Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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/shm.h>
37#include <linux/signal.h>
38#include <linux/compiler.h>
39#include <linux/completion.h>
40#include <linux/pid.h>
41#include <linux/percpu.h>
42#include <linux/topology.h>
43#include <linux/proportions.h>
44#include <linux/seccomp.h>
45#include <linux/rcupdate.h>
46#include <linux/rculist.h>
47#include <linux/rtmutex.h>
48
49#include <linux/time.h>
50#include <linux/param.h>
51#include <linux/resource.h>
52#include <linux/timer.h>
53#include <linux/hrtimer.h>
54#include <linux/task_io_accounting.h>
55#include <linux/latencytop.h>
56#include <linux/cred.h>
57#include <linux/llist.h>
58#include <linux/uidgid.h>
59#include <linux/gfp.h>
60
61#include <asm/processor.h>
62
63#define SCHED_ATTR_SIZE_VER0 48 /* sizeof first published struct */
64
65/*
66 * Extended scheduling parameters data structure.
67 *
68 * This is needed because the original struct sched_param can not be
69 * altered without introducing ABI issues with legacy applications
70 * (e.g., in sched_getparam()).
71 *
72 * However, the possibility of specifying more than just a priority for
73 * the tasks may be useful for a wide variety of application fields, e.g.,
74 * multimedia, streaming, automation and control, and many others.
75 *
76 * This variant (sched_attr) is meant at describing a so-called
77 * sporadic time-constrained task. In such model a task is specified by:
78 * - the activation period or minimum instance inter-arrival time;
79 * - the maximum (or average, depending on the actual scheduling
80 * discipline) computation time of all instances, a.k.a. runtime;
81 * - the deadline (relative to the actual activation time) of each
82 * instance.
83 * Very briefly, a periodic (sporadic) task asks for the execution of
84 * some specific computation --which is typically called an instance--
85 * (at most) every period. Moreover, each instance typically lasts no more
86 * than the runtime and must be completed by time instant t equal to
87 * the instance activation time + the deadline.
88 *
89 * This is reflected by the actual fields of the sched_attr structure:
90 *
91 * @size size of the structure, for fwd/bwd compat.
92 *
93 * @sched_policy task's scheduling policy
94 * @sched_flags for customizing the scheduler behaviour
95 * @sched_nice task's nice value (SCHED_NORMAL/BATCH)
96 * @sched_priority task's static priority (SCHED_FIFO/RR)
97 * @sched_deadline representative of the task's deadline
98 * @sched_runtime representative of the task's runtime
99 * @sched_period representative of the task's period
100 *
101 * Given this task model, there are a multiplicity of scheduling algorithms
102 * and policies, that can be used to ensure all the tasks will make their
103 * timing constraints.
104 *
105 * As of now, the SCHED_DEADLINE policy (sched_dl scheduling class) is the
106 * only user of this new interface. More information about the algorithm
107 * available in the scheduling class file or in Documentation/.
108 */
109struct sched_attr {
110 u32 size;
111
112 u32 sched_policy;
113 u64 sched_flags;
114
115 /* SCHED_NORMAL, SCHED_BATCH */
116 s32 sched_nice;
117
118 /* SCHED_FIFO, SCHED_RR */
119 u32 sched_priority;
120
121 /* SCHED_DEADLINE */
122 u64 sched_runtime;
123 u64 sched_deadline;
124 u64 sched_period;
125};
126
127struct exec_domain;
128struct futex_pi_state;
129struct robust_list_head;
130struct bio_list;
131struct fs_struct;
132struct perf_event_context;
133struct blk_plug;
134struct filename;
135
136#define VMACACHE_BITS 2
137#define VMACACHE_SIZE (1U << VMACACHE_BITS)
138#define VMACACHE_MASK (VMACACHE_SIZE - 1)
139
140/*
141 * These are the constant used to fake the fixed-point load-average
142 * counting. Some notes:
143 * - 11 bit fractions expand to 22 bits by the multiplies: this gives
144 * a load-average precision of 10 bits integer + 11 bits fractional
145 * - if you want to count load-averages more often, you need more
146 * precision, or rounding will get you. With 2-second counting freq,
147 * the EXP_n values would be 1981, 2034 and 2043 if still using only
148 * 11 bit fractions.
149 */
150extern unsigned long avenrun[]; /* Load averages */
151extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift);
152
153#define FSHIFT 11 /* nr of bits of precision */
154#define FIXED_1 (1<<FSHIFT) /* 1.0 as fixed-point */
155#define LOAD_FREQ (5*HZ+1) /* 5 sec intervals */
156#define EXP_1 1884 /* 1/exp(5sec/1min) as fixed-point */
157#define EXP_5 2014 /* 1/exp(5sec/5min) */
158#define EXP_15 2037 /* 1/exp(5sec/15min) */
159
160#define CALC_LOAD(load,exp,n) \
161 load *= exp; \
162 load += n*(FIXED_1-exp); \
163 load >>= FSHIFT;
164
165extern unsigned long total_forks;
166extern int nr_threads;
167DECLARE_PER_CPU(unsigned long, process_counts);
168extern int nr_processes(void);
169extern unsigned long nr_running(void);
170extern unsigned long nr_iowait(void);
171extern unsigned long nr_iowait_cpu(int cpu);
172extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load);
173
174extern void calc_global_load(unsigned long ticks);
175extern void update_cpu_load_nohz(void);
176
177extern unsigned long get_parent_ip(unsigned long addr);
178
179extern void dump_cpu_task(int cpu);
180
181struct seq_file;
182struct cfs_rq;
183struct task_group;
184#ifdef CONFIG_SCHED_DEBUG
185extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
186extern void proc_sched_set_task(struct task_struct *p);
187extern void
188print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq);
189#endif
190
191/*
192 * Task state bitmask. NOTE! These bits are also
193 * encoded in fs/proc/array.c: get_task_state().
194 *
195 * We have two separate sets of flags: task->state
196 * is about runnability, while task->exit_state are
197 * about the task exiting. Confusing, but this way
198 * modifying one set can't modify the other one by
199 * mistake.
200 */
201#define TASK_RUNNING 0
202#define TASK_INTERRUPTIBLE 1
203#define TASK_UNINTERRUPTIBLE 2
204#define __TASK_STOPPED 4
205#define __TASK_TRACED 8
206/* in tsk->exit_state */
207#define EXIT_DEAD 16
208#define EXIT_ZOMBIE 32
209#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
210/* in tsk->state again */
211#define TASK_DEAD 64
212#define TASK_WAKEKILL 128
213#define TASK_WAKING 256
214#define TASK_PARKED 512
215#define TASK_STATE_MAX 1024
216
217#define TASK_STATE_TO_CHAR_STR "RSDTtXZxKWP"
218
219extern char ___assert_task_state[1 - 2*!!(
220 sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
221
222/* Convenience macros for the sake of set_task_state */
223#define TASK_KILLABLE (TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
224#define TASK_STOPPED (TASK_WAKEKILL | __TASK_STOPPED)
225#define TASK_TRACED (TASK_WAKEKILL | __TASK_TRACED)
226
227/* Convenience macros for the sake of wake_up */
228#define TASK_NORMAL (TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)
229#define TASK_ALL (TASK_NORMAL | __TASK_STOPPED | __TASK_TRACED)
230
231/* get_task_state() */
232#define TASK_REPORT (TASK_RUNNING | TASK_INTERRUPTIBLE | \
233 TASK_UNINTERRUPTIBLE | __TASK_STOPPED | \
234 __TASK_TRACED | EXIT_ZOMBIE | EXIT_DEAD)
235
236#define task_is_traced(task) ((task->state & __TASK_TRACED) != 0)
237#define task_is_stopped(task) ((task->state & __TASK_STOPPED) != 0)
238#define task_is_stopped_or_traced(task) \
239 ((task->state & (__TASK_STOPPED | __TASK_TRACED)) != 0)
240#define task_contributes_to_load(task) \
241 ((task->state & TASK_UNINTERRUPTIBLE) != 0 && \
242 (task->flags & PF_FROZEN) == 0)
243
244#define __set_task_state(tsk, state_value) \
245 do { (tsk)->state = (state_value); } while (0)
246#define set_task_state(tsk, state_value) \
247 set_mb((tsk)->state, (state_value))
248
249/*
250 * set_current_state() includes a barrier so that the write of current->state
251 * is correctly serialised wrt the caller's subsequent test of whether to
252 * actually sleep:
253 *
254 * set_current_state(TASK_UNINTERRUPTIBLE);
255 * if (do_i_need_to_sleep())
256 * schedule();
257 *
258 * If the caller does not need such serialisation then use __set_current_state()
259 */
260#define __set_current_state(state_value) \
261 do { current->state = (state_value); } while (0)
262#define set_current_state(state_value) \
263 set_mb(current->state, (state_value))
264
265/* Task command name length */
266#define TASK_COMM_LEN 16
267
268#include <linux/spinlock.h>
269
270/*
271 * This serializes "schedule()" and also protects
272 * the run-queue from deletions/modifications (but
273 * _adding_ to the beginning of the run-queue has
274 * a separate lock).
275 */
276extern rwlock_t tasklist_lock;
277extern spinlock_t mmlist_lock;
278
279struct task_struct;
280
281#ifdef CONFIG_PROVE_RCU
282extern int lockdep_tasklist_lock_is_held(void);
283#endif /* #ifdef CONFIG_PROVE_RCU */
284
285extern void sched_init(void);
286extern void sched_init_smp(void);
287extern asmlinkage void schedule_tail(struct task_struct *prev);
288extern void init_idle(struct task_struct *idle, int cpu);
289extern void init_idle_bootup_task(struct task_struct *idle);
290
291extern int runqueue_is_locked(int cpu);
292
293#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
294extern void nohz_balance_enter_idle(int cpu);
295extern void set_cpu_sd_state_idle(void);
296extern int get_nohz_timer_target(int pinned);
297#else
298static inline void nohz_balance_enter_idle(int cpu) { }
299static inline void set_cpu_sd_state_idle(void) { }
300static inline int get_nohz_timer_target(int pinned)
301{
302 return smp_processor_id();
303}
304#endif
305
306/*
307 * Only dump TASK_* tasks. (0 for all tasks)
308 */
309extern void show_state_filter(unsigned long state_filter);
310
311static inline void show_state(void)
312{
313 show_state_filter(0);
314}
315
316extern void show_regs(struct pt_regs *);
317
318/*
319 * TASK is a pointer to the task whose backtrace we want to see (or NULL for current
320 * task), SP is the stack pointer of the first frame that should be shown in the back
321 * trace (or NULL if the entire call-chain of the task should be shown).
322 */
323extern void show_stack(struct task_struct *task, unsigned long *sp);
324
325void io_schedule(void);
326long io_schedule_timeout(long timeout);
327
328extern void cpu_init (void);
329extern void trap_init(void);
330extern void update_process_times(int user);
331extern void scheduler_tick(void);
332
333extern void sched_show_task(struct task_struct *p);
334
335#ifdef CONFIG_LOCKUP_DETECTOR
336extern void touch_softlockup_watchdog(void);
337extern void touch_softlockup_watchdog_sync(void);
338extern void touch_all_softlockup_watchdogs(void);
339extern int proc_dowatchdog_thresh(struct ctl_table *table, int write,
340 void __user *buffer,
341 size_t *lenp, loff_t *ppos);
342extern unsigned int softlockup_panic;
343void lockup_detector_init(void);
344#else
345static inline void touch_softlockup_watchdog(void)
346{
347}
348static inline void touch_softlockup_watchdog_sync(void)
349{
350}
351static inline void touch_all_softlockup_watchdogs(void)
352{
353}
354static inline void lockup_detector_init(void)
355{
356}
357#endif
358
359#ifdef CONFIG_DETECT_HUNG_TASK
360void reset_hung_task_detector(void);
361#else
362static inline void reset_hung_task_detector(void)
363{
364}
365#endif
366
367/* Attach to any functions which should be ignored in wchan output. */
368#define __sched __attribute__((__section__(".sched.text")))
369
370/* Linker adds these: start and end of __sched functions */
371extern char __sched_text_start[], __sched_text_end[];
372
373/* Is this address in the __sched functions? */
374extern int in_sched_functions(unsigned long addr);
375
376#define MAX_SCHEDULE_TIMEOUT LONG_MAX
377extern signed long schedule_timeout(signed long timeout);
378extern signed long schedule_timeout_interruptible(signed long timeout);
379extern signed long schedule_timeout_killable(signed long timeout);
380extern signed long schedule_timeout_uninterruptible(signed long timeout);
381asmlinkage void schedule(void);
382extern void schedule_preempt_disabled(void);
383
384struct nsproxy;
385struct user_namespace;
386
387#ifdef CONFIG_MMU
388extern void arch_pick_mmap_layout(struct mm_struct *mm);
389extern unsigned long
390arch_get_unmapped_area(struct file *, unsigned long, unsigned long,
391 unsigned long, unsigned long);
392extern unsigned long
393arch_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
394 unsigned long len, unsigned long pgoff,
395 unsigned long flags);
396#else
397static inline void arch_pick_mmap_layout(struct mm_struct *mm) {}
398#endif
399
400#define SUID_DUMP_DISABLE 0 /* No setuid dumping */
401#define SUID_DUMP_USER 1 /* Dump as user of process */
402#define SUID_DUMP_ROOT 2 /* Dump as root */
403
404/* mm flags */
405
406/* for SUID_DUMP_* above */
407#define MMF_DUMPABLE_BITS 2
408#define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
409
410extern void set_dumpable(struct mm_struct *mm, int value);
411/*
412 * This returns the actual value of the suid_dumpable flag. For things
413 * that are using this for checking for privilege transitions, it must
414 * test against SUID_DUMP_USER rather than treating it as a boolean
415 * value.
416 */
417static inline int __get_dumpable(unsigned long mm_flags)
418{
419 return mm_flags & MMF_DUMPABLE_MASK;
420}
421
422static inline int get_dumpable(struct mm_struct *mm)
423{
424 return __get_dumpable(mm->flags);
425}
426
427/* coredump filter bits */
428#define MMF_DUMP_ANON_PRIVATE 2
429#define MMF_DUMP_ANON_SHARED 3
430#define MMF_DUMP_MAPPED_PRIVATE 4
431#define MMF_DUMP_MAPPED_SHARED 5
432#define MMF_DUMP_ELF_HEADERS 6
433#define MMF_DUMP_HUGETLB_PRIVATE 7
434#define MMF_DUMP_HUGETLB_SHARED 8
435
436#define MMF_DUMP_FILTER_SHIFT MMF_DUMPABLE_BITS
437#define MMF_DUMP_FILTER_BITS 7
438#define MMF_DUMP_FILTER_MASK \
439 (((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
440#define MMF_DUMP_FILTER_DEFAULT \
441 ((1 << MMF_DUMP_ANON_PRIVATE) | (1 << MMF_DUMP_ANON_SHARED) |\
442 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
443
444#ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
445# define MMF_DUMP_MASK_DEFAULT_ELF (1 << MMF_DUMP_ELF_HEADERS)
446#else
447# define MMF_DUMP_MASK_DEFAULT_ELF 0
448#endif
449 /* leave room for more dump flags */
450#define MMF_VM_MERGEABLE 16 /* KSM may merge identical pages */
451#define MMF_VM_HUGEPAGE 17 /* set when VM_HUGEPAGE is set on vma */
452#define MMF_EXE_FILE_CHANGED 18 /* see prctl_set_mm_exe_file() */
453
454#define MMF_HAS_UPROBES 19 /* has uprobes */
455#define MMF_RECALC_UPROBES 20 /* MMF_HAS_UPROBES can be wrong */
456
457#define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
458
459struct sighand_struct {
460 atomic_t count;
461 struct k_sigaction action[_NSIG];
462 spinlock_t siglock;
463 wait_queue_head_t signalfd_wqh;
464};
465
466struct pacct_struct {
467 int ac_flag;
468 long ac_exitcode;
469 unsigned long ac_mem;
470 cputime_t ac_utime, ac_stime;
471 unsigned long ac_minflt, ac_majflt;
472};
473
474struct cpu_itimer {
475 cputime_t expires;
476 cputime_t incr;
477 u32 error;
478 u32 incr_error;
479};
480
481/**
482 * struct cputime - snaphsot of system and user cputime
483 * @utime: time spent in user mode
484 * @stime: time spent in system mode
485 *
486 * Gathers a generic snapshot of user and system time.
487 */
488struct cputime {
489 cputime_t utime;
490 cputime_t stime;
491};
492
493/**
494 * struct task_cputime - collected CPU time counts
495 * @utime: time spent in user mode, in &cputime_t units
496 * @stime: time spent in kernel mode, in &cputime_t units
497 * @sum_exec_runtime: total time spent on the CPU, in nanoseconds
498 *
499 * This is an extension of struct cputime that includes the total runtime
500 * spent by the task from the scheduler point of view.
501 *
502 * As a result, this structure groups together three kinds of CPU time
503 * that are tracked for threads and thread groups. Most things considering
504 * CPU time want to group these counts together and treat all three
505 * of them in parallel.
506 */
507struct task_cputime {
508 cputime_t utime;
509 cputime_t stime;
510 unsigned long long sum_exec_runtime;
511};
512/* Alternate field names when used to cache expirations. */
513#define prof_exp stime
514#define virt_exp utime
515#define sched_exp sum_exec_runtime
516
517#define INIT_CPUTIME \
518 (struct task_cputime) { \
519 .utime = 0, \
520 .stime = 0, \
521 .sum_exec_runtime = 0, \
522 }
523
524#ifdef CONFIG_PREEMPT_COUNT
525#define PREEMPT_DISABLED (1 + PREEMPT_ENABLED)
526#else
527#define PREEMPT_DISABLED PREEMPT_ENABLED
528#endif
529
530/*
531 * Disable preemption until the scheduler is running.
532 * Reset by start_kernel()->sched_init()->init_idle().
533 *
534 * We include PREEMPT_ACTIVE to avoid cond_resched() from working
535 * before the scheduler is active -- see should_resched().
536 */
537#define INIT_PREEMPT_COUNT (PREEMPT_DISABLED + PREEMPT_ACTIVE)
538
539/**
540 * struct thread_group_cputimer - thread group interval timer counts
541 * @cputime: thread group interval timers.
542 * @running: non-zero when there are timers running and
543 * @cputime receives updates.
544 * @lock: lock for fields in this struct.
545 *
546 * This structure contains the version of task_cputime, above, that is
547 * used for thread group CPU timer calculations.
548 */
549struct thread_group_cputimer {
550 struct task_cputime cputime;
551 int running;
552 raw_spinlock_t lock;
553};
554
555#include <linux/rwsem.h>
556struct autogroup;
557
558/*
559 * NOTE! "signal_struct" does not have its own
560 * locking, because a shared signal_struct always
561 * implies a shared sighand_struct, so locking
562 * sighand_struct is always a proper superset of
563 * the locking of signal_struct.
564 */
565struct signal_struct {
566 atomic_t sigcnt;
567 atomic_t live;
568 int nr_threads;
569 struct list_head thread_head;
570
571 wait_queue_head_t wait_chldexit; /* for wait4() */
572
573 /* current thread group signal load-balancing target: */
574 struct task_struct *curr_target;
575
576 /* shared signal handling: */
577 struct sigpending shared_pending;
578
579 /* thread group exit support */
580 int group_exit_code;
581 /* overloaded:
582 * - notify group_exit_task when ->count is equal to notify_count
583 * - everyone except group_exit_task is stopped during signal delivery
584 * of fatal signals, group_exit_task processes the signal.
585 */
586 int notify_count;
587 struct task_struct *group_exit_task;
588
589 /* thread group stop support, overloads group_exit_code too */
590 int group_stop_count;
591 unsigned int flags; /* see SIGNAL_* flags below */
592
593 /*
594 * PR_SET_CHILD_SUBREAPER marks a process, like a service
595 * manager, to re-parent orphan (double-forking) child processes
596 * to this process instead of 'init'. The service manager is
597 * able to receive SIGCHLD signals and is able to investigate
598 * the process until it calls wait(). All children of this
599 * process will inherit a flag if they should look for a
600 * child_subreaper process at exit.
601 */
602 unsigned int is_child_subreaper:1;
603 unsigned int has_child_subreaper:1;
604
605 /* POSIX.1b Interval Timers */
606 int posix_timer_id;
607 struct list_head posix_timers;
608
609 /* ITIMER_REAL timer for the process */
610 struct hrtimer real_timer;
611 struct pid *leader_pid;
612 ktime_t it_real_incr;
613
614 /*
615 * ITIMER_PROF and ITIMER_VIRTUAL timers for the process, we use
616 * CPUCLOCK_PROF and CPUCLOCK_VIRT for indexing array as these
617 * values are defined to 0 and 1 respectively
618 */
619 struct cpu_itimer it[2];
620
621 /*
622 * Thread group totals for process CPU timers.
623 * See thread_group_cputimer(), et al, for details.
624 */
625 struct thread_group_cputimer cputimer;
626
627 /* Earliest-expiration cache. */
628 struct task_cputime cputime_expires;
629
630 struct list_head cpu_timers[3];
631
632 struct pid *tty_old_pgrp;
633
634 /* boolean value for session group leader */
635 int leader;
636
637 struct tty_struct *tty; /* NULL if no tty */
638
639#ifdef CONFIG_SCHED_AUTOGROUP
640 struct autogroup *autogroup;
641#endif
642 /*
643 * Cumulative resource counters for dead threads in the group,
644 * and for reaped dead child processes forked by this group.
645 * Live threads maintain their own counters and add to these
646 * in __exit_signal, except for the group leader.
647 */
648 cputime_t utime, stime, cutime, cstime;
649 cputime_t gtime;
650 cputime_t cgtime;
651#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
652 struct cputime prev_cputime;
653#endif
654 unsigned long nvcsw, nivcsw, cnvcsw, cnivcsw;
655 unsigned long min_flt, maj_flt, cmin_flt, cmaj_flt;
656 unsigned long inblock, oublock, cinblock, coublock;
657 unsigned long maxrss, cmaxrss;
658 struct task_io_accounting ioac;
659
660 /*
661 * Cumulative ns of schedule CPU time fo dead threads in the
662 * group, not including a zombie group leader, (This only differs
663 * from jiffies_to_ns(utime + stime) if sched_clock uses something
664 * other than jiffies.)
665 */
666 unsigned long long sum_sched_runtime;
667
668 /*
669 * We don't bother to synchronize most readers of this at all,
670 * because there is no reader checking a limit that actually needs
671 * to get both rlim_cur and rlim_max atomically, and either one
672 * alone is a single word that can safely be read normally.
673 * getrlimit/setrlimit use task_lock(current->group_leader) to
674 * protect this instead of the siglock, because they really
675 * have no need to disable irqs.
676 */
677 struct rlimit rlim[RLIM_NLIMITS];
678
679#ifdef CONFIG_BSD_PROCESS_ACCT
680 struct pacct_struct pacct; /* per-process accounting information */
681#endif
682#ifdef CONFIG_TASKSTATS
683 struct taskstats *stats;
684#endif
685#ifdef CONFIG_AUDIT
686 unsigned audit_tty;
687 unsigned audit_tty_log_passwd;
688 struct tty_audit_buf *tty_audit_buf;
689#endif
690#ifdef CONFIG_CGROUPS
691 /*
692 * group_rwsem prevents new tasks from entering the threadgroup and
693 * member tasks from exiting,a more specifically, setting of
694 * PF_EXITING. fork and exit paths are protected with this rwsem
695 * using threadgroup_change_begin/end(). Users which require
696 * threadgroup to remain stable should use threadgroup_[un]lock()
697 * which also takes care of exec path. Currently, cgroup is the
698 * only user.
699 */
700 struct rw_semaphore group_rwsem;
701#endif
702
703 oom_flags_t oom_flags;
704 short oom_score_adj; /* OOM kill score adjustment */
705 short oom_score_adj_min; /* OOM kill score adjustment min value.
706 * Only settable by CAP_SYS_RESOURCE. */
707
708 struct mutex cred_guard_mutex; /* guard against foreign influences on
709 * credential calculations
710 * (notably. ptrace) */
711};
712
713/*
714 * Bits in flags field of signal_struct.
715 */
716#define SIGNAL_STOP_STOPPED 0x00000001 /* job control stop in effect */
717#define SIGNAL_STOP_CONTINUED 0x00000002 /* SIGCONT since WCONTINUED reap */
718#define SIGNAL_GROUP_EXIT 0x00000004 /* group exit in progress */
719#define SIGNAL_GROUP_COREDUMP 0x00000008 /* coredump in progress */
720/*
721 * Pending notifications to parent.
722 */
723#define SIGNAL_CLD_STOPPED 0x00000010
724#define SIGNAL_CLD_CONTINUED 0x00000020
725#define SIGNAL_CLD_MASK (SIGNAL_CLD_STOPPED|SIGNAL_CLD_CONTINUED)
726
727#define SIGNAL_UNKILLABLE 0x00000040 /* for init: ignore fatal signals */
728
729/* If true, all threads except ->group_exit_task have pending SIGKILL */
730static inline int signal_group_exit(const struct signal_struct *sig)
731{
732 return (sig->flags & SIGNAL_GROUP_EXIT) ||
733 (sig->group_exit_task != NULL);
734}
735
736/*
737 * Some day this will be a full-fledged user tracking system..
738 */
739struct user_struct {
740 atomic_t __count; /* reference count */
741 atomic_t processes; /* How many processes does this user have? */
742 atomic_t sigpending; /* How many pending signals does this user have? */
743#ifdef CONFIG_INOTIFY_USER
744 atomic_t inotify_watches; /* How many inotify watches does this user have? */
745 atomic_t inotify_devs; /* How many inotify devs does this user have opened? */
746#endif
747#ifdef CONFIG_FANOTIFY
748 atomic_t fanotify_listeners;
749#endif
750#ifdef CONFIG_EPOLL
751 atomic_long_t epoll_watches; /* The number of file descriptors currently watched */
752#endif
753#ifdef CONFIG_POSIX_MQUEUE
754 /* protected by mq_lock */
755 unsigned long mq_bytes; /* How many bytes can be allocated to mqueue? */
756#endif
757 unsigned long locked_shm; /* How many pages of mlocked shm ? */
758
759#ifdef CONFIG_KEYS
760 struct key *uid_keyring; /* UID specific keyring */
761 struct key *session_keyring; /* UID's default session keyring */
762#endif
763
764 /* Hash table maintenance information */
765 struct hlist_node uidhash_node;
766 kuid_t uid;
767
768#ifdef CONFIG_PERF_EVENTS
769 atomic_long_t locked_vm;
770#endif
771};
772
773extern int uids_sysfs_init(void);
774
775extern struct user_struct *find_user(kuid_t);
776
777extern struct user_struct root_user;
778#define INIT_USER (&root_user)
779
780
781struct backing_dev_info;
782struct reclaim_state;
783
784#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
785struct sched_info {
786 /* cumulative counters */
787 unsigned long pcount; /* # of times run on this cpu */
788 unsigned long long run_delay; /* time spent waiting on a runqueue */
789
790 /* timestamps */
791 unsigned long long last_arrival,/* when we last ran on a cpu */
792 last_queued; /* when we were last queued to run */
793};
794#endif /* defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) */
795
796#ifdef CONFIG_TASK_DELAY_ACCT
797struct task_delay_info {
798 spinlock_t lock;
799 unsigned int flags; /* Private per-task flags */
800
801 /* For each stat XXX, add following, aligned appropriately
802 *
803 * struct timespec XXX_start, XXX_end;
804 * u64 XXX_delay;
805 * u32 XXX_count;
806 *
807 * Atomicity of updates to XXX_delay, XXX_count protected by
808 * single lock above (split into XXX_lock if contention is an issue).
809 */
810
811 /*
812 * XXX_count is incremented on every XXX operation, the delay
813 * associated with the operation is added to XXX_delay.
814 * XXX_delay contains the accumulated delay time in nanoseconds.
815 */
816 u64 blkio_start; /* Shared by blkio, swapin */
817 u64 blkio_delay; /* wait for sync block io completion */
818 u64 swapin_delay; /* wait for swapin block io completion */
819 u32 blkio_count; /* total count of the number of sync block */
820 /* io operations performed */
821 u32 swapin_count; /* total count of the number of swapin block */
822 /* io operations performed */
823
824 u64 freepages_start;
825 u64 freepages_delay; /* wait for memory reclaim */
826 u32 freepages_count; /* total count of memory reclaim */
827};
828#endif /* CONFIG_TASK_DELAY_ACCT */
829
830static inline int sched_info_on(void)
831{
832#ifdef CONFIG_SCHEDSTATS
833 return 1;
834#elif defined(CONFIG_TASK_DELAY_ACCT)
835 extern int delayacct_on;
836 return delayacct_on;
837#else
838 return 0;
839#endif
840}
841
842enum cpu_idle_type {
843 CPU_IDLE,
844 CPU_NOT_IDLE,
845 CPU_NEWLY_IDLE,
846 CPU_MAX_IDLE_TYPES
847};
848
849/*
850 * Increase resolution of cpu_capacity calculations
851 */
852#define SCHED_CAPACITY_SHIFT 10
853#define SCHED_CAPACITY_SCALE (1L << SCHED_CAPACITY_SHIFT)
854
855/*
856 * sched-domains (multiprocessor balancing) declarations:
857 */
858#ifdef CONFIG_SMP
859#define SD_LOAD_BALANCE 0x0001 /* Do load balancing on this domain. */
860#define SD_BALANCE_NEWIDLE 0x0002 /* Balance when about to become idle */
861#define SD_BALANCE_EXEC 0x0004 /* Balance on exec */
862#define SD_BALANCE_FORK 0x0008 /* Balance on fork, clone */
863#define SD_BALANCE_WAKE 0x0010 /* Balance on wakeup */
864#define SD_WAKE_AFFINE 0x0020 /* Wake task to waking CPU */
865#define SD_SHARE_CPUCAPACITY 0x0080 /* Domain members share cpu power */
866#define SD_SHARE_POWERDOMAIN 0x0100 /* Domain members share power domain */
867#define SD_SHARE_PKG_RESOURCES 0x0200 /* Domain members share cpu pkg resources */
868#define SD_SERIALIZE 0x0400 /* Only a single load balancing instance */
869#define SD_ASYM_PACKING 0x0800 /* Place busy groups earlier in the domain */
870#define SD_PREFER_SIBLING 0x1000 /* Prefer to place tasks in a sibling domain */
871#define SD_OVERLAP 0x2000 /* sched_domains of this level overlap */
872#define SD_NUMA 0x4000 /* cross-node balancing */
873
874#ifdef CONFIG_SCHED_SMT
875static inline int cpu_smt_flags(void)
876{
877 return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
878}
879#endif
880
881#ifdef CONFIG_SCHED_MC
882static inline int cpu_core_flags(void)
883{
884 return SD_SHARE_PKG_RESOURCES;
885}
886#endif
887
888#ifdef CONFIG_NUMA
889static inline int cpu_numa_flags(void)
890{
891 return SD_NUMA;
892}
893#endif
894
895struct sched_domain_attr {
896 int relax_domain_level;
897};
898
899#define SD_ATTR_INIT (struct sched_domain_attr) { \
900 .relax_domain_level = -1, \
901}
902
903extern int sched_domain_level_max;
904
905struct sched_group;
906
907struct sched_domain {
908 /* These fields must be setup */
909 struct sched_domain *parent; /* top domain must be null terminated */
910 struct sched_domain *child; /* bottom domain must be null terminated */
911 struct sched_group *groups; /* the balancing groups of the domain */
912 unsigned long min_interval; /* Minimum balance interval ms */
913 unsigned long max_interval; /* Maximum balance interval ms */
914 unsigned int busy_factor; /* less balancing by factor if busy */
915 unsigned int imbalance_pct; /* No balance until over watermark */
916 unsigned int cache_nice_tries; /* Leave cache hot tasks for # tries */
917 unsigned int busy_idx;
918 unsigned int idle_idx;
919 unsigned int newidle_idx;
920 unsigned int wake_idx;
921 unsigned int forkexec_idx;
922 unsigned int smt_gain;
923
924 int nohz_idle; /* NOHZ IDLE status */
925 int flags; /* See SD_* */
926 int level;
927
928 /* Runtime fields. */
929 unsigned long last_balance; /* init to jiffies. units in jiffies */
930 unsigned int balance_interval; /* initialise to 1. units in ms. */
931 unsigned int nr_balance_failed; /* initialise to 0 */
932
933 /* idle_balance() stats */
934 u64 max_newidle_lb_cost;
935 unsigned long next_decay_max_lb_cost;
936
937#ifdef CONFIG_SCHEDSTATS
938 /* load_balance() stats */
939 unsigned int lb_count[CPU_MAX_IDLE_TYPES];
940 unsigned int lb_failed[CPU_MAX_IDLE_TYPES];
941 unsigned int lb_balanced[CPU_MAX_IDLE_TYPES];
942 unsigned int lb_imbalance[CPU_MAX_IDLE_TYPES];
943 unsigned int lb_gained[CPU_MAX_IDLE_TYPES];
944 unsigned int lb_hot_gained[CPU_MAX_IDLE_TYPES];
945 unsigned int lb_nobusyg[CPU_MAX_IDLE_TYPES];
946 unsigned int lb_nobusyq[CPU_MAX_IDLE_TYPES];
947
948 /* Active load balancing */
949 unsigned int alb_count;
950 unsigned int alb_failed;
951 unsigned int alb_pushed;
952
953 /* SD_BALANCE_EXEC stats */
954 unsigned int sbe_count;
955 unsigned int sbe_balanced;
956 unsigned int sbe_pushed;
957
958 /* SD_BALANCE_FORK stats */
959 unsigned int sbf_count;
960 unsigned int sbf_balanced;
961 unsigned int sbf_pushed;
962
963 /* try_to_wake_up() stats */
964 unsigned int ttwu_wake_remote;
965 unsigned int ttwu_move_affine;
966 unsigned int ttwu_move_balance;
967#endif
968#ifdef CONFIG_SCHED_DEBUG
969 char *name;
970#endif
971 union {
972 void *private; /* used during construction */
973 struct rcu_head rcu; /* used during destruction */
974 };
975
976 unsigned int span_weight;
977 /*
978 * Span of all CPUs in this domain.
979 *
980 * NOTE: this field is variable length. (Allocated dynamically
981 * by attaching extra space to the end of the structure,
982 * depending on how many CPUs the kernel has booted up with)
983 */
984 unsigned long span[0];
985};
986
987static inline struct cpumask *sched_domain_span(struct sched_domain *sd)
988{
989 return to_cpumask(sd->span);
990}
991
992extern void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
993 struct sched_domain_attr *dattr_new);
994
995/* Allocate an array of sched domains, for partition_sched_domains(). */
996cpumask_var_t *alloc_sched_domains(unsigned int ndoms);
997void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
998
999bool cpus_share_cache(int this_cpu, int that_cpu);
1000
1001typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
1002typedef int (*sched_domain_flags_f)(void);
1003
1004#define SDTL_OVERLAP 0x01
1005
1006struct sd_data {
1007 struct sched_domain **__percpu sd;
1008 struct sched_group **__percpu sg;
1009 struct sched_group_capacity **__percpu sgc;
1010};
1011
1012struct sched_domain_topology_level {
1013 sched_domain_mask_f mask;
1014 sched_domain_flags_f sd_flags;
1015 int flags;
1016 int numa_level;
1017 struct sd_data data;
1018#ifdef CONFIG_SCHED_DEBUG
1019 char *name;
1020#endif
1021};
1022
1023extern struct sched_domain_topology_level *sched_domain_topology;
1024
1025extern void set_sched_topology(struct sched_domain_topology_level *tl);
1026
1027#ifdef CONFIG_SCHED_DEBUG
1028# define SD_INIT_NAME(type) .name = #type
1029#else
1030# define SD_INIT_NAME(type)
1031#endif
1032
1033#else /* CONFIG_SMP */
1034
1035struct sched_domain_attr;
1036
1037static inline void
1038partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
1039 struct sched_domain_attr *dattr_new)
1040{
1041}
1042
1043static inline bool cpus_share_cache(int this_cpu, int that_cpu)
1044{
1045 return true;
1046}
1047
1048#endif /* !CONFIG_SMP */
1049
1050
1051struct io_context; /* See blkdev.h */
1052
1053
1054#ifdef ARCH_HAS_PREFETCH_SWITCH_STACK
1055extern void prefetch_stack(struct task_struct *t);
1056#else
1057static inline void prefetch_stack(struct task_struct *t) { }
1058#endif
1059
1060struct audit_context; /* See audit.c */
1061struct mempolicy;
1062struct pipe_inode_info;
1063struct uts_namespace;
1064
1065struct load_weight {
1066 unsigned long weight;
1067 u32 inv_weight;
1068};
1069
1070struct sched_avg {
1071 /*
1072 * These sums represent an infinite geometric series and so are bound
1073 * above by 1024/(1-y). Thus we only need a u32 to store them for all
1074 * choices of y < 1-2^(-32)*1024.
1075 */
1076 u32 runnable_avg_sum, runnable_avg_period;
1077 u64 last_runnable_update;
1078 s64 decay_count;
1079 unsigned long load_avg_contrib;
1080};
1081
1082#ifdef CONFIG_SCHEDSTATS
1083struct sched_statistics {
1084 u64 wait_start;
1085 u64 wait_max;
1086 u64 wait_count;
1087 u64 wait_sum;
1088 u64 iowait_count;
1089 u64 iowait_sum;
1090
1091 u64 sleep_start;
1092 u64 sleep_max;
1093 s64 sum_sleep_runtime;
1094
1095 u64 block_start;
1096 u64 block_max;
1097 u64 exec_max;
1098 u64 slice_max;
1099
1100 u64 nr_migrations_cold;
1101 u64 nr_failed_migrations_affine;
1102 u64 nr_failed_migrations_running;
1103 u64 nr_failed_migrations_hot;
1104 u64 nr_forced_migrations;
1105
1106 u64 nr_wakeups;
1107 u64 nr_wakeups_sync;
1108 u64 nr_wakeups_migrate;
1109 u64 nr_wakeups_local;
1110 u64 nr_wakeups_remote;
1111 u64 nr_wakeups_affine;
1112 u64 nr_wakeups_affine_attempts;
1113 u64 nr_wakeups_passive;
1114 u64 nr_wakeups_idle;
1115};
1116#endif
1117
1118struct sched_entity {
1119 struct load_weight load; /* for load-balancing */
1120 struct rb_node run_node;
1121 struct list_head group_node;
1122 unsigned int on_rq;
1123
1124 u64 exec_start;
1125 u64 sum_exec_runtime;
1126 u64 vruntime;
1127 u64 prev_sum_exec_runtime;
1128
1129 u64 nr_migrations;
1130
1131#ifdef CONFIG_SCHEDSTATS
1132 struct sched_statistics statistics;
1133#endif
1134
1135#ifdef CONFIG_FAIR_GROUP_SCHED
1136 int depth;
1137 struct sched_entity *parent;
1138 /* rq on which this entity is (to be) queued: */
1139 struct cfs_rq *cfs_rq;
1140 /* rq "owned" by this entity/group: */
1141 struct cfs_rq *my_q;
1142#endif
1143
1144#ifdef CONFIG_SMP
1145 /* Per-entity load-tracking */
1146 struct sched_avg avg;
1147#endif
1148};
1149
1150struct sched_rt_entity {
1151 struct list_head run_list;
1152 unsigned long timeout;
1153 unsigned long watchdog_stamp;
1154 unsigned int time_slice;
1155
1156 struct sched_rt_entity *back;
1157#ifdef CONFIG_RT_GROUP_SCHED
1158 struct sched_rt_entity *parent;
1159 /* rq on which this entity is (to be) queued: */
1160 struct rt_rq *rt_rq;
1161 /* rq "owned" by this entity/group: */
1162 struct rt_rq *my_q;
1163#endif
1164};
1165
1166struct sched_dl_entity {
1167 struct rb_node rb_node;
1168
1169 /*
1170 * Original scheduling parameters. Copied here from sched_attr
1171 * during sched_setattr(), they will remain the same until
1172 * the next sched_setattr().
1173 */
1174 u64 dl_runtime; /* maximum runtime for each instance */
1175 u64 dl_deadline; /* relative deadline of each instance */
1176 u64 dl_period; /* separation of two instances (period) */
1177 u64 dl_bw; /* dl_runtime / dl_deadline */
1178
1179 /*
1180 * Actual scheduling parameters. Initialized with the values above,
1181 * they are continously updated during task execution. Note that
1182 * the remaining runtime could be < 0 in case we are in overrun.
1183 */
1184 s64 runtime; /* remaining runtime for this instance */
1185 u64 deadline; /* absolute deadline for this instance */
1186 unsigned int flags; /* specifying the scheduler behaviour */
1187
1188 /*
1189 * Some bool flags:
1190 *
1191 * @dl_throttled tells if we exhausted the runtime. If so, the
1192 * task has to wait for a replenishment to be performed at the
1193 * next firing of dl_timer.
1194 *
1195 * @dl_new tells if a new instance arrived. If so we must
1196 * start executing it with full runtime and reset its absolute
1197 * deadline;
1198 *
1199 * @dl_boosted tells if we are boosted due to DI. If so we are
1200 * outside bandwidth enforcement mechanism (but only until we
1201 * exit the critical section);
1202 *
1203 * @dl_yielded tells if task gave up the cpu before consuming
1204 * all its available runtime during the last job.
1205 */
1206 int dl_throttled, dl_new, dl_boosted, dl_yielded;
1207
1208 /*
1209 * Bandwidth enforcement timer. Each -deadline task has its
1210 * own bandwidth to be enforced, thus we need one timer per task.
1211 */
1212 struct hrtimer dl_timer;
1213};
1214
1215struct rcu_node;
1216
1217enum perf_event_task_context {
1218 perf_invalid_context = -1,
1219 perf_hw_context = 0,
1220 perf_sw_context,
1221 perf_nr_task_contexts,
1222};
1223
1224struct task_struct {
1225 volatile long state; /* -1 unrunnable, 0 runnable, >0 stopped */
1226 void *stack;
1227 atomic_t usage;
1228 unsigned int flags; /* per process flags, defined below */
1229 unsigned int ptrace;
1230
1231#ifdef CONFIG_SMP
1232 struct llist_node wake_entry;
1233 int on_cpu;
1234 struct task_struct *last_wakee;
1235 unsigned long wakee_flips;
1236 unsigned long wakee_flip_decay_ts;
1237
1238 int wake_cpu;
1239#endif
1240 int on_rq;
1241
1242 int prio, static_prio, normal_prio;
1243 unsigned int rt_priority;
1244 const struct sched_class *sched_class;
1245 struct sched_entity se;
1246 struct sched_rt_entity rt;
1247#ifdef CONFIG_CGROUP_SCHED
1248 struct task_group *sched_task_group;
1249#endif
1250 struct sched_dl_entity dl;
1251
1252#ifdef CONFIG_PREEMPT_NOTIFIERS
1253 /* list of struct preempt_notifier: */
1254 struct hlist_head preempt_notifiers;
1255#endif
1256
1257#ifdef CONFIG_BLK_DEV_IO_TRACE
1258 unsigned int btrace_seq;
1259#endif
1260
1261 unsigned int policy;
1262 int nr_cpus_allowed;
1263 cpumask_t cpus_allowed;
1264
1265#ifdef CONFIG_PREEMPT_RCU
1266 int rcu_read_lock_nesting;
1267 char rcu_read_unlock_special;
1268 struct list_head rcu_node_entry;
1269#endif /* #ifdef CONFIG_PREEMPT_RCU */
1270#ifdef CONFIG_TREE_PREEMPT_RCU
1271 struct rcu_node *rcu_blocked_node;
1272#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
1273
1274#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
1275 struct sched_info sched_info;
1276#endif
1277
1278 struct list_head tasks;
1279#ifdef CONFIG_SMP
1280 struct plist_node pushable_tasks;
1281 struct rb_node pushable_dl_tasks;
1282#endif
1283
1284 struct mm_struct *mm, *active_mm;
1285#ifdef CONFIG_COMPAT_BRK
1286 unsigned brk_randomized:1;
1287#endif
1288 /* per-thread vma caching */
1289 u32 vmacache_seqnum;
1290 struct vm_area_struct *vmacache[VMACACHE_SIZE];
1291#if defined(SPLIT_RSS_COUNTING)
1292 struct task_rss_stat rss_stat;
1293#endif
1294/* task state */
1295 int exit_state;
1296 int exit_code, exit_signal;
1297 int pdeath_signal; /* The signal sent when the parent dies */
1298 unsigned int jobctl; /* JOBCTL_*, siglock protected */
1299
1300 /* Used for emulating ABI behavior of previous Linux versions */
1301 unsigned int personality;
1302
1303 unsigned in_execve:1; /* Tell the LSMs that the process is doing an
1304 * execve */
1305 unsigned in_iowait:1;
1306
1307 /* Revert to default priority/policy when forking */
1308 unsigned sched_reset_on_fork:1;
1309 unsigned sched_contributes_to_load:1;
1310
1311 unsigned long atomic_flags; /* Flags needing atomic access. */
1312
1313 pid_t pid;
1314 pid_t tgid;
1315
1316#ifdef CONFIG_CC_STACKPROTECTOR
1317 /* Canary value for the -fstack-protector gcc feature */
1318 unsigned long stack_canary;
1319#endif
1320 /*
1321 * pointers to (original) parent process, youngest child, younger sibling,
1322 * older sibling, respectively. (p->father can be replaced with
1323 * p->real_parent->pid)
1324 */
1325 struct task_struct __rcu *real_parent; /* real parent process */
1326 struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
1327 /*
1328 * children/sibling forms the list of my natural children
1329 */
1330 struct list_head children; /* list of my children */
1331 struct list_head sibling; /* linkage in my parent's children list */
1332 struct task_struct *group_leader; /* threadgroup leader */
1333
1334 /*
1335 * ptraced is the list of tasks this task is using ptrace on.
1336 * This includes both natural children and PTRACE_ATTACH targets.
1337 * p->ptrace_entry is p's link on the p->parent->ptraced list.
1338 */
1339 struct list_head ptraced;
1340 struct list_head ptrace_entry;
1341
1342 /* PID/PID hash table linkage. */
1343 struct pid_link pids[PIDTYPE_MAX];
1344 struct list_head thread_group;
1345 struct list_head thread_node;
1346
1347 struct completion *vfork_done; /* for vfork() */
1348 int __user *set_child_tid; /* CLONE_CHILD_SETTID */
1349 int __user *clear_child_tid; /* CLONE_CHILD_CLEARTID */
1350
1351 cputime_t utime, stime, utimescaled, stimescaled;
1352 cputime_t gtime;
1353#ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
1354 struct cputime prev_cputime;
1355#endif
1356#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1357 seqlock_t vtime_seqlock;
1358 unsigned long long vtime_snap;
1359 enum {
1360 VTIME_SLEEPING = 0,
1361 VTIME_USER,
1362 VTIME_SYS,
1363 } vtime_snap_whence;
1364#endif
1365 unsigned long nvcsw, nivcsw; /* context switch counts */
1366 u64 start_time; /* monotonic time in nsec */
1367 u64 real_start_time; /* boot based time in nsec */
1368/* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
1369 unsigned long min_flt, maj_flt;
1370
1371 struct task_cputime cputime_expires;
1372 struct list_head cpu_timers[3];
1373
1374/* process credentials */
1375 const struct cred __rcu *real_cred; /* objective and real subjective task
1376 * credentials (COW) */
1377 const struct cred __rcu *cred; /* effective (overridable) subjective task
1378 * credentials (COW) */
1379 char comm[TASK_COMM_LEN]; /* executable name excluding path
1380 - access with [gs]et_task_comm (which lock
1381 it with task_lock())
1382 - initialized normally by setup_new_exec */
1383/* file system info */
1384 int link_count, total_link_count;
1385#ifdef CONFIG_SYSVIPC
1386/* ipc stuff */
1387 struct sysv_sem sysvsem;
1388 struct sysv_shm sysvshm;
1389#endif
1390#ifdef CONFIG_DETECT_HUNG_TASK
1391/* hung task detection */
1392 unsigned long last_switch_count;
1393#endif
1394/* CPU-specific state of this task */
1395 struct thread_struct thread;
1396/* filesystem information */
1397 struct fs_struct *fs;
1398/* open file information */
1399 struct files_struct *files;
1400/* namespaces */
1401 struct nsproxy *nsproxy;
1402/* signal handlers */
1403 struct signal_struct *signal;
1404 struct sighand_struct *sighand;
1405
1406 sigset_t blocked, real_blocked;
1407 sigset_t saved_sigmask; /* restored if set_restore_sigmask() was used */
1408 struct sigpending pending;
1409
1410 unsigned long sas_ss_sp;
1411 size_t sas_ss_size;
1412 int (*notifier)(void *priv);
1413 void *notifier_data;
1414 sigset_t *notifier_mask;
1415 struct callback_head *task_works;
1416
1417 struct audit_context *audit_context;
1418#ifdef CONFIG_AUDITSYSCALL
1419 kuid_t loginuid;
1420 unsigned int sessionid;
1421#endif
1422 struct seccomp seccomp;
1423
1424/* Thread group tracking */
1425 u32 parent_exec_id;
1426 u32 self_exec_id;
1427/* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
1428 * mempolicy */
1429 spinlock_t alloc_lock;
1430
1431 /* Protection of the PI data structures: */
1432 raw_spinlock_t pi_lock;
1433
1434#ifdef CONFIG_RT_MUTEXES
1435 /* PI waiters blocked on a rt_mutex held by this task */
1436 struct rb_root pi_waiters;
1437 struct rb_node *pi_waiters_leftmost;
1438 /* Deadlock detection and priority inheritance handling */
1439 struct rt_mutex_waiter *pi_blocked_on;
1440#endif
1441
1442#ifdef CONFIG_DEBUG_MUTEXES
1443 /* mutex deadlock detection */
1444 struct mutex_waiter *blocked_on;
1445#endif
1446#ifdef CONFIG_TRACE_IRQFLAGS
1447 unsigned int irq_events;
1448 unsigned long hardirq_enable_ip;
1449 unsigned long hardirq_disable_ip;
1450 unsigned int hardirq_enable_event;
1451 unsigned int hardirq_disable_event;
1452 int hardirqs_enabled;
1453 int hardirq_context;
1454 unsigned long softirq_disable_ip;
1455 unsigned long softirq_enable_ip;
1456 unsigned int softirq_disable_event;
1457 unsigned int softirq_enable_event;
1458 int softirqs_enabled;
1459 int softirq_context;
1460#endif
1461#ifdef CONFIG_LOCKDEP
1462# define MAX_LOCK_DEPTH 48UL
1463 u64 curr_chain_key;
1464 int lockdep_depth;
1465 unsigned int lockdep_recursion;
1466 struct held_lock held_locks[MAX_LOCK_DEPTH];
1467 gfp_t lockdep_reclaim_gfp;
1468#endif
1469
1470/* journalling filesystem info */
1471 void *journal_info;
1472
1473/* stacked block device info */
1474 struct bio_list *bio_list;
1475
1476#ifdef CONFIG_BLOCK
1477/* stack plugging */
1478 struct blk_plug *plug;
1479#endif
1480
1481/* VM state */
1482 struct reclaim_state *reclaim_state;
1483
1484 struct backing_dev_info *backing_dev_info;
1485
1486 struct io_context *io_context;
1487
1488 unsigned long ptrace_message;
1489 siginfo_t *last_siginfo; /* For ptrace use. */
1490 struct task_io_accounting ioac;
1491#if defined(CONFIG_TASK_XACCT)
1492 u64 acct_rss_mem1; /* accumulated rss usage */
1493 u64 acct_vm_mem1; /* accumulated virtual memory usage */
1494 cputime_t acct_timexpd; /* stime + utime since last update */
1495#endif
1496#ifdef CONFIG_CPUSETS
1497 nodemask_t mems_allowed; /* Protected by alloc_lock */
1498 seqcount_t mems_allowed_seq; /* Seqence no to catch updates */
1499 int cpuset_mem_spread_rotor;
1500 int cpuset_slab_spread_rotor;
1501#endif
1502#ifdef CONFIG_CGROUPS
1503 /* Control Group info protected by css_set_lock */
1504 struct css_set __rcu *cgroups;
1505 /* cg_list protected by css_set_lock and tsk->alloc_lock */
1506 struct list_head cg_list;
1507#endif
1508#ifdef CONFIG_FUTEX
1509 struct robust_list_head __user *robust_list;
1510#ifdef CONFIG_COMPAT
1511 struct compat_robust_list_head __user *compat_robust_list;
1512#endif
1513 struct list_head pi_state_list;
1514 struct futex_pi_state *pi_state_cache;
1515#endif
1516#ifdef CONFIG_PERF_EVENTS
1517 struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
1518 struct mutex perf_event_mutex;
1519 struct list_head perf_event_list;
1520#endif
1521#ifdef CONFIG_DEBUG_PREEMPT
1522 unsigned long preempt_disable_ip;
1523#endif
1524#ifdef CONFIG_NUMA
1525 struct mempolicy *mempolicy; /* Protected by alloc_lock */
1526 short il_next;
1527 short pref_node_fork;
1528#endif
1529#ifdef CONFIG_NUMA_BALANCING
1530 int numa_scan_seq;
1531 unsigned int numa_scan_period;
1532 unsigned int numa_scan_period_max;
1533 int numa_preferred_nid;
1534 unsigned long numa_migrate_retry;
1535 u64 node_stamp; /* migration stamp */
1536 u64 last_task_numa_placement;
1537 u64 last_sum_exec_runtime;
1538 struct callback_head numa_work;
1539
1540 struct list_head numa_entry;
1541 struct numa_group *numa_group;
1542
1543 /*
1544 * Exponential decaying average of faults on a per-node basis.
1545 * Scheduling placement decisions are made based on the these counts.
1546 * The values remain static for the duration of a PTE scan
1547 */
1548 unsigned long *numa_faults_memory;
1549 unsigned long total_numa_faults;
1550
1551 /*
1552 * numa_faults_buffer records faults per node during the current
1553 * scan window. When the scan completes, the counts in
1554 * numa_faults_memory decay and these values are copied.
1555 */
1556 unsigned long *numa_faults_buffer_memory;
1557
1558 /*
1559 * Track the nodes the process was running on when a NUMA hinting
1560 * fault was incurred.
1561 */
1562 unsigned long *numa_faults_cpu;
1563 unsigned long *numa_faults_buffer_cpu;
1564
1565 /*
1566 * numa_faults_locality tracks if faults recorded during the last
1567 * scan window were remote/local. The task scan period is adapted
1568 * based on the locality of the faults with different weights
1569 * depending on whether they were shared or private faults
1570 */
1571 unsigned long numa_faults_locality[2];
1572
1573 unsigned long numa_pages_migrated;
1574#endif /* CONFIG_NUMA_BALANCING */
1575
1576 struct rcu_head rcu;
1577
1578 /*
1579 * cache last used pipe for splice
1580 */
1581 struct pipe_inode_info *splice_pipe;
1582
1583 struct page_frag task_frag;
1584
1585#ifdef CONFIG_TASK_DELAY_ACCT
1586 struct task_delay_info *delays;
1587#endif
1588#ifdef CONFIG_FAULT_INJECTION
1589 int make_it_fail;
1590#endif
1591 /*
1592 * when (nr_dirtied >= nr_dirtied_pause), it's time to call
1593 * balance_dirty_pages() for some dirty throttling pause
1594 */
1595 int nr_dirtied;
1596 int nr_dirtied_pause;
1597 unsigned long dirty_paused_when; /* start of a write-and-pause period */
1598
1599#ifdef CONFIG_LATENCYTOP
1600 int latency_record_count;
1601 struct latency_record latency_record[LT_SAVECOUNT];
1602#endif
1603 /*
1604 * time slack values; these are used to round up poll() and
1605 * select() etc timeout values. These are in nanoseconds.
1606 */
1607 unsigned long timer_slack_ns;
1608 unsigned long default_timer_slack_ns;
1609
1610#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1611 /* Index of current stored address in ret_stack */
1612 int curr_ret_stack;
1613 /* Stack of return addresses for return function tracing */
1614 struct ftrace_ret_stack *ret_stack;
1615 /* time stamp for last schedule */
1616 unsigned long long ftrace_timestamp;
1617 /*
1618 * Number of functions that haven't been traced
1619 * because of depth overrun.
1620 */
1621 atomic_t trace_overrun;
1622 /* Pause for the tracing */
1623 atomic_t tracing_graph_pause;
1624#endif
1625#ifdef CONFIG_TRACING
1626 /* state flags for use by tracers */
1627 unsigned long trace;
1628 /* bitmask and counter of trace recursion */
1629 unsigned long trace_recursion;
1630#endif /* CONFIG_TRACING */
1631#ifdef CONFIG_MEMCG /* memcg uses this to do batch job */
1632 unsigned int memcg_kmem_skip_account;
1633 struct memcg_oom_info {
1634 struct mem_cgroup *memcg;
1635 gfp_t gfp_mask;
1636 int order;
1637 unsigned int may_oom:1;
1638 } memcg_oom;
1639#endif
1640#ifdef CONFIG_UPROBES
1641 struct uprobe_task *utask;
1642#endif
1643#if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
1644 unsigned int sequential_io;
1645 unsigned int sequential_io_avg;
1646#endif
1647};
1648
1649/* Future-safe accessor for struct task_struct's cpus_allowed. */
1650#define tsk_cpus_allowed(tsk) (&(tsk)->cpus_allowed)
1651
1652#define TNF_MIGRATED 0x01
1653#define TNF_NO_GROUP 0x02
1654#define TNF_SHARED 0x04
1655#define TNF_FAULT_LOCAL 0x08
1656
1657#ifdef CONFIG_NUMA_BALANCING
1658extern void task_numa_fault(int last_node, int node, int pages, int flags);
1659extern pid_t task_numa_group_id(struct task_struct *p);
1660extern void set_numabalancing_state(bool enabled);
1661extern void task_numa_free(struct task_struct *p);
1662extern bool should_numa_migrate_memory(struct task_struct *p, struct page *page,
1663 int src_nid, int dst_cpu);
1664#else
1665static inline void task_numa_fault(int last_node, int node, int pages,
1666 int flags)
1667{
1668}
1669static inline pid_t task_numa_group_id(struct task_struct *p)
1670{
1671 return 0;
1672}
1673static inline void set_numabalancing_state(bool enabled)
1674{
1675}
1676static inline void task_numa_free(struct task_struct *p)
1677{
1678}
1679static inline bool should_numa_migrate_memory(struct task_struct *p,
1680 struct page *page, int src_nid, int dst_cpu)
1681{
1682 return true;
1683}
1684#endif
1685
1686static inline struct pid *task_pid(struct task_struct *task)
1687{
1688 return task->pids[PIDTYPE_PID].pid;
1689}
1690
1691static inline struct pid *task_tgid(struct task_struct *task)
1692{
1693 return task->group_leader->pids[PIDTYPE_PID].pid;
1694}
1695
1696/*
1697 * Without tasklist or rcu lock it is not safe to dereference
1698 * the result of task_pgrp/task_session even if task == current,
1699 * we can race with another thread doing sys_setsid/sys_setpgid.
1700 */
1701static inline struct pid *task_pgrp(struct task_struct *task)
1702{
1703 return task->group_leader->pids[PIDTYPE_PGID].pid;
1704}
1705
1706static inline struct pid *task_session(struct task_struct *task)
1707{
1708 return task->group_leader->pids[PIDTYPE_SID].pid;
1709}
1710
1711struct pid_namespace;
1712
1713/*
1714 * the helpers to get the task's different pids as they are seen
1715 * from various namespaces
1716 *
1717 * task_xid_nr() : global id, i.e. the id seen from the init namespace;
1718 * task_xid_vnr() : virtual id, i.e. the id seen from the pid namespace of
1719 * current.
1720 * task_xid_nr_ns() : id seen from the ns specified;
1721 *
1722 * set_task_vxid() : assigns a virtual id to a task;
1723 *
1724 * see also pid_nr() etc in include/linux/pid.h
1725 */
1726pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
1727 struct pid_namespace *ns);
1728
1729static inline pid_t task_pid_nr(struct task_struct *tsk)
1730{
1731 return tsk->pid;
1732}
1733
1734static inline pid_t task_pid_nr_ns(struct task_struct *tsk,
1735 struct pid_namespace *ns)
1736{
1737 return __task_pid_nr_ns(tsk, PIDTYPE_PID, ns);
1738}
1739
1740static inline pid_t task_pid_vnr(struct task_struct *tsk)
1741{
1742 return __task_pid_nr_ns(tsk, PIDTYPE_PID, NULL);
1743}
1744
1745
1746static inline pid_t task_tgid_nr(struct task_struct *tsk)
1747{
1748 return tsk->tgid;
1749}
1750
1751pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns);
1752
1753static inline pid_t task_tgid_vnr(struct task_struct *tsk)
1754{
1755 return pid_vnr(task_tgid(tsk));
1756}
1757
1758
1759static inline int pid_alive(const struct task_struct *p);
1760static inline pid_t task_ppid_nr_ns(const struct task_struct *tsk, struct pid_namespace *ns)
1761{
1762 pid_t pid = 0;
1763
1764 rcu_read_lock();
1765 if (pid_alive(tsk))
1766 pid = task_tgid_nr_ns(rcu_dereference(tsk->real_parent), ns);
1767 rcu_read_unlock();
1768
1769 return pid;
1770}
1771
1772static inline pid_t task_ppid_nr(const struct task_struct *tsk)
1773{
1774 return task_ppid_nr_ns(tsk, &init_pid_ns);
1775}
1776
1777static inline pid_t task_pgrp_nr_ns(struct task_struct *tsk,
1778 struct pid_namespace *ns)
1779{
1780 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, ns);
1781}
1782
1783static inline pid_t task_pgrp_vnr(struct task_struct *tsk)
1784{
1785 return __task_pid_nr_ns(tsk, PIDTYPE_PGID, NULL);
1786}
1787
1788
1789static inline pid_t task_session_nr_ns(struct task_struct *tsk,
1790 struct pid_namespace *ns)
1791{
1792 return __task_pid_nr_ns(tsk, PIDTYPE_SID, ns);
1793}
1794
1795static inline pid_t task_session_vnr(struct task_struct *tsk)
1796{
1797 return __task_pid_nr_ns(tsk, PIDTYPE_SID, NULL);
1798}
1799
1800/* obsolete, do not use */
1801static inline pid_t task_pgrp_nr(struct task_struct *tsk)
1802{
1803 return task_pgrp_nr_ns(tsk, &init_pid_ns);
1804}
1805
1806/**
1807 * pid_alive - check that a task structure is not stale
1808 * @p: Task structure to be checked.
1809 *
1810 * Test if a process is not yet dead (at most zombie state)
1811 * If pid_alive fails, then pointers within the task structure
1812 * can be stale and must not be dereferenced.
1813 *
1814 * Return: 1 if the process is alive. 0 otherwise.
1815 */
1816static inline int pid_alive(const struct task_struct *p)
1817{
1818 return p->pids[PIDTYPE_PID].pid != NULL;
1819}
1820
1821/**
1822 * is_global_init - check if a task structure is init
1823 * @tsk: Task structure to be checked.
1824 *
1825 * Check if a task structure is the first user space task the kernel created.
1826 *
1827 * Return: 1 if the task structure is init. 0 otherwise.
1828 */
1829static inline int is_global_init(struct task_struct *tsk)
1830{
1831 return tsk->pid == 1;
1832}
1833
1834extern struct pid *cad_pid;
1835
1836extern void free_task(struct task_struct *tsk);
1837#define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
1838
1839extern void __put_task_struct(struct task_struct *t);
1840
1841static inline void put_task_struct(struct task_struct *t)
1842{
1843 if (atomic_dec_and_test(&t->usage))
1844 __put_task_struct(t);
1845}
1846
1847#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1848extern void task_cputime(struct task_struct *t,
1849 cputime_t *utime, cputime_t *stime);
1850extern void task_cputime_scaled(struct task_struct *t,
1851 cputime_t *utimescaled, cputime_t *stimescaled);
1852extern cputime_t task_gtime(struct task_struct *t);
1853#else
1854static inline void task_cputime(struct task_struct *t,
1855 cputime_t *utime, cputime_t *stime)
1856{
1857 if (utime)
1858 *utime = t->utime;
1859 if (stime)
1860 *stime = t->stime;
1861}
1862
1863static inline void task_cputime_scaled(struct task_struct *t,
1864 cputime_t *utimescaled,
1865 cputime_t *stimescaled)
1866{
1867 if (utimescaled)
1868 *utimescaled = t->utimescaled;
1869 if (stimescaled)
1870 *stimescaled = t->stimescaled;
1871}
1872
1873static inline cputime_t task_gtime(struct task_struct *t)
1874{
1875 return t->gtime;
1876}
1877#endif
1878extern void task_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
1879extern void thread_group_cputime_adjusted(struct task_struct *p, cputime_t *ut, cputime_t *st);
1880
1881/*
1882 * Per process flags
1883 */
1884#define PF_EXITING 0x00000004 /* getting shut down */
1885#define PF_EXITPIDONE 0x00000008 /* pi exit done on shut down */
1886#define PF_VCPU 0x00000010 /* I'm a virtual CPU */
1887#define PF_WQ_WORKER 0x00000020 /* I'm a workqueue worker */
1888#define PF_FORKNOEXEC 0x00000040 /* forked but didn't exec */
1889#define PF_MCE_PROCESS 0x00000080 /* process policy on mce errors */
1890#define PF_SUPERPRIV 0x00000100 /* used super-user privileges */
1891#define PF_DUMPCORE 0x00000200 /* dumped core */
1892#define PF_SIGNALED 0x00000400 /* killed by a signal */
1893#define PF_MEMALLOC 0x00000800 /* Allocating memory */
1894#define PF_NPROC_EXCEEDED 0x00001000 /* set_user noticed that RLIMIT_NPROC was exceeded */
1895#define PF_USED_MATH 0x00002000 /* if unset the fpu must be initialized before use */
1896#define PF_USED_ASYNC 0x00004000 /* used async_schedule*(), used by module init */
1897#define PF_NOFREEZE 0x00008000 /* this thread should not be frozen */
1898#define PF_FROZEN 0x00010000 /* frozen for system suspend */
1899#define PF_FSTRANS 0x00020000 /* inside a filesystem transaction */
1900#define PF_KSWAPD 0x00040000 /* I am kswapd */
1901#define PF_MEMALLOC_NOIO 0x00080000 /* Allocating memory without IO involved */
1902#define PF_LESS_THROTTLE 0x00100000 /* Throttle me less: I clean memory */
1903#define PF_KTHREAD 0x00200000 /* I am a kernel thread */
1904#define PF_RANDOMIZE 0x00400000 /* randomize virtual address space */
1905#define PF_SWAPWRITE 0x00800000 /* Allowed to write to swap */
1906#define PF_NO_SETAFFINITY 0x04000000 /* Userland is not allowed to meddle with cpus_allowed */
1907#define PF_MCE_EARLY 0x08000000 /* Early kill for mce process policy */
1908#define PF_MUTEX_TESTER 0x20000000 /* Thread belongs to the rt mutex tester */
1909#define PF_FREEZER_SKIP 0x40000000 /* Freezer should not count it as freezable */
1910#define PF_SUSPEND_TASK 0x80000000 /* this thread called freeze_processes and should not be frozen */
1911
1912/*
1913 * Only the _current_ task can read/write to tsk->flags, but other
1914 * tasks can access tsk->flags in readonly mode for example
1915 * with tsk_used_math (like during threaded core dumping).
1916 * There is however an exception to this rule during ptrace
1917 * or during fork: the ptracer task is allowed to write to the
1918 * child->flags of its traced child (same goes for fork, the parent
1919 * can write to the child->flags), because we're guaranteed the
1920 * child is not running and in turn not changing child->flags
1921 * at the same time the parent does it.
1922 */
1923#define clear_stopped_child_used_math(child) do { (child)->flags &= ~PF_USED_MATH; } while (0)
1924#define set_stopped_child_used_math(child) do { (child)->flags |= PF_USED_MATH; } while (0)
1925#define clear_used_math() clear_stopped_child_used_math(current)
1926#define set_used_math() set_stopped_child_used_math(current)
1927#define conditional_stopped_child_used_math(condition, child) \
1928 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= (condition) ? PF_USED_MATH : 0; } while (0)
1929#define conditional_used_math(condition) \
1930 conditional_stopped_child_used_math(condition, current)
1931#define copy_to_stopped_child_used_math(child) \
1932 do { (child)->flags &= ~PF_USED_MATH, (child)->flags |= current->flags & PF_USED_MATH; } while (0)
1933/* NOTE: this will return 0 or PF_USED_MATH, it will never return 1 */
1934#define tsk_used_math(p) ((p)->flags & PF_USED_MATH)
1935#define used_math() tsk_used_math(current)
1936
1937/* __GFP_IO isn't allowed if PF_MEMALLOC_NOIO is set in current->flags */
1938static inline gfp_t memalloc_noio_flags(gfp_t flags)
1939{
1940 if (unlikely(current->flags & PF_MEMALLOC_NOIO))
1941 flags &= ~__GFP_IO;
1942 return flags;
1943}
1944
1945static inline unsigned int memalloc_noio_save(void)
1946{
1947 unsigned int flags = current->flags & PF_MEMALLOC_NOIO;
1948 current->flags |= PF_MEMALLOC_NOIO;
1949 return flags;
1950}
1951
1952static inline void memalloc_noio_restore(unsigned int flags)
1953{
1954 current->flags = (current->flags & ~PF_MEMALLOC_NOIO) | flags;
1955}
1956
1957/* Per-process atomic flags. */
1958#define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */
1959#define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */
1960#define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */
1961
1962
1963#define TASK_PFA_TEST(name, func) \
1964 static inline bool task_##func(struct task_struct *p) \
1965 { return test_bit(PFA_##name, &p->atomic_flags); }
1966#define TASK_PFA_SET(name, func) \
1967 static inline void task_set_##func(struct task_struct *p) \
1968 { set_bit(PFA_##name, &p->atomic_flags); }
1969#define TASK_PFA_CLEAR(name, func) \
1970 static inline void task_clear_##func(struct task_struct *p) \
1971 { clear_bit(PFA_##name, &p->atomic_flags); }
1972
1973TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
1974TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
1975
1976TASK_PFA_TEST(SPREAD_PAGE, spread_page)
1977TASK_PFA_SET(SPREAD_PAGE, spread_page)
1978TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
1979
1980TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
1981TASK_PFA_SET(SPREAD_SLAB, spread_slab)
1982TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
1983
1984/*
1985 * task->jobctl flags
1986 */
1987#define JOBCTL_STOP_SIGMASK 0xffff /* signr of the last group stop */
1988
1989#define JOBCTL_STOP_DEQUEUED_BIT 16 /* stop signal dequeued */
1990#define JOBCTL_STOP_PENDING_BIT 17 /* task should stop for group stop */
1991#define JOBCTL_STOP_CONSUME_BIT 18 /* consume group stop count */
1992#define JOBCTL_TRAP_STOP_BIT 19 /* trap for STOP */
1993#define JOBCTL_TRAP_NOTIFY_BIT 20 /* trap for NOTIFY */
1994#define JOBCTL_TRAPPING_BIT 21 /* switching to TRACED */
1995#define JOBCTL_LISTENING_BIT 22 /* ptracer is listening for events */
1996
1997#define JOBCTL_STOP_DEQUEUED (1 << JOBCTL_STOP_DEQUEUED_BIT)
1998#define JOBCTL_STOP_PENDING (1 << JOBCTL_STOP_PENDING_BIT)
1999#define JOBCTL_STOP_CONSUME (1 << JOBCTL_STOP_CONSUME_BIT)
2000#define JOBCTL_TRAP_STOP (1 << JOBCTL_TRAP_STOP_BIT)
2001#define JOBCTL_TRAP_NOTIFY (1 << JOBCTL_TRAP_NOTIFY_BIT)
2002#define JOBCTL_TRAPPING (1 << JOBCTL_TRAPPING_BIT)
2003#define JOBCTL_LISTENING (1 << JOBCTL_LISTENING_BIT)
2004
2005#define JOBCTL_TRAP_MASK (JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
2006#define JOBCTL_PENDING_MASK (JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
2007
2008extern bool task_set_jobctl_pending(struct task_struct *task,
2009 unsigned int mask);
2010extern void task_clear_jobctl_trapping(struct task_struct *task);
2011extern void task_clear_jobctl_pending(struct task_struct *task,
2012 unsigned int mask);
2013
2014#ifdef CONFIG_PREEMPT_RCU
2015
2016#define RCU_READ_UNLOCK_BLOCKED (1 << 0) /* blocked while in RCU read-side. */
2017#define RCU_READ_UNLOCK_NEED_QS (1 << 1) /* RCU core needs CPU response. */
2018
2019static inline void rcu_copy_process(struct task_struct *p)
2020{
2021 p->rcu_read_lock_nesting = 0;
2022 p->rcu_read_unlock_special = 0;
2023#ifdef CONFIG_TREE_PREEMPT_RCU
2024 p->rcu_blocked_node = NULL;
2025#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
2026 INIT_LIST_HEAD(&p->rcu_node_entry);
2027}
2028
2029#else
2030
2031static inline void rcu_copy_process(struct task_struct *p)
2032{
2033}
2034
2035#endif
2036
2037static inline void tsk_restore_flags(struct task_struct *task,
2038 unsigned long orig_flags, unsigned long flags)
2039{
2040 task->flags &= ~flags;
2041 task->flags |= orig_flags & flags;
2042}
2043
2044#ifdef CONFIG_SMP
2045extern void do_set_cpus_allowed(struct task_struct *p,
2046 const struct cpumask *new_mask);
2047
2048extern int set_cpus_allowed_ptr(struct task_struct *p,
2049 const struct cpumask *new_mask);
2050#else
2051static inline void do_set_cpus_allowed(struct task_struct *p,
2052 const struct cpumask *new_mask)
2053{
2054}
2055static inline int set_cpus_allowed_ptr(struct task_struct *p,
2056 const struct cpumask *new_mask)
2057{
2058 if (!cpumask_test_cpu(0, new_mask))
2059 return -EINVAL;
2060 return 0;
2061}
2062#endif
2063
2064#ifdef CONFIG_NO_HZ_COMMON
2065void calc_load_enter_idle(void);
2066void calc_load_exit_idle(void);
2067#else
2068static inline void calc_load_enter_idle(void) { }
2069static inline void calc_load_exit_idle(void) { }
2070#endif /* CONFIG_NO_HZ_COMMON */
2071
2072#ifndef CONFIG_CPUMASK_OFFSTACK
2073static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask)
2074{
2075 return set_cpus_allowed_ptr(p, &new_mask);
2076}
2077#endif
2078
2079/*
2080 * Do not use outside of architecture code which knows its limitations.
2081 *
2082 * sched_clock() has no promise of monotonicity or bounded drift between
2083 * CPUs, use (which you should not) requires disabling IRQs.
2084 *
2085 * Please use one of the three interfaces below.
2086 */
2087extern unsigned long long notrace sched_clock(void);
2088/*
2089 * See the comment in kernel/sched/clock.c
2090 */
2091extern u64 cpu_clock(int cpu);
2092extern u64 local_clock(void);
2093extern u64 sched_clock_cpu(int cpu);
2094
2095
2096extern void sched_clock_init(void);
2097
2098#ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
2099static inline void sched_clock_tick(void)
2100{
2101}
2102
2103static inline void sched_clock_idle_sleep_event(void)
2104{
2105}
2106
2107static inline void sched_clock_idle_wakeup_event(u64 delta_ns)
2108{
2109}
2110#else
2111/*
2112 * Architectures can set this to 1 if they have specified
2113 * CONFIG_HAVE_UNSTABLE_SCHED_CLOCK in their arch Kconfig,
2114 * but then during bootup it turns out that sched_clock()
2115 * is reliable after all:
2116 */
2117extern int sched_clock_stable(void);
2118extern void set_sched_clock_stable(void);
2119extern void clear_sched_clock_stable(void);
2120
2121extern void sched_clock_tick(void);
2122extern void sched_clock_idle_sleep_event(void);
2123extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2124#endif
2125
2126#ifdef CONFIG_IRQ_TIME_ACCOUNTING
2127/*
2128 * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
2129 * The reason for this explicit opt-in is not to have perf penalty with
2130 * slow sched_clocks.
2131 */
2132extern void enable_sched_clock_irqtime(void);
2133extern void disable_sched_clock_irqtime(void);
2134#else
2135static inline void enable_sched_clock_irqtime(void) {}
2136static inline void disable_sched_clock_irqtime(void) {}
2137#endif
2138
2139extern unsigned long long
2140task_sched_runtime(struct task_struct *task);
2141
2142/* sched_exec is called by processes performing an exec */
2143#ifdef CONFIG_SMP
2144extern void sched_exec(void);
2145#else
2146#define sched_exec() {}
2147#endif
2148
2149extern void sched_clock_idle_sleep_event(void);
2150extern void sched_clock_idle_wakeup_event(u64 delta_ns);
2151
2152#ifdef CONFIG_HOTPLUG_CPU
2153extern void idle_task_exit(void);
2154#else
2155static inline void idle_task_exit(void) {}
2156#endif
2157
2158#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
2159extern void wake_up_nohz_cpu(int cpu);
2160#else
2161static inline void wake_up_nohz_cpu(int cpu) { }
2162#endif
2163
2164#ifdef CONFIG_NO_HZ_FULL
2165extern bool sched_can_stop_tick(void);
2166extern u64 scheduler_tick_max_deferment(void);
2167#else
2168static inline bool sched_can_stop_tick(void) { return false; }
2169#endif
2170
2171#ifdef CONFIG_SCHED_AUTOGROUP
2172extern void sched_autogroup_create_attach(struct task_struct *p);
2173extern void sched_autogroup_detach(struct task_struct *p);
2174extern void sched_autogroup_fork(struct signal_struct *sig);
2175extern void sched_autogroup_exit(struct signal_struct *sig);
2176#ifdef CONFIG_PROC_FS
2177extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
2178extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
2179#endif
2180#else
2181static inline void sched_autogroup_create_attach(struct task_struct *p) { }
2182static inline void sched_autogroup_detach(struct task_struct *p) { }
2183static inline void sched_autogroup_fork(struct signal_struct *sig) { }
2184static inline void sched_autogroup_exit(struct signal_struct *sig) { }
2185#endif
2186
2187extern int yield_to(struct task_struct *p, bool preempt);
2188extern void set_user_nice(struct task_struct *p, long nice);
2189extern int task_prio(const struct task_struct *p);
2190/**
2191 * task_nice - return the nice value of a given task.
2192 * @p: the task in question.
2193 *
2194 * Return: The nice value [ -20 ... 0 ... 19 ].
2195 */
2196static inline int task_nice(const struct task_struct *p)
2197{
2198 return PRIO_TO_NICE((p)->static_prio);
2199}
2200extern int can_nice(const struct task_struct *p, const int nice);
2201extern int task_curr(const struct task_struct *p);
2202extern int idle_cpu(int cpu);
2203extern int sched_setscheduler(struct task_struct *, int,
2204 const struct sched_param *);
2205extern int sched_setscheduler_nocheck(struct task_struct *, int,
2206 const struct sched_param *);
2207extern int sched_setattr(struct task_struct *,
2208 const struct sched_attr *);
2209extern struct task_struct *idle_task(int cpu);
2210/**
2211 * is_idle_task - is the specified task an idle task?
2212 * @p: the task in question.
2213 *
2214 * Return: 1 if @p is an idle task. 0 otherwise.
2215 */
2216static inline bool is_idle_task(const struct task_struct *p)
2217{
2218 return p->pid == 0;
2219}
2220extern struct task_struct *curr_task(int cpu);
2221extern void set_curr_task(int cpu, struct task_struct *p);
2222
2223void yield(void);
2224
2225/*
2226 * The default (Linux) execution domain.
2227 */
2228extern struct exec_domain default_exec_domain;
2229
2230union thread_union {
2231 struct thread_info thread_info;
2232 unsigned long stack[THREAD_SIZE/sizeof(long)];
2233};
2234
2235#ifndef __HAVE_ARCH_KSTACK_END
2236static inline int kstack_end(void *addr)
2237{
2238 /* Reliable end of stack detection:
2239 * Some APM bios versions misalign the stack
2240 */
2241 return !(((unsigned long)addr+sizeof(void*)-1) & (THREAD_SIZE-sizeof(void*)));
2242}
2243#endif
2244
2245extern union thread_union init_thread_union;
2246extern struct task_struct init_task;
2247
2248extern struct mm_struct init_mm;
2249
2250extern struct pid_namespace init_pid_ns;
2251
2252/*
2253 * find a task by one of its numerical ids
2254 *
2255 * find_task_by_pid_ns():
2256 * finds a task by its pid in the specified namespace
2257 * find_task_by_vpid():
2258 * finds a task by its virtual pid
2259 *
2260 * see also find_vpid() etc in include/linux/pid.h
2261 */
2262
2263extern struct task_struct *find_task_by_vpid(pid_t nr);
2264extern struct task_struct *find_task_by_pid_ns(pid_t nr,
2265 struct pid_namespace *ns);
2266
2267/* per-UID process charging. */
2268extern struct user_struct * alloc_uid(kuid_t);
2269static inline struct user_struct *get_uid(struct user_struct *u)
2270{
2271 atomic_inc(&u->__count);
2272 return u;
2273}
2274extern void free_uid(struct user_struct *);
2275
2276#include <asm/current.h>
2277
2278extern void xtime_update(unsigned long ticks);
2279
2280extern int wake_up_state(struct task_struct *tsk, unsigned int state);
2281extern int wake_up_process(struct task_struct *tsk);
2282extern void wake_up_new_task(struct task_struct *tsk);
2283#ifdef CONFIG_SMP
2284 extern void kick_process(struct task_struct *tsk);
2285#else
2286 static inline void kick_process(struct task_struct *tsk) { }
2287#endif
2288extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
2289extern void sched_dead(struct task_struct *p);
2290
2291extern void proc_caches_init(void);
2292extern void flush_signals(struct task_struct *);
2293extern void __flush_signals(struct task_struct *);
2294extern void ignore_signals(struct task_struct *);
2295extern void flush_signal_handlers(struct task_struct *, int force_default);
2296extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info);
2297
2298static inline int dequeue_signal_lock(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
2299{
2300 unsigned long flags;
2301 int ret;
2302
2303 spin_lock_irqsave(&tsk->sighand->siglock, flags);
2304 ret = dequeue_signal(tsk, mask, info);
2305 spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
2306
2307 return ret;
2308}
2309
2310extern void block_all_signals(int (*notifier)(void *priv), void *priv,
2311 sigset_t *mask);
2312extern void unblock_all_signals(void);
2313extern void release_task(struct task_struct * p);
2314extern int send_sig_info(int, struct siginfo *, struct task_struct *);
2315extern int force_sigsegv(int, struct task_struct *);
2316extern int force_sig_info(int, struct siginfo *, struct task_struct *);
2317extern int __kill_pgrp_info(int sig, struct siginfo *info, struct pid *pgrp);
2318extern int kill_pid_info(int sig, struct siginfo *info, struct pid *pid);
2319extern int kill_pid_info_as_cred(int, struct siginfo *, struct pid *,
2320 const struct cred *, u32);
2321extern int kill_pgrp(struct pid *pid, int sig, int priv);
2322extern int kill_pid(struct pid *pid, int sig, int priv);
2323extern int kill_proc_info(int, struct siginfo *, pid_t);
2324extern __must_check bool do_notify_parent(struct task_struct *, int);
2325extern void __wake_up_parent(struct task_struct *p, struct task_struct *parent);
2326extern void force_sig(int, struct task_struct *);
2327extern int send_sig(int, struct task_struct *, int);
2328extern int zap_other_threads(struct task_struct *p);
2329extern struct sigqueue *sigqueue_alloc(void);
2330extern void sigqueue_free(struct sigqueue *);
2331extern int send_sigqueue(struct sigqueue *, struct task_struct *, int group);
2332extern int do_sigaction(int, struct k_sigaction *, struct k_sigaction *);
2333
2334static inline void restore_saved_sigmask(void)
2335{
2336 if (test_and_clear_restore_sigmask())
2337 __set_current_blocked(¤t->saved_sigmask);
2338}
2339
2340static inline sigset_t *sigmask_to_save(void)
2341{
2342 sigset_t *res = ¤t->blocked;
2343 if (unlikely(test_restore_sigmask()))
2344 res = ¤t->saved_sigmask;
2345 return res;
2346}
2347
2348static inline int kill_cad_pid(int sig, int priv)
2349{
2350 return kill_pid(cad_pid, sig, priv);
2351}
2352
2353/* These can be the second arg to send_sig_info/send_group_sig_info. */
2354#define SEND_SIG_NOINFO ((struct siginfo *) 0)
2355#define SEND_SIG_PRIV ((struct siginfo *) 1)
2356#define SEND_SIG_FORCED ((struct siginfo *) 2)
2357
2358/*
2359 * True if we are on the alternate signal stack.
2360 */
2361static inline int on_sig_stack(unsigned long sp)
2362{
2363#ifdef CONFIG_STACK_GROWSUP
2364 return sp >= current->sas_ss_sp &&
2365 sp - current->sas_ss_sp < current->sas_ss_size;
2366#else
2367 return sp > current->sas_ss_sp &&
2368 sp - current->sas_ss_sp <= current->sas_ss_size;
2369#endif
2370}
2371
2372static inline int sas_ss_flags(unsigned long sp)
2373{
2374 if (!current->sas_ss_size)
2375 return SS_DISABLE;
2376
2377 return on_sig_stack(sp) ? SS_ONSTACK : 0;
2378}
2379
2380static inline unsigned long sigsp(unsigned long sp, struct ksignal *ksig)
2381{
2382 if (unlikely((ksig->ka.sa.sa_flags & SA_ONSTACK)) && ! sas_ss_flags(sp))
2383#ifdef CONFIG_STACK_GROWSUP
2384 return current->sas_ss_sp;
2385#else
2386 return current->sas_ss_sp + current->sas_ss_size;
2387#endif
2388 return sp;
2389}
2390
2391/*
2392 * Routines for handling mm_structs
2393 */
2394extern struct mm_struct * mm_alloc(void);
2395
2396/* mmdrop drops the mm and the page tables */
2397extern void __mmdrop(struct mm_struct *);
2398static inline void mmdrop(struct mm_struct * mm)
2399{
2400 if (unlikely(atomic_dec_and_test(&mm->mm_count)))
2401 __mmdrop(mm);
2402}
2403
2404/* mmput gets rid of the mappings and all user-space */
2405extern void mmput(struct mm_struct *);
2406/* Grab a reference to a task's mm, if it is not already going away */
2407extern struct mm_struct *get_task_mm(struct task_struct *task);
2408/*
2409 * Grab a reference to a task's mm, if it is not already going away
2410 * and ptrace_may_access with the mode parameter passed to it
2411 * succeeds.
2412 */
2413extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
2414/* Remove the current tasks stale references to the old mm_struct */
2415extern void mm_release(struct task_struct *, struct mm_struct *);
2416
2417extern int copy_thread(unsigned long, unsigned long, unsigned long,
2418 struct task_struct *);
2419extern void flush_thread(void);
2420extern void exit_thread(void);
2421
2422extern void exit_files(struct task_struct *);
2423extern void __cleanup_sighand(struct sighand_struct *);
2424
2425extern void exit_itimers(struct signal_struct *);
2426extern void flush_itimer_signals(void);
2427
2428extern void do_group_exit(int);
2429
2430extern int do_execve(struct filename *,
2431 const char __user * const __user *,
2432 const char __user * const __user *);
2433extern long do_fork(unsigned long, unsigned long, unsigned long, int __user *, int __user *);
2434struct task_struct *fork_idle(int);
2435extern pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags);
2436
2437extern void __set_task_comm(struct task_struct *tsk, const char *from, bool exec);
2438static inline void set_task_comm(struct task_struct *tsk, const char *from)
2439{
2440 __set_task_comm(tsk, from, false);
2441}
2442extern char *get_task_comm(char *to, struct task_struct *tsk);
2443
2444#ifdef CONFIG_SMP
2445void scheduler_ipi(void);
2446extern unsigned long wait_task_inactive(struct task_struct *, long match_state);
2447#else
2448static inline void scheduler_ipi(void) { }
2449static inline unsigned long wait_task_inactive(struct task_struct *p,
2450 long match_state)
2451{
2452 return 1;
2453}
2454#endif
2455
2456#define next_task(p) \
2457 list_entry_rcu((p)->tasks.next, struct task_struct, tasks)
2458
2459#define for_each_process(p) \
2460 for (p = &init_task ; (p = next_task(p)) != &init_task ; )
2461
2462extern bool current_is_single_threaded(void);
2463
2464/*
2465 * Careful: do_each_thread/while_each_thread is a double loop so
2466 * 'break' will not work as expected - use goto instead.
2467 */
2468#define do_each_thread(g, t) \
2469 for (g = t = &init_task ; (g = t = next_task(g)) != &init_task ; ) do
2470
2471#define while_each_thread(g, t) \
2472 while ((t = next_thread(t)) != g)
2473
2474#define __for_each_thread(signal, t) \
2475 list_for_each_entry_rcu(t, &(signal)->thread_head, thread_node)
2476
2477#define for_each_thread(p, t) \
2478 __for_each_thread((p)->signal, t)
2479
2480/* Careful: this is a double loop, 'break' won't work as expected. */
2481#define for_each_process_thread(p, t) \
2482 for_each_process(p) for_each_thread(p, t)
2483
2484static inline int get_nr_threads(struct task_struct *tsk)
2485{
2486 return tsk->signal->nr_threads;
2487}
2488
2489static inline bool thread_group_leader(struct task_struct *p)
2490{
2491 return p->exit_signal >= 0;
2492}
2493
2494/* Do to the insanities of de_thread it is possible for a process
2495 * to have the pid of the thread group leader without actually being
2496 * the thread group leader. For iteration through the pids in proc
2497 * all we care about is that we have a task with the appropriate
2498 * pid, we don't actually care if we have the right task.
2499 */
2500static inline bool has_group_leader_pid(struct task_struct *p)
2501{
2502 return task_pid(p) == p->signal->leader_pid;
2503}
2504
2505static inline
2506bool same_thread_group(struct task_struct *p1, struct task_struct *p2)
2507{
2508 return p1->signal == p2->signal;
2509}
2510
2511static inline struct task_struct *next_thread(const struct task_struct *p)
2512{
2513 return list_entry_rcu(p->thread_group.next,
2514 struct task_struct, thread_group);
2515}
2516
2517static inline int thread_group_empty(struct task_struct *p)
2518{
2519 return list_empty(&p->thread_group);
2520}
2521
2522#define delay_group_leader(p) \
2523 (thread_group_leader(p) && !thread_group_empty(p))
2524
2525/*
2526 * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
2527 * subscriptions and synchronises with wait4(). Also used in procfs. Also
2528 * pins the final release of task.io_context. Also protects ->cpuset and
2529 * ->cgroup.subsys[]. And ->vfork_done.
2530 *
2531 * Nests both inside and outside of read_lock(&tasklist_lock).
2532 * It must not be nested with write_lock_irq(&tasklist_lock),
2533 * neither inside nor outside.
2534 */
2535static inline void task_lock(struct task_struct *p)
2536{
2537 spin_lock(&p->alloc_lock);
2538}
2539
2540static inline void task_unlock(struct task_struct *p)
2541{
2542 spin_unlock(&p->alloc_lock);
2543}
2544
2545extern struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
2546 unsigned long *flags);
2547
2548static inline struct sighand_struct *lock_task_sighand(struct task_struct *tsk,
2549 unsigned long *flags)
2550{
2551 struct sighand_struct *ret;
2552
2553 ret = __lock_task_sighand(tsk, flags);
2554 (void)__cond_lock(&tsk->sighand->siglock, ret);
2555 return ret;
2556}
2557
2558static inline void unlock_task_sighand(struct task_struct *tsk,
2559 unsigned long *flags)
2560{
2561 spin_unlock_irqrestore(&tsk->sighand->siglock, *flags);
2562}
2563
2564#ifdef CONFIG_CGROUPS
2565static inline void threadgroup_change_begin(struct task_struct *tsk)
2566{
2567 down_read(&tsk->signal->group_rwsem);
2568}
2569static inline void threadgroup_change_end(struct task_struct *tsk)
2570{
2571 up_read(&tsk->signal->group_rwsem);
2572}
2573
2574/**
2575 * threadgroup_lock - lock threadgroup
2576 * @tsk: member task of the threadgroup to lock
2577 *
2578 * Lock the threadgroup @tsk belongs to. No new task is allowed to enter
2579 * and member tasks aren't allowed to exit (as indicated by PF_EXITING) or
2580 * change ->group_leader/pid. This is useful for cases where the threadgroup
2581 * needs to stay stable across blockable operations.
2582 *
2583 * fork and exit paths explicitly call threadgroup_change_{begin|end}() for
2584 * synchronization. While held, no new task will be added to threadgroup
2585 * and no existing live task will have its PF_EXITING set.
2586 *
2587 * de_thread() does threadgroup_change_{begin|end}() when a non-leader
2588 * sub-thread becomes a new leader.
2589 */
2590static inline void threadgroup_lock(struct task_struct *tsk)
2591{
2592 down_write(&tsk->signal->group_rwsem);
2593}
2594
2595/**
2596 * threadgroup_unlock - unlock threadgroup
2597 * @tsk: member task of the threadgroup to unlock
2598 *
2599 * Reverse threadgroup_lock().
2600 */
2601static inline void threadgroup_unlock(struct task_struct *tsk)
2602{
2603 up_write(&tsk->signal->group_rwsem);
2604}
2605#else
2606static inline void threadgroup_change_begin(struct task_struct *tsk) {}
2607static inline void threadgroup_change_end(struct task_struct *tsk) {}
2608static inline void threadgroup_lock(struct task_struct *tsk) {}
2609static inline void threadgroup_unlock(struct task_struct *tsk) {}
2610#endif
2611
2612#ifndef __HAVE_THREAD_FUNCTIONS
2613
2614#define task_thread_info(task) ((struct thread_info *)(task)->stack)
2615#define task_stack_page(task) ((task)->stack)
2616
2617static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
2618{
2619 *task_thread_info(p) = *task_thread_info(org);
2620 task_thread_info(p)->task = p;
2621}
2622
2623/*
2624 * Return the address of the last usable long on the stack.
2625 *
2626 * When the stack grows down, this is just above the thread
2627 * info struct. Going any lower will corrupt the threadinfo.
2628 *
2629 * When the stack grows up, this is the highest address.
2630 * Beyond that position, we corrupt data on the next page.
2631 */
2632static inline unsigned long *end_of_stack(struct task_struct *p)
2633{
2634#ifdef CONFIG_STACK_GROWSUP
2635 return (unsigned long *)((unsigned long)task_thread_info(p) + THREAD_SIZE) - 1;
2636#else
2637 return (unsigned long *)(task_thread_info(p) + 1);
2638#endif
2639}
2640
2641#endif
2642
2643static inline int object_is_on_stack(void *obj)
2644{
2645 void *stack = task_stack_page(current);
2646
2647 return (obj >= stack) && (obj < (stack + THREAD_SIZE));
2648}
2649
2650extern void thread_info_cache_init(void);
2651
2652#ifdef CONFIG_DEBUG_STACK_USAGE
2653static inline unsigned long stack_not_used(struct task_struct *p)
2654{
2655 unsigned long *n = end_of_stack(p);
2656
2657 do { /* Skip over canary */
2658 n++;
2659 } while (!*n);
2660
2661 return (unsigned long)n - (unsigned long)end_of_stack(p);
2662}
2663#endif
2664
2665/* set thread flags in other task's structures
2666 * - see asm/thread_info.h for TIF_xxxx flags available
2667 */
2668static inline void set_tsk_thread_flag(struct task_struct *tsk, int flag)
2669{
2670 set_ti_thread_flag(task_thread_info(tsk), flag);
2671}
2672
2673static inline void clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2674{
2675 clear_ti_thread_flag(task_thread_info(tsk), flag);
2676}
2677
2678static inline int test_and_set_tsk_thread_flag(struct task_struct *tsk, int flag)
2679{
2680 return test_and_set_ti_thread_flag(task_thread_info(tsk), flag);
2681}
2682
2683static inline int test_and_clear_tsk_thread_flag(struct task_struct *tsk, int flag)
2684{
2685 return test_and_clear_ti_thread_flag(task_thread_info(tsk), flag);
2686}
2687
2688static inline int test_tsk_thread_flag(struct task_struct *tsk, int flag)
2689{
2690 return test_ti_thread_flag(task_thread_info(tsk), flag);
2691}
2692
2693static inline void set_tsk_need_resched(struct task_struct *tsk)
2694{
2695 set_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2696}
2697
2698static inline void clear_tsk_need_resched(struct task_struct *tsk)
2699{
2700 clear_tsk_thread_flag(tsk,TIF_NEED_RESCHED);
2701}
2702
2703static inline int test_tsk_need_resched(struct task_struct *tsk)
2704{
2705 return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED));
2706}
2707
2708static inline int restart_syscall(void)
2709{
2710 set_tsk_thread_flag(current, TIF_SIGPENDING);
2711 return -ERESTARTNOINTR;
2712}
2713
2714static inline int signal_pending(struct task_struct *p)
2715{
2716 return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));
2717}
2718
2719static inline int __fatal_signal_pending(struct task_struct *p)
2720{
2721 return unlikely(sigismember(&p->pending.signal, SIGKILL));
2722}
2723
2724static inline int fatal_signal_pending(struct task_struct *p)
2725{
2726 return signal_pending(p) && __fatal_signal_pending(p);
2727}
2728
2729static inline int signal_pending_state(long state, struct task_struct *p)
2730{
2731 if (!(state & (TASK_INTERRUPTIBLE | TASK_WAKEKILL)))
2732 return 0;
2733 if (!signal_pending(p))
2734 return 0;
2735
2736 return (state & TASK_INTERRUPTIBLE) || __fatal_signal_pending(p);
2737}
2738
2739/*
2740 * cond_resched() and cond_resched_lock(): latency reduction via
2741 * explicit rescheduling in places that are safe. The return
2742 * value indicates whether a reschedule was done in fact.
2743 * cond_resched_lock() will drop the spinlock before scheduling,
2744 * cond_resched_softirq() will enable bhs before scheduling.
2745 */
2746extern int _cond_resched(void);
2747
2748#define cond_resched() ({ \
2749 __might_sleep(__FILE__, __LINE__, 0); \
2750 _cond_resched(); \
2751})
2752
2753extern int __cond_resched_lock(spinlock_t *lock);
2754
2755#ifdef CONFIG_PREEMPT_COUNT
2756#define PREEMPT_LOCK_OFFSET PREEMPT_OFFSET
2757#else
2758#define PREEMPT_LOCK_OFFSET 0
2759#endif
2760
2761#define cond_resched_lock(lock) ({ \
2762 __might_sleep(__FILE__, __LINE__, PREEMPT_LOCK_OFFSET); \
2763 __cond_resched_lock(lock); \
2764})
2765
2766extern int __cond_resched_softirq(void);
2767
2768#define cond_resched_softirq() ({ \
2769 __might_sleep(__FILE__, __LINE__, SOFTIRQ_DISABLE_OFFSET); \
2770 __cond_resched_softirq(); \
2771})
2772
2773static inline void cond_resched_rcu(void)
2774{
2775#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
2776 rcu_read_unlock();
2777 cond_resched();
2778 rcu_read_lock();
2779#endif
2780}
2781
2782/*
2783 * Does a critical section need to be broken due to another
2784 * task waiting?: (technically does not depend on CONFIG_PREEMPT,
2785 * but a general need for low latency)
2786 */
2787static inline int spin_needbreak(spinlock_t *lock)
2788{
2789#ifdef CONFIG_PREEMPT
2790 return spin_is_contended(lock);
2791#else
2792 return 0;
2793#endif
2794}
2795
2796/*
2797 * Idle thread specific functions to determine the need_resched
2798 * polling state.
2799 */
2800#ifdef TIF_POLLING_NRFLAG
2801static inline int tsk_is_polling(struct task_struct *p)
2802{
2803 return test_tsk_thread_flag(p, TIF_POLLING_NRFLAG);
2804}
2805
2806static inline void __current_set_polling(void)
2807{
2808 set_thread_flag(TIF_POLLING_NRFLAG);
2809}
2810
2811static inline bool __must_check current_set_polling_and_test(void)
2812{
2813 __current_set_polling();
2814
2815 /*
2816 * Polling state must be visible before we test NEED_RESCHED,
2817 * paired by resched_curr()
2818 */
2819 smp_mb__after_atomic();
2820
2821 return unlikely(tif_need_resched());
2822}
2823
2824static inline void __current_clr_polling(void)
2825{
2826 clear_thread_flag(TIF_POLLING_NRFLAG);
2827}
2828
2829static inline bool __must_check current_clr_polling_and_test(void)
2830{
2831 __current_clr_polling();
2832
2833 /*
2834 * Polling state must be visible before we test NEED_RESCHED,
2835 * paired by resched_curr()
2836 */
2837 smp_mb__after_atomic();
2838
2839 return unlikely(tif_need_resched());
2840}
2841
2842#else
2843static inline int tsk_is_polling(struct task_struct *p) { return 0; }
2844static inline void __current_set_polling(void) { }
2845static inline void __current_clr_polling(void) { }
2846
2847static inline bool __must_check current_set_polling_and_test(void)
2848{
2849 return unlikely(tif_need_resched());
2850}
2851static inline bool __must_check current_clr_polling_and_test(void)
2852{
2853 return unlikely(tif_need_resched());
2854}
2855#endif
2856
2857static inline void current_clr_polling(void)
2858{
2859 __current_clr_polling();
2860
2861 /*
2862 * Ensure we check TIF_NEED_RESCHED after we clear the polling bit.
2863 * Once the bit is cleared, we'll get IPIs with every new
2864 * TIF_NEED_RESCHED and the IPI handler, scheduler_ipi(), will also
2865 * fold.
2866 */
2867 smp_mb(); /* paired with resched_curr() */
2868
2869 preempt_fold_need_resched();
2870}
2871
2872static __always_inline bool need_resched(void)
2873{
2874 return unlikely(tif_need_resched());
2875}
2876
2877/*
2878 * Thread group CPU time accounting.
2879 */
2880void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times);
2881void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times);
2882
2883static inline void thread_group_cputime_init(struct signal_struct *sig)
2884{
2885 raw_spin_lock_init(&sig->cputimer.lock);
2886}
2887
2888/*
2889 * Reevaluate whether the task has signals pending delivery.
2890 * Wake the task if so.
2891 * This is required every time the blocked sigset_t changes.
2892 * callers must hold sighand->siglock.
2893 */
2894extern void recalc_sigpending_and_wake(struct task_struct *t);
2895extern void recalc_sigpending(void);
2896
2897extern void signal_wake_up_state(struct task_struct *t, unsigned int state);
2898
2899static inline void signal_wake_up(struct task_struct *t, bool resume)
2900{
2901 signal_wake_up_state(t, resume ? TASK_WAKEKILL : 0);
2902}
2903static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
2904{
2905 signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
2906}
2907
2908/*
2909 * Wrappers for p->thread_info->cpu access. No-op on UP.
2910 */
2911#ifdef CONFIG_SMP
2912
2913static inline unsigned int task_cpu(const struct task_struct *p)
2914{
2915 return task_thread_info(p)->cpu;
2916}
2917
2918static inline int task_node(const struct task_struct *p)
2919{
2920 return cpu_to_node(task_cpu(p));
2921}
2922
2923extern void set_task_cpu(struct task_struct *p, unsigned int cpu);
2924
2925#else
2926
2927static inline unsigned int task_cpu(const struct task_struct *p)
2928{
2929 return 0;
2930}
2931
2932static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)
2933{
2934}
2935
2936#endif /* CONFIG_SMP */
2937
2938extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask);
2939extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
2940
2941#ifdef CONFIG_CGROUP_SCHED
2942extern struct task_group root_task_group;
2943#endif /* CONFIG_CGROUP_SCHED */
2944
2945extern int task_can_switch_user(struct user_struct *up,
2946 struct task_struct *tsk);
2947
2948#ifdef CONFIG_TASK_XACCT
2949static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
2950{
2951 tsk->ioac.rchar += amt;
2952}
2953
2954static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
2955{
2956 tsk->ioac.wchar += amt;
2957}
2958
2959static inline void inc_syscr(struct task_struct *tsk)
2960{
2961 tsk->ioac.syscr++;
2962}
2963
2964static inline void inc_syscw(struct task_struct *tsk)
2965{
2966 tsk->ioac.syscw++;
2967}
2968#else
2969static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
2970{
2971}
2972
2973static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
2974{
2975}
2976
2977static inline void inc_syscr(struct task_struct *tsk)
2978{
2979}
2980
2981static inline void inc_syscw(struct task_struct *tsk)
2982{
2983}
2984#endif
2985
2986#ifndef TASK_SIZE_OF
2987#define TASK_SIZE_OF(tsk) TASK_SIZE
2988#endif
2989
2990#ifdef CONFIG_MEMCG
2991extern void mm_update_next_owner(struct mm_struct *mm);
2992#else
2993static inline void mm_update_next_owner(struct mm_struct *mm)
2994{
2995}
2996#endif /* CONFIG_MEMCG */
2997
2998static inline unsigned long task_rlimit(const struct task_struct *tsk,
2999 unsigned int limit)
3000{
3001 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_cur);
3002}
3003
3004static inline unsigned long task_rlimit_max(const struct task_struct *tsk,
3005 unsigned int limit)
3006{
3007 return ACCESS_ONCE(tsk->signal->rlim[limit].rlim_max);
3008}
3009
3010static inline unsigned long rlimit(unsigned int limit)
3011{
3012 return task_rlimit(current, limit);
3013}
3014
3015static inline unsigned long rlimit_max(unsigned int limit)
3016{
3017 return task_rlimit_max(current, limit);
3018}
3019
3020#endif