed long timeout);
387asmlinkage void schedule(void);
388extern void schedule_preempt_disabled(void);
389
390struct nsproxy;
391struct user_namespace;
392
393#ifdef CONFIG_MMU
394extern void arch_pick_mmap_layout(struct mm_struct *mm);
395extern unsigned long
396arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
397 unsigned long, unsigned long);
398extern unsigned long
399arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
400 unsigned long len, unsigned long pgoff,
401 unsigned long flags);
402#else
403static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
404#endif
405
406#define SUID_DUMP_DISABLE 0 /* No setuid dumping */
407#define SUID_DUMP_USER 1 /* Dump as user of process */
408#define SUID_DUMP_ROOT 2 /* Dump as root */
409
410/* mm flags */
411
412/* for SUID_DUMP_* above */
413#define MMF_DUMPABLE_BITS 2
414#define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
415
416extern void set_dumpable(struct mm_struct *mm, int value);
417/*
418 * This returns the actual value of the suid_dumpable flag. For things
419 * that are using this for checking for privilege transitions, it must
420 * test against SUID_DUMP_USER rather than treating it as a boolean
421 * value.
422 */
423static inline int __get_dumpable(unsigned long mm_flags)
424{
425 return mm_flags & MMF_DUMPABLE_MASK;
426}
427
428static inline int get_dumpable(struct mm_struct *mm)
429{
430 return __get_dumpable(mm->flags);
431}
432
433/* coredump filter bits */
434#define MMF_DUMP_ANON_PRIVATE 2
435#define MMF_DUMP_ANON_SHARED 3
436#define MMF_DUMP_MAPPED_PRIVATE 4
437#define MMF_DUMP_MAPPED_SHARED 5
438#define MMF_DUMP_ELF_HEADERS 6
439#define MMF_DUMP_HUGETLB_PRIVATE 7
440#define MMF_DUMP_HUGETLB_SHARED 8
441
442#define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
443#define MMF_DUMP_FILTER_BITS 7
444#define MMF_DUMP_FILTER_MASK \
445 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
446#define MMF_DUMP_FILTER_DEFAULT \
447 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
448 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
449
450#ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
451# define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
452#else
453# define MMF_DUMP_MASK_DEFAULT_ELF 0
454#endif
455 /* leave room for more dump flags */
456#define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
457#define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
458#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
459
460#define MMF_HAS_UPROBES 19 /* has uprobes */
461#define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
462
463#define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
464
465struct sighand_struct {
466 atomic_t count;
467 struct k_sigaction action[_NSIG];
468 spinlock_t siglock;
469 wait_queue_head_t signalfd_wqh;
470};
471
472struct pacct_struct {
473 int ac_flag;
474 long ac_exitcode;
475 unsigned long ac_mem;
476 cputime_t ac_utime, ac_stime;
477 unsigned long ac_minflt, ac_majflt;
478};
479
480struct cpu_itimer {
481 cputime_t expires;
482 cputime_t incr;
483 u32 error;
484 u32 incr_error;
485};
486
487/**
488 * struct cputime - snaphsot of system and user cputime
489 * @utime: time spent in user mode
490 * @stime: time spent in system mode
491 *
492 * Gathers a generic snapshot of user and system time.
493 */
494struct cputime {
495 cputime_t utime;
496 cputime_t stime;
497};
498
499/**
500 * struct task_cputime - collected CPU time counts
501 * @utime: time spent in user mode, in &cputime_t units
502 * @stime: time spent in kernel mode, in &cputime_t units
503 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
504 *
505 * This is an extension of struct cputime that includes the total runtime
506 * spent by the task from the scheduler point of view.
507 *
508 * As a result, this structure groups together three kinds of CPU time
509 * that are tracked for threads and thread groups. Most things considering
510 * CPU time want to group these counts together and treat all three
511 * of them in parallel.
512 */
513struct task_cputime {
514 cputime_t utime;
515 cputime_t stime;
516 unsigned long long sum_exec_runtime;
517};
518/* Alternate field names when used to cache expirations. */
519#define prof_exp stime
520#define virt_exp utime
521#define sched_exp sum_exec_runtime
522
523#define INIT_CPUTIME \
524 (struct task_cputime) { \
525 .utime = 0, \
526 .stime = 0, \
527 .sum_exec_runtime = 0, \
528 }
529
530#ifdef CONFIG_PREEMPT_COUNT
531#define PREEMPT_DISABLED (1 + PREEMPT_ENABLED)
532#else
533#define PREEMPT_DISABLED PREEMPT_ENABLED
534#endif
535
536/*
537 * Disable preemption until the scheduler is running.
538 * Reset by start_kernel()->sched_init()->init_idle().
539 *
540 * We include PREEMPT_ACTIVE to avoid cond_resched() from working
541 * before the scheduler is active -- see should_resched().
542 */
543#define INIT_PREEMPT_COUNT (PREEMPT_DISABLED + PREEMPT_ACTIVE)
544
545/**
546 * struct thread_group_cputimer - thread group interval timer counts
547 * @cputime: thread group interval timers.
548 * @running: non-zero when there are timers running and
549 * @cputime receives updates.
550 * @lock: lock for fields in this struct.
551 *
552 * This structure contains the version of task_cputime, above, that is
553 * used for thread group CPU timer calculations.
554 */
555struct thread_group_cputimer {
556 struct task_cputime cputime;
557 int running;
558 raw_spinlock_t lock;
559};
560
561#include <linux/rwsem.h>
562struct autogroup;
563
564/*
565 * NOTE! "signal_struct" does not have its own
566 * locking, because a shared signal_struct always
567 * implies a shared sighand_struct, so locking
568 * sighand_struct is always a proper superset of
569 * the locking of signal_struct.
570 */
571struct signal_struct {
572 atomic_t sigcnt;
573 atomic_t live;
574 int nr_threads;
575 struct list_head thread_head;
576
577 wait_queue_head_t wait_chldexit; /* for wait4() */
578
579 /* current thread group signal load-balancing target: */
580 struct task_struct *curr_target;
581
582 /* shared signal handling: */
583 struct sigpending shared_pending;
584
585 /* thread group exit support */
586 int group_exit_code;
587 /* overloaded:
588 * - notify group_exit_task when ->count is equal to notify_count
589 * - everyone except group_exit_task is stopped during signal delivery
590 * of fatal signals, group_exit_task processes the signal.
591 */
592 int notify_count;
593 struct task_struct *group_exit_task;
594
595 /* thread group stop support, overloads group_exit_code too */
596 int group_stop_count;
597 unsigned int flags; /* see SIGNAL_* flags below */
598
599 /*
600 * PR_SET_CHILD_SUBREAPER marks a process, like a service
601 * manager, to re-parent orphan (double-forking) child processes
602 * to this process instead of 'init'. The service manager is
603 * able to receive SIGCHLD signals and is able to investigate
604 * the process until it calls wait(). All children of this
605 * process will inherit a flag if they should look for a
606 * child_subreaper process at exit.
607 */
608 unsigned int is_child_subreaper:1;
609 unsigned int has_child_subreaper:1;
610
611 /* POSIX.1b Interval Timers */
612 int posix_timer_id;
613 struct list_head posix_timers;
614
615 /* ITIMER_REAL timer for the process */
616 struct hrtimer real_timer;
617 struct pid *leader_pid;
618 ktime_t it_real_incr;
619
620 /*
621 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
622 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
623 * values are defined to 0 and 1 respectively
624 */
625 struct cpu_itimer it[2];
626
627 /*
628 * Thread group totals for process CPU timers.
629 * See thread_group_cputimer(), et al, for details.
630 */
631 struct thread_group_cputimer cputimer;
632
633 /* Earliest-expiration cache. */
634 struct task_cputime cputime_expires;
635
636 struct list_head cpu_timers[3];
637
638 struct pid *tty_old_pgrp;
639
640 /* boolean value for session group leader */
641 int leader;
642
643 struct tty_struct *tty; /* NULL if no tty */
644
645#ifdef CONFIG_SCHED_AUTOGROUP
646 struct autogroup *autogroup;
647#endif
648 /*
649 * Cumulative resource counters for dead threads in the group,
650 * and for reaped dead child processes forked by this group.
651 * Live threads maintain their own counters and add to these
652 * in __exit_signal, except for the group leader.
653 */
654 cputime_t utime, stime, cutime, cstime;
655 cputime_t gtime;
656 cputime_t cgtime;
657#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
658 struct cputime prev_cputime;
659#endif
660 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
661 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
662 unsigned long inblock, oublock, cinblock, coublock;
663 unsigned long maxrss, cmaxrss;
664 struct task_io_accounting ioac;
665
666 /*
667 * Cumulative ns of schedule CPU time fo dead threads in the
668 * group, not including a zombie group leader, (This only differs
669 * from jiffies_to_ns(utime + stime) if sched_clock uses something
670 * other than jiffies.)
671 */
672 unsigned long long sum_sched_runtime;
673
674 /*
675 * We don't bother to synchronize most readers of this at all,
676 * because there is no reader checking a limit that actually needs
677 * to get both rlim_cur and rlim_max atomically, and either one
678 * alone is a single word that can safely be read normally.
679 * getrlimit/setrlimit use task_lock(current->group_leader) to
680 * protect this instead of the siglock, because they really
681 * have no need to disable irqs.
682 */
683 struct rlimit rlim[RLIM_NLIMITS];
684
685#ifdef CONFIG_BSD_PROCESS_ACCT
686 struct pacct_struct pacct; /* per-process accounting information */
687#endif
688#ifdef CONFIG_TASKSTATS
689 struct taskstats *stats;
690#endif
691#ifdef CONFIG_AUDIT
692 unsigned audit_tty;
693 unsigned audit_tty_log_passwd;
694 struct tty_audit_buf *tty_audit_buf;
695#endif
696#ifdef CONFIG_CGROUPS
697 /*
698 * group_rwsem prevents new tasks from entering the threadgroup and
699 * member tasks from exiting,a more specifically, setting of
700 * PF_EXITING. fork and exit paths are protected with this rwsem
701 * using threadgroup_change_begin/end(). Users which require
702 * threadgroup to remain stable should use threadgroup_[un]lock()
703 * which also takes care of exec path. Currently, cgroup is the
704 * only user.
705 */
706 struct rw_semaphore group_rwsem;
707#endif
708
709 oom_flags_t oom_flags;
710 short oom_score_adj; /* OOM kill score adjustment */
711 short oom_score_adj_min; /* OOM kill score adjustment min value.
712 * Only settable by CAP_SYS_RESOURCE. */
713
714 struct mutex cred_guard_mutex; /* guard against foreign influences on
715 * credential calculations
716 * (notably. ptrace) */
717};
718
719/*
720 * Bits in flags field of signal_struct.
721 */
722#define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
723#define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
724#define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
725#define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
726/*
727 * Pending notifications to parent.
728 */
729#define SIGNAL_CLD_STOPPED 0x00000010
730#define SIGNAL_CLD_CONTINUED 0x00000020
731#define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
732
733#define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
734
735/* If true, all threads except ->group_exit_task have pending SIGKILL */
736static inline int signal_group_exit(const struct signal_struct *sig)
737{
738 return (sig->flags & SIGNAL_GROUP_EXIT) ||
739 (sig->group_exit_task != NULL);
740}
741
742/*
743 * Some day this will be a full-fledged user tracking system..
744 */
745struct user_struct {
746 atomic_t __count; /* reference count */
747 atomic_t processes; /* How many processes does this user have? */
748 atomic_t files; /* How many open files does this user have? */
749 atomic_t sigpending; /* How many pending signals does this user have? */
750#ifdef CONFIG_INOTIFY_USER
751 atomic_t inotify_watches; /* How many inotify watches does this user have? */
752 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
753#endif
754#ifdef CONFIG_FANOTIFY
755 atomic_t fanotify_listeners;
756#endif
757#ifdef CONFIG_EPOLL
758 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
759#endif
760#ifdef CONFIG_POSIX_MQUEUE
761 /* protected by mq_lock */
762 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
763#endif
764 unsigned long locked_shm; /* How many pages of mlocked shm ? */
765
766#ifdef CONFIG_KEYS
767 struct key *uid_keyring; /* UID specific keyring */
768 struct key *session_keyring; /* UID's default session keyring */
769#endif
770
771 /* Hash table maintenance information */
772 struct hlist_node uidhash_node;
773 kuid_t uid;
774
775#ifdef CONFIG_PERF_EVENTS
776 atomic_long_t locked_vm;
777#endif
778};
779
780extern int uids_sysfs_init(void);
781
782extern struct user_struct *find_user(kuid_t);
783
784extern struct user_struct root_user;
785#define INIT_USER (&root_user)
786
787
788struct backing_dev_info;
789struct reclaim_state;
790
791#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
792struct sched_info {
793 /* cumulative counters */
794 unsigned long pcount; /* # of times run on this cpu */
795 unsigned long long run_delay; /* time spent waiting on a runqueue */
796
797 /* timestamps */
798 unsigned long long last_arrival,/* when we last ran on a cpu */
799 last_queued; /* when we were last queued to run */
800};
801#endif /* defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) */
802
803#ifdef CONFIG_TASK_DELAY_ACCT
804struct task_delay_info {
805 spinlock_t lock;
806 unsigned int flags; /* Private per-task flags */
807
808 /* For each stat XXX, add following, aligned appropriately
809 *
810 * struct timespec XXX_start, XXX_end;
811 * u64 XXX_delay;
812 * u32 XXX_count;
813 *
814 * Atomicity of updates to XXX_delay, XXX_count protected by
815 * single lock above (split into XXX_lock if contention is an issue).
816 */
817
818 /*
819 * XXX_count is incremented on every XXX operation, the delay
820 * associated with the operation is added to XXX_delay.
821 * XXX_delay contains the accumulated delay time in nanoseconds.
822 */
823 struct timespec blkio_start, blkio_end; /* Shared by blkio, swapin */
824 u64 blkio_delay; /* wait for sync block io completion */
825 u64 swapin_delay; /* wait for swapin block io completion */
826 u32 blkio_count; /* total count of the number of sync block */
827 /* io operations performed */
828 u32 swapin_count; /* total count of the number of swapin block */
829 /* io operations performed */
830
831 struct timespec freepages_start, freepages_end;
832 u64 freepages_delay; /* wait for memory reclaim */
833 u32 freepages_count; /* total count of memory reclaim */
834};
835#endif /* CONFIG_TASK_DELAY_ACCT */
836
837static inline int sched_info_on(void)
838{
839#ifdef CONFIG_SCHEDSTATS
840 return 1;
841#elif defined(CONFIG_TASK_DELAY_ACCT)
842 extern int delayacct_on;
843 return delayacct_on;
844#else
845 return 0;
846#endif
847}
848
849enum cpu_idle_type {
850 CPU_IDLE,
851 CPU_NOT_IDLE,
852 CPU_NEWLY_IDLE,
853 CPU_MAX_IDLE_TYPES
854};
855
856/*
857 * Increase resolution of cpu_power calculations
858 */
859#define SCHED_POWER_SHIFT 10
860#define SCHED_POWER_SCALE (1L << SCHED_POWER_SHIFT)
861
862/*
863 * sched-domains (multiprocessor balancing) declarations:
864 */
865#ifdef CONFIG_SMP
866#define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
867#define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
868#define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
869#define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
870#define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
871#define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
872#define SD_SHARE_CPUPOWER 0x0080 /* Domain members share cpu power */
873#define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
874#define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
875#define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
876#define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
877#define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
878#define SD_NUMA 0x4000 /* cross-node balancing */
879
880extern int __weak arch_sd_sibiling_asym_packing(void);
881
882struct sched_domain_attr {
883 int relax_domain_level;
884};
885
886#define SD_ATTR_INIT (struct sched_domain_attr) { \
887 .relax_domain_level = -1, \
888}
889
890extern int sched_domain_level_max;
891
892struct sched_group;
893
894struct sched_domain {
895 /* These fields must be setup */
896 struct sched_domain *parent; /* top domain must be null terminated */
897 struct sched_domain *child; /* bottom domain must be null terminated */
898 struct sched_group *groups; /* the balancing groups of the domain */
899 unsigned long min_interval; /* Minimum balance interval ms */
900 unsigned long max_interval; /* Maximum balance interval ms */
901 unsigned int busy_factor; /* less balancing by factor if busy */
902 unsigned int imbalance_pct; /* No balance until over watermark */
903 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
904 unsigned int busy_idx;
905 unsigned int idle_idx;
906 unsigned int newidle_idx;
907 unsigned int wake_idx;
908 unsigned int forkexec_idx;
909 unsigned int smt_gain;
910
911 int nohz_idle; /* NOHZ IDLE status */
912 int flags; /* See SD_* */
913 int level;
914
915 /* Runtime fields. */
916 unsigned long last_balance; /* init to jiffies. units in jiffies */
917 unsigned int balance_interval; /* initialise to 1. units in ms. */
918 unsigned int nr_balance_failed; /* initialise to 0 */
919
920 /* idle_balance() stats */
921 u64 max_newidle_lb_cost;
922 unsigned long next_decay_max_lb_cost;
923
924#ifdef CONFIG_SCHEDSTATS
925 /* load_balance() stats */
926 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
927 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
928 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
929 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
930 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
931 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
932 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
933 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
934
935 /* Active load balancing */
936 unsigned int alb_count;
937 unsigned int alb_failed;
938 unsigned int alb_pushed;
939
940 /* SD_BALANCE_EXEC stats */
941 unsigned int sbe_count;
942 unsigned int sbe_balanced;
943 unsigned int sbe_pushed;
944
945 /* SD_BALANCE_FORK stats */
946 unsigned int sbf_count;
947 unsigned int sbf_balanced;
948 unsigned int sbf_pushed;
949
950 /* try_to_wake_up() stats */
951 unsigned int ttwu_wake_remote;
952 unsigned int ttwu_move_affine;
953 unsigned int ttwu_move_balance;
954#endif
955#ifdef CONFIG_SCHED_DEBUG
956 char *name;
957#endif
958 union {
959 void *private; /* used during construction */
960 struct rcu_head rcu; /* used during destruction */
961 };
962
963 unsigned int span_weight;
964 /*
965 * Span of all CPUs in this domain.
966 *
967 * NOTE: this field is variable length. (Allocated dynamically
968 * by attaching extra space to the end of the structure,
969 * depending on how many CPUs the kernel has booted up with)
970 */
971 unsigned long span[0];
972};
973
974static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
975{
976 return to_cpumask(sd->span);
977}
978
979extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
980 struct sched_domain_attr *dattr_new);
981
982/* Allocate an array of sched domains, for partition_sched_domains(). */
983cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
984void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
985
986bool cpus_share_cache(int this_cpu, int that_cpu);
987
988#else /* CONFIG_SMP */
989
990struct sched_domain_attr;
991
992static inline void
993partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
994 struct sched_domain_attr *dattr_new)
995{
996}
997
998static inline bool cpus_share_cache(int this_cpu, int that_cpu)
999{
1000 return true;
1001}
1002
1003#endif /* !CONFIG_SMP */
1004
1005
1006struct io_context; /* See blkdev.h */
1007
1008
1009#ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1010extern void prefetch_stack(struct task_struct *t);
1011#else
1012static inline void prefetch_stack(struct task_struct *t) { }
1013#endif
1014
1015struct audit_context; /* See audit.c */
1016struct mempolicy;
1017struct pipe_inode_info;
1018struct uts_namespace;
1019
1020struct load_weight {
1021 unsigned long weight;
1022 u32 inv_weight;
1023};
1024
1025struct sched_avg {
1026 /*
1027 * These sums represent an infinite geometric series and so are bound
1028 * above by 1024/(1-y). Thus we only need a u32 to store them for all
1029 * choices of y < 1-2^(-32)*1024.
1030 */
1031 u32 runnable_avg_sum, runnable_avg_period;
1032 u64 last_runnable_update;
1033 s64 decay_count;
1034 unsigned long load_avg_contrib;
1035};
1036
1037#ifdef CONFIG_SCHEDSTATS
1038struct sched_statistics {
1039 u64 wait_start;
1040 u64 wait_max;
1041 u64 wait_count;
1042 u64 wait_sum;
1043 u64 iowait_count;
1044 u64 iowait_sum;
1045
1046 u64 sleep_start;
1047 u64 sleep_max;
1048 s64 sum_sleep_runtime;
1049
1050 u64 block_start;
1051 u64 block_max;
1052 u64 exec_max;
1053 u64 slice_max;
1054
1055 u64 nr_migrations_cold;
1056 u64 nr_failed_migrations_affine;
1057 u64 nr_failed_migrations_running;
1058 u64 nr_failed_migrations_hot;
1059 u64 nr_forced_migrations;
1060
1061 u64 nr_wakeups;
1062 u64 nr_wakeups_sync;
1063 u64 nr_wakeups_migrate;
1064 u64 nr_wakeups_local;
1065 u64 nr_wakeups_remote;
1066 u64 nr_wakeups_affine;
1067 u64 nr_wakeups_affine_attempts;
1068 u64 nr_wakeups_passive;
1069 u64 nr_wakeups_idle;
1070};
1071#endif
1072
1073struct sched_entity {
1074 struct load_weight load; /* for load-balancing */
1075 struct rb_node run_node;
1076 struct list_head group_node;
1077 unsigned int on_rq;
1078
1079 u64 exec_start;
1080 u64 sum_exec_runtime;
1081 u64 vruntime;
1082 u64 prev_sum_exec_runtime;
1083
1084 u64 nr_migrations;
1085
1086#ifdef CONFIG_SCHEDSTATS
1087 struct sched_statistics statistics;
1088#endif
1089
1090#ifdef CONFIG_FAIR_GROUP_SCHED
1091 int depth;
1092 struct sched_entity *parent;
1093 /* rq on which this entity is (to be) queued: */
1094 struct cfs_rq *cfs_rq;
1095 /* rq "owned" by this entity/group: */
1096 struct cfs_rq *my_q;
1097#endif
1098
1099#ifdef CONFIG_SMP
1100 /* Per-entity load-tracking */
1101 struct sched_avg avg;
1102#endif
1103};
1104
1105struct sched_rt_entity {
1106 struct list_head run_list;
1107 unsigned long timeout;
1108 unsigned long watchdog_stamp;
1109 unsigned int time_slice;
1110
1111 struct sched_rt_entity *back;
1112#ifdef CONFIG_RT_GROUP_SCHED
1113 struct sched_rt_entity *parent;
1114 /* rq on which this entity is (to be) queued: */
1115 struct rt_rq *rt_rq;
1116 /* rq "owned" by this entity/group: */
1117 struct rt_rq *my_q;
1118#endif
1119};
1120
1121struct sched_dl_entity {
1122 struct rb_node rb_node;
1123
1124 /*
1125 * Original scheduling parameters. Copied here from sched_attr
1126 * during sched_setscheduler2(), they will remain the same until
1127 * the next sched_setscheduler2().
1128 */
1129 u64 dl_runtime; /* maximum runtime for each instance */
1130 u64 dl_deadline; /* relative deadline of each instance */
1131 u64 dl_period; /* separation of two instances (period) */
1132 u64 dl_bw; /* dl_runtime / dl_deadline */
1133
1134 /*
1135 * Actual scheduling parameters. Initialized with the values above,
1136 * they are continously updated during task execution. Note that
1137 * the remaining runtime could be < 0 in case we are in overrun.
1138 */
1139 s64 runtime; /* remaining runtime for this instance */
1140 u64 deadline; /* absolute deadline for this instance */
1141 unsigned int flags; /* specifying the scheduler behaviour */
1142
1143 /*
1144 * Some bool flags:
1145 *
1146 * @dl_throttled tells if we exhausted the runtime. If so, the
1147 * task has to wait for a replenishment to be performed at the
1148 * next firing of dl_timer.
1149 *
1150 * @dl_new tells if a new instance arrived. If so we must
1151 * start executing it with full runtime and reset its absolute
1152 * deadline;
1153 *
1154 * @dl_boosted tells if we are boosted due to DI. If so we are
1155 * outside bandwidth enforcement mechanism (but only until we
1156 * exit the critical section);
1157 *
1158 * @dl_yielded tells if task gave up the cpu before consuming
1159 * all its available runtime during the last job.
1160 */
1161 int dl_throttled, dl_new, dl_boosted, dl_yielded;
1162
1163 /*
1164 * Bandwidth enforcement timer. Each -deadline task has its
1165 * own bandwidth to be enforced, thus we need one timer per task.
1166 */
1167 struct hrtimer dl_timer;
1168};
1169
1170struct rcu_node;
1171
1172enum perf_event_task_context {
1173 perf_invalid_context = -1,
1174 perf_hw_context = 0,
1175 perf_sw_context,
1176 perf_nr_task_contexts,
1177};
1178
1179struct task_struct {
1180 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1181 void *stack;
1182 atomic_t usage;
1183 unsigned int flags; /* per process flags, defined below */
1184 unsigned int ptrace;
1185
1186#ifdef CONFIG_SMP
1187 struct llist_node wake_entry;
1188 int on_cpu;
1189 struct task_struct *last_wakee;
1190 unsigned long wakee_flips;
1191 unsigned long wakee_flip_decay_ts;
1192
1193 int wake_cpu;
1194#endif
1195 int on_rq;
1196
1197 int prio, static_prio, normal_prio;
1198 unsigned int rt_priority;
1199 const struct sched_class *sched_class;
1200 struct sched_entity se;
1201 struct sched_rt_entity rt;
1202#ifdef CONFIG_CGROUP_SCHED
1203 struct task_group *sched_task_group;
1204#endif
1205 struct sched_dl_entity dl;
1206
1207#ifdef CONFIG_PREEMPT_NOTIFIERS
1208 /* list of struct preempt_notifier: */
1209 struct hlist_head preempt_notifiers;
1210#endif
1211
1212#ifdef CONFIG_BLK_DEV_IO_TRACE
1213 unsigned int btrace_seq;
1214#endif
1215
1216 unsigned int policy;
1217 int nr_cpus_allowed;
1218 cpumask_t cpus_allowed;
1219
1220#ifdef CONFIG_PREEMPT_RCU
1221 int rcu_read_lock_nesting;
1222 char rcu_read_unlock_special;
1223 struct list_head rcu_node_entry;
1224#endif /* #ifdef CONFIG_PREEMPT_RCU */
1225#ifdef CONFIG_TREE_PREEMPT_RCU
1226 struct rcu_node *rcu_blocked_node;
1227#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
1228#ifdef CONFIG_RCU_BOOST
1229 struct rt_mutex *rcu_boost_mutex;
1230#endif /* #ifdef CONFIG_RCU_BOOST */
1231
1232#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
1233 struct sched_info sched_info;
1234#endif
1235
1236 struct list_head tasks;
1237#ifdef CONFIG_SMP
1238 struct plist_node pushable_tasks;
1239 struct rb_node pushable_dl_tasks;
1240#endif
1241
1242 struct mm_struct *mm, *active_mm;
1243#ifdef CONFIG_COMPAT_BRK
1244 unsigned brk_randomized:1;
1245#endif
1246 /* per-thread vma caching */
1247 u32 vmacache_seqnum;
1248 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1249#if defined(SPLIT_RSS_COUNTING)
1250 struct task_rss_stat rss_stat;
1251#endif
1252/* task state */
1253 int exit_state;
1254 int exit_code, exit_signal;
1255 int pdeath_signal; /* The signal sent when the parent dies */
1256 unsigned int jobctl; /* JOBCTL_*, siglock protected */
1257
1258 /* Used for emulating ABI behavior of previous Linux versions */
1259 unsigned int personality;
1260
1261 unsigned in_execve:1; /* Tell the LSMs that the process is doing an
1262 * execve */
1263 unsigned in_iowait:1;
1264
1265 /* task may not gain privileges */
1266 unsigned no_new_privs:1;
1267
1268 /* Revert to default priority/policy when forking */
1269 unsigned sched_reset_on_fork:1;
1270 unsigned sched_contributes_to_load:1;
1271
1272 pid_t pid;
1273 pid_t tgid;
1274
1275#ifdef CONFIG_CC_STACKPROTECTOR
1276 /* Canary value for the -fstack-protector gcc feature */
1277 unsigned long stack_canary;
1278#endif
1279 /*
1280 * pointers to (original) parent process, youngest child, younger sibling,
1281 * older sibling, respectively. (p->father can be replaced with
1282 * p->real_parent->pid)
1283 */
1284 struct task_struct __rcu *real_parent; /* real parent process */
1285 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1286 /*
1287 * children/sibling forms the list of my natural children
1288 */
1289 struct list_head children; /* list of my children */
1290 struct list_head sibling; /* linkage in my parent's children list */
1291 struct task_struct *group_leader; /* threadgroup leader */
1292
1293 /*
1294 * ptraced is the list of tasks this task is using ptrace on.
1295 * This includes both natural children and PTRACE_ATTACH targets.
1296 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1297 */
1298 struct list_head ptraced;
1299 struct list_head ptrace_entry;
1300
1301 /* PID/PID hash table linkage. */
1302 struct pid_link pids[PIDTYPE_MAX];
1303 struct list_head thread_group;
1304 struct list_head thread_node;
1305
1306 struct completion *vfork_done; /* for vfork() */
1307 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1308 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1309
1310 cputime_t utime, stime, utimescaled, stimescaled;
1311 cputime_t gtime;
1312#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
1313 struct cputime prev_cputime;
1314#endif
1315#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1316 seqlock_t vtime_seqlock;
1317 unsigned long long vtime_snap;
1318 enum {
1319 VTIME_SLEEPING = 0,
1320 VTIME_USER,
1321 VTIME_SYS,
1322 } vtime_snap_whence;
1323#endif
1324 unsigned long nvcsw, nivcsw; /* context switch counts */
1325 struct timespec start_time; /* monotonic time */
1326 struct timespec real_start_time; /* boot based time */
1327/* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1328 unsigned long min_flt, maj_flt;
1329
1330 struct task_cputime cputime_expires;
1331 struct list_head cpu_timers[3];
1332
1333/* process credentials */
1334 const struct cred __rcu *real_cred; /* objective and real subjective task
1335 * credentials (COW) */
1336 const struct cred __rcu *cred; /* effective (overridable) subjective task
1337 * credentials (COW) */
1338 char comm[TASK_COMM_LEN]; /* executable name excluding path
1339 - access with [gs]et_task_comm (which lock
1340 it with task_lock())
1341 - initialized normally by setup_new_exec */
1342/* file system info */
1343 int link_count, total_link_count;
1344#ifdef CONFIG_SYSVIPC
1345/* ipc stuff */
1346 struct sysv_sem sysvsem;
1347#endif
1348#ifdef CONFIG_DETECT_HUNG_TASK
1349/* hung task detection */
1350 unsigned long last_switch_count;
1351#endif
1352/* CPU-specific state of this task */
1353 struct thread_struct thread;
1354/* filesystem information */
1355 struct fs_struct *fs;
1356/* open file information */
1357 struct files_struct *files;
1358/* namespaces */
1359 struct nsproxy *nsproxy;
1360/* signal handlers */
1361 struct signal_struct *signal;
1362 struct sighand_struct *sighand;
1363
1364 sigset_t blocked, real_blocked;
1365 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1366 struct sigpending pending;
1367
1368 unsigned long sas_ss_sp;
1369 size_t sas_ss_size;
1370 int (*notifier)(void *priv);
1371 void *notifier_data;
1372 sigset_t *notifier_mask;
1373 struct callback_head *task_works;
1374
1375 struct audit_context *audit_context;
1376#ifdef CONFIG_AUDITSYSCALL
1377 kuid_t loginuid;
1378 unsigned int sessionid;
1379#endif
1380 struct seccomp seccomp;
1381
1382/* Thread group tracking */
1383 u32 parent_exec_id;
1384 u32 self_exec_id;
1385/* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1386 * mempolicy */
1387 spinlock_t alloc_lock;
1388
1389 /* Protection of the PI data structures: */
1390 raw_spinlock_t pi_lock;
1391
1392#ifdef CONFIG_RT_MUTEXES
1393 /* PI waiters blocked on a rt_mutex held by this task */
1394 struct rb_root pi_waiters;
1395 struct rb_node *pi_waiters_leftmost;
1396 /* Deadlock detection and priority inheritance handling */
1397 struct rt_mutex_waiter *pi_blocked_on;
1398 /* Top pi_waiters task */
1399 struct task_struct *pi_top_task;
1400#endif
1401
1402#ifdef CONFIG_DEBUG_MUTEXES
1403 /* mutex deadlock detection */
1404 struct mutex_waiter *blocked_on;
1405#endif
1406#ifdef CONFIG_TRACE_IRQFLAGS
1407 unsigned int irq_events;
1408 unsigned long hardirq_enable_ip;
1409 unsigned long hardirq_disable_ip;
1410 unsigned int hardirq_enable_event;
1411 unsigned int hardirq_disable_event;
1412 int hardirqs_enabled;
1413 int hardirq_context;
1414 unsigned long softirq_disable_ip;
1415 unsigned long softirq_enable_ip;
1416 unsigned int softirq_disable_event;
1417 unsigned int softirq_enable_event;
1418 int softirqs_enabled;
1419 int softirq_context;
1420#endif
1421#ifdef CONFIG_LOCKDEP
1422# define MAX_LOCK_DEPTH 48UL
1423 u64 curr_chain_key;
1424 int lockdep_depth;
1425 unsigned int lockdep_recursion;
1426 struct held_lock held_locks[MAX_LOCK_DEPTH];
1427 gfp_t lockdep_reclaim_gfp;
1428#endif
1429
1430/* journalling filesystem info */
1431 void *journal_info;
1432
1433/* stacked block device info */
1434 struct bio_list *bio_list;
1435
1436#ifdef CONFIG_BLOCK
1437/* stack plugging */
1438 struct blk_plug *plug;
1439#endif
1440
1441/* VM state */
1442 struct reclaim_state *reclaim_state;
1443
1444 struct backing_dev_info *backing_dev_info;
1445
1446 struct io_context *io_context;
1447
1448 unsigned long ptrace_message;
1449 siginfo_t *last_siginfo; /* For ptrace use. */
1450 struct task_io_accounting ioac;
1451#if defined(CONFIG_TASK_XACCT)
1452 u64 acct_rss_mem1; /* accumulated rss usage */
1453 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1454 cputime_t acct_timexpd; /* stime + utime since last update */
1455#endif
1456#ifdef CONFIG_CPUSETS
1457 nodemask_t mems_allowed; /* Protected by alloc_lock */
1458 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1459 int cpuset_mem_spread_rotor;
1460 int cpuset_slab_spread_rotor;
1461#endif
1462#ifdef CONFIG_CGROUPS
1463 /* Control Group info protected by css_set_lock */
1464 struct css_set __rcu *cgroups;
1465 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1466 struct list_head cg_list;
1467#endif
1468#ifdef CONFIG_FUTEX
1469 struct robust_list_head __user *robust_list;
1470#ifdef CONFIG_COMPAT
1471 struct compat_robust_list_head __user *compat_robust_list;
1472#endif
1473 struct list_head pi_state_list;
1474 struct futex_pi_state *pi_state_cache;
1475#endif
1476#ifdef CONFIG_PERF_EVENTS
1477 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1478 struct mutex perf_event_mutex;
1479 struct list_head perf_event_list;
1480#endif
1481#ifdef CONFIG_DEBUG_PREEMPT
1482 unsigned long preempt_disable_ip;
1483#endif
1484#ifdef CONFIG_NUMA
1485 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1486 short il_next;
1487 short pref_node_fork;
1488#endif
1489#ifdef CONFIG_NUMA_BALANCING
1490 int numa_scan_seq;
1491 unsigned int numa_scan_period;
1492 unsigned int numa_scan_period_max;
1493 int numa_preferred_nid;
1494 unsigned long numa_migrate_retry;
1495 u64 node_stamp; /* migration stamp */
1496 u64 last_task_numa_placement;
1497 u64 last_sum_exec_runtime;
1498 struct callback_head numa_work;
1499
1500 struct list_head numa_entry;
1501 struct numa_group *numa_group;
1502
1503 /*
1504 * Exponential decaying average of faults on a per-node basis.
1505 * Scheduling placement decisions are made based on the these counts.
1506 * The values remain static for the duration of a PTE scan
1507 */
1508 unsigned long *numa_faults_memory;
1509 unsigned long total_numa_faults;
1510
1511 /*
1512 * numa_faults_buffer records faults per node during the current
1513 * scan window. When the scan completes, the counts in
1514 * numa_faults_memory decay and these values are copied.
1515 */
1516 unsigned long *numa_faults_buffer_memory;
1517
1518 /*
1519 * Track the nodes the process was running on when a NUMA hinting
1520 * fault was incurred.
1521 */
1522 unsigned long *numa_faults_cpu;
1523 unsigned long *numa_faults_buffer_cpu;
1524
1525 /*
1526 * numa_faults_locality tracks if faults recorded during the last
1527 * scan window were remote/local. The task scan period is adapted
1528 * based on the locality of the faults with different weights
1529 * depending on whether they were shared or private faults
1530 */
1531 unsigned long numa_faults_locality[2];
1532
1533 unsigned long numa_pages_migrated;
1534#endif /* CONFIG_NUMA_BALANCING */
1535
1536 struct rcu_head rcu;
1537
1538 /*
1539 * cache last used pipe for splice
1540 */
1541 struct pipe_inode_info *splice_pipe;
1542
1543 struct page_frag task_frag;
1544
1545#ifdef CONFIG_TASK_DELAY_ACCT
1546 struct task_delay_info *delays;
1547#endif
1548#ifdef CONFIG_FAULT_INJECTION
1549 int make_it_fail;
1550#endif
1551 /*
1552 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1553 * balance_dirty_pages() for some dirty throttling pause
1554 */
1555 int nr_dirtied;
1556 int nr_dirtied_pause;
1557 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1558
1559#ifdef CONFIG_LATENCYTOP
1560 int latency_record_count;
1561 struct latency_record latency_record[LT_SAVECOUNT];
1562#endif
1563 /*
1564 * time slack values; these are used to round up poll() and
1565 * select() etc timeout values. These are in nanoseconds.
1566 */
1567 unsigned long timer_slack_ns;
1568 unsigned long default_timer_slack_ns;
1569
1570#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1571 /* Index of current stored address in ret_stack */
1572 int curr_ret_stack;
1573 /* Stack of return addresses for return function tracing */
1574 struct ftrace_ret_stack *ret_stack;
1575 /* time stamp for last schedule */
1576 unsigned long long ftrace_timestamp;
1577 /*
1578 * Number of functions that haven't been traced
1579 * because of depth overrun.
1580 */
1581 atomic_t trace_overrun;
1582 /* Pause for the tracing */
1583 atomic_t tracing_graph_pause;
1584#endif
1585#ifdef CONFIG_TRACING
1586 /* state flags for use by tracers */
1587 unsigned long trace;
1588 /* bitmask and counter of trace recursion */
1589 unsigned long trace_recursion;
1590#endif /* CONFIG_TRACING */
1591#ifdef CONFIG_MEMCG /* memcg uses this to do batch job */
1592 struct memcg_batch_info {
1593 int do_batch; /* incremented when batch uncharge started */
1594 struct mem_cgroup *memcg; /* target memcg of uncharge */
1595 unsigned long nr_pages; /* uncharged usage */
1596 unsigned long memsw_nr_pages; /* uncharged mem+swap usage */
1597 } memcg_batch;
1598 unsigned int memcg_kmem_skip_account;
1599 struct memcg_oom_info {
1600 struct mem_cgroup *memcg;
1601 gfp_t gfp_mask;
1602 int order;
1603 unsigned int may_oom:1;
1604 } memcg_oom;
1605#endif
1606#ifdef CONFIG_UPROBES
1607 struct uprobe_task *utask;
1608#endif
1609#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1610 unsigned int sequential_io;
1611 unsigned int sequential_io_avg;
1612#endif
1613};
1614
1615/* Future-safe accessor for struct task_struct's cpus_allowed. */
1616#define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1617
1618#define TNF_MIGRATED 0x01
1619#define TNF_NO_GROUP 0x02
1620#define TNF_SHARED 0x04
1621#define TNF_FAULT_LOCAL 0x08
1622
1623#ifdef CONFIG_NUMA_BALANCING
1624extern void task_numa_fault(int last_node, int node, int pages, int flags);
1625extern pid_t task_numa_group_id(struct task_struct *p);
1626extern void set_numabalancing_state(bool enabled);
1627extern void task_numa_free(struct task_struct *p);
1628extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
1629 int src_nid, int dst_cpu);
1630#else
1631static inline void task_numa_fault(int last_node, int node, int pages,
1632 int flags)
1633{
1634}
1635static inline pid_t task_numa_group_id(struct task_struct *p)
1636{
1637 return 0;
1638}
1639static inline void set_numabalancing_state(bool enabled)
1640{
1641}
1642static inline void task_numa_free(struct task_struct *p)
1643{
1644}
1645static inline bool should_numa_migrate_memory(struct task_struct *p,
1646 struct page *page, int src_nid, int dst_cpu)
1647{
1648 return true;
1649}
1650#endif
1651
1652static inline struct pid *task_pid(struct task_struct *task)
1653{
1654 return task->pids[PIDTYPE_PID].pid;
1655}
1656
1657static inline struct pid *task_tgid(struct task_struct *task)
1658{
1659 return task->group_leader->pids[PIDTYPE_PID].pid;
1660}
1661
1662/*
1663 * Without tasklist or rcu lock it is not safe to dereference
1664 * the result of task_pgrp/task_session even if task == current,
1665 * we can race with another thread doing sys_setsid/sys_setpgid.
1666 */
1667static inline struct pid *task_pgrp(struct task_struct *task)
1668{
1669 return task->group_leader->pids[PIDTYPE_PGID].pid;
1670}
1671
1672static inline struct pid *task_session(struct task_struct *task)
1673{
1674 return task->group_leader->pids[PIDTYPE_SID].pid;
1675}
1676
1677struct pid_namespace;
1678
1679/*
1680 * the helpers to get the task's different pids as they are seen
1681 * from various namespaces
1682 *
1683 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1684 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1685 * current.
1686 * task_xid_nr_ns() : id seen from the ns specified;
1687 *
1688 * set_task_vxid() : assigns a virtual id to a task;
1689 *
1690 * see also pid_nr() etc in include/linux/pid.h
1691 */
1692pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1693 struct pid_namespace *ns);
1694
1695static inline pid_t task_pid_nr(struct task_struct *tsk)
1696{
1697 return tsk->pid;
1698}
1699
1700static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1701 struct pid_namespace *ns)
1702{
1703 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1704}
1705
1706static inline pid_t task_pid_vnr(struct task_struct *tsk)
1707{
1708 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1709}
1710
1711
1712static inline pid_t task_tgid_nr(struct task_struct *tsk)
1713{
1714 return tsk->tgid;
1715}
1716
1717pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1718
1719static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1720{
1721 return pid_vnr(task_tgid(tsk));
1722}
1723
1724
1725static inline int pid_alive(const struct task_struct *p);
1726static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1727{
1728 pid_t pid = 0;
1729
1730 rcu_read_lock();
1731 if (pid_alive(tsk))
1732 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1733 rcu_read_unlock();
1734
1735 return pid;
1736}
1737
1738static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1739{
1740 return task_ppid_nr_ns(tsk, &init_pid_ns);
1741}
1742
1743static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1744 struct pid_namespace *ns)
1745{
1746 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1747}
1748
1749static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1750{
1751 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1752}
1753
1754
1755static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1756 struct pid_namespace *ns)
1757{
1758 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1759}
1760
1761static inline pid_t task_session_vnr(struct task_struct *tsk)
1762{
1763 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1764}
1765
1766/* obsolete, do not use */
1767static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1768{
1769 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1770}
1771
1772/**
1773 * pid_alive - check that a task structure is not stale
1774 * @p: Task structure to be checked.
1775 *
1776 * Test if a process is not yet dead (at most zombie state)
1777 * If pid_alive fails, then pointers within the task structure
1778 * can be stale and must not be dereferenced.
1779 *
1780 * Return: 1 if the process is alive. 0 otherwise.
1781 */
1782static inline int pid_alive(const struct task_struct *p)
1783{
1784 return p->pids[PIDTYPE_PID].pid != NULL;
1785}
1786
1787/**
1788 * is_global_init - check if a task structure is init
1789 * @tsk: Task structure to be checked.
1790 *
1791 * Check if a task structure is the first user space task the kernel created.
1792 *
1793 * Return: 1 if the task structure is init. 0 otherwise.
1794 */
1795static inline int is_global_init(struct task_struct *tsk)
1796{
1797 return tsk->pid == 1;
1798}
1799
1800extern struct pid *cad_pid;
1801
1802extern void free_task(struct task_struct *tsk);
1803#define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
1804
1805extern void __put_task_struct(struct task_struct *t);
1806
1807static inline void put_task_struct(struct task_struct *t)
1808{
1809 if (atomic_dec_and_test(&t->usage))
1810 __put_task_struct(t);
1811}
1812
1813#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1814extern void task_cputime(struct task_struct *t,
1815 cputime_t *utime, cputime_t *stime);
1816extern void task_cputime_scaled(struct task_struct *t,
1817 cputime_t *utimescaled, cputime_t *stimescaled);
1818extern cputime_t task_gtime(struct task_struct *t);
1819#else
1820static inline void task_cputime(struct task_struct *t,
1821 cputime_t *utime, cputime_t *stime)
1822{
1823 if (utime)
1824 *utime = t->utime;
1825 if (stime)
1826 *stime = t->stime;
1827}
1828
1829static inline void task_cputime_scaled(struct task_struct *t,
1830 cputime_t *utimescaled,
1831 cputime_t *stimescaled)
1832{
1833 if (utimescaled)
1834 *utimescaled = t->utimescaled;
1835 if (stimescaled)
1836 *stimescaled = t->stimescaled;
1837}
1838
1839static inline cputime_t task_gtime(struct task_struct *t)
1840{
1841 return t->gtime;
1842}
1843#endif
1844extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
1845extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
1846
1847/*
1848 * Per process flags
1849 */
1850#define PF_EXITING 0x00000004 /* getting shut down */
1851#define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1852#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1853#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1854#define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1855#define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1856#define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1857#define PF_DUMPCORE 0x00000200 /* dumped core */
1858#define PF_SIGNALED 0x00000400 /* killed by a signal */
1859#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1860#define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1861#define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1862#define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1863#define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1864#define PF_FROZEN 0x00010000 /* frozen for system suspend */
1865#define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1866#define PF_KSWAPD 0x00040000 /* I am kswapd */
1867#define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1868#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1869#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1870#define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1871#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1872#define PF_SPREAD_PAGE 0x01000000 /* Spread page cache over cpuset */
1873#define PF_SPREAD_SLAB 0x02000000 /* Spread some slab caches over cpuset */
1874#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1875#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1876#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1877#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1878#define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1879
1880/*
1881 * Only the _current_ task can read/write to tsk->flags, but other
1882 * tasks can access tsk->flags in readonly mode for example
1883 * with tsk_used_math (like during threaded core dumping).
1884 * There is however an exception to this rule during ptrace
1885 * or during fork: the ptracer task is allowed to write to the
1886 * child->flags of its traced child (same goes for fork, the parent
1887 * can write to the child->flags), because we're guaranteed the
1888 * child is not running and in turn not changing child->flags
1889 * at the same time the parent does it.
1890 */
1891#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1892#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1893#define clear_used_math() clear_stopped_child_used_math(current)
1894#define set_used_math() set_stopped_child_used_math(current)
1895#define conditional_stopped_child_used_math(condition, child) \
1896 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1897#define conditional_used_math(condition) \
1898 conditional_stopped_child_used_math(condition, current)
1899#define copy_to_stopped_child_used_math(child) \
1900 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1901/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1902#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1903#define used_math() tsk_used_math(current)
1904
1905/* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags */
1906static inline gfp_t memalloc_noio_flags(gfp_t flags)
1907{
1908 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
1909 flags &= ~__GFP_IO;
1910 return flags;
1911}
1912
1913static inline unsigned int memalloc_noio_save(void)
1914{
1915 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
1916 current->flags |= PF_MEMALLOC_NOIO;
1917 return flags;
1918}
1919
1920static inline void memalloc_noio_restore(unsigned int flags)
1921{
1922 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
1923}
1924
1925/*
1926 * task->jobctl flags
1927 */
1928#define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
1929
1930#define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
1931#define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
1932#define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
1933#define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
1934#define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
1935#define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
1936#define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
1937
1938#define JOBCTL_STOP_DEQUEUED (1 << JOBCTL_STOP_DEQUEUED_BIT)
1939#define JOBCTL_STOP_PENDING (1 << JOBCTL_STOP_PENDING_BIT)
1940#define JOBCTL_STOP_CONSUME (1 << JOBCTL_STOP_CONSUME_BIT)
1941#define JOBCTL_TRAP_STOP (1 << JOBCTL_TRAP_STOP_BIT)
1942#define JOBCTL_TRAP_NOTIFY (1 << JOBCTL_TRAP_NOTIFY_BIT)
1943#define JOBCTL_TRAPPING (1 << JOBCTL_TRAPPING_BIT)
1944#define JOBCTL_LISTENING (1 << JOBCTL_LISTENING_BIT)
1945
1946#define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
1947#define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
1948
1949extern bool task_set_jobctl_pending(struct task_struct *task,
1950 unsigned int mask);
1951extern void task_clear_jobctl_trapping(struct task_struct *task);
1952extern void task_clear_jobctl_pending(struct task_struct *task,
1953 unsigned int mask);
1954
1955#ifdef CONFIG_PREEMPT_RCU
1956
1957#define RCU_READ_UNLOCK_BLOCKED (1 << 0) /* blocked while in RCU read-side. */
1958#define RCU_READ_UNLOCK_NEED_QS (1 << 1) /* RCU core needs CPU response. */
1959
1960static inline void rcu_copy_process(struct task_struct *p)
1961{
1962 p->rcu_read_lock_nesting = 0;
1963 p->rcu_read_unlock_special = 0;
1964#ifdef CONFIG_TREE_PREEMPT_RCU
1965 p->rcu_blocked_node = NULL;
1966#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
1967#ifdef CONFIG_RCU_BOOST
1968 p->rcu_boost_mutex = NULL;
1969#endif /* #ifdef CONFIG_RCU_BOOST */
1970 INIT_LIST_HEAD(&p->rcu_node_entry);
1971}
1972
1973#else
1974
1975static inline void rcu_copy_process(struct task_struct *p)
1976{
1977}
1978
1979#endif
1980
1981static inline void tsk_restore_flags(struct task_struct *task,
1982 unsigned long orig_flags, unsigned long flags)
1983{
1984 task->flags &= ~flags;
1985 task->flags |= orig_flags & flags;
1986}
1987
1988#ifdef CONFIG_SMP
1989extern void do_set_cpus_allowed(struct task_struct *p,
1990 const struct cpumask *new_mask);
1991
1992extern int set_cpus_allowed_ptr(struct task_struct *p,
1993 const struct cpumask *new_mask);
1994#else
1995static inline void do_set_cpus_allowed(struct task_struct *p,
1996 const struct cpumask *new_mask)
1997{
1998}
1999static inline int set_cpus_allowed_ptr(struct task_struct *p,
2000 const struct cpumask *new_mask)
2001{
2002 if (!cpumask_test_cpu(0, new_mask))
2003 return -EINVAL;
2004 return 0;
2005}
2006#endif
2007
2008#ifdef CONFIG_NO_HZ_COMMON
2009void calc_load_enter_idle(void);
2010void calc_load_exit_idle(void);
2011#else
2012static inline void calc_load_enter_idle(void) { }
2013static inline void calc_load_exit_idle(void) { }
2014#endif /* CONFIG_NO_HZ_COMMON */
2015
2016#ifndef CONFIG_CPUMASK_OFFSTACK
2017static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask)
2018{
2019 return set_cpus_allowed_ptr(p, &new_mask);
2020}
2021#endif
2022
2023/*
2024 * Do not use outside of architecture code which knows its limitations.
2025 *
2026 * sched_clock() has no promise of monotonicity or bounded drift between
2027 * CPUs, use (which you should not) requires disabling IRQs.
2028 *
2029 * Please use one of the three interfaces below.
2030 */
2031extern unsigned long long notrace sched_clock(void);
2032/*
2033 * See the comment in kernel/sched/clock.c
2034 */
2035extern u64 cpu_clock(int cpu);
2036extern u64 local_clock(void);
2037extern u64 sched_clock_cpu(int cpu);
2038
2039
2040extern void sched_clock_init(void);
2041
2042#ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2043static inline void sched_clock_tick(void)
2044{
2045}
2046
2047static inline void sched_clock_idle_sleep_event(void)
2048{
2049}
2050
2051static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2052{
2053}
2054#else
2055/*
2056 * Architectures can set this to 1 if they have specified
2057 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2058 * but then during bootup it turns out that sched_clock()
2059 * is reliable after all:
2060 */
2061extern int sched_clock_stable(void);
2062extern void set_sched_clock_stable(void);
2063extern void clear_sched_clock_stable(void);
2064
2065extern void sched_clock_tick(void);
2066extern void sched_clock_idle_sleep_event(void);
2067extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2068#endif
2069
2070#ifdef CONFIG_IRQ_TIME_ACCOUNTING
2071/*
2072 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2073 * The reason for this explicit opt-in is not to have perf penalty with
2074 * slow sched_clocks.
2075 */
2076extern void enable_sched_clock_irqtime(void);
2077extern void disable_sched_clock_irqtime(void);
2078#else
2079static inline void enable_sched_clock_irqtime(void) {}
2080static inline void disable_sched_clock_irqtime(void) {}
2081#endif
2082
2083extern unsigned long long
2084task_sched_runtime(struct task_struct *task);
2085
2086/* sched_exec is called by processes performing an exec */
2087#ifdef CONFIG_SMP
2088extern void sched_exec(void);
2089#else
2090#define sched_exec() {}
2091#endif
2092
2093extern void sched_clock_idle_sleep_event(void);
2094extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2095
2096#ifdef CONFIG_HOTPLUG_CPU
2097extern void idle_task_exit(void);
2098#else
2099static inline void idle_task_exit(void) {}
2100#endif
2101
2102#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2103extern void wake_up_nohz_cpu(int cpu);
2104#else
2105static inline void wake_up_nohz_cpu(int cpu) { }
2106#endif
2107
2108#ifdef CONFIG_NO_HZ_FULL
2109extern bool sched_can_stop_tick(void);
2110extern u64 scheduler_tick_max_deferment(void);
2111#else
2112static inline bool sched_can_stop_tick(void) { return false; }
2113#endif
2114
2115#ifdef CONFIG_SCHED_AUTOGROUP
2116extern void sched_autogroup_create_attach(struct task_struct *p);
2117extern void sched_autogroup_detach(struct task_struct *p);
2118extern void sched_autogroup_fork(struct signal_struct *sig);
2119extern void sched_autogroup_exit(struct signal_struct *sig);
2120#ifdef CONFIG_PROC_FS
2121extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2122extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2123#endif
2124#else
2125static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2126static inline void sched_autogroup_detach(struct task_struct *p) { }
2127static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2128static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2129#endif
2130
2131extern bool yield_to(struct task_struct *p, bool preempt);
2132extern void set_user_nice(struct task_struct *p, long nice);
2133extern int task_prio(const struct task_struct *p);
2134/**
2135 * task_nice - return the nice value of a given task.
2136 * @p: the task in question.
2137 *
2138 * Return: The nice value [ -20 ... 0 ... 19 ].
2139 */
2140static inline int task_nice(const struct task_struct *p)
2141{
2142 return PRIO_TO_NICE((p)->static_prio);
2143}
2144extern int can_nice(const struct task_struct *p, const int nice);
2145extern int task_curr(const struct task_struct *p);
2146extern int idle_cpu(int cpu);
2147extern int sched_setscheduler(struct task_struct *, int,
2148 const struct sched_param *);
2149extern int sched_setscheduler_nocheck(struct task_struct *, int,
2150 const struct sched_param *);
2151extern int sched_setattr(struct task_struct *,
2152 const struct sched_attr *);
2153extern struct task_struct *idle_task(int cpu);
2154/**
2155 * is_idle_task - is the specified task an idle task?
2156 * @p: the task in question.
2157 *
2158 * Return: 1 if @p is an idle task. 0 otherwise.
2159 */
2160static inline bool is_idle_task(const struct task_struct *p)
2161{
2162 return p->pid == 0;
2163}
2164extern struct task_struct *curr_task(int cpu);
2165extern void set_curr_task(int cpu, struct task_struct *p);
2166
2167void yield(void);
2168
2169/*
2170 * The default (Linux) execution domain.
2171 */
2172extern struct exec_domain default_exec_domain;
2173
2174union thread_union {
2175 struct thread_info thread_info;
2176 unsigned long stack[THREAD_SIZE/sizeof(long)];
2177};
2178
2179#ifndef __HAVE_ARCH_KSTACK_END
2180static inline int kstack_end(void *addr)
2181{
2182 /* Reliable end of stack detection:
2183 * Some APM bios versions misalign the stack
2184 */
2185 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2186}
2187#endif
2188
2189extern union thread_union init_thread_union;
2190extern struct task_struct init_task;
2191
2192extern struct mm_struct init_mm;
2193
2194extern struct pid_namespace init_pid_ns;
2195
2196/*
2197 * find a task by one of its numerical ids
2198 *
2199 * find_task_by_pid_ns():
2200 * finds a task by its pid in the specified namespace
2201 * find_task_by_vpid():
2202 * finds a task by its virtual pid
2203 *
2204 * see also find_vpid() etc in include/linux/pid.h
2205 */
2206
2207extern struct task_struct *find_task_by_vpid(pid_t nr);
2208extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2209 struct pid_namespace *ns);
2210
2211/* per-UID process charging. */
2212extern struct user_struct * alloc_uid(kuid_t);
2213static inline struct user_struct *get_uid(struct user_struct *u)
2214{
2215 atomic_inc(&u->__count);
2216 return u;
2217}
2218extern void free_uid(struct user_struct *);
2219
2220#include <asm/current.h>
2221
2222extern void xtime_update(unsigned long ticks);
2223
2224extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2225extern int wake_up_process(struct task_struct *tsk);
2226extern void wake_up_new_task(struct task_struct *tsk);
2227#ifdef CONFIG_SMP
2228 extern void kick_process(struct task_struct *tsk);
2229#else
2230 static inline void kick_process(struct task_struct *tsk) { }
2231#endif
2232extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2233extern void sched_dead(struct task_struct *p);
2234
2235extern void proc_caches_init(void);
2236extern void flush_signals(struct task_struct *);
2237extern void __flush_signals(struct task_struct *);
2238extern void ignore_signals(struct task_struct *);
2239extern void flush_signal_handlers(struct task_struct *, int force_default);
2240extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2241
2242static inline int dequeue_signal_lock(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
2243{
2244 unsigned long flags;
2245 int ret;
2246
2247 spin_lock_irqsave(&tsk->sighand->siglock, flags);
2248 ret = dequeue_signal(tsk, mask, info);
2249 spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
2250
2251 return ret;
2252}
2253
2254extern void block_all_signals(int (*notifier)(void *priv), void *priv,
2255 sigset_t *mask);
2256extern void unblock_all_signals(void);
2257extern void release_task(struct task_struct * p);
2258extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2259extern int force_sigsegv(int, struct task_struct *);
2260extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2261extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2262extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2263extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2264 const struct cred *, u32);
2265extern int kill_pgrp(struct pid *pid, int sig, int priv);
2266extern int kill_pid(struct pid *pid, int sig, int priv);
2267extern int kill_proc_info(int, struct siginfo *, pid_t);
2268extern __must_check bool do_notify_parent(struct task_struct *, int);
2269extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2270extern void force_sig(int, struct task_struct *);
2271extern int send_sig(int, struct task_struct *, int);
2272extern int zap_other_threads(struct task_struct *p);
2273extern struct sigqueue *sigqueue_alloc(void);
2274extern void sigqueue_free(struct sigqueue *);
2275extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2276extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2277
2278static inline void restore_saved_sigmask(void)
2279{
2280 if (test_and_clear_restore_sigmask())
2281 __set_current_blocked(¤t->saved_sigmask);
2282}
2283
2284static inline sigset_t *sigmask_to_save(void)
2285{
2286 sigset_t *res = ¤t->blocked;
2287 if (unlikely(test_restore_sigmask()))
2288 res = ¤t->saved_sigmask;
2289 return res;
2290}
2291
2292static inline int kill_cad_pid(int sig, int priv)
2293{
2294 return kill_pid(cad_pid, sig, priv);
2295}
2296
2297/* These can be the second arg to send_sig_info/send_group_sig_info. */
2298#define SEND_SIG_NOINFO ((struct siginfo *) 0)
2299#define SEND_SIG_PRIV ((struct siginfo *) 1)
2300#define SEND_SIG_FORCED ((struct siginfo *) 2)
2301
2302/*
2303 * True if we are on the alternate signal stack.
2304 */
2305static inline int on_sig_stack(unsigned long sp)
2306{
2307#ifdef CONFIG_STACK_GROWSUP
2308 return sp >= current->sas_ss_sp &&
2309 sp - current->sas_ss_sp < current->sas_ss_size;
2310#else
2311 return sp > current->sas_ss_sp &&
2312 sp - current->sas_ss_sp <= current->sas_ss_size;
2313#endif
2314}
2315
2316static inline int sas_ss_flags(unsigned long sp)
2317{
2318 return (current->sas_ss_size == 0 ? SS_DISABLE
2319 : on_sig_stack(sp) ? SS_ONSTACK : 0);
2320}
2321
2322static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2323{
2324 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2325#ifdef CONFIG_STACK_GROWSUP
2326 return current->sas_ss_sp;
2327#else
2328 return current->sas_ss_sp + current->sas_ss_size;
2329#endif
2330 return sp;
2331}
2332
2333/*
2334 * Routines for handling mm_structs
2335 */
2336extern struct mm_struct * mm_alloc(void);
2337
2338/* mmdrop drops the mm and the page tables */
2339extern void __mmdrop(struct mm_struct *);
2340static inline void mmdrop(struct mm_struct * mm)
2341{
2342 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2343 __mmdrop(mm);
2344}
2345
2346/* mmput gets rid of the mappings and all user-space */
2347extern void mmput(struct mm_struct *);
2348/* Grab a reference to a task's mm, if it is not already going away */
2349extern struct mm_struct *get_task_mm(struct task_struct *task);
2350/*
2351 * Grab a reference to a task's mm, if it is not already going away
2352 * and ptrace_may_access with the mode parameter passed to it
2353 * succeeds.
2354 */
2355extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2356/* Remove the current tasks stale references to the old mm_struct */
2357extern void mm_release(struct task_struct *, struct mm_struct *);
2358
2359extern int copy_thread(unsigned long, unsigned long, unsigned long,
2360 struct task_struct *);
2361extern void flush_thread(void);
2362extern void exit_thread(void);
2363
2364extern void exit_files(struct task_struct *);
2365extern void __cleanup_sighand(struct sighand_struct *);
2366
2367extern void exit_itimers(struct signal_struct *);
2368extern void flush_itimer_signals(void);
2369
2370extern void do_group_exit(int);
2371
2372extern int allow_signal(int);
2373extern int disallow_signal(int);
2374
2375extern int do_execve(struct filename *,
2376 const char __user * const __user *,
2377 const char __user * const __user *);
2378extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2379struct task_struct *fork_idle(int);
2380extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2381
2382extern void set_task_comm(struct task_struct *tsk, const char *from);
2383extern char *get_task_comm(char *to, struct task_struct *tsk);
2384
2385#ifdef CONFIG_SMP
2386void scheduler_ipi(void);
2387extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2388#else
2389static inline void scheduler_ipi(void) { }
2390static inline unsigned long wait_task_inactive(struct task_struct *p,
2391 long match_state)
2392{
2393 return 1;
2394}
2395#endif
2396
2397#define next_task(p) \
2398 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2399
2400#define for_each_process(p) \
2401 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2402
2403extern bool current_is_single_threaded(void);
2404
2405/*
2406 * Careful: do_each_thread/while_each_thread is a double loop so
2407 * 'break' will not work as expected - use goto instead.
2408 */
2409#define do_each_thread(g, t) \
2410 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
2411
2412#define while_each_thread(g, t) \
2413 while ((t = next_thread(t)) != g)
2414
2415#define __for_each_thread(signal, t) \
2416 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
2417
2418#define for_each_thread(p, t) \
2419 __for_each_thread((p)->signal, t)
2420
2421/* Careful: this is a double loop, 'break' won't work as expected. */
2422#define for_each_process_thread(p, t) \
2423 for_each_process(p) for_each_thread(p, t)
2424
2425static inline int get_nr_threads(struct task_struct *tsk)
2426{
2427 return tsk->signal->nr_threads;
2428}
2429
2430static inline bool thread_group_leader(struct task_struct *p)
2431{
2432 return p->exit_signal >= 0;
2433}
2434
2435/* Do to the insanities of de_thread it is possible for a process
2436 * to have the pid of the thread group leader without actually being
2437 * the thread group leader. For iteration through the pids in proc
2438 * all we care about is that we have a task with the appropriate
2439 * pid, we don't actually care if we have the right task.
2440 */
2441static inline bool has_group_leader_pid(struct task_struct *p)
2442{
2443 return task_pid(p) == p->signal->leader_pid;
2444}
2445
2446static inline
2447bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
2448{
2449 return p1->signal == p2->signal;
2450}
2451
2452static inline struct task_struct *next_thread(const struct task_struct *p)
2453{
2454 return list_entry_rcu(p->thread_group.next,
2455 struct task_struct, thread_group);
2456}
2457
2458static inline int thread_group_empty(struct task_struct *p)
2459{
2460 return list_empty(&p->thread_group);
2461}
2462
2463#define delay_group_leader(p) \
2464 (thread_group_leader(p) && !thread_group_empty(p))
2465
2466/*
2467 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
2468 * subscriptions and synchronises with wait4(). Also used in procfs. Also
2469 * pins the final release of task.io_context. Also protects ->cpuset and
2470 * ->cgroup.subsys[]. And ->vfork_done.
2471 *
2472 * Nests both inside and outside of read_lock(&tasklist_lock).
2473 * It must not be nested with write_lock_irq(&tasklist_lock),
2474 * neither inside nor outside.
2475 */
2476static inline void task_lock(struct task_struct *p)
2477{
2478 spin_lock(&p->alloc_lock);
2479}
2480
2481static inline void task_unlock(struct task_struct *p)
2482{
2483 spin_unlock(&p->alloc_lock);
2484}
2485
2486extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
2487 unsigned long *flags);
2488
2489static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
2490 unsigned long *flags)
2491{
2492 struct sighand_struct *ret;
2493
2494 ret = __lock_task_sighand(tsk, flags);
2495 (void)__cond_lock(&tsk->sighand->siglock, ret);
2496 return ret;
2497}
2498
2499static inline void unlock_task_sighand(struct task_struct *tsk,
2500 unsigned long *flags)
2501{
2502 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
2503}
2504
2505#ifdef CONFIG_CGROUPS
2506static inline void threadgroup_change_begin(struct task_struct *tsk)
2507{
2508 down_read(&tsk->signal->group_rwsem);
2509}
2510static inline void threadgroup_change_end(struct task_struct *tsk)
2511{
2512 up_read(&tsk->signal->group_rwsem);
2513}
2514
2515/**
2516 * threadgroup_lock - lock threadgroup
2517 * @tsk: member task of the threadgroup to lock
2518 *
2519 * Lock the threadgroup @tsk belongs to. No new task is allowed to enter
2520 * and member tasks aren't allowed to exit (as indicated by PF_EXITING) or
2521 * change ->group_leader/pid. This is useful for cases where the threadgroup
2522 * needs to stay stable across blockable operations.
2523 *
2524 * fork and exit paths explicitly call threadgroup_change_{begin|end}() for
2525 * synchronization. While held, no new task will be added to threadgroup
2526 * and no existing live task will have its PF_EXITING set.
2527 *
2528 * de_thread() does threadgroup_change_{begin|end}() when a non-leader
2529 * sub-thread becomes a new leader.
2530 */
2531static inline void threadgroup_lock(struct task_struct *tsk)
2532{
2533 down_write(&tsk->signal->group_rwsem);
2534}
2535
2536/**
2537 * threadgroup_unlock - unlock threadgroup
2538 * @tsk: member task of the threadgroup to unlock
2539 *
2540 * Reverse threadgroup_lock().
2541 */
2542static inline void threadgroup_unlock(struct task_struct *tsk)
2543{
2544 up_write(&tsk->signal->group_rwsem);
2545}
2546#else
2547static inline void threadgroup_change_begin(struct task_struct *tsk) {}
2548static inline void threadgroup_change_end(struct task_struct *tsk) {}
2549static inline void threadgroup_lock(struct task_struct *tsk) {}
2550static inline void threadgroup_unlock(struct task_struct *tsk) {}
2551#endif
2552
2553#ifndef __HAVE_THREAD_FUNCTIONS
2554
2555#define task_thread_info(task) ((struct thread_info *)(task)->stack)
2556#define task_stack_page(task) ((task)->stack)
2557
2558static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
2559{
2560 *task_thread_info(p) = *task_thread_info(org);
2561 task_thread_info(p)->task = p;
2562}
2563
2564static inline unsigned long *end_of_stack(struct task_struct *p)
2565{
2566 return (unsigned long *)(task_thread_info(p) + 1);
2567}
2568
2569#endif
2570
2571static inline int object_is_on_stack(void *obj)
2572{
2573 void *stack = task_stack_page(current);
2574
2575 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
2576}
2577
2578extern void thread_info_cache_init(void);
2579
2580#ifdef CONFIG_DEBUG_STACK_USAGE
2581static inline unsigned long stack_not_used(struct task_struct *p)
2582{
2583 unsigned long *n = end_of_stack(p);
2584
2585 do { /* Skip over canary */
2586 n++;
2587 } while (!*n);
2588
2589 return (unsigned long)n - (unsigned long)end_of_stack(p);
2590}
2591#endif
2592
2593/* set thread flags in other task's structures
2594 * - see asm/thread_info.h for TIF_xxxx flags available
2595 */
2596static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2597{
2598 set_ti_thread_flag(task_thread_info(tsk), flag);
2599}
2600
2601static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2602{
2603 clear_ti_thread_flag(task_thread_info(tsk), flag);
2604}
2605
2606static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2607{
2608 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2609}
2610
2611static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2612{
2613 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2614}
2615
2616static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2617{
2618 return test_ti_thread_flag(task_thread_info(tsk), flag);
2619}
2620
2621static inline void set_tsk_need_resched(struct task_struct *tsk)
2622{
2623 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2624}
2625
2626static inline void clear_tsk_need_resched(struct task_struct *tsk)
2627{
2628 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2629}
2630
2631static inline int test_tsk_need_resched(struct task_struct *tsk)
2632{
2633 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2634}
2635
2636static inline int restart_syscall(void)
2637{
2638 set_tsk_thread_flag(current, TIF_SIGPENDING);
2639 return -ERESTARTNOINTR;
2640}
2641
2642static inline int signal_pending(struct task_struct *p)
2643{
2644 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
2645}
2646
2647static inline int __fatal_signal_pending(struct task_struct *p)
2648{
2649 return unlikely(sigismember(&p->pending.signal, SIGKILL));
2650}
2651
2652static inline int fatal_signal_pending(struct task_struct *p)
2653{
2654 return signal_pending(p) && __fatal_signal_pending(p);
2655}
2656
2657static inline int signal_pending_state(long state, struct task_struct *p)
2658{
2659 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
2660 return 0;
2661 if (!signal_pending(p))
2662 return 0;
2663
2664 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
2665}
2666
2667/*
2668 * cond_resched() and cond_resched_lock(): latency reduction via
2669 * explicit rescheduling in places that are safe. The return
2670 * value indicates whether a reschedule was done in fact.
2671 * cond_resched_lock() will drop the spinlock before scheduling,
2672 * cond_resched_softirq() will enable bhs before scheduling.
2673 */
2674extern int _cond_resched(void);
2675
2676#define cond_resched() ({ \
2677 __might_sleep(__FILE__, __LINE__, 0); \
2678 _cond_resched(); \
2679})
2680
2681extern int __cond_resched_lock(spinlock_t *lock);
2682
2683#ifdef CONFIG_PREEMPT_COUNT
2684#define PREEMPT_LOCK_OFFSET PREEMPT_OFFSET
2685#else
2686#define PREEMPT_LOCK_OFFSET 0
2687#endif
2688
2689#define cond_resched_lock(lock) ({ \
2690 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2691 __cond_resched_lock(lock); \
2692})
2693
2694extern int __cond_resched_softirq(void);
2695
2696#define cond_resched_softirq() ({ \
2697 __might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
2698 __cond_resched_softirq(); \
2699})
2700
2701static inline void cond_resched_rcu(void)
2702{
2703#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2704 rcu_read_unlock();
2705 cond_resched();
2706 rcu_read_lock();
2707#endif
2708}
2709
2710/*
2711 * Does a critical section need to be broken due to another
2712 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
2713 * but a general need for low latency)
2714 */
2715static inline int spin_needbreak(spinlock_t *lock)
2716{
2717#ifdef CONFIG_PREEMPT
2718 return spin_is_contended(lock);
2719#else
2720 return 0;
2721#endif
2722}
2723
2724/*
2725 * Idle thread specific functions to determine the need_resched
2726 * polling state. We have two versions, one based on TS_POLLING in
2727 * thread_info.status and one based on TIF_POLLING_NRFLAG in
2728 * thread_info.flags
2729 */
2730#ifdef TS_POLLING
2731static inline int tsk_is_polling(struct task_struct *p)
2732{
2733 return task_thread_info(p)->status & TS_POLLING;
2734}
2735static inline void __current_set_polling(void)
2736{
2737 current_thread_info()->status |= TS_POLLING;
2738}
2739
2740static inline bool __must_check current_set_polling_and_test(void)
2741{
2742 __current_set_polling();
2743
2744 /*
2745 * Polling state must be visible before we test NEED_RESCHED,
2746 * paired by resched_task()
2747 */
2748 smp_mb();
2749
2750 return unlikely(tif_need_resched());
2751}
2752
2753static inline void __current_clr_polling(void)
2754{
2755 current_thread_info()->status &= ~TS_POLLING;
2756}
2757
2758static inline bool __must_check current_clr_polling_and_test(void)
2759{
2760 __current_clr_polling();
2761
2762 /*
2763 * Polling state must be visible before we test NEED_RESCHED,
2764 * paired by resched_task()
2765 */
2766 smp_mb();
2767
2768 return unlikely(tif_need_resched());
2769}
2770#elif defined(TIF_POLLING_NRFLAG)
2771static inline int tsk_is_polling(struct task_struct *p)
2772{
2773 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
2774}
2775
2776static inline void __current_set_polling(void)
2777{
2778 set_thread_flag(TIF_POLLING_NRFLAG);
2779}
2780
2781static inline bool __must_check current_set_polling_and_test(void)
2782{
2783 __current_set_polling();
2784
2785 /*
2786 * Polling state must be visible before we test NEED_RESCHED,
2787 * paired by resched_task()
2788 *
2789 * XXX: assumes set/clear bit are identical barrier wise.
2790 */
2791 smp_mb__after_clear_bit();
2792
2793 return unlikely(tif_need_resched());
2794}
2795
2796static inline void __current_clr_polling(void)
2797{
2798 clear_thread_flag(TIF_POLLING_NRFLAG);
2799}
2800
2801static inline bool __must_check current_clr_polling_and_test(void)
2802{
2803 __current_clr_polling();
2804
2805 /*
2806 * Polling state must be visible before we test NEED_RESCHED,
2807 * paired by resched_task()
2808 */
2809 smp_mb__after_clear_bit();
2810
2811 return unlikely(tif_need_resched());
2812}
2813
2814#else
2815static inline int tsk_is_polling(struct task_struct *p) { return 0; }
2816static inline void __current_set_polling(void) { }
2817static inline void __current_clr_polling(void) { }
2818
2819static inline bool __must_check current_set_polling_and_test(void)
2820{
2821 return unlikely(tif_need_resched());
2822}
2823static inline bool __must_check current_clr_polling_and_test(void)
2824{
2825 return unlikely(tif_need_resched());
2826}
2827#endif
2828
2829static inline void current_clr_polling(void)
2830{
2831 __current_clr_polling();
2832
2833 /*
2834 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
2835 * Once the bit is cleared, we'll get IPIs with every new
2836 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
2837 * fold.
2838 */
2839 smp_mb(); /* paired with resched_task() */
2840
2841 preempt_fold_need_resched();
2842}
2843
2844static __always_inline bool need_resched(void)
2845{
2846 return unlikely(tif_need_resched());
2847}
2848
2849/*
2850 * Thread group CPU time accounting.
2851 */
2852void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
2853void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
2854
2855static inline void thread_group_cputime_init(struct signal_struct *sig)
2856{
2857 raw_spin_lock_init(&sig->cputimer.lock);
2858}
2859
2860/*
2861 * Reevaluate whether the task has signals pending delivery.
2862 * Wake the task if so.
2863 * This is required every time the blocked sigset_t changes.
2864 * callers must hold sighand->siglock.
2865 */
2866extern void recalc_sigpending_and_wake(struct task_struct *t);
2867extern void recalc_sigpending(void);
2868
2869extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
2870
2871static inline void signal_wake_up(struct task_struct *t, bool resume)
2872{
2873 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
2874}
2875static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
2876{
2877 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
2878}
2879
2880/*
2881 * Wrappers for p->thread_info->cpu access. No-op on UP.
2882 */
2883#ifdef CONFIG_SMP
2884
2885static inline unsigned int task_cpu(const struct task_struct *p)
2886{
2887 return task_thread_info(p)->cpu;
2888}
2889
2890static inline int task_node(const struct task_struct *p)
2891{
2892 return cpu_to_node(task_cpu(p));
2893}
2894
2895extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2896
2897#else
2898
2899static inline unsigned int task_cpu(const struct task_struct *p)
2900{
2901 return 0;
2902}
2903
2904static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2905{
2906}
2907
2908#endif /* CONFIG_SMP */
2909
2910extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2911extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2912
2913#ifdef CONFIG_CGROUP_SCHED
2914extern struct task_group root_task_group;
2915#endif /* CONFIG_CGROUP_SCHED */
2916
2917extern int task_can_switch_user(struct user_struct *up,
2918 struct task_struct *tsk);
2919
2920#ifdef CONFIG_TASK_XACCT
2921static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
2922{
2923 tsk->ioac.rchar += amt;
2924}
2925
2926static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
2927{
2928 tsk->ioac.wchar += amt;
2929}
2930
2931static inline void inc_syscr(struct task_struct *tsk)
2932{
2933 tsk->ioac.syscr++;
2934}
2935
2936static inline void inc_syscw(struct task_struct *tsk)
2937{
2938 tsk->ioac.syscw++;
2939}
2940#else
2941static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
2942{
2943}
2944
2945static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
2946{
2947}
2948
2949static inline void inc_syscr(struct task_struct *tsk)
2950{
2951}
2952
2953static inline void inc_syscw(struct task_struct *tsk)
2954{
2955}
2956#endif
2957
2958#ifndef TASK_SIZE_OF
2959#define TASK_SIZE_OF(tsk) TASK_SIZE
2960#endif
2961
2962#ifdef CONFIG_MM_OWNER
2963extern void mm_update_next_owner(struct mm_struct *mm);
2964extern void mm_init_owner(struct mm_struct *mm, struct task_struct *p);
2965#else
2966static inline void mm_update_next_owner(struct mm_struct *mm)
2967{
2968}
2969
2970static inline void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
2971{
2972}
2973#endif /* CONFIG_MM_OWNER */
2974
2975static inline unsigned long task_rlimit(const struct task_struct *tsk,
2976 unsigned int limit)
2977{
2978 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_cur);
2979}
2980
2981static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
2982 unsigned int limit)
2983{
2984 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_max);
2985}
2986
2987static inline unsigned long rlimit(unsigned int limit)
2988{
2989 return task_rlimit(current, limit);
2990}
2991
2992static inline unsigned long rlimit_max(unsigned int limit)
2993{
2994 return task_rlimit_max(current, limit);
2995}
2996
2997#endif
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#include <asm/processor.h>
14#include <linux/thread_info.h>
15#include <linux/preempt.h>
16#include <linux/cpumask.h>
17
18#include <linux/cache.h>
19#include <linux/irqflags_types.h>
20#include <linux/smp_types.h>
21#include <linux/pid_types.h>
22#include <linux/sem_types.h>
23#include <linux/shm.h>
24#include <linux/kmsan_types.h>
25#include <linux/mutex_types.h>
26#include <linux/plist_types.h>
27#include <linux/hrtimer_types.h>
28#include <linux/timer_types.h>
29#include <linux/seccomp_types.h>
30#include <linux/nodemask_types.h>
31#include <linux/refcount_types.h>
32#include <linux/resource.h>
33#include <linux/latencytop.h>
34#include <linux/sched/prio.h>
35#include <linux/sched/types.h>
36#include <linux/signal_types.h>
37#include <linux/syscall_user_dispatch_types.h>
38#include <linux/mm_types_task.h>
39#include <linux/task_io_accounting.h>
40#include <linux/posix-timers_types.h>
41#include <linux/restart_block.h>
42#include <uapi/linux/rseq.h>
43#include <linux/seqlock_types.h>
44#include <linux/kcsan.h>
45#include <linux/rv.h>
46#include <linux/livepatch_sched.h>
47#include <linux/uidgid_types.h>
48#include <asm/kmap_size.h>
49
50/* task_struct member predeclarations (sorted alphabetically): */
51struct audit_context;
52struct bio_list;
53struct blk_plug;
54struct bpf_local_storage;
55struct bpf_run_ctx;
56struct capture_control;
57struct cfs_rq;
58struct fs_struct;
59struct futex_pi_state;
60struct io_context;
61struct io_uring_task;
62struct mempolicy;
63struct nameidata;
64struct nsproxy;
65struct perf_event_context;
66struct pid_namespace;
67struct pipe_inode_info;
68struct rcu_node;
69struct reclaim_state;
70struct robust_list_head;
71struct root_domain;
72struct rq;
73struct sched_attr;
74struct sched_dl_entity;
75struct seq_file;
76struct sighand_struct;
77struct signal_struct;
78struct task_delay_info;
79struct task_group;
80struct task_struct;
81struct user_event_mm;
82
83/*
84 * Task state bitmask. NOTE! These bits are also
85 * encoded in fs/proc/array.c: get_task_state().
86 *
87 * We have two separate sets of flags: task->__state
88 * is about runnability, while task->exit_state are
89 * about the task exiting. Confusing, but this way
90 * modifying one set can't modify the other one by
91 * mistake.
92 */
93
94/* Used in tsk->__state: */
95#define TASK_RUNNING 0x00000000
96#define TASK_INTERRUPTIBLE 0x00000001
97#define TASK_UNINTERRUPTIBLE 0x00000002
98#define __TASK_STOPPED 0x00000004
99#define __TASK_TRACED 0x00000008
100/* Used in tsk->exit_state: */
101#define EXIT_DEAD 0x00000010
102#define EXIT_ZOMBIE 0x00000020
103#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
104/* Used in tsk->__state again: */
105#define TASK_PARKED 0x00000040
106#define TASK_DEAD 0x00000080
107#define TASK_WAKEKILL 0x00000100
108#define TASK_WAKING 0x00000200
109#define TASK_NOLOAD 0x00000400
110#define TASK_NEW 0x00000800
111#define TASK_RTLOCK_WAIT 0x00001000
112#define TASK_FREEZABLE 0x00002000
113#define __TASK_FREEZABLE_UNSAFE (0x00004000 * IS_ENABLED(CONFIG_LOCKDEP))
114#define TASK_FROZEN 0x00008000
115#define TASK_STATE_MAX 0x00010000
116
117#define TASK_ANY (TASK_STATE_MAX-1)
118
119/*
120 * DO NOT ADD ANY NEW USERS !
121 */
122#define TASK_FREEZABLE_UNSAFE (TASK_FREEZABLE | __TASK_FREEZABLE_UNSAFE)
123
124/* Convenience macros for the sake of set_current_state: */
125#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
126#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
127#define TASK_TRACED __TASK_TRACED
128
129#define TASK_IDLE (TASK_UNINTERRUPTIBLE | TASK_NOLOAD)
130
131/* Convenience macros for the sake of wake_up(): */
132#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
133
134/* get_task_state(): */
135#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
136 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
137 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
138 TASK_PARKED)
139
140#define task_is_running(task) (READ_ONCE((task)->__state) == TASK_RUNNING)
141
142#define task_is_traced(task) ((READ_ONCE(task->jobctl) & JOBCTL_TRACED) != 0)
143#define task_is_stopped(task) ((READ_ONCE(task->jobctl) & JOBCTL_STOPPED) != 0)
144#define task_is_stopped_or_traced(task) ((READ_ONCE(task->jobctl) & (JOBCTL_STOPPED | JOBCTL_TRACED)) != 0)
145
146/*
147 * Special states are those that do not use the normal wait-loop pattern. See
148 * the comment with set_special_state().
149 */
150#define is_special_task_state(state) \
151 ((state) & (__TASK_STOPPED | __TASK_TRACED | TASK_PARKED | TASK_DEAD))
152
153#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
154# define debug_normal_state_change(state_value) \
155 do { \
156 WARN_ON_ONCE(is_special_task_state(state_value)); \
157 current->task_state_change = _THIS_IP_; \
158 } while (0)
159
160# define debug_special_state_change(state_value) \
161 do { \
162 WARN_ON_ONCE(!is_special_task_state(state_value)); \
163 current->task_state_change = _THIS_IP_; \
164 } while (0)
165
166# define debug_rtlock_wait_set_state() \
167 do { \
168 current->saved_state_change = current->task_state_change;\
169 current->task_state_change = _THIS_IP_; \
170 } while (0)
171
172# define debug_rtlock_wait_restore_state() \
173 do { \
174 current->task_state_change = current->saved_state_change;\
175 } while (0)
176
177#else
178# define debug_normal_state_change(cond) do { } while (0)
179# define debug_special_state_change(cond) do { } while (0)
180# define debug_rtlock_wait_set_state() do { } while (0)
181# define debug_rtlock_wait_restore_state() do { } while (0)
182#endif
183
184/*
185 * set_current_state() includes a barrier so that the write of current->__state
186 * is correctly serialised wrt the caller's subsequent test of whether to
187 * actually sleep:
188 *
189 * for (;;) {
190 * set_current_state(TASK_UNINTERRUPTIBLE);
191 * if (CONDITION)
192 * break;
193 *
194 * schedule();
195 * }
196 * __set_current_state(TASK_RUNNING);
197 *
198 * If the caller does not need such serialisation (because, for instance, the
199 * CONDITION test and condition change and wakeup are under the same lock) then
200 * use __set_current_state().
201 *
202 * The above is typically ordered against the wakeup, which does:
203 *
204 * CONDITION = 1;
205 * wake_up_state(p, TASK_UNINTERRUPTIBLE);
206 *
207 * where wake_up_state()/try_to_wake_up() executes a full memory barrier before
208 * accessing p->__state.
209 *
210 * Wakeup will do: if (@state & p->__state) p->__state = TASK_RUNNING, that is,
211 * once it observes the TASK_UNINTERRUPTIBLE store the waking CPU can issue a
212 * TASK_RUNNING store which can collide with __set_current_state(TASK_RUNNING).
213 *
214 * However, with slightly different timing the wakeup TASK_RUNNING store can
215 * also collide with the TASK_UNINTERRUPTIBLE store. Losing that store is not
216 * a problem either because that will result in one extra go around the loop
217 * and our @cond test will save the day.
218 *
219 * Also see the comments of try_to_wake_up().
220 */
221#define __set_current_state(state_value) \
222 do { \
223 debug_normal_state_change((state_value)); \
224 WRITE_ONCE(current->__state, (state_value)); \
225 } while (0)
226
227#define set_current_state(state_value) \
228 do { \
229 debug_normal_state_change((state_value)); \
230 smp_store_mb(current->__state, (state_value)); \
231 } while (0)
232
233/*
234 * set_special_state() should be used for those states when the blocking task
235 * can not use the regular condition based wait-loop. In that case we must
236 * serialize against wakeups such that any possible in-flight TASK_RUNNING
237 * stores will not collide with our state change.
238 */
239#define set_special_state(state_value) \
240 do { \
241 unsigned long flags; /* may shadow */ \
242 \
243 raw_spin_lock_irqsave(¤t->pi_lock, flags); \
244 debug_special_state_change((state_value)); \
245 WRITE_ONCE(current->__state, (state_value)); \
246 raw_spin_unlock_irqrestore(¤t->pi_lock, flags); \
247 } while (0)
248
249/*
250 * PREEMPT_RT specific variants for "sleeping" spin/rwlocks
251 *
252 * RT's spin/rwlock substitutions are state preserving. The state of the
253 * task when blocking on the lock is saved in task_struct::saved_state and
254 * restored after the lock has been acquired. These operations are
255 * serialized by task_struct::pi_lock against try_to_wake_up(). Any non RT
256 * lock related wakeups while the task is blocked on the lock are
257 * redirected to operate on task_struct::saved_state to ensure that these
258 * are not dropped. On restore task_struct::saved_state is set to
259 * TASK_RUNNING so any wakeup attempt redirected to saved_state will fail.
260 *
261 * The lock operation looks like this:
262 *
263 * current_save_and_set_rtlock_wait_state();
264 * for (;;) {
265 * if (try_lock())
266 * break;
267 * raw_spin_unlock_irq(&lock->wait_lock);
268 * schedule_rtlock();
269 * raw_spin_lock_irq(&lock->wait_lock);
270 * set_current_state(TASK_RTLOCK_WAIT);
271 * }
272 * current_restore_rtlock_saved_state();
273 */
274#define current_save_and_set_rtlock_wait_state() \
275 do { \
276 lockdep_assert_irqs_disabled(); \
277 raw_spin_lock(¤t->pi_lock); \
278 current->saved_state = current->__state; \
279 debug_rtlock_wait_set_state(); \
280 WRITE_ONCE(current->__state, TASK_RTLOCK_WAIT); \
281 raw_spin_unlock(¤t->pi_lock); \
282 } while (0);
283
284#define current_restore_rtlock_saved_state() \
285 do { \
286 lockdep_assert_irqs_disabled(); \
287 raw_spin_lock(¤t->pi_lock); \
288 debug_rtlock_wait_restore_state(); \
289 WRITE_ONCE(current->__state, current->saved_state); \
290 current->saved_state = TASK_RUNNING; \
291 raw_spin_unlock(¤t->pi_lock); \
292 } while (0);
293
294#define get_current_state() READ_ONCE(current->__state)
295
296/*
297 * Define the task command name length as enum, then it can be visible to
298 * BPF programs.
299 */
300enum {
301 TASK_COMM_LEN = 16,
302};
303
304extern void scheduler_tick(void);
305
306#define MAX_SCHEDULE_TIMEOUT LONG_MAX
307
308extern long schedule_timeout(long timeout);
309extern long schedule_timeout_interruptible(long timeout);
310extern long schedule_timeout_killable(long timeout);
311extern long schedule_timeout_uninterruptible(long timeout);
312extern long schedule_timeout_idle(long timeout);
313asmlinkage void schedule(void);
314extern void schedule_preempt_disabled(void);
315asmlinkage void preempt_schedule_irq(void);
316#ifdef CONFIG_PREEMPT_RT
317 extern void schedule_rtlock(void);
318#endif
319
320extern int __must_check io_schedule_prepare(void);
321extern void io_schedule_finish(int token);
322extern long io_schedule_timeout(long timeout);
323extern void io_schedule(void);
324
325/**
326 * struct prev_cputime - snapshot of system and user cputime
327 * @utime: time spent in user mode
328 * @stime: time spent in system mode
329 * @lock: protects the above two fields
330 *
331 * Stores previous user/system time values such that we can guarantee
332 * monotonicity.
333 */
334struct prev_cputime {
335#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
336 u64 utime;
337 u64 stime;
338 raw_spinlock_t lock;
339#endif
340};
341
342enum vtime_state {
343 /* Task is sleeping or running in a CPU with VTIME inactive: */
344 VTIME_INACTIVE = 0,
345 /* Task is idle */
346 VTIME_IDLE,
347 /* Task runs in kernelspace in a CPU with VTIME active: */
348 VTIME_SYS,
349 /* Task runs in userspace in a CPU with VTIME active: */
350 VTIME_USER,
351 /* Task runs as guests in a CPU with VTIME active: */
352 VTIME_GUEST,
353};
354
355struct vtime {
356 seqcount_t seqcount;
357 unsigned long long starttime;
358 enum vtime_state state;
359 unsigned int cpu;
360 u64 utime;
361 u64 stime;
362 u64 gtime;
363};
364
365/*
366 * Utilization clamp constraints.
367 * @UCLAMP_MIN: Minimum utilization
368 * @UCLAMP_MAX: Maximum utilization
369 * @UCLAMP_CNT: Utilization clamp constraints count
370 */
371enum uclamp_id {
372 UCLAMP_MIN = 0,
373 UCLAMP_MAX,
374 UCLAMP_CNT
375};
376
377#ifdef CONFIG_SMP
378extern struct root_domain def_root_domain;
379extern struct mutex sched_domains_mutex;
380#endif
381
382struct sched_param {
383 int sched_priority;
384};
385
386struct sched_info {
387#ifdef CONFIG_SCHED_INFO
388 /* Cumulative counters: */
389
390 /* # of times we have run on this CPU: */
391 unsigned long pcount;
392
393 /* Time spent waiting on a runqueue: */
394 unsigned long long run_delay;
395
396 /* Timestamps: */
397
398 /* When did we last run on a CPU? */
399 unsigned long long last_arrival;
400
401 /* When were we last queued to run? */
402 unsigned long long last_queued;
403
404#endif /* CONFIG_SCHED_INFO */
405};
406
407/*
408 * Integer metrics need fixed point arithmetic, e.g., sched/fair
409 * has a few: load, load_avg, util_avg, freq, and capacity.
410 *
411 * We define a basic fixed point arithmetic range, and then formalize
412 * all these metrics based on that basic range.
413 */
414# define SCHED_FIXEDPOINT_SHIFT 10
415# define SCHED_FIXEDPOINT_SCALE (1L << SCHED_FIXEDPOINT_SHIFT)
416
417/* Increase resolution of cpu_capacity calculations */
418# define SCHED_CAPACITY_SHIFT SCHED_FIXEDPOINT_SHIFT
419# define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
420
421struct load_weight {
422 unsigned long weight;
423 u32 inv_weight;
424};
425
426/*
427 * The load/runnable/util_avg accumulates an infinite geometric series
428 * (see __update_load_avg_cfs_rq() in kernel/sched/pelt.c).
429 *
430 * [load_avg definition]
431 *
432 * load_avg = runnable% * scale_load_down(load)
433 *
434 * [runnable_avg definition]
435 *
436 * runnable_avg = runnable% * SCHED_CAPACITY_SCALE
437 *
438 * [util_avg definition]
439 *
440 * util_avg = running% * SCHED_CAPACITY_SCALE
441 *
442 * where runnable% is the time ratio that a sched_entity is runnable and
443 * running% the time ratio that a sched_entity is running.
444 *
445 * For cfs_rq, they are the aggregated values of all runnable and blocked
446 * sched_entities.
447 *
448 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
449 * capacity scaling. The scaling is done through the rq_clock_pelt that is used
450 * for computing those signals (see update_rq_clock_pelt())
451 *
452 * N.B., the above ratios (runnable% and running%) themselves are in the
453 * range of [0, 1]. To do fixed point arithmetics, we therefore scale them
454 * to as large a range as necessary. This is for example reflected by
455 * util_avg's SCHED_CAPACITY_SCALE.
456 *
457 * [Overflow issue]
458 *
459 * The 64-bit load_sum can have 4353082796 (=2^64/47742/88761) entities
460 * with the highest load (=88761), always runnable on a single cfs_rq,
461 * and should not overflow as the number already hits PID_MAX_LIMIT.
462 *
463 * For all other cases (including 32-bit kernels), struct load_weight's
464 * weight will overflow first before we do, because:
465 *
466 * Max(load_avg) <= Max(load.weight)
467 *
468 * Then it is the load_weight's responsibility to consider overflow
469 * issues.
470 */
471struct sched_avg {
472 u64 last_update_time;
473 u64 load_sum;
474 u64 runnable_sum;
475 u32 util_sum;
476 u32 period_contrib;
477 unsigned long load_avg;
478 unsigned long runnable_avg;
479 unsigned long util_avg;
480 unsigned int util_est;
481} ____cacheline_aligned;
482
483/*
484 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
485 * updates. When a task is dequeued, its util_est should not be updated if its
486 * util_avg has not been updated in the meantime.
487 * This information is mapped into the MSB bit of util_est at dequeue time.
488 * Since max value of util_est for a task is 1024 (PELT util_avg for a task)
489 * it is safe to use MSB.
490 */
491#define UTIL_EST_WEIGHT_SHIFT 2
492#define UTIL_AVG_UNCHANGED 0x80000000
493
494struct sched_statistics {
495#ifdef CONFIG_SCHEDSTATS
496 u64 wait_start;
497 u64 wait_max;
498 u64 wait_count;
499 u64 wait_sum;
500 u64 iowait_count;
501 u64 iowait_sum;
502
503 u64 sleep_start;
504 u64 sleep_max;
505 s64 sum_sleep_runtime;
506
507 u64 block_start;
508 u64 block_max;
509 s64 sum_block_runtime;
510
511 s64 exec_max;
512 u64 slice_max;
513
514 u64 nr_migrations_cold;
515 u64 nr_failed_migrations_affine;
516 u64 nr_failed_migrations_running;
517 u64 nr_failed_migrations_hot;
518 u64 nr_forced_migrations;
519
520 u64 nr_wakeups;
521 u64 nr_wakeups_sync;
522 u64 nr_wakeups_migrate;
523 u64 nr_wakeups_local;
524 u64 nr_wakeups_remote;
525 u64 nr_wakeups_affine;
526 u64 nr_wakeups_affine_attempts;
527 u64 nr_wakeups_passive;
528 u64 nr_wakeups_idle;
529
530#ifdef CONFIG_SCHED_CORE
531 u64 core_forceidle_sum;
532#endif
533#endif /* CONFIG_SCHEDSTATS */
534} ____cacheline_aligned;
535
536struct sched_entity {
537 /* For load-balancing: */
538 struct load_weight load;
539 struct rb_node run_node;
540 u64 deadline;
541 u64 min_vruntime;
542
543 struct list_head group_node;
544 unsigned int on_rq;
545
546 u64 exec_start;
547 u64 sum_exec_runtime;
548 u64 prev_sum_exec_runtime;
549 u64 vruntime;
550 s64 vlag;
551 u64 slice;
552
553 u64 nr_migrations;
554
555#ifdef CONFIG_FAIR_GROUP_SCHED
556 int depth;
557 struct sched_entity *parent;
558 /* rq on which this entity is (to be) queued: */
559 struct cfs_rq *cfs_rq;
560 /* rq "owned" by this entity/group: */
561 struct cfs_rq *my_q;
562 /* cached value of my_q->h_nr_running */
563 unsigned long runnable_weight;
564#endif
565
566#ifdef CONFIG_SMP
567 /*
568 * Per entity load average tracking.
569 *
570 * Put into separate cache line so it does not
571 * collide with read-mostly values above.
572 */
573 struct sched_avg avg;
574#endif
575};
576
577struct sched_rt_entity {
578 struct list_head run_list;
579 unsigned long timeout;
580 unsigned long watchdog_stamp;
581 unsigned int time_slice;
582 unsigned short on_rq;
583 unsigned short on_list;
584
585 struct sched_rt_entity *back;
586#ifdef CONFIG_RT_GROUP_SCHED
587 struct sched_rt_entity *parent;
588 /* rq on which this entity is (to be) queued: */
589 struct rt_rq *rt_rq;
590 /* rq "owned" by this entity/group: */
591 struct rt_rq *my_q;
592#endif
593} __randomize_layout;
594
595typedef bool (*dl_server_has_tasks_f)(struct sched_dl_entity *);
596typedef struct task_struct *(*dl_server_pick_f)(struct sched_dl_entity *);
597
598struct sched_dl_entity {
599 struct rb_node rb_node;
600
601 /*
602 * Original scheduling parameters. Copied here from sched_attr
603 * during sched_setattr(), they will remain the same until
604 * the next sched_setattr().
605 */
606 u64 dl_runtime; /* Maximum runtime for each instance */
607 u64 dl_deadline; /* Relative deadline of each instance */
608 u64 dl_period; /* Separation of two instances (period) */
609 u64 dl_bw; /* dl_runtime / dl_period */
610 u64 dl_density; /* dl_runtime / dl_deadline */
611
612 /*
613 * Actual scheduling parameters. Initialized with the values above,
614 * they are continuously updated during task execution. Note that
615 * the remaining runtime could be < 0 in case we are in overrun.
616 */
617 s64 runtime; /* Remaining runtime for this instance */
618 u64 deadline; /* Absolute deadline for this instance */
619 unsigned int flags; /* Specifying the scheduler behaviour */
620
621 /*
622 * Some bool flags:
623 *
624 * @dl_throttled tells if we exhausted the runtime. If so, the
625 * task has to wait for a replenishment to be performed at the
626 * next firing of dl_timer.
627 *
628 * @dl_yielded tells if task gave up the CPU before consuming
629 * all its available runtime during the last job.
630 *
631 * @dl_non_contending tells if the task is inactive while still
632 * contributing to the active utilization. In other words, it
633 * indicates if the inactive timer has been armed and its handler
634 * has not been executed yet. This flag is useful to avoid race
635 * conditions between the inactive timer handler and the wakeup
636 * code.
637 *
638 * @dl_overrun tells if the task asked to be informed about runtime
639 * overruns.
640 */
641 unsigned int dl_throttled : 1;
642 unsigned int dl_yielded : 1;
643 unsigned int dl_non_contending : 1;
644 unsigned int dl_overrun : 1;
645 unsigned int dl_server : 1;
646
647 /*
648 * Bandwidth enforcement timer. Each -deadline task has its
649 * own bandwidth to be enforced, thus we need one timer per task.
650 */
651 struct hrtimer dl_timer;
652
653 /*
654 * Inactive timer, responsible for decreasing the active utilization
655 * at the "0-lag time". When a -deadline task blocks, it contributes
656 * to GRUB's active utilization until the "0-lag time", hence a
657 * timer is needed to decrease the active utilization at the correct
658 * time.
659 */
660 struct hrtimer inactive_timer;
661
662 /*
663 * Bits for DL-server functionality. Also see the comment near
664 * dl_server_update().
665 *
666 * @rq the runqueue this server is for
667 *
668 * @server_has_tasks() returns true if @server_pick return a
669 * runnable task.
670 */
671 struct rq *rq;
672 dl_server_has_tasks_f server_has_tasks;
673 dl_server_pick_f server_pick;
674
675#ifdef CONFIG_RT_MUTEXES
676 /*
677 * Priority Inheritance. When a DEADLINE scheduling entity is boosted
678 * pi_se points to the donor, otherwise points to the dl_se it belongs
679 * to (the original one/itself).
680 */
681 struct sched_dl_entity *pi_se;
682#endif
683};
684
685#ifdef CONFIG_UCLAMP_TASK
686/* Number of utilization clamp buckets (shorter alias) */
687#define UCLAMP_BUCKETS CONFIG_UCLAMP_BUCKETS_COUNT
688
689/*
690 * Utilization clamp for a scheduling entity
691 * @value: clamp value "assigned" to a se
692 * @bucket_id: bucket index corresponding to the "assigned" value
693 * @active: the se is currently refcounted in a rq's bucket
694 * @user_defined: the requested clamp value comes from user-space
695 *
696 * The bucket_id is the index of the clamp bucket matching the clamp value
697 * which is pre-computed and stored to avoid expensive integer divisions from
698 * the fast path.
699 *
700 * The active bit is set whenever a task has got an "effective" value assigned,
701 * which can be different from the clamp value "requested" from user-space.
702 * This allows to know a task is refcounted in the rq's bucket corresponding
703 * to the "effective" bucket_id.
704 *
705 * The user_defined bit is set whenever a task has got a task-specific clamp
706 * value requested from userspace, i.e. the system defaults apply to this task
707 * just as a restriction. This allows to relax default clamps when a less
708 * restrictive task-specific value has been requested, thus allowing to
709 * implement a "nice" semantic. For example, a task running with a 20%
710 * default boost can still drop its own boosting to 0%.
711 */
712struct uclamp_se {
713 unsigned int value : bits_per(SCHED_CAPACITY_SCALE);
714 unsigned int bucket_id : bits_per(UCLAMP_BUCKETS);
715 unsigned int active : 1;
716 unsigned int user_defined : 1;
717};
718#endif /* CONFIG_UCLAMP_TASK */
719
720union rcu_special {
721 struct {
722 u8 blocked;
723 u8 need_qs;
724 u8 exp_hint; /* Hint for performance. */
725 u8 need_mb; /* Readers need smp_mb(). */
726 } b; /* Bits. */
727 u32 s; /* Set of bits. */
728};
729
730enum perf_event_task_context {
731 perf_invalid_context = -1,
732 perf_hw_context = 0,
733 perf_sw_context,
734 perf_nr_task_contexts,
735};
736
737struct wake_q_node {
738 struct wake_q_node *next;
739};
740
741struct kmap_ctrl {
742#ifdef CONFIG_KMAP_LOCAL
743 int idx;
744 pte_t pteval[KM_MAX_IDX];
745#endif
746};
747
748struct task_struct {
749#ifdef CONFIG_THREAD_INFO_IN_TASK
750 /*
751 * For reasons of header soup (see current_thread_info()), this
752 * must be the first element of task_struct.
753 */
754 struct thread_info thread_info;
755#endif
756 unsigned int __state;
757
758 /* saved state for "spinlock sleepers" */
759 unsigned int saved_state;
760
761 /*
762 * This begins the randomizable portion of task_struct. Only
763 * scheduling-critical items should be added above here.
764 */
765 randomized_struct_fields_start
766
767 void *stack;
768 refcount_t usage;
769 /* Per task flags (PF_*), defined further below: */
770 unsigned int flags;
771 unsigned int ptrace;
772
773#ifdef CONFIG_SMP
774 int on_cpu;
775 struct __call_single_node wake_entry;
776 unsigned int wakee_flips;
777 unsigned long wakee_flip_decay_ts;
778 struct task_struct *last_wakee;
779
780 /*
781 * recent_used_cpu is initially set as the last CPU used by a task
782 * that wakes affine another task. Waker/wakee relationships can
783 * push tasks around a CPU where each wakeup moves to the next one.
784 * Tracking a recently used CPU allows a quick search for a recently
785 * used CPU that may be idle.
786 */
787 int recent_used_cpu;
788 int wake_cpu;
789#endif
790 int on_rq;
791
792 int prio;
793 int static_prio;
794 int normal_prio;
795 unsigned int rt_priority;
796
797 struct sched_entity se;
798 struct sched_rt_entity rt;
799 struct sched_dl_entity dl;
800 struct sched_dl_entity *dl_server;
801 const struct sched_class *sched_class;
802
803#ifdef CONFIG_SCHED_CORE
804 struct rb_node core_node;
805 unsigned long core_cookie;
806 unsigned int core_occupation;
807#endif
808
809#ifdef CONFIG_CGROUP_SCHED
810 struct task_group *sched_task_group;
811#endif
812
813#ifdef CONFIG_UCLAMP_TASK
814 /*
815 * Clamp values requested for a scheduling entity.
816 * Must be updated with task_rq_lock() held.
817 */
818 struct uclamp_se uclamp_req[UCLAMP_CNT];
819 /*
820 * Effective clamp values used for a scheduling entity.
821 * Must be updated with task_rq_lock() held.
822 */
823 struct uclamp_se uclamp[UCLAMP_CNT];
824#endif
825
826 struct sched_statistics stats;
827
828#ifdef CONFIG_PREEMPT_NOTIFIERS
829 /* List of struct preempt_notifier: */
830 struct hlist_head preempt_notifiers;
831#endif
832
833#ifdef CONFIG_BLK_DEV_IO_TRACE
834 unsigned int btrace_seq;
835#endif
836
837 unsigned int policy;
838 int nr_cpus_allowed;
839 const cpumask_t *cpus_ptr;
840 cpumask_t *user_cpus_ptr;
841 cpumask_t cpus_mask;
842 void *migration_pending;
843#ifdef CONFIG_SMP
844 unsigned short migration_disabled;
845#endif
846 unsigned short migration_flags;
847
848#ifdef CONFIG_PREEMPT_RCU
849 int rcu_read_lock_nesting;
850 union rcu_special rcu_read_unlock_special;
851 struct list_head rcu_node_entry;
852 struct rcu_node *rcu_blocked_node;
853#endif /* #ifdef CONFIG_PREEMPT_RCU */
854
855#ifdef CONFIG_TASKS_RCU
856 unsigned long rcu_tasks_nvcsw;
857 u8 rcu_tasks_holdout;
858 u8 rcu_tasks_idx;
859 int rcu_tasks_idle_cpu;
860 struct list_head rcu_tasks_holdout_list;
861#endif /* #ifdef CONFIG_TASKS_RCU */
862
863#ifdef CONFIG_TASKS_TRACE_RCU
864 int trc_reader_nesting;
865 int trc_ipi_to_cpu;
866 union rcu_special trc_reader_special;
867 struct list_head trc_holdout_list;
868 struct list_head trc_blkd_node;
869 int trc_blkd_cpu;
870#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
871
872 struct sched_info sched_info;
873
874 struct list_head tasks;
875#ifdef CONFIG_SMP
876 struct plist_node pushable_tasks;
877 struct rb_node pushable_dl_tasks;
878#endif
879
880 struct mm_struct *mm;
881 struct mm_struct *active_mm;
882 struct address_space *faults_disabled_mapping;
883
884 int exit_state;
885 int exit_code;
886 int exit_signal;
887 /* The signal sent when the parent dies: */
888 int pdeath_signal;
889 /* JOBCTL_*, siglock protected: */
890 unsigned long jobctl;
891
892 /* Used for emulating ABI behavior of previous Linux versions: */
893 unsigned int personality;
894
895 /* Scheduler bits, serialized by scheduler locks: */
896 unsigned sched_reset_on_fork:1;
897 unsigned sched_contributes_to_load:1;
898 unsigned sched_migrated:1;
899
900 /* Force alignment to the next boundary: */
901 unsigned :0;
902
903 /* Unserialized, strictly 'current' */
904
905 /*
906 * This field must not be in the scheduler word above due to wakelist
907 * queueing no longer being serialized by p->on_cpu. However:
908 *
909 * p->XXX = X; ttwu()
910 * schedule() if (p->on_rq && ..) // false
911 * smp_mb__after_spinlock(); if (smp_load_acquire(&p->on_cpu) && //true
912 * deactivate_task() ttwu_queue_wakelist())
913 * p->on_rq = 0; p->sched_remote_wakeup = Y;
914 *
915 * guarantees all stores of 'current' are visible before
916 * ->sched_remote_wakeup gets used, so it can be in this word.
917 */
918 unsigned sched_remote_wakeup:1;
919#ifdef CONFIG_RT_MUTEXES
920 unsigned sched_rt_mutex:1;
921#endif
922
923 /* Bit to tell TOMOYO we're in execve(): */
924 unsigned in_execve:1;
925 unsigned in_iowait:1;
926#ifndef TIF_RESTORE_SIGMASK
927 unsigned restore_sigmask:1;
928#endif
929#ifdef CONFIG_MEMCG
930 unsigned in_user_fault:1;
931#endif
932#ifdef CONFIG_LRU_GEN
933 /* whether the LRU algorithm may apply to this access */
934 unsigned in_lru_fault:1;
935#endif
936#ifdef CONFIG_COMPAT_BRK
937 unsigned brk_randomized:1;
938#endif
939#ifdef CONFIG_CGROUPS
940 /* disallow userland-initiated cgroup migration */
941 unsigned no_cgroup_migration:1;
942 /* task is frozen/stopped (used by the cgroup freezer) */
943 unsigned frozen:1;
944#endif
945#ifdef CONFIG_BLK_CGROUP
946 unsigned use_memdelay:1;
947#endif
948#ifdef CONFIG_PSI
949 /* Stalled due to lack of memory */
950 unsigned in_memstall:1;
951#endif
952#ifdef CONFIG_PAGE_OWNER
953 /* Used by page_owner=on to detect recursion in page tracking. */
954 unsigned in_page_owner:1;
955#endif
956#ifdef CONFIG_EVENTFD
957 /* Recursion prevention for eventfd_signal() */
958 unsigned in_eventfd:1;
959#endif
960#ifdef CONFIG_ARCH_HAS_CPU_PASID
961 unsigned pasid_activated:1;
962#endif
963#ifdef CONFIG_CPU_SUP_INTEL
964 unsigned reported_split_lock:1;
965#endif
966#ifdef CONFIG_TASK_DELAY_ACCT
967 /* delay due to memory thrashing */
968 unsigned in_thrashing:1;
969#endif
970
971 unsigned long atomic_flags; /* Flags requiring atomic access. */
972
973 struct restart_block restart_block;
974
975 pid_t pid;
976 pid_t tgid;
977
978#ifdef CONFIG_STACKPROTECTOR
979 /* Canary value for the -fstack-protector GCC feature: */
980 unsigned long stack_canary;
981#endif
982 /*
983 * Pointers to the (original) parent process, youngest child, younger sibling,
984 * older sibling, respectively. (p->father can be replaced with
985 * p->real_parent->pid)
986 */
987
988 /* Real parent process: */
989 struct task_struct __rcu *real_parent;
990
991 /* Recipient of SIGCHLD, wait4() reports: */
992 struct task_struct __rcu *parent;
993
994 /*
995 * Children/sibling form the list of natural children:
996 */
997 struct list_head children;
998 struct list_head sibling;
999 struct task_struct *group_leader;
1000
1001 /*
1002 * 'ptraced' is the list of tasks this task is using ptrace() on.
1003 *
1004 * This includes both natural children and PTRACE_ATTACH targets.
1005 * 'ptrace_entry' is this task's link on the p->parent->ptraced list.
1006 */
1007 struct list_head ptraced;
1008 struct list_head ptrace_entry;
1009
1010 /* PID/PID hash table linkage. */
1011 struct pid *thread_pid;
1012 struct hlist_node pid_links[PIDTYPE_MAX];
1013 struct list_head thread_node;
1014
1015 struct completion *vfork_done;
1016
1017 /* CLONE_CHILD_SETTID: */
1018 int __user *set_child_tid;
1019
1020 /* CLONE_CHILD_CLEARTID: */
1021 int __user *clear_child_tid;
1022
1023 /* PF_KTHREAD | PF_IO_WORKER */
1024 void *worker_private;
1025
1026 u64 utime;
1027 u64 stime;
1028#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1029 u64 utimescaled;
1030 u64 stimescaled;
1031#endif
1032 u64 gtime;
1033 struct prev_cputime prev_cputime;
1034#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1035 struct vtime vtime;
1036#endif
1037
1038#ifdef CONFIG_NO_HZ_FULL
1039 atomic_t tick_dep_mask;
1040#endif
1041 /* Context switch counts: */
1042 unsigned long nvcsw;
1043 unsigned long nivcsw;
1044
1045 /* Monotonic time in nsecs: */
1046 u64 start_time;
1047
1048 /* Boot based time in nsecs: */
1049 u64 start_boottime;
1050
1051 /* MM fault and swap info: this can arguably be seen as either mm-specific or thread-specific: */
1052 unsigned long min_flt;
1053 unsigned long maj_flt;
1054
1055 /* Empty if CONFIG_POSIX_CPUTIMERS=n */
1056 struct posix_cputimers posix_cputimers;
1057
1058#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
1059 struct posix_cputimers_work posix_cputimers_work;
1060#endif
1061
1062 /* Process credentials: */
1063
1064 /* Tracer's credentials at attach: */
1065 const struct cred __rcu *ptracer_cred;
1066
1067 /* Objective and real subjective task credentials (COW): */
1068 const struct cred __rcu *real_cred;
1069
1070 /* Effective (overridable) subjective task credentials (COW): */
1071 const struct cred __rcu *cred;
1072
1073#ifdef CONFIG_KEYS
1074 /* Cached requested key. */
1075 struct key *cached_requested_key;
1076#endif
1077
1078 /*
1079 * executable name, excluding path.
1080 *
1081 * - normally initialized setup_new_exec()
1082 * - access it with [gs]et_task_comm()
1083 * - lock it with task_lock()
1084 */
1085 char comm[TASK_COMM_LEN];
1086
1087 struct nameidata *nameidata;
1088
1089#ifdef CONFIG_SYSVIPC
1090 struct sysv_sem sysvsem;
1091 struct sysv_shm sysvshm;
1092#endif
1093#ifdef CONFIG_DETECT_HUNG_TASK
1094 unsigned long last_switch_count;
1095 unsigned long last_switch_time;
1096#endif
1097 /* Filesystem information: */
1098 struct fs_struct *fs;
1099
1100 /* Open file information: */
1101 struct files_struct *files;
1102
1103#ifdef CONFIG_IO_URING
1104 struct io_uring_task *io_uring;
1105#endif
1106
1107 /* Namespaces: */
1108 struct nsproxy *nsproxy;
1109
1110 /* Signal handlers: */
1111 struct signal_struct *signal;
1112 struct sighand_struct __rcu *sighand;
1113 sigset_t blocked;
1114 sigset_t real_blocked;
1115 /* Restored if set_restore_sigmask() was used: */
1116 sigset_t saved_sigmask;
1117 struct sigpending pending;
1118 unsigned long sas_ss_sp;
1119 size_t sas_ss_size;
1120 unsigned int sas_ss_flags;
1121
1122 struct callback_head *task_works;
1123
1124#ifdef CONFIG_AUDIT
1125#ifdef CONFIG_AUDITSYSCALL
1126 struct audit_context *audit_context;
1127#endif
1128 kuid_t loginuid;
1129 unsigned int sessionid;
1130#endif
1131 struct seccomp seccomp;
1132 struct syscall_user_dispatch syscall_dispatch;
1133
1134 /* Thread group tracking: */
1135 u64 parent_exec_id;
1136 u64 self_exec_id;
1137
1138 /* Protection against (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, mempolicy: */
1139 spinlock_t alloc_lock;
1140
1141 /* Protection of the PI data structures: */
1142 raw_spinlock_t pi_lock;
1143
1144 struct wake_q_node wake_q;
1145
1146#ifdef CONFIG_RT_MUTEXES
1147 /* PI waiters blocked on a rt_mutex held by this task: */
1148 struct rb_root_cached pi_waiters;
1149 /* Updated under owner's pi_lock and rq lock */
1150 struct task_struct *pi_top_task;
1151 /* Deadlock detection and priority inheritance handling: */
1152 struct rt_mutex_waiter *pi_blocked_on;
1153#endif
1154
1155#ifdef CONFIG_DEBUG_MUTEXES
1156 /* Mutex deadlock detection: */
1157 struct mutex_waiter *blocked_on;
1158#endif
1159
1160#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1161 int non_block_count;
1162#endif
1163
1164#ifdef CONFIG_TRACE_IRQFLAGS
1165 struct irqtrace_events irqtrace;
1166 unsigned int hardirq_threaded;
1167 u64 hardirq_chain_key;
1168 int softirqs_enabled;
1169 int softirq_context;
1170 int irq_config;
1171#endif
1172#ifdef CONFIG_PREEMPT_RT
1173 int softirq_disable_cnt;
1174#endif
1175
1176#ifdef CONFIG_LOCKDEP
1177# define MAX_LOCK_DEPTH 48UL
1178 u64 curr_chain_key;
1179 int lockdep_depth;
1180 unsigned int lockdep_recursion;
1181 struct held_lock held_locks[MAX_LOCK_DEPTH];
1182#endif
1183
1184#if defined(CONFIG_UBSAN) && !defined(CONFIG_UBSAN_TRAP)
1185 unsigned int in_ubsan;
1186#endif
1187
1188 /* Journalling filesystem info: */
1189 void *journal_info;
1190
1191 /* Stacked block device info: */
1192 struct bio_list *bio_list;
1193
1194 /* Stack plugging: */
1195 struct blk_plug *plug;
1196
1197 /* VM state: */
1198 struct reclaim_state *reclaim_state;
1199
1200 struct io_context *io_context;
1201
1202#ifdef CONFIG_COMPACTION
1203 struct capture_control *capture_control;
1204#endif
1205 /* Ptrace state: */
1206 unsigned long ptrace_message;
1207 kernel_siginfo_t *last_siginfo;
1208
1209 struct task_io_accounting ioac;
1210#ifdef CONFIG_PSI
1211 /* Pressure stall state */
1212 unsigned int psi_flags;
1213#endif
1214#ifdef CONFIG_TASK_XACCT
1215 /* Accumulated RSS usage: */
1216 u64 acct_rss_mem1;
1217 /* Accumulated virtual memory usage: */
1218 u64 acct_vm_mem1;
1219 /* stime + utime since last update: */
1220 u64 acct_timexpd;
1221#endif
1222#ifdef CONFIG_CPUSETS
1223 /* Protected by ->alloc_lock: */
1224 nodemask_t mems_allowed;
1225 /* Sequence number to catch updates: */
1226 seqcount_spinlock_t mems_allowed_seq;
1227 int cpuset_mem_spread_rotor;
1228 int cpuset_slab_spread_rotor;
1229#endif
1230#ifdef CONFIG_CGROUPS
1231 /* Control Group info protected by css_set_lock: */
1232 struct css_set __rcu *cgroups;
1233 /* cg_list protected by css_set_lock and tsk->alloc_lock: */
1234 struct list_head cg_list;
1235#endif
1236#ifdef CONFIG_X86_CPU_RESCTRL
1237 u32 closid;
1238 u32 rmid;
1239#endif
1240#ifdef CONFIG_FUTEX
1241 struct robust_list_head __user *robust_list;
1242#ifdef CONFIG_COMPAT
1243 struct compat_robust_list_head __user *compat_robust_list;
1244#endif
1245 struct list_head pi_state_list;
1246 struct futex_pi_state *pi_state_cache;
1247 struct mutex futex_exit_mutex;
1248 unsigned int futex_state;
1249#endif
1250#ifdef CONFIG_PERF_EVENTS
1251 struct perf_event_context *perf_event_ctxp;
1252 struct mutex perf_event_mutex;
1253 struct list_head perf_event_list;
1254#endif
1255#ifdef CONFIG_DEBUG_PREEMPT
1256 unsigned long preempt_disable_ip;
1257#endif
1258#ifdef CONFIG_NUMA
1259 /* Protected by alloc_lock: */
1260 struct mempolicy *mempolicy;
1261 short il_prev;
1262 short pref_node_fork;
1263#endif
1264#ifdef CONFIG_NUMA_BALANCING
1265 int numa_scan_seq;
1266 unsigned int numa_scan_period;
1267 unsigned int numa_scan_period_max;
1268 int numa_preferred_nid;
1269 unsigned long numa_migrate_retry;
1270 /* Migration stamp: */
1271 u64 node_stamp;
1272 u64 last_task_numa_placement;
1273 u64 last_sum_exec_runtime;
1274 struct callback_head numa_work;
1275
1276 /*
1277 * This pointer is only modified for current in syscall and
1278 * pagefault context (and for tasks being destroyed), so it can be read
1279 * from any of the following contexts:
1280 * - RCU read-side critical section
1281 * - current->numa_group from everywhere
1282 * - task's runqueue locked, task not running
1283 */
1284 struct numa_group __rcu *numa_group;
1285
1286 /*
1287 * numa_faults is an array split into four regions:
1288 * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
1289 * in this precise order.
1290 *
1291 * faults_memory: Exponential decaying average of faults on a per-node
1292 * basis. Scheduling placement decisions are made based on these
1293 * counts. The values remain static for the duration of a PTE scan.
1294 * faults_cpu: Track the nodes the process was running on when a NUMA
1295 * hinting fault was incurred.
1296 * faults_memory_buffer and faults_cpu_buffer: Record faults per node
1297 * during the current scan window. When the scan completes, the counts
1298 * in faults_memory and faults_cpu decay and these values are copied.
1299 */
1300 unsigned long *numa_faults;
1301 unsigned long total_numa_faults;
1302
1303 /*
1304 * numa_faults_locality tracks if faults recorded during the last
1305 * scan window were remote/local or failed to migrate. The task scan
1306 * period is adapted based on the locality of the faults with different
1307 * weights depending on whether they were shared or private faults
1308 */
1309 unsigned long numa_faults_locality[3];
1310
1311 unsigned long numa_pages_migrated;
1312#endif /* CONFIG_NUMA_BALANCING */
1313
1314#ifdef CONFIG_RSEQ
1315 struct rseq __user *rseq;
1316 u32 rseq_len;
1317 u32 rseq_sig;
1318 /*
1319 * RmW on rseq_event_mask must be performed atomically
1320 * with respect to preemption.
1321 */
1322 unsigned long rseq_event_mask;
1323#endif
1324
1325#ifdef CONFIG_SCHED_MM_CID
1326 int mm_cid; /* Current cid in mm */
1327 int last_mm_cid; /* Most recent cid in mm */
1328 int migrate_from_cpu;
1329 int mm_cid_active; /* Whether cid bitmap is active */
1330 struct callback_head cid_work;
1331#endif
1332
1333 struct tlbflush_unmap_batch tlb_ubc;
1334
1335 /* Cache last used pipe for splice(): */
1336 struct pipe_inode_info *splice_pipe;
1337
1338 struct page_frag task_frag;
1339
1340#ifdef CONFIG_TASK_DELAY_ACCT
1341 struct task_delay_info *delays;
1342#endif
1343
1344#ifdef CONFIG_FAULT_INJECTION
1345 int make_it_fail;
1346 unsigned int fail_nth;
1347#endif
1348 /*
1349 * When (nr_dirtied >= nr_dirtied_pause), it's time to call
1350 * balance_dirty_pages() for a dirty throttling pause:
1351 */
1352 int nr_dirtied;
1353 int nr_dirtied_pause;
1354 /* Start of a write-and-pause period: */
1355 unsigned long dirty_paused_when;
1356
1357#ifdef CONFIG_LATENCYTOP
1358 int latency_record_count;
1359 struct latency_record latency_record[LT_SAVECOUNT];
1360#endif
1361 /*
1362 * Time slack values; these are used to round up poll() and
1363 * select() etc timeout values. These are in nanoseconds.
1364 */
1365 u64 timer_slack_ns;
1366 u64 default_timer_slack_ns;
1367
1368#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
1369 unsigned int kasan_depth;
1370#endif
1371
1372#ifdef CONFIG_KCSAN
1373 struct kcsan_ctx kcsan_ctx;
1374#ifdef CONFIG_TRACE_IRQFLAGS
1375 struct irqtrace_events kcsan_save_irqtrace;
1376#endif
1377#ifdef CONFIG_KCSAN_WEAK_MEMORY
1378 int kcsan_stack_depth;
1379#endif
1380#endif
1381
1382#ifdef CONFIG_KMSAN
1383 struct kmsan_ctx kmsan_ctx;
1384#endif
1385
1386#if IS_ENABLED(CONFIG_KUNIT)
1387 struct kunit *kunit_test;
1388#endif
1389
1390#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1391 /* Index of current stored address in ret_stack: */
1392 int curr_ret_stack;
1393 int curr_ret_depth;
1394
1395 /* Stack of return addresses for return function tracing: */
1396 struct ftrace_ret_stack *ret_stack;
1397
1398 /* Timestamp for last schedule: */
1399 unsigned long long ftrace_timestamp;
1400
1401 /*
1402 * Number of functions that haven't been traced
1403 * because of depth overrun:
1404 */
1405 atomic_t trace_overrun;
1406
1407 /* Pause tracing: */
1408 atomic_t tracing_graph_pause;
1409#endif
1410
1411#ifdef CONFIG_TRACING
1412 /* Bitmask and counter of trace recursion: */
1413 unsigned long trace_recursion;
1414#endif /* CONFIG_TRACING */
1415
1416#ifdef CONFIG_KCOV
1417 /* See kernel/kcov.c for more details. */
1418
1419 /* Coverage collection mode enabled for this task (0 if disabled): */
1420 unsigned int kcov_mode;
1421
1422 /* Size of the kcov_area: */
1423 unsigned int kcov_size;
1424
1425 /* Buffer for coverage collection: */
1426 void *kcov_area;
1427
1428 /* KCOV descriptor wired with this task or NULL: */
1429 struct kcov *kcov;
1430
1431 /* KCOV common handle for remote coverage collection: */
1432 u64 kcov_handle;
1433
1434 /* KCOV sequence number: */
1435 int kcov_sequence;
1436
1437 /* Collect coverage from softirq context: */
1438 unsigned int kcov_softirq;
1439#endif
1440
1441#ifdef CONFIG_MEMCG
1442 struct mem_cgroup *memcg_in_oom;
1443 gfp_t memcg_oom_gfp_mask;
1444 int memcg_oom_order;
1445
1446 /* Number of pages to reclaim on returning to userland: */
1447 unsigned int memcg_nr_pages_over_high;
1448
1449 /* Used by memcontrol for targeted memcg charge: */
1450 struct mem_cgroup *active_memcg;
1451#endif
1452
1453#ifdef CONFIG_MEMCG_KMEM
1454 struct obj_cgroup *objcg;
1455#endif
1456
1457#ifdef CONFIG_BLK_CGROUP
1458 struct gendisk *throttle_disk;
1459#endif
1460
1461#ifdef CONFIG_UPROBES
1462 struct uprobe_task *utask;
1463#endif
1464#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1465 unsigned int sequential_io;
1466 unsigned int sequential_io_avg;
1467#endif
1468 struct kmap_ctrl kmap_ctrl;
1469#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
1470 unsigned long task_state_change;
1471# ifdef CONFIG_PREEMPT_RT
1472 unsigned long saved_state_change;
1473# endif
1474#endif
1475 struct rcu_head rcu;
1476 refcount_t rcu_users;
1477 int pagefault_disabled;
1478#ifdef CONFIG_MMU
1479 struct task_struct *oom_reaper_list;
1480 struct timer_list oom_reaper_timer;
1481#endif
1482#ifdef CONFIG_VMAP_STACK
1483 struct vm_struct *stack_vm_area;
1484#endif
1485#ifdef CONFIG_THREAD_INFO_IN_TASK
1486 /* A live task holds one reference: */
1487 refcount_t stack_refcount;
1488#endif
1489#ifdef CONFIG_LIVEPATCH
1490 int patch_state;
1491#endif
1492#ifdef CONFIG_SECURITY
1493 /* Used by LSM modules for access restriction: */
1494 void *security;
1495#endif
1496#ifdef CONFIG_BPF_SYSCALL
1497 /* Used by BPF task local storage */
1498 struct bpf_local_storage __rcu *bpf_storage;
1499 /* Used for BPF run context */
1500 struct bpf_run_ctx *bpf_ctx;
1501#endif
1502
1503#ifdef CONFIG_GCC_PLUGIN_STACKLEAK
1504 unsigned long lowest_stack;
1505 unsigned long prev_lowest_stack;
1506#endif
1507
1508#ifdef CONFIG_X86_MCE
1509 void __user *mce_vaddr;
1510 __u64 mce_kflags;
1511 u64 mce_addr;
1512 __u64 mce_ripv : 1,
1513 mce_whole_page : 1,
1514 __mce_reserved : 62;
1515 struct callback_head mce_kill_me;
1516 int mce_count;
1517#endif
1518
1519#ifdef CONFIG_KRETPROBES
1520 struct llist_head kretprobe_instances;
1521#endif
1522#ifdef CONFIG_RETHOOK
1523 struct llist_head rethooks;
1524#endif
1525
1526#ifdef CONFIG_ARCH_HAS_PARANOID_L1D_FLUSH
1527 /*
1528 * If L1D flush is supported on mm context switch
1529 * then we use this callback head to queue kill work
1530 * to kill tasks that are not running on SMT disabled
1531 * cores
1532 */
1533 struct callback_head l1d_flush_kill;
1534#endif
1535
1536#ifdef CONFIG_RV
1537 /*
1538 * Per-task RV monitor. Nowadays fixed in RV_PER_TASK_MONITORS.
1539 * If we find justification for more monitors, we can think
1540 * about adding more or developing a dynamic method. So far,
1541 * none of these are justified.
1542 */
1543 union rv_task_monitor rv[RV_PER_TASK_MONITORS];
1544#endif
1545
1546#ifdef CONFIG_USER_EVENTS
1547 struct user_event_mm *user_event_mm;
1548#endif
1549
1550 /*
1551 * New fields for task_struct should be added above here, so that
1552 * they are included in the randomized portion of task_struct.
1553 */
1554 randomized_struct_fields_end
1555
1556 /* CPU-specific state of this task: */
1557 struct thread_struct thread;
1558
1559 /*
1560 * WARNING: on x86, 'thread_struct' contains a variable-sized
1561 * structure. It *MUST* be at the end of 'task_struct'.
1562 *
1563 * Do not put anything below here!
1564 */
1565};
1566
1567#define TASK_REPORT_IDLE (TASK_REPORT + 1)
1568#define TASK_REPORT_MAX (TASK_REPORT_IDLE << 1)
1569
1570static inline unsigned int __task_state_index(unsigned int tsk_state,
1571 unsigned int tsk_exit_state)
1572{
1573 unsigned int state = (tsk_state | tsk_exit_state) & TASK_REPORT;
1574
1575 BUILD_BUG_ON_NOT_POWER_OF_2(TASK_REPORT_MAX);
1576
1577 if ((tsk_state & TASK_IDLE) == TASK_IDLE)
1578 state = TASK_REPORT_IDLE;
1579
1580 /*
1581 * We're lying here, but rather than expose a completely new task state
1582 * to userspace, we can make this appear as if the task has gone through
1583 * a regular rt_mutex_lock() call.
1584 */
1585 if (tsk_state & TASK_RTLOCK_WAIT)
1586 state = TASK_UNINTERRUPTIBLE;
1587
1588 return fls(state);
1589}
1590
1591static inline unsigned int task_state_index(struct task_struct *tsk)
1592{
1593 return __task_state_index(READ_ONCE(tsk->__state), tsk->exit_state);
1594}
1595
1596static inline char task_index_to_char(unsigned int state)
1597{
1598 static const char state_char[] = "RSDTtXZPI";
1599
1600 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != sizeof(state_char) - 1);
1601
1602 return state_char[state];
1603}
1604
1605static inline char task_state_to_char(struct task_struct *tsk)
1606{
1607 return task_index_to_char(task_state_index(tsk));
1608}
1609
1610extern struct pid *cad_pid;
1611
1612/*
1613 * Per process flags
1614 */
1615#define PF_VCPU 0x00000001 /* I'm a virtual CPU */
1616#define PF_IDLE 0x00000002 /* I am an IDLE thread */
1617#define PF_EXITING 0x00000004 /* Getting shut down */
1618#define PF_POSTCOREDUMP 0x00000008 /* Coredumps should ignore this task */
1619#define PF_IO_WORKER 0x00000010 /* Task is an IO worker */
1620#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1621#define PF_FORKNOEXEC 0x00000040 /* Forked but didn't exec */
1622#define PF_MCE_PROCESS 0x00000080 /* Process policy on mce errors */
1623#define PF_SUPERPRIV 0x00000100 /* Used super-user privileges */
1624#define PF_DUMPCORE 0x00000200 /* Dumped core */
1625#define PF_SIGNALED 0x00000400 /* Killed by a signal */
1626#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1627#define PF_NPROC_EXCEEDED 0x00001000 /* set_user() noticed that RLIMIT_NPROC was exceeded */
1628#define PF_USED_MATH 0x00002000 /* If unset the fpu must be initialized before use */
1629#define PF_USER_WORKER 0x00004000 /* Kernel thread cloned from userspace thread */
1630#define PF_NOFREEZE 0x00008000 /* This thread should not be frozen */
1631#define PF__HOLE__00010000 0x00010000
1632#define PF_KSWAPD 0x00020000 /* I am kswapd */
1633#define PF_MEMALLOC_NOFS 0x00040000 /* All allocation requests will inherit GFP_NOFS */
1634#define PF_MEMALLOC_NOIO 0x00080000 /* All allocation requests will inherit GFP_NOIO */
1635#define PF_LOCAL_THROTTLE 0x00100000 /* Throttle writes only against the bdi I write to,
1636 * I am cleaning dirty pages from some other bdi. */
1637#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1638#define PF_RANDOMIZE 0x00400000 /* Randomize virtual address space */
1639#define PF__HOLE__00800000 0x00800000
1640#define PF__HOLE__01000000 0x01000000
1641#define PF__HOLE__02000000 0x02000000
1642#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_mask */
1643#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1644#define PF_MEMALLOC_PIN 0x10000000 /* Allocation context constrained to zones which allow long term pinning. */
1645#define PF__HOLE__20000000 0x20000000
1646#define PF__HOLE__40000000 0x40000000
1647#define PF_SUSPEND_TASK 0x80000000 /* This thread called freeze_processes() and should not be frozen */
1648
1649/*
1650 * Only the _current_ task can read/write to tsk->flags, but other
1651 * tasks can access tsk->flags in readonly mode for example
1652 * with tsk_used_math (like during threaded core dumping).
1653 * There is however an exception to this rule during ptrace
1654 * or during fork: the ptracer task is allowed to write to the
1655 * child->flags of its traced child (same goes for fork, the parent
1656 * can write to the child->flags), because we're guaranteed the
1657 * child is not running and in turn not changing child->flags
1658 * at the same time the parent does it.
1659 */
1660#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1661#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1662#define clear_used_math() clear_stopped_child_used_math(current)
1663#define set_used_math() set_stopped_child_used_math(current)
1664
1665#define conditional_stopped_child_used_math(condition, child) \
1666 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1667
1668#define conditional_used_math(condition) conditional_stopped_child_used_math(condition, current)
1669
1670#define copy_to_stopped_child_used_math(child) \
1671 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1672
1673/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1674#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1675#define used_math() tsk_used_math(current)
1676
1677static __always_inline bool is_percpu_thread(void)
1678{
1679#ifdef CONFIG_SMP
1680 return (current->flags & PF_NO_SETAFFINITY) &&
1681 (current->nr_cpus_allowed == 1);
1682#else
1683 return true;
1684#endif
1685}
1686
1687/* Per-process atomic flags. */
1688#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1689#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1690#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1691#define PFA_SPEC_SSB_DISABLE 3 /* Speculative Store Bypass disabled */
1692#define PFA_SPEC_SSB_FORCE_DISABLE 4 /* Speculative Store Bypass force disabled*/
1693#define PFA_SPEC_IB_DISABLE 5 /* Indirect branch speculation restricted */
1694#define PFA_SPEC_IB_FORCE_DISABLE 6 /* Indirect branch speculation permanently restricted */
1695#define PFA_SPEC_SSB_NOEXEC 7 /* Speculative Store Bypass clear on execve() */
1696
1697#define TASK_PFA_TEST(name, func) \
1698 static inline bool task_##func(struct task_struct *p) \
1699 { return test_bit(PFA_##name, &p->atomic_flags); }
1700
1701#define TASK_PFA_SET(name, func) \
1702 static inline void task_set_##func(struct task_struct *p) \
1703 { set_bit(PFA_##name, &p->atomic_flags); }
1704
1705#define TASK_PFA_CLEAR(name, func) \
1706 static inline void task_clear_##func(struct task_struct *p) \
1707 { clear_bit(PFA_##name, &p->atomic_flags); }
1708
1709TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1710TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1711
1712TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1713TASK_PFA_SET(SPREAD_PAGE, spread_page)
1714TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1715
1716TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1717TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1718TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1719
1720TASK_PFA_TEST(SPEC_SSB_DISABLE, spec_ssb_disable)
1721TASK_PFA_SET(SPEC_SSB_DISABLE, spec_ssb_disable)
1722TASK_PFA_CLEAR(SPEC_SSB_DISABLE, spec_ssb_disable)
1723
1724TASK_PFA_TEST(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1725TASK_PFA_SET(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1726TASK_PFA_CLEAR(SPEC_SSB_NOEXEC, spec_ssb_noexec)
1727
1728TASK_PFA_TEST(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1729TASK_PFA_SET(SPEC_SSB_FORCE_DISABLE, spec_ssb_force_disable)
1730
1731TASK_PFA_TEST(SPEC_IB_DISABLE, spec_ib_disable)
1732TASK_PFA_SET(SPEC_IB_DISABLE, spec_ib_disable)
1733TASK_PFA_CLEAR(SPEC_IB_DISABLE, spec_ib_disable)
1734
1735TASK_PFA_TEST(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1736TASK_PFA_SET(SPEC_IB_FORCE_DISABLE, spec_ib_force_disable)
1737
1738static inline void
1739current_restore_flags(unsigned long orig_flags, unsigned long flags)
1740{
1741 current->flags &= ~flags;
1742 current->flags |= orig_flags & flags;
1743}
1744
1745extern int cpuset_cpumask_can_shrink(const struct cpumask *cur, const struct cpumask *trial);
1746extern int task_can_attach(struct task_struct *p);
1747extern int dl_bw_alloc(int cpu, u64 dl_bw);
1748extern void dl_bw_free(int cpu, u64 dl_bw);
1749#ifdef CONFIG_SMP
1750
1751/* do_set_cpus_allowed() - consider using set_cpus_allowed_ptr() instead */
1752extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
1753
1754/**
1755 * set_cpus_allowed_ptr - set CPU affinity mask of a task
1756 * @p: the task
1757 * @new_mask: CPU affinity mask
1758 *
1759 * Return: zero if successful, or a negative error code
1760 */
1761extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
1762extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
1763extern void release_user_cpus_ptr(struct task_struct *p);
1764extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
1765extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
1766extern void relax_compatible_cpus_allowed_ptr(struct task_struct *p);
1767#else
1768static inline void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
1769{
1770}
1771static inline int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
1772{
1773 if (!cpumask_test_cpu(0, new_mask))
1774 return -EINVAL;
1775 return 0;
1776}
1777static inline int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node)
1778{
1779 if (src->user_cpus_ptr)
1780 return -EINVAL;
1781 return 0;
1782}
1783static inline void release_user_cpus_ptr(struct task_struct *p)
1784{
1785 WARN_ON(p->user_cpus_ptr);
1786}
1787
1788static inline int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
1789{
1790 return 0;
1791}
1792#endif
1793
1794extern int yield_to(struct task_struct *p, bool preempt);
1795extern void set_user_nice(struct task_struct *p, long nice);
1796extern int task_prio(const struct task_struct *p);
1797
1798/**
1799 * task_nice - return the nice value of a given task.
1800 * @p: the task in question.
1801 *
1802 * Return: The nice value [ -20 ... 0 ... 19 ].
1803 */
1804static inline int task_nice(const struct task_struct *p)
1805{
1806 return PRIO_TO_NICE((p)->static_prio);
1807}
1808
1809extern int can_nice(const struct task_struct *p, const int nice);
1810extern int task_curr(const struct task_struct *p);
1811extern int idle_cpu(int cpu);
1812extern int available_idle_cpu(int cpu);
1813extern int sched_setscheduler(struct task_struct *, int, const struct sched_param *);
1814extern int sched_setscheduler_nocheck(struct task_struct *, int, const struct sched_param *);
1815extern void sched_set_fifo(struct task_struct *p);
1816extern void sched_set_fifo_low(struct task_struct *p);
1817extern void sched_set_normal(struct task_struct *p, int nice);
1818extern int sched_setattr(struct task_struct *, const struct sched_attr *);
1819extern int sched_setattr_nocheck(struct task_struct *, const struct sched_attr *);
1820extern struct task_struct *idle_task(int cpu);
1821
1822/**
1823 * is_idle_task - is the specified task an idle task?
1824 * @p: the task in question.
1825 *
1826 * Return: 1 if @p is an idle task. 0 otherwise.
1827 */
1828static __always_inline bool is_idle_task(const struct task_struct *p)
1829{
1830 return !!(p->flags & PF_IDLE);
1831}
1832
1833extern struct task_struct *curr_task(int cpu);
1834extern void ia64_set_curr_task(int cpu, struct task_struct *p);
1835
1836void yield(void);
1837
1838union thread_union {
1839 struct task_struct task;
1840#ifndef CONFIG_THREAD_INFO_IN_TASK
1841 struct thread_info thread_info;
1842#endif
1843 unsigned long stack[THREAD_SIZE/sizeof(long)];
1844};
1845
1846#ifndef CONFIG_THREAD_INFO_IN_TASK
1847extern struct thread_info init_thread_info;
1848#endif
1849
1850extern unsigned long init_stack[THREAD_SIZE / sizeof(unsigned long)];
1851
1852#ifdef CONFIG_THREAD_INFO_IN_TASK
1853# define task_thread_info(task) (&(task)->thread_info)
1854#elif !defined(__HAVE_THREAD_FUNCTIONS)
1855# define task_thread_info(task) ((struct thread_info *)(task)->stack)
1856#endif
1857
1858/*
1859 * find a task by one of its numerical ids
1860 *
1861 * find_task_by_pid_ns():
1862 * finds a task by its pid in the specified namespace
1863 * find_task_by_vpid():
1864 * finds a task by its virtual pid
1865 *
1866 * see also find_vpid() etc in include/linux/pid.h
1867 */
1868
1869extern struct task_struct *find_task_by_vpid(pid_t nr);
1870extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns);
1871
1872/*
1873 * find a task by its virtual pid and get the task struct
1874 */
1875extern struct task_struct *find_get_task_by_vpid(pid_t nr);
1876
1877extern int wake_up_state(struct task_struct *tsk, unsigned int state);
1878extern int wake_up_process(struct task_struct *tsk);
1879extern void wake_up_new_task(struct task_struct *tsk);
1880
1881#ifdef CONFIG_SMP
1882extern void kick_process(struct task_struct *tsk);
1883#else
1884static inline void kick_process(struct task_struct *tsk) { }
1885#endif
1886
1887extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
1888
1889static inline void set_task_comm(struct task_struct *tsk, const char *from)
1890{
1891 __set_task_comm(tsk, from, false);
1892}
1893
1894extern char *__get_task_comm(char *to, size_t len, struct task_struct *tsk);
1895#define get_task_comm(buf, tsk) ({ \
1896 BUILD_BUG_ON(sizeof(buf) != TASK_COMM_LEN); \
1897 __get_task_comm(buf, sizeof(buf), tsk); \
1898})
1899
1900#ifdef CONFIG_SMP
1901static __always_inline void scheduler_ipi(void)
1902{
1903 /*
1904 * Fold TIF_NEED_RESCHED into the preempt_count; anybody setting
1905 * TIF_NEED_RESCHED remotely (for the first time) will also send
1906 * this IPI.
1907 */
1908 preempt_fold_need_resched();
1909}
1910#else
1911static inline void scheduler_ipi(void) { }
1912#endif
1913
1914extern unsigned long wait_task_inactive(struct task_struct *, unsigned int match_state);
1915
1916/*
1917 * Set thread flags in other task's structures.
1918 * See asm/thread_info.h for TIF_xxxx flags available:
1919 */
1920static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
1921{
1922 set_ti_thread_flag(task_thread_info(tsk), flag);
1923}
1924
1925static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1926{
1927 clear_ti_thread_flag(task_thread_info(tsk), flag);
1928}
1929
1930static inline void update_tsk_thread_flag(struct task_struct *tsk, int flag,
1931 bool value)
1932{
1933 update_ti_thread_flag(task_thread_info(tsk), flag, value);
1934}
1935
1936static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
1937{
1938 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
1939}
1940
1941static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
1942{
1943 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
1944}
1945
1946static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
1947{
1948 return test_ti_thread_flag(task_thread_info(tsk), flag);
1949}
1950
1951static inline void set_tsk_need_resched(struct task_struct *tsk)
1952{
1953 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1954}
1955
1956static inline void clear_tsk_need_resched(struct task_struct *tsk)
1957{
1958 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
1959}
1960
1961static inline int test_tsk_need_resched(struct task_struct *tsk)
1962{
1963 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
1964}
1965
1966/*
1967 * cond_resched() and cond_resched_lock(): latency reduction via
1968 * explicit rescheduling in places that are safe. The return
1969 * value indicates whether a reschedule was done in fact.
1970 * cond_resched_lock() will drop the spinlock before scheduling,
1971 */
1972#if !defined(CONFIG_PREEMPTION) || defined(CONFIG_PREEMPT_DYNAMIC)
1973extern int __cond_resched(void);
1974
1975#if defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
1976
1977void sched_dynamic_klp_enable(void);
1978void sched_dynamic_klp_disable(void);
1979
1980DECLARE_STATIC_CALL(cond_resched, __cond_resched);
1981
1982static __always_inline int _cond_resched(void)
1983{
1984 return static_call_mod(cond_resched)();
1985}
1986
1987#elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_KEY)
1988
1989extern int dynamic_cond_resched(void);
1990
1991static __always_inline int _cond_resched(void)
1992{
1993 return dynamic_cond_resched();
1994}
1995
1996#else /* !CONFIG_PREEMPTION */
1997
1998static inline int _cond_resched(void)
1999{
2000 klp_sched_try_switch();
2001 return __cond_resched();
2002}
2003
2004#endif /* PREEMPT_DYNAMIC && CONFIG_HAVE_PREEMPT_DYNAMIC_CALL */
2005
2006#else /* CONFIG_PREEMPTION && !CONFIG_PREEMPT_DYNAMIC */
2007
2008static inline int _cond_resched(void)
2009{
2010 klp_sched_try_switch();
2011 return 0;
2012}
2013
2014#endif /* !CONFIG_PREEMPTION || CONFIG_PREEMPT_DYNAMIC */
2015
2016#define cond_resched() ({ \
2017 __might_resched(__FILE__, __LINE__, 0); \
2018 _cond_resched(); \
2019})
2020
2021extern int __cond_resched_lock(spinlock_t *lock);
2022extern int __cond_resched_rwlock_read(rwlock_t *lock);
2023extern int __cond_resched_rwlock_write(rwlock_t *lock);
2024
2025#define MIGHT_RESCHED_RCU_SHIFT 8
2026#define MIGHT_RESCHED_PREEMPT_MASK ((1U << MIGHT_RESCHED_RCU_SHIFT) - 1)
2027
2028#ifndef CONFIG_PREEMPT_RT
2029/*
2030 * Non RT kernels have an elevated preempt count due to the held lock,
2031 * but are not allowed to be inside a RCU read side critical section
2032 */
2033# define PREEMPT_LOCK_RESCHED_OFFSETS PREEMPT_LOCK_OFFSET
2034#else
2035/*
2036 * spin/rw_lock() on RT implies rcu_read_lock(). The might_sleep() check in
2037 * cond_resched*lock() has to take that into account because it checks for
2038 * preempt_count() and rcu_preempt_depth().
2039 */
2040# define PREEMPT_LOCK_RESCHED_OFFSETS \
2041 (PREEMPT_LOCK_OFFSET + (1U << MIGHT_RESCHED_RCU_SHIFT))
2042#endif
2043
2044#define cond_resched_lock(lock) ({ \
2045 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2046 __cond_resched_lock(lock); \
2047})
2048
2049#define cond_resched_rwlock_read(lock) ({ \
2050 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2051 __cond_resched_rwlock_read(lock); \
2052})
2053
2054#define cond_resched_rwlock_write(lock) ({ \
2055 __might_resched(__FILE__, __LINE__, PREEMPT_LOCK_RESCHED_OFFSETS); \
2056 __cond_resched_rwlock_write(lock); \
2057})
2058
2059#ifdef CONFIG_PREEMPT_DYNAMIC
2060
2061extern bool preempt_model_none(void);
2062extern bool preempt_model_voluntary(void);
2063extern bool preempt_model_full(void);
2064
2065#else
2066
2067static inline bool preempt_model_none(void)
2068{
2069 return IS_ENABLED(CONFIG_PREEMPT_NONE);
2070}
2071static inline bool preempt_model_voluntary(void)
2072{
2073 return IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY);
2074}
2075static inline bool preempt_model_full(void)
2076{
2077 return IS_ENABLED(CONFIG_PREEMPT);
2078}
2079
2080#endif
2081
2082static inline bool preempt_model_rt(void)
2083{
2084 return IS_ENABLED(CONFIG_PREEMPT_RT);
2085}
2086
2087/*
2088 * Does the preemption model allow non-cooperative preemption?
2089 *
2090 * For !CONFIG_PREEMPT_DYNAMIC kernels this is an exact match with
2091 * CONFIG_PREEMPTION; for CONFIG_PREEMPT_DYNAMIC this doesn't work as the
2092 * kernel is *built* with CONFIG_PREEMPTION=y but may run with e.g. the
2093 * PREEMPT_NONE model.
2094 */
2095static inline bool preempt_model_preemptible(void)
2096{
2097 return preempt_model_full() || preempt_model_rt();
2098}
2099
2100static __always_inline bool need_resched(void)
2101{
2102 return unlikely(tif_need_resched());
2103}
2104
2105/*
2106 * Wrappers for p->thread_info->cpu access. No-op on UP.
2107 */
2108#ifdef CONFIG_SMP
2109
2110static inline unsigned int task_cpu(const struct task_struct *p)
2111{
2112 return READ_ONCE(task_thread_info(p)->cpu);
2113}
2114
2115extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2116
2117#else
2118
2119static inline unsigned int task_cpu(const struct task_struct *p)
2120{
2121 return 0;
2122}
2123
2124static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2125{
2126}
2127
2128#endif /* CONFIG_SMP */
2129
2130extern bool sched_task_on_rq(struct task_struct *p);
2131extern unsigned long get_wchan(struct task_struct *p);
2132extern struct task_struct *cpu_curr_snapshot(int cpu);
2133
2134#include <linux/spinlock.h>
2135
2136/*
2137 * In order to reduce various lock holder preemption latencies provide an
2138 * interface to see if a vCPU is currently running or not.
2139 *
2140 * This allows us to terminate optimistic spin loops and block, analogous to
2141 * the native optimistic spin heuristic of testing if the lock owner task is
2142 * running or not.
2143 */
2144#ifndef vcpu_is_preempted
2145static inline bool vcpu_is_preempted(int cpu)
2146{
2147 return false;
2148}
2149#endif
2150
2151extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2152extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2153
2154#ifndef TASK_SIZE_OF
2155#define TASK_SIZE_OF(tsk) TASK_SIZE
2156#endif
2157
2158#ifdef CONFIG_SMP
2159static inline bool owner_on_cpu(struct task_struct *owner)
2160{
2161 /*
2162 * As lock holder preemption issue, we both skip spinning if
2163 * task is not on cpu or its cpu is preempted
2164 */
2165 return READ_ONCE(owner->on_cpu) && !vcpu_is_preempted(task_cpu(owner));
2166}
2167
2168/* Returns effective CPU energy utilization, as seen by the scheduler */
2169unsigned long sched_cpu_util(int cpu);
2170#endif /* CONFIG_SMP */
2171
2172#ifdef CONFIG_SCHED_CORE
2173extern void sched_core_free(struct task_struct *tsk);
2174extern void sched_core_fork(struct task_struct *p);
2175extern int sched_core_share_pid(unsigned int cmd, pid_t pid, enum pid_type type,
2176 unsigned long uaddr);
2177extern int sched_core_idle_cpu(int cpu);
2178#else
2179static inline void sched_core_free(struct task_struct *tsk) { }
2180static inline void sched_core_fork(struct task_struct *p) { }
2181static inline int sched_core_idle_cpu(int cpu) { return idle_cpu(cpu); }
2182#endif
2183
2184extern void sched_set_stop_task(int cpu, struct task_struct *stop);
2185
2186#endif